Update bcachefs sources to 99175e5712 bcachefs: Fix bch2_check_discard_freespace_key()

11 files changed, 92 insertions, 47 deletions

diff --git a/.bcachefs_revision b/.bcachefs_revision
index b9058b5e..6ec4d5d7 100644
--- a/.bcachefs_revision
+++ b/.bcachefs_revision
@@ -1 +1 @@
-01d7ad6d95c85cf5434a891e6dd7971797e0e1fa
+99175e5712ecc323930db29565b43d5d0e55a276
diff --git a/libbcachefs/alloc_background.c b/libbcachefs/alloc_background.c
index f774a660..00cdfcb8 100644
--- a/libbcachefs/alloc_background.c
+++ b/libbcachefs/alloc_background.c
@@ -788,10 +788,12 @@ static int bch2_bucket_do_index(struct btree_trans *trans,
 	if (ca->mi.freespace_initialized &&
 	    test_bit(BCH_FS_CHECK_ALLOC_DONE, &c->flags) &&
 	    bch2_trans_inconsistent_on(old.k->type != old_type, trans,
-			"incorrect key when %s %s btree (got %s should be %s)\n"
+			"incorrect key when %s %s:%llu:%llu:0 (got %s should be %s)\n"
 			"  for %s",
 			set ? "setting" : "clearing",
 			bch2_btree_ids[btree],
+			iter.pos.inode,
+			iter.pos.offset,
 			bch2_bkey_types[old.k->type],
 			bch2_bkey_types[old_type],
 			(bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) {
@@ -1278,8 +1280,8 @@ fsck_err:
 	return ret;
 }
 
-static int bch2_check_discard_freespace_key(struct btree_trans *trans,
-					    struct btree_iter *iter)
+static int __bch2_check_discard_freespace_key(struct btree_trans *trans,
+					      struct btree_iter *iter)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter alloc_iter;
@@ -1321,6 +1323,7 @@ static int bch2_check_discard_freespace_key(struct btree_trans *trans,
 		goto delete;
 out:
 fsck_err:
+	set_btree_iter_dontneed(&alloc_iter);
 	bch2_trans_iter_exit(trans, &alloc_iter);
 	printbuf_exit(&buf);
 	return ret;
@@ -1330,6 +1333,24 @@ delete:
 	goto out;
 }
 
+static int bch2_check_discard_freespace_key(struct btree_trans *trans,
+					    struct btree_iter *iter,
+					    struct bpos end)
+{
+	if (!btree_node_type_is_extents(iter->btree_id)) {
+		return __bch2_check_discard_freespace_key(trans, iter);
+	} else {
+		int ret;
+
+		while (!bkey_eq(iter->pos, end) &&
+		       !(ret = btree_trans_too_many_iters(trans) ?:
+			       __bch2_check_discard_freespace_key(trans, iter)))
+			bch2_btree_iter_set_pos(iter, bpos_nosnap_successor(iter->pos));
+
+		return ret;
+	}
+}
+
 /*
  * We've already checked that generation numbers in the bucket_gens btree are
  * valid for buckets that exist; this just checks for keys for nonexistent
@@ -1485,12 +1506,12 @@ bkey_err:
 			BTREE_ID_need_discard, POS_MIN,
 			BTREE_ITER_PREFETCH, k,
 			NULL, NULL, BTREE_INSERT_NOFAIL|BTREE_INSERT_LAZY_RW,
-		bch2_check_discard_freespace_key(&trans, &iter)) ?:
+		bch2_check_discard_freespace_key(&trans, &iter, k.k->p)) ?:
 	      for_each_btree_key_commit(&trans, iter,
 			BTREE_ID_freespace, POS_MIN,
 			BTREE_ITER_PREFETCH, k,
 			NULL, NULL, BTREE_INSERT_NOFAIL|BTREE_INSERT_LAZY_RW,
-		bch2_check_discard_freespace_key(&trans, &iter)) ?:
+		bch2_check_discard_freespace_key(&trans, &iter, k.k->p)) ?:
 	      for_each_btree_key_commit(&trans, iter,
 			BTREE_ID_bucket_gens, POS_MIN,
 			BTREE_ITER_PREFETCH, k,
diff --git a/libbcachefs/btree_cache.c b/libbcachefs/btree_cache.c
index 661b7667..e5816efe 100644
--- a/libbcachefs/btree_cache.c
+++ b/libbcachefs/btree_cache.c
@@ -793,7 +793,7 @@ static noinline struct btree *bch2_btree_node_fill(struct btree_trans *trans,
 
 	/* Unlock before doing IO: */
 	if (trans && sync)
-		bch2_trans_unlock(trans);
+		bch2_trans_unlock_noassert(trans);
 
 	bch2_btree_node_read(c, b, sync);
 
diff --git a/libbcachefs/btree_iter.c b/libbcachefs/btree_iter.c
index 0e32247f..b78bae43 100644
--- a/libbcachefs/btree_iter.c
+++ b/libbcachefs/btree_iter.c
@@ -2859,6 +2859,7 @@ static noinline void bch2_trans_reset_srcu_lock(struct btree_trans *trans)
 u32 bch2_trans_begin(struct btree_trans *trans)
 {
 	struct btree_path *path;
+	u64 now;
 
 	bch2_trans_reset_updates(trans);
 
@@ -2887,10 +2888,14 @@ u32 bch2_trans_begin(struct btree_trans *trans)
 			path->preserve = false;
 	}
 
+	now = local_clock();
 	if (!trans->restarted &&
 	    (need_resched() ||
-	     local_clock() - trans->last_begin_time > BTREE_TRANS_MAX_LOCK_HOLD_TIME_NS))
+	     now - trans->last_begin_time > BTREE_TRANS_MAX_LOCK_HOLD_TIME_NS)) {
 		drop_locks_do(trans, (cond_resched(), 0));
+		now = local_clock();
+	}
+	trans->last_begin_time = now;
 
 	if (unlikely(time_after(jiffies, trans->srcu_lock_time + msecs_to_jiffies(10))))
 		bch2_trans_reset_srcu_lock(trans);
@@ -2901,7 +2906,6 @@ u32 bch2_trans_begin(struct btree_trans *trans)
 		trans->notrace_relock_fail = false;
 	}
 
-	trans->last_begin_time = local_clock();
 	return trans->restart_count;
 }
 
diff --git a/libbcachefs/btree_locking.c b/libbcachefs/btree_locking.c
index 5b290324..7d9e01c1 100644
--- a/libbcachefs/btree_locking.c
+++ b/libbcachefs/btree_locking.c
@@ -714,6 +714,14 @@ int bch2_trans_relock_notrace(struct btree_trans *trans)
 	return 0;
 }
 
+void bch2_trans_unlock_noassert(struct btree_trans *trans)
+{
+	struct btree_path *path;
+
+	trans_for_each_path(trans, path)
+		__bch2_btree_path_unlock(trans, path);
+}
+
 void bch2_trans_unlock(struct btree_trans *trans)
 {
 	struct btree_path *path;
diff --git a/libbcachefs/btree_locking.h b/libbcachefs/btree_locking.h
index b341cc89..0ad8fd44 100644
--- a/libbcachefs/btree_locking.h
+++ b/libbcachefs/btree_locking.h
@@ -22,6 +22,8 @@ void bch2_assert_btree_nodes_not_locked(void);
 static inline void bch2_assert_btree_nodes_not_locked(void) {}
 #endif
 
+void bch2_trans_unlock_noassert(struct btree_trans *);
+
 static inline bool is_btree_node(struct btree_path *path, unsigned l)
 {
 	return l < BTREE_MAX_DEPTH && !IS_ERR_OR_NULL(path->l[l].b);
diff --git a/libbcachefs/btree_update_interior.c b/libbcachefs/btree_update_interior.c
index d2811c47..e95e4885 100644
--- a/libbcachefs/btree_update_interior.c
+++ b/libbcachefs/btree_update_interior.c
@@ -1077,7 +1077,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
 					BKEY_BTREE_PTR_U64s_MAX * (1 + split)))
 			break;
 
-		split = true;
+		split = path->l[update_level].b->nr.live_u64s > BTREE_SPLIT_THRESHOLD(c);
 	}
 
 	if (flags & BTREE_INSERT_GC_LOCK_HELD)
diff --git a/libbcachefs/data_update.c b/libbcachefs/data_update.c
index c89ee14f..9f7a30c7 100644
--- a/libbcachefs/data_update.c
+++ b/libbcachefs/data_update.c
@@ -474,7 +474,7 @@ int bch2_data_update_init(struct btree_trans *trans,
 			if (crc_is_compressed(p.crc))
 				reserve_sectors += k.k->size;
 
-			m->op.nr_replicas += bch2_extent_ptr_durability(c, &p);
+			m->op.nr_replicas += bch2_extent_ptr_desired_durability(c, &p);
 		} else if (!p.ptr.cached) {
 			bch2_dev_list_add_dev(&m->op.devs_have, p.ptr.dev);
 		}
diff --git a/libbcachefs/extents.c b/libbcachefs/extents.c
index e2b126ad..7e005509 100644
--- a/libbcachefs/extents.c
+++ b/libbcachefs/extents.c
@@ -641,9 +641,8 @@ unsigned bch2_bkey_replicas(struct bch_fs *c, struct bkey_s_c k)
 	return replicas;
 }
 
-unsigned bch2_extent_ptr_durability(struct bch_fs *c, struct extent_ptr_decoded *p)
+unsigned bch2_extent_ptr_desired_durability(struct bch_fs *c, struct extent_ptr_decoded *p)
 {
-	unsigned durability = 0;
 	struct bch_dev *ca;
 
 	if (p->ptr.cached)
@@ -651,13 +650,28 @@ unsigned bch2_extent_ptr_durability(struct bch_fs *c, struct extent_ptr_decoded
 
 	ca = bch_dev_bkey_exists(c, p->ptr.dev);
 
-	if (ca->mi.state != BCH_MEMBER_STATE_failed)
-		durability = max_t(unsigned, durability, ca->mi.durability);
+	return ca->mi.durability +
+		(p->has_ec
+		 ? p->ec.redundancy
+		 : 0);
+}
 
-	if (p->has_ec)
-		durability += p->ec.redundancy;
+unsigned bch2_extent_ptr_durability(struct bch_fs *c, struct extent_ptr_decoded *p)
+{
+	struct bch_dev *ca;
 
-	return durability;
+	if (p->ptr.cached)
+		return 0;
+
+	ca = bch_dev_bkey_exists(c, p->ptr.dev);
+
+	if (ca->mi.state == BCH_MEMBER_STATE_failed)
+		return 0;
+
+	return ca->mi.durability +
+		(p->has_ec
+		 ? p->ec.redundancy
+		 : 0);
 }
 
 unsigned bch2_bkey_durability(struct bch_fs *c, struct bkey_s_c k)
diff --git a/libbcachefs/extents.h b/libbcachefs/extents.h
index 31c81409..3ba41e37 100644
--- a/libbcachefs/extents.h
+++ b/libbcachefs/extents.h
@@ -610,6 +610,7 @@ bool bch2_bkey_is_incompressible(struct bkey_s_c);
 unsigned bch2_bkey_sectors_compressed(struct bkey_s_c);
 
 unsigned bch2_bkey_replicas(struct bch_fs *, struct bkey_s_c);
+unsigned bch2_extent_ptr_desired_durability(struct bch_fs *, struct extent_ptr_decoded *);
 unsigned bch2_extent_ptr_durability(struct bch_fs *, struct extent_ptr_decoded *);
 unsigned bch2_bkey_durability(struct bch_fs *, struct bkey_s_c);
 
diff --git a/linux/six.c b/linux/six.c
index 5b81c3fc..0b9c4bb7 100644
--- a/linux/six.c
+++ b/linux/six.c
@@ -47,26 +47,26 @@ struct six_lock_vals {
 	enum six_lock_type	unlock_wakeup;
 };
 
-#define LOCK_VALS {							\
-	[SIX_LOCK_read] = {						\
-		.lock_val	= 1U << SIX_LOCK_HELD_read_OFFSET,	\
-		.lock_fail	= SIX_LOCK_HELD_write,			\
-		.held_mask	= SIX_LOCK_HELD_read,			\
-		.unlock_wakeup	= SIX_LOCK_write,			\
-	},								\
-	[SIX_LOCK_intent] = {						\
-		.lock_val	= SIX_LOCK_HELD_intent,			\
-		.lock_fail	= SIX_LOCK_HELD_intent,			\
-		.held_mask	= SIX_LOCK_HELD_intent,			\
-		.unlock_wakeup	= SIX_LOCK_intent,			\
-	},								\
-	[SIX_LOCK_write] = {						\
-		.lock_val	= SIX_LOCK_HELD_write,			\
-		.lock_fail	= SIX_LOCK_HELD_read,			\
-		.held_mask	= SIX_LOCK_HELD_write,			\
-		.unlock_wakeup	= SIX_LOCK_read,			\
-	},								\
-}
+static const struct six_lock_vals l[] = {
+	[SIX_LOCK_read] = {
+		.lock_val	= 1U << SIX_LOCK_HELD_read_OFFSET,
+		.lock_fail	= SIX_LOCK_HELD_write,
+		.held_mask	= SIX_LOCK_HELD_read,
+		.unlock_wakeup	= SIX_LOCK_write,
+	},
+	[SIX_LOCK_intent] = {
+		.lock_val	= SIX_LOCK_HELD_intent,
+		.lock_fail	= SIX_LOCK_HELD_intent,
+		.held_mask	= SIX_LOCK_HELD_intent,
+		.unlock_wakeup	= SIX_LOCK_intent,
+	},
+	[SIX_LOCK_write] = {
+		.lock_val	= SIX_LOCK_HELD_write,
+		.lock_fail	= SIX_LOCK_HELD_read,
+		.held_mask	= SIX_LOCK_HELD_write,
+		.unlock_wakeup	= SIX_LOCK_read,
+	},
+};
 
 static inline void six_set_bitmask(struct six_lock *lock, u32 mask)
 {
@@ -116,7 +116,6 @@ static inline unsigned pcpu_read_count(struct six_lock *lock)
 static int __do_six_trylock(struct six_lock *lock, enum six_lock_type type,
 			    struct task_struct *task, bool try)
 {
-	const struct six_lock_vals l[] = LOCK_VALS;
 	int ret;
 	u32 old;
 
@@ -301,10 +300,10 @@ EXPORT_SYMBOL_GPL(six_trylock_ip);
 bool six_relock_ip(struct six_lock *lock, enum six_lock_type type,
 		   unsigned seq, unsigned long ip)
 {
-	if (lock->seq != seq || !six_trylock_ip(lock, type, ip))
+	if (six_lock_seq(lock) != seq || !six_trylock_ip(lock, type, ip))
 		return false;
 
-	if (lock->seq != seq) {
+	if (six_lock_seq(lock) != seq) {
 		six_unlock_ip(lock, type, ip);
 		return false;
 	}
@@ -588,7 +587,6 @@ EXPORT_SYMBOL_GPL(six_lock_ip_waiter);
 __always_inline
 static void do_six_unlock_type(struct six_lock *lock, enum six_lock_type type)
 {
-	const struct six_lock_vals l[] = LOCK_VALS;
 	u32 state;
 
 	if (type == SIX_LOCK_intent)
@@ -638,6 +636,8 @@ void six_unlock_ip(struct six_lock *lock, enum six_lock_type type, unsigned long
 
 	if (type != SIX_LOCK_write)
 		six_release(&lock->dep_map, ip);
+	else
+		lock->seq++;
 
 	if (type == SIX_LOCK_intent &&
 	    lock->intent_lock_recurse) {
@@ -645,8 +645,6 @@ void six_unlock_ip(struct six_lock *lock, enum six_lock_type type, unsigned long
 		return;
 	}
 
-	lock->seq += type == SIX_LOCK_write;
-
 	do_six_unlock_type(lock, type);
 }
 EXPORT_SYMBOL_GPL(six_unlock_ip);
@@ -675,7 +673,6 @@ EXPORT_SYMBOL_GPL(six_lock_downgrade);
  */
 bool six_lock_tryupgrade(struct six_lock *lock)
 {
-	const struct six_lock_vals l[] = LOCK_VALS;
 	u32 old = atomic_read(&lock->state), new;
 
 	do {
@@ -743,8 +740,6 @@ EXPORT_SYMBOL_GPL(six_trylock_convert);
  */
 void six_lock_increment(struct six_lock *lock, enum six_lock_type type)
 {
-	const struct six_lock_vals l[] = LOCK_VALS;
-
 	six_acquire(&lock->dep_map, 0, type == SIX_LOCK_read, _RET_IP_);
 
 	/* XXX: assert already locked, and that we don't overflow: */