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" */
112 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
115 PG_private, /* If pagecache, has fs-private data */
116 PG_private_2, /* If pagecache, has fs aux data */
117 PG_mappedtodisk, /* Has blocks allocated on-disk */
118 PG_reclaim, /* To be reclaimed asap */
119 PG_swapbacked, /* Page is backed by RAM/swap */
120 PG_unevictable, /* Page is "unevictable" */
122 PG_mlocked, /* Page is vma mlocked */
124 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
125 PG_uncached, /* Page has been mapped as uncached */
127 #ifdef CONFIG_MEMORY_FAILURE
128 PG_hwpoison, /* hardware poisoned page. Don't touch */
130 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
134 #ifdef CONFIG_ARCH_USES_PG_ARCH_X
140 PG_readahead = PG_reclaim,
143 * Depending on the way an anonymous folio can be mapped into a page
144 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped
145 * THP), PG_anon_exclusive may be set only for the head page or for
146 * tail pages of an anonymous folio. For now, we only expect it to be
147 * set on tail pages for PTE-mapped THP.
149 PG_anon_exclusive = PG_mappedtodisk,
152 PG_checked = PG_owner_priv_1,
155 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
157 /* Two page bits are conscripted by FS-Cache to maintain local caching
158 * state. These bits are set on pages belonging to the netfs's inodes
159 * when those inodes are being locally cached.
161 PG_fscache = PG_private_2, /* page backed by cache */
164 /* Pinned in Xen as a read-only pagetable page. */
165 PG_pinned = PG_owner_priv_1,
166 /* Pinned as part of domain save (see xen_mm_pin_all()). */
167 PG_savepinned = PG_dirty,
168 /* Has a grant mapping of another (foreign) domain's page. */
169 PG_foreign = PG_owner_priv_1,
170 /* Remapped by swiotlb-xen. */
171 PG_xen_remapped = PG_owner_priv_1,
173 /* non-lru isolated movable page */
174 PG_isolated = PG_reclaim,
176 /* Only valid for buddy pages. Used to track pages that are reported */
177 PG_reported = PG_uptodate,
179 #ifdef CONFIG_MEMORY_HOTPLUG
180 /* For self-hosted memmap pages */
181 PG_vmemmap_self_hosted = PG_owner_priv_1,
185 * Flags only valid for compound pages. Stored in first tail page's
186 * flags word. Cannot use the first 8 flags or any flag marked as
190 /* At least one page in this folio has the hwpoison flag set */
191 PG_has_hwpoisoned = PG_error,
192 PG_large_rmappable = PG_workingset, /* anon or file-backed */
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(const 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(const struct page *page)
284 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page);
287 static __always_inline int PageCompound(const 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 const unsigned long *const_folio_flags(const struct folio *folio,
310 const struct page *page = &folio->page;
312 VM_BUG_ON_PGFLAGS(PageTail(page), page);
313 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
314 return &page[n].flags;
317 static unsigned long *folio_flags(struct folio *folio, unsigned n)
319 struct page *page = &folio->page;
321 VM_BUG_ON_PGFLAGS(PageTail(page), page);
322 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
323 return &page[n].flags;
327 * Page flags policies wrt compound pages
330 * check if this struct page poisoned/uninitialized
333 * the page flag is relevant for small, head and tail pages.
336 * for compound page all operations related to the page flag applied to
340 * modifications of the page flag must be done on small or head pages,
341 * checks can be done on tail pages too.
344 * the page flag is not relevant for compound pages.
347 * the page flag is stored in the first tail page.
349 #define PF_POISONED_CHECK(page) ({ \
350 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
352 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
353 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
354 #define PF_NO_TAIL(page, enforce) ({ \
355 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
356 PF_POISONED_CHECK(compound_head(page)); })
357 #define PF_NO_COMPOUND(page, enforce) ({ \
358 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
359 PF_POISONED_CHECK(page); })
360 #define PF_SECOND(page, enforce) ({ \
361 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
362 PF_POISONED_CHECK(&page[1]); })
364 /* Which page is the flag stored in */
365 #define FOLIO_PF_ANY 0
366 #define FOLIO_PF_HEAD 0
367 #define FOLIO_PF_NO_TAIL 0
368 #define FOLIO_PF_NO_COMPOUND 0
369 #define FOLIO_PF_SECOND 1
371 #define FOLIO_HEAD_PAGE 0
372 #define FOLIO_SECOND_PAGE 1
375 * Macros to create function definitions for page flags
377 #define FOLIO_TEST_FLAG(name, page) \
378 static __always_inline bool folio_test_##name(const struct folio *folio) \
379 { return test_bit(PG_##name, const_folio_flags(folio, page)); }
381 #define FOLIO_SET_FLAG(name, page) \
382 static __always_inline void folio_set_##name(struct folio *folio) \
383 { set_bit(PG_##name, folio_flags(folio, page)); }
385 #define FOLIO_CLEAR_FLAG(name, page) \
386 static __always_inline void folio_clear_##name(struct folio *folio) \
387 { clear_bit(PG_##name, folio_flags(folio, page)); }
389 #define __FOLIO_SET_FLAG(name, page) \
390 static __always_inline void __folio_set_##name(struct folio *folio) \
391 { __set_bit(PG_##name, folio_flags(folio, page)); }
393 #define __FOLIO_CLEAR_FLAG(name, page) \
394 static __always_inline void __folio_clear_##name(struct folio *folio) \
395 { __clear_bit(PG_##name, folio_flags(folio, page)); }
397 #define FOLIO_TEST_SET_FLAG(name, page) \
398 static __always_inline bool folio_test_set_##name(struct folio *folio) \
399 { return test_and_set_bit(PG_##name, folio_flags(folio, page)); }
401 #define FOLIO_TEST_CLEAR_FLAG(name, page) \
402 static __always_inline bool folio_test_clear_##name(struct folio *folio) \
403 { return test_and_clear_bit(PG_##name, folio_flags(folio, page)); }
405 #define FOLIO_FLAG(name, page) \
406 FOLIO_TEST_FLAG(name, page) \
407 FOLIO_SET_FLAG(name, page) \
408 FOLIO_CLEAR_FLAG(name, page)
410 #define TESTPAGEFLAG(uname, lname, policy) \
411 FOLIO_TEST_FLAG(lname, FOLIO_##policy) \
412 static __always_inline int Page##uname(const struct page *page) \
413 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
415 #define SETPAGEFLAG(uname, lname, policy) \
416 FOLIO_SET_FLAG(lname, FOLIO_##policy) \
417 static __always_inline void SetPage##uname(struct page *page) \
418 { set_bit(PG_##lname, &policy(page, 1)->flags); }
420 #define CLEARPAGEFLAG(uname, lname, policy) \
421 FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \
422 static __always_inline void ClearPage##uname(struct page *page) \
423 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
425 #define __SETPAGEFLAG(uname, lname, policy) \
426 __FOLIO_SET_FLAG(lname, 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 __FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \
432 static __always_inline void __ClearPage##uname(struct page *page) \
433 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
435 #define TESTSETFLAG(uname, lname, policy) \
436 FOLIO_TEST_SET_FLAG(lname, FOLIO_##policy) \
437 static __always_inline int TestSetPage##uname(struct page *page) \
438 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
440 #define TESTCLEARFLAG(uname, lname, policy) \
441 FOLIO_TEST_CLEAR_FLAG(lname, FOLIO_##policy) \
442 static __always_inline int TestClearPage##uname(struct page *page) \
443 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
445 #define PAGEFLAG(uname, lname, policy) \
446 TESTPAGEFLAG(uname, lname, policy) \
447 SETPAGEFLAG(uname, lname, policy) \
448 CLEARPAGEFLAG(uname, lname, policy)
450 #define __PAGEFLAG(uname, lname, policy) \
451 TESTPAGEFLAG(uname, lname, policy) \
452 __SETPAGEFLAG(uname, lname, policy) \
453 __CLEARPAGEFLAG(uname, lname, policy)
455 #define TESTSCFLAG(uname, lname, policy) \
456 TESTSETFLAG(uname, lname, policy) \
457 TESTCLEARFLAG(uname, lname, policy)
459 #define FOLIO_TEST_FLAG_FALSE(name) \
460 static inline bool folio_test_##name(const struct folio *folio) \
462 #define FOLIO_SET_FLAG_NOOP(name) \
463 static inline void folio_set_##name(struct folio *folio) { }
464 #define FOLIO_CLEAR_FLAG_NOOP(name) \
465 static inline void folio_clear_##name(struct folio *folio) { }
466 #define __FOLIO_SET_FLAG_NOOP(name) \
467 static inline void __folio_set_##name(struct folio *folio) { }
468 #define __FOLIO_CLEAR_FLAG_NOOP(name) \
469 static inline void __folio_clear_##name(struct folio *folio) { }
470 #define FOLIO_TEST_SET_FLAG_FALSE(name) \
471 static inline bool folio_test_set_##name(struct folio *folio) \
473 #define FOLIO_TEST_CLEAR_FLAG_FALSE(name) \
474 static inline bool folio_test_clear_##name(struct folio *folio) \
477 #define FOLIO_FLAG_FALSE(name) \
478 FOLIO_TEST_FLAG_FALSE(name) \
479 FOLIO_SET_FLAG_NOOP(name) \
480 FOLIO_CLEAR_FLAG_NOOP(name)
482 #define TESTPAGEFLAG_FALSE(uname, lname) \
483 FOLIO_TEST_FLAG_FALSE(lname) \
484 static inline int Page##uname(const struct page *page) { return 0; }
486 #define SETPAGEFLAG_NOOP(uname, lname) \
487 FOLIO_SET_FLAG_NOOP(lname) \
488 static inline void SetPage##uname(struct page *page) { }
490 #define CLEARPAGEFLAG_NOOP(uname, lname) \
491 FOLIO_CLEAR_FLAG_NOOP(lname) \
492 static inline void ClearPage##uname(struct page *page) { }
494 #define __CLEARPAGEFLAG_NOOP(uname, lname) \
495 __FOLIO_CLEAR_FLAG_NOOP(lname) \
496 static inline void __ClearPage##uname(struct page *page) { }
498 #define TESTSETFLAG_FALSE(uname, lname) \
499 FOLIO_TEST_SET_FLAG_FALSE(lname) \
500 static inline int TestSetPage##uname(struct page *page) { return 0; }
502 #define TESTCLEARFLAG_FALSE(uname, lname) \
503 FOLIO_TEST_CLEAR_FLAG_FALSE(lname) \
504 static inline int TestClearPage##uname(struct page *page) { return 0; }
506 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \
507 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
509 #define TESTSCFLAG_FALSE(uname, lname) \
510 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
512 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
513 FOLIO_FLAG(waiters, FOLIO_HEAD_PAGE)
514 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
515 FOLIO_FLAG(referenced, FOLIO_HEAD_PAGE)
516 FOLIO_TEST_CLEAR_FLAG(referenced, FOLIO_HEAD_PAGE)
517 __FOLIO_SET_FLAG(referenced, FOLIO_HEAD_PAGE)
518 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
519 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
520 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
521 TESTCLEARFLAG(LRU, lru, PF_HEAD)
522 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
523 TESTCLEARFLAG(Active, active, PF_HEAD)
524 PAGEFLAG(Workingset, workingset, PF_HEAD)
525 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
526 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
529 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
530 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
531 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
532 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
533 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
534 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
536 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
537 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
538 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
539 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
540 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
541 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
544 * Private page markings that may be used by the filesystem that owns the page
545 * for its own purposes.
546 * - PG_private and PG_private_2 cause release_folio() and co to be invoked
548 PAGEFLAG(Private, private, PF_ANY)
549 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
550 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
551 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
554 * Only test-and-set exist for PG_writeback. The unconditional operators are
555 * risky: they bypass page accounting.
557 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
558 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
559 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
561 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
562 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
563 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
564 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND)
565 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND)
567 #ifdef CONFIG_HIGHMEM
569 * Must use a macro here due to header dependency issues. page_zone() is not
570 * available at this point.
572 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
573 #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f))
575 PAGEFLAG_FALSE(HighMem, highmem)
579 static __always_inline bool folio_test_swapcache(const struct folio *folio)
581 return folio_test_swapbacked(folio) &&
582 test_bit(PG_swapcache, const_folio_flags(folio, 0));
585 static __always_inline bool PageSwapCache(const struct page *page)
587 return folio_test_swapcache(page_folio(page));
590 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
591 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
593 PAGEFLAG_FALSE(SwapCache, swapcache)
596 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
597 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
598 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
601 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
602 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
603 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
605 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked)
606 TESTSCFLAG_FALSE(Mlocked, mlocked)
609 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
610 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
612 PAGEFLAG_FALSE(Uncached, uncached)
615 #ifdef CONFIG_MEMORY_FAILURE
616 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
617 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
618 #define __PG_HWPOISON (1UL << PG_hwpoison)
619 #define MAGIC_HWPOISON 0x48575053U /* HWPS */
620 extern void SetPageHWPoisonTakenOff(struct page *page);
621 extern void ClearPageHWPoisonTakenOff(struct page *page);
622 extern bool take_page_off_buddy(struct page *page);
623 extern bool put_page_back_buddy(struct page *page);
625 PAGEFLAG_FALSE(HWPoison, hwpoison)
626 #define __PG_HWPOISON 0
629 #ifdef CONFIG_PAGE_IDLE_FLAG
631 FOLIO_TEST_FLAG(young, FOLIO_HEAD_PAGE)
632 FOLIO_SET_FLAG(young, FOLIO_HEAD_PAGE)
633 FOLIO_TEST_CLEAR_FLAG(young, FOLIO_HEAD_PAGE)
634 FOLIO_FLAG(idle, FOLIO_HEAD_PAGE)
636 /* See page_idle.h for !64BIT workaround */
637 #else /* !CONFIG_PAGE_IDLE_FLAG */
638 FOLIO_FLAG_FALSE(young)
639 FOLIO_TEST_CLEAR_FLAG_FALSE(young)
640 FOLIO_FLAG_FALSE(idle)
644 * PageReported() is used to track reported free pages within the Buddy
645 * allocator. We can use the non-atomic version of the test and set
646 * operations as both should be shielded with the zone lock to prevent
647 * any possible races on the setting or clearing of the bit.
649 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
651 #ifdef CONFIG_MEMORY_HOTPLUG
652 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
654 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
658 * On an anonymous page mapped into a user virtual memory area,
659 * page->mapping points to its anon_vma, not to a struct address_space;
660 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
662 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
663 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
664 * bit; and then page->mapping points, not to an anon_vma, but to a private
665 * structure which KSM associates with that merged page. See ksm.h.
667 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
668 * page and then page->mapping points to a struct movable_operations.
670 * Please note that, confusingly, "page_mapping" refers to the inode
671 * address_space which maps the page from disk; whereas "page_mapped"
672 * refers to user virtual address space into which the page is mapped.
674 * For slab pages, since slab reuses the bits in struct page to store its
675 * internal states, the page->mapping does not exist as such, nor do these
676 * flags below. So in order to avoid testing non-existent bits, please
677 * make sure that PageSlab(page) actually evaluates to false before calling
678 * the following functions (e.g., PageAnon). See mm/slab.h.
680 #define PAGE_MAPPING_ANON 0x1
681 #define PAGE_MAPPING_MOVABLE 0x2
682 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
683 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
686 * Different with flags above, this flag is used only for fsdax mode. It
687 * indicates that this page->mapping is now under reflink case.
689 #define PAGE_MAPPING_DAX_SHARED ((void *)0x1)
691 static __always_inline bool folio_mapping_flags(const struct folio *folio)
693 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0;
696 static __always_inline bool PageMappingFlags(const struct page *page)
698 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
701 static __always_inline bool folio_test_anon(const struct folio *folio)
703 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0;
706 static __always_inline bool PageAnon(const struct page *page)
708 return folio_test_anon(page_folio(page));
711 static __always_inline bool __folio_test_movable(const struct folio *folio)
713 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
714 PAGE_MAPPING_MOVABLE;
717 static __always_inline bool __PageMovable(const struct page *page)
719 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
720 PAGE_MAPPING_MOVABLE;
725 * A KSM page is one of those write-protected "shared pages" or "merged pages"
726 * which KSM maps into multiple mms, wherever identical anonymous page content
727 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
728 * anon_vma, but to that page's node of the stable tree.
730 static __always_inline bool folio_test_ksm(const struct folio *folio)
732 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
736 static __always_inline bool PageKsm(const struct page *page)
738 return folio_test_ksm(page_folio(page));
741 TESTPAGEFLAG_FALSE(Ksm, ksm)
744 u64 stable_page_flags(const struct page *page);
747 * folio_xor_flags_has_waiters - Change some folio flags.
749 * @mask: Bits set in this word will be changed.
751 * This must only be used for flags which are changed with the folio
752 * lock held. For example, it is unsafe to use for PG_dirty as that
753 * can be set without the folio lock held. It can also only be used
754 * on flags which are in the range 0-6 as some of the implementations
755 * only affect those bits.
757 * Return: Whether there are tasks waiting on the folio.
759 static inline bool folio_xor_flags_has_waiters(struct folio *folio,
762 return xor_unlock_is_negative_byte(mask, folio_flags(folio, 0));
766 * folio_test_uptodate - Is this folio up to date?
769 * The uptodate flag is set on a folio when every byte in the folio is
770 * at least as new as the corresponding bytes on storage. Anonymous
771 * and CoW folios are always uptodate. If the folio is not uptodate,
772 * some of the bytes in it may be; see the is_partially_uptodate()
773 * address_space operation.
775 static inline bool folio_test_uptodate(const struct folio *folio)
777 bool ret = test_bit(PG_uptodate, const_folio_flags(folio, 0));
779 * Must ensure that the data we read out of the folio is loaded
780 * _after_ we've loaded folio->flags to check the uptodate bit.
781 * We can skip the barrier if the folio is not uptodate, because
782 * we wouldn't be reading anything from it.
784 * See folio_mark_uptodate() for the other side of the story.
792 static inline bool PageUptodate(const struct page *page)
794 return folio_test_uptodate(page_folio(page));
797 static __always_inline void __folio_mark_uptodate(struct folio *folio)
800 __set_bit(PG_uptodate, folio_flags(folio, 0));
803 static __always_inline void folio_mark_uptodate(struct folio *folio)
806 * Memory barrier must be issued before setting the PG_uptodate bit,
807 * so that all previous stores issued in order to bring the folio
808 * uptodate are actually visible before folio_test_uptodate becomes true.
811 set_bit(PG_uptodate, folio_flags(folio, 0));
814 static __always_inline void __SetPageUptodate(struct page *page)
816 __folio_mark_uptodate((struct folio *)page);
819 static __always_inline void SetPageUptodate(struct page *page)
821 folio_mark_uptodate((struct folio *)page);
824 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
826 void __folio_start_writeback(struct folio *folio, bool keep_write);
827 void set_page_writeback(struct page *page);
829 #define folio_start_writeback(folio) \
830 __folio_start_writeback(folio, false)
831 #define folio_start_writeback_keepwrite(folio) \
832 __folio_start_writeback(folio, true)
834 static __always_inline bool folio_test_head(const struct folio *folio)
836 return test_bit(PG_head, const_folio_flags(folio, FOLIO_PF_ANY));
839 static __always_inline int PageHead(const struct page *page)
841 PF_POISONED_CHECK(page);
842 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page);
845 __SETPAGEFLAG(Head, head, PF_ANY)
846 __CLEARPAGEFLAG(Head, head, PF_ANY)
847 CLEARPAGEFLAG(Head, head, PF_ANY)
850 * folio_test_large() - Does this folio contain more than one page?
851 * @folio: The folio to test.
853 * Return: True if the folio is larger than one page.
855 static inline bool folio_test_large(const struct folio *folio)
857 return folio_test_head(folio);
860 static __always_inline void set_compound_head(struct page *page, struct page *head)
862 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
865 static __always_inline void clear_compound_head(struct page *page)
867 WRITE_ONCE(page->compound_head, 0);
870 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
871 static inline void ClearPageCompound(struct page *page)
873 BUG_ON(!PageHead(page));
876 FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE)
878 FOLIO_FLAG_FALSE(large_rmappable)
881 #define PG_head_mask ((1UL << PG_head))
883 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
885 * PageHuge() only returns true for hugetlbfs pages, but not for
886 * normal or transparent huge pages.
888 * PageTransHuge() returns true for both transparent huge and
889 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
890 * called only in the core VM paths where hugetlbfs pages can't exist.
892 static inline int PageTransHuge(const struct page *page)
894 VM_BUG_ON_PAGE(PageTail(page), page);
895 return PageHead(page);
899 * PageTransCompound returns true for both transparent huge pages
900 * and hugetlbfs pages, so it should only be called when it's known
901 * that hugetlbfs pages aren't involved.
903 static inline int PageTransCompound(const struct page *page)
905 return PageCompound(page);
909 * PageTransTail returns true for both transparent huge pages
910 * and hugetlbfs pages, so it should only be called when it's known
911 * that hugetlbfs pages aren't involved.
913 static inline int PageTransTail(const struct page *page)
915 return PageTail(page);
918 TESTPAGEFLAG_FALSE(TransHuge, transhuge)
919 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
920 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap)
921 TESTPAGEFLAG_FALSE(TransTail, transtail)
924 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
926 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
929 * This flag is set by hwpoison handler. Cleared by THP split or free page.
931 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
932 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
934 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
935 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
939 * For pages that are never mapped to userspace,
940 * page_type may be used. Because it is initialised to -1, we invert the
941 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
942 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
943 * low bits so that an underflow or overflow of _mapcount won't be
944 * mistaken for a page type value.
947 #define PAGE_TYPE_BASE 0xf0000000
948 /* Reserve 0x0000007f to catch underflows of _mapcount */
949 #define PAGE_MAPCOUNT_RESERVE -128
950 #define PG_buddy 0x00000080
951 #define PG_offline 0x00000100
952 #define PG_table 0x00000200
953 #define PG_guard 0x00000400
954 #define PG_hugetlb 0x00000800
955 #define PG_slab 0x00001000
957 #define PageType(page, flag) \
958 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
959 #define folio_test_type(folio, flag) \
960 ((folio->page.page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
962 static inline int page_type_has_type(unsigned int page_type)
964 return (int)page_type < PAGE_MAPCOUNT_RESERVE;
967 static inline int page_has_type(const struct page *page)
969 return page_type_has_type(page->page_type);
972 #define FOLIO_TYPE_OPS(lname, fname) \
973 static __always_inline bool folio_test_##fname(const struct folio *folio)\
975 return folio_test_type(folio, PG_##lname); \
977 static __always_inline void __folio_set_##fname(struct folio *folio) \
979 VM_BUG_ON_FOLIO(!folio_test_type(folio, 0), folio); \
980 folio->page.page_type &= ~PG_##lname; \
982 static __always_inline void __folio_clear_##fname(struct folio *folio) \
984 VM_BUG_ON_FOLIO(!folio_test_##fname(folio), folio); \
985 folio->page.page_type |= PG_##lname; \
988 #define PAGE_TYPE_OPS(uname, lname, fname) \
989 FOLIO_TYPE_OPS(lname, fname) \
990 static __always_inline int Page##uname(const struct page *page) \
992 return PageType(page, PG_##lname); \
994 static __always_inline void __SetPage##uname(struct page *page) \
996 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
997 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; \
1006 * PageBuddy() indicates that the page is free and in the buddy system
1007 * (see mm/page_alloc.c).
1009 PAGE_TYPE_OPS(Buddy, buddy, buddy)
1012 * PageOffline() indicates that the page is logically offline although the
1013 * containing section is online. (e.g. inflated in a balloon driver or
1014 * not onlined when onlining the section).
1015 * The content of these pages is effectively stale. Such pages should not
1016 * be touched (read/write/dump/save) except by their owner.
1018 * If a driver wants to allow to offline unmovable PageOffline() pages without
1019 * putting them back to the buddy, it can do so via the memory notifier by
1020 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
1021 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
1022 * pages (now with a reference count of zero) are treated like free pages,
1023 * allowing the containing memory block to get offlined. A driver that
1024 * relies on this feature is aware that re-onlining the memory block will
1025 * require to re-set the pages PageOffline() and not giving them to the
1026 * buddy via online_page_callback_t.
1028 * There are drivers that mark a page PageOffline() and expect there won't be
1029 * any further access to page content. PFN walkers that read content of random
1030 * pages should check PageOffline() and synchronize with such drivers using
1031 * page_offline_freeze()/page_offline_thaw().
1033 PAGE_TYPE_OPS(Offline, offline, offline)
1035 extern void page_offline_freeze(void);
1036 extern void page_offline_thaw(void);
1037 extern void page_offline_begin(void);
1038 extern void page_offline_end(void);
1041 * Marks pages in use as page tables.
1043 PAGE_TYPE_OPS(Table, table, pgtable)
1046 * Marks guardpages used with debug_pagealloc.
1048 PAGE_TYPE_OPS(Guard, guard, guard)
1050 FOLIO_TYPE_OPS(slab, slab)
1053 * PageSlab - Determine if the page belongs to the slab allocator
1054 * @page: The page to test.
1056 * Context: Any context.
1057 * Return: True for slab pages, false for any other kind of page.
1059 static inline bool PageSlab(const struct page *page)
1061 return folio_test_slab(page_folio(page));
1064 #ifdef CONFIG_HUGETLB_PAGE
1065 FOLIO_TYPE_OPS(hugetlb, hugetlb)
1067 FOLIO_TEST_FLAG_FALSE(hugetlb)
1071 * PageHuge - Determine if the page belongs to hugetlbfs
1072 * @page: The page to test.
1074 * Context: Any context.
1075 * Return: True for hugetlbfs pages, false for anon pages or pages
1076 * belonging to other filesystems.
1078 static inline bool PageHuge(const struct page *page)
1080 return folio_test_hugetlb(page_folio(page));
1084 * Check if a page is currently marked HWPoisoned. Note that this check is
1085 * best effort only and inherently racy: there is no way to synchronize with
1088 static inline bool is_page_hwpoison(const struct page *page)
1090 const struct folio *folio;
1092 if (PageHWPoison(page))
1094 folio = page_folio(page);
1095 return folio_test_hugetlb(folio) && PageHWPoison(&folio->page);
1098 bool is_free_buddy_page(const struct page *page);
1100 PAGEFLAG(Isolated, isolated, PF_ANY);
1102 static __always_inline int PageAnonExclusive(const struct page *page)
1104 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1106 * HugeTLB stores this information on the head page; THP keeps it per
1110 page = compound_head(page);
1111 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1114 static __always_inline void SetPageAnonExclusive(struct page *page)
1116 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1117 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1118 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1121 static __always_inline void ClearPageAnonExclusive(struct page *page)
1123 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1124 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1125 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1128 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1130 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1131 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1132 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1136 #define __PG_MLOCKED (1UL << PG_mlocked)
1138 #define __PG_MLOCKED 0
1142 * Flags checked when a page is freed. Pages being freed should not have
1143 * these flags set. If they are, there is a problem.
1145 #define PAGE_FLAGS_CHECK_AT_FREE \
1146 (1UL << PG_lru | 1UL << PG_locked | \
1147 1UL << PG_private | 1UL << PG_private_2 | \
1148 1UL << PG_writeback | 1UL << PG_reserved | \
1149 1UL << PG_active | \
1150 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK)
1153 * Flags checked when a page is prepped for return by the page allocator.
1154 * Pages being prepped should not have these flags set. If they are set,
1155 * there has been a kernel bug or struct page corruption.
1157 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1158 * alloc-free cycle to prevent from reusing the page.
1160 #define PAGE_FLAGS_CHECK_AT_PREP \
1161 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK)
1164 * Flags stored in the second page of a compound page. They may overlap
1165 * the CHECK_AT_FREE flags above, so need to be cleared.
1167 #define PAGE_FLAGS_SECOND \
1168 (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \
1169 1UL << PG_large_rmappable)
1171 #define PAGE_FLAGS_PRIVATE \
1172 (1UL << PG_private | 1UL << PG_private_2)
1174 * page_has_private - Determine if page has private stuff
1175 * @page: The page to be checked
1177 * Determine if a page has private stuff, indicating that release routines
1178 * should be invoked upon it.
1180 static inline int page_has_private(const struct page *page)
1182 return !!(page->flags & PAGE_FLAGS_PRIVATE);
1185 static inline bool folio_has_private(const struct folio *folio)
1187 return page_has_private(&folio->page);
1193 #undef PF_NO_COMPOUND
1195 #endif /* !__GENERATING_BOUNDS_H */
1197 #endif /* PAGE_FLAGS_H */