1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Macros for manipulating and testing page->flags
9 #include <linux/types.h>
10 #include <linux/bug.h>
11 #include <linux/mmdebug.h>
12 #ifndef __GENERATING_BOUNDS_H
13 #include <linux/mm_types.h>
14 #include <generated/bounds.h>
15 #endif /* !__GENERATING_BOUNDS_H */
18 * Various page->flags bits:
20 * PG_reserved is set for special pages. The "struct page" of such a page
21 * should in general not be touched (e.g. set dirty) except by its owner.
22 * Pages marked as PG_reserved include:
23 * - Pages part of the kernel image (including vDSO) and similar (e.g. BIOS,
25 * - Pages reserved or allocated early during boot (before the page allocator
26 * was initialized). This includes (depending on the architecture) the
27 * initial vmemmap, initial page tables, crashkernel, elfcorehdr, and much
28 * much more. Once (if ever) freed, PG_reserved is cleared and they will
29 * be given to the page allocator.
30 * - Pages falling into physical memory gaps - not IORESOURCE_SYSRAM. Trying
31 * to read/write these pages might end badly. Don't touch!
33 * - Pages not added to the page allocator when onlining a section because
34 * they were excluded via the online_page_callback() or because they are
36 * - Pages allocated in the context of kexec/kdump (loaded kernel image,
37 * control pages, vmcoreinfo)
38 * - MMIO/DMA pages. Some architectures don't allow to ioremap pages that are
39 * not marked PG_reserved (as they might be in use by somebody else who does
40 * not respect the caching strategy).
41 * - Pages part of an offline section (struct pages of offline sections should
42 * not be trusted as they will be initialized when first onlined).
44 * - Pages holding CPU notes for POWER Firmware Assisted Dump
45 * - Device memory (e.g. PMEM, DAX, HMM)
46 * Some PG_reserved pages will be excluded from the hibernation image.
47 * PG_reserved does in general not hinder anybody from dumping or swapping
48 * and is no longer required for remap_pfn_range(). ioremap might require it.
49 * Consequently, PG_reserved for a page mapped into user space can indicate
50 * the zero page, the vDSO, MMIO pages or device memory.
52 * The PG_private bitflag is set on pagecache pages if they contain filesystem
53 * specific data (which is normally at page->private). It can be used by
54 * private allocations for its own usage.
56 * During initiation of disk I/O, PG_locked is set. This bit is set before I/O
57 * and cleared when writeback _starts_ or when read _completes_. PG_writeback
58 * is set before writeback starts and cleared when it finishes.
60 * PG_locked also pins a page in pagecache, and blocks truncation of the file
63 * page_waitqueue(page) is a wait queue of all tasks waiting for the page
66 * PG_swapbacked is set when a page uses swap as a backing storage. This are
67 * usually PageAnon or shmem pages but please note that even anonymous pages
68 * might lose their PG_swapbacked flag when they simply can be dropped (e.g. as
69 * a result of MADV_FREE).
71 * PG_referenced, PG_reclaim are used for page reclaim for anonymous and
72 * file-backed pagecache (see mm/vmscan.c).
74 * PG_error is set to indicate that an I/O error occurred on this page.
76 * PG_arch_1 is an architecture specific page state bit. The generic code
77 * guarantees that this bit is cleared for a page when it first is entered into
80 * PG_hwpoison indicates that a page got corrupted in hardware and contains
81 * data with incorrect ECC bits that triggered a machine check. Accessing is
82 * not safe since it may cause another machine check. Don't touch!
86 * Don't use the pageflags directly. Use the PageFoo macros.
88 * The page flags field is split into two parts, the main flags area
89 * which extends from the low bits upwards, and the fields area which
90 * extends from the high bits downwards.
92 * | FIELD | ... | FLAGS |
96 * The fields area is reserved for fields mapping zone, node (for NUMA) and
97 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like
98 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP).
101 PG_locked, /* Page is locked. Don't touch. */
108 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
111 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
114 PG_private, /* If pagecache, has fs-private data */
115 PG_private_2, /* If pagecache, has fs aux data */
116 PG_writeback, /* Page is under writeback */
117 PG_head, /* A head page */
118 PG_mappedtodisk, /* Has blocks allocated on-disk */
119 PG_reclaim, /* To be reclaimed asap */
120 PG_swapbacked, /* Page is backed by RAM/swap */
121 PG_unevictable, /* Page is "unevictable" */
123 PG_mlocked, /* Page is vma mlocked */
125 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
126 PG_uncached, /* Page has been mapped as uncached */
128 #ifdef CONFIG_MEMORY_FAILURE
129 PG_hwpoison, /* hardware poisoned page. Don't touch */
131 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
138 #ifdef CONFIG_KASAN_HW_TAGS
139 PG_skip_kasan_poison,
143 PG_readahead = PG_reclaim,
146 * Depending on the way an anonymous folio can be mapped into a page
147 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped
148 * THP), PG_anon_exclusive may be set only for the head page or for
149 * tail pages of an anonymous folio. For now, we only expect it to be
150 * set on tail pages for PTE-mapped THP.
152 PG_anon_exclusive = PG_mappedtodisk,
155 PG_checked = PG_owner_priv_1,
158 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
160 /* Two page bits are conscripted by FS-Cache to maintain local caching
161 * state. These bits are set on pages belonging to the netfs's inodes
162 * when those inodes are being locally cached.
164 PG_fscache = PG_private_2, /* page backed by cache */
167 /* Pinned in Xen as a read-only pagetable page. */
168 PG_pinned = PG_owner_priv_1,
169 /* Pinned as part of domain save (see xen_mm_pin_all()). */
170 PG_savepinned = PG_dirty,
171 /* Has a grant mapping of another (foreign) domain's page. */
172 PG_foreign = PG_owner_priv_1,
173 /* Remapped by swiotlb-xen. */
174 PG_xen_remapped = PG_owner_priv_1,
177 PG_slob_free = PG_private,
179 /* Compound pages. Stored in first tail page's flags */
180 PG_double_map = PG_workingset,
182 #ifdef CONFIG_MEMORY_FAILURE
184 * Compound pages. Stored in first tail page's flags.
185 * Indicates that at least one subpage is hwpoisoned in the
188 PG_has_hwpoisoned = PG_error,
191 /* non-lru isolated movable page */
192 PG_isolated = PG_reclaim,
194 /* Only valid for buddy pages. Used to track pages that are reported */
195 PG_reported = PG_uptodate,
197 #ifdef CONFIG_MEMORY_HOTPLUG
198 /* For self-hosted memmap pages */
199 PG_vmemmap_self_hosted = PG_owner_priv_1,
203 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
205 #ifndef __GENERATING_BOUNDS_H
207 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
208 DECLARE_STATIC_KEY_MAYBE(CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON,
209 hugetlb_optimize_vmemmap_key);
211 static __always_inline bool hugetlb_optimize_vmemmap_enabled(void)
213 return static_branch_maybe(CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON,
214 &hugetlb_optimize_vmemmap_key);
218 * If the feature of optimizing vmemmap pages associated with each HugeTLB
219 * page is enabled, the head vmemmap page frame is reused and all of the tail
220 * vmemmap addresses map to the head vmemmap page frame (furture details can
221 * refer to the figure at the head of the mm/hugetlb_vmemmap.c). In other
222 * words, there are more than one page struct with PG_head associated with each
223 * HugeTLB page. We __know__ that there is only one head page struct, the tail
224 * page structs with PG_head are fake head page structs. We need an approach
225 * to distinguish between those two different types of page structs so that
226 * compound_head() can return the real head page struct when the parameter is
227 * the tail page struct but with PG_head.
229 * The page_fixed_fake_head() returns the real head page struct if the @page is
230 * fake page head, otherwise, returns @page which can either be a true page
233 static __always_inline const struct page *page_fixed_fake_head(const struct page *page)
235 if (!hugetlb_optimize_vmemmap_enabled())
239 * Only addresses aligned with PAGE_SIZE of struct page may be fake head
240 * struct page. The alignment check aims to avoid access the fields (
241 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly)
242 * cold cacheline in some cases.
244 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) &&
245 test_bit(PG_head, &page->flags)) {
247 * We can safely access the field of the @page[1] with PG_head
248 * because the @page is a compound page composed with at least
249 * two contiguous pages.
251 unsigned long head = READ_ONCE(page[1].compound_head);
253 if (likely(head & 1))
254 return (const struct page *)(head - 1);
259 static inline const struct page *page_fixed_fake_head(const struct page *page)
264 static inline bool hugetlb_optimize_vmemmap_enabled(void)
270 static __always_inline int page_is_fake_head(struct page *page)
272 return page_fixed_fake_head(page) != page;
275 static inline unsigned long _compound_head(const struct page *page)
277 unsigned long head = READ_ONCE(page->compound_head);
279 if (unlikely(head & 1))
281 return (unsigned long)page_fixed_fake_head(page);
284 #define compound_head(page) ((typeof(page))_compound_head(page))
287 * page_folio - Converts from page to folio.
290 * Every page is part of a folio. This function cannot be called on a
293 * Context: No reference, nor lock is required on @page. If the caller
294 * does not hold a reference, this call may race with a folio split, so
295 * it should re-check the folio still contains this page after gaining
296 * a reference on the folio.
297 * Return: The folio which contains this page.
299 #define page_folio(p) (_Generic((p), \
300 const struct page *: (const struct folio *)_compound_head(p), \
301 struct page *: (struct folio *)_compound_head(p)))
304 * folio_page - Return a page from a folio.
306 * @n: The page number to return.
308 * @n is relative to the start of the folio. This function does not
309 * check that the page number lies within @folio; the caller is presumed
310 * to have a reference to the page.
312 #define folio_page(folio, n) nth_page(&(folio)->page, n)
314 static __always_inline int PageTail(struct page *page)
316 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page);
319 static __always_inline int PageCompound(struct page *page)
321 return test_bit(PG_head, &page->flags) ||
322 READ_ONCE(page->compound_head) & 1;
325 #define PAGE_POISON_PATTERN -1l
326 static inline int PagePoisoned(const struct page *page)
328 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN;
331 #ifdef CONFIG_DEBUG_VM
332 void page_init_poison(struct page *page, size_t size);
334 static inline void page_init_poison(struct page *page, size_t size)
339 static unsigned long *folio_flags(struct folio *folio, unsigned n)
341 struct page *page = &folio->page;
343 VM_BUG_ON_PGFLAGS(PageTail(page), page);
344 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
345 return &page[n].flags;
349 * Page flags policies wrt compound pages
352 * check if this struct page poisoned/uninitialized
355 * the page flag is relevant for small, head and tail pages.
358 * for compound page all operations related to the page flag applied to
362 * for compound page, callers only ever operate on the head page.
365 * modifications of the page flag must be done on small or head pages,
366 * checks can be done on tail pages too.
369 * the page flag is not relevant for compound pages.
372 * the page flag is stored in the first tail page.
374 #define PF_POISONED_CHECK(page) ({ \
375 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
377 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
378 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
379 #define PF_ONLY_HEAD(page, enforce) ({ \
380 VM_BUG_ON_PGFLAGS(PageTail(page), page); \
381 PF_POISONED_CHECK(page); })
382 #define PF_NO_TAIL(page, enforce) ({ \
383 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
384 PF_POISONED_CHECK(compound_head(page)); })
385 #define PF_NO_COMPOUND(page, enforce) ({ \
386 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
387 PF_POISONED_CHECK(page); })
388 #define PF_SECOND(page, enforce) ({ \
389 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
390 PF_POISONED_CHECK(&page[1]); })
392 /* Which page is the flag stored in */
393 #define FOLIO_PF_ANY 0
394 #define FOLIO_PF_HEAD 0
395 #define FOLIO_PF_ONLY_HEAD 0
396 #define FOLIO_PF_NO_TAIL 0
397 #define FOLIO_PF_NO_COMPOUND 0
398 #define FOLIO_PF_SECOND 1
401 * Macros to create function definitions for page flags
403 #define TESTPAGEFLAG(uname, lname, policy) \
404 static __always_inline bool folio_test_##lname(struct folio *folio) \
405 { return test_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
406 static __always_inline int Page##uname(struct page *page) \
407 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
409 #define SETPAGEFLAG(uname, lname, policy) \
410 static __always_inline \
411 void folio_set_##lname(struct folio *folio) \
412 { set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
413 static __always_inline void SetPage##uname(struct page *page) \
414 { set_bit(PG_##lname, &policy(page, 1)->flags); }
416 #define CLEARPAGEFLAG(uname, lname, policy) \
417 static __always_inline \
418 void folio_clear_##lname(struct folio *folio) \
419 { clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
420 static __always_inline void ClearPage##uname(struct page *page) \
421 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
423 #define __SETPAGEFLAG(uname, lname, policy) \
424 static __always_inline \
425 void __folio_set_##lname(struct folio *folio) \
426 { __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
427 static __always_inline void __SetPage##uname(struct page *page) \
428 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
430 #define __CLEARPAGEFLAG(uname, lname, policy) \
431 static __always_inline \
432 void __folio_clear_##lname(struct folio *folio) \
433 { __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
434 static __always_inline void __ClearPage##uname(struct page *page) \
435 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
437 #define TESTSETFLAG(uname, lname, policy) \
438 static __always_inline \
439 bool folio_test_set_##lname(struct folio *folio) \
440 { return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
441 static __always_inline int TestSetPage##uname(struct page *page) \
442 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
444 #define TESTCLEARFLAG(uname, lname, policy) \
445 static __always_inline \
446 bool folio_test_clear_##lname(struct folio *folio) \
447 { return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
448 static __always_inline int TestClearPage##uname(struct page *page) \
449 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
451 #define PAGEFLAG(uname, lname, policy) \
452 TESTPAGEFLAG(uname, lname, policy) \
453 SETPAGEFLAG(uname, lname, policy) \
454 CLEARPAGEFLAG(uname, lname, policy)
456 #define __PAGEFLAG(uname, lname, policy) \
457 TESTPAGEFLAG(uname, lname, policy) \
458 __SETPAGEFLAG(uname, lname, policy) \
459 __CLEARPAGEFLAG(uname, lname, policy)
461 #define TESTSCFLAG(uname, lname, policy) \
462 TESTSETFLAG(uname, lname, policy) \
463 TESTCLEARFLAG(uname, lname, policy)
465 #define TESTPAGEFLAG_FALSE(uname, lname) \
466 static inline bool folio_test_##lname(const struct folio *folio) { return false; } \
467 static inline int Page##uname(const struct page *page) { return 0; }
469 #define SETPAGEFLAG_NOOP(uname, lname) \
470 static inline void folio_set_##lname(struct folio *folio) { } \
471 static inline void SetPage##uname(struct page *page) { }
473 #define CLEARPAGEFLAG_NOOP(uname, lname) \
474 static inline void folio_clear_##lname(struct folio *folio) { } \
475 static inline void ClearPage##uname(struct page *page) { }
477 #define __CLEARPAGEFLAG_NOOP(uname, lname) \
478 static inline void __folio_clear_##lname(struct folio *folio) { } \
479 static inline void __ClearPage##uname(struct page *page) { }
481 #define TESTSETFLAG_FALSE(uname, lname) \
482 static inline bool folio_test_set_##lname(struct folio *folio) \
484 static inline int TestSetPage##uname(struct page *page) { return 0; }
486 #define TESTCLEARFLAG_FALSE(uname, lname) \
487 static inline bool folio_test_clear_##lname(struct folio *folio) \
489 static inline int TestClearPage##uname(struct page *page) { return 0; }
491 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \
492 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
494 #define TESTSCFLAG_FALSE(uname, lname) \
495 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
497 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
498 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD)
499 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
500 PAGEFLAG(Referenced, referenced, PF_HEAD)
501 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
502 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
503 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
504 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
505 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
506 TESTCLEARFLAG(LRU, lru, PF_HEAD)
507 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
508 TESTCLEARFLAG(Active, active, PF_HEAD)
509 PAGEFLAG(Workingset, workingset, PF_HEAD)
510 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
511 __PAGEFLAG(Slab, slab, PF_NO_TAIL)
512 __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL)
513 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
516 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
517 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
518 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
519 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
520 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
521 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
523 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
524 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
525 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
526 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
527 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
528 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
531 * Private page markings that may be used by the filesystem that owns the page
532 * for its own purposes.
533 * - PG_private and PG_private_2 cause release_folio() and co to be invoked
535 PAGEFLAG(Private, private, PF_ANY)
536 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
537 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
538 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
541 * Only test-and-set exist for PG_writeback. The unconditional operators are
542 * risky: they bypass page accounting.
544 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
545 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
546 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
548 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
549 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
550 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
551 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND)
552 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND)
554 #ifdef CONFIG_HIGHMEM
556 * Must use a macro here due to header dependency issues. page_zone() is not
557 * available at this point.
559 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
561 PAGEFLAG_FALSE(HighMem, highmem)
565 static __always_inline bool folio_test_swapcache(struct folio *folio)
567 return folio_test_swapbacked(folio) &&
568 test_bit(PG_swapcache, folio_flags(folio, 0));
571 static __always_inline bool PageSwapCache(struct page *page)
573 return folio_test_swapcache(page_folio(page));
576 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
577 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
579 PAGEFLAG_FALSE(SwapCache, swapcache)
582 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
583 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
584 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
587 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
588 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
589 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
591 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked)
592 TESTSCFLAG_FALSE(Mlocked, mlocked)
595 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
596 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
598 PAGEFLAG_FALSE(Uncached, uncached)
601 #ifdef CONFIG_MEMORY_FAILURE
602 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
603 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
604 #define __PG_HWPOISON (1UL << PG_hwpoison)
605 #define MAGIC_HWPOISON 0x48575053U /* HWPS */
606 extern void SetPageHWPoisonTakenOff(struct page *page);
607 extern void ClearPageHWPoisonTakenOff(struct page *page);
608 extern bool take_page_off_buddy(struct page *page);
609 extern bool put_page_back_buddy(struct page *page);
611 PAGEFLAG_FALSE(HWPoison, hwpoison)
612 #define __PG_HWPOISON 0
615 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
616 TESTPAGEFLAG(Young, young, PF_ANY)
617 SETPAGEFLAG(Young, young, PF_ANY)
618 TESTCLEARFLAG(Young, young, PF_ANY)
619 PAGEFLAG(Idle, idle, PF_ANY)
622 #ifdef CONFIG_KASAN_HW_TAGS
623 PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD)
625 PAGEFLAG_FALSE(SkipKASanPoison, skip_kasan_poison)
629 * PageReported() is used to track reported free pages within the Buddy
630 * allocator. We can use the non-atomic version of the test and set
631 * operations as both should be shielded with the zone lock to prevent
632 * any possible races on the setting or clearing of the bit.
634 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
636 #ifdef CONFIG_MEMORY_HOTPLUG
637 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
639 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
643 * On an anonymous page mapped into a user virtual memory area,
644 * page->mapping points to its anon_vma, not to a struct address_space;
645 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
647 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
648 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
649 * bit; and then page->mapping points, not to an anon_vma, but to a private
650 * structure which KSM associates with that merged page. See ksm.h.
652 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
653 * page and then page->mapping points to a struct movable_operations.
655 * Please note that, confusingly, "page_mapping" refers to the inode
656 * address_space which maps the page from disk; whereas "page_mapped"
657 * refers to user virtual address space into which the page is mapped.
659 #define PAGE_MAPPING_ANON 0x1
660 #define PAGE_MAPPING_MOVABLE 0x2
661 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
662 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
665 * Different with flags above, this flag is used only for fsdax mode. It
666 * indicates that this page->mapping is now under reflink case.
668 #define PAGE_MAPPING_DAX_COW 0x1
670 static __always_inline bool folio_mapping_flags(struct folio *folio)
672 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0;
675 static __always_inline int PageMappingFlags(struct page *page)
677 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
680 static __always_inline bool folio_test_anon(struct folio *folio)
682 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0;
685 static __always_inline bool PageAnon(struct page *page)
687 return folio_test_anon(page_folio(page));
690 static __always_inline bool __folio_test_movable(const struct folio *folio)
692 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
693 PAGE_MAPPING_MOVABLE;
696 static __always_inline int __PageMovable(struct page *page)
698 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
699 PAGE_MAPPING_MOVABLE;
704 * A KSM page is one of those write-protected "shared pages" or "merged pages"
705 * which KSM maps into multiple mms, wherever identical anonymous page content
706 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
707 * anon_vma, but to that page's node of the stable tree.
709 static __always_inline bool folio_test_ksm(struct folio *folio)
711 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
715 static __always_inline bool PageKsm(struct page *page)
717 return folio_test_ksm(page_folio(page));
720 TESTPAGEFLAG_FALSE(Ksm, ksm)
723 u64 stable_page_flags(struct page *page);
726 * folio_test_uptodate - Is this folio up to date?
729 * The uptodate flag is set on a folio when every byte in the folio is
730 * at least as new as the corresponding bytes on storage. Anonymous
731 * and CoW folios are always uptodate. If the folio is not uptodate,
732 * some of the bytes in it may be; see the is_partially_uptodate()
733 * address_space operation.
735 static inline bool folio_test_uptodate(struct folio *folio)
737 bool ret = test_bit(PG_uptodate, folio_flags(folio, 0));
739 * Must ensure that the data we read out of the folio is loaded
740 * _after_ we've loaded folio->flags to check the uptodate bit.
741 * We can skip the barrier if the folio is not uptodate, because
742 * we wouldn't be reading anything from it.
744 * See folio_mark_uptodate() for the other side of the story.
752 static inline int PageUptodate(struct page *page)
754 return folio_test_uptodate(page_folio(page));
757 static __always_inline void __folio_mark_uptodate(struct folio *folio)
760 __set_bit(PG_uptodate, folio_flags(folio, 0));
763 static __always_inline void folio_mark_uptodate(struct folio *folio)
766 * Memory barrier must be issued before setting the PG_uptodate bit,
767 * so that all previous stores issued in order to bring the folio
768 * uptodate are actually visible before folio_test_uptodate becomes true.
771 set_bit(PG_uptodate, folio_flags(folio, 0));
774 static __always_inline void __SetPageUptodate(struct page *page)
776 __folio_mark_uptodate((struct folio *)page);
779 static __always_inline void SetPageUptodate(struct page *page)
781 folio_mark_uptodate((struct folio *)page);
784 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
786 bool __folio_start_writeback(struct folio *folio, bool keep_write);
787 bool set_page_writeback(struct page *page);
789 #define folio_start_writeback(folio) \
790 __folio_start_writeback(folio, false)
791 #define folio_start_writeback_keepwrite(folio) \
792 __folio_start_writeback(folio, true)
794 static inline void set_page_writeback_keepwrite(struct page *page)
796 folio_start_writeback_keepwrite(page_folio(page));
799 static inline bool test_set_page_writeback(struct page *page)
801 return set_page_writeback(page);
804 static __always_inline bool folio_test_head(struct folio *folio)
806 return test_bit(PG_head, folio_flags(folio, FOLIO_PF_ANY));
809 static __always_inline int PageHead(struct page *page)
811 PF_POISONED_CHECK(page);
812 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page);
815 __SETPAGEFLAG(Head, head, PF_ANY)
816 __CLEARPAGEFLAG(Head, head, PF_ANY)
817 CLEARPAGEFLAG(Head, head, PF_ANY)
820 * folio_test_large() - Does this folio contain more than one page?
821 * @folio: The folio to test.
823 * Return: True if the folio is larger than one page.
825 static inline bool folio_test_large(struct folio *folio)
827 return folio_test_head(folio);
830 static __always_inline void set_compound_head(struct page *page, struct page *head)
832 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
835 static __always_inline void clear_compound_head(struct page *page)
837 WRITE_ONCE(page->compound_head, 0);
840 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
841 static inline void ClearPageCompound(struct page *page)
843 BUG_ON(!PageHead(page));
848 #define PG_head_mask ((1UL << PG_head))
850 #ifdef CONFIG_HUGETLB_PAGE
851 int PageHuge(struct page *page);
852 int PageHeadHuge(struct page *page);
853 static inline bool folio_test_hugetlb(struct folio *folio)
855 return PageHeadHuge(&folio->page);
858 TESTPAGEFLAG_FALSE(Huge, hugetlb)
859 TESTPAGEFLAG_FALSE(HeadHuge, headhuge)
862 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
864 * PageHuge() only returns true for hugetlbfs pages, but not for
865 * normal or transparent huge pages.
867 * PageTransHuge() returns true for both transparent huge and
868 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
869 * called only in the core VM paths where hugetlbfs pages can't exist.
871 static inline int PageTransHuge(struct page *page)
873 VM_BUG_ON_PAGE(PageTail(page), page);
874 return PageHead(page);
877 static inline bool folio_test_transhuge(struct folio *folio)
879 return folio_test_head(folio);
883 * PageTransCompound returns true for both transparent huge pages
884 * and hugetlbfs pages, so it should only be called when it's known
885 * that hugetlbfs pages aren't involved.
887 static inline int PageTransCompound(struct page *page)
889 return PageCompound(page);
893 * PageTransTail returns true for both transparent huge pages
894 * and hugetlbfs pages, so it should only be called when it's known
895 * that hugetlbfs pages aren't involved.
897 static inline int PageTransTail(struct page *page)
899 return PageTail(page);
903 * PageDoubleMap indicates that the compound page is mapped with PTEs as well
906 * This is required for optimization of rmap operations for THP: we can postpone
907 * per small page mapcount accounting (and its overhead from atomic operations)
908 * until the first PMD split.
910 * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up
911 * by one. This reference will go away with last compound_mapcount.
913 * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap().
915 PAGEFLAG(DoubleMap, double_map, PF_SECOND)
916 TESTSCFLAG(DoubleMap, double_map, PF_SECOND)
918 TESTPAGEFLAG_FALSE(TransHuge, transhuge)
919 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
920 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap)
921 TESTPAGEFLAG_FALSE(TransTail, transtail)
922 PAGEFLAG_FALSE(DoubleMap, double_map)
923 TESTSCFLAG_FALSE(DoubleMap, double_map)
926 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
928 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
931 * This flag is set by hwpoison handler. Cleared by THP split or free page.
933 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
934 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
936 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
937 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
941 * Check if a page is currently marked HWPoisoned. Note that this check is
942 * best effort only and inherently racy: there is no way to synchronize with
945 static inline bool is_page_hwpoison(struct page *page)
947 if (PageHWPoison(page))
949 return PageHuge(page) && PageHWPoison(compound_head(page));
953 * For pages that are never mapped to userspace (and aren't PageSlab),
954 * page_type may be used. Because it is initialised to -1, we invert the
955 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
956 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
957 * low bits so that an underflow or overflow of page_mapcount() won't be
958 * mistaken for a page type value.
961 #define PAGE_TYPE_BASE 0xf0000000
962 /* Reserve 0x0000007f to catch underflows of page_mapcount */
963 #define PAGE_MAPCOUNT_RESERVE -128
964 #define PG_buddy 0x00000080
965 #define PG_offline 0x00000100
966 #define PG_table 0x00000200
967 #define PG_guard 0x00000400
969 #define PageType(page, flag) \
970 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
972 static inline int page_has_type(struct page *page)
974 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE;
977 #define PAGE_TYPE_OPS(uname, lname) \
978 static __always_inline int Page##uname(struct page *page) \
980 return PageType(page, PG_##lname); \
982 static __always_inline void __SetPage##uname(struct page *page) \
984 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
985 page->page_type &= ~PG_##lname; \
987 static __always_inline void __ClearPage##uname(struct page *page) \
989 VM_BUG_ON_PAGE(!Page##uname(page), page); \
990 page->page_type |= PG_##lname; \
994 * PageBuddy() indicates that the page is free and in the buddy system
995 * (see mm/page_alloc.c).
997 PAGE_TYPE_OPS(Buddy, buddy)
1000 * PageOffline() indicates that the page is logically offline although the
1001 * containing section is online. (e.g. inflated in a balloon driver or
1002 * not onlined when onlining the section).
1003 * The content of these pages is effectively stale. Such pages should not
1004 * be touched (read/write/dump/save) except by their owner.
1006 * If a driver wants to allow to offline unmovable PageOffline() pages without
1007 * putting them back to the buddy, it can do so via the memory notifier by
1008 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
1009 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
1010 * pages (now with a reference count of zero) are treated like free pages,
1011 * allowing the containing memory block to get offlined. A driver that
1012 * relies on this feature is aware that re-onlining the memory block will
1013 * require to re-set the pages PageOffline() and not giving them to the
1014 * buddy via online_page_callback_t.
1016 * There are drivers that mark a page PageOffline() and expect there won't be
1017 * any further access to page content. PFN walkers that read content of random
1018 * pages should check PageOffline() and synchronize with such drivers using
1019 * page_offline_freeze()/page_offline_thaw().
1021 PAGE_TYPE_OPS(Offline, offline)
1023 extern void page_offline_freeze(void);
1024 extern void page_offline_thaw(void);
1025 extern void page_offline_begin(void);
1026 extern void page_offline_end(void);
1029 * Marks pages in use as page tables.
1031 PAGE_TYPE_OPS(Table, table)
1034 * Marks guardpages used with debug_pagealloc.
1036 PAGE_TYPE_OPS(Guard, guard)
1038 extern bool is_free_buddy_page(struct page *page);
1040 PAGEFLAG(Isolated, isolated, PF_ANY);
1042 static __always_inline int PageAnonExclusive(struct page *page)
1044 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1045 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1046 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1049 static __always_inline void SetPageAnonExclusive(struct page *page)
1051 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1052 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1053 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1056 static __always_inline void ClearPageAnonExclusive(struct page *page)
1058 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1059 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1060 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1063 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1065 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1066 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1067 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1071 #define __PG_MLOCKED (1UL << PG_mlocked)
1073 #define __PG_MLOCKED 0
1077 * Flags checked when a page is freed. Pages being freed should not have
1078 * these flags set. If they are, there is a problem.
1080 #define PAGE_FLAGS_CHECK_AT_FREE \
1081 (1UL << PG_lru | 1UL << PG_locked | \
1082 1UL << PG_private | 1UL << PG_private_2 | \
1083 1UL << PG_writeback | 1UL << PG_reserved | \
1084 1UL << PG_slab | 1UL << PG_active | \
1085 1UL << PG_unevictable | __PG_MLOCKED)
1088 * Flags checked when a page is prepped for return by the page allocator.
1089 * Pages being prepped should not have these flags set. If they are set,
1090 * there has been a kernel bug or struct page corruption.
1092 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1093 * alloc-free cycle to prevent from reusing the page.
1095 #define PAGE_FLAGS_CHECK_AT_PREP \
1096 (PAGEFLAGS_MASK & ~__PG_HWPOISON)
1098 #define PAGE_FLAGS_PRIVATE \
1099 (1UL << PG_private | 1UL << PG_private_2)
1101 * page_has_private - Determine if page has private stuff
1102 * @page: The page to be checked
1104 * Determine if a page has private stuff, indicating that release routines
1105 * should be invoked upon it.
1107 static inline int page_has_private(struct page *page)
1109 return !!(page->flags & PAGE_FLAGS_PRIVATE);
1112 static inline bool folio_has_private(struct folio *folio)
1114 return page_has_private(&folio->page);
1121 #undef PF_NO_COMPOUND
1123 #endif /* !__GENERATING_BOUNDS_H */
1125 #endif /* PAGE_FLAGS_H */