* RE: Intel laptop: Starting with `maxcpus=1` and then bringing other CPUs online freezes system
[not found] <125e904a-088e-f111-00a4-95c3b18d882f@molgen.mpg.de>
@ 2020-06-11 13:13 ` Mario.Limonciello
2020-06-11 16:17 ` Borislav Petkov
0 siblings, 1 reply; 2+ messages in thread
From: Mario.Limonciello @ 2020-06-11 13:13 UTC (permalink / raw)
To: pmenzel, tglx, mingo, bp; +Cc: x86, linux-kernel, hdegoede, ckellner, linux-pm
> -----Original Message-----
> From: Paul Menzel <pmenzel@molgen.mpg.de>
> Sent: Thursday, June 11, 2020 3:59 AM
> To: Thomas Gleixner; Ingo Molnar; Borislav Petkov
> Cc: x86@kernel.org; LKML; Limonciello, Mario; Hans de Goede; Christian
> Kellner; linux-pm@vger.kernel.org
> Subject: Intel laptop: Starting with `maxcpus=1` and then bringing other
> CPUs online freezes system
>
>
> [EXTERNAL EMAIL]
>
> Dear Linux folks,
>
>
> This is an interesting one. On a Dell Precision 3540 with a dedicated
> AMD graphics card, and with Debian Sid/unstable and Linux 5.6.14, there
> are several strange issues (one example [1]).
>
> Starting the system, and adding `maxcpus=1` to the Linux command line,
> the system starts.
>
> Trying to bring up more CPUs, works for the first one, but fails for the
> second. The system freezes then. Nothing is dumped on the Linux virtual
> terminal.
>
> In my limited tests, this always happens, when the system runs on
> batteries, that means, without the power cord plugged in. But I was also
> able to reproduce it with the USB Type-C power cable plugged in. But
> sometimes it works (to bring all other seven CPUs online).
>
> Please find the Linux messages up to bringing CPU 1 online attached with
> no power cord plugged in, and the messages of a successful operation.
>
> Is this expected? I’d appreciated tips, if it’s possible to debug this
> further.
>
>
> Kind regards,
>
> Paul
>
>
> [1]: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1883065
From some other similar reports I've heard this could be related to the ucode
loading causing a freeze. I would suggest trying to disable the Intel ucode
loader with the initramfs as a debugging tactic.
^ permalink raw reply [flat|nested] 2+ messages in thread