Bug 12712 - Nvidia fail to allocate memory on resume (so X crash) if more than a couple programs are running
Summary: Nvidia fail to allocate memory on resume (so X crash) if more than a couple p...
Status: RESOLVED FIXED
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: 4
Hardware: x86_64 Linux
Priority: Normal critical
Target Milestone: ---
Assignee: Mageia Bug Squad
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-02-10 23:54 CET by Morgan Leijström
Modified: 2014-10-27 22:01 CET (History)
1 user (show)

See Also:
Source RPM: x11-server-xorg-1.14.5-2.mga4
CVE:
Status comment:


Attachments
Xorg.0.log(.old) of crashing session (23.41 KB, text/plain)
2014-02-11 20:29 CET, Morgan Leijström
Details

Description Morgan Leijström 2014-02-10 23:54:17 CET
When having many programs *1) loaded, X crash *2) after resume from suspend or hibernation *3)

*1) example: opera, firefox, chrome (the former with a couple pages each, the ones after this parenthesis are enough without data files for simple test:), showfoto, gwenview, scribus, libreoffice draw+writer+calc+impress. 

*2) i assume, nvidia logo briefly shows, then normal KDE login, then KDE loads desktop from last logout/shotdown/reboot - not as when hibernated/suspended.

*3) restore from hibernation is extremely slow, bug 12706

Affected machines:  Lenovo Thinkpad R61 and T61
lspci -vv from R61, se: https://bugs.mageia.org/attachment.cgi?id=4973


I set it critical as user may loose work.

Any idea what to look for in what log?
Comment 1 Morgan Leijström 2014-02-11 14:11:26 CET
Contrast: My old mga4 desktop system 4xOpteron, Radeon, resumes nicely from hibernation even with more programs running.  (Can not test suspend there, probably not compatible hardware)
Comment 2 Morgan Leijström 2014-02-11 20:29:39 CET
Created attachment 4978 [details]
Xorg.0.log(.old) of crashing session

Excerpt:
[  3835.638] (EE) NVIDIA(0): Failed to allocate primary buffer: out of memory.
[  3835.640] (EE) NVIDIA(0):  *** Aborting ***
Morgan Leijström 2014-02-11 20:36:36 CET

Source RPM: (none) => x11-server-xorg-1.14.5-2.mga4

Comment 3 Morgan Leijström 2014-05-08 12:25:00 CEST
I searched teh web for that X error and found some who solved by adding kernel parameter nopat.
I tried but that did not help in my case...
Comment 4 Morgan Leijström 2014-05-10 21:56:18 CEST
cc thierry ; maybe this is a kernel bug ?

Maybe some configuration for reserving enough memory?
-but why should it need more memory after resume??
   - is it not reusing/restoring it correctly?
Any setting i could try to change?

If you have time to dig into this and think cauldron is a better test bed i can make a disk available for swapping in and install cauldron as testing system.



        Additional info:

§ this happen both for noveau and nvidia proprietary driver.

§ most often just the X session crash and restart, but sometimes it is a full hang with mga background, sometimes black, sometimes but seldom reboot

§ brightness is not working - maybe related but probably not

§ I believe at least the R61 also had this problem om mga3, but I did not document it.

§ i have tried with desktops cinnamon and KDE, on mga4, same problem.

§ My (wifes) R61 is now running fedora 20 - same problem, but expresses itself a bit different; i do not get to KDE login instead it either reboots or hang. And seem less rare than on mga4.  And brightness works.  Only tried with noveau driver on fedora, with KDE.  Also on fedora there are some - pretty different - bug reports on resuming.


The Thinkpad T61 is still on mga4.
Tell me what log or other info could be helpful to track this down.

CC: (none) => thierry.vignaud
Summary: X crash on resume if many programs are running => Nvidia fail to allocate memory on resume (so X crash) if more than a couple programs are running

Comment 5 Morgan Leijström 2014-05-13 00:04:12 CEST
corresponding bug on fedora
https://bugzilla.redhat.com/show_bug.cgi?id=1096989
Comment 6 Morgan Leijström 2014-07-29 23:55:28 CEST
Hurray!
This problem seem gone in mga5 x86_64 , tested in fresh network install shortly after alpha 1 isos release, and also at updates until now on both thinkpad R61 and T61.  Kernel 3.15.6-desktop-1.mga5

It was still a problem earlier today on fedora 20 which we have now left.  mga4 have not been running lately on this hardware but i suppose nothing have changed, so I guess we can not set it as solved, but actually it works nicely on cauldron.

Production systems on these laptops are now switched to mga5 :)
Comment 7 Morgan Leijström 2014-10-27 22:01:16 CET
Now i think there is another version of hibernation problem: Bug 14392

Status: NEW => RESOLVED
Resolution: (none) => FIXED


Note You need to log in before you can comment on or make changes to this bug.