Why was a malloc with heap_size = 0 successful?

[Ken]

We started with the default project that have no heap. Added mallocs to the system and they seemed to succeed, without realizing heap was set to 0. They did not return NULL.

I would have expected the mallocs to fail since there is no heap at all.

Can anyone explain why the malloc succeeded?

[Håkon Alseth]

What compiler/stdlib are you using? malloc is normally provided by an in-built library, unless you're brave and declare your own.

I know that the malloc implementation in libnosys (-lnosys in your linker file, if using GCC) does not have any overflow check in function _sbrk(), so if you're using GCC + nosys; this will happen and it will most likely eat out of the bottom of the .stack. Here's the _sbrk() from libnosys, which basically has no checks: github.com/.../sbrk.c

Cheers, Håkon

[Ken]

We are using the default libnosys, so that explains why it was successful. Should we be using the malloc functions from nordic's mem_manager library rather than the libnosys malloc functions. Would this add the heap overflow protection?

Or Is there a better nordic recommended library for use with the GCC compiler?

[Håkon Alseth]

Hi! The mem-manager has checks for > 0 && < max_block_size: infocenter.nordicsemi.com/.../lib_mem_manager.html

We tend to avoid dynamic memory allocation, unless absolutely needed, as there are many "gotcha's" with it, like memory fragmentation etc. Unfortunately, I have little to no experience with dynamic allocation on embedded platforms, so I cannot give a recommendation on which library you should use.

Cheers, Håkon