Re: Re: Remotely exploitable crash in dhcpcd
Jason A. Donenfeld
Mon Jun 26 11:57:00 2017
I saw this via tcpdump. I'm away til Weds but will send more extensive
debugging when home.
On Jun 26, 2017 13:07, "Roy Marples" <roy@xxxxxxxxxxxx> wrote:
> Hi Jason
>
> On 25/06/2017 17:48, Jason A. Donenfeld wrote:
> > Both interfaces now get both address, which is great.
> >
> > However, I've found two small bugs in this initial stab at the
> implementation:
> >
> > - eth0 is metric 200
> > - wlan0 is metric 300
> > - set eth0 up --> eth0 gets 10.10.10.100
> > - set wlan0 up --> wlan0 gets 10.10.10.100
> > - unplug eth0 --> ARP sent wlan0 so smooth transition
> > - plug eth0 --> ARP sent on wlan0 instead of on eth0, so transition is
> > not smooth, until gateway does its own ARP
> >
> > and
> >
> > - eth0 is metric 200
> > - wlan0 is metric 300
> > - set eth0 up --> eth0 gets 10.10.10.100
> > - set wlan0 up --> wlan0 gets 10.10.10.100
> > - unplug wlan0 --> ARP sent on eth0 even though it isn't needed, I think
> > - plug wlan0 --> ARP sent on wlan0, which is incorrect, so pings of
> > eth0 get interrupted temporarily
>
> I can't replicate this.
>
> If you run `dhcpcd -dB4` in a window, you'll see debug logs across all
> interfaces as they happen.
> If you do that and go through the above motions, do you see ARP
> announcements on the correct interfaces? If not, please post your
> dhcpcd.conf, if so can you verify they are leaving the correct interface
> via tcmpdump?
>
> Also, what kernel version are you running?
>
> Roy
>
Archive administrator: postmaster@marples.name