linux/include/linux/mm.h
<<
>>
Prefs
   1/* SPDX-License-Identifier: GPL-2.0 */
   2#ifndef _LINUX_MM_H
   3#define _LINUX_MM_H
   4
   5#include <linux/errno.h>
   6
   7#ifdef __KERNEL__
   8
   9#include <linux/mmdebug.h>
  10#include <linux/gfp.h>
  11#include <linux/bug.h>
  12#include <linux/list.h>
  13#include <linux/mmzone.h>
  14#include <linux/rbtree.h>
  15#include <linux/atomic.h>
  16#include <linux/debug_locks.h>
  17#include <linux/mm_types.h>
  18#include <linux/range.h>
  19#include <linux/pfn.h>
  20#include <linux/percpu-refcount.h>
  21#include <linux/bit_spinlock.h>
  22#include <linux/shrinker.h>
  23#include <linux/resource.h>
  24#include <linux/page_ext.h>
  25#include <linux/err.h>
  26#include <linux/page_ref.h>
  27#include <linux/memremap.h>
  28#include <linux/overflow.h>
  29#include <linux/sizes.h>
  30
  31struct mempolicy;
  32struct anon_vma;
  33struct anon_vma_chain;
  34struct file_ra_state;
  35struct user_struct;
  36struct writeback_control;
  37struct bdi_writeback;
  38
  39void init_mm_internals(void);
  40
  41#ifndef CONFIG_NEED_MULTIPLE_NODES      /* Don't use mapnrs, do it properly */
  42extern unsigned long max_mapnr;
  43
  44static inline void set_max_mapnr(unsigned long limit)
  45{
  46        max_mapnr = limit;
  47}
  48#else
  49static inline void set_max_mapnr(unsigned long limit) { }
  50#endif
  51
  52extern atomic_long_t _totalram_pages;
  53static inline unsigned long totalram_pages(void)
  54{
  55        return (unsigned long)atomic_long_read(&_totalram_pages);
  56}
  57
  58static inline void totalram_pages_inc(void)
  59{
  60        atomic_long_inc(&_totalram_pages);
  61}
  62
  63static inline void totalram_pages_dec(void)
  64{
  65        atomic_long_dec(&_totalram_pages);
  66}
  67
  68static inline void totalram_pages_add(long count)
  69{
  70        atomic_long_add(count, &_totalram_pages);
  71}
  72
  73static inline void totalram_pages_set(long val)
  74{
  75        atomic_long_set(&_totalram_pages, val);
  76}
  77
  78extern void * high_memory;
  79extern int page_cluster;
  80
  81#ifdef CONFIG_SYSCTL
  82extern int sysctl_legacy_va_layout;
  83#else
  84#define sysctl_legacy_va_layout 0
  85#endif
  86
  87#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
  88extern const int mmap_rnd_bits_min;
  89extern const int mmap_rnd_bits_max;
  90extern int mmap_rnd_bits __read_mostly;
  91#endif
  92#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
  93extern const int mmap_rnd_compat_bits_min;
  94extern const int mmap_rnd_compat_bits_max;
  95extern int mmap_rnd_compat_bits __read_mostly;
  96#endif
  97
  98#include <asm/page.h>
  99#include <asm/pgtable.h>
 100#include <asm/processor.h>
 101
 102/*
 103 * Architectures that support memory tagging (assigning tags to memory regions,
 104 * embedding these tags into addresses that point to these memory regions, and
 105 * checking that the memory and the pointer tags match on memory accesses)
 106 * redefine this macro to strip tags from pointers.
 107 * It's defined as noop for arcitectures that don't support memory tagging.
 108 */
 109#ifndef untagged_addr
 110#define untagged_addr(addr) (addr)
 111#endif
 112
 113#ifndef __pa_symbol
 114#define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
 115#endif
 116
 117#ifndef page_to_virt
 118#define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
 119#endif
 120
 121#ifndef lm_alias
 122#define lm_alias(x)     __va(__pa_symbol(x))
 123#endif
 124
 125/*
 126 * To prevent common memory management code establishing
 127 * a zero page mapping on a read fault.
 128 * This macro should be defined within <asm/pgtable.h>.
 129 * s390 does this to prevent multiplexing of hardware bits
 130 * related to the physical page in case of virtualization.
 131 */
 132#ifndef mm_forbids_zeropage
 133#define mm_forbids_zeropage(X)  (0)
 134#endif
 135
 136/*
 137 * On some architectures it is expensive to call memset() for small sizes.
 138 * If an architecture decides to implement their own version of
 139 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
 140 * define their own version of this macro in <asm/pgtable.h>
 141 */
 142#if BITS_PER_LONG == 64
 143/* This function must be updated when the size of struct page grows above 80
 144 * or reduces below 56. The idea that compiler optimizes out switch()
 145 * statement, and only leaves move/store instructions. Also the compiler can
 146 * combine write statments if they are both assignments and can be reordered,
 147 * this can result in several of the writes here being dropped.
 148 */
 149#define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
 150static inline void __mm_zero_struct_page(struct page *page)
 151{
 152        unsigned long *_pp = (void *)page;
 153
 154         /* Check that struct page is either 56, 64, 72, or 80 bytes */
 155        BUILD_BUG_ON(sizeof(struct page) & 7);
 156        BUILD_BUG_ON(sizeof(struct page) < 56);
 157        BUILD_BUG_ON(sizeof(struct page) > 80);
 158
 159        switch (sizeof(struct page)) {
 160        case 80:
 161                _pp[9] = 0;     /* fallthrough */
 162        case 72:
 163                _pp[8] = 0;     /* fallthrough */
 164        case 64:
 165                _pp[7] = 0;     /* fallthrough */
 166        case 56:
 167                _pp[6] = 0;
 168                _pp[5] = 0;
 169                _pp[4] = 0;
 170                _pp[3] = 0;
 171                _pp[2] = 0;
 172                _pp[1] = 0;
 173                _pp[0] = 0;
 174        }
 175}
 176#else
 177#define mm_zero_struct_page(pp)  ((void)memset((pp), 0, sizeof(struct page)))
 178#endif
 179
 180/*
 181 * Default maximum number of active map areas, this limits the number of vmas
 182 * per mm struct. Users can overwrite this number by sysctl but there is a
 183 * problem.
 184 *
 185 * When a program's coredump is generated as ELF format, a section is created
 186 * per a vma. In ELF, the number of sections is represented in unsigned short.
 187 * This means the number of sections should be smaller than 65535 at coredump.
 188 * Because the kernel adds some informative sections to a image of program at
 189 * generating coredump, we need some margin. The number of extra sections is
 190 * 1-3 now and depends on arch. We use "5" as safe margin, here.
 191 *
 192 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
 193 * not a hard limit any more. Although some userspace tools can be surprised by
 194 * that.
 195 */
 196#define MAPCOUNT_ELF_CORE_MARGIN        (5)
 197#define DEFAULT_MAX_MAP_COUNT   (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
 198
 199extern int sysctl_max_map_count;
 200
 201extern unsigned long sysctl_user_reserve_kbytes;
 202extern unsigned long sysctl_admin_reserve_kbytes;
 203
 204extern int sysctl_overcommit_memory;
 205extern int sysctl_overcommit_ratio;
 206extern unsigned long sysctl_overcommit_kbytes;
 207
 208extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
 209                                    size_t *, loff_t *);
 210extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
 211                                    size_t *, loff_t *);
 212
 213#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
 214
 215/* to align the pointer to the (next) page boundary */
 216#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
 217
 218/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
 219#define PAGE_ALIGNED(addr)      IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
 220
 221#define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
 222
 223/*
 224 * Linux kernel virtual memory manager primitives.
 225 * The idea being to have a "virtual" mm in the same way
 226 * we have a virtual fs - giving a cleaner interface to the
 227 * mm details, and allowing different kinds of memory mappings
 228 * (from shared memory to executable loading to arbitrary
 229 * mmap() functions).
 230 */
 231
 232struct vm_area_struct *vm_area_alloc(struct mm_struct *);
 233struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
 234void vm_area_free(struct vm_area_struct *);
 235
 236#ifndef CONFIG_MMU
 237extern struct rb_root nommu_region_tree;
 238extern struct rw_semaphore nommu_region_sem;
 239
 240extern unsigned int kobjsize(const void *objp);
 241#endif
 242
 243/*
 244 * vm_flags in vm_area_struct, see mm_types.h.
 245 * When changing, update also include/trace/events/mmflags.h
 246 */
 247#define VM_NONE         0x00000000
 248
 249#define VM_READ         0x00000001      /* currently active flags */
 250#define VM_WRITE        0x00000002
 251#define VM_EXEC         0x00000004
 252#define VM_SHARED       0x00000008
 253
 254/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
 255#define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
 256#define VM_MAYWRITE     0x00000020
 257#define VM_MAYEXEC      0x00000040
 258#define VM_MAYSHARE     0x00000080
 259
 260#define VM_GROWSDOWN    0x00000100      /* general info on the segment */
 261#define VM_UFFD_MISSING 0x00000200      /* missing pages tracking */
 262#define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
 263#define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
 264#define VM_UFFD_WP      0x00001000      /* wrprotect pages tracking */
 265
 266#define VM_LOCKED       0x00002000
 267#define VM_IO           0x00004000      /* Memory mapped I/O or similar */
 268
 269                                        /* Used by sys_madvise() */
 270#define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
 271#define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
 272
 273#define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
 274#define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
 275#define VM_LOCKONFAULT  0x00080000      /* Lock the pages covered when they are faulted in */
 276#define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
 277#define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
 278#define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
 279#define VM_SYNC         0x00800000      /* Synchronous page faults */
 280#define VM_ARCH_1       0x01000000      /* Architecture-specific flag */
 281#define VM_WIPEONFORK   0x02000000      /* Wipe VMA contents in child. */
 282#define VM_DONTDUMP     0x04000000      /* Do not include in the core dump */
 283
 284#ifdef CONFIG_MEM_SOFT_DIRTY
 285# define VM_SOFTDIRTY   0x08000000      /* Not soft dirty clean area */
 286#else
 287# define VM_SOFTDIRTY   0
 288#endif
 289
 290#define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
 291#define VM_HUGEPAGE     0x20000000      /* MADV_HUGEPAGE marked this vma */
 292#define VM_NOHUGEPAGE   0x40000000      /* MADV_NOHUGEPAGE marked this vma */
 293#define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
 294
 295#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
 296#define VM_HIGH_ARCH_BIT_0      32      /* bit only usable on 64-bit architectures */
 297#define VM_HIGH_ARCH_BIT_1      33      /* bit only usable on 64-bit architectures */
 298#define VM_HIGH_ARCH_BIT_2      34      /* bit only usable on 64-bit architectures */
 299#define VM_HIGH_ARCH_BIT_3      35      /* bit only usable on 64-bit architectures */
 300#define VM_HIGH_ARCH_BIT_4      36      /* bit only usable on 64-bit architectures */
 301#define VM_HIGH_ARCH_0  BIT(VM_HIGH_ARCH_BIT_0)
 302#define VM_HIGH_ARCH_1  BIT(VM_HIGH_ARCH_BIT_1)
 303#define VM_HIGH_ARCH_2  BIT(VM_HIGH_ARCH_BIT_2)
 304#define VM_HIGH_ARCH_3  BIT(VM_HIGH_ARCH_BIT_3)
 305#define VM_HIGH_ARCH_4  BIT(VM_HIGH_ARCH_BIT_4)
 306#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
 307
 308#ifdef CONFIG_ARCH_HAS_PKEYS
 309# define VM_PKEY_SHIFT  VM_HIGH_ARCH_BIT_0
 310# define VM_PKEY_BIT0   VM_HIGH_ARCH_0  /* A protection key is a 4-bit value */
 311# define VM_PKEY_BIT1   VM_HIGH_ARCH_1  /* on x86 and 5-bit value on ppc64   */
 312# define VM_PKEY_BIT2   VM_HIGH_ARCH_2
 313# define VM_PKEY_BIT3   VM_HIGH_ARCH_3
 314#ifdef CONFIG_PPC
 315# define VM_PKEY_BIT4  VM_HIGH_ARCH_4
 316#else
 317# define VM_PKEY_BIT4  0
 318#endif
 319#endif /* CONFIG_ARCH_HAS_PKEYS */
 320
 321#if defined(CONFIG_X86)
 322# define VM_PAT         VM_ARCH_1       /* PAT reserves whole VMA at once (x86) */
 323#elif defined(CONFIG_PPC)
 324# define VM_SAO         VM_ARCH_1       /* Strong Access Ordering (powerpc) */
 325#elif defined(CONFIG_PARISC)
 326# define VM_GROWSUP     VM_ARCH_1
 327#elif defined(CONFIG_IA64)
 328# define VM_GROWSUP     VM_ARCH_1
 329#elif defined(CONFIG_SPARC64)
 330# define VM_SPARC_ADI   VM_ARCH_1       /* Uses ADI tag for access control */
 331# define VM_ARCH_CLEAR  VM_SPARC_ADI
 332#elif !defined(CONFIG_MMU)
 333# define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
 334#endif
 335
 336#if defined(CONFIG_X86_INTEL_MPX)
 337/* MPX specific bounds table or bounds directory */
 338# define VM_MPX         VM_HIGH_ARCH_4
 339#else
 340# define VM_MPX         VM_NONE
 341#endif
 342
 343#ifndef VM_GROWSUP
 344# define VM_GROWSUP     VM_NONE
 345#endif
 346
 347/* Bits set in the VMA until the stack is in its final location */
 348#define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
 349
 350#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 351#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 352#endif
 353
 354#ifdef CONFIG_STACK_GROWSUP
 355#define VM_STACK        VM_GROWSUP
 356#else
 357#define VM_STACK        VM_GROWSDOWN
 358#endif
 359
 360#define VM_STACK_FLAGS  (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 361
 362/*
 363 * Special vmas that are non-mergable, non-mlock()able.
 364 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
 365 */
 366#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
 367
 368/* This mask defines which mm->def_flags a process can inherit its parent */
 369#define VM_INIT_DEF_MASK        VM_NOHUGEPAGE
 370
 371/* This mask is used to clear all the VMA flags used by mlock */
 372#define VM_LOCKED_CLEAR_MASK    (~(VM_LOCKED | VM_LOCKONFAULT))
 373
 374/* Arch-specific flags to clear when updating VM flags on protection change */
 375#ifndef VM_ARCH_CLEAR
 376# define VM_ARCH_CLEAR  VM_NONE
 377#endif
 378#define VM_FLAGS_CLEAR  (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
 379
 380/*
 381 * mapping from the currently active vm_flags protection bits (the
 382 * low four bits) to a page protection mask..
 383 */
 384extern pgprot_t protection_map[16];
 385
 386#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 387#define FAULT_FLAG_MKWRITE      0x02    /* Fault was mkwrite of existing pte */
 388#define FAULT_FLAG_ALLOW_RETRY  0x04    /* Retry fault if blocking */
 389#define FAULT_FLAG_RETRY_NOWAIT 0x08    /* Don't drop mmap_sem and wait when retrying */
 390#define FAULT_FLAG_KILLABLE     0x10    /* The fault task is in SIGKILL killable region */
 391#define FAULT_FLAG_TRIED        0x20    /* Second try */
 392#define FAULT_FLAG_USER         0x40    /* The fault originated in userspace */
 393#define FAULT_FLAG_REMOTE       0x80    /* faulting for non current tsk/mm */
 394#define FAULT_FLAG_INSTRUCTION  0x100   /* The fault was during an instruction fetch */
 395
 396#define FAULT_FLAG_TRACE \
 397        { FAULT_FLAG_WRITE,             "WRITE" }, \
 398        { FAULT_FLAG_MKWRITE,           "MKWRITE" }, \
 399        { FAULT_FLAG_ALLOW_RETRY,       "ALLOW_RETRY" }, \
 400        { FAULT_FLAG_RETRY_NOWAIT,      "RETRY_NOWAIT" }, \
 401        { FAULT_FLAG_KILLABLE,          "KILLABLE" }, \
 402        { FAULT_FLAG_TRIED,             "TRIED" }, \
 403        { FAULT_FLAG_USER,              "USER" }, \
 404        { FAULT_FLAG_REMOTE,            "REMOTE" }, \
 405        { FAULT_FLAG_INSTRUCTION,       "INSTRUCTION" }
 406
 407/*
 408 * vm_fault is filled by the the pagefault handler and passed to the vma's
 409 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 410 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 411 *
 412 * MM layer fills up gfp_mask for page allocations but fault handler might
 413 * alter it if its implementation requires a different allocation context.
 414 *
 415 * pgoff should be used in favour of virtual_address, if possible.
 416 */
 417struct vm_fault {
 418        struct vm_area_struct *vma;     /* Target VMA */
 419        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 420        gfp_t gfp_mask;                 /* gfp mask to be used for allocations */
 421        pgoff_t pgoff;                  /* Logical page offset based on vma */
 422        unsigned long address;          /* Faulting virtual address */
 423        pmd_t *pmd;                     /* Pointer to pmd entry matching
 424                                         * the 'address' */
 425        pud_t *pud;                     /* Pointer to pud entry matching
 426                                         * the 'address'
 427                                         */
 428        pte_t orig_pte;                 /* Value of PTE at the time of fault */
 429
 430        struct page *cow_page;          /* Page handler may use for COW fault */
 431        struct mem_cgroup *memcg;       /* Cgroup cow_page belongs to */
 432        struct page *page;              /* ->fault handlers should return a
 433                                         * page here, unless VM_FAULT_NOPAGE
 434                                         * is set (which is also implied by
 435                                         * VM_FAULT_ERROR).
 436                                         */
 437        /* These three entries are valid only while holding ptl lock */
 438        pte_t *pte;                     /* Pointer to pte entry matching
 439                                         * the 'address'. NULL if the page
 440                                         * table hasn't been allocated.
 441                                         */
 442        spinlock_t *ptl;                /* Page table lock.
 443                                         * Protects pte page table if 'pte'
 444                                         * is not NULL, otherwise pmd.
 445                                         */
 446        pgtable_t prealloc_pte;         /* Pre-allocated pte page table.
 447                                         * vm_ops->map_pages() calls
 448                                         * alloc_set_pte() from atomic context.
 449                                         * do_fault_around() pre-allocates
 450                                         * page table to avoid allocation from
 451                                         * atomic context.
 452                                         */
 453};
 454
 455/* page entry size for vm->huge_fault() */
 456enum page_entry_size {
 457        PE_SIZE_PTE = 0,
 458        PE_SIZE_PMD,
 459        PE_SIZE_PUD,
 460};
 461
 462/*
 463 * These are the virtual MM functions - opening of an area, closing and
 464 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 465 * to the functions called when a no-page or a wp-page exception occurs.
 466 */
 467struct vm_operations_struct {
 468        void (*open)(struct vm_area_struct * area);
 469        void (*close)(struct vm_area_struct * area);
 470        int (*split)(struct vm_area_struct * area, unsigned long addr);
 471        int (*mremap)(struct vm_area_struct * area);
 472        vm_fault_t (*fault)(struct vm_fault *vmf);
 473        vm_fault_t (*huge_fault)(struct vm_fault *vmf,
 474                        enum page_entry_size pe_size);
 475        void (*map_pages)(struct vm_fault *vmf,
 476                        pgoff_t start_pgoff, pgoff_t end_pgoff);
 477        unsigned long (*pagesize)(struct vm_area_struct * area);
 478
 479        /* notification that a previously read-only page is about to become
 480         * writable, if an error is returned it will cause a SIGBUS */
 481        vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
 482
 483        /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
 484        vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
 485
 486        /* called by access_process_vm when get_user_pages() fails, typically
 487         * for use by special VMAs that can switch between memory and hardware
 488         */
 489        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 490                      void *buf, int len, int write);
 491
 492        /* Called by the /proc/PID/maps code to ask the vma whether it
 493         * has a special name.  Returning non-NULL will also cause this
 494         * vma to be dumped unconditionally. */
 495        const char *(*name)(struct vm_area_struct *vma);
 496
 497#ifdef CONFIG_NUMA
 498        /*
 499         * set_policy() op must add a reference to any non-NULL @new mempolicy
 500         * to hold the policy upon return.  Caller should pass NULL @new to
 501         * remove a policy and fall back to surrounding context--i.e. do not
 502         * install a MPOL_DEFAULT policy, nor the task or system default
 503         * mempolicy.
 504         */
 505        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 506
 507        /*
 508         * get_policy() op must add reference [mpol_get()] to any policy at
 509         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 510         * in mm/mempolicy.c will do this automatically.
 511         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 512         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 513         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 514         * must return NULL--i.e., do not "fallback" to task or system default
 515         * policy.
 516         */
 517        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 518                                        unsigned long addr);
 519#endif
 520        /*
 521         * Called by vm_normal_page() for special PTEs to find the
 522         * page for @addr.  This is useful if the default behavior
 523         * (using pte_page()) would not find the correct page.
 524         */
 525        struct page *(*find_special_page)(struct vm_area_struct *vma,
 526                                          unsigned long addr);
 527};
 528
 529static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
 530{
 531        static const struct vm_operations_struct dummy_vm_ops = {};
 532
 533        memset(vma, 0, sizeof(*vma));
 534        vma->vm_mm = mm;
 535        vma->vm_ops = &dummy_vm_ops;
 536        INIT_LIST_HEAD(&vma->anon_vma_chain);
 537}
 538
 539static inline void vma_set_anonymous(struct vm_area_struct *vma)
 540{
 541        vma->vm_ops = NULL;
 542}
 543
 544static inline bool vma_is_anonymous(struct vm_area_struct *vma)
 545{
 546        return !vma->vm_ops;
 547}
 548
 549#ifdef CONFIG_SHMEM
 550/*
 551 * The vma_is_shmem is not inline because it is used only by slow
 552 * paths in userfault.
 553 */
 554bool vma_is_shmem(struct vm_area_struct *vma);
 555#else
 556static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
 557#endif
 558
 559int vma_is_stack_for_current(struct vm_area_struct *vma);
 560
 561/* flush_tlb_range() takes a vma, not a mm, and can care about flags */
 562#define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
 563
 564struct mmu_gather;
 565struct inode;
 566
 567#if !defined(CONFIG_ARCH_HAS_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
 568static inline int pmd_devmap(pmd_t pmd)
 569{
 570        return 0;
 571}
 572static inline int pud_devmap(pud_t pud)
 573{
 574        return 0;
 575}
 576static inline int pgd_devmap(pgd_t pgd)
 577{
 578        return 0;
 579}
 580#endif
 581
 582/*
 583 * FIXME: take this include out, include page-flags.h in
 584 * files which need it (119 of them)
 585 */
 586#include <linux/page-flags.h>
 587#include <linux/huge_mm.h>
 588
 589/*
 590 * Methods to modify the page usage count.
 591 *
 592 * What counts for a page usage:
 593 * - cache mapping   (page->mapping)
 594 * - private data    (page->private)
 595 * - page mapped in a task's page tables, each mapping
 596 *   is counted separately
 597 *
 598 * Also, many kernel routines increase the page count before a critical
 599 * routine so they can be sure the page doesn't go away from under them.
 600 */
 601
 602/*
 603 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 604 */
 605static inline int put_page_testzero(struct page *page)
 606{
 607        VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
 608        return page_ref_dec_and_test(page);
 609}
 610
 611/*
 612 * Try to grab a ref unless the page has a refcount of zero, return false if
 613 * that is the case.
 614 * This can be called when MMU is off so it must not access
 615 * any of the virtual mappings.
 616 */
 617static inline int get_page_unless_zero(struct page *page)
 618{
 619        return page_ref_add_unless(page, 1, 0);
 620}
 621
 622extern int page_is_ram(unsigned long pfn);
 623
 624enum {
 625        REGION_INTERSECTS,
 626        REGION_DISJOINT,
 627        REGION_MIXED,
 628};
 629
 630int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
 631                      unsigned long desc);
 632
 633/* Support for virtually mapped pages */
 634struct page *vmalloc_to_page(const void *addr);
 635unsigned long vmalloc_to_pfn(const void *addr);
 636
 637/*
 638 * Determine if an address is within the vmalloc range
 639 *
 640 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 641 * is no special casing required.
 642 */
 643static inline bool is_vmalloc_addr(const void *x)
 644{
 645#ifdef CONFIG_MMU
 646        unsigned long addr = (unsigned long)x;
 647
 648        return addr >= VMALLOC_START && addr < VMALLOC_END;
 649#else
 650        return false;
 651#endif
 652}
 653
 654#ifndef is_ioremap_addr
 655#define is_ioremap_addr(x) is_vmalloc_addr(x)
 656#endif
 657
 658#ifdef CONFIG_MMU
 659extern int is_vmalloc_or_module_addr(const void *x);
 660#else
 661static inline int is_vmalloc_or_module_addr(const void *x)
 662{
 663        return 0;
 664}
 665#endif
 666
 667extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
 668static inline void *kvmalloc(size_t size, gfp_t flags)
 669{
 670        return kvmalloc_node(size, flags, NUMA_NO_NODE);
 671}
 672static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
 673{
 674        return kvmalloc_node(size, flags | __GFP_ZERO, node);
 675}
 676static inline void *kvzalloc(size_t size, gfp_t flags)
 677{
 678        return kvmalloc(size, flags | __GFP_ZERO);
 679}
 680
 681static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
 682{
 683        size_t bytes;
 684
 685        if (unlikely(check_mul_overflow(n, size, &bytes)))
 686                return NULL;
 687
 688        return kvmalloc(bytes, flags);
 689}
 690
 691static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
 692{
 693        return kvmalloc_array(n, size, flags | __GFP_ZERO);
 694}
 695
 696extern void kvfree(const void *addr);
 697
 698static inline int compound_mapcount(struct page *page)
 699{
 700        VM_BUG_ON_PAGE(!PageCompound(page), page);
 701        page = compound_head(page);
 702        return atomic_read(compound_mapcount_ptr(page)) + 1;
 703}
 704
 705/*
 706 * The atomic page->_mapcount, starts from -1: so that transitions
 707 * both from it and to it can be tracked, using atomic_inc_and_test
 708 * and atomic_add_negative(-1).
 709 */
 710static inline void page_mapcount_reset(struct page *page)
 711{
 712        atomic_set(&(page)->_mapcount, -1);
 713}
 714
 715int __page_mapcount(struct page *page);
 716
 717static inline int page_mapcount(struct page *page)
 718{
 719        VM_BUG_ON_PAGE(PageSlab(page), page);
 720
 721        if (unlikely(PageCompound(page)))
 722                return __page_mapcount(page);
 723        return atomic_read(&page->_mapcount) + 1;
 724}
 725
 726#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 727int total_mapcount(struct page *page);
 728int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
 729#else
 730static inline int total_mapcount(struct page *page)
 731{
 732        return page_mapcount(page);
 733}
 734static inline int page_trans_huge_mapcount(struct page *page,
 735                                           int *total_mapcount)
 736{
 737        int mapcount = page_mapcount(page);
 738        if (total_mapcount)
 739                *total_mapcount = mapcount;
 740        return mapcount;
 741}
 742#endif
 743
 744static inline struct page *virt_to_head_page(const void *x)
 745{
 746        struct page *page = virt_to_page(x);
 747
 748        return compound_head(page);
 749}
 750
 751void __put_page(struct page *page);
 752
 753void put_pages_list(struct list_head *pages);
 754
 755void split_page(struct page *page, unsigned int order);
 756
 757/*
 758 * Compound pages have a destructor function.  Provide a
 759 * prototype for that function and accessor functions.
 760 * These are _only_ valid on the head of a compound page.
 761 */
 762typedef void compound_page_dtor(struct page *);
 763
 764/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
 765enum compound_dtor_id {
 766        NULL_COMPOUND_DTOR,
 767        COMPOUND_PAGE_DTOR,
 768#ifdef CONFIG_HUGETLB_PAGE
 769        HUGETLB_PAGE_DTOR,
 770#endif
 771#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 772        TRANSHUGE_PAGE_DTOR,
 773#endif
 774        NR_COMPOUND_DTORS,
 775};
 776extern compound_page_dtor * const compound_page_dtors[];
 777
 778static inline void set_compound_page_dtor(struct page *page,
 779                enum compound_dtor_id compound_dtor)
 780{
 781        VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
 782        page[1].compound_dtor = compound_dtor;
 783}
 784
 785static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 786{
 787        VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
 788        return compound_page_dtors[page[1].compound_dtor];
 789}
 790
 791static inline unsigned int compound_order(struct page *page)
 792{
 793        if (!PageHead(page))
 794                return 0;
 795        return page[1].compound_order;
 796}
 797
 798static inline void set_compound_order(struct page *page, unsigned int order)
 799{
 800        page[1].compound_order = order;
 801}
 802
 803/* Returns the number of pages in this potentially compound page. */
 804static inline unsigned long compound_nr(struct page *page)
 805{
 806        return 1UL << compound_order(page);
 807}
 808
 809/* Returns the number of bytes in this potentially compound page. */
 810static inline unsigned long page_size(struct page *page)
 811{
 812        return PAGE_SIZE << compound_order(page);
 813}
 814
 815/* Returns the number of bits needed for the number of bytes in a page */
 816static inline unsigned int page_shift(struct page *page)
 817{
 818        return PAGE_SHIFT + compound_order(page);
 819}
 820
 821void free_compound_page(struct page *page);
 822
 823#ifdef CONFIG_MMU
 824/*
 825 * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
 826 * servicing faults for write access.  In the normal case, do always want
 827 * pte_mkwrite.  But get_user_pages can cause write faults for mappings
 828 * that do not have writing enabled, when used by access_process_vm.
 829 */
 830static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 831{
 832        if (likely(vma->vm_flags & VM_WRITE))
 833                pte = pte_mkwrite(pte);
 834        return pte;
 835}
 836
 837vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
 838                struct page *page);
 839vm_fault_t finish_fault(struct vm_fault *vmf);
 840vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
 841#endif
 842
 843/*
 844 * Multiple processes may "see" the same page. E.g. for untouched
 845 * mappings of /dev/null, all processes see the same page full of
 846 * zeroes, and text pages of executables and shared libraries have
 847 * only one copy in memory, at most, normally.
 848 *
 849 * For the non-reserved pages, page_count(page) denotes a reference count.
 850 *   page_count() == 0 means the page is free. page->lru is then used for
 851 *   freelist management in the buddy allocator.
 852 *   page_count() > 0  means the page has been allocated.
 853 *
 854 * Pages are allocated by the slab allocator in order to provide memory
 855 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 856 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 857 * unless a particular usage is carefully commented. (the responsibility of
 858 * freeing the kmalloc memory is the caller's, of course).
 859 *
 860 * A page may be used by anyone else who does a __get_free_page().
 861 * In this case, page_count still tracks the references, and should only
 862 * be used through the normal accessor functions. The top bits of page->flags
 863 * and page->virtual store page management information, but all other fields
 864 * are unused and could be used privately, carefully. The management of this
 865 * page is the responsibility of the one who allocated it, and those who have
 866 * subsequently been given references to it.
 867 *
 868 * The other pages (we may call them "pagecache pages") are completely
 869 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 870 * The following discussion applies only to them.
 871 *
 872 * A pagecache page contains an opaque `private' member, which belongs to the
 873 * page's address_space. Usually, this is the address of a circular list of
 874 * the page's disk buffers. PG_private must be set to tell the VM to call
 875 * into the filesystem to release these pages.
 876 *
 877 * A page may belong to an inode's memory mapping. In this case, page->mapping
 878 * is the pointer to the inode, and page->index is the file offset of the page,
 879 * in units of PAGE_SIZE.
 880 *
 881 * If pagecache pages are not associated with an inode, they are said to be
 882 * anonymous pages. These may become associated with the swapcache, and in that
 883 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 884 *
 885 * In either case (swapcache or inode backed), the pagecache itself holds one
 886 * reference to the page. Setting PG_private should also increment the
 887 * refcount. The each user mapping also has a reference to the page.
 888 *
 889 * The pagecache pages are stored in a per-mapping radix tree, which is
 890 * rooted at mapping->i_pages, and indexed by offset.
 891 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 892 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 893 *
 894 * All pagecache pages may be subject to I/O:
 895 * - inode pages may need to be read from disk,
 896 * - inode pages which have been modified and are MAP_SHARED may need
 897 *   to be written back to the inode on disk,
 898 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 899 *   modified may need to be swapped out to swap space and (later) to be read
 900 *   back into memory.
 901 */
 902
 903/*
 904 * The zone field is never updated after free_area_init_core()
 905 * sets it, so none of the operations on it need to be atomic.
 906 */
 907
 908/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
 909#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 910#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 911#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 912#define LAST_CPUPID_PGOFF       (ZONES_PGOFF - LAST_CPUPID_WIDTH)
 913#define KASAN_TAG_PGOFF         (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
 914
 915/*
 916 * Define the bit shifts to access each section.  For non-existent
 917 * sections we define the shift as 0; that plus a 0 mask ensures
 918 * the compiler will optimise away reference to them.
 919 */
 920#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 921#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 922#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 923#define LAST_CPUPID_PGSHIFT     (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
 924#define KASAN_TAG_PGSHIFT       (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
 925
 926/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
 927#ifdef NODE_NOT_IN_PAGE_FLAGS
 928#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 929#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 930                                                SECTIONS_PGOFF : ZONES_PGOFF)
 931#else
 932#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 933#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 934                                                NODES_PGOFF : ZONES_PGOFF)
 935#endif
 936
 937#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 938
 939#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 940#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 941#endif
 942
 943#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 944#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 945#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 946#define LAST_CPUPID_MASK        ((1UL << LAST_CPUPID_SHIFT) - 1)
 947#define KASAN_TAG_MASK          ((1UL << KASAN_TAG_WIDTH) - 1)
 948#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 949
 950static inline enum zone_type page_zonenum(const struct page *page)
 951{
 952        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 953}
 954
 955#ifdef CONFIG_ZONE_DEVICE
 956static inline bool is_zone_device_page(const struct page *page)
 957{
 958        return page_zonenum(page) == ZONE_DEVICE;
 959}
 960extern void memmap_init_zone_device(struct zone *, unsigned long,
 961                                    unsigned long, struct dev_pagemap *);
 962#else
 963static inline bool is_zone_device_page(const struct page *page)
 964{
 965        return false;
 966}
 967#endif
 968
 969#ifdef CONFIG_DEV_PAGEMAP_OPS
 970void __put_devmap_managed_page(struct page *page);
 971DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
 972static inline bool put_devmap_managed_page(struct page *page)
 973{
 974        if (!static_branch_unlikely(&devmap_managed_key))
 975                return false;
 976        if (!is_zone_device_page(page))
 977                return false;
 978        switch (page->pgmap->type) {
 979        case MEMORY_DEVICE_PRIVATE:
 980        case MEMORY_DEVICE_FS_DAX:
 981                __put_devmap_managed_page(page);
 982                return true;
 983        default:
 984                break;
 985        }
 986        return false;
 987}
 988
 989#else /* CONFIG_DEV_PAGEMAP_OPS */
 990static inline bool put_devmap_managed_page(struct page *page)
 991{
 992        return false;
 993}
 994#endif /* CONFIG_DEV_PAGEMAP_OPS */
 995
 996static inline bool is_device_private_page(const struct page *page)
 997{
 998        return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
 999                IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1000                is_zone_device_page(page) &&
1001                page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1002}
1003
1004static inline bool is_pci_p2pdma_page(const struct page *page)
1005{
1006        return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1007                IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1008                is_zone_device_page(page) &&
1009                page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1010}
1011
1012/* 127: arbitrary random number, small enough to assemble well */
1013#define page_ref_zero_or_close_to_overflow(page) \
1014        ((unsigned int) page_ref_count(page) + 127u <= 127u)
1015
1016static inline void get_page(struct page *page)
1017{
1018        page = compound_head(page);
1019        /*
1020         * Getting a normal page or the head of a compound page
1021         * requires to already have an elevated page->_refcount.
1022         */
1023        VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1024        page_ref_inc(page);
1025}
1026
1027static inline __must_check bool try_get_page(struct page *page)
1028{
1029        page = compound_head(page);
1030        if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1031                return false;
1032        page_ref_inc(page);
1033        return true;
1034}
1035
1036static inline void put_page(struct page *page)
1037{
1038        page = compound_head(page);
1039
1040        /*
1041         * For devmap managed pages we need to catch refcount transition from
1042         * 2 to 1, when refcount reach one it means the page is free and we
1043         * need to inform the device driver through callback. See
1044         * include/linux/memremap.h and HMM for details.
1045         */
1046        if (put_devmap_managed_page(page))
1047                return;
1048
1049        if (put_page_testzero(page))
1050                __put_page(page);
1051}
1052
1053/**
1054 * put_user_page() - release a gup-pinned page
1055 * @page:            pointer to page to be released
1056 *
1057 * Pages that were pinned via get_user_pages*() must be released via
1058 * either put_user_page(), or one of the put_user_pages*() routines
1059 * below. This is so that eventually, pages that are pinned via
1060 * get_user_pages*() can be separately tracked and uniquely handled. In
1061 * particular, interactions with RDMA and filesystems need special
1062 * handling.
1063 *
1064 * put_user_page() and put_page() are not interchangeable, despite this early
1065 * implementation that makes them look the same. put_user_page() calls must
1066 * be perfectly matched up with get_user_page() calls.
1067 */
1068static inline void put_user_page(struct page *page)
1069{
1070        put_page(page);
1071}
1072
1073void put_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1074                               bool make_dirty);
1075
1076void put_user_pages(struct page **pages, unsigned long npages);
1077
1078#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1079#define SECTION_IN_PAGE_FLAGS
1080#endif
1081
1082/*
1083 * The identification function is mainly used by the buddy allocator for
1084 * determining if two pages could be buddies. We are not really identifying
1085 * the zone since we could be using the section number id if we do not have
1086 * node id available in page flags.
1087 * We only guarantee that it will return the same value for two combinable
1088 * pages in a zone.
1089 */
1090static inline int page_zone_id(struct page *page)
1091{
1092        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1093}
1094
1095#ifdef NODE_NOT_IN_PAGE_FLAGS
1096extern int page_to_nid(const struct page *page);
1097#else
1098static inline int page_to_nid(const struct page *page)
1099{
1100        struct page *p = (struct page *)page;
1101
1102        return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1103}
1104#endif
1105
1106#ifdef CONFIG_NUMA_BALANCING
1107static inline int cpu_pid_to_cpupid(int cpu, int pid)
1108{
1109        return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1110}
1111
1112static inline int cpupid_to_pid(int cpupid)
1113{
1114        return cpupid & LAST__PID_MASK;
1115}
1116
1117static inline int cpupid_to_cpu(int cpupid)
1118{
1119        return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1120}
1121
1122static inline int cpupid_to_nid(int cpupid)
1123{
1124        return cpu_to_node(cpupid_to_cpu(cpupid));
1125}
1126
1127static inline bool cpupid_pid_unset(int cpupid)
1128{
1129        return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1130}
1131
1132static inline bool cpupid_cpu_unset(int cpupid)
1133{
1134        return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1135}
1136
1137static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1138{
1139        return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1140}
1141
1142#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1143#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1144static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1145{
1146        return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1147}
1148
1149static inline int page_cpupid_last(struct page *page)
1150{
1151        return page->_last_cpupid;
1152}
1153static inline void page_cpupid_reset_last(struct page *page)
1154{
1155        page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1156}
1157#else
1158static inline int page_cpupid_last(struct page *page)
1159{
1160        return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1161}
1162
1163extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1164
1165static inline void page_cpupid_reset_last(struct page *page)
1166{
1167        page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1168}
1169#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1170#else /* !CONFIG_NUMA_BALANCING */
1171static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1172{
1173        return page_to_nid(page); /* XXX */
1174}
1175
1176static inline int page_cpupid_last(struct page *page)
1177{
1178        return page_to_nid(page); /* XXX */
1179}
1180
1181static inline int cpupid_to_nid(int cpupid)
1182{
1183        return -1;
1184}
1185
1186static inline int cpupid_to_pid(int cpupid)
1187{
1188        return -1;
1189}
1190
1191static inline int cpupid_to_cpu(int cpupid)
1192{
1193        return -1;
1194}
1195
1196static inline int cpu_pid_to_cpupid(int nid, int pid)
1197{
1198        return -1;
1199}
1200
1201static inline bool cpupid_pid_unset(int cpupid)
1202{
1203        return 1;
1204}
1205
1206static inline void page_cpupid_reset_last(struct page *page)
1207{
1208}
1209
1210static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1211{
1212        return false;
1213}
1214#endif /* CONFIG_NUMA_BALANCING */
1215
1216#ifdef CONFIG_KASAN_SW_TAGS
1217static inline u8 page_kasan_tag(const struct page *page)
1218{
1219        return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1220}
1221
1222static inline void page_kasan_tag_set(struct page *page, u8 tag)
1223{
1224        page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1225        page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1226}
1227
1228static inline void page_kasan_tag_reset(struct page *page)
1229{
1230        page_kasan_tag_set(page, 0xff);
1231}
1232#else
1233static inline u8 page_kasan_tag(const struct page *page)
1234{
1235        return 0xff;
1236}
1237
1238static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1239static inline void page_kasan_tag_reset(struct page *page) { }
1240#endif
1241
1242static inline struct zone *page_zone(const struct page *page)
1243{
1244        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1245}
1246
1247static inline pg_data_t *page_pgdat(const struct page *page)
1248{
1249        return NODE_DATA(page_to_nid(page));
1250}
1251
1252#ifdef SECTION_IN_PAGE_FLAGS
1253static inline void set_page_section(struct page *page, unsigned long section)
1254{
1255        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1256        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1257}
1258
1259static inline unsigned long page_to_section(const struct page *page)
1260{
1261        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1262}
1263#endif
1264
1265static inline void set_page_zone(struct page *page, enum zone_type zone)
1266{
1267        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1268        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1269}
1270
1271static inline void set_page_node(struct page *page, unsigned long node)
1272{
1273        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1274        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1275}
1276
1277static inline void set_page_links(struct page *page, enum zone_type zone,
1278        unsigned long node, unsigned long pfn)
1279{
1280        set_page_zone(page, zone);
1281        set_page_node(page, node);
1282#ifdef SECTION_IN_PAGE_FLAGS
1283        set_page_section(page, pfn_to_section_nr(pfn));
1284#endif
1285}
1286
1287#ifdef CONFIG_MEMCG
1288static inline struct mem_cgroup *page_memcg(struct page *page)
1289{
1290        return page->mem_cgroup;
1291}
1292static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1293{
1294        WARN_ON_ONCE(!rcu_read_lock_held());
1295        return READ_ONCE(page->mem_cgroup);
1296}
1297#else
1298static inline struct mem_cgroup *page_memcg(struct page *page)
1299{
1300        return NULL;
1301}
1302static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1303{
1304        WARN_ON_ONCE(!rcu_read_lock_held());
1305        return NULL;
1306}
1307#endif
1308
1309/*
1310 * Some inline functions in vmstat.h depend on page_zone()
1311 */
1312#include <linux/vmstat.h>
1313
1314static __always_inline void *lowmem_page_address(const struct page *page)
1315{
1316        return page_to_virt(page);
1317}
1318
1319#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1320#define HASHED_PAGE_VIRTUAL
1321#endif
1322
1323#if defined(WANT_PAGE_VIRTUAL)
1324static inline void *page_address(const struct page *page)
1325{
1326        return page->virtual;
1327}
1328static inline void set_page_address(struct page *page, void *address)
1329{
1330        page->virtual = address;
1331}
1332#define page_address_init()  do { } while(0)
1333#endif
1334
1335#if defined(HASHED_PAGE_VIRTUAL)
1336void *page_address(const struct page *page);
1337void set_page_address(struct page *page, void *virtual);
1338void page_address_init(void);
1339#endif
1340
1341#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1342#define page_address(page) lowmem_page_address(page)
1343#define set_page_address(page, address)  do { } while(0)
1344#define page_address_init()  do { } while(0)
1345#endif
1346
1347extern void *page_rmapping(struct page *page);
1348extern struct anon_vma *page_anon_vma(struct page *page);
1349extern struct address_space *page_mapping(struct page *page);
1350
1351extern struct address_space *__page_file_mapping(struct page *);
1352
1353static inline
1354struct address_space *page_file_mapping(struct page *page)
1355{
1356        if (unlikely(PageSwapCache(page)))
1357                return __page_file_mapping(page);
1358
1359        return page->mapping;
1360}
1361
1362extern pgoff_t __page_file_index(struct page *page);
1363
1364/*
1365 * Return the pagecache index of the passed page.  Regular pagecache pages
1366 * use ->index whereas swapcache pages use swp_offset(->private)
1367 */
1368static inline pgoff_t page_index(struct page *page)
1369{
1370        if (unlikely(PageSwapCache(page)))
1371                return __page_file_index(page);
1372        return page->index;
1373}
1374
1375bool page_mapped(struct page *page);
1376struct address_space *page_mapping(struct page *page);
1377struct address_space *page_mapping_file(struct page *page);
1378
1379/*
1380 * Return true only if the page has been allocated with
1381 * ALLOC_NO_WATERMARKS and the low watermark was not
1382 * met implying that the system is under some pressure.
1383 */
1384static inline bool page_is_pfmemalloc(struct page *page)
1385{
1386        /*
1387         * Page index cannot be this large so this must be
1388         * a pfmemalloc page.
1389         */
1390        return page->index == -1UL;
1391}
1392
1393/*
1394 * Only to be called by the page allocator on a freshly allocated
1395 * page.
1396 */
1397static inline void set_page_pfmemalloc(struct page *page)
1398{
1399        page->index = -1UL;
1400}
1401
1402static inline void clear_page_pfmemalloc(struct page *page)
1403{
1404        page->index = 0;
1405}
1406
1407/*
1408 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1409 */
1410extern void pagefault_out_of_memory(void);
1411
1412#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
1413
1414/*
1415 * Flags passed to show_mem() and show_free_areas() to suppress output in
1416 * various contexts.
1417 */
1418#define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
1419
1420extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1421
1422#ifdef CONFIG_MMU
1423extern bool can_do_mlock(void);
1424#else
1425static inline bool can_do_mlock(void) { return false; }
1426#endif
1427extern int user_shm_lock(size_t, struct user_struct *);
1428extern void user_shm_unlock(size_t, struct user_struct *);
1429
1430/*
1431 * Parameter block passed down to zap_pte_range in exceptional cases.
1432 */
1433struct zap_details {
1434        struct address_space *check_mapping;    /* Check page->mapping if set */
1435        pgoff_t first_index;                    /* Lowest page->index to unmap */
1436        pgoff_t last_index;                     /* Highest page->index to unmap */
1437};
1438
1439struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1440                             pte_t pte);
1441struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1442                                pmd_t pmd);
1443
1444void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1445                  unsigned long size);
1446void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1447                    unsigned long size);
1448void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1449                unsigned long start, unsigned long end);
1450
1451struct mmu_notifier_range;
1452
1453void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1454                unsigned long end, unsigned long floor, unsigned long ceiling);
1455int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1456                        struct vm_area_struct *vma);
1457int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1458                   struct mmu_notifier_range *range,
1459                   pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1460int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1461        unsigned long *pfn);
1462int follow_phys(struct vm_area_struct *vma, unsigned long address,
1463                unsigned int flags, unsigned long *prot, resource_size_t *phys);
1464int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1465                        void *buf, int len, int write);
1466
1467extern void truncate_pagecache(struct inode *inode, loff_t new);
1468extern void truncate_setsize(struct inode *inode, loff_t newsize);
1469void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1470void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1471int truncate_inode_page(struct address_space *mapping, struct page *page);
1472int generic_error_remove_page(struct address_space *mapping, struct page *page);
1473int invalidate_inode_page(struct page *page);
1474
1475#ifdef CONFIG_MMU
1476extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1477                        unsigned long address, unsigned int flags);
1478extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1479                            unsigned long address, unsigned int fault_flags,
1480                            bool *unlocked);
1481void unmap_mapping_pages(struct address_space *mapping,
1482                pgoff_t start, pgoff_t nr, bool even_cows);
1483void unmap_mapping_range(struct address_space *mapping,
1484                loff_t const holebegin, loff_t const holelen, int even_cows);
1485#else
1486static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1487                unsigned long address, unsigned int flags)
1488{
1489        /* should never happen if there's no MMU */
1490        BUG();
1491        return VM_FAULT_SIGBUS;
1492}
1493static inline int fixup_user_fault(struct task_struct *tsk,
1494                struct mm_struct *mm, unsigned long address,
1495                unsigned int fault_flags, bool *unlocked)
1496{
1497        /* should never happen if there's no MMU */
1498        BUG();
1499        return -EFAULT;
1500}
1501static inline void unmap_mapping_pages(struct address_space *mapping,
1502                pgoff_t start, pgoff_t nr, bool even_cows) { }
1503static inline void unmap_mapping_range(struct address_space *mapping,
1504                loff_t const holebegin, loff_t const holelen, int even_cows) { }
1505#endif
1506
1507static inline void unmap_shared_mapping_range(struct address_space *mapping,
1508                loff_t const holebegin, loff_t const holelen)
1509{
1510        unmap_mapping_range(mapping, holebegin, holelen, 0);
1511}
1512
1513extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1514                void *buf, int len, unsigned int gup_flags);
1515extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1516                void *buf, int len, unsigned int gup_flags);
1517extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1518                unsigned long addr, void *buf, int len, unsigned int gup_flags);
1519
1520long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1521                            unsigned long start, unsigned long nr_pages,
1522                            unsigned int gup_flags, struct page **pages,
1523                            struct vm_area_struct **vmas, int *locked);
1524long get_user_pages(unsigned long start, unsigned long nr_pages,
1525                            unsigned int gup_flags, struct page **pages,
1526                            struct vm_area_struct **vmas);
1527long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1528                    unsigned int gup_flags, struct page **pages, int *locked);
1529long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1530                    struct page **pages, unsigned int gup_flags);
1531
1532int get_user_pages_fast(unsigned long start, int nr_pages,
1533                        unsigned int gup_flags, struct page **pages);
1534
1535int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1536int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1537                        struct task_struct *task, bool bypass_rlim);
1538
1539/* Container for pinned pfns / pages */
1540struct frame_vector {
1541        unsigned int nr_allocated;      /* Number of frames we have space for */
1542        unsigned int nr_frames; /* Number of frames stored in ptrs array */
1543        bool got_ref;           /* Did we pin pages by getting page ref? */
1544        bool is_pfns;           /* Does array contain pages or pfns? */
1545        void *ptrs[0];          /* Array of pinned pfns / pages. Use
1546                                 * pfns_vector_pages() or pfns_vector_pfns()
1547                                 * for access */
1548};
1549
1550struct frame_vector *frame_vector_create(unsigned int nr_frames);
1551void frame_vector_destroy(struct frame_vector *vec);
1552int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1553                     unsigned int gup_flags, struct frame_vector *vec);
1554void put_vaddr_frames(struct frame_vector *vec);
1555int frame_vector_to_pages(struct frame_vector *vec);
1556void frame_vector_to_pfns(struct frame_vector *vec);
1557
1558static inline unsigned int frame_vector_count(struct frame_vector *vec)
1559{
1560        return vec->nr_frames;
1561}
1562
1563static inline struct page **frame_vector_pages(struct frame_vector *vec)
1564{
1565        if (vec->is_pfns) {
1566                int err = frame_vector_to_pages(vec);
1567
1568                if (err)
1569                        return ERR_PTR(err);
1570        }
1571        return (struct page **)(vec->ptrs);
1572}
1573
1574static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1575{
1576        if (!vec->is_pfns)
1577                frame_vector_to_pfns(vec);
1578        return (unsigned long *)(vec->ptrs);
1579}
1580
1581struct kvec;
1582int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1583                        struct page **pages);
1584int get_kernel_page(unsigned long start, int write, struct page **pages);
1585struct page *get_dump_page(unsigned long addr);
1586
1587extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1588extern void do_invalidatepage(struct page *page, unsigned int offset,
1589                              unsigned int length);
1590
1591void __set_page_dirty(struct page *, struct address_space *, int warn);
1592int __set_page_dirty_nobuffers(struct page *page);
1593int __set_page_dirty_no_writeback(struct page *page);
1594int redirty_page_for_writepage(struct writeback_control *wbc,
1595                                struct page *page);
1596void account_page_dirtied(struct page *page, struct address_space *mapping);
1597void account_page_cleaned(struct page *page, struct address_space *mapping,
1598                          struct bdi_writeback *wb);
1599int set_page_dirty(struct page *page);
1600int set_page_dirty_lock(struct page *page);
1601void __cancel_dirty_page(struct page *page);
1602static inline void cancel_dirty_page(struct page *page)
1603{
1604        /* Avoid atomic ops, locking, etc. when not actually needed. */
1605        if (PageDirty(page))
1606                __cancel_dirty_page(page);
1607}
1608int clear_page_dirty_for_io(struct page *page);
1609
1610int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1611
1612extern unsigned long move_page_tables(struct vm_area_struct *vma,
1613                unsigned long old_addr, struct vm_area_struct *new_vma,
1614                unsigned long new_addr, unsigned long len,
1615                bool need_rmap_locks);
1616extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1617                              unsigned long end, pgprot_t newprot,
1618                              int dirty_accountable, int prot_numa);
1619extern int mprotect_fixup(struct vm_area_struct *vma,
1620                          struct vm_area_struct **pprev, unsigned long start,
1621                          unsigned long end, unsigned long newflags);
1622
1623/*
1624 * doesn't attempt to fault and will return short.
1625 */
1626int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1627                          struct page **pages);
1628/*
1629 * per-process(per-mm_struct) statistics.
1630 */
1631static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1632{
1633        long val = atomic_long_read(&mm->rss_stat.count[member]);
1634
1635#ifdef SPLIT_RSS_COUNTING
1636        /*
1637         * counter is updated in asynchronous manner and may go to minus.
1638         * But it's never be expected number for users.
1639         */
1640        if (val < 0)
1641                val = 0;
1642#endif
1643        return (unsigned long)val;
1644}
1645
1646static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1647{
1648        atomic_long_add(value, &mm->rss_stat.count[member]);
1649}
1650
1651static inline void inc_mm_counter(struct mm_struct *mm, int member)
1652{
1653        atomic_long_inc(&mm->rss_stat.count[member]);
1654}
1655
1656static inline void dec_mm_counter(struct mm_struct *mm, int member)
1657{
1658        atomic_long_dec(&mm->rss_stat.count[member]);
1659}
1660
1661/* Optimized variant when page is already known not to be PageAnon */
1662static inline int mm_counter_file(struct page *page)
1663{
1664        if (PageSwapBacked(page))
1665                return MM_SHMEMPAGES;
1666        return MM_FILEPAGES;
1667}
1668
1669static inline int mm_counter(struct page *page)
1670{
1671        if (PageAnon(page))
1672                return MM_ANONPAGES;
1673        return mm_counter_file(page);
1674}
1675
1676static inline unsigned long get_mm_rss(struct mm_struct *mm)
1677{
1678        return get_mm_counter(mm, MM_FILEPAGES) +
1679                get_mm_counter(mm, MM_ANONPAGES) +
1680                get_mm_counter(mm, MM_SHMEMPAGES);
1681}
1682
1683static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1684{
1685        return max(mm->hiwater_rss, get_mm_rss(mm));
1686}
1687
1688static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1689{
1690        return max(mm->hiwater_vm, mm->total_vm);
1691}
1692
1693static inline void update_hiwater_rss(struct mm_struct *mm)
1694{
1695        unsigned long _rss = get_mm_rss(mm);
1696
1697        if ((mm)->hiwater_rss < _rss)
1698                (mm)->hiwater_rss = _rss;
1699}
1700
1701static inline void update_hiwater_vm(struct mm_struct *mm)
1702{
1703        if (mm->hiwater_vm < mm->total_vm)
1704                mm->hiwater_vm = mm->total_vm;
1705}
1706
1707static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1708{
1709        mm->hiwater_rss = get_mm_rss(mm);
1710}
1711
1712static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1713                                         struct mm_struct *mm)
1714{
1715        unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1716
1717        if (*maxrss < hiwater_rss)
1718                *maxrss = hiwater_rss;
1719}
1720
1721#if defined(SPLIT_RSS_COUNTING)
1722void sync_mm_rss(struct mm_struct *mm);
1723#else
1724static inline void sync_mm_rss(struct mm_struct *mm)
1725{
1726}
1727#endif
1728
1729#ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1730static inline int pte_devmap(pte_t pte)
1731{
1732        return 0;
1733}
1734#endif
1735
1736int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1737
1738extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1739                               spinlock_t **ptl);
1740static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1741                                    spinlock_t **ptl)
1742{
1743        pte_t *ptep;
1744        __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1745        return ptep;
1746}
1747
1748#ifdef __PAGETABLE_P4D_FOLDED
1749static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1750                                                unsigned long address)
1751{
1752        return 0;
1753}
1754#else
1755int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1756#endif
1757
1758#if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1759static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1760                                                unsigned long address)
1761{
1762        return 0;
1763}
1764static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1765static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1766
1767#else
1768int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1769
1770static inline void mm_inc_nr_puds(struct mm_struct *mm)
1771{
1772        if (mm_pud_folded(mm))
1773                return;
1774        atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1775}
1776
1777static inline void mm_dec_nr_puds(struct mm_struct *mm)
1778{
1779        if (mm_pud_folded(mm))
1780                return;
1781        atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1782}
1783#endif
1784
1785#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1786static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1787                                                unsigned long address)
1788{
1789        return 0;
1790}
1791
1792static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1793static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1794
1795#else
1796int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1797
1798static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1799{
1800        if (mm_pmd_folded(mm))
1801                return;
1802        atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1803}
1804
1805static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1806{
1807        if (mm_pmd_folded(mm))
1808                return;
1809        atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1810}
1811#endif
1812
1813#ifdef CONFIG_MMU
1814static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1815{
1816        atomic_long_set(&mm->pgtables_bytes, 0);
1817}
1818
1819static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1820{
1821        return atomic_long_read(&mm->pgtables_bytes);
1822}
1823
1824static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1825{
1826        atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1827}
1828
1829static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1830{
1831        atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1832}
1833#else
1834
1835static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1836static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1837{
1838        return 0;
1839}
1840
1841static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1842static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1843#endif
1844
1845int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1846int __pte_alloc_kernel(pmd_t *pmd);
1847
1848/*
1849 * The following ifdef needed to get the 4level-fixup.h header to work.
1850 * Remove it when 4level-fixup.h has been removed.
1851 */
1852#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1853
1854#ifndef __ARCH_HAS_5LEVEL_HACK
1855static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1856                unsigned long address)
1857{
1858        return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1859                NULL : p4d_offset(pgd, address);
1860}
1861
1862static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1863                unsigned long address)
1864{
1865        return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1866                NULL : pud_offset(p4d, address);
1867}
1868#endif /* !__ARCH_HAS_5LEVEL_HACK */
1869
1870static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1871{
1872        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1873                NULL: pmd_offset(pud, address);
1874}
1875#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1876
1877#if USE_SPLIT_PTE_PTLOCKS
1878#if ALLOC_SPLIT_PTLOCKS
1879void __init ptlock_cache_init(void);
1880extern bool ptlock_alloc(struct page *page);
1881extern void ptlock_free(struct page *page);
1882
1883static inline spinlock_t *ptlock_ptr(struct page *page)
1884{
1885        return page->ptl;
1886}
1887#else /* ALLOC_SPLIT_PTLOCKS */
1888static inline void ptlock_cache_init(void)
1889{
1890}
1891
1892static inline bool ptlock_alloc(struct page *page)
1893{
1894        return true;
1895}
1896
1897static inline void ptlock_free(struct page *page)
1898{
1899}
1900
1901static inline spinlock_t *ptlock_ptr(struct page *page)
1902{
1903        return &page->ptl;
1904}
1905#endif /* ALLOC_SPLIT_PTLOCKS */
1906
1907static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1908{
1909        return ptlock_ptr(pmd_page(*pmd));
1910}
1911
1912static inline bool ptlock_init(struct page *page)
1913{
1914        /*
1915         * prep_new_page() initialize page->private (and therefore page->ptl)
1916         * with 0. Make sure nobody took it in use in between.
1917         *
1918         * It can happen if arch try to use slab for page table allocation:
1919         * slab code uses page->slab_cache, which share storage with page->ptl.
1920         */
1921        VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1922        if (!ptlock_alloc(page))
1923                return false;
1924        spin_lock_init(ptlock_ptr(page));
1925        return true;
1926}
1927
1928#else   /* !USE_SPLIT_PTE_PTLOCKS */
1929/*
1930 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1931 */
1932static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1933{
1934        return &mm->page_table_lock;
1935}
1936static inline void ptlock_cache_init(void) {}
1937static inline bool ptlock_init(struct page *page) { return true; }
1938static inline void ptlock_free(struct page *page) {}
1939#endif /* USE_SPLIT_PTE_PTLOCKS */
1940
1941static inline void pgtable_init(void)
1942{
1943        ptlock_cache_init();
1944        pgtable_cache_init();
1945}
1946
1947static inline bool pgtable_pte_page_ctor(struct page *page)
1948{
1949        if (!ptlock_init(page))
1950                return false;
1951        __SetPageTable(page);
1952        inc_zone_page_state(page, NR_PAGETABLE);
1953        return true;
1954}
1955
1956static inline void pgtable_pte_page_dtor(struct page *page)
1957{
1958        ptlock_free(page);
1959        __ClearPageTable(page);
1960        dec_zone_page_state(page, NR_PAGETABLE);
1961}
1962
1963#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1964({                                                      \
1965        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1966        pte_t *__pte = pte_offset_map(pmd, address);    \
1967        *(ptlp) = __ptl;                                \
1968        spin_lock(__ptl);                               \
1969        __pte;                                          \
1970})
1971
1972#define pte_unmap_unlock(pte, ptl)      do {            \
1973        spin_unlock(ptl);                               \
1974        pte_unmap(pte);                                 \
1975} while (0)
1976
1977#define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
1978
1979#define pte_alloc_map(mm, pmd, address)                 \
1980        (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
1981
1982#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1983        (pte_alloc(mm, pmd) ?                   \
1984                 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1985
1986#define pte_alloc_kernel(pmd, address)                  \
1987        ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
1988                NULL: pte_offset_kernel(pmd, address))
1989
1990#if USE_SPLIT_PMD_PTLOCKS
1991
1992static struct page *pmd_to_page(pmd_t *pmd)
1993{
1994        unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1995        return virt_to_page((void *)((unsigned long) pmd & mask));
1996}
1997
1998static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1999{
2000        return ptlock_ptr(pmd_to_page(pmd));
2001}
2002
2003static inline bool pgtable_pmd_page_ctor(struct page *page)
2004{
2005#ifdef CONFIG_TRANSPARENT_HUGEPAGE
2006        page->pmd_huge_pte = NULL;
2007#endif
2008        return ptlock_init(page);
2009}
2010
2011static inline void pgtable_pmd_page_dtor(struct page *page)
2012{
2013#ifdef CONFIG_TRANSPARENT_HUGEPAGE
2014        VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2015#endif
2016        ptlock_free(page);
2017}
2018
2019#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2020
2021#else
2022
2023static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2024{
2025        return &mm->page_table_lock;
2026}
2027
2028static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2029static inline void pgtable_pmd_page_dtor(struct page *page) {}
2030
2031#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2032
2033#endif
2034
2035static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2036{
2037        spinlock_t *ptl = pmd_lockptr(mm, pmd);
2038        spin_lock(ptl);
2039        return ptl;
2040}
2041
2042/*
2043 * No scalability reason to split PUD locks yet, but follow the same pattern
2044 * as the PMD locks to make it easier if we decide to.  The VM should not be
2045 * considered ready to switch to split PUD locks yet; there may be places
2046 * which need to be converted from page_table_lock.
2047 */
2048static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2049{
2050        return &mm->page_table_lock;
2051}
2052
2053static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2054{
2055        spinlock_t *ptl = pud_lockptr(mm, pud);
2056
2057        spin_lock(ptl);
2058        return ptl;
2059}
2060
2061extern void __init pagecache_init(void);
2062extern void free_area_init(unsigned long * zones_size);
2063extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2064                unsigned long zone_start_pfn, unsigned long *zholes_size);
2065extern void free_initmem(void);
2066
2067/*
2068 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2069 * into the buddy system. The freed pages will be poisoned with pattern
2070 * "poison" if it's within range [0, UCHAR_MAX].
2071 * Return pages freed into the buddy system.
2072 */
2073extern unsigned long free_reserved_area(void *start, void *end,
2074                                        int poison, const char *s);
2075
2076#ifdef  CONFIG_HIGHMEM
2077/*
2078 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2079 * and totalram_pages.
2080 */
2081extern void free_highmem_page(struct page *page);
2082#endif
2083
2084extern void adjust_managed_page_count(struct page *page, long count);
2085extern void mem_init_print_info(const char *str);
2086
2087extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2088
2089/* Free the reserved page into the buddy system, so it gets managed. */
2090static inline void __free_reserved_page(struct page *page)
2091{
2092        ClearPageReserved(page);
2093        init_page_count(page);
2094        __free_page(page);
2095}
2096
2097static inline void free_reserved_page(struct page *page)
2098{
2099        __free_reserved_page(page);
2100        adjust_managed_page_count(page, 1);
2101}
2102
2103static inline void mark_page_reserved(struct page *page)
2104{
2105        SetPageReserved(page);
2106        adjust_managed_page_count(page, -1);
2107}
2108
2109/*
2110 * Default method to free all the __init memory into the buddy system.
2111 * The freed pages will be poisoned with pattern "poison" if it's within
2112 * range [0, UCHAR_MAX].
2113 * Return pages freed into the buddy system.
2114 */
2115static inline unsigned long free_initmem_default(int poison)
2116{
2117        extern char __init_begin[], __init_end[];
2118
2119        return free_reserved_area(&__init_begin, &__init_end,
2120                                  poison, "unused kernel");
2121}
2122
2123static inline unsigned long get_num_physpages(void)
2124{
2125        int nid;
2126        unsigned long phys_pages = 0;
2127
2128        for_each_online_node(nid)
2129                phys_pages += node_present_pages(nid);
2130
2131        return phys_pages;
2132}
2133
2134#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2135/*
2136 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2137 * zones, allocate the backing mem_map and account for memory holes in a more
2138 * architecture independent manner. This is a substitute for creating the
2139 * zone_sizes[] and zholes_size[] arrays and passing them to
2140 * free_area_init_node()
2141 *
2142 * An architecture is expected to register range of page frames backed by
2143 * physical memory with memblock_add[_node]() before calling
2144 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2145 * usage, an architecture is expected to do something like
2146 *
2147 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2148 *                                                       max_highmem_pfn};
2149 * for_each_valid_physical_page_range()
2150 *      memblock_add_node(base, size, nid)
2151 * free_area_init_nodes(max_zone_pfns);
2152 *
2153 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2154 * registered physical page range.  Similarly
2155 * sparse_memory_present_with_active_regions() calls memory_present() for
2156 * each range when SPARSEMEM is enabled.
2157 *
2158 * See mm/page_alloc.c for more information on each function exposed by
2159 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2160 */
2161extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2162unsigned long node_map_pfn_alignment(void);
2163unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2164                                                unsigned long end_pfn);
2165extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2166                                                unsigned long end_pfn);
2167extern void get_pfn_range_for_nid(unsigned int nid,
2168                        unsigned long *start_pfn, unsigned long *end_pfn);
2169extern unsigned long find_min_pfn_with_active_regions(void);
2170extern void free_bootmem_with_active_regions(int nid,
2171                                                unsigned long max_low_pfn);
2172extern void sparse_memory_present_with_active_regions(int nid);
2173
2174#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2175
2176#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2177    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2178static inline int __early_pfn_to_nid(unsigned long pfn,
2179                                        struct mminit_pfnnid_cache *state)
2180{
2181        return 0;
2182}
2183#else
2184/* please see mm/page_alloc.c */
2185extern int __meminit early_pfn_to_nid(unsigned long pfn);
2186/* there is a per-arch backend function. */
2187extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2188                                        struct mminit_pfnnid_cache *state);
2189#endif
2190
2191#if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2192void zero_resv_unavail(void);
2193#else
2194static inline void zero_resv_unavail(void) {}
2195#endif
2196
2197extern void set_dma_reserve(unsigned long new_dma_reserve);
2198extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2199                enum memmap_context, struct vmem_altmap *);
2200extern void setup_per_zone_wmarks(void);
2201extern int __meminit init_per_zone_wmark_min(void);
2202extern void mem_init(void);
2203extern void __init mmap_init(void);
2204extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2205extern long si_mem_available(void);
2206extern void si_meminfo(struct sysinfo * val);
2207extern void si_meminfo_node(struct sysinfo *val, int nid);
2208#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2209extern unsigned long arch_reserved_kernel_pages(void);
2210#endif
2211
2212extern __printf(3, 4)
2213void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2214
2215extern void setup_per_cpu_pageset(void);
2216
2217extern void zone_pcp_update(struct zone *zone);
2218extern void zone_pcp_reset(struct zone *zone);
2219
2220/* page_alloc.c */
2221extern int min_free_kbytes;
2222extern int watermark_boost_factor;
2223extern int watermark_scale_factor;
2224
2225/* nommu.c */
2226extern atomic_long_t mmap_pages_allocated;
2227extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2228
2229/* interval_tree.c */
2230void vma_interval_tree_insert(struct vm_area_struct *node,
2231                              struct rb_root_cached *root);
2232void vma_interval_tree_insert_after(struct vm_area_struct *node,
2233                                    struct vm_area_struct *prev,
2234                                    struct rb_root_cached *root);
2235void vma_interval_tree_remove(struct vm_area_struct *node,
2236                              struct rb_root_cached *root);
2237struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2238                                unsigned long start, unsigned long last);
2239struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2240                                unsigned long start, unsigned long last);
2241
2242#define vma_interval_tree_foreach(vma, root, start, last)               \
2243        for (vma = vma_interval_tree_iter_first(root, start, last);     \
2244             vma; vma = vma_interval_tree_iter_next(vma, start, last))
2245
2246void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2247                                   struct rb_root_cached *root);
2248void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2249                                   struct rb_root_cached *root);
2250struct anon_vma_chain *
2251anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2252                                  unsigned long start, unsigned long last);
2253struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2254        struct anon_vma_chain *node, unsigned long start, unsigned long last);
2255#ifdef CONFIG_DEBUG_VM_RB
2256void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2257#endif
2258
2259#define anon_vma_interval_tree_foreach(avc, root, start, last)           \
2260        for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2261             avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2262
2263/* mmap.c */
2264extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2265extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2266        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2267        struct vm_area_struct *expand);
2268static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2269        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2270{
2271        return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2272}
2273extern struct vm_area_struct *vma_merge(struct mm_struct *,
2274        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2275        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2276        struct mempolicy *, struct vm_userfaultfd_ctx);
2277extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2278extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2279        unsigned long addr, int new_below);
2280extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2281        unsigned long addr, int new_below);
2282extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2283extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2284        struct rb_node **, struct rb_node *);
2285extern void unlink_file_vma(struct vm_area_struct *);
2286extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2287        unsigned long addr, unsigned long len, pgoff_t pgoff,
2288        bool *need_rmap_locks);
2289extern void exit_mmap(struct mm_struct *);
2290
2291static inline int check_data_rlimit(unsigned long rlim,
2292                                    unsigned long new,
2293                                    unsigned long start,
2294                                    unsigned long end_data,
2295                                    unsigned long start_data)
2296{
2297        if (rlim < RLIM_INFINITY) {
2298                if (((new - start) + (end_data - start_data)) > rlim)
2299                        return -ENOSPC;
2300        }
2301
2302        return 0;
2303}
2304
2305extern int mm_take_all_locks(struct mm_struct *mm);
2306extern void mm_drop_all_locks(struct mm_struct *mm);
2307
2308extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2309extern struct file *get_mm_exe_file(struct mm_struct *mm);
2310extern struct file *get_task_exe_file(struct task_struct *task);
2311
2312extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2313extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2314
2315extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2316                                   const struct vm_special_mapping *sm);
2317extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2318                                   unsigned long addr, unsigned long len,
2319                                   unsigned long flags,
2320                                   const struct vm_special_mapping *spec);
2321/* This is an obsolete alternative to _install_special_mapping. */
2322extern int install_special_mapping(struct mm_struct *mm,
2323                                   unsigned long addr, unsigned long len,
2324                                   unsigned long flags, struct page **pages);
2325
2326unsigned long randomize_stack_top(unsigned long stack_top);
2327
2328extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2329
2330extern unsigned long mmap_region(struct file *file, unsigned long addr,
2331        unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2332        struct list_head *uf);
2333extern unsigned long do_mmap(struct file *file, unsigned long addr,
2334        unsigned long len, unsigned long prot, unsigned long flags,
2335        vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2336        struct list_head *uf);
2337extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2338                       struct list_head *uf, bool downgrade);
2339extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2340                     struct list_head *uf);
2341
2342static inline unsigned long
2343do_mmap_pgoff(struct file *file, unsigned long addr,
2344        unsigned long len, unsigned long prot, unsigned long flags,
2345        unsigned long pgoff, unsigned long *populate,
2346        struct list_head *uf)
2347{
2348        return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2349}
2350
2351#ifdef CONFIG_MMU
2352extern int __mm_populate(unsigned long addr, unsigned long len,
2353                         int ignore_errors);
2354static inline void mm_populate(unsigned long addr, unsigned long len)
2355{
2356        /* Ignore errors */
2357        (void) __mm_populate(addr, len, 1);
2358}
2359#else
2360static inline void mm_populate(unsigned long addr, unsigned long len) {}
2361#endif
2362
2363/* These take the mm semaphore themselves */
2364extern int __must_check vm_brk(unsigned long, unsigned long);
2365extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2366extern int vm_munmap(unsigned long, size_t);
2367extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2368        unsigned long, unsigned long,
2369        unsigned long, unsigned long);
2370
2371struct vm_unmapped_area_info {
2372#define VM_UNMAPPED_AREA_TOPDOWN 1
2373        unsigned long flags;
2374        unsigned long length;
2375        unsigned long low_limit;
2376        unsigned long high_limit;
2377        unsigned long align_mask;
2378        unsigned long align_offset;
2379};
2380
2381extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2382extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2383
2384/*
2385 * Search for an unmapped address range.
2386 *
2387 * We are looking for a range that:
2388 * - does not intersect with any VMA;
2389 * - is contained within the [low_limit, high_limit) interval;
2390 * - is at least the desired size.
2391 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2392 */
2393static inline unsigned long
2394vm_unmapped_area(struct vm_unmapped_area_info *info)
2395{
2396        if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2397                return unmapped_area_topdown(info);
2398        else
2399                return unmapped_area(info);
2400}
2401
2402/* truncate.c */
2403extern void truncate_inode_pages(struct address_space *, loff_t);
2404extern void truncate_inode_pages_range(struct address_space *,
2405                                       loff_t lstart, loff_t lend);
2406extern void truncate_inode_pages_final(struct address_space *);
2407
2408/* generic vm_area_ops exported for stackable file systems */
2409extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2410extern void filemap_map_pages(struct vm_fault *vmf,
2411                pgoff_t start_pgoff, pgoff_t end_pgoff);
2412extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2413
2414/* mm/page-writeback.c */
2415int __must_check write_one_page(struct page *page);
2416void task_dirty_inc(struct task_struct *tsk);
2417
2418/* readahead.c */
2419#define VM_READAHEAD_PAGES      (SZ_128K / PAGE_SIZE)
2420
2421int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2422                        pgoff_t offset, unsigned long nr_to_read);
2423
2424void page_cache_sync_readahead(struct address_space *mapping,
2425                               struct file_ra_state *ra,
2426                               struct file *filp,
2427                               pgoff_t offset,
2428                               unsigned long size);
2429
2430void page_cache_async_readahead(struct address_space *mapping,
2431                                struct file_ra_state *ra,
2432                                struct file *filp,
2433                                struct page *pg,
2434                                pgoff_t offset,
2435                                unsigned long size);
2436
2437extern unsigned long stack_guard_gap;
2438/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2439extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2440
2441/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2442extern int expand_downwards(struct vm_area_struct *vma,
2443                unsigned long address);
2444#if VM_GROWSUP
2445extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2446#else
2447  #define expand_upwards(vma, address) (0)
2448#endif
2449
2450/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2451extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2452extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2453                                             struct vm_area_struct **pprev);
2454
2455/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2456   NULL if none.  Assume start_addr < end_addr. */
2457static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2458{
2459        struct vm_area_struct * vma = find_vma(mm,start_addr);
2460
2461        if (vma && end_addr <= vma->vm_start)
2462                vma = NULL;
2463        return vma;
2464}
2465
2466static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2467{
2468        unsigned long vm_start = vma->vm_start;
2469
2470        if (vma->vm_flags & VM_GROWSDOWN) {
2471                vm_start -= stack_guard_gap;
2472                if (vm_start > vma->vm_start)
2473                        vm_start = 0;
2474        }
2475        return vm_start;
2476}
2477
2478static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2479{
2480        unsigned long vm_end = vma->vm_end;
2481
2482        if (vma->vm_flags & VM_GROWSUP) {
2483                vm_end += stack_guard_gap;
2484                if (vm_end < vma->vm_end)
2485                        vm_end = -PAGE_SIZE;
2486        }
2487        return vm_end;
2488}
2489
2490static inline unsigned long vma_pages(struct vm_area_struct *vma)
2491{
2492        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2493}
2494
2495/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2496static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2497                                unsigned long vm_start, unsigned long vm_end)
2498{
2499        struct vm_area_struct *vma = find_vma(mm, vm_start);
2500
2501        if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2502                vma = NULL;
2503
2504        return vma;
2505}
2506
2507static inline bool range_in_vma(struct vm_area_struct *vma,
2508                                unsigned long start, unsigned long end)
2509{
2510        return (vma && vma->vm_start <= start && end <= vma->vm_end);
2511}
2512
2513#ifdef CONFIG_MMU
2514pgprot_t vm_get_page_prot(unsigned long vm_flags);
2515void vma_set_page_prot(struct vm_area_struct *vma);
2516#else
2517static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2518{
2519        return __pgprot(0);
2520}
2521static inline void vma_set_page_prot(struct vm_area_struct *vma)
2522{
2523        vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2524}
2525#endif
2526
2527#ifdef CONFIG_NUMA_BALANCING
2528unsigned long change_prot_numa(struct vm_area_struct *vma,
2529                        unsigned long start, unsigned long end);
2530#endif
2531
2532struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2533int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2534                        unsigned long pfn, unsigned long size, pgprot_t);
2535int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2536int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2537                                unsigned long num);
2538int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2539                                unsigned long num);
2540vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2541                        unsigned long pfn);
2542vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2543                        unsigned long pfn, pgprot_t pgprot);
2544vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2545                        pfn_t pfn);
2546vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2547                unsigned long addr, pfn_t pfn);
2548int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2549
2550static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2551                                unsigned long addr, struct page *page)
2552{
2553        int err = vm_insert_page(vma, addr, page);
2554
2555        if (err == -ENOMEM)
2556                return VM_FAULT_OOM;
2557        if (err < 0 && err != -EBUSY)
2558                return VM_FAULT_SIGBUS;
2559
2560        return VM_FAULT_NOPAGE;
2561}
2562
2563static inline vm_fault_t vmf_error(int err)
2564{
2565        if (err == -ENOMEM)
2566                return VM_FAULT_OOM;
2567        return VM_FAULT_SIGBUS;
2568}
2569
2570struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2571                         unsigned int foll_flags);
2572
2573#define FOLL_WRITE      0x01    /* check pte is writable */
2574#define FOLL_TOUCH      0x02    /* mark page accessed */
2575#define FOLL_GET        0x04    /* do get_page on page */
2576#define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
2577#define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
2578#define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
2579                                 * and return without waiting upon it */
2580#define FOLL_POPULATE   0x40    /* fault in page */
2581#define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
2582#define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
2583#define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
2584#define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
2585#define FOLL_TRIED      0x800   /* a retry, previous pass started an IO */
2586#define FOLL_MLOCK      0x1000  /* lock present pages */
2587#define FOLL_REMOTE     0x2000  /* we are working on non-current tsk/mm */
2588#define FOLL_COW        0x4000  /* internal GUP flag */
2589#define FOLL_ANON       0x8000  /* don't do file mappings */
2590#define FOLL_LONGTERM   0x10000 /* mapping lifetime is indefinite: see below */
2591#define FOLL_SPLIT_PMD  0x20000 /* split huge pmd before returning */
2592
2593/*
2594 * NOTE on FOLL_LONGTERM:
2595 *
2596 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2597 * period _often_ under userspace control.  This is contrasted with
2598 * iov_iter_get_pages() where usages which are transient.
2599 *
2600 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2601 * lifetime enforced by the filesystem and we need guarantees that longterm
2602 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2603 * the filesystem.  Ideas for this coordination include revoking the longterm
2604 * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
2605 * added after the problem with filesystems was found FS DAX VMAs are
2606 * specifically failed.  Filesystem pages are still subject to bugs and use of
2607 * FOLL_LONGTERM should be avoided on those pages.
2608 *
2609 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2610 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2611 * and calls to get_user_pages_[un]locked are specifically not allowed.  This
2612 * is due to an incompatibility with the FS DAX check and
2613 * FAULT_FLAG_ALLOW_RETRY
2614 *
2615 * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2616 * that region.  And so CMA attempts to migrate the page before pinning when
2617 * FOLL_LONGTERM is specified.
2618 */
2619
2620static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2621{
2622        if (vm_fault & VM_FAULT_OOM)
2623                return -ENOMEM;
2624        if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2625                return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2626        if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2627                return -EFAULT;
2628        return 0;
2629}
2630
2631typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2632extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2633                               unsigned long size, pte_fn_t fn, void *data);
2634
2635
2636#ifdef CONFIG_PAGE_POISONING
2637extern bool page_poisoning_enabled(void);
2638extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2639#else
2640static inline bool page_poisoning_enabled(void) { return false; }
2641static inline void kernel_poison_pages(struct page *page, int numpages,
2642                                        int enable) { }
2643#endif
2644
2645#ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2646DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2647#else
2648DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2649#endif
2650static inline bool want_init_on_alloc(gfp_t flags)
2651{
2652        if (static_branch_unlikely(&init_on_alloc) &&
2653            !page_poisoning_enabled())
2654                return true;
2655        return flags & __GFP_ZERO;
2656}
2657
2658#ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2659DECLARE_STATIC_KEY_TRUE(init_on_free);
2660#else
2661DECLARE_STATIC_KEY_FALSE(init_on_free);
2662#endif
2663static inline bool want_init_on_free(void)
2664{
2665        return static_branch_unlikely(&init_on_free) &&
2666               !page_poisoning_enabled();
2667}
2668
2669#ifdef CONFIG_DEBUG_PAGEALLOC_ENABLE_DEFAULT
2670DECLARE_STATIC_KEY_TRUE(_debug_pagealloc_enabled);
2671#else
2672DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2673#endif
2674
2675static inline bool debug_pagealloc_enabled(void)
2676{
2677        if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2678                return false;
2679
2680        return static_branch_unlikely(&_debug_pagealloc_enabled);
2681}
2682
2683#if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2684extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2685
2686static inline void
2687kernel_map_pages(struct page *page, int numpages, int enable)
2688{
2689        __kernel_map_pages(page, numpages, enable);
2690}
2691#ifdef CONFIG_HIBERNATION
2692extern bool kernel_page_present(struct page *page);
2693#endif  /* CONFIG_HIBERNATION */
2694#else   /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2695static inline void
2696kernel_map_pages(struct page *page, int numpages, int enable) {}
2697#ifdef CONFIG_HIBERNATION
2698static inline bool kernel_page_present(struct page *page) { return true; }
2699#endif  /* CONFIG_HIBERNATION */
2700#endif  /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2701
2702#ifdef __HAVE_ARCH_GATE_AREA
2703extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2704extern int in_gate_area_no_mm(unsigned long addr);
2705extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2706#else
2707static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2708{
2709        return NULL;
2710}
2711static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2712static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2713{
2714        return 0;
2715}
2716#endif  /* __HAVE_ARCH_GATE_AREA */
2717
2718extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2719
2720#ifdef CONFIG_SYSCTL
2721extern int sysctl_drop_caches;
2722int drop_caches_sysctl_handler(struct ctl_table *, int,
2723                                        void __user *, size_t *, loff_t *);
2724#endif
2725
2726void drop_slab(void);
2727void drop_slab_node(int nid);
2728
2729#ifndef CONFIG_MMU
2730#define randomize_va_space 0
2731#else
2732extern int randomize_va_space;
2733#endif
2734
2735const char * arch_vma_name(struct vm_area_struct *vma);
2736#ifdef CONFIG_MMU
2737void print_vma_addr(char *prefix, unsigned long rip);
2738#else
2739static inline void print_vma_addr(char *prefix, unsigned long rip)
2740{
2741}
2742#endif
2743
2744void *sparse_buffer_alloc(unsigned long size);
2745struct page * __populate_section_memmap(unsigned long pfn,
2746                unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2747pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2748p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2749pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2750pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2751pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2752void *vmemmap_alloc_block(unsigned long size, int node);
2753struct vmem_altmap;
2754void *vmemmap_alloc_block_buf(unsigned long size, int node);
2755void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2756void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2757int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2758                               int node);
2759int vmemmap_populate(unsigned long start, unsigned long end, int node,
2760                struct vmem_altmap *altmap);
2761void vmemmap_populate_print_last(void);
2762#ifdef CONFIG_MEMORY_HOTPLUG
2763void vmemmap_free(unsigned long start, unsigned long end,
2764                struct vmem_altmap *altmap);
2765#endif
2766void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2767                                  unsigned long nr_pages);
2768
2769enum mf_flags {
2770        MF_COUNT_INCREASED = 1 << 0,
2771        MF_ACTION_REQUIRED = 1 << 1,
2772        MF_MUST_KILL = 1 << 2,
2773        MF_SOFT_OFFLINE = 1 << 3,
2774};
2775extern int memory_failure(unsigned long pfn, int flags);
2776extern void memory_failure_queue(unsigned long pfn, int flags);
2777extern int unpoison_memory(unsigned long pfn);
2778extern int get_hwpoison_page(struct page *page);
2779#define put_hwpoison_page(page) put_page(page)
2780extern int sysctl_memory_failure_early_kill;
2781extern int sysctl_memory_failure_recovery;
2782extern void shake_page(struct page *p, int access);
2783extern atomic_long_t num_poisoned_pages __read_mostly;
2784extern int soft_offline_page(struct page *page, int flags);
2785
2786
2787/*
2788 * Error handlers for various types of pages.
2789 */
2790enum mf_result {
2791        MF_IGNORED,     /* Error: cannot be handled */
2792        MF_FAILED,      /* Error: handling failed */
2793        MF_DELAYED,     /* Will be handled later */
2794        MF_RECOVERED,   /* Successfully recovered */
2795};
2796
2797enum mf_action_page_type {
2798        MF_MSG_KERNEL,
2799        MF_MSG_KERNEL_HIGH_ORDER,
2800        MF_MSG_SLAB,
2801        MF_MSG_DIFFERENT_COMPOUND,
2802        MF_MSG_POISONED_HUGE,
2803        MF_MSG_HUGE,
2804        MF_MSG_FREE_HUGE,
2805        MF_MSG_NON_PMD_HUGE,
2806        MF_MSG_UNMAP_FAILED,
2807        MF_MSG_DIRTY_SWAPCACHE,
2808        MF_MSG_CLEAN_SWAPCACHE,
2809        MF_MSG_DIRTY_MLOCKED_LRU,
2810        MF_MSG_CLEAN_MLOCKED_LRU,
2811        MF_MSG_DIRTY_UNEVICTABLE_LRU,
2812        MF_MSG_CLEAN_UNEVICTABLE_LRU,
2813        MF_MSG_DIRTY_LRU,
2814        MF_MSG_CLEAN_LRU,
2815        MF_MSG_TRUNCATED_LRU,
2816        MF_MSG_BUDDY,
2817        MF_MSG_BUDDY_2ND,
2818        MF_MSG_DAX,
2819        MF_MSG_UNKNOWN,
2820};
2821
2822#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2823extern void clear_huge_page(struct page *page,
2824                            unsigned long addr_hint,
2825                            unsigned int pages_per_huge_page);
2826extern void copy_user_huge_page(struct page *dst, struct page *src,
2827                                unsigned long addr_hint,
2828                                struct vm_area_struct *vma,
2829                                unsigned int pages_per_huge_page);
2830extern long copy_huge_page_from_user(struct page *dst_page,
2831                                const void __user *usr_src,
2832                                unsigned int pages_per_huge_page,
2833                                bool allow_pagefault);
2834#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2835
2836#ifdef CONFIG_DEBUG_PAGEALLOC
2837extern unsigned int _debug_guardpage_minorder;
2838DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2839
2840static inline unsigned int debug_guardpage_minorder(void)
2841{
2842        return _debug_guardpage_minorder;
2843}
2844
2845static inline bool debug_guardpage_enabled(void)
2846{
2847        return static_branch_unlikely(&_debug_guardpage_enabled);
2848}
2849
2850static inline bool page_is_guard(struct page *page)
2851{
2852        if (!debug_guardpage_enabled())
2853                return false;
2854
2855        return PageGuard(page);
2856}
2857#else
2858static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2859static inline bool debug_guardpage_enabled(void) { return false; }
2860static inline bool page_is_guard(struct page *page) { return false; }
2861#endif /* CONFIG_DEBUG_PAGEALLOC */
2862
2863#if MAX_NUMNODES > 1
2864void __init setup_nr_node_ids(void);
2865#else
2866static inline void setup_nr_node_ids(void) {}
2867#endif
2868
2869extern int memcmp_pages(struct page *page1, struct page *page2);
2870
2871static inline int pages_identical(struct page *page1, struct page *page2)
2872{
2873        return !memcmp_pages(page1, page2);
2874}
2875
2876#endif /* __KERNEL__ */
2877#endif /* _LINUX_MM_H */
2878