Bug 28627

Summary:

2.6.31.6 is the last kernel where KMS works well on an RV515 card for regular PCI

Product:

DRI

Reporter:

Connor Behan <connor.behan>

Component:

DRM/Radeon

Assignee:

Default DRI bug account <dri-devel>

Status:

RESOLVED WORKSFORME

QA Contact:

Severity:

major

Priority:

medium

Version:

unspecified

Hardware:

x86 (IA32)

OS:

All

Whiteboard:

i915 platform:

i915 features:

Attachments:

Description	Flags
dmesg.good	none
glxinfo.good	none
xorglog.good	none
dmesg.2.6.33.4.bad	none
glxinfo.2.6.33.4.bad	none
xorglog.2.6.33.4.bad	none
dmesg.2.6.33.4.reallybad	none
glxinfo.2.6.33.4.reallybad	none
xorglog.2.6.33.4.reallybad	none
lspci.good	none
2.6.34.terrible	none
dmesg.35	none
glxinfo.35	none
xorglog.35	none

Description Connor Behan 2010-06-20 09:45:54 UTC

After constant regressions, I thought I would post this.

dmesg.good: http://pastebin.com/kL3iXTU3
xorglog.good: http://pastebin.com/n7diwRCL
glxinfo.good: http://pastebin.com/aDajgYQC

dmesg.bad: http://pastebin.com/JSKSD8Dn
xorglog.bad: http://pastebin.com/QdDhfknr
glxinfo.bad: http://pastebin.com/AWabZKN8

dmesg.reallybad: http://pastebin.com/Jc4sqySK
xorglog.reallybad: http://pastebin.com/fzMEJfZH
glxinfo.reallybad: http://pastebin.com/tkJgTgKa

Good is kernel 2.6.31.6 where I get glxgears frame rates of ~300fps. This is all I can expect from a slow PCI bus.

Bad is kenel 2.6.33.4 where the glxgears frame rates are a little slower, maybe 250fps and quitting X fails. Logging out of my desktop environment, Ctrl+Alt+F2, Ctrl+Alt+Backspace, chvt commands, all of these freeze my system.

Reallybad is kernel 2.6.33.4 where glxgears has no consistent frame rate. It jumps around between 5fps and 60fps and all graphical programs are incredibly slow. This is in addition to the zapping issue described above.

I am not sure why 2.6.33.4 is sometimes bad and sometimes reallybad. It might have to do with what USB devices are connected when it boots up.

Please do not tell me to upgrade my userspace programs. I have tried every combination of xorg-server, mesa, xf86-video-ati and libdrm and none of them have any effect. The kernel is the ONLY package that changes how bad this bug is. Also please do not tell me to upgrade the kernel unless you are very sure that the version you are suggesting fixes this bug and you can tell me why it fixes it.

I am not looking for a kernel that works. I KNOW which kernel works. I am trying to understand WHY it is the only one that works and hopefully help you fix this. Thank-you very much.

Comment 1 Connor Behan 2010-06-20 09:55:27 UTC

Created attachment 36383 [details]
dmesg.good

Comment 2 Connor Behan 2010-06-20 09:57:00 UTC

Created attachment 36384 [details]
glxinfo.good

Comment 3 Connor Behan 2010-06-20 09:57:52 UTC

Created attachment 36385 [details]
xorglog.good

Comment 4 Connor Behan 2010-06-20 09:58:22 UTC

Created attachment 36386 [details]
dmesg.2.6.33.4.bad

Comment 5 Connor Behan 2010-06-20 09:58:53 UTC

Created attachment 36387 [details]
glxinfo.2.6.33.4.bad

Comment 6 Connor Behan 2010-06-20 10:00:11 UTC

Created attachment 36388 [details]
xorglog.2.6.33.4.bad

Comment 7 Connor Behan 2010-06-20 10:01:25 UTC

Created attachment 36389 [details]
dmesg.2.6.33.4.reallybad

Comment 8 Connor Behan 2010-06-20 10:02:03 UTC

Created attachment 36390 [details]
glxinfo.2.6.33.4.reallybad

Comment 9 Connor Behan 2010-06-20 10:02:36 UTC

Created attachment 36391 [details]
xorglog.2.6.33.4.reallybad

Comment 10 Connor Behan 2010-06-20 10:04:41 UTC

Created attachment 36392 [details]
lspci.good

Comment 11 Connor Behan 2010-06-20 10:11:02 UTC

Created attachment 36393 [details]
2.6.34.terrible

Comment 12 Connor Behan 2010-06-20 10:12:46 UTC

Comment on attachment 36393 [details]
2.6.34.terrible

2.6.34 is even worse. I will continue to add pastes for new kernels as they come out. I would prefer not to go back and do 2.6.32 because I remember that being just like 2.6.33.4 and compiling it would be a huge pain.

Comment 13 Connor Behan 2010-06-20 10:32:28 UTC

Replacing these will not change a thing but in case you're wondering:

libdrm 2.4.21
xf86-video-ati 6.13.0
mesa 7.8.1
xorg-server 1.8.1

Also my distro patches ONLY the following directories:

drivers/video/logo/
fs/
include/linux/
security/

Comment 14 Alex Deucher 2010-06-21 08:28:12 UTC

Any chance you can use git to bisect what kernel commit is the problematic one?

Comment 15 Connor Behan 2010-06-29 16:48:19 UTC

That's a good idea, if a lengthy one.

Is there a way to see a sequential list of commits in one of the "main" branches? And is there a way to compile only the radeon module if I'm moving from one kernel to a very similar one?

Comment 16 Alex Deucher 2010-06-29 16:53:45 UTC

You can browse commits in via the git web interface:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git
You can also narrow the bisect range by just bisecting drm commits, e.g.,
git bisect start -- drivers/gpu/drm
Although this won't help if the problematic commit is not drm related.

Comment 17 Connor Behan 2010-08-20 19:33:13 UTC

Created attachment 38032 [details]
dmesg.35

Comment 18 Connor Behan 2010-08-20 19:33:44 UTC

Created attachment 38033 [details]
glxinfo.35

Comment 19 Connor Behan 2010-08-20 19:34:09 UTC

Created attachment 38034 [details]
xorglog.35

Comment 20 Connor Behan 2010-08-20 19:35:23 UTC

2.6.35.2 is acting the way 2.6.33 did. Slow graphical apps, inconsistent glxgears frame rate and having to use Alt + SysRq + REISUB every time I log out of X.

Comment 21 Connor Behan 2011-04-13 21:40:10 UTC

This bug largely goes away if I use kernels 2.6.37 and 2.6.38 with the Gallium Radeon/DRI driver. In fact the glxgears framerates I get that way are slightly better. Some things to note are that the framerates become awful again if I turn "EXAPixmaps" "off" and that I still have trouble logging out of X. This is surely a topic for another bug.

Thanks for all the work you've been doing!

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.