On Mon, Sep 05, 2022 at 10:24:40AM +0530, Athira Rajeev wrote:
> The affinity code in "affinity_set" function access array
> named "sched_cpus". The size for this array is allocated in
> affinity_setup function which is nothing but value from
> get_cpu_set_size. This is used to contain the cpumask value
> for each cpu. While setting bit for each cpu, it calls
> "set_bit" function which access index in sched_cpus array.
> If we provide a command-line option to -C which is more than
> the number of CPU's present in the system, the set_bit could
> access an array member which is out-of the array size. This
> is because currently, there is no boundary check for the CPU.
> This will result in seg fault:
>
> <<>>
> ./perf stat -C 12323431 ls
> Perf can support 2048 CPUs. Consider raising MAX_NR_CPUS
> Segmentation fault (core dumped)
> <<>>
>
> Fix this by adding boundary check for the array.
>
> After the fix from powerpc system:
>
> <<>>
> ./perf stat -C 12323431 ls 1>out
> Perf can support 2048 CPUs. Consider raising MAX_NR_CPUS
>
> Performance counter stats for 'CPU(s) 12323431':
>
> <not supported> msec cpu-clock
> <not supported> context-switches
> <not supported> cpu-migrations
> <not supported> page-faults
> <not supported> cycles
> <not supported> instructions
> <not supported> branches
> <not supported> branch-misses
>
> 0.001192373 seconds time elapsed
> <<>>
>
> Reported-by: Nageswara Sastry <rnsas...@linux.ibm.com>
> Signed-off-by: Athira Rajeev <atraj...@linux.vnet.ibm.com>
> ---
> tools/perf/util/affinity.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/affinity.c b/tools/perf/util/affinity.c
> index 4d216c0dc425..a1dd37347abc 100644
> --- a/tools/perf/util/affinity.c
> +++ b/tools/perf/util/affinity.c
> @@ -49,8 +49,14 @@ void affinity__set(struct affinity *a, int cpu)
> {
> int cpu_set_size = get_cpu_set_size();
>
> - if (cpu == -1)
> + /*
> + * Return:
> + * - if cpu is -1
> + * - restrict out of bound access to sched_cpus
> + */
> + if (cpu == -1 || ((cpu / __BITS_PER_LONG) >= (cpu_set_size / 8)))
hm, there's __BITS_PER_LONG in one case, but then there's hardcoded 8
would this be simpler:
if (cpu == -1 || ((cpu >= (cpu_set_size * 8))))
return;
jirka
> return;
> +
> a->changed = true;
> set_bit(cpu, a->sched_cpus);
> /*
> --
> 2.35.1
>
