I borrowed a trick from the TI lwIP implementation. I set the ethernet Rx interrupt flag periodically to cause an interrupt where timeouts are checked, and it checks for Rx packets. Things work pretty well with 2.0.3, no out of memory errors.
On Wednesday, March 28, 2018, 7:01:34 AM CDT, Sergio R. Caprile <address@hidden> wrote:
... and talking about memory getting freed slower than it is alloced... since HTTP runs on top of TCP and you said "there is only one entry point in the Rx", I wonder, how do you call sys_check_timeouts()? TCP relies on that. If you don't have a main loop, you need to manage to call it frequently without interrupting the Rx ISR.