GCC builds with FreeRTOS and newlib re-entrancy

RE: GCC builds with FreeRTOS and newlib re-entrancy

RobMeades — Wed, 24 Jun 2020 15:25:42 GMT

FYI, the construction in the attached file, a variant of Dave Nadler's solution, fixed the problem for me.

devzone.nordicsemi.com/.../heap_5F00_useNewlib.c

RE: GCC builds with FreeRTOS and newlib re-entrancy

RobMeades — Tue, 23 Jun 2020 08:52:48 GMT

Where the printf() goes to is not the problem, it's the fact of calling it, or many other stdlib function, at all. See here for a list of all the stdlib functions that WILL cause memory corruption if you call them from an RTOS task without implementing the necessary memory protection hooks for newlib. Please don't underestimate the danger of this hole, it is a very big one and you've only not noticed it because the example code you provide with RTOS included are very few in number.

And it is NOT fixed by simply throwing the switch in the FreeRTOS config file, you actually have to write code to connect the two, see Dave Nadler's link above where he has fixed exactly this problem on Freescale and ST: they hadn't spotted it either, so you're in good (? :-)) company.

RE: GCC builds with FreeRTOS and newlib re-entrancy

Susheel Nuguru — Tue, 23 Jun 2020 08:44:57 GMT

Rob,

We normally override printf to redirect to UART instead of stdio. We do this by adding SDK\components\libraries\uart\retarget.c file to the project.

[quote user=""]I also notice that NONE of the NRF5 examples that use FreeRTOS set configUSE_NEWLIB_REENTRANT to 1 in their FreeRTOSConfig.h files, which is kind of strange when you are using newlib.[/quote]

We have not experimented with this config yet, also the documentation for this is very thin to get confidence to experiment with it.