perf evsel: Handle EACCESS + perf_event_paranoid=2 in fallback()

Now with the default for the kernel.perf_event_paranoid sysctl being 2 [1]
we need to fall back to :u, i.e. to set perf_event_attr.exclude_kernel
to 1.

Before:

  [acme@jouet linux]$ perf record usleep 1
  Error:
  You may not have permission to collect stats.

  Consider tweaking /proc/sys/kernel/perf_event_paranoid,
  which controls use of the performance events system by
  unprivileged users (without CAP_SYS_ADMIN).

  The current value is 2:

    -1: Allow use of (almost) all events by all users
  >= 0: Disallow raw tracepoint access by users without CAP_IOC_LOCK
  >= 1: Disallow CPU event access by users without CAP_SYS_ADMIN
  >= 2: Disallow kernel profiling by users without CAP_SYS_ADMIN
  [acme@jouet linux]$

After:

  [acme@jouet linux]$ perf record usleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.016 MB perf.data (7 samples) ]
  [acme@jouet linux]$ perf evlist
  cycles:u
  [acme@jouet linux]$ perf evlist -v
  cycles:u: size: 112, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|PERIOD, disabled: 1, inherit: 1, exclude_kernel: 1, mmap: 1, comm: 1, freq: 1, enable_on_exec: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1
  [acme@jouet linux]$

And if the user turns on verbose mode, an explanation will appear:

  [acme@jouet linux]$ perf record -v usleep 1
  Warning:
  kernel.perf_event_paranoid=2, trying to fall back to excluding kernel samples
  mmap size 528384B
  [ perf record: Woken up 1 times to write data ]
  Looking at the vmlinux_path (8 entries long)
  Using /lib/modules/4.6.0-rc7+/build/vmlinux for symbols
  [ perf record: Captured and wrote 0.016 MB perf.data (7 samples) ]
  [acme@jouet linux]$

[1] 0161028b7c8a ("perf/core: Change the default paranoia level to 2")

Reported-by: Ingo Molnar <mingo@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-b20jmx4dxt5hpaa9t2rroi0o@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index a5f339d..645dc18 100644 (file)
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -2345,6 +2345,8 @@ out:
 bool perf_evsel__fallback(struct perf_evsel *evsel, int err,
                          char *msg, size_t msgsize)
 {
+       int paranoid;
+
        if ((err == ENOENT || err == ENXIO || err == ENODEV) &&
            evsel->attr.type   == PERF_TYPE_HARDWARE &&
            evsel->attr.config == PERF_COUNT_HW_CPU_CYCLES) {
@@ -2363,6 +2365,22 @@ bool perf_evsel__fallback(struct perf_evsel *evsel, int err,
                evsel->attr.config = PERF_COUNT_SW_CPU_CLOCK;
 
                zfree(&evsel->name);
+               return true;
+       } else if (err == EACCES && !evsel->attr.exclude_kernel &&
+                  (paranoid = perf_event_paranoid()) > 1) {
+               const char *name = perf_evsel__name(evsel);
+               char *new_name;
+
+               if (asprintf(&new_name, "%s%su", name, strchr(name, ':') ? "" : ":") < 0)
+                       return false;
+
+               if (evsel->name)
+                       free(evsel->name);
+               evsel->name = new_name;
+               scnprintf(msg, msgsize,
+"kernel.perf_event_paranoid=%d, trying to fall back to excluding kernel samples", paranoid);
+               evsel->attr.exclude_kernel = 1;
+
                return true;
        }