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 /*
202 * For tail pages, if the size of struct page is power-of-2 ->compound_info
203 * encodes the mask that converts the address of the tail page address to
204 * the head page address.
205 *
206 * Otherwise, ->compound_info has direct pointer to head pages.
207 */
compound_info_has_mask(void)208 static __always_inline bool compound_info_has_mask(void)
209 {
210 /*
211 * Limit mask usage to HugeTLB vmemmap optimization (HVO) where it
212 * makes a difference.
213 *
214 * The approach with mask would work in the wider set of conditions,
215 * but it requires validating that struct pages are naturally aligned
216 * for all orders up to the MAX_FOLIO_ORDER, which can be tricky.
217 */
218 if (!IS_ENABLED(CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP))
219 return false;
220
221 return is_power_of_2(sizeof(struct page));
222 }
223
_compound_head(const struct page * page)224 static __always_inline unsigned long _compound_head(const struct page *page)
225 {
226 unsigned long info = READ_ONCE(page->compound_info);
227 unsigned long mask;
228
229 if (!compound_info_has_mask()) {
230 /* Bit 0 encodes PageTail() */
231 if (info & 1)
232 return info - 1;
233
234 return (unsigned long)page;
235 }
236
237 /*
238 * If compound_info_has_mask() is true the rest of the info encodes
239 * the mask that converts the address of the tail page to the head page.
240 *
241 * No need to clear bit 0 in the mask as 'page' always has it clear.
242 *
243 * Let's do it in a branchless manner.
244 */
245
246 /* Non-tail: -1UL, Tail: 0 */
247 mask = (info & 1) - 1;
248
249 /* Non-tail: -1UL, Tail: info */
250 mask |= info;
251
252 return (unsigned long)page & mask;
253 }
254
255 #define compound_head(page) ((typeof(page))_compound_head(page))
256
set_compound_head(struct page * tail,const struct page * head,unsigned int order)257 static __always_inline void set_compound_head(struct page *tail,
258 const struct page *head, unsigned int order)
259 {
260 unsigned int shift;
261 unsigned long mask;
262
263 if (!compound_info_has_mask()) {
264 WRITE_ONCE(tail->compound_info, (unsigned long)head | 1);
265 return;
266 }
267
268 /*
269 * If the size of struct page is power-of-2, bits [shift:0] of the
270 * virtual address of compound head are zero.
271 *
272 * Calculate mask that can be applied to the virtual address of
273 * the tail page to get address of the head page.
274 */
275 shift = order + order_base_2(sizeof(struct page));
276 mask = GENMASK(BITS_PER_LONG - 1, shift);
277
278 /* Bit 0 encodes PageTail() */
279 WRITE_ONCE(tail->compound_info, mask | 1);
280 }
281
clear_compound_head(struct page * page)282 static __always_inline void clear_compound_head(struct page *page)
283 {
284 WRITE_ONCE(page->compound_info, 0);
285 }
286
287 /**
288 * page_folio - Converts from page to folio.
289 * @p: The page.
290 *
291 * Every page is part of a folio. This function cannot be called on a
292 * NULL pointer.
293 *
294 * Context: No reference, nor lock is required on @page. If the caller
295 * does not hold a reference, this call may race with a folio split, so
296 * it should re-check the folio still contains this page after gaining
297 * a reference on the folio.
298 * Return: The folio which contains this page.
299 */
300 #define page_folio(p) (_Generic((p), \
301 const struct page *: (const struct folio *)_compound_head(p), \
302 struct page *: (struct folio *)_compound_head(p)))
303
304 /**
305 * folio_page - Return a page from a folio.
306 * @folio: The folio.
307 * @n: The page number to return.
308 *
309 * @n is relative to the start of the folio. This function does not
310 * check that the page number lies within @folio; the caller is presumed
311 * to have a reference to the page.
312 */
313 #define folio_page(folio, n) (&(folio)->page + (n))
314
PageTail(const struct page * page)315 static __always_inline int PageTail(const struct page *page)
316 {
317 return READ_ONCE(page->compound_info) & 1;
318 }
319
PageCompound(const struct page * page)320 static __always_inline int PageCompound(const struct page *page)
321 {
322 return test_bit(PG_head, &page->flags.f) ||
323 READ_ONCE(page->compound_info) & 1;
324 }
325
326 #define PAGE_POISON_PATTERN -1l
PagePoisoned(const struct page * page)327 static inline int PagePoisoned(const struct page *page)
328 {
329 return READ_ONCE(page->flags.f) == PAGE_POISON_PATTERN;
330 }
331
332 #ifdef CONFIG_DEBUG_VM
333 void page_init_poison(struct page *page, size_t size);
334 #else
page_init_poison(struct page * page,size_t size)335 static inline void page_init_poison(struct page *page, size_t size)
336 {
337 }
338 #endif
339
const_folio_flags(const struct folio * folio,unsigned n)340 static const unsigned long *const_folio_flags(const struct folio *folio,
341 unsigned n)
342 {
343 const struct page *page = &folio->page;
344
345 VM_BUG_ON_PGFLAGS(page->compound_info & 1, page);
346 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags.f), page);
347 return &page[n].flags.f;
348 }
349
folio_flags(struct folio * folio,unsigned n)350 static unsigned long *folio_flags(struct folio *folio, unsigned n)
351 {
352 struct page *page = &folio->page;
353
354 VM_BUG_ON_PGFLAGS(page->compound_info & 1, page);
355 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags.f), page);
356 return &page[n].flags.f;
357 }
358
359 /*
360 * Page flags policies wrt compound pages
361 *
362 * PF_POISONED_CHECK
363 * check if this struct page poisoned/uninitialized
364 *
365 * PF_ANY:
366 * the page flag is relevant for small, head and tail pages.
367 *
368 * PF_HEAD:
369 * for compound page all operations related to the page flag applied to
370 * head page.
371 *
372 * PF_NO_TAIL:
373 * modifications of the page flag must be done on small or head pages,
374 * checks can be done on tail pages too.
375 *
376 * PF_NO_COMPOUND:
377 * the page flag is not relevant for compound pages.
378 *
379 * PF_SECOND:
380 * the page flag is stored in the first tail page.
381 */
382 #define PF_POISONED_CHECK(page) ({ \
383 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
384 page; })
385 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
386 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
387 #define PF_NO_TAIL(page, enforce) ({ \
388 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
389 PF_POISONED_CHECK(compound_head(page)); })
390 #define PF_NO_COMPOUND(page, enforce) ({ \
391 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
392 PF_POISONED_CHECK(page); })
393 #define PF_SECOND(page, enforce) ({ \
394 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
395 PF_POISONED_CHECK(&page[1]); })
396
397 /* Which page is the flag stored in */
398 #define FOLIO_PF_ANY 0
399 #define FOLIO_PF_HEAD 0
400 #define FOLIO_PF_NO_TAIL 0
401 #define FOLIO_PF_NO_COMPOUND 0
402 #define FOLIO_PF_SECOND 1
403
404 #define FOLIO_HEAD_PAGE 0
405 #define FOLIO_SECOND_PAGE 1
406
407 /*
408 * Macros to create function definitions for page flags
409 */
410 #define FOLIO_TEST_FLAG(name, page) \
411 static __always_inline bool folio_test_##name(const struct folio *folio) \
412 { return test_bit(PG_##name, const_folio_flags(folio, page)); }
413
414 #define FOLIO_SET_FLAG(name, page) \
415 static __always_inline void folio_set_##name(struct folio *folio) \
416 { set_bit(PG_##name, folio_flags(folio, page)); }
417
418 #define FOLIO_CLEAR_FLAG(name, page) \
419 static __always_inline void folio_clear_##name(struct folio *folio) \
420 { clear_bit(PG_##name, folio_flags(folio, page)); }
421
422 #define __FOLIO_SET_FLAG(name, page) \
423 static __always_inline void __folio_set_##name(struct folio *folio) \
424 { __set_bit(PG_##name, folio_flags(folio, page)); }
425
426 #define __FOLIO_CLEAR_FLAG(name, page) \
427 static __always_inline void __folio_clear_##name(struct folio *folio) \
428 { __clear_bit(PG_##name, folio_flags(folio, page)); }
429
430 #define FOLIO_TEST_SET_FLAG(name, page) \
431 static __always_inline bool folio_test_set_##name(struct folio *folio) \
432 { return test_and_set_bit(PG_##name, folio_flags(folio, page)); }
433
434 #define FOLIO_TEST_CLEAR_FLAG(name, page) \
435 static __always_inline bool folio_test_clear_##name(struct folio *folio) \
436 { return test_and_clear_bit(PG_##name, folio_flags(folio, page)); }
437
438 #define FOLIO_FLAG(name, page) \
439 FOLIO_TEST_FLAG(name, page) \
440 FOLIO_SET_FLAG(name, page) \
441 FOLIO_CLEAR_FLAG(name, page)
442
443 #define TESTPAGEFLAG(uname, lname, policy) \
444 FOLIO_TEST_FLAG(lname, FOLIO_##policy) \
445 static __always_inline int Page##uname(const struct page *page) \
446 { return test_bit(PG_##lname, &policy(page, 0)->flags.f); }
447
448 #define SETPAGEFLAG(uname, lname, policy) \
449 FOLIO_SET_FLAG(lname, FOLIO_##policy) \
450 static __always_inline void SetPage##uname(struct page *page) \
451 { set_bit(PG_##lname, &policy(page, 1)->flags.f); }
452
453 #define CLEARPAGEFLAG(uname, lname, policy) \
454 FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \
455 static __always_inline void ClearPage##uname(struct page *page) \
456 { clear_bit(PG_##lname, &policy(page, 1)->flags.f); }
457
458 #define __SETPAGEFLAG(uname, lname, policy) \
459 __FOLIO_SET_FLAG(lname, FOLIO_##policy) \
460 static __always_inline void __SetPage##uname(struct page *page) \
461 { __set_bit(PG_##lname, &policy(page, 1)->flags.f); }
462
463 #define __CLEARPAGEFLAG(uname, lname, policy) \
464 __FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \
465 static __always_inline void __ClearPage##uname(struct page *page) \
466 { __clear_bit(PG_##lname, &policy(page, 1)->flags.f); }
467
468 #define TESTSETFLAG(uname, lname, policy) \
469 FOLIO_TEST_SET_FLAG(lname, FOLIO_##policy) \
470 static __always_inline int TestSetPage##uname(struct page *page) \
471 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags.f); }
472
473 #define TESTCLEARFLAG(uname, lname, policy) \
474 FOLIO_TEST_CLEAR_FLAG(lname, FOLIO_##policy) \
475 static __always_inline int TestClearPage##uname(struct page *page) \
476 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags.f); }
477
478 #define PAGEFLAG(uname, lname, policy) \
479 TESTPAGEFLAG(uname, lname, policy) \
480 SETPAGEFLAG(uname, lname, policy) \
481 CLEARPAGEFLAG(uname, lname, policy)
482
483 #define __PAGEFLAG(uname, lname, policy) \
484 TESTPAGEFLAG(uname, lname, policy) \
485 __SETPAGEFLAG(uname, lname, policy) \
486 __CLEARPAGEFLAG(uname, lname, policy)
487
488 #define TESTSCFLAG(uname, lname, policy) \
489 TESTSETFLAG(uname, lname, policy) \
490 TESTCLEARFLAG(uname, lname, policy)
491
492 #define FOLIO_TEST_FLAG_FALSE(name) \
493 static inline bool folio_test_##name(const struct folio *folio) \
494 { return false; }
495 #define FOLIO_SET_FLAG_NOOP(name) \
496 static inline void folio_set_##name(struct folio *folio) { }
497 #define FOLIO_CLEAR_FLAG_NOOP(name) \
498 static inline void folio_clear_##name(struct folio *folio) { }
499 #define __FOLIO_SET_FLAG_NOOP(name) \
500 static inline void __folio_set_##name(struct folio *folio) { }
501 #define __FOLIO_CLEAR_FLAG_NOOP(name) \
502 static inline void __folio_clear_##name(struct folio *folio) { }
503 #define FOLIO_TEST_SET_FLAG_FALSE(name) \
504 static inline bool folio_test_set_##name(struct folio *folio) \
505 { return false; }
506 #define FOLIO_TEST_CLEAR_FLAG_FALSE(name) \
507 static inline bool folio_test_clear_##name(struct folio *folio) \
508 { return false; }
509
510 #define FOLIO_FLAG_FALSE(name) \
511 FOLIO_TEST_FLAG_FALSE(name) \
512 FOLIO_SET_FLAG_NOOP(name) \
513 FOLIO_CLEAR_FLAG_NOOP(name)
514
515 #define TESTPAGEFLAG_FALSE(uname, lname) \
516 FOLIO_TEST_FLAG_FALSE(lname) \
517 static inline int Page##uname(const struct page *page) { return 0; }
518
519 #define SETPAGEFLAG_NOOP(uname, lname) \
520 FOLIO_SET_FLAG_NOOP(lname) \
521 static inline void SetPage##uname(struct page *page) { }
522
523 #define CLEARPAGEFLAG_NOOP(uname, lname) \
524 FOLIO_CLEAR_FLAG_NOOP(lname) \
525 static inline void ClearPage##uname(struct page *page) { }
526
527 #define __CLEARPAGEFLAG_NOOP(uname, lname) \
528 __FOLIO_CLEAR_FLAG_NOOP(lname) \
529 static inline void __ClearPage##uname(struct page *page) { }
530
531 #define TESTSETFLAG_FALSE(uname, lname) \
532 FOLIO_TEST_SET_FLAG_FALSE(lname) \
533 static inline int TestSetPage##uname(struct page *page) { return 0; }
534
535 #define TESTCLEARFLAG_FALSE(uname, lname) \
536 FOLIO_TEST_CLEAR_FLAG_FALSE(lname) \
537 static inline int TestClearPage##uname(struct page *page) { return 0; }
538
539 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \
540 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
541
542 #define TESTSCFLAG_FALSE(uname, lname) \
543 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
544
545 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
546 FOLIO_FLAG(waiters, FOLIO_HEAD_PAGE)
547 FOLIO_FLAG(referenced, FOLIO_HEAD_PAGE)
548 FOLIO_TEST_CLEAR_FLAG(referenced, FOLIO_HEAD_PAGE)
549 __FOLIO_SET_FLAG(referenced, FOLIO_HEAD_PAGE)
550 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
551 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
552 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
553 TESTCLEARFLAG(LRU, lru, PF_HEAD)
554 FOLIO_FLAG(active, FOLIO_HEAD_PAGE)
555 __FOLIO_CLEAR_FLAG(active, FOLIO_HEAD_PAGE)
556 FOLIO_TEST_CLEAR_FLAG(active, FOLIO_HEAD_PAGE)
557 PAGEFLAG(Workingset, workingset, PF_HEAD)
558 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
559 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
560
561 /* Xen */
562 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
563 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
564 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
565 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
PAGEFLAG(XenRemapped,xen_remapped,PF_NO_COMPOUND)566 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
567 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
568
569 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
570 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
571 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
572 FOLIO_FLAG(swapbacked, FOLIO_HEAD_PAGE)
573 __FOLIO_CLEAR_FLAG(swapbacked, FOLIO_HEAD_PAGE)
574 __FOLIO_SET_FLAG(swapbacked, FOLIO_HEAD_PAGE)
575
576 /*
577 * Private page markings that may be used by the filesystem that owns the page
578 * for its own purposes.
579 * - PG_private and PG_private_2 cause release_folio() and co to be invoked
580 */
581 PAGEFLAG(Private, private, PF_ANY)
582 FOLIO_FLAG(private_2, FOLIO_HEAD_PAGE)
583
584 /* owner_2 can be set on tail pages for anon memory */
585 FOLIO_FLAG(owner_2, FOLIO_HEAD_PAGE)
586
587 /*
588 * Only test-and-set exist for PG_writeback. The unconditional operators are
589 * risky: they bypass page accounting.
590 */
591 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
592 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
593 FOLIO_FLAG(mappedtodisk, FOLIO_HEAD_PAGE)
594
595 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
596 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
597 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
598 FOLIO_FLAG(readahead, FOLIO_HEAD_PAGE)
599 FOLIO_TEST_CLEAR_FLAG(readahead, FOLIO_HEAD_PAGE)
600
601 FOLIO_FLAG(dropbehind, FOLIO_HEAD_PAGE)
602 FOLIO_TEST_CLEAR_FLAG(dropbehind, FOLIO_HEAD_PAGE)
603 __FOLIO_SET_FLAG(dropbehind, FOLIO_HEAD_PAGE)
604
605 #ifdef CONFIG_HIGHMEM
606 /*
607 * Must use a macro here due to header dependency issues. page_zone() is not
608 * available at this point.
609 */
610 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
611 #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f))
612 #else
613 PAGEFLAG_FALSE(HighMem, highmem)
614 #endif
615 #define PhysHighMem(__p) (PageHighMem(phys_to_page(__p)))
616
617 /* Does kmap_local_folio() only allow access to one page of the folio? */
618 #ifdef CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP
619 #define folio_test_partial_kmap(f) true
620 #else
621 #define folio_test_partial_kmap(f) folio_test_highmem(f)
622 #endif
623
624 #ifdef CONFIG_SWAP
625 static __always_inline bool folio_test_swapcache(const struct folio *folio)
626 {
627 return folio_test_swapbacked(folio) &&
628 test_bit(PG_swapcache, const_folio_flags(folio, 0));
629 }
630
FOLIO_SET_FLAG(swapcache,FOLIO_HEAD_PAGE)631 FOLIO_SET_FLAG(swapcache, FOLIO_HEAD_PAGE)
632 FOLIO_CLEAR_FLAG(swapcache, FOLIO_HEAD_PAGE)
633 #else
634 FOLIO_FLAG_FALSE(swapcache)
635 #endif
636
637 FOLIO_FLAG(unevictable, FOLIO_HEAD_PAGE)
638 __FOLIO_CLEAR_FLAG(unevictable, FOLIO_HEAD_PAGE)
639 FOLIO_TEST_CLEAR_FLAG(unevictable, FOLIO_HEAD_PAGE)
640
641 #ifdef CONFIG_MMU
642 FOLIO_FLAG(mlocked, FOLIO_HEAD_PAGE)
643 __FOLIO_CLEAR_FLAG(mlocked, FOLIO_HEAD_PAGE)
644 FOLIO_TEST_CLEAR_FLAG(mlocked, FOLIO_HEAD_PAGE)
645 FOLIO_TEST_SET_FLAG(mlocked, FOLIO_HEAD_PAGE)
646 #else
647 FOLIO_FLAG_FALSE(mlocked)
648 __FOLIO_CLEAR_FLAG_NOOP(mlocked)
649 FOLIO_TEST_CLEAR_FLAG_FALSE(mlocked)
650 FOLIO_TEST_SET_FLAG_FALSE(mlocked)
651 #endif
652
653 #ifdef CONFIG_MEMORY_FAILURE
654 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
655 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
656 #define __PG_HWPOISON (1UL << PG_hwpoison)
657 #else
658 PAGEFLAG_FALSE(HWPoison, hwpoison)
659 #define __PG_HWPOISON 0
660 #endif
661
662 #ifdef CONFIG_PAGE_IDLE_FLAG
663 #ifdef CONFIG_64BIT
664 FOLIO_TEST_FLAG(young, FOLIO_HEAD_PAGE)
665 FOLIO_SET_FLAG(young, FOLIO_HEAD_PAGE)
666 FOLIO_TEST_CLEAR_FLAG(young, FOLIO_HEAD_PAGE)
667 FOLIO_FLAG(idle, FOLIO_HEAD_PAGE)
668 #endif
669 /* See page_idle.h for !64BIT workaround */
670 #else /* !CONFIG_PAGE_IDLE_FLAG */
671 FOLIO_FLAG_FALSE(young)
672 FOLIO_TEST_CLEAR_FLAG_FALSE(young)
673 FOLIO_FLAG_FALSE(idle)
674 #endif
675
676 /*
677 * PageReported() is used to track reported free pages within the Buddy
678 * allocator. We can use the non-atomic version of the test and set
679 * operations as both should be shielded with the zone lock to prevent
680 * any possible races on the setting or clearing of the bit.
681 */
682 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
683
684 #ifdef CONFIG_MEMORY_HOTPLUG
685 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
686 #else
687 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
688 #endif
689
690 /*
691 * On an anonymous folio mapped into a user virtual memory area,
692 * folio->mapping points to its anon_vma, not to a struct address_space;
693 * with the FOLIO_MAPPING_ANON bit set to distinguish it. See rmap.h.
694 *
695 * On an anonymous folio in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
696 * the FOLIO_MAPPING_ANON_KSM bit may be set along with the FOLIO_MAPPING_ANON
697 * bit; and then folio->mapping points, not to an anon_vma, but to a private
698 * structure which KSM associates with that merged folio. See ksm.h.
699 *
700 * Please note that, confusingly, "folio_mapping" refers to the inode
701 * address_space which maps the folio from disk; whereas "folio_mapped"
702 * refers to user virtual address space into which the folio is mapped.
703 *
704 * For slab pages, since slab reuses the bits in struct page to store its
705 * internal states, the folio->mapping does not exist as such, nor do
706 * these flags below. So in order to avoid testing non-existent bits,
707 * please make sure that folio_test_slab(folio) actually evaluates to
708 * false before calling the following functions (e.g., folio_test_anon).
709 * See mm/slab.h.
710 */
711 #define FOLIO_MAPPING_ANON 0x1
712 #define FOLIO_MAPPING_ANON_KSM 0x2
713 #define FOLIO_MAPPING_KSM (FOLIO_MAPPING_ANON | FOLIO_MAPPING_ANON_KSM)
714 #define FOLIO_MAPPING_FLAGS (FOLIO_MAPPING_ANON | FOLIO_MAPPING_ANON_KSM)
715
716 static __always_inline bool folio_test_anon(const struct folio *folio)
717 {
718 return ((unsigned long)folio->mapping & FOLIO_MAPPING_ANON) != 0;
719 }
720
folio_test_lazyfree(const struct folio * folio)721 static __always_inline bool folio_test_lazyfree(const struct folio *folio)
722 {
723 return folio_test_anon(folio) && !folio_test_swapbacked(folio);
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.f);
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
867 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
ClearPageCompound(struct page * page)868 static inline void ClearPageCompound(struct page *page)
869 {
870 BUG_ON(!PageHead(page));
871 ClearPageHead(page);
872 }
FOLIO_FLAG(large_rmappable,FOLIO_SECOND_PAGE)873 FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE)
874 FOLIO_FLAG(partially_mapped, FOLIO_SECOND_PAGE)
875 #else
876 FOLIO_FLAG_FALSE(large_rmappable)
877 FOLIO_FLAG_FALSE(partially_mapped)
878 #endif
879
880 #define PG_head_mask ((1UL << PG_head))
881
882 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
883 /*
884 * PageTransCompound returns true for both transparent huge pages
885 * and hugetlbfs pages, so it should only be called when it's known
886 * that hugetlbfs pages aren't involved.
887 */
888 static inline int PageTransCompound(const struct page *page)
889 {
890 return PageCompound(page);
891 }
892 #else
893 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
894 #endif
895
896 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
897 /*
898 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
899 * compound page.
900 *
901 * This flag is set by hwpoison handler. Cleared by THP split or free page.
902 */
903 FOLIO_FLAG(has_hwpoisoned, FOLIO_SECOND_PAGE)
904 #else
905 FOLIO_FLAG_FALSE(has_hwpoisoned)
906 #endif
907
908 /*
909 * For pages that do not use mapcount, page_type may be used.
910 * The low 24 bits of pagetype may be used for your own purposes, as long
911 * as you are careful to not affect the top 8 bits. The low bits of
912 * pagetype will be overwritten when you clear the page_type from the page.
913 */
914 enum pagetype {
915 /* 0x00-0x7f are positive numbers, ie mapcount */
916 /* Reserve 0x80-0xef for mapcount overflow. */
917 PGTY_buddy = 0xf0,
918 PGTY_offline = 0xf1,
919 PGTY_table = 0xf2,
920 PGTY_guard = 0xf3,
921 PGTY_hugetlb = 0xf4,
922 PGTY_slab = 0xf5,
923 PGTY_zsmalloc = 0xf6,
924 PGTY_unaccepted = 0xf7,
925 PGTY_large_kmalloc = 0xf8,
926 PGTY_netpp = 0xf9,
927
928 PGTY_mapcount_underflow = 0xff
929 };
930
page_type_has_type(int page_type)931 static inline bool page_type_has_type(int page_type)
932 {
933 return page_type < (PGTY_mapcount_underflow << 24);
934 }
935
936 /* This takes a mapcount which is one more than page->_mapcount */
page_mapcount_is_type(unsigned int mapcount)937 static inline bool page_mapcount_is_type(unsigned int mapcount)
938 {
939 return page_type_has_type(mapcount - 1);
940 }
941
page_has_type(const struct page * page)942 static inline bool page_has_type(const struct page *page)
943 {
944 return page_type_has_type(data_race(page->page_type));
945 }
946
947 #define FOLIO_TYPE_OPS(lname, fname) \
948 static __always_inline bool folio_test_##fname(const struct folio *folio) \
949 { \
950 return data_race(folio->page.page_type >> 24) == PGTY_##lname; \
951 } \
952 static __always_inline void __folio_set_##fname(struct folio *folio) \
953 { \
954 if (folio_test_##fname(folio)) \
955 return; \
956 VM_BUG_ON_FOLIO(data_race(folio->page.page_type) != UINT_MAX, \
957 folio); \
958 folio->page.page_type = (unsigned int)PGTY_##lname << 24; \
959 } \
960 static __always_inline void __folio_clear_##fname(struct folio *folio) \
961 { \
962 if (folio->page.page_type == UINT_MAX) \
963 return; \
964 VM_BUG_ON_FOLIO(!folio_test_##fname(folio), folio); \
965 folio->page.page_type = UINT_MAX; \
966 }
967
968 #define PAGE_TYPE_OPS(uname, lname, fname) \
969 FOLIO_TYPE_OPS(lname, fname) \
970 static __always_inline int Page##uname(const struct page *page) \
971 { \
972 return data_race(page->page_type >> 24) == PGTY_##lname; \
973 } \
974 static __always_inline void __SetPage##uname(struct page *page) \
975 { \
976 if (Page##uname(page)) \
977 return; \
978 VM_BUG_ON_PAGE(data_race(page->page_type) != UINT_MAX, page); \
979 page->page_type = (unsigned int)PGTY_##lname << 24; \
980 } \
981 static __always_inline void __ClearPage##uname(struct page *page) \
982 { \
983 if (page->page_type == UINT_MAX) \
984 return; \
985 VM_BUG_ON_PAGE(!Page##uname(page), page); \
986 page->page_type = UINT_MAX; \
987 }
988
989 /*
990 * PageBuddy() indicates that the page is free and in the buddy system
991 * (see mm/page_alloc.c).
992 */
993 PAGE_TYPE_OPS(Buddy, buddy, buddy)
994
995 /*
996 * PageOffline() indicates that the page is logically offline although the
997 * containing section is online. (e.g. inflated in a balloon driver or
998 * not onlined when onlining the section).
999 * The content of these pages is effectively stale. Such pages should not
1000 * be touched (read/write/dump/save) except by their owner.
1001 *
1002 * When a memory block gets onlined, all pages are initialized with a
1003 * refcount of 1 and PageOffline(). generic_online_page() will
1004 * take care of clearing PageOffline().
1005 *
1006 * If a driver wants to allow to offline unmovable PageOffline() pages without
1007 * putting them back to the buddy, it can do so via the memory notifier by
1008 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
1009 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
1010 * pages (now with a reference count of zero) are treated like free (unmanaged)
1011 * pages, allowing the containing memory block to get offlined. A driver that
1012 * relies on this feature is aware that re-onlining the memory block will
1013 * require not giving them to the buddy via generic_online_page().
1014 *
1015 * Memory offlining code will not adjust the managed page count for any
1016 * PageOffline() pages, treating them like they were never exposed to the
1017 * buddy using generic_online_page().
1018 *
1019 * There are drivers that mark a page PageOffline() and expect there won't be
1020 * any further access to page content. PFN walkers that read content of random
1021 * pages should check PageOffline() and synchronize with such drivers using
1022 * page_offline_freeze()/page_offline_thaw().
1023 */
1024 PAGE_TYPE_OPS(Offline, offline, offline)
1025
1026 extern void page_offline_freeze(void);
1027 extern void page_offline_thaw(void);
1028 extern void page_offline_begin(void);
1029 extern void page_offline_end(void);
1030
1031 /*
1032 * Marks pages in use as page tables.
1033 */
PAGE_TYPE_OPS(Table,table,pgtable)1034 PAGE_TYPE_OPS(Table, table, pgtable)
1035
1036 /*
1037 * Marks guardpages used with debug_pagealloc.
1038 */
1039 PAGE_TYPE_OPS(Guard, guard, guard)
1040
1041 PAGE_TYPE_OPS(Slab, slab, slab)
1042
1043 #ifdef CONFIG_HUGETLB_PAGE
1044 FOLIO_TYPE_OPS(hugetlb, hugetlb)
1045 #else
1046 FOLIO_TEST_FLAG_FALSE(hugetlb)
1047 #endif
1048
1049 PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
1050
1051 /*
1052 * Mark pages that has to be accepted before touched for the first time.
1053 *
1054 * Serialized with zone lock.
1055 */
1056 PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
1057 PAGE_TYPE_OPS(LargeKmalloc, large_kmalloc, large_kmalloc)
1058
1059 /*
1060 * Marks page_pool allocated pages.
1061 */
1062 PAGE_TYPE_OPS(Netpp, netpp, netpp)
1063
1064 /**
1065 * PageHuge - Determine if the page belongs to hugetlbfs
1066 * @page: The page to test.
1067 *
1068 * Context: Any context.
1069 * Return: True for hugetlbfs pages, false for anon pages or pages
1070 * belonging to other filesystems.
1071 */
1072 static inline bool PageHuge(const struct page *page)
1073 {
1074 return folio_test_hugetlb(page_folio(page));
1075 }
1076
1077 /*
1078 * Check if a page is currently marked HWPoisoned. Note that this check is
1079 * best effort only and inherently racy: there is no way to synchronize with
1080 * failing hardware.
1081 */
is_page_hwpoison(const struct page * page)1082 static inline bool is_page_hwpoison(const struct page *page)
1083 {
1084 const struct folio *folio;
1085
1086 if (PageHWPoison(page))
1087 return true;
1088 folio = page_folio(page);
1089 return folio_test_hugetlb(folio) && PageHWPoison(&folio->page);
1090 }
1091
folio_contain_hwpoisoned_page(struct folio * folio)1092 static inline bool folio_contain_hwpoisoned_page(struct folio *folio)
1093 {
1094 return folio_test_hwpoison(folio) ||
1095 (folio_test_large(folio) && folio_test_has_hwpoisoned(folio));
1096 }
1097
1098 bool is_free_buddy_page(const struct page *page);
1099
1100 #ifdef CONFIG_MIGRATION
1101 /*
1102 * This page is migratable through movable_ops (for selected typed pages
1103 * only).
1104 *
1105 * Page migration of such pages might fail, for example, if the page is
1106 * already isolated by somebody else, or if the page is about to get freed.
1107 *
1108 * While a subsystem might set selected typed pages that support page migration
1109 * as being movable through movable_ops, it must never clear this flag.
1110 *
1111 * This flag is only cleared when the page is freed back to the buddy.
1112 *
1113 * Only selected page types support this flag (see page_movable_ops()) and
1114 * the flag might be used in other context for other pages. Always use
1115 * page_has_movable_ops() instead.
1116 */
1117 TESTPAGEFLAG(MovableOps, movable_ops, PF_NO_TAIL);
1118 SETPAGEFLAG(MovableOps, movable_ops, PF_NO_TAIL);
1119 /*
1120 * A movable_ops page has this flag set while it is isolated for migration.
1121 * This flag primarily protects against concurrent migration attempts.
1122 *
1123 * Once migration ended (success or failure), the flag is cleared. The
1124 * flag is managed by the migration core.
1125 */
1126 PAGEFLAG(MovableOpsIsolated, movable_ops_isolated, PF_NO_TAIL);
1127 #else /* !CONFIG_MIGRATION */
1128 TESTPAGEFLAG_FALSE(MovableOps, movable_ops);
1129 SETPAGEFLAG_NOOP(MovableOps, movable_ops);
1130 PAGEFLAG_FALSE(MovableOpsIsolated, movable_ops_isolated);
1131 #endif /* CONFIG_MIGRATION */
1132
1133 /**
1134 * page_has_movable_ops - test for a movable_ops page
1135 * @page: The page to test.
1136 *
1137 * Test whether this is a movable_ops page. Such pages will stay that
1138 * way until freed.
1139 *
1140 * Returns true if this is a movable_ops page, otherwise false.
1141 */
page_has_movable_ops(const struct page * page)1142 static inline bool page_has_movable_ops(const struct page *page)
1143 {
1144 return PageMovableOps(page) &&
1145 (PageOffline(page) || PageZsmalloc(page));
1146 }
1147
PageAnonExclusive(const struct page * page)1148 static __always_inline int PageAnonExclusive(const struct page *page)
1149 {
1150 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1151 /*
1152 * HugeTLB stores this information on the head page; THP keeps it per
1153 * page
1154 */
1155 if (PageHuge(page))
1156 page = compound_head(page);
1157 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags.f);
1158 }
1159
SetPageAnonExclusive(struct page * page)1160 static __always_inline void SetPageAnonExclusive(struct page *page)
1161 {
1162 VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page);
1163 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1164 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags.f);
1165 }
1166
ClearPageAnonExclusive(struct page * page)1167 static __always_inline void ClearPageAnonExclusive(struct page *page)
1168 {
1169 VM_BUG_ON_PGFLAGS(!PageAnonNotKsm(page), page);
1170 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1171 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags.f);
1172 }
1173
__ClearPageAnonExclusive(struct page * page)1174 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1175 {
1176 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1177 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1178 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags.f);
1179 }
1180
1181 #ifdef CONFIG_MMU
1182 #define __PG_MLOCKED (1UL << PG_mlocked)
1183 #else
1184 #define __PG_MLOCKED 0
1185 #endif
1186
1187 /*
1188 * Flags checked when a page is freed. Pages being freed should not have
1189 * these flags set. If they are, there is a problem.
1190 */
1191 #define PAGE_FLAGS_CHECK_AT_FREE \
1192 (1UL << PG_lru | 1UL << PG_locked | \
1193 1UL << PG_private | 1UL << PG_private_2 | \
1194 1UL << PG_writeback | 1UL << PG_reserved | \
1195 1UL << PG_active | \
1196 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK)
1197
1198 /*
1199 * Flags checked when a page is prepped for return by the page allocator.
1200 * Pages being prepped should not have these flags set. If they are set,
1201 * there has been a kernel bug or struct page corruption.
1202 *
1203 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1204 * alloc-free cycle to prevent from reusing the page.
1205 */
1206 #define PAGE_FLAGS_CHECK_AT_PREP \
1207 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK)
1208
1209 /*
1210 * Flags stored in the second page of a compound page. They may overlap
1211 * the CHECK_AT_FREE flags above, so need to be cleared.
1212 */
1213 #define PAGE_FLAGS_SECOND \
1214 (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \
1215 1UL << PG_large_rmappable | 1UL << PG_partially_mapped)
1216
1217 #define PAGE_FLAGS_PRIVATE \
1218 (1UL << PG_private | 1UL << PG_private_2)
1219 /**
1220 * folio_has_private - Determine if folio has private stuff
1221 * @folio: The folio to be checked
1222 *
1223 * Determine if a folio has private stuff, indicating that release routines
1224 * should be invoked upon it.
1225 */
folio_has_private(const struct folio * folio)1226 static inline int folio_has_private(const struct folio *folio)
1227 {
1228 return !!(folio->flags.f & PAGE_FLAGS_PRIVATE);
1229 }
1230
1231 #undef PF_ANY
1232 #undef PF_HEAD
1233 #undef PF_NO_TAIL
1234 #undef PF_NO_COMPOUND
1235 #undef PF_SECOND
1236 #endif /* !__GENERATING_BOUNDS_H */
1237
1238 #endif /* PAGE_FLAGS_H */
1239