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 pageflags directly. Use the PageFoo macros.
91 * The page flags field is split into two parts, the main flags area
92 * which extends from the low bits upwards, and the fields area which
93 * extends from the high bits downwards.
95 * | FIELD | ... | FLAGS |
99 * The fields area is reserved for fields mapping zone, node (for NUMA) and
100 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like
101 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP).
104 PG_locked, /* Page is locked. Don't touch. */
111 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
114 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
117 PG_private, /* If pagecache, has fs-private data */
118 PG_private_2, /* If pagecache, has fs aux data */
119 PG_writeback, /* Page is under writeback */
120 PG_head, /* A head page */
121 PG_mappedtodisk, /* Has blocks allocated on-disk */
122 PG_reclaim, /* To be reclaimed asap */
123 PG_swapbacked, /* Page is backed by RAM/swap */
124 PG_unevictable, /* Page is "unevictable" */
126 PG_mlocked, /* Page is vma mlocked */
128 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
129 PG_uncached, /* Page has been mapped as uncached */
131 #ifdef CONFIG_MEMORY_FAILURE
132 PG_hwpoison, /* hardware poisoned page. Don't touch */
134 #if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT)
141 #ifdef CONFIG_KASAN_HW_TAGS
142 PG_skip_kasan_poison,
147 PG_checked = PG_owner_priv_1,
150 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
152 /* Two page bits are conscripted by FS-Cache to maintain local caching
153 * state. These bits are set on pages belonging to the netfs's inodes
154 * when those inodes are being locally cached.
156 PG_fscache = PG_private_2, /* page backed by cache */
159 /* Pinned in Xen as a read-only pagetable page. */
160 PG_pinned = PG_owner_priv_1,
161 /* Pinned as part of domain save (see xen_mm_pin_all()). */
162 PG_savepinned = PG_dirty,
163 /* Has a grant mapping of another (foreign) domain's page. */
164 PG_foreign = PG_owner_priv_1,
165 /* Remapped by swiotlb-xen. */
166 PG_xen_remapped = PG_owner_priv_1,
169 PG_slob_free = PG_private,
171 /* Compound pages. Stored in first tail page's flags */
172 PG_double_map = PG_workingset,
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,
181 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
183 #ifndef __GENERATING_BOUNDS_H
185 static inline unsigned long _compound_head(const struct page *page)
187 unsigned long head = READ_ONCE(page->compound_head);
189 if (unlikely(head & 1))
191 return (unsigned long)page;
194 #define compound_head(page) ((typeof(page))_compound_head(page))
196 static __always_inline int PageTail(struct page *page)
198 return READ_ONCE(page->compound_head) & 1;
201 static __always_inline int PageCompound(struct page *page)
203 return test_bit(PG_head, &page->flags) || PageTail(page);
206 #define PAGE_POISON_PATTERN -1l
207 static inline int PagePoisoned(const struct page *page)
209 return page->flags == PAGE_POISON_PATTERN;
212 #ifdef CONFIG_DEBUG_VM
213 void page_init_poison(struct page *page, size_t size);
215 static inline void page_init_poison(struct page *page, size_t size)
221 * Page flags policies wrt compound pages
224 * check if this struct page poisoned/uninitialized
227 * the page flag is relevant for small, head and tail pages.
230 * for compound page all operations related to the page flag applied to
234 * for compound page, callers only ever operate on the head page.
237 * modifications of the page flag must be done on small or head pages,
238 * checks can be done on tail pages too.
241 * the page flag is not relevant for compound pages.
244 * the page flag is stored in the first tail page.
246 #define PF_POISONED_CHECK(page) ({ \
247 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
249 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
250 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
251 #define PF_ONLY_HEAD(page, enforce) ({ \
252 VM_BUG_ON_PGFLAGS(PageTail(page), page); \
253 PF_POISONED_CHECK(page); })
254 #define PF_NO_TAIL(page, enforce) ({ \
255 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
256 PF_POISONED_CHECK(compound_head(page)); })
257 #define PF_NO_COMPOUND(page, enforce) ({ \
258 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
259 PF_POISONED_CHECK(page); })
260 #define PF_SECOND(page, enforce) ({ \
261 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
262 PF_POISONED_CHECK(&page[1]); })
265 * Macros to create function definitions for page flags
267 #define TESTPAGEFLAG(uname, lname, policy) \
268 static __always_inline int Page##uname(struct page *page) \
269 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
271 #define SETPAGEFLAG(uname, lname, policy) \
272 static __always_inline void SetPage##uname(struct page *page) \
273 { set_bit(PG_##lname, &policy(page, 1)->flags); }
275 #define CLEARPAGEFLAG(uname, lname, policy) \
276 static __always_inline void ClearPage##uname(struct page *page) \
277 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
279 #define __SETPAGEFLAG(uname, lname, policy) \
280 static __always_inline void __SetPage##uname(struct page *page) \
281 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
283 #define __CLEARPAGEFLAG(uname, lname, policy) \
284 static __always_inline void __ClearPage##uname(struct page *page) \
285 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
287 #define TESTSETFLAG(uname, lname, policy) \
288 static __always_inline int TestSetPage##uname(struct page *page) \
289 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
291 #define TESTCLEARFLAG(uname, lname, policy) \
292 static __always_inline int TestClearPage##uname(struct page *page) \
293 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
295 #define PAGEFLAG(uname, lname, policy) \
296 TESTPAGEFLAG(uname, lname, policy) \
297 SETPAGEFLAG(uname, lname, policy) \
298 CLEARPAGEFLAG(uname, lname, policy)
300 #define __PAGEFLAG(uname, lname, policy) \
301 TESTPAGEFLAG(uname, lname, policy) \
302 __SETPAGEFLAG(uname, lname, policy) \
303 __CLEARPAGEFLAG(uname, lname, policy)
305 #define TESTSCFLAG(uname, lname, policy) \
306 TESTSETFLAG(uname, lname, policy) \
307 TESTCLEARFLAG(uname, lname, policy)
309 #define TESTPAGEFLAG_FALSE(uname) \
310 static inline int Page##uname(const struct page *page) { return 0; }
312 #define SETPAGEFLAG_NOOP(uname) \
313 static inline void SetPage##uname(struct page *page) { }
315 #define CLEARPAGEFLAG_NOOP(uname) \
316 static inline void ClearPage##uname(struct page *page) { }
318 #define __CLEARPAGEFLAG_NOOP(uname) \
319 static inline void __ClearPage##uname(struct page *page) { }
321 #define TESTSETFLAG_FALSE(uname) \
322 static inline int TestSetPage##uname(struct page *page) { return 0; }
324 #define TESTCLEARFLAG_FALSE(uname) \
325 static inline int TestClearPage##uname(struct page *page) { return 0; }
327 #define PAGEFLAG_FALSE(uname) TESTPAGEFLAG_FALSE(uname) \
328 SETPAGEFLAG_NOOP(uname) CLEARPAGEFLAG_NOOP(uname)
330 #define TESTSCFLAG_FALSE(uname) \
331 TESTSETFLAG_FALSE(uname) TESTCLEARFLAG_FALSE(uname)
333 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
334 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) __CLEARPAGEFLAG(Waiters, waiters, PF_ONLY_HEAD)
335 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
336 PAGEFLAG(Referenced, referenced, PF_HEAD)
337 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
338 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
339 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
340 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
341 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
342 TESTCLEARFLAG(LRU, lru, PF_HEAD)
343 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
344 TESTCLEARFLAG(Active, active, PF_HEAD)
345 PAGEFLAG(Workingset, workingset, PF_HEAD)
346 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
347 __PAGEFLAG(Slab, slab, PF_NO_TAIL)
348 __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL)
349 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
352 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
353 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
354 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
355 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
356 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
357 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
359 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
360 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
361 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
362 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
363 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
364 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
367 * Private page markings that may be used by the filesystem that owns the page
368 * for its own purposes.
369 * - PG_private and PG_private_2 cause releasepage() and co to be invoked
371 PAGEFLAG(Private, private, PF_ANY)
372 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
373 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
374 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
377 * Only test-and-set exist for PG_writeback. The unconditional operators are
378 * risky: they bypass page accounting.
380 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
381 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
382 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
384 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
385 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
386 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
387 PAGEFLAG(Readahead, reclaim, PF_NO_COMPOUND)
388 TESTCLEARFLAG(Readahead, reclaim, PF_NO_COMPOUND)
390 #ifdef CONFIG_HIGHMEM
392 * Must use a macro here due to header dependency issues. page_zone() is not
393 * available at this point.
395 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
397 PAGEFLAG_FALSE(HighMem)
401 static __always_inline int PageSwapCache(struct page *page)
403 #ifdef CONFIG_THP_SWAP
404 page = compound_head(page);
406 return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags);
409 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
410 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
412 PAGEFLAG_FALSE(SwapCache)
415 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
416 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
417 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
420 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
421 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
422 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
424 PAGEFLAG_FALSE(Mlocked) __CLEARPAGEFLAG_NOOP(Mlocked)
425 TESTSCFLAG_FALSE(Mlocked)
428 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
429 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
431 PAGEFLAG_FALSE(Uncached)
434 #ifdef CONFIG_MEMORY_FAILURE
435 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
436 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
437 #define __PG_HWPOISON (1UL << PG_hwpoison)
438 extern bool take_page_off_buddy(struct page *page);
440 PAGEFLAG_FALSE(HWPoison)
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)
451 #ifdef CONFIG_KASAN_HW_TAGS
452 PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD)
454 PAGEFLAG_FALSE(SkipKASanPoison)
458 * PageReported() is used to track reported free pages within the Buddy
459 * allocator. We can use the non-atomic version of the test and set
460 * operations as both should be shielded with the zone lock to prevent
461 * any possible races on the setting or clearing of the bit.
463 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
466 * On an anonymous page mapped into a user virtual memory area,
467 * page->mapping points to its anon_vma, not to a struct address_space;
468 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
470 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
471 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
472 * bit; and then page->mapping points, not to an anon_vma, but to a private
473 * structure which KSM associates with that merged page. See ksm.h.
475 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
476 * page and then page->mapping points a struct address_space.
478 * Please note that, confusingly, "page_mapping" refers to the inode
479 * address_space which maps the page from disk; whereas "page_mapped"
480 * refers to user virtual address space into which the page is mapped.
482 #define PAGE_MAPPING_ANON 0x1
483 #define PAGE_MAPPING_MOVABLE 0x2
484 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
485 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
487 static __always_inline int PageMappingFlags(struct page *page)
489 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
492 static __always_inline int PageAnon(struct page *page)
494 page = compound_head(page);
495 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
498 static __always_inline int __PageMovable(struct page *page)
500 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
501 PAGE_MAPPING_MOVABLE;
506 * A KSM page is one of those write-protected "shared pages" or "merged pages"
507 * which KSM maps into multiple mms, wherever identical anonymous page content
508 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
509 * anon_vma, but to that page's node of the stable tree.
511 static __always_inline int PageKsm(struct page *page)
513 page = compound_head(page);
514 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
518 TESTPAGEFLAG_FALSE(Ksm)
521 u64 stable_page_flags(struct page *page);
523 static inline int PageUptodate(struct page *page)
526 page = compound_head(page);
527 ret = test_bit(PG_uptodate, &(page)->flags);
529 * Must ensure that the data we read out of the page is loaded
530 * _after_ we've loaded page->flags to check for PageUptodate.
531 * We can skip the barrier if the page is not uptodate, because
532 * we wouldn't be reading anything from it.
534 * See SetPageUptodate() for the other side of the story.
542 static __always_inline void __SetPageUptodate(struct page *page)
544 VM_BUG_ON_PAGE(PageTail(page), page);
546 __set_bit(PG_uptodate, &page->flags);
549 static __always_inline void SetPageUptodate(struct page *page)
551 VM_BUG_ON_PAGE(PageTail(page), page);
553 * Memory barrier must be issued before setting the PG_uptodate bit,
554 * so that all previous stores issued in order to bring the page
555 * uptodate are actually visible before PageUptodate becomes true.
558 set_bit(PG_uptodate, &page->flags);
561 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
563 int test_clear_page_writeback(struct page *page);
564 int __test_set_page_writeback(struct page *page, bool keep_write);
566 #define test_set_page_writeback(page) \
567 __test_set_page_writeback(page, false)
568 #define test_set_page_writeback_keepwrite(page) \
569 __test_set_page_writeback(page, true)
571 static inline void set_page_writeback(struct page *page)
573 test_set_page_writeback(page);
576 static inline void set_page_writeback_keepwrite(struct page *page)
578 test_set_page_writeback_keepwrite(page);
581 __PAGEFLAG(Head, head, PF_ANY) CLEARPAGEFLAG(Head, head, PF_ANY)
583 static __always_inline void set_compound_head(struct page *page, struct page *head)
585 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
588 static __always_inline void clear_compound_head(struct page *page)
590 WRITE_ONCE(page->compound_head, 0);
593 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
594 static inline void ClearPageCompound(struct page *page)
596 BUG_ON(!PageHead(page));
601 #define PG_head_mask ((1UL << PG_head))
603 #ifdef CONFIG_HUGETLB_PAGE
604 int PageHuge(struct page *page);
605 int PageHeadHuge(struct page *page);
607 TESTPAGEFLAG_FALSE(Huge)
608 TESTPAGEFLAG_FALSE(HeadHuge)
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 * Check if a page is currently marked HWPoisoned. Note that this check is
710 * best effort only and inherently racy: there is no way to synchronize with
713 static inline bool is_page_hwpoison(struct page *page)
715 if (PageHWPoison(page))
717 return PageHuge(page) && PageHWPoison(compound_head(page));
721 * For pages that are never mapped to userspace (and aren't PageSlab),
722 * page_type may be used. Because it is initialised to -1, we invert the
723 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
724 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
725 * low bits so that an underflow or overflow of page_mapcount() won't be
726 * mistaken for a page type value.
729 #define PAGE_TYPE_BASE 0xf0000000
730 /* Reserve 0x0000007f to catch underflows of page_mapcount */
731 #define PAGE_MAPCOUNT_RESERVE -128
732 #define PG_buddy 0x00000080
733 #define PG_offline 0x00000100
734 #define PG_table 0x00000200
735 #define PG_guard 0x00000400
737 #define PageType(page, flag) \
738 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
740 static inline int page_has_type(struct page *page)
742 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE;
745 #define PAGE_TYPE_OPS(uname, lname) \
746 static __always_inline int Page##uname(struct page *page) \
748 return PageType(page, PG_##lname); \
750 static __always_inline void __SetPage##uname(struct page *page) \
752 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
753 page->page_type &= ~PG_##lname; \
755 static __always_inline void __ClearPage##uname(struct page *page) \
757 VM_BUG_ON_PAGE(!Page##uname(page), page); \
758 page->page_type |= PG_##lname; \
762 * PageBuddy() indicates that the page is free and in the buddy system
763 * (see mm/page_alloc.c).
765 PAGE_TYPE_OPS(Buddy, buddy)
768 * PageOffline() indicates that the page is logically offline although the
769 * containing section is online. (e.g. inflated in a balloon driver or
770 * not onlined when onlining the section).
771 * The content of these pages is effectively stale. Such pages should not
772 * be touched (read/write/dump/save) except by their owner.
774 * If a driver wants to allow to offline unmovable PageOffline() pages without
775 * putting them back to the buddy, it can do so via the memory notifier by
776 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
777 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
778 * pages (now with a reference count of zero) are treated like free pages,
779 * allowing the containing memory block to get offlined. A driver that
780 * relies on this feature is aware that re-onlining the memory block will
781 * require to re-set the pages PageOffline() and not giving them to the
782 * buddy via online_page_callback_t.
784 * There are drivers that mark a page PageOffline() and expect there won't be
785 * any further access to page content. PFN walkers that read content of random
786 * pages should check PageOffline() and synchronize with such drivers using
787 * page_offline_freeze()/page_offline_thaw().
789 PAGE_TYPE_OPS(Offline, offline)
791 extern void page_offline_freeze(void);
792 extern void page_offline_thaw(void);
793 extern void page_offline_begin(void);
794 extern void page_offline_end(void);
797 * Marks pages in use as page tables.
799 PAGE_TYPE_OPS(Table, table)
802 * Marks guardpages used with debug_pagealloc.
804 PAGE_TYPE_OPS(Guard, guard)
806 extern bool is_free_buddy_page(struct page *page);
808 __PAGEFLAG(Isolated, isolated, PF_ANY);
811 * If network-based swap is enabled, sl*b must keep track of whether pages
812 * were allocated from pfmemalloc reserves.
814 static inline int PageSlabPfmemalloc(struct page *page)
816 VM_BUG_ON_PAGE(!PageSlab(page), page);
817 return PageActive(page);
820 static inline void SetPageSlabPfmemalloc(struct page *page)
822 VM_BUG_ON_PAGE(!PageSlab(page), page);
826 static inline void __ClearPageSlabPfmemalloc(struct page *page)
828 VM_BUG_ON_PAGE(!PageSlab(page), page);
829 __ClearPageActive(page);
832 static inline void ClearPageSlabPfmemalloc(struct page *page)
834 VM_BUG_ON_PAGE(!PageSlab(page), page);
835 ClearPageActive(page);
839 #define __PG_MLOCKED (1UL << PG_mlocked)
841 #define __PG_MLOCKED 0
845 * Flags checked when a page is freed. Pages being freed should not have
846 * these flags set. If they are, there is a problem.
848 #define PAGE_FLAGS_CHECK_AT_FREE \
849 (1UL << PG_lru | 1UL << PG_locked | \
850 1UL << PG_private | 1UL << PG_private_2 | \
851 1UL << PG_writeback | 1UL << PG_reserved | \
852 1UL << PG_slab | 1UL << PG_active | \
853 1UL << PG_unevictable | __PG_MLOCKED)
856 * Flags checked when a page is prepped for return by the page allocator.
857 * Pages being prepped should not have these flags set. If they are set,
858 * there has been a kernel bug or struct page corruption.
860 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
861 * alloc-free cycle to prevent from reusing the page.
863 #define PAGE_FLAGS_CHECK_AT_PREP \
864 (PAGEFLAGS_MASK & ~__PG_HWPOISON)
866 #define PAGE_FLAGS_PRIVATE \
867 (1UL << PG_private | 1UL << PG_private_2)
869 * page_has_private - Determine if page has private stuff
870 * @page: The page to be checked
872 * Determine if a page has private stuff, indicating that release routines
873 * should be invoked upon it.
875 static inline int page_has_private(struct page *page)
877 return !!(page->flags & PAGE_FLAGS_PRIVATE);
884 #undef PF_NO_COMPOUND
886 #endif /* !__GENERATING_BOUNDS_H */
888 #endif /* PAGE_FLAGS_H */