Hello, On Wed, Jan 25, 2017 at 09:15:19AM -0800, Shaohua Li wrote: > On Tue, Jan 24, 2017 at 11:32:12AM +0900, Minchan Kim wrote: > > Hi Shaohua, > > > > On Mon, Jan 23, 2017 at 03:15:52PM -0800, Shaohua Li wrote: > > > The page reclaim has an assumption writting to a page with clean pte > > > should trigger a page fault, because there is a window between pte zero > > > and tlb flush where a new write could come. If the new write doesn't > > > trigger page fault, page reclaim will not notice it and think the page > > > is clean and reclaim it. The MADV_FREE pages don't comply with the rule > > > and the pte is just cleaned without writeprotect, so there will be no > > > pagefault for new write. This will cause data corruption. > > > > It's hard to understand. > > Could you show me exact scenario seqence you have in mind? > Sorry for the delay, for some reason, I didn't receive the mail. > in try_to_unmap_one: > CPU 1: CPU2: > 1. pteval = ptep_get_and_clear(mm, address, pte); > 2. write to the address > 3. tlb flush > > step 1 will get a clean pteval, step2 dirty it, but the unmap missed the dirty > bit so discard the page without pageout. step2 doesn't trigger a page fault, I thought about that when Mel introduced deferred flush and concluded it should be no problem from theses discussion: 1. https://lkml.org/lkml/2015/4/15/565 2. https://lkml.org/lkml/2015/4/16/136 So, shouldn't it make trap? Ccing Mel. > because the tlb cache still has the pte entry. The defer flush makes the window > bigger actually. There are comments about this in try_to_unmap_one too. > > Thanks, > Shaohua > > > > Cc: Minchan Kim <minchan@xxxxxxxxxx> > > > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> > > > Cc: Hugh Dickins <hughd@xxxxxxxxxx> > > > Cc: Rik van Riel <riel@xxxxxxxxxxx> > > > Cc: stable@xxxxxxxxxx > > > Signed-off-by: Shaohua Li <shli@xxxxxx> > > > --- > > > mm/huge_memory.c | 1 + > > > mm/madvise.c | 1 + > > > 2 files changed, 2 insertions(+) > > > > > > diff --git a/mm/huge_memory.c b/mm/huge_memory.c > > > index 9a6bd6c..9cc5de5 100644 > > > --- a/mm/huge_memory.c > > > +++ b/mm/huge_memory.c > > > @@ -1381,6 +1381,7 @@ bool madvise_free_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma, > > > tlb->fullmm); > > > orig_pmd = pmd_mkold(orig_pmd); > > > orig_pmd = pmd_mkclean(orig_pmd); > > > + orig_pmd = pmd_wrprotect(orig_pmd); > > > > > > set_pmd_at(mm, addr, pmd, orig_pmd); > > > tlb_remove_pmd_tlb_entry(tlb, pmd, addr); > > > diff --git a/mm/madvise.c b/mm/madvise.c > > > index 0e3828e..bfb6800 100644 > > > --- a/mm/madvise.c > > > +++ b/mm/madvise.c > > > @@ -373,6 +373,7 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr, > > > > > > ptent = pte_mkold(ptent); > > > ptent = pte_mkclean(ptent); > > > + ptent = pte_wrprotect(ptent); > > > set_pte_at(mm, addr, pte, ptent); > > > if (PageActive(page)) > > > deactivate_page(page); > > > -- > > > 2.9.3 > > > > > > -- > > > To unsubscribe, send a message with 'unsubscribe linux-mm' in > > > the body to majordomo@xxxxxxxxx. For more info on Linux MM, > > > see: http://www.linux-mm.org/ . > > > Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a> > > > > -- > > To unsubscribe, send a message with 'unsubscribe linux-mm' in > > the body to majordomo@xxxxxxxxx. For more info on Linux MM, > > see: http://www.linux-mm.org/ . > > Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a> > > -- > To unsubscribe, send a message with 'unsubscribe linux-mm' in > the body to majordomo@xxxxxxxxx. For more info on Linux MM, > see: http://www.linux-mm.org/ . > Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a> -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
