1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* internal.h: mm/ internal definitions 3 * 4 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. 5 * Written by David Howells (dhowells@redhat.com) 6 */ 7 #ifndef __MM_INTERNAL_H 8 #define __MM_INTERNAL_H 9 10 #include <linux/fs.h> 11 #include <linux/mm.h> 12 #include <linux/pagemap.h> 13 #include <linux/rmap.h> 14 #include <linux/tracepoint-defs.h> 15 16 struct folio_batch; 17 18 /* 19 * The set of flags that only affect watermark checking and reclaim 20 * behaviour. This is used by the MM to obey the caller constraints 21 * about IO, FS and watermark checking while ignoring placement 22 * hints such as HIGHMEM usage. 23 */ 24 #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\ 25 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\ 26 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\ 27 __GFP_NOLOCKDEP) 28 29 /* The GFP flags allowed during early boot */ 30 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS)) 31 32 /* Control allocation cpuset and node placement constraints */ 33 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) 34 35 /* Do not use these with a slab allocator */ 36 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) 37 38 /* 39 * Different from WARN_ON_ONCE(), no warning will be issued 40 * when we specify __GFP_NOWARN. 41 */ 42 #define WARN_ON_ONCE_GFP(cond, gfp) ({ \ 43 static bool __section(".data.once") __warned; \ 44 int __ret_warn_once = !!(cond); \ 45 \ 46 if (unlikely(!(gfp & __GFP_NOWARN) && __ret_warn_once && !__warned)) { \ 47 __warned = true; \ 48 WARN_ON(1); \ 49 } \ 50 unlikely(__ret_warn_once); \ 51 }) 52 53 void page_writeback_init(void); 54 55 /* 56 * If a 16GB hugetlb folio were mapped by PTEs of all of its 4kB pages, 57 * its nr_pages_mapped would be 0x400000: choose the ENTIRELY_MAPPED bit 58 * above that range, instead of 2*(PMD_SIZE/PAGE_SIZE). Hugetlb currently 59 * leaves nr_pages_mapped at 0, but avoid surprise if it participates later. 60 */ 61 #define ENTIRELY_MAPPED 0x800000 62 #define FOLIO_PAGES_MAPPED (ENTIRELY_MAPPED - 1) 63 64 /* 65 * Flags passed to __show_mem() and show_free_areas() to suppress output in 66 * various contexts. 67 */ 68 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ 69 70 /* 71 * How many individual pages have an elevated _mapcount. Excludes 72 * the folio's entire_mapcount. 73 */ 74 static inline int folio_nr_pages_mapped(const struct folio *folio) 75 { 76 return atomic_read(&folio->_nr_pages_mapped) & FOLIO_PAGES_MAPPED; 77 } 78 79 /* 80 * Retrieve the first entry of a folio based on a provided entry within the 81 * folio. We cannot rely on folio->swap as there is no guarantee that it has 82 * been initialized. Used for calling arch_swap_restore() 83 */ 84 static inline swp_entry_t folio_swap(swp_entry_t entry, 85 const struct folio *folio) 86 { 87 swp_entry_t swap = { 88 .val = ALIGN_DOWN(entry.val, folio_nr_pages(folio)), 89 }; 90 91 return swap; 92 } 93 94 static inline void *folio_raw_mapping(const struct folio *folio) 95 { 96 unsigned long mapping = (unsigned long)folio->mapping; 97 98 return (void *)(mapping & ~PAGE_MAPPING_FLAGS); 99 } 100 101 #ifdef CONFIG_MMU 102 103 /* Flags for folio_pte_batch(). */ 104 typedef int __bitwise fpb_t; 105 106 /* Compare PTEs after pte_mkclean(), ignoring the dirty bit. */ 107 #define FPB_IGNORE_DIRTY ((__force fpb_t)BIT(0)) 108 109 /* Compare PTEs after pte_clear_soft_dirty(), ignoring the soft-dirty bit. */ 110 #define FPB_IGNORE_SOFT_DIRTY ((__force fpb_t)BIT(1)) 111 112 static inline pte_t __pte_batch_clear_ignored(pte_t pte, fpb_t flags) 113 { 114 if (flags & FPB_IGNORE_DIRTY) 115 pte = pte_mkclean(pte); 116 if (likely(flags & FPB_IGNORE_SOFT_DIRTY)) 117 pte = pte_clear_soft_dirty(pte); 118 return pte_wrprotect(pte_mkold(pte)); 119 } 120 121 /** 122 * folio_pte_batch - detect a PTE batch for a large folio 123 * @folio: The large folio to detect a PTE batch for. 124 * @addr: The user virtual address the first page is mapped at. 125 * @start_ptep: Page table pointer for the first entry. 126 * @pte: Page table entry for the first page. 127 * @max_nr: The maximum number of table entries to consider. 128 * @flags: Flags to modify the PTE batch semantics. 129 * @any_writable: Optional pointer to indicate whether any entry except the 130 * first one is writable. 131 * 132 * Detect a PTE batch: consecutive (present) PTEs that map consecutive 133 * pages of the same large folio. 134 * 135 * All PTEs inside a PTE batch have the same PTE bits set, excluding the PFN, 136 * the accessed bit, writable bit, dirty bit (with FPB_IGNORE_DIRTY) and 137 * soft-dirty bit (with FPB_IGNORE_SOFT_DIRTY). 138 * 139 * start_ptep must map any page of the folio. max_nr must be at least one and 140 * must be limited by the caller so scanning cannot exceed a single page table. 141 * 142 * Return: the number of table entries in the batch. 143 */ 144 static inline int folio_pte_batch(struct folio *folio, unsigned long addr, 145 pte_t *start_ptep, pte_t pte, int max_nr, fpb_t flags, 146 bool *any_writable) 147 { 148 unsigned long folio_end_pfn = folio_pfn(folio) + folio_nr_pages(folio); 149 const pte_t *end_ptep = start_ptep + max_nr; 150 pte_t expected_pte, *ptep; 151 bool writable; 152 int nr; 153 154 if (any_writable) 155 *any_writable = false; 156 157 VM_WARN_ON_FOLIO(!pte_present(pte), folio); 158 VM_WARN_ON_FOLIO(!folio_test_large(folio) || max_nr < 1, folio); 159 VM_WARN_ON_FOLIO(page_folio(pfn_to_page(pte_pfn(pte))) != folio, folio); 160 161 nr = pte_batch_hint(start_ptep, pte); 162 expected_pte = __pte_batch_clear_ignored(pte_advance_pfn(pte, nr), flags); 163 ptep = start_ptep + nr; 164 165 while (ptep < end_ptep) { 166 pte = ptep_get(ptep); 167 if (any_writable) 168 writable = !!pte_write(pte); 169 pte = __pte_batch_clear_ignored(pte, flags); 170 171 if (!pte_same(pte, expected_pte)) 172 break; 173 174 /* 175 * Stop immediately once we reached the end of the folio. In 176 * corner cases the next PFN might fall into a different 177 * folio. 178 */ 179 if (pte_pfn(pte) >= folio_end_pfn) 180 break; 181 182 if (any_writable) 183 *any_writable |= writable; 184 185 nr = pte_batch_hint(ptep, pte); 186 expected_pte = pte_advance_pfn(expected_pte, nr); 187 ptep += nr; 188 } 189 190 return min(ptep - start_ptep, max_nr); 191 } 192 #endif /* CONFIG_MMU */ 193 194 void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio, 195 int nr_throttled); 196 static inline void acct_reclaim_writeback(struct folio *folio) 197 { 198 pg_data_t *pgdat = folio_pgdat(folio); 199 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled); 200 201 if (nr_throttled) 202 __acct_reclaim_writeback(pgdat, folio, nr_throttled); 203 } 204 205 static inline void wake_throttle_isolated(pg_data_t *pgdat) 206 { 207 wait_queue_head_t *wqh; 208 209 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED]; 210 if (waitqueue_active(wqh)) 211 wake_up(wqh); 212 } 213 214 vm_fault_t vmf_anon_prepare(struct vm_fault *vmf); 215 vm_fault_t do_swap_page(struct vm_fault *vmf); 216 void folio_rotate_reclaimable(struct folio *folio); 217 bool __folio_end_writeback(struct folio *folio); 218 void deactivate_file_folio(struct folio *folio); 219 void folio_activate(struct folio *folio); 220 221 void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas, 222 struct vm_area_struct *start_vma, unsigned long floor, 223 unsigned long ceiling, bool mm_wr_locked); 224 void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte); 225 226 struct zap_details; 227 void unmap_page_range(struct mmu_gather *tlb, 228 struct vm_area_struct *vma, 229 unsigned long addr, unsigned long end, 230 struct zap_details *details); 231 232 void page_cache_ra_order(struct readahead_control *, struct file_ra_state *, 233 unsigned int order); 234 void force_page_cache_ra(struct readahead_control *, unsigned long nr); 235 static inline void force_page_cache_readahead(struct address_space *mapping, 236 struct file *file, pgoff_t index, unsigned long nr_to_read) 237 { 238 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index); 239 force_page_cache_ra(&ractl, nr_to_read); 240 } 241 242 unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start, 243 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); 244 unsigned find_get_entries(struct address_space *mapping, pgoff_t *start, 245 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); 246 void filemap_free_folio(struct address_space *mapping, struct folio *folio); 247 int truncate_inode_folio(struct address_space *mapping, struct folio *folio); 248 bool truncate_inode_partial_folio(struct folio *folio, loff_t start, 249 loff_t end); 250 long mapping_evict_folio(struct address_space *mapping, struct folio *folio); 251 unsigned long mapping_try_invalidate(struct address_space *mapping, 252 pgoff_t start, pgoff_t end, unsigned long *nr_failed); 253 254 /** 255 * folio_evictable - Test whether a folio is evictable. 256 * @folio: The folio to test. 257 * 258 * Test whether @folio is evictable -- i.e., should be placed on 259 * active/inactive lists vs unevictable list. 260 * 261 * Reasons folio might not be evictable: 262 * 1. folio's mapping marked unevictable 263 * 2. One of the pages in the folio is part of an mlocked VMA 264 */ 265 static inline bool folio_evictable(struct folio *folio) 266 { 267 bool ret; 268 269 /* Prevent address_space of inode and swap cache from being freed */ 270 rcu_read_lock(); 271 ret = !mapping_unevictable(folio_mapping(folio)) && 272 !folio_test_mlocked(folio); 273 rcu_read_unlock(); 274 return ret; 275 } 276 277 /* 278 * Turn a non-refcounted page (->_refcount == 0) into refcounted with 279 * a count of one. 280 */ 281 static inline void set_page_refcounted(struct page *page) 282 { 283 VM_BUG_ON_PAGE(PageTail(page), page); 284 VM_BUG_ON_PAGE(page_ref_count(page), page); 285 set_page_count(page, 1); 286 } 287 288 /* 289 * Return true if a folio needs ->release_folio() calling upon it. 290 */ 291 static inline bool folio_needs_release(struct folio *folio) 292 { 293 struct address_space *mapping = folio_mapping(folio); 294 295 return folio_has_private(folio) || 296 (mapping && mapping_release_always(mapping)); 297 } 298 299 extern unsigned long highest_memmap_pfn; 300 301 /* 302 * Maximum number of reclaim retries without progress before the OOM 303 * killer is consider the only way forward. 304 */ 305 #define MAX_RECLAIM_RETRIES 16 306 307 /* 308 * in mm/vmscan.c: 309 */ 310 bool isolate_lru_page(struct page *page); 311 bool folio_isolate_lru(struct folio *folio); 312 void putback_lru_page(struct page *page); 313 void folio_putback_lru(struct folio *folio); 314 extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason); 315 316 /* 317 * in mm/rmap.c: 318 */ 319 pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); 320 321 /* 322 * in mm/page_alloc.c 323 */ 324 #define K(x) ((x) << (PAGE_SHIFT-10)) 325 326 extern char * const zone_names[MAX_NR_ZONES]; 327 328 /* perform sanity checks on struct pages being allocated or freed */ 329 DECLARE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled); 330 331 extern int min_free_kbytes; 332 333 void setup_per_zone_wmarks(void); 334 void calculate_min_free_kbytes(void); 335 int __meminit init_per_zone_wmark_min(void); 336 void page_alloc_sysctl_init(void); 337 338 /* 339 * Structure for holding the mostly immutable allocation parameters passed 340 * between functions involved in allocations, including the alloc_pages* 341 * family of functions. 342 * 343 * nodemask, migratetype and highest_zoneidx are initialized only once in 344 * __alloc_pages() and then never change. 345 * 346 * zonelist, preferred_zone and highest_zoneidx are set first in 347 * __alloc_pages() for the fast path, and might be later changed 348 * in __alloc_pages_slowpath(). All other functions pass the whole structure 349 * by a const pointer. 350 */ 351 struct alloc_context { 352 struct zonelist *zonelist; 353 nodemask_t *nodemask; 354 struct zoneref *preferred_zoneref; 355 int migratetype; 356 357 /* 358 * highest_zoneidx represents highest usable zone index of 359 * the allocation request. Due to the nature of the zone, 360 * memory on lower zone than the highest_zoneidx will be 361 * protected by lowmem_reserve[highest_zoneidx]. 362 * 363 * highest_zoneidx is also used by reclaim/compaction to limit 364 * the target zone since higher zone than this index cannot be 365 * usable for this allocation request. 366 */ 367 enum zone_type highest_zoneidx; 368 bool spread_dirty_pages; 369 }; 370 371 /* 372 * This function returns the order of a free page in the buddy system. In 373 * general, page_zone(page)->lock must be held by the caller to prevent the 374 * page from being allocated in parallel and returning garbage as the order. 375 * If a caller does not hold page_zone(page)->lock, it must guarantee that the 376 * page cannot be allocated or merged in parallel. Alternatively, it must 377 * handle invalid values gracefully, and use buddy_order_unsafe() below. 378 */ 379 static inline unsigned int buddy_order(struct page *page) 380 { 381 /* PageBuddy() must be checked by the caller */ 382 return page_private(page); 383 } 384 385 /* 386 * Like buddy_order(), but for callers who cannot afford to hold the zone lock. 387 * PageBuddy() should be checked first by the caller to minimize race window, 388 * and invalid values must be handled gracefully. 389 * 390 * READ_ONCE is used so that if the caller assigns the result into a local 391 * variable and e.g. tests it for valid range before using, the compiler cannot 392 * decide to remove the variable and inline the page_private(page) multiple 393 * times, potentially observing different values in the tests and the actual 394 * use of the result. 395 */ 396 #define buddy_order_unsafe(page) READ_ONCE(page_private(page)) 397 398 /* 399 * This function checks whether a page is free && is the buddy 400 * we can coalesce a page and its buddy if 401 * (a) the buddy is not in a hole (check before calling!) && 402 * (b) the buddy is in the buddy system && 403 * (c) a page and its buddy have the same order && 404 * (d) a page and its buddy are in the same zone. 405 * 406 * For recording whether a page is in the buddy system, we set PageBuddy. 407 * Setting, clearing, and testing PageBuddy is serialized by zone->lock. 408 * 409 * For recording page's order, we use page_private(page). 410 */ 411 static inline bool page_is_buddy(struct page *page, struct page *buddy, 412 unsigned int order) 413 { 414 if (!page_is_guard(buddy) && !PageBuddy(buddy)) 415 return false; 416 417 if (buddy_order(buddy) != order) 418 return false; 419 420 /* 421 * zone check is done late to avoid uselessly calculating 422 * zone/node ids for pages that could never merge. 423 */ 424 if (page_zone_id(page) != page_zone_id(buddy)) 425 return false; 426 427 VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy); 428 429 return true; 430 } 431 432 /* 433 * Locate the struct page for both the matching buddy in our 434 * pair (buddy1) and the combined O(n+1) page they form (page). 435 * 436 * 1) Any buddy B1 will have an order O twin B2 which satisfies 437 * the following equation: 438 * B2 = B1 ^ (1 << O) 439 * For example, if the starting buddy (buddy2) is #8 its order 440 * 1 buddy is #10: 441 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 442 * 443 * 2) Any buddy B will have an order O+1 parent P which 444 * satisfies the following equation: 445 * P = B & ~(1 << O) 446 * 447 * Assumption: *_mem_map is contiguous at least up to MAX_PAGE_ORDER 448 */ 449 static inline unsigned long 450 __find_buddy_pfn(unsigned long page_pfn, unsigned int order) 451 { 452 return page_pfn ^ (1 << order); 453 } 454 455 /* 456 * Find the buddy of @page and validate it. 457 * @page: The input page 458 * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the 459 * function is used in the performance-critical __free_one_page(). 460 * @order: The order of the page 461 * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to 462 * page_to_pfn(). 463 * 464 * The found buddy can be a non PageBuddy, out of @page's zone, or its order is 465 * not the same as @page. The validation is necessary before use it. 466 * 467 * Return: the found buddy page or NULL if not found. 468 */ 469 static inline struct page *find_buddy_page_pfn(struct page *page, 470 unsigned long pfn, unsigned int order, unsigned long *buddy_pfn) 471 { 472 unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order); 473 struct page *buddy; 474 475 buddy = page + (__buddy_pfn - pfn); 476 if (buddy_pfn) 477 *buddy_pfn = __buddy_pfn; 478 479 if (page_is_buddy(page, buddy, order)) 480 return buddy; 481 return NULL; 482 } 483 484 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, 485 unsigned long end_pfn, struct zone *zone); 486 487 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn, 488 unsigned long end_pfn, struct zone *zone) 489 { 490 if (zone->contiguous) 491 return pfn_to_page(start_pfn); 492 493 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone); 494 } 495 496 void set_zone_contiguous(struct zone *zone); 497 498 static inline void clear_zone_contiguous(struct zone *zone) 499 { 500 zone->contiguous = false; 501 } 502 503 extern int __isolate_free_page(struct page *page, unsigned int order); 504 extern void __putback_isolated_page(struct page *page, unsigned int order, 505 int mt); 506 extern void memblock_free_pages(struct page *page, unsigned long pfn, 507 unsigned int order); 508 extern void __free_pages_core(struct page *page, unsigned int order); 509 extern void kernel_init_pages(struct page *page, int numpages); 510 511 /* 512 * This will have no effect, other than possibly generating a warning, if the 513 * caller passes in a non-large folio. 514 */ 515 static inline void folio_set_order(struct folio *folio, unsigned int order) 516 { 517 if (WARN_ON_ONCE(!order || !folio_test_large(folio))) 518 return; 519 520 folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order; 521 #ifdef CONFIG_64BIT 522 folio->_folio_nr_pages = 1U << order; 523 #endif 524 } 525 526 void folio_undo_large_rmappable(struct folio *folio); 527 528 static inline struct folio *page_rmappable_folio(struct page *page) 529 { 530 struct folio *folio = (struct folio *)page; 531 532 if (folio && folio_test_large(folio)) 533 folio_set_large_rmappable(folio); 534 return folio; 535 } 536 537 static inline void prep_compound_head(struct page *page, unsigned int order) 538 { 539 struct folio *folio = (struct folio *)page; 540 541 folio_set_order(folio, order); 542 atomic_set(&folio->_entire_mapcount, -1); 543 atomic_set(&folio->_nr_pages_mapped, 0); 544 atomic_set(&folio->_pincount, 0); 545 if (order > 1) 546 INIT_LIST_HEAD(&folio->_deferred_list); 547 } 548 549 static inline void prep_compound_tail(struct page *head, int tail_idx) 550 { 551 struct page *p = head + tail_idx; 552 553 p->mapping = TAIL_MAPPING; 554 set_compound_head(p, head); 555 set_page_private(p, 0); 556 } 557 558 extern void prep_compound_page(struct page *page, unsigned int order); 559 560 extern void post_alloc_hook(struct page *page, unsigned int order, 561 gfp_t gfp_flags); 562 extern bool free_pages_prepare(struct page *page, unsigned int order); 563 564 extern int user_min_free_kbytes; 565 566 void free_unref_page(struct page *page, unsigned int order); 567 void free_unref_folios(struct folio_batch *fbatch); 568 569 extern void zone_pcp_reset(struct zone *zone); 570 extern void zone_pcp_disable(struct zone *zone); 571 extern void zone_pcp_enable(struct zone *zone); 572 extern void zone_pcp_init(struct zone *zone); 573 574 extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, 575 phys_addr_t min_addr, 576 int nid, bool exact_nid); 577 578 void memmap_init_range(unsigned long, int, unsigned long, unsigned long, 579 unsigned long, enum meminit_context, struct vmem_altmap *, int); 580 581 #if defined CONFIG_COMPACTION || defined CONFIG_CMA 582 583 /* 584 * in mm/compaction.c 585 */ 586 /* 587 * compact_control is used to track pages being migrated and the free pages 588 * they are being migrated to during memory compaction. The free_pfn starts 589 * at the end of a zone and migrate_pfn begins at the start. Movable pages 590 * are moved to the end of a zone during a compaction run and the run 591 * completes when free_pfn <= migrate_pfn 592 */ 593 struct compact_control { 594 struct list_head freepages[NR_PAGE_ORDERS]; /* List of free pages to migrate to */ 595 struct list_head migratepages; /* List of pages being migrated */ 596 unsigned int nr_freepages; /* Number of isolated free pages */ 597 unsigned int nr_migratepages; /* Number of pages to migrate */ 598 unsigned long free_pfn; /* isolate_freepages search base */ 599 /* 600 * Acts as an in/out parameter to page isolation for migration. 601 * isolate_migratepages uses it as a search base. 602 * isolate_migratepages_block will update the value to the next pfn 603 * after the last isolated one. 604 */ 605 unsigned long migrate_pfn; 606 unsigned long fast_start_pfn; /* a pfn to start linear scan from */ 607 struct zone *zone; 608 unsigned long total_migrate_scanned; 609 unsigned long total_free_scanned; 610 unsigned short fast_search_fail;/* failures to use free list searches */ 611 short search_order; /* order to start a fast search at */ 612 const gfp_t gfp_mask; /* gfp mask of a direct compactor */ 613 int order; /* order a direct compactor needs */ 614 int migratetype; /* migratetype of direct compactor */ 615 const unsigned int alloc_flags; /* alloc flags of a direct compactor */ 616 const int highest_zoneidx; /* zone index of a direct compactor */ 617 enum migrate_mode mode; /* Async or sync migration mode */ 618 bool ignore_skip_hint; /* Scan blocks even if marked skip */ 619 bool no_set_skip_hint; /* Don't mark blocks for skipping */ 620 bool ignore_block_suitable; /* Scan blocks considered unsuitable */ 621 bool direct_compaction; /* False from kcompactd or /proc/... */ 622 bool proactive_compaction; /* kcompactd proactive compaction */ 623 bool whole_zone; /* Whole zone should/has been scanned */ 624 bool contended; /* Signal lock contention */ 625 bool finish_pageblock; /* Scan the remainder of a pageblock. Used 626 * when there are potentially transient 627 * isolation or migration failures to 628 * ensure forward progress. 629 */ 630 bool alloc_contig; /* alloc_contig_range allocation */ 631 }; 632 633 /* 634 * Used in direct compaction when a page should be taken from the freelists 635 * immediately when one is created during the free path. 636 */ 637 struct capture_control { 638 struct compact_control *cc; 639 struct page *page; 640 }; 641 642 unsigned long 643 isolate_freepages_range(struct compact_control *cc, 644 unsigned long start_pfn, unsigned long end_pfn); 645 int 646 isolate_migratepages_range(struct compact_control *cc, 647 unsigned long low_pfn, unsigned long end_pfn); 648 649 int __alloc_contig_migrate_range(struct compact_control *cc, 650 unsigned long start, unsigned long end, 651 int migratetype); 652 653 /* Free whole pageblock and set its migration type to MIGRATE_CMA. */ 654 void init_cma_reserved_pageblock(struct page *page); 655 656 #endif /* CONFIG_COMPACTION || CONFIG_CMA */ 657 658 int find_suitable_fallback(struct free_area *area, unsigned int order, 659 int migratetype, bool only_stealable, bool *can_steal); 660 661 static inline bool free_area_empty(struct free_area *area, int migratetype) 662 { 663 return list_empty(&area->free_list[migratetype]); 664 } 665 666 /* 667 * These three helpers classifies VMAs for virtual memory accounting. 668 */ 669 670 /* 671 * Executable code area - executable, not writable, not stack 672 */ 673 static inline bool is_exec_mapping(vm_flags_t flags) 674 { 675 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; 676 } 677 678 /* 679 * Stack area (including shadow stacks) 680 * 681 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: 682 * do_mmap() forbids all other combinations. 683 */ 684 static inline bool is_stack_mapping(vm_flags_t flags) 685 { 686 return ((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK); 687 } 688 689 /* 690 * Data area - private, writable, not stack 691 */ 692 static inline bool is_data_mapping(vm_flags_t flags) 693 { 694 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; 695 } 696 697 /* mm/util.c */ 698 struct anon_vma *folio_anon_vma(struct folio *folio); 699 700 #ifdef CONFIG_MMU 701 void unmap_mapping_folio(struct folio *folio); 702 extern long populate_vma_page_range(struct vm_area_struct *vma, 703 unsigned long start, unsigned long end, int *locked); 704 extern long faultin_page_range(struct mm_struct *mm, unsigned long start, 705 unsigned long end, bool write, int *locked); 706 extern bool mlock_future_ok(struct mm_struct *mm, unsigned long flags, 707 unsigned long bytes); 708 709 /* 710 * NOTE: This function can't tell whether the folio is "fully mapped" in the 711 * range. 712 * "fully mapped" means all the pages of folio is associated with the page 713 * table of range while this function just check whether the folio range is 714 * within the range [start, end). Function caller needs to do page table 715 * check if it cares about the page table association. 716 * 717 * Typical usage (like mlock or madvise) is: 718 * Caller knows at least 1 page of folio is associated with page table of VMA 719 * and the range [start, end) is intersect with the VMA range. Caller wants 720 * to know whether the folio is fully associated with the range. It calls 721 * this function to check whether the folio is in the range first. Then checks 722 * the page table to know whether the folio is fully mapped to the range. 723 */ 724 static inline bool 725 folio_within_range(struct folio *folio, struct vm_area_struct *vma, 726 unsigned long start, unsigned long end) 727 { 728 pgoff_t pgoff, addr; 729 unsigned long vma_pglen = vma_pages(vma); 730 731 VM_WARN_ON_FOLIO(folio_test_ksm(folio), folio); 732 if (start > end) 733 return false; 734 735 if (start < vma->vm_start) 736 start = vma->vm_start; 737 738 if (end > vma->vm_end) 739 end = vma->vm_end; 740 741 pgoff = folio_pgoff(folio); 742 743 /* if folio start address is not in vma range */ 744 if (!in_range(pgoff, vma->vm_pgoff, vma_pglen)) 745 return false; 746 747 addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); 748 749 return !(addr < start || end - addr < folio_size(folio)); 750 } 751 752 static inline bool 753 folio_within_vma(struct folio *folio, struct vm_area_struct *vma) 754 { 755 return folio_within_range(folio, vma, vma->vm_start, vma->vm_end); 756 } 757 758 /* 759 * mlock_vma_folio() and munlock_vma_folio(): 760 * should be called with vma's mmap_lock held for read or write, 761 * under page table lock for the pte/pmd being added or removed. 762 * 763 * mlock is usually called at the end of folio_add_*_rmap_*(), munlock at 764 * the end of folio_remove_rmap_*(); but new anon folios are managed by 765 * folio_add_lru_vma() calling mlock_new_folio(). 766 */ 767 void mlock_folio(struct folio *folio); 768 static inline void mlock_vma_folio(struct folio *folio, 769 struct vm_area_struct *vma) 770 { 771 /* 772 * The VM_SPECIAL check here serves two purposes. 773 * 1) VM_IO check prevents migration from double-counting during mlock. 774 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED 775 * is never left set on a VM_SPECIAL vma, there is an interval while 776 * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may 777 * still be set while VM_SPECIAL bits are added: so ignore it then. 778 */ 779 if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED)) 780 mlock_folio(folio); 781 } 782 783 void munlock_folio(struct folio *folio); 784 static inline void munlock_vma_folio(struct folio *folio, 785 struct vm_area_struct *vma) 786 { 787 /* 788 * munlock if the function is called. Ideally, we should only 789 * do munlock if any page of folio is unmapped from VMA and 790 * cause folio not fully mapped to VMA. 791 * 792 * But it's not easy to confirm that's the situation. So we 793 * always munlock the folio and page reclaim will correct it 794 * if it's wrong. 795 */ 796 if (unlikely(vma->vm_flags & VM_LOCKED)) 797 munlock_folio(folio); 798 } 799 800 void mlock_new_folio(struct folio *folio); 801 bool need_mlock_drain(int cpu); 802 void mlock_drain_local(void); 803 void mlock_drain_remote(int cpu); 804 805 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); 806 807 /* 808 * Return the start of user virtual address at the specific offset within 809 * a vma. 810 */ 811 static inline unsigned long 812 vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages, 813 struct vm_area_struct *vma) 814 { 815 unsigned long address; 816 817 if (pgoff >= vma->vm_pgoff) { 818 address = vma->vm_start + 819 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); 820 /* Check for address beyond vma (or wrapped through 0?) */ 821 if (address < vma->vm_start || address >= vma->vm_end) 822 address = -EFAULT; 823 } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) { 824 /* Test above avoids possibility of wrap to 0 on 32-bit */ 825 address = vma->vm_start; 826 } else { 827 address = -EFAULT; 828 } 829 return address; 830 } 831 832 /* 833 * Return the start of user virtual address of a page within a vma. 834 * Returns -EFAULT if all of the page is outside the range of vma. 835 * If page is a compound head, the entire compound page is considered. 836 */ 837 static inline unsigned long 838 vma_address(struct page *page, struct vm_area_struct *vma) 839 { 840 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */ 841 return vma_pgoff_address(page_to_pgoff(page), compound_nr(page), vma); 842 } 843 844 /* 845 * Then at what user virtual address will none of the range be found in vma? 846 * Assumes that vma_address() already returned a good starting address. 847 */ 848 static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw) 849 { 850 struct vm_area_struct *vma = pvmw->vma; 851 pgoff_t pgoff; 852 unsigned long address; 853 854 /* Common case, plus ->pgoff is invalid for KSM */ 855 if (pvmw->nr_pages == 1) 856 return pvmw->address + PAGE_SIZE; 857 858 pgoff = pvmw->pgoff + pvmw->nr_pages; 859 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); 860 /* Check for address beyond vma (or wrapped through 0?) */ 861 if (address < vma->vm_start || address > vma->vm_end) 862 address = vma->vm_end; 863 return address; 864 } 865 866 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf, 867 struct file *fpin) 868 { 869 int flags = vmf->flags; 870 871 if (fpin) 872 return fpin; 873 874 /* 875 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or 876 * anything, so we only pin the file and drop the mmap_lock if only 877 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt. 878 */ 879 if (fault_flag_allow_retry_first(flags) && 880 !(flags & FAULT_FLAG_RETRY_NOWAIT)) { 881 fpin = get_file(vmf->vma->vm_file); 882 release_fault_lock(vmf); 883 } 884 return fpin; 885 } 886 #else /* !CONFIG_MMU */ 887 static inline void unmap_mapping_folio(struct folio *folio) { } 888 static inline void mlock_new_folio(struct folio *folio) { } 889 static inline bool need_mlock_drain(int cpu) { return false; } 890 static inline void mlock_drain_local(void) { } 891 static inline void mlock_drain_remote(int cpu) { } 892 static inline void vunmap_range_noflush(unsigned long start, unsigned long end) 893 { 894 } 895 #endif /* !CONFIG_MMU */ 896 897 /* Memory initialisation debug and verification */ 898 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT 899 DECLARE_STATIC_KEY_TRUE(deferred_pages); 900 901 bool __init deferred_grow_zone(struct zone *zone, unsigned int order); 902 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ 903 904 enum mminit_level { 905 MMINIT_WARNING, 906 MMINIT_VERIFY, 907 MMINIT_TRACE 908 }; 909 910 #ifdef CONFIG_DEBUG_MEMORY_INIT 911 912 extern int mminit_loglevel; 913 914 #define mminit_dprintk(level, prefix, fmt, arg...) \ 915 do { \ 916 if (level < mminit_loglevel) { \ 917 if (level <= MMINIT_WARNING) \ 918 pr_warn("mminit::" prefix " " fmt, ##arg); \ 919 else \ 920 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ 921 } \ 922 } while (0) 923 924 extern void mminit_verify_pageflags_layout(void); 925 extern void mminit_verify_zonelist(void); 926 #else 927 928 static inline void mminit_dprintk(enum mminit_level level, 929 const char *prefix, const char *fmt, ...) 930 { 931 } 932 933 static inline void mminit_verify_pageflags_layout(void) 934 { 935 } 936 937 static inline void mminit_verify_zonelist(void) 938 { 939 } 940 #endif /* CONFIG_DEBUG_MEMORY_INIT */ 941 942 #define NODE_RECLAIM_NOSCAN -2 943 #define NODE_RECLAIM_FULL -1 944 #define NODE_RECLAIM_SOME 0 945 #define NODE_RECLAIM_SUCCESS 1 946 947 #ifdef CONFIG_NUMA 948 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); 949 extern int find_next_best_node(int node, nodemask_t *used_node_mask); 950 #else 951 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask, 952 unsigned int order) 953 { 954 return NODE_RECLAIM_NOSCAN; 955 } 956 static inline int find_next_best_node(int node, nodemask_t *used_node_mask) 957 { 958 return NUMA_NO_NODE; 959 } 960 #endif 961 962 /* 963 * mm/memory-failure.c 964 */ 965 extern int hwpoison_filter(struct page *p); 966 967 extern u32 hwpoison_filter_dev_major; 968 extern u32 hwpoison_filter_dev_minor; 969 extern u64 hwpoison_filter_flags_mask; 970 extern u64 hwpoison_filter_flags_value; 971 extern u64 hwpoison_filter_memcg; 972 extern u32 hwpoison_filter_enable; 973 974 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long, 975 unsigned long, unsigned long, 976 unsigned long, unsigned long); 977 978 extern void set_pageblock_order(void); 979 unsigned long reclaim_pages(struct list_head *folio_list, bool ignore_references); 980 unsigned int reclaim_clean_pages_from_list(struct zone *zone, 981 struct list_head *folio_list); 982 /* The ALLOC_WMARK bits are used as an index to zone->watermark */ 983 #define ALLOC_WMARK_MIN WMARK_MIN 984 #define ALLOC_WMARK_LOW WMARK_LOW 985 #define ALLOC_WMARK_HIGH WMARK_HIGH 986 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ 987 988 /* Mask to get the watermark bits */ 989 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) 990 991 /* 992 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we 993 * cannot assume a reduced access to memory reserves is sufficient for 994 * !MMU 995 */ 996 #ifdef CONFIG_MMU 997 #define ALLOC_OOM 0x08 998 #else 999 #define ALLOC_OOM ALLOC_NO_WATERMARKS 1000 #endif 1001 1002 #define ALLOC_NON_BLOCK 0x10 /* Caller cannot block. Allow access 1003 * to 25% of the min watermark or 1004 * 62.5% if __GFP_HIGH is set. 1005 */ 1006 #define ALLOC_MIN_RESERVE 0x20 /* __GFP_HIGH set. Allow access to 50% 1007 * of the min watermark. 1008 */ 1009 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ 1010 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ 1011 #ifdef CONFIG_ZONE_DMA32 1012 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */ 1013 #else 1014 #define ALLOC_NOFRAGMENT 0x0 1015 #endif 1016 #define ALLOC_HIGHATOMIC 0x200 /* Allows access to MIGRATE_HIGHATOMIC */ 1017 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */ 1018 1019 /* Flags that allow allocations below the min watermark. */ 1020 #define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM) 1021 1022 enum ttu_flags; 1023 struct tlbflush_unmap_batch; 1024 1025 1026 /* 1027 * only for MM internal work items which do not depend on 1028 * any allocations or locks which might depend on allocations 1029 */ 1030 extern struct workqueue_struct *mm_percpu_wq; 1031 1032 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH 1033 void try_to_unmap_flush(void); 1034 void try_to_unmap_flush_dirty(void); 1035 void flush_tlb_batched_pending(struct mm_struct *mm); 1036 #else 1037 static inline void try_to_unmap_flush(void) 1038 { 1039 } 1040 static inline void try_to_unmap_flush_dirty(void) 1041 { 1042 } 1043 static inline void flush_tlb_batched_pending(struct mm_struct *mm) 1044 { 1045 } 1046 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ 1047 1048 extern const struct trace_print_flags pageflag_names[]; 1049 extern const struct trace_print_flags pagetype_names[]; 1050 extern const struct trace_print_flags vmaflag_names[]; 1051 extern const struct trace_print_flags gfpflag_names[]; 1052 1053 static inline bool is_migrate_highatomic(enum migratetype migratetype) 1054 { 1055 return migratetype == MIGRATE_HIGHATOMIC; 1056 } 1057 1058 void setup_zone_pageset(struct zone *zone); 1059 1060 struct migration_target_control { 1061 int nid; /* preferred node id */ 1062 nodemask_t *nmask; 1063 gfp_t gfp_mask; 1064 enum migrate_reason reason; 1065 }; 1066 1067 /* 1068 * mm/filemap.c 1069 */ 1070 size_t splice_folio_into_pipe(struct pipe_inode_info *pipe, 1071 struct folio *folio, loff_t fpos, size_t size); 1072 1073 /* 1074 * mm/vmalloc.c 1075 */ 1076 #ifdef CONFIG_MMU 1077 void __init vmalloc_init(void); 1078 int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end, 1079 pgprot_t prot, struct page **pages, unsigned int page_shift); 1080 #else 1081 static inline void vmalloc_init(void) 1082 { 1083 } 1084 1085 static inline 1086 int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end, 1087 pgprot_t prot, struct page **pages, unsigned int page_shift) 1088 { 1089 return -EINVAL; 1090 } 1091 #endif 1092 1093 int __must_check __vmap_pages_range_noflush(unsigned long addr, 1094 unsigned long end, pgprot_t prot, 1095 struct page **pages, unsigned int page_shift); 1096 1097 void vunmap_range_noflush(unsigned long start, unsigned long end); 1098 1099 void __vunmap_range_noflush(unsigned long start, unsigned long end); 1100 1101 int numa_migrate_prep(struct folio *folio, struct vm_fault *vmf, 1102 unsigned long addr, int page_nid, int *flags); 1103 1104 void free_zone_device_page(struct page *page); 1105 int migrate_device_coherent_page(struct page *page); 1106 1107 /* 1108 * mm/gup.c 1109 */ 1110 struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags); 1111 int __must_check try_grab_page(struct page *page, unsigned int flags); 1112 1113 /* 1114 * mm/huge_memory.c 1115 */ 1116 void touch_pud(struct vm_area_struct *vma, unsigned long addr, 1117 pud_t *pud, bool write); 1118 void touch_pmd(struct vm_area_struct *vma, unsigned long addr, 1119 pmd_t *pmd, bool write); 1120 1121 /* 1122 * mm/mmap.c 1123 */ 1124 struct vm_area_struct *vma_merge_extend(struct vma_iterator *vmi, 1125 struct vm_area_struct *vma, 1126 unsigned long delta); 1127 1128 enum { 1129 /* mark page accessed */ 1130 FOLL_TOUCH = 1 << 16, 1131 /* a retry, previous pass started an IO */ 1132 FOLL_TRIED = 1 << 17, 1133 /* we are working on non-current tsk/mm */ 1134 FOLL_REMOTE = 1 << 18, 1135 /* pages must be released via unpin_user_page */ 1136 FOLL_PIN = 1 << 19, 1137 /* gup_fast: prevent fall-back to slow gup */ 1138 FOLL_FAST_ONLY = 1 << 20, 1139 /* allow unlocking the mmap lock */ 1140 FOLL_UNLOCKABLE = 1 << 21, 1141 /* VMA lookup+checks compatible with MADV_POPULATE_(READ|WRITE) */ 1142 FOLL_MADV_POPULATE = 1 << 22, 1143 }; 1144 1145 #define INTERNAL_GUP_FLAGS (FOLL_TOUCH | FOLL_TRIED | FOLL_REMOTE | FOLL_PIN | \ 1146 FOLL_FAST_ONLY | FOLL_UNLOCKABLE | \ 1147 FOLL_MADV_POPULATE) 1148 1149 /* 1150 * Indicates for which pages that are write-protected in the page table, 1151 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the 1152 * GUP pin will remain consistent with the pages mapped into the page tables 1153 * of the MM. 1154 * 1155 * Temporary unmapping of PageAnonExclusive() pages or clearing of 1156 * PageAnonExclusive() has to protect against concurrent GUP: 1157 * * Ordinary GUP: Using the PT lock 1158 * * GUP-fast and fork(): mm->write_protect_seq 1159 * * GUP-fast and KSM or temporary unmapping (swap, migration): see 1160 * folio_try_share_anon_rmap_*() 1161 * 1162 * Must be called with the (sub)page that's actually referenced via the 1163 * page table entry, which might not necessarily be the head page for a 1164 * PTE-mapped THP. 1165 * 1166 * If the vma is NULL, we're coming from the GUP-fast path and might have 1167 * to fallback to the slow path just to lookup the vma. 1168 */ 1169 static inline bool gup_must_unshare(struct vm_area_struct *vma, 1170 unsigned int flags, struct page *page) 1171 { 1172 /* 1173 * FOLL_WRITE is implicitly handled correctly as the page table entry 1174 * has to be writable -- and if it references (part of) an anonymous 1175 * folio, that part is required to be marked exclusive. 1176 */ 1177 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN) 1178 return false; 1179 /* 1180 * Note: PageAnon(page) is stable until the page is actually getting 1181 * freed. 1182 */ 1183 if (!PageAnon(page)) { 1184 /* 1185 * We only care about R/O long-term pining: R/O short-term 1186 * pinning does not have the semantics to observe successive 1187 * changes through the process page tables. 1188 */ 1189 if (!(flags & FOLL_LONGTERM)) 1190 return false; 1191 1192 /* We really need the vma ... */ 1193 if (!vma) 1194 return true; 1195 1196 /* 1197 * ... because we only care about writable private ("COW") 1198 * mappings where we have to break COW early. 1199 */ 1200 return is_cow_mapping(vma->vm_flags); 1201 } 1202 1203 /* Paired with a memory barrier in folio_try_share_anon_rmap_*(). */ 1204 if (IS_ENABLED(CONFIG_HAVE_FAST_GUP)) 1205 smp_rmb(); 1206 1207 /* 1208 * Note that PageKsm() pages cannot be exclusive, and consequently, 1209 * cannot get pinned. 1210 */ 1211 return !PageAnonExclusive(page); 1212 } 1213 1214 extern bool mirrored_kernelcore; 1215 extern bool memblock_has_mirror(void); 1216 1217 static __always_inline void vma_set_range(struct vm_area_struct *vma, 1218 unsigned long start, unsigned long end, 1219 pgoff_t pgoff) 1220 { 1221 vma->vm_start = start; 1222 vma->vm_end = end; 1223 vma->vm_pgoff = pgoff; 1224 } 1225 1226 static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma) 1227 { 1228 /* 1229 * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty 1230 * enablements, because when without soft-dirty being compiled in, 1231 * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY) 1232 * will be constantly true. 1233 */ 1234 if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY)) 1235 return false; 1236 1237 /* 1238 * Soft-dirty is kind of special: its tracking is enabled when the 1239 * vma flags not set. 1240 */ 1241 return !(vma->vm_flags & VM_SOFTDIRTY); 1242 } 1243 1244 static inline void vma_iter_config(struct vma_iterator *vmi, 1245 unsigned long index, unsigned long last) 1246 { 1247 __mas_set_range(&vmi->mas, index, last - 1); 1248 } 1249 1250 static inline void vma_iter_reset(struct vma_iterator *vmi) 1251 { 1252 mas_reset(&vmi->mas); 1253 } 1254 1255 static inline 1256 struct vm_area_struct *vma_iter_prev_range_limit(struct vma_iterator *vmi, unsigned long min) 1257 { 1258 return mas_prev_range(&vmi->mas, min); 1259 } 1260 1261 static inline 1262 struct vm_area_struct *vma_iter_next_range_limit(struct vma_iterator *vmi, unsigned long max) 1263 { 1264 return mas_next_range(&vmi->mas, max); 1265 } 1266 1267 static inline int vma_iter_area_lowest(struct vma_iterator *vmi, unsigned long min, 1268 unsigned long max, unsigned long size) 1269 { 1270 return mas_empty_area(&vmi->mas, min, max - 1, size); 1271 } 1272 1273 static inline int vma_iter_area_highest(struct vma_iterator *vmi, unsigned long min, 1274 unsigned long max, unsigned long size) 1275 { 1276 return mas_empty_area_rev(&vmi->mas, min, max - 1, size); 1277 } 1278 1279 /* 1280 * VMA Iterator functions shared between nommu and mmap 1281 */ 1282 static inline int vma_iter_prealloc(struct vma_iterator *vmi, 1283 struct vm_area_struct *vma) 1284 { 1285 return mas_preallocate(&vmi->mas, vma, GFP_KERNEL); 1286 } 1287 1288 static inline void vma_iter_clear(struct vma_iterator *vmi) 1289 { 1290 mas_store_prealloc(&vmi->mas, NULL); 1291 } 1292 1293 static inline struct vm_area_struct *vma_iter_load(struct vma_iterator *vmi) 1294 { 1295 return mas_walk(&vmi->mas); 1296 } 1297 1298 /* Store a VMA with preallocated memory */ 1299 static inline void vma_iter_store(struct vma_iterator *vmi, 1300 struct vm_area_struct *vma) 1301 { 1302 1303 #if defined(CONFIG_DEBUG_VM_MAPLE_TREE) 1304 if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start && 1305 vmi->mas.index > vma->vm_start)) { 1306 pr_warn("%lx > %lx\n store vma %lx-%lx\n into slot %lx-%lx\n", 1307 vmi->mas.index, vma->vm_start, vma->vm_start, 1308 vma->vm_end, vmi->mas.index, vmi->mas.last); 1309 } 1310 if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start && 1311 vmi->mas.last < vma->vm_start)) { 1312 pr_warn("%lx < %lx\nstore vma %lx-%lx\ninto slot %lx-%lx\n", 1313 vmi->mas.last, vma->vm_start, vma->vm_start, vma->vm_end, 1314 vmi->mas.index, vmi->mas.last); 1315 } 1316 #endif 1317 1318 if (vmi->mas.status != ma_start && 1319 ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start))) 1320 vma_iter_invalidate(vmi); 1321 1322 __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1); 1323 mas_store_prealloc(&vmi->mas, vma); 1324 } 1325 1326 static inline int vma_iter_store_gfp(struct vma_iterator *vmi, 1327 struct vm_area_struct *vma, gfp_t gfp) 1328 { 1329 if (vmi->mas.status != ma_start && 1330 ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start))) 1331 vma_iter_invalidate(vmi); 1332 1333 __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1); 1334 mas_store_gfp(&vmi->mas, vma, gfp); 1335 if (unlikely(mas_is_err(&vmi->mas))) 1336 return -ENOMEM; 1337 1338 return 0; 1339 } 1340 1341 /* 1342 * VMA lock generalization 1343 */ 1344 struct vma_prepare { 1345 struct vm_area_struct *vma; 1346 struct vm_area_struct *adj_next; 1347 struct file *file; 1348 struct address_space *mapping; 1349 struct anon_vma *anon_vma; 1350 struct vm_area_struct *insert; 1351 struct vm_area_struct *remove; 1352 struct vm_area_struct *remove2; 1353 }; 1354 1355 void __meminit __init_single_page(struct page *page, unsigned long pfn, 1356 unsigned long zone, int nid); 1357 1358 /* shrinker related functions */ 1359 unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, 1360 int priority); 1361 1362 #ifdef CONFIG_SHRINKER_DEBUG 1363 static inline __printf(2, 0) int shrinker_debugfs_name_alloc( 1364 struct shrinker *shrinker, const char *fmt, va_list ap) 1365 { 1366 shrinker->name = kvasprintf_const(GFP_KERNEL, fmt, ap); 1367 1368 return shrinker->name ? 0 : -ENOMEM; 1369 } 1370 1371 static inline void shrinker_debugfs_name_free(struct shrinker *shrinker) 1372 { 1373 kfree_const(shrinker->name); 1374 shrinker->name = NULL; 1375 } 1376 1377 extern int shrinker_debugfs_add(struct shrinker *shrinker); 1378 extern struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker, 1379 int *debugfs_id); 1380 extern void shrinker_debugfs_remove(struct dentry *debugfs_entry, 1381 int debugfs_id); 1382 #else /* CONFIG_SHRINKER_DEBUG */ 1383 static inline int shrinker_debugfs_add(struct shrinker *shrinker) 1384 { 1385 return 0; 1386 } 1387 static inline int shrinker_debugfs_name_alloc(struct shrinker *shrinker, 1388 const char *fmt, va_list ap) 1389 { 1390 return 0; 1391 } 1392 static inline void shrinker_debugfs_name_free(struct shrinker *shrinker) 1393 { 1394 } 1395 static inline struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker, 1396 int *debugfs_id) 1397 { 1398 *debugfs_id = -1; 1399 return NULL; 1400 } 1401 static inline void shrinker_debugfs_remove(struct dentry *debugfs_entry, 1402 int debugfs_id) 1403 { 1404 } 1405 #endif /* CONFIG_SHRINKER_DEBUG */ 1406 1407 /* Only track the nodes of mappings with shadow entries */ 1408 void workingset_update_node(struct xa_node *node); 1409 extern struct list_lru shadow_nodes; 1410 1411 #endif /* __MM_INTERNAL_H */ 1412