Re: [PATCH] mm: Save soft-dirty bits on file pages

From: Pavel Emelyanov
Date: Mon Jul 29 2013 - 10:09:25 EST


On 07/27/2013 12:18 AM, Cyrill Gorcunov wrote:
> Andy reported that if file page get reclaimed we loose soft-dirty bit
> if it was there, so save _PAGE_BIT_SOFT_DIRTY bit when page address
> get encoded into pte entry. Thus when #pf happens on such non-present
> pte we can restore it back.
>
> Reported-by: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
> Signed-off-by: Cyrill Gorcunov <gorcunov@xxxxxxxxxx>
> Cc: Pavel Emelyanov <xemul@xxxxxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: Matt Mackall <mpm@xxxxxxxxxxx>
> Cc: Xiao Guangrong <xiaoguangrong@xxxxxxxxxxxxxxxxxx>
> Cc: Marcelo Tosatti <mtosatti@xxxxxxxxxx>
> Cc: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxx>
> Cc: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
> ---

> @@ -57,17 +57,25 @@ static int install_file_pte(struct mm_st
> unsigned long addr, unsigned long pgoff, pgprot_t prot)
> {
> int err = -ENOMEM;
> - pte_t *pte;
> + pte_t *pte, ptfile;
> spinlock_t *ptl;
>
> pte = get_locked_pte(mm, addr, &ptl);
> if (!pte)
> goto out;
>
> - if (!pte_none(*pte))
> + ptfile = pgoff_to_pte(pgoff);
> +
> + if (!pte_none(*pte)) {
> +#ifdef CONFIG_MEM_SOFT_DIRTY
> + if (pte_present(*pte) &&
> + pte_soft_dirty(*pte))

I think there's no need in wrapping every such if () inside #ifdef CONFIG_...,
since the pte_soft_dirty() routine itself would be 0 for non-soft-dirty case
and compiler would optimize this code out.

> + pte_file_mksoft_dirty(ptfile);
> +#endif
> zap_pte(mm, vma, addr, pte);
> + }
>
> - set_pte_at(mm, addr, pte, pgoff_to_pte(pgoff));
> + set_pte_at(mm, addr, pte, ptfile);
> /*
> * We don't need to run update_mmu_cache() here because the "file pte"
> * being installed by install_file_pte() is not a real pte - it's a
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/