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_ATOMIC|__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 void page_writeback_init(void); 39 40 static inline void *folio_raw_mapping(struct folio *folio) 41 { 42 unsigned long mapping = (unsigned long)folio->mapping; 43 44 return (void *)(mapping & ~PAGE_MAPPING_FLAGS); 45 } 46 47 void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio, 48 int nr_throttled); 49 static inline void acct_reclaim_writeback(struct folio *folio) 50 { 51 pg_data_t *pgdat = folio_pgdat(folio); 52 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled); 53 54 if (nr_throttled) 55 __acct_reclaim_writeback(pgdat, folio, nr_throttled); 56 } 57 58 static inline void wake_throttle_isolated(pg_data_t *pgdat) 59 { 60 wait_queue_head_t *wqh; 61 62 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED]; 63 if (waitqueue_active(wqh)) 64 wake_up(wqh); 65 } 66 67 vm_fault_t do_swap_page(struct vm_fault *vmf); 68 void folio_rotate_reclaimable(struct folio *folio); 69 bool __folio_end_writeback(struct folio *folio); 70 void deactivate_file_folio(struct folio *folio); 71 72 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, 73 unsigned long floor, unsigned long ceiling); 74 void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte); 75 76 struct zap_details; 77 void unmap_page_range(struct mmu_gather *tlb, 78 struct vm_area_struct *vma, 79 unsigned long addr, unsigned long end, 80 struct zap_details *details); 81 82 void page_cache_ra_order(struct readahead_control *, struct file_ra_state *, 83 unsigned int order); 84 void force_page_cache_ra(struct readahead_control *, unsigned long nr); 85 static inline void force_page_cache_readahead(struct address_space *mapping, 86 struct file *file, pgoff_t index, unsigned long nr_to_read) 87 { 88 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index); 89 force_page_cache_ra(&ractl, nr_to_read); 90 } 91 92 unsigned find_lock_entries(struct address_space *mapping, pgoff_t start, 93 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); 94 unsigned find_get_entries(struct address_space *mapping, pgoff_t start, 95 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); 96 void filemap_free_folio(struct address_space *mapping, struct folio *folio); 97 int truncate_inode_folio(struct address_space *mapping, struct folio *folio); 98 bool truncate_inode_partial_folio(struct folio *folio, loff_t start, 99 loff_t end); 100 long invalidate_inode_page(struct page *page); 101 unsigned long invalidate_mapping_pagevec(struct address_space *mapping, 102 pgoff_t start, pgoff_t end, unsigned long *nr_pagevec); 103 104 /** 105 * folio_evictable - Test whether a folio is evictable. 106 * @folio: The folio to test. 107 * 108 * Test whether @folio is evictable -- i.e., should be placed on 109 * active/inactive lists vs unevictable list. 110 * 111 * Reasons folio might not be evictable: 112 * 1. folio's mapping marked unevictable 113 * 2. One of the pages in the folio is part of an mlocked VMA 114 */ 115 static inline bool folio_evictable(struct folio *folio) 116 { 117 bool ret; 118 119 /* Prevent address_space of inode and swap cache from being freed */ 120 rcu_read_lock(); 121 ret = !mapping_unevictable(folio_mapping(folio)) && 122 !folio_test_mlocked(folio); 123 rcu_read_unlock(); 124 return ret; 125 } 126 127 static inline bool page_evictable(struct page *page) 128 { 129 bool ret; 130 131 /* Prevent address_space of inode and swap cache from being freed */ 132 rcu_read_lock(); 133 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page); 134 rcu_read_unlock(); 135 return ret; 136 } 137 138 /* 139 * Turn a non-refcounted page (->_refcount == 0) into refcounted with 140 * a count of one. 141 */ 142 static inline void set_page_refcounted(struct page *page) 143 { 144 VM_BUG_ON_PAGE(PageTail(page), page); 145 VM_BUG_ON_PAGE(page_ref_count(page), page); 146 set_page_count(page, 1); 147 } 148 149 extern unsigned long highest_memmap_pfn; 150 151 /* 152 * Maximum number of reclaim retries without progress before the OOM 153 * killer is consider the only way forward. 154 */ 155 #define MAX_RECLAIM_RETRIES 16 156 157 /* 158 * in mm/early_ioremap.c 159 */ 160 pgprot_t __init early_memremap_pgprot_adjust(resource_size_t phys_addr, 161 unsigned long size, pgprot_t prot); 162 163 /* 164 * in mm/vmscan.c: 165 */ 166 int isolate_lru_page(struct page *page); 167 int folio_isolate_lru(struct folio *folio); 168 void putback_lru_page(struct page *page); 169 void folio_putback_lru(struct folio *folio); 170 extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason); 171 172 /* 173 * in mm/rmap.c: 174 */ 175 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); 176 177 /* 178 * in mm/page_alloc.c 179 */ 180 181 /* 182 * Structure for holding the mostly immutable allocation parameters passed 183 * between functions involved in allocations, including the alloc_pages* 184 * family of functions. 185 * 186 * nodemask, migratetype and highest_zoneidx are initialized only once in 187 * __alloc_pages() and then never change. 188 * 189 * zonelist, preferred_zone and highest_zoneidx are set first in 190 * __alloc_pages() for the fast path, and might be later changed 191 * in __alloc_pages_slowpath(). All other functions pass the whole structure 192 * by a const pointer. 193 */ 194 struct alloc_context { 195 struct zonelist *zonelist; 196 nodemask_t *nodemask; 197 struct zoneref *preferred_zoneref; 198 int migratetype; 199 200 /* 201 * highest_zoneidx represents highest usable zone index of 202 * the allocation request. Due to the nature of the zone, 203 * memory on lower zone than the highest_zoneidx will be 204 * protected by lowmem_reserve[highest_zoneidx]. 205 * 206 * highest_zoneidx is also used by reclaim/compaction to limit 207 * the target zone since higher zone than this index cannot be 208 * usable for this allocation request. 209 */ 210 enum zone_type highest_zoneidx; 211 bool spread_dirty_pages; 212 }; 213 214 /* 215 * This function returns the order of a free page in the buddy system. In 216 * general, page_zone(page)->lock must be held by the caller to prevent the 217 * page from being allocated in parallel and returning garbage as the order. 218 * If a caller does not hold page_zone(page)->lock, it must guarantee that the 219 * page cannot be allocated or merged in parallel. Alternatively, it must 220 * handle invalid values gracefully, and use buddy_order_unsafe() below. 221 */ 222 static inline unsigned int buddy_order(struct page *page) 223 { 224 /* PageBuddy() must be checked by the caller */ 225 return page_private(page); 226 } 227 228 /* 229 * Like buddy_order(), but for callers who cannot afford to hold the zone lock. 230 * PageBuddy() should be checked first by the caller to minimize race window, 231 * and invalid values must be handled gracefully. 232 * 233 * READ_ONCE is used so that if the caller assigns the result into a local 234 * variable and e.g. tests it for valid range before using, the compiler cannot 235 * decide to remove the variable and inline the page_private(page) multiple 236 * times, potentially observing different values in the tests and the actual 237 * use of the result. 238 */ 239 #define buddy_order_unsafe(page) READ_ONCE(page_private(page)) 240 241 /* 242 * This function checks whether a page is free && is the buddy 243 * we can coalesce a page and its buddy if 244 * (a) the buddy is not in a hole (check before calling!) && 245 * (b) the buddy is in the buddy system && 246 * (c) a page and its buddy have the same order && 247 * (d) a page and its buddy are in the same zone. 248 * 249 * For recording whether a page is in the buddy system, we set PageBuddy. 250 * Setting, clearing, and testing PageBuddy is serialized by zone->lock. 251 * 252 * For recording page's order, we use page_private(page). 253 */ 254 static inline bool page_is_buddy(struct page *page, struct page *buddy, 255 unsigned int order) 256 { 257 if (!page_is_guard(buddy) && !PageBuddy(buddy)) 258 return false; 259 260 if (buddy_order(buddy) != order) 261 return false; 262 263 /* 264 * zone check is done late to avoid uselessly calculating 265 * zone/node ids for pages that could never merge. 266 */ 267 if (page_zone_id(page) != page_zone_id(buddy)) 268 return false; 269 270 VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy); 271 272 return true; 273 } 274 275 /* 276 * Locate the struct page for both the matching buddy in our 277 * pair (buddy1) and the combined O(n+1) page they form (page). 278 * 279 * 1) Any buddy B1 will have an order O twin B2 which satisfies 280 * the following equation: 281 * B2 = B1 ^ (1 << O) 282 * For example, if the starting buddy (buddy2) is #8 its order 283 * 1 buddy is #10: 284 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 285 * 286 * 2) Any buddy B will have an order O+1 parent P which 287 * satisfies the following equation: 288 * P = B & ~(1 << O) 289 * 290 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER 291 */ 292 static inline unsigned long 293 __find_buddy_pfn(unsigned long page_pfn, unsigned int order) 294 { 295 return page_pfn ^ (1 << order); 296 } 297 298 /* 299 * Find the buddy of @page and validate it. 300 * @page: The input page 301 * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the 302 * function is used in the performance-critical __free_one_page(). 303 * @order: The order of the page 304 * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to 305 * page_to_pfn(). 306 * 307 * The found buddy can be a non PageBuddy, out of @page's zone, or its order is 308 * not the same as @page. The validation is necessary before use it. 309 * 310 * Return: the found buddy page or NULL if not found. 311 */ 312 static inline struct page *find_buddy_page_pfn(struct page *page, 313 unsigned long pfn, unsigned int order, unsigned long *buddy_pfn) 314 { 315 unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order); 316 struct page *buddy; 317 318 buddy = page + (__buddy_pfn - pfn); 319 if (buddy_pfn) 320 *buddy_pfn = __buddy_pfn; 321 322 if (page_is_buddy(page, buddy, order)) 323 return buddy; 324 return NULL; 325 } 326 327 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, 328 unsigned long end_pfn, struct zone *zone); 329 330 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn, 331 unsigned long end_pfn, struct zone *zone) 332 { 333 if (zone->contiguous) 334 return pfn_to_page(start_pfn); 335 336 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone); 337 } 338 339 extern int __isolate_free_page(struct page *page, unsigned int order); 340 extern void __putback_isolated_page(struct page *page, unsigned int order, 341 int mt); 342 extern void memblock_free_pages(struct page *page, unsigned long pfn, 343 unsigned int order); 344 extern void __free_pages_core(struct page *page, unsigned int order); 345 extern void prep_compound_page(struct page *page, unsigned int order); 346 extern void post_alloc_hook(struct page *page, unsigned int order, 347 gfp_t gfp_flags); 348 extern int user_min_free_kbytes; 349 350 extern void free_unref_page(struct page *page, unsigned int order); 351 extern void free_unref_page_list(struct list_head *list); 352 353 extern void zone_pcp_update(struct zone *zone, int cpu_online); 354 extern void zone_pcp_reset(struct zone *zone); 355 extern void zone_pcp_disable(struct zone *zone); 356 extern void zone_pcp_enable(struct zone *zone); 357 358 extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, 359 phys_addr_t min_addr, 360 int nid, bool exact_nid); 361 362 #if defined CONFIG_COMPACTION || defined CONFIG_CMA 363 364 /* 365 * in mm/compaction.c 366 */ 367 /* 368 * compact_control is used to track pages being migrated and the free pages 369 * they are being migrated to during memory compaction. The free_pfn starts 370 * at the end of a zone and migrate_pfn begins at the start. Movable pages 371 * are moved to the end of a zone during a compaction run and the run 372 * completes when free_pfn <= migrate_pfn 373 */ 374 struct compact_control { 375 struct list_head freepages; /* List of free pages to migrate to */ 376 struct list_head migratepages; /* List of pages being migrated */ 377 unsigned int nr_freepages; /* Number of isolated free pages */ 378 unsigned int nr_migratepages; /* Number of pages to migrate */ 379 unsigned long free_pfn; /* isolate_freepages search base */ 380 /* 381 * Acts as an in/out parameter to page isolation for migration. 382 * isolate_migratepages uses it as a search base. 383 * isolate_migratepages_block will update the value to the next pfn 384 * after the last isolated one. 385 */ 386 unsigned long migrate_pfn; 387 unsigned long fast_start_pfn; /* a pfn to start linear scan from */ 388 struct zone *zone; 389 unsigned long total_migrate_scanned; 390 unsigned long total_free_scanned; 391 unsigned short fast_search_fail;/* failures to use free list searches */ 392 short search_order; /* order to start a fast search at */ 393 const gfp_t gfp_mask; /* gfp mask of a direct compactor */ 394 int order; /* order a direct compactor needs */ 395 int migratetype; /* migratetype of direct compactor */ 396 const unsigned int alloc_flags; /* alloc flags of a direct compactor */ 397 const int highest_zoneidx; /* zone index of a direct compactor */ 398 enum migrate_mode mode; /* Async or sync migration mode */ 399 bool ignore_skip_hint; /* Scan blocks even if marked skip */ 400 bool no_set_skip_hint; /* Don't mark blocks for skipping */ 401 bool ignore_block_suitable; /* Scan blocks considered unsuitable */ 402 bool direct_compaction; /* False from kcompactd or /proc/... */ 403 bool proactive_compaction; /* kcompactd proactive compaction */ 404 bool whole_zone; /* Whole zone should/has been scanned */ 405 bool contended; /* Signal lock or sched contention */ 406 bool rescan; /* Rescanning the same pageblock */ 407 bool alloc_contig; /* alloc_contig_range allocation */ 408 }; 409 410 /* 411 * Used in direct compaction when a page should be taken from the freelists 412 * immediately when one is created during the free path. 413 */ 414 struct capture_control { 415 struct compact_control *cc; 416 struct page *page; 417 }; 418 419 unsigned long 420 isolate_freepages_range(struct compact_control *cc, 421 unsigned long start_pfn, unsigned long end_pfn); 422 int 423 isolate_migratepages_range(struct compact_control *cc, 424 unsigned long low_pfn, unsigned long end_pfn); 425 #endif 426 int find_suitable_fallback(struct free_area *area, unsigned int order, 427 int migratetype, bool only_stealable, bool *can_steal); 428 429 /* 430 * These three helpers classifies VMAs for virtual memory accounting. 431 */ 432 433 /* 434 * Executable code area - executable, not writable, not stack 435 */ 436 static inline bool is_exec_mapping(vm_flags_t flags) 437 { 438 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; 439 } 440 441 /* 442 * Stack area - automatically grows in one direction 443 * 444 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: 445 * do_mmap() forbids all other combinations. 446 */ 447 static inline bool is_stack_mapping(vm_flags_t flags) 448 { 449 return (flags & VM_STACK) == VM_STACK; 450 } 451 452 /* 453 * Data area - private, writable, not stack 454 */ 455 static inline bool is_data_mapping(vm_flags_t flags) 456 { 457 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; 458 } 459 460 /* mm/util.c */ 461 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, 462 struct vm_area_struct *prev); 463 void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma); 464 struct anon_vma *folio_anon_vma(struct folio *folio); 465 466 #ifdef CONFIG_MMU 467 void unmap_mapping_folio(struct folio *folio); 468 extern long populate_vma_page_range(struct vm_area_struct *vma, 469 unsigned long start, unsigned long end, int *locked); 470 extern long faultin_vma_page_range(struct vm_area_struct *vma, 471 unsigned long start, unsigned long end, 472 bool write, int *locked); 473 extern int mlock_future_check(struct mm_struct *mm, unsigned long flags, 474 unsigned long len); 475 /* 476 * mlock_vma_page() and munlock_vma_page(): 477 * should be called with vma's mmap_lock held for read or write, 478 * under page table lock for the pte/pmd being added or removed. 479 * 480 * mlock is usually called at the end of page_add_*_rmap(), 481 * munlock at the end of page_remove_rmap(); but new anon 482 * pages are managed by lru_cache_add_inactive_or_unevictable() 483 * calling mlock_new_page(). 484 * 485 * @compound is used to include pmd mappings of THPs, but filter out 486 * pte mappings of THPs, which cannot be consistently counted: a pte 487 * mapping of the THP head cannot be distinguished by the page alone. 488 */ 489 void mlock_folio(struct folio *folio); 490 static inline void mlock_vma_folio(struct folio *folio, 491 struct vm_area_struct *vma, bool compound) 492 { 493 /* 494 * The VM_SPECIAL check here serves two purposes. 495 * 1) VM_IO check prevents migration from double-counting during mlock. 496 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED 497 * is never left set on a VM_SPECIAL vma, there is an interval while 498 * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may 499 * still be set while VM_SPECIAL bits are added: so ignore it then. 500 */ 501 if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED) && 502 (compound || !folio_test_large(folio))) 503 mlock_folio(folio); 504 } 505 506 static inline void mlock_vma_page(struct page *page, 507 struct vm_area_struct *vma, bool compound) 508 { 509 mlock_vma_folio(page_folio(page), vma, compound); 510 } 511 512 void munlock_page(struct page *page); 513 static inline void munlock_vma_page(struct page *page, 514 struct vm_area_struct *vma, bool compound) 515 { 516 if (unlikely(vma->vm_flags & VM_LOCKED) && 517 (compound || !PageTransCompound(page))) 518 munlock_page(page); 519 } 520 void mlock_new_page(struct page *page); 521 bool need_mlock_page_drain(int cpu); 522 void mlock_page_drain_local(void); 523 void mlock_page_drain_remote(int cpu); 524 525 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); 526 527 /* 528 * At what user virtual address is page expected in vma? 529 * Returns -EFAULT if all of the page is outside the range of vma. 530 * If page is a compound head, the entire compound page is considered. 531 */ 532 static inline unsigned long 533 vma_address(struct page *page, struct vm_area_struct *vma) 534 { 535 pgoff_t pgoff; 536 unsigned long address; 537 538 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */ 539 pgoff = page_to_pgoff(page); 540 if (pgoff >= vma->vm_pgoff) { 541 address = vma->vm_start + 542 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); 543 /* Check for address beyond vma (or wrapped through 0?) */ 544 if (address < vma->vm_start || address >= vma->vm_end) 545 address = -EFAULT; 546 } else if (PageHead(page) && 547 pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) { 548 /* Test above avoids possibility of wrap to 0 on 32-bit */ 549 address = vma->vm_start; 550 } else { 551 address = -EFAULT; 552 } 553 return address; 554 } 555 556 /* 557 * Then at what user virtual address will none of the range be found in vma? 558 * Assumes that vma_address() already returned a good starting address. 559 */ 560 static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw) 561 { 562 struct vm_area_struct *vma = pvmw->vma; 563 pgoff_t pgoff; 564 unsigned long address; 565 566 /* Common case, plus ->pgoff is invalid for KSM */ 567 if (pvmw->nr_pages == 1) 568 return pvmw->address + PAGE_SIZE; 569 570 pgoff = pvmw->pgoff + pvmw->nr_pages; 571 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); 572 /* Check for address beyond vma (or wrapped through 0?) */ 573 if (address < vma->vm_start || address > vma->vm_end) 574 address = vma->vm_end; 575 return address; 576 } 577 578 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf, 579 struct file *fpin) 580 { 581 int flags = vmf->flags; 582 583 if (fpin) 584 return fpin; 585 586 /* 587 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or 588 * anything, so we only pin the file and drop the mmap_lock if only 589 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt. 590 */ 591 if (fault_flag_allow_retry_first(flags) && 592 !(flags & FAULT_FLAG_RETRY_NOWAIT)) { 593 fpin = get_file(vmf->vma->vm_file); 594 mmap_read_unlock(vmf->vma->vm_mm); 595 } 596 return fpin; 597 } 598 #else /* !CONFIG_MMU */ 599 static inline void unmap_mapping_folio(struct folio *folio) { } 600 static inline void mlock_vma_page(struct page *page, 601 struct vm_area_struct *vma, bool compound) { } 602 static inline void munlock_vma_page(struct page *page, 603 struct vm_area_struct *vma, bool compound) { } 604 static inline void mlock_new_page(struct page *page) { } 605 static inline bool need_mlock_page_drain(int cpu) { return false; } 606 static inline void mlock_page_drain_local(void) { } 607 static inline void mlock_page_drain_remote(int cpu) { } 608 static inline void vunmap_range_noflush(unsigned long start, unsigned long end) 609 { 610 } 611 #endif /* !CONFIG_MMU */ 612 613 /* 614 * Return the mem_map entry representing the 'offset' subpage within 615 * the maximally aligned gigantic page 'base'. Handle any discontiguity 616 * in the mem_map at MAX_ORDER_NR_PAGES boundaries. 617 */ 618 static inline struct page *mem_map_offset(struct page *base, int offset) 619 { 620 if (unlikely(offset >= MAX_ORDER_NR_PAGES)) 621 return nth_page(base, offset); 622 return base + offset; 623 } 624 625 /* 626 * Iterator over all subpages within the maximally aligned gigantic 627 * page 'base'. Handle any discontiguity in the mem_map. 628 */ 629 static inline struct page *mem_map_next(struct page *iter, 630 struct page *base, int offset) 631 { 632 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { 633 unsigned long pfn = page_to_pfn(base) + offset; 634 if (!pfn_valid(pfn)) 635 return NULL; 636 return pfn_to_page(pfn); 637 } 638 return iter + 1; 639 } 640 641 /* Memory initialisation debug and verification */ 642 enum mminit_level { 643 MMINIT_WARNING, 644 MMINIT_VERIFY, 645 MMINIT_TRACE 646 }; 647 648 #ifdef CONFIG_DEBUG_MEMORY_INIT 649 650 extern int mminit_loglevel; 651 652 #define mminit_dprintk(level, prefix, fmt, arg...) \ 653 do { \ 654 if (level < mminit_loglevel) { \ 655 if (level <= MMINIT_WARNING) \ 656 pr_warn("mminit::" prefix " " fmt, ##arg); \ 657 else \ 658 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ 659 } \ 660 } while (0) 661 662 extern void mminit_verify_pageflags_layout(void); 663 extern void mminit_verify_zonelist(void); 664 #else 665 666 static inline void mminit_dprintk(enum mminit_level level, 667 const char *prefix, const char *fmt, ...) 668 { 669 } 670 671 static inline void mminit_verify_pageflags_layout(void) 672 { 673 } 674 675 static inline void mminit_verify_zonelist(void) 676 { 677 } 678 #endif /* CONFIG_DEBUG_MEMORY_INIT */ 679 680 #define NODE_RECLAIM_NOSCAN -2 681 #define NODE_RECLAIM_FULL -1 682 #define NODE_RECLAIM_SOME 0 683 #define NODE_RECLAIM_SUCCESS 1 684 685 #ifdef CONFIG_NUMA 686 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); 687 extern int find_next_best_node(int node, nodemask_t *used_node_mask); 688 #else 689 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask, 690 unsigned int order) 691 { 692 return NODE_RECLAIM_NOSCAN; 693 } 694 static inline int find_next_best_node(int node, nodemask_t *used_node_mask) 695 { 696 return NUMA_NO_NODE; 697 } 698 #endif 699 700 extern int hwpoison_filter(struct page *p); 701 702 extern u32 hwpoison_filter_dev_major; 703 extern u32 hwpoison_filter_dev_minor; 704 extern u64 hwpoison_filter_flags_mask; 705 extern u64 hwpoison_filter_flags_value; 706 extern u64 hwpoison_filter_memcg; 707 extern u32 hwpoison_filter_enable; 708 709 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long, 710 unsigned long, unsigned long, 711 unsigned long, unsigned long); 712 713 extern void set_pageblock_order(void); 714 unsigned int reclaim_clean_pages_from_list(struct zone *zone, 715 struct list_head *page_list); 716 /* The ALLOC_WMARK bits are used as an index to zone->watermark */ 717 #define ALLOC_WMARK_MIN WMARK_MIN 718 #define ALLOC_WMARK_LOW WMARK_LOW 719 #define ALLOC_WMARK_HIGH WMARK_HIGH 720 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ 721 722 /* Mask to get the watermark bits */ 723 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) 724 725 /* 726 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we 727 * cannot assume a reduced access to memory reserves is sufficient for 728 * !MMU 729 */ 730 #ifdef CONFIG_MMU 731 #define ALLOC_OOM 0x08 732 #else 733 #define ALLOC_OOM ALLOC_NO_WATERMARKS 734 #endif 735 736 #define ALLOC_HARDER 0x10 /* try to alloc harder */ 737 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ 738 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ 739 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ 740 #ifdef CONFIG_ZONE_DMA32 741 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */ 742 #else 743 #define ALLOC_NOFRAGMENT 0x0 744 #endif 745 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */ 746 747 enum ttu_flags; 748 struct tlbflush_unmap_batch; 749 750 751 /* 752 * only for MM internal work items which do not depend on 753 * any allocations or locks which might depend on allocations 754 */ 755 extern struct workqueue_struct *mm_percpu_wq; 756 757 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH 758 void try_to_unmap_flush(void); 759 void try_to_unmap_flush_dirty(void); 760 void flush_tlb_batched_pending(struct mm_struct *mm); 761 #else 762 static inline void try_to_unmap_flush(void) 763 { 764 } 765 static inline void try_to_unmap_flush_dirty(void) 766 { 767 } 768 static inline void flush_tlb_batched_pending(struct mm_struct *mm) 769 { 770 } 771 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ 772 773 extern const struct trace_print_flags pageflag_names[]; 774 extern const struct trace_print_flags vmaflag_names[]; 775 extern const struct trace_print_flags gfpflag_names[]; 776 777 static inline bool is_migrate_highatomic(enum migratetype migratetype) 778 { 779 return migratetype == MIGRATE_HIGHATOMIC; 780 } 781 782 static inline bool is_migrate_highatomic_page(struct page *page) 783 { 784 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC; 785 } 786 787 void setup_zone_pageset(struct zone *zone); 788 789 struct migration_target_control { 790 int nid; /* preferred node id */ 791 nodemask_t *nmask; 792 gfp_t gfp_mask; 793 }; 794 795 /* 796 * mm/vmalloc.c 797 */ 798 #ifdef CONFIG_MMU 799 int vmap_pages_range_noflush(unsigned long addr, unsigned long end, 800 pgprot_t prot, struct page **pages, unsigned int page_shift); 801 #else 802 static inline 803 int vmap_pages_range_noflush(unsigned long addr, unsigned long end, 804 pgprot_t prot, struct page **pages, unsigned int page_shift) 805 { 806 return -EINVAL; 807 } 808 #endif 809 810 void vunmap_range_noflush(unsigned long start, unsigned long end); 811 812 int numa_migrate_prep(struct page *page, struct vm_area_struct *vma, 813 unsigned long addr, int page_nid, int *flags); 814 815 void free_zone_device_page(struct page *page); 816 817 /* 818 * mm/gup.c 819 */ 820 struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags); 821 822 DECLARE_PER_CPU(struct per_cpu_nodestat, boot_nodestats); 823 824 #endif /* __MM_INTERNAL_H */ 825