Created attachment 70558 [details] Dmesg log - after third pass of piglit. I was giving piglit a trial run, and noticed that there was a few errors about the radeon driver being able to allocate gem objects. Piglit package: 1.0.0~git20121122.cf47699-0edgers~quantal1 Piglit Git revision:cf47699087f74a6fd94c59a74837fc6c3c6422a8 Mesa Package: 9.1~git20121110.c6f426c0-0ubuntu0ricotz~quantal Mesa Git Revision: c6f426c02d14fcdfdbdd03473d098d4c9b81a8b2 Libdrm package: 2.4.40+git20121109.e01d68f9-0ubuntu0ricotz~quantal libdrm revision: e01d68f9f3acfc35fe164283904b5d058c2ab378 Notes: In the kernel log, the first pass errors started at: 11166.715813, second pass started at 11742.655557 and third pass had errors start 13408.932142.
Created attachment 70559 [details] Xorg log
Created attachment 70560 [details] Piglit test results for all 3 passes
Some piglit tests are about allocating as large textures as possible, which can result in such intermittent allocation failures. It might be feasible to handle this more gracefully in userspace, but otherwise it should be mostly harmless, assuming there are no other, real problems associated with it.
AFAIK, allocation failures are correctly reported to applications as the GL_OUT_OF_MEMORY errors.
Totally agree with that. Some piglit tests even allocate all the memory on purpose to test that the driver can handle that gracefully, and the expectation in the good case is that those error messages start to appear. As long as you don't experience any crash, hang or something like this everything should be fine.
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.