btrfs: scrub: refine the error messages
Get series with b4
b4 am [email protected]Versions newest first
- v1 Latest0 of 6 patches 2 repliesIn review
Patches in v1 No applies recorded yet
None of the 6 patches the cover letter announces are in this archive.
Reviews trailers per patch
| Patch | Anand Jain | Filipe Manana |
|---|---|---|
| Series | R-b | R-b R-b R-b R-b |
Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.
Cover letter [PATCH 0/6] btrfs: scrub: refine the error messages
During some support sessions, I found older kernels are always report very unuseful scrub error messages like:
BTRFS error (device dm-0): bdev /dev/mapper/sys-rootlv errs: wr 0, rd 0, flush 0, corrupt 2823, gen 0 BTRFS error (device dm-0): unable to fixup (regular) error at logical 1563504640 on dev /dev/mapper/sys-rootlv BTRFS error (device dm-0): bdev /dev/mapper/sys-rootlv errs: wr 0, rd 0, flush 0, corrupt 2824, gen 0 BTRFS error (device dm-0): unable to fixup (regular) error at logical 1579016192 on dev /dev/mapper/sys-rootlv
There are two problems:
- No proper details about the corruption
No metadata or data indication, nor the filename or the tree id. Even the involved kernel (and newer kernels after the scrub rework) has the ability to do backref walk and find out the file path or the tree backref.
My guess is, for data sometimes the corrupted sector is no longer referred by any data extent.
- Too noisy and useless error message from btrfs_dev_stat_inc_and_print()
I'd argue we should not output any error message just for btrfs_dev_stat_inc_and_print().
After the series, the error message would look like this:
BTRFS warning (device dm-2): checksum error at inode 5/257(file1) fileoff 16384, logical 13647872(1) physical 1(/dev/mapper/test-scratch1)13647872
This involves the following enhancement:
- Single line
And we ensure we output at least one line for the error we hit. No more unrelated btrfs_dev_stat_inc_and_print() output.
- Proper fall backs
We have the following different fall back options
* Repaired
Just a line of something repaired for logical/physical address.
* Detailed data info
Including the following elements (if possible), and if higher priority ones can not be fetched, it would be skipped and try lower priority items:
+ file path (can have multiple ones)
+ root/inode number and offset
+ logical/physical address (always output)
* Detaile metadata info
The priority list is:
+ root ref/level
+ logical/physical address (always output)
For the worst case of data corruption, we would still have some like:
BTRFS warning (device dm-2): checksum error at data, logical 13647872(1) physical 1(/dev/mapper/test-scratch1)13647872
And similar ones for metadata:
BTRFS warning (device dm-2): checksum error at meta, logical 13647872(1) physical 1(/dev/mapper/test-scratch1)13647872
The first patch is fixing a regression in the error message, which leads to bad logical/physical bytenr.
The second one would reduce the log level for btrfs_dev_stat_inc_and_print().
Path 3~4 are cleanups to remove unnecessary parameters.
The remaining reworks the format and refine the error message frequency.
Qu Wenruo (7):
btrfs: scrub: fix incorrectly reported logical/physical address
btrfs: reduce the log level for btrfs_dev_stat_inc_and_print()
btrfs: scrub: remove unused is_super parameter from
scrub_print_common_warning()
btrfs: scrub: remove unnecessary dev/physical lookup for
scrub_stripe_report_errors()
btrfs: scrub: simplify the inode iteration output
btrfs: scrub: unify and shorten the error message
btrfs: scrub: fix the frequency of error messages
fs/btrfs/scrub.c | 185 ++++++++++++++++----------------------------- fs/btrfs/volumes.c | 2 +- 2 files changed, 66 insertions(+), 121 deletions(-)
Signature · 1 line
2.44.0