| 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 | |
| 17 | /* |
| 18 | * The set of flags that only affect watermark checking and reclaim |
| 19 | * behaviour. This is used by the MM to obey the caller constraints |
| 20 | * about IO, FS and watermark checking while ignoring placement |
| 21 | * hints such as HIGHMEM usage. |
| 22 | */ |
| 23 | #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\ |
| 24 | __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\ |
| 25 | __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC) |
| 26 | |
| 27 | /* The GFP flags allowed during early boot */ |
| 28 | #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS)) |
| 29 | |
| 30 | /* Control allocation cpuset and node placement constraints */ |
| 31 | #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) |
| 32 | |
| 33 | /* Do not use these with a slab allocator */ |
| 34 | #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) |
| 35 | |
| 36 | void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
| 37 | unsigned long floor, unsigned long ceiling); |
| 38 | |
| 39 | static inline void set_page_count(struct page *page, int v) |
| 40 | { |
| 41 | atomic_set(&page->_count, v); |
| 42 | } |
| 43 | |
| 44 | extern int __do_page_cache_readahead(struct address_space *mapping, |
| 45 | struct file *filp, pgoff_t offset, unsigned long nr_to_read, |
| 46 | unsigned long lookahead_size); |
| 47 | |
| 48 | /* |
| 49 | * Submit IO for the read-ahead request in file_ra_state. |
| 50 | */ |
| 51 | static inline unsigned long ra_submit(struct file_ra_state *ra, |
| 52 | struct address_space *mapping, struct file *filp) |
| 53 | { |
| 54 | return __do_page_cache_readahead(mapping, filp, |
| 55 | ra->start, ra->size, ra->async_size); |
| 56 | } |
| 57 | |
| 58 | /* |
| 59 | * Turn a non-refcounted page (->_count == 0) into refcounted with |
| 60 | * a count of one. |
| 61 | */ |
| 62 | static inline void set_page_refcounted(struct page *page) |
| 63 | { |
| 64 | VM_BUG_ON_PAGE(PageTail(page), page); |
| 65 | VM_BUG_ON_PAGE(atomic_read(&page->_count), page); |
| 66 | set_page_count(page, 1); |
| 67 | } |
| 68 | |
| 69 | static inline void __get_page_tail_foll(struct page *page, |
| 70 | bool get_page_head) |
| 71 | { |
| 72 | /* |
| 73 | * If we're getting a tail page, the elevated page->_count is |
| 74 | * required only in the head page and we will elevate the head |
| 75 | * page->_count and tail page->_mapcount. |
| 76 | * |
| 77 | * We elevate page_tail->_mapcount for tail pages to force |
| 78 | * page_tail->_count to be zero at all times to avoid getting |
| 79 | * false positives from get_page_unless_zero() with |
| 80 | * speculative page access (like in |
| 81 | * page_cache_get_speculative()) on tail pages. |
| 82 | */ |
| 83 | VM_BUG_ON_PAGE(atomic_read(&compound_head(page)->_count) <= 0, page); |
| 84 | if (get_page_head) |
| 85 | atomic_inc(&compound_head(page)->_count); |
| 86 | get_huge_page_tail(page); |
| 87 | } |
| 88 | |
| 89 | /* |
| 90 | * This is meant to be called as the FOLL_GET operation of |
| 91 | * follow_page() and it must be called while holding the proper PT |
| 92 | * lock while the pte (or pmd_trans_huge) is still mapping the page. |
| 93 | */ |
| 94 | static inline void get_page_foll(struct page *page) |
| 95 | { |
| 96 | if (unlikely(PageTail(page))) |
| 97 | /* |
| 98 | * This is safe only because |
| 99 | * __split_huge_page_refcount() can't run under |
| 100 | * get_page_foll() because we hold the proper PT lock. |
| 101 | */ |
| 102 | __get_page_tail_foll(page, true); |
| 103 | else { |
| 104 | /* |
| 105 | * Getting a normal page or the head of a compound page |
| 106 | * requires to already have an elevated page->_count. |
| 107 | */ |
| 108 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); |
| 109 | atomic_inc(&page->_count); |
| 110 | } |
| 111 | } |
| 112 | |
| 113 | extern unsigned long highest_memmap_pfn; |
| 114 | |
| 115 | /* |
| 116 | * in mm/vmscan.c: |
| 117 | */ |
| 118 | extern int isolate_lru_page(struct page *page); |
| 119 | extern void putback_lru_page(struct page *page); |
| 120 | extern bool zone_reclaimable(struct zone *zone); |
| 121 | |
| 122 | /* |
| 123 | * in mm/rmap.c: |
| 124 | */ |
| 125 | extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); |
| 126 | |
| 127 | /* |
| 128 | * in mm/page_alloc.c |
| 129 | */ |
| 130 | |
| 131 | /* |
| 132 | * Structure for holding the mostly immutable allocation parameters passed |
| 133 | * between functions involved in allocations, including the alloc_pages* |
| 134 | * family of functions. |
| 135 | * |
| 136 | * nodemask, migratetype and high_zoneidx are initialized only once in |
| 137 | * __alloc_pages_nodemask() and then never change. |
| 138 | * |
| 139 | * zonelist, preferred_zone and classzone_idx are set first in |
| 140 | * __alloc_pages_nodemask() for the fast path, and might be later changed |
| 141 | * in __alloc_pages_slowpath(). All other functions pass the whole strucure |
| 142 | * by a const pointer. |
| 143 | */ |
| 144 | struct alloc_context { |
| 145 | struct zonelist *zonelist; |
| 146 | nodemask_t *nodemask; |
| 147 | struct zone *preferred_zone; |
| 148 | int classzone_idx; |
| 149 | int migratetype; |
| 150 | enum zone_type high_zoneidx; |
| 151 | bool spread_dirty_pages; |
| 152 | }; |
| 153 | |
| 154 | /* |
| 155 | * Locate the struct page for both the matching buddy in our |
| 156 | * pair (buddy1) and the combined O(n+1) page they form (page). |
| 157 | * |
| 158 | * 1) Any buddy B1 will have an order O twin B2 which satisfies |
| 159 | * the following equation: |
| 160 | * B2 = B1 ^ (1 << O) |
| 161 | * For example, if the starting buddy (buddy2) is #8 its order |
| 162 | * 1 buddy is #10: |
| 163 | * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 |
| 164 | * |
| 165 | * 2) Any buddy B will have an order O+1 parent P which |
| 166 | * satisfies the following equation: |
| 167 | * P = B & ~(1 << O) |
| 168 | * |
| 169 | * Assumption: *_mem_map is contiguous at least up to MAX_ORDER |
| 170 | */ |
| 171 | static inline unsigned long |
| 172 | __find_buddy_index(unsigned long page_idx, unsigned int order) |
| 173 | { |
| 174 | return page_idx ^ (1 << order); |
| 175 | } |
| 176 | |
| 177 | extern int __isolate_free_page(struct page *page, unsigned int order); |
| 178 | extern void __free_pages_bootmem(struct page *page, unsigned long pfn, |
| 179 | unsigned int order); |
| 180 | extern void prep_compound_page(struct page *page, unsigned int order); |
| 181 | #ifdef CONFIG_MEMORY_FAILURE |
| 182 | extern bool is_free_buddy_page(struct page *page); |
| 183 | #endif |
| 184 | extern int user_min_free_kbytes; |
| 185 | |
| 186 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
| 187 | |
| 188 | /* |
| 189 | * in mm/compaction.c |
| 190 | */ |
| 191 | /* |
| 192 | * compact_control is used to track pages being migrated and the free pages |
| 193 | * they are being migrated to during memory compaction. The free_pfn starts |
| 194 | * at the end of a zone and migrate_pfn begins at the start. Movable pages |
| 195 | * are moved to the end of a zone during a compaction run and the run |
| 196 | * completes when free_pfn <= migrate_pfn |
| 197 | */ |
| 198 | struct compact_control { |
| 199 | struct list_head freepages; /* List of free pages to migrate to */ |
| 200 | struct list_head migratepages; /* List of pages being migrated */ |
| 201 | unsigned long nr_freepages; /* Number of isolated free pages */ |
| 202 | unsigned long nr_migratepages; /* Number of pages to migrate */ |
| 203 | unsigned long free_pfn; /* isolate_freepages search base */ |
| 204 | unsigned long migrate_pfn; /* isolate_migratepages search base */ |
| 205 | unsigned long last_migrated_pfn;/* Not yet flushed page being freed */ |
| 206 | enum migrate_mode mode; /* Async or sync migration mode */ |
| 207 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
| 208 | int order; /* order a direct compactor needs */ |
| 209 | const gfp_t gfp_mask; /* gfp mask of a direct compactor */ |
| 210 | const int alloc_flags; /* alloc flags of a direct compactor */ |
| 211 | const int classzone_idx; /* zone index of a direct compactor */ |
| 212 | struct zone *zone; |
| 213 | int contended; /* Signal need_sched() or lock |
| 214 | * contention detected during |
| 215 | * compaction |
| 216 | */ |
| 217 | }; |
| 218 | |
| 219 | unsigned long |
| 220 | isolate_freepages_range(struct compact_control *cc, |
| 221 | unsigned long start_pfn, unsigned long end_pfn); |
| 222 | unsigned long |
| 223 | isolate_migratepages_range(struct compact_control *cc, |
| 224 | unsigned long low_pfn, unsigned long end_pfn); |
| 225 | int find_suitable_fallback(struct free_area *area, unsigned int order, |
| 226 | int migratetype, bool only_stealable, bool *can_steal); |
| 227 | |
| 228 | #endif |
| 229 | |
| 230 | /* |
| 231 | * This function returns the order of a free page in the buddy system. In |
| 232 | * general, page_zone(page)->lock must be held by the caller to prevent the |
| 233 | * page from being allocated in parallel and returning garbage as the order. |
| 234 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the |
| 235 | * page cannot be allocated or merged in parallel. Alternatively, it must |
| 236 | * handle invalid values gracefully, and use page_order_unsafe() below. |
| 237 | */ |
| 238 | static inline unsigned int page_order(struct page *page) |
| 239 | { |
| 240 | /* PageBuddy() must be checked by the caller */ |
| 241 | return page_private(page); |
| 242 | } |
| 243 | |
| 244 | /* |
| 245 | * Like page_order(), but for callers who cannot afford to hold the zone lock. |
| 246 | * PageBuddy() should be checked first by the caller to minimize race window, |
| 247 | * and invalid values must be handled gracefully. |
| 248 | * |
| 249 | * READ_ONCE is used so that if the caller assigns the result into a local |
| 250 | * variable and e.g. tests it for valid range before using, the compiler cannot |
| 251 | * decide to remove the variable and inline the page_private(page) multiple |
| 252 | * times, potentially observing different values in the tests and the actual |
| 253 | * use of the result. |
| 254 | */ |
| 255 | #define page_order_unsafe(page) READ_ONCE(page_private(page)) |
| 256 | |
| 257 | static inline bool is_cow_mapping(vm_flags_t flags) |
| 258 | { |
| 259 | return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE; |
| 260 | } |
| 261 | |
| 262 | /* mm/util.c */ |
| 263 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, |
| 264 | struct vm_area_struct *prev, struct rb_node *rb_parent); |
| 265 | |
| 266 | #ifdef CONFIG_MMU |
| 267 | extern long populate_vma_page_range(struct vm_area_struct *vma, |
| 268 | unsigned long start, unsigned long end, int *nonblocking); |
| 269 | extern void munlock_vma_pages_range(struct vm_area_struct *vma, |
| 270 | unsigned long start, unsigned long end); |
| 271 | static inline void munlock_vma_pages_all(struct vm_area_struct *vma) |
| 272 | { |
| 273 | munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end); |
| 274 | } |
| 275 | |
| 276 | /* |
| 277 | * must be called with vma's mmap_sem held for read or write, and page locked. |
| 278 | */ |
| 279 | extern void mlock_vma_page(struct page *page); |
| 280 | extern unsigned int munlock_vma_page(struct page *page); |
| 281 | |
| 282 | /* |
| 283 | * Clear the page's PageMlocked(). This can be useful in a situation where |
| 284 | * we want to unconditionally remove a page from the pagecache -- e.g., |
| 285 | * on truncation or freeing. |
| 286 | * |
| 287 | * It is legal to call this function for any page, mlocked or not. |
| 288 | * If called for a page that is still mapped by mlocked vmas, all we do |
| 289 | * is revert to lazy LRU behaviour -- semantics are not broken. |
| 290 | */ |
| 291 | extern void clear_page_mlock(struct page *page); |
| 292 | |
| 293 | /* |
| 294 | * mlock_migrate_page - called only from migrate_misplaced_transhuge_page() |
| 295 | * (because that does not go through the full procedure of migration ptes): |
| 296 | * to migrate the Mlocked page flag; update statistics. |
| 297 | */ |
| 298 | static inline void mlock_migrate_page(struct page *newpage, struct page *page) |
| 299 | { |
| 300 | if (TestClearPageMlocked(page)) { |
| 301 | int nr_pages = hpage_nr_pages(page); |
| 302 | |
| 303 | /* Holding pmd lock, no change in irq context: __mod is safe */ |
| 304 | __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages); |
| 305 | SetPageMlocked(newpage); |
| 306 | __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages); |
| 307 | } |
| 308 | } |
| 309 | |
| 310 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
| 311 | |
| 312 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 313 | extern unsigned long vma_address(struct page *page, |
| 314 | struct vm_area_struct *vma); |
| 315 | #endif |
| 316 | #else /* !CONFIG_MMU */ |
| 317 | static inline void clear_page_mlock(struct page *page) { } |
| 318 | static inline void mlock_vma_page(struct page *page) { } |
| 319 | static inline void mlock_migrate_page(struct page *new, struct page *old) { } |
| 320 | |
| 321 | #endif /* !CONFIG_MMU */ |
| 322 | |
| 323 | /* |
| 324 | * Return the mem_map entry representing the 'offset' subpage within |
| 325 | * the maximally aligned gigantic page 'base'. Handle any discontiguity |
| 326 | * in the mem_map at MAX_ORDER_NR_PAGES boundaries. |
| 327 | */ |
| 328 | static inline struct page *mem_map_offset(struct page *base, int offset) |
| 329 | { |
| 330 | if (unlikely(offset >= MAX_ORDER_NR_PAGES)) |
| 331 | return nth_page(base, offset); |
| 332 | return base + offset; |
| 333 | } |
| 334 | |
| 335 | /* |
| 336 | * Iterator over all subpages within the maximally aligned gigantic |
| 337 | * page 'base'. Handle any discontiguity in the mem_map. |
| 338 | */ |
| 339 | static inline struct page *mem_map_next(struct page *iter, |
| 340 | struct page *base, int offset) |
| 341 | { |
| 342 | if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { |
| 343 | unsigned long pfn = page_to_pfn(base) + offset; |
| 344 | if (!pfn_valid(pfn)) |
| 345 | return NULL; |
| 346 | return pfn_to_page(pfn); |
| 347 | } |
| 348 | return iter + 1; |
| 349 | } |
| 350 | |
| 351 | /* |
| 352 | * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node, |
| 353 | * so all functions starting at paging_init should be marked __init |
| 354 | * in those cases. SPARSEMEM, however, allows for memory hotplug, |
| 355 | * and alloc_bootmem_node is not used. |
| 356 | */ |
| 357 | #ifdef CONFIG_SPARSEMEM |
| 358 | #define __paginginit __meminit |
| 359 | #else |
| 360 | #define __paginginit __init |
| 361 | #endif |
| 362 | |
| 363 | /* Memory initialisation debug and verification */ |
| 364 | enum mminit_level { |
| 365 | MMINIT_WARNING, |
| 366 | MMINIT_VERIFY, |
| 367 | MMINIT_TRACE |
| 368 | }; |
| 369 | |
| 370 | #ifdef CONFIG_DEBUG_MEMORY_INIT |
| 371 | |
| 372 | extern int mminit_loglevel; |
| 373 | |
| 374 | #define mminit_dprintk(level, prefix, fmt, arg...) \ |
| 375 | do { \ |
| 376 | if (level < mminit_loglevel) { \ |
| 377 | if (level <= MMINIT_WARNING) \ |
| 378 | printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \ |
| 379 | else \ |
| 380 | printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ |
| 381 | } \ |
| 382 | } while (0) |
| 383 | |
| 384 | extern void mminit_verify_pageflags_layout(void); |
| 385 | extern void mminit_verify_zonelist(void); |
| 386 | #else |
| 387 | |
| 388 | static inline void mminit_dprintk(enum mminit_level level, |
| 389 | const char *prefix, const char *fmt, ...) |
| 390 | { |
| 391 | } |
| 392 | |
| 393 | static inline void mminit_verify_pageflags_layout(void) |
| 394 | { |
| 395 | } |
| 396 | |
| 397 | static inline void mminit_verify_zonelist(void) |
| 398 | { |
| 399 | } |
| 400 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
| 401 | |
| 402 | /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */ |
| 403 | #if defined(CONFIG_SPARSEMEM) |
| 404 | extern void mminit_validate_memmodel_limits(unsigned long *start_pfn, |
| 405 | unsigned long *end_pfn); |
| 406 | #else |
| 407 | static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn, |
| 408 | unsigned long *end_pfn) |
| 409 | { |
| 410 | } |
| 411 | #endif /* CONFIG_SPARSEMEM */ |
| 412 | |
| 413 | #define ZONE_RECLAIM_NOSCAN -2 |
| 414 | #define ZONE_RECLAIM_FULL -1 |
| 415 | #define ZONE_RECLAIM_SOME 0 |
| 416 | #define ZONE_RECLAIM_SUCCESS 1 |
| 417 | |
| 418 | extern int hwpoison_filter(struct page *p); |
| 419 | |
| 420 | extern u32 hwpoison_filter_dev_major; |
| 421 | extern u32 hwpoison_filter_dev_minor; |
| 422 | extern u64 hwpoison_filter_flags_mask; |
| 423 | extern u64 hwpoison_filter_flags_value; |
| 424 | extern u64 hwpoison_filter_memcg; |
| 425 | extern u32 hwpoison_filter_enable; |
| 426 | |
| 427 | extern unsigned long vm_mmap_pgoff(struct file *, unsigned long, |
| 428 | unsigned long, unsigned long, |
| 429 | unsigned long, unsigned long); |
| 430 | |
| 431 | extern void set_pageblock_order(void); |
| 432 | unsigned long reclaim_clean_pages_from_list(struct zone *zone, |
| 433 | struct list_head *page_list); |
| 434 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
| 435 | #define ALLOC_WMARK_MIN WMARK_MIN |
| 436 | #define ALLOC_WMARK_LOW WMARK_LOW |
| 437 | #define ALLOC_WMARK_HIGH WMARK_HIGH |
| 438 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ |
| 439 | |
| 440 | /* Mask to get the watermark bits */ |
| 441 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) |
| 442 | |
| 443 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ |
| 444 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ |
| 445 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ |
| 446 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ |
| 447 | #define ALLOC_FAIR 0x100 /* fair zone allocation */ |
| 448 | |
| 449 | enum ttu_flags; |
| 450 | struct tlbflush_unmap_batch; |
| 451 | |
| 452 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
| 453 | void try_to_unmap_flush(void); |
| 454 | void try_to_unmap_flush_dirty(void); |
| 455 | #else |
| 456 | static inline void try_to_unmap_flush(void) |
| 457 | { |
| 458 | } |
| 459 | static inline void try_to_unmap_flush_dirty(void) |
| 460 | { |
| 461 | } |
| 462 | |
| 463 | #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ |
| 464 | #endif /* __MM_INTERNAL_H */ |