Re: [lwip-users] Multi-Threaded netconn API TCP application

[Darius Babrauskas]

Congratulations, Mihai!


From: Pîrvu Mihai
Sent: Wednesday, September 23, 2015 12:58 PM
To: Mailing list for lwIP users
Subject: Re: [lwip-users] Multi-Threaded netconn API TCP application

> Yeah, the problem was that my MEMP_NUM_NETBUF was set to only 2. I have 
> increased this and now it works.

> Thanks for the help guys, keep the good code running :)

> Mihai

On Tue, Sep 22, 2015 at 11:10 PM, Sylvain Rochet <address@hidden> 
wrote:

> > Hi Darius,


On Tue, Sep 22, 2015 at 05:01:27PM +0300, Darius Babrauskas wrote:
> > Hi Mihai,
> >
> > It looks that you have problems with systimeouts. Each tread must have
> > self systimeout stuct.

> I disagree, only the lwIP core thread handle timeouts. Maybe your design
> outside of lwIP require a timeout callout system in each thread but
> that's not the case for lwIP.

Yes, you are right.
About systimeouts I found her(old info):
http://lwip.wikia.com/wiki/Porting_for_an_OS

But I see in version LWIP  1.4.1  lwip not using systimeouts  in sys.c file. 
It's Ok, because user don’t must create socket tread  with sys_thread_new 
function, which create systimeouts struct. And after delete thread not need 
release  systimeouts.  In linux ports, not need use  sys_thread_new , 
because it adding when call currentthread function.




> > create_thread("client", client, (void *)&session_index);
> >
> > //PS: The application runs on Linux atm, using lwip 1.4.0 (last stable 
> > one), the unix port from contrib directory and the clients connect using 
> > netcat <ip> <port>.
> > If you using Linux, why you use LWIP?  Are you want testing?

> Of course he want to test and understand how to use the stack before
> putting it to a microcontroller with boring flash-time and less debug
> capabilities, and he is right to do so ;-)


> > It seem to me, that linux port not very good implemented. Maybe for
> > simple testing.

> You can't claim that without explaining why, could you tell why it is
> "not very good implemented" ?
> Sylvain

Its my opinion, because some year ago my experience was not successful with 
linux port (works but unstable) .

In current linux  port I have a sighting in function "cond_wait" which using 
gettimeofday:
We have depends to system time adjustments (ntp, user ...)
Maybe better solution use CLOCK_MONOTONIC... example her:
http://www.qnx.com/developers/docs/6.5.0/index.jsp?topic=%2Fcom.qnx.doc.neutrino_lib_ref%2Fs%2Fsem_timedwait.html

Darius