Bug 16165 - Multi-gpu Quadro NVS 400 (NV17a x2) fails to initialise xserver
Summary: Multi-gpu Quadro NVS 400 (NV17a x2) fails to initialise xserver
Status: RESOLVED DUPLICATE of bug 16897
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/nouveau (show other bugs)
Version: git
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: Nouveau Project
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-05-30 07:35 UTC by Pete Beardmore
Modified: 2009-04-11 13:31 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
log (203.42 KB, application/octet-stream)
2008-05-30 07:35 UTC, Pete Beardmore
no flags Details
posting gpu 2 first.. (107.69 KB, application/octet-stream)
2008-05-30 07:36 UTC, Pete Beardmore
no flags Details
nouveau-Xorg.log-isolate-success-same-as-without-isolate (20.53 KB, application/octet-stream)
2009-04-11 12:25 UTC, Pete Beardmore
no flags Details
nouveau-Xorg.log-isolate-fail (6.46 KB, application/octet-stream)
2009-04-11 12:25 UTC, Pete Beardmore
no flags Details
nouveau-isolate-fail-gdb-bt (2.13 KB, application/octet-stream)
2009-04-11 12:26 UTC, Pete Beardmore
no flags Details
nouveau-Xorg.log-explicit.single.second.gpu.and.isolate-fail (6.18 KB, application/octet-stream)
2009-04-11 12:40 UTC, Pete Beardmore
no flags Details
xorg.conf (10.78 KB, application/octet-stream)
2009-04-11 12:42 UTC, Pete Beardmore
no flags Details

Description Pete Beardmore 2008-05-30 07:35:21 UTC
Created attachment 16825 [details]
log

Multi-gpu Quadro NVS 400 (NV17a x2) fails to initialise xserver.

When both of the separate pci bus addresses are used in seperate devices/screens,  the xserver instance fails.

Also, making the second gpu post first (i think that's what I've managed to do - see log3) fails when it is the sole device in use (tried IsolateDevice = bus-id and SingleCard = true options).

The noint10 option makes the loading of the driver / connection detection get further.

Dual screen using randr on DVI-A-0 and 1 is working.
Comment 1 Pete Beardmore 2008-05-30 07:36:34 UTC
Created attachment 16826 [details]
posting gpu 2 first..
Comment 2 Pete Beardmore 2009-04-11 12:25:01 UTC
Created attachment 24721 [details]
nouveau-Xorg.log-isolate-success-same-as-without-isolate
Comment 3 Pete Beardmore 2009-04-11 12:25:33 UTC
Created attachment 24722 [details]
nouveau-Xorg.log-isolate-fail
Comment 4 Pete Beardmore 2009-04-11 12:26:25 UTC
Created attachment 24723 [details]
nouveau-isolate-fail-gdb-bt
Comment 5 Pete Beardmore 2009-04-11 12:40:46 UTC
Created attachment 24724 [details]
nouveau-Xorg.log-explicit.single.second.gpu.and.isolate-fail
Comment 6 Pete Beardmore 2009-04-11 12:42:38 UTC
Created attachment 24725 [details]
xorg.conf
Comment 7 Pete Beardmore 2009-04-11 13:05:22 UTC
hi.

looking at my title now, i suppose someone could argue it's fixed, although i would be a little upset! the server doesn't crash when one specifies use of both gpus in a server layout, it just completely ignores the second screen and hence i get a dual-screen. any attempts to get the xserver to initiate the second-gpu have not worked.

originally, back in May 2008, the other DAC device and it's two outputs DVI-A-0 and DVI-A-1 were recognised, if not usable.

now, when isolateDevice '2nd gpu' is used in a quad setup, the xserver crashes:

#0  NVGetArchitecture (regs=0x0) at nv_driver.c:375
#1  0xb7b37365 in NVPciProbe (drv=0x81eefb0, entity_num=0, dev=0x81eac60, match_data=0) at nv_driver.c:435

...when isolating the 2nd gpu for a dual screen only, the X server bails with "no devices attached" -> queue a quick build of the proprietory blob just to ensure the card's 2nd gpu was still alive (it is!! ..the blob is still useless though).

i'm around on irc too if anyone fancies a semi-interactive debug session?!

cheers.
Comment 8 Pete Beardmore 2009-04-11 13:07:09 UTC
using git 'everything' as of ~ start of Apr 2009.
Comment 9 Younes Manton 2009-04-11 13:31:42 UTC
Server has issues with multiple cards. There's another bug open for this.

(Something to do with libpciaccess if I recall correctly?)

*** This bug has been marked as a duplicate of bug 16897 ***


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.