RE: [lwip-devel] Problem with netconn_write()

[Mountifield, Tony]

Hi Peter,

Thanks for your comments...

> Years ago, I ran into a timinig-dependent problem with 
> netconn_write(), which has probably the same roots.
> IIRC I came to the same conclusion, 
> the semaphore should exist until netconn_close().

I see that netconn_delete() also deletes the semaphore if it still exists, 
which is a much more logical place, so I have commented out the semaphore 
deletion in netconn_write(). Now my application works fine.

> >In any case, assuming sys_sem_signal() and sys_sem_free() 
> >can't block, there should perhaps be some exclusion to make 
> >sure that a test of conn->sem and its subsequent use are 
> >indivisble, for semaphores that can be dynamically deleted.
> >  
> >
> A usual implementation of sys_sem_signal()  is likely to 
> leave through the OSes scheduler, switching to a different
> task, so I guess the check and the call can not be made
> indivisible by a simple solution like the "lightweight
> protection" here (which disables task-switching).

Yes, you're right.

> I hesitate to propose a more complicated solution. The best
> would be to remove the need for the check.

Yes, I agree.

Cheers
Tony


***********************************************************************************
This email, its content and any attachments is PRIVATE AND
CONFIDENTIAL to TANDBERG Television. If received in error please
notify the sender and destroy the original message and attachments.

www.tandbergtv.com
***********************************************************************************