1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Macros for manipulating and testing page->flags 4 */ 5 6 #ifndef PAGE_FLAGS_H 7 #define PAGE_FLAGS_H 8 9 #include <linux/types.h> 10 #include <linux/bug.h> 11 #include <linux/mmdebug.h> 12 #ifndef __GENERATING_BOUNDS_H 13 #include <linux/mm_types.h> 14 #include <generated/bounds.h> 15 #endif /* !__GENERATING_BOUNDS_H */ 16 17 /* 18 * Various page->flags bits: 19 * 20 * PG_reserved is set for special pages. The "struct page" of such a page 21 * should in general not be touched (e.g. set dirty) except by its owner. 22 * Pages marked as PG_reserved include: 23 * - Pages part of the kernel image (including vDSO) and similar (e.g. BIOS, 24 * initrd, HW tables) 25 * - Pages reserved or allocated early during boot (before the page allocator 26 * was initialized). This includes (depending on the architecture) the 27 * initial vmemmap, initial page tables, crashkernel, elfcorehdr, and much 28 * much more. Once (if ever) freed, PG_reserved is cleared and they will 29 * be given to the page allocator. 30 * - Pages falling into physical memory gaps - not IORESOURCE_SYSRAM. Trying 31 * to read/write these pages might end badly. Don't touch! 32 * - The zero page(s) 33 * - Pages not added to the page allocator when onlining a section because 34 * they were excluded via the online_page_callback() or because they are 35 * PG_hwpoison. 36 * - Pages allocated in the context of kexec/kdump (loaded kernel image, 37 * control pages, vmcoreinfo) 38 * - MMIO/DMA pages. Some architectures don't allow to ioremap pages that are 39 * not marked PG_reserved (as they might be in use by somebody else who does 40 * not respect the caching strategy). 41 * - Pages part of an offline section (struct pages of offline sections should 42 * not be trusted as they will be initialized when first onlined). 43 * - MCA pages on ia64 44 * - Pages holding CPU notes for POWER Firmware Assisted Dump 45 * - Device memory (e.g. PMEM, DAX, HMM) 46 * Some PG_reserved pages will be excluded from the hibernation image. 47 * PG_reserved does in general not hinder anybody from dumping or swapping 48 * and is no longer required for remap_pfn_range(). ioremap might require it. 49 * Consequently, PG_reserved for a page mapped into user space can indicate 50 * the zero page, the vDSO, MMIO pages or device memory. 51 * 52 * The PG_private bitflag is set on pagecache pages if they contain filesystem 53 * specific data (which is normally at page->private). It can be used by 54 * private allocations for its own usage. 55 * 56 * During initiation of disk I/O, PG_locked is set. This bit is set before I/O 57 * and cleared when writeback _starts_ or when read _completes_. PG_writeback 58 * is set before writeback starts and cleared when it finishes. 59 * 60 * PG_locked also pins a page in pagecache, and blocks truncation of the file 61 * while it is held. 62 * 63 * page_waitqueue(page) is a wait queue of all tasks waiting for the page 64 * to become unlocked. 65 * 66 * PG_swapbacked is set when a page uses swap as a backing storage. This are 67 * usually PageAnon or shmem pages but please note that even anonymous pages 68 * might lose their PG_swapbacked flag when they simply can be dropped (e.g. as 69 * a result of MADV_FREE). 70 * 71 * PG_referenced, PG_reclaim are used for page reclaim for anonymous and 72 * file-backed pagecache (see mm/vmscan.c). 73 * 74 * PG_error is set to indicate that an I/O error occurred on this page. 75 * 76 * PG_arch_1 is an architecture specific page state bit. The generic code 77 * guarantees that this bit is cleared for a page when it first is entered into 78 * the page cache. 79 * 80 * PG_hwpoison indicates that a page got corrupted in hardware and contains 81 * data with incorrect ECC bits that triggered a machine check. Accessing is 82 * not safe since it may cause another machine check. Don't touch! 83 */ 84 85 /* 86 * Don't use the pageflags directly. Use the PageFoo macros. 87 * 88 * The page flags field is split into two parts, the main flags area 89 * which extends from the low bits upwards, and the fields area which 90 * extends from the high bits downwards. 91 * 92 * | FIELD | ... | FLAGS | 93 * N-1 ^ 0 94 * (NR_PAGEFLAGS) 95 * 96 * The fields area is reserved for fields mapping zone, node (for NUMA) and 97 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like 98 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP). 99 */ 100 enum pageflags { 101 PG_locked, /* Page is locked. Don't touch. */ 102 PG_referenced, 103 PG_uptodate, 104 PG_dirty, 105 PG_lru, 106 PG_active, 107 PG_workingset, 108 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */ 109 PG_error, 110 PG_slab, 111 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/ 112 PG_arch_1, 113 PG_reserved, 114 PG_private, /* If pagecache, has fs-private data */ 115 PG_private_2, /* If pagecache, has fs aux data */ 116 PG_writeback, /* Page is under writeback */ 117 PG_head, /* A head page */ 118 PG_mappedtodisk, /* Has blocks allocated on-disk */ 119 PG_reclaim, /* To be reclaimed asap */ 120 PG_swapbacked, /* Page is backed by RAM/swap */ 121 PG_unevictable, /* Page is "unevictable" */ 122 #ifdef CONFIG_MMU 123 PG_mlocked, /* Page is vma mlocked */ 124 #endif 125 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 126 PG_uncached, /* Page has been mapped as uncached */ 127 #endif 128 #ifdef CONFIG_MEMORY_FAILURE 129 PG_hwpoison, /* hardware poisoned page. Don't touch */ 130 #endif 131 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 132 PG_young, 133 PG_idle, 134 #endif 135 #ifdef CONFIG_64BIT 136 PG_arch_2, 137 #endif 138 #ifdef CONFIG_KASAN_HW_TAGS 139 PG_skip_kasan_poison, 140 #endif 141 __NR_PAGEFLAGS, 142 143 PG_readahead = PG_reclaim, 144 145 /* 146 * Depending on the way an anonymous folio can be mapped into a page 147 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped 148 * THP), PG_anon_exclusive may be set only for the head page or for 149 * tail pages of an anonymous folio. For now, we only expect it to be 150 * set on tail pages for PTE-mapped THP. 151 */ 152 PG_anon_exclusive = PG_mappedtodisk, 153 154 /* Filesystems */ 155 PG_checked = PG_owner_priv_1, 156 157 /* SwapBacked */ 158 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */ 159 160 /* Two page bits are conscripted by FS-Cache to maintain local caching 161 * state. These bits are set on pages belonging to the netfs's inodes 162 * when those inodes are being locally cached. 163 */ 164 PG_fscache = PG_private_2, /* page backed by cache */ 165 166 /* XEN */ 167 /* Pinned in Xen as a read-only pagetable page. */ 168 PG_pinned = PG_owner_priv_1, 169 /* Pinned as part of domain save (see xen_mm_pin_all()). */ 170 PG_savepinned = PG_dirty, 171 /* Has a grant mapping of another (foreign) domain's page. */ 172 PG_foreign = PG_owner_priv_1, 173 /* Remapped by swiotlb-xen. */ 174 PG_xen_remapped = PG_owner_priv_1, 175 176 /* SLOB */ 177 PG_slob_free = PG_private, 178 179 #ifdef CONFIG_MEMORY_FAILURE 180 /* 181 * Compound pages. Stored in first tail page's flags. 182 * Indicates that at least one subpage is hwpoisoned in the 183 * THP. 184 */ 185 PG_has_hwpoisoned = PG_error, 186 #endif 187 188 /* non-lru isolated movable page */ 189 PG_isolated = PG_reclaim, 190 191 /* Only valid for buddy pages. Used to track pages that are reported */ 192 PG_reported = PG_uptodate, 193 194 #ifdef CONFIG_MEMORY_HOTPLUG 195 /* For self-hosted memmap pages */ 196 PG_vmemmap_self_hosted = PG_owner_priv_1, 197 #endif 198 }; 199 200 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1) 201 202 #ifndef __GENERATING_BOUNDS_H 203 204 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 205 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key); 206 207 /* 208 * Return the real head page struct iff the @page is a fake head page, otherwise 209 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst. 210 */ 211 static __always_inline const struct page *page_fixed_fake_head(const struct page *page) 212 { 213 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key)) 214 return page; 215 216 /* 217 * Only addresses aligned with PAGE_SIZE of struct page may be fake head 218 * struct page. The alignment check aims to avoid access the fields ( 219 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly) 220 * cold cacheline in some cases. 221 */ 222 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) && 223 test_bit(PG_head, &page->flags)) { 224 /* 225 * We can safely access the field of the @page[1] with PG_head 226 * because the @page is a compound page composed with at least 227 * two contiguous pages. 228 */ 229 unsigned long head = READ_ONCE(page[1].compound_head); 230 231 if (likely(head & 1)) 232 return (const struct page *)(head - 1); 233 } 234 return page; 235 } 236 #else 237 static inline const struct page *page_fixed_fake_head(const struct page *page) 238 { 239 return page; 240 } 241 #endif 242 243 static __always_inline int page_is_fake_head(struct page *page) 244 { 245 return page_fixed_fake_head(page) != page; 246 } 247 248 static inline unsigned long _compound_head(const struct page *page) 249 { 250 unsigned long head = READ_ONCE(page->compound_head); 251 252 if (unlikely(head & 1)) 253 return head - 1; 254 return (unsigned long)page_fixed_fake_head(page); 255 } 256 257 #define compound_head(page) ((typeof(page))_compound_head(page)) 258 259 /** 260 * page_folio - Converts from page to folio. 261 * @p: The page. 262 * 263 * Every page is part of a folio. This function cannot be called on a 264 * NULL pointer. 265 * 266 * Context: No reference, nor lock is required on @page. If the caller 267 * does not hold a reference, this call may race with a folio split, so 268 * it should re-check the folio still contains this page after gaining 269 * a reference on the folio. 270 * Return: The folio which contains this page. 271 */ 272 #define page_folio(p) (_Generic((p), \ 273 const struct page *: (const struct folio *)_compound_head(p), \ 274 struct page *: (struct folio *)_compound_head(p))) 275 276 /** 277 * folio_page - Return a page from a folio. 278 * @folio: The folio. 279 * @n: The page number to return. 280 * 281 * @n is relative to the start of the folio. This function does not 282 * check that the page number lies within @folio; the caller is presumed 283 * to have a reference to the page. 284 */ 285 #define folio_page(folio, n) nth_page(&(folio)->page, n) 286 287 static __always_inline int PageTail(struct page *page) 288 { 289 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page); 290 } 291 292 static __always_inline int PageCompound(struct page *page) 293 { 294 return test_bit(PG_head, &page->flags) || 295 READ_ONCE(page->compound_head) & 1; 296 } 297 298 #define PAGE_POISON_PATTERN -1l 299 static inline int PagePoisoned(const struct page *page) 300 { 301 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN; 302 } 303 304 #ifdef CONFIG_DEBUG_VM 305 void page_init_poison(struct page *page, size_t size); 306 #else 307 static inline void page_init_poison(struct page *page, size_t size) 308 { 309 } 310 #endif 311 312 static unsigned long *folio_flags(struct folio *folio, unsigned n) 313 { 314 struct page *page = &folio->page; 315 316 VM_BUG_ON_PGFLAGS(PageTail(page), page); 317 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); 318 return &page[n].flags; 319 } 320 321 /* 322 * Page flags policies wrt compound pages 323 * 324 * PF_POISONED_CHECK 325 * check if this struct page poisoned/uninitialized 326 * 327 * PF_ANY: 328 * the page flag is relevant for small, head and tail pages. 329 * 330 * PF_HEAD: 331 * for compound page all operations related to the page flag applied to 332 * head page. 333 * 334 * PF_ONLY_HEAD: 335 * for compound page, callers only ever operate on the head page. 336 * 337 * PF_NO_TAIL: 338 * modifications of the page flag must be done on small or head pages, 339 * checks can be done on tail pages too. 340 * 341 * PF_NO_COMPOUND: 342 * the page flag is not relevant for compound pages. 343 * 344 * PF_SECOND: 345 * the page flag is stored in the first tail page. 346 */ 347 #define PF_POISONED_CHECK(page) ({ \ 348 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ 349 page; }) 350 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page) 351 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) 352 #define PF_ONLY_HEAD(page, enforce) ({ \ 353 VM_BUG_ON_PGFLAGS(PageTail(page), page); \ 354 PF_POISONED_CHECK(page); }) 355 #define PF_NO_TAIL(page, enforce) ({ \ 356 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ 357 PF_POISONED_CHECK(compound_head(page)); }) 358 #define PF_NO_COMPOUND(page, enforce) ({ \ 359 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ 360 PF_POISONED_CHECK(page); }) 361 #define PF_SECOND(page, enforce) ({ \ 362 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \ 363 PF_POISONED_CHECK(&page[1]); }) 364 365 /* Which page is the flag stored in */ 366 #define FOLIO_PF_ANY 0 367 #define FOLIO_PF_HEAD 0 368 #define FOLIO_PF_ONLY_HEAD 0 369 #define FOLIO_PF_NO_TAIL 0 370 #define FOLIO_PF_NO_COMPOUND 0 371 #define FOLIO_PF_SECOND 1 372 373 /* 374 * Macros to create function definitions for page flags 375 */ 376 #define TESTPAGEFLAG(uname, lname, policy) \ 377 static __always_inline bool folio_test_##lname(struct folio *folio) \ 378 { return test_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 379 static __always_inline int Page##uname(struct page *page) \ 380 { return test_bit(PG_##lname, &policy(page, 0)->flags); } 381 382 #define SETPAGEFLAG(uname, lname, policy) \ 383 static __always_inline \ 384 void folio_set_##lname(struct folio *folio) \ 385 { set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 386 static __always_inline void SetPage##uname(struct page *page) \ 387 { set_bit(PG_##lname, &policy(page, 1)->flags); } 388 389 #define CLEARPAGEFLAG(uname, lname, policy) \ 390 static __always_inline \ 391 void folio_clear_##lname(struct folio *folio) \ 392 { clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 393 static __always_inline void ClearPage##uname(struct page *page) \ 394 { clear_bit(PG_##lname, &policy(page, 1)->flags); } 395 396 #define __SETPAGEFLAG(uname, lname, policy) \ 397 static __always_inline \ 398 void __folio_set_##lname(struct folio *folio) \ 399 { __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 400 static __always_inline void __SetPage##uname(struct page *page) \ 401 { __set_bit(PG_##lname, &policy(page, 1)->flags); } 402 403 #define __CLEARPAGEFLAG(uname, lname, policy) \ 404 static __always_inline \ 405 void __folio_clear_##lname(struct folio *folio) \ 406 { __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 407 static __always_inline void __ClearPage##uname(struct page *page) \ 408 { __clear_bit(PG_##lname, &policy(page, 1)->flags); } 409 410 #define TESTSETFLAG(uname, lname, policy) \ 411 static __always_inline \ 412 bool folio_test_set_##lname(struct folio *folio) \ 413 { return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 414 static __always_inline int TestSetPage##uname(struct page *page) \ 415 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } 416 417 #define TESTCLEARFLAG(uname, lname, policy) \ 418 static __always_inline \ 419 bool folio_test_clear_##lname(struct folio *folio) \ 420 { return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 421 static __always_inline int TestClearPage##uname(struct page *page) \ 422 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } 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 __PAGEFLAG(uname, lname, policy) \ 430 TESTPAGEFLAG(uname, lname, policy) \ 431 __SETPAGEFLAG(uname, lname, policy) \ 432 __CLEARPAGEFLAG(uname, lname, policy) 433 434 #define TESTSCFLAG(uname, lname, policy) \ 435 TESTSETFLAG(uname, lname, policy) \ 436 TESTCLEARFLAG(uname, lname, policy) 437 438 #define TESTPAGEFLAG_FALSE(uname, lname) \ 439 static inline bool folio_test_##lname(const struct folio *folio) { return false; } \ 440 static inline int Page##uname(const struct page *page) { return 0; } 441 442 #define SETPAGEFLAG_NOOP(uname, lname) \ 443 static inline void folio_set_##lname(struct folio *folio) { } \ 444 static inline void SetPage##uname(struct page *page) { } 445 446 #define CLEARPAGEFLAG_NOOP(uname, lname) \ 447 static inline void folio_clear_##lname(struct folio *folio) { } \ 448 static inline void ClearPage##uname(struct page *page) { } 449 450 #define __CLEARPAGEFLAG_NOOP(uname, lname) \ 451 static inline void __folio_clear_##lname(struct folio *folio) { } \ 452 static inline void __ClearPage##uname(struct page *page) { } 453 454 #define TESTSETFLAG_FALSE(uname, lname) \ 455 static inline bool folio_test_set_##lname(struct folio *folio) \ 456 { return 0; } \ 457 static inline int TestSetPage##uname(struct page *page) { return 0; } 458 459 #define TESTCLEARFLAG_FALSE(uname, lname) \ 460 static inline bool folio_test_clear_##lname(struct folio *folio) \ 461 { return 0; } \ 462 static inline int TestClearPage##uname(struct page *page) { return 0; } 463 464 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \ 465 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname) 466 467 #define TESTSCFLAG_FALSE(uname, lname) \ 468 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname) 469 470 __PAGEFLAG(Locked, locked, PF_NO_TAIL) 471 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) 472 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL) 473 PAGEFLAG(Referenced, referenced, PF_HEAD) 474 TESTCLEARFLAG(Referenced, referenced, PF_HEAD) 475 __SETPAGEFLAG(Referenced, referenced, PF_HEAD) 476 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) 477 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) 478 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) 479 TESTCLEARFLAG(LRU, lru, PF_HEAD) 480 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) 481 TESTCLEARFLAG(Active, active, PF_HEAD) 482 PAGEFLAG(Workingset, workingset, PF_HEAD) 483 TESTCLEARFLAG(Workingset, workingset, PF_HEAD) 484 __PAGEFLAG(Slab, slab, PF_NO_TAIL) 485 __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL) 486 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ 487 488 /* Xen */ 489 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) 490 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) 491 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); 492 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); 493 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 494 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 495 496 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 497 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 498 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 499 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 500 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 501 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 502 503 /* 504 * Private page markings that may be used by the filesystem that owns the page 505 * for its own purposes. 506 * - PG_private and PG_private_2 cause release_folio() and co to be invoked 507 */ 508 PAGEFLAG(Private, private, PF_ANY) 509 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) 510 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 511 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 512 513 /* 514 * Only test-and-set exist for PG_writeback. The unconditional operators are 515 * risky: they bypass page accounting. 516 */ 517 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) 518 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) 519 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) 520 521 /* PG_readahead is only used for reads; PG_reclaim is only for writes */ 522 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) 523 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) 524 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND) 525 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND) 526 527 #ifdef CONFIG_HIGHMEM 528 /* 529 * Must use a macro here due to header dependency issues. page_zone() is not 530 * available at this point. 531 */ 532 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) 533 #else 534 PAGEFLAG_FALSE(HighMem, highmem) 535 #endif 536 537 #ifdef CONFIG_SWAP 538 static __always_inline bool folio_test_swapcache(struct folio *folio) 539 { 540 return folio_test_swapbacked(folio) && 541 test_bit(PG_swapcache, folio_flags(folio, 0)); 542 } 543 544 static __always_inline bool PageSwapCache(struct page *page) 545 { 546 return folio_test_swapcache(page_folio(page)); 547 } 548 549 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 550 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 551 #else 552 PAGEFLAG_FALSE(SwapCache, swapcache) 553 #endif 554 555 PAGEFLAG(Unevictable, unevictable, PF_HEAD) 556 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) 557 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) 558 559 #ifdef CONFIG_MMU 560 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 561 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 562 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) 563 #else 564 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked) 565 TESTSCFLAG_FALSE(Mlocked, mlocked) 566 #endif 567 568 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 569 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) 570 #else 571 PAGEFLAG_FALSE(Uncached, uncached) 572 #endif 573 574 #ifdef CONFIG_MEMORY_FAILURE 575 PAGEFLAG(HWPoison, hwpoison, PF_ANY) 576 TESTSCFLAG(HWPoison, hwpoison, PF_ANY) 577 #define __PG_HWPOISON (1UL << PG_hwpoison) 578 #define MAGIC_HWPOISON 0x48575053U /* HWPS */ 579 extern void SetPageHWPoisonTakenOff(struct page *page); 580 extern void ClearPageHWPoisonTakenOff(struct page *page); 581 extern bool take_page_off_buddy(struct page *page); 582 extern bool put_page_back_buddy(struct page *page); 583 #else 584 PAGEFLAG_FALSE(HWPoison, hwpoison) 585 #define __PG_HWPOISON 0 586 #endif 587 588 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 589 TESTPAGEFLAG(Young, young, PF_ANY) 590 SETPAGEFLAG(Young, young, PF_ANY) 591 TESTCLEARFLAG(Young, young, PF_ANY) 592 PAGEFLAG(Idle, idle, PF_ANY) 593 #endif 594 595 #ifdef CONFIG_KASAN_HW_TAGS 596 PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD) 597 #else 598 PAGEFLAG_FALSE(SkipKASanPoison, skip_kasan_poison) 599 #endif 600 601 /* 602 * PageReported() is used to track reported free pages within the Buddy 603 * allocator. We can use the non-atomic version of the test and set 604 * operations as both should be shielded with the zone lock to prevent 605 * any possible races on the setting or clearing of the bit. 606 */ 607 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND) 608 609 #ifdef CONFIG_MEMORY_HOTPLUG 610 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY) 611 #else 612 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted) 613 #endif 614 615 /* 616 * On an anonymous page mapped into a user virtual memory area, 617 * page->mapping points to its anon_vma, not to a struct address_space; 618 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 619 * 620 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 621 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON 622 * bit; and then page->mapping points, not to an anon_vma, but to a private 623 * structure which KSM associates with that merged page. See ksm.h. 624 * 625 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable 626 * page and then page->mapping points to a struct movable_operations. 627 * 628 * Please note that, confusingly, "page_mapping" refers to the inode 629 * address_space which maps the page from disk; whereas "page_mapped" 630 * refers to user virtual address space into which the page is mapped. 631 */ 632 #define PAGE_MAPPING_ANON 0x1 633 #define PAGE_MAPPING_MOVABLE 0x2 634 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 635 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 636 637 /* 638 * Different with flags above, this flag is used only for fsdax mode. It 639 * indicates that this page->mapping is now under reflink case. 640 */ 641 #define PAGE_MAPPING_DAX_SHARED ((void *)0x1) 642 643 static __always_inline bool folio_mapping_flags(struct folio *folio) 644 { 645 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0; 646 } 647 648 static __always_inline int PageMappingFlags(struct page *page) 649 { 650 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; 651 } 652 653 static __always_inline bool folio_test_anon(struct folio *folio) 654 { 655 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0; 656 } 657 658 static __always_inline bool PageAnon(struct page *page) 659 { 660 return folio_test_anon(page_folio(page)); 661 } 662 663 static __always_inline bool __folio_test_movable(const struct folio *folio) 664 { 665 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 666 PAGE_MAPPING_MOVABLE; 667 } 668 669 static __always_inline int __PageMovable(struct page *page) 670 { 671 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 672 PAGE_MAPPING_MOVABLE; 673 } 674 675 #ifdef CONFIG_KSM 676 /* 677 * A KSM page is one of those write-protected "shared pages" or "merged pages" 678 * which KSM maps into multiple mms, wherever identical anonymous page content 679 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any 680 * anon_vma, but to that page's node of the stable tree. 681 */ 682 static __always_inline bool folio_test_ksm(struct folio *folio) 683 { 684 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 685 PAGE_MAPPING_KSM; 686 } 687 688 static __always_inline bool PageKsm(struct page *page) 689 { 690 return folio_test_ksm(page_folio(page)); 691 } 692 #else 693 TESTPAGEFLAG_FALSE(Ksm, ksm) 694 #endif 695 696 u64 stable_page_flags(struct page *page); 697 698 /** 699 * folio_test_uptodate - Is this folio up to date? 700 * @folio: The folio. 701 * 702 * The uptodate flag is set on a folio when every byte in the folio is 703 * at least as new as the corresponding bytes on storage. Anonymous 704 * and CoW folios are always uptodate. If the folio is not uptodate, 705 * some of the bytes in it may be; see the is_partially_uptodate() 706 * address_space operation. 707 */ 708 static inline bool folio_test_uptodate(struct folio *folio) 709 { 710 bool ret = test_bit(PG_uptodate, folio_flags(folio, 0)); 711 /* 712 * Must ensure that the data we read out of the folio is loaded 713 * _after_ we've loaded folio->flags to check the uptodate bit. 714 * We can skip the barrier if the folio is not uptodate, because 715 * we wouldn't be reading anything from it. 716 * 717 * See folio_mark_uptodate() for the other side of the story. 718 */ 719 if (ret) 720 smp_rmb(); 721 722 return ret; 723 } 724 725 static inline int PageUptodate(struct page *page) 726 { 727 return folio_test_uptodate(page_folio(page)); 728 } 729 730 static __always_inline void __folio_mark_uptodate(struct folio *folio) 731 { 732 smp_wmb(); 733 __set_bit(PG_uptodate, folio_flags(folio, 0)); 734 } 735 736 static __always_inline void folio_mark_uptodate(struct folio *folio) 737 { 738 /* 739 * Memory barrier must be issued before setting the PG_uptodate bit, 740 * so that all previous stores issued in order to bring the folio 741 * uptodate are actually visible before folio_test_uptodate becomes true. 742 */ 743 smp_wmb(); 744 set_bit(PG_uptodate, folio_flags(folio, 0)); 745 } 746 747 static __always_inline void __SetPageUptodate(struct page *page) 748 { 749 __folio_mark_uptodate((struct folio *)page); 750 } 751 752 static __always_inline void SetPageUptodate(struct page *page) 753 { 754 folio_mark_uptodate((struct folio *)page); 755 } 756 757 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) 758 759 bool __folio_start_writeback(struct folio *folio, bool keep_write); 760 bool set_page_writeback(struct page *page); 761 762 #define folio_start_writeback(folio) \ 763 __folio_start_writeback(folio, false) 764 #define folio_start_writeback_keepwrite(folio) \ 765 __folio_start_writeback(folio, true) 766 767 static inline void set_page_writeback_keepwrite(struct page *page) 768 { 769 folio_start_writeback_keepwrite(page_folio(page)); 770 } 771 772 static inline bool test_set_page_writeback(struct page *page) 773 { 774 return set_page_writeback(page); 775 } 776 777 static __always_inline bool folio_test_head(struct folio *folio) 778 { 779 return test_bit(PG_head, folio_flags(folio, FOLIO_PF_ANY)); 780 } 781 782 static __always_inline int PageHead(struct page *page) 783 { 784 PF_POISONED_CHECK(page); 785 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page); 786 } 787 788 __SETPAGEFLAG(Head, head, PF_ANY) 789 __CLEARPAGEFLAG(Head, head, PF_ANY) 790 CLEARPAGEFLAG(Head, head, PF_ANY) 791 792 /** 793 * folio_test_large() - Does this folio contain more than one page? 794 * @folio: The folio to test. 795 * 796 * Return: True if the folio is larger than one page. 797 */ 798 static inline bool folio_test_large(struct folio *folio) 799 { 800 return folio_test_head(folio); 801 } 802 803 static __always_inline void set_compound_head(struct page *page, struct page *head) 804 { 805 WRITE_ONCE(page->compound_head, (unsigned long)head + 1); 806 } 807 808 static __always_inline void clear_compound_head(struct page *page) 809 { 810 WRITE_ONCE(page->compound_head, 0); 811 } 812 813 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 814 static inline void ClearPageCompound(struct page *page) 815 { 816 BUG_ON(!PageHead(page)); 817 ClearPageHead(page); 818 } 819 #endif 820 821 #define PG_head_mask ((1UL << PG_head)) 822 823 #ifdef CONFIG_HUGETLB_PAGE 824 int PageHuge(struct page *page); 825 int PageHeadHuge(struct page *page); 826 static inline bool folio_test_hugetlb(struct folio *folio) 827 { 828 return PageHeadHuge(&folio->page); 829 } 830 #else 831 TESTPAGEFLAG_FALSE(Huge, hugetlb) 832 TESTPAGEFLAG_FALSE(HeadHuge, headhuge) 833 #endif 834 835 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 836 /* 837 * PageHuge() only returns true for hugetlbfs pages, but not for 838 * normal or transparent huge pages. 839 * 840 * PageTransHuge() returns true for both transparent huge and 841 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be 842 * called only in the core VM paths where hugetlbfs pages can't exist. 843 */ 844 static inline int PageTransHuge(struct page *page) 845 { 846 VM_BUG_ON_PAGE(PageTail(page), page); 847 return PageHead(page); 848 } 849 850 static inline bool folio_test_transhuge(struct folio *folio) 851 { 852 return folio_test_head(folio); 853 } 854 855 /* 856 * PageTransCompound returns true for both transparent huge pages 857 * and hugetlbfs pages, so it should only be called when it's known 858 * that hugetlbfs pages aren't involved. 859 */ 860 static inline int PageTransCompound(struct page *page) 861 { 862 return PageCompound(page); 863 } 864 865 /* 866 * PageTransTail returns true for both transparent huge pages 867 * and hugetlbfs pages, so it should only be called when it's known 868 * that hugetlbfs pages aren't involved. 869 */ 870 static inline int PageTransTail(struct page *page) 871 { 872 return PageTail(page); 873 } 874 #else 875 TESTPAGEFLAG_FALSE(TransHuge, transhuge) 876 TESTPAGEFLAG_FALSE(TransCompound, transcompound) 877 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap) 878 TESTPAGEFLAG_FALSE(TransTail, transtail) 879 #endif 880 881 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE) 882 /* 883 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the 884 * compound page. 885 * 886 * This flag is set by hwpoison handler. Cleared by THP split or free page. 887 */ 888 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 889 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 890 #else 891 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 892 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 893 #endif 894 895 /* 896 * Check if a page is currently marked HWPoisoned. Note that this check is 897 * best effort only and inherently racy: there is no way to synchronize with 898 * failing hardware. 899 */ 900 static inline bool is_page_hwpoison(struct page *page) 901 { 902 if (PageHWPoison(page)) 903 return true; 904 return PageHuge(page) && PageHWPoison(compound_head(page)); 905 } 906 907 /* 908 * For pages that are never mapped to userspace (and aren't PageSlab), 909 * page_type may be used. Because it is initialised to -1, we invert the 910 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and 911 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and 912 * low bits so that an underflow or overflow of page_mapcount() won't be 913 * mistaken for a page type value. 914 */ 915 916 #define PAGE_TYPE_BASE 0xf0000000 917 /* Reserve 0x0000007f to catch underflows of page_mapcount */ 918 #define PAGE_MAPCOUNT_RESERVE -128 919 #define PG_buddy 0x00000080 920 #define PG_offline 0x00000100 921 #define PG_table 0x00000200 922 #define PG_guard 0x00000400 923 924 #define PageType(page, flag) \ 925 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) 926 927 static inline int page_has_type(struct page *page) 928 { 929 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE; 930 } 931 932 #define PAGE_TYPE_OPS(uname, lname) \ 933 static __always_inline int Page##uname(struct page *page) \ 934 { \ 935 return PageType(page, PG_##lname); \ 936 } \ 937 static __always_inline void __SetPage##uname(struct page *page) \ 938 { \ 939 VM_BUG_ON_PAGE(!PageType(page, 0), page); \ 940 page->page_type &= ~PG_##lname; \ 941 } \ 942 static __always_inline void __ClearPage##uname(struct page *page) \ 943 { \ 944 VM_BUG_ON_PAGE(!Page##uname(page), page); \ 945 page->page_type |= PG_##lname; \ 946 } 947 948 /* 949 * PageBuddy() indicates that the page is free and in the buddy system 950 * (see mm/page_alloc.c). 951 */ 952 PAGE_TYPE_OPS(Buddy, buddy) 953 954 /* 955 * PageOffline() indicates that the page is logically offline although the 956 * containing section is online. (e.g. inflated in a balloon driver or 957 * not onlined when onlining the section). 958 * The content of these pages is effectively stale. Such pages should not 959 * be touched (read/write/dump/save) except by their owner. 960 * 961 * If a driver wants to allow to offline unmovable PageOffline() pages without 962 * putting them back to the buddy, it can do so via the memory notifier by 963 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the 964 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline() 965 * pages (now with a reference count of zero) are treated like free pages, 966 * allowing the containing memory block to get offlined. A driver that 967 * relies on this feature is aware that re-onlining the memory block will 968 * require to re-set the pages PageOffline() and not giving them to the 969 * buddy via online_page_callback_t. 970 * 971 * There are drivers that mark a page PageOffline() and expect there won't be 972 * any further access to page content. PFN walkers that read content of random 973 * pages should check PageOffline() and synchronize with such drivers using 974 * page_offline_freeze()/page_offline_thaw(). 975 */ 976 PAGE_TYPE_OPS(Offline, offline) 977 978 extern void page_offline_freeze(void); 979 extern void page_offline_thaw(void); 980 extern void page_offline_begin(void); 981 extern void page_offline_end(void); 982 983 /* 984 * Marks pages in use as page tables. 985 */ 986 PAGE_TYPE_OPS(Table, table) 987 988 /* 989 * Marks guardpages used with debug_pagealloc. 990 */ 991 PAGE_TYPE_OPS(Guard, guard) 992 993 extern bool is_free_buddy_page(struct page *page); 994 995 PAGEFLAG(Isolated, isolated, PF_ANY); 996 997 static __always_inline int PageAnonExclusive(struct page *page) 998 { 999 VM_BUG_ON_PGFLAGS(!PageAnon(page), page); 1000 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1001 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1002 } 1003 1004 static __always_inline void SetPageAnonExclusive(struct page *page) 1005 { 1006 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page); 1007 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1008 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1009 } 1010 1011 static __always_inline void ClearPageAnonExclusive(struct page *page) 1012 { 1013 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page); 1014 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1015 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1016 } 1017 1018 static __always_inline void __ClearPageAnonExclusive(struct page *page) 1019 { 1020 VM_BUG_ON_PGFLAGS(!PageAnon(page), page); 1021 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1022 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1023 } 1024 1025 #ifdef CONFIG_MMU 1026 #define __PG_MLOCKED (1UL << PG_mlocked) 1027 #else 1028 #define __PG_MLOCKED 0 1029 #endif 1030 1031 /* 1032 * Flags checked when a page is freed. Pages being freed should not have 1033 * these flags set. If they are, there is a problem. 1034 */ 1035 #define PAGE_FLAGS_CHECK_AT_FREE \ 1036 (1UL << PG_lru | 1UL << PG_locked | \ 1037 1UL << PG_private | 1UL << PG_private_2 | \ 1038 1UL << PG_writeback | 1UL << PG_reserved | \ 1039 1UL << PG_slab | 1UL << PG_active | \ 1040 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK) 1041 1042 /* 1043 * Flags checked when a page is prepped for return by the page allocator. 1044 * Pages being prepped should not have these flags set. If they are set, 1045 * there has been a kernel bug or struct page corruption. 1046 * 1047 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's 1048 * alloc-free cycle to prevent from reusing the page. 1049 */ 1050 #define PAGE_FLAGS_CHECK_AT_PREP \ 1051 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK) 1052 1053 #define PAGE_FLAGS_PRIVATE \ 1054 (1UL << PG_private | 1UL << PG_private_2) 1055 /** 1056 * page_has_private - Determine if page has private stuff 1057 * @page: The page to be checked 1058 * 1059 * Determine if a page has private stuff, indicating that release routines 1060 * should be invoked upon it. 1061 */ 1062 static inline int page_has_private(struct page *page) 1063 { 1064 return !!(page->flags & PAGE_FLAGS_PRIVATE); 1065 } 1066 1067 static inline bool folio_has_private(struct folio *folio) 1068 { 1069 return page_has_private(&folio->page); 1070 } 1071 1072 #undef PF_ANY 1073 #undef PF_HEAD 1074 #undef PF_ONLY_HEAD 1075 #undef PF_NO_TAIL 1076 #undef PF_NO_COMPOUND 1077 #undef PF_SECOND 1078 #endif /* !__GENERATING_BOUNDS_H */ 1079 1080 #endif /* PAGE_FLAGS_H */ 1081