On Fri, Jan 31, 2020 at 10:00:52AM +0000, Thomas Hellström (VMware) wrote:
> From: Thomas Hellstrom <thellstrom@xxxxxxxxxx>
>
> Following the update of pagewalk code
> commit a07984d48146 ("mm: pagewalk: add p4d_entry() and pgd_entry()")
> we can modify the mapping_dirty_helpers' huge page-table entry callbacks
> to avoid splitting when a huge pud or -pmd is encountered.
>
> Signed-off-by: Thomas Hellstrom <thellstrom@xxxxxxxxxx>
> Cc: Steven Price <steven.price@xxxxxxx>
LGTM
Reviewed-by: Steven Price <steven.price@xxxxxxx>
> ---
> mm/mapping_dirty_helpers.c | 42 ++++++++++++++++++++++++++++++++++----
> 1 file changed, 38 insertions(+), 4 deletions(-)
>
> diff --git a/mm/mapping_dirty_helpers.c b/mm/mapping_dirty_helpers.c
> index 71070dda9643..2c7d03675903 100644
> --- a/mm/mapping_dirty_helpers.c
> +++ b/mm/mapping_dirty_helpers.c
> @@ -111,26 +111,60 @@ static int clean_record_pte(pte_t *pte, unsigned long addr,
> return 0;
> }
>
> -/* wp_clean_pmd_entry - The pagewalk pmd callback. */
> +/*
> + * wp_clean_pmd_entry - The pagewalk pmd callback.
> + *
> + * Dirty-tracking should take place on the PTE level, so
> + * WARN() if encountering a dirty huge pmd.
> + * Furthermore, never split huge pmds, since that currently
> + * causes dirty info loss. The pagefault handler should do
> + * that if needed.
> + */
> static int wp_clean_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long end,
> struct mm_walk *walk)
> {
> - /* Dirty-tracking should be handled on the pte level */
> pmd_t pmdval = pmd_read_atomic(pmd);
>
> + if (!pmd_trans_unstable(&pmdval))
> + return 0;
> +
> + if (pmd_none(pmdval)) {
> + walk->action = ACTION_AGAIN;
> + return 0;
> + }
> +
> + /* Huge pmd, present or migrated */
> + walk->action = ACTION_CONTINUE;
> if (pmd_trans_huge(pmdval) || pmd_devmap(pmdval))
> WARN_ON(pmd_write(pmdval) || pmd_dirty(pmdval));
>
> return 0;
> }
>
> -/* wp_clean_pud_entry - The pagewalk pud callback. */
> +/*
> + * wp_clean_pud_entry - The pagewalk pud callback.
> + *
> + * Dirty-tracking should take place on the PTE level, so
> + * WARN() if encountering a dirty huge puds.
> + * Furthermore, never split huge puds, since that currently
> + * causes dirty info loss. The pagefault handler should do
> + * that if needed.
> + */
> static int wp_clean_pud_entry(pud_t *pud, unsigned long addr, unsigned long end,
> struct mm_walk *walk)
> {
> - /* Dirty-tracking should be handled on the pte level */
> pud_t pudval = READ_ONCE(*pud);
>
> + if (!pud_trans_unstable(&pudval))
> + return 0;
> +
> + if (pud_none(pudval)) {
> + walk->action = ACTION_AGAIN;
> + return 0;
> + }
> +
> + /* Huge pud */
> + walk->action = ACTION_CONTINUE;
> if (pud_trans_huge(pudval) || pud_devmap(pudval))
> WARN_ON(pud_write(pudval) || pud_dirty(pudval));
>
> --
> 2.21.1
>
