Arch Linux

jamespharvey20

Member

Registered: 2015-06-09

Posts: 129

nouveau causes Arch to lock up every 1-3 days

(Believe this is the fault of nouveau, the open source nVidia driver, see second post for updates)

I have been experiencing a lockup on my workstation approximately every 1-3 days.  I think this has been happening since I switched from Fedora to Arch 3 months ago.  (I wish I had kept better track to be sure - everything I run autosaves well, so haven't had data loss, so haven't had it be a priority.)  My server has been running Arch without lockups.  I previously ran Windows 7 on my workstation, and until I complete a kvm install of Windows 7 with a VGA passthrough, I'm periodically switching to Windows 7 for Photoshop and gaming.  Windows 7 runs completely solid.  Before running both, I'd have Windows 7 up for months between reboots.

Keyboard LED's don't flash
Num/Caps/Scroll Lock doesn't toggle the keyboard LED's
Can't switch to text tty's
Can't ping its IP or ssh into it from another machine (sshd is setup)
Mouse cursor moves, but can't interact with anything.  System seems totally frozen except for mouse cursor.

I've set kernel.sysrq to 1, so I'll be able to try SysRq commands when it happens next.

Although I'm hoping someone might be able to give help at this point, my main reason for writing now is so I'm prepared to collect helpful information to track this down, when the next crash occurs.  What commands should I be running and logs should I be saving, on the successful boot after toggling power, next time it locks up, to add to this post later?  Anything I should change to logging settings, before it locks up, to have better data to post?

I already have a window open running top, sorted by amount of virtual memory each process uses, that is set to be on top of all other windows.

System Notes
microcode: Microcode Update Driver: v2.00 <tigran@aivazian.fsnet.co.uk>, Peter Oruba
No overclocking
Inside thoroughly vacuumed regularly, no dust blocking heatsink, all temperatures in good range
lspci -k viewable here
lsusb viewable here
lsusb -t viewable here
Single hard drive, passes extensive testing, btrfs
Most recent motherboard BIOS - MSI X99S SLI PLUS (1.8 3/20/2015)
Most recent VGA BIOS - XFX nVidia GeForce GT 640 2GB (7/14/2012)

System Specifications
DMI: MSI MS-7885/X99S SLI PLUS (MS-7885), BIOS 1.80 03/20/2015
Haswell-E -- smpboot: CPU0: Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz (fam: 06, model: 3f, stepping: 02)
Linux version 4.1.6-1-ARCH (builduser@tobias) (gcc version 5.2.0 (GCC) ) #1 SMP PREEMPT Mon Aug 17 08:52:28 CEST 2015
Memory: 32840088K/33447336K available (5699K kernel code, 893K rwdata, 1732K rodata, 1180K init, 1152K bss, 607248K reserved, 0K cma-reserved)
06:00.0 VGA compatible controller: NVIDIA Corporation GK107 [GeForce GT 640] (rev a1)
Model=TOSHIBA MD04ACA500, FwRev=FP2A --- Plugged into motherboard SATA, no SATA controllers
KDE plasma-desktop/workspace 5.4.1-2
nouveau driver 1.0.11-3

Last edited by jamespharvey20 (2015-09-13 00:18:00)

jamespharvey20

Member

Registered: 2015-06-09

Posts: 129

Re: nouveau causes Arch to lock up every 1-3 days

First time I started looking into this, journalctl didn't get the chance to log anything.  Most recent crash, it did...

JOURNALCTL SHOWS (SOMETIMES VIEWABLE AFTER REBOOT - SOMETIMES SHOWS NOTHING)
Sep 12 08:45:39 kvm kernel: nouveau E[   PFIFO][0000:06:00.0] SCHED_ERROR [ CTXSW_TIMEOUT ]
Sep 12 08:45:39 kvm kernel: nouveau E[   PFIFO][0000:06:00.0] PGRAPH engine fault on channel 7, recovering...
Sep 12 09:22:18 kvm kernel: nouveau E[   PFIFO][0000:06:00.0] PBDMA0: ACQUIRE
Sep 12 09:22:18 kvm kernel: nouveau E[   PFIFO][0000:06:00.0] PBDMA0: ch 2 [Xorg[655]] subc 0 mthd 0x001c data 0x00001004
{{{ then the last 2 lines repeat for about 50 lines, the nothing more is logged until the reboot }}}

Upstream bug report filed here

Looks like others have been reporting similar crashes for months without resolution or it really being worked on.  Looks like current workarounds are to go with the closed source proprietary nvidia driver, or to switch to AMD(ATI).

mich41

Member

Registered: 2012-06-22

Posts: 796

Re: nouveau causes Arch to lock up every 1-3 days

Switching to linux-lts may or may not help