Re: [lwip-users] threading issues with sockets

[Åke Forslund]

Another update.

The problem has been found and corrected. A high level debug message had snuck 
in to the registered netif_set_status_callback... this kind of debug message 
would try to echo to the terminal instead of serial out if a terminal is 
connected. And since the callback is called from lwip's tcpip-thread calling 
sockets functions will result in a deadlock.

/Åke

-----Ursprungligt meddelande-----
Från: address@hidden [mailto:address@hidden För Åke Forslund
Skickat: den 20 juni 2012 14:32
Till: Mailing list for lwIP users
Ämne: Re: [lwip-users] threading issues with sockets

I'll try to get some debug-information shortly I'm not sure my debugger will 
let me have stack-traces without a fight... In the meantime I can add to my 
previous statement that my problems seem to originate when the network-handling 
thread does the following:

  netifapi_netif_set_down(emmyNetif);
  ip4_addr_set_u32(&ipAddr, *(unsigned long *)ip);
  netif_set_ipaddr(emmyNetif, &ipAddr);
  netifapi_netif_set_up(emmyNetif); <--Hangs here

If I remove the netifapi_netif_set_down/up()  I get better results, everything 
works and lwip_read() returns as you say. I'll see if a little general cleanup 
of the functions for handling both manually set ip and dhcp will solve my 
troubles.

Thanks for the quick reply.

/Åke

-----Ursprungligt meddelande-----
Från: address@hidden [mailto:address@hidden För Simon Goldschmidt
Skickat: den 20 juni 2012 13:35
Till: Mailing list for lwIP users
Ämne: Re: [lwip-users] threading issues with sockets

"Åke Forslund" <address@hidden> wrote:
> I'm not quite sure how to handle the situation since when using 
> blocking sockets lwip_read() doesn't return even if the link goes down 
> (? is this expected behavior?)

A link-down event shouldn't cause read() to abort, I guess, since you could 
re-plug the device into the same network again.

A netif-down event (which is what stopping DHCP does) however should cause the 
read() to abort since the IP address is gone.

> so I can't use a semaphore to ensure that socket functions can't be 
> accessed at the same time as the netifapi_*()-functions.
> 
> Any advice on how to handle link-status changes together with sockets 
> in a safe manner would be great.

I'm not aware of problems here. As far as I know, there should be no problems 
with what you described. As you are having problems, I don't have any 
suggestions except for debugging to see which threads are pending on which 
semaphores, I'm afraid. A stack trace of all threads at the time of suspected 
deadlock would be helpful to give more hints.

Simon
--
Empfehlen Sie GMX DSL Ihren Freunden und Bekannten und wir belohnen Sie mit bis 
zu 50,- Euro! https://freundschaftswerbung.gmx.de

_______________________________________________
lwip-users mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/lwip-users
_______________________________________________
lwip-users mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/lwip-users