Re: [PATCH v2 00/10] Misc powerpc selftests kernel fixes and cleanups

2026-03-30 Thread Madhavan Srinivasan
On Mon, 09 Mar 2026 23:44:23 +0530, Ritesh Harjani (IBM) wrote:
> v1->v2:
> 
> - dropped debug_vm_pgtable patch which adds a testcase to simulate the
>   failure scenario. Since it belongs to linux-mm, I will send that out
>   separately.
> - Modified Patch-2 in this series to also cover PMD device migration
>   entry (in addition to PMD THP migration entry). Hence dropped the
>   previous RB tag.
> - Added a new Patch-3 to fix another selftests WARNING.
> - Fixed commit subject of Patch-10.
> - Changed subject pre-fix of few patches to be consistent with others
>   (powerpc/64s)
> - Added RB tags
> 
> [...]

Applied to powerpc/next.

[01/10] powerpc/pgtable-frag: Fix bad page state in pte_frag_destroy

https://git.kernel.org/powerpc/c/fda4d71651f71c44b35829d13f3c8bf920032f77
[02/10] powerpc/64s: Fix unmap race with PMD migration entries

https://git.kernel.org/powerpc/c/bbcbf045d6c778e82b47a35fc8728387708e9a3d
[03/10] powerpc/64s: Fix _HPAGE_CHG_MASK to include _PAGE_SPECIAL bit

https://git.kernel.org/powerpc/c/68b1fa0ed5c84769e4e60d58f6a5af37e7273b51
[04/10] powerpc/64s/tlbflush-radix: Remove unused radix__flush_tlb_pwc()

https://git.kernel.org/powerpc/c/4a342f3e6f6848c816a661d8d7b10c75430598cf
[05/10] powerpc/64s: Move serialize_against_pte_lookup() to hash_pgtable.c

https://git.kernel.org/powerpc/c/bf7c1497d2568ff803a0b0fc6728a1c06d11bf6e
[06/10] powerpc/64s: Kill the unused argument of exit_lazy_flush_tlb

https://git.kernel.org/powerpc/c/4894e2fb7b9a25cef843ee2c3b2ac49fd808647d
[07/10] powerpc/64s: Rename tlbie_va_lpid to tlbie_va_pid_lpid

https://git.kernel.org/powerpc/c/7bcfba20e946ec160fd72c3a0b4cf6e3e845d629
[08/10] powerpc/64s: Rename tlbie_lpid_va to tlbie_va_lpid

https://git.kernel.org/powerpc/c/f074059c7a4d4b93914eee404391dcdb0fd60aa6
[09/10] powerpc/64s: Make use of H_RPTI_TYPE_ALL macro

https://git.kernel.org/powerpc/c/24eb6378408fc125eacc4ad498d120ecf7becc35
[10/10] powerpc: Print MMU_FTRS_POSSIBLE & MMU_FTRS_ALWAYS at startup

https://git.kernel.org/powerpc/c/07791ff060dd3aa270cc03861f2599d81a77b97f

cheers



Re: [PATCH v2 00/10] Misc powerpc selftests kernel fixes and cleanups

2026-03-10 Thread IBM
Venkat Rao Bagalkote  writes:

> On 09/03/26 11:44 pm, Ritesh Harjani (IBM) wrote:
>> v1->v2:
>>
>> - dropped debug_vm_pgtable patch which adds a testcase to simulate the
>>failure scenario. Since it belongs to linux-mm, I will send that out
>>separately.
>> - Modified Patch-2 in this series to also cover PMD device migration
>>entry (in addition to PMD THP migration entry). Hence dropped the
>>previous RB tag.
>> - Added a new Patch-3 to fix another selftests WARNING.
>> - Fixed commit subject of Patch-10.
>> - Changed subject pre-fix of few patches to be consistent with others
>>(powerpc/64s)
>> - Added RB tags
>>
>> This patch series addresses selftests issues w.r.t warnings or
>> VM_BUG_ONs seen mainly on book3s64 powerpc kernel. This also carries
>> cleanups and refactoring changes which I identified while reviewing
>> other's patches and/or during code walkthrough.
>>
>> Suggestions and feedback are welcome!
>>
>> Ritesh Harjani (IBM) (10):
>>powerpc/pgtable-frag: Fix bad page state in pte_frag_destroy
>>powerpc/64s: Fix unmap race with PMD migration entries
>>powerpc/64s: Fix _HPAGE_CHG_MASK to include _PAGE_SPECIAL bit
>>powerpc/64s/tlbflush-radix: Remove unused radix__flush_tlb_pwc()
>>powerpc/64s: Move serialize_against_pte_lookup() to hash_pgtable.c
>>powerpc/64s: Kill the unused argument of exit_lazy_flush_tlb
>>powerpc/64s: Rename tlbie_va_lpid to tlbie_va_pid_lpid
>>powerpc/64s: Rename tlbie_lpid_va to tlbie_va_lpid
>>powerpc/64s: Make use of H_RPTI_TYPE_ALL macro
>>powerpc: Print MMU_FTRS_POSSIBLE & MMU_FTRS_ALWAYS at startup
>>
>>   arch/powerpc/include/asm/book3s/64/pgtable.h  | 20 +-
>>   .../include/asm/book3s/64/tlbflush-radix.h|  1 -
>>   arch/powerpc/kernel/setup-common.c|  4 ++
>>   arch/powerpc/mm/book3s64/hash_pgtable.c   | 21 +++
>>   arch/powerpc/mm/book3s64/internal.h   |  2 -
>>   arch/powerpc/mm/book3s64/pgtable.c| 40 +++-
>>   arch/powerpc/mm/book3s64/radix_tlb.c  | 61 ---
>>   arch/powerpc/mm/pgtable-frag.c|  1 +
>>   8 files changed, 79 insertions(+), 71 deletions(-)
>>
> Tested-by: Venkat Rao Bagalkote 

Thanks a lot!

>
> I applied the patch series on top of mainline and verified that the kernel
> builds and boots successfully.
>
> I also ran the following test suites on both RADIX (POWER11) and HASH 
> (POWER9)
> MMU configurations:
>
>    - tools/testing/selftests/mm
>    - tools/testing/selftests/memory-hotplug
>    - tools/testing/selftests/powerpc/mm
>    - tools/testing/selftests/powerpc/cache_shape
>    - tools/testing/selftests/powerpc/copyloops
>
> In addition, I executed basic sanity and stress tests, including:
> stutter, eatmemory, hugepage_sanity, fork_mem, memory_api mprotect,
> vatest, and several transparent-hugepage sanity checks.

Thanks Venkat for verifying this extensively.
So other than couple of hmm tests, there shouldn't be any other kernel
warnings or any VM_BUG_ONs() hitting after this patch series.
(we discussed this internally too!)

And as for those warnings with hmm tests fixes - will fix them in a
separate patch series later (as those looks to be non-powerpc fixes).

This should also enable Venkat and other CI systems to run mm selftests
on book3s64 PowerPC without any kernel issues.

>
> All tests passed without regressions.

Thanks! 

>
> Regards,
> Venkat



Re: [PATCH v2 00/10] Misc powerpc selftests kernel fixes and cleanups

2026-03-10 Thread Venkat Rao Bagalkote



On 09/03/26 11:44 pm, Ritesh Harjani (IBM) wrote:

v1->v2:

- dropped debug_vm_pgtable patch which adds a testcase to simulate the
   failure scenario. Since it belongs to linux-mm, I will send that out
   separately.
- Modified Patch-2 in this series to also cover PMD device migration
   entry (in addition to PMD THP migration entry). Hence dropped the
   previous RB tag.
- Added a new Patch-3 to fix another selftests WARNING.
- Fixed commit subject of Patch-10.
- Changed subject pre-fix of few patches to be consistent with others
   (powerpc/64s)
- Added RB tags

This patch series addresses selftests issues w.r.t warnings or
VM_BUG_ONs seen mainly on book3s64 powerpc kernel. This also carries
cleanups and refactoring changes which I identified while reviewing
other's patches and/or during code walkthrough.

Suggestions and feedback are welcome!

Ritesh Harjani (IBM) (10):
   powerpc/pgtable-frag: Fix bad page state in pte_frag_destroy
   powerpc/64s: Fix unmap race with PMD migration entries
   powerpc/64s: Fix _HPAGE_CHG_MASK to include _PAGE_SPECIAL bit
   powerpc/64s/tlbflush-radix: Remove unused radix__flush_tlb_pwc()
   powerpc/64s: Move serialize_against_pte_lookup() to hash_pgtable.c
   powerpc/64s: Kill the unused argument of exit_lazy_flush_tlb
   powerpc/64s: Rename tlbie_va_lpid to tlbie_va_pid_lpid
   powerpc/64s: Rename tlbie_lpid_va to tlbie_va_lpid
   powerpc/64s: Make use of H_RPTI_TYPE_ALL macro
   powerpc: Print MMU_FTRS_POSSIBLE & MMU_FTRS_ALWAYS at startup

  arch/powerpc/include/asm/book3s/64/pgtable.h  | 20 +-
  .../include/asm/book3s/64/tlbflush-radix.h|  1 -
  arch/powerpc/kernel/setup-common.c|  4 ++
  arch/powerpc/mm/book3s64/hash_pgtable.c   | 21 +++
  arch/powerpc/mm/book3s64/internal.h   |  2 -
  arch/powerpc/mm/book3s64/pgtable.c| 40 +++-
  arch/powerpc/mm/book3s64/radix_tlb.c  | 61 ---
  arch/powerpc/mm/pgtable-frag.c|  1 +
  8 files changed, 79 insertions(+), 71 deletions(-)


Tested-by: Venkat Rao Bagalkote 

I applied the patch series on top of mainline and verified that the kernel
builds and boots successfully.

I also ran the following test suites on both RADIX (POWER11) and HASH 
(POWER9)

MMU configurations:

  - tools/testing/selftests/mm
  - tools/testing/selftests/memory-hotplug
  - tools/testing/selftests/powerpc/mm
  - tools/testing/selftests/powerpc/cache_shape
  - tools/testing/selftests/powerpc/copyloops

In addition, I executed basic sanity and stress tests, including:
stutter, eatmemory, hugepage_sanity, fork_mem, memory_api mprotect,
vatest, and several transparent-hugepage sanity checks.

All tests passed without regressions.

Regards,
Venkat