The patch titled ksm: share anon page without allocating has been added to the -mm tree. Its filename is ksm-share-anon-page-without-allocating.patch Before you just go and hit "reply", please: a) Consider who else should be cc'ed b) Prefer to cc a suitable mailing list as well c) Ideally: find the original patch on the mailing list and do a reply-to-all to that, adding suitable additional cc's *** Remember to use Documentation/SubmitChecklist when testing your code *** See http://userweb.kernel.org/~akpm/stuff/added-to-mm.txt to find out what to do about this The current -mm tree may be found at http://userweb.kernel.org/~akpm/mmotm/ ------------------------------------------------------ Subject: ksm: share anon page without allocating From: Hugh Dickins <hugh.dickins@xxxxxxxxxxxxx> When ksm pages were unswappable, it made no sense to include them in mem cgroup accounting; but now that they are swappable (although I see no strict logical connection) the principle of least surprise implies that they should be accounted (with the usual dissatisfaction, that a shared page is accounted to only one of the cgroups using it). This patch was intended to add mem cgroup accounting where necessary; but turned inside out, it now avoids allocating a ksm page, instead upgrading an anon page to ksm - which brings its existing mem cgroup accounting with it. Thus mem cgroups don't appear in the patch at all. This upgrade from PageAnon to PageKsm takes place under page lock (via a somewhat hacky NULL kpage interface), and audit showed only one place which needed to cope with the race - page_referenced() is sometimes used without page lock, so page_lock_anon_vma() needs an ACCESS_ONCE() to be sure of getting anon_vma and flags together (no problem if the page goes ksm an instant after, the integrity of that anon_vma list is unaffected). Signed-off-by: Hugh Dickins <hugh.dickins@xxxxxxxxxxxxx> Cc: Izik Eidus <ieidus@xxxxxxxxxx> Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx> Cc: Chris Wright <chrisw@xxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/ksm.c | 67 ++++++++++++++++------------------------------------ mm/rmap.c | 6 +++- 2 files changed, 25 insertions(+), 48 deletions(-) diff -puN mm/ksm.c~ksm-share-anon-page-without-allocating mm/ksm.c --- a/mm/ksm.c~ksm-share-anon-page-without-allocating +++ a/mm/ksm.c @@ -831,7 +831,8 @@ out: * try_to_merge_one_page - take two pages and merge them into one * @vma: the vma that holds the pte pointing to page * @page: the PageAnon page that we want to replace with kpage - * @kpage: the PageKsm page that we want to map instead of page + * @kpage: the PageKsm page that we want to map instead of page, + * or NULL the first time when we want to use page as kpage. * * This function returns 0 if the pages were merged, -EFAULT otherwise. */ @@ -864,15 +865,24 @@ static int try_to_merge_one_page(struct * ptes are necessarily already write-protected. But in either * case, we need to lock and check page_count is not raised. */ - if (write_protect_page(vma, page, &orig_pte) == 0 && - pages_identical(page, kpage)) - err = replace_page(vma, page, kpage, orig_pte); + if (write_protect_page(vma, page, &orig_pte) == 0) { + if (!kpage) { + /* + * While we hold page lock, upgrade page from + * PageAnon+anon_vma to PageKsm+NULL stable_node: + * stable_tree_insert() will update stable_node. + */ + set_page_stable_node(page, NULL); + mark_page_accessed(page); + err = 0; + } else if (pages_identical(page, kpage)) + err = replace_page(vma, page, kpage, orig_pte); + } - if ((vma->vm_flags & VM_LOCKED) && !err) { + if ((vma->vm_flags & VM_LOCKED) && kpage && !err) { munlock_vma_page(page); if (!PageMlocked(kpage)) { unlock_page(page); - lru_add_drain(); lock_page(kpage); mlock_vma_page(kpage); page = kpage; /* for final unlock */ @@ -922,7 +932,7 @@ out: * This function returns the kpage if we successfully merged two identical * pages into one ksm page, NULL otherwise. * - * Note that this function allocates a new kernel page: if one of the pages + * Note that this function upgrades page to ksm page: if one of the pages * is already a ksm page, try_to_merge_with_ksm_page should be used. */ static struct page *try_to_merge_two_pages(struct rmap_item *rmap_item, @@ -930,10 +940,7 @@ static struct page *try_to_merge_two_pag struct rmap_item *tree_rmap_item, struct page *tree_page) { - struct mm_struct *mm = rmap_item->mm; - struct vm_area_struct *vma; - struct page *kpage; - int err = -EFAULT; + int err; /* * The number of nodes in the stable tree @@ -943,37 +950,10 @@ static struct page *try_to_merge_two_pag ksm_max_kernel_pages <= ksm_pages_shared) return NULL; - kpage = alloc_page(GFP_HIGHUSER); - if (!kpage) - return NULL; - - down_read(&mm->mmap_sem); - if (ksm_test_exit(mm)) - goto up; - vma = find_vma(mm, rmap_item->address); - if (!vma || vma->vm_start > rmap_item->address) - goto up; - - copy_user_highpage(kpage, page, rmap_item->address, vma); - - SetPageDirty(kpage); - __SetPageUptodate(kpage); - SetPageSwapBacked(kpage); - set_page_stable_node(kpage, NULL); /* mark it PageKsm */ - lru_cache_add_lru(kpage, LRU_ACTIVE_ANON); - - err = try_to_merge_one_page(vma, page, kpage); - if (err) - goto up; - - /* Must get reference to anon_vma while still holding mmap_sem */ - hold_anon_vma(rmap_item, vma->anon_vma); -up: - up_read(&mm->mmap_sem); - + err = try_to_merge_with_ksm_page(rmap_item, page, NULL); if (!err) { err = try_to_merge_with_ksm_page(tree_rmap_item, - tree_page, kpage); + tree_page, page); /* * If that fails, we have a ksm page with only one pte * pointing to it: so break it. @@ -981,11 +961,7 @@ up: if (err) break_cow(rmap_item); } - if (err) { - put_page(kpage); - kpage = NULL; - } - return kpage; + return err ? NULL : page; } /* @@ -1244,7 +1220,6 @@ static void cmp_and_merge_page(struct pa stable_tree_append(rmap_item, stable_node); } unlock_page(kpage); - put_page(kpage); /* * If we fail to insert the page into the stable tree, diff -puN mm/rmap.c~ksm-share-anon-page-without-allocating mm/rmap.c --- a/mm/rmap.c~ksm-share-anon-page-without-allocating +++ a/mm/rmap.c @@ -204,7 +204,7 @@ struct anon_vma *page_lock_anon_vma(stru unsigned long anon_mapping; rcu_read_lock(); - anon_mapping = (unsigned long) page->mapping; + anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping); if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON) goto out; if (!page_mapped(page)) @@ -666,7 +666,9 @@ static void __page_check_anon_rmap(struc * @address: the user virtual address mapped * * The caller needs to hold the pte lock, and the page must be locked in - * the anon_vma case: to serialize mapping,index checking after setting. + * the anon_vma case: to serialize mapping,index checking after setting, + * and to ensure that PageAnon is not being upgraded racily to PageKsm + * (but PageKsm is never downgraded to PageAnon). */ void page_add_anon_rmap(struct page *page, struct vm_area_struct *vma, unsigned long address) _ Patches currently in -mm which might be from hugh.dickins@xxxxxxxxxxxxx are mmap-dont-return-enomem-when-mapcount-is-temporarily-exceeded-in-munmap.patch mmap-dont-return-enomem-when-mapcount-is-temporarily-exceeded-in-munmap-checkpatch-fixes.patch vmalloc-adjust-gfp-mask-passed-on-nested-vmalloc-invocation.patch swap_info-private-to-swapfilec.patch swap_info-change-to-array-of-pointers.patch swap_info-include-first_swap_extent.patch swap_info-include-first_swap_extent-fix.patch swap_info-include-first_swap_extent-fix-fix.patch swap_info-miscellaneous-minor-cleanups.patch swap_info-swap_has_cache-cleanups.patch swap_info-swap_map-of-chars-not-shorts.patch swap_info-swap-count-continuations.patch swap_info-note-swap_map_shmem.patch swap_info-reorder-its-fields.patch rmap-fix-the-comment-for-try_to_unmap_anon.patch oom_kill-use-rss-value-instead-of-vm-size-for-badness.patch mm-define-page_mapping_flags.patch mm-mlocking-in-try_to_unmap_one.patch mm-mlocking-in-try_to_unmap_one-fix.patch mm-mlocking-in-try_to_unmap_one-fix-fix.patch mm-config_mmu-for-pg_mlocked.patch mm-pass-address-down-to-rmap-ones.patch mm-stop-ptlock-enlarging-struct-page.patch mm-sigbus-instead-of-abusing-oom.patch ksm-three-remove_rmap_item_from_tree-cleanups.patch ksm-remove-redundancies-when-merging-page.patch ksm-cleanup-some-function-arguments.patch ksm-singly-linked-rmap_list.patch ksm-separate-stable_node.patch ksm-stable_node-point-to-page-and-back.patch ksm-fix-mlockfreed-to-munlocked.patch ksm-let-shared-pages-be-swappable.patch ksm-hold-anon_vma-in-rmap_item.patch ksm-take-keyhole-reference-to-page.patch ksm-share-anon-page-without-allocating.patch ksm-mem-cgroup-charge-swapin-copy.patch ksm-rmap_walk-to-remove_migation_ptes.patch ksm-memory-hotremove-migration-only.patch ksm-remove-unswappable-max_kernel_pages.patch hugetlb-prevent-deadlock-in-__unmap_hugepage_range-when-alloc_huge_page-fails-2.patch mm-simplify-try_to_unmap_one.patch mm-simplify-try_to_unmap_one-fix.patch elf-kill-use_elf_core_dump.patch prio_tree-debugging-patch.patch -- To unsubscribe from this list: send the line "unsubscribe mm-commits" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html