Malloc failed in with pic32MX port

chaabanemalki wrote on Monday, February 22, 2016:

Hello,
Sometimes when I try to allocate 2048 bytes using pvPortMalloc() it fails and reported that remaining space heap size is greated then 11k in the vApplicationMallocFailedHook()

I use heap4.c and I can’t really find an explanation for this ?

Regards
Chaabane

richard_damon wrote on Monday, February 22, 2016:

One danger with using heap memory is that the heap space can become fragmented and hard to find large blocks. For instance, if you allocate a number of 1k blocks, then release every other one, the space left normally can’t be used to allocate a 2k block. This is one reason many coding standards restrict the use of heap memory to startup and ‘non-critical’ operations that are allowed to ‘fail’.

chaabanemalki wrote on Monday, February 22, 2016:

I have allocated 36k for the kernel total heap, after few allocations of 38,32,512… bytes (doesn’t exceed 4k), the pvPortMalloc() systematically failes to allocate 2K space yet it reported that it has more than 11,5K free space.

I reduced the total heap to 30K and everything is working just fine now !!!

Any idea why i’m having this issue ?

rtel wrote on Monday, February 22, 2016:

I don’t know of any reason this should be a problem.

How much RAM does your PIC32 actually have? Is anything else walking
over the heap space? Is your linker script correct for your device and
are any linker warnings generated?

chaabanemalki wrote on Monday, February 22, 2016:

My PIC has 128K RAM and I don’t have any linker warning.

What do you mean by “Is anything else walking over the heap space?” I’m not sure how to check that !!

rtel wrote on Monday, February 22, 2016:

What do you mean by “Is anything else walking over the heap space?” I’m
not sure how to check that !!

The heap includes some meta data that describes its state (which blocks
are free, which allocated, etc.). If there is a good old fashioned data
corruption by another part of the application writing over this data
then the heap implementation will misbehave because the meta data will
be corrupt.

It might be that a certain part of your RAM is getting used by something
else, either intentionally or unintentionally, and once you set the heap
size large enough so it hits that RAM region, you get into problems of
corruption.

chaabanemalki wrote on Wednesday, February 24, 2016:

It could be the case, Is there a way to check this ?