On Fri, 20 Jul 2018 09:44:48 +0100
Janosch Frank <frankja@xxxxxxxxxxxxx> wrote:
> Similarly to the pte skey handling, where we set the storage key to
> the default key for each newly mapped pte, we have to also do that for
> huge pmds.
>
> With the PG_arch_1 flag we keep track if the area has already been
> cleared of its skeys.
>
> Signed-off-by: Janosch Frank <frankja@xxxxxxxxxxxxx>
> ---
> arch/s390/mm/gmap.c | 2 ++
> arch/s390/mm/hugetlbpage.c | 24 ++++++++++++++++++++++++
> 2 files changed, 26 insertions(+)
>
> diff --git a/arch/s390/mm/gmap.c b/arch/s390/mm/gmap.c
> index 6194a8a62aa0..409bc8f9d5b6 100644
> --- a/arch/s390/mm/gmap.c
> +++ b/arch/s390/mm/gmap.c
> @@ -2554,6 +2554,7 @@ static int __s390_enable_skey_hugetlb(pte_t *pte, unsigned long addr,
> {
> pmd_t *pmd = (pmd_t *)pte;
> unsigned long start, end;
> + struct page *page = pmd_page(*pmd);
>
> /*
> * The write check makes sure we do not set a key on shared
> @@ -2568,6 +2569,7 @@ static int __s390_enable_skey_hugetlb(pte_t *pte, unsigned long addr,
> start = pmd_val(*pmd) & HPAGE_MASK;
> end = start + HPAGE_SIZE - 1;
> __storage_key_init_range(start, end);
> + set_bit(PG_arch_1, &page->flags);
> return 0;
> }
>
> diff --git a/arch/s390/mm/hugetlbpage.c b/arch/s390/mm/hugetlbpage.c
> index e804090f4470..b0246c705a19 100644
> --- a/arch/s390/mm/hugetlbpage.c
> +++ b/arch/s390/mm/hugetlbpage.c
> @@ -123,6 +123,29 @@ static inline pte_t __rste_to_pte(unsigned long rste)
> return pte;
> }
>
> +static void clear_huge_pte_skeys(struct mm_struct *mm, unsigned long rste)
> +{
> + struct page *page;
> + unsigned long size, paddr;
> +
> + if (!mm_uses_skeys(mm) ||
> + rste & _SEGMENT_ENTRY_INVALID)
> + return;
> +
> + if ((rste & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3) {
> + page = pud_page(__pud(rste));
> + size = PUD_SIZE;
> + paddr = rste & PUD_MASK;
> + } else {
> + page = pmd_page(__pmd(rste));
> + size = PMD_SIZE;
> + paddr = rste & PMD_MASK;
> + }
> +
> + if (!test_and_set_bit(PG_arch_1, &page->flags))
> + __storage_key_init_range(paddr, paddr + size - 1);
> +}
> +
> void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
> pte_t *ptep, pte_t pte)
> {
> @@ -137,6 +160,7 @@ void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
> rste |= _REGION_ENTRY_TYPE_R3 | _REGION3_ENTRY_LARGE;
> else
> rste |= _SEGMENT_ENTRY_LARGE;
> + clear_huge_pte_skeys(mm, rste);
> pte_val(*ptep) = rste;
> }
>
That looks promising but I miss the hunk for arch_clear_hugepage_flags() to reset
the PG_arch_1 bit again when the huge page is freed.
--
blue skies,
Martin.
"Reality continues to ruin my life." - Calvin.
