[RV350 9600] [Needs quirk] X freezes in drmCommandWrite() when Compiz is enabled

I'm adding only Xorg.0.log because Xorg.0.log.old appears to be empty.

Looks good to me, I'm coinfirming

I got another backtrace, almost identical (from what I remember, where did it go?), although I'm running Gutsy on a X800 (not a mobility)

lspci -vvnn

I've got a similar problem with my almost identical ThinkPad T42 / Mobility Radeon 9600 chip.
Well, with the latest version of the fglrx driver compiz runs perfectly well and the machine is also able to suspend properly. But however, when I try to resume - with compiz previously enabled - I a get a grey and somewhat speckled slim bar on the lower right of the screen with some artifacts around the mouse cursor. The mouse can be moved but nothing else will work - compiz desktop will just not awake a second time!
When I previously disable compiz and then set my machine in suspend mode, this grey bar shows up as well, but only for a couple of milliseconds and then the desktop appears to start normally.
There is another problem, I don't know if it is related to this one: when I log-off and try to lag back on (with compiz enabled) X freezes totally and the only thing I can do is a hard reboot! This happens every time I try to start X for a second time.
I can provide you with more information - configuration files and logs - just say what you need and where I can find it!

I just tried to follow the step in https://wiki.ubuntu.com/DebuggingXorg, but when I try to restart X it freezes, and when I pgrep Xorg, I get two process numbers. If I try to attach process No 1, I get the following error message:

This GDB was configured as "i486-linux-gnu"...
Using host libthread_db library "/lib/tls/i686/cmov/libthread_db.so.1".
(gdb) attach 5488
Attaching to program: /usr/bin/Xorg, process 5488
warning: process 5488 is a cloned process
/build/buildd/gdb-6.7.1/gdb/linux-nat.c:979: internal-error: linux_nat_attach: Assertion `pid == GET_PID (inferior_ptid) && WIFSTOPPED (status) && WSTOPSIG (status) == SIGSTOP' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.

If I try to attach the second process ID, the computer freezes again!

I have to add to my previous post, that when I log off and try to log back in, the computer seems to freeze once the audio gets initialized, because I hear a freezing typical audio loop (just like "zzzzzz"). I don'T know if this helps.

Do I just have to wait for the release of Hardy, or is this a problem related to the proprietary fglrx driver? Is this a problem of my installation, or did you have suimilar problems on your machines running hardy+fglrx(+ati mobility radeon graphics)???

For info : I have the same bug with radeon mobility X700 : enabling compiz freezes X (but the movement of the mouse cursor) after a few minutes/hours (randomly), disabling compiz solve the problem.

I'll try to find the log/debug info and attach them (no promises, Linux newbie)

These problems seem to be fixed in hardy, at least when using the ati driver. I suspect fglrx users will still have issues but that's fglrx for you.

Is anyone still having this problem in hardy while using the 'ati' driver?

No, logout/login and suspend/resume seem to work, but with the ati-driver everything is white with no textures while all compiz effects seem to work (at least slowly)...
So what are we - the (newbie) users supposed to do? I guess we just have to wait once again - maybe once in a while everything will just work out of the box with ubuntu. I love it already but its not yet enough to be a full replacement for the proprietary stuff:'-(

> Is anyone still having this problem in hardy while using the 'ati' driver?

I'm sorry I can't test this before next week.
Isn't the next alpha release around the corner anyway?
I'm going to report back on a fresh install of Hardy Alpha6 then.

We may end up just blacklisting all people using a laptop with the 'ati' driver since other bugs I read after this one show that people are still having problems.

Please, not all.
I have a laptop with open source "ati" driver and no freezes at all
with compiz. Never. (IGP 340M)
Just a problem with suspend but it's not related to compiz.

2008/3/5, Travis Watkins <email address hidden>:
> We may end up just blacklisting all people using a laptop with the 'ati'
> driver since other bugs I read after this one show that people are still
> having problems.
>
>
> --
> X freezes when Compiz is enabled on ATI Radeon Mobility
> https://bugs.launchpad.net/bugs/195051
> You received this bug notification because you are subscribed to compiz
> in ubuntu.
>

Hi all,
when switched to proprietary ati drivers, I got no more freezing on my ATI Mobility Radeon X600 laptop.
Still compiz-fusion was working much better without proprietary drivers: now effects seem to be slightly less rapid and, most of all, I'm subject to the slow-scrolling bug which is very very annoying.
I have the feeling that with such proprietary drivers some computation are no more in charge of the GPU and are processed by the real CPU, bringing more slowness over all the system.
Sorry, but let me say that it's very annoying thinking that with the kubuntu 7.04 beryl was working great and that it is since almost one whole year that I cannot use such great effects I got used - addicted - to.

Is anyone still having this problem in hardy while using the 'ati' driver?

Yes. I use the open-ATI driver with X700 and if I use any 3D application (compiz or small 3D game) I get desktop freezes with mouse movement, so it is not related to Compiz only. It smells of some 3D buffer overun perhaps in the ATI driver, just guessing :-)

We may end up just blacklisting all people using a laptop with the 'ati' driver since other bugs I read after this one show that people are still having problems.

I certainly hope not. Limiting the use of Ubuntu to fix various problems is hopefully not the way forward. I mean what about next time a problem of this size occurs, do we then blacklist our way out again. Hopefully the problem will be analysed and fixed before the official release of Hardy :-)

I think a lot of us first experienced the problem in Oct. 07 just before the official release of Gutsy, when both Compiz and the 'ati' driver was updated. It should be possible to go back and see the few changes to better isolate the current problem. I can easily reproduce the problem especially with Compiz enabled and I'd be happy to do help with any tracing or bug catching is necessary :-)

>We may end up just blacklisting all people using a laptop with
>the 'ati' driver since other bugs I read after this one show that
>people are still having problems.

This is frightening. I can't imagine any situation (unless you're Apple) where a valid "fix" would be to blacklist hardware that is otherwise completely supported and functional.

As Arngren said, I'd be glad to help with any tracing that is necessary to find a fix. I can't stand to think that blacklisting hardware/software combinations is a valid bug fix.

I'm not a driver developer and I don't even have ATI hardware. If you can find a fix by all means, go for it. We'd be happy to get compiz working on these systems. If not, the only other answer is to make sure people don't have these problems in the default install by blocking compiz from starting.

This has been broken for two releases now, I don't think we're going to get a fix in the next month.

The ATI X700 hardware has previously been blacklisted, but only temporarely - If that is the suggestion, then I don't mind, as long as a permanent fix, where compiz is fully functional, is the aim.

I also believe from the various bug reports on this issue, that Compiz is not part of the problem. It is most likely the ATI driver running 3D which causes the X server to freeze. Compiz just happens to use a lot of 3D effects and thereby provoke the problem faster than any other application.

Perhaps the responsibility of this bug should also be considered by the ATI driver team? Just an idea to make the bugfixing more efficient :-)

This bug is assigned to the ati driver, I'm just talking about what we can do to make sure most users don't have this problem as compiz will probably be the only 3D application they would run.

Of course a permanent fix would be preferred but that doesn't seem possible at this point.

I can confirm the bug is still there on a fresh install of Alpha6 with all updates.

I started compiz via SKIP_CHECKS and was able to use it for some minutes, then X froze and this time I couldn't even use SysRq to communicate with the kernel in order to shut it down.

I have to say I'm surprised there hasn't been any feedback from the people in charge. Is nobody working on this?
Has it been reported upstream?

I would be glad to help but it seems like there's very little interest in working on this.
Anyway... I did some investigation myself and installed Feisty again as well as Debian testing on this machine. Both are running fine with Compiz and the ati/radeon driver.
But both are using version 1:6.6.3-2 of xserver-xorg-video-ati so that may be no surprise after all.

If there's anything I can do to help investigate some more, please let me know.

I have the same problem with ati rv515. Hardy Heron alpha 6

I have this one since gutsy (feisty was working fine) and I was hoping that for a LTS something will be done to correct it. It seems that I'll have to try another distribution for my next upgrade.

Nicolas: This is not something Ubuntu has broken, it's a regression in the ati driver.

Ok so the choice is going back to an old driver very buggy and slow, use a new one also buggy or use the proprietary buggy one...

Arglll I hate ATI!

Nick, thanks for collecting a backtrace, however it looks like you did not install the debug symbols (the -dbg or -dbgsym packages), as you can see in the following where there are ?? instead of the functions. Please install them and collect another backtrace. Also, please do it by '(gdb) backtrace full' which collects more data than the default backtrace. I'll up the priority on this bug, thanks for filing it separate from 108527 which has gotten too crowded with unrelated issues.

For everyone else, I want to focus this bug on Nick's issue, which is with lockups with the ATI 9600. Unless you have the same card and exact symptoms, or unless you have *exactly* the same backtrace as him, please file a separate bug with backtrace and Xorg.0.log. Even if these turn out to be the same underlying issue it will make investigation easier to have them organized this way. We can always dupe them in LP if they end up being the same issue.

Program received signal SIGINT, Interrupt.
[Switching to Thread 0xb7c3b6b0 (LWP 6074)]
0xb7f63410 in __kernel_vsyscall ()
(gdb) backtrace full
#0 0xb7f63410 in __kernel_vsyscall ()
No symbol table info available.
#1 0xb7db1929 in ioctl () from /lib/tls/i686/cmov/libc.so.6
No symbol table info available.
#2 0xb7b3ea8b in drmCommandWrite () from /usr/lib/libdrm.so.2
No symbol table info available.
#3 0xb2dee3c2 in ?? () from /usr/lib/dri/r300_dri.so
No symbol table info available.
#4 0x00000008 in ?? ()
No symbol table info available.
#5 0x00000017 in ?? ()
No symbol table info available.
#6 0x083e0754 in ?? ()
No symbol table info available.
#7 0x00000004 in ?? ()
No symbol table info available.
#8 0x083e7010 in ?? ()
No symbol table info available.
#9 0x00000001 in ?? ()
No symbol table info available.
#10 0x083e0700 in ?? ()
No symbol table info available.
#11 0xb2fc3e1c in ?? () from /usr/lib/dri/r300_dri.so
---Type <return> to continue, or q <return> to quit---
No symbol table info available.
#12 0x083e7010 in ?? ()
No symbol table info available.
#13 0x083df750 in ?? ()
No symbol table info available.
#14 0xbfc4bb98 in ?? ()
No symbol table info available.
#15 0xb2e139c8 in _mesa_Finish () from /usr/lib/dri/r300_dri.so
No symbol table info available.
Backtrace stopped: frame did not save the PC

Sorry, I see you did do backtrace full - great. So you just need the debug symbols installed. drmCommandWrite() is already a useful hint - evidently this has something to do with libdrm?

Bryce,

thanks for looking into this.
I did have xserver-xorg-core-dbgsym installed by the time I did the backtrace.

Now I installed xserver-xorg-core-dbg (instead) as well as xserver-xorg-video-ati-dbg

$ dpkg -l | grep dbg
ii xserver-xorg-core-dbg 2:1.4.1~git20080131-1ubuntu5
ii xserver-xorg-video-ati-dbg 1:6.8.0-1

Unfortunately it doesn't make a difference (see attachment).

Do you have an idea what dbg/dbgsym packages are missing?

Just a heads-up: I filed a similar bug (at Bryce's suggestion) at https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/206845

I'm running a similar setup (ThinkPad with ATI Radeon Mobility 9000), but I'm getting some different output in the backtrace. I've also discovered that the freeze with mouse activity can be reproduced on my system by switching to the "Braid" screensaver in Gnome Screensaver Preferences. Nick, could you give it a try?

When I switch to Braid, everything but the mouse locks up. Strangely, the system will recover if I choose a different screensaver (even though the system appears unresponsive to clicks). This freeze will persist anywhere from 5 minutes to an hour and then POOF, the system is again responsive with the new screensaver selected. I should also warn that I get broken pipe if I select random screensaver and Braid happens to come up (that's how I stumbled across the problem with Braid in the first place).

Anyway, all the backtraces I managed to collect during the "freeze" ended with problems in Dispatch() in dix/dispatch.c and dix/main.c. No problems with drmCommandWrite() here.

Kris, thanks for taking the time to work on this.
Like I said in the upstream report I filed over at fd.o (http://bugs.freedesktop.org/show_bug.cgi?id=15096) this issue is not limited to Compiz but 3D apps in general.

I tried your suggestion and I do not have the same problem with the Braid screensaver. I can watch it in the preview just fine for several minutes (although I noticed it does not work properly in fullscreen – no freeze, though).
It's not a 3D screensaver anyway, so I guess your problem is not exactly the same. X never recovered here from the freeze also (at least not within an hour, which was the longest time I waited one time).

But I tried the Bubble3D screensaver next to it and the X almost instantly froze.
Interestingly the backtrace is more verbose this time (attached).

Some more testing, this time with 'extremetuxracer' (3D game from Ubuntu sources).

This game works fine on my desktop machine (Radeon 9600 XT) with the open radon driver.
On the Thinkpad the game appears to freezes after 2-5 seconds. The music continues, though.

The backtrace is similar to the last one from the 3D screensaver (attached).
I tried to kill the game remotely which only worked via 'kill -9'
The screen stays the same but the music stops, so obviously it's not the game that froze in the first place but X.

Then I tried to restart X via:
$ sudo /etc/init.d/gdm restart
 * Stopping GNOME Display Manager... [ OK ]
 * Starting GNOME Display Manager... [ OK ]

Although it looks good in the remote shell, the screen stays the same and there's nothing I can do.
Finally 'sudo shutdown -h now' did work.

Perhaps is it related not only to ATI but also to Thinkpad. The majority of people touch bit it does have a laptop from thsis brand. (Mine is a Thinkpad Z60m).

I know that you cannot install on windows the ATI catalyst version but you have to use the Lenovo one's...

ps: by the way I saw exactly the same things than Nick. I have been able to connect by ssh on the laptop but nothing to restart X worked.

I have a desktop system running an ATI X600 that previously exhibited the all-but-mouse freeze, I'll run the debugger on that system to see if it produces anything different than the ThinkPads.

Hi,

I also a problem with my ati card : X hangs shortly after login, with our without compiz. I solved it by adding
Option "AGPMode" "1"
in my xorg.conf. However, it's not good, as it really slows graphics display. I reported a bug here :

http://bugs.freedesktop.org/show_bug.cgi?id=14892

could anyone try this option to see if those bugs are related?

kede, a quirk for your setup to use AGPMode 1 is included in 1:6.9.0+git20080826.a3cc1d7a-2ubuntu1.

Thank you very much!

I happened to install Windows XP and used this opportunity to check the graphic card with the tool "powerstrip".
Here are the readings:

Interestingly, kede's Windows XP uses AGP x4 (and even FastWrites enabled). One might start to suspect there is a bug in Xorg AGP handling, even though upstream so far says it's a hardware setting incompatibility (card vs bridge).

Hi Nick, it's been some time since we last heard from you - are you still interested in this bug? If so, it'd be great if you could re-test it against intrepid, and provide the info mentioned above. Otherwise, we'll close this as expired; the bug report has wandered a bit from topic so would be nice to get closed out one way or the other.

Hi,

I have just found this bug as until now my ati t60 worked perfectly with ubuntu/compiz. Today I installed Intrepid and I was pretty shoked to find that after I enabled compiz it froze completely in a few minutes of working. No ctrl alt del or ctrl alt backspace would work only a button shutdown. Now I am using no desktop effects and it's a pitty.

Can you advice what else info is needed perhaps I can help more... .

realthor, please see https://wiki.ubuntu.com/X/Quirks#ATI%20AGP%20Mode%20Quirk and file a new bug with Xorg.0.log attached.

We're closing this bug since it is has been some time with no response from the original reporter. However, if the issue still exists please feel free to reopen with the requested information. Also, if you could, please test against the latest version of Ubuntu, since this confirms the bug is one we may be able to pass upstream for help.

Just wanted to report that my problem ( which was similar to this one, but not the same) went away in Intrepid and I am now enjoying Compiz on my ATI Radeon 9600 without any problems. So thank you to all the contributers who are ironing out the ATI issues! I am enjoying my Linux more and more!

Hi Bryce and all,

I'm very sorry that I am reporting so late on this but I didn't have the time to upgrade to Intrepid on my Thinkpad.
Now that I did, I am sorry to report that the freezing problem still exist.

BUT there is a workaround for anyone who's affected (pointed ou to me by Alex Deucher):
You need to set AGP mode to 1 in your Xorg configuration file.

Just add this line to the Section "Device" of your /etc/X11/xorg.conf

 Option "AGPMode" "1"

I have added my xorg.conf for reference.

Unfortunately, it seems there can not be done much about this problem by Ubuntu.
Quoting Alex Deucher:
"There are just too many combinations of agp cards and chipsets that only work at specific settings (some only work at 1x, others only work at 4x or 8x, etc.). We've tried both defaults in the driver, but someone's combination always breaks."

I have the same problems with a
01:00.0 VGA compatible controller: ATI Technologies Inc RV350 [Mobility Radeon 9600 M10]
I have not yet tested the AGPMode 1 option.

Nick, can you please see https://wiki.ubuntu.com/X/Quirks#ATI%20AGP%20Mode%20Quirk and post the required information here? You probably have given most of the information already, but since this bug report has gotten so long and untidy it would be nice if we could get just the needed information in one block.

jcdutton and all you others, please also see the wiki page and file your own bugs. This bug report only tracks Nick's issue, and your comments here will be kind of lost when we close the report.

For unresolved AGPMode issues, we would like separate bug reports for each card/motherboard combination, so don't hesitate to file your own report.

We're closing this bug since it is has been some time with no response from the original reporter. However, if the issue still exists please feel free to reopen with the requested information. Also, if you could, please test against the latest development version of Ubuntu, since this confirms the bug is one we may be able to pass upstream for help.

Actually I think we have enough info now to do a quirk.

I made test packages for Jaunty in my PPA https://edge.launchpad.net/~tormodvolden/+archive/ppa (-ati and -radeon packages version 6.10.99.99+git20090218.9a108f0a-0ubuntu0tormod2) which you can test - without using AGPMode in xorg.conf of course. Please confirm that it works, and we will push the change upstream and get it into Jaunty.

Nick, did you get a chance to try my test packages?

Sorry, the laptop is mainly my wife's. I do not have Jaunty installed on it.

You can also test 1:6.10.99.99+git20090218.9a108f0a-0ubuntu0tormod2~intrepid for Intrepid.

Patch applied upstream http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=8f7c17a6b77ca368f067c2acb803cc8f9ef19351

Fixed in Jaunty with 1:6.11.0+git20090310.945ccbbd-0ubuntu1

Just wanted to tell that I have exactly the same experience with ATI Mobility Radeon 9600 on current Jaunty Beta with all current updates installed.
Without compiz the radeon-driver seems to run fine.

cljk, then you have to file a new bug, see the wiki page mentioned above.

Is it really the right way to workaround this bug with quirks?

As far as I understand, a quirk changes the default behavior for a specific combination of hardware so that such notebook-models use a certain agp-mode by default.

For my acer this is AGP1x, other modes result in freezes or corrupted graphics. By default, Ubuntu selects AGP4x which is correct because it is the best mode supported by the hardware.
It supports AGP4x and runs fine with fastwrites enabled on Windows XP.

So the problem is not that Ubuntu selects an unsupported AGP mode, the problem is that the highest mode supported by the hardware does not work with Ubuntu.

Are there realy no hints where the real problem is?

kede, I have asked the same but concluded that AGP is black magic. Some proprietary drivers seem to get away with it, but I think the problem is the hardware does not fully comply with the AGP specifications and only the manufacturers know how. I was hoping AMD/ATI opening up would help with this, but almost all their resources (logically) goes towards newer hardware using PCI-E etc.

I can confirm this is fixed in Karmic. See my comment on fd.o

Unfortunately, it just froze on me using Openoffice. Sorry, still not fixed.

10.10 has the problem, 10.04 did not seem to have exhibited the same issue.