Testcases fixes #60567
Testcases fixes #60567
Conversation
povergoing
requested review from
rlubos,
tbursztyka,
dcpleung,
andyross,
nashif and
cfriedt
as code owners
zephyrbot
added
area: LWM2M
area: POSIX
| @@ -1,6 +1,7 @@ | |||
| CONFIG_ZTEST=y | |||
| CONFIG_ZTEST_NEW_API=y | |||
| CONFIG_POSIX_API=y | |||
| CONFIG_TEST_STACK_SIZE=2048 | |||
There was a problem hiding this comment.
Can you provide some evidence with reproducibility that the test thread stack overflows? Normally that would be part of a bug report.
There was a problem hiding this comment.
I might make a mistake about this CONFIG_TEST_STACK_SIZE, I cannot find out why I set it, now remove it.
tests/kernel/spinlock/src/main.c
Outdated
| trylock_failures = 0; | ||
| trylock_successes = 0; | ||
|
|
There was a problem hiding this comment.
These are already zero-initialized with bss. There is no need to do it explicitly here (since they are not shared between tests.
Arguably, the correct place to do this though would be in a setup function as part of the testsuite.
There was a problem hiding this comment.
This should do it
diff --git a/tests/kernel/spinlock/src/main.c b/tests/kernel/spinlock/src/main.c
index 4ad13373d1..0ab132bf2b 100644
--- a/tests/kernel/spinlock/src/main.c
+++ b/tests/kernel/spinlock/src/main.c
@@ -223,4 +223,11 @@ ZTEST(spinlock, test_trylock)
zassert_true(trylock_successes > 0);
}
-ZTEST_SUITE(spinlock, NULL, NULL, NULL, NULL, NULL);
+static void before(void *ctx)
+{
+ ARG_UNUSED(ctx);
+
+ bounce_done = 0;
+}
+
+ZTEST_SUITE(spinlock, NULL, NULL, before, NULL, NULL);
Fix the testcases failures by increasing the stack size. Signed-off-by: Jaxson Han <[email protected]>
The test_trylock reuses the cpu1_thread, but there is no way for it to exit. This will cause the thread created twice, as a result, two cpu running the same thread simultaneously cause an unexpected crash. Fix this by adding initialization of resources and also the exit for the cpu1_thread. Signed-off-by: Jaxson Han <[email protected]>
These testcases are not SMP-safe, will fail on all SMP boards. Simply turning them to 1cpu test can not fix the issue. So, setting CONFIG_MP_MAX_NUM_CPUS to 1 as a workaround. Signed-off-by: Jaxson Han <[email protected]>
povergoing
force-pushed
the
testcases_fixes
branch
from
d16f204
to
e305a3c
Compare
Added more details in the description |
@povergoing Does executing ztest in SMP environment mean that several tests can run in parallel on different CPUs? If so, then indeed those test may not work in such environment. |
@rlubos Yes it does. Creating 4 threads at a 4 CPU SMP platform means 4 threads running in parallel and even it does not care about the priority of the threads. So I just set it to one CPU to pass the tests. |
Stack overflows on some arm64 platforms, so increase the stack as it's likely reaching the limit of the stack for most of the platforms.
setting
CONFIG_MAIN_STACK_SIZE=2048because we observe the following platforms are reaching (some already exceeded) the stack limit (1152):fvp_baser_aemv8r_smp:
qemu_cortex_a53:
qemu_riscv64_smp:
qemu_x86_64:
tests: kernel: spinlock: Fix test_trylock thread reusable issue
The test_trylock reuses the cpu1_thread, but there is no way for it to exit. This will cause the thread created twice, as a result, two CPU running the same thread simultaneously cause an unexpected crash.
Analysis:
cpu1_threadandcpu1_fncpu1_fnnever exitcpu1_threadagain but withtrylock_fn. Remembercpu1_fnmight be running at another core.cpu1_threadtwice but the first one never callsk_thread_abort. Thus, two samethread structliving in two_cpu->currentat the same time will cause unexpected behavior.tests: net: lib: lwm2m: Use 1 cpu only as a workaround
These testcases are not SMP-safe, and will fail on all SMP boards. Simply turning them to 1cpu test can not fix the issue. So, setting CONFIG_MP_MAX_NUM_CPUS to 1 as a workaround.
To reproduce: