Discussion: the ati/radeon open source drivers & radeon repository

MuPuF · 2010-02-13 11:11:03

Neuro wrote:

Ok, everyone using open source radeon drivers, with mesa from git and are using a 2.6.33-rc kernel or 2.6.33-git kenel, please listen:
We're trying to trace a bug which causes kernel hangs (http://bugzilla.kernel.org/show_bug.cgi), but can't get anything that's 100% per cent reproducible. Things that are known to cause the bug sometimes are:
1. Playing flash games (namely Hex Empire)
2. Using KWin present windows
3. Using thumbnails in KDE's taskbar.
4. Other stuff.
Symptoms:
- Black screen
- System unresponsive to keyboard or SSH
- Monitor states "no signal"
If you experience such a crash, please post your experience here or to the bug comments. If you are able to deterministically reproduce the bug, we'd all be gratefull.

Do you get this behavior using KMS or UMS ?

I'm still stuck in UMS and can't move to KMS because of the screen corruption I get in KMS. I own an HD4770 (rv740) and this card is not really supported. For instance, I'm still stuck to opengl 1.4.

Vi0L0 · 2010-02-13 11:45:56

Neuro wrote:

Right, apart from the random crash bug, what about slow flash video movies?
For example:
http://vimeo.com/6194911
This works fine with intel graphics card, but the radeon driver plays it (non-smoothly) and causes firefox to hardly respond. Anyone with a similar problem?

I can feel a little lag when playing this as site content, also flash buttons are showing after some time and lagging, one of my cores (2,4GHz) catches 100% usage, ff responds with some very very small lag, but on fullscreen everything is smooth. Anyhow it's my first laggy flash video.
x86_64, 2.6.33-rc7-git1, hd4850, KMS on, tested on both [nightly] 12.02 and own mesa-full 31.01
umm... im using ff 3.5.7 with nspluginwrapper-flash (32-bit flash plugin) - which was always ~2x faster than 64bit flashplugin for me
also my ff is locked in a sandbox, called sandfox http://bbs.archlinux.org/viewtopic.php?pid=703283 which i can recommand

Neuro · 2010-02-13 12:51:00

@MuPuF
I'm talking about KMS.
I've had corruption in Kwin, but it's fixed in 2.6.33-rc7. You could give it a try.

Vi0L0 wrote:

I can feel a little lag when playing this as site content, also flash buttons are showing after some time and lagging, one of my cores (2,4GHz) catches 100% usage, ff responds with some very very small lag, but on fullscreen everything is smooth. Anyhow it's my first laggy flash video.
x86_64, 2.6.33-rc7-git1, hd4850, KMS on, tested on both [nightly] 12.02 and own mesa-full 31.01
umm... im using ff 3.5.7 with nspluginwrapper-flash (32-bit flash plugin) - which was always ~2x faster than 64bit flashplugin for me
also my ff is locked in a sandbox, called sandfox http://bbs.archlinux.org/viewtopic.php?pid=703283 which i can recommand

This is exactly what I'm talking about. Laggy Firefox, laggy buttons on flash video. Everything is ok in full screen. All this happens with composite both on and off.

Black Mage · 2010-02-13 12:58:10

neuro i think this bug also happened to me every time i tried to run vegastrike, so check it out.
my 3d doesnt work right now so i cant confirm.

Neuro · 2010-02-13 13:04:36

Black Mage wrote:

neuro i think this bug also happened to me every time i tried to run vegastrike, so check it out.
my 3d doesnt work right now so i cant confirm.

You mean the crash-freeze bug?
Also:
1.What kernel were you using?
2. What packages should I install?
3. At which point it had crashed for you?

EDIT:
I couldn't reproduce the ug in vegastrike.

@Vi0L0
As for the flash video bug, switching from KMS to UMS fixed the problem. Everything is smooth in UMS (no Firefox lag, no button lag). I think about reporting a bug, but I don't know where. Since it's KMS, would it fit in the kernel bugzilla?

Last edited by Neuro (2010-02-13 14:41:30)

Vi0L0 · 2010-02-13 13:21:32

Neuro wrote:

All this happens with composite both on and off.

Same here.

Neuro wrote:

Since it's KMS, would it fit in the kernel bugzilla?

I think so.

Neuro · 2010-02-13 14:54:09

Vi0L0 wrote:

I can feel a little lag when playing this as site content, also flash buttons are showing after some time and lagging, one of my cores (2,4GHz) catches 100% usage, ff responds with some very very small lag, but on fullscreen everything is smooth. Anyhow it's my first laggy flash video.
x86_64, 2.6.33-rc7-git1, hd4850, KMS on, tested on both [nightly] 12.02 and own mesa-full 31.01
umm... im using ff 3.5.7 with nspluginwrapper-flash (32-bit flash plugin) - which was always ~2x faster than 64bit flashplugin for me
also my ff is locked in a sandbox, called sandfox http://bbs.archlinux.org/viewtopic.php?pid=703283 which i can recommand

I've reported the bug:
http://bugzilla.kernel.org/show_bug.cgi?id=15293

Please post there if you're experiencing similar problems.

handy · 2010-02-14 00:50:48

Perry3D wrote:

What kernel are you using at the moment? IRQ support has been introduced with kernel 2.6.33. And that is just a warning.
And your output says OpenGL 2.0 (shading language is 1.10).
Don't use glxgears as a benchmark!

I installed the PST yesterday.

Re, your kernel question I'm using kernel26-git 20100117-1, though the following Phoronix Test Suite identifies it as .32?

& 20100129 git packages.

I ran a test this morning, with Firefox running 13 tabs open, 2 torrents were happening with Transmission, Worker was open, & there were 4 tabs going in Sakura; one of them being htop. The machine had been running for 50 hours plus, & there is still a memory leak in Ghostery, so I had a little more than half of the 1GB of RAM available. Both CPUs were just about idling before the tests though.

When I named the test 1st.phoronix.test, I was still under the impression that the test results were for my own benefit only. If I had of known that the results could be uploaded, I would have named it more appropriately.:rolleyes:

http://global.phoronix-test-suite.com/i … 26987-2834

Last edited by handy (2010-02-14 00:55:29)

korpenkraxar · 2010-02-14 01:09:56

KDE SC 4.4 test case:

i) Use compositing & Present Windows as window switching effect

ii) Start 4 Konsole terminal windows

iii) Press ALT + TAB to show all four windows at the same time (down-scaled)
iv) release TAB, but keep holding ALT so that you do not fall back to the regular desktop view. You should see the four terminals.
v) Move the mouse pointer in a circle (I did this anti-clockwise) over each window, so that they take turns being highlighted. Try complete five or six circles. If you suffer from the same kind of bug I do, your system is expected to crash here.
vi) Release ALT and return to the regular desktop view. Go to iii)

I just repeated this crash five times now, but not always in the first loop. It can crash almost immediately or take anything up to four loops (my record so far).

Have fun!

Neuro · 2010-02-14 07:49:49

korpenkraxar wrote:

KDE SC 4.4 test case:
i) Use compositing & Present Windows as window switching effect
ii) Start 4 Konsole terminal windows
iii) Press ALT + TAB to show all four windows at the same time (down-scaled)
iv) release TAB, but keep holding ALT so that you do not fall back to the regular desktop view. You should see the four terminals.
v) Move the mouse pointer in a circle (I did this anti-clockwise) over each window, so that they take turns being highlighted. Try complete five or six circles. If you suffer from the same kind of bug I do, your system is expected to crash here.
vi) Release ALT and return to the regular desktop view. Go to iii)
I just repeated this crash five times now, but not always in the first loop. It can crash almost immediately or take anything up to four loops (my record so far).
Have fun!

I'm using KDEmod 4.4 core here, Ati 3450, radeon KMS (mesa stuff from 20100212), effects on (OpenGL, thumbnails only for shown windows, texture from pixmap, bilinear, direct rendering, use vsync, smooth scaling) and can't reproduce it.

I've created 4 semi-transparent konsole windows, przsed Win+Esc (my binding for present windows) and did circles around it many times. I've tried it 10 times with 8-10 cricles per each try, nothing happened.

You could post it though in the kernel bugtracker. Maybe the kernel devs have a way of acquiring some crash logs and the fact that you've got it reproducible (thoug, only for your machine) may be helpful.

EDIT:
I'm sorry, you were right. I reverted to UMS bacause of the crashes. While testing in KMS I indeed were able to reproduce the crash twice. Thank you very much.

Last edited by Neuro (2010-02-14 08:33:35)

xheyther · 2010-02-15 16:20:29

korpenkraxar wrote:

KDE SC 4.4 test case:
i) Use compositing & Present Windows as window switching effect
ii) Start 4 Konsole terminal windows
iii) Press ALT + TAB to show all four windows at the same time (down-scaled)
iv) release TAB, but keep holding ALT so that you do not fall back to the regular desktop view. You should see the four terminals.
v) Move the mouse pointer in a circle (I did this anti-clockwise) over each window, so that they take turns being highlighted. Try complete five or six circles. If you suffer from the same kind of bug I do, your system is expected to crash here.
vi) Release ALT and return to the regular desktop view. Go to iii)
I just repeated this crash five times now, but not always in the first loop. It can crash almost immediately or take anything up to four loops (my record so far).
Have fun!

You should post that on the kernel bugzilla since it would allow a dev to reproduce while profiling and/or debugging.

korpenkraxar · 2010-02-15 19:14:41

xheyther wrote:

You should post that on the kernel bugzilla since it would allow a dev to reproduce while profiling and/or debugging.

It is already posted.

Cory · 2010-02-16 09:29:28

After following all the procedure I get the following error

(EE) AIGLX error: dlopen of /usr/lib/xorg/modules/dri/r600_dri.so failed (/usr/lib/xorg/modules/dri/r600_dri.so: undefined symbol: _mesa_framebuffer_renderbuffer)
(EE) AIGLX: reverting to software rendering
(II) AIGLX: Screen 0 is not DRI capable
(EE) AIGLX error: dlopen of /usr/lib/xorg/modules/dri/swrast_dri.so failed (/usr/lib/xorg/modules/dri/swrast_dri.so: undefined symbol: _mesa_framebuffer_renderbuffer)
(EE) GLX: could not load software renderer

Someone know where is the problem? KMS works wonders, but I have no 3d acceleration

Perry3D · 2010-02-16 17:38:21

Update your drivers: Tear-free opengl is back. And the xv - acceleration is faster: Phoronix: Radeon KMS Gets Faster X-Video Support
And: http://cgit.freedesktop.org/xorg/driver … d5278a1118

@Cory: Looks like you have compiled the packages in the wrong order.

It should be something like this:

libdrm ----------|
                 |---------> mesa -------> xf86-video-ati
dri2proto -------|

venky80 · 2010-02-16 18:30:28

does anyone know of a repository which maintains the latest kernel26-git?

Vi0L0 · 2010-02-16 19:28:51

Did anybody noticed something like this while playing Sauerbraten: Cube2?:
(edit: and by "while playing" i mean "all the time")

Kinda Cool i know , but i believe that's not what sauerbraten devs wanted to achieve
Im getting those bugs since 2 weeks now, although before that time there were no problems - so it's regression.
Tested with many versions of mesa-full and [nightly].
Those screens was taken today with today's mesa-full git's builds. x86_64, 2.6.33-rc8, hd4850, both kms and ums, composite on/off, i've even removed ~/.sauerbraten
fps is so low cuz of high settings and shaders (yep - tested also on lowests settings).

Last edited by Vi0L0 (2010-02-16 20:12:05)

Cory · 2010-02-16 20:51:10

I have a HD4670 graphics card. Update all packages and install them in the correct order but still keep getting the same error, no matter whether KMS enables or disables the error is the same.

Install the packages in this order:
kernel26-firmware-git+kernel26-git+linux-api-headers+radeon_ucode+libdrm-git+glrptoto-git+dri2proto-git+libgl-git+ati-dri-git+mesa-git+xf86-vide-ati-git add my user to the group video and reboot

When I run glxinfo | grep-i opengl gives me the following error
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Error: couldn't find RGB GLX visual or fbconfig
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".
Xlib: extension "GLX" missing on display ":0.0".

These are my outputs from dmesg and Xorg.0.log

http://pastebin.com/m4503cc3f
http://pastebin.com/m1f3489b4

I'm running the system without Xorg.conf if I try to generate with Xorg-configure the system hangs completely, any help is appreciated thanks

Perry3D · 2010-02-16 20:56:07

I suggest to use mesa-full instead of ati-dri-git, libgl-git and mesa-git. It produces less problems as the splitted packages.

/edit: Are you using the packages from the nightly repo?

Last edited by Perry3D (2010-02-16 20:56:48)

Cory · 2010-02-16 21:18:49

mesa-full give me error when i try to install It has conflicts with freeglut. So, use mesa-split to install all packages at once

No, I compile packages myself

Vi0L0 · 2010-02-16 21:45:03

Cory wrote:

mesa-full give me error when i try to install It has conflicts with freeglut. So, use mesa-split to install all packages at once
No, I compile packages myself

mesa-full provides freeglut

Comment by: bjacob on Sat, 23 Jan 2010 17:41:12 +0000
The provides= line is outdated: it should also mention that it provides freeglut. Indeed, other packages like jasper are now requiring freeglut, not glut. So add "freeglut=${_realver}".

after building you may safely remove freeglut from your arch, and install mesa-full, packages will no longer complain about missing freeglut

Last edited by Vi0L0 (2010-02-16 21:45:25)

Vi0L0 · 2010-02-16 22:11:11

Is there any way to force vertical refresh in KMS?
Example: i want to have 75Hz all the time - i may handy set it in KDE, but when i will try to switch to tty* (where refresh=60Hz) i must to wait some time. It maybe strange but it looks like my kde is working a little bit faster with higher refresh rate.
Also i haven't got xorg.conf

Cory · 2010-02-16 22:21:29

Thank you very much to the two just reinstall everything and switched mesa-split for mesa-full and works great thanks again. At last I have 3d acceleration, KMS and opengl=tearfree

Perry3D · 2010-02-16 22:39:30

Vi0L0 wrote:

after building you may safely remove freeglut from your arch, and install mesa-full, packages will no longer complain about missing freeglut

Did you also add glew to the provides line? There is a conflict.

Vi0L0 · 2010-02-16 22:39:54

Perry3D wrote:

Update your drivers: Tear-free opengl is back. And the xv - acceleration is faster: Phoronix: Radeon KMS Gets Faster X-Video Support
And: http://cgit.freedesktop.org/xorg/driver … d5278a1118

So true, tear-free opengl is kinda cool, no more video flickering even in window
But this could be very annoying when you try to play fps with it - even with 75Hz/fps its too slow, everything seems to flow... although it's still faster than catalyst's v-sync!
Is there a way to enable/disable this kind of v-sync dynamically (without reboot)?

-- sorry for maybe lame questions - i performed quick search for this but found nothing (maybe i dont know how to arrange right question)...

Edit: Awww even flash video when playing it as a site content (windowed) are tear-free - sweet, i haven't seen this before . Tearing is back after switching it to fullscreen - but i guess its adobe's fault...

Last edited by Vi0L0 (2010-02-16 23:12:34)

Perry3D · 2010-02-16 23:31:20

Vi0L0 wrote:

Perry3D wrote:
Update your drivers: Tear-free opengl is back. And the xv - acceleration is faster: Phoronix: Radeon KMS Gets Faster X-Video Support
And: http://cgit.freedesktop.org/xorg/driver … d5278a1118
So true, tear-free opengl is kinda cool, no more video flickering even in window
But this could be very annoying when you try to play fps with it - even with 75Hz/fps its too slow, everything seems to flow... although it's still faster than catalyst's v-sync!
Is there a way to enable/disable this kind of v-sync dynamically (without reboot)?
-- sorry for maybe lame questions - i performed quick search for this but found nothing (maybe i dont know how to arrange right question)...
Edit: Awww even flash video when playing it as a site content (windowed) are tear-free - sweet, i haven't seen this before . Tearing is back after switching it to fullscreen - but i guess its adobe's fault...

I think you can manually set the resolution with an option for the radeon module. Something like video=.... Maybe you can set the vertical refresh rate also.
I have no idea if you can disable vsync. Maybe it is application-based.
About flash videos: i think it is the fault of your window manager. If you use compiz try diabling "unridirect_fullscreen_videos" in ccsm/general.

Arch Linux

#826 2010-02-13 11:11:03

Re: Discussion: the ati/radeon open source drivers & radeon repository

#827 2010-02-13 11:45:56

Re: Discussion: the ati/radeon open source drivers & radeon repository

#828 2010-02-13 12:51:00

Re: Discussion: the ati/radeon open source drivers & radeon repository

#829 2010-02-13 12:58:10

Re: Discussion: the ati/radeon open source drivers & radeon repository

#830 2010-02-13 13:04:36

Re: Discussion: the ati/radeon open source drivers & radeon repository

#831 2010-02-13 13:21:32

Re: Discussion: the ati/radeon open source drivers & radeon repository

#832 2010-02-13 14:54:09

Re: Discussion: the ati/radeon open source drivers & radeon repository

#833 2010-02-14 00:50:48

Re: Discussion: the ati/radeon open source drivers & radeon repository

#834 2010-02-14 01:09:56

Re: Discussion: the ati/radeon open source drivers & radeon repository

#835 2010-02-14 07:49:49

Re: Discussion: the ati/radeon open source drivers & radeon repository

#836 2010-02-15 16:20:29

Re: Discussion: the ati/radeon open source drivers & radeon repository

#837 2010-02-15 19:14:41

Re: Discussion: the ati/radeon open source drivers & radeon repository

#838 2010-02-16 09:29:28

Re: Discussion: the ati/radeon open source drivers & radeon repository

#839 2010-02-16 17:38:21

Re: Discussion: the ati/radeon open source drivers & radeon repository

#840 2010-02-16 18:30:28

Re: Discussion: the ati/radeon open source drivers & radeon repository

#841 2010-02-16 19:28:51

Re: Discussion: the ati/radeon open source drivers & radeon repository

#842 2010-02-16 20:51:10

Re: Discussion: the ati/radeon open source drivers & radeon repository

#843 2010-02-16 20:56:07

Re: Discussion: the ati/radeon open source drivers & radeon repository

#844 2010-02-16 21:18:49

Re: Discussion: the ati/radeon open source drivers & radeon repository

#845 2010-02-16 21:45:03

Re: Discussion: the ati/radeon open source drivers & radeon repository

#846 2010-02-16 22:11:11

Re: Discussion: the ati/radeon open source drivers & radeon repository

#847 2010-02-16 22:21:29

Re: Discussion: the ati/radeon open source drivers & radeon repository

#848 2010-02-16 22:39:30

Re: Discussion: the ati/radeon open source drivers & radeon repository

#849 2010-02-16 22:39:54

Re: Discussion: the ati/radeon open source drivers & radeon repository

#850 2010-02-16 23:31:20

Re: Discussion: the ati/radeon open source drivers & radeon repository

Board footer