[PATCH v2] selftests/mm: Fix memleak in migration benchmark

[PATCH v2] selftests/mm: Fix memleak in migration benchmark

[Hongfu Li]

Several early return paths in run_migration_benchmark() skip
hmm_buffer_free(), leaking the buffer. Replace with a single cleanup
label.

Fixes: 271a7b2e3c13 ("selftests/mm/hmm-tests: new throughput tests including THP")
Signed-off-by: Hongfu Li <lihongfu@kylinos.cn>
Reviewed-by: David Hildenbrand (Arm) <david@kernel.org>
Reviewed-by: SJ Park <sj@kernel.org>
---
v2:
 - Set buffer->ptr to NULL on mmap failure to avoid passing MAP_FAILED
   to munmap() in hmm_buffer_free()
 - Drop the redundant "/* Cleanup */" comment
 - Add Fixes tag for the commit that introduced the leak
 - Collect Reviewed-by tags from David Hildenbrand and SJ Park
---
 tools/testing/selftests/mm/hmm-tests.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

diff --git a/tools/testing/selftests/mm/hmm-tests.c b/tools/testing/selftests/mm/hmm-tests.c
index e4c49699f3f7..402c6f6f5f09 100644
--- a/tools/testing/selftests/mm/hmm-tests.c
+++ b/tools/testing/selftests/mm/hmm-tests.c
@@ -2828,8 +2828,11 @@ static inline int run_migration_benchmark(int fd, int use_thp, size_t buffer_siz
 	buffer->ptr = mmap(NULL, buffer_size, PROT_READ | PROT_WRITE,
 			  MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
 
-	if (buffer->ptr == MAP_FAILED)
-		return -1;
+	if (buffer->ptr == MAP_FAILED) {
+		buffer->ptr = NULL;
+		ret = -1;
+		goto cleanup;
+	}
 
 	/* Apply THP hint if requested */
 	if (use_thp)
@@ -2838,7 +2841,7 @@ static inline int run_migration_benchmark(int fd, int use_thp, size_t buffer_siz
 		ret = madvise(buffer->ptr, buffer_size, MADV_NOHUGEPAGE);
 
 	if (ret)
-		return ret;
+		goto cleanup;
 
 	/* Initialize memory to make sure pages are allocated */
 	ptr = (int *)buffer->ptr;
@@ -2848,11 +2851,11 @@ static inline int run_migration_benchmark(int fd, int use_thp, size_t buffer_siz
 	/* Warmup iteration */
 	ret = hmm_migrate_sys_to_dev(fd, buffer, npages);
 	if (ret)
-		return ret;
+		goto cleanup;
 
 	ret = hmm_migrate_dev_to_sys(fd, buffer, npages);
 	if (ret)
-		return ret;
+		goto cleanup;
 
 	/* Benchmark iterations */
 	for (i = 0; i < iterations; i++) {
@@ -2861,7 +2864,7 @@ static inline int run_migration_benchmark(int fd, int use_thp, size_t buffer_siz
 
 		ret = hmm_migrate_sys_to_dev(fd, buffer, npages);
 		if (ret)
-			return ret;
+			goto cleanup;
 
 		end = get_time_ms();
 		s2d_total += (end - start);
@@ -2871,7 +2874,7 @@ static inline int run_migration_benchmark(int fd, int use_thp, size_t buffer_siz
 
 		ret = hmm_migrate_dev_to_sys(fd, buffer, npages);
 		if (ret)
-			return ret;
+			goto cleanup;
 
 		end = get_time_ms();
 		d2s_total += (end - start);
@@ -2885,9 +2888,9 @@ static inline int run_migration_benchmark(int fd, int use_thp, size_t buffer_siz
 	results->throughput_d2s = (buffer_size / (1024.0 * 1024.0 * 1024.0)) /
 				 (results->dev_to_sys_time / 1000.0);
 
-	/* Cleanup */
+cleanup:
 	hmm_buffer_free(buffer);
-	return 0;
+	return ret;
 }
 
 /*
-- 
2.25.1

Re: [PATCH v2] selftests/mm: Fix memleak in migration benchmark

[Balbir Singh]

On Thu, Jul 09, 2026 at 04:18:43PM +0800, Hongfu Li wrote:
> Several early return paths in run_migration_benchmark() skip
> hmm_buffer_free(), leaking the buffer. Replace with a single cleanup
> label.
> 
> Fixes: 271a7b2e3c13 ("selftests/mm/hmm-tests: new throughput tests including THP")
> Signed-off-by: Hongfu Li <lihongfu@kylinos.cn>
> Reviewed-by: David Hildenbrand (Arm) <david@kernel.org>
> Reviewed-by: SJ Park <sj@kernel.org>
> ---

Thanks!

Acked-by: Balbir Singh <balbirs@nvidia.com>