Re: [RFC] libperf: Add support for user space counter access

From: Peter Zijlstra
Date: Mon Aug 10 2020 - 16:25:05 EST


On Fri, Aug 07, 2020 at 05:05:17PM -0600, Rob Herring wrote:
> x86 and arm64 can both support direct access of event counters in
> userspace. The access sequence is less than trivial and currently exists
> in perf test code (tools/perf/arch/x86/tests/rdpmc.c) with copies in
> projects such as PAPI and libpfm4.
>
> Patches to add arm64 userspace support are pending[1].
>
> For this RFC, looking for a yes, seems like a good idea, or no, go away we
> don't want this in libperf.

I'd like it lots better if you'd at least take an optimized version of
this, also see this thread:

https://lkml.kernel.org/r/20200322101848.GF2452@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Also, I usually strip out all the multiplexing crud out (and use pinned
counters), which saves a bunch.


static inline u64 mmap_read_pinned(void *addr)
{
struct perf_event_mmap_page *pc = addr;
u32 seq, idx, width = 0;
u64 count;
s64 pmc = 0;

do {
seq = pc->lock;
barrier();

idx = pc->index;
count = pc->offset;
if (pc->cap_user_rdpmc && idx) {
width = pc->pmc_width;
pmc = rdpmc(idx - 1);
}

barrier();
} while (pc->lock != seq);

if (idx) {
pmc <<= 64 - width;
pmc >>= 64 - width; /* shift right signed */
count += pmc;
}

return count;
}