Re: [GIT PULL] arm64 fixes for 6.13-rc3

[Mark Rutland <mark.rutland@arm.com>]

On Mon, Dec 16, 2024 at 09:59:12AM -0500, Sasha Levin wrote:
> On Mon, Dec 16, 2024 at 10:02:34AM +0000, Mark Rutland wrote:
> > On Sun, Dec 15, 2024 at 12:06:19PM -0500, Sasha Levin wrote:
> > > On Fri, Dec 13, 2024 at 05:41:38PM +0000, Catalin Marinas wrote:
> > > > - arm64 stacktrace: address some fallout from the recent changes to
> > > >  unwinding across exception boundaries
> > > 
> > > Hi Catalin, Mark,
> > > 
> > > It seems that kselftests can hit a warning that was addressed in the
> > > commit above:
> > 
> > I assume you're referring to commit:
> > 
> >  32ed1205682ec42a ("arm64: stacktrace: Skip reporting LR at exception boundaries")
> > 
> > ... ?
> 
> I was actually thinking 65ac33bed8b9 ("arm64: stacktrace: Don't WARN
> when unwinding other tasks"), sorry for the confusion

That commit should have no impact the case you report, which is seen
when unwinding the current task, not another task.

The commit immediately before that (32ed1205682ec42a) *should* fix the
issue triggered by the ftrace kselftests.

> > That was intended to fix this specific issue, as explained in the commit
> > message (with a very similar splat from running the ftrace kselftests).
> > 
> > Am I missing something? i.e. are you saying it *doesn't* fix that?
> 
> Right - LKFT is able to trigger the warning I copied on Linus's tree.

Is that "Right" as in "Right, that commit fixes it", or "Right, that
commit does not fix it"?

The warning you reported says:

	Not tainted 6.13.0-rc2 #1

... which is *before* either of the two fixes above, since the arm64
for-next/fixes branch is based on v6.13-rc2.

Are you seeing the warning with those applied? e.g. do you see this in
v6.13-rc3? I cannot reproduce the issue on v6.13-rc3 so far.

Mark.