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. */
102 PG_writeback, /* Page is under writeback */
107 PG_head, /* Must be in bit 6 */
108 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
113 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
116 PG_private, /* If pagecache, has fs-private data */
117 PG_private_2, /* If pagecache, has fs aux data */
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 /* non-lru isolated movable page */
175 PG_isolated = PG_reclaim,
177 /* Only valid for buddy pages. Used to track pages that are reported */
178 PG_reported = PG_uptodate,
180 #ifdef CONFIG_MEMORY_HOTPLUG
181 /* For self-hosted memmap pages */
182 PG_vmemmap_self_hosted = PG_owner_priv_1,
186 * Flags only valid for compound pages. Stored in first tail page's
187 * flags word. Cannot use the first 8 flags or any flag marked as
191 /* At least one page in this folio has the hwpoison flag set */
192 PG_has_hwpoisoned = PG_error,
193 PG_hugetlb = PG_active,
194 PG_large_rmappable = PG_workingset, /* anon or file-backed */
197 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
199 #ifndef __GENERATING_BOUNDS_H
201 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
202 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key);
205 * Return the real head page struct iff the @page is a fake head page, otherwise
206 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst.
208 static __always_inline const struct page *page_fixed_fake_head(const struct page *page)
210 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key))
214 * Only addresses aligned with PAGE_SIZE of struct page may be fake head
215 * struct page. The alignment check aims to avoid access the fields (
216 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly)
217 * cold cacheline in some cases.
219 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) &&
220 test_bit(PG_head, &page->flags)) {
222 * We can safely access the field of the @page[1] with PG_head
223 * because the @page is a compound page composed with at least
224 * two contiguous pages.
226 unsigned long head = READ_ONCE(page[1].compound_head);
228 if (likely(head & 1))
229 return (const struct page *)(head - 1);
234 static inline const struct page *page_fixed_fake_head(const struct page *page)
240 static __always_inline int page_is_fake_head(const struct page *page)
242 return page_fixed_fake_head(page) != page;
245 static inline unsigned long _compound_head(const struct page *page)
247 unsigned long head = READ_ONCE(page->compound_head);
249 if (unlikely(head & 1))
251 return (unsigned long)page_fixed_fake_head(page);
254 #define compound_head(page) ((typeof(page))_compound_head(page))
257 * page_folio - Converts from page to folio.
260 * Every page is part of a folio. This function cannot be called on a
263 * Context: No reference, nor lock is required on @page. If the caller
264 * does not hold a reference, this call may race with a folio split, so
265 * it should re-check the folio still contains this page after gaining
266 * a reference on the folio.
267 * Return: The folio which contains this page.
269 #define page_folio(p) (_Generic((p), \
270 const struct page *: (const struct folio *)_compound_head(p), \
271 struct page *: (struct folio *)_compound_head(p)))
274 * folio_page - Return a page from a folio.
276 * @n: The page number to return.
278 * @n is relative to the start of the folio. This function does not
279 * check that the page number lies within @folio; the caller is presumed
280 * to have a reference to the page.
282 #define folio_page(folio, n) nth_page(&(folio)->page, n)
284 static __always_inline int PageTail(const struct page *page)
286 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page);
289 static __always_inline int PageCompound(const struct page *page)
291 return test_bit(PG_head, &page->flags) ||
292 READ_ONCE(page->compound_head) & 1;
295 #define PAGE_POISON_PATTERN -1l
296 static inline int PagePoisoned(const struct page *page)
298 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN;
301 #ifdef CONFIG_DEBUG_VM
302 void page_init_poison(struct page *page, size_t size);
304 static inline void page_init_poison(struct page *page, size_t size)
309 static const unsigned long *const_folio_flags(const struct folio *folio,
312 const struct page *page = &folio->page;
314 VM_BUG_ON_PGFLAGS(PageTail(page), page);
315 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
316 return &page[n].flags;
319 static unsigned long *folio_flags(struct folio *folio, unsigned n)
321 struct page *page = &folio->page;
323 VM_BUG_ON_PGFLAGS(PageTail(page), page);
324 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
325 return &page[n].flags;
329 * Page flags policies wrt compound pages
332 * check if this struct page poisoned/uninitialized
335 * the page flag is relevant for small, head and tail pages.
338 * for compound page all operations related to the page flag applied to
342 * modifications of the page flag must be done on small or head pages,
343 * checks can be done on tail pages too.
346 * the page flag is not relevant for compound pages.
349 * the page flag is stored in the first tail page.
351 #define PF_POISONED_CHECK(page) ({ \
352 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
354 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
355 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
356 #define PF_NO_TAIL(page, enforce) ({ \
357 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
358 PF_POISONED_CHECK(compound_head(page)); })
359 #define PF_NO_COMPOUND(page, enforce) ({ \
360 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
361 PF_POISONED_CHECK(page); })
362 #define PF_SECOND(page, enforce) ({ \
363 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
364 PF_POISONED_CHECK(&page[1]); })
366 /* Which page is the flag stored in */
367 #define FOLIO_PF_ANY 0
368 #define FOLIO_PF_HEAD 0
369 #define FOLIO_PF_NO_TAIL 0
370 #define FOLIO_PF_NO_COMPOUND 0
371 #define FOLIO_PF_SECOND 1
373 #define FOLIO_HEAD_PAGE 0
374 #define FOLIO_SECOND_PAGE 1
377 * Macros to create function definitions for page flags
379 #define FOLIO_TEST_FLAG(name, page) \
380 static __always_inline bool folio_test_##name(const struct folio *folio) \
381 { return test_bit(PG_##name, const_folio_flags(folio, page)); }
383 #define FOLIO_SET_FLAG(name, page) \
384 static __always_inline void folio_set_##name(struct folio *folio) \
385 { set_bit(PG_##name, folio_flags(folio, page)); }
387 #define FOLIO_CLEAR_FLAG(name, page) \
388 static __always_inline void folio_clear_##name(struct folio *folio) \
389 { clear_bit(PG_##name, folio_flags(folio, page)); }
391 #define __FOLIO_SET_FLAG(name, page) \
392 static __always_inline void __folio_set_##name(struct folio *folio) \
393 { __set_bit(PG_##name, folio_flags(folio, page)); }
395 #define __FOLIO_CLEAR_FLAG(name, page) \
396 static __always_inline void __folio_clear_##name(struct folio *folio) \
397 { __clear_bit(PG_##name, folio_flags(folio, page)); }
399 #define FOLIO_TEST_SET_FLAG(name, page) \
400 static __always_inline bool folio_test_set_##name(struct folio *folio) \
401 { return test_and_set_bit(PG_##name, folio_flags(folio, page)); }
403 #define FOLIO_TEST_CLEAR_FLAG(name, page) \
404 static __always_inline bool folio_test_clear_##name(struct folio *folio) \
405 { return test_and_clear_bit(PG_##name, folio_flags(folio, page)); }
407 #define FOLIO_FLAG(name, page) \
408 FOLIO_TEST_FLAG(name, page) \
409 FOLIO_SET_FLAG(name, page) \
410 FOLIO_CLEAR_FLAG(name, page)
412 #define TESTPAGEFLAG(uname, lname, policy) \
413 FOLIO_TEST_FLAG(lname, FOLIO_##policy) \
414 static __always_inline int Page##uname(const struct page *page) \
415 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
417 #define SETPAGEFLAG(uname, lname, policy) \
418 FOLIO_SET_FLAG(lname, FOLIO_##policy) \
419 static __always_inline void SetPage##uname(struct page *page) \
420 { set_bit(PG_##lname, &policy(page, 1)->flags); }
422 #define CLEARPAGEFLAG(uname, lname, policy) \
423 FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \
424 static __always_inline void ClearPage##uname(struct page *page) \
425 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
427 #define __SETPAGEFLAG(uname, lname, policy) \
428 __FOLIO_SET_FLAG(lname, FOLIO_##policy) \
429 static __always_inline void __SetPage##uname(struct page *page) \
430 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
432 #define __CLEARPAGEFLAG(uname, lname, policy) \
433 __FOLIO_CLEAR_FLAG(lname, 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 FOLIO_TEST_SET_FLAG(lname, FOLIO_##policy) \
439 static __always_inline int TestSetPage##uname(struct page *page) \
440 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
442 #define TESTCLEARFLAG(uname, lname, policy) \
443 FOLIO_TEST_CLEAR_FLAG(lname, FOLIO_##policy) \
444 static __always_inline int TestClearPage##uname(struct page *page) \
445 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
447 #define PAGEFLAG(uname, lname, policy) \
448 TESTPAGEFLAG(uname, lname, policy) \
449 SETPAGEFLAG(uname, lname, policy) \
450 CLEARPAGEFLAG(uname, lname, policy)
452 #define __PAGEFLAG(uname, lname, policy) \
453 TESTPAGEFLAG(uname, lname, policy) \
454 __SETPAGEFLAG(uname, lname, policy) \
455 __CLEARPAGEFLAG(uname, lname, policy)
457 #define TESTSCFLAG(uname, lname, policy) \
458 TESTSETFLAG(uname, lname, policy) \
459 TESTCLEARFLAG(uname, lname, policy)
461 #define TESTPAGEFLAG_FALSE(uname, lname) \
462 static inline bool folio_test_##lname(const struct folio *folio) { return false; } \
463 static inline int Page##uname(const struct page *page) { return 0; }
465 #define SETPAGEFLAG_NOOP(uname, lname) \
466 static inline void folio_set_##lname(struct folio *folio) { } \
467 static inline void SetPage##uname(struct page *page) { }
469 #define CLEARPAGEFLAG_NOOP(uname, lname) \
470 static inline void folio_clear_##lname(struct folio *folio) { } \
471 static inline void ClearPage##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 TESTSETFLAG_FALSE(uname, lname) \
478 static inline bool folio_test_set_##lname(struct folio *folio) \
480 static inline int TestSetPage##uname(struct page *page) { return 0; }
482 #define TESTCLEARFLAG_FALSE(uname, lname) \
483 static inline bool folio_test_clear_##lname(struct folio *folio) \
485 static inline int TestClearPage##uname(struct page *page) { return 0; }
487 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \
488 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
490 #define TESTSCFLAG_FALSE(uname, lname) \
491 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
493 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
494 FOLIO_FLAG(waiters, FOLIO_HEAD_PAGE)
495 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
496 PAGEFLAG(Referenced, referenced, PF_HEAD)
497 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
498 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
499 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
500 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
501 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
502 TESTCLEARFLAG(LRU, lru, PF_HEAD)
503 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
504 TESTCLEARFLAG(Active, active, PF_HEAD)
505 PAGEFLAG(Workingset, workingset, PF_HEAD)
506 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
507 __PAGEFLAG(Slab, slab, PF_NO_TAIL)
508 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
511 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
512 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
513 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
514 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
515 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
516 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
518 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
519 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
520 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
521 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
522 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
523 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
526 * Private page markings that may be used by the filesystem that owns the page
527 * for its own purposes.
528 * - PG_private and PG_private_2 cause release_folio() and co to be invoked
530 PAGEFLAG(Private, private, PF_ANY)
531 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
532 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
533 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
536 * Only test-and-set exist for PG_writeback. The unconditional operators are
537 * risky: they bypass page accounting.
539 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
540 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
541 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
543 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
544 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
545 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
546 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND)
547 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND)
549 #ifdef CONFIG_HIGHMEM
551 * Must use a macro here due to header dependency issues. page_zone() is not
552 * available at this point.
554 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
555 #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f))
557 PAGEFLAG_FALSE(HighMem, highmem)
561 static __always_inline bool folio_test_swapcache(const struct folio *folio)
563 return folio_test_swapbacked(folio) &&
564 test_bit(PG_swapcache, const_folio_flags(folio, 0));
567 static __always_inline bool PageSwapCache(const struct page *page)
569 return folio_test_swapcache(page_folio(page));
572 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
573 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
575 PAGEFLAG_FALSE(SwapCache, swapcache)
578 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
579 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
580 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
583 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
584 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
585 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
587 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked)
588 TESTSCFLAG_FALSE(Mlocked, mlocked)
591 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
592 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
594 PAGEFLAG_FALSE(Uncached, uncached)
597 #ifdef CONFIG_MEMORY_FAILURE
598 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
599 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
600 #define __PG_HWPOISON (1UL << PG_hwpoison)
601 #define MAGIC_HWPOISON 0x48575053U /* HWPS */
602 extern void SetPageHWPoisonTakenOff(struct page *page);
603 extern void ClearPageHWPoisonTakenOff(struct page *page);
604 extern bool take_page_off_buddy(struct page *page);
605 extern bool put_page_back_buddy(struct page *page);
607 PAGEFLAG_FALSE(HWPoison, hwpoison)
608 #define __PG_HWPOISON 0
611 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
612 FOLIO_TEST_FLAG(young, FOLIO_HEAD_PAGE)
613 FOLIO_SET_FLAG(young, FOLIO_HEAD_PAGE)
614 FOLIO_TEST_CLEAR_FLAG(young, FOLIO_HEAD_PAGE)
615 FOLIO_FLAG(idle, FOLIO_HEAD_PAGE)
619 * PageReported() is used to track reported free pages within the Buddy
620 * allocator. We can use the non-atomic version of the test and set
621 * operations as both should be shielded with the zone lock to prevent
622 * any possible races on the setting or clearing of the bit.
624 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
626 #ifdef CONFIG_MEMORY_HOTPLUG
627 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
629 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
633 * On an anonymous page mapped into a user virtual memory area,
634 * page->mapping points to its anon_vma, not to a struct address_space;
635 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
637 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
638 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
639 * bit; and then page->mapping points, not to an anon_vma, but to a private
640 * structure which KSM associates with that merged page. See ksm.h.
642 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
643 * page and then page->mapping points to a struct movable_operations.
645 * Please note that, confusingly, "page_mapping" refers to the inode
646 * address_space which maps the page from disk; whereas "page_mapped"
647 * refers to user virtual address space into which the page is mapped.
649 * For slab pages, since slab reuses the bits in struct page to store its
650 * internal states, the page->mapping does not exist as such, nor do these
651 * flags below. So in order to avoid testing non-existent bits, please
652 * make sure that PageSlab(page) actually evaluates to false before calling
653 * the following functions (e.g., PageAnon). See mm/slab.h.
655 #define PAGE_MAPPING_ANON 0x1
656 #define PAGE_MAPPING_MOVABLE 0x2
657 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
658 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
661 * Different with flags above, this flag is used only for fsdax mode. It
662 * indicates that this page->mapping is now under reflink case.
664 #define PAGE_MAPPING_DAX_SHARED ((void *)0x1)
666 static __always_inline bool folio_mapping_flags(const struct folio *folio)
668 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0;
671 static __always_inline int PageMappingFlags(const struct page *page)
673 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
676 static __always_inline bool folio_test_anon(const struct folio *folio)
678 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0;
681 static __always_inline bool PageAnon(const struct page *page)
683 return folio_test_anon(page_folio(page));
686 static __always_inline bool __folio_test_movable(const struct folio *folio)
688 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
689 PAGE_MAPPING_MOVABLE;
692 static __always_inline int __PageMovable(const struct page *page)
694 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
695 PAGE_MAPPING_MOVABLE;
700 * A KSM page is one of those write-protected "shared pages" or "merged pages"
701 * which KSM maps into multiple mms, wherever identical anonymous page content
702 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
703 * anon_vma, but to that page's node of the stable tree.
705 static __always_inline bool folio_test_ksm(const struct folio *folio)
707 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
711 static __always_inline bool PageKsm(const struct page *page)
713 return folio_test_ksm(page_folio(page));
716 TESTPAGEFLAG_FALSE(Ksm, ksm)
719 u64 stable_page_flags(struct page *page);
722 * folio_xor_flags_has_waiters - Change some folio flags.
724 * @mask: Bits set in this word will be changed.
726 * This must only be used for flags which are changed with the folio
727 * lock held. For example, it is unsafe to use for PG_dirty as that
728 * can be set without the folio lock held. It can also only be used
729 * on flags which are in the range 0-6 as some of the implementations
730 * only affect those bits.
732 * Return: Whether there are tasks waiting on the folio.
734 static inline bool folio_xor_flags_has_waiters(struct folio *folio,
737 return xor_unlock_is_negative_byte(mask, folio_flags(folio, 0));
741 * folio_test_uptodate - Is this folio up to date?
744 * The uptodate flag is set on a folio when every byte in the folio is
745 * at least as new as the corresponding bytes on storage. Anonymous
746 * and CoW folios are always uptodate. If the folio is not uptodate,
747 * some of the bytes in it may be; see the is_partially_uptodate()
748 * address_space operation.
750 static inline bool folio_test_uptodate(const struct folio *folio)
752 bool ret = test_bit(PG_uptodate, const_folio_flags(folio, 0));
754 * Must ensure that the data we read out of the folio is loaded
755 * _after_ we've loaded folio->flags to check the uptodate bit.
756 * We can skip the barrier if the folio is not uptodate, because
757 * we wouldn't be reading anything from it.
759 * See folio_mark_uptodate() for the other side of the story.
767 static inline int PageUptodate(const struct page *page)
769 return folio_test_uptodate(page_folio(page));
772 static __always_inline void __folio_mark_uptodate(struct folio *folio)
775 __set_bit(PG_uptodate, folio_flags(folio, 0));
778 static __always_inline void folio_mark_uptodate(struct folio *folio)
781 * Memory barrier must be issued before setting the PG_uptodate bit,
782 * so that all previous stores issued in order to bring the folio
783 * uptodate are actually visible before folio_test_uptodate becomes true.
786 set_bit(PG_uptodate, folio_flags(folio, 0));
789 static __always_inline void __SetPageUptodate(struct page *page)
791 __folio_mark_uptodate((struct folio *)page);
794 static __always_inline void SetPageUptodate(struct page *page)
796 folio_mark_uptodate((struct folio *)page);
799 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
801 void __folio_start_writeback(struct folio *folio, bool keep_write);
802 void set_page_writeback(struct page *page);
804 #define folio_start_writeback(folio) \
805 __folio_start_writeback(folio, false)
806 #define folio_start_writeback_keepwrite(folio) \
807 __folio_start_writeback(folio, true)
809 static __always_inline bool folio_test_head(const struct folio *folio)
811 return test_bit(PG_head, const_folio_flags(folio, FOLIO_PF_ANY));
814 static __always_inline int PageHead(const struct page *page)
816 PF_POISONED_CHECK(page);
817 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page);
820 __SETPAGEFLAG(Head, head, PF_ANY)
821 __CLEARPAGEFLAG(Head, head, PF_ANY)
822 CLEARPAGEFLAG(Head, head, PF_ANY)
825 * folio_test_large() - Does this folio contain more than one page?
826 * @folio: The folio to test.
828 * Return: True if the folio is larger than one page.
830 static inline bool folio_test_large(const struct folio *folio)
832 return folio_test_head(folio);
835 static __always_inline void set_compound_head(struct page *page, struct page *head)
837 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
840 static __always_inline void clear_compound_head(struct page *page)
842 WRITE_ONCE(page->compound_head, 0);
845 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
846 static inline void ClearPageCompound(struct page *page)
848 BUG_ON(!PageHead(page));
851 PAGEFLAG(LargeRmappable, large_rmappable, PF_SECOND)
853 TESTPAGEFLAG_FALSE(LargeRmappable, large_rmappable)
856 #define PG_head_mask ((1UL << PG_head))
858 #ifdef CONFIG_HUGETLB_PAGE
859 int PageHuge(const struct page *page);
860 SETPAGEFLAG(HugeTLB, hugetlb, PF_SECOND)
861 CLEARPAGEFLAG(HugeTLB, hugetlb, PF_SECOND)
864 * folio_test_hugetlb - Determine if the folio belongs to hugetlbfs
865 * @folio: The folio to test.
867 * Context: Any context. Caller should have a reference on the folio to
868 * prevent it from being turned into a tail page.
869 * Return: True for hugetlbfs folios, false for anon folios or folios
870 * belonging to other filesystems.
872 static inline bool folio_test_hugetlb(const struct folio *folio)
874 return folio_test_large(folio) &&
875 test_bit(PG_hugetlb, const_folio_flags(folio, 1));
878 TESTPAGEFLAG_FALSE(Huge, hugetlb)
881 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
883 * PageHuge() only returns true for hugetlbfs pages, but not for
884 * normal or transparent huge pages.
886 * PageTransHuge() returns true for both transparent huge and
887 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
888 * called only in the core VM paths where hugetlbfs pages can't exist.
890 static inline int PageTransHuge(const struct page *page)
892 VM_BUG_ON_PAGE(PageTail(page), page);
893 return PageHead(page);
897 * PageTransCompound returns true for both transparent huge pages
898 * and hugetlbfs pages, so it should only be called when it's known
899 * that hugetlbfs pages aren't involved.
901 static inline int PageTransCompound(const struct page *page)
903 return PageCompound(page);
907 * PageTransTail returns true for both transparent huge pages
908 * and hugetlbfs pages, so it should only be called when it's known
909 * that hugetlbfs pages aren't involved.
911 static inline int PageTransTail(const struct page *page)
913 return PageTail(page);
916 TESTPAGEFLAG_FALSE(TransHuge, transhuge)
917 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
918 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap)
919 TESTPAGEFLAG_FALSE(TransTail, transtail)
922 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
924 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
927 * This flag is set by hwpoison handler. Cleared by THP split or free page.
929 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
930 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
932 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
933 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
937 * Check if a page is currently marked HWPoisoned. Note that this check is
938 * best effort only and inherently racy: there is no way to synchronize with
941 static inline bool is_page_hwpoison(struct page *page)
943 if (PageHWPoison(page))
945 return PageHuge(page) && PageHWPoison(compound_head(page));
949 * For pages that are never mapped to userspace (and aren't PageSlab),
950 * page_type may be used. Because it is initialised to -1, we invert the
951 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
952 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
953 * low bits so that an underflow or overflow of page_mapcount() won't be
954 * mistaken for a page type value.
957 #define PAGE_TYPE_BASE 0xf0000000
958 /* Reserve 0x0000007f to catch underflows of page_mapcount */
959 #define PAGE_MAPCOUNT_RESERVE -128
960 #define PG_buddy 0x00000080
961 #define PG_offline 0x00000100
962 #define PG_table 0x00000200
963 #define PG_guard 0x00000400
965 #define PageType(page, flag) \
966 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
967 #define folio_test_type(folio, flag) \
968 ((folio->page.page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
970 static inline int page_type_has_type(unsigned int page_type)
972 return (int)page_type < PAGE_MAPCOUNT_RESERVE;
975 static inline int page_has_type(const struct page *page)
977 return page_type_has_type(page->page_type);
980 #define PAGE_TYPE_OPS(uname, lname, fname) \
981 static __always_inline int Page##uname(const struct page *page) \
983 return PageType(page, PG_##lname); \
985 static __always_inline int folio_test_##fname(const struct folio *folio)\
987 return folio_test_type(folio, PG_##lname); \
989 static __always_inline void __SetPage##uname(struct page *page) \
991 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
992 page->page_type &= ~PG_##lname; \
994 static __always_inline void __folio_set_##fname(struct folio *folio) \
996 VM_BUG_ON_FOLIO(!folio_test_type(folio, 0), folio); \
997 folio->page.page_type &= ~PG_##lname; \
999 static __always_inline void __ClearPage##uname(struct page *page) \
1001 VM_BUG_ON_PAGE(!Page##uname(page), page); \
1002 page->page_type |= PG_##lname; \
1004 static __always_inline void __folio_clear_##fname(struct folio *folio) \
1006 VM_BUG_ON_FOLIO(!folio_test_##fname(folio), folio); \
1007 folio->page.page_type |= PG_##lname; \
1011 * PageBuddy() indicates that the page is free and in the buddy system
1012 * (see mm/page_alloc.c).
1014 PAGE_TYPE_OPS(Buddy, buddy, buddy)
1017 * PageOffline() indicates that the page is logically offline although the
1018 * containing section is online. (e.g. inflated in a balloon driver or
1019 * not onlined when onlining the section).
1020 * The content of these pages is effectively stale. Such pages should not
1021 * be touched (read/write/dump/save) except by their owner.
1023 * If a driver wants to allow to offline unmovable PageOffline() pages without
1024 * putting them back to the buddy, it can do so via the memory notifier by
1025 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
1026 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
1027 * pages (now with a reference count of zero) are treated like free pages,
1028 * allowing the containing memory block to get offlined. A driver that
1029 * relies on this feature is aware that re-onlining the memory block will
1030 * require to re-set the pages PageOffline() and not giving them to the
1031 * buddy via online_page_callback_t.
1033 * There are drivers that mark a page PageOffline() and expect there won't be
1034 * any further access to page content. PFN walkers that read content of random
1035 * pages should check PageOffline() and synchronize with such drivers using
1036 * page_offline_freeze()/page_offline_thaw().
1038 PAGE_TYPE_OPS(Offline, offline, offline)
1040 extern void page_offline_freeze(void);
1041 extern void page_offline_thaw(void);
1042 extern void page_offline_begin(void);
1043 extern void page_offline_end(void);
1046 * Marks pages in use as page tables.
1048 PAGE_TYPE_OPS(Table, table, pgtable)
1051 * Marks guardpages used with debug_pagealloc.
1053 PAGE_TYPE_OPS(Guard, guard, guard)
1055 extern bool is_free_buddy_page(struct page *page);
1057 PAGEFLAG(Isolated, isolated, PF_ANY);
1059 static __always_inline int PageAnonExclusive(const struct page *page)
1061 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1062 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1063 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1066 static __always_inline void SetPageAnonExclusive(struct page *page)
1068 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1069 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1070 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1073 static __always_inline void ClearPageAnonExclusive(struct page *page)
1075 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1076 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1077 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1080 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1082 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1083 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1084 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1088 #define __PG_MLOCKED (1UL << PG_mlocked)
1090 #define __PG_MLOCKED 0
1094 * Flags checked when a page is freed. Pages being freed should not have
1095 * these flags set. If they are, there is a problem.
1097 #define PAGE_FLAGS_CHECK_AT_FREE \
1098 (1UL << PG_lru | 1UL << PG_locked | \
1099 1UL << PG_private | 1UL << PG_private_2 | \
1100 1UL << PG_writeback | 1UL << PG_reserved | \
1101 1UL << PG_slab | 1UL << PG_active | \
1102 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK)
1105 * Flags checked when a page is prepped for return by the page allocator.
1106 * Pages being prepped should not have these flags set. If they are set,
1107 * there has been a kernel bug or struct page corruption.
1109 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1110 * alloc-free cycle to prevent from reusing the page.
1112 #define PAGE_FLAGS_CHECK_AT_PREP \
1113 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK)
1116 * Flags stored in the second page of a compound page. They may overlap
1117 * the CHECK_AT_FREE flags above, so need to be cleared.
1119 #define PAGE_FLAGS_SECOND \
1120 (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \
1121 1UL << PG_hugetlb | 1UL << PG_large_rmappable)
1123 #define PAGE_FLAGS_PRIVATE \
1124 (1UL << PG_private | 1UL << PG_private_2)
1126 * page_has_private - Determine if page has private stuff
1127 * @page: The page to be checked
1129 * Determine if a page has private stuff, indicating that release routines
1130 * should be invoked upon it.
1132 static inline int page_has_private(const struct page *page)
1134 return !!(page->flags & PAGE_FLAGS_PRIVATE);
1137 static inline bool folio_has_private(const struct folio *folio)
1139 return page_has_private(&folio->page);
1145 #undef PF_NO_COMPOUND
1147 #endif /* !__GENERATING_BOUNDS_H */
1149 #endif /* PAGE_FLAGS_H */