Re: net: rps: support 802.1Q

[Ben Hutchings <bhutchings@solarflare.com>]

On Fri, 2011-08-19 at 23:05 +0800, Changli Gao wrote:
> On Fri, Aug 19, 2011 at 7:54 PM, Ben Hutchings
> <bhutchings@solarflare.com> wrote:
> >
> > Should this really be reading an unlimited number of tags?
> 
> Not unlimited, but it won't stop until reaching the end of the packet.

Right, I understand that the parsing is properly range-checked against
the length of the packet.

> >  What if an
> > attacker starts sending packets full of VLAN tags?  Since this runs
> > before netfilter, there would be no way to prevent those packets burning
> > our CPU time.  And if there are legitimately multiple VLAN tags, they
> > presumably won't all have the 802.1q Ethertype.
> >
> 
> Do we need to limit the number of rounds to stop this kind of "bad"
> packets from burning our CPU time?

Well, maybe.  Then again, the most effective way for an attacker to
waste a target's CPU time (aside from application-level vulnerabilities)
will often be just to send minimum size packets.

> Then,  __netif_receive_skb() has to
> be update too, so the inspection of tunnel in __skb_get_rxhash() does.

Yes, if we agree this is something worth defending against then we would
need to be consistent in limiting any such parsing loop in pre-netfilter
processing.

> Is there a such limitation in xfrm?

It appears to be limited to 6 levels of encapsulation.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.