xref: /linux/include/linux/page-flags.h (revision 204e9a18f1b9685476d6480d4f26d5d7f7e2d505)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * Macros for manipulating and testing page->flags
4  */
5 
6 #ifndef PAGE_FLAGS_H
7 #define PAGE_FLAGS_H
8 
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 */
16 
17 /*
18  * Various page->flags bits:
19  *
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,
24  *   initrd, HW tables)
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!
32  * - The zero page(s)
33  * - Pages allocated in the context of kexec/kdump (loaded kernel image,
34  *   control pages, vmcoreinfo)
35  * - MMIO/DMA pages. Some architectures don't allow to ioremap pages that are
36  *   not marked PG_reserved (as they might be in use by somebody else who does
37  *   not respect the caching strategy).
38  * - MCA pages on ia64
39  * - Pages holding CPU notes for POWER Firmware Assisted Dump
40  * - Device memory (e.g. PMEM, DAX, HMM)
41  * Some PG_reserved pages will be excluded from the hibernation image.
42  * PG_reserved does in general not hinder anybody from dumping or swapping
43  * and is no longer required for remap_pfn_range(). ioremap might require it.
44  * Consequently, PG_reserved for a page mapped into user space can indicate
45  * the zero page, the vDSO, MMIO pages or device memory.
46  *
47  * The PG_private bitflag is set on pagecache pages if they contain filesystem
48  * specific data (which is normally at page->private). It can be used by
49  * private allocations for its own usage.
50  *
51  * During initiation of disk I/O, PG_locked is set. This bit is set before I/O
52  * and cleared when writeback _starts_ or when read _completes_. PG_writeback
53  * is set before writeback starts and cleared when it finishes.
54  *
55  * PG_locked also pins a page in pagecache, and blocks truncation of the file
56  * while it is held.
57  *
58  * page_waitqueue(page) is a wait queue of all tasks waiting for the page
59  * to become unlocked.
60  *
61  * PG_swapbacked is set when a page uses swap as a backing storage.  This are
62  * usually PageAnon or shmem pages but please note that even anonymous pages
63  * might lose their PG_swapbacked flag when they simply can be dropped (e.g. as
64  * a result of MADV_FREE).
65  *
66  * PG_referenced, PG_reclaim are used for page reclaim for anonymous and
67  * file-backed pagecache (see mm/vmscan.c).
68  *
69  * PG_arch_1 is an architecture specific page state bit.  The generic code
70  * guarantees that this bit is cleared for a page when it first is entered into
71  * the page cache.
72  *
73  * PG_hwpoison indicates that a page got corrupted in hardware and contains
74  * data with incorrect ECC bits that triggered a machine check. Accessing is
75  * not safe since it may cause another machine check. Don't touch!
76  */
77 
78 /*
79  * Don't use the pageflags directly.  Use the PageFoo macros.
80  *
81  * The page flags field is split into two parts, the main flags area
82  * which extends from the low bits upwards, and the fields area which
83  * extends from the high bits downwards.
84  *
85  *  | FIELD | ... | FLAGS |
86  *  N-1           ^       0
87  *               (NR_PAGEFLAGS)
88  *
89  * The fields area is reserved for fields mapping zone, node (for NUMA) and
90  * SPARSEMEM section (for variants of SPARSEMEM that require section ids like
91  * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP).
92  */
93 enum pageflags {
94 	PG_locked,		/* Page is locked. Don't touch. */
95 	PG_writeback,		/* Page is under writeback */
96 	PG_referenced,
97 	PG_uptodate,
98 	PG_dirty,
99 	PG_lru,
100 	PG_head,		/* Must be in bit 6 */
101 	PG_waiters,		/* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
102 	PG_active,
103 	PG_workingset,
104 	PG_owner_priv_1,	/* Owner use. If pagecache, fs may use */
105 	PG_owner_2,		/* Owner use. If pagecache, fs may use */
106 	PG_arch_1,
107 	PG_reserved,
108 	PG_private,		/* If pagecache, has fs-private data */
109 	PG_private_2,		/* If pagecache, has fs aux data */
110 	PG_reclaim,		/* To be reclaimed asap */
111 	PG_swapbacked,		/* Page is backed by RAM/swap */
112 	PG_unevictable,		/* Page is "unevictable"  */
113 	PG_dropbehind,		/* drop pages on IO completion */
114 #ifdef CONFIG_MMU
115 	PG_mlocked,		/* Page is vma mlocked */
116 #endif
117 #ifdef CONFIG_MEMORY_FAILURE
118 	PG_hwpoison,		/* hardware poisoned page. Don't touch */
119 #endif
120 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
121 	PG_young,
122 	PG_idle,
123 #endif
124 #ifdef CONFIG_ARCH_USES_PG_ARCH_2
125 	PG_arch_2,
126 #endif
127 #ifdef CONFIG_ARCH_USES_PG_ARCH_3
128 	PG_arch_3,
129 #endif
130 	__NR_PAGEFLAGS,
131 
132 	PG_readahead = PG_reclaim,
133 
134 	/* Anonymous memory (and shmem) */
135 	PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
136 	/* Some filesystems */
137 	PG_checked = PG_owner_priv_1,
138 
139 	/*
140 	 * Depending on the way an anonymous folio can be mapped into a page
141 	 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped
142 	 * THP), PG_anon_exclusive may be set only for the head page or for
143 	 * tail pages of an anonymous folio. For now, we only expect it to be
144 	 * set on tail pages for PTE-mapped THP.
145 	 */
146 	PG_anon_exclusive = PG_owner_2,
147 
148 	/*
149 	 * Set if all buffer heads in the folio are mapped.
150 	 * Filesystems which do not use BHs can use it for their own purpose.
151 	 */
152 	PG_mappedtodisk = PG_owner_2,
153 
154 	/* Two page bits are conscripted by FS-Cache to maintain local caching
155 	 * state.  These bits are set on pages belonging to the netfs's inodes
156 	 * when those inodes are being locally cached.
157 	 */
158 	PG_fscache = PG_private_2,	/* page backed by cache */
159 
160 	/* XEN */
161 	/* Pinned in Xen as a read-only pagetable page. */
162 	PG_pinned = PG_owner_priv_1,
163 	/* Pinned as part of domain save (see xen_mm_pin_all()). */
164 	PG_savepinned = PG_dirty,
165 	/* Has a grant mapping of another (foreign) domain's page. */
166 	PG_foreign = PG_owner_priv_1,
167 	/* Remapped by swiotlb-xen. */
168 	PG_xen_remapped = PG_owner_priv_1,
169 
170 	/* non-lru isolated movable page */
171 	PG_isolated = PG_reclaim,
172 
173 	/* Only valid for buddy pages. Used to track pages that are reported */
174 	PG_reported = PG_uptodate,
175 
176 #ifdef CONFIG_MEMORY_HOTPLUG
177 	/* For self-hosted memmap pages */
178 	PG_vmemmap_self_hosted = PG_owner_priv_1,
179 #endif
180 
181 	/*
182 	 * Flags only valid for compound pages.  Stored in first tail page's
183 	 * flags word.  Cannot use the first 8 flags or any flag marked as
184 	 * PF_ANY.
185 	 */
186 
187 	/* At least one page in this folio has the hwpoison flag set */
188 	PG_has_hwpoisoned = PG_active,
189 	PG_large_rmappable = PG_workingset, /* anon or file-backed */
190 	PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */
191 };
192 
193 #define PAGEFLAGS_MASK		((1UL << NR_PAGEFLAGS) - 1)
194 
195 #ifndef __GENERATING_BOUNDS_H
196 
197 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
198 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key);
199 
200 /*
201  * Return the real head page struct iff the @page is a fake head page, otherwise
202  * return the @page itself. See Documentation/mm/vmemmap_dedup.rst.
203  */
page_fixed_fake_head(const struct page * page)204 static __always_inline const struct page *page_fixed_fake_head(const struct page *page)
205 {
206 	if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key))
207 		return page;
208 
209 	/*
210 	 * Only addresses aligned with PAGE_SIZE of struct page may be fake head
211 	 * struct page. The alignment check aims to avoid access the fields (
212 	 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly)
213 	 * cold cacheline in some cases.
214 	 */
215 	if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) &&
216 	    test_bit(PG_head, &page->flags)) {
217 		/*
218 		 * We can safely access the field of the @page[1] with PG_head
219 		 * because the @page is a compound page composed with at least
220 		 * two contiguous pages.
221 		 */
222 		unsigned long head = READ_ONCE(page[1].compound_head);
223 
224 		if (likely(head & 1))
225 			return (const struct page *)(head - 1);
226 	}
227 	return page;
228 }
229 
page_count_writable(const struct page * page,int u)230 static __always_inline bool page_count_writable(const struct page *page, int u)
231 {
232 	if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key))
233 		return true;
234 
235 	/*
236 	 * The refcount check is ordered before the fake-head check to prevent
237 	 * the following race:
238 	 *   CPU 1 (HVO)                     CPU 2 (speculative PFN walker)
239 	 *
240 	 *   page_ref_freeze()
241 	 *   synchronize_rcu()
242 	 *                                   rcu_read_lock()
243 	 *                                   page_is_fake_head() is false
244 	 *   vmemmap_remap_pte()
245 	 *   XXX: struct page[] becomes r/o
246 	 *
247 	 *   page_ref_unfreeze()
248 	 *                                   page_ref_count() is not zero
249 	 *
250 	 *                                   atomic_add_unless(&page->_refcount)
251 	 *                                   XXX: try to modify r/o struct page[]
252 	 *
253 	 * The refcount check also prevents modification attempts to other (r/o)
254 	 * tail pages that are not fake heads.
255 	 */
256 	if (atomic_read_acquire(&page->_refcount) == u)
257 		return false;
258 
259 	return page_fixed_fake_head(page) == page;
260 }
261 #else
page_fixed_fake_head(const struct page * page)262 static inline const struct page *page_fixed_fake_head(const struct page *page)
263 {
264 	return page;
265 }
266 
page_count_writable(const struct page * page,int u)267 static inline bool page_count_writable(const struct page *page, int u)
268 {
269 	return true;
270 }
271 #endif
272 
page_is_fake_head(const struct page * page)273 static __always_inline int page_is_fake_head(const struct page *page)
274 {
275 	return page_fixed_fake_head(page) != page;
276 }
277 
_compound_head(const struct page * page)278 static __always_inline unsigned long _compound_head(const struct page *page)
279 {
280 	unsigned long head = READ_ONCE(page->compound_head);
281 
282 	if (unlikely(head & 1))
283 		return head - 1;
284 	return (unsigned long)page_fixed_fake_head(page);
285 }
286 
287 #define compound_head(page)	((typeof(page))_compound_head(page))
288 
289 /**
290  * page_folio - Converts from page to folio.
291  * @p: The page.
292  *
293  * Every page is part of a folio.  This function cannot be called on a
294  * NULL pointer.
295  *
296  * Context: No reference, nor lock is required on @page.  If the caller
297  * does not hold a reference, this call may race with a folio split, so
298  * it should re-check the folio still contains this page after gaining
299  * a reference on the folio.
300  * Return: The folio which contains this page.
301  */
302 #define page_folio(p)		(_Generic((p),				\
303 	const struct page *:	(const struct folio *)_compound_head(p), \
304 	struct page *:		(struct folio *)_compound_head(p)))
305 
306 /**
307  * folio_page - Return a page from a folio.
308  * @folio: The folio.
309  * @n: The page number to return.
310  *
311  * @n is relative to the start of the folio.  This function does not
312  * check that the page number lies within @folio; the caller is presumed
313  * to have a reference to the page.
314  */
315 #define folio_page(folio, n)	nth_page(&(folio)->page, n)
316 
PageTail(const struct page * page)317 static __always_inline int PageTail(const struct page *page)
318 {
319 	return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page);
320 }
321 
PageCompound(const struct page * page)322 static __always_inline int PageCompound(const struct page *page)
323 {
324 	return test_bit(PG_head, &page->flags) ||
325 	       READ_ONCE(page->compound_head) & 1;
326 }
327 
328 #define	PAGE_POISON_PATTERN	-1l
PagePoisoned(const struct page * page)329 static inline int PagePoisoned(const struct page *page)
330 {
331 	return READ_ONCE(page->flags) == PAGE_POISON_PATTERN;
332 }
333 
334 #ifdef CONFIG_DEBUG_VM
335 void page_init_poison(struct page *page, size_t size);
336 #else
page_init_poison(struct page * page,size_t size)337 static inline void page_init_poison(struct page *page, size_t size)
338 {
339 }
340 #endif
341 
const_folio_flags(const struct folio * folio,unsigned n)342 static const unsigned long *const_folio_flags(const struct folio *folio,
343 		unsigned n)
344 {
345 	const struct page *page = &folio->page;
346 
347 	VM_BUG_ON_PGFLAGS(page->compound_head & 1, page);
348 	VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
349 	return &page[n].flags;
350 }
351 
folio_flags(struct folio * folio,unsigned n)352 static unsigned long *folio_flags(struct folio *folio, unsigned n)
353 {
354 	struct page *page = &folio->page;
355 
356 	VM_BUG_ON_PGFLAGS(page->compound_head & 1, page);
357 	VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
358 	return &page[n].flags;
359 }
360 
361 /*
362  * Page flags policies wrt compound pages
363  *
364  * PF_POISONED_CHECK
365  *     check if this struct page poisoned/uninitialized
366  *
367  * PF_ANY:
368  *     the page flag is relevant for small, head and tail pages.
369  *
370  * PF_HEAD:
371  *     for compound page all operations related to the page flag applied to
372  *     head page.
373  *
374  * PF_NO_TAIL:
375  *     modifications of the page flag must be done on small or head pages,
376  *     checks can be done on tail pages too.
377  *
378  * PF_NO_COMPOUND:
379  *     the page flag is not relevant for compound pages.
380  *
381  * PF_SECOND:
382  *     the page flag is stored in the first tail page.
383  */
384 #define PF_POISONED_CHECK(page) ({					\
385 		VM_BUG_ON_PGFLAGS(PagePoisoned(page), page);		\
386 		page; })
387 #define PF_ANY(page, enforce)	PF_POISONED_CHECK(page)
388 #define PF_HEAD(page, enforce)	PF_POISONED_CHECK(compound_head(page))
389 #define PF_NO_TAIL(page, enforce) ({					\
390 		VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page);	\
391 		PF_POISONED_CHECK(compound_head(page)); })
392 #define PF_NO_COMPOUND(page, enforce) ({				\
393 		VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page);	\
394 		PF_POISONED_CHECK(page); })
395 #define PF_SECOND(page, enforce) ({					\
396 		VM_BUG_ON_PGFLAGS(!PageHead(page), page);		\
397 		PF_POISONED_CHECK(&page[1]); })
398 
399 /* Which page is the flag stored in */
400 #define FOLIO_PF_ANY		0
401 #define FOLIO_PF_HEAD		0
402 #define FOLIO_PF_NO_TAIL	0
403 #define FOLIO_PF_NO_COMPOUND	0
404 #define FOLIO_PF_SECOND		1
405 
406 #define FOLIO_HEAD_PAGE		0
407 #define FOLIO_SECOND_PAGE	1
408 
409 /*
410  * Macros to create function definitions for page flags
411  */
412 #define FOLIO_TEST_FLAG(name, page)					\
413 static __always_inline bool folio_test_##name(const struct folio *folio) \
414 { return test_bit(PG_##name, const_folio_flags(folio, page)); }
415 
416 #define FOLIO_SET_FLAG(name, page)					\
417 static __always_inline void folio_set_##name(struct folio *folio)	\
418 { set_bit(PG_##name, folio_flags(folio, page)); }
419 
420 #define FOLIO_CLEAR_FLAG(name, page)					\
421 static __always_inline void folio_clear_##name(struct folio *folio)	\
422 { clear_bit(PG_##name, folio_flags(folio, page)); }
423 
424 #define __FOLIO_SET_FLAG(name, page)					\
425 static __always_inline void __folio_set_##name(struct folio *folio)	\
426 { __set_bit(PG_##name, folio_flags(folio, page)); }
427 
428 #define __FOLIO_CLEAR_FLAG(name, page)					\
429 static __always_inline void __folio_clear_##name(struct folio *folio)	\
430 { __clear_bit(PG_##name, folio_flags(folio, page)); }
431 
432 #define FOLIO_TEST_SET_FLAG(name, page)					\
433 static __always_inline bool folio_test_set_##name(struct folio *folio)	\
434 { return test_and_set_bit(PG_##name, folio_flags(folio, page)); }
435 
436 #define FOLIO_TEST_CLEAR_FLAG(name, page)				\
437 static __always_inline bool folio_test_clear_##name(struct folio *folio) \
438 { return test_and_clear_bit(PG_##name, folio_flags(folio, page)); }
439 
440 #define FOLIO_FLAG(name, page)						\
441 FOLIO_TEST_FLAG(name, page)						\
442 FOLIO_SET_FLAG(name, page)						\
443 FOLIO_CLEAR_FLAG(name, page)
444 
445 #define TESTPAGEFLAG(uname, lname, policy)				\
446 FOLIO_TEST_FLAG(lname, FOLIO_##policy)					\
447 static __always_inline int Page##uname(const struct page *page)		\
448 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
449 
450 #define SETPAGEFLAG(uname, lname, policy)				\
451 FOLIO_SET_FLAG(lname, FOLIO_##policy)					\
452 static __always_inline void SetPage##uname(struct page *page)		\
453 { set_bit(PG_##lname, &policy(page, 1)->flags); }
454 
455 #define CLEARPAGEFLAG(uname, lname, policy)				\
456 FOLIO_CLEAR_FLAG(lname, FOLIO_##policy)					\
457 static __always_inline void ClearPage##uname(struct page *page)		\
458 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
459 
460 #define __SETPAGEFLAG(uname, lname, policy)				\
461 __FOLIO_SET_FLAG(lname, FOLIO_##policy)					\
462 static __always_inline void __SetPage##uname(struct page *page)		\
463 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
464 
465 #define __CLEARPAGEFLAG(uname, lname, policy)				\
466 __FOLIO_CLEAR_FLAG(lname, FOLIO_##policy)				\
467 static __always_inline void __ClearPage##uname(struct page *page)	\
468 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
469 
470 #define TESTSETFLAG(uname, lname, policy)				\
471 FOLIO_TEST_SET_FLAG(lname, FOLIO_##policy)				\
472 static __always_inline int TestSetPage##uname(struct page *page)	\
473 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
474 
475 #define TESTCLEARFLAG(uname, lname, policy)				\
476 FOLIO_TEST_CLEAR_FLAG(lname, FOLIO_##policy)				\
477 static __always_inline int TestClearPage##uname(struct page *page)	\
478 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
479 
480 #define PAGEFLAG(uname, lname, policy)					\
481 	TESTPAGEFLAG(uname, lname, policy)				\
482 	SETPAGEFLAG(uname, lname, policy)				\
483 	CLEARPAGEFLAG(uname, lname, policy)
484 
485 #define __PAGEFLAG(uname, lname, policy)				\
486 	TESTPAGEFLAG(uname, lname, policy)				\
487 	__SETPAGEFLAG(uname, lname, policy)				\
488 	__CLEARPAGEFLAG(uname, lname, policy)
489 
490 #define TESTSCFLAG(uname, lname, policy)				\
491 	TESTSETFLAG(uname, lname, policy)				\
492 	TESTCLEARFLAG(uname, lname, policy)
493 
494 #define FOLIO_TEST_FLAG_FALSE(name)					\
495 static inline bool folio_test_##name(const struct folio *folio)		\
496 { return false; }
497 #define FOLIO_SET_FLAG_NOOP(name)					\
498 static inline void folio_set_##name(struct folio *folio) { }
499 #define FOLIO_CLEAR_FLAG_NOOP(name)					\
500 static inline void folio_clear_##name(struct folio *folio) { }
501 #define __FOLIO_SET_FLAG_NOOP(name)					\
502 static inline void __folio_set_##name(struct folio *folio) { }
503 #define __FOLIO_CLEAR_FLAG_NOOP(name)					\
504 static inline void __folio_clear_##name(struct folio *folio) { }
505 #define FOLIO_TEST_SET_FLAG_FALSE(name)					\
506 static inline bool folio_test_set_##name(struct folio *folio)		\
507 { return false; }
508 #define FOLIO_TEST_CLEAR_FLAG_FALSE(name)				\
509 static inline bool folio_test_clear_##name(struct folio *folio)		\
510 { return false; }
511 
512 #define FOLIO_FLAG_FALSE(name)						\
513 FOLIO_TEST_FLAG_FALSE(name)						\
514 FOLIO_SET_FLAG_NOOP(name)						\
515 FOLIO_CLEAR_FLAG_NOOP(name)
516 
517 #define TESTPAGEFLAG_FALSE(uname, lname)				\
518 FOLIO_TEST_FLAG_FALSE(lname)						\
519 static inline int Page##uname(const struct page *page) { return 0; }
520 
521 #define SETPAGEFLAG_NOOP(uname, lname)					\
522 FOLIO_SET_FLAG_NOOP(lname)						\
523 static inline void SetPage##uname(struct page *page) {  }
524 
525 #define CLEARPAGEFLAG_NOOP(uname, lname)				\
526 FOLIO_CLEAR_FLAG_NOOP(lname)						\
527 static inline void ClearPage##uname(struct page *page) {  }
528 
529 #define __CLEARPAGEFLAG_NOOP(uname, lname)				\
530 __FOLIO_CLEAR_FLAG_NOOP(lname)						\
531 static inline void __ClearPage##uname(struct page *page) {  }
532 
533 #define TESTSETFLAG_FALSE(uname, lname)					\
534 FOLIO_TEST_SET_FLAG_FALSE(lname)					\
535 static inline int TestSetPage##uname(struct page *page) { return 0; }
536 
537 #define TESTCLEARFLAG_FALSE(uname, lname)				\
538 FOLIO_TEST_CLEAR_FLAG_FALSE(lname)					\
539 static inline int TestClearPage##uname(struct page *page) { return 0; }
540 
541 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname)	\
542 	SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
543 
544 #define TESTSCFLAG_FALSE(uname, lname)					\
545 	TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
546 
547 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
548 FOLIO_FLAG(waiters, FOLIO_HEAD_PAGE)
549 FOLIO_FLAG(referenced, FOLIO_HEAD_PAGE)
550 	FOLIO_TEST_CLEAR_FLAG(referenced, FOLIO_HEAD_PAGE)
551 	__FOLIO_SET_FLAG(referenced, FOLIO_HEAD_PAGE)
552 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
553 	__CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
554 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
555 	TESTCLEARFLAG(LRU, lru, PF_HEAD)
556 FOLIO_FLAG(active, FOLIO_HEAD_PAGE)
557 	__FOLIO_CLEAR_FLAG(active, FOLIO_HEAD_PAGE)
558 	FOLIO_TEST_CLEAR_FLAG(active, FOLIO_HEAD_PAGE)
559 PAGEFLAG(Workingset, workingset, PF_HEAD)
560 	TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
561 PAGEFLAG(Checked, checked, PF_NO_COMPOUND)	   /* Used by some filesystems */
562 
563 /* Xen */
564 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
565 	TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
566 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
567 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
PAGEFLAG(XenRemapped,xen_remapped,PF_NO_COMPOUND)568 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
569 	TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
570 
571 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
572 	__CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
573 	__SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
574 FOLIO_FLAG(swapbacked, FOLIO_HEAD_PAGE)
575 	__FOLIO_CLEAR_FLAG(swapbacked, FOLIO_HEAD_PAGE)
576 	__FOLIO_SET_FLAG(swapbacked, FOLIO_HEAD_PAGE)
577 
578 /*
579  * Private page markings that may be used by the filesystem that owns the page
580  * for its own purposes.
581  * - PG_private and PG_private_2 cause release_folio() and co to be invoked
582  */
583 PAGEFLAG(Private, private, PF_ANY)
584 FOLIO_FLAG(private_2, FOLIO_HEAD_PAGE)
585 
586 /* owner_2 can be set on tail pages for anon memory */
587 FOLIO_FLAG(owner_2, FOLIO_HEAD_PAGE)
588 
589 /*
590  * Only test-and-set exist for PG_writeback.  The unconditional operators are
591  * risky: they bypass page accounting.
592  */
593 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
594 	TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
595 FOLIO_FLAG(mappedtodisk, FOLIO_HEAD_PAGE)
596 
597 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
598 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
599 	TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
600 FOLIO_FLAG(readahead, FOLIO_HEAD_PAGE)
601 	FOLIO_TEST_CLEAR_FLAG(readahead, FOLIO_HEAD_PAGE)
602 
603 FOLIO_FLAG(dropbehind, FOLIO_HEAD_PAGE)
604 	FOLIO_TEST_CLEAR_FLAG(dropbehind, FOLIO_HEAD_PAGE)
605 	__FOLIO_SET_FLAG(dropbehind, FOLIO_HEAD_PAGE)
606 
607 #ifdef CONFIG_HIGHMEM
608 /*
609  * Must use a macro here due to header dependency issues. page_zone() is not
610  * available at this point.
611  */
612 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
613 #define folio_test_highmem(__f)	is_highmem_idx(folio_zonenum(__f))
614 #else
615 PAGEFLAG_FALSE(HighMem, highmem)
616 #endif
617 
618 #ifdef CONFIG_SWAP
619 static __always_inline bool folio_test_swapcache(const struct folio *folio)
620 {
621 	return folio_test_swapbacked(folio) &&
622 			test_bit(PG_swapcache, const_folio_flags(folio, 0));
623 }
624 
FOLIO_SET_FLAG(swapcache,FOLIO_HEAD_PAGE)625 FOLIO_SET_FLAG(swapcache, FOLIO_HEAD_PAGE)
626 FOLIO_CLEAR_FLAG(swapcache, FOLIO_HEAD_PAGE)
627 #else
628 FOLIO_FLAG_FALSE(swapcache)
629 #endif
630 
631 FOLIO_FLAG(unevictable, FOLIO_HEAD_PAGE)
632 	__FOLIO_CLEAR_FLAG(unevictable, FOLIO_HEAD_PAGE)
633 	FOLIO_TEST_CLEAR_FLAG(unevictable, FOLIO_HEAD_PAGE)
634 
635 #ifdef CONFIG_MMU
636 FOLIO_FLAG(mlocked, FOLIO_HEAD_PAGE)
637 	__FOLIO_CLEAR_FLAG(mlocked, FOLIO_HEAD_PAGE)
638 	FOLIO_TEST_CLEAR_FLAG(mlocked, FOLIO_HEAD_PAGE)
639 	FOLIO_TEST_SET_FLAG(mlocked, FOLIO_HEAD_PAGE)
640 #else
641 FOLIO_FLAG_FALSE(mlocked)
642 	__FOLIO_CLEAR_FLAG_NOOP(mlocked)
643 	FOLIO_TEST_CLEAR_FLAG_FALSE(mlocked)
644 	FOLIO_TEST_SET_FLAG_FALSE(mlocked)
645 #endif
646 
647 #ifdef CONFIG_MEMORY_FAILURE
648 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
649 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
650 #define __PG_HWPOISON (1UL << PG_hwpoison)
651 #else
652 PAGEFLAG_FALSE(HWPoison, hwpoison)
653 #define __PG_HWPOISON 0
654 #endif
655 
656 #ifdef CONFIG_PAGE_IDLE_FLAG
657 #ifdef CONFIG_64BIT
658 FOLIO_TEST_FLAG(young, FOLIO_HEAD_PAGE)
659 FOLIO_SET_FLAG(young, FOLIO_HEAD_PAGE)
660 FOLIO_TEST_CLEAR_FLAG(young, FOLIO_HEAD_PAGE)
661 FOLIO_FLAG(idle, FOLIO_HEAD_PAGE)
662 #endif
663 /* See page_idle.h for !64BIT workaround */
664 #else /* !CONFIG_PAGE_IDLE_FLAG */
665 FOLIO_FLAG_FALSE(young)
666 FOLIO_TEST_CLEAR_FLAG_FALSE(young)
667 FOLIO_FLAG_FALSE(idle)
668 #endif
669 
670 /*
671  * PageReported() is used to track reported free pages within the Buddy
672  * allocator. We can use the non-atomic version of the test and set
673  * operations as both should be shielded with the zone lock to prevent
674  * any possible races on the setting or clearing of the bit.
675  */
676 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
677 
678 #ifdef CONFIG_MEMORY_HOTPLUG
679 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
680 #else
681 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
682 #endif
683 
684 /*
685  * On an anonymous folio mapped into a user virtual memory area,
686  * folio->mapping points to its anon_vma, not to a struct address_space;
687  * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
688  *
689  * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
690  * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
691  * bit; and then folio->mapping points, not to an anon_vma, but to a private
692  * structure which KSM associates with that merged page.  See ksm.h.
693  *
694  * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
695  * page and then folio->mapping points to a struct movable_operations.
696  *
697  * Please note that, confusingly, "folio_mapping" refers to the inode
698  * address_space which maps the folio from disk; whereas "folio_mapped"
699  * refers to user virtual address space into which the folio is mapped.
700  *
701  * For slab pages, since slab reuses the bits in struct page to store its
702  * internal states, the folio->mapping does not exist as such, nor do
703  * these flags below.  So in order to avoid testing non-existent bits,
704  * please make sure that folio_test_slab(folio) actually evaluates to
705  * false before calling the following functions (e.g., folio_test_anon).
706  * See mm/slab.h.
707  */
708 #define PAGE_MAPPING_ANON	0x1
709 #define PAGE_MAPPING_MOVABLE	0x2
710 #define PAGE_MAPPING_KSM	(PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
711 #define PAGE_MAPPING_FLAGS	(PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
712 
713 static __always_inline bool folio_mapping_flags(const struct folio *folio)
714 {
715 	return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0;
716 }
717 
PageMappingFlags(const struct page * page)718 static __always_inline bool PageMappingFlags(const struct page *page)
719 {
720 	return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
721 }
722 
folio_test_anon(const struct folio * folio)723 static __always_inline bool folio_test_anon(const struct folio *folio)
724 {
725 	return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0;
726 }
727 
PageAnonNotKsm(const struct page * page)728 static __always_inline bool PageAnonNotKsm(const struct page *page)
729 {
730 	unsigned long flags = (unsigned long)page_folio(page)->mapping;
731 
732 	return (flags & PAGE_MAPPING_FLAGS) == PAGE_MAPPING_ANON;
733 }
734 
PageAnon(const struct page * page)735 static __always_inline bool PageAnon(const struct page *page)
736 {
737 	return folio_test_anon(page_folio(page));
738 }
739 
__folio_test_movable(const struct folio * folio)740 static __always_inline bool __folio_test_movable(const struct folio *folio)
741 {
742 	return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
743 			PAGE_MAPPING_MOVABLE;
744 }
745 
__PageMovable(const struct page * page)746 static __always_inline bool __PageMovable(const struct page *page)
747 {
748 	return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
749 				PAGE_MAPPING_MOVABLE;
750 }
751 
752 #ifdef CONFIG_KSM
753 /*
754  * A KSM page is one of those write-protected "shared pages" or "merged pages"
755  * which KSM maps into multiple mms, wherever identical anonymous page content
756  * is found in VM_MERGEABLE vmas.  It's a PageAnon page, pointing not to any
757  * anon_vma, but to that page's node of the stable tree.
758  */
folio_test_ksm(const struct folio * folio)759 static __always_inline bool folio_test_ksm(const struct folio *folio)
760 {
761 	return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
762 				PAGE_MAPPING_KSM;
763 }
764 #else
765 FOLIO_TEST_FLAG_FALSE(ksm)
766 #endif
767 
768 u64 stable_page_flags(const struct page *page);
769 
770 /**
771  * folio_xor_flags_has_waiters - Change some folio flags.
772  * @folio: The folio.
773  * @mask: Bits set in this word will be changed.
774  *
775  * This must only be used for flags which are changed with the folio
776  * lock held.  For example, it is unsafe to use for PG_dirty as that
777  * can be set without the folio lock held.  It can also only be used
778  * on flags which are in the range 0-6 as some of the implementations
779  * only affect those bits.
780  *
781  * Return: Whether there are tasks waiting on the folio.
782  */
folio_xor_flags_has_waiters(struct folio * folio,unsigned long mask)783 static inline bool folio_xor_flags_has_waiters(struct folio *folio,
784 		unsigned long mask)
785 {
786 	return xor_unlock_is_negative_byte(mask, folio_flags(folio, 0));
787 }
788 
789 /**
790  * folio_test_uptodate - Is this folio up to date?
791  * @folio: The folio.
792  *
793  * The uptodate flag is set on a folio when every byte in the folio is
794  * at least as new as the corresponding bytes on storage.  Anonymous
795  * and CoW folios are always uptodate.  If the folio is not uptodate,
796  * some of the bytes in it may be; see the is_partially_uptodate()
797  * address_space operation.
798  */
folio_test_uptodate(const struct folio * folio)799 static inline bool folio_test_uptodate(const struct folio *folio)
800 {
801 	bool ret = test_bit(PG_uptodate, const_folio_flags(folio, 0));
802 	/*
803 	 * Must ensure that the data we read out of the folio is loaded
804 	 * _after_ we've loaded folio->flags to check the uptodate bit.
805 	 * We can skip the barrier if the folio is not uptodate, because
806 	 * we wouldn't be reading anything from it.
807 	 *
808 	 * See folio_mark_uptodate() for the other side of the story.
809 	 */
810 	if (ret)
811 		smp_rmb();
812 
813 	return ret;
814 }
815 
PageUptodate(const struct page * page)816 static inline bool PageUptodate(const struct page *page)
817 {
818 	return folio_test_uptodate(page_folio(page));
819 }
820 
__folio_mark_uptodate(struct folio * folio)821 static __always_inline void __folio_mark_uptodate(struct folio *folio)
822 {
823 	smp_wmb();
824 	__set_bit(PG_uptodate, folio_flags(folio, 0));
825 }
826 
folio_mark_uptodate(struct folio * folio)827 static __always_inline void folio_mark_uptodate(struct folio *folio)
828 {
829 	/*
830 	 * Memory barrier must be issued before setting the PG_uptodate bit,
831 	 * so that all previous stores issued in order to bring the folio
832 	 * uptodate are actually visible before folio_test_uptodate becomes true.
833 	 */
834 	smp_wmb();
835 	set_bit(PG_uptodate, folio_flags(folio, 0));
836 }
837 
__SetPageUptodate(struct page * page)838 static __always_inline void __SetPageUptodate(struct page *page)
839 {
840 	__folio_mark_uptodate((struct folio *)page);
841 }
842 
SetPageUptodate(struct page * page)843 static __always_inline void SetPageUptodate(struct page *page)
844 {
845 	folio_mark_uptodate((struct folio *)page);
846 }
847 
848 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
849 
850 void __folio_start_writeback(struct folio *folio, bool keep_write);
851 void set_page_writeback(struct page *page);
852 
853 #define folio_start_writeback(folio)			\
854 	__folio_start_writeback(folio, false)
855 #define folio_start_writeback_keepwrite(folio)	\
856 	__folio_start_writeback(folio, true)
857 
folio_test_head(const struct folio * folio)858 static __always_inline bool folio_test_head(const struct folio *folio)
859 {
860 	return test_bit(PG_head, const_folio_flags(folio, FOLIO_PF_ANY));
861 }
862 
PageHead(const struct page * page)863 static __always_inline int PageHead(const struct page *page)
864 {
865 	PF_POISONED_CHECK(page);
866 	return test_bit(PG_head, &page->flags) && !page_is_fake_head(page);
867 }
868 
__SETPAGEFLAG(Head,head,PF_ANY)869 __SETPAGEFLAG(Head, head, PF_ANY)
870 __CLEARPAGEFLAG(Head, head, PF_ANY)
871 CLEARPAGEFLAG(Head, head, PF_ANY)
872 
873 /**
874  * folio_test_large() - Does this folio contain more than one page?
875  * @folio: The folio to test.
876  *
877  * Return: True if the folio is larger than one page.
878  */
879 static inline bool folio_test_large(const struct folio *folio)
880 {
881 	return folio_test_head(folio);
882 }
883 
set_compound_head(struct page * page,struct page * head)884 static __always_inline void set_compound_head(struct page *page, struct page *head)
885 {
886 	WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
887 }
888 
clear_compound_head(struct page * page)889 static __always_inline void clear_compound_head(struct page *page)
890 {
891 	WRITE_ONCE(page->compound_head, 0);
892 }
893 
894 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
ClearPageCompound(struct page * page)895 static inline void ClearPageCompound(struct page *page)
896 {
897 	BUG_ON(!PageHead(page));
898 	ClearPageHead(page);
899 }
FOLIO_FLAG(large_rmappable,FOLIO_SECOND_PAGE)900 FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE)
901 FOLIO_FLAG(partially_mapped, FOLIO_SECOND_PAGE)
902 #else
903 FOLIO_FLAG_FALSE(large_rmappable)
904 FOLIO_FLAG_FALSE(partially_mapped)
905 #endif
906 
907 #define PG_head_mask ((1UL << PG_head))
908 
909 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
910 /*
911  * PageHuge() only returns true for hugetlbfs pages, but not for
912  * normal or transparent huge pages.
913  *
914  * PageTransHuge() returns true for both transparent huge and
915  * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
916  * called only in the core VM paths where hugetlbfs pages can't exist.
917  */
918 static inline int PageTransHuge(const struct page *page)
919 {
920 	VM_BUG_ON_PAGE(PageTail(page), page);
921 	return PageHead(page);
922 }
923 
924 /*
925  * PageTransCompound returns true for both transparent huge pages
926  * and hugetlbfs pages, so it should only be called when it's known
927  * that hugetlbfs pages aren't involved.
928  */
PageTransCompound(const struct page * page)929 static inline int PageTransCompound(const struct page *page)
930 {
931 	return PageCompound(page);
932 }
933 #else
934 TESTPAGEFLAG_FALSE(TransHuge, transhuge)
935 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
936 #endif
937 
938 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
939 /*
940  * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
941  * compound page.
942  *
943  * This flag is set by hwpoison handler.  Cleared by THP split or free page.
944  */
945 FOLIO_FLAG(has_hwpoisoned, FOLIO_SECOND_PAGE)
946 #else
947 FOLIO_FLAG_FALSE(has_hwpoisoned)
948 #endif
949 
950 /*
951  * For pages that do not use mapcount, page_type may be used.
952  * The low 24 bits of pagetype may be used for your own purposes, as long
953  * as you are careful to not affect the top 8 bits.  The low bits of
954  * pagetype will be overwritten when you clear the page_type from the page.
955  */
956 enum pagetype {
957 	/* 0x00-0x7f are positive numbers, ie mapcount */
958 	/* Reserve 0x80-0xef for mapcount overflow. */
959 	PGTY_buddy		= 0xf0,
960 	PGTY_offline		= 0xf1,
961 	PGTY_table		= 0xf2,
962 	PGTY_guard		= 0xf3,
963 	PGTY_hugetlb		= 0xf4,
964 	PGTY_slab		= 0xf5,
965 	PGTY_zsmalloc		= 0xf6,
966 	PGTY_unaccepted		= 0xf7,
967 	PGTY_large_kmalloc	= 0xf8,
968 
969 	PGTY_mapcount_underflow = 0xff
970 };
971 
page_type_has_type(int page_type)972 static inline bool page_type_has_type(int page_type)
973 {
974 	return page_type < (PGTY_mapcount_underflow << 24);
975 }
976 
977 /* This takes a mapcount which is one more than page->_mapcount */
page_mapcount_is_type(unsigned int mapcount)978 static inline bool page_mapcount_is_type(unsigned int mapcount)
979 {
980 	return page_type_has_type(mapcount - 1);
981 }
982 
page_has_type(const struct page * page)983 static inline bool page_has_type(const struct page *page)
984 {
985 	return page_mapcount_is_type(data_race(page->page_type));
986 }
987 
988 #define FOLIO_TYPE_OPS(lname, fname)					\
989 static __always_inline bool folio_test_##fname(const struct folio *folio) \
990 {									\
991 	return data_race(folio->page.page_type >> 24) == PGTY_##lname;	\
992 }									\
993 static __always_inline void __folio_set_##fname(struct folio *folio)	\
994 {									\
995 	if (folio_test_##fname(folio))					\
996 		return;							\
997 	VM_BUG_ON_FOLIO(data_race(folio->page.page_type) != UINT_MAX,	\
998 			folio);						\
999 	folio->page.page_type = (unsigned int)PGTY_##lname << 24;	\
1000 }									\
1001 static __always_inline void __folio_clear_##fname(struct folio *folio)	\
1002 {									\
1003 	if (folio->page.page_type == UINT_MAX)				\
1004 		return;							\
1005 	VM_BUG_ON_FOLIO(!folio_test_##fname(folio), folio);		\
1006 	folio->page.page_type = UINT_MAX;				\
1007 }
1008 
1009 #define PAGE_TYPE_OPS(uname, lname, fname)				\
1010 FOLIO_TYPE_OPS(lname, fname)						\
1011 static __always_inline int Page##uname(const struct page *page)		\
1012 {									\
1013 	return data_race(page->page_type >> 24) == PGTY_##lname;	\
1014 }									\
1015 static __always_inline void __SetPage##uname(struct page *page)		\
1016 {									\
1017 	if (Page##uname(page))						\
1018 		return;							\
1019 	VM_BUG_ON_PAGE(data_race(page->page_type) != UINT_MAX, page);	\
1020 	page->page_type = (unsigned int)PGTY_##lname << 24;		\
1021 }									\
1022 static __always_inline void __ClearPage##uname(struct page *page)	\
1023 {									\
1024 	if (page->page_type == UINT_MAX)				\
1025 		return;							\
1026 	VM_BUG_ON_PAGE(!Page##uname(page), page);			\
1027 	page->page_type = UINT_MAX;					\
1028 }
1029 
1030 /*
1031  * PageBuddy() indicates that the page is free and in the buddy system
1032  * (see mm/page_alloc.c).
1033  */
1034 PAGE_TYPE_OPS(Buddy, buddy, buddy)
1035 
1036 /*
1037  * PageOffline() indicates that the page is logically offline although the
1038  * containing section is online. (e.g. inflated in a balloon driver or
1039  * not onlined when onlining the section).
1040  * The content of these pages is effectively stale. Such pages should not
1041  * be touched (read/write/dump/save) except by their owner.
1042  *
1043  * When a memory block gets onlined, all pages are initialized with a
1044  * refcount of 1 and PageOffline(). generic_online_page() will
1045  * take care of clearing PageOffline().
1046  *
1047  * If a driver wants to allow to offline unmovable PageOffline() pages without
1048  * putting them back to the buddy, it can do so via the memory notifier by
1049  * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
1050  * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
1051  * pages (now with a reference count of zero) are treated like free (unmanaged)
1052  * pages, allowing the containing memory block to get offlined. A driver that
1053  * relies on this feature is aware that re-onlining the memory block will
1054  * require not giving them to the buddy via generic_online_page().
1055  *
1056  * Memory offlining code will not adjust the managed page count for any
1057  * PageOffline() pages, treating them like they were never exposed to the
1058  * buddy using generic_online_page().
1059  *
1060  * There are drivers that mark a page PageOffline() and expect there won't be
1061  * any further access to page content. PFN walkers that read content of random
1062  * pages should check PageOffline() and synchronize with such drivers using
1063  * page_offline_freeze()/page_offline_thaw().
1064  */
1065 PAGE_TYPE_OPS(Offline, offline, offline)
1066 
1067 extern void page_offline_freeze(void);
1068 extern void page_offline_thaw(void);
1069 extern void page_offline_begin(void);
1070 extern void page_offline_end(void);
1071 
1072 /*
1073  * Marks pages in use as page tables.
1074  */
PAGE_TYPE_OPS(Table,table,pgtable)1075 PAGE_TYPE_OPS(Table, table, pgtable)
1076 
1077 /*
1078  * Marks guardpages used with debug_pagealloc.
1079  */
1080 PAGE_TYPE_OPS(Guard, guard, guard)
1081 
1082 FOLIO_TYPE_OPS(slab, slab)
1083 
1084 /**
1085  * PageSlab - Determine if the page belongs to the slab allocator
1086  * @page: The page to test.
1087  *
1088  * Context: Any context.
1089  * Return: True for slab pages, false for any other kind of page.
1090  */
1091 static inline bool PageSlab(const struct page *page)
1092 {
1093 	return folio_test_slab(page_folio(page));
1094 }
1095 
1096 #ifdef CONFIG_HUGETLB_PAGE
FOLIO_TYPE_OPS(hugetlb,hugetlb)1097 FOLIO_TYPE_OPS(hugetlb, hugetlb)
1098 #else
1099 FOLIO_TEST_FLAG_FALSE(hugetlb)
1100 #endif
1101 
1102 PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
1103 
1104 /*
1105  * Mark pages that has to be accepted before touched for the first time.
1106  *
1107  * Serialized with zone lock.
1108  */
1109 PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
1110 FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
1111 
1112 /**
1113  * PageHuge - Determine if the page belongs to hugetlbfs
1114  * @page: The page to test.
1115  *
1116  * Context: Any context.
1117  * Return: True for hugetlbfs pages, false for anon pages or pages
1118  * belonging to other filesystems.
1119  */
1120 static inline bool PageHuge(const struct page *page)
1121 {
1122 	return folio_test_hugetlb(page_folio(page));
1123 }
1124 
1125 /*
1126  * Check if a page is currently marked HWPoisoned. Note that this check is
1127  * best effort only and inherently racy: there is no way to synchronize with
1128  * failing hardware.
1129  */
is_page_hwpoison(const struct page * page)1130 static inline bool is_page_hwpoison(const struct page *page)
1131 {
1132 	const struct folio *folio;
1133 
1134 	if (PageHWPoison(page))
1135 		return true;
1136 	folio = page_folio(page);
1137 	return folio_test_hugetlb(folio) && PageHWPoison(&folio->page);
1138 }
1139 
folio_contain_hwpoisoned_page(struct folio * folio)1140 static inline bool folio_contain_hwpoisoned_page(struct folio *folio)
1141 {
1142 	return folio_test_hwpoison(folio) ||
1143 	    (folio_test_large(folio) && folio_test_has_hwpoisoned(folio));
1144 }
1145 
1146 bool is_free_buddy_page(const struct page *page);
1147 
1148 PAGEFLAG(Isolated, isolated, PF_ANY);
1149 
PageAnonExclusive(const struct page * page)1150 static __always_inline int PageAnonExclusive(const struct page *page)
1151 {
1152 	VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1153 	/*
1154 	 * HugeTLB stores this information on the head page; THP keeps it per
1155 	 * page
1156 	 */
1157 	if (PageHuge(page))
1158 		page = compound_head(page);
1159 	return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1160 }
1161 
SetPageAnonExclusive(struct page * page)1162 static __always_inline void SetPageAnonExclusive(struct page *page)
1163 {
1164 	VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page);
1165 	VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1166 	set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1167 }
1168 
ClearPageAnonExclusive(struct page * page)1169 static __always_inline void ClearPageAnonExclusive(struct page *page)
1170 {
1171 	VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page);
1172 	VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1173 	clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1174 }
1175 
__ClearPageAnonExclusive(struct page * page)1176 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1177 {
1178 	VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1179 	VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1180 	__clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1181 }
1182 
1183 #ifdef CONFIG_MMU
1184 #define __PG_MLOCKED		(1UL << PG_mlocked)
1185 #else
1186 #define __PG_MLOCKED		0
1187 #endif
1188 
1189 /*
1190  * Flags checked when a page is freed.  Pages being freed should not have
1191  * these flags set.  If they are, there is a problem.
1192  */
1193 #define PAGE_FLAGS_CHECK_AT_FREE				\
1194 	(1UL << PG_lru		| 1UL << PG_locked	|	\
1195 	 1UL << PG_private	| 1UL << PG_private_2	|	\
1196 	 1UL << PG_writeback	| 1UL << PG_reserved	|	\
1197 	 1UL << PG_active 	|				\
1198 	 1UL << PG_unevictable	| __PG_MLOCKED | LRU_GEN_MASK)
1199 
1200 /*
1201  * Flags checked when a page is prepped for return by the page allocator.
1202  * Pages being prepped should not have these flags set.  If they are set,
1203  * there has been a kernel bug or struct page corruption.
1204  *
1205  * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1206  * alloc-free cycle to prevent from reusing the page.
1207  */
1208 #define PAGE_FLAGS_CHECK_AT_PREP	\
1209 	((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK)
1210 
1211 /*
1212  * Flags stored in the second page of a compound page.  They may overlap
1213  * the CHECK_AT_FREE flags above, so need to be cleared.
1214  */
1215 #define PAGE_FLAGS_SECOND						\
1216 	(0xffUL /* order */		| 1UL << PG_has_hwpoisoned |	\
1217 	 1UL << PG_large_rmappable	| 1UL << PG_partially_mapped)
1218 
1219 #define PAGE_FLAGS_PRIVATE				\
1220 	(1UL << PG_private | 1UL << PG_private_2)
1221 /**
1222  * folio_has_private - Determine if folio has private stuff
1223  * @folio: The folio to be checked
1224  *
1225  * Determine if a folio has private stuff, indicating that release routines
1226  * should be invoked upon it.
1227  */
folio_has_private(const struct folio * folio)1228 static inline int folio_has_private(const struct folio *folio)
1229 {
1230 	return !!(folio->flags & PAGE_FLAGS_PRIVATE);
1231 }
1232 
folio_test_large_maybe_mapped_shared(const struct folio * folio)1233 static inline bool folio_test_large_maybe_mapped_shared(const struct folio *folio)
1234 {
1235 	return test_bit(FOLIO_MM_IDS_SHARED_BITNUM, &folio->_mm_ids);
1236 }
1237 #undef PF_ANY
1238 #undef PF_HEAD
1239 #undef PF_NO_TAIL
1240 #undef PF_NO_COMPOUND
1241 #undef PF_SECOND
1242 #endif /* !__GENERATING_BOUNDS_H */
1243 
1244 #endif	/* PAGE_FLAGS_H */
1245