Clearing PXD_TABLE_BIT i.e bit[1] on a page table entry always operates on
the assumption that subsequent PXD_VALID i.e bit[0] is set. That's because
bits[1:0]="01" makes a block mapping. So it is prudent to treat bits[1:0]
as a single register field, which should be updated as block or table etc.
Although mk_[pmd|pud]_sect_prot() helpers go to some extent in using these
PXD_TYPE_SECT macros, their usage is not really consistent else where.

This series removes these table bit clearing for block mapping creation and
eventually completely drops off those table macros.

This series applies on v6.12-rc1.

Cc: Marc Zyngier <maz@kernel.org>
Cc: Oliver Upton <oliver.upton@linux.dev>
Cc: James Morse <james.morse@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: kvmarm@lists.linux.dev
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org

Anshuman Khandual (5):
  arm64/mm: Drop pte_mkhuge()
  arm64/mm: Replace PXD_TABLE_BIT with PXD_TYPE_[MASK|SECT]
  arm64/ptdump: Test PMD_TYPE_MASK for block mapping
  KVM: arm64: ptdump: Test PMD_TYPE_MASK for block mapping
  arm64/mm: Drop PXD_TABLE_BIT

 arch/arm64/include/asm/pgtable-hwdef.h |  6 +-----
 arch/arm64/include/asm/pgtable.h       | 20 ++++++++------------
 arch/arm64/kvm/ptdump.c                |  4 ++--
 arch/arm64/mm/hugetlbpage.c            |  2 +-
 arch/arm64/mm/ptdump.c                 |  8 ++++----
 5 files changed, 16 insertions(+), 24 deletions(-)

-- 
2.25.1

On 05/10/2024 13:38, Anshuman Khandual wrote:
> Clearing PXD_TABLE_BIT i.e bit[1] on a page table entry always operates on
> the assumption that subsequent PXD_VALID i.e bit[0] is set. That's because
> bits[1:0]="01" makes a block mapping. So it is prudent to treat bits[1:0]
> as a single register field, which should be updated as block or table etc.
> Although mk_[pmd|pud]_sect_prot() helpers go to some extent in using these
> PXD_TYPE_SECT macros, their usage is not really consistent else where.
> 
> This series removes these table bit clearing for block mapping creation and
> eventually completely drops off those table macros.

Given the issue I just noticed in patch 2, I'm not sure if it's going to be
practical to remove the table bit after all? Sorry I didn't spot this before.

> 
> This series applies on v6.12-rc1.
> 
> Cc: Marc Zyngier <maz@kernel.org>
> Cc: Oliver Upton <oliver.upton@linux.dev>
> Cc: James Morse <james.morse@arm.com>
> Cc: Catalin Marinas <catalin.marinas@arm.com>
> Cc: Will Deacon <will@kernel.org>
> Cc: Ard Biesheuvel <ardb@kernel.org>
> Cc: Ryan Roberts <ryan.roberts@arm.com>
> Cc: Mark Rutland <mark.rutland@arm.com>
> Cc: kvmarm@lists.linux.dev
> Cc: linux-arm-kernel@lists.infradead.org
> Cc: linux-kernel@vger.kernel.org
> 
> Anshuman Khandual (5):
>   arm64/mm: Drop pte_mkhuge()
>   arm64/mm: Replace PXD_TABLE_BIT with PXD_TYPE_[MASK|SECT]
>   arm64/ptdump: Test PMD_TYPE_MASK for block mapping
>   KVM: arm64: ptdump: Test PMD_TYPE_MASK for block mapping
>   arm64/mm: Drop PXD_TABLE_BIT
> 
>  arch/arm64/include/asm/pgtable-hwdef.h |  6 +-----
>  arch/arm64/include/asm/pgtable.h       | 20 ++++++++------------
>  arch/arm64/kvm/ptdump.c                |  4 ++--
>  arch/arm64/mm/hugetlbpage.c            |  2 +-
>  arch/arm64/mm/ptdump.c                 |  8 ++++----
>  5 files changed, 16 insertions(+), 24 deletions(-)
>

On 10/9/24 19:02, Ryan Roberts wrote:
> On 05/10/2024 13:38, Anshuman Khandual wrote:
>> Clearing PXD_TABLE_BIT i.e bit[1] on a page table entry always operates on
>> the assumption that subsequent PXD_VALID i.e bit[0] is set. That's because
>> bits[1:0]="01" makes a block mapping. So it is prudent to treat bits[1:0]
>> as a single register field, which should be updated as block or table etc.
>> Although mk_[pmd|pud]_sect_prot() helpers go to some extent in using these
>> PXD_TYPE_SECT macros, their usage is not really consistent else where.
>>
>> This series removes these table bit clearing for block mapping creation and
>> eventually completely drops off those table macros.
> Given the issue I just noticed in patch 2, I'm not sure if it's going to be
> practical to remove the table bit after all? Sorry I didn't spot this before.

Now that arch_make_huge_pte() seems to be all good in patch 1, and there might
be a solution for pmd_present() via de-coupled pmd_mkinvalid(), would it still
not possible to drop PXD_TABLE_BIT ?