Re: [PATCH v3 6/6] selftests: prctl: introduce tests for disabling THPs except for madvise

[Usama Arif <usamaarif642@gmail.com>]

On 05/08/2025 11:36, David Hildenbrand wrote:
> On 04.08.25 17:40, Usama Arif wrote:
>> The test will set the global system THP setting to never, madvise
>> or always depending on the fixture variant and the 2M setting to
>> inherit before it starts (and reset to original at teardown)
>>
>> This tests if the process can:
>> - successfully set and get the policy to disable THPs expect for madvise.
>> - get hugepages only on MADV_HUGE and MADV_COLLAPSE if the global policy
>>    is madvise/always and only with MADV_COLLAPSE if the global policy is
>>    never.
>> - successfully reset the policy of the process.
>> - after reset, only get hugepages with:
>>    - MADV_COLLAPSE when policy is set to never.
>>    - MADV_HUGE and MADV_COLLAPSE when policy is set to madvise.
>>    - always when policy is set to "always".
>> - repeat the above tests in a forked process to make sure  the policy is
>>    carried across forks.
>>
>> Signed-off-by: Usama Arif <usamaarif642@gmail.com>
>> ---
> 
> [...]
> 
>> +FIXTURE_VARIANT(prctl_thp_disable_except_madvise)
>> +{
>> +    enum thp_enabled thp_policy;
>> +};
>> +
>> +FIXTURE_VARIANT_ADD(prctl_thp_disable_except_madvise, never)
>> +{
>> +    .thp_policy = THP_NEVER,
>> +};
>> +
>> +FIXTURE_VARIANT_ADD(prctl_thp_disable_except_madvise, madvise)
>> +{
>> +    .thp_policy = THP_MADVISE,
>> +};
>> +
>> +FIXTURE_VARIANT_ADD(prctl_thp_disable_except_madvise, always)
>> +{
>> +    .thp_policy = THP_ALWAYS,
>> +};
>> +
>> +FIXTURE_SETUP(prctl_thp_disable_except_madvise)
>> +{
>> +    if (!thp_available())
>> +        SKIP(return, "Transparent Hugepages not available\n");
>> +
>> +    self->pmdsize = read_pmd_pagesize();
>> +    if (!self->pmdsize)
>> +        SKIP(return, "Unable to read PMD size\n");
> 
> Should we test here if the kernel knows PR_THP_DISABLE_EXCEPT_ADVISED, and if not, skip?
> 
> Might be as simple as trying issuing two prctl, and making sure the first disabling attempt doesn't fail. If so, SKIP.
> 
> Nothing else jumped at me. Can you include a test run result in the patch description?
> 

Instead of 2 prctls, I think doing just the below should be enough:

diff --git a/tools/testing/selftests/mm/prctl_thp_disable.c b/tools/testing/selftests/mm/prctl_thp_disable.c
index 93cedaa59854..da28bc4441ed 100644
--- a/tools/testing/selftests/mm/prctl_thp_disable.c
+++ b/tools/testing/selftests/mm/prctl_thp_disable.c
@@ -236,6 +236,9 @@ FIXTURE_SETUP(prctl_thp_disable_except_madvise)
        if (!self->pmdsize)
                SKIP(return, "Unable to read PMD size\n");
 
+       if (prctl(PR_SET_THP_DISABLE, 1, PR_THP_DISABLE_EXCEPT_ADVISED, NULL, NULL))
+               SKIP(return, "Unable to set PR_THP_DISABLE_EXCEPT_ADVISED\n");
+
        thp_save_settings();
        thp_read_settings(&self->settings);
        self->settings.thp_enabled = variant->thp_policy;



Will include the test run result in the last patch description. Just adding it here as well:

./prctl_thp_disable
TAP version 13
1..12
# Starting 12 tests from 6 test cases.
#  RUN           prctl_thp_disable_completely.never.nofork ...
#            OK  prctl_thp_disable_completely.never.nofork
ok 1 prctl_thp_disable_completely.never.nofork
#  RUN           prctl_thp_disable_completely.never.fork ...
#            OK  prctl_thp_disable_completely.never.fork
ok 2 prctl_thp_disable_completely.never.fork
#  RUN           prctl_thp_disable_completely.madvise.nofork ...
#            OK  prctl_thp_disable_completely.madvise.nofork
ok 3 prctl_thp_disable_completely.madvise.nofork
#  RUN           prctl_thp_disable_completely.madvise.fork ...
#            OK  prctl_thp_disable_completely.madvise.fork
ok 4 prctl_thp_disable_completely.madvise.fork
#  RUN           prctl_thp_disable_completely.always.nofork ...
#            OK  prctl_thp_disable_completely.always.nofork
ok 5 prctl_thp_disable_completely.always.nofork
#  RUN           prctl_thp_disable_completely.always.fork ...
#            OK  prctl_thp_disable_completely.always.fork
ok 6 prctl_thp_disable_completely.always.fork
#  RUN           prctl_thp_disable_except_madvise.never.nofork ...
#            OK  prctl_thp_disable_except_madvise.never.nofork
ok 7 prctl_thp_disable_except_madvise.never.nofork
#  RUN           prctl_thp_disable_except_madvise.never.fork ...
#            OK  prctl_thp_disable_except_madvise.never.fork
ok 8 prctl_thp_disable_except_madvise.never.fork
#  RUN           prctl_thp_disable_except_madvise.madvise.nofork ...
#            OK  prctl_thp_disable_except_madvise.madvise.nofork
ok 9 prctl_thp_disable_except_madvise.madvise.nofork
#  RUN           prctl_thp_disable_except_madvise.madvise.fork ...
#            OK  prctl_thp_disable_except_madvise.madvise.fork
ok 10 prctl_thp_disable_except_madvise.madvise.fork
#  RUN           prctl_thp_disable_except_madvise.always.nofork ...
#            OK  prctl_thp_disable_except_madvise.always.nofork
ok 11 prctl_thp_disable_except_madvise.always.nofork
#  RUN           prctl_thp_disable_except_madvise.always.fork ...
#            OK  prctl_thp_disable_except_madvise.always.fork
ok 12 prctl_thp_disable_except_madvise.always.fork
# PASSED: 12 / 12 tests passed.
# Totals: pass:12 fail:0 xfail:0 xpass:0 skip:0 error:0