On Fri, Jan 13, 2017 at 10:40:42AM -0600, Kim Phillips wrote:

On Fri, 13 Jan 2017 16:03:48 +0000
Will Deacon [off-list ref] wrote:

quoted

+#define DRVNAME				"arm_spe_pmu"

PMU is implied.  "arm_spe"?

As stated before, I'm going for consistency here. Is it causing any
real issues on the tooling side?

quoted

+	if (is_kernel_in_hyp_mode()) {
+		if (attr->exclude_kernel != attr->exclude_hv)
+			return -EOPNOTSUPP;
+	} else if (!attr->exclude_hv) {
+		return -EOPNOTSUPP;
+	}
+
+	reg = arm_spe_event_to_pmsfcr(event);
+	if ((reg & BIT(PMSFCR_EL1_FE_SHIFT)) &&
+	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_EVT))
+		return -EOPNOTSUPP;
+
+	if ((reg & BIT(PMSFCR_EL1_FT_SHIFT)) &&
+	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_TYP))
+		return -EOPNOTSUPP;
+
+	if ((reg & BIT(PMSFCR_EL1_FL_SHIFT)) &&
+	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_LAT))
+		return -EOPNOTSUPP;
+
+	return 0;
+}

Please insert pr_* statements before blindly returning errors before a
better facility becomes available.

That was discussed in the thread I linked to last time:

https://lkml.org/lkml/2015/8/26/661

and there are good reasons not to add those prints.

Will