On Wed, Mar 31, 2021 at 02:23:44PM -0400, Zack Rusin wrote:
> The pagetable walk callbacks in mm/mapping_dirty_helpers.c depend
> on a set of helpers from which pud_dirty(pud) was missing. I'm
> assuming mapping_dirty_helpers weren't used on ARM64 before
> because the missing pud_dirty is causing a compilation error.
>
> The drivers/gpu/drm/vmwgfx code uses mapping_dirty_helpers and
> has been ported to ARM64 but it depends on this code getting in
> first in order to compile/work.
>
> Cc: Catalin Marinas <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> Cc: [email protected]
> Signed-off-by: Zack Rusin <[email protected]>
> ---
> arch/arm64/include/asm/pgtable.h | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
> index 47027796c2f9..ecd80f87a996 100644
> --- a/arch/arm64/include/asm/pgtable.h
> +++ b/arch/arm64/include/asm/pgtable.h
> @@ -451,6 +451,7 @@ static inline pmd_t pmd_mkdevmap(pmd_t pmd)
> #define pfn_pmd(pfn,prot) __pmd(__phys_to_pmd_val((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot))
> #define mk_pmd(page,prot) pfn_pmd(page_to_pfn(page),prot)
>
> +#define pud_dirty(pud) pte_dirty(pud_pte(pud))
> #define pud_young(pud) pte_young(pud_pte(pud))
> #define pud_mkyoung(pud) pte_pud(pte_mkyoung(pud_pte(pud)))
> #define pud_write(pud) pte_write(pud_pte(pud))
I think pud_dirty() should only be defined if
CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD which is not the case on
arm64. Well, the pud_young/mkyoung/write should probably also be removed
until we have pud pages on arm64.
The wp_clean_pud_entry() function bails out early since
pud_trans_unstable() returns 0 on arm64. So we either bracket this
function by some #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD or we
make some generic definitions for pud_* accessors when we don't have
this config.
--
Catalin
On 4/2/21 7:26 AM, Catalin Marinas wrote:
> On Wed, Mar 31, 2021 at 02:23:44PM -0400, Zack Rusin wrote:
>> The pagetable walk callbacks in mm/mapping_dirty_helpers.c depend
>> on a set of helpers from which pud_dirty(pud) was missing. I'm
>> assuming mapping_dirty_helpers weren't used on ARM64 before
>> because the missing pud_dirty is causing a compilation error.
>>
>> The drivers/gpu/drm/vmwgfx code uses mapping_dirty_helpers and
>> has been ported to ARM64 but it depends on this code getting in
>> first in order to compile/work.
>>
>> Cc: Catalin Marinas <[email protected]>
>> Cc: Will Deacon <[email protected]>
>> Cc: Peter Zijlstra <[email protected]>
>> Cc: [email protected]
>> Signed-off-by: Zack Rusin <[email protected]>
>> ---
>> arch/arm64/include/asm/pgtable.h | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
>> index 47027796c2f9..ecd80f87a996 100644
>> --- a/arch/arm64/include/asm/pgtable.h
>> +++ b/arch/arm64/include/asm/pgtable.h
>> @@ -451,6 +451,7 @@ static inline pmd_t pmd_mkdevmap(pmd_t pmd)
>> #define pfn_pmd(pfn,prot) __pmd(__phys_to_pmd_val((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot))
>> #define mk_pmd(page,prot) pfn_pmd(page_to_pfn(page),prot)
>>
>> +#define pud_dirty(pud) pte_dirty(pud_pte(pud))
>> #define pud_young(pud) pte_young(pud_pte(pud))
>> #define pud_mkyoung(pud) pte_pud(pte_mkyoung(pud_pte(pud)))
>> #define pud_write(pud) pte_write(pud_pte(pud))
>
> I think pud_dirty() should only be defined if
> CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD which is not the case on
> arm64. Well, the pud_young/mkyoung/write should probably also be removed
> until we have pud pages on arm64.
Agreed. I got confused by pud_young and co being defined here even
though they're all shielded by CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
when they're used and incorrectly assumed the practice was to define
them for arm64 here whether CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
was defined or not.
I'll go ahead and shield the pud_dirty with
CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD in mapping_dirty_helpers.c
instead.
z