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)
135 #ifdef CONFIG_ARCH_USES_PG_ARCH_X
141 PG_readahead = PG_reclaim,
144 * Depending on the way an anonymous folio can be mapped into a page
145 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped
146 * THP), PG_anon_exclusive may be set only for the head page or for
147 * tail pages of an anonymous folio. For now, we only expect it to be
148 * set on tail pages for PTE-mapped THP.
150 PG_anon_exclusive = PG_mappedtodisk,
153 PG_checked = PG_owner_priv_1,
156 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
158 /* Two page bits are conscripted by FS-Cache to maintain local caching
159 * state. These bits are set on pages belonging to the netfs's inodes
160 * when those inodes are being locally cached.
162 PG_fscache = PG_private_2, /* page backed by cache */
165 /* Pinned in Xen as a read-only pagetable page. */
166 PG_pinned = PG_owner_priv_1,
167 /* Pinned as part of domain save (see xen_mm_pin_all()). */
168 PG_savepinned = PG_dirty,
169 /* Has a grant mapping of another (foreign) domain's page. */
170 PG_foreign = PG_owner_priv_1,
171 /* Remapped by swiotlb-xen. */
172 PG_xen_remapped = PG_owner_priv_1,
174 #ifdef CONFIG_MEMORY_FAILURE
176 * Compound pages. Stored in first tail page's flags.
177 * Indicates that at least one subpage is hwpoisoned in the
180 PG_has_hwpoisoned = PG_error,
183 /* non-lru isolated movable page */
184 PG_isolated = PG_reclaim,
186 /* Only valid for buddy pages. Used to track pages that are reported */
187 PG_reported = PG_uptodate,
189 #ifdef CONFIG_MEMORY_HOTPLUG
190 /* For self-hosted memmap pages */
191 PG_vmemmap_self_hosted = PG_owner_priv_1,
195 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
197 #ifndef __GENERATING_BOUNDS_H
199 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
200 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key);
203 * Return the real head page struct iff the @page is a fake head page, otherwise
204 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst.
206 static __always_inline const struct page *page_fixed_fake_head(const struct page *page)
208 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key))
212 * Only addresses aligned with PAGE_SIZE of struct page may be fake head
213 * struct page. The alignment check aims to avoid access the fields (
214 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly)
215 * cold cacheline in some cases.
217 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) &&
218 test_bit(PG_head, &page->flags)) {
220 * We can safely access the field of the @page[1] with PG_head
221 * because the @page is a compound page composed with at least
222 * two contiguous pages.
224 unsigned long head = READ_ONCE(page[1].compound_head);
226 if (likely(head & 1))
227 return (const struct page *)(head - 1);
232 static inline const struct page *page_fixed_fake_head(const struct page *page)
238 static __always_inline int page_is_fake_head(struct page *page)
240 return page_fixed_fake_head(page) != page;
243 static inline unsigned long _compound_head(const struct page *page)
245 unsigned long head = READ_ONCE(page->compound_head);
247 if (unlikely(head & 1))
249 return (unsigned long)page_fixed_fake_head(page);
252 #define compound_head(page) ((typeof(page))_compound_head(page))
255 * page_folio - Converts from page to folio.
258 * Every page is part of a folio. This function cannot be called on a
261 * Context: No reference, nor lock is required on @page. If the caller
262 * does not hold a reference, this call may race with a folio split, so
263 * it should re-check the folio still contains this page after gaining
264 * a reference on the folio.
265 * Return: The folio which contains this page.
267 #define page_folio(p) (_Generic((p), \
268 const struct page *: (const struct folio *)_compound_head(p), \
269 struct page *: (struct folio *)_compound_head(p)))
272 * folio_page - Return a page from a folio.
274 * @n: The page number to return.
276 * @n is relative to the start of the folio. This function does not
277 * check that the page number lies within @folio; the caller is presumed
278 * to have a reference to the page.
280 #define folio_page(folio, n) nth_page(&(folio)->page, n)
282 static __always_inline int PageTail(struct page *page)
284 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page);
287 static __always_inline int PageCompound(struct page *page)
289 return test_bit(PG_head, &page->flags) ||
290 READ_ONCE(page->compound_head) & 1;
293 #define PAGE_POISON_PATTERN -1l
294 static inline int PagePoisoned(const struct page *page)
296 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN;
299 #ifdef CONFIG_DEBUG_VM
300 void page_init_poison(struct page *page, size_t size);
302 static inline void page_init_poison(struct page *page, size_t size)
307 static unsigned long *folio_flags(struct folio *folio, unsigned n)
309 struct page *page = &folio->page;
311 VM_BUG_ON_PGFLAGS(PageTail(page), page);
312 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
313 return &page[n].flags;
317 * Page flags policies wrt compound pages
320 * check if this struct page poisoned/uninitialized
323 * the page flag is relevant for small, head and tail pages.
326 * for compound page all operations related to the page flag applied to
330 * for compound page, callers only ever operate on the head page.
333 * modifications of the page flag must be done on small or head pages,
334 * checks can be done on tail pages too.
337 * the page flag is not relevant for compound pages.
340 * the page flag is stored in the first tail page.
342 #define PF_POISONED_CHECK(page) ({ \
343 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
345 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
346 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
347 #define PF_ONLY_HEAD(page, enforce) ({ \
348 VM_BUG_ON_PGFLAGS(PageTail(page), page); \
349 PF_POISONED_CHECK(page); })
350 #define PF_NO_TAIL(page, enforce) ({ \
351 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
352 PF_POISONED_CHECK(compound_head(page)); })
353 #define PF_NO_COMPOUND(page, enforce) ({ \
354 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
355 PF_POISONED_CHECK(page); })
356 #define PF_SECOND(page, enforce) ({ \
357 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
358 PF_POISONED_CHECK(&page[1]); })
360 /* Which page is the flag stored in */
361 #define FOLIO_PF_ANY 0
362 #define FOLIO_PF_HEAD 0
363 #define FOLIO_PF_ONLY_HEAD 0
364 #define FOLIO_PF_NO_TAIL 0
365 #define FOLIO_PF_NO_COMPOUND 0
366 #define FOLIO_PF_SECOND 1
369 * Macros to create function definitions for page flags
371 #define TESTPAGEFLAG(uname, lname, policy) \
372 static __always_inline bool folio_test_##lname(struct folio *folio) \
373 { return test_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
374 static __always_inline int Page##uname(struct page *page) \
375 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
377 #define SETPAGEFLAG(uname, lname, policy) \
378 static __always_inline \
379 void folio_set_##lname(struct folio *folio) \
380 { set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
381 static __always_inline void SetPage##uname(struct page *page) \
382 { set_bit(PG_##lname, &policy(page, 1)->flags); }
384 #define CLEARPAGEFLAG(uname, lname, policy) \
385 static __always_inline \
386 void folio_clear_##lname(struct folio *folio) \
387 { clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
388 static __always_inline void ClearPage##uname(struct page *page) \
389 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
391 #define __SETPAGEFLAG(uname, lname, policy) \
392 static __always_inline \
393 void __folio_set_##lname(struct folio *folio) \
394 { __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
395 static __always_inline void __SetPage##uname(struct page *page) \
396 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
398 #define __CLEARPAGEFLAG(uname, lname, policy) \
399 static __always_inline \
400 void __folio_clear_##lname(struct folio *folio) \
401 { __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
402 static __always_inline void __ClearPage##uname(struct page *page) \
403 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
405 #define TESTSETFLAG(uname, lname, policy) \
406 static __always_inline \
407 bool folio_test_set_##lname(struct folio *folio) \
408 { return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
409 static __always_inline int TestSetPage##uname(struct page *page) \
410 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
412 #define TESTCLEARFLAG(uname, lname, policy) \
413 static __always_inline \
414 bool folio_test_clear_##lname(struct folio *folio) \
415 { return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
416 static __always_inline int TestClearPage##uname(struct page *page) \
417 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
419 #define PAGEFLAG(uname, lname, policy) \
420 TESTPAGEFLAG(uname, lname, policy) \
421 SETPAGEFLAG(uname, lname, policy) \
422 CLEARPAGEFLAG(uname, lname, policy)
424 #define __PAGEFLAG(uname, lname, policy) \
425 TESTPAGEFLAG(uname, lname, policy) \
426 __SETPAGEFLAG(uname, lname, policy) \
427 __CLEARPAGEFLAG(uname, lname, policy)
429 #define TESTSCFLAG(uname, lname, policy) \
430 TESTSETFLAG(uname, lname, policy) \
431 TESTCLEARFLAG(uname, lname, policy)
433 #define TESTPAGEFLAG_FALSE(uname, lname) \
434 static inline bool folio_test_##lname(const struct folio *folio) { return false; } \
435 static inline int Page##uname(const struct page *page) { return 0; }
437 #define SETPAGEFLAG_NOOP(uname, lname) \
438 static inline void folio_set_##lname(struct folio *folio) { } \
439 static inline void SetPage##uname(struct page *page) { }
441 #define CLEARPAGEFLAG_NOOP(uname, lname) \
442 static inline void folio_clear_##lname(struct folio *folio) { } \
443 static inline void ClearPage##uname(struct page *page) { }
445 #define __CLEARPAGEFLAG_NOOP(uname, lname) \
446 static inline void __folio_clear_##lname(struct folio *folio) { } \
447 static inline void __ClearPage##uname(struct page *page) { }
449 #define TESTSETFLAG_FALSE(uname, lname) \
450 static inline bool folio_test_set_##lname(struct folio *folio) \
452 static inline int TestSetPage##uname(struct page *page) { return 0; }
454 #define TESTCLEARFLAG_FALSE(uname, lname) \
455 static inline bool folio_test_clear_##lname(struct folio *folio) \
457 static inline int TestClearPage##uname(struct page *page) { return 0; }
459 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \
460 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
462 #define TESTSCFLAG_FALSE(uname, lname) \
463 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
465 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
466 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD)
467 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
468 PAGEFLAG(Referenced, referenced, PF_HEAD)
469 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
470 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
471 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
472 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
473 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
474 TESTCLEARFLAG(LRU, lru, PF_HEAD)
475 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
476 TESTCLEARFLAG(Active, active, PF_HEAD)
477 PAGEFLAG(Workingset, workingset, PF_HEAD)
478 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
479 __PAGEFLAG(Slab, slab, PF_NO_TAIL)
480 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
483 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
484 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
485 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
486 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
487 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
488 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
490 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
491 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
492 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
493 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
494 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
495 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
498 * Private page markings that may be used by the filesystem that owns the page
499 * for its own purposes.
500 * - PG_private and PG_private_2 cause release_folio() and co to be invoked
502 PAGEFLAG(Private, private, PF_ANY)
503 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
504 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
505 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
508 * Only test-and-set exist for PG_writeback. The unconditional operators are
509 * risky: they bypass page accounting.
511 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
512 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
513 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
515 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
516 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
517 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
518 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND)
519 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND)
521 #ifdef CONFIG_HIGHMEM
523 * Must use a macro here due to header dependency issues. page_zone() is not
524 * available at this point.
526 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
527 #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f))
529 PAGEFLAG_FALSE(HighMem, highmem)
533 static __always_inline bool folio_test_swapcache(struct folio *folio)
535 return folio_test_swapbacked(folio) &&
536 test_bit(PG_swapcache, folio_flags(folio, 0));
539 static __always_inline bool PageSwapCache(struct page *page)
541 return folio_test_swapcache(page_folio(page));
544 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
545 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
547 PAGEFLAG_FALSE(SwapCache, swapcache)
550 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
551 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
552 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
555 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
556 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
557 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
559 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked)
560 TESTSCFLAG_FALSE(Mlocked, mlocked)
563 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
564 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
566 PAGEFLAG_FALSE(Uncached, uncached)
569 #ifdef CONFIG_MEMORY_FAILURE
570 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
571 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
572 #define __PG_HWPOISON (1UL << PG_hwpoison)
573 #define MAGIC_HWPOISON 0x48575053U /* HWPS */
574 extern void SetPageHWPoisonTakenOff(struct page *page);
575 extern void ClearPageHWPoisonTakenOff(struct page *page);
576 extern bool take_page_off_buddy(struct page *page);
577 extern bool put_page_back_buddy(struct page *page);
579 PAGEFLAG_FALSE(HWPoison, hwpoison)
580 #define __PG_HWPOISON 0
583 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
584 TESTPAGEFLAG(Young, young, PF_ANY)
585 SETPAGEFLAG(Young, young, PF_ANY)
586 TESTCLEARFLAG(Young, young, PF_ANY)
587 PAGEFLAG(Idle, idle, PF_ANY)
591 * PageReported() is used to track reported free pages within the Buddy
592 * allocator. We can use the non-atomic version of the test and set
593 * operations as both should be shielded with the zone lock to prevent
594 * any possible races on the setting or clearing of the bit.
596 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
598 #ifdef CONFIG_MEMORY_HOTPLUG
599 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
601 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
605 * On an anonymous page mapped into a user virtual memory area,
606 * page->mapping points to its anon_vma, not to a struct address_space;
607 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
609 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
610 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
611 * bit; and then page->mapping points, not to an anon_vma, but to a private
612 * structure which KSM associates with that merged page. See ksm.h.
614 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
615 * page and then page->mapping points to a struct movable_operations.
617 * Please note that, confusingly, "page_mapping" refers to the inode
618 * address_space which maps the page from disk; whereas "page_mapped"
619 * refers to user virtual address space into which the page is mapped.
621 * For slab pages, since slab reuses the bits in struct page to store its
622 * internal states, the page->mapping does not exist as such, nor do these
623 * flags below. So in order to avoid testing non-existent bits, please
624 * make sure that PageSlab(page) actually evaluates to false before calling
625 * the following functions (e.g., PageAnon). See mm/slab.h.
627 #define PAGE_MAPPING_ANON 0x1
628 #define PAGE_MAPPING_MOVABLE 0x2
629 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
630 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
633 * Different with flags above, this flag is used only for fsdax mode. It
634 * indicates that this page->mapping is now under reflink case.
636 #define PAGE_MAPPING_DAX_SHARED ((void *)0x1)
638 static __always_inline bool folio_mapping_flags(struct folio *folio)
640 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0;
643 static __always_inline int PageMappingFlags(struct page *page)
645 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
648 static __always_inline bool folio_test_anon(struct folio *folio)
650 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0;
653 static __always_inline bool PageAnon(struct page *page)
655 return folio_test_anon(page_folio(page));
658 static __always_inline bool __folio_test_movable(const struct folio *folio)
660 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
661 PAGE_MAPPING_MOVABLE;
664 static __always_inline int __PageMovable(struct page *page)
666 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
667 PAGE_MAPPING_MOVABLE;
672 * A KSM page is one of those write-protected "shared pages" or "merged pages"
673 * which KSM maps into multiple mms, wherever identical anonymous page content
674 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
675 * anon_vma, but to that page's node of the stable tree.
677 static __always_inline bool folio_test_ksm(struct folio *folio)
679 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
683 static __always_inline bool PageKsm(struct page *page)
685 return folio_test_ksm(page_folio(page));
688 TESTPAGEFLAG_FALSE(Ksm, ksm)
691 u64 stable_page_flags(struct page *page);
694 * folio_test_uptodate - Is this folio up to date?
697 * The uptodate flag is set on a folio when every byte in the folio is
698 * at least as new as the corresponding bytes on storage. Anonymous
699 * and CoW folios are always uptodate. If the folio is not uptodate,
700 * some of the bytes in it may be; see the is_partially_uptodate()
701 * address_space operation.
703 static inline bool folio_test_uptodate(struct folio *folio)
705 bool ret = test_bit(PG_uptodate, folio_flags(folio, 0));
707 * Must ensure that the data we read out of the folio is loaded
708 * _after_ we've loaded folio->flags to check the uptodate bit.
709 * We can skip the barrier if the folio is not uptodate, because
710 * we wouldn't be reading anything from it.
712 * See folio_mark_uptodate() for the other side of the story.
720 static inline int PageUptodate(struct page *page)
722 return folio_test_uptodate(page_folio(page));
725 static __always_inline void __folio_mark_uptodate(struct folio *folio)
728 __set_bit(PG_uptodate, folio_flags(folio, 0));
731 static __always_inline void folio_mark_uptodate(struct folio *folio)
734 * Memory barrier must be issued before setting the PG_uptodate bit,
735 * so that all previous stores issued in order to bring the folio
736 * uptodate are actually visible before folio_test_uptodate becomes true.
739 set_bit(PG_uptodate, folio_flags(folio, 0));
742 static __always_inline void __SetPageUptodate(struct page *page)
744 __folio_mark_uptodate((struct folio *)page);
747 static __always_inline void SetPageUptodate(struct page *page)
749 folio_mark_uptodate((struct folio *)page);
752 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
754 bool __folio_start_writeback(struct folio *folio, bool keep_write);
755 bool set_page_writeback(struct page *page);
757 #define folio_start_writeback(folio) \
758 __folio_start_writeback(folio, false)
759 #define folio_start_writeback_keepwrite(folio) \
760 __folio_start_writeback(folio, true)
762 static inline bool test_set_page_writeback(struct page *page)
764 return set_page_writeback(page);
767 static __always_inline bool folio_test_head(struct folio *folio)
769 return test_bit(PG_head, folio_flags(folio, FOLIO_PF_ANY));
772 static __always_inline int PageHead(struct page *page)
774 PF_POISONED_CHECK(page);
775 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page);
778 __SETPAGEFLAG(Head, head, PF_ANY)
779 __CLEARPAGEFLAG(Head, head, PF_ANY)
780 CLEARPAGEFLAG(Head, head, PF_ANY)
783 * folio_test_large() - Does this folio contain more than one page?
784 * @folio: The folio to test.
786 * Return: True if the folio is larger than one page.
788 static inline bool folio_test_large(struct folio *folio)
790 return folio_test_head(folio);
793 static __always_inline void set_compound_head(struct page *page, struct page *head)
795 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
798 static __always_inline void clear_compound_head(struct page *page)
800 WRITE_ONCE(page->compound_head, 0);
803 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
804 static inline void ClearPageCompound(struct page *page)
806 BUG_ON(!PageHead(page));
811 #define PG_head_mask ((1UL << PG_head))
813 #ifdef CONFIG_HUGETLB_PAGE
814 int PageHuge(struct page *page);
815 bool folio_test_hugetlb(struct folio *folio);
817 TESTPAGEFLAG_FALSE(Huge, hugetlb)
820 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
822 * PageHuge() only returns true for hugetlbfs pages, but not for
823 * normal or transparent huge pages.
825 * PageTransHuge() returns true for both transparent huge and
826 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
827 * called only in the core VM paths where hugetlbfs pages can't exist.
829 static inline int PageTransHuge(struct page *page)
831 VM_BUG_ON_PAGE(PageTail(page), page);
832 return PageHead(page);
835 static inline bool folio_test_transhuge(struct folio *folio)
837 return folio_test_head(folio);
841 * PageTransCompound returns true for both transparent huge pages
842 * and hugetlbfs pages, so it should only be called when it's known
843 * that hugetlbfs pages aren't involved.
845 static inline int PageTransCompound(struct page *page)
847 return PageCompound(page);
851 * PageTransTail returns true for both transparent huge pages
852 * and hugetlbfs pages, so it should only be called when it's known
853 * that hugetlbfs pages aren't involved.
855 static inline int PageTransTail(struct page *page)
857 return PageTail(page);
860 TESTPAGEFLAG_FALSE(TransHuge, transhuge)
861 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
862 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap)
863 TESTPAGEFLAG_FALSE(TransTail, transtail)
866 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
868 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
871 * This flag is set by hwpoison handler. Cleared by THP split or free page.
873 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
874 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
876 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
877 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
881 * Check if a page is currently marked HWPoisoned. Note that this check is
882 * best effort only and inherently racy: there is no way to synchronize with
885 static inline bool is_page_hwpoison(struct page *page)
887 if (PageHWPoison(page))
889 return PageHuge(page) && PageHWPoison(compound_head(page));
893 * For pages that are never mapped to userspace (and aren't PageSlab),
894 * page_type may be used. Because it is initialised to -1, we invert the
895 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
896 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
897 * low bits so that an underflow or overflow of page_mapcount() won't be
898 * mistaken for a page type value.
901 #define PAGE_TYPE_BASE 0xf0000000
902 /* Reserve 0x0000007f to catch underflows of page_mapcount */
903 #define PAGE_MAPCOUNT_RESERVE -128
904 #define PG_buddy 0x00000080
905 #define PG_offline 0x00000100
906 #define PG_table 0x00000200
907 #define PG_guard 0x00000400
909 #define PageType(page, flag) \
910 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
912 static inline int page_type_has_type(unsigned int page_type)
914 return (int)page_type < PAGE_MAPCOUNT_RESERVE;
917 static inline int page_has_type(struct page *page)
919 return page_type_has_type(page->page_type);
922 #define PAGE_TYPE_OPS(uname, lname) \
923 static __always_inline int Page##uname(struct page *page) \
925 return PageType(page, PG_##lname); \
927 static __always_inline void __SetPage##uname(struct page *page) \
929 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
930 page->page_type &= ~PG_##lname; \
932 static __always_inline void __ClearPage##uname(struct page *page) \
934 VM_BUG_ON_PAGE(!Page##uname(page), page); \
935 page->page_type |= PG_##lname; \
939 * PageBuddy() indicates that the page is free and in the buddy system
940 * (see mm/page_alloc.c).
942 PAGE_TYPE_OPS(Buddy, buddy)
945 * PageOffline() indicates that the page is logically offline although the
946 * containing section is online. (e.g. inflated in a balloon driver or
947 * not onlined when onlining the section).
948 * The content of these pages is effectively stale. Such pages should not
949 * be touched (read/write/dump/save) except by their owner.
951 * If a driver wants to allow to offline unmovable PageOffline() pages without
952 * putting them back to the buddy, it can do so via the memory notifier by
953 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
954 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
955 * pages (now with a reference count of zero) are treated like free pages,
956 * allowing the containing memory block to get offlined. A driver that
957 * relies on this feature is aware that re-onlining the memory block will
958 * require to re-set the pages PageOffline() and not giving them to the
959 * buddy via online_page_callback_t.
961 * There are drivers that mark a page PageOffline() and expect there won't be
962 * any further access to page content. PFN walkers that read content of random
963 * pages should check PageOffline() and synchronize with such drivers using
964 * page_offline_freeze()/page_offline_thaw().
966 PAGE_TYPE_OPS(Offline, offline)
968 extern void page_offline_freeze(void);
969 extern void page_offline_thaw(void);
970 extern void page_offline_begin(void);
971 extern void page_offline_end(void);
974 * Marks pages in use as page tables.
976 PAGE_TYPE_OPS(Table, table)
979 * Marks guardpages used with debug_pagealloc.
981 PAGE_TYPE_OPS(Guard, guard)
983 extern bool is_free_buddy_page(struct page *page);
985 PAGEFLAG(Isolated, isolated, PF_ANY);
987 static __always_inline int PageAnonExclusive(struct page *page)
989 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
990 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
991 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
994 static __always_inline void SetPageAnonExclusive(struct page *page)
996 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
997 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
998 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1001 static __always_inline void ClearPageAnonExclusive(struct page *page)
1003 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1004 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1005 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1008 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1010 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1011 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1012 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1016 #define __PG_MLOCKED (1UL << PG_mlocked)
1018 #define __PG_MLOCKED 0
1022 * Flags checked when a page is freed. Pages being freed should not have
1023 * these flags set. If they are, there is a problem.
1025 #define PAGE_FLAGS_CHECK_AT_FREE \
1026 (1UL << PG_lru | 1UL << PG_locked | \
1027 1UL << PG_private | 1UL << PG_private_2 | \
1028 1UL << PG_writeback | 1UL << PG_reserved | \
1029 1UL << PG_slab | 1UL << PG_active | \
1030 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK)
1033 * Flags checked when a page is prepped for return by the page allocator.
1034 * Pages being prepped should not have these flags set. If they are set,
1035 * there has been a kernel bug or struct page corruption.
1037 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1038 * alloc-free cycle to prevent from reusing the page.
1040 #define PAGE_FLAGS_CHECK_AT_PREP \
1041 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK)
1043 #define PAGE_FLAGS_PRIVATE \
1044 (1UL << PG_private | 1UL << PG_private_2)
1046 * page_has_private - Determine if page has private stuff
1047 * @page: The page to be checked
1049 * Determine if a page has private stuff, indicating that release routines
1050 * should be invoked upon it.
1052 static inline int page_has_private(struct page *page)
1054 return !!(page->flags & PAGE_FLAGS_PRIVATE);
1057 static inline bool folio_has_private(struct folio *folio)
1059 return page_has_private(&folio->page);
1066 #undef PF_NO_COMPOUND
1068 #endif /* !__GENERATING_BOUNDS_H */
1070 #endif /* PAGE_FLAGS_H */