1 /* internal.h: mm/ internal definitions 2 * 3 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. 4 * Written by David Howells (dhowells@redhat.com) 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License 8 * as published by the Free Software Foundation; either version 9 * 2 of the License, or (at your option) any later version. 10 */ 11 #ifndef __MM_INTERNAL_H 12 #define __MM_INTERNAL_H 13 14 #include <linux/fs.h> 15 #include <linux/mm.h> 16 #include <linux/pagemap.h> 17 #include <linux/tracepoint-defs.h> 18 19 /* 20 * The set of flags that only affect watermark checking and reclaim 21 * behaviour. This is used by the MM to obey the caller constraints 22 * about IO, FS and watermark checking while ignoring placement 23 * hints such as HIGHMEM usage. 24 */ 25 #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\ 26 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\ 27 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC) 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 free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, 39 unsigned long floor, unsigned long ceiling); 40 41 void unmap_page_range(struct mmu_gather *tlb, 42 struct vm_area_struct *vma, 43 unsigned long addr, unsigned long end, 44 struct zap_details *details); 45 46 extern int __do_page_cache_readahead(struct address_space *mapping, 47 struct file *filp, pgoff_t offset, unsigned long nr_to_read, 48 unsigned long lookahead_size); 49 50 /* 51 * Submit IO for the read-ahead request in file_ra_state. 52 */ 53 static inline unsigned long ra_submit(struct file_ra_state *ra, 54 struct address_space *mapping, struct file *filp) 55 { 56 return __do_page_cache_readahead(mapping, filp, 57 ra->start, ra->size, ra->async_size); 58 } 59 60 /* 61 * Turn a non-refcounted page (->_count == 0) into refcounted with 62 * a count of one. 63 */ 64 static inline void set_page_refcounted(struct page *page) 65 { 66 VM_BUG_ON_PAGE(PageTail(page), page); 67 VM_BUG_ON_PAGE(page_ref_count(page), page); 68 set_page_count(page, 1); 69 } 70 71 extern unsigned long highest_memmap_pfn; 72 73 /* 74 * in mm/vmscan.c: 75 */ 76 extern int isolate_lru_page(struct page *page); 77 extern void putback_lru_page(struct page *page); 78 extern bool zone_reclaimable(struct zone *zone); 79 80 /* 81 * in mm/rmap.c: 82 */ 83 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); 84 85 /* 86 * in mm/page_alloc.c 87 */ 88 89 /* 90 * Structure for holding the mostly immutable allocation parameters passed 91 * between functions involved in allocations, including the alloc_pages* 92 * family of functions. 93 * 94 * nodemask, migratetype and high_zoneidx are initialized only once in 95 * __alloc_pages_nodemask() and then never change. 96 * 97 * zonelist, preferred_zone and classzone_idx are set first in 98 * __alloc_pages_nodemask() for the fast path, and might be later changed 99 * in __alloc_pages_slowpath(). All other functions pass the whole strucure 100 * by a const pointer. 101 */ 102 struct alloc_context { 103 struct zonelist *zonelist; 104 nodemask_t *nodemask; 105 struct zone *preferred_zone; 106 int classzone_idx; 107 int migratetype; 108 enum zone_type high_zoneidx; 109 bool spread_dirty_pages; 110 }; 111 112 /* 113 * Locate the struct page for both the matching buddy in our 114 * pair (buddy1) and the combined O(n+1) page they form (page). 115 * 116 * 1) Any buddy B1 will have an order O twin B2 which satisfies 117 * the following equation: 118 * B2 = B1 ^ (1 << O) 119 * For example, if the starting buddy (buddy2) is #8 its order 120 * 1 buddy is #10: 121 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 122 * 123 * 2) Any buddy B will have an order O+1 parent P which 124 * satisfies the following equation: 125 * P = B & ~(1 << O) 126 * 127 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER 128 */ 129 static inline unsigned long 130 __find_buddy_index(unsigned long page_idx, unsigned int order) 131 { 132 return page_idx ^ (1 << order); 133 } 134 135 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, 136 unsigned long end_pfn, struct zone *zone); 137 138 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn, 139 unsigned long end_pfn, struct zone *zone) 140 { 141 if (zone->contiguous) 142 return pfn_to_page(start_pfn); 143 144 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone); 145 } 146 147 extern int __isolate_free_page(struct page *page, unsigned int order); 148 extern void __free_pages_bootmem(struct page *page, unsigned long pfn, 149 unsigned int order); 150 extern void prep_compound_page(struct page *page, unsigned int order); 151 extern int user_min_free_kbytes; 152 153 #if defined CONFIG_COMPACTION || defined CONFIG_CMA 154 155 /* 156 * in mm/compaction.c 157 */ 158 /* 159 * compact_control is used to track pages being migrated and the free pages 160 * they are being migrated to during memory compaction. The free_pfn starts 161 * at the end of a zone and migrate_pfn begins at the start. Movable pages 162 * are moved to the end of a zone during a compaction run and the run 163 * completes when free_pfn <= migrate_pfn 164 */ 165 struct compact_control { 166 struct list_head freepages; /* List of free pages to migrate to */ 167 struct list_head migratepages; /* List of pages being migrated */ 168 unsigned long nr_freepages; /* Number of isolated free pages */ 169 unsigned long nr_migratepages; /* Number of pages to migrate */ 170 unsigned long free_pfn; /* isolate_freepages search base */ 171 unsigned long migrate_pfn; /* isolate_migratepages search base */ 172 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */ 173 enum migrate_mode mode; /* Async or sync migration mode */ 174 bool ignore_skip_hint; /* Scan blocks even if marked skip */ 175 bool direct_compaction; /* False from kcompactd or /proc/... */ 176 int order; /* order a direct compactor needs */ 177 const gfp_t gfp_mask; /* gfp mask of a direct compactor */ 178 const int alloc_flags; /* alloc flags of a direct compactor */ 179 const int classzone_idx; /* zone index of a direct compactor */ 180 struct zone *zone; 181 int contended; /* Signal need_sched() or lock 182 * contention detected during 183 * compaction 184 */ 185 }; 186 187 unsigned long 188 isolate_freepages_range(struct compact_control *cc, 189 unsigned long start_pfn, unsigned long end_pfn); 190 unsigned long 191 isolate_migratepages_range(struct compact_control *cc, 192 unsigned long low_pfn, unsigned long end_pfn); 193 int find_suitable_fallback(struct free_area *area, unsigned int order, 194 int migratetype, bool only_stealable, bool *can_steal); 195 196 #endif 197 198 /* 199 * This function returns the order of a free page in the buddy system. In 200 * general, page_zone(page)->lock must be held by the caller to prevent the 201 * page from being allocated in parallel and returning garbage as the order. 202 * If a caller does not hold page_zone(page)->lock, it must guarantee that the 203 * page cannot be allocated or merged in parallel. Alternatively, it must 204 * handle invalid values gracefully, and use page_order_unsafe() below. 205 */ 206 static inline unsigned int page_order(struct page *page) 207 { 208 /* PageBuddy() must be checked by the caller */ 209 return page_private(page); 210 } 211 212 /* 213 * Like page_order(), but for callers who cannot afford to hold the zone lock. 214 * PageBuddy() should be checked first by the caller to minimize race window, 215 * and invalid values must be handled gracefully. 216 * 217 * READ_ONCE is used so that if the caller assigns the result into a local 218 * variable and e.g. tests it for valid range before using, the compiler cannot 219 * decide to remove the variable and inline the page_private(page) multiple 220 * times, potentially observing different values in the tests and the actual 221 * use of the result. 222 */ 223 #define page_order_unsafe(page) READ_ONCE(page_private(page)) 224 225 static inline bool is_cow_mapping(vm_flags_t flags) 226 { 227 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE; 228 } 229 230 /* 231 * These three helpers classifies VMAs for virtual memory accounting. 232 */ 233 234 /* 235 * Executable code area - executable, not writable, not stack 236 */ 237 static inline bool is_exec_mapping(vm_flags_t flags) 238 { 239 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; 240 } 241 242 /* 243 * Stack area - atomatically grows in one direction 244 * 245 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: 246 * do_mmap() forbids all other combinations. 247 */ 248 static inline bool is_stack_mapping(vm_flags_t flags) 249 { 250 return (flags & VM_STACK) == VM_STACK; 251 } 252 253 /* 254 * Data area - private, writable, not stack 255 */ 256 static inline bool is_data_mapping(vm_flags_t flags) 257 { 258 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; 259 } 260 261 /* mm/util.c */ 262 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, 263 struct vm_area_struct *prev, struct rb_node *rb_parent); 264 265 #ifdef CONFIG_MMU 266 extern long populate_vma_page_range(struct vm_area_struct *vma, 267 unsigned long start, unsigned long end, int *nonblocking); 268 extern void munlock_vma_pages_range(struct vm_area_struct *vma, 269 unsigned long start, unsigned long end); 270 static inline void munlock_vma_pages_all(struct vm_area_struct *vma) 271 { 272 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end); 273 } 274 275 /* 276 * must be called with vma's mmap_sem held for read or write, and page locked. 277 */ 278 extern void mlock_vma_page(struct page *page); 279 extern unsigned int munlock_vma_page(struct page *page); 280 281 /* 282 * Clear the page's PageMlocked(). This can be useful in a situation where 283 * we want to unconditionally remove a page from the pagecache -- e.g., 284 * on truncation or freeing. 285 * 286 * It is legal to call this function for any page, mlocked or not. 287 * If called for a page that is still mapped by mlocked vmas, all we do 288 * is revert to lazy LRU behaviour -- semantics are not broken. 289 */ 290 extern void clear_page_mlock(struct page *page); 291 292 /* 293 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page() 294 * (because that does not go through the full procedure of migration ptes): 295 * to migrate the Mlocked page flag; update statistics. 296 */ 297 static inline void mlock_migrate_page(struct page *newpage, struct page *page) 298 { 299 if (TestClearPageMlocked(page)) { 300 int nr_pages = hpage_nr_pages(page); 301 302 /* Holding pmd lock, no change in irq context: __mod is safe */ 303 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages); 304 SetPageMlocked(newpage); 305 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages); 306 } 307 } 308 309 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); 310 311 /* 312 * At what user virtual address is page expected in @vma? 313 */ 314 static inline unsigned long 315 __vma_address(struct page *page, struct vm_area_struct *vma) 316 { 317 pgoff_t pgoff = page_to_pgoff(page); 318 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); 319 } 320 321 static inline unsigned long 322 vma_address(struct page *page, struct vm_area_struct *vma) 323 { 324 unsigned long address = __vma_address(page, vma); 325 326 /* page should be within @vma mapping range */ 327 VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma); 328 329 return address; 330 } 331 332 #else /* !CONFIG_MMU */ 333 static inline void clear_page_mlock(struct page *page) { } 334 static inline void mlock_vma_page(struct page *page) { } 335 static inline void mlock_migrate_page(struct page *new, struct page *old) { } 336 337 #endif /* !CONFIG_MMU */ 338 339 /* 340 * Return the mem_map entry representing the 'offset' subpage within 341 * the maximally aligned gigantic page 'base'. Handle any discontiguity 342 * in the mem_map at MAX_ORDER_NR_PAGES boundaries. 343 */ 344 static inline struct page *mem_map_offset(struct page *base, int offset) 345 { 346 if (unlikely(offset >= MAX_ORDER_NR_PAGES)) 347 return nth_page(base, offset); 348 return base + offset; 349 } 350 351 /* 352 * Iterator over all subpages within the maximally aligned gigantic 353 * page 'base'. Handle any discontiguity in the mem_map. 354 */ 355 static inline struct page *mem_map_next(struct page *iter, 356 struct page *base, int offset) 357 { 358 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { 359 unsigned long pfn = page_to_pfn(base) + offset; 360 if (!pfn_valid(pfn)) 361 return NULL; 362 return pfn_to_page(pfn); 363 } 364 return iter + 1; 365 } 366 367 /* 368 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node, 369 * so all functions starting at paging_init should be marked __init 370 * in those cases. SPARSEMEM, however, allows for memory hotplug, 371 * and alloc_bootmem_node is not used. 372 */ 373 #ifdef CONFIG_SPARSEMEM 374 #define __paginginit __meminit 375 #else 376 #define __paginginit __init 377 #endif 378 379 /* Memory initialisation debug and verification */ 380 enum mminit_level { 381 MMINIT_WARNING, 382 MMINIT_VERIFY, 383 MMINIT_TRACE 384 }; 385 386 #ifdef CONFIG_DEBUG_MEMORY_INIT 387 388 extern int mminit_loglevel; 389 390 #define mminit_dprintk(level, prefix, fmt, arg...) \ 391 do { \ 392 if (level < mminit_loglevel) { \ 393 if (level <= MMINIT_WARNING) \ 394 pr_warn("mminit::" prefix " " fmt, ##arg); \ 395 else \ 396 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ 397 } \ 398 } while (0) 399 400 extern void mminit_verify_pageflags_layout(void); 401 extern void mminit_verify_zonelist(void); 402 #else 403 404 static inline void mminit_dprintk(enum mminit_level level, 405 const char *prefix, const char *fmt, ...) 406 { 407 } 408 409 static inline void mminit_verify_pageflags_layout(void) 410 { 411 } 412 413 static inline void mminit_verify_zonelist(void) 414 { 415 } 416 #endif /* CONFIG_DEBUG_MEMORY_INIT */ 417 418 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */ 419 #if defined(CONFIG_SPARSEMEM) 420 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn, 421 unsigned long *end_pfn); 422 #else 423 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn, 424 unsigned long *end_pfn) 425 { 426 } 427 #endif /* CONFIG_SPARSEMEM */ 428 429 #define ZONE_RECLAIM_NOSCAN -2 430 #define ZONE_RECLAIM_FULL -1 431 #define ZONE_RECLAIM_SOME 0 432 #define ZONE_RECLAIM_SUCCESS 1 433 434 extern int hwpoison_filter(struct page *p); 435 436 extern u32 hwpoison_filter_dev_major; 437 extern u32 hwpoison_filter_dev_minor; 438 extern u64 hwpoison_filter_flags_mask; 439 extern u64 hwpoison_filter_flags_value; 440 extern u64 hwpoison_filter_memcg; 441 extern u32 hwpoison_filter_enable; 442 443 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long, 444 unsigned long, unsigned long, 445 unsigned long, unsigned long); 446 447 extern void set_pageblock_order(void); 448 unsigned long reclaim_clean_pages_from_list(struct zone *zone, 449 struct list_head *page_list); 450 /* The ALLOC_WMARK bits are used as an index to zone->watermark */ 451 #define ALLOC_WMARK_MIN WMARK_MIN 452 #define ALLOC_WMARK_LOW WMARK_LOW 453 #define ALLOC_WMARK_HIGH WMARK_HIGH 454 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ 455 456 /* Mask to get the watermark bits */ 457 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) 458 459 #define ALLOC_HARDER 0x10 /* try to alloc harder */ 460 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ 461 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ 462 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ 463 #define ALLOC_FAIR 0x100 /* fair zone allocation */ 464 465 enum ttu_flags; 466 struct tlbflush_unmap_batch; 467 468 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH 469 void try_to_unmap_flush(void); 470 void try_to_unmap_flush_dirty(void); 471 #else 472 static inline void try_to_unmap_flush(void) 473 { 474 } 475 static inline void try_to_unmap_flush_dirty(void) 476 { 477 } 478 479 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ 480 481 extern const struct trace_print_flags pageflag_names[]; 482 extern const struct trace_print_flags vmaflag_names[]; 483 extern const struct trace_print_flags gfpflag_names[]; 484 485 #endif /* __MM_INTERNAL_H */ 486