rtalloc()/rtfree() problems on CURRENT

[Gleb Smirnoff]

On Tue, Feb 24, 2004 at 01:20:18PM +0100, Andre Oppermann wrote:
A> > Thanks for this hint. So if using rtalloc_ign() I do not need to call rtfree()?
A> You still have to call rtfree() to decrement the refcount of the rtentry.

Thanks.

A> > A> Maybe you do some nasty things to rt here?
A> > 
A> > Really, I don't :) Exact code is given at end of the mail.
A> 
A> You indeed do some nasty things.

Please point me at 'em. I'm only reading some values from parent structure.

A> > Here is exact code:
A> > 
A> >         bzero((caddr_t)&ro, sizeof(ro));
A> >         sin = (struct sockaddr_in *)&ro.ro_dst;
A> >         sin->sin_len = sizeof(*sin);
A> >         sin->sin_family = AF_INET;
A> >         sin->sin_addr = fle->r.r_dst;
A> >         rtalloc(&ro);
A> >         if (ro.ro_rt != NULL) {
A> >                 struct rtentry *rt = ro.ro_rt;
A> > 
A> >                 /* This is cloned route, use its parent */
A> >                 if (ro.ro_rt->rt_flags & RTF_WASCLONED &&
A> >                     ro.ro_rt->rt_parent)
A> >                         rt = ro.ro_rt->rt_parent;
A> 
A> With rtalloc() you get a cloned route essentially every time.  Use
A> rtalloc_ign(&ro, RTF_CLONING) to get the parent route directly.
A> 
A> >                 fle->o_ifx = rt->rt_ifp->if_index;
A> > 
A> >                 if (rt->rt_flags & RTF_GATEWAY &&
A> >                     rt->rt_gateway->sa_family == AF_INET)
A> >                         fle->next_hop =
A> >                             ((struct sockaddr_in *)(rt->rt_gateway))->sin_addr;
A> > 
A> >                 if (rt_mask(rt))
A> >                         fle->dst_mask =
A> >                             bit_count(((struct sockaddr_in *)rt_mask(rt))->sin_addr.s_addr);
A> >                 else if (rt->rt_flags & RTF_HOST)
A> >                         /* Give up. We can't determine mask :( */
A> >                         fle->dst_mask = 32;
A> > 
A> >                 rtfree(ro.ro_rt);
A> 
A> Use the macro RTFREE() instead of rtfree(), it will take care of some
A> locking issues.

This should be present in manpage, too.


-- 
Totus tuus, Glebius.
GLEBIUS-RIPN GLEB-RIPE