Kirill A. Shutemov wrote:
> Something like this?
>
>
> From 251e124630da82482e8b320c73162ce89af04d5d Mon Sep 17 00:00:00 2001
> From: "Kirill A. Shutemov" <kirill.shute...@linux.intel.com>
> Date: Thu, 18 Jan 2018 18:24:07 +0300
> Subject: [PATCH] mm, page_vma_mapped: Fix pointer arithmetics in check_pte()
>
> Tetsuo reported random crashes under memory pressure on 32-bit x86
> system and tracked down to change that introduced
> page_vma_mapped_walk().
>
> The root cause of the issue is the faulty pointer math in check_pte().
> As ->pte may point to an arbitrary page we have to check that they are
> belong to the section before doing math. Otherwise it may lead to weird
> results.
>
> It wasn't noticed until now as mem_map[] is virtually contiguous on flatmem or
> vmemmap sparsemem. Pointer arithmetic just works against all 'struct page'
> pointers. But with classic sparsemem, it doesn't.
>
> Let's restructure code a bit and add necessary check.
>
> Signed-off-by: Kirill A. Shutemov <kirill.shute...@linux.intel.com>
> Reported-by: Tetsuo Handa <penguin-ker...@i-love.sakura.ne.jp>
> Fixes: ace71a19cec5 ("mm: introduce page_vma_mapped_walk()")
> Cc: sta...@vger.kernel.org
This patch solves the problem. Thank you.
Tested-by: Tetsuo Handa <penguin-ker...@i-love.sakura.ne.jp>
> ---
> mm/page_vma_mapped.c | 66
> +++++++++++++++++++++++++++++++++++-----------------
> 1 file changed, 45 insertions(+), 21 deletions(-)
>
> diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
> index d22b84310f6d..de195dcdfbd8 100644
> --- a/mm/page_vma_mapped.c
> +++ b/mm/page_vma_mapped.c
> @@ -30,8 +30,28 @@ static bool map_pte(struct page_vma_mapped_walk *pvmw)
> return true;
> }
>
> +/**
> + * check_pte - check if @pvmw->page is mapped at the @pvmw->pte
> + *
> + * page_vma_mapped_walk() found a place where @pvmw->page is *potentially*
> + * mapped. check_pte() has to validate this.
> + *
> + * @pvmw->pte may point to empty PTE, swap PTE or PTE pointing to arbitrary
> + * page.
> + *
> + * If PVMW_MIGRATION flag is set, returns true if @pvmw->pte contains
> migration
> + * entry that points to @pvmw->page or any subpage in case of THP.
> + *
> + * If PVMW_MIGRATION flag is not set, returns true if @pvmw->pte points to
> + * @pvmw->page or any subpage in case of THP.
> + *
> + * Otherwise, return false.
> + *
> + */
> static bool check_pte(struct page_vma_mapped_walk *pvmw)
> {
> + struct page *page;
> +
> if (pvmw->flags & PVMW_MIGRATION) {
> #ifdef CONFIG_MIGRATION
> swp_entry_t entry;
> @@ -41,37 +61,41 @@ static bool check_pte(struct page_vma_mapped_walk *pvmw)
>
> if (!is_migration_entry(entry))
> return false;
> - if (migration_entry_to_page(entry) - pvmw->page >=
> - hpage_nr_pages(pvmw->page)) {
> - return false;
> - }
> - if (migration_entry_to_page(entry) < pvmw->page)
> - return false;
> +
> + page = migration_entry_to_page(entry);
> #else
> WARN_ON_ONCE(1);
> #endif
> - } else {
> - if (is_swap_pte(*pvmw->pte)) {
> - swp_entry_t entry;
> + } else if (is_swap_pte(*pvmw->pte)) {
> + swp_entry_t entry;
>
> - entry = pte_to_swp_entry(*pvmw->pte);
> - if (is_device_private_entry(entry) &&
> - device_private_entry_to_page(entry) == pvmw->page)
> - return true;
> - }
> + /* Handle un-addressable ZONE_DEVICE memory */
> + entry = pte_to_swp_entry(*pvmw->pte);
> + if (!is_device_private_entry(entry))
> + return false;
>
> + page = device_private_entry_to_page(entry);
> + } else {
> if (!pte_present(*pvmw->pte))
> return false;
>
> - /* THP can be referenced by any subpage */
> - if (pte_page(*pvmw->pte) - pvmw->page >=
> - hpage_nr_pages(pvmw->page)) {
> - return false;
> - }
> - if (pte_page(*pvmw->pte) < pvmw->page)
> - return false;
> + page = pte_page(*pvmw->pte);
> }
>
> + /*
> + * Make sure that pages are in the same section before doing pointer
> + * arithmetics.
> + */
> + if (page_to_section(pvmw->page) != page_to_section(page))
> + return false;
> +
> + if (page < pvmw->page)
> + return false;
> +
> + /* THP can be referenced by any subpage */
> + if (page - pvmw->page >= hpage_nr_pages(pvmw->page))
> + return false;
> +
> return true;
> }
