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