Re: [lwip-devel] possible issue in sockets.c

[Simon Goldschmidt]

Albert Huitsing wrote:

> as far as I understand things correctly then (note I'm on ARM / FreeRTOS)

> 1) SYS_ARCH_UNPROTECT() will call taskENTER_CRITICAL() and block IRQs -> blocking pre-emptive task switching. But this has no consequence for

> the SWI instruction on ARM:

 

I'm on embOS/Cortex-M and this one blocks the scheduler (in addition to blocking interrupts) while in a critical region.

 

> During a quick scan through the lwip code, I couldn't find the same situation again: every time

> SYS_ARCH_PROTECT() / SYS_ARCH_UNPROTECT() is called there is only plain vanilla code in-between

> (i.e.: no calls to RTOS functions) which strongs my belief that this *really* is a bug inside sockets.c only!

 

That would mainly be because the core code doesn't use threads and the socket layer is one of the only files using multithreading.

 

I think I remember that we had a bug exactly along these lines some time, so I'm not at all enthusiastic about changing it again. I suggest having a look in the file's history if it wasn't as you suggested.

 

Anyway, the fix is easy: just block your scheduler in SYS_ARCH_UNPROTECT(). I'll stick to "it's a bug in the FreeRTOS/lwIP port", even if it triggers nowhere else :-)

 

 

Simon