On Tue, 2009-01-06 at 10:41 -0800, Jesse Barnes wrote: > This patch adds a sanity check to drmWaitVBlank to prevent hangs. Since the > server interrupts syscalls pretty frequently with SIGALM, we tend to restart > syscalls when EINTR is returned. That means when a vblank timeout happens, > we probably won't catch it (it happens after 3s, and SIGALM happens every few > ms), so we can fairly easily get stuck restarting the vblank wait ioctl if > something fishy is going on (like vblank interrupts are disabled for some > reason). > > So this patch removes the top level restart code, pushing it into the vblank > wait ioctl wrapper, while adding a timeout. If there are cases where more > than 1s waits are desirable, we'd probably need to check the sequence numbers > and come up with a more reasonable value, or add a new call that takes a > timeout parameter.
Uh, changing drmIoctl to return on EINTR seems like a bad plan. I'd hack up drmWaitVBlank to do the ioctl directly instead. > Any thoughts here? I chose to use clock_gettime(CLOCK_MONOTONIC..) to avoid > any issues with the TOD clock getting reset under us; I'm not sure if that's > safe everywhere and a simpler method is probably possible too... -- keith.pack...@intel.com
signature.asc
Description: This is a digitally signed message part
------------------------------------------------------------------------------ Check out the new SourceForge.net Marketplace. It is the best place to buy or sell services for just about anything Open Source. http://p.sf.net/sfu/Xq1LFB
-- _______________________________________________ Dri-devel mailing list Dri-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dri-devel
