Go Back   nV News Forums > Linux Support Forums > NVIDIA Linux

Newegg Daily Deals

Reply
 
Thread Tools
Old 05-17-10, 06:39 PM   #13
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: Repeatable X Crash, dual FX580 on HP Z600

Hi Brad,

I've been trying to get this to happen on a similar setup (RHEL 5.2, Z600, 2x FX 580, triple-head) but haven't seen any crashes so far. I've been using comment #3 on this thread as a guideline and starting glxgears once on each screen, moving one slightly off screen, starting new glxgears instances, and moving/resizing windows on all three screens.

Do you have any further information or more detailed steps on how to produce a crash? Do you think you might be able to provide a test app that can reliably crash your system?
danix is offline   Reply With Quote
Old 05-17-10, 08:01 PM   #14
Brad.Scalio
Registered User
 
Join Date: Mar 2007
Posts: 20
Default Re: Repeatable X Crash, dual FX580 on HP Z600

Quote:
Originally Posted by danix View Post
Do you have any further information or more detailed steps on how to produce a crash? Do you think you might be able to provide a test app that can reliably crash your system?
Thanks "danix" for looking into this ... unfortunately, I cannot reproduce the problem in our lab using the same hardware setup and applications that cause X to crash in the field.

We have two applications, neither uses openGL calls, one is called D2d and one GFE. When a user logs into the workstation, and launches a D2d on each screen (3 total) then maximizes the D2d to take up the whole screen. The user the moves a single D2d in any window off screen almost entirely and launches the GFE application. Then, any resizing of the window or movement of the GFE window causes the crash. The applications unfortunately are hard to launch outside of our environment since they require certain NFS mounts for their configuration files and libraries.

I can't repeat the problems in my lab either. However, we have problems at every location with the new HP Z600 and FX580 cards. These workstations are replacing HP 4400 workstations with GeForce Go 7600 cards. This is the first Quadro we have used outside of the NVS245? (can't rememer the model).

There are no openGL calls in either application, they are approx 10 year old apps that have been rock solid since their inception. The only reason we can't disable GLX though is that a standalone application requires it.

One thing of interest...if the user launches the GFE application first, then the three D2ds, X will not crash. It seems that the order must be D2d (times three) and then GFE in order to produce the crash.

You can grab a core file from

http://pafg.arh.noaa.gov/download/lx...00517_1735.tgz

this was with DAMAGE and RENDER extension disabled. I realize this is hard to troubleshoot, it has been for us as well since we can't reproduce it here. This has halted our hardware refresh of the workstations so we have about 1,000 right now sitting in warehouses until this gets resolved somehow ;-)

The applications, D2d and GFE, render meteorological data, you can see an image here:

http://en.wikipedia.org/wiki/File:AW...orkstation.jpg

If you read the wikipedia entry for AWIPS that might help shed some light as well on the applications, they do render a lot of images.

Thanks and email me if you have any further queries. It is my username here, then at noaa dot gov

Last edited by Brad.Scalio; 05-17-10 at 08:04 PM. Reason: Added links to AWIPS on wikipedia
Brad.Scalio is offline   Reply With Quote
Old 05-18-10, 09:44 AM   #15
Brad.Scalio
Registered User
 
Join Date: Mar 2007
Posts: 20
Default Re: Repeatable X Crash, dual FX580 on HP Z600

Our previous setup had a 1GB GeForce Go card and a Quadro NVS card ... we didn't see this problem ... since the order of starting applications appears to matter, could the shape and size of the windows/pixmaps as well as order of allocation of VRAM be an issue? If so, how can one go about controlling that. The application that causes the crash, GFE, does not have the ability to control VRAM directly however.

It is looking more and more like the NVIDIA driver and X may be tripping over each other on some graphics context fields but since it only is occuring under certain circumstances it is hard to get a grasp on this.

I really do not know what to try next!
Brad.Scalio is offline   Reply With Quote
Old 05-18-10, 11:38 AM   #16
Brad.Scalio
Registered User
 
Join Date: Mar 2007
Posts: 20
Default Re: Repeatable X Crash, dual FX580 on HP Z600

I've noticed a lot of chatter on Xinerama inplementations and was wondering if similar issues without Xinerama may be present on multi-headed displays:

http://www.nvnews.net/vbulletin/showthread.php?t=142656

Granted not the same issue, but perhaps a problem similar?
Brad.Scalio is offline   Reply With Quote
Old 05-18-10, 12:50 PM   #17
Brad.Scalio
Registered User
 
Join Date: Mar 2007
Posts: 20
Default Re: Repeatable X Crash, dual FX580 on HP Z600

any chance we can get debuginfo or help with the stack and core so we can figure out what is going on?

I believe HP vendor is also pursuing this with you guys (NVIDIA) - I can't login to the customer care ticket, I forgot the passwd and it won't reset so I am relinquished to use this thread for communication
Brad.Scalio is offline   Reply With Quote
Old 05-18-10, 05:23 PM   #18
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: Repeatable X Crash, dual FX580 on HP Z600

The VGA Arbiter issue is unrelated. You're not using a new enough kernel to be affected by this.

I've had no luck investigating this with my setup so far. HP has their own investigation on their end, and hopefully they'll be able to help us with a repro of this issue.
danix is offline   Reply With Quote
Old 05-19-10, 09:40 AM   #19
Brad.Scalio
Registered User
 
Join Date: Mar 2007
Posts: 20
Default Re: Repeatable X Crash, dual FX580 on HP Z600

We have a call today however some more information that is interesting:

I can now reproduce the problem in our lab.

We have HP Z600 workstations and HP4400 workstations

HP4400 -- 1x GeForce7800GT (256MB) + 1x Quadro NVS285 (128MB) = NO CRASH
Z600 -- 2x FX580 (512MB) = CRASH

HP4400 -- 2xFX580 (512MB) = CRASH
Z600 -- 1x GeForce7800GT (256MB) + 1x Quadro NVS285 (128MB) = NO CRASH

I tried a FX1800 (1GB) card as well in both the HP4400 and Z600 and that crashes as well.

It appears to follow the FX series video cards
Brad.Scalio is offline   Reply With Quote
Old 05-19-10, 02:23 PM   #20
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: Repeatable X Crash, dual FX580 on HP Z600

Hi Brad, can you please provide more information about how to reproduce the crash?
danix is offline   Reply With Quote

Old 05-19-10, 02:27 PM   #21
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: Repeatable X Crash, dual FX580 on HP Z600

Also, is this definitely limited to a triple head configuration? Do you see it with dual head, dual GPU? How about dual head, single GPU? single head?

Since the apps that cause the crash don't require OpenGL, do we still see this crash with other drivers, ex. vesa/nv? (to rule out possible problems in X)
danix is offline   Reply With Quote
Old 05-20-10, 07:43 AM   #22
Brad.Scalio
Registered User
 
Join Date: Mar 2007
Posts: 20
Default Re: Repeatable X Crash, dual FX580 on HP Z600

Thanks for the reply -- we currently are working with HP in conjunction with nVIDIA but to update here for those who may also experience an issue, and for sake of completeness, here are some test results:

RHEL 5 update 3 with 185.19.26 in 3-head 2xFX580 -- CRASH
RHEL 5 update 3 with 195.26.15 in 3-head 2xFX580 -- CRASH

The rest are using RHEL 5 update 2 and nVIDIA driver 195.26.15 with the 2xFX580
3-head -- CRASH
2-head (same GPU) -- CRASH
2-head (different GPU) -- CRASH
Single head -- CRASH

We had some Quadro NVS 285 cards lying around, I replaced the Quadro FX580 cards with dual Quadro NVS 285 cards and we DO NOT get a crash.

I will be testing using the FOSS drivers today as well and will let you know how that goes.

So is there any implementation differences between these two as it pertains to pixmap drawing, allocation, rendering, and the moving of shaped-windows?

Last edited by Brad.Scalio; 05-20-10 at 07:44 AM. Reason: Added FOSS driver comments
Brad.Scalio is offline   Reply With Quote
Old 05-24-10, 06:07 PM   #23
Brad.Scalio
Registered User
 
Join Date: Mar 2007
Posts: 20
Default Re: Repeatable X Crash, dual FX580 on HP Z600

For the sake of completeness will continue to update here until resolution.

Tried various nVIDIA driver options, --no-ubb, no-render-accel, no-composite, etc ...

All continue to reproduce the crash except --no-render-accel, however, as expected, this causes the system to be extremely unresponsive, load times increase from an average of 10-20 seconds to 5 minutes. However, there is no crash.

There is also NO CRASH when using a single-head configuration and the vesa driver. This same configuration, using the nVIDIA driver causes a crash.

64bit OS/driver causes a crash as well

I beginning to suspect something with the UBB extension, feature, in the Quadro series cards causing our problem.

Will post any final thoughts once they become available
Brad.Scalio is offline   Reply With Quote
Old 05-25-10, 12:05 AM   #24
AaronP
NVIDIA Corporation
 
AaronP's Avatar
 
Join Date: Mar 2005
Posts: 2,487
Default Re: Repeatable X Crash, dual FX580 on HP Z600

Hi Brad,

Could you please try disabling the DOUBLE-BUFFER extension by changing 'Load "dbe"' to 'Disable "dbe"' in the Module section of /etc/X11/xorg.conf? I found some suspicious code in that part of the X server that could cause a crash like the one you're seeing.
AaronP is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 02:45 PM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.