[v1] selftests/mm: skip several tests when thp is not available

[RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Chunyu Hu 3 weeks, 5 days ago

When thp is not enabled on some kernel config such as realtime kernel, the
test will report failure. Fix the false positive by skipping the test
directly when thp is not enabled.

There's a naming conflict on write_file() function, both thp_settings.h and
split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
the thp_settings.h, rename this local write_file to safe_write_file to
avoid the conflict. The reason to use 'safe_' is it does some error check.

Tested with thp disabled kernel:
Before The fix:
  # --------------------------------------------------
  # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
  # --------------------------------------------------
  # TAP version 13
  # Bail out! Reading PMD pagesize failed
  # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
  # [FAIL]
  not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1

After the fix:
  # --------------------------------------------------
  # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
  # --------------------------------------------------
  # TAP version 13
  # 1..0 # SKIP Transparent Hugepages not available
  # [SKIP]
  ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP

CC: Li Wang <liwang@redhat.com>
Signed-off-by: Chunyu Hu <chuhu@redhat.com>
---
 tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
index e0167111bdd1..615b75ca62cc 100644
--- a/tools/testing/selftests/mm/split_huge_page_test.c
+++ b/tools/testing/selftests/mm/split_huge_page_test.c
@@ -21,6 +21,7 @@
 #include <time.h>
 #include "vm_util.h"
 #include "kselftest.h"
+#include "thp_settings.h"
 
 uint64_t pagesize;
 unsigned int pageshift;
@@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
 	return status;
 }
 
-static void write_file(const char *path, const char *buf, size_t buflen)
+static void safe_write_file(const char *path, const char *buf, size_t buflen)
 {
 	int fd;
 	ssize_t numwritten;
@@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
 	if (ret >= INPUT_MAX)
 		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
 
-	write_file(SPLIT_DEBUGFS, input, ret + 1);
+	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
 }
 
 static char *allocate_zero_filled_hugepage(size_t len)
@@ -772,6 +773,10 @@ int main(int argc, char **argv)
 		ksft_finished();
 	}
 
+	if (!thp_is_enabled()) {
+		ksft_exit_skip("Transparent Hugepages not available\n");
+	}
+
 	if (argc > 1)
 		optional_xfs_path = argv[1];
 
-- 
2.53.0

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by David Hildenbrand (Arm) 3 weeks, 4 days ago

>  
> +	if (!thp_is_enabled()) {
> +		ksft_exit_skip("Transparent Hugepages not available\n");
> +	}
> +

No need for the {}

-- 
Cheers,

David

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Chunyu Hu 3 weeks, 4 days ago

On Thu, Mar 12, 2026 at 08:44:46PM +0100, David Hildenbrand (Arm) wrote:
> 
> >  
> > +	if (!thp_is_enabled()) {
> > +		ksft_exit_skip("Transparent Hugepages not available\n");
> > +	}
> > +
> 
> No need for the {}

Oops, my bad. I overlooked it when checking patch. Will fix in v2.

> 
> -- 
> Cheers,
> 
> David
>

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Zi Yan 3 weeks, 5 days ago

On 12 Mar 2026, at 7:40, Chunyu Hu wrote:

> When thp is not enabled on some kernel config such as realtime kernel, the
> test will report failure. Fix the false positive by skipping the test
> directly when thp is not enabled.
>
> There's a naming conflict on write_file() function, both thp_settings.h and
> split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
> the thp_settings.h, rename this local write_file to safe_write_file to
> avoid the conflict. The reason to use 'safe_' is it does some error check.

Both write_file() do safe checks. Actually the code of both functions looks
almost identical except that the thp_settings one does exit() instead of
ksft_exit_fail_msg().

Can you rename the split_huge_page_test’s write_file() to write_file_local()
and add a comment above the function like
/* add _local to avoid a function conflict with thp_settings.h */?

With that, feel free to add:

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

Thanks.

>
> Tested with thp disabled kernel:
> Before The fix:
>   # --------------------------------------------------
>   # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
>   # --------------------------------------------------
>   # TAP version 13
>   # Bail out! Reading PMD pagesize failed
>   # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
>   # [FAIL]
>   not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1
>
> After the fix:
>   # --------------------------------------------------
>   # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
>   # --------------------------------------------------
>   # TAP version 13
>   # 1..0 # SKIP Transparent Hugepages not available
>   # [SKIP]
>   ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP
>
> CC: Li Wang <liwang@redhat.com>
> Signed-off-by: Chunyu Hu <chuhu@redhat.com>
> ---
>  tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
> index e0167111bdd1..615b75ca62cc 100644
> --- a/tools/testing/selftests/mm/split_huge_page_test.c
> +++ b/tools/testing/selftests/mm/split_huge_page_test.c
> @@ -21,6 +21,7 @@
>  #include <time.h>
>  #include "vm_util.h"
>  #include "kselftest.h"
> +#include "thp_settings.h"
>
>  uint64_t pagesize;
>  unsigned int pageshift;
> @@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
>  	return status;
>  }
>
> -static void write_file(const char *path, const char *buf, size_t buflen)
> +static void safe_write_file(const char *path, const char *buf, size_t buflen)
>  {
>  	int fd;
>  	ssize_t numwritten;
> @@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
>  	if (ret >= INPUT_MAX)
>  		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
>
> -	write_file(SPLIT_DEBUGFS, input, ret + 1);
> +	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
>  }
>
>  static char *allocate_zero_filled_hugepage(size_t len)
> @@ -772,6 +773,10 @@ int main(int argc, char **argv)
>  		ksft_finished();
>  	}
>
> +	if (!thp_is_enabled()) {
> +		ksft_exit_skip("Transparent Hugepages not available\n");
> +	}
> +
>  	if (argc > 1)
>  		optional_xfs_path = argv[1];
>
> -- 
> 2.53.0


Best Regards,
Yan, Zi

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Mike Rapoport 3 weeks, 4 days ago

On Thu, Mar 12, 2026 at 10:11:12AM -0400, Zi Yan wrote:
> On 12 Mar 2026, at 7:40, Chunyu Hu wrote:
> 
> > When thp is not enabled on some kernel config such as realtime kernel, the
> > test will report failure. Fix the false positive by skipping the test
> > directly when thp is not enabled.
> >
> > There's a naming conflict on write_file() function, both thp_settings.h and
> > split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
> > the thp_settings.h, rename this local write_file to safe_write_file to
> > avoid the conflict. The reason to use 'safe_' is it does some error check.
> 
> Both write_file() do safe checks. Actually the code of both functions looks
> almost identical except that the thp_settings one does exit() instead of
> ksft_exit_fail_msg().
> 
> Can you rename the split_huge_page_test’s write_file() to write_file_local()
> and add a comment above the function like
> /* add _local to avoid a function conflict with thp_settings.h */?

Looks like we can move the version that uses ksft_exit_fail_msg() to
vm_util and drop the other one.
 
> With that, feel free to add:
> 
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> 
> Thanks.
> 
> >
> > Tested with thp disabled kernel:
> > Before The fix:
> >   # --------------------------------------------------
> >   # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
> >   # --------------------------------------------------
> >   # TAP version 13
> >   # Bail out! Reading PMD pagesize failed
> >   # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
> >   # [FAIL]
> >   not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1
> >
> > After the fix:
> >   # --------------------------------------------------
> >   # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
> >   # --------------------------------------------------
> >   # TAP version 13
> >   # 1..0 # SKIP Transparent Hugepages not available
> >   # [SKIP]
> >   ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP
> >
> > CC: Li Wang <liwang@redhat.com>
> > Signed-off-by: Chunyu Hu <chuhu@redhat.com>
> > ---
> >  tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
> > index e0167111bdd1..615b75ca62cc 100644
> > --- a/tools/testing/selftests/mm/split_huge_page_test.c
> > +++ b/tools/testing/selftests/mm/split_huge_page_test.c
> > @@ -21,6 +21,7 @@
> >  #include <time.h>
> >  #include "vm_util.h"
> >  #include "kselftest.h"
> > +#include "thp_settings.h"
> >
> >  uint64_t pagesize;
> >  unsigned int pageshift;
> > @@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
> >  	return status;
> >  }
> >
> > -static void write_file(const char *path, const char *buf, size_t buflen)
> > +static void safe_write_file(const char *path, const char *buf, size_t buflen)
> >  {
> >  	int fd;
> >  	ssize_t numwritten;
> > @@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
> >  	if (ret >= INPUT_MAX)
> >  		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
> >
> > -	write_file(SPLIT_DEBUGFS, input, ret + 1);
> > +	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
> >  }
> >
> >  static char *allocate_zero_filled_hugepage(size_t len)
> > @@ -772,6 +773,10 @@ int main(int argc, char **argv)
> >  		ksft_finished();
> >  	}
> >
> > +	if (!thp_is_enabled()) {
> > +		ksft_exit_skip("Transparent Hugepages not available\n");
> > +	}
> > +
> >  	if (argc > 1)
> >  		optional_xfs_path = argv[1];
> >
> > -- 
> > 2.53.0
> 
> 
> Best Regards,
> Yan, Zi

-- 
Sincerely yours,
Mike.

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Chunyu Hu 3 weeks, 1 day ago

On Fri, Mar 13, 2026 at 05:15:57PM +0200, Mike Rapoport wrote:
> On Thu, Mar 12, 2026 at 10:11:12AM -0400, Zi Yan wrote:
> > On 12 Mar 2026, at 7:40, Chunyu Hu wrote:
> > 
> > > When thp is not enabled on some kernel config such as realtime kernel, the
> > > test will report failure. Fix the false positive by skipping the test
> > > directly when thp is not enabled.
> > >
> > > There's a naming conflict on write_file() function, both thp_settings.h and
> > > split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
> > > the thp_settings.h, rename this local write_file to safe_write_file to
> > > avoid the conflict. The reason to use 'safe_' is it does some error check.
> > 
> > Both write_file() do safe checks. Actually the code of both functions looks
> > almost identical except that the thp_settings one does exit() instead of
> > ksft_exit_fail_msg().
> > 
> > Can you rename the split_huge_page_test’s write_file() to write_file_local()
> > and add a comment above the function like
> > /* add _local to avoid a function conflict with thp_settings.h */?
> 
> Looks like we can move the version that uses ksft_exit_fail_msg() to
> vm_util and drop the other one.

Good point. I'll try this in the v2. And I think we can keep return
value.

>  
> > With that, feel free to add:
> > 
> > Reviewed-by: Zi Yan <ziy@nvidia.com>
> > 
> > Thanks.
> > 
> > >
> > > Tested with thp disabled kernel:
> > > Before The fix:
> > >   # --------------------------------------------------
> > >   # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
> > >   # --------------------------------------------------
> > >   # TAP version 13
> > >   # Bail out! Reading PMD pagesize failed
> > >   # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
> > >   # [FAIL]
> > >   not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1
> > >
> > > After the fix:
> > >   # --------------------------------------------------
> > >   # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
> > >   # --------------------------------------------------
> > >   # TAP version 13
> > >   # 1..0 # SKIP Transparent Hugepages not available
> > >   # [SKIP]
> > >   ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP
> > >
> > > CC: Li Wang <liwang@redhat.com>
> > > Signed-off-by: Chunyu Hu <chuhu@redhat.com>
> > > ---
> > >  tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
> > >  1 file changed, 7 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
> > > index e0167111bdd1..615b75ca62cc 100644
> > > --- a/tools/testing/selftests/mm/split_huge_page_test.c
> > > +++ b/tools/testing/selftests/mm/split_huge_page_test.c
> > > @@ -21,6 +21,7 @@
> > >  #include <time.h>
> > >  #include "vm_util.h"
> > >  #include "kselftest.h"
> > > +#include "thp_settings.h"
> > >
> > >  uint64_t pagesize;
> > >  unsigned int pageshift;
> > > @@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
> > >  	return status;
> > >  }
> > >
> > > -static void write_file(const char *path, const char *buf, size_t buflen)
> > > +static void safe_write_file(const char *path, const char *buf, size_t buflen)
> > >  {
> > >  	int fd;
> > >  	ssize_t numwritten;
> > > @@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
> > >  	if (ret >= INPUT_MAX)
> > >  		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
> > >
> > > -	write_file(SPLIT_DEBUGFS, input, ret + 1);
> > > +	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
> > >  }
> > >
> > >  static char *allocate_zero_filled_hugepage(size_t len)
> > > @@ -772,6 +773,10 @@ int main(int argc, char **argv)
> > >  		ksft_finished();
> > >  	}
> > >
> > > +	if (!thp_is_enabled()) {
> > > +		ksft_exit_skip("Transparent Hugepages not available\n");
> > > +	}
> > > +
> > >  	if (argc > 1)
> > >  		optional_xfs_path = argv[1];
> > >
> > > -- 
> > > 2.53.0
> > 
> > 
> > Best Regards,
> > Yan, Zi
> 
> -- 
> Sincerely yours,
> Mike.
>

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Zi Yan 3 weeks, 4 days ago

On 13 Mar 2026, at 11:15, Mike Rapoport wrote:

> On Thu, Mar 12, 2026 at 10:11:12AM -0400, Zi Yan wrote:
>> On 12 Mar 2026, at 7:40, Chunyu Hu wrote:
>>
>>> When thp is not enabled on some kernel config such as realtime kernel, the
>>> test will report failure. Fix the false positive by skipping the test
>>> directly when thp is not enabled.
>>>
>>> There's a naming conflict on write_file() function, both thp_settings.h and
>>> split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
>>> the thp_settings.h, rename this local write_file to safe_write_file to
>>> avoid the conflict. The reason to use 'safe_' is it does some error check.
>>
>> Both write_file() do safe checks. Actually the code of both functions looks
>> almost identical except that the thp_settings one does exit() instead of
>> ksft_exit_fail_msg().
>>
>> Can you rename the split_huge_page_test’s write_file() to write_file_local()
>> and add a comment above the function like
>> /* add _local to avoid a function conflict with thp_settings.h */?
>
> Looks like we can move the version that uses ksft_exit_fail_msg() to
> vm_util and drop the other one.

Could work.

tools/testing/selftests/mm/khugepaged.c uses the write_file() from
thp_settings.h in file_setup_area(), which is part of struct mem_ops.
I thought struct mem_ops was passed to some syscall, but it turns out
that it is used locally. The only downside is that file_setup_area() now
has this ksefltest friendly code. Someone might want to clean up
tools/testing/selftests/mm/khugepaged.c with kselftests APIs. :)


>
>> With that, feel free to add:
>>
>> Reviewed-by: Zi Yan <ziy@nvidia.com>
>>
>> Thanks.
>>
>>>
>>> Tested with thp disabled kernel:
>>> Before The fix:
>>>   # --------------------------------------------------
>>>   # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
>>>   # --------------------------------------------------
>>>   # TAP version 13
>>>   # Bail out! Reading PMD pagesize failed
>>>   # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
>>>   # [FAIL]
>>>   not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1
>>>
>>> After the fix:
>>>   # --------------------------------------------------
>>>   # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
>>>   # --------------------------------------------------
>>>   # TAP version 13
>>>   # 1..0 # SKIP Transparent Hugepages not available
>>>   # [SKIP]
>>>   ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP
>>>
>>> CC: Li Wang <liwang@redhat.com>
>>> Signed-off-by: Chunyu Hu <chuhu@redhat.com>
>>> ---
>>>  tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
>>>  1 file changed, 7 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
>>> index e0167111bdd1..615b75ca62cc 100644
>>> --- a/tools/testing/selftests/mm/split_huge_page_test.c
>>> +++ b/tools/testing/selftests/mm/split_huge_page_test.c
>>> @@ -21,6 +21,7 @@
>>>  #include <time.h>
>>>  #include "vm_util.h"
>>>  #include "kselftest.h"
>>> +#include "thp_settings.h"
>>>
>>>  uint64_t pagesize;
>>>  unsigned int pageshift;
>>> @@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
>>>  	return status;
>>>  }
>>>
>>> -static void write_file(const char *path, const char *buf, size_t buflen)
>>> +static void safe_write_file(const char *path, const char *buf, size_t buflen)
>>>  {
>>>  	int fd;
>>>  	ssize_t numwritten;
>>> @@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
>>>  	if (ret >= INPUT_MAX)
>>>  		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
>>>
>>> -	write_file(SPLIT_DEBUGFS, input, ret + 1);
>>> +	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
>>>  }
>>>
>>>  static char *allocate_zero_filled_hugepage(size_t len)
>>> @@ -772,6 +773,10 @@ int main(int argc, char **argv)
>>>  		ksft_finished();
>>>  	}
>>>
>>> +	if (!thp_is_enabled()) {
>>> +		ksft_exit_skip("Transparent Hugepages not available\n");
>>> +	}
>>> +
>>>  	if (argc > 1)
>>>  		optional_xfs_path = argv[1];
>>>
>>> -- 
>>> 2.53.0
>>
>>
>> Best Regards,
>> Yan, Zi
>
> -- 
> Sincerely yours,
> Mike.


Best Regards,
Yan, Zi

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Chunyu Hu 3 weeks, 1 day ago

On Fri, Mar 13, 2026 at 11:27:30AM -0400, Zi Yan wrote:
> On 13 Mar 2026, at 11:15, Mike Rapoport wrote:
> 
> > On Thu, Mar 12, 2026 at 10:11:12AM -0400, Zi Yan wrote:
> >> On 12 Mar 2026, at 7:40, Chunyu Hu wrote:
> >>
> >>> When thp is not enabled on some kernel config such as realtime kernel, the
> >>> test will report failure. Fix the false positive by skipping the test
> >>> directly when thp is not enabled.
> >>>
> >>> There's a naming conflict on write_file() function, both thp_settings.h and
> >>> split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
> >>> the thp_settings.h, rename this local write_file to safe_write_file to
> >>> avoid the conflict. The reason to use 'safe_' is it does some error check.
> >>
> >> Both write_file() do safe checks. Actually the code of both functions looks
> >> almost identical except that the thp_settings one does exit() instead of
> >> ksft_exit_fail_msg().
> >>
> >> Can you rename the split_huge_page_test’s write_file() to write_file_local()
> >> and add a comment above the function like
> >> /* add _local to avoid a function conflict with thp_settings.h */?
> >
> > Looks like we can move the version that uses ksft_exit_fail_msg() to
> > vm_util and drop the other one.
> 
> Could work.
> 
> tools/testing/selftests/mm/khugepaged.c uses the write_file() from
> thp_settings.h in file_setup_area(), which is part of struct mem_ops.
> I thought struct mem_ops was passed to some syscall, but it turns out
> that it is used locally. The only downside is that file_setup_area() now
> has this ksefltest friendly code. Someone might want to clean up
> tools/testing/selftests/mm/khugepaged.c with kselftests APIs. :)

It looks like work in my try, even before the cleanup with the
kselftest APIs.

> 
> 
> >
> >> With that, feel free to add:
> >>
> >> Reviewed-by: Zi Yan <ziy@nvidia.com>
> >>
> >> Thanks.
> >>
> >>>
> >>> Tested with thp disabled kernel:
> >>> Before The fix:
> >>>   # --------------------------------------------------
> >>>   # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
> >>>   # --------------------------------------------------
> >>>   # TAP version 13
> >>>   # Bail out! Reading PMD pagesize failed
> >>>   # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
> >>>   # [FAIL]
> >>>   not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1
> >>>
> >>> After the fix:
> >>>   # --------------------------------------------------
> >>>   # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
> >>>   # --------------------------------------------------
> >>>   # TAP version 13
> >>>   # 1..0 # SKIP Transparent Hugepages not available
> >>>   # [SKIP]
> >>>   ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP
> >>>
> >>> CC: Li Wang <liwang@redhat.com>
> >>> Signed-off-by: Chunyu Hu <chuhu@redhat.com>
> >>> ---
> >>>  tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
> >>>  1 file changed, 7 insertions(+), 2 deletions(-)
> >>>
> >>> diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
> >>> index e0167111bdd1..615b75ca62cc 100644
> >>> --- a/tools/testing/selftests/mm/split_huge_page_test.c
> >>> +++ b/tools/testing/selftests/mm/split_huge_page_test.c
> >>> @@ -21,6 +21,7 @@
> >>>  #include <time.h>
> >>>  #include "vm_util.h"
> >>>  #include "kselftest.h"
> >>> +#include "thp_settings.h"
> >>>
> >>>  uint64_t pagesize;
> >>>  unsigned int pageshift;
> >>> @@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
> >>>  	return status;
> >>>  }
> >>>
> >>> -static void write_file(const char *path, const char *buf, size_t buflen)
> >>> +static void safe_write_file(const char *path, const char *buf, size_t buflen)
> >>>  {
> >>>  	int fd;
> >>>  	ssize_t numwritten;
> >>> @@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
> >>>  	if (ret >= INPUT_MAX)
> >>>  		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
> >>>
> >>> -	write_file(SPLIT_DEBUGFS, input, ret + 1);
> >>> +	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
> >>>  }
> >>>
> >>>  static char *allocate_zero_filled_hugepage(size_t len)
> >>> @@ -772,6 +773,10 @@ int main(int argc, char **argv)
> >>>  		ksft_finished();
> >>>  	}
> >>>
> >>> +	if (!thp_is_enabled()) {
> >>> +		ksft_exit_skip("Transparent Hugepages not available\n");
> >>> +	}
> >>> +
> >>>  	if (argc > 1)
> >>>  		optional_xfs_path = argv[1];
> >>>
> >>> -- 
> >>> 2.53.0
> >>
> >>
> >> Best Regards,
> >> Yan, Zi
> >
> > -- 
> > Sincerely yours,
> > Mike.
> 
> 
> Best Regards,
> Yan, Zi
>

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Lorenzo Stoakes (Oracle) 3 weeks, 3 days ago

On Fri, Mar 13, 2026 at 11:27:30AM -0400, Zi Yan wrote:
> On 13 Mar 2026, at 11:15, Mike Rapoport wrote:
>
> > On Thu, Mar 12, 2026 at 10:11:12AM -0400, Zi Yan wrote:
> >> On 12 Mar 2026, at 7:40, Chunyu Hu wrote:
> >>
> >>> When thp is not enabled on some kernel config such as realtime kernel, the
> >>> test will report failure. Fix the false positive by skipping the test
> >>> directly when thp is not enabled.
> >>>
> >>> There's a naming conflict on write_file() function, both thp_settings.h and
> >>> split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
> >>> the thp_settings.h, rename this local write_file to safe_write_file to
> >>> avoid the conflict. The reason to use 'safe_' is it does some error check.
> >>
> >> Both write_file() do safe checks. Actually the code of both functions looks
> >> almost identical except that the thp_settings one does exit() instead of
> >> ksft_exit_fail_msg().
> >>
> >> Can you rename the split_huge_page_test’s write_file() to write_file_local()
> >> and add a comment above the function like
> >> /* add _local to avoid a function conflict with thp_settings.h */?
> >
> > Looks like we can move the version that uses ksft_exit_fail_msg() to
> > vm_util and drop the other one.
>
> Could work.
>
> tools/testing/selftests/mm/khugepaged.c uses the write_file() from
> thp_settings.h in file_setup_area(), which is part of struct mem_ops.
> I thought struct mem_ops was passed to some syscall, but it turns out
> that it is used locally. The only downside is that file_setup_area() now
> has this ksefltest friendly code. Someone might want to clean up
> tools/testing/selftests/mm/khugepaged.c with kselftests APIs. :)

Yes please :)

With Mike's suggestion, LGTM so:

Reviewed-by: Lorenzo Stoakes (Oracle) <ljs@kernel.org>

>
>
> >
> >> With that, feel free to add:
> >>
> >> Reviewed-by: Zi Yan <ziy@nvidia.com>
> >>
> >> Thanks.
> >>
> >>>
> >>> Tested with thp disabled kernel:
> >>> Before The fix:
> >>>   # --------------------------------------------------
> >>>   # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
> >>>   # --------------------------------------------------
> >>>   # TAP version 13
> >>>   # Bail out! Reading PMD pagesize failed
> >>>   # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
> >>>   # [FAIL]
> >>>   not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1
> >>>
> >>> After the fix:
> >>>   # --------------------------------------------------
> >>>   # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
> >>>   # --------------------------------------------------
> >>>   # TAP version 13
> >>>   # 1..0 # SKIP Transparent Hugepages not available
> >>>   # [SKIP]
> >>>   ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP
> >>>
> >>> CC: Li Wang <liwang@redhat.com>
> >>> Signed-off-by: Chunyu Hu <chuhu@redhat.com>
> >>> ---
> >>>  tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
> >>>  1 file changed, 7 insertions(+), 2 deletions(-)
> >>>
> >>> diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
> >>> index e0167111bdd1..615b75ca62cc 100644
> >>> --- a/tools/testing/selftests/mm/split_huge_page_test.c
> >>> +++ b/tools/testing/selftests/mm/split_huge_page_test.c
> >>> @@ -21,6 +21,7 @@
> >>>  #include <time.h>
> >>>  #include "vm_util.h"
> >>>  #include "kselftest.h"
> >>> +#include "thp_settings.h"
> >>>
> >>>  uint64_t pagesize;
> >>>  unsigned int pageshift;
> >>> @@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
> >>>  	return status;
> >>>  }
> >>>
> >>> -static void write_file(const char *path, const char *buf, size_t buflen)
> >>> +static void safe_write_file(const char *path, const char *buf, size_t buflen)
> >>>  {
> >>>  	int fd;
> >>>  	ssize_t numwritten;
> >>> @@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
> >>>  	if (ret >= INPUT_MAX)
> >>>  		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
> >>>
> >>> -	write_file(SPLIT_DEBUGFS, input, ret + 1);
> >>> +	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
> >>>  }
> >>>
> >>>  static char *allocate_zero_filled_hugepage(size_t len)
> >>> @@ -772,6 +773,10 @@ int main(int argc, char **argv)
> >>>  		ksft_finished();
> >>>  	}
> >>>
> >>> +	if (!thp_is_enabled()) {
> >>> +		ksft_exit_skip("Transparent Hugepages not available\n");
> >>> +	}
> >>> +
> >>>  	if (argc > 1)
> >>>  		optional_xfs_path = argv[1];
> >>>
> >>> --
> >>> 2.53.0
> >>
> >>
> >> Best Regards,
> >> Yan, Zi
> >
> > --
> > Sincerely yours,
> > Mike.
>
>
> Best Regards,
> Yan, Zi

Re: [RESEND PATCH 3/4] selftests/mm: split_huge_page_test: skip the test when thp is not available

Posted by Chunyu Hu 3 weeks, 4 days ago

On Thu, Mar 12, 2026 at 10:11:12AM -0400, Zi Yan wrote:
> On 12 Mar 2026, at 7:40, Chunyu Hu wrote:
> 
> > When thp is not enabled on some kernel config such as realtime kernel, the
> > test will report failure. Fix the false positive by skipping the test
> > directly when thp is not enabled.
> >
> > There's a naming conflict on write_file() function, both thp_settings.h and
> > split_huge_page_test.c define it. To make use of thp_is_enabled() helper in
> > the thp_settings.h, rename this local write_file to safe_write_file to
> > avoid the conflict. The reason to use 'safe_' is it does some error check.
> 
> Both write_file() do safe checks. Actually the code of both functions looks
> almost identical except that the thp_settings one does exit() instead of
> ksft_exit_fail_msg().

indeed.

> 
> Can you rename the split_huge_page_test’s write_file() to write_file_local()
> and add a comment above the function like
> /* add _local to avoid a function conflict with thp_settings.h */?

Good point. I'll add that in the v2.

> 
> With that, feel free to add:
> 
> Reviewed-by: Zi Yan <ziy@nvidia.com>

Thank you for the rewview. Will add.

> 
> Thanks.
> 
> >
> > Tested with thp disabled kernel:
> > Before The fix:
> >   # --------------------------------------------------
> >   # running ./split_huge_page_test /tmp/xfs_dir_Ywup9p
> >   # --------------------------------------------------
> >   # TAP version 13
> >   # Bail out! Reading PMD pagesize failed
> >   # # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
> >   # [FAIL]
> >   not ok 61 split_huge_page_test /tmp/xfs_dir_Ywup9p # exit=1
> >
> > After the fix:
> >   # --------------------------------------------------
> >   # running ./split_huge_page_test /tmp/xfs_dir_YHPUPl
> >   # --------------------------------------------------
> >   # TAP version 13
> >   # 1..0 # SKIP Transparent Hugepages not available
> >   # [SKIP]
> >   ok 6 split_huge_page_test /tmp/xfs_dir_YHPUPl # SKIP
> >
> > CC: Li Wang <liwang@redhat.com>
> > Signed-off-by: Chunyu Hu <chuhu@redhat.com>
> > ---
> >  tools/testing/selftests/mm/split_huge_page_test.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c
> > index e0167111bdd1..615b75ca62cc 100644
> > --- a/tools/testing/selftests/mm/split_huge_page_test.c
> > +++ b/tools/testing/selftests/mm/split_huge_page_test.c
> > @@ -21,6 +21,7 @@
> >  #include <time.h>
> >  #include "vm_util.h"
> >  #include "kselftest.h"
> > +#include "thp_settings.h"
> >
> >  uint64_t pagesize;
> >  unsigned int pageshift;
> > @@ -255,7 +256,7 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len,
> >  	return status;
> >  }
> >
> > -static void write_file(const char *path, const char *buf, size_t buflen)
> > +static void safe_write_file(const char *path, const char *buf, size_t buflen)
> >  {
> >  	int fd;
> >  	ssize_t numwritten;
> > @@ -283,7 +284,7 @@ static void write_debugfs(const char *fmt, ...)
> >  	if (ret >= INPUT_MAX)
> >  		ksft_exit_fail_msg("%s: Debugfs input is too long\n", __func__);
> >
> > -	write_file(SPLIT_DEBUGFS, input, ret + 1);
> > +	safe_write_file(SPLIT_DEBUGFS, input, ret + 1);
> >  }
> >
> >  static char *allocate_zero_filled_hugepage(size_t len)
> > @@ -772,6 +773,10 @@ int main(int argc, char **argv)
> >  		ksft_finished();
> >  	}
> >
> > +	if (!thp_is_enabled()) {
> > +		ksft_exit_skip("Transparent Hugepages not available\n");
> > +	}
> > +
> >  	if (argc > 1)
> >  		optional_xfs_path = argv[1];
> >
> > -- 
> > 2.53.0
> 
> 
> Best Regards,
> Yan, Zi
>