FS#30330 - [nvidia] 302.17 breaks suspend/resume

Attached to Project: Arch Linux
Opened by Andrej Podzimek (andrej) - Sunday, 17 June 2012, 16:10 GMT
Last edited by Ionut Biru (wonder) - Monday, 06 August 2012, 22:11 GMT
Task Type Bug Report
Category Upstream Bugs
Status Closed
Assigned To Ionut Biru (wonder)
Architecture All
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 11
Private No

Details

Description:
Neither suspend-to-disk nor suspend-to-ram work with 302.17. The machine wakes up somehow, but the GPU remains in an undefined state (black screen, X server frozen).

Additional info:
* package version(s)
The previous version worked just fine.

Steps to reproduce:
Boot a machine with an NVidia Quadro FX 880 and try to suspend/resume. (Don't know whether other models are affected.)
This task depends upon

Closed by  Ionut Biru (wonder)
Monday, 06 August 2012, 22:11 GMT
Reason for closing:  Fixed
Comment by Ionut Biru (wonder) - Sunday, 17 June 2012, 16:13 GMT
report to nvidia. there is nothing we can do here since nvidia is closed source.
Comment by Kevin (anonymous_user) - Sunday, 17 June 2012, 17:23 GMT
Looks like this issue will be fixed in 304.xx:

http://www.nvnews.net/vbulletin/showpost.php?p=2564669&postcount=51
Comment by ajs124 (andy123) - Wednesday, 20 June 2012, 13:44 GMT
same problem here, when will the 304.xx driver be released?
Comment by Soup (Soup) - Wednesday, 04 July 2012, 09:19 GMT
With the /lib removal, those of us who have downgraded (and asked pacman to ignore nvidia packages in order to keep the working version of nvidia) might end up with upgrade problems.

Would the best way to deal with this, when the "final leg of /lib removal" is completed, be to remove nvidia, perform the upgrade and then reinstall the 295.59-1 version of nvidia?
Comment by Ionut Biru (wonder) - Wednesday, 04 July 2012, 13:51 GMT
those of you who have downgraded, should upgrade to the latest release.
Comment by ajs124 (andy123) - Wednesday, 04 July 2012, 14:00 GMT
But we downgraded, because the latest release doesn't work for us.

I think I'll write a PKGBUILD for 295.59 that installs to /usr/lib/modules
looks like there are only 5 or 6 lines to change.
Comment by Ionut Biru (wonder) - Wednesday, 04 July 2012, 14:08 GMT
i don't really care about old and unmaintained stuff.
Comment by Soup (Soup) - Wednesday, 04 July 2012, 14:18 GMT
I agree with you completely, but considering 302.17 killed suspend for everyone with nvidia; the old and unmaintained is the only option for people wanting a working system. That's 37.80% of arch users according to https://www.archlinux.de/?page=PackageStatistics

Edit: + 2.09% for the -lts people
Comment by ajs124 (andy123) - Wednesday, 04 July 2012, 14:23 GMT
@wonder
It's just that this bug isn't fixed upstream yet…
But that's not your problem, you're right.

@Soup, I think installing my modified version before rebooting after the /lib removal will fix this…
or quick and dirty: move the nvidia kernel module manually, that should work too but I think that will cause conflicts when upgrading to a fixed version.
Comment by Ionut Biru (wonder) - Wednesday, 04 July 2012, 14:23 GMT
i still don't care.
Comment by Stefan Brenner (balert) - Sunday, 15 July 2012, 10:20 GMT
I also downgraded the nvidia package and got the described problems caused by the new /lib symlink. For me the following steps worked quite well.

1) remove ignoring the nvidia 302 packages by IgnorePkg from pacman.conf
2) do the full system upgrade initially ignoring glibc according to the solution from the archlinux forum "pacman -Syu --ignore glibc" plus "pacman -Su"
3) after the /lib directory has been moved and the symlink has been created again downgrade nvidia.

Now everything seems to work again like before, using nvidia 295, while 302 still breaks standby.
Comment by ajs124 (andy123) - Sunday, 15 July 2012, 11:00 GMT
The bug was fixed upstream in 304.22
If you don't want to ignore every nvidia package when upgrading, you can install nvidia-ll or nvidia-beta from the aur until 304.xx is in extra.
Comment by Alexandr Fedorchuk (KOL-93) - Wednesday, 25 July 2012, 12:09 GMT
I have the same problem on my notebook.
As experiment I've tried to switch video driver to VESA. And I've got the same problem. I think, it's not only bug of NVIDIA.
Comment by Jeremy M. (jskier) - Wednesday, 25 July 2012, 12:17 GMT
Alex, your problem is probably something else. I work with several Nvidia GPUs and the problem is fixed entirely with beta 304 in AUR.
Comment by Musikolo (Musikolo) - Saturday, 28 July 2012, 17:22 GMT
For those wanting to downgrade nvidia drivers, the most-recent-version dependent packages that you must install is this:

* Nvidia drivers:
- nvidia-295.53-2
- nvidia-utils-295.53-1
- opencl-nvidia-295.53-1

* Latest kernel providing support to the above nvidia drivers:
- linux-3.4.5-1

* If you have virtualbox, the most recent version compatible with the above linux package is this:
- virtualbox-modules-4.1.18-3

Finally, you have to add a line like this in your /etc/pacman.conf file:

- IgnorePkg = nvidia nvidia-utils opencl-nvidia linux virtualbox-modules

Now you can run 'pacman -Syu' seamlessly.

I hope this work around helps anyone while a new version of nvidia driver isn't released.

Regards.
Comment by Daniel Wallace (gtmanfred) - Saturday, 28 July 2012, 17:38 GMT
this is also fixed in nvidia-beta in aur, 304.22 http://www.nvidia.com/object/linux-display-amd64-304.22-driver.html
Comment by Musikolo (Musikolo) - Monday, 06 August 2012, 21:52 GMT
New nvidia driver 304.32 available in Extra repository is fixing this issue.

Regards.

Loading...