Btrfs: add a recovery_errors mount option to limp along with a corrupted FSrecovery_errors

When btrfs isn't able to read a tree root, it generally aborts the
mount.  This patch adds mount -o recovery_errors to force it to
try and continue along.

Right now it only deals with corrupt csum roots, but it will also
gain support for extent allocation tree roots as well.  In this mode,
the FS is forced readonly and we ignore all csums.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

6 files changed, 64 insertions, 11 deletions

diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index 14f1c5a0b2d29..c2b0c987c8c31 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -113,7 +113,8 @@ static int check_compressed_csum(struct inode *inode,
 	u32 csum;
 	u32 *cb_sum = &cb->sums;
 
-	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
+	if ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM) ||
+	    btrfs_test_opt(root, NODATASUM))
 		return 0;
 
 	for (i = 0; i < cb->nr_pages; i++) {
@@ -577,8 +578,13 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 	u64 em_start;
 	struct extent_map *em;
 	int ret = -ENOMEM;
+	int skip_sum = 0;
 	u32 *sums;
 
+	if ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM) ||
+	    btrfs_test_opt(root, NODATASUM))
+		skip_sum = 1;
+
 	tree = &BTRFS_I(inode)->io_tree;
 	em_tree = &BTRFS_I(inode)->extent_tree;
 
@@ -670,7 +676,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 			 */
 			atomic_inc(&cb->pending_bios);
 
-			if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
+			if (!skip_sum) {
 				ret = btrfs_lookup_bio_sums(root, inode,
 							comp_bio, sums);
 				BUG_ON(ret);
@@ -698,7 +704,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 	ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0);
 	BUG_ON(ret);
 
-	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
+	if (!skip_sum) {
 		ret = btrfs_lookup_bio_sums(root, inode, comp_bio, sums);
 		BUG_ON(ret);
 	}
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index b9ba59ff92925..660364d8da1bb 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1410,6 +1410,7 @@ struct btrfs_ioctl_defrag_range_args {
 #define BTRFS_MOUNT_AUTO_DEFRAG		(1 << 16)
 #define BTRFS_MOUNT_INODE_MAP_CACHE	(1 << 17)
 #define BTRFS_MOUNT_RECOVERY		(1 << 18)
+#define BTRFS_MOUNT_RECOVERY_ERR	(1 << 19)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index b6a5c0dd0dd8e..5b9d3808fb30d 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1773,11 +1773,14 @@ static void backup_super_roots(struct btrfs_fs_info *info)
 	btrfs_set_backup_dev_root_level(root_backup,
 				       btrfs_header_level(info->dev_root->node));
 
-	btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
-	btrfs_set_backup_csum_root_gen(root_backup,
+	if (info->csum_root->node) {
+		btrfs_set_backup_csum_root(root_backup,
+					   info->csum_root->node->start);
+		btrfs_set_backup_csum_root_gen(root_backup,
 			       btrfs_header_generation(info->csum_root->node));
-	btrfs_set_backup_csum_root_level(root_backup,
+		btrfs_set_backup_csum_root_level(root_backup,
 			       btrfs_header_level(info->csum_root->node));
+	}
 
 	btrfs_set_backup_total_bytes(root_backup,
 			     btrfs_super_total_bytes(info->super_copy));
@@ -2306,8 +2309,25 @@ retry_root_backup:
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_CSUM_TREE_OBJECTID, csum_root);
-	if (ret)
-		goto recovery_tree_root;
+	if (ret) {
+		if (!btrfs_test_opt(tree_root, RECOVERY_ERR))
+			goto recovery_tree_root;
+
+		/* don't use the log in recovery mode, it won't be valid */
+		btrfs_set_super_log_root(disk_super, 0);
+
+		/* we can't trust the free space cache either */
+		btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
+
+		/* we must set ourselves readonly */
+		sb->s_flags |= MS_RDONLY;
+
+		/* the crc tree is dead, don't use it */
+		btrfs_set_opt(fs_info->mount_opt, NODATASUM);
+
+		printk("btrfs failed to read the csum root, forcing readonly\n");
+		ret = 0;
+	}
 
 	csum_root->track_dirty = 1;
 
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index c7fb3a4247d3a..1ae703b826eb1 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -167,6 +167,9 @@ static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
 	struct btrfs_csum_item *item = NULL;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 
+	if (!root->fs_info->csum_root->node)
+		return -EIO;
+
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
@@ -550,6 +553,8 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans,
 	int blocksize_bits = root->fs_info->sb->s_blocksize_bits;
 
 	root = root->fs_info->csum_root;
+	if (!root->node)
+		return -EIO;
 
 	path = btrfs_alloc_path();
 	if (!path)
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index e16215f480d00..c182bf4ad42cd 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1479,6 +1479,10 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 	int ret = 0;
 	int skip_sum;
 
+	/*
+	 * even if we're mounted nodatasum, once the inode has sums we have
+	 * to keep summing it for writes.
+	 */
 	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
 
 	if (btrfs_is_free_space_inode(root, inode))
@@ -1488,10 +1492,11 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 	BUG_ON(ret);
 
 	if (!(rw & REQ_WRITE)) {
+		/* for reads, we can skip the crc if we're mounted NODATASUM */
 		if (bio_flags & EXTENT_BIO_COMPRESSED) {
 			return btrfs_submit_compressed_read(inode, bio,
 						    mirror_num, bio_flags);
-		} else if (!skip_sum) {
+		} else if (!skip_sum && !btrfs_test_opt(root, NODATASUM)) {
 			ret = btrfs_lookup_bio_sums(root, inode, bio, NULL);
 			if (ret)
 				return ret;
@@ -1849,6 +1854,9 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
 		goto good;
 
+	if (btrfs_test_opt(root, NODATASUM))
+		goto good;
+
 	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
 	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
 		clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
@@ -5580,10 +5588,15 @@ static void btrfs_endio_direct_read(struct bio *bio, int err)
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	u64 start;
 	u32 *private = dip->csums;
+	int check_sums = 1;
+
+	if ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM) ||
+	    btrfs_test_opt(root, NODATASUM))
+		check_sums = 0;
 
 	start = dip->logical_offset;
 	do {
-		if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
+		if (check_sums) {
 			struct page *page = bvec->bv_page;
 			char *kaddr;
 			u32 csum = ~(u32)0;
@@ -5949,6 +5962,8 @@ static void btrfs_submit_direct(int rw, struct bio *bio, struct inode *inode,
 	int ret = 0;
 
 	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
+	if (!write)
+		skip_sum |= btrfs_test_opt(root, NODATASUM);
 
 	dip = kmalloc(sizeof(*dip), GFP_NOFS);
 	if (!dip) {
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 8bd9d6d0e07ae..a95ebb9427176 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -164,7 +164,8 @@ enum {
 	Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 	Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 	Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
-	Opt_inode_cache, Opt_no_space_cache, Opt_recovery, Opt_err,
+	Opt_inode_cache, Opt_no_space_cache, Opt_recovery, Opt_recovery_errors,
+	Opt_err,
 };
 
 static match_table_t tokens = {
@@ -199,6 +200,7 @@ static match_table_t tokens = {
 	{Opt_inode_cache, "inode_cache"},
 	{Opt_no_space_cache, "nospace_cache"},
 	{Opt_recovery, "recovery"},
+	{Opt_recovery_errors, "recovery_errors"},
 	{Opt_err, NULL},
 };
 
@@ -397,6 +399,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			printk(KERN_INFO "btrfs: enabling auto recovery");
 			btrfs_set_opt(info->mount_opt, RECOVERY);
 			break;
+		case Opt_recovery_errors:
+			printk(KERN_INFO "btrfs: enabling recovery failure mode");
+			btrfs_set_opt(info->mount_opt, RECOVERY_ERR);
+			break;
 		case Opt_err:
 			printk(KERN_INFO "btrfs: unrecognized mount option "
 			       "'%s'\n", p);