Received-SPF: pass (google.com: domain of linux-kernel+bounces-15859-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Date: Wed, 3 Jan 2024 18:24:19 +0000
From: Mark Brown <broonie@kernel.org>
To: Will Deacon <will@kernel.org>
Cc: syzbot <syzbot+95ffb6a83b20ea7f4f55@syzkaller.appspotmail.com>,
	catalin.marinas@arm.com, linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org, syzkaller-bugs@googlegroups.com,
	mark.rutland@arm.com
Subject: Re: [syzbot] [arm?] WARNING in do_sve_acc
Message-ID: <ZZWmU9zFdIqfQL8x@finisterre.sirena.org.uk>
References: <000000000000511e9a060ce5a45c@google.com>
 <20240103180548.GB6200@willie-the-truck>
Precedence: bulk
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="0bf6vRTcG7q0E1me"
Content-Disposition: inline
In-Reply-To: <20240103180548.GB6200@willie-the-truck>


--0bf6vRTcG7q0E1me
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

On Wed, Jan 03, 2024 at 06:05:49PM +0000, Will Deacon wrote:
> [+Broonie and Mark R as I'm sure we've seen this one before, possibly
>  in a report from Naresh? Was it a QEMU issue?]

It's been reported by the LKFT people, but it occurs with *extremely*
low frequency, there's no reproducer and nothing showed up from code
inspection by myself and Mark.  The reports suggest it's been there
since ~v5.15 (that's the oldest stable it was seen in and Linaro do
still cover older), though Linaro's reports were conflating multiple
potentially related symptoms so it's a little unclear what exactly was
seen in what version.

I suspect to the extent it's qemu it's just that that's mostly where
people are exercising SVE.  This was shown with a userspace with SVE
usage in glibc, I *think* that upgrading to a similar glibc version was
what started Linaro seeing things too based on the release dates for the
particular stable versions they mentioned but didn't get them to confirm
yet.

I *might've* been able to show something with current code just before
Christmas but the symptoms are a bit different, I was going to start
looking when I get back.

> > Unfortunately, I don't have any reproducer for this issue yet.

Indeed...

> > compiler:       aarch64-linux-gnu-gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40

> > pc : do_sve_acc+0x6c/0x10c arch/arm64/kernel/fpsimd.c:1394
> > lr : do_sve_acc+0x5c/0x10c arch/arm64/kernel/fpsimd.c:1391

--0bf6vRTcG7q0E1me
Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----

iQEzBAABCgAdFiEEreZoqmdXGLWf4p/qJNaLcl1Uh9AFAmWVplAACgkQJNaLcl1U
h9AKLQf/RF8z9wVHPXxMFkRTCcBegv3D1GlkFGzCGdaxBr98Yzgyc1LPn1X835gs
bNPZSL7oFU0HjizrPPI+khfT43aWlhrHFTXtLWO2ETcjxUqqhnUYvh7y2L3nrVmP
CbgSEzWMMGjz/PZmWGgj98W2yi2Xps+KmrMqCJM1RrXTa4wavW+tLV7/T9QugLUq
kF5fSWivhQTrSLRaLGYyE27gZkyCpBMsbwSuSQGH9RrUYO0JUBwgRZZHHaobInZM
TpqOh4RctVboXY6aFF4Z1UXBqhK3E0K8hlW3WjIJYbksG/MIk+BVTlIdjtLE0CmT
e4Akq7Nkm48Aj1tHZ8MDcpSWaPuh+g==
=LjvY
-----END PGP SIGNATURE-----

--0bf6vRTcG7q0E1me--