tests/kernel/mem_pool/mem_pool_threadsafe/kernel.memory_pool fails sporadically

[nashif]

Describe the bug
test fails sporadically on both qemu_x86 and mps2_an385

To Reproduce
sanitycheck -T tests/kernel/mem_pool/mem_pool_threadsafe

Booting from ROM..***** Booting Zephyr OS zephyr-v1.13.0-3307-g12f984995f *****
Running test suite mpool_threadsafe
===================================================================
starting test - test_mpool_threadsafe
ASSERTION FAIL [block_fits(p, block, lsz)] @ /home/nashif/github/zephyr/lib/mempool/mempool.c:162

on mps2_an385 (qemu):

***** Booting Zephyr OS zephyr-v1.13.0-3307-g12f984995f *****
Running test suite mpool_threadsafe
===================================================================
starting test - test_mpool_threadsafe
***** BUS FAULT *****
  Precise data bus error
  BFAR Address: 0x300005b0
***** Hardware exception *****
Current thread ID = 0x2000009c
Faulting instruction address = 0x4752
Fatal fault in ISR! Spinning...

[andrewboie]

Seen this also on qemu_nios2

[andyross]

Partially addressed by #13770. Seems like all emulation platforms but ARM are fine now. There, though, the test shows a hilarious inversion of our normal pattern: it works well enough in a big sanitycheck, but if I run it in isolation I get pretty reproducible deadlocks. Still looking.

[andyross]

Seems to be down to the swap-is-async priority bug detailed in #12553. Goes away with the trial patch in that bug.

[andyross]

See extensive notes in #12553. This test, on mps2_an385, is the only spurious qemu failure that I can't make go away in volume testing with careful control over machine load and host kernel tick rate. It's likely a real bug somewhere.

[andyross]

Fix up in #14539 has crossed 5500 separate test cases now (790+ runs across the 8 default boards/variants) with no failures. Think this one is fixed.

[andyross]

Fixed merged, but forgot "Fixes ..." in commit. Oops.