Re: [PATCH ipsec-next v12 1/4] xfrm: Add Direction to the SA in or out

[Antony Antony <antony@phenome.org>]

On Wed, Apr 24, 2024 at 11:38:08AM +0200, Sabrina Dubroca via Devel wrote:
> 2024-04-23, 14:49:17 +0200, Antony Antony wrote:
> > diff --git a/net/xfrm/xfrm_state.c b/net/xfrm/xfrm_state.c
> > index 0c306473a79d..c8c5fc47c431 100644
> > --- a/net/xfrm/xfrm_state.c
> > +++ b/net/xfrm/xfrm_state.c
> > @@ -1292,6 +1292,7 @@ xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
> >  		if (km_query(x, tmpl, pol) == 0) {
> >  			spin_lock_bh(&net->xfrm.xfrm_state_lock);
> >  			x->km.state = XFRM_STATE_ACQ;
> > +			x->dir = XFRM_SA_DIR_OUT;
> 
> Would that make updates fail if userspace isn't setting SA_DIR
> afterwards?

Not in typical, operation of IKE daemons. I haven't update acquire state.
I don't think iproute2 supports that.
UPDSA is called for input where there is no acquire; acquire is for output.  
For output SA, added using NEWSA, will find the output acquire and delete it 
if p->seq matches. It won't update.

May be some other userspace update acquire? I extend the code to accomodate 
this possible case, in v13.

> > diff --git a/net/xfrm/xfrm_user.c b/net/xfrm/xfrm_user.c
> > index 810b520493f3..d34ac467a219 100644
> > --- a/net/xfrm/xfrm_user.c
> > +++ b/net/xfrm/xfrm_user.c
> [...]
> > @@ -176,6 +200,7 @@ static int verify_newsa_info(struct xfrm_usersa_info *p,
> >  			     struct netlink_ext_ack *extack)
> >  {
> >  	int err;
> > +	u8 sa_dir = attrs[XFRMA_SA_DIR] ?  nla_get_u8(attrs[XFRMA_SA_DIR]) : 0;
> 
> nit: extra ' ' after '?', only one is needed.

fixed.

> > @@ -358,6 +383,64 @@ static int verify_newsa_info(struct 
> > xfrm_usersa_info *p,
> >  			err = -EINVAL;
> >  			goto out;
> >  		}
> > +
> > +		if (sa_dir == XFRM_SA_DIR_OUT) {
> > +			NL_SET_ERR_MSG(extack,
> > +				       "MTIMER_THRESH attribute should not be set on output SA");
> > +			err = -EINVAL;
> > +			goto out;
> > +		}
> > +	}
> > +
> > +	if (sa_dir == XFRM_SA_DIR_OUT) {
> > +		if (p->flags & XFRM_STATE_DECAP_DSCP) {
> > +			NL_SET_ERR_MSG(extack, "Flag NDECAP_DSCP should not be set for output SA");
> 
> That typo in the error string is still here (extra N in flag name).

sorry.Fixed. I will send v13 soon.

-antony