Le 29/05/2024 à 11:23, Oscar Salvador a écrit :
> On Mon, May 27, 2024 at 03:30:12PM +0200, Christophe Leroy wrote:
>> On book3s/64, the only user of hugepd is hash in 4k mode.
>>
>> All other setups (hash-64, radix-4, radix-64) use leaf PMD/PUD.
>>
>> Rework hash-4k to use contiguous PMD and PUD instead.
>>
>> In that setup there are only two huge page sizes: 16M and 16G.
>>
>> 16M sits at PMD level and 16G at PUD level.
>
>
> On 4k mode, PMD_SIZE is 2MB and PUD_SIZE is 256MB, right?
Correct, as documented in arch/powerpc/include/asm/book3s/64/hash-4k.h
>
>> +static inline unsigned long hash__pte_update(struct mm_struct *mm,
>> + unsigned long addr,
>> + pte_t *ptep, unsigned long clr,
>> + unsigned long set,
>> + int huge)
>> +{
>> + unsigned long old;
>> +
>> + old = hash__pte_update_one(ptep, clr, set);
>> +
>> + if (IS_ENABLED(CONFIG_PPC_4K_PAGES) && huge) {
>> + unsigned int psize = get_slice_psize(mm, addr);
>> + int nb, i;
>> +
>> + if (psize == MMU_PAGE_16M)
>> + nb = SZ_16M / PMD_SIZE;
>> + else if (psize == MMU_PAGE_16G)
>> + nb = SZ_16G / PUD_SIZE;
>> + else
>> + nb = 1;
>
> On 4K, hugepages are either 16M or 16G. How can we end up in a situation
> whwere the is pte is huge, but is is neither MMU_PAGE_16G nor MMU_PAGE_16M?
We can't but I didn't want to leave nb undefined or with a value that
might lead to writing in the weed. Value 1 seems a safe default.
>
>> diff --git a/arch/powerpc/mm/book3s64/hugetlbpage.c b/arch/powerpc/mm/book3s64/hugetlbpage.c
>> index 5a2e512e96db..83c3361b358b 100644
>> --- a/arch/powerpc/mm/book3s64/hugetlbpage.c
>> +++ b/arch/powerpc/mm/book3s64/hugetlbpage.c
>> @@ -53,6 +53,16 @@ int __hash_page_huge(unsigned long ea, unsigned long access, unsigned long vsid,
>> /* If PTE permissions don't match, take page fault */
>> if (unlikely(!check_pte_access(access, old_pte)))
>> return 1;
>> + /*
>> + * If hash-4k, hugepages use seeral contiguous PxD entries
> 'several'
>> + * so bail out and let mm make the page young or dirty
>> + */
>> + if (IS_ENABLED(CONFIG_PPC_4K_PAGES)) {
>> + if (!(old_pte & _PAGE_ACCESSED))
>> + return 1;
>> + if ((access & _PAGE_WRITE) && !(old_pte & _PAGE_DIRTY))
>> + return 1;
>
> I have 0 clue about this code. What would happen if we do not bail out?
>
In that case the pte_xchg() in the while () will only set ACCESS or
DIRTY bit on the first PxD entry, not on all cont-PxD entries.
Christophe