Bug 76998 - hang on CEDAR right after log on
Summary: hang on CEDAR right after log on
Status: RESOLVED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Radeon (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium critical
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-04-03 12:25 UTC by tcl_de
Modified: 2014-07-23 06:44 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Xorg log (47.31 KB, text/plain)
2014-04-04 11:43 UTC, tcl_de
no flags Details
dmesg (59.72 KB, text/plain)
2014-04-04 11:51 UTC, tcl_de
no flags Details
possible fix (1.89 KB, patch)
2014-07-07 22:03 UTC, Alex Deucher
no flags Details | Splinter Review

Description tcl_de 2014-04-03 12:25:24 UTC
Using recent kernel versions, my Debian testing machine always hangs after the graphical log in. The X server and lightdm start normally but a few seconds after entering the credentials and pressing log in, the machine hangs. 

Kernel 3.8.1 works perfectly, but the Debian 3.11-2-amd64, 3.12-1-amd64 and 3.13-1-amd64 kernels don't.

In order to ensure that nothing has gone wrong with my installation, I also tried the openSUSE 13.1 KDE Live DVD but it also hangs shortly after the desktop has fully loaded.

My graphics card is a Radeon HD5450 (CEDAR).
Comment 1 Alex Deucher 2014-04-03 20:04:24 UTC
please attach your xorg log and dmesg output.  Can you use git to bisect the kernel and find out what commit caused the regression?
Comment 2 tcl_de 2014-04-04 11:43:00 UTC
Created attachment 96897 [details]
Xorg log
Comment 3 tcl_de 2014-04-04 11:51:19 UTC
Created attachment 96898 [details]
dmesg
Comment 4 tcl_de 2014-04-04 11:59:52 UTC
I did a small "bisection" and installed the 3.9.0-030900-generic, 3.10.0-031000-generic and 3.14.0-031400-generic kernels from the Ubuntu MainlineBuilds.

3.9 works, 3.10 and 3.14 hang.

The attached logs are for the 3.9 kernel.

I'm sorry, I prefer not to do a git bisect because I can only spend a limited amount of time on this bug.
Comment 5 tcl_de 2014-06-27 09:50:52 UTC
The result of the bisect is
d4788db30a1a66255b592dd12613dda80c1443f7 is the first bad commit
commit d4788db30a1a66255b592dd12613dda80c1443f7
Author: Alex Deucher
Date:   Thu Feb 28 14:40:09 2013 -0500

    drm/radeon/evergreen: add support for golden register init
    
    Signed-off-by: Alex Deucher
Comment 6 Alex Deucher 2014-07-01 15:12:30 UTC
(In reply to comment #5)
> The result of the bisect is
> d4788db30a1a66255b592dd12613dda80c1443f7 is the first bad commit
> commit d4788db30a1a66255b592dd12613dda80c1443f7
> Author: Alex Deucher
> Date:   Thu Feb 28 14:40:09 2013 -0500
> 
>     drm/radeon/evergreen: add support for golden register init
>     
>     Signed-off-by: Alex Deucher

Do you think you could go through the golden register arrays for cedar and narrow down which register(s) are problematic?
Comment 7 tcl_de 2014-07-06 21:11:54 UTC
Yes, I think I can.
 
In fact, I already tried and managed to go through the register arrays. 
There is a single problematic register in cedar_golden_registers, namely 0x5cc. The complete corresponding row of the array is 0x5cc, 0xffffffff, 0x00000001 (line 384).
Comment 8 Alex Deucher 2014-07-07 22:03:48 UTC
Created attachment 102392 [details] [review]
possible fix

The attached patch should fix the issue.
Comment 9 tcl_de 2014-07-13 11:11:51 UTC
I can confirm that this patch fixes the issue.
Comment 10 tcl_de 2014-07-23 06:44:54 UTC
I am closing this bug as the fix has landed in kernels 3.16-rc5, 3.15.6, 3.14.13, 3.12.25 and 3.10.49 .


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.