| 1 | #ifndef _LINUX_MM_H |
| 2 | #define _LINUX_MM_H |
| 3 | |
| 4 | #include <linux/errno.h> |
| 5 | |
| 6 | #ifdef __KERNEL__ |
| 7 | |
| 8 | #include <linux/mmdebug.h> |
| 9 | #include <linux/gfp.h> |
| 10 | #include <linux/bug.h> |
| 11 | #include <linux/list.h> |
| 12 | #include <linux/mmzone.h> |
| 13 | #include <linux/rbtree.h> |
| 14 | #include <linux/atomic.h> |
| 15 | #include <linux/debug_locks.h> |
| 16 | #include <linux/mm_types.h> |
| 17 | #include <linux/range.h> |
| 18 | #include <linux/pfn.h> |
| 19 | #include <linux/percpu-refcount.h> |
| 20 | #include <linux/bit_spinlock.h> |
| 21 | #include <linux/shrinker.h> |
| 22 | #include <linux/resource.h> |
| 23 | #include <linux/page_ext.h> |
| 24 | #include <linux/err.h> |
| 25 | |
| 26 | struct mempolicy; |
| 27 | struct anon_vma; |
| 28 | struct anon_vma_chain; |
| 29 | struct file_ra_state; |
| 30 | struct user_struct; |
| 31 | struct writeback_control; |
| 32 | struct bdi_writeback; |
| 33 | |
| 34 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ |
| 35 | extern unsigned long max_mapnr; |
| 36 | |
| 37 | static inline void set_max_mapnr(unsigned long limit) |
| 38 | { |
| 39 | max_mapnr = limit; |
| 40 | } |
| 41 | #else |
| 42 | static inline void set_max_mapnr(unsigned long limit) { } |
| 43 | #endif |
| 44 | |
| 45 | extern unsigned long totalram_pages; |
| 46 | extern void * high_memory; |
| 47 | extern int page_cluster; |
| 48 | |
| 49 | #ifdef CONFIG_SYSCTL |
| 50 | extern int sysctl_legacy_va_layout; |
| 51 | #else |
| 52 | #define sysctl_legacy_va_layout 0 |
| 53 | #endif |
| 54 | |
| 55 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS |
| 56 | extern const int mmap_rnd_bits_min; |
| 57 | extern const int mmap_rnd_bits_max; |
| 58 | extern int mmap_rnd_bits __read_mostly; |
| 59 | #endif |
| 60 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS |
| 61 | extern const int mmap_rnd_compat_bits_min; |
| 62 | extern const int mmap_rnd_compat_bits_max; |
| 63 | extern int mmap_rnd_compat_bits __read_mostly; |
| 64 | #endif |
| 65 | |
| 66 | #include <asm/page.h> |
| 67 | #include <asm/pgtable.h> |
| 68 | #include <asm/processor.h> |
| 69 | |
| 70 | #ifndef __pa_symbol |
| 71 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) |
| 72 | #endif |
| 73 | |
| 74 | /* |
| 75 | * To prevent common memory management code establishing |
| 76 | * a zero page mapping on a read fault. |
| 77 | * This macro should be defined within <asm/pgtable.h>. |
| 78 | * s390 does this to prevent multiplexing of hardware bits |
| 79 | * related to the physical page in case of virtualization. |
| 80 | */ |
| 81 | #ifndef mm_forbids_zeropage |
| 82 | #define mm_forbids_zeropage(X) (0) |
| 83 | #endif |
| 84 | |
| 85 | extern unsigned long sysctl_user_reserve_kbytes; |
| 86 | extern unsigned long sysctl_admin_reserve_kbytes; |
| 87 | |
| 88 | extern int sysctl_overcommit_memory; |
| 89 | extern int sysctl_overcommit_ratio; |
| 90 | extern unsigned long sysctl_overcommit_kbytes; |
| 91 | |
| 92 | extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, |
| 93 | size_t *, loff_t *); |
| 94 | extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, |
| 95 | size_t *, loff_t *); |
| 96 | |
| 97 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
| 98 | |
| 99 | /* to align the pointer to the (next) page boundary */ |
| 100 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) |
| 101 | |
| 102 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ |
| 103 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE) |
| 104 | |
| 105 | /* |
| 106 | * Linux kernel virtual memory manager primitives. |
| 107 | * The idea being to have a "virtual" mm in the same way |
| 108 | * we have a virtual fs - giving a cleaner interface to the |
| 109 | * mm details, and allowing different kinds of memory mappings |
| 110 | * (from shared memory to executable loading to arbitrary |
| 111 | * mmap() functions). |
| 112 | */ |
| 113 | |
| 114 | extern struct kmem_cache *vm_area_cachep; |
| 115 | |
| 116 | #ifndef CONFIG_MMU |
| 117 | extern struct rb_root nommu_region_tree; |
| 118 | extern struct rw_semaphore nommu_region_sem; |
| 119 | |
| 120 | extern unsigned int kobjsize(const void *objp); |
| 121 | #endif |
| 122 | |
| 123 | /* |
| 124 | * vm_flags in vm_area_struct, see mm_types.h. |
| 125 | */ |
| 126 | #define VM_NONE 0x00000000 |
| 127 | |
| 128 | #define VM_READ 0x00000001 /* currently active flags */ |
| 129 | #define VM_WRITE 0x00000002 |
| 130 | #define VM_EXEC 0x00000004 |
| 131 | #define VM_SHARED 0x00000008 |
| 132 | |
| 133 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
| 134 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
| 135 | #define VM_MAYWRITE 0x00000020 |
| 136 | #define VM_MAYEXEC 0x00000040 |
| 137 | #define VM_MAYSHARE 0x00000080 |
| 138 | |
| 139 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ |
| 140 | #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */ |
| 141 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
| 142 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
| 143 | #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */ |
| 144 | |
| 145 | #define VM_LOCKED 0x00002000 |
| 146 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ |
| 147 | |
| 148 | /* Used by sys_madvise() */ |
| 149 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ |
| 150 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ |
| 151 | |
| 152 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ |
| 153 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ |
| 154 | #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */ |
| 155 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
| 156 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
| 157 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
| 158 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ |
| 159 | #define VM_ARCH_2 0x02000000 |
| 160 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ |
| 161 | |
| 162 | #ifdef CONFIG_MEM_SOFT_DIRTY |
| 163 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ |
| 164 | #else |
| 165 | # define VM_SOFTDIRTY 0 |
| 166 | #endif |
| 167 | |
| 168 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
| 169 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ |
| 170 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ |
| 171 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
| 172 | |
| 173 | #if defined(CONFIG_X86) |
| 174 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ |
| 175 | #elif defined(CONFIG_PPC) |
| 176 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ |
| 177 | #elif defined(CONFIG_PARISC) |
| 178 | # define VM_GROWSUP VM_ARCH_1 |
| 179 | #elif defined(CONFIG_METAG) |
| 180 | # define VM_GROWSUP VM_ARCH_1 |
| 181 | #elif defined(CONFIG_IA64) |
| 182 | # define VM_GROWSUP VM_ARCH_1 |
| 183 | #elif !defined(CONFIG_MMU) |
| 184 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ |
| 185 | #endif |
| 186 | |
| 187 | #if defined(CONFIG_X86) |
| 188 | /* MPX specific bounds table or bounds directory */ |
| 189 | # define VM_MPX VM_ARCH_2 |
| 190 | #endif |
| 191 | |
| 192 | #ifndef VM_GROWSUP |
| 193 | # define VM_GROWSUP VM_NONE |
| 194 | #endif |
| 195 | |
| 196 | /* Bits set in the VMA until the stack is in its final location */ |
| 197 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) |
| 198 | |
| 199 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
| 200 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS |
| 201 | #endif |
| 202 | |
| 203 | #ifdef CONFIG_STACK_GROWSUP |
| 204 | #define VM_STACK VM_GROWSUP |
| 205 | #else |
| 206 | #define VM_STACK VM_GROWSDOWN |
| 207 | #endif |
| 208 | |
| 209 | #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 210 | |
| 211 | /* |
| 212 | * Special vmas that are non-mergable, non-mlock()able. |
| 213 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. |
| 214 | */ |
| 215 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) |
| 216 | |
| 217 | /* This mask defines which mm->def_flags a process can inherit its parent */ |
| 218 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE |
| 219 | |
| 220 | /* This mask is used to clear all the VMA flags used by mlock */ |
| 221 | #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT)) |
| 222 | |
| 223 | /* |
| 224 | * mapping from the currently active vm_flags protection bits (the |
| 225 | * low four bits) to a page protection mask.. |
| 226 | */ |
| 227 | extern pgprot_t protection_map[16]; |
| 228 | |
| 229 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ |
| 230 | #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */ |
| 231 | #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */ |
| 232 | #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */ |
| 233 | #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */ |
| 234 | #define FAULT_FLAG_TRIED 0x20 /* Second try */ |
| 235 | #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */ |
| 236 | |
| 237 | /* |
| 238 | * vm_fault is filled by the the pagefault handler and passed to the vma's |
| 239 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
| 240 | * of VM_FAULT_xxx flags that give details about how the fault was handled. |
| 241 | * |
| 242 | * MM layer fills up gfp_mask for page allocations but fault handler might |
| 243 | * alter it if its implementation requires a different allocation context. |
| 244 | * |
| 245 | * pgoff should be used in favour of virtual_address, if possible. |
| 246 | */ |
| 247 | struct vm_fault { |
| 248 | unsigned int flags; /* FAULT_FLAG_xxx flags */ |
| 249 | gfp_t gfp_mask; /* gfp mask to be used for allocations */ |
| 250 | pgoff_t pgoff; /* Logical page offset based on vma */ |
| 251 | void __user *virtual_address; /* Faulting virtual address */ |
| 252 | |
| 253 | struct page *cow_page; /* Handler may choose to COW */ |
| 254 | struct page *page; /* ->fault handlers should return a |
| 255 | * page here, unless VM_FAULT_NOPAGE |
| 256 | * is set (which is also implied by |
| 257 | * VM_FAULT_ERROR). |
| 258 | */ |
| 259 | /* for ->map_pages() only */ |
| 260 | pgoff_t max_pgoff; /* map pages for offset from pgoff till |
| 261 | * max_pgoff inclusive */ |
| 262 | pte_t *pte; /* pte entry associated with ->pgoff */ |
| 263 | }; |
| 264 | |
| 265 | /* |
| 266 | * These are the virtual MM functions - opening of an area, closing and |
| 267 | * unmapping it (needed to keep files on disk up-to-date etc), pointer |
| 268 | * to the functions called when a no-page or a wp-page exception occurs. |
| 269 | */ |
| 270 | struct vm_operations_struct { |
| 271 | void (*open)(struct vm_area_struct * area); |
| 272 | void (*close)(struct vm_area_struct * area); |
| 273 | int (*mremap)(struct vm_area_struct * area); |
| 274 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); |
| 275 | int (*pmd_fault)(struct vm_area_struct *, unsigned long address, |
| 276 | pmd_t *, unsigned int flags); |
| 277 | void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf); |
| 278 | |
| 279 | /* notification that a previously read-only page is about to become |
| 280 | * writable, if an error is returned it will cause a SIGBUS */ |
| 281 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); |
| 282 | |
| 283 | /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ |
| 284 | int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); |
| 285 | |
| 286 | /* called by access_process_vm when get_user_pages() fails, typically |
| 287 | * for use by special VMAs that can switch between memory and hardware |
| 288 | */ |
| 289 | int (*access)(struct vm_area_struct *vma, unsigned long addr, |
| 290 | void *buf, int len, int write); |
| 291 | |
| 292 | /* Called by the /proc/PID/maps code to ask the vma whether it |
| 293 | * has a special name. Returning non-NULL will also cause this |
| 294 | * vma to be dumped unconditionally. */ |
| 295 | const char *(*name)(struct vm_area_struct *vma); |
| 296 | |
| 297 | #ifdef CONFIG_NUMA |
| 298 | /* |
| 299 | * set_policy() op must add a reference to any non-NULL @new mempolicy |
| 300 | * to hold the policy upon return. Caller should pass NULL @new to |
| 301 | * remove a policy and fall back to surrounding context--i.e. do not |
| 302 | * install a MPOL_DEFAULT policy, nor the task or system default |
| 303 | * mempolicy. |
| 304 | */ |
| 305 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
| 306 | |
| 307 | /* |
| 308 | * get_policy() op must add reference [mpol_get()] to any policy at |
| 309 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure |
| 310 | * in mm/mempolicy.c will do this automatically. |
| 311 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not |
| 312 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. |
| 313 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op |
| 314 | * must return NULL--i.e., do not "fallback" to task or system default |
| 315 | * policy. |
| 316 | */ |
| 317 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
| 318 | unsigned long addr); |
| 319 | #endif |
| 320 | /* |
| 321 | * Called by vm_normal_page() for special PTEs to find the |
| 322 | * page for @addr. This is useful if the default behavior |
| 323 | * (using pte_page()) would not find the correct page. |
| 324 | */ |
| 325 | struct page *(*find_special_page)(struct vm_area_struct *vma, |
| 326 | unsigned long addr); |
| 327 | }; |
| 328 | |
| 329 | struct mmu_gather; |
| 330 | struct inode; |
| 331 | |
| 332 | #define page_private(page) ((page)->private) |
| 333 | #define set_page_private(page, v) ((page)->private = (v)) |
| 334 | |
| 335 | #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE) |
| 336 | static inline int pmd_devmap(pmd_t pmd) |
| 337 | { |
| 338 | return 0; |
| 339 | } |
| 340 | #endif |
| 341 | |
| 342 | /* |
| 343 | * FIXME: take this include out, include page-flags.h in |
| 344 | * files which need it (119 of them) |
| 345 | */ |
| 346 | #include <linux/page-flags.h> |
| 347 | #include <linux/huge_mm.h> |
| 348 | |
| 349 | /* |
| 350 | * Methods to modify the page usage count. |
| 351 | * |
| 352 | * What counts for a page usage: |
| 353 | * - cache mapping (page->mapping) |
| 354 | * - private data (page->private) |
| 355 | * - page mapped in a task's page tables, each mapping |
| 356 | * is counted separately |
| 357 | * |
| 358 | * Also, many kernel routines increase the page count before a critical |
| 359 | * routine so they can be sure the page doesn't go away from under them. |
| 360 | */ |
| 361 | |
| 362 | /* |
| 363 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
| 364 | */ |
| 365 | static inline int put_page_testzero(struct page *page) |
| 366 | { |
| 367 | VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page); |
| 368 | return atomic_dec_and_test(&page->_count); |
| 369 | } |
| 370 | |
| 371 | /* |
| 372 | * Try to grab a ref unless the page has a refcount of zero, return false if |
| 373 | * that is the case. |
| 374 | * This can be called when MMU is off so it must not access |
| 375 | * any of the virtual mappings. |
| 376 | */ |
| 377 | static inline int get_page_unless_zero(struct page *page) |
| 378 | { |
| 379 | return atomic_inc_not_zero(&page->_count); |
| 380 | } |
| 381 | |
| 382 | extern int page_is_ram(unsigned long pfn); |
| 383 | |
| 384 | enum { |
| 385 | REGION_INTERSECTS, |
| 386 | REGION_DISJOINT, |
| 387 | REGION_MIXED, |
| 388 | }; |
| 389 | |
| 390 | int region_intersects(resource_size_t offset, size_t size, const char *type); |
| 391 | |
| 392 | /* Support for virtually mapped pages */ |
| 393 | struct page *vmalloc_to_page(const void *addr); |
| 394 | unsigned long vmalloc_to_pfn(const void *addr); |
| 395 | |
| 396 | /* |
| 397 | * Determine if an address is within the vmalloc range |
| 398 | * |
| 399 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there |
| 400 | * is no special casing required. |
| 401 | */ |
| 402 | static inline int is_vmalloc_addr(const void *x) |
| 403 | { |
| 404 | #ifdef CONFIG_MMU |
| 405 | unsigned long addr = (unsigned long)x; |
| 406 | |
| 407 | return addr >= VMALLOC_START && addr < VMALLOC_END; |
| 408 | #else |
| 409 | return 0; |
| 410 | #endif |
| 411 | } |
| 412 | #ifdef CONFIG_MMU |
| 413 | extern int is_vmalloc_or_module_addr(const void *x); |
| 414 | #else |
| 415 | static inline int is_vmalloc_or_module_addr(const void *x) |
| 416 | { |
| 417 | return 0; |
| 418 | } |
| 419 | #endif |
| 420 | |
| 421 | extern void kvfree(const void *addr); |
| 422 | |
| 423 | static inline atomic_t *compound_mapcount_ptr(struct page *page) |
| 424 | { |
| 425 | return &page[1].compound_mapcount; |
| 426 | } |
| 427 | |
| 428 | static inline int compound_mapcount(struct page *page) |
| 429 | { |
| 430 | if (!PageCompound(page)) |
| 431 | return 0; |
| 432 | page = compound_head(page); |
| 433 | return atomic_read(compound_mapcount_ptr(page)) + 1; |
| 434 | } |
| 435 | |
| 436 | /* |
| 437 | * The atomic page->_mapcount, starts from -1: so that transitions |
| 438 | * both from it and to it can be tracked, using atomic_inc_and_test |
| 439 | * and atomic_add_negative(-1). |
| 440 | */ |
| 441 | static inline void page_mapcount_reset(struct page *page) |
| 442 | { |
| 443 | atomic_set(&(page)->_mapcount, -1); |
| 444 | } |
| 445 | |
| 446 | int __page_mapcount(struct page *page); |
| 447 | |
| 448 | static inline int page_mapcount(struct page *page) |
| 449 | { |
| 450 | VM_BUG_ON_PAGE(PageSlab(page), page); |
| 451 | |
| 452 | if (unlikely(PageCompound(page))) |
| 453 | return __page_mapcount(page); |
| 454 | return atomic_read(&page->_mapcount) + 1; |
| 455 | } |
| 456 | |
| 457 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 458 | int total_mapcount(struct page *page); |
| 459 | #else |
| 460 | static inline int total_mapcount(struct page *page) |
| 461 | { |
| 462 | return page_mapcount(page); |
| 463 | } |
| 464 | #endif |
| 465 | |
| 466 | static inline int page_count(struct page *page) |
| 467 | { |
| 468 | return atomic_read(&compound_head(page)->_count); |
| 469 | } |
| 470 | |
| 471 | static inline struct page *virt_to_head_page(const void *x) |
| 472 | { |
| 473 | struct page *page = virt_to_page(x); |
| 474 | |
| 475 | return compound_head(page); |
| 476 | } |
| 477 | |
| 478 | /* |
| 479 | * Setup the page count before being freed into the page allocator for |
| 480 | * the first time (boot or memory hotplug) |
| 481 | */ |
| 482 | static inline void init_page_count(struct page *page) |
| 483 | { |
| 484 | atomic_set(&page->_count, 1); |
| 485 | } |
| 486 | |
| 487 | void __put_page(struct page *page); |
| 488 | |
| 489 | void put_pages_list(struct list_head *pages); |
| 490 | |
| 491 | void split_page(struct page *page, unsigned int order); |
| 492 | int split_free_page(struct page *page); |
| 493 | |
| 494 | /* |
| 495 | * Compound pages have a destructor function. Provide a |
| 496 | * prototype for that function and accessor functions. |
| 497 | * These are _only_ valid on the head of a compound page. |
| 498 | */ |
| 499 | typedef void compound_page_dtor(struct page *); |
| 500 | |
| 501 | /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */ |
| 502 | enum compound_dtor_id { |
| 503 | NULL_COMPOUND_DTOR, |
| 504 | COMPOUND_PAGE_DTOR, |
| 505 | #ifdef CONFIG_HUGETLB_PAGE |
| 506 | HUGETLB_PAGE_DTOR, |
| 507 | #endif |
| 508 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 509 | TRANSHUGE_PAGE_DTOR, |
| 510 | #endif |
| 511 | NR_COMPOUND_DTORS, |
| 512 | }; |
| 513 | extern compound_page_dtor * const compound_page_dtors[]; |
| 514 | |
| 515 | static inline void set_compound_page_dtor(struct page *page, |
| 516 | enum compound_dtor_id compound_dtor) |
| 517 | { |
| 518 | VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page); |
| 519 | page[1].compound_dtor = compound_dtor; |
| 520 | } |
| 521 | |
| 522 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) |
| 523 | { |
| 524 | VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page); |
| 525 | return compound_page_dtors[page[1].compound_dtor]; |
| 526 | } |
| 527 | |
| 528 | static inline unsigned int compound_order(struct page *page) |
| 529 | { |
| 530 | if (!PageHead(page)) |
| 531 | return 0; |
| 532 | return page[1].compound_order; |
| 533 | } |
| 534 | |
| 535 | static inline void set_compound_order(struct page *page, unsigned int order) |
| 536 | { |
| 537 | page[1].compound_order = order; |
| 538 | } |
| 539 | |
| 540 | void free_compound_page(struct page *page); |
| 541 | |
| 542 | #ifdef CONFIG_MMU |
| 543 | /* |
| 544 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when |
| 545 | * servicing faults for write access. In the normal case, do always want |
| 546 | * pte_mkwrite. But get_user_pages can cause write faults for mappings |
| 547 | * that do not have writing enabled, when used by access_process_vm. |
| 548 | */ |
| 549 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) |
| 550 | { |
| 551 | if (likely(vma->vm_flags & VM_WRITE)) |
| 552 | pte = pte_mkwrite(pte); |
| 553 | return pte; |
| 554 | } |
| 555 | |
| 556 | void do_set_pte(struct vm_area_struct *vma, unsigned long address, |
| 557 | struct page *page, pte_t *pte, bool write, bool anon); |
| 558 | #endif |
| 559 | |
| 560 | /* |
| 561 | * Multiple processes may "see" the same page. E.g. for untouched |
| 562 | * mappings of /dev/null, all processes see the same page full of |
| 563 | * zeroes, and text pages of executables and shared libraries have |
| 564 | * only one copy in memory, at most, normally. |
| 565 | * |
| 566 | * For the non-reserved pages, page_count(page) denotes a reference count. |
| 567 | * page_count() == 0 means the page is free. page->lru is then used for |
| 568 | * freelist management in the buddy allocator. |
| 569 | * page_count() > 0 means the page has been allocated. |
| 570 | * |
| 571 | * Pages are allocated by the slab allocator in order to provide memory |
| 572 | * to kmalloc and kmem_cache_alloc. In this case, the management of the |
| 573 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c |
| 574 | * unless a particular usage is carefully commented. (the responsibility of |
| 575 | * freeing the kmalloc memory is the caller's, of course). |
| 576 | * |
| 577 | * A page may be used by anyone else who does a __get_free_page(). |
| 578 | * In this case, page_count still tracks the references, and should only |
| 579 | * be used through the normal accessor functions. The top bits of page->flags |
| 580 | * and page->virtual store page management information, but all other fields |
| 581 | * are unused and could be used privately, carefully. The management of this |
| 582 | * page is the responsibility of the one who allocated it, and those who have |
| 583 | * subsequently been given references to it. |
| 584 | * |
| 585 | * The other pages (we may call them "pagecache pages") are completely |
| 586 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
| 587 | * The following discussion applies only to them. |
| 588 | * |
| 589 | * A pagecache page contains an opaque `private' member, which belongs to the |
| 590 | * page's address_space. Usually, this is the address of a circular list of |
| 591 | * the page's disk buffers. PG_private must be set to tell the VM to call |
| 592 | * into the filesystem to release these pages. |
| 593 | * |
| 594 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
| 595 | * is the pointer to the inode, and page->index is the file offset of the page, |
| 596 | * in units of PAGE_CACHE_SIZE. |
| 597 | * |
| 598 | * If pagecache pages are not associated with an inode, they are said to be |
| 599 | * anonymous pages. These may become associated with the swapcache, and in that |
| 600 | * case PG_swapcache is set, and page->private is an offset into the swapcache. |
| 601 | * |
| 602 | * In either case (swapcache or inode backed), the pagecache itself holds one |
| 603 | * reference to the page. Setting PG_private should also increment the |
| 604 | * refcount. The each user mapping also has a reference to the page. |
| 605 | * |
| 606 | * The pagecache pages are stored in a per-mapping radix tree, which is |
| 607 | * rooted at mapping->page_tree, and indexed by offset. |
| 608 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
| 609 | * lists, we instead now tag pages as dirty/writeback in the radix tree. |
| 610 | * |
| 611 | * All pagecache pages may be subject to I/O: |
| 612 | * - inode pages may need to be read from disk, |
| 613 | * - inode pages which have been modified and are MAP_SHARED may need |
| 614 | * to be written back to the inode on disk, |
| 615 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been |
| 616 | * modified may need to be swapped out to swap space and (later) to be read |
| 617 | * back into memory. |
| 618 | */ |
| 619 | |
| 620 | /* |
| 621 | * The zone field is never updated after free_area_init_core() |
| 622 | * sets it, so none of the operations on it need to be atomic. |
| 623 | */ |
| 624 | |
| 625 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ |
| 626 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
| 627 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
| 628 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) |
| 629 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) |
| 630 | |
| 631 | /* |
| 632 | * Define the bit shifts to access each section. For non-existent |
| 633 | * sections we define the shift as 0; that plus a 0 mask ensures |
| 634 | * the compiler will optimise away reference to them. |
| 635 | */ |
| 636 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
| 637 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) |
| 638 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) |
| 639 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) |
| 640 | |
| 641 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ |
| 642 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
| 643 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
| 644 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
| 645 | SECTIONS_PGOFF : ZONES_PGOFF) |
| 646 | #else |
| 647 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
| 648 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
| 649 | NODES_PGOFF : ZONES_PGOFF) |
| 650 | #endif |
| 651 | |
| 652 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
| 653 | |
| 654 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
| 655 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
| 656 | #endif |
| 657 | |
| 658 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
| 659 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) |
| 660 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) |
| 661 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1) |
| 662 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
| 663 | |
| 664 | static inline enum zone_type page_zonenum(const struct page *page) |
| 665 | { |
| 666 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
| 667 | } |
| 668 | |
| 669 | #ifdef CONFIG_ZONE_DEVICE |
| 670 | void get_zone_device_page(struct page *page); |
| 671 | void put_zone_device_page(struct page *page); |
| 672 | static inline bool is_zone_device_page(const struct page *page) |
| 673 | { |
| 674 | return page_zonenum(page) == ZONE_DEVICE; |
| 675 | } |
| 676 | #else |
| 677 | static inline void get_zone_device_page(struct page *page) |
| 678 | { |
| 679 | } |
| 680 | static inline void put_zone_device_page(struct page *page) |
| 681 | { |
| 682 | } |
| 683 | static inline bool is_zone_device_page(const struct page *page) |
| 684 | { |
| 685 | return false; |
| 686 | } |
| 687 | #endif |
| 688 | |
| 689 | static inline void get_page(struct page *page) |
| 690 | { |
| 691 | page = compound_head(page); |
| 692 | /* |
| 693 | * Getting a normal page or the head of a compound page |
| 694 | * requires to already have an elevated page->_count. |
| 695 | */ |
| 696 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); |
| 697 | atomic_inc(&page->_count); |
| 698 | |
| 699 | if (unlikely(is_zone_device_page(page))) |
| 700 | get_zone_device_page(page); |
| 701 | } |
| 702 | |
| 703 | static inline void put_page(struct page *page) |
| 704 | { |
| 705 | page = compound_head(page); |
| 706 | |
| 707 | if (put_page_testzero(page)) |
| 708 | __put_page(page); |
| 709 | |
| 710 | if (unlikely(is_zone_device_page(page))) |
| 711 | put_zone_device_page(page); |
| 712 | } |
| 713 | |
| 714 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
| 715 | #define SECTION_IN_PAGE_FLAGS |
| 716 | #endif |
| 717 | |
| 718 | /* |
| 719 | * The identification function is mainly used by the buddy allocator for |
| 720 | * determining if two pages could be buddies. We are not really identifying |
| 721 | * the zone since we could be using the section number id if we do not have |
| 722 | * node id available in page flags. |
| 723 | * We only guarantee that it will return the same value for two combinable |
| 724 | * pages in a zone. |
| 725 | */ |
| 726 | static inline int page_zone_id(struct page *page) |
| 727 | { |
| 728 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
| 729 | } |
| 730 | |
| 731 | static inline int zone_to_nid(struct zone *zone) |
| 732 | { |
| 733 | #ifdef CONFIG_NUMA |
| 734 | return zone->node; |
| 735 | #else |
| 736 | return 0; |
| 737 | #endif |
| 738 | } |
| 739 | |
| 740 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
| 741 | extern int page_to_nid(const struct page *page); |
| 742 | #else |
| 743 | static inline int page_to_nid(const struct page *page) |
| 744 | { |
| 745 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; |
| 746 | } |
| 747 | #endif |
| 748 | |
| 749 | #ifdef CONFIG_NUMA_BALANCING |
| 750 | static inline int cpu_pid_to_cpupid(int cpu, int pid) |
| 751 | { |
| 752 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); |
| 753 | } |
| 754 | |
| 755 | static inline int cpupid_to_pid(int cpupid) |
| 756 | { |
| 757 | return cpupid & LAST__PID_MASK; |
| 758 | } |
| 759 | |
| 760 | static inline int cpupid_to_cpu(int cpupid) |
| 761 | { |
| 762 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; |
| 763 | } |
| 764 | |
| 765 | static inline int cpupid_to_nid(int cpupid) |
| 766 | { |
| 767 | return cpu_to_node(cpupid_to_cpu(cpupid)); |
| 768 | } |
| 769 | |
| 770 | static inline bool cpupid_pid_unset(int cpupid) |
| 771 | { |
| 772 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); |
| 773 | } |
| 774 | |
| 775 | static inline bool cpupid_cpu_unset(int cpupid) |
| 776 | { |
| 777 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); |
| 778 | } |
| 779 | |
| 780 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) |
| 781 | { |
| 782 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); |
| 783 | } |
| 784 | |
| 785 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) |
| 786 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
| 787 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) |
| 788 | { |
| 789 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); |
| 790 | } |
| 791 | |
| 792 | static inline int page_cpupid_last(struct page *page) |
| 793 | { |
| 794 | return page->_last_cpupid; |
| 795 | } |
| 796 | static inline void page_cpupid_reset_last(struct page *page) |
| 797 | { |
| 798 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; |
| 799 | } |
| 800 | #else |
| 801 | static inline int page_cpupid_last(struct page *page) |
| 802 | { |
| 803 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; |
| 804 | } |
| 805 | |
| 806 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); |
| 807 | |
| 808 | static inline void page_cpupid_reset_last(struct page *page) |
| 809 | { |
| 810 | int cpupid = (1 << LAST_CPUPID_SHIFT) - 1; |
| 811 | |
| 812 | page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT); |
| 813 | page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT; |
| 814 | } |
| 815 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ |
| 816 | #else /* !CONFIG_NUMA_BALANCING */ |
| 817 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) |
| 818 | { |
| 819 | return page_to_nid(page); /* XXX */ |
| 820 | } |
| 821 | |
| 822 | static inline int page_cpupid_last(struct page *page) |
| 823 | { |
| 824 | return page_to_nid(page); /* XXX */ |
| 825 | } |
| 826 | |
| 827 | static inline int cpupid_to_nid(int cpupid) |
| 828 | { |
| 829 | return -1; |
| 830 | } |
| 831 | |
| 832 | static inline int cpupid_to_pid(int cpupid) |
| 833 | { |
| 834 | return -1; |
| 835 | } |
| 836 | |
| 837 | static inline int cpupid_to_cpu(int cpupid) |
| 838 | { |
| 839 | return -1; |
| 840 | } |
| 841 | |
| 842 | static inline int cpu_pid_to_cpupid(int nid, int pid) |
| 843 | { |
| 844 | return -1; |
| 845 | } |
| 846 | |
| 847 | static inline bool cpupid_pid_unset(int cpupid) |
| 848 | { |
| 849 | return 1; |
| 850 | } |
| 851 | |
| 852 | static inline void page_cpupid_reset_last(struct page *page) |
| 853 | { |
| 854 | } |
| 855 | |
| 856 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) |
| 857 | { |
| 858 | return false; |
| 859 | } |
| 860 | #endif /* CONFIG_NUMA_BALANCING */ |
| 861 | |
| 862 | static inline struct zone *page_zone(const struct page *page) |
| 863 | { |
| 864 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; |
| 865 | } |
| 866 | |
| 867 | #ifdef SECTION_IN_PAGE_FLAGS |
| 868 | static inline void set_page_section(struct page *page, unsigned long section) |
| 869 | { |
| 870 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); |
| 871 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; |
| 872 | } |
| 873 | |
| 874 | static inline unsigned long page_to_section(const struct page *page) |
| 875 | { |
| 876 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; |
| 877 | } |
| 878 | #endif |
| 879 | |
| 880 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
| 881 | { |
| 882 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); |
| 883 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; |
| 884 | } |
| 885 | |
| 886 | static inline void set_page_node(struct page *page, unsigned long node) |
| 887 | { |
| 888 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); |
| 889 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; |
| 890 | } |
| 891 | |
| 892 | static inline void set_page_links(struct page *page, enum zone_type zone, |
| 893 | unsigned long node, unsigned long pfn) |
| 894 | { |
| 895 | set_page_zone(page, zone); |
| 896 | set_page_node(page, node); |
| 897 | #ifdef SECTION_IN_PAGE_FLAGS |
| 898 | set_page_section(page, pfn_to_section_nr(pfn)); |
| 899 | #endif |
| 900 | } |
| 901 | |
| 902 | #ifdef CONFIG_MEMCG |
| 903 | static inline struct mem_cgroup *page_memcg(struct page *page) |
| 904 | { |
| 905 | return page->mem_cgroup; |
| 906 | } |
| 907 | |
| 908 | static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg) |
| 909 | { |
| 910 | page->mem_cgroup = memcg; |
| 911 | } |
| 912 | #else |
| 913 | static inline struct mem_cgroup *page_memcg(struct page *page) |
| 914 | { |
| 915 | return NULL; |
| 916 | } |
| 917 | |
| 918 | static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg) |
| 919 | { |
| 920 | } |
| 921 | #endif |
| 922 | |
| 923 | /* |
| 924 | * Some inline functions in vmstat.h depend on page_zone() |
| 925 | */ |
| 926 | #include <linux/vmstat.h> |
| 927 | |
| 928 | static __always_inline void *lowmem_page_address(const struct page *page) |
| 929 | { |
| 930 | return __va(PFN_PHYS(page_to_pfn(page))); |
| 931 | } |
| 932 | |
| 933 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) |
| 934 | #define HASHED_PAGE_VIRTUAL |
| 935 | #endif |
| 936 | |
| 937 | #if defined(WANT_PAGE_VIRTUAL) |
| 938 | static inline void *page_address(const struct page *page) |
| 939 | { |
| 940 | return page->virtual; |
| 941 | } |
| 942 | static inline void set_page_address(struct page *page, void *address) |
| 943 | { |
| 944 | page->virtual = address; |
| 945 | } |
| 946 | #define page_address_init() do { } while(0) |
| 947 | #endif |
| 948 | |
| 949 | #if defined(HASHED_PAGE_VIRTUAL) |
| 950 | void *page_address(const struct page *page); |
| 951 | void set_page_address(struct page *page, void *virtual); |
| 952 | void page_address_init(void); |
| 953 | #endif |
| 954 | |
| 955 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) |
| 956 | #define page_address(page) lowmem_page_address(page) |
| 957 | #define set_page_address(page, address) do { } while(0) |
| 958 | #define page_address_init() do { } while(0) |
| 959 | #endif |
| 960 | |
| 961 | extern void *page_rmapping(struct page *page); |
| 962 | extern struct anon_vma *page_anon_vma(struct page *page); |
| 963 | extern struct address_space *page_mapping(struct page *page); |
| 964 | |
| 965 | extern struct address_space *__page_file_mapping(struct page *); |
| 966 | |
| 967 | static inline |
| 968 | struct address_space *page_file_mapping(struct page *page) |
| 969 | { |
| 970 | if (unlikely(PageSwapCache(page))) |
| 971 | return __page_file_mapping(page); |
| 972 | |
| 973 | return page->mapping; |
| 974 | } |
| 975 | |
| 976 | /* |
| 977 | * Return the pagecache index of the passed page. Regular pagecache pages |
| 978 | * use ->index whereas swapcache pages use ->private |
| 979 | */ |
| 980 | static inline pgoff_t page_index(struct page *page) |
| 981 | { |
| 982 | if (unlikely(PageSwapCache(page))) |
| 983 | return page_private(page); |
| 984 | return page->index; |
| 985 | } |
| 986 | |
| 987 | extern pgoff_t __page_file_index(struct page *page); |
| 988 | |
| 989 | /* |
| 990 | * Return the file index of the page. Regular pagecache pages use ->index |
| 991 | * whereas swapcache pages use swp_offset(->private) |
| 992 | */ |
| 993 | static inline pgoff_t page_file_index(struct page *page) |
| 994 | { |
| 995 | if (unlikely(PageSwapCache(page))) |
| 996 | return __page_file_index(page); |
| 997 | |
| 998 | return page->index; |
| 999 | } |
| 1000 | |
| 1001 | /* |
| 1002 | * Return true if this page is mapped into pagetables. |
| 1003 | * For compound page it returns true if any subpage of compound page is mapped. |
| 1004 | */ |
| 1005 | static inline bool page_mapped(struct page *page) |
| 1006 | { |
| 1007 | int i; |
| 1008 | if (likely(!PageCompound(page))) |
| 1009 | return atomic_read(&page->_mapcount) >= 0; |
| 1010 | page = compound_head(page); |
| 1011 | if (atomic_read(compound_mapcount_ptr(page)) >= 0) |
| 1012 | return true; |
| 1013 | for (i = 0; i < hpage_nr_pages(page); i++) { |
| 1014 | if (atomic_read(&page[i]._mapcount) >= 0) |
| 1015 | return true; |
| 1016 | } |
| 1017 | return false; |
| 1018 | } |
| 1019 | |
| 1020 | /* |
| 1021 | * Return true only if the page has been allocated with |
| 1022 | * ALLOC_NO_WATERMARKS and the low watermark was not |
| 1023 | * met implying that the system is under some pressure. |
| 1024 | */ |
| 1025 | static inline bool page_is_pfmemalloc(struct page *page) |
| 1026 | { |
| 1027 | /* |
| 1028 | * Page index cannot be this large so this must be |
| 1029 | * a pfmemalloc page. |
| 1030 | */ |
| 1031 | return page->index == -1UL; |
| 1032 | } |
| 1033 | |
| 1034 | /* |
| 1035 | * Only to be called by the page allocator on a freshly allocated |
| 1036 | * page. |
| 1037 | */ |
| 1038 | static inline void set_page_pfmemalloc(struct page *page) |
| 1039 | { |
| 1040 | page->index = -1UL; |
| 1041 | } |
| 1042 | |
| 1043 | static inline void clear_page_pfmemalloc(struct page *page) |
| 1044 | { |
| 1045 | page->index = 0; |
| 1046 | } |
| 1047 | |
| 1048 | /* |
| 1049 | * Different kinds of faults, as returned by handle_mm_fault(). |
| 1050 | * Used to decide whether a process gets delivered SIGBUS or |
| 1051 | * just gets major/minor fault counters bumped up. |
| 1052 | */ |
| 1053 | |
| 1054 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ |
| 1055 | |
| 1056 | #define VM_FAULT_OOM 0x0001 |
| 1057 | #define VM_FAULT_SIGBUS 0x0002 |
| 1058 | #define VM_FAULT_MAJOR 0x0004 |
| 1059 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ |
| 1060 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ |
| 1061 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ |
| 1062 | #define VM_FAULT_SIGSEGV 0x0040 |
| 1063 | |
| 1064 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ |
| 1065 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ |
| 1066 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ |
| 1067 | #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */ |
| 1068 | |
| 1069 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ |
| 1070 | |
| 1071 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \ |
| 1072 | VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \ |
| 1073 | VM_FAULT_FALLBACK) |
| 1074 | |
| 1075 | /* Encode hstate index for a hwpoisoned large page */ |
| 1076 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) |
| 1077 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) |
| 1078 | |
| 1079 | /* |
| 1080 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. |
| 1081 | */ |
| 1082 | extern void pagefault_out_of_memory(void); |
| 1083 | |
| 1084 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
| 1085 | |
| 1086 | /* |
| 1087 | * Flags passed to show_mem() and show_free_areas() to suppress output in |
| 1088 | * various contexts. |
| 1089 | */ |
| 1090 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ |
| 1091 | |
| 1092 | extern void show_free_areas(unsigned int flags); |
| 1093 | extern bool skip_free_areas_node(unsigned int flags, int nid); |
| 1094 | |
| 1095 | int shmem_zero_setup(struct vm_area_struct *); |
| 1096 | #ifdef CONFIG_SHMEM |
| 1097 | bool shmem_mapping(struct address_space *mapping); |
| 1098 | #else |
| 1099 | static inline bool shmem_mapping(struct address_space *mapping) |
| 1100 | { |
| 1101 | return false; |
| 1102 | } |
| 1103 | #endif |
| 1104 | |
| 1105 | extern bool can_do_mlock(void); |
| 1106 | extern int user_shm_lock(size_t, struct user_struct *); |
| 1107 | extern void user_shm_unlock(size_t, struct user_struct *); |
| 1108 | |
| 1109 | /* |
| 1110 | * Parameter block passed down to zap_pte_range in exceptional cases. |
| 1111 | */ |
| 1112 | struct zap_details { |
| 1113 | struct address_space *check_mapping; /* Check page->mapping if set */ |
| 1114 | pgoff_t first_index; /* Lowest page->index to unmap */ |
| 1115 | pgoff_t last_index; /* Highest page->index to unmap */ |
| 1116 | }; |
| 1117 | |
| 1118 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
| 1119 | pte_t pte); |
| 1120 | |
| 1121 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
| 1122 | unsigned long size); |
| 1123 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, |
| 1124 | unsigned long size, struct zap_details *); |
| 1125 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
| 1126 | unsigned long start, unsigned long end); |
| 1127 | |
| 1128 | /** |
| 1129 | * mm_walk - callbacks for walk_page_range |
| 1130 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry |
| 1131 | * this handler is required to be able to handle |
| 1132 | * pmd_trans_huge() pmds. They may simply choose to |
| 1133 | * split_huge_page() instead of handling it explicitly. |
| 1134 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry |
| 1135 | * @pte_hole: if set, called for each hole at all levels |
| 1136 | * @hugetlb_entry: if set, called for each hugetlb entry |
| 1137 | * @test_walk: caller specific callback function to determine whether |
| 1138 | * we walk over the current vma or not. A positive returned |
| 1139 | * value means "do page table walk over the current vma," |
| 1140 | * and a negative one means "abort current page table walk |
| 1141 | * right now." 0 means "skip the current vma." |
| 1142 | * @mm: mm_struct representing the target process of page table walk |
| 1143 | * @vma: vma currently walked (NULL if walking outside vmas) |
| 1144 | * @private: private data for callbacks' usage |
| 1145 | * |
| 1146 | * (see the comment on walk_page_range() for more details) |
| 1147 | */ |
| 1148 | struct mm_walk { |
| 1149 | int (*pmd_entry)(pmd_t *pmd, unsigned long addr, |
| 1150 | unsigned long next, struct mm_walk *walk); |
| 1151 | int (*pte_entry)(pte_t *pte, unsigned long addr, |
| 1152 | unsigned long next, struct mm_walk *walk); |
| 1153 | int (*pte_hole)(unsigned long addr, unsigned long next, |
| 1154 | struct mm_walk *walk); |
| 1155 | int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, |
| 1156 | unsigned long addr, unsigned long next, |
| 1157 | struct mm_walk *walk); |
| 1158 | int (*test_walk)(unsigned long addr, unsigned long next, |
| 1159 | struct mm_walk *walk); |
| 1160 | struct mm_struct *mm; |
| 1161 | struct vm_area_struct *vma; |
| 1162 | void *private; |
| 1163 | }; |
| 1164 | |
| 1165 | int walk_page_range(unsigned long addr, unsigned long end, |
| 1166 | struct mm_walk *walk); |
| 1167 | int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk); |
| 1168 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
| 1169 | unsigned long end, unsigned long floor, unsigned long ceiling); |
| 1170 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
| 1171 | struct vm_area_struct *vma); |
| 1172 | void unmap_mapping_range(struct address_space *mapping, |
| 1173 | loff_t const holebegin, loff_t const holelen, int even_cows); |
| 1174 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
| 1175 | unsigned long *pfn); |
| 1176 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
| 1177 | unsigned int flags, unsigned long *prot, resource_size_t *phys); |
| 1178 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
| 1179 | void *buf, int len, int write); |
| 1180 | |
| 1181 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
| 1182 | loff_t const holebegin, loff_t const holelen) |
| 1183 | { |
| 1184 | unmap_mapping_range(mapping, holebegin, holelen, 0); |
| 1185 | } |
| 1186 | |
| 1187 | extern void truncate_pagecache(struct inode *inode, loff_t new); |
| 1188 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
| 1189 | void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); |
| 1190 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); |
| 1191 | int truncate_inode_page(struct address_space *mapping, struct page *page); |
| 1192 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
| 1193 | int invalidate_inode_page(struct page *page); |
| 1194 | |
| 1195 | #ifdef CONFIG_MMU |
| 1196 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, |
| 1197 | unsigned long address, unsigned int flags); |
| 1198 | extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, |
| 1199 | unsigned long address, unsigned int fault_flags, |
| 1200 | bool *unlocked); |
| 1201 | #else |
| 1202 | static inline int handle_mm_fault(struct mm_struct *mm, |
| 1203 | struct vm_area_struct *vma, unsigned long address, |
| 1204 | unsigned int flags) |
| 1205 | { |
| 1206 | /* should never happen if there's no MMU */ |
| 1207 | BUG(); |
| 1208 | return VM_FAULT_SIGBUS; |
| 1209 | } |
| 1210 | static inline int fixup_user_fault(struct task_struct *tsk, |
| 1211 | struct mm_struct *mm, unsigned long address, |
| 1212 | unsigned int fault_flags, bool *unlocked) |
| 1213 | { |
| 1214 | /* should never happen if there's no MMU */ |
| 1215 | BUG(); |
| 1216 | return -EFAULT; |
| 1217 | } |
| 1218 | #endif |
| 1219 | |
| 1220 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); |
| 1221 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
| 1222 | void *buf, int len, int write); |
| 1223 | |
| 1224 | long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
| 1225 | unsigned long start, unsigned long nr_pages, |
| 1226 | unsigned int foll_flags, struct page **pages, |
| 1227 | struct vm_area_struct **vmas, int *nonblocking); |
| 1228 | long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm, |
| 1229 | unsigned long start, unsigned long nr_pages, |
| 1230 | int write, int force, struct page **pages, |
| 1231 | struct vm_area_struct **vmas); |
| 1232 | long get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
| 1233 | unsigned long start, unsigned long nr_pages, |
| 1234 | int write, int force, struct page **pages, |
| 1235 | struct vm_area_struct **vmas); |
| 1236 | long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm, |
| 1237 | unsigned long start, unsigned long nr_pages, |
| 1238 | int write, int force, struct page **pages, |
| 1239 | int *locked); |
| 1240 | long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm, |
| 1241 | unsigned long start, unsigned long nr_pages, |
| 1242 | int write, int force, struct page **pages, |
| 1243 | unsigned int gup_flags); |
| 1244 | long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm, |
| 1245 | unsigned long start, unsigned long nr_pages, |
| 1246 | int write, int force, struct page **pages); |
| 1247 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, |
| 1248 | struct page **pages); |
| 1249 | |
| 1250 | /* Container for pinned pfns / pages */ |
| 1251 | struct frame_vector { |
| 1252 | unsigned int nr_allocated; /* Number of frames we have space for */ |
| 1253 | unsigned int nr_frames; /* Number of frames stored in ptrs array */ |
| 1254 | bool got_ref; /* Did we pin pages by getting page ref? */ |
| 1255 | bool is_pfns; /* Does array contain pages or pfns? */ |
| 1256 | void *ptrs[0]; /* Array of pinned pfns / pages. Use |
| 1257 | * pfns_vector_pages() or pfns_vector_pfns() |
| 1258 | * for access */ |
| 1259 | }; |
| 1260 | |
| 1261 | struct frame_vector *frame_vector_create(unsigned int nr_frames); |
| 1262 | void frame_vector_destroy(struct frame_vector *vec); |
| 1263 | int get_vaddr_frames(unsigned long start, unsigned int nr_pfns, |
| 1264 | bool write, bool force, struct frame_vector *vec); |
| 1265 | void put_vaddr_frames(struct frame_vector *vec); |
| 1266 | int frame_vector_to_pages(struct frame_vector *vec); |
| 1267 | void frame_vector_to_pfns(struct frame_vector *vec); |
| 1268 | |
| 1269 | static inline unsigned int frame_vector_count(struct frame_vector *vec) |
| 1270 | { |
| 1271 | return vec->nr_frames; |
| 1272 | } |
| 1273 | |
| 1274 | static inline struct page **frame_vector_pages(struct frame_vector *vec) |
| 1275 | { |
| 1276 | if (vec->is_pfns) { |
| 1277 | int err = frame_vector_to_pages(vec); |
| 1278 | |
| 1279 | if (err) |
| 1280 | return ERR_PTR(err); |
| 1281 | } |
| 1282 | return (struct page **)(vec->ptrs); |
| 1283 | } |
| 1284 | |
| 1285 | static inline unsigned long *frame_vector_pfns(struct frame_vector *vec) |
| 1286 | { |
| 1287 | if (!vec->is_pfns) |
| 1288 | frame_vector_to_pfns(vec); |
| 1289 | return (unsigned long *)(vec->ptrs); |
| 1290 | } |
| 1291 | |
| 1292 | struct kvec; |
| 1293 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, |
| 1294 | struct page **pages); |
| 1295 | int get_kernel_page(unsigned long start, int write, struct page **pages); |
| 1296 | struct page *get_dump_page(unsigned long addr); |
| 1297 | |
| 1298 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
| 1299 | extern void do_invalidatepage(struct page *page, unsigned int offset, |
| 1300 | unsigned int length); |
| 1301 | |
| 1302 | int __set_page_dirty_nobuffers(struct page *page); |
| 1303 | int __set_page_dirty_no_writeback(struct page *page); |
| 1304 | int redirty_page_for_writepage(struct writeback_control *wbc, |
| 1305 | struct page *page); |
| 1306 | void account_page_dirtied(struct page *page, struct address_space *mapping, |
| 1307 | struct mem_cgroup *memcg); |
| 1308 | void account_page_cleaned(struct page *page, struct address_space *mapping, |
| 1309 | struct mem_cgroup *memcg, struct bdi_writeback *wb); |
| 1310 | int set_page_dirty(struct page *page); |
| 1311 | int set_page_dirty_lock(struct page *page); |
| 1312 | void cancel_dirty_page(struct page *page); |
| 1313 | int clear_page_dirty_for_io(struct page *page); |
| 1314 | |
| 1315 | int get_cmdline(struct task_struct *task, char *buffer, int buflen); |
| 1316 | |
| 1317 | /* Is the vma a continuation of the stack vma above it? */ |
| 1318 | static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) |
| 1319 | { |
| 1320 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); |
| 1321 | } |
| 1322 | |
| 1323 | static inline bool vma_is_anonymous(struct vm_area_struct *vma) |
| 1324 | { |
| 1325 | return !vma->vm_ops; |
| 1326 | } |
| 1327 | |
| 1328 | static inline int stack_guard_page_start(struct vm_area_struct *vma, |
| 1329 | unsigned long addr) |
| 1330 | { |
| 1331 | return (vma->vm_flags & VM_GROWSDOWN) && |
| 1332 | (vma->vm_start == addr) && |
| 1333 | !vma_growsdown(vma->vm_prev, addr); |
| 1334 | } |
| 1335 | |
| 1336 | /* Is the vma a continuation of the stack vma below it? */ |
| 1337 | static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) |
| 1338 | { |
| 1339 | return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); |
| 1340 | } |
| 1341 | |
| 1342 | static inline int stack_guard_page_end(struct vm_area_struct *vma, |
| 1343 | unsigned long addr) |
| 1344 | { |
| 1345 | return (vma->vm_flags & VM_GROWSUP) && |
| 1346 | (vma->vm_end == addr) && |
| 1347 | !vma_growsup(vma->vm_next, addr); |
| 1348 | } |
| 1349 | |
| 1350 | int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t); |
| 1351 | |
| 1352 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
| 1353 | unsigned long old_addr, struct vm_area_struct *new_vma, |
| 1354 | unsigned long new_addr, unsigned long len, |
| 1355 | bool need_rmap_locks); |
| 1356 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, |
| 1357 | unsigned long end, pgprot_t newprot, |
| 1358 | int dirty_accountable, int prot_numa); |
| 1359 | extern int mprotect_fixup(struct vm_area_struct *vma, |
| 1360 | struct vm_area_struct **pprev, unsigned long start, |
| 1361 | unsigned long end, unsigned long newflags); |
| 1362 | |
| 1363 | /* |
| 1364 | * doesn't attempt to fault and will return short. |
| 1365 | */ |
| 1366 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, |
| 1367 | struct page **pages); |
| 1368 | /* |
| 1369 | * per-process(per-mm_struct) statistics. |
| 1370 | */ |
| 1371 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) |
| 1372 | { |
| 1373 | long val = atomic_long_read(&mm->rss_stat.count[member]); |
| 1374 | |
| 1375 | #ifdef SPLIT_RSS_COUNTING |
| 1376 | /* |
| 1377 | * counter is updated in asynchronous manner and may go to minus. |
| 1378 | * But it's never be expected number for users. |
| 1379 | */ |
| 1380 | if (val < 0) |
| 1381 | val = 0; |
| 1382 | #endif |
| 1383 | return (unsigned long)val; |
| 1384 | } |
| 1385 | |
| 1386 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) |
| 1387 | { |
| 1388 | atomic_long_add(value, &mm->rss_stat.count[member]); |
| 1389 | } |
| 1390 | |
| 1391 | static inline void inc_mm_counter(struct mm_struct *mm, int member) |
| 1392 | { |
| 1393 | atomic_long_inc(&mm->rss_stat.count[member]); |
| 1394 | } |
| 1395 | |
| 1396 | static inline void dec_mm_counter(struct mm_struct *mm, int member) |
| 1397 | { |
| 1398 | atomic_long_dec(&mm->rss_stat.count[member]); |
| 1399 | } |
| 1400 | |
| 1401 | /* Optimized variant when page is already known not to be PageAnon */ |
| 1402 | static inline int mm_counter_file(struct page *page) |
| 1403 | { |
| 1404 | if (PageSwapBacked(page)) |
| 1405 | return MM_SHMEMPAGES; |
| 1406 | return MM_FILEPAGES; |
| 1407 | } |
| 1408 | |
| 1409 | static inline int mm_counter(struct page *page) |
| 1410 | { |
| 1411 | if (PageAnon(page)) |
| 1412 | return MM_ANONPAGES; |
| 1413 | return mm_counter_file(page); |
| 1414 | } |
| 1415 | |
| 1416 | static inline unsigned long get_mm_rss(struct mm_struct *mm) |
| 1417 | { |
| 1418 | return get_mm_counter(mm, MM_FILEPAGES) + |
| 1419 | get_mm_counter(mm, MM_ANONPAGES) + |
| 1420 | get_mm_counter(mm, MM_SHMEMPAGES); |
| 1421 | } |
| 1422 | |
| 1423 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) |
| 1424 | { |
| 1425 | return max(mm->hiwater_rss, get_mm_rss(mm)); |
| 1426 | } |
| 1427 | |
| 1428 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) |
| 1429 | { |
| 1430 | return max(mm->hiwater_vm, mm->total_vm); |
| 1431 | } |
| 1432 | |
| 1433 | static inline void update_hiwater_rss(struct mm_struct *mm) |
| 1434 | { |
| 1435 | unsigned long _rss = get_mm_rss(mm); |
| 1436 | |
| 1437 | if ((mm)->hiwater_rss < _rss) |
| 1438 | (mm)->hiwater_rss = _rss; |
| 1439 | } |
| 1440 | |
| 1441 | static inline void update_hiwater_vm(struct mm_struct *mm) |
| 1442 | { |
| 1443 | if (mm->hiwater_vm < mm->total_vm) |
| 1444 | mm->hiwater_vm = mm->total_vm; |
| 1445 | } |
| 1446 | |
| 1447 | static inline void reset_mm_hiwater_rss(struct mm_struct *mm) |
| 1448 | { |
| 1449 | mm->hiwater_rss = get_mm_rss(mm); |
| 1450 | } |
| 1451 | |
| 1452 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, |
| 1453 | struct mm_struct *mm) |
| 1454 | { |
| 1455 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); |
| 1456 | |
| 1457 | if (*maxrss < hiwater_rss) |
| 1458 | *maxrss = hiwater_rss; |
| 1459 | } |
| 1460 | |
| 1461 | #if defined(SPLIT_RSS_COUNTING) |
| 1462 | void sync_mm_rss(struct mm_struct *mm); |
| 1463 | #else |
| 1464 | static inline void sync_mm_rss(struct mm_struct *mm) |
| 1465 | { |
| 1466 | } |
| 1467 | #endif |
| 1468 | |
| 1469 | #ifndef __HAVE_ARCH_PTE_DEVMAP |
| 1470 | static inline int pte_devmap(pte_t pte) |
| 1471 | { |
| 1472 | return 0; |
| 1473 | } |
| 1474 | #endif |
| 1475 | |
| 1476 | int vma_wants_writenotify(struct vm_area_struct *vma); |
| 1477 | |
| 1478 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
| 1479 | spinlock_t **ptl); |
| 1480 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, |
| 1481 | spinlock_t **ptl) |
| 1482 | { |
| 1483 | pte_t *ptep; |
| 1484 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); |
| 1485 | return ptep; |
| 1486 | } |
| 1487 | |
| 1488 | #ifdef __PAGETABLE_PUD_FOLDED |
| 1489 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, |
| 1490 | unsigned long address) |
| 1491 | { |
| 1492 | return 0; |
| 1493 | } |
| 1494 | #else |
| 1495 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
| 1496 | #endif |
| 1497 | |
| 1498 | #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU) |
| 1499 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
| 1500 | unsigned long address) |
| 1501 | { |
| 1502 | return 0; |
| 1503 | } |
| 1504 | |
| 1505 | static inline void mm_nr_pmds_init(struct mm_struct *mm) {} |
| 1506 | |
| 1507 | static inline unsigned long mm_nr_pmds(struct mm_struct *mm) |
| 1508 | { |
| 1509 | return 0; |
| 1510 | } |
| 1511 | |
| 1512 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) {} |
| 1513 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) {} |
| 1514 | |
| 1515 | #else |
| 1516 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
| 1517 | |
| 1518 | static inline void mm_nr_pmds_init(struct mm_struct *mm) |
| 1519 | { |
| 1520 | atomic_long_set(&mm->nr_pmds, 0); |
| 1521 | } |
| 1522 | |
| 1523 | static inline unsigned long mm_nr_pmds(struct mm_struct *mm) |
| 1524 | { |
| 1525 | return atomic_long_read(&mm->nr_pmds); |
| 1526 | } |
| 1527 | |
| 1528 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) |
| 1529 | { |
| 1530 | atomic_long_inc(&mm->nr_pmds); |
| 1531 | } |
| 1532 | |
| 1533 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) |
| 1534 | { |
| 1535 | atomic_long_dec(&mm->nr_pmds); |
| 1536 | } |
| 1537 | #endif |
| 1538 | |
| 1539 | int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, |
| 1540 | pmd_t *pmd, unsigned long address); |
| 1541 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); |
| 1542 | |
| 1543 | /* |
| 1544 | * The following ifdef needed to get the 4level-fixup.h header to work. |
| 1545 | * Remove it when 4level-fixup.h has been removed. |
| 1546 | */ |
| 1547 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
| 1548 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
| 1549 | { |
| 1550 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? |
| 1551 | NULL: pud_offset(pgd, address); |
| 1552 | } |
| 1553 | |
| 1554 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
| 1555 | { |
| 1556 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
| 1557 | NULL: pmd_offset(pud, address); |
| 1558 | } |
| 1559 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
| 1560 | |
| 1561 | #if USE_SPLIT_PTE_PTLOCKS |
| 1562 | #if ALLOC_SPLIT_PTLOCKS |
| 1563 | void __init ptlock_cache_init(void); |
| 1564 | extern bool ptlock_alloc(struct page *page); |
| 1565 | extern void ptlock_free(struct page *page); |
| 1566 | |
| 1567 | static inline spinlock_t *ptlock_ptr(struct page *page) |
| 1568 | { |
| 1569 | return page->ptl; |
| 1570 | } |
| 1571 | #else /* ALLOC_SPLIT_PTLOCKS */ |
| 1572 | static inline void ptlock_cache_init(void) |
| 1573 | { |
| 1574 | } |
| 1575 | |
| 1576 | static inline bool ptlock_alloc(struct page *page) |
| 1577 | { |
| 1578 | return true; |
| 1579 | } |
| 1580 | |
| 1581 | static inline void ptlock_free(struct page *page) |
| 1582 | { |
| 1583 | } |
| 1584 | |
| 1585 | static inline spinlock_t *ptlock_ptr(struct page *page) |
| 1586 | { |
| 1587 | return &page->ptl; |
| 1588 | } |
| 1589 | #endif /* ALLOC_SPLIT_PTLOCKS */ |
| 1590 | |
| 1591 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 1592 | { |
| 1593 | return ptlock_ptr(pmd_page(*pmd)); |
| 1594 | } |
| 1595 | |
| 1596 | static inline bool ptlock_init(struct page *page) |
| 1597 | { |
| 1598 | /* |
| 1599 | * prep_new_page() initialize page->private (and therefore page->ptl) |
| 1600 | * with 0. Make sure nobody took it in use in between. |
| 1601 | * |
| 1602 | * It can happen if arch try to use slab for page table allocation: |
| 1603 | * slab code uses page->slab_cache, which share storage with page->ptl. |
| 1604 | */ |
| 1605 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); |
| 1606 | if (!ptlock_alloc(page)) |
| 1607 | return false; |
| 1608 | spin_lock_init(ptlock_ptr(page)); |
| 1609 | return true; |
| 1610 | } |
| 1611 | |
| 1612 | /* Reset page->mapping so free_pages_check won't complain. */ |
| 1613 | static inline void pte_lock_deinit(struct page *page) |
| 1614 | { |
| 1615 | page->mapping = NULL; |
| 1616 | ptlock_free(page); |
| 1617 | } |
| 1618 | |
| 1619 | #else /* !USE_SPLIT_PTE_PTLOCKS */ |
| 1620 | /* |
| 1621 | * We use mm->page_table_lock to guard all pagetable pages of the mm. |
| 1622 | */ |
| 1623 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 1624 | { |
| 1625 | return &mm->page_table_lock; |
| 1626 | } |
| 1627 | static inline void ptlock_cache_init(void) {} |
| 1628 | static inline bool ptlock_init(struct page *page) { return true; } |
| 1629 | static inline void pte_lock_deinit(struct page *page) {} |
| 1630 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
| 1631 | |
| 1632 | static inline void pgtable_init(void) |
| 1633 | { |
| 1634 | ptlock_cache_init(); |
| 1635 | pgtable_cache_init(); |
| 1636 | } |
| 1637 | |
| 1638 | static inline bool pgtable_page_ctor(struct page *page) |
| 1639 | { |
| 1640 | if (!ptlock_init(page)) |
| 1641 | return false; |
| 1642 | inc_zone_page_state(page, NR_PAGETABLE); |
| 1643 | return true; |
| 1644 | } |
| 1645 | |
| 1646 | static inline void pgtable_page_dtor(struct page *page) |
| 1647 | { |
| 1648 | pte_lock_deinit(page); |
| 1649 | dec_zone_page_state(page, NR_PAGETABLE); |
| 1650 | } |
| 1651 | |
| 1652 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
| 1653 | ({ \ |
| 1654 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
| 1655 | pte_t *__pte = pte_offset_map(pmd, address); \ |
| 1656 | *(ptlp) = __ptl; \ |
| 1657 | spin_lock(__ptl); \ |
| 1658 | __pte; \ |
| 1659 | }) |
| 1660 | |
| 1661 | #define pte_unmap_unlock(pte, ptl) do { \ |
| 1662 | spin_unlock(ptl); \ |
| 1663 | pte_unmap(pte); \ |
| 1664 | } while (0) |
| 1665 | |
| 1666 | #define pte_alloc_map(mm, vma, pmd, address) \ |
| 1667 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ |
| 1668 | pmd, address))? \ |
| 1669 | NULL: pte_offset_map(pmd, address)) |
| 1670 | |
| 1671 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
| 1672 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ |
| 1673 | pmd, address))? \ |
| 1674 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) |
| 1675 | |
| 1676 | #define pte_alloc_kernel(pmd, address) \ |
| 1677 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ |
| 1678 | NULL: pte_offset_kernel(pmd, address)) |
| 1679 | |
| 1680 | #if USE_SPLIT_PMD_PTLOCKS |
| 1681 | |
| 1682 | static struct page *pmd_to_page(pmd_t *pmd) |
| 1683 | { |
| 1684 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); |
| 1685 | return virt_to_page((void *)((unsigned long) pmd & mask)); |
| 1686 | } |
| 1687 | |
| 1688 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 1689 | { |
| 1690 | return ptlock_ptr(pmd_to_page(pmd)); |
| 1691 | } |
| 1692 | |
| 1693 | static inline bool pgtable_pmd_page_ctor(struct page *page) |
| 1694 | { |
| 1695 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 1696 | page->pmd_huge_pte = NULL; |
| 1697 | #endif |
| 1698 | return ptlock_init(page); |
| 1699 | } |
| 1700 | |
| 1701 | static inline void pgtable_pmd_page_dtor(struct page *page) |
| 1702 | { |
| 1703 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 1704 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); |
| 1705 | #endif |
| 1706 | ptlock_free(page); |
| 1707 | } |
| 1708 | |
| 1709 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) |
| 1710 | |
| 1711 | #else |
| 1712 | |
| 1713 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 1714 | { |
| 1715 | return &mm->page_table_lock; |
| 1716 | } |
| 1717 | |
| 1718 | static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } |
| 1719 | static inline void pgtable_pmd_page_dtor(struct page *page) {} |
| 1720 | |
| 1721 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) |
| 1722 | |
| 1723 | #endif |
| 1724 | |
| 1725 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) |
| 1726 | { |
| 1727 | spinlock_t *ptl = pmd_lockptr(mm, pmd); |
| 1728 | spin_lock(ptl); |
| 1729 | return ptl; |
| 1730 | } |
| 1731 | |
| 1732 | extern void free_area_init(unsigned long * zones_size); |
| 1733 | extern void free_area_init_node(int nid, unsigned long * zones_size, |
| 1734 | unsigned long zone_start_pfn, unsigned long *zholes_size); |
| 1735 | extern void free_initmem(void); |
| 1736 | |
| 1737 | /* |
| 1738 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) |
| 1739 | * into the buddy system. The freed pages will be poisoned with pattern |
| 1740 | * "poison" if it's within range [0, UCHAR_MAX]. |
| 1741 | * Return pages freed into the buddy system. |
| 1742 | */ |
| 1743 | extern unsigned long free_reserved_area(void *start, void *end, |
| 1744 | int poison, char *s); |
| 1745 | |
| 1746 | #ifdef CONFIG_HIGHMEM |
| 1747 | /* |
| 1748 | * Free a highmem page into the buddy system, adjusting totalhigh_pages |
| 1749 | * and totalram_pages. |
| 1750 | */ |
| 1751 | extern void free_highmem_page(struct page *page); |
| 1752 | #endif |
| 1753 | |
| 1754 | extern void adjust_managed_page_count(struct page *page, long count); |
| 1755 | extern void mem_init_print_info(const char *str); |
| 1756 | |
| 1757 | extern void reserve_bootmem_region(unsigned long start, unsigned long end); |
| 1758 | |
| 1759 | /* Free the reserved page into the buddy system, so it gets managed. */ |
| 1760 | static inline void __free_reserved_page(struct page *page) |
| 1761 | { |
| 1762 | ClearPageReserved(page); |
| 1763 | init_page_count(page); |
| 1764 | __free_page(page); |
| 1765 | } |
| 1766 | |
| 1767 | static inline void free_reserved_page(struct page *page) |
| 1768 | { |
| 1769 | __free_reserved_page(page); |
| 1770 | adjust_managed_page_count(page, 1); |
| 1771 | } |
| 1772 | |
| 1773 | static inline void mark_page_reserved(struct page *page) |
| 1774 | { |
| 1775 | SetPageReserved(page); |
| 1776 | adjust_managed_page_count(page, -1); |
| 1777 | } |
| 1778 | |
| 1779 | /* |
| 1780 | * Default method to free all the __init memory into the buddy system. |
| 1781 | * The freed pages will be poisoned with pattern "poison" if it's within |
| 1782 | * range [0, UCHAR_MAX]. |
| 1783 | * Return pages freed into the buddy system. |
| 1784 | */ |
| 1785 | static inline unsigned long free_initmem_default(int poison) |
| 1786 | { |
| 1787 | extern char __init_begin[], __init_end[]; |
| 1788 | |
| 1789 | return free_reserved_area(&__init_begin, &__init_end, |
| 1790 | poison, "unused kernel"); |
| 1791 | } |
| 1792 | |
| 1793 | static inline unsigned long get_num_physpages(void) |
| 1794 | { |
| 1795 | int nid; |
| 1796 | unsigned long phys_pages = 0; |
| 1797 | |
| 1798 | for_each_online_node(nid) |
| 1799 | phys_pages += node_present_pages(nid); |
| 1800 | |
| 1801 | return phys_pages; |
| 1802 | } |
| 1803 | |
| 1804 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
| 1805 | /* |
| 1806 | * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its |
| 1807 | * zones, allocate the backing mem_map and account for memory holes in a more |
| 1808 | * architecture independent manner. This is a substitute for creating the |
| 1809 | * zone_sizes[] and zholes_size[] arrays and passing them to |
| 1810 | * free_area_init_node() |
| 1811 | * |
| 1812 | * An architecture is expected to register range of page frames backed by |
| 1813 | * physical memory with memblock_add[_node]() before calling |
| 1814 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic |
| 1815 | * usage, an architecture is expected to do something like |
| 1816 | * |
| 1817 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, |
| 1818 | * max_highmem_pfn}; |
| 1819 | * for_each_valid_physical_page_range() |
| 1820 | * memblock_add_node(base, size, nid) |
| 1821 | * free_area_init_nodes(max_zone_pfns); |
| 1822 | * |
| 1823 | * free_bootmem_with_active_regions() calls free_bootmem_node() for each |
| 1824 | * registered physical page range. Similarly |
| 1825 | * sparse_memory_present_with_active_regions() calls memory_present() for |
| 1826 | * each range when SPARSEMEM is enabled. |
| 1827 | * |
| 1828 | * See mm/page_alloc.c for more information on each function exposed by |
| 1829 | * CONFIG_HAVE_MEMBLOCK_NODE_MAP. |
| 1830 | */ |
| 1831 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); |
| 1832 | unsigned long node_map_pfn_alignment(void); |
| 1833 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, |
| 1834 | unsigned long end_pfn); |
| 1835 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
| 1836 | unsigned long end_pfn); |
| 1837 | extern void get_pfn_range_for_nid(unsigned int nid, |
| 1838 | unsigned long *start_pfn, unsigned long *end_pfn); |
| 1839 | extern unsigned long find_min_pfn_with_active_regions(void); |
| 1840 | extern void free_bootmem_with_active_regions(int nid, |
| 1841 | unsigned long max_low_pfn); |
| 1842 | extern void sparse_memory_present_with_active_regions(int nid); |
| 1843 | |
| 1844 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ |
| 1845 | |
| 1846 | #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ |
| 1847 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) |
| 1848 | static inline int __early_pfn_to_nid(unsigned long pfn, |
| 1849 | struct mminit_pfnnid_cache *state) |
| 1850 | { |
| 1851 | return 0; |
| 1852 | } |
| 1853 | #else |
| 1854 | /* please see mm/page_alloc.c */ |
| 1855 | extern int __meminit early_pfn_to_nid(unsigned long pfn); |
| 1856 | /* there is a per-arch backend function. */ |
| 1857 | extern int __meminit __early_pfn_to_nid(unsigned long pfn, |
| 1858 | struct mminit_pfnnid_cache *state); |
| 1859 | #endif |
| 1860 | |
| 1861 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
| 1862 | extern void memmap_init_zone(unsigned long, int, unsigned long, |
| 1863 | unsigned long, enum memmap_context); |
| 1864 | extern void setup_per_zone_wmarks(void); |
| 1865 | extern int __meminit init_per_zone_wmark_min(void); |
| 1866 | extern void mem_init(void); |
| 1867 | extern void __init mmap_init(void); |
| 1868 | extern void show_mem(unsigned int flags); |
| 1869 | extern void si_meminfo(struct sysinfo * val); |
| 1870 | extern void si_meminfo_node(struct sysinfo *val, int nid); |
| 1871 | |
| 1872 | extern __printf(3, 4) |
| 1873 | void warn_alloc_failed(gfp_t gfp_mask, unsigned int order, |
| 1874 | const char *fmt, ...); |
| 1875 | |
| 1876 | extern void setup_per_cpu_pageset(void); |
| 1877 | |
| 1878 | extern void zone_pcp_update(struct zone *zone); |
| 1879 | extern void zone_pcp_reset(struct zone *zone); |
| 1880 | |
| 1881 | /* page_alloc.c */ |
| 1882 | extern int min_free_kbytes; |
| 1883 | |
| 1884 | /* nommu.c */ |
| 1885 | extern atomic_long_t mmap_pages_allocated; |
| 1886 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
| 1887 | |
| 1888 | /* interval_tree.c */ |
| 1889 | void vma_interval_tree_insert(struct vm_area_struct *node, |
| 1890 | struct rb_root *root); |
| 1891 | void vma_interval_tree_insert_after(struct vm_area_struct *node, |
| 1892 | struct vm_area_struct *prev, |
| 1893 | struct rb_root *root); |
| 1894 | void vma_interval_tree_remove(struct vm_area_struct *node, |
| 1895 | struct rb_root *root); |
| 1896 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root, |
| 1897 | unsigned long start, unsigned long last); |
| 1898 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, |
| 1899 | unsigned long start, unsigned long last); |
| 1900 | |
| 1901 | #define vma_interval_tree_foreach(vma, root, start, last) \ |
| 1902 | for (vma = vma_interval_tree_iter_first(root, start, last); \ |
| 1903 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) |
| 1904 | |
| 1905 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, |
| 1906 | struct rb_root *root); |
| 1907 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, |
| 1908 | struct rb_root *root); |
| 1909 | struct anon_vma_chain *anon_vma_interval_tree_iter_first( |
| 1910 | struct rb_root *root, unsigned long start, unsigned long last); |
| 1911 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( |
| 1912 | struct anon_vma_chain *node, unsigned long start, unsigned long last); |
| 1913 | #ifdef CONFIG_DEBUG_VM_RB |
| 1914 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); |
| 1915 | #endif |
| 1916 | |
| 1917 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ |
| 1918 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ |
| 1919 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) |
| 1920 | |
| 1921 | /* mmap.c */ |
| 1922 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
| 1923 | extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, |
| 1924 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); |
| 1925 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
| 1926 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, |
| 1927 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, |
| 1928 | struct mempolicy *, struct vm_userfaultfd_ctx); |
| 1929 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
| 1930 | extern int split_vma(struct mm_struct *, |
| 1931 | struct vm_area_struct *, unsigned long addr, int new_below); |
| 1932 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
| 1933 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, |
| 1934 | struct rb_node **, struct rb_node *); |
| 1935 | extern void unlink_file_vma(struct vm_area_struct *); |
| 1936 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
| 1937 | unsigned long addr, unsigned long len, pgoff_t pgoff, |
| 1938 | bool *need_rmap_locks); |
| 1939 | extern void exit_mmap(struct mm_struct *); |
| 1940 | |
| 1941 | static inline int check_data_rlimit(unsigned long rlim, |
| 1942 | unsigned long new, |
| 1943 | unsigned long start, |
| 1944 | unsigned long end_data, |
| 1945 | unsigned long start_data) |
| 1946 | { |
| 1947 | if (rlim < RLIM_INFINITY) { |
| 1948 | if (((new - start) + (end_data - start_data)) > rlim) |
| 1949 | return -ENOSPC; |
| 1950 | } |
| 1951 | |
| 1952 | return 0; |
| 1953 | } |
| 1954 | |
| 1955 | extern int mm_take_all_locks(struct mm_struct *mm); |
| 1956 | extern void mm_drop_all_locks(struct mm_struct *mm); |
| 1957 | |
| 1958 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); |
| 1959 | extern struct file *get_mm_exe_file(struct mm_struct *mm); |
| 1960 | |
| 1961 | extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages); |
| 1962 | extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages); |
| 1963 | |
| 1964 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, |
| 1965 | unsigned long addr, unsigned long len, |
| 1966 | unsigned long flags, |
| 1967 | const struct vm_special_mapping *spec); |
| 1968 | /* This is an obsolete alternative to _install_special_mapping. */ |
| 1969 | extern int install_special_mapping(struct mm_struct *mm, |
| 1970 | unsigned long addr, unsigned long len, |
| 1971 | unsigned long flags, struct page **pages); |
| 1972 | |
| 1973 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
| 1974 | |
| 1975 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
| 1976 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff); |
| 1977 | extern unsigned long do_mmap(struct file *file, unsigned long addr, |
| 1978 | unsigned long len, unsigned long prot, unsigned long flags, |
| 1979 | vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate); |
| 1980 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); |
| 1981 | |
| 1982 | static inline unsigned long |
| 1983 | do_mmap_pgoff(struct file *file, unsigned long addr, |
| 1984 | unsigned long len, unsigned long prot, unsigned long flags, |
| 1985 | unsigned long pgoff, unsigned long *populate) |
| 1986 | { |
| 1987 | return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate); |
| 1988 | } |
| 1989 | |
| 1990 | #ifdef CONFIG_MMU |
| 1991 | extern int __mm_populate(unsigned long addr, unsigned long len, |
| 1992 | int ignore_errors); |
| 1993 | static inline void mm_populate(unsigned long addr, unsigned long len) |
| 1994 | { |
| 1995 | /* Ignore errors */ |
| 1996 | (void) __mm_populate(addr, len, 1); |
| 1997 | } |
| 1998 | #else |
| 1999 | static inline void mm_populate(unsigned long addr, unsigned long len) {} |
| 2000 | #endif |
| 2001 | |
| 2002 | /* These take the mm semaphore themselves */ |
| 2003 | extern unsigned long vm_brk(unsigned long, unsigned long); |
| 2004 | extern int vm_munmap(unsigned long, size_t); |
| 2005 | extern unsigned long vm_mmap(struct file *, unsigned long, |
| 2006 | unsigned long, unsigned long, |
| 2007 | unsigned long, unsigned long); |
| 2008 | |
| 2009 | struct vm_unmapped_area_info { |
| 2010 | #define VM_UNMAPPED_AREA_TOPDOWN 1 |
| 2011 | unsigned long flags; |
| 2012 | unsigned long length; |
| 2013 | unsigned long low_limit; |
| 2014 | unsigned long high_limit; |
| 2015 | unsigned long align_mask; |
| 2016 | unsigned long align_offset; |
| 2017 | }; |
| 2018 | |
| 2019 | extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); |
| 2020 | extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); |
| 2021 | |
| 2022 | /* |
| 2023 | * Search for an unmapped address range. |
| 2024 | * |
| 2025 | * We are looking for a range that: |
| 2026 | * - does not intersect with any VMA; |
| 2027 | * - is contained within the [low_limit, high_limit) interval; |
| 2028 | * - is at least the desired size. |
| 2029 | * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) |
| 2030 | */ |
| 2031 | static inline unsigned long |
| 2032 | vm_unmapped_area(struct vm_unmapped_area_info *info) |
| 2033 | { |
| 2034 | if (info->flags & VM_UNMAPPED_AREA_TOPDOWN) |
| 2035 | return unmapped_area_topdown(info); |
| 2036 | else |
| 2037 | return unmapped_area(info); |
| 2038 | } |
| 2039 | |
| 2040 | /* truncate.c */ |
| 2041 | extern void truncate_inode_pages(struct address_space *, loff_t); |
| 2042 | extern void truncate_inode_pages_range(struct address_space *, |
| 2043 | loff_t lstart, loff_t lend); |
| 2044 | extern void truncate_inode_pages_final(struct address_space *); |
| 2045 | |
| 2046 | /* generic vm_area_ops exported for stackable file systems */ |
| 2047 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); |
| 2048 | extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf); |
| 2049 | extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); |
| 2050 | |
| 2051 | /* mm/page-writeback.c */ |
| 2052 | int write_one_page(struct page *page, int wait); |
| 2053 | void task_dirty_inc(struct task_struct *tsk); |
| 2054 | |
| 2055 | /* readahead.c */ |
| 2056 | #define VM_MAX_READAHEAD 128 /* kbytes */ |
| 2057 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ |
| 2058 | |
| 2059 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
| 2060 | pgoff_t offset, unsigned long nr_to_read); |
| 2061 | |
| 2062 | void page_cache_sync_readahead(struct address_space *mapping, |
| 2063 | struct file_ra_state *ra, |
| 2064 | struct file *filp, |
| 2065 | pgoff_t offset, |
| 2066 | unsigned long size); |
| 2067 | |
| 2068 | void page_cache_async_readahead(struct address_space *mapping, |
| 2069 | struct file_ra_state *ra, |
| 2070 | struct file *filp, |
| 2071 | struct page *pg, |
| 2072 | pgoff_t offset, |
| 2073 | unsigned long size); |
| 2074 | |
| 2075 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ |
| 2076 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
| 2077 | |
| 2078 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ |
| 2079 | extern int expand_downwards(struct vm_area_struct *vma, |
| 2080 | unsigned long address); |
| 2081 | #if VM_GROWSUP |
| 2082 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
| 2083 | #else |
| 2084 | #define expand_upwards(vma, address) (0) |
| 2085 | #endif |
| 2086 | |
| 2087 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ |
| 2088 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); |
| 2089 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, |
| 2090 | struct vm_area_struct **pprev); |
| 2091 | |
| 2092 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, |
| 2093 | NULL if none. Assume start_addr < end_addr. */ |
| 2094 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) |
| 2095 | { |
| 2096 | struct vm_area_struct * vma = find_vma(mm,start_addr); |
| 2097 | |
| 2098 | if (vma && end_addr <= vma->vm_start) |
| 2099 | vma = NULL; |
| 2100 | return vma; |
| 2101 | } |
| 2102 | |
| 2103 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
| 2104 | { |
| 2105 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
| 2106 | } |
| 2107 | |
| 2108 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ |
| 2109 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, |
| 2110 | unsigned long vm_start, unsigned long vm_end) |
| 2111 | { |
| 2112 | struct vm_area_struct *vma = find_vma(mm, vm_start); |
| 2113 | |
| 2114 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) |
| 2115 | vma = NULL; |
| 2116 | |
| 2117 | return vma; |
| 2118 | } |
| 2119 | |
| 2120 | #ifdef CONFIG_MMU |
| 2121 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
| 2122 | void vma_set_page_prot(struct vm_area_struct *vma); |
| 2123 | #else |
| 2124 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) |
| 2125 | { |
| 2126 | return __pgprot(0); |
| 2127 | } |
| 2128 | static inline void vma_set_page_prot(struct vm_area_struct *vma) |
| 2129 | { |
| 2130 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
| 2131 | } |
| 2132 | #endif |
| 2133 | |
| 2134 | #ifdef CONFIG_NUMA_BALANCING |
| 2135 | unsigned long change_prot_numa(struct vm_area_struct *vma, |
| 2136 | unsigned long start, unsigned long end); |
| 2137 | #endif |
| 2138 | |
| 2139 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
| 2140 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
| 2141 | unsigned long pfn, unsigned long size, pgprot_t); |
| 2142 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
| 2143 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
| 2144 | unsigned long pfn); |
| 2145 | int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr, |
| 2146 | unsigned long pfn, pgprot_t pgprot); |
| 2147 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
| 2148 | pfn_t pfn); |
| 2149 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); |
| 2150 | |
| 2151 | |
| 2152 | struct page *follow_page_mask(struct vm_area_struct *vma, |
| 2153 | unsigned long address, unsigned int foll_flags, |
| 2154 | unsigned int *page_mask); |
| 2155 | |
| 2156 | static inline struct page *follow_page(struct vm_area_struct *vma, |
| 2157 | unsigned long address, unsigned int foll_flags) |
| 2158 | { |
| 2159 | unsigned int unused_page_mask; |
| 2160 | return follow_page_mask(vma, address, foll_flags, &unused_page_mask); |
| 2161 | } |
| 2162 | |
| 2163 | #define FOLL_WRITE 0x01 /* check pte is writable */ |
| 2164 | #define FOLL_TOUCH 0x02 /* mark page accessed */ |
| 2165 | #define FOLL_GET 0x04 /* do get_page on page */ |
| 2166 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ |
| 2167 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ |
| 2168 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO |
| 2169 | * and return without waiting upon it */ |
| 2170 | #define FOLL_POPULATE 0x40 /* fault in page */ |
| 2171 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ |
| 2172 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ |
| 2173 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ |
| 2174 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ |
| 2175 | #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */ |
| 2176 | #define FOLL_MLOCK 0x1000 /* lock present pages */ |
| 2177 | #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */ |
| 2178 | |
| 2179 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, |
| 2180 | void *data); |
| 2181 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, |
| 2182 | unsigned long size, pte_fn_t fn, void *data); |
| 2183 | |
| 2184 | |
| 2185 | #ifdef CONFIG_DEBUG_PAGEALLOC |
| 2186 | extern bool _debug_pagealloc_enabled; |
| 2187 | extern void __kernel_map_pages(struct page *page, int numpages, int enable); |
| 2188 | |
| 2189 | static inline bool debug_pagealloc_enabled(void) |
| 2190 | { |
| 2191 | return _debug_pagealloc_enabled; |
| 2192 | } |
| 2193 | |
| 2194 | static inline void |
| 2195 | kernel_map_pages(struct page *page, int numpages, int enable) |
| 2196 | { |
| 2197 | if (!debug_pagealloc_enabled()) |
| 2198 | return; |
| 2199 | |
| 2200 | __kernel_map_pages(page, numpages, enable); |
| 2201 | } |
| 2202 | #ifdef CONFIG_HIBERNATION |
| 2203 | extern bool kernel_page_present(struct page *page); |
| 2204 | #endif /* CONFIG_HIBERNATION */ |
| 2205 | #else |
| 2206 | static inline void |
| 2207 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
| 2208 | #ifdef CONFIG_HIBERNATION |
| 2209 | static inline bool kernel_page_present(struct page *page) { return true; } |
| 2210 | #endif /* CONFIG_HIBERNATION */ |
| 2211 | #endif |
| 2212 | |
| 2213 | #ifdef __HAVE_ARCH_GATE_AREA |
| 2214 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
| 2215 | extern int in_gate_area_no_mm(unsigned long addr); |
| 2216 | extern int in_gate_area(struct mm_struct *mm, unsigned long addr); |
| 2217 | #else |
| 2218 | static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) |
| 2219 | { |
| 2220 | return NULL; |
| 2221 | } |
| 2222 | static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } |
| 2223 | static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) |
| 2224 | { |
| 2225 | return 0; |
| 2226 | } |
| 2227 | #endif /* __HAVE_ARCH_GATE_AREA */ |
| 2228 | |
| 2229 | #ifdef CONFIG_SYSCTL |
| 2230 | extern int sysctl_drop_caches; |
| 2231 | int drop_caches_sysctl_handler(struct ctl_table *, int, |
| 2232 | void __user *, size_t *, loff_t *); |
| 2233 | #endif |
| 2234 | |
| 2235 | void drop_slab(void); |
| 2236 | void drop_slab_node(int nid); |
| 2237 | |
| 2238 | #ifndef CONFIG_MMU |
| 2239 | #define randomize_va_space 0 |
| 2240 | #else |
| 2241 | extern int randomize_va_space; |
| 2242 | #endif |
| 2243 | |
| 2244 | const char * arch_vma_name(struct vm_area_struct *vma); |
| 2245 | void print_vma_addr(char *prefix, unsigned long rip); |
| 2246 | |
| 2247 | void sparse_mem_maps_populate_node(struct page **map_map, |
| 2248 | unsigned long pnum_begin, |
| 2249 | unsigned long pnum_end, |
| 2250 | unsigned long map_count, |
| 2251 | int nodeid); |
| 2252 | |
| 2253 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); |
| 2254 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
| 2255 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); |
| 2256 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); |
| 2257 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); |
| 2258 | void *vmemmap_alloc_block(unsigned long size, int node); |
| 2259 | struct vmem_altmap; |
| 2260 | void *__vmemmap_alloc_block_buf(unsigned long size, int node, |
| 2261 | struct vmem_altmap *altmap); |
| 2262 | static inline void *vmemmap_alloc_block_buf(unsigned long size, int node) |
| 2263 | { |
| 2264 | return __vmemmap_alloc_block_buf(size, node, NULL); |
| 2265 | } |
| 2266 | |
| 2267 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
| 2268 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, |
| 2269 | int node); |
| 2270 | int vmemmap_populate(unsigned long start, unsigned long end, int node); |
| 2271 | void vmemmap_populate_print_last(void); |
| 2272 | #ifdef CONFIG_MEMORY_HOTPLUG |
| 2273 | void vmemmap_free(unsigned long start, unsigned long end); |
| 2274 | #endif |
| 2275 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, |
| 2276 | unsigned long size); |
| 2277 | |
| 2278 | enum mf_flags { |
| 2279 | MF_COUNT_INCREASED = 1 << 0, |
| 2280 | MF_ACTION_REQUIRED = 1 << 1, |
| 2281 | MF_MUST_KILL = 1 << 2, |
| 2282 | MF_SOFT_OFFLINE = 1 << 3, |
| 2283 | }; |
| 2284 | extern int memory_failure(unsigned long pfn, int trapno, int flags); |
| 2285 | extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); |
| 2286 | extern int unpoison_memory(unsigned long pfn); |
| 2287 | extern int get_hwpoison_page(struct page *page); |
| 2288 | #define put_hwpoison_page(page) put_page(page) |
| 2289 | extern int sysctl_memory_failure_early_kill; |
| 2290 | extern int sysctl_memory_failure_recovery; |
| 2291 | extern void shake_page(struct page *p, int access); |
| 2292 | extern atomic_long_t num_poisoned_pages; |
| 2293 | extern int soft_offline_page(struct page *page, int flags); |
| 2294 | |
| 2295 | |
| 2296 | /* |
| 2297 | * Error handlers for various types of pages. |
| 2298 | */ |
| 2299 | enum mf_result { |
| 2300 | MF_IGNORED, /* Error: cannot be handled */ |
| 2301 | MF_FAILED, /* Error: handling failed */ |
| 2302 | MF_DELAYED, /* Will be handled later */ |
| 2303 | MF_RECOVERED, /* Successfully recovered */ |
| 2304 | }; |
| 2305 | |
| 2306 | enum mf_action_page_type { |
| 2307 | MF_MSG_KERNEL, |
| 2308 | MF_MSG_KERNEL_HIGH_ORDER, |
| 2309 | MF_MSG_SLAB, |
| 2310 | MF_MSG_DIFFERENT_COMPOUND, |
| 2311 | MF_MSG_POISONED_HUGE, |
| 2312 | MF_MSG_HUGE, |
| 2313 | MF_MSG_FREE_HUGE, |
| 2314 | MF_MSG_UNMAP_FAILED, |
| 2315 | MF_MSG_DIRTY_SWAPCACHE, |
| 2316 | MF_MSG_CLEAN_SWAPCACHE, |
| 2317 | MF_MSG_DIRTY_MLOCKED_LRU, |
| 2318 | MF_MSG_CLEAN_MLOCKED_LRU, |
| 2319 | MF_MSG_DIRTY_UNEVICTABLE_LRU, |
| 2320 | MF_MSG_CLEAN_UNEVICTABLE_LRU, |
| 2321 | MF_MSG_DIRTY_LRU, |
| 2322 | MF_MSG_CLEAN_LRU, |
| 2323 | MF_MSG_TRUNCATED_LRU, |
| 2324 | MF_MSG_BUDDY, |
| 2325 | MF_MSG_BUDDY_2ND, |
| 2326 | MF_MSG_UNKNOWN, |
| 2327 | }; |
| 2328 | |
| 2329 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
| 2330 | extern void clear_huge_page(struct page *page, |
| 2331 | unsigned long addr, |
| 2332 | unsigned int pages_per_huge_page); |
| 2333 | extern void copy_user_huge_page(struct page *dst, struct page *src, |
| 2334 | unsigned long addr, struct vm_area_struct *vma, |
| 2335 | unsigned int pages_per_huge_page); |
| 2336 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ |
| 2337 | |
| 2338 | extern struct page_ext_operations debug_guardpage_ops; |
| 2339 | extern struct page_ext_operations page_poisoning_ops; |
| 2340 | |
| 2341 | #ifdef CONFIG_DEBUG_PAGEALLOC |
| 2342 | extern unsigned int _debug_guardpage_minorder; |
| 2343 | extern bool _debug_guardpage_enabled; |
| 2344 | |
| 2345 | static inline unsigned int debug_guardpage_minorder(void) |
| 2346 | { |
| 2347 | return _debug_guardpage_minorder; |
| 2348 | } |
| 2349 | |
| 2350 | static inline bool debug_guardpage_enabled(void) |
| 2351 | { |
| 2352 | return _debug_guardpage_enabled; |
| 2353 | } |
| 2354 | |
| 2355 | static inline bool page_is_guard(struct page *page) |
| 2356 | { |
| 2357 | struct page_ext *page_ext; |
| 2358 | |
| 2359 | if (!debug_guardpage_enabled()) |
| 2360 | return false; |
| 2361 | |
| 2362 | page_ext = lookup_page_ext(page); |
| 2363 | return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags); |
| 2364 | } |
| 2365 | #else |
| 2366 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } |
| 2367 | static inline bool debug_guardpage_enabled(void) { return false; } |
| 2368 | static inline bool page_is_guard(struct page *page) { return false; } |
| 2369 | #endif /* CONFIG_DEBUG_PAGEALLOC */ |
| 2370 | |
| 2371 | #if MAX_NUMNODES > 1 |
| 2372 | void __init setup_nr_node_ids(void); |
| 2373 | #else |
| 2374 | static inline void setup_nr_node_ids(void) {} |
| 2375 | #endif |
| 2376 | |
| 2377 | #endif /* __KERNEL__ */ |
| 2378 | #endif /* _LINUX_MM_H */ |