Re: [PATCH v2 bpf-next] libbpf: Use sysconf to simplify libbpf_num_possible_cpus

[Andrii Nakryiko]

On Wed, Sep 22, 2021 at 2:22 PM Daniel Borkmann <daniel@xxxxxxxxxxxxx> wrote:
>
> On 9/22/21 9:07 AM, Muhammad Falak R Wani wrote:
> > Simplify libbpf_num_possible_cpus by using sysconf(_SC_NPROCESSORS_CONF)
> > instead of parsing a file.
> > This patch is a part ([0]) of libbpf-1.0 milestone.
> >
> > [0] Closes: https://github.com/libbpf/libbpf/issues/383
> >
> > Signed-off-by: Muhammad Falak R Wani <falakreyaz@xxxxxxxxx>
> > ---
> >   tools/lib/bpf/libbpf.c | 17 ++++-------------
> >   1 file changed, 4 insertions(+), 13 deletions(-)
> >
> > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> > index ef5db34bf913..f1c0abe5b58d 100644
> > --- a/tools/lib/bpf/libbpf.c
> > +++ b/tools/lib/bpf/libbpf.c
> > @@ -10898,25 +10898,16 @@ int parse_cpu_mask_file(const char *fcpu, bool **mask, int *mask_sz)
> >
> >   int libbpf_num_possible_cpus(void)
> >   {
> > -     static const char *fcpu = "/sys/devices/system/cpu/possible";
> >       static int cpus;
> > -     int err, n, i, tmp_cpus;
> > -     bool *mask;
> > +     int tmp_cpus;
> >
> >       tmp_cpus = READ_ONCE(cpus);
> >       if (tmp_cpus > 0)
> >               return tmp_cpus;
> >
> > -     err = parse_cpu_mask_file(fcpu, &mask, &n);
> > -     if (err)
> > -             return libbpf_err(err);
> > -
> > -     tmp_cpus = 0;
> > -     for (i = 0; i < n; i++) {
> > -             if (mask[i])
> > -                     tmp_cpus++;
> > -     }
> > -     free(mask);
> > +     tmp_cpus = sysconf(_SC_NPROCESSORS_CONF);
> > +     if (tmp_cpus < 1)
> > +             return libbpf_err(-EINVAL);
>
> This approach is unfortunately broken, see also commit e00c7b216f34 ("bpf: fix
> multiple issues in selftest suite and samples") for more details:

Oh, that predates me. Thanks, Daniel!

Sorry, Muhammad, seems like current implementation is there for a
reason and will have to stay. Thanks a lot for working on this,
though. Hopefully you can help with other issues, though.

[...]

>
> Thanks,
> Daniel
>
> >       WRITE_ONCE(cpus, tmp_cpus);
> >       return tmp_cpus;
> >
>