linux/include/linux/mm.h
<<
>>
Prefs
   1#ifndef _LINUX_MM_H
   2#define _LINUX_MM_H
   3
   4#include <linux/errno.h>
   5
   6#ifdef __KERNEL__
   7
   8#include <linux/gfp.h>
   9#include <linux/list.h>
  10#include <linux/mmdebug.h>
  11#include <linux/mmzone.h>
  12#include <linux/rbtree.h>
  13#include <linux/prio_tree.h>
  14#include <linux/debug_locks.h>
  15#include <linux/mm_types.h>
  16
  17struct mempolicy;
  18struct anon_vma;
  19struct file_ra_state;
  20struct user_struct;
  21struct writeback_control;
  22
  23#ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
  24extern unsigned long max_mapnr;
  25#endif
  26
  27extern unsigned long num_physpages;
  28extern void * high_memory;
  29extern int page_cluster;
  30
  31#ifdef CONFIG_SYSCTL
  32extern int sysctl_legacy_va_layout;
  33#else
  34#define sysctl_legacy_va_layout 0
  35#endif
  36
  37extern unsigned long mmap_min_addr;
  38
  39#include <asm/page.h>
  40#include <asm/pgtable.h>
  41#include <asm/processor.h>
  42
  43#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  44
  45/* to align the pointer to the (next) page boundary */
  46#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
  47
  48/*
  49 * Linux kernel virtual memory manager primitives.
  50 * The idea being to have a "virtual" mm in the same way
  51 * we have a virtual fs - giving a cleaner interface to the
  52 * mm details, and allowing different kinds of memory mappings
  53 * (from shared memory to executable loading to arbitrary
  54 * mmap() functions).
  55 */
  56
  57extern struct kmem_cache *vm_area_cachep;
  58
  59/*
  60 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
  61 * disabled, then there's a single shared list of VMAs maintained by the
  62 * system, and mm's subscribe to these individually
  63 */
  64struct vm_list_struct {
  65        struct vm_list_struct   *next;
  66        struct vm_area_struct   *vma;
  67};
  68
  69#ifndef CONFIG_MMU
  70extern struct rb_root nommu_vma_tree;
  71extern struct rw_semaphore nommu_vma_sem;
  72
  73extern unsigned int kobjsize(const void *objp);
  74#endif
  75
  76/*
  77 * vm_flags in vm_area_struct, see mm_types.h.
  78 */
  79#define VM_READ         0x00000001      /* currently active flags */
  80#define VM_WRITE        0x00000002
  81#define VM_EXEC         0x00000004
  82#define VM_SHARED       0x00000008
  83
  84/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  85#define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
  86#define VM_MAYWRITE     0x00000020
  87#define VM_MAYEXEC      0x00000040
  88#define VM_MAYSHARE     0x00000080
  89
  90#define VM_GROWSDOWN    0x00000100      /* general info on the segment */
  91#define VM_GROWSUP      0x00000200
  92#define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
  93#define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
  94
  95#define VM_EXECUTABLE   0x00001000
  96#define VM_LOCKED       0x00002000
  97#define VM_IO           0x00004000      /* Memory mapped I/O or similar */
  98
  99                                        /* Used by sys_madvise() */
 100#define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
 101#define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
 102
 103#define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
 104#define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
 105#define VM_RESERVED     0x00080000      /* Count as reserved_vm like IO */
 106#define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
 107#define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
 108#define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
 109#define VM_NONLINEAR    0x00800000      /* Is non-linear (remap_file_pages) */
 110#define VM_MAPPED_COPY  0x01000000      /* T if mapped copy of data (nommu mmap) */
 111#define VM_INSERTPAGE   0x02000000      /* The vma has had "vm_insert_page()" done on it */
 112#define VM_ALWAYSDUMP   0x04000000      /* Always include in core dumps */
 113
 114#define VM_CAN_NONLINEAR 0x08000000     /* Has ->fault & does nonlinear pages */
 115#define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
 116#define VM_SAO          0x20000000      /* Strong Access Ordering (powerpc) */
 117
 118#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 119#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 120#endif
 121
 122#ifdef CONFIG_STACK_GROWSUP
 123#define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 124#else
 125#define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 126#endif
 127
 128#define VM_READHINTMASK                 (VM_SEQ_READ | VM_RAND_READ)
 129#define VM_ClearReadHint(v)             (v)->vm_flags &= ~VM_READHINTMASK
 130#define VM_NormalReadHint(v)            (!((v)->vm_flags & VM_READHINTMASK))
 131#define VM_SequentialReadHint(v)        ((v)->vm_flags & VM_SEQ_READ)
 132#define VM_RandomReadHint(v)            ((v)->vm_flags & VM_RAND_READ)
 133
 134/*
 135 * special vmas that are non-mergable, non-mlock()able
 136 */
 137#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
 138
 139/*
 140 * mapping from the currently active vm_flags protection bits (the
 141 * low four bits) to a page protection mask..
 142 */
 143extern pgprot_t protection_map[16];
 144
 145#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 146#define FAULT_FLAG_NONLINEAR    0x02    /* Fault was via a nonlinear mapping */
 147
 148
 149/*
 150 * vm_fault is filled by the the pagefault handler and passed to the vma's
 151 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 152 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 153 *
 154 * pgoff should be used in favour of virtual_address, if possible. If pgoff
 155 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
 156 * mapping support.
 157 */
 158struct vm_fault {
 159        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 160        pgoff_t pgoff;                  /* Logical page offset based on vma */
 161        void __user *virtual_address;   /* Faulting virtual address */
 162
 163        struct page *page;              /* ->fault handlers should return a
 164                                         * page here, unless VM_FAULT_NOPAGE
 165                                         * is set (which is also implied by
 166                                         * VM_FAULT_ERROR).
 167                                         */
 168};
 169
 170/*
 171 * These are the virtual MM functions - opening of an area, closing and
 172 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 173 * to the functions called when a no-page or a wp-page exception occurs. 
 174 */
 175struct vm_operations_struct {
 176        void (*open)(struct vm_area_struct * area);
 177        void (*close)(struct vm_area_struct * area);
 178        int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 179
 180        /* notification that a previously read-only page is about to become
 181         * writable, if an error is returned it will cause a SIGBUS */
 182        int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
 183
 184        /* called by access_process_vm when get_user_pages() fails, typically
 185         * for use by special VMAs that can switch between memory and hardware
 186         */
 187        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 188                      void *buf, int len, int write);
 189#ifdef CONFIG_NUMA
 190        /*
 191         * set_policy() op must add a reference to any non-NULL @new mempolicy
 192         * to hold the policy upon return.  Caller should pass NULL @new to
 193         * remove a policy and fall back to surrounding context--i.e. do not
 194         * install a MPOL_DEFAULT policy, nor the task or system default
 195         * mempolicy.
 196         */
 197        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 198
 199        /*
 200         * get_policy() op must add reference [mpol_get()] to any policy at
 201         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 202         * in mm/mempolicy.c will do this automatically.
 203         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 204         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 205         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 206         * must return NULL--i.e., do not "fallback" to task or system default
 207         * policy.
 208         */
 209        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 210                                        unsigned long addr);
 211        int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 212                const nodemask_t *to, unsigned long flags);
 213#endif
 214};
 215
 216struct mmu_gather;
 217struct inode;
 218
 219#define page_private(page)              ((page)->private)
 220#define set_page_private(page, v)       ((page)->private = (v))
 221
 222/*
 223 * FIXME: take this include out, include page-flags.h in
 224 * files which need it (119 of them)
 225 */
 226#include <linux/page-flags.h>
 227
 228/*
 229 * Methods to modify the page usage count.
 230 *
 231 * What counts for a page usage:
 232 * - cache mapping   (page->mapping)
 233 * - private data    (page->private)
 234 * - page mapped in a task's page tables, each mapping
 235 *   is counted separately
 236 *
 237 * Also, many kernel routines increase the page count before a critical
 238 * routine so they can be sure the page doesn't go away from under them.
 239 */
 240
 241/*
 242 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 243 */
 244static inline int put_page_testzero(struct page *page)
 245{
 246        VM_BUG_ON(atomic_read(&page->_count) == 0);
 247        return atomic_dec_and_test(&page->_count);
 248}
 249
 250/*
 251 * Try to grab a ref unless the page has a refcount of zero, return false if
 252 * that is the case.
 253 */
 254static inline int get_page_unless_zero(struct page *page)
 255{
 256        return atomic_inc_not_zero(&page->_count);
 257}
 258
 259/* Support for virtually mapped pages */
 260struct page *vmalloc_to_page(const void *addr);
 261unsigned long vmalloc_to_pfn(const void *addr);
 262
 263/*
 264 * Determine if an address is within the vmalloc range
 265 *
 266 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 267 * is no special casing required.
 268 */
 269static inline int is_vmalloc_addr(const void *x)
 270{
 271#ifdef CONFIG_MMU
 272        unsigned long addr = (unsigned long)x;
 273
 274        return addr >= VMALLOC_START && addr < VMALLOC_END;
 275#else
 276        return 0;
 277#endif
 278}
 279
 280static inline struct page *compound_head(struct page *page)
 281{
 282        if (unlikely(PageTail(page)))
 283                return page->first_page;
 284        return page;
 285}
 286
 287static inline int page_count(struct page *page)
 288{
 289        return atomic_read(&compound_head(page)->_count);
 290}
 291
 292static inline void get_page(struct page *page)
 293{
 294        page = compound_head(page);
 295        VM_BUG_ON(atomic_read(&page->_count) == 0);
 296        atomic_inc(&page->_count);
 297}
 298
 299static inline struct page *virt_to_head_page(const void *x)
 300{
 301        struct page *page = virt_to_page(x);
 302        return compound_head(page);
 303}
 304
 305/*
 306 * Setup the page count before being freed into the page allocator for
 307 * the first time (boot or memory hotplug)
 308 */
 309static inline void init_page_count(struct page *page)
 310{
 311        atomic_set(&page->_count, 1);
 312}
 313
 314void put_page(struct page *page);
 315void put_pages_list(struct list_head *pages);
 316
 317void split_page(struct page *page, unsigned int order);
 318
 319/*
 320 * Compound pages have a destructor function.  Provide a
 321 * prototype for that function and accessor functions.
 322 * These are _only_ valid on the head of a PG_compound page.
 323 */
 324typedef void compound_page_dtor(struct page *);
 325
 326static inline void set_compound_page_dtor(struct page *page,
 327                                                compound_page_dtor *dtor)
 328{
 329        page[1].lru.next = (void *)dtor;
 330}
 331
 332static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 333{
 334        return (compound_page_dtor *)page[1].lru.next;
 335}
 336
 337static inline int compound_order(struct page *page)
 338{
 339        if (!PageHead(page))
 340                return 0;
 341        return (unsigned long)page[1].lru.prev;
 342}
 343
 344static inline void set_compound_order(struct page *page, unsigned long order)
 345{
 346        page[1].lru.prev = (void *)order;
 347}
 348
 349/*
 350 * Multiple processes may "see" the same page. E.g. for untouched
 351 * mappings of /dev/null, all processes see the same page full of
 352 * zeroes, and text pages of executables and shared libraries have
 353 * only one copy in memory, at most, normally.
 354 *
 355 * For the non-reserved pages, page_count(page) denotes a reference count.
 356 *   page_count() == 0 means the page is free. page->lru is then used for
 357 *   freelist management in the buddy allocator.
 358 *   page_count() > 0  means the page has been allocated.
 359 *
 360 * Pages are allocated by the slab allocator in order to provide memory
 361 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 362 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 363 * unless a particular usage is carefully commented. (the responsibility of
 364 * freeing the kmalloc memory is the caller's, of course).
 365 *
 366 * A page may be used by anyone else who does a __get_free_page().
 367 * In this case, page_count still tracks the references, and should only
 368 * be used through the normal accessor functions. The top bits of page->flags
 369 * and page->virtual store page management information, but all other fields
 370 * are unused and could be used privately, carefully. The management of this
 371 * page is the responsibility of the one who allocated it, and those who have
 372 * subsequently been given references to it.
 373 *
 374 * The other pages (we may call them "pagecache pages") are completely
 375 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 376 * The following discussion applies only to them.
 377 *
 378 * A pagecache page contains an opaque `private' member, which belongs to the
 379 * page's address_space. Usually, this is the address of a circular list of
 380 * the page's disk buffers. PG_private must be set to tell the VM to call
 381 * into the filesystem to release these pages.
 382 *
 383 * A page may belong to an inode's memory mapping. In this case, page->mapping
 384 * is the pointer to the inode, and page->index is the file offset of the page,
 385 * in units of PAGE_CACHE_SIZE.
 386 *
 387 * If pagecache pages are not associated with an inode, they are said to be
 388 * anonymous pages. These may become associated with the swapcache, and in that
 389 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 390 *
 391 * In either case (swapcache or inode backed), the pagecache itself holds one
 392 * reference to the page. Setting PG_private should also increment the
 393 * refcount. The each user mapping also has a reference to the page.
 394 *
 395 * The pagecache pages are stored in a per-mapping radix tree, which is
 396 * rooted at mapping->page_tree, and indexed by offset.
 397 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 398 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 399 *
 400 * All pagecache pages may be subject to I/O:
 401 * - inode pages may need to be read from disk,
 402 * - inode pages which have been modified and are MAP_SHARED may need
 403 *   to be written back to the inode on disk,
 404 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 405 *   modified may need to be swapped out to swap space and (later) to be read
 406 *   back into memory.
 407 */
 408
 409/*
 410 * The zone field is never updated after free_area_init_core()
 411 * sets it, so none of the operations on it need to be atomic.
 412 */
 413
 414
 415/*
 416 * page->flags layout:
 417 *
 418 * There are three possibilities for how page->flags get
 419 * laid out.  The first is for the normal case, without
 420 * sparsemem.  The second is for sparsemem when there is
 421 * plenty of space for node and section.  The last is when
 422 * we have run out of space and have to fall back to an
 423 * alternate (slower) way of determining the node.
 424 *
 425 * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
 426 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
 427 * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
 428 */
 429#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 430#define SECTIONS_WIDTH          SECTIONS_SHIFT
 431#else
 432#define SECTIONS_WIDTH          0
 433#endif
 434
 435#define ZONES_WIDTH             ZONES_SHIFT
 436
 437#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
 438#define NODES_WIDTH             NODES_SHIFT
 439#else
 440#ifdef CONFIG_SPARSEMEM_VMEMMAP
 441#error "Vmemmap: No space for nodes field in page flags"
 442#endif
 443#define NODES_WIDTH             0
 444#endif
 445
 446/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
 447#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 448#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 449#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 450
 451/*
 452 * We are going to use the flags for the page to node mapping if its in
 453 * there.  This includes the case where there is no node, so it is implicit.
 454 */
 455#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
 456#define NODE_NOT_IN_PAGE_FLAGS
 457#endif
 458
 459#ifndef PFN_SECTION_SHIFT
 460#define PFN_SECTION_SHIFT 0
 461#endif
 462
 463/*
 464 * Define the bit shifts to access each section.  For non-existant
 465 * sections we define the shift as 0; that plus a 0 mask ensures
 466 * the compiler will optimise away reference to them.
 467 */
 468#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 469#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 470#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 471
 472/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
 473#ifdef NODE_NOT_IN_PAGEFLAGS
 474#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 475#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 476                                                SECTIONS_PGOFF : ZONES_PGOFF)
 477#else
 478#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 479#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 480                                                NODES_PGOFF : ZONES_PGOFF)
 481#endif
 482
 483#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 484
 485#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 486#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 487#endif
 488
 489#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 490#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 491#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 492#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 493
 494static inline enum zone_type page_zonenum(struct page *page)
 495{
 496        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 497}
 498
 499/*
 500 * The identification function is only used by the buddy allocator for
 501 * determining if two pages could be buddies. We are not really
 502 * identifying a zone since we could be using a the section number
 503 * id if we have not node id available in page flags.
 504 * We guarantee only that it will return the same value for two
 505 * combinable pages in a zone.
 506 */
 507static inline int page_zone_id(struct page *page)
 508{
 509        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 510}
 511
 512static inline int zone_to_nid(struct zone *zone)
 513{
 514#ifdef CONFIG_NUMA
 515        return zone->node;
 516#else
 517        return 0;
 518#endif
 519}
 520
 521#ifdef NODE_NOT_IN_PAGE_FLAGS
 522extern int page_to_nid(struct page *page);
 523#else
 524static inline int page_to_nid(struct page *page)
 525{
 526        return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 527}
 528#endif
 529
 530static inline struct zone *page_zone(struct page *page)
 531{
 532        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 533}
 534
 535#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 536static inline unsigned long page_to_section(struct page *page)
 537{
 538        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 539}
 540#endif
 541
 542static inline void set_page_zone(struct page *page, enum zone_type zone)
 543{
 544        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 545        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 546}
 547
 548static inline void set_page_node(struct page *page, unsigned long node)
 549{
 550        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 551        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 552}
 553
 554static inline void set_page_section(struct page *page, unsigned long section)
 555{
 556        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 557        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 558}
 559
 560static inline void set_page_links(struct page *page, enum zone_type zone,
 561        unsigned long node, unsigned long pfn)
 562{
 563        set_page_zone(page, zone);
 564        set_page_node(page, node);
 565        set_page_section(page, pfn_to_section_nr(pfn));
 566}
 567
 568/*
 569 * If a hint addr is less than mmap_min_addr change hint to be as
 570 * low as possible but still greater than mmap_min_addr
 571 */
 572static inline unsigned long round_hint_to_min(unsigned long hint)
 573{
 574#ifdef CONFIG_SECURITY
 575        hint &= PAGE_MASK;
 576        if (((void *)hint != NULL) &&
 577            (hint < mmap_min_addr))
 578                return PAGE_ALIGN(mmap_min_addr);
 579#endif
 580        return hint;
 581}
 582
 583/*
 584 * Some inline functions in vmstat.h depend on page_zone()
 585 */
 586#include <linux/vmstat.h>
 587
 588static __always_inline void *lowmem_page_address(struct page *page)
 589{
 590        return __va(page_to_pfn(page) << PAGE_SHIFT);
 591}
 592
 593#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 594#define HASHED_PAGE_VIRTUAL
 595#endif
 596
 597#if defined(WANT_PAGE_VIRTUAL)
 598#define page_address(page) ((page)->virtual)
 599#define set_page_address(page, address)                 \
 600        do {                                            \
 601                (page)->virtual = (address);            \
 602        } while(0)
 603#define page_address_init()  do { } while(0)
 604#endif
 605
 606#if defined(HASHED_PAGE_VIRTUAL)
 607void *page_address(struct page *page);
 608void set_page_address(struct page *page, void *virtual);
 609void page_address_init(void);
 610#endif
 611
 612#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 613#define page_address(page) lowmem_page_address(page)
 614#define set_page_address(page, address)  do { } while(0)
 615#define page_address_init()  do { } while(0)
 616#endif
 617
 618/*
 619 * On an anonymous page mapped into a user virtual memory area,
 620 * page->mapping points to its anon_vma, not to a struct address_space;
 621 * with the PAGE_MAPPING_ANON bit set to distinguish it.
 622 *
 623 * Please note that, confusingly, "page_mapping" refers to the inode
 624 * address_space which maps the page from disk; whereas "page_mapped"
 625 * refers to user virtual address space into which the page is mapped.
 626 */
 627#define PAGE_MAPPING_ANON       1
 628
 629extern struct address_space swapper_space;
 630static inline struct address_space *page_mapping(struct page *page)
 631{
 632        struct address_space *mapping = page->mapping;
 633
 634        VM_BUG_ON(PageSlab(page));
 635#ifdef CONFIG_SWAP
 636        if (unlikely(PageSwapCache(page)))
 637                mapping = &swapper_space;
 638        else
 639#endif
 640        if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
 641                mapping = NULL;
 642        return mapping;
 643}
 644
 645static inline int PageAnon(struct page *page)
 646{
 647        return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
 648}
 649
 650/*
 651 * Return the pagecache index of the passed page.  Regular pagecache pages
 652 * use ->index whereas swapcache pages use ->private
 653 */
 654static inline pgoff_t page_index(struct page *page)
 655{
 656        if (unlikely(PageSwapCache(page)))
 657                return page_private(page);
 658        return page->index;
 659}
 660
 661/*
 662 * The atomic page->_mapcount, like _count, starts from -1:
 663 * so that transitions both from it and to it can be tracked,
 664 * using atomic_inc_and_test and atomic_add_negative(-1).
 665 */
 666static inline void reset_page_mapcount(struct page *page)
 667{
 668        atomic_set(&(page)->_mapcount, -1);
 669}
 670
 671static inline int page_mapcount(struct page *page)
 672{
 673        return atomic_read(&(page)->_mapcount) + 1;
 674}
 675
 676/*
 677 * Return true if this page is mapped into pagetables.
 678 */
 679static inline int page_mapped(struct page *page)
 680{
 681        return atomic_read(&(page)->_mapcount) >= 0;
 682}
 683
 684/*
 685 * Different kinds of faults, as returned by handle_mm_fault().
 686 * Used to decide whether a process gets delivered SIGBUS or
 687 * just gets major/minor fault counters bumped up.
 688 */
 689
 690#define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
 691
 692#define VM_FAULT_OOM    0x0001
 693#define VM_FAULT_SIGBUS 0x0002
 694#define VM_FAULT_MAJOR  0x0004
 695#define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
 696
 697#define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
 698#define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
 699
 700#define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS)
 701
 702#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
 703
 704extern void show_free_areas(void);
 705
 706#ifdef CONFIG_SHMEM
 707extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
 708#else
 709static inline int shmem_lock(struct file *file, int lock,
 710                            struct user_struct *user)
 711{
 712        return 0;
 713}
 714#endif
 715struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
 716
 717int shmem_zero_setup(struct vm_area_struct *);
 718
 719#ifndef CONFIG_MMU
 720extern unsigned long shmem_get_unmapped_area(struct file *file,
 721                                             unsigned long addr,
 722                                             unsigned long len,
 723                                             unsigned long pgoff,
 724                                             unsigned long flags);
 725#endif
 726
 727extern int can_do_mlock(void);
 728extern int user_shm_lock(size_t, struct user_struct *);
 729extern void user_shm_unlock(size_t, struct user_struct *);
 730
 731/*
 732 * Parameter block passed down to zap_pte_range in exceptional cases.
 733 */
 734struct zap_details {
 735        struct vm_area_struct *nonlinear_vma;   /* Check page->index if set */
 736        struct address_space *check_mapping;    /* Check page->mapping if set */
 737        pgoff_t first_index;                    /* Lowest page->index to unmap */
 738        pgoff_t last_index;                     /* Highest page->index to unmap */
 739        spinlock_t *i_mmap_lock;                /* For unmap_mapping_range: */
 740        unsigned long truncate_count;           /* Compare vm_truncate_count */
 741};
 742
 743struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
 744                pte_t pte);
 745
 746int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 747                unsigned long size);
 748unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 749                unsigned long size, struct zap_details *);
 750unsigned long unmap_vmas(struct mmu_gather **tlb,
 751                struct vm_area_struct *start_vma, unsigned long start_addr,
 752                unsigned long end_addr, unsigned long *nr_accounted,
 753                struct zap_details *);
 754
 755/**
 756 * mm_walk - callbacks for walk_page_range
 757 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
 758 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
 759 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
 760 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
 761 * @pte_hole: if set, called for each hole at all levels
 762 *
 763 * (see walk_page_range for more details)
 764 */
 765struct mm_walk {
 766        int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
 767        int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
 768        int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
 769        int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
 770        int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
 771        struct mm_struct *mm;
 772        void *private;
 773};
 774
 775int walk_page_range(unsigned long addr, unsigned long end,
 776                struct mm_walk *walk);
 777void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
 778                unsigned long end, unsigned long floor, unsigned long ceiling);
 779int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
 780                        struct vm_area_struct *vma);
 781void unmap_mapping_range(struct address_space *mapping,
 782                loff_t const holebegin, loff_t const holelen, int even_cows);
 783int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
 784                        void *buf, int len, int write);
 785
 786static inline void unmap_shared_mapping_range(struct address_space *mapping,
 787                loff_t const holebegin, loff_t const holelen)
 788{
 789        unmap_mapping_range(mapping, holebegin, holelen, 0);
 790}
 791
 792extern int vmtruncate(struct inode * inode, loff_t offset);
 793extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
 794
 795#ifdef CONFIG_MMU
 796extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 797                        unsigned long address, int write_access);
 798#else
 799static inline int handle_mm_fault(struct mm_struct *mm,
 800                        struct vm_area_struct *vma, unsigned long address,
 801                        int write_access)
 802{
 803        /* should never happen if there's no MMU */
 804        BUG();
 805        return VM_FAULT_SIGBUS;
 806}
 807#endif
 808
 809extern int make_pages_present(unsigned long addr, unsigned long end);
 810extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
 811
 812int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
 813                int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
 814
 815extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
 816extern void do_invalidatepage(struct page *page, unsigned long offset);
 817
 818int __set_page_dirty_nobuffers(struct page *page);
 819int __set_page_dirty_no_writeback(struct page *page);
 820int redirty_page_for_writepage(struct writeback_control *wbc,
 821                                struct page *page);
 822int set_page_dirty(struct page *page);
 823int set_page_dirty_lock(struct page *page);
 824int clear_page_dirty_for_io(struct page *page);
 825
 826extern unsigned long move_page_tables(struct vm_area_struct *vma,
 827                unsigned long old_addr, struct vm_area_struct *new_vma,
 828                unsigned long new_addr, unsigned long len);
 829extern unsigned long do_mremap(unsigned long addr,
 830                               unsigned long old_len, unsigned long new_len,
 831                               unsigned long flags, unsigned long new_addr);
 832extern int mprotect_fixup(struct vm_area_struct *vma,
 833                          struct vm_area_struct **pprev, unsigned long start,
 834                          unsigned long end, unsigned long newflags);
 835
 836/*
 837 * get_user_pages_fast provides equivalent functionality to get_user_pages,
 838 * operating on current and current->mm (force=0 and doesn't return any vmas).
 839 *
 840 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
 841 * can be made about locking. get_user_pages_fast is to be implemented in a
 842 * way that is advantageous (vs get_user_pages()) when the user memory area is
 843 * already faulted in and present in ptes. However if the pages have to be
 844 * faulted in, it may turn out to be slightly slower).
 845 */
 846int get_user_pages_fast(unsigned long start, int nr_pages, int write,
 847                        struct page **pages);
 848
 849/*
 850 * A callback you can register to apply pressure to ageable caches.
 851 *
 852 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
 853 * look through the least-recently-used 'nr_to_scan' entries and
 854 * attempt to free them up.  It should return the number of objects
 855 * which remain in the cache.  If it returns -1, it means it cannot do
 856 * any scanning at this time (eg. there is a risk of deadlock).
 857 *
 858 * The 'gfpmask' refers to the allocation we are currently trying to
 859 * fulfil.
 860 *
 861 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
 862 * querying the cache size, so a fastpath for that case is appropriate.
 863 */
 864struct shrinker {
 865        int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
 866        int seeks;      /* seeks to recreate an obj */
 867
 868        /* These are for internal use */
 869        struct list_head list;
 870        long nr;        /* objs pending delete */
 871};
 872#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
 873extern void register_shrinker(struct shrinker *);
 874extern void unregister_shrinker(struct shrinker *);
 875
 876int vma_wants_writenotify(struct vm_area_struct *vma);
 877
 878extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
 879
 880#ifdef __PAGETABLE_PUD_FOLDED
 881static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
 882                                                unsigned long address)
 883{
 884        return 0;
 885}
 886#else
 887int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
 888#endif
 889
 890#ifdef __PAGETABLE_PMD_FOLDED
 891static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
 892                                                unsigned long address)
 893{
 894        return 0;
 895}
 896#else
 897int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
 898#endif
 899
 900int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
 901int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
 902
 903/*
 904 * The following ifdef needed to get the 4level-fixup.h header to work.
 905 * Remove it when 4level-fixup.h has been removed.
 906 */
 907#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
 908static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
 909{
 910        return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
 911                NULL: pud_offset(pgd, address);
 912}
 913
 914static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 915{
 916        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
 917                NULL: pmd_offset(pud, address);
 918}
 919#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
 920
 921#if USE_SPLIT_PTLOCKS
 922/*
 923 * We tuck a spinlock to guard each pagetable page into its struct page,
 924 * at page->private, with BUILD_BUG_ON to make sure that this will not
 925 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
 926 * When freeing, reset page->mapping so free_pages_check won't complain.
 927 */
 928#define __pte_lockptr(page)     &((page)->ptl)
 929#define pte_lock_init(_page)    do {                                    \
 930        spin_lock_init(__pte_lockptr(_page));                           \
 931} while (0)
 932#define pte_lock_deinit(page)   ((page)->mapping = NULL)
 933#define pte_lockptr(mm, pmd)    ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
 934#else   /* !USE_SPLIT_PTLOCKS */
 935/*
 936 * We use mm->page_table_lock to guard all pagetable pages of the mm.
 937 */
 938#define pte_lock_init(page)     do {} while (0)
 939#define pte_lock_deinit(page)   do {} while (0)
 940#define pte_lockptr(mm, pmd)    ({(void)(pmd); &(mm)->page_table_lock;})
 941#endif /* USE_SPLIT_PTLOCKS */
 942
 943static inline void pgtable_page_ctor(struct page *page)
 944{
 945        pte_lock_init(page);
 946        inc_zone_page_state(page, NR_PAGETABLE);
 947}
 948
 949static inline void pgtable_page_dtor(struct page *page)
 950{
 951        pte_lock_deinit(page);
 952        dec_zone_page_state(page, NR_PAGETABLE);
 953}
 954
 955#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
 956({                                                      \
 957        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
 958        pte_t *__pte = pte_offset_map(pmd, address);    \
 959        *(ptlp) = __ptl;                                \
 960        spin_lock(__ptl);                               \
 961        __pte;                                          \
 962})
 963
 964#define pte_unmap_unlock(pte, ptl)      do {            \
 965        spin_unlock(ptl);                               \
 966        pte_unmap(pte);                                 \
 967} while (0)
 968
 969#define pte_alloc_map(mm, pmd, address)                 \
 970        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
 971                NULL: pte_offset_map(pmd, address))
 972
 973#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
 974        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
 975                NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
 976
 977#define pte_alloc_kernel(pmd, address)                  \
 978        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
 979                NULL: pte_offset_kernel(pmd, address))
 980
 981extern void free_area_init(unsigned long * zones_size);
 982extern void free_area_init_node(int nid, unsigned long * zones_size,
 983                unsigned long zone_start_pfn, unsigned long *zholes_size);
 984#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
 985/*
 986 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
 987 * zones, allocate the backing mem_map and account for memory holes in a more
 988 * architecture independent manner. This is a substitute for creating the
 989 * zone_sizes[] and zholes_size[] arrays and passing them to
 990 * free_area_init_node()
 991 *
 992 * An architecture is expected to register range of page frames backed by
 993 * physical memory with add_active_range() before calling
 994 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
 995 * usage, an architecture is expected to do something like
 996 *
 997 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
 998 *                                                       max_highmem_pfn};
 999 * for_each_valid_physical_page_range()
1000 *      add_active_range(node_id, start_pfn, end_pfn)
1001 * free_area_init_nodes(max_zone_pfns);
1002 *
1003 * If the architecture guarantees that there are no holes in the ranges
1004 * registered with add_active_range(), free_bootmem_active_regions()
1005 * will call free_bootmem_node() for each registered physical page range.
1006 * Similarly sparse_memory_present_with_active_regions() calls
1007 * memory_present() for each range when SPARSEMEM is enabled.
1008 *
1009 * See mm/page_alloc.c for more information on each function exposed by
1010 * CONFIG_ARCH_POPULATES_NODE_MAP
1011 */
1012extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1013extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1014                                        unsigned long end_pfn);
1015extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1016                                        unsigned long end_pfn);
1017extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1018                                        unsigned long end_pfn);
1019extern void remove_all_active_ranges(void);
1020extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1021                                                unsigned long end_pfn);
1022extern void get_pfn_range_for_nid(unsigned int nid,
1023                        unsigned long *start_pfn, unsigned long *end_pfn);
1024extern unsigned long find_min_pfn_with_active_regions(void);
1025extern void free_bootmem_with_active_regions(int nid,
1026                                                unsigned long max_low_pfn);
1027typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1028extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1029extern void sparse_memory_present_with_active_regions(int nid);
1030#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1031
1032#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1033    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1034static inline int __early_pfn_to_nid(unsigned long pfn)
1035{
1036        return 0;
1037}
1038#else
1039/* please see mm/page_alloc.c */
1040extern int __meminit early_pfn_to_nid(unsigned long pfn);
1041#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1042/* there is a per-arch backend function. */
1043extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1044#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1045#endif
1046
1047extern void set_dma_reserve(unsigned long new_dma_reserve);
1048extern void memmap_init_zone(unsigned long, int, unsigned long,
1049                                unsigned long, enum memmap_context);
1050extern void setup_per_zone_pages_min(void);
1051extern void mem_init(void);
1052extern void show_mem(void);
1053extern void si_meminfo(struct sysinfo * val);
1054extern void si_meminfo_node(struct sysinfo *val, int nid);
1055extern int after_bootmem;
1056
1057#ifdef CONFIG_NUMA
1058extern void setup_per_cpu_pageset(void);
1059#else
1060static inline void setup_per_cpu_pageset(void) {}
1061#endif
1062
1063/* prio_tree.c */
1064void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1065void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1066void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1067struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1068        struct prio_tree_iter *iter);
1069
1070#define vma_prio_tree_foreach(vma, iter, root, begin, end)      \
1071        for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;   \
1072                (vma = vma_prio_tree_next(vma, iter)); )
1073
1074static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1075                                        struct list_head *list)
1076{
1077        vma->shared.vm_set.parent = NULL;
1078        list_add_tail(&vma->shared.vm_set.list, list);
1079}
1080
1081/* mmap.c */
1082extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1083extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1084        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1085extern struct vm_area_struct *vma_merge(struct mm_struct *,
1086        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1087        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1088        struct mempolicy *);
1089extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1090extern int split_vma(struct mm_struct *,
1091        struct vm_area_struct *, unsigned long addr, int new_below);
1092extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1093extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1094        struct rb_node **, struct rb_node *);
1095extern void unlink_file_vma(struct vm_area_struct *);
1096extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1097        unsigned long addr, unsigned long len, pgoff_t pgoff);
1098extern void exit_mmap(struct mm_struct *);
1099
1100extern int mm_take_all_locks(struct mm_struct *mm);
1101extern void mm_drop_all_locks(struct mm_struct *mm);
1102
1103#ifdef CONFIG_PROC_FS
1104/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1105extern void added_exe_file_vma(struct mm_struct *mm);
1106extern void removed_exe_file_vma(struct mm_struct *mm);
1107#else
1108static inline void added_exe_file_vma(struct mm_struct *mm)
1109{}
1110
1111static inline void removed_exe_file_vma(struct mm_struct *mm)
1112{}
1113#endif /* CONFIG_PROC_FS */
1114
1115extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1116extern int install_special_mapping(struct mm_struct *mm,
1117                                   unsigned long addr, unsigned long len,
1118                                   unsigned long flags, struct page **pages);
1119
1120extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1121
1122extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1123        unsigned long len, unsigned long prot,
1124        unsigned long flag, unsigned long pgoff);
1125extern unsigned long mmap_region(struct file *file, unsigned long addr,
1126        unsigned long len, unsigned long flags,
1127        unsigned int vm_flags, unsigned long pgoff,
1128        int accountable);
1129
1130static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1131        unsigned long len, unsigned long prot,
1132        unsigned long flag, unsigned long offset)
1133{
1134        unsigned long ret = -EINVAL;
1135        if ((offset + PAGE_ALIGN(len)) < offset)
1136                goto out;
1137        if (!(offset & ~PAGE_MASK))
1138                ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1139out:
1140        return ret;
1141}
1142
1143extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1144
1145extern unsigned long do_brk(unsigned long, unsigned long);
1146
1147/* filemap.c */
1148extern unsigned long page_unuse(struct page *);
1149extern void truncate_inode_pages(struct address_space *, loff_t);
1150extern void truncate_inode_pages_range(struct address_space *,
1151                                       loff_t lstart, loff_t lend);
1152
1153/* generic vm_area_ops exported for stackable file systems */
1154extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1155
1156/* mm/page-writeback.c */
1157int write_one_page(struct page *page, int wait);
1158
1159/* readahead.c */
1160#define VM_MAX_READAHEAD        128     /* kbytes */
1161#define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1162
1163int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1164                        pgoff_t offset, unsigned long nr_to_read);
1165int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1166                        pgoff_t offset, unsigned long nr_to_read);
1167
1168void page_cache_sync_readahead(struct address_space *mapping,
1169                               struct file_ra_state *ra,
1170                               struct file *filp,
1171                               pgoff_t offset,
1172                               unsigned long size);
1173
1174void page_cache_async_readahead(struct address_space *mapping,
1175                                struct file_ra_state *ra,
1176                                struct file *filp,
1177                                struct page *pg,
1178                                pgoff_t offset,
1179                                unsigned long size);
1180
1181unsigned long max_sane_readahead(unsigned long nr);
1182
1183/* Do stack extension */
1184extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1185#ifdef CONFIG_IA64
1186extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1187#endif
1188extern int expand_stack_downwards(struct vm_area_struct *vma,
1189                                  unsigned long address);
1190
1191/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1192extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1193extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1194                                             struct vm_area_struct **pprev);
1195
1196/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1197   NULL if none.  Assume start_addr < end_addr. */
1198static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1199{
1200        struct vm_area_struct * vma = find_vma(mm,start_addr);
1201
1202        if (vma && end_addr <= vma->vm_start)
1203                vma = NULL;
1204        return vma;
1205}
1206
1207static inline unsigned long vma_pages(struct vm_area_struct *vma)
1208{
1209        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1210}
1211
1212pgprot_t vm_get_page_prot(unsigned long vm_flags);
1213struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1214int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1215                        unsigned long pfn, unsigned long size, pgprot_t);
1216int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1217int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1218                        unsigned long pfn);
1219int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1220                        unsigned long pfn);
1221
1222struct page *follow_page(struct vm_area_struct *, unsigned long address,
1223                        unsigned int foll_flags);
1224#define FOLL_WRITE      0x01    /* check pte is writable */
1225#define FOLL_TOUCH      0x02    /* mark page accessed */
1226#define FOLL_GET        0x04    /* do get_page on page */
1227#define FOLL_ANON       0x08    /* give ZERO_PAGE if no pgtable */
1228
1229typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1230                        void *data);
1231extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1232                               unsigned long size, pte_fn_t fn, void *data);
1233
1234#ifdef CONFIG_PROC_FS
1235void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1236#else
1237static inline void vm_stat_account(struct mm_struct *mm,
1238                        unsigned long flags, struct file *file, long pages)
1239{
1240}
1241#endif /* CONFIG_PROC_FS */
1242
1243#ifdef CONFIG_DEBUG_PAGEALLOC
1244extern int debug_pagealloc_enabled;
1245
1246extern void kernel_map_pages(struct page *page, int numpages, int enable);
1247
1248static inline void enable_debug_pagealloc(void)
1249{
1250        debug_pagealloc_enabled = 1;
1251}
1252#ifdef CONFIG_HIBERNATION
1253extern bool kernel_page_present(struct page *page);
1254#endif /* CONFIG_HIBERNATION */
1255#else
1256static inline void
1257kernel_map_pages(struct page *page, int numpages, int enable) {}
1258static inline void enable_debug_pagealloc(void)
1259{
1260}
1261#ifdef CONFIG_HIBERNATION
1262static inline bool kernel_page_present(struct page *page) { return true; }
1263#endif /* CONFIG_HIBERNATION */
1264#endif
1265
1266extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1267#ifdef  __HAVE_ARCH_GATE_AREA
1268int in_gate_area_no_task(unsigned long addr);
1269int in_gate_area(struct task_struct *task, unsigned long addr);
1270#else
1271int in_gate_area_no_task(unsigned long addr);
1272#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1273#endif  /* __HAVE_ARCH_GATE_AREA */
1274
1275int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1276                                        void __user *, size_t *, loff_t *);
1277unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1278                        unsigned long lru_pages);
1279
1280#ifndef CONFIG_MMU
1281#define randomize_va_space 0
1282#else
1283extern int randomize_va_space;
1284#endif
1285
1286const char * arch_vma_name(struct vm_area_struct *vma);
1287void print_vma_addr(char *prefix, unsigned long rip);
1288
1289struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1290pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1291pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1292pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1293pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1294void *vmemmap_alloc_block(unsigned long size, int node);
1295void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1296int vmemmap_populate_basepages(struct page *start_page,
1297                                                unsigned long pages, int node);
1298int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1299void vmemmap_populate_print_last(void);
1300
1301#endif /* __KERNEL__ */
1302#endif /* _LINUX_MM_H */
1303
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.