Re: [PATCH] net: Check skb->rxhash in gro_receive

[Ben Hutchings <ben@decadent.org.uk>]

[-- Attachment #1: Type: text/plain, Size: 1621 bytes --]

On Mon, 2014-01-13 at 13:37 -0800, Eric Dumazet wrote:
> On Mon, 2014-01-13 at 20:50 +0000, Ben Hutchings wrote:
> 
> > According to the original report of that vulnerability:
> > 
> > > skb_flow_dissect() were used by several places:                         
> > > - packet scheduler that want classify flows                             
> > > - skb_get_rxhash() that will be used by RPS, vxlan, multiqueue          
> > > tap,macvtap packet fanout                                               
> > > - skb_probe_transport_header() which was used for probing transport     
> > > header for DODGY packets                                                
> > > - __skb_get_poff() which will be used by socket filter             
> > 
> > So flow dissector is already part of the attack surface for both local
> > and remote users in common configurations.
> 
> Take a debian or Android distro, and this bug is not hit on 'common
> configurations'. Send them a 'packet of death', they will not hang,
> unless some admin set up RPS/RFS on the incoming device.
[...]

When I investigated the scope of this for Debian, I tried sending a
'packet of death' to a VM and actually triggered the lockup in the TX
path of the *host*, running Debian unstable with Linux 3.11.  I didn't
track down exactly why that was but I think that libvirt's default
networking configuration includes multiqueue devices that use flow
dissector.

Ben.

-- 
Ben Hutchings
Life is what happens to you while you're busy making other plans.
                                                               - John Lennon

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]