[PATCH v3] selftests/mm: ksm-functional-tests: fix partial write handling

[PATCH v3] selftests/mm: ksm-functional-tests: fix partial write handling

[Vineet Agarwal]

Update write() checks to properly detect and handle partial writes.

Previously, partial writes (ret > 0 && ret != len) could be treated
as success because write() does not set errno in this case and the
code returned -errno. This could result in returning 0 and
incorrectly signaling success.

Fix this by verifying that write() returns the full expected length
and treating any mismatch as failure.

Signed-off-by: Vineet Agarwal <agarwal.vineet2006@gmail.com>

Changes in v3:
- Simplify error handling as suggested by reviewer
- Return -1 when write() does not complete fully
- Rebase patch against base tree
---
 .../selftests/mm/ksm_functional_tests.c       | 27 +++++++++++++------
 1 file changed, 19 insertions(+), 8 deletions(-)

diff --git a/tools/testing/selftests/mm/ksm_functional_tests.c b/tools/testing/selftests/mm/ksm_functional_tests.c
index 8d874c4754f3..4abd98be3f59 100644
--- a/tools/testing/selftests/mm/ksm_functional_tests.c
+++ b/tools/testing/selftests/mm/ksm_functional_tests.c
@@ -498,6 +498,8 @@ static void test_prctl_fork(void)
 static int start_ksmd_and_set_frequency(char *pages_to_scan, char *sleep_ms)
 {
 	int ksm_fd;
+	ssize_t ret;
+	size_t len, sleep_len;
 
 	ksm_fd = open("/sys/kernel/mm/ksm/run", O_RDWR);
 	if (ksm_fd < 0)
@@ -506,11 +508,17 @@ static int start_ksmd_and_set_frequency(char *pages_to_scan, char *sleep_ms)
 	if (write(ksm_fd, "1", 1) != 1)
 		return -errno;
 
-	if (write(pages_to_scan_fd, pages_to_scan, strlen(pages_to_scan)) <= 0)
-		return -errno;
+	len = strlen(pages_to_scan);
 
-	if (write(sleep_millisecs_fd, sleep_ms, strlen(sleep_ms)) <= 0)
-		return -errno;
+	ret = write(pages_to_scan_fd, pages_to_scan, len);
+	if (ret != len)
+		return -1;
+
+	sleep_len = strlen(sleep_ms);
+
+	ret = write(sleep_millisecs_fd, sleep_ms, sleep_len);
+	if (ret != sleep_len)
+		return -1;
 
 	return 0;
 }
@@ -518,6 +526,7 @@ static int start_ksmd_and_set_frequency(char *pages_to_scan, char *sleep_ms)
 static int stop_ksmd_and_restore_frequency(void)
 {
 	int ksm_fd;
+	ssize_t ret;
 
 	ksm_fd = open("/sys/kernel/mm/ksm/run", O_RDWR);
 	if (ksm_fd < 0)
@@ -526,11 +535,13 @@ static int stop_ksmd_and_restore_frequency(void)
 	if (write(ksm_fd, "2", 1) != 1)
 		return -errno;
 
-	if (write(pages_to_scan_fd, "100", 3) <= 0)
-		return -errno;
+	ret = write(pages_to_scan_fd, "100", 3);
+	if (ret != 3)
+		return -1;
 
-	if (write(sleep_millisecs_fd, "20", 2) <= 0)
-		return -errno;
+	ret = write(sleep_millisecs_fd, "20", 2);
+	if (ret != 2)
+		return -1;
 
 	return 0;
 }
-- 
2.54.0

Re: [PATCH v3] selftests/mm: ksm-functional-tests: fix partial write handling

[Andrew Morton]

On Sun,  3 May 2026 15:38:40 +0530 Vineet Agarwal <agarwal.vineet2006@gmail.com> wrote:

> Update write() checks to properly detect and handle partial writes.
> 
> Previously, partial writes (ret > 0 && ret != len) could be treated
> as success because write() does not set errno in this case and the
> code returned -errno. This could result in returning 0 and
> incorrectly signaling success.
> 
> Fix this by verifying that write() returns the full expected length
> and treating any mismatch as failure.

Thanks.  AI review asked some questions:
	https://sashiko.dev/#/patchset/20260503101229.654763-1-agarwal.vineet2006@gmail.com

Re: [PATCH v3] selftests/mm: ksm-functional-tests: fix partial write handling

[Vineet Agarwal]

[-- Attachment #1: Type: text/plain, Size: 1469 bytes --]

Hi,

Thanks for taking a look.

Regarding the AI review comments:

   -

   The issue in the original code is that partial writes (ret > 0 && ret <
   len)
   were not detected because the code only checked for write() <= 0. As a
   result,
   partial writes were treated as success.
   -

   The current check ret != len intentionally covers both partial writes and
   error cases (ret < 0) in a simple way, which follows the earlier
   suggestion
   to keep the logic straightforward for selftests.
   -

   For these tests, distinguishing specific error codes is not required, so
   returning -1 on any mismatch is sufficient.

Please let me know if you'd prefer any adjustments to the commit message.

Thanks,
Vineet


On Sun, 3 May 2026 at 18:53, Andrew Morton <akpm@linux-foundation.org>
wrote:

> On Sun,  3 May 2026 15:38:40 +0530 Vineet Agarwal <
> agarwal.vineet2006@gmail.com> wrote:
>
> > Update write() checks to properly detect and handle partial writes.
> >
> > Previously, partial writes (ret > 0 && ret != len) could be treated
> > as success because write() does not set errno in this case and the
> > code returned -errno. This could result in returning 0 and
> > incorrectly signaling success.
> >
> > Fix this by verifying that write() returns the full expected length
> > and treating any mismatch as failure.
>
> Thanks.  AI review asked some questions:
>
> https://sashiko.dev/#/patchset/20260503101229.654763-1-agarwal.vineet2006@gmail.com
>
>

[-- Attachment #2: Type: text/html, Size: 2148 bytes --]

Re: [PATCH v3] selftests/mm: ksm-functional-tests: fix partial write handling

[Mike Rapoport]

Hi,

On Sun, May 03, 2026 at 03:38:40PM +0530, Vineet Agarwal wrote:
> Update write() checks to properly detect and handle partial writes.
> 
> Previously, partial writes (ret > 0 && ret != len) could be treated
> as success because write() does not set errno in this case and the
> code returned -errno. This could result in returning 0 and
> incorrectly signaling success.
> 
> Fix this by verifying that write() returns the full expected length
> and treating any mismatch as failure.
> 
> Signed-off-by: Vineet Agarwal <agarwal.vineet2006@gmail.com>
> 
> Changes in v3:
> - Simplify error handling as suggested by reviewer
> - Return -1 when write() does not complete fully
> - Rebase patch against base tree
> ---
>  .../selftests/mm/ksm_functional_tests.c       | 27 +++++++++++++------
>  1 file changed, 19 insertions(+), 8 deletions(-)
> 
> diff --git a/tools/testing/selftests/mm/ksm_functional_tests.c b/tools/testing/selftests/mm/ksm_functional_tests.c
> index 8d874c4754f3..4abd98be3f59 100644
> --- a/tools/testing/selftests/mm/ksm_functional_tests.c
> +++ b/tools/testing/selftests/mm/ksm_functional_tests.c
> @@ -498,6 +498,8 @@ static void test_prctl_fork(void)
>  static int start_ksmd_and_set_frequency(char *pages_to_scan, char *sleep_ms)
>  {
>  	int ksm_fd;
> +	ssize_t ret;
> +	size_t len, sleep_len;

ret and len should be of the same type and one len variable is enough.
  
>  	ksm_fd = open("/sys/kernel/mm/ksm/run", O_RDWR);
>  	if (ksm_fd < 0)
> @@ -506,11 +508,17 @@ static int start_ksmd_and_set_frequency(char *pages_to_scan, char *sleep_ms)
>  	if (write(ksm_fd, "1", 1) != 1)
>  		return -errno;
>  
> -	if (write(pages_to_scan_fd, pages_to_scan, strlen(pages_to_scan)) <= 0)
> -		return -errno;
> +	len = strlen(pages_to_scan);
>  
> -	if (write(sleep_millisecs_fd, sleep_ms, strlen(sleep_ms)) <= 0)
> -		return -errno;
> +	ret = write(pages_to_scan_fd, pages_to_scan, len);

	if (write(pages_to_scan_fd, pages_to_scan, len) != len)
		return -1;

would be enough here and below.

> +	if (ret != len)
> +		return -1;
> +

-- 
Sincerely yours,
Mike.