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 #ifdef CONFIG_MIGRATION
171 /* movable_ops page that is isolated for migration */
172 PG_movable_ops_isolated = PG_reclaim,
173 /* this is a movable_ops page (for selected typed pages only) */
174 PG_movable_ops = PG_uptodate,
175 #endif
176
177 /* Only valid for buddy pages. Used to track pages that are reported */
178 PG_reported = PG_uptodate,
179
180 #ifdef CONFIG_MEMORY_HOTPLUG
181 /* For self-hosted memmap pages */
182 PG_vmemmap_self_hosted = PG_owner_priv_1,
183 #endif
184
185 /*
186 * Flags only valid for compound pages. Stored in first tail page's
187 * flags word. Cannot use the first 8 flags or any flag marked as
188 * PF_ANY.
189 */
190
191 /* At least one page in this folio has the hwpoison flag set */
192 PG_has_hwpoisoned = PG_active,
193 PG_large_rmappable = PG_workingset, /* anon or file-backed */
194 PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */
195 };
196
197 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
198
199 #ifndef __GENERATING_BOUNDS_H
200
201 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
202 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key);
203
204 /*
205 * Return the real head page struct iff the @page is a fake head page, otherwise
206 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst.
207 */
page_fixed_fake_head(const struct page * page)208 static __always_inline const struct page *page_fixed_fake_head(const struct page *page)
209 {
210 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key))
211 return page;
212
213 /*
214 * Only addresses aligned with PAGE_SIZE of struct page may be fake head
215 * struct page. The alignment check aims to avoid access the fields (
216 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly)
217 * cold cacheline in some cases.
218 */
219 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) &&
220 test_bit(PG_head, &page->flags)) {
221 /*
222 * We can safely access the field of the @page[1] with PG_head
223 * because the @page is a compound page composed with at least
224 * two contiguous pages.
225 */
226 unsigned long head = READ_ONCE(page[1].compound_head);
227
228 if (likely(head & 1))
229 return (const struct page *)(head - 1);
230 }
231 return page;
232 }
233
page_count_writable(const struct page * page,int u)234 static __always_inline bool page_count_writable(const struct page *page, int u)
235 {
236 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key))
237 return true;
238
239 /*
240 * The refcount check is ordered before the fake-head check to prevent
241 * the following race:
242 * CPU 1 (HVO) CPU 2 (speculative PFN walker)
243 *
244 * page_ref_freeze()
245 * synchronize_rcu()
246 * rcu_read_lock()
247 * page_is_fake_head() is false
248 * vmemmap_remap_pte()
249 * XXX: struct page[] becomes r/o
250 *
251 * page_ref_unfreeze()
252 * page_ref_count() is not zero
253 *
254 * atomic_add_unless(&page->_refcount)
255 * XXX: try to modify r/o struct page[]
256 *
257 * The refcount check also prevents modification attempts to other (r/o)
258 * tail pages that are not fake heads.
259 */
260 if (atomic_read_acquire(&page->_refcount) == u)
261 return false;
262
263 return page_fixed_fake_head(page) == page;
264 }
265 #else
page_fixed_fake_head(const struct page * page)266 static inline const struct page *page_fixed_fake_head(const struct page *page)
267 {
268 return page;
269 }
270
page_count_writable(const struct page * page,int u)271 static inline bool page_count_writable(const struct page *page, int u)
272 {
273 return true;
274 }
275 #endif
276
page_is_fake_head(const struct page * page)277 static __always_inline int page_is_fake_head(const struct page *page)
278 {
279 return page_fixed_fake_head(page) != page;
280 }
281
_compound_head(const struct page * page)282 static __always_inline unsigned long _compound_head(const struct page *page)
283 {
284 unsigned long head = READ_ONCE(page->compound_head);
285
286 if (unlikely(head & 1))
287 return head - 1;
288 return (unsigned long)page_fixed_fake_head(page);
289 }
290
291 #define compound_head(page) ((typeof(page))_compound_head(page))
292
293 /**
294 * page_folio - Converts from page to folio.
295 * @p: The page.
296 *
297 * Every page is part of a folio. This function cannot be called on a
298 * NULL pointer.
299 *
300 * Context: No reference, nor lock is required on @page. If the caller
301 * does not hold a reference, this call may race with a folio split, so
302 * it should re-check the folio still contains this page after gaining
303 * a reference on the folio.
304 * Return: The folio which contains this page.
305 */
306 #define page_folio(p) (_Generic((p), \
307 const struct page *: (const struct folio *)_compound_head(p), \
308 struct page *: (struct folio *)_compound_head(p)))
309
310 /**
311 * folio_page - Return a page from a folio.
312 * @folio: The folio.
313 * @n: The page number to return.
314 *
315 * @n is relative to the start of the folio. This function does not
316 * check that the page number lies within @folio; the caller is presumed
317 * to have a reference to the page.
318 */
319 #define folio_page(folio, n) nth_page(&(folio)->page, n)
320
PageTail(const struct page * page)321 static __always_inline int PageTail(const struct page *page)
322 {
323 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page);
324 }
325
PageCompound(const struct page * page)326 static __always_inline int PageCompound(const struct page *page)
327 {
328 return test_bit(PG_head, &page->flags) ||
329 READ_ONCE(page->compound_head) & 1;
330 }
331
332 #define PAGE_POISON_PATTERN -1l
PagePoisoned(const struct page * page)333 static inline int PagePoisoned(const struct page *page)
334 {
335 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN;
336 }
337
338 #ifdef CONFIG_DEBUG_VM
339 void page_init_poison(struct page *page, size_t size);
340 #else
page_init_poison(struct page * page,size_t size)341 static inline void page_init_poison(struct page *page, size_t size)
342 {
343 }
344 #endif
345
const_folio_flags(const struct folio * folio,unsigned n)346 static const unsigned long *const_folio_flags(const struct folio *folio,
347 unsigned n)
348 {
349 const struct page *page = &folio->page;
350
351 VM_BUG_ON_PGFLAGS(page->compound_head & 1, page);
352 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
353 return &page[n].flags;
354 }
355
folio_flags(struct folio * folio,unsigned n)356 static unsigned long *folio_flags(struct folio *folio, unsigned n)
357 {
358 struct page *page = &folio->page;
359
360 VM_BUG_ON_PGFLAGS(page->compound_head & 1, page);
361 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
362 return &page[n].flags;
363 }
364
365 /*
366 * Page flags policies wrt compound pages
367 *
368 * PF_POISONED_CHECK
369 * check if this struct page poisoned/uninitialized
370 *
371 * PF_ANY:
372 * the page flag is relevant for small, head and tail pages.
373 *
374 * PF_HEAD:
375 * for compound page all operations related to the page flag applied to
376 * head page.
377 *
378 * PF_NO_TAIL:
379 * modifications of the page flag must be done on small or head pages,
380 * checks can be done on tail pages too.
381 *
382 * PF_NO_COMPOUND:
383 * the page flag is not relevant for compound pages.
384 *
385 * PF_SECOND:
386 * the page flag is stored in the first tail page.
387 */
388 #define PF_POISONED_CHECK(page) ({ \
389 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
390 page; })
391 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
392 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
393 #define PF_NO_TAIL(page, enforce) ({ \
394 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
395 PF_POISONED_CHECK(compound_head(page)); })
396 #define PF_NO_COMPOUND(page, enforce) ({ \
397 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
398 PF_POISONED_CHECK(page); })
399 #define PF_SECOND(page, enforce) ({ \
400 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
401 PF_POISONED_CHECK(&page[1]); })
402
403 /* Which page is the flag stored in */
404 #define FOLIO_PF_ANY 0
405 #define FOLIO_PF_HEAD 0
406 #define FOLIO_PF_NO_TAIL 0
407 #define FOLIO_PF_NO_COMPOUND 0
408 #define FOLIO_PF_SECOND 1
409
410 #define FOLIO_HEAD_PAGE 0
411 #define FOLIO_SECOND_PAGE 1
412
413 /*
414 * Macros to create function definitions for page flags
415 */
416 #define FOLIO_TEST_FLAG(name, page) \
417 static __always_inline bool folio_test_##name(const struct folio *folio) \
418 { return test_bit(PG_##name, const_folio_flags(folio, page)); }
419
420 #define FOLIO_SET_FLAG(name, page) \
421 static __always_inline void folio_set_##name(struct folio *folio) \
422 { set_bit(PG_##name, folio_flags(folio, page)); }
423
424 #define FOLIO_CLEAR_FLAG(name, page) \
425 static __always_inline void folio_clear_##name(struct folio *folio) \
426 { clear_bit(PG_##name, folio_flags(folio, page)); }
427
428 #define __FOLIO_SET_FLAG(name, page) \
429 static __always_inline void __folio_set_##name(struct folio *folio) \
430 { __set_bit(PG_##name, folio_flags(folio, page)); }
431
432 #define __FOLIO_CLEAR_FLAG(name, page) \
433 static __always_inline void __folio_clear_##name(struct folio *folio) \
434 { __clear_bit(PG_##name, folio_flags(folio, page)); }
435
436 #define FOLIO_TEST_SET_FLAG(name, page) \
437 static __always_inline bool folio_test_set_##name(struct folio *folio) \
438 { return test_and_set_bit(PG_##name, folio_flags(folio, page)); }
439
440 #define FOLIO_TEST_CLEAR_FLAG(name, page) \
441 static __always_inline bool folio_test_clear_##name(struct folio *folio) \
442 { return test_and_clear_bit(PG_##name, folio_flags(folio, page)); }
443
444 #define FOLIO_FLAG(name, page) \
445 FOLIO_TEST_FLAG(name, page) \
446 FOLIO_SET_FLAG(name, page) \
447 FOLIO_CLEAR_FLAG(name, page)
448
449 #define TESTPAGEFLAG(uname, lname, policy) \
450 FOLIO_TEST_FLAG(lname, FOLIO_##policy) \
451 static __always_inline int Page##uname(const struct page *page) \
452 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
453
454 #define SETPAGEFLAG(uname, lname, policy) \
455 FOLIO_SET_FLAG(lname, FOLIO_##policy) \
456 static __always_inline void SetPage##uname(struct page *page) \
457 { set_bit(PG_##lname, &policy(page, 1)->flags); }
458
459 #define CLEARPAGEFLAG(uname, lname, policy) \
460 FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \
461 static __always_inline void ClearPage##uname(struct page *page) \
462 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
463
464 #define __SETPAGEFLAG(uname, lname, policy) \
465 __FOLIO_SET_FLAG(lname, FOLIO_##policy) \
466 static __always_inline void __SetPage##uname(struct page *page) \
467 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
468
469 #define __CLEARPAGEFLAG(uname, lname, policy) \
470 __FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \
471 static __always_inline void __ClearPage##uname(struct page *page) \
472 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
473
474 #define TESTSETFLAG(uname, lname, policy) \
475 FOLIO_TEST_SET_FLAG(lname, FOLIO_##policy) \
476 static __always_inline int TestSetPage##uname(struct page *page) \
477 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
478
479 #define TESTCLEARFLAG(uname, lname, policy) \
480 FOLIO_TEST_CLEAR_FLAG(lname, FOLIO_##policy) \
481 static __always_inline int TestClearPage##uname(struct page *page) \
482 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
483
484 #define PAGEFLAG(uname, lname, policy) \
485 TESTPAGEFLAG(uname, lname, policy) \
486 SETPAGEFLAG(uname, lname, policy) \
487 CLEARPAGEFLAG(uname, lname, policy)
488
489 #define __PAGEFLAG(uname, lname, policy) \
490 TESTPAGEFLAG(uname, lname, policy) \
491 __SETPAGEFLAG(uname, lname, policy) \
492 __CLEARPAGEFLAG(uname, lname, policy)
493
494 #define TESTSCFLAG(uname, lname, policy) \
495 TESTSETFLAG(uname, lname, policy) \
496 TESTCLEARFLAG(uname, lname, policy)
497
498 #define FOLIO_TEST_FLAG_FALSE(name) \
499 static inline bool folio_test_##name(const struct folio *folio) \
500 { return false; }
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_SET_FLAG_NOOP(name) \
506 static inline void __folio_set_##name(struct folio *folio) { }
507 #define __FOLIO_CLEAR_FLAG_NOOP(name) \
508 static inline void __folio_clear_##name(struct folio *folio) { }
509 #define FOLIO_TEST_SET_FLAG_FALSE(name) \
510 static inline bool folio_test_set_##name(struct folio *folio) \
511 { return false; }
512 #define FOLIO_TEST_CLEAR_FLAG_FALSE(name) \
513 static inline bool folio_test_clear_##name(struct folio *folio) \
514 { return false; }
515
516 #define FOLIO_FLAG_FALSE(name) \
517 FOLIO_TEST_FLAG_FALSE(name) \
518 FOLIO_SET_FLAG_NOOP(name) \
519 FOLIO_CLEAR_FLAG_NOOP(name)
520
521 #define TESTPAGEFLAG_FALSE(uname, lname) \
522 FOLIO_TEST_FLAG_FALSE(lname) \
523 static inline int Page##uname(const struct page *page) { return 0; }
524
525 #define SETPAGEFLAG_NOOP(uname, lname) \
526 FOLIO_SET_FLAG_NOOP(lname) \
527 static inline void SetPage##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 __CLEARPAGEFLAG_NOOP(uname, lname) \
534 __FOLIO_CLEAR_FLAG_NOOP(lname) \
535 static inline void __ClearPage##uname(struct page *page) { }
536
537 #define TESTSETFLAG_FALSE(uname, lname) \
538 FOLIO_TEST_SET_FLAG_FALSE(lname) \
539 static inline int TestSetPage##uname(struct page *page) { return 0; }
540
541 #define TESTCLEARFLAG_FALSE(uname, lname) \
542 FOLIO_TEST_CLEAR_FLAG_FALSE(lname) \
543 static inline int TestClearPage##uname(struct page *page) { return 0; }
544
545 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \
546 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
547
548 #define TESTSCFLAG_FALSE(uname, lname) \
549 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
550
551 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
552 FOLIO_FLAG(waiters, FOLIO_HEAD_PAGE)
553 FOLIO_FLAG(referenced, FOLIO_HEAD_PAGE)
554 FOLIO_TEST_CLEAR_FLAG(referenced, FOLIO_HEAD_PAGE)
555 __FOLIO_SET_FLAG(referenced, FOLIO_HEAD_PAGE)
556 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
557 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
558 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
559 TESTCLEARFLAG(LRU, lru, PF_HEAD)
560 FOLIO_FLAG(active, FOLIO_HEAD_PAGE)
561 __FOLIO_CLEAR_FLAG(active, FOLIO_HEAD_PAGE)
562 FOLIO_TEST_CLEAR_FLAG(active, FOLIO_HEAD_PAGE)
563 PAGEFLAG(Workingset, workingset, PF_HEAD)
564 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
565 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
566
567 /* Xen */
568 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
569 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
570 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
571 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
PAGEFLAG(XenRemapped,xen_remapped,PF_NO_COMPOUND)572 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
573 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
574
575 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
576 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
577 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
578 FOLIO_FLAG(swapbacked, FOLIO_HEAD_PAGE)
579 __FOLIO_CLEAR_FLAG(swapbacked, FOLIO_HEAD_PAGE)
580 __FOLIO_SET_FLAG(swapbacked, FOLIO_HEAD_PAGE)
581
582 /*
583 * Private page markings that may be used by the filesystem that owns the page
584 * for its own purposes.
585 * - PG_private and PG_private_2 cause release_folio() and co to be invoked
586 */
587 PAGEFLAG(Private, private, PF_ANY)
588 FOLIO_FLAG(private_2, FOLIO_HEAD_PAGE)
589
590 /* owner_2 can be set on tail pages for anon memory */
591 FOLIO_FLAG(owner_2, FOLIO_HEAD_PAGE)
592
593 /*
594 * Only test-and-set exist for PG_writeback. The unconditional operators are
595 * risky: they bypass page accounting.
596 */
597 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
598 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
599 FOLIO_FLAG(mappedtodisk, FOLIO_HEAD_PAGE)
600
601 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
602 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
603 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
604 FOLIO_FLAG(readahead, FOLIO_HEAD_PAGE)
605 FOLIO_TEST_CLEAR_FLAG(readahead, FOLIO_HEAD_PAGE)
606
607 FOLIO_FLAG(dropbehind, FOLIO_HEAD_PAGE)
608 FOLIO_TEST_CLEAR_FLAG(dropbehind, FOLIO_HEAD_PAGE)
609 __FOLIO_SET_FLAG(dropbehind, FOLIO_HEAD_PAGE)
610
611 #ifdef CONFIG_HIGHMEM
612 /*
613 * Must use a macro here due to header dependency issues. page_zone() is not
614 * available at this point.
615 */
616 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
617 #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f))
618 #else
619 PAGEFLAG_FALSE(HighMem, highmem)
620 #endif
621
622 /* Does kmap_local_folio() only allow access to one page of the folio? */
623 #ifdef CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP
624 #define folio_test_partial_kmap(f) true
625 #else
626 #define folio_test_partial_kmap(f) folio_test_highmem(f)
627 #endif
628
629 #ifdef CONFIG_SWAP
630 static __always_inline bool folio_test_swapcache(const struct folio *folio)
631 {
632 return folio_test_swapbacked(folio) &&
633 test_bit(PG_swapcache, const_folio_flags(folio, 0));
634 }
635
FOLIO_SET_FLAG(swapcache,FOLIO_HEAD_PAGE)636 FOLIO_SET_FLAG(swapcache, FOLIO_HEAD_PAGE)
637 FOLIO_CLEAR_FLAG(swapcache, FOLIO_HEAD_PAGE)
638 #else
639 FOLIO_FLAG_FALSE(swapcache)
640 #endif
641
642 FOLIO_FLAG(unevictable, FOLIO_HEAD_PAGE)
643 __FOLIO_CLEAR_FLAG(unevictable, FOLIO_HEAD_PAGE)
644 FOLIO_TEST_CLEAR_FLAG(unevictable, FOLIO_HEAD_PAGE)
645
646 #ifdef CONFIG_MMU
647 FOLIO_FLAG(mlocked, FOLIO_HEAD_PAGE)
648 __FOLIO_CLEAR_FLAG(mlocked, FOLIO_HEAD_PAGE)
649 FOLIO_TEST_CLEAR_FLAG(mlocked, FOLIO_HEAD_PAGE)
650 FOLIO_TEST_SET_FLAG(mlocked, FOLIO_HEAD_PAGE)
651 #else
652 FOLIO_FLAG_FALSE(mlocked)
653 __FOLIO_CLEAR_FLAG_NOOP(mlocked)
654 FOLIO_TEST_CLEAR_FLAG_FALSE(mlocked)
655 FOLIO_TEST_SET_FLAG_FALSE(mlocked)
656 #endif
657
658 #ifdef CONFIG_MEMORY_FAILURE
659 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
660 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
661 #define __PG_HWPOISON (1UL << PG_hwpoison)
662 #else
663 PAGEFLAG_FALSE(HWPoison, hwpoison)
664 #define __PG_HWPOISON 0
665 #endif
666
667 #ifdef CONFIG_PAGE_IDLE_FLAG
668 #ifdef CONFIG_64BIT
669 FOLIO_TEST_FLAG(young, FOLIO_HEAD_PAGE)
670 FOLIO_SET_FLAG(young, FOLIO_HEAD_PAGE)
671 FOLIO_TEST_CLEAR_FLAG(young, FOLIO_HEAD_PAGE)
672 FOLIO_FLAG(idle, FOLIO_HEAD_PAGE)
673 #endif
674 /* See page_idle.h for !64BIT workaround */
675 #else /* !CONFIG_PAGE_IDLE_FLAG */
676 FOLIO_FLAG_FALSE(young)
677 FOLIO_TEST_CLEAR_FLAG_FALSE(young)
678 FOLIO_FLAG_FALSE(idle)
679 #endif
680
681 /*
682 * PageReported() is used to track reported free pages within the Buddy
683 * allocator. We can use the non-atomic version of the test and set
684 * operations as both should be shielded with the zone lock to prevent
685 * any possible races on the setting or clearing of the bit.
686 */
687 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
688
689 #ifdef CONFIG_MEMORY_HOTPLUG
690 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
691 #else
692 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
693 #endif
694
695 /*
696 * On an anonymous folio mapped into a user virtual memory area,
697 * folio->mapping points to its anon_vma, not to a struct address_space;
698 * with the FOLIO_MAPPING_ANON bit set to distinguish it. See rmap.h.
699 *
700 * On an anonymous folio in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
701 * the FOLIO_MAPPING_ANON_KSM bit may be set along with the FOLIO_MAPPING_ANON
702 * bit; and then folio->mapping points, not to an anon_vma, but to a private
703 * structure which KSM associates with that merged folio. See ksm.h.
704 *
705 * Please note that, confusingly, "folio_mapping" refers to the inode
706 * address_space which maps the folio from disk; whereas "folio_mapped"
707 * refers to user virtual address space into which the folio is mapped.
708 *
709 * For slab pages, since slab reuses the bits in struct page to store its
710 * internal states, the folio->mapping does not exist as such, nor do
711 * these flags below. So in order to avoid testing non-existent bits,
712 * please make sure that folio_test_slab(folio) actually evaluates to
713 * false before calling the following functions (e.g., folio_test_anon).
714 * See mm/slab.h.
715 */
716 #define FOLIO_MAPPING_ANON 0x1
717 #define FOLIO_MAPPING_ANON_KSM 0x2
718 #define FOLIO_MAPPING_KSM (FOLIO_MAPPING_ANON | FOLIO_MAPPING_ANON_KSM)
719 #define FOLIO_MAPPING_FLAGS (FOLIO_MAPPING_ANON | FOLIO_MAPPING_ANON_KSM)
720
721 static __always_inline bool folio_test_anon(const struct folio *folio)
722 {
723 return ((unsigned long)folio->mapping & FOLIO_MAPPING_ANON) != 0;
724 }
725
PageAnonNotKsm(const struct page * page)726 static __always_inline bool PageAnonNotKsm(const struct page *page)
727 {
728 unsigned long flags = (unsigned long)page_folio(page)->mapping;
729
730 return (flags & FOLIO_MAPPING_FLAGS) == FOLIO_MAPPING_ANON;
731 }
732
PageAnon(const struct page * page)733 static __always_inline bool PageAnon(const struct page *page)
734 {
735 return folio_test_anon(page_folio(page));
736 }
737 #ifdef CONFIG_KSM
738 /*
739 * A KSM page is one of those write-protected "shared pages" or "merged pages"
740 * which KSM maps into multiple mms, wherever identical anonymous page content
741 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
742 * anon_vma, but to that page's node of the stable tree.
743 */
folio_test_ksm(const struct folio * folio)744 static __always_inline bool folio_test_ksm(const struct folio *folio)
745 {
746 return ((unsigned long)folio->mapping & FOLIO_MAPPING_FLAGS) ==
747 FOLIO_MAPPING_KSM;
748 }
749 #else
750 FOLIO_TEST_FLAG_FALSE(ksm)
751 #endif
752
753 u64 stable_page_flags(const struct page *page);
754
755 /**
756 * folio_xor_flags_has_waiters - Change some folio flags.
757 * @folio: The folio.
758 * @mask: Bits set in this word will be changed.
759 *
760 * This must only be used for flags which are changed with the folio
761 * lock held. For example, it is unsafe to use for PG_dirty as that
762 * can be set without the folio lock held. It can also only be used
763 * on flags which are in the range 0-6 as some of the implementations
764 * only affect those bits.
765 *
766 * Return: Whether there are tasks waiting on the folio.
767 */
folio_xor_flags_has_waiters(struct folio * folio,unsigned long mask)768 static inline bool folio_xor_flags_has_waiters(struct folio *folio,
769 unsigned long mask)
770 {
771 return xor_unlock_is_negative_byte(mask, folio_flags(folio, 0));
772 }
773
774 /**
775 * folio_test_uptodate - Is this folio up to date?
776 * @folio: The folio.
777 *
778 * The uptodate flag is set on a folio when every byte in the folio is
779 * at least as new as the corresponding bytes on storage. Anonymous
780 * and CoW folios are always uptodate. If the folio is not uptodate,
781 * some of the bytes in it may be; see the is_partially_uptodate()
782 * address_space operation.
783 */
folio_test_uptodate(const struct folio * folio)784 static inline bool folio_test_uptodate(const struct folio *folio)
785 {
786 bool ret = test_bit(PG_uptodate, const_folio_flags(folio, 0));
787 /*
788 * Must ensure that the data we read out of the folio is loaded
789 * _after_ we've loaded folio->flags to check the uptodate bit.
790 * We can skip the barrier if the folio is not uptodate, because
791 * we wouldn't be reading anything from it.
792 *
793 * See folio_mark_uptodate() for the other side of the story.
794 */
795 if (ret)
796 smp_rmb();
797
798 return ret;
799 }
800
PageUptodate(const struct page * page)801 static inline bool PageUptodate(const struct page *page)
802 {
803 return folio_test_uptodate(page_folio(page));
804 }
805
__folio_mark_uptodate(struct folio * folio)806 static __always_inline void __folio_mark_uptodate(struct folio *folio)
807 {
808 smp_wmb();
809 __set_bit(PG_uptodate, folio_flags(folio, 0));
810 }
811
folio_mark_uptodate(struct folio * folio)812 static __always_inline void folio_mark_uptodate(struct folio *folio)
813 {
814 /*
815 * Memory barrier must be issued before setting the PG_uptodate bit,
816 * so that all previous stores issued in order to bring the folio
817 * uptodate are actually visible before folio_test_uptodate becomes true.
818 */
819 smp_wmb();
820 set_bit(PG_uptodate, folio_flags(folio, 0));
821 }
822
__SetPageUptodate(struct page * page)823 static __always_inline void __SetPageUptodate(struct page *page)
824 {
825 __folio_mark_uptodate((struct folio *)page);
826 }
827
SetPageUptodate(struct page * page)828 static __always_inline void SetPageUptodate(struct page *page)
829 {
830 folio_mark_uptodate((struct folio *)page);
831 }
832
833 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
834
835 void __folio_start_writeback(struct folio *folio, bool keep_write);
836 void set_page_writeback(struct page *page);
837
838 #define folio_start_writeback(folio) \
839 __folio_start_writeback(folio, false)
840
folio_test_head(const struct folio * folio)841 static __always_inline bool folio_test_head(const struct folio *folio)
842 {
843 return test_bit(PG_head, const_folio_flags(folio, FOLIO_PF_ANY));
844 }
845
PageHead(const struct page * page)846 static __always_inline int PageHead(const struct page *page)
847 {
848 PF_POISONED_CHECK(page);
849 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page);
850 }
851
__SETPAGEFLAG(Head,head,PF_ANY)852 __SETPAGEFLAG(Head, head, PF_ANY)
853 __CLEARPAGEFLAG(Head, head, PF_ANY)
854 CLEARPAGEFLAG(Head, head, PF_ANY)
855
856 /**
857 * folio_test_large() - Does this folio contain more than one page?
858 * @folio: The folio to test.
859 *
860 * Return: True if the folio is larger than one page.
861 */
862 static inline bool folio_test_large(const struct folio *folio)
863 {
864 return folio_test_head(folio);
865 }
866
set_compound_head(struct page * page,struct page * head)867 static __always_inline void set_compound_head(struct page *page, struct page *head)
868 {
869 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
870 }
871
clear_compound_head(struct page * page)872 static __always_inline void clear_compound_head(struct page *page)
873 {
874 WRITE_ONCE(page->compound_head, 0);
875 }
876
877 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
ClearPageCompound(struct page * page)878 static inline void ClearPageCompound(struct page *page)
879 {
880 BUG_ON(!PageHead(page));
881 ClearPageHead(page);
882 }
FOLIO_FLAG(large_rmappable,FOLIO_SECOND_PAGE)883 FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE)
884 FOLIO_FLAG(partially_mapped, FOLIO_SECOND_PAGE)
885 #else
886 FOLIO_FLAG_FALSE(large_rmappable)
887 FOLIO_FLAG_FALSE(partially_mapped)
888 #endif
889
890 #define PG_head_mask ((1UL << PG_head))
891
892 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
893 /*
894 * PageTransCompound returns true for both transparent huge pages
895 * and hugetlbfs pages, so it should only be called when it's known
896 * that hugetlbfs pages aren't involved.
897 */
898 static inline int PageTransCompound(const struct page *page)
899 {
900 return PageCompound(page);
901 }
902 #else
903 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
904 #endif
905
906 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
907 /*
908 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
909 * compound page.
910 *
911 * This flag is set by hwpoison handler. Cleared by THP split or free page.
912 */
913 FOLIO_FLAG(has_hwpoisoned, FOLIO_SECOND_PAGE)
914 #else
915 FOLIO_FLAG_FALSE(has_hwpoisoned)
916 #endif
917
918 /*
919 * For pages that do not use mapcount, page_type may be used.
920 * The low 24 bits of pagetype may be used for your own purposes, as long
921 * as you are careful to not affect the top 8 bits. The low bits of
922 * pagetype will be overwritten when you clear the page_type from the page.
923 */
924 enum pagetype {
925 /* 0x00-0x7f are positive numbers, ie mapcount */
926 /* Reserve 0x80-0xef for mapcount overflow. */
927 PGTY_buddy = 0xf0,
928 PGTY_offline = 0xf1,
929 PGTY_table = 0xf2,
930 PGTY_guard = 0xf3,
931 PGTY_hugetlb = 0xf4,
932 PGTY_slab = 0xf5,
933 PGTY_zsmalloc = 0xf6,
934 PGTY_unaccepted = 0xf7,
935 PGTY_large_kmalloc = 0xf8,
936
937 PGTY_mapcount_underflow = 0xff
938 };
939
page_type_has_type(int page_type)940 static inline bool page_type_has_type(int page_type)
941 {
942 return page_type < (PGTY_mapcount_underflow << 24);
943 }
944
945 /* This takes a mapcount which is one more than page->_mapcount */
page_mapcount_is_type(unsigned int mapcount)946 static inline bool page_mapcount_is_type(unsigned int mapcount)
947 {
948 return page_type_has_type(mapcount - 1);
949 }
950
page_has_type(const struct page * page)951 static inline bool page_has_type(const struct page *page)
952 {
953 return page_type_has_type(data_race(page->page_type));
954 }
955
956 #define FOLIO_TYPE_OPS(lname, fname) \
957 static __always_inline bool folio_test_##fname(const struct folio *folio) \
958 { \
959 return data_race(folio->page.page_type >> 24) == PGTY_##lname; \
960 } \
961 static __always_inline void __folio_set_##fname(struct folio *folio) \
962 { \
963 if (folio_test_##fname(folio)) \
964 return; \
965 VM_BUG_ON_FOLIO(data_race(folio->page.page_type) != UINT_MAX, \
966 folio); \
967 folio->page.page_type = (unsigned int)PGTY_##lname << 24; \
968 } \
969 static __always_inline void __folio_clear_##fname(struct folio *folio) \
970 { \
971 if (folio->page.page_type == UINT_MAX) \
972 return; \
973 VM_BUG_ON_FOLIO(!folio_test_##fname(folio), folio); \
974 folio->page.page_type = UINT_MAX; \
975 }
976
977 #define PAGE_TYPE_OPS(uname, lname, fname) \
978 FOLIO_TYPE_OPS(lname, fname) \
979 static __always_inline int Page##uname(const struct page *page) \
980 { \
981 return data_race(page->page_type >> 24) == PGTY_##lname; \
982 } \
983 static __always_inline void __SetPage##uname(struct page *page) \
984 { \
985 if (Page##uname(page)) \
986 return; \
987 VM_BUG_ON_PAGE(data_race(page->page_type) != UINT_MAX, page); \
988 page->page_type = (unsigned int)PGTY_##lname << 24; \
989 } \
990 static __always_inline void __ClearPage##uname(struct page *page) \
991 { \
992 if (page->page_type == UINT_MAX) \
993 return; \
994 VM_BUG_ON_PAGE(!Page##uname(page), page); \
995 page->page_type = UINT_MAX; \
996 }
997
998 /*
999 * PageBuddy() indicates that the page is free and in the buddy system
1000 * (see mm/page_alloc.c).
1001 */
1002 PAGE_TYPE_OPS(Buddy, buddy, buddy)
1003
1004 /*
1005 * PageOffline() indicates that the page is logically offline although the
1006 * containing section is online. (e.g. inflated in a balloon driver or
1007 * not onlined when onlining the section).
1008 * The content of these pages is effectively stale. Such pages should not
1009 * be touched (read/write/dump/save) except by their owner.
1010 *
1011 * When a memory block gets onlined, all pages are initialized with a
1012 * refcount of 1 and PageOffline(). generic_online_page() will
1013 * take care of clearing PageOffline().
1014 *
1015 * If a driver wants to allow to offline unmovable PageOffline() pages without
1016 * putting them back to the buddy, it can do so via the memory notifier by
1017 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
1018 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
1019 * pages (now with a reference count of zero) are treated like free (unmanaged)
1020 * pages, allowing the containing memory block to get offlined. A driver that
1021 * relies on this feature is aware that re-onlining the memory block will
1022 * require not giving them to the buddy via generic_online_page().
1023 *
1024 * Memory offlining code will not adjust the managed page count for any
1025 * PageOffline() pages, treating them like they were never exposed to the
1026 * buddy using generic_online_page().
1027 *
1028 * There are drivers that mark a page PageOffline() and expect there won't be
1029 * any further access to page content. PFN walkers that read content of random
1030 * pages should check PageOffline() and synchronize with such drivers using
1031 * page_offline_freeze()/page_offline_thaw().
1032 */
1033 PAGE_TYPE_OPS(Offline, offline, offline)
1034
1035 extern void page_offline_freeze(void);
1036 extern void page_offline_thaw(void);
1037 extern void page_offline_begin(void);
1038 extern void page_offline_end(void);
1039
1040 /*
1041 * Marks pages in use as page tables.
1042 */
PAGE_TYPE_OPS(Table,table,pgtable)1043 PAGE_TYPE_OPS(Table, table, pgtable)
1044
1045 /*
1046 * Marks guardpages used with debug_pagealloc.
1047 */
1048 PAGE_TYPE_OPS(Guard, guard, guard)
1049
1050 FOLIO_TYPE_OPS(slab, slab)
1051
1052 /**
1053 * PageSlab - Determine if the page belongs to the slab allocator
1054 * @page: The page to test.
1055 *
1056 * Context: Any context.
1057 * Return: True for slab pages, false for any other kind of page.
1058 */
1059 static inline bool PageSlab(const struct page *page)
1060 {
1061 return folio_test_slab(page_folio(page));
1062 }
1063
1064 #ifdef CONFIG_HUGETLB_PAGE
FOLIO_TYPE_OPS(hugetlb,hugetlb)1065 FOLIO_TYPE_OPS(hugetlb, hugetlb)
1066 #else
1067 FOLIO_TEST_FLAG_FALSE(hugetlb)
1068 #endif
1069
1070 PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
1071
1072 /*
1073 * Mark pages that has to be accepted before touched for the first time.
1074 *
1075 * Serialized with zone lock.
1076 */
1077 PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
1078 FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
1079
1080 /**
1081 * PageHuge - Determine if the page belongs to hugetlbfs
1082 * @page: The page to test.
1083 *
1084 * Context: Any context.
1085 * Return: True for hugetlbfs pages, false for anon pages or pages
1086 * belonging to other filesystems.
1087 */
1088 static inline bool PageHuge(const struct page *page)
1089 {
1090 return folio_test_hugetlb(page_folio(page));
1091 }
1092
1093 /*
1094 * Check if a page is currently marked HWPoisoned. Note that this check is
1095 * best effort only and inherently racy: there is no way to synchronize with
1096 * failing hardware.
1097 */
is_page_hwpoison(const struct page * page)1098 static inline bool is_page_hwpoison(const struct page *page)
1099 {
1100 const struct folio *folio;
1101
1102 if (PageHWPoison(page))
1103 return true;
1104 folio = page_folio(page);
1105 return folio_test_hugetlb(folio) && PageHWPoison(&folio->page);
1106 }
1107
folio_contain_hwpoisoned_page(struct folio * folio)1108 static inline bool folio_contain_hwpoisoned_page(struct folio *folio)
1109 {
1110 return folio_test_hwpoison(folio) ||
1111 (folio_test_large(folio) && folio_test_has_hwpoisoned(folio));
1112 }
1113
1114 bool is_free_buddy_page(const struct page *page);
1115
1116 #ifdef CONFIG_MIGRATION
1117 /*
1118 * This page is migratable through movable_ops (for selected typed pages
1119 * only).
1120 *
1121 * Page migration of such pages might fail, for example, if the page is
1122 * already isolated by somebody else, or if the page is about to get freed.
1123 *
1124 * While a subsystem might set selected typed pages that support page migration
1125 * as being movable through movable_ops, it must never clear this flag.
1126 *
1127 * This flag is only cleared when the page is freed back to the buddy.
1128 *
1129 * Only selected page types support this flag (see page_movable_ops()) and
1130 * the flag might be used in other context for other pages. Always use
1131 * page_has_movable_ops() instead.
1132 */
1133 TESTPAGEFLAG(MovableOps, movable_ops, PF_NO_TAIL);
1134 SETPAGEFLAG(MovableOps, movable_ops, PF_NO_TAIL);
1135 /*
1136 * A movable_ops page has this flag set while it is isolated for migration.
1137 * This flag primarily protects against concurrent migration attempts.
1138 *
1139 * Once migration ended (success or failure), the flag is cleared. The
1140 * flag is managed by the migration core.
1141 */
1142 PAGEFLAG(MovableOpsIsolated, movable_ops_isolated, PF_NO_TAIL);
1143 #else /* !CONFIG_MIGRATION */
1144 TESTPAGEFLAG_FALSE(MovableOps, movable_ops);
1145 SETPAGEFLAG_NOOP(MovableOps, movable_ops);
1146 PAGEFLAG_FALSE(MovableOpsIsolated, movable_ops_isolated);
1147 #endif /* CONFIG_MIGRATION */
1148
1149 /**
1150 * page_has_movable_ops - test for a movable_ops page
1151 * @page: The page to test.
1152 *
1153 * Test whether this is a movable_ops page. Such pages will stay that
1154 * way until freed.
1155 *
1156 * Returns true if this is a movable_ops page, otherwise false.
1157 */
page_has_movable_ops(const struct page * page)1158 static inline bool page_has_movable_ops(const struct page *page)
1159 {
1160 return PageMovableOps(page) &&
1161 (PageOffline(page) || PageZsmalloc(page));
1162 }
1163
PageAnonExclusive(const struct page * page)1164 static __always_inline int PageAnonExclusive(const struct page *page)
1165 {
1166 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1167 /*
1168 * HugeTLB stores this information on the head page; THP keeps it per
1169 * page
1170 */
1171 if (PageHuge(page))
1172 page = compound_head(page);
1173 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1174 }
1175
SetPageAnonExclusive(struct page * page)1176 static __always_inline void SetPageAnonExclusive(struct page *page)
1177 {
1178 VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page);
1179 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1180 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1181 }
1182
ClearPageAnonExclusive(struct page * page)1183 static __always_inline void ClearPageAnonExclusive(struct page *page)
1184 {
1185 VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page);
1186 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1187 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1188 }
1189
__ClearPageAnonExclusive(struct page * page)1190 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1191 {
1192 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1193 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1194 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1195 }
1196
1197 #ifdef CONFIG_MMU
1198 #define __PG_MLOCKED (1UL << PG_mlocked)
1199 #else
1200 #define __PG_MLOCKED 0
1201 #endif
1202
1203 /*
1204 * Flags checked when a page is freed. Pages being freed should not have
1205 * these flags set. If they are, there is a problem.
1206 */
1207 #define PAGE_FLAGS_CHECK_AT_FREE \
1208 (1UL << PG_lru | 1UL << PG_locked | \
1209 1UL << PG_private | 1UL << PG_private_2 | \
1210 1UL << PG_writeback | 1UL << PG_reserved | \
1211 1UL << PG_active | \
1212 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK)
1213
1214 /*
1215 * Flags checked when a page is prepped for return by the page allocator.
1216 * Pages being prepped should not have these flags set. If they are set,
1217 * there has been a kernel bug or struct page corruption.
1218 *
1219 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1220 * alloc-free cycle to prevent from reusing the page.
1221 */
1222 #define PAGE_FLAGS_CHECK_AT_PREP \
1223 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK)
1224
1225 /*
1226 * Flags stored in the second page of a compound page. They may overlap
1227 * the CHECK_AT_FREE flags above, so need to be cleared.
1228 */
1229 #define PAGE_FLAGS_SECOND \
1230 (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \
1231 1UL << PG_large_rmappable | 1UL << PG_partially_mapped)
1232
1233 #define PAGE_FLAGS_PRIVATE \
1234 (1UL << PG_private | 1UL << PG_private_2)
1235 /**
1236 * folio_has_private - Determine if folio has private stuff
1237 * @folio: The folio to be checked
1238 *
1239 * Determine if a folio has private stuff, indicating that release routines
1240 * should be invoked upon it.
1241 */
folio_has_private(const struct folio * folio)1242 static inline int folio_has_private(const struct folio *folio)
1243 {
1244 return !!(folio->flags & PAGE_FLAGS_PRIVATE);
1245 }
1246
1247 #undef PF_ANY
1248 #undef PF_HEAD
1249 #undef PF_NO_TAIL
1250 #undef PF_NO_COMPOUND
1251 #undef PF_SECOND
1252 #endif /* !__GENERATING_BOUNDS_H */
1253
1254 #endif /* PAGE_FLAGS_H */
1255