[PATCH] selftests/mm: skip split_huge_page_test without the split_huge_pages file

[PATCH] selftests/mm: skip split_huge_page_test without the split_huge_pages file

[Sasha Levin]

Every case in split_huge_page_test goes through
/sys/kernel/debug/split_huge_pages. When that file does not exist,
because debugfs is not mounted or the kernel has no CONFIG_DEBUG_FS, the
first write bails out of the whole test and it is reported as a failure:

  Bail out! write_file(/sys/kernel/debug/split_huge_pages): No such file or directory

Skip the test in that case.

Fixes: fa6c02315f74 ("mm: huge_memory: a new debugfs interface for splitting THP tests")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
Found with KernelCI's NVIDIA lab, which runs the tests in a chroot
without debugfs. Tested on an AMD EPYC 9655P with KernelCI's x86_64
build of next-20261104: without debugfs the test now skips, and with
debugfs mounted it still passes all 62 cases.

 tools/testing/selftests/mm/split_huge_page_test.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
index 33a2473327bb7..deabb7db06192 100644
--- a/tools/testing/selftests/mm/split_huge_page_test.c
+++ b/tools/testing/selftests/mm/split_huge_page_test.c
@@ -610,6 +610,9 @@ int main(int argc, char **argv)
 	if (!thp_is_enabled())
 		ksft_exit_skip("Transparent Hugepages not available\n");
 
+	if (access(SPLIT_DEBUGFS, W_OK))
+		ksft_exit_skip("%s: %s\n", SPLIT_DEBUGFS, strerror(errno));
+
 	if (argc > 1)
 		optional_xfs_path = argv[1];
 
-- 
2.53.0

Re: [PATCH] selftests/mm: skip split_huge_page_test without the split_huge_pages file

[Kiryl Shutsemau]

On Wed, Nov 04, 2026 at 10:02:42PM -0500, Sasha Levin wrote:
> Every case in split_huge_page_test goes through
> /sys/kernel/debug/split_huge_pages. When that file does not exist,
> because debugfs is not mounted or the kernel has no CONFIG_DEBUG_FS, the
> first write bails out of the whole test and it is reported as a failure:
> 
>   Bail out! write_file(/sys/kernel/debug/split_huge_pages): No such file or directory
> 
> Skip the test in that case.
> 
> Fixes: fa6c02315f74 ("mm: huge_memory: a new debugfs interface for splitting THP tests")
> Signed-off-by: Sasha Levin <sashal@kernel.org>

Acked-by: Kiryl Shutsemau (Meta) <kas@kernel.org>

-- 
  Kiryl Shutsemau / Kirill A. Shutemov

Re: [PATCH] selftests/mm: skip split_huge_page_test without the split_huge_pages file

[Zi Yan]

On 4 Nov 2026, at 22:02, Sasha Levin wrote:

> Every case in split_huge_page_test goes through
> /sys/kernel/debug/split_huge_pages. When that file does not exist,
> because debugfs is not mounted or the kernel has no CONFIG_DEBUG_FS, the
> first write bails out of the whole test and it is reported as a failure:
>
>   Bail out! write_file(/sys/kernel/debug/split_huge_pages): No such file or directory
>
> Skip the test in that case.
>
> Fixes: fa6c02315f74 ("mm: huge_memory: a new debugfs interface for splitting THP tests")
> Signed-off-by: Sasha Levin <sashal@kernel.org>
> ---
> Found with KernelCI's NVIDIA lab, which runs the tests in a chroot
> without debugfs. Tested on an AMD EPYC 9655P with KernelCI's x86_64
> build of next-20261104: without debugfs the test now skips, and with
> debugfs mounted it still passes all 62 cases.
>
>  tools/testing/selftests/mm/split_huge_page_test.c | 3 +++
>  1 file changed, 3 insertions(+)
>

Thanks. But I do think we need a way of specifying the necessary
Kconfig options for selftests. Oh, we do have one:
https://elixir.bootlin.com/linux/v7.3.1/source/tools/testing/selftests/mm/config. Can you add CONFIG_DEBUG_FS=y to it while at it?

Otherwise, LGTM.

Acked-by: Zi Yan <ziy@nvidia.com>

Best Regards,
Yan, Zi

Re: [PATCH] selftests/mm: skip split_huge_page_test without the split_huge_pages file

[David Hildenbrand (Arm)]

On 11/5/26 16:19, Zi Yan wrote:
> On 4 Nov 2026, at 22:02, Sasha Levin wrote:
> 
>> Every case in split_huge_page_test goes through
>> /sys/kernel/debug/split_huge_pages. When that file does not exist,
>> because debugfs is not mounted or the kernel has no CONFIG_DEBUG_FS, the
>> first write bails out of the whole test and it is reported as a failure:
>>
>>   Bail out! write_file(/sys/kernel/debug/split_huge_pages): No such file or directory
>>
>> Skip the test in that case.
>>
>> Fixes: fa6c02315f74 ("mm: huge_memory: a new debugfs interface for splitting THP tests")
>> Signed-off-by: Sasha Levin <sashal@kernel.org>
>> ---
>> Found with KernelCI's NVIDIA lab, which runs the tests in a chroot
>> without debugfs. Tested on an AMD EPYC 9655P with KernelCI's x86_64
>> build of next-20261104: without debugfs the test now skips, and with
>> debugfs mounted it still passes all 62 cases.
>>
>>  tools/testing/selftests/mm/split_huge_page_test.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
> 
> Thanks. But I do think we need a way of specifying the necessary
> Kconfig options for selftests. Oh, we do have one:
> https://elixir.bootlin.com/linux/v7.3.1/source/tools/testing/selftests/mm/config. Can you add CONFIG_DEBUG_FS=y to it while at it?

Just what I was thinking.

-- 
Cheers,

David

[PATCH] selftests/mm: add CONFIG_DEBUG_FS to the config fragment

[Sasha Levin]

The split_huge_page and gup tests need debugfs, and CONFIG_GUP_TEST
depends on it. Add CONFIG_DEBUG_FS=y to the fragment.

Suggested-by: Zi Yan <ziy@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
Merged into an x86_64 tinyconfig, the fragment leaves CONFIG_DEBUG_FS
and CONFIG_GUP_TEST unset without this change and sets both with it.
x86_64_defconfig already has CONFIG_DEBUG_FS=y and is unaffected.

 tools/testing/selftests/mm/config | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tools/testing/selftests/mm/config b/tools/testing/selftests/mm/config
index 06f78bd232e2e..ff85b358d04cc 100644
--- a/tools/testing/selftests/mm/config
+++ b/tools/testing/selftests/mm/config
@@ -4,6 +4,7 @@ CONFIG_PTE_MARKER_UFFD_WP=y
 CONFIG_TEST_VMALLOC=m
 CONFIG_DEVICE_PRIVATE=y
 CONFIG_TEST_HMM=m
+CONFIG_DEBUG_FS=y
 CONFIG_GUP_TEST=y
 CONFIG_TRANSPARENT_HUGEPAGE=y
 CONFIG_MEM_SOFT_DIRTY=y

base-commit: dba8aba0a9a00ff781c03d82ae534fc6f8c2e263
-- 
2.53.0

Re: [PATCH] selftests/mm: add CONFIG_DEBUG_FS to the config fragment

[Zi Yan]

On 5 Nov 2026, at 12:18, Sasha Levin wrote:

> The split_huge_page and gup tests need debugfs, and CONFIG_GUP_TEST
> depends on it. Add CONFIG_DEBUG_FS=y to the fragment.
>
> Suggested-by: Zi Yan <ziy@nvidia.com>
> Signed-off-by: Sasha Levin <sashal@kernel.org>
> ---
> Merged into an x86_64 tinyconfig, the fragment leaves CONFIG_DEBUG_FS
> and CONFIG_GUP_TEST unset without this change and sets both with it.
> x86_64_defconfig already has CONFIG_DEBUG_FS=y and is unaffected.
>

Thanks.

Reviewed-by: Zi Yan <ziy@nvidia.com>

Best Regards,
Yan, Zi

Re: [PATCH] selftests/mm: add CONFIG_DEBUG_FS to the config fragment

[David Hildenbrand (Arm)]

On 11/5/26 18:21, Zi Yan wrote:
> On 5 Nov 2026, at 12:18, Sasha Levin wrote:
> 
>> The split_huge_page and gup tests need debugfs, and CONFIG_GUP_TEST
>> depends on it. Add CONFIG_DEBUG_FS=y to the fragment.
>>
>> Suggested-by: Zi Yan <ziy@nvidia.com>
>> Signed-off-by: Sasha Levin <sashal@kernel.org>
>> ---
>> Merged into an x86_64 tinyconfig, the fragment leaves CONFIG_DEBUG_FS
>> and CONFIG_GUP_TEST unset without this change and sets both with it.
>> x86_64_defconfig already has CONFIG_DEBUG_FS=y and is unaffected.
>>
> 
> Thanks.
> 
> Reviewed-by: Zi Yan <ziy@nvidia.com>

Thanks, I'll wait a bit and then queue both.

-- 
Cheers,

David