Some more misc fsmap fixes

[PATCH 0/2] Some more misc fsmap fixes

Posted by Ojaswin Mujoo 5 months ago

These 2 patches mostly fixup a couple edge case inconsistent behavior
we were seeing in fsmap. Further, convert fsmap path to use block size
units wherever possible to avoid issues due to block to cluster to block
conversions like [1].

[1] https://lore.kernel.org/linux-ext4/e7472c8535c9c5ec10f425f495366864ea12c9da.1754377641.git.ojaswin@linux.ibm.com/

Tested with some custom edge cases (which I'll see how to consistently
convert to xfstests) as well as generic/365 on 4k, 64k and bigalloc

Ojaswin Mujoo (2):
  ext4: Correctly handle queries for metadata mappings
  fsmap: use blocksize units instead of cluster units

 fs/ext4/fsmap.c   | 56 +++++++++++++++++++++++++++--------------------
 fs/ext4/mballoc.c | 25 ++++++++++++++-------
 2 files changed, 49 insertions(+), 32 deletions(-)

-- 
2.49.0

Re: [PATCH 0/2] Some more misc fsmap fixes

Posted by Theodore Ts'o 4 months, 1 week ago

On Fri, 05 Sep 2025 13:44:45 +0530, Ojaswin Mujoo wrote:
> These 2 patches mostly fixup a couple edge case inconsistent behavior
> we were seeing in fsmap. Further, convert fsmap path to use block size
> units wherever possible to avoid issues due to block to cluster to block
> conversions like [1].
> 
> [1] https://lore.kernel.org/linux-ext4/e7472c8535c9c5ec10f425f495366864ea12c9da.1754377641.git.ojaswin@linux.ibm.com/
> 
> [...]

Applied, thanks!

[1/2] ext4: Correctly handle queries for metadata mappings
      commit: 46c22a8bb4cb03211da1100d7ee4a2005bf77c70

([2/2] fsmap: use blocksize units instead of cluster units was dropped
 because it introduced a regression.)

Best regards,
-- 
Theodore Ts'o <tytso@mit.edu>