powerpc/perf: Ignore separate BHRB privilege state filter request
authorAnshuman Khandual <khandual@linux.vnet.ibm.com>
Mon, 10 Jun 2013 05:53:28 +0000 (11:23 +0530)
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>
Wed, 24 Jul 2013 04:42:31 +0000 (14:42 +1000)
commit7689bdcab1dca061c4c91f0e1703cef1b7b67e71
tree96995e475dbe8601429008133d000e84c11c3d64
parent679750054ae8129a536734bccf526ff6da35376a
powerpc/perf: Ignore separate BHRB privilege state filter request

Completely ignore BHRB privilege state filter request as we are
already configuring that with privilege state filtering attribute
for the accompanying PMU event. This would help achieve cleaner
user space interaction for BHRB.

This patch fixes a situation like this

Before patch:-
------------
./perf record -j any -e branch-misses:k ls
Error:
The sys_perf_event_open() syscall returned with 95 (Operation not
supported) for event (branch-misses:k).
/bin/dmesg may provide additional information.
No CONFIG_PERF_EVENTS=y kernel support configured?

Here 'perf record' actually copies over ':k' filter request into BHRB
privilege state filter config and our previous check in kernel would
fail that.

After patch:-
-------------
./perf record -j any -e branch-misses:k ls
perf  perf.data  perf.data.old  test-mmap-ring
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.002 MB perf.data (~102 samples)]

Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/perf/power8-pmu.c