On 10/28/2015 03:20 AM, Alex Smith wrote:
On 27 October 2015 at 20:46, Leonid Yegoshin <Leonid.Yegoshin@xxxxxxxxxx> wrote:
I believe, until this issue is fixed the R4K only CPU should be excluded
from VDSO timing acceleration.
The VDSO code will currently use the CP0 count whenever the kernel is
using it as its primary clocksource, aside from the case where RDHWR
is broken as it is on old QEMUs.
1) I don't see that in code - there is no check that kernel uses
actually uses R4K clocksource as primary (A), and if kernel uses R4K
count as a clocksource and later switches to some more precise
clocksource then there is no change in VDSO gettimeofday handling (B).
2) The fact that MIPS kernel as today uses CP0_COUNT in any core as the
same clocksource is correct only until first power saving event with CPU
clock disabled (skipping Octeon). After that it is an incorrect use
without an accurate synchronization and that synchronization doesn't exist.
And I remember that today kernel uses only CPU0 CP0_COUNT to update
time... may be wrong, need to check, but that could be a good code.
Maybe I'm missing something but I don't see anything in the generic
timekeeping code that handles the same clocksource being
unsynchronised or running at a different rate on different CPUs.
(I would like to skip here the generic timekeeping code capabilities,
just to restrict a discussion to HW capabilities)
Given that, if you think there is an issue that prevents the VDSO from
using it then that would surely affect the kernel as well and needs to
be fixed separately?
If it really is necessary to prevent the VDSO from using a certain
clocksource even though the kernel is using it, it should be a simple
matter of setting clocksource.archdata.vdso_clock_mode to
VDSO_CLOCK_NONE. This is how this patch stops it using the CP0 count
when RDHWR is broken.
OK, just put kernel-build time check that it is not SMP without GIC
clocksource or it is Octeon. It would be enough to stop a mess.
- Leonid
