Re: [stable 4.9] PANIC: double fault, error_code: 0x0 - clang boot failed on x86_64

From: Nick Desaulniers <ndesaulniers@google.com>
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Naresh Kamboju <naresh.kamboju@linaro.org>
Cc: linux-stable <stable@vger.kernel.org>,
	open list <linux-kernel@vger.kernel.org>,
	lkft-triage@lists.linaro.org, Sasha Levin <sashal@kernel.org>,
	clang-built-linux <clang-built-linux@googlegroups.com>,
	x86@vger.kernel.org, Matthias Kaehlcke <mka@chromium.org>
Subject: Re: [stable 4.9] PANIC: double fault, error_code: 0x0 - clang boot failed on x86_64
Date: Tue, 1 Dec 2020 12:10:04 -0800	[thread overview]
Message-ID: <CAKwvOdkYWBsL-QXcfPztsCTzyAHLiSodzDznDxOz1MPkktWS1w@mail.gmail.com> (raw)
In-Reply-To: <X8X8y4j9Ip+C5DwS@kroah.com>

On Tue, Dec 1, 2020 at 12:19 AM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
>
> On Mon, Nov 30, 2020 at 12:12:39PM -0800, Nick Desaulniers wrote:
> > On Wed, Nov 25, 2020 at 10:38 PM Greg Kroah-Hartman
> > <gregkh@linuxfoundation.org> wrote:
> > >
> > > Is the mainline 4.9 tree supposed to work with clang?  I didn't think
> > > that upstream effort started until 4.19 or so.
> >
> > (For historical records, separate from the initial bug report that
> > started this thread)
> >
> > I consider 785f11aa595b ("kbuild: Add better clang cross build
> > support") to be the starting point of a renewed effort to upstream
> > clang support. 785f11aa595b landed in v4.12-rc1.  I think most patches
> > landed between there and 4.15 (would have been my guess).  From there,
> > support was backported to 4.14, 4.9, and 4.4 for x86_64 and aarch64.
> > We still have CI coverage of those branches+arches with Clang today.
> > Pixel 2 shipped with 4.4+clang, Pixel 3 and 3a with 4.9+clang, Pixel 4
> > and 4a with 4.14+clang.  CrOS has also shipped clang built kernels
> > since 4.4+.
>
> Thanks for the info.  Naresh, does this help explain why maybe testing
> these kernel branches with clang might not be the best thing to do?

On the contrary, I think it's very much worthwhile to test these
branches with Clang.  Particularly since CrOS is shipping x86_64
devices built with Clang since 4.4.y.  This looks like a problem
that's potentially been fixed but the fix not yet identified and
backported.  It would be good for us to identify and fix the issue
before it becomes a problem for CrOS.

Though, it looks like CrOS just skipped 4.9...? Looking at:
https://chromium.googlesource.com/chromiumos/third_party/kernel/+refs
I don't see a chromeos-4.9 branch.

That said, I still find such reports helpful to track.
-- 
Thanks,
~Nick Desaulniers