On Thu, Feb 04, 2010 at 03:37:58PM +0100, Émeric Maschino wrote: > 2010/2/3 Stephane Marchesin <stephane.marche...@gmail.com>: > > No, you are right they don't trigger MCA. Hmm I didn't have any of > > those back then, my lockups came from the bus mostly... > > Thank you for clarifying this point. > > > Really if you have such lockups they may also happen on x86, did you > > try the card there? > > Yes, I have no problem with this (AGP Pro 4x) graphics adapter (ATI > FireGL X1) in x86 hardware. > > > At this point your best bet is probably replay the crashing sequence > > until you can reduce it to the offending couple of commands. > > OK. Are the commands you're talking about the argument passed in the > various ioctl() calls logged when stracing the offending OpenGL > application? For example, strace glxgears gives lines like: > ioctl(4, 0xc0106451, 0x60000fffffd52d30) = 0 > ioctl(4, 0xc0186419, 0x60000fffffd52d30) = 0 > ioctl(4, 0x40106459, 0x60000fffffd52d58) = 0 > where 4 is the file descriptor of /dev/dri/card0. Are 0xc0106451, > 0xc0186419 or 0x40106459 the commands passed to the GPU? > > I don't know if it's related to GPU lockup or not (I mean, being the > cause or a consequence), but I've also noticed in the strace glxgears > logs (or even simple application like glxinfo) that most of the read() > calls to /tmp/.X11-unix/X0 fail, whereas the writev() calls seem to > succeed: > poll([{fd=3, events=POLLIN|POLLOUT}], 1, -1) = 1 ([{fd=3, revents=POLLOUT}]) > writev(3, [{"\222\0\3\0\4\0\0\0\0\0\0\0", 12}, {NULL, 0}, {"", 0}], 3) = 12 > poll([{fd=3, events=POLLIN}], 1, -1) = 1 ([{fd=3, revents=POLLIN}]) > read(3, "\1\0*\0\0\0\0\0\4\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0", > 4096) = 32 > read(3, 0x60000000000093e4, 4096) = -1 EAGAIN (Resource > temporarily unavailable) > where 3 is the file descriptor of /tmp/.X11-unix/X0 > > Émeric >
IIRC old radeon drm doesn't have any thing to dump GPU command stream. Look at http://www.x.org/docs/AMD/R5xx_Acceleration_v1.4.pdf to see what radeon GPU stream command looks like (packet pm4 stuff). Note that dump GPU command stream can quickly eat Gigs of data and finding what is causing the lockup is then very cumberstone especialy as in your case it sounds like it's a timing issue. You might want to force your card into pci mode to see if it's agp related. Cheers, Jerome ------------------------------------------------------------------------------ The Planet: dedicated and managed hosting, cloud storage, colocation Stay online with enterprise data centers and the best network in the business Choose flexible plans and management services without long-term contracts Personal 24x7 support from experience hosting pros just a phone call away. http://p.sf.net/sfu/theplanet-com -- _______________________________________________ Dri-devel mailing list Dri-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dri-devel