2012-11-17 19:38:06

by Mikael Pettersson

[permalink] [raw]
Subject: [3.7-rc5/rc6 regression] "drm/nvc0/disp: fix regression in vblank semaphore release" broke nouveau driver and mplayer

mplayer worked fine on my Dell Latitude E6510 (nVidia GT218 [NVS 3100M] graphics)
up to and including kernel 3.7-rc4. However, with 3.7-rc5 or -rc6, any attempt to
run mplayer just blanks the screen, shows some stray white pixels in the upper left
corner, kills the X server, and spews the following errors from the kernel:

nouveau E[ PGRAPH][0000:01:00.0] TRAP_M2MF NOTIFY
nouveau E[ PGRAPH][0000:01:00.0] TRAP_M2MF 00304041 000043e0 00000000 06000434
nouveau [ PGRAPH][0000:01:00.0] TRAP
nouveau E[ PGRAPH][0000:01:00.0] ch 2 [0x001fb44000] subc 3 class 0x5039 mthd 0x0328 data 0x00000000
nouveau E[ PFB][0000:01:00.0] trapped read at 0x002001a020 on channel 0x0001fb44 SEMAPHORE_BG/PFIFO_READ/00 reason: PAGE_NOT_PRESENT
nouveau [ PFIFO][0000:01:00.0] CACHE_ERROR - Ch 2/4 Mthd 0x0068 Data 0x11111111
nouveau E[ PGRAPH][0000:01:00.0] TRAP_DISPATCH (unknown 0x00000004)
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS
nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 3 class 0x0000 mthd 0x0860 data 0x00000000
nouveau E[ PFB][0000:01:00.0] trapped read at 0x002001a024 on channel 0x0001fb44 PFIFO/PFIFO_READ/SEMAPHORE reason: PAGE_NOT_PRESENT
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS
nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 3 class 0x0000 mthd 0x0860 data 0x00000000
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS
nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 3 class 0x0000 mthd 0x0860 data 0x00000000
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS

(this bit repeats itself for 700+ lines)

nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 3 class 0x0000 mthd 0x0860 data 0x00000000
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS
nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 5 class 0x0000 mthd 0x0860 data 0x00000000
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS

(this bit repeats itself for 30+ lines)

nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 5 class 0x0000 mthd 0x03c4 data 0x40000000
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS
nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 5 class 0x0000 mthd 0x03c8 data 0x00000000
nouveau E[ PFB][0000:01:00.0] trapped write at 0x0000000000 on channel 0x0001fb44 PGRAPH/DISPATCH/GRCTX reason: DMAOBJ_LIMIT
nouveau E[ PGRAPH][0000:01:00.0] TRAP_DISPATCH (unknown 0x00000004)
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS
nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 5 class 0x0000 mthd 0x03cc data 0x40000000
nouveau E[ PFB][0000:01:00.0] trapped write at 0x0000000420 on channel 0x0001fb44 PGRAPH/DISPATCH/GRCTX reason: DMAOBJ_LIMIT
nouveau [ PGRAPH][0000:01:00.0] ILLEGAL_MTHD ILLEGAL_CLASS
nouveau E[ PGRAPH][0000:01:00.0] ch -1 [0x001fb44000] subc 3 class 0x502d mthd 0x0860 data 0x00000000
nouveau E[ PFB][0000:01:00.0] trapped write at 0x000002b000 on channel 0x0001fcb0 PGRAPH/DISPATCH/GRCTX reason: DMAOBJ_LIMIT

The error is 100% repeatable.

git bisect identified the following culprit:

11d92561c81be2f4a7af37f035e1af294b960abe is the first bad commit
commit 11d92561c81be2f4a7af37f035e1af294b960abe
Author: Kelly Doran <[email protected]>
Date: Wed Nov 7 10:02:04 2012 +1000

drm/nvc0/disp: fix regression in vblank semaphore release

Signed-off-by: Kelly Doran <[email protected]>
Reviewed-by: Maarten Lankhorst <[email protected]>
Signed-off-by: Ben Skeggs <[email protected]>

:040000 040000 e539bc754b029da133f89f3bcf5bf31495cb07c5 4e779444a976c40cb07aafefc8e6e7b1e64f092c M drivers

I've confirmed that reverting this from -rc5 and -rc6 allows mplayer to work again.

User-space is Fedora 15 x86_64 w/ final updates, plus mplayer-1.0-0.129.20110917svn from rpmfusion.
There are no binary-only or otherwise out-of-tree kernel or X drivers anywhere on the machine.

Please revert or fix this breakage before kernel 3.7.0 final.

/Mikael


2012-11-17 21:45:30

by Jamie Heilman

[permalink] [raw]
Subject: Re: [3.7-rc5/rc6 regression] "drm/nvc0/disp: fix regression in vblank semaphore release" broke nouveau driver and mplayer

Mikael Pettersson wrote:
> mplayer worked fine on my Dell Latitude E6510 (nVidia GT218 [NVS
> 3100M] graphics) up to and including kernel 3.7-rc4. However, with
> 3.7-rc5 or -rc6, any attempt to run mplayer just blanks the screen,
> shows some stray white pixels in the upper left corner, kills the X
> server, and spews the following errors from the kernel:
...
>
> The error is 100% repeatable.

I see the same problem on my workstation with a G86 [Quadro NVS 290] (rev a1)
running Debian (sid) with:
xserver-xorg-core 2:1.12.4-3
libdrm-nouveau1a:amd64 2.4.33-3
xserver-xorg-video-nouveau 1:1.0.1-3
and mplayer2 2.0-600-g95e81df w/the xv video output driver

> git bisect identified the following culprit:
>
> 11d92561c81be2f4a7af37f035e1af294b960abe is the first bad commit

I bisected to the same commit as well.


--
Jamie Heilman http://audible.transient.net/~jamie/

2012-11-17 22:03:11

by Marcin Slusarz

[permalink] [raw]
Subject: Re: [3.7-rc5/rc6 regression] "drm/nvc0/disp: fix regression in vblank semaphore release" broke nouveau driver and mplayer

On Sat, Nov 17, 2012 at 08:35:18PM +0100, Mikael Pettersson wrote:
> mplayer worked fine on my Dell Latitude E6510 (nVidia GT218 [NVS 3100M] graphics)
> up to and including kernel 3.7-rc4. However, with 3.7-rc5 or -rc6, any attempt to
> run mplayer just blanks the screen, shows some stray white pixels in the upper left
> corner, kills the X server, and spews the following errors from the kernel:

Fix was already posted and should be merged soon.

http://cgit.freedesktop.org/nouveau/linux-2.6/commit/?id=7a259e65569bd7593ad541c84982027969ec9c45

2012-11-18 11:44:29

by Mikael Pettersson

[permalink] [raw]
Subject: Re: [3.7-rc5/rc6 regression] "drm/nvc0/disp: fix regression in vblank semaphore release" broke nouveau driver and mplayer

Marcin Slusarz writes:
> On Sat, Nov 17, 2012 at 08:35:18PM +0100, Mikael Pettersson wrote:
> > mplayer worked fine on my Dell Latitude E6510 (nVidia GT218 [NVS 3100M] graphics)
> > up to and including kernel 3.7-rc4. However, with 3.7-rc5 or -rc6, any attempt to
> > run mplayer just blanks the screen, shows some stray white pixels in the upper left
> > corner, kills the X server, and spews the following errors from the kernel:
>
> Fix was already posted and should be merged soon.
>
> http://cgit.freedesktop.org/nouveau/linux-2.6/commit/?id=7a259e65569bd7593ad541c84982027969ec9c45

That patch fixes the bug I reported. Thanks.

2012-11-21 17:03:14

by Joseph Salisbury

[permalink] [raw]
Subject: Re: [3.7-rc5/rc6 regression] "drm/nvc0/disp: fix regression in vblank semaphore release" broke nouveau driver and mplayer

On 11/17/2012 05:03 PM, Marcin Slusarz wrote:
> On Sat, Nov 17, 2012 at 08:35:18PM +0100, Mikael Pettersson wrote:
>> mplayer worked fine on my Dell Latitude E6510 (nVidia GT218 [NVS 3100M] graphics)
>> up to and including kernel 3.7-rc4. However, with 3.7-rc5 or -rc6, any attempt to
>> run mplayer just blanks the screen, shows some stray white pixels in the upper left
>> corner, kills the X server, and spews the following errors from the kernel:
> Fix was already posted and should be merged soon.
>
> http://cgit.freedesktop.org/nouveau/linux-2.6/commit/?id=7a259e65569bd7593ad541c84982027969ec9c45
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
Do you happen to know if this fix will make it into 3.7, or do you think
it won't happen until 3.8?

Thanks,

Joe