perf evlist: Do not use hard coded value for a mmap_pages default
authorArnaldo Carvalho de Melo <acme@redhat.com>
Fri, 12 Dec 2014 20:46:45 +0000 (17:46 -0300)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Tue, 16 Dec 2014 16:38:26 +0000 (13:38 -0300)
So far what is in there by default is what we were using: 512KB + the
control page, but the admin may change that, and if it does to a smaller
value, all calls to tooling for non root users start failing, requiring
that the user manually set --mmap_pages/-m.

Use instead what is in /proc/sys/kernel/perf_event_mlock_kb.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-2f6mtm8xu3wo5lhkql6jdblh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/util/evlist.c

index 7847f388508119e469dfbec09998fe9727c117b0..ac808680e61c7e34713a8dc10d346c6be86c0b3c 100644 (file)
@@ -893,10 +893,22 @@ out_unmap:
 
 static size_t perf_evlist__mmap_size(unsigned long pages)
 {
-       /* 512 kiB: default amount of unprivileged mlocked memory */
-       if (pages == UINT_MAX)
-               pages = (512 * 1024) / page_size;
-       else if (!is_power_of_2(pages))
+       if (pages == UINT_MAX) {
+               int max;
+
+               if (sysctl__read_int("kernel/perf_event_mlock_kb", &max) < 0) {
+                       /*
+                        * Pick a once upon a time good value, i.e. things look
+                        * strange since we can't read a sysctl value, but lets not
+                        * die yet...
+                        */
+                       max = 512;
+               } else {
+                       max -= (page_size / 1024);
+               }
+
+               pages = (max * 1024) / page_size;
+       } else if (!is_power_of_2(pages))
                return 0;
 
        return (pages + 1) * page_size;