On Tue, May 12, 2020 at 02:41:03PM +0200, Peter Zijlstra wrote:
> This completes the ARM64 cap_user_time support.
>
> Signed-off-by: Peter Zijlstra (Intel) <pet...@infradead.org>
> ---
> arch/arm64/kernel/perf_event.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
> --- a/arch/arm64/kernel/perf_event.c
> +++ b/arch/arm64/kernel/perf_event.c
> @@ -1173,6 +1173,7 @@ void arch_perf_update_userpage(struct pe
>
> userpg->cap_user_time = 0;
> userpg->cap_user_time_zero = 0;
> + userpg->cap_user_time_short = 0;
>
> do {
> rd = sched_clock_read_begin(&seq);
> @@ -1183,13 +1184,13 @@ void arch_perf_update_userpage(struct pe
> userpg->time_mult = rd->mult;
> userpg->time_shift = rd->shift;
> userpg->time_zero = rd->epoch_ns;
> + userpg->time_cycle = rd->epoch_cyc;
s/time_cycle/time_cycles, maybe consider to change the naming to
'time_cycle'.
This patch set looks good to me after I tested it on Arm64 board.
Thanks,
Leo
> + userpg->time_mask = rd->sched_clock_mask;
>
> /*
> - * This isn't strictly correct, the ARM64 counter can be
> - * 'short' and then we get funnies when it wraps. The correct
> - * thing would be to extend the perf ABI with a cycle and mask
> - * value, but because wrapping on ARM64 is very rare in
> - * practise this 'works'.
> + * Subtract the cycle base, such that software that
> + * doesn't know about cap_user_time_short still 'works'
> + * assuming no wraps.
> */
> userpg->time_zero -= (rd->epoch_cyc * rd->mult) >> rd->shift;
>
> @@ -1214,4 +1215,5 @@ void arch_perf_update_userpage(struct pe
> */
> userpg->cap_user_time = 1;
> userpg->cap_user_time_zero = 1;
> + userpg->cap_user_time_short = 1;
> }
>
>
