| Message ID | 20191111075059.30352-2-wqu@suse.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/2] btrfs-progs: Reduce error level from error to warning for OPEN_CTREE_PARTIAL | expand |
On Mon, Nov 11, 2019 at 03:50:59PM +0800, Qu Wenruo wrote: > [BUG] > Even "btrfs rescue zero-log" only reset btrfs_super_block::log_root and > btrfs_super_block::log_root_level, we still use trasction to write all > super blocks for all devices. > > This means we can't handle things like corrupted extent tree: > > checksum verify failed on 2172747776 found 000000B6 wanted 00000000 > checksum verify failed on 2172747776 found 000000B6 wanted 00000000 > bad tree block 2172747776, bytenr mismatch, want=2172747776, have=0 > WARNING: could not setup extent tree, skipping it > Clearing log on /dev/nvme/btrfs, previous log_root 0, level 0 > ERROR: Corrupted fs, no valid METADATA block group found > ERROR: attempt to start transaction over already running one > > [CAUSE] > Because we have extra check in transaction code to ensure we have valid > METADATA block groups. > > In fact we don't really need transaction at all. > > [FIX] > Instead of commit transaction, we can just call write_all_supers() > manually, so we can still handle multi-device fs while avoid above > error. > > Also, add OPEN_CTREE_NO_BLOCK_GROUPS open ctree flag to make it more > robust. > > Reported-by: Christian Pernegger <pernegger@gmail.com> > Signed-off-by: Qu Wenruo <wqu@suse.com> Thanks, v1 has been replaced.
diff --git a/cmds/rescue.c b/cmds/rescue.c index e8eab6808bc3..087c33befeff 100644 --- a/cmds/rescue.c +++ b/cmds/rescue.c @@ -165,7 +165,6 @@ static int cmd_rescue_zero_log(const struct cmd_struct *cmd, int argc, char **argv) { struct btrfs_root *root; - struct btrfs_trans_handle *trans; struct btrfs_super_block *sb; char *devname; int ret; @@ -187,7 +186,8 @@ static int cmd_rescue_zero_log(const struct cmd_struct *cmd, goto out; } - root = open_ctree(devname, 0, OPEN_CTREE_WRITES | OPEN_CTREE_PARTIAL); + root = open_ctree(devname, 0, OPEN_CTREE_WRITES | OPEN_CTREE_PARTIAL | + OPEN_CTREE_NO_BLOCK_GROUPS); if (!root) { error("could not open ctree"); return 1; @@ -198,13 +198,14 @@ static int cmd_rescue_zero_log(const struct cmd_struct *cmd, devname, (unsigned long long)btrfs_super_log_root(sb), (unsigned)btrfs_super_log_root_level(sb)); - trans = btrfs_start_transaction(root, 1); - BUG_ON(IS_ERR(trans)); btrfs_set_super_log_root(sb, 0); btrfs_set_super_log_root_level(sb, 0); - btrfs_commit_transaction(trans, root); + ret = write_all_supers(root->fs_info); + if (ret < 0) { + errno = -ret; + error("failed to write dev supers: %m"); + } close_ctree(root); - out: return !!ret; }
[BUG] Even "btrfs rescue zero-log" only reset btrfs_super_block::log_root and btrfs_super_block::log_root_level, we still use trasction to write all super blocks for all devices. This means we can't handle things like corrupted extent tree: checksum verify failed on 2172747776 found 000000B6 wanted 00000000 checksum verify failed on 2172747776 found 000000B6 wanted 00000000 bad tree block 2172747776, bytenr mismatch, want=2172747776, have=0 WARNING: could not setup extent tree, skipping it Clearing log on /dev/nvme/btrfs, previous log_root 0, level 0 ERROR: Corrupted fs, no valid METADATA block group found ERROR: attempt to start transaction over already running one [CAUSE] Because we have extra check in transaction code to ensure we have valid METADATA block groups. In fact we don't really need transaction at all. [FIX] Instead of commit transaction, we can just call write_all_supers() manually, so we can still handle multi-device fs while avoid above error. Also, add OPEN_CTREE_NO_BLOCK_GROUPS open ctree flag to make it more robust. Reported-by: Christian Pernegger <pernegger@gmail.com> Signed-off-by: Qu Wenruo <wqu@suse.com> --- cmds/rescue.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-)