Attached is a patch to fix the drmRadeonCmdBuffer: -22
problem in the SW TCL path (used on Radeon VE & M6 cards) in the Radeon driver. It turns out that not every place in the code was using the right packet sizes. I believe that this was the only missed place. I have tested with with a Q3 demo (that hit the lock at the same spot everytime) on both a Radeon M6 and a Radeon DDR with RADEON_TCL_FORCE_DISABLE set. See my previous message in this thread for the complete description of the problem. The lock in bzflag seems to be caused by something different. I have not had a chance to track that one down. -- Smile! http://antwrp.gsfc.nasa.gov/apod/ap990315.html
radeon-swtcl-hardlock-fix.patch.gz
Description: application/gunzip
