Re: [lwip-users] upgrade to 0.6.2

[Ed Sutter]

Dave,
Thanks for the response.
My priorities look ok (tcpip_thread is higher than all others).
I'm finding that lwip_close() is hanging up in netconn_delete()
because netconn_delete() is draining the receive mailbox by
looping on a non-zero return from sys_arch_mbox_fetch().

At this point in my system, sys_arch_mbox_fetch() is always
returning 0xffffffff because there is no mail so it times out.
I haven't looked at this code much, but it seems like there
should be a differentiation between 0xffffffff and all other
non-zero values returned here.  However, since this works for
everyone else, I must have a bug in my sys_arch_mbox_fetch()
return code logic.

Ed
    
David Haas wrote:
> 
> Hi Ed,
> 
> Ed Sutter wrote:
> 
> >In sys_arch.c...
> >  sys_arch_mbox_fetch() and sys_arch_sem_wait() have a slightly
> >  different return values than in previous releases.
> >In ethernetif.c...
> >  make sure that netif->hwaddr_len is initialized to 6
> >In sys_arch.txt...
> >  just read this and follow the instructions.
> >
> >
> This sounds about right.
> 
> >In my system I have three TCP servers and two UDP servers and with the
> >older version of LWIP I had no known problems.  With the latest version
> >integrated into my system, I now notice that quite frequently when my
> >TCP server is blocked on lwip_recv() it does not consistently detect the
> >fact that the other end (the client) has closed the connection.
> >
> >
> >I had been assuming that a return value of <= 0 from lwip_recv() was
> >a reasonable indication that the other end had closed the connection.
> >Is this wrong, or can anyone think of anything else that I need to do
> >to transition to 0.6.2?
> >
> >
> This should be the way it works. I'm not quite sure why it would'nt
> work. I'm using a version very similar to 0.6.2, since I stopped
> upgrading about then and it has been very very stable. I am also using
> several TCP and UDP servers. I have not seen a problem with hanging
> connections (although I had seen some issues before 0.6.2).
> 
> Just for the heck of it, I just checked the thread priorities in my code
> and some of my servers are running at higher priorities than TCP thread
> and some are running at lower priorities. I actually have'nt tuned my
> system very well yet. But at any rate, priorities did not seem to make a
> difference for me.
> 
> Take a look at the PCB and see what the state is.
> 
> David