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_ARCH_USES_PG_ARCH_X 136 PG_arch_2, 137 PG_arch_3, 138 #endif 139 #ifdef CONFIG_KASAN_HW_TAGS 140 PG_skip_kasan_poison, 141 #endif 142 __NR_PAGEFLAGS, 143 144 PG_readahead = PG_reclaim, 145 146 /* 147 * Depending on the way an anonymous folio can be mapped into a page 148 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped 149 * THP), PG_anon_exclusive may be set only for the head page or for 150 * tail pages of an anonymous folio. For now, we only expect it to be 151 * set on tail pages for PTE-mapped THP. 152 */ 153 PG_anon_exclusive = PG_mappedtodisk, 154 155 /* Filesystems */ 156 PG_checked = PG_owner_priv_1, 157 158 /* SwapBacked */ 159 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */ 160 161 /* Two page bits are conscripted by FS-Cache to maintain local caching 162 * state. These bits are set on pages belonging to the netfs's inodes 163 * when those inodes are being locally cached. 164 */ 165 PG_fscache = PG_private_2, /* page backed by cache */ 166 167 /* XEN */ 168 /* Pinned in Xen as a read-only pagetable page. */ 169 PG_pinned = PG_owner_priv_1, 170 /* Pinned as part of domain save (see xen_mm_pin_all()). */ 171 PG_savepinned = PG_dirty, 172 /* Has a grant mapping of another (foreign) domain's page. */ 173 PG_foreign = PG_owner_priv_1, 174 /* Remapped by swiotlb-xen. */ 175 PG_xen_remapped = PG_owner_priv_1, 176 177 #ifdef CONFIG_MEMORY_FAILURE 178 /* 179 * Compound pages. Stored in first tail page's flags. 180 * Indicates that at least one subpage is hwpoisoned in the 181 * THP. 182 */ 183 PG_has_hwpoisoned = PG_error, 184 #endif 185 186 /* non-lru isolated movable page */ 187 PG_isolated = PG_reclaim, 188 189 /* Only valid for buddy pages. Used to track pages that are reported */ 190 PG_reported = PG_uptodate, 191 192 #ifdef CONFIG_MEMORY_HOTPLUG 193 /* For self-hosted memmap pages */ 194 PG_vmemmap_self_hosted = PG_owner_priv_1, 195 #endif 196 }; 197 198 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1) 199 200 #ifndef __GENERATING_BOUNDS_H 201 202 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 203 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key); 204 205 /* 206 * Return the real head page struct iff the @page is a fake head page, otherwise 207 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst. 208 */ 209 static __always_inline const struct page *page_fixed_fake_head(const struct page *page) 210 { 211 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key)) 212 return page; 213 214 /* 215 * Only addresses aligned with PAGE_SIZE of struct page may be fake head 216 * struct page. The alignment check aims to avoid access the fields ( 217 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly) 218 * cold cacheline in some cases. 219 */ 220 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) && 221 test_bit(PG_head, &page->flags)) { 222 /* 223 * We can safely access the field of the @page[1] with PG_head 224 * because the @page is a compound page composed with at least 225 * two contiguous pages. 226 */ 227 unsigned long head = READ_ONCE(page[1].compound_head); 228 229 if (likely(head & 1)) 230 return (const struct page *)(head - 1); 231 } 232 return page; 233 } 234 #else 235 static inline const struct page *page_fixed_fake_head(const struct page *page) 236 { 237 return page; 238 } 239 #endif 240 241 static __always_inline int page_is_fake_head(struct page *page) 242 { 243 return page_fixed_fake_head(page) != page; 244 } 245 246 static inline unsigned long _compound_head(const struct page *page) 247 { 248 unsigned long head = READ_ONCE(page->compound_head); 249 250 if (unlikely(head & 1)) 251 return head - 1; 252 return (unsigned long)page_fixed_fake_head(page); 253 } 254 255 #define compound_head(page) ((typeof(page))_compound_head(page)) 256 257 /** 258 * page_folio - Converts from page to folio. 259 * @p: The page. 260 * 261 * Every page is part of a folio. This function cannot be called on a 262 * NULL pointer. 263 * 264 * Context: No reference, nor lock is required on @page. If the caller 265 * does not hold a reference, this call may race with a folio split, so 266 * it should re-check the folio still contains this page after gaining 267 * a reference on the folio. 268 * Return: The folio which contains this page. 269 */ 270 #define page_folio(p) (_Generic((p), \ 271 const struct page *: (const struct folio *)_compound_head(p), \ 272 struct page *: (struct folio *)_compound_head(p))) 273 274 /** 275 * folio_page - Return a page from a folio. 276 * @folio: The folio. 277 * @n: The page number to return. 278 * 279 * @n is relative to the start of the folio. This function does not 280 * check that the page number lies within @folio; the caller is presumed 281 * to have a reference to the page. 282 */ 283 #define folio_page(folio, n) nth_page(&(folio)->page, n) 284 285 static __always_inline int PageTail(struct page *page) 286 { 287 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page); 288 } 289 290 static __always_inline int PageCompound(struct page *page) 291 { 292 return test_bit(PG_head, &page->flags) || 293 READ_ONCE(page->compound_head) & 1; 294 } 295 296 #define PAGE_POISON_PATTERN -1l 297 static inline int PagePoisoned(const struct page *page) 298 { 299 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN; 300 } 301 302 #ifdef CONFIG_DEBUG_VM 303 void page_init_poison(struct page *page, size_t size); 304 #else 305 static inline void page_init_poison(struct page *page, size_t size) 306 { 307 } 308 #endif 309 310 static unsigned long *folio_flags(struct folio *folio, unsigned n) 311 { 312 struct page *page = &folio->page; 313 314 VM_BUG_ON_PGFLAGS(PageTail(page), page); 315 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); 316 return &page[n].flags; 317 } 318 319 /* 320 * Page flags policies wrt compound pages 321 * 322 * PF_POISONED_CHECK 323 * check if this struct page poisoned/uninitialized 324 * 325 * PF_ANY: 326 * the page flag is relevant for small, head and tail pages. 327 * 328 * PF_HEAD: 329 * for compound page all operations related to the page flag applied to 330 * head page. 331 * 332 * PF_ONLY_HEAD: 333 * for compound page, callers only ever operate on the head page. 334 * 335 * PF_NO_TAIL: 336 * modifications of the page flag must be done on small or head pages, 337 * checks can be done on tail pages too. 338 * 339 * PF_NO_COMPOUND: 340 * the page flag is not relevant for compound pages. 341 * 342 * PF_SECOND: 343 * the page flag is stored in the first tail page. 344 */ 345 #define PF_POISONED_CHECK(page) ({ \ 346 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ 347 page; }) 348 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page) 349 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) 350 #define PF_ONLY_HEAD(page, enforce) ({ \ 351 VM_BUG_ON_PGFLAGS(PageTail(page), page); \ 352 PF_POISONED_CHECK(page); }) 353 #define PF_NO_TAIL(page, enforce) ({ \ 354 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ 355 PF_POISONED_CHECK(compound_head(page)); }) 356 #define PF_NO_COMPOUND(page, enforce) ({ \ 357 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ 358 PF_POISONED_CHECK(page); }) 359 #define PF_SECOND(page, enforce) ({ \ 360 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \ 361 PF_POISONED_CHECK(&page[1]); }) 362 363 /* Which page is the flag stored in */ 364 #define FOLIO_PF_ANY 0 365 #define FOLIO_PF_HEAD 0 366 #define FOLIO_PF_ONLY_HEAD 0 367 #define FOLIO_PF_NO_TAIL 0 368 #define FOLIO_PF_NO_COMPOUND 0 369 #define FOLIO_PF_SECOND 1 370 371 /* 372 * Macros to create function definitions for page flags 373 */ 374 #define TESTPAGEFLAG(uname, lname, policy) \ 375 static __always_inline bool folio_test_##lname(struct folio *folio) \ 376 { return test_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 377 static __always_inline int Page##uname(struct page *page) \ 378 { return test_bit(PG_##lname, &policy(page, 0)->flags); } 379 380 #define SETPAGEFLAG(uname, lname, policy) \ 381 static __always_inline \ 382 void folio_set_##lname(struct folio *folio) \ 383 { set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 384 static __always_inline void SetPage##uname(struct page *page) \ 385 { set_bit(PG_##lname, &policy(page, 1)->flags); } 386 387 #define CLEARPAGEFLAG(uname, lname, policy) \ 388 static __always_inline \ 389 void folio_clear_##lname(struct folio *folio) \ 390 { clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 391 static __always_inline void ClearPage##uname(struct page *page) \ 392 { clear_bit(PG_##lname, &policy(page, 1)->flags); } 393 394 #define __SETPAGEFLAG(uname, lname, policy) \ 395 static __always_inline \ 396 void __folio_set_##lname(struct folio *folio) \ 397 { __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 398 static __always_inline void __SetPage##uname(struct page *page) \ 399 { __set_bit(PG_##lname, &policy(page, 1)->flags); } 400 401 #define __CLEARPAGEFLAG(uname, lname, policy) \ 402 static __always_inline \ 403 void __folio_clear_##lname(struct folio *folio) \ 404 { __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 405 static __always_inline void __ClearPage##uname(struct page *page) \ 406 { __clear_bit(PG_##lname, &policy(page, 1)->flags); } 407 408 #define TESTSETFLAG(uname, lname, policy) \ 409 static __always_inline \ 410 bool folio_test_set_##lname(struct folio *folio) \ 411 { return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 412 static __always_inline int TestSetPage##uname(struct page *page) \ 413 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } 414 415 #define TESTCLEARFLAG(uname, lname, policy) \ 416 static __always_inline \ 417 bool folio_test_clear_##lname(struct folio *folio) \ 418 { return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 419 static __always_inline int TestClearPage##uname(struct page *page) \ 420 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } 421 422 #define PAGEFLAG(uname, lname, policy) \ 423 TESTPAGEFLAG(uname, lname, policy) \ 424 SETPAGEFLAG(uname, lname, policy) \ 425 CLEARPAGEFLAG(uname, lname, policy) 426 427 #define __PAGEFLAG(uname, lname, policy) \ 428 TESTPAGEFLAG(uname, lname, policy) \ 429 __SETPAGEFLAG(uname, lname, policy) \ 430 __CLEARPAGEFLAG(uname, lname, policy) 431 432 #define TESTSCFLAG(uname, lname, policy) \ 433 TESTSETFLAG(uname, lname, policy) \ 434 TESTCLEARFLAG(uname, lname, policy) 435 436 #define TESTPAGEFLAG_FALSE(uname, lname) \ 437 static inline bool folio_test_##lname(const struct folio *folio) { return false; } \ 438 static inline int Page##uname(const struct page *page) { return 0; } 439 440 #define SETPAGEFLAG_NOOP(uname, lname) \ 441 static inline void folio_set_##lname(struct folio *folio) { } \ 442 static inline void SetPage##uname(struct page *page) { } 443 444 #define CLEARPAGEFLAG_NOOP(uname, lname) \ 445 static inline void folio_clear_##lname(struct folio *folio) { } \ 446 static inline void ClearPage##uname(struct page *page) { } 447 448 #define __CLEARPAGEFLAG_NOOP(uname, lname) \ 449 static inline void __folio_clear_##lname(struct folio *folio) { } \ 450 static inline void __ClearPage##uname(struct page *page) { } 451 452 #define TESTSETFLAG_FALSE(uname, lname) \ 453 static inline bool folio_test_set_##lname(struct folio *folio) \ 454 { return 0; } \ 455 static inline int TestSetPage##uname(struct page *page) { return 0; } 456 457 #define TESTCLEARFLAG_FALSE(uname, lname) \ 458 static inline bool folio_test_clear_##lname(struct folio *folio) \ 459 { return 0; } \ 460 static inline int TestClearPage##uname(struct page *page) { return 0; } 461 462 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \ 463 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname) 464 465 #define TESTSCFLAG_FALSE(uname, lname) \ 466 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname) 467 468 __PAGEFLAG(Locked, locked, PF_NO_TAIL) 469 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) 470 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL) 471 PAGEFLAG(Referenced, referenced, PF_HEAD) 472 TESTCLEARFLAG(Referenced, referenced, PF_HEAD) 473 __SETPAGEFLAG(Referenced, referenced, PF_HEAD) 474 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) 475 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) 476 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) 477 TESTCLEARFLAG(LRU, lru, PF_HEAD) 478 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) 479 TESTCLEARFLAG(Active, active, PF_HEAD) 480 PAGEFLAG(Workingset, workingset, PF_HEAD) 481 TESTCLEARFLAG(Workingset, workingset, PF_HEAD) 482 __PAGEFLAG(Slab, slab, PF_NO_TAIL) 483 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ 484 485 /* Xen */ 486 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) 487 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) 488 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); 489 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); 490 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 491 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 492 493 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 494 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 495 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 496 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 497 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 498 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 499 500 /* 501 * Private page markings that may be used by the filesystem that owns the page 502 * for its own purposes. 503 * - PG_private and PG_private_2 cause release_folio() and co to be invoked 504 */ 505 PAGEFLAG(Private, private, PF_ANY) 506 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) 507 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 508 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 509 510 /* 511 * Only test-and-set exist for PG_writeback. The unconditional operators are 512 * risky: they bypass page accounting. 513 */ 514 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) 515 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) 516 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) 517 518 /* PG_readahead is only used for reads; PG_reclaim is only for writes */ 519 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) 520 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) 521 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND) 522 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND) 523 524 #ifdef CONFIG_HIGHMEM 525 /* 526 * Must use a macro here due to header dependency issues. page_zone() is not 527 * available at this point. 528 */ 529 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) 530 #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f)) 531 #else 532 PAGEFLAG_FALSE(HighMem, highmem) 533 #endif 534 535 #ifdef CONFIG_SWAP 536 static __always_inline bool folio_test_swapcache(struct folio *folio) 537 { 538 return folio_test_swapbacked(folio) && 539 test_bit(PG_swapcache, folio_flags(folio, 0)); 540 } 541 542 static __always_inline bool PageSwapCache(struct page *page) 543 { 544 return folio_test_swapcache(page_folio(page)); 545 } 546 547 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 548 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 549 #else 550 PAGEFLAG_FALSE(SwapCache, swapcache) 551 #endif 552 553 PAGEFLAG(Unevictable, unevictable, PF_HEAD) 554 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) 555 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) 556 557 #ifdef CONFIG_MMU 558 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 559 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 560 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) 561 #else 562 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked) 563 TESTSCFLAG_FALSE(Mlocked, mlocked) 564 #endif 565 566 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 567 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) 568 #else 569 PAGEFLAG_FALSE(Uncached, uncached) 570 #endif 571 572 #ifdef CONFIG_MEMORY_FAILURE 573 PAGEFLAG(HWPoison, hwpoison, PF_ANY) 574 TESTSCFLAG(HWPoison, hwpoison, PF_ANY) 575 #define __PG_HWPOISON (1UL << PG_hwpoison) 576 #define MAGIC_HWPOISON 0x48575053U /* HWPS */ 577 extern void SetPageHWPoisonTakenOff(struct page *page); 578 extern void ClearPageHWPoisonTakenOff(struct page *page); 579 extern bool take_page_off_buddy(struct page *page); 580 extern bool put_page_back_buddy(struct page *page); 581 #else 582 PAGEFLAG_FALSE(HWPoison, hwpoison) 583 #define __PG_HWPOISON 0 584 #endif 585 586 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 587 TESTPAGEFLAG(Young, young, PF_ANY) 588 SETPAGEFLAG(Young, young, PF_ANY) 589 TESTCLEARFLAG(Young, young, PF_ANY) 590 PAGEFLAG(Idle, idle, PF_ANY) 591 #endif 592 593 #ifdef CONFIG_KASAN_HW_TAGS 594 PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD) 595 #else 596 PAGEFLAG_FALSE(SkipKASanPoison, skip_kasan_poison) 597 #endif 598 599 /* 600 * PageReported() is used to track reported free pages within the Buddy 601 * allocator. We can use the non-atomic version of the test and set 602 * operations as both should be shielded with the zone lock to prevent 603 * any possible races on the setting or clearing of the bit. 604 */ 605 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND) 606 607 #ifdef CONFIG_MEMORY_HOTPLUG 608 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY) 609 #else 610 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted) 611 #endif 612 613 /* 614 * On an anonymous page mapped into a user virtual memory area, 615 * page->mapping points to its anon_vma, not to a struct address_space; 616 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 617 * 618 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 619 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON 620 * bit; and then page->mapping points, not to an anon_vma, but to a private 621 * structure which KSM associates with that merged page. See ksm.h. 622 * 623 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable 624 * page and then page->mapping points to a struct movable_operations. 625 * 626 * Please note that, confusingly, "page_mapping" refers to the inode 627 * address_space which maps the page from disk; whereas "page_mapped" 628 * refers to user virtual address space into which the page is mapped. 629 */ 630 #define PAGE_MAPPING_ANON 0x1 631 #define PAGE_MAPPING_MOVABLE 0x2 632 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 633 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 634 635 /* 636 * Different with flags above, this flag is used only for fsdax mode. It 637 * indicates that this page->mapping is now under reflink case. 638 */ 639 #define PAGE_MAPPING_DAX_SHARED ((void *)0x1) 640 641 static __always_inline bool folio_mapping_flags(struct folio *folio) 642 { 643 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0; 644 } 645 646 static __always_inline int PageMappingFlags(struct page *page) 647 { 648 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; 649 } 650 651 static __always_inline bool folio_test_anon(struct folio *folio) 652 { 653 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0; 654 } 655 656 static __always_inline bool PageAnon(struct page *page) 657 { 658 return folio_test_anon(page_folio(page)); 659 } 660 661 static __always_inline bool __folio_test_movable(const struct folio *folio) 662 { 663 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 664 PAGE_MAPPING_MOVABLE; 665 } 666 667 static __always_inline int __PageMovable(struct page *page) 668 { 669 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 670 PAGE_MAPPING_MOVABLE; 671 } 672 673 #ifdef CONFIG_KSM 674 /* 675 * A KSM page is one of those write-protected "shared pages" or "merged pages" 676 * which KSM maps into multiple mms, wherever identical anonymous page content 677 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any 678 * anon_vma, but to that page's node of the stable tree. 679 */ 680 static __always_inline bool folio_test_ksm(struct folio *folio) 681 { 682 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 683 PAGE_MAPPING_KSM; 684 } 685 686 static __always_inline bool PageKsm(struct page *page) 687 { 688 return folio_test_ksm(page_folio(page)); 689 } 690 #else 691 TESTPAGEFLAG_FALSE(Ksm, ksm) 692 #endif 693 694 u64 stable_page_flags(struct page *page); 695 696 /** 697 * folio_test_uptodate - Is this folio up to date? 698 * @folio: The folio. 699 * 700 * The uptodate flag is set on a folio when every byte in the folio is 701 * at least as new as the corresponding bytes on storage. Anonymous 702 * and CoW folios are always uptodate. If the folio is not uptodate, 703 * some of the bytes in it may be; see the is_partially_uptodate() 704 * address_space operation. 705 */ 706 static inline bool folio_test_uptodate(struct folio *folio) 707 { 708 bool ret = test_bit(PG_uptodate, folio_flags(folio, 0)); 709 /* 710 * Must ensure that the data we read out of the folio is loaded 711 * _after_ we've loaded folio->flags to check the uptodate bit. 712 * We can skip the barrier if the folio is not uptodate, because 713 * we wouldn't be reading anything from it. 714 * 715 * See folio_mark_uptodate() for the other side of the story. 716 */ 717 if (ret) 718 smp_rmb(); 719 720 return ret; 721 } 722 723 static inline int PageUptodate(struct page *page) 724 { 725 return folio_test_uptodate(page_folio(page)); 726 } 727 728 static __always_inline void __folio_mark_uptodate(struct folio *folio) 729 { 730 smp_wmb(); 731 __set_bit(PG_uptodate, folio_flags(folio, 0)); 732 } 733 734 static __always_inline void folio_mark_uptodate(struct folio *folio) 735 { 736 /* 737 * Memory barrier must be issued before setting the PG_uptodate bit, 738 * so that all previous stores issued in order to bring the folio 739 * uptodate are actually visible before folio_test_uptodate becomes true. 740 */ 741 smp_wmb(); 742 set_bit(PG_uptodate, folio_flags(folio, 0)); 743 } 744 745 static __always_inline void __SetPageUptodate(struct page *page) 746 { 747 __folio_mark_uptodate((struct folio *)page); 748 } 749 750 static __always_inline void SetPageUptodate(struct page *page) 751 { 752 folio_mark_uptodate((struct folio *)page); 753 } 754 755 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) 756 757 bool __folio_start_writeback(struct folio *folio, bool keep_write); 758 bool set_page_writeback(struct page *page); 759 760 #define folio_start_writeback(folio) \ 761 __folio_start_writeback(folio, false) 762 #define folio_start_writeback_keepwrite(folio) \ 763 __folio_start_writeback(folio, true) 764 765 static inline bool test_set_page_writeback(struct page *page) 766 { 767 return set_page_writeback(page); 768 } 769 770 static __always_inline bool folio_test_head(struct folio *folio) 771 { 772 return test_bit(PG_head, folio_flags(folio, FOLIO_PF_ANY)); 773 } 774 775 static __always_inline int PageHead(struct page *page) 776 { 777 PF_POISONED_CHECK(page); 778 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page); 779 } 780 781 __SETPAGEFLAG(Head, head, PF_ANY) 782 __CLEARPAGEFLAG(Head, head, PF_ANY) 783 CLEARPAGEFLAG(Head, head, PF_ANY) 784 785 /** 786 * folio_test_large() - Does this folio contain more than one page? 787 * @folio: The folio to test. 788 * 789 * Return: True if the folio is larger than one page. 790 */ 791 static inline bool folio_test_large(struct folio *folio) 792 { 793 return folio_test_head(folio); 794 } 795 796 static __always_inline void set_compound_head(struct page *page, struct page *head) 797 { 798 WRITE_ONCE(page->compound_head, (unsigned long)head + 1); 799 } 800 801 static __always_inline void clear_compound_head(struct page *page) 802 { 803 WRITE_ONCE(page->compound_head, 0); 804 } 805 806 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 807 static inline void ClearPageCompound(struct page *page) 808 { 809 BUG_ON(!PageHead(page)); 810 ClearPageHead(page); 811 } 812 #endif 813 814 #define PG_head_mask ((1UL << PG_head)) 815 816 #ifdef CONFIG_HUGETLB_PAGE 817 int PageHuge(struct page *page); 818 int PageHeadHuge(struct page *page); 819 static inline bool folio_test_hugetlb(struct folio *folio) 820 { 821 return PageHeadHuge(&folio->page); 822 } 823 #else 824 TESTPAGEFLAG_FALSE(Huge, hugetlb) 825 TESTPAGEFLAG_FALSE(HeadHuge, headhuge) 826 #endif 827 828 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 829 /* 830 * PageHuge() only returns true for hugetlbfs pages, but not for 831 * normal or transparent huge pages. 832 * 833 * PageTransHuge() returns true for both transparent huge and 834 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be 835 * called only in the core VM paths where hugetlbfs pages can't exist. 836 */ 837 static inline int PageTransHuge(struct page *page) 838 { 839 VM_BUG_ON_PAGE(PageTail(page), page); 840 return PageHead(page); 841 } 842 843 static inline bool folio_test_transhuge(struct folio *folio) 844 { 845 return folio_test_head(folio); 846 } 847 848 /* 849 * PageTransCompound returns true for both transparent huge pages 850 * and hugetlbfs pages, so it should only be called when it's known 851 * that hugetlbfs pages aren't involved. 852 */ 853 static inline int PageTransCompound(struct page *page) 854 { 855 return PageCompound(page); 856 } 857 858 /* 859 * PageTransTail returns true for both transparent huge pages 860 * and hugetlbfs pages, so it should only be called when it's known 861 * that hugetlbfs pages aren't involved. 862 */ 863 static inline int PageTransTail(struct page *page) 864 { 865 return PageTail(page); 866 } 867 #else 868 TESTPAGEFLAG_FALSE(TransHuge, transhuge) 869 TESTPAGEFLAG_FALSE(TransCompound, transcompound) 870 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap) 871 TESTPAGEFLAG_FALSE(TransTail, transtail) 872 #endif 873 874 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE) 875 /* 876 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the 877 * compound page. 878 * 879 * This flag is set by hwpoison handler. Cleared by THP split or free page. 880 */ 881 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 882 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 883 #else 884 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 885 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 886 #endif 887 888 /* 889 * Check if a page is currently marked HWPoisoned. Note that this check is 890 * best effort only and inherently racy: there is no way to synchronize with 891 * failing hardware. 892 */ 893 static inline bool is_page_hwpoison(struct page *page) 894 { 895 if (PageHWPoison(page)) 896 return true; 897 return PageHuge(page) && PageHWPoison(compound_head(page)); 898 } 899 900 /* 901 * For pages that are never mapped to userspace (and aren't PageSlab), 902 * page_type may be used. Because it is initialised to -1, we invert the 903 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and 904 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and 905 * low bits so that an underflow or overflow of page_mapcount() won't be 906 * mistaken for a page type value. 907 */ 908 909 #define PAGE_TYPE_BASE 0xf0000000 910 /* Reserve 0x0000007f to catch underflows of page_mapcount */ 911 #define PAGE_MAPCOUNT_RESERVE -128 912 #define PG_buddy 0x00000080 913 #define PG_offline 0x00000100 914 #define PG_table 0x00000200 915 #define PG_guard 0x00000400 916 917 #define PageType(page, flag) \ 918 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) 919 920 static inline int page_has_type(struct page *page) 921 { 922 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE; 923 } 924 925 #define PAGE_TYPE_OPS(uname, lname) \ 926 static __always_inline int Page##uname(struct page *page) \ 927 { \ 928 return PageType(page, PG_##lname); \ 929 } \ 930 static __always_inline void __SetPage##uname(struct page *page) \ 931 { \ 932 VM_BUG_ON_PAGE(!PageType(page, 0), page); \ 933 page->page_type &= ~PG_##lname; \ 934 } \ 935 static __always_inline void __ClearPage##uname(struct page *page) \ 936 { \ 937 VM_BUG_ON_PAGE(!Page##uname(page), page); \ 938 page->page_type |= PG_##lname; \ 939 } 940 941 /* 942 * PageBuddy() indicates that the page is free and in the buddy system 943 * (see mm/page_alloc.c). 944 */ 945 PAGE_TYPE_OPS(Buddy, buddy) 946 947 /* 948 * PageOffline() indicates that the page is logically offline although the 949 * containing section is online. (e.g. inflated in a balloon driver or 950 * not onlined when onlining the section). 951 * The content of these pages is effectively stale. Such pages should not 952 * be touched (read/write/dump/save) except by their owner. 953 * 954 * If a driver wants to allow to offline unmovable PageOffline() pages without 955 * putting them back to the buddy, it can do so via the memory notifier by 956 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the 957 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline() 958 * pages (now with a reference count of zero) are treated like free pages, 959 * allowing the containing memory block to get offlined. A driver that 960 * relies on this feature is aware that re-onlining the memory block will 961 * require to re-set the pages PageOffline() and not giving them to the 962 * buddy via online_page_callback_t. 963 * 964 * There are drivers that mark a page PageOffline() and expect there won't be 965 * any further access to page content. PFN walkers that read content of random 966 * pages should check PageOffline() and synchronize with such drivers using 967 * page_offline_freeze()/page_offline_thaw(). 968 */ 969 PAGE_TYPE_OPS(Offline, offline) 970 971 extern void page_offline_freeze(void); 972 extern void page_offline_thaw(void); 973 extern void page_offline_begin(void); 974 extern void page_offline_end(void); 975 976 /* 977 * Marks pages in use as page tables. 978 */ 979 PAGE_TYPE_OPS(Table, table) 980 981 /* 982 * Marks guardpages used with debug_pagealloc. 983 */ 984 PAGE_TYPE_OPS(Guard, guard) 985 986 extern bool is_free_buddy_page(struct page *page); 987 988 PAGEFLAG(Isolated, isolated, PF_ANY); 989 990 static __always_inline int PageAnonExclusive(struct page *page) 991 { 992 VM_BUG_ON_PGFLAGS(!PageAnon(page), page); 993 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 994 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 995 } 996 997 static __always_inline void SetPageAnonExclusive(struct page *page) 998 { 999 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page); 1000 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1001 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1002 } 1003 1004 static __always_inline void ClearPageAnonExclusive(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 clear_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), 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 #ifdef CONFIG_MMU 1019 #define __PG_MLOCKED (1UL << PG_mlocked) 1020 #else 1021 #define __PG_MLOCKED 0 1022 #endif 1023 1024 /* 1025 * Flags checked when a page is freed. Pages being freed should not have 1026 * these flags set. If they are, there is a problem. 1027 */ 1028 #define PAGE_FLAGS_CHECK_AT_FREE \ 1029 (1UL << PG_lru | 1UL << PG_locked | \ 1030 1UL << PG_private | 1UL << PG_private_2 | \ 1031 1UL << PG_writeback | 1UL << PG_reserved | \ 1032 1UL << PG_slab | 1UL << PG_active | \ 1033 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK) 1034 1035 /* 1036 * Flags checked when a page is prepped for return by the page allocator. 1037 * Pages being prepped should not have these flags set. If they are set, 1038 * there has been a kernel bug or struct page corruption. 1039 * 1040 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's 1041 * alloc-free cycle to prevent from reusing the page. 1042 */ 1043 #define PAGE_FLAGS_CHECK_AT_PREP \ 1044 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK) 1045 1046 #define PAGE_FLAGS_PRIVATE \ 1047 (1UL << PG_private | 1UL << PG_private_2) 1048 /** 1049 * page_has_private - Determine if page has private stuff 1050 * @page: The page to be checked 1051 * 1052 * Determine if a page has private stuff, indicating that release routines 1053 * should be invoked upon it. 1054 */ 1055 static inline int page_has_private(struct page *page) 1056 { 1057 return !!(page->flags & PAGE_FLAGS_PRIVATE); 1058 } 1059 1060 static inline bool folio_has_private(struct folio *folio) 1061 { 1062 return page_has_private(&folio->page); 1063 } 1064 1065 #undef PF_ANY 1066 #undef PF_HEAD 1067 #undef PF_ONLY_HEAD 1068 #undef PF_NO_TAIL 1069 #undef PF_NO_COMPOUND 1070 #undef PF_SECOND 1071 #endif /* !__GENERATING_BOUNDS_H */ 1072 1073 #endif /* PAGE_FLAGS_H */ 1074