Bug 54133 - [radeon][RV620] Resuming from suspend/hibernation randomly fails since kernel 3.5.0
[radeon][RV620] Resuming from suspend/hibernation randomly fails since kernel...
Status: NEW
Product: DRI
Classification: Unclassified
Component: DRM/Radeon
unspecified
x86-64 (AMD64) Linux (All)
: medium major
Assigned To: Default DRI bug account
https://bugzilla.redhat.com/show_bug....
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2012-08-27 21:15 UTC by Loïc Yhuel
Modified: 2013-02-01 11:44 UTC (History)
6 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Loïc Yhuel 2012-08-27 21:15:23 UTC
Since 3.5.0 kernel (currently 3.5.2-3.fc17.x86_64), resuming from suspend or hibernation randomly fails. There is no display, and the screen flashes every few seconds (probably due to the backlight).

Hardware:
Toshiba Satellite Pro A300 laptop
01:00.0 VGA compatible controller: ATI Technologies Inc Mobility Radeon HD 3400 Series

Here are the relevant kernel messages :
[drm] PCIE GART of 512M enabled (table at 0x0000000000040000).
radeon 0000:01:00.0: WB enabled
radeon 0000:01:00.0: fence driver on ring 0 use gpu addr 0x0000000010000c00 and cpu addr 0xffff880134cc6c00
[drm] ring test on 0 succeeded in 0 usecs
[drm] ib test on ring 0 succeeded in 0 usecs
[drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -35!
radeon 0000:01:00.0: GPU reset succeed
[drm] PCIE GART of 512M enabled (table at 0x0000000000040000).
...
Comment 1 Alex Deucher 2012-08-27 21:28:29 UTC
Can you bisect?
Comment 2 Loïc Yhuel 2012-08-27 22:05:43 UTC
I will try, but it can be long as the problem is not systematic.
Comment 3 aaannz 2012-10-27 16:34:27 UTC
I am experiencing same problem. My log repeats this messages almost 4times/second:

[drm] PCIE GART of 512M enabled (table at 0x0000000000040000).
radeon 0000:01:00.0: WB enabled
radeon 0000:01:00.0: fence driver on ring 0 use gpu addr 0x0000000020000c00 and cpu addr 0xffff880220418c00
[drm] ring test on 0 succeeded in 1 usecs
[drm] ib test on ring 0 succeeded in 0 usecs
[drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -35!
radeon 0000:01:00.0: GPU reset succeeded, trying to resume
[drm] probing gen 2 caps for device 8086:29e1 = 2/0
[drm] enabling PCIE gen 2 link speeds, disable with radeon.pcie_gen2=0

HW is desktop pc with GPU
01:00.0 VGA compatible controller: Advanced Micro Devices [AMD] nee ATI RV770 [Radeon HD 4850]

using openSUSE 12.2 with kernel 3.6.3

But I think I didn't encounter this issue back in august when this report was created and I always try to use the latest kernel released. I start to hitting this after upgrading to xorg server 1.13 (default openSUSE has 1.12) and Mesa 9.0 (default openSUSE is Mesa 8.x)
Comment 4 aaannz 2012-11-11 15:21:46 UTC
Noticed that this can be easily triggered by switching between text and X console few times.

My current versions:
# uname -a
Linux Cerberos 3.6.5-10-desktop #1 SMP PREEMPT Wed Oct 31 20:15:15 UTC 2012 (cefb3b0) x86_64 x86_64 x86_64 GNU/Linux
# rpm -q libdrm_radeon1
libdrm_radeon1-2.4.39-94.1.x86_64
Comment 5 aaannz 2013-01-05 17:17:49 UTC
Since upgrade to kernel 3.7.1 (from openSUSE Tumbleweed) I can no longer replicate this. For me this problem appears to be fixed.
Comment 6 Ash 2013-01-18 00:54:19 UTC
I am using kernel 3.7.2-201.fc18.x86_64 (Fedora 18) and no libdrm_radeon and am also experiencing this problem. Problem not present in Fedora 17.
Comment 7 Jean-François Fortin Tam 2013-01-22 15:14:28 UTC
Downstream reports:
https://bugzilla.redhat.com/show_bug.cgi?id=883536
https://bugzilla.redhat.com/show_bug.cgi?id=849347

I'm kinda wondering if this has something to do with 64-bits (probably not, as comment 6 hints that this was not present in Fedora 17 64-bits).
Comment 8 Alex Deucher 2013-01-22 15:29:14 UTC
Can anyone bisect and track down what broke it?
Comment 9 Jean-François Fortin Tam 2013-01-24 14:10:17 UTC
I don't think I have the knowledge to build (and git bisect!) my own driver, however there is some additional clue I can give from https://bugzilla.redhat.com/show_bug.cgi?id=883536 :

the issue does not seem to be strictly related to suspend & resume. It is very easy to trigger the problem in Fedora 18 by booting up to the GDM login screen, switching to another virtual terminal by pressing ctrl+alt+F3, then trying to switch back to GDM by doing ctrl+alt+F1.