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_writeback, /* Page is under writeback */ |
103 | PG_referenced, |
104 | PG_uptodate, |
105 | PG_dirty, |
106 | PG_lru, |
107 | PG_head, /* Must be in bit 6 */ |
108 | PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */ |
109 | PG_active, |
110 | PG_workingset, |
111 | PG_error, |
112 | PG_slab, |
113 | PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/ |
114 | PG_arch_1, |
115 | PG_reserved, |
116 | PG_private, /* If pagecache, has fs-private data */ |
117 | PG_private_2, /* If pagecache, has fs aux data */ |
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_ARCH_USES_PG_ARCH_X |
136 | PG_arch_2, |
137 | PG_arch_3, |
138 | #endif |
139 | __NR_PAGEFLAGS, |
140 | |
141 | PG_readahead = PG_reclaim, |
142 | |
143 | /* |
144 | * Depending on the way an anonymous folio can be mapped into a page |
145 | * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped |
146 | * THP), PG_anon_exclusive may be set only for the head page or for |
147 | * tail pages of an anonymous folio. For now, we only expect it to be |
148 | * set on tail pages for PTE-mapped THP. |
149 | */ |
150 | PG_anon_exclusive = PG_mappedtodisk, |
151 | |
152 | /* Filesystems */ |
153 | PG_checked = PG_owner_priv_1, |
154 | |
155 | /* SwapBacked */ |
156 | PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */ |
157 | |
158 | /* Two page bits are conscripted by FS-Cache to maintain local caching |
159 | * state. These bits are set on pages belonging to the netfs's inodes |
160 | * when those inodes are being locally cached. |
161 | */ |
162 | PG_fscache = PG_private_2, /* page backed by cache */ |
163 | |
164 | /* XEN */ |
165 | /* Pinned in Xen as a read-only pagetable page. */ |
166 | PG_pinned = PG_owner_priv_1, |
167 | /* Pinned as part of domain save (see xen_mm_pin_all()). */ |
168 | PG_savepinned = PG_dirty, |
169 | /* Has a grant mapping of another (foreign) domain's page. */ |
170 | PG_foreign = PG_owner_priv_1, |
171 | /* Remapped by swiotlb-xen. */ |
172 | PG_xen_remapped = PG_owner_priv_1, |
173 | |
174 | /* non-lru isolated movable page */ |
175 | PG_isolated = PG_reclaim, |
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_error, |
193 | PG_hugetlb = PG_active, |
194 | PG_large_rmappable = PG_workingset, /* anon or file-backed */ |
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 | */ |
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 | #else |
234 | static inline const struct page *page_fixed_fake_head(const struct page *page) |
235 | { |
236 | return page; |
237 | } |
238 | #endif |
239 | |
240 | static __always_inline int page_is_fake_head(const struct page *page) |
241 | { |
242 | return page_fixed_fake_head(page) != page; |
243 | } |
244 | |
245 | static inline unsigned long _compound_head(const struct page *page) |
246 | { |
247 | unsigned long head = READ_ONCE(page->compound_head); |
248 | |
249 | if (unlikely(head & 1)) |
250 | return head - 1; |
251 | return (unsigned long)page_fixed_fake_head(page); |
252 | } |
253 | |
254 | #define compound_head(page) ((typeof(page))_compound_head(page)) |
255 | |
256 | /** |
257 | * page_folio - Converts from page to folio. |
258 | * @p: The page. |
259 | * |
260 | * Every page is part of a folio. This function cannot be called on a |
261 | * NULL pointer. |
262 | * |
263 | * Context: No reference, nor lock is required on @page. If the caller |
264 | * does not hold a reference, this call may race with a folio split, so |
265 | * it should re-check the folio still contains this page after gaining |
266 | * a reference on the folio. |
267 | * Return: The folio which contains this page. |
268 | */ |
269 | #define page_folio(p) (_Generic((p), \ |
270 | const struct page *: (const struct folio *)_compound_head(p), \ |
271 | struct page *: (struct folio *)_compound_head(p))) |
272 | |
273 | /** |
274 | * folio_page - Return a page from a folio. |
275 | * @folio: The folio. |
276 | * @n: The page number to return. |
277 | * |
278 | * @n is relative to the start of the folio. This function does not |
279 | * check that the page number lies within @folio; the caller is presumed |
280 | * to have a reference to the page. |
281 | */ |
282 | #define folio_page(folio, n) nth_page(&(folio)->page, n) |
283 | |
284 | static __always_inline int PageTail(const struct page *page) |
285 | { |
286 | return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page); |
287 | } |
288 | |
289 | static __always_inline int PageCompound(const struct page *page) |
290 | { |
291 | return test_bit(PG_head, &page->flags) || |
292 | READ_ONCE(page->compound_head) & 1; |
293 | } |
294 | |
295 | #define PAGE_POISON_PATTERN -1l |
296 | static inline int PagePoisoned(const struct page *page) |
297 | { |
298 | return READ_ONCE(page->flags) == PAGE_POISON_PATTERN; |
299 | } |
300 | |
301 | #ifdef CONFIG_DEBUG_VM |
302 | void page_init_poison(struct page *page, size_t size); |
303 | #else |
304 | static inline void page_init_poison(struct page *page, size_t size) |
305 | { |
306 | } |
307 | #endif |
308 | |
309 | static const unsigned long *const_folio_flags(const struct folio *folio, |
310 | unsigned n) |
311 | { |
312 | const struct page *page = &folio->page; |
313 | |
314 | VM_BUG_ON_PGFLAGS(PageTail(page), page); |
315 | VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); |
316 | return &page[n].flags; |
317 | } |
318 | |
319 | static unsigned long *folio_flags(struct folio *folio, unsigned n) |
320 | { |
321 | struct page *page = &folio->page; |
322 | |
323 | VM_BUG_ON_PGFLAGS(PageTail(page), page); |
324 | VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); |
325 | return &page[n].flags; |
326 | } |
327 | |
328 | /* |
329 | * Page flags policies wrt compound pages |
330 | * |
331 | * PF_POISONED_CHECK |
332 | * check if this struct page poisoned/uninitialized |
333 | * |
334 | * PF_ANY: |
335 | * the page flag is relevant for small, head and tail pages. |
336 | * |
337 | * PF_HEAD: |
338 | * for compound page all operations related to the page flag applied to |
339 | * head page. |
340 | * |
341 | * PF_NO_TAIL: |
342 | * modifications of the page flag must be done on small or head pages, |
343 | * checks can be done on tail pages too. |
344 | * |
345 | * PF_NO_COMPOUND: |
346 | * the page flag is not relevant for compound pages. |
347 | * |
348 | * PF_SECOND: |
349 | * the page flag is stored in the first tail page. |
350 | */ |
351 | #define PF_POISONED_CHECK(page) ({ \ |
352 | VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ |
353 | page; }) |
354 | #define PF_ANY(page, enforce) PF_POISONED_CHECK(page) |
355 | #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) |
356 | #define PF_NO_TAIL(page, enforce) ({ \ |
357 | VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ |
358 | PF_POISONED_CHECK(compound_head(page)); }) |
359 | #define PF_NO_COMPOUND(page, enforce) ({ \ |
360 | VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ |
361 | PF_POISONED_CHECK(page); }) |
362 | #define PF_SECOND(page, enforce) ({ \ |
363 | VM_BUG_ON_PGFLAGS(!PageHead(page), page); \ |
364 | PF_POISONED_CHECK(&page[1]); }) |
365 | |
366 | /* Which page is the flag stored in */ |
367 | #define FOLIO_PF_ANY 0 |
368 | #define FOLIO_PF_HEAD 0 |
369 | #define FOLIO_PF_NO_TAIL 0 |
370 | #define FOLIO_PF_NO_COMPOUND 0 |
371 | #define FOLIO_PF_SECOND 1 |
372 | |
373 | #define FOLIO_HEAD_PAGE 0 |
374 | #define FOLIO_SECOND_PAGE 1 |
375 | |
376 | /* |
377 | * Macros to create function definitions for page flags |
378 | */ |
379 | #define FOLIO_TEST_FLAG(name, page) \ |
380 | static __always_inline bool folio_test_##name(const struct folio *folio) \ |
381 | { return test_bit(PG_##name, const_folio_flags(folio, page)); } |
382 | |
383 | #define FOLIO_SET_FLAG(name, page) \ |
384 | static __always_inline void folio_set_##name(struct folio *folio) \ |
385 | { set_bit(PG_##name, folio_flags(folio, page)); } |
386 | |
387 | #define FOLIO_CLEAR_FLAG(name, page) \ |
388 | static __always_inline void folio_clear_##name(struct folio *folio) \ |
389 | { clear_bit(PG_##name, folio_flags(folio, page)); } |
390 | |
391 | #define __FOLIO_SET_FLAG(name, page) \ |
392 | static __always_inline void __folio_set_##name(struct folio *folio) \ |
393 | { __set_bit(PG_##name, folio_flags(folio, page)); } |
394 | |
395 | #define __FOLIO_CLEAR_FLAG(name, page) \ |
396 | static __always_inline void __folio_clear_##name(struct folio *folio) \ |
397 | { __clear_bit(PG_##name, folio_flags(folio, page)); } |
398 | |
399 | #define FOLIO_TEST_SET_FLAG(name, page) \ |
400 | static __always_inline bool folio_test_set_##name(struct folio *folio) \ |
401 | { return test_and_set_bit(PG_##name, folio_flags(folio, page)); } |
402 | |
403 | #define FOLIO_TEST_CLEAR_FLAG(name, page) \ |
404 | static __always_inline bool folio_test_clear_##name(struct folio *folio) \ |
405 | { return test_and_clear_bit(PG_##name, folio_flags(folio, page)); } |
406 | |
407 | #define FOLIO_FLAG(name, page) \ |
408 | FOLIO_TEST_FLAG(name, page) \ |
409 | FOLIO_SET_FLAG(name, page) \ |
410 | FOLIO_CLEAR_FLAG(name, page) |
411 | |
412 | #define TESTPAGEFLAG(uname, lname, policy) \ |
413 | FOLIO_TEST_FLAG(lname, FOLIO_##policy) \ |
414 | static __always_inline int Page##uname(const struct page *page) \ |
415 | { return test_bit(PG_##lname, &policy(page, 0)->flags); } |
416 | |
417 | #define SETPAGEFLAG(uname, lname, policy) \ |
418 | FOLIO_SET_FLAG(lname, FOLIO_##policy) \ |
419 | static __always_inline void SetPage##uname(struct page *page) \ |
420 | { set_bit(PG_##lname, &policy(page, 1)->flags); } |
421 | |
422 | #define CLEARPAGEFLAG(uname, lname, policy) \ |
423 | FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \ |
424 | static __always_inline void ClearPage##uname(struct page *page) \ |
425 | { clear_bit(PG_##lname, &policy(page, 1)->flags); } |
426 | |
427 | #define __SETPAGEFLAG(uname, lname, policy) \ |
428 | __FOLIO_SET_FLAG(lname, FOLIO_##policy) \ |
429 | static __always_inline void __SetPage##uname(struct page *page) \ |
430 | { __set_bit(PG_##lname, &policy(page, 1)->flags); } |
431 | |
432 | #define __CLEARPAGEFLAG(uname, lname, policy) \ |
433 | __FOLIO_CLEAR_FLAG(lname, FOLIO_##policy) \ |
434 | static __always_inline void __ClearPage##uname(struct page *page) \ |
435 | { __clear_bit(PG_##lname, &policy(page, 1)->flags); } |
436 | |
437 | #define TESTSETFLAG(uname, lname, policy) \ |
438 | FOLIO_TEST_SET_FLAG(lname, FOLIO_##policy) \ |
439 | static __always_inline int TestSetPage##uname(struct page *page) \ |
440 | { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } |
441 | |
442 | #define TESTCLEARFLAG(uname, lname, policy) \ |
443 | FOLIO_TEST_CLEAR_FLAG(lname, FOLIO_##policy) \ |
444 | static __always_inline int TestClearPage##uname(struct page *page) \ |
445 | { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } |
446 | |
447 | #define PAGEFLAG(uname, lname, policy) \ |
448 | TESTPAGEFLAG(uname, lname, policy) \ |
449 | SETPAGEFLAG(uname, lname, policy) \ |
450 | CLEARPAGEFLAG(uname, lname, policy) |
451 | |
452 | #define __PAGEFLAG(uname, lname, policy) \ |
453 | TESTPAGEFLAG(uname, lname, policy) \ |
454 | __SETPAGEFLAG(uname, lname, policy) \ |
455 | __CLEARPAGEFLAG(uname, lname, policy) |
456 | |
457 | #define TESTSCFLAG(uname, lname, policy) \ |
458 | TESTSETFLAG(uname, lname, policy) \ |
459 | TESTCLEARFLAG(uname, lname, policy) |
460 | |
461 | #define TESTPAGEFLAG_FALSE(uname, lname) \ |
462 | static inline bool folio_test_##lname(const struct folio *folio) { return false; } \ |
463 | static inline int Page##uname(const struct page *page) { return 0; } |
464 | |
465 | #define SETPAGEFLAG_NOOP(uname, lname) \ |
466 | static inline void folio_set_##lname(struct folio *folio) { } \ |
467 | static inline void SetPage##uname(struct page *page) { } |
468 | |
469 | #define CLEARPAGEFLAG_NOOP(uname, lname) \ |
470 | static inline void folio_clear_##lname(struct folio *folio) { } \ |
471 | static inline void ClearPage##uname(struct page *page) { } |
472 | |
473 | #define __CLEARPAGEFLAG_NOOP(uname, lname) \ |
474 | static inline void __folio_clear_##lname(struct folio *folio) { } \ |
475 | static inline void __ClearPage##uname(struct page *page) { } |
476 | |
477 | #define TESTSETFLAG_FALSE(uname, lname) \ |
478 | static inline bool folio_test_set_##lname(struct folio *folio) \ |
479 | { return 0; } \ |
480 | static inline int TestSetPage##uname(struct page *page) { return 0; } |
481 | |
482 | #define TESTCLEARFLAG_FALSE(uname, lname) \ |
483 | static inline bool folio_test_clear_##lname(struct folio *folio) \ |
484 | { return 0; } \ |
485 | static inline int TestClearPage##uname(struct page *page) { return 0; } |
486 | |
487 | #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \ |
488 | SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname) |
489 | |
490 | #define TESTSCFLAG_FALSE(uname, lname) \ |
491 | TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname) |
492 | |
493 | __PAGEFLAG(Locked, locked, PF_NO_TAIL) |
494 | FOLIO_FLAG(waiters, FOLIO_HEAD_PAGE) |
495 | PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL) |
496 | PAGEFLAG(Referenced, referenced, PF_HEAD) |
497 | TESTCLEARFLAG(Referenced, referenced, PF_HEAD) |
498 | __SETPAGEFLAG(Referenced, referenced, PF_HEAD) |
499 | PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) |
500 | __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) |
501 | PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) |
502 | TESTCLEARFLAG(LRU, lru, PF_HEAD) |
503 | PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) |
504 | TESTCLEARFLAG(Active, active, PF_HEAD) |
505 | PAGEFLAG(Workingset, workingset, PF_HEAD) |
506 | TESTCLEARFLAG(Workingset, workingset, PF_HEAD) |
507 | __PAGEFLAG(Slab, slab, PF_NO_TAIL) |
508 | PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ |
509 | |
510 | /* Xen */ |
511 | PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) |
512 | TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) |
513 | PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); |
514 | PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); |
515 | PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) |
516 | TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) |
517 | |
518 | PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) |
519 | __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) |
520 | __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) |
521 | PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) |
522 | __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) |
523 | __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) |
524 | |
525 | /* |
526 | * Private page markings that may be used by the filesystem that owns the page |
527 | * for its own purposes. |
528 | * - PG_private and PG_private_2 cause release_folio() and co to be invoked |
529 | */ |
530 | PAGEFLAG(Private, private, PF_ANY) |
531 | PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) |
532 | PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) |
533 | TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) |
534 | |
535 | /* |
536 | * Only test-and-set exist for PG_writeback. The unconditional operators are |
537 | * risky: they bypass page accounting. |
538 | */ |
539 | TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) |
540 | TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) |
541 | PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) |
542 | |
543 | /* PG_readahead is only used for reads; PG_reclaim is only for writes */ |
544 | PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) |
545 | TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) |
546 | PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND) |
547 | TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND) |
548 | |
549 | #ifdef CONFIG_HIGHMEM |
550 | /* |
551 | * Must use a macro here due to header dependency issues. page_zone() is not |
552 | * available at this point. |
553 | */ |
554 | #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) |
555 | #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f)) |
556 | #else |
557 | PAGEFLAG_FALSE(HighMem, highmem) |
558 | #endif |
559 | |
560 | #ifdef CONFIG_SWAP |
561 | static __always_inline bool folio_test_swapcache(const struct folio *folio) |
562 | { |
563 | return folio_test_swapbacked(folio) && |
564 | test_bit(PG_swapcache, const_folio_flags(folio, 0)); |
565 | } |
566 | |
567 | static __always_inline bool PageSwapCache(const struct page *page) |
568 | { |
569 | return folio_test_swapcache(page_folio(page)); |
570 | } |
571 | |
572 | SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) |
573 | CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) |
574 | #else |
575 | PAGEFLAG_FALSE(SwapCache, swapcache) |
576 | #endif |
577 | |
578 | PAGEFLAG(Unevictable, unevictable, PF_HEAD) |
579 | __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) |
580 | TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) |
581 | |
582 | #ifdef CONFIG_MMU |
583 | PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) |
584 | __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) |
585 | TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) |
586 | #else |
587 | PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked) |
588 | TESTSCFLAG_FALSE(Mlocked, mlocked) |
589 | #endif |
590 | |
591 | #ifdef CONFIG_ARCH_USES_PG_UNCACHED |
592 | PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) |
593 | #else |
594 | PAGEFLAG_FALSE(Uncached, uncached) |
595 | #endif |
596 | |
597 | #ifdef CONFIG_MEMORY_FAILURE |
598 | PAGEFLAG(HWPoison, hwpoison, PF_ANY) |
599 | TESTSCFLAG(HWPoison, hwpoison, PF_ANY) |
600 | #define __PG_HWPOISON (1UL << PG_hwpoison) |
601 | #define MAGIC_HWPOISON 0x48575053U /* HWPS */ |
602 | extern void SetPageHWPoisonTakenOff(struct page *page); |
603 | extern void ClearPageHWPoisonTakenOff(struct page *page); |
604 | extern bool take_page_off_buddy(struct page *page); |
605 | extern bool put_page_back_buddy(struct page *page); |
606 | #else |
607 | PAGEFLAG_FALSE(HWPoison, hwpoison) |
608 | #define __PG_HWPOISON 0 |
609 | #endif |
610 | |
611 | #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) |
612 | FOLIO_TEST_FLAG(young, FOLIO_HEAD_PAGE) |
613 | FOLIO_SET_FLAG(young, FOLIO_HEAD_PAGE) |
614 | FOLIO_TEST_CLEAR_FLAG(young, FOLIO_HEAD_PAGE) |
615 | FOLIO_FLAG(idle, FOLIO_HEAD_PAGE) |
616 | #endif |
617 | |
618 | /* |
619 | * PageReported() is used to track reported free pages within the Buddy |
620 | * allocator. We can use the non-atomic version of the test and set |
621 | * operations as both should be shielded with the zone lock to prevent |
622 | * any possible races on the setting or clearing of the bit. |
623 | */ |
624 | __PAGEFLAG(Reported, reported, PF_NO_COMPOUND) |
625 | |
626 | #ifdef CONFIG_MEMORY_HOTPLUG |
627 | PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY) |
628 | #else |
629 | PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted) |
630 | #endif |
631 | |
632 | /* |
633 | * On an anonymous page mapped into a user virtual memory area, |
634 | * page->mapping points to its anon_vma, not to a struct address_space; |
635 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. |
636 | * |
637 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, |
638 | * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON |
639 | * bit; and then page->mapping points, not to an anon_vma, but to a private |
640 | * structure which KSM associates with that merged page. See ksm.h. |
641 | * |
642 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable |
643 | * page and then page->mapping points to a struct movable_operations. |
644 | * |
645 | * Please note that, confusingly, "page_mapping" refers to the inode |
646 | * address_space which maps the page from disk; whereas "page_mapped" |
647 | * refers to user virtual address space into which the page is mapped. |
648 | * |
649 | * For slab pages, since slab reuses the bits in struct page to store its |
650 | * internal states, the page->mapping does not exist as such, nor do these |
651 | * flags below. So in order to avoid testing non-existent bits, please |
652 | * make sure that PageSlab(page) actually evaluates to false before calling |
653 | * the following functions (e.g., PageAnon). See mm/slab.h. |
654 | */ |
655 | #define PAGE_MAPPING_ANON 0x1 |
656 | #define PAGE_MAPPING_MOVABLE 0x2 |
657 | #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) |
658 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) |
659 | |
660 | /* |
661 | * Different with flags above, this flag is used only for fsdax mode. It |
662 | * indicates that this page->mapping is now under reflink case. |
663 | */ |
664 | #define PAGE_MAPPING_DAX_SHARED ((void *)0x1) |
665 | |
666 | static __always_inline bool folio_mapping_flags(const struct folio *folio) |
667 | { |
668 | return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0; |
669 | } |
670 | |
671 | static __always_inline int PageMappingFlags(const struct page *page) |
672 | { |
673 | return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; |
674 | } |
675 | |
676 | static __always_inline bool folio_test_anon(const struct folio *folio) |
677 | { |
678 | return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0; |
679 | } |
680 | |
681 | static __always_inline bool PageAnon(const struct page *page) |
682 | { |
683 | return folio_test_anon(page_folio(page)); |
684 | } |
685 | |
686 | static __always_inline bool __folio_test_movable(const struct folio *folio) |
687 | { |
688 | return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == |
689 | PAGE_MAPPING_MOVABLE; |
690 | } |
691 | |
692 | static __always_inline int __PageMovable(const struct page *page) |
693 | { |
694 | return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == |
695 | PAGE_MAPPING_MOVABLE; |
696 | } |
697 | |
698 | #ifdef CONFIG_KSM |
699 | /* |
700 | * A KSM page is one of those write-protected "shared pages" or "merged pages" |
701 | * which KSM maps into multiple mms, wherever identical anonymous page content |
702 | * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any |
703 | * anon_vma, but to that page's node of the stable tree. |
704 | */ |
705 | static __always_inline bool folio_test_ksm(const struct folio *folio) |
706 | { |
707 | return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == |
708 | PAGE_MAPPING_KSM; |
709 | } |
710 | |
711 | static __always_inline bool PageKsm(const struct page *page) |
712 | { |
713 | return folio_test_ksm(page_folio(page)); |
714 | } |
715 | #else |
716 | TESTPAGEFLAG_FALSE(Ksm, ksm) |
717 | #endif |
718 | |
719 | u64 stable_page_flags(struct page *page); |
720 | |
721 | /** |
722 | * folio_xor_flags_has_waiters - Change some folio flags. |
723 | * @folio: The folio. |
724 | * @mask: Bits set in this word will be changed. |
725 | * |
726 | * This must only be used for flags which are changed with the folio |
727 | * lock held. For example, it is unsafe to use for PG_dirty as that |
728 | * can be set without the folio lock held. It can also only be used |
729 | * on flags which are in the range 0-6 as some of the implementations |
730 | * only affect those bits. |
731 | * |
732 | * Return: Whether there are tasks waiting on the folio. |
733 | */ |
734 | static inline bool folio_xor_flags_has_waiters(struct folio *folio, |
735 | unsigned long mask) |
736 | { |
737 | return xor_unlock_is_negative_byte(mask, addr: folio_flags(folio, n: 0)); |
738 | } |
739 | |
740 | /** |
741 | * folio_test_uptodate - Is this folio up to date? |
742 | * @folio: The folio. |
743 | * |
744 | * The uptodate flag is set on a folio when every byte in the folio is |
745 | * at least as new as the corresponding bytes on storage. Anonymous |
746 | * and CoW folios are always uptodate. If the folio is not uptodate, |
747 | * some of the bytes in it may be; see the is_partially_uptodate() |
748 | * address_space operation. |
749 | */ |
750 | static inline bool folio_test_uptodate(const struct folio *folio) |
751 | { |
752 | bool ret = test_bit(PG_uptodate, const_folio_flags(folio, 0)); |
753 | /* |
754 | * Must ensure that the data we read out of the folio is loaded |
755 | * _after_ we've loaded folio->flags to check the uptodate bit. |
756 | * We can skip the barrier if the folio is not uptodate, because |
757 | * we wouldn't be reading anything from it. |
758 | * |
759 | * See folio_mark_uptodate() for the other side of the story. |
760 | */ |
761 | if (ret) |
762 | smp_rmb(); |
763 | |
764 | return ret; |
765 | } |
766 | |
767 | static inline int PageUptodate(const struct page *page) |
768 | { |
769 | return folio_test_uptodate(page_folio(page)); |
770 | } |
771 | |
772 | static __always_inline void __folio_mark_uptodate(struct folio *folio) |
773 | { |
774 | smp_wmb(); |
775 | __set_bit(PG_uptodate, folio_flags(folio, 0)); |
776 | } |
777 | |
778 | static __always_inline void folio_mark_uptodate(struct folio *folio) |
779 | { |
780 | /* |
781 | * Memory barrier must be issued before setting the PG_uptodate bit, |
782 | * so that all previous stores issued in order to bring the folio |
783 | * uptodate are actually visible before folio_test_uptodate becomes true. |
784 | */ |
785 | smp_wmb(); |
786 | set_bit(nr: PG_uptodate, addr: folio_flags(folio, n: 0)); |
787 | } |
788 | |
789 | static __always_inline void __SetPageUptodate(struct page *page) |
790 | { |
791 | __folio_mark_uptodate(folio: (struct folio *)page); |
792 | } |
793 | |
794 | static __always_inline void SetPageUptodate(struct page *page) |
795 | { |
796 | folio_mark_uptodate(folio: (struct folio *)page); |
797 | } |
798 | |
799 | CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) |
800 | |
801 | void __folio_start_writeback(struct folio *folio, bool keep_write); |
802 | void set_page_writeback(struct page *page); |
803 | |
804 | #define folio_start_writeback(folio) \ |
805 | __folio_start_writeback(folio, false) |
806 | #define folio_start_writeback_keepwrite(folio) \ |
807 | __folio_start_writeback(folio, true) |
808 | |
809 | static __always_inline bool folio_test_head(const struct folio *folio) |
810 | { |
811 | return test_bit(PG_head, const_folio_flags(folio, FOLIO_PF_ANY)); |
812 | } |
813 | |
814 | static __always_inline int PageHead(const struct page *page) |
815 | { |
816 | PF_POISONED_CHECK(page); |
817 | return test_bit(PG_head, &page->flags) && !page_is_fake_head(page); |
818 | } |
819 | |
820 | __SETPAGEFLAG(Head, head, PF_ANY) |
821 | __CLEARPAGEFLAG(Head, head, PF_ANY) |
822 | CLEARPAGEFLAG(Head, head, PF_ANY) |
823 | |
824 | /** |
825 | * folio_test_large() - Does this folio contain more than one page? |
826 | * @folio: The folio to test. |
827 | * |
828 | * Return: True if the folio is larger than one page. |
829 | */ |
830 | static inline bool folio_test_large(const struct folio *folio) |
831 | { |
832 | return folio_test_head(folio); |
833 | } |
834 | |
835 | static __always_inline void set_compound_head(struct page *page, struct page *head) |
836 | { |
837 | WRITE_ONCE(page->compound_head, (unsigned long)head + 1); |
838 | } |
839 | |
840 | static __always_inline void clear_compound_head(struct page *page) |
841 | { |
842 | WRITE_ONCE(page->compound_head, 0); |
843 | } |
844 | |
845 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
846 | static inline void ClearPageCompound(struct page *page) |
847 | { |
848 | BUG_ON(!PageHead(page)); |
849 | ClearPageHead(page); |
850 | } |
851 | PAGEFLAG(LargeRmappable, large_rmappable, PF_SECOND) |
852 | #else |
853 | TESTPAGEFLAG_FALSE(LargeRmappable, large_rmappable) |
854 | #endif |
855 | |
856 | #define PG_head_mask ((1UL << PG_head)) |
857 | |
858 | #ifdef CONFIG_HUGETLB_PAGE |
859 | int PageHuge(const struct page *page); |
860 | SETPAGEFLAG(HugeTLB, hugetlb, PF_SECOND) |
861 | CLEARPAGEFLAG(HugeTLB, hugetlb, PF_SECOND) |
862 | |
863 | /** |
864 | * folio_test_hugetlb - Determine if the folio belongs to hugetlbfs |
865 | * @folio: The folio to test. |
866 | * |
867 | * Context: Any context. Caller should have a reference on the folio to |
868 | * prevent it from being turned into a tail page. |
869 | * Return: True for hugetlbfs folios, false for anon folios or folios |
870 | * belonging to other filesystems. |
871 | */ |
872 | static inline bool folio_test_hugetlb(const struct folio *folio) |
873 | { |
874 | return folio_test_large(folio) && |
875 | test_bit(PG_hugetlb, const_folio_flags(folio, 1)); |
876 | } |
877 | #else |
878 | TESTPAGEFLAG_FALSE(Huge, hugetlb) |
879 | #endif |
880 | |
881 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
882 | /* |
883 | * PageHuge() only returns true for hugetlbfs pages, but not for |
884 | * normal or transparent huge pages. |
885 | * |
886 | * PageTransHuge() returns true for both transparent huge and |
887 | * hugetlbfs pages, but not normal pages. PageTransHuge() can only be |
888 | * called only in the core VM paths where hugetlbfs pages can't exist. |
889 | */ |
890 | static inline int PageTransHuge(const struct page *page) |
891 | { |
892 | VM_BUG_ON_PAGE(PageTail(page), page); |
893 | return PageHead(page); |
894 | } |
895 | |
896 | /* |
897 | * PageTransCompound returns true for both transparent huge pages |
898 | * and hugetlbfs pages, so it should only be called when it's known |
899 | * that hugetlbfs pages aren't involved. |
900 | */ |
901 | static inline int PageTransCompound(const struct page *page) |
902 | { |
903 | return PageCompound(page); |
904 | } |
905 | |
906 | /* |
907 | * PageTransTail returns true for both transparent huge pages |
908 | * and hugetlbfs pages, so it should only be called when it's known |
909 | * that hugetlbfs pages aren't involved. |
910 | */ |
911 | static inline int PageTransTail(const struct page *page) |
912 | { |
913 | return PageTail(page); |
914 | } |
915 | #else |
916 | TESTPAGEFLAG_FALSE(TransHuge, transhuge) |
917 | TESTPAGEFLAG_FALSE(TransCompound, transcompound) |
918 | TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap) |
919 | TESTPAGEFLAG_FALSE(TransTail, transtail) |
920 | #endif |
921 | |
922 | #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE) |
923 | /* |
924 | * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the |
925 | * compound page. |
926 | * |
927 | * This flag is set by hwpoison handler. Cleared by THP split or free page. |
928 | */ |
929 | |
---|