When trying to play doom3 ( https://bugs.freedesktop.org/show_bug.cgi?id=23545 ) my system instantly lockups after starting a new game. I am using current git compiled without libdrm_radeon and without KMS: GL_RENDERER = Mesa DRI R300 (RV530 71C5) 20090101 x86/MMX/SSE2 TCL
Correct doom3 download link: http://zerowing.idsoftware.com/
Could be dupe of https://bugs.freedesktop.org/show_bug.cgi?id=22372 and just not recovering from the assert due to doom3's signal handler (so it's not actually locked up you could just kill doom3 from another box for instance).
If your assessment is correct, then it may not a driver bug, but that doesn't really help users who are affected by the problem. Is there some piece of advice we can give in that situation? I.e. some magic key combination like Ctrl+Alt+Del on Windows that people could hit to recover their desktop session? Obviously the now removed Ctrl+Alt+Backspace is not a solution since it kills the desktop session, and Ctrl+Alt+Fn-ing to a terminal plus invoking magic commands is also not a good solution.
OK, the game can indeed be killed from another box after sshing from it. I also tried the patch attached to bug #22372 comment #8, but it doesn't change anything. Other bugs that may be related to this are: bug #22742, bug #22741.
Just want to add that after killing the game without the patch I get this assert: doom.x86: radeon_mipmap_tree.c:117: compute_tex_image_offset: Assertion `lvl->size > 0' failed. while when killing it with the patch I get this: doom.x86: radeon_texture.c:874: migrate_image_to_miptree: Assertion `dstlvl->width == image->base.Width' failed.
The hang is probably related to the DRI1 lock, and may not happen with DRI2. Not sure anything can be done for DRI1, though this may have been made worse by a recent signal related DRM change which helped for starting a DRI1 enabled X server in gdb.
The patch at bug #22742 comment #7 also fixes this issue.
The original issue is fixed and the game starts fine. See bug #25179 for a new minor issue.
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.