Re: [lwip-users] Mailboxes and Semaphores

[address@hidden]

Timmy Brolin wrote:
> How about mailboxes? Does LwIP require multiple waiting threads on a
> mailbox? That would require some substantial rewriting of our OS mailbox
> system.
>   

I don't think that's necessary: mailboxes are currently used
a) for the tcpip_thread to process input packets and other actions from 
other threads in the main lwip thread and
b) for queueing input packets on normal connections / newly accepted 
connections on tcp listening connections

Since lwip does not support multiple threads to use one socket/netconn 
simultaneously, you won't have any problem here.

However, this is not an official statement about mbox requirements, so 
it could be (although not very likely) that things change, especially 
when it comes to multithread-usage of sockets. But even if so, you might 
still be able to avoid those situations. After all, having multiple 
listeners on an mbox does not make sense to me in the situation of this 
stack, as results are not predictable.

Simon

> I think it would be pretty straight forward to add support for
> semaphore/mailbox deallocation to our OS, so that is not a very big issue.
>
> Regards,
> Timmy Brolin
>
> address@hidden wrote:
>   
> > If I remember correctly, there are many places in the stack where your
> > simple semaphore can be used (tcpip_thread, netconn semaphores, etc.)
> > but there are also other usages where multiple waiting threads have to
> > be supported (i.e. locking the heap in mem_malloc). So I guess you are
> > right that the generic sys_sem_t must be the multiple-type.
> >
> > Speaking of deallocation: I also ran lwIP on an OS which didn't
> > support this. I ended up in having a preallocated list/array of
> > mboxes/semaphores and returning an instance of that list in
> > sys_sem_new().
> >
> > Simon
> >
> >
> > Timmy Brolin wrote:
> >     
> > > Hello,
> > > I have A few questions about LwIP OS requirements..
> > >
> > > Our operating system provides two kinds of semaphores. The normal (and
> > > resource-light) semaphore only allows a single task to block on it at
> > > any one time. There is also a second type which allows multiple tasks to
> > > block on it at the same time, but it requires a bit more memory
> > > resources.
> > > Am I correct in assuming that LwIP requires multiple tasks to block on
> > > the same semaphore? Are there any semaphores in LwIP which could do with
> > > the simpler semaphore type?
> > >
> > > The mailbox system in our operating system is based on the standard
> > > semaphores, so only one task may block on a mailbox at any one time. Is
> > > this acceptable for LwIP?
> > >
> > > Our OS has a limitation in that semaphores and mailboxes never can be
> > > deallocated, and I was hoping that the addition of a deallocation
> > > funtion for semaphores and mailboxes would be enough to get LwIP running
> > > on our system. I would prefer not having to write a new mailbox system
> > > as well.. :-)
> > >
> > > Regards,
> > > Timmy Brolin
> > >
> > >
> > > _______________________________________________
> > > lwip-users mailing list
> > > address@hidden
> > > http://lists.nongnu.org/mailman/listinfo/lwip-users
> > >
> > >   
> > >       
> >
> > _______________________________________________
> > lwip-users mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/lwip-users
> >
> >     
>
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users
>
>