| 1 | #ifndef _LINUX_PAGEMAP_H |
| 2 | #define _LINUX_PAGEMAP_H |
| 3 | |
| 4 | /* |
| 5 | * Copyright 1995 Linus Torvalds |
| 6 | */ |
| 7 | #include <linux/mm.h> |
| 8 | #include <linux/fs.h> |
| 9 | #include <linux/list.h> |
| 10 | #include <linux/highmem.h> |
| 11 | #include <linux/compiler.h> |
| 12 | #include <asm/uaccess.h> |
| 13 | #include <linux/gfp.h> |
| 14 | #include <linux/bitops.h> |
| 15 | #include <linux/hardirq.h> /* for in_interrupt() */ |
| 16 | #include <linux/hugetlb_inline.h> |
| 17 | |
| 18 | /* |
| 19 | * Bits in mapping->flags. The lower __GFP_BITS_SHIFT bits are the page |
| 20 | * allocation mode flags. |
| 21 | */ |
| 22 | enum mapping_flags { |
| 23 | AS_EIO = __GFP_BITS_SHIFT + 0, /* IO error on async write */ |
| 24 | AS_ENOSPC = __GFP_BITS_SHIFT + 1, /* ENOSPC on async write */ |
| 25 | AS_MM_ALL_LOCKS = __GFP_BITS_SHIFT + 2, /* under mm_take_all_locks() */ |
| 26 | AS_UNEVICTABLE = __GFP_BITS_SHIFT + 3, /* e.g., ramdisk, SHM_LOCK */ |
| 27 | AS_EXITING = __GFP_BITS_SHIFT + 4, /* final truncate in progress */ |
| 28 | }; |
| 29 | |
| 30 | static inline void mapping_set_error(struct address_space *mapping, int error) |
| 31 | { |
| 32 | if (unlikely(error)) { |
| 33 | if (error == -ENOSPC) |
| 34 | set_bit(AS_ENOSPC, &mapping->flags); |
| 35 | else |
| 36 | set_bit(AS_EIO, &mapping->flags); |
| 37 | } |
| 38 | } |
| 39 | |
| 40 | static inline void mapping_set_unevictable(struct address_space *mapping) |
| 41 | { |
| 42 | set_bit(AS_UNEVICTABLE, &mapping->flags); |
| 43 | } |
| 44 | |
| 45 | static inline void mapping_clear_unevictable(struct address_space *mapping) |
| 46 | { |
| 47 | clear_bit(AS_UNEVICTABLE, &mapping->flags); |
| 48 | } |
| 49 | |
| 50 | static inline int mapping_unevictable(struct address_space *mapping) |
| 51 | { |
| 52 | if (mapping) |
| 53 | return test_bit(AS_UNEVICTABLE, &mapping->flags); |
| 54 | return !!mapping; |
| 55 | } |
| 56 | |
| 57 | static inline void mapping_set_exiting(struct address_space *mapping) |
| 58 | { |
| 59 | set_bit(AS_EXITING, &mapping->flags); |
| 60 | } |
| 61 | |
| 62 | static inline int mapping_exiting(struct address_space *mapping) |
| 63 | { |
| 64 | return test_bit(AS_EXITING, &mapping->flags); |
| 65 | } |
| 66 | |
| 67 | static inline gfp_t mapping_gfp_mask(struct address_space * mapping) |
| 68 | { |
| 69 | return (__force gfp_t)mapping->flags & __GFP_BITS_MASK; |
| 70 | } |
| 71 | |
| 72 | /* Restricts the given gfp_mask to what the mapping allows. */ |
| 73 | static inline gfp_t mapping_gfp_constraint(struct address_space *mapping, |
| 74 | gfp_t gfp_mask) |
| 75 | { |
| 76 | return mapping_gfp_mask(mapping) & gfp_mask; |
| 77 | } |
| 78 | |
| 79 | /* |
| 80 | * This is non-atomic. Only to be used before the mapping is activated. |
| 81 | * Probably needs a barrier... |
| 82 | */ |
| 83 | static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask) |
| 84 | { |
| 85 | m->flags = (m->flags & ~(__force unsigned long)__GFP_BITS_MASK) | |
| 86 | (__force unsigned long)mask; |
| 87 | } |
| 88 | |
| 89 | void release_pages(struct page **pages, int nr, bool cold); |
| 90 | |
| 91 | /* |
| 92 | * speculatively take a reference to a page. |
| 93 | * If the page is free (_count == 0), then _count is untouched, and 0 |
| 94 | * is returned. Otherwise, _count is incremented by 1 and 1 is returned. |
| 95 | * |
| 96 | * This function must be called inside the same rcu_read_lock() section as has |
| 97 | * been used to lookup the page in the pagecache radix-tree (or page table): |
| 98 | * this allows allocators to use a synchronize_rcu() to stabilize _count. |
| 99 | * |
| 100 | * Unless an RCU grace period has passed, the count of all pages coming out |
| 101 | * of the allocator must be considered unstable. page_count may return higher |
| 102 | * than expected, and put_page must be able to do the right thing when the |
| 103 | * page has been finished with, no matter what it is subsequently allocated |
| 104 | * for (because put_page is what is used here to drop an invalid speculative |
| 105 | * reference). |
| 106 | * |
| 107 | * This is the interesting part of the lockless pagecache (and lockless |
| 108 | * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page) |
| 109 | * has the following pattern: |
| 110 | * 1. find page in radix tree |
| 111 | * 2. conditionally increment refcount |
| 112 | * 3. check the page is still in pagecache (if no, goto 1) |
| 113 | * |
| 114 | * Remove-side that cares about stability of _count (eg. reclaim) has the |
| 115 | * following (with tree_lock held for write): |
| 116 | * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg) |
| 117 | * B. remove page from pagecache |
| 118 | * C. free the page |
| 119 | * |
| 120 | * There are 2 critical interleavings that matter: |
| 121 | * - 2 runs before A: in this case, A sees elevated refcount and bails out |
| 122 | * - A runs before 2: in this case, 2 sees zero refcount and retries; |
| 123 | * subsequently, B will complete and 1 will find no page, causing the |
| 124 | * lookup to return NULL. |
| 125 | * |
| 126 | * It is possible that between 1 and 2, the page is removed then the exact same |
| 127 | * page is inserted into the same position in pagecache. That's OK: the |
| 128 | * old find_get_page using tree_lock could equally have run before or after |
| 129 | * such a re-insertion, depending on order that locks are granted. |
| 130 | * |
| 131 | * Lookups racing against pagecache insertion isn't a big problem: either 1 |
| 132 | * will find the page or it will not. Likewise, the old find_get_page could run |
| 133 | * either before the insertion or afterwards, depending on timing. |
| 134 | */ |
| 135 | static inline int page_cache_get_speculative(struct page *page) |
| 136 | { |
| 137 | VM_BUG_ON(in_interrupt()); |
| 138 | |
| 139 | #ifdef CONFIG_TINY_RCU |
| 140 | # ifdef CONFIG_PREEMPT_COUNT |
| 141 | VM_BUG_ON(!in_atomic()); |
| 142 | # endif |
| 143 | /* |
| 144 | * Preempt must be disabled here - we rely on rcu_read_lock doing |
| 145 | * this for us. |
| 146 | * |
| 147 | * Pagecache won't be truncated from interrupt context, so if we have |
| 148 | * found a page in the radix tree here, we have pinned its refcount by |
| 149 | * disabling preempt, and hence no need for the "speculative get" that |
| 150 | * SMP requires. |
| 151 | */ |
| 152 | VM_BUG_ON_PAGE(page_count(page) == 0, page); |
| 153 | page_ref_inc(page); |
| 154 | |
| 155 | #else |
| 156 | if (unlikely(!get_page_unless_zero(page))) { |
| 157 | /* |
| 158 | * Either the page has been freed, or will be freed. |
| 159 | * In either case, retry here and the caller should |
| 160 | * do the right thing (see comments above). |
| 161 | */ |
| 162 | return 0; |
| 163 | } |
| 164 | #endif |
| 165 | VM_BUG_ON_PAGE(PageTail(page), page); |
| 166 | |
| 167 | return 1; |
| 168 | } |
| 169 | |
| 170 | /* |
| 171 | * Same as above, but add instead of inc (could just be merged) |
| 172 | */ |
| 173 | static inline int page_cache_add_speculative(struct page *page, int count) |
| 174 | { |
| 175 | VM_BUG_ON(in_interrupt()); |
| 176 | |
| 177 | #if !defined(CONFIG_SMP) && defined(CONFIG_TREE_RCU) |
| 178 | # ifdef CONFIG_PREEMPT_COUNT |
| 179 | VM_BUG_ON(!in_atomic()); |
| 180 | # endif |
| 181 | VM_BUG_ON_PAGE(page_count(page) == 0, page); |
| 182 | page_ref_add(page, count); |
| 183 | |
| 184 | #else |
| 185 | if (unlikely(!page_ref_add_unless(page, count, 0))) |
| 186 | return 0; |
| 187 | #endif |
| 188 | VM_BUG_ON_PAGE(PageCompound(page) && page != compound_head(page), page); |
| 189 | |
| 190 | return 1; |
| 191 | } |
| 192 | |
| 193 | #ifdef CONFIG_NUMA |
| 194 | extern struct page *__page_cache_alloc(gfp_t gfp); |
| 195 | #else |
| 196 | static inline struct page *__page_cache_alloc(gfp_t gfp) |
| 197 | { |
| 198 | return alloc_pages(gfp, 0); |
| 199 | } |
| 200 | #endif |
| 201 | |
| 202 | static inline struct page *page_cache_alloc(struct address_space *x) |
| 203 | { |
| 204 | return __page_cache_alloc(mapping_gfp_mask(x)); |
| 205 | } |
| 206 | |
| 207 | static inline struct page *page_cache_alloc_cold(struct address_space *x) |
| 208 | { |
| 209 | return __page_cache_alloc(mapping_gfp_mask(x)|__GFP_COLD); |
| 210 | } |
| 211 | |
| 212 | static inline struct page *page_cache_alloc_readahead(struct address_space *x) |
| 213 | { |
| 214 | return __page_cache_alloc(mapping_gfp_mask(x) | |
| 215 | __GFP_COLD | __GFP_NORETRY | __GFP_NOWARN); |
| 216 | } |
| 217 | |
| 218 | typedef int filler_t(void *, struct page *); |
| 219 | |
| 220 | pgoff_t page_cache_next_hole(struct address_space *mapping, |
| 221 | pgoff_t index, unsigned long max_scan); |
| 222 | pgoff_t page_cache_prev_hole(struct address_space *mapping, |
| 223 | pgoff_t index, unsigned long max_scan); |
| 224 | |
| 225 | #define FGP_ACCESSED 0x00000001 |
| 226 | #define FGP_LOCK 0x00000002 |
| 227 | #define FGP_CREAT 0x00000004 |
| 228 | #define FGP_WRITE 0x00000008 |
| 229 | #define FGP_NOFS 0x00000010 |
| 230 | #define FGP_NOWAIT 0x00000020 |
| 231 | |
| 232 | struct page *pagecache_get_page(struct address_space *mapping, pgoff_t offset, |
| 233 | int fgp_flags, gfp_t cache_gfp_mask); |
| 234 | |
| 235 | /** |
| 236 | * find_get_page - find and get a page reference |
| 237 | * @mapping: the address_space to search |
| 238 | * @offset: the page index |
| 239 | * |
| 240 | * Looks up the page cache slot at @mapping & @offset. If there is a |
| 241 | * page cache page, it is returned with an increased refcount. |
| 242 | * |
| 243 | * Otherwise, %NULL is returned. |
| 244 | */ |
| 245 | static inline struct page *find_get_page(struct address_space *mapping, |
| 246 | pgoff_t offset) |
| 247 | { |
| 248 | return pagecache_get_page(mapping, offset, 0, 0); |
| 249 | } |
| 250 | |
| 251 | static inline struct page *find_get_page_flags(struct address_space *mapping, |
| 252 | pgoff_t offset, int fgp_flags) |
| 253 | { |
| 254 | return pagecache_get_page(mapping, offset, fgp_flags, 0); |
| 255 | } |
| 256 | |
| 257 | /** |
| 258 | * find_lock_page - locate, pin and lock a pagecache page |
| 259 | * pagecache_get_page - find and get a page reference |
| 260 | * @mapping: the address_space to search |
| 261 | * @offset: the page index |
| 262 | * |
| 263 | * Looks up the page cache slot at @mapping & @offset. If there is a |
| 264 | * page cache page, it is returned locked and with an increased |
| 265 | * refcount. |
| 266 | * |
| 267 | * Otherwise, %NULL is returned. |
| 268 | * |
| 269 | * find_lock_page() may sleep. |
| 270 | */ |
| 271 | static inline struct page *find_lock_page(struct address_space *mapping, |
| 272 | pgoff_t offset) |
| 273 | { |
| 274 | return pagecache_get_page(mapping, offset, FGP_LOCK, 0); |
| 275 | } |
| 276 | |
| 277 | /** |
| 278 | * find_or_create_page - locate or add a pagecache page |
| 279 | * @mapping: the page's address_space |
| 280 | * @index: the page's index into the mapping |
| 281 | * @gfp_mask: page allocation mode |
| 282 | * |
| 283 | * Looks up the page cache slot at @mapping & @offset. If there is a |
| 284 | * page cache page, it is returned locked and with an increased |
| 285 | * refcount. |
| 286 | * |
| 287 | * If the page is not present, a new page is allocated using @gfp_mask |
| 288 | * and added to the page cache and the VM's LRU list. The page is |
| 289 | * returned locked and with an increased refcount. |
| 290 | * |
| 291 | * On memory exhaustion, %NULL is returned. |
| 292 | * |
| 293 | * find_or_create_page() may sleep, even if @gfp_flags specifies an |
| 294 | * atomic allocation! |
| 295 | */ |
| 296 | static inline struct page *find_or_create_page(struct address_space *mapping, |
| 297 | pgoff_t offset, gfp_t gfp_mask) |
| 298 | { |
| 299 | return pagecache_get_page(mapping, offset, |
| 300 | FGP_LOCK|FGP_ACCESSED|FGP_CREAT, |
| 301 | gfp_mask); |
| 302 | } |
| 303 | |
| 304 | /** |
| 305 | * grab_cache_page_nowait - returns locked page at given index in given cache |
| 306 | * @mapping: target address_space |
| 307 | * @index: the page index |
| 308 | * |
| 309 | * Same as grab_cache_page(), but do not wait if the page is unavailable. |
| 310 | * This is intended for speculative data generators, where the data can |
| 311 | * be regenerated if the page couldn't be grabbed. This routine should |
| 312 | * be safe to call while holding the lock for another page. |
| 313 | * |
| 314 | * Clear __GFP_FS when allocating the page to avoid recursion into the fs |
| 315 | * and deadlock against the caller's locked page. |
| 316 | */ |
| 317 | static inline struct page *grab_cache_page_nowait(struct address_space *mapping, |
| 318 | pgoff_t index) |
| 319 | { |
| 320 | return pagecache_get_page(mapping, index, |
| 321 | FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, |
| 322 | mapping_gfp_mask(mapping)); |
| 323 | } |
| 324 | |
| 325 | struct page *find_get_entry(struct address_space *mapping, pgoff_t offset); |
| 326 | struct page *find_lock_entry(struct address_space *mapping, pgoff_t offset); |
| 327 | unsigned find_get_entries(struct address_space *mapping, pgoff_t start, |
| 328 | unsigned int nr_entries, struct page **entries, |
| 329 | pgoff_t *indices); |
| 330 | unsigned find_get_pages(struct address_space *mapping, pgoff_t start, |
| 331 | unsigned int nr_pages, struct page **pages); |
| 332 | unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start, |
| 333 | unsigned int nr_pages, struct page **pages); |
| 334 | unsigned find_get_pages_tag(struct address_space *mapping, pgoff_t *index, |
| 335 | int tag, unsigned int nr_pages, struct page **pages); |
| 336 | unsigned find_get_entries_tag(struct address_space *mapping, pgoff_t start, |
| 337 | int tag, unsigned int nr_entries, |
| 338 | struct page **entries, pgoff_t *indices); |
| 339 | |
| 340 | struct page *grab_cache_page_write_begin(struct address_space *mapping, |
| 341 | pgoff_t index, unsigned flags); |
| 342 | |
| 343 | /* |
| 344 | * Returns locked page at given index in given cache, creating it if needed. |
| 345 | */ |
| 346 | static inline struct page *grab_cache_page(struct address_space *mapping, |
| 347 | pgoff_t index) |
| 348 | { |
| 349 | return find_or_create_page(mapping, index, mapping_gfp_mask(mapping)); |
| 350 | } |
| 351 | |
| 352 | extern struct page * read_cache_page(struct address_space *mapping, |
| 353 | pgoff_t index, filler_t *filler, void *data); |
| 354 | extern struct page * read_cache_page_gfp(struct address_space *mapping, |
| 355 | pgoff_t index, gfp_t gfp_mask); |
| 356 | extern int read_cache_pages(struct address_space *mapping, |
| 357 | struct list_head *pages, filler_t *filler, void *data); |
| 358 | |
| 359 | static inline struct page *read_mapping_page(struct address_space *mapping, |
| 360 | pgoff_t index, void *data) |
| 361 | { |
| 362 | filler_t *filler = (filler_t *)mapping->a_ops->readpage; |
| 363 | return read_cache_page(mapping, index, filler, data); |
| 364 | } |
| 365 | |
| 366 | /* |
| 367 | * Get the offset in PAGE_SIZE. |
| 368 | * (TODO: hugepage should have ->index in PAGE_SIZE) |
| 369 | */ |
| 370 | static inline pgoff_t page_to_pgoff(struct page *page) |
| 371 | { |
| 372 | pgoff_t pgoff; |
| 373 | |
| 374 | if (unlikely(PageHeadHuge(page))) |
| 375 | return page->index << compound_order(page); |
| 376 | |
| 377 | if (likely(!PageTransTail(page))) |
| 378 | return page->index; |
| 379 | |
| 380 | /* |
| 381 | * We don't initialize ->index for tail pages: calculate based on |
| 382 | * head page |
| 383 | */ |
| 384 | pgoff = compound_head(page)->index; |
| 385 | pgoff += page - compound_head(page); |
| 386 | return pgoff; |
| 387 | } |
| 388 | |
| 389 | /* |
| 390 | * Return byte-offset into filesystem object for page. |
| 391 | */ |
| 392 | static inline loff_t page_offset(struct page *page) |
| 393 | { |
| 394 | return ((loff_t)page->index) << PAGE_SHIFT; |
| 395 | } |
| 396 | |
| 397 | static inline loff_t page_file_offset(struct page *page) |
| 398 | { |
| 399 | return ((loff_t)page_file_index(page)) << PAGE_SHIFT; |
| 400 | } |
| 401 | |
| 402 | extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma, |
| 403 | unsigned long address); |
| 404 | |
| 405 | static inline pgoff_t linear_page_index(struct vm_area_struct *vma, |
| 406 | unsigned long address) |
| 407 | { |
| 408 | pgoff_t pgoff; |
| 409 | if (unlikely(is_vm_hugetlb_page(vma))) |
| 410 | return linear_hugepage_index(vma, address); |
| 411 | pgoff = (address - vma->vm_start) >> PAGE_SHIFT; |
| 412 | pgoff += vma->vm_pgoff; |
| 413 | return pgoff; |
| 414 | } |
| 415 | |
| 416 | extern void __lock_page(struct page *page); |
| 417 | extern int __lock_page_killable(struct page *page); |
| 418 | extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm, |
| 419 | unsigned int flags); |
| 420 | extern void unlock_page(struct page *page); |
| 421 | |
| 422 | static inline int trylock_page(struct page *page) |
| 423 | { |
| 424 | page = compound_head(page); |
| 425 | return (likely(!test_and_set_bit_lock(PG_locked, &page->flags))); |
| 426 | } |
| 427 | |
| 428 | /* |
| 429 | * lock_page may only be called if we have the page's inode pinned. |
| 430 | */ |
| 431 | static inline void lock_page(struct page *page) |
| 432 | { |
| 433 | might_sleep(); |
| 434 | if (!trylock_page(page)) |
| 435 | __lock_page(page); |
| 436 | } |
| 437 | |
| 438 | /* |
| 439 | * lock_page_killable is like lock_page but can be interrupted by fatal |
| 440 | * signals. It returns 0 if it locked the page and -EINTR if it was |
| 441 | * killed while waiting. |
| 442 | */ |
| 443 | static inline int lock_page_killable(struct page *page) |
| 444 | { |
| 445 | might_sleep(); |
| 446 | if (!trylock_page(page)) |
| 447 | return __lock_page_killable(page); |
| 448 | return 0; |
| 449 | } |
| 450 | |
| 451 | /* |
| 452 | * lock_page_or_retry - Lock the page, unless this would block and the |
| 453 | * caller indicated that it can handle a retry. |
| 454 | * |
| 455 | * Return value and mmap_sem implications depend on flags; see |
| 456 | * __lock_page_or_retry(). |
| 457 | */ |
| 458 | static inline int lock_page_or_retry(struct page *page, struct mm_struct *mm, |
| 459 | unsigned int flags) |
| 460 | { |
| 461 | might_sleep(); |
| 462 | return trylock_page(page) || __lock_page_or_retry(page, mm, flags); |
| 463 | } |
| 464 | |
| 465 | /* |
| 466 | * This is exported only for wait_on_page_locked/wait_on_page_writeback, |
| 467 | * and for filesystems which need to wait on PG_private. |
| 468 | */ |
| 469 | extern void wait_on_page_bit(struct page *page, int bit_nr); |
| 470 | |
| 471 | extern int wait_on_page_bit_killable(struct page *page, int bit_nr); |
| 472 | extern int wait_on_page_bit_killable_timeout(struct page *page, |
| 473 | int bit_nr, unsigned long timeout); |
| 474 | |
| 475 | static inline int wait_on_page_locked_killable(struct page *page) |
| 476 | { |
| 477 | if (!PageLocked(page)) |
| 478 | return 0; |
| 479 | return wait_on_page_bit_killable(compound_head(page), PG_locked); |
| 480 | } |
| 481 | |
| 482 | extern wait_queue_head_t *page_waitqueue(struct page *page); |
| 483 | static inline void wake_up_page(struct page *page, int bit) |
| 484 | { |
| 485 | __wake_up_bit(page_waitqueue(page), &page->flags, bit); |
| 486 | } |
| 487 | |
| 488 | /* |
| 489 | * Wait for a page to be unlocked. |
| 490 | * |
| 491 | * This must be called with the caller "holding" the page, |
| 492 | * ie with increased "page->count" so that the page won't |
| 493 | * go away during the wait.. |
| 494 | */ |
| 495 | static inline void wait_on_page_locked(struct page *page) |
| 496 | { |
| 497 | if (PageLocked(page)) |
| 498 | wait_on_page_bit(compound_head(page), PG_locked); |
| 499 | } |
| 500 | |
| 501 | /* |
| 502 | * Wait for a page to complete writeback |
| 503 | */ |
| 504 | static inline void wait_on_page_writeback(struct page *page) |
| 505 | { |
| 506 | if (PageWriteback(page)) |
| 507 | wait_on_page_bit(page, PG_writeback); |
| 508 | } |
| 509 | |
| 510 | extern void end_page_writeback(struct page *page); |
| 511 | void wait_for_stable_page(struct page *page); |
| 512 | |
| 513 | void page_endio(struct page *page, int rw, int err); |
| 514 | |
| 515 | /* |
| 516 | * Add an arbitrary waiter to a page's wait queue |
| 517 | */ |
| 518 | extern void add_page_wait_queue(struct page *page, wait_queue_t *waiter); |
| 519 | |
| 520 | /* |
| 521 | * Fault a userspace page into pagetables. Return non-zero on a fault. |
| 522 | * |
| 523 | * This assumes that two userspace pages are always sufficient. |
| 524 | */ |
| 525 | static inline int fault_in_pages_writeable(char __user *uaddr, int size) |
| 526 | { |
| 527 | int ret; |
| 528 | |
| 529 | if (unlikely(size == 0)) |
| 530 | return 0; |
| 531 | |
| 532 | /* |
| 533 | * Writing zeroes into userspace here is OK, because we know that if |
| 534 | * the zero gets there, we'll be overwriting it. |
| 535 | */ |
| 536 | ret = __put_user(0, uaddr); |
| 537 | if (ret == 0) { |
| 538 | char __user *end = uaddr + size - 1; |
| 539 | |
| 540 | /* |
| 541 | * If the page was already mapped, this will get a cache miss |
| 542 | * for sure, so try to avoid doing it. |
| 543 | */ |
| 544 | if (((unsigned long)uaddr & PAGE_MASK) != |
| 545 | ((unsigned long)end & PAGE_MASK)) |
| 546 | ret = __put_user(0, end); |
| 547 | } |
| 548 | return ret; |
| 549 | } |
| 550 | |
| 551 | static inline int fault_in_pages_readable(const char __user *uaddr, int size) |
| 552 | { |
| 553 | volatile char c; |
| 554 | int ret; |
| 555 | |
| 556 | if (unlikely(size == 0)) |
| 557 | return 0; |
| 558 | |
| 559 | ret = __get_user(c, uaddr); |
| 560 | if (ret == 0) { |
| 561 | const char __user *end = uaddr + size - 1; |
| 562 | |
| 563 | if (((unsigned long)uaddr & PAGE_MASK) != |
| 564 | ((unsigned long)end & PAGE_MASK)) { |
| 565 | ret = __get_user(c, end); |
| 566 | (void)c; |
| 567 | } |
| 568 | } |
| 569 | return ret; |
| 570 | } |
| 571 | |
| 572 | /* |
| 573 | * Multipage variants of the above prefault helpers, useful if more than |
| 574 | * PAGE_SIZE of data needs to be prefaulted. These are separate from the above |
| 575 | * functions (which only handle up to PAGE_SIZE) to avoid clobbering the |
| 576 | * filemap.c hotpaths. |
| 577 | */ |
| 578 | static inline int fault_in_multipages_writeable(char __user *uaddr, int size) |
| 579 | { |
| 580 | int ret = 0; |
| 581 | char __user *end = uaddr + size - 1; |
| 582 | |
| 583 | if (unlikely(size == 0)) |
| 584 | return ret; |
| 585 | |
| 586 | /* |
| 587 | * Writing zeroes into userspace here is OK, because we know that if |
| 588 | * the zero gets there, we'll be overwriting it. |
| 589 | */ |
| 590 | while (uaddr <= end) { |
| 591 | ret = __put_user(0, uaddr); |
| 592 | if (ret != 0) |
| 593 | return ret; |
| 594 | uaddr += PAGE_SIZE; |
| 595 | } |
| 596 | |
| 597 | /* Check whether the range spilled into the next page. */ |
| 598 | if (((unsigned long)uaddr & PAGE_MASK) == |
| 599 | ((unsigned long)end & PAGE_MASK)) |
| 600 | ret = __put_user(0, end); |
| 601 | |
| 602 | return ret; |
| 603 | } |
| 604 | |
| 605 | static inline int fault_in_multipages_readable(const char __user *uaddr, |
| 606 | int size) |
| 607 | { |
| 608 | volatile char c; |
| 609 | int ret = 0; |
| 610 | const char __user *end = uaddr + size - 1; |
| 611 | |
| 612 | if (unlikely(size == 0)) |
| 613 | return ret; |
| 614 | |
| 615 | while (uaddr <= end) { |
| 616 | ret = __get_user(c, uaddr); |
| 617 | if (ret != 0) |
| 618 | return ret; |
| 619 | uaddr += PAGE_SIZE; |
| 620 | } |
| 621 | |
| 622 | /* Check whether the range spilled into the next page. */ |
| 623 | if (((unsigned long)uaddr & PAGE_MASK) == |
| 624 | ((unsigned long)end & PAGE_MASK)) { |
| 625 | ret = __get_user(c, end); |
| 626 | (void)c; |
| 627 | } |
| 628 | |
| 629 | return ret; |
| 630 | } |
| 631 | |
| 632 | int add_to_page_cache_locked(struct page *page, struct address_space *mapping, |
| 633 | pgoff_t index, gfp_t gfp_mask); |
| 634 | int add_to_page_cache_lru(struct page *page, struct address_space *mapping, |
| 635 | pgoff_t index, gfp_t gfp_mask); |
| 636 | extern void delete_from_page_cache(struct page *page); |
| 637 | extern void __delete_from_page_cache(struct page *page, void *shadow); |
| 638 | int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask); |
| 639 | |
| 640 | /* |
| 641 | * Like add_to_page_cache_locked, but used to add newly allocated pages: |
| 642 | * the page is new, so we can just run __SetPageLocked() against it. |
| 643 | */ |
| 644 | static inline int add_to_page_cache(struct page *page, |
| 645 | struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask) |
| 646 | { |
| 647 | int error; |
| 648 | |
| 649 | __SetPageLocked(page); |
| 650 | error = add_to_page_cache_locked(page, mapping, offset, gfp_mask); |
| 651 | if (unlikely(error)) |
| 652 | __ClearPageLocked(page); |
| 653 | return error; |
| 654 | } |
| 655 | |
| 656 | static inline unsigned long dir_pages(struct inode *inode) |
| 657 | { |
| 658 | return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >> |
| 659 | PAGE_SHIFT; |
| 660 | } |
| 661 | |
| 662 | #endif /* _LINUX_PAGEMAP_H */ |