Michael McConville wrote:
> Michael McConville wrote:
> > Does this make sense?
> 
> I just realized that the allocation failure checks earlier in the
> function return ENOBUFS. This probably makes more sense for the sake of
> consistency.

The best I can tell, the only use of this function is in
sys/crypto/crypto.c:157. It's accessed through a pointer stored in a
struct by crypto_register(). That usage doesn't seem to be affected by
the below change, considering that the outcome would be no different
than that of the other ENOBUFS failures above it.

> > Index: sys/crypto/cryptosoft.c
> > ===================================================================
> > RCS file: /cvs/src/sys/crypto/cryptosoft.c,v
> > retrieving revision 1.80
> > diff -u -p -r1.80 cryptosoft.c
> > --- sys/crypto/cryptosoft.c 10 Dec 2015 21:00:51 -0000      1.80
> > +++ sys/crypto/cryptosoft.c 26 Feb 2016 17:21:00 -0000
> > @@ -826,7 +826,7 @@ swcr_newsession(u_int32_t *sid, struct c
> >                                 M_CRYPTO_DATA, M_NOWAIT | M_ZERO);
> >                             if ((*swd)->sw_kschedule == NULL) {
> >                                     swcr_freesession(i);
> > -                                   return EINVAL;
> > +                                   return ENOMEM;
> >                             }
> >                     }
> >                     if (txf->setkey((*swd)->sw_kschedule, cri->cri_key,
> > 
> 

Reply via email to