at v6.4-rc2 1064 lines 35 kB view raw
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 */ 100enum pageflags { 101 PG_locked, /* Page is locked. Don't touch. */ 102 PG_referenced, 103 PG_uptodate, 104 PG_dirty, 105 PG_lru, 106 PG_active, 107 PG_workingset, 108 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */ 109 PG_error, 110 PG_slab, 111 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/ 112 PG_arch_1, 113 PG_reserved, 114 PG_private, /* If pagecache, has fs-private data */ 115 PG_private_2, /* If pagecache, has fs aux data */ 116 PG_writeback, /* Page is under writeback */ 117 PG_head, /* A head page */ 118 PG_mappedtodisk, /* Has blocks allocated on-disk */ 119 PG_reclaim, /* To be reclaimed asap */ 120 PG_swapbacked, /* Page is backed by RAM/swap */ 121 PG_unevictable, /* Page is "unevictable" */ 122#ifdef CONFIG_MMU 123 PG_mlocked, /* Page is vma mlocked */ 124#endif 125#ifdef CONFIG_ARCH_USES_PG_UNCACHED 126 PG_uncached, /* Page has been mapped as uncached */ 127#endif 128#ifdef CONFIG_MEMORY_FAILURE 129 PG_hwpoison, /* hardware poisoned page. Don't touch */ 130#endif 131#if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 132 PG_young, 133 PG_idle, 134#endif 135#ifdef CONFIG_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#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_error, 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#ifdef CONFIG_MEMORY_HOTPLUG 190 /* For self-hosted memmap pages */ 191 PG_vmemmap_self_hosted = PG_owner_priv_1, 192#endif 193}; 194 195#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1) 196 197#ifndef __GENERATING_BOUNDS_H 198 199#ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 200DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key); 201 202/* 203 * Return the real head page struct iff the @page is a fake head page, otherwise 204 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst. 205 */ 206static __always_inline const struct page *page_fixed_fake_head(const struct page *page) 207{ 208 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key)) 209 return page; 210 211 /* 212 * Only addresses aligned with PAGE_SIZE of struct page may be fake head 213 * struct page. The alignment check aims to avoid access the fields ( 214 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly) 215 * cold cacheline in some cases. 216 */ 217 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) && 218 test_bit(PG_head, &page->flags)) { 219 /* 220 * We can safely access the field of the @page[1] with PG_head 221 * because the @page is a compound page composed with at least 222 * two contiguous pages. 223 */ 224 unsigned long head = READ_ONCE(page[1].compound_head); 225 226 if (likely(head & 1)) 227 return (const struct page *)(head - 1); 228 } 229 return page; 230} 231#else 232static inline const struct page *page_fixed_fake_head(const struct page *page) 233{ 234 return page; 235} 236#endif 237 238static __always_inline int page_is_fake_head(struct page *page) 239{ 240 return page_fixed_fake_head(page) != page; 241} 242 243static inline unsigned long _compound_head(const struct page *page) 244{ 245 unsigned long head = READ_ONCE(page->compound_head); 246 247 if (unlikely(head & 1)) 248 return head - 1; 249 return (unsigned long)page_fixed_fake_head(page); 250} 251 252#define compound_head(page) ((typeof(page))_compound_head(page)) 253 254/** 255 * page_folio - Converts from page to folio. 256 * @p: The page. 257 * 258 * Every page is part of a folio. This function cannot be called on a 259 * NULL pointer. 260 * 261 * Context: No reference, nor lock is required on @page. If the caller 262 * does not hold a reference, this call may race with a folio split, so 263 * it should re-check the folio still contains this page after gaining 264 * a reference on the folio. 265 * Return: The folio which contains this page. 266 */ 267#define page_folio(p) (_Generic((p), \ 268 const struct page *: (const struct folio *)_compound_head(p), \ 269 struct page *: (struct folio *)_compound_head(p))) 270 271/** 272 * folio_page - Return a page from a folio. 273 * @folio: The folio. 274 * @n: The page number to return. 275 * 276 * @n is relative to the start of the folio. This function does not 277 * check that the page number lies within @folio; the caller is presumed 278 * to have a reference to the page. 279 */ 280#define folio_page(folio, n) nth_page(&(folio)->page, n) 281 282static __always_inline int PageTail(struct page *page) 283{ 284 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page); 285} 286 287static __always_inline int PageCompound(struct page *page) 288{ 289 return test_bit(PG_head, &page->flags) || 290 READ_ONCE(page->compound_head) & 1; 291} 292 293#define PAGE_POISON_PATTERN -1l 294static inline int PagePoisoned(const struct page *page) 295{ 296 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN; 297} 298 299#ifdef CONFIG_DEBUG_VM 300void page_init_poison(struct page *page, size_t size); 301#else 302static inline void page_init_poison(struct page *page, size_t size) 303{ 304} 305#endif 306 307static unsigned long *folio_flags(struct folio *folio, unsigned n) 308{ 309 struct page *page = &folio->page; 310 311 VM_BUG_ON_PGFLAGS(PageTail(page), page); 312 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); 313 return &page[n].flags; 314} 315 316/* 317 * Page flags policies wrt compound pages 318 * 319 * PF_POISONED_CHECK 320 * check if this struct page poisoned/uninitialized 321 * 322 * PF_ANY: 323 * the page flag is relevant for small, head and tail pages. 324 * 325 * PF_HEAD: 326 * for compound page all operations related to the page flag applied to 327 * head page. 328 * 329 * PF_ONLY_HEAD: 330 * for compound page, callers only ever operate on the head page. 331 * 332 * PF_NO_TAIL: 333 * modifications of the page flag must be done on small or head pages, 334 * checks can be done on tail pages too. 335 * 336 * PF_NO_COMPOUND: 337 * the page flag is not relevant for compound pages. 338 * 339 * PF_SECOND: 340 * the page flag is stored in the first tail page. 341 */ 342#define PF_POISONED_CHECK(page) ({ \ 343 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ 344 page; }) 345#define PF_ANY(page, enforce) PF_POISONED_CHECK(page) 346#define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) 347#define PF_ONLY_HEAD(page, enforce) ({ \ 348 VM_BUG_ON_PGFLAGS(PageTail(page), page); \ 349 PF_POISONED_CHECK(page); }) 350#define PF_NO_TAIL(page, enforce) ({ \ 351 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ 352 PF_POISONED_CHECK(compound_head(page)); }) 353#define PF_NO_COMPOUND(page, enforce) ({ \ 354 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ 355 PF_POISONED_CHECK(page); }) 356#define PF_SECOND(page, enforce) ({ \ 357 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \ 358 PF_POISONED_CHECK(&page[1]); }) 359 360/* Which page is the flag stored in */ 361#define FOLIO_PF_ANY 0 362#define FOLIO_PF_HEAD 0 363#define FOLIO_PF_ONLY_HEAD 0 364#define FOLIO_PF_NO_TAIL 0 365#define FOLIO_PF_NO_COMPOUND 0 366#define FOLIO_PF_SECOND 1 367 368/* 369 * Macros to create function definitions for page flags 370 */ 371#define TESTPAGEFLAG(uname, lname, policy) \ 372static __always_inline bool folio_test_##lname(struct folio *folio) \ 373{ return test_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 374static __always_inline int Page##uname(struct page *page) \ 375{ return test_bit(PG_##lname, &policy(page, 0)->flags); } 376 377#define SETPAGEFLAG(uname, lname, policy) \ 378static __always_inline \ 379void folio_set_##lname(struct folio *folio) \ 380{ set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 381static __always_inline void SetPage##uname(struct page *page) \ 382{ set_bit(PG_##lname, &policy(page, 1)->flags); } 383 384#define CLEARPAGEFLAG(uname, lname, policy) \ 385static __always_inline \ 386void folio_clear_##lname(struct folio *folio) \ 387{ clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 388static __always_inline void ClearPage##uname(struct page *page) \ 389{ clear_bit(PG_##lname, &policy(page, 1)->flags); } 390 391#define __SETPAGEFLAG(uname, lname, policy) \ 392static __always_inline \ 393void __folio_set_##lname(struct folio *folio) \ 394{ __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 395static __always_inline void __SetPage##uname(struct page *page) \ 396{ __set_bit(PG_##lname, &policy(page, 1)->flags); } 397 398#define __CLEARPAGEFLAG(uname, lname, policy) \ 399static __always_inline \ 400void __folio_clear_##lname(struct folio *folio) \ 401{ __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 402static __always_inline void __ClearPage##uname(struct page *page) \ 403{ __clear_bit(PG_##lname, &policy(page, 1)->flags); } 404 405#define TESTSETFLAG(uname, lname, policy) \ 406static __always_inline \ 407bool folio_test_set_##lname(struct folio *folio) \ 408{ return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 409static __always_inline int TestSetPage##uname(struct page *page) \ 410{ return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } 411 412#define TESTCLEARFLAG(uname, lname, policy) \ 413static __always_inline \ 414bool folio_test_clear_##lname(struct folio *folio) \ 415{ return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 416static __always_inline int TestClearPage##uname(struct page *page) \ 417{ return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } 418 419#define PAGEFLAG(uname, lname, policy) \ 420 TESTPAGEFLAG(uname, lname, policy) \ 421 SETPAGEFLAG(uname, lname, policy) \ 422 CLEARPAGEFLAG(uname, lname, policy) 423 424#define __PAGEFLAG(uname, lname, policy) \ 425 TESTPAGEFLAG(uname, lname, policy) \ 426 __SETPAGEFLAG(uname, lname, policy) \ 427 __CLEARPAGEFLAG(uname, lname, policy) 428 429#define TESTSCFLAG(uname, lname, policy) \ 430 TESTSETFLAG(uname, lname, policy) \ 431 TESTCLEARFLAG(uname, lname, policy) 432 433#define TESTPAGEFLAG_FALSE(uname, lname) \ 434static inline bool folio_test_##lname(const struct folio *folio) { return false; } \ 435static inline int Page##uname(const struct page *page) { return 0; } 436 437#define SETPAGEFLAG_NOOP(uname, lname) \ 438static inline void folio_set_##lname(struct folio *folio) { } \ 439static inline void SetPage##uname(struct page *page) { } 440 441#define CLEARPAGEFLAG_NOOP(uname, lname) \ 442static inline void folio_clear_##lname(struct folio *folio) { } \ 443static inline void ClearPage##uname(struct page *page) { } 444 445#define __CLEARPAGEFLAG_NOOP(uname, lname) \ 446static inline void __folio_clear_##lname(struct folio *folio) { } \ 447static inline void __ClearPage##uname(struct page *page) { } 448 449#define TESTSETFLAG_FALSE(uname, lname) \ 450static inline bool folio_test_set_##lname(struct folio *folio) \ 451{ return 0; } \ 452static inline int TestSetPage##uname(struct page *page) { return 0; } 453 454#define TESTCLEARFLAG_FALSE(uname, lname) \ 455static inline bool folio_test_clear_##lname(struct folio *folio) \ 456{ return 0; } \ 457static inline int TestClearPage##uname(struct page *page) { return 0; } 458 459#define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \ 460 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname) 461 462#define TESTSCFLAG_FALSE(uname, lname) \ 463 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname) 464 465__PAGEFLAG(Locked, locked, PF_NO_TAIL) 466PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) 467PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL) 468PAGEFLAG(Referenced, referenced, PF_HEAD) 469 TESTCLEARFLAG(Referenced, referenced, PF_HEAD) 470 __SETPAGEFLAG(Referenced, referenced, PF_HEAD) 471PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) 472 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) 473PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) 474 TESTCLEARFLAG(LRU, lru, PF_HEAD) 475PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) 476 TESTCLEARFLAG(Active, active, PF_HEAD) 477PAGEFLAG(Workingset, workingset, PF_HEAD) 478 TESTCLEARFLAG(Workingset, workingset, PF_HEAD) 479__PAGEFLAG(Slab, slab, PF_NO_TAIL) 480PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ 481 482/* Xen */ 483PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) 484 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) 485PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); 486PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); 487PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 488 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 489 490PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 491 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 492 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 493PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 494 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 495 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 496 497/* 498 * Private page markings that may be used by the filesystem that owns the page 499 * for its own purposes. 500 * - PG_private and PG_private_2 cause release_folio() and co to be invoked 501 */ 502PAGEFLAG(Private, private, PF_ANY) 503PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) 504PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 505 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 506 507/* 508 * Only test-and-set exist for PG_writeback. The unconditional operators are 509 * risky: they bypass page accounting. 510 */ 511TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) 512 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) 513PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) 514 515/* PG_readahead is only used for reads; PG_reclaim is only for writes */ 516PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) 517 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) 518PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND) 519 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND) 520 521#ifdef CONFIG_HIGHMEM 522/* 523 * Must use a macro here due to header dependency issues. page_zone() is not 524 * available at this point. 525 */ 526#define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) 527#define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f)) 528#else 529PAGEFLAG_FALSE(HighMem, highmem) 530#endif 531 532#ifdef CONFIG_SWAP 533static __always_inline bool folio_test_swapcache(struct folio *folio) 534{ 535 return folio_test_swapbacked(folio) && 536 test_bit(PG_swapcache, folio_flags(folio, 0)); 537} 538 539static __always_inline bool PageSwapCache(struct page *page) 540{ 541 return folio_test_swapcache(page_folio(page)); 542} 543 544SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 545CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 546#else 547PAGEFLAG_FALSE(SwapCache, swapcache) 548#endif 549 550PAGEFLAG(Unevictable, unevictable, PF_HEAD) 551 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) 552 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) 553 554#ifdef CONFIG_MMU 555PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 556 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 557 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) 558#else 559PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked) 560 TESTSCFLAG_FALSE(Mlocked, mlocked) 561#endif 562 563#ifdef CONFIG_ARCH_USES_PG_UNCACHED 564PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) 565#else 566PAGEFLAG_FALSE(Uncached, uncached) 567#endif 568 569#ifdef CONFIG_MEMORY_FAILURE 570PAGEFLAG(HWPoison, hwpoison, PF_ANY) 571TESTSCFLAG(HWPoison, hwpoison, PF_ANY) 572#define __PG_HWPOISON (1UL << PG_hwpoison) 573#define MAGIC_HWPOISON 0x48575053U /* HWPS */ 574extern void SetPageHWPoisonTakenOff(struct page *page); 575extern void ClearPageHWPoisonTakenOff(struct page *page); 576extern bool take_page_off_buddy(struct page *page); 577extern bool put_page_back_buddy(struct page *page); 578#else 579PAGEFLAG_FALSE(HWPoison, hwpoison) 580#define __PG_HWPOISON 0 581#endif 582 583#if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 584TESTPAGEFLAG(Young, young, PF_ANY) 585SETPAGEFLAG(Young, young, PF_ANY) 586TESTCLEARFLAG(Young, young, PF_ANY) 587PAGEFLAG(Idle, idle, PF_ANY) 588#endif 589 590/* 591 * PageReported() is used to track reported free pages within the Buddy 592 * allocator. We can use the non-atomic version of the test and set 593 * operations as both should be shielded with the zone lock to prevent 594 * any possible races on the setting or clearing of the bit. 595 */ 596__PAGEFLAG(Reported, reported, PF_NO_COMPOUND) 597 598#ifdef CONFIG_MEMORY_HOTPLUG 599PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY) 600#else 601PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted) 602#endif 603 604/* 605 * On an anonymous page mapped into a user virtual memory area, 606 * page->mapping points to its anon_vma, not to a struct address_space; 607 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 608 * 609 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 610 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON 611 * bit; and then page->mapping points, not to an anon_vma, but to a private 612 * structure which KSM associates with that merged page. See ksm.h. 613 * 614 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable 615 * page and then page->mapping points to a struct movable_operations. 616 * 617 * Please note that, confusingly, "page_mapping" refers to the inode 618 * address_space which maps the page from disk; whereas "page_mapped" 619 * refers to user virtual address space into which the page is mapped. 620 */ 621#define PAGE_MAPPING_ANON 0x1 622#define PAGE_MAPPING_MOVABLE 0x2 623#define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 624#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 625 626/* 627 * Different with flags above, this flag is used only for fsdax mode. It 628 * indicates that this page->mapping is now under reflink case. 629 */ 630#define PAGE_MAPPING_DAX_SHARED ((void *)0x1) 631 632static __always_inline bool folio_mapping_flags(struct folio *folio) 633{ 634 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0; 635} 636 637static __always_inline int PageMappingFlags(struct page *page) 638{ 639 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; 640} 641 642static __always_inline bool folio_test_anon(struct folio *folio) 643{ 644 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0; 645} 646 647static __always_inline bool PageAnon(struct page *page) 648{ 649 return folio_test_anon(page_folio(page)); 650} 651 652static __always_inline bool __folio_test_movable(const struct folio *folio) 653{ 654 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 655 PAGE_MAPPING_MOVABLE; 656} 657 658static __always_inline int __PageMovable(struct page *page) 659{ 660 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 661 PAGE_MAPPING_MOVABLE; 662} 663 664#ifdef CONFIG_KSM 665/* 666 * A KSM page is one of those write-protected "shared pages" or "merged pages" 667 * which KSM maps into multiple mms, wherever identical anonymous page content 668 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any 669 * anon_vma, but to that page's node of the stable tree. 670 */ 671static __always_inline bool folio_test_ksm(struct folio *folio) 672{ 673 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 674 PAGE_MAPPING_KSM; 675} 676 677static __always_inline bool PageKsm(struct page *page) 678{ 679 return folio_test_ksm(page_folio(page)); 680} 681#else 682TESTPAGEFLAG_FALSE(Ksm, ksm) 683#endif 684 685u64 stable_page_flags(struct page *page); 686 687/** 688 * folio_test_uptodate - Is this folio up to date? 689 * @folio: The folio. 690 * 691 * The uptodate flag is set on a folio when every byte in the folio is 692 * at least as new as the corresponding bytes on storage. Anonymous 693 * and CoW folios are always uptodate. If the folio is not uptodate, 694 * some of the bytes in it may be; see the is_partially_uptodate() 695 * address_space operation. 696 */ 697static inline bool folio_test_uptodate(struct folio *folio) 698{ 699 bool ret = test_bit(PG_uptodate, folio_flags(folio, 0)); 700 /* 701 * Must ensure that the data we read out of the folio is loaded 702 * _after_ we've loaded folio->flags to check the uptodate bit. 703 * We can skip the barrier if the folio is not uptodate, because 704 * we wouldn't be reading anything from it. 705 * 706 * See folio_mark_uptodate() for the other side of the story. 707 */ 708 if (ret) 709 smp_rmb(); 710 711 return ret; 712} 713 714static inline int PageUptodate(struct page *page) 715{ 716 return folio_test_uptodate(page_folio(page)); 717} 718 719static __always_inline void __folio_mark_uptodate(struct folio *folio) 720{ 721 smp_wmb(); 722 __set_bit(PG_uptodate, folio_flags(folio, 0)); 723} 724 725static __always_inline void folio_mark_uptodate(struct folio *folio) 726{ 727 /* 728 * Memory barrier must be issued before setting the PG_uptodate bit, 729 * so that all previous stores issued in order to bring the folio 730 * uptodate are actually visible before folio_test_uptodate becomes true. 731 */ 732 smp_wmb(); 733 set_bit(PG_uptodate, folio_flags(folio, 0)); 734} 735 736static __always_inline void __SetPageUptodate(struct page *page) 737{ 738 __folio_mark_uptodate((struct folio *)page); 739} 740 741static __always_inline void SetPageUptodate(struct page *page) 742{ 743 folio_mark_uptodate((struct folio *)page); 744} 745 746CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) 747 748bool __folio_start_writeback(struct folio *folio, bool keep_write); 749bool set_page_writeback(struct page *page); 750 751#define folio_start_writeback(folio) \ 752 __folio_start_writeback(folio, false) 753#define folio_start_writeback_keepwrite(folio) \ 754 __folio_start_writeback(folio, true) 755 756static inline bool test_set_page_writeback(struct page *page) 757{ 758 return set_page_writeback(page); 759} 760 761static __always_inline bool folio_test_head(struct folio *folio) 762{ 763 return test_bit(PG_head, folio_flags(folio, FOLIO_PF_ANY)); 764} 765 766static __always_inline int PageHead(struct page *page) 767{ 768 PF_POISONED_CHECK(page); 769 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page); 770} 771 772__SETPAGEFLAG(Head, head, PF_ANY) 773__CLEARPAGEFLAG(Head, head, PF_ANY) 774CLEARPAGEFLAG(Head, head, PF_ANY) 775 776/** 777 * folio_test_large() - Does this folio contain more than one page? 778 * @folio: The folio to test. 779 * 780 * Return: True if the folio is larger than one page. 781 */ 782static inline bool folio_test_large(struct folio *folio) 783{ 784 return folio_test_head(folio); 785} 786 787static __always_inline void set_compound_head(struct page *page, struct page *head) 788{ 789 WRITE_ONCE(page->compound_head, (unsigned long)head + 1); 790} 791 792static __always_inline void clear_compound_head(struct page *page) 793{ 794 WRITE_ONCE(page->compound_head, 0); 795} 796 797#ifdef CONFIG_TRANSPARENT_HUGEPAGE 798static inline void ClearPageCompound(struct page *page) 799{ 800 BUG_ON(!PageHead(page)); 801 ClearPageHead(page); 802} 803#endif 804 805#define PG_head_mask ((1UL << PG_head)) 806 807#ifdef CONFIG_HUGETLB_PAGE 808int PageHuge(struct page *page); 809bool folio_test_hugetlb(struct folio *folio); 810#else 811TESTPAGEFLAG_FALSE(Huge, hugetlb) 812#endif 813 814#ifdef CONFIG_TRANSPARENT_HUGEPAGE 815/* 816 * PageHuge() only returns true for hugetlbfs pages, but not for 817 * normal or transparent huge pages. 818 * 819 * PageTransHuge() returns true for both transparent huge and 820 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be 821 * called only in the core VM paths where hugetlbfs pages can't exist. 822 */ 823static inline int PageTransHuge(struct page *page) 824{ 825 VM_BUG_ON_PAGE(PageTail(page), page); 826 return PageHead(page); 827} 828 829static inline bool folio_test_transhuge(struct folio *folio) 830{ 831 return folio_test_head(folio); 832} 833 834/* 835 * PageTransCompound returns true for both transparent huge pages 836 * and hugetlbfs pages, so it should only be called when it's known 837 * that hugetlbfs pages aren't involved. 838 */ 839static inline int PageTransCompound(struct page *page) 840{ 841 return PageCompound(page); 842} 843 844/* 845 * PageTransTail returns true for both transparent huge pages 846 * and hugetlbfs pages, so it should only be called when it's known 847 * that hugetlbfs pages aren't involved. 848 */ 849static inline int PageTransTail(struct page *page) 850{ 851 return PageTail(page); 852} 853#else 854TESTPAGEFLAG_FALSE(TransHuge, transhuge) 855TESTPAGEFLAG_FALSE(TransCompound, transcompound) 856TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap) 857TESTPAGEFLAG_FALSE(TransTail, transtail) 858#endif 859 860#if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE) 861/* 862 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the 863 * compound page. 864 * 865 * This flag is set by hwpoison handler. Cleared by THP split or free page. 866 */ 867PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 868 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 869#else 870PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 871 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 872#endif 873 874/* 875 * Check if a page is currently marked HWPoisoned. Note that this check is 876 * best effort only and inherently racy: there is no way to synchronize with 877 * failing hardware. 878 */ 879static inline bool is_page_hwpoison(struct page *page) 880{ 881 if (PageHWPoison(page)) 882 return true; 883 return PageHuge(page) && PageHWPoison(compound_head(page)); 884} 885 886/* 887 * For pages that are never mapped to userspace (and aren't PageSlab), 888 * page_type may be used. Because it is initialised to -1, we invert the 889 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and 890 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and 891 * low bits so that an underflow or overflow of page_mapcount() won't be 892 * mistaken for a page type value. 893 */ 894 895#define PAGE_TYPE_BASE 0xf0000000 896/* Reserve 0x0000007f to catch underflows of page_mapcount */ 897#define PAGE_MAPCOUNT_RESERVE -128 898#define PG_buddy 0x00000080 899#define PG_offline 0x00000100 900#define PG_table 0x00000200 901#define PG_guard 0x00000400 902 903#define PageType(page, flag) \ 904 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) 905 906static inline int page_type_has_type(unsigned int page_type) 907{ 908 return (int)page_type < PAGE_MAPCOUNT_RESERVE; 909} 910 911static inline int page_has_type(struct page *page) 912{ 913 return page_type_has_type(page->page_type); 914} 915 916#define PAGE_TYPE_OPS(uname, lname) \ 917static __always_inline int Page##uname(struct page *page) \ 918{ \ 919 return PageType(page, PG_##lname); \ 920} \ 921static __always_inline void __SetPage##uname(struct page *page) \ 922{ \ 923 VM_BUG_ON_PAGE(!PageType(page, 0), page); \ 924 page->page_type &= ~PG_##lname; \ 925} \ 926static __always_inline void __ClearPage##uname(struct page *page) \ 927{ \ 928 VM_BUG_ON_PAGE(!Page##uname(page), page); \ 929 page->page_type |= PG_##lname; \ 930} 931 932/* 933 * PageBuddy() indicates that the page is free and in the buddy system 934 * (see mm/page_alloc.c). 935 */ 936PAGE_TYPE_OPS(Buddy, buddy) 937 938/* 939 * PageOffline() indicates that the page is logically offline although the 940 * containing section is online. (e.g. inflated in a balloon driver or 941 * not onlined when onlining the section). 942 * The content of these pages is effectively stale. Such pages should not 943 * be touched (read/write/dump/save) except by their owner. 944 * 945 * If a driver wants to allow to offline unmovable PageOffline() pages without 946 * putting them back to the buddy, it can do so via the memory notifier by 947 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the 948 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline() 949 * pages (now with a reference count of zero) are treated like free pages, 950 * allowing the containing memory block to get offlined. A driver that 951 * relies on this feature is aware that re-onlining the memory block will 952 * require to re-set the pages PageOffline() and not giving them to the 953 * buddy via online_page_callback_t. 954 * 955 * There are drivers that mark a page PageOffline() and expect there won't be 956 * any further access to page content. PFN walkers that read content of random 957 * pages should check PageOffline() and synchronize with such drivers using 958 * page_offline_freeze()/page_offline_thaw(). 959 */ 960PAGE_TYPE_OPS(Offline, offline) 961 962extern void page_offline_freeze(void); 963extern void page_offline_thaw(void); 964extern void page_offline_begin(void); 965extern void page_offline_end(void); 966 967/* 968 * Marks pages in use as page tables. 969 */ 970PAGE_TYPE_OPS(Table, table) 971 972/* 973 * Marks guardpages used with debug_pagealloc. 974 */ 975PAGE_TYPE_OPS(Guard, guard) 976 977extern bool is_free_buddy_page(struct page *page); 978 979PAGEFLAG(Isolated, isolated, PF_ANY); 980 981static __always_inline int PageAnonExclusive(struct page *page) 982{ 983 VM_BUG_ON_PGFLAGS(!PageAnon(page), page); 984 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 985 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 986} 987 988static __always_inline void SetPageAnonExclusive(struct page *page) 989{ 990 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page); 991 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 992 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 993} 994 995static __always_inline void ClearPageAnonExclusive(struct page *page) 996{ 997 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page); 998 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 999 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1000} 1001 1002static __always_inline void __ClearPageAnonExclusive(struct page *page) 1003{ 1004 VM_BUG_ON_PGFLAGS(!PageAnon(page), page); 1005 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1006 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1007} 1008 1009#ifdef CONFIG_MMU 1010#define __PG_MLOCKED (1UL << PG_mlocked) 1011#else 1012#define __PG_MLOCKED 0 1013#endif 1014 1015/* 1016 * Flags checked when a page is freed. Pages being freed should not have 1017 * these flags set. If they are, there is a problem. 1018 */ 1019#define PAGE_FLAGS_CHECK_AT_FREE \ 1020 (1UL << PG_lru | 1UL << PG_locked | \ 1021 1UL << PG_private | 1UL << PG_private_2 | \ 1022 1UL << PG_writeback | 1UL << PG_reserved | \ 1023 1UL << PG_slab | 1UL << PG_active | \ 1024 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK) 1025 1026/* 1027 * Flags checked when a page is prepped for return by the page allocator. 1028 * Pages being prepped should not have these flags set. If they are set, 1029 * there has been a kernel bug or struct page corruption. 1030 * 1031 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's 1032 * alloc-free cycle to prevent from reusing the page. 1033 */ 1034#define PAGE_FLAGS_CHECK_AT_PREP \ 1035 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK) 1036 1037#define PAGE_FLAGS_PRIVATE \ 1038 (1UL << PG_private | 1UL << PG_private_2) 1039/** 1040 * page_has_private - Determine if page has private stuff 1041 * @page: The page to be checked 1042 * 1043 * Determine if a page has private stuff, indicating that release routines 1044 * should be invoked upon it. 1045 */ 1046static inline int page_has_private(struct page *page) 1047{ 1048 return !!(page->flags & PAGE_FLAGS_PRIVATE); 1049} 1050 1051static inline bool folio_has_private(struct folio *folio) 1052{ 1053 return page_has_private(&folio->page); 1054} 1055 1056#undef PF_ANY 1057#undef PF_HEAD 1058#undef PF_ONLY_HEAD 1059#undef PF_NO_TAIL 1060#undef PF_NO_COMPOUND 1061#undef PF_SECOND 1062#endif /* !__GENERATING_BOUNDS_H */ 1063 1064#endif /* PAGE_FLAGS_H */