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 not added to the page allocator when onlining a section because
34 * they were excluded via the online_page_callback() or because they are
35 * PG_hwpoison.
36 * - Pages allocated in the context of kexec/kdump (loaded kernel image,
37 * control pages, vmcoreinfo)
38 * - MMIO/DMA pages. Some architectures don't allow to ioremap pages that are
39 * not marked PG_reserved (as they might be in use by somebody else who does
40 * not respect the caching strategy).
41 * - Pages part of an offline section (struct pages of offline sections should
42 * not be trusted as they will be initialized when first onlined).
43 * - MCA pages on ia64
44 * - Pages holding CPU notes for POWER Firmware Assisted Dump
45 * - Device memory (e.g. PMEM, DAX, HMM)
46 * Some PG_reserved pages will be excluded from the hibernation image.
47 * PG_reserved does in general not hinder anybody from dumping or swapping
48 * and is no longer required for remap_pfn_range(). ioremap might require it.
49 * Consequently, PG_reserved for a page mapped into user space can indicate
50 * the zero page, the vDSO, MMIO pages or device memory.
51 *
52 * The PG_private bitflag is set on pagecache pages if they contain filesystem
53 * specific data (which is normally at page->private). It can be used by
54 * private allocations for its own usage.
55 *
56 * During initiation of disk I/O, PG_locked is set. This bit is set before I/O
57 * and cleared when writeback _starts_ or when read _completes_. PG_writeback
58 * is set before writeback starts and cleared when it finishes.
59 *
60 * PG_locked also pins a page in pagecache, and blocks truncation of the file
61 * while it is held.
62 *
63 * page_waitqueue(page) is a wait queue of all tasks waiting for the page
64 * to become unlocked.
65 *
66 * PG_swapbacked is set when a page uses swap as a backing storage. This are
67 * usually PageAnon or shmem pages but please note that even anonymous pages
68 * might lose their PG_swapbacked flag when they simply can be dropped (e.g. as
69 * a result of MADV_FREE).
70 *
71 * PG_referenced, PG_reclaim are used for page reclaim for anonymous and
72 * file-backed pagecache (see mm/vmscan.c).
73 *
74 * PG_error is set to indicate that an I/O error occurred on this page.
75 *
76 * PG_arch_1 is an architecture specific page state bit. The generic code
77 * guarantees that this bit is cleared for a page when it first is entered into
78 * the page cache.
79 *
80 * PG_hwpoison indicates that a page got corrupted in hardware and contains
81 * data with incorrect ECC bits that triggered a machine check. Accessing is
82 * not safe since it may cause another machine check. Don't touch!
83 */
84
85 /*
86 * Don't use the pageflags directly. Use the PageFoo macros.
87 *
88 * The page flags field is split into two parts, the main flags area
89 * which extends from the low bits upwards, and the fields area which
90 * extends from the high bits downwards.
91 *
92 * | FIELD | ... | FLAGS |
93 * N-1 ^ 0
94 * (NR_PAGEFLAGS)
95 *
96 * The fields area is reserved for fields mapping zone, node (for NUMA) and
97 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like
98 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP).
99 */
100 enum pageflags {
101 PG_locked, /* Page is locked. Don't touch. */
102 PG_referenced,
103 PG_uptodate,
104 PG_dirty,
105 PG_lru,
106 PG_active,
107 PG_workingset,
108 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
109 PG_error,
110 PG_slab,
111 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
112 PG_arch_1,
113 PG_reserved,
114 PG_private, /* If pagecache, has fs-private data */
115 PG_private_2, /* If pagecache, has fs aux data */
116 PG_writeback, /* Page is under writeback */
117 PG_head, /* A head page */
118 PG_mappedtodisk, /* Has blocks allocated on-disk */
119 PG_reclaim, /* To be reclaimed asap */
120 PG_swapbacked, /* Page is backed by RAM/swap */
121 PG_unevictable, /* Page is "unevictable" */
122 #ifdef CONFIG_MMU
123 PG_mlocked, /* Page is vma mlocked */
124 #endif
125 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
126 PG_uncached, /* Page has been mapped as uncached */
127 #endif
128 #ifdef CONFIG_MEMORY_FAILURE
129 PG_hwpoison, /* hardware poisoned page. Don't touch */
130 #endif
131 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
132 PG_young,
133 PG_idle,
134 #endif
135 #ifdef CONFIG_64BIT
136 PG_arch_2,
137 #endif
138 #ifdef CONFIG_KASAN_HW_TAGS
139 PG_skip_kasan_poison,
140 #endif
141 __NR_PAGEFLAGS,
142
143 PG_readahead = PG_reclaim,
144
145 /*
146 * Depending on the way an anonymous folio can be mapped into a page
147 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped
148 * THP), PG_anon_exclusive may be set only for the head page or for
149 * tail pages of an anonymous folio. For now, we only expect it to be
150 * set on tail pages for PTE-mapped THP.
151 */
152 PG_anon_exclusive = PG_mappedtodisk,
153
154 /* Filesystems */
155 PG_checked = PG_owner_priv_1,
156
157 /* SwapBacked */
158 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
159
160 /* Two page bits are conscripted by FS-Cache to maintain local caching
161 * state. These bits are set on pages belonging to the netfs's inodes
162 * when those inodes are being locally cached.
163 */
164 PG_fscache = PG_private_2, /* page backed by cache */
165
166 /* XEN */
167 /* Pinned in Xen as a read-only pagetable page. */
168 PG_pinned = PG_owner_priv_1,
169 /* Pinned as part of domain save (see xen_mm_pin_all()). */
170 PG_savepinned = PG_dirty,
171 /* Has a grant mapping of another (foreign) domain's page. */
172 PG_foreign = PG_owner_priv_1,
173 /* Remapped by swiotlb-xen. */
174 PG_xen_remapped = PG_owner_priv_1,
175
176 /* SLOB */
177 PG_slob_free = PG_private,
178
179 /* Compound pages. Stored in first tail page's flags */
180 PG_double_map = PG_workingset,
181
182 #ifdef CONFIG_MEMORY_FAILURE
183 /*
184 * Compound pages. Stored in first tail page's flags.
185 * Indicates that at least one subpage is hwpoisoned in the
186 * THP.
187 */
188 PG_has_hwpoisoned = PG_error,
189 #endif
190
191 /* non-lru isolated movable page */
192 PG_isolated = PG_reclaim,
193
194 /* Only valid for buddy pages. Used to track pages that are reported */
195 PG_reported = PG_uptodate,
196
197 #ifdef CONFIG_MEMORY_HOTPLUG
198 /* For self-hosted memmap pages */
199 PG_vmemmap_self_hosted = PG_owner_priv_1,
200 #endif
201 };
202
203 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
204
205 #ifndef __GENERATING_BOUNDS_H
206
207 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
208 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key);
209
210 /*
211 * Return the real head page struct iff the @page is a fake head page, otherwise
212 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst.
213 */
page_fixed_fake_head(const struct page * page)214 static __always_inline const struct page *page_fixed_fake_head(const struct page *page)
215 {
216 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key))
217 return page;
218
219 /*
220 * Only addresses aligned with PAGE_SIZE of struct page may be fake head
221 * struct page. The alignment check aims to avoid access the fields (
222 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly)
223 * cold cacheline in some cases.
224 */
225 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) &&
226 test_bit(PG_head, &page->flags)) {
227 /*
228 * We can safely access the field of the @page[1] with PG_head
229 * because the @page is a compound page composed with at least
230 * two contiguous pages.
231 */
232 unsigned long head = READ_ONCE(page[1].compound_head);
233
234 if (likely(head & 1))
235 return (const struct page *)(head - 1);
236 }
237 return page;
238 }
239 #else
page_fixed_fake_head(const struct page * page)240 static inline const struct page *page_fixed_fake_head(const struct page *page)
241 {
242 return page;
243 }
244 #endif
245
page_is_fake_head(struct page * page)246 static __always_inline int page_is_fake_head(struct page *page)
247 {
248 return page_fixed_fake_head(page) != page;
249 }
250
_compound_head(const struct page * page)251 static inline unsigned long _compound_head(const struct page *page)
252 {
253 unsigned long head = READ_ONCE(page->compound_head);
254
255 if (unlikely(head & 1))
256 return head - 1;
257 return (unsigned long)page_fixed_fake_head(page);
258 }
259
260 #define compound_head(page) ((typeof(page))_compound_head(page))
261
262 /**
263 * page_folio - Converts from page to folio.
264 * @p: The page.
265 *
266 * Every page is part of a folio. This function cannot be called on a
267 * NULL pointer.
268 *
269 * Context: No reference, nor lock is required on @page. If the caller
270 * does not hold a reference, this call may race with a folio split, so
271 * it should re-check the folio still contains this page after gaining
272 * a reference on the folio.
273 * Return: The folio which contains this page.
274 */
275 #define page_folio(p) (_Generic((p), \
276 const struct page *: (const struct folio *)_compound_head(p), \
277 struct page *: (struct folio *)_compound_head(p)))
278
279 /**
280 * folio_page - Return a page from a folio.
281 * @folio: The folio.
282 * @n: The page number to return.
283 *
284 * @n is relative to the start of the folio. This function does not
285 * check that the page number lies within @folio; the caller is presumed
286 * to have a reference to the page.
287 */
288 #define folio_page(folio, n) nth_page(&(folio)->page, n)
289
PageTail(struct page * page)290 static __always_inline int PageTail(struct page *page)
291 {
292 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page);
293 }
294
PageCompound(struct page * page)295 static __always_inline int PageCompound(struct page *page)
296 {
297 return test_bit(PG_head, &page->flags) ||
298 READ_ONCE(page->compound_head) & 1;
299 }
300
301 #define PAGE_POISON_PATTERN -1l
PagePoisoned(const struct page * page)302 static inline int PagePoisoned(const struct page *page)
303 {
304 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN;
305 }
306
307 #ifdef CONFIG_DEBUG_VM
308 void page_init_poison(struct page *page, size_t size);
309 #else
page_init_poison(struct page * page,size_t size)310 static inline void page_init_poison(struct page *page, size_t size)
311 {
312 }
313 #endif
314
folio_flags(struct folio * folio,unsigned n)315 static unsigned long *folio_flags(struct folio *folio, unsigned n)
316 {
317 struct page *page = &folio->page;
318
319 VM_BUG_ON_PGFLAGS(PageTail(page), page);
320 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page);
321 return &page[n].flags;
322 }
323
324 /*
325 * Page flags policies wrt compound pages
326 *
327 * PF_POISONED_CHECK
328 * check if this struct page poisoned/uninitialized
329 *
330 * PF_ANY:
331 * the page flag is relevant for small, head and tail pages.
332 *
333 * PF_HEAD:
334 * for compound page all operations related to the page flag applied to
335 * head page.
336 *
337 * PF_ONLY_HEAD:
338 * for compound page, callers only ever operate on the head page.
339 *
340 * PF_NO_TAIL:
341 * modifications of the page flag must be done on small or head pages,
342 * checks can be done on tail pages too.
343 *
344 * PF_NO_COMPOUND:
345 * the page flag is not relevant for compound pages.
346 *
347 * PF_SECOND:
348 * the page flag is stored in the first tail page.
349 */
350 #define PF_POISONED_CHECK(page) ({ \
351 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
352 page; })
353 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
354 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
355 #define PF_ONLY_HEAD(page, enforce) ({ \
356 VM_BUG_ON_PGFLAGS(PageTail(page), page); \
357 PF_POISONED_CHECK(page); })
358 #define PF_NO_TAIL(page, enforce) ({ \
359 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
360 PF_POISONED_CHECK(compound_head(page)); })
361 #define PF_NO_COMPOUND(page, enforce) ({ \
362 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
363 PF_POISONED_CHECK(page); })
364 #define PF_SECOND(page, enforce) ({ \
365 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
366 PF_POISONED_CHECK(&page[1]); })
367
368 /* Which page is the flag stored in */
369 #define FOLIO_PF_ANY 0
370 #define FOLIO_PF_HEAD 0
371 #define FOLIO_PF_ONLY_HEAD 0
372 #define FOLIO_PF_NO_TAIL 0
373 #define FOLIO_PF_NO_COMPOUND 0
374 #define FOLIO_PF_SECOND 1
375
376 /*
377 * Macros to create function definitions for page flags
378 */
379 #define TESTPAGEFLAG(uname, lname, policy) \
380 static __always_inline bool folio_test_##lname(struct folio *folio) \
381 { return test_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
382 static __always_inline int Page##uname(struct page *page) \
383 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
384
385 #define SETPAGEFLAG(uname, lname, policy) \
386 static __always_inline \
387 void folio_set_##lname(struct folio *folio) \
388 { set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
389 static __always_inline void SetPage##uname(struct page *page) \
390 { set_bit(PG_##lname, &policy(page, 1)->flags); }
391
392 #define CLEARPAGEFLAG(uname, lname, policy) \
393 static __always_inline \
394 void folio_clear_##lname(struct folio *folio) \
395 { clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
396 static __always_inline void ClearPage##uname(struct page *page) \
397 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
398
399 #define __SETPAGEFLAG(uname, lname, policy) \
400 static __always_inline \
401 void __folio_set_##lname(struct folio *folio) \
402 { __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
403 static __always_inline void __SetPage##uname(struct page *page) \
404 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
405
406 #define __CLEARPAGEFLAG(uname, lname, policy) \
407 static __always_inline \
408 void __folio_clear_##lname(struct folio *folio) \
409 { __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
410 static __always_inline void __ClearPage##uname(struct page *page) \
411 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
412
413 #define TESTSETFLAG(uname, lname, policy) \
414 static __always_inline \
415 bool folio_test_set_##lname(struct folio *folio) \
416 { return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
417 static __always_inline int TestSetPage##uname(struct page *page) \
418 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
419
420 #define TESTCLEARFLAG(uname, lname, policy) \
421 static __always_inline \
422 bool folio_test_clear_##lname(struct folio *folio) \
423 { return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
424 static __always_inline int TestClearPage##uname(struct page *page) \
425 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
426
427 #define PAGEFLAG(uname, lname, policy) \
428 TESTPAGEFLAG(uname, lname, policy) \
429 SETPAGEFLAG(uname, lname, policy) \
430 CLEARPAGEFLAG(uname, lname, policy)
431
432 #define __PAGEFLAG(uname, lname, policy) \
433 TESTPAGEFLAG(uname, lname, policy) \
434 __SETPAGEFLAG(uname, lname, policy) \
435 __CLEARPAGEFLAG(uname, lname, policy)
436
437 #define TESTSCFLAG(uname, lname, policy) \
438 TESTSETFLAG(uname, lname, policy) \
439 TESTCLEARFLAG(uname, lname, policy)
440
441 #define TESTPAGEFLAG_FALSE(uname, lname) \
442 static inline bool folio_test_##lname(const struct folio *folio) { return false; } \
443 static inline int Page##uname(const struct page *page) { return 0; }
444
445 #define SETPAGEFLAG_NOOP(uname, lname) \
446 static inline void folio_set_##lname(struct folio *folio) { } \
447 static inline void SetPage##uname(struct page *page) { }
448
449 #define CLEARPAGEFLAG_NOOP(uname, lname) \
450 static inline void folio_clear_##lname(struct folio *folio) { } \
451 static inline void ClearPage##uname(struct page *page) { }
452
453 #define __CLEARPAGEFLAG_NOOP(uname, lname) \
454 static inline void __folio_clear_##lname(struct folio *folio) { } \
455 static inline void __ClearPage##uname(struct page *page) { }
456
457 #define TESTSETFLAG_FALSE(uname, lname) \
458 static inline bool folio_test_set_##lname(struct folio *folio) \
459 { return 0; } \
460 static inline int TestSetPage##uname(struct page *page) { return 0; }
461
462 #define TESTCLEARFLAG_FALSE(uname, lname) \
463 static inline bool folio_test_clear_##lname(struct folio *folio) \
464 { return 0; } \
465 static inline int TestClearPage##uname(struct page *page) { return 0; }
466
467 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \
468 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname)
469
470 #define TESTSCFLAG_FALSE(uname, lname) \
471 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname)
472
473 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
474 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD)
475 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
476 PAGEFLAG(Referenced, referenced, PF_HEAD)
477 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
478 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
479 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
480 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
481 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
482 TESTCLEARFLAG(LRU, lru, PF_HEAD)
483 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
484 TESTCLEARFLAG(Active, active, PF_HEAD)
485 PAGEFLAG(Workingset, workingset, PF_HEAD)
486 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
487 __PAGEFLAG(Slab, slab, PF_NO_TAIL)
488 __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL)
489 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
490
491 /* Xen */
492 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
493 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
494 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
495 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
PAGEFLAG(XenRemapped,xen_remapped,PF_NO_COMPOUND)496 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
497 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
498
499 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
500 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
501 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
502 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
503 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
504 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
505
506 /*
507 * Private page markings that may be used by the filesystem that owns the page
508 * for its own purposes.
509 * - PG_private and PG_private_2 cause release_folio() and co to be invoked
510 */
511 PAGEFLAG(Private, private, PF_ANY)
512 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
513 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
514 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
515
516 /*
517 * Only test-and-set exist for PG_writeback. The unconditional operators are
518 * risky: they bypass page accounting.
519 */
520 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
521 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
522 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
523
524 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
525 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
526 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
527 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND)
528 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND)
529
530 #ifdef CONFIG_HIGHMEM
531 /*
532 * Must use a macro here due to header dependency issues. page_zone() is not
533 * available at this point.
534 */
535 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
536 #else
537 PAGEFLAG_FALSE(HighMem, highmem)
538 #endif
539
540 #ifdef CONFIG_SWAP
541 static __always_inline bool folio_test_swapcache(struct folio *folio)
542 {
543 return folio_test_swapbacked(folio) &&
544 test_bit(PG_swapcache, folio_flags(folio, 0));
545 }
546
PageSwapCache(struct page * page)547 static __always_inline bool PageSwapCache(struct page *page)
548 {
549 return folio_test_swapcache(page_folio(page));
550 }
551
552 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
553 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
554 #else
555 PAGEFLAG_FALSE(SwapCache, swapcache)
556 #endif
557
558 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
559 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
560 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
561
562 #ifdef CONFIG_MMU
563 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
564 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
565 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
566 #else
567 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked)
568 TESTSCFLAG_FALSE(Mlocked, mlocked)
569 #endif
570
571 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
572 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
573 #else
574 PAGEFLAG_FALSE(Uncached, uncached)
575 #endif
576
577 #ifdef CONFIG_MEMORY_FAILURE
578 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
579 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
580 #define __PG_HWPOISON (1UL << PG_hwpoison)
581 #define MAGIC_HWPOISON 0x48575053U /* HWPS */
582 extern void SetPageHWPoisonTakenOff(struct page *page);
583 extern void ClearPageHWPoisonTakenOff(struct page *page);
584 extern bool take_page_off_buddy(struct page *page);
585 extern bool put_page_back_buddy(struct page *page);
586 #else
587 PAGEFLAG_FALSE(HWPoison, hwpoison)
588 #define __PG_HWPOISON 0
589 #endif
590
591 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
TESTPAGEFLAG(Young,young,PF_ANY)592 TESTPAGEFLAG(Young, young, PF_ANY)
593 SETPAGEFLAG(Young, young, PF_ANY)
594 TESTCLEARFLAG(Young, young, PF_ANY)
595 PAGEFLAG(Idle, idle, PF_ANY)
596 #endif
597
598 #ifdef CONFIG_KASAN_HW_TAGS
599 PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD)
600 #else
601 PAGEFLAG_FALSE(SkipKASanPoison, skip_kasan_poison)
602 #endif
603
604 /*
605 * PageReported() is used to track reported free pages within the Buddy
606 * allocator. We can use the non-atomic version of the test and set
607 * operations as both should be shielded with the zone lock to prevent
608 * any possible races on the setting or clearing of the bit.
609 */
610 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
611
612 #ifdef CONFIG_MEMORY_HOTPLUG
613 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY)
614 #else
615 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted)
616 #endif
617
618 /*
619 * On an anonymous page mapped into a user virtual memory area,
620 * page->mapping points to its anon_vma, not to a struct address_space;
621 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
622 *
623 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
624 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
625 * bit; and then page->mapping points, not to an anon_vma, but to a private
626 * structure which KSM associates with that merged page. See ksm.h.
627 *
628 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
629 * page and then page->mapping points to a struct movable_operations.
630 *
631 * Please note that, confusingly, "page_mapping" refers to the inode
632 * address_space which maps the page from disk; whereas "page_mapped"
633 * refers to user virtual address space into which the page is mapped.
634 */
635 #define PAGE_MAPPING_ANON 0x1
636 #define PAGE_MAPPING_MOVABLE 0x2
637 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
638 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
639
640 /*
641 * Different with flags above, this flag is used only for fsdax mode. It
642 * indicates that this page->mapping is now under reflink case.
643 */
644 #define PAGE_MAPPING_DAX_COW 0x1
645
646 static __always_inline bool folio_mapping_flags(struct folio *folio)
647 {
648 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0;
649 }
650
PageMappingFlags(struct page * page)651 static __always_inline int PageMappingFlags(struct page *page)
652 {
653 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
654 }
655
folio_test_anon(struct folio * folio)656 static __always_inline bool folio_test_anon(struct folio *folio)
657 {
658 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0;
659 }
660
PageAnon(struct page * page)661 static __always_inline bool PageAnon(struct page *page)
662 {
663 return folio_test_anon(page_folio(page));
664 }
665
__folio_test_movable(const struct folio * folio)666 static __always_inline bool __folio_test_movable(const struct folio *folio)
667 {
668 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
669 PAGE_MAPPING_MOVABLE;
670 }
671
__PageMovable(struct page * page)672 static __always_inline int __PageMovable(struct page *page)
673 {
674 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
675 PAGE_MAPPING_MOVABLE;
676 }
677
678 #ifdef CONFIG_KSM
679 /*
680 * A KSM page is one of those write-protected "shared pages" or "merged pages"
681 * which KSM maps into multiple mms, wherever identical anonymous page content
682 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
683 * anon_vma, but to that page's node of the stable tree.
684 */
folio_test_ksm(struct folio * folio)685 static __always_inline bool folio_test_ksm(struct folio *folio)
686 {
687 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) ==
688 PAGE_MAPPING_KSM;
689 }
690
PageKsm(struct page * page)691 static __always_inline bool PageKsm(struct page *page)
692 {
693 return folio_test_ksm(page_folio(page));
694 }
695 #else
696 TESTPAGEFLAG_FALSE(Ksm, ksm)
697 #endif
698
699 u64 stable_page_flags(struct page *page);
700
701 /**
702 * folio_test_uptodate - Is this folio up to date?
703 * @folio: The folio.
704 *
705 * The uptodate flag is set on a folio when every byte in the folio is
706 * at least as new as the corresponding bytes on storage. Anonymous
707 * and CoW folios are always uptodate. If the folio is not uptodate,
708 * some of the bytes in it may be; see the is_partially_uptodate()
709 * address_space operation.
710 */
folio_test_uptodate(struct folio * folio)711 static inline bool folio_test_uptodate(struct folio *folio)
712 {
713 bool ret = test_bit(PG_uptodate, folio_flags(folio, 0));
714 /*
715 * Must ensure that the data we read out of the folio is loaded
716 * _after_ we've loaded folio->flags to check the uptodate bit.
717 * We can skip the barrier if the folio is not uptodate, because
718 * we wouldn't be reading anything from it.
719 *
720 * See folio_mark_uptodate() for the other side of the story.
721 */
722 if (ret)
723 smp_rmb();
724
725 return ret;
726 }
727
PageUptodate(struct page * page)728 static inline int PageUptodate(struct page *page)
729 {
730 return folio_test_uptodate(page_folio(page));
731 }
732
__folio_mark_uptodate(struct folio * folio)733 static __always_inline void __folio_mark_uptodate(struct folio *folio)
734 {
735 smp_wmb();
736 __set_bit(PG_uptodate, folio_flags(folio, 0));
737 }
738
folio_mark_uptodate(struct folio * folio)739 static __always_inline void folio_mark_uptodate(struct folio *folio)
740 {
741 /*
742 * Memory barrier must be issued before setting the PG_uptodate bit,
743 * so that all previous stores issued in order to bring the folio
744 * uptodate are actually visible before folio_test_uptodate becomes true.
745 */
746 smp_wmb();
747 set_bit(PG_uptodate, folio_flags(folio, 0));
748 }
749
__SetPageUptodate(struct page * page)750 static __always_inline void __SetPageUptodate(struct page *page)
751 {
752 __folio_mark_uptodate((struct folio *)page);
753 }
754
SetPageUptodate(struct page * page)755 static __always_inline void SetPageUptodate(struct page *page)
756 {
757 folio_mark_uptodate((struct folio *)page);
758 }
759
760 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
761
762 bool __folio_start_writeback(struct folio *folio, bool keep_write);
763 bool set_page_writeback(struct page *page);
764
765 #define folio_start_writeback(folio) \
766 __folio_start_writeback(folio, false)
767 #define folio_start_writeback_keepwrite(folio) \
768 __folio_start_writeback(folio, true)
769
set_page_writeback_keepwrite(struct page * page)770 static inline void set_page_writeback_keepwrite(struct page *page)
771 {
772 folio_start_writeback_keepwrite(page_folio(page));
773 }
774
test_set_page_writeback(struct page * page)775 static inline bool test_set_page_writeback(struct page *page)
776 {
777 return set_page_writeback(page);
778 }
779
folio_test_head(struct folio * folio)780 static __always_inline bool folio_test_head(struct folio *folio)
781 {
782 return test_bit(PG_head, folio_flags(folio, FOLIO_PF_ANY));
783 }
784
PageHead(struct page * page)785 static __always_inline int PageHead(struct page *page)
786 {
787 PF_POISONED_CHECK(page);
788 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page);
789 }
790
__SETPAGEFLAG(Head,head,PF_ANY)791 __SETPAGEFLAG(Head, head, PF_ANY)
792 __CLEARPAGEFLAG(Head, head, PF_ANY)
793 CLEARPAGEFLAG(Head, head, PF_ANY)
794
795 /**
796 * folio_test_large() - Does this folio contain more than one page?
797 * @folio: The folio to test.
798 *
799 * Return: True if the folio is larger than one page.
800 */
801 static inline bool folio_test_large(struct folio *folio)
802 {
803 return folio_test_head(folio);
804 }
805
set_compound_head(struct page * page,struct page * head)806 static __always_inline void set_compound_head(struct page *page, struct page *head)
807 {
808 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
809 }
810
clear_compound_head(struct page * page)811 static __always_inline void clear_compound_head(struct page *page)
812 {
813 WRITE_ONCE(page->compound_head, 0);
814 }
815
816 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
ClearPageCompound(struct page * page)817 static inline void ClearPageCompound(struct page *page)
818 {
819 BUG_ON(!PageHead(page));
820 ClearPageHead(page);
821 }
822 #endif
823
824 #define PG_head_mask ((1UL << PG_head))
825
826 #ifdef CONFIG_HUGETLB_PAGE
827 int PageHuge(struct page *page);
828 int PageHeadHuge(struct page *page);
folio_test_hugetlb(struct folio * folio)829 static inline bool folio_test_hugetlb(struct folio *folio)
830 {
831 return PageHeadHuge(&folio->page);
832 }
833 #else
TESTPAGEFLAG_FALSE(Huge,hugetlb)834 TESTPAGEFLAG_FALSE(Huge, hugetlb)
835 TESTPAGEFLAG_FALSE(HeadHuge, headhuge)
836 #endif
837
838 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
839 /*
840 * PageHuge() only returns true for hugetlbfs pages, but not for
841 * normal or transparent huge pages.
842 *
843 * PageTransHuge() returns true for both transparent huge and
844 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
845 * called only in the core VM paths where hugetlbfs pages can't exist.
846 */
847 static inline int PageTransHuge(struct page *page)
848 {
849 VM_BUG_ON_PAGE(PageTail(page), page);
850 return PageHead(page);
851 }
852
folio_test_transhuge(struct folio * folio)853 static inline bool folio_test_transhuge(struct folio *folio)
854 {
855 return folio_test_head(folio);
856 }
857
858 /*
859 * PageTransCompound returns true for both transparent huge pages
860 * and hugetlbfs pages, so it should only be called when it's known
861 * that hugetlbfs pages aren't involved.
862 */
PageTransCompound(struct page * page)863 static inline int PageTransCompound(struct page *page)
864 {
865 return PageCompound(page);
866 }
867
868 /*
869 * PageTransTail returns true for both transparent huge pages
870 * and hugetlbfs pages, so it should only be called when it's known
871 * that hugetlbfs pages aren't involved.
872 */
PageTransTail(struct page * page)873 static inline int PageTransTail(struct page *page)
874 {
875 return PageTail(page);
876 }
877
878 /*
879 * PageDoubleMap indicates that the compound page is mapped with PTEs as well
880 * as PMDs.
881 *
882 * This is required for optimization of rmap operations for THP: we can postpone
883 * per small page mapcount accounting (and its overhead from atomic operations)
884 * until the first PMD split.
885 *
886 * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up
887 * by one. This reference will go away with last compound_mapcount.
888 *
889 * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap().
890 */
PAGEFLAG(DoubleMap,double_map,PF_SECOND)891 PAGEFLAG(DoubleMap, double_map, PF_SECOND)
892 TESTSCFLAG(DoubleMap, double_map, PF_SECOND)
893 #else
894 TESTPAGEFLAG_FALSE(TransHuge, transhuge)
895 TESTPAGEFLAG_FALSE(TransCompound, transcompound)
896 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap)
897 TESTPAGEFLAG_FALSE(TransTail, transtail)
898 PAGEFLAG_FALSE(DoubleMap, double_map)
899 TESTSCFLAG_FALSE(DoubleMap, double_map)
900 #endif
901
902 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
903 /*
904 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
905 * compound page.
906 *
907 * This flag is set by hwpoison handler. Cleared by THP split or free page.
908 */
909 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
910 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
911 #else
912 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
913 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
914 #endif
915
916 /*
917 * Check if a page is currently marked HWPoisoned. Note that this check is
918 * best effort only and inherently racy: there is no way to synchronize with
919 * failing hardware.
920 */
921 static inline bool is_page_hwpoison(struct page *page)
922 {
923 if (PageHWPoison(page))
924 return true;
925 return PageHuge(page) && PageHWPoison(compound_head(page));
926 }
927
928 /*
929 * For pages that are never mapped to userspace (and aren't PageSlab),
930 * page_type may be used. Because it is initialised to -1, we invert the
931 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
932 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
933 * low bits so that an underflow or overflow of page_mapcount() won't be
934 * mistaken for a page type value.
935 */
936
937 #define PAGE_TYPE_BASE 0xf0000000
938 /* Reserve 0x0000007f to catch underflows of page_mapcount */
939 #define PAGE_MAPCOUNT_RESERVE -128
940 #define PG_buddy 0x00000080
941 #define PG_offline 0x00000100
942 #define PG_table 0x00000200
943 #define PG_guard 0x00000400
944
945 #define PageType(page, flag) \
946 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
947
page_has_type(struct page * page)948 static inline int page_has_type(struct page *page)
949 {
950 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE;
951 }
952
953 #define PAGE_TYPE_OPS(uname, lname) \
954 static __always_inline int Page##uname(struct page *page) \
955 { \
956 return PageType(page, PG_##lname); \
957 } \
958 static __always_inline void __SetPage##uname(struct page *page) \
959 { \
960 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
961 page->page_type &= ~PG_##lname; \
962 } \
963 static __always_inline void __ClearPage##uname(struct page *page) \
964 { \
965 VM_BUG_ON_PAGE(!Page##uname(page), page); \
966 page->page_type |= PG_##lname; \
967 }
968
969 /*
970 * PageBuddy() indicates that the page is free and in the buddy system
971 * (see mm/page_alloc.c).
972 */
973 PAGE_TYPE_OPS(Buddy, buddy)
974
975 /*
976 * PageOffline() indicates that the page is logically offline although the
977 * containing section is online. (e.g. inflated in a balloon driver or
978 * not onlined when onlining the section).
979 * The content of these pages is effectively stale. Such pages should not
980 * be touched (read/write/dump/save) except by their owner.
981 *
982 * If a driver wants to allow to offline unmovable PageOffline() pages without
983 * putting them back to the buddy, it can do so via the memory notifier by
984 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
985 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
986 * pages (now with a reference count of zero) are treated like free pages,
987 * allowing the containing memory block to get offlined. A driver that
988 * relies on this feature is aware that re-onlining the memory block will
989 * require to re-set the pages PageOffline() and not giving them to the
990 * buddy via online_page_callback_t.
991 *
992 * There are drivers that mark a page PageOffline() and expect there won't be
993 * any further access to page content. PFN walkers that read content of random
994 * pages should check PageOffline() and synchronize with such drivers using
995 * page_offline_freeze()/page_offline_thaw().
996 */
997 PAGE_TYPE_OPS(Offline, offline)
998
999 extern void page_offline_freeze(void);
1000 extern void page_offline_thaw(void);
1001 extern void page_offline_begin(void);
1002 extern void page_offline_end(void);
1003
1004 /*
1005 * Marks pages in use as page tables.
1006 */
1007 PAGE_TYPE_OPS(Table, table)
1008
1009 /*
1010 * Marks guardpages used with debug_pagealloc.
1011 */
1012 PAGE_TYPE_OPS(Guard, guard)
1013
1014 extern bool is_free_buddy_page(struct page *page);
1015
1016 PAGEFLAG(Isolated, isolated, PF_ANY);
1017
PageAnonExclusive(struct page * page)1018 static __always_inline int PageAnonExclusive(struct page *page)
1019 {
1020 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1021 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1022 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1023 }
1024
SetPageAnonExclusive(struct page * page)1025 static __always_inline void SetPageAnonExclusive(struct page *page)
1026 {
1027 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1028 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1029 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1030 }
1031
ClearPageAnonExclusive(struct page * page)1032 static __always_inline void ClearPageAnonExclusive(struct page *page)
1033 {
1034 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page);
1035 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1036 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1037 }
1038
__ClearPageAnonExclusive(struct page * page)1039 static __always_inline void __ClearPageAnonExclusive(struct page *page)
1040 {
1041 VM_BUG_ON_PGFLAGS(!PageAnon(page), page);
1042 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page);
1043 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags);
1044 }
1045
1046 #ifdef CONFIG_MMU
1047 #define __PG_MLOCKED (1UL << PG_mlocked)
1048 #else
1049 #define __PG_MLOCKED 0
1050 #endif
1051
1052 /*
1053 * Flags checked when a page is freed. Pages being freed should not have
1054 * these flags set. If they are, there is a problem.
1055 */
1056 #define PAGE_FLAGS_CHECK_AT_FREE \
1057 (1UL << PG_lru | 1UL << PG_locked | \
1058 1UL << PG_private | 1UL << PG_private_2 | \
1059 1UL << PG_writeback | 1UL << PG_reserved | \
1060 1UL << PG_slab | 1UL << PG_active | \
1061 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK)
1062
1063 /*
1064 * Flags checked when a page is prepped for return by the page allocator.
1065 * Pages being prepped should not have these flags set. If they are set,
1066 * there has been a kernel bug or struct page corruption.
1067 *
1068 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
1069 * alloc-free cycle to prevent from reusing the page.
1070 */
1071 #define PAGE_FLAGS_CHECK_AT_PREP \
1072 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK)
1073
1074 #define PAGE_FLAGS_PRIVATE \
1075 (1UL << PG_private | 1UL << PG_private_2)
1076 /**
1077 * page_has_private - Determine if page has private stuff
1078 * @page: The page to be checked
1079 *
1080 * Determine if a page has private stuff, indicating that release routines
1081 * should be invoked upon it.
1082 */
page_has_private(struct page * page)1083 static inline int page_has_private(struct page *page)
1084 {
1085 return !!(page->flags & PAGE_FLAGS_PRIVATE);
1086 }
1087
folio_has_private(struct folio * folio)1088 static inline bool folio_has_private(struct folio *folio)
1089 {
1090 return page_has_private(&folio->page);
1091 }
1092
1093 #undef PF_ANY
1094 #undef PF_HEAD
1095 #undef PF_ONLY_HEAD
1096 #undef PF_NO_TAIL
1097 #undef PF_NO_COMPOUND
1098 #undef PF_SECOND
1099 #endif /* !__GENERATING_BOUNDS_H */
1100
1101 #endif /* PAGE_FLAGS_H */
1102