perf kvm: Guest userspace samples should not be lumped with host uspace
authorDavid Ahern <dsahern@gmail.com>
Fri, 20 Jul 2012 23:25:48 +0000 (17:25 -0600)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Wed, 25 Jul 2012 14:29:46 +0000 (11:29 -0300)
e.g., perf kvm --host  --guest report -i perf.data --stdio -D
shows:

599127912065356 0x143b8 [0x48]: PERF_RECORD_SAMPLE(IP, 5): 5671/5676: 0x7fdf95a061c0 period: 1 addr: 0
... chain: nr:2
.....  0: ffffffffffffff80
.....  1: fffffffffffffe00
 ... thread: qemu-kvm:5671
 ...... dso: <not found>

(IP, 5) means sample in guest userspace. Those samples should not be lumped
into the VMM's host thread. i.e, the report output:

    56.86%  qemu-kvm  [unknown]         [u] 0x00007fdf95a061c0

With this patch the output emphasizes it is a guest userspace hit:

    56.86%  [guest/5671]  [unknown]         [u] 0x00007fdf95a061c0

Looking at 3 VMs (2 64-bit, 1 32-bit) with each running a CPU bound
process (openssl speed), perf report currently shows:

  93.84%  117726   qemu-kvm  [unknown]   [u] 0x00007fd7dcaea8e5

which is wrong. With this patch you get:

  31.50%   39258   [guest/18772]  [unknown]   [u] 0x00007fd7dcaea8e5
  31.50%   39236   [guest/11230]  [unknown]   [u] 0x0000000000a57340
  30.84%   39232   [guest/18395]  [unknown]   [u] 0x00007f66f641e107

Tested-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1342826756-64663-4-git-send-email-dsahern@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/util/session.c

index 8e48559..90ee39d 100644 (file)
@@ -918,7 +918,9 @@ static struct machine *
 {
        const u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
 
-       if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL && perf_guest) {
+       if (perf_guest &&
+           ((cpumode == PERF_RECORD_MISC_GUEST_KERNEL) ||
+            (cpumode == PERF_RECORD_MISC_GUEST_USER))) {
                u32 pid;
 
                if (event->header.type == PERF_RECORD_MMAP)