Re: [PATCH 6.1 000/481] 6.1.167-rc1 review

["Theodore Tso" <tytso@mit.edu>]

On Tue, Mar 24, 2026 at 03:36:21PM +0000, Mark Brown wrote:
> I have a bisect for an ext4 issue in v6.1 which comes out at:
> 
> # first bad commit: [29897d75d6491ffe23cdc9d96caba9282a20dfc3] ext4: convert bd_bitmap_page to bd_bitmap_folio
> 
> For an oops which looks very similar (but on arm64):

I can confirm this bisection; I was testing on x86_64, and using
"kvm-xfstests -c ext4/4k generic/001" on a failure, it would crash
before running the first test (in my test runner infrastructure when
running syncfs on the results directory --- go figure).

Unfortunately, you can't just revert this commit because of merge
conflicts.  In order to get a clean revert, you have to revert (or
drop) three commits:

% git log -3
commit b12a69d9770b58fb02d3b4f72abe5acd28aa7e76 (HEAD)
Author: Theodore Ts'o <tytso@mit.edu>
Date:   Tue Mar 24 23:46:15 2026 -0400

    Revert "ext4: convert bd_bitmap_page to bd_bitmap_folio"

    This reverts commit 29897d75d6491ffe23cdc9d96caba9282a20dfc3.

commit 9c95c376c79f47fe9ee8ce562249d3630a50ab12
Author: Theodore Ts'o <tytso@mit.edu>
Date:   Tue Mar 24 23:44:23 2026 -0400

    Revert "ext4: convert bd_buddy_page to bd_buddy_folio"

    This reverts commit fe80bba8f76f9f0995cdc64fc89b65173e1ae828.

commit 98f5de80114f6194af4d9fae572b73440efa67c2
Author: Theodore Ts'o <tytso@mit.edu>
Date:   Tue Mar 24 23:43:40 2026 -0400

    Revert "ext4: fix e4b bitmap inconsistency reports"

    This reverts commit cb45b6209aa53979b054bd026d938107d5a3031b.

I haven't had time to investigate this more closely, but I'm assuming
the automated stable picker was trying to backport cb45b6209aa5
("ext4: fix e4b bitmap inconsistency reports"), and determined that
the fe80bba8f76f ("ext4: convert bd_buddy_page to bd_buddy_folio") and
29897d75d649 ("ext4: convert bd_bitmap_page to bd_bitmap_folio") were
prerequisite commits --- and while 29897d75d649 cherry picked
correctly, either the git scrwed up the cherry pick, or there was some
additional prerequisite commit needed, but wasn't caught by the "it
patches cleanly, ship it!" algorithm.

I don't have time to investigate further, but Greg, if you could drop
these three patches, that should address this issue.

Cheers,

					- Ted