Re: [PATCH 05/18] perf evlist: Do not use hard coded value for a mmap_pages default

[Jiri Olsa]

On Tue, Dec 16, 2014 at 01:57:07PM -0300, Arnaldo Carvalho de Melo wrote:
> From: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
> 
> 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@xxxxxxxxx>
> Cc: Borislav Petkov <bp@xxxxxxx>
> Cc: David Ahern <dsahern@xxxxxxxxx>
> Cc: Don Zickus <dzickus@xxxxxxxxxx>
> Cc: Frederic Weisbecker <fweisbec@xxxxxxxxx>
> Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
> Cc: Mike Galbraith <efault@xxxxxx>
> Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
> Cc: Paul Mackerras <paulus@xxxxxxxxx>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: Stephane Eranian <eranian@xxxxxxxxxx>
> Link: http://lkml.kernel.org/n/tip-2f6mtm8xu3wo5lhkql6jdblh@xxxxxxxxxxxxxx
> Signed-off-by: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
> ---
>  tools/perf/util/evlist.c | 20 ++++++++++++++++----
>  1 file changed, 16 insertions(+), 4 deletions(-)
> 
> diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
> index 7847f3885081..ac808680e61c 100644
> --- a/tools/perf/util/evlist.c
> +++ b/tools/perf/util/evlist.c
> @@ -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);
> +		}

so this way you depend on value in perf_event_mlock_kb being power 2,
otherwise:

# echo 1000 > /proc/sys/kernel/perf_event_mlock_kb
$ perf record ls
failed to mmap with 22 (Invalid argument)

maybe you could use the logic/code from parse_pages_arg function

jirka
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/