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/bug.h>
  10#include <linux/list.h>
  11#include <linux/mmzone.h>
  12#include <linux/rbtree.h>
  13#include <linux/prio_tree.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/bit_spinlock.h>
  20#include <linux/shrinker.h>
  21
  22struct mempolicy;
  23struct anon_vma;
  24struct file_ra_state;
  25struct user_struct;
  26struct writeback_control;
  27
  28#ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
  29extern unsigned long max_mapnr;
  30#endif
  31
  32extern unsigned long num_physpages;
  33extern unsigned long totalram_pages;
  34extern void * high_memory;
  35extern int page_cluster;
  36
  37#ifdef CONFIG_SYSCTL
  38extern int sysctl_legacy_va_layout;
  39#else
  40#define sysctl_legacy_va_layout 0
  41#endif
  42
  43#include <asm/page.h>
  44#include <asm/pgtable.h>
  45#include <asm/processor.h>
  46
  47#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  48
  49/* to align the pointer to the (next) page boundary */
  50#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
  51
  52/*
  53 * Linux kernel virtual memory manager primitives.
  54 * The idea being to have a "virtual" mm in the same way
  55 * we have a virtual fs - giving a cleaner interface to the
  56 * mm details, and allowing different kinds of memory mappings
  57 * (from shared memory to executable loading to arbitrary
  58 * mmap() functions).
  59 */
  60
  61extern struct kmem_cache *vm_area_cachep;
  62
  63#ifndef CONFIG_MMU
  64extern struct rb_root nommu_region_tree;
  65extern struct rw_semaphore nommu_region_sem;
  66
  67extern unsigned int kobjsize(const void *objp);
  68#endif
  69
  70/*
  71 * vm_flags in vm_area_struct, see mm_types.h.
  72 */
  73#define VM_READ         0x00000001      /* currently active flags */
  74#define VM_WRITE        0x00000002
  75#define VM_EXEC         0x00000004
  76#define VM_SHARED       0x00000008
  77
  78/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  79#define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
  80#define VM_MAYWRITE     0x00000020
  81#define VM_MAYEXEC      0x00000040
  82#define VM_MAYSHARE     0x00000080
  83
  84#define VM_GROWSDOWN    0x00000100      /* general info on the segment */
  85#if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
  86#define VM_GROWSUP      0x00000200
  87#else
  88#define VM_GROWSUP      0x00000000
  89#define VM_NOHUGEPAGE   0x00000200      /* MADV_NOHUGEPAGE marked this vma */
  90#endif
  91#define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
  92#define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
  93
  94#define VM_EXECUTABLE   0x00001000
  95#define VM_LOCKED       0x00002000
  96#define VM_IO           0x00004000      /* Memory mapped I/O or similar */
  97
  98                                        /* Used by sys_madvise() */
  99#define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
 100#define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
 101
 102#define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
 103#define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
 104#define VM_RESERVED     0x00080000      /* Count as reserved_vm like IO */
 105#define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
 106#define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
 107#define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
 108#define VM_NONLINEAR    0x00800000      /* Is non-linear (remap_file_pages) */
 109#ifndef CONFIG_TRANSPARENT_HUGEPAGE
 110#define VM_MAPPED_COPY  0x01000000      /* T if mapped copy of data (nommu mmap) */
 111#else
 112#define VM_HUGEPAGE     0x01000000      /* MADV_HUGEPAGE marked this vma */
 113#endif
 114#define VM_INSERTPAGE   0x02000000      /* The vma has had "vm_insert_page()" done on it */
 115#define VM_NODUMP       0x04000000      /* Do not include in the core dump */
 116
 117#define VM_CAN_NONLINEAR 0x08000000     /* Has ->fault & does nonlinear pages */
 118#define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
 119#define VM_SAO          0x20000000      /* Strong Access Ordering (powerpc) */
 120#define VM_PFN_AT_MMAP  0x40000000      /* PFNMAP vma that is fully mapped at mmap time */
 121#define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
 122
 123/* Bits set in the VMA until the stack is in its final location */
 124#define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
 125
 126#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 127#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 128#endif
 129
 130#ifdef CONFIG_STACK_GROWSUP
 131#define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 132#else
 133#define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 134#endif
 135
 136#define VM_READHINTMASK                 (VM_SEQ_READ | VM_RAND_READ)
 137#define VM_ClearReadHint(v)             (v)->vm_flags &= ~VM_READHINTMASK
 138#define VM_NormalReadHint(v)            (!((v)->vm_flags & VM_READHINTMASK))
 139#define VM_SequentialReadHint(v)        ((v)->vm_flags & VM_SEQ_READ)
 140#define VM_RandomReadHint(v)            ((v)->vm_flags & VM_RAND_READ)
 141
 142/*
 143 * Special vmas that are non-mergable, non-mlock()able.
 144 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
 145 */
 146#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
 147
 148/*
 149 * mapping from the currently active vm_flags protection bits (the
 150 * low four bits) to a page protection mask..
 151 */
 152extern pgprot_t protection_map[16];
 153
 154#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 155#define FAULT_FLAG_NONLINEAR    0x02    /* Fault was via a nonlinear mapping */
 156#define FAULT_FLAG_MKWRITE      0x04    /* Fault was mkwrite of existing pte */
 157#define FAULT_FLAG_ALLOW_RETRY  0x08    /* Retry fault if blocking */
 158#define FAULT_FLAG_RETRY_NOWAIT 0x10    /* Don't drop mmap_sem and wait when retrying */
 159#define FAULT_FLAG_KILLABLE     0x20    /* The fault task is in SIGKILL killable region */
 160
 161/*
 162 * This interface is used by x86 PAT code to identify a pfn mapping that is
 163 * linear over entire vma. This is to optimize PAT code that deals with
 164 * marking the physical region with a particular prot. This is not for generic
 165 * mm use. Note also that this check will not work if the pfn mapping is
 166 * linear for a vma starting at physical address 0. In which case PAT code
 167 * falls back to slow path of reserving physical range page by page.
 168 */
 169static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
 170{
 171        return !!(vma->vm_flags & VM_PFN_AT_MMAP);
 172}
 173
 174static inline int is_pfn_mapping(struct vm_area_struct *vma)
 175{
 176        return !!(vma->vm_flags & VM_PFNMAP);
 177}
 178
 179/*
 180 * vm_fault is filled by the the pagefault handler and passed to the vma's
 181 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 182 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 183 *
 184 * pgoff should be used in favour of virtual_address, if possible. If pgoff
 185 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
 186 * mapping support.
 187 */
 188struct vm_fault {
 189        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 190        pgoff_t pgoff;                  /* Logical page offset based on vma */
 191        void __user *virtual_address;   /* Faulting virtual address */
 192
 193        struct page *page;              /* ->fault handlers should return a
 194                                         * page here, unless VM_FAULT_NOPAGE
 195                                         * is set (which is also implied by
 196                                         * VM_FAULT_ERROR).
 197                                         */
 198};
 199
 200/*
 201 * These are the virtual MM functions - opening of an area, closing and
 202 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 203 * to the functions called when a no-page or a wp-page exception occurs. 
 204 */
 205struct vm_operations_struct {
 206        void (*open)(struct vm_area_struct * area);
 207        void (*close)(struct vm_area_struct * area);
 208        int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 209
 210        /* notification that a previously read-only page is about to become
 211         * writable, if an error is returned it will cause a SIGBUS */
 212        int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 213
 214        /* called by access_process_vm when get_user_pages() fails, typically
 215         * for use by special VMAs that can switch between memory and hardware
 216         */
 217        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 218                      void *buf, int len, int write);
 219#ifdef CONFIG_NUMA
 220        /*
 221         * set_policy() op must add a reference to any non-NULL @new mempolicy
 222         * to hold the policy upon return.  Caller should pass NULL @new to
 223         * remove a policy and fall back to surrounding context--i.e. do not
 224         * install a MPOL_DEFAULT policy, nor the task or system default
 225         * mempolicy.
 226         */
 227        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 228
 229        /*
 230         * get_policy() op must add reference [mpol_get()] to any policy at
 231         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 232         * in mm/mempolicy.c will do this automatically.
 233         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 234         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 235         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 236         * must return NULL--i.e., do not "fallback" to task or system default
 237         * policy.
 238         */
 239        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 240                                        unsigned long addr);
 241        int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 242                const nodemask_t *to, unsigned long flags);
 243#endif
 244};
 245
 246struct mmu_gather;
 247struct inode;
 248
 249#define page_private(page)              ((page)->private)
 250#define set_page_private(page, v)       ((page)->private = (v))
 251
 252/*
 253 * FIXME: take this include out, include page-flags.h in
 254 * files which need it (119 of them)
 255 */
 256#include <linux/page-flags.h>
 257#include <linux/huge_mm.h>
 258
 259/*
 260 * Methods to modify the page usage count.
 261 *
 262 * What counts for a page usage:
 263 * - cache mapping   (page->mapping)
 264 * - private data    (page->private)
 265 * - page mapped in a task's page tables, each mapping
 266 *   is counted separately
 267 *
 268 * Also, many kernel routines increase the page count before a critical
 269 * routine so they can be sure the page doesn't go away from under them.
 270 */
 271
 272/*
 273 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 274 */
 275static inline int put_page_testzero(struct page *page)
 276{
 277        VM_BUG_ON(atomic_read(&page->_count) == 0);
 278        return atomic_dec_and_test(&page->_count);
 279}
 280
 281/*
 282 * Try to grab a ref unless the page has a refcount of zero, return false if
 283 * that is the case.
 284 */
 285static inline int get_page_unless_zero(struct page *page)
 286{
 287        return atomic_inc_not_zero(&page->_count);
 288}
 289
 290extern int page_is_ram(unsigned long pfn);
 291
 292/* Support for virtually mapped pages */
 293struct page *vmalloc_to_page(const void *addr);
 294unsigned long vmalloc_to_pfn(const void *addr);
 295
 296/*
 297 * Determine if an address is within the vmalloc range
 298 *
 299 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 300 * is no special casing required.
 301 */
 302static inline int is_vmalloc_addr(const void *x)
 303{
 304#ifdef CONFIG_MMU
 305        unsigned long addr = (unsigned long)x;
 306
 307        return addr >= VMALLOC_START && addr < VMALLOC_END;
 308#else
 309        return 0;
 310#endif
 311}
 312#ifdef CONFIG_MMU
 313extern int is_vmalloc_or_module_addr(const void *x);
 314#else
 315static inline int is_vmalloc_or_module_addr(const void *x)
 316{
 317        return 0;
 318}
 319#endif
 320
 321static inline void compound_lock(struct page *page)
 322{
 323#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 324        VM_BUG_ON(PageSlab(page));
 325        bit_spin_lock(PG_compound_lock, &page->flags);
 326#endif
 327}
 328
 329static inline void compound_unlock(struct page *page)
 330{
 331#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 332        VM_BUG_ON(PageSlab(page));
 333        bit_spin_unlock(PG_compound_lock, &page->flags);
 334#endif
 335}
 336
 337static inline unsigned long compound_lock_irqsave(struct page *page)
 338{
 339        unsigned long uninitialized_var(flags);
 340#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 341        local_irq_save(flags);
 342        compound_lock(page);
 343#endif
 344        return flags;
 345}
 346
 347static inline void compound_unlock_irqrestore(struct page *page,
 348                                              unsigned long flags)
 349{
 350#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 351        compound_unlock(page);
 352        local_irq_restore(flags);
 353#endif
 354}
 355
 356static inline struct page *compound_head(struct page *page)
 357{
 358        if (unlikely(PageTail(page)))
 359                return page->first_page;
 360        return page;
 361}
 362
 363/*
 364 * The atomic page->_mapcount, starts from -1: so that transitions
 365 * both from it and to it can be tracked, using atomic_inc_and_test
 366 * and atomic_add_negative(-1).
 367 */
 368static inline void reset_page_mapcount(struct page *page)
 369{
 370        atomic_set(&(page)->_mapcount, -1);
 371}
 372
 373static inline int page_mapcount(struct page *page)
 374{
 375        return atomic_read(&(page)->_mapcount) + 1;
 376}
 377
 378static inline int page_count(struct page *page)
 379{
 380        return atomic_read(&compound_head(page)->_count);
 381}
 382
 383static inline void get_huge_page_tail(struct page *page)
 384{
 385        /*
 386         * __split_huge_page_refcount() cannot run
 387         * from under us.
 388         */
 389        VM_BUG_ON(page_mapcount(page) < 0);
 390        VM_BUG_ON(atomic_read(&page->_count) != 0);
 391        atomic_inc(&page->_mapcount);
 392}
 393
 394extern bool __get_page_tail(struct page *page);
 395
 396static inline void get_page(struct page *page)
 397{
 398        if (unlikely(PageTail(page)))
 399                if (likely(__get_page_tail(page)))
 400                        return;
 401        /*
 402         * Getting a normal page or the head of a compound page
 403         * requires to already have an elevated page->_count.
 404         */
 405        VM_BUG_ON(atomic_read(&page->_count) <= 0);
 406        atomic_inc(&page->_count);
 407}
 408
 409static inline struct page *virt_to_head_page(const void *x)
 410{
 411        struct page *page = virt_to_page(x);
 412        return compound_head(page);
 413}
 414
 415/*
 416 * Setup the page count before being freed into the page allocator for
 417 * the first time (boot or memory hotplug)
 418 */
 419static inline void init_page_count(struct page *page)
 420{
 421        atomic_set(&page->_count, 1);
 422}
 423
 424/*
 425 * PageBuddy() indicate that the page is free and in the buddy system
 426 * (see mm/page_alloc.c).
 427 *
 428 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
 429 * -2 so that an underflow of the page_mapcount() won't be mistaken
 430 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
 431 * efficiently by most CPU architectures.
 432 */
 433#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
 434
 435static inline int PageBuddy(struct page *page)
 436{
 437        return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
 438}
 439
 440static inline void __SetPageBuddy(struct page *page)
 441{
 442        VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
 443        atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
 444}
 445
 446static inline void __ClearPageBuddy(struct page *page)
 447{
 448        VM_BUG_ON(!PageBuddy(page));
 449        atomic_set(&page->_mapcount, -1);
 450}
 451
 452void put_page(struct page *page);
 453void put_pages_list(struct list_head *pages);
 454
 455void split_page(struct page *page, unsigned int order);
 456int split_free_page(struct page *page);
 457
 458/*
 459 * Compound pages have a destructor function.  Provide a
 460 * prototype for that function and accessor functions.
 461 * These are _only_ valid on the head of a PG_compound page.
 462 */
 463typedef void compound_page_dtor(struct page *);
 464
 465static inline void set_compound_page_dtor(struct page *page,
 466                                                compound_page_dtor *dtor)
 467{
 468        page[1].lru.next = (void *)dtor;
 469}
 470
 471static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 472{
 473        return (compound_page_dtor *)page[1].lru.next;
 474}
 475
 476static inline int compound_order(struct page *page)
 477{
 478        if (!PageHead(page))
 479                return 0;
 480        return (unsigned long)page[1].lru.prev;
 481}
 482
 483static inline int compound_trans_order(struct page *page)
 484{
 485        int order;
 486        unsigned long flags;
 487
 488        if (!PageHead(page))
 489                return 0;
 490
 491        flags = compound_lock_irqsave(page);
 492        order = compound_order(page);
 493        compound_unlock_irqrestore(page, flags);
 494        return order;
 495}
 496
 497static inline void set_compound_order(struct page *page, unsigned long order)
 498{
 499        page[1].lru.prev = (void *)order;
 500}
 501
 502#ifdef CONFIG_MMU
 503/*
 504 * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
 505 * servicing faults for write access.  In the normal case, do always want
 506 * pte_mkwrite.  But get_user_pages can cause write faults for mappings
 507 * that do not have writing enabled, when used by access_process_vm.
 508 */
 509static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 510{
 511        if (likely(vma->vm_flags & VM_WRITE))
 512                pte = pte_mkwrite(pte);
 513        return pte;
 514}
 515#endif
 516
 517/*
 518 * Multiple processes may "see" the same page. E.g. for untouched
 519 * mappings of /dev/null, all processes see the same page full of
 520 * zeroes, and text pages of executables and shared libraries have
 521 * only one copy in memory, at most, normally.
 522 *
 523 * For the non-reserved pages, page_count(page) denotes a reference count.
 524 *   page_count() == 0 means the page is free. page->lru is then used for
 525 *   freelist management in the buddy allocator.
 526 *   page_count() > 0  means the page has been allocated.
 527 *
 528 * Pages are allocated by the slab allocator in order to provide memory
 529 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 530 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 531 * unless a particular usage is carefully commented. (the responsibility of
 532 * freeing the kmalloc memory is the caller's, of course).
 533 *
 534 * A page may be used by anyone else who does a __get_free_page().
 535 * In this case, page_count still tracks the references, and should only
 536 * be used through the normal accessor functions. The top bits of page->flags
 537 * and page->virtual store page management information, but all other fields
 538 * are unused and could be used privately, carefully. The management of this
 539 * page is the responsibility of the one who allocated it, and those who have
 540 * subsequently been given references to it.
 541 *
 542 * The other pages (we may call them "pagecache pages") are completely
 543 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 544 * The following discussion applies only to them.
 545 *
 546 * A pagecache page contains an opaque `private' member, which belongs to the
 547 * page's address_space. Usually, this is the address of a circular list of
 548 * the page's disk buffers. PG_private must be set to tell the VM to call
 549 * into the filesystem to release these pages.
 550 *
 551 * A page may belong to an inode's memory mapping. In this case, page->mapping
 552 * is the pointer to the inode, and page->index is the file offset of the page,
 553 * in units of PAGE_CACHE_SIZE.
 554 *
 555 * If pagecache pages are not associated with an inode, they are said to be
 556 * anonymous pages. These may become associated with the swapcache, and in that
 557 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 558 *
 559 * In either case (swapcache or inode backed), the pagecache itself holds one
 560 * reference to the page. Setting PG_private should also increment the
 561 * refcount. The each user mapping also has a reference to the page.
 562 *
 563 * The pagecache pages are stored in a per-mapping radix tree, which is
 564 * rooted at mapping->page_tree, and indexed by offset.
 565 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 566 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 567 *
 568 * All pagecache pages may be subject to I/O:
 569 * - inode pages may need to be read from disk,
 570 * - inode pages which have been modified and are MAP_SHARED may need
 571 *   to be written back to the inode on disk,
 572 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 573 *   modified may need to be swapped out to swap space and (later) to be read
 574 *   back into memory.
 575 */
 576
 577/*
 578 * The zone field is never updated after free_area_init_core()
 579 * sets it, so none of the operations on it need to be atomic.
 580 */
 581
 582
 583/*
 584 * page->flags layout:
 585 *
 586 * There are three possibilities for how page->flags get
 587 * laid out.  The first is for the normal case, without
 588 * sparsemem.  The second is for sparsemem when there is
 589 * plenty of space for node and section.  The last is when
 590 * we have run out of space and have to fall back to an
 591 * alternate (slower) way of determining the node.
 592 *
 593 * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
 594 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
 595 * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
 596 */
 597#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 598#define SECTIONS_WIDTH          SECTIONS_SHIFT
 599#else
 600#define SECTIONS_WIDTH          0
 601#endif
 602
 603#define ZONES_WIDTH             ZONES_SHIFT
 604
 605#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
 606#define NODES_WIDTH             NODES_SHIFT
 607#else
 608#ifdef CONFIG_SPARSEMEM_VMEMMAP
 609#error "Vmemmap: No space for nodes field in page flags"
 610#endif
 611#define NODES_WIDTH             0
 612#endif
 613
 614/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
 615#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 616#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 617#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 618
 619/*
 620 * We are going to use the flags for the page to node mapping if its in
 621 * there.  This includes the case where there is no node, so it is implicit.
 622 */
 623#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
 624#define NODE_NOT_IN_PAGE_FLAGS
 625#endif
 626
 627/*
 628 * Define the bit shifts to access each section.  For non-existent
 629 * sections we define the shift as 0; that plus a 0 mask ensures
 630 * the compiler will optimise away reference to them.
 631 */
 632#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 633#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 634#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 635
 636/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
 637#ifdef NODE_NOT_IN_PAGE_FLAGS
 638#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 639#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 640                                                SECTIONS_PGOFF : ZONES_PGOFF)
 641#else
 642#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 643#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 644                                                NODES_PGOFF : ZONES_PGOFF)
 645#endif
 646
 647#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 648
 649#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 650#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 651#endif
 652
 653#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 654#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 655#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 656#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 657
 658static inline enum zone_type page_zonenum(const struct page *page)
 659{
 660        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 661}
 662
 663/*
 664 * The identification function is only used by the buddy allocator for
 665 * determining if two pages could be buddies. We are not really
 666 * identifying a zone since we could be using a the section number
 667 * id if we have not node id available in page flags.
 668 * We guarantee only that it will return the same value for two
 669 * combinable pages in a zone.
 670 */
 671static inline int page_zone_id(struct page *page)
 672{
 673        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 674}
 675
 676static inline int zone_to_nid(struct zone *zone)
 677{
 678#ifdef CONFIG_NUMA
 679        return zone->node;
 680#else
 681        return 0;
 682#endif
 683}
 684
 685#ifdef NODE_NOT_IN_PAGE_FLAGS
 686extern int page_to_nid(const struct page *page);
 687#else
 688static inline int page_to_nid(const struct page *page)
 689{
 690        return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 691}
 692#endif
 693
 694static inline struct zone *page_zone(const struct page *page)
 695{
 696        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 697}
 698
 699#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 700static inline void set_page_section(struct page *page, unsigned long section)
 701{
 702        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 703        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 704}
 705
 706static inline unsigned long page_to_section(const struct page *page)
 707{
 708        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 709}
 710#endif
 711
 712static inline void set_page_zone(struct page *page, enum zone_type zone)
 713{
 714        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 715        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 716}
 717
 718static inline void set_page_node(struct page *page, unsigned long node)
 719{
 720        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 721        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 722}
 723
 724static inline void set_page_links(struct page *page, enum zone_type zone,
 725        unsigned long node, unsigned long pfn)
 726{
 727        set_page_zone(page, zone);
 728        set_page_node(page, node);
 729#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 730        set_page_section(page, pfn_to_section_nr(pfn));
 731#endif
 732}
 733
 734/*
 735 * Some inline functions in vmstat.h depend on page_zone()
 736 */
 737#include <linux/vmstat.h>
 738
 739static __always_inline void *lowmem_page_address(const struct page *page)
 740{
 741        return __va(PFN_PHYS(page_to_pfn(page)));
 742}
 743
 744#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 745#define HASHED_PAGE_VIRTUAL
 746#endif
 747
 748#if defined(WANT_PAGE_VIRTUAL)
 749#define page_address(page) ((page)->virtual)
 750#define set_page_address(page, address)                 \
 751        do {                                            \
 752                (page)->virtual = (address);            \
 753        } while(0)
 754#define page_address_init()  do { } while(0)
 755#endif
 756
 757#if defined(HASHED_PAGE_VIRTUAL)
 758void *page_address(const struct page *page);
 759void set_page_address(struct page *page, void *virtual);
 760void page_address_init(void);
 761#endif
 762
 763#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 764#define page_address(page) lowmem_page_address(page)
 765#define set_page_address(page, address)  do { } while(0)
 766#define page_address_init()  do { } while(0)
 767#endif
 768
 769/*
 770 * On an anonymous page mapped into a user virtual memory area,
 771 * page->mapping points to its anon_vma, not to a struct address_space;
 772 * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
 773 *
 774 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
 775 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
 776 * and then page->mapping points, not to an anon_vma, but to a private
 777 * structure which KSM associates with that merged page.  See ksm.h.
 778 *
 779 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
 780 *
 781 * Please note that, confusingly, "page_mapping" refers to the inode
 782 * address_space which maps the page from disk; whereas "page_mapped"
 783 * refers to user virtual address space into which the page is mapped.
 784 */
 785#define PAGE_MAPPING_ANON       1
 786#define PAGE_MAPPING_KSM        2
 787#define PAGE_MAPPING_FLAGS      (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
 788
 789extern struct address_space swapper_space;
 790static inline struct address_space *page_mapping(struct page *page)
 791{
 792        struct address_space *mapping = page->mapping;
 793
 794        VM_BUG_ON(PageSlab(page));
 795        if (unlikely(PageSwapCache(page)))
 796                mapping = &swapper_space;
 797        else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 798                mapping = NULL;
 799        return mapping;
 800}
 801
 802/* Neutral page->mapping pointer to address_space or anon_vma or other */
 803static inline void *page_rmapping(struct page *page)
 804{
 805        return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
 806}
 807
 808extern struct address_space *__page_file_mapping(struct page *);
 809
 810static inline
 811struct address_space *page_file_mapping(struct page *page)
 812{
 813        if (unlikely(PageSwapCache(page)))
 814                return __page_file_mapping(page);
 815
 816        return page->mapping;
 817}
 818
 819static inline int PageAnon(struct page *page)
 820{
 821        return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
 822}
 823
 824/*
 825 * Return the pagecache index of the passed page.  Regular pagecache pages
 826 * use ->index whereas swapcache pages use ->private
 827 */
 828static inline pgoff_t page_index(struct page *page)
 829{
 830        if (unlikely(PageSwapCache(page)))
 831                return page_private(page);
 832        return page->index;
 833}
 834
 835extern pgoff_t __page_file_index(struct page *page);
 836
 837/*
 838 * Return the file index of the page. Regular pagecache pages use ->index
 839 * whereas swapcache pages use swp_offset(->private)
 840 */
 841static inline pgoff_t page_file_index(struct page *page)
 842{
 843        if (unlikely(PageSwapCache(page)))
 844                return __page_file_index(page);
 845
 846        return page->index;
 847}
 848
 849/*
 850 * Return true if this page is mapped into pagetables.
 851 */
 852static inline int page_mapped(struct page *page)
 853{
 854        return atomic_read(&(page)->_mapcount) >= 0;
 855}
 856
 857/*
 858 * Different kinds of faults, as returned by handle_mm_fault().
 859 * Used to decide whether a process gets delivered SIGBUS or
 860 * just gets major/minor fault counters bumped up.
 861 */
 862
 863#define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
 864
 865#define VM_FAULT_OOM    0x0001
 866#define VM_FAULT_SIGBUS 0x0002
 867#define VM_FAULT_MAJOR  0x0004
 868#define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
 869#define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
 870#define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
 871
 872#define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
 873#define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
 874#define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
 875
 876#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
 877
 878#define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
 879                         VM_FAULT_HWPOISON_LARGE)
 880
 881/* Encode hstate index for a hwpoisoned large page */
 882#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
 883#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
 884
 885/*
 886 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
 887 */
 888extern void pagefault_out_of_memory(void);
 889
 890#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
 891
 892/*
 893 * Flags passed to show_mem() and show_free_areas() to suppress output in
 894 * various contexts.
 895 */
 896#define SHOW_MEM_FILTER_NODES   (0x0001u)       /* filter disallowed nodes */
 897
 898extern void show_free_areas(unsigned int flags);
 899extern bool skip_free_areas_node(unsigned int flags, int nid);
 900
 901int shmem_zero_setup(struct vm_area_struct *);
 902
 903extern int can_do_mlock(void);
 904extern int user_shm_lock(size_t, struct user_struct *);
 905extern void user_shm_unlock(size_t, struct user_struct *);
 906
 907/*
 908 * Parameter block passed down to zap_pte_range in exceptional cases.
 909 */
 910struct zap_details {
 911        struct vm_area_struct *nonlinear_vma;   /* Check page->index if set */
 912        struct address_space *check_mapping;    /* Check page->mapping if set */
 913        pgoff_t first_index;                    /* Lowest page->index to unmap */
 914        pgoff_t last_index;                     /* Highest page->index to unmap */
 915};
 916
 917struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
 918                pte_t pte);
 919
 920int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 921                unsigned long size);
 922void zap_page_range(struct vm_area_struct *vma, unsigned long address,
 923                unsigned long size, struct zap_details *);
 924void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 925                unsigned long start, unsigned long end);
 926
 927/**
 928 * mm_walk - callbacks for walk_page_range
 929 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
 930 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
 931 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
 932 *             this handler is required to be able to handle
 933 *             pmd_trans_huge() pmds.  They may simply choose to
 934 *             split_huge_page() instead of handling it explicitly.
 935 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
 936 * @pte_hole: if set, called for each hole at all levels
 937 * @hugetlb_entry: if set, called for each hugetlb entry
 938 *                 *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
 939 *                            is used.
 940 *
 941 * (see walk_page_range for more details)
 942 */
 943struct mm_walk {
 944        int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
 945        int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
 946        int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
 947        int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
 948        int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
 949        int (*hugetlb_entry)(pte_t *, unsigned long,
 950                             unsigned long, unsigned long, struct mm_walk *);
 951        struct mm_struct *mm;
 952        void *private;
 953};
 954
 955int walk_page_range(unsigned long addr, unsigned long end,
 956                struct mm_walk *walk);
 957void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
 958                unsigned long end, unsigned long floor, unsigned long ceiling);
 959int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
 960                        struct vm_area_struct *vma);
 961void unmap_mapping_range(struct address_space *mapping,
 962                loff_t const holebegin, loff_t const holelen, int even_cows);
 963int follow_pfn(struct vm_area_struct *vma, unsigned long address,
 964        unsigned long *pfn);
 965int follow_phys(struct vm_area_struct *vma, unsigned long address,
 966                unsigned int flags, unsigned long *prot, resource_size_t *phys);
 967int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
 968                        void *buf, int len, int write);
 969
 970static inline void unmap_shared_mapping_range(struct address_space *mapping,
 971                loff_t const holebegin, loff_t const holelen)
 972{
 973        unmap_mapping_range(mapping, holebegin, holelen, 0);
 974}
 975
 976extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
 977extern void truncate_setsize(struct inode *inode, loff_t newsize);
 978extern int vmtruncate(struct inode *inode, loff_t offset);
 979void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
 980int truncate_inode_page(struct address_space *mapping, struct page *page);
 981int generic_error_remove_page(struct address_space *mapping, struct page *page);
 982int invalidate_inode_page(struct page *page);
 983
 984#ifdef CONFIG_MMU
 985extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 986                        unsigned long address, unsigned int flags);
 987extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
 988                            unsigned long address, unsigned int fault_flags);
 989#else
 990static inline int handle_mm_fault(struct mm_struct *mm,
 991                        struct vm_area_struct *vma, unsigned long address,
 992                        unsigned int flags)
 993{
 994        /* should never happen if there's no MMU */
 995        BUG();
 996        return VM_FAULT_SIGBUS;
 997}
 998static inline int fixup_user_fault(struct task_struct *tsk,
 999                struct mm_struct *mm, unsigned long address,
1000                unsigned int fault_flags)
1001{
1002        /* should never happen if there's no MMU */
1003        BUG();
1004        return -EFAULT;
1005}
1006#endif
1007
1008extern int make_pages_present(unsigned long addr, unsigned long end);
1009extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1010extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1011                void *buf, int len, int write);
1012
1013int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1014                     unsigned long start, int len, unsigned int foll_flags,
1015                     struct page **pages, struct vm_area_struct **vmas,
1016                     int *nonblocking);
1017int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1018                        unsigned long start, int nr_pages, int write, int force,
1019                        struct page **pages, struct vm_area_struct **vmas);
1020int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1021                        struct page **pages);
1022struct kvec;
1023int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1024                        struct page **pages);
1025int get_kernel_page(unsigned long start, int write, struct page **pages);
1026struct page *get_dump_page(unsigned long addr);
1027
1028extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1029extern void do_invalidatepage(struct page *page, unsigned long offset);
1030
1031int __set_page_dirty_nobuffers(struct page *page);
1032int __set_page_dirty_no_writeback(struct page *page);
1033int redirty_page_for_writepage(struct writeback_control *wbc,
1034                                struct page *page);
1035void account_page_dirtied(struct page *page, struct address_space *mapping);
1036void account_page_writeback(struct page *page);
1037int set_page_dirty(struct page *page);
1038int set_page_dirty_lock(struct page *page);
1039int clear_page_dirty_for_io(struct page *page);
1040
1041/* Is the vma a continuation of the stack vma above it? */
1042static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1043{
1044        return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1045}
1046
1047static inline int stack_guard_page_start(struct vm_area_struct *vma,
1048                                             unsigned long addr)
1049{
1050        return (vma->vm_flags & VM_GROWSDOWN) &&
1051                (vma->vm_start == addr) &&
1052                !vma_growsdown(vma->vm_prev, addr);
1053}
1054
1055/* Is the vma a continuation of the stack vma below it? */
1056static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1057{
1058        return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1059}
1060
1061static inline int stack_guard_page_end(struct vm_area_struct *vma,
1062                                           unsigned long addr)
1063{
1064        return (vma->vm_flags & VM_GROWSUP) &&
1065                (vma->vm_end == addr) &&
1066                !vma_growsup(vma->vm_next, addr);
1067}
1068
1069extern pid_t
1070vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1071
1072extern unsigned long move_page_tables(struct vm_area_struct *vma,
1073                unsigned long old_addr, struct vm_area_struct *new_vma,
1074                unsigned long new_addr, unsigned long len);
1075extern unsigned long do_mremap(unsigned long addr,
1076                               unsigned long old_len, unsigned long new_len,
1077                               unsigned long flags, unsigned long new_addr);
1078extern int mprotect_fixup(struct vm_area_struct *vma,
1079                          struct vm_area_struct **pprev, unsigned long start,
1080                          unsigned long end, unsigned long newflags);
1081
1082/*
1083 * doesn't attempt to fault and will return short.
1084 */
1085int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1086                          struct page **pages);
1087/*
1088 * per-process(per-mm_struct) statistics.
1089 */
1090static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1091{
1092        long val = atomic_long_read(&mm->rss_stat.count[member]);
1093
1094#ifdef SPLIT_RSS_COUNTING
1095        /*
1096         * counter is updated in asynchronous manner and may go to minus.
1097         * But it's never be expected number for users.
1098         */
1099        if (val < 0)
1100                val = 0;
1101#endif
1102        return (unsigned long)val;
1103}
1104
1105static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1106{
1107        atomic_long_add(value, &mm->rss_stat.count[member]);
1108}
1109
1110static inline void inc_mm_counter(struct mm_struct *mm, int member)
1111{
1112        atomic_long_inc(&mm->rss_stat.count[member]);
1113}
1114
1115static inline void dec_mm_counter(struct mm_struct *mm, int member)
1116{
1117        atomic_long_dec(&mm->rss_stat.count[member]);
1118}
1119
1120static inline unsigned long get_mm_rss(struct mm_struct *mm)
1121{
1122        return get_mm_counter(mm, MM_FILEPAGES) +
1123                get_mm_counter(mm, MM_ANONPAGES);
1124}
1125
1126static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1127{
1128        return max(mm->hiwater_rss, get_mm_rss(mm));
1129}
1130
1131static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1132{
1133        return max(mm->hiwater_vm, mm->total_vm);
1134}
1135
1136static inline void update_hiwater_rss(struct mm_struct *mm)
1137{
1138        unsigned long _rss = get_mm_rss(mm);
1139
1140        if ((mm)->hiwater_rss < _rss)
1141                (mm)->hiwater_rss = _rss;
1142}
1143
1144static inline void update_hiwater_vm(struct mm_struct *mm)
1145{
1146        if (mm->hiwater_vm < mm->total_vm)
1147                mm->hiwater_vm = mm->total_vm;
1148}
1149
1150static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1151                                         struct mm_struct *mm)
1152{
1153        unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1154
1155        if (*maxrss < hiwater_rss)
1156                *maxrss = hiwater_rss;
1157}
1158
1159#if defined(SPLIT_RSS_COUNTING)
1160void sync_mm_rss(struct mm_struct *mm);
1161#else
1162static inline void sync_mm_rss(struct mm_struct *mm)
1163{
1164}
1165#endif
1166
1167int vma_wants_writenotify(struct vm_area_struct *vma);
1168
1169extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1170                               spinlock_t **ptl);
1171static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1172                                    spinlock_t **ptl)
1173{
1174        pte_t *ptep;
1175        __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1176        return ptep;
1177}
1178
1179#ifdef __PAGETABLE_PUD_FOLDED
1180static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1181                                                unsigned long address)
1182{
1183        return 0;
1184}
1185#else
1186int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1187#endif
1188
1189#ifdef __PAGETABLE_PMD_FOLDED
1190static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1191                                                unsigned long address)
1192{
1193        return 0;
1194}
1195#else
1196int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1197#endif
1198
1199int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1200                pmd_t *pmd, unsigned long address);
1201int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1202
1203/*
1204 * The following ifdef needed to get the 4level-fixup.h header to work.
1205 * Remove it when 4level-fixup.h has been removed.
1206 */
1207#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1208static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1209{
1210        return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1211                NULL: pud_offset(pgd, address);
1212}
1213
1214static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1215{
1216        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1217                NULL: pmd_offset(pud, address);
1218}
1219#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1220
1221#if USE_SPLIT_PTLOCKS
1222/*
1223 * We tuck a spinlock to guard each pagetable page into its struct page,
1224 * at page->private, with BUILD_BUG_ON to make sure that this will not
1225 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1226 * When freeing, reset page->mapping so free_pages_check won't complain.
1227 */
1228#define __pte_lockptr(page)     &((page)->ptl)
1229#define pte_lock_init(_page)    do {                                    \
1230        spin_lock_init(__pte_lockptr(_page));                           \
1231} while (0)
1232#define pte_lock_deinit(page)   ((page)->mapping = NULL)
1233#define pte_lockptr(mm, pmd)    ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1234#else   /* !USE_SPLIT_PTLOCKS */
1235/*
1236 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1237 */
1238#define pte_lock_init(page)     do {} while (0)
1239#define pte_lock_deinit(page)   do {} while (0)
1240#define pte_lockptr(mm, pmd)    ({(void)(pmd); &(mm)->page_table_lock;})
1241#endif /* USE_SPLIT_PTLOCKS */
1242
1243static inline void pgtable_page_ctor(struct page *page)
1244{
1245        pte_lock_init(page);
1246        inc_zone_page_state(page, NR_PAGETABLE);
1247}
1248
1249static inline void pgtable_page_dtor(struct page *page)
1250{
1251        pte_lock_deinit(page);
1252        dec_zone_page_state(page, NR_PAGETABLE);
1253}
1254
1255#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1256({                                                      \
1257        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1258        pte_t *__pte = pte_offset_map(pmd, address);    \
1259        *(ptlp) = __ptl;                                \
1260        spin_lock(__ptl);                               \
1261        __pte;                                          \
1262})
1263
1264#define pte_unmap_unlock(pte, ptl)      do {            \
1265        spin_unlock(ptl);                               \
1266        pte_unmap(pte);                                 \
1267} while (0)
1268
1269#define pte_alloc_map(mm, vma, pmd, address)                            \
1270        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1271                                                        pmd, address))? \
1272         NULL: pte_offset_map(pmd, address))
1273
1274#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1275        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1276                                                        pmd, address))? \
1277                NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1278
1279#define pte_alloc_kernel(pmd, address)                  \
1280        ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1281                NULL: pte_offset_kernel(pmd, address))
1282
1283extern void free_area_init(unsigned long * zones_size);
1284extern void free_area_init_node(int nid, unsigned long * zones_size,
1285                unsigned long zone_start_pfn, unsigned long *zholes_size);
1286extern void free_initmem(void);
1287
1288#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1289/*
1290 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1291 * zones, allocate the backing mem_map and account for memory holes in a more
1292 * architecture independent manner. This is a substitute for creating the
1293 * zone_sizes[] and zholes_size[] arrays and passing them to
1294 * free_area_init_node()
1295 *
1296 * An architecture is expected to register range of page frames backed by
1297 * physical memory with memblock_add[_node]() before calling
1298 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1299 * usage, an architecture is expected to do something like
1300 *
1301 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1302 *                                                       max_highmem_pfn};
1303 * for_each_valid_physical_page_range()
1304 *      memblock_add_node(base, size, nid)
1305 * free_area_init_nodes(max_zone_pfns);
1306 *
1307 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1308 * registered physical page range.  Similarly
1309 * sparse_memory_present_with_active_regions() calls memory_present() for
1310 * each range when SPARSEMEM is enabled.
1311 *
1312 * See mm/page_alloc.c for more information on each function exposed by
1313 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1314 */
1315extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1316unsigned long node_map_pfn_alignment(void);
1317unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1318                                                unsigned long end_pfn);
1319extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1320                                                unsigned long end_pfn);
1321extern void get_pfn_range_for_nid(unsigned int nid,
1322                        unsigned long *start_pfn, unsigned long *end_pfn);
1323extern unsigned long find_min_pfn_with_active_regions(void);
1324extern void free_bootmem_with_active_regions(int nid,
1325                                                unsigned long max_low_pfn);
1326extern void sparse_memory_present_with_active_regions(int nid);
1327
1328#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1329
1330#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1331    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1332static inline int __early_pfn_to_nid(unsigned long pfn)
1333{
1334        return 0;
1335}
1336#else
1337/* please see mm/page_alloc.c */
1338extern int __meminit early_pfn_to_nid(unsigned long pfn);
1339#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1340/* there is a per-arch backend function. */
1341extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1342#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1343#endif
1344
1345extern void set_dma_reserve(unsigned long new_dma_reserve);
1346extern void memmap_init_zone(unsigned long, int, unsigned long,
1347                                unsigned long, enum memmap_context);
1348extern void setup_per_zone_wmarks(void);
1349extern int __meminit init_per_zone_wmark_min(void);
1350extern void mem_init(void);
1351extern void __init mmap_init(void);
1352extern void show_mem(unsigned int flags);
1353extern void si_meminfo(struct sysinfo * val);
1354extern void si_meminfo_node(struct sysinfo *val, int nid);
1355extern int after_bootmem;
1356
1357extern __printf(3, 4)
1358void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1359
1360extern void setup_per_cpu_pageset(void);
1361
1362extern void zone_pcp_update(struct zone *zone);
1363extern void zone_pcp_reset(struct zone *zone);
1364
1365/* nommu.c */
1366extern atomic_long_t mmap_pages_allocated;
1367extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1368
1369/* prio_tree.c */
1370void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1371void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1372void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1373struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1374        struct prio_tree_iter *iter);
1375
1376#define vma_prio_tree_foreach(vma, iter, root, begin, end)      \
1377        for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;   \
1378                (vma = vma_prio_tree_next(vma, iter)); )
1379
1380static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1381                                        struct list_head *list)
1382{
1383        vma->shared.vm_set.parent = NULL;
1384        list_add_tail(&vma->shared.vm_set.list, list);
1385}
1386
1387/* mmap.c */
1388extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1389extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1390        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1391extern struct vm_area_struct *vma_merge(struct mm_struct *,
1392        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1393        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1394        struct mempolicy *);
1395extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1396extern int split_vma(struct mm_struct *,
1397        struct vm_area_struct *, unsigned long addr, int new_below);
1398extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1399extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1400        struct rb_node **, struct rb_node *);
1401extern void unlink_file_vma(struct vm_area_struct *);
1402extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1403        unsigned long addr, unsigned long len, pgoff_t pgoff);
1404extern void exit_mmap(struct mm_struct *);
1405
1406extern int mm_take_all_locks(struct mm_struct *mm);
1407extern void mm_drop_all_locks(struct mm_struct *mm);
1408
1409/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1410extern void added_exe_file_vma(struct mm_struct *mm);
1411extern void removed_exe_file_vma(struct mm_struct *mm);
1412extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1413extern struct file *get_mm_exe_file(struct mm_struct *mm);
1414
1415extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1416extern int install_special_mapping(struct mm_struct *mm,
1417                                   unsigned long addr, unsigned long len,
1418                                   unsigned long flags, struct page **pages);
1419
1420extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1421
1422extern unsigned long mmap_region(struct file *file, unsigned long addr,
1423        unsigned long len, unsigned long flags,
1424        vm_flags_t vm_flags, unsigned long pgoff);
1425extern unsigned long do_mmap_pgoff(struct file *, unsigned long,
1426        unsigned long, unsigned long,
1427        unsigned long, unsigned long);
1428extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1429
1430/* These take the mm semaphore themselves */
1431extern unsigned long vm_brk(unsigned long, unsigned long);
1432extern int vm_munmap(unsigned long, size_t);
1433extern unsigned long vm_mmap(struct file *, unsigned long,
1434        unsigned long, unsigned long,
1435        unsigned long, unsigned long);
1436
1437/* truncate.c */
1438extern void truncate_inode_pages(struct address_space *, loff_t);
1439extern void truncate_inode_pages_range(struct address_space *,
1440                                       loff_t lstart, loff_t lend);
1441
1442/* generic vm_area_ops exported for stackable file systems */
1443extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1444extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1445
1446/* mm/page-writeback.c */
1447int write_one_page(struct page *page, int wait);
1448void task_dirty_inc(struct task_struct *tsk);
1449
1450/* readahead.c */
1451#define VM_MAX_READAHEAD        128     /* kbytes */
1452#define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1453
1454int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1455                        pgoff_t offset, unsigned long nr_to_read);
1456
1457void page_cache_sync_readahead(struct address_space *mapping,
1458                               struct file_ra_state *ra,
1459                               struct file *filp,
1460                               pgoff_t offset,
1461                               unsigned long size);
1462
1463void page_cache_async_readahead(struct address_space *mapping,
1464                                struct file_ra_state *ra,
1465                                struct file *filp,
1466                                struct page *pg,
1467                                pgoff_t offset,
1468                                unsigned long size);
1469
1470unsigned long max_sane_readahead(unsigned long nr);
1471unsigned long ra_submit(struct file_ra_state *ra,
1472                        struct address_space *mapping,
1473                        struct file *filp);
1474
1475/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1476extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1477
1478/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1479extern int expand_downwards(struct vm_area_struct *vma,
1480                unsigned long address);
1481#if VM_GROWSUP
1482extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1483#else
1484  #define expand_upwards(vma, address) do { } while (0)
1485#endif
1486
1487/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1488extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1489extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1490                                             struct vm_area_struct **pprev);
1491
1492/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1493   NULL if none.  Assume start_addr < end_addr. */
1494static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1495{
1496        struct vm_area_struct * vma = find_vma(mm,start_addr);
1497
1498        if (vma && end_addr <= vma->vm_start)
1499                vma = NULL;
1500        return vma;
1501}
1502
1503static inline unsigned long vma_pages(struct vm_area_struct *vma)
1504{
1505        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1506}
1507
1508/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1509static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1510                                unsigned long vm_start, unsigned long vm_end)
1511{
1512        struct vm_area_struct *vma = find_vma(mm, vm_start);
1513
1514        if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1515                vma = NULL;
1516
1517        return vma;
1518}
1519
1520#ifdef CONFIG_MMU
1521pgprot_t vm_get_page_prot(unsigned long vm_flags);
1522#else
1523static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1524{
1525        return __pgprot(0);
1526}
1527#endif
1528
1529struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1530int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1531                        unsigned long pfn, unsigned long size, pgprot_t);
1532int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1533int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1534                        unsigned long pfn);
1535int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1536                        unsigned long pfn);
1537
1538struct page *follow_page(struct vm_area_struct *, unsigned long address,
1539                        unsigned int foll_flags);
1540#define FOLL_WRITE      0x01    /* check pte is writable */
1541#define FOLL_TOUCH      0x02    /* mark page accessed */
1542#define FOLL_GET        0x04    /* do get_page on page */
1543#define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
1544#define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
1545#define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
1546                                 * and return without waiting upon it */
1547#define FOLL_MLOCK      0x40    /* mark page as mlocked */
1548#define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
1549#define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
1550
1551typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1552                        void *data);
1553extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1554                               unsigned long size, pte_fn_t fn, void *data);
1555
1556#ifdef CONFIG_PROC_FS
1557void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1558#else
1559static inline void vm_stat_account(struct mm_struct *mm,
1560                        unsigned long flags, struct file *file, long pages)
1561{
1562        mm->total_vm += pages;
1563}
1564#endif /* CONFIG_PROC_FS */
1565
1566#ifdef CONFIG_DEBUG_PAGEALLOC
1567extern void kernel_map_pages(struct page *page, int numpages, int enable);
1568#ifdef CONFIG_HIBERNATION
1569extern bool kernel_page_present(struct page *page);
1570#endif /* CONFIG_HIBERNATION */
1571#else
1572static inline void
1573kernel_map_pages(struct page *page, int numpages, int enable) {}
1574#ifdef CONFIG_HIBERNATION
1575static inline bool kernel_page_present(struct page *page) { return true; }
1576#endif /* CONFIG_HIBERNATION */
1577#endif
1578
1579extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1580#ifdef  __HAVE_ARCH_GATE_AREA
1581int in_gate_area_no_mm(unsigned long addr);
1582int in_gate_area(struct mm_struct *mm, unsigned long addr);
1583#else
1584int in_gate_area_no_mm(unsigned long addr);
1585#define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1586#endif  /* __HAVE_ARCH_GATE_AREA */
1587
1588int drop_caches_sysctl_handler(struct ctl_table *, int,
1589                                        void __user *, size_t *, loff_t *);
1590unsigned long shrink_slab(struct shrink_control *shrink,
1591                          unsigned long nr_pages_scanned,
1592                          unsigned long lru_pages);
1593
1594#ifndef CONFIG_MMU
1595#define randomize_va_space 0
1596#else
1597extern int randomize_va_space;
1598#endif
1599
1600const char * arch_vma_name(struct vm_area_struct *vma);
1601void print_vma_addr(char *prefix, unsigned long rip);
1602
1603void sparse_mem_maps_populate_node(struct page **map_map,
1604                                   unsigned long pnum_begin,
1605                                   unsigned long pnum_end,
1606                                   unsigned long map_count,
1607                                   int nodeid);
1608
1609struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1610pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1611pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1612pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1613pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1614void *vmemmap_alloc_block(unsigned long size, int node);
1615void *vmemmap_alloc_block_buf(unsigned long size, int node);
1616void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1617int vmemmap_populate_basepages(struct page *start_page,
1618                                                unsigned long pages, int node);
1619int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1620void vmemmap_populate_print_last(void);
1621
1622
1623enum mf_flags {
1624        MF_COUNT_INCREASED = 1 << 0,
1625        MF_ACTION_REQUIRED = 1 << 1,
1626        MF_MUST_KILL = 1 << 2,
1627};
1628extern int memory_failure(unsigned long pfn, int trapno, int flags);
1629extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1630extern int unpoison_memory(unsigned long pfn);
1631extern int sysctl_memory_failure_early_kill;
1632extern int sysctl_memory_failure_recovery;
1633extern void shake_page(struct page *p, int access);
1634extern atomic_long_t mce_bad_pages;
1635extern int soft_offline_page(struct page *page, int flags);
1636
1637extern void dump_page(struct page *page);
1638
1639#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1640extern void clear_huge_page(struct page *page,
1641                            unsigned long addr,
1642                            unsigned int pages_per_huge_page);
1643extern void copy_user_huge_page(struct page *dst, struct page *src,
1644                                unsigned long addr, struct vm_area_struct *vma,
1645                                unsigned int pages_per_huge_page);
1646#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1647
1648#ifdef CONFIG_DEBUG_PAGEALLOC
1649extern unsigned int _debug_guardpage_minorder;
1650
1651static inline unsigned int debug_guardpage_minorder(void)
1652{
1653        return _debug_guardpage_minorder;
1654}
1655
1656static inline bool page_is_guard(struct page *page)
1657{
1658        return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1659}
1660#else
1661static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1662static inline bool page_is_guard(struct page *page) { return false; }
1663#endif /* CONFIG_DEBUG_PAGEALLOC */
1664
1665#endif /* __KERNEL__ */
1666#endif /* _LINUX_MM_H */
1667
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.