[opensuse-kernel] TSC question for kernel developers
Hi I know it's slightly offtopic, but I guess people here might know the answer ;-) On a modern CPU, with constant_tsc feature, how well synchronized are the TSCs when synchronized on bootup by the linux kernel? And do they stay like that, or do they drift? Thanks, Stefan -- Stefan Seyfried "Any ideas, John?" "Well, surrounding them's out." -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On Thursday 14 October 2010 09:05:56 pm Stefan Seyfried wrote:
Hi
I know it's slightly offtopic, but I guess people here might know the answer ;-)
On a modern CPU, with constant_tsc feature, how well synchronized are the TSCs when synchronized on bootup by the linux kernel? They are not synchronized by the kernel because this would mess up synchronization already. On i386, when x86 was still split, it was tried to sync TSC on OS level and it caused grief. And do they stay like that, or do they drift? They must not drift, otherwise you would not be able to use it as a clocksource.
Thomas -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On Thursday 14 October 2010 09:05:56 pm Stefan Seyfried wrote:
Hi
I know it's slightly offtopic, but I guess people here might know the answer ;-)
On a modern CPU, with constant_tsc feature, how well synchronized are the TSCs Never tried it myself, but heard it detects warps rather well: http://people.redhat.com/mingo/time-warp-test/time-warp-test.c
The kernel has some boot time sync check mechanisms for some time to detect bad behaving CPUs or clustered systems with warping TSCs and falls back (hopefully with an obvious message) to hpet then. Thomas -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
participants (2)
-
Stefan Seyfried -
Thomas Renninger