Bug 68665

Summary:	[NVE4] Crashing games when set AA to x2 on GTX760
Product:	Mesa	Reporter:	Igor Gnatenko <i.gnatenko.brain>
Component:	Drivers/DRI/nouveau	Assignee:	Nouveau Project <nouveau>
Status:	RESOLVED FIXED	QA Contact:
Severity:	normal
Priority:	medium	CC:	i.gnatenko.brain
Version:	9.2
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:

Description Igor Gnatenko 2013-08-28 14:24:40 UTC

Version: 9.2

I see many bugreports with this problem:
https://bugzilla.redhat.com/show_bug.cgi?id=1001714
https://bugzilla.redhat.com/show_bug.cgi?id=1001698
https://bugzilla.redhat.com/show_bug.cgi?id=1001740

I see always problem w/ pushbuf_dump. It's crashing w/ SIGSEGV. The same of:
Thread no. 1 (10 frames)
 #0 pushbuf_dump at pushbuf.c:296
 #1 pushbuf_submit at pushbuf.c:361
 #2 pushbuf_flush at pushbuf.c:402
 #3 nouveau_pushbuf_kick at pushbuf.c:773
 #4 PUSH_KICK at ../../../../src/gallium/drivers/nouveau/nouveau_winsys.h:56
 #5 nvc0_flush at nvc0_context.c:45
 #6 st_glFlush at ../../src/mesa/state_tracker/st_cb_flush.c:121
 #7 _mesa_make_current at ../../src/mesa/main/context.c:1455
 #8 st_api_make_current at ../../src/mesa/state_tracker/st_manager.c:737
 #9 dri_unbind_context at dri_context.c:217

All backtraces, some info in links to RHBZ. If you need additional info you're welcome.

Comment 1 Ilia Mirkin 2013-08-28 14:38:42 UTC

Do you see anything in dmesg when this happens?

Comment 2 Igor Gnatenko 2013-08-28 14:41:22 UTC

(In reply to comment #1)
> Do you see anything in dmesg when this happens?
Aug 27 14:26:43 Elias kernel: [15584.798799] nouveau E[  PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 5 [0x000007f73d supertuxkart[6691]] subc 1 class 0xa0c0 mthd 0x02e8 data 0x0040aaaa
Aug 27 14:26:43 Elias kernel: [15584.798816] nouveau E[  PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 5 [0x000007f73d supertuxkart[6691]] subc 1 class 0xa0c0 mthd 0x02f4 data 0x0040aaaa
Aug 27 14:27:22 Elias kernel: [15623.033892] nouveau E[  PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 5 [0x000007f73d supertuxkart[6709]] subc 1 class 0xa0c0 mthd 0x02e8 data 0x0040aaaa
Aug 27 14:27:22 Elias kernel: [15623.033912] nouveau E[  PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 5 [0x000007f73d supertuxkart[6709]] subc 1 class 0xa0c0 mthd 0x02f4 data 0x0040aaaa
Aug 27 15:51:56 Elias kernel: [20697.098486] nouveau E[  PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 5 [0x000007f73d supertuxkart[6867]] subc 1 class 0xa0c0 mthd 0x02e8 data 0x0040aaaa
Aug 27 15:51:56 Elias kernel: [20697.098503] nouveau E[  PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 5 [0x000007f73d supertuxkart[6867]] subc 1 class 0xa0c0 mthd 0x02f4 data 0x0040aaaa
Aug 27 15:51:56 Elias kernel: [20697.328287] traps: supertuxkart[6867] general protection ip:7fed62c742af sp:7fff81412af0 error:0 in libdrm_nouveau.so.2.0.0[7fed62c71000+6000]
Aug 27 15:51:57 Elias abrt[6878]: Saved core dump of pid 6867 (/usr/bin/supertuxkart) to /var/tmp/abrt/ccpp-2013-08-27-15:51:56-6867 (36970496 bytes)

In other bugreports I see the similar.

Comment 3 Ilia Mirkin 2013-08-28 14:49:45 UTC

To answer my own question, from the linked bugs, there is:

Aug 27 16:51:15 Elias kernel: [24255.888373] nouveau E[  PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 5 [0x000007f73d xonotic-sdl[7519]] subc 1 class 0xa0c0 mthd 0x02f4 data 0x0040aaaa

Unfortunately, Igor, you're not the original reporter of these bugs, as I understand it. (And more importantly, can't test potential fixes.) Hopefully someone on the nouveau team has the requisite hardware to be able to test this out. Given that it's a kepler card, I think it's important to make sure that the latest git kernel is used, as there are fixes not available in 3.10 regarding context switching and initialization. (Not sure if those are at fault here, but it's good to get that out of the way.)

Comment 4 Igor Gnatenko 2013-08-28 14:58:45 UTC

(In reply to comment #3)
> Unfortunately, Igor, you're not the original reporter of these bugs, as I
> understand it. (And more importantly, can't test potential fixes.) 
You can write there and I can provide info for RHBZ reporter and conversely.

Comment 5 Maarten Lankhorst 2013-08-29 10:33:15 UTC

Already fixed in 9.2, your snapshot is too old.

Comment 6 Igor Gnatenko 2013-08-29 10:52:17 UTC

(In reply to comment #5)
> Already fixed in 9.2, your snapshot is too old.
Ok. We will update our mesa to 9.2. If we can reproduce this I will re-open this bug. Thanks.

Comment 7 Igor Gnatenko 2013-09-05 08:18:43 UTC

Ok. We snapshotted git from 02 sep 2013. Problem is still present.
https://bugzilla.redhat.com/show_bug.cgi?id=1001714
https://bugzilla.redhat.com/show_bug.cgi?id=1004674
and older.

Comment 8 Ilia Mirkin 2013-09-05 08:23:35 UTC

There is no way to get the data ending with aaaa in the new code. If they're still seeing it, it's still an old snapshot. (Of course, the problem may be entirely unrelated to that error.) Fresh logs would be nice. And a debug mesa build wouldn't hurt either, as that enables all sorts of extra checks.

Comment 9 Maarten Lankhorst 2013-09-05 10:35:43 UTC

revert b77316ad7594fa5873717992b6986cb4c0179d23, that should fix it

Comment 10 Igor Gnatenko 2013-09-06 18:34:55 UTC

(In reply to comment #9)
> revert b77316ad7594fa5873717992b6986cb4c0179d23, that should fix it
yes. thats fixed problem.

Comment 11 Maarten Lankhorst 2013-09-11 07:05:45 UTC

Fixed in master branch with commit b217d48364f368f541e53006af5dd56f664be24d.
"st/dri: do not create a new context for msaa copy"

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.