[SOLVED] Steam causing kernel panic / system freeze

coralfang · 01-04-2018, 07:15 PM

Hi has anyone had issues with steam since the Jan 3rd updates in -current?

I'm using alienbobs steamclient package. I launched steam after updating the recent packages and reinstalling the nvidia driver... The entire computer just hangs, having to literally pull the plug.

After booting back up, the scripts in ~/.local/share/Steam are all 0 in size. I decided to delete the entire local steam folder and start clean, running steam again.

The prompt for steam downloading an update is visible as normal, and the local folder is repopulated, but then once this finishes... The system completely locks up.

I believe this could be related to the new kernel package, but I'm not sure. I am getting no other issues besides steam completely crashing my computer.

Is anyone else getting this problem and what is the best way to find out what is actually happening here?

----
Edit 1: also steam worked fine just before I updated to the latest -current packages from Jan 3rd.

----
Edit 2: i've just tried looking through every log file in /var/log and cannot find anything about a panic, crash or even a segfault. Oddly, my keyboard doesn't have a sysrq key, but upon the freeze after steam is run, no input works at all, not even toggling the caps lock LED for example.

coralfang · 01-04-2018, 10:39 PM

SOLVED. This problem is related to the recent CPU security patches involving Intel (despite me having an amd fx 8320 cpu...)

Add this to your kernel boot commands:

Code:

nopti

Steam now loads up as per usual and does not cause the system to hang/freeze. This command disables whatever those patches added for the farce involving Intel cpus as far as i know.

coralfang · 01-05-2018, 12:09 AM

Just spotted this patch included in Arch's latest kernel of 4.14.11, https://git.archlinux.org/svntogit/p...3ba403ff798529

It disables the pti related patch on AMD CPUS.

Might be worth applying that on Slackware if you have an AMD CPU for the time being, although the bootloader option for "nopti" seems to be adequate.

bassmadrigal · 01-05-2018, 02:35 AM

Quote:

Originally Posted by coralfang

Just spotted this patch included in Arch's latest kernel of 4.14.11, https://git.archlinux.org/svntogit/p...3ba403ff798529

It disables the pti related patch on AMD CPUS.

Might be worth applying that on Slackware if you have an AMD CPU for the time being, although the bootloader option for "nopti" seems to be adequate.

This fix will likely be included in the next point release of the 4.14 series kernel. The details of the patch state:

Code:

   - Exclude AMD from the PTI enforcement. Not necessarily a fix, but if
     AMD is so confident that they are not affected, then we should not
     burden users with the overhead"

SOURCE: https://git.kernel.org/pub/scm/linux...utm_source=anz

orbea · 01-05-2018, 12:41 PM

After updating to 4.14.11 with PTI enforcement built in I encountered a reproducible full system crash after running slackpkg in a 32-bit 14.2 chroot.

As I have an amd cpu I added 'nopti' to my /etc/lilo.conf and the crashes went away...

coralfang · 01-05-2018, 12:58 PM

Quote:

Originally Posted by bassmadrigal

This fix will likely be included in the next point release of the 4.14 series kernel. The details of the patch state:

Yep, appears to already be in the new 4.14.12
https://cdn.kernel.org/pub/linux/ker...ngeLog-4.14.12

Quote:

commit 151d7039757b71ebd9d170af0944562f51149372
Author: Tom Lendacky <thomas.lendacky@amd.com>
Date: Tue Dec 26 23:43:54 2017 -0600

x86/cpu, x86/pti: Do not enable PTI on AMD processors

commit 694d99d40972f12e59a3696effee8a376b79d7c8 upstream.

AMD processors are not subject to the types of attacks that the kernel
page table isolation feature protects against. The AMD microarchitecture
does not allow memory references, including speculative references, that
access higher privileged data when running in a lesser privileged mode
when that access would result in a page fault.

Disable page table isolation by default on AMD processors by not setting
the X86_BUG_CPU_INSECURE feature, which controls whether X86_FEATURE_PTI
is set.

orbea · 01-05-2018, 02:10 PM

No crash in 4.14.12 for me, but I suppose this is because PTI was masked for amd.

Paulo2 · 01-06-2018, 07:46 AM

Quote:

Originally Posted by coralfang

After booting back up, the scripts in ~/.local/share/Steam are all 0 in size.

It can be more dangerous. I launched Steam from the whiskermenu and the system froze.
After reboot, whiskermenu was with default configurations and the hierarchy in the Wine menu was lost,
so I think other files opened at the time of freeze could get corrupted too.