tests/kernel/fifo/fifo_timeout fails on nrf52840_pca10056

[arun1joshi]

The subtest test_timeout_threads_pend_on_dual_fifos in tests/kernel/fifo/fifo_timeout fail with Assertion failure on nrf52840_pca10056
Note: if I comment out this test case, there is no kernel panic and all other test cases pass in the suite; this test suite is passing on nrf52_pca10040

The test started failing from this commit: [5fb4ac9]

Arch: arm
Platforms: nrf52840_pca10056

Error on console log:

***** delaying boot 1000ms (per build configuration) *****
***** Booting Zephyr OS zephyr-v1.13.0-4573-g012d404125 (delayed boot 1000ms) *****
Running test suite test_fifo_timeout
===================================================================
starting test - test_timeout_threads_pend_on_dual_fifos
 thread (q order: 0, t/o: 0, fifo 0x20000134)
 thread (q order: 5, t/o: 10, fifo 0x20000120)
 thread (q order: 1, t/o: 30, fifo 0x20000120)
 thread (q order: 7, t/o: 20, fifo 0x20000120)
 thread (q order: 2, t/o: 50, fifo 0x20000120)
 thread (q order: 8, t/o: 40, fifo 0x20000134)

    Assertion failed at zephyr.git/tests/kernel/fifo/fifo_timeout/src/main.c:130: test_thread_pend_and_timeout: is_timeout_in_range(start_time, d->timeout) is false

 thread (q order: 4, t/o: 70, fifo 0x20000134)
 thread (q order: 3, t/o: 80, fifo 0x20000134)
FAIL - test_timeout_threads_pend_on_dual_fifos
===================================================================
starting test - test_timeout_threads_pend_fail_on_fifo
***** BUS FAULT *****
  Precise data bus error
  BFAR Address: 0x4553534f
***** Hardware exception *****
Current thread ID = 0x20000c68
Faulting instruction address = 0x4a3c
Fatal fault in thread 0x20000c68! Aborting.
***** MPU FAULT *****
  Data Access Violation
  MMFAR Address: 0x4553534f
***** Hardware exception *****
Current thread ID = 0x20000210
Faulting instruction address = 0x7482
Fatal fault in thread 0x20000210! Aborting.
ASSERTION FAIL [z_spin_lock_valid(l)] @ zephyr.git/include/spinlock.h:66
	Recursive spinlock
***** Kernel Panic! *****
Current thread ID = 0x20000294
Faulting instruction address = 0x1cba
evaluation failed
clean passed
COMPLETION failed

[ioannisg]

@lemrey FYI

[lemrey]

Oh, this is exactly what I have been hitting, though somewhere else entirely.
I would say this is a high priority bug.

[andrewboie]

it is reproducible on x86 and ARM.

Can you post a direct link to your test case which demonstrates this?

[andyross]

Is that bisection accurate above? I can totally imagine this being exposed by the spinlock validation patch or the later spinlockification series that went in last Friday. I have a harder time seeing how double precision floating point is involved. :)

Can you re-bisect and check that it worked as of 552a03b?

[andyross]

Actually reading that more carefully: it seems like the original root cause is the timing related failure reported via the test. The recursive spinlock (which is a new assert) is more likely due to a bug in the fault handling.

[lemrey]

Neither I or @ioannisg have been able to reproduce the original issue locally.
I think it was related to #13411 anyway, in which case I suppose it could be closed.