Re: general protection fault in af_alg_free_areq_sgls

[Eric Biggers]

On Tue, Nov 28, 2017 at 10:10:55AM +0100, Stephan Mueller wrote:
> > diff --git a/crypto/af_alg.c b/crypto/af_alg.c
> > index 358749c38894..415a54ced4d6 100644
> > --- a/crypto/af_alg.c
> > +++ b/crypto/af_alg.c
> > @@ -672,14 +672,15 @@ void af_alg_free_areq_sgls(struct af_alg_async_req
> > *areq) }
> > 
> >  	tsgl = areq->tsgl;
> > -	for_each_sg(tsgl, sg, areq->tsgl_entries, i) {
> > -		if (!sg_page(sg))
> > -			continue;
> > -		put_page(sg_page(sg));
> > -	}
> > +	if (tsgl) {
> > +		for_each_sg(tsgl, sg, areq->tsgl_entries, i) {
> > +			if (!sg_page(sg))
> > +				continue;
> > +			put_page(sg_page(sg));
> > +		}
> > 
> > -	if (areq->tsgl && areq->tsgl_entries)
> 
> Why do you want to remove the check for areq->tsgl_entries? I know in the 
> current code that cannot happen. But it should be caught in case of a 
> programming error.
> 
> Thus, should we add a BUG_ON(!areq->tsgl_entries)?
> 

sock_kfree_s() works even if the size is 0.  So there's no reason to check.

Eric