On 21/09/2018 16:08, Shameer Kolothum wrote:
+
+ssize_t smmu_pmu_event_show(struct device *dev,
+ struct device_attribute *attr, char *page)
+{
+ struct perf_pmu_events_attr *pmu_attr;
+
+ pmu_attr = container_of(attr, struct perf_pmu_events_attr, attr);
+
+ return sprintf(page, "event=0x%02llx\n", pmu_attr->id);
+}
Is there some reason PMU drivers have their own edition of this
function? I see a few similar ones here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/kernel/perf_event.c#n309
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/powerpc/perf/core-book3s.c#n1986
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/s390/kernel/perf_event.c#n239
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/perf/qcom_l2_pmu.c#n731
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/perf/qcom_l3_pmu.c#n655
That is apart from the event id width format.
But the leading zeroes seem to be cropped anyway in the perf tool,
according to this:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/util/pmu.c#n335
(this was added in 0c24d6fb7db, in June 18)
John
+
+#define SMMU_EVENT_ATTR(_name, _id) \
+ (&((struct perf_pmu_events_attr[]) { \
+ { .attr = __ATTR(_name, 0444, smmu_pmu_event_show, NULL), \
+ .id = _id, } \
+ })[0].attr.attr)
+
+static struct attribute *smmu_pmu_events[] = {
+ SMMU_EVENT_ATTR(cycles, 0),
+ SMMU_EVENT_ATTR(transaction, 1),
+ SMMU_EVENT_ATTR(tlb_miss, 2),
+ SMMU_EVENT_ATTR(config_cache_miss, 3),
+ SMMU_EVENT_ATTR(trans_table_walk, 4),
+ SMMU_EVENT_ATTR(config_struct_access, 5),
+ SMMU_EVENT_ATTR(pcie_ats_trans_rq, 6),
+ SMMU_EVENT_ATTR(pcie_ats_trans_passed, 7),
+ NULL
+};
