Hi all,

Powerpc systems with a 64K base page size exposed several issues while
running mm selftests. Some tests assume specific hugetlb configurations,
use incorrect interfaces, or fail instead of skipping when the required
kernel features are not available.

This series fixes these issues and improves test robustness.

Please review the patches and provide any feedback or suggestions for
improvement.

Thanks,
Sayali

---
v6->v7
  - Rebased onto the latest mm-new branch, top commit of the base is
    commit 0cec77cfd531 ("mm/filemap: fix page_cache_prev_miss() when no hole is found")
  - Dropped "selftest/mm: align memory size to huge page size in
    hugepage-mremap test":
    Similar patch has already been merged.
  - For "selftests/mm: skip uffd-stress test when nr_pages_per_cpu is zero":
    Changed format specifier of bytes to %zu as per review
    comment.
  - For "selftests/mm: move hwpoison setup into run_test() and silence
    modprobe output for memory-failure category":
    Added condition to avoid executing the test pipeline for skipped tests and assign
    ret directly.

v6: https://lore.kernel.org/all/cover.1777877814.git.sayalip@linux.ibm.com/
---
v5->v6
  - For "selftests/mm: restore default nr_hugepages value via exit trap
    in charge_reserved_hugetlb.sh":
    Extended trap to include INT and TERM in hugetlb tests to ensure
    nr_hugepages is restored on signal-based exits.
  - For "selftests/mm: restore default nr_hugepages value via exit trap
    in hugetlb_reparenting_test.sh":
    Extended trap to include INT and TERM in hugetlb tests to ensure
    nr_hugepages is restored on signal-based exits.
  - For "selftests/mm: fix cgroup task placement and drop memory.current
    checks in hugetlb_reparenting_test.sh":
    Updated the subshell invocation to pass variables as positional arguments
    instead of interpolating them into the command string as per review comment.
  - For "selftests/mm: move hwpoison setup into run_test() and silence modprobe
    output for memory-failure category":
    Updated the condition to avoid unconditional module removal and ensures it is
    only unloaded if loaded by the test.

v5: https://lore.kernel.org/all/cover.1776150071.git.sayalip@linux.ibm.com/
---
---
v4->v5
  - For "selftests/mm: move hwpoison setup into run_test() and silence modprobe
    output for memory-failure category":
    Renamed LOADED_MOD to LOADED_HWPOISON_INJECT_MOD as per review comment.
  - Dropped "selftests/cgroup: extend test_hugetlb_memcg.c to support all huge
    page sizes":
    Due to intermittent failures observed on some systems.
  - Added Tested-by from Venkat for all patches.
  - Added Reviewed-by from Zi yan for patches 7,8 and 13.
  - Added Acked-by from Zi yan for patches 1 and 3.

v4: https://lore.kernel.org/all/cover.1775466329.git.sayalip@linux.ibm.com/
---
v3->v4

  - selftests/mm: restore default nr_hugepages value via EXIT
    trap in charge_reserved_hugetlb.sh:
    Updated to use an EXIT trap to restore the original nr_hugepages
    once at script termination, ensuring reliable restoration on all
    exit paths.

  - selftest/mm: fix cgroup task placement and drop memory.current checks
    in hugetlb_reparenting_test.sh:
    Updated to use a deterministic synchronization method by migrating the
    workload process within a wrapper subshell before calling exec() to 
    ensure correct cgroup accounting.

  - selftest/mm: align memory size to huge page size in hugepage-mremap
    test:
    Update test code directly to align memory allocations to the hugepage
    size, rather than modifying run_vmtests.sh. This replaces the previous
    commit “adjust hugepage-mremap test size for large huge pages” with
    a cleaner, direct approach in the test itself.

  - selftests/mm: ensure destination is hugetlb-backed in hugepage-mremap:
    Updated to remove MAP_POPULATE flag for the mmap call as per review
    comment.

  - selftests/mm: skip uffd-wp-mremap if UFFD write-protect is unsupported:
    Updated as per review comment to initialize features to zero and exit
    on all uffd_get_features() failures, avoiding spurious test errors.

  - selftests/mm: move hwpoison setup into run_test() and silence
    modprobe output for memory-failure category:
    Declared LOADED_MOD as local variable and avoided a redundant 
    skip message when the module is not found.
    Updated logic to set exitcode to ksft_skip only when no failure has
    been recorded.

  - selftests/cgroup: extend test_hugetlb_memcg.c to support all huge
    page sizes:
    Updated to skip on memory constraints instead of returning
    EXIT_FAILURE.
    Updated the logic to fully account for the scenario when the
    per-CPU stock is empty and a refill charges MEMCG_CHARGE_BATCH. 
    Updated to avoid overflow on 32-bit systems for memory.max value.

  - Included "selftests/mm: restore default nr_hugepages value via
    EXIT trap in hugetlb_reparenting_test.sh"

  - Included "selftests/mm: free dynamically allocated PMD-sized
    buffers in split_huge_page_test"

  - Included "selftests/mm: clarify alternate unmapping in compaction_test"

  - Dropped "selftests/mm: fix double increment in linked list cleanup in
    compaction_test":
    The behaviour is intentional.

v3: https://lore.kernel.org/all/cover.1774591179.git.sayalip@linux.ibm.com/

---
v2->v3
  - selftests/mm: skip uffd-wp-mremap if UFFD write-protect is unsupported:
    Rename function to check_uffd_wp_feature_supported() as suggested in review.
  - selftest/mm: fix cgroup task placement and drop memory.current checks
    in hugetlb_reparenting_test.sh:
    Drop memory.current validation from the hugetlb reparenting test.
    Keep tolerance at 7MB (reverting earlier increase to 8MB in v1).
  - Included "selftests/mm: allow PUD-level entries in compound testcase of hmm
    tests" patch:
    Extend the compound testcase checks to accept PUD-level mappings.
  - Included "selftests/mm: replace hardcoded THP size with runtime PMD page size in
    hmm tests" patch:
    Use read_pmd_pagesize() instead of TWOMEG and cap maximum THPs in
    benchmarks to avoid integer overflow.

v2: https://lore.kernel.org/all/cover.1773305677.git.sayalip@linux.ibm.com/

---
v1->v2
  - For "selftests/mm: ensure destination is hugetlb-backed in hugepage-mremap":
    update FLAGS definition to MAP_HUGETLB | MAP_SHARED | MAP_POPULATE and
    used it for mmap() calls as suggested during review.

v1: https://lore.kernel.org/all/cover.1773134177.git.sayalip@linux.ibm.com/
---

Sayali Patil (13):
  selftests/mm: restore default nr_hugepages value via exit trap in
    charge_reserved_hugetlb.sh
  selftests/mm: fix hugetlb pathname construction in
    charge_reserved_hugetlb.sh
  selftests/mm: restore default nr_hugepages value via exit trap in
    hugetlb_reparenting_test.sh
  selftests/mm: fix hugetlb pathname construction in
    hugetlb_reparenting_test.sh
  selftests/mm: fix cgroup task placement and drop memory.current checks
    in hugetlb_reparenting_test.sh
  selftests/mm: size tmpfs according to PMD page size in
    split_huge_page_test
  selftests/mm: free dynamically allocated PMD-sized buffers in
    split_huge_page_test
  selftest/mm: register existing mapping with userfaultfd in
    hugetlb-mremap
  selftests/mm: ensure destination is hugetlb-backed in hugetlb-mremap
  selftests/mm: skip uffd-wp-mremap if UFFD write-protect is unsupported
  selftests/mm: skip uffd-stress test when nr_pages_per_cpu is zero
  selftests/mm: move hwpoison setup into run_test() and silence modprobe
    output for memory-failure category
  selftests/mm: clarify alternate unmapping in compaction_test

 .../selftests/mm/charge_reserved_hugetlb.sh   | 45 +++++++++-----
 tools/testing/selftests/mm/compaction_test.c  |  3 +
 tools/testing/selftests/mm/hugetlb-mremap.c   | 32 +++-------
 .../selftests/mm/hugetlb_reparenting_test.sh  | 60 +++++++++---------
 tools/testing/selftests/mm/run_vmtests.sh     | 62 ++++++++++++-------
 .../selftests/mm/split_huge_page_test.c       | 27 +++++---
 tools/testing/selftests/mm/uffd-stress.c      |  6 +-
 tools/testing/selftests/mm/uffd-wp-mremap.c   | 13 ++++
 8 files changed, 151 insertions(+), 97 deletions(-)

-- 
2.52.0

On 21/05/26 12:03, Sayali Patil wrote:
> 
> Hi all,
> 
> Powerpc systems with a 64K base page size exposed several issues while
> running mm selftests. Some tests assume specific hugetlb configurations,
> use incorrect interfaces, or fail instead of skipping when the required
> kernel features are not available.
> 
> This series fixes these issues and improves test robustness.
> 
> Please review the patches and provide any feedback or suggestions for
> improvement.
> 
> Thanks,
> Sayali
> 
> ---
> v6->v7
>    - Rebased onto the latest mm-new branch, top commit of the base is
>      commit 0cec77cfd531 ("mm/filemap: fix page_cache_prev_miss() when no hole is found")
>    - Dropped "selftest/mm: align memory size to huge page size in
>      hugepage-mremap test":
>      Similar patch has already been merged.
>    - For "selftests/mm: skip uffd-stress test when nr_pages_per_cpu is zero":
>      Changed format specifier of bytes to %zu as per review
>      comment.
>    - For "selftests/mm: move hwpoison setup into run_test() and silence
>      modprobe output for memory-failure category":
>      Added condition to avoid executing the test pipeline for skipped tests and assign
>      ret directly.
> 
> v6: https://lore.kernel.org/all/cover.1777877814.git.sayalip@linux.ibm.com/
> ---
> v5->v6
>    - For "selftests/mm: restore default nr_hugepages value via exit trap
>      in charge_reserved_hugetlb.sh":
>      Extended trap to include INT and TERM in hugetlb tests to ensure
>      nr_hugepages is restored on signal-based exits.
>    - For "selftests/mm: restore default nr_hugepages value via exit trap
>      in hugetlb_reparenting_test.sh":
>      Extended trap to include INT and TERM in hugetlb tests to ensure
>      nr_hugepages is restored on signal-based exits.
>    - For "selftests/mm: fix cgroup task placement and drop memory.current
>      checks in hugetlb_reparenting_test.sh":
>      Updated the subshell invocation to pass variables as positional arguments
>      instead of interpolating them into the command string as per review comment.
>    - For "selftests/mm: move hwpoison setup into run_test() and silence modprobe
>      output for memory-failure category":
>      Updated the condition to avoid unconditional module removal and ensures it is
>      only unloaded if loaded by the test.
> 
> v5: https://lore.kernel.org/all/cover.1776150071.git.sayalip@linux.ibm.com/
> ---
> ---
> v4->v5
>    - For "selftests/mm: move hwpoison setup into run_test() and silence modprobe
>      output for memory-failure category":
>      Renamed LOADED_MOD to LOADED_HWPOISON_INJECT_MOD as per review comment.
>    - Dropped "selftests/cgroup: extend test_hugetlb_memcg.c to support all huge
>      page sizes":
>      Due to intermittent failures observed on some systems.
>    - Added Tested-by from Venkat for all patches.
>    - Added Reviewed-by from Zi yan for patches 7,8 and 13.
>    - Added Acked-by from Zi yan for patches 1 and 3.
> 
> v4: https://lore.kernel.org/all/cover.1775466329.git.sayalip@linux.ibm.com/
> ---
> v3->v4
> 
>    - selftests/mm: restore default nr_hugepages value via EXIT
>      trap in charge_reserved_hugetlb.sh:
>      Updated to use an EXIT trap to restore the original nr_hugepages
>      once at script termination, ensuring reliable restoration on all
>      exit paths.
> 
>    - selftest/mm: fix cgroup task placement and drop memory.current checks
>      in hugetlb_reparenting_test.sh:
>      Updated to use a deterministic synchronization method by migrating the
>      workload process within a wrapper subshell before calling exec() to
>      ensure correct cgroup accounting.
> 
>    - selftest/mm: align memory size to huge page size in hugepage-mremap
>      test:
>      Update test code directly to align memory allocations to the hugepage
>      size, rather than modifying run_vmtests.sh. This replaces the previous
>      commit “adjust hugepage-mremap test size for large huge pages” with
>      a cleaner, direct approach in the test itself.
> 
>    - selftests/mm: ensure destination is hugetlb-backed in hugepage-mremap:
>      Updated to remove MAP_POPULATE flag for the mmap call as per review
>      comment.
> 
>    - selftests/mm: skip uffd-wp-mremap if UFFD write-protect is unsupported:
>      Updated as per review comment to initialize features to zero and exit
>      on all uffd_get_features() failures, avoiding spurious test errors.
> 
>    - selftests/mm: move hwpoison setup into run_test() and silence
>      modprobe output for memory-failure category:
>      Declared LOADED_MOD as local variable and avoided a redundant
>      skip message when the module is not found.
>      Updated logic to set exitcode to ksft_skip only when no failure has
>      been recorded.
> 
>    - selftests/cgroup: extend test_hugetlb_memcg.c to support all huge
>      page sizes:
>      Updated to skip on memory constraints instead of returning
>      EXIT_FAILURE.
>      Updated the logic to fully account for the scenario when the
>      per-CPU stock is empty and a refill charges MEMCG_CHARGE_BATCH.
>      Updated to avoid overflow on 32-bit systems for memory.max value.
> 
>    - Included "selftests/mm: restore default nr_hugepages value via
>      EXIT trap in hugetlb_reparenting_test.sh"
> 
>    - Included "selftests/mm: free dynamically allocated PMD-sized
>      buffers in split_huge_page_test"
> 
>    - Included "selftests/mm: clarify alternate unmapping in compaction_test"
> 
>    - Dropped "selftests/mm: fix double increment in linked list cleanup in
>      compaction_test":
>      The behaviour is intentional.
> 
> v3: https://lore.kernel.org/all/cover.1774591179.git.sayalip@linux.ibm.com/
> 
> ---
> v2->v3
>    - selftests/mm: skip uffd-wp-mremap if UFFD write-protect is unsupported:
>      Rename function to check_uffd_wp_feature_supported() as suggested in review.
>    - selftest/mm: fix cgroup task placement and drop memory.current checks
>      in hugetlb_reparenting_test.sh:
>      Drop memory.current validation from the hugetlb reparenting test.
>      Keep tolerance at 7MB (reverting earlier increase to 8MB in v1).
>    - Included "selftests/mm: allow PUD-level entries in compound testcase of hmm
>      tests" patch:
>      Extend the compound testcase checks to accept PUD-level mappings.
>    - Included "selftests/mm: replace hardcoded THP size with runtime PMD page size in
>      hmm tests" patch:
>      Use read_pmd_pagesize() instead of TWOMEG and cap maximum THPs in
>      benchmarks to avoid integer overflow.
> 
> v2: https://lore.kernel.org/all/cover.1773305677.git.sayalip@linux.ibm.com/
> 
> ---
> v1->v2
>    - For "selftests/mm: ensure destination is hugetlb-backed in hugepage-mremap":
>      update FLAGS definition to MAP_HUGETLB | MAP_SHARED | MAP_POPULATE and
>      used it for mmap() calls as suggested during review.
> 
> v1: https://lore.kernel.org/all/cover.1773134177.git.sayalip@linux.ibm.com/
> ---
> 
> Sayali Patil (13):
>    selftests/mm: restore default nr_hugepages value via exit trap in
>      charge_reserved_hugetlb.sh
>    selftests/mm: fix hugetlb pathname construction in
>      charge_reserved_hugetlb.sh
>    selftests/mm: restore default nr_hugepages value via exit trap in
>      hugetlb_reparenting_test.sh
>    selftests/mm: fix hugetlb pathname construction in
>      hugetlb_reparenting_test.sh
>    selftests/mm: fix cgroup task placement and drop memory.current checks
>      in hugetlb_reparenting_test.sh
>    selftests/mm: size tmpfs according to PMD page size in
>      split_huge_page_test
>    selftests/mm: free dynamically allocated PMD-sized buffers in
>      split_huge_page_test
>    selftest/mm: register existing mapping with userfaultfd in
>      hugetlb-mremap
>    selftests/mm: ensure destination is hugetlb-backed in hugetlb-mremap
>    selftests/mm: skip uffd-wp-mremap if UFFD write-protect is unsupported
>    selftests/mm: skip uffd-stress test when nr_pages_per_cpu is zero
>    selftests/mm: move hwpoison setup into run_test() and silence modprobe
>      output for memory-failure category
>    selftests/mm: clarify alternate unmapping in compaction_test
> 
>   .../selftests/mm/charge_reserved_hugetlb.sh   | 45 +++++++++-----
>   tools/testing/selftests/mm/compaction_test.c  |  3 +
>   tools/testing/selftests/mm/hugetlb-mremap.c   | 32 +++-------
>   .../selftests/mm/hugetlb_reparenting_test.sh  | 60 +++++++++---------
>   tools/testing/selftests/mm/run_vmtests.sh     | 62 ++++++++++++-------
>   .../selftests/mm/split_huge_page_test.c       | 27 +++++---
>   tools/testing/selftests/mm/uffd-stress.c      |  6 +-
>   tools/testing/selftests/mm/uffd-wp-mremap.c   | 13 ++++
>   8 files changed, 151 insertions(+), 97 deletions(-)
> 
Hi Andrew,

I checked the AI review comments. Most of them appear to be relatively 
minor.

I have sent one fixup patch ("selftests/mm: use ksft_exit_skip() instead 
of KSFT_SKIP in uffd-stress") for the comment in "selftests/mm: skip 
uffd-stress test when nr_pages_per_cpu is zero". The remaining comments 
were either already discussed earlier or can be addressed later if needed.

Please let me know if you have any additional comments or concerns.

Thanks,
Sayali

On Thu, 21 May 2026 12:03:26 +0530 Sayali Patil <sayalip@linux.ibm.com> wrote:

> Powerpc systems with a 64K base page size exposed several issues while
> running mm selftests. Some tests assume specific hugetlb configurations,
> use incorrect interfaces, or fail instead of skipping when the required
> kernel features are not available.
> 
> This series fixes these issues and improves test robustness.

Thanks, I've re-added this series to mm.git's mm-new branch.

> I checked the AI review comments. Most of them appear to be relatively 
> minor.
> 
> I have sent one fixup patch ("selftests/mm: use ksft_exit_skip() instead 
> of KSFT_SKIP in uffd-stress") for the comment in "selftests/mm: skip 
> uffd-stress test when nr_pages_per_cpu is zero". The remaining comments 
> were either already discussed earlier or can be addressed later if needed.

Cool, thanks for letting us know where we stand with the AI review. 
It helps!