Merge branch 'for-next' of https://evilpiepirate.org/git/bcachefs.gitNotice: this object is not reachable from any branch.

Notice: this object is not reachable from any branch.

103 files changed, 3807 insertions, 3666 deletions

diff --git a/Documentation/filesystems/bcachefs/CodingStyle.rst b/Documentation/filesystems/bcachefs/CodingStyle.rst
new file mode 100644
index 00000000000000..0c45829a4899af
--- /dev/null
+++ b/Documentation/filesystems/bcachefs/CodingStyle.rst
@@ -0,0 +1,186 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+bcachefs coding style
+=====================
+
+Good development is like gardening, and codebases are our gardens. Tend to them
+every day; look for little things that are out of place or in need of tidying.
+A little weeding here and there goes a long way; don't wait until things have
+spiraled out of control.
+
+Things don't always have to be perfect - nitpicking often does more harm than
+good. But appreciate beauty when you see it - and let people know.
+
+The code that you are afraid to touch is the code most in need of refactoring.
+
+A little organizing here and there goes a long way.
+
+Put real thought into how you organize things.
+
+Good code is readable code, where the structure is simple and leaves nowhere
+for bugs to hide.
+
+Assertions are one of our most important tools for writing reliable code. If in
+the course of writing a patchset you encounter a condition that shouldn't
+happen (and will have unpredictable or undefined behaviour if it does), or
+you're not sure if it can happen and not sure how to handle it yet - make it a
+BUG_ON(). Don't leave undefined or unspecified behavior lurking in the codebase.
+
+By the time you finish the patchset, you should understand better which
+assertions need to be handled and turned into checks with error paths, and
+which should be logically impossible. Leave the BUG_ON()s in for the ones which
+are logically impossible. (Or, make them debug mode assertions if they're
+expensive - but don't turn everything into a debug mode assertion, so that
+we're not stuck debugging undefined behaviour should it turn out that you were
+wrong).
+
+Assertions are documentation that can't go out of date. Good assertions are
+wonderful.
+
+Good assertions drastically and dramatically reduce the amount of testing
+required to shake out bugs.
+
+Good assertions are based on state, not logic. To write good assertions, you
+have to think about what the invariants on your state are.
+
+Good invariants and assertions will hold everywhere in your codebase. This
+means that you can run them in only a few places in the checked in version, but
+should you need to debug something that caused the assertion to fail, you can
+quickly shotgun them everywhere to find the codepath that broke the invariant.
+
+A good assertion checks something that the compiler could check for us, and
+elide - if we were working in a language with embedded correctness proofs that
+the compiler could check. This is something that exists today, but it'll likely
+still be a few decades before it comes to systems programming languages. But we
+can still incorporate that kind of thinking into our code and document the
+invariants with runtime checks - much like the way people working in
+dynamically typed languages may add type annotations, gradually making their
+code statically typed.
+
+Looking for ways to make your assertions simpler - and higher level - will
+often nudge you towards making the entire system simpler and more robust.
+
+Good code is code where you can poke around and see what it's doing -
+introspection. We can't debug anything if we can't see what's going on.
+
+Whenever we're debugging, and the solution isn't immediately obvious, if the
+issue is that we don't know where the issue is because we can't see what's
+going on - fix that first.
+
+We have the tools to make anything visible at runtime, efficiently - RCU and
+percpu data structures among them. Don't let things stay hidden.
+
+The most important tool for introspection is the humble pretty printer - in
+bcachefs, this means `*_to_text()` functions, which output to printbufs.
+
+Pretty printers are wonderful, because they compose and you can use them
+everywhere. Having functions to print whatever object you're working with will
+make your error messages much easier to write (therefore they will actually
+exist) and much more informative. And they can be used from sysfs/debugfs, as
+well as tracepoints.
+
+Runtime info and debugging tools should come with clear descriptions and
+labels, and good structure - we don't want files with a list of bare integers,
+like in procfs. Part of the job of the debugging tools is to educate users and
+new developers as to how the system works.
+
+Error messages should, whenever possible, tell you everything you need to debug
+the issue. It's worth putting effort into them.
+
+Tracepoints shouldn't be the first thing you reach for. They're an important
+tool, but always look for more immediate ways to make things visible. When we
+have to rely on tracing, we have to know which tracepoints we're looking for,
+and then we have to run the troublesome workload, and then we have to sift
+through logs. This is a lot of steps to go through when a user is hitting
+something, and if it's intermittent it may not even be possible.
+
+The humble counter is an incredibly useful tool. They're cheap and simple to
+use, and many complicated internal operations with lots of things that can
+behave weirdly (anything involving memory reclaim, for example) become
+shockingly easy to debug once you have counters on every distinct codepath.
+
+Persistent counters are even better.
+
+When debugging, try to get the most out of every bug you come across; don't
+rush to fix the initial issue. Look for things that will make related bugs
+easier the next time around - introspection, new assertions, better error
+messages, new debug tools, and do those first. Look for ways to make the system
+better behaved; often one bug will uncover several other bugs through
+downstream effects.
+
+Fix all that first, and then the original bug last - even if that means keeping
+a user waiting. They'll thank you in the long run, and when they understand
+what you're doing you'll be amazed at how patient they're happy to be. Users
+like to help - otherwise they wouldn't be reporting the bug in the first place.
+
+Talk to your users. Don't isolate yourself.
+
+Users notice all sorts of interesting things, and by just talking to them and
+interacting with them you can benefit from their experience.
+
+Spend time doing support and helpdesk stuff. Don't just write code - code isn't
+finished until it's being used trouble free.
+
+This will also motivate you to make your debugging tools as good as possible,
+and perhaps even your documentation, too. Like anything else in life, the more
+time you spend at it the better you'll get, and you the developer are the
+person most able to improve the tools to make debugging quick and easy.
+
+Be wary of how you take on and commit to big projects. Don't let development
+become product-manager focused. Often time an idea is a good one but needs to
+wait for its proper time - but you won't know if it's the proper time for an
+idea until you start writing code.
+
+Expect to throw a lot of things away, or leave them half finished for later.
+Nobody writes all perfect code that all gets shipped, and you'll be much more
+productive in the long run if you notice this early and shift to something
+else. The experience gained and lessons learned will be valuable for all the
+other work you do.
+
+But don't be afraid to tackle projects that require significant rework of
+existing code. Sometimes these can be the best projects, because they can lead
+us to make existing code more general, more flexible, more multipurpose and
+perhaps more robust. Just don't hesitate to abandon the idea if it looks like
+it's going to make a mess of things.
+
+Complicated features can often be done as a series of refactorings, with the
+final change that actually implements the feature as a quite small patch at the
+end. It's wonderful when this happens, especially when those refactorings are
+things that improve the codebase in their own right. When that happens there's
+much less risk of wasted effort if the feature you were going for doesn't work
+out.
+
+Always strive to work incrementally. Always strive to turn the big projects
+into little bite sized projects that can prove their own merits.
+
+Instead of always tackling those big projects, look for little things that
+will be useful, and make the big projects easier.
+
+The question of what's likely to be useful is where junior developers most
+often go astray - doing something because it seems like it'll be useful often
+leads to overengineering. Knowing what's useful comes from many years of
+experience, or talking with people who have that experience - or from simply
+reading lots of code and looking for common patterns and issues. Don't be
+afraid to throw things away and do something simpler.
+
+Talk about your ideas with your fellow developers; often times the best things
+come from relaxed conversations where people aren't afraid to say "what if?".
+
+Don't neglect your tools.
+
+The most important tools (besides the compiler and our text editor) are the
+tools we use for testing. The shortest possible edit/test/debug cycle is
+essential for working productively. We learn, gain experience, and discover the
+errors in our thinking by running our code and seeing what happens. If your
+time is being wasted because your tools are bad or too slow - don't accept it,
+fix it.
+
+Put effort into your documentation, commmit messages, and code comments - but
+don't go overboard. A good commit message is wonderful - but if the information
+was important enough to go in a commit message, ask yourself if it would be
+even better as a code comment.
+
+A good code comment is wonderful, but even better is the comment that didn't
+need to exist because the code was so straightforward as to be obvious;
+organized into small clean and tidy modules, with clear and descriptive names
+for functions and variable, where every line of code has a clear purpose.
diff --git a/Documentation/filesystems/bcachefs/index.rst b/Documentation/filesystems/bcachefs/index.rst
index e2bd61ccd96fff..95fc4b90739ed5 100644
--- a/Documentation/filesystems/bcachefs/index.rst
+++ b/Documentation/filesystems/bcachefs/index.rst
@@ -8,4 +8,5 @@ bcachefs Documentation
    :maxdepth: 2
    :numbered:
 
+   CodingStyle
    errorcodes
diff --git a/fs/bcachefs/acl.c b/fs/bcachefs/acl.c
index 5c180fdc3efbdf..250d6c6d3a3a1c 100644
--- a/fs/bcachefs/acl.c
+++ b/fs/bcachefs/acl.c
@@ -282,18 +282,12 @@ struct posix_acl *bch2_get_acl(struct mnt_idmap *idmap,
 	struct btree_trans *trans = bch2_trans_get(c);
 	struct btree_iter iter = { NULL };
 	struct posix_acl *acl = NULL;
-	struct bkey_s_c k;
-	int ret;
 retry:
 	bch2_trans_begin(trans);
 
-	ret = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
-			&hash, inode_inum(inode), &search, 0);
-	if (ret)
-		goto err;
-
-	k = bch2_btree_iter_peek_slot(&iter);
-	ret = bkey_err(k);
+	struct bkey_s_c k = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
+					     &hash, inode_inum(inode), &search, 0);
+	int ret = bkey_err(k);
 	if (ret)
 		goto err;
 
@@ -366,7 +360,7 @@ retry:
 
 	ret   = bch2_subvol_is_ro_trans(trans, inode->ei_subvol) ?:
 		bch2_inode_peek(trans, &inode_iter, &inode_u, inode_inum(inode),
-			      BTREE_ITER_INTENT);
+			      BTREE_ITER_intent);
 	if (ret)
 		goto btree_err;
 
@@ -414,39 +408,30 @@ int bch2_acl_chmod(struct btree_trans *trans, subvol_inum inum,
 	struct bch_hash_info hash_info = bch2_hash_info_init(trans->c, inode);
 	struct xattr_search_key search = X_SEARCH(KEY_TYPE_XATTR_INDEX_POSIX_ACL_ACCESS, "", 0);
 	struct btree_iter iter;
-	struct bkey_s_c_xattr xattr;
-	struct bkey_i_xattr *new;
 	struct posix_acl *acl = NULL;
-	struct bkey_s_c k;
-	int ret;
 
-	ret = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
-			       &hash_info, inum, &search, BTREE_ITER_INTENT);
+	struct bkey_s_c k = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
+			       &hash_info, inum, &search, BTREE_ITER_intent);
+	int ret = bkey_err(k);
 	if (ret)
 		return bch2_err_matches(ret, ENOENT) ? 0 : ret;
 
-	k = bch2_btree_iter_peek_slot(&iter);
-	ret = bkey_err(k);
-	if (ret)
-		goto err;
-	xattr = bkey_s_c_to_xattr(k);
+	struct bkey_s_c_xattr xattr = bkey_s_c_to_xattr(k);
 
 	acl = bch2_acl_from_disk(trans, xattr_val(xattr.v),
 			le16_to_cpu(xattr.v->x_val_len));
 	ret = PTR_ERR_OR_ZERO(acl);
-	if (IS_ERR_OR_NULL(acl))
+	if (ret)
 		goto err;
 
-	ret = allocate_dropping_locks_errcode(trans,
-				__posix_acl_chmod(&acl, _gfp, mode));
+	ret = allocate_dropping_locks_errcode(trans, __posix_acl_chmod(&acl, _gfp, mode));
 	if (ret)
 		goto err;
 
-	new = bch2_acl_to_xattr(trans, acl, ACL_TYPE_ACCESS);
-	if (IS_ERR(new)) {
-		ret = PTR_ERR(new);
+	struct bkey_i_xattr *new = bch2_acl_to_xattr(trans, acl, ACL_TYPE_ACCESS);
+	ret = PTR_ERR_OR_ZERO(new);
+	if (ret)
 		goto err;
-	}
 
 	new->k.p = iter.pos;
 	ret = bch2_trans_update(trans, &iter, &new->k_i, 0);
diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 4ff56fa4d53920..0f68c9dd775954 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -263,7 +263,7 @@ int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
 	case BCH_DATA_free:
 	case BCH_DATA_need_gc_gens:
 	case BCH_DATA_need_discard:
-		bkey_fsck_err_on(bch2_bucket_sectors(*a.v) || a.v->stripe,
+		bkey_fsck_err_on(bch2_bucket_sectors_total(*a.v) || a.v->stripe,
 				 c, err, alloc_key_empty_but_have_data,
 				 "empty data type free but have data");
 		break;
@@ -330,27 +330,17 @@ void bch2_alloc_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c
 	prt_printf(out, "gen %u oldest_gen %u data_type ", a->gen, a->oldest_gen);
 	bch2_prt_data_type(out, a->data_type);
 	prt_newline(out);
-	prt_printf(out, "journal_seq       %llu",	a->journal_seq);
-	prt_newline(out);
-	prt_printf(out, "need_discard      %llu",	BCH_ALLOC_V4_NEED_DISCARD(a));
-	prt_newline(out);
-	prt_printf(out, "need_inc_gen      %llu",	BCH_ALLOC_V4_NEED_INC_GEN(a));
-	prt_newline(out);
-	prt_printf(out, "dirty_sectors     %u",	a->dirty_sectors);
-	prt_newline(out);
-	prt_printf(out, "cached_sectors    %u",	a->cached_sectors);
-	prt_newline(out);
-	prt_printf(out, "stripe            %u",	a->stripe);
-	prt_newline(out);
-	prt_printf(out, "stripe_redundancy %u",	a->stripe_redundancy);
-	prt_newline(out);
-	prt_printf(out, "io_time[READ]     %llu",	a->io_time[READ]);
-	prt_newline(out);
-	prt_printf(out, "io_time[WRITE]    %llu",	a->io_time[WRITE]);
-	prt_newline(out);
-	prt_printf(out, "fragmentation     %llu",	a->fragmentation_lru);
-	prt_newline(out);
-	prt_printf(out, "bp_start          %llu", BCH_ALLOC_V4_BACKPOINTERS_START(a));
+	prt_printf(out, "journal_seq       %llu\n",	a->journal_seq);
+	prt_printf(out, "need_discard      %llu\n",	BCH_ALLOC_V4_NEED_DISCARD(a));
+	prt_printf(out, "need_inc_gen      %llu\n",	BCH_ALLOC_V4_NEED_INC_GEN(a));
+	prt_printf(out, "dirty_sectors     %u\n",	a->dirty_sectors);
+	prt_printf(out, "cached_sectors    %u\n",	a->cached_sectors);
+	prt_printf(out, "stripe            %u\n",	a->stripe);
+	prt_printf(out, "stripe_redundancy %u\n",	a->stripe_redundancy);
+	prt_printf(out, "io_time[READ]     %llu\n",	a->io_time[READ]);
+	prt_printf(out, "io_time[WRITE]    %llu\n",	a->io_time[WRITE]);
+	prt_printf(out, "fragmentation     %llu\n",	a->fragmentation_lru);
+	prt_printf(out, "bp_start          %llu\n", BCH_ALLOC_V4_BACKPOINTERS_START(a));
 	printbuf_indent_sub(out, 2);
 }
 
@@ -439,22 +429,18 @@ struct bkey_i_alloc_v4 *bch2_alloc_to_v4_mut(struct btree_trans *trans, struct b
 }
 
 struct bkey_i_alloc_v4 *
-bch2_trans_start_alloc_update(struct btree_trans *trans, struct btree_iter *iter,
-			      struct bpos pos)
+bch2_trans_start_alloc_update_noupdate(struct btree_trans *trans, struct btree_iter *iter,
+				       struct bpos pos)
 {
-	struct bkey_s_c k;
-	struct bkey_i_alloc_v4 *a;
-	int ret;
-
-	k = bch2_bkey_get_iter(trans, iter, BTREE_ID_alloc, pos,
-			     BTREE_ITER_WITH_UPDATES|
-			     BTREE_ITER_CACHED|
-			     BTREE_ITER_INTENT);
-	ret = bkey_err(k);
+	struct bkey_s_c k = bch2_bkey_get_iter(trans, iter, BTREE_ID_alloc, pos,
+					       BTREE_ITER_with_updates|
+					       BTREE_ITER_cached|
+					       BTREE_ITER_intent);
+	int ret = bkey_err(k);
 	if (unlikely(ret))
 		return ERR_PTR(ret);
 
-	a = bch2_alloc_to_v4_mut_inlined(trans, k);
+	struct bkey_i_alloc_v4 *a = bch2_alloc_to_v4_mut_inlined(trans, k);
 	ret = PTR_ERR_OR_ZERO(a);
 	if (unlikely(ret))
 		goto err;
@@ -464,6 +450,20 @@ err:
 	return ERR_PTR(ret);
 }
 
+__flatten
+struct bkey_i_alloc_v4 *bch2_trans_start_alloc_update(struct btree_trans *trans, struct bpos pos)
+{
+	struct btree_iter iter;
+	struct bkey_i_alloc_v4 *a = bch2_trans_start_alloc_update_noupdate(trans, &iter, pos);
+	int ret = PTR_ERR_OR_ZERO(a);
+	if (ret)
+		return ERR_PTR(ret);
+
+	ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
+	bch2_trans_iter_exit(trans, &iter);
+	return unlikely(ret) ? ERR_PTR(ret) : a;
+}
+
 static struct bpos alloc_gens_pos(struct bpos pos, unsigned *offset)
 {
 	*offset = pos.offset & KEY_TYPE_BUCKET_GENS_MASK;
@@ -520,7 +520,7 @@ int bch2_bucket_gens_init(struct bch_fs *c)
 	int ret;
 
 	ret = for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
-				 BTREE_ITER_PREFETCH, k, ({
+				 BTREE_ITER_prefetch, k, ({
 		/*
 		 * Not a fsck error because this is checked/repaired by
 		 * bch2_check_alloc_key() which runs later:
@@ -567,29 +567,31 @@ iter_err:
 int bch2_alloc_read(struct bch_fs *c)
 {
 	struct btree_trans *trans = bch2_trans_get(c);
+	struct bch_dev *ca = NULL;
 	int ret;
 
 	down_read(&c->gc_lock);
 
 	if (c->sb.version_upgrade_complete >= bcachefs_metadata_version_bucket_gens) {
 		ret = for_each_btree_key(trans, iter, BTREE_ID_bucket_gens, POS_MIN,
-					 BTREE_ITER_PREFETCH, k, ({
+					 BTREE_ITER_prefetch, k, ({
 			u64 start = bucket_gens_pos_to_alloc(k.k->p, 0).offset;
 			u64 end = bucket_gens_pos_to_alloc(bpos_nosnap_successor(k.k->p), 0).offset;
 
 			if (k.k->type != KEY_TYPE_bucket_gens)
 				continue;
 
-			const struct bch_bucket_gens *g = bkey_s_c_to_bucket_gens(k).v;
-
+			ca = bch2_dev_iterate(c, ca, k.k->p.inode);
 			/*
 			 * Not a fsck error because this is checked/repaired by
 			 * bch2_check_alloc_key() which runs later:
 			 */
-			if (!bch2_dev_exists2(c, k.k->p.inode))
+			if (!ca) {
+				bch2_btree_iter_set_pos(&iter, POS(k.k->p.inode + 1, 0));
 				continue;
+			}
 
-			struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode);
+			const struct bch_bucket_gens *g = bkey_s_c_to_bucket_gens(k).v;
 
 			for (u64 b = max_t(u64, ca->mi.first_bucket, start);
 			     b < min_t(u64, ca->mi.nbuckets, end);
@@ -599,15 +601,16 @@ int bch2_alloc_read(struct bch_fs *c)
 		}));
 	} else {
 		ret = for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
-					 BTREE_ITER_PREFETCH, k, ({
+					 BTREE_ITER_prefetch, k, ({
+			ca = bch2_dev_iterate(c, ca, k.k->p.inode);
 			/*
 			 * Not a fsck error because this is checked/repaired by
 			 * bch2_check_alloc_key() which runs later:
 			 */
-			if (!bch2_dev_bucket_exists(c, k.k->p))
+			if (!ca) {
+				bch2_btree_iter_set_pos(&iter, POS(k.k->p.inode + 1, 0));
 				continue;
-
-			struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode);
+			}
 
 			struct bch_alloc_v4 a;
 			*bucket_gen(ca, k.k->p.offset) = bch2_alloc_to_v4(k, &a)->gen;
@@ -615,6 +618,7 @@ int bch2_alloc_read(struct bch_fs *c)
 		}));
 	}
 
+	bch2_dev_put(ca);
 	bch2_trans_put(trans);
 	up_read(&c->gc_lock);
 
@@ -625,12 +629,12 @@ int bch2_alloc_read(struct bch_fs *c)
 /* Free space/discard btree: */
 
 static int bch2_bucket_do_index(struct btree_trans *trans,
+				struct bch_dev *ca,
 				struct bkey_s_c alloc_k,
 				const struct bch_alloc_v4 *a,
 				bool set)
 {
 	struct bch_fs *c = trans->c;
-	struct bch_dev *ca = bch_dev_bkey_exists(c, alloc_k.k->p.inode);
 	struct btree_iter iter;
 	struct bkey_s_c old;
 	struct bkey_i *k;
@@ -667,7 +671,7 @@ static int bch2_bucket_do_index(struct btree_trans *trans,
 
 	old = bch2_bkey_get_iter(trans, &iter, btree,
 			     bkey_start_pos(&k->k),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	ret = bkey_err(old);
 	if (ret)
 		return ret;
@@ -711,8 +715,8 @@ static noinline int bch2_bucket_gen_update(struct btree_trans *trans,
 		return ret;
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_bucket_gens, pos,
-			       BTREE_ITER_INTENT|
-			       BTREE_ITER_WITH_UPDATES);
+			       BTREE_ITER_intent|
+			       BTREE_ITER_with_updates);
 	ret = bkey_err(k);
 	if (ret)
 		return ret;
@@ -734,26 +738,24 @@ static noinline int bch2_bucket_gen_update(struct btree_trans *trans,
 int bch2_trigger_alloc(struct btree_trans *trans,
 		       enum btree_id btree, unsigned level,
 		       struct bkey_s_c old, struct bkey_s new,
-		       unsigned flags)
+		       enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_fs *c = trans->c;
 	int ret = 0;
 
-	if (bch2_trans_inconsistent_on(!bch2_dev_bucket_exists(c, new.k->p), trans,
-				       "alloc key for invalid device or bucket"))
+	struct bch_dev *ca = bch2_dev_bucket_tryget(c, new.k->p);
+	if (!ca)
 		return -EIO;
 
-	struct bch_dev *ca = bch_dev_bkey_exists(c, new.k->p.inode);
-
 	struct bch_alloc_v4 old_a_convert;
 	const struct bch_alloc_v4 *old_a = bch2_alloc_to_v4(old, &old_a_convert);
 
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (flags & BTREE_TRIGGER_transactional) {
 		struct bch_alloc_v4 *new_a = bkey_s_to_alloc_v4(new).v;
 
-		new_a->data_type = alloc_data_type(*new_a, new_a->data_type);
+		alloc_data_type_set(new_a, new_a->data_type);
 
-		if (bch2_bucket_sectors(*new_a) > bch2_bucket_sectors(*old_a)) {
+		if (bch2_bucket_sectors_total(*new_a) > bch2_bucket_sectors_total(*old_a)) {
 			new_a->io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now));
 			new_a->io_time[WRITE]= max_t(u64, 1, atomic64_read(&c->io_clock[WRITE].now));
 			SET_BCH_ALLOC_V4_NEED_INC_GEN(new_a, true);
@@ -770,10 +772,10 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 		if (old_a->data_type != new_a->data_type ||
 		    (new_a->data_type == BCH_DATA_free &&
 		     alloc_freespace_genbits(*old_a) != alloc_freespace_genbits(*new_a))) {
-			ret =   bch2_bucket_do_index(trans, old, old_a, false) ?:
-				bch2_bucket_do_index(trans, new.s_c, new_a, true);
+			ret =   bch2_bucket_do_index(trans, ca, old, old_a, false) ?:
+				bch2_bucket_do_index(trans, ca, new.s_c, new_a, true);
 			if (ret)
-				return ret;
+				goto err;
 		}
 
 		if (new_a->data_type == BCH_DATA_cached &&
@@ -787,24 +789,23 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 					      bucket_to_u64(new.k->p),
 					      old_lru, new_lru);
 			if (ret)
-				return ret;
+				goto err;
 		}
 
-		new_a->fragmentation_lru = alloc_lru_idx_fragmentation(*new_a,
-						bch_dev_bkey_exists(c, new.k->p.inode));
+		new_a->fragmentation_lru = alloc_lru_idx_fragmentation(*new_a, ca);
 		if (old_a->fragmentation_lru != new_a->fragmentation_lru) {
 			ret = bch2_lru_change(trans,
 					BCH_LRU_FRAGMENTATION_START,
 					bucket_to_u64(new.k->p),
 					old_a->fragmentation_lru, new_a->fragmentation_lru);
 			if (ret)
-				return ret;
+				goto err;
 		}
 
 		if (old_a->gen != new_a->gen) {
 			ret = bch2_bucket_gen_update(trans, new.k->p, new_a->gen);
 			if (ret)
-				return ret;
+				goto err;
 		}
 
 		/*
@@ -812,21 +813,21 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 		 * not:
 		 */
 
-		if ((flags & BTREE_TRIGGER_BUCKET_INVALIDATE) &&
+		if ((flags & BTREE_TRIGGER_bucket_invalidate) &&
 		    old_a->cached_sectors) {
 			ret = bch2_update_cached_sectors_list(trans, new.k->p.inode,
 							      -((s64) old_a->cached_sectors));
 			if (ret)
-				return ret;
+				goto err;
 		}
 	}
 
-	if ((flags & BTREE_TRIGGER_ATOMIC) && (flags & BTREE_TRIGGER_INSERT)) {
+	if ((flags & BTREE_TRIGGER_atomic) && (flags & BTREE_TRIGGER_insert)) {
 		struct bch_alloc_v4 *new_a = bkey_s_to_alloc_v4(new).v;
 		u64 journal_seq = trans->journal_res.seq;
 		u64 bucket_journal_seq = new_a->journal_seq;
 
-		if ((flags & BTREE_TRIGGER_INSERT) &&
+		if ((flags & BTREE_TRIGGER_insert) &&
 		    data_type_is_empty(old_a->data_type) !=
 		    data_type_is_empty(new_a->data_type) &&
 		    new.k->type == KEY_TYPE_alloc_v4) {
@@ -854,7 +855,7 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 			if (ret) {
 				bch2_fs_fatal_error(c,
 					"setting bucket_needs_journal_commit: %s", bch2_err_str(ret));
-				return ret;
+				goto err;
 			}
 		}
 
@@ -884,11 +885,11 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 			bch2_do_invalidates(c);
 
 		if (statechange(a->data_type == BCH_DATA_need_gc_gens))
-			bch2_do_gc_gens(c);
+			bch2_gc_gens_async(c);
 	}
 
-	if ((flags & BTREE_TRIGGER_GC) &&
-	    (flags & BTREE_TRIGGER_BUCKET_INVALIDATE)) {
+	if ((flags & BTREE_TRIGGER_gc) &&
+	    (flags & BTREE_TRIGGER_bucket_invalidate)) {
 		struct bch_alloc_v4 new_a_convert;
 		const struct bch_alloc_v4 *new_a = bch2_alloc_to_v4(new.s_c, &new_a_convert);
 
@@ -908,12 +909,13 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 		bucket_unlock(g);
 		percpu_up_read(&c->mark_lock);
 	}
-
-	return 0;
+err:
+	bch2_dev_put(ca);
+	return ret;
 }
 
 /*
- * This synthesizes deleted extents for holes, similar to BTREE_ITER_SLOTS for
+ * This synthesizes deleted extents for holes, similar to BTREE_ITER_slots for
  * extents style btrees, but works on non-extents btrees:
  */
 static struct bkey_s_c bch2_get_key_or_hole(struct btree_iter *iter, struct bpos end, struct bkey *hole)
@@ -965,8 +967,8 @@ static bool next_bucket(struct bch_fs *c, struct bpos *bucket)
 	if (bch2_dev_bucket_exists(c, *bucket))
 		return true;
 
-	if (bch2_dev_exists2(c, bucket->inode)) {
-		ca = bch_dev_bkey_exists(c, bucket->inode);
+	if (bch2_dev_exists(c, bucket->inode)) {
+		ca = bch2_dev_bkey_exists(c, bucket->inode);
 
 		if (bucket->offset < ca->mi.first_bucket) {
 			bucket->offset = ca->mi.first_bucket;
@@ -1007,7 +1009,7 @@ again:
 		}
 
 		if (!bch2_dev_bucket_exists(c, k.k->p)) {
-			struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
+			struct bch_dev *ca = bch2_dev_bkey_exists(c, bucket.inode);
 
 			bch2_key_resize(hole, ca->mi.nbuckets - bucket.offset);
 		}
@@ -1025,24 +1027,25 @@ int bch2_check_alloc_key(struct btree_trans *trans,
 			 struct btree_iter *bucket_gens_iter)
 {
 	struct bch_fs *c = trans->c;
-	struct bch_dev *ca;
 	struct bch_alloc_v4 a_convert;
 	const struct bch_alloc_v4 *a;
 	unsigned discard_key_type, freespace_key_type;
 	unsigned gens_offset;
 	struct bkey_s_c k;
 	struct printbuf buf = PRINTBUF;
-	int ret;
+	int ret = 0;
 
-	if (fsck_err_on(!bch2_dev_bucket_exists(c, alloc_k.k->p), c,
-			alloc_key_to_missing_dev_bucket,
+	struct bch_dev *ca = bch2_dev_bucket_tryget_noerror(c, alloc_k.k->p);
+	if (fsck_err_on(!ca,
+			c, alloc_key_to_missing_dev_bucket,
 			"alloc key for invalid device:bucket %llu:%llu",
 			alloc_k.k->p.inode, alloc_k.k->p.offset))
-		return bch2_btree_delete_at(trans, alloc_iter, 0);
+		ret = bch2_btree_delete_at(trans, alloc_iter, 0);
+	if (!ca)
+		return ret;
 
-	ca = bch_dev_bkey_exists(c, alloc_k.k->p.inode);
 	if (!ca->mi.freespace_initialized)
-		return 0;
+		goto out;
 
 	a = bch2_alloc_to_v4(alloc_k, &a_convert);
 
@@ -1141,8 +1144,10 @@ int bch2_check_alloc_key(struct btree_trans *trans,
 		if (ret)
 			goto err;
 	}
+out:
 err:
 fsck_err:
+	bch2_dev_put(ca);
 	printbuf_exit(&buf);
 	return ret;
 }
@@ -1159,7 +1164,7 @@ int bch2_check_alloc_hole_freespace(struct btree_trans *trans,
 	struct printbuf buf = PRINTBUF;
 	int ret;
 
-	ca = bch_dev_bkey_exists(c, start.inode);
+	ca = bch2_dev_bkey_exists(c, start.inode);
 	if (!ca->mi.freespace_initialized)
 		return 0;
 
@@ -1313,7 +1318,7 @@ static noinline_for_stack int bch2_check_discard_freespace_key(struct btree_tran
 		goto delete;
 out:
 fsck_err:
-	set_btree_iter_dontneed(&alloc_iter);
+	bch2_set_btree_iter_dontneed(&alloc_iter);
 	bch2_trans_iter_exit(trans, &alloc_iter);
 	printbuf_exit(&buf);
 	return ret;
@@ -1349,7 +1354,7 @@ int bch2_check_bucket_gens_key(struct btree_trans *trans,
 	bkey_reassemble(&g.k_i, k);
 
 	/* if no bch_dev, skip out whether we repair or not */
-	dev_exists = bch2_dev_exists2(c, k.k->p.inode);
+	dev_exists = bch2_dev_exists(c, k.k->p.inode);
 	if (!dev_exists) {
 		if (fsck_err_on(!dev_exists, c,
 				bucket_gens_to_invalid_dev,
@@ -1360,7 +1365,7 @@ int bch2_check_bucket_gens_key(struct btree_trans *trans,
 		goto out;
 	}
 
-	ca = bch_dev_bkey_exists(c, k.k->p.inode);
+	ca = bch2_dev_bkey_exists(c, k.k->p.inode);
 	if (fsck_err_on(end <= ca->mi.first_bucket ||
 			start >= ca->mi.nbuckets, c,
 			bucket_gens_to_invalid_buckets,
@@ -1411,13 +1416,13 @@ int bch2_check_alloc_info(struct bch_fs *c)
 	int ret = 0;
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, POS_MIN,
-			     BTREE_ITER_PREFETCH);
+			     BTREE_ITER_prefetch);
 	bch2_trans_iter_init(trans, &discard_iter, BTREE_ID_need_discard, POS_MIN,
-			     BTREE_ITER_PREFETCH);
+			     BTREE_ITER_prefetch);
 	bch2_trans_iter_init(trans, &freespace_iter, BTREE_ID_freespace, POS_MIN,
-			     BTREE_ITER_PREFETCH);
+			     BTREE_ITER_prefetch);
 	bch2_trans_iter_init(trans, &bucket_gens_iter, BTREE_ID_bucket_gens, POS_MIN,
-			     BTREE_ITER_PREFETCH);
+			     BTREE_ITER_prefetch);
 
 	while (1) {
 		struct bpos next;
@@ -1479,13 +1484,13 @@ bkey_err:
 
 	ret = for_each_btree_key(trans, iter,
 			BTREE_ID_need_discard, POS_MIN,
-			BTREE_ITER_PREFETCH, k,
+			BTREE_ITER_prefetch, k,
 		bch2_check_discard_freespace_key(trans, &iter));
 	if (ret)
 		goto err;
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_freespace, POS_MIN,
-			     BTREE_ITER_PREFETCH);
+			     BTREE_ITER_prefetch);
 	while (1) {
 		bch2_trans_begin(trans);
 		k = bch2_btree_iter_peek(&iter);
@@ -1515,7 +1520,7 @@ bkey_err:
 
 	ret = for_each_btree_key_commit(trans, iter,
 			BTREE_ID_bucket_gens, POS_MIN,
-			BTREE_ITER_PREFETCH, k,
+			BTREE_ITER_prefetch, k,
 			NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 		bch2_check_bucket_gens_key(trans, &iter, k));
 err:
@@ -1562,7 +1567,7 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 
 		a_mut->v.io_time[READ] = atomic64_read(&c->io_clock[READ].now);
 		ret = bch2_trans_update(trans, alloc_iter,
-					&a_mut->k_i, BTREE_TRIGGER_NORUN);
+					&a_mut->k_i, BTREE_TRIGGER_norun);
 		if (ret)
 			goto err;
 
@@ -1601,7 +1606,7 @@ int bch2_check_alloc_to_lru_refs(struct bch_fs *c)
 {
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_alloc,
-				POS_MIN, BTREE_ITER_PREFETCH, k,
+				POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			bch2_check_alloc_to_lru_ref(trans, &iter)));
 	bch_err_fn(c, ret);
@@ -1679,7 +1684,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 	bool discard_locked = false;
 	int ret = 0;
 
-	ca = bch_dev_bkey_exists(c, pos.inode);
+	ca = bch2_dev_bkey_exists(c, pos.inode);
 
 	if (!percpu_ref_tryget(&ca->io_ref)) {
 		bch2_btree_iter_set_pos(need_discard_iter, POS(pos.inode + 1, 0));
@@ -1703,7 +1708,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
 			       need_discard_iter->pos,
-			       BTREE_ITER_CACHED);
+			       BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (ret)
 		goto out;
@@ -1713,7 +1718,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 	if (ret)
 		goto out;
 
-	if (a->v.dirty_sectors) {
+	if (bch2_bucket_sectors_total(a->v)) {
 		if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info,
 					       trans, "attempting to discard bucket with dirty data\n%s",
 					       (bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
@@ -1771,7 +1776,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 	}
 
 	SET_BCH_ALLOC_V4_NEED_DISCARD(&a->v, false);
-	a->v.data_type = alloc_data_type(a->v, a->v.data_type);
+	alloc_data_type_set(&a->v, a->v.data_type);
 write:
 	ret =   bch2_trans_update(trans, &iter, &a->k_i, 0) ?:
 		bch2_trans_commit(trans, NULL, NULL,
@@ -1827,7 +1832,7 @@ void bch2_do_discards(struct bch_fs *c)
 static int bch2_clear_bucket_needs_discard(struct btree_trans *trans, struct bpos bucket)
 {
 	struct btree_iter iter;
-	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, bucket, BTREE_ITER_INTENT);
+	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, bucket, BTREE_ITER_intent);
 	struct bkey_s_c k = bch2_btree_iter_peek_slot(&iter);
 	int ret = bkey_err(k);
 	if (ret)
@@ -1840,7 +1845,7 @@ static int bch2_clear_bucket_needs_discard(struct btree_trans *trans, struct bpo
 
 	BUG_ON(a->v.dirty_sectors);
 	SET_BCH_ALLOC_V4_NEED_DISCARD(&a->v, false);
-	a->v.data_type = alloc_data_type(a->v, a->v.data_type);
+	alloc_data_type_set(&a->v, a->v.data_type);
 
 	ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
 err:
@@ -1862,7 +1867,7 @@ static void bch2_do_discards_fast_work(struct work_struct *work)
 			if (i->snapshot)
 				continue;
 
-			ca = bch_dev_bkey_exists(c, i->inode);
+			ca = bch2_dev_bkey_exists(c, i->inode);
 
 			if (!percpu_ref_tryget(&ca->io_ref)) {
 				darray_remove_item(&c->discard_buckets_in_flight, i);
@@ -1903,7 +1908,7 @@ static void bch2_do_discards_fast_work(struct work_struct *work)
 
 static void bch2_discard_one_bucket_fast(struct bch_fs *c, struct bpos bucket)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
+	struct bch_dev *ca = bch2_dev_bkey_exists(c, bucket.inode);
 
 	if (!percpu_ref_is_dying(&ca->io_ref) &&
 	    !discard_in_flight_add(c, bucket) &&
@@ -1918,7 +1923,6 @@ static int invalidate_one_bucket(struct btree_trans *trans,
 				 s64 *nr_to_invalidate)
 {
 	struct bch_fs *c = trans->c;
-	struct btree_iter alloc_iter = { NULL };
 	struct bkey_i_alloc_v4 *a = NULL;
 	struct printbuf buf = PRINTBUF;
 	struct bpos bucket = u64_to_bucket(lru_k.k->p.offset);
@@ -1936,7 +1940,7 @@ static int invalidate_one_bucket(struct btree_trans *trans,
 	if (bch2_bucket_is_open_safe(c, bucket.inode, bucket.offset))
 		return 0;
 
-	a = bch2_trans_start_alloc_update(trans, &alloc_iter, bucket);
+	a = bch2_trans_start_alloc_update(trans, bucket);
 	ret = PTR_ERR_OR_ZERO(a);
 	if (ret)
 		goto out;
@@ -1961,18 +1965,15 @@ static int invalidate_one_bucket(struct btree_trans *trans,
 	a->v.io_time[READ]	= atomic64_read(&c->io_clock[READ].now);
 	a->v.io_time[WRITE]	= atomic64_read(&c->io_clock[WRITE].now);
 
-	ret =   bch2_trans_update(trans, &alloc_iter, &a->k_i,
-				BTREE_TRIGGER_BUCKET_INVALIDATE) ?:
-		bch2_trans_commit(trans, NULL, NULL,
-				  BCH_WATERMARK_btree|
-				  BCH_TRANS_COMMIT_no_enospc);
+	ret = bch2_trans_commit(trans, NULL, NULL,
+				BCH_WATERMARK_btree|
+				BCH_TRANS_COMMIT_no_enospc);
 	if (ret)
 		goto out;
 
 	trace_and_count(c, bucket_invalidate, c, bucket.inode, bucket.offset, cached_sectors);
 	--*nr_to_invalidate;
 out:
-	bch2_trans_iter_exit(trans, &alloc_iter);
 	printbuf_exit(&buf);
 	return ret;
 err:
@@ -2014,7 +2015,7 @@ static void bch2_do_invalidates_work(struct work_struct *work)
 		ret = for_each_btree_key_upto(trans, iter, BTREE_ID_lru,
 				lru_pos(ca->dev_idx, 0, 0),
 				lru_pos(ca->dev_idx, U64_MAX, LRU_TIME_MAX),
-				BTREE_ITER_INTENT, k,
+				BTREE_ITER_intent, k,
 			invalidate_one_bucket(trans, &iter, k, &nr_to_invalidate));
 
 		if (ret < 0) {
@@ -2051,7 +2052,7 @@ int bch2_dev_freespace_init(struct bch_fs *c, struct bch_dev *ca,
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
 		POS(ca->dev_idx, max_t(u64, ca->mi.first_bucket, bucket_start)),
-		BTREE_ITER_PREFETCH);
+		BTREE_ITER_prefetch);
 	/*
 	 * Scan the alloc btree for every bucket on @ca, and add buckets to the
 	 * freespace/need_discard/need_gc_gens btrees as needed:
@@ -2083,7 +2084,7 @@ int bch2_dev_freespace_init(struct bch_fs *c, struct bch_dev *ca,
 			struct bch_alloc_v4 a_convert;
 			const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert);
 
-			ret =   bch2_bucket_do_index(trans, k, a, true) ?:
+			ret =   bch2_bucket_do_index(trans, ca, k, a, true) ?:
 				bch2_trans_commit(trans, NULL, NULL,
 						  BCH_TRANS_COMMIT_no_enospc);
 			if (ret)
@@ -2182,7 +2183,10 @@ int bch2_bucket_io_time_reset(struct btree_trans *trans, unsigned dev,
 	u64 now;
 	int ret = 0;
 
-	a = bch2_trans_start_alloc_update(trans, &iter,  POS(dev, bucket_nr));
+	if (bch2_trans_relock(trans))
+		bch2_trans_begin(trans);
+
+	a = bch2_trans_start_alloc_update_noupdate(trans, &iter, POS(dev, bucket_nr));
 	ret = PTR_ERR_OR_ZERO(a);
 	if (ret)
 		return ret;
diff --git a/fs/bcachefs/alloc_background.h b/fs/bcachefs/alloc_background.h
index 052b2fac25d693..3b78e35c111e85 100644
--- a/fs/bcachefs/alloc_background.h
+++ b/fs/bcachefs/alloc_background.h
@@ -15,14 +15,11 @@ enum bkey_invalid_flags;
 
 static inline bool bch2_dev_bucket_exists(struct bch_fs *c, struct bpos pos)
 {
-	struct bch_dev *ca;
-
-	if (!bch2_dev_exists2(c, pos.inode))
-		return false;
-
-	ca = bch_dev_bkey_exists(c, pos.inode);
-	return pos.offset >= ca->mi.first_bucket &&
-		pos.offset < ca->mi.nbuckets;
+	rcu_read_lock();
+	struct bch_dev *ca = bch2_dev_rcu(c, pos.inode);
+	bool ret = ca && bucket_valid(ca, pos.offset);
+	rcu_read_unlock();
+	return ret;
 }
 
 static inline u64 bucket_to_u64(struct bpos bucket)
@@ -40,38 +37,50 @@ static inline u8 alloc_gc_gen(struct bch_alloc_v4 a)
 	return a.gen - a.oldest_gen;
 }
 
-static inline enum bch_data_type __alloc_data_type(u32 dirty_sectors,
-						   u32 cached_sectors,
-						   u32 stripe,
-						   struct bch_alloc_v4 a,
-						   enum bch_data_type data_type)
+static inline void alloc_to_bucket(struct bucket *dst, struct bch_alloc_v4 src)
 {
-	if (stripe)
-		return data_type == BCH_DATA_parity ? data_type : BCH_DATA_stripe;
-	if (dirty_sectors)
-		return data_type;
-	if (cached_sectors)
-		return BCH_DATA_cached;
-	if (BCH_ALLOC_V4_NEED_DISCARD(&a))
-		return BCH_DATA_need_discard;
-	if (alloc_gc_gen(a) >= BUCKET_GC_GEN_MAX)
-		return BCH_DATA_need_gc_gens;
-	return BCH_DATA_free;
+	dst->gen		= src.gen;
+	dst->data_type		= src.data_type;
+	dst->dirty_sectors	= src.dirty_sectors;
+	dst->cached_sectors	= src.cached_sectors;
+	dst->stripe		= src.stripe;
 }
 
-static inline enum bch_data_type alloc_data_type(struct bch_alloc_v4 a,
-						 enum bch_data_type data_type)
+static inline void __bucket_m_to_alloc(struct bch_alloc_v4 *dst, struct bucket src)
 {
-	return __alloc_data_type(a.dirty_sectors, a.cached_sectors,
-				 a.stripe, a, data_type);
+	dst->gen		= src.gen;
+	dst->data_type		= src.data_type;
+	dst->dirty_sectors	= src.dirty_sectors;
+	dst->cached_sectors	= src.cached_sectors;
+	dst->stripe		= src.stripe;
+}
+
+static inline struct bch_alloc_v4 bucket_m_to_alloc(struct bucket b)
+{
+	struct bch_alloc_v4 ret = {};
+	__bucket_m_to_alloc(&ret, b);
+	return ret;
 }
 
 static inline enum bch_data_type bucket_data_type(enum bch_data_type data_type)
 {
-	return data_type == BCH_DATA_stripe ? BCH_DATA_user : data_type;
+	switch (data_type) {
+	case BCH_DATA_cached:
+	case BCH_DATA_stripe:
+		return BCH_DATA_user;
+	default:
+		return data_type;
+	}
+}
+
+static inline bool bucket_data_type_mismatch(enum bch_data_type bucket,
+					     enum bch_data_type ptr)
+{
+	return !data_type_is_empty(bucket) &&
+		bucket_data_type(bucket) != bucket_data_type(ptr);
 }
 
-static inline unsigned bch2_bucket_sectors(struct bch_alloc_v4 a)
+static inline unsigned bch2_bucket_sectors_total(struct bch_alloc_v4 a)
 {
 	return a.dirty_sectors + a.cached_sectors;
 }
@@ -89,6 +98,27 @@ static inline unsigned bch2_bucket_sectors_fragmented(struct bch_dev *ca,
 	return d ? max(0, ca->mi.bucket_size - d) : 0;
 }
 
+static inline enum bch_data_type alloc_data_type(struct bch_alloc_v4 a,
+						 enum bch_data_type data_type)
+{
+	if (a.stripe)
+		return data_type == BCH_DATA_parity ? data_type : BCH_DATA_stripe;
+	if (a.dirty_sectors)
+		return data_type;
+	if (a.cached_sectors)
+		return BCH_DATA_cached;
+	if (BCH_ALLOC_V4_NEED_DISCARD(&a))
+		return BCH_DATA_need_discard;
+	if (alloc_gc_gen(a) >= BUCKET_GC_GEN_MAX)
+		return BCH_DATA_need_gc_gens;
+	return BCH_DATA_free;
+}
+
+static inline void alloc_data_type_set(struct bch_alloc_v4 *a, enum bch_data_type data_type)
+{
+	a->data_type = alloc_data_type(*a, data_type);
+}
+
 static inline u64 alloc_lru_idx_read(struct bch_alloc_v4 a)
 {
 	return a.data_type == BCH_DATA_cached ? a.io_time[READ] : 0;
@@ -143,7 +173,9 @@ static inline void set_alloc_v4_u64s(struct bkey_i_alloc_v4 *a)
 }
 
 struct bkey_i_alloc_v4 *
-bch2_trans_start_alloc_update(struct btree_trans *, struct btree_iter *, struct bpos);
+bch2_trans_start_alloc_update_noupdate(struct btree_trans *, struct btree_iter *, struct bpos);
+struct bkey_i_alloc_v4 *
+bch2_trans_start_alloc_update(struct btree_trans *, struct bpos);
 
 void __bch2_alloc_to_v4(struct bkey_s_c, struct bch_alloc_v4 *);
 
@@ -229,7 +261,8 @@ static inline bool bkey_is_alloc(const struct bkey *k)
 int bch2_alloc_read(struct bch_fs *);
 
 int bch2_trigger_alloc(struct btree_trans *, enum btree_id, unsigned,
-		       struct bkey_s_c, struct bkey_s, unsigned);
+		       struct bkey_s_c, struct bkey_s,
+		       enum btree_iter_update_trigger_flags);
 int bch2_check_alloc_info(struct bch_fs *);
 int bch2_check_alloc_to_lru_refs(struct bch_fs *);
 void bch2_do_discards(struct bch_fs *);
diff --git a/fs/bcachefs/alloc_foreground.c b/fs/bcachefs/alloc_foreground.c
index a1fc30adf9129d..fa6bf5b7cb2749 100644
--- a/fs/bcachefs/alloc_foreground.c
+++ b/fs/bcachefs/alloc_foreground.c
@@ -100,7 +100,7 @@ static void bch2_open_bucket_hash_remove(struct bch_fs *c, struct open_bucket *o
 
 void __bch2_open_bucket_put(struct bch_fs *c, struct open_bucket *ob)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
+	struct bch_dev *ca = ob_dev(c, ob);
 
 	if (ob->ec) {
 		ec_stripe_new_put(c, ob->ec, STRIPE_REF_io);
@@ -203,7 +203,7 @@ static inline unsigned open_buckets_reserved(enum bch_watermark watermark)
 }
 
 static struct open_bucket *__try_alloc_bucket(struct bch_fs *c, struct bch_dev *ca,
-					      u64 bucket,
+					      u64 *bucket,
 					      enum bch_watermark watermark,
 					      const struct bch_alloc_v4 *a,
 					      struct bucket_alloc_state *s,
@@ -211,23 +211,34 @@ static struct open_bucket *__try_alloc_bucket(struct bch_fs *c, struct bch_dev *
 {
 	struct open_bucket *ob;
 
-	if (unlikely(ca->buckets_nouse && test_bit(bucket, ca->buckets_nouse))) {
+	if (unlikely(ca->buckets_nouse && test_bit(*bucket, ca->buckets_nouse))) {
 		s->skipped_nouse++;
 		return NULL;
 	}
 
-	if (bch2_bucket_is_open(c, ca->dev_idx, bucket)) {
+	if (s->mi_btree_bitmap_only &&
+	    unlikely(!bch2_dev_btree_bitmap_marked_sectors(ca,
+					bucket_to_sector(ca, *bucket), ca->mi.bucket_size))) {
+
+		u64 sector = bucket_to_sector(ca, *bucket);
+		sector = round_up(sector, 1ULL << ca->mi.btree_bitmap_shift);
+		*bucket = sector_to_bucket(ca, sector);
+		s->skipped_mi_btree_bitmap++;
+		return NULL;
+	}
+
+	if (bch2_bucket_is_open(c, ca->dev_idx, *bucket)) {
 		s->skipped_open++;
 		return NULL;
 	}
 
 	if (bch2_bucket_needs_journal_commit(&c->buckets_waiting_for_journal,
-			c->journal.flushed_seq_ondisk, ca->dev_idx, bucket)) {
+			c->journal.flushed_seq_ondisk, ca->dev_idx, *bucket)) {
 		s->skipped_need_journal_commit++;
 		return NULL;
 	}
 
-	if (bch2_bucket_nocow_is_locked(&c->nocow_locks, POS(ca->dev_idx, bucket))) {
+	if (bch2_bucket_nocow_is_locked(&c->nocow_locks, POS(ca->dev_idx, *bucket))) {
 		s->skipped_nocow++;
 		return NULL;
 	}
@@ -244,7 +255,7 @@ static struct open_bucket *__try_alloc_bucket(struct bch_fs *c, struct bch_dev *
 	}
 
 	/* Recheck under lock: */
-	if (bch2_bucket_is_open(c, ca->dev_idx, bucket)) {
+	if (bch2_bucket_is_open(c, ca->dev_idx, *bucket)) {
 		spin_unlock(&c->freelist_lock);
 		s->skipped_open++;
 		return NULL;
@@ -258,7 +269,7 @@ static struct open_bucket *__try_alloc_bucket(struct bch_fs *c, struct bch_dev *
 	ob->sectors_free = ca->mi.bucket_size;
 	ob->dev		= ca->dev_idx;
 	ob->gen		= a->gen;
-	ob->bucket	= bucket;
+	ob->bucket	= *bucket;
 	spin_unlock(&ob->lock);
 
 	ca->nr_open_buckets++;
@@ -272,7 +283,8 @@ static struct open_bucket *__try_alloc_bucket(struct bch_fs *c, struct bch_dev *
 }
 
 static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bch_dev *ca,
-					    enum bch_watermark watermark, u64 free_entry,
+					    enum bch_watermark watermark,
+					    u64 *b, unsigned genbits,
 					    struct bucket_alloc_state *s,
 					    struct bkey_s_c freespace_k,
 					    struct closure *cl)
@@ -283,12 +295,10 @@ static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bc
 	struct open_bucket *ob;
 	struct bch_alloc_v4 a_convert;
 	const struct bch_alloc_v4 *a;
-	u64 b = free_entry & ~(~0ULL << 56);
-	unsigned genbits = free_entry >> 56;
 	struct printbuf buf = PRINTBUF;
 	int ret;
 
-	if (b < ca->mi.first_bucket || b >= ca->mi.nbuckets) {
+	if (*b < ca->mi.first_bucket || *b >= ca->mi.nbuckets) {
 		prt_printf(&buf, "freespace btree has bucket outside allowed range %u-%llu\n"
 		       "  freespace key ",
 			ca->mi.first_bucket, ca->mi.nbuckets);
@@ -299,8 +309,8 @@ static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bc
 	}
 
 	k = bch2_bkey_get_iter(trans, &iter,
-			       BTREE_ID_alloc, POS(ca->dev_idx, b),
-			       BTREE_ITER_CACHED);
+			       BTREE_ID_alloc, POS(ca->dev_idx, *b),
+			       BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (ret) {
 		ob = ERR_PTR(ret);
@@ -342,9 +352,9 @@ static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bc
 		struct bch_backpointer bp;
 		struct bpos bp_pos = POS_MIN;
 
-		ret = bch2_get_next_backpointer(trans, POS(ca->dev_idx, b), -1,
+		ret = bch2_get_next_backpointer(trans, ca, POS(ca->dev_idx, *b), -1,
 						&bp_pos, &bp,
-						BTREE_ITER_NOPRESERVE);
+						BTREE_ITER_nopreserve);
 		if (ret) {
 			ob = ERR_PTR(ret);
 			goto err;
@@ -363,10 +373,10 @@ static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bc
 
 	ob = __try_alloc_bucket(c, ca, b, watermark, a, s, cl);
 	if (!ob)
-		set_btree_iter_dontneed(&iter);
+		bch2_set_btree_iter_dontneed(&iter);
 err:
 	if (iter.path)
-		set_btree_iter_dontneed(&iter);
+		bch2_set_btree_iter_dontneed(&iter);
 	bch2_trans_iter_exit(trans, &iter);
 	printbuf_exit(&buf);
 	return ob;
@@ -404,9 +414,10 @@ bch2_bucket_alloc_early(struct btree_trans *trans,
 	 */
 again:
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_alloc, POS(ca->dev_idx, alloc_cursor),
-			   BTREE_ITER_SLOTS, k, ret) {
+			   BTREE_ITER_slots, k, ret) {
 		struct bch_alloc_v4 a_convert;
 		const struct bch_alloc_v4 *a;
+		u64 bucket = k.k->p.offset;
 
 		if (bkey_ge(k.k->p, POS(ca->dev_idx, ca->mi.nbuckets)))
 			break;
@@ -420,7 +431,7 @@ again:
 			continue;
 
 		/* now check the cached key to serialize concurrent allocs of the bucket */
-		ck = bch2_bkey_get_iter(trans, &citer, BTREE_ID_alloc, k.k->p, BTREE_ITER_CACHED);
+		ck = bch2_bkey_get_iter(trans, &citer, BTREE_ID_alloc, k.k->p, BTREE_ITER_cached);
 		ret = bkey_err(ck);
 		if (ret)
 			break;
@@ -431,12 +442,20 @@ again:
 
 		s->buckets_seen++;
 
-		ob = __try_alloc_bucket(trans->c, ca, k.k->p.offset, watermark, a, s, cl);
+		ob = __try_alloc_bucket(trans->c, ca, &bucket, watermark, a, s, cl);
 next:
-		set_btree_iter_dontneed(&citer);
+		bch2_set_btree_iter_dontneed(&citer);
 		bch2_trans_iter_exit(trans, &citer);
 		if (ob)
 			break;
+
+		/*
+		 * are we skipping over a range due to the mi_btree_allocated
+		 * check?
+		 * -1 because the iterator will still be advanced
+		 */
+		if (bucket > k.k->p.offset)
+			bch2_btree_iter_set_pos(&iter, POS(ca->dev_idx, bucket - 1));
 	}
 	bch2_trans_iter_exit(trans, &iter);
 
@@ -485,12 +504,22 @@ again:
 
 			s->buckets_seen++;
 
+			u64 b = alloc_cursor & ~(~0ULL << 56);
+			u64 genbits = alloc_cursor >> 56;
+
 			ob = try_alloc_bucket(trans, ca, watermark,
-					      alloc_cursor, s, k, cl);
+					      &b, genbits, s, k, cl);
 			if (ob) {
-				set_btree_iter_dontneed(&iter);
+				bch2_set_btree_iter_dontneed(&iter);
 				break;
 			}
+
+			/*
+			 * are we skipping over a range due to the
+			 * mi_btree_allocated check?
+			 */
+			if (b > (alloc_cursor & ~(~0ULL << 56)))
+				alloc_cursor = (b - 1) | (genbits << 56);
 		}
 
 		if (ob || ret)
@@ -516,6 +545,7 @@ again:
  * @trans:	transaction object
  * @ca:		device to allocate from
  * @watermark:	how important is this allocation?
+ * @data_type:	BCH_DATA_journal, btree, user...
  * @cl:		if not NULL, closure to be used to wait if buckets not available
  * @usage:	for secondarily also returning the current device usage
  *
@@ -524,6 +554,7 @@ again:
 static struct open_bucket *bch2_bucket_alloc_trans(struct btree_trans *trans,
 				      struct bch_dev *ca,
 				      enum bch_watermark watermark,
+				      enum bch_data_type data_type,
 				      struct closure *cl,
 				      struct bch_dev_usage *usage)
 {
@@ -531,7 +562,9 @@ static struct open_bucket *bch2_bucket_alloc_trans(struct btree_trans *trans,
 	struct open_bucket *ob = NULL;
 	bool freespace = READ_ONCE(ca->mi.freespace_initialized);
 	u64 avail;
-	struct bucket_alloc_state s = { 0 };
+	struct bucket_alloc_state s = {
+		.mi_btree_bitmap_only = data_type == BCH_DATA_btree,
+	};
 	bool waiting = false;
 again:
 	bch2_dev_usage_read_fast(ca, usage);
@@ -541,7 +574,7 @@ again:
 		bch2_do_discards(c);
 
 	if (usage->d[BCH_DATA_need_gc_gens].buckets > avail)
-		bch2_do_gc_gens(c);
+		bch2_gc_gens_async(c);
 
 	if (should_invalidate_buckets(ca, *usage))
 		bch2_do_invalidates(c);
@@ -569,6 +602,11 @@ alloc:
 	if (s.skipped_need_journal_commit * 2 > avail)
 		bch2_journal_flush_async(&c->journal, NULL);
 
+	if (!ob && s.mi_btree_bitmap_only) {
+		s.mi_btree_bitmap_only = false;
+		goto alloc;
+	}
+
 	if (!ob && freespace && c->curr_recovery_pass <= BCH_RECOVERY_PASS_check_alloc_info) {
 		freespace = false;
 		goto alloc;
@@ -578,6 +616,9 @@ err:
 		ob = ERR_PTR(-BCH_ERR_no_buckets_found);
 
 	if (!IS_ERR(ob))
+		ob->data_type = data_type;
+
+	if (!IS_ERR(ob))
 		trace_and_count(c, bucket_alloc, ca,
 				bch2_watermarks[watermark],
 				ob->bucket,
@@ -605,6 +646,7 @@ err:
 
 struct open_bucket *bch2_bucket_alloc(struct bch_fs *c, struct bch_dev *ca,
 				      enum bch_watermark watermark,
+				      enum bch_data_type data_type,
 				      struct closure *cl)
 {
 	struct bch_dev_usage usage;
@@ -612,7 +654,7 @@ struct open_bucket *bch2_bucket_alloc(struct bch_fs *c, struct bch_dev *ca,
 
 	bch2_trans_do(c, NULL, NULL, 0,
 		      PTR_ERR_OR_ZERO(ob = bch2_bucket_alloc_trans(trans, ca, watermark,
-							cl, &usage)));
+							data_type, cl, &usage)));
 	return ob;
 }
 
@@ -678,8 +720,7 @@ static int add_new_bucket(struct bch_fs *c,
 			   unsigned flags,
 			   struct open_bucket *ob)
 {
-	unsigned durability =
-		bch_dev_bkey_exists(c, ob->dev)->mi.durability;
+	unsigned durability = ob_dev(c, ob)->mi.durability;
 
 	BUG_ON(*nr_effective >= nr_replicas);
 
@@ -738,7 +779,7 @@ int bch2_bucket_alloc_set_trans(struct btree_trans *trans,
 			continue;
 		}
 
-		ob = bch2_bucket_alloc_trans(trans, ca, watermark, cl, &usage);
+		ob = bch2_bucket_alloc_trans(trans, ca, watermark, data_type, cl, &usage);
 		if (!IS_ERR(ob))
 			bch2_dev_stripe_increment_inlined(ca, stripe, &usage);
 		percpu_ref_put(&ca->ref);
@@ -750,8 +791,6 @@ int bch2_bucket_alloc_set_trans(struct btree_trans *trans,
 			continue;
 		}
 
-		ob->data_type = data_type;
-
 		if (add_new_bucket(c, ptrs, devs_may_alloc,
 				   nr_replicas, nr_effective,
 				   have_cache, flags, ob)) {
@@ -836,7 +875,7 @@ static bool want_bucket(struct bch_fs *c,
 			bool *have_cache, bool ec,
 			struct open_bucket *ob)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
+	struct bch_dev *ca = ob_dev(c, ob);
 
 	if (!test_bit(ob->dev, devs_may_alloc->d))
 		return false;
@@ -906,7 +945,7 @@ static int bucket_alloc_set_partial(struct bch_fs *c,
 		struct open_bucket *ob = c->open_buckets + c->open_buckets_partial[i];
 
 		if (want_bucket(c, wp, devs_may_alloc, have_cache, ec, ob)) {
-			struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
+			struct bch_dev *ca = ob_dev(c, ob);
 			struct bch_dev_usage usage;
 			u64 avail;
 
@@ -1291,7 +1330,7 @@ deallocate_extra_replicas(struct bch_fs *c,
 	unsigned i;
 
 	open_bucket_for_each(c, ptrs, ob, i) {
-		unsigned d = bch_dev_bkey_exists(c, ob->dev)->mi.durability;
+		unsigned d = ob_dev(c, ob)->mi.durability;
 
 		if (d && d <= extra_replicas) {
 			extra_replicas -= d;
@@ -1342,6 +1381,10 @@ retry:
 
 	*wp_ret = wp = writepoint_find(trans, write_point.v);
 
+	ret = bch2_trans_relock(trans);
+	if (ret)
+		goto err;
+
 	/* metadata may not allocate on cache devices: */
 	if (wp->data_type != BCH_DATA_user)
 		have_cache = true;
@@ -1444,7 +1487,7 @@ err:
 
 struct bch_extent_ptr bch2_ob_ptr(struct bch_fs *c, struct open_bucket *ob)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
+	struct bch_dev *ca = ob_dev(c, ob);
 
 	return (struct bch_extent_ptr) {
 		.type	= 1 << BCH_EXTENT_ENTRY_ptr,
@@ -1520,7 +1563,7 @@ void bch2_fs_allocator_foreground_init(struct bch_fs *c)
 
 static void bch2_open_bucket_to_text(struct printbuf *out, struct bch_fs *c, struct open_bucket *ob)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
+	struct bch_dev *ca = ob_dev(c, ob);
 	unsigned data_type = ob->data_type;
 	barrier(); /* READ_ONCE() doesn't work on bitfields */
 
diff --git a/fs/bcachefs/alloc_foreground.h b/fs/bcachefs/alloc_foreground.h
index 7aaeec44c7466c..fa6b70749ca099 100644
--- a/fs/bcachefs/alloc_foreground.h
+++ b/fs/bcachefs/alloc_foreground.h
@@ -30,8 +30,14 @@ void bch2_dev_stripe_increment(struct bch_dev *, struct dev_stripe_state *);
 
 long bch2_bucket_alloc_new_fs(struct bch_dev *);
 
+static inline struct bch_dev *ob_dev(struct bch_fs *c, struct open_bucket *ob)
+{
+	return bch2_dev_have_ref(c, ob->dev);
+}
+
 struct open_bucket *bch2_bucket_alloc(struct bch_fs *, struct bch_dev *,
-				      enum bch_watermark, struct closure *);
+				      enum bch_watermark, enum bch_data_type,
+				      struct closure *);
 
 static inline void ob_push(struct bch_fs *c, struct open_buckets *obs,
 			   struct open_bucket *ob)
@@ -184,7 +190,7 @@ bch2_alloc_sectors_append_ptrs_inlined(struct bch_fs *c, struct write_point *wp,
 	wp->sectors_allocated	+= sectors;
 
 	open_bucket_for_each(c, &wp->ptrs, ob, i) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
+		struct bch_dev *ca = ob_dev(c, ob);
 		struct bch_extent_ptr ptr = bch2_ob_ptr(c, ob);
 
 		ptr.cached = cached ||
diff --git a/fs/bcachefs/alloc_types.h b/fs/bcachefs/alloc_types.h
index c2226e947c41fb..72b1c8e054c42d 100644
--- a/fs/bcachefs/alloc_types.h
+++ b/fs/bcachefs/alloc_types.h
@@ -9,11 +9,13 @@
 #include "fifo.h"
 
 struct bucket_alloc_state {
+	bool	mi_btree_bitmap_only;
 	u64	buckets_seen;
 	u64	skipped_open;
 	u64	skipped_need_journal_commit;
 	u64	skipped_nocow;
 	u64	skipped_nouse;
+	u64	skipped_mi_btree_bitmap;
 };
 
 #define BCH_WATERMARKS()		\
diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c
index a200442010025a..629903d1cb1b75 100644
--- a/fs/bcachefs/backpointers.c
+++ b/fs/bcachefs/backpointers.c
@@ -23,6 +23,7 @@ static bool extent_matches_bp(struct bch_fs *c,
 	const union bch_extent_entry *entry;
 	struct extent_ptr_decoded p;
 
+	rcu_read_lock();
 	bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
 		struct bpos bucket2;
 		struct bch_backpointer bp2;
@@ -30,11 +31,18 @@ static bool extent_matches_bp(struct bch_fs *c,
 		if (p.ptr.cached)
 			continue;
 
-		bch2_extent_ptr_to_bp(c, btree_id, level, k, p, entry, &bucket2, &bp2);
+		struct bch_dev *ca = bch2_dev_rcu(c, p.ptr.dev);
+		if (!ca)
+			continue;
+
+		bch2_extent_ptr_to_bp(c, ca, btree_id, level, k, p, entry, &bucket2, &bp2);
 		if (bpos_eq(bucket, bucket2) &&
-		    !memcmp(&bp, &bp2, sizeof(bp)))
+		    !memcmp(&bp, &bp2, sizeof(bp))) {
+			rcu_read_unlock();
 			return true;
+		}
 	}
+	rcu_read_unlock();
 
 	return false;
 }
@@ -45,16 +53,21 @@ int bch2_backpointer_invalid(struct bch_fs *c, struct bkey_s_c k,
 {
 	struct bkey_s_c_backpointer bp = bkey_s_c_to_backpointer(k);
 
-	/* these will be caught by fsck */
-	if (!bch2_dev_exists2(c, bp.k->p.inode))
+	rcu_read_lock();
+	struct bch_dev *ca = bch2_dev_rcu(c, bp.k->p.inode);
+	if (!ca) {
+		/* these will be caught by fsck */
+		rcu_read_unlock();
 		return 0;
+	}
 
-	struct bch_dev *ca = bch_dev_bkey_exists(c, bp.k->p.inode);
-	struct bpos bucket = bp_pos_to_bucket(c, bp.k->p);
+	struct bpos bucket = bp_pos_to_bucket(ca, bp.k->p);
+	struct bpos bp_pos = bucket_pos_to_bp(ca, bucket, bp.v->bucket_offset);
+	rcu_read_unlock();
 	int ret = 0;
 
 	bkey_fsck_err_on((bp.v->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT) >= ca->mi.bucket_size ||
-			 !bpos_eq(bp.k->p, bucket_pos_to_bp(c, bucket, bp.v->bucket_offset)),
+			 !bpos_eq(bp.k->p, bp_pos),
 			 c, err,
 			 backpointer_bucket_offset_wrong,
 			 "backpointer bucket_offset wrong");
@@ -75,10 +88,16 @@ void bch2_backpointer_to_text(struct printbuf *out, const struct bch_backpointer
 
 void bch2_backpointer_k_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c k)
 {
-	if (bch2_dev_exists2(c, k.k->p.inode)) {
+	rcu_read_lock();
+	struct bch_dev *ca = bch2_dev_rcu(c, k.k->p.inode);
+	if (ca) {
+		struct bpos bucket = bp_pos_to_bucket(ca, k.k->p);
+		rcu_read_unlock();
 		prt_str(out, "bucket=");
-		bch2_bpos_to_text(out, bp_pos_to_bucket(c, k.k->p));
+		bch2_bpos_to_text(out, bucket);
 		prt_str(out, " ");
+	} else {
+		rcu_read_unlock();
 	}
 
 	bch2_backpointer_to_text(out, bkey_s_c_to_backpointer(k).v);
@@ -117,8 +136,7 @@ static noinline int backpointer_mod_err(struct btree_trans *trans,
 
 		bch_err(c, "%s", buf.buf);
 	} else if (c->curr_recovery_pass > BCH_RECOVERY_PASS_check_extents_to_backpointers) {
-		prt_printf(&buf, "backpointer not found when deleting");
-		prt_newline(&buf);
+		prt_printf(&buf, "backpointer not found when deleting\n");
 		printbuf_indent_add(&buf, 2);
 
 		prt_printf(&buf, "searching for ");
@@ -145,6 +163,7 @@ static noinline int backpointer_mod_err(struct btree_trans *trans,
 }
 
 int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *trans,
+				struct bch_dev *ca,
 				struct bpos bucket,
 				struct bch_backpointer bp,
 				struct bkey_s_c orig_k,
@@ -161,7 +180,7 @@ int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *trans,
 		return ret;
 
 	bkey_backpointer_init(&bp_k->k_i);
-	bp_k->k.p = bucket_pos_to_bp(trans->c, bucket, bp.bucket_offset);
+	bp_k->k.p = bucket_pos_to_bp(ca, bucket, bp.bucket_offset);
 	bp_k->v = bp;
 
 	if (!insert) {
@@ -171,9 +190,9 @@ int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *trans,
 
 	k = bch2_bkey_get_iter(trans, &bp_iter, BTREE_ID_backpointers,
 			       bp_k->k.p,
-			       BTREE_ITER_INTENT|
-			       BTREE_ITER_SLOTS|
-			       BTREE_ITER_WITH_UPDATES);
+			       BTREE_ITER_intent|
+			       BTREE_ITER_slots|
+			       BTREE_ITER_with_updates);
 	ret = bkey_err(k);
 	if (ret)
 		goto err;
@@ -197,13 +216,13 @@ err:
  * Find the next backpointer >= *bp_offset:
  */
 int bch2_get_next_backpointer(struct btree_trans *trans,
+			      struct bch_dev *ca,
 			      struct bpos bucket, int gen,
 			      struct bpos *bp_pos,
 			      struct bch_backpointer *bp,
 			      unsigned iter_flags)
 {
-	struct bch_fs *c = trans->c;
-	struct bpos bp_end_pos = bucket_pos_to_bp(c, bpos_nosnap_successor(bucket), 0);
+	struct bpos bp_end_pos = bucket_pos_to_bp(ca, bpos_nosnap_successor(bucket), 0);
 	struct btree_iter alloc_iter = { NULL }, bp_iter = { NULL };
 	struct bkey_s_c k;
 	int ret = 0;
@@ -213,7 +232,7 @@ int bch2_get_next_backpointer(struct btree_trans *trans,
 
 	if (gen >= 0) {
 		k = bch2_bkey_get_iter(trans, &alloc_iter, BTREE_ID_alloc,
-				       bucket, BTREE_ITER_CACHED|iter_flags);
+				       bucket, BTREE_ITER_cached|iter_flags);
 		ret = bkey_err(k);
 		if (ret)
 			goto out;
@@ -223,7 +242,7 @@ int bch2_get_next_backpointer(struct btree_trans *trans,
 			goto done;
 	}
 
-	*bp_pos = bpos_max(*bp_pos, bucket_pos_to_bp(c, bucket, 0));
+	*bp_pos = bpos_max(*bp_pos, bucket_pos_to_bp(ca, bucket, 0));
 
 	for_each_btree_key_norestart(trans, bp_iter, BTREE_ID_backpointers,
 				     *bp_pos, iter_flags, k, ret) {
@@ -249,7 +268,6 @@ static void backpointer_not_found(struct btree_trans *trans,
 {
 	struct bch_fs *c = trans->c;
 	struct printbuf buf = PRINTBUF;
-	struct bpos bucket = bp_pos_to_bucket(c, bp_pos);
 
 	/*
 	 * If we're using the btree write buffer, the backpointer we were
@@ -259,6 +277,10 @@ static void backpointer_not_found(struct btree_trans *trans,
 	if (likely(!bch2_backpointers_no_use_write_buffer))
 		return;
 
+	struct bpos bucket;
+	if (!bp_pos_to_bucket_nodev(c, bp_pos, &bucket))
+		return;
+
 	prt_printf(&buf, "backpointer doesn't match %s it points to:\n  ",
 		   bp.level ? "btree node" : "extent");
 	prt_printf(&buf, "bucket: ");
@@ -288,15 +310,17 @@ struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
 {
 	if (likely(!bp.level)) {
 		struct bch_fs *c = trans->c;
-		struct bpos bucket = bp_pos_to_bucket(c, bp_pos);
-		struct bkey_s_c k;
+
+		struct bpos bucket;
+		if (!bp_pos_to_bucket_nodev(c, bp_pos, &bucket))
+			return bkey_s_c_err(-EIO);
 
 		bch2_trans_node_iter_init(trans, iter,
 					  bp.btree_id,
 					  bp.pos,
 					  0, 0,
 					  iter_flags);
-		k = bch2_btree_iter_peek_slot(iter);
+		struct bkey_s_c k = bch2_btree_iter_peek_slot(iter);
 		if (bkey_err(k)) {
 			bch2_trans_iter_exit(trans, iter);
 			return k;
@@ -325,18 +349,20 @@ struct btree *bch2_backpointer_get_node(struct btree_trans *trans,
 					struct bch_backpointer bp)
 {
 	struct bch_fs *c = trans->c;
-	struct bpos bucket = bp_pos_to_bucket(c, bp_pos);
-	struct btree *b;
 
 	BUG_ON(!bp.level);
 
+	struct bpos bucket;
+	if (!bp_pos_to_bucket_nodev(c, bp_pos, &bucket))
+		return ERR_PTR(-EIO);
+
 	bch2_trans_node_iter_init(trans, iter,
 				  bp.btree_id,
 				  bp.pos,
 				  0,
 				  bp.level - 1,
 				  0);
-	b = bch2_btree_iter_peek_node(iter);
+	struct btree *b = bch2_btree_iter_peek_node(iter);
 	if (IS_ERR_OR_NULL(b))
 		goto err;
 
@@ -367,16 +393,16 @@ static int bch2_check_btree_backpointer(struct btree_trans *trans, struct btree_
 	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
-	if (fsck_err_on(!bch2_dev_exists2(c, k.k->p.inode), c,
-			backpointer_to_missing_device,
-			"backpointer for missing device:\n%s",
-			(bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
-		ret = bch2_btree_delete_at(trans, bp_iter, 0);
+	struct bpos bucket;
+	if (!bp_pos_to_bucket_nodev_noerror(c, k.k->p, &bucket)) {
+		if (fsck_err(c, backpointer_to_missing_device,
+			     "backpointer for missing device:\n%s",
+			     (bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+			ret = bch2_btree_delete_at(trans, bp_iter, 0);
 		goto out;
 	}
 
-	alloc_k = bch2_bkey_get_iter(trans, &alloc_iter, BTREE_ID_alloc,
-				     bp_pos_to_bucket(c, k.k->p), 0);
+	alloc_k = bch2_bkey_get_iter(trans, &alloc_iter, BTREE_ID_alloc, bucket, 0);
 	ret = bkey_err(alloc_k);
 	if (ret)
 		goto out;
@@ -460,7 +486,7 @@ found:
 
 	bytes = p.crc.compressed_size << 9;
 
-	struct bch_dev *ca = bch_dev_bkey_exists(c, dev);
+	struct bch_dev *ca = bch2_dev_bkey_exists(c, dev);
 	if (!bch2_dev_get_ioref(ca, READ))
 		return false;
 
@@ -511,25 +537,27 @@ static int check_bp_exists(struct btree_trans *trans,
 	struct printbuf buf = PRINTBUF;
 	struct bkey_s_c bp_k;
 	struct bkey_buf tmp;
-	int ret;
+	int ret = 0;
 
 	bch2_bkey_buf_init(&tmp);
 
-	if (!bch2_dev_bucket_exists(c, bucket)) {
+	struct bch_dev *ca = bch2_dev_bucket_tryget(c, bucket);
+	if (!ca) {
 		prt_str(&buf, "extent for nonexistent device:bucket ");
 		bch2_bpos_to_text(&buf, bucket);
 		prt_str(&buf, "\n  ");
 		bch2_bkey_val_to_text(&buf, c, orig_k);
 		bch_err(c, "%s", buf.buf);
-		return -BCH_ERR_fsck_repair_unimplemented;
+		ret = -BCH_ERR_fsck_repair_unimplemented;
+		goto err;
 	}
 
 	if (bpos_lt(bucket, s->bucket_start) ||
 	    bpos_gt(bucket, s->bucket_end))
-		return 0;
+		goto out;
 
 	bp_k = bch2_bkey_get_iter(trans, &bp_iter, BTREE_ID_backpointers,
-				  bucket_pos_to_bp(c, bucket, bp.bucket_offset),
+				  bucket_pos_to_bp(ca, bucket, bp.bucket_offset),
 				  0);
 	ret = bkey_err(bp_k);
 	if (ret)
@@ -562,6 +590,7 @@ fsck_err:
 	bch2_trans_iter_exit(trans, &other_extent_iter);
 	bch2_trans_iter_exit(trans, &bp_iter);
 	bch2_bkey_buf_exit(&tmp, c);
+	bch2_dev_put(ca);
 	printbuf_exit(&buf);
 	return ret;
 check_existing_bp:
@@ -637,13 +666,13 @@ missing:
 
 	struct bkey_i_backpointer n_bp_k;
 	bkey_backpointer_init(&n_bp_k.k_i);
-	n_bp_k.k.p = bucket_pos_to_bp(trans->c, bucket, bp.bucket_offset);
+	n_bp_k.k.p = bucket_pos_to_bp(ca, bucket, bp.bucket_offset);
 	n_bp_k.v = bp;
 	prt_printf(&buf, "\n  want:  ");
 	bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&n_bp_k.k_i));
 
 	if (fsck_err(c, ptr_to_missing_backpointer, "%s", buf.buf))
-		ret = bch2_bucket_backpointer_mod(trans, bucket, bp, orig_k, true);
+		ret = bch2_bucket_backpointer_mod(trans, ca, bucket, bp, orig_k, true);
 
 	goto out;
 }
@@ -667,7 +696,14 @@ static int check_extent_to_backpointers(struct btree_trans *trans,
 		if (p.ptr.cached)
 			continue;
 
-		bch2_extent_ptr_to_bp(c, btree, level, k, p, entry, &bucket_pos, &bp);
+		rcu_read_lock();
+		struct bch_dev *ca = bch2_dev_rcu(c, p.ptr.dev);
+		if (ca)
+			bch2_extent_ptr_to_bp(c, ca, btree, level, k, p, entry, &bucket_pos, &bp);
+		rcu_read_unlock();
+
+		if (!ca)
+			continue;
 
 		ret = check_bp_exists(trans, s, bucket_pos, bp, k);
 		if (ret)
@@ -760,7 +796,7 @@ static int bch2_get_btree_in_memory_pos(struct btree_trans *trans,
 
 		__for_each_btree_node(trans, iter, btree,
 				      btree == start.btree ? start.pos : POS_MIN,
-				      0, depth, BTREE_ITER_PREFETCH, b, ret) {
+				      0, depth, BTREE_ITER_prefetch, b, ret) {
 			mem_may_pin -= btree_buf_bytes(b);
 			if (mem_may_pin <= 0) {
 				c->btree_cache.pinned_nodes_end = *end =
@@ -794,31 +830,13 @@ static int bch2_check_extents_to_backpointers_pass(struct btree_trans *trans,
 
 		while (level >= depth) {
 			struct btree_iter iter;
-			bch2_trans_node_iter_init(trans, &iter, btree_id, POS_MIN, 0,
-						  level,
-						  BTREE_ITER_PREFETCH);
-			while (1) {
-				bch2_trans_begin(trans);
-
-				struct bkey_s_c k = bch2_btree_iter_peek(&iter);
-				if (!k.k)
-					break;
-				ret = bkey_err(k) ?:
-					check_extent_to_backpointers(trans, s, btree_id, level, k) ?:
-					bch2_trans_commit(trans, NULL, NULL,
-							  BCH_TRANS_COMMIT_no_enospc);
-				if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) {
-					ret = 0;
-					continue;
-				}
-				if (ret)
-					break;
-				if (bpos_eq(iter.pos, SPOS_MAX))
-					break;
-				bch2_btree_iter_advance(&iter);
-			}
-			bch2_trans_iter_exit(trans, &iter);
+			bch2_trans_node_iter_init(trans, &iter, btree_id, POS_MIN, 0, level,
+						  BTREE_ITER_prefetch);
 
+			ret = for_each_btree_key_continue(trans, iter, 0, k, ({
+				check_extent_to_backpointers(trans, s, btree_id, level, k) ?:
+				bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
+			}));
 			if (ret)
 				return ret;
 
@@ -936,7 +954,7 @@ static int bch2_check_backpointers_to_extents_pass(struct btree_trans *trans,
 	struct bpos last_flushed_pos = SPOS_MAX;
 
 	return for_each_btree_key_commit(trans, iter, BTREE_ID_backpointers,
-				  POS_MIN, BTREE_ITER_PREFETCH, k,
+				  POS_MIN, BTREE_ITER_prefetch, k,
 				  NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 		check_one_backpointer(trans, start, end,
 				      bkey_s_c_to_backpointer(k),
diff --git a/fs/bcachefs/backpointers.h b/fs/bcachefs/backpointers.h
index 85949b9fd880ce..f0cd5d4e63c296 100644
--- a/fs/bcachefs/backpointers.h
+++ b/fs/bcachefs/backpointers.h
@@ -6,6 +6,7 @@
 #include "btree_iter.h"
 #include "btree_update.h"
 #include "buckets.h"
+#include "error.h"
 #include "super.h"
 
 static inline u64 swab40(u64 x)
@@ -36,47 +37,62 @@ void bch2_backpointer_swab(struct bkey_s);
  * Convert from pos in backpointer btree to pos of corresponding bucket in alloc
  * btree:
  */
-static inline struct bpos bp_pos_to_bucket(const struct bch_fs *c,
+static inline struct bpos bp_pos_to_bucket(const struct bch_dev *ca,
 					   struct bpos bp_pos)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, bp_pos.inode);
 	u64 bucket_sector = bp_pos.offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT;
 
 	return POS(bp_pos.inode, sector_to_bucket(ca, bucket_sector));
 }
 
+static inline bool bp_pos_to_bucket_nodev_noerror(struct bch_fs *c, struct bpos bp_pos, struct bpos *bucket)
+{
+	rcu_read_lock();
+	struct bch_dev *ca = bch2_dev_rcu(c, bp_pos.inode);
+	if (ca)
+		*bucket = bp_pos_to_bucket(ca, bp_pos);
+	rcu_read_unlock();
+	return ca != NULL;
+}
+
+static inline bool bp_pos_to_bucket_nodev(struct bch_fs *c, struct bpos bp_pos, struct bpos *bucket)
+{
+	return !bch2_fs_inconsistent_on(!bp_pos_to_bucket_nodev_noerror(c, bp_pos, bucket),
+					c, "backpointer for missing device %llu", bp_pos.inode);
+}
+
 /*
  * Convert from pos in alloc btree + bucket offset to pos in backpointer btree:
  */
-static inline struct bpos bucket_pos_to_bp(const struct bch_fs *c,
+static inline struct bpos bucket_pos_to_bp(const struct bch_dev *ca,
 					   struct bpos bucket,
 					   u64 bucket_offset)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
 	struct bpos ret = POS(bucket.inode,
 			      (bucket_to_sector(ca, bucket.offset) <<
 			       MAX_EXTENT_COMPRESS_RATIO_SHIFT) + bucket_offset);
 
-	EBUG_ON(!bkey_eq(bucket, bp_pos_to_bucket(c, ret)));
+	EBUG_ON(!bkey_eq(bucket, bp_pos_to_bucket(ca, ret)));
 	return ret;
 }
 
-int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *, struct bpos bucket,
-				struct bch_backpointer, struct bkey_s_c, bool);
+int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *, struct bch_dev *,
+				struct bpos bucket, struct bch_backpointer, struct bkey_s_c, bool);
 
 static inline int bch2_bucket_backpointer_mod(struct btree_trans *trans,
+				struct bch_dev *ca,
 				struct bpos bucket,
 				struct bch_backpointer bp,
 				struct bkey_s_c orig_k,
 				bool insert)
 {
 	if (unlikely(bch2_backpointers_no_use_write_buffer))
-		return bch2_bucket_backpointer_mod_nowritebuffer(trans, bucket, bp, orig_k, insert);
+		return bch2_bucket_backpointer_mod_nowritebuffer(trans, ca, bucket, bp, orig_k, insert);
 
 	struct bkey_i_backpointer bp_k;
 
 	bkey_backpointer_init(&bp_k.k_i);
-	bp_k.k.p = bucket_pos_to_bp(trans->c, bucket, bp.bucket_offset);
+	bp_k.k.p = bucket_pos_to_bp(ca, bucket, bp.bucket_offset);
 	bp_k.v = bp;
 
 	if (!insert) {
@@ -114,7 +130,7 @@ static inline enum bch_data_type bch2_bkey_ptr_data_type(struct bkey_s_c k,
 	}
 }
 
-static inline void bch2_extent_ptr_to_bp(struct bch_fs *c,
+static inline void bch2_extent_ptr_to_bp(struct bch_fs *c, struct bch_dev *ca,
 			   enum btree_id btree_id, unsigned level,
 			   struct bkey_s_c k, struct extent_ptr_decoded p,
 			   const union bch_extent_entry *entry,
@@ -124,7 +140,7 @@ static inline void bch2_extent_ptr_to_bp(struct bch_fs *c,
 	s64 sectors = level ? btree_sectors(c) : k.k->size;
 	u32 bucket_offset;
 
-	*bucket_pos = PTR_BUCKET_POS_OFFSET(c, &p.ptr, &bucket_offset);
+	*bucket_pos = PTR_BUCKET_POS_OFFSET(ca, &p.ptr, &bucket_offset);
 	*bp = (struct bch_backpointer) {
 		.btree_id	= btree_id,
 		.level		= level,
@@ -136,7 +152,7 @@ static inline void bch2_extent_ptr_to_bp(struct bch_fs *c,
 	};
 }
 
-int bch2_get_next_backpointer(struct btree_trans *, struct bpos, int,
+int bch2_get_next_backpointer(struct btree_trans *, struct bch_dev *ca, struct bpos, int,
 			      struct bpos *, struct bch_backpointer *, unsigned);
 struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *, struct btree_iter *,
 					 struct bpos, struct bch_backpointer,
diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h
index 91c3c1fef233d1..a1ee178de090aa 100644
--- a/fs/bcachefs/bcachefs.h
+++ b/fs/bcachefs/bcachefs.h
@@ -359,6 +359,8 @@ do {									\
 #define BCH_DEBUG_PARAMS_ALWAYS()					\
 	BCH_DEBUG_PARAM(key_merging_disabled,				\
 		"Disables merging of extents")				\
+	BCH_DEBUG_PARAM(btree_node_merging_disabled,			\
+		"Disables merging of btree nodes")			\
 	BCH_DEBUG_PARAM(btree_gc_always_rewrite,			\
 		"Causes mark and sweep to compact and rewrite every "	\
 		"btree node it traverses")				\
@@ -468,6 +470,7 @@ enum bch_time_stats {
 #include "quota_types.h"
 #include "rebalance_types.h"
 #include "replicas_types.h"
+#include "sb-members_types.h"
 #include "subvolume_types.h"
 #include "super_types.h"
 #include "thread_with_file_types.h"
@@ -516,8 +519,8 @@ enum gc_phase {
 
 struct gc_pos {
 	enum gc_phase		phase;
+	u16			level;
 	struct bpos		pos;
-	unsigned		level;
 };
 
 struct reflink_gc {
@@ -560,14 +563,11 @@ struct bch_dev {
 
 	struct bch_devs_mask	self;
 
-	/* biosets used in cloned bios for writing multiple replicas */
-	struct bio_set		replica_set;
-
 	/*
 	 * Buckets:
 	 * Per-bucket arrays are protected by c->mark_lock, bucket_lock and
 	 * gc_lock, for device resize - holding any is sufficient for access:
-	 * Or rcu_read_lock(), but only for ptr_stale():
+	 * Or rcu_read_lock(), but only for dev_ptr_stale():
 	 */
 	struct bucket_array __rcu *buckets_gc;
 	struct bucket_gens __rcu *bucket_gens;
@@ -627,7 +627,6 @@ struct bch_dev {
 	x(clean_shutdown)		\
 	x(fsck_running)			\
 	x(initial_gc_unfixed)		\
-	x(need_another_gc)		\
 	x(need_delete_dead_snapshots)	\
 	x(error)			\
 	x(topology_error)		\
@@ -715,6 +714,7 @@ struct btree_trans_buf {
 	x(discard_fast)							\
 	x(invalidate)							\
 	x(delete_dead_snapshots)					\
+	x(gc_gens)							\
 	x(snapshot_delete_pagecache)					\
 	x(sysfs)							\
 	x(btree_write_buffer)
@@ -926,7 +926,6 @@ struct bch_fs {
 	/* JOURNAL SEQ BLACKLIST */
 	struct journal_seq_blacklist_table *
 				journal_seq_blacklist_table;
-	struct work_struct	journal_seq_blacklist_gc_work;
 
 	/* ALLOCATOR */
 	spinlock_t		freelist_lock;
@@ -957,8 +956,7 @@ struct bch_fs {
 	struct work_struct	discard_fast_work;
 
 	/* GARBAGE COLLECTION */
-	struct task_struct	*gc_thread;
-	atomic_t		kick_gc;
+	struct work_struct	gc_gens_work;
 	unsigned long		gc_count;
 
 	enum btree_id		gc_gens_btree;
@@ -988,6 +986,7 @@ struct bch_fs {
 	struct bio_set		bio_read;
 	struct bio_set		bio_read_split;
 	struct bio_set		bio_write;
+	struct bio_set		replica_set;
 	struct mutex		bio_bounce_pages_lock;
 	mempool_t		bio_bounce_pages;
 	struct bucket_nocow_lock_table
@@ -1115,7 +1114,6 @@ struct bch_fs {
 	u64			counters_on_mount[BCH_COUNTER_NR];
 	u64 __percpu		*counters;
 
-	unsigned		btree_gc_periodic:1;
 	unsigned		copy_gc_enabled:1;
 	bool			promote_whole_extents;
 
@@ -1250,11 +1248,6 @@ static inline s64 bch2_current_time(const struct bch_fs *c)
 	return timespec_to_bch2_time(c, now);
 }
 
-static inline bool bch2_dev_exists2(const struct bch_fs *c, unsigned dev)
-{
-	return dev < c->sb.nr_devices && c->devs[dev];
-}
-
 static inline struct stdio_redirect *bch2_fs_stdio_redirect(struct bch_fs *c)
 {
 	struct stdio_redirect *stdio = c->stdio;
diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h
index f7fbfccd2b1e4d..24bceb344941fe 100644
--- a/fs/bcachefs/bcachefs_format.h
+++ b/fs/bcachefs/bcachefs_format.h
@@ -589,6 +589,13 @@ struct bch_member {
 	__le64			errors_reset_time;
 	__le64			seq;
 	__le64			btree_allocated_bitmap;
+	/*
+	 * On recovery from a clean shutdown we don't normally read the journal,
+	 * but we still want to resume writing from where we left off so we
+	 * don't overwrite more than is necessary, for list journal debugging:
+	 */
+	__le32			last_journal_bucket;
+	__le32			last_journal_bucket_offset;
 };
 
 #define BCH_MEMBER_V1_BYTES	56
diff --git a/fs/bcachefs/bkey_methods.h b/fs/bcachefs/bkey_methods.h
index 03efe8ee565a90..7351c0d38534fc 100644
--- a/fs/bcachefs/bkey_methods.h
+++ b/fs/bcachefs/bkey_methods.h
@@ -29,7 +29,8 @@ struct bkey_ops {
 	bool		(*key_normalize)(struct bch_fs *, struct bkey_s);
 	bool		(*key_merge)(struct bch_fs *, struct bkey_s, struct bkey_s_c);
 	int		(*trigger)(struct btree_trans *, enum btree_id, unsigned,
-				   struct bkey_s_c, struct bkey_s, unsigned);
+				   struct bkey_s_c, struct bkey_s,
+				   enum btree_iter_update_trigger_flags);
 	void		(*compat)(enum btree_id id, unsigned version,
 				  unsigned big_endian, int write,
 				  struct bkey_s);
@@ -76,56 +77,10 @@ static inline bool bch2_bkey_maybe_mergable(const struct bkey *l, const struct b
 
 bool bch2_bkey_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
 
-enum btree_update_flags {
-	__BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE = __BTREE_ITER_FLAGS_END,
-	__BTREE_UPDATE_NOJOURNAL,
-	__BTREE_UPDATE_KEY_CACHE_RECLAIM,
-
-	__BTREE_TRIGGER_NORUN,
-	__BTREE_TRIGGER_TRANSACTIONAL,
-	__BTREE_TRIGGER_ATOMIC,
-	__BTREE_TRIGGER_GC,
-	__BTREE_TRIGGER_INSERT,
-	__BTREE_TRIGGER_OVERWRITE,
-	__BTREE_TRIGGER_BUCKET_INVALIDATE,
-};
-
-#define BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE (1U << __BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE)
-#define BTREE_UPDATE_NOJOURNAL		(1U << __BTREE_UPDATE_NOJOURNAL)
-#define BTREE_UPDATE_KEY_CACHE_RECLAIM	(1U << __BTREE_UPDATE_KEY_CACHE_RECLAIM)
-
-/* Don't run triggers at all */
-#define BTREE_TRIGGER_NORUN		(1U << __BTREE_TRIGGER_NORUN)
-
-/*
- * If set, we're running transactional triggers as part of a transaction commit:
- * triggers may generate new updates
- *
- * If cleared, and either BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE are set,
- * we're running atomic triggers during a transaction commit: we have our
- * journal reservation, we're holding btree node write locks, and we know the
- * transaction is going to commit (returning an error here is a fatal error,
- * causing us to go emergency read-only)
- */
-#define BTREE_TRIGGER_TRANSACTIONAL	(1U << __BTREE_TRIGGER_TRANSACTIONAL)
-#define BTREE_TRIGGER_ATOMIC		(1U << __BTREE_TRIGGER_ATOMIC)
-
-/* We're in gc/fsck: running triggers to recalculate e.g. disk usage */
-#define BTREE_TRIGGER_GC		(1U << __BTREE_TRIGGER_GC)
-
-/* @new is entering the btree */
-#define BTREE_TRIGGER_INSERT		(1U << __BTREE_TRIGGER_INSERT)
-
-/* @old is leaving the btree */
-#define BTREE_TRIGGER_OVERWRITE		(1U << __BTREE_TRIGGER_OVERWRITE)
-
-/* signal from bucket invalidate path to alloc trigger */
-#define BTREE_TRIGGER_BUCKET_INVALIDATE	(1U << __BTREE_TRIGGER_BUCKET_INVALIDATE)
-
 static inline int bch2_key_trigger(struct btree_trans *trans,
 		enum btree_id btree, unsigned level,
 		struct bkey_s_c old, struct bkey_s new,
-		unsigned flags)
+		enum btree_iter_update_trigger_flags flags)
 {
 	const struct bkey_ops *ops = bch2_bkey_type_ops(old.k->type ?: new.k->type);
 
@@ -135,8 +90,9 @@ static inline int bch2_key_trigger(struct btree_trans *trans,
 }
 
 static inline int bch2_key_trigger_old(struct btree_trans *trans,
-				       enum btree_id btree_id, unsigned level,
-				       struct bkey_s_c old, unsigned flags)
+			enum btree_id btree_id, unsigned level,
+			struct bkey_s_c old,
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bkey_i deleted;
 
@@ -144,12 +100,13 @@ static inline int bch2_key_trigger_old(struct btree_trans *trans,
 	deleted.k.p = old.k->p;
 
 	return bch2_key_trigger(trans, btree_id, level, old, bkey_i_to_s(&deleted),
-				BTREE_TRIGGER_OVERWRITE|flags);
+				BTREE_TRIGGER_overwrite|flags);
 }
 
 static inline int bch2_key_trigger_new(struct btree_trans *trans,
-				       enum btree_id btree_id, unsigned level,
-				       struct bkey_s new, unsigned flags)
+			enum btree_id btree_id, unsigned level,
+			struct bkey_s new,
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bkey_i deleted;
 
@@ -157,7 +114,7 @@ static inline int bch2_key_trigger_new(struct btree_trans *trans,
 	deleted.k.p = new.k->p;
 
 	return bch2_key_trigger(trans, btree_id, level, bkey_i_to_s_c(&deleted), new,
-				BTREE_TRIGGER_INSERT|flags);
+				BTREE_TRIGGER_insert|flags);
 }
 
 void bch2_bkey_renumber(enum btree_node_type, struct bkey_packed *, int);
diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index 02c70e813face0..debdd7dc04778e 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -190,6 +190,35 @@ int bch2_btree_node_hash_insert(struct btree_cache *bc, struct btree *b,
 	return ret;
 }
 
+void bch2_btree_node_update_key_early(struct btree_trans *trans,
+				      enum btree_id btree, unsigned level,
+				      struct bkey_s_c old, struct bkey_i *new)
+{
+	struct bch_fs *c = trans->c;
+	struct btree *b;
+	struct bkey_buf tmp;
+	int ret;
+
+	bch2_bkey_buf_init(&tmp);
+	bch2_bkey_buf_reassemble(&tmp, c, old);
+
+	b = bch2_btree_node_get_noiter(trans, tmp.k, btree, level, true);
+	if (!IS_ERR_OR_NULL(b)) {
+		mutex_lock(&c->btree_cache.lock);
+
+		bch2_btree_node_hash_remove(&c->btree_cache, b);
+
+		bkey_copy(&b->key, new);
+		ret = __bch2_btree_node_hash_insert(&c->btree_cache, b);
+		BUG_ON(ret);
+
+		mutex_unlock(&c->btree_cache.lock);
+		six_unlock_read(&b->c.lock);
+	}
+
+	bch2_bkey_buf_exit(&tmp, c);
+}
+
 __flatten
 static inline struct btree *btree_cache_find(struct btree_cache *bc,
 				     const struct bkey_i *k)
diff --git a/fs/bcachefs/btree_cache.h b/fs/bcachefs/btree_cache.h
index 6d33885fdbde0d..6fe91d1c0fd4c9 100644
--- a/fs/bcachefs/btree_cache.h
+++ b/fs/bcachefs/btree_cache.h
@@ -17,6 +17,9 @@ int __bch2_btree_node_hash_insert(struct btree_cache *, struct btree *);
 int bch2_btree_node_hash_insert(struct btree_cache *, struct btree *,
 				unsigned, enum btree_id);
 
+void bch2_btree_node_update_key_early(struct btree_trans *, enum btree_id, unsigned,
+				      struct bkey_s_c, struct bkey_i *);
+
 void bch2_btree_cache_cannibalize_unlock(struct btree_trans *);
 int bch2_btree_cache_cannibalize_lock(struct btree_trans *, struct closure *);
 
diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index 791470b0c65455..9186d880d31885 100644
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -52,12 +52,6 @@ static struct bkey_s unsafe_bkey_s_c_to_s(struct bkey_s_c k)
 	}}};
 }
 
-static bool should_restart_for_topology_repair(struct bch_fs *c)
-{
-	return c->opts.fix_errors != FSCK_FIX_no &&
-		!(c->recovery_passes_complete & BIT_ULL(BCH_RECOVERY_PASS_check_topology));
-}
-
 static inline void __gc_pos_set(struct bch_fs *c, struct gc_pos new_pos)
 {
 	preempt_disable();
@@ -69,7 +63,7 @@ static inline void __gc_pos_set(struct bch_fs *c, struct gc_pos new_pos)
 
 static inline void gc_pos_set(struct bch_fs *c, struct gc_pos new_pos)
 {
-	BUG_ON(gc_pos_cmp(new_pos, c->gc_pos) <= 0);
+	BUG_ON(gc_pos_cmp(new_pos, c->gc_pos) < 0);
 	__gc_pos_set(c, new_pos);
 }
 
@@ -97,35 +91,6 @@ static void btree_ptr_to_v2(struct btree *b, struct bkey_i_btree_ptr_v2 *dst)
 	}
 }
 
-static void bch2_btree_node_update_key_early(struct btree_trans *trans,
-					     enum btree_id btree, unsigned level,
-					     struct bkey_s_c old, struct bkey_i *new)
-{
-	struct bch_fs *c = trans->c;
-	struct btree *b;
-	struct bkey_buf tmp;
-	int ret;
-
-	bch2_bkey_buf_init(&tmp);
-	bch2_bkey_buf_reassemble(&tmp, c, old);
-
-	b = bch2_btree_node_get_noiter(trans, tmp.k, btree, level, true);
-	if (!IS_ERR_OR_NULL(b)) {
-		mutex_lock(&c->btree_cache.lock);
-
-		bch2_btree_node_hash_remove(&c->btree_cache, b);
-
-		bkey_copy(&b->key, new);
-		ret = __bch2_btree_node_hash_insert(&c->btree_cache, b);
-		BUG_ON(ret);
-
-		mutex_unlock(&c->btree_cache.lock);
-		six_unlock_read(&b->c.lock);
-	}
-
-	bch2_bkey_buf_exit(&tmp, c);
-}
-
 static int set_node_min(struct bch_fs *c, struct btree *b, struct bpos new_min)
 {
 	struct bkey_i_btree_ptr_v2 *new;
@@ -546,9 +511,9 @@ reconstruct_root:
 			if (!bch2_btree_has_scanned_nodes(c, i)) {
 				mustfix_fsck_err(c, btree_root_unreadable_and_scan_found_nothing,
 						 "no nodes found for btree %s, continue?", bch2_btree_id_str(i));
-				bch2_btree_root_alloc_fake(c, i, 0);
+				bch2_btree_root_alloc_fake_trans(trans, i, 0);
 			} else {
-				bch2_btree_root_alloc_fake(c, i, 1);
+				bch2_btree_root_alloc_fake_trans(trans, i, 1);
 				bch2_shoot_down_journal_keys(c, i, 1, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX);
 				ret = bch2_get_scanned_nodes(c, i, 0, POS_MIN, SPOS_MAX);
 				if (ret)
@@ -576,7 +541,7 @@ reconstruct_root:
 				goto reconstruct_root;
 
 			bch_err(c, "empty btree root %s", bch2_btree_id_str(i));
-			bch2_btree_root_alloc_fake(c, i, 0);
+			bch2_btree_root_alloc_fake_trans(trans, i, 0);
 			r->alive = false;
 			ret = 0;
 		}
@@ -586,495 +551,123 @@ fsck_err:
 	return ret;
 }
 
-static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id,
-			       unsigned level, bool is_root,
-			       struct bkey_s_c *k)
+/* marking of btree keys/nodes: */
+
+static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id,
+			    unsigned level, struct btree **prev,
+			    struct btree_iter *iter, struct bkey_s_c k,
+			    bool initial)
 {
 	struct bch_fs *c = trans->c;
-	struct bkey_ptrs_c ptrs_c = bch2_bkey_ptrs_c(*k);
-	const union bch_extent_entry *entry_c;
-	struct extent_ptr_decoded p = { 0 };
-	bool do_update = false;
-	struct printbuf buf = PRINTBUF;
-	int ret = 0;
-
-	/*
-	 * XXX
-	 * use check_bucket_ref here
-	 */
-	bkey_for_each_ptr_decode(k->k, ptrs_c, p, entry_c) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
-		struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
-		enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, p, entry_c);
-
-		if (fsck_err_on(!g->gen_valid,
-				c, ptr_to_missing_alloc_key,
-				"bucket %u:%zu data type %s ptr gen %u missing in alloc btree\n"
-				"while marking %s",
-				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr),
-				bch2_data_type_str(ptr_data_type(k->k, &p.ptr)),
-				p.ptr.gen,
-				(printbuf_reset(&buf),
-				 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) {
-			if (!p.ptr.cached) {
-				g->gen_valid		= true;
-				g->gen			= p.ptr.gen;
-			} else {
-				do_update = true;
-			}
-		}
 
-		if (fsck_err_on(gen_cmp(p.ptr.gen, g->gen) > 0,
-				c, ptr_gen_newer_than_bucket_gen,
-				"bucket %u:%zu data type %s ptr gen in the future: %u > %u\n"
-				"while marking %s",
-				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr),
-				bch2_data_type_str(ptr_data_type(k->k, &p.ptr)),
-				p.ptr.gen, g->gen,
-				(printbuf_reset(&buf),
-				 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) {
-			if (!p.ptr.cached) {
-				g->gen_valid		= true;
-				g->gen			= p.ptr.gen;
-				g->data_type		= 0;
-				g->dirty_sectors	= 0;
-				g->cached_sectors	= 0;
-				set_bit(BCH_FS_need_another_gc, &c->flags);
-			} else {
-				do_update = true;
-			}
-		}
-
-		if (fsck_err_on(gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX,
-				c, ptr_gen_newer_than_bucket_gen,
-				"bucket %u:%zu gen %u data type %s: ptr gen %u too stale\n"
-				"while marking %s",
-				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), g->gen,
-				bch2_data_type_str(ptr_data_type(k->k, &p.ptr)),
-				p.ptr.gen,
-				(printbuf_reset(&buf),
-				 bch2_bkey_val_to_text(&buf, c, *k), buf.buf)))
-			do_update = true;
-
-		if (fsck_err_on(!p.ptr.cached && gen_cmp(p.ptr.gen, g->gen) < 0,
-				c, stale_dirty_ptr,
-				"bucket %u:%zu data type %s stale dirty ptr: %u < %u\n"
-				"while marking %s",
-				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr),
-				bch2_data_type_str(ptr_data_type(k->k, &p.ptr)),
-				p.ptr.gen, g->gen,
-				(printbuf_reset(&buf),
-				 bch2_bkey_val_to_text(&buf, c, *k), buf.buf)))
-			do_update = true;
-
-		if (data_type != BCH_DATA_btree && p.ptr.gen != g->gen)
-			continue;
-
-		if (fsck_err_on(bucket_data_type(g->data_type) &&
-				bucket_data_type(g->data_type) !=
-				bucket_data_type(data_type), c,
-				ptr_bucket_data_type_mismatch,
-				"bucket %u:%zu different types of data in same bucket: %s, %s\n"
-				"while marking %s",
-				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr),
-				bch2_data_type_str(g->data_type),
-				bch2_data_type_str(data_type),
-				(printbuf_reset(&buf),
-				 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) {
-			if (data_type == BCH_DATA_btree) {
-				g->data_type	= data_type;
-				set_bit(BCH_FS_need_another_gc, &c->flags);
-			} else {
-				do_update = true;
-			}
-		}
+	if (iter) {
+		struct btree_path *path = btree_iter_path(trans, iter);
+		struct btree *b = path_l(path)->b;
 
-		if (p.has_ec) {
-			struct gc_stripe *m = genradix_ptr(&c->gc_stripes, p.ec.idx);
-
-			if (fsck_err_on(!m || !m->alive, c,
-					ptr_to_missing_stripe,
-					"pointer to nonexistent stripe %llu\n"
-					"while marking %s",
-					(u64) p.ec.idx,
-					(printbuf_reset(&buf),
-					 bch2_bkey_val_to_text(&buf, c, *k), buf.buf)))
-				do_update = true;
-
-			if (fsck_err_on(m && m->alive && !bch2_ptr_matches_stripe_m(m, p), c,
-					ptr_to_incorrect_stripe,
-					"pointer does not match stripe %llu\n"
-					"while marking %s",
-					(u64) p.ec.idx,
-					(printbuf_reset(&buf),
-					 bch2_bkey_val_to_text(&buf, c, *k), buf.buf)))
-				do_update = true;
-		}
-	}
-
-	if (do_update) {
-		if (is_root) {
-			bch_err(c, "cannot update btree roots yet");
-			ret = -EINVAL;
-			goto err;
-		}
-
-		struct bkey_i *new = kmalloc(bkey_bytes(k->k), GFP_KERNEL);
-		if (!new) {
-			ret = -BCH_ERR_ENOMEM_gc_repair_key;
-			bch_err_msg(c, ret, "allocating new key");
-			goto err;
-		}
-
-		bkey_reassemble(new, *k);
-
-		if (level) {
-			/*
-			 * We don't want to drop btree node pointers - if the
-			 * btree node isn't there anymore, the read path will
-			 * sort it out:
-			 */
-			struct bkey_ptrs ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
-			bkey_for_each_ptr(ptrs, ptr) {
-				struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
-				struct bucket *g = PTR_GC_BUCKET(ca, ptr);
-
-				ptr->gen = g->gen;
-			}
-		} else {
-			struct bkey_ptrs ptrs;
-			union bch_extent_entry *entry;
-restart_drop_ptrs:
-			ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
-			bkey_for_each_ptr_decode(bkey_i_to_s(new).k, ptrs, p, entry) {
-				struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
-				struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
-				enum bch_data_type data_type = bch2_bkey_ptr_data_type(bkey_i_to_s_c(new), p, entry);
-
-				if ((p.ptr.cached &&
-				     (!g->gen_valid || gen_cmp(p.ptr.gen, g->gen) > 0)) ||
-				    (!p.ptr.cached &&
-				     gen_cmp(p.ptr.gen, g->gen) < 0) ||
-				    gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX ||
-				    (g->data_type &&
-				     g->data_type != data_type)) {
-					bch2_bkey_drop_ptr(bkey_i_to_s(new), &entry->ptr);
-					goto restart_drop_ptrs;
-				}
-			}
-again:
-			ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
-			bkey_extent_entry_for_each(ptrs, entry) {
-				if (extent_entry_type(entry) == BCH_EXTENT_ENTRY_stripe_ptr) {
-					struct gc_stripe *m = genradix_ptr(&c->gc_stripes,
-									entry->stripe_ptr.idx);
-					union bch_extent_entry *next_ptr;
-
-					bkey_extent_entry_for_each_from(ptrs, next_ptr, entry)
-						if (extent_entry_type(next_ptr) == BCH_EXTENT_ENTRY_ptr)
-							goto found;
-					next_ptr = NULL;
-found:
-					if (!next_ptr) {
-						bch_err(c, "aieee, found stripe ptr with no data ptr");
-						continue;
-					}
-
-					if (!m || !m->alive ||
-					    !__bch2_ptr_matches_stripe(&m->ptrs[entry->stripe_ptr.block],
-								       &next_ptr->ptr,
-								       m->sectors)) {
-						bch2_bkey_extent_entry_drop(new, entry);
-						goto again;
-					}
-				}
-			}
-		}
-
-		if (level)
-			bch2_btree_node_update_key_early(trans, btree_id, level - 1, *k, new);
-
-		if (0) {
-			printbuf_reset(&buf);
-			bch2_bkey_val_to_text(&buf, c, *k);
-			bch_info(c, "updated %s", buf.buf);
-
-			printbuf_reset(&buf);
-			bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(new));
-			bch_info(c, "new key %s", buf.buf);
-		}
-
-		ret = bch2_journal_key_insert_take(c, btree_id, level, new);
-		if (ret) {
-			kfree(new);
-			goto err;
+		if (*prev != b) {
+			int ret = bch2_btree_node_check_topology(trans, b);
+			if (ret)
+				return ret;
 		}
-
-		*k = bkey_i_to_s_c(new);
+		*prev = b;
 	}
-err:
-fsck_err:
-	printbuf_exit(&buf);
-	return ret;
-}
-
-/* marking of btree keys/nodes: */
 
-static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id,
-			    unsigned level, bool is_root,
-			    struct bkey_s_c *k,
-			    bool initial)
-{
-	struct bch_fs *c = trans->c;
 	struct bkey deleted = KEY(0, 0, 0);
 	struct bkey_s_c old = (struct bkey_s_c) { &deleted, NULL };
 	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
-	deleted.p = k->k->p;
+	deleted.p = k.k->p;
 
 	if (initial) {
 		BUG_ON(bch2_journal_seq_verify &&
-		       k->k->version.lo > atomic64_read(&c->journal.seq));
+		       k.k->version.lo > atomic64_read(&c->journal.seq));
 
-		if (fsck_err_on(k->k->version.lo > atomic64_read(&c->key_version), c,
+		if (fsck_err_on(k.k->version.lo > atomic64_read(&c->key_version), c,
 				bkey_version_in_future,
 				"key version number higher than recorded: %llu > %llu",
-				k->k->version.lo,
+				k.k->version.lo,
 				atomic64_read(&c->key_version)))
-			atomic64_set(&c->key_version, k->k->version.lo);
+			atomic64_set(&c->key_version, k.k->version.lo);
 	}
 
-	ret = bch2_check_fix_ptrs(trans, btree_id, level, is_root, k);
-	if (ret)
-		goto err;
-
-	if (mustfix_fsck_err_on(level && !bch2_dev_btree_bitmap_marked(c, *k),
+	if (mustfix_fsck_err_on(level && !bch2_dev_btree_bitmap_marked(c, k),
 				c, btree_bitmap_not_marked,
 				"btree ptr not marked in member info btree allocated bitmap\n  %s",
-				(bch2_bkey_val_to_text(&buf, c, *k),
+				(bch2_bkey_val_to_text(&buf, c, k),
 				 buf.buf))) {
 		mutex_lock(&c->sb_lock);
-		bch2_dev_btree_bitmap_mark(c, *k);
+		bch2_dev_btree_bitmap_mark(c, k);
 		bch2_write_super(c);
 		mutex_unlock(&c->sb_lock);
 	}
 
-	ret = commit_do(trans, NULL, NULL, 0,
-			bch2_key_trigger(trans, btree_id, level, old,
-					 unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC));
-fsck_err:
-err:
-	printbuf_exit(&buf);
-	bch_err_fn(c, ret);
-	return ret;
-}
-
-static int btree_gc_mark_node(struct btree_trans *trans, struct btree *b, bool initial)
-{
-	struct btree_node_iter iter;
-	struct bkey unpacked;
-	struct bkey_s_c k;
-	int ret = 0;
+	/*
+	 * We require a commit before key_trigger() because
+	 * key_trigger(BTREE_TRIGGER_GC) is not idempotant; we'll calculate the
+	 * wrong result if we run it multiple times.
+	 */
+	unsigned flags = !iter ? BTREE_TRIGGER_is_root : 0;
 
-	ret = bch2_btree_node_check_topology(trans, b);
+	ret = bch2_key_trigger(trans, btree_id, level, old, unsafe_bkey_s_c_to_s(k),
+			       BTREE_TRIGGER_check_repair|flags);
 	if (ret)
-		return ret;
-
-	if (!btree_node_type_needs_gc(btree_node_type(b)))
-		return 0;
-
-	bch2_btree_node_iter_init_from_start(&iter, b);
-
-	while ((k = bch2_btree_node_iter_peek_unpack(&iter, b, &unpacked)).k) {
-		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, false,
-				       &k, initial);
-		if (ret)
-			return ret;
+		goto out;
 
-		bch2_btree_node_iter_advance(&iter, b);
+	if (trans->nr_updates) {
+		ret = bch2_trans_commit(trans, NULL, NULL, 0) ?:
+			-BCH_ERR_transaction_restart_nested;
+		goto out;
 	}
 
-	return 0;
+	ret = bch2_key_trigger(trans, btree_id, level, old, unsafe_bkey_s_c_to_s(k),
+			       BTREE_TRIGGER_gc|flags);
+out:
+fsck_err:
+	printbuf_exit(&buf);
+	bch_err_fn(c, ret);
+	return ret;
 }
 
-static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree_id,
-			 bool initial, bool metadata_only)
+static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree, bool initial)
 {
 	struct bch_fs *c = trans->c;
-	struct btree_iter iter;
-	struct btree *b;
-	unsigned depth = metadata_only ? 1 : 0;
+	int level = 0, target_depth = btree_node_type_needs_gc(__btree_node_type(0, btree)) ? 0 : 1;
 	int ret = 0;
 
-	gc_pos_set(c, gc_pos_btree(btree_id, POS_MIN, 0));
-
-	__for_each_btree_node(trans, iter, btree_id, POS_MIN,
-			      0, depth, BTREE_ITER_PREFETCH, b, ret) {
-		bch2_verify_btree_nr_keys(b);
-
-		gc_pos_set(c, gc_pos_btree_node(b));
-
-		ret = btree_gc_mark_node(trans, b, initial);
-		if (ret)
-			break;
-	}
-	bch2_trans_iter_exit(trans, &iter);
-
-	if (ret)
-		return ret;
+	/* We need to make sure every leaf node is readable before going RW */
+	if (initial)
+		target_depth = 0;
 
+	/* root */
 	mutex_lock(&c->btree_root_lock);
-	b = bch2_btree_id_root(c, btree_id)->b;
+	struct btree *b = bch2_btree_id_root(c, btree)->b;
 	if (!btree_node_fake(b)) {
-		struct bkey_s_c k = bkey_i_to_s_c(&b->key);
-
-		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1,
-				       true, &k, initial);
+		gc_pos_set(c, gc_pos_btree(btree, b->c.level + 1, SPOS_MAX));
+		ret = lockrestart_do(trans,
+			bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1,
+					 NULL, NULL, bkey_i_to_s_c(&b->key), initial));
+		level = b->c.level;
 	}
-	gc_pos_set(c, gc_pos_btree_root(b->c.btree_id));
 	mutex_unlock(&c->btree_root_lock);
 
-	return ret;
-}
-
-static int bch2_gc_btree_init_recurse(struct btree_trans *trans, struct btree *b,
-				      unsigned target_depth)
-{
-	struct bch_fs *c = trans->c;
-	struct btree_and_journal_iter iter;
-	struct bkey_s_c k;
-	struct bkey_buf cur;
-	struct printbuf buf = PRINTBUF;
-	int ret = 0;
-
-	ret = bch2_btree_node_check_topology(trans, b);
 	if (ret)
 		return ret;
 
-	bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b);
-	bch2_bkey_buf_init(&cur);
-
-	while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) {
-		BUG_ON(bpos_lt(k.k->p, b->data->min_key));
-		BUG_ON(bpos_gt(k.k->p, b->data->max_key));
+	for (; level >= target_depth; --level) {
+		struct btree *prev = NULL;
+		struct btree_iter iter;
+		bch2_trans_node_iter_init(trans, &iter, btree, POS_MIN, 0, level,
+					  BTREE_ITER_prefetch);
 
-		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level,
-				       false, &k, true);
+		ret = for_each_btree_key_continue(trans, iter, 0, k, ({
+			gc_pos_set(c, gc_pos_btree(btree, level, k.k->p));
+			bch2_gc_mark_key(trans, btree, level, &prev, &iter, k, initial);
+		}));
 		if (ret)
-			goto fsck_err;
-
-		bch2_btree_and_journal_iter_advance(&iter);
-	}
-
-	if (b->c.level > target_depth) {
-		bch2_btree_and_journal_iter_exit(&iter);
-		bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b);
-		iter.prefetch = true;
-
-		while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) {
-			struct btree *child;
-
-			bch2_bkey_buf_reassemble(&cur, c, k);
-			bch2_btree_and_journal_iter_advance(&iter);
-
-			child = bch2_btree_node_get_noiter(trans, cur.k,
-						b->c.btree_id, b->c.level - 1,
-						false);
-			ret = PTR_ERR_OR_ZERO(child);
-
-			if (bch2_err_matches(ret, EIO)) {
-				bch2_topology_error(c);
-
-				if (__fsck_err(c,
-					  FSCK_CAN_FIX|
-					  FSCK_CAN_IGNORE|
-					  FSCK_NO_RATELIMIT,
-					  btree_node_read_error,
-					  "Unreadable btree node at btree %s level %u:\n"
-					  "  %s",
-					  bch2_btree_id_str(b->c.btree_id),
-					  b->c.level - 1,
-					  (printbuf_reset(&buf),
-					   bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur.k)), buf.buf)) &&
-				    should_restart_for_topology_repair(c)) {
-					bch_info(c, "Halting mark and sweep to start topology repair pass");
-					ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology);
-					goto fsck_err;
-				} else {
-					/* Continue marking when opted to not
-					 * fix the error: */
-					ret = 0;
-					set_bit(BCH_FS_initial_gc_unfixed, &c->flags);
-					continue;
-				}
-			} else if (ret) {
-				bch_err_msg(c, ret, "getting btree node");
-				break;
-			}
-
-			ret = bch2_gc_btree_init_recurse(trans, child,
-							 target_depth);
-			six_unlock_read(&child->c.lock);
-
-			if (ret)
-				break;
-		}
-	}
-fsck_err:
-	bch2_bkey_buf_exit(&cur, c);
-	bch2_btree_and_journal_iter_exit(&iter);
-	printbuf_exit(&buf);
-	return ret;
-}
-
-static int bch2_gc_btree_init(struct btree_trans *trans,
-			      enum btree_id btree_id,
-			      bool metadata_only)
-{
-	struct bch_fs *c = trans->c;
-	struct btree *b;
-	unsigned target_depth = metadata_only ? 1 : 0;
-	struct printbuf buf = PRINTBUF;
-	int ret = 0;
-
-	b = bch2_btree_id_root(c, btree_id)->b;
-
-	six_lock_read(&b->c.lock, NULL, NULL);
-	printbuf_reset(&buf);
-	bch2_bpos_to_text(&buf, b->data->min_key);
-	if (mustfix_fsck_err_on(!bpos_eq(b->data->min_key, POS_MIN), c,
-				btree_root_bad_min_key,
-			"btree root with incorrect min_key: %s", buf.buf)) {
-		bch_err(c, "repair unimplemented");
-		ret = -BCH_ERR_fsck_repair_unimplemented;
-		goto fsck_err;
-	}
-
-	printbuf_reset(&buf);
-	bch2_bpos_to_text(&buf, b->data->max_key);
-	if (mustfix_fsck_err_on(!bpos_eq(b->data->max_key, SPOS_MAX), c,
-				btree_root_bad_max_key,
-			"btree root with incorrect max_key: %s", buf.buf)) {
-		bch_err(c, "repair unimplemented");
-		ret = -BCH_ERR_fsck_repair_unimplemented;
-		goto fsck_err;
-	}
-
-	if (b->c.level >= target_depth)
-		ret = bch2_gc_btree_init_recurse(trans, b, target_depth);
-
-	if (!ret) {
-		struct bkey_s_c k = bkey_i_to_s_c(&b->key);
-
-		ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1, true,
-				       &k, true);
+			break;
 	}
-fsck_err:
-	six_unlock_read(&b->c.lock);
 
-	bch_err_fn(c, ret);
-	printbuf_exit(&buf);
 	return ret;
 }
 
@@ -1084,7 +677,7 @@ static inline int btree_id_gc_phase_cmp(enum btree_id l, enum btree_id r)
 		(int) btree_id_to_gc_phase(r);
 }
 
-static int bch2_gc_btrees(struct bch_fs *c, bool initial, bool metadata_only)
+static int bch2_gc_btrees(struct bch_fs *c)
 {
 	struct btree_trans *trans = bch2_trans_get(c);
 	enum btree_id ids[BTREE_ID_NR];
@@ -1095,98 +688,36 @@ static int bch2_gc_btrees(struct bch_fs *c, bool initial, bool metadata_only)
 		ids[i] = i;
 	bubble_sort(ids, BTREE_ID_NR, btree_id_gc_phase_cmp);
 
-	for (i = 0; i < BTREE_ID_NR && !ret; i++)
-		ret = initial
-			? bch2_gc_btree_init(trans, ids[i], metadata_only)
-			: bch2_gc_btree(trans, ids[i], initial, metadata_only);
+	for (i = 0; i < btree_id_nr_alive(c) && !ret; i++) {
+		unsigned btree = i < BTREE_ID_NR ? ids[i] : i;
 
-	for (i = BTREE_ID_NR; i < btree_id_nr_alive(c) && !ret; i++) {
-		if (!bch2_btree_id_root(c, i)->alive)
+		if (IS_ERR_OR_NULL(bch2_btree_id_root(c, btree)->b))
 			continue;
 
-		ret = initial
-			? bch2_gc_btree_init(trans, i, metadata_only)
-			: bch2_gc_btree(trans, i, initial, metadata_only);
-	}
+		ret = bch2_gc_btree(trans, btree, true);
 
+		if (mustfix_fsck_err_on(bch2_err_matches(ret, EIO),
+					c, btree_node_read_error,
+			       "btree node read error for %s",
+			       bch2_btree_id_str(btree)))
+			ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology);
+	}
+fsck_err:
 	bch2_trans_put(trans);
 	bch_err_fn(c, ret);
 	return ret;
 }
 
-static void mark_metadata_sectors(struct bch_fs *c, struct bch_dev *ca,
-				  u64 start, u64 end,
-				  enum bch_data_type type,
-				  unsigned flags)
-{
-	u64 b = sector_to_bucket(ca, start);
-
-	do {
-		unsigned sectors =
-			min_t(u64, bucket_to_sector(ca, b + 1), end) - start;
-
-		bch2_mark_metadata_bucket(c, ca, b, type, sectors,
-					  gc_phase(GC_PHASE_SB), flags);
-		b++;
-		start += sectors;
-	} while (start < end);
-}
-
-static void bch2_mark_dev_superblock(struct bch_fs *c, struct bch_dev *ca,
-				     unsigned flags)
-{
-	struct bch_sb_layout *layout = &ca->disk_sb.sb->layout;
-	unsigned i;
-	u64 b;
-
-	for (i = 0; i < layout->nr_superblocks; i++) {
-		u64 offset = le64_to_cpu(layout->sb_offset[i]);
-
-		if (offset == BCH_SB_SECTOR)
-			mark_metadata_sectors(c, ca, 0, BCH_SB_SECTOR,
-					      BCH_DATA_sb, flags);
-
-		mark_metadata_sectors(c, ca, offset,
-				      offset + (1 << layout->sb_max_size_bits),
-				      BCH_DATA_sb, flags);
-	}
-
-	for (i = 0; i < ca->journal.nr; i++) {
-		b = ca->journal.buckets[i];
-		bch2_mark_metadata_bucket(c, ca, b, BCH_DATA_journal,
-					  ca->mi.bucket_size,
-					  gc_phase(GC_PHASE_SB), flags);
-	}
-}
-
-static void bch2_mark_superblocks(struct bch_fs *c)
+static int bch2_mark_superblocks(struct bch_fs *c)
 {
 	mutex_lock(&c->sb_lock);
 	gc_pos_set(c, gc_phase(GC_PHASE_SB));
 
-	for_each_online_member(c, ca)
-		bch2_mark_dev_superblock(c, ca, BTREE_TRIGGER_GC);
+	int ret = bch2_trans_mark_dev_sbs_flags(c, BTREE_TRIGGER_gc);
 	mutex_unlock(&c->sb_lock);
+	return ret;
 }
 
-#if 0
-/* Also see bch2_pending_btree_node_free_insert_done() */
-static void bch2_mark_pending_btree_node_frees(struct bch_fs *c)
-{
-	struct btree_update *as;
-	struct pending_btree_node_free *d;
-
-	mutex_lock(&c->btree_interior_update_lock);
-	gc_pos_set(c, gc_phase(GC_PHASE_PENDING_DELETE));
-
-	for_each_pending_btree_node_free(c, as, d)
-		if (d->index_update_done)
-			bch2_mark_key(c, bkey_i_to_s_c(&d->key), BTREE_TRIGGER_GC);
-
-	mutex_unlock(&c->btree_interior_update_lock);
-}
-#endif
-
 static void bch2_gc_free(struct bch_fs *c)
 {
 	genradix_free(&c->reflink_gc_table);
@@ -1204,28 +735,23 @@ static void bch2_gc_free(struct bch_fs *c)
 	c->usage_gc = NULL;
 }
 
-static int bch2_gc_done(struct bch_fs *c,
-			bool initial, bool metadata_only)
+static int bch2_gc_done(struct bch_fs *c)
 {
 	struct bch_dev *ca = NULL;
 	struct printbuf buf = PRINTBUF;
-	bool verify = !metadata_only &&
-		!c->opts.reconstruct_alloc &&
-		(!initial || (c->sb.compat & (1ULL << BCH_COMPAT_alloc_info)));
 	unsigned i;
 	int ret = 0;
 
 	percpu_down_write(&c->mark_lock);
 
-#define copy_field(_err, _f, _msg, ...)					\
-	if (dst->_f != src->_f &&					\
-	    (!verify ||							\
-	     fsck_err(c, _err, _msg ": got %llu, should be %llu"	\
-		      , ##__VA_ARGS__, dst->_f, src->_f)))		\
+#define copy_field(_err, _f, _msg, ...)						\
+	if (fsck_err_on(dst->_f != src->_f, c, _err,				\
+			_msg ": got %llu, should be %llu" , ##__VA_ARGS__,	\
+			dst->_f, src->_f))					\
 		dst->_f = src->_f
-#define copy_dev_field(_err, _f, _msg, ...)				\
+#define copy_dev_field(_err, _f, _msg, ...)					\
 	copy_field(_err, _f, "dev %u has wrong " _msg, ca->dev_idx, ##__VA_ARGS__)
-#define copy_fs_field(_err, _f, _msg, ...)				\
+#define copy_fs_field(_err, _f, _msg, ...)					\
 	copy_field(_err, _f, "fs has wrong " _msg, ##__VA_ARGS__)
 
 	for (i = 0; i < ARRAY_SIZE(c->usage); i++)
@@ -1258,31 +784,24 @@ static int bch2_gc_done(struct bch_fs *c,
 		copy_fs_field(fs_usage_btree_wrong,
 			      b.btree,		"btree");
 
-		if (!metadata_only) {
-			copy_fs_field(fs_usage_data_wrong,
-				      b.data,	"data");
-			copy_fs_field(fs_usage_cached_wrong,
-				      b.cached,	"cached");
-			copy_fs_field(fs_usage_reserved_wrong,
-				      b.reserved,	"reserved");
-			copy_fs_field(fs_usage_nr_inodes_wrong,
-				      b.nr_inodes,"nr_inodes");
-
-			for (i = 0; i < BCH_REPLICAS_MAX; i++)
-				copy_fs_field(fs_usage_persistent_reserved_wrong,
-					      persistent_reserved[i],
-					      "persistent_reserved[%i]", i);
-		}
+		copy_fs_field(fs_usage_data_wrong,
+			      b.data,	"data");
+		copy_fs_field(fs_usage_cached_wrong,
+			      b.cached,	"cached");
+		copy_fs_field(fs_usage_reserved_wrong,
+			      b.reserved,	"reserved");
+		copy_fs_field(fs_usage_nr_inodes_wrong,
+			      b.nr_inodes,"nr_inodes");
+
+		for (i = 0; i < BCH_REPLICAS_MAX; i++)
+			copy_fs_field(fs_usage_persistent_reserved_wrong,
+				      persistent_reserved[i],
+				      "persistent_reserved[%i]", i);
 
 		for (i = 0; i < c->replicas.nr; i++) {
 			struct bch_replicas_entry_v1 *e =
 				cpu_replicas_entry(&c->replicas, i);
 
-			if (metadata_only &&
-			    (e->data_type == BCH_DATA_user ||
-			     e->data_type == BCH_DATA_cached))
-				continue;
-
 			printbuf_reset(&buf);
 			bch2_replicas_entry_to_text(&buf, e);
 
@@ -1333,19 +852,6 @@ static int bch2_gc_start(struct bch_fs *c)
 	return 0;
 }
 
-static int bch2_gc_reset(struct bch_fs *c)
-{
-	for_each_member_device(c, ca) {
-		free_percpu(ca->usage_gc);
-		ca->usage_gc = NULL;
-	}
-
-	free_percpu(c->usage_gc);
-	c->usage_gc = NULL;
-
-	return bch2_gc_start(c);
-}
-
 /* returns true if not equal */
 static inline bool bch2_alloc_v4_cmp(struct bch_alloc_v4 l,
 				     struct bch_alloc_v4 r)
@@ -1361,56 +867,41 @@ static inline bool bch2_alloc_v4_cmp(struct bch_alloc_v4 l,
 
 static int bch2_alloc_write_key(struct btree_trans *trans,
 				struct btree_iter *iter,
-				struct bkey_s_c k,
-				bool metadata_only)
+				struct bch_dev *ca,
+				struct bkey_s_c k)
 {
 	struct bch_fs *c = trans->c;
-	struct bch_dev *ca = bch_dev_bkey_exists(c, iter->pos.inode);
-	struct bucket old_gc, gc, *b;
 	struct bkey_i_alloc_v4 *a;
-	struct bch_alloc_v4 old_convert, new;
+	struct bch_alloc_v4 old_gc, gc, old_convert, new;
 	const struct bch_alloc_v4 *old;
 	int ret;
 
 	old = bch2_alloc_to_v4(k, &old_convert);
-	new = *old;
+	gc = new = *old;
 
 	percpu_down_read(&c->mark_lock);
-	b = gc_bucket(ca, iter->pos.offset);
-	old_gc = *b;
+	__bucket_m_to_alloc(&gc, *gc_bucket(ca, iter->pos.offset));
+
+	old_gc = gc;
 
 	if ((old->data_type == BCH_DATA_sb ||
 	     old->data_type == BCH_DATA_journal) &&
 	    !bch2_dev_is_online(ca)) {
-		b->data_type = old->data_type;
-		b->dirty_sectors = old->dirty_sectors;
+		gc.data_type = old->data_type;
+		gc.dirty_sectors = old->dirty_sectors;
 	}
 
 	/*
-	 * b->data_type doesn't yet include need_discard & need_gc_gen states -
+	 * gc.data_type doesn't yet include need_discard & need_gc_gen states -
 	 * fix that here:
 	 */
-	b->data_type = __alloc_data_type(b->dirty_sectors,
-					 b->cached_sectors,
-					 b->stripe,
-					 *old,
-					 b->data_type);
-	gc = *b;
+	alloc_data_type_set(&gc, gc.data_type);
 
 	if (gc.data_type != old_gc.data_type ||
 	    gc.dirty_sectors != old_gc.dirty_sectors)
-		bch2_dev_usage_update_m(c, ca, &old_gc, &gc);
+		bch2_dev_usage_update(c, ca, &old_gc, &gc, 0, true);
 	percpu_up_read(&c->mark_lock);
 
-	if (metadata_only &&
-	    gc.data_type != BCH_DATA_sb &&
-	    gc.data_type != BCH_DATA_journal &&
-	    gc.data_type != BCH_DATA_btree)
-		return 0;
-
-	if (gen_after(old->gen, gc.gen))
-		return 0;
-
 	if (fsck_err_on(new.data_type != gc.data_type, c,
 			alloc_key_data_type_wrong,
 			"bucket %llu:%llu gen %u has wrong data_type"
@@ -1460,12 +951,12 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 	if (a->v.data_type == BCH_DATA_cached && !a->v.io_time[READ])
 		a->v.io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now));
 
-	ret = bch2_trans_update(trans, iter, &a->k_i, BTREE_TRIGGER_NORUN);
+	ret = bch2_trans_update(trans, iter, &a->k_i, BTREE_TRIGGER_norun);
 fsck_err:
 	return ret;
 }
 
-static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_alloc_done(struct bch_fs *c)
 {
 	int ret = 0;
 
@@ -1474,9 +965,9 @@ static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only)
 			for_each_btree_key_upto_commit(trans, iter, BTREE_ID_alloc,
 					POS(ca->dev_idx, ca->mi.first_bucket),
 					POS(ca->dev_idx, ca->mi.nbuckets - 1),
-					BTREE_ITER_SLOTS|BTREE_ITER_PREFETCH, k,
+					BTREE_ITER_slots|BTREE_ITER_prefetch, k,
 					NULL, NULL, BCH_TRANS_COMMIT_lazy_rw,
-				bch2_alloc_write_key(trans, &iter, k, metadata_only)));
+				bch2_alloc_write_key(trans, &iter, ca, k)));
 		if (ret) {
 			percpu_ref_put(&ca->ref);
 			break;
@@ -1487,7 +978,7 @@ static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only)
 	return ret;
 }
 
-static int bch2_gc_alloc_start(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_alloc_start(struct bch_fs *c)
 {
 	for_each_member_device(c, ca) {
 		struct bucket_array *buckets = kvmalloc(sizeof(struct bucket_array) +
@@ -1504,54 +995,29 @@ static int bch2_gc_alloc_start(struct bch_fs *c, bool metadata_only)
 		rcu_assign_pointer(ca->buckets_gc, buckets);
 	}
 
+	struct bch_dev *ca = NULL;
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
-					 BTREE_ITER_PREFETCH, k, ({
-			struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode);
-			struct bucket *g = gc_bucket(ca, k.k->p.offset);
+					 BTREE_ITER_prefetch, k, ({
+			ca = bch2_dev_iterate(c, ca, k.k->p.inode);
+			if (!ca) {
+				bch2_btree_iter_set_pos(&iter, POS(k.k->p.inode + 1, 0));
+				continue;
+			}
 
 			struct bch_alloc_v4 a_convert;
 			const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert);
 
+			struct bucket *g = gc_bucket(ca, k.k->p.offset);
 			g->gen_valid	= 1;
 			g->gen		= a->gen;
-
-			if (metadata_only &&
-			    (a->data_type == BCH_DATA_user ||
-			     a->data_type == BCH_DATA_cached ||
-			     a->data_type == BCH_DATA_parity)) {
-				g->data_type		= a->data_type;
-				g->dirty_sectors	= a->dirty_sectors;
-				g->cached_sectors	= a->cached_sectors;
-				g->stripe		= a->stripe;
-				g->stripe_redundancy	= a->stripe_redundancy;
-			}
-
 			0;
 		})));
+	bch2_dev_put(ca);
 	bch_err_fn(c, ret);
 	return ret;
 }
 
-static void bch2_gc_alloc_reset(struct bch_fs *c, bool metadata_only)
-{
-	for_each_member_device(c, ca) {
-		struct bucket_array *buckets = gc_bucket_array(ca);
-		struct bucket *g;
-
-		for_each_bucket(g, buckets) {
-			if (metadata_only &&
-			    (g->data_type == BCH_DATA_user ||
-			     g->data_type == BCH_DATA_cached ||
-			     g->data_type == BCH_DATA_parity))
-				continue;
-			g->data_type = 0;
-			g->dirty_sectors = 0;
-			g->cached_sectors = 0;
-		}
-	}
-}
-
 static int bch2_gc_write_reflink_key(struct btree_trans *trans,
 				     struct btree_iter *iter,
 				     struct bkey_s_c k,
@@ -1601,35 +1067,27 @@ fsck_err:
 	return ret;
 }
 
-static int bch2_gc_reflink_done(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_reflink_done(struct bch_fs *c)
 {
 	size_t idx = 0;
 
-	if (metadata_only)
-		return 0;
-
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
 				BTREE_ID_reflink, POS_MIN,
-				BTREE_ITER_PREFETCH, k,
+				BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			bch2_gc_write_reflink_key(trans, &iter, k, &idx)));
 	c->reflink_gc_nr = 0;
 	return ret;
 }
 
-static int bch2_gc_reflink_start(struct bch_fs *c,
-				 bool metadata_only)
+static int bch2_gc_reflink_start(struct bch_fs *c)
 {
-
-	if (metadata_only)
-		return 0;
-
 	c->reflink_gc_nr = 0;
 
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_reflink, POS_MIN,
-				   BTREE_ITER_PREFETCH, k, ({
+				   BTREE_ITER_prefetch, k, ({
 			const __le64 *refcount = bkey_refcount_c(k);
 
 			if (!refcount)
@@ -1652,15 +1110,6 @@ static int bch2_gc_reflink_start(struct bch_fs *c,
 	return ret;
 }
 
-static void bch2_gc_reflink_reset(struct bch_fs *c, bool metadata_only)
-{
-	struct genradix_iter iter;
-	struct reflink_gc *r;
-
-	genradix_for_each(&c->reflink_gc_table, iter, r)
-		r->refcount = 0;
-}
-
 static int bch2_gc_write_stripes_key(struct btree_trans *trans,
 				     struct btree_iter *iter,
 				     struct bkey_s_c k)
@@ -1714,30 +1163,20 @@ fsck_err:
 	return ret;
 }
 
-static int bch2_gc_stripes_done(struct bch_fs *c, bool metadata_only)
+static int bch2_gc_stripes_done(struct bch_fs *c)
 {
-	if (metadata_only)
-		return 0;
-
 	return bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
 				BTREE_ID_stripes, POS_MIN,
-				BTREE_ITER_PREFETCH, k,
+				BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			bch2_gc_write_stripes_key(trans, &iter, k)));
 }
 
-static void bch2_gc_stripes_reset(struct bch_fs *c, bool metadata_only)
-{
-	genradix_free(&c->gc_stripes);
-}
-
 /**
- * bch2_gc - walk _all_ references to buckets, and recompute them:
+ * bch2_check_allocations - walk all references to buckets, and recompute them:
  *
  * @c:			filesystem object
- * @initial:		are we in recovery?
- * @metadata_only:	are we just checking metadata references, or everything?
  *
  * Returns: 0 on success, or standard errcode on failure
  *
@@ -1756,9 +1195,8 @@ static void bch2_gc_stripes_reset(struct bch_fs *c, bool metadata_only)
  *    move around - if references move backwards in the ordering GC
  *    uses, GC could skip past them
  */
-int bch2_gc(struct bch_fs *c, bool initial, bool metadata_only)
+int bch2_check_allocations(struct bch_fs *c)
 {
-	unsigned iter = 0;
 	int ret;
 
 	lockdep_assert_held(&c->state_lock);
@@ -1768,62 +1206,30 @@ int bch2_gc(struct bch_fs *c, bool initial, bool metadata_only)
 	bch2_btree_interior_updates_flush(c);
 
 	ret   = bch2_gc_start(c) ?:
-		bch2_gc_alloc_start(c, metadata_only) ?:
-		bch2_gc_reflink_start(c, metadata_only);
+		bch2_gc_alloc_start(c) ?:
+		bch2_gc_reflink_start(c);
 	if (ret)
 		goto out;
-again:
-	gc_pos_set(c, gc_phase(GC_PHASE_START));
 
-	bch2_mark_superblocks(c);
+	gc_pos_set(c, gc_phase(GC_PHASE_START));
 
-	ret = bch2_gc_btrees(c, initial, metadata_only);
+	ret = bch2_mark_superblocks(c);
+	BUG_ON(ret);
 
+	ret = bch2_gc_btrees(c);
 	if (ret)
 		goto out;
 
-#if 0
-	bch2_mark_pending_btree_node_frees(c);
-#endif
 	c->gc_count++;
 
-	if (test_bit(BCH_FS_need_another_gc, &c->flags) ||
-	    (!iter && bch2_test_restart_gc)) {
-		if (iter++ > 2) {
-			bch_info(c, "Unable to fix bucket gens, looping");
-			ret = -EINVAL;
-			goto out;
-		}
-
-		/*
-		 * XXX: make sure gens we fixed got saved
-		 */
-		bch_info(c, "Second GC pass needed, restarting:");
-		clear_bit(BCH_FS_need_another_gc, &c->flags);
-		__gc_pos_set(c, gc_phase(GC_PHASE_NOT_RUNNING));
-
-		bch2_gc_stripes_reset(c, metadata_only);
-		bch2_gc_alloc_reset(c, metadata_only);
-		bch2_gc_reflink_reset(c, metadata_only);
-		ret = bch2_gc_reset(c);
-		if (ret)
-			goto out;
-
-		/* flush fsck errors, reset counters */
-		bch2_flush_fsck_errs(c);
-		goto again;
-	}
+	bch2_journal_block(&c->journal);
 out:
-	if (!ret) {
-		bch2_journal_block(&c->journal);
+	ret   = bch2_gc_alloc_done(c) ?:
+		bch2_gc_done(c) ?:
+		bch2_gc_stripes_done(c) ?:
+		bch2_gc_reflink_done(c);
 
-		ret   = bch2_gc_alloc_done(c, metadata_only) ?:
-			bch2_gc_done(c, initial, metadata_only) ?:
-			bch2_gc_stripes_done(c, metadata_only) ?:
-			bch2_gc_reflink_done(c, metadata_only);
-
-		bch2_journal_unblock(&c->journal);
-	}
+	bch2_journal_unblock(&c->journal);
 
 	percpu_down_write(&c->mark_lock);
 	/* Indicates that gc is no longer in progress: */
@@ -1852,23 +1258,33 @@ static int gc_btree_gens_key(struct btree_trans *trans,
 	struct bkey_i *u;
 	int ret;
 
+	if (unlikely(test_bit(BCH_FS_going_ro, &c->flags)))
+		return -EROFS;
+
 	percpu_down_read(&c->mark_lock);
+	rcu_read_lock();
 	bkey_for_each_ptr(ptrs, ptr) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+		struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
+		if (!ca)
+			continue;
 
-		if (ptr_stale(ca, ptr) > 16) {
+		if (dev_ptr_stale(ca, ptr) > 16) {
+			rcu_read_unlock();
 			percpu_up_read(&c->mark_lock);
 			goto update;
 		}
 	}
 
 	bkey_for_each_ptr(ptrs, ptr) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
-		u8 *gen = &ca->oldest_gen[PTR_BUCKET_NR(ca, ptr)];
+		struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
+		if (!ca)
+			continue;
 
+		u8 *gen = &ca->oldest_gen[PTR_BUCKET_NR(ca, ptr)];
 		if (gen_after(*gen, ptr->gen))
 			*gen = ptr->gen;
 	}
+	rcu_read_unlock();
 	percpu_up_read(&c->mark_lock);
 	return 0;
 update:
@@ -1881,10 +1297,9 @@ update:
 	return 0;
 }
 
-static int bch2_alloc_write_oldest_gen(struct btree_trans *trans, struct btree_iter *iter,
-				       struct bkey_s_c k)
+static int bch2_alloc_write_oldest_gen(struct btree_trans *trans, struct bch_dev *ca,
+				       struct btree_iter *iter, struct bkey_s_c k)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(trans->c, iter->pos.inode);
 	struct bch_alloc_v4 a_convert;
 	const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert);
 	struct bkey_i_alloc_v4 *a_mut;
@@ -1899,7 +1314,7 @@ static int bch2_alloc_write_oldest_gen(struct btree_trans *trans, struct btree_i
 		return ret;
 
 	a_mut->v.oldest_gen = ca->oldest_gen[iter->pos.offset];
-	a_mut->v.data_type = alloc_data_type(a_mut->v, a_mut->v.data_type);
+	alloc_data_type_set(&a_mut->v, a_mut->v.data_type);
 
 	return bch2_trans_update(trans, iter, &a_mut->k_i, 0);
 }
@@ -1945,7 +1360,7 @@ int bch2_gc_gens(struct bch_fs *c)
 			ret = bch2_trans_run(c,
 				for_each_btree_key_commit(trans, iter, i,
 						POS_MIN,
-						BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
+						BTREE_ITER_prefetch|BTREE_ITER_all_snapshots,
 						k,
 						NULL, NULL,
 						BCH_TRANS_COMMIT_no_enospc,
@@ -1954,14 +1369,23 @@ int bch2_gc_gens(struct bch_fs *c)
 				goto err;
 		}
 
+	struct bch_dev *ca = NULL;
 	ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_alloc,
 				POS_MIN,
-				BTREE_ITER_PREFETCH,
+				BTREE_ITER_prefetch,
 				k,
 				NULL, NULL,
-				BCH_TRANS_COMMIT_no_enospc,
-			bch2_alloc_write_oldest_gen(trans, &iter, k)));
+				BCH_TRANS_COMMIT_no_enospc, ({
+			ca = bch2_dev_iterate(c, ca, k.k->p.inode);
+			if (!ca) {
+				bch2_btree_iter_set_pos(&iter, POS(k.k->p.inode + 1, 0));
+				continue;
+			}
+			bch2_alloc_write_oldest_gen(trans, ca, &iter, k);
+		})));
+	bch2_dev_put(ca);
+
 	if (ret)
 		goto err;
 
@@ -1985,87 +1409,23 @@ err:
 	return ret;
 }
 
-static int bch2_gc_thread(void *arg)
+static void bch2_gc_gens_work(struct work_struct *work)
 {
-	struct bch_fs *c = arg;
-	struct io_clock *clock = &c->io_clock[WRITE];
-	unsigned long last = atomic64_read(&clock->now);
-	unsigned last_kick = atomic_read(&c->kick_gc);
-
-	set_freezable();
-
-	while (1) {
-		while (1) {
-			set_current_state(TASK_INTERRUPTIBLE);
-
-			if (kthread_should_stop()) {
-				__set_current_state(TASK_RUNNING);
-				return 0;
-			}
-
-			if (atomic_read(&c->kick_gc) != last_kick)
-				break;
-
-			if (c->btree_gc_periodic) {
-				unsigned long next = last + c->capacity / 16;
-
-				if (atomic64_read(&clock->now) >= next)
-					break;
-
-				bch2_io_clock_schedule_timeout(clock, next);
-			} else {
-				schedule();
-			}
-
-			try_to_freeze();
-		}
-		__set_current_state(TASK_RUNNING);
-
-		last = atomic64_read(&clock->now);
-		last_kick = atomic_read(&c->kick_gc);
-
-		/*
-		 * Full gc is currently incompatible with btree key cache:
-		 */
-#if 0
-		ret = bch2_gc(c, false, false);
-#else
-		bch2_gc_gens(c);
-#endif
-		debug_check_no_locks_held();
-	}
-
-	return 0;
+	struct bch_fs *c = container_of(work, struct bch_fs, gc_gens_work);
+	bch2_gc_gens(c);
+	bch2_write_ref_put(c, BCH_WRITE_REF_gc_gens);
 }
 
-void bch2_gc_thread_stop(struct bch_fs *c)
+void bch2_gc_gens_async(struct bch_fs *c)
 {
-	struct task_struct *p;
-
-	p = c->gc_thread;
-	c->gc_thread = NULL;
-
-	if (p) {
-		kthread_stop(p);
-		put_task_struct(p);
-	}
+	if (bch2_write_ref_tryget(c, BCH_WRITE_REF_gc_gens) &&
+	    !queue_work(c->write_ref_wq, &c->gc_gens_work))
+		bch2_write_ref_put(c, BCH_WRITE_REF_gc_gens);
 }
 
-int bch2_gc_thread_start(struct bch_fs *c)
+void bch2_fs_gc_init(struct bch_fs *c)
 {
-	struct task_struct *p;
+	seqcount_init(&c->gc_pos_lock);
 
-	if (c->gc_thread)
-		return 0;
-
-	p = kthread_create(bch2_gc_thread, c, "bch-gc/%s", c->name);
-	if (IS_ERR(p)) {
-		bch_err_fn(c, PTR_ERR(p));
-		return PTR_ERR(p);
-	}
-
-	get_task_struct(p);
-	c->gc_thread = p;
-	wake_up_process(p);
-	return 0;
+	INIT_WORK(&c->gc_gens_work, bch2_gc_gens_work);
 }
diff --git a/fs/bcachefs/btree_gc.h b/fs/bcachefs/btree_gc.h
index 607575f83a0023..1b6489d8e0f4fa 100644
--- a/fs/bcachefs/btree_gc.h
+++ b/fs/bcachefs/btree_gc.h
@@ -6,10 +6,7 @@
 #include "btree_types.h"
 
 int bch2_check_topology(struct bch_fs *);
-int bch2_gc(struct bch_fs *, bool, bool);
-int bch2_gc_gens(struct bch_fs *);
-void bch2_gc_thread_stop(struct bch_fs *);
-int bch2_gc_thread_start(struct bch_fs *);
+int bch2_check_allocations(struct bch_fs *);
 
 /*
  * For concurrent mark and sweep (with other index updates), we define a total
@@ -37,16 +34,16 @@ static inline struct gc_pos gc_phase(enum gc_phase phase)
 {
 	return (struct gc_pos) {
 		.phase	= phase,
-		.pos	= POS_MIN,
 		.level	= 0,
+		.pos	= POS_MIN,
 	};
 }
 
 static inline int gc_pos_cmp(struct gc_pos l, struct gc_pos r)
 {
-	return  cmp_int(l.phase, r.phase) ?:
-		bpos_cmp(l.pos, r.pos) ?:
-		cmp_int(l.level, r.level);
+	return   cmp_int(l.phase, r.phase) ?:
+		-cmp_int(l.level, r.level) ?:
+		 bpos_cmp(l.pos, r.pos);
 }
 
 static inline enum gc_phase btree_id_to_gc_phase(enum btree_id id)
@@ -60,13 +57,13 @@ static inline enum gc_phase btree_id_to_gc_phase(enum btree_id id)
 	}
 }
 
-static inline struct gc_pos gc_pos_btree(enum btree_id id,
-					 struct bpos pos, unsigned level)
+static inline struct gc_pos gc_pos_btree(enum btree_id btree, unsigned level,
+					 struct bpos pos)
 {
 	return (struct gc_pos) {
-		.phase	= btree_id_to_gc_phase(id),
-		.pos	= pos,
+		.phase	= btree_id_to_gc_phase(btree),
 		.level	= level,
+		.pos	= pos,
 	};
 }
 
@@ -76,19 +73,7 @@ static inline struct gc_pos gc_pos_btree(enum btree_id id,
  */
 static inline struct gc_pos gc_pos_btree_node(struct btree *b)
 {
-	return gc_pos_btree(b->c.btree_id, b->key.k.p, b->c.level);
-}
-
-/*
- * GC position of the pointer to a btree root: we don't use
- * gc_pos_pointer_to_btree_node() here to avoid a potential race with
- * btree_split() increasing the tree depth - the new root will have level > the
- * old root and thus have a greater gc position than the old root, but that
- * would be incorrect since once gc has marked the root it's not coming back.
- */
-static inline struct gc_pos gc_pos_btree_root(enum btree_id id)
-{
-	return gc_pos_btree(id, SPOS_MAX, BTREE_MAX_DEPTH);
+	return gc_pos_btree(b->c.btree_id, b->c.level, b->key.k.p);
 }
 
 static inline bool gc_visited(struct bch_fs *c, struct gc_pos pos)
@@ -104,11 +89,8 @@ static inline bool gc_visited(struct bch_fs *c, struct gc_pos pos)
 	return ret;
 }
 
-static inline void bch2_do_gc_gens(struct bch_fs *c)
-{
-	atomic_inc(&c->kick_gc);
-	if (c->gc_thread)
-		wake_up_process(c->gc_thread);
-}
+int bch2_gc_gens(struct bch_fs *);
+void bch2_gc_gens_async(struct bch_fs *);
+void bch2_fs_gc_init(struct bch_fs *);
 
 #endif /* _BCACHEFS_BTREE_GC_H */
diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c
index debb0edc3455af..055e91f0e4cb5b 100644
--- a/fs/bcachefs/btree_io.c
+++ b/fs/bcachefs/btree_io.c
@@ -23,6 +23,18 @@
 
 #include <linux/sched/mm.h>
 
+static void bch2_btree_node_header_to_text(struct printbuf *out, struct btree_node *bn)
+{
+	prt_printf(out, "btree=%s l=%u seq %llux\n",
+		   bch2_btree_id_str(BTREE_NODE_ID(bn)),
+		   (unsigned) BTREE_NODE_LEVEL(bn), bn->keys.seq);
+	prt_str(out, "min: ");
+	bch2_bpos_to_text(out, bn->min_key);
+	prt_newline(out);
+	prt_str(out, "max: ");
+	bch2_bpos_to_text(out, bn->max_key);
+}
+
 void bch2_btree_node_io_unlock(struct btree *b)
 {
 	EBUG_ON(!btree_node_write_in_flight(b));
@@ -524,7 +536,9 @@ static void btree_err_msg(struct printbuf *out, struct bch_fs *c,
 	prt_printf(out, "at btree ");
 	bch2_btree_pos_to_text(out, c, b);
 
-	prt_printf(out, "\n  node offset %u/%u",
+	printbuf_indent_add(out, 2);
+
+	prt_printf(out, "\nnode offset %u/%u",
 		   b->written, btree_ptr_sectors_written(&b->key));
 	if (i)
 		prt_printf(out, " bset u64s %u", le16_to_cpu(i->u64s));
@@ -543,6 +557,7 @@ static int __btree_err(int ret,
 		       const char *fmt, ...)
 {
 	struct printbuf out = PRINTBUF;
+	bool silent = c->curr_recovery_pass == BCH_RECOVERY_PASS_scan_for_btree_nodes;
 	va_list args;
 
 	btree_err_msg(&out, c, ca, b, i, b->written, write);
@@ -564,12 +579,14 @@ static int __btree_err(int ret,
 	if (!have_retry && ret == -BCH_ERR_btree_node_read_err_must_retry)
 		ret = -BCH_ERR_btree_node_read_err_bad_node;
 
-	if (ret != -BCH_ERR_btree_node_read_err_fixable)
+	if (!silent && ret != -BCH_ERR_btree_node_read_err_fixable)
 		bch2_sb_error_count(c, err_type);
 
 	switch (ret) {
 	case -BCH_ERR_btree_node_read_err_fixable:
-		ret = bch2_fsck_err(c, FSCK_CAN_FIX, err_type, "%s", out.buf);
+		ret = !silent
+			? bch2_fsck_err(c, FSCK_CAN_FIX, err_type, "%s", out.buf)
+			: -BCH_ERR_fsck_fix;
 		if (ret != -BCH_ERR_fsck_fix &&
 		    ret != -BCH_ERR_fsck_ignore)
 			goto fsck_err;
@@ -577,14 +594,17 @@ static int __btree_err(int ret,
 		break;
 	case -BCH_ERR_btree_node_read_err_want_retry:
 	case -BCH_ERR_btree_node_read_err_must_retry:
-		bch2_print_string_as_lines(KERN_ERR, out.buf);
+		if (!silent)
+			bch2_print_string_as_lines(KERN_ERR, out.buf);
 		break;
 	case -BCH_ERR_btree_node_read_err_bad_node:
-		bch2_print_string_as_lines(KERN_ERR, out.buf);
+		if (!silent)
+			bch2_print_string_as_lines(KERN_ERR, out.buf);
 		ret = bch2_topology_error(c);
 		break;
 	case -BCH_ERR_btree_node_read_err_incompatible:
-		bch2_print_string_as_lines(KERN_ERR, out.buf);
+		if (!silent)
+			bch2_print_string_as_lines(KERN_ERR, out.buf);
 		ret = -BCH_ERR_fsck_errors_not_fixed;
 		break;
 	default:
@@ -1021,18 +1041,19 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
 			     -BCH_ERR_btree_node_read_err_must_retry,
 			     c, ca, b, NULL,
 			     btree_node_bad_seq,
-			     "got wrong btree node (want %llx got %llx)\n"
-			     "got btree %s level %llu pos %s",
-			     bp->seq, b->data->keys.seq,
-			     bch2_btree_id_str(BTREE_NODE_ID(b->data)),
-			     BTREE_NODE_LEVEL(b->data),
-			     buf.buf);
+			     "got wrong btree node: got\n%s",
+			     (printbuf_reset(&buf),
+			      bch2_btree_node_header_to_text(&buf, b->data),
+			      buf.buf));
 	} else {
 		btree_err_on(!b->data->keys.seq,
 			     -BCH_ERR_btree_node_read_err_must_retry,
 			     c, ca, b, NULL,
 			     btree_node_bad_seq,
-			     "bad btree header: seq 0");
+			     "bad btree header: seq 0\n%s",
+			     (printbuf_reset(&buf),
+			      bch2_btree_node_header_to_text(&buf, b->data),
+			      buf.buf));
 	}
 
 	while (b->written < (ptr_written ?: btree_sectors(c))) {
@@ -1250,7 +1271,7 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
 	btree_node_reset_sib_u64s(b);
 
 	bkey_for_each_ptr(bch2_bkey_ptrs(bkey_i_to_s(&b->key)), ptr) {
-		struct bch_dev *ca2 = bch_dev_bkey_exists(c, ptr->dev);
+		struct bch_dev *ca2 = bch2_dev_bkey_exists(c, ptr->dev);
 
 		if (ca2->mi.state != BCH_MEMBER_STATE_rw)
 			set_btree_node_need_rewrite(b);
@@ -1280,7 +1301,7 @@ static void btree_node_read_work(struct work_struct *work)
 		container_of(work, struct btree_read_bio, work);
 	struct bch_fs *c	= rb->c;
 	struct btree *b		= rb->b;
-	struct bch_dev *ca	= bch_dev_bkey_exists(c, rb->pick.ptr.dev);
+	struct bch_dev *ca	= bch2_dev_bkey_exists(c, rb->pick.ptr.dev);
 	struct bio *bio		= &rb->bio;
 	struct bch_io_failures failed = { .nr = 0 };
 	struct printbuf buf = PRINTBUF;
@@ -1292,7 +1313,7 @@ static void btree_node_read_work(struct work_struct *work)
 	while (1) {
 		retry = true;
 		bch_info(c, "retrying read");
-		ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
+		ca = bch2_dev_bkey_exists(c, rb->pick.ptr.dev);
 		rb->have_ioref		= bch2_dev_get_ioref(ca, READ);
 		bio_reset(bio, NULL, REQ_OP_READ|REQ_SYNC|REQ_META);
 		bio->bi_iter.bi_sector	= rb->pick.ptr.offset;
@@ -1363,7 +1384,7 @@ static void btree_node_read_endio(struct bio *bio)
 	struct bch_fs *c	= rb->c;
 
 	if (rb->have_ioref) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, rb->pick.ptr.dev);
 
 		bch2_latency_acct(ca, rb->start_time, READ);
 	}
@@ -1560,7 +1581,7 @@ static void btree_node_read_all_replicas_endio(struct bio *bio)
 	struct btree_node_read_all *ra = rb->ra;
 
 	if (rb->have_ioref) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, rb->pick.ptr.dev);
 
 		bch2_latency_acct(ca, rb->start_time, READ);
 	}
@@ -1602,7 +1623,7 @@ static int btree_node_read_all_replicas(struct bch_fs *c, struct btree *b, bool
 
 	i = 0;
 	bkey_for_each_ptr_decode(k.k, ptrs, pick, entry) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, pick.ptr.dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, pick.ptr.dev);
 		struct btree_read_bio *rb =
 			container_of(ra->bio[i], struct btree_read_bio, bio);
 		rb->c			= c;
@@ -1679,7 +1700,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b,
 		return;
 	}
 
-	ca = bch_dev_bkey_exists(c, pick.ptr.dev);
+	ca = bch2_dev_bkey_exists(c, pick.ptr.dev);
 
 	bio = bio_alloc_bioset(NULL,
 			       buf_pages(b->data, btree_buf_bytes(b)),
@@ -1896,7 +1917,7 @@ static void btree_node_write_endio(struct bio *bio)
 	struct btree_write_bio *wb	= container_of(orig, struct btree_write_bio, wbio);
 	struct bch_fs *c		= wbio->c;
 	struct btree *b			= wbio->bio.bi_private;
-	struct bch_dev *ca		= bch_dev_bkey_exists(c, wbio->dev);
+	struct bch_dev *ca		= bch2_dev_bkey_exists(c, wbio->dev);
 	unsigned long flags;
 
 	if (wbio->have_ioref)
@@ -2346,20 +2367,13 @@ void bch2_btree_write_stats_to_text(struct printbuf *out, struct bch_fs *c)
 	printbuf_tabstop_push(out, 20);
 	printbuf_tabstop_push(out, 10);
 
-	prt_tab(out);
-	prt_str(out, "nr");
-	prt_tab(out);
-	prt_str(out, "size");
-	prt_newline(out);
+	prt_printf(out, "\tnr\tsize\n");
 
 	for (unsigned i = 0; i < BTREE_WRITE_TYPE_NR; i++) {
 		u64 nr		= atomic64_read(&c->btree_write_stats[i].nr);
 		u64 bytes	= atomic64_read(&c->btree_write_stats[i].bytes);
 
-		prt_printf(out, "%s:", bch2_btree_write_types[i]);
-		prt_tab(out);
-		prt_u64(out, nr);
-		prt_tab(out);
+		prt_printf(out, "%s:\t%llu\t", bch2_btree_write_types[i], nr);
 		prt_human_readable_u64(out, nr ? div64_u64(bytes, nr) : 0);
 		prt_newline(out);
 	}
diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index 2a211a4bebd153..5bf98cb8b15da6 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -61,7 +61,7 @@ static inline int btree_path_cmp(const struct btree_path *l,
 static inline struct bpos bkey_successor(struct btree_iter *iter, struct bpos p)
 {
 	/* Are we iterating over keys in all snapshots? */
-	if (iter->flags & BTREE_ITER_ALL_SNAPSHOTS) {
+	if (iter->flags & BTREE_ITER_all_snapshots) {
 		p = bpos_successor(p);
 	} else {
 		p = bpos_nosnap_successor(p);
@@ -74,7 +74,7 @@ static inline struct bpos bkey_successor(struct btree_iter *iter, struct bpos p)
 static inline struct bpos bkey_predecessor(struct btree_iter *iter, struct bpos p)
 {
 	/* Are we iterating over keys in all snapshots? */
-	if (iter->flags & BTREE_ITER_ALL_SNAPSHOTS) {
+	if (iter->flags & BTREE_ITER_all_snapshots) {
 		p = bpos_predecessor(p);
 	} else {
 		p = bpos_nosnap_predecessor(p);
@@ -88,7 +88,7 @@ static inline struct bpos btree_iter_search_key(struct btree_iter *iter)
 {
 	struct bpos pos = iter->pos;
 
-	if ((iter->flags & BTREE_ITER_IS_EXTENTS) &&
+	if ((iter->flags & BTREE_ITER_is_extents) &&
 	    !bkey_eq(pos, POS_MAX))
 		pos = bkey_successor(iter, pos);
 	return pos;
@@ -253,13 +253,13 @@ static void bch2_btree_iter_verify(struct btree_iter *iter)
 
 	BUG_ON(iter->btree_id >= BTREE_ID_NR);
 
-	BUG_ON(!!(iter->flags & BTREE_ITER_CACHED) != btree_iter_path(trans, iter)->cached);
+	BUG_ON(!!(iter->flags & BTREE_ITER_cached) != btree_iter_path(trans, iter)->cached);
 
-	BUG_ON((iter->flags & BTREE_ITER_IS_EXTENTS) &&
-	       (iter->flags & BTREE_ITER_ALL_SNAPSHOTS));
+	BUG_ON((iter->flags & BTREE_ITER_is_extents) &&
+	       (iter->flags & BTREE_ITER_all_snapshots));
 
-	BUG_ON(!(iter->flags & __BTREE_ITER_ALL_SNAPSHOTS) &&
-	       (iter->flags & BTREE_ITER_ALL_SNAPSHOTS) &&
+	BUG_ON(!(iter->flags & BTREE_ITER_snapshot_field) &&
+	       (iter->flags & BTREE_ITER_all_snapshots) &&
 	       !btree_type_has_snapshot_field(iter->btree_id));
 
 	if (iter->update_path)
@@ -269,10 +269,10 @@ static void bch2_btree_iter_verify(struct btree_iter *iter)
 
 static void bch2_btree_iter_verify_entry_exit(struct btree_iter *iter)
 {
-	BUG_ON((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) &&
+	BUG_ON((iter->flags & BTREE_ITER_filter_snapshots) &&
 	       !iter->pos.snapshot);
 
-	BUG_ON(!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS) &&
+	BUG_ON(!(iter->flags & BTREE_ITER_all_snapshots) &&
 	       iter->pos.snapshot != iter->snapshot);
 
 	BUG_ON(bkey_lt(iter->pos, bkey_start_pos(&iter->k)) ||
@@ -289,7 +289,7 @@ static int bch2_btree_iter_verify_ret(struct btree_iter *iter, struct bkey_s_c k
 	if (!bch2_debug_check_iterators)
 		return 0;
 
-	if (!(iter->flags & BTREE_ITER_FILTER_SNAPSHOTS))
+	if (!(iter->flags & BTREE_ITER_filter_snapshots))
 		return 0;
 
 	if (bkey_err(k) || !k.k)
@@ -300,8 +300,8 @@ static int bch2_btree_iter_verify_ret(struct btree_iter *iter, struct bkey_s_c k
 					  k.k->p.snapshot));
 
 	bch2_trans_iter_init(trans, &copy, iter->btree_id, iter->pos,
-			     BTREE_ITER_NOPRESERVE|
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_nopreserve|
+			     BTREE_ITER_all_snapshots);
 	prev = bch2_btree_iter_prev(&copy);
 	if (!prev.k)
 		goto out;
@@ -897,7 +897,7 @@ static noinline int btree_node_iter_and_journal_peek(struct btree_trans *trans,
 
 	bch2_bkey_buf_reassemble(out, c, k);
 
-	if ((flags & BTREE_ITER_PREFETCH) &&
+	if ((flags & BTREE_ITER_prefetch) &&
 	    c->opts.btree_node_prefetch)
 		ret = btree_path_prefetch_j(trans, path, &jiter);
 
@@ -944,7 +944,7 @@ static __always_inline int btree_path_down(struct btree_trans *trans,
 
 		bch2_bkey_buf_unpack(&tmp, c, l->b, k);
 
-		if ((flags & BTREE_ITER_PREFETCH) &&
+		if ((flags & BTREE_ITER_prefetch) &&
 		    c->opts.btree_node_prefetch) {
 			ret = btree_path_prefetch(trans, path);
 			if (ret)
@@ -999,6 +999,7 @@ retry_all:
 
 	bch2_trans_unlock(trans);
 	cond_resched();
+	trans->locked = true;
 
 	if (unlikely(trans->memory_allocation_failure)) {
 		struct closure cl;
@@ -1162,6 +1163,7 @@ int bch2_btree_path_traverse_one(struct btree_trans *trans,
 		goto out_uptodate;
 
 	path->level = btree_path_up_until_good_node(trans, path, 0);
+	unsigned max_level = path->level;
 
 	EBUG_ON(btree_path_node(path, path->level) &&
 		!btree_node_locked(path, path->level));
@@ -1192,6 +1194,16 @@ int bch2_btree_path_traverse_one(struct btree_trans *trans,
 			goto out;
 		}
 	}
+
+	if (unlikely(max_level > path->level)) {
+		struct btree_path *linked;
+		unsigned iter;
+
+		trans_for_each_path_with_node(trans, path_l(path)->b, linked, iter)
+			for (unsigned j = path->level + 1; j < max_level; j++)
+				linked->l[j] = path->l[j];
+	}
+
 out_uptodate:
 	path->uptodate = BTREE_ITER_UPTODATE;
 out:
@@ -1221,11 +1233,14 @@ static inline void btree_path_copy(struct btree_trans *trans, struct btree_path
 }
 
 static btree_path_idx_t btree_path_clone(struct btree_trans *trans, btree_path_idx_t src,
-					 bool intent)
+					 bool intent, unsigned long ip)
 {
 	btree_path_idx_t new = btree_path_alloc(trans, src);
 	btree_path_copy(trans, trans->paths + new, trans->paths + src);
 	__btree_path_get(trans->paths + new, intent);
+#ifdef TRACK_PATH_ALLOCATED
+	trans->paths[new].ip_allocated = ip;
+#endif
 	return new;
 }
 
@@ -1234,7 +1249,7 @@ btree_path_idx_t __bch2_btree_path_make_mut(struct btree_trans *trans,
 			btree_path_idx_t path, bool intent, unsigned long ip)
 {
 	__btree_path_put(trans->paths + path, intent);
-	path = btree_path_clone(trans, path, intent);
+	path = btree_path_clone(trans, path, intent, ip);
 	trans->paths[path].preserve = false;
 	return path;
 }
@@ -1334,6 +1349,26 @@ static inline void __bch2_path_free(struct btree_trans *trans, btree_path_idx_t
 	__clear_bit(path, trans->paths_allocated);
 }
 
+static bool bch2_btree_path_can_relock(struct btree_trans *trans, struct btree_path *path)
+{
+	unsigned l = path->level;
+
+	do {
+		if (!btree_path_node(path, l))
+			break;
+
+		if (!is_btree_node(path, l))
+			return false;
+
+		if (path->l[l].lock_seq != path->l[l].b->c.lock.seq)
+			return false;
+
+		l++;
+	} while (l < path->locks_want);
+
+	return true;
+}
+
 void bch2_path_put(struct btree_trans *trans, btree_path_idx_t path_idx, bool intent)
 {
 	struct btree_path *path = trans->paths + path_idx, *dup;
@@ -1348,10 +1383,15 @@ void bch2_path_put(struct btree_trans *trans, btree_path_idx_t path_idx, bool in
 	if (!dup && !(!path->preserve && !is_btree_node(path, path->level)))
 		return;
 
-	if (path->should_be_locked &&
-	    !trans->restarted &&
-	    (!dup || !bch2_btree_path_relock_norestart(trans, dup)))
-		return;
+	if (path->should_be_locked && !trans->restarted) {
+		if (!dup)
+			return;
+
+		if (!(trans->locked
+		      ? bch2_btree_path_relock_norestart(trans, dup)
+		      : bch2_btree_path_can_relock(trans, dup)))
+			return;
+	}
 
 	if (dup) {
 		dup->preserve		|= path->preserve;
@@ -1384,22 +1424,26 @@ void __noreturn bch2_trans_in_restart_error(struct btree_trans *trans)
 	      (void *) trans->last_restarted_ip);
 }
 
+void __noreturn bch2_trans_unlocked_error(struct btree_trans *trans)
+{
+	panic("trans should be locked, unlocked by %pS\n",
+	      (void *) trans->last_unlock_ip);
+}
+
 noinline __cold
 void bch2_trans_updates_to_text(struct printbuf *buf, struct btree_trans *trans)
 {
-	prt_printf(buf, "transaction updates for %s journal seq %llu",
+	prt_printf(buf, "transaction updates for %s journal seq %llu\n",
 	       trans->fn, trans->journal_res.seq);
-	prt_newline(buf);
 	printbuf_indent_add(buf, 2);
 
 	trans_for_each_update(trans, i) {
 		struct bkey_s_c old = { &i->old_k, i->old_v };
 
-		prt_printf(buf, "update: btree=%s cached=%u %pS",
+		prt_printf(buf, "update: btree=%s cached=%u %pS\n",
 		       bch2_btree_id_str(i->btree_id),
 		       i->cached,
 		       (void *) i->ip_allocated);
-		prt_newline(buf);
 
 		prt_printf(buf, "  old ");
 		bch2_bkey_val_to_text(buf, trans->c, old);
@@ -1428,23 +1472,63 @@ void bch2_dump_trans_updates(struct btree_trans *trans)
 	printbuf_exit(&buf);
 }
 
-static void bch2_btree_path_to_text(struct printbuf *out, struct btree_trans *trans, btree_path_idx_t path_idx)
+static void bch2_btree_path_to_text_short(struct printbuf *out, struct btree_trans *trans, btree_path_idx_t path_idx)
 {
 	struct btree_path *path = trans->paths + path_idx;
 
-	prt_printf(out, "path: idx %2u ref %u:%u %c %c btree=%s l=%u pos ",
+	prt_printf(out, "path: idx %2u ref %u:%u %c %c %c btree=%s l=%u pos ",
 		   path_idx, path->ref, path->intent_ref,
 		   path->preserve ? 'P' : ' ',
 		   path->should_be_locked ? 'S' : ' ',
+		   path->cached ? 'C' : 'B',
 		   bch2_btree_id_str(path->btree_id),
 		   path->level);
 	bch2_bpos_to_text(out, path->pos);
 
-	prt_printf(out, " locks %u", path->nodes_locked);
 #ifdef TRACK_PATH_ALLOCATED
 	prt_printf(out, " %pS", (void *) path->ip_allocated);
 #endif
+}
+
+static const char *btree_node_locked_str(enum btree_node_locked_type t)
+{
+	switch (t) {
+	case BTREE_NODE_UNLOCKED:
+		return "unlocked";
+	case BTREE_NODE_READ_LOCKED:
+		return "read";
+	case BTREE_NODE_INTENT_LOCKED:
+		return "intent";
+	case BTREE_NODE_WRITE_LOCKED:
+		return "write";
+	default:
+		return NULL;
+	}
+}
+
+void bch2_btree_path_to_text(struct printbuf *out, struct btree_trans *trans, btree_path_idx_t path_idx)
+{
+	bch2_btree_path_to_text_short(out, trans, path_idx);
+
+	struct btree_path *path = trans->paths + path_idx;
+
+	prt_printf(out, " uptodate %u locks_want %u", path->uptodate, path->locks_want);
 	prt_newline(out);
+
+	printbuf_indent_add(out, 2);
+	for (unsigned l = 0; l < BTREE_MAX_DEPTH; l++) {
+		prt_printf(out, "l=%u locks %s seq %u node ", l,
+			   btree_node_locked_str(btree_node_locked_type(path, l)),
+			   path->l[l].lock_seq);
+
+		int ret = PTR_ERR_OR_ZERO(path->l[l].b);
+		if (ret)
+			prt_str(out, bch2_err_str(ret));
+		else
+			prt_printf(out, "%px", path->l[l].b);
+		prt_newline(out);
+	}
+	printbuf_indent_sub(out, 2);
 }
 
 static noinline __cold
@@ -1456,8 +1540,10 @@ void __bch2_trans_paths_to_text(struct printbuf *out, struct btree_trans *trans,
 	if (!nosort)
 		btree_trans_sort_paths(trans);
 
-	trans_for_each_path_idx_inorder(trans, iter)
-		bch2_btree_path_to_text(out, trans, iter.path_idx);
+	trans_for_each_path_idx_inorder(trans, iter) {
+		bch2_btree_path_to_text_short(out, trans, iter.path_idx);
+		prt_newline(out);
+	}
 }
 
 noinline __cold
@@ -1608,11 +1694,12 @@ btree_path_idx_t bch2_path_get(struct btree_trans *trans,
 			     unsigned flags, unsigned long ip)
 {
 	struct btree_path *path;
-	bool cached = flags & BTREE_ITER_CACHED;
-	bool intent = flags & BTREE_ITER_INTENT;
+	bool cached = flags & BTREE_ITER_cached;
+	bool intent = flags & BTREE_ITER_intent;
 	struct trans_for_each_path_inorder_iter iter;
 	btree_path_idx_t path_pos = 0, path_idx;
 
+	bch2_trans_verify_not_unlocked(trans);
 	bch2_trans_verify_not_in_restart(trans);
 	bch2_trans_verify_locks(trans);
 
@@ -1657,7 +1744,7 @@ btree_path_idx_t bch2_path_get(struct btree_trans *trans,
 		trans->paths_sorted		= false;
 	}
 
-	if (!(flags & BTREE_ITER_NOPRESERVE))
+	if (!(flags & BTREE_ITER_nopreserve))
 		path->preserve = true;
 
 	if (path->intent_ref)
@@ -1678,6 +1765,22 @@ btree_path_idx_t bch2_path_get(struct btree_trans *trans,
 	return path_idx;
 }
 
+btree_path_idx_t bch2_path_get_unlocked_mut(struct btree_trans *trans,
+					    enum btree_id btree_id,
+					    unsigned level,
+					    struct bpos pos)
+{
+	btree_path_idx_t path_idx = bch2_path_get(trans, btree_id, pos, level + 1, level,
+			     BTREE_ITER_nopreserve|
+			     BTREE_ITER_intent, _RET_IP_);
+	path_idx = bch2_btree_path_make_mut(trans, path_idx, true, _RET_IP_);
+
+	struct btree_path *path = trans->paths + path_idx;
+	bch2_btree_path_downgrade(trans, path);
+	__bch2_btree_path_unlock(trans, path);
+	return path_idx;
+}
+
 struct bkey_s_c bch2_btree_path_peek_slot(struct btree_path *path, struct bkey *u)
 {
 
@@ -1719,6 +1822,19 @@ hole:
 	return (struct bkey_s_c) { u, NULL };
 }
 
+
+void bch2_set_btree_iter_dontneed(struct btree_iter *iter)
+{
+	struct btree_trans *trans = iter->trans;
+
+	if (!iter->path || trans->restarted)
+		return;
+
+	struct btree_path *path = btree_iter_path(trans, iter);
+	path->preserve		= false;
+	if (path->ref == 1)
+		path->should_be_locked	= false;
+}
 /* Btree iterators: */
 
 int __must_check
@@ -1733,9 +1849,11 @@ bch2_btree_iter_traverse(struct btree_iter *iter)
 	struct btree_trans *trans = iter->trans;
 	int ret;
 
+	bch2_trans_verify_not_unlocked(trans);
+
 	iter->path = bch2_btree_path_set_pos(trans, iter->path,
 					btree_iter_search_key(iter),
-					iter->flags & BTREE_ITER_INTENT,
+					iter->flags & BTREE_ITER_intent,
 					btree_iter_ip_allocated(iter));
 
 	ret = bch2_btree_path_traverse(iter->trans, iter->path, iter->flags);
@@ -1774,7 +1892,7 @@ struct btree *bch2_btree_iter_peek_node(struct btree_iter *iter)
 	iter->k.p = iter->pos = b->key.k.p;
 
 	iter->path = bch2_btree_path_set_pos(trans, iter->path, b->key.k.p,
-					iter->flags & BTREE_ITER_INTENT,
+					iter->flags & BTREE_ITER_intent,
 					btree_iter_ip_allocated(iter));
 	btree_path_set_should_be_locked(btree_iter_path(trans, iter));
 out:
@@ -1835,13 +1953,16 @@ struct btree *bch2_btree_iter_next_node(struct btree_iter *iter)
 	if (bpos_eq(iter->pos, b->key.k.p)) {
 		__btree_path_set_level_up(trans, path, path->level++);
 	} else {
+		if (btree_lock_want(path, path->level + 1) == BTREE_NODE_UNLOCKED)
+			btree_node_unlock(trans, path, path->level + 1);
+
 		/*
 		 * Haven't gotten to the end of the parent node: go back down to
 		 * the next child node
 		 */
 		iter->path = bch2_btree_path_set_pos(trans, iter->path,
 					bpos_successor(iter->pos),
-					iter->flags & BTREE_ITER_INTENT,
+					iter->flags & BTREE_ITER_intent,
 					btree_iter_ip_allocated(iter));
 
 		path = btree_iter_path(trans, iter);
@@ -1859,7 +1980,7 @@ struct btree *bch2_btree_iter_next_node(struct btree_iter *iter)
 	iter->k.p = iter->pos = b->key.k.p;
 
 	iter->path = bch2_btree_path_set_pos(trans, iter->path, b->key.k.p,
-					iter->flags & BTREE_ITER_INTENT,
+					iter->flags & BTREE_ITER_intent,
 					btree_iter_ip_allocated(iter));
 	btree_path_set_should_be_locked(btree_iter_path(trans, iter));
 	EBUG_ON(btree_iter_path(trans, iter)->uptodate);
@@ -1878,11 +1999,11 @@ err:
 inline bool bch2_btree_iter_advance(struct btree_iter *iter)
 {
 	struct bpos pos = iter->k.p;
-	bool ret = !(iter->flags & BTREE_ITER_ALL_SNAPSHOTS
+	bool ret = !(iter->flags & BTREE_ITER_all_snapshots
 		     ? bpos_eq(pos, SPOS_MAX)
 		     : bkey_eq(pos, SPOS_MAX));
 
-	if (ret && !(iter->flags & BTREE_ITER_IS_EXTENTS))
+	if (ret && !(iter->flags & BTREE_ITER_is_extents))
 		pos = bkey_successor(iter, pos);
 	bch2_btree_iter_set_pos(iter, pos);
 	return ret;
@@ -1891,11 +2012,11 @@ inline bool bch2_btree_iter_advance(struct btree_iter *iter)
 inline bool bch2_btree_iter_rewind(struct btree_iter *iter)
 {
 	struct bpos pos = bkey_start_pos(&iter->k);
-	bool ret = !(iter->flags & BTREE_ITER_ALL_SNAPSHOTS
+	bool ret = !(iter->flags & BTREE_ITER_all_snapshots
 		     ? bpos_eq(pos, POS_MIN)
 		     : bkey_eq(pos, POS_MIN));
 
-	if (ret && !(iter->flags & BTREE_ITER_IS_EXTENTS))
+	if (ret && !(iter->flags & BTREE_ITER_is_extents))
 		pos = bkey_predecessor(iter, pos);
 	bch2_btree_iter_set_pos(iter, pos);
 	return ret;
@@ -2006,7 +2127,10 @@ struct bkey_s_c btree_trans_peek_key_cache(struct btree_iter *iter, struct bpos
 	struct bkey_s_c k;
 	int ret;
 
-	if ((iter->flags & BTREE_ITER_KEY_CACHE_FILL) &&
+	bch2_trans_verify_not_in_restart(trans);
+	bch2_trans_verify_not_unlocked(trans);
+
+	if ((iter->flags & BTREE_ITER_key_cache_fill) &&
 	    bpos_eq(iter->pos, pos))
 		return bkey_s_c_null;
 
@@ -2015,17 +2139,17 @@ struct bkey_s_c btree_trans_peek_key_cache(struct btree_iter *iter, struct bpos
 
 	if (!iter->key_cache_path)
 		iter->key_cache_path = bch2_path_get(trans, iter->btree_id, pos,
-						     iter->flags & BTREE_ITER_INTENT, 0,
-						     iter->flags|BTREE_ITER_CACHED|
-						     BTREE_ITER_CACHED_NOFILL,
+						     iter->flags & BTREE_ITER_intent, 0,
+						     iter->flags|BTREE_ITER_cached|
+						     BTREE_ITER_cached_nofill,
 						     _THIS_IP_);
 
 	iter->key_cache_path = bch2_btree_path_set_pos(trans, iter->key_cache_path, pos,
-					iter->flags & BTREE_ITER_INTENT,
+					iter->flags & BTREE_ITER_intent,
 					btree_iter_ip_allocated(iter));
 
 	ret =   bch2_btree_path_traverse(trans, iter->key_cache_path,
-					 iter->flags|BTREE_ITER_CACHED) ?:
+					 iter->flags|BTREE_ITER_cached) ?:
 		bch2_btree_path_relock(trans, btree_iter_path(trans, iter), _THIS_IP_);
 	if (unlikely(ret))
 		return bkey_s_c_err(ret);
@@ -2053,7 +2177,7 @@ static struct bkey_s_c __bch2_btree_iter_peek(struct btree_iter *iter, struct bp
 		struct btree_path_level *l;
 
 		iter->path = bch2_btree_path_set_pos(trans, iter->path, search_key,
-					iter->flags & BTREE_ITER_INTENT,
+					iter->flags & BTREE_ITER_intent,
 					btree_iter_ip_allocated(iter));
 
 		ret = bch2_btree_path_traverse(trans, iter->path, iter->flags);
@@ -2078,7 +2202,7 @@ static struct bkey_s_c __bch2_btree_iter_peek(struct btree_iter *iter, struct bp
 
 		k = btree_path_level_peek_all(trans->c, l, &iter->k);
 
-		if (unlikely(iter->flags & BTREE_ITER_WITH_KEY_CACHE) &&
+		if (unlikely(iter->flags & BTREE_ITER_with_key_cache) &&
 		    k.k &&
 		    (k2 = btree_trans_peek_key_cache(iter, k.k->p)).k) {
 			k = k2;
@@ -2089,10 +2213,10 @@ static struct bkey_s_c __bch2_btree_iter_peek(struct btree_iter *iter, struct bp
 			}
 		}
 
-		if (unlikely(iter->flags & BTREE_ITER_WITH_JOURNAL))
+		if (unlikely(iter->flags & BTREE_ITER_with_journal))
 			k = btree_trans_peek_journal(trans, iter, k);
 
-		if (unlikely((iter->flags & BTREE_ITER_WITH_UPDATES) &&
+		if (unlikely((iter->flags & BTREE_ITER_with_updates) &&
 			     trans->nr_updates))
 			bch2_btree_trans_peek_updates(trans, iter, &k);
 
@@ -2144,11 +2268,12 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 	struct bpos iter_pos;
 	int ret;
 
-	EBUG_ON((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) && bkey_eq(end, POS_MAX));
+	bch2_trans_verify_not_unlocked(trans);
+	EBUG_ON((iter->flags & BTREE_ITER_filter_snapshots) && bkey_eq(end, POS_MAX));
 
 	if (iter->update_path) {
 		bch2_path_put_nokeep(trans, iter->update_path,
-				     iter->flags & BTREE_ITER_INTENT);
+				     iter->flags & BTREE_ITER_intent);
 		iter->update_path = 0;
 	}
 
@@ -2171,7 +2296,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 		 * isn't monotonically increasing before FILTER_SNAPSHOTS, and
 		 * that's what we check against in extents mode:
 		 */
-		if (unlikely(!(iter->flags & BTREE_ITER_IS_EXTENTS)
+		if (unlikely(!(iter->flags & BTREE_ITER_is_extents)
 			     ? bkey_gt(k.k->p, end)
 			     : k.k->p.inode > end.inode))
 			goto end;
@@ -2179,13 +2304,13 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 		if (iter->update_path &&
 		    !bkey_eq(trans->paths[iter->update_path].pos, k.k->p)) {
 			bch2_path_put_nokeep(trans, iter->update_path,
-					     iter->flags & BTREE_ITER_INTENT);
+					     iter->flags & BTREE_ITER_intent);
 			iter->update_path = 0;
 		}
 
-		if ((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) &&
-		    (iter->flags & BTREE_ITER_INTENT) &&
-		    !(iter->flags & BTREE_ITER_IS_EXTENTS) &&
+		if ((iter->flags & BTREE_ITER_filter_snapshots) &&
+		    (iter->flags & BTREE_ITER_intent) &&
+		    !(iter->flags & BTREE_ITER_is_extents) &&
 		    !iter->update_path) {
 			struct bpos pos = k.k->p;
 
@@ -2200,12 +2325,12 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 			 * advance, same as on exit for iter->path, but only up
 			 * to snapshot
 			 */
-			__btree_path_get(trans->paths + iter->path, iter->flags & BTREE_ITER_INTENT);
+			__btree_path_get(trans->paths + iter->path, iter->flags & BTREE_ITER_intent);
 			iter->update_path = iter->path;
 
 			iter->update_path = bch2_btree_path_set_pos(trans,
 						iter->update_path, pos,
-						iter->flags & BTREE_ITER_INTENT,
+						iter->flags & BTREE_ITER_intent,
 						_THIS_IP_);
 			ret = bch2_btree_path_traverse(trans, iter->update_path, iter->flags);
 			if (unlikely(ret)) {
@@ -2218,7 +2343,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 		 * We can never have a key in a leaf node at POS_MAX, so
 		 * we don't have to check these successor() calls:
 		 */
-		if ((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) &&
+		if ((iter->flags & BTREE_ITER_filter_snapshots) &&
 		    !bch2_snapshot_is_ancestor(trans->c,
 					       iter->snapshot,
 					       k.k->p.snapshot)) {
@@ -2227,7 +2352,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 		}
 
 		if (bkey_whiteout(k.k) &&
-		    !(iter->flags & BTREE_ITER_ALL_SNAPSHOTS)) {
+		    !(iter->flags & BTREE_ITER_all_snapshots)) {
 			search_key = bkey_successor(iter, k.k->p);
 			continue;
 		}
@@ -2237,12 +2362,12 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 		 * equal to the key we just returned - except extents can
 		 * straddle iter->pos:
 		 */
-		if (!(iter->flags & BTREE_ITER_IS_EXTENTS))
+		if (!(iter->flags & BTREE_ITER_is_extents))
 			iter_pos = k.k->p;
 		else
 			iter_pos = bkey_max(iter->pos, bkey_start_pos(k.k));
 
-		if (unlikely(!(iter->flags & BTREE_ITER_IS_EXTENTS)
+		if (unlikely(!(iter->flags & BTREE_ITER_is_extents)
 			     ? bkey_gt(iter_pos, end)
 			     : bkey_ge(iter_pos, end)))
 			goto end;
@@ -2253,7 +2378,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 	iter->pos = iter_pos;
 
 	iter->path = bch2_btree_path_set_pos(trans, iter->path, k.k->p,
-				iter->flags & BTREE_ITER_INTENT,
+				iter->flags & BTREE_ITER_intent,
 				btree_iter_ip_allocated(iter));
 
 	btree_path_set_should_be_locked(btree_iter_path(trans, iter));
@@ -2266,7 +2391,7 @@ out_no_locked:
 			btree_path_set_should_be_locked(trans->paths + iter->update_path);
 	}
 
-	if (!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS))
+	if (!(iter->flags & BTREE_ITER_all_snapshots))
 		iter->pos.snapshot = iter->snapshot;
 
 	ret = bch2_btree_iter_verify_ret(iter, k);
@@ -2316,21 +2441,22 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
 	btree_path_idx_t saved_path = 0;
 	int ret;
 
+	bch2_trans_verify_not_unlocked(trans);
 	EBUG_ON(btree_iter_path(trans, iter)->cached ||
 		btree_iter_path(trans, iter)->level);
 
-	if (iter->flags & BTREE_ITER_WITH_JOURNAL)
+	if (iter->flags & BTREE_ITER_with_journal)
 		return bkey_s_c_err(-BCH_ERR_btree_iter_with_journal_not_supported);
 
 	bch2_btree_iter_verify(iter);
 	bch2_btree_iter_verify_entry_exit(iter);
 
-	if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)
+	if (iter->flags & BTREE_ITER_filter_snapshots)
 		search_key.snapshot = U32_MAX;
 
 	while (1) {
 		iter->path = bch2_btree_path_set_pos(trans, iter->path, search_key,
-						iter->flags & BTREE_ITER_INTENT,
+						iter->flags & BTREE_ITER_intent,
 						btree_iter_ip_allocated(iter));
 
 		ret = bch2_btree_path_traverse(trans, iter->path, iter->flags);
@@ -2345,17 +2471,17 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
 
 		k = btree_path_level_peek(trans, path, &path->l[0], &iter->k);
 		if (!k.k ||
-		    ((iter->flags & BTREE_ITER_IS_EXTENTS)
+		    ((iter->flags & BTREE_ITER_is_extents)
 		     ? bpos_ge(bkey_start_pos(k.k), search_key)
 		     : bpos_gt(k.k->p, search_key)))
 			k = btree_path_level_prev(trans, path, &path->l[0], &iter->k);
 
-		if (unlikely((iter->flags & BTREE_ITER_WITH_UPDATES) &&
+		if (unlikely((iter->flags & BTREE_ITER_with_updates) &&
 			     trans->nr_updates))
 			bch2_btree_trans_peek_prev_updates(trans, iter, &k);
 
 		if (likely(k.k)) {
-			if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) {
+			if (iter->flags & BTREE_ITER_filter_snapshots) {
 				if (k.k->p.snapshot == iter->snapshot)
 					goto got_key;
 
@@ -2366,7 +2492,7 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
 				 */
 				if (saved_path && !bkey_eq(k.k->p, saved_k.p)) {
 					bch2_path_put_nokeep(trans, iter->path,
-						      iter->flags & BTREE_ITER_INTENT);
+						      iter->flags & BTREE_ITER_intent);
 					iter->path = saved_path;
 					saved_path = 0;
 					iter->k	= saved_k;
@@ -2379,9 +2505,10 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
 							      k.k->p.snapshot)) {
 					if (saved_path)
 						bch2_path_put_nokeep(trans, saved_path,
-						      iter->flags & BTREE_ITER_INTENT);
+						      iter->flags & BTREE_ITER_intent);
 					saved_path = btree_path_clone(trans, iter->path,
-								iter->flags & BTREE_ITER_INTENT);
+								iter->flags & BTREE_ITER_intent,
+								_THIS_IP_);
 					path = btree_iter_path(trans, iter);
 					saved_k = *k.k;
 					saved_v = k.v;
@@ -2392,9 +2519,9 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
 			}
 got_key:
 			if (bkey_whiteout(k.k) &&
-			    !(iter->flags & BTREE_ITER_ALL_SNAPSHOTS)) {
+			    !(iter->flags & BTREE_ITER_all_snapshots)) {
 				search_key = bkey_predecessor(iter, k.k->p);
-				if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)
+				if (iter->flags & BTREE_ITER_filter_snapshots)
 					search_key.snapshot = U32_MAX;
 				continue;
 			}
@@ -2418,11 +2545,11 @@ got_key:
 	if (bkey_lt(k.k->p, iter->pos))
 		iter->pos = k.k->p;
 
-	if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)
+	if (iter->flags & BTREE_ITER_filter_snapshots)
 		iter->pos.snapshot = iter->snapshot;
 out_no_locked:
 	if (saved_path)
-		bch2_path_put_nokeep(trans, saved_path, iter->flags & BTREE_ITER_INTENT);
+		bch2_path_put_nokeep(trans, saved_path, iter->flags & BTREE_ITER_intent);
 
 	bch2_btree_iter_verify_entry_exit(iter);
 	bch2_btree_iter_verify(iter);
@@ -2452,12 +2579,13 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
 	struct bkey_s_c k;
 	int ret;
 
+	bch2_trans_verify_not_unlocked(trans);
 	bch2_btree_iter_verify(iter);
 	bch2_btree_iter_verify_entry_exit(iter);
-	EBUG_ON(btree_iter_path(trans, iter)->level && (iter->flags & BTREE_ITER_WITH_KEY_CACHE));
+	EBUG_ON(btree_iter_path(trans, iter)->level && (iter->flags & BTREE_ITER_with_key_cache));
 
 	/* extents can't span inode numbers: */
-	if ((iter->flags & BTREE_ITER_IS_EXTENTS) &&
+	if ((iter->flags & BTREE_ITER_is_extents) &&
 	    unlikely(iter->pos.offset == KEY_OFFSET_MAX)) {
 		if (iter->pos.inode == KEY_INODE_MAX)
 			return bkey_s_c_null;
@@ -2467,7 +2595,7 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
 
 	search_key = btree_iter_search_key(iter);
 	iter->path = bch2_btree_path_set_pos(trans, iter->path, search_key,
-					iter->flags & BTREE_ITER_INTENT,
+					iter->flags & BTREE_ITER_intent,
 					btree_iter_ip_allocated(iter));
 
 	ret = bch2_btree_path_traverse(trans, iter->path, iter->flags);
@@ -2476,22 +2604,22 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
 		goto out_no_locked;
 	}
 
-	if ((iter->flags & BTREE_ITER_CACHED) ||
-	    !(iter->flags & (BTREE_ITER_IS_EXTENTS|BTREE_ITER_FILTER_SNAPSHOTS))) {
+	if ((iter->flags & BTREE_ITER_cached) ||
+	    !(iter->flags & (BTREE_ITER_is_extents|BTREE_ITER_filter_snapshots))) {
 		k = bkey_s_c_null;
 
-		if (unlikely((iter->flags & BTREE_ITER_WITH_UPDATES) &&
+		if (unlikely((iter->flags & BTREE_ITER_with_updates) &&
 			     trans->nr_updates)) {
 			bch2_btree_trans_peek_slot_updates(trans, iter, &k);
 			if (k.k)
 				goto out;
 		}
 
-		if (unlikely(iter->flags & BTREE_ITER_WITH_JOURNAL) &&
+		if (unlikely(iter->flags & BTREE_ITER_with_journal) &&
 		    (k = btree_trans_peek_slot_journal(trans, iter)).k)
 			goto out;
 
-		if (unlikely(iter->flags & BTREE_ITER_WITH_KEY_CACHE) &&
+		if (unlikely(iter->flags & BTREE_ITER_with_key_cache) &&
 		    (k = btree_trans_peek_key_cache(iter, iter->pos)).k) {
 			if (!bkey_err(k))
 				iter->k = *k.k;
@@ -2506,12 +2634,12 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
 		struct bpos next;
 		struct bpos end = iter->pos;
 
-		if (iter->flags & BTREE_ITER_IS_EXTENTS)
+		if (iter->flags & BTREE_ITER_is_extents)
 			end.offset = U64_MAX;
 
 		EBUG_ON(btree_iter_path(trans, iter)->level);
 
-		if (iter->flags & BTREE_ITER_INTENT) {
+		if (iter->flags & BTREE_ITER_intent) {
 			struct btree_iter iter2;
 
 			bch2_trans_copy_iter(&iter2, iter);
@@ -2542,7 +2670,7 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
 			bkey_init(&iter->k);
 			iter->k.p = iter->pos;
 
-			if (iter->flags & BTREE_ITER_IS_EXTENTS) {
+			if (iter->flags & BTREE_ITER_is_extents) {
 				bch2_key_resize(&iter->k,
 						min_t(u64, KEY_SIZE_MAX,
 						      (next.inode == iter->pos.inode
@@ -2726,13 +2854,13 @@ void bch2_trans_iter_exit(struct btree_trans *trans, struct btree_iter *iter)
 {
 	if (iter->update_path)
 		bch2_path_put_nokeep(trans, iter->update_path,
-			      iter->flags & BTREE_ITER_INTENT);
+			      iter->flags & BTREE_ITER_intent);
 	if (iter->path)
 		bch2_path_put(trans, iter->path,
-			      iter->flags & BTREE_ITER_INTENT);
+			      iter->flags & BTREE_ITER_intent);
 	if (iter->key_cache_path)
 		bch2_path_put(trans, iter->key_cache_path,
-			      iter->flags & BTREE_ITER_INTENT);
+			      iter->flags & BTREE_ITER_intent);
 	iter->path		= 0;
 	iter->update_path	= 0;
 	iter->key_cache_path	= 0;
@@ -2757,9 +2885,9 @@ void bch2_trans_node_iter_init(struct btree_trans *trans,
 			       unsigned depth,
 			       unsigned flags)
 {
-	flags |= BTREE_ITER_NOT_EXTENTS;
-	flags |= __BTREE_ITER_ALL_SNAPSHOTS;
-	flags |= BTREE_ITER_ALL_SNAPSHOTS;
+	flags |= BTREE_ITER_not_extents;
+	flags |= BTREE_ITER_snapshot_field;
+	flags |= BTREE_ITER_all_snapshots;
 
 	bch2_trans_iter_init_common(trans, iter, btree_id, pos, locks_want, depth,
 			       __bch2_btree_iter_flags(trans, btree_id, flags),
@@ -2782,9 +2910,9 @@ void bch2_trans_copy_iter(struct btree_iter *dst, struct btree_iter *src)
 	dst->ip_allocated = _RET_IP_;
 #endif
 	if (src->path)
-		__btree_path_get(trans->paths + src->path, src->flags & BTREE_ITER_INTENT);
+		__btree_path_get(trans->paths + src->path, src->flags & BTREE_ITER_intent);
 	if (src->update_path)
-		__btree_path_get(trans->paths + src->update_path, src->flags & BTREE_ITER_INTENT);
+		__btree_path_get(trans->paths + src->update_path, src->flags & BTREE_ITER_intent);
 	dst->key_cache_path = 0;
 }
 
@@ -2953,7 +3081,8 @@ u32 bch2_trans_begin(struct btree_trans *trans)
 	if (!trans->restarted &&
 	    (need_resched() ||
 	     time_after64(now, trans->last_begin_time + BTREE_TRANS_MAX_LOCK_HOLD_TIME_NS))) {
-		drop_locks_do(trans, (cond_resched(), 0));
+		bch2_trans_unlock(trans);
+		cond_resched();
 		now = local_clock();
 	}
 	trans->last_begin_time = now;
@@ -2963,11 +3092,14 @@ u32 bch2_trans_begin(struct btree_trans *trans)
 		bch2_trans_srcu_unlock(trans);
 
 	trans->last_begin_ip = _RET_IP_;
+	trans->locked  = true;
+
 	if (trans->restarted) {
 		bch2_btree_path_traverse_all(trans);
 		trans->notrace_relock_fail = false;
 	}
 
+	bch2_trans_verify_not_unlocked(trans);
 	return trans->restart_count;
 }
 
@@ -3020,7 +3152,7 @@ struct btree_trans *__bch2_trans_get(struct bch_fs *c, unsigned fn_idx)
 			 */
 			BUG_ON(pos_task &&
 			       pid == pos_task->pid &&
-			       bch2_trans_locked(pos));
+			       pos->locked);
 
 			if (pos_task && pid < pos_task->pid) {
 				list_add_tail(&trans->list, &pos->list);
@@ -3036,8 +3168,9 @@ got_trans:
 	trans->last_begin_time	= local_clock();
 	trans->fn_idx		= fn_idx;
 	trans->locking_wait.task = current;
+	trans->locked		= true;
 	trans->journal_replay_not_finished =
-		unlikely(!test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)) &&
+		unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)) &&
 		atomic_inc_not_zero(&c->journal_keys.ref);
 	trans->nr_paths		= ARRAY_SIZE(trans->_paths);
 	trans->paths_allocated	= trans->_paths_allocated;
@@ -3166,13 +3299,11 @@ bch2_btree_bkey_cached_common_to_text(struct printbuf *out,
 	pid = owner ? owner->pid : 0;
 	rcu_read_unlock();
 
-	prt_tab(out);
-	prt_printf(out, "%px %c l=%u %s:", b, b->cached ? 'c' : 'b',
+	prt_printf(out, "\t%px %c l=%u %s:", b, b->cached ? 'c' : 'b',
 		   b->level, bch2_btree_id_str(b->btree_id));
 	bch2_bpos_to_text(out, btree_node_pos(b));
 
-	prt_tab(out);
-	prt_printf(out, " locks %u:%u:%u held by pid %u",
+	prt_printf(out, "\t locks %u:%u:%u held by pid %u",
 		   c.n[0], c.n[1], c.n[2], pid);
 }
 
@@ -3229,10 +3360,8 @@ void bch2_btree_trans_to_text(struct printbuf *out, struct btree_trans *trans)
 
 	b = READ_ONCE(trans->locking);
 	if (b) {
-		prt_printf(out, "  blocked for %lluus on",
-			   div_u64(local_clock() - trans->locking_wait.start_time,
-				   1000));
-		prt_newline(out);
+		prt_printf(out, "  blocked for %lluus on\n",
+			   div_u64(local_clock() - trans->locking_wait.start_time, 1000));
 		prt_printf(out, "    %c", lock_types[trans->locking_wait.lock_want]);
 		bch2_btree_bkey_cached_common_to_text(out, b);
 		prt_newline(out);
diff --git a/fs/bcachefs/btree_iter.h b/fs/bcachefs/btree_iter.h
index 1c70836dd7cce4..eab2a25bdc7a26 100644
--- a/fs/bcachefs/btree_iter.h
+++ b/fs/bcachefs/btree_iter.h
@@ -216,9 +216,13 @@ int __must_check bch2_btree_path_traverse_one(struct btree_trans *,
 					      btree_path_idx_t,
 					      unsigned, unsigned long);
 
+static inline void bch2_trans_verify_not_unlocked(struct btree_trans *);
+
 static inline int __must_check bch2_btree_path_traverse(struct btree_trans *trans,
 					  btree_path_idx_t path, unsigned flags)
 {
+	bch2_trans_verify_not_unlocked(trans);
+
 	if (trans->paths[path].uptodate < BTREE_ITER_NEED_RELOCK)
 		return 0;
 
@@ -227,6 +231,9 @@ static inline int __must_check bch2_btree_path_traverse(struct btree_trans *tran
 
 btree_path_idx_t bch2_path_get(struct btree_trans *, enum btree_id, struct bpos,
 				 unsigned, unsigned, unsigned, unsigned long);
+btree_path_idx_t bch2_path_get_unlocked_mut(struct btree_trans *, enum btree_id,
+					    unsigned, struct bpos);
+
 struct bkey_s_c bch2_btree_path_peek_slot(struct btree_path *, struct bkey *);
 
 /*
@@ -283,7 +290,6 @@ int bch2_trans_relock(struct btree_trans *);
 int bch2_trans_relock_notrace(struct btree_trans *);
 void bch2_trans_unlock(struct btree_trans *);
 void bch2_trans_unlock_long(struct btree_trans *);
-bool bch2_trans_locked(struct btree_trans *);
 
 static inline int trans_was_restarted(struct btree_trans *trans, u32 restart_count)
 {
@@ -309,6 +315,14 @@ static inline void bch2_trans_verify_not_in_restart(struct btree_trans *trans)
 		bch2_trans_in_restart_error(trans);
 }
 
+void __noreturn bch2_trans_unlocked_error(struct btree_trans *);
+
+static inline void bch2_trans_verify_not_unlocked(struct btree_trans *trans)
+{
+	if (!trans->locked)
+		bch2_trans_unlocked_error(trans);
+}
+
 __always_inline
 static int btree_trans_restart_nounlock(struct btree_trans *trans, int err)
 {
@@ -386,10 +400,10 @@ static inline void bch2_btree_iter_set_pos(struct btree_iter *iter, struct bpos
 
 	if (unlikely(iter->update_path))
 		bch2_path_put(trans, iter->update_path,
-			      iter->flags & BTREE_ITER_INTENT);
+			      iter->flags & BTREE_ITER_intent);
 	iter->update_path = 0;
 
-	if (!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS))
+	if (!(iter->flags & BTREE_ITER_all_snapshots))
 		new_pos.snapshot = iter->snapshot;
 
 	__bch2_btree_iter_set_pos(iter, new_pos);
@@ -397,7 +411,7 @@ static inline void bch2_btree_iter_set_pos(struct btree_iter *iter, struct bpos
 
 static inline void bch2_btree_iter_set_pos_to_extent_start(struct btree_iter *iter)
 {
-	BUG_ON(!(iter->flags & BTREE_ITER_IS_EXTENTS));
+	BUG_ON(!(iter->flags & BTREE_ITER_is_extents));
 	iter->pos = bkey_start_pos(&iter->k);
 }
 
@@ -416,20 +430,20 @@ static inline unsigned __bch2_btree_iter_flags(struct btree_trans *trans,
 					       unsigned btree_id,
 					       unsigned flags)
 {
-	if (!(flags & (BTREE_ITER_ALL_SNAPSHOTS|BTREE_ITER_NOT_EXTENTS)) &&
+	if (!(flags & (BTREE_ITER_all_snapshots|BTREE_ITER_not_extents)) &&
 	    btree_id_is_extents(btree_id))
-		flags |= BTREE_ITER_IS_EXTENTS;
+		flags |= BTREE_ITER_is_extents;
 
-	if (!(flags & __BTREE_ITER_ALL_SNAPSHOTS) &&
+	if (!(flags & BTREE_ITER_snapshot_field) &&
 	    !btree_type_has_snapshot_field(btree_id))
-		flags &= ~BTREE_ITER_ALL_SNAPSHOTS;
+		flags &= ~BTREE_ITER_all_snapshots;
 
-	if (!(flags & BTREE_ITER_ALL_SNAPSHOTS) &&
+	if (!(flags & BTREE_ITER_all_snapshots) &&
 	    btree_type_has_snapshots(btree_id))
-		flags |= BTREE_ITER_FILTER_SNAPSHOTS;
+		flags |= BTREE_ITER_filter_snapshots;
 
 	if (trans->journal_replay_not_finished)
-		flags |= BTREE_ITER_WITH_JOURNAL;
+		flags |= BTREE_ITER_with_journal;
 
 	return flags;
 }
@@ -439,10 +453,10 @@ static inline unsigned bch2_btree_iter_flags(struct btree_trans *trans,
 					     unsigned flags)
 {
 	if (!btree_id_cached(trans->c, btree_id)) {
-		flags &= ~BTREE_ITER_CACHED;
-		flags &= ~BTREE_ITER_WITH_KEY_CACHE;
-	} else if (!(flags & BTREE_ITER_CACHED))
-		flags |= BTREE_ITER_WITH_KEY_CACHE;
+		flags &= ~BTREE_ITER_cached;
+		flags &= ~BTREE_ITER_with_key_cache;
+	} else if (!(flags & BTREE_ITER_cached))
+		flags |= BTREE_ITER_with_key_cache;
 
 	return __bch2_btree_iter_flags(trans, btree_id, flags);
 }
@@ -494,18 +508,7 @@ void bch2_trans_node_iter_init(struct btree_trans *, struct btree_iter *,
 			       unsigned, unsigned, unsigned);
 void bch2_trans_copy_iter(struct btree_iter *, struct btree_iter *);
 
-static inline void set_btree_iter_dontneed(struct btree_iter *iter)
-{
-	struct btree_trans *trans = iter->trans;
-
-	if (!iter->path || trans->restarted)
-		return;
-
-	struct btree_path *path = btree_iter_path(trans, iter);
-	path->preserve		= false;
-	if (path->ref == 1)
-		path->should_be_locked	= false;
-}
+void bch2_set_btree_iter_dontneed(struct btree_iter *);
 
 void *__bch2_trans_kmalloc(struct btree_trans *, size_t);
 
@@ -619,14 +622,14 @@ u32 bch2_trans_begin(struct btree_trans *);
 static inline struct bkey_s_c bch2_btree_iter_peek_prev_type(struct btree_iter *iter,
 							     unsigned flags)
 {
-	return  flags & BTREE_ITER_SLOTS      ? bch2_btree_iter_peek_slot(iter) :
+	return  flags & BTREE_ITER_slots      ? bch2_btree_iter_peek_slot(iter) :
 						bch2_btree_iter_peek_prev(iter);
 }
 
 static inline struct bkey_s_c bch2_btree_iter_peek_type(struct btree_iter *iter,
 							unsigned flags)
 {
-	return  flags & BTREE_ITER_SLOTS      ? bch2_btree_iter_peek_slot(iter) :
+	return  flags & BTREE_ITER_slots      ? bch2_btree_iter_peek_slot(iter) :
 						bch2_btree_iter_peek(iter);
 }
 
@@ -634,7 +637,7 @@ static inline struct bkey_s_c bch2_btree_iter_peek_upto_type(struct btree_iter *
 							     struct bpos end,
 							     unsigned flags)
 {
-	if (!(flags & BTREE_ITER_SLOTS))
+	if (!(flags & BTREE_ITER_slots))
 		return bch2_btree_iter_peek_upto(iter, end);
 
 	if (bkey_gt(iter->pos, end))
@@ -699,16 +702,12 @@ transaction_restart:							\
 	_ret2 ?: trans_was_restarted(_trans, _restart_count);		\
 })
 
-#define for_each_btree_key_upto(_trans, _iter, _btree_id,		\
-				_start, _end, _flags, _k, _do)		\
+#define for_each_btree_key_upto_continue(_trans, _iter,			\
+					 _end, _flags, _k, _do)		\
 ({									\
-	struct btree_iter _iter;					\
 	struct bkey_s_c _k;						\
 	int _ret3 = 0;							\
 									\
-	bch2_trans_iter_init((_trans), &(_iter), (_btree_id),		\
-			     (_start), (_flags));			\
-									\
 	do {								\
 		_ret3 = lockrestart_do(_trans, ({			\
 			(_k) = bch2_btree_iter_peek_upto_type(&(_iter),	\
@@ -724,6 +723,21 @@ transaction_restart:							\
 	_ret3;								\
 })
 
+#define for_each_btree_key_continue(_trans, _iter, _flags, _k, _do)	\
+	for_each_btree_key_upto_continue(_trans, _iter, SPOS_MAX, _flags, _k, _do)
+
+#define for_each_btree_key_upto(_trans, _iter, _btree_id,		\
+				_start, _end, _flags, _k, _do)		\
+({									\
+	bch2_trans_begin(trans);					\
+									\
+	struct btree_iter _iter;					\
+	bch2_trans_iter_init((_trans), &(_iter), (_btree_id),		\
+			     (_start), (_flags));			\
+									\
+	for_each_btree_key_upto_continue(_trans, _iter, _end, _flags, _k, _do);\
+})
+
 #define for_each_btree_key(_trans, _iter, _btree_id,			\
 			   _start, _flags, _k, _do)			\
 	for_each_btree_key_upto(_trans, _iter, _btree_id, _start,	\
@@ -794,14 +808,6 @@ __bch2_btree_iter_peek_and_restart(struct btree_trans *trans,
 	return k;
 }
 
-#define for_each_btree_key_old(_trans, _iter, _btree_id,		\
-			   _start, _flags, _k, _ret)			\
-	for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id),	\
-				  (_start), (_flags));			\
-	     (_k) = __bch2_btree_iter_peek_and_restart((_trans), &(_iter), _flags),\
-	     !((_ret) = bkey_err(_k)) && (_k).k;			\
-	     bch2_btree_iter_advance(&(_iter)))
-
 #define for_each_btree_key_upto_norestart(_trans, _iter, _btree_id,	\
 			   _start, _end, _flags, _k, _ret)		\
 	for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id),	\
@@ -861,6 +867,7 @@ __bch2_btree_iter_peek_and_restart(struct btree_trans *trans,
 })
 
 void bch2_trans_updates_to_text(struct printbuf *, struct btree_trans *);
+void bch2_btree_path_to_text(struct printbuf *, struct btree_trans *, btree_path_idx_t);
 void bch2_trans_paths_to_text(struct printbuf *, struct btree_trans *);
 void bch2_dump_trans_updates(struct btree_trans *);
 void bch2_dump_trans_paths_updates(struct btree_trans *);
diff --git a/fs/bcachefs/btree_journal_iter.c b/fs/bcachefs/btree_journal_iter.c
index 1e8cf49a693531..332dbf164929ed 100644
--- a/fs/bcachefs/btree_journal_iter.c
+++ b/fs/bcachefs/btree_journal_iter.c
@@ -623,3 +623,20 @@ void bch2_shoot_down_journal_keys(struct bch_fs *c, enum btree_id btree,
 			keys->data[dst++] = *i;
 	keys->nr = keys->gap = dst;
 }
+
+void bch2_journal_keys_dump(struct bch_fs *c)
+{
+	struct journal_keys *keys = &c->journal_keys;
+	struct printbuf buf = PRINTBUF;
+
+	pr_info("%zu keys:", keys->nr);
+
+	move_gap(keys, keys->nr);
+
+	darray_for_each(*keys, i) {
+		printbuf_reset(&buf);
+		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(i->k));
+		pr_err("%s l=%u %s", bch2_btree_id_str(i->btree_id), i->level, buf.buf);
+	}
+	printbuf_exit(&buf);
+}
diff --git a/fs/bcachefs/btree_journal_iter.h b/fs/bcachefs/btree_journal_iter.h
index af25046ebcaa76..1ba4a79b0ef905 100644
--- a/fs/bcachefs/btree_journal_iter.h
+++ b/fs/bcachefs/btree_journal_iter.h
@@ -70,4 +70,6 @@ void bch2_shoot_down_journal_keys(struct bch_fs *, enum btree_id,
 				  unsigned, unsigned,
 				  struct bpos, struct bpos);
 
+void bch2_journal_keys_dump(struct bch_fs *);
+
 #endif /* _BCACHEFS_BTREE_JOURNAL_ITER_H */
diff --git a/fs/bcachefs/btree_key_cache.c b/fs/bcachefs/btree_key_cache.c
index e8c1c530cd95f5..9dc0bf361da60b 100644
--- a/fs/bcachefs/btree_key_cache.c
+++ b/fs/bcachefs/btree_key_cache.c
@@ -383,9 +383,9 @@ static int btree_key_cache_fill(struct btree_trans *trans,
 	int ret;
 
 	bch2_trans_iter_init(trans, &iter, ck->key.btree_id, ck->key.pos,
-			     BTREE_ITER_KEY_CACHE_FILL|
-			     BTREE_ITER_CACHED_NOFILL);
-	iter.flags &= ~BTREE_ITER_WITH_JOURNAL;
+			     BTREE_ITER_key_cache_fill|
+			     BTREE_ITER_cached_nofill);
+	iter.flags &= ~BTREE_ITER_with_journal;
 	k = bch2_btree_iter_peek_slot(&iter);
 	ret = bkey_err(k);
 	if (ret)
@@ -456,7 +456,7 @@ static int btree_key_cache_fill(struct btree_trans *trans,
 	bch2_btree_node_unlock_write(trans, ck_path, ck_path->l[0].b);
 
 	/* We're not likely to need this iterator again: */
-	set_btree_iter_dontneed(&iter);
+	bch2_set_btree_iter_dontneed(&iter);
 err:
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
@@ -515,23 +515,10 @@ retry:
 fill:
 	path->uptodate = BTREE_ITER_UPTODATE;
 
-	if (!ck->valid && !(flags & BTREE_ITER_CACHED_NOFILL)) {
-		/*
-		 * Using the underscore version because we haven't set
-		 * path->uptodate yet:
-		 */
-		if (!path->locks_want &&
-		    !__bch2_btree_path_upgrade(trans, path, 1, NULL)) {
-			trace_and_count(trans->c, trans_restart_key_cache_upgrade, trans, _THIS_IP_);
-			ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_upgrade);
-			goto err;
-		}
-
-		ret = btree_key_cache_fill(trans, path, ck);
-		if (ret)
-			goto err;
-
-		ret = bch2_btree_path_relock(trans, path, _THIS_IP_);
+	if (!ck->valid && !(flags & BTREE_ITER_cached_nofill)) {
+		ret =   bch2_btree_path_upgrade(trans, path, 1) ?:
+			btree_key_cache_fill(trans, path, ck) ?:
+			bch2_btree_path_relock(trans, path, _THIS_IP_);
 		if (ret)
 			goto err;
 
@@ -622,13 +609,13 @@ static int btree_key_cache_flush_pos(struct btree_trans *trans,
 	int ret;
 
 	bch2_trans_iter_init(trans, &b_iter, key.btree_id, key.pos,
-			     BTREE_ITER_SLOTS|
-			     BTREE_ITER_INTENT|
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_slots|
+			     BTREE_ITER_intent|
+			     BTREE_ITER_all_snapshots);
 	bch2_trans_iter_init(trans, &c_iter, key.btree_id, key.pos,
-			     BTREE_ITER_CACHED|
-			     BTREE_ITER_INTENT);
-	b_iter.flags &= ~BTREE_ITER_WITH_KEY_CACHE;
+			     BTREE_ITER_cached|
+			     BTREE_ITER_intent);
+	b_iter.flags &= ~BTREE_ITER_with_key_cache;
 
 	ret = bch2_btree_iter_traverse(&c_iter);
 	if (ret)
@@ -661,14 +648,14 @@ static int btree_key_cache_flush_pos(struct btree_trans *trans,
 		commit_flags |= BCH_WATERMARK_reclaim;
 
 	if (ck->journal.seq != journal_last_seq(j) ||
-	    !test_bit(JOURNAL_SPACE_LOW, &c->journal.flags))
+	    !test_bit(JOURNAL_space_low, &c->journal.flags))
 		commit_flags |= BCH_TRANS_COMMIT_no_journal_res;
 
 	ret   = bch2_btree_iter_traverse(&b_iter) ?:
 		bch2_trans_update(trans, &b_iter, ck->k,
-				  BTREE_UPDATE_KEY_CACHE_RECLAIM|
-				  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
-				  BTREE_TRIGGER_NORUN) ?:
+				  BTREE_UPDATE_key_cache_reclaim|
+				  BTREE_UPDATE_internal_snapshot_node|
+				  BTREE_TRIGGER_norun) ?:
 		bch2_trans_commit(trans, NULL, NULL,
 				  BCH_TRANS_COMMIT_no_check_rw|
 				  BCH_TRANS_COMMIT_no_enospc|
@@ -790,7 +777,7 @@ bool bch2_btree_insert_key_cached(struct btree_trans *trans,
 	 * flushing. The flush callback will not proceed unless ->seq matches
 	 * the latest pin, so make sure it starts with a consistent value.
 	 */
-	if (!(insert_entry->flags & BTREE_UPDATE_NOJOURNAL) ||
+	if (!(insert_entry->flags & BTREE_UPDATE_nojournal) ||
 	    !journal_pin_active(&ck->journal)) {
 		ck->seq = trans->journal_res.seq;
 	}
@@ -835,6 +822,8 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
 	int srcu_idx;
 
 	mutex_lock(&bc->lock);
+	bc->requested_to_free += sc->nr_to_scan;
+
 	srcu_idx = srcu_read_lock(&c->btree_trans_barrier);
 	flags = memalloc_nofs_save();
 
@@ -853,6 +842,7 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
 		atomic_long_dec(&bc->nr_freed);
 		freed++;
 		bc->nr_freed_nonpcpu--;
+		bc->freed++;
 	}
 
 	list_for_each_entry_safe(ck, t, &bc->freed_pcpu, list) {
@@ -866,6 +856,7 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
 		atomic_long_dec(&bc->nr_freed);
 		freed++;
 		bc->nr_freed_pcpu--;
+		bc->freed++;
 	}
 
 	rcu_read_lock();
@@ -884,13 +875,18 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
 			ck = container_of(pos, struct bkey_cached, hash);
 
 			if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
+				bc->skipped_dirty++;
 				goto next;
 			} else if (test_bit(BKEY_CACHED_ACCESSED, &ck->flags)) {
 				clear_bit(BKEY_CACHED_ACCESSED, &ck->flags);
+				bc->skipped_accessed++;
 				goto next;
 			} else if (bkey_cached_lock_for_evict(ck)) {
 				bkey_cached_evict(bc, ck);
 				bkey_cached_free(bc, ck);
+				bc->moved_to_freelist++;
+			} else {
+				bc->skipped_lock_fail++;
 			}
 
 			scanned++;
@@ -1035,14 +1031,47 @@ int bch2_fs_btree_key_cache_init(struct btree_key_cache *bc)
 	return 0;
 }
 
-void bch2_btree_key_cache_to_text(struct printbuf *out, struct btree_key_cache *c)
+void bch2_btree_key_cache_to_text(struct printbuf *out, struct btree_key_cache *bc)
 {
-	prt_printf(out, "nr_freed:\t%lu",	atomic_long_read(&c->nr_freed));
-	prt_newline(out);
-	prt_printf(out, "nr_keys:\t%lu",	atomic_long_read(&c->nr_keys));
-	prt_newline(out);
-	prt_printf(out, "nr_dirty:\t%lu",	atomic_long_read(&c->nr_dirty));
-	prt_newline(out);
+	struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
+
+	printbuf_tabstop_push(out, 24);
+	printbuf_tabstop_push(out, 12);
+
+	unsigned flags = memalloc_nofs_save();
+	mutex_lock(&bc->lock);
+	prt_printf(out, "keys:\t%lu\r\n",		atomic_long_read(&bc->nr_keys));
+	prt_printf(out, "dirty:\t%lu\r\n",		atomic_long_read(&bc->nr_dirty));
+	prt_printf(out, "freelist:\t%lu\r\n",		atomic_long_read(&bc->nr_freed));
+	prt_printf(out, "nonpcpu freelist:\t%zu\r\n",	bc->nr_freed_nonpcpu);
+	prt_printf(out, "pcpu freelist:\t%zu\r\n",	bc->nr_freed_pcpu);
+
+	prt_printf(out, "\nshrinker:\n");
+	prt_printf(out, "requested_to_free:\t%lu\r\n",	bc->requested_to_free);
+	prt_printf(out, "freed:\t%lu\r\n",		bc->freed);
+	prt_printf(out, "moved_to_freelist:\t%lu\r\n",	bc->moved_to_freelist);
+	prt_printf(out, "skipped_dirty:\t%lu\r\n",	bc->skipped_dirty);
+	prt_printf(out, "skipped_accessed:\t%lu\r\n",	bc->skipped_accessed);
+	prt_printf(out, "skipped_lock_fail:\t%lu\r\n",	bc->skipped_lock_fail);
+
+	prt_printf(out, "srcu seq:\t%lu\r\n",		get_state_synchronize_srcu(&c->btree_trans_barrier));
+
+	struct bkey_cached *ck;
+	unsigned iter = 0;
+	list_for_each_entry(ck, &bc->freed_nonpcpu, list) {
+		prt_printf(out, "freed_nonpcpu:\t%lu\r\n", ck->btree_trans_barrier_seq);
+		if (++iter > 10)
+			break;
+	}
+
+	iter = 0;
+	list_for_each_entry(ck, &bc->freed_pcpu, list) {
+		prt_printf(out, "freed_pcpu:\t%lu\r\n", ck->btree_trans_barrier_seq);
+		if (++iter > 10)
+			break;
+	}
+	mutex_unlock(&bc->lock);
+	memalloc_flags_restore(flags);
 }
 
 void bch2_btree_key_cache_exit(void)
diff --git a/fs/bcachefs/btree_key_cache_types.h b/fs/bcachefs/btree_key_cache_types.h
index 290e4e57df5bbc..237e8bb3ac407f 100644
--- a/fs/bcachefs/btree_key_cache_types.h
+++ b/fs/bcachefs/btree_key_cache_types.h
@@ -24,6 +24,14 @@ struct btree_key_cache {
 	atomic_long_t		nr_freed;
 	atomic_long_t		nr_keys;
 	atomic_long_t		nr_dirty;
+
+	/* shrinker stats */
+	unsigned long		requested_to_free;
+	unsigned long		freed;
+	unsigned long		moved_to_freelist;
+	unsigned long		skipped_dirty;
+	unsigned long		skipped_accessed;
+	unsigned long		skipped_lock_fail;
 };
 
 struct bkey_cached_key {
diff --git a/fs/bcachefs/btree_locking.c b/fs/bcachefs/btree_locking.c
index f2caf491957efc..c3e9b0cc7bbdc9 100644
--- a/fs/bcachefs/btree_locking.c
+++ b/fs/bcachefs/btree_locking.c
@@ -83,8 +83,7 @@ static noinline void print_cycle(struct printbuf *out, struct lock_graph *g)
 {
 	struct trans_waiting_for_lock *i;
 
-	prt_printf(out, "Found lock cycle (%u entries):", g->nr);
-	prt_newline(out);
+	prt_printf(out, "Found lock cycle (%u entries):\n", g->nr);
 
 	for (i = g->g; i < g->g + g->nr; i++) {
 		struct task_struct *task = READ_ONCE(i->trans->locking_wait.task);
@@ -224,8 +223,7 @@ static noinline int break_cycle(struct lock_graph *g, struct printbuf *cycle)
 
 			bch2_btree_trans_to_text(&buf, trans);
 
-			prt_printf(&buf, "backtrace:");
-			prt_newline(&buf);
+			prt_printf(&buf, "backtrace:\n");
 			printbuf_indent_add(&buf, 2);
 			bch2_prt_task_backtrace(&buf, trans->locking_wait.task, 2, GFP_NOWAIT);
 			printbuf_indent_sub(&buf, 2);
@@ -492,8 +490,6 @@ static inline bool btree_path_get_locks(struct btree_trans *trans,
 	if (path->uptodate == BTREE_ITER_NEED_RELOCK)
 		path->uptodate = BTREE_ITER_UPTODATE;
 
-	bch2_trans_verify_locks(trans);
-
 	return path->uptodate < BTREE_ITER_NEED_RELOCK;
 }
 
@@ -609,7 +605,9 @@ bool bch2_btree_path_relock_norestart(struct btree_trans *trans, struct btree_pa
 {
 	struct get_locks_fail f;
 
-	return btree_path_get_locks(trans, path, false, &f);
+	bool ret = btree_path_get_locks(trans, path, false, &f);
+	bch2_trans_verify_locks(trans);
+	return ret;
 }
 
 int __bch2_btree_path_relock(struct btree_trans *trans,
@@ -632,7 +630,9 @@ bool bch2_btree_path_upgrade_noupgrade_sibs(struct btree_trans *trans,
 
 	path->locks_want = new_locks_want;
 
-	return btree_path_get_locks(trans, path, true, f);
+	bool ret = btree_path_get_locks(trans, path, true, f);
+	bch2_trans_verify_locks(trans);
+	return ret;
 }
 
 bool __bch2_btree_path_upgrade(struct btree_trans *trans,
@@ -640,8 +640,9 @@ bool __bch2_btree_path_upgrade(struct btree_trans *trans,
 			       unsigned new_locks_want,
 			       struct get_locks_fail *f)
 {
-	if (bch2_btree_path_upgrade_noupgrade_sibs(trans, path, new_locks_want, f))
-		return true;
+	bool ret = bch2_btree_path_upgrade_noupgrade_sibs(trans, path, new_locks_want, f);
+	if (ret)
+		goto out;
 
 	/*
 	 * XXX: this is ugly - we'd prefer to not be mucking with other
@@ -675,8 +676,9 @@ bool __bch2_btree_path_upgrade(struct btree_trans *trans,
 				btree_path_get_locks(trans, linked, true, NULL);
 			}
 	}
-
-	return false;
+out:
+	bch2_trans_verify_locks(trans);
+	return ret;
 }
 
 void __bch2_btree_path_downgrade(struct btree_trans *trans,
@@ -725,82 +727,100 @@ void bch2_trans_downgrade(struct btree_trans *trans)
 			bch2_btree_path_downgrade(trans, path);
 }
 
-int bch2_trans_relock(struct btree_trans *trans)
+static inline void __bch2_trans_unlock(struct btree_trans *trans)
 {
 	struct btree_path *path;
 	unsigned i;
 
-	if (unlikely(trans->restarted))
-		return -((int) trans->restarted);
+	trans_for_each_path(trans, path, i)
+		__bch2_btree_path_unlock(trans, path);
+}
 
-	trans_for_each_path(trans, path, i) {
-		struct get_locks_fail f;
+static noinline __cold int bch2_trans_relock_fail(struct btree_trans *trans, struct btree_path *path,
+						  struct get_locks_fail *f, bool trace)
+{
+	if (!trace)
+		goto out;
 
-		if (path->should_be_locked &&
-		    !btree_path_get_locks(trans, path, false, &f)) {
-			if (trace_trans_restart_relock_enabled()) {
-				struct printbuf buf = PRINTBUF;
-
-				bch2_bpos_to_text(&buf, path->pos);
-				prt_printf(&buf, " l=%u seq=%u node seq=",
-					   f.l, path->l[f.l].lock_seq);
-				if (IS_ERR_OR_NULL(f.b)) {
-					prt_str(&buf, bch2_err_str(PTR_ERR(f.b)));
-				} else {
-					prt_printf(&buf, "%u", f.b->c.lock.seq);
-
-					struct six_lock_count c =
-						bch2_btree_node_lock_counts(trans, NULL, &f.b->c, f.l);
-					prt_printf(&buf, " self locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
-
-					c = six_lock_counts(&f.b->c.lock);
-					prt_printf(&buf, " total locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
-				}
+	if (trace_trans_restart_relock_enabled()) {
+		struct printbuf buf = PRINTBUF;
 
-				trace_trans_restart_relock(trans, _RET_IP_, buf.buf);
-				printbuf_exit(&buf);
-			}
+		bch2_bpos_to_text(&buf, path->pos);
+		prt_printf(&buf, " l=%u seq=%u node seq=", f->l, path->l[f->l].lock_seq);
+		if (IS_ERR_OR_NULL(f->b)) {
+			prt_str(&buf, bch2_err_str(PTR_ERR(f->b)));
+		} else {
+			prt_printf(&buf, "%u", f->b->c.lock.seq);
 
-			count_event(trans->c, trans_restart_relock);
-			return btree_trans_restart(trans, BCH_ERR_transaction_restart_relock);
+			struct six_lock_count c =
+				bch2_btree_node_lock_counts(trans, NULL, &f->b->c, f->l);
+			prt_printf(&buf, " self locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
+
+			c = six_lock_counts(&f->b->c.lock);
+			prt_printf(&buf, " total locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
 		}
+
+		trace_trans_restart_relock(trans, _RET_IP_, buf.buf);
+		printbuf_exit(&buf);
 	}
 
-	return 0;
+	count_event(trans->c, trans_restart_relock);
+out:
+	__bch2_trans_unlock(trans);
+	bch2_trans_verify_locks(trans);
+	return btree_trans_restart(trans, BCH_ERR_transaction_restart_relock);
 }
 
-int bch2_trans_relock_notrace(struct btree_trans *trans)
+static inline int __bch2_trans_relock(struct btree_trans *trans, bool trace)
 {
-	struct btree_path *path;
-	unsigned i;
+	bch2_trans_verify_locks(trans);
 
 	if (unlikely(trans->restarted))
 		return -((int) trans->restarted);
+	if (unlikely(trans->locked))
+		goto out;
+
+	struct btree_path *path;
+	unsigned i;
+
+	trans_for_each_path(trans, path, i) {
+		struct get_locks_fail f;
 
-	trans_for_each_path(trans, path, i)
 		if (path->should_be_locked &&
-		    !bch2_btree_path_relock_norestart(trans, path)) {
-			return btree_trans_restart(trans, BCH_ERR_transaction_restart_relock);
-		}
+		    !btree_path_get_locks(trans, path, false, &f))
+			return bch2_trans_relock_fail(trans, path, &f, trace);
+	}
+
+	trans->locked = true;
+out:
+	bch2_trans_verify_locks(trans);
 	return 0;
 }
 
+int bch2_trans_relock(struct btree_trans *trans)
+{
+	return __bch2_trans_relock(trans, true);
+}
+
+int bch2_trans_relock_notrace(struct btree_trans *trans)
+{
+	return __bch2_trans_relock(trans, false);
+}
+
 void bch2_trans_unlock_noassert(struct btree_trans *trans)
 {
-	struct btree_path *path;
-	unsigned i;
+	__bch2_trans_unlock(trans);
 
-	trans_for_each_path(trans, path, i)
-		__bch2_btree_path_unlock(trans, path);
+	trans->locked = false;
+	trans->last_unlock_ip = _RET_IP_;
 }
 
 void bch2_trans_unlock(struct btree_trans *trans)
 {
-	struct btree_path *path;
-	unsigned i;
+	__bch2_trans_unlock(trans);
 
-	trans_for_each_path(trans, path, i)
-		__bch2_btree_path_unlock(trans, path);
+	trans->locked = false;
+	trans->last_unlock_ip = _RET_IP_;
 }
 
 void bch2_trans_unlock_long(struct btree_trans *trans)
@@ -809,17 +829,6 @@ void bch2_trans_unlock_long(struct btree_trans *trans)
 	bch2_trans_srcu_unlock(trans);
 }
 
-bool bch2_trans_locked(struct btree_trans *trans)
-{
-	struct btree_path *path;
-	unsigned i;
-
-	trans_for_each_path(trans, path, i)
-		if (path->nodes_locked)
-			return true;
-	return false;
-}
-
 int __bch2_trans_mutex_lock(struct btree_trans *trans,
 			    struct mutex *lock)
 {
@@ -836,15 +845,19 @@ int __bch2_trans_mutex_lock(struct btree_trans *trans,
 
 void bch2_btree_path_verify_locks(struct btree_path *path)
 {
-	unsigned l;
+	/*
+	 * A path may be uptodate and yet have nothing locked if and only if
+	 * there is no node at path->level, which generally means we were
+	 * iterating over all nodes and got to the end of the btree
+	 */
+	BUG_ON(path->uptodate == BTREE_ITER_UPTODATE &&
+	       btree_path_node(path, path->level) &&
+	       !path->nodes_locked);
 
-	if (!path->nodes_locked) {
-		BUG_ON(path->uptodate == BTREE_ITER_UPTODATE &&
-		       btree_path_node(path, path->level));
+	if (!path->nodes_locked)
 		return;
-	}
 
-	for (l = 0; l < BTREE_MAX_DEPTH; l++) {
+	for (unsigned l = 0; l < BTREE_MAX_DEPTH; l++) {
 		int want = btree_lock_want(path, l);
 		int have = btree_node_locked_type(path, l);
 
@@ -857,8 +870,24 @@ void bch2_btree_path_verify_locks(struct btree_path *path)
 	}
 }
 
+static bool bch2_trans_locked(struct btree_trans *trans)
+{
+	struct btree_path *path;
+	unsigned i;
+
+	trans_for_each_path(trans, path, i)
+		if (path->nodes_locked)
+			return true;
+	return false;
+}
+
 void bch2_trans_verify_locks(struct btree_trans *trans)
 {
+	if (!trans->locked) {
+		BUG_ON(bch2_trans_locked(trans));
+		return;
+	}
+
 	struct btree_path *path;
 	unsigned i;
 
diff --git a/fs/bcachefs/btree_locking.h b/fs/bcachefs/btree_locking.h
index 4bd72c855da1a4..7f41545b9147f8 100644
--- a/fs/bcachefs/btree_locking.h
+++ b/fs/bcachefs/btree_locking.h
@@ -364,14 +364,14 @@ static inline int bch2_btree_path_upgrade(struct btree_trans *trans,
 					  struct btree_path *path,
 					  unsigned new_locks_want)
 {
-	struct get_locks_fail f;
+	struct get_locks_fail f = {};
 	unsigned old_locks_want = path->locks_want;
 
 	new_locks_want = min(new_locks_want, BTREE_MAX_DEPTH);
 
 	if (path->locks_want < new_locks_want
 	    ? __bch2_btree_path_upgrade(trans, path, new_locks_want, &f)
-	    : path->uptodate == BTREE_ITER_UPTODATE)
+	    : path->nodes_locked)
 		return 0;
 
 	trace_and_count(trans->c, trans_restart_upgrade, trans, _THIS_IP_, path,
diff --git a/fs/bcachefs/btree_trans_commit.c b/fs/bcachefs/btree_trans_commit.c
index bbec91e8e6506f..a219a87e9ef079 100644
--- a/fs/bcachefs/btree_trans_commit.c
+++ b/fs/bcachefs/btree_trans_commit.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
+#include "alloc_foreground.h"
 #include "btree_gc.h"
 #include "btree_io.h"
 #include "btree_iter.h"
@@ -19,6 +20,26 @@
 
 #include <linux/prefetch.h>
 
+static const char * const trans_commit_flags_strs[] = {
+#define x(n, ...) #n,
+	BCH_TRANS_COMMIT_FLAGS()
+#undef x
+	NULL
+};
+
+void bch2_trans_commit_flags_to_text(struct printbuf *out, enum bch_trans_commit_flags flags)
+{
+	enum bch_watermark watermark = flags & BCH_WATERMARK_MASK;
+
+	prt_printf(out, "watermark=%s", bch2_watermarks[watermark]);
+
+	flags >>= BCH_WATERMARK_BITS;
+	if (flags) {
+		prt_char(out, ' ');
+		bch2_prt_bitflags(out, trans_commit_flags_strs, flags);
+	}
+}
+
 static void verify_update_old_key(struct btree_trans *trans, struct btree_insert_entry *i)
 {
 #ifdef CONFIG_BCACHEFS_DEBUG
@@ -315,8 +336,8 @@ static inline void btree_insert_entry_checks(struct btree_trans *trans,
 	BUG_ON(i->btree_id	!= path->btree_id);
 	EBUG_ON(!i->level &&
 		btree_type_has_snapshots(i->btree_id) &&
-		!(i->flags & BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) &&
-		test_bit(JOURNAL_REPLAY_DONE, &trans->c->journal.flags) &&
+		!(i->flags & BTREE_UPDATE_internal_snapshot_node) &&
+		test_bit(JOURNAL_replay_done, &trans->c->journal.flags) &&
 		i->k->k.p.snapshot &&
 		bch2_snapshot_is_internal_node(trans->c, i->k->k.p.snapshot) > 0);
 }
@@ -443,13 +464,13 @@ static int run_one_mem_trigger(struct btree_trans *trans,
 
 	verify_update_old_key(trans, i);
 
-	if (unlikely(flags & BTREE_TRIGGER_NORUN))
+	if (unlikely(flags & BTREE_TRIGGER_norun))
 		return 0;
 
 	if (old_ops->trigger == new_ops->trigger) {
 		ret   = bch2_key_trigger(trans, i->btree_id, i->level,
 				old, bkey_i_to_s(new),
-				BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE|flags);
+				BTREE_TRIGGER_insert|BTREE_TRIGGER_overwrite|flags);
 	} else {
 		ret   = bch2_key_trigger_new(trans, i->btree_id, i->level,
 				bkey_i_to_s(new), flags) ?:
@@ -472,11 +493,11 @@ static int run_one_trans_trigger(struct btree_trans *trans, struct btree_insert_
 	struct bkey_s_c old = { &old_k, i->old_v };
 	const struct bkey_ops *old_ops = bch2_bkey_type_ops(old.k->type);
 	const struct bkey_ops *new_ops = bch2_bkey_type_ops(i->k->k.type);
-	unsigned flags = i->flags|BTREE_TRIGGER_TRANSACTIONAL;
+	unsigned flags = i->flags|BTREE_TRIGGER_transactional;
 
 	verify_update_old_key(trans, i);
 
-	if ((i->flags & BTREE_TRIGGER_NORUN) ||
+	if ((i->flags & BTREE_TRIGGER_norun) ||
 	    !(BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS & (1U << i->bkey_type)))
 		return 0;
 
@@ -486,8 +507,8 @@ static int run_one_trans_trigger(struct btree_trans *trans, struct btree_insert_
 		i->overwrite_trigger_run = true;
 		i->insert_trigger_run = true;
 		return bch2_key_trigger(trans, i->btree_id, i->level, old, bkey_i_to_s(i->k),
-					BTREE_TRIGGER_INSERT|
-					BTREE_TRIGGER_OVERWRITE|flags) ?: 1;
+					BTREE_TRIGGER_insert|
+					BTREE_TRIGGER_overwrite|flags) ?: 1;
 	} else if (overwrite && !i->overwrite_trigger_run) {
 		i->overwrite_trigger_run = true;
 		return bch2_key_trigger_old(trans, i->btree_id, i->level, old, flags) ?: 1;
@@ -572,7 +593,7 @@ static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
 
 #ifdef CONFIG_BCACHEFS_DEBUG
 	trans_for_each_update(trans, i)
-		BUG_ON(!(i->flags & BTREE_TRIGGER_NORUN) &&
+		BUG_ON(!(i->flags & BTREE_TRIGGER_norun) &&
 		       (BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS & (1U << i->bkey_type)) &&
 		       (!i->insert_trigger_run || !i->overwrite_trigger_run));
 #endif
@@ -590,7 +611,7 @@ static noinline int bch2_trans_commit_run_gc_triggers(struct btree_trans *trans)
 
 		if (btree_node_type_needs_gc(__btree_node_type(i->level, i->btree_id)) &&
 		    gc_visited(trans->c, gc_pos_btree_node(insert_l(trans, i)->b))) {
-			int ret = run_one_mem_trigger(trans, i, i->flags|BTREE_TRIGGER_GC);
+			int ret = run_one_mem_trigger(trans, i, i->flags|BTREE_TRIGGER_gc);
 			if (ret)
 				return ret;
 		}
@@ -609,6 +630,9 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
 	unsigned u64s = 0;
 	int ret;
 
+	bch2_trans_verify_not_unlocked(trans);
+	bch2_trans_verify_not_in_restart(trans);
+
 	if (race_fault()) {
 		trace_and_count(c, trans_restart_fault_inject, trans, trace_ip);
 		return btree_trans_restart_nounlock(trans, BCH_ERR_transaction_restart_fault_inject);
@@ -686,7 +710,7 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
 
 	trans_for_each_update(trans, i)
 		if (BTREE_NODE_TYPE_HAS_ATOMIC_TRIGGERS & (1U << i->bkey_type)) {
-			ret = run_one_mem_trigger(trans, i, BTREE_TRIGGER_ATOMIC|i->flags);
+			ret = run_one_mem_trigger(trans, i, BTREE_TRIGGER_atomic|i->flags);
 			if (ret)
 				goto fatal_err;
 		}
@@ -705,7 +729,7 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
 			if (i->key_cache_already_flushed)
 				continue;
 
-			if (i->flags & BTREE_UPDATE_NOJOURNAL)
+			if (i->flags & BTREE_UPDATE_nojournal)
 				continue;
 
 			verify_update_old_key(trans, i);
@@ -773,9 +797,8 @@ static noinline int bch2_trans_commit_bkey_invalid(struct btree_trans *trans,
 	struct bch_fs *c = trans->c;
 
 	printbuf_reset(err);
-	prt_printf(err, "invalid bkey on insert from %s -> %ps",
+	prt_printf(err, "invalid bkey on insert from %s -> %ps\n",
 		   trans->fn, (void *) i->ip_allocated);
-	prt_newline(err);
 	printbuf_indent_add(err, 2);
 
 	bch2_bkey_val_to_text(err, c, bkey_i_to_s_c(i->k));
@@ -796,8 +819,7 @@ static noinline int bch2_trans_commit_journal_entry_invalid(struct btree_trans *
 	struct bch_fs *c = trans->c;
 	struct printbuf buf = PRINTBUF;
 
-	prt_printf(&buf, "invalid bkey on insert from %s", trans->fn);
-	prt_newline(&buf);
+	prt_printf(&buf, "invalid bkey on insert from %s\n", trans->fn);
 	printbuf_indent_add(&buf, 2);
 
 	bch2_journal_entry_to_text(&buf, c, i);
@@ -988,6 +1010,9 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
 	struct bch_fs *c = trans->c;
 	int ret = 0;
 
+	bch2_trans_verify_not_unlocked(trans);
+	bch2_trans_verify_not_in_restart(trans);
+
 	if (!trans->nr_updates &&
 	    !trans->journal_entries_u64s)
 		goto out_reset;
@@ -1065,7 +1090,7 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
 		if (i->key_cache_already_flushed)
 			continue;
 
-		if (i->flags & BTREE_UPDATE_NOJOURNAL)
+		if (i->flags & BTREE_UPDATE_nojournal)
 			continue;
 
 		/* we're going to journal the key being updated: */
@@ -1086,6 +1111,7 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
 	}
 retry:
 	errored_at = NULL;
+	bch2_trans_verify_not_unlocked(trans);
 	bch2_trans_verify_not_in_restart(trans);
 	if (likely(!(flags & BCH_TRANS_COMMIT_no_journal_res)))
 		memset(&trans->journal_res, 0, sizeof(trans->journal_res));
diff --git a/fs/bcachefs/btree_types.h b/fs/bcachefs/btree_types.h
index c69b233c41bb3d..4ff5213219a50e 100644
--- a/fs/bcachefs/btree_types.h
+++ b/fs/bcachefs/btree_types.h
@@ -187,36 +187,89 @@ struct btree_node_iter {
 	} data[MAX_BSETS];
 };
 
+#define BTREE_ITER_FLAGS()			\
+	x(slots)				\
+	x(intent)				\
+	x(prefetch)				\
+	x(is_extents)				\
+	x(not_extents)				\
+	x(cached)				\
+	x(with_key_cache)			\
+	x(with_updates)				\
+	x(with_journal)				\
+	x(snapshot_field)			\
+	x(all_snapshots)			\
+	x(filter_snapshots)			\
+	x(nopreserve)				\
+	x(cached_nofill)			\
+	x(key_cache_fill)			\
+
+#define STR_HASH_FLAGS()			\
+	x(must_create)				\
+	x(must_replace)
+
+#define BTREE_UPDATE_FLAGS()			\
+	x(internal_snapshot_node)		\
+	x(nojournal)				\
+	x(key_cache_reclaim)
+
+
 /*
- * Iterate over all possible positions, synthesizing deleted keys for holes:
- */
-static const __maybe_unused u16 BTREE_ITER_SLOTS		= 1 << 0;
-/*
- * Indicates that intent locks should be taken on leaf nodes, because we expect
- * to be doing updates:
- */
-static const __maybe_unused u16 BTREE_ITER_INTENT		= 1 << 1;
-/*
- * Causes the btree iterator code to prefetch additional btree nodes from disk:
- */
-static const __maybe_unused u16 BTREE_ITER_PREFETCH		= 1 << 2;
-/*
- * Used in bch2_btree_iter_traverse(), to indicate whether we're searching for
- * @pos or the first key strictly greater than @pos
+ * BTREE_TRIGGER_norun - don't run triggers at all
+ *
+ * BTREE_TRIGGER_transactional - we're running transactional triggers as part of
+ * a transaction commit: triggers may generate new updates
+ *
+ * BTREE_TRIGGER_atomic - we're running atomic triggers during a transaction
+ * commit: we have our journal reservation, we're holding btree node write
+ * locks, and we know the transaction is going to commit (returning an error
+ * here is a fatal error, causing us to go emergency read-only)
+ *
+ * BTREE_TRIGGER_gc - we're in gc/fsck: running triggers to recalculate e.g. disk usage
+ *
+ * BTREE_TRIGGER_insert - @new is entering the btree
+ * BTREE_TRIGGER_overwrite - @old is leaving the btree
+ *
+ * BTREE_TRIGGER_bucket_invalidate - signal from bucket invalidate path to alloc
+ * trigger
  */
-static const __maybe_unused u16 BTREE_ITER_IS_EXTENTS		= 1 << 3;
-static const __maybe_unused u16 BTREE_ITER_NOT_EXTENTS		= 1 << 4;
-static const __maybe_unused u16 BTREE_ITER_CACHED		= 1 << 5;
-static const __maybe_unused u16 BTREE_ITER_WITH_KEY_CACHE	= 1 << 6;
-static const __maybe_unused u16 BTREE_ITER_WITH_UPDATES		= 1 << 7;
-static const __maybe_unused u16 BTREE_ITER_WITH_JOURNAL		= 1 << 8;
-static const __maybe_unused u16 __BTREE_ITER_ALL_SNAPSHOTS	= 1 << 9;
-static const __maybe_unused u16 BTREE_ITER_ALL_SNAPSHOTS	= 1 << 10;
-static const __maybe_unused u16 BTREE_ITER_FILTER_SNAPSHOTS	= 1 << 11;
-static const __maybe_unused u16 BTREE_ITER_NOPRESERVE		= 1 << 12;
-static const __maybe_unused u16 BTREE_ITER_CACHED_NOFILL	= 1 << 13;
-static const __maybe_unused u16 BTREE_ITER_KEY_CACHE_FILL	= 1 << 14;
-#define __BTREE_ITER_FLAGS_END					       15
+#define BTREE_TRIGGER_FLAGS()			\
+	x(norun)				\
+	x(transactional)			\
+	x(atomic)				\
+	x(check_repair)				\
+	x(gc)					\
+	x(insert)				\
+	x(overwrite)				\
+	x(is_root)				\
+	x(bucket_invalidate)
+
+enum {
+#define x(n) BTREE_ITER_FLAG_BIT_##n,
+	BTREE_ITER_FLAGS()
+	STR_HASH_FLAGS()
+	BTREE_UPDATE_FLAGS()
+	BTREE_TRIGGER_FLAGS()
+#undef x
+};
+
+/* iter flags must fit in a u16: */
+//BUILD_BUG_ON(BTREE_ITER_FLAG_BIT_key_cache_fill > 15);
+
+enum btree_iter_update_trigger_flags {
+#define x(n) BTREE_ITER_##n	= 1U << BTREE_ITER_FLAG_BIT_##n,
+	BTREE_ITER_FLAGS()
+#undef x
+#define x(n) STR_HASH_##n	= 1U << BTREE_ITER_FLAG_BIT_##n,
+	STR_HASH_FLAGS()
+#undef x
+#define x(n) BTREE_UPDATE_##n	= 1U << BTREE_ITER_FLAG_BIT_##n,
+	BTREE_UPDATE_FLAGS()
+#undef x
+#define x(n) BTREE_TRIGGER_##n	= 1U << BTREE_ITER_FLAG_BIT_##n,
+	BTREE_TRIGGER_FLAGS()
+#undef x
+};
 
 enum btree_path_uptodate {
 	BTREE_ITER_UPTODATE		= 0,
@@ -307,7 +360,7 @@ struct btree_iter {
 	 */
 	struct bkey		k;
 
-	/* BTREE_ITER_WITH_JOURNAL: */
+	/* BTREE_ITER_with_journal: */
 	size_t			journal_idx;
 #ifdef TRACK_PATH_ALLOCATED
 	unsigned long		ip_allocated;
@@ -418,6 +471,8 @@ struct btree_trans {
 	u8			lock_must_abort;
 	bool			lock_may_not_fail:1;
 	bool			srcu_held:1;
+	bool			locked:1;
+	bool			write_locked:1;
 	bool			used_mempool:1;
 	bool			in_traverse_all:1;
 	bool			paths_sorted:1;
@@ -425,13 +480,13 @@ struct btree_trans {
 	bool			journal_transaction_names:1;
 	bool			journal_replay_not_finished:1;
 	bool			notrace_relock_fail:1;
-	bool			write_locked:1;
 	enum bch_errcode	restarted:16;
 	u32			restart_count;
 
 	u64			last_begin_time;
 	unsigned long		last_begin_ip;
 	unsigned long		last_restarted_ip;
+	unsigned long		last_unlock_ip;
 	unsigned long		srcu_lock_time;
 
 	const char		*fn;
diff --git a/fs/bcachefs/btree_update.c b/fs/bcachefs/btree_update.c
index 8e47e260eba59b..f3c645a43dcbae 100644
--- a/fs/bcachefs/btree_update.c
+++ b/fs/bcachefs/btree_update.c
@@ -25,14 +25,14 @@ static inline int btree_insert_entry_cmp(const struct btree_insert_entry *l,
 
 static int __must_check
 bch2_trans_update_by_path(struct btree_trans *, btree_path_idx_t,
-			  struct bkey_i *, enum btree_update_flags,
+			  struct bkey_i *, enum btree_iter_update_trigger_flags,
 			  unsigned long ip);
 
 static noinline int extent_front_merge(struct btree_trans *trans,
 				       struct btree_iter *iter,
 				       struct bkey_s_c k,
 				       struct bkey_i **insert,
-				       enum btree_update_flags flags)
+				       enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_fs *c = trans->c;
 	struct bkey_i *update;
@@ -104,8 +104,8 @@ static int need_whiteout_for_snapshot(struct btree_trans *trans,
 	pos.snapshot++;
 
 	for_each_btree_key_norestart(trans, iter, btree_id, pos,
-			   BTREE_ITER_ALL_SNAPSHOTS|
-			   BTREE_ITER_NOPRESERVE, k, ret) {
+			   BTREE_ITER_all_snapshots|
+			   BTREE_ITER_nopreserve, k, ret) {
 		if (!bkey_eq(k.k->p, pos))
 			break;
 
@@ -138,8 +138,8 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
 	darray_init(&s);
 
 	bch2_trans_iter_init(trans, &old_iter, id, old_pos,
-			     BTREE_ITER_NOT_EXTENTS|
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_not_extents|
+			     BTREE_ITER_all_snapshots);
 	while ((old_k = bch2_btree_iter_prev(&old_iter)).k &&
 	       !(ret = bkey_err(old_k)) &&
 	       bkey_eq(old_pos, old_k.k->p)) {
@@ -151,8 +151,8 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
 			continue;
 
 		new_k = bch2_bkey_get_iter(trans, &new_iter, id, whiteout_pos,
-					   BTREE_ITER_NOT_EXTENTS|
-					   BTREE_ITER_INTENT);
+					   BTREE_ITER_not_extents|
+					   BTREE_ITER_intent);
 		ret = bkey_err(new_k);
 		if (ret)
 			break;
@@ -168,7 +168,7 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
 			update->k.type		= KEY_TYPE_whiteout;
 
 			ret = bch2_trans_update(trans, &new_iter, update,
-						BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+						BTREE_UPDATE_internal_snapshot_node);
 		}
 		bch2_trans_iter_exit(trans, &new_iter);
 
@@ -185,7 +185,7 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
 
 int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 				       struct btree_iter *iter,
-				       enum btree_update_flags flags,
+				       enum btree_iter_update_trigger_flags flags,
 				       struct bkey_s_c old,
 				       struct bkey_s_c new)
 {
@@ -218,7 +218,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 		ret =   bch2_insert_snapshot_whiteouts(trans, btree_id,
 					old.k->p, update->k.p) ?:
 			bch2_btree_insert_nonextent(trans, btree_id, update,
-					BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
+					BTREE_UPDATE_internal_snapshot_node|flags);
 		if (ret)
 			return ret;
 	}
@@ -235,7 +235,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 		ret =   bch2_insert_snapshot_whiteouts(trans, btree_id,
 					old.k->p, update->k.p) ?:
 			bch2_btree_insert_nonextent(trans, btree_id, update,
-					  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
+					  BTREE_UPDATE_internal_snapshot_node|flags);
 		if (ret)
 			return ret;
 	}
@@ -260,7 +260,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 		}
 
 		ret = bch2_btree_insert_nonextent(trans, btree_id, update,
-					  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
+					  BTREE_UPDATE_internal_snapshot_node|flags);
 		if (ret)
 			return ret;
 	}
@@ -273,7 +273,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 		bch2_cut_front(new.k->p, update);
 
 		ret = bch2_trans_update_by_path(trans, iter->path, update,
-					  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
+					  BTREE_UPDATE_internal_snapshot_node|
 					  flags, _RET_IP_);
 		if (ret)
 			return ret;
@@ -285,7 +285,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 static int bch2_trans_update_extent(struct btree_trans *trans,
 				    struct btree_iter *orig_iter,
 				    struct bkey_i *insert,
-				    enum btree_update_flags flags)
+				    enum btree_iter_update_trigger_flags flags)
 {
 	struct btree_iter iter;
 	struct bkey_s_c k;
@@ -293,9 +293,9 @@ static int bch2_trans_update_extent(struct btree_trans *trans,
 	int ret = 0;
 
 	bch2_trans_iter_init(trans, &iter, btree_id, bkey_start_pos(&insert->k),
-			     BTREE_ITER_INTENT|
-			     BTREE_ITER_WITH_UPDATES|
-			     BTREE_ITER_NOT_EXTENTS);
+			     BTREE_ITER_intent|
+			     BTREE_ITER_with_updates|
+			     BTREE_ITER_not_extents);
 	k = bch2_btree_iter_peek_upto(&iter, POS(insert->k.p.inode, U64_MAX));
 	if ((ret = bkey_err(k)))
 		goto err;
@@ -346,7 +346,7 @@ err:
 
 static noinline int flush_new_cached_update(struct btree_trans *trans,
 					    struct btree_insert_entry *i,
-					    enum btree_update_flags flags,
+					    enum btree_iter_update_trigger_flags flags,
 					    unsigned long ip)
 {
 	struct bkey k;
@@ -354,7 +354,7 @@ static noinline int flush_new_cached_update(struct btree_trans *trans,
 
 	btree_path_idx_t path_idx =
 		bch2_path_get(trans, i->btree_id, i->old_k.p, 1, 0,
-			      BTREE_ITER_INTENT, _THIS_IP_);
+			      BTREE_ITER_intent, _THIS_IP_);
 	ret = bch2_btree_path_traverse(trans, path_idx, 0);
 	if (ret)
 		goto out;
@@ -372,7 +372,7 @@ static noinline int flush_new_cached_update(struct btree_trans *trans,
 		goto out;
 
 	i->key_cache_already_flushed = true;
-	i->flags |= BTREE_TRIGGER_NORUN;
+	i->flags |= BTREE_TRIGGER_norun;
 
 	btree_path_set_should_be_locked(btree_path);
 	ret = bch2_trans_update_by_path(trans, path_idx, i->k, flags, ip);
@@ -383,7 +383,7 @@ out:
 
 static int __must_check
 bch2_trans_update_by_path(struct btree_trans *trans, btree_path_idx_t path_idx,
-			  struct bkey_i *k, enum btree_update_flags flags,
+			  struct bkey_i *k, enum btree_iter_update_trigger_flags flags,
 			  unsigned long ip)
 {
 	struct bch_fs *c = trans->c;
@@ -479,15 +479,15 @@ static noinline int bch2_trans_update_get_key_cache(struct btree_trans *trans,
 		if (!iter->key_cache_path)
 			iter->key_cache_path =
 				bch2_path_get(trans, path->btree_id, path->pos, 1, 0,
-					      BTREE_ITER_INTENT|
-					      BTREE_ITER_CACHED, _THIS_IP_);
+					      BTREE_ITER_intent|
+					      BTREE_ITER_cached, _THIS_IP_);
 
 		iter->key_cache_path =
 			bch2_btree_path_set_pos(trans, iter->key_cache_path, path->pos,
-						iter->flags & BTREE_ITER_INTENT,
+						iter->flags & BTREE_ITER_intent,
 						_THIS_IP_);
 
-		ret = bch2_btree_path_traverse(trans, iter->key_cache_path, BTREE_ITER_CACHED);
+		ret = bch2_btree_path_traverse(trans, iter->key_cache_path, BTREE_ITER_cached);
 		if (unlikely(ret))
 			return ret;
 
@@ -505,17 +505,17 @@ static noinline int bch2_trans_update_get_key_cache(struct btree_trans *trans,
 }
 
 int __must_check bch2_trans_update(struct btree_trans *trans, struct btree_iter *iter,
-				   struct bkey_i *k, enum btree_update_flags flags)
+				   struct bkey_i *k, enum btree_iter_update_trigger_flags flags)
 {
 	btree_path_idx_t path_idx = iter->update_path ?: iter->path;
 	int ret;
 
-	if (iter->flags & BTREE_ITER_IS_EXTENTS)
+	if (iter->flags & BTREE_ITER_is_extents)
 		return bch2_trans_update_extent(trans, iter, k, flags);
 
 	if (bkey_deleted(&k->k) &&
-	    !(flags & BTREE_UPDATE_KEY_CACHE_RECLAIM) &&
-	    (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)) {
+	    !(flags & BTREE_UPDATE_key_cache_reclaim) &&
+	    (iter->flags & BTREE_ITER_filter_snapshots)) {
 		ret = need_whiteout_for_snapshot(trans, iter->btree_id, k->k.p);
 		if (unlikely(ret < 0))
 			return ret;
@@ -528,7 +528,7 @@ int __must_check bch2_trans_update(struct btree_trans *trans, struct btree_iter
 	 * Ensure that updates to cached btrees go to the key cache:
 	 */
 	struct btree_path *path = trans->paths + path_idx;
-	if (!(flags & BTREE_UPDATE_KEY_CACHE_RECLAIM) &&
+	if (!(flags & BTREE_UPDATE_key_cache_reclaim) &&
 	    !path->cached &&
 	    !path->level &&
 	    btree_id_cached(trans->c, path->btree_id)) {
@@ -587,7 +587,7 @@ int bch2_bkey_get_empty_slot(struct btree_trans *trans, struct btree_iter *iter,
 	struct bkey_s_c k;
 	int ret = 0;
 
-	bch2_trans_iter_init(trans, iter, btree, POS_MAX, BTREE_ITER_INTENT);
+	bch2_trans_iter_init(trans, iter, btree, POS_MAX, BTREE_ITER_intent);
 	k = bch2_btree_iter_prev(iter);
 	ret = bkey_err(k);
 	if (ret)
@@ -621,15 +621,15 @@ void bch2_trans_commit_hook(struct btree_trans *trans,
 
 int bch2_btree_insert_nonextent(struct btree_trans *trans,
 				enum btree_id btree, struct bkey_i *k,
-				enum btree_update_flags flags)
+				enum btree_iter_update_trigger_flags flags)
 {
 	struct btree_iter iter;
 	int ret;
 
 	bch2_trans_iter_init(trans, &iter, btree, k->k.p,
-			     BTREE_ITER_CACHED|
-			     BTREE_ITER_NOT_EXTENTS|
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_cached|
+			     BTREE_ITER_not_extents|
+			     BTREE_ITER_intent);
 	ret   = bch2_btree_iter_traverse(&iter) ?:
 		bch2_trans_update(trans, &iter, k, flags);
 	bch2_trans_iter_exit(trans, &iter);
@@ -637,16 +637,13 @@ int bch2_btree_insert_nonextent(struct btree_trans *trans,
 }
 
 int bch2_btree_insert_trans(struct btree_trans *trans, enum btree_id id,
-			    struct bkey_i *k, enum btree_update_flags flags)
+			    struct bkey_i *k, enum btree_iter_update_trigger_flags flags)
 {
 	struct btree_iter iter;
-	int ret;
-
 	bch2_trans_iter_init(trans, &iter, id, bkey_start_pos(&k->k),
-			     BTREE_ITER_CACHED|
-			     BTREE_ITER_INTENT);
-	ret   = bch2_btree_iter_traverse(&iter) ?:
-		bch2_trans_update(trans, &iter, k, flags);
+			     BTREE_ITER_intent|flags);
+	int ret = bch2_btree_iter_traverse(&iter) ?:
+		  bch2_trans_update(trans, &iter, k, flags);
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }
@@ -698,8 +695,8 @@ int bch2_btree_delete(struct btree_trans *trans,
 	int ret;
 
 	bch2_trans_iter_init(trans, &iter, btree, pos,
-			     BTREE_ITER_CACHED|
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_cached|
+			     BTREE_ITER_intent);
 	ret   = bch2_btree_iter_traverse(&iter) ?:
 		bch2_btree_delete_at(trans, &iter, update_flags);
 	bch2_trans_iter_exit(trans, &iter);
@@ -717,7 +714,7 @@ int bch2_btree_delete_range_trans(struct btree_trans *trans, enum btree_id id,
 	struct bkey_s_c k;
 	int ret = 0;
 
-	bch2_trans_iter_init(trans, &iter, id, start, BTREE_ITER_INTENT);
+	bch2_trans_iter_init(trans, &iter, id, start, BTREE_ITER_intent);
 	while ((k = bch2_btree_iter_peek_upto(&iter, end)).k) {
 		struct disk_reservation disk_res =
 			bch2_disk_reservation_init(trans->c, 0);
@@ -745,7 +742,7 @@ int bch2_btree_delete_range_trans(struct btree_trans *trans, enum btree_id id,
 		 */
 		delete.k.p = iter.pos;
 
-		if (iter.flags & BTREE_ITER_IS_EXTENTS)
+		if (iter.flags & BTREE_ITER_is_extents)
 			bch2_key_resize(&delete.k,
 					bpos_min(end, k.k->p).offset -
 					iter.pos.offset);
@@ -804,7 +801,7 @@ int bch2_btree_bit_mod(struct btree_trans *trans, enum btree_id btree,
 	k->k.p = pos;
 
 	struct btree_iter iter;
-	bch2_trans_iter_init(trans, &iter, btree, pos, BTREE_ITER_INTENT);
+	bch2_trans_iter_init(trans, &iter, btree, pos, BTREE_ITER_intent);
 
 	ret   = bch2_btree_iter_traverse(&iter) ?:
 		bch2_trans_update(trans, &iter, k, 0);
@@ -852,7 +849,7 @@ __bch2_fs_log_msg(struct bch_fs *c, unsigned commit_flags, const char *fmt,
 	if (ret)
 		goto err;
 
-	if (!test_bit(JOURNAL_STARTED, &c->journal.flags)) {
+	if (!test_bit(JOURNAL_running, &c->journal.flags)) {
 		ret = darray_make_room(&c->journal.early_journal_entries, jset_u64s(u64s));
 		if (ret)
 			goto err;
diff --git a/fs/bcachefs/btree_update.h b/fs/bcachefs/btree_update.h
index cc7c53e83f89dd..b4894e4d5447fd 100644
--- a/fs/bcachefs/btree_update.h
+++ b/fs/bcachefs/btree_update.h
@@ -44,16 +44,18 @@ enum bch_trans_commit_flags {
 #undef x
 };
 
+void bch2_trans_commit_flags_to_text(struct printbuf *, enum bch_trans_commit_flags);
+
 int bch2_btree_delete_extent_at(struct btree_trans *, struct btree_iter *,
 				unsigned, unsigned);
 int bch2_btree_delete_at(struct btree_trans *, struct btree_iter *, unsigned);
 int bch2_btree_delete(struct btree_trans *, enum btree_id, struct bpos, unsigned);
 
 int bch2_btree_insert_nonextent(struct btree_trans *, enum btree_id,
-				struct bkey_i *, enum btree_update_flags);
+				struct bkey_i *, enum btree_iter_update_trigger_flags);
 
 int bch2_btree_insert_trans(struct btree_trans *, enum btree_id, struct bkey_i *,
-			enum btree_update_flags);
+			enum btree_iter_update_trigger_flags);
 int bch2_btree_insert(struct bch_fs *, enum btree_id, struct bkey_i *,
 		     struct disk_reservation *, int flags);
 
@@ -94,14 +96,14 @@ static inline int bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
 }
 
 int bch2_trans_update_extent_overwrite(struct btree_trans *, struct btree_iter *,
-				       enum btree_update_flags,
+				       enum btree_iter_update_trigger_flags,
 				       struct bkey_s_c, struct bkey_s_c);
 
 int bch2_bkey_get_empty_slot(struct btree_trans *, struct btree_iter *,
 			     enum btree_id, struct bpos);
 
 int __must_check bch2_trans_update(struct btree_trans *, struct btree_iter *,
-				   struct bkey_i *, enum btree_update_flags);
+				   struct bkey_i *, enum btree_iter_update_trigger_flags);
 
 struct jset_entry *__bch2_trans_jset_entry_alloc(struct btree_trans *, unsigned);
 
@@ -276,7 +278,7 @@ static inline struct bkey_i *__bch2_bkey_get_mut_noupdate(struct btree_trans *tr
 					 unsigned flags, unsigned type, unsigned min_bytes)
 {
 	struct bkey_s_c k = __bch2_bkey_get_iter(trans, iter,
-				btree_id, pos, flags|BTREE_ITER_INTENT, type);
+				btree_id, pos, flags|BTREE_ITER_intent, type);
 	struct bkey_i *ret = IS_ERR(k.k)
 		? ERR_CAST(k.k)
 		: __bch2_bkey_make_mut_noupdate(trans, k, 0, min_bytes);
@@ -299,7 +301,7 @@ static inline struct bkey_i *__bch2_bkey_get_mut(struct btree_trans *trans,
 					 unsigned flags, unsigned type, unsigned min_bytes)
 {
 	struct bkey_i *mut = __bch2_bkey_get_mut_noupdate(trans, iter,
-				btree_id, pos, flags|BTREE_ITER_INTENT, type, min_bytes);
+				btree_id, pos, flags|BTREE_ITER_intent, type, min_bytes);
 	int ret;
 
 	if (IS_ERR(mut))
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index b4efd8cc4d1a2b..2e8b092a182064 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -38,22 +38,6 @@ static int bch2_btree_insert_node(struct btree_update *, struct btree_trans *,
 				  btree_path_idx_t, struct btree *, struct keylist *);
 static void bch2_btree_update_add_new_node(struct btree_update *, struct btree *);
 
-static btree_path_idx_t get_unlocked_mut_path(struct btree_trans *trans,
-					      enum btree_id btree_id,
-					      unsigned level,
-					      struct bpos pos)
-{
-	btree_path_idx_t path_idx = bch2_path_get(trans, btree_id, pos, level + 1, level,
-			     BTREE_ITER_NOPRESERVE|
-			     BTREE_ITER_INTENT, _RET_IP_);
-	path_idx = bch2_btree_path_make_mut(trans, path_idx, true, _RET_IP_);
-
-	struct btree_path *path = trans->paths + path_idx;
-	bch2_btree_path_downgrade(trans, path);
-	__bch2_btree_path_unlock(trans, path);
-	return path_idx;
-}
-
 /*
  * Verify that child nodes correctly span parent node's range:
  */
@@ -73,6 +57,24 @@ int bch2_btree_node_check_topology(struct btree_trans *trans, struct btree *b)
 	       !bpos_eq(bkey_i_to_btree_ptr_v2(&b->key)->v.min_key,
 			b->data->min_key));
 
+	if (b == btree_node_root(c, b)) {
+		if (!bpos_eq(b->data->min_key, POS_MIN)) {
+			printbuf_reset(&buf);
+			bch2_bpos_to_text(&buf, b->data->min_key);
+			need_fsck_err(c, btree_root_bad_min_key,
+				      "btree root with incorrect min_key: %s", buf.buf);
+			goto topology_repair;
+		}
+
+		if (!bpos_eq(b->data->max_key, SPOS_MAX)) {
+			printbuf_reset(&buf);
+			bch2_bpos_to_text(&buf, b->data->max_key);
+			need_fsck_err(c, btree_root_bad_max_key,
+				      "btree root with incorrect max_key: %s", buf.buf);
+			goto topology_repair;
+		}
+	}
+
 	if (!b->c.level)
 		return 0;
 
@@ -646,7 +648,7 @@ static int btree_update_nodes_written_trans(struct btree_trans *trans,
 		unsigned level = bkey_i_to_btree_ptr_v2(k)->v.mem_ptr;
 
 		ret = bch2_key_trigger_old(trans, as->btree_id, level, bkey_i_to_s_c(k),
-					   BTREE_TRIGGER_TRANSACTIONAL);
+					   BTREE_TRIGGER_transactional);
 		if (ret)
 			return ret;
 	}
@@ -655,7 +657,7 @@ static int btree_update_nodes_written_trans(struct btree_trans *trans,
 		unsigned level = bkey_i_to_btree_ptr_v2(k)->v.mem_ptr;
 
 		ret = bch2_key_trigger_new(trans, as->btree_id, level, bkey_i_to_s(k),
-					   BTREE_TRIGGER_TRANSACTIONAL);
+					   BTREE_TRIGGER_transactional);
 		if (ret)
 			return ret;
 	}
@@ -735,9 +737,6 @@ err:
 	 */
 	b = READ_ONCE(as->b);
 	if (b) {
-		btree_path_idx_t path_idx = get_unlocked_mut_path(trans,
-						as->btree_id, b->c.level, b->key.k.p);
-		struct btree_path *path = trans->paths + path_idx;
 		/*
 		 * @b is the node we did the final insert into:
 		 *
@@ -755,12 +754,16 @@ err:
 		 * btree_node_lock_nopath() (the use of which is always suspect,
 		 * we need to work on removing this in the future)
 		 *
-		 * It should be, but get_unlocked_mut_path() -> bch2_path_get()
+		 * It should be, but bch2_path_get_unlocked_mut() -> bch2_path_get()
 		 * calls bch2_path_upgrade(), before we call path_make_mut(), so
 		 * we may rarely end up with a locked path besides the one we
 		 * have here:
 		 */
 		bch2_trans_unlock(trans);
+		bch2_trans_begin(trans);
+		btree_path_idx_t path_idx = bch2_path_get_unlocked_mut(trans,
+						as->btree_id, b->c.level, b->key.k.p);
+		struct btree_path *path = trans->paths + path_idx;
 		btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_intent);
 		mark_btree_node_locked(trans, path, b->c.level, BTREE_NODE_INTENT_LOCKED);
 		path->l[b->c.level].lock_seq = six_lock_seq(&b->c.lock);
@@ -1154,13 +1157,12 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
 	flags |= watermark;
 
 	if (watermark < BCH_WATERMARK_reclaim &&
-	    test_bit(JOURNAL_SPACE_LOW, &c->journal.flags)) {
+	    test_bit(JOURNAL_space_low, &c->journal.flags)) {
 		if (flags & BCH_TRANS_COMMIT_journal_reclaim)
 			return ERR_PTR(-BCH_ERR_journal_reclaim_would_deadlock);
 
-		bch2_trans_unlock(trans);
-		wait_event(c->journal.wait, !test_bit(JOURNAL_SPACE_LOW, &c->journal.flags));
-		ret = bch2_trans_relock(trans);
+		ret = drop_locks_do(trans,
+			({ wait_event(c->journal.wait, !test_bit(JOURNAL_space_low, &c->journal.flags)); 0; }));
 		if (ret)
 			return ERR_PTR(ret);
 	}
@@ -1206,7 +1208,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
 	as->start_time		= start_time;
 	as->ip_started		= _RET_IP_;
 	as->mode		= BTREE_UPDATE_none;
-	as->watermark		= watermark;
+	as->flags		= flags;
 	as->took_gc_lock	= true;
 	as->btree_id		= path->btree_id;
 	as->update_level_start	= level_start;
@@ -1360,7 +1362,7 @@ static void bch2_insert_fixup_btree_ptr(struct btree_update *as,
 	BUG_ON(insert->k.type == KEY_TYPE_btree_ptr_v2 &&
 	       !btree_ptr_sectors_written(insert));
 
-	if (unlikely(!test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)))
+	if (unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)))
 		bch2_journal_key_overwritten(c, b->c.btree_id, b->c.level, insert->k.p);
 
 	if (bch2_bkey_invalid(c, bkey_i_to_s_c(insert),
@@ -1619,12 +1621,12 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
 		six_unlock_write(&n2->c.lock);
 		six_unlock_write(&n1->c.lock);
 
-		path1 = get_unlocked_mut_path(trans, as->btree_id, n1->c.level, n1->key.k.p);
+		path1 = bch2_path_get_unlocked_mut(trans, as->btree_id, n1->c.level, n1->key.k.p);
 		six_lock_increment(&n1->c.lock, SIX_LOCK_intent);
 		mark_btree_node_locked(trans, trans->paths + path1, n1->c.level, BTREE_NODE_INTENT_LOCKED);
 		bch2_btree_path_level_init(trans, trans->paths + path1, n1);
 
-		path2 = get_unlocked_mut_path(trans, as->btree_id, n2->c.level, n2->key.k.p);
+		path2 = bch2_path_get_unlocked_mut(trans, as->btree_id, n2->c.level, n2->key.k.p);
 		six_lock_increment(&n2->c.lock, SIX_LOCK_intent);
 		mark_btree_node_locked(trans, trans->paths + path2, n2->c.level, BTREE_NODE_INTENT_LOCKED);
 		bch2_btree_path_level_init(trans, trans->paths + path2, n2);
@@ -1669,7 +1671,7 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
 		bch2_btree_update_add_new_node(as, n1);
 		six_unlock_write(&n1->c.lock);
 
-		path1 = get_unlocked_mut_path(trans, as->btree_id, n1->c.level, n1->key.k.p);
+		path1 = bch2_path_get_unlocked_mut(trans, as->btree_id, n1->c.level, n1->key.k.p);
 		six_lock_increment(&n1->c.lock, SIX_LOCK_intent);
 		mark_btree_node_locked(trans, trans->paths + path1, n1->c.level, BTREE_NODE_INTENT_LOCKED);
 		bch2_btree_path_level_init(trans, trans->paths + path1, n1);
@@ -1947,6 +1949,8 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
 	u64 start_time = local_clock();
 	int ret = 0;
 
+	bch2_trans_verify_not_in_restart(trans);
+	bch2_trans_verify_not_unlocked(trans);
 	BUG_ON(!trans->paths[path].should_be_locked);
 	BUG_ON(!btree_node_locked(&trans->paths[path], level));
 
@@ -1979,7 +1983,7 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
 		: bpos_successor(b->data->max_key);
 
 	sib_path = bch2_path_get(trans, btree, sib_pos,
-				 U8_MAX, level, BTREE_ITER_INTENT, _THIS_IP_);
+				 U8_MAX, level, BTREE_ITER_intent, _THIS_IP_);
 	ret = bch2_btree_path_traverse(trans, sib_path, false);
 	if (ret)
 		goto err;
@@ -2072,7 +2076,7 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
 	bch2_btree_update_add_new_node(as, n);
 	six_unlock_write(&n->c.lock);
 
-	new_path = get_unlocked_mut_path(trans, btree, n->c.level, n->key.k.p);
+	new_path = bch2_path_get_unlocked_mut(trans, btree, n->c.level, n->key.k.p);
 	six_lock_increment(&n->c.lock, SIX_LOCK_intent);
 	mark_btree_node_locked(trans, trans->paths + new_path, n->c.level, BTREE_NODE_INTENT_LOCKED);
 	bch2_btree_path_level_init(trans, trans->paths + new_path, n);
@@ -2150,7 +2154,7 @@ int bch2_btree_node_rewrite(struct btree_trans *trans,
 	bch2_btree_update_add_new_node(as, n);
 	six_unlock_write(&n->c.lock);
 
-	new_path = get_unlocked_mut_path(trans, iter->btree_id, n->c.level, n->key.k.p);
+	new_path = bch2_path_get_unlocked_mut(trans, iter->btree_id, n->c.level, n->key.k.p);
 	six_lock_increment(&n->c.lock, SIX_LOCK_intent);
 	mark_btree_node_locked(trans, trans->paths + new_path, n->c.level, BTREE_NODE_INTENT_LOCKED);
 	bch2_btree_path_level_init(trans, trans->paths + new_path, n);
@@ -2333,10 +2337,10 @@ static int __bch2_btree_node_update_key(struct btree_trans *trans,
 	if (!skip_triggers) {
 		ret   = bch2_key_trigger_old(trans, b->c.btree_id, b->c.level + 1,
 					     bkey_i_to_s_c(&b->key),
-					     BTREE_TRIGGER_TRANSACTIONAL) ?:
+					     BTREE_TRIGGER_transactional) ?:
 			bch2_key_trigger_new(trans, b->c.btree_id, b->c.level + 1,
 					     bkey_i_to_s(new_key),
-					     BTREE_TRIGGER_TRANSACTIONAL);
+					     BTREE_TRIGGER_transactional);
 		if (ret)
 			return ret;
 	}
@@ -2353,7 +2357,7 @@ static int __bch2_btree_node_update_key(struct btree_trans *trans,
 		bch2_trans_copy_iter(&iter2, iter);
 
 		iter2.path = bch2_btree_path_make_mut(trans, iter2.path,
-				iter2.flags & BTREE_ITER_INTENT,
+				iter2.flags & BTREE_ITER_intent,
 				_THIS_IP_);
 
 		struct btree_path *path2 = btree_iter_path(trans, &iter2);
@@ -2365,7 +2369,7 @@ static int __bch2_btree_node_update_key(struct btree_trans *trans,
 		trans->paths_sorted = false;
 
 		ret   = bch2_btree_iter_traverse(&iter2) ?:
-			bch2_trans_update(trans, &iter2, new_key, BTREE_TRIGGER_NORUN);
+			bch2_trans_update(trans, &iter2, new_key, BTREE_TRIGGER_norun);
 		if (ret)
 			goto err;
 	} else {
@@ -2473,7 +2477,7 @@ int bch2_btree_node_update_key_get_iter(struct btree_trans *trans,
 
 	bch2_trans_node_iter_init(trans, &iter, b->c.btree_id, b->key.k.p,
 				  BTREE_MAX_DEPTH, b->c.level,
-				  BTREE_ITER_INTENT);
+				  BTREE_ITER_intent);
 	ret = bch2_btree_iter_traverse(&iter);
 	if (ret)
 		goto out;
@@ -2511,7 +2515,7 @@ void bch2_btree_set_root_for_read(struct bch_fs *c, struct btree *b)
 	bch2_btree_set_root_inmem(c, b);
 }
 
-static int __bch2_btree_root_alloc_fake(struct btree_trans *trans, enum btree_id id, unsigned level)
+int bch2_btree_root_alloc_fake_trans(struct btree_trans *trans, enum btree_id id, unsigned level)
 {
 	struct bch_fs *c = trans->c;
 	struct closure cl;
@@ -2559,17 +2563,18 @@ static int __bch2_btree_root_alloc_fake(struct btree_trans *trans, enum btree_id
 
 void bch2_btree_root_alloc_fake(struct bch_fs *c, enum btree_id id, unsigned level)
 {
-	bch2_trans_run(c, __bch2_btree_root_alloc_fake(trans, id, level));
+	bch2_trans_run(c, bch2_btree_root_alloc_fake_trans(trans, id, level));
 }
 
 static void bch2_btree_update_to_text(struct printbuf *out, struct btree_update *as)
 {
-	prt_printf(out, "%ps: btree=%s l=%u-%u watermark=%s mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
-		   (void *) as->ip_started,
+	prt_printf(out, "%ps: ", (void *) as->ip_started);
+	bch2_trans_commit_flags_to_text(out, as->flags);
+
+	prt_printf(out, " btree=%s l=%u-%u mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
 		   bch2_btree_id_str(as->btree_id),
 		   as->update_level_start,
 		   as->update_level_end,
-		   bch2_watermarks[as->watermark],
 		   bch2_btree_update_modes[as->mode],
 		   as->nodes_written,
 		   closure_nr_remaining(&as->cl),
diff --git a/fs/bcachefs/btree_update_interior.h b/fs/bcachefs/btree_update_interior.h
index c1a479ebaad121..b5b76ce01cfcb6 100644
--- a/fs/bcachefs/btree_update_interior.h
+++ b/fs/bcachefs/btree_update_interior.h
@@ -52,7 +52,7 @@ struct btree_update {
 	struct list_head		unwritten_list;
 
 	enum btree_update_mode		mode;
-	enum bch_watermark		watermark;
+	enum bch_trans_commit_flags	flags;
 	unsigned			nodes_written:1;
 	unsigned			took_gc_lock:1;
 
@@ -144,6 +144,9 @@ static inline int bch2_foreground_maybe_merge_sibling(struct btree_trans *trans,
 
 	EBUG_ON(!btree_node_locked(path, level));
 
+	if (bch2_btree_node_merging_disabled)
+		return 0;
+
 	b = path->l[level].b;
 	if (b->sib_u64s[sib] > trans->c->btree_foreground_merge_threshold)
 		return 0;
@@ -172,6 +175,8 @@ int bch2_btree_node_update_key_get_iter(struct btree_trans *, struct btree *,
 					struct bkey_i *, unsigned, bool);
 
 void bch2_btree_set_root_for_read(struct bch_fs *, struct btree *);
+
+int bch2_btree_root_alloc_fake_trans(struct btree_trans *, enum btree_id, unsigned);
 void bch2_btree_root_alloc_fake(struct bch_fs *, enum btree_id, unsigned);
 
 static inline unsigned btree_update_reserve_required(struct bch_fs *c,
diff --git a/fs/bcachefs/btree_write_buffer.c b/fs/bcachefs/btree_write_buffer.c
index 36a6f42aba5e6f..75c8a196b3f63e 100644
--- a/fs/bcachefs/btree_write_buffer.c
+++ b/fs/bcachefs/btree_write_buffer.c
@@ -122,7 +122,7 @@ static noinline int wb_flush_one_slowpath(struct btree_trans *trans,
 	trans->journal_res.seq = wb->journal_seq;
 
 	return bch2_trans_update(trans, iter, &wb->k,
-				 BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+				 BTREE_UPDATE_internal_snapshot_node) ?:
 		bch2_trans_commit(trans, NULL, NULL,
 				  BCH_TRANS_COMMIT_no_enospc|
 				  BCH_TRANS_COMMIT_no_check_rw|
@@ -191,13 +191,13 @@ btree_write_buffered_insert(struct btree_trans *trans,
 	int ret;
 
 	bch2_trans_iter_init(trans, &iter, wb->btree, bkey_start_pos(&wb->k.k),
-			     BTREE_ITER_CACHED|BTREE_ITER_INTENT);
+			     BTREE_ITER_cached|BTREE_ITER_intent);
 
 	trans->journal_res.seq = wb->journal_seq;
 
 	ret   = bch2_btree_iter_traverse(&iter) ?:
 		bch2_trans_update(trans, &iter, &wb->k,
-				  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				  BTREE_UPDATE_internal_snapshot_node);
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }
@@ -332,7 +332,7 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans)
 		if (!iter.path || iter.btree_id != k->btree) {
 			bch2_trans_iter_exit(trans, &iter);
 			bch2_trans_iter_init(trans, &iter, k->btree, k->k.k.p,
-					     BTREE_ITER_INTENT|BTREE_ITER_ALL_SNAPSHOTS);
+					     BTREE_ITER_intent|BTREE_ITER_all_snapshots);
 		}
 
 		bch2_btree_iter_set_pos(&iter, k->k.k.p);
diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index 82f179258867b7..46b9083bf0265d 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -274,25 +274,14 @@ void bch2_dev_usage_init(struct bch_dev *ca)
 
 void bch2_dev_usage_to_text(struct printbuf *out, struct bch_dev_usage *usage)
 {
-	prt_tab(out);
-	prt_str(out, "buckets");
-	prt_tab_rjust(out);
-	prt_str(out, "sectors");
-	prt_tab_rjust(out);
-	prt_str(out, "fragmented");
-	prt_tab_rjust(out);
-	prt_newline(out);
+	prt_printf(out, "\tbuckets\rsectors\rfragmented\r\n");
 
 	for (unsigned i = 0; i < BCH_DATA_NR; i++) {
 		bch2_prt_data_type(out, i);
-		prt_tab(out);
-		prt_u64(out, usage->d[i].buckets);
-		prt_tab_rjust(out);
-		prt_u64(out, usage->d[i].sectors);
-		prt_tab_rjust(out);
-		prt_u64(out, usage->d[i].fragmented);
-		prt_tab_rjust(out);
-		prt_newline(out);
+		prt_printf(out, "\t%llu\r%llu\r%llu\r\n",
+			usage->d[i].buckets,
+			usage->d[i].sectors,
+			usage->d[i].fragmented);
 	}
 }
 
@@ -329,26 +318,6 @@ void bch2_dev_usage_update(struct bch_fs *c, struct bch_dev *ca,
 	preempt_enable();
 }
 
-static inline struct bch_alloc_v4 bucket_m_to_alloc(struct bucket b)
-{
-	return (struct bch_alloc_v4) {
-		.gen		= b.gen,
-		.data_type	= b.data_type,
-		.dirty_sectors	= b.dirty_sectors,
-		.cached_sectors	= b.cached_sectors,
-		.stripe		= b.stripe,
-	};
-}
-
-void bch2_dev_usage_update_m(struct bch_fs *c, struct bch_dev *ca,
-			     struct bucket *old, struct bucket *new)
-{
-	struct bch_alloc_v4 old_a = bucket_m_to_alloc(*old);
-	struct bch_alloc_v4 new_a = bucket_m_to_alloc(*new);
-
-	bch2_dev_usage_update(c, ca, &old_a, &new_a, 0, true);
-}
-
 static inline int __update_replicas(struct bch_fs *c,
 				    struct bch_fs_usage *fs_usage,
 				    struct bch_replicas_entry_v1 *r,
@@ -496,78 +465,255 @@ int bch2_update_cached_sectors_list(struct btree_trans *trans, unsigned dev, s64
 	return bch2_update_replicas_list(trans, &r.e, sectors);
 }
 
-int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
-			      size_t b, enum bch_data_type data_type,
-			      unsigned sectors, struct gc_pos pos,
-			      unsigned flags)
+int bch2_check_fix_ptrs(struct btree_trans *trans,
+			enum btree_id btree, unsigned level, struct bkey_s_c k,
+			enum btree_iter_update_trigger_flags flags)
 {
-	struct bucket old, new, *g;
+	struct bch_fs *c = trans->c;
+	struct bkey_ptrs_c ptrs_c = bch2_bkey_ptrs_c(k);
+	const union bch_extent_entry *entry_c;
+	struct extent_ptr_decoded p = { 0 };
+	bool do_update = false;
+	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
-	BUG_ON(!(flags & BTREE_TRIGGER_GC));
-	BUG_ON(data_type != BCH_DATA_sb &&
-	       data_type != BCH_DATA_journal);
+	percpu_down_read(&c->mark_lock);
 
-	/*
-	 * Backup superblock might be past the end of our normal usable space:
-	 */
-	if (b >= ca->mi.nbuckets)
-		return 0;
+	bkey_for_each_ptr_decode(k.k, ptrs_c, p, entry_c) {
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, p.ptr.dev);
+		struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
+		enum bch_data_type data_type = bch2_bkey_ptr_data_type(k, p, entry_c);
+
+		if (fsck_err_on(!g->gen_valid,
+				c, ptr_to_missing_alloc_key,
+				"bucket %u:%zu data type %s ptr gen %u missing in alloc btree\n"
+				"while marking %s",
+				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr),
+				bch2_data_type_str(ptr_data_type(k.k, &p.ptr)),
+				p.ptr.gen,
+				(printbuf_reset(&buf),
+				 bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+			if (!p.ptr.cached) {
+				g->gen_valid		= true;
+				g->gen			= p.ptr.gen;
+			} else {
+				do_update = true;
+			}
+		}
 
-	percpu_down_read(&c->mark_lock);
-	g = gc_bucket(ca, b);
+		if (fsck_err_on(gen_cmp(p.ptr.gen, g->gen) > 0,
+				c, ptr_gen_newer_than_bucket_gen,
+				"bucket %u:%zu data type %s ptr gen in the future: %u > %u\n"
+				"while marking %s",
+				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr),
+				bch2_data_type_str(ptr_data_type(k.k, &p.ptr)),
+				p.ptr.gen, g->gen,
+				(printbuf_reset(&buf),
+				 bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+			if (!p.ptr.cached &&
+			    (g->data_type != BCH_DATA_btree ||
+			     data_type == BCH_DATA_btree)) {
+				g->gen_valid		= true;
+				g->gen			= p.ptr.gen;
+				g->data_type		= 0;
+				g->dirty_sectors	= 0;
+				g->cached_sectors	= 0;
+			} else {
+				do_update = true;
+			}
+		}
 
-	bucket_lock(g);
-	old = *g;
+		if (fsck_err_on(gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX,
+				c, ptr_gen_newer_than_bucket_gen,
+				"bucket %u:%zu gen %u data type %s: ptr gen %u too stale\n"
+				"while marking %s",
+				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), g->gen,
+				bch2_data_type_str(ptr_data_type(k.k, &p.ptr)),
+				p.ptr.gen,
+				(printbuf_reset(&buf),
+				 bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+			do_update = true;
+
+		if (fsck_err_on(!p.ptr.cached && gen_cmp(p.ptr.gen, g->gen) < 0,
+				c, stale_dirty_ptr,
+				"bucket %u:%zu data type %s stale dirty ptr: %u < %u\n"
+				"while marking %s",
+				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr),
+				bch2_data_type_str(ptr_data_type(k.k, &p.ptr)),
+				p.ptr.gen, g->gen,
+				(printbuf_reset(&buf),
+				 bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+			do_update = true;
+
+		if (data_type != BCH_DATA_btree && p.ptr.gen != g->gen)
+			continue;
+
+		if (fsck_err_on(bucket_data_type_mismatch(g->data_type, data_type),
+				c, ptr_bucket_data_type_mismatch,
+				"bucket %u:%zu gen %u different types of data in same bucket: %s, %s\n"
+				"while marking %s",
+				p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), g->gen,
+				bch2_data_type_str(g->data_type),
+				bch2_data_type_str(data_type),
+				(printbuf_reset(&buf),
+				 bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+			if (data_type == BCH_DATA_btree) {
+				g->gen_valid		= true;
+				g->gen			= p.ptr.gen;
+				g->data_type		= data_type;
+				g->dirty_sectors	= 0;
+				g->cached_sectors	= 0;
+			} else {
+				do_update = true;
+			}
+		}
 
-	if (bch2_fs_inconsistent_on(g->data_type &&
-			g->data_type != data_type, c,
-			"different types of data in same bucket: %s, %s",
-			bch2_data_type_str(g->data_type),
-			bch2_data_type_str(data_type))) {
-		ret = -EIO;
-		goto err;
+		if (p.has_ec) {
+			struct gc_stripe *m = genradix_ptr(&c->gc_stripes, p.ec.idx);
+
+			if (fsck_err_on(!m || !m->alive, c,
+					ptr_to_missing_stripe,
+					"pointer to nonexistent stripe %llu\n"
+					"while marking %s",
+					(u64) p.ec.idx,
+					(printbuf_reset(&buf),
+					 bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+				do_update = true;
+
+			if (fsck_err_on(m && m->alive && !bch2_ptr_matches_stripe_m(m, p), c,
+					ptr_to_incorrect_stripe,
+					"pointer does not match stripe %llu\n"
+					"while marking %s",
+					(u64) p.ec.idx,
+					(printbuf_reset(&buf),
+					 bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+				do_update = true;
+		}
 	}
 
-	if (bch2_fs_inconsistent_on((u64) g->dirty_sectors + sectors > ca->mi.bucket_size, c,
-			"bucket %u:%zu gen %u data type %s sector count overflow: %u + %u > bucket size",
-			ca->dev_idx, b, g->gen,
-			bch2_data_type_str(g->data_type ?: data_type),
-			g->dirty_sectors, sectors)) {
-		ret = -EIO;
-		goto err;
-	}
+	if (do_update) {
+		if (flags & BTREE_TRIGGER_is_root) {
+			bch_err(c, "cannot update btree roots yet");
+			ret = -EINVAL;
+			goto err;
+		}
 
-	g->data_type = data_type;
-	g->dirty_sectors += sectors;
-	new = *g;
+		struct bkey_i *new = bch2_bkey_make_mut_noupdate(trans, k);
+		ret = PTR_ERR_OR_ZERO(new);
+		if (ret)
+			goto err;
+
+		if (level) {
+			/*
+			 * We don't want to drop btree node pointers - if the
+			 * btree node isn't there anymore, the read path will
+			 * sort it out:
+			 */
+			struct bkey_ptrs ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
+			bkey_for_each_ptr(ptrs, ptr) {
+				struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
+				struct bucket *g = PTR_GC_BUCKET(ca, ptr);
+
+				ptr->gen = g->gen;
+			}
+		} else {
+			struct bkey_ptrs ptrs;
+			union bch_extent_entry *entry;
+restart_drop_ptrs:
+			ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
+			bkey_for_each_ptr_decode(bkey_i_to_s(new).k, ptrs, p, entry) {
+				struct bch_dev *ca = bch2_dev_bkey_exists(c, p.ptr.dev);
+				struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
+				enum bch_data_type data_type = bch2_bkey_ptr_data_type(bkey_i_to_s_c(new), p, entry);
+
+				if ((p.ptr.cached &&
+				     (!g->gen_valid || gen_cmp(p.ptr.gen, g->gen) > 0)) ||
+				    (!p.ptr.cached &&
+				     gen_cmp(p.ptr.gen, g->gen) < 0) ||
+				    gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX ||
+				    (g->data_type &&
+				     g->data_type != data_type)) {
+					bch2_bkey_drop_ptr(bkey_i_to_s(new), &entry->ptr);
+					goto restart_drop_ptrs;
+				}
+			}
+again:
+			ptrs = bch2_bkey_ptrs(bkey_i_to_s(new));
+			bkey_extent_entry_for_each(ptrs, entry) {
+				if (extent_entry_type(entry) == BCH_EXTENT_ENTRY_stripe_ptr) {
+					struct gc_stripe *m = genradix_ptr(&c->gc_stripes,
+									entry->stripe_ptr.idx);
+					union bch_extent_entry *next_ptr;
+
+					bkey_extent_entry_for_each_from(ptrs, next_ptr, entry)
+						if (extent_entry_type(next_ptr) == BCH_EXTENT_ENTRY_ptr)
+							goto found;
+					next_ptr = NULL;
+found:
+					if (!next_ptr) {
+						bch_err(c, "aieee, found stripe ptr with no data ptr");
+						continue;
+					}
+
+					if (!m || !m->alive ||
+					    !__bch2_ptr_matches_stripe(&m->ptrs[entry->stripe_ptr.block],
+								       &next_ptr->ptr,
+								       m->sectors)) {
+						bch2_bkey_extent_entry_drop(new, entry);
+						goto again;
+					}
+				}
+			}
+		}
+
+		if (0) {
+			printbuf_reset(&buf);
+			bch2_bkey_val_to_text(&buf, c, k);
+			bch_info(c, "updated %s", buf.buf);
+
+			printbuf_reset(&buf);
+			bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(new));
+			bch_info(c, "new key %s", buf.buf);
+		}
+
+		percpu_up_read(&c->mark_lock);
+		struct btree_iter iter;
+		bch2_trans_node_iter_init(trans, &iter, btree, new->k.p, 0, level,
+					  BTREE_ITER_intent|BTREE_ITER_all_snapshots);
+		ret =   bch2_btree_iter_traverse(&iter) ?:
+			bch2_trans_update(trans, &iter, new,
+					  BTREE_UPDATE_internal_snapshot_node|
+					  BTREE_TRIGGER_norun);
+		bch2_trans_iter_exit(trans, &iter);
+		percpu_down_read(&c->mark_lock);
+
+		if (ret)
+			goto err;
+
+		if (level)
+			bch2_btree_node_update_key_early(trans, btree, level - 1, k, new);
+	}
 err:
-	bucket_unlock(g);
-	if (!ret)
-		bch2_dev_usage_update_m(c, ca, &old, &new);
+fsck_err:
 	percpu_up_read(&c->mark_lock);
+	printbuf_exit(&buf);
 	return ret;
 }
 
-int bch2_check_bucket_ref(struct btree_trans *trans,
-			  struct bkey_s_c k,
-			  const struct bch_extent_ptr *ptr,
-			  s64 sectors, enum bch_data_type ptr_data_type,
-			  u8 b_gen, u8 bucket_data_type,
-			  u32 bucket_sectors)
+int bch2_bucket_ref_update(struct btree_trans *trans, struct bch_dev *ca,
+			   struct bkey_s_c k,
+			   const struct bch_extent_ptr *ptr,
+			   s64 sectors, enum bch_data_type ptr_data_type,
+			   u8 b_gen, u8 bucket_data_type,
+			   u32 *bucket_sectors)
 {
 	struct bch_fs *c = trans->c;
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
 	size_t bucket_nr = PTR_BUCKET_NR(ca, ptr);
 	struct printbuf buf = PRINTBUF;
+	bool inserting = sectors > 0;
 	int ret = 0;
 
-	if (bucket_data_type == BCH_DATA_cached)
-		bucket_data_type = BCH_DATA_user;
-
-	if ((bucket_data_type == BCH_DATA_stripe && ptr_data_type == BCH_DATA_user) ||
-	    (bucket_data_type == BCH_DATA_user   && ptr_data_type == BCH_DATA_stripe))
-		bucket_data_type = ptr_data_type = BCH_DATA_stripe;
+	BUG_ON(!sectors);
 
 	if (gen_after(ptr->gen, b_gen)) {
 		bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
@@ -578,8 +724,9 @@ int bch2_check_bucket_ref(struct btree_trans *trans,
 			bch2_data_type_str(bucket_data_type ?: ptr_data_type),
 			ptr->gen,
 			(bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-		ret = -EIO;
-		goto err;
+		if (inserting)
+			goto err;
+		goto out;
 	}
 
 	if (gen_cmp(b_gen, ptr->gen) > BUCKET_GC_GEN_MAX) {
@@ -592,11 +739,17 @@ int bch2_check_bucket_ref(struct btree_trans *trans,
 			ptr->gen,
 			(printbuf_reset(&buf),
 			 bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-		ret = -EIO;
-		goto err;
+		if (inserting)
+			goto err;
+		goto out;
 	}
 
-	if (b_gen != ptr->gen && !ptr->cached) {
+	if (b_gen != ptr->gen && ptr->cached) {
+		ret = 1;
+		goto out;
+	}
+
+	if (b_gen != ptr->gen) {
 		bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
 			      BCH_FSCK_ERR_stale_dirty_ptr,
 			"bucket %u:%zu gen %u (mem gen %u) data type %s: stale dirty ptr (gen %u)\n"
@@ -607,18 +760,12 @@ int bch2_check_bucket_ref(struct btree_trans *trans,
 			ptr->gen,
 			(printbuf_reset(&buf),
 			 bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-		ret = -EIO;
-		goto err;
-	}
-
-	if (b_gen != ptr->gen) {
-		ret = 1;
+		if (inserting)
+			goto err;
 		goto out;
 	}
 
-	if (!data_type_is_empty(bucket_data_type) &&
-	    ptr_data_type &&
-	    bucket_data_type != ptr_data_type) {
+	if (bucket_data_type_mismatch(bucket_data_type, ptr_data_type)) {
 		bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
 			      BCH_FSCK_ERR_ptr_bucket_data_type_mismatch,
 			"bucket %u:%zu gen %u different types of data in same bucket: %s, %s\n"
@@ -628,28 +775,33 @@ int bch2_check_bucket_ref(struct btree_trans *trans,
 			bch2_data_type_str(ptr_data_type),
 			(printbuf_reset(&buf),
 			 bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-		ret = -EIO;
-		goto err;
+		if (inserting)
+			goto err;
+		goto out;
 	}
 
-	if ((u64) bucket_sectors + sectors > U32_MAX) {
+	if ((u64) *bucket_sectors + sectors > U32_MAX) {
 		bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
 			      BCH_FSCK_ERR_bucket_sector_count_overflow,
 			"bucket %u:%zu gen %u data type %s sector count overflow: %u + %lli > U32_MAX\n"
 			"while marking %s",
 			ptr->dev, bucket_nr, b_gen,
 			bch2_data_type_str(bucket_data_type ?: ptr_data_type),
-			bucket_sectors, sectors,
+			*bucket_sectors, sectors,
 			(printbuf_reset(&buf),
 			 bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-		ret = -EIO;
-		goto err;
+		if (inserting)
+			goto err;
+		sectors = -*bucket_sectors;
 	}
+
+	*bucket_sectors += sectors;
 out:
 	printbuf_exit(&buf);
 	return ret;
 err:
 	bch2_dump_trans_updates(trans);
+	ret = -EIO;
 	goto out;
 }
 
@@ -786,29 +938,22 @@ need_mark:
 
 /* KEY_TYPE_extent: */
 
-static int __mark_pointer(struct btree_trans *trans,
+static int __mark_pointer(struct btree_trans *trans, struct bch_dev *ca,
 			  struct bkey_s_c k,
 			  const struct bch_extent_ptr *ptr,
 			  s64 sectors, enum bch_data_type ptr_data_type,
-			  u8 bucket_gen, u8 *bucket_data_type,
-			  u32 *dirty_sectors, u32 *cached_sectors)
+			  struct bch_alloc_v4 *a)
 {
 	u32 *dst_sectors = !ptr->cached
-		? dirty_sectors
-		: cached_sectors;
-	int ret = bch2_check_bucket_ref(trans, k, ptr, sectors, ptr_data_type,
-				   bucket_gen, *bucket_data_type, *dst_sectors);
+		? &a->dirty_sectors
+		: &a->cached_sectors;
+	int ret = bch2_bucket_ref_update(trans, ca, k, ptr, sectors, ptr_data_type,
+					 a->gen, a->data_type, dst_sectors);
 
 	if (ret)
 		return ret;
 
-	*dst_sectors += sectors;
-
-	if (!*dirty_sectors && !*cached_sectors)
-		*bucket_data_type = 0;
-	else if (*bucket_data_type != BCH_DATA_stripe)
-		*bucket_data_type = ptr_data_type;
-
+	alloc_data_type_set(a, ptr_data_type);
 	return 0;
 }
 
@@ -816,81 +961,69 @@ static int bch2_trigger_pointer(struct btree_trans *trans,
 			enum btree_id btree_id, unsigned level,
 			struct bkey_s_c k, struct extent_ptr_decoded p,
 			const union bch_extent_entry *entry,
-			s64 *sectors, unsigned flags)
+			s64 *sectors,
+			enum btree_iter_update_trigger_flags flags)
 {
-	bool insert = !(flags & BTREE_TRIGGER_OVERWRITE);
+	bool insert = !(flags & BTREE_TRIGGER_overwrite);
+	int ret = 0;
+
+	struct bch_fs *c = trans->c;
+	struct bch_dev *ca = bch2_dev_tryget(c, p.ptr.dev);
+	if (unlikely(!ca)) {
+		if (insert)
+			ret = -EIO;
+		goto err;
+	}
+
 	struct bpos bucket;
 	struct bch_backpointer bp;
-
-	bch2_extent_ptr_to_bp(trans->c, btree_id, level, k, p, entry, &bucket, &bp);
+	bch2_extent_ptr_to_bp(trans->c, ca, btree_id, level, k, p, entry, &bucket, &bp);
 	*sectors = insert ? bp.bucket_len : -((s64) bp.bucket_len);
 
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
-		struct btree_iter iter;
-		struct bkey_i_alloc_v4 *a = bch2_trans_start_alloc_update(trans, &iter, bucket);
-		int ret = PTR_ERR_OR_ZERO(a);
-		if (ret)
-			return ret;
-
-		ret = __mark_pointer(trans, k, &p.ptr, *sectors, bp.data_type,
-				     a->v.gen, &a->v.data_type,
-				     &a->v.dirty_sectors, &a->v.cached_sectors) ?:
-			bch2_trans_update(trans, &iter, &a->k_i, 0);
-		bch2_trans_iter_exit(trans, &iter);
-
+	if (flags & BTREE_TRIGGER_transactional) {
+		struct bkey_i_alloc_v4 *a = bch2_trans_start_alloc_update(trans, bucket);
+		ret = PTR_ERR_OR_ZERO(a) ?:
+			__mark_pointer(trans, ca, k, &p.ptr, *sectors, bp.data_type, &a->v);
 		if (ret)
-			return ret;
+			goto err;
 
 		if (!p.ptr.cached) {
-			ret = bch2_bucket_backpointer_mod(trans, bucket, bp, k, insert);
+			ret = bch2_bucket_backpointer_mod(trans, ca, bucket, bp, k, insert);
 			if (ret)
-				return ret;
+				goto err;
 		}
 	}
 
-	if (flags & BTREE_TRIGGER_GC) {
-		struct bch_fs *c = trans->c;
-		struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev);
-		enum bch_data_type data_type = bch2_bkey_ptr_data_type(k, p, entry);
-
+	if (flags & BTREE_TRIGGER_gc) {
 		percpu_down_read(&c->mark_lock);
-		struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr);
+		struct bucket *g = gc_bucket(ca, bucket.offset);
 		bucket_lock(g);
-		struct bucket old = *g;
-
-		u8 bucket_data_type = g->data_type;
-		int ret = __mark_pointer(trans, k, &p.ptr, *sectors,
-				     data_type, g->gen,
-				     &bucket_data_type,
-				     &g->dirty_sectors,
-				     &g->cached_sectors);
-		if (ret) {
-			bucket_unlock(g);
-			percpu_up_read(&c->mark_lock);
-			return ret;
+		struct bch_alloc_v4 old = bucket_m_to_alloc(*g), new = old;
+		ret = __mark_pointer(trans, ca, k, &p.ptr, *sectors, bp.data_type, &new);
+		if (!ret) {
+			alloc_to_bucket(g, new);
+			bch2_dev_usage_update(c, ca, &old, &new, 0, true);
 		}
-
-		g->data_type = bucket_data_type;
-		struct bucket new = *g;
 		bucket_unlock(g);
-		bch2_dev_usage_update_m(c, ca, &old, &new);
 		percpu_up_read(&c->mark_lock);
 	}
-
-	return 0;
+err:
+	bch2_dev_put(ca);
+	return ret;
 }
 
 static int bch2_trigger_stripe_ptr(struct btree_trans *trans,
 				struct bkey_s_c k,
 				struct extent_ptr_decoded p,
 				enum bch_data_type data_type,
-				s64 sectors, unsigned flags)
+				s64 sectors,
+				enum btree_iter_update_trigger_flags flags)
 {
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (flags & BTREE_TRIGGER_transactional) {
 		struct btree_iter iter;
 		struct bkey_i_stripe *s = bch2_bkey_get_mut_typed(trans, &iter,
 				BTREE_ID_stripes, POS(0, p.ec.idx),
-				BTREE_ITER_WITH_UPDATES, stripe);
+				BTREE_ITER_with_updates, stripe);
 		int ret = PTR_ERR_OR_ZERO(s);
 		if (unlikely(ret)) {
 			bch2_trans_inconsistent_on(bch2_err_matches(ret, ENOENT), trans,
@@ -920,10 +1053,10 @@ err:
 		return ret;
 	}
 
-	if (flags & BTREE_TRIGGER_GC) {
+	if (flags & BTREE_TRIGGER_gc) {
 		struct bch_fs *c = trans->c;
 
-		BUG_ON(!(flags & BTREE_TRIGGER_GC));
+		BUG_ON(!(flags & BTREE_TRIGGER_gc));
 
 		struct gc_stripe *m = genradix_ptr_alloc(&c->gc_stripes, p.ec.idx, GFP_KERNEL);
 		if (!m) {
@@ -959,9 +1092,10 @@ err:
 
 static int __trigger_extent(struct btree_trans *trans,
 			    enum btree_id btree_id, unsigned level,
-			    struct bkey_s_c k, unsigned flags)
+			    struct bkey_s_c k,
+			    enum btree_iter_update_trigger_flags flags)
 {
-	bool gc = flags & BTREE_TRIGGER_GC;
+	bool gc = flags & BTREE_TRIGGER_gc;
 	struct bch_fs *c = trans->c;
 	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
 	const union bch_extent_entry *entry;
@@ -970,7 +1104,7 @@ static int __trigger_extent(struct btree_trans *trans,
 	enum bch_data_type data_type = bkey_is_btree_ptr(k.k)
 		? BCH_DATA_btree
 		: BCH_DATA_user;
-	s64 dirty_sectors = 0;
+	s64 replicas_sectors = 0;
 	int ret = 0;
 
 	r.e.data_type	= data_type;
@@ -996,7 +1130,7 @@ static int __trigger_extent(struct btree_trans *trans,
 					return ret;
 			}
 		} else if (!p.has_ec) {
-			dirty_sectors	       += disk_sectors;
+			replicas_sectors       += disk_sectors;
 			r.e.devs[r.e.nr_devs++]	= p.ptr.dev;
 		} else {
 			ret = bch2_trigger_stripe_ptr(trans, k, p, data_type, disk_sectors, flags);
@@ -1014,8 +1148,8 @@ static int __trigger_extent(struct btree_trans *trans,
 
 	if (r.e.nr_devs) {
 		ret = !gc
-			? bch2_update_replicas_list(trans, &r.e, dirty_sectors)
-			: bch2_update_replicas(c, k, &r.e, dirty_sectors, 0, true);
+			? bch2_update_replicas_list(trans, &r.e, replicas_sectors)
+			: bch2_update_replicas(c, k, &r.e, replicas_sectors, 0, true);
 		if (unlikely(ret && gc)) {
 			struct printbuf buf = PRINTBUF;
 
@@ -1031,15 +1165,18 @@ static int __trigger_extent(struct btree_trans *trans,
 }
 
 int bch2_trigger_extent(struct btree_trans *trans,
-			enum btree_id btree_id, unsigned level,
+			enum btree_id btree, unsigned level,
 			struct bkey_s_c old, struct bkey_s new,
-			unsigned flags)
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bkey_ptrs_c new_ptrs = bch2_bkey_ptrs_c(new.s_c);
 	struct bkey_ptrs_c old_ptrs = bch2_bkey_ptrs_c(old);
 	unsigned new_ptrs_bytes = (void *) new_ptrs.end - (void *) new_ptrs.start;
 	unsigned old_ptrs_bytes = (void *) old_ptrs.end - (void *) old_ptrs.start;
 
+	if (unlikely(flags & BTREE_TRIGGER_check_repair))
+		return bch2_check_fix_ptrs(trans, btree, level, new.s_c, flags);
+
 	/* if pointers aren't changing - nothing to do: */
 	if (new_ptrs_bytes == old_ptrs_bytes &&
 	    !memcmp(new_ptrs.start,
@@ -1047,7 +1184,7 @@ int bch2_trigger_extent(struct btree_trans *trans,
 		    new_ptrs_bytes))
 		return 0;
 
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (flags & BTREE_TRIGGER_transactional) {
 		struct bch_fs *c = trans->c;
 		int mod = (int) bch2_bkey_needs_rebalance(c, new.s_c) -
 			  (int) bch2_bkey_needs_rebalance(c, old);
@@ -1060,8 +1197,8 @@ int bch2_trigger_extent(struct btree_trans *trans,
 		}
 	}
 
-	if (flags & (BTREE_TRIGGER_TRANSACTIONAL|BTREE_TRIGGER_GC))
-		return trigger_run_overwrite_then_insert(__trigger_extent, trans, btree_id, level, old, new, flags);
+	if (flags & (BTREE_TRIGGER_transactional|BTREE_TRIGGER_gc))
+		return trigger_run_overwrite_then_insert(__trigger_extent, trans, btree, level, old, new, flags);
 
 	return 0;
 }
@@ -1069,17 +1206,17 @@ int bch2_trigger_extent(struct btree_trans *trans,
 /* KEY_TYPE_reservation */
 
 static int __trigger_reservation(struct btree_trans *trans,
-				 enum btree_id btree_id, unsigned level,
-				 struct bkey_s_c k, unsigned flags)
+			enum btree_id btree_id, unsigned level, struct bkey_s_c k,
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_fs *c = trans->c;
 	unsigned replicas = bkey_s_c_to_reservation(k).v->nr_replicas;
 	s64 sectors = (s64) k.k->size * replicas;
 
-	if (flags & BTREE_TRIGGER_OVERWRITE)
+	if (flags & BTREE_TRIGGER_overwrite)
 		sectors = -sectors;
 
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (flags & BTREE_TRIGGER_transactional) {
 		int ret = bch2_replicas_deltas_realloc(trans, 0);
 		if (ret)
 			return ret;
@@ -1090,7 +1227,7 @@ static int __trigger_reservation(struct btree_trans *trans,
 		d->persistent_reserved[replicas - 1] += sectors;
 	}
 
-	if (flags & BTREE_TRIGGER_GC) {
+	if (flags & BTREE_TRIGGER_gc) {
 		percpu_down_read(&c->mark_lock);
 		preempt_disable();
 
@@ -1110,7 +1247,7 @@ static int __trigger_reservation(struct btree_trans *trans,
 int bch2_trigger_reservation(struct btree_trans *trans,
 			  enum btree_id btree_id, unsigned level,
 			  struct bkey_s_c old, struct bkey_s new,
-			  unsigned flags)
+			  enum btree_iter_update_trigger_flags flags)
 {
 	return trigger_run_overwrite_then_insert(__trigger_reservation, trans, btree_id, level, old, new, flags);
 }
@@ -1118,22 +1255,16 @@ int bch2_trigger_reservation(struct btree_trans *trans,
 /* Mark superblocks: */
 
 static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
-				    struct bch_dev *ca, size_t b,
+				    struct bch_dev *ca, u64 b,
 				    enum bch_data_type type,
 				    unsigned sectors)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
-	struct bkey_i_alloc_v4 *a;
 	int ret = 0;
 
-	/*
-	 * Backup superblock might be past the end of our normal usable space:
-	 */
-	if (b >= ca->mi.nbuckets)
-		return 0;
-
-	a = bch2_trans_start_alloc_update(trans, &iter, POS(ca->dev_idx, b));
+	struct bkey_i_alloc_v4 *a =
+		bch2_trans_start_alloc_update_noupdate(trans, &iter, POS(ca->dev_idx, b));
 	if (IS_ERR(a))
 		return PTR_ERR(a);
 
@@ -1161,20 +1292,75 @@ err:
 	return ret;
 }
 
+static int bch2_mark_metadata_bucket(struct bch_fs *c, struct bch_dev *ca,
+			u64 b, enum bch_data_type data_type, unsigned sectors,
+			enum btree_iter_update_trigger_flags flags)
+{
+	int ret = 0;
+
+	percpu_down_read(&c->mark_lock);
+	struct bucket *g = gc_bucket(ca, b);
+
+	bucket_lock(g);
+	struct bch_alloc_v4 old = bucket_m_to_alloc(*g);
+
+	if (bch2_fs_inconsistent_on(g->data_type &&
+			g->data_type != data_type, c,
+			"different types of data in same bucket: %s, %s",
+			bch2_data_type_str(g->data_type),
+			bch2_data_type_str(data_type))) {
+		ret = -EIO;
+		goto err;
+	}
+
+	if (bch2_fs_inconsistent_on((u64) g->dirty_sectors + sectors > ca->mi.bucket_size, c,
+			"bucket %u:%llu gen %u data type %s sector count overflow: %u + %u > bucket size",
+			ca->dev_idx, b, g->gen,
+			bch2_data_type_str(g->data_type ?: data_type),
+			g->dirty_sectors, sectors)) {
+		ret = -EIO;
+		goto err;
+	}
+
+	g->data_type = data_type;
+	g->dirty_sectors += sectors;
+	struct bch_alloc_v4 new = bucket_m_to_alloc(*g);
+err:
+	bucket_unlock(g);
+	if (!ret)
+		bch2_dev_usage_update(c, ca, &old, &new, 0, true);
+	percpu_up_read(&c->mark_lock);
+	return ret;
+}
+
 int bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
-				    struct bch_dev *ca, size_t b,
-				    enum bch_data_type type,
-				    unsigned sectors)
+			struct bch_dev *ca, u64 b,
+			enum bch_data_type type, unsigned sectors,
+			enum btree_iter_update_trigger_flags flags)
 {
-	return commit_do(trans, NULL, NULL, 0,
-			__bch2_trans_mark_metadata_bucket(trans, ca, b, type, sectors));
+	BUG_ON(type != BCH_DATA_free &&
+	       type != BCH_DATA_sb &&
+	       type != BCH_DATA_journal);
+
+	/*
+	 * Backup superblock might be past the end of our normal usable space:
+	 */
+	if (b >= ca->mi.nbuckets)
+		return 0;
+
+	if (flags & BTREE_TRIGGER_gc)
+		return bch2_mark_metadata_bucket(trans->c, ca, b, type, sectors, flags);
+	else if (flags & BTREE_TRIGGER_transactional)
+		return commit_do(trans, NULL, NULL, 0,
+				 __bch2_trans_mark_metadata_bucket(trans, ca, b, type, sectors));
+	else
+		BUG();
 }
 
 static int bch2_trans_mark_metadata_sectors(struct btree_trans *trans,
-					    struct bch_dev *ca,
-					    u64 start, u64 end,
-					    enum bch_data_type type,
-					    u64 *bucket, unsigned *bucket_sectors)
+			struct bch_dev *ca, u64 start, u64 end,
+			enum bch_data_type type, u64 *bucket, unsigned *bucket_sectors,
+			enum btree_iter_update_trigger_flags flags)
 {
 	do {
 		u64 b = sector_to_bucket(ca, start);
@@ -1183,7 +1369,7 @@ static int bch2_trans_mark_metadata_sectors(struct btree_trans *trans,
 
 		if (b != *bucket && *bucket_sectors) {
 			int ret = bch2_trans_mark_metadata_bucket(trans, ca, *bucket,
-								  type, *bucket_sectors);
+							type, *bucket_sectors, flags);
 			if (ret)
 				return ret;
 
@@ -1198,8 +1384,8 @@ static int bch2_trans_mark_metadata_sectors(struct btree_trans *trans,
 	return 0;
 }
 
-static int __bch2_trans_mark_dev_sb(struct btree_trans *trans,
-				    struct bch_dev *ca)
+static int __bch2_trans_mark_dev_sb(struct btree_trans *trans, struct bch_dev *ca,
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_sb_layout *layout = &ca->disk_sb.sb->layout;
 	u64 bucket = 0;
@@ -1212,21 +1398,21 @@ static int __bch2_trans_mark_dev_sb(struct btree_trans *trans,
 		if (offset == BCH_SB_SECTOR) {
 			ret = bch2_trans_mark_metadata_sectors(trans, ca,
 						0, BCH_SB_SECTOR,
-						BCH_DATA_sb, &bucket, &bucket_sectors);
+						BCH_DATA_sb, &bucket, &bucket_sectors, flags);
 			if (ret)
 				return ret;
 		}
 
 		ret = bch2_trans_mark_metadata_sectors(trans, ca, offset,
 				      offset + (1 << layout->sb_max_size_bits),
-				      BCH_DATA_sb, &bucket, &bucket_sectors);
+				      BCH_DATA_sb, &bucket, &bucket_sectors, flags);
 		if (ret)
 			return ret;
 	}
 
 	if (bucket_sectors) {
 		ret = bch2_trans_mark_metadata_bucket(trans, ca,
-				bucket, BCH_DATA_sb, bucket_sectors);
+				bucket, BCH_DATA_sb, bucket_sectors, flags);
 		if (ret)
 			return ret;
 	}
@@ -1234,7 +1420,7 @@ static int __bch2_trans_mark_dev_sb(struct btree_trans *trans,
 	for (i = 0; i < ca->journal.nr; i++) {
 		ret = bch2_trans_mark_metadata_bucket(trans, ca,
 				ca->journal.buckets[i],
-				BCH_DATA_journal, ca->mi.bucket_size);
+				BCH_DATA_journal, ca->mi.bucket_size, flags);
 		if (ret)
 			return ret;
 	}
@@ -1242,18 +1428,20 @@ static int __bch2_trans_mark_dev_sb(struct btree_trans *trans,
 	return 0;
 }
 
-int bch2_trans_mark_dev_sb(struct bch_fs *c, struct bch_dev *ca)
+int bch2_trans_mark_dev_sb(struct bch_fs *c, struct bch_dev *ca,
+			enum btree_iter_update_trigger_flags flags)
 {
-	int ret = bch2_trans_run(c, __bch2_trans_mark_dev_sb(trans, ca));
-
+	int ret = bch2_trans_run(c,
+		__bch2_trans_mark_dev_sb(trans, ca, flags));
 	bch_err_fn(c, ret);
 	return ret;
 }
 
-int bch2_trans_mark_dev_sbs(struct bch_fs *c)
+int bch2_trans_mark_dev_sbs_flags(struct bch_fs *c,
+			enum btree_iter_update_trigger_flags flags)
 {
 	for_each_online_member(c, ca) {
-		int ret = bch2_trans_mark_dev_sb(c, ca);
+		int ret = bch2_trans_mark_dev_sb(c, ca, flags);
 		if (ret) {
 			percpu_ref_put(&ca->ref);
 			return ret;
@@ -1263,6 +1451,11 @@ int bch2_trans_mark_dev_sbs(struct bch_fs *c)
 	return 0;
 }
 
+int bch2_trans_mark_dev_sbs(struct bch_fs *c)
+{
+	return bch2_trans_mark_dev_sbs_flags(c, BTREE_TRIGGER_transactional);
+}
+
 /* Disk reservations: */
 
 #define SECTORS_CACHE	1024
@@ -1331,6 +1524,31 @@ recalculate:
 
 /* Startup/shutdown: */
 
+void bch2_buckets_nouse_free(struct bch_fs *c)
+{
+	for_each_member_device(c, ca) {
+		kvfree_rcu_mightsleep(ca->buckets_nouse);
+		ca->buckets_nouse = NULL;
+	}
+}
+
+int bch2_buckets_nouse_alloc(struct bch_fs *c)
+{
+	for_each_member_device(c, ca) {
+		BUG_ON(ca->buckets_nouse);
+
+		ca->buckets_nouse = kvmalloc(BITS_TO_LONGS(ca->mi.nbuckets) *
+					    sizeof(unsigned long),
+					    GFP_KERNEL|__GFP_ZERO);
+		if (!ca->buckets_nouse) {
+			percpu_ref_put(&ca->ref);
+			return -BCH_ERR_ENOMEM_buckets_nouse;
+		}
+	}
+
+	return 0;
+}
+
 static void bucket_gens_free_rcu(struct rcu_head *rcu)
 {
 	struct bucket_gens *buckets =
@@ -1342,24 +1560,17 @@ static void bucket_gens_free_rcu(struct rcu_head *rcu)
 int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 {
 	struct bucket_gens *bucket_gens = NULL, *old_bucket_gens = NULL;
-	unsigned long *buckets_nouse = NULL;
 	bool resize = ca->bucket_gens != NULL;
 	int ret;
 
+	BUG_ON(resize && ca->buckets_nouse);
+
 	if (!(bucket_gens	= kvmalloc(sizeof(struct bucket_gens) + nbuckets,
 					   GFP_KERNEL|__GFP_ZERO))) {
 		ret = -BCH_ERR_ENOMEM_bucket_gens;
 		goto err;
 	}
 
-	if ((c->opts.buckets_nouse &&
-	     !(buckets_nouse	= kvmalloc(BITS_TO_LONGS(nbuckets) *
-					   sizeof(unsigned long),
-					   GFP_KERNEL|__GFP_ZERO)))) {
-		ret = -BCH_ERR_ENOMEM_buckets_nouse;
-		goto err;
-	}
-
 	bucket_gens->first_bucket = ca->mi.first_bucket;
 	bucket_gens->nbuckets	= nbuckets;
 
@@ -1377,17 +1588,11 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 		memcpy(bucket_gens->b,
 		       old_bucket_gens->b,
 		       n);
-		if (buckets_nouse)
-			memcpy(buckets_nouse,
-			       ca->buckets_nouse,
-			       BITS_TO_LONGS(n) * sizeof(unsigned long));
 	}
 
 	rcu_assign_pointer(ca->bucket_gens, bucket_gens);
 	bucket_gens	= old_bucket_gens;
 
-	swap(ca->buckets_nouse, buckets_nouse);
-
 	nbuckets = ca->mi.nbuckets;
 
 	if (resize) {
@@ -1398,7 +1603,6 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 
 	ret = 0;
 err:
-	kvfree(buckets_nouse);
 	if (bucket_gens)
 		call_rcu(&bucket_gens->rcu, bucket_gens_free_rcu);
 
diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h
index f9af5adabe8363..617ffde2fb7ad9 100644
--- a/fs/bcachefs/buckets.h
+++ b/fs/bcachefs/buckets.h
@@ -12,7 +12,7 @@
 #include "extents.h"
 #include "sb-members.h"
 
-static inline size_t sector_to_bucket(const struct bch_dev *ca, sector_t s)
+static inline u64 sector_to_bucket(const struct bch_dev *ca, sector_t s)
 {
 	return div_u64(s, ca->mi.bucket_size);
 }
@@ -30,8 +30,7 @@ static inline sector_t bucket_remainder(const struct bch_dev *ca, sector_t s)
 	return remainder;
 }
 
-static inline size_t sector_to_bucket_and_offset(const struct bch_dev *ca, sector_t s,
-						 u32 *offset)
+static inline u64 sector_to_bucket_and_offset(const struct bch_dev *ca, sector_t s, u32 *offset)
 {
 	return div_u64_rem(s, ca->mi.bucket_size, offset);
 }
@@ -94,7 +93,7 @@ static inline struct bucket *gc_bucket(struct bch_dev *ca, size_t b)
 {
 	struct bucket_array *buckets = gc_bucket_array(ca);
 
-	BUG_ON(b < buckets->first_bucket || b >= buckets->nbuckets);
+	BUG_ON(!bucket_valid(ca, b));
 	return buckets->b + b;
 }
 
@@ -111,7 +110,7 @@ static inline u8 *bucket_gen(struct bch_dev *ca, size_t b)
 {
 	struct bucket_gens *gens = bucket_gens(ca);
 
-	BUG_ON(b < gens->first_bucket || b >= gens->nbuckets);
+	BUG_ON(!bucket_valid(ca, b));
 	return gens->b + b;
 }
 
@@ -121,20 +120,16 @@ static inline size_t PTR_BUCKET_NR(const struct bch_dev *ca,
 	return sector_to_bucket(ca, ptr->offset);
 }
 
-static inline struct bpos PTR_BUCKET_POS(const struct bch_fs *c,
-				   const struct bch_extent_ptr *ptr)
+static inline struct bpos PTR_BUCKET_POS(const struct bch_dev *ca,
+					 const struct bch_extent_ptr *ptr)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
-
 	return POS(ptr->dev, PTR_BUCKET_NR(ca, ptr));
 }
 
-static inline struct bpos PTR_BUCKET_POS_OFFSET(const struct bch_fs *c,
+static inline struct bpos PTR_BUCKET_POS_OFFSET(const struct bch_dev *ca,
 						const struct bch_extent_ptr *ptr,
 						u32 *bucket_offset)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
-
 	return POS(ptr->dev, sector_to_bucket_and_offset(ca, ptr->offset, bucket_offset));
 }
 
@@ -175,17 +170,19 @@ static inline int gen_after(u8 a, u8 b)
 	return r > 0 ? r : 0;
 }
 
+static inline u8 dev_ptr_stale_rcu(struct bch_dev *ca, const struct bch_extent_ptr *ptr)
+{
+	return gen_after(*bucket_gen(ca, PTR_BUCKET_NR(ca, ptr)), ptr->gen);
+}
+
 /**
- * ptr_stale() - check if a pointer points into a bucket that has been
+ * dev_ptr_stale() - check if a pointer points into a bucket that has been
  * invalidated.
  */
-static inline u8 ptr_stale(struct bch_dev *ca,
-			   const struct bch_extent_ptr *ptr)
+static inline u8 dev_ptr_stale(struct bch_dev *ca, const struct bch_extent_ptr *ptr)
 {
-	u8 ret;
-
 	rcu_read_lock();
-	ret = gen_after(*bucket_gen(ca, PTR_BUCKET_NR(ca, ptr)), ptr->gen);
+	u8 ret = dev_ptr_stale_rcu(ca, ptr);
 	rcu_read_unlock();
 
 	return ret;
@@ -306,8 +303,6 @@ bch2_fs_usage_read_short(struct bch_fs *);
 void bch2_dev_usage_update(struct bch_fs *, struct bch_dev *,
 			   const struct bch_alloc_v4 *,
 			   const struct bch_alloc_v4 *, u64, bool);
-void bch2_dev_usage_update_m(struct bch_fs *, struct bch_dev *,
-			     struct bucket *, struct bucket *);
 
 /* key/bucket marking: */
 
@@ -333,27 +328,29 @@ int bch2_replicas_deltas_realloc(struct btree_trans *, unsigned);
 
 void bch2_fs_usage_initialize(struct bch_fs *);
 
-int bch2_check_bucket_ref(struct btree_trans *, struct bkey_s_c,
-			  const struct bch_extent_ptr *,
-			  s64, enum bch_data_type, u8, u8, u32);
+int bch2_bucket_ref_update(struct btree_trans *, struct bch_dev *,
+			   struct bkey_s_c, const struct bch_extent_ptr *,
+			   s64, enum bch_data_type, u8, u8, u32 *);
 
-int bch2_mark_metadata_bucket(struct bch_fs *, struct bch_dev *,
-			      size_t, enum bch_data_type, unsigned,
-			      struct gc_pos, unsigned);
+int bch2_check_fix_ptrs(struct btree_trans *,
+			enum btree_id, unsigned, struct bkey_s_c,
+			enum btree_iter_update_trigger_flags);
 
 int bch2_trigger_extent(struct btree_trans *, enum btree_id, unsigned,
-			struct bkey_s_c, struct bkey_s, unsigned);
+			struct bkey_s_c, struct bkey_s,
+			enum btree_iter_update_trigger_flags);
 int bch2_trigger_reservation(struct btree_trans *, enum btree_id, unsigned,
-			  struct bkey_s_c, struct bkey_s, unsigned);
+			  struct bkey_s_c, struct bkey_s,
+			  enum btree_iter_update_trigger_flags);
 
 #define trigger_run_overwrite_then_insert(_fn, _trans, _btree_id, _level, _old, _new, _flags)\
 ({												\
 	int ret = 0;										\
 												\
 	if (_old.k->type)									\
-		ret = _fn(_trans, _btree_id, _level, _old, _flags & ~BTREE_TRIGGER_INSERT);	\
+		ret = _fn(_trans, _btree_id, _level, _old, _flags & ~BTREE_TRIGGER_insert);	\
 	if (!ret && _new.k->type)								\
-		ret = _fn(_trans, _btree_id, _level, _new.s_c, _flags & ~BTREE_TRIGGER_OVERWRITE);\
+		ret = _fn(_trans, _btree_id, _level, _new.s_c, _flags & ~BTREE_TRIGGER_overwrite);\
 	ret;											\
 })
 
@@ -362,9 +359,13 @@ void bch2_trans_account_disk_usage_change(struct btree_trans *);
 void bch2_trans_fs_usage_revert(struct btree_trans *, struct replicas_delta_list *);
 int bch2_trans_fs_usage_apply(struct btree_trans *, struct replicas_delta_list *);
 
-int bch2_trans_mark_metadata_bucket(struct btree_trans *, struct bch_dev *,
-				    size_t, enum bch_data_type, unsigned);
-int bch2_trans_mark_dev_sb(struct bch_fs *, struct bch_dev *);
+int bch2_trans_mark_metadata_bucket(struct btree_trans *, struct bch_dev *, u64,
+				    enum bch_data_type, unsigned,
+				    enum btree_iter_update_trigger_flags);
+int bch2_trans_mark_dev_sb(struct bch_fs *, struct bch_dev *,
+				    enum btree_iter_update_trigger_flags);
+int bch2_trans_mark_dev_sbs_flags(struct bch_fs *,
+				    enum btree_iter_update_trigger_flags);
 int bch2_trans_mark_dev_sbs(struct bch_fs *);
 
 static inline bool is_superblock_bucket(struct bch_dev *ca, u64 b)
@@ -464,6 +465,9 @@ static inline u64 avail_factor(u64 r)
 	return div_u64(r << RESERVE_FACTOR, (1 << RESERVE_FACTOR) + 1);
 }
 
+void bch2_buckets_nouse_free(struct bch_fs *);
+int bch2_buckets_nouse_alloc(struct bch_fs *);
+
 int bch2_dev_buckets_resize(struct bch_fs *, struct bch_dev *, u64);
 void bch2_dev_buckets_free(struct bch_dev *);
 int bch2_dev_buckets_alloc(struct bch_fs *, struct bch_dev *);
diff --git a/fs/bcachefs/chardev.c b/fs/bcachefs/chardev.c
index 4d14f19f51850e..dbcd433db00498 100644
--- a/fs/bcachefs/chardev.c
+++ b/fs/bcachefs/chardev.c
@@ -961,7 +961,9 @@ static const struct file_operations bch_chardev_fops = {
 };
 
 static int bch_chardev_major;
-static struct class *bch_chardev_class;
+static const struct class bch_chardev_class = {
+	.name = "bcachefs",
+};
 static struct device *bch_chardev;
 
 void bch2_fs_chardev_exit(struct bch_fs *c)
@@ -978,7 +980,7 @@ int bch2_fs_chardev_init(struct bch_fs *c)
 	if (c->minor < 0)
 		return c->minor;
 
-	c->chardev = device_create(bch_chardev_class, NULL,
+	c->chardev = device_create(&bch_chardev_class, NULL,
 				   MKDEV(bch_chardev_major, c->minor), c,
 				   "bcachefs%u-ctl", c->minor);
 	if (IS_ERR(c->chardev))
@@ -989,32 +991,39 @@ int bch2_fs_chardev_init(struct bch_fs *c)
 
 void bch2_chardev_exit(void)
 {
-	if (!IS_ERR_OR_NULL(bch_chardev_class))
-		device_destroy(bch_chardev_class,
-			       MKDEV(bch_chardev_major, U8_MAX));
-	if (!IS_ERR_OR_NULL(bch_chardev_class))
-		class_destroy(bch_chardev_class);
+	device_destroy(&bch_chardev_class, MKDEV(bch_chardev_major, U8_MAX));
+	class_unregister(&bch_chardev_class);
 	if (bch_chardev_major > 0)
 		unregister_chrdev(bch_chardev_major, "bcachefs");
 }
 
 int __init bch2_chardev_init(void)
 {
+	int ret;
+
 	bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
 	if (bch_chardev_major < 0)
 		return bch_chardev_major;
 
-	bch_chardev_class = class_create("bcachefs");
-	if (IS_ERR(bch_chardev_class))
-		return PTR_ERR(bch_chardev_class);
+	ret = class_register(&bch_chardev_class);
+	if (ret)
+		goto major_out;
 
-	bch_chardev = device_create(bch_chardev_class, NULL,
+	bch_chardev = device_create(&bch_chardev_class, NULL,
 				    MKDEV(bch_chardev_major, U8_MAX),
 				    NULL, "bcachefs-ctl");
-	if (IS_ERR(bch_chardev))
-		return PTR_ERR(bch_chardev);
+	if (IS_ERR(bch_chardev)) {
+		ret = PTR_ERR(bch_chardev);
+		goto class_out;
+	}
 
 	return 0;
+
+class_out:
+	class_unregister(&bch_chardev_class);
+major_out:
+	unregister_chrdev(bch_chardev_major, "bcachefs-ctl");
+	return ret;
 }
 
 #endif /* NO_BCACHEFS_CHARDEV */
diff --git a/fs/bcachefs/checksum.c b/fs/bcachefs/checksum.c
index 7ed779b411f61e..8eccd7dd878730 100644
--- a/fs/bcachefs/checksum.c
+++ b/fs/bcachefs/checksum.c
@@ -493,14 +493,10 @@ static void bch2_sb_crypt_to_text(struct printbuf *out, struct bch_sb *sb,
 {
 	struct bch_sb_field_crypt *crypt = field_to_type(f, crypt);
 
-	prt_printf(out, "KFD:               %llu", BCH_CRYPT_KDF_TYPE(crypt));
-	prt_newline(out);
-	prt_printf(out, "scrypt n:          %llu", BCH_KDF_SCRYPT_N(crypt));
-	prt_newline(out);
-	prt_printf(out, "scrypt r:          %llu", BCH_KDF_SCRYPT_R(crypt));
-	prt_newline(out);
-	prt_printf(out, "scrypt p:          %llu", BCH_KDF_SCRYPT_P(crypt));
-	prt_newline(out);
+	prt_printf(out, "KFD:               %llu\n", BCH_CRYPT_KDF_TYPE(crypt));
+	prt_printf(out, "scrypt n:          %llu\n", BCH_KDF_SCRYPT_N(crypt));
+	prt_printf(out, "scrypt r:          %llu\n", BCH_KDF_SCRYPT_R(crypt));
+	prt_printf(out, "scrypt p:          %llu\n", BCH_KDF_SCRYPT_P(crypt));
 }
 
 const struct bch_sb_field_ops bch_sb_field_ops_crypt = {
diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c
index 0022b51ce3c09c..d72ca19665066c 100644
--- a/fs/bcachefs/data_update.c
+++ b/fs/bcachefs/data_update.c
@@ -106,7 +106,7 @@ static int __bch2_data_update_index_update(struct btree_trans *trans,
 
 	bch2_trans_iter_init(trans, &iter, m->btree_id,
 			     bkey_start_pos(&bch2_keylist_front(keys)->k),
-			     BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
+			     BTREE_ITER_slots|BTREE_ITER_intent);
 
 	while (1) {
 		struct bkey_s_c k;
@@ -203,6 +203,8 @@ restart_drop_conflicting_replicas:
 
 		/* Now, drop excess replicas: */
 restart_drop_extra_replicas:
+
+		rcu_read_lock();
 		bkey_for_each_ptr_decode(old.k, bch2_bkey_ptrs(bkey_i_to_s(insert)), p, entry) {
 			unsigned ptr_durability = bch2_extent_ptr_durability(c, &p);
 
@@ -214,6 +216,7 @@ restart_drop_extra_replicas:
 				goto restart_drop_extra_replicas;
 			}
 		}
+		rcu_read_unlock();
 
 		/* Finally, add the pointers we just wrote: */
 		extent_for_each_ptr_decode(extent_i_to_s(new), p, entry)
@@ -288,7 +291,7 @@ restart_drop_extra_replicas:
 						k.k->p, insert->k.p) ?:
 			bch2_bkey_set_needs_rebalance(c, insert, &op->opts) ?:
 			bch2_trans_update(trans, &iter, insert,
-				BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+				BTREE_UPDATE_internal_snapshot_node) ?:
 			bch2_trans_commit(trans, &op->res,
 				NULL,
 				BCH_TRANS_COMMIT_no_check_rw|
@@ -357,10 +360,11 @@ void bch2_data_update_exit(struct data_update *update)
 		bch2_bkey_ptrs_c(bkey_i_to_s_c(update->k.k));
 
 	bkey_for_each_ptr(ptrs, ptr) {
+		struct bch_dev *ca = bch2_dev_have_ref(c, ptr->dev);
 		if (c->opts.nocow_enabled)
 			bch2_bucket_nocow_unlock(&c->nocow_locks,
-						 PTR_BUCKET_POS(c, ptr), 0);
-		percpu_ref_put(&bch_dev_bkey_exists(c, ptr->dev)->ref);
+						 PTR_BUCKET_POS(ca, ptr), 0);
+		percpu_ref_put(&ca->ref);
 	}
 
 	bch2_bkey_buf_exit(&update->k, c);
@@ -386,8 +390,10 @@ static void bch2_update_unwritten_extent(struct btree_trans *trans,
 	while (bio_sectors(bio)) {
 		unsigned sectors = bio_sectors(bio);
 
+		bch2_trans_begin(trans);
+
 		bch2_trans_iter_init(trans, &iter, update->btree_id, update->op.pos,
-				     BTREE_ITER_SLOTS);
+				     BTREE_ITER_slots);
 		ret = lockrestart_do(trans, ({
 			k = bch2_btree_iter_peek_slot(&iter);
 			bkey_err(k);
@@ -480,15 +486,15 @@ int bch2_extent_drop_ptrs(struct btree_trans *trans,
 
 	/*
 	 * Since we're not inserting through an extent iterator
-	 * (BTREE_ITER_ALL_SNAPSHOTS iterators aren't extent iterators),
+	 * (BTREE_ITER_all_snapshots iterators aren't extent iterators),
 	 * we aren't using the extent overwrite path to delete, we're
 	 * just using the normal key deletion path:
 	 */
-	if (bkey_deleted(&n->k) && !(iter->flags & BTREE_ITER_IS_EXTENTS))
+	if (bkey_deleted(&n->k) && !(iter->flags & BTREE_ITER_is_extents))
 		n->k.size = 0;
 
 	return bch2_trans_relock(trans) ?:
-		bch2_trans_update(trans, iter, n, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+		bch2_trans_update(trans, iter, n, BTREE_UPDATE_internal_snapshot_node) ?:
 		bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
 }
 
@@ -540,14 +546,17 @@ int bch2_data_update_init(struct btree_trans *trans,
 	m->op.watermark		= m->data_opts.btree_insert_flags & BCH_WATERMARK_MASK;
 
 	bkey_for_each_ptr(ptrs, ptr)
-		percpu_ref_get(&bch_dev_bkey_exists(c, ptr->dev)->ref);
+		percpu_ref_get(&bch2_dev_bkey_exists(c, ptr->dev)->ref);
 
 	unsigned durability_have = 0, durability_removing = 0;
 
 	i = 0;
 	bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
+		struct bch_dev *ca = bch2_dev_have_ref(c, p.ptr.dev);
+		struct bpos bucket = PTR_BUCKET_POS(ca, &p.ptr);
 		bool locked;
 
+		rcu_read_lock();
 		if (((1U << i) & m->data_opts.rewrite_ptrs)) {
 			BUG_ON(p.ptr.cached);
 
@@ -561,6 +570,7 @@ int bch2_data_update_init(struct btree_trans *trans,
 			bch2_dev_list_add_dev(&m->op.devs_have, p.ptr.dev);
 			durability_have += bch2_extent_ptr_durability(c, &p);
 		}
+		rcu_read_unlock();
 
 		/*
 		 * op->csum_type is normally initialized from the fs/file's
@@ -579,15 +589,13 @@ int bch2_data_update_init(struct btree_trans *trans,
 			if (ctxt) {
 				move_ctxt_wait_event(ctxt,
 						(locked = bch2_bucket_nocow_trylock(&c->nocow_locks,
-									  PTR_BUCKET_POS(c, &p.ptr), 0)) ||
+									  bucket, 0)) ||
 						list_empty(&ctxt->ios));
 
 				if (!locked)
-					bch2_bucket_nocow_lock(&c->nocow_locks,
-							       PTR_BUCKET_POS(c, &p.ptr), 0);
+					bch2_bucket_nocow_lock(&c->nocow_locks, bucket, 0);
 			} else {
-				if (!bch2_bucket_nocow_trylock(&c->nocow_locks,
-							       PTR_BUCKET_POS(c, &p.ptr), 0)) {
+				if (!bch2_bucket_nocow_trylock(&c->nocow_locks, bucket, 0)) {
 					ret = -BCH_ERR_nocow_lock_blocked;
 					goto err;
 				}
@@ -649,10 +657,11 @@ int bch2_data_update_init(struct btree_trans *trans,
 err:
 	i = 0;
 	bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
+		struct bch_dev *ca = bch2_dev_have_ref(c, p.ptr.dev);
+		struct bpos bucket = PTR_BUCKET_POS(ca, &p.ptr);
 		if ((1U << i) & ptrs_locked)
-			bch2_bucket_nocow_unlock(&c->nocow_locks,
-						 PTR_BUCKET_POS(c, &p.ptr), 0);
-		percpu_ref_put(&bch_dev_bkey_exists(c, p.ptr.dev)->ref);
+			bch2_bucket_nocow_unlock(&c->nocow_locks, bucket, 0);
+		percpu_ref_put(&ca->ref);
 		i++;
 	}
 
diff --git a/fs/bcachefs/debug.c b/fs/bcachefs/debug.c
index cd99b739941447..b5aa4e5b224b6c 100644
--- a/fs/bcachefs/debug.c
+++ b/fs/bcachefs/debug.c
@@ -37,7 +37,7 @@ static bool bch2_btree_verify_replica(struct bch_fs *c, struct btree *b,
 	struct btree_node *n_ondisk = c->verify_ondisk;
 	struct btree_node *n_sorted = c->verify_data->data;
 	struct bset *sorted, *inmemory = &b->data->keys;
-	struct bch_dev *ca = bch_dev_bkey_exists(c, pick.ptr.dev);
+	struct bch_dev *ca = bch2_dev_bkey_exists(c, pick.ptr.dev);
 	struct bio *bio;
 	bool failed = false, saw_error = false;
 
@@ -194,7 +194,7 @@ void bch2_btree_node_ondisk_to_text(struct printbuf *out, struct bch_fs *c,
 		return;
 	}
 
-	ca = bch_dev_bkey_exists(c, pick.ptr.dev);
+	ca = bch2_dev_bkey_exists(c, pick.ptr.dev);
 	if (!bch2_dev_get_ioref(ca, READ)) {
 		prt_printf(out, "error getting device to read from: not online\n");
 		return;
@@ -375,8 +375,8 @@ static ssize_t bch2_read_btree(struct file *file, char __user *buf,
 	return flush_buf(i) ?:
 		bch2_trans_run(i->c,
 			for_each_btree_key(trans, iter, i->id, i->from,
-					   BTREE_ITER_PREFETCH|
-					   BTREE_ITER_ALL_SNAPSHOTS, k, ({
+					   BTREE_ITER_prefetch|
+					   BTREE_ITER_all_snapshots, k, ({
 				bch2_bkey_val_to_text(&i->buf, i->c, k);
 				prt_newline(&i->buf);
 				bch2_trans_unlock(trans);
@@ -459,8 +459,8 @@ static ssize_t bch2_read_bfloat_failed(struct file *file, char __user *buf,
 	return flush_buf(i) ?:
 		bch2_trans_run(i->c,
 			for_each_btree_key(trans, iter, i->id, i->from,
-					   BTREE_ITER_PREFETCH|
-					   BTREE_ITER_ALL_SNAPSHOTS, k, ({
+					   BTREE_ITER_prefetch|
+					   BTREE_ITER_all_snapshots, k, ({
 				struct btree_path_level *l =
 					&btree_iter_path(trans, &iter)->l[0];
 				struct bkey_packed *_k =
@@ -492,51 +492,26 @@ static void bch2_cached_btree_node_to_text(struct printbuf *out, struct bch_fs *
 	if (!out->nr_tabstops)
 		printbuf_tabstop_push(out, 32);
 
-	prt_printf(out, "%px btree=%s l=%u ",
-	       b,
-	       bch2_btree_id_str(b->c.btree_id),
-	       b->c.level);
-	prt_newline(out);
+	prt_printf(out, "%px btree=%s l=%u\n", b, bch2_btree_id_str(b->c.btree_id), b->c.level);
 
 	printbuf_indent_add(out, 2);
 
 	bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(&b->key));
 	prt_newline(out);
 
-	prt_printf(out, "flags: ");
-	prt_tab(out);
+	prt_printf(out, "flags:\t");
 	prt_bitflags(out, bch2_btree_node_flags, b->flags);
 	prt_newline(out);
 
-	prt_printf(out, "pcpu read locks: ");
-	prt_tab(out);
-	prt_printf(out, "%u", b->c.lock.readers != NULL);
-	prt_newline(out);
-
-	prt_printf(out, "written:");
-	prt_tab(out);
-	prt_printf(out, "%u", b->written);
-	prt_newline(out);
-
-	prt_printf(out, "writes blocked:");
-	prt_tab(out);
-	prt_printf(out, "%u", !list_empty_careful(&b->write_blocked));
-	prt_newline(out);
-
-	prt_printf(out, "will make reachable:");
-	prt_tab(out);
-	prt_printf(out, "%lx", b->will_make_reachable);
-	prt_newline(out);
-
-	prt_printf(out, "journal pin %px:", &b->writes[0].journal);
-	prt_tab(out);
-	prt_printf(out, "%llu", b->writes[0].journal.seq);
-	prt_newline(out);
+	prt_printf(out, "pcpu read locks:\t%u\n",	b->c.lock.readers != NULL);
+	prt_printf(out, "written:\t%u\n",		b->written);
+	prt_printf(out, "writes blocked:\t%u\n",	!list_empty_careful(&b->write_blocked));
+	prt_printf(out, "will make reachable:\t%lx\n",	b->will_make_reachable);
 
-	prt_printf(out, "journal pin %px:", &b->writes[1].journal);
-	prt_tab(out);
-	prt_printf(out, "%llu", b->writes[1].journal.seq);
-	prt_newline(out);
+	prt_printf(out, "journal pin %px:\t%llu\n",
+		   &b->writes[0].journal, b->writes[0].journal.seq);
+	prt_printf(out, "journal pin %px:\t%llu\n",
+		   &b->writes[1].journal, b->writes[1].journal.seq);
 
 	printbuf_indent_sub(out, 2);
 }
@@ -625,8 +600,7 @@ restart:
 
 		bch2_btree_trans_to_text(&i->buf, trans);
 
-		prt_printf(&i->buf, "backtrace:");
-		prt_newline(&i->buf);
+		prt_printf(&i->buf, "backtrace:\n");
 		printbuf_indent_add(&i->buf, 2);
 		bch2_prt_task_backtrace(&i->buf, task, 0, GFP_KERNEL);
 		printbuf_indent_sub(&i->buf, 2);
@@ -782,25 +756,20 @@ static ssize_t btree_transaction_stats_read(struct file *file, char __user *buf,
 		    !bch2_btree_transaction_fns[i->iter])
 			break;
 
-		prt_printf(&i->buf, "%s: ", bch2_btree_transaction_fns[i->iter]);
-		prt_newline(&i->buf);
+		prt_printf(&i->buf, "%s:\n", bch2_btree_transaction_fns[i->iter]);
 		printbuf_indent_add(&i->buf, 2);
 
 		mutex_lock(&s->lock);
 
-		prt_printf(&i->buf, "Max mem used: %u", s->max_mem);
-		prt_newline(&i->buf);
-
-		prt_printf(&i->buf, "Transaction duration:");
-		prt_newline(&i->buf);
+		prt_printf(&i->buf, "Max mem used: %u\n", s->max_mem);
+		prt_printf(&i->buf, "Transaction duration:\n");
 
 		printbuf_indent_add(&i->buf, 2);
 		bch2_time_stats_to_text(&i->buf, &s->duration);
 		printbuf_indent_sub(&i->buf, 2);
 
 		if (IS_ENABLED(CONFIG_BCACHEFS_LOCK_TIME_STATS)) {
-			prt_printf(&i->buf, "Lock hold times:");
-			prt_newline(&i->buf);
+			prt_printf(&i->buf, "Lock hold times:\n");
 
 			printbuf_indent_add(&i->buf, 2);
 			bch2_time_stats_to_text(&i->buf, &s->lock_hold_times);
@@ -808,8 +777,7 @@ static ssize_t btree_transaction_stats_read(struct file *file, char __user *buf,
 		}
 
 		if (s->max_paths_text) {
-			prt_printf(&i->buf, "Maximum allocated btree paths (%u):", s->nr_max_paths);
-			prt_newline(&i->buf);
+			prt_printf(&i->buf, "Maximum allocated btree paths (%u):\n", s->nr_max_paths);
 
 			printbuf_indent_add(&i->buf, 2);
 			prt_str_indented(&i->buf, s->max_paths_text);
diff --git a/fs/bcachefs/dirent.c b/fs/bcachefs/dirent.c
index d37bd07afbfe40..dd9de8e5a28a98 100644
--- a/fs/bcachefs/dirent.c
+++ b/fs/bcachefs/dirent.c
@@ -205,7 +205,7 @@ int bch2_dirent_create_snapshot(struct btree_trans *trans,
 			const struct bch_hash_info *hash_info,
 			u8 type, const struct qstr *name, u64 dst_inum,
 			u64 *dir_offset,
-			bch_str_hash_flags_t str_hash_flags)
+			enum btree_iter_update_trigger_flags flags)
 {
 	subvol_inum dir_inum = { .subvol = dir_subvol, .inum = dir };
 	struct bkey_i_dirent *dirent;
@@ -220,9 +220,8 @@ int bch2_dirent_create_snapshot(struct btree_trans *trans,
 	dirent->k.p.snapshot	= snapshot;
 
 	ret = bch2_hash_set_in_snapshot(trans, bch2_dirent_hash_desc, hash_info,
-					dir_inum, snapshot,
-					&dirent->k_i, str_hash_flags,
-					BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+					dir_inum, snapshot, &dirent->k_i,
+					flags|BTREE_UPDATE_internal_snapshot_node);
 	*dir_offset = dirent->k.p.offset;
 
 	return ret;
@@ -232,7 +231,7 @@ int bch2_dirent_create(struct btree_trans *trans, subvol_inum dir,
 		       const struct bch_hash_info *hash_info,
 		       u8 type, const struct qstr *name, u64 dst_inum,
 		       u64 *dir_offset,
-		       bch_str_hash_flags_t str_hash_flags)
+		       enum btree_iter_update_trigger_flags flags)
 {
 	struct bkey_i_dirent *dirent;
 	int ret;
@@ -243,7 +242,7 @@ int bch2_dirent_create(struct btree_trans *trans, subvol_inum dir,
 		return ret;
 
 	ret = bch2_hash_set(trans, bch2_dirent_hash_desc, hash_info,
-			    dir, &dirent->k_i, str_hash_flags);
+			    dir, &dirent->k_i, flags);
 	*dir_offset = dirent->k.p.offset;
 
 	return ret;
@@ -272,7 +271,7 @@ int bch2_dirent_read_target(struct btree_trans *trans, subvol_inum dir,
 	} else {
 		target->subvol	= le32_to_cpu(d.v->d_child_subvol);
 
-		ret = bch2_subvolume_get(trans, target->subvol, true, BTREE_ITER_CACHED, &s);
+		ret = bch2_subvolume_get(trans, target->subvol, true, BTREE_ITER_cached, &s);
 
 		target->inum	= le64_to_cpu(s.inode);
 	}
@@ -301,13 +300,9 @@ int bch2_dirent_rename(struct btree_trans *trans,
 	memset(dst_inum, 0, sizeof(*dst_inum));
 
 	/* Lookup src: */
-	ret = bch2_hash_lookup(trans, &src_iter, bch2_dirent_hash_desc,
-			       src_hash, src_dir, src_name,
-			       BTREE_ITER_INTENT);
-	if (ret)
-		goto out;
-
-	old_src = bch2_btree_iter_peek_slot(&src_iter);
+	old_src = bch2_hash_lookup(trans, &src_iter, bch2_dirent_hash_desc,
+				   src_hash, src_dir, src_name,
+				   BTREE_ITER_intent);
 	ret = bkey_err(old_src);
 	if (ret)
 		goto out;
@@ -329,13 +324,9 @@ int bch2_dirent_rename(struct btree_trans *trans,
 		if (ret)
 			goto out;
 	} else {
-		ret = bch2_hash_lookup(trans, &dst_iter, bch2_dirent_hash_desc,
-				       dst_hash, dst_dir, dst_name,
-				       BTREE_ITER_INTENT);
-		if (ret)
-			goto out;
-
-		old_dst = bch2_btree_iter_peek_slot(&dst_iter);
+		old_dst = bch2_hash_lookup(trans, &dst_iter, bch2_dirent_hash_desc,
+					    dst_hash, dst_dir, dst_name,
+					    BTREE_ITER_intent);
 		ret = bkey_err(old_dst);
 		if (ret)
 			goto out;
@@ -450,7 +441,7 @@ out_set_src:
 	if (delete_src) {
 		bch2_btree_iter_set_snapshot(&src_iter, old_src.k->p.snapshot);
 		ret =   bch2_btree_iter_traverse(&src_iter) ?:
-			bch2_btree_delete_at(trans, &src_iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+			bch2_btree_delete_at(trans, &src_iter, BTREE_UPDATE_internal_snapshot_node);
 		if (ret)
 			goto out;
 	}
@@ -458,7 +449,7 @@ out_set_src:
 	if (delete_dst) {
 		bch2_btree_iter_set_snapshot(&dst_iter, old_dst.k->p.snapshot);
 		ret =   bch2_btree_iter_traverse(&dst_iter) ?:
-			bch2_btree_delete_at(trans, &dst_iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+			bch2_btree_delete_at(trans, &dst_iter, BTREE_UPDATE_internal_snapshot_node);
 		if (ret)
 			goto out;
 	}
@@ -479,13 +470,9 @@ int bch2_dirent_lookup_trans(struct btree_trans *trans,
 			     const struct qstr *name, subvol_inum *inum,
 			     unsigned flags)
 {
-	int ret = bch2_hash_lookup(trans, iter, bch2_dirent_hash_desc,
-				   hash_info, dir, name, flags);
-	if (ret)
-		return ret;
-
-	struct bkey_s_c k = bch2_btree_iter_peek_slot(iter);
-	ret = bkey_err(k);
+	struct bkey_s_c k = bch2_hash_lookup(trans, iter, bch2_dirent_hash_desc,
+					     hash_info, dir, name, flags);
+	int ret = bkey_err(k);
 	if (ret)
 		goto err;
 
@@ -541,16 +528,26 @@ int bch2_empty_dir_trans(struct btree_trans *trans, subvol_inum dir)
 		bch2_empty_dir_snapshot(trans, dir.inum, dir.subvol, snapshot);
 }
 
+static int bch2_dir_emit(struct dir_context *ctx, struct bkey_s_c_dirent d, subvol_inum target)
+{
+	struct qstr name = bch2_dirent_get_name(d);
+	bool ret = dir_emit(ctx, name.name,
+		      name.len,
+		      target.inum,
+		      vfs_d_type(d.v->d_type));
+	if (ret)
+		ctx->pos = d.k->p.offset + 1;
+	return ret;
+}
+
 int bch2_readdir(struct bch_fs *c, subvol_inum inum, struct dir_context *ctx)
 {
 	struct btree_trans *trans = bch2_trans_get(c);
 	struct btree_iter iter;
 	struct bkey_s_c k;
-	struct bkey_s_c_dirent dirent;
 	subvol_inum target;
 	u32 snapshot;
 	struct bkey_buf sk;
-	struct qstr name;
 	int ret;
 
 	bch2_bkey_buf_init(&sk);
@@ -567,7 +564,9 @@ retry:
 		if (k.k->type != KEY_TYPE_dirent)
 			continue;
 
-		dirent = bkey_s_c_to_dirent(k);
+		/* dir_emit() can fault and block: */
+		bch2_bkey_buf_reassemble(&sk, c, k);
+		struct bkey_s_c_dirent dirent = bkey_i_to_s_c_dirent(sk.k);
 
 		ret = bch2_dirent_read_target(trans, inum, dirent, &target);
 		if (ret < 0)
@@ -575,28 +574,22 @@ retry:
 		if (ret)
 			continue;
 
-		/* dir_emit() can fault and block: */
-		bch2_bkey_buf_reassemble(&sk, c, k);
-		dirent = bkey_i_to_s_c_dirent(sk.k);
-		bch2_trans_unlock(trans);
-
-		name = bch2_dirent_get_name(dirent);
-
-		ctx->pos = dirent.k->p.offset;
-		if (!dir_emit(ctx, name.name,
-			      name.len,
-			      target.inum,
-			      vfs_d_type(dirent.v->d_type)))
-			break;
-		ctx->pos = dirent.k->p.offset + 1;
-
 		/*
 		 * read_target looks up subvolumes, we can overflow paths if the
 		 * directory has many subvolumes in it
+		 *
+		 * XXX: btree_trans_too_many_iters() is something we'd like to
+		 * get rid of, and there's no good reason to be using it here
+		 * except that we don't yet have a for_each_btree_key() helper
+		 * that does subvolume_get_snapshot().
 		 */
-		ret = btree_trans_too_many_iters(trans);
-		if (ret)
+		ret =   drop_locks_do(trans,
+				bch2_dir_emit(ctx, dirent, target)) ?:
+			btree_trans_too_many_iters(trans);
+		if (ret) {
+			ret = ret < 0 ? ret : 0;
 			break;
+		}
 	}
 	bch2_trans_iter_exit(trans, &iter);
 err:
diff --git a/fs/bcachefs/dirent.h b/fs/bcachefs/dirent.h
index bee55cca2aa0dd..635ceb3c07fbc8 100644
--- a/fs/bcachefs/dirent.h
+++ b/fs/bcachefs/dirent.h
@@ -38,11 +38,11 @@ int bch2_dirent_read_target(struct btree_trans *, subvol_inum,
 int bch2_dirent_create_snapshot(struct btree_trans *, u32, u64, u32,
 			const struct bch_hash_info *, u8,
 			const struct qstr *, u64, u64 *,
-			bch_str_hash_flags_t);
+			enum btree_iter_update_trigger_flags);
 int bch2_dirent_create(struct btree_trans *, subvol_inum,
 		       const struct bch_hash_info *, u8,
 		       const struct qstr *, u64, u64 *,
-		       bch_str_hash_flags_t);
+		       enum btree_iter_update_trigger_flags);
 
 static inline unsigned vfs_d_type(unsigned type)
 {
diff --git a/fs/bcachefs/disk_groups.c b/fs/bcachefs/disk_groups.c
index 06a7df529b401c..04b6aa4144765e 100644
--- a/fs/bcachefs/disk_groups.c
+++ b/fs/bcachefs/disk_groups.c
@@ -177,7 +177,7 @@ int bch2_sb_disk_groups_to_cpu(struct bch_fs *c)
 		struct bch_member m = bch2_sb_member_get(c->disk_sb.sb, i);
 		struct bch_disk_group_cpu *dst;
 
-		if (!bch2_member_exists(&m))
+		if (!bch2_member_alive(&m))
 			continue;
 
 		g = BCH_MEMBER_GROUP(&m);
@@ -588,7 +588,7 @@ static void bch2_target_to_text_sb(struct printbuf *out, struct bch_sb *sb, unsi
 	case TARGET_DEV: {
 		struct bch_member m = bch2_sb_member_get(sb, t.dev);
 
-		if (bch2_dev_exists(sb, t.dev)) {
+		if (bch2_member_exists(sb, t.dev)) {
 			prt_printf(out, "Device ");
 			pr_uuid(out, m.uuid.b);
 			prt_printf(out, " (%u)", t.dev);
diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index 556a217108d32e..271b876b7be9c5 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -163,146 +163,189 @@ void bch2_stripe_to_text(struct printbuf *out, struct bch_fs *c,
 
 /* Triggers: */
 
-static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans,
-					 struct bkey_s_c_stripe s,
-					 unsigned idx, bool deleting)
-{
-	struct bch_fs *c = trans->c;
-	const struct bch_extent_ptr *ptr = &s.v->ptrs[idx];
-	struct btree_iter iter;
-	struct bkey_i_alloc_v4 *a;
-	enum bch_data_type data_type = idx >= s.v->nr_blocks - s.v->nr_redundant
-		? BCH_DATA_parity : 0;
-	s64 sectors = data_type ? le16_to_cpu(s.v->sectors) : 0;
+static int __mark_stripe_bucket(struct btree_trans *trans,
+				struct bch_dev *ca,
+				struct bkey_s_c_stripe s,
+				unsigned ptr_idx, bool deleting,
+				struct bpos bucket,
+				struct bch_alloc_v4 *a,
+				enum btree_iter_update_trigger_flags flags)
+{
+	const struct bch_extent_ptr *ptr = s.v->ptrs + ptr_idx;
+	unsigned nr_data = s.v->nr_blocks - s.v->nr_redundant;
+	bool parity = ptr_idx >= nr_data;
+	enum bch_data_type data_type = parity ? BCH_DATA_parity : BCH_DATA_stripe;
+	s64 sectors = parity ? le16_to_cpu(s.v->sectors) : 0;
+	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
+	struct bch_fs *c = trans->c;
 	if (deleting)
 		sectors = -sectors;
 
-	a = bch2_trans_start_alloc_update(trans, &iter, PTR_BUCKET_POS(c, ptr));
-	if (IS_ERR(a))
-		return PTR_ERR(a);
-
-	ret = bch2_check_bucket_ref(trans, s.s_c, ptr, sectors, data_type,
-				    a->v.gen, a->v.data_type,
-				    a->v.dirty_sectors);
-	if (ret)
-		goto err;
-
 	if (!deleting) {
-		if (bch2_trans_inconsistent_on(a->v.stripe ||
-					       a->v.stripe_redundancy, trans,
-				"bucket %llu:%llu gen %u data type %s dirty_sectors %u: multiple stripes using same bucket (%u, %llu)",
-				iter.pos.inode, iter.pos.offset, a->v.gen,
-				bch2_data_type_str(a->v.data_type),
-				a->v.dirty_sectors,
-				a->v.stripe, s.k->p.offset)) {
+		if (bch2_trans_inconsistent_on(a->stripe ||
+					       a->stripe_redundancy, trans,
+				"bucket %llu:%llu gen %u data type %s dirty_sectors %u: multiple stripes using same bucket (%u, %llu)\n%s",
+				bucket.inode, bucket.offset, a->gen,
+				bch2_data_type_str(a->data_type),
+				a->dirty_sectors,
+				a->stripe, s.k->p.offset,
+				(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
 			ret = -EIO;
 			goto err;
 		}
 
-		if (bch2_trans_inconsistent_on(data_type && a->v.dirty_sectors, trans,
-				"bucket %llu:%llu gen %u data type %s dirty_sectors %u: data already in stripe bucket %llu",
-				iter.pos.inode, iter.pos.offset, a->v.gen,
-				bch2_data_type_str(a->v.data_type),
-				a->v.dirty_sectors,
-				s.k->p.offset)) {
+		if (bch2_trans_inconsistent_on(parity && bch2_bucket_sectors_total(*a), trans,
+				"bucket %llu:%llu gen %u data type %s dirty_sectors %u cached_sectors %u: data already in parity bucket\n%s",
+				bucket.inode, bucket.offset, a->gen,
+				bch2_data_type_str(a->data_type),
+				a->dirty_sectors,
+				a->cached_sectors,
+				(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
 			ret = -EIO;
 			goto err;
 		}
-
-		a->v.stripe		= s.k->p.offset;
-		a->v.stripe_redundancy	= s.v->nr_redundant;
-		a->v.data_type		= BCH_DATA_stripe;
 	} else {
-		if (bch2_trans_inconsistent_on(a->v.stripe != s.k->p.offset ||
-					       a->v.stripe_redundancy != s.v->nr_redundant, trans,
-				"bucket %llu:%llu gen %u: not marked as stripe when deleting stripe %llu (got %u)",
-				iter.pos.inode, iter.pos.offset, a->v.gen,
-				s.k->p.offset, a->v.stripe)) {
+		if (bch2_trans_inconsistent_on(a->stripe != s.k->p.offset ||
+					       a->stripe_redundancy != s.v->nr_redundant, trans,
+				"bucket %llu:%llu gen %u: not marked as stripe when deleting stripe (got %u)\n%s",
+				bucket.inode, bucket.offset, a->gen,
+				a->stripe,
+				(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
 			ret = -EIO;
 			goto err;
 		}
 
-		a->v.stripe		= 0;
-		a->v.stripe_redundancy	= 0;
-		a->v.data_type		= alloc_data_type(a->v, BCH_DATA_user);
+		if (bch2_trans_inconsistent_on(a->data_type != data_type, trans,
+				"bucket %llu:%llu gen %u data type %s: wrong data type when stripe, should be %s\n%s",
+				bucket.inode, bucket.offset, a->gen,
+				bch2_data_type_str(a->data_type),
+				bch2_data_type_str(data_type),
+				(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+			ret = -EIO;
+			goto err;
+		}
+
+		if (bch2_trans_inconsistent_on(parity &&
+					       (a->dirty_sectors != -sectors ||
+						a->cached_sectors), trans,
+				"bucket %llu:%llu gen %u dirty_sectors %u cached_sectors %u: wrong sectors when deleting parity block of stripe\n%s",
+				bucket.inode, bucket.offset, a->gen,
+				a->dirty_sectors,
+				a->cached_sectors,
+				(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
+			ret = -EIO;
+			goto err;
+		}
 	}
 
-	a->v.dirty_sectors += sectors;
-	if (data_type)
-		a->v.data_type = !deleting ? data_type : 0;
+	if (sectors) {
+		ret = bch2_bucket_ref_update(trans, ca, s.s_c, ptr, sectors, data_type,
+					     a->gen, a->data_type, &a->dirty_sectors);
+		if (ret)
+			goto err;
+	}
 
-	ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
-	if (ret)
-		goto err;
+	if (!deleting) {
+		a->stripe		= s.k->p.offset;
+		a->stripe_redundancy	= s.v->nr_redundant;
+	} else {
+		a->stripe		= 0;
+		a->stripe_redundancy	= 0;
+	}
+
+	alloc_data_type_set(a, data_type);
 err:
-	bch2_trans_iter_exit(trans, &iter);
+	printbuf_exit(&buf);
 	return ret;
 }
 
 static int mark_stripe_bucket(struct btree_trans *trans,
-			      struct bkey_s_c k,
-			      unsigned ptr_idx,
-			      unsigned flags)
+			      struct bkey_s_c_stripe s,
+			      unsigned ptr_idx, bool deleting,
+			      enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_fs *c = trans->c;
-	const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
-	unsigned nr_data = s->nr_blocks - s->nr_redundant;
-	bool parity = ptr_idx >= nr_data;
-	enum bch_data_type data_type = parity ? BCH_DATA_parity : BCH_DATA_stripe;
-	s64 sectors = parity ? le16_to_cpu(s->sectors) : 0;
-	const struct bch_extent_ptr *ptr = s->ptrs + ptr_idx;
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
-	struct bucket old, new, *g;
-	struct printbuf buf = PRINTBUF;
+	const struct bch_extent_ptr *ptr = s.v->ptrs + ptr_idx;
 	int ret = 0;
 
-	BUG_ON(!(flags & BTREE_TRIGGER_GC));
+	struct bch_dev *ca = bch2_dev_tryget(c, ptr->dev);
+	if (unlikely(!ca)) {
+		if (!(flags & BTREE_TRIGGER_overwrite))
+			ret = -EIO;
+		goto err;
+	}
 
-	/* * XXX doesn't handle deletion */
+	struct bpos bucket = PTR_BUCKET_POS(ca, ptr);
 
-	percpu_down_read(&c->mark_lock);
-	g = PTR_GC_BUCKET(ca, ptr);
+	if (flags & BTREE_TRIGGER_transactional) {
+		struct bkey_i_alloc_v4 *a =
+			bch2_trans_start_alloc_update(trans, bucket);
+		ret = PTR_ERR_OR_ZERO(a) ?:
+			__mark_stripe_bucket(trans, ca, s, ptr_idx, deleting, bucket, &a->v, flags);
+	}
 
-	if (g->dirty_sectors ||
-	    (g->stripe && g->stripe != k.k->p.offset)) {
-		bch2_fs_inconsistent(c,
-			      "bucket %u:%zu gen %u: multiple stripes using same bucket\n%s",
-			      ptr->dev, PTR_BUCKET_NR(ca, ptr), g->gen,
-			      (bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-		ret = -EINVAL;
-		goto err;
+	if (flags & BTREE_TRIGGER_gc) {
+		percpu_down_read(&c->mark_lock);
+		struct bucket *g = gc_bucket(ca, bucket.offset);
+		bucket_lock(g);
+		struct bch_alloc_v4 old = bucket_m_to_alloc(*g), new = old;
+		ret = __mark_stripe_bucket(trans, ca, s, ptr_idx, deleting, bucket, &new, flags);
+		if (!ret) {
+			alloc_to_bucket(g, new);
+			bch2_dev_usage_update(c, ca, &old, &new, 0, true);
+		}
+		bucket_unlock(g);
+		percpu_up_read(&c->mark_lock);
 	}
+err:
+	bch2_dev_put(ca);
+	return ret;
+}
+
+static int mark_stripe_buckets(struct btree_trans *trans,
+			       struct bkey_s_c old, struct bkey_s_c new,
+			       enum btree_iter_update_trigger_flags flags)
+{
+	const struct bch_stripe *old_s = old.k->type == KEY_TYPE_stripe
+		? bkey_s_c_to_stripe(old).v : NULL;
+	const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe
+		? bkey_s_c_to_stripe(new).v : NULL;
 
-	bucket_lock(g);
-	old = *g;
+	BUG_ON(old_s && new_s && old_s->nr_blocks != new_s->nr_blocks);
 
-	ret = bch2_check_bucket_ref(trans, k, ptr, sectors, data_type,
-				    g->gen, g->data_type,
-				    g->dirty_sectors);
-	if (ret)
-		goto err;
+	unsigned nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
 
-	g->data_type = data_type;
-	g->dirty_sectors += sectors;
+	for (unsigned i = 0; i < nr_blocks; i++) {
+		if (new_s && old_s &&
+		    !memcmp(&new_s->ptrs[i],
+			    &old_s->ptrs[i],
+			    sizeof(new_s->ptrs[i])))
+			continue;
 
-	g->stripe		= k.k->p.offset;
-	g->stripe_redundancy	= s->nr_redundant;
-	new = *g;
-err:
-	bucket_unlock(g);
-	if (!ret)
-		bch2_dev_usage_update_m(c, ca, &old, &new);
-	percpu_up_read(&c->mark_lock);
-	printbuf_exit(&buf);
-	return ret;
+		if (new_s) {
+			int ret = mark_stripe_bucket(trans,
+					bkey_s_c_to_stripe(new), i, false, flags);
+			if (ret)
+				return ret;
+		}
+
+		if (old_s) {
+			int ret = mark_stripe_bucket(trans,
+					bkey_s_c_to_stripe(old), i, true, flags);
+			if (ret)
+				return ret;
+		}
+	}
+
+	return 0;
 }
 
 int bch2_trigger_stripe(struct btree_trans *trans,
-			enum btree_id btree_id, unsigned level,
+			enum btree_id btree, unsigned level,
 			struct bkey_s_c old, struct bkey_s _new,
-			unsigned flags)
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bkey_s_c new = _new.s_c;
 	struct bch_fs *c = trans->c;
@@ -312,7 +355,10 @@ int bch2_trigger_stripe(struct btree_trans *trans,
 	const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe
 		? bkey_s_c_to_stripe(new).v : NULL;
 
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (unlikely(flags & BTREE_TRIGGER_check_repair))
+		return bch2_check_fix_ptrs(trans, btree, level, _new.s_c, flags);
+
+	if (flags & BTREE_TRIGGER_transactional) {
 		/*
 		 * If the pointers aren't changing, we don't need to do anything:
 		 */
@@ -347,31 +393,12 @@ int bch2_trigger_stripe(struct btree_trans *trans,
 				return ret;
 		}
 
-		unsigned nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
-		for (unsigned i = 0; i < nr_blocks; i++) {
-			if (new_s && old_s &&
-			    !memcmp(&new_s->ptrs[i],
-				    &old_s->ptrs[i],
-				    sizeof(new_s->ptrs[i])))
-				continue;
-
-			if (new_s) {
-				int ret = bch2_trans_mark_stripe_bucket(trans,
-						bkey_s_c_to_stripe(new), i, false);
-				if (ret)
-					return ret;
-			}
-
-			if (old_s) {
-				int ret = bch2_trans_mark_stripe_bucket(trans,
-						bkey_s_c_to_stripe(old), i, true);
-				if (ret)
-					return ret;
-			}
-		}
+		int ret = mark_stripe_buckets(trans, old, new, flags);
+		if (ret)
+			return ret;
 	}
 
-	if (flags & BTREE_TRIGGER_ATOMIC) {
+	if (flags & BTREE_TRIGGER_atomic) {
 		struct stripe *m = genradix_ptr(&c->stripes, idx);
 
 		if (!m) {
@@ -410,7 +437,7 @@ int bch2_trigger_stripe(struct btree_trans *trans,
 		}
 	}
 
-	if (flags & BTREE_TRIGGER_GC) {
+	if (flags & BTREE_TRIGGER_gc) {
 		struct gc_stripe *m =
 			genradix_ptr_alloc(&c->gc_stripes, idx, GFP_KERNEL);
 
@@ -439,13 +466,11 @@ int bch2_trigger_stripe(struct btree_trans *trans,
 		 */
 		memset(m->block_sectors, 0, sizeof(m->block_sectors));
 
-		for (unsigned i = 0; i < new_s->nr_blocks; i++) {
-			int ret = mark_stripe_bucket(trans, new, i, flags);
-			if (ret)
-				return ret;
-		}
+		int ret = mark_stripe_buckets(trans, old, new, flags);
+		if (ret)
+			return ret;
 
-		int ret = bch2_update_replicas(c, new, &m->r.e,
+		ret = bch2_update_replicas(c, new, &m->r.e,
 				      ((s64) m->sectors * m->nr_redundant),
 				      0, true);
 		if (ret) {
@@ -608,19 +633,21 @@ static void ec_validate_checksums(struct bch_fs *c, struct ec_stripe_buf *buf)
 			struct bch_csum got = ec_block_checksum(buf, i, offset);
 
 			if (bch2_crc_cmp(want, got)) {
-				struct printbuf err = PRINTBUF;
-				struct bch_dev *ca = bch_dev_bkey_exists(c, v->ptrs[i].dev);
+				struct bch_dev *ca = bch2_dev_tryget(c, v->ptrs[i].dev);
+				if (ca) {
+					struct printbuf err = PRINTBUF;
 
-				prt_str(&err, "stripe ");
-				bch2_csum_err_msg(&err, v->csum_type, want, got);
-				prt_printf(&err, "  for %ps at %u of\n  ", (void *) _RET_IP_, i);
-				bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
-				bch_err_ratelimited(ca, "%s", err.buf);
-				printbuf_exit(&err);
+					prt_str(&err, "stripe ");
+					bch2_csum_err_msg(&err, v->csum_type, want, got);
+					prt_printf(&err, "  for %ps at %u of\n  ", (void *) _RET_IP_, i);
+					bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
+					bch_err_ratelimited(ca, "%s", err.buf);
+					printbuf_exit(&err);
 
-				clear_bit(i, buf->valid);
+					bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
+				}
 
-				bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
+				clear_bit(i, buf->valid);
 				break;
 			}
 
@@ -687,7 +714,7 @@ static void ec_block_endio(struct bio *bio)
 			       bch2_blk_status_to_str(bio->bi_status)))
 		clear_bit(ec_bio->idx, ec_bio->buf->valid);
 
-	if (ptr_stale(ca, ptr)) {
+	if (dev_ptr_stale(ca, ptr)) {
 		bch_err_ratelimited(ca->fs,
 				    "error %s stripe: stale pointer after io",
 				    bio_data_dir(bio) == READ ? "reading from" : "writing to");
@@ -705,13 +732,13 @@ static void ec_block_io(struct bch_fs *c, struct ec_stripe_buf *buf,
 	struct bch_stripe *v = &bkey_i_to_stripe(&buf->key)->v;
 	unsigned offset = 0, bytes = buf->size << 9;
 	struct bch_extent_ptr *ptr = &v->ptrs[idx];
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+	struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 	enum bch_data_type data_type = idx < v->nr_blocks - v->nr_redundant
 		? BCH_DATA_user
 		: BCH_DATA_parity;
 	int rw = op_is_write(opf);
 
-	if (ptr_stale(ca, ptr)) {
+	if (dev_ptr_stale(ca, ptr)) {
 		bch_err_ratelimited(c,
 				    "error %s stripe: stale pointer",
 				    rw == READ ? "reading from" : "writing to");
@@ -769,7 +796,7 @@ static int get_stripe_key_trans(struct btree_trans *trans, u64 idx,
 	int ret;
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes,
-			       POS(0, idx), BTREE_ITER_SLOTS);
+			       POS(0, idx), BTREE_ITER_slots);
 	ret = bkey_err(k);
 	if (ret)
 		goto err;
@@ -1060,7 +1087,7 @@ static int ec_stripe_delete(struct btree_trans *trans, u64 idx)
 	int ret;
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes, POS(0, idx),
-			       BTREE_ITER_INTENT);
+			       BTREE_ITER_intent);
 	ret = bkey_err(k);
 	if (ret)
 		goto err;
@@ -1131,7 +1158,7 @@ static int ec_stripe_key_update(struct btree_trans *trans,
 	int ret;
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes,
-			       new->k.p, BTREE_ITER_INTENT);
+			       new->k.p, BTREE_ITER_intent);
 	ret = bkey_err(k);
 	if (ret)
 		goto err;
@@ -1173,6 +1200,7 @@ err:
 }
 
 static int ec_stripe_update_extent(struct btree_trans *trans,
+				   struct bch_dev *ca,
 				   struct bpos bucket, u8 gen,
 				   struct ec_stripe_buf *s,
 				   struct bpos *bp_pos)
@@ -1188,8 +1216,8 @@ static int ec_stripe_update_extent(struct btree_trans *trans,
 	struct bkey_i *n;
 	int ret, dev, block;
 
-	ret = bch2_get_next_backpointer(trans, bucket, gen,
-				bp_pos, &bp, BTREE_ITER_CACHED);
+	ret = bch2_get_next_backpointer(trans, ca, bucket, gen,
+				bp_pos, &bp, BTREE_ITER_cached);
 	if (ret)
 		return ret;
 	if (bpos_eq(*bp_pos, SPOS_MAX))
@@ -1214,7 +1242,7 @@ static int ec_stripe_update_extent(struct btree_trans *trans,
 		return -EIO;
 	}
 
-	k = bch2_backpointer_get_key(trans, &iter, *bp_pos, bp, BTREE_ITER_INTENT);
+	k = bch2_backpointer_get_key(trans, &iter, *bp_pos, bp, BTREE_ITER_intent);
 	ret = bkey_err(k);
 	if (ret)
 		return ret;
@@ -1272,17 +1300,21 @@ static int ec_stripe_update_bucket(struct btree_trans *trans, struct ec_stripe_b
 {
 	struct bch_fs *c = trans->c;
 	struct bch_stripe *v = &bkey_i_to_stripe(&s->key)->v;
-	struct bch_extent_ptr bucket = v->ptrs[block];
-	struct bpos bucket_pos = PTR_BUCKET_POS(c, &bucket);
+	struct bch_extent_ptr ptr = v->ptrs[block];
 	struct bpos bp_pos = POS_MIN;
 	int ret = 0;
 
+	struct bch_dev *ca = bch2_dev_tryget(c, ptr.dev);
+	if (!ca)
+		return -EIO;
+
+	struct bpos bucket_pos = PTR_BUCKET_POS(ca, &ptr);
+
 	while (1) {
 		ret = commit_do(trans, NULL, NULL,
 				BCH_TRANS_COMMIT_no_check_rw|
 				BCH_TRANS_COMMIT_no_enospc,
-			ec_stripe_update_extent(trans, bucket_pos, bucket.gen,
-						s, &bp_pos));
+			ec_stripe_update_extent(trans, ca, bucket_pos, ptr.gen, s, &bp_pos));
 		if (ret)
 			break;
 		if (bkey_eq(bp_pos, POS_MAX))
@@ -1291,6 +1323,7 @@ static int ec_stripe_update_bucket(struct btree_trans *trans, struct ec_stripe_b
 		bp_pos = bpos_nosnap_successor(bp_pos);
 	}
 
+	bch2_dev_put(ca);
 	return ret;
 }
 
@@ -1321,7 +1354,7 @@ static void zero_out_rest_of_ec_bucket(struct bch_fs *c,
 				       unsigned block,
 				       struct open_bucket *ob)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
+	struct bch_dev *ca = bch2_dev_bkey_exists(c, ob->dev);
 	unsigned offset = ca->mi.bucket_size - ob->sectors_free;
 	int ret;
 
@@ -1519,16 +1552,13 @@ void bch2_ec_bucket_cancel(struct bch_fs *c, struct open_bucket *ob)
 void *bch2_writepoint_ec_buf(struct bch_fs *c, struct write_point *wp)
 {
 	struct open_bucket *ob = ec_open_bucket(c, &wp->ptrs);
-	struct bch_dev *ca;
-	unsigned offset;
-
 	if (!ob)
 		return NULL;
 
 	BUG_ON(!ob->ec->new_stripe.data[ob->ec_idx]);
 
-	ca	= bch_dev_bkey_exists(c, ob->dev);
-	offset	= ca->mi.bucket_size - ob->sectors_free;
+	struct bch_dev *ca	= ob_dev(c, ob);
+	unsigned offset		= ca->mi.bucket_size - ob->sectors_free;
 
 	return ob->ec->new_stripe.data[ob->ec_idx] + (offset << 9);
 }
@@ -1937,7 +1967,7 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
 	}
 
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_stripes, start_pos,
-			   BTREE_ITER_SLOTS|BTREE_ITER_INTENT, k, ret) {
+			   BTREE_ITER_slots|BTREE_ITER_intent, k, ret) {
 		if (bkey_gt(k.k->p, POS(0, U32_MAX))) {
 			if (start_pos.offset) {
 				start_pos = min_pos;
@@ -2127,7 +2157,7 @@ int bch2_stripes_read(struct bch_fs *c)
 {
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_stripes, POS_MIN,
-				   BTREE_ITER_PREFETCH, k, ({
+				   BTREE_ITER_prefetch, k, ({
 			if (k.k->type != KEY_TYPE_stripe)
 				continue;
 
diff --git a/fs/bcachefs/ec.h b/fs/bcachefs/ec.h
index f042616888b0a1..a1c2bc74ef4582 100644
--- a/fs/bcachefs/ec.h
+++ b/fs/bcachefs/ec.h
@@ -13,7 +13,8 @@ int bch2_stripe_invalid(struct bch_fs *, struct bkey_s_c,
 void bch2_stripe_to_text(struct printbuf *, struct bch_fs *,
 			 struct bkey_s_c);
 int bch2_trigger_stripe(struct btree_trans *, enum btree_id, unsigned,
-			struct bkey_s_c, struct bkey_s, unsigned);
+			struct bkey_s_c, struct bkey_s,
+			enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_stripe ((struct bkey_ops) {	\
 	.key_invalid	= bch2_stripe_invalid,		\
diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c
index 82a6656c941c5f..c66eeffcd7f2a8 100644
--- a/fs/bcachefs/error.c
+++ b/fs/bcachefs/error.c
@@ -176,6 +176,21 @@ static struct fsck_err_state *fsck_err_get(struct bch_fs *c, const char *fmt)
 	return s;
 }
 
+/* s/fix?/fixing/ s/recreate?/recreating/ */
+static void prt_actioning(struct printbuf *out, const char *action)
+{
+	unsigned len = strlen(action);
+
+	BUG_ON(action[len - 1] != '?');
+	--len;
+
+	if (action[len - 1] == 'e')
+		--len;
+
+	prt_bytes(out, action, len);
+	prt_str(out, "ing");
+}
+
 int bch2_fsck_err(struct bch_fs *c,
 		  enum bch_fsck_flags flags,
 		  enum bch_sb_error_id err,
@@ -186,6 +201,7 @@ int bch2_fsck_err(struct bch_fs *c,
 	bool print = true, suppressing = false, inconsistent = false;
 	struct printbuf buf = PRINTBUF, *out = &buf;
 	int ret = -BCH_ERR_fsck_ignore;
+	const char *action_orig = "fix?", *action = action_orig;
 
 	if ((flags & FSCK_CAN_FIX) &&
 	    test_bit(err, c->sb.errors_silent))
@@ -197,6 +213,19 @@ int bch2_fsck_err(struct bch_fs *c,
 	prt_vprintf(out, fmt, args);
 	va_end(args);
 
+	/* Custom fix/continue/recreate/etc.? */
+	if (out->buf[out->pos - 1] == '?') {
+		const char *p = strrchr(out->buf, ',');
+		if (p) {
+			out->pos = p - out->buf;
+			action = kstrdup(p + 2, GFP_KERNEL);
+			if (!action) {
+				ret = -ENOMEM;
+				goto err;
+			}
+		}
+	}
+
 	mutex_lock(&c->fsck_error_msgs_lock);
 	s = fsck_err_get(c, fmt);
 	if (s) {
@@ -208,12 +237,16 @@ int bch2_fsck_err(struct bch_fs *c,
 		if (s->last_msg && !strcmp(buf.buf, s->last_msg)) {
 			ret = s->ret;
 			mutex_unlock(&c->fsck_error_msgs_lock);
-			printbuf_exit(&buf);
-			return ret;
+			goto err;
 		}
 
 		kfree(s->last_msg);
 		s->last_msg = kstrdup(buf.buf, GFP_KERNEL);
+		if (!s->last_msg) {
+			mutex_unlock(&c->fsck_error_msgs_lock);
+			ret = -ENOMEM;
+			goto err;
+		}
 
 		if (c->opts.ratelimit_errors &&
 		    !(flags & FSCK_NO_RATELIMIT) &&
@@ -239,7 +272,8 @@ int bch2_fsck_err(struct bch_fs *c,
 			inconsistent = true;
 			ret = -BCH_ERR_fsck_errors_not_fixed;
 		} else if (flags & FSCK_CAN_FIX) {
-			prt_str(out, ", fixing");
+			prt_str(out, ", ");
+			prt_actioning(out, action);
 			ret = -BCH_ERR_fsck_fix;
 		} else {
 			prt_str(out, ", continuing");
@@ -254,16 +288,16 @@ int bch2_fsck_err(struct bch_fs *c,
 			: c->opts.fix_errors;
 
 		if (fix == FSCK_FIX_ask) {
-			int ask;
+			prt_str(out, ", ");
+			prt_str(out, action);
 
-			prt_str(out, ": fix?");
 			if (bch2_fs_stdio_redirect(c))
 				bch2_print(c, "%s", out->buf);
 			else
 				bch2_print_string_as_lines(KERN_ERR, out->buf);
 			print = false;
 
-			ask = bch2_fsck_ask_yn(c);
+			int ask = bch2_fsck_ask_yn(c);
 
 			if (ask >= YN_ALLNO && s)
 				s->fix = ask == YN_ALLNO
@@ -276,10 +310,12 @@ int bch2_fsck_err(struct bch_fs *c,
 		} else if (fix == FSCK_FIX_yes ||
 			   (c->opts.nochanges &&
 			    !(flags & FSCK_CAN_IGNORE))) {
-			prt_str(out, ", fixing");
+			prt_str(out, ", ");
+			prt_actioning(out, action);
 			ret = -BCH_ERR_fsck_fix;
 		} else {
-			prt_str(out, ", not fixing");
+			prt_str(out, ", not ");
+			prt_actioning(out, action);
 		}
 	} else if (flags & FSCK_NEED_FSCK) {
 		prt_str(out, " (run fsck to correct)");
@@ -311,8 +347,6 @@ int bch2_fsck_err(struct bch_fs *c,
 
 	mutex_unlock(&c->fsck_error_msgs_lock);
 
-	printbuf_exit(&buf);
-
 	if (inconsistent)
 		bch2_inconsistent_error(c);
 
@@ -322,7 +356,10 @@ int bch2_fsck_err(struct bch_fs *c,
 		set_bit(BCH_FS_errors_not_fixed, &c->flags);
 		set_bit(BCH_FS_error, &c->flags);
 	}
-
+err:
+	if (action != action_orig)
+		kfree(action);
+	printbuf_exit(&buf);
 	return ret;
 }
 
diff --git a/fs/bcachefs/extent_update.c b/fs/bcachefs/extent_update.c
index b9033bb4f11cf3..5f4fecb358da03 100644
--- a/fs/bcachefs/extent_update.c
+++ b/fs/bcachefs/extent_update.c
@@ -72,7 +72,7 @@ static int count_iters_for_insert(struct btree_trans *trans,
 
 		for_each_btree_key_norestart(trans, iter,
 				   BTREE_ID_reflink, POS(0, idx + offset),
-				   BTREE_ITER_SLOTS, r_k, ret2) {
+				   BTREE_ITER_slots, r_k, ret2) {
 			if (bkey_ge(bkey_start_pos(r_k.k), POS(0, idx + sectors)))
 				break;
 
diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c
index 1a331e53920485..39036f179f64f0 100644
--- a/fs/bcachefs/extents.c
+++ b/fs/bcachefs/extents.c
@@ -71,6 +71,12 @@ void bch2_mark_io_failure(struct bch_io_failures *failed,
 	}
 }
 
+static inline u64 dev_latency(struct bch_fs *c, unsigned dev)
+{
+	struct bch_dev *ca = bch2_dev_rcu(c, dev);
+	return ca ? atomic64_read(&ca->cur_latency[READ]) : S64_MAX;
+}
+
 /*
  * returns true if p1 is better than p2:
  */
@@ -79,11 +85,10 @@ static inline bool ptr_better(struct bch_fs *c,
 			      const struct extent_ptr_decoded p2)
 {
 	if (likely(!p1.idx && !p2.idx)) {
-		struct bch_dev *dev1 = bch_dev_bkey_exists(c, p1.ptr.dev);
-		struct bch_dev *dev2 = bch_dev_bkey_exists(c, p2.ptr.dev);
-
-		u64 l1 = atomic64_read(&dev1->cur_latency[READ]);
-		u64 l2 = atomic64_read(&dev2->cur_latency[READ]);
+		rcu_read_lock();
+		u64 l1 = dev_latency(c, p1.ptr.dev);
+		u64 l2 = dev_latency(c, p2.ptr.dev);
+		rcu_read_unlock();
 
 		/* Pick at random, biased in favor of the faster device: */
 
@@ -123,7 +128,7 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
 		if (p.ptr.unwritten)
 			return 0;
 
-		ca = bch_dev_bkey_exists(c, p.ptr.dev);
+		ca = bch2_dev_bkey_exists(c, p.ptr.dev);
 
 		/*
 		 * If there are any dirty pointers it's an error if we can't
@@ -132,7 +137,7 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
 		if (!ret && !p.ptr.cached)
 			ret = -EIO;
 
-		if (p.ptr.cached && ptr_stale(ca, &p.ptr))
+		if (p.ptr.cached && dev_ptr_stale(ca, &p.ptr))
 			continue;
 
 		f = failed ? dev_io_failures(failed, p.ptr.dev) : NULL;
@@ -201,6 +206,11 @@ int bch2_btree_ptr_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
 			 c, err, btree_ptr_v2_min_key_bad,
 			 "min_key > key");
 
+	if (flags & BKEY_INVALID_WRITE)
+		bkey_fsck_err_on(!bp.v->sectors_written,
+				 c, err, btree_ptr_v2_written_0,
+				 "sectors_written == 0");
+
 	ret = bch2_bkey_ptrs_invalid(c, k, flags, err);
 fsck_err:
 	return ret;
@@ -247,7 +257,6 @@ bool bch2_extent_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
 	const union bch_extent_entry *en_r;
 	struct extent_ptr_decoded lp, rp;
 	bool use_right_ptr;
-	struct bch_dev *ca;
 
 	en_l = l_ptrs.start;
 	en_r = r_ptrs.start;
@@ -278,8 +287,12 @@ bool bch2_extent_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
 			return false;
 
 		/* Extents may not straddle buckets: */
-		ca = bch_dev_bkey_exists(c, lp.ptr.dev);
-		if (PTR_BUCKET_NR(ca, &lp.ptr) != PTR_BUCKET_NR(ca, &rp.ptr))
+		rcu_read_lock();
+		struct bch_dev *ca = bch2_dev_rcu(c, lp.ptr.dev);
+		bool same_bucket = ca && PTR_BUCKET_NR(ca, &lp.ptr) == PTR_BUCKET_NR(ca, &rp.ptr);
+		rcu_read_unlock();
+
+		if (!same_bucket)
 			return false;
 
 		if (lp.has_ec			!= rp.has_ec ||
@@ -667,16 +680,16 @@ static inline unsigned __extent_ptr_durability(struct bch_dev *ca, struct extent
 
 unsigned bch2_extent_ptr_desired_durability(struct bch_fs *c, struct extent_ptr_decoded *p)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, p->ptr.dev);
+	struct bch_dev *ca = bch2_dev_rcu(c, p->ptr.dev);
 
-	return __extent_ptr_durability(ca, p);
+	return ca ? __extent_ptr_durability(ca, p) : 0;
 }
 
 unsigned bch2_extent_ptr_durability(struct bch_fs *c, struct extent_ptr_decoded *p)
 {
-	struct bch_dev *ca = bch_dev_bkey_exists(c, p->ptr.dev);
+	struct bch_dev *ca = bch2_dev_rcu(c, p->ptr.dev);
 
-	if (ca->mi.state == BCH_MEMBER_STATE_failed)
+	if (!ca || ca->mi.state == BCH_MEMBER_STATE_failed)
 		return 0;
 
 	return __extent_ptr_durability(ca, p);
@@ -689,8 +702,10 @@ unsigned bch2_bkey_durability(struct bch_fs *c, struct bkey_s_c k)
 	struct extent_ptr_decoded p;
 	unsigned durability = 0;
 
+	rcu_read_lock();
 	bkey_for_each_ptr_decode(k.k, ptrs, p, entry)
 		durability += bch2_extent_ptr_durability(c, &p);
+	rcu_read_unlock();
 
 	return durability;
 }
@@ -702,9 +717,11 @@ static unsigned bch2_bkey_durability_safe(struct bch_fs *c, struct bkey_s_c k)
 	struct extent_ptr_decoded p;
 	unsigned durability = 0;
 
+	rcu_read_lock();
 	bkey_for_each_ptr_decode(k.k, ptrs, p, entry)
 		if (p.ptr.dev < c->sb.nr_devices && c->devs[p.ptr.dev])
 			durability += bch2_extent_ptr_durability(c, &p);
+	rcu_read_unlock();
 
 	return durability;
 }
@@ -860,14 +877,21 @@ const struct bch_extent_ptr *bch2_bkey_has_device_c(struct bkey_s_c k, unsigned
 bool bch2_bkey_has_target(struct bch_fs *c, struct bkey_s_c k, unsigned target)
 {
 	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+	struct bch_dev *ca;
+	bool ret = false;
 
+	rcu_read_lock();
 	bkey_for_each_ptr(ptrs, ptr)
 		if (bch2_dev_in_target(c, ptr->dev, target) &&
+		    (ca = bch2_dev_rcu(c, ptr->dev)) &&
 		    (!ptr->cached ||
-		     !ptr_stale(bch_dev_bkey_exists(c, ptr->dev), ptr)))
-			return true;
+		     !dev_ptr_stale_rcu(ca, ptr))) {
+			ret = true;
+			break;
+		}
+	rcu_read_unlock();
 
-	return false;
+	return ret;
 }
 
 bool bch2_bkey_matches_ptr(struct bch_fs *c, struct bkey_s_c k,
@@ -970,20 +994,23 @@ void bch2_extent_ptr_set_cached(struct bkey_s k, struct bch_extent_ptr *ptr)
 bool bch2_extent_normalize(struct bch_fs *c, struct bkey_s k)
 {
 	struct bch_extent_ptr *ptr;
+	struct bch_dev *ca;
 
+	rcu_read_lock();
 	bch2_bkey_drop_ptrs(k, ptr,
 		ptr->cached &&
-		ptr_stale(bch_dev_bkey_exists(c, ptr->dev), ptr));
+		(ca = bch2_dev_rcu(c, ptr->dev)) &&
+		dev_ptr_stale_rcu(ca, ptr));
+	rcu_read_unlock();
 
 	return bkey_deleted(k.k);
 }
 
 void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *c, const struct bch_extent_ptr *ptr)
 {
-	struct bch_dev *ca = c && ptr->dev < c->sb.nr_devices && c->devs[ptr->dev]
-		? bch_dev_bkey_exists(c, ptr->dev)
-		: NULL;
-
+	out->atomic++;
+	rcu_read_lock();
+	struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
 	if (!ca) {
 		prt_printf(out, "ptr: %u:%llu gen %u%s", ptr->dev,
 			   (u64) ptr->offset, ptr->gen,
@@ -998,11 +1025,11 @@ void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *c, const struc
 			prt_str(out, " cached");
 		if (ptr->unwritten)
 			prt_str(out, " unwritten");
-		if (b >= ca->mi.first_bucket &&
-		    b <  ca->mi.nbuckets &&
-		    ptr_stale(ca, ptr))
+		if (bucket_valid(ca, b) && dev_ptr_stale_rcu(ca, ptr))
 			prt_printf(out, " stale");
 	}
+	rcu_read_unlock();
+	--out->atomic;
 }
 
 void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
@@ -1076,12 +1103,9 @@ static int extent_ptr_invalid(struct bch_fs *c,
 			      struct printbuf *err)
 {
 	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
-	u64 bucket;
-	u32 bucket_offset;
-	struct bch_dev *ca;
 	int ret = 0;
 
-	if (!bch2_dev_exists2(c, ptr->dev)) {
+	if (!bch2_dev_exists(c, ptr->dev)) {
 		/*
 		 * If we're in the write path this key might have already been
 		 * overwritten, and we could be seeing a device that doesn't
@@ -1094,24 +1118,35 @@ static int extent_ptr_invalid(struct bch_fs *c,
 			   "pointer to invalid device (%u)", ptr->dev);
 	}
 
-	ca = bch_dev_bkey_exists(c, ptr->dev);
+	rcu_read_lock();
+	struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
+	if (!ca) {
+		rcu_read_unlock();
+		return 0;
+	}
+	u32 bucket_offset;
+	u64 bucket = sector_to_bucket_and_offset(ca, ptr->offset, &bucket_offset);
+	unsigned first_bucket	= ca->mi.first_bucket;
+	u64 nbuckets		= ca->mi.nbuckets;
+	unsigned bucket_size	= ca->mi.bucket_size;
+	rcu_read_unlock();
+
 	bkey_for_each_ptr(ptrs, ptr2)
 		bkey_fsck_err_on(ptr != ptr2 && ptr->dev == ptr2->dev, c, err,
 				 ptr_to_duplicate_device,
 				 "multiple pointers to same device (%u)", ptr->dev);
 
-	bucket = sector_to_bucket_and_offset(ca, ptr->offset, &bucket_offset);
 
-	bkey_fsck_err_on(bucket >= ca->mi.nbuckets, c, err,
+	bkey_fsck_err_on(bucket >= nbuckets, c, err,
 			 ptr_after_last_bucket,
-			 "pointer past last bucket (%llu > %llu)", bucket, ca->mi.nbuckets);
-	bkey_fsck_err_on(ptr->offset < bucket_to_sector(ca, ca->mi.first_bucket), c, err,
+			 "pointer past last bucket (%llu > %llu)", bucket, nbuckets);
+	bkey_fsck_err_on(bucket < first_bucket, c, err,
 			 ptr_before_first_bucket,
-			 "pointer before first bucket (%llu < %u)", bucket, ca->mi.first_bucket);
-	bkey_fsck_err_on(bucket_offset + size_ondisk > ca->mi.bucket_size, c, err,
+			 "pointer before first bucket (%llu < %u)", bucket, first_bucket);
+	bkey_fsck_err_on(bucket_offset + size_ondisk > bucket_size, c, err,
 			 ptr_spans_multiple_buckets,
 			 "pointer spans multiple buckets (%u + %u > %u)",
-		       bucket_offset, size_ondisk, ca->mi.bucket_size);
+		       bucket_offset, size_ondisk, bucket_size);
 fsck_err:
 	return ret;
 }
diff --git a/fs/bcachefs/eytzinger.c b/fs/bcachefs/eytzinger.c
index 0f955c3c76a7bc..2eaffe37b5e76a 100644
--- a/fs/bcachefs/eytzinger.c
+++ b/fs/bcachefs/eytzinger.c
@@ -171,7 +171,7 @@ void eytzinger0_sort_r(void *base, size_t n, size_t size,
 		       swap_r_func_t swap_func,
 		       const void *priv)
 {
-	int i, c, r;
+	int i, j, k;
 
 	/* called from 'sort' without swap function, let's pick the default */
 	if (swap_func == SWAP_WRAPPER && !((struct wrapper *)priv)->swap_func)
@@ -188,17 +188,22 @@ void eytzinger0_sort_r(void *base, size_t n, size_t size,
 
 	/* heapify */
 	for (i = n / 2 - 1; i >= 0; --i) {
-		for (r = i; r * 2 + 1 < n; r = c) {
-			c = r * 2 + 1;
+		/* Find the sift-down path all the way to the leaves. */
+		for (j = i; k = j * 2 + 1, k + 1 < n;)
+			j = eytzinger0_do_cmp(base, n, size, cmp_func, priv, k, k + 1) > 0 ? k : k + 1;
 
-			if (c + 1 < n &&
-			    eytzinger0_do_cmp(base, n, size, cmp_func, priv, c, c + 1) < 0)
-				c++;
+		/* Special case for the last leaf with no sibling. */
+		if (j * 2 + 2 == n)
+			j = j * 2 + 1;
 
-			if (eytzinger0_do_cmp(base, n, size, cmp_func, priv, r, c) >= 0)
-				break;
+		/* Backtrack to the correct location. */
+		while (j != i && eytzinger0_do_cmp(base, n, size, cmp_func, priv, i, j) >= 0)
+			j = (j - 1) / 2;
 
-			eytzinger0_do_swap(base, n, size, swap_func, priv, r, c);
+		/* Shift the element into its correct place. */
+		for (k = j; j != i;) {
+			j = (j - 1) / 2;
+			eytzinger0_do_swap(base, n, size, swap_func, priv, j, k);
 		}
 	}
 
@@ -206,17 +211,22 @@ void eytzinger0_sort_r(void *base, size_t n, size_t size,
 	for (i = n - 1; i > 0; --i) {
 		eytzinger0_do_swap(base, n, size, swap_func, priv, 0, i);
 
-		for (r = 0; r * 2 + 1 < i; r = c) {
-			c = r * 2 + 1;
+		/* Find the sift-down path all the way to the leaves. */
+		for (j = 0; k = j * 2 + 1, k + 1 < i;)
+			j = eytzinger0_do_cmp(base, n, size, cmp_func, priv, k, k + 1) > 0 ? k : k + 1;
 
-			if (c + 1 < i &&
-			    eytzinger0_do_cmp(base, n, size, cmp_func, priv, c, c + 1) < 0)
-				c++;
+		/* Special case for the last leaf with no sibling. */
+		if (j * 2 + 2 == i)
+			j = j * 2 + 1;
 
-			if (eytzinger0_do_cmp(base, n, size, cmp_func, priv, r, c) >= 0)
-				break;
+		/* Backtrack to the correct location. */
+		while (j && eytzinger0_do_cmp(base, n, size, cmp_func, priv, 0, j) >= 0)
+			j = (j - 1) / 2;
 
-			eytzinger0_do_swap(base, n, size, swap_func, priv, r, c);
+		/* Shift the element into its correct place. */
+		for (k = j; j;) {
+			j = (j - 1) / 2;
+			eytzinger0_do_swap(base, n, size, swap_func, priv, j, k);
 		}
 	}
 }
@@ -232,3 +242,64 @@ void eytzinger0_sort(void *base, size_t n, size_t size,
 
 	return eytzinger0_sort_r(base, n, size, _CMP_WRAPPER, SWAP_WRAPPER, &w);
 }
+
+#if 0
+#include <linux/slab.h>
+#include <linux/random.h>
+#include <linux/ktime.h>
+
+static u64 cmp_count;
+
+static int mycmp(const void *a, const void *b)
+{
+	u32 _a = *(u32 *)a;
+	u32 _b = *(u32 *)b;
+
+	cmp_count++;
+	if (_a < _b)
+		return -1;
+	else if (_a > _b)
+		return 1;
+	else
+		return 0;
+}
+
+static int test(void)
+{
+	size_t N, i;
+	ktime_t start, end;
+	s64 delta;
+	u32 *arr;
+
+	for (N = 10000; N <= 100000; N += 10000) {
+		arr = kmalloc_array(N, sizeof(u32), GFP_KERNEL);
+		cmp_count = 0;
+
+		for (i = 0; i < N; i++)
+			arr[i] = get_random_u32();
+
+		start = ktime_get();
+		eytzinger0_sort(arr, N, sizeof(u32), mycmp, NULL);
+		end = ktime_get();
+
+		delta = ktime_us_delta(end, start);
+		printk(KERN_INFO "time: %lld\n", delta);
+		printk(KERN_INFO "comparisons: %lld\n", cmp_count);
+
+		u32 prev = 0;
+
+		eytzinger0_for_each(i, N) {
+			if (prev > arr[i])
+				goto err;
+			prev = arr[i];
+		}
+
+		kfree(arr);
+	}
+	return 0;
+
+err:
+	kfree(arr);
+	return -1;
+}
+#endif
diff --git a/fs/bcachefs/fs-common.c b/fs/bcachefs/fs-common.c
index 624e6f963240f8..508d029ac53d06 100644
--- a/fs/bcachefs/fs-common.c
+++ b/fs/bcachefs/fs-common.c
@@ -42,7 +42,7 @@ int bch2_create_trans(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_INTENT);
+	ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_intent);
 	if (ret)
 		goto err;
 
@@ -70,7 +70,7 @@ int bch2_create_trans(struct btree_trans *trans,
 			struct bch_subvolume s;
 
 			ret = bch2_subvolume_get(trans, snapshot_src.subvol, true,
-						 BTREE_ITER_CACHED, &s);
+						 BTREE_ITER_cached, &s);
 			if (ret)
 				goto err;
 
@@ -78,7 +78,7 @@ int bch2_create_trans(struct btree_trans *trans,
 		}
 
 		ret = bch2_inode_peek(trans, &inode_iter, new_inode, snapshot_src,
-				      BTREE_ITER_INTENT);
+				      BTREE_ITER_intent);
 		if (ret)
 			goto err;
 
@@ -163,7 +163,7 @@ int bch2_create_trans(struct btree_trans *trans,
 					 name,
 					 dir_target,
 					 &dir_offset,
-					 BCH_HASH_SET_MUST_CREATE);
+					 STR_HASH_must_create);
 		if (ret)
 			goto err;
 
@@ -171,7 +171,7 @@ int bch2_create_trans(struct btree_trans *trans,
 		new_inode->bi_dir_offset	= dir_offset;
 	}
 
-	inode_iter.flags &= ~BTREE_ITER_ALL_SNAPSHOTS;
+	inode_iter.flags &= ~BTREE_ITER_all_snapshots;
 	bch2_btree_iter_set_snapshot(&inode_iter, snapshot);
 
 	ret   = bch2_btree_iter_traverse(&inode_iter) ?:
@@ -198,16 +198,16 @@ int bch2_link_trans(struct btree_trans *trans,
 	if (dir.subvol != inum.subvol)
 		return -EXDEV;
 
-	ret = bch2_inode_peek(trans, &inode_iter, inode_u, inum, BTREE_ITER_INTENT);
+	ret = bch2_inode_peek(trans, &inode_iter, inode_u, inum, BTREE_ITER_intent);
 	if (ret)
-		goto err;
+		return ret;
 
 	inode_u->bi_ctime = now;
 	ret = bch2_inode_nlink_inc(inode_u);
 	if (ret)
-		return ret;
+		goto err;
 
-	ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_INTENT);
+	ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_intent);
 	if (ret)
 		goto err;
 
@@ -223,7 +223,7 @@ int bch2_link_trans(struct btree_trans *trans,
 	ret = bch2_dirent_create(trans, dir, &dir_hash,
 				 mode_to_type(inode_u->bi_mode),
 				 name, inum.inum, &dir_offset,
-				 BCH_HASH_SET_MUST_CREATE);
+				 STR_HASH_must_create);
 	if (ret)
 		goto err;
 
@@ -255,19 +255,19 @@ int bch2_unlink_trans(struct btree_trans *trans,
 	struct bkey_s_c k;
 	int ret;
 
-	ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_INTENT);
+	ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_intent);
 	if (ret)
 		goto err;
 
 	dir_hash = bch2_hash_info_init(c, dir_u);
 
 	ret = bch2_dirent_lookup_trans(trans, &dirent_iter, dir, &dir_hash,
-				       name, &inum, BTREE_ITER_INTENT);
+				       name, &inum, BTREE_ITER_intent);
 	if (ret)
 		goto err;
 
 	ret = bch2_inode_peek(trans, &inode_iter, inode_u, inum,
-			      BTREE_ITER_INTENT);
+			      BTREE_ITER_intent);
 	if (ret)
 		goto err;
 
@@ -322,7 +322,7 @@ int bch2_unlink_trans(struct btree_trans *trans,
 
 	ret =   bch2_hash_delete_at(trans, bch2_dirent_hash_desc,
 				    &dir_hash, &dirent_iter,
-				    BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+				    BTREE_UPDATE_internal_snapshot_node) ?:
 		bch2_inode_write(trans, &dir_iter, dir_u) ?:
 		bch2_inode_write(trans, &inode_iter, inode_u);
 err:
@@ -363,7 +363,7 @@ static int subvol_update_parent(struct btree_trans *trans, u32 subvol, u32 new_p
 	struct bkey_i_subvolume *s =
 		bch2_bkey_get_mut_typed(trans, &iter,
 			BTREE_ID_subvolumes, POS(0, subvol),
-			BTREE_ITER_CACHED, subvolume);
+			BTREE_ITER_cached, subvolume);
 	int ret = PTR_ERR_OR_ZERO(s);
 	if (ret)
 		return ret;
@@ -394,7 +394,7 @@ int bch2_rename_trans(struct btree_trans *trans,
 	int ret;
 
 	ret = bch2_inode_peek(trans, &src_dir_iter, src_dir_u, src_dir,
-			      BTREE_ITER_INTENT);
+			      BTREE_ITER_intent);
 	if (ret)
 		goto err;
 
@@ -403,7 +403,7 @@ int bch2_rename_trans(struct btree_trans *trans,
 	if (dst_dir.inum	!= src_dir.inum ||
 	    dst_dir.subvol	!= src_dir.subvol) {
 		ret = bch2_inode_peek(trans, &dst_dir_iter, dst_dir_u, dst_dir,
-				      BTREE_ITER_INTENT);
+				      BTREE_ITER_intent);
 		if (ret)
 			goto err;
 
@@ -423,13 +423,13 @@ int bch2_rename_trans(struct btree_trans *trans,
 		goto err;
 
 	ret = bch2_inode_peek(trans, &src_inode_iter, src_inode_u, src_inum,
-			      BTREE_ITER_INTENT);
+			      BTREE_ITER_intent);
 	if (ret)
 		goto err;
 
 	if (dst_inum.inum) {
 		ret = bch2_inode_peek(trans, &dst_inode_iter, dst_inode_u, dst_inum,
-				      BTREE_ITER_INTENT);
+				      BTREE_ITER_intent);
 		if (ret)
 			goto err;
 	}
diff --git a/fs/bcachefs/fs-io-buffered.c b/fs/bcachefs/fs-io-buffered.c
index 39292e7ef342c5..b0a33fabadf81c 100644
--- a/fs/bcachefs/fs-io-buffered.c
+++ b/fs/bcachefs/fs-io-buffered.c
@@ -30,15 +30,8 @@ static void bch2_readpages_end_io(struct bio *bio)
 {
 	struct folio_iter fi;
 
-	bio_for_each_folio_all(fi, bio) {
-		if (!bio->bi_status) {
-			folio_mark_uptodate(fi.folio);
-		} else {
-			folio_clear_uptodate(fi.folio);
-			folio_set_error(fi.folio);
-		}
-		folio_unlock(fi.folio);
-	}
+	bio_for_each_folio_all(fi, bio)
+		folio_end_read(fi.folio, bio->bi_status == BLK_STS_OK);
 
 	bio_put(bio);
 }
@@ -176,7 +169,7 @@ retry:
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     SPOS(inum.inum, rbio->bio.bi_iter.bi_sector, snapshot),
-			     BTREE_ITER_SLOTS);
+			     BTREE_ITER_slots);
 	while (1) {
 		struct bkey_s_c k;
 		unsigned bytes, sectors, offset_into_extent;
@@ -408,7 +401,6 @@ static void bch2_writepage_io_done(struct bch_write_op *op)
 		bio_for_each_folio_all(fi, bio) {
 			struct bch_folio *s;
 
-			folio_set_error(fi.folio);
 			mapping_set_error(fi.folio->mapping, -EIO);
 
 			s = __bch2_folio(fi.folio);
diff --git a/fs/bcachefs/fs-io-direct.c b/fs/bcachefs/fs-io-direct.c
index b889370a508811..09d21aef879af8 100644
--- a/fs/bcachefs/fs-io-direct.c
+++ b/fs/bcachefs/fs-io-direct.c
@@ -254,7 +254,7 @@ retry:
 
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_extents,
 			   SPOS(inum.inum, offset, snapshot),
-			   BTREE_ITER_SLOTS, k, err) {
+			   BTREE_ITER_slots, k, err) {
 		if (bkey_ge(bkey_start_pos(k.k), POS(inum.inum, end)))
 			break;
 
diff --git a/fs/bcachefs/fs-io-pagecache.c b/fs/bcachefs/fs-io-pagecache.c
index d359aa9b33b828..872283e5bd1e22 100644
--- a/fs/bcachefs/fs-io-pagecache.c
+++ b/fs/bcachefs/fs-io-pagecache.c
@@ -214,7 +214,7 @@ retry:
 
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_extents,
 			   SPOS(inum.inum, offset, snapshot),
-			   BTREE_ITER_SLOTS, k, ret) {
+			   BTREE_ITER_slots, k, ret) {
 		unsigned nr_ptrs = bch2_bkey_nr_ptrs_fully_allocated(k);
 		unsigned state = bkey_to_sector_state(k);
 
diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c
index 20b40477425f49..442bcb0793c412 100644
--- a/fs/bcachefs/fs-io.c
+++ b/fs/bcachefs/fs-io.c
@@ -594,7 +594,7 @@ static int __bchfs_fallocate(struct bch_inode_info *inode, int mode,
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			POS(inode->v.i_ino, start_sector),
-			BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
+			BTREE_ITER_slots|BTREE_ITER_intent);
 
 	while (!ret && bkey_lt(iter.pos, end_pos)) {
 		s64 i_sectors_delta = 0;
@@ -1009,7 +1009,7 @@ retry:
 
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_extents,
 			   SPOS(inode->v.i_ino, offset >> 9, snapshot),
-			   BTREE_ITER_SLOTS, k, ret) {
+			   BTREE_ITER_slots, k, ret) {
 		if (k.k->p.inode != inode->v.i_ino) {
 			next_hole = bch2_seek_pagecache_hole(&inode->v,
 					offset, MAX_LFS_FILESIZE, 0, false);
diff --git a/fs/bcachefs/fs-ioctl.c b/fs/bcachefs/fs-ioctl.c
index 3dc8630ff9fe13..205a323ffc6ddd 100644
--- a/fs/bcachefs/fs-ioctl.c
+++ b/fs/bcachefs/fs-ioctl.c
@@ -548,7 +548,7 @@ long bch2_compat_fs_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 {
 	/* These are just misnamed, they actually get/put from/to user an int */
 	switch (cmd) {
-	case FS_IOC_GETFLAGS:
+	case FS_IOC32_GETFLAGS:
 		cmd = FS_IOC_GETFLAGS;
 		break;
 	case FS_IOC32_SETFLAGS:
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index fce690007edfce..4e4442bc71e364 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -90,7 +90,7 @@ retry:
 	bch2_trans_begin(trans);
 
 	ret   = bch2_inode_peek(trans, &iter, &inode_u, inode_inum(inode),
-				BTREE_ITER_INTENT) ?:
+				BTREE_ITER_intent) ?:
 		(set ? set(trans, inode, &inode_u, p) : 0) ?:
 		bch2_inode_write(trans, &iter, &inode_u) ?:
 		bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
@@ -213,19 +213,43 @@ static struct bch_inode_info *bch2_inode_insert(struct bch_fs *c, struct bch_ino
 	_ret;									\
 })
 
+static struct inode *bch2_alloc_inode(struct super_block *sb)
+{
+	BUG();
+}
+
+static struct bch_inode_info *__bch2_new_inode(struct bch_fs *c)
+{
+	struct bch_inode_info *inode = kmem_cache_alloc(bch2_inode_cache, GFP_NOFS);
+	if (!inode)
+		return NULL;
+
+	inode_init_once(&inode->v);
+	mutex_init(&inode->ei_update_lock);
+	two_state_lock_init(&inode->ei_pagecache_lock);
+	INIT_LIST_HEAD(&inode->ei_vfs_inode_list);
+	mutex_init(&inode->ei_quota_lock);
+	inode->v.i_state = 0;
+
+	if (unlikely(inode_init_always(c->vfs_sb, &inode->v))) {
+		kmem_cache_free(bch2_inode_cache, inode);
+		return NULL;
+	}
+
+	return inode;
+}
+
 /*
  * Allocate a new inode, dropping/retaking btree locks if necessary:
  */
 static struct bch_inode_info *bch2_new_inode(struct btree_trans *trans)
 {
-	struct bch_fs *c = trans->c;
-
 	struct bch_inode_info *inode =
 		memalloc_flags_do(PF_MEMALLOC_NORECLAIM|PF_MEMALLOC_NOWARN,
-				  to_bch_ei(new_inode(c->vfs_sb)));
+				  __bch2_new_inode(trans->c));
 
 	if (unlikely(!inode)) {
-		int ret = drop_locks_do(trans, (inode = to_bch_ei(new_inode(c->vfs_sb))) ? 0 : -ENOMEM);
+		int ret = drop_locks_do(trans, (inode = __bch2_new_inode(trans->c)) ? 0 : -ENOMEM);
 		if (ret && inode) {
 			__destroy_inode(&inode->v);
 			kmem_cache_free(bch2_inode_cache, inode);
@@ -290,7 +314,7 @@ __bch2_create(struct mnt_idmap *idmap,
 	if (ret)
 		return ERR_PTR(ret);
 #endif
-	inode = to_bch_ei(new_inode(c->vfs_sb));
+	inode = __bch2_new_inode(c);
 	if (unlikely(!inode)) {
 		inode = ERR_PTR(-ENOMEM);
 		goto err;
@@ -323,7 +347,7 @@ retry:
 	inum.inum = inode_u.bi_inum;
 
 	ret   = bch2_subvolume_get(trans, inum.subvol, true,
-				   BTREE_ITER_WITH_UPDATES, &subvol) ?:
+				   BTREE_ITER_with_updates, &subvol) ?:
 		bch2_trans_commit(trans, NULL, &journal_seq, 0);
 	if (unlikely(ret)) {
 		bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1,
@@ -376,17 +400,14 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
 	struct bch_fs *c = trans->c;
 	struct btree_iter dirent_iter = {};
 	subvol_inum inum = {};
+	struct printbuf buf = PRINTBUF;
 
-	int ret = bch2_hash_lookup(trans, &dirent_iter, bch2_dirent_hash_desc,
-				   dir_hash_info, dir, name, 0);
+	struct bkey_s_c k = bch2_hash_lookup(trans, &dirent_iter, bch2_dirent_hash_desc,
+					     dir_hash_info, dir, name, 0);
+	int ret = bkey_err(k);
 	if (ret)
 		return ERR_PTR(ret);
 
-	struct bkey_s_c k = bch2_btree_iter_peek_slot(&dirent_iter);
-	ret = bkey_err(k);
-	if (ret)
-		goto err;
-
 	ret = bch2_dirent_read_target(trans, dir, bkey_s_c_to_dirent(k), &inum);
 	if (ret > 0)
 		ret = -ENOENT;
@@ -406,20 +427,31 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
 	ret =   bch2_subvolume_get(trans, inum.subvol, true, 0, &subvol) ?:
 		bch2_inode_find_by_inum_nowarn_trans(trans, inum, &inode_u) ?:
 		PTR_ERR_OR_ZERO(inode = bch2_new_inode(trans));
-	if (bch2_err_matches(ret, ENOENT)) {
-		struct printbuf buf = PRINTBUF;
 
-		bch2_bkey_val_to_text(&buf, c, k);
-		bch_err(c, "%s points to missing inode", buf.buf);
-		printbuf_exit(&buf);
-	}
+	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT),
+				c, "dirent to missing inode:\n  %s",
+				(bch2_bkey_val_to_text(&buf, c, k), buf.buf));
 	if (ret)
 		goto err;
 
+	/* regular files may have hardlinks: */
+	if (bch2_fs_inconsistent_on(bch2_inode_should_have_bp(&inode_u) &&
+				    !bkey_eq(k.k->p, POS(inode_u.bi_dir, inode_u.bi_dir_offset)),
+				    c,
+				    "dirent points to inode that does not point back:\n  %s",
+				    (bch2_bkey_val_to_text(&buf, c, k),
+				     prt_printf(&buf, "\n  "),
+				     bch2_inode_unpacked_to_text(&buf, &inode_u),
+				     buf.buf))) {
+		ret = -ENOENT;
+		goto err;
+	}
+
 	bch2_vfs_inode_init(trans, inum, inode, &inode_u, &subvol);
 	inode = bch2_inode_insert(c, inode);
 out:
 	bch2_trans_iter_exit(trans, &dirent_iter);
+	printbuf_exit(&buf);
 	return inode;
 err:
 	inode = ERR_PTR(ret);
@@ -787,7 +819,7 @@ retry:
 	acl = NULL;
 
 	ret = bch2_inode_peek(trans, &inode_iter, &inode_u, inode_inum(inode),
-			      BTREE_ITER_INTENT);
+			      BTREE_ITER_intent);
 	if (ret)
 		goto btree_err;
 
@@ -1040,6 +1072,10 @@ retry:
 
 		bch2_btree_iter_set_pos(&iter,
 			POS(iter.pos.inode, iter.pos.offset + sectors));
+
+		ret = bch2_trans_relock(trans);
+		if (ret)
+			break;
 	}
 	start = iter.pos.offset;
 	bch2_trans_iter_exit(trans, &iter);
@@ -1487,34 +1523,9 @@ static void bch2_vfs_inode_init(struct btree_trans *trans, subvol_inum inum,
 	mapping_set_large_folios(inode->v.i_mapping);
 }
 
-static struct inode *bch2_alloc_inode(struct super_block *sb)
-{
-	struct bch_inode_info *inode;
-
-	inode = kmem_cache_alloc(bch2_inode_cache, GFP_NOFS);
-	if (!inode)
-		return NULL;
-
-	inode_init_once(&inode->v);
-	mutex_init(&inode->ei_update_lock);
-	two_state_lock_init(&inode->ei_pagecache_lock);
-	INIT_LIST_HEAD(&inode->ei_vfs_inode_list);
-	mutex_init(&inode->ei_quota_lock);
-
-	return &inode->v;
-}
-
-static void bch2_i_callback(struct rcu_head *head)
-{
-	struct inode *vinode = container_of(head, struct inode, i_rcu);
-	struct bch_inode_info *inode = to_bch_ei(vinode);
-
-	kmem_cache_free(bch2_inode_cache, inode);
-}
-
-static void bch2_destroy_inode(struct inode *vinode)
+static void bch2_free_inode(struct inode *vinode)
 {
-	call_rcu(&vinode->i_rcu, bch2_i_callback);
+	kmem_cache_free(bch2_inode_cache, to_bch_ei(vinode));
 }
 
 static int inode_update_times_fn(struct btree_trans *trans,
@@ -1822,7 +1833,7 @@ static int bch2_unfreeze(struct super_block *sb)
 
 static const struct super_operations bch_super_operations = {
 	.alloc_inode	= bch2_alloc_inode,
-	.destroy_inode	= bch2_destroy_inode,
+	.free_inode	= bch2_free_inode,
 	.write_inode	= bch2_vfs_write_inode,
 	.evict_inode	= bch2_evict_inode,
 	.sync_fs	= bch2_sync_fs,
diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c
index 8e2010212cc371..c8f57465131c54 100644
--- a/fs/bcachefs/fsck.c
+++ b/fs/bcachefs/fsck.c
@@ -79,7 +79,7 @@ static int lookup_first_inode(struct btree_trans *trans, u64 inode_nr,
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_inodes,
 			     POS(0, inode_nr),
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_all_snapshots);
 	k = bch2_btree_iter_peek(&iter);
 	ret = bkey_err(k);
 	if (ret)
@@ -127,13 +127,13 @@ static int lookup_dirent_in_snapshot(struct btree_trans *trans,
 			   u64 *target, unsigned *type, u32 snapshot)
 {
 	struct btree_iter iter;
-	struct bkey_s_c_dirent d;
-	int ret = bch2_hash_lookup_in_snapshot(trans, &iter, bch2_dirent_hash_desc,
-			       &hash_info, dir, name, 0, snapshot);
+	struct bkey_s_c k = bch2_hash_lookup_in_snapshot(trans, &iter, bch2_dirent_hash_desc,
+							 &hash_info, dir, name, 0, snapshot);
+	int ret = bkey_err(k);
 	if (ret)
 		return ret;
 
-	d = bkey_s_c_to_dirent(bch2_btree_iter_peek_slot(&iter));
+	struct bkey_s_c_dirent d = bkey_s_c_to_dirent(bch2_btree_iter_peek_slot(&iter));
 	*target = le64_to_cpu(d.v->d_inum);
 	*type = d.v->d_type;
 	bch2_trans_iter_exit(trans, &iter);
@@ -154,12 +154,12 @@ static int __remove_dirent(struct btree_trans *trans, struct bpos pos)
 
 	dir_hash_info = bch2_hash_info_init(c, &dir_inode);
 
-	bch2_trans_iter_init(trans, &iter, BTREE_ID_dirents, pos, BTREE_ITER_INTENT);
+	bch2_trans_iter_init(trans, &iter, BTREE_ID_dirents, pos, BTREE_ITER_intent);
 
 	ret =   bch2_btree_iter_traverse(&iter) ?:
 		bch2_hash_delete_at(trans, bch2_dirent_hash_desc,
 				    &dir_hash_info, &iter,
-				    BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				    BTREE_UPDATE_internal_snapshot_node);
 	bch2_trans_iter_exit(trans, &iter);
 err:
 	bch_err_fn(c, ret);
@@ -274,9 +274,9 @@ create_lostfound:
 				&lostfound_str,
 				lostfound->bi_inum,
 				&lostfound->bi_dir_offset,
-				BCH_HASH_SET_MUST_CREATE) ?:
+				STR_HASH_must_create) ?:
 		bch2_inode_write_flags(trans, &lostfound_iter, lostfound,
-				       BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				       BTREE_UPDATE_internal_snapshot_node);
 err:
 	bch_err_msg(c, ret, "creating lost+found");
 	bch2_trans_iter_exit(trans, &lostfound_iter);
@@ -333,7 +333,7 @@ static int reattach_inode(struct btree_trans *trans,
 				&name,
 				inode->bi_subvol ?: inode->bi_inum,
 				&dir_offset,
-				BCH_HASH_SET_MUST_CREATE);
+				STR_HASH_must_create);
 	if (ret)
 		return ret;
 
@@ -486,14 +486,9 @@ static int reconstruct_reg_inode(struct btree_trans *trans, u32 snapshot, u64 in
 	return reconstruct_inode(trans, snapshot, inum, k.k->p.offset << 9, S_IFREG);
 }
 
-struct snapshots_seen_entry {
-	u32				id;
-	u32				equiv;
-};
-
 struct snapshots_seen {
 	struct bpos			pos;
-	DARRAY(struct snapshots_seen_entry) ids;
+	snapshot_id_list		ids;
 };
 
 static inline void snapshots_seen_exit(struct snapshots_seen *s)
@@ -508,20 +503,15 @@ static inline void snapshots_seen_init(struct snapshots_seen *s)
 
 static int snapshots_seen_add_inorder(struct bch_fs *c, struct snapshots_seen *s, u32 id)
 {
-	struct snapshots_seen_entry *i, n = {
-		.id	= id,
-		.equiv	= bch2_snapshot_equiv(c, id),
-	};
-	int ret = 0;
-
+	u32 *i;
 	__darray_for_each(s->ids, i) {
-		if (i->id == id)
+		if (*i == id)
 			return 0;
-		if (i->id > id)
+		if (*i > id)
 			break;
 	}
 
-	ret = darray_insert_item(&s->ids, i - s->ids.data, n);
+	int ret = darray_insert_item(&s->ids, i - s->ids.data, id);
 	if (ret)
 		bch_err(c, "error reallocating snapshots_seen table (size %zu)",
 			s->ids.size);
@@ -531,42 +521,11 @@ static int snapshots_seen_add_inorder(struct bch_fs *c, struct snapshots_seen *s
 static int snapshots_seen_update(struct bch_fs *c, struct snapshots_seen *s,
 				 enum btree_id btree_id, struct bpos pos)
 {
-	struct snapshots_seen_entry n = {
-		.id	= pos.snapshot,
-		.equiv	= bch2_snapshot_equiv(c, pos.snapshot),
-	};
-	int ret = 0;
-
 	if (!bkey_eq(s->pos, pos))
 		s->ids.nr = 0;
-
 	s->pos = pos;
-	s->pos.snapshot = n.equiv;
 
-	darray_for_each(s->ids, i) {
-		if (i->id == n.id)
-			return 0;
-
-		/*
-		 * We currently don't rigorously track for snapshot cleanup
-		 * needing to be run, so it shouldn't be a fsck error yet:
-		 */
-		if (i->equiv == n.equiv) {
-			bch_err(c, "snapshot deletion did not finish:\n"
-				"  duplicate keys in btree %s at %llu:%llu snapshots %u, %u (equiv %u)\n",
-				bch2_btree_id_str(btree_id),
-				pos.inode, pos.offset,
-				i->id, n.id, n.equiv);
-			set_bit(BCH_FS_need_delete_dead_snapshots, &c->flags);
-			return bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_delete_dead_snapshots);
-		}
-	}
-
-	ret = darray_push(&s->ids, n);
-	if (ret)
-		bch_err(c, "error reallocating snapshots_seen table (size %zu)",
-			s->ids.size);
-	return ret;
+	return snapshot_list_add_nodup(c, &s->ids, pos.snapshot);
 }
 
 /**
@@ -586,12 +545,10 @@ static bool key_visible_in_snapshot(struct bch_fs *c, struct snapshots_seen *see
 	ssize_t i;
 
 	EBUG_ON(id > ancestor);
-	EBUG_ON(!bch2_snapshot_is_equiv(c, id));
-	EBUG_ON(!bch2_snapshot_is_equiv(c, ancestor));
 
 	/* @ancestor should be the snapshot most recently added to @seen */
 	EBUG_ON(ancestor != seen->pos.snapshot);
-	EBUG_ON(ancestor != seen->ids.data[seen->ids.nr - 1].equiv);
+	EBUG_ON(ancestor != darray_last(seen->ids));
 
 	if (id == ancestor)
 		return true;
@@ -610,9 +567,9 @@ static bool key_visible_in_snapshot(struct bch_fs *c, struct snapshots_seen *see
 	 */
 
 	for (i = seen->ids.nr - 2;
-	     i >= 0 && seen->ids.data[i].equiv >= id;
+	     i >= 0 && seen->ids.data[i] >= id;
 	     --i)
-		if (bch2_snapshot_is_ancestor(c, id, seen->ids.data[i].equiv))
+		if (bch2_snapshot_is_ancestor(c, id, seen->ids.data[i]))
 			return false;
 
 	return true;
@@ -643,9 +600,6 @@ static int ref_visible2(struct bch_fs *c,
 			u32 src, struct snapshots_seen *src_seen,
 			u32 dst, struct snapshots_seen *dst_seen)
 {
-	src = bch2_snapshot_equiv(c, src);
-	dst = bch2_snapshot_equiv(c, dst);
-
 	if (dst > src) {
 		swap(dst, src);
 		swap(dst_seen, src_seen);
@@ -692,7 +646,7 @@ static int add_inode(struct bch_fs *c, struct inode_walker *w,
 
 	return darray_push(&w->inodes, ((struct inode_walker_entry) {
 		.inode		= u,
-		.snapshot	= bch2_snapshot_equiv(c, inode.k->p.snapshot),
+		.snapshot	= inode.k->p.snapshot,
 	}));
 }
 
@@ -708,7 +662,7 @@ static int get_inodes_all_snapshots(struct btree_trans *trans,
 	w->inodes.nr = 0;
 
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_inodes, POS(0, inum),
-				     BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+				     BTREE_ITER_all_snapshots, k, ret) {
 		if (k.k->p.offset != inum)
 			break;
 
@@ -728,21 +682,20 @@ static struct inode_walker_entry *
 lookup_inode_for_snapshot(struct bch_fs *c, struct inode_walker *w, struct bkey_s_c k)
 {
 	bool is_whiteout = k.k->type == KEY_TYPE_whiteout;
-	u32 snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
 
 	struct inode_walker_entry *i;
 	__darray_for_each(w->inodes, i)
-		if (bch2_snapshot_is_ancestor(c, snapshot, i->snapshot))
+		if (bch2_snapshot_is_ancestor(c, k.k->p.snapshot, i->snapshot))
 			goto found;
 
 	return NULL;
 found:
-	BUG_ON(snapshot > i->snapshot);
+	BUG_ON(k.k->p.snapshot > i->snapshot);
 
-	if (snapshot != i->snapshot && !is_whiteout) {
+	if (k.k->p.snapshot != i->snapshot && !is_whiteout) {
 		struct inode_walker_entry new = *i;
 
-		new.snapshot = snapshot;
+		new.snapshot = k.k->p.snapshot;
 		new.count = 0;
 
 		struct printbuf buf = PRINTBUF;
@@ -751,10 +704,10 @@ found:
 		bch_info(c, "have key for inode %llu:%u but have inode in ancestor snapshot %u\n"
 			 "unexpected because we should always update the inode when we update a key in that inode\n"
 			 "%s",
-			 w->last_pos.inode, snapshot, i->snapshot, buf.buf);
+			 w->last_pos.inode, k.k->p.snapshot, i->snapshot, buf.buf);
 		printbuf_exit(&buf);
 
-		while (i > w->inodes.data && i[-1].snapshot > snapshot)
+		while (i > w->inodes.data && i[-1].snapshot > k.k->p.snapshot)
 			--i;
 
 		size_t pos = i - w->inodes.data;
@@ -786,10 +739,10 @@ static struct inode_walker_entry *walk_inode(struct btree_trans *trans,
 	return lookup_inode_for_snapshot(trans->c, w, k);
 }
 
-static int __get_visible_inodes(struct btree_trans *trans,
-				struct inode_walker *w,
-				struct snapshots_seen *s,
-				u64 inum)
+static int get_visible_inodes(struct btree_trans *trans,
+			      struct inode_walker *w,
+			      struct snapshots_seen *s,
+			      u64 inum)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
@@ -799,19 +752,17 @@ static int __get_visible_inodes(struct btree_trans *trans,
 	w->inodes.nr = 0;
 
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_inodes, POS(0, inum),
-			   BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
-		u32 equiv = bch2_snapshot_equiv(c, k.k->p.snapshot);
-
+			   BTREE_ITER_all_snapshots, k, ret) {
 		if (k.k->p.offset != inum)
 			break;
 
-		if (!ref_visible(c, s, s->pos.snapshot, equiv))
+		if (!ref_visible(c, s, s->pos.snapshot, k.k->p.snapshot))
 			continue;
 
 		if (bkey_is_inode(k.k))
 			add_inode(c, w, k);
 
-		if (equiv >= s->pos.snapshot)
+		if (k.k->p.snapshot >= s->pos.snapshot)
 			break;
 	}
 	bch2_trans_iter_exit(trans, &iter);
@@ -832,7 +783,7 @@ static int check_key_has_snapshot(struct btree_trans *trans,
 				"key in missing snapshot: %s",
 				(bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
 		ret = bch2_btree_delete_at(trans, iter,
-					    BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?: 1;
+					    BTREE_UPDATE_internal_snapshot_node) ?: 1;
 fsck_err:
 	printbuf_exit(&buf);
 	return ret;
@@ -861,8 +812,8 @@ static int hash_redo_key(struct btree_trans *trans,
 		bch2_hash_set_in_snapshot(trans, desc, hash_info,
 				       (subvol_inum) { 0, k.k->p.inode },
 				       k.k->p.snapshot, tmp,
-				       BCH_HASH_SET_MUST_CREATE,
-				       BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+				       STR_HASH_must_create|
+				       BTREE_UPDATE_internal_snapshot_node) ?:
 		bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
 }
 
@@ -891,7 +842,7 @@ static int hash_check_key(struct btree_trans *trans,
 
 	for_each_btree_key_norestart(trans, iter, desc.btree_id,
 				     SPOS(hash_k.k->p.inode, hash, hash_k.k->p.snapshot),
-				     BTREE_ITER_SLOTS, k, ret) {
+				     BTREE_ITER_slots, k, ret) {
 		if (bkey_eq(k.k->p, hash_k.k->p))
 			break;
 
@@ -1233,7 +1184,7 @@ int bch2_check_inodes(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_inodes,
 				POS_MIN,
-				BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+				BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			check_inode(trans, &iter, k, &prev, &s, full)));
 
@@ -1362,8 +1313,8 @@ static int overlapping_extents_found(struct btree_trans *trans,
 	BUG_ON(bkey_le(pos1, bkey_start_pos(&pos2)));
 
 	bch2_trans_iter_init(trans, &iter1, btree, pos1,
-			     BTREE_ITER_ALL_SNAPSHOTS|
-			     BTREE_ITER_NOT_EXTENTS);
+			     BTREE_ITER_all_snapshots|
+			     BTREE_ITER_not_extents);
 	k1 = bch2_btree_iter_peek_upto(&iter1, POS(pos1.inode, U64_MAX));
 	ret = bkey_err(k1);
 	if (ret)
@@ -1425,7 +1376,7 @@ static int overlapping_extents_found(struct btree_trans *trans,
 		trans->extra_disk_res += bch2_bkey_sectors_compressed(k2);
 
 		ret =   bch2_trans_update_extent_overwrite(trans, old_iter,
-				BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE,
+				BTREE_UPDATE_internal_snapshot_node,
 				k1, k2) ?:
 			bch2_trans_commit(trans, &res, NULL, BCH_TRANS_COMMIT_no_enospc);
 		bch2_disk_reservation_put(c, &res);
@@ -1466,7 +1417,6 @@ static int check_overlapping_extents(struct btree_trans *trans,
 			      struct snapshots_seen *seen,
 			      struct extent_ends *extent_ends,
 			      struct bkey_s_c k,
-			      u32 equiv,
 			      struct btree_iter *iter,
 			      bool *fixed)
 {
@@ -1535,11 +1485,8 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
 	struct bch_fs *c = trans->c;
 	struct inode_walker_entry *i;
 	struct printbuf buf = PRINTBUF;
-	struct bpos equiv = k.k->p;
 	int ret = 0;
 
-	equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
-
 	ret = check_key_has_snapshot(trans, iter, k);
 	if (ret) {
 		ret = ret < 0 ? ret : 0;
@@ -1589,8 +1536,7 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
 				 bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
 			goto delete;
 
-		ret = check_overlapping_extents(trans, s, extent_ends, k,
-						equiv.snapshot, iter,
+		ret = check_overlapping_extents(trans, s, extent_ends, k, iter,
 						&inode->recalculate_sums);
 		if (ret)
 			goto err;
@@ -1607,8 +1553,8 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
 	for (;
 	     inode->inodes.data && i >= inode->inodes.data;
 	     --i) {
-		if (i->snapshot > equiv.snapshot ||
-		    !key_visible_in_snapshot(c, s, i->snapshot, equiv.snapshot))
+		if (i->snapshot > k.k->p.snapshot ||
+		    !key_visible_in_snapshot(c, s, i->snapshot, k.k->p.snapshot))
 			continue;
 
 		if (k.k->type != KEY_TYPE_whiteout) {
@@ -1625,7 +1571,7 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
 				bch2_btree_iter_set_snapshot(&iter2, i->snapshot);
 				ret =   bch2_btree_iter_traverse(&iter2) ?:
 					bch2_btree_delete_at(trans, &iter2,
-						BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+						BTREE_UPDATE_internal_snapshot_node);
 				bch2_trans_iter_exit(trans, &iter2);
 				if (ret)
 					goto err;
@@ -1652,7 +1598,7 @@ fsck_err:
 	bch_err_fn(c, ret);
 	return ret;
 delete:
-	ret = bch2_btree_delete_at(trans, iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+	ret = bch2_btree_delete_at(trans, iter, BTREE_UPDATE_internal_snapshot_node);
 	goto out;
 }
 
@@ -1673,7 +1619,7 @@ int bch2_check_extents(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_extents,
 				POS(BCACHEFS_ROOT_INO, 0),
-				BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+				BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 				&res, NULL,
 				BCH_TRANS_COMMIT_no_enospc, ({
 			bch2_disk_reservation_put(c, &res);
@@ -1698,7 +1644,7 @@ int bch2_check_indirect_extents(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_reflink,
 				POS_MIN,
-				BTREE_ITER_PREFETCH, k,
+				BTREE_ITER_prefetch, k,
 				&res, NULL,
 				BCH_TRANS_COMMIT_no_enospc, ({
 			bch2_disk_reservation_put(c, &res);
@@ -1767,6 +1713,15 @@ static int check_dirent_inode_dirent(struct btree_trans *trans,
 	if (inode_points_to_dirent(target, d))
 		return 0;
 
+	if (bch2_inode_should_have_bp(target) &&
+	    !fsck_err(c, inode_wrong_backpointer,
+		      "dirent points to inode that does not point back:\n  %s",
+		      (bch2_bkey_val_to_text(&buf, c, d.s_c),
+		       prt_printf(&buf, "\n  "),
+		       bch2_inode_unpacked_to_text(&buf, target),
+		       buf.buf)))
+		goto out_noiter;
+
 	if (!target->bi_dir &&
 	    !target->bi_dir_offset) {
 		target->bi_dir		= d.k->p.inode;
@@ -1835,6 +1790,7 @@ out:
 err:
 fsck_err:
 	bch2_trans_iter_exit(trans, &bp_iter);
+out_noiter:
 	printbuf_exit(&buf);
 	bch_err_fn(c, ret);
 	return ret;
@@ -2052,7 +2008,6 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
 	struct bch_fs *c = trans->c;
 	struct inode_walker_entry *i;
 	struct printbuf buf = PRINTBUF;
-	struct bpos equiv;
 	int ret = 0;
 
 	ret = check_key_has_snapshot(trans, iter, k);
@@ -2061,9 +2016,6 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
 		goto out;
 	}
 
-	equiv = k.k->p;
-	equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
-
 	ret = snapshots_seen_update(c, s, iter->btree_id, k.k->p);
 	if (ret)
 		goto err;
@@ -2104,7 +2056,7 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
 			(printbuf_reset(&buf),
 			 bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
 		ret = bch2_btree_delete_at(trans, iter,
-				BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				BTREE_UPDATE_internal_snapshot_node);
 		goto out;
 	}
 
@@ -2140,14 +2092,13 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
 		if (ret)
 			goto err;
 	} else {
-		ret = __get_visible_inodes(trans, target, s, le64_to_cpu(d.v->d_inum));
+		ret = get_visible_inodes(trans, target, s, le64_to_cpu(d.v->d_inum));
 		if (ret)
 			goto err;
 
 		if (fsck_err_on(!target->inodes.nr,
 				c, dirent_to_missing_inode,
-				"dirent points to missing inode: (equiv %u)\n%s",
-				equiv.snapshot,
+				"dirent points to missing inode:\n%s",
 				(printbuf_reset(&buf),
 				 bch2_bkey_val_to_text(&buf, c, k),
 				 buf.buf))) {
@@ -2164,7 +2115,7 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
 		}
 
 		if (d.v->d_type == DT_DIR)
-			for_each_visible_inode(c, s, dir, equiv.snapshot, i)
+			for_each_visible_inode(c, s, dir, d.k->p.snapshot, i)
 				i->count++;
 	}
 out:
@@ -2191,7 +2142,7 @@ int bch2_check_dirents(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_dirents,
 				POS(BCACHEFS_ROOT_INO, 0),
-				BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
+				BTREE_ITER_prefetch|BTREE_ITER_all_snapshots,
 				k,
 				NULL, NULL,
 				BCH_TRANS_COMMIT_no_enospc,
@@ -2255,7 +2206,7 @@ int bch2_check_xattrs(struct bch_fs *c)
 	ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_xattrs,
 			POS(BCACHEFS_ROOT_INO, 0),
-			BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
+			BTREE_ITER_prefetch|BTREE_ITER_all_snapshots,
 			k,
 			NULL, NULL,
 			BCH_TRANS_COMMIT_no_enospc,
@@ -2422,7 +2373,7 @@ int bch2_check_subvolume_structure(struct bch_fs *c)
 {
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
-				BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_PREFETCH, k,
+				BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			check_subvol_path(trans, &iter, k)));
 	bch_err_fn(c, ret);
@@ -2457,7 +2408,7 @@ static int check_path(struct btree_trans *trans, pathbuf *p, struct bkey_s_c ino
 	struct btree_iter inode_iter = {};
 	struct bch_inode_unpacked inode;
 	struct printbuf buf = PRINTBUF;
-	u32 snapshot = bch2_snapshot_equiv(c, inode_k.k->p.snapshot);
+	u32 snapshot = inode_k.k->p.snapshot;
 	int ret = 0;
 
 	p->nr = 0;
@@ -2559,9 +2510,9 @@ int bch2_check_directory_structure(struct bch_fs *c)
 
 	ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_inodes, POS_MIN,
-					  BTREE_ITER_INTENT|
-					  BTREE_ITER_PREFETCH|
-					  BTREE_ITER_ALL_SNAPSHOTS, k,
+					  BTREE_ITER_intent|
+					  BTREE_ITER_prefetch|
+					  BTREE_ITER_all_snapshots, k,
 					  NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({
 			if (!bkey_is_inode(k.k))
 				continue;
@@ -2661,9 +2612,9 @@ static int check_nlinks_find_hardlinks(struct bch_fs *c,
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_inodes,
 				   POS(0, start),
-				   BTREE_ITER_INTENT|
-				   BTREE_ITER_PREFETCH|
-				   BTREE_ITER_ALL_SNAPSHOTS, k, ({
+				   BTREE_ITER_intent|
+				   BTREE_ITER_prefetch|
+				   BTREE_ITER_all_snapshots, k, ({
 			if (!bkey_is_inode(k.k))
 				continue;
 
@@ -2704,9 +2655,9 @@ static int check_nlinks_walk_dirents(struct bch_fs *c, struct nlink_table *links
 
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_dirents, POS_MIN,
-				   BTREE_ITER_INTENT|
-				   BTREE_ITER_PREFETCH|
-				   BTREE_ITER_ALL_SNAPSHOTS, k, ({
+				   BTREE_ITER_intent|
+				   BTREE_ITER_prefetch|
+				   BTREE_ITER_all_snapshots, k, ({
 			ret = snapshots_seen_update(c, &s, iter.btree_id, k.k->p);
 			if (ret)
 				break;
@@ -2717,8 +2668,7 @@ static int check_nlinks_walk_dirents(struct bch_fs *c, struct nlink_table *links
 				if (d.v->d_type != DT_DIR &&
 				    d.v->d_type != DT_SUBVOL)
 					inc_link(c, &s, links, range_start, range_end,
-						 le64_to_cpu(d.v->d_inum),
-						 bch2_snapshot_equiv(c, d.k->p.snapshot));
+						 le64_to_cpu(d.v->d_inum), d.k->p.snapshot);
 			}
 			0;
 		})));
@@ -2781,7 +2731,7 @@ static int check_nlinks_update_hardlinks(struct bch_fs *c,
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_inodes,
 				POS(0, range_start),
-				BTREE_ITER_INTENT|BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+				BTREE_ITER_intent|BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			check_nlinks_update_inode(trans, &iter, k, links, &idx, range_end)));
 	if (ret < 0) {
@@ -2849,7 +2799,7 @@ static int fix_reflink_p_key(struct btree_trans *trans, struct btree_iter *iter,
 	u->v.front_pad	= 0;
 	u->v.back_pad	= 0;
 
-	return bch2_trans_update(trans, iter, &u->k_i, BTREE_TRIGGER_NORUN);
+	return bch2_trans_update(trans, iter, &u->k_i, BTREE_TRIGGER_norun);
 }
 
 int bch2_fix_reflink_p(struct bch_fs *c)
@@ -2860,8 +2810,8 @@ int bch2_fix_reflink_p(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
 				BTREE_ID_extents, POS_MIN,
-				BTREE_ITER_INTENT|BTREE_ITER_PREFETCH|
-				BTREE_ITER_ALL_SNAPSHOTS, k,
+				BTREE_ITER_intent|BTREE_ITER_prefetch|
+				BTREE_ITER_all_snapshots, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			fix_reflink_p_key(trans, &iter, k)));
 	bch_err_fn(c, ret);
diff --git a/fs/bcachefs/inode.c b/fs/bcachefs/inode.c
index 0f95d7fb5ec0bd..91572e05d5f06b 100644
--- a/fs/bcachefs/inode.c
+++ b/fs/bcachefs/inode.c
@@ -339,7 +339,7 @@ int bch2_inode_peek_nowarn(struct btree_trans *trans,
 
 	k = bch2_bkey_get_iter(trans, iter, BTREE_ID_inodes,
 			       SPOS(0, inum.inum, snapshot),
-			       flags|BTREE_ITER_CACHED);
+			       flags|BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (ret)
 		return ret;
@@ -371,7 +371,7 @@ int bch2_inode_peek(struct btree_trans *trans,
 int bch2_inode_write_flags(struct btree_trans *trans,
 		     struct btree_iter *iter,
 		     struct bch_inode_unpacked *inode,
-		     enum btree_update_flags flags)
+		     enum btree_iter_update_trigger_flags flags)
 {
 	struct bkey_inode_buf *inode_p;
 
@@ -399,7 +399,7 @@ int __bch2_fsck_write_inode(struct btree_trans *trans,
 
 	return bch2_btree_insert_nonextent(trans, BTREE_ID_inodes,
 				&inode_p->inode.k_i,
-				BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				BTREE_UPDATE_internal_snapshot_node);
 }
 
 int bch2_fsck_write_inode(struct btree_trans *trans,
@@ -535,29 +535,19 @@ static void __bch2_inode_unpacked_to_text(struct printbuf *out,
 					  struct bch_inode_unpacked *inode)
 {
 	printbuf_indent_add(out, 2);
-	prt_printf(out, "mode=%o", inode->bi_mode);
-	prt_newline(out);
+	prt_printf(out, "mode=%o\n", inode->bi_mode);
 
 	prt_str(out, "flags=");
 	prt_bitflags(out, bch2_inode_flag_strs, inode->bi_flags & ((1U << 20) - 1));
-	prt_printf(out, " (%x)", inode->bi_flags);
-	prt_newline(out);
+	prt_printf(out, " (%x)\n", inode->bi_flags);
 
-	prt_printf(out, "journal_seq=%llu", inode->bi_journal_seq);
-	prt_newline(out);
-
-	prt_printf(out, "bi_size=%llu", inode->bi_size);
-	prt_newline(out);
-
-	prt_printf(out, "bi_sectors=%llu", inode->bi_sectors);
-	prt_newline(out);
-
-	prt_printf(out, "bi_version=%llu", inode->bi_version);
-	prt_newline(out);
+	prt_printf(out, "journal_seq=%llu\n",	inode->bi_journal_seq);
+	prt_printf(out, "bi_size=%llu\n",	inode->bi_size);
+	prt_printf(out, "bi_sectors=%llu\n",	inode->bi_sectors);
+	prt_printf(out, "bi_version=%llu\n",	inode->bi_version);
 
 #define x(_name, _bits)						\
-	prt_printf(out, #_name "=%llu", (u64) inode->_name);	\
-	prt_newline(out);
+	prt_printf(out, #_name "=%llu\n", (u64) inode->_name);
 	BCH_INODE_FIELDS_v3()
 #undef  x
 	printbuf_indent_sub(out, 2);
@@ -604,11 +594,11 @@ int bch2_trigger_inode(struct btree_trans *trans,
 		       enum btree_id btree_id, unsigned level,
 		       struct bkey_s_c old,
 		       struct bkey_s new,
-		       unsigned flags)
+		       enum btree_iter_update_trigger_flags flags)
 {
 	s64 nr = (s64) bkey_is_inode(new.k) - (s64) bkey_is_inode(old.k);
 
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (flags & BTREE_TRIGGER_transactional) {
 		if (nr) {
 			int ret = bch2_replicas_deltas_realloc(trans, 0);
 			if (ret)
@@ -627,13 +617,13 @@ int bch2_trigger_inode(struct btree_trans *trans,
 		}
 	}
 
-	if ((flags & BTREE_TRIGGER_ATOMIC) && (flags & BTREE_TRIGGER_INSERT)) {
+	if ((flags & BTREE_TRIGGER_atomic) && (flags & BTREE_TRIGGER_insert)) {
 		BUG_ON(!trans->journal_res.seq);
 
 		bkey_s_to_inode_v3(new).v->bi_journal_seq = cpu_to_le64(trans->journal_res.seq);
 	}
 
-	if (flags & BTREE_TRIGGER_GC) {
+	if (flags & BTREE_TRIGGER_gc) {
 		struct bch_fs *c = trans->c;
 
 		percpu_down_read(&c->mark_lock);
@@ -762,8 +752,8 @@ int bch2_inode_create(struct btree_trans *trans,
 
 	pos = start;
 	bch2_trans_iter_init(trans, iter, BTREE_ID_inodes, POS(0, pos),
-			     BTREE_ITER_ALL_SNAPSHOTS|
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_all_snapshots|
+			     BTREE_ITER_intent);
 again:
 	while ((k = bch2_btree_iter_peek(iter)).k &&
 	       !(ret = bkey_err(k)) &&
@@ -824,7 +814,7 @@ static int bch2_inode_delete_keys(struct btree_trans *trans,
 	 * extent iterator:
 	 */
 	bch2_trans_iter_init(trans, &iter, id, POS(inum.inum, 0),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	while (1) {
 		bch2_trans_begin(trans);
@@ -846,7 +836,7 @@ static int bch2_inode_delete_keys(struct btree_trans *trans,
 		bkey_init(&delete.k);
 		delete.k.p = iter.pos;
 
-		if (iter.flags & BTREE_ITER_IS_EXTENTS)
+		if (iter.flags & BTREE_ITER_is_extents)
 			bch2_key_resize(&delete.k,
 					bpos_min(end, k.k->p).offset -
 					iter.pos.offset);
@@ -895,7 +885,7 @@ retry:
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
 			       SPOS(0, inum.inum, snapshot),
-			       BTREE_ITER_INTENT|BTREE_ITER_CACHED);
+			       BTREE_ITER_intent|BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (ret)
 		goto err;
@@ -1055,7 +1045,7 @@ retry:
 	bch2_trans_begin(trans);
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
-			       SPOS(0, inum, snapshot), BTREE_ITER_INTENT);
+			       SPOS(0, inum, snapshot), BTREE_ITER_intent);
 	ret = bkey_err(k);
 	if (ret)
 		goto err;
@@ -1100,7 +1090,7 @@ static int may_delete_deleted_inode(struct btree_trans *trans,
 	struct bch_inode_unpacked inode;
 	int ret;
 
-	k = bch2_bkey_get_iter(trans, &inode_iter, BTREE_ID_inodes, pos, BTREE_ITER_CACHED);
+	k = bch2_bkey_get_iter(trans, &inode_iter, BTREE_ID_inodes, pos, BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (ret)
 		return ret;
@@ -1152,7 +1142,7 @@ static int may_delete_deleted_inode(struct btree_trans *trans,
 		inode.bi_flags &= ~BCH_INODE_unlinked;
 
 		ret = bch2_inode_write_flags(trans, &inode_iter, &inode,
-					     BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+					     BTREE_UPDATE_internal_snapshot_node);
 		bch_err_msg(c, ret, "clearing inode unlinked flag");
 		if (ret)
 			goto out;
@@ -1199,7 +1189,7 @@ again:
 	 * flushed and we'd spin:
 	 */
 	ret = for_each_btree_key_commit(trans, iter, BTREE_ID_deleted_inodes, POS_MIN,
-					BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+					BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 					NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({
 		ret = may_delete_deleted_inode(trans, &iter, k.k->p, &need_another_pass);
 		if (ret > 0) {
diff --git a/fs/bcachefs/inode.h b/fs/bcachefs/inode.h
index 056298050550f9..abdfb1f3e4d08f 100644
--- a/fs/bcachefs/inode.h
+++ b/fs/bcachefs/inode.h
@@ -18,7 +18,8 @@ int bch2_inode_v3_invalid(struct bch_fs *, struct bkey_s_c,
 void bch2_inode_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 int bch2_trigger_inode(struct btree_trans *, enum btree_id, unsigned,
-			  struct bkey_s_c, struct bkey_s, unsigned);
+		       struct bkey_s_c, struct bkey_s,
+		       enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_inode ((struct bkey_ops) {	\
 	.key_invalid	= bch2_inode_invalid,		\
@@ -101,7 +102,7 @@ int bch2_inode_peek(struct btree_trans *, struct btree_iter *,
 		    struct bch_inode_unpacked *, subvol_inum, unsigned);
 
 int bch2_inode_write_flags(struct btree_trans *, struct btree_iter *,
-		     struct bch_inode_unpacked *, enum btree_update_flags);
+		     struct bch_inode_unpacked *, enum btree_iter_update_trigger_flags);
 
 static inline int bch2_inode_write(struct btree_trans *trans,
 		     struct btree_iter *iter,
@@ -220,6 +221,14 @@ static inline void bch2_inode_nlink_set(struct bch_inode_unpacked *bi,
 int bch2_inode_nlink_inc(struct bch_inode_unpacked *);
 void bch2_inode_nlink_dec(struct btree_trans *, struct bch_inode_unpacked *);
 
+static inline bool bch2_inode_should_have_bp(struct bch_inode_unpacked *inode)
+{
+	bool inode_has_bp = inode->bi_dir || inode->bi_dir_offset;
+
+	return S_ISDIR(inode->bi_mode) ||
+		(!inode->bi_nlink && inode_has_bp);
+}
+
 struct bch_opts bch2_inode_opts_to_opts(struct bch_inode_unpacked *);
 void bch2_inode_opts_get(struct bch_io_opts *, struct bch_fs *,
 			 struct bch_inode_unpacked *);
diff --git a/fs/bcachefs/io_misc.c b/fs/bcachefs/io_misc.c
index 82f9170dab3fdc..4ec979b4b23e4b 100644
--- a/fs/bcachefs/io_misc.c
+++ b/fs/bcachefs/io_misc.c
@@ -198,7 +198,7 @@ int bch2_fpunch(struct bch_fs *c, subvol_inum inum, u64 start, u64 end,
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     POS(inum.inum, start),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	ret = bch2_fpunch_at(trans, &iter, inum, end, i_sectors_delta);
 
@@ -230,7 +230,7 @@ static int truncate_set_isize(struct btree_trans *trans,
 	struct bch_inode_unpacked inode_u;
 	int ret;
 
-	ret   = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_INTENT) ?:
+	ret   = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent) ?:
 		(inode_u.bi_size = new_i_size, 0) ?:
 		bch2_inode_write(trans, &iter, &inode_u);
 
@@ -256,7 +256,7 @@ static int __bch2_resume_logged_op_truncate(struct btree_trans *trans,
 
 	bch2_trans_iter_init(trans, &fpunch_iter, BTREE_ID_extents,
 			     POS(inum.inum, round_up(new_i_size, block_bytes(c)) >> 9),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	ret = bch2_fpunch_at(trans, &fpunch_iter, inum, U64_MAX, i_sectors_delta);
 	bch2_trans_iter_exit(trans, &fpunch_iter);
 
@@ -317,7 +317,7 @@ static int adjust_i_size(struct btree_trans *trans, subvol_inum inum, u64 offset
 	offset	<<= 9;
 	len	<<= 9;
 
-	ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_INTENT);
+	ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent);
 	if (ret)
 		return ret;
 
@@ -365,7 +365,7 @@ static int __bch2_resume_logged_op_finsert(struct btree_trans *trans,
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     POS(inum.inum, 0),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	switch (op->v.state) {
 case LOGGED_OP_FINSERT_start:
diff --git a/fs/bcachefs/io_read.c b/fs/bcachefs/io_read.c
index 8a556e6d1ab6f2..91260d3dbf5037 100644
--- a/fs/bcachefs/io_read.c
+++ b/fs/bcachefs/io_read.c
@@ -378,7 +378,7 @@ static void bch2_read_retry_nodecode(struct bch_fs *c, struct bch_read_bio *rbio
 	bch2_bkey_buf_init(&sk);
 
 	bch2_trans_iter_init(trans, &iter, rbio->data_btree,
-			     rbio->read_pos, BTREE_ITER_SLOTS);
+			     rbio->read_pos, BTREE_ITER_slots);
 retry:
 	rbio->bio.bi_status = 0;
 
@@ -487,7 +487,7 @@ static int __bch2_rbio_narrow_crcs(struct btree_trans *trans,
 		return 0;
 
 	k = bch2_bkey_get_iter(trans, &iter, rbio->data_btree, rbio->data_pos,
-			       BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
+			       BTREE_ITER_slots|BTREE_ITER_intent);
 	if ((ret = bkey_err(k)))
 		goto out;
 
@@ -523,7 +523,7 @@ static int __bch2_rbio_narrow_crcs(struct btree_trans *trans,
 		goto out;
 
 	ret = bch2_trans_update(trans, &iter, new,
-				BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				BTREE_UPDATE_internal_snapshot_node);
 out:
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
@@ -541,7 +541,7 @@ static void __bch2_read_endio(struct work_struct *work)
 	struct bch_read_bio *rbio =
 		container_of(work, struct bch_read_bio, work);
 	struct bch_fs *c	= rbio->c;
-	struct bch_dev *ca	= bch_dev_bkey_exists(c, rbio->pick.ptr.dev);
+	struct bch_dev *ca	= bch2_dev_bkey_exists(c, rbio->pick.ptr.dev);
 	struct bio *src		= &rbio->bio;
 	struct bio *dst		= &bch2_rbio_parent(rbio)->bio;
 	struct bvec_iter dst_iter = rbio->bvec_iter;
@@ -675,7 +675,7 @@ static void bch2_read_endio(struct bio *bio)
 	struct bch_read_bio *rbio =
 		container_of(bio, struct bch_read_bio, bio);
 	struct bch_fs *c	= rbio->c;
-	struct bch_dev *ca	= bch_dev_bkey_exists(c, rbio->pick.ptr.dev);
+	struct bch_dev *ca	= bch2_dev_bkey_exists(c, rbio->pick.ptr.dev);
 	struct workqueue_struct *wq = NULL;
 	enum rbio_context context = RBIO_CONTEXT_NULL;
 
@@ -697,7 +697,7 @@ static void bch2_read_endio(struct bio *bio)
 	}
 
 	if (((rbio->flags & BCH_READ_RETRY_IF_STALE) && race_fault()) ||
-	    ptr_stale(ca, &rbio->pick.ptr)) {
+	    dev_ptr_stale(ca, &rbio->pick.ptr)) {
 		trace_and_count(c, read_reuse_race, &rbio->bio);
 
 		if (rbio->flags & BCH_READ_RETRY_IF_STALE)
@@ -758,22 +758,21 @@ err:
 }
 
 static noinline void read_from_stale_dirty_pointer(struct btree_trans *trans,
+						   struct bch_dev *ca,
 						   struct bkey_s_c k,
 						   struct bch_extent_ptr ptr)
 {
 	struct bch_fs *c = trans->c;
-	struct bch_dev *ca = bch_dev_bkey_exists(c, ptr.dev);
 	struct btree_iter iter;
 	struct printbuf buf = PRINTBUF;
 	int ret;
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
-			     PTR_BUCKET_POS(c, &ptr),
-			     BTREE_ITER_CACHED);
+			     PTR_BUCKET_POS(ca, &ptr),
+			     BTREE_ITER_cached);
 
-	prt_printf(&buf, "Attempting to read from stale dirty pointer:");
+	prt_printf(&buf, "Attempting to read from stale dirty pointer:\n");
 	printbuf_indent_add(&buf, 2);
-	prt_newline(&buf);
 
 	bch2_bkey_val_to_text(&buf, c, k);
 	prt_newline(&buf);
@@ -832,7 +831,7 @@ retry_pick:
 		goto err;
 	}
 
-	ca = bch_dev_bkey_exists(c, pick.ptr.dev);
+	ca = bch2_dev_bkey_exists(c, pick.ptr.dev);
 
 	/*
 	 * Stale dirty pointers are treated as IO errors, but @failed isn't
@@ -842,8 +841,8 @@ retry_pick:
 	 */
 	if ((flags & BCH_READ_IN_RETRY) &&
 	    !pick.ptr.cached &&
-	    unlikely(ptr_stale(ca, &pick.ptr))) {
-		read_from_stale_dirty_pointer(trans, k, pick.ptr);
+	    unlikely(dev_ptr_stale(ca, &pick.ptr))) {
+		read_from_stale_dirty_pointer(trans, ca, k, pick.ptr);
 		bch2_mark_io_failure(failed, &pick);
 		goto retry_pick;
 	}
@@ -1113,7 +1112,7 @@ retry:
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     SPOS(inum.inum, bvec_iter.bi_sector, snapshot),
-			     BTREE_ITER_SLOTS);
+			     BTREE_ITER_slots);
 	while (1) {
 		unsigned bytes, sectors, offset_into_extent;
 		enum btree_id data_btree = BTREE_ID_extents;
diff --git a/fs/bcachefs/io_write.c b/fs/bcachefs/io_write.c
index f137252bccc575..fa92a60b0a5da5 100644
--- a/fs/bcachefs/io_write.c
+++ b/fs/bcachefs/io_write.c
@@ -166,7 +166,7 @@ int bch2_sum_sector_overwrites(struct btree_trans *trans,
 	bch2_trans_copy_iter(&iter, extent_iter);
 
 	for_each_btree_key_upto_continue_norestart(iter,
-				new->k.p, BTREE_ITER_SLOTS, old, ret) {
+				new->k.p, BTREE_ITER_slots, old, ret) {
 		s64 sectors = min(new->k.p.offset, old.k->p.offset) -
 			max(bkey_start_offset(&new->k),
 			    bkey_start_offset(old.k));
@@ -213,14 +213,14 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans,
 	 * to be journalled - if we crash, the bi_journal_seq update will be
 	 * lost, but that's fine.
 	 */
-	unsigned inode_update_flags = BTREE_UPDATE_NOJOURNAL;
+	unsigned inode_update_flags = BTREE_UPDATE_nojournal;
 	int ret;
 
 	k = bch2_bkey_get_mut_noupdate(trans, &iter, BTREE_ID_inodes,
 			      SPOS(0,
 				   extent_iter->pos.inode,
 				   extent_iter->snapshot),
-			      BTREE_ITER_CACHED);
+			      BTREE_ITER_cached);
 	ret = PTR_ERR_OR_ZERO(k);
 	if (unlikely(ret))
 		return ret;
@@ -251,7 +251,7 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans,
 	}
 
 	ret = bch2_trans_update(trans, &iter, &inode->k_i,
-				BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
+				BTREE_UPDATE_internal_snapshot_node|
 				inode_update_flags);
 err:
 	bch2_trans_iter_exit(trans, &iter);
@@ -360,7 +360,7 @@ static int bch2_write_index_default(struct bch_write_op *op)
 
 		bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 				     bkey_start_pos(&sk.k->k),
-				     BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
+				     BTREE_ITER_slots|BTREE_ITER_intent);
 
 		ret =   bch2_bkey_set_needs_rebalance(c, sk.k, &op->opts) ?:
 			bch2_extent_update(trans, inum, &iter, sk.k,
@@ -399,13 +399,12 @@ void bch2_submit_wbio_replicas(struct bch_write_bio *wbio, struct bch_fs *c,
 	BUG_ON(c->opts.nochanges);
 
 	bkey_for_each_ptr(ptrs, ptr) {
-		BUG_ON(!bch2_dev_exists2(c, ptr->dev));
+		BUG_ON(!bch2_dev_exists(c, ptr->dev));
 
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 
 		if (to_entry(ptr + 1) < ptrs.end) {
-			n = to_wbio(bio_alloc_clone(NULL, &wbio->bio,
-						GFP_NOFS, &ca->replica_set));
+			n = to_wbio(bio_alloc_clone(NULL, &wbio->bio, GFP_NOFS, &c->replica_set));
 
 			n->bio.bi_end_io	= wbio->bio.bi_end_io;
 			n->bio.bi_private	= wbio->bio.bi_private;
@@ -642,7 +641,7 @@ static void bch2_write_endio(struct bio *bio)
 	struct bch_write_bio *wbio	= to_wbio(bio);
 	struct bch_write_bio *parent	= wbio->split ? wbio->parent : NULL;
 	struct bch_fs *c		= wbio->c;
-	struct bch_dev *ca		= bch_dev_bkey_exists(c, wbio->dev);
+	struct bch_dev *ca		= bch2_dev_bkey_exists(c, wbio->dev);
 
 	if (bch2_dev_inum_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_write,
 				    op->pos.inode,
@@ -1093,12 +1092,17 @@ static bool bch2_extent_is_writeable(struct bch_write_op *op,
 		return false;
 
 	e = bkey_s_c_to_extent(k);
+
+	rcu_read_lock();
 	extent_for_each_ptr_decode(e, p, entry) {
-		if (crc_is_encoded(p.crc) || p.has_ec)
+		if (crc_is_encoded(p.crc) || p.has_ec) {
+			rcu_read_unlock();
 			return false;
+		}
 
 		replicas += bch2_extent_ptr_durability(c, &p);
 	}
+	rcu_read_unlock();
 
 	return replicas >= op->opts.data_replicas;
 }
@@ -1110,10 +1114,12 @@ static inline void bch2_nocow_write_unlock(struct bch_write_op *op)
 	for_each_keylist_key(&op->insert_keys, k) {
 		struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(bkey_i_to_s_c(k));
 
-		bkey_for_each_ptr(ptrs, ptr)
+		bkey_for_each_ptr(ptrs, ptr) {
+			struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 			bch2_bucket_nocow_unlock(&c->nocow_locks,
-						 PTR_BUCKET_POS(c, ptr),
+						 PTR_BUCKET_POS(ca, ptr),
 						 BUCKET_NOCOW_LOCK_UPDATE);
+		}
 	}
 }
 
@@ -1150,7 +1156,7 @@ static int bch2_nocow_write_convert_one_unwritten(struct btree_trans *trans,
 	return  bch2_extent_update_i_size_sectors(trans, iter,
 					min(new->k.p.offset << 9, new_i_size), 0) ?:
 		bch2_trans_update(trans, iter, new,
-				  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				  BTREE_UPDATE_internal_snapshot_node);
 }
 
 static void bch2_nocow_write_convert_unwritten(struct bch_write_op *op)
@@ -1161,7 +1167,7 @@ static void bch2_nocow_write_convert_unwritten(struct bch_write_op *op)
 	for_each_keylist_key(&op->insert_keys, orig) {
 		int ret = for_each_btree_key_upto_commit(trans, iter, BTREE_ID_extents,
 				     bkey_start_pos(&orig->k), orig->k.p,
-				     BTREE_ITER_INTENT, k,
+				     BTREE_ITER_intent, k,
 				     NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({
 			bch2_nocow_write_convert_one_unwritten(trans, &iter, orig, k, op->new_i_size);
 		}));
@@ -1234,12 +1240,16 @@ retry:
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     SPOS(op->pos.inode, op->pos.offset, snapshot),
-			     BTREE_ITER_SLOTS);
+			     BTREE_ITER_slots);
 	while (1) {
 		struct bio *bio = &op->wbio.bio;
 
 		buckets.nr = 0;
 
+		ret = bch2_trans_relock(trans);
+		if (ret)
+			break;
+
 		k = bch2_btree_iter_peek_slot(&iter);
 		ret = bkey_err(k);
 		if (ret)
@@ -1259,12 +1269,13 @@ retry:
 		/* Get iorefs before dropping btree locks: */
 		struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
 		bkey_for_each_ptr(ptrs, ptr) {
-			struct bpos b = PTR_BUCKET_POS(c, ptr);
+			struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
+			struct bpos b = PTR_BUCKET_POS(ca, ptr);
 			struct nocow_lock_bucket *l =
 				bucket_nocow_lock(&c->nocow_locks, bucket_to_u64(b));
 			prefetch(l);
 
-			if (unlikely(!bch2_dev_get_ioref(bch_dev_bkey_exists(c, ptr->dev), WRITE)))
+			if (unlikely(!bch2_dev_get_ioref(ca, WRITE)))
 				goto err_get_ioref;
 
 			/* XXX allocating memory with btree locks held - rare */
@@ -1285,7 +1296,7 @@ retry:
 			bch2_cut_back(POS(op->pos.inode, op->pos.offset + bio_sectors(bio)), op->insert_keys.top);
 
 		darray_for_each(buckets, i) {
-			struct bch_dev *ca = bch_dev_bkey_exists(c, i->b.inode);
+			struct bch_dev *ca = bch2_dev_have_ref(c, i->b.inode);
 
 			__bch2_bucket_nocow_lock(&c->nocow_locks, i->l,
 						 bucket_to_u64(i->b),
@@ -1362,7 +1373,7 @@ err:
 	return;
 err_get_ioref:
 	darray_for_each(buckets, i)
-		percpu_ref_put(&bch_dev_bkey_exists(c, i->b.inode)->io_ref);
+		percpu_ref_put(&bch2_dev_have_ref(c, i->b.inode)->io_ref);
 
 	/* Fall back to COW path: */
 	goto out;
@@ -1639,8 +1650,7 @@ void bch2_write_op_to_text(struct printbuf *out, struct bch_write_op *op)
 	prt_bitflags(out, bch2_write_flags, op->flags);
 	prt_newline(out);
 
-	prt_printf(out, "ref: %u", closure_nr_remaining(&op->cl));
-	prt_newline(out);
+	prt_printf(out, "ref: %u\n", closure_nr_remaining(&op->cl));
 
 	printbuf_indent_sub(out, 2);
 }
@@ -1648,13 +1658,14 @@ void bch2_write_op_to_text(struct printbuf *out, struct bch_write_op *op)
 void bch2_fs_io_write_exit(struct bch_fs *c)
 {
 	mempool_exit(&c->bio_bounce_pages);
+	bioset_exit(&c->replica_set);
 	bioset_exit(&c->bio_write);
 }
 
 int bch2_fs_io_write_init(struct bch_fs *c)
 {
-	if (bioset_init(&c->bio_write, 1, offsetof(struct bch_write_bio, bio),
-			BIOSET_NEED_BVECS))
+	if (bioset_init(&c->bio_write,   1, offsetof(struct bch_write_bio, bio), BIOSET_NEED_BVECS) ||
+	    bioset_init(&c->replica_set, 4, offsetof(struct bch_write_bio, bio), 0))
 		return -BCH_ERR_ENOMEM_bio_write_init;
 
 	if (mempool_init_page_pool(&c->bio_bounce_pages,
diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index 9c9a25dbd6137a..3a42222cc1c539 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -53,29 +53,19 @@ static void bch2_journal_buf_to_text(struct printbuf *out, struct journal *j, u6
 	unsigned i = seq & JOURNAL_BUF_MASK;
 	struct journal_buf *buf = j->buf + i;
 
-	prt_str(out, "seq:");
-	prt_tab(out);
-	prt_printf(out, "%llu", seq);
-	prt_newline(out);
+	prt_printf(out, "seq:\t%llu\n", seq);
 	printbuf_indent_add(out, 2);
 
-	prt_str(out, "refcount:");
-	prt_tab(out);
-	prt_printf(out, "%u", journal_state_count(s, i));
-	prt_newline(out);
+	prt_printf(out, "refcount:\t%u\n", journal_state_count(s, i));
 
-	prt_str(out, "size:");
-	prt_tab(out);
+	prt_printf(out, "size:\t");
 	prt_human_readable_u64(out, vstruct_bytes(buf->data));
 	prt_newline(out);
 
-	prt_str(out, "expires:");
-	prt_tab(out);
-	prt_printf(out, "%li jiffies", buf->expires - jiffies);
-	prt_newline(out);
+	prt_printf(out, "expires:\t");
+	prt_printf(out, "%li jiffies\n", buf->expires - jiffies);
 
-	prt_str(out, "flags:");
-	prt_tab(out);
+	prt_printf(out, "flags:\t");
 	if (buf->noflush)
 		prt_str(out, "noflush ");
 	if (buf->must_flush)
@@ -87,9 +77,9 @@ static void bch2_journal_buf_to_text(struct printbuf *out, struct journal *j, u6
 	if (buf->write_started)
 		prt_str(out, "write_started ");
 	if (buf->write_allocated)
-		prt_str(out, "write allocated ");
+		prt_str(out, "write_allocated ");
 	if (buf->write_done)
-		prt_str(out, "write done");
+		prt_str(out, "write_done");
 	prt_newline(out);
 
 	printbuf_indent_sub(out, 2);
@@ -940,7 +930,8 @@ static int __bch2_set_nr_journal_buckets(struct bch_dev *ca, unsigned nr,
 				break;
 			}
 		} else {
-			ob[nr_got] = bch2_bucket_alloc(c, ca, BCH_WATERMARK_normal, cl);
+			ob[nr_got] = bch2_bucket_alloc(c, ca, BCH_WATERMARK_normal,
+						       BCH_DATA_journal, cl);
 			ret = PTR_ERR_OR_ZERO(ob[nr_got]);
 			if (ret)
 				break;
@@ -948,7 +939,7 @@ static int __bch2_set_nr_journal_buckets(struct bch_dev *ca, unsigned nr,
 			ret = bch2_trans_run(c,
 				bch2_trans_mark_metadata_bucket(trans, ca,
 						ob[nr_got]->bucket, BCH_DATA_journal,
-						ca->mi.bucket_size));
+						ca->mi.bucket_size, BTREE_TRIGGER_transactional));
 			if (ret) {
 				bch2_open_bucket_put(c, ob[nr_got]);
 				bch_err_msg(c, ret, "marking new journal buckets");
@@ -1028,7 +1019,8 @@ err_unblock:
 		for (i = 0; i < nr_got; i++)
 			bch2_trans_run(c,
 				bch2_trans_mark_metadata_bucket(trans, ca,
-						bu[i], BCH_DATA_free, 0));
+						bu[i], BCH_DATA_free, 0,
+						BTREE_TRIGGER_transactional));
 err_free:
 	if (!new_fs)
 		for (i = 0; i < nr_got; i++)
@@ -1179,12 +1171,14 @@ void bch2_fs_journal_stop(struct journal *j)
 	bch2_journal_meta(j);
 
 	journal_quiesce(j);
+	cancel_delayed_work_sync(&j->write_work);
 
 	BUG_ON(!bch2_journal_error(j) &&
-	       test_bit(JOURNAL_REPLAY_DONE, &j->flags) &&
+	       test_bit(JOURNAL_replay_done, &j->flags) &&
 	       j->last_empty_seq != journal_cur_seq(j));
 
-	cancel_delayed_work_sync(&j->write_work);
+	if (!bch2_journal_error(j))
+		clear_bit(JOURNAL_running, &j->flags);
 }
 
 int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
@@ -1258,7 +1252,7 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
 
 	spin_lock(&j->lock);
 
-	set_bit(JOURNAL_STARTED, &j->flags);
+	set_bit(JOURNAL_running, &j->flags);
 	j->last_flush_write = jiffies;
 
 	j->reservations.idx = j->reservations.unwritten_idx = journal_cur_seq(j);
@@ -1399,6 +1393,13 @@ int bch2_fs_journal_init(struct journal *j)
 
 /* debug: */
 
+static const char * const bch2_journal_flags_strs[] = {
+#define x(n)	#n,
+	JOURNAL_FLAGS()
+#undef x
+	NULL
+};
+
 void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 {
 	struct bch_fs *c = container_of(j, struct bch_fs, journal);
@@ -1413,13 +1414,16 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 	rcu_read_lock();
 	s = READ_ONCE(j->reservations);
 
+	prt_printf(out, "flags:\t");
+	prt_bitflags(out, bch2_journal_flags_strs, j->flags);
+	prt_newline(out);
 	prt_printf(out, "dirty journal entries:\t%llu/%llu\n",	fifo_used(&j->pin), j->pin.size);
-	prt_printf(out, "seq:\t\t\t%llu\n",			journal_cur_seq(j));
-	prt_printf(out, "seq_ondisk:\t\t%llu\n",		j->seq_ondisk);
-	prt_printf(out, "last_seq:\t\t%llu\n",			journal_last_seq(j));
+	prt_printf(out, "seq:\t%llu\n",				journal_cur_seq(j));
+	prt_printf(out, "seq_ondisk:\t%llu\n",			j->seq_ondisk);
+	prt_printf(out, "last_seq:\t%llu\n",			journal_last_seq(j));
 	prt_printf(out, "last_seq_ondisk:\t%llu\n",		j->last_seq_ondisk);
 	prt_printf(out, "flushed_seq_ondisk:\t%llu\n",		j->flushed_seq_ondisk);
-	prt_printf(out, "watermark:\t\t%s\n",			bch2_watermarks[j->watermark]);
+	prt_printf(out, "watermark:\t%s\n",			bch2_watermarks[j->watermark]);
 	prt_printf(out, "each entry reserved:\t%u\n",		j->entry_u64s_reserved);
 	prt_printf(out, "nr flush writes:\t%llu\n",		j->nr_flush_writes);
 	prt_printf(out, "nr noflush writes:\t%llu\n",		j->nr_noflush_writes);
@@ -1428,48 +1432,44 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 	prt_newline(out);
 	prt_printf(out, "nr direct reclaim:\t%llu\n",		j->nr_direct_reclaim);
 	prt_printf(out, "nr background reclaim:\t%llu\n",	j->nr_background_reclaim);
-	prt_printf(out, "reclaim kicked:\t\t%u\n",		j->reclaim_kicked);
+	prt_printf(out, "reclaim kicked:\t%u\n",		j->reclaim_kicked);
 	prt_printf(out, "reclaim runs in:\t%u ms\n",		time_after(j->next_reclaim, now)
 	       ? jiffies_to_msecs(j->next_reclaim - jiffies) : 0);
-	prt_printf(out, "blocked:\t\t%u\n",			j->blocked);
+	prt_printf(out, "blocked:\t%u\n",			j->blocked);
 	prt_printf(out, "current entry sectors:\t%u\n",		j->cur_entry_sectors);
 	prt_printf(out, "current entry error:\t%s\n",		bch2_journal_errors[j->cur_entry_error]);
-	prt_printf(out, "current entry:\t\t");
+	prt_printf(out, "current entry:\t");
 
 	switch (s.cur_entry_offset) {
 	case JOURNAL_ENTRY_ERROR_VAL:
-		prt_printf(out, "error");
+		prt_printf(out, "error\n");
 		break;
 	case JOURNAL_ENTRY_CLOSED_VAL:
-		prt_printf(out, "closed");
+		prt_printf(out, "closed\n");
 		break;
 	default:
-		prt_printf(out, "%u/%u", s.cur_entry_offset, j->cur_entry_u64s);
+		prt_printf(out, "%u/%u\n", s.cur_entry_offset, j->cur_entry_u64s);
 		break;
 	}
 
-	prt_newline(out);
-	prt_printf(out, "unwritten entries:");
-	prt_newline(out);
+	prt_printf(out, "unwritten entries:\n");
 	bch2_journal_bufs_to_text(out, j);
 
-	prt_printf(out,
-	       "replay done:\t\t%i\n",
-	       test_bit(JOURNAL_REPLAY_DONE,	&j->flags));
-
 	prt_printf(out, "space:\n");
-	prt_printf(out, "\tdiscarded\t%u:%u\n",
+	printbuf_indent_add(out, 2);
+	prt_printf(out, "discarded\t%u:%u\n",
 	       j->space[journal_space_discarded].next_entry,
 	       j->space[journal_space_discarded].total);
-	prt_printf(out, "\tclean ondisk\t%u:%u\n",
+	prt_printf(out, "clean ondisk\t%u:%u\n",
 	       j->space[journal_space_clean_ondisk].next_entry,
 	       j->space[journal_space_clean_ondisk].total);
-	prt_printf(out, "\tclean\t\t%u:%u\n",
+	prt_printf(out, "clean\t%u:%u\n",
 	       j->space[journal_space_clean].next_entry,
 	       j->space[journal_space_clean].total);
-	prt_printf(out, "\ttotal\t\t%u:%u\n",
+	prt_printf(out, "total\t%u:%u\n",
 	       j->space[journal_space_total].next_entry,
 	       j->space[journal_space_total].total);
+	printbuf_indent_sub(out, 2);
 
 	for_each_member_device_rcu(c, ca, &c->rw_devs[BCH_DATA_journal]) {
 		struct journal_device *ja = &ca->journal;
@@ -1480,14 +1480,16 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 		if (!ja->nr)
 			continue;
 
-		prt_printf(out, "dev %u:\n",		ca->dev_idx);
-		prt_printf(out, "\tnr\t\t%u\n",		ja->nr);
-		prt_printf(out, "\tbucket size\t%u\n",	ca->mi.bucket_size);
-		prt_printf(out, "\tavailable\t%u:%u\n",	bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
-		prt_printf(out, "\tdiscard_idx\t%u\n",	ja->discard_idx);
-		prt_printf(out, "\tdirty_ondisk\t%u (seq %llu)\n", ja->dirty_idx_ondisk,	ja->bucket_seq[ja->dirty_idx_ondisk]);
-		prt_printf(out, "\tdirty_idx\t%u (seq %llu)\n", ja->dirty_idx,		ja->bucket_seq[ja->dirty_idx]);
-		prt_printf(out, "\tcur_idx\t\t%u (seq %llu)\n", ja->cur_idx,		ja->bucket_seq[ja->cur_idx]);
+		prt_printf(out, "dev %u:\n",			ca->dev_idx);
+		printbuf_indent_add(out, 2);
+		prt_printf(out, "nr\t%u\n",			ja->nr);
+		prt_printf(out, "bucket size\t%u\n",		ca->mi.bucket_size);
+		prt_printf(out, "available\t%u:%u\n",		bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
+		prt_printf(out, "discard_idx\t%u\n",		ja->discard_idx);
+		prt_printf(out, "dirty_ondisk\t%u (seq %llu)\n",ja->dirty_idx_ondisk,	ja->bucket_seq[ja->dirty_idx_ondisk]);
+		prt_printf(out, "dirty_idx\t%u (seq %llu)\n",	ja->dirty_idx,		ja->bucket_seq[ja->dirty_idx]);
+		prt_printf(out, "cur_idx\t%u (seq %llu)\n",	ja->cur_idx,		ja->bucket_seq[ja->cur_idx]);
+		printbuf_indent_sub(out, 2);
 	}
 
 	rcu_read_unlock();
@@ -1519,25 +1521,18 @@ bool bch2_journal_seq_pins_to_text(struct printbuf *out, struct journal *j, u64
 
 	pin_list = journal_seq_pin(j, *seq);
 
-	prt_printf(out, "%llu: count %u", *seq, atomic_read(&pin_list->count));
-	prt_newline(out);
+	prt_printf(out, "%llu: count %u\n", *seq, atomic_read(&pin_list->count));
 	printbuf_indent_add(out, 2);
 
 	for (unsigned i = 0; i < ARRAY_SIZE(pin_list->list); i++)
-		list_for_each_entry(pin, &pin_list->list[i], list) {
-			prt_printf(out, "\t%px %ps", pin, pin->flush);
-			prt_newline(out);
-		}
+		list_for_each_entry(pin, &pin_list->list[i], list)
+			prt_printf(out, "\t%px %ps\n", pin, pin->flush);
 
-	if (!list_empty(&pin_list->flushed)) {
-		prt_printf(out, "flushed:");
-		prt_newline(out);
-	}
+	if (!list_empty(&pin_list->flushed))
+		prt_printf(out, "flushed:\n");
 
-	list_for_each_entry(pin, &pin_list->flushed, list) {
-		prt_printf(out, "\t%px %ps", pin, pin->flush);
-		prt_newline(out);
-	}
+	list_for_each_entry(pin, &pin_list->flushed, list)
+		prt_printf(out, "\t%px %ps\n", pin, pin->flush);
 
 	printbuf_indent_sub(out, 2);
 
diff --git a/fs/bcachefs/journal.h b/fs/bcachefs/journal.h
index 7c7528f839c567..fd1f7cdaa8bc62 100644
--- a/fs/bcachefs/journal.h
+++ b/fs/bcachefs/journal.h
@@ -372,7 +372,7 @@ static inline int bch2_journal_res_get(struct journal *j, struct journal_res *re
 	int ret;
 
 	EBUG_ON(res->ref);
-	EBUG_ON(!test_bit(JOURNAL_STARTED, &j->flags));
+	EBUG_ON(!test_bit(JOURNAL_running, &j->flags));
 
 	res->u64s = u64s;
 
@@ -418,8 +418,8 @@ struct bch_dev;
 
 static inline void bch2_journal_set_replay_done(struct journal *j)
 {
-	BUG_ON(!test_bit(JOURNAL_STARTED, &j->flags));
-	set_bit(JOURNAL_REPLAY_DONE, &j->flags);
+	BUG_ON(!test_bit(JOURNAL_running, &j->flags));
+	set_bit(JOURNAL_replay_done, &j->flags);
 }
 
 void bch2_journal_unblock(struct journal *);
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index eb1f9d6f5a196e..e76eb98af74a20 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -17,11 +17,39 @@
 #include "sb-clean.h"
 #include "trace.h"
 
+void bch2_journal_pos_from_member_info_set(struct bch_fs *c)
+{
+	lockdep_assert_held(&c->sb_lock);
+
+	for_each_member_device(c, ca) {
+		struct bch_member *m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
+
+		m->last_journal_bucket = cpu_to_le32(ca->journal.cur_idx);
+		m->last_journal_bucket_offset = cpu_to_le32(ca->mi.bucket_size - ca->journal.sectors_free);
+	}
+}
+
+void bch2_journal_pos_from_member_info_resume(struct bch_fs *c)
+{
+	mutex_lock(&c->sb_lock);
+	for_each_member_device(c, ca) {
+		struct bch_member m = bch2_sb_member_get(c->disk_sb.sb, ca->dev_idx);
+
+		unsigned idx = le32_to_cpu(m.last_journal_bucket);
+		if (idx < ca->journal.nr)
+			ca->journal.cur_idx = idx;
+		unsigned offset = le32_to_cpu(m.last_journal_bucket_offset);
+		if (offset <= ca->mi.bucket_size)
+			ca->journal.sectors_free = ca->mi.bucket_size - offset;
+	}
+	mutex_unlock(&c->sb_lock);
+}
+
 void bch2_journal_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
 			       struct journal_replay *j)
 {
 	darray_for_each(j->ptrs, i) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, i->dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, i->dev);
 		u64 offset;
 
 		div64_u64_rem(i->sector, ca->mi.bucket_size, &offset);
@@ -122,6 +150,10 @@ static int journal_entry_add(struct bch_fs *c, struct bch_dev *ca,
 	struct printbuf buf = PRINTBUF;
 	int ret = JOURNAL_ENTRY_ADD_OK;
 
+	if (!c->journal.oldest_seq_found_ondisk ||
+	    le64_to_cpu(j->seq) < c->journal.oldest_seq_found_ondisk)
+		c->journal.oldest_seq_found_ondisk = le64_to_cpu(j->seq);
+
 	/* Is this entry older than the range we need? */
 	if (!c->opts.read_entire_journal &&
 	    le64_to_cpu(j->seq) < jlist->last_seq)
@@ -677,7 +709,7 @@ static int journal_entry_dev_usage_validate(struct bch_fs *c,
 
 	dev = le32_to_cpu(u->dev);
 
-	if (journal_entry_err_on(!bch2_dev_exists2(c, dev),
+	if (journal_entry_err_on(!bch2_dev_exists(c, dev),
 				 c, version, jset, entry,
 				 journal_entry_dev_usage_bad_dev,
 				 "bad dev")) {
@@ -1057,6 +1089,13 @@ reread:
 			goto err;
 		}
 
+		if (le64_to_cpu(j->seq) > ja->highest_seq_found) {
+			ja->highest_seq_found = le64_to_cpu(j->seq);
+			ja->cur_idx = bucket;
+			ja->sectors_free = ca->mi.bucket_size -
+				bucket_remainder(ca, offset) - sectors;
+		}
+
 		/*
 		 * This happens sometimes if we don't have discards on -
 		 * when we've partially overwritten a bucket with new
@@ -1125,8 +1164,6 @@ static CLOSURE_CALLBACK(bch2_journal_read_device)
 	struct bch_fs *c = ca->fs;
 	struct journal_list *jlist =
 		container_of(cl->parent, struct journal_list, cl);
-	struct journal_replay *r, **_r;
-	struct genradix_iter iter;
 	struct journal_read_buf buf = { NULL, 0 };
 	unsigned i;
 	int ret = 0;
@@ -1146,47 +1183,6 @@ static CLOSURE_CALLBACK(bch2_journal_read_device)
 			goto err;
 	}
 
-	ja->sectors_free = ca->mi.bucket_size;
-
-	mutex_lock(&jlist->lock);
-	genradix_for_each_reverse(&c->journal_entries, iter, _r) {
-		r = *_r;
-
-		if (!r)
-			continue;
-
-		darray_for_each(r->ptrs, i)
-			if (i->dev == ca->dev_idx) {
-				unsigned wrote = bucket_remainder(ca, i->sector) +
-					vstruct_sectors(&r->j, c->block_bits);
-
-				ja->cur_idx = i->bucket;
-				ja->sectors_free = ca->mi.bucket_size - wrote;
-				goto found;
-			}
-	}
-found:
-	mutex_unlock(&jlist->lock);
-
-	if (ja->bucket_seq[ja->cur_idx] &&
-	    ja->sectors_free == ca->mi.bucket_size) {
-#if 0
-		/*
-		 * Debug code for ZNS support, where we (probably) want to be
-		 * correlated where we stopped in the journal to the zone write
-		 * points:
-		 */
-		bch_err(c, "ja->sectors_free == ca->mi.bucket_size");
-		bch_err(c, "cur_idx %u/%u", ja->cur_idx, ja->nr);
-		for (i = 0; i < 3; i++) {
-			unsigned idx = (ja->cur_idx + ja->nr - 1 + i) % ja->nr;
-
-			bch_err(c, "bucket_seq[%u] = %llu", idx, ja->bucket_seq[idx]);
-		}
-#endif
-		ja->sectors_free = 0;
-	}
-
 	/*
 	 * Set dirty_idx to indicate the entire journal is full and needs to be
 	 * reclaimed - journal reclaim will immediately reclaim whatever isn't
@@ -1366,7 +1362,7 @@ int bch2_journal_read(struct bch_fs *c,
 			fsck_err(c, journal_entries_missing,
 				 "journal entries %llu-%llu missing! (replaying %llu-%llu)\n"
 				 "  prev at %s\n"
-				 "  next at %s",
+				 "  next at %s, continue?",
 				 missing_start, missing_end,
 				 *last_seq, *blacklist_seq - 1,
 				 buf1.buf, buf2.buf);
@@ -1390,7 +1386,7 @@ int bch2_journal_read(struct bch_fs *c,
 			continue;
 
 		darray_for_each(i->ptrs, ptr) {
-			struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+			struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 
 			if (!ptr->csum_good)
 				bch_err_dev_offset(ca, ptr->sector,
@@ -1400,7 +1396,7 @@ int bch2_journal_read(struct bch_fs *c,
 		}
 
 		ret = jset_validate(c,
-				    bch_dev_bkey_exists(c, i->ptrs.data[0].dev),
+				    bch2_dev_bkey_exists(c, i->ptrs.data[0].dev),
 				    &i->j,
 				    i->ptrs.data[0].sector,
 				    READ);
@@ -1731,7 +1727,7 @@ static CLOSURE_CALLBACK(journal_write_submit)
 	unsigned sectors = vstruct_sectors(w->data, c->block_bits);
 
 	extent_for_each_ptr(bkey_i_to_s_extent(&w->key), ptr) {
-		struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
+		struct bch_dev *ca = bch2_dev_bkey_exists(c, ptr->dev);
 		struct journal_device *ja = &ca->journal;
 
 		if (!percpu_ref_tryget(&ca->io_ref)) {
@@ -1958,14 +1954,14 @@ static int bch2_journal_write_pick_flush(struct journal *j, struct journal_buf *
 	 * So if we're in an error state, and we're still starting up, we don't
 	 * write anything at all.
 	 */
-	if (error && test_bit(JOURNAL_NEED_FLUSH_WRITE, &j->flags))
+	if (error && test_bit(JOURNAL_need_flush_write, &j->flags))
 		return -EIO;
 
 	if (error ||
 	    w->noflush ||
 	    (!w->must_flush &&
 	     (jiffies - j->last_flush_write) < msecs_to_jiffies(c->opts.journal_flush_delay) &&
-	     test_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags))) {
+	     test_bit(JOURNAL_may_skip_flush, &j->flags))) {
 		w->noflush = true;
 		SET_JSET_NO_FLUSH(w->data, true);
 		w->data->last_seq	= 0;
@@ -1976,7 +1972,7 @@ static int bch2_journal_write_pick_flush(struct journal *j, struct journal_buf *
 		w->must_flush = true;
 		j->last_flush_write = jiffies;
 		j->nr_flush_writes++;
-		clear_bit(JOURNAL_NEED_FLUSH_WRITE, &j->flags);
+		clear_bit(JOURNAL_need_flush_write, &j->flags);
 	}
 
 	return 0;
diff --git a/fs/bcachefs/journal_io.h b/fs/bcachefs/journal_io.h
index 4f1e763ab50600..43cfed58ef33f0 100644
--- a/fs/bcachefs/journal_io.h
+++ b/fs/bcachefs/journal_io.h
@@ -4,6 +4,9 @@
 
 #include "darray.h"
 
+void bch2_journal_pos_from_member_info_set(struct bch_fs *);
+void bch2_journal_pos_from_member_info_resume(struct bch_fs *);
+
 struct journal_ptr {
 	bool		csum_good;
 	u8		dev;
diff --git a/fs/bcachefs/journal_reclaim.c b/fs/bcachefs/journal_reclaim.c
index 04a577848b015c..79be0eaddfa0d1 100644
--- a/fs/bcachefs/journal_reclaim.c
+++ b/fs/bcachefs/journal_reclaim.c
@@ -67,7 +67,7 @@ void bch2_journal_set_watermark(struct journal *j)
 	    track_event_change(&c->times[BCH_TIME_blocked_write_buffer_full], low_on_wb))
 		trace_and_count(c, journal_full, c);
 
-	mod_bit(JOURNAL_SPACE_LOW, &j->flags, low_on_space || low_on_pin);
+	mod_bit(JOURNAL_space_low, &j->flags, low_on_space || low_on_pin);
 
 	swap(watermark, j->watermark);
 	if (watermark > j->watermark)
@@ -225,9 +225,9 @@ void bch2_journal_space_available(struct journal *j)
 	     j->space[journal_space_clean_ondisk].total) &&
 	    (clean - clean_ondisk <= total / 8) &&
 	    (clean_ondisk * 2 > clean))
-		set_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags);
+		set_bit(JOURNAL_may_skip_flush, &j->flags);
 	else
-		clear_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags);
+		clear_bit(JOURNAL_may_skip_flush, &j->flags);
 
 	bch2_journal_set_watermark(j);
 out:
@@ -818,7 +818,7 @@ static int journal_flush_done(struct journal *j, u64 seq_to_flush,
 	 * If journal replay hasn't completed, the unreplayed journal entries
 	 * hold refs on their corresponding sequence numbers
 	 */
-	ret = !test_bit(JOURNAL_REPLAY_DONE, &j->flags) ||
+	ret = !test_bit(JOURNAL_replay_done, &j->flags) ||
 		journal_last_seq(j) > seq_to_flush ||
 		!fifo_used(&j->pin);
 
@@ -833,7 +833,7 @@ bool bch2_journal_flush_pins(struct journal *j, u64 seq_to_flush)
 	/* time_stats this */
 	bool did_work = false;
 
-	if (!test_bit(JOURNAL_STARTED, &j->flags))
+	if (!test_bit(JOURNAL_running, &j->flags))
 		return false;
 
 	closure_wait_event(&j->async_wait,
diff --git a/fs/bcachefs/journal_seq_blacklist.c b/fs/bcachefs/journal_seq_blacklist.c
index 37a024e034d495..d8d40d46a27a2a 100644
--- a/fs/bcachefs/journal_seq_blacklist.c
+++ b/fs/bcachefs/journal_seq_blacklist.c
@@ -1,8 +1,8 @@
 // SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
-#include "btree_iter.h"
 #include "eytzinger.h"
+#include "journal.h"
 #include "journal_seq_blacklist.h"
 #include "super-io.h"
 
@@ -217,78 +217,40 @@ const struct bch_sb_field_ops bch_sb_field_ops_journal_seq_blacklist = {
 	.to_text	= bch2_sb_journal_seq_blacklist_to_text
 };
 
-void bch2_blacklist_entries_gc(struct work_struct *work)
+bool bch2_blacklist_entries_gc(struct bch_fs *c)
 {
-	struct bch_fs *c = container_of(work, struct bch_fs,
-					journal_seq_blacklist_gc_work);
-	struct journal_seq_blacklist_table *t;
-	struct bch_sb_field_journal_seq_blacklist *bl;
 	struct journal_seq_blacklist_entry *src, *dst;
-	struct btree_trans *trans = bch2_trans_get(c);
-	unsigned i, nr, new_nr;
-	int ret;
-
-	for (i = 0; i < BTREE_ID_NR; i++) {
-		struct btree_iter iter;
-		struct btree *b;
-
-		bch2_trans_node_iter_init(trans, &iter, i, POS_MIN,
-					  0, 0, BTREE_ITER_PREFETCH);
-retry:
-		bch2_trans_begin(trans);
-
-		b = bch2_btree_iter_peek_node(&iter);
-
-		while (!(ret = PTR_ERR_OR_ZERO(b)) &&
-		       b &&
-		       !test_bit(BCH_FS_stopping, &c->flags))
-			b = bch2_btree_iter_next_node(&iter);
-
-		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
-			goto retry;
 
-		bch2_trans_iter_exit(trans, &iter);
-	}
-
-	bch2_trans_put(trans);
-	if (ret)
-		return;
-
-	mutex_lock(&c->sb_lock);
-	bl = bch2_sb_field_get(c->disk_sb.sb, journal_seq_blacklist);
+	struct bch_sb_field_journal_seq_blacklist *bl =
+		bch2_sb_field_get(c->disk_sb.sb, journal_seq_blacklist);
 	if (!bl)
-		goto out;
+		return false;
 
-	nr = blacklist_nr_entries(bl);
+	unsigned nr = blacklist_nr_entries(bl);
 	dst = bl->start;
 
-	t = c->journal_seq_blacklist_table;
+	struct journal_seq_blacklist_table *t = c->journal_seq_blacklist_table;
 	BUG_ON(nr != t->nr);
 
+	unsigned i;
 	for (src = bl->start, i = eytzinger0_first(t->nr);
 	     src < bl->start + nr;
 	     src++, i = eytzinger0_next(i, nr)) {
 		BUG_ON(t->entries[i].start	!= le64_to_cpu(src->start));
 		BUG_ON(t->entries[i].end	!= le64_to_cpu(src->end));
 
-		if (t->entries[i].dirty)
+		if (t->entries[i].dirty || t->entries[i].end >= c->journal.oldest_seq_found_ondisk)
 			*dst++ = *src;
 	}
 
-	new_nr = dst - bl->start;
-
-	bch_info(c, "nr blacklist entries was %u, now %u", nr, new_nr);
-
-	if (new_nr != nr) {
-		bl = bch2_sb_field_resize(&c->disk_sb, journal_seq_blacklist,
-				new_nr ? sb_blacklist_u64s(new_nr) : 0);
-		BUG_ON(new_nr && !bl);
+	unsigned new_nr = dst - bl->start;
+	if (new_nr == nr)
+		return false;
 
-		if (!new_nr)
-			c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_journal_seq_blacklist_v3));
+	bch_verbose(c, "nr blacklist entries was %u, now %u", nr, new_nr);
 
-		bch2_write_super(c);
-	}
-out:
-	mutex_unlock(&c->sb_lock);
+	bl = bch2_sb_field_resize(&c->disk_sb, journal_seq_blacklist,
+				  new_nr ? sb_blacklist_u64s(new_nr) : 0);
+	BUG_ON(new_nr && !bl);
+	return true;
 }
diff --git a/fs/bcachefs/journal_seq_blacklist.h b/fs/bcachefs/journal_seq_blacklist.h
index afb886ec8e254b..d47636f96fdc61 100644
--- a/fs/bcachefs/journal_seq_blacklist.h
+++ b/fs/bcachefs/journal_seq_blacklist.h
@@ -17,6 +17,6 @@ int bch2_blacklist_table_initialize(struct bch_fs *);
 
 extern const struct bch_sb_field_ops bch_sb_field_ops_journal_seq_blacklist;
 
-void bch2_blacklist_entries_gc(struct work_struct *);
+bool bch2_blacklist_entries_gc(struct bch_fs *);
 
 #endif /* _BCACHEFS_JOURNAL_SEQ_BLACKLIST_H */
diff --git a/fs/bcachefs/journal_types.h b/fs/bcachefs/journal_types.h
index b5161b5d76a008..19183fcf7ad7fd 100644
--- a/fs/bcachefs/journal_types.h
+++ b/fs/bcachefs/journal_types.h
@@ -129,12 +129,17 @@ enum journal_space_from {
 	journal_space_nr,
 };
 
+#define JOURNAL_FLAGS()			\
+	x(replay_done)			\
+	x(running)			\
+	x(may_skip_flush)		\
+	x(need_flush_write)		\
+	x(space_low)
+
 enum journal_flags {
-	JOURNAL_REPLAY_DONE,
-	JOURNAL_STARTED,
-	JOURNAL_MAY_SKIP_FLUSH,
-	JOURNAL_NEED_FLUSH_WRITE,
-	JOURNAL_SPACE_LOW,
+#define x(n)	JOURNAL_##n,
+	JOURNAL_FLAGS()
+#undef x
 };
 
 /* Reasons we may fail to get a journal reservation: */
@@ -229,6 +234,7 @@ struct journal {
 	u64			last_seq_ondisk;
 	u64			err_seq;
 	u64			last_empty_seq;
+	u64			oldest_seq_found_ondisk;
 
 	/*
 	 * FIFO of journal entries whose btree updates have not yet been
@@ -326,6 +332,7 @@ struct journal_device {
 
 	/* for bch_journal_read_device */
 	struct closure		read;
+	u64			highest_seq_found;
 };
 
 /*
diff --git a/fs/bcachefs/logged_ops.c b/fs/bcachefs/logged_ops.c
index b82f8209041ffb..f49fdca1d07dca 100644
--- a/fs/bcachefs/logged_ops.c
+++ b/fs/bcachefs/logged_ops.c
@@ -56,7 +56,7 @@ int bch2_resume_logged_ops(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter,
 				   BTREE_ID_logged_ops, POS_MIN,
-				   BTREE_ITER_PREFETCH, k,
+				   BTREE_ITER_prefetch, k,
 			resume_logged_op(trans, &iter, k)));
 	bch_err_fn(c, ret);
 	return ret;
diff --git a/fs/bcachefs/lru.c b/fs/bcachefs/lru.c
index 26569043e36803..19285fcd252c5f 100644
--- a/fs/bcachefs/lru.c
+++ b/fs/bcachefs/lru.c
@@ -149,7 +149,7 @@ int bch2_check_lrus(struct bch_fs *c)
 	struct bpos last_flushed_pos = POS_MIN;
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
-				BTREE_ID_lru, POS_MIN, BTREE_ITER_PREFETCH, k,
+				BTREE_ID_lru, POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc|BCH_TRANS_COMMIT_lazy_rw,
 			bch2_check_lru_key(trans, &iter, k, &last_flushed_pos)));
 	bch_err_fn(c, ret);
diff --git a/fs/bcachefs/migrate.c b/fs/bcachefs/migrate.c
index 69098eeb5d48e3..ddc187fb693d14 100644
--- a/fs/bcachefs/migrate.c
+++ b/fs/bcachefs/migrate.c
@@ -49,7 +49,7 @@ static int bch2_dev_usrdata_drop_key(struct btree_trans *trans,
 	if (!bch2_bkey_has_device_c(k, dev_idx))
 		return 0;
 
-	n = bch2_bkey_make_mut(trans, iter, &k, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+	n = bch2_bkey_make_mut(trans, iter, &k, BTREE_UPDATE_internal_snapshot_node);
 	ret = PTR_ERR_OR_ZERO(n);
 	if (ret)
 		return ret;
@@ -67,7 +67,7 @@ static int bch2_dev_usrdata_drop_key(struct btree_trans *trans,
 
 	/*
 	 * Since we're not inserting through an extent iterator
-	 * (BTREE_ITER_ALL_SNAPSHOTS iterators aren't extent iterators),
+	 * (BTREE_ITER_all_snapshots iterators aren't extent iterators),
 	 * we aren't using the extent overwrite path to delete, we're
 	 * just using the normal key deletion path:
 	 */
@@ -87,7 +87,7 @@ static int bch2_dev_usrdata_drop(struct bch_fs *c, unsigned dev_idx, int flags)
 			continue;
 
 		ret = for_each_btree_key_commit(trans, iter, id, POS_MIN,
-				BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+				BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			bch2_dev_usrdata_drop_key(trans, &iter, k, dev_idx, flags));
 		if (ret)
@@ -119,7 +119,7 @@ static int bch2_dev_metadata_drop(struct bch_fs *c, unsigned dev_idx, int flags)
 
 	for (id = 0; id < BTREE_ID_NR; id++) {
 		bch2_trans_node_iter_init(trans, &iter, id, POS_MIN, 0, 0,
-					  BTREE_ITER_PREFETCH);
+					  BTREE_ITER_prefetch);
 retry:
 		ret = 0;
 		while (bch2_trans_begin(trans),
diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index bf68ea49447b95..fbd4df3d8725e2 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -41,28 +41,23 @@ static void bch2_data_update_opts_to_text(struct printbuf *out, struct bch_fs *c
 					  struct data_update_opts *data_opts)
 {
 	printbuf_tabstop_push(out, 20);
-	prt_str(out, "rewrite ptrs:");
-	prt_tab(out);
+	prt_str(out, "rewrite ptrs:\t");
 	bch2_prt_u64_base2(out, data_opts->rewrite_ptrs);
 	prt_newline(out);
 
-	prt_str(out, "kill ptrs: ");
-	prt_tab(out);
+	prt_str(out, "kill ptrs:\t");
 	bch2_prt_u64_base2(out, data_opts->kill_ptrs);
 	prt_newline(out);
 
-	prt_str(out, "target: ");
-	prt_tab(out);
+	prt_str(out, "target:\t");
 	bch2_target_to_text(out, c, data_opts->target);
 	prt_newline(out);
 
-	prt_str(out, "compression: ");
-	prt_tab(out);
+	prt_str(out, "compression:\t");
 	bch2_compression_opt_to_text(out, background_compression(*io_opts));
 	prt_newline(out);
 
-	prt_str(out, "extra replicas: ");
-	prt_tab(out);
+	prt_str(out, "extra replicas:\t");
 	prt_u64(out, data_opts->extra_replicas);
 }
 
@@ -421,7 +416,7 @@ struct bch_io_opts *bch2_move_get_io_opts(struct btree_trans *trans,
 		io_opts->d.nr = 0;
 
 		ret = for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, extent_k.k->p.inode),
-					 BTREE_ITER_ALL_SNAPSHOTS, k, ({
+					 BTREE_ITER_all_snapshots, k, ({
 			if (k.k->p.offset != extent_k.k->p.inode)
 				break;
 
@@ -467,7 +462,7 @@ int bch2_move_get_io_opts_one(struct btree_trans *trans,
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
 			       SPOS(0, extent_k.k->p.inode, extent_k.k->p.snapshot),
-			       BTREE_ITER_CACHED);
+			       BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 		return ret;
@@ -553,8 +548,8 @@ static int bch2_move_data_btree(struct moving_context *ctxt,
 	}
 
 	bch2_trans_iter_init(trans, &iter, btree_id, start,
-			     BTREE_ITER_PREFETCH|
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_prefetch|
+			     BTREE_ITER_all_snapshots);
 
 	if (ctxt->rate)
 		bch2_ratelimit_reset(ctxt->rate);
@@ -695,6 +690,10 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
 	struct bpos bp_pos = POS_MIN;
 	int ret = 0;
 
+	struct bch_dev *ca = bch2_dev_tryget(c, bucket.inode);
+	if (!ca)
+		return 0;
+
 	trace_bucket_evacuate(c, &bucket);
 
 	bch2_bkey_buf_init(&sk);
@@ -705,7 +704,7 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
 	bch2_trans_begin(trans);
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
-			     bucket, BTREE_ITER_CACHED);
+			     bucket, BTREE_ITER_cached);
 	ret = lockrestart_do(trans,
 			bkey_err(k = bch2_btree_iter_peek_slot(&iter)));
 	bch2_trans_iter_exit(trans, &iter);
@@ -716,7 +715,7 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
 
 	a = bch2_alloc_to_v4(k, &a_convert);
 	dirty_sectors = bch2_bucket_sectors_dirty(*a);
-	bucket_size = bch_dev_bkey_exists(c, bucket.inode)->mi.bucket_size;
+	bucket_size = ca->mi.bucket_size;
 	fragmentation = a->fragmentation_lru;
 
 	ret = bch2_btree_write_buffer_tryflush(trans);
@@ -730,9 +729,9 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
 
 		bch2_trans_begin(trans);
 
-		ret = bch2_get_next_backpointer(trans, bucket, gen,
+		ret = bch2_get_next_backpointer(trans, ca, bucket, gen,
 						&bp_pos, &bp,
-						BTREE_ITER_CACHED);
+						BTREE_ITER_cached);
 		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 			continue;
 		if (ret)
@@ -828,6 +827,7 @@ next:
 
 	trace_evacuate_bucket(c, &bucket, dirty_sectors, bucket_size, fragmentation, ret);
 err:
+	bch2_dev_put(ca);
 	bch2_bkey_buf_exit(&sk, c);
 	return ret;
 }
@@ -868,7 +868,7 @@ static int bch2_move_btree(struct bch_fs *c,
 			continue;
 
 		bch2_trans_node_iter_init(trans, &iter, btree, POS_MIN, 0, 0,
-					  BTREE_ITER_PREFETCH);
+					  BTREE_ITER_prefetch);
 retry:
 		ret = 0;
 		while (bch2_trans_begin(trans),
@@ -1043,6 +1043,7 @@ static bool drop_extra_replicas_pred(struct bch_fs *c, void *arg,
 	struct extent_ptr_decoded p;
 	unsigned i = 0;
 
+	rcu_read_lock();
 	bkey_for_each_ptr_decode(k.k, bch2_bkey_ptrs_c(k), p, entry) {
 		unsigned d = bch2_extent_ptr_durability(c, &p);
 
@@ -1053,6 +1054,7 @@ static bool drop_extra_replicas_pred(struct bch_fs *c, void *arg,
 
 		i++;
 	}
+	rcu_read_unlock();
 
 	return data_opts->kill_ptrs != 0;
 }
@@ -1137,23 +1139,17 @@ void bch2_move_stats_to_text(struct printbuf *out, struct bch_move_stats *stats)
 	prt_newline(out);
 	printbuf_indent_add(out, 2);
 
-	prt_str(out, "keys moved:  ");
-	prt_u64(out, atomic64_read(&stats->keys_moved));
-	prt_newline(out);
-
-	prt_str(out, "keys raced:  ");
-	prt_u64(out, atomic64_read(&stats->keys_raced));
-	prt_newline(out);
-
-	prt_str(out, "bytes seen:  ");
+	prt_printf(out, "keys moved:  %llu\n",	atomic64_read(&stats->keys_moved));
+	prt_printf(out, "keys raced:  %llu\n",	atomic64_read(&stats->keys_raced));
+	prt_printf(out, "bytes seen:  ");
 	prt_human_readable_u64(out, atomic64_read(&stats->sectors_seen) << 9);
 	prt_newline(out);
 
-	prt_str(out, "bytes moved: ");
+	prt_printf(out, "bytes moved: ");
 	prt_human_readable_u64(out, atomic64_read(&stats->sectors_moved) << 9);
 	prt_newline(out);
 
-	prt_str(out, "bytes raced: ");
+	prt_printf(out, "bytes raced: ");
 	prt_human_readable_u64(out, atomic64_read(&stats->sectors_raced) << 9);
 	prt_newline(out);
 
@@ -1167,19 +1163,17 @@ static void bch2_moving_ctxt_to_text(struct printbuf *out, struct bch_fs *c, str
 	bch2_move_stats_to_text(out, ctxt->stats);
 	printbuf_indent_add(out, 2);
 
-	prt_printf(out, "reads: ios %u/%u sectors %u/%u",
+	prt_printf(out, "reads: ios %u/%u sectors %u/%u\n",
 		   atomic_read(&ctxt->read_ios),
 		   c->opts.move_ios_in_flight,
 		   atomic_read(&ctxt->read_sectors),
 		   c->opts.move_bytes_in_flight >> 9);
-	prt_newline(out);
 
-	prt_printf(out, "writes: ios %u/%u sectors %u/%u",
+	prt_printf(out, "writes: ios %u/%u sectors %u/%u\n",
 		   atomic_read(&ctxt->write_ios),
 		   c->opts.move_ios_in_flight,
 		   atomic_read(&ctxt->write_sectors),
 		   c->opts.move_bytes_in_flight >> 9);
-	prt_newline(out);
 
 	printbuf_indent_add(out, 2);
 
diff --git a/fs/bcachefs/movinggc.c b/fs/bcachefs/movinggc.c
index 0d2b82d8d11f39..10bfb31c151b20 100644
--- a/fs/bcachefs/movinggc.c
+++ b/fs/bcachefs/movinggc.c
@@ -84,7 +84,7 @@ static int bch2_bucket_is_movable(struct btree_trans *trans,
 		return 0;
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
-			       b->k.bucket, BTREE_ITER_CACHED);
+			       b->k.bucket, BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (ret)
 		return ret;
@@ -158,6 +158,8 @@ static int bch2_copygc_get_buckets(struct moving_context *ctxt,
 	if (bch2_fs_fatal_err_on(ret, c, "%s: from bch2_btree_write_buffer_tryflush()", bch2_err_str(ret)))
 		return ret;
 
+	bch2_trans_begin(trans);
+
 	ret = for_each_btree_key_upto(trans, iter, BTREE_ID_lru,
 				  lru_pos(BCH_LRU_FRAGMENTATION_START, 0, 0),
 				  lru_pos(BCH_LRU_FRAGMENTATION_START, U64_MAX, LRU_TIME_MAX),
diff --git a/fs/bcachefs/opts.h b/fs/bcachefs/opts.h
index 84e452835a17d8..25530e0bb2f386 100644
--- a/fs/bcachefs/opts.h
+++ b/fs/bcachefs/opts.h
@@ -426,11 +426,6 @@ enum fsck_err_opts {
 	  BCH_SB_VERSION_UPGRADE,	BCH_VERSION_UPGRADE_compatible,	\
 	  NULL,		"Set superblock to latest version,\n"		\
 			"allowing any new features to be used")		\
-	x(buckets_nouse,		u8,				\
-	  0,								\
-	  OPT_BOOL(),							\
-	  BCH2_NO_SB_OPT,		false,				\
-	  NULL,		"Allocate the buckets_nouse bitmap")		\
 	x(stdio,			u64,				\
 	  0,								\
 	  OPT_UINT(0, S64_MAX),						\
@@ -480,7 +475,7 @@ enum fsck_err_opts {
 	  OPT_FS|OPT_MOUNT|OPT_RUNTIME,					\
 	  OPT_BOOL(),							\
 	  BCH2_NO_SB_OPT,		true,				\
-	  NULL,		"BTREE_ITER_PREFETCH casuse btree nodes to be\n"\
+	  NULL,		"BTREE_ITER_prefetch casuse btree nodes to be\n"\
 	  " prefetched sequentially")
 
 struct bch_opts {
diff --git a/fs/bcachefs/printbuf.c b/fs/bcachefs/printbuf.c
index b27d22925929a6..8b0369185f5cfd 100644
--- a/fs/bcachefs/printbuf.c
+++ b/fs/bcachefs/printbuf.c
@@ -10,35 +10,50 @@
 
 #include "printbuf.h"
 
-static inline unsigned printbuf_linelen(struct printbuf *buf)
+static inline unsigned __printbuf_linelen(struct printbuf *buf, unsigned pos)
 {
-	return buf->pos - buf->last_newline;
+	return pos - buf->last_newline;
 }
 
-int bch2_printbuf_make_room(struct printbuf *out, unsigned extra)
+static inline unsigned printbuf_linelen(struct printbuf *buf)
 {
-	unsigned new_size;
-	char *buf;
+	return __printbuf_linelen(buf, buf->pos);
+}
 
-	if (!out->heap_allocated)
-		return 0;
+/*
+ * Returns spaces from start of line, if set, or 0 if unset:
+ */
+static inline unsigned cur_tabstop(struct printbuf *buf)
+{
+	return buf->cur_tabstop < buf->nr_tabstops
+		? buf->_tabstops[buf->cur_tabstop]
+		: 0;
+}
 
+int bch2_printbuf_make_room(struct printbuf *out, unsigned extra)
+{
 	/* Reserved space for terminating nul: */
 	extra += 1;
 
-	if (out->pos + extra < out->size)
+	if (out->pos + extra <= out->size)
 		return 0;
 
-	new_size = roundup_pow_of_two(out->size + extra);
+	if (!out->heap_allocated) {
+		out->overflow = true;
+		return 0;
+	}
+
+	unsigned new_size = roundup_pow_of_two(out->size + extra);
 
 	/*
 	 * Note: output buffer must be freeable with kfree(), it's not required
 	 * that the user use printbuf_exit().
 	 */
-	buf = krealloc(out->buf, new_size, !out->atomic ? GFP_KERNEL : GFP_NOWAIT);
+	char *buf = krealloc(out->buf, new_size, !out->atomic ? GFP_KERNEL : GFP_NOWAIT);
 
 	if (!buf) {
 		out->allocation_failure = true;
+		out->overflow = true;
 		return -ENOMEM;
 	}
 
@@ -47,6 +62,92 @@ int bch2_printbuf_make_room(struct printbuf *out, unsigned extra)
 	return 0;
 }
 
+static void printbuf_advance_pos(struct printbuf *out, unsigned len)
+{
+	out->pos += min(len, printbuf_remaining(out));
+}
+
+static void printbuf_insert_spaces(struct printbuf *out, unsigned pos, unsigned nr)
+{
+	unsigned move = out->pos - pos;
+
+	bch2_printbuf_make_room(out, nr);
+
+	if (pos + nr < out->size)
+		memmove(out->buf + pos + nr,
+			out->buf + pos,
+			min(move, out->size - 1 - pos - nr));
+
+	if (pos < out->size)
+		memset(out->buf + pos, ' ', min(nr, out->size - pos));
+
+	printbuf_advance_pos(out, nr);
+	printbuf_nul_terminate_reserved(out);
+}
+
+static void __printbuf_do_indent(struct printbuf *out, unsigned pos)
+{
+	while (true) {
+		int pad;
+		unsigned len = out->pos - pos;
+		char *p = out->buf + pos;
+		char *n = memscan(p, '\n', len);
+		if (cur_tabstop(out)) {
+			n = min(n, (char *) memscan(p, '\r', len));
+			n = min(n, (char *) memscan(p, '\t', len));
+		}
+
+		pos = n - out->buf;
+		if (pos == out->pos)
+			break;
+
+		switch (*n) {
+		case '\n':
+			pos++;
+			out->last_newline = pos;
+
+			printbuf_insert_spaces(out, pos, out->indent);
+
+			pos = min(pos + out->indent, out->pos);
+			out->last_field = pos;
+			out->cur_tabstop = 0;
+			break;
+		case '\r':
+			memmove(n, n + 1, out->pos - pos);
+			--out->pos;
+			pad = (int) cur_tabstop(out) - (int) __printbuf_linelen(out, pos);
+			if (pad > 0) {
+				printbuf_insert_spaces(out, out->last_field, pad);
+				pos += pad;
+			}
+
+			out->last_field = pos;
+			out->cur_tabstop++;
+			break;
+		case '\t':
+			pad = (int) cur_tabstop(out) - (int) __printbuf_linelen(out, pos) - 1;
+			if (pad > 0) {
+				*n = ' ';
+				printbuf_insert_spaces(out, pos, pad - 1);
+				pos += pad;
+			} else {
+				memmove(n, n + 1, out->pos - pos);
+				--out->pos;
+			}
+
+			out->last_field = pos;
+			out->cur_tabstop++;
+			break;
+		}
+	}
+}
+
+static inline void printbuf_do_indent(struct printbuf *out, unsigned pos)
+{
+	if (out->has_indent_or_tabstops && !out->suppress_indent_tabstop_handling)
+		__printbuf_do_indent(out, pos);
+}
+
 void bch2_prt_vprintf(struct printbuf *out, const char *fmt, va_list args)
 {
 	int len;
@@ -55,14 +156,14 @@ void bch2_prt_vprintf(struct printbuf *out, const char *fmt, va_list args)
 		va_list args2;
 
 		va_copy(args2, args);
-		len = vsnprintf(out->buf + out->pos, printbuf_remaining(out), fmt, args2);
+		len = vsnprintf(out->buf + out->pos, printbuf_remaining_size(out), fmt, args2);
 		va_end(args2);
-	} while (len + 1 >= printbuf_remaining(out) &&
-		 !bch2_printbuf_make_room(out, len + 1));
+	} while (len > printbuf_remaining(out) &&
+		 !bch2_printbuf_make_room(out, len));
 
-	len = min_t(size_t, len,
-		  printbuf_remaining(out) ? printbuf_remaining(out) - 1 : 0);
-	out->pos += len;
+	unsigned indent_pos = out->pos;
+	printbuf_advance_pos(out, len);
+	printbuf_do_indent(out, indent_pos);
 }
 
 void bch2_prt_printf(struct printbuf *out, const char *fmt, ...)
@@ -72,14 +173,14 @@ void bch2_prt_printf(struct printbuf *out, const char *fmt, ...)
 
 	do {
 		va_start(args, fmt);
-		len = vsnprintf(out->buf + out->pos, printbuf_remaining(out), fmt, args);
+		len = vsnprintf(out->buf + out->pos, printbuf_remaining_size(out), fmt, args);
 		va_end(args);
-	} while (len + 1 >= printbuf_remaining(out) &&
-		 !bch2_printbuf_make_room(out, len + 1));
+	} while (len > printbuf_remaining(out) &&
+		 !bch2_printbuf_make_room(out, len));
 
-	len = min_t(size_t, len,
-		  printbuf_remaining(out) ? printbuf_remaining(out) - 1 : 0);
-	out->pos += len;
+	unsigned indent_pos = out->pos;
+	printbuf_advance_pos(out, len);
+	printbuf_do_indent(out, indent_pos);
 }
 
 /**
@@ -194,33 +295,20 @@ void bch2_printbuf_indent_sub(struct printbuf *buf, unsigned spaces)
 
 void bch2_prt_newline(struct printbuf *buf)
 {
-	unsigned i;
-
 	bch2_printbuf_make_room(buf, 1 + buf->indent);
 
-	__prt_char(buf, '\n');
+	__prt_char_reserved(buf, '\n');
 
 	buf->last_newline	= buf->pos;
 
-	for (i = 0; i < buf->indent; i++)
-		__prt_char(buf, ' ');
+	__prt_chars_reserved(buf, ' ', buf->indent);
 
-	printbuf_nul_terminate(buf);
+	printbuf_nul_terminate_reserved(buf);
 
 	buf->last_field		= buf->pos;
 	buf->cur_tabstop	= 0;
 }
 
-/*
- * Returns spaces from start of line, if set, or 0 if unset:
- */
-static inline unsigned cur_tabstop(struct printbuf *buf)
-{
-	return buf->cur_tabstop < buf->nr_tabstops
-		? buf->_tabstops[buf->cur_tabstop]
-		: 0;
-}
-
 static void __prt_tab(struct printbuf *out)
 {
 	int spaces = max_t(int, 0, cur_tabstop(out) - printbuf_linelen(out));
@@ -247,24 +335,9 @@ void bch2_prt_tab(struct printbuf *out)
 
 static void __prt_tab_rjust(struct printbuf *buf)
 {
-	unsigned move = buf->pos - buf->last_field;
 	int pad = (int) cur_tabstop(buf) - (int) printbuf_linelen(buf);
-
-	if (pad > 0) {
-		bch2_printbuf_make_room(buf, pad);
-
-		if (buf->last_field + pad < buf->size)
-			memmove(buf->buf + buf->last_field + pad,
-				buf->buf + buf->last_field,
-				min(move, buf->size - 1 - buf->last_field - pad));
-
-		if (buf->last_field < buf->size)
-			memset(buf->buf + buf->last_field, ' ',
-			       min((unsigned) pad, buf->size - buf->last_field));
-
-		buf->pos += pad;
-		printbuf_nul_terminate(buf);
-	}
+	if (pad > 0)
+		printbuf_insert_spaces(buf, buf->last_field, pad);
 
 	buf->last_field = buf->pos;
 	buf->cur_tabstop++;
@@ -301,41 +374,9 @@ void bch2_prt_tab_rjust(struct printbuf *buf)
  */
 void bch2_prt_bytes_indented(struct printbuf *out, const char *str, unsigned count)
 {
-	const char *unprinted_start = str;
-	const char *end = str + count;
-
-	if (!out->has_indent_or_tabstops || out->suppress_indent_tabstop_handling) {
-		prt_bytes(out, str, count);
-		return;
-	}
-
-	while (str != end) {
-		switch (*str) {
-		case '\n':
-			prt_bytes(out, unprinted_start, str - unprinted_start);
-			unprinted_start = str + 1;
-			bch2_prt_newline(out);
-			break;
-		case '\t':
-			if (likely(cur_tabstop(out))) {
-				prt_bytes(out, unprinted_start, str - unprinted_start);
-				unprinted_start = str + 1;
-				__prt_tab(out);
-			}
-			break;
-		case '\r':
-			if (likely(cur_tabstop(out))) {
-				prt_bytes(out, unprinted_start, str - unprinted_start);
-				unprinted_start = str + 1;
-				__prt_tab_rjust(out);
-			}
-			break;
-		}
-
-		str++;
-	}
-
-	prt_bytes(out, unprinted_start, str - unprinted_start);
+	unsigned indent_pos = out->pos;
+	prt_bytes(out, str, count);
+	printbuf_do_indent(out, indent_pos);
 }
 
 /**
@@ -348,9 +389,10 @@ void bch2_prt_bytes_indented(struct printbuf *out, const char *str, unsigned cou
 void bch2_prt_human_readable_u64(struct printbuf *out, u64 v)
 {
 	bch2_printbuf_make_room(out, 10);
-	out->pos += string_get_size(v, 1, !out->si_units,
-				    out->buf + out->pos,
-				    printbuf_remaining_size(out));
+	unsigned len = string_get_size(v, 1, !out->si_units,
+				       out->buf + out->pos,
+				       printbuf_remaining_size(out));
+	printbuf_advance_pos(out, len);
 }
 
 /**
@@ -402,9 +444,7 @@ void bch2_prt_string_option(struct printbuf *out,
 			    const char * const list[],
 			    size_t selected)
 {
-	size_t i;
-
-	for (i = 0; list[i]; i++)
+	for (size_t i = 0; list[i]; i++)
 		bch2_prt_printf(out, i == selected ? "[%s] " : "%s ", list[i]);
 }
 
diff --git a/fs/bcachefs/printbuf.h b/fs/bcachefs/printbuf.h
index 9a4a56c4093715..9ecc56bc96359a 100644
--- a/fs/bcachefs/printbuf.h
+++ b/fs/bcachefs/printbuf.h
@@ -86,6 +86,7 @@ struct printbuf {
 	u8			atomic;
 	bool			allocation_failure:1;
 	bool			heap_allocated:1;
+	bool			overflow:1;
 	enum printbuf_si	si_units:1;
 	bool			human_readable_units:1;
 	bool			has_indent_or_tabstops:1;
@@ -142,7 +143,9 @@ void bch2_prt_bitflags_vector(struct printbuf *, const char * const[],
  */
 static inline unsigned printbuf_remaining_size(struct printbuf *out)
 {
-	return out->pos < out->size ? out->size - out->pos : 0;
+	if (WARN_ON(out->size && out->pos >= out->size))
+		out->pos = out->size - 1;
+	return out->size - out->pos;
 }
 
 /*
@@ -151,7 +154,7 @@ static inline unsigned printbuf_remaining_size(struct printbuf *out)
  */
 static inline unsigned printbuf_remaining(struct printbuf *out)
 {
-	return out->pos < out->size ? out->size - out->pos - 1 : 0;
+	return out->size ? printbuf_remaining_size(out) - 1 : 0;
 }
 
 static inline unsigned printbuf_written(struct printbuf *out)
@@ -159,30 +162,25 @@ static inline unsigned printbuf_written(struct printbuf *out)
 	return out->size ? min(out->pos, out->size - 1) : 0;
 }
 
-/*
- * Returns true if output was truncated:
- */
-static inline bool printbuf_overflowed(struct printbuf *out)
+static inline void printbuf_nul_terminate_reserved(struct printbuf *out)
 {
-	return out->pos >= out->size;
+	if (WARN_ON(out->size && out->pos >= out->size))
+		out->pos = out->size - 1;
+	if (out->size)
+		out->buf[out->pos] = 0;
 }
 
 static inline void printbuf_nul_terminate(struct printbuf *out)
 {
 	bch2_printbuf_make_room(out, 1);
-
-	if (out->pos < out->size)
-		out->buf[out->pos] = 0;
-	else if (out->size)
-		out->buf[out->size - 1] = 0;
+	printbuf_nul_terminate_reserved(out);
 }
 
 /* Doesn't call bch2_printbuf_make_room(), doesn't nul terminate: */
 static inline void __prt_char_reserved(struct printbuf *out, char c)
 {
 	if (printbuf_remaining(out))
-		out->buf[out->pos] = c;
-	out->pos++;
+		out->buf[out->pos++] = c;
 }
 
 /* Doesn't nul terminate: */
@@ -194,37 +192,34 @@ static inline void __prt_char(struct printbuf *out, char c)
 
 static inline void prt_char(struct printbuf *out, char c)
 {
-	__prt_char(out, c);
-	printbuf_nul_terminate(out);
+	bch2_printbuf_make_room(out, 2);
+	__prt_char_reserved(out, c);
+	printbuf_nul_terminate_reserved(out);
 }
 
 static inline void __prt_chars_reserved(struct printbuf *out, char c, unsigned n)
 {
-	unsigned i, can_print = min(n, printbuf_remaining(out));
+	unsigned can_print = min(n, printbuf_remaining(out));
 
-	for (i = 0; i < can_print; i++)
+	for (unsigned i = 0; i < can_print; i++)
 		out->buf[out->pos++] = c;
-	out->pos += n - can_print;
 }
 
 static inline void prt_chars(struct printbuf *out, char c, unsigned n)
 {
 	bch2_printbuf_make_room(out, n);
 	__prt_chars_reserved(out, c, n);
-	printbuf_nul_terminate(out);
+	printbuf_nul_terminate_reserved(out);
 }
 
 static inline void prt_bytes(struct printbuf *out, const void *b, unsigned n)
 {
-	unsigned i, can_print;
-
 	bch2_printbuf_make_room(out, n);
 
-	can_print = min(n, printbuf_remaining(out));
+	unsigned can_print = min(n, printbuf_remaining(out));
 
-	for (i = 0; i < can_print; i++)
+	for (unsigned i = 0; i < can_print; i++)
 		out->buf[out->pos++] = ((char *) b)[i];
-	out->pos += n - can_print;
 
 	printbuf_nul_terminate(out);
 }
@@ -241,18 +236,18 @@ static inline void prt_str_indented(struct printbuf *out, const char *str)
 
 static inline void prt_hex_byte(struct printbuf *out, u8 byte)
 {
-	bch2_printbuf_make_room(out, 2);
+	bch2_printbuf_make_room(out, 3);
 	__prt_char_reserved(out, hex_asc_hi(byte));
 	__prt_char_reserved(out, hex_asc_lo(byte));
-	printbuf_nul_terminate(out);
+	printbuf_nul_terminate_reserved(out);
 }
 
 static inline void prt_hex_byte_upper(struct printbuf *out, u8 byte)
 {
-	bch2_printbuf_make_room(out, 2);
+	bch2_printbuf_make_room(out, 3);
 	__prt_char_reserved(out, hex_asc_upper_hi(byte));
 	__prt_char_reserved(out, hex_asc_upper_lo(byte));
-	printbuf_nul_terminate(out);
+	printbuf_nul_terminate_reserved(out);
 }
 
 /**
diff --git a/fs/bcachefs/quota.c b/fs/bcachefs/quota.c
index e68b34eab90a91..da7c8fd61bbfed 100644
--- a/fs/bcachefs/quota.c
+++ b/fs/bcachefs/quota.c
@@ -97,45 +97,14 @@ static void qc_info_to_text(struct printbuf *out, struct qc_info *i)
 	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, 20);
 
-	prt_str(out, "i_fieldmask");
-	prt_tab(out);
-	prt_printf(out, "%x", i->i_fieldmask);
-	prt_newline(out);
-
-	prt_str(out, "i_flags");
-	prt_tab(out);
-	prt_printf(out, "%u", i->i_flags);
-	prt_newline(out);
-
-	prt_str(out, "i_spc_timelimit");
-	prt_tab(out);
-	prt_printf(out, "%u", i->i_spc_timelimit);
-	prt_newline(out);
-
-	prt_str(out, "i_ino_timelimit");
-	prt_tab(out);
-	prt_printf(out, "%u", i->i_ino_timelimit);
-	prt_newline(out);
-
-	prt_str(out, "i_rt_spc_timelimit");
-	prt_tab(out);
-	prt_printf(out, "%u", i->i_rt_spc_timelimit);
-	prt_newline(out);
-
-	prt_str(out, "i_spc_warnlimit");
-	prt_tab(out);
-	prt_printf(out, "%u", i->i_spc_warnlimit);
-	prt_newline(out);
-
-	prt_str(out, "i_ino_warnlimit");
-	prt_tab(out);
-	prt_printf(out, "%u", i->i_ino_warnlimit);
-	prt_newline(out);
-
-	prt_str(out, "i_rt_spc_warnlimit");
-	prt_tab(out);
-	prt_printf(out, "%u", i->i_rt_spc_warnlimit);
-	prt_newline(out);
+	prt_printf(out, "i_fieldmask\t%x\n",		i->i_fieldmask);
+	prt_printf(out, "i_flags\t%u\n",		i->i_flags);
+	prt_printf(out, "i_spc_timelimit\t%u\n",	i->i_spc_timelimit);
+	prt_printf(out, "i_ino_timelimit\t%u\n",	i->i_ino_timelimit);
+	prt_printf(out, "i_rt_spc_timelimit\t%u\n",	i->i_rt_spc_timelimit);
+	prt_printf(out, "i_spc_warnlimit\t%u\n",	i->i_spc_warnlimit);
+	prt_printf(out, "i_ino_warnlimit\t%u\n",	i->i_ino_warnlimit);
+	prt_printf(out, "i_rt_spc_warnlimit\t%u\n",	i->i_rt_spc_warnlimit);
 }
 
 static void qc_dqblk_to_text(struct printbuf *out, struct qc_dqblk *q)
@@ -143,60 +112,17 @@ static void qc_dqblk_to_text(struct printbuf *out, struct qc_dqblk *q)
 	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, 20);
 
-	prt_str(out, "d_fieldmask");
-	prt_tab(out);
-	prt_printf(out, "%x", q->d_fieldmask);
-	prt_newline(out);
-
-	prt_str(out, "d_spc_hardlimit");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_spc_hardlimit);
-	prt_newline(out);
-
-	prt_str(out, "d_spc_softlimit");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_spc_softlimit);
-	prt_newline(out);
-
-	prt_str(out, "d_ino_hardlimit");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_ino_hardlimit);
-	prt_newline(out);
-
-	prt_str(out, "d_ino_softlimit");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_ino_softlimit);
-	prt_newline(out);
-
-	prt_str(out, "d_space");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_space);
-	prt_newline(out);
-
-	prt_str(out, "d_ino_count");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_ino_count);
-	prt_newline(out);
-
-	prt_str(out, "d_ino_timer");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_ino_timer);
-	prt_newline(out);
-
-	prt_str(out, "d_spc_timer");
-	prt_tab(out);
-	prt_printf(out, "%llu", q->d_spc_timer);
-	prt_newline(out);
-
-	prt_str(out, "d_ino_warns");
-	prt_tab(out);
-	prt_printf(out, "%i", q->d_ino_warns);
-	prt_newline(out);
-
-	prt_str(out, "d_spc_warns");
-	prt_tab(out);
-	prt_printf(out, "%i", q->d_spc_warns);
-	prt_newline(out);
+	prt_printf(out, "d_fieldmask\t%x\n",		q->d_fieldmask);
+	prt_printf(out, "d_spc_hardlimit\t%llu\n",	q->d_spc_hardlimit);
+	prt_printf(out, "d_spc_softlimit\t%llu\n",	q->d_spc_softlimit);
+	prt_printf(out, "d_ino_hardlimit\%llu\n",	q->d_ino_hardlimit);
+	prt_printf(out, "d_ino_softlimit\t%llu\n",	q->d_ino_softlimit);
+	prt_printf(out, "d_space\t%llu\n",		q->d_space);
+	prt_printf(out, "d_ino_count\t%llu\n",		q->d_ino_count);
+	prt_printf(out, "d_ino_timer\t%llu\n",		q->d_ino_timer);
+	prt_printf(out, "d_spc_timer\t%llu\n",		q->d_spc_timer);
+	prt_printf(out, "d_ino_warns\t%i\n",		q->d_ino_warns);
+	prt_printf(out, "d_spc_warns\t%i\n",		q->d_spc_warns);
 }
 
 static inline unsigned __next_qtype(unsigned i, unsigned qtypes)
@@ -612,10 +538,10 @@ int bch2_fs_quota_read(struct bch_fs *c)
 
 	int ret = bch2_trans_run(c,
 		for_each_btree_key(trans, iter, BTREE_ID_quotas, POS_MIN,
-				   BTREE_ITER_PREFETCH, k,
+				   BTREE_ITER_prefetch, k,
 			__bch2_quota_set(c, k, NULL)) ?:
 		for_each_btree_key(trans, iter, BTREE_ID_inodes, POS_MIN,
-				   BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+				   BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 			bch2_fs_quota_read_inode(trans, &iter, k)));
 	bch_err_fn(c, ret);
 	return ret;
@@ -902,7 +828,7 @@ static int bch2_set_quota_trans(struct btree_trans *trans,
 	int ret;
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_quotas, new_quota->k.p,
-			       BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
+			       BTREE_ITER_slots|BTREE_ITER_intent);
 	ret = bkey_err(k);
 	if (unlikely(ret))
 		return ret;
diff --git a/fs/bcachefs/rebalance.c b/fs/bcachefs/rebalance.c
index 56336f3dd1d077..cf81e5128c3abb 100644
--- a/fs/bcachefs/rebalance.c
+++ b/fs/bcachefs/rebalance.c
@@ -42,7 +42,7 @@ static int __bch2_set_rebalance_needs_scan(struct btree_trans *trans, u64 inum)
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_rebalance_work,
 			     SPOS(inum, REBALANCE_WORK_SCAN_OFFSET, U32_MAX),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	k = bch2_btree_iter_peek_slot(&iter);
 	ret = bkey_err(k);
 	if (ret)
@@ -89,7 +89,7 @@ static int bch2_clear_rebalance_needs_scan(struct btree_trans *trans, u64 inum,
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_rebalance_work,
 			     SPOS(inum, REBALANCE_WORK_SCAN_OFFSET, U32_MAX),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	k = bch2_btree_iter_peek_slot(&iter);
 	ret = bkey_err(k);
 	if (ret)
@@ -140,7 +140,7 @@ static struct bkey_s_c next_rebalance_extent(struct btree_trans *trans,
 	bch2_trans_iter_init(trans, extent_iter,
 			     work_pos.inode ? BTREE_ID_extents : BTREE_ID_reflink,
 			     work_pos,
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_all_snapshots);
 	k = bch2_btree_iter_peek_slot(extent_iter);
 	if (bkey_err(k))
 		return k;
@@ -323,12 +323,14 @@ static int do_rebalance(struct moving_context *ctxt)
 	struct bkey_s_c k;
 	int ret = 0;
 
+	bch2_trans_begin(trans);
+
 	bch2_move_stats_init(&r->work_stats, "rebalance_work");
 	bch2_move_stats_init(&r->scan_stats, "rebalance_scan");
 
 	bch2_trans_iter_init(trans, &rebalance_work_iter,
 			     BTREE_ID_rebalance_work, POS_MIN,
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_all_snapshots);
 
 	while (!bch2_move_ratelimit(ctxt)) {
 		if (!r->enabled) {
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index be5b4761932700..5eeb6762c1f5a8 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -65,9 +65,20 @@ static void bch2_reconstruct_alloc(struct bch_fs *c)
 	__set_bit_le64(BCH_FSCK_ERR_ptr_to_missing_alloc_key, ext->errors_silent);
 	__set_bit_le64(BCH_FSCK_ERR_ptr_gen_newer_than_bucket_gen, ext->errors_silent);
 	__set_bit_le64(BCH_FSCK_ERR_stale_dirty_ptr, ext->errors_silent);
+
+	__set_bit_le64(BCH_FSCK_ERR_dev_usage_buckets_wrong, ext->errors_silent);
+	__set_bit_le64(BCH_FSCK_ERR_dev_usage_sectors_wrong, ext->errors_silent);
+	__set_bit_le64(BCH_FSCK_ERR_dev_usage_fragmented_wrong, ext->errors_silent);
+
+	__set_bit_le64(BCH_FSCK_ERR_fs_usage_btree_wrong, ext->errors_silent);
+	__set_bit_le64(BCH_FSCK_ERR_fs_usage_cached_wrong, ext->errors_silent);
+	__set_bit_le64(BCH_FSCK_ERR_fs_usage_persistent_reserved_wrong, ext->errors_silent);
+	__set_bit_le64(BCH_FSCK_ERR_fs_usage_replicas_wrong, ext->errors_silent);
+
 	__set_bit_le64(BCH_FSCK_ERR_alloc_key_data_type_wrong, ext->errors_silent);
 	__set_bit_le64(BCH_FSCK_ERR_alloc_key_gen_wrong, ext->errors_silent);
 	__set_bit_le64(BCH_FSCK_ERR_alloc_key_dirty_sectors_wrong, ext->errors_silent);
+	__set_bit_le64(BCH_FSCK_ERR_alloc_key_cached_sectors_wrong, ext->errors_silent);
 	__set_bit_le64(BCH_FSCK_ERR_alloc_key_stripe_wrong, ext->errors_silent);
 	__set_bit_le64(BCH_FSCK_ERR_alloc_key_stripe_redundancy_wrong, ext->errors_silent);
 	__set_bit_le64(BCH_FSCK_ERR_need_discard_key_wrong, ext->errors_silent);
@@ -125,9 +136,9 @@ static int bch2_journal_replay_key(struct btree_trans *trans,
 {
 	struct btree_iter iter;
 	unsigned iter_flags =
-		BTREE_ITER_INTENT|
-		BTREE_ITER_NOT_EXTENTS;
-	unsigned update_flags = BTREE_TRIGGER_NORUN;
+		BTREE_ITER_intent|
+		BTREE_ITER_not_extents;
+	unsigned update_flags = BTREE_TRIGGER_norun;
 	int ret;
 
 	if (k->overwritten)
@@ -136,17 +147,17 @@ static int bch2_journal_replay_key(struct btree_trans *trans,
 	trans->journal_res.seq = k->journal_seq;
 
 	/*
-	 * BTREE_UPDATE_KEY_CACHE_RECLAIM disables key cache lookup/update to
+	 * BTREE_UPDATE_key_cache_reclaim disables key cache lookup/update to
 	 * keep the key cache coherent with the underlying btree. Nothing
 	 * besides the allocator is doing updates yet so we don't need key cache
 	 * coherency for non-alloc btrees, and key cache fills for snapshots
-	 * btrees use BTREE_ITER_FILTER_SNAPSHOTS, which isn't available until
+	 * btrees use BTREE_ITER_filter_snapshots, which isn't available until
 	 * the snapshots recovery pass runs.
 	 */
 	if (!k->level && k->btree_id == BTREE_ID_alloc)
-		iter_flags |= BTREE_ITER_CACHED;
+		iter_flags |= BTREE_ITER_cached;
 	else
-		update_flags |= BTREE_UPDATE_KEY_CACHE_RECLAIM;
+		update_flags |= BTREE_UPDATE_key_cache_reclaim;
 
 	bch2_trans_node_iter_init(trans, &iter, k->btree_id, k->k->k.p,
 				  BTREE_MAX_DEPTH, k->level,
@@ -191,7 +202,7 @@ int bch2_journal_replay(struct bch_fs *c)
 	struct journal *j = &c->journal;
 	u64 start_seq	= c->journal_replay_seq_start;
 	u64 end_seq	= c->journal_replay_seq_start;
-	struct btree_trans *trans = bch2_trans_get(c);
+	struct btree_trans *trans = NULL;
 	bool immediate_flush = false;
 	int ret = 0;
 
@@ -205,6 +216,7 @@ int bch2_journal_replay(struct bch_fs *c)
 	BUG_ON(!atomic_read(&keys->ref));
 
 	move_gap(keys, keys->nr);
+	trans = bch2_trans_get(c);
 
 	/*
 	 * First, attempt to replay keys in sorted order. This is more
@@ -361,14 +373,17 @@ static int journal_replay_entry_early(struct bch_fs *c,
 	case BCH_JSET_ENTRY_dev_usage: {
 		struct jset_entry_dev_usage *u =
 			container_of(entry, struct jset_entry_dev_usage, entry);
-		struct bch_dev *ca = bch_dev_bkey_exists(c, le32_to_cpu(u->dev));
-		unsigned i, nr_types = jset_entry_dev_usage_nr_types(u);
-
-		for (i = 0; i < min_t(unsigned, nr_types, BCH_DATA_NR); i++) {
-			ca->usage_base->d[i].buckets	= le64_to_cpu(u->d[i].buckets);
-			ca->usage_base->d[i].sectors	= le64_to_cpu(u->d[i].sectors);
-			ca->usage_base->d[i].fragmented	= le64_to_cpu(u->d[i].fragmented);
-		}
+		unsigned nr_types = jset_entry_dev_usage_nr_types(u);
+
+		rcu_read_lock();
+		struct bch_dev *ca = bch2_dev_rcu(c, le32_to_cpu(u->dev));
+		if (ca)
+			for (unsigned i = 0; i < min_t(unsigned, nr_types, BCH_DATA_NR); i++) {
+				ca->usage_base->d[i].buckets	= le64_to_cpu(u->d[i].buckets);
+				ca->usage_base->d[i].sectors	= le64_to_cpu(u->d[i].sectors);
+				ca->usage_base->d[i].fragmented	= le64_to_cpu(u->d[i].fragmented);
+			}
+		rcu_read_unlock();
 
 		break;
 	}
@@ -597,56 +612,54 @@ int bch2_fs_recovery(struct bch_fs *c)
 	if (c->opts.norecovery)
 		c->opts.recovery_pass_last = BCH_RECOVERY_PASS_journal_replay - 1;
 
-	if (!c->opts.nochanges) {
-		mutex_lock(&c->sb_lock);
-		struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
-		bool write_sb = false;
-
-		if (BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb)) {
-			ext->recovery_passes_required[0] |=
-				cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(BCH_RECOVERY_PASS_check_topology)));
-			write_sb = true;
-		}
+	mutex_lock(&c->sb_lock);
+	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
+	bool write_sb = false;
 
-		u64 sb_passes = bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
-		if (sb_passes) {
-			struct printbuf buf = PRINTBUF;
-			prt_str(&buf, "superblock requires following recovery passes to be run:\n  ");
-			prt_bitflags(&buf, bch2_recovery_passes, sb_passes);
-			bch_info(c, "%s", buf.buf);
-			printbuf_exit(&buf);
-		}
+	if (BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb)) {
+		ext->recovery_passes_required[0] |=
+			cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(BCH_RECOVERY_PASS_check_topology)));
+		write_sb = true;
+	}
 
-		if (bch2_check_version_downgrade(c)) {
-			struct printbuf buf = PRINTBUF;
+	u64 sb_passes = bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
+	if (sb_passes) {
+		struct printbuf buf = PRINTBUF;
+		prt_str(&buf, "superblock requires following recovery passes to be run:\n  ");
+		prt_bitflags(&buf, bch2_recovery_passes, sb_passes);
+		bch_info(c, "%s", buf.buf);
+		printbuf_exit(&buf);
+	}
 
-			prt_str(&buf, "Version downgrade required:");
+	if (bch2_check_version_downgrade(c)) {
+		struct printbuf buf = PRINTBUF;
 
-			__le64 passes = ext->recovery_passes_required[0];
-			bch2_sb_set_downgrade(c,
-					BCH_VERSION_MINOR(bcachefs_metadata_version_current),
-					BCH_VERSION_MINOR(c->sb.version));
-			passes = ext->recovery_passes_required[0] & ~passes;
-			if (passes) {
-				prt_str(&buf, "\n  running recovery passes: ");
-				prt_bitflags(&buf, bch2_recovery_passes,
-					     bch2_recovery_passes_from_stable(le64_to_cpu(passes)));
-			}
+		prt_str(&buf, "Version downgrade required:");
 
-			bch_info(c, "%s", buf.buf);
-			printbuf_exit(&buf);
-			write_sb = true;
+		__le64 passes = ext->recovery_passes_required[0];
+		bch2_sb_set_downgrade(c,
+				      BCH_VERSION_MINOR(bcachefs_metadata_version_current),
+				      BCH_VERSION_MINOR(c->sb.version));
+		passes = ext->recovery_passes_required[0] & ~passes;
+		if (passes) {
+			prt_str(&buf, "\n  running recovery passes: ");
+			prt_bitflags(&buf, bch2_recovery_passes,
+				     bch2_recovery_passes_from_stable(le64_to_cpu(passes)));
 		}
 
-		if (check_version_upgrade(c))
-			write_sb = true;
+		bch_info(c, "%s", buf.buf);
+		printbuf_exit(&buf);
+		write_sb = true;
+	}
 
-		if (write_sb)
-			bch2_write_super(c);
+	if (check_version_upgrade(c))
+		write_sb = true;
 
-		c->recovery_passes_explicit |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
-		mutex_unlock(&c->sb_lock);
-	}
+	if (write_sb)
+		bch2_write_super(c);
+
+	c->recovery_passes_explicit |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
+	mutex_unlock(&c->sb_lock);
 
 	if (c->opts.fsck && IS_ENABLED(CONFIG_BCACHEFS_DEBUG))
 		c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_topology);
@@ -660,7 +673,9 @@ int bch2_fs_recovery(struct bch_fs *c)
 		goto err;
 	}
 
-	if (!c->sb.clean || c->opts.fsck || c->opts.retain_recovery_info) {
+	bch2_journal_pos_from_member_info_resume(c);
+
+	if (!c->sb.clean || c->opts.retain_recovery_info) {
 		struct genradix_iter iter;
 		struct journal_replay **i;
 
@@ -832,8 +847,8 @@ use_clean:
 	}
 
 	mutex_lock(&c->sb_lock);
-	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
-	bool write_sb = false;
+	ext = bch2_sb_field_get(c->disk_sb.sb, ext);
+	write_sb = false;
 
 	if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) != le16_to_cpu(c->disk_sb.sb->version)) {
 		SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, le16_to_cpu(c->disk_sb.sb->version));
@@ -868,6 +883,9 @@ use_clean:
 		write_sb = true;
 	}
 
+	if (bch2_blacklist_entries_gc(c))
+		write_sb = true;
+
 	if (write_sb)
 		bch2_write_super(c);
 	mutex_unlock(&c->sb_lock);
@@ -890,10 +908,6 @@ use_clean:
 		bch_info(c, "scanning for old btree nodes done");
 	}
 
-	if (c->journal_seq_blacklist_table &&
-	    c->journal_seq_blacklist_table->nr > 128)
-		queue_work(system_long_wq, &c->journal_seq_blacklist_gc_work);
-
 	ret = 0;
 out:
 	bch2_flush_fsck_errs(c);
diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index 0cec0f7d970352..4a9eb9582b6e52 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -26,11 +26,6 @@ const char * const bch2_recovery_passes[] = {
 	NULL
 };
 
-static int bch2_check_allocations(struct bch_fs *c)
-{
-	return bch2_gc(c, true, false);
-}
-
 static int bch2_set_may_go_rw(struct bch_fs *c)
 {
 	struct journal_keys *keys = &c->journal_keys;
@@ -227,7 +222,8 @@ int bch2_run_recovery_passes(struct bch_fs *c)
 		if (should_run_recovery_pass(c, c->curr_recovery_pass)) {
 			unsigned pass = c->curr_recovery_pass;
 
-			ret = bch2_run_recovery_pass(c, c->curr_recovery_pass);
+			ret =   bch2_run_recovery_pass(c, c->curr_recovery_pass) ?:
+				bch2_journal_flush(&c->journal);
 			if (bch2_err_matches(ret, BCH_ERR_restart_recovery) ||
 			    (ret && c->curr_recovery_pass < pass))
 				continue;
diff --git a/fs/bcachefs/reflink.c b/fs/bcachefs/reflink.c
index ff7864731a073d..fb31a68148ff98 100644
--- a/fs/bcachefs/reflink.c
+++ b/fs/bcachefs/reflink.c
@@ -74,20 +74,20 @@ bool bch2_reflink_p_merge(struct bch_fs *c, struct bkey_s _l, struct bkey_s_c _r
 }
 
 static int trans_trigger_reflink_p_segment(struct btree_trans *trans,
-			struct bkey_s_c_reflink_p p,
-			u64 *idx, unsigned flags)
+			struct bkey_s_c_reflink_p p, u64 *idx,
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
 	struct bkey_i *k;
 	__le64 *refcount;
-	int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
+	int add = !(flags & BTREE_TRIGGER_overwrite) ? 1 : -1;
 	struct printbuf buf = PRINTBUF;
 	int ret;
 
 	k = bch2_bkey_get_mut_noupdate(trans, &iter,
 			BTREE_ID_reflink, POS(0, *idx),
-			BTREE_ITER_WITH_UPDATES);
+			BTREE_ITER_with_updates);
 	ret = PTR_ERR_OR_ZERO(k);
 	if (ret)
 		goto err;
@@ -102,7 +102,7 @@ static int trans_trigger_reflink_p_segment(struct btree_trans *trans,
 		goto err;
 	}
 
-	if (!*refcount && (flags & BTREE_TRIGGER_OVERWRITE)) {
+	if (!*refcount && (flags & BTREE_TRIGGER_overwrite)) {
 		bch2_bkey_val_to_text(&buf, c, p.s_c);
 		bch2_trans_inconsistent(trans,
 			"indirect extent refcount underflow at %llu while marking\n  %s",
@@ -111,7 +111,7 @@ static int trans_trigger_reflink_p_segment(struct btree_trans *trans,
 		goto err;
 	}
 
-	if (flags & BTREE_TRIGGER_INSERT) {
+	if (flags & BTREE_TRIGGER_insert) {
 		struct bch_reflink_p *v = (struct bch_reflink_p *) p.v;
 		u64 pad;
 
@@ -141,12 +141,13 @@ err:
 }
 
 static s64 gc_trigger_reflink_p_segment(struct btree_trans *trans,
-				struct bkey_s_c_reflink_p p,
-				u64 *idx, unsigned flags, size_t r_idx)
+				struct bkey_s_c_reflink_p p, u64 *idx,
+				enum btree_iter_update_trigger_flags flags,
+				size_t r_idx)
 {
 	struct bch_fs *c = trans->c;
 	struct reflink_gc *r;
-	int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
+	int add = !(flags & BTREE_TRIGGER_overwrite) ? 1 : -1;
 	u64 start = le64_to_cpu(p.v->idx);
 	u64 end = le64_to_cpu(p.v->idx) + p.k->size;
 	u64 next_idx = end + le32_to_cpu(p.v->back_pad);
@@ -163,10 +164,13 @@ static s64 gc_trigger_reflink_p_segment(struct btree_trans *trans,
 
 	BUG_ON((s64) r->refcount + add < 0);
 
-	r->refcount += add;
+	if (flags & BTREE_TRIGGER_gc)
+		r->refcount += add;
 	*idx = r->offset;
 	return 0;
 not_found:
+	BUG_ON(!(flags & BTREE_TRIGGER_check_repair));
+
 	if (fsck_err(c, reflink_p_to_missing_reflink_v,
 		     "pointer to missing indirect extent\n"
 		     "  %s\n"
@@ -189,7 +193,7 @@ not_found:
 			set_bkey_val_u64s(&update->k, 0);
 		}
 
-		ret = bch2_btree_insert_trans(trans, BTREE_ID_extents, update, BTREE_TRIGGER_NORUN);
+		ret = bch2_btree_insert_trans(trans, BTREE_ID_extents, update, BTREE_TRIGGER_norun);
 	}
 
 	*idx = next_idx;
@@ -200,8 +204,8 @@ fsck_err:
 }
 
 static int __trigger_reflink_p(struct btree_trans *trans,
-			    enum btree_id btree_id, unsigned level,
-			    struct bkey_s_c k, unsigned flags)
+		enum btree_id btree_id, unsigned level, struct bkey_s_c k,
+		enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_fs *c = trans->c;
 	struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
@@ -210,12 +214,12 @@ static int __trigger_reflink_p(struct btree_trans *trans,
 	u64 idx = le64_to_cpu(p.v->idx) - le32_to_cpu(p.v->front_pad);
 	u64 end = le64_to_cpu(p.v->idx) + p.k->size + le32_to_cpu(p.v->back_pad);
 
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (flags & BTREE_TRIGGER_transactional) {
 		while (idx < end && !ret)
 			ret = trans_trigger_reflink_p_segment(trans, p, &idx, flags);
 	}
 
-	if (flags & BTREE_TRIGGER_GC) {
+	if (flags & (BTREE_TRIGGER_check_repair|BTREE_TRIGGER_gc)) {
 		size_t l = 0, r = c->reflink_gc_nr;
 
 		while (l < r) {
@@ -238,10 +242,10 @@ int bch2_trigger_reflink_p(struct btree_trans *trans,
 			   enum btree_id btree_id, unsigned level,
 			   struct bkey_s_c old,
 			   struct bkey_s new,
-			   unsigned flags)
+			   enum btree_iter_update_trigger_flags flags)
 {
-	if ((flags & BTREE_TRIGGER_TRANSACTIONAL) &&
-	    (flags & BTREE_TRIGGER_INSERT)) {
+	if ((flags & BTREE_TRIGGER_transactional) &&
+	    (flags & BTREE_TRIGGER_insert)) {
 		struct bch_reflink_p *v = bkey_s_to_reflink_p(new).v;
 
 		v->front_pad = v->back_pad = 0;
@@ -281,23 +285,25 @@ bool bch2_reflink_v_merge(struct bch_fs *c, struct bkey_s _l, struct bkey_s_c _r
 }
 #endif
 
-static inline void check_indirect_extent_deleting(struct bkey_s new, unsigned *flags)
+static inline void
+check_indirect_extent_deleting(struct bkey_s new,
+			       enum btree_iter_update_trigger_flags *flags)
 {
-	if ((*flags & BTREE_TRIGGER_INSERT) && !*bkey_refcount(new)) {
+	if ((*flags & BTREE_TRIGGER_insert) && !*bkey_refcount(new)) {
 		new.k->type = KEY_TYPE_deleted;
 		new.k->size = 0;
 		set_bkey_val_u64s(new.k, 0);
-		*flags &= ~BTREE_TRIGGER_INSERT;
+		*flags &= ~BTREE_TRIGGER_insert;
 	}
 }
 
 int bch2_trigger_reflink_v(struct btree_trans *trans,
 			   enum btree_id btree_id, unsigned level,
 			   struct bkey_s_c old, struct bkey_s new,
-			   unsigned flags)
+			   enum btree_iter_update_trigger_flags flags)
 {
-	if ((flags & BTREE_TRIGGER_TRANSACTIONAL) &&
-	    (flags & BTREE_TRIGGER_INSERT))
+	if ((flags & BTREE_TRIGGER_transactional) &&
+	    (flags & BTREE_TRIGGER_insert))
 		check_indirect_extent_deleting(new, &flags);
 
 	return bch2_trigger_extent(trans, btree_id, level, old, new, flags);
@@ -326,7 +332,7 @@ void bch2_indirect_inline_data_to_text(struct printbuf *out,
 int bch2_trigger_indirect_inline_data(struct btree_trans *trans,
 			      enum btree_id btree_id, unsigned level,
 			      struct bkey_s_c old, struct bkey_s new,
-			      unsigned flags)
+			      enum btree_iter_update_trigger_flags flags)
 {
 	check_indirect_extent_deleting(new, &flags);
 
@@ -349,7 +355,7 @@ static int bch2_make_extent_indirect(struct btree_trans *trans,
 		bch2_check_set_feature(c, BCH_FEATURE_reflink_inline_data);
 
 	bch2_trans_iter_init(trans, &reflink_iter, BTREE_ID_reflink, POS_MAX,
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	k = bch2_btree_iter_peek_prev(&reflink_iter);
 	ret = bkey_err(k);
 	if (ret)
@@ -394,7 +400,7 @@ static int bch2_make_extent_indirect(struct btree_trans *trans,
 	r_p->v.idx = cpu_to_le64(bkey_start_offset(&r_v->k));
 
 	ret = bch2_trans_update(trans, extent_iter, &r_p->k_i,
-				BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+				BTREE_UPDATE_internal_snapshot_node);
 err:
 	bch2_trans_iter_exit(trans, &reflink_iter);
 
@@ -455,9 +461,9 @@ s64 bch2_remap_range(struct bch_fs *c,
 		goto err;
 
 	bch2_trans_iter_init(trans, &src_iter, BTREE_ID_extents, src_start,
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	bch2_trans_iter_init(trans, &dst_iter, BTREE_ID_extents, dst_start,
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	while ((ret == 0 ||
 		bch2_err_matches(ret, BCH_ERR_transaction_restart)) &&
@@ -567,7 +573,7 @@ s64 bch2_remap_range(struct bch_fs *c,
 		bch2_trans_begin(trans);
 
 		ret2 = bch2_inode_peek(trans, &inode_iter, &inode_u,
-				       dst_inum, BTREE_ITER_INTENT);
+				       dst_inum, BTREE_ITER_intent);
 
 		if (!ret2 &&
 		    inode_u.bi_size < new_i_size) {
diff --git a/fs/bcachefs/reflink.h b/fs/bcachefs/reflink.h
index 4d8867289717bf..59d4dbae3fb1f4 100644
--- a/fs/bcachefs/reflink.h
+++ b/fs/bcachefs/reflink.h
@@ -10,7 +10,8 @@ void bch2_reflink_p_to_text(struct printbuf *, struct bch_fs *,
 			    struct bkey_s_c);
 bool bch2_reflink_p_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
 int bch2_trigger_reflink_p(struct btree_trans *, enum btree_id, unsigned,
-			   struct bkey_s_c, struct bkey_s, unsigned);
+			   struct bkey_s_c, struct bkey_s,
+			   enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_reflink_p ((struct bkey_ops) {		\
 	.key_invalid	= bch2_reflink_p_invalid,		\
@@ -25,7 +26,8 @@ int bch2_reflink_v_invalid(struct bch_fs *, struct bkey_s_c,
 void bch2_reflink_v_to_text(struct printbuf *, struct bch_fs *,
 			    struct bkey_s_c);
 int bch2_trigger_reflink_v(struct btree_trans *, enum btree_id, unsigned,
-			      struct bkey_s_c, struct bkey_s, unsigned);
+			   struct bkey_s_c, struct bkey_s,
+			   enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_reflink_v ((struct bkey_ops) {		\
 	.key_invalid	= bch2_reflink_v_invalid,		\
@@ -42,7 +44,7 @@ void bch2_indirect_inline_data_to_text(struct printbuf *,
 int bch2_trigger_indirect_inline_data(struct btree_trans *,
 					 enum btree_id, unsigned,
 			      struct bkey_s_c, struct bkey_s,
-			      unsigned);
+			      enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_indirect_inline_data ((struct bkey_ops) {	\
 	.key_invalid	= bch2_indirect_inline_data_invalid,	\
diff --git a/fs/bcachefs/replicas.c b/fs/bcachefs/replicas.c
index 678b9c20e2514b..f8ff7c8bb05ee5 100644
--- a/fs/bcachefs/replicas.c
+++ b/fs/bcachefs/replicas.c
@@ -84,7 +84,7 @@ int bch2_replicas_entry_validate(struct bch_replicas_entry_v1 *r,
 	}
 
 	for (unsigned i = 0; i < r->nr_devs; i++)
-		if (!bch2_dev_exists(sb, r->devs[i])) {
+		if (!bch2_member_exists(sb, r->devs[i])) {
 			prt_printf(err, "invalid device %u in entry ", r->devs[i]);
 			goto bad;
 		}
@@ -200,7 +200,7 @@ cpu_replicas_add_entry(struct bch_fs *c,
 	};
 
 	for (i = 0; i < new_entry->nr_devs; i++)
-		BUG_ON(!bch2_dev_exists2(c, new_entry->devs[i]));
+		BUG_ON(!bch2_dev_exists(c, new_entry->devs[i]));
 
 	BUG_ON(!new_entry->data_type);
 	verify_replicas_entry(new_entry);
@@ -947,18 +947,20 @@ bool bch2_have_enough_devs(struct bch_fs *c, struct bch_devs_mask devs,
 
 	percpu_down_read(&c->mark_lock);
 	for_each_cpu_replicas_entry(&c->replicas, e) {
-		unsigned i, nr_online = 0, nr_failed = 0, dflags = 0;
+		unsigned nr_online = 0, nr_failed = 0, dflags = 0;
 		bool metadata = e->data_type < BCH_DATA_user;
 
 		if (e->data_type == BCH_DATA_cached)
 			continue;
 
-		for (i = 0; i < e->nr_devs; i++) {
-			struct bch_dev *ca = bch_dev_bkey_exists(c, e->devs[i]);
-
+		rcu_read_lock();
+		for (unsigned i = 0; i < e->nr_devs; i++) {
 			nr_online += test_bit(e->devs[i], devs.d);
-			nr_failed += ca->mi.state == BCH_MEMBER_STATE_failed;
+
+			struct bch_dev *ca = bch2_dev_rcu(c, e->devs[i]);
+			nr_failed += ca && ca->mi.state == BCH_MEMBER_STATE_failed;
 		}
+		rcu_read_unlock();
 
 		if (nr_failed == e->nr_devs)
 			continue;
diff --git a/fs/bcachefs/sb-clean.c b/fs/bcachefs/sb-clean.c
index 35ca3f138de6fa..0d531603784110 100644
--- a/fs/bcachefs/sb-clean.c
+++ b/fs/bcachefs/sb-clean.c
@@ -287,10 +287,8 @@ static void bch2_sb_clean_to_text(struct printbuf *out, struct bch_sb *sb,
 	struct bch_sb_field_clean *clean = field_to_type(f, clean);
 	struct jset_entry *entry;
 
-	prt_printf(out, "flags:          %x",	le32_to_cpu(clean->flags));
-	prt_newline(out);
-	prt_printf(out, "journal_seq:    %llu",	le64_to_cpu(clean->journal_seq));
-	prt_newline(out);
+	prt_printf(out, "flags:          %x\n",		le32_to_cpu(clean->flags));
+	prt_printf(out, "journal_seq:    %llu\n",	le64_to_cpu(clean->journal_seq));
 
 	for (entry = clean->start;
 	     entry != vstruct_end(&clean->field);
@@ -378,6 +376,8 @@ void bch2_fs_mark_clean(struct bch_fs *c)
 		goto out;
 	}
 
+	bch2_journal_pos_from_member_info_set(c);
+
 	bch2_write_super(c);
 out:
 	mutex_unlock(&c->sb_lock);
diff --git a/fs/bcachefs/sb-counters.c b/fs/bcachefs/sb-counters.c
index 7dc898761bb312..7282f78861926e 100644
--- a/fs/bcachefs/sb-counters.c
+++ b/fs/bcachefs/sb-counters.c
@@ -31,19 +31,12 @@ static void bch2_sb_counters_to_text(struct printbuf *out, struct bch_sb *sb,
 			      struct bch_sb_field *f)
 {
 	struct bch_sb_field_counters *ctrs = field_to_type(f, counters);
-	unsigned int i;
 	unsigned int nr = bch2_sb_counter_nr_entries(ctrs);
 
-	for (i = 0; i < nr; i++) {
-		if (i < BCH_COUNTER_NR)
-			prt_printf(out, "%s ", bch2_counter_names[i]);
-		else
-			prt_printf(out, "(unknown)");
-
-		prt_tab(out);
-		prt_printf(out, "%llu", le64_to_cpu(ctrs->d[i]));
-		prt_newline(out);
-	}
+	for (unsigned i = 0; i < nr; i++)
+		prt_printf(out, "%s \t%llu\n",
+			   i < BCH_COUNTER_NR ? bch2_counter_names[i] : "(unknown)",
+			   le64_to_cpu(ctrs->d[i]));
 };
 
 int bch2_sb_counters_to_cpu(struct bch_fs *c)
diff --git a/fs/bcachefs/sb-downgrade.c b/fs/bcachefs/sb-downgrade.c
index a98ef940b7a328..90b06f8aa1df08 100644
--- a/fs/bcachefs/sb-downgrade.c
+++ b/fs/bcachefs/sb-downgrade.c
@@ -164,19 +164,16 @@ static void bch2_sb_downgrade_to_text(struct printbuf *out, struct bch_sb *sb,
 		printbuf_tabstop_push(out, 16);
 
 	for_each_downgrade_entry(e, i) {
-		prt_str(out, "version:");
-		prt_tab(out);
+		prt_str(out, "version:\t");
 		bch2_version_to_text(out, le16_to_cpu(i->version));
 		prt_newline(out);
 
-		prt_str(out, "recovery passes:");
-		prt_tab(out);
+		prt_str(out, "recovery passes:\t");
 		prt_bitflags(out, bch2_recovery_passes,
 			     bch2_recovery_passes_from_stable(le64_to_cpu(i->recovery_passes[0])));
 		prt_newline(out);
 
-		prt_str(out, "errors:");
-		prt_tab(out);
+		prt_str(out, "errors:\t");
 		bool first = true;
 		for (unsigned j = 0; j < le16_to_cpu(i->nr_errors); j++) {
 			if (!first)
diff --git a/fs/bcachefs/sb-errors_types.h b/fs/bcachefs/sb-errors_types.h
index 06c7a644f4a442..87324747351aae 100644
--- a/fs/bcachefs/sb-errors_types.h
+++ b/fs/bcachefs/sb-errors_types.h
@@ -272,7 +272,8 @@
 	x(snapshot_node_missing,				264)	\
 	x(dup_backpointer_to_bad_csum_extent,			265)	\
 	x(btree_bitmap_not_marked,				266)	\
-	x(sb_clean_entry_overrun,				267)
+	x(sb_clean_entry_overrun,				267)	\
+	x(btree_ptr_v2_written_0,				268)
 
 enum bch_sb_error_id {
 #define x(t, n) BCH_FSCK_ERR_##t = n,
diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c
index 5b8e621ac5eb57..9def2abba6db5b 100644
--- a/fs/bcachefs/sb-members.c
+++ b/fs/bcachefs/sb-members.c
@@ -3,11 +3,22 @@
 #include "bcachefs.h"
 #include "btree_cache.h"
 #include "disk_groups.h"
+#include "error.h"
 #include "opts.h"
 #include "replicas.h"
 #include "sb-members.h"
 #include "super-io.h"
 
+void bch2_dev_missing(struct bch_fs *c, unsigned dev)
+{
+	bch2_fs_inconsistent(c, "pointer to nonexistent device %u", dev);
+}
+
+void bch2_dev_bucket_missing(struct bch_fs *c, struct bpos bucket)
+{
+	bch2_fs_inconsistent(c, "pointer to nonexistent bucket %llu:%llu", bucket.inode, bucket.offset);
+}
+
 #define x(t, n, ...) [n] = #t,
 static const char * const bch2_iops_measurements[] = {
 	BCH_IOPS_MEASUREMENTS()
@@ -164,18 +175,14 @@ static void member_to_text(struct printbuf *out,
 	u64 bucket_size = le16_to_cpu(m.bucket_size);
 	u64 device_size = le64_to_cpu(m.nbuckets) * bucket_size;
 
-	if (!bch2_member_exists(&m))
+	if (!bch2_member_alive(&m))
 		return;
 
-	prt_printf(out, "Device:");
-	prt_tab(out);
-	prt_printf(out, "%u", i);
-	prt_newline(out);
+	prt_printf(out, "Device:\t%u\n", i);
 
 	printbuf_indent_add(out, 2);
 
-	prt_printf(out, "Label:");
-	prt_tab(out);
+	prt_printf(out, "Label:\t");
 	if (BCH_MEMBER_GROUP(&m)) {
 		unsigned idx = BCH_MEMBER_GROUP(&m) - 1;
 
@@ -189,96 +196,67 @@ static void member_to_text(struct printbuf *out,
 	}
 	prt_newline(out);
 
-	prt_printf(out, "UUID:");
-	prt_tab(out);
+	prt_printf(out, "UUID:\t");
 	pr_uuid(out, m.uuid.b);
 	prt_newline(out);
 
-	prt_printf(out, "Size:");
-	prt_tab(out);
+	prt_printf(out, "Size:\t");
 	prt_units_u64(out, device_size << 9);
 	prt_newline(out);
 
-	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
-		prt_printf(out, "%s errors:", bch2_member_error_strs[i]);
-		prt_tab(out);
-		prt_u64(out, le64_to_cpu(m.errors[i]));
-		prt_newline(out);
-	}
+	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
+		prt_printf(out, "%s errors:\t%llu\n", bch2_member_error_strs[i], le64_to_cpu(m.errors[i]));
 
-	for (unsigned i = 0; i < BCH_IOPS_NR; i++) {
-		prt_printf(out, "%s iops:", bch2_iops_measurements[i]);
-		prt_tab(out);
-		prt_printf(out, "%u", le32_to_cpu(m.iops[i]));
-		prt_newline(out);
-	}
+	for (unsigned i = 0; i < BCH_IOPS_NR; i++)
+		prt_printf(out, "%s iops:\t%u\n", bch2_iops_measurements[i], le32_to_cpu(m.iops[i]));
 
-	prt_printf(out, "Bucket size:");
-	prt_tab(out);
+	prt_printf(out, "Bucket size:\t");
 	prt_units_u64(out, bucket_size << 9);
 	prt_newline(out);
 
-	prt_printf(out, "First bucket:");
-	prt_tab(out);
-	prt_printf(out, "%u", le16_to_cpu(m.first_bucket));
-	prt_newline(out);
-
-	prt_printf(out, "Buckets:");
-	prt_tab(out);
-	prt_printf(out, "%llu", le64_to_cpu(m.nbuckets));
-	prt_newline(out);
+	prt_printf(out, "First bucket:\t%u\n", le16_to_cpu(m.first_bucket));
+	prt_printf(out, "Buckets:\t%llu\n", le64_to_cpu(m.nbuckets));
 
-	prt_printf(out, "Last mount:");
-	prt_tab(out);
+	prt_printf(out, "Last mount:\t");
 	if (m.last_mount)
 		bch2_prt_datetime(out, le64_to_cpu(m.last_mount));
 	else
 		prt_printf(out, "(never)");
 	prt_newline(out);
 
-	prt_printf(out, "Last superblock write:");
-	prt_tab(out);
-	prt_u64(out, le64_to_cpu(m.seq));
-	prt_newline(out);
+	prt_printf(out, "Last superblock write:\t%llu\n", le64_to_cpu(m.seq));
 
-	prt_printf(out, "State:");
-	prt_tab(out);
-	prt_printf(out, "%s",
+	prt_printf(out, "State:\t%s\n",
 		   BCH_MEMBER_STATE(&m) < BCH_MEMBER_STATE_NR
 		   ? bch2_member_states[BCH_MEMBER_STATE(&m)]
 		   : "unknown");
-	prt_newline(out);
 
-	prt_printf(out, "Data allowed:");
-	prt_tab(out);
+	prt_printf(out, "Data allowed:\t");
 	if (BCH_MEMBER_DATA_ALLOWED(&m))
 		prt_bitflags(out, __bch2_data_types, BCH_MEMBER_DATA_ALLOWED(&m));
 	else
 		prt_printf(out, "(none)");
 	prt_newline(out);
 
-	prt_printf(out, "Has data:");
-	prt_tab(out);
+	prt_printf(out, "Has data:\t");
 	if (data_have)
 		prt_bitflags(out, __bch2_data_types, data_have);
 	else
 		prt_printf(out, "(none)");
 	prt_newline(out);
 
-	prt_str(out, "Durability:");
-	prt_tab(out);
-	prt_printf(out, "%llu", BCH_MEMBER_DURABILITY(&m) ? BCH_MEMBER_DURABILITY(&m) - 1 : 1);
+	prt_printf(out, "Btree allocated bitmap blocksize:\t");
+	prt_units_u64(out, 1ULL << m.btree_bitmap_shift);
 	prt_newline(out);
 
-	prt_printf(out, "Discard:");
-	prt_tab(out);
-	prt_printf(out, "%llu", BCH_MEMBER_DISCARD(&m));
+	prt_printf(out, "Btree allocated bitmap:\t");
+	bch2_prt_u64_base2_nbits(out, le64_to_cpu(m.btree_allocated_bitmap), 64);
 	prt_newline(out);
 
-	prt_printf(out, "Freespace initialized:");
-	prt_tab(out);
-	prt_printf(out, "%llu", BCH_MEMBER_FREESPACE_INITIALIZED(&m));
-	prt_newline(out);
+	prt_printf(out, "Durability:\t%llu\n", BCH_MEMBER_DURABILITY(&m) ? BCH_MEMBER_DURABILITY(&m) - 1 : 1);
+
+	prt_printf(out, "Discard:\t%llu\n", BCH_MEMBER_DISCARD(&m));
+	prt_printf(out, "Freespace initialized:\t%llu\n", BCH_MEMBER_FREESPACE_INITIALIZED(&m));
 
 	printbuf_indent_sub(out, 2);
 }
@@ -390,12 +368,8 @@ void bch2_dev_io_errors_to_text(struct printbuf *out, struct bch_dev *ca)
 	prt_newline(out);
 
 	printbuf_indent_add(out, 2);
-	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
-		prt_printf(out, "%s:", bch2_member_error_strs[i]);
-		prt_tab(out);
-		prt_u64(out, atomic64_read(&ca->errors[i]));
-		prt_newline(out);
-	}
+	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
+		prt_printf(out, "%s:\t%llu\n", bch2_member_error_strs[i], atomic64_read(&ca->errors[i]));
 	printbuf_indent_sub(out, 2);
 
 	prt_str(out, "IO errors since ");
@@ -404,12 +378,9 @@ void bch2_dev_io_errors_to_text(struct printbuf *out, struct bch_dev *ca)
 	prt_newline(out);
 
 	printbuf_indent_add(out, 2);
-	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
-		prt_printf(out, "%s:", bch2_member_error_strs[i]);
-		prt_tab(out);
-		prt_u64(out, atomic64_read(&ca->errors[i]) - le64_to_cpu(m.errors_at_reset[i]));
-		prt_newline(out);
-	}
+	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
+		prt_printf(out, "%s:\t%llu\n", bch2_member_error_strs[i],
+			   atomic64_read(&ca->errors[i]) - le64_to_cpu(m.errors_at_reset[i]));
 	printbuf_indent_sub(out, 2);
 }
 
@@ -437,11 +408,20 @@ void bch2_dev_errors_reset(struct bch_dev *ca)
 
 bool bch2_dev_btree_bitmap_marked(struct bch_fs *c, struct bkey_s_c k)
 {
-	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
-		if (!bch2_dev_btree_bitmap_marked_sectors(bch_dev_bkey_exists(c, ptr->dev),
-							  ptr->offset, btree_sectors(c)))
-			return false;
-	return true;
+	bool ret = true;
+	rcu_read_lock();
+	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr) {
+		struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
+		if (!ca)
+			continue;
+
+		if (!bch2_dev_btree_bitmap_marked_sectors(ca, ptr->offset, btree_sectors(c))) {
+			ret = false;
+			break;
+		}
+	}
+	rcu_read_unlock();
+	return ret;
 }
 
 static void __bch2_dev_btree_bitmap_mark(struct bch_sb_field_members_v2 *mi, unsigned dev,
@@ -463,6 +443,9 @@ static void __bch2_dev_btree_bitmap_mark(struct bch_sb_field_members_v2 *mi, uns
 		m->btree_bitmap_shift += resize;
 	}
 
+	BUG_ON(m->btree_bitmap_shift > 57);
+	BUG_ON(end > 64ULL << m->btree_bitmap_shift);
+
 	for (unsigned bit = start >> m->btree_bitmap_shift;
 	     (u64) bit << m->btree_bitmap_shift < end;
 	     bit++)
@@ -476,6 +459,10 @@ void bch2_dev_btree_bitmap_mark(struct bch_fs *c, struct bkey_s_c k)
 	lockdep_assert_held(&c->sb_lock);
 
 	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
-	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
+	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr) {
+		if (!bch2_member_exists(c->disk_sb.sb, ptr->dev))
+			continue;
+
 		__bch2_dev_btree_bitmap_mark(mi, ptr->dev, ptr->offset, btree_sectors(c));
+	}
 }
diff --git a/fs/bcachefs/sb-members.h b/fs/bcachefs/sb-members.h
index 5efa64eca5f85a..f221c38970a60a 100644
--- a/fs/bcachefs/sb-members.h
+++ b/fs/bcachefs/sb-members.h
@@ -158,26 +158,105 @@ static inline struct bch_dev *bch2_get_next_online_dev(struct bch_fs *c,
 #define for_each_readable_member(c, ca)				\
 	__for_each_online_member(c, ca,	BIT( BCH_MEMBER_STATE_rw)|BIT(BCH_MEMBER_STATE_ro))
 
+static inline bool bch2_dev_exists(const struct bch_fs *c, unsigned dev)
+{
+	return dev < c->sb.nr_devices && c->devs[dev];
+}
+
+static inline bool bucket_valid(const struct bch_dev *ca, u64 b)
+{
+	return b - ca->mi.first_bucket < ca->mi.nbuckets_minus_first;
+}
+
+static inline struct bch_dev *bch2_dev_have_ref(const struct bch_fs *c, unsigned dev)
+{
+	EBUG_ON(!bch2_dev_exists(c, dev));
+
+	return rcu_dereference_check(c->devs[dev], 1);
+}
+
 /*
  * If a key exists that references a device, the device won't be going away and
  * we can omit rcu_read_lock():
  */
-static inline struct bch_dev *bch_dev_bkey_exists(const struct bch_fs *c, unsigned idx)
+static inline struct bch_dev *bch2_dev_bkey_exists(const struct bch_fs *c, unsigned dev)
 {
-	EBUG_ON(idx >= c->sb.nr_devices || !c->devs[idx]);
+	EBUG_ON(!bch2_dev_exists(c, dev));
 
-	return rcu_dereference_check(c->devs[idx], 1);
+	return rcu_dereference_check(c->devs[dev], 1);
 }
 
-static inline struct bch_dev *bch_dev_locked(struct bch_fs *c, unsigned idx)
+static inline struct bch_dev *bch2_dev_locked(struct bch_fs *c, unsigned dev)
 {
-	EBUG_ON(idx >= c->sb.nr_devices || !c->devs[idx]);
+	EBUG_ON(!bch2_dev_exists(c, dev));
 
-	return rcu_dereference_protected(c->devs[idx],
+	return rcu_dereference_protected(c->devs[dev],
 					 lockdep_is_held(&c->sb_lock) ||
 					 lockdep_is_held(&c->state_lock));
 }
 
+static inline struct bch_dev *bch2_dev_rcu(struct bch_fs *c, unsigned dev)
+{
+	return c && dev < c->sb.nr_devices
+		? rcu_dereference(c->devs[dev])
+		: NULL;
+}
+
+static inline void bch2_dev_put(struct bch_dev *ca)
+{
+	if (ca)
+		percpu_ref_put(&ca->ref);
+}
+
+static inline struct bch_dev *bch2_dev_tryget_noerror(struct bch_fs *c, unsigned dev)
+{
+	rcu_read_lock();
+	struct bch_dev *ca = bch2_dev_rcu(c, dev);
+	if (ca && !percpu_ref_tryget(&ca->ref))
+		ca = NULL;
+	rcu_read_unlock();
+	return ca;
+}
+
+void bch2_dev_missing(struct bch_fs *, unsigned);
+
+static inline struct bch_dev *bch2_dev_tryget(struct bch_fs *c, unsigned dev)
+{
+	struct bch_dev *ca = bch2_dev_tryget_noerror(c, dev);
+	if (!ca)
+		bch2_dev_missing(c, dev);
+	return ca;
+}
+
+static inline struct bch_dev *bch2_dev_bucket_tryget_noerror(struct bch_fs *c, struct bpos bucket)
+{
+	struct bch_dev *ca = bch2_dev_tryget_noerror(c, bucket.inode);
+	if (ca && !bucket_valid(ca, bucket.offset)) {
+		percpu_ref_put(&ca->ref);
+		ca = NULL;
+	}
+	return ca;
+}
+
+void bch2_dev_bucket_missing(struct bch_fs *, struct bpos);
+
+static inline struct bch_dev *bch2_dev_bucket_tryget(struct bch_fs *c, struct bpos bucket)
+{
+	struct bch_dev *ca = bch2_dev_bucket_tryget_noerror(c, bucket);
+	if (!ca)
+		bch2_dev_bucket_missing(c, bucket);
+	return ca;
+}
+
+static inline struct bch_dev *bch2_dev_iterate(struct bch_fs *c, struct bch_dev *ca, unsigned dev_idx)
+{
+	if (ca && ca->dev_idx == dev_idx)
+		return ca;
+	if (ca)
+		percpu_ref_put(&ca->ref);
+	return bch2_dev_tryget(c, dev_idx);
+}
+
 /* XXX kill, move to struct bch_fs */
 static inline struct bch_devs_mask bch2_online_devs(struct bch_fs *c)
 {
@@ -192,16 +271,16 @@ static inline struct bch_devs_mask bch2_online_devs(struct bch_fs *c)
 extern const struct bch_sb_field_ops bch_sb_field_ops_members_v1;
 extern const struct bch_sb_field_ops bch_sb_field_ops_members_v2;
 
-static inline bool bch2_member_exists(struct bch_member *m)
+static inline bool bch2_member_alive(struct bch_member *m)
 {
 	return !bch2_is_zero(&m->uuid, sizeof(m->uuid));
 }
 
-static inline bool bch2_dev_exists(struct bch_sb *sb, unsigned dev)
+static inline bool bch2_member_exists(struct bch_sb *sb, unsigned dev)
 {
 	if (dev < sb->nr_devices) {
 		struct bch_member m = bch2_sb_member_get(sb, dev);
-		return bch2_member_exists(&m);
+		return bch2_member_alive(&m);
 	}
 	return false;
 }
@@ -210,6 +289,8 @@ static inline struct bch_member_cpu bch2_mi_to_cpu(struct bch_member *mi)
 {
 	return (struct bch_member_cpu) {
 		.nbuckets	= le64_to_cpu(mi->nbuckets),
+		.nbuckets_minus_first = le64_to_cpu(mi->nbuckets) -
+			le16_to_cpu(mi->first_bucket),
 		.first_bucket	= le16_to_cpu(mi->first_bucket),
 		.bucket_size	= le16_to_cpu(mi->bucket_size),
 		.group		= BCH_MEMBER_GROUP(mi),
@@ -220,7 +301,7 @@ static inline struct bch_member_cpu bch2_mi_to_cpu(struct bch_member *mi)
 			? BCH_MEMBER_DURABILITY(mi) - 1
 			: 1,
 		.freespace_initialized = BCH_MEMBER_FREESPACE_INITIALIZED(mi),
-		.valid		= bch2_member_exists(mi),
+		.valid		= bch2_member_alive(mi),
 		.btree_bitmap_shift	= mi->btree_bitmap_shift,
 		.btree_allocated_bitmap = le64_to_cpu(mi->btree_allocated_bitmap),
 	};
diff --git a/fs/bcachefs/sb-members_types.h b/fs/bcachefs/sb-members_types.h
new file mode 100644
index 00000000000000..c0eda888fe39a0
--- /dev/null
+++ b/fs/bcachefs/sb-members_types.h
@@ -0,0 +1,21 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _BCACHEFS_SB_MEMBERS_TYPES_H
+#define _BCACHEFS_SB_MEMBERS_TYPES_H
+
+struct bch_member_cpu {
+	u64			nbuckets;	/* device size */
+	u64			nbuckets_minus_first;
+	u16			first_bucket;   /* index of first bucket used */
+	u16			bucket_size;	/* sectors */
+	u16			group;
+	u8			state;
+	u8			discard;
+	u8			data_allowed;
+	u8			durability;
+	u8			freespace_initialized;
+	u8			valid;
+	u8			btree_bitmap_shift;
+	u64			btree_allocated_bitmap;
+};
+
+#endif /* _BCACHEFS_SB_MEMBERS_H */
diff --git a/fs/bcachefs/snapshot.c b/fs/bcachefs/snapshot.c
index 544322d5c25170..3aa4686cc71fcc 100644
--- a/fs/bcachefs/snapshot.c
+++ b/fs/bcachefs/snapshot.c
@@ -49,7 +49,7 @@ int bch2_snapshot_tree_lookup(struct btree_trans *trans, u32 id,
 			      struct bch_snapshot_tree *s)
 {
 	int ret = bch2_bkey_get_val_typed(trans, BTREE_ID_snapshot_trees, POS(0, id),
-					  BTREE_ITER_WITH_UPDATES, snapshot_tree, s);
+					  BTREE_ITER_with_updates, snapshot_tree, s);
 
 	if (bch2_err_matches(ret, ENOENT))
 		ret = -BCH_ERR_ENOENT_snapshot_tree;
@@ -298,7 +298,7 @@ static void set_is_ancestor_bitmap(struct bch_fs *c, u32 id)
 static int __bch2_mark_snapshot(struct btree_trans *trans,
 		       enum btree_id btree, unsigned level,
 		       struct bkey_s_c old, struct bkey_s_c new,
-		       unsigned flags)
+		       enum btree_iter_update_trigger_flags flags)
 {
 	struct bch_fs *c = trans->c;
 	struct snapshot_t *t;
@@ -352,7 +352,7 @@ err:
 int bch2_mark_snapshot(struct btree_trans *trans,
 		       enum btree_id btree, unsigned level,
 		       struct bkey_s_c old, struct bkey_s new,
-		       unsigned flags)
+		       enum btree_iter_update_trigger_flags flags)
 {
 	return __bch2_mark_snapshot(trans, btree, level, old, new.s_c, flags);
 }
@@ -361,7 +361,7 @@ int bch2_snapshot_lookup(struct btree_trans *trans, u32 id,
 			 struct bch_snapshot *s)
 {
 	return bch2_bkey_get_val_typed(trans, BTREE_ID_snapshots, POS(0, id),
-				       BTREE_ITER_WITH_UPDATES, snapshot, s);
+				       BTREE_ITER_with_updates, snapshot, s);
 }
 
 static int bch2_snapshot_live(struct btree_trans *trans, u32 id)
@@ -618,7 +618,7 @@ int bch2_check_snapshot_trees(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
 			BTREE_ID_snapshot_trees, POS_MIN,
-			BTREE_ITER_PREFETCH, k,
+			BTREE_ITER_prefetch, k,
 			NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 		check_snapshot_tree(trans, &iter, k)));
 	bch_err_fn(c, ret);
@@ -695,7 +695,7 @@ static int snapshot_tree_ptr_repair(struct btree_trans *trans,
 
 	root = bch2_bkey_get_iter_typed(trans, &root_iter,
 			       BTREE_ID_snapshots, POS(0, root_id),
-			       BTREE_ITER_WITH_UPDATES, snapshot);
+			       BTREE_ITER_with_updates, snapshot);
 	ret = bkey_err(root);
 	if (ret)
 		goto err;
@@ -886,7 +886,7 @@ int bch2_check_snapshots(struct bch_fs *c)
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_reverse_commit(trans, iter,
 				BTREE_ID_snapshots, POS_MAX,
-				BTREE_ITER_PREFETCH, k,
+				BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			check_snapshot(trans, &iter, k)));
 	bch_err_fn(c, ret);
@@ -900,7 +900,8 @@ static int check_snapshot_exists(struct btree_trans *trans, u32 id)
 	if (bch2_snapshot_equiv(c, id))
 		return 0;
 
-	u32 tree_id;
+	/* 0 is an invalid tree ID */
+	u32 tree_id = 0;
 	int ret = bch2_snapshot_tree_create(trans, id, 0, &tree_id);
 	if (ret)
 		return ret;
@@ -1001,7 +1002,7 @@ int bch2_reconstruct_snapshots(struct bch_fs *c)
 			r.btree = btree;
 
 			ret = for_each_btree_key(trans, iter, btree, POS_MIN,
-					BTREE_ITER_ALL_SNAPSHOTS|BTREE_ITER_PREFETCH, k, ({
+					BTREE_ITER_all_snapshots|BTREE_ITER_prefetch, k, ({
 				get_snapshot_trees(c, &r, k.k->p);
 			}));
 			if (ret)
@@ -1018,7 +1019,7 @@ int bch2_reconstruct_snapshots(struct bch_fs *c)
 		darray_for_each(*t, id) {
 			if (fsck_err_on(!bch2_snapshot_equiv(c, *id),
 					c, snapshot_node_missing,
-					"snapshot node %u from tree %s missing", *id, buf.buf)) {
+					"snapshot node %u from tree %s missing, recreate?", *id, buf.buf)) {
 				if (t->nr > 1) {
 					bch_err(c, "cannot reconstruct snapshot trees with multiple nodes");
 					ret = -BCH_ERR_fsck_repair_unimplemented;
@@ -1090,7 +1091,7 @@ static int bch2_snapshot_node_delete(struct btree_trans *trans, u32 id)
 	int ret = 0;
 
 	s = bch2_bkey_get_iter_typed(trans, &iter, BTREE_ID_snapshots, POS(0, id),
-				     BTREE_ITER_INTENT, snapshot);
+				     BTREE_ITER_intent, snapshot);
 	ret = bkey_err(s);
 	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), c,
 				"missing snapshot %u", id);
@@ -1199,7 +1200,7 @@ static int create_snapids(struct btree_trans *trans, u32 parent, u32 tree,
 	int ret;
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_snapshots,
-			     POS_MIN, BTREE_ITER_INTENT);
+			     POS_MIN, BTREE_ITER_intent);
 	k = bch2_btree_iter_peek(&iter);
 	ret = bkey_err(k);
 	if (ret)
@@ -1367,7 +1368,7 @@ static int snapshot_delete_key(struct btree_trans *trans,
 	if (snapshot_list_has_id(deleted, k.k->p.snapshot) ||
 	    snapshot_list_has_id(equiv_seen, equiv)) {
 		return bch2_btree_delete_at(trans, iter,
-					    BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+					    BTREE_UPDATE_internal_snapshot_node);
 	} else {
 		return snapshot_list_add(c, equiv_seen, equiv);
 	}
@@ -1404,15 +1405,15 @@ static int move_key_to_correct_snapshot(struct btree_trans *trans,
 		new->k.p.snapshot = equiv;
 
 		bch2_trans_iter_init(trans, &new_iter, iter->btree_id, new->k.p,
-				     BTREE_ITER_ALL_SNAPSHOTS|
-				     BTREE_ITER_CACHED|
-				     BTREE_ITER_INTENT);
+				     BTREE_ITER_all_snapshots|
+				     BTREE_ITER_cached|
+				     BTREE_ITER_intent);
 
 		ret =   bch2_btree_iter_traverse(&new_iter) ?:
 			bch2_trans_update(trans, &new_iter, new,
-					BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+					BTREE_UPDATE_internal_snapshot_node) ?:
 			bch2_btree_delete_at(trans, iter,
-					BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+					BTREE_UPDATE_internal_snapshot_node);
 		bch2_trans_iter_exit(trans, &new_iter);
 		if (ret)
 			return ret;
@@ -1603,12 +1604,12 @@ int bch2_delete_dead_snapshots(struct bch_fs *c)
 
 		ret = for_each_btree_key_commit(trans, iter,
 				id, POS_MIN,
-				BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+				BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 				&res, NULL, BCH_TRANS_COMMIT_no_enospc,
 			snapshot_delete_key(trans, &iter, k, &deleted, &equiv_seen, &last_pos)) ?:
 		      for_each_btree_key_commit(trans, iter,
 				id, POS_MIN,
-				BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+				BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
 				&res, NULL, BCH_TRANS_COMMIT_no_enospc,
 			move_key_to_correct_snapshot(trans, &iter, k));
 
@@ -1643,7 +1644,7 @@ int bch2_delete_dead_snapshots(struct bch_fs *c)
 	 * nodes some depth fields will be off:
 	 */
 	ret = for_each_btree_key_commit(trans, iter, BTREE_ID_snapshots, POS_MIN,
-				  BTREE_ITER_INTENT, k,
+				  BTREE_ITER_intent, k,
 				  NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 		bch2_fix_child_of_deleted_snapshot(trans, &iter, k, &deleted_interior));
 	if (ret)
@@ -1699,8 +1700,8 @@ int __bch2_key_has_snapshot_overwrites(struct btree_trans *trans,
 	int ret;
 
 	bch2_trans_iter_init(trans, &iter, id, pos,
-			     BTREE_ITER_NOT_EXTENTS|
-			     BTREE_ITER_ALL_SNAPSHOTS);
+			     BTREE_ITER_not_extents|
+			     BTREE_ITER_all_snapshots);
 	while (1) {
 		k = bch2_btree_iter_prev(&iter);
 		ret = bkey_err(k);
@@ -1752,7 +1753,7 @@ static int bch2_propagate_key_to_snapshot_leaf(struct btree_trans *trans,
 
 	pos.snapshot = leaf_id;
 
-	bch2_trans_iter_init(trans, &iter, btree, pos, BTREE_ITER_INTENT);
+	bch2_trans_iter_init(trans, &iter, btree, pos, BTREE_ITER_intent);
 	k = bch2_btree_iter_peek_slot(&iter);
 	ret = bkey_err(k);
 	if (ret)
diff --git a/fs/bcachefs/snapshot.h b/fs/bcachefs/snapshot.h
index b7d2fed37c4f31..b3c43d57954028 100644
--- a/fs/bcachefs/snapshot.h
+++ b/fs/bcachefs/snapshot.h
@@ -22,7 +22,8 @@ void bch2_snapshot_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 int bch2_snapshot_invalid(struct bch_fs *, struct bkey_s_c,
 			  enum bkey_invalid_flags, struct printbuf *);
 int bch2_mark_snapshot(struct btree_trans *, enum btree_id, unsigned,
-		       struct bkey_s_c, struct bkey_s, unsigned);
+		       struct bkey_s_c, struct bkey_s,
+		       enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_snapshot ((struct bkey_ops) {		\
 	.key_invalid	= bch2_snapshot_invalid,		\
@@ -77,7 +78,7 @@ static inline u32 __bch2_snapshot_parent(struct bch_fs *c, u32 id)
 		return 0;
 
 	u32 parent = s->parent;
-	if (IS_ENABLED(CONFIG_BCACHEFS_DEBU) &&
+	if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG) &&
 	    parent &&
 	    s->depth != snapshot_t(c, parent)->depth + 1)
 		panic("id %u depth=%u parent %u depth=%u\n",
@@ -135,11 +136,6 @@ static inline u32 bch2_snapshot_equiv(struct bch_fs *c, u32 id)
 	return id;
 }
 
-static inline bool bch2_snapshot_is_equiv(struct bch_fs *c, u32 id)
-{
-	return id == bch2_snapshot_equiv(c, id);
-}
-
 static inline int bch2_snapshot_is_internal_node(struct bch_fs *c, u32 id)
 {
 	rcu_read_lock();
diff --git a/fs/bcachefs/str_hash.h b/fs/bcachefs/str_hash.h
index 3976f80721bf1b..cbad9b27874fed 100644
--- a/fs/bcachefs/str_hash.h
+++ b/fs/bcachefs/str_hash.h
@@ -15,16 +15,6 @@
 #include <crypto/hash.h>
 #include <crypto/sha2.h>
 
-typedef unsigned __bitwise bch_str_hash_flags_t;
-
-enum bch_str_hash_flags {
-	__BCH_HASH_SET_MUST_CREATE,
-	__BCH_HASH_SET_MUST_REPLACE,
-};
-
-#define BCH_HASH_SET_MUST_CREATE	(__force bch_str_hash_flags_t) BIT(__BCH_HASH_SET_MUST_CREATE)
-#define BCH_HASH_SET_MUST_REPLACE	(__force bch_str_hash_flags_t) BIT(__BCH_HASH_SET_MUST_REPLACE)
-
 static inline enum bch_str_hash_type
 bch2_str_hash_opt_to_type(struct bch_fs *c, enum bch_str_hash_opts opt)
 {
@@ -159,13 +149,14 @@ static inline bool is_visible_key(struct bch_hash_desc desc, subvol_inum inum, s
 		 desc.is_visible(inum, k));
 }
 
-static __always_inline int
+static __always_inline struct bkey_s_c
 bch2_hash_lookup_in_snapshot(struct btree_trans *trans,
 		 struct btree_iter *iter,
 		 const struct bch_hash_desc desc,
 		 const struct bch_hash_info *info,
 		 subvol_inum inum, const void *key,
-		 unsigned flags, u32 snapshot)
+		 enum btree_iter_update_trigger_flags flags,
+		 u32 snapshot)
 {
 	struct bkey_s_c k;
 	int ret;
@@ -173,10 +164,10 @@ bch2_hash_lookup_in_snapshot(struct btree_trans *trans,
 	for_each_btree_key_upto_norestart(trans, *iter, desc.btree_id,
 			   SPOS(inum.inum, desc.hash_key(info, key), snapshot),
 			   POS(inum.inum, U64_MAX),
-			   BTREE_ITER_SLOTS|flags, k, ret) {
+			   BTREE_ITER_slots|flags, k, ret) {
 		if (is_visible_key(desc, inum, k)) {
 			if (!desc.cmp_key(k, key))
-				return 0;
+				return k;
 		} else if (k.k->type == KEY_TYPE_hash_whiteout) {
 			;
 		} else {
@@ -186,20 +177,23 @@ bch2_hash_lookup_in_snapshot(struct btree_trans *trans,
 	}
 	bch2_trans_iter_exit(trans, iter);
 
-	return ret ?: -BCH_ERR_ENOENT_str_hash_lookup;
+	return bkey_s_c_err(ret ?: -BCH_ERR_ENOENT_str_hash_lookup);
 }
 
-static __always_inline int
+static __always_inline struct bkey_s_c
 bch2_hash_lookup(struct btree_trans *trans,
 		 struct btree_iter *iter,
 		 const struct bch_hash_desc desc,
 		 const struct bch_hash_info *info,
 		 subvol_inum inum, const void *key,
-		 unsigned flags)
+		 enum btree_iter_update_trigger_flags flags)
 {
 	u32 snapshot;
-	return  bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot) ?:
-		bch2_hash_lookup_in_snapshot(trans, iter, desc, info, inum, key, flags, snapshot);
+	int ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+	if (ret)
+		return bkey_s_c_err(ret);
+
+	return bch2_hash_lookup_in_snapshot(trans, iter, desc, info, inum, key, flags, snapshot);
 }
 
 static __always_inline int
@@ -220,7 +214,7 @@ bch2_hash_hole(struct btree_trans *trans,
 	for_each_btree_key_upto_norestart(trans, *iter, desc.btree_id,
 			   SPOS(inum.inum, desc.hash_key(info, key), snapshot),
 			   POS(inum.inum, U64_MAX),
-			   BTREE_ITER_SLOTS|BTREE_ITER_INTENT, k, ret)
+			   BTREE_ITER_slots|BTREE_ITER_intent, k, ret)
 		if (!is_visible_key(desc, inum, k))
 			return 0;
 	bch2_trans_iter_exit(trans, iter);
@@ -242,7 +236,7 @@ int bch2_hash_needs_whiteout(struct btree_trans *trans,
 
 	bch2_btree_iter_advance(&iter);
 
-	for_each_btree_key_continue_norestart(iter, BTREE_ITER_SLOTS, k, ret) {
+	for_each_btree_key_continue_norestart(iter, BTREE_ITER_slots, k, ret) {
 		if (k.k->type != desc.key_type &&
 		    k.k->type != KEY_TYPE_hash_whiteout)
 			break;
@@ -264,8 +258,7 @@ int bch2_hash_set_in_snapshot(struct btree_trans *trans,
 			   const struct bch_hash_info *info,
 			   subvol_inum inum, u32 snapshot,
 			   struct bkey_i *insert,
-			   bch_str_hash_flags_t str_hash_flags,
-			   int update_flags)
+			   enum btree_iter_update_trigger_flags flags)
 {
 	struct btree_iter iter, slot = { NULL };
 	struct bkey_s_c k;
@@ -277,7 +270,7 @@ int bch2_hash_set_in_snapshot(struct btree_trans *trans,
 				desc.hash_bkey(info, bkey_i_to_s_c(insert)),
 				snapshot),
 			   POS(insert->k.p.inode, U64_MAX),
-			   BTREE_ITER_SLOTS|BTREE_ITER_INTENT, k, ret) {
+			   BTREE_ITER_slots|BTREE_ITER_intent, k, ret) {
 		if (is_visible_key(desc, inum, k)) {
 			if (!desc.cmp_bkey(k, bkey_i_to_s_c(insert)))
 				goto found;
@@ -286,8 +279,7 @@ int bch2_hash_set_in_snapshot(struct btree_trans *trans,
 			continue;
 		}
 
-		if (!slot.path &&
-		    !(str_hash_flags & BCH_HASH_SET_MUST_REPLACE))
+		if (!slot.path && !(flags & STR_HASH_must_replace))
 			bch2_trans_copy_iter(&slot, &iter);
 
 		if (k.k->type != KEY_TYPE_hash_whiteout)
@@ -305,16 +297,16 @@ found:
 	found = true;
 not_found:
 
-	if (!found && (str_hash_flags & BCH_HASH_SET_MUST_REPLACE)) {
+	if (!found && (flags & STR_HASH_must_replace)) {
 		ret = -BCH_ERR_ENOENT_str_hash_set_must_replace;
-	} else if (found && (str_hash_flags & BCH_HASH_SET_MUST_CREATE)) {
+	} else if (found && (flags & STR_HASH_must_create)) {
 		ret = -EEXIST;
 	} else {
 		if (!found && slot.path)
 			swap(iter, slot);
 
 		insert->k.p = iter.pos;
-		ret = bch2_trans_update(trans, &iter, insert, update_flags);
+		ret = bch2_trans_update(trans, &iter, insert, flags);
 	}
 
 	goto out;
@@ -326,14 +318,14 @@ int bch2_hash_set(struct btree_trans *trans,
 		  const struct bch_hash_info *info,
 		  subvol_inum inum,
 		  struct bkey_i *insert,
-		  bch_str_hash_flags_t str_hash_flags)
+		  enum btree_iter_update_trigger_flags flags)
 {
 	insert->k.p.inode = inum.inum;
 
 	u32 snapshot;
 	return  bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot) ?:
 		bch2_hash_set_in_snapshot(trans, desc, info, inum,
-					  snapshot, insert, str_hash_flags, 0);
+					  snapshot, insert, flags);
 }
 
 static __always_inline
@@ -341,7 +333,7 @@ int bch2_hash_delete_at(struct btree_trans *trans,
 			const struct bch_hash_desc desc,
 			const struct bch_hash_info *info,
 			struct btree_iter *iter,
-			unsigned update_flags)
+			enum btree_iter_update_trigger_flags flags)
 {
 	struct bkey_i *delete;
 	int ret;
@@ -359,7 +351,7 @@ int bch2_hash_delete_at(struct btree_trans *trans,
 	delete->k.p = iter->pos;
 	delete->k.type = ret ? KEY_TYPE_hash_whiteout : KEY_TYPE_deleted;
 
-	return bch2_trans_update(trans, iter, delete, update_flags);
+	return bch2_trans_update(trans, iter, delete, flags);
 }
 
 static __always_inline
@@ -369,14 +361,10 @@ int bch2_hash_delete(struct btree_trans *trans,
 		     subvol_inum inum, const void *key)
 {
 	struct btree_iter iter;
-	int ret;
-
-	ret = bch2_hash_lookup(trans, &iter, desc, info, inum, key,
-				BTREE_ITER_INTENT);
-	if (ret)
-		return ret;
-
-	ret = bch2_hash_delete_at(trans, desc, info, &iter, 0);
+	struct bkey_s_c k = bch2_hash_lookup(trans, &iter, desc, info, inum, key,
+					     BTREE_ITER_intent);
+	int ret = bkey_err(k) ?:
+		  bch2_hash_delete_at(trans, desc, info, &iter, 0);
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }
diff --git a/fs/bcachefs/subvolume.c b/fs/bcachefs/subvolume.c
index 88a79c82327687..3341285e75317a 100644
--- a/fs/bcachefs/subvolume.c
+++ b/fs/bcachefs/subvolume.c
@@ -162,7 +162,7 @@ int bch2_check_subvols(struct bch_fs *c)
 {
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
-				BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_PREFETCH, k,
+				BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			check_subvol(trans, &iter, k)));
 	bch_err_fn(c, ret);
@@ -198,7 +198,7 @@ int bch2_check_subvol_children(struct bch_fs *c)
 {
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter,
-				BTREE_ID_subvolume_children, POS_MIN, BTREE_ITER_PREFETCH, k,
+				BTREE_ID_subvolume_children, POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			check_subvol_child(trans, &iter, k)));
 	bch_err_fn(c, ret);
@@ -245,9 +245,9 @@ static int subvolume_children_mod(struct btree_trans *trans, struct bpos pos, bo
 int bch2_subvolume_trigger(struct btree_trans *trans,
 			   enum btree_id btree_id, unsigned level,
 			   struct bkey_s_c old, struct bkey_s new,
-			   unsigned flags)
+			   enum btree_iter_update_trigger_flags flags)
 {
-	if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
+	if (flags & BTREE_TRIGGER_transactional) {
 		struct bpos children_pos_old = subvolume_children_pos(old);
 		struct bpos children_pos_new = subvolume_children_pos(new.s_c);
 
@@ -333,7 +333,7 @@ int bch2_subvolume_get_snapshot(struct btree_trans *trans, u32 subvolid,
 
 	subvol = bch2_bkey_get_iter_typed(trans, &iter,
 					  BTREE_ID_subvolumes, POS(0, subvolid),
-					  BTREE_ITER_CACHED|BTREE_ITER_WITH_UPDATES,
+					  BTREE_ITER_cached|BTREE_ITER_with_updates,
 					  subvolume);
 	ret = bkey_err(subvol);
 	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), trans->c,
@@ -383,9 +383,9 @@ static int bch2_subvolumes_reparent(struct btree_trans *trans, u32 subvolid_to_d
 
 	return lockrestart_do(trans,
 			bch2_subvolume_get(trans, subvolid_to_delete, true,
-				   BTREE_ITER_CACHED, &s)) ?:
+				   BTREE_ITER_cached, &s)) ?:
 		for_each_btree_key_commit(trans, iter,
-				BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_PREFETCH, k,
+				BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			bch2_subvolume_reparent(trans, &iter, k,
 					subvolid_to_delete, le32_to_cpu(s.creation_parent)));
@@ -404,7 +404,7 @@ static int __bch2_subvolume_delete(struct btree_trans *trans, u32 subvolid)
 
 	subvol = bch2_bkey_get_iter_typed(trans, &iter,
 				BTREE_ID_subvolumes, POS(0, subvolid),
-				BTREE_ITER_CACHED|BTREE_ITER_INTENT,
+				BTREE_ITER_cached|BTREE_ITER_intent,
 				subvolume);
 	ret = bkey_err(subvol);
 	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), trans->c,
@@ -505,7 +505,7 @@ int bch2_subvolume_unlink(struct btree_trans *trans, u32 subvolid)
 
 	n = bch2_bkey_get_mut_typed(trans, &iter,
 			BTREE_ID_subvolumes, POS(0, subvolid),
-			BTREE_ITER_CACHED, subvolume);
+			BTREE_ITER_cached, subvolume);
 	ret = PTR_ERR_OR_ZERO(n);
 	if (unlikely(ret)) {
 		bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), trans->c,
@@ -547,7 +547,7 @@ int bch2_subvolume_create(struct btree_trans *trans, u64 inode,
 
 		src_subvol = bch2_bkey_get_mut_typed(trans, &src_iter,
 				BTREE_ID_subvolumes, POS(0, src_subvolid),
-				BTREE_ITER_CACHED, subvolume);
+				BTREE_ITER_cached, subvolume);
 		ret = PTR_ERR_OR_ZERO(src_subvol);
 		if (unlikely(ret)) {
 			bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), c,
diff --git a/fs/bcachefs/subvolume.h b/fs/bcachefs/subvolume.h
index d2015d549bd2a3..1530a10327ab36 100644
--- a/fs/bcachefs/subvolume.h
+++ b/fs/bcachefs/subvolume.h
@@ -14,7 +14,8 @@ int bch2_subvolume_invalid(struct bch_fs *, struct bkey_s_c,
 			   enum bkey_invalid_flags, struct printbuf *);
 void bch2_subvolume_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 int bch2_subvolume_trigger(struct btree_trans *, enum btree_id, unsigned,
-			   struct bkey_s_c, struct bkey_s, unsigned);
+			   struct bkey_s_c, struct bkey_s,
+			   enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_subvolume ((struct bkey_ops) {		\
 	.key_invalid	= bch2_subvolume_invalid,		\
diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c
index 08ea3dbbbe97ce..af2d1b1a445cfe 100644
--- a/fs/bcachefs/super-io.c
+++ b/fs/bcachefs/super-io.c
@@ -1150,8 +1150,7 @@ static void bch2_sb_ext_to_text(struct printbuf *out, struct bch_sb *sb,
 {
 	struct bch_sb_field_ext *e = field_to_type(f, ext);
 
-	prt_printf(out, "Recovery passes required:");
-	prt_tab(out);
+	prt_printf(out, "Recovery passes required:\t");
 	prt_bitflags(out, bch2_recovery_passes,
 		     bch2_recovery_passes_from_stable(le64_to_cpu(e->recovery_passes_required[0])));
 	prt_newline(out);
@@ -1160,16 +1159,14 @@ static void bch2_sb_ext_to_text(struct printbuf *out, struct bch_sb *sb,
 	if (errors_silent) {
 		le_bitvector_to_cpu(errors_silent, (void *) e->errors_silent, sizeof(e->errors_silent) * 8);
 
-		prt_printf(out, "Errors to silently fix:");
-		prt_tab(out);
+		prt_printf(out, "Errors to silently fix:\t");
 		prt_bitflags_vector(out, bch2_sb_error_strs, errors_silent, sizeof(e->errors_silent) * 8);
 		prt_newline(out);
 
 		kfree(errors_silent);
 	}
 
-	prt_printf(out, "Btrees with missing data:");
-	prt_tab(out);
+	prt_printf(out, "Btrees with missing data:\t");
 	prt_bitflags(out, __bch2_btree_ids, le64_to_cpu(e->btrees_lost_data));
 	prt_newline(out);
 }
@@ -1277,97 +1274,73 @@ void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb,
 		printbuf_tabstop_push(out, 44);
 
 	for (int i = 0; i < sb->nr_devices; i++)
-		nr_devices += bch2_dev_exists(sb, i);
+		nr_devices += bch2_member_exists(sb, i);
 
-	prt_printf(out, "External UUID:");
-	prt_tab(out);
+	prt_printf(out, "External UUID:\t");
 	pr_uuid(out, sb->user_uuid.b);
 	prt_newline(out);
 
-	prt_printf(out, "Internal UUID:");
-	prt_tab(out);
+	prt_printf(out, "Internal UUID:\t");
 	pr_uuid(out, sb->uuid.b);
 	prt_newline(out);
 
-	prt_printf(out, "Magic number:");
-	prt_tab(out);
+	prt_printf(out, "Magic number:\t");
 	pr_uuid(out, sb->magic.b);
 	prt_newline(out);
 
-	prt_str(out, "Device index:");
-	prt_tab(out);
-	prt_printf(out, "%u", sb->dev_idx);
-	prt_newline(out);
+	prt_printf(out, "Device index:\t%u\n", sb->dev_idx);
 
-	prt_str(out, "Label:");
-	prt_tab(out);
+	prt_str(out, "Label:\t");
 	prt_printf(out, "%.*s", (int) sizeof(sb->label), sb->label);
 	prt_newline(out);
 
-	prt_str(out, "Version:");
-	prt_tab(out);
+	prt_str(out, "Version:\t");
 	bch2_version_to_text(out, le16_to_cpu(sb->version));
 	prt_newline(out);
 
-	prt_str(out, "Version upgrade complete:");
-	prt_tab(out);
+	prt_str(out, "Version upgrade complete:\t");
 	bch2_version_to_text(out, BCH_SB_VERSION_UPGRADE_COMPLETE(sb));
 	prt_newline(out);
 
-	prt_printf(out, "Oldest version on disk:");
-	prt_tab(out);
+	prt_printf(out, "Oldest version on disk:\t");
 	bch2_version_to_text(out, le16_to_cpu(sb->version_min));
 	prt_newline(out);
 
-	prt_printf(out, "Created:");
-	prt_tab(out);
+	prt_printf(out, "Created:\t");
 	if (sb->time_base_lo)
 		bch2_prt_datetime(out, div_u64(le64_to_cpu(sb->time_base_lo), NSEC_PER_SEC));
 	else
 		prt_printf(out, "(not set)");
 	prt_newline(out);
 
-	prt_printf(out, "Sequence number:");
-	prt_tab(out);
+	prt_printf(out, "Sequence number:\t");
 	prt_printf(out, "%llu", le64_to_cpu(sb->seq));
 	prt_newline(out);
 
-	prt_printf(out, "Time of last write:");
-	prt_tab(out);
+	prt_printf(out, "Time of last write:\t");
 	bch2_prt_datetime(out, le64_to_cpu(sb->write_time));
 	prt_newline(out);
 
-	prt_printf(out, "Superblock size:");
-	prt_tab(out);
+	prt_printf(out, "Superblock size:\t");
 	prt_units_u64(out, vstruct_bytes(sb));
 	prt_str(out, "/");
 	prt_units_u64(out, 512ULL << sb->layout.sb_max_size_bits);
 	prt_newline(out);
 
-	prt_printf(out, "Clean:");
-	prt_tab(out);
-	prt_printf(out, "%llu", BCH_SB_CLEAN(sb));
-	prt_newline(out);
-
-	prt_printf(out, "Devices:");
-	prt_tab(out);
-	prt_printf(out, "%u", nr_devices);
-	prt_newline(out);
+	prt_printf(out, "Clean:\t%llu\n", BCH_SB_CLEAN(sb));
+	prt_printf(out, "Devices:\t%u\n", nr_devices);
 
-	prt_printf(out, "Sections:");
+	prt_printf(out, "Sections:\t");
 	vstruct_for_each(sb, f)
 		fields_have |= 1 << le32_to_cpu(f->type);
-	prt_tab(out);
 	prt_bitflags(out, bch2_sb_fields, fields_have);
 	prt_newline(out);
 
-	prt_printf(out, "Features:");
-	prt_tab(out);
+	prt_printf(out, "Features:\t");
 	prt_bitflags(out, bch2_sb_features, le64_to_cpu(sb->features[0]));
 	prt_newline(out);
 
-	prt_printf(out, "Compat features:");
-	prt_tab(out);
+	prt_printf(out, "Compat features:\t");
 	prt_bitflags(out, bch2_sb_compat, le64_to_cpu(sb->compat[0]));
 	prt_newline(out);
 
@@ -1384,8 +1357,7 @@ void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb,
 			if (opt->get_sb != BCH2_NO_SB_OPT) {
 				u64 v = bch2_opt_from_sb(sb, id);
 
-				prt_printf(out, "%s:", opt->attr.name);
-				prt_tab(out);
+				prt_printf(out, "%s:\t", opt->attr.name);
 				bch2_opt_to_text(out, NULL, sb, opt, v,
 						 OPT_HUMAN_READABLE|OPT_SHOW_FULL_LIST);
 				prt_newline(out);
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 88e214c609bb2b..24d673a14cf3b3 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -264,7 +264,6 @@ static void __bch2_fs_read_only(struct bch_fs *c)
 	bch2_open_buckets_stop(c, NULL, true);
 	bch2_rebalance_stop(c);
 	bch2_copygc_stop(c);
-	bch2_gc_thread_stop(c);
 	bch2_fs_ec_flush(c);
 
 	bch_verbose(c, "flushing journal and stopping allocators, journal seq %llu",
@@ -285,7 +284,7 @@ static void __bch2_fs_read_only(struct bch_fs *c)
 	bch_verbose(c, "flushing journal and stopping allocators complete, journal seq %llu",
 		    journal_cur_seq(&c->journal));
 
-	if (test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags) &&
+	if (test_bit(JOURNAL_replay_done, &c->journal.flags) &&
 	    !test_bit(BCH_FS_emergency_ro, &c->flags))
 		set_bit(BCH_FS_clean_shutdown, &c->flags);
 
@@ -467,7 +466,8 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early)
 	 * overwriting whatever was there previously, and there must always be
 	 * at least one non-flush write in the journal or recovery will fail:
 	 */
-	set_bit(JOURNAL_NEED_FLUSH_WRITE, &c->journal.flags);
+	set_bit(JOURNAL_need_flush_write, &c->journal.flags);
+	set_bit(JOURNAL_running, &c->journal.flags);
 
 	for_each_rw_member(c, ca)
 		bch2_dev_allocator_add(c, ca);
@@ -485,12 +485,6 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early)
 	}
 #endif
 
-	ret = bch2_gc_thread_start(c);
-	if (ret) {
-		bch_err(c, "error starting gc thread");
-		return ret;
-	}
-
 	ret = bch2_journal_reclaim_start(&c->journal);
 	if (ret)
 		goto err;
@@ -537,9 +531,7 @@ int bch2_fs_read_write_early(struct bch_fs *c)
 
 static void __bch2_fs_free(struct bch_fs *c)
 {
-	unsigned i;
-
-	for (i = 0; i < BCH_TIME_STAT_NR; i++)
+	for (unsigned i = 0; i < BCH_TIME_STAT_NR; i++)
 		bch2_time_stats_exit(&c->times[i]);
 
 	bch2_find_btree_nodes_exit(&c->found_btree_nodes);
@@ -572,6 +564,7 @@ static void __bch2_fs_free(struct bch_fs *c)
 	BUG_ON(atomic_read(&c->journal_keys.ref));
 	bch2_fs_btree_write_buffer_exit(c);
 	percpu_free_rwsem(&c->mark_lock);
+	EBUG_ON(percpu_u64_get(c->online_reserved));
 	free_percpu(c->online_reserved);
 
 	darray_exit(&c->btree_roots_extra);
@@ -616,8 +609,6 @@ void __bch2_fs_stop(struct bch_fs *c)
 
 	set_bit(BCH_FS_stopping, &c->flags);
 
-	cancel_work_sync(&c->journal_seq_blacklist_gc_work);
-
 	down_write(&c->state_lock);
 	bch2_fs_read_only(c);
 	up_write(&c->state_lock);
@@ -778,6 +769,7 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 	for (i = 0; i < BCH_TIME_STAT_NR; i++)
 		bch2_time_stats_init(&c->times[i]);
 
+	bch2_fs_gc_init(c);
 	bch2_fs_copygc_init(c);
 	bch2_fs_btree_key_cache_init_early(&c->btree_key_cache);
 	bch2_fs_btree_iter_init_early(c);
@@ -800,16 +792,11 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 
 	spin_lock_init(&c->btree_write_error_lock);
 
-	INIT_WORK(&c->journal_seq_blacklist_gc_work,
-		  bch2_blacklist_entries_gc);
-
 	INIT_LIST_HEAD(&c->journal_iters);
 
 	INIT_LIST_HEAD(&c->fsck_error_msgs);
 	mutex_init(&c->fsck_error_msgs_lock);
 
-	seqcount_init(&c->gc_pos_lock);
-
 	seqcount_init(&c->usage_lock);
 
 	sema_init(&c->io_in_flight, 128);
@@ -940,7 +927,7 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 		goto err;
 
 	for (i = 0; i < c->sb.nr_devices; i++)
-		if (bch2_dev_exists(c->disk_sb.sb, i) &&
+		if (bch2_member_exists(c->disk_sb.sb, i) &&
 		    bch2_dev_alloc(c, i)) {
 			ret = -EEXIST;
 			goto err;
@@ -1101,7 +1088,7 @@ static int bch2_dev_in_fs(struct bch_sb_handle *fs,
 	if (!uuid_equal(&fs->sb->uuid, &sb->sb->uuid))
 		return -BCH_ERR_device_not_a_member_of_filesystem;
 
-	if (!bch2_dev_exists(fs->sb, sb->sb->dev_idx))
+	if (!bch2_member_exists(fs->sb, sb->sb->dev_idx))
 		return -BCH_ERR_device_has_been_removed;
 
 	if (fs->sb->block_size != sb->sb->block_size)
@@ -1200,11 +1187,11 @@ static void bch2_dev_free(struct bch_dev *ca)
 	if (ca->kobj.state_in_sysfs)
 		kobject_del(&ca->kobj);
 
+	kfree(ca->buckets_nouse);
 	bch2_free_super(&ca->disk_sb);
 	bch2_dev_journal_exit(ca);
 
 	free_percpu(ca->io_done);
-	bioset_exit(&ca->replica_set);
 	bch2_dev_buckets_free(ca);
 	free_page((unsigned long) ca->sb_read_scratch);
 
@@ -1319,8 +1306,6 @@ static struct bch_dev *__bch2_dev_alloc(struct bch_fs *c,
 			    PERCPU_REF_INIT_DEAD, GFP_KERNEL) ||
 	    !(ca->sb_read_scratch = (void *) __get_free_page(GFP_KERNEL)) ||
 	    bch2_dev_buckets_alloc(c, ca) ||
-	    bioset_init(&ca->replica_set, 4,
-			offsetof(struct bch_write_bio, bio), 0) ||
 	    !(ca->io_done	= alloc_percpu(*ca->io_done)))
 		goto err;
 
@@ -1411,10 +1396,9 @@ static int bch2_dev_attach_bdev(struct bch_fs *c, struct bch_sb_handle *sb)
 	    le64_to_cpu(c->disk_sb.sb->seq))
 		bch2_sb_to_fs(c, sb->sb);
 
-	BUG_ON(sb->sb->dev_idx >= c->sb.nr_devices ||
-	       !c->devs[sb->sb->dev_idx]);
+	BUG_ON(!bch2_dev_exists(c, sb->sb->dev_idx));
 
-	ca = bch_dev_locked(c, sb->sb->dev_idx);
+	ca = bch2_dev_locked(c, sb->sb->dev_idx);
 
 	ret = __bch2_dev_attach_bdev(ca, sb);
 	if (ret)
@@ -1506,10 +1490,10 @@ static bool bch2_fs_may_start(struct bch_fs *c)
 		mutex_lock(&c->sb_lock);
 
 		for (i = 0; i < c->disk_sb.sb->nr_devices; i++) {
-			if (!bch2_dev_exists(c->disk_sb.sb, i))
+			if (!bch2_member_exists(c->disk_sb.sb, i))
 				continue;
 
-			ca = bch_dev_locked(c, i);
+			ca = bch2_dev_locked(c, i);
 
 			if (!bch2_dev_is_online(ca) &&
 			    (ca->mi.state == BCH_MEMBER_STATE_rw ||
@@ -1599,17 +1583,17 @@ static int bch2_dev_remove_alloc(struct bch_fs *c, struct bch_dev *ca)
 	 * with bch2_do_invalidates() and bch2_do_discards()
 	 */
 	ret =   bch2_btree_delete_range(c, BTREE_ID_lru, start, end,
-					BTREE_TRIGGER_NORUN, NULL) ?:
+					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_need_discard, start, end,
-					BTREE_TRIGGER_NORUN, NULL) ?:
+					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_freespace, start, end,
-					BTREE_TRIGGER_NORUN, NULL) ?:
+					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_backpointers, start, end,
-					BTREE_TRIGGER_NORUN, NULL) ?:
+					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_alloc, start, end,
-					BTREE_TRIGGER_NORUN, NULL) ?:
+					BTREE_TRIGGER_norun, NULL) ?:
 		bch2_btree_delete_range(c, BTREE_ID_bucket_gens, start, end,
-					BTREE_TRIGGER_NORUN, NULL);
+					BTREE_TRIGGER_norun, NULL);
 	bch_err_msg(c, ret, "removing dev alloc info");
 	return ret;
 }
@@ -1777,9 +1761,28 @@ int bch2_dev_add(struct bch_fs *c, const char *path)
 	if (dynamic_fault("bcachefs:add:no_slot"))
 		goto no_slot;
 
-	for (dev_idx = 0; dev_idx < BCH_SB_MEMBERS_MAX; dev_idx++)
-		if (!bch2_dev_exists(c->disk_sb.sb, dev_idx))
-			goto have_slot;
+	if (c->sb.nr_devices < BCH_SB_MEMBERS_MAX) {
+		dev_idx = c->sb.nr_devices;
+		goto have_slot;
+	}
+
+	int best = -1;
+	u64 best_last_mount = 0;
+	for (dev_idx = 0; dev_idx < BCH_SB_MEMBERS_MAX; dev_idx++) {
+		struct bch_member m = bch2_sb_member_get(c->disk_sb.sb, dev_idx);
+		if (bch2_member_alive(&m))
+			continue;
+
+		u64 last_mount = le64_to_cpu(m.last_mount);
+		if (best < 0 || last_mount < best_last_mount) {
+			best = dev_idx;
+			best_last_mount = last_mount;
+		}
+	}
+	if (best >= 0) {
+		dev_idx = best;
+		goto have_slot;
+	}
 no_slot:
 	ret = -BCH_ERR_ENOSPC_sb_members;
 	bch_err_msg(c, ret, "setting up new superblock");
@@ -1821,7 +1824,7 @@ have_slot:
 
 	bch2_dev_usage_journal_reserve(c);
 
-	ret = bch2_trans_mark_dev_sb(c, ca);
+	ret = bch2_trans_mark_dev_sb(c, ca, BTREE_TRIGGER_transactional);
 	bch_err_msg(ca, ret, "marking new superblock");
 	if (ret)
 		goto err_late;
@@ -1884,9 +1887,9 @@ int bch2_dev_online(struct bch_fs *c, const char *path)
 	if (ret)
 		goto err;
 
-	ca = bch_dev_locked(c, dev_idx);
+	ca = bch2_dev_locked(c, dev_idx);
 
-	ret = bch2_trans_mark_dev_sb(c, ca);
+	ret = bch2_trans_mark_dev_sb(c, ca, BTREE_TRIGGER_transactional);
 	bch_err_msg(c, ret, "bringing %s online: error from bch2_trans_mark_dev_sb", path);
 	if (ret)
 		goto err;
@@ -1972,7 +1975,7 @@ int bch2_dev_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 	if (ret)
 		goto err;
 
-	ret = bch2_trans_mark_dev_sb(c, ca);
+	ret = bch2_trans_mark_dev_sb(c, ca, BTREE_TRIGGER_transactional);
 	if (ret)
 		goto err;
 
diff --git a/fs/bcachefs/super_types.h b/fs/bcachefs/super_types.h
index 11bcef170c2c22..368a63d938cfc9 100644
--- a/fs/bcachefs/super_types.h
+++ b/fs/bcachefs/super_types.h
@@ -26,19 +26,4 @@ struct bch_devs_list {
 	u8			data[BCH_BKEY_PTRS_MAX];
 };
 
-struct bch_member_cpu {
-	u64			nbuckets;	/* device size */
-	u16			first_bucket;   /* index of first bucket used */
-	u16			bucket_size;	/* sectors */
-	u16			group;
-	u8			state;
-	u8			discard;
-	u8			data_allowed;
-	u8			durability;
-	u8			freespace_initialized;
-	u8			valid;
-	u8			btree_bitmap_shift;
-	u64			btree_allocated_bitmap;
-};
-
 #endif /* _BCACHEFS_SUPER_TYPES_H */
diff --git a/fs/bcachefs/sysfs.c b/fs/bcachefs/sysfs.c
index 5be92fe3f4ea4e..df020e6c485470 100644
--- a/fs/bcachefs/sysfs.c
+++ b/fs/bcachefs/sysfs.c
@@ -140,9 +140,8 @@ write_attribute(trigger_gc);
 write_attribute(trigger_discards);
 write_attribute(trigger_invalidates);
 write_attribute(trigger_journal_flush);
-write_attribute(prune_cache);
-write_attribute(btree_wakeup);
-rw_attribute(btree_gc_periodic);
+write_attribute(trigger_btree_cache_shrink);
+write_attribute(trigger_btree_key_cache_shrink);
 rw_attribute(gc_gens_pos);
 
 read_attribute(uuid);
@@ -189,12 +188,8 @@ static void bch2_write_refs_to_text(struct printbuf *out, struct bch_fs *c)
 {
 	bch2_printbuf_tabstop_push(out, 24);
 
-	for (unsigned i = 0; i < ARRAY_SIZE(c->writes); i++) {
-		prt_str(out, bch2_write_refs[i]);
-		prt_tab(out);
-		prt_printf(out, "%li", atomic_long_read(&c->writes[i]));
-		prt_newline(out);
-	}
+	for (unsigned i = 0; i < ARRAY_SIZE(c->writes); i++)
+		prt_printf(out, "%s\t%li\n", bch2_write_refs[i], atomic_long_read(&c->writes[i]));
 }
 #endif
 
@@ -278,7 +273,7 @@ static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c
 			continue;
 
 		ret = for_each_btree_key(trans, iter, id, POS_MIN,
-					 BTREE_ITER_ALL_SNAPSHOTS, k, ({
+					 BTREE_ITER_all_snapshots, k, ({
 			struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
 			struct bch_extent_crc_unpacked crc;
 			const union bch_extent_entry *entry;
@@ -313,22 +308,11 @@ static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c
 	if (ret)
 		return ret;
 
-	prt_str(out, "type");
 	printbuf_tabstop_push(out, 12);
-	prt_tab(out);
-
-	prt_str(out, "compressed");
 	printbuf_tabstop_push(out, 16);
-	prt_tab_rjust(out);
-
-	prt_str(out, "uncompressed");
 	printbuf_tabstop_push(out, 16);
-	prt_tab_rjust(out);
-
-	prt_str(out, "average extent size");
 	printbuf_tabstop_push(out, 24);
-	prt_tab_rjust(out);
-	prt_newline(out);
+	prt_printf(out, "type\tcompressed\runcompressed\raverage extent size\r\n");
 
 	for (unsigned i = 0; i < ARRAY_SIZE(s); i++) {
 		bch2_prt_compression_type(out, i);
@@ -362,19 +346,32 @@ static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 	prt_printf(out, "\n");
 }
 
-static void bch2_btree_wakeup_all(struct bch_fs *c)
+static void fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c)
 {
-	struct btree_trans *trans;
+	unsigned nr[BCH_DATA_NR];
 
-	seqmutex_lock(&c->btree_trans_lock);
-	list_for_each_entry(trans, &c->btree_trans_list, list) {
-		struct btree_bkey_cached_common *b = READ_ONCE(trans->locking);
+	memset(nr, 0, sizeof(nr));
 
-		if (b)
-			six_lock_wakeup_all(&b->lock);
+	for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
+		nr[c->open_buckets[i].data_type]++;
 
-	}
-	seqmutex_unlock(&c->btree_trans_lock);
+	printbuf_tabstop_push(out, 24);
+	prt_printf(out, "hidden\t%llu\n",		bch2_fs_usage_read_one(c, &c->usage_base->b.hidden));
+	prt_printf(out, "btree\t%llu\n",		bch2_fs_usage_read_one(c, &c->usage_base->b.btree));
+	prt_printf(out, "data\t%llu\n",			bch2_fs_usage_read_one(c, &c->usage_base->b.data));
+	prt_printf(out, "cached\t%llu\n",		bch2_fs_usage_read_one(c, &c->usage_base->b.cached));
+	prt_printf(out, "reserved\t%llu\n",		bch2_fs_usage_read_one(c, &c->usage_base->b.reserved));
+	prt_printf(out, "online_reserved\t%llu\n",	percpu_u64_get(c->online_reserved));
+	prt_printf(out, "nr_inodes\t%llu\n",		bch2_fs_usage_read_one(c, &c->usage_base->b.nr_inodes));
+
+	prt_newline(out);
+	prt_printf(out, "freelist_wait\t%s\n",			c->freelist_wait.list.first ? "waiting" : "empty");
+	prt_printf(out, "open buckets allocated\t%i\n",		OPEN_BUCKETS_COUNT - c->open_buckets_nr_free);
+	prt_printf(out, "open buckets total\t%u\n",		OPEN_BUCKETS_COUNT);
+	prt_printf(out, "open_buckets_wait\t%s\n",		c->open_buckets_wait.list.first ? "waiting" : "empty");
+	prt_printf(out, "open_buckets_btree\t%u\n",		nr[BCH_DATA_btree]);
+	prt_printf(out, "open_buckets_user\t%u\n",		nr[BCH_DATA_user]);
+	prt_printf(out, "btree reserve cache\t%u\n",		c->btree_reserve_cache_nr);
 }
 
 SHOW(bch2_fs)
@@ -392,8 +389,6 @@ SHOW(bch2_fs)
 	if (attr == &sysfs_btree_write_stats)
 		bch2_btree_write_stats_to_text(out, c);
 
-	sysfs_printf(btree_gc_periodic, "%u",	(int) c->btree_gc_periodic);
-
 	if (attr == &sysfs_gc_gens_pos)
 		bch2_gc_gens_pos_to_text(out, c);
 
@@ -459,6 +454,9 @@ SHOW(bch2_fs)
 	if (attr == &sysfs_disk_groups)
 		bch2_disk_groups_to_text(out, c);
 
+	if (attr == &sysfs_alloc_debug)
+		fs_alloc_debug_to_text(out, c);
+
 	return 0;
 }
 
@@ -466,14 +464,6 @@ STORE(bch2_fs)
 {
 	struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 
-	if (attr == &sysfs_btree_gc_periodic) {
-		ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
-			?: (ssize_t) size;
-
-		wake_up_process(c->gc_thread);
-		return ret;
-	}
-
 	if (attr == &sysfs_copy_gc_enabled) {
 		ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 			?: (ssize_t) size;
@@ -505,7 +495,7 @@ STORE(bch2_fs)
 	if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_sysfs))
 		return -EROFS;
 
-	if (attr == &sysfs_prune_cache) {
+	if (attr == &sysfs_trigger_btree_cache_shrink) {
 		struct shrink_control sc;
 
 		sc.gfp_mask = GFP_KERNEL;
@@ -513,22 +503,17 @@ STORE(bch2_fs)
 		c->btree_cache.shrink->scan_objects(c->btree_cache.shrink, &sc);
 	}
 
-	if (attr == &sysfs_btree_wakeup)
-		bch2_btree_wakeup_all(c);
-
-	if (attr == &sysfs_trigger_gc) {
-		/*
-		 * Full gc is currently incompatible with btree key cache:
-		 */
-#if 0
-		down_read(&c->state_lock);
-		bch2_gc(c, false, false);
-		up_read(&c->state_lock);
-#else
-		bch2_gc_gens(c);
-#endif
+	if (attr == &sysfs_trigger_btree_key_cache_shrink) {
+		struct shrink_control sc;
+
+		sc.gfp_mask = GFP_KERNEL;
+		sc.nr_to_scan = strtoul_or_return(buf);
+		c->btree_key_cache.shrink->scan_objects(c->btree_cache.shrink, &sc);
 	}
 
+	if (attr == &sysfs_trigger_gc)
+		bch2_gc_gens(c);
+
 	if (attr == &sysfs_trigger_discards)
 		bch2_do_discards(c);
 
@@ -594,13 +579,11 @@ SHOW(bch2_fs_counters)
 		if (attr == &sysfs_##t) {					\
 			counter             = percpu_u64_get(&c->counters[BCH_COUNTER_##t]);\
 			counter_since_mount = counter - c->counters_on_mount[BCH_COUNTER_##t];\
-			prt_printf(out, "since mount:");				\
-			prt_tab(out);						\
+			prt_printf(out, "since mount:\t");			\
 			prt_human_readable_u64(out, counter_since_mount);	\
 			prt_newline(out);					\
 										\
-			prt_printf(out, "since filesystem creation:");		\
-			prt_tab(out);						\
+			prt_printf(out, "since filesystem creation:\t");	\
 			prt_human_readable_u64(out, counter);			\
 			prt_newline(out);					\
 		}
@@ -660,8 +643,8 @@ struct attribute *bch2_fs_internal_files[] = {
 	&sysfs_trigger_discards,
 	&sysfs_trigger_invalidates,
 	&sysfs_trigger_journal_flush,
-	&sysfs_prune_cache,
-	&sysfs_btree_wakeup,
+	&sysfs_trigger_btree_cache_shrink,
+	&sysfs_trigger_btree_key_cache_shrink,
 
 	&sysfs_gc_gens_pos,
 
@@ -677,6 +660,7 @@ struct attribute *bch2_fs_internal_files[] = {
 	&sysfs_internal_uuid,
 
 	&sysfs_disk_groups,
+	&sysfs_alloc_debug,
 	NULL
 };
 
@@ -796,11 +780,11 @@ static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 {
 	struct bch_fs *c = ca->fs;
 	struct bch_dev_usage stats = bch2_dev_usage_read(ca);
-	unsigned i, nr[BCH_DATA_NR];
+	unsigned nr[BCH_DATA_NR];
 
 	memset(nr, 0, sizeof(nr));
 
-	for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
+	for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 		nr[c->open_buckets[i].data_type]++;
 
 	printbuf_tabstop_push(out, 8);
@@ -813,65 +797,24 @@ static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 
 	prt_newline(out);
 
-	prt_printf(out, "reserves:");
-	prt_newline(out);
-	for (i = 0; i < BCH_WATERMARK_NR; i++) {
-		prt_str(out, bch2_watermarks[i]);
-		prt_tab(out);
-		prt_u64(out, bch2_dev_buckets_reserved(ca, i));
-		prt_tab_rjust(out);
-		prt_newline(out);
-	}
+	prt_printf(out, "reserves:\n");
+	for (unsigned i = 0; i < BCH_WATERMARK_NR; i++)
+		prt_printf(out, "%s\t%llu\r\n", bch2_watermarks[i], bch2_dev_buckets_reserved(ca, i));
 
 	prt_newline(out);
 
 	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, 24);
 
-	prt_str(out, "freelist_wait");
-	prt_tab(out);
-	prt_str(out, c->freelist_wait.list.first ? "waiting" : "empty");
-	prt_newline(out);
-
-	prt_str(out, "open buckets allocated");
-	prt_tab(out);
-	prt_u64(out, OPEN_BUCKETS_COUNT - c->open_buckets_nr_free);
-	prt_newline(out);
-
-	prt_str(out, "open buckets this dev");
-	prt_tab(out);
-	prt_u64(out, ca->nr_open_buckets);
-	prt_newline(out);
-
-	prt_str(out, "open buckets total");
-	prt_tab(out);
-	prt_u64(out, OPEN_BUCKETS_COUNT);
-	prt_newline(out);
-
-	prt_str(out, "open_buckets_wait");
-	prt_tab(out);
-	prt_str(out, c->open_buckets_wait.list.first ? "waiting" : "empty");
-	prt_newline(out);
-
-	prt_str(out, "open_buckets_btree");
-	prt_tab(out);
-	prt_u64(out, nr[BCH_DATA_btree]);
-	prt_newline(out);
-
-	prt_str(out, "open_buckets_user");
-	prt_tab(out);
-	prt_u64(out, nr[BCH_DATA_user]);
-	prt_newline(out);
-
-	prt_str(out, "buckets_to_invalidate");
-	prt_tab(out);
-	prt_u64(out, should_invalidate_buckets(ca, stats));
-	prt_newline(out);
-
-	prt_str(out, "btree reserve cache");
-	prt_tab(out);
-	prt_u64(out, c->btree_reserve_cache_nr);
-	prt_newline(out);
+	prt_printf(out, "freelist_wait\t%s\n",			c->freelist_wait.list.first ? "waiting" : "empty");
+	prt_printf(out, "open buckets allocated\t%i\n",		OPEN_BUCKETS_COUNT - c->open_buckets_nr_free);
+	prt_printf(out, "open buckets this dev\t%i\n",		ca->nr_open_buckets);
+	prt_printf(out, "open buckets total\t%u\n",		OPEN_BUCKETS_COUNT);
+	prt_printf(out, "open_buckets_wait\t%s\n",		c->open_buckets_wait.list.first ? "waiting" : "empty");
+	prt_printf(out, "open_buckets_btree\t%u\n",		nr[BCH_DATA_btree]);
+	prt_printf(out, "open_buckets_user\t%u\n",		nr[BCH_DATA_user]);
+	prt_printf(out, "buckets_to_invalidate\t%llu\n",	should_invalidate_buckets(ca, stats));
+	prt_printf(out, "btree reserve cache\t%u\n",		c->btree_reserve_cache_nr);
 }
 
 static const char * const bch2_rw[] = {
diff --git a/fs/bcachefs/tests.c b/fs/bcachefs/tests.c
index bfec656f94c075..68104b2056d94d 100644
--- a/fs/bcachefs/tests.c
+++ b/fs/bcachefs/tests.c
@@ -40,7 +40,7 @@ static int test_delete(struct bch_fs *c, u64 nr)
 	k.k.p.snapshot = U32_MAX;
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_xattrs, k.k.p,
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	ret = commit_do(trans, NULL, NULL, 0,
 		bch2_btree_iter_traverse(&iter) ?:
@@ -81,7 +81,7 @@ static int test_delete_written(struct bch_fs *c, u64 nr)
 	k.k.p.snapshot = U32_MAX;
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_xattrs, k.k.p,
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	ret = commit_do(trans, NULL, NULL, 0,
 		bch2_btree_iter_traverse(&iter) ?:
@@ -261,7 +261,7 @@ static int test_iterate_slots(struct bch_fs *c, u64 nr)
 	ret = bch2_trans_run(c,
 		for_each_btree_key_upto(trans, iter, BTREE_ID_xattrs,
 					SPOS(0, 0, U32_MAX), POS(0, U64_MAX),
-					BTREE_ITER_SLOTS, k, ({
+					BTREE_ITER_slots, k, ({
 			if (i >= nr * 2)
 				break;
 
@@ -322,7 +322,7 @@ static int test_iterate_slots_extents(struct bch_fs *c, u64 nr)
 	ret = bch2_trans_run(c,
 		for_each_btree_key_upto(trans, iter, BTREE_ID_extents,
 					SPOS(0, 0, U32_MAX), POS(0, U64_MAX),
-					BTREE_ITER_SLOTS, k, ({
+					BTREE_ITER_slots, k, ({
 			if (i == nr)
 				break;
 			BUG_ON(bkey_deleted(k.k) != !(i % 16));
@@ -452,7 +452,7 @@ static int insert_test_overlapping_extent(struct bch_fs *c, u64 inum, u64 start,
 
 	ret = bch2_trans_do(c, NULL, NULL, 0,
 		bch2_btree_insert_nonextent(trans, BTREE_ID_extents, &k.k_i,
-					    BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE));
+					    BTREE_UPDATE_internal_snapshot_node));
 	bch_err_fn(c, ret);
 	return ret;
 }
@@ -671,7 +671,7 @@ static int __do_delete(struct btree_trans *trans, struct bpos pos)
 	int ret = 0;
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_xattrs, pos,
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	k = bch2_btree_iter_peek_upto(&iter, POS(0, U64_MAX));
 	ret = bkey_err(k);
 	if (ret)
@@ -714,7 +714,7 @@ static int seq_insert(struct bch_fs *c, u64 nr)
 	return bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_xattrs,
 					SPOS(0, 0, U32_MAX),
-					BTREE_ITER_SLOTS|BTREE_ITER_INTENT, k,
+					BTREE_ITER_slots|BTREE_ITER_intent, k,
 					NULL, NULL, 0, ({
 			if (iter.pos.offset >= nr)
 				break;
@@ -737,7 +737,7 @@ static int seq_overwrite(struct bch_fs *c, u64 nr)
 	return bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_xattrs,
 					SPOS(0, 0, U32_MAX),
-					BTREE_ITER_INTENT, k,
+					BTREE_ITER_intent, k,
 					NULL, NULL, 0, ({
 			struct bkey_i_cookie u;
 
diff --git a/fs/bcachefs/util.c b/fs/bcachefs/util.c
index 92c6ad75e702ab..de331dec2a99cf 100644
--- a/fs/bcachefs/util.c
+++ b/fs/bcachefs/util.c
@@ -348,15 +348,12 @@ static void bch2_pr_time_units_aligned(struct printbuf *out, u64 ns)
 {
 	const struct time_unit *u = bch2_pick_time_units(ns);
 
-	prt_printf(out, "%llu ", div64_u64(ns, u->nsecs));
-	prt_tab_rjust(out);
-	prt_printf(out, "%s", u->name);
+	prt_printf(out, "%llu \r%s", div64_u64(ns, u->nsecs), u->name);
 }
 
 static inline void pr_name_and_units(struct printbuf *out, const char *name, u64 ns)
 {
-	prt_str(out, name);
-	prt_tab(out);
+	prt_printf(out, "%s\t", name);
 	bch2_pr_time_units_aligned(out, ns);
 	prt_newline(out);
 }
@@ -389,12 +386,8 @@ void bch2_time_stats_to_text(struct printbuf *out, struct bch2_time_stats *stats
 	}
 
 	printbuf_tabstop_push(out, out->indent + TABSTOP_SIZE);
-	prt_printf(out, "count:");
-	prt_tab(out);
-	prt_printf(out, "%llu ",
-			 stats->duration_stats.n);
+	prt_printf(out, "count:\t%llu\n", stats->duration_stats.n);
 	printbuf_tabstop_pop(out);
-	prt_newline(out);
 
 	printbuf_tabstops_reset(out);
 
@@ -403,13 +396,8 @@ void bch2_time_stats_to_text(struct printbuf *out, struct bch2_time_stats *stats
 	printbuf_tabstop_push(out, 0);
 	printbuf_tabstop_push(out, TABSTOP_SIZE + 2);
 
-	prt_tab(out);
-	prt_printf(out, "since mount");
-	prt_tab_rjust(out);
-	prt_tab(out);
+	prt_printf(out, "\tsince mount\r\trecent\r\n");
 	prt_printf(out, "recent");
-	prt_tab_rjust(out);
-	prt_newline(out);
 
 	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, out->indent + 20);
@@ -417,23 +405,20 @@ void bch2_time_stats_to_text(struct printbuf *out, struct bch2_time_stats *stats
 	printbuf_tabstop_push(out, 2);
 	printbuf_tabstop_push(out, TABSTOP_SIZE);
 
-	prt_printf(out, "duration of events");
-	prt_newline(out);
+	prt_printf(out, "duration of events\n");
 	printbuf_indent_add(out, 2);
 
 	pr_name_and_units(out, "min:", stats->min_duration);
 	pr_name_and_units(out, "max:", stats->max_duration);
 	pr_name_and_units(out, "total:", stats->total_duration);
 
-	prt_printf(out, "mean:");
-	prt_tab(out);
+	prt_printf(out, "mean:\t");
 	bch2_pr_time_units_aligned(out, d_mean);
 	prt_tab(out);
 	bch2_pr_time_units_aligned(out, mean_and_variance_weighted_get_mean(stats->duration_stats_weighted, TIME_STATS_MV_WEIGHT));
 	prt_newline(out);
 
-	prt_printf(out, "stddev:");
-	prt_tab(out);
+	prt_printf(out, "stddev:\t");
 	bch2_pr_time_units_aligned(out, d_stddev);
 	prt_tab(out);
 	bch2_pr_time_units_aligned(out, mean_and_variance_weighted_get_stddev(stats->duration_stats_weighted, TIME_STATS_MV_WEIGHT));
@@ -441,22 +426,19 @@ void bch2_time_stats_to_text(struct printbuf *out, struct bch2_time_stats *stats
 	printbuf_indent_sub(out, 2);
 	prt_newline(out);
 
-	prt_printf(out, "time between events");
-	prt_newline(out);
+	prt_printf(out, "time between events\n");
 	printbuf_indent_add(out, 2);
 
 	pr_name_and_units(out, "min:", stats->min_freq);
 	pr_name_and_units(out, "max:", stats->max_freq);
 
-	prt_printf(out, "mean:");
-	prt_tab(out);
+	prt_printf(out, "mean:\t");
 	bch2_pr_time_units_aligned(out, f_mean);
 	prt_tab(out);
 	bch2_pr_time_units_aligned(out, mean_and_variance_weighted_get_mean(stats->freq_stats_weighted, TIME_STATS_MV_WEIGHT));
 	prt_newline(out);
 
-	prt_printf(out, "stddev:");
-	prt_tab(out);
+	prt_printf(out, "stddev:\t");
 	bch2_pr_time_units_aligned(out, f_stddev);
 	prt_tab(out);
 	bch2_pr_time_units_aligned(out, mean_and_variance_weighted_get_stddev(stats->freq_stats_weighted, TIME_STATS_MV_WEIGHT));
@@ -589,40 +571,31 @@ void bch2_pd_controller_debug_to_text(struct printbuf *out, struct bch_pd_contro
 	if (!out->nr_tabstops)
 		printbuf_tabstop_push(out, 20);
 
-	prt_printf(out, "rate:");
-	prt_tab(out);
+	prt_printf(out, "rate:\t");
 	prt_human_readable_s64(out, pd->rate.rate);
 	prt_newline(out);
 
-	prt_printf(out, "target:");
-	prt_tab(out);
+	prt_printf(out, "target:\t");
 	prt_human_readable_u64(out, pd->last_target);
 	prt_newline(out);
 
-	prt_printf(out, "actual:");
-	prt_tab(out);
+	prt_printf(out, "actual:\t");
 	prt_human_readable_u64(out, pd->last_actual);
 	prt_newline(out);
 
-	prt_printf(out, "proportional:");
-	prt_tab(out);
+	prt_printf(out, "proportional:\t");
 	prt_human_readable_s64(out, pd->last_proportional);
 	prt_newline(out);
 
-	prt_printf(out, "derivative:");
-	prt_tab(out);
+	prt_printf(out, "derivative:\t");
 	prt_human_readable_s64(out, pd->last_derivative);
 	prt_newline(out);
 
-	prt_printf(out, "change:");
-	prt_tab(out);
+	prt_printf(out, "change:\t");
 	prt_human_readable_s64(out, pd->last_change);
 	prt_newline(out);
 
-	prt_printf(out, "next io:");
-	prt_tab(out);
-	prt_printf(out, "%llims", div64_s64(pd->rate.next - local_clock(), NSEC_PER_MSEC));
-	prt_newline(out);
+	prt_printf(out, "next io:\t%llims\n", div64_s64(pd->rate.next - local_clock(), NSEC_PER_MSEC));
 }
 
 /* misc: */
diff --git a/fs/bcachefs/xattr.c b/fs/bcachefs/xattr.c
index 754f17bba68edb..e81db0614436da 100644
--- a/fs/bcachefs/xattr.c
+++ b/fs/bcachefs/xattr.c
@@ -138,21 +138,13 @@ static int bch2_xattr_get_trans(struct btree_trans *trans, struct bch_inode_info
 	struct bch_hash_info hash = bch2_hash_info_init(trans->c, &inode->ei_inode);
 	struct xattr_search_key search = X_SEARCH(type, name, strlen(name));
 	struct btree_iter iter;
-	struct bkey_s_c_xattr xattr;
-	struct bkey_s_c k;
-	int ret;
-
-	ret = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc, &hash,
-			       inode_inum(inode), &search, 0);
-	if (ret)
-		goto err1;
-
-	k = bch2_btree_iter_peek_slot(&iter);
-	ret = bkey_err(k);
+	struct bkey_s_c k = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc, &hash,
+					     inode_inum(inode), &search, 0);
+	int ret = bkey_err(k);
 	if (ret)
-		goto err2;
+		return ret;
 
-	xattr = bkey_s_c_to_xattr(k);
+	struct bkey_s_c_xattr xattr = bkey_s_c_to_xattr(k);
 	ret = le16_to_cpu(xattr.v->x_val_len);
 	if (buffer) {
 		if (ret > size)
@@ -160,10 +152,8 @@ static int bch2_xattr_get_trans(struct btree_trans *trans, struct bch_inode_info
 		else
 			memcpy(buffer, xattr_val(xattr.v), ret);
 	}
-err2:
 	bch2_trans_iter_exit(trans, &iter);
-err1:
-	return ret < 0 && bch2_err_matches(ret, ENOENT) ? -ENODATA : ret;
+	return ret;
 }
 
 int bch2_xattr_set(struct btree_trans *trans, subvol_inum inum,
@@ -177,7 +167,7 @@ int bch2_xattr_set(struct btree_trans *trans, subvol_inum inum,
 	int ret;
 
 	ret   = bch2_subvol_is_ro_trans(trans, inum.subvol) ?:
-		bch2_inode_peek(trans, &inode_iter, inode_u, inum, BTREE_ITER_INTENT);
+		bch2_inode_peek(trans, &inode_iter, inode_u, inum, BTREE_ITER_intent);
 	if (ret)
 		return ret;
 
@@ -212,8 +202,8 @@ int bch2_xattr_set(struct btree_trans *trans, subvol_inum inum,
 
 		ret = bch2_hash_set(trans, bch2_xattr_hash_desc, hash_info,
 			      inum, &xattr->k_i,
-			      (flags & XATTR_CREATE ? BCH_HASH_SET_MUST_CREATE : 0)|
-			      (flags & XATTR_REPLACE ? BCH_HASH_SET_MUST_REPLACE : 0));
+			      (flags & XATTR_CREATE ? STR_HASH_must_create : 0)|
+			      (flags & XATTR_REPLACE ? STR_HASH_must_replace : 0));
 	} else {
 		struct xattr_search_key search =
 			X_SEARCH(type, name, strlen(name));
@@ -359,6 +349,9 @@ static int bch2_xattr_get_handler(const struct xattr_handler *handler,
 	int ret = bch2_trans_do(c, NULL, NULL, 0,
 		bch2_xattr_get_trans(trans, inode, name, buffer, size, handler->flags));
 
+	if (ret < 0 && bch2_err_matches(ret, ENOENT))
+		ret = -ENODATA;
+
 	return bch2_err_class(ret);
 }