Hi all,

A couple more small fixes. The first patch fixes a byte order problem in
the boundary code between bcachefs and the crc32c library related to
extent (and thus checksum) merges. The second patch addresses a small
logic wart that leads to more split writeback behavior than necessary,
which is how the problem addressed by patch 1 was observed. Patch 2 was
more of a "I'm not clear on why this size check is here, so let's try to
remove it and see what happens" patch. ;)

>From chatting a bit in the meeting yesterday, I think we've already
established this isn't exactly what we want to do here. Instead, we
should still have some sort of bound on the size of a writeback op.
This is discussed a bit in the commit log description, but the consensus
seemed to be that this should just be replaced with a moderately less
conservative size check, such as 16MB or so. That's notably better than
the current limit of 1MB, so unless discussion on this patch leads in a
different direction, I'll probably follow up with a v2 to do that once
I've had a bit more chance to test it. Therefore, patch 2 is included
here mainly for posterity.

Thoughts, reviews, flames appreciated.

Brian

Brian Foster (2):
  bcachefs: fix crc32c checksum merge byte order problem
  bcachefs: remove writeback bio size limit

 fs/bcachefs/checksum.c       | 4 ++--
 fs/bcachefs/fs-io-buffered.c | 2 --
 2 files changed, 2 insertions(+), 4 deletions(-)

-- 
2.41.0

Reply via email to