Some update:

I've enabled "LWIP_NETCONN_FULLDUPLEX" - needed some work as it requires
LWIP_NETCON_SEM_PER_THREAD.

So I needed to implement(getting a unique semaphore per thread):

   - LWIP_NETCONN_THREAD_SEM_GET()
   - LWIP_NETCONN_THREAD_SEM_ALLOC()
   - LWIP_NETCONN_THREAD_SEM_FREE()


Now I cant even reach the corruption issue(not sure if solved or not), as I
constantly getting another assert after a while : "Assertion "sock->fd_used
!= 0" failed at line 344 in lwip/src/api/sockets.c.".

I'm guessing thats why this feature is in alpha level :)





On Tue, Nov 7, 2017 at 7:09 PM, Itzik Levi <[email protected]> wrote:

> Hi Sylvain,
>
> Although I am reading and writing from separated threads, I am not doing
> that in parallel(mutex protected), lwip_recv is never called in parallel
> to lwip_send.
> I will try and enable LWIP_NETCONN_FULLDUPLEX to see if it does anything,
> but I remember its highly experimental at this point.
>
> Will report back.
>
> On Tue, Nov 7, 2017 at 7:02 PM, Sylvain Rochet <[email protected]>
> wrote:
>
>> Hi Itzik,
>>
>> On Tue, Nov 07, 2017 at 06:40:52PM +0200, Itzik Levi wrote:
>> > Hi Sylvain,
>> >
>> > Thanks for the response!
>> >
>> > Just tried disabling both CORE_LOCKING and CORE_LOCKING_INPUT and
>> > unfortunately encountered the same problem.
>> >
>> > I'm probably missing something, but what... loss-less stream works
>> > excellent, without any corruption, might be some kind of internal tcp
>> queue
>> > corruption while congested?
>>
>> It looks like a thread safety issue, but you seem to do things well
>> (PPPAPI, pppos_input_tcpip, socket API, rx/tx mutex), but could you try
>> setting LWIP_NETCONN_FULLDUPLEX since you are using rx and tx from
>> multiple threads ?
>>
>> Sylvain
>>
>> _______________________________________________
>> lwip-users mailing list
>> [email protected]
>> https://lists.nongnu.org/mailman/listinfo/lwip-users
>>
>
>
_______________________________________________
lwip-users mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/lwip-users

Reply via email to