Re: [PATCH v3 4/6] perf/util/session: Replace MAX_NR_CPUS with nr_cpus_online

From: Jiri Olsa
Date: Tue Aug 20 2019 - 05:18:57 EST


On Mon, Aug 19, 2019 at 03:23:33PM -0500, Kyle Meyer wrote:
> nr_cpus_online, the number of CPUs online during a record session, can be
> used as a dynamic alternative for MAX_NR_CPUS in perf_session__cpu_bitmap.
>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
> Cc: Alexander Shishkin <alexander.shishkin@xxxxxxxxxxxxxxx>
> Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
> Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> Cc: Russ Anderson <russ.anderson@xxxxxxx>
> Signed-off-by: Kyle Meyer <kyle.meyer@xxxxxxx>
> ---
> tools/perf/util/session.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> Index: tip/tools/perf/util/session.c
> ===================================================================
> --- tip.orig/tools/perf/util/session.c
> +++ tip/tools/perf/util/session.c
> @@ -2284,6 +2284,7 @@ int perf_session__cpu_bitmap(struct perf
> {
> int i, err = -1;
> struct perf_cpu_map *map;
> + int nr_cpus_online = session->header.env.nr_cpus_online;

so all those bitmaps that use this function are initialized with
MAX_NR_CPUS length, are we sure that session->header.env.nr_cpus_online
is always smaller than MAX_NR_CPUS?

jirka

>
> for (i = 0; i < PERF_TYPE_MAX; ++i) {
> struct evsel *evsel;
> @@ -2308,9 +2309,8 @@ int perf_session__cpu_bitmap(struct perf
> for (i = 0; i < map->nr; i++) {
> int cpu = map->map[i];
>
> - if (cpu >= MAX_NR_CPUS) {
> - pr_err("Requested CPU %d too large. "
> - "Consider raising MAX_NR_CPUS\n", cpu);
> + if (cpu >= nr_cpus_online) {
> + pr_err("Requested CPU %d too large\n", cpu);
> goto out_delete_map;
> }
>