On 3/21/24 15:24, Matthew Wilcox (Oracle) wrote:
> The current folio_test_hugetlb() can be fooled by a concurrent folio split
> into returning true for a folio which has never belonged to hugetlbfs.
> This can't happen if the caller holds a refcount on it, but we have a
> few places (memory-failure, compaction, procfs) which do not and should
> not take a speculative reference.
In compaction and with CONFIG_DEBUG_VM enabled, the current implementation
can result in an oops, as reported by Luis. This happens since 9c5ccf2db04b
("mm: remove HUGETLB_PAGE_DTOR") effectively added some VM_BUG_ON() checks
in the PageHuge() testing path.
> Since hugetlb pages do not use individual page mapcounts (they are always
> fully mapped and use the entire_mapcount field to record the number
> of mappings), the PageType field is available now that page_mapcount()
> ignores the value in this field.
Reported-by: Luis Chamberlain <mcgrof@xxxxxxxxxx>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=218227
Fixes: 9c5ccf2db04b ("mm: remove HUGETLB_PAGE_DTOR")
Cc: <stable@xxxxxxxxxxxxxxx>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
> Reviewed-by: David Hildenbrand <david@xxxxxxxxxx>
> ---
> include/linux/page-flags.h | 70 ++++++++++++++++------------------
> include/trace/events/mmflags.h | 1 +
> mm/hugetlb.c | 22 ++---------
> 3 files changed, 37 insertions(+), 56 deletions(-)
>
> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> index 5852f967c640..6fb3cd42ee59 100644
> --- a/include/linux/page-flags.h
> +++ b/include/linux/page-flags.h
> @@ -190,7 +190,6 @@ enum pageflags {
>
> /* At least one page in this folio has the hwpoison flag set */
> PG_has_hwpoisoned = PG_error,
> - PG_hugetlb = PG_active,
> PG_large_rmappable = PG_workingset, /* anon or file-backed */
> };
>
> @@ -876,29 +875,6 @@ FOLIO_FLAG_FALSE(large_rmappable)
>
> #define PG_head_mask ((1UL << PG_head))
>
> -#ifdef CONFIG_HUGETLB_PAGE
> -int PageHuge(const struct page *page);
> -SETPAGEFLAG(HugeTLB, hugetlb, PF_SECOND)
> -CLEARPAGEFLAG(HugeTLB, hugetlb, PF_SECOND)
> -
> -/**
> - * folio_test_hugetlb - Determine if the folio belongs to hugetlbfs
> - * @folio: The folio to test.
> - *
> - * Context: Any context. Caller should have a reference on the folio to
> - * prevent it from being turned into a tail page.
> - * Return: True for hugetlbfs folios, false for anon folios or folios
> - * belonging to other filesystems.
> - */
> -static inline bool folio_test_hugetlb(const struct folio *folio)
> -{
> - return folio_test_large(folio) &&
> - test_bit(PG_hugetlb, const_folio_flags(folio, 1));
> -}
> -#else
> -TESTPAGEFLAG_FALSE(Huge, hugetlb)
> -#endif
> -
> #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> /*
> * PageHuge() only returns true for hugetlbfs pages, but not for
> @@ -954,18 +930,6 @@ PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
> TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
> #endif
>
> -/*
> - * Check if a page is currently marked HWPoisoned. Note that this check is
> - * best effort only and inherently racy: there is no way to synchronize with
> - * failing hardware.
> - */
> -static inline bool is_page_hwpoison(struct page *page)
> -{
> - if (PageHWPoison(page))
> - return true;
> - return PageHuge(page) && PageHWPoison(compound_head(page));
> -}
> -
> /*
> * For pages that are never mapped to userspace (and aren't PageSlab),
> * page_type may be used. Because it is initialised to -1, we invert the
> @@ -982,6 +946,7 @@ static inline bool is_page_hwpoison(struct page *page)
> #define PG_offline 0x00000100
> #define PG_table 0x00000200
> #define PG_guard 0x00000400
> +#define PG_hugetlb 0x00000800
>
> #define PageType(page, flag) \
> ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
> @@ -1076,6 +1041,37 @@ PAGE_TYPE_OPS(Table, table, pgtable)
> */
> PAGE_TYPE_OPS(Guard, guard, guard)
>
> +#ifdef CONFIG_HUGETLB_PAGE
> +FOLIO_TYPE_OPS(hugetlb, hugetlb)
> +#else
> +FOLIO_TEST_FLAG_FALSE(hugetlb)
> +#endif
> +
> +/**
> + * PageHuge - Determine if the page belongs to hugetlbfs
> + * @page: The page to test.
> + *
> + * Context: Any context.
> + * Return: True for hugetlbfs pages, false for anon pages or pages
> + * belonging to other filesystems.
> + */
> +static inline bool PageHuge(const struct page *page)
> +{
> + return folio_test_hugetlb(page_folio(page));
> +}
> +
> +/*
> + * Check if a page is currently marked HWPoisoned. Note that this check is
> + * best effort only and inherently racy: there is no way to synchronize with
> + * failing hardware.
> + */
> +static inline bool is_page_hwpoison(struct page *page)
> +{
> + if (PageHWPoison(page))
> + return true;
> + return PageHuge(page) && PageHWPoison(compound_head(page));
> +}
> +
> extern bool is_free_buddy_page(struct page *page);
>
> PAGEFLAG(Isolated, isolated, PF_ANY);
> @@ -1142,7 +1138,7 @@ static __always_inline void __ClearPageAnonExclusive(struct page *page)
> */
> #define PAGE_FLAGS_SECOND \
> (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \
> - 1UL << PG_hugetlb | 1UL << PG_large_rmappable)
> + 1UL << PG_large_rmappable)
>
> #define PAGE_FLAGS_PRIVATE \
> (1UL << PG_private | 1UL << PG_private_2)
> diff --git a/include/trace/events/mmflags.h b/include/trace/events/mmflags.h
> index d801409b33cf..d55e53ac91bd 100644
> --- a/include/trace/events/mmflags.h
> +++ b/include/trace/events/mmflags.h
> @@ -135,6 +135,7 @@ IF_HAVE_PG_ARCH_X(arch_3)
> #define DEF_PAGETYPE_NAME(_name) { PG_##_name, __stringify(_name) }
>
> #define __def_pagetype_names \
> + DEF_PAGETYPE_NAME(hugetlb), \
> DEF_PAGETYPE_NAME(offline), \
> DEF_PAGETYPE_NAME(guard), \
> DEF_PAGETYPE_NAME(table), \
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 7e9a766059aa..bdcbb62096cf 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -1624,7 +1624,7 @@ static inline void __clear_hugetlb_destructor(struct hstate *h,
> {
> lockdep_assert_held(&hugetlb_lock);
>
> - folio_clear_hugetlb(folio);
> + __folio_clear_hugetlb(folio);
> }
>
> /*
> @@ -1711,7 +1711,7 @@ static void add_hugetlb_folio(struct hstate *h, struct folio *folio,
> h->surplus_huge_pages_node[nid]++;
> }
>
> - folio_set_hugetlb(folio);
> + __folio_set_hugetlb(folio);
> folio_change_private(folio, NULL);
> /*
> * We have to set hugetlb_vmemmap_optimized again as above
> @@ -2050,7 +2050,7 @@ static void __prep_account_new_huge_page(struct hstate *h, int nid)
>
> static void init_new_hugetlb_folio(struct hstate *h, struct folio *folio)
> {
> - folio_set_hugetlb(folio);
> + __folio_set_hugetlb(folio);
> INIT_LIST_HEAD(&folio->lru);
> hugetlb_set_folio_subpool(folio, NULL);
> set_hugetlb_cgroup(folio, NULL);
> @@ -2160,22 +2160,6 @@ static bool prep_compound_gigantic_folio_for_demote(struct folio *folio,
> return __prep_compound_gigantic_folio(folio, order, true);
> }
>
> -/*
> - * PageHuge() only returns true for hugetlbfs pages, but not for normal or
> - * transparent huge pages. See the PageTransHuge() documentation for more
> - * details.
> - */
> -int PageHuge(const struct page *page)
> -{
> - const struct folio *folio;
> -
> - if (!PageCompound(page))
> - return 0;
> - folio = page_folio(page);
> - return folio_test_hugetlb(folio);
> -}
> -EXPORT_SYMBOL_GPL(PageHuge);
> -
> /*
> * Find and lock address space (mapping) in write mode.
> *
