k_mem_pool crash with larger values of n_max

[andrewboie]

Steps to reproduce:

At toplevel declare:
K_MEM_POOL_DEFINE(my_pool, 64, 1024, 61, 4);

Where n_max is 61.

Then somewhere in code:

	struct k_mem_block blk;

	mem = k_mem_pool_alloc(&my_pool, &blk, 1024, 0);

This results in a page fault on x86, on further inspection, the dlist structures in pool->levels are being corrupted:

***** CPU Page Fault (error code 0x00000000)
Supervisor thread read address 0x20000003

If this is not a bug, we need to clearly document what the upper bounds for various parameters to K_MEM_POOL_DEFINE are, and add assertions to enforce them.

[andyross]

Yeah, there's nothing wrong or out of bounds with that code. This should work, it's a real bug. Will investigate.

[linkmeyer]

@andyross , what's the status?

[andyross]

Been sitting behind higher priority scheduler bugs and now a sam_e70 performance issue. Both look resolved at this point, so should have some time this afternoon

[nashif]

@andyross any update on this for 1.13?

[andyross]

Last I tried this, I couldn't reproduce it. But it needs investigation. Will be fixed or closed for 1.13.