On Mon, Aug 16, 2021 at 3:53 PM Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Mon, Aug 16, 2021 at 03:38:29PM +0800, Dongliang Mu wrote:
> >
> > - for_each_possible_cpu(i)
> > - vfree(*per_cpu_ptr(scratches, i));
> > + for_each_possible_cpu(i) {
> > + void *scratch = *per_cpu_ptr(scratches, i);
> > + if (!scratch)
> > + vfree(scratch);
> > + }
>
> This patch is unnecessary. Please check the implementation of
> vfree, it already checks for NULL pointers just like most of our
> free primitives.
Hi Herbert,
since there is no reproducer in the syzbot, I guess the problem might be:
if vmalloc_node in the ipcomp_alloc_scratches returns a NULL pointer,
it directly returns NULL without updating the per_cpu_ptr(scratches,
i).
Therefore, in the ipcomp_free_scratches, vfree will take an invalid
and outdated per_cpu_ptr as its argument, leading to the bug - BUG:
unable to handle kernel paging request in ipcomp_free_scratches.
Any idea?
>
> Cheers,
> --
> Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>
> Home Page: http://gondor.apana.org.au/~herbert/
> PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
