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_uptodate tells whether the page's contents is valid. When a read
72 * completes, the page becomes uptodate, unless a disk I/O error happened.
74 * PG_referenced, PG_reclaim are used for page reclaim for anonymous and
75 * file-backed pagecache (see mm/vmscan.c).
77 * PG_error is set to indicate that an I/O error occurred on this page.
79 * PG_arch_1 is an architecture specific page state bit. The generic code
80 * guarantees that this bit is cleared for a page when it first is entered into
83 * PG_hwpoison indicates that a page got corrupted in hardware and contains
84 * data with incorrect ECC bits that triggered a machine check. Accessing is
85 * not safe since it may cause another machine check. Don't touch!
89 * Don't use the *_dontuse flags. Use the macros. Otherwise you'll break
90 * locked- and dirty-page accounting.
92 * The page flags field is split into two parts, the main flags area
93 * which extends from the low bits upwards, and the fields area which
94 * extends from the high bits downwards.
96 * | FIELD | ... | FLAGS |
100 * The fields area is reserved for fields mapping zone, node (for NUMA) and
101 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like
102 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP).
105 PG_locked, /* Page is locked. Don't touch. */
112 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
115 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
118 PG_private, /* If pagecache, has fs-private data */
119 PG_private_2, /* If pagecache, has fs aux data */
120 PG_writeback, /* Page is under writeback */
121 PG_head, /* A head page */
122 PG_mappedtodisk, /* Has blocks allocated on-disk */
123 PG_reclaim, /* To be reclaimed asap */
124 PG_swapbacked, /* Page is backed by RAM/swap */
125 PG_unevictable, /* Page is "unevictable" */
127 PG_mlocked, /* Page is vma mlocked */
129 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
130 PG_uncached, /* Page has been mapped as uncached */
132 #ifdef CONFIG_MEMORY_FAILURE
133 PG_hwpoison, /* hardware poisoned page. Don't touch */
135 #if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT)
145 PG_checked = PG_owner_priv_1,
148 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
150 /* Two page bits are conscripted by FS-Cache to maintain local caching
151 * state. These bits are set on pages belonging to the netfs's inodes
152 * when those inodes are being locally cached.
154 PG_fscache = PG_private_2, /* page backed by cache */
157 /* Pinned in Xen as a read-only pagetable page. */
158 PG_pinned = PG_owner_priv_1,
159 /* Pinned as part of domain save (see xen_mm_pin_all()). */
160 PG_savepinned = PG_dirty,
161 /* Has a grant mapping of another (foreign) domain's page. */
162 PG_foreign = PG_owner_priv_1,
163 /* Remapped by swiotlb-xen. */
164 PG_xen_remapped = PG_owner_priv_1,
167 PG_slob_free = PG_private,
169 /* Compound pages. Stored in first tail page's flags */
170 PG_double_map = PG_workingset,
172 /* non-lru isolated movable page */
173 PG_isolated = PG_reclaim,
175 /* Only valid for buddy pages. Used to track pages that are reported */
176 PG_reported = PG_uptodate,
179 #ifndef __GENERATING_BOUNDS_H
181 struct page; /* forward declaration */
183 static inline struct page *compound_head(struct page *page)
185 unsigned long head = READ_ONCE(page->compound_head);
187 if (unlikely(head & 1))
188 return (struct page *) (head - 1);
192 static __always_inline int PageTail(struct page *page)
194 return READ_ONCE(page->compound_head) & 1;
197 static __always_inline int PageCompound(struct page *page)
199 return test_bit(PG_head, &page->flags) || PageTail(page);
202 #define PAGE_POISON_PATTERN -1l
203 static inline int PagePoisoned(const struct page *page)
205 return page->flags == PAGE_POISON_PATTERN;
208 #ifdef CONFIG_DEBUG_VM
209 void page_init_poison(struct page *page, size_t size);
211 static inline void page_init_poison(struct page *page, size_t size)
217 * Page flags policies wrt compound pages
220 * check if this struct page poisoned/uninitialized
223 * the page flag is relevant for small, head and tail pages.
226 * for compound page all operations related to the page flag applied to
230 * for compound page, callers only ever operate on the head page.
233 * modifications of the page flag must be done on small or head pages,
234 * checks can be done on tail pages too.
237 * the page flag is not relevant for compound pages.
240 * the page flag is stored in the first tail page.
242 #define PF_POISONED_CHECK(page) ({ \
243 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
245 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
246 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
247 #define PF_ONLY_HEAD(page, enforce) ({ \
248 VM_BUG_ON_PGFLAGS(PageTail(page), page); \
249 PF_POISONED_CHECK(page); })
250 #define PF_NO_TAIL(page, enforce) ({ \
251 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
252 PF_POISONED_CHECK(compound_head(page)); })
253 #define PF_NO_COMPOUND(page, enforce) ({ \
254 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
255 PF_POISONED_CHECK(page); })
256 #define PF_SECOND(page, enforce) ({ \
257 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
258 PF_POISONED_CHECK(&page[1]); })
261 * Macros to create function definitions for page flags
263 #define TESTPAGEFLAG(uname, lname, policy) \
264 static __always_inline int Page##uname(struct page *page) \
265 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
267 #define SETPAGEFLAG(uname, lname, policy) \
268 static __always_inline void SetPage##uname(struct page *page) \
269 { set_bit(PG_##lname, &policy(page, 1)->flags); }
271 #define CLEARPAGEFLAG(uname, lname, policy) \
272 static __always_inline void ClearPage##uname(struct page *page) \
273 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
275 #define __SETPAGEFLAG(uname, lname, policy) \
276 static __always_inline void __SetPage##uname(struct page *page) \
277 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
279 #define __CLEARPAGEFLAG(uname, lname, policy) \
280 static __always_inline void __ClearPage##uname(struct page *page) \
281 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
283 #define TESTSETFLAG(uname, lname, policy) \
284 static __always_inline int TestSetPage##uname(struct page *page) \
285 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
287 #define TESTCLEARFLAG(uname, lname, policy) \
288 static __always_inline int TestClearPage##uname(struct page *page) \
289 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
291 #define PAGEFLAG(uname, lname, policy) \
292 TESTPAGEFLAG(uname, lname, policy) \
293 SETPAGEFLAG(uname, lname, policy) \
294 CLEARPAGEFLAG(uname, lname, policy)
296 #define __PAGEFLAG(uname, lname, policy) \
297 TESTPAGEFLAG(uname, lname, policy) \
298 __SETPAGEFLAG(uname, lname, policy) \
299 __CLEARPAGEFLAG(uname, lname, policy)
301 #define TESTSCFLAG(uname, lname, policy) \
302 TESTSETFLAG(uname, lname, policy) \
303 TESTCLEARFLAG(uname, lname, policy)
305 #define TESTPAGEFLAG_FALSE(uname) \
306 static inline int Page##uname(const struct page *page) { return 0; }
308 #define SETPAGEFLAG_NOOP(uname) \
309 static inline void SetPage##uname(struct page *page) { }
311 #define CLEARPAGEFLAG_NOOP(uname) \
312 static inline void ClearPage##uname(struct page *page) { }
314 #define __CLEARPAGEFLAG_NOOP(uname) \
315 static inline void __ClearPage##uname(struct page *page) { }
317 #define TESTSETFLAG_FALSE(uname) \
318 static inline int TestSetPage##uname(struct page *page) { return 0; }
320 #define TESTCLEARFLAG_FALSE(uname) \
321 static inline int TestClearPage##uname(struct page *page) { return 0; }
323 #define PAGEFLAG_FALSE(uname) TESTPAGEFLAG_FALSE(uname) \
324 SETPAGEFLAG_NOOP(uname) CLEARPAGEFLAG_NOOP(uname)
326 #define TESTSCFLAG_FALSE(uname) \
327 TESTSETFLAG_FALSE(uname) TESTCLEARFLAG_FALSE(uname)
329 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
330 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) __CLEARPAGEFLAG(Waiters, waiters, PF_ONLY_HEAD)
331 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
332 PAGEFLAG(Referenced, referenced, PF_HEAD)
333 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
334 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
335 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
336 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
337 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
338 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
339 TESTCLEARFLAG(Active, active, PF_HEAD)
340 PAGEFLAG(Workingset, workingset, PF_HEAD)
341 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
342 __PAGEFLAG(Slab, slab, PF_NO_TAIL)
343 __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL)
344 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
347 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
348 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
349 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
350 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
351 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
352 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
354 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
355 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
356 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
357 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
358 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
359 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
362 * Private page markings that may be used by the filesystem that owns the page
363 * for its own purposes.
364 * - PG_private and PG_private_2 cause releasepage() and co to be invoked
366 PAGEFLAG(Private, private, PF_ANY) __SETPAGEFLAG(Private, private, PF_ANY)
367 __CLEARPAGEFLAG(Private, private, PF_ANY)
368 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
369 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
370 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
373 * Only test-and-set exist for PG_writeback. The unconditional operators are
374 * risky: they bypass page accounting.
376 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
377 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
378 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
380 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
381 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
382 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
383 PAGEFLAG(Readahead, reclaim, PF_NO_COMPOUND)
384 TESTCLEARFLAG(Readahead, reclaim, PF_NO_COMPOUND)
386 #ifdef CONFIG_HIGHMEM
388 * Must use a macro here due to header dependency issues. page_zone() is not
389 * available at this point.
391 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
393 PAGEFLAG_FALSE(HighMem)
397 static __always_inline int PageSwapCache(struct page *page)
399 #ifdef CONFIG_THP_SWAP
400 page = compound_head(page);
402 return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags);
405 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
406 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
408 PAGEFLAG_FALSE(SwapCache)
411 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
412 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
413 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
416 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
417 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
418 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
420 PAGEFLAG_FALSE(Mlocked) __CLEARPAGEFLAG_NOOP(Mlocked)
421 TESTSCFLAG_FALSE(Mlocked)
424 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
425 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
427 PAGEFLAG_FALSE(Uncached)
430 #ifdef CONFIG_MEMORY_FAILURE
431 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
432 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
433 #define __PG_HWPOISON (1UL << PG_hwpoison)
434 extern bool set_hwpoison_free_buddy_page(struct page *page);
436 PAGEFLAG_FALSE(HWPoison)
437 static inline bool set_hwpoison_free_buddy_page(struct page *page)
441 #define __PG_HWPOISON 0
444 #if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT)
445 TESTPAGEFLAG(Young, young, PF_ANY)
446 SETPAGEFLAG(Young, young, PF_ANY)
447 TESTCLEARFLAG(Young, young, PF_ANY)
448 PAGEFLAG(Idle, idle, PF_ANY)
452 * PageReported() is used to track reported free pages within the Buddy
453 * allocator. We can use the non-atomic version of the test and set
454 * operations as both should be shielded with the zone lock to prevent
455 * any possible races on the setting or clearing of the bit.
457 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
460 * On an anonymous page mapped into a user virtual memory area,
461 * page->mapping points to its anon_vma, not to a struct address_space;
462 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
464 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
465 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
466 * bit; and then page->mapping points, not to an anon_vma, but to a private
467 * structure which KSM associates with that merged page. See ksm.h.
469 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
470 * page and then page->mapping points a struct address_space.
472 * Please note that, confusingly, "page_mapping" refers to the inode
473 * address_space which maps the page from disk; whereas "page_mapped"
474 * refers to user virtual address space into which the page is mapped.
476 #define PAGE_MAPPING_ANON 0x1
477 #define PAGE_MAPPING_MOVABLE 0x2
478 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
479 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
481 static __always_inline int PageMappingFlags(struct page *page)
483 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
486 static __always_inline int PageAnon(struct page *page)
488 page = compound_head(page);
489 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
492 static __always_inline int __PageMovable(struct page *page)
494 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
495 PAGE_MAPPING_MOVABLE;
500 * A KSM page is one of those write-protected "shared pages" or "merged pages"
501 * which KSM maps into multiple mms, wherever identical anonymous page content
502 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
503 * anon_vma, but to that page's node of the stable tree.
505 static __always_inline int PageKsm(struct page *page)
507 page = compound_head(page);
508 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
512 TESTPAGEFLAG_FALSE(Ksm)
515 u64 stable_page_flags(struct page *page);
517 static inline int PageUptodate(struct page *page)
520 page = compound_head(page);
521 ret = test_bit(PG_uptodate, &(page)->flags);
523 * Must ensure that the data we read out of the page is loaded
524 * _after_ we've loaded page->flags to check for PageUptodate.
525 * We can skip the barrier if the page is not uptodate, because
526 * we wouldn't be reading anything from it.
528 * See SetPageUptodate() for the other side of the story.
536 static __always_inline void __SetPageUptodate(struct page *page)
538 VM_BUG_ON_PAGE(PageTail(page), page);
540 __set_bit(PG_uptodate, &page->flags);
543 static __always_inline void SetPageUptodate(struct page *page)
545 VM_BUG_ON_PAGE(PageTail(page), page);
547 * Memory barrier must be issued before setting the PG_uptodate bit,
548 * so that all previous stores issued in order to bring the page
549 * uptodate are actually visible before PageUptodate becomes true.
552 set_bit(PG_uptodate, &page->flags);
555 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
557 int test_clear_page_writeback(struct page *page);
558 int __test_set_page_writeback(struct page *page, bool keep_write);
560 #define test_set_page_writeback(page) \
561 __test_set_page_writeback(page, false)
562 #define test_set_page_writeback_keepwrite(page) \
563 __test_set_page_writeback(page, true)
565 static inline void set_page_writeback(struct page *page)
567 test_set_page_writeback(page);
570 static inline void set_page_writeback_keepwrite(struct page *page)
572 test_set_page_writeback_keepwrite(page);
575 __PAGEFLAG(Head, head, PF_ANY) CLEARPAGEFLAG(Head, head, PF_ANY)
577 static __always_inline void set_compound_head(struct page *page, struct page *head)
579 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
582 static __always_inline void clear_compound_head(struct page *page)
584 WRITE_ONCE(page->compound_head, 0);
587 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
588 static inline void ClearPageCompound(struct page *page)
590 BUG_ON(!PageHead(page));
595 #define PG_head_mask ((1UL << PG_head))
597 #ifdef CONFIG_HUGETLB_PAGE
598 int PageHuge(struct page *page);
599 int PageHeadHuge(struct page *page);
600 bool page_huge_active(struct page *page);
602 TESTPAGEFLAG_FALSE(Huge)
603 TESTPAGEFLAG_FALSE(HeadHuge)
605 static inline bool page_huge_active(struct page *page)
612 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
614 * PageHuge() only returns true for hugetlbfs pages, but not for
615 * normal or transparent huge pages.
617 * PageTransHuge() returns true for both transparent huge and
618 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
619 * called only in the core VM paths where hugetlbfs pages can't exist.
621 static inline int PageTransHuge(struct page *page)
623 VM_BUG_ON_PAGE(PageTail(page), page);
624 return PageHead(page);
628 * PageTransCompound returns true for both transparent huge pages
629 * and hugetlbfs pages, so it should only be called when it's known
630 * that hugetlbfs pages aren't involved.
632 static inline int PageTransCompound(struct page *page)
634 return PageCompound(page);
638 * PageTransCompoundMap is the same as PageTransCompound, but it also
639 * guarantees the primary MMU has the entire compound page mapped
640 * through pmd_trans_huge, which in turn guarantees the secondary MMUs
641 * can also map the entire compound page. This allows the secondary
642 * MMUs to call get_user_pages() only once for each compound page and
643 * to immediately map the entire compound page with a single secondary
644 * MMU fault. If there will be a pmd split later, the secondary MMUs
645 * will get an update through the MMU notifier invalidation through
648 * Unlike PageTransCompound, this is safe to be called only while
649 * split_huge_pmd() cannot run from under us, like if protected by the
650 * MMU notifier, otherwise it may result in page->_mapcount check false
653 * We have to treat page cache THP differently since every subpage of it
654 * would get _mapcount inc'ed once it is PMD mapped. But, it may be PTE
655 * mapped in the current process so comparing subpage's _mapcount to
656 * compound_mapcount to filter out PTE mapped case.
658 static inline int PageTransCompoundMap(struct page *page)
662 if (!PageTransCompound(page))
666 return atomic_read(&page->_mapcount) < 0;
668 head = compound_head(page);
669 /* File THP is PMD mapped and not PTE mapped */
670 return atomic_read(&page->_mapcount) ==
671 atomic_read(compound_mapcount_ptr(head));
675 * PageTransTail returns true for both transparent huge pages
676 * and hugetlbfs pages, so it should only be called when it's known
677 * that hugetlbfs pages aren't involved.
679 static inline int PageTransTail(struct page *page)
681 return PageTail(page);
685 * PageDoubleMap indicates that the compound page is mapped with PTEs as well
688 * This is required for optimization of rmap operations for THP: we can postpone
689 * per small page mapcount accounting (and its overhead from atomic operations)
690 * until the first PMD split.
692 * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up
693 * by one. This reference will go away with last compound_mapcount.
695 * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap().
697 PAGEFLAG(DoubleMap, double_map, PF_SECOND)
698 TESTSCFLAG(DoubleMap, double_map, PF_SECOND)
700 TESTPAGEFLAG_FALSE(TransHuge)
701 TESTPAGEFLAG_FALSE(TransCompound)
702 TESTPAGEFLAG_FALSE(TransCompoundMap)
703 TESTPAGEFLAG_FALSE(TransTail)
704 PAGEFLAG_FALSE(DoubleMap)
705 TESTSCFLAG_FALSE(DoubleMap)
709 * For pages that are never mapped to userspace (and aren't PageSlab),
710 * page_type may be used. Because it is initialised to -1, we invert the
711 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
712 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
713 * low bits so that an underflow or overflow of page_mapcount() won't be
714 * mistaken for a page type value.
717 #define PAGE_TYPE_BASE 0xf0000000
718 /* Reserve 0x0000007f to catch underflows of page_mapcount */
719 #define PAGE_MAPCOUNT_RESERVE -128
720 #define PG_buddy 0x00000080
721 #define PG_offline 0x00000100
722 #define PG_kmemcg 0x00000200
723 #define PG_table 0x00000400
724 #define PG_guard 0x00000800
726 #define PageType(page, flag) \
727 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
729 static inline int page_has_type(struct page *page)
731 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE;
734 #define PAGE_TYPE_OPS(uname, lname) \
735 static __always_inline int Page##uname(struct page *page) \
737 return PageType(page, PG_##lname); \
739 static __always_inline void __SetPage##uname(struct page *page) \
741 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
742 page->page_type &= ~PG_##lname; \
744 static __always_inline void __ClearPage##uname(struct page *page) \
746 VM_BUG_ON_PAGE(!Page##uname(page), page); \
747 page->page_type |= PG_##lname; \
751 * PageBuddy() indicates that the page is free and in the buddy system
752 * (see mm/page_alloc.c).
754 PAGE_TYPE_OPS(Buddy, buddy)
757 * PageOffline() indicates that the page is logically offline although the
758 * containing section is online. (e.g. inflated in a balloon driver or
759 * not onlined when onlining the section).
760 * The content of these pages is effectively stale. Such pages should not
761 * be touched (read/write/dump/save) except by their owner.
763 * If a driver wants to allow to offline unmovable PageOffline() pages without
764 * putting them back to the buddy, it can do so via the memory notifier by
765 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
766 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
767 * pages (now with a reference count of zero) are treated like free pages,
768 * allowing the containing memory block to get offlined. A driver that
769 * relies on this feature is aware that re-onlining the memory block will
770 * require to re-set the pages PageOffline() and not giving them to the
771 * buddy via online_page_callback_t.
773 PAGE_TYPE_OPS(Offline, offline)
776 * If kmemcg is enabled, the buddy allocator will set PageKmemcg() on
777 * pages allocated with __GFP_ACCOUNT. It gets cleared on page free.
779 PAGE_TYPE_OPS(Kmemcg, kmemcg)
782 * Marks pages in use as page tables.
784 PAGE_TYPE_OPS(Table, table)
787 * Marks guardpages used with debug_pagealloc.
789 PAGE_TYPE_OPS(Guard, guard)
791 extern bool is_free_buddy_page(struct page *page);
793 __PAGEFLAG(Isolated, isolated, PF_ANY);
796 * If network-based swap is enabled, sl*b must keep track of whether pages
797 * were allocated from pfmemalloc reserves.
799 static inline int PageSlabPfmemalloc(struct page *page)
801 VM_BUG_ON_PAGE(!PageSlab(page), page);
802 return PageActive(page);
805 static inline void SetPageSlabPfmemalloc(struct page *page)
807 VM_BUG_ON_PAGE(!PageSlab(page), page);
811 static inline void __ClearPageSlabPfmemalloc(struct page *page)
813 VM_BUG_ON_PAGE(!PageSlab(page), page);
814 __ClearPageActive(page);
817 static inline void ClearPageSlabPfmemalloc(struct page *page)
819 VM_BUG_ON_PAGE(!PageSlab(page), page);
820 ClearPageActive(page);
824 #define __PG_MLOCKED (1UL << PG_mlocked)
826 #define __PG_MLOCKED 0
830 * Flags checked when a page is freed. Pages being freed should not have
831 * these flags set. It they are, there is a problem.
833 #define PAGE_FLAGS_CHECK_AT_FREE \
834 (1UL << PG_lru | 1UL << PG_locked | \
835 1UL << PG_private | 1UL << PG_private_2 | \
836 1UL << PG_writeback | 1UL << PG_reserved | \
837 1UL << PG_slab | 1UL << PG_active | \
838 1UL << PG_unevictable | __PG_MLOCKED)
841 * Flags checked when a page is prepped for return by the page allocator.
842 * Pages being prepped should not have these flags set. It they are set,
843 * there has been a kernel bug or struct page corruption.
845 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
846 * alloc-free cycle to prevent from reusing the page.
848 #define PAGE_FLAGS_CHECK_AT_PREP \
849 (((1UL << NR_PAGEFLAGS) - 1) & ~__PG_HWPOISON)
851 #define PAGE_FLAGS_PRIVATE \
852 (1UL << PG_private | 1UL << PG_private_2)
854 * page_has_private - Determine if page has private stuff
855 * @page: The page to be checked
857 * Determine if a page has private stuff, indicating that release routines
858 * should be invoked upon it.
860 static inline int page_has_private(struct page *page)
862 return !!(page->flags & PAGE_FLAGS_PRIVATE);
869 #undef PF_NO_COMPOUND
871 #endif /* !__GENERATING_BOUNDS_H */
873 #endif /* PAGE_FLAGS_H */