Re: [PATCH 2/2] selftests/mm: rewrite gup_test as a standalone harness-based selftest

[Sarthak Sharma <sarthak.sharma@arm.com>]

Hi David!

On 5/18/26 9:10 PM, David Hildenbrand (Arm) wrote:
> On 5/15/26 10:48, Sarthak Sharma wrote:
>> Rewrite gup_test.c using kselftest_harness.h. The new test uses
>> FIXTURE_VARIANT to cover seven configurations (private/shared,
>> read/write, THP, hugetlb) and runs four test cases per variant
>> (GUP_BASIC_TEST, PIN_BASIC_TEST, DUMP_USER_PAGES_TEST with get and
>> pin), giving 28 TAP-reported cases in total without requiring any
>> command-line arguments.
>>
>> Update run_vmtests.sh: remove run_gup_matrix() and the multiple flagged
>> invocations of gup_test, replacing them with a single unconditional
>> invocation. Benchmark functionality is handled by tools/mm/gup_bench
>> introduced in the previous patch.
>>
> 
> Thanks for working on this!
> 
> Compared to the old test (e.g., matrix) are we now performing less tests? What's
> the runtime of the new test?

Yes, in the current version we are not covering pin-longterm, gup-fast,
and pin-fast tests. However, after going through Sashiko's review, I
realized these can be added as well since they exercise separate code paths.

I'll also restore the coverage for all page counts similar to
run_gup_matrix() and send a v2.

> 
> Would be good to incorporate that information in the patch description.

Sure, I'll add that.