Re: [PATCH 6.12 000/385] 6.12.75-rc1 review

[Sasha Levin]

On Sun, Mar 01, 2026 at 10:05:02PM -0800, Barry K. Nathan wrote:
> On 2/28/26 10:00, Sasha Levin wrote:
> > This is the start of the stable review cycle for the 6.12.75 release.
> > There are 385 patches in this series, all will be posted as a response
> > to this one.  If anyone has any issues with these being applied, please
> > let me know.
> >
> > Responses should be made by Mon Mar  2 05:59:55 PM UTC 2026.
> > Anything received after that time might be too late.
> >
> > The whole patch series can be found in one patch at:
> >         https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git/patch/?id=linux-6.12.y&id2=v6.12.74
> > or in the git tree and branch at:
> >         git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git linux-6.12.y
> > and the diffstat can be found below.
> >
> > Thanks,
> > Sasha
>
> I just now noticed a sizable discrepancy between what's in the
> stable-queue and what's in -rc1, for 5.10.y through 6.12.y. (6.18.y
> and 6.19.y appear unaffected.)
>
> To make sure this is an apples-to-apples comparison, I'll compare with
> the stable-queue as of commit 2370009958172f632d48973387e7b6ae116086b1
> ("Drop a broken ACPI patch"); I'd expect the queue as of that commit to
> match the -rc1 patches, if I'm not mistaken.
>
>
>                      # of patches in         # of patches in
>                      stable mailing list     stable-queue git
>                      thread                  @ 237000995817
>
> 5.10.252-rc1          147                     334
> 5.15.202-rc1          164                     411
> 6.1.165-rc1           232                     533
> 6.6.128-rc1           283                     683
> 6.12.75-rc1           385                     953
> 6.18.16-rc1           752                     751
> 6.19.6-rc1            844                     843
>
> The off-by-one difference for 6.18.y/6.19.y is expected, since
> (unlike the stable-queue itself) the -rc1 patch and the mailing
> list thread include a Makefile patch to update the version number.
>
> For the other kernels, though, it looks to me like something
> went wrong somewhere. Of course I could be mistaken, but that's
> how it appears to me.
>
> In any case, I figured I should bring this to your attention.

Barry, this is a great catch. Thank you!

The root cause turned out to be a bug in git-quiltimport. One of the
patches queued has the literal text "\0" in its subject line:

  selftests: tc_actions: don't dump 2MB of \0 to stdout

git-quiltimport constructs commit messages using echo(1):

  commit=$( { echo "$SUBJECT"; echo; cat "$tmp_msg"; } | git commit-tree $tree -p $commit)

The problem is that echo interprets backslash escape sequences, so
"\0" gets expanded into an actual NUL byte (0x00). git commit-tree
then rejects the commit with:

  error: a NUL byte in commit log message not allowed.

This caused git-quiltimport to bail out mid-way through building
several trees during -rc construction. The trees that had this patch
queued (5.10 through 6.12) only got a partial set of patches into
the -rc branch, while 6.18 and 6.19 were unaffected because they
hadn't hit the problematic patch yet.

6.18 and 6.19 were also previously released by Greg, who uses actual
quilt rather than git-quiltimport, so he wouldn't have run into this.

This also explains the bogus FAILED emails that went out for patches
that actually apply cleanly - since the patches never made it into the
-rc branch, my scripts concluded they had failed to apply.

I'll propose a fix for git to use printf '%s\n' instead of echo on
that line, which doesn't interpret escape sequences.

--
Thanks,
Sasha