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