tests: interrupt: refine the offload case not rely on delay timing

[enjiamai]

The interrupt offload testcases fail on some boards because the timing
of the delay is too short. Refine the testcases and make it not rely
on the delay timing.

Fixes #35097
Fixes #35241

These testcases reported failure in those board:

• verified pass
  intel_adsp_cavs15

• need help to verify, I do not have these boards, could you please help to verify? Appreciate your help.
  @MaureenHelm @hakehuang
  frdm_kl25z

  @ABOSTM
  nucleo_f091rc
  nucleo_l073
  nucleo_l152re

Signed-off-by: Enjia Mai enjiax.mai@intel.com

[enjiamai]

Excuse me, @MaureenHelm @hakehuang @ABOSTM , could you please help to verify if this issue got fixed on those boards when you got a chance, thank you so much!

And thanks for the solution that @MaureenHelm provided~

[hakehuang]

Excuse me, @MaureenHelm @hakehuang @ABOSTM , could you please help to verify if this issue got fixed on those boards when you got a chance, thank you so much!

And thanks for the solution that @MaureenHelm provided~

update with your patch below

diff --git a/tests/kernel/interrupt/src/interrupt_offload.c b/tests/kernel/interrupt/src/interrupt_offload.c
index d00a6167c5..4b741dd813 100644
--- a/tests/kernel/interrupt/src/interrupt_offload.c
+++ b/tests/kernel/interrupt/src/interrupt_offload.c
@@ -11,6 +11,9 @@
 #define STACK_SIZE     1024
 #define NUM_WORK       4

+#define DELAY_US       (1000000UL * CONFIG_SYS_CLOCK_TICKS_PER_SEC  \
+                               / CONFIG_SYS_CLOCK_HW_CYCLES_PER_SEC)
+
 static struct k_work offload_work[NUM_WORK];
 static struct k_work_q wq_queue;
 static K_THREAD_STACK_DEFINE(wq_stack, STACK_SIZE);
@@ -127,7 +130,7 @@ static void t_running(void *p1, void *p2, void *p3)
 {
        while (1) {
                orig_t_keep_run = 1;
-               k_usleep(1);
+               k_usleep(DELAY_US);
        }
 }

@@ -173,7 +176,7 @@ static void run_test_offload(int case_type, int real_irq)
        }

        /* wait for thread start */
-       k_usleep(10);
+       k_usleep(DELAY_US);

        for (int i = 0; i < NUM_WORK; i++) {

@@ -194,7 +197,7 @@ static void run_test_offload(int case_type, int real_irq)
        /* wait for all offload job complete */
        k_sem_take(&end_sem, K_FOREVER);

-       k_usleep(1);
+       k_usleep(DELAY_US);

        zassert_equal(orig_t_keep_run, 1,
                        "offload job done, the original thread run");
@@ -219,6 +222,8 @@ static void run_test_offload(int case_type, int real_irq)
  */
 void test_isr_offload_job_multiple(void)
 {
+       TC_PRINT("delay unit is %ld us\n", DELAY_US);
+
        offload_job_prio_higher = false;
        run_test_offload(TEST_OFFLOAD_MULTI_JOBS, false);

the issues is still there

*** Booting Zephyr OS build zephyr-v2.5.0-3845-gf8ac3a49ec95  ***
Running test suite interrupt_feature
===================================================================
START - test_isr_dynamic
installing dynamic ISR for IRQ 0
 PASS - test_isr_dynamic in 0.3 seconds
===================================================================
START - test_nested_isr
Triggering irq : 29
isr0: Enter
Triggering irq : 28
isr1: Enter
isr1: Leave
isr0: Leave
 PASS - test_nested_isr in 0.19 seconds
===================================================================
START - test_prevent_interruption
locking interrupts
unlocking interrupts
timer fired
 PASS - test_prevent_interruption in 2.5 seconds
===================================================================
START - test_isr_regular
 SKIP - test_isr_regular in 0.1 seconds
===================================================================
START - test_isr_offload_job_multiple
delay unit is 29 us
case 0
case 0

ASSERTION FAIL [work->queue != ((void *)0)] @ WEST_TOPDIR/zephyr/kernel/work.c:334
E: ***** HARD FAULT *****
E: ARCH_EXCEPT with reason 4

E: r0/a1:  0x00000004  r1/a2:  0x0000014e  r2/a3:  0x00006e1c
E: r3/a4:  0x000035c5 r12/ip:  0x00000003 r14/lr:  0x000035cf
E:  xpsr:  0x0100000b
E: Faulting instruction address (r15/pc): 0x00005782
E: >>> ZEPHYR FATAL ERROR 4: Kernel panic on CPU 0
E: Fault during interrupt handling

E: Current thread: 0x1ffff480 (ztest_thread)
E: Halting system

and looks like the delay is not enough,

delay unit is 29 us

and the value is here for frdm_kl25z

#define CONFIG_SYS_CLOCK_TICKS_PER_SEC 10000
#define CONFIG_SYS_CLOCK_HW_CYCLES_PER_SEC 48000000

when I manually change below as Maureen suggested, it works.

@@ -194,7 +197,7 @@ static void run_test_offload(int case_type, int real_irq)
        /* wait for all offload job complete */
        k_sem_take(&end_sem, K_FOREVER);

-       k_usleep(1);
+       k_usleep(1000);

IMHO, the interrupt handling is better to controled by semephore, or lock, rather than use different delay, which will be difficult to maintain over time.

[enjiamai]

@hakehuang ok, thanks for the help, let me update it.

[enjiamai]

@hakehuang I make the testcase not relying on the timing, base on your suggestion. Could you please help to verify it again, thanks!

[hakehuang]

LGTM

[ABOSTM]

@enjiamai ,
All 3 boards: nucleo_f091rc, nucleo_l073 and nucleo_l152re
are now passed.
Thanks

[enjiamai]

@hakehuang @ABOSTM Thanks for the help on verifying them!

[PerMac]

I checked this PR at nrf boards and it solves the issue also on our platforms

[enjiamai]

I checked this PR at nrf boards and it solves the issue also on our platforms

@PerMac , thanks for your help with this!

[MaureenHelm]

I verified that this PR allows the test to pass on frdm_kl25z, but would like @andyross or @ioannisg to review the functional changes to the test before we merge.

[MaureenHelm]

@andyross will you please review?

[stephanosio]

First pass

[enjiamai]

Hi, @stephanosio Thanks for your review! I have addressed the comments you gave. Could you please help to take look at this again?

[stephanosio]

Looks ok to me.

p.s. verified working on nrf52840dk_nrf52840 and nrf51dk_nrf51422.