2022-07-22 17:25:38

by Daniel Latypov

Subject: [PATCH v2 5/5] kunit: make kunit_kfree(NULL) a no-op to match kfree()

The real kfree() function will silently return when given a NULL.
So a user might reasonably think they can write the following code:
char *buffer = NULL;
if (param->use_buffer) buffer = kunit_kzalloc(test, 10, GFP_KERNEL);
...
kunit_kfree(test, buffer);

As-is, kunit_kfree() will mark the test as FAILED when buffer is NULL.
(And in earlier times, it would segfault).

Let's match the semantics of kfree().

Suggested-by: David Gow <[email protected]>
Signed-off-by: Daniel Latypov <[email protected]>
---
v1 -> v2: add this patch to the series.
---
lib/kunit/test.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/lib/kunit/test.c b/lib/kunit/test.c
index c7ca87484968..879c8db36cb5 100644
--- a/lib/kunit/test.c
+++ b/lib/kunit/test.c
@@ -698,6 +698,9 @@ static inline bool kunit_kfree_match(struct kunit *test,

void kunit_kfree(struct kunit *test, const void *ptr)
{
+ if (!ptr)
+ return;
+
if (kunit_destroy_resource(test, kunit_kfree_match, (void *)ptr))
KUNIT_FAIL(test, "kunit_kfree: %px already freed or not allocated by kunit", ptr);
}
--
2.37.1.359.gd136c6c3e2-goog

2022-07-23 06:51:17

by David Gow

Subject: Re: [PATCH v2 5/5] kunit: make kunit_kfree(NULL) a no-op to match kfree()

On Sat, Jul 23, 2022 at 1:15 AM Daniel Latypov <[email protected]> wrote:
>
> The real kfree() function will silently return when given a NULL.
> So a user might reasonably think they can write the following code:
> char *buffer = NULL;
> if (param->use_buffer) buffer = kunit_kzalloc(test, 10, GFP_KERNEL);
> ...
> kunit_kfree(test, buffer);
>
> As-is, kunit_kfree() will mark the test as FAILED when buffer is NULL.
> (And in earlier times, it would segfault).
>
> Let's match the semantics of kfree().
>
> Suggested-by: David Gow <[email protected]>
> Signed-off-by: Daniel Latypov <[email protected]>
> ---
> v1 -> v2: add this patch to the series.
> ---

Thanks! This looks good to me, and worked with a basic test.

Reviewed-by: David Gow <[email protected]>

Cheers,
-- David

> lib/kunit/test.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/lib/kunit/test.c b/lib/kunit/test.c
> index c7ca87484968..879c8db36cb5 100644
> --- a/lib/kunit/test.c
> +++ b/lib/kunit/test.c
> @@ -698,6 +698,9 @@ static inline bool kunit_kfree_match(struct kunit *test,
>
> void kunit_kfree(struct kunit *test, const void *ptr)
> {
> + if (!ptr)
> + return;
> +
> if (kunit_destroy_resource(test, kunit_kfree_match, (void *)ptr))
> KUNIT_FAIL(test, "kunit_kfree: %px already freed or not allocated by kunit", ptr);
> }
> --
> 2.37.1.359.gd136c6c3e2-goog
>

2022-10-05 21:21:06

by Brendan Higgins

Subject: Re: [PATCH v2 5/5] kunit: make kunit_kfree(NULL) a no-op to match kfree()

On Fri, Jul 22, 2022 at 1:15 PM Daniel Latypov <[email protected]> wrote:
>
> The real kfree() function will silently return when given a NULL.
> So a user might reasonably think they can write the following code:
> char *buffer = NULL;
> if (param->use_buffer) buffer = kunit_kzalloc(test, 10, GFP_KERNEL);
> ...
> kunit_kfree(test, buffer);
>
> As-is, kunit_kfree() will mark the test as FAILED when buffer is NULL.
> (And in earlier times, it would segfault).
>
> Let's match the semantics of kfree().
>
> Suggested-by: David Gow <[email protected]>
> Signed-off-by: Daniel Latypov <[email protected]>

Reviewed-by: Brendan Higgins <[email protected]>