linux-kernel - Re: [PATCH v1] mm: kmsan: add tests for high-order page freeing

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <aa38c43c-0907-4b49-8e76-a2ade35089cd@arm.com>
Date: Mon, 12 Jan 2026 15:18:00 +0000
From: Ryan Roberts <ryan.roberts@....com>
To: Alexander Potapenko <glider@...gle.com>
Cc: akpm@...ux-foundation.org, linux-mm@...ck.org,
 linux-kernel@...r.kernel.org, elver@...gle.com, dvyukov@...gle.com,
 kasan-dev@...glegroups.com
Subject: Re: [PATCH v1] mm: kmsan: add tests for high-order page freeing

On 12/01/2026 14:51, Alexander Potapenko wrote:
> Add regression tests to verify that KMSAN correctly poisons the full memory
> range when freeing pages.
> 
> Specifically, verify that accessing the tail pages of a high-order
> non-compound allocation triggers a use-after-free report. This ensures
> that the fix "mm: kmsan: Fix poisoning of high-order non-compound pages"
> is working as expected.
> 
> Also add a test for standard order-0 pages for completeness.
> 
> Link: https://lore.kernel.org/all/20260104134348.3544298-1-ryan.roberts@arm.com/
> Signed-off-by: Alexander Potapenko <glider@...gle.com>
> ---
>  mm/kmsan/kmsan_test.c | 48 ++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 47 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/kmsan/kmsan_test.c b/mm/kmsan/kmsan_test.c
> index 902ec48b1e3e6..25cfba0db2cfb 100644
> --- a/mm/kmsan/kmsan_test.c
> +++ b/mm/kmsan/kmsan_test.c
> @@ -361,7 +361,7 @@ static void test_init_vmalloc(struct kunit *test)
>  	KUNIT_EXPECT_TRUE(test, report_matches(&expect));
>  }
>  
> -/* Test case: ensure that use-after-free reporting works. */
> +/* Test case: ensure that use-after-free reporting works for kmalloc. */
>  static void test_uaf(struct kunit *test)
>  {
>  	EXPECTATION_USE_AFTER_FREE(expect);
> @@ -378,6 +378,50 @@ static void test_uaf(struct kunit *test)
>  	KUNIT_EXPECT_TRUE(test, report_matches(&expect));
>  }
>  
> +/* Test case: ensure that use-after-free reporting works for freed pages. */
> +static void test_uaf_pages(struct kunit *test)
> +{
> +	EXPECTATION_USE_AFTER_FREE(expect);
> +	const int order = 0;
> +	volatile char value;
> +	struct page *page;
> +	volatile char *var;
> +
> +	kunit_info(test, "use-after-free on a freed page (UMR report)\n");
> +
> +	/* Memory is initialized up until __free_pages() thanks to __GFP_ZERO. */
> +	page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
> +	var = page_address(page);
> +	__free_pages(page, order);
> +
> +	/* Copy the invalid value before checking it. */
> +	value = var[3];
> +	USE(value);
> +	KUNIT_EXPECT_TRUE(test, report_matches(&expect));
> +}
> +
> +/* Test case: ensure that use-after-free reporting works for alloc_pages. */
> +static void test_uaf_high_order_pages(struct kunit *test)
> +{
> +	EXPECTATION_USE_AFTER_FREE(expect);
> +	const int order = 1;
> +	volatile char value;
> +	struct page *page;
> +	volatile char *var;
> +
> +	kunit_info(test,
> +		   "use-after-free on a freed high-order page (UMR report)\n");
> +
> +	page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
> +	var = page_address(page) + PAGE_SIZE;
> +	__free_pages(page, order);
> +
> +	/* Copy the invalid value before checking it. */
> +	value = var[3];
> +	USE(value);
> +	KUNIT_EXPECT_TRUE(test, report_matches(&expect));
> +}

test_uaf_pages() and test_uaf_high_order_pages() are the same except for the
value of order. Does it make sense to create a single parameterized helper that
gets called from the 2 tests wrappers?

Functionally looks correct though, so either way, feel free to add:

Reviewed-by: Ryan Roberts <ryan.roberts@....com>

> +
>  /*
>   * Test case: ensure that uninitialized values are propagated through per-CPU
>   * memory.
> @@ -683,6 +727,8 @@ static struct kunit_case kmsan_test_cases[] = {
>  	KUNIT_CASE(test_init_kmsan_vmap_vunmap),
>  	KUNIT_CASE(test_init_vmalloc),
>  	KUNIT_CASE(test_uaf),
> +	KUNIT_CASE(test_uaf_pages),
> +	KUNIT_CASE(test_uaf_high_order_pages),
>  	KUNIT_CASE(test_percpu_propagate),
>  	KUNIT_CASE(test_printk),
>  	KUNIT_CASE(test_init_memcpy),