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