Hi John,

在 2021/1/26 23:56, John Garry 写道:

On 26/01/2021 12:33, John Garry wrote:

quoted

quoted

  }

@@ -332,8 +506,16 @@ static int hisi_ddrc_pmu_probe(struct platform_device *pdev)

          return ret;
      }
-    name = devm_kasprintf(&pdev->dev, GFP_KERNEL, "hisi_sccl%u_ddrc%u",
-                  ddrc_pmu->sccl_id, ddrc_pmu->index_id);
+    if (ddrc_pmu->identifier >= HISI_PMU_V2)
+        name = devm_kasprintf(&pdev->dev, GFP_KERNEL,
+                      "hisi_sccl%u_ddrc%u_%u",
+                      ddrc_pmu->sccl_id, ddrc_pmu->index_id,
+                      ddrc_pmu->sub_id);

I think that this is actually ok for perf.

Sounds good,

Checking pmu_uncore_alias_match():

bool pmu_uncore_alias_match(const char *pmu_name, const char *name)
{
    char *tmp = NULL, *tok, *str;
    bool res;

...

    for (; tok; name += strlen(tok), tok = strtok_r(NULL, ",", &tmp)) {
        name = strstr(name, tok);
        if (!name) {
            res = false;
            goto out;
        }
    }

    ...
}

We match each token in @pmu_name, "hisi_sccl,ddrc". So we would match "hisi_sccl%u_" part with
"hisi_sccl" token and then match "ddrc%u_%u" with "ddrc" token.

But it would be good to triple check. We need the JSON files created at some stage anyway :)

Sure, will do it.

Thanks,
Shaokun

quoted

I don't think perf can support event aliasing in this format (hisi_scclX_Y_Z), but I need to
double-check.

quoted

+    else
+        name = devm_kasprintf(&pdev->dev, GFP_KERNEL, 

.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel