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_writeback, /* Page is under writeback */ 103 PG_referenced, 104 PG_uptodate, 105 PG_dirty, 106 PG_lru, 107 PG_head, /* Must be in bit 6 */ 108 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */ 109 PG_active, 110 PG_workingset, 111 PG_error, 112 PG_slab, 113 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/ 114 PG_arch_1, 115 PG_reserved, 116 PG_private, /* If pagecache, has fs-private data */ 117 PG_private_2, /* If pagecache, has fs aux data */ 118 PG_mappedtodisk, /* Has blocks allocated on-disk */ 119 PG_reclaim, /* To be reclaimed asap */ 120 PG_swapbacked, /* Page is backed by RAM/swap */ 121 PG_unevictable, /* Page is "unevictable" */ 122 #ifdef CONFIG_MMU 123 PG_mlocked, /* Page is vma mlocked */ 124 #endif 125 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 126 PG_uncached, /* Page has been mapped as uncached */ 127 #endif 128 #ifdef CONFIG_MEMORY_FAILURE 129 PG_hwpoison, /* hardware poisoned page. Don't touch */ 130 #endif 131 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 132 PG_young, 133 PG_idle, 134 #endif 135 #ifdef CONFIG_ARCH_USES_PG_ARCH_X 136 PG_arch_2, 137 PG_arch_3, 138 #endif 139 __NR_PAGEFLAGS, 140 141 PG_readahead = PG_reclaim, 142 143 /* 144 * Depending on the way an anonymous folio can be mapped into a page 145 * table (e.g., single PMD/PUD/CONT of the head page vs. PTE-mapped 146 * THP), PG_anon_exclusive may be set only for the head page or for 147 * tail pages of an anonymous folio. For now, we only expect it to be 148 * set on tail pages for PTE-mapped THP. 149 */ 150 PG_anon_exclusive = PG_mappedtodisk, 151 152 /* Filesystems */ 153 PG_checked = PG_owner_priv_1, 154 155 /* SwapBacked */ 156 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */ 157 158 /* Two page bits are conscripted by FS-Cache to maintain local caching 159 * state. These bits are set on pages belonging to the netfs's inodes 160 * when those inodes are being locally cached. 161 */ 162 PG_fscache = PG_private_2, /* page backed by cache */ 163 164 /* XEN */ 165 /* Pinned in Xen as a read-only pagetable page. */ 166 PG_pinned = PG_owner_priv_1, 167 /* Pinned as part of domain save (see xen_mm_pin_all()). */ 168 PG_savepinned = PG_dirty, 169 /* Has a grant mapping of another (foreign) domain's page. */ 170 PG_foreign = PG_owner_priv_1, 171 /* Remapped by swiotlb-xen. */ 172 PG_xen_remapped = PG_owner_priv_1, 173 174 /* non-lru isolated movable page */ 175 PG_isolated = PG_reclaim, 176 177 /* Only valid for buddy pages. Used to track pages that are reported */ 178 PG_reported = PG_uptodate, 179 180 #ifdef CONFIG_MEMORY_HOTPLUG 181 /* For self-hosted memmap pages */ 182 PG_vmemmap_self_hosted = PG_owner_priv_1, 183 #endif 184 185 /* 186 * Flags only valid for compound pages. Stored in first tail page's 187 * flags word. Cannot use the first 8 flags or any flag marked as 188 * PF_ANY. 189 */ 190 191 /* At least one page in this folio has the hwpoison flag set */ 192 PG_has_hwpoisoned = PG_error, 193 PG_hugetlb = PG_active, 194 PG_large_rmappable = PG_workingset, /* anon or file-backed */ 195 }; 196 197 #define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1) 198 199 #ifndef __GENERATING_BOUNDS_H 200 201 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 202 DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key); 203 204 /* 205 * Return the real head page struct iff the @page is a fake head page, otherwise 206 * return the @page itself. See Documentation/mm/vmemmap_dedup.rst. 207 */ 208 static __always_inline const struct page *page_fixed_fake_head(const struct page *page) 209 { 210 if (!static_branch_unlikely(&hugetlb_optimize_vmemmap_key)) 211 return page; 212 213 /* 214 * Only addresses aligned with PAGE_SIZE of struct page may be fake head 215 * struct page. The alignment check aims to avoid access the fields ( 216 * e.g. compound_head) of the @page[1]. It can avoid touch a (possibly) 217 * cold cacheline in some cases. 218 */ 219 if (IS_ALIGNED((unsigned long)page, PAGE_SIZE) && 220 test_bit(PG_head, &page->flags)) { 221 /* 222 * We can safely access the field of the @page[1] with PG_head 223 * because the @page is a compound page composed with at least 224 * two contiguous pages. 225 */ 226 unsigned long head = READ_ONCE(page[1].compound_head); 227 228 if (likely(head & 1)) 229 return (const struct page *)(head - 1); 230 } 231 return page; 232 } 233 #else 234 static inline const struct page *page_fixed_fake_head(const struct page *page) 235 { 236 return page; 237 } 238 #endif 239 240 static __always_inline int page_is_fake_head(struct page *page) 241 { 242 return page_fixed_fake_head(page) != page; 243 } 244 245 static inline unsigned long _compound_head(const struct page *page) 246 { 247 unsigned long head = READ_ONCE(page->compound_head); 248 249 if (unlikely(head & 1)) 250 return head - 1; 251 return (unsigned long)page_fixed_fake_head(page); 252 } 253 254 #define compound_head(page) ((typeof(page))_compound_head(page)) 255 256 /** 257 * page_folio - Converts from page to folio. 258 * @p: The page. 259 * 260 * Every page is part of a folio. This function cannot be called on a 261 * NULL pointer. 262 * 263 * Context: No reference, nor lock is required on @page. If the caller 264 * does not hold a reference, this call may race with a folio split, so 265 * it should re-check the folio still contains this page after gaining 266 * a reference on the folio. 267 * Return: The folio which contains this page. 268 */ 269 #define page_folio(p) (_Generic((p), \ 270 const struct page *: (const struct folio *)_compound_head(p), \ 271 struct page *: (struct folio *)_compound_head(p))) 272 273 /** 274 * folio_page - Return a page from a folio. 275 * @folio: The folio. 276 * @n: The page number to return. 277 * 278 * @n is relative to the start of the folio. This function does not 279 * check that the page number lies within @folio; the caller is presumed 280 * to have a reference to the page. 281 */ 282 #define folio_page(folio, n) nth_page(&(folio)->page, n) 283 284 static __always_inline int PageTail(struct page *page) 285 { 286 return READ_ONCE(page->compound_head) & 1 || page_is_fake_head(page); 287 } 288 289 static __always_inline int PageCompound(struct page *page) 290 { 291 return test_bit(PG_head, &page->flags) || 292 READ_ONCE(page->compound_head) & 1; 293 } 294 295 #define PAGE_POISON_PATTERN -1l 296 static inline int PagePoisoned(const struct page *page) 297 { 298 return READ_ONCE(page->flags) == PAGE_POISON_PATTERN; 299 } 300 301 #ifdef CONFIG_DEBUG_VM 302 void page_init_poison(struct page *page, size_t size); 303 #else 304 static inline void page_init_poison(struct page *page, size_t size) 305 { 306 } 307 #endif 308 309 static unsigned long *folio_flags(struct folio *folio, unsigned n) 310 { 311 struct page *page = &folio->page; 312 313 VM_BUG_ON_PGFLAGS(PageTail(page), page); 314 VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); 315 return &page[n].flags; 316 } 317 318 /* 319 * Page flags policies wrt compound pages 320 * 321 * PF_POISONED_CHECK 322 * check if this struct page poisoned/uninitialized 323 * 324 * PF_ANY: 325 * the page flag is relevant for small, head and tail pages. 326 * 327 * PF_HEAD: 328 * for compound page all operations related to the page flag applied to 329 * head page. 330 * 331 * PF_ONLY_HEAD: 332 * for compound page, callers only ever operate on the head page. 333 * 334 * PF_NO_TAIL: 335 * modifications of the page flag must be done on small or head pages, 336 * checks can be done on tail pages too. 337 * 338 * PF_NO_COMPOUND: 339 * the page flag is not relevant for compound pages. 340 * 341 * PF_SECOND: 342 * the page flag is stored in the first tail page. 343 */ 344 #define PF_POISONED_CHECK(page) ({ \ 345 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ 346 page; }) 347 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page) 348 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) 349 #define PF_ONLY_HEAD(page, enforce) ({ \ 350 VM_BUG_ON_PGFLAGS(PageTail(page), page); \ 351 PF_POISONED_CHECK(page); }) 352 #define PF_NO_TAIL(page, enforce) ({ \ 353 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ 354 PF_POISONED_CHECK(compound_head(page)); }) 355 #define PF_NO_COMPOUND(page, enforce) ({ \ 356 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ 357 PF_POISONED_CHECK(page); }) 358 #define PF_SECOND(page, enforce) ({ \ 359 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \ 360 PF_POISONED_CHECK(&page[1]); }) 361 362 /* Which page is the flag stored in */ 363 #define FOLIO_PF_ANY 0 364 #define FOLIO_PF_HEAD 0 365 #define FOLIO_PF_ONLY_HEAD 0 366 #define FOLIO_PF_NO_TAIL 0 367 #define FOLIO_PF_NO_COMPOUND 0 368 #define FOLIO_PF_SECOND 1 369 370 /* 371 * Macros to create function definitions for page flags 372 */ 373 #define TESTPAGEFLAG(uname, lname, policy) \ 374 static __always_inline bool folio_test_##lname(struct folio *folio) \ 375 { return test_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 376 static __always_inline int Page##uname(struct page *page) \ 377 { return test_bit(PG_##lname, &policy(page, 0)->flags); } 378 379 #define SETPAGEFLAG(uname, lname, policy) \ 380 static __always_inline \ 381 void folio_set_##lname(struct folio *folio) \ 382 { set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 383 static __always_inline void SetPage##uname(struct page *page) \ 384 { set_bit(PG_##lname, &policy(page, 1)->flags); } 385 386 #define CLEARPAGEFLAG(uname, lname, policy) \ 387 static __always_inline \ 388 void folio_clear_##lname(struct folio *folio) \ 389 { clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 390 static __always_inline void ClearPage##uname(struct page *page) \ 391 { clear_bit(PG_##lname, &policy(page, 1)->flags); } 392 393 #define __SETPAGEFLAG(uname, lname, policy) \ 394 static __always_inline \ 395 void __folio_set_##lname(struct folio *folio) \ 396 { __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 397 static __always_inline void __SetPage##uname(struct page *page) \ 398 { __set_bit(PG_##lname, &policy(page, 1)->flags); } 399 400 #define __CLEARPAGEFLAG(uname, lname, policy) \ 401 static __always_inline \ 402 void __folio_clear_##lname(struct folio *folio) \ 403 { __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 404 static __always_inline void __ClearPage##uname(struct page *page) \ 405 { __clear_bit(PG_##lname, &policy(page, 1)->flags); } 406 407 #define TESTSETFLAG(uname, lname, policy) \ 408 static __always_inline \ 409 bool folio_test_set_##lname(struct folio *folio) \ 410 { return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 411 static __always_inline int TestSetPage##uname(struct page *page) \ 412 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } 413 414 #define TESTCLEARFLAG(uname, lname, policy) \ 415 static __always_inline \ 416 bool folio_test_clear_##lname(struct folio *folio) \ 417 { return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \ 418 static __always_inline int TestClearPage##uname(struct page *page) \ 419 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } 420 421 #define PAGEFLAG(uname, lname, policy) \ 422 TESTPAGEFLAG(uname, lname, policy) \ 423 SETPAGEFLAG(uname, lname, policy) \ 424 CLEARPAGEFLAG(uname, lname, policy) 425 426 #define __PAGEFLAG(uname, lname, policy) \ 427 TESTPAGEFLAG(uname, lname, policy) \ 428 __SETPAGEFLAG(uname, lname, policy) \ 429 __CLEARPAGEFLAG(uname, lname, policy) 430 431 #define TESTSCFLAG(uname, lname, policy) \ 432 TESTSETFLAG(uname, lname, policy) \ 433 TESTCLEARFLAG(uname, lname, policy) 434 435 #define TESTPAGEFLAG_FALSE(uname, lname) \ 436 static inline bool folio_test_##lname(const struct folio *folio) { return false; } \ 437 static inline int Page##uname(const struct page *page) { return 0; } 438 439 #define SETPAGEFLAG_NOOP(uname, lname) \ 440 static inline void folio_set_##lname(struct folio *folio) { } \ 441 static inline void SetPage##uname(struct page *page) { } 442 443 #define CLEARPAGEFLAG_NOOP(uname, lname) \ 444 static inline void folio_clear_##lname(struct folio *folio) { } \ 445 static inline void ClearPage##uname(struct page *page) { } 446 447 #define __CLEARPAGEFLAG_NOOP(uname, lname) \ 448 static inline void __folio_clear_##lname(struct folio *folio) { } \ 449 static inline void __ClearPage##uname(struct page *page) { } 450 451 #define TESTSETFLAG_FALSE(uname, lname) \ 452 static inline bool folio_test_set_##lname(struct folio *folio) \ 453 { return 0; } \ 454 static inline int TestSetPage##uname(struct page *page) { return 0; } 455 456 #define TESTCLEARFLAG_FALSE(uname, lname) \ 457 static inline bool folio_test_clear_##lname(struct folio *folio) \ 458 { return 0; } \ 459 static inline int TestClearPage##uname(struct page *page) { return 0; } 460 461 #define PAGEFLAG_FALSE(uname, lname) TESTPAGEFLAG_FALSE(uname, lname) \ 462 SETPAGEFLAG_NOOP(uname, lname) CLEARPAGEFLAG_NOOP(uname, lname) 463 464 #define TESTSCFLAG_FALSE(uname, lname) \ 465 TESTSETFLAG_FALSE(uname, lname) TESTCLEARFLAG_FALSE(uname, lname) 466 467 __PAGEFLAG(Locked, locked, PF_NO_TAIL) 468 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) 469 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL) 470 PAGEFLAG(Referenced, referenced, PF_HEAD) 471 TESTCLEARFLAG(Referenced, referenced, PF_HEAD) 472 __SETPAGEFLAG(Referenced, referenced, PF_HEAD) 473 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) 474 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) 475 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) 476 TESTCLEARFLAG(LRU, lru, PF_HEAD) 477 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) 478 TESTCLEARFLAG(Active, active, PF_HEAD) 479 PAGEFLAG(Workingset, workingset, PF_HEAD) 480 TESTCLEARFLAG(Workingset, workingset, PF_HEAD) 481 __PAGEFLAG(Slab, slab, PF_NO_TAIL) 482 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ 483 484 /* Xen */ 485 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) 486 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) 487 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); 488 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); 489 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 490 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 491 492 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 493 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 494 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 495 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 496 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 497 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 498 499 /* 500 * Private page markings that may be used by the filesystem that owns the page 501 * for its own purposes. 502 * - PG_private and PG_private_2 cause release_folio() and co to be invoked 503 */ 504 PAGEFLAG(Private, private, PF_ANY) 505 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) 506 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 507 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 508 509 /* 510 * Only test-and-set exist for PG_writeback. The unconditional operators are 511 * risky: they bypass page accounting. 512 */ 513 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) 514 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) 515 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) 516 517 /* PG_readahead is only used for reads; PG_reclaim is only for writes */ 518 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) 519 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) 520 PAGEFLAG(Readahead, readahead, PF_NO_COMPOUND) 521 TESTCLEARFLAG(Readahead, readahead, PF_NO_COMPOUND) 522 523 #ifdef CONFIG_HIGHMEM 524 /* 525 * Must use a macro here due to header dependency issues. page_zone() is not 526 * available at this point. 527 */ 528 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) 529 #define folio_test_highmem(__f) is_highmem_idx(folio_zonenum(__f)) 530 #else 531 PAGEFLAG_FALSE(HighMem, highmem) 532 #endif 533 534 #ifdef CONFIG_SWAP 535 static __always_inline bool folio_test_swapcache(struct folio *folio) 536 { 537 return folio_test_swapbacked(folio) && 538 test_bit(PG_swapcache, folio_flags(folio, 0)); 539 } 540 541 static __always_inline bool PageSwapCache(struct page *page) 542 { 543 return folio_test_swapcache(page_folio(page)); 544 } 545 546 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 547 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 548 #else 549 PAGEFLAG_FALSE(SwapCache, swapcache) 550 #endif 551 552 PAGEFLAG(Unevictable, unevictable, PF_HEAD) 553 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) 554 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) 555 556 #ifdef CONFIG_MMU 557 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 558 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 559 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) 560 #else 561 PAGEFLAG_FALSE(Mlocked, mlocked) __CLEARPAGEFLAG_NOOP(Mlocked, mlocked) 562 TESTSCFLAG_FALSE(Mlocked, mlocked) 563 #endif 564 565 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 566 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) 567 #else 568 PAGEFLAG_FALSE(Uncached, uncached) 569 #endif 570 571 #ifdef CONFIG_MEMORY_FAILURE 572 PAGEFLAG(HWPoison, hwpoison, PF_ANY) 573 TESTSCFLAG(HWPoison, hwpoison, PF_ANY) 574 #define __PG_HWPOISON (1UL << PG_hwpoison) 575 #define MAGIC_HWPOISON 0x48575053U /* HWPS */ 576 extern void SetPageHWPoisonTakenOff(struct page *page); 577 extern void ClearPageHWPoisonTakenOff(struct page *page); 578 extern bool take_page_off_buddy(struct page *page); 579 extern bool put_page_back_buddy(struct page *page); 580 #else 581 PAGEFLAG_FALSE(HWPoison, hwpoison) 582 #define __PG_HWPOISON 0 583 #endif 584 585 #if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 586 TESTPAGEFLAG(Young, young, PF_ANY) 587 SETPAGEFLAG(Young, young, PF_ANY) 588 TESTCLEARFLAG(Young, young, PF_ANY) 589 PAGEFLAG(Idle, idle, PF_ANY) 590 #endif 591 592 /* 593 * PageReported() is used to track reported free pages within the Buddy 594 * allocator. We can use the non-atomic version of the test and set 595 * operations as both should be shielded with the zone lock to prevent 596 * any possible races on the setting or clearing of the bit. 597 */ 598 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND) 599 600 #ifdef CONFIG_MEMORY_HOTPLUG 601 PAGEFLAG(VmemmapSelfHosted, vmemmap_self_hosted, PF_ANY) 602 #else 603 PAGEFLAG_FALSE(VmemmapSelfHosted, vmemmap_self_hosted) 604 #endif 605 606 /* 607 * On an anonymous page mapped into a user virtual memory area, 608 * page->mapping points to its anon_vma, not to a struct address_space; 609 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 610 * 611 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 612 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON 613 * bit; and then page->mapping points, not to an anon_vma, but to a private 614 * structure which KSM associates with that merged page. See ksm.h. 615 * 616 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable 617 * page and then page->mapping points to a struct movable_operations. 618 * 619 * Please note that, confusingly, "page_mapping" refers to the inode 620 * address_space which maps the page from disk; whereas "page_mapped" 621 * refers to user virtual address space into which the page is mapped. 622 * 623 * For slab pages, since slab reuses the bits in struct page to store its 624 * internal states, the page->mapping does not exist as such, nor do these 625 * flags below. So in order to avoid testing non-existent bits, please 626 * make sure that PageSlab(page) actually evaluates to false before calling 627 * the following functions (e.g., PageAnon). See mm/slab.h. 628 */ 629 #define PAGE_MAPPING_ANON 0x1 630 #define PAGE_MAPPING_MOVABLE 0x2 631 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 632 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 633 634 /* 635 * Different with flags above, this flag is used only for fsdax mode. It 636 * indicates that this page->mapping is now under reflink case. 637 */ 638 #define PAGE_MAPPING_DAX_SHARED ((void *)0x1) 639 640 static __always_inline bool folio_mapping_flags(struct folio *folio) 641 { 642 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) != 0; 643 } 644 645 static __always_inline int PageMappingFlags(struct page *page) 646 { 647 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; 648 } 649 650 static __always_inline bool folio_test_anon(struct folio *folio) 651 { 652 return ((unsigned long)folio->mapping & PAGE_MAPPING_ANON) != 0; 653 } 654 655 static __always_inline bool PageAnon(struct page *page) 656 { 657 return folio_test_anon(page_folio(page)); 658 } 659 660 static __always_inline bool __folio_test_movable(const struct folio *folio) 661 { 662 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 663 PAGE_MAPPING_MOVABLE; 664 } 665 666 static __always_inline int __PageMovable(struct page *page) 667 { 668 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 669 PAGE_MAPPING_MOVABLE; 670 } 671 672 #ifdef CONFIG_KSM 673 /* 674 * A KSM page is one of those write-protected "shared pages" or "merged pages" 675 * which KSM maps into multiple mms, wherever identical anonymous page content 676 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any 677 * anon_vma, but to that page's node of the stable tree. 678 */ 679 static __always_inline bool folio_test_ksm(struct folio *folio) 680 { 681 return ((unsigned long)folio->mapping & PAGE_MAPPING_FLAGS) == 682 PAGE_MAPPING_KSM; 683 } 684 685 static __always_inline bool PageKsm(struct page *page) 686 { 687 return folio_test_ksm(page_folio(page)); 688 } 689 #else 690 TESTPAGEFLAG_FALSE(Ksm, ksm) 691 #endif 692 693 u64 stable_page_flags(struct page *page); 694 695 /** 696 * folio_xor_flags_has_waiters - Change some folio flags. 697 * @folio: The folio. 698 * @mask: Bits set in this word will be changed. 699 * 700 * This must only be used for flags which are changed with the folio 701 * lock held. For example, it is unsafe to use for PG_dirty as that 702 * can be set without the folio lock held. It can also only be used 703 * on flags which are in the range 0-6 as some of the implementations 704 * only affect those bits. 705 * 706 * Return: Whether there are tasks waiting on the folio. 707 */ 708 static inline bool folio_xor_flags_has_waiters(struct folio *folio, 709 unsigned long mask) 710 { 711 return xor_unlock_is_negative_byte(mask, folio_flags(folio, 0)); 712 } 713 714 /** 715 * folio_test_uptodate - Is this folio up to date? 716 * @folio: The folio. 717 * 718 * The uptodate flag is set on a folio when every byte in the folio is 719 * at least as new as the corresponding bytes on storage. Anonymous 720 * and CoW folios are always uptodate. If the folio is not uptodate, 721 * some of the bytes in it may be; see the is_partially_uptodate() 722 * address_space operation. 723 */ 724 static inline bool folio_test_uptodate(struct folio *folio) 725 { 726 bool ret = test_bit(PG_uptodate, folio_flags(folio, 0)); 727 /* 728 * Must ensure that the data we read out of the folio is loaded 729 * _after_ we've loaded folio->flags to check the uptodate bit. 730 * We can skip the barrier if the folio is not uptodate, because 731 * we wouldn't be reading anything from it. 732 * 733 * See folio_mark_uptodate() for the other side of the story. 734 */ 735 if (ret) 736 smp_rmb(); 737 738 return ret; 739 } 740 741 static inline int PageUptodate(struct page *page) 742 { 743 return folio_test_uptodate(page_folio(page)); 744 } 745 746 static __always_inline void __folio_mark_uptodate(struct folio *folio) 747 { 748 smp_wmb(); 749 __set_bit(PG_uptodate, folio_flags(folio, 0)); 750 } 751 752 static __always_inline void folio_mark_uptodate(struct folio *folio) 753 { 754 /* 755 * Memory barrier must be issued before setting the PG_uptodate bit, 756 * so that all previous stores issued in order to bring the folio 757 * uptodate are actually visible before folio_test_uptodate becomes true. 758 */ 759 smp_wmb(); 760 set_bit(PG_uptodate, folio_flags(folio, 0)); 761 } 762 763 static __always_inline void __SetPageUptodate(struct page *page) 764 { 765 __folio_mark_uptodate((struct folio *)page); 766 } 767 768 static __always_inline void SetPageUptodate(struct page *page) 769 { 770 folio_mark_uptodate((struct folio *)page); 771 } 772 773 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) 774 775 bool __folio_start_writeback(struct folio *folio, bool keep_write); 776 bool set_page_writeback(struct page *page); 777 778 #define folio_start_writeback(folio) \ 779 __folio_start_writeback(folio, false) 780 #define folio_start_writeback_keepwrite(folio) \ 781 __folio_start_writeback(folio, true) 782 783 static inline bool test_set_page_writeback(struct page *page) 784 { 785 return set_page_writeback(page); 786 } 787 788 static __always_inline bool folio_test_head(struct folio *folio) 789 { 790 return test_bit(PG_head, folio_flags(folio, FOLIO_PF_ANY)); 791 } 792 793 static __always_inline int PageHead(struct page *page) 794 { 795 PF_POISONED_CHECK(page); 796 return test_bit(PG_head, &page->flags) && !page_is_fake_head(page); 797 } 798 799 __SETPAGEFLAG(Head, head, PF_ANY) 800 __CLEARPAGEFLAG(Head, head, PF_ANY) 801 CLEARPAGEFLAG(Head, head, PF_ANY) 802 803 /** 804 * folio_test_large() - Does this folio contain more than one page? 805 * @folio: The folio to test. 806 * 807 * Return: True if the folio is larger than one page. 808 */ 809 static inline bool folio_test_large(struct folio *folio) 810 { 811 return folio_test_head(folio); 812 } 813 814 static __always_inline void set_compound_head(struct page *page, struct page *head) 815 { 816 WRITE_ONCE(page->compound_head, (unsigned long)head + 1); 817 } 818 819 static __always_inline void clear_compound_head(struct page *page) 820 { 821 WRITE_ONCE(page->compound_head, 0); 822 } 823 824 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 825 static inline void ClearPageCompound(struct page *page) 826 { 827 BUG_ON(!PageHead(page)); 828 ClearPageHead(page); 829 } 830 PAGEFLAG(LargeRmappable, large_rmappable, PF_SECOND) 831 #else 832 TESTPAGEFLAG_FALSE(LargeRmappable, large_rmappable) 833 #endif 834 835 #define PG_head_mask ((1UL << PG_head)) 836 837 #ifdef CONFIG_HUGETLB_PAGE 838 int PageHuge(struct page *page); 839 SETPAGEFLAG(HugeTLB, hugetlb, PF_SECOND) 840 CLEARPAGEFLAG(HugeTLB, hugetlb, PF_SECOND) 841 842 /** 843 * folio_test_hugetlb - Determine if the folio belongs to hugetlbfs 844 * @folio: The folio to test. 845 * 846 * Context: Any context. Caller should have a reference on the folio to 847 * prevent it from being turned into a tail page. 848 * Return: True for hugetlbfs folios, false for anon folios or folios 849 * belonging to other filesystems. 850 */ 851 static inline bool folio_test_hugetlb(struct folio *folio) 852 { 853 return folio_test_large(folio) && 854 test_bit(PG_hugetlb, folio_flags(folio, 1)); 855 } 856 #else 857 TESTPAGEFLAG_FALSE(Huge, hugetlb) 858 #endif 859 860 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 861 /* 862 * PageHuge() only returns true for hugetlbfs pages, but not for 863 * normal or transparent huge pages. 864 * 865 * PageTransHuge() returns true for both transparent huge and 866 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be 867 * called only in the core VM paths where hugetlbfs pages can't exist. 868 */ 869 static inline int PageTransHuge(struct page *page) 870 { 871 VM_BUG_ON_PAGE(PageTail(page), page); 872 return PageHead(page); 873 } 874 875 /* 876 * PageTransCompound returns true for both transparent huge pages 877 * and hugetlbfs pages, so it should only be called when it's known 878 * that hugetlbfs pages aren't involved. 879 */ 880 static inline int PageTransCompound(struct page *page) 881 { 882 return PageCompound(page); 883 } 884 885 /* 886 * PageTransTail returns true for both transparent huge pages 887 * and hugetlbfs pages, so it should only be called when it's known 888 * that hugetlbfs pages aren't involved. 889 */ 890 static inline int PageTransTail(struct page *page) 891 { 892 return PageTail(page); 893 } 894 #else 895 TESTPAGEFLAG_FALSE(TransHuge, transhuge) 896 TESTPAGEFLAG_FALSE(TransCompound, transcompound) 897 TESTPAGEFLAG_FALSE(TransCompoundMap, transcompoundmap) 898 TESTPAGEFLAG_FALSE(TransTail, transtail) 899 #endif 900 901 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_TRANSPARENT_HUGEPAGE) 902 /* 903 * PageHasHWPoisoned indicates that at least one subpage is hwpoisoned in the 904 * compound page. 905 * 906 * This flag is set by hwpoison handler. Cleared by THP split or free page. 907 */ 908 PAGEFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 909 TESTSCFLAG(HasHWPoisoned, has_hwpoisoned, PF_SECOND) 910 #else 911 PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 912 TESTSCFLAG_FALSE(HasHWPoisoned, has_hwpoisoned) 913 #endif 914 915 /* 916 * Check if a page is currently marked HWPoisoned. Note that this check is 917 * best effort only and inherently racy: there is no way to synchronize with 918 * failing hardware. 919 */ 920 static inline bool is_page_hwpoison(struct page *page) 921 { 922 if (PageHWPoison(page)) 923 return true; 924 return PageHuge(page) && PageHWPoison(compound_head(page)); 925 } 926 927 /* 928 * For pages that are never mapped to userspace (and aren't PageSlab), 929 * page_type may be used. Because it is initialised to -1, we invert the 930 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and 931 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and 932 * low bits so that an underflow or overflow of page_mapcount() won't be 933 * mistaken for a page type value. 934 */ 935 936 #define PAGE_TYPE_BASE 0xf0000000 937 /* Reserve 0x0000007f to catch underflows of page_mapcount */ 938 #define PAGE_MAPCOUNT_RESERVE -128 939 #define PG_buddy 0x00000080 940 #define PG_offline 0x00000100 941 #define PG_table 0x00000200 942 #define PG_guard 0x00000400 943 944 #define PageType(page, flag) \ 945 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) 946 #define folio_test_type(folio, flag) \ 947 ((folio->page.page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) 948 949 static inline int page_type_has_type(unsigned int page_type) 950 { 951 return (int)page_type < PAGE_MAPCOUNT_RESERVE; 952 } 953 954 static inline int page_has_type(struct page *page) 955 { 956 return page_type_has_type(page->page_type); 957 } 958 959 #define PAGE_TYPE_OPS(uname, lname, fname) \ 960 static __always_inline int Page##uname(const struct page *page) \ 961 { \ 962 return PageType(page, PG_##lname); \ 963 } \ 964 static __always_inline int folio_test_##fname(const struct folio *folio)\ 965 { \ 966 return folio_test_type(folio, PG_##lname); \ 967 } \ 968 static __always_inline void __SetPage##uname(struct page *page) \ 969 { \ 970 VM_BUG_ON_PAGE(!PageType(page, 0), page); \ 971 page->page_type &= ~PG_##lname; \ 972 } \ 973 static __always_inline void __folio_set_##fname(struct folio *folio) \ 974 { \ 975 VM_BUG_ON_FOLIO(!folio_test_type(folio, 0), folio); \ 976 folio->page.page_type &= ~PG_##lname; \ 977 } \ 978 static __always_inline void __ClearPage##uname(struct page *page) \ 979 { \ 980 VM_BUG_ON_PAGE(!Page##uname(page), page); \ 981 page->page_type |= PG_##lname; \ 982 } \ 983 static __always_inline void __folio_clear_##fname(struct folio *folio) \ 984 { \ 985 VM_BUG_ON_FOLIO(!folio_test_##fname(folio), folio); \ 986 folio->page.page_type |= PG_##lname; \ 987 } \ 988 989 /* 990 * PageBuddy() indicates that the page is free and in the buddy system 991 * (see mm/page_alloc.c). 992 */ 993 PAGE_TYPE_OPS(Buddy, buddy, buddy) 994 995 /* 996 * PageOffline() indicates that the page is logically offline although the 997 * containing section is online. (e.g. inflated in a balloon driver or 998 * not onlined when onlining the section). 999 * The content of these pages is effectively stale. Such pages should not 1000 * be touched (read/write/dump/save) except by their owner. 1001 * 1002 * If a driver wants to allow to offline unmovable PageOffline() pages without 1003 * putting them back to the buddy, it can do so via the memory notifier by 1004 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the 1005 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline() 1006 * pages (now with a reference count of zero) are treated like free pages, 1007 * allowing the containing memory block to get offlined. A driver that 1008 * relies on this feature is aware that re-onlining the memory block will 1009 * require to re-set the pages PageOffline() and not giving them to the 1010 * buddy via online_page_callback_t. 1011 * 1012 * There are drivers that mark a page PageOffline() and expect there won't be 1013 * any further access to page content. PFN walkers that read content of random 1014 * pages should check PageOffline() and synchronize with such drivers using 1015 * page_offline_freeze()/page_offline_thaw(). 1016 */ 1017 PAGE_TYPE_OPS(Offline, offline, offline) 1018 1019 extern void page_offline_freeze(void); 1020 extern void page_offline_thaw(void); 1021 extern void page_offline_begin(void); 1022 extern void page_offline_end(void); 1023 1024 /* 1025 * Marks pages in use as page tables. 1026 */ 1027 PAGE_TYPE_OPS(Table, table, pgtable) 1028 1029 /* 1030 * Marks guardpages used with debug_pagealloc. 1031 */ 1032 PAGE_TYPE_OPS(Guard, guard, guard) 1033 1034 extern bool is_free_buddy_page(struct page *page); 1035 1036 PAGEFLAG(Isolated, isolated, PF_ANY); 1037 1038 static __always_inline int PageAnonExclusive(struct page *page) 1039 { 1040 VM_BUG_ON_PGFLAGS(!PageAnon(page), page); 1041 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1042 return test_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1043 } 1044 1045 static __always_inline void SetPageAnonExclusive(struct page *page) 1046 { 1047 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page); 1048 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1049 set_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1050 } 1051 1052 static __always_inline void ClearPageAnonExclusive(struct page *page) 1053 { 1054 VM_BUG_ON_PGFLAGS(!PageAnon(page) || PageKsm(page), page); 1055 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1056 clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1057 } 1058 1059 static __always_inline void __ClearPageAnonExclusive(struct page *page) 1060 { 1061 VM_BUG_ON_PGFLAGS(!PageAnon(page), page); 1062 VM_BUG_ON_PGFLAGS(PageHuge(page) && !PageHead(page), page); 1063 __clear_bit(PG_anon_exclusive, &PF_ANY(page, 1)->flags); 1064 } 1065 1066 #ifdef CONFIG_MMU 1067 #define __PG_MLOCKED (1UL << PG_mlocked) 1068 #else 1069 #define __PG_MLOCKED 0 1070 #endif 1071 1072 /* 1073 * Flags checked when a page is freed. Pages being freed should not have 1074 * these flags set. If they are, there is a problem. 1075 */ 1076 #define PAGE_FLAGS_CHECK_AT_FREE \ 1077 (1UL << PG_lru | 1UL << PG_locked | \ 1078 1UL << PG_private | 1UL << PG_private_2 | \ 1079 1UL << PG_writeback | 1UL << PG_reserved | \ 1080 1UL << PG_slab | 1UL << PG_active | \ 1081 1UL << PG_unevictable | __PG_MLOCKED | LRU_GEN_MASK) 1082 1083 /* 1084 * Flags checked when a page is prepped for return by the page allocator. 1085 * Pages being prepped should not have these flags set. If they are set, 1086 * there has been a kernel bug or struct page corruption. 1087 * 1088 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's 1089 * alloc-free cycle to prevent from reusing the page. 1090 */ 1091 #define PAGE_FLAGS_CHECK_AT_PREP \ 1092 ((PAGEFLAGS_MASK & ~__PG_HWPOISON) | LRU_GEN_MASK | LRU_REFS_MASK) 1093 1094 /* 1095 * Flags stored in the second page of a compound page. They may overlap 1096 * the CHECK_AT_FREE flags above, so need to be cleared. 1097 */ 1098 #define PAGE_FLAGS_SECOND \ 1099 (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \ 1100 1UL << PG_hugetlb | 1UL << PG_large_rmappable) 1101 1102 #define PAGE_FLAGS_PRIVATE \ 1103 (1UL << PG_private | 1UL << PG_private_2) 1104 /** 1105 * page_has_private - Determine if page has private stuff 1106 * @page: The page to be checked 1107 * 1108 * Determine if a page has private stuff, indicating that release routines 1109 * should be invoked upon it. 1110 */ 1111 static inline int page_has_private(struct page *page) 1112 { 1113 return !!(page->flags & PAGE_FLAGS_PRIVATE); 1114 } 1115 1116 static inline bool folio_has_private(struct folio *folio) 1117 { 1118 return page_has_private(&folio->page); 1119 } 1120 1121 #undef PF_ANY 1122 #undef PF_HEAD 1123 #undef PF_ONLY_HEAD 1124 #undef PF_NO_TAIL 1125 #undef PF_NO_COMPOUND 1126 #undef PF_SECOND 1127 #endif /* !__GENERATING_BOUNDS_H */ 1128 1129 #endif /* PAGE_FLAGS_H */ 1130