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_uptodate tells whether the page's contents is valid. When a read 72 * completes, the page becomes uptodate, unless a disk I/O error happened. 73 * 74 * PG_referenced, PG_reclaim are used for page reclaim for anonymous and 75 * file-backed pagecache (see mm/vmscan.c). 76 * 77 * PG_error is set to indicate that an I/O error occurred on this page. 78 * 79 * PG_arch_1 is an architecture specific page state bit. The generic code 80 * guarantees that this bit is cleared for a page when it first is entered into 81 * the page cache. 82 * 83 * PG_hwpoison indicates that a page got corrupted in hardware and contains 84 * data with incorrect ECC bits that triggered a machine check. Accessing is 85 * not safe since it may cause another machine check. Don't touch! 86 */ 87 88 /* 89 * Don't use the pageflags directly. Use the PageFoo macros. 90 * 91 * The page flags field is split into two parts, the main flags area 92 * which extends from the low bits upwards, and the fields area which 93 * extends from the high bits downwards. 94 * 95 * | FIELD | ... | FLAGS | 96 * N-1 ^ 0 97 * (NR_PAGEFLAGS) 98 * 99 * The fields area is reserved for fields mapping zone, node (for NUMA) and 100 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like 101 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP). 102 */ 103 enum pageflags { 104 PG_locked, /* Page is locked. Don't touch. */ 105 PG_referenced, 106 PG_uptodate, 107 PG_dirty, 108 PG_lru, 109 PG_active, 110 PG_workingset, 111 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */ 112 PG_error, 113 PG_slab, 114 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/ 115 PG_arch_1, 116 PG_reserved, 117 PG_private, /* If pagecache, has fs-private data */ 118 PG_private_2, /* If pagecache, has fs aux data */ 119 PG_writeback, /* Page is under writeback */ 120 PG_head, /* A head page */ 121 PG_mappedtodisk, /* Has blocks allocated on-disk */ 122 PG_reclaim, /* To be reclaimed asap */ 123 PG_swapbacked, /* Page is backed by RAM/swap */ 124 PG_unevictable, /* Page is "unevictable" */ 125 #ifdef CONFIG_MMU 126 PG_mlocked, /* Page is vma mlocked */ 127 #endif 128 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 129 PG_uncached, /* Page has been mapped as uncached */ 130 #endif 131 #ifdef CONFIG_MEMORY_FAILURE 132 PG_hwpoison, /* hardware poisoned page. Don't touch */ 133 #endif 134 #if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT) 135 PG_young, 136 PG_idle, 137 #endif 138 #ifdef CONFIG_64BIT 139 PG_arch_2, 140 #endif 141 #ifdef CONFIG_KASAN_HW_TAGS 142 PG_skip_kasan_poison, 143 #endif 144 __NR_PAGEFLAGS, 145 146 /* Filesystems */ 147 PG_checked = PG_owner_priv_1, 148 149 /* SwapBacked */ 150 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */ 151 152 /* Two page bits are conscripted by FS-Cache to maintain local caching 153 * state. These bits are set on pages belonging to the netfs's inodes 154 * when those inodes are being locally cached. 155 */ 156 PG_fscache = PG_private_2, /* page backed by cache */ 157 158 /* XEN */ 159 /* Pinned in Xen as a read-only pagetable page. */ 160 PG_pinned = PG_owner_priv_1, 161 /* Pinned as part of domain save (see xen_mm_pin_all()). */ 162 PG_savepinned = PG_dirty, 163 /* Has a grant mapping of another (foreign) domain's page. */ 164 PG_foreign = PG_owner_priv_1, 165 /* Remapped by swiotlb-xen. */ 166 PG_xen_remapped = PG_owner_priv_1, 167 168 /* SLOB */ 169 PG_slob_free = PG_private, 170 171 /* Compound pages. Stored in first tail page's flags */ 172 PG_double_map = PG_workingset, 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 }; 180 181 #ifndef __GENERATING_BOUNDS_H 182 183 static inline unsigned long _compound_head(const struct page *page) 184 { 185 unsigned long head = READ_ONCE(page->compound_head); 186 187 if (unlikely(head & 1)) 188 return head - 1; 189 return (unsigned long)page; 190 } 191 192 #define compound_head(page) ((typeof(page))_compound_head(page)) 193 194 static __always_inline int PageTail(struct page *page) 195 { 196 return READ_ONCE(page->compound_head) & 1; 197 } 198 199 static __always_inline int PageCompound(struct page *page) 200 { 201 return test_bit(PG_head, &page->flags) || PageTail(page); 202 } 203 204 #define PAGE_POISON_PATTERN -1l 205 static inline int PagePoisoned(const struct page *page) 206 { 207 return page->flags == PAGE_POISON_PATTERN; 208 } 209 210 #ifdef CONFIG_DEBUG_VM 211 void page_init_poison(struct page *page, size_t size); 212 #else 213 static inline void page_init_poison(struct page *page, size_t size) 214 { 215 } 216 #endif 217 218 /* 219 * Page flags policies wrt compound pages 220 * 221 * PF_POISONED_CHECK 222 * check if this struct page poisoned/uninitialized 223 * 224 * PF_ANY: 225 * the page flag is relevant for small, head and tail pages. 226 * 227 * PF_HEAD: 228 * for compound page all operations related to the page flag applied to 229 * head page. 230 * 231 * PF_ONLY_HEAD: 232 * for compound page, callers only ever operate on the head page. 233 * 234 * PF_NO_TAIL: 235 * modifications of the page flag must be done on small or head pages, 236 * checks can be done on tail pages too. 237 * 238 * PF_NO_COMPOUND: 239 * the page flag is not relevant for compound pages. 240 * 241 * PF_SECOND: 242 * the page flag is stored in the first tail page. 243 */ 244 #define PF_POISONED_CHECK(page) ({ \ 245 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ 246 page; }) 247 #define PF_ANY(page, enforce) PF_POISONED_CHECK(page) 248 #define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) 249 #define PF_ONLY_HEAD(page, enforce) ({ \ 250 VM_BUG_ON_PGFLAGS(PageTail(page), page); \ 251 PF_POISONED_CHECK(page); }) 252 #define PF_NO_TAIL(page, enforce) ({ \ 253 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ 254 PF_POISONED_CHECK(compound_head(page)); }) 255 #define PF_NO_COMPOUND(page, enforce) ({ \ 256 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ 257 PF_POISONED_CHECK(page); }) 258 #define PF_SECOND(page, enforce) ({ \ 259 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \ 260 PF_POISONED_CHECK(&page[1]); }) 261 262 /* 263 * Macros to create function definitions for page flags 264 */ 265 #define TESTPAGEFLAG(uname, lname, policy) \ 266 static __always_inline int Page##uname(struct page *page) \ 267 { return test_bit(PG_##lname, &policy(page, 0)->flags); } 268 269 #define SETPAGEFLAG(uname, lname, policy) \ 270 static __always_inline void SetPage##uname(struct page *page) \ 271 { set_bit(PG_##lname, &policy(page, 1)->flags); } 272 273 #define CLEARPAGEFLAG(uname, lname, policy) \ 274 static __always_inline void ClearPage##uname(struct page *page) \ 275 { clear_bit(PG_##lname, &policy(page, 1)->flags); } 276 277 #define __SETPAGEFLAG(uname, lname, policy) \ 278 static __always_inline void __SetPage##uname(struct page *page) \ 279 { __set_bit(PG_##lname, &policy(page, 1)->flags); } 280 281 #define __CLEARPAGEFLAG(uname, lname, policy) \ 282 static __always_inline void __ClearPage##uname(struct page *page) \ 283 { __clear_bit(PG_##lname, &policy(page, 1)->flags); } 284 285 #define TESTSETFLAG(uname, lname, policy) \ 286 static __always_inline int TestSetPage##uname(struct page *page) \ 287 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } 288 289 #define TESTCLEARFLAG(uname, lname, policy) \ 290 static __always_inline int TestClearPage##uname(struct page *page) \ 291 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } 292 293 #define PAGEFLAG(uname, lname, policy) \ 294 TESTPAGEFLAG(uname, lname, policy) \ 295 SETPAGEFLAG(uname, lname, policy) \ 296 CLEARPAGEFLAG(uname, lname, policy) 297 298 #define __PAGEFLAG(uname, lname, policy) \ 299 TESTPAGEFLAG(uname, lname, policy) \ 300 __SETPAGEFLAG(uname, lname, policy) \ 301 __CLEARPAGEFLAG(uname, lname, policy) 302 303 #define TESTSCFLAG(uname, lname, policy) \ 304 TESTSETFLAG(uname, lname, policy) \ 305 TESTCLEARFLAG(uname, lname, policy) 306 307 #define TESTPAGEFLAG_FALSE(uname) \ 308 static inline int Page##uname(const struct page *page) { return 0; } 309 310 #define SETPAGEFLAG_NOOP(uname) \ 311 static inline void SetPage##uname(struct page *page) { } 312 313 #define CLEARPAGEFLAG_NOOP(uname) \ 314 static inline void ClearPage##uname(struct page *page) { } 315 316 #define __CLEARPAGEFLAG_NOOP(uname) \ 317 static inline void __ClearPage##uname(struct page *page) { } 318 319 #define TESTSETFLAG_FALSE(uname) \ 320 static inline int TestSetPage##uname(struct page *page) { return 0; } 321 322 #define TESTCLEARFLAG_FALSE(uname) \ 323 static inline int TestClearPage##uname(struct page *page) { return 0; } 324 325 #define PAGEFLAG_FALSE(uname) TESTPAGEFLAG_FALSE(uname) \ 326 SETPAGEFLAG_NOOP(uname) CLEARPAGEFLAG_NOOP(uname) 327 328 #define TESTSCFLAG_FALSE(uname) \ 329 TESTSETFLAG_FALSE(uname) TESTCLEARFLAG_FALSE(uname) 330 331 __PAGEFLAG(Locked, locked, PF_NO_TAIL) 332 PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) __CLEARPAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) 333 PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL) 334 PAGEFLAG(Referenced, referenced, PF_HEAD) 335 TESTCLEARFLAG(Referenced, referenced, PF_HEAD) 336 __SETPAGEFLAG(Referenced, referenced, PF_HEAD) 337 PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) 338 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) 339 PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) 340 TESTCLEARFLAG(LRU, lru, PF_HEAD) 341 PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) 342 TESTCLEARFLAG(Active, active, PF_HEAD) 343 PAGEFLAG(Workingset, workingset, PF_HEAD) 344 TESTCLEARFLAG(Workingset, workingset, PF_HEAD) 345 __PAGEFLAG(Slab, slab, PF_NO_TAIL) 346 __PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL) 347 PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ 348 349 /* Xen */ 350 PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) 351 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) 352 PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); 353 PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); 354 PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 355 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 356 357 PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 358 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 359 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 360 PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 361 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 362 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 363 364 /* 365 * Private page markings that may be used by the filesystem that owns the page 366 * for its own purposes. 367 * - PG_private and PG_private_2 cause releasepage() and co to be invoked 368 */ 369 PAGEFLAG(Private, private, PF_ANY) 370 PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) 371 PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 372 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 373 374 /* 375 * Only test-and-set exist for PG_writeback. The unconditional operators are 376 * risky: they bypass page accounting. 377 */ 378 TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) 379 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) 380 PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) 381 382 /* PG_readahead is only used for reads; PG_reclaim is only for writes */ 383 PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) 384 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) 385 PAGEFLAG(Readahead, reclaim, PF_NO_COMPOUND) 386 TESTCLEARFLAG(Readahead, reclaim, PF_NO_COMPOUND) 387 388 #ifdef CONFIG_HIGHMEM 389 /* 390 * Must use a macro here due to header dependency issues. page_zone() is not 391 * available at this point. 392 */ 393 #define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) 394 #else 395 PAGEFLAG_FALSE(HighMem) 396 #endif 397 398 #ifdef CONFIG_SWAP 399 static __always_inline int PageSwapCache(struct page *page) 400 { 401 #ifdef CONFIG_THP_SWAP 402 page = compound_head(page); 403 #endif 404 return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags); 405 406 } 407 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 408 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 409 #else 410 PAGEFLAG_FALSE(SwapCache) 411 #endif 412 413 PAGEFLAG(Unevictable, unevictable, PF_HEAD) 414 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) 415 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) 416 417 #ifdef CONFIG_MMU 418 PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 419 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 420 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) 421 #else 422 PAGEFLAG_FALSE(Mlocked) __CLEARPAGEFLAG_NOOP(Mlocked) 423 TESTSCFLAG_FALSE(Mlocked) 424 #endif 425 426 #ifdef CONFIG_ARCH_USES_PG_UNCACHED 427 PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) 428 #else 429 PAGEFLAG_FALSE(Uncached) 430 #endif 431 432 #ifdef CONFIG_MEMORY_FAILURE 433 PAGEFLAG(HWPoison, hwpoison, PF_ANY) 434 TESTSCFLAG(HWPoison, hwpoison, PF_ANY) 435 #define __PG_HWPOISON (1UL << PG_hwpoison) 436 extern bool take_page_off_buddy(struct page *page); 437 #else 438 PAGEFLAG_FALSE(HWPoison) 439 #define __PG_HWPOISON 0 440 #endif 441 442 #if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT) 443 TESTPAGEFLAG(Young, young, PF_ANY) 444 SETPAGEFLAG(Young, young, PF_ANY) 445 TESTCLEARFLAG(Young, young, PF_ANY) 446 PAGEFLAG(Idle, idle, PF_ANY) 447 #endif 448 449 #ifdef CONFIG_KASAN_HW_TAGS 450 PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD) 451 #else 452 PAGEFLAG_FALSE(SkipKASanPoison) 453 #endif 454 455 /* 456 * PageReported() is used to track reported free pages within the Buddy 457 * allocator. We can use the non-atomic version of the test and set 458 * operations as both should be shielded with the zone lock to prevent 459 * any possible races on the setting or clearing of the bit. 460 */ 461 __PAGEFLAG(Reported, reported, PF_NO_COMPOUND) 462 463 /* 464 * On an anonymous page mapped into a user virtual memory area, 465 * page->mapping points to its anon_vma, not to a struct address_space; 466 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 467 * 468 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 469 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON 470 * bit; and then page->mapping points, not to an anon_vma, but to a private 471 * structure which KSM associates with that merged page. See ksm.h. 472 * 473 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable 474 * page and then page->mapping points a struct address_space. 475 * 476 * Please note that, confusingly, "page_mapping" refers to the inode 477 * address_space which maps the page from disk; whereas "page_mapped" 478 * refers to user virtual address space into which the page is mapped. 479 */ 480 #define PAGE_MAPPING_ANON 0x1 481 #define PAGE_MAPPING_MOVABLE 0x2 482 #define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 483 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 484 485 static __always_inline int PageMappingFlags(struct page *page) 486 { 487 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; 488 } 489 490 static __always_inline int PageAnon(struct page *page) 491 { 492 page = compound_head(page); 493 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 494 } 495 496 static __always_inline int __PageMovable(struct page *page) 497 { 498 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 499 PAGE_MAPPING_MOVABLE; 500 } 501 502 #ifdef CONFIG_KSM 503 /* 504 * A KSM page is one of those write-protected "shared pages" or "merged pages" 505 * which KSM maps into multiple mms, wherever identical anonymous page content 506 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any 507 * anon_vma, but to that page's node of the stable tree. 508 */ 509 static __always_inline int PageKsm(struct page *page) 510 { 511 page = compound_head(page); 512 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 513 PAGE_MAPPING_KSM; 514 } 515 #else 516 TESTPAGEFLAG_FALSE(Ksm) 517 #endif 518 519 u64 stable_page_flags(struct page *page); 520 521 static inline int PageUptodate(struct page *page) 522 { 523 int ret; 524 page = compound_head(page); 525 ret = test_bit(PG_uptodate, &(page)->flags); 526 /* 527 * Must ensure that the data we read out of the page is loaded 528 * _after_ we've loaded page->flags to check for PageUptodate. 529 * We can skip the barrier if the page is not uptodate, because 530 * we wouldn't be reading anything from it. 531 * 532 * See SetPageUptodate() for the other side of the story. 533 */ 534 if (ret) 535 smp_rmb(); 536 537 return ret; 538 } 539 540 static __always_inline void __SetPageUptodate(struct page *page) 541 { 542 VM_BUG_ON_PAGE(PageTail(page), page); 543 smp_wmb(); 544 __set_bit(PG_uptodate, &page->flags); 545 } 546 547 static __always_inline void SetPageUptodate(struct page *page) 548 { 549 VM_BUG_ON_PAGE(PageTail(page), page); 550 /* 551 * Memory barrier must be issued before setting the PG_uptodate bit, 552 * so that all previous stores issued in order to bring the page 553 * uptodate are actually visible before PageUptodate becomes true. 554 */ 555 smp_wmb(); 556 set_bit(PG_uptodate, &page->flags); 557 } 558 559 CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) 560 561 int test_clear_page_writeback(struct page *page); 562 int __test_set_page_writeback(struct page *page, bool keep_write); 563 564 #define test_set_page_writeback(page) \ 565 __test_set_page_writeback(page, false) 566 #define test_set_page_writeback_keepwrite(page) \ 567 __test_set_page_writeback(page, true) 568 569 static inline void set_page_writeback(struct page *page) 570 { 571 test_set_page_writeback(page); 572 } 573 574 static inline void set_page_writeback_keepwrite(struct page *page) 575 { 576 test_set_page_writeback_keepwrite(page); 577 } 578 579 __PAGEFLAG(Head, head, PF_ANY) CLEARPAGEFLAG(Head, head, PF_ANY) 580 581 static __always_inline void set_compound_head(struct page *page, struct page *head) 582 { 583 WRITE_ONCE(page->compound_head, (unsigned long)head + 1); 584 } 585 586 static __always_inline void clear_compound_head(struct page *page) 587 { 588 WRITE_ONCE(page->compound_head, 0); 589 } 590 591 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 592 static inline void ClearPageCompound(struct page *page) 593 { 594 BUG_ON(!PageHead(page)); 595 ClearPageHead(page); 596 } 597 #endif 598 599 #define PG_head_mask ((1UL << PG_head)) 600 601 #ifdef CONFIG_HUGETLB_PAGE 602 int PageHuge(struct page *page); 603 int PageHeadHuge(struct page *page); 604 #else 605 TESTPAGEFLAG_FALSE(Huge) 606 TESTPAGEFLAG_FALSE(HeadHuge) 607 #endif 608 609 610 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 611 /* 612 * PageHuge() only returns true for hugetlbfs pages, but not for 613 * normal or transparent huge pages. 614 * 615 * PageTransHuge() returns true for both transparent huge and 616 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be 617 * called only in the core VM paths where hugetlbfs pages can't exist. 618 */ 619 static inline int PageTransHuge(struct page *page) 620 { 621 VM_BUG_ON_PAGE(PageTail(page), page); 622 return PageHead(page); 623 } 624 625 /* 626 * PageTransCompound returns true for both transparent huge pages 627 * and hugetlbfs pages, so it should only be called when it's known 628 * that hugetlbfs pages aren't involved. 629 */ 630 static inline int PageTransCompound(struct page *page) 631 { 632 return PageCompound(page); 633 } 634 635 /* 636 * PageTransCompoundMap is the same as PageTransCompound, but it also 637 * guarantees the primary MMU has the entire compound page mapped 638 * through pmd_trans_huge, which in turn guarantees the secondary MMUs 639 * can also map the entire compound page. This allows the secondary 640 * MMUs to call get_user_pages() only once for each compound page and 641 * to immediately map the entire compound page with a single secondary 642 * MMU fault. If there will be a pmd split later, the secondary MMUs 643 * will get an update through the MMU notifier invalidation through 644 * split_huge_pmd(). 645 * 646 * Unlike PageTransCompound, this is safe to be called only while 647 * split_huge_pmd() cannot run from under us, like if protected by the 648 * MMU notifier, otherwise it may result in page->_mapcount check false 649 * positives. 650 * 651 * We have to treat page cache THP differently since every subpage of it 652 * would get _mapcount inc'ed once it is PMD mapped. But, it may be PTE 653 * mapped in the current process so comparing subpage's _mapcount to 654 * compound_mapcount to filter out PTE mapped case. 655 */ 656 static inline int PageTransCompoundMap(struct page *page) 657 { 658 struct page *head; 659 660 if (!PageTransCompound(page)) 661 return 0; 662 663 if (PageAnon(page)) 664 return atomic_read(&page->_mapcount) < 0; 665 666 head = compound_head(page); 667 /* File THP is PMD mapped and not PTE mapped */ 668 return atomic_read(&page->_mapcount) == 669 atomic_read(compound_mapcount_ptr(head)); 670 } 671 672 /* 673 * PageTransTail returns true for both transparent huge pages 674 * and hugetlbfs pages, so it should only be called when it's known 675 * that hugetlbfs pages aren't involved. 676 */ 677 static inline int PageTransTail(struct page *page) 678 { 679 return PageTail(page); 680 } 681 682 /* 683 * PageDoubleMap indicates that the compound page is mapped with PTEs as well 684 * as PMDs. 685 * 686 * This is required for optimization of rmap operations for THP: we can postpone 687 * per small page mapcount accounting (and its overhead from atomic operations) 688 * until the first PMD split. 689 * 690 * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up 691 * by one. This reference will go away with last compound_mapcount. 692 * 693 * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap(). 694 */ 695 PAGEFLAG(DoubleMap, double_map, PF_SECOND) 696 TESTSCFLAG(DoubleMap, double_map, PF_SECOND) 697 #else 698 TESTPAGEFLAG_FALSE(TransHuge) 699 TESTPAGEFLAG_FALSE(TransCompound) 700 TESTPAGEFLAG_FALSE(TransCompoundMap) 701 TESTPAGEFLAG_FALSE(TransTail) 702 PAGEFLAG_FALSE(DoubleMap) 703 TESTSCFLAG_FALSE(DoubleMap) 704 #endif 705 706 /* 707 * Check if a page is currently marked HWPoisoned. Note that this check is 708 * best effort only and inherently racy: there is no way to synchronize with 709 * failing hardware. 710 */ 711 static inline bool is_page_hwpoison(struct page *page) 712 { 713 if (PageHWPoison(page)) 714 return true; 715 return PageHuge(page) && PageHWPoison(compound_head(page)); 716 } 717 718 /* 719 * For pages that are never mapped to userspace (and aren't PageSlab), 720 * page_type may be used. Because it is initialised to -1, we invert the 721 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and 722 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and 723 * low bits so that an underflow or overflow of page_mapcount() won't be 724 * mistaken for a page type value. 725 */ 726 727 #define PAGE_TYPE_BASE 0xf0000000 728 /* Reserve 0x0000007f to catch underflows of page_mapcount */ 729 #define PAGE_MAPCOUNT_RESERVE -128 730 #define PG_buddy 0x00000080 731 #define PG_offline 0x00000100 732 #define PG_table 0x00000200 733 #define PG_guard 0x00000400 734 735 #define PageType(page, flag) \ 736 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) 737 738 static inline int page_has_type(struct page *page) 739 { 740 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE; 741 } 742 743 #define PAGE_TYPE_OPS(uname, lname) \ 744 static __always_inline int Page##uname(struct page *page) \ 745 { \ 746 return PageType(page, PG_##lname); \ 747 } \ 748 static __always_inline void __SetPage##uname(struct page *page) \ 749 { \ 750 VM_BUG_ON_PAGE(!PageType(page, 0), page); \ 751 page->page_type &= ~PG_##lname; \ 752 } \ 753 static __always_inline void __ClearPage##uname(struct page *page) \ 754 { \ 755 VM_BUG_ON_PAGE(!Page##uname(page), page); \ 756 page->page_type |= PG_##lname; \ 757 } 758 759 /* 760 * PageBuddy() indicates that the page is free and in the buddy system 761 * (see mm/page_alloc.c). 762 */ 763 PAGE_TYPE_OPS(Buddy, buddy) 764 765 /* 766 * PageOffline() indicates that the page is logically offline although the 767 * containing section is online. (e.g. inflated in a balloon driver or 768 * not onlined when onlining the section). 769 * The content of these pages is effectively stale. Such pages should not 770 * be touched (read/write/dump/save) except by their owner. 771 * 772 * If a driver wants to allow to offline unmovable PageOffline() pages without 773 * putting them back to the buddy, it can do so via the memory notifier by 774 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the 775 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline() 776 * pages (now with a reference count of zero) are treated like free pages, 777 * allowing the containing memory block to get offlined. A driver that 778 * relies on this feature is aware that re-onlining the memory block will 779 * require to re-set the pages PageOffline() and not giving them to the 780 * buddy via online_page_callback_t. 781 * 782 * There are drivers that mark a page PageOffline() and expect there won't be 783 * any further access to page content. PFN walkers that read content of random 784 * pages should check PageOffline() and synchronize with such drivers using 785 * page_offline_freeze()/page_offline_thaw(). 786 */ 787 PAGE_TYPE_OPS(Offline, offline) 788 789 extern void page_offline_freeze(void); 790 extern void page_offline_thaw(void); 791 extern void page_offline_begin(void); 792 extern void page_offline_end(void); 793 794 /* 795 * Marks pages in use as page tables. 796 */ 797 PAGE_TYPE_OPS(Table, table) 798 799 /* 800 * Marks guardpages used with debug_pagealloc. 801 */ 802 PAGE_TYPE_OPS(Guard, guard) 803 804 extern bool is_free_buddy_page(struct page *page); 805 806 __PAGEFLAG(Isolated, isolated, PF_ANY); 807 808 /* 809 * If network-based swap is enabled, sl*b must keep track of whether pages 810 * were allocated from pfmemalloc reserves. 811 */ 812 static inline int PageSlabPfmemalloc(struct page *page) 813 { 814 VM_BUG_ON_PAGE(!PageSlab(page), page); 815 return PageActive(page); 816 } 817 818 static inline void SetPageSlabPfmemalloc(struct page *page) 819 { 820 VM_BUG_ON_PAGE(!PageSlab(page), page); 821 SetPageActive(page); 822 } 823 824 static inline void __ClearPageSlabPfmemalloc(struct page *page) 825 { 826 VM_BUG_ON_PAGE(!PageSlab(page), page); 827 __ClearPageActive(page); 828 } 829 830 static inline void ClearPageSlabPfmemalloc(struct page *page) 831 { 832 VM_BUG_ON_PAGE(!PageSlab(page), page); 833 ClearPageActive(page); 834 } 835 836 #ifdef CONFIG_MMU 837 #define __PG_MLOCKED (1UL << PG_mlocked) 838 #else 839 #define __PG_MLOCKED 0 840 #endif 841 842 /* 843 * Flags checked when a page is freed. Pages being freed should not have 844 * these flags set. If they are, there is a problem. 845 */ 846 #define PAGE_FLAGS_CHECK_AT_FREE \ 847 (1UL << PG_lru | 1UL << PG_locked | \ 848 1UL << PG_private | 1UL << PG_private_2 | \ 849 1UL << PG_writeback | 1UL << PG_reserved | \ 850 1UL << PG_slab | 1UL << PG_active | \ 851 1UL << PG_unevictable | __PG_MLOCKED) 852 853 /* 854 * Flags checked when a page is prepped for return by the page allocator. 855 * Pages being prepped should not have these flags set. If they are set, 856 * there has been a kernel bug or struct page corruption. 857 * 858 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's 859 * alloc-free cycle to prevent from reusing the page. 860 */ 861 #define PAGE_FLAGS_CHECK_AT_PREP \ 862 (((1UL << NR_PAGEFLAGS) - 1) & ~__PG_HWPOISON) 863 864 #define PAGE_FLAGS_PRIVATE \ 865 (1UL << PG_private | 1UL << PG_private_2) 866 /** 867 * page_has_private - Determine if page has private stuff 868 * @page: The page to be checked 869 * 870 * Determine if a page has private stuff, indicating that release routines 871 * should be invoked upon it. 872 */ 873 static inline int page_has_private(struct page *page) 874 { 875 return !!(page->flags & PAGE_FLAGS_PRIVATE); 876 } 877 878 #undef PF_ANY 879 #undef PF_HEAD 880 #undef PF_ONLY_HEAD 881 #undef PF_NO_TAIL 882 #undef PF_NO_COMPOUND 883 #undef PF_SECOND 884 #endif /* !__GENERATING_BOUNDS_H */ 885 886 #endif /* PAGE_FLAGS_H */ 887