Re: [lwip-users] Packets stop sending after a while when MEMP_OVERFLOW_CHECK isn't 2

[address@hidden]

On 25.04.2018 20:41, Sebastian Gniazdowski wrote:
> [..]
> However, now I've discovered that if I change the *_CHECK define to 0 
> or 1, then ca. 13000 packets are correctly sent, and after that the 
> only packets sent are frequent (in each second) ARP queries (discovery 
> of gate's MAC), as captured by Wireshark, confirmed by MCU's Ethernet 
> transfer diode.
>
> Why such correlation between a debug option and crash-behavior? What 
> can be wrong?

I'd say you're violating lwIP's concurrency/threading requirements and 
the slowdown of MEMP_OVERFLOW_CHECK==2 hides this.

See http://www.nongnu.org/lwip/2_0_x/pitfalls.html
or try the new LWIP_ASSERT_CORE_LOCKED() check macro. Although for that, 
you need a recent git master version and need to implement the check 
correctly...

Simon