Created attachment 128422 [details] valgrind output There seems to be some memory corruption (use after free) after running windows version of INSIDE game demo through wine and Steam, which leads to random crashes. Crashes do not happen if I use modesetting_drv instead. Valgrind output attached. I'm not entirely sure it's amdgpu's fault, as it only participates in some of the traces. I'm using my distro's version of Xorg (1.18.4-0ubuntu0.2), it would be bad if Ubuntu LTS is shipping broken Xorg, so I hope it's not that.
Created attachment 128423 [details] Xorg.log
Yeah, looks like an xserver present code issue, not a driver one, reassigning. Please attach the output of valgrind with debugging symbols available for /usr/lib/xorg/Xorg (make sure the xserver-xorg-core-dbg package is installed).
xserver-xorg-core-dbg is installed, but for some reason valgrind doesn't like those debug symbols (see it's complaints ant the beginning of the log). I've tested this on iGPU with xf86-video-intel and it doesn't trigger the issue. As modesetting is also fine, it strongly suggests it's amdgpu that's somehow provoking it. It looks like I'll have to live with modesetting for now...
(In reply to Grazvydas Ignotas from comment #3) > xserver-xorg-core-dbg is installed, but for some reason valgrind doesn't > like those debug symbols (see it's complaints ant the beginning of the log). Looks like the xserver-xorg-core-dbg package is missing the debugging symbols for /usr/lib/xorg/Xorg. Any chance you can build the xorg-server source package locally and replace /usr/lib/xorg/Xorg with the unstripped version from your build? It's hard to make sense of the valgrind output without line numbers for the Xorg entries. > I've tested this on iGPU with xf86-video-intel and it doesn't trigger the > issue. Was DRI3 enabled when testing that as well? > As modesetting is also fine, it strongly suggests it's amdgpu that's > somehow provoking it. valgrind doesn't complain either in that case?
(In reply to Michel Dänzer from comment #4) > > Any chance you can build the xorg-server source package locally and replace > /usr/lib/xorg/Xorg with the unstripped version from your build? It's hard to > make sense of the valgrind output without line numbers for the Xorg entries. I'll try that after work. > > I've tested this on iGPU with xf86-video-intel and it doesn't trigger the > > issue. > > Was DRI3 enabled when testing that as well? Yes. > > As modesetting is also fine, it strongly suggests it's amdgpu that's > > somehow provoking it. > > valgrind doesn't complain either in that case? No, pretty much clean output.
Today I'm no longer able to reproduce the issue. I guess my GPU was in some bad state, as failed flips can be seen in Xorg.log from that day (see Comment 1), I think there were some VM faults in dmesg too. Anyway it all works now, closing until I can reproduce this reliably again.
I was able to reproduce crashes like this by making the driver flip hook artificially return FALSE. I'll investigate more after the holidays if nobody beats me to it.
-- GitLab Migration Automatic Message -- This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity. You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/xorg/xserver/issues/511.
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.