Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
From:   Rui Salvaterra <rsalvaterra@gmail.com>
Date:   Wed, 5 Apr 2023 23:14:45 +0100
Message-ID: <CALjTZvakMZQsqzzjjXh_fw4+Zra5y5o-fhc5gDVTYM1uM40RNQ@mail.gmail.com>
Subject: [BUG?] unchecked MSR access error: WRMSR to 0x19c
To:     x86@kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi, everyone,

I have a Haswell (Core i7-4770R) machine running Linux 6.3-rc5 on
which, after a while under load (say, compiling the kernel), I get
this trace=E2=80=A6

[  832.549630] unchecked MSR access error: WRMSR to 0x19c (tried to
write 0x000000000000aaa8) at rIP: 0xffffffff816f66a6
(throttle_active_work+0xa6/0x1d0)
[  832.549652] Call Trace:
[  832.549654]  <TASK>
[  832.549655]  process_one_work+0x1ab/0x300
[  832.549661]  worker_thread+0x4b/0x340
[  832.549664]  ? process_one_work+0x300/0x300
[  832.549676]  kthread+0xac/0xc0
[  832.549679]  ? kthread_exit+0x20/0x20
[  832.549682]  ret_from_fork+0x1f/0x30
[  832.549693]  </TASK>

=E2=80=A6 after which I get these from time to time in dmesg.

[  836.709562] CPU7: Core temperature is above threshold, cpu clock is
throttled (total events =3D 219)
[  836.709569] CPU3: Core temperature is above threshold, cpu clock is
throttled (total events =3D 219)
[ 1272.792138] CPU2: Core temperature is above threshold, cpu clock is
throttled (total events =3D 1)
[ 1272.792156] CPU6: Core temperature is above threshold, cpu clock is
throttled (total events =3D 1)

This is the microcode revision on the CPU.

[    0.000000] microcode: updated early: 0xe -> 0x1c, date =3D 2019-11-12

Note that I have the exact same issue on an Ivy Bridge (Core
i7-3720QM) machine, but not on an Ivy Bridge laptop (Celeron 1007U).
Maybe this is a legitimate warning, but please note that I've
thorughly cleaned the machines before retesting to see if, by
coincidence, I had any airway/cooling issues. The fact that it started
happening recently (since Linux 6.1, I believe), and the fact that
running stress-ng --cpu 16 before the unchecked WRMSR error happens
doesn't cause any thermal throttling events, lead me to believe this
is possibly some unintended oversight.

Please let me know if you need any additional information (.config, or
anything else).

Thanks in advance,
Rui