Re: [PATCH RFC v2 13/20] x86: mm: carve out the generic compile-time folded pgtable case in effective_prot()

["David Hildenbrand (Arm)" <david@kernel.org>]

On 7/22/26 23:00, Yeoreum Yun wrote:
> On Wed, Jul 22, 2026 at 01:20:39PM -0700, Dave Hansen wrote:
>> On 7/22/26 10:37, Yeoreum Yun wrote:
>>> However, mm_pXd_folded() requires to mm for other architecture like
>>> s390. might save the mm instead of first_level and calls the
>>> ptdump_pt_level_first() with static inline version would make the
>>> constant comparison. So it requires to save *mm* structure in here.
>>
>> I'm having a really hard time parsing that.
>>
>> I think you're trying to say that the effective_prot_p*() functions need
>> to know the first level but they don't (today) need the mm_struct. If
>> they don't get the (calculated) first_level passed in, they need the
>> mm_struct instead.
>>
>> I think you're arguing 'pg_state' needs a ->first_level or a ->mm.
>> Having a pg_state->mm doesn't seem bad to me at all.
>>
>> But, it's also a little bit silly. It would not be rocket science to
>> pass an mm_struct down to the effective_prot_p*() functions. It makes a
>> whole lot of sense to me for a page table walking function to need
>> metadata from the mm_struct to walk correctly.
> 
> Yes. I mean to add pg_state->mm like:
> 
> diff --git a/arch/x86/mm/dump_pagetables.c b/arch/x86/mm/dump_pagetables.c
> index 2afa7a23340e..aaf00f0c6624 100644
> --- a/arch/x86/mm/dump_pagetables.c
> +++ b/arch/x86/mm/dump_pagetables.c
> @@ -38,6 +38,7 @@ struct pg_state {
>         bool check_wx;
>         unsigned long wx_pages;
>         struct seq_file *seq;
> +       struct mm_struct *mm;
>  };
> 
>  struct addr_marker {
> @@ -254,7 +255,7 @@ static void effective_prot(struct ptdump_state *pt_st, int level, u64 val)
>         pgprotval_t prot = val & PTE_FLAGS_MASK;
>         pgprotval_t effective;
> 
> -       if (level > 0) {
> +       if (level > pgtable_first_level(st->mm)) {
>                 pgprotval_t higher_prot = st->prot_levels[level - 1];
> 
>                 effective = (higher_prot & prot & (_PAGE_USER | _PAGE_RW)) |
> @@ -452,7 +453,8 @@ bool ptdump_walk_pgd_level_core(struct seq_file *m,
>                 .level = -1,
>                 .to_dmesg       = dmesg,
>                 .check_wx       = checkwx,
> -               .seq            = m
> +               .seq            = m,
> +               .mm             = mm,
>         };
> 
>         ptdump_walk_pgd(&st.ptdump, mm, pgd);
> diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
> index 8c093c119e5a..6e7d0580db99 100644
> --- a/include/linux/pgtable.h
> +++ b/include/linux/pgtable.h
> @@ -2490,4 +2490,15 @@ pgprot_t vm_get_page_prot(vm_flags_t vm_flags)                           \
>  }                                                                      \
>  EXPORT_SYMBOL(vm_get_page_prot);
> 
> +static inline int pgtable_first_level(struct mm_struct *mm)
> +{
> +       if (mm_pmd_folded(mm))
> +               return 3;
> +       if (mm_pud_folded(mm))
> +               return 2;
> +       if (mm_p4d_folded(mm))
> +               return 1;
> +       return 0;
> +}

In that case the function should probably be called

"mm_first_pgtable_level"

But now it gets confusing, because we have

enum pgtable_level {
	PGTABLE_LEVEL_PTE = 0,
	PGTABLE_LEVEL_PMD,
	PGTABLE_LEVEL_PUD,
	PGTABLE_LEVEL_P4D,
	PGTABLE_LEVEL_PGD,
};


But maybe we can make sense of it and do

/*
 * The enum values correspond to the numerical page table level,
 * starting with the highest level being level 0.
 */
enum pgtable_level {
	PGTABLE_LEVEL_PGD = 0,
	PGTABLE_LEVEL_P4D,
	PGTABLE_LEVEL_PUD,
	PGTABLE_LEVEL_PMD,
	PGTABLE_LEVEL_PTE,
};

static inline enum pgtable_level mm_first_pgtable_level(struct mm_struct *mm)
{
	if (mm_pmd_folded(mm))
		return PGTABLE_LEVEL_PMD;
	if (mm_pud_folded(mm))
		return PGTABLE_LEVEL_PUD;
	if (mm_p4d_folded(mm))
		return PGTABLE_LEVEL_P4D;
	return PGTABLE_LEVEL_PGD;
}


We could even teach effective_prot() and friends to consume enum pgtable_level
now and have it all be a bit cleaner?

-- 
Cheers,

David