| Summary: | perf test shows multiple errors and fails on qemux86 | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] BSPs | Reporter: | Alexandru Georgescu <alexandru.c.georgescu> |
| Component: | bsps-oe-core | Assignee: | Tom Zanussi <tom.zanussi> |
| Status: | VERIFIED NOTABUG | QA Contact: | |
| Severity: | normal | ||
| Priority: | Medium | ||
| Version: | unspecified | ||
| Target Milestone: | 1.6 M2 | ||
| Hardware: | Other | ||
| OS: | x86 | ||
| Whiteboard: | |||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | No (bug/feature does not impact docs) | |
|
Description
Alexandru Georgescu
2013-12-05 10:31:12 UTC
Tom, can you have a look at this please? Do we expect this to work on all BSPs or is there a debug/tracing feature required? A quick look at the output seems to show it trying to parse e.g. scsi events, which may not be there if scsi isn't enabled in the kernel. My first question is if this is a regression - I don't recall this test being run before - or is this a new QA test? (In reply to comment #2) > A quick look at the output seems to show it trying to parse e.g. scsi > events, which may not be there if scsi isn't enabled in the kernel. > > My first question is if this is a regression - I don't recall this test > being run before - or is this a new QA test? Hi Tom, We don't usually test this, but I stumbled on it by testing eclipse-pluign. Eclipse has a tool that gives the developers the ability to use perf. So it was me just trying "perf test" in the Eclipse plugin and see what's the output. I cannot tell if it's a regression or not since. I also tested it on qemux86. On qemux86-64 gives slightly similar issues. I don't know for the other arches but I suspect similar output as well. perf testing shouldn't be done using qemu, which doesn't properly emulate the PMU. From qemux86 dmesg: Performance Events: Broken PMU hardware detected, using software events only. See previous entry - not expected to work on qemu. verified as per Tom's comments. |