Created attachment 359739 [details] dmesg Description of problem: After upgrading to Fedora 12 / rawhide my laptop crashes regularly. It usually happens when some "heavy" redraw operations happens. Usually when switching to Firefox. I regularly see redraw errors - especially white lines in firefox tab titles and dropouts of text in gnometerminal. I am running nomodeset because booting with KMS doesn't work - it gives "fading psycedelic colors" when KMS kicks in (reported for Fedora 11 at bug 496333). Once when the machine was hanging I also saw the colors faded away. Suspend/resume worked fine with Fedora 11 (nomodeset). With Fedora 12 the display stays black. It seems like it keeps running for some time (caps lock toggles and text console "find /" activates hd), but later it hangs. (I have however seen "possible circular locking dependency" messages, bug 519536.) I assume that these issues are related and most of them really not interessting because KMS is the way to go and should be fixed first. Please let me know if I can help test or provide more information. Version-Release number of selected component (if applicable): kernel-PAE-2.6.31-0.199.rc8.git2.fc12.i686 xorg-x11-drv-ati-6.13.0-0.2.20090821gitb1b77a4d6.fc12.i686 01:00.0 VGA compatible controller: ATI Technologies Inc M22 [Mobility Radeon X300] http://www.smolts.org/client/show/pub_6ee23250-8a78-44b8-a4ec-9a0b999ebbe1
Created attachment 359740 [details] X log X log - probably not that relevant
Created attachment 359933 [details] dmesg from 2.6.31-0.204.rc9.fc12.i686.PAE with KMS (and thus unusable display) As suggested by nanonyme: dmesg from 2.6.31-0.204.rc9.fc12.i686.PAE with KMS (and thus unusable display) He also noted that nomodeset X had (EE) RADEON(0): [dri] DRIScreenInit failed. Disabling DRI. - but I suggest focussing on making KMS work first.
*** Bug 501140 has been marked as a duplicate of this bug. ***
Do you still have issue with lastest fedora 12 kernel ?
*** Bug 496333 has been marked as a duplicate of this bug. ***
Re comment 4: Yes, I generally use the latest kernels and still see the issues. Of course I work around them by - not using KMS - not suspending (fortunately hibernate (suspend-to-disk) has started to work) - ignoring redraw errors - restart when it hangs (- and as en extra bonus I have bug 528390 ;-) Should I test something in a specific kernel version?
So to sumup KMS: - computer hangs after a while, mostly when "heavily" using firefox - suspend/resume is broken For suspend resume i suspect that : https://bugzilla.redhat.com/show_bug.cgi?id=519536 is the root cause of the issue. Can you login on the computer through ssh after resume (wait a bit to give time for network to comeback) ?
No, the KMS sumup is: KMS doesn't work - it gives "fading psycedelic colors" when KMS kicks in (reported for Fedora 11 at bug 496333). The issues I correlate with firefox might of course be bug 528390. I have only seen few hangs where I was sure firefox wasn't involved. I will retest suspend/resume ASAP and see if 2.6.31.5-121 has helped. Thanks for looking into this. Sorry for the messy report, but I'm not sure these issues can be separated.
Please test with KMS enabled. Do you have a bug open for your KMS issue ? If you don't i suggest to open one and i will switch this one to UMS. Right now we are focussing on fixing KMS as it's were we want to go.
Re comment #9: I have created 540139 for the KMS issue, so what is left here is nomodeset issues: - X crashes (probably bug 528390, mostly avoidable with care) - redraw errors (not every day, but regularly seen on tortoisehg logs, alt-tab or mouse-hoovering often helps) - resume after suspend hangs (hibernate-to-disk works as a workaround) Re comment #7: (See also bug 519536) With kernel-PAE-2.6.31.6-145.fc12.i686 nomodeset (latest from koji) resume doesn't even get as far as this issue. It locks up really hard: Resuming turns the display on, but nothing is written to /var/log/messages, apparently no disc activity, no reaction to caps-lock, ctrl-alt-f2 ctrl-alt-del does nothing, only a hard power off works. (I don't have another machine easily available for ssh'ing, but I assume that the description above is enough?)
I am having similar issues, after consulting the common fedora 12 bugs it suggested posting a report on bug #498457 but this one seemed closer to the problems I was having. I am running a dual desktop display with an ATI RV370. I have had no lockups of X but there are a lot of redraw artifacts. I can get most of them to go away for a particular window by minimizing it and bringing it back. Following the suggestions of the Fedora 12 common bugs page I added 'nomodeset' to my kernel parameters and so far that seems to have resolved the issue. Before I added 'nomodeset' the most common applications to see the issues with was thunderbird and firefox. It would also happen when loading up KDE after login when I was using gdm. Switching to kdm+kde seems to have resolved that issue before I started using 'nomodeset'. I guess this could be a separate bug but figured the more information the better. I dont ever suspend/resume my computer so I do not know how it would handle that. I am happy to provide any logs or files requested, and will perform any test on my system that is requested. Good luck to all those hard working developers out there :)
Created attachment 373099 [details] Redraw problems in gnome-terminal tab titles The crash problem might be solved with xorg-x11-drv-ati-6.13.0-0.12.20091119git437113124.fc12.i686, but I just verified that the redraw problems also remains.
I just reported a hang on bug 540511. So the hangs and the crashes I have mentioned here might be different issues.
xorg-x11-drv-ati-6.13.0-0.15.20091127gita8dbf7c23 without XAA still shows the redraw problems shown in comment #12. Using 'Option "AccelMethod" "XAA"' in xorg.conf as described on https://fedoraproject.org/wiki/Common_F12_bugs#radeon-misc-gfx solves the redraw problems (and the firefox crashes and gives better vlc performance, but it doesn't solve resume).
With kernel-PAE-2.6.32.3-24.fc12.i686 xorg-x11-drv-ati-6.13.0-0.20.20091221git4b05c47ac.fc12.i686 - KMS works - suspend/resume works - no redrawerrors so far - no hangs so far - fine performance and stability without XAA hack The situation has changed so much that if I see similar problems again I will file a new report. Closing.
Update: I do see corruption now - probably bug 529081.