Re: [PATCH 2/2] kunit: Add example of test suite that can be skipped at runtime

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] The new example promotes a `suite_init` skip pattern that triggers multiple existing KUnit infrastructure bugs, leading to resource leaks and malformed KTAP output.
- [Low] Macro duck-typing API misuse: `kunit_mark_skipped()` is intended for `struct kunit *` but is passed a `struct kunit_suite *`.
--

commit bb04d5d94394b78bbd531471f54e3d41268deac5
Author: Vaibhav Jain <vaibhav@linux.ibm.com>

kunit: Add example of test suite that can be skipped at runtime

This patch adds an example test suite to kunit-example-test.c demonstrating
how to skip an entire test suite based on runtime conditions. It uses the
suite_init callback to mark the suite as skipped using kunit_mark_skipped().

> diff --git a/lib/kunit/kunit-example-test.c b/lib/kunit/kunit-example-test.c
> index 0bae7b7ca0b05..b8ded54fa46d8 100644
> --- a/lib/kunit/kunit-example-test.c
> +++ b/lib/kunit/kunit-example-test.c
> @@ -591,5 +591,34 @@ static struct kunit_suite example_init_test_suite = {
>   */
>  kunit_test_init_section_suites(&example_init_test_suite);
>  

[ ... ]

> +static int example_skip_suite_init(struct kunit_suite *suite)
> +{
> +	kunit_mark_skipped(suite, "Test suite expected to be skipped");

[Severity: Medium]
Does this example promote a pattern that triggers underlying KUnit
infrastructure bugs?

The underlying framework bugs predate this patch, but promoting this pattern
could lead to resource leaks. In kunit_run_tests(), skipping a suite causes
it to completely bypass suite_exit():

lib/kunit/test.c:kunit_run_tests() {
    ...
	} else if (suite->status == KUNIT_SKIPPED) {
		/* Skip this kunit suite */
		goto suite_end;
	}
    ...
}

This contradicts the KUnit API documentation (which guarantees suite_exit
will run even if suite_init fails) and structurally guarantees a memory leak
if test authors follow this example and allocate resources prior to checking
the skip condition.

Additionally, skipping the suite here leads to malformed KTAP output because
debugfs_print_results() unconditionally prints the test plan without the
inner test results. It also permanently locks the suite in the skipped state
across debugfs re-runs because kunit_init_suite() fails to reset
suite->status back to KUNIT_SUCCESS.

[Severity: Low]
Is it safe to pass a struct kunit_suite pointer to kunit_mark_skipped()?

The kunit_mark_skipped() macro is designed and explicitly documented to
accept a struct kunit pointer (the individual test case context), rather
than a suite:

include/kunit/test.h:
#define kunit_mark_skipped(test, fmt, ...)				\
	do {								\
		WRITE_ONCE((test)->status, KUNIT_SKIPPED);		\

This only compiles because both structs coincidentally share identically
named status and status_comment fields, relying on C macro structural
duck-typing.

If this macro is ever refactored into a type-safe inline function, this
example and any downstream code that copies it will fail to compile. Should
a proper type-safe API like kunit_suite_mark_skipped() be introduced instead?

> +	return 0;
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260604162805.556135-1-vaibhav@linux.ibm.com?part=2