From: Felix <fx...@gm...> - 2006-03-18 19:45:59
|
Am Mittwoch, den 08.03.2006, 07:02 +0100 schrieb Alexandre Gouaillard: > hi all. >=20 > I have read about X hanging when a client would die holding the lock No, when a client dies (is killed, aborts, segfaults, ...) while holding the hardware lock, the lock is released automatically. > , or=20 > when a client would get a freeze command, but I don't seem to be able t= o=20 > reproduce it under recent linux distros (fedora). If you're talking about SIGSTOP or a debugger breakpoint, you're right. However, that should not happen in normal operation. A user debugging a DRI driver should know what he's doing and run gdb through ssh. Another situation where a client can hold the lock indefinitely is, when the GPU is locked up and the client holding the lock is waiting for the GPU to become available, which never happens. The symptoms are always the same, but the exact cause is different every time and usually hard to track down. If you're referring to a specific problem, please be more ... specific. >=20 > what is the curent status of this problem? Is it solve in DRi framework= , or=20 > any workaround has been found? You said, you can't reproduce it, so I guess I don't understand your problem. Regards, Felix >=20 > alex. >=20 --=20 | Felix K=C3=BChling <fx...@gm...> http://fxk.de.v= u | | PGP Fingerprint: 6A3C 9566 5B30 DDED 73C3 B152 151C 5CC1 D888 E595 | |