arm64: perf: use architected event for CPU cycle counter

We currently use a fake event encoding (0xFF) to indicate CPU cycles so
that we don't waste an event counter and can target the hardware cycle
counter instead.

The problem with this approach is that the event space defined by the
architecture permits an implementation to allocate 0xFF for some other
event.

This patch uses the architected cycle counter encoding (0x11) so that
we avoid potentially clashing with event encodings on future CPU
implementations.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
This commit is contained in:
Will Deacon 2012-11-05 12:34:47 +00:00 committed by Catalin Marinas
parent 3fd9396af8
commit f46f979fda

View File

@ -613,17 +613,11 @@ enum armv8_pmuv3_perf_types {
ARMV8_PMUV3_PERFCTR_BUS_ACCESS = 0x19,
ARMV8_PMUV3_PERFCTR_MEM_ERROR = 0x1A,
ARMV8_PMUV3_PERFCTR_BUS_CYCLES = 0x1D,
/*
* This isn't an architected event.
* We detect this event number and use the cycle counter instead.
*/
ARMV8_PMUV3_PERFCTR_CPU_CYCLES = 0xFF,
};
/* PMUv3 HW events mapping. */
static const unsigned armv8_pmuv3_perf_map[PERF_COUNT_HW_MAX] = {
[PERF_COUNT_HW_CPU_CYCLES] = ARMV8_PMUV3_PERFCTR_CPU_CYCLES,
[PERF_COUNT_HW_CPU_CYCLES] = ARMV8_PMUV3_PERFCTR_CLOCK_CYCLES,
[PERF_COUNT_HW_INSTRUCTIONS] = ARMV8_PMUV3_PERFCTR_INSTR_EXECUTED,
[PERF_COUNT_HW_CACHE_REFERENCES] = ARMV8_PMUV3_PERFCTR_L1_DCACHE_ACCESS,
[PERF_COUNT_HW_CACHE_MISSES] = ARMV8_PMUV3_PERFCTR_L1_DCACHE_REFILL,
@ -1106,7 +1100,7 @@ static int armv8pmu_get_event_idx(struct pmu_hw_events *cpuc,
unsigned long evtype = event->config_base & ARMV8_EVTYPE_EVENT;
/* Always place a cycle counter into the cycle counter. */
if (evtype == ARMV8_PMUV3_PERFCTR_CPU_CYCLES) {
if (evtype == ARMV8_PMUV3_PERFCTR_CLOCK_CYCLES) {
if (test_and_set_bit(ARMV8_IDX_CYCLE_COUNTER, cpuc->used_mask))
return -EAGAIN;