Roland Lezuo reported on the Debian BTS that radeonhd 1.2.0 (and latest upstream git) causes random freezes on his RV530. He says: "I could reliable trigger the bug by activating the screensaver. The hack running is a python based gnome-screensaver plugin written by me. The freeze occured after the screen faded out, at or shortly after the the first few seconds of the hack running." The config and log is available at http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=5;bug=476309 Driver 1.1.0 didn't have any problem. He bisected and is rather confident that the first bad commit is: commit af0c64d897f7f2288419b6e43468850e1037bb59 Author: Luc Verhaegen <libv@skynet.be> Date: Mon Jan 28 13:09:55 2008 +0100 R5xx accel: Add XAA support. MMIO path only, no CP support yet.
Before blaming the XAA code in our driver, please verify that Option "AccelMethod" "ShadowFB" mitigates the problem.
Hi, I'm the original reporter, and I tried with Option "AccelMethod" "ShadowFB" and the system runs stable (using radeonhd 1.2.0), so it seems to be the XAA code. Let me know if I need to run testcases for you. regards roland lezuo
Does 'freeze' mean the system is totally unresponsive and unaccessible from remote?
Yes, this is what "freeze" means. Does not answer pings any more, only a hard reset leaves the state.
Hi, I had some spare time today and tried with todays git version - still the system freezes. I also checkout the ati git repository and gave that driver a quick try, it recognized my hardware but the system also froze. I have figured out that starting the screensaver (by locking the screen(gnome) ) is a very reliable test for freezing the system. How do you debug such freezes?
ATM this is difficult. Also, we're replacing the MMIO based acceleration code with a CP (command processor) based version soon, which could expose a different behavior. As we do not see many reports about freezes this is difficult to debug, and I would urge you to use ShadowFB until we have switched over. When the switch is done, please retest.
@Brice: Does the problem still exist with the latest GIT driver? If the hardware isn't broken this screensaver plugin would be a good test for the 2d driver.
> --- Comment #7 from Egbert Eich <eich@pdx.freedesktop.org> 2008-07-20 06:58:30 PST --- > @Brice: Does the problem still exist with the latest GIT driver? > If the hardware isn't broken this screensaver plugin would be a good test for > the 2d driver. > Roland (the original bug reported) is subscribed to this bug, I'll let him reply directly. Brice
I've tried to build b0f470aa76cc156092754187fc8c6c41c4125ce4 but failed: r5xx_xaa.c:132: error: 'struct RHDRec' has no member named 'TwoDPrivate' If you give me an version i should try, i'll do so happily. regards roland lezuo
You should do a rebiuld from scratch. The easiest way would be to make a: ./autogen.sh; make in the top directory.
thanks this did it. I used the driver now for 2 days and had no stability issues. The speedup is great too. From my point of view the bug can be closed (I do so). Btw, if tried version 708e31a4d8e882fb4b448e05073c258b7bc12551. thanks for the great work you do, roland lezuo
Maybe you can give us a pointer to the screensaver plugin used here. This way we can use it as a test to avoid this problem in the future.
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.