Re: [PATCH v2] zram: fix possible use after free in zcomp_create()

[Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>]

On (09/08/15 10:07), Minchan Kim wrote:
[..]
> > +	int ret;
> 
> For the clarification, I want to call it as 'error' instead of ret.
> 
> >  
> >  	backend = find_backend(compress);
> >  	if (!backend)
> > @@ -347,10 +348,10 @@ struct zcomp *zcomp_create(const char *compress, int max_strm)
> >  
> >  	comp->backend = backend;
> >  	if (max_strm > 1)
> > -		zcomp_strm_multi_create(comp, max_strm);
> > +		ret = zcomp_strm_multi_create(comp, max_strm);
> >  	else
> > -		zcomp_strm_single_create(comp);
> > -	if (!comp->stream) {
> > +		ret = zcomp_strm_single_create(comp);
> > +	if (ret) {
> >  		kfree(comp);
> >  		return ERR_PTR(-ENOMEM);
> >  	}
> 
> And we could return ERR_PTR(error) rather than fixed -ENOMEM to propagate

yep, I thought about that; looks to be minor so I didn't insist
on changing that. don't mind to change it to ERR_PTR(error).

> other errors potentially could be happen in future(ex, crypto support).
> Of course, we should change description of the function about error return.

on the other hand, this is zcomp_strm_FOO_create(), which is mostly
about memory allocation. but yeah, who knows. We can change this as
a part of crypto api rework; ERR_PTR(error) does not fix anything
per se, so may be we can avoid pushing this change into stable.

	-ss