Re: [lwip-users] Repeated DHCP requests.

[Rogier Wolff]

On Fri, Oct 06, 2023 at 11:29:43AM -0500, Nathan Martin wrote:
> Hi Rogier,
> 
> You should enable the DHCP log messages in the project's lwipopts.h file by
> setting the following defines:
> 
> #define LWIP_DEBUG
> #define DHCP_DEBUG LWIP_DBG_ON
> 
> The DHCP debug messages should give you an idea as to what lwIP's DHCP is
> doing and hopefully why.
> 
> It is possible that your DHCP server has short term IP address leases
> configured.

No, about(*) a day IIRC. And if that were the case, all others on my
network would also be "DHCP-ing" a lot. This issue was noticed when my
logs had hundreds of my test-project DHCPing repeatedly (with just
the normal amount of other devices inbetween).

> And as to why multiple times per second, it is normal to initially send out
> a few discover packets in case the first packet doesn't arrive to the DHCP
> server as everything is exchanged as UDP broadcasts. The acquiring DHCP
> lease does involve multiple packets being sent and received.
> 
> The DHCP code is located in lwip-dir/src/core/ipv4/dhcp.c

At one point I thought the problem had gone away and decided to leave
it at that.... But it hadn't. 

I debugged it my own way and found that for the DHCP code it seems to
start with

  dhcp_network_changed 1856.940025

being called (1856 is the timestamp (in seconds since boot). 

That seems to be called from src/core/netif.c (that's the only call I
found), and that in turn doesn't seem to be called except when loop or
a few other specific devices come up.

Ok... Just added a debug message to netif_set_link_up in
src/core/netif.c and... Nope: dhcp_network_changed is NOT called from
there, at least I'm not seeing the debug message, and I don't see any
other place where it gets called.

Checked the disassembly of the binary and netif_set_link_up is
the only place where dhcp_network_changed is being called. 

And I see the 
  bl      1000feb4 <debug_calls>

that would generate my debug message in the disassembly for
netif_set_link_up. (i.e. I didnt put the debug message in some #ifdef
part that didn't get compiled.)

Further suggestions? 

        Roger. 



(*) So half a day, a day or two days. Something like that. Not 'about
a day" as in "23 hours".

> 
>     -Nathan Martin
> 
> 
> On 10/6/2023 6:51 AM, Rogier Wolff wrote:
> > Hi all,
> > 
> > I have a project, and I noticed a while back that it would repeatedly
> > repeat its DHCP request.
> > 
> > Yesterday I had intended to solve this and also do some other
> > maintenance on the code. I started with the latter and... the DHCP
> > problem disappeared.
> > 
> > However, this morning it had started again.
> > 
> > Now you might say: it is supposed to re-validate the DHCP lease every
> > now and then. And I'd agree. However multiple times a second is WAY
> > too much.
> > 
> > I have now reverted the board to more stock "example" code: Way less
> > of my code in there. But it seems that now the problem only
> > happens after a while.
> > 
> > This is running on an RP2040 with hardwired ethernet through a
> > lan8720 PHY.
> > 
> > Anybody seen such behaviour? What was the problem then?
> > 
> > Why would LWIP restart the DHCP sequence? What could cause this?
> > Where would I look to find the code responsible?
> > 
> >     Roger.
> > 
> 
> 
> _______________________________________________
> lwip-users mailing list
> lwip-users@nongnu.org
> https://lists.nongnu.org/mailman/listinfo/lwip-users

-- 
** R.E.Wolff@BitWizard.nl ** https://www.BitWizard.nl/ ** +31-15-2049110 **
**    Delftechpark 11 2628 XJ  Delft, The Netherlands.  KVK: 27239233    **
f equals m times a. When your f is steady, and your m is going down
your a is going up.  -- Chris Hadfield about flying up the space shuttle.