Bug 89985

Summary:	system video hangs on video switch to fullscreen
Product:	Mesa	Reporter:	Mario Barrera <mbarrera>
Component:	Drivers/DRI/nouveau	Assignee:	Nouveau Project <nouveau>
Status:	RESOLVED MOVED	QA Contact:	Nouveau Project <nouveau>
Severity:	blocker
Priority:	medium
Version:	unspecified
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:

Description Mario Barrera 2015-04-11 20:01:48 UTC

Occasionally, when switching any video playback into full-screen, all system video, except for the mouse pointer, hangs until reboot. It is possible to SSH into the machine for a while until it freezes completely.

Video card: NVIDIA Corporation GK104M [GeForce GTX 670MX] (rev a1)
Operating system: Archlinux with linux 3.19.3-3-ARCH

nouveau-fw is installed and nouveau.config=NvGrUseFW=1 is in the kernel parameters.

I logged in via SSH into the blocked machine and dumped dmesg's output, it produced a 6MiB sized file of only:

[  432.267634] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267645] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267647] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267658] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267661] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267672] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267674] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267685] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267687] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267698] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267701] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267712] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267714] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267725] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267727] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267738] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267741] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267752] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267754] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267765] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: METHOD
[  432.267767] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 6 [gstglcontext[1984]] subc 0 mthd 0x000c data 0x2001054e
[  432.267777] systemd-journald[276]: /dev/kmsg buffer overrun, some messages lost.

Comment 1 Ilia Mirkin 2015-04-11 20:06:02 UTC

You shouldn't need to use blob context switching firmware... I'd start by removing

nouveau.config=NvGrUseFW=1

Also, if this is recent, and you're using libdrm-2.4.60, try downgrading to libdrm-2.4.59.

Comment 2 Mario Barrera 2015-04-11 23:46:06 UTC

(In reply to Ilia Mirkin from comment #1)
> You shouldn't need to use blob context switching firmware... I'd start by
> removing
> 
> nouveau.config=NvGrUseFW=1
> 
> Also, if this is recent, and you're using libdrm-2.4.60, try downgrading to
> libdrm-2.4.59.

Effectively this is recent, I downgraded to libdrm-2.4.59 and removed the kernel parameter. I still could reproduce the issue with a secondary screen and with the primary only, It tried that because I though it could be because of the secondary display.

relevant output of dmesg when reproducing with two displays:

[  195.394185] compiz[1209]: segfault at 7f23ad5c7580 ip 00007f23ad5c7580 sp 00007ffec56131a8 error 14 in card0[7f23ad7cd000+200000]
[  518.479158] gstglcontext[2295]: segfault at 8 ip 00007f5d39a4272e sp 00007f5b9bffe950 error 4 in libdrm_nouveau.so.2.0.0[7f5d39a3f000+6000]
[  586.592466] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: PBENTRY
[  586.592475] nouveau E[   PFIFO][0000:01:00.0] PBDMA0: ch 10 [gstglcontext[3129]] subc 0 mthd 0x0000 data 0x00000000

relevant output of dmesg when reproducing with a single display:

[  376.359606] nouveau E[   PFIFO][0000:01:00.0] read fault at 0x01b3f00000 [UNSUPPORTED_KIND] from CE2/GR_COPY on channel 0x00bf449000 [unknown]

Comment 3 GitLab Migration User 2019-09-18 20:40:01 UTC

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/1072.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.