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

Newegg Daily Deals

Reply
 
Thread Tools
Old 08-14-08, 04:11 PM   #25
AaronP
NVIDIA Corporation
 
AaronP's Avatar
 
Join Date: Mar 2005
Posts: 2,487
Default Re: Spawning new session leads to crash

Please check the value of the serial number in pDrawable and pGC in damagePolyFillRect. If they don't match, then the X server incorrectly failed to revalidate the drawables Please also check pDrawable->type and pGC->pCompositeClip. If type is 0 (i.e. DRAWABLE_WINDOW), then the clip region should have been set to NULL during the switch.
AaronP is offline   Reply With Quote
Old 08-14-08, 06:47 PM   #26
devsk
Registered User
 
Join Date: Jun 2006
Posts: 46
Default Re: Spawning new session leads to crash

Quote:
Originally Posted by AaronP View Post
Please check the value of the serial number in pDrawable and pGC in damagePolyFillRect. If they don't match, then the X server incorrectly failed to revalidate the drawables
This can be seen in the prints that I posted. They match.

Quote:
Originally Posted by AaronP View Post
Please also check pDrawable->type and pGC->pCompositeClip. If type is 0 (i.e. DRAWABLE_WINDOW), then the clip region should have been set to NULL during the switch.
now, this is not true. The type is 0 and pGC->pCompositeClip is not NULL. Who is supposed to set it to NULL?
devsk is offline   Reply With Quote
Old 08-15-08, 02:41 AM   #27
AaronP
NVIDIA Corporation
 
AaronP's Avatar
 
Join Date: Mar 2005
Posts: 2,487
Default Re: Spawning new session leads to crash

Quote:
Originally Posted by devsk View Post
This can be seen in the prints that I posted. They match.
Sorry, I missed the pGC one somehow.

Quote:
Originally Posted by devsk View Post
now, this is not true. The type is 0 and pGC->pCompositeClip is not NULL. Who is supposed to set it to NULL?
It should have happened during xf86SetRootClip, which is called during a VT switch by xf86EnableDisableFBAccess and should have also changed the window serial number.
AaronP is offline   Reply With Quote
Old 08-15-08, 09:42 AM   #28
devsk
Registered User
 
Join Date: Jun 2006
Posts: 46
Default Re: Spawning new session leads to crash

Quote:
Originally Posted by AaronP View Post
Sorry, I missed the pGC one somehow.


It should have happened during xf86SetRootClip, which is called during a VT switch by xf86EnableDisableFBAccess and should have also changed the window serial number.
where are these functions that you refer to? Are they in xorg-server code base? Does this mean that its an Xorg bug? If it is, then why doesn't this happen on every VT switch?
devsk is offline   Reply With Quote
Old 08-15-08, 01:46 PM   #29
AaronP
NVIDIA Corporation
 
AaronP's Avatar
 
Join Date: Mar 2005
Posts: 2,487
Default Re: Spawning new session leads to crash

I'm trying to determine whose bug it is, exactly. The xf86* functions are indeed in the core X server. Your backtrace indicates that the X server is receiving a rendering request from a client trying to render a filled rectangle to a window. Since the server isn't on the VT, that rendering should have been clipped away but instead it's calling into the driver to perform the rendering with a non-NULL clip region. If the GC had been properly revalidated, the driver would have called down to the software rendering library (libwfb) immediately but it's crashing somewhere instead, probably trying to access the screen's video memory which was freed during the VT switch.

I'll see if I can reproduce the problem with some directed testing, but I'm on vacation right now so it'll be a while until I can get to it.
AaronP is offline   Reply With Quote
Old 08-20-08, 12:06 PM   #30
devsk
Registered User
 
Join Date: Jun 2006
Posts: 46
Default Re: Spawning new session leads to crash

Quote:
Originally Posted by AaronP View Post
I'm trying to determine whose bug it is, exactly. The xf86* functions are indeed in the core X server. Your backtrace indicates that the X server is receiving a rendering request from a client trying to render a filled rectangle to a window. Since the server isn't on the VT, that rendering should have been clipped away but instead it's calling into the driver to perform the rendering with a non-NULL clip region. If the GC had been properly revalidated, the driver would have called down to the software rendering library (libwfb) immediately but it's crashing somewhere instead, probably trying to access the screen's video memory which was freed during the VT switch.

I'll see if I can reproduce the problem with some directed testing, but I'm on vacation right now so it'll be a while until I can get to it.
@AaronP: I still have the debug enabled X running. Please let me know if you need any more information or debug info. This still happens with latest beta drivers.
devsk is offline   Reply With Quote
Old 08-23-08, 03:10 AM   #31
devsk
Registered User
 
Join Date: Jun 2006
Posts: 46
Default Re: Spawning new session leads to crash

Hey, I am hanging in here...did you make any progress with this? Live debugging session with a debugger attached is a rare opportunity to solve a problem because one can see all state.

Let me know if you are not able to reproduce this. It is little tough. But what I do is to start two X sessions in two different users and then from one session as root in a xterm, I start:

while true;do sleep 5;chvt 7;sleep 5;chvt 9;done

Before this I ssh from a different box and attach GDB to X of the session which started the second session.

In each X session, I have a command to switch desktop to next (6 desktops in one session and 4 in the other, all of them with variety of text and images) just throw random text and images to the driver to display: while true;do sleep 6;changeDesktop 1;done

Then, I go back to my ssh session, sip coffee and wait for SEGV....:-)
devsk is offline   Reply With Quote
Old 09-13-08, 11:16 PM   #32
Alejandro Nova
Registered User
 
Join Date: Sep 2006
Posts: 26
Default Re: Spawning new session leads to crash

Quote:
Originally Posted by AaronP View Post
All of the reports of this problem seem to be with xserver 1.4.0.90, so my guess is that some change in that particular server version broke something. Also, I've seen reports of BackingStore causing crashes due to missing GC validation in the server somewhere. For those that have BackingStore enabled, does disabling it fix the problem? Also, please try the nv or vesa drivers to see if they have a similar problem.
Half-off-topic (using your quote: that's why I am posting here): Will you fix those BackingStore crashes with XServer 1.5 (released) in a next 173.xx.xx driver release? Will those fixes come only for 177 series? I'm stuck here with a FX 5500, and BackingStore is all I need by now.
Alejandro Nova is offline   Reply With Quote

Old 09-15-08, 09:43 PM   #33
AaronP
NVIDIA Corporation
 
AaronP's Avatar
 
Join Date: Mar 2005
Posts: 2,487
Default Re: Spawning new session leads to crash

Quote:
Originally Posted by Alejandro Nova View Post
Half-off-topic (using your quote: that's why I am posting here): Will you fix those BackingStore crashes with XServer 1.5 (released) in a next 173.xx.xx driver release? Will those fixes come only for 177 series? I'm stuck here with a FX 5500, and BackingStore is all I need by now.
If there really is a GC validation problem with BackingStore, it would be an X server bug and would be difficult to work around in the driver. BackingStore is pretty much a transparent layer on top of the driver, and the driver itself doesn't really have much to do with it.

I seem to recall some of the X developers questioning the usefulness of BackingStore in general, and I think it might be removed in favor of Composite in future X servers. I would recommend asking on the xorg@lists.freedesktop.org mailing list.
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


Similar Threads
Thread Thread Starter Forum Replies Last Post
Passion Leads Army: DX 11 and GPU PhysX Benchmark News Archived News Items 0 05-28-12 08:00 PM
Titan Supercomputer Session Showcases Science on GPUs News Archived News Items 0 05-16-12 03:30 AM
gf 6150se crash! oddhornedant NVIDIA Linux 2 05-05-12 05:30 AM
xvideo crash elanthis NVIDIA Linux 5 10-29-02 07:12 PM

All times are GMT -5. The time now is 09:25 PM.


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