btrfs: btrfs_lookup_bio_sums() related fixes
Get series with b4
b4 am [email protected]Versions newest first · linked by author and title
- v2 Latest3 patches 9 repliesIn review
- v13 patches 8 repliesSuperseded
Changes in v2 from the cover letter
- Remove the @file_offset parameter completely
- Remove btrfs_find_ordered_sum() completely
- Introduce data reloc inode specific file_offset lookup.
Patches in v2 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/3 | btrfs: file-item: use nodesize to determine whether we need readhead for btrfs_lookup_bio_sums()fs/ | +5 | 2 replies | |
| 2/3 | btrfs: file-item: remove the btrfs_find_ordered_sum() call in btrfs_lookup_bio_sums()fs/ fs/ fs/ | +11 | 3 replies | |
| 3/3 | btrfs: file-item: refactor btrfs_lookup_bio_sums() to handle out-of-order bvecsfs/ fs/ fs/ +1 more | +171 | 4 replies |
Reviews trailers per patch
No Reviewed-by, Acked-by or Tested-by trailers for v2 yet.
Cover letter [PATCH v2 0/3] btrfs: btrfs_lookup_bio_sums() related fixes
This is the final piece to make subpage read/write patchset to pass all fsstress tests.
Before this 3 patches, there is a very low chance to hit the following errors:
- False bad data csum error
The expected csum is alwasy 0x00000000, which should represents something.
- WARN_ON_ONCE(count) hit
Something is definitely wrong in btrfs_lookup_bio_sums().
After more debugging, even on X86_64, there is definitely something to notice, the most important thing is, the bvec in btrfs_lookup_bio_sums() is not always in linear bytenr order:
btrfs_lookup_bio_sums: file_offset=-1 expected_bvec_offset=532480 page_offset=716800 bv_offset=0
btrfs_lookup_bio_sums: orig_file_offset=520192 bvec_index=3 root=259 ino=260 page_owner_ino=260
This is from x86_64 run, with extra debug info from the bio_for_each_segment() loop.
It turns out that, bio can be merged as long as their disk_bytenr can be merged, no need for their page offset to be adjacent.
This means, the file_offset in btrfs_lookup_bio_sums() doesn't make much sense.
To address the problem, the 3rd patch is introduced to do bio unrelated csum search.
This does not only simplify the main loop (just one small main loop), but also free us from the out-of-order bvec problems.
The other two patches are mostly small enhancement found during the development.
With the patchset, the subpage can survive infinite fsstress run and the regular page size case can still pass all existing fstests.
Since it has nothing special to subpage at all, and by nature they are mostly renames and refactor, they can be submitted right now, with or without subpage patches.
Changelog:
v2:
- Remove the @file_offset parameter completely
- Remove btrfs_find_ordered_sum() completely
- Introduce data reloc inode specific file_offset lookup.
Qu Wenruo (3):
btrfs: file-item: use nodesize to determine whether we need readhead
for btrfs_lookup_bio_sums()
btrfs: file-item: remove the btrfs_find_ordered_sum() call in
btrfs_lookup_bio_sums()
btrfs: file-item: refactor btrfs_lookup_bio_sums() to handle
out-of-order bvecs
fs/btrfs/compression.c | 4 +- fs/btrfs/ctree.h | 2 +- fs/btrfs/file-item.c | 274 ++++++++++++++++++++++++++-------------- fs/btrfs/inode.c | 5 +- fs/btrfs/ordered-data.c | 46 ------- fs/btrfs/ordered-data.h | 2 - 6 files changed, 185 insertions(+), 148 deletions(-)
Signature · 1 line
2.29.1