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_uptodate tells whether the page's contents is valid. When a read |
67 | * completes, the page becomes uptodate, unless a disk I/O error happened. |
68 | * |
69 | * PG_referenced, PG_reclaim are used for page reclaim for anonymous and |
70 | * file-backed pagecache (see mm/vmscan.c). |
71 | * |
72 | * PG_error is set to indicate that an I/O error occurred on this page. |
73 | * |
74 | * PG_arch_1 is an architecture specific page state bit. The generic code |
75 | * guarantees that this bit is cleared for a page when it first is entered into |
76 | * the page cache. |
77 | * |
78 | * PG_hwpoison indicates that a page got corrupted in hardware and contains |
79 | * data with incorrect ECC bits that triggered a machine check. Accessing is |
80 | * not safe since it may cause another machine check. Don't touch! |
81 | */ |
82 | |
83 | /* |
84 | * Don't use the *_dontuse flags. Use the macros. Otherwise you'll break |
85 | * locked- and dirty-page accounting. |
86 | * |
87 | * The page flags field is split into two parts, the main flags area |
88 | * which extends from the low bits upwards, and the fields area which |
89 | * extends from the high bits downwards. |
90 | * |
91 | * | FIELD | ... | FLAGS | |
92 | * N-1 ^ 0 |
93 | * (NR_PAGEFLAGS) |
94 | * |
95 | * The fields area is reserved for fields mapping zone, node (for NUMA) and |
96 | * SPARSEMEM section (for variants of SPARSEMEM that require section ids like |
97 | * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP). |
98 | */ |
99 | enum pageflags { |
100 | PG_locked, /* Page is locked. Don't touch. */ |
101 | PG_referenced, |
102 | PG_uptodate, |
103 | PG_dirty, |
104 | PG_lru, |
105 | PG_active, |
106 | PG_workingset, |
107 | PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */ |
108 | PG_error, |
109 | PG_slab, |
110 | PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/ |
111 | PG_arch_1, |
112 | PG_reserved, |
113 | PG_private, /* If pagecache, has fs-private data */ |
114 | PG_private_2, /* If pagecache, has fs aux data */ |
115 | PG_writeback, /* Page is under writeback */ |
116 | PG_head, /* A head page */ |
117 | PG_mappedtodisk, /* Has blocks allocated on-disk */ |
118 | PG_reclaim, /* To be reclaimed asap */ |
119 | PG_swapbacked, /* Page is backed by RAM/swap */ |
120 | PG_unevictable, /* Page is "unevictable" */ |
121 | #ifdef CONFIG_MMU |
122 | PG_mlocked, /* Page is vma mlocked */ |
123 | #endif |
124 | #ifdef CONFIG_ARCH_USES_PG_UNCACHED |
125 | PG_uncached, /* Page has been mapped as uncached */ |
126 | #endif |
127 | #ifdef CONFIG_MEMORY_FAILURE |
128 | PG_hwpoison, /* hardware poisoned page. Don't touch */ |
129 | #endif |
130 | #if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT) |
131 | PG_young, |
132 | PG_idle, |
133 | #endif |
134 | __NR_PAGEFLAGS, |
135 | |
136 | /* Filesystems */ |
137 | PG_checked = PG_owner_priv_1, |
138 | |
139 | /* SwapBacked */ |
140 | PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */ |
141 | |
142 | /* Two page bits are conscripted by FS-Cache to maintain local caching |
143 | * state. These bits are set on pages belonging to the netfs's inodes |
144 | * when those inodes are being locally cached. |
145 | */ |
146 | PG_fscache = PG_private_2, /* page backed by cache */ |
147 | |
148 | /* XEN */ |
149 | /* Pinned in Xen as a read-only pagetable page. */ |
150 | PG_pinned = PG_owner_priv_1, |
151 | /* Pinned as part of domain save (see xen_mm_pin_all()). */ |
152 | PG_savepinned = PG_dirty, |
153 | /* Has a grant mapping of another (foreign) domain's page. */ |
154 | PG_foreign = PG_owner_priv_1, |
155 | |
156 | /* SLOB */ |
157 | PG_slob_free = PG_private, |
158 | |
159 | /* Compound pages. Stored in first tail page's flags */ |
160 | PG_double_map = PG_private_2, |
161 | |
162 | /* non-lru isolated movable page */ |
163 | PG_isolated = PG_reclaim, |
164 | }; |
165 | |
166 | #ifndef __GENERATING_BOUNDS_H |
167 | |
168 | struct page; /* forward declaration */ |
169 | |
170 | static inline struct page *compound_head(struct page *page) |
171 | { |
172 | unsigned long head = READ_ONCE(page->compound_head); |
173 | |
174 | if (unlikely(head & 1)) |
175 | return (struct page *) (head - 1); |
176 | return page; |
177 | } |
178 | |
179 | static __always_inline int PageTail(struct page *page) |
180 | { |
181 | return READ_ONCE(page->compound_head) & 1; |
182 | } |
183 | |
184 | static __always_inline int PageCompound(struct page *page) |
185 | { |
186 | return test_bit(PG_head, &page->flags) || PageTail(page); |
187 | } |
188 | |
189 | #define PAGE_POISON_PATTERN -1l |
190 | static inline int PagePoisoned(const struct page *page) |
191 | { |
192 | return page->flags == PAGE_POISON_PATTERN; |
193 | } |
194 | |
195 | #ifdef CONFIG_DEBUG_VM |
196 | void page_init_poison(struct page *page, size_t size); |
197 | #else |
198 | static inline void page_init_poison(struct page *page, size_t size) |
199 | { |
200 | } |
201 | #endif |
202 | |
203 | /* |
204 | * Page flags policies wrt compound pages |
205 | * |
206 | * PF_POISONED_CHECK |
207 | * check if this struct page poisoned/uninitialized |
208 | * |
209 | * PF_ANY: |
210 | * the page flag is relevant for small, head and tail pages. |
211 | * |
212 | * PF_HEAD: |
213 | * for compound page all operations related to the page flag applied to |
214 | * head page. |
215 | * |
216 | * PF_ONLY_HEAD: |
217 | * for compound page, callers only ever operate on the head page. |
218 | * |
219 | * PF_NO_TAIL: |
220 | * modifications of the page flag must be done on small or head pages, |
221 | * checks can be done on tail pages too. |
222 | * |
223 | * PF_NO_COMPOUND: |
224 | * the page flag is not relevant for compound pages. |
225 | */ |
226 | #define PF_POISONED_CHECK(page) ({ \ |
227 | VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ |
228 | page; }) |
229 | #define PF_ANY(page, enforce) PF_POISONED_CHECK(page) |
230 | #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) |
231 | #define PF_ONLY_HEAD(page, enforce) ({ \ |
232 | VM_BUG_ON_PGFLAGS(PageTail(page), page); \ |
233 | PF_POISONED_CHECK(page); }) |
234 | #define PF_NO_TAIL(page, enforce) ({ \ |
235 | VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ |
236 | PF_POISONED_CHECK(compound_head(page)); }) |
237 | #define PF_NO_COMPOUND(page, enforce) ({ \ |
238 | VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ |
239 | PF_POISONED_CHECK(page); }) |
240 | |
241 | /* |
242 | * Macros to create function definitions for page flags |
243 | */ |
244 | #define TESTPAGEFLAG(uname, lname, policy) \ |
245 | static __always_inline int Page##uname(struct page *page) \ |
246 | { return test_bit(PG_##lname, &policy(page, 0)->flags); } |
247 | |
248 | #define SETPAGEFLAG(uname, lname, policy) \ |
249 | static __always_inline void SetPage##uname(struct page *page) \ |
250 | { set_bit(PG_##lname, &policy(page, 1)->flags); } |
251 | |
252 | #define CLEARPAGEFLAG(uname, lname, policy) \ |
253 | static __always_inline void ClearPage##uname(struct page *page) \ |
254 | { clear_bit(PG_##lname, &policy(page, 1)->flags); } |
255 | |
256 | #define __SETPAGEFLAG(uname, lname, policy) \ |
257 | static __always_inline void __SetPage##uname(struct page *page) \ |
258 | { __set_bit(PG_##lname, &policy(page, 1)->flags); } |
259 | |
260 | #define __CLEARPAGEFLAG(uname, lname, policy) \ |
261 | static __always_inline void __ClearPage##uname(struct page *page) \ |
262 | { __clear_bit(PG_##lname, &policy(page, 1)->flags); } |
263 | |
264 | #define TESTSETFLAG(uname, lname, policy) \ |
265 | static __always_inline int TestSetPage##uname(struct page *page) \ |
266 | { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } |
267 | |
268 | #define TESTCLEARFLAG(uname, lname, policy) \ |
269 | static __always_inline int TestClearPage##uname(struct page *page) \ |
270 | { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } |
271 | |
272 | #define PAGEFLAG(uname, lname, policy) \ |
273 | TESTPAGEFLAG(uname, lname, policy) \ |
274 | SETPAGEFLAG(uname, lname, policy) \ |
275 | CLEARPAGEFLAG(uname, lname, policy) |
276 | |
277 | #define __PAGEFLAG(uname, lname, policy) \ |
278 | TESTPAGEFLAG(uname, lname, policy) \ |
279 | __SETPAGEFLAG(uname, lname, policy) \ |
280 | __CLEARPAGEFLAG(uname, lname, policy) |
281 | |
282 | #define TESTSCFLAG(uname, lname, policy) \ |
283 | TESTSETFLAG(uname, lname, policy) \ |
284 | TESTCLEARFLAG(uname, lname, policy) |
285 | |
286 | #define TESTPAGEFLAG_FALSE(uname) \ |
287 | static inline int Page##uname(const struct page *page) { return 0; } |
288 | |
289 | #define SETPAGEFLAG_NOOP(uname) \ |
290 | static inline void SetPage##uname(struct page *page) { } |
291 | |
292 | #define CLEARPAGEFLAG_NOOP(uname) \ |
293 | static inline void ClearPage##uname(struct page *page) { } |
294 | |
295 | #define __CLEARPAGEFLAG_NOOP(uname) \ |
296 | static inline void __ClearPage##uname(struct page *page) { } |
297 | |
298 | #define TESTSETFLAG_FALSE(uname) \ |
299 | static inline int TestSetPage##uname(struct page *page) { return 0; } |
300 | |
301 | #define TESTCLEARFLAG_FALSE(uname) \ |
302 | static inline int TestClearPage##uname(struct page *page) { return 0; } |
303 | |
304 | #define PAGEFLAG_FALSE(uname) TESTPAGEFLAG_FALSE(uname) \ |
305 | SETPAGEFLAG_NOOP(uname) CLEARPAGEFLAG_NOOP(uname) |
306 | |
307 | #define TESTSCFLAG_FALSE(uname) \ |
308 | TESTSETFLAG_FALSE(uname) TESTCLEARFLAG_FALSE(uname) |
309 | |
310 | __PAGEFLAG(Locked, locked, PF_NO_TAIL) |
311 | PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) __CLEARPAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) |
312 | PAGEFLAG(Error, error, PF_NO_COMPOUND) TESTCLEARFLAG(Error, error, PF_NO_COMPOUND) |
313 | PAGEFLAG(Referenced, referenced, PF_HEAD) |
314 | TESTCLEARFLAG(Referenced, referenced, PF_HEAD) |
315 | __SETPAGEFLAG(Referenced, referenced, PF_HEAD) |
316 | PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) |
317 | __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) |
318 | PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) |
319 | PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) |
320 | TESTCLEARFLAG(Active, active, PF_HEAD) |
321 | PAGEFLAG(Workingset, workingset, PF_HEAD) |
322 | TESTCLEARFLAG(Workingset, workingset, PF_HEAD) |
323 | __PAGEFLAG(Slab, slab, PF_NO_TAIL) |
324 | __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL) |
325 | PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ |
326 | |
327 | /* Xen */ |
328 | PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) |
329 | TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) |
330 | PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); |
331 | PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); |
332 | |
333 | PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) |
334 | __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) |
335 | __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) |
336 | PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) |
337 | __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) |
338 | __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) |
339 | |
340 | /* |
341 | * Private page markings that may be used by the filesystem that owns the page |
342 | * for its own purposes. |
343 | * - PG_private and PG_private_2 cause releasepage() and co to be invoked |
344 | */ |
345 | PAGEFLAG(Private, private, PF_ANY) __SETPAGEFLAG(Private, private, PF_ANY) |
346 | __CLEARPAGEFLAG(Private, private, PF_ANY) |
347 | PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) |
348 | PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) |
349 | TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) |
350 | |
351 | /* |
352 | * Only test-and-set exist for PG_writeback. The unconditional operators are |
353 | * risky: they bypass page accounting. |
354 | */ |
355 | TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) |
356 | TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) |
357 | PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) |
358 | |
359 | /* PG_readahead is only used for reads; PG_reclaim is only for writes */ |
360 | PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) |
361 | TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) |
362 | PAGEFLAG(Readahead, reclaim, PF_NO_COMPOUND) |
363 | TESTCLEARFLAG(Readahead, reclaim, PF_NO_COMPOUND) |
364 | |
365 | #ifdef CONFIG_HIGHMEM |
366 | /* |
367 | * Must use a macro here due to header dependency issues. page_zone() is not |
368 | * available at this point. |
369 | */ |
370 | #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) |
371 | #else |
372 | PAGEFLAG_FALSE(HighMem) |
373 | #endif |
374 | |
375 | #ifdef CONFIG_SWAP |
376 | static __always_inline int PageSwapCache(struct page *page) |
377 | { |
378 | #ifdef CONFIG_THP_SWAP |
379 | page = compound_head(page); |
380 | #endif |
381 | return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags); |
382 | |
383 | } |
384 | SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) |
385 | CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) |
386 | #else |
387 | PAGEFLAG_FALSE(SwapCache) |
388 | #endif |
389 | |
390 | PAGEFLAG(Unevictable, unevictable, PF_HEAD) |
391 | __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) |
392 | TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) |
393 | |
394 | #ifdef CONFIG_MMU |
395 | PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) |
396 | __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) |
397 | TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) |
398 | #else |
399 | PAGEFLAG_FALSE(Mlocked) __CLEARPAGEFLAG_NOOP(Mlocked) |
400 | TESTSCFLAG_FALSE(Mlocked) |
401 | #endif |
402 | |
403 | #ifdef CONFIG_ARCH_USES_PG_UNCACHED |
404 | PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) |
405 | #else |
406 | PAGEFLAG_FALSE(Uncached) |
407 | #endif |
408 | |
409 | #ifdef CONFIG_MEMORY_FAILURE |
410 | PAGEFLAG(HWPoison, hwpoison, PF_ANY) |
411 | TESTSCFLAG(HWPoison, hwpoison, PF_ANY) |
412 | #define __PG_HWPOISON (1UL << PG_hwpoison) |
413 | extern bool set_hwpoison_free_buddy_page(struct page *page); |
414 | #else |
415 | PAGEFLAG_FALSE(HWPoison) |
416 | static inline bool set_hwpoison_free_buddy_page(struct page *page) |
417 | { |
418 | return 0; |
419 | } |
420 | #define __PG_HWPOISON 0 |
421 | #endif |
422 | |
423 | #if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT) |
424 | TESTPAGEFLAG(Young, young, PF_ANY) |
425 | SETPAGEFLAG(Young, young, PF_ANY) |
426 | TESTCLEARFLAG(Young, young, PF_ANY) |
427 | PAGEFLAG(Idle, idle, PF_ANY) |
428 | #endif |
429 | |
430 | /* |
431 | * On an anonymous page mapped into a user virtual memory area, |
432 | * page->mapping points to its anon_vma, not to a struct address_space; |
433 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. |
434 | * |
435 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, |
436 | * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON |
437 | * bit; and then page->mapping points, not to an anon_vma, but to a private |
438 | * structure which KSM associates with that merged page. See ksm.h. |
439 | * |
440 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable |
441 | * page and then page->mapping points a struct address_space. |
442 | * |
443 | * Please note that, confusingly, "page_mapping" refers to the inode |
444 | * address_space which maps the page from disk; whereas "page_mapped" |
445 | * refers to user virtual address space into which the page is mapped. |
446 | */ |
447 | #define PAGE_MAPPING_ANON 0x1 |
448 | #define PAGE_MAPPING_MOVABLE 0x2 |
449 | #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) |
450 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) |
451 | |
452 | static __always_inline int PageMappingFlags(struct page *page) |
453 | { |
454 | return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; |
455 | } |
456 | |
457 | static __always_inline int PageAnon(struct page *page) |
458 | { |
459 | page = compound_head(page); |
460 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; |
461 | } |
462 | |
463 | static __always_inline int __PageMovable(struct page *page) |
464 | { |
465 | return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == |
466 | PAGE_MAPPING_MOVABLE; |
467 | } |
468 | |
469 | #ifdef CONFIG_KSM |
470 | /* |
471 | * A KSM page is one of those write-protected "shared pages" or "merged pages" |
472 | * which KSM maps into multiple mms, wherever identical anonymous page content |
473 | * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any |
474 | * anon_vma, but to that page's node of the stable tree. |
475 | */ |
476 | static __always_inline int PageKsm(struct page *page) |
477 | { |
478 | page = compound_head(page); |
479 | return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == |
480 | PAGE_MAPPING_KSM; |
481 | } |
482 | #else |
483 | TESTPAGEFLAG_FALSE(Ksm) |
484 | #endif |
485 | |
486 | u64 stable_page_flags(struct page *page); |
487 | |
488 | static inline int PageUptodate(struct page *page) |
489 | { |
490 | int ret; |
491 | page = compound_head(page); |
492 | ret = test_bit(PG_uptodate, &(page)->flags); |
493 | /* |
494 | * Must ensure that the data we read out of the page is loaded |
495 | * _after_ we've loaded page->flags to check for PageUptodate. |
496 | * We can skip the barrier if the page is not uptodate, because |
497 | * we wouldn't be reading anything from it. |
498 | * |
499 | * See SetPageUptodate() for the other side of the story. |
500 | */ |
501 | if (ret) |
502 | smp_rmb(); |
503 | |
504 | return ret; |
505 | } |
506 | |
507 | static __always_inline void __SetPageUptodate(struct page *page) |
508 | { |
509 | VM_BUG_ON_PAGE(PageTail(page), page); |
510 | smp_wmb(); |
511 | __set_bit(PG_uptodate, &page->flags); |
512 | } |
513 | |
514 | static __always_inline void SetPageUptodate(struct page *page) |
515 | { |
516 | VM_BUG_ON_PAGE(PageTail(page), page); |
517 | /* |
518 | * Memory barrier must be issued before setting the PG_uptodate bit, |
519 | * so that all previous stores issued in order to bring the page |
520 | * uptodate are actually visible before PageUptodate becomes true. |
521 | */ |
522 | smp_wmb(); |
523 | set_bit(PG_uptodate, &page->flags); |
524 | } |
525 | |
526 | CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) |
527 | |
528 | int test_clear_page_writeback(struct page *page); |
529 | int __test_set_page_writeback(struct page *page, bool keep_write); |
530 | |
531 | #define test_set_page_writeback(page) \ |
532 | __test_set_page_writeback(page, false) |
533 | #define test_set_page_writeback_keepwrite(page) \ |
534 | __test_set_page_writeback(page, true) |
535 | |
536 | static inline void set_page_writeback(struct page *page) |
537 | { |
538 | test_set_page_writeback(page); |
539 | } |
540 | |
541 | static inline void set_page_writeback_keepwrite(struct page *page) |
542 | { |
543 | test_set_page_writeback_keepwrite(page); |
544 | } |
545 | |
546 | __PAGEFLAG(Head, head, PF_ANY) CLEARPAGEFLAG(Head, head, PF_ANY) |
547 | |
548 | static __always_inline void set_compound_head(struct page *page, struct page *head) |
549 | { |
550 | WRITE_ONCE(page->compound_head, (unsigned long)head + 1); |
551 | } |
552 | |
553 | static __always_inline void clear_compound_head(struct page *page) |
554 | { |
555 | WRITE_ONCE(page->compound_head, 0); |
556 | } |
557 | |
558 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
559 | static inline void ClearPageCompound(struct page *page) |
560 | { |
561 | BUG_ON(!PageHead(page)); |
562 | ClearPageHead(page); |
563 | } |
564 | #endif |
565 | |
566 | #define PG_head_mask ((1UL << PG_head)) |
567 | |
568 | #ifdef CONFIG_HUGETLB_PAGE |
569 | int PageHuge(struct page *page); |
570 | int PageHeadHuge(struct page *page); |
571 | bool page_huge_active(struct page *page); |
572 | #else |
573 | TESTPAGEFLAG_FALSE(Huge) |
574 | TESTPAGEFLAG_FALSE(HeadHuge) |
575 | |
576 | static inline bool page_huge_active(struct page *page) |
577 | { |
578 | return 0; |
579 | } |
580 | #endif |
581 | |
582 | |
583 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
584 | /* |
585 | * PageHuge() only returns true for hugetlbfs pages, but not for |
586 | * normal or transparent huge pages. |
587 | * |
588 | * PageTransHuge() returns true for both transparent huge and |
589 | * hugetlbfs pages, but not normal pages. PageTransHuge() can only be |
590 | * called only in the core VM paths where hugetlbfs pages can't exist. |
591 | */ |
592 | static inline int PageTransHuge(struct page *page) |
593 | { |
594 | VM_BUG_ON_PAGE(PageTail(page), page); |
595 | return PageHead(page); |
596 | } |
597 | |
598 | /* |
599 | * PageTransCompound returns true for both transparent huge pages |
600 | * and hugetlbfs pages, so it should only be called when it's known |
601 | * that hugetlbfs pages aren't involved. |
602 | */ |
603 | static inline int PageTransCompound(struct page *page) |
604 | { |
605 | return PageCompound(page); |
606 | } |
607 | |
608 | /* |
609 | * PageTransCompoundMap is the same as PageTransCompound, but it also |
610 | * guarantees the primary MMU has the entire compound page mapped |
611 | * through pmd_trans_huge, which in turn guarantees the secondary MMUs |
612 | * can also map the entire compound page. This allows the secondary |
613 | * MMUs to call get_user_pages() only once for each compound page and |
614 | * to immediately map the entire compound page with a single secondary |
615 | * MMU fault. If there will be a pmd split later, the secondary MMUs |
616 | * will get an update through the MMU notifier invalidation through |
617 | * split_huge_pmd(). |
618 | * |
619 | * Unlike PageTransCompound, this is safe to be called only while |
620 | * split_huge_pmd() cannot run from under us, like if protected by the |
621 | * MMU notifier, otherwise it may result in page->_mapcount < 0 false |
622 | * positives. |
623 | */ |
624 | static inline int PageTransCompoundMap(struct page *page) |
625 | { |
626 | return PageTransCompound(page) && atomic_read(&page->_mapcount) < 0; |
627 | } |
628 | |
629 | /* |
630 | * PageTransTail returns true for both transparent huge pages |
631 | * and hugetlbfs pages, so it should only be called when it's known |
632 | * that hugetlbfs pages aren't involved. |
633 | */ |
634 | static inline int PageTransTail(struct page *page) |
635 | { |
636 | return PageTail(page); |
637 | } |
638 | |
639 | /* |
640 | * PageDoubleMap indicates that the compound page is mapped with PTEs as well |
641 | * as PMDs. |
642 | * |
643 | * This is required for optimization of rmap operations for THP: we can postpone |
644 | * per small page mapcount accounting (and its overhead from atomic operations) |
645 | * until the first PMD split. |
646 | * |
647 | * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up |
648 | * by one. This reference will go away with last compound_mapcount. |
649 | * |
650 | * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap(). |
651 | */ |
652 | static inline int PageDoubleMap(struct page *page) |
653 | { |
654 | return PageHead(page) && test_bit(PG_double_map, &page[1].flags); |
655 | } |
656 | |
657 | static inline void SetPageDoubleMap(struct page *page) |
658 | { |
659 | VM_BUG_ON_PAGE(!PageHead(page), page); |
660 | set_bit(PG_double_map, &page[1].flags); |
661 | } |
662 | |
663 | static inline void ClearPageDoubleMap(struct page *page) |
664 | { |
665 | VM_BUG_ON_PAGE(!PageHead(page), page); |
666 | clear_bit(PG_double_map, &page[1].flags); |
667 | } |
668 | static inline int TestSetPageDoubleMap(struct page *page) |
669 | { |
670 | VM_BUG_ON_PAGE(!PageHead(page), page); |
671 | return test_and_set_bit(PG_double_map, &page[1].flags); |
672 | } |
673 | |
674 | static inline int TestClearPageDoubleMap(struct page *page) |
675 | { |
676 | VM_BUG_ON_PAGE(!PageHead(page), page); |
677 | return test_and_clear_bit(PG_double_map, &page[1].flags); |
678 | } |
679 | |
680 | #else |
681 | TESTPAGEFLAG_FALSE(TransHuge) |
682 | TESTPAGEFLAG_FALSE(TransCompound) |
683 | TESTPAGEFLAG_FALSE(TransCompoundMap) |
684 | TESTPAGEFLAG_FALSE(TransTail) |
685 | PAGEFLAG_FALSE(DoubleMap) |
686 | TESTSETFLAG_FALSE(DoubleMap) |
687 | TESTCLEARFLAG_FALSE(DoubleMap) |
688 | #endif |
689 | |
690 | /* |
691 | * For pages that are never mapped to userspace (and aren't PageSlab), |
692 | * page_type may be used. Because it is initialised to -1, we invert the |
693 | * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and |
694 | * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and |
695 | * low bits so that an underflow or overflow of page_mapcount() won't be |
696 | * mistaken for a page type value. |
697 | */ |
698 | |
699 | #define PAGE_TYPE_BASE 0xf0000000 |
700 | /* Reserve 0x0000007f to catch underflows of page_mapcount */ |
701 | #define PAGE_MAPCOUNT_RESERVE -128 |
702 | #define PG_buddy 0x00000080 |
703 | #define PG_offline 0x00000100 |
704 | #define PG_kmemcg 0x00000200 |
705 | #define PG_table 0x00000400 |
706 | |
707 | #define PageType(page, flag) \ |
708 | ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) |
709 | |
710 | static inline int page_has_type(struct page *page) |
711 | { |
712 | return (int)page->page_type < PAGE_MAPCOUNT_RESERVE; |
713 | } |
714 | |
715 | #define PAGE_TYPE_OPS(uname, lname) \ |
716 | static __always_inline int Page##uname(struct page *page) \ |
717 | { \ |
718 | return PageType(page, PG_##lname); \ |
719 | } \ |
720 | static __always_inline void __SetPage##uname(struct page *page) \ |
721 | { \ |
722 | VM_BUG_ON_PAGE(!PageType(page, 0), page); \ |
723 | page->page_type &= ~PG_##lname; \ |
724 | } \ |
725 | static __always_inline void __ClearPage##uname(struct page *page) \ |
726 | { \ |
727 | VM_BUG_ON_PAGE(!Page##uname(page), page); \ |
728 | page->page_type |= PG_##lname; \ |
729 | } |
730 | |
731 | /* |
732 | * PageBuddy() indicates that the page is free and in the buddy system |
733 | * (see mm/page_alloc.c). |
734 | */ |
735 | PAGE_TYPE_OPS(Buddy, buddy) |
736 | |
737 | /* |
738 | * PageOffline() indicates that the page is logically offline although the |
739 | * containing section is online. (e.g. inflated in a balloon driver or |
740 | * not onlined when onlining the section). |
741 | * The content of these pages is effectively stale. Such pages should not |
742 | * be touched (read/write/dump/save) except by their owner. |
743 | */ |
744 | PAGE_TYPE_OPS(Offline, offline) |
745 | |
746 | /* |
747 | * If kmemcg is enabled, the buddy allocator will set PageKmemcg() on |
748 | * pages allocated with __GFP_ACCOUNT. It gets cleared on page free. |
749 | */ |
750 | PAGE_TYPE_OPS(Kmemcg, kmemcg) |
751 | |
752 | /* |
753 | * Marks pages in use as page tables. |
754 | */ |
755 | PAGE_TYPE_OPS(Table, table) |
756 | |
757 | extern bool is_free_buddy_page(struct page *page); |
758 | |
759 | __PAGEFLAG(Isolated, isolated, PF_ANY); |
760 | |
761 | /* |
762 | * If network-based swap is enabled, sl*b must keep track of whether pages |
763 | * were allocated from pfmemalloc reserves. |
764 | */ |
765 | static inline int PageSlabPfmemalloc(struct page *page) |
766 | { |
767 | VM_BUG_ON_PAGE(!PageSlab(page), page); |
768 | return PageActive(page); |
769 | } |
770 | |
771 | static inline void SetPageSlabPfmemalloc(struct page *page) |
772 | { |
773 | VM_BUG_ON_PAGE(!PageSlab(page), page); |
774 | SetPageActive(page); |
775 | } |
776 | |
777 | static inline void __ClearPageSlabPfmemalloc(struct page *page) |
778 | { |
779 | VM_BUG_ON_PAGE(!PageSlab(page), page); |
780 | __ClearPageActive(page); |
781 | } |
782 | |
783 | static inline void ClearPageSlabPfmemalloc(struct page *page) |
784 | { |
785 | VM_BUG_ON_PAGE(!PageSlab(page), page); |
786 | ClearPageActive(page); |
787 | } |
788 | |
789 | #ifdef CONFIG_MMU |
790 | #define __PG_MLOCKED (1UL << PG_mlocked) |
791 | #else |
792 | #define __PG_MLOCKED 0 |
793 | #endif |
794 | |
795 | /* |
796 | * Flags checked when a page is freed. Pages being freed should not have |
797 | * these flags set. It they are, there is a problem. |
798 | */ |
799 | #define PAGE_FLAGS_CHECK_AT_FREE \ |
800 | (1UL << PG_lru | 1UL << PG_locked | \ |
801 | 1UL << PG_private | 1UL << PG_private_2 | \ |
802 | 1UL << PG_writeback | 1UL << PG_reserved | \ |
803 | 1UL << PG_slab | 1UL << PG_active | \ |
804 | 1UL << PG_unevictable | __PG_MLOCKED) |
805 | |
806 | /* |
807 | * Flags checked when a page is prepped for return by the page allocator. |
808 | * Pages being prepped should not have these flags set. It they are set, |
809 | * there has been a kernel bug or struct page corruption. |
810 | * |
811 | * __PG_HWPOISON is exceptional because it needs to be kept beyond page's |
812 | * alloc-free cycle to prevent from reusing the page. |
813 | */ |
814 | #define PAGE_FLAGS_CHECK_AT_PREP \ |
815 | (((1UL << NR_PAGEFLAGS) - 1) & ~__PG_HWPOISON) |
816 | |
817 | #define PAGE_FLAGS_PRIVATE \ |
818 | (1UL << PG_private | 1UL << PG_private_2) |
819 | /** |
820 | * page_has_private - Determine if page has private stuff |
821 | * @page: The page to be checked |
822 | * |
823 | * Determine if a page has private stuff, indicating that release routines |
824 | * should be invoked upon it. |
825 | */ |
826 | static inline int page_has_private(struct page *page) |
827 | { |
828 | return !!(page->flags & PAGE_FLAGS_PRIVATE); |
829 | } |
830 | |
831 | #undef PF_ANY |
832 | #undef PF_HEAD |
833 | #undef PF_ONLY_HEAD |
834 | #undef PF_NO_TAIL |
835 | #undef PF_NO_COMPOUND |
836 | #endif /* !__GENERATING_BOUNDS_H */ |
837 | |
838 | #endif /* PAGE_FLAGS_H */ |
839 | |