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_uptodate tells whether the page's contents is valid. When a read
72 * completes, the page becomes uptodate, unless a disk I/O error happened.
73 *
74 * PG_referenced, PG_reclaim are used for page reclaim for anonymous and
75 * file-backed pagecache (see mm/vmscan.c).
76 *
77 * PG_error is set to indicate that an I/O error occurred on this page.
78 *
79 * PG_arch_1 is an architecture specific page state bit. The generic code
80 * guarantees that this bit is cleared for a page when it first is entered into
81 * the page cache.
82 *
83 * PG_hwpoison indicates that a page got corrupted in hardware and contains
84 * data with incorrect ECC bits that triggered a machine check. Accessing is
85 * not safe since it may cause another machine check. Don't touch!
86 */
87
88 /*
89 * Don't use the pageflags directly. Use the PageFoo macros.
90 *
91 * The page flags field is split into two parts, the main flags area
92 * which extends from the low bits upwards, and the fields area which
93 * extends from the high bits downwards.
94 *
95 * | FIELD | ... | FLAGS |
96 * N-1 ^ 0
97 * (NR_PAGEFLAGS)
98 *
99 * The fields area is reserved for fields mapping zone, node (for NUMA) and
100 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like
101 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP).
102 */
103 enum pageflags {
104 PG_locked, /* Page is locked. Don't touch. */
105 PG_referenced,
106 PG_uptodate,
107 PG_dirty,
108 PG_lru,
109 PG_active,
110 PG_workingset,
111 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
112 PG_error,
113 PG_slab,
114 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
115 PG_arch_1,
116 PG_reserved,
117 PG_private, /* If pagecache, has fs-private data */
118 PG_private_2, /* If pagecache, has fs aux data */
119 PG_writeback, /* Page is under writeback */
120 PG_head, /* A head page */
121 PG_mappedtodisk, /* Has blocks allocated on-disk */
122 PG_reclaim, /* To be reclaimed asap */
123 PG_swapbacked, /* Page is backed by RAM/swap */
124 PG_unevictable, /* Page is "unevictable" */
125 #ifdef CONFIG_MMU
126 PG_mlocked, /* Page is vma mlocked */
127 #endif
128 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
129 PG_uncached, /* Page has been mapped as uncached */
130 #endif
131 #ifdef CONFIG_MEMORY_FAILURE
132 PG_hwpoison, /* hardware poisoned page. Don't touch */
133 #endif
134 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
135 PG_young,
136 PG_idle,
137 #endif
138 #ifdef CONFIG_64BIT
139 PG_arch_2,
140 #endif
141 #ifdef CONFIG_KASAN_HW_TAGS
142 PG_skip_kasan_poison,
143 #endif
144 __NR_PAGEFLAGS,
145
146 /* Filesystems */
147 PG_checked = PG_owner_priv_1,
148
149 /* SwapBacked */
150 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
151
152 /* Two page bits are conscripted by FS-Cache to maintain local caching
153 * state. These bits are set on pages belonging to the netfs's inodes
154 * when those inodes are being locally cached.
155 */
156 PG_fscache = PG_private_2, /* page backed by cache */
157
158 /* XEN */
159 /* Pinned in Xen as a read-only pagetable page. */
160 PG_pinned = PG_owner_priv_1,
161 /* Pinned as part of domain save (see xen_mm_pin_all()). */
162 PG_savepinned = PG_dirty,
163 /* Has a grant mapping of another (foreign) domain's page. */
164 PG_foreign = PG_owner_priv_1,
165 /* Remapped by swiotlb-xen. */
166 PG_xen_remapped = PG_owner_priv_1,
167
168 /* SLOB */
169 PG_slob_free = PG_private,
170
171 /* Compound pages. Stored in first tail page's flags */
172 PG_double_map = PG_workingset,
173
174 #ifdef CONFIG_MEMORY_FAILURE
175 /*
176 * Compound pages. Stored in first tail page's flags.
177 * Indicates that at least one subpage is hwpoisoned in the
178 * THP.
179 */
180 PG_has_hwpoisoned = PG_mappedtodisk,
181 #endif
182
183 /* non-lru isolated movable page */
184 PG_isolated = PG_reclaim,
185
186 /* Only valid for buddy pages. Used to track pages that are reported */
187 PG_reported = PG_uptodate,
188 };
189
190 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
191
192 #ifndef __GENERATING_BOUNDS_H
193
_compound_head(const struct page * page)194 static inline unsigned long _compound_head(const struct page *page)
195 {
196 unsigned long head = READ_ONCE(page->compound_head);
197
198 if (unlikely(head & 1))
199 return head - 1;
200 return (unsigned long)page;
201 }
202
203 #define compound_head(page) ((typeof(page))_compound_head(page))
204
PageTail(struct page * page)205 static __always_inline int PageTail(struct page *page)
206 {
207 return READ_ONCE(page->compound_head) & 1;
208 }
209
PageCompound(struct page * page)210 static __always_inline int PageCompound(struct page *page)
211 {
212 return test_bit(PG_head, &page->flags) || PageTail(page);
213 }
214
215 #define PAGE_POISON_PATTERN -1l
PagePoisoned(const struct page * page)216 static inline int PagePoisoned(const struct page *page)
217 {
218 return page->flags == PAGE_POISON_PATTERN;
219 }
220
221 #ifdef CONFIG_DEBUG_VM
222 void page_init_poison(struct page *page, size_t size);
223 #else
page_init_poison(struct page * page,size_t size)224 static inline void page_init_poison(struct page *page, size_t size)
225 {
226 }
227 #endif
228
229 /*
230 * Page flags policies wrt compound pages
231 *
232 * PF_POISONED_CHECK
233 * check if this struct page poisoned/uninitialized
234 *
235 * PF_ANY:
236 * the page flag is relevant for small, head and tail pages.
237 *
238 * PF_HEAD:
239 * for compound page all operations related to the page flag applied to
240 * head page.
241 *
242 * PF_ONLY_HEAD:
243 * for compound page, callers only ever operate on the head page.
244 *
245 * PF_NO_TAIL:
246 * modifications of the page flag must be done on small or head pages,
247 * checks can be done on tail pages too.
248 *
249 * PF_NO_COMPOUND:
250 * the page flag is not relevant for compound pages.
251 *
252 * PF_SECOND:
253 * the page flag is stored in the first tail page.
254 */
255 #define PF_POISONED_CHECK(page) ({ \
256 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
257 page; })
258 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
259 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
260 #define PF_ONLY_HEAD(page, enforce) ({ \
261 VM_BUG_ON_PGFLAGS(PageTail(page), page); \
262 PF_POISONED_CHECK(page); })
263 #define PF_NO_TAIL(page, enforce) ({ \
264 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
265 PF_POISONED_CHECK(compound_head(page)); })
266 #define PF_NO_COMPOUND(page, enforce) ({ \
267 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
268 PF_POISONED_CHECK(page); })
269 #define PF_SECOND(page, enforce) ({ \
270 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \
271 PF_POISONED_CHECK(&page[1]); })
272
273 /*
274 * Macros to create function definitions for page flags
275 */
276 #define TESTPAGEFLAG(uname, lname, policy) \
277 static __always_inline int Page##uname(struct page *page) \
278 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
279
280 #define SETPAGEFLAG(uname, lname, policy) \
281 static __always_inline void SetPage##uname(struct page *page) \
282 { set_bit(PG_##lname, &policy(page, 1)->flags); }
283
284 #define CLEARPAGEFLAG(uname, lname, policy) \
285 static __always_inline void ClearPage##uname(struct page *page) \
286 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
287
288 #define __SETPAGEFLAG(uname, lname, policy) \
289 static __always_inline void __SetPage##uname(struct page *page) \
290 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
291
292 #define __CLEARPAGEFLAG(uname, lname, policy) \
293 static __always_inline void __ClearPage##uname(struct page *page) \
294 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
295
296 #define TESTSETFLAG(uname, lname, policy) \
297 static __always_inline int TestSetPage##uname(struct page *page) \
298 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
299
300 #define TESTCLEARFLAG(uname, lname, policy) \
301 static __always_inline int TestClearPage##uname(struct page *page) \
302 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
303
304 #define PAGEFLAG(uname, lname, policy) \
305 TESTPAGEFLAG(uname, lname, policy) \
306 SETPAGEFLAG(uname, lname, policy) \
307 CLEARPAGEFLAG(uname, lname, policy)
308
309 #define __PAGEFLAG(uname, lname, policy) \
310 TESTPAGEFLAG(uname, lname, policy) \
311 __SETPAGEFLAG(uname, lname, policy) \
312 __CLEARPAGEFLAG(uname, lname, policy)
313
314 #define TESTSCFLAG(uname, lname, policy) \
315 TESTSETFLAG(uname, lname, policy) \
316 TESTCLEARFLAG(uname, lname, policy)
317
318 #define TESTPAGEFLAG_FALSE(uname) \
319 static inline int Page##uname(const struct page *page) { return 0; }
320
321 #define SETPAGEFLAG_NOOP(uname) \
322 static inline void SetPage##uname(struct page *page) { }
323
324 #define CLEARPAGEFLAG_NOOP(uname) \
325 static inline void ClearPage##uname(struct page *page) { }
326
327 #define __CLEARPAGEFLAG_NOOP(uname) \
328 static inline void __ClearPage##uname(struct page *page) { }
329
330 #define TESTSETFLAG_FALSE(uname) \
331 static inline int TestSetPage##uname(struct page *page) { return 0; }
332
333 #define TESTCLEARFLAG_FALSE(uname) \
334 static inline int TestClearPage##uname(struct page *page) { return 0; }
335
336 #define PAGEFLAG_FALSE(uname) TESTPAGEFLAG_FALSE(uname) \
337 SETPAGEFLAG_NOOP(uname) CLEARPAGEFLAG_NOOP(uname)
338
339 #define TESTSCFLAG_FALSE(uname) \
340 TESTSETFLAG_FALSE(uname) TESTCLEARFLAG_FALSE(uname)
341
342 __PAGEFLAG(Locked, locked, PF_NO_TAIL)
343 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) __CLEARPAGEFLAG(Waiters, waiters, PF_ONLY_HEAD)
344 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
345 PAGEFLAG(Referenced, referenced, PF_HEAD)
346 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
347 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
348 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
349 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
350 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
351 TESTCLEARFLAG(LRU, lru, PF_HEAD)
352 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
353 TESTCLEARFLAG(Active, active, PF_HEAD)
354 PAGEFLAG(Workingset, workingset, PF_HEAD)
355 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
356 __PAGEFLAG(Slab, slab, PF_NO_TAIL)
357 __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL)
358 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
359
360 /* Xen */
361 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
362 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
363 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
364 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
PAGEFLAG(XenRemapped,xen_remapped,PF_NO_COMPOUND)365 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
366 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
367
368 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
369 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
370 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
371 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
372 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
373 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
374
375 /*
376 * Private page markings that may be used by the filesystem that owns the page
377 * for its own purposes.
378 * - PG_private and PG_private_2 cause releasepage() and co to be invoked
379 */
380 PAGEFLAG(Private, private, PF_ANY)
381 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
382 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
383 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
384
385 /*
386 * Only test-and-set exist for PG_writeback. The unconditional operators are
387 * risky: they bypass page accounting.
388 */
389 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
390 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
391 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
392
393 /* PG_readahead is only used for reads; PG_reclaim is only for writes */
394 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
395 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
396 PAGEFLAG(Readahead, reclaim, PF_NO_COMPOUND)
397 TESTCLEARFLAG(Readahead, reclaim, PF_NO_COMPOUND)
398
399 #ifdef CONFIG_HIGHMEM
400 /*
401 * Must use a macro here due to header dependency issues. page_zone() is not
402 * available at this point.
403 */
404 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
405 #else
406 PAGEFLAG_FALSE(HighMem)
407 #endif
408
409 #ifdef CONFIG_SWAP
410 static __always_inline int PageSwapCache(struct page *page)
411 {
412 #ifdef CONFIG_THP_SWAP
413 page = compound_head(page);
414 #endif
415 return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags);
416
417 }
418 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
419 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
420 #else
421 PAGEFLAG_FALSE(SwapCache)
422 #endif
423
424 PAGEFLAG(Unevictable, unevictable, PF_HEAD)
425 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
426 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
427
428 #ifdef CONFIG_MMU
429 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
430 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
431 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
432 #else
433 PAGEFLAG_FALSE(Mlocked) __CLEARPAGEFLAG_NOOP(Mlocked)
434 TESTSCFLAG_FALSE(Mlocked)
435 #endif
436
437 #ifdef CONFIG_ARCH_USES_PG_UNCACHED
438 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
439 #else
440 PAGEFLAG_FALSE(Uncached)
441 #endif
442
443 #ifdef CONFIG_MEMORY_FAILURE
444 PAGEFLAG(HWPoison, hwpoison, PF_ANY)
445 TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
446 #define __PG_HWPOISON (1UL << PG_hwpoison)
447 extern bool take_page_off_buddy(struct page *page);
448 #else
449 PAGEFLAG_FALSE(HWPoison)
450 #define __PG_HWPOISON 0
451 #endif
452
453 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT)
TESTPAGEFLAG(Young,young,PF_ANY)454 TESTPAGEFLAG(Young, young, PF_ANY)
455 SETPAGEFLAG(Young, young, PF_ANY)
456 TESTCLEARFLAG(Young, young, PF_ANY)
457 PAGEFLAG(Idle, idle, PF_ANY)
458 #endif
459
460 #ifdef CONFIG_KASAN_HW_TAGS
461 PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD)
462 #else
463 PAGEFLAG_FALSE(SkipKASanPoison)
464 #endif
465
466 /*
467 * PageReported() is used to track reported free pages within the Buddy
468 * allocator. We can use the non-atomic version of the test and set
469 * operations as both should be shielded with the zone lock to prevent
470 * any possible races on the setting or clearing of the bit.
471 */
472 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
473
474 /*
475 * On an anonymous page mapped into a user virtual memory area,
476 * page->mapping points to its anon_vma, not to a struct address_space;
477 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
478 *
479 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
480 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
481 * bit; and then page->mapping points, not to an anon_vma, but to a private
482 * structure which KSM associates with that merged page. See ksm.h.
483 *
484 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
485 * page and then page->mapping points a struct address_space.
486 *
487 * Please note that, confusingly, "page_mapping" refers to the inode
488 * address_space which maps the page from disk; whereas "page_mapped"
489 * refers to user virtual address space into which the page is mapped.
490 */
491 #define PAGE_MAPPING_ANON 0x1
492 #define PAGE_MAPPING_MOVABLE 0x2
493 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
494 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
495
496 static __always_inline int PageMappingFlags(struct page *page)
497 {
498 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
499 }
500
PageAnon(struct page * page)501 static __always_inline int PageAnon(struct page *page)
502 {
503 page = compound_head(page);
504 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
505 }
506
__PageMovable(struct page * page)507 static __always_inline int __PageMovable(struct page *page)
508 {
509 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
510 PAGE_MAPPING_MOVABLE;
511 }
512
513 #ifdef CONFIG_KSM
514 /*
515 * A KSM page is one of those write-protected "shared pages" or "merged pages"
516 * which KSM maps into multiple mms, wherever identical anonymous page content
517 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
518 * anon_vma, but to that page's node of the stable tree.
519 */
PageKsm(struct page * page)520 static __always_inline int PageKsm(struct page *page)
521 {
522 page = compound_head(page);
523 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
524 PAGE_MAPPING_KSM;
525 }
526 #else
527 TESTPAGEFLAG_FALSE(Ksm)
528 #endif
529
530 u64 stable_page_flags(struct page *page);
531
PageUptodate(struct page * page)532 static inline int PageUptodate(struct page *page)
533 {
534 int ret;
535 page = compound_head(page);
536 ret = test_bit(PG_uptodate, &(page)->flags);
537 /*
538 * Must ensure that the data we read out of the page is loaded
539 * _after_ we've loaded page->flags to check for PageUptodate.
540 * We can skip the barrier if the page is not uptodate, because
541 * we wouldn't be reading anything from it.
542 *
543 * See SetPageUptodate() for the other side of the story.
544 */
545 if (ret)
546 smp_rmb();
547
548 return ret;
549 }
550
__SetPageUptodate(struct page * page)551 static __always_inline void __SetPageUptodate(struct page *page)
552 {
553 VM_BUG_ON_PAGE(PageTail(page), page);
554 smp_wmb();
555 __set_bit(PG_uptodate, &page->flags);
556 }
557
SetPageUptodate(struct page * page)558 static __always_inline void SetPageUptodate(struct page *page)
559 {
560 VM_BUG_ON_PAGE(PageTail(page), page);
561 /*
562 * Memory barrier must be issued before setting the PG_uptodate bit,
563 * so that all previous stores issued in order to bring the page
564 * uptodate are actually visible before PageUptodate becomes true.
565 */
566 smp_wmb();
567 set_bit(PG_uptodate, &page->flags);
568 }
569
570 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
571
572 int test_clear_page_writeback(struct page *page);
573 int __test_set_page_writeback(struct page *page, bool keep_write);
574
575 #define test_set_page_writeback(page) \
576 __test_set_page_writeback(page, false)
577 #define test_set_page_writeback_keepwrite(page) \
578 __test_set_page_writeback(page, true)
579
set_page_writeback(struct page * page)580 static inline void set_page_writeback(struct page *page)
581 {
582 test_set_page_writeback(page);
583 }
584
set_page_writeback_keepwrite(struct page * page)585 static inline void set_page_writeback_keepwrite(struct page *page)
586 {
587 test_set_page_writeback_keepwrite(page);
588 }
589
__PAGEFLAG(Head,head,PF_ANY)590 __PAGEFLAG(Head, head, PF_ANY) CLEARPAGEFLAG(Head, head, PF_ANY)
591
592 static __always_inline void set_compound_head(struct page *page, struct page *head)
593 {
594 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
595 }
596
clear_compound_head(struct page * page)597 static __always_inline void clear_compound_head(struct page *page)
598 {
599 WRITE_ONCE(page->compound_head, 0);
600 }
601
602 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
ClearPageCompound(struct page * page)603 static inline void ClearPageCompound(struct page *page)
604 {
605 BUG_ON(!PageHead(page));
606 ClearPageHead(page);
607 }
608 #endif
609
610 #define PG_head_mask ((1UL << PG_head))
611
612 #ifdef CONFIG_HUGETLB_PAGE
613 int PageHuge(struct page *page);
614 int PageHeadHuge(struct page *page);
615 #else
616 TESTPAGEFLAG_FALSE(Huge)
TESTPAGEFLAG_FALSE(HeadHuge)617 TESTPAGEFLAG_FALSE(HeadHuge)
618 #endif
619
620
621 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
622 /*
623 * PageHuge() only returns true for hugetlbfs pages, but not for
624 * normal or transparent huge pages.
625 *
626 * PageTransHuge() returns true for both transparent huge and
627 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
628 * called only in the core VM paths where hugetlbfs pages can't exist.
629 */
630 static inline int PageTransHuge(struct page *page)
631 {
632 VM_BUG_ON_PAGE(PageTail(page), page);
633 return PageHead(page);
634 }
635
636 /*
637 * PageTransCompound returns true for both transparent huge pages
638 * and hugetlbfs pages, so it should only be called when it's known
639 * that hugetlbfs pages aren't involved.
640 */
PageTransCompound(struct page * page)641 static inline int PageTransCompound(struct page *page)
642 {
643 return PageCompound(page);
644 }
645
646 /*
647 * PageTransTail returns true for both transparent huge pages
648 * and hugetlbfs pages, so it should only be called when it's known
649 * that hugetlbfs pages aren't involved.
650 */
PageTransTail(struct page * page)651 static inline int PageTransTail(struct page *page)
652 {
653 return PageTail(page);
654 }
655
656 /*
657 * PageDoubleMap indicates that the compound page is mapped with PTEs as well
658 * as PMDs.
659 *
660 * This is required for optimization of rmap operations for THP: we can postpone
661 * per small page mapcount accounting (and its overhead from atomic operations)
662 * until the first PMD split.
663 *
664 * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up
665 * by one. This reference will go away with last compound_mapcount.
666 *
667 * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap().
668 */
PAGEFLAG(DoubleMap,double_map,PF_SECOND)669 PAGEFLAG(DoubleMap, double_map, PF_SECOND)
670 TESTSCFLAG(DoubleMap, double_map, PF_SECOND)
671 #else
672 TESTPAGEFLAG_FALSE(TransHuge)
673 TESTPAGEFLAG_FALSE(TransCompound)
674 TESTPAGEFLAG_FALSE(TransCompoundMap)
675 TESTPAGEFLAG_FALSE(TransTail)
676 PAGEFLAG_FALSE(DoubleMap)
677 TESTSCFLAG_FALSE(DoubleMap)
678 #endif
679
680 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
681 /*
682 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the
683 * compound page.
684 *
685 * This flag is set by hwpoison handler. Cleared by THP split or free page.
686 */
687 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
688 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND)
689 #else
690 PAGEFLAG_FALSE(HasHWPoisoned)
691 TESTSCFLAG_FALSE(HasHWPoisoned)
692 #endif
693
694 /*
695 * Check if a page is currently marked HWPoisoned. Note that this check is
696 * best effort only and inherently racy: there is no way to synchronize with
697 * failing hardware.
698 */
699 static inline bool is_page_hwpoison(struct page *page)
700 {
701 if (PageHWPoison(page))
702 return true;
703 return PageHuge(page) && PageHWPoison(compound_head(page));
704 }
705
706 /*
707 * For pages that are never mapped to userspace (and aren't PageSlab),
708 * page_type may be used. Because it is initialised to -1, we invert the
709 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
710 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
711 * low bits so that an underflow or overflow of page_mapcount() won't be
712 * mistaken for a page type value.
713 */
714
715 #define PAGE_TYPE_BASE 0xf0000000
716 /* Reserve 0x0000007f to catch underflows of page_mapcount */
717 #define PAGE_MAPCOUNT_RESERVE -128
718 #define PG_buddy 0x00000080
719 #define PG_offline 0x00000100
720 #define PG_table 0x00000200
721 #define PG_guard 0x00000400
722
723 #define PageType(page, flag) \
724 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
725
page_has_type(struct page * page)726 static inline int page_has_type(struct page *page)
727 {
728 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE;
729 }
730
731 #define PAGE_TYPE_OPS(uname, lname) \
732 static __always_inline int Page##uname(struct page *page) \
733 { \
734 return PageType(page, PG_##lname); \
735 } \
736 static __always_inline void __SetPage##uname(struct page *page) \
737 { \
738 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
739 page->page_type &= ~PG_##lname; \
740 } \
741 static __always_inline void __ClearPage##uname(struct page *page) \
742 { \
743 VM_BUG_ON_PAGE(!Page##uname(page), page); \
744 page->page_type |= PG_##lname; \
745 }
746
747 /*
748 * PageBuddy() indicates that the page is free and in the buddy system
749 * (see mm/page_alloc.c).
750 */
751 PAGE_TYPE_OPS(Buddy, buddy)
752
753 /*
754 * PageOffline() indicates that the page is logically offline although the
755 * containing section is online. (e.g. inflated in a balloon driver or
756 * not onlined when onlining the section).
757 * The content of these pages is effectively stale. Such pages should not
758 * be touched (read/write/dump/save) except by their owner.
759 *
760 * If a driver wants to allow to offline unmovable PageOffline() pages without
761 * putting them back to the buddy, it can do so via the memory notifier by
762 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
763 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
764 * pages (now with a reference count of zero) are treated like free pages,
765 * allowing the containing memory block to get offlined. A driver that
766 * relies on this feature is aware that re-onlining the memory block will
767 * require to re-set the pages PageOffline() and not giving them to the
768 * buddy via online_page_callback_t.
769 *
770 * There are drivers that mark a page PageOffline() and expect there won't be
771 * any further access to page content. PFN walkers that read content of random
772 * pages should check PageOffline() and synchronize with such drivers using
773 * page_offline_freeze()/page_offline_thaw().
774 */
775 PAGE_TYPE_OPS(Offline, offline)
776
777 extern void page_offline_freeze(void);
778 extern void page_offline_thaw(void);
779 extern void page_offline_begin(void);
780 extern void page_offline_end(void);
781
782 /*
783 * Marks pages in use as page tables.
784 */
785 PAGE_TYPE_OPS(Table, table)
786
787 /*
788 * Marks guardpages used with debug_pagealloc.
789 */
790 PAGE_TYPE_OPS(Guard, guard)
791
792 extern bool is_free_buddy_page(struct page *page);
793
794 __PAGEFLAG(Isolated, isolated, PF_ANY);
795
796 /*
797 * If network-based swap is enabled, sl*b must keep track of whether pages
798 * were allocated from pfmemalloc reserves.
799 */
PageSlabPfmemalloc(struct page * page)800 static inline int PageSlabPfmemalloc(struct page *page)
801 {
802 VM_BUG_ON_PAGE(!PageSlab(page), page);
803 return PageActive(page);
804 }
805
806 /*
807 * A version of PageSlabPfmemalloc() for opportunistic checks where the page
808 * might have been freed under us and not be a PageSlab anymore.
809 */
__PageSlabPfmemalloc(struct page * page)810 static inline int __PageSlabPfmemalloc(struct page *page)
811 {
812 return PageActive(page);
813 }
814
SetPageSlabPfmemalloc(struct page * page)815 static inline void SetPageSlabPfmemalloc(struct page *page)
816 {
817 VM_BUG_ON_PAGE(!PageSlab(page), page);
818 SetPageActive(page);
819 }
820
__ClearPageSlabPfmemalloc(struct page * page)821 static inline void __ClearPageSlabPfmemalloc(struct page *page)
822 {
823 VM_BUG_ON_PAGE(!PageSlab(page), page);
824 __ClearPageActive(page);
825 }
826
ClearPageSlabPfmemalloc(struct page * page)827 static inline void ClearPageSlabPfmemalloc(struct page *page)
828 {
829 VM_BUG_ON_PAGE(!PageSlab(page), page);
830 ClearPageActive(page);
831 }
832
833 #ifdef CONFIG_MMU
834 #define __PG_MLOCKED (1UL << PG_mlocked)
835 #else
836 #define __PG_MLOCKED 0
837 #endif
838
839 /*
840 * Flags checked when a page is freed. Pages being freed should not have
841 * these flags set. If they are, there is a problem.
842 */
843 #define PAGE_FLAGS_CHECK_AT_FREE \
844 (1UL << PG_lru | 1UL << PG_locked | \
845 1UL << PG_private | 1UL << PG_private_2 | \
846 1UL << PG_writeback | 1UL << PG_reserved | \
847 1UL << PG_slab | 1UL << PG_active | \
848 1UL << PG_unevictable | __PG_MLOCKED)
849
850 /*
851 * Flags checked when a page is prepped for return by the page allocator.
852 * Pages being prepped should not have these flags set. If they are set,
853 * there has been a kernel bug or struct page corruption.
854 *
855 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
856 * alloc-free cycle to prevent from reusing the page.
857 */
858 #define PAGE_FLAGS_CHECK_AT_PREP \
859 (PAGEFLAGS_MASK & ~__PG_HWPOISON)
860
861 #define PAGE_FLAGS_PRIVATE \
862 (1UL << PG_private | 1UL << PG_private_2)
863 /**
864 * page_has_private - Determine if page has private stuff
865 * @page: The page to be checked
866 *
867 * Determine if a page has private stuff, indicating that release routines
868 * should be invoked upon it.
869 */
page_has_private(struct page * page)870 static inline int page_has_private(struct page *page)
871 {
872 return !!(page->flags & PAGE_FLAGS_PRIVATE);
873 }
874
875 #undef PF_ANY
876 #undef PF_HEAD
877 #undef PF_ONLY_HEAD
878 #undef PF_NO_TAIL
879 #undef PF_NO_COMPOUND
880 #undef PF_SECOND
881 #endif /* !__GENERATING_BOUNDS_H */
882
883 #endif /* PAGE_FLAGS_H */
884
