Re: [PATCH v4] af_netlink: ensure that NLMSG_DONE never fails in dumps

[Johannes Berg <johannes@sipsolutions.net>]

On Sat, 2017-11-11 at 23:09 +0900, David Miller wrote:
> From: "Jason A. Donenfeld" <Jason@zx2c4.com>
> Date: Thu,  9 Nov 2017 13:04:44 +0900
> 
> > @@ -2195,13 +2197,15 @@ static int netlink_dump(struct sock *sk)
> >               return 0;
> >       }
> >  
> > -     nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
> > -     if (!nlh)
> > +     nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE,
> > +                            sizeof(nlk->dump_done_errno), NLM_F_MULTI);
> > +     if (WARN_ON(!nlh))
> >               goto errout_skb;
> 
> If you're handling this by forcing another read() to procude the
> NLMSG_DONE, then you have no reason to WARN_ON() here.
> 
> In fact you are adding a WARN_ON() which is trivially triggerable by
> any user.

I added this in my suggestion for how this could work, but I don't
think you're right, since we previously check if there's enough space.
The patch is missing the full context, but this is:


+       if (nlk->dump_done_errno > 0 ||
+           skb_tailroom(skb) < nlmsg_total_size(sizeof(nlk->dump_done_errno))) {
                mutex_unlock(nlk->cb_mutex);
 
                if (sk_filter(sk, skb))
                        kfree_skb(skb);
                else
                        __netlink_sendskb(sk, skb);
                return 0;
        }

-       nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
-       if (!nlh)
+       nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE,
+                              sizeof(nlk->dump_done_errno), NLM_F_MULTI);
+       if (WARN_ON(!nlh))


So unless the nlmsg_total_size() vs. nlmsg_put_answer() suddenly gets a
different idea of how much space is needed, nlh shouldn't ever be NULL
once we get here.

johannes