[v2,0/4] Fixes for hugetlb and vmalloc on arm64

Message

Ryan Roberts Feb. 17, 2025, 2:04 p.m. UTC

Hi All,

This series contains some fixes for hugetlb on arm64, and is split out from v1
of a wider series at [1]. While the last patch is technically targetting core-mm
and is not directly related to arm64, I'd like to to go via the arm64 tree so
that the wider performance improvement series (v2 to be posted shortly) that
depends on this series doesn't have to be robust to the fix not being present.

I've included maintainers/reviewers for all the arches that are (trivially)
touched due to the API changes, hoping for some ACKs.

Changes since v1 [1]
====================
- Added Rb from Anshuman - Thanks!
- Added "#ifndef __PAGETABLE_PMD_FOLDED" around PUD_SIZE in flush_hugetlb_tlb_range()

I've marked all of these as candidates for backport to stable.

Applies on top of v6.14-rc3. All mm selftests run and pass.

[1] https://lore.kernel.org/linux-arm-kernel/[e-mail]/

Thanks,
Ryan

Ryan Roberts (4):
  mm: hugetlb: Add huge page size param to huge_ptep_get_and_clear()
  arm64: hugetlb: Fix huge_ptep_get_and_clear() for non-present ptes
  arm64: hugetlb: Fix flush_hugetlb_tlb_range() invalidation level
  mm: Don't skip arch_sync_kernel_mappings() in error paths

 arch/arm64/include/asm/hugetlb.h     | 26 ++++++++++-----
 arch/arm64/mm/hugetlbpage.c          | 48 +++++++++++++---------------
 arch/loongarch/include/asm/hugetlb.h |  6 ++--
 arch/mips/include/asm/hugetlb.h      |  6 ++--
 arch/parisc/include/asm/hugetlb.h    |  2 +-
 arch/parisc/mm/hugetlbpage.c         |  2 +-
 arch/powerpc/include/asm/hugetlb.h   |  6 ++--
 arch/riscv/include/asm/hugetlb.h     |  3 +-
 arch/riscv/mm/hugetlbpage.c          |  2 +-
 arch/s390/include/asm/hugetlb.h      | 12 ++++---
 arch/s390/mm/hugetlbpage.c           | 10 ++++--
 arch/sparc/include/asm/hugetlb.h     |  2 +-
 arch/sparc/mm/hugetlbpage.c          |  2 +-
 include/asm-generic/hugetlb.h        |  2 +-
 include/linux/hugetlb.h              |  4 ++-
 mm/hugetlb.c                         |  4 +--
 mm/memory.c                          |  6 ++--
 mm/vmalloc.c                         |  4 +--
 18 files changed, 87 insertions(+), 60 deletions(-)

--
2.43.0

Comments

Catalin Marinas Feb. 19, 2025, 7:10 p.m. UTC | #1

Will, Ryan,

On Mon, Feb 17, 2025 at 02:04:13PM +0000, Ryan Roberts wrote:
> This series contains some fixes for hugetlb on arm64, and is split out from v1
> of a wider series at [1]. While the last patch is technically targetting core-mm
> and is not directly related to arm64, I'd like to to go via the arm64 tree so
> that the wider performance improvement series (v2 to be posted shortly) that
> depends on this series doesn't have to be robust to the fix not being present.
> 
> I've included maintainers/reviewers for all the arches that are (trivially)
> touched due to the API changes, hoping for some ACKs.

These fixes look fine to me and I think we should get them in for 6.14.
I think Andrew was ok with them going in via the arm64 tree:

https://lore.kernel.org/linux-arm-kernel/[e-mail]/

Will Deacon Feb. 21, 2025, 3:35 p.m. UTC | #2

On Wed, Feb 19, 2025 at 07:10:32PM +0000, Catalin Marinas wrote:
> Will, Ryan,
> 
> On Mon, Feb 17, 2025 at 02:04:13PM +0000, Ryan Roberts wrote:
> > This series contains some fixes for hugetlb on arm64, and is split out from v1
> > of a wider series at [1]. While the last patch is technically targetting core-mm
> > and is not directly related to arm64, I'd like to to go via the arm64 tree so
> > that the wider performance improvement series (v2 to be posted shortly) that
> > depends on this series doesn't have to be robust to the fix not being present.
> > 
> > I've included maintainers/reviewers for all the arches that are (trivially)
> > touched due to the API changes, hoping for some ACKs.
> 
> These fixes look fine to me and I think we should get them in for 6.14.
> I think Andrew was ok with them going in via the arm64 tree:
> 
> https://lore.kernel.org/linux-arm-kernel/[e-mail]/

I think the diffstat looks worse than it really is, as the arch changes
are reasonably mechanical. I'd like to let it sit in next for a few days
though, so I'll pick this up once we've resolved my comment on patch #2.

Ryan -- did you find all of these issues by inspection, or are you aware
of anybody hitting them in practice?

Will

Ryan Roberts Feb. 24, 2025, 12:13 p.m. UTC | #3

On 21/02/2025 15:35, Will Deacon wrote:
> On Wed, Feb 19, 2025 at 07:10:32PM +0000, Catalin Marinas wrote:
>> Will, Ryan,
>>
>> On Mon, Feb 17, 2025 at 02:04:13PM +0000, Ryan Roberts wrote:
>>> This series contains some fixes for hugetlb on arm64, and is split out from v1
>>> of a wider series at [1]. While the last patch is technically targetting core-mm
>>> and is not directly related to arm64, I'd like to to go via the arm64 tree so
>>> that the wider performance improvement series (v2 to be posted shortly) that
>>> depends on this series doesn't have to be robust to the fix not being present.
>>>
>>> I've included maintainers/reviewers for all the arches that are (trivially)
>>> touched due to the API changes, hoping for some ACKs.
>>
>> These fixes look fine to me and I think we should get them in for 6.14.
>> I think Andrew was ok with them going in via the arm64 tree:
>>
>> https://lore.kernel.org/linux-arm-kernel/[e-mail]/
> 
> I think the diffstat looks worse than it really is, as the arch changes
> are reasonably mechanical. I'd like to let it sit in next for a few days
> though, so I'll pick this up once we've resolved my comment on patch #2.

Sounds good to me; if you're happy with my proposal in the patch 2 thread, then
I'll respin.

> 
> Ryan -- did you find all of these issues by inspection, or are you aware
> of anybody hitting them in practice?

All by inspection. So I guess they are not urgent to fix from that perspective.

> 
> Will