Re: [lwip-users] PPP proxy arp support [fixed typo]

[Patrick Klos]

On 10/10/2018 8:59 AM, Patrick Klos wrote:
> On 10/9/2018 8:24 PM, Bob Jones wrote:
> > Hello,
> >
> > I'm trying to resolve an issue with regards to PPP and ARP requests. 
> > My network topology is as follows,
> >
> > Laptop (192.168.1.3, ethernet) <-> MCU #1 (192.168.1.4, ethernet) <-> 
> > MCU #1 (192.168.1.126, PPP serial [server]) <-> MCU #2 (192.168.1.127 
> > PPP serial [client])
> >
> > The issue I'm running into is that when I try to send a packet to MCU 
> > #2 (192.168.1.127) from the laptop, the ARP request the laptop sends 
> > is never satisfied. This makes sense to me, as the instance of lwip 
> > running on MCU #1 can't respond to this ARP request as it doesn't 
> > have MCU #2's IP address in its routing table (and shouldn't). In 
> > this case, one solution is to add a static route to the laptop's 
> > routing table to send all requests destined for 192.168.1.127 out of 
> > the interface corresponding to 192.168.1.4. Voila, no ARP request for 
> > 192.168.1.127, no problem.
> >
> > Unfortunately, however, our team is working with some antiquated 
> > hardware in place of the "laptop" in the above example that doesn't 
> > support adding static routes.
> >
> > So, in doing some Googling around and reading of past posts on this 
> > forum, it seems like layer 2 forwarding is one possible solution here 
> > (MCU #1 would be responsible for this). I was thinking of using PPP 
> > proxy arp support on MCU #1, but noticed that it was commented out of 
> > the lwip ppp implementation in 2015 (commit hash 99bcce78...). The 
> > commit message simply reads "PPP, IPCP, removed proxy ARP support". 
> > Was curious to know if this is something that, if the surrounding 
> > `#define`s were removed, would likely just work, or if there was some 
> > underlying implementation issue that led to its removal. Also, any 
> > other ideas about how this problem might be resolved otherwise, given 
> > the constraints on network topology?
>
> Hello Bob,
>
> What version of LwIP are you using?  What are your subnet masks on the 
> various devices?
>
> Scenario #1:
>     The subnet mask includes all the devices in question (i.e. 
> 192.168.1.0/24).  In this case, MCU#1 will have to answer ARP requests 
> for MCU#2.
>
> Scenario #2:
>     The subnets are truly broken up (i.e. ethernet is 192.168.1.0/22 
> and PPP is 192.168.1.64/22) and MCU#1 is acting as a router 

Ooops... my mistake.  Subnets would 192.168.1.0/26 and 192.168.1.64/26 
(26 bits instead of 22 bits).

> between the 2 subnets.  In this case, the laptop would have MCU#1's IP 
> address in it's route table to 192.168.1.64/22 (or it's default route 
> if necessary).
>
> I'm guessing you're using scenario #1?  If that's the case, when the 
> laptop sends an ARP request for MCU#2's IP address, MCU#1 will need to 
> respond with it's own MAC address (and MCU#2's IP address) in the ARP 
> response.  Maybe that's what the PPP proxy ARP code on MCU#1 is 
> supposed to do - I haven't looked at that code?  If the code is 
> commented out because it doesn't currently work, that doesn't mean it 
> can't be made to work.
>
> You say the "antiquated hardware" has limited configuration 
> capabilities.  What are they?  Can you set it's IP address, subnet 
> mask, and default gateway?  Does it support DHCP or is it manually 
> configured?  Does that "antiquated hardware" just need to talk to 
> MCU#2 or does it need to communicate with other devices on the network 
> (or on the Internet) as well?

Patrick