Re: [RESEND PATCH V2 1/4] perf/x86/intel: fix event update for auto-reload

From: Liang, Kan
Date: Wed Jan 10 2018 - 09:31:10 EST




On 1/10/2018 5:22 AM, Jiri Olsa wrote:
On Mon, Jan 08, 2018 at 07:15:13AM -0800, kan.liang@xxxxxxxxx wrote:

SNIP

There is nothing need to do in x86_perf_event_set_period(). Because it
is fixed period. The period_left is already adjusted.

Signed-off-by: Kan Liang <kan.liang@xxxxxxxxxxxxxxx>
---
arch/x86/events/intel/ds.c | 69 ++++++++++++++++++++++++++++++++++++++++++++--
1 file changed, 67 insertions(+), 2 deletions(-)

diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c
index 3674a4b..cc1f373 100644
--- a/arch/x86/events/intel/ds.c
+++ b/arch/x86/events/intel/ds.c
@@ -1251,17 +1251,82 @@ get_next_pebs_record_by_bit(void *base, void *top, int bit)
return NULL;
}
+/*
+ * Specific intel_pmu_save_and_restart() for auto-reload.
+ */
+static int intel_pmu_save_and_restart_reload(struct perf_event *event,
+ u64 reload_val,
+ int reload_times)
+{
+ struct hw_perf_event *hwc = &event->hw;
+ int shift = 64 - x86_pmu.cntval_bits;
+ u64 prev_raw_count, new_raw_count;
+ u64 delta;
+
+ if ((reload_times == 0) || (reload_val == 0))
+ return intel_pmu_save_and_restart(event);

why is this check needed? AFAICS __intel_pmu_pebs_event is
called only if reload_times != 0 and reload_val is always
non zero for sampling


Here is a sanity check for reload_times and reload_val.
Right, usually they are non zero.
I think it should not bring any issues. Right?
If so, I think we may still keep it?

Thanks,
Kan