Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
fork

Configure Feed

Select the types of activity you want to include in your feed.

at v5.15-rc6 860 lines 28 kB view raw
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 */ 103enum 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_PAGE_IDLE_FLAG) && 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#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1) 182 183#ifndef __GENERATING_BOUNDS_H 184 185static inline unsigned long _compound_head(const struct page *page) 186{ 187 unsigned long head = READ_ONCE(page->compound_head); 188 189 if (unlikely(head & 1)) 190 return head - 1; 191 return (unsigned long)page; 192} 193 194#define compound_head(page) ((typeof(page))_compound_head(page)) 195 196static __always_inline int PageTail(struct page *page) 197{ 198 return READ_ONCE(page->compound_head) & 1; 199} 200 201static __always_inline int PageCompound(struct page *page) 202{ 203 return test_bit(PG_head, &page->flags) || PageTail(page); 204} 205 206#define PAGE_POISON_PATTERN -1l 207static inline int PagePoisoned(const struct page *page) 208{ 209 return page->flags == PAGE_POISON_PATTERN; 210} 211 212#ifdef CONFIG_DEBUG_VM 213void page_init_poison(struct page *page, size_t size); 214#else 215static inline void page_init_poison(struct page *page, size_t size) 216{ 217} 218#endif 219 220/* 221 * Page flags policies wrt compound pages 222 * 223 * PF_POISONED_CHECK 224 * check if this struct page poisoned/uninitialized 225 * 226 * PF_ANY: 227 * the page flag is relevant for small, head and tail pages. 228 * 229 * PF_HEAD: 230 * for compound page all operations related to the page flag applied to 231 * head page. 232 * 233 * PF_ONLY_HEAD: 234 * for compound page, callers only ever operate on the head page. 235 * 236 * PF_NO_TAIL: 237 * modifications of the page flag must be done on small or head pages, 238 * checks can be done on tail pages too. 239 * 240 * PF_NO_COMPOUND: 241 * the page flag is not relevant for compound pages. 242 * 243 * PF_SECOND: 244 * the page flag is stored in the first tail page. 245 */ 246#define PF_POISONED_CHECK(page) ({ \ 247 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \ 248 page; }) 249#define PF_ANY(page, enforce) PF_POISONED_CHECK(page) 250#define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page)) 251#define PF_ONLY_HEAD(page, enforce) ({ \ 252 VM_BUG_ON_PGFLAGS(PageTail(page), page); \ 253 PF_POISONED_CHECK(page); }) 254#define PF_NO_TAIL(page, enforce) ({ \ 255 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \ 256 PF_POISONED_CHECK(compound_head(page)); }) 257#define PF_NO_COMPOUND(page, enforce) ({ \ 258 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \ 259 PF_POISONED_CHECK(page); }) 260#define PF_SECOND(page, enforce) ({ \ 261 VM_BUG_ON_PGFLAGS(!PageHead(page), page); \ 262 PF_POISONED_CHECK(&page[1]); }) 263 264/* 265 * Macros to create function definitions for page flags 266 */ 267#define TESTPAGEFLAG(uname, lname, policy) \ 268static __always_inline int Page##uname(struct page *page) \ 269 { return test_bit(PG_##lname, &policy(page, 0)->flags); } 270 271#define SETPAGEFLAG(uname, lname, policy) \ 272static __always_inline void SetPage##uname(struct page *page) \ 273 { set_bit(PG_##lname, &policy(page, 1)->flags); } 274 275#define CLEARPAGEFLAG(uname, lname, policy) \ 276static __always_inline void ClearPage##uname(struct page *page) \ 277 { clear_bit(PG_##lname, &policy(page, 1)->flags); } 278 279#define __SETPAGEFLAG(uname, lname, policy) \ 280static __always_inline void __SetPage##uname(struct page *page) \ 281 { __set_bit(PG_##lname, &policy(page, 1)->flags); } 282 283#define __CLEARPAGEFLAG(uname, lname, policy) \ 284static __always_inline void __ClearPage##uname(struct page *page) \ 285 { __clear_bit(PG_##lname, &policy(page, 1)->flags); } 286 287#define TESTSETFLAG(uname, lname, policy) \ 288static __always_inline int TestSetPage##uname(struct page *page) \ 289 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); } 290 291#define TESTCLEARFLAG(uname, lname, policy) \ 292static __always_inline int TestClearPage##uname(struct page *page) \ 293 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); } 294 295#define PAGEFLAG(uname, lname, policy) \ 296 TESTPAGEFLAG(uname, lname, policy) \ 297 SETPAGEFLAG(uname, lname, policy) \ 298 CLEARPAGEFLAG(uname, lname, policy) 299 300#define __PAGEFLAG(uname, lname, policy) \ 301 TESTPAGEFLAG(uname, lname, policy) \ 302 __SETPAGEFLAG(uname, lname, policy) \ 303 __CLEARPAGEFLAG(uname, lname, policy) 304 305#define TESTSCFLAG(uname, lname, policy) \ 306 TESTSETFLAG(uname, lname, policy) \ 307 TESTCLEARFLAG(uname, lname, policy) 308 309#define TESTPAGEFLAG_FALSE(uname) \ 310static inline int Page##uname(const struct page *page) { return 0; } 311 312#define SETPAGEFLAG_NOOP(uname) \ 313static inline void SetPage##uname(struct page *page) { } 314 315#define CLEARPAGEFLAG_NOOP(uname) \ 316static inline void ClearPage##uname(struct page *page) { } 317 318#define __CLEARPAGEFLAG_NOOP(uname) \ 319static inline void __ClearPage##uname(struct page *page) { } 320 321#define TESTSETFLAG_FALSE(uname) \ 322static inline int TestSetPage##uname(struct page *page) { return 0; } 323 324#define TESTCLEARFLAG_FALSE(uname) \ 325static inline int TestClearPage##uname(struct page *page) { return 0; } 326 327#define PAGEFLAG_FALSE(uname) TESTPAGEFLAG_FALSE(uname) \ 328 SETPAGEFLAG_NOOP(uname) CLEARPAGEFLAG_NOOP(uname) 329 330#define TESTSCFLAG_FALSE(uname) \ 331 TESTSETFLAG_FALSE(uname) TESTCLEARFLAG_FALSE(uname) 332 333__PAGEFLAG(Locked, locked, PF_NO_TAIL) 334PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) __CLEARPAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) 335PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL) 336PAGEFLAG(Referenced, referenced, PF_HEAD) 337 TESTCLEARFLAG(Referenced, referenced, PF_HEAD) 338 __SETPAGEFLAG(Referenced, referenced, PF_HEAD) 339PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD) 340 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD) 341PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD) 342 TESTCLEARFLAG(LRU, lru, PF_HEAD) 343PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD) 344 TESTCLEARFLAG(Active, active, PF_HEAD) 345PAGEFLAG(Workingset, workingset, PF_HEAD) 346 TESTCLEARFLAG(Workingset, workingset, PF_HEAD) 347__PAGEFLAG(Slab, slab, PF_NO_TAIL) 348__PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL) 349PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */ 350 351/* Xen */ 352PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND) 353 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND) 354PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND); 355PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND); 356PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 357 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND) 358 359PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 360 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 361 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND) 362PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 363 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 364 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL) 365 366/* 367 * Private page markings that may be used by the filesystem that owns the page 368 * for its own purposes. 369 * - PG_private and PG_private_2 cause releasepage() and co to be invoked 370 */ 371PAGEFLAG(Private, private, PF_ANY) 372PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY) 373PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 374 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY) 375 376/* 377 * Only test-and-set exist for PG_writeback. The unconditional operators are 378 * risky: they bypass page accounting. 379 */ 380TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL) 381 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL) 382PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL) 383 384/* PG_readahead is only used for reads; PG_reclaim is only for writes */ 385PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL) 386 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL) 387PAGEFLAG(Readahead, reclaim, PF_NO_COMPOUND) 388 TESTCLEARFLAG(Readahead, reclaim, PF_NO_COMPOUND) 389 390#ifdef CONFIG_HIGHMEM 391/* 392 * Must use a macro here due to header dependency issues. page_zone() is not 393 * available at this point. 394 */ 395#define PageHighMem(__p) is_highmem_idx(page_zonenum(__p)) 396#else 397PAGEFLAG_FALSE(HighMem) 398#endif 399 400#ifdef CONFIG_SWAP 401static __always_inline int PageSwapCache(struct page *page) 402{ 403#ifdef CONFIG_THP_SWAP 404 page = compound_head(page); 405#endif 406 return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags); 407 408} 409SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 410CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL) 411#else 412PAGEFLAG_FALSE(SwapCache) 413#endif 414 415PAGEFLAG(Unevictable, unevictable, PF_HEAD) 416 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD) 417 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD) 418 419#ifdef CONFIG_MMU 420PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 421 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL) 422 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL) 423#else 424PAGEFLAG_FALSE(Mlocked) __CLEARPAGEFLAG_NOOP(Mlocked) 425 TESTSCFLAG_FALSE(Mlocked) 426#endif 427 428#ifdef CONFIG_ARCH_USES_PG_UNCACHED 429PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND) 430#else 431PAGEFLAG_FALSE(Uncached) 432#endif 433 434#ifdef CONFIG_MEMORY_FAILURE 435PAGEFLAG(HWPoison, hwpoison, PF_ANY) 436TESTSCFLAG(HWPoison, hwpoison, PF_ANY) 437#define __PG_HWPOISON (1UL << PG_hwpoison) 438extern bool take_page_off_buddy(struct page *page); 439#else 440PAGEFLAG_FALSE(HWPoison) 441#define __PG_HWPOISON 0 442#endif 443 444#if defined(CONFIG_PAGE_IDLE_FLAG) && defined(CONFIG_64BIT) 445TESTPAGEFLAG(Young, young, PF_ANY) 446SETPAGEFLAG(Young, young, PF_ANY) 447TESTCLEARFLAG(Young, young, PF_ANY) 448PAGEFLAG(Idle, idle, PF_ANY) 449#endif 450 451#ifdef CONFIG_KASAN_HW_TAGS 452PAGEFLAG(SkipKASanPoison, skip_kasan_poison, PF_HEAD) 453#else 454PAGEFLAG_FALSE(SkipKASanPoison) 455#endif 456 457/* 458 * PageReported() is used to track reported free pages within the Buddy 459 * allocator. We can use the non-atomic version of the test and set 460 * operations as both should be shielded with the zone lock to prevent 461 * any possible races on the setting or clearing of the bit. 462 */ 463__PAGEFLAG(Reported, reported, PF_NO_COMPOUND) 464 465/* 466 * On an anonymous page mapped into a user virtual memory area, 467 * page->mapping points to its anon_vma, not to a struct address_space; 468 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 469 * 470 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 471 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON 472 * bit; and then page->mapping points, not to an anon_vma, but to a private 473 * structure which KSM associates with that merged page. See ksm.h. 474 * 475 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable 476 * page and then page->mapping points a struct address_space. 477 * 478 * Please note that, confusingly, "page_mapping" refers to the inode 479 * address_space which maps the page from disk; whereas "page_mapped" 480 * refers to user virtual address space into which the page is mapped. 481 */ 482#define PAGE_MAPPING_ANON 0x1 483#define PAGE_MAPPING_MOVABLE 0x2 484#define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 485#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE) 486 487static __always_inline int PageMappingFlags(struct page *page) 488{ 489 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0; 490} 491 492static __always_inline int PageAnon(struct page *page) 493{ 494 page = compound_head(page); 495 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 496} 497 498static __always_inline int __PageMovable(struct page *page) 499{ 500 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 501 PAGE_MAPPING_MOVABLE; 502} 503 504#ifdef CONFIG_KSM 505/* 506 * A KSM page is one of those write-protected "shared pages" or "merged pages" 507 * which KSM maps into multiple mms, wherever identical anonymous page content 508 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any 509 * anon_vma, but to that page's node of the stable tree. 510 */ 511static __always_inline int PageKsm(struct page *page) 512{ 513 page = compound_head(page); 514 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) == 515 PAGE_MAPPING_KSM; 516} 517#else 518TESTPAGEFLAG_FALSE(Ksm) 519#endif 520 521u64 stable_page_flags(struct page *page); 522 523static inline int PageUptodate(struct page *page) 524{ 525 int ret; 526 page = compound_head(page); 527 ret = test_bit(PG_uptodate, &(page)->flags); 528 /* 529 * Must ensure that the data we read out of the page is loaded 530 * _after_ we've loaded page->flags to check for PageUptodate. 531 * We can skip the barrier if the page is not uptodate, because 532 * we wouldn't be reading anything from it. 533 * 534 * See SetPageUptodate() for the other side of the story. 535 */ 536 if (ret) 537 smp_rmb(); 538 539 return ret; 540} 541 542static __always_inline void __SetPageUptodate(struct page *page) 543{ 544 VM_BUG_ON_PAGE(PageTail(page), page); 545 smp_wmb(); 546 __set_bit(PG_uptodate, &page->flags); 547} 548 549static __always_inline void SetPageUptodate(struct page *page) 550{ 551 VM_BUG_ON_PAGE(PageTail(page), page); 552 /* 553 * Memory barrier must be issued before setting the PG_uptodate bit, 554 * so that all previous stores issued in order to bring the page 555 * uptodate are actually visible before PageUptodate becomes true. 556 */ 557 smp_wmb(); 558 set_bit(PG_uptodate, &page->flags); 559} 560 561CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL) 562 563int test_clear_page_writeback(struct page *page); 564int __test_set_page_writeback(struct page *page, bool keep_write); 565 566#define test_set_page_writeback(page) \ 567 __test_set_page_writeback(page, false) 568#define test_set_page_writeback_keepwrite(page) \ 569 __test_set_page_writeback(page, true) 570 571static inline void set_page_writeback(struct page *page) 572{ 573 test_set_page_writeback(page); 574} 575 576static inline void set_page_writeback_keepwrite(struct page *page) 577{ 578 test_set_page_writeback_keepwrite(page); 579} 580 581__PAGEFLAG(Head, head, PF_ANY) CLEARPAGEFLAG(Head, head, PF_ANY) 582 583static __always_inline void set_compound_head(struct page *page, struct page *head) 584{ 585 WRITE_ONCE(page->compound_head, (unsigned long)head + 1); 586} 587 588static __always_inline void clear_compound_head(struct page *page) 589{ 590 WRITE_ONCE(page->compound_head, 0); 591} 592 593#ifdef CONFIG_TRANSPARENT_HUGEPAGE 594static inline void ClearPageCompound(struct page *page) 595{ 596 BUG_ON(!PageHead(page)); 597 ClearPageHead(page); 598} 599#endif 600 601#define PG_head_mask ((1UL << PG_head)) 602 603#ifdef CONFIG_HUGETLB_PAGE 604int PageHuge(struct page *page); 605int PageHeadHuge(struct page *page); 606#else 607TESTPAGEFLAG_FALSE(Huge) 608TESTPAGEFLAG_FALSE(HeadHuge) 609#endif 610 611 612#ifdef CONFIG_TRANSPARENT_HUGEPAGE 613/* 614 * PageHuge() only returns true for hugetlbfs pages, but not for 615 * normal or transparent huge pages. 616 * 617 * PageTransHuge() returns true for both transparent huge and 618 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be 619 * called only in the core VM paths where hugetlbfs pages can't exist. 620 */ 621static inline int PageTransHuge(struct page *page) 622{ 623 VM_BUG_ON_PAGE(PageTail(page), page); 624 return PageHead(page); 625} 626 627/* 628 * PageTransCompound returns true for both transparent huge pages 629 * and hugetlbfs pages, so it should only be called when it's known 630 * that hugetlbfs pages aren't involved. 631 */ 632static inline int PageTransCompound(struct page *page) 633{ 634 return PageCompound(page); 635} 636 637/* 638 * PageTransTail returns true for both transparent huge pages 639 * and hugetlbfs pages, so it should only be called when it's known 640 * that hugetlbfs pages aren't involved. 641 */ 642static inline int PageTransTail(struct page *page) 643{ 644 return PageTail(page); 645} 646 647/* 648 * PageDoubleMap indicates that the compound page is mapped with PTEs as well 649 * as PMDs. 650 * 651 * This is required for optimization of rmap operations for THP: we can postpone 652 * per small page mapcount accounting (and its overhead from atomic operations) 653 * until the first PMD split. 654 * 655 * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up 656 * by one. This reference will go away with last compound_mapcount. 657 * 658 * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap(). 659 */ 660PAGEFLAG(DoubleMap, double_map, PF_SECOND) 661 TESTSCFLAG(DoubleMap, double_map, PF_SECOND) 662#else 663TESTPAGEFLAG_FALSE(TransHuge) 664TESTPAGEFLAG_FALSE(TransCompound) 665TESTPAGEFLAG_FALSE(TransCompoundMap) 666TESTPAGEFLAG_FALSE(TransTail) 667PAGEFLAG_FALSE(DoubleMap) 668 TESTSCFLAG_FALSE(DoubleMap) 669#endif 670 671/* 672 * Check if a page is currently marked HWPoisoned. Note that this check is 673 * best effort only and inherently racy: there is no way to synchronize with 674 * failing hardware. 675 */ 676static inline bool is_page_hwpoison(struct page *page) 677{ 678 if (PageHWPoison(page)) 679 return true; 680 return PageHuge(page) && PageHWPoison(compound_head(page)); 681} 682 683/* 684 * For pages that are never mapped to userspace (and aren't PageSlab), 685 * page_type may be used. Because it is initialised to -1, we invert the 686 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and 687 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and 688 * low bits so that an underflow or overflow of page_mapcount() won't be 689 * mistaken for a page type value. 690 */ 691 692#define PAGE_TYPE_BASE 0xf0000000 693/* Reserve 0x0000007f to catch underflows of page_mapcount */ 694#define PAGE_MAPCOUNT_RESERVE -128 695#define PG_buddy 0x00000080 696#define PG_offline 0x00000100 697#define PG_table 0x00000200 698#define PG_guard 0x00000400 699 700#define PageType(page, flag) \ 701 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE) 702 703static inline int page_has_type(struct page *page) 704{ 705 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE; 706} 707 708#define PAGE_TYPE_OPS(uname, lname) \ 709static __always_inline int Page##uname(struct page *page) \ 710{ \ 711 return PageType(page, PG_##lname); \ 712} \ 713static __always_inline void __SetPage##uname(struct page *page) \ 714{ \ 715 VM_BUG_ON_PAGE(!PageType(page, 0), page); \ 716 page->page_type &= ~PG_##lname; \ 717} \ 718static __always_inline void __ClearPage##uname(struct page *page) \ 719{ \ 720 VM_BUG_ON_PAGE(!Page##uname(page), page); \ 721 page->page_type |= PG_##lname; \ 722} 723 724/* 725 * PageBuddy() indicates that the page is free and in the buddy system 726 * (see mm/page_alloc.c). 727 */ 728PAGE_TYPE_OPS(Buddy, buddy) 729 730/* 731 * PageOffline() indicates that the page is logically offline although the 732 * containing section is online. (e.g. inflated in a balloon driver or 733 * not onlined when onlining the section). 734 * The content of these pages is effectively stale. Such pages should not 735 * be touched (read/write/dump/save) except by their owner. 736 * 737 * If a driver wants to allow to offline unmovable PageOffline() pages without 738 * putting them back to the buddy, it can do so via the memory notifier by 739 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the 740 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline() 741 * pages (now with a reference count of zero) are treated like free pages, 742 * allowing the containing memory block to get offlined. A driver that 743 * relies on this feature is aware that re-onlining the memory block will 744 * require to re-set the pages PageOffline() and not giving them to the 745 * buddy via online_page_callback_t. 746 * 747 * There are drivers that mark a page PageOffline() and expect there won't be 748 * any further access to page content. PFN walkers that read content of random 749 * pages should check PageOffline() and synchronize with such drivers using 750 * page_offline_freeze()/page_offline_thaw(). 751 */ 752PAGE_TYPE_OPS(Offline, offline) 753 754extern void page_offline_freeze(void); 755extern void page_offline_thaw(void); 756extern void page_offline_begin(void); 757extern void page_offline_end(void); 758 759/* 760 * Marks pages in use as page tables. 761 */ 762PAGE_TYPE_OPS(Table, table) 763 764/* 765 * Marks guardpages used with debug_pagealloc. 766 */ 767PAGE_TYPE_OPS(Guard, guard) 768 769extern bool is_free_buddy_page(struct page *page); 770 771__PAGEFLAG(Isolated, isolated, PF_ANY); 772 773/* 774 * If network-based swap is enabled, sl*b must keep track of whether pages 775 * were allocated from pfmemalloc reserves. 776 */ 777static inline int PageSlabPfmemalloc(struct page *page) 778{ 779 VM_BUG_ON_PAGE(!PageSlab(page), page); 780 return PageActive(page); 781} 782 783/* 784 * A version of PageSlabPfmemalloc() for opportunistic checks where the page 785 * might have been freed under us and not be a PageSlab anymore. 786 */ 787static inline int __PageSlabPfmemalloc(struct page *page) 788{ 789 return PageActive(page); 790} 791 792static inline void SetPageSlabPfmemalloc(struct page *page) 793{ 794 VM_BUG_ON_PAGE(!PageSlab(page), page); 795 SetPageActive(page); 796} 797 798static inline void __ClearPageSlabPfmemalloc(struct page *page) 799{ 800 VM_BUG_ON_PAGE(!PageSlab(page), page); 801 __ClearPageActive(page); 802} 803 804static inline void ClearPageSlabPfmemalloc(struct page *page) 805{ 806 VM_BUG_ON_PAGE(!PageSlab(page), page); 807 ClearPageActive(page); 808} 809 810#ifdef CONFIG_MMU 811#define __PG_MLOCKED (1UL << PG_mlocked) 812#else 813#define __PG_MLOCKED 0 814#endif 815 816/* 817 * Flags checked when a page is freed. Pages being freed should not have 818 * these flags set. If they are, there is a problem. 819 */ 820#define PAGE_FLAGS_CHECK_AT_FREE \ 821 (1UL << PG_lru | 1UL << PG_locked | \ 822 1UL << PG_private | 1UL << PG_private_2 | \ 823 1UL << PG_writeback | 1UL << PG_reserved | \ 824 1UL << PG_slab | 1UL << PG_active | \ 825 1UL << PG_unevictable | __PG_MLOCKED) 826 827/* 828 * Flags checked when a page is prepped for return by the page allocator. 829 * Pages being prepped should not have these flags set. If they are set, 830 * there has been a kernel bug or struct page corruption. 831 * 832 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's 833 * alloc-free cycle to prevent from reusing the page. 834 */ 835#define PAGE_FLAGS_CHECK_AT_PREP \ 836 (PAGEFLAGS_MASK & ~__PG_HWPOISON) 837 838#define PAGE_FLAGS_PRIVATE \ 839 (1UL << PG_private | 1UL << PG_private_2) 840/** 841 * page_has_private - Determine if page has private stuff 842 * @page: The page to be checked 843 * 844 * Determine if a page has private stuff, indicating that release routines 845 * should be invoked upon it. 846 */ 847static inline int page_has_private(struct page *page) 848{ 849 return !!(page->flags & PAGE_FLAGS_PRIVATE); 850} 851 852#undef PF_ANY 853#undef PF_HEAD 854#undef PF_ONLY_HEAD 855#undef PF_NO_TAIL 856#undef PF_NO_COMPOUND 857#undef PF_SECOND 858#endif /* !__GENERATING_BOUNDS_H */ 859 860#endif /* PAGE_FLAGS_H */