Linux Audio

Check our new training course

Loading...
v4.10.11
   1#ifndef _LINUX_MM_H
   2#define _LINUX_MM_H
   3
   4#include <linux/errno.h>
   5
   6#ifdef __KERNEL__
   7
   8#include <linux/mmdebug.h>
   9#include <linux/gfp.h>
  10#include <linux/bug.h>
  11#include <linux/list.h>
  12#include <linux/mmzone.h>
  13#include <linux/rbtree.h>
  14#include <linux/atomic.h>
  15#include <linux/debug_locks.h>
  16#include <linux/mm_types.h>
  17#include <linux/range.h>
  18#include <linux/pfn.h>
  19#include <linux/percpu-refcount.h>
  20#include <linux/bit_spinlock.h>
  21#include <linux/shrinker.h>
  22#include <linux/resource.h>
  23#include <linux/page_ext.h>
  24#include <linux/err.h>
  25#include <linux/page_ref.h>
  26
  27struct mempolicy;
  28struct anon_vma;
  29struct anon_vma_chain;
  30struct file_ra_state;
  31struct user_struct;
  32struct writeback_control;
  33struct bdi_writeback;
  34
  35#ifndef CONFIG_NEED_MULTIPLE_NODES	/* Don't use mapnrs, do it properly */
  36extern unsigned long max_mapnr;
  37
  38static inline void set_max_mapnr(unsigned long limit)
  39{
  40	max_mapnr = limit;
  41}
  42#else
  43static inline void set_max_mapnr(unsigned long limit) { }
  44#endif
  45
  46extern unsigned long totalram_pages;
  47extern void * high_memory;
  48extern int page_cluster;
  49
  50#ifdef CONFIG_SYSCTL
  51extern int sysctl_legacy_va_layout;
  52#else
  53#define sysctl_legacy_va_layout 0
  54#endif
  55
  56#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
  57extern const int mmap_rnd_bits_min;
  58extern const int mmap_rnd_bits_max;
  59extern int mmap_rnd_bits __read_mostly;
  60#endif
  61#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
  62extern const int mmap_rnd_compat_bits_min;
  63extern const int mmap_rnd_compat_bits_max;
  64extern int mmap_rnd_compat_bits __read_mostly;
  65#endif
  66
  67#include <asm/page.h>
  68#include <asm/pgtable.h>
  69#include <asm/processor.h>
  70
  71#ifndef __pa_symbol
  72#define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
  73#endif
  74
  75#ifndef page_to_virt
  76#define page_to_virt(x)	__va(PFN_PHYS(page_to_pfn(x)))
  77#endif
  78
  79/*
  80 * To prevent common memory management code establishing
  81 * a zero page mapping on a read fault.
  82 * This macro should be defined within <asm/pgtable.h>.
  83 * s390 does this to prevent multiplexing of hardware bits
  84 * related to the physical page in case of virtualization.
  85 */
  86#ifndef mm_forbids_zeropage
  87#define mm_forbids_zeropage(X)	(0)
  88#endif
  89
  90/*
  91 * Default maximum number of active map areas, this limits the number of vmas
  92 * per mm struct. Users can overwrite this number by sysctl but there is a
  93 * problem.
  94 *
  95 * When a program's coredump is generated as ELF format, a section is created
  96 * per a vma. In ELF, the number of sections is represented in unsigned short.
  97 * This means the number of sections should be smaller than 65535 at coredump.
  98 * Because the kernel adds some informative sections to a image of program at
  99 * generating coredump, we need some margin. The number of extra sections is
 100 * 1-3 now and depends on arch. We use "5" as safe margin, here.
 101 *
 102 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
 103 * not a hard limit any more. Although some userspace tools can be surprised by
 104 * that.
 105 */
 106#define MAPCOUNT_ELF_CORE_MARGIN	(5)
 107#define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
 108
 109extern int sysctl_max_map_count;
 110
 111extern unsigned long sysctl_user_reserve_kbytes;
 112extern unsigned long sysctl_admin_reserve_kbytes;
 113
 114extern int sysctl_overcommit_memory;
 115extern int sysctl_overcommit_ratio;
 116extern unsigned long sysctl_overcommit_kbytes;
 117
 118extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
 119				    size_t *, loff_t *);
 120extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
 121				    size_t *, loff_t *);
 122
 123#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
 124
 125/* to align the pointer to the (next) page boundary */
 126#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
 127
 128/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
 129#define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
 130
 131/*
 132 * Linux kernel virtual memory manager primitives.
 133 * The idea being to have a "virtual" mm in the same way
 134 * we have a virtual fs - giving a cleaner interface to the
 135 * mm details, and allowing different kinds of memory mappings
 136 * (from shared memory to executable loading to arbitrary
 137 * mmap() functions).
 138 */
 139
 140extern struct kmem_cache *vm_area_cachep;
 141
 142#ifndef CONFIG_MMU
 143extern struct rb_root nommu_region_tree;
 144extern struct rw_semaphore nommu_region_sem;
 145
 146extern unsigned int kobjsize(const void *objp);
 147#endif
 148
 149/*
 150 * vm_flags in vm_area_struct, see mm_types.h.
 151 * When changing, update also include/trace/events/mmflags.h
 152 */
 153#define VM_NONE		0x00000000
 154
 155#define VM_READ		0x00000001	/* currently active flags */
 156#define VM_WRITE	0x00000002
 157#define VM_EXEC		0x00000004
 158#define VM_SHARED	0x00000008
 159
 160/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
 161#define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
 162#define VM_MAYWRITE	0x00000020
 163#define VM_MAYEXEC	0x00000040
 164#define VM_MAYSHARE	0x00000080
 165
 166#define VM_GROWSDOWN	0x00000100	/* general info on the segment */
 167#define VM_UFFD_MISSING	0x00000200	/* missing pages tracking */
 168#define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
 169#define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
 170#define VM_UFFD_WP	0x00001000	/* wrprotect pages tracking */
 171
 172#define VM_LOCKED	0x00002000
 173#define VM_IO           0x00004000	/* Memory mapped I/O or similar */
 174
 175					/* Used by sys_madvise() */
 176#define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
 177#define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
 178
 179#define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
 180#define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
 181#define VM_LOCKONFAULT	0x00080000	/* Lock the pages covered when they are faulted in */
 182#define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
 183#define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
 184#define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
 185#define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
 186#define VM_ARCH_2	0x02000000
 187#define VM_DONTDUMP	0x04000000	/* Do not include in the core dump */
 188
 189#ifdef CONFIG_MEM_SOFT_DIRTY
 190# define VM_SOFTDIRTY	0x08000000	/* Not soft dirty clean area */
 191#else
 192# define VM_SOFTDIRTY	0
 193#endif
 194
 195#define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
 196#define VM_HUGEPAGE	0x20000000	/* MADV_HUGEPAGE marked this vma */
 197#define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
 198#define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
 199
 200#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
 201#define VM_HIGH_ARCH_BIT_0	32	/* bit only usable on 64-bit architectures */
 202#define VM_HIGH_ARCH_BIT_1	33	/* bit only usable on 64-bit architectures */
 203#define VM_HIGH_ARCH_BIT_2	34	/* bit only usable on 64-bit architectures */
 204#define VM_HIGH_ARCH_BIT_3	35	/* bit only usable on 64-bit architectures */
 205#define VM_HIGH_ARCH_0	BIT(VM_HIGH_ARCH_BIT_0)
 206#define VM_HIGH_ARCH_1	BIT(VM_HIGH_ARCH_BIT_1)
 207#define VM_HIGH_ARCH_2	BIT(VM_HIGH_ARCH_BIT_2)
 208#define VM_HIGH_ARCH_3	BIT(VM_HIGH_ARCH_BIT_3)
 209#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
 210
 211#if defined(CONFIG_X86)
 212# define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
 213#if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
 214# define VM_PKEY_SHIFT	VM_HIGH_ARCH_BIT_0
 215# define VM_PKEY_BIT0	VM_HIGH_ARCH_0	/* A protection key is a 4-bit value */
 216# define VM_PKEY_BIT1	VM_HIGH_ARCH_1
 217# define VM_PKEY_BIT2	VM_HIGH_ARCH_2
 218# define VM_PKEY_BIT3	VM_HIGH_ARCH_3
 219#endif
 220#elif defined(CONFIG_PPC)
 221# define VM_SAO		VM_ARCH_1	/* Strong Access Ordering (powerpc) */
 222#elif defined(CONFIG_PARISC)
 223# define VM_GROWSUP	VM_ARCH_1
 224#elif defined(CONFIG_METAG)
 225# define VM_GROWSUP	VM_ARCH_1
 226#elif defined(CONFIG_IA64)
 227# define VM_GROWSUP	VM_ARCH_1
 228#elif !defined(CONFIG_MMU)
 229# define VM_MAPPED_COPY	VM_ARCH_1	/* T if mapped copy of data (nommu mmap) */
 230#endif
 231
 232#if defined(CONFIG_X86)
 233/* MPX specific bounds table or bounds directory */
 234# define VM_MPX		VM_ARCH_2
 235#endif
 236
 237#ifndef VM_GROWSUP
 238# define VM_GROWSUP	VM_NONE
 239#endif
 240
 241/* Bits set in the VMA until the stack is in its final location */
 242#define VM_STACK_INCOMPLETE_SETUP	(VM_RAND_READ | VM_SEQ_READ)
 243
 244#ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
 245#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 246#endif
 247
 248#ifdef CONFIG_STACK_GROWSUP
 249#define VM_STACK	VM_GROWSUP
 250#else
 251#define VM_STACK	VM_GROWSDOWN
 252#endif
 253
 254#define VM_STACK_FLAGS	(VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 255
 256/*
 257 * Special vmas that are non-mergable, non-mlock()able.
 258 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
 259 */
 260#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
 261
 262/* This mask defines which mm->def_flags a process can inherit its parent */
 263#define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
 264
 265/* This mask is used to clear all the VMA flags used by mlock */
 266#define VM_LOCKED_CLEAR_MASK	(~(VM_LOCKED | VM_LOCKONFAULT))
 267
 268/*
 269 * mapping from the currently active vm_flags protection bits (the
 270 * low four bits) to a page protection mask..
 271 */
 272extern pgprot_t protection_map[16];
 273
 274#define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
 275#define FAULT_FLAG_MKWRITE	0x02	/* Fault was mkwrite of existing pte */
 276#define FAULT_FLAG_ALLOW_RETRY	0x04	/* Retry fault if blocking */
 277#define FAULT_FLAG_RETRY_NOWAIT	0x08	/* Don't drop mmap_sem and wait when retrying */
 278#define FAULT_FLAG_KILLABLE	0x10	/* The fault task is in SIGKILL killable region */
 279#define FAULT_FLAG_TRIED	0x20	/* Second try */
 280#define FAULT_FLAG_USER		0x40	/* The fault originated in userspace */
 281#define FAULT_FLAG_REMOTE	0x80	/* faulting for non current tsk/mm */
 282#define FAULT_FLAG_INSTRUCTION  0x100	/* The fault was during an instruction fetch */
 283
 284/*
 285 * vm_fault is filled by the the pagefault handler and passed to the vma's
 286 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 287 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 288 *
 289 * MM layer fills up gfp_mask for page allocations but fault handler might
 290 * alter it if its implementation requires a different allocation context.
 291 *
 292 * pgoff should be used in favour of virtual_address, if possible.
 293 */
 294struct vm_fault {
 295	struct vm_area_struct *vma;	/* Target VMA */
 296	unsigned int flags;		/* FAULT_FLAG_xxx flags */
 297	gfp_t gfp_mask;			/* gfp mask to be used for allocations */
 298	pgoff_t pgoff;			/* Logical page offset based on vma */
 299	unsigned long address;		/* Faulting virtual address */
 300	pmd_t *pmd;			/* Pointer to pmd entry matching
 301					 * the 'address' */
 302	pte_t orig_pte;			/* Value of PTE at the time of fault */
 303
 304	struct page *cow_page;		/* Page handler may use for COW fault */
 305	struct mem_cgroup *memcg;	/* Cgroup cow_page belongs to */
 306	struct page *page;		/* ->fault handlers should return a
 307					 * page here, unless VM_FAULT_NOPAGE
 308					 * is set (which is also implied by
 309					 * VM_FAULT_ERROR).
 310					 */
 311	/* These three entries are valid only while holding ptl lock */
 312	pte_t *pte;			/* Pointer to pte entry matching
 313					 * the 'address'. NULL if the page
 314					 * table hasn't been allocated.
 315					 */
 316	spinlock_t *ptl;		/* Page table lock.
 317					 * Protects pte page table if 'pte'
 318					 * is not NULL, otherwise pmd.
 319					 */
 320	pgtable_t prealloc_pte;		/* Pre-allocated pte page table.
 321					 * vm_ops->map_pages() calls
 322					 * alloc_set_pte() from atomic context.
 323					 * do_fault_around() pre-allocates
 324					 * page table to avoid allocation from
 325					 * atomic context.
 326					 */
 327};
 328
 329/*
 330 * These are the virtual MM functions - opening of an area, closing and
 331 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 332 * to the functions called when a no-page or a wp-page exception occurs. 
 333 */
 334struct vm_operations_struct {
 335	void (*open)(struct vm_area_struct * area);
 336	void (*close)(struct vm_area_struct * area);
 337	int (*mremap)(struct vm_area_struct * area);
 338	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 339	int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
 340						pmd_t *, unsigned int flags);
 341	void (*map_pages)(struct vm_fault *vmf,
 342			pgoff_t start_pgoff, pgoff_t end_pgoff);
 343
 344	/* notification that a previously read-only page is about to become
 345	 * writable, if an error is returned it will cause a SIGBUS */
 346	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 347
 348	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
 349	int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 350
 351	/* called by access_process_vm when get_user_pages() fails, typically
 352	 * for use by special VMAs that can switch between memory and hardware
 353	 */
 354	int (*access)(struct vm_area_struct *vma, unsigned long addr,
 355		      void *buf, int len, int write);
 356
 357	/* Called by the /proc/PID/maps code to ask the vma whether it
 358	 * has a special name.  Returning non-NULL will also cause this
 359	 * vma to be dumped unconditionally. */
 360	const char *(*name)(struct vm_area_struct *vma);
 361
 362#ifdef CONFIG_NUMA
 363	/*
 364	 * set_policy() op must add a reference to any non-NULL @new mempolicy
 365	 * to hold the policy upon return.  Caller should pass NULL @new to
 366	 * remove a policy and fall back to surrounding context--i.e. do not
 367	 * install a MPOL_DEFAULT policy, nor the task or system default
 368	 * mempolicy.
 369	 */
 370	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 371
 372	/*
 373	 * get_policy() op must add reference [mpol_get()] to any policy at
 374	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 375	 * in mm/mempolicy.c will do this automatically.
 376	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 377	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 378	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 379	 * must return NULL--i.e., do not "fallback" to task or system default
 380	 * policy.
 381	 */
 382	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 383					unsigned long addr);
 384#endif
 385	/*
 386	 * Called by vm_normal_page() for special PTEs to find the
 387	 * page for @addr.  This is useful if the default behavior
 388	 * (using pte_page()) would not find the correct page.
 389	 */
 390	struct page *(*find_special_page)(struct vm_area_struct *vma,
 391					  unsigned long addr);
 392};
 393
 394struct mmu_gather;
 395struct inode;
 396
 397#define page_private(page)		((page)->private)
 398#define set_page_private(page, v)	((page)->private = (v))
 399
 400#if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
 401static inline int pmd_devmap(pmd_t pmd)
 402{
 403	return 0;
 404}
 405#endif
 406
 407/*
 408 * FIXME: take this include out, include page-flags.h in
 409 * files which need it (119 of them)
 410 */
 411#include <linux/page-flags.h>
 412#include <linux/huge_mm.h>
 413
 414/*
 415 * Methods to modify the page usage count.
 416 *
 417 * What counts for a page usage:
 418 * - cache mapping   (page->mapping)
 419 * - private data    (page->private)
 420 * - page mapped in a task's page tables, each mapping
 421 *   is counted separately
 422 *
 423 * Also, many kernel routines increase the page count before a critical
 424 * routine so they can be sure the page doesn't go away from under them.
 425 */
 426
 427/*
 428 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 429 */
 430static inline int put_page_testzero(struct page *page)
 431{
 432	VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
 433	return page_ref_dec_and_test(page);
 434}
 435
 436/*
 437 * Try to grab a ref unless the page has a refcount of zero, return false if
 438 * that is the case.
 439 * This can be called when MMU is off so it must not access
 440 * any of the virtual mappings.
 441 */
 442static inline int get_page_unless_zero(struct page *page)
 443{
 444	return page_ref_add_unless(page, 1, 0);
 445}
 446
 447extern int page_is_ram(unsigned long pfn);
 448
 449enum {
 450	REGION_INTERSECTS,
 451	REGION_DISJOINT,
 452	REGION_MIXED,
 453};
 454
 455int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
 456		      unsigned long desc);
 457
 458/* Support for virtually mapped pages */
 459struct page *vmalloc_to_page(const void *addr);
 460unsigned long vmalloc_to_pfn(const void *addr);
 461
 462/*
 463 * Determine if an address is within the vmalloc range
 464 *
 465 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 466 * is no special casing required.
 467 */
 468static inline bool is_vmalloc_addr(const void *x)
 469{
 470#ifdef CONFIG_MMU
 471	unsigned long addr = (unsigned long)x;
 472
 473	return addr >= VMALLOC_START && addr < VMALLOC_END;
 474#else
 475	return false;
 476#endif
 477}
 478#ifdef CONFIG_MMU
 479extern int is_vmalloc_or_module_addr(const void *x);
 480#else
 481static inline int is_vmalloc_or_module_addr(const void *x)
 482{
 483	return 0;
 484}
 485#endif
 486
 487extern void kvfree(const void *addr);
 488
 489static inline atomic_t *compound_mapcount_ptr(struct page *page)
 490{
 491	return &page[1].compound_mapcount;
 492}
 493
 494static inline int compound_mapcount(struct page *page)
 495{
 496	VM_BUG_ON_PAGE(!PageCompound(page), page);
 
 497	page = compound_head(page);
 498	return atomic_read(compound_mapcount_ptr(page)) + 1;
 499}
 500
 501/*
 502 * The atomic page->_mapcount, starts from -1: so that transitions
 503 * both from it and to it can be tracked, using atomic_inc_and_test
 504 * and atomic_add_negative(-1).
 505 */
 506static inline void page_mapcount_reset(struct page *page)
 507{
 508	atomic_set(&(page)->_mapcount, -1);
 509}
 510
 511int __page_mapcount(struct page *page);
 512
 513static inline int page_mapcount(struct page *page)
 514{
 515	VM_BUG_ON_PAGE(PageSlab(page), page);
 516
 517	if (unlikely(PageCompound(page)))
 518		return __page_mapcount(page);
 519	return atomic_read(&page->_mapcount) + 1;
 520}
 521
 522#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 523int total_mapcount(struct page *page);
 524int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
 525#else
 526static inline int total_mapcount(struct page *page)
 527{
 528	return page_mapcount(page);
 529}
 530static inline int page_trans_huge_mapcount(struct page *page,
 531					   int *total_mapcount)
 532{
 533	int mapcount = page_mapcount(page);
 534	if (total_mapcount)
 535		*total_mapcount = mapcount;
 536	return mapcount;
 537}
 538#endif
 539
 540static inline struct page *virt_to_head_page(const void *x)
 541{
 542	struct page *page = virt_to_page(x);
 543
 544	return compound_head(page);
 545}
 546
 547void __put_page(struct page *page);
 548
 549void put_pages_list(struct list_head *pages);
 550
 551void split_page(struct page *page, unsigned int order);
 
 552
 553/*
 554 * Compound pages have a destructor function.  Provide a
 555 * prototype for that function and accessor functions.
 556 * These are _only_ valid on the head of a compound page.
 557 */
 558typedef void compound_page_dtor(struct page *);
 559
 560/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
 561enum compound_dtor_id {
 562	NULL_COMPOUND_DTOR,
 563	COMPOUND_PAGE_DTOR,
 564#ifdef CONFIG_HUGETLB_PAGE
 565	HUGETLB_PAGE_DTOR,
 566#endif
 567#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 568	TRANSHUGE_PAGE_DTOR,
 569#endif
 570	NR_COMPOUND_DTORS,
 571};
 572extern compound_page_dtor * const compound_page_dtors[];
 573
 574static inline void set_compound_page_dtor(struct page *page,
 575		enum compound_dtor_id compound_dtor)
 576{
 577	VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
 578	page[1].compound_dtor = compound_dtor;
 579}
 580
 581static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 582{
 583	VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
 584	return compound_page_dtors[page[1].compound_dtor];
 585}
 586
 587static inline unsigned int compound_order(struct page *page)
 588{
 589	if (!PageHead(page))
 590		return 0;
 591	return page[1].compound_order;
 592}
 593
 594static inline void set_compound_order(struct page *page, unsigned int order)
 595{
 596	page[1].compound_order = order;
 597}
 598
 599void free_compound_page(struct page *page);
 600
 601#ifdef CONFIG_MMU
 602/*
 603 * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
 604 * servicing faults for write access.  In the normal case, do always want
 605 * pte_mkwrite.  But get_user_pages can cause write faults for mappings
 606 * that do not have writing enabled, when used by access_process_vm.
 607 */
 608static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 609{
 610	if (likely(vma->vm_flags & VM_WRITE))
 611		pte = pte_mkwrite(pte);
 612	return pte;
 613}
 614
 615int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
 616		struct page *page);
 617int finish_fault(struct vm_fault *vmf);
 618int finish_mkwrite_fault(struct vm_fault *vmf);
 619#endif
 620
 621/*
 622 * Multiple processes may "see" the same page. E.g. for untouched
 623 * mappings of /dev/null, all processes see the same page full of
 624 * zeroes, and text pages of executables and shared libraries have
 625 * only one copy in memory, at most, normally.
 626 *
 627 * For the non-reserved pages, page_count(page) denotes a reference count.
 628 *   page_count() == 0 means the page is free. page->lru is then used for
 629 *   freelist management in the buddy allocator.
 630 *   page_count() > 0  means the page has been allocated.
 631 *
 632 * Pages are allocated by the slab allocator in order to provide memory
 633 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 634 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 635 * unless a particular usage is carefully commented. (the responsibility of
 636 * freeing the kmalloc memory is the caller's, of course).
 637 *
 638 * A page may be used by anyone else who does a __get_free_page().
 639 * In this case, page_count still tracks the references, and should only
 640 * be used through the normal accessor functions. The top bits of page->flags
 641 * and page->virtual store page management information, but all other fields
 642 * are unused and could be used privately, carefully. The management of this
 643 * page is the responsibility of the one who allocated it, and those who have
 644 * subsequently been given references to it.
 645 *
 646 * The other pages (we may call them "pagecache pages") are completely
 647 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 648 * The following discussion applies only to them.
 649 *
 650 * A pagecache page contains an opaque `private' member, which belongs to the
 651 * page's address_space. Usually, this is the address of a circular list of
 652 * the page's disk buffers. PG_private must be set to tell the VM to call
 653 * into the filesystem to release these pages.
 654 *
 655 * A page may belong to an inode's memory mapping. In this case, page->mapping
 656 * is the pointer to the inode, and page->index is the file offset of the page,
 657 * in units of PAGE_SIZE.
 658 *
 659 * If pagecache pages are not associated with an inode, they are said to be
 660 * anonymous pages. These may become associated with the swapcache, and in that
 661 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 662 *
 663 * In either case (swapcache or inode backed), the pagecache itself holds one
 664 * reference to the page. Setting PG_private should also increment the
 665 * refcount. The each user mapping also has a reference to the page.
 666 *
 667 * The pagecache pages are stored in a per-mapping radix tree, which is
 668 * rooted at mapping->page_tree, and indexed by offset.
 669 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 670 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 671 *
 672 * All pagecache pages may be subject to I/O:
 673 * - inode pages may need to be read from disk,
 674 * - inode pages which have been modified and are MAP_SHARED may need
 675 *   to be written back to the inode on disk,
 676 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 677 *   modified may need to be swapped out to swap space and (later) to be read
 678 *   back into memory.
 679 */
 680
 681/*
 682 * The zone field is never updated after free_area_init_core()
 683 * sets it, so none of the operations on it need to be atomic.
 684 */
 685
 686/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
 687#define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 688#define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
 689#define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
 690#define LAST_CPUPID_PGOFF	(ZONES_PGOFF - LAST_CPUPID_WIDTH)
 691
 692/*
 693 * Define the bit shifts to access each section.  For non-existent
 694 * sections we define the shift as 0; that plus a 0 mask ensures
 695 * the compiler will optimise away reference to them.
 696 */
 697#define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 698#define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
 699#define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
 700#define LAST_CPUPID_PGSHIFT	(LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
 701
 702/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
 703#ifdef NODE_NOT_IN_PAGE_FLAGS
 704#define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
 705#define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
 706						SECTIONS_PGOFF : ZONES_PGOFF)
 707#else
 708#define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
 709#define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
 710						NODES_PGOFF : ZONES_PGOFF)
 711#endif
 712
 713#define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 714
 715#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 716#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 717#endif
 718
 719#define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
 720#define NODES_MASK		((1UL << NODES_WIDTH) - 1)
 721#define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
 722#define LAST_CPUPID_MASK	((1UL << LAST_CPUPID_SHIFT) - 1)
 723#define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
 724
 725static inline enum zone_type page_zonenum(const struct page *page)
 726{
 727	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 728}
 729
 730#ifdef CONFIG_ZONE_DEVICE
 731void get_zone_device_page(struct page *page);
 732void put_zone_device_page(struct page *page);
 733static inline bool is_zone_device_page(const struct page *page)
 734{
 735	return page_zonenum(page) == ZONE_DEVICE;
 736}
 737#else
 738static inline void get_zone_device_page(struct page *page)
 739{
 740}
 741static inline void put_zone_device_page(struct page *page)
 742{
 743}
 744static inline bool is_zone_device_page(const struct page *page)
 745{
 746	return false;
 747}
 748#endif
 749
 750static inline void get_page(struct page *page)
 751{
 752	page = compound_head(page);
 753	/*
 754	 * Getting a normal page or the head of a compound page
 755	 * requires to already have an elevated page->_refcount.
 756	 */
 757	VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
 758	page_ref_inc(page);
 759
 760	if (unlikely(is_zone_device_page(page)))
 761		get_zone_device_page(page);
 762}
 763
 764static inline void put_page(struct page *page)
 765{
 766	page = compound_head(page);
 767
 768	if (put_page_testzero(page))
 769		__put_page(page);
 770
 771	if (unlikely(is_zone_device_page(page)))
 772		put_zone_device_page(page);
 773}
 774
 775#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 776#define SECTION_IN_PAGE_FLAGS
 777#endif
 778
 779/*
 780 * The identification function is mainly used by the buddy allocator for
 781 * determining if two pages could be buddies. We are not really identifying
 782 * the zone since we could be using the section number id if we do not have
 783 * node id available in page flags.
 784 * We only guarantee that it will return the same value for two combinable
 785 * pages in a zone.
 786 */
 787static inline int page_zone_id(struct page *page)
 788{
 789	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 790}
 791
 792static inline int zone_to_nid(struct zone *zone)
 793{
 794#ifdef CONFIG_NUMA
 795	return zone->node;
 796#else
 797	return 0;
 798#endif
 799}
 800
 801#ifdef NODE_NOT_IN_PAGE_FLAGS
 802extern int page_to_nid(const struct page *page);
 803#else
 804static inline int page_to_nid(const struct page *page)
 805{
 806	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 807}
 808#endif
 809
 810#ifdef CONFIG_NUMA_BALANCING
 811static inline int cpu_pid_to_cpupid(int cpu, int pid)
 812{
 813	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
 814}
 815
 816static inline int cpupid_to_pid(int cpupid)
 817{
 818	return cpupid & LAST__PID_MASK;
 819}
 820
 821static inline int cpupid_to_cpu(int cpupid)
 822{
 823	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
 824}
 825
 826static inline int cpupid_to_nid(int cpupid)
 827{
 828	return cpu_to_node(cpupid_to_cpu(cpupid));
 829}
 830
 831static inline bool cpupid_pid_unset(int cpupid)
 832{
 833	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
 834}
 835
 836static inline bool cpupid_cpu_unset(int cpupid)
 837{
 838	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
 839}
 840
 841static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
 842{
 843	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
 844}
 845
 846#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
 847#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
 848static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
 849{
 850	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
 851}
 852
 853static inline int page_cpupid_last(struct page *page)
 854{
 855	return page->_last_cpupid;
 856}
 857static inline void page_cpupid_reset_last(struct page *page)
 858{
 859	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
 860}
 861#else
 862static inline int page_cpupid_last(struct page *page)
 863{
 864	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
 865}
 866
 867extern int page_cpupid_xchg_last(struct page *page, int cpupid);
 868
 869static inline void page_cpupid_reset_last(struct page *page)
 870{
 871	page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
 
 
 
 872}
 873#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
 874#else /* !CONFIG_NUMA_BALANCING */
 875static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
 876{
 877	return page_to_nid(page); /* XXX */
 878}
 879
 880static inline int page_cpupid_last(struct page *page)
 881{
 882	return page_to_nid(page); /* XXX */
 883}
 884
 885static inline int cpupid_to_nid(int cpupid)
 886{
 887	return -1;
 888}
 889
 890static inline int cpupid_to_pid(int cpupid)
 891{
 892	return -1;
 893}
 894
 895static inline int cpupid_to_cpu(int cpupid)
 896{
 897	return -1;
 898}
 899
 900static inline int cpu_pid_to_cpupid(int nid, int pid)
 901{
 902	return -1;
 903}
 904
 905static inline bool cpupid_pid_unset(int cpupid)
 906{
 907	return 1;
 908}
 909
 910static inline void page_cpupid_reset_last(struct page *page)
 911{
 912}
 913
 914static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
 915{
 916	return false;
 917}
 918#endif /* CONFIG_NUMA_BALANCING */
 919
 920static inline struct zone *page_zone(const struct page *page)
 921{
 922	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 923}
 924
 925static inline pg_data_t *page_pgdat(const struct page *page)
 926{
 927	return NODE_DATA(page_to_nid(page));
 928}
 929
 930#ifdef SECTION_IN_PAGE_FLAGS
 931static inline void set_page_section(struct page *page, unsigned long section)
 932{
 933	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 934	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 935}
 936
 937static inline unsigned long page_to_section(const struct page *page)
 938{
 939	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 940}
 941#endif
 942
 943static inline void set_page_zone(struct page *page, enum zone_type zone)
 944{
 945	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 946	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 947}
 948
 949static inline void set_page_node(struct page *page, unsigned long node)
 950{
 951	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 952	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 953}
 954
 955static inline void set_page_links(struct page *page, enum zone_type zone,
 956	unsigned long node, unsigned long pfn)
 957{
 958	set_page_zone(page, zone);
 959	set_page_node(page, node);
 960#ifdef SECTION_IN_PAGE_FLAGS
 961	set_page_section(page, pfn_to_section_nr(pfn));
 962#endif
 963}
 964
 965#ifdef CONFIG_MEMCG
 966static inline struct mem_cgroup *page_memcg(struct page *page)
 967{
 968	return page->mem_cgroup;
 969}
 970static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
 971{
 972	WARN_ON_ONCE(!rcu_read_lock_held());
 973	return READ_ONCE(page->mem_cgroup);
 974}
 975#else
 976static inline struct mem_cgroup *page_memcg(struct page *page)
 977{
 978	return NULL;
 979}
 980static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
 981{
 982	WARN_ON_ONCE(!rcu_read_lock_held());
 983	return NULL;
 984}
 985#endif
 986
 987/*
 988 * Some inline functions in vmstat.h depend on page_zone()
 989 */
 990#include <linux/vmstat.h>
 991
 992static __always_inline void *lowmem_page_address(const struct page *page)
 993{
 994	return page_to_virt(page);
 995}
 996
 997#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 998#define HASHED_PAGE_VIRTUAL
 999#endif
1000
1001#if defined(WANT_PAGE_VIRTUAL)
1002static inline void *page_address(const struct page *page)
1003{
1004	return page->virtual;
1005}
1006static inline void set_page_address(struct page *page, void *address)
1007{
1008	page->virtual = address;
1009}
1010#define page_address_init()  do { } while(0)
1011#endif
1012
1013#if defined(HASHED_PAGE_VIRTUAL)
1014void *page_address(const struct page *page);
1015void set_page_address(struct page *page, void *virtual);
1016void page_address_init(void);
1017#endif
1018
1019#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1020#define page_address(page) lowmem_page_address(page)
1021#define set_page_address(page, address)  do { } while(0)
1022#define page_address_init()  do { } while(0)
1023#endif
1024
1025extern void *page_rmapping(struct page *page);
1026extern struct anon_vma *page_anon_vma(struct page *page);
1027extern struct address_space *page_mapping(struct page *page);
1028
1029extern struct address_space *__page_file_mapping(struct page *);
1030
1031static inline
1032struct address_space *page_file_mapping(struct page *page)
1033{
1034	if (unlikely(PageSwapCache(page)))
1035		return __page_file_mapping(page);
1036
1037	return page->mapping;
1038}
1039
1040extern pgoff_t __page_file_index(struct page *page);
1041
1042/*
1043 * Return the pagecache index of the passed page.  Regular pagecache pages
1044 * use ->index whereas swapcache pages use swp_offset(->private)
1045 */
1046static inline pgoff_t page_index(struct page *page)
1047{
1048	if (unlikely(PageSwapCache(page)))
 
 
 
 
 
 
 
 
 
 
 
 
 
1049		return __page_file_index(page);
 
1050	return page->index;
1051}
1052
1053bool page_mapped(struct page *page);
1054struct address_space *page_mapping(struct page *page);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1055
1056/*
1057 * Return true only if the page has been allocated with
1058 * ALLOC_NO_WATERMARKS and the low watermark was not
1059 * met implying that the system is under some pressure.
1060 */
1061static inline bool page_is_pfmemalloc(struct page *page)
1062{
1063	/*
1064	 * Page index cannot be this large so this must be
1065	 * a pfmemalloc page.
1066	 */
1067	return page->index == -1UL;
1068}
1069
1070/*
1071 * Only to be called by the page allocator on a freshly allocated
1072 * page.
1073 */
1074static inline void set_page_pfmemalloc(struct page *page)
1075{
1076	page->index = -1UL;
1077}
1078
1079static inline void clear_page_pfmemalloc(struct page *page)
1080{
1081	page->index = 0;
1082}
1083
1084/*
1085 * Different kinds of faults, as returned by handle_mm_fault().
1086 * Used to decide whether a process gets delivered SIGBUS or
1087 * just gets major/minor fault counters bumped up.
1088 */
1089
1090#define VM_FAULT_OOM	0x0001
1091#define VM_FAULT_SIGBUS	0x0002
1092#define VM_FAULT_MAJOR	0x0004
1093#define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
1094#define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
1095#define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1096#define VM_FAULT_SIGSEGV 0x0040
1097
1098#define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
1099#define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
1100#define VM_FAULT_RETRY	0x0400	/* ->fault blocked, must retry */
1101#define VM_FAULT_FALLBACK 0x0800	/* huge page fault failed, fall back to small */
1102#define VM_FAULT_DONE_COW   0x1000	/* ->fault has fully handled COW */
1103
1104#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1105
1106#define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1107			 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1108			 VM_FAULT_FALLBACK)
1109
1110/* Encode hstate index for a hwpoisoned large page */
1111#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1112#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1113
1114/*
1115 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1116 */
1117extern void pagefault_out_of_memory(void);
1118
1119#define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
1120
1121/*
1122 * Flags passed to show_mem() and show_free_areas() to suppress output in
1123 * various contexts.
1124 */
1125#define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
1126
1127extern void show_free_areas(unsigned int flags);
1128extern bool skip_free_areas_node(unsigned int flags, int nid);
1129
1130int shmem_zero_setup(struct vm_area_struct *);
1131#ifdef CONFIG_SHMEM
1132bool shmem_mapping(struct address_space *mapping);
1133#else
1134static inline bool shmem_mapping(struct address_space *mapping)
1135{
1136	return false;
1137}
1138#endif
1139
1140extern bool can_do_mlock(void);
1141extern int user_shm_lock(size_t, struct user_struct *);
1142extern void user_shm_unlock(size_t, struct user_struct *);
1143
1144/*
1145 * Parameter block passed down to zap_pte_range in exceptional cases.
1146 */
1147struct zap_details {
1148	struct address_space *check_mapping;	/* Check page->mapping if set */
1149	pgoff_t	first_index;			/* Lowest page->index to unmap */
1150	pgoff_t last_index;			/* Highest page->index to unmap */
1151	bool ignore_dirty;			/* Ignore dirty pages */
1152	bool check_swap_entries;		/* Check also swap entries */
1153};
1154
1155struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1156		pte_t pte);
1157struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1158				pmd_t pmd);
1159
1160int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1161		unsigned long size);
1162void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1163		unsigned long size, struct zap_details *);
1164void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1165		unsigned long start, unsigned long end);
1166
1167/**
1168 * mm_walk - callbacks for walk_page_range
1169 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1170 *	       this handler is required to be able to handle
1171 *	       pmd_trans_huge() pmds.  They may simply choose to
1172 *	       split_huge_page() instead of handling it explicitly.
1173 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1174 * @pte_hole: if set, called for each hole at all levels
1175 * @hugetlb_entry: if set, called for each hugetlb entry
1176 * @test_walk: caller specific callback function to determine whether
1177 *             we walk over the current vma or not. Returning 0
1178 *             value means "do page table walk over the current vma,"
1179 *             and a negative one means "abort current page table walk
1180 *             right now." 1 means "skip the current vma."
1181 * @mm:        mm_struct representing the target process of page table walk
1182 * @vma:       vma currently walked (NULL if walking outside vmas)
1183 * @private:   private data for callbacks' usage
1184 *
1185 * (see the comment on walk_page_range() for more details)
1186 */
1187struct mm_walk {
1188	int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1189			 unsigned long next, struct mm_walk *walk);
1190	int (*pte_entry)(pte_t *pte, unsigned long addr,
1191			 unsigned long next, struct mm_walk *walk);
1192	int (*pte_hole)(unsigned long addr, unsigned long next,
1193			struct mm_walk *walk);
1194	int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1195			     unsigned long addr, unsigned long next,
1196			     struct mm_walk *walk);
1197	int (*test_walk)(unsigned long addr, unsigned long next,
1198			struct mm_walk *walk);
1199	struct mm_struct *mm;
1200	struct vm_area_struct *vma;
1201	void *private;
1202};
1203
1204int walk_page_range(unsigned long addr, unsigned long end,
1205		struct mm_walk *walk);
1206int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1207void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1208		unsigned long end, unsigned long floor, unsigned long ceiling);
1209int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1210			struct vm_area_struct *vma);
1211void unmap_mapping_range(struct address_space *mapping,
1212		loff_t const holebegin, loff_t const holelen, int even_cows);
1213int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1214			     pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1215int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1216	unsigned long *pfn);
1217int follow_phys(struct vm_area_struct *vma, unsigned long address,
1218		unsigned int flags, unsigned long *prot, resource_size_t *phys);
1219int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1220			void *buf, int len, int write);
1221
1222static inline void unmap_shared_mapping_range(struct address_space *mapping,
1223		loff_t const holebegin, loff_t const holelen)
1224{
1225	unmap_mapping_range(mapping, holebegin, holelen, 0);
1226}
1227
1228extern void truncate_pagecache(struct inode *inode, loff_t new);
1229extern void truncate_setsize(struct inode *inode, loff_t newsize);
1230void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1231void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1232int truncate_inode_page(struct address_space *mapping, struct page *page);
1233int generic_error_remove_page(struct address_space *mapping, struct page *page);
1234int invalidate_inode_page(struct page *page);
1235
1236#ifdef CONFIG_MMU
1237extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1238		unsigned int flags);
1239extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1240			    unsigned long address, unsigned int fault_flags,
1241			    bool *unlocked);
1242#else
1243static inline int handle_mm_fault(struct vm_area_struct *vma,
1244		unsigned long address, unsigned int flags)
 
1245{
1246	/* should never happen if there's no MMU */
1247	BUG();
1248	return VM_FAULT_SIGBUS;
1249}
1250static inline int fixup_user_fault(struct task_struct *tsk,
1251		struct mm_struct *mm, unsigned long address,
1252		unsigned int fault_flags, bool *unlocked)
1253{
1254	/* should never happen if there's no MMU */
1255	BUG();
1256	return -EFAULT;
1257}
1258#endif
1259
1260extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1261		unsigned int gup_flags);
1262extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1263		void *buf, int len, unsigned int gup_flags);
1264extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1265		unsigned long addr, void *buf, int len, unsigned int gup_flags);
1266
 
 
 
 
1267long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1268			    unsigned long start, unsigned long nr_pages,
1269			    unsigned int gup_flags, struct page **pages,
1270			    struct vm_area_struct **vmas, int *locked);
1271long get_user_pages(unsigned long start, unsigned long nr_pages,
1272			    unsigned int gup_flags, struct page **pages,
1273			    struct vm_area_struct **vmas);
1274long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1275		    unsigned int gup_flags, struct page **pages, int *locked);
 
 
 
 
1276long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1277		    struct page **pages, unsigned int gup_flags);
1278int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1279			struct page **pages);
1280
1281/* Container for pinned pfns / pages */
1282struct frame_vector {
1283	unsigned int nr_allocated;	/* Number of frames we have space for */
1284	unsigned int nr_frames;	/* Number of frames stored in ptrs array */
1285	bool got_ref;		/* Did we pin pages by getting page ref? */
1286	bool is_pfns;		/* Does array contain pages or pfns? */
1287	void *ptrs[0];		/* Array of pinned pfns / pages. Use
1288				 * pfns_vector_pages() or pfns_vector_pfns()
1289				 * for access */
1290};
1291
1292struct frame_vector *frame_vector_create(unsigned int nr_frames);
1293void frame_vector_destroy(struct frame_vector *vec);
1294int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1295		     unsigned int gup_flags, struct frame_vector *vec);
1296void put_vaddr_frames(struct frame_vector *vec);
1297int frame_vector_to_pages(struct frame_vector *vec);
1298void frame_vector_to_pfns(struct frame_vector *vec);
1299
1300static inline unsigned int frame_vector_count(struct frame_vector *vec)
1301{
1302	return vec->nr_frames;
1303}
1304
1305static inline struct page **frame_vector_pages(struct frame_vector *vec)
1306{
1307	if (vec->is_pfns) {
1308		int err = frame_vector_to_pages(vec);
1309
1310		if (err)
1311			return ERR_PTR(err);
1312	}
1313	return (struct page **)(vec->ptrs);
1314}
1315
1316static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1317{
1318	if (!vec->is_pfns)
1319		frame_vector_to_pfns(vec);
1320	return (unsigned long *)(vec->ptrs);
1321}
1322
1323struct kvec;
1324int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1325			struct page **pages);
1326int get_kernel_page(unsigned long start, int write, struct page **pages);
1327struct page *get_dump_page(unsigned long addr);
1328
1329extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1330extern void do_invalidatepage(struct page *page, unsigned int offset,
1331			      unsigned int length);
1332
1333int __set_page_dirty_nobuffers(struct page *page);
1334int __set_page_dirty_no_writeback(struct page *page);
1335int redirty_page_for_writepage(struct writeback_control *wbc,
1336				struct page *page);
1337void account_page_dirtied(struct page *page, struct address_space *mapping);
1338void account_page_cleaned(struct page *page, struct address_space *mapping,
1339			  struct bdi_writeback *wb);
1340int set_page_dirty(struct page *page);
1341int set_page_dirty_lock(struct page *page);
1342void cancel_dirty_page(struct page *page);
1343int clear_page_dirty_for_io(struct page *page);
1344
1345int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1346
1347/* Is the vma a continuation of the stack vma above it? */
1348static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1349{
1350	return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1351}
1352
1353static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1354{
1355	return !vma->vm_ops;
1356}
1357
1358static inline int stack_guard_page_start(struct vm_area_struct *vma,
1359					     unsigned long addr)
1360{
1361	return (vma->vm_flags & VM_GROWSDOWN) &&
1362		(vma->vm_start == addr) &&
1363		!vma_growsdown(vma->vm_prev, addr);
1364}
1365
1366/* Is the vma a continuation of the stack vma below it? */
1367static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1368{
1369	return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1370}
1371
1372static inline int stack_guard_page_end(struct vm_area_struct *vma,
1373					   unsigned long addr)
1374{
1375	return (vma->vm_flags & VM_GROWSUP) &&
1376		(vma->vm_end == addr) &&
1377		!vma_growsup(vma->vm_next, addr);
1378}
1379
1380int vma_is_stack_for_current(struct vm_area_struct *vma);
1381
1382extern unsigned long move_page_tables(struct vm_area_struct *vma,
1383		unsigned long old_addr, struct vm_area_struct *new_vma,
1384		unsigned long new_addr, unsigned long len,
1385		bool need_rmap_locks);
1386extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1387			      unsigned long end, pgprot_t newprot,
1388			      int dirty_accountable, int prot_numa);
1389extern int mprotect_fixup(struct vm_area_struct *vma,
1390			  struct vm_area_struct **pprev, unsigned long start,
1391			  unsigned long end, unsigned long newflags);
1392
1393/*
1394 * doesn't attempt to fault and will return short.
1395 */
1396int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1397			  struct page **pages);
1398/*
1399 * per-process(per-mm_struct) statistics.
1400 */
1401static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1402{
1403	long val = atomic_long_read(&mm->rss_stat.count[member]);
1404
1405#ifdef SPLIT_RSS_COUNTING
1406	/*
1407	 * counter is updated in asynchronous manner and may go to minus.
1408	 * But it's never be expected number for users.
1409	 */
1410	if (val < 0)
1411		val = 0;
1412#endif
1413	return (unsigned long)val;
1414}
1415
1416static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1417{
1418	atomic_long_add(value, &mm->rss_stat.count[member]);
1419}
1420
1421static inline void inc_mm_counter(struct mm_struct *mm, int member)
1422{
1423	atomic_long_inc(&mm->rss_stat.count[member]);
1424}
1425
1426static inline void dec_mm_counter(struct mm_struct *mm, int member)
1427{
1428	atomic_long_dec(&mm->rss_stat.count[member]);
1429}
1430
1431/* Optimized variant when page is already known not to be PageAnon */
1432static inline int mm_counter_file(struct page *page)
1433{
1434	if (PageSwapBacked(page))
1435		return MM_SHMEMPAGES;
1436	return MM_FILEPAGES;
1437}
1438
1439static inline int mm_counter(struct page *page)
1440{
1441	if (PageAnon(page))
1442		return MM_ANONPAGES;
1443	return mm_counter_file(page);
1444}
1445
1446static inline unsigned long get_mm_rss(struct mm_struct *mm)
1447{
1448	return get_mm_counter(mm, MM_FILEPAGES) +
1449		get_mm_counter(mm, MM_ANONPAGES) +
1450		get_mm_counter(mm, MM_SHMEMPAGES);
1451}
1452
1453static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1454{
1455	return max(mm->hiwater_rss, get_mm_rss(mm));
1456}
1457
1458static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1459{
1460	return max(mm->hiwater_vm, mm->total_vm);
1461}
1462
1463static inline void update_hiwater_rss(struct mm_struct *mm)
1464{
1465	unsigned long _rss = get_mm_rss(mm);
1466
1467	if ((mm)->hiwater_rss < _rss)
1468		(mm)->hiwater_rss = _rss;
1469}
1470
1471static inline void update_hiwater_vm(struct mm_struct *mm)
1472{
1473	if (mm->hiwater_vm < mm->total_vm)
1474		mm->hiwater_vm = mm->total_vm;
1475}
1476
1477static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1478{
1479	mm->hiwater_rss = get_mm_rss(mm);
1480}
1481
1482static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1483					 struct mm_struct *mm)
1484{
1485	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1486
1487	if (*maxrss < hiwater_rss)
1488		*maxrss = hiwater_rss;
1489}
1490
1491#if defined(SPLIT_RSS_COUNTING)
1492void sync_mm_rss(struct mm_struct *mm);
1493#else
1494static inline void sync_mm_rss(struct mm_struct *mm)
1495{
1496}
1497#endif
1498
1499#ifndef __HAVE_ARCH_PTE_DEVMAP
1500static inline int pte_devmap(pte_t pte)
1501{
1502	return 0;
1503}
1504#endif
1505
1506int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1507
1508extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1509			       spinlock_t **ptl);
1510static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1511				    spinlock_t **ptl)
1512{
1513	pte_t *ptep;
1514	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1515	return ptep;
1516}
1517
1518#ifdef __PAGETABLE_PUD_FOLDED
1519static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1520						unsigned long address)
1521{
1522	return 0;
1523}
1524#else
1525int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1526#endif
1527
1528#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1529static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1530						unsigned long address)
1531{
1532	return 0;
1533}
1534
1535static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1536
1537static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1538{
1539	return 0;
1540}
1541
1542static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1543static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1544
1545#else
1546int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1547
1548static inline void mm_nr_pmds_init(struct mm_struct *mm)
1549{
1550	atomic_long_set(&mm->nr_pmds, 0);
1551}
1552
1553static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1554{
1555	return atomic_long_read(&mm->nr_pmds);
1556}
1557
1558static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1559{
1560	atomic_long_inc(&mm->nr_pmds);
1561}
1562
1563static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1564{
1565	atomic_long_dec(&mm->nr_pmds);
1566}
1567#endif
1568
1569int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1570int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1571
1572/*
1573 * The following ifdef needed to get the 4level-fixup.h header to work.
1574 * Remove it when 4level-fixup.h has been removed.
1575 */
1576#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1577static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1578{
1579	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1580		NULL: pud_offset(pgd, address);
1581}
1582
1583static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1584{
1585	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1586		NULL: pmd_offset(pud, address);
1587}
1588#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1589
1590#if USE_SPLIT_PTE_PTLOCKS
1591#if ALLOC_SPLIT_PTLOCKS
1592void __init ptlock_cache_init(void);
1593extern bool ptlock_alloc(struct page *page);
1594extern void ptlock_free(struct page *page);
1595
1596static inline spinlock_t *ptlock_ptr(struct page *page)
1597{
1598	return page->ptl;
1599}
1600#else /* ALLOC_SPLIT_PTLOCKS */
1601static inline void ptlock_cache_init(void)
1602{
1603}
1604
1605static inline bool ptlock_alloc(struct page *page)
1606{
1607	return true;
1608}
1609
1610static inline void ptlock_free(struct page *page)
1611{
1612}
1613
1614static inline spinlock_t *ptlock_ptr(struct page *page)
1615{
1616	return &page->ptl;
1617}
1618#endif /* ALLOC_SPLIT_PTLOCKS */
1619
1620static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1621{
1622	return ptlock_ptr(pmd_page(*pmd));
1623}
1624
1625static inline bool ptlock_init(struct page *page)
1626{
1627	/*
1628	 * prep_new_page() initialize page->private (and therefore page->ptl)
1629	 * with 0. Make sure nobody took it in use in between.
1630	 *
1631	 * It can happen if arch try to use slab for page table allocation:
1632	 * slab code uses page->slab_cache, which share storage with page->ptl.
1633	 */
1634	VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1635	if (!ptlock_alloc(page))
1636		return false;
1637	spin_lock_init(ptlock_ptr(page));
1638	return true;
1639}
1640
1641/* Reset page->mapping so free_pages_check won't complain. */
1642static inline void pte_lock_deinit(struct page *page)
1643{
1644	page->mapping = NULL;
1645	ptlock_free(page);
1646}
1647
1648#else	/* !USE_SPLIT_PTE_PTLOCKS */
1649/*
1650 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1651 */
1652static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1653{
1654	return &mm->page_table_lock;
1655}
1656static inline void ptlock_cache_init(void) {}
1657static inline bool ptlock_init(struct page *page) { return true; }
1658static inline void pte_lock_deinit(struct page *page) {}
1659#endif /* USE_SPLIT_PTE_PTLOCKS */
1660
1661static inline void pgtable_init(void)
1662{
1663	ptlock_cache_init();
1664	pgtable_cache_init();
1665}
1666
1667static inline bool pgtable_page_ctor(struct page *page)
1668{
1669	if (!ptlock_init(page))
1670		return false;
1671	inc_zone_page_state(page, NR_PAGETABLE);
1672	return true;
1673}
1674
1675static inline void pgtable_page_dtor(struct page *page)
1676{
1677	pte_lock_deinit(page);
1678	dec_zone_page_state(page, NR_PAGETABLE);
1679}
1680
1681#define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1682({							\
1683	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1684	pte_t *__pte = pte_offset_map(pmd, address);	\
1685	*(ptlp) = __ptl;				\
1686	spin_lock(__ptl);				\
1687	__pte;						\
1688})
1689
1690#define pte_unmap_unlock(pte, ptl)	do {		\
1691	spin_unlock(ptl);				\
1692	pte_unmap(pte);					\
1693} while (0)
1694
1695#define pte_alloc(mm, pmd, address)			\
1696	(unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1697
1698#define pte_alloc_map(mm, pmd, address)			\
1699	(pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1700
1701#define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1702	(pte_alloc(mm, pmd, address) ?			\
1703		 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1704
1705#define pte_alloc_kernel(pmd, address)			\
1706	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1707		NULL: pte_offset_kernel(pmd, address))
1708
1709#if USE_SPLIT_PMD_PTLOCKS
1710
1711static struct page *pmd_to_page(pmd_t *pmd)
1712{
1713	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1714	return virt_to_page((void *)((unsigned long) pmd & mask));
1715}
1716
1717static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1718{
1719	return ptlock_ptr(pmd_to_page(pmd));
1720}
1721
1722static inline bool pgtable_pmd_page_ctor(struct page *page)
1723{
1724#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1725	page->pmd_huge_pte = NULL;
1726#endif
1727	return ptlock_init(page);
1728}
1729
1730static inline void pgtable_pmd_page_dtor(struct page *page)
1731{
1732#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1733	VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1734#endif
1735	ptlock_free(page);
1736}
1737
1738#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1739
1740#else
1741
1742static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1743{
1744	return &mm->page_table_lock;
1745}
1746
1747static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1748static inline void pgtable_pmd_page_dtor(struct page *page) {}
1749
1750#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1751
1752#endif
1753
1754static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1755{
1756	spinlock_t *ptl = pmd_lockptr(mm, pmd);
1757	spin_lock(ptl);
1758	return ptl;
1759}
1760
1761extern void __init pagecache_init(void);
1762
1763extern void free_area_init(unsigned long * zones_size);
1764extern void free_area_init_node(int nid, unsigned long * zones_size,
1765		unsigned long zone_start_pfn, unsigned long *zholes_size);
1766extern void free_initmem(void);
1767
1768/*
1769 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1770 * into the buddy system. The freed pages will be poisoned with pattern
1771 * "poison" if it's within range [0, UCHAR_MAX].
1772 * Return pages freed into the buddy system.
1773 */
1774extern unsigned long free_reserved_area(void *start, void *end,
1775					int poison, char *s);
1776
1777#ifdef	CONFIG_HIGHMEM
1778/*
1779 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1780 * and totalram_pages.
1781 */
1782extern void free_highmem_page(struct page *page);
1783#endif
1784
1785extern void adjust_managed_page_count(struct page *page, long count);
1786extern void mem_init_print_info(const char *str);
1787
1788extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1789
1790/* Free the reserved page into the buddy system, so it gets managed. */
1791static inline void __free_reserved_page(struct page *page)
1792{
1793	ClearPageReserved(page);
1794	init_page_count(page);
1795	__free_page(page);
1796}
1797
1798static inline void free_reserved_page(struct page *page)
1799{
1800	__free_reserved_page(page);
1801	adjust_managed_page_count(page, 1);
1802}
1803
1804static inline void mark_page_reserved(struct page *page)
1805{
1806	SetPageReserved(page);
1807	adjust_managed_page_count(page, -1);
1808}
1809
1810/*
1811 * Default method to free all the __init memory into the buddy system.
1812 * The freed pages will be poisoned with pattern "poison" if it's within
1813 * range [0, UCHAR_MAX].
1814 * Return pages freed into the buddy system.
1815 */
1816static inline unsigned long free_initmem_default(int poison)
1817{
1818	extern char __init_begin[], __init_end[];
1819
1820	return free_reserved_area(&__init_begin, &__init_end,
1821				  poison, "unused kernel");
1822}
1823
1824static inline unsigned long get_num_physpages(void)
1825{
1826	int nid;
1827	unsigned long phys_pages = 0;
1828
1829	for_each_online_node(nid)
1830		phys_pages += node_present_pages(nid);
1831
1832	return phys_pages;
1833}
1834
1835#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1836/*
1837 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1838 * zones, allocate the backing mem_map and account for memory holes in a more
1839 * architecture independent manner. This is a substitute for creating the
1840 * zone_sizes[] and zholes_size[] arrays and passing them to
1841 * free_area_init_node()
1842 *
1843 * An architecture is expected to register range of page frames backed by
1844 * physical memory with memblock_add[_node]() before calling
1845 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1846 * usage, an architecture is expected to do something like
1847 *
1848 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1849 * 							 max_highmem_pfn};
1850 * for_each_valid_physical_page_range()
1851 * 	memblock_add_node(base, size, nid)
1852 * free_area_init_nodes(max_zone_pfns);
1853 *
1854 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1855 * registered physical page range.  Similarly
1856 * sparse_memory_present_with_active_regions() calls memory_present() for
1857 * each range when SPARSEMEM is enabled.
1858 *
1859 * See mm/page_alloc.c for more information on each function exposed by
1860 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1861 */
1862extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1863unsigned long node_map_pfn_alignment(void);
1864unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1865						unsigned long end_pfn);
1866extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1867						unsigned long end_pfn);
1868extern void get_pfn_range_for_nid(unsigned int nid,
1869			unsigned long *start_pfn, unsigned long *end_pfn);
1870extern unsigned long find_min_pfn_with_active_regions(void);
1871extern void free_bootmem_with_active_regions(int nid,
1872						unsigned long max_low_pfn);
1873extern void sparse_memory_present_with_active_regions(int nid);
1874
1875#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1876
1877#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1878    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1879static inline int __early_pfn_to_nid(unsigned long pfn,
1880					struct mminit_pfnnid_cache *state)
1881{
1882	return 0;
1883}
1884#else
1885/* please see mm/page_alloc.c */
1886extern int __meminit early_pfn_to_nid(unsigned long pfn);
1887/* there is a per-arch backend function. */
1888extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1889					struct mminit_pfnnid_cache *state);
1890#endif
1891
1892extern void set_dma_reserve(unsigned long new_dma_reserve);
1893extern void memmap_init_zone(unsigned long, int, unsigned long,
1894				unsigned long, enum memmap_context);
1895extern void setup_per_zone_wmarks(void);
1896extern int __meminit init_per_zone_wmark_min(void);
1897extern void mem_init(void);
1898extern void __init mmap_init(void);
1899extern void show_mem(unsigned int flags);
1900extern long si_mem_available(void);
1901extern void si_meminfo(struct sysinfo * val);
1902extern void si_meminfo_node(struct sysinfo *val, int nid);
1903#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
1904extern unsigned long arch_reserved_kernel_pages(void);
1905#endif
1906
1907extern __printf(2, 3)
1908void warn_alloc(gfp_t gfp_mask, const char *fmt, ...);
 
1909
1910extern void setup_per_cpu_pageset(void);
1911
1912extern void zone_pcp_update(struct zone *zone);
1913extern void zone_pcp_reset(struct zone *zone);
1914
1915/* page_alloc.c */
1916extern int min_free_kbytes;
1917extern int watermark_scale_factor;
1918
1919/* nommu.c */
1920extern atomic_long_t mmap_pages_allocated;
1921extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1922
1923/* interval_tree.c */
1924void vma_interval_tree_insert(struct vm_area_struct *node,
1925			      struct rb_root *root);
1926void vma_interval_tree_insert_after(struct vm_area_struct *node,
1927				    struct vm_area_struct *prev,
1928				    struct rb_root *root);
1929void vma_interval_tree_remove(struct vm_area_struct *node,
1930			      struct rb_root *root);
1931struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1932				unsigned long start, unsigned long last);
1933struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1934				unsigned long start, unsigned long last);
1935
1936#define vma_interval_tree_foreach(vma, root, start, last)		\
1937	for (vma = vma_interval_tree_iter_first(root, start, last);	\
1938	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
1939
1940void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1941				   struct rb_root *root);
1942void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1943				   struct rb_root *root);
1944struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1945	struct rb_root *root, unsigned long start, unsigned long last);
1946struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1947	struct anon_vma_chain *node, unsigned long start, unsigned long last);
1948#ifdef CONFIG_DEBUG_VM_RB
1949void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1950#endif
1951
1952#define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
1953	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1954	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1955
1956/* mmap.c */
1957extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1958extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
1959	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
1960	struct vm_area_struct *expand);
1961static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1962	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
1963{
1964	return __vma_adjust(vma, start, end, pgoff, insert, NULL);
1965}
1966extern struct vm_area_struct *vma_merge(struct mm_struct *,
1967	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1968	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1969	struct mempolicy *, struct vm_userfaultfd_ctx);
1970extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1971extern int split_vma(struct mm_struct *,
1972	struct vm_area_struct *, unsigned long addr, int new_below);
1973extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1974extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1975	struct rb_node **, struct rb_node *);
1976extern void unlink_file_vma(struct vm_area_struct *);
1977extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1978	unsigned long addr, unsigned long len, pgoff_t pgoff,
1979	bool *need_rmap_locks);
1980extern void exit_mmap(struct mm_struct *);
1981
1982static inline int check_data_rlimit(unsigned long rlim,
1983				    unsigned long new,
1984				    unsigned long start,
1985				    unsigned long end_data,
1986				    unsigned long start_data)
1987{
1988	if (rlim < RLIM_INFINITY) {
1989		if (((new - start) + (end_data - start_data)) > rlim)
1990			return -ENOSPC;
1991	}
1992
1993	return 0;
1994}
1995
1996extern int mm_take_all_locks(struct mm_struct *mm);
1997extern void mm_drop_all_locks(struct mm_struct *mm);
1998
1999extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2000extern struct file *get_mm_exe_file(struct mm_struct *mm);
2001extern struct file *get_task_exe_file(struct task_struct *task);
2002
2003extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2004extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2005
2006extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2007				   const struct vm_special_mapping *sm);
2008extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2009				   unsigned long addr, unsigned long len,
2010				   unsigned long flags,
2011				   const struct vm_special_mapping *spec);
2012/* This is an obsolete alternative to _install_special_mapping. */
2013extern int install_special_mapping(struct mm_struct *mm,
2014				   unsigned long addr, unsigned long len,
2015				   unsigned long flags, struct page **pages);
2016
2017extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2018
2019extern unsigned long mmap_region(struct file *file, unsigned long addr,
2020	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2021extern unsigned long do_mmap(struct file *file, unsigned long addr,
2022	unsigned long len, unsigned long prot, unsigned long flags,
2023	vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2024extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2025
2026static inline unsigned long
2027do_mmap_pgoff(struct file *file, unsigned long addr,
2028	unsigned long len, unsigned long prot, unsigned long flags,
2029	unsigned long pgoff, unsigned long *populate)
2030{
2031	return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2032}
2033
2034#ifdef CONFIG_MMU
2035extern int __mm_populate(unsigned long addr, unsigned long len,
2036			 int ignore_errors);
2037static inline void mm_populate(unsigned long addr, unsigned long len)
2038{
2039	/* Ignore errors */
2040	(void) __mm_populate(addr, len, 1);
2041}
2042#else
2043static inline void mm_populate(unsigned long addr, unsigned long len) {}
2044#endif
2045
2046/* These take the mm semaphore themselves */
2047extern int __must_check vm_brk(unsigned long, unsigned long);
2048extern int vm_munmap(unsigned long, size_t);
2049extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2050        unsigned long, unsigned long,
2051        unsigned long, unsigned long);
2052
2053struct vm_unmapped_area_info {
2054#define VM_UNMAPPED_AREA_TOPDOWN 1
2055	unsigned long flags;
2056	unsigned long length;
2057	unsigned long low_limit;
2058	unsigned long high_limit;
2059	unsigned long align_mask;
2060	unsigned long align_offset;
2061};
2062
2063extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2064extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2065
2066/*
2067 * Search for an unmapped address range.
2068 *
2069 * We are looking for a range that:
2070 * - does not intersect with any VMA;
2071 * - is contained within the [low_limit, high_limit) interval;
2072 * - is at least the desired size.
2073 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2074 */
2075static inline unsigned long
2076vm_unmapped_area(struct vm_unmapped_area_info *info)
2077{
2078	if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2079		return unmapped_area_topdown(info);
2080	else
2081		return unmapped_area(info);
2082}
2083
2084/* truncate.c */
2085extern void truncate_inode_pages(struct address_space *, loff_t);
2086extern void truncate_inode_pages_range(struct address_space *,
2087				       loff_t lstart, loff_t lend);
2088extern void truncate_inode_pages_final(struct address_space *);
2089
2090/* generic vm_area_ops exported for stackable file systems */
2091extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2092extern void filemap_map_pages(struct vm_fault *vmf,
2093		pgoff_t start_pgoff, pgoff_t end_pgoff);
2094extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2095
2096/* mm/page-writeback.c */
2097int write_one_page(struct page *page, int wait);
2098void task_dirty_inc(struct task_struct *tsk);
2099
2100/* readahead.c */
2101#define VM_MAX_READAHEAD	128	/* kbytes */
2102#define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
2103
2104int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2105			pgoff_t offset, unsigned long nr_to_read);
2106
2107void page_cache_sync_readahead(struct address_space *mapping,
2108			       struct file_ra_state *ra,
2109			       struct file *filp,
2110			       pgoff_t offset,
2111			       unsigned long size);
2112
2113void page_cache_async_readahead(struct address_space *mapping,
2114				struct file_ra_state *ra,
2115				struct file *filp,
2116				struct page *pg,
2117				pgoff_t offset,
2118				unsigned long size);
2119
2120/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2121extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2122
2123/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2124extern int expand_downwards(struct vm_area_struct *vma,
2125		unsigned long address);
2126#if VM_GROWSUP
2127extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2128#else
2129  #define expand_upwards(vma, address) (0)
2130#endif
2131
2132/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2133extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2134extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2135					     struct vm_area_struct **pprev);
2136
2137/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2138   NULL if none.  Assume start_addr < end_addr. */
2139static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2140{
2141	struct vm_area_struct * vma = find_vma(mm,start_addr);
2142
2143	if (vma && end_addr <= vma->vm_start)
2144		vma = NULL;
2145	return vma;
2146}
2147
2148static inline unsigned long vma_pages(struct vm_area_struct *vma)
2149{
2150	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2151}
2152
2153/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2154static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2155				unsigned long vm_start, unsigned long vm_end)
2156{
2157	struct vm_area_struct *vma = find_vma(mm, vm_start);
2158
2159	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2160		vma = NULL;
2161
2162	return vma;
2163}
2164
2165#ifdef CONFIG_MMU
2166pgprot_t vm_get_page_prot(unsigned long vm_flags);
2167void vma_set_page_prot(struct vm_area_struct *vma);
2168#else
2169static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2170{
2171	return __pgprot(0);
2172}
2173static inline void vma_set_page_prot(struct vm_area_struct *vma)
2174{
2175	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2176}
2177#endif
2178
2179#ifdef CONFIG_NUMA_BALANCING
2180unsigned long change_prot_numa(struct vm_area_struct *vma,
2181			unsigned long start, unsigned long end);
2182#endif
2183
2184struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2185int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2186			unsigned long pfn, unsigned long size, pgprot_t);
2187int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2188int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2189			unsigned long pfn);
2190int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2191			unsigned long pfn, pgprot_t pgprot);
2192int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2193			pfn_t pfn);
2194int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2195
2196
2197struct page *follow_page_mask(struct vm_area_struct *vma,
2198			      unsigned long address, unsigned int foll_flags,
2199			      unsigned int *page_mask);
2200
2201static inline struct page *follow_page(struct vm_area_struct *vma,
2202		unsigned long address, unsigned int foll_flags)
2203{
2204	unsigned int unused_page_mask;
2205	return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2206}
2207
2208#define FOLL_WRITE	0x01	/* check pte is writable */
2209#define FOLL_TOUCH	0x02	/* mark page accessed */
2210#define FOLL_GET	0x04	/* do get_page on page */
2211#define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
2212#define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
2213#define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
2214				 * and return without waiting upon it */
2215#define FOLL_POPULATE	0x40	/* fault in page */
2216#define FOLL_SPLIT	0x80	/* don't return transhuge pages, split them */
2217#define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
2218#define FOLL_NUMA	0x200	/* force NUMA hinting page fault */
2219#define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
2220#define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
2221#define FOLL_MLOCK	0x1000	/* lock present pages */
2222#define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
2223#define FOLL_COW	0x4000	/* internal GUP flag */
2224
2225typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2226			void *data);
2227extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2228			       unsigned long size, pte_fn_t fn, void *data);
2229
2230
2231#ifdef CONFIG_PAGE_POISONING
2232extern bool page_poisoning_enabled(void);
2233extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2234extern bool page_is_poisoned(struct page *page);
2235#else
2236static inline bool page_poisoning_enabled(void) { return false; }
2237static inline void kernel_poison_pages(struct page *page, int numpages,
2238					int enable) { }
2239static inline bool page_is_poisoned(struct page *page) { return false; }
2240#endif
2241
2242#ifdef CONFIG_DEBUG_PAGEALLOC
2243extern bool _debug_pagealloc_enabled;
2244extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2245
2246static inline bool debug_pagealloc_enabled(void)
2247{
2248	return _debug_pagealloc_enabled;
2249}
2250
2251static inline void
2252kernel_map_pages(struct page *page, int numpages, int enable)
2253{
2254	if (!debug_pagealloc_enabled())
2255		return;
2256
2257	__kernel_map_pages(page, numpages, enable);
2258}
2259#ifdef CONFIG_HIBERNATION
2260extern bool kernel_page_present(struct page *page);
2261#endif	/* CONFIG_HIBERNATION */
2262#else	/* CONFIG_DEBUG_PAGEALLOC */
2263static inline void
2264kernel_map_pages(struct page *page, int numpages, int enable) {}
2265#ifdef CONFIG_HIBERNATION
2266static inline bool kernel_page_present(struct page *page) { return true; }
2267#endif	/* CONFIG_HIBERNATION */
2268static inline bool debug_pagealloc_enabled(void)
2269{
2270	return false;
2271}
2272#endif	/* CONFIG_DEBUG_PAGEALLOC */
2273
2274#ifdef __HAVE_ARCH_GATE_AREA
2275extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2276extern int in_gate_area_no_mm(unsigned long addr);
2277extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2278#else
2279static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2280{
2281	return NULL;
2282}
2283static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2284static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2285{
2286	return 0;
2287}
2288#endif	/* __HAVE_ARCH_GATE_AREA */
2289
2290extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2291
2292#ifdef CONFIG_SYSCTL
2293extern int sysctl_drop_caches;
2294int drop_caches_sysctl_handler(struct ctl_table *, int,
2295					void __user *, size_t *, loff_t *);
2296#endif
2297
2298void drop_slab(void);
2299void drop_slab_node(int nid);
2300
2301#ifndef CONFIG_MMU
2302#define randomize_va_space 0
2303#else
2304extern int randomize_va_space;
2305#endif
2306
2307const char * arch_vma_name(struct vm_area_struct *vma);
2308void print_vma_addr(char *prefix, unsigned long rip);
2309
2310void sparse_mem_maps_populate_node(struct page **map_map,
2311				   unsigned long pnum_begin,
2312				   unsigned long pnum_end,
2313				   unsigned long map_count,
2314				   int nodeid);
2315
2316struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2317pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2318pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2319pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2320pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2321void *vmemmap_alloc_block(unsigned long size, int node);
2322struct vmem_altmap;
2323void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2324		struct vmem_altmap *altmap);
2325static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2326{
2327	return __vmemmap_alloc_block_buf(size, node, NULL);
2328}
2329
2330void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2331int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2332			       int node);
2333int vmemmap_populate(unsigned long start, unsigned long end, int node);
2334void vmemmap_populate_print_last(void);
2335#ifdef CONFIG_MEMORY_HOTPLUG
2336void vmemmap_free(unsigned long start, unsigned long end);
2337#endif
2338void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2339				  unsigned long size);
2340
2341enum mf_flags {
2342	MF_COUNT_INCREASED = 1 << 0,
2343	MF_ACTION_REQUIRED = 1 << 1,
2344	MF_MUST_KILL = 1 << 2,
2345	MF_SOFT_OFFLINE = 1 << 3,
2346};
2347extern int memory_failure(unsigned long pfn, int trapno, int flags);
2348extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2349extern int unpoison_memory(unsigned long pfn);
2350extern int get_hwpoison_page(struct page *page);
2351#define put_hwpoison_page(page)	put_page(page)
2352extern int sysctl_memory_failure_early_kill;
2353extern int sysctl_memory_failure_recovery;
2354extern void shake_page(struct page *p, int access);
2355extern atomic_long_t num_poisoned_pages;
2356extern int soft_offline_page(struct page *page, int flags);
2357
2358
2359/*
2360 * Error handlers for various types of pages.
2361 */
2362enum mf_result {
2363	MF_IGNORED,	/* Error: cannot be handled */
2364	MF_FAILED,	/* Error: handling failed */
2365	MF_DELAYED,	/* Will be handled later */
2366	MF_RECOVERED,	/* Successfully recovered */
2367};
2368
2369enum mf_action_page_type {
2370	MF_MSG_KERNEL,
2371	MF_MSG_KERNEL_HIGH_ORDER,
2372	MF_MSG_SLAB,
2373	MF_MSG_DIFFERENT_COMPOUND,
2374	MF_MSG_POISONED_HUGE,
2375	MF_MSG_HUGE,
2376	MF_MSG_FREE_HUGE,
2377	MF_MSG_UNMAP_FAILED,
2378	MF_MSG_DIRTY_SWAPCACHE,
2379	MF_MSG_CLEAN_SWAPCACHE,
2380	MF_MSG_DIRTY_MLOCKED_LRU,
2381	MF_MSG_CLEAN_MLOCKED_LRU,
2382	MF_MSG_DIRTY_UNEVICTABLE_LRU,
2383	MF_MSG_CLEAN_UNEVICTABLE_LRU,
2384	MF_MSG_DIRTY_LRU,
2385	MF_MSG_CLEAN_LRU,
2386	MF_MSG_TRUNCATED_LRU,
2387	MF_MSG_BUDDY,
2388	MF_MSG_BUDDY_2ND,
2389	MF_MSG_UNKNOWN,
2390};
2391
2392#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2393extern void clear_huge_page(struct page *page,
2394			    unsigned long addr,
2395			    unsigned int pages_per_huge_page);
2396extern void copy_user_huge_page(struct page *dst, struct page *src,
2397				unsigned long addr, struct vm_area_struct *vma,
2398				unsigned int pages_per_huge_page);
2399#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2400
2401extern struct page_ext_operations debug_guardpage_ops;
2402extern struct page_ext_operations page_poisoning_ops;
2403
2404#ifdef CONFIG_DEBUG_PAGEALLOC
2405extern unsigned int _debug_guardpage_minorder;
2406extern bool _debug_guardpage_enabled;
2407
2408static inline unsigned int debug_guardpage_minorder(void)
2409{
2410	return _debug_guardpage_minorder;
2411}
2412
2413static inline bool debug_guardpage_enabled(void)
2414{
2415	return _debug_guardpage_enabled;
2416}
2417
2418static inline bool page_is_guard(struct page *page)
2419{
2420	struct page_ext *page_ext;
2421
2422	if (!debug_guardpage_enabled())
2423		return false;
2424
2425	page_ext = lookup_page_ext(page);
2426	if (unlikely(!page_ext))
2427		return false;
2428
2429	return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2430}
2431#else
2432static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2433static inline bool debug_guardpage_enabled(void) { return false; }
2434static inline bool page_is_guard(struct page *page) { return false; }
2435#endif /* CONFIG_DEBUG_PAGEALLOC */
2436
2437#if MAX_NUMNODES > 1
2438void __init setup_nr_node_ids(void);
2439#else
2440static inline void setup_nr_node_ids(void) {}
2441#endif
2442
2443#endif /* __KERNEL__ */
2444#endif /* _LINUX_MM_H */
v4.6
   1#ifndef _LINUX_MM_H
   2#define _LINUX_MM_H
   3
   4#include <linux/errno.h>
   5
   6#ifdef __KERNEL__
   7
   8#include <linux/mmdebug.h>
   9#include <linux/gfp.h>
  10#include <linux/bug.h>
  11#include <linux/list.h>
  12#include <linux/mmzone.h>
  13#include <linux/rbtree.h>
  14#include <linux/atomic.h>
  15#include <linux/debug_locks.h>
  16#include <linux/mm_types.h>
  17#include <linux/range.h>
  18#include <linux/pfn.h>
  19#include <linux/percpu-refcount.h>
  20#include <linux/bit_spinlock.h>
  21#include <linux/shrinker.h>
  22#include <linux/resource.h>
  23#include <linux/page_ext.h>
  24#include <linux/err.h>
  25#include <linux/page_ref.h>
  26
  27struct mempolicy;
  28struct anon_vma;
  29struct anon_vma_chain;
  30struct file_ra_state;
  31struct user_struct;
  32struct writeback_control;
  33struct bdi_writeback;
  34
  35#ifndef CONFIG_NEED_MULTIPLE_NODES	/* Don't use mapnrs, do it properly */
  36extern unsigned long max_mapnr;
  37
  38static inline void set_max_mapnr(unsigned long limit)
  39{
  40	max_mapnr = limit;
  41}
  42#else
  43static inline void set_max_mapnr(unsigned long limit) { }
  44#endif
  45
  46extern unsigned long totalram_pages;
  47extern void * high_memory;
  48extern int page_cluster;
  49
  50#ifdef CONFIG_SYSCTL
  51extern int sysctl_legacy_va_layout;
  52#else
  53#define sysctl_legacy_va_layout 0
  54#endif
  55
  56#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
  57extern const int mmap_rnd_bits_min;
  58extern const int mmap_rnd_bits_max;
  59extern int mmap_rnd_bits __read_mostly;
  60#endif
  61#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
  62extern const int mmap_rnd_compat_bits_min;
  63extern const int mmap_rnd_compat_bits_max;
  64extern int mmap_rnd_compat_bits __read_mostly;
  65#endif
  66
  67#include <asm/page.h>
  68#include <asm/pgtable.h>
  69#include <asm/processor.h>
  70
  71#ifndef __pa_symbol
  72#define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
  73#endif
  74
 
 
 
 
  75/*
  76 * To prevent common memory management code establishing
  77 * a zero page mapping on a read fault.
  78 * This macro should be defined within <asm/pgtable.h>.
  79 * s390 does this to prevent multiplexing of hardware bits
  80 * related to the physical page in case of virtualization.
  81 */
  82#ifndef mm_forbids_zeropage
  83#define mm_forbids_zeropage(X)	(0)
  84#endif
  85
  86/*
  87 * Default maximum number of active map areas, this limits the number of vmas
  88 * per mm struct. Users can overwrite this number by sysctl but there is a
  89 * problem.
  90 *
  91 * When a program's coredump is generated as ELF format, a section is created
  92 * per a vma. In ELF, the number of sections is represented in unsigned short.
  93 * This means the number of sections should be smaller than 65535 at coredump.
  94 * Because the kernel adds some informative sections to a image of program at
  95 * generating coredump, we need some margin. The number of extra sections is
  96 * 1-3 now and depends on arch. We use "5" as safe margin, here.
  97 *
  98 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
  99 * not a hard limit any more. Although some userspace tools can be surprised by
 100 * that.
 101 */
 102#define MAPCOUNT_ELF_CORE_MARGIN	(5)
 103#define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
 104
 105extern int sysctl_max_map_count;
 106
 107extern unsigned long sysctl_user_reserve_kbytes;
 108extern unsigned long sysctl_admin_reserve_kbytes;
 109
 110extern int sysctl_overcommit_memory;
 111extern int sysctl_overcommit_ratio;
 112extern unsigned long sysctl_overcommit_kbytes;
 113
 114extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
 115				    size_t *, loff_t *);
 116extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
 117				    size_t *, loff_t *);
 118
 119#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
 120
 121/* to align the pointer to the (next) page boundary */
 122#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
 123
 124/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
 125#define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
 126
 127/*
 128 * Linux kernel virtual memory manager primitives.
 129 * The idea being to have a "virtual" mm in the same way
 130 * we have a virtual fs - giving a cleaner interface to the
 131 * mm details, and allowing different kinds of memory mappings
 132 * (from shared memory to executable loading to arbitrary
 133 * mmap() functions).
 134 */
 135
 136extern struct kmem_cache *vm_area_cachep;
 137
 138#ifndef CONFIG_MMU
 139extern struct rb_root nommu_region_tree;
 140extern struct rw_semaphore nommu_region_sem;
 141
 142extern unsigned int kobjsize(const void *objp);
 143#endif
 144
 145/*
 146 * vm_flags in vm_area_struct, see mm_types.h.
 147 * When changing, update also include/trace/events/mmflags.h
 148 */
 149#define VM_NONE		0x00000000
 150
 151#define VM_READ		0x00000001	/* currently active flags */
 152#define VM_WRITE	0x00000002
 153#define VM_EXEC		0x00000004
 154#define VM_SHARED	0x00000008
 155
 156/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
 157#define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
 158#define VM_MAYWRITE	0x00000020
 159#define VM_MAYEXEC	0x00000040
 160#define VM_MAYSHARE	0x00000080
 161
 162#define VM_GROWSDOWN	0x00000100	/* general info on the segment */
 163#define VM_UFFD_MISSING	0x00000200	/* missing pages tracking */
 164#define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
 165#define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
 166#define VM_UFFD_WP	0x00001000	/* wrprotect pages tracking */
 167
 168#define VM_LOCKED	0x00002000
 169#define VM_IO           0x00004000	/* Memory mapped I/O or similar */
 170
 171					/* Used by sys_madvise() */
 172#define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
 173#define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
 174
 175#define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
 176#define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
 177#define VM_LOCKONFAULT	0x00080000	/* Lock the pages covered when they are faulted in */
 178#define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
 179#define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
 180#define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
 181#define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
 182#define VM_ARCH_2	0x02000000
 183#define VM_DONTDUMP	0x04000000	/* Do not include in the core dump */
 184
 185#ifdef CONFIG_MEM_SOFT_DIRTY
 186# define VM_SOFTDIRTY	0x08000000	/* Not soft dirty clean area */
 187#else
 188# define VM_SOFTDIRTY	0
 189#endif
 190
 191#define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
 192#define VM_HUGEPAGE	0x20000000	/* MADV_HUGEPAGE marked this vma */
 193#define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
 194#define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
 195
 196#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
 197#define VM_HIGH_ARCH_BIT_0	32	/* bit only usable on 64-bit architectures */
 198#define VM_HIGH_ARCH_BIT_1	33	/* bit only usable on 64-bit architectures */
 199#define VM_HIGH_ARCH_BIT_2	34	/* bit only usable on 64-bit architectures */
 200#define VM_HIGH_ARCH_BIT_3	35	/* bit only usable on 64-bit architectures */
 201#define VM_HIGH_ARCH_0	BIT(VM_HIGH_ARCH_BIT_0)
 202#define VM_HIGH_ARCH_1	BIT(VM_HIGH_ARCH_BIT_1)
 203#define VM_HIGH_ARCH_2	BIT(VM_HIGH_ARCH_BIT_2)
 204#define VM_HIGH_ARCH_3	BIT(VM_HIGH_ARCH_BIT_3)
 205#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
 206
 207#if defined(CONFIG_X86)
 208# define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
 209#if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
 210# define VM_PKEY_SHIFT	VM_HIGH_ARCH_BIT_0
 211# define VM_PKEY_BIT0	VM_HIGH_ARCH_0	/* A protection key is a 4-bit value */
 212# define VM_PKEY_BIT1	VM_HIGH_ARCH_1
 213# define VM_PKEY_BIT2	VM_HIGH_ARCH_2
 214# define VM_PKEY_BIT3	VM_HIGH_ARCH_3
 215#endif
 216#elif defined(CONFIG_PPC)
 217# define VM_SAO		VM_ARCH_1	/* Strong Access Ordering (powerpc) */
 218#elif defined(CONFIG_PARISC)
 219# define VM_GROWSUP	VM_ARCH_1
 220#elif defined(CONFIG_METAG)
 221# define VM_GROWSUP	VM_ARCH_1
 222#elif defined(CONFIG_IA64)
 223# define VM_GROWSUP	VM_ARCH_1
 224#elif !defined(CONFIG_MMU)
 225# define VM_MAPPED_COPY	VM_ARCH_1	/* T if mapped copy of data (nommu mmap) */
 226#endif
 227
 228#if defined(CONFIG_X86)
 229/* MPX specific bounds table or bounds directory */
 230# define VM_MPX		VM_ARCH_2
 231#endif
 232
 233#ifndef VM_GROWSUP
 234# define VM_GROWSUP	VM_NONE
 235#endif
 236
 237/* Bits set in the VMA until the stack is in its final location */
 238#define VM_STACK_INCOMPLETE_SETUP	(VM_RAND_READ | VM_SEQ_READ)
 239
 240#ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
 241#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 242#endif
 243
 244#ifdef CONFIG_STACK_GROWSUP
 245#define VM_STACK	VM_GROWSUP
 246#else
 247#define VM_STACK	VM_GROWSDOWN
 248#endif
 249
 250#define VM_STACK_FLAGS	(VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 251
 252/*
 253 * Special vmas that are non-mergable, non-mlock()able.
 254 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
 255 */
 256#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
 257
 258/* This mask defines which mm->def_flags a process can inherit its parent */
 259#define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
 260
 261/* This mask is used to clear all the VMA flags used by mlock */
 262#define VM_LOCKED_CLEAR_MASK	(~(VM_LOCKED | VM_LOCKONFAULT))
 263
 264/*
 265 * mapping from the currently active vm_flags protection bits (the
 266 * low four bits) to a page protection mask..
 267 */
 268extern pgprot_t protection_map[16];
 269
 270#define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
 271#define FAULT_FLAG_MKWRITE	0x02	/* Fault was mkwrite of existing pte */
 272#define FAULT_FLAG_ALLOW_RETRY	0x04	/* Retry fault if blocking */
 273#define FAULT_FLAG_RETRY_NOWAIT	0x08	/* Don't drop mmap_sem and wait when retrying */
 274#define FAULT_FLAG_KILLABLE	0x10	/* The fault task is in SIGKILL killable region */
 275#define FAULT_FLAG_TRIED	0x20	/* Second try */
 276#define FAULT_FLAG_USER		0x40	/* The fault originated in userspace */
 277#define FAULT_FLAG_REMOTE	0x80	/* faulting for non current tsk/mm */
 278#define FAULT_FLAG_INSTRUCTION  0x100	/* The fault was during an instruction fetch */
 279
 280/*
 281 * vm_fault is filled by the the pagefault handler and passed to the vma's
 282 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 283 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 284 *
 285 * MM layer fills up gfp_mask for page allocations but fault handler might
 286 * alter it if its implementation requires a different allocation context.
 287 *
 288 * pgoff should be used in favour of virtual_address, if possible.
 289 */
 290struct vm_fault {
 
 291	unsigned int flags;		/* FAULT_FLAG_xxx flags */
 292	gfp_t gfp_mask;			/* gfp mask to be used for allocations */
 293	pgoff_t pgoff;			/* Logical page offset based on vma */
 294	void __user *virtual_address;	/* Faulting virtual address */
 
 
 
 295
 296	struct page *cow_page;		/* Handler may choose to COW */
 
 297	struct page *page;		/* ->fault handlers should return a
 298					 * page here, unless VM_FAULT_NOPAGE
 299					 * is set (which is also implied by
 300					 * VM_FAULT_ERROR).
 301					 */
 302	/* for ->map_pages() only */
 303	pgoff_t max_pgoff;		/* map pages for offset from pgoff till
 304					 * max_pgoff inclusive */
 305	pte_t *pte;			/* pte entry associated with ->pgoff */
 
 
 
 
 
 
 
 
 
 
 
 
 306};
 307
 308/*
 309 * These are the virtual MM functions - opening of an area, closing and
 310 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 311 * to the functions called when a no-page or a wp-page exception occurs. 
 312 */
 313struct vm_operations_struct {
 314	void (*open)(struct vm_area_struct * area);
 315	void (*close)(struct vm_area_struct * area);
 316	int (*mremap)(struct vm_area_struct * area);
 317	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 318	int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
 319						pmd_t *, unsigned int flags);
 320	void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
 
 321
 322	/* notification that a previously read-only page is about to become
 323	 * writable, if an error is returned it will cause a SIGBUS */
 324	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 325
 326	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
 327	int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 328
 329	/* called by access_process_vm when get_user_pages() fails, typically
 330	 * for use by special VMAs that can switch between memory and hardware
 331	 */
 332	int (*access)(struct vm_area_struct *vma, unsigned long addr,
 333		      void *buf, int len, int write);
 334
 335	/* Called by the /proc/PID/maps code to ask the vma whether it
 336	 * has a special name.  Returning non-NULL will also cause this
 337	 * vma to be dumped unconditionally. */
 338	const char *(*name)(struct vm_area_struct *vma);
 339
 340#ifdef CONFIG_NUMA
 341	/*
 342	 * set_policy() op must add a reference to any non-NULL @new mempolicy
 343	 * to hold the policy upon return.  Caller should pass NULL @new to
 344	 * remove a policy and fall back to surrounding context--i.e. do not
 345	 * install a MPOL_DEFAULT policy, nor the task or system default
 346	 * mempolicy.
 347	 */
 348	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 349
 350	/*
 351	 * get_policy() op must add reference [mpol_get()] to any policy at
 352	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 353	 * in mm/mempolicy.c will do this automatically.
 354	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 355	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 356	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 357	 * must return NULL--i.e., do not "fallback" to task or system default
 358	 * policy.
 359	 */
 360	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 361					unsigned long addr);
 362#endif
 363	/*
 364	 * Called by vm_normal_page() for special PTEs to find the
 365	 * page for @addr.  This is useful if the default behavior
 366	 * (using pte_page()) would not find the correct page.
 367	 */
 368	struct page *(*find_special_page)(struct vm_area_struct *vma,
 369					  unsigned long addr);
 370};
 371
 372struct mmu_gather;
 373struct inode;
 374
 375#define page_private(page)		((page)->private)
 376#define set_page_private(page, v)	((page)->private = (v))
 377
 378#if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
 379static inline int pmd_devmap(pmd_t pmd)
 380{
 381	return 0;
 382}
 383#endif
 384
 385/*
 386 * FIXME: take this include out, include page-flags.h in
 387 * files which need it (119 of them)
 388 */
 389#include <linux/page-flags.h>
 390#include <linux/huge_mm.h>
 391
 392/*
 393 * Methods to modify the page usage count.
 394 *
 395 * What counts for a page usage:
 396 * - cache mapping   (page->mapping)
 397 * - private data    (page->private)
 398 * - page mapped in a task's page tables, each mapping
 399 *   is counted separately
 400 *
 401 * Also, many kernel routines increase the page count before a critical
 402 * routine so they can be sure the page doesn't go away from under them.
 403 */
 404
 405/*
 406 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 407 */
 408static inline int put_page_testzero(struct page *page)
 409{
 410	VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
 411	return page_ref_dec_and_test(page);
 412}
 413
 414/*
 415 * Try to grab a ref unless the page has a refcount of zero, return false if
 416 * that is the case.
 417 * This can be called when MMU is off so it must not access
 418 * any of the virtual mappings.
 419 */
 420static inline int get_page_unless_zero(struct page *page)
 421{
 422	return page_ref_add_unless(page, 1, 0);
 423}
 424
 425extern int page_is_ram(unsigned long pfn);
 426
 427enum {
 428	REGION_INTERSECTS,
 429	REGION_DISJOINT,
 430	REGION_MIXED,
 431};
 432
 433int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
 434		      unsigned long desc);
 435
 436/* Support for virtually mapped pages */
 437struct page *vmalloc_to_page(const void *addr);
 438unsigned long vmalloc_to_pfn(const void *addr);
 439
 440/*
 441 * Determine if an address is within the vmalloc range
 442 *
 443 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 444 * is no special casing required.
 445 */
 446static inline int is_vmalloc_addr(const void *x)
 447{
 448#ifdef CONFIG_MMU
 449	unsigned long addr = (unsigned long)x;
 450
 451	return addr >= VMALLOC_START && addr < VMALLOC_END;
 452#else
 453	return 0;
 454#endif
 455}
 456#ifdef CONFIG_MMU
 457extern int is_vmalloc_or_module_addr(const void *x);
 458#else
 459static inline int is_vmalloc_or_module_addr(const void *x)
 460{
 461	return 0;
 462}
 463#endif
 464
 465extern void kvfree(const void *addr);
 466
 467static inline atomic_t *compound_mapcount_ptr(struct page *page)
 468{
 469	return &page[1].compound_mapcount;
 470}
 471
 472static inline int compound_mapcount(struct page *page)
 473{
 474	if (!PageCompound(page))
 475		return 0;
 476	page = compound_head(page);
 477	return atomic_read(compound_mapcount_ptr(page)) + 1;
 478}
 479
 480/*
 481 * The atomic page->_mapcount, starts from -1: so that transitions
 482 * both from it and to it can be tracked, using atomic_inc_and_test
 483 * and atomic_add_negative(-1).
 484 */
 485static inline void page_mapcount_reset(struct page *page)
 486{
 487	atomic_set(&(page)->_mapcount, -1);
 488}
 489
 490int __page_mapcount(struct page *page);
 491
 492static inline int page_mapcount(struct page *page)
 493{
 494	VM_BUG_ON_PAGE(PageSlab(page), page);
 495
 496	if (unlikely(PageCompound(page)))
 497		return __page_mapcount(page);
 498	return atomic_read(&page->_mapcount) + 1;
 499}
 500
 501#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 502int total_mapcount(struct page *page);
 503int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
 504#else
 505static inline int total_mapcount(struct page *page)
 506{
 507	return page_mapcount(page);
 508}
 509static inline int page_trans_huge_mapcount(struct page *page,
 510					   int *total_mapcount)
 511{
 512	int mapcount = page_mapcount(page);
 513	if (total_mapcount)
 514		*total_mapcount = mapcount;
 515	return mapcount;
 516}
 517#endif
 518
 519static inline struct page *virt_to_head_page(const void *x)
 520{
 521	struct page *page = virt_to_page(x);
 522
 523	return compound_head(page);
 524}
 525
 526void __put_page(struct page *page);
 527
 528void put_pages_list(struct list_head *pages);
 529
 530void split_page(struct page *page, unsigned int order);
 531int split_free_page(struct page *page);
 532
 533/*
 534 * Compound pages have a destructor function.  Provide a
 535 * prototype for that function and accessor functions.
 536 * These are _only_ valid on the head of a compound page.
 537 */
 538typedef void compound_page_dtor(struct page *);
 539
 540/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
 541enum compound_dtor_id {
 542	NULL_COMPOUND_DTOR,
 543	COMPOUND_PAGE_DTOR,
 544#ifdef CONFIG_HUGETLB_PAGE
 545	HUGETLB_PAGE_DTOR,
 546#endif
 547#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 548	TRANSHUGE_PAGE_DTOR,
 549#endif
 550	NR_COMPOUND_DTORS,
 551};
 552extern compound_page_dtor * const compound_page_dtors[];
 553
 554static inline void set_compound_page_dtor(struct page *page,
 555		enum compound_dtor_id compound_dtor)
 556{
 557	VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
 558	page[1].compound_dtor = compound_dtor;
 559}
 560
 561static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 562{
 563	VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
 564	return compound_page_dtors[page[1].compound_dtor];
 565}
 566
 567static inline unsigned int compound_order(struct page *page)
 568{
 569	if (!PageHead(page))
 570		return 0;
 571	return page[1].compound_order;
 572}
 573
 574static inline void set_compound_order(struct page *page, unsigned int order)
 575{
 576	page[1].compound_order = order;
 577}
 578
 579void free_compound_page(struct page *page);
 580
 581#ifdef CONFIG_MMU
 582/*
 583 * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
 584 * servicing faults for write access.  In the normal case, do always want
 585 * pte_mkwrite.  But get_user_pages can cause write faults for mappings
 586 * that do not have writing enabled, when used by access_process_vm.
 587 */
 588static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 589{
 590	if (likely(vma->vm_flags & VM_WRITE))
 591		pte = pte_mkwrite(pte);
 592	return pte;
 593}
 594
 595void do_set_pte(struct vm_area_struct *vma, unsigned long address,
 596		struct page *page, pte_t *pte, bool write, bool anon);
 
 
 597#endif
 598
 599/*
 600 * Multiple processes may "see" the same page. E.g. for untouched
 601 * mappings of /dev/null, all processes see the same page full of
 602 * zeroes, and text pages of executables and shared libraries have
 603 * only one copy in memory, at most, normally.
 604 *
 605 * For the non-reserved pages, page_count(page) denotes a reference count.
 606 *   page_count() == 0 means the page is free. page->lru is then used for
 607 *   freelist management in the buddy allocator.
 608 *   page_count() > 0  means the page has been allocated.
 609 *
 610 * Pages are allocated by the slab allocator in order to provide memory
 611 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 612 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 613 * unless a particular usage is carefully commented. (the responsibility of
 614 * freeing the kmalloc memory is the caller's, of course).
 615 *
 616 * A page may be used by anyone else who does a __get_free_page().
 617 * In this case, page_count still tracks the references, and should only
 618 * be used through the normal accessor functions. The top bits of page->flags
 619 * and page->virtual store page management information, but all other fields
 620 * are unused and could be used privately, carefully. The management of this
 621 * page is the responsibility of the one who allocated it, and those who have
 622 * subsequently been given references to it.
 623 *
 624 * The other pages (we may call them "pagecache pages") are completely
 625 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 626 * The following discussion applies only to them.
 627 *
 628 * A pagecache page contains an opaque `private' member, which belongs to the
 629 * page's address_space. Usually, this is the address of a circular list of
 630 * the page's disk buffers. PG_private must be set to tell the VM to call
 631 * into the filesystem to release these pages.
 632 *
 633 * A page may belong to an inode's memory mapping. In this case, page->mapping
 634 * is the pointer to the inode, and page->index is the file offset of the page,
 635 * in units of PAGE_SIZE.
 636 *
 637 * If pagecache pages are not associated with an inode, they are said to be
 638 * anonymous pages. These may become associated with the swapcache, and in that
 639 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 640 *
 641 * In either case (swapcache or inode backed), the pagecache itself holds one
 642 * reference to the page. Setting PG_private should also increment the
 643 * refcount. The each user mapping also has a reference to the page.
 644 *
 645 * The pagecache pages are stored in a per-mapping radix tree, which is
 646 * rooted at mapping->page_tree, and indexed by offset.
 647 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 648 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 649 *
 650 * All pagecache pages may be subject to I/O:
 651 * - inode pages may need to be read from disk,
 652 * - inode pages which have been modified and are MAP_SHARED may need
 653 *   to be written back to the inode on disk,
 654 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 655 *   modified may need to be swapped out to swap space and (later) to be read
 656 *   back into memory.
 657 */
 658
 659/*
 660 * The zone field is never updated after free_area_init_core()
 661 * sets it, so none of the operations on it need to be atomic.
 662 */
 663
 664/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
 665#define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 666#define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
 667#define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
 668#define LAST_CPUPID_PGOFF	(ZONES_PGOFF - LAST_CPUPID_WIDTH)
 669
 670/*
 671 * Define the bit shifts to access each section.  For non-existent
 672 * sections we define the shift as 0; that plus a 0 mask ensures
 673 * the compiler will optimise away reference to them.
 674 */
 675#define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 676#define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
 677#define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
 678#define LAST_CPUPID_PGSHIFT	(LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
 679
 680/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
 681#ifdef NODE_NOT_IN_PAGE_FLAGS
 682#define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
 683#define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
 684						SECTIONS_PGOFF : ZONES_PGOFF)
 685#else
 686#define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
 687#define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
 688						NODES_PGOFF : ZONES_PGOFF)
 689#endif
 690
 691#define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 692
 693#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 694#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 695#endif
 696
 697#define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
 698#define NODES_MASK		((1UL << NODES_WIDTH) - 1)
 699#define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
 700#define LAST_CPUPID_MASK	((1UL << LAST_CPUPID_SHIFT) - 1)
 701#define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
 702
 703static inline enum zone_type page_zonenum(const struct page *page)
 704{
 705	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 706}
 707
 708#ifdef CONFIG_ZONE_DEVICE
 709void get_zone_device_page(struct page *page);
 710void put_zone_device_page(struct page *page);
 711static inline bool is_zone_device_page(const struct page *page)
 712{
 713	return page_zonenum(page) == ZONE_DEVICE;
 714}
 715#else
 716static inline void get_zone_device_page(struct page *page)
 717{
 718}
 719static inline void put_zone_device_page(struct page *page)
 720{
 721}
 722static inline bool is_zone_device_page(const struct page *page)
 723{
 724	return false;
 725}
 726#endif
 727
 728static inline void get_page(struct page *page)
 729{
 730	page = compound_head(page);
 731	/*
 732	 * Getting a normal page or the head of a compound page
 733	 * requires to already have an elevated page->_count.
 734	 */
 735	VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
 736	page_ref_inc(page);
 737
 738	if (unlikely(is_zone_device_page(page)))
 739		get_zone_device_page(page);
 740}
 741
 742static inline void put_page(struct page *page)
 743{
 744	page = compound_head(page);
 745
 746	if (put_page_testzero(page))
 747		__put_page(page);
 748
 749	if (unlikely(is_zone_device_page(page)))
 750		put_zone_device_page(page);
 751}
 752
 753#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 754#define SECTION_IN_PAGE_FLAGS
 755#endif
 756
 757/*
 758 * The identification function is mainly used by the buddy allocator for
 759 * determining if two pages could be buddies. We are not really identifying
 760 * the zone since we could be using the section number id if we do not have
 761 * node id available in page flags.
 762 * We only guarantee that it will return the same value for two combinable
 763 * pages in a zone.
 764 */
 765static inline int page_zone_id(struct page *page)
 766{
 767	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 768}
 769
 770static inline int zone_to_nid(struct zone *zone)
 771{
 772#ifdef CONFIG_NUMA
 773	return zone->node;
 774#else
 775	return 0;
 776#endif
 777}
 778
 779#ifdef NODE_NOT_IN_PAGE_FLAGS
 780extern int page_to_nid(const struct page *page);
 781#else
 782static inline int page_to_nid(const struct page *page)
 783{
 784	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 785}
 786#endif
 787
 788#ifdef CONFIG_NUMA_BALANCING
 789static inline int cpu_pid_to_cpupid(int cpu, int pid)
 790{
 791	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
 792}
 793
 794static inline int cpupid_to_pid(int cpupid)
 795{
 796	return cpupid & LAST__PID_MASK;
 797}
 798
 799static inline int cpupid_to_cpu(int cpupid)
 800{
 801	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
 802}
 803
 804static inline int cpupid_to_nid(int cpupid)
 805{
 806	return cpu_to_node(cpupid_to_cpu(cpupid));
 807}
 808
 809static inline bool cpupid_pid_unset(int cpupid)
 810{
 811	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
 812}
 813
 814static inline bool cpupid_cpu_unset(int cpupid)
 815{
 816	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
 817}
 818
 819static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
 820{
 821	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
 822}
 823
 824#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
 825#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
 826static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
 827{
 828	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
 829}
 830
 831static inline int page_cpupid_last(struct page *page)
 832{
 833	return page->_last_cpupid;
 834}
 835static inline void page_cpupid_reset_last(struct page *page)
 836{
 837	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
 838}
 839#else
 840static inline int page_cpupid_last(struct page *page)
 841{
 842	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
 843}
 844
 845extern int page_cpupid_xchg_last(struct page *page, int cpupid);
 846
 847static inline void page_cpupid_reset_last(struct page *page)
 848{
 849	int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
 850
 851	page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
 852	page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
 853}
 854#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
 855#else /* !CONFIG_NUMA_BALANCING */
 856static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
 857{
 858	return page_to_nid(page); /* XXX */
 859}
 860
 861static inline int page_cpupid_last(struct page *page)
 862{
 863	return page_to_nid(page); /* XXX */
 864}
 865
 866static inline int cpupid_to_nid(int cpupid)
 867{
 868	return -1;
 869}
 870
 871static inline int cpupid_to_pid(int cpupid)
 872{
 873	return -1;
 874}
 875
 876static inline int cpupid_to_cpu(int cpupid)
 877{
 878	return -1;
 879}
 880
 881static inline int cpu_pid_to_cpupid(int nid, int pid)
 882{
 883	return -1;
 884}
 885
 886static inline bool cpupid_pid_unset(int cpupid)
 887{
 888	return 1;
 889}
 890
 891static inline void page_cpupid_reset_last(struct page *page)
 892{
 893}
 894
 895static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
 896{
 897	return false;
 898}
 899#endif /* CONFIG_NUMA_BALANCING */
 900
 901static inline struct zone *page_zone(const struct page *page)
 902{
 903	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 904}
 905
 
 
 
 
 
 906#ifdef SECTION_IN_PAGE_FLAGS
 907static inline void set_page_section(struct page *page, unsigned long section)
 908{
 909	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 910	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 911}
 912
 913static inline unsigned long page_to_section(const struct page *page)
 914{
 915	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 916}
 917#endif
 918
 919static inline void set_page_zone(struct page *page, enum zone_type zone)
 920{
 921	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 922	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 923}
 924
 925static inline void set_page_node(struct page *page, unsigned long node)
 926{
 927	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 928	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 929}
 930
 931static inline void set_page_links(struct page *page, enum zone_type zone,
 932	unsigned long node, unsigned long pfn)
 933{
 934	set_page_zone(page, zone);
 935	set_page_node(page, node);
 936#ifdef SECTION_IN_PAGE_FLAGS
 937	set_page_section(page, pfn_to_section_nr(pfn));
 938#endif
 939}
 940
 941#ifdef CONFIG_MEMCG
 942static inline struct mem_cgroup *page_memcg(struct page *page)
 943{
 944	return page->mem_cgroup;
 945}
 
 
 
 
 
 946#else
 947static inline struct mem_cgroup *page_memcg(struct page *page)
 948{
 949	return NULL;
 950}
 
 
 
 
 
 951#endif
 952
 953/*
 954 * Some inline functions in vmstat.h depend on page_zone()
 955 */
 956#include <linux/vmstat.h>
 957
 958static __always_inline void *lowmem_page_address(const struct page *page)
 959{
 960	return __va(PFN_PHYS(page_to_pfn(page)));
 961}
 962
 963#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 964#define HASHED_PAGE_VIRTUAL
 965#endif
 966
 967#if defined(WANT_PAGE_VIRTUAL)
 968static inline void *page_address(const struct page *page)
 969{
 970	return page->virtual;
 971}
 972static inline void set_page_address(struct page *page, void *address)
 973{
 974	page->virtual = address;
 975}
 976#define page_address_init()  do { } while(0)
 977#endif
 978
 979#if defined(HASHED_PAGE_VIRTUAL)
 980void *page_address(const struct page *page);
 981void set_page_address(struct page *page, void *virtual);
 982void page_address_init(void);
 983#endif
 984
 985#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 986#define page_address(page) lowmem_page_address(page)
 987#define set_page_address(page, address)  do { } while(0)
 988#define page_address_init()  do { } while(0)
 989#endif
 990
 991extern void *page_rmapping(struct page *page);
 992extern struct anon_vma *page_anon_vma(struct page *page);
 993extern struct address_space *page_mapping(struct page *page);
 994
 995extern struct address_space *__page_file_mapping(struct page *);
 996
 997static inline
 998struct address_space *page_file_mapping(struct page *page)
 999{
1000	if (unlikely(PageSwapCache(page)))
1001		return __page_file_mapping(page);
1002
1003	return page->mapping;
1004}
1005
 
 
1006/*
1007 * Return the pagecache index of the passed page.  Regular pagecache pages
1008 * use ->index whereas swapcache pages use ->private
1009 */
1010static inline pgoff_t page_index(struct page *page)
1011{
1012	if (unlikely(PageSwapCache(page)))
1013		return page_private(page);
1014	return page->index;
1015}
1016
1017extern pgoff_t __page_file_index(struct page *page);
1018
1019/*
1020 * Return the file index of the page. Regular pagecache pages use ->index
1021 * whereas swapcache pages use swp_offset(->private)
1022 */
1023static inline pgoff_t page_file_index(struct page *page)
1024{
1025	if (unlikely(PageSwapCache(page)))
1026		return __page_file_index(page);
1027
1028	return page->index;
1029}
1030
1031/*
1032 * Return true if this page is mapped into pagetables.
1033 * For compound page it returns true if any subpage of compound page is mapped.
1034 */
1035static inline bool page_mapped(struct page *page)
1036{
1037	int i;
1038	if (likely(!PageCompound(page)))
1039		return atomic_read(&page->_mapcount) >= 0;
1040	page = compound_head(page);
1041	if (atomic_read(compound_mapcount_ptr(page)) >= 0)
1042		return true;
1043	if (PageHuge(page))
1044		return false;
1045	for (i = 0; i < hpage_nr_pages(page); i++) {
1046		if (atomic_read(&page[i]._mapcount) >= 0)
1047			return true;
1048	}
1049	return false;
1050}
1051
1052/*
1053 * Return true only if the page has been allocated with
1054 * ALLOC_NO_WATERMARKS and the low watermark was not
1055 * met implying that the system is under some pressure.
1056 */
1057static inline bool page_is_pfmemalloc(struct page *page)
1058{
1059	/*
1060	 * Page index cannot be this large so this must be
1061	 * a pfmemalloc page.
1062	 */
1063	return page->index == -1UL;
1064}
1065
1066/*
1067 * Only to be called by the page allocator on a freshly allocated
1068 * page.
1069 */
1070static inline void set_page_pfmemalloc(struct page *page)
1071{
1072	page->index = -1UL;
1073}
1074
1075static inline void clear_page_pfmemalloc(struct page *page)
1076{
1077	page->index = 0;
1078}
1079
1080/*
1081 * Different kinds of faults, as returned by handle_mm_fault().
1082 * Used to decide whether a process gets delivered SIGBUS or
1083 * just gets major/minor fault counters bumped up.
1084 */
1085
1086#define VM_FAULT_OOM	0x0001
1087#define VM_FAULT_SIGBUS	0x0002
1088#define VM_FAULT_MAJOR	0x0004
1089#define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
1090#define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
1091#define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1092#define VM_FAULT_SIGSEGV 0x0040
1093
1094#define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
1095#define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
1096#define VM_FAULT_RETRY	0x0400	/* ->fault blocked, must retry */
1097#define VM_FAULT_FALLBACK 0x0800	/* huge page fault failed, fall back to small */
 
1098
1099#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1100
1101#define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1102			 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1103			 VM_FAULT_FALLBACK)
1104
1105/* Encode hstate index for a hwpoisoned large page */
1106#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1107#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1108
1109/*
1110 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1111 */
1112extern void pagefault_out_of_memory(void);
1113
1114#define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
1115
1116/*
1117 * Flags passed to show_mem() and show_free_areas() to suppress output in
1118 * various contexts.
1119 */
1120#define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
1121
1122extern void show_free_areas(unsigned int flags);
1123extern bool skip_free_areas_node(unsigned int flags, int nid);
1124
1125int shmem_zero_setup(struct vm_area_struct *);
1126#ifdef CONFIG_SHMEM
1127bool shmem_mapping(struct address_space *mapping);
1128#else
1129static inline bool shmem_mapping(struct address_space *mapping)
1130{
1131	return false;
1132}
1133#endif
1134
1135extern bool can_do_mlock(void);
1136extern int user_shm_lock(size_t, struct user_struct *);
1137extern void user_shm_unlock(size_t, struct user_struct *);
1138
1139/*
1140 * Parameter block passed down to zap_pte_range in exceptional cases.
1141 */
1142struct zap_details {
1143	struct address_space *check_mapping;	/* Check page->mapping if set */
1144	pgoff_t	first_index;			/* Lowest page->index to unmap */
1145	pgoff_t last_index;			/* Highest page->index to unmap */
1146	bool ignore_dirty;			/* Ignore dirty pages */
1147	bool check_swap_entries;		/* Check also swap entries */
1148};
1149
1150struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1151		pte_t pte);
1152struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1153				pmd_t pmd);
1154
1155int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1156		unsigned long size);
1157void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1158		unsigned long size, struct zap_details *);
1159void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1160		unsigned long start, unsigned long end);
1161
1162/**
1163 * mm_walk - callbacks for walk_page_range
1164 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1165 *	       this handler is required to be able to handle
1166 *	       pmd_trans_huge() pmds.  They may simply choose to
1167 *	       split_huge_page() instead of handling it explicitly.
1168 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1169 * @pte_hole: if set, called for each hole at all levels
1170 * @hugetlb_entry: if set, called for each hugetlb entry
1171 * @test_walk: caller specific callback function to determine whether
1172 *             we walk over the current vma or not. A positive returned
1173 *             value means "do page table walk over the current vma,"
1174 *             and a negative one means "abort current page table walk
1175 *             right now." 0 means "skip the current vma."
1176 * @mm:        mm_struct representing the target process of page table walk
1177 * @vma:       vma currently walked (NULL if walking outside vmas)
1178 * @private:   private data for callbacks' usage
1179 *
1180 * (see the comment on walk_page_range() for more details)
1181 */
1182struct mm_walk {
1183	int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1184			 unsigned long next, struct mm_walk *walk);
1185	int (*pte_entry)(pte_t *pte, unsigned long addr,
1186			 unsigned long next, struct mm_walk *walk);
1187	int (*pte_hole)(unsigned long addr, unsigned long next,
1188			struct mm_walk *walk);
1189	int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1190			     unsigned long addr, unsigned long next,
1191			     struct mm_walk *walk);
1192	int (*test_walk)(unsigned long addr, unsigned long next,
1193			struct mm_walk *walk);
1194	struct mm_struct *mm;
1195	struct vm_area_struct *vma;
1196	void *private;
1197};
1198
1199int walk_page_range(unsigned long addr, unsigned long end,
1200		struct mm_walk *walk);
1201int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1202void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1203		unsigned long end, unsigned long floor, unsigned long ceiling);
1204int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1205			struct vm_area_struct *vma);
1206void unmap_mapping_range(struct address_space *mapping,
1207		loff_t const holebegin, loff_t const holelen, int even_cows);
 
 
1208int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1209	unsigned long *pfn);
1210int follow_phys(struct vm_area_struct *vma, unsigned long address,
1211		unsigned int flags, unsigned long *prot, resource_size_t *phys);
1212int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1213			void *buf, int len, int write);
1214
1215static inline void unmap_shared_mapping_range(struct address_space *mapping,
1216		loff_t const holebegin, loff_t const holelen)
1217{
1218	unmap_mapping_range(mapping, holebegin, holelen, 0);
1219}
1220
1221extern void truncate_pagecache(struct inode *inode, loff_t new);
1222extern void truncate_setsize(struct inode *inode, loff_t newsize);
1223void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1224void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1225int truncate_inode_page(struct address_space *mapping, struct page *page);
1226int generic_error_remove_page(struct address_space *mapping, struct page *page);
1227int invalidate_inode_page(struct page *page);
1228
1229#ifdef CONFIG_MMU
1230extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1231			unsigned long address, unsigned int flags);
1232extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1233			    unsigned long address, unsigned int fault_flags,
1234			    bool *unlocked);
1235#else
1236static inline int handle_mm_fault(struct mm_struct *mm,
1237			struct vm_area_struct *vma, unsigned long address,
1238			unsigned int flags)
1239{
1240	/* should never happen if there's no MMU */
1241	BUG();
1242	return VM_FAULT_SIGBUS;
1243}
1244static inline int fixup_user_fault(struct task_struct *tsk,
1245		struct mm_struct *mm, unsigned long address,
1246		unsigned int fault_flags, bool *unlocked)
1247{
1248	/* should never happen if there's no MMU */
1249	BUG();
1250	return -EFAULT;
1251}
1252#endif
1253
1254extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
 
1255extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1256		void *buf, int len, int write);
 
 
1257
1258long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1259		      unsigned long start, unsigned long nr_pages,
1260		      unsigned int foll_flags, struct page **pages,
1261		      struct vm_area_struct **vmas, int *nonblocking);
1262long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1263			    unsigned long start, unsigned long nr_pages,
1264			    int write, int force, struct page **pages,
1265			    struct vm_area_struct **vmas);
1266long get_user_pages(unsigned long start, unsigned long nr_pages,
1267			    int write, int force, struct page **pages,
1268			    struct vm_area_struct **vmas);
1269long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1270		    int write, int force, struct page **pages, int *locked);
1271long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1272			       unsigned long start, unsigned long nr_pages,
1273			       int write, int force, struct page **pages,
1274			       unsigned int gup_flags);
1275long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1276		    int write, int force, struct page **pages);
1277int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1278			struct page **pages);
1279
1280/* Container for pinned pfns / pages */
1281struct frame_vector {
1282	unsigned int nr_allocated;	/* Number of frames we have space for */
1283	unsigned int nr_frames;	/* Number of frames stored in ptrs array */
1284	bool got_ref;		/* Did we pin pages by getting page ref? */
1285	bool is_pfns;		/* Does array contain pages or pfns? */
1286	void *ptrs[0];		/* Array of pinned pfns / pages. Use
1287				 * pfns_vector_pages() or pfns_vector_pfns()
1288				 * for access */
1289};
1290
1291struct frame_vector *frame_vector_create(unsigned int nr_frames);
1292void frame_vector_destroy(struct frame_vector *vec);
1293int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1294		     bool write, bool force, struct frame_vector *vec);
1295void put_vaddr_frames(struct frame_vector *vec);
1296int frame_vector_to_pages(struct frame_vector *vec);
1297void frame_vector_to_pfns(struct frame_vector *vec);
1298
1299static inline unsigned int frame_vector_count(struct frame_vector *vec)
1300{
1301	return vec->nr_frames;
1302}
1303
1304static inline struct page **frame_vector_pages(struct frame_vector *vec)
1305{
1306	if (vec->is_pfns) {
1307		int err = frame_vector_to_pages(vec);
1308
1309		if (err)
1310			return ERR_PTR(err);
1311	}
1312	return (struct page **)(vec->ptrs);
1313}
1314
1315static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1316{
1317	if (!vec->is_pfns)
1318		frame_vector_to_pfns(vec);
1319	return (unsigned long *)(vec->ptrs);
1320}
1321
1322struct kvec;
1323int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1324			struct page **pages);
1325int get_kernel_page(unsigned long start, int write, struct page **pages);
1326struct page *get_dump_page(unsigned long addr);
1327
1328extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1329extern void do_invalidatepage(struct page *page, unsigned int offset,
1330			      unsigned int length);
1331
1332int __set_page_dirty_nobuffers(struct page *page);
1333int __set_page_dirty_no_writeback(struct page *page);
1334int redirty_page_for_writepage(struct writeback_control *wbc,
1335				struct page *page);
1336void account_page_dirtied(struct page *page, struct address_space *mapping);
1337void account_page_cleaned(struct page *page, struct address_space *mapping,
1338			  struct bdi_writeback *wb);
1339int set_page_dirty(struct page *page);
1340int set_page_dirty_lock(struct page *page);
1341void cancel_dirty_page(struct page *page);
1342int clear_page_dirty_for_io(struct page *page);
1343
1344int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1345
1346/* Is the vma a continuation of the stack vma above it? */
1347static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1348{
1349	return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1350}
1351
1352static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1353{
1354	return !vma->vm_ops;
1355}
1356
1357static inline int stack_guard_page_start(struct vm_area_struct *vma,
1358					     unsigned long addr)
1359{
1360	return (vma->vm_flags & VM_GROWSDOWN) &&
1361		(vma->vm_start == addr) &&
1362		!vma_growsdown(vma->vm_prev, addr);
1363}
1364
1365/* Is the vma a continuation of the stack vma below it? */
1366static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1367{
1368	return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1369}
1370
1371static inline int stack_guard_page_end(struct vm_area_struct *vma,
1372					   unsigned long addr)
1373{
1374	return (vma->vm_flags & VM_GROWSUP) &&
1375		(vma->vm_end == addr) &&
1376		!vma_growsup(vma->vm_next, addr);
1377}
1378
1379int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1380
1381extern unsigned long move_page_tables(struct vm_area_struct *vma,
1382		unsigned long old_addr, struct vm_area_struct *new_vma,
1383		unsigned long new_addr, unsigned long len,
1384		bool need_rmap_locks);
1385extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1386			      unsigned long end, pgprot_t newprot,
1387			      int dirty_accountable, int prot_numa);
1388extern int mprotect_fixup(struct vm_area_struct *vma,
1389			  struct vm_area_struct **pprev, unsigned long start,
1390			  unsigned long end, unsigned long newflags);
1391
1392/*
1393 * doesn't attempt to fault and will return short.
1394 */
1395int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1396			  struct page **pages);
1397/*
1398 * per-process(per-mm_struct) statistics.
1399 */
1400static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1401{
1402	long val = atomic_long_read(&mm->rss_stat.count[member]);
1403
1404#ifdef SPLIT_RSS_COUNTING
1405	/*
1406	 * counter is updated in asynchronous manner and may go to minus.
1407	 * But it's never be expected number for users.
1408	 */
1409	if (val < 0)
1410		val = 0;
1411#endif
1412	return (unsigned long)val;
1413}
1414
1415static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1416{
1417	atomic_long_add(value, &mm->rss_stat.count[member]);
1418}
1419
1420static inline void inc_mm_counter(struct mm_struct *mm, int member)
1421{
1422	atomic_long_inc(&mm->rss_stat.count[member]);
1423}
1424
1425static inline void dec_mm_counter(struct mm_struct *mm, int member)
1426{
1427	atomic_long_dec(&mm->rss_stat.count[member]);
1428}
1429
1430/* Optimized variant when page is already known not to be PageAnon */
1431static inline int mm_counter_file(struct page *page)
1432{
1433	if (PageSwapBacked(page))
1434		return MM_SHMEMPAGES;
1435	return MM_FILEPAGES;
1436}
1437
1438static inline int mm_counter(struct page *page)
1439{
1440	if (PageAnon(page))
1441		return MM_ANONPAGES;
1442	return mm_counter_file(page);
1443}
1444
1445static inline unsigned long get_mm_rss(struct mm_struct *mm)
1446{
1447	return get_mm_counter(mm, MM_FILEPAGES) +
1448		get_mm_counter(mm, MM_ANONPAGES) +
1449		get_mm_counter(mm, MM_SHMEMPAGES);
1450}
1451
1452static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1453{
1454	return max(mm->hiwater_rss, get_mm_rss(mm));
1455}
1456
1457static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1458{
1459	return max(mm->hiwater_vm, mm->total_vm);
1460}
1461
1462static inline void update_hiwater_rss(struct mm_struct *mm)
1463{
1464	unsigned long _rss = get_mm_rss(mm);
1465
1466	if ((mm)->hiwater_rss < _rss)
1467		(mm)->hiwater_rss = _rss;
1468}
1469
1470static inline void update_hiwater_vm(struct mm_struct *mm)
1471{
1472	if (mm->hiwater_vm < mm->total_vm)
1473		mm->hiwater_vm = mm->total_vm;
1474}
1475
1476static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1477{
1478	mm->hiwater_rss = get_mm_rss(mm);
1479}
1480
1481static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1482					 struct mm_struct *mm)
1483{
1484	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1485
1486	if (*maxrss < hiwater_rss)
1487		*maxrss = hiwater_rss;
1488}
1489
1490#if defined(SPLIT_RSS_COUNTING)
1491void sync_mm_rss(struct mm_struct *mm);
1492#else
1493static inline void sync_mm_rss(struct mm_struct *mm)
1494{
1495}
1496#endif
1497
1498#ifndef __HAVE_ARCH_PTE_DEVMAP
1499static inline int pte_devmap(pte_t pte)
1500{
1501	return 0;
1502}
1503#endif
1504
1505int vma_wants_writenotify(struct vm_area_struct *vma);
1506
1507extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1508			       spinlock_t **ptl);
1509static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1510				    spinlock_t **ptl)
1511{
1512	pte_t *ptep;
1513	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1514	return ptep;
1515}
1516
1517#ifdef __PAGETABLE_PUD_FOLDED
1518static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1519						unsigned long address)
1520{
1521	return 0;
1522}
1523#else
1524int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1525#endif
1526
1527#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1528static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1529						unsigned long address)
1530{
1531	return 0;
1532}
1533
1534static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1535
1536static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1537{
1538	return 0;
1539}
1540
1541static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1542static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1543
1544#else
1545int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1546
1547static inline void mm_nr_pmds_init(struct mm_struct *mm)
1548{
1549	atomic_long_set(&mm->nr_pmds, 0);
1550}
1551
1552static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1553{
1554	return atomic_long_read(&mm->nr_pmds);
1555}
1556
1557static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1558{
1559	atomic_long_inc(&mm->nr_pmds);
1560}
1561
1562static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1563{
1564	atomic_long_dec(&mm->nr_pmds);
1565}
1566#endif
1567
1568int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1569int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1570
1571/*
1572 * The following ifdef needed to get the 4level-fixup.h header to work.
1573 * Remove it when 4level-fixup.h has been removed.
1574 */
1575#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1576static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1577{
1578	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1579		NULL: pud_offset(pgd, address);
1580}
1581
1582static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1583{
1584	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1585		NULL: pmd_offset(pud, address);
1586}
1587#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1588
1589#if USE_SPLIT_PTE_PTLOCKS
1590#if ALLOC_SPLIT_PTLOCKS
1591void __init ptlock_cache_init(void);
1592extern bool ptlock_alloc(struct page *page);
1593extern void ptlock_free(struct page *page);
1594
1595static inline spinlock_t *ptlock_ptr(struct page *page)
1596{
1597	return page->ptl;
1598}
1599#else /* ALLOC_SPLIT_PTLOCKS */
1600static inline void ptlock_cache_init(void)
1601{
1602}
1603
1604static inline bool ptlock_alloc(struct page *page)
1605{
1606	return true;
1607}
1608
1609static inline void ptlock_free(struct page *page)
1610{
1611}
1612
1613static inline spinlock_t *ptlock_ptr(struct page *page)
1614{
1615	return &page->ptl;
1616}
1617#endif /* ALLOC_SPLIT_PTLOCKS */
1618
1619static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1620{
1621	return ptlock_ptr(pmd_page(*pmd));
1622}
1623
1624static inline bool ptlock_init(struct page *page)
1625{
1626	/*
1627	 * prep_new_page() initialize page->private (and therefore page->ptl)
1628	 * with 0. Make sure nobody took it in use in between.
1629	 *
1630	 * It can happen if arch try to use slab for page table allocation:
1631	 * slab code uses page->slab_cache, which share storage with page->ptl.
1632	 */
1633	VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1634	if (!ptlock_alloc(page))
1635		return false;
1636	spin_lock_init(ptlock_ptr(page));
1637	return true;
1638}
1639
1640/* Reset page->mapping so free_pages_check won't complain. */
1641static inline void pte_lock_deinit(struct page *page)
1642{
1643	page->mapping = NULL;
1644	ptlock_free(page);
1645}
1646
1647#else	/* !USE_SPLIT_PTE_PTLOCKS */
1648/*
1649 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1650 */
1651static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1652{
1653	return &mm->page_table_lock;
1654}
1655static inline void ptlock_cache_init(void) {}
1656static inline bool ptlock_init(struct page *page) { return true; }
1657static inline void pte_lock_deinit(struct page *page) {}
1658#endif /* USE_SPLIT_PTE_PTLOCKS */
1659
1660static inline void pgtable_init(void)
1661{
1662	ptlock_cache_init();
1663	pgtable_cache_init();
1664}
1665
1666static inline bool pgtable_page_ctor(struct page *page)
1667{
1668	if (!ptlock_init(page))
1669		return false;
1670	inc_zone_page_state(page, NR_PAGETABLE);
1671	return true;
1672}
1673
1674static inline void pgtable_page_dtor(struct page *page)
1675{
1676	pte_lock_deinit(page);
1677	dec_zone_page_state(page, NR_PAGETABLE);
1678}
1679
1680#define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1681({							\
1682	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1683	pte_t *__pte = pte_offset_map(pmd, address);	\
1684	*(ptlp) = __ptl;				\
1685	spin_lock(__ptl);				\
1686	__pte;						\
1687})
1688
1689#define pte_unmap_unlock(pte, ptl)	do {		\
1690	spin_unlock(ptl);				\
1691	pte_unmap(pte);					\
1692} while (0)
1693
1694#define pte_alloc(mm, pmd, address)			\
1695	(unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1696
1697#define pte_alloc_map(mm, pmd, address)			\
1698	(pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1699
1700#define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1701	(pte_alloc(mm, pmd, address) ?			\
1702		 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1703
1704#define pte_alloc_kernel(pmd, address)			\
1705	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1706		NULL: pte_offset_kernel(pmd, address))
1707
1708#if USE_SPLIT_PMD_PTLOCKS
1709
1710static struct page *pmd_to_page(pmd_t *pmd)
1711{
1712	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1713	return virt_to_page((void *)((unsigned long) pmd & mask));
1714}
1715
1716static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1717{
1718	return ptlock_ptr(pmd_to_page(pmd));
1719}
1720
1721static inline bool pgtable_pmd_page_ctor(struct page *page)
1722{
1723#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1724	page->pmd_huge_pte = NULL;
1725#endif
1726	return ptlock_init(page);
1727}
1728
1729static inline void pgtable_pmd_page_dtor(struct page *page)
1730{
1731#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1732	VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1733#endif
1734	ptlock_free(page);
1735}
1736
1737#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1738
1739#else
1740
1741static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1742{
1743	return &mm->page_table_lock;
1744}
1745
1746static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1747static inline void pgtable_pmd_page_dtor(struct page *page) {}
1748
1749#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1750
1751#endif
1752
1753static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1754{
1755	spinlock_t *ptl = pmd_lockptr(mm, pmd);
1756	spin_lock(ptl);
1757	return ptl;
1758}
1759
 
 
1760extern void free_area_init(unsigned long * zones_size);
1761extern void free_area_init_node(int nid, unsigned long * zones_size,
1762		unsigned long zone_start_pfn, unsigned long *zholes_size);
1763extern void free_initmem(void);
1764
1765/*
1766 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1767 * into the buddy system. The freed pages will be poisoned with pattern
1768 * "poison" if it's within range [0, UCHAR_MAX].
1769 * Return pages freed into the buddy system.
1770 */
1771extern unsigned long free_reserved_area(void *start, void *end,
1772					int poison, char *s);
1773
1774#ifdef	CONFIG_HIGHMEM
1775/*
1776 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1777 * and totalram_pages.
1778 */
1779extern void free_highmem_page(struct page *page);
1780#endif
1781
1782extern void adjust_managed_page_count(struct page *page, long count);
1783extern void mem_init_print_info(const char *str);
1784
1785extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1786
1787/* Free the reserved page into the buddy system, so it gets managed. */
1788static inline void __free_reserved_page(struct page *page)
1789{
1790	ClearPageReserved(page);
1791	init_page_count(page);
1792	__free_page(page);
1793}
1794
1795static inline void free_reserved_page(struct page *page)
1796{
1797	__free_reserved_page(page);
1798	adjust_managed_page_count(page, 1);
1799}
1800
1801static inline void mark_page_reserved(struct page *page)
1802{
1803	SetPageReserved(page);
1804	adjust_managed_page_count(page, -1);
1805}
1806
1807/*
1808 * Default method to free all the __init memory into the buddy system.
1809 * The freed pages will be poisoned with pattern "poison" if it's within
1810 * range [0, UCHAR_MAX].
1811 * Return pages freed into the buddy system.
1812 */
1813static inline unsigned long free_initmem_default(int poison)
1814{
1815	extern char __init_begin[], __init_end[];
1816
1817	return free_reserved_area(&__init_begin, &__init_end,
1818				  poison, "unused kernel");
1819}
1820
1821static inline unsigned long get_num_physpages(void)
1822{
1823	int nid;
1824	unsigned long phys_pages = 0;
1825
1826	for_each_online_node(nid)
1827		phys_pages += node_present_pages(nid);
1828
1829	return phys_pages;
1830}
1831
1832#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1833/*
1834 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1835 * zones, allocate the backing mem_map and account for memory holes in a more
1836 * architecture independent manner. This is a substitute for creating the
1837 * zone_sizes[] and zholes_size[] arrays and passing them to
1838 * free_area_init_node()
1839 *
1840 * An architecture is expected to register range of page frames backed by
1841 * physical memory with memblock_add[_node]() before calling
1842 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1843 * usage, an architecture is expected to do something like
1844 *
1845 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1846 * 							 max_highmem_pfn};
1847 * for_each_valid_physical_page_range()
1848 * 	memblock_add_node(base, size, nid)
1849 * free_area_init_nodes(max_zone_pfns);
1850 *
1851 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1852 * registered physical page range.  Similarly
1853 * sparse_memory_present_with_active_regions() calls memory_present() for
1854 * each range when SPARSEMEM is enabled.
1855 *
1856 * See mm/page_alloc.c for more information on each function exposed by
1857 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1858 */
1859extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1860unsigned long node_map_pfn_alignment(void);
1861unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1862						unsigned long end_pfn);
1863extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1864						unsigned long end_pfn);
1865extern void get_pfn_range_for_nid(unsigned int nid,
1866			unsigned long *start_pfn, unsigned long *end_pfn);
1867extern unsigned long find_min_pfn_with_active_regions(void);
1868extern void free_bootmem_with_active_regions(int nid,
1869						unsigned long max_low_pfn);
1870extern void sparse_memory_present_with_active_regions(int nid);
1871
1872#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1873
1874#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1875    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1876static inline int __early_pfn_to_nid(unsigned long pfn,
1877					struct mminit_pfnnid_cache *state)
1878{
1879	return 0;
1880}
1881#else
1882/* please see mm/page_alloc.c */
1883extern int __meminit early_pfn_to_nid(unsigned long pfn);
1884/* there is a per-arch backend function. */
1885extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1886					struct mminit_pfnnid_cache *state);
1887#endif
1888
1889extern void set_dma_reserve(unsigned long new_dma_reserve);
1890extern void memmap_init_zone(unsigned long, int, unsigned long,
1891				unsigned long, enum memmap_context);
1892extern void setup_per_zone_wmarks(void);
1893extern int __meminit init_per_zone_wmark_min(void);
1894extern void mem_init(void);
1895extern void __init mmap_init(void);
1896extern void show_mem(unsigned int flags);
1897extern long si_mem_available(void);
1898extern void si_meminfo(struct sysinfo * val);
1899extern void si_meminfo_node(struct sysinfo *val, int nid);
 
 
 
1900
1901extern __printf(3, 4)
1902void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1903		const char *fmt, ...);
1904
1905extern void setup_per_cpu_pageset(void);
1906
1907extern void zone_pcp_update(struct zone *zone);
1908extern void zone_pcp_reset(struct zone *zone);
1909
1910/* page_alloc.c */
1911extern int min_free_kbytes;
1912extern int watermark_scale_factor;
1913
1914/* nommu.c */
1915extern atomic_long_t mmap_pages_allocated;
1916extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1917
1918/* interval_tree.c */
1919void vma_interval_tree_insert(struct vm_area_struct *node,
1920			      struct rb_root *root);
1921void vma_interval_tree_insert_after(struct vm_area_struct *node,
1922				    struct vm_area_struct *prev,
1923				    struct rb_root *root);
1924void vma_interval_tree_remove(struct vm_area_struct *node,
1925			      struct rb_root *root);
1926struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1927				unsigned long start, unsigned long last);
1928struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1929				unsigned long start, unsigned long last);
1930
1931#define vma_interval_tree_foreach(vma, root, start, last)		\
1932	for (vma = vma_interval_tree_iter_first(root, start, last);	\
1933	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
1934
1935void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1936				   struct rb_root *root);
1937void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1938				   struct rb_root *root);
1939struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1940	struct rb_root *root, unsigned long start, unsigned long last);
1941struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1942	struct anon_vma_chain *node, unsigned long start, unsigned long last);
1943#ifdef CONFIG_DEBUG_VM_RB
1944void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1945#endif
1946
1947#define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
1948	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1949	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1950
1951/* mmap.c */
1952extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1953extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1954	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
 
 
 
 
 
 
1955extern struct vm_area_struct *vma_merge(struct mm_struct *,
1956	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1957	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1958	struct mempolicy *, struct vm_userfaultfd_ctx);
1959extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1960extern int split_vma(struct mm_struct *,
1961	struct vm_area_struct *, unsigned long addr, int new_below);
1962extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1963extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1964	struct rb_node **, struct rb_node *);
1965extern void unlink_file_vma(struct vm_area_struct *);
1966extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1967	unsigned long addr, unsigned long len, pgoff_t pgoff,
1968	bool *need_rmap_locks);
1969extern void exit_mmap(struct mm_struct *);
1970
1971static inline int check_data_rlimit(unsigned long rlim,
1972				    unsigned long new,
1973				    unsigned long start,
1974				    unsigned long end_data,
1975				    unsigned long start_data)
1976{
1977	if (rlim < RLIM_INFINITY) {
1978		if (((new - start) + (end_data - start_data)) > rlim)
1979			return -ENOSPC;
1980	}
1981
1982	return 0;
1983}
1984
1985extern int mm_take_all_locks(struct mm_struct *mm);
1986extern void mm_drop_all_locks(struct mm_struct *mm);
1987
1988extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1989extern struct file *get_mm_exe_file(struct mm_struct *mm);
 
1990
1991extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
1992extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
1993
 
 
1994extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1995				   unsigned long addr, unsigned long len,
1996				   unsigned long flags,
1997				   const struct vm_special_mapping *spec);
1998/* This is an obsolete alternative to _install_special_mapping. */
1999extern int install_special_mapping(struct mm_struct *mm,
2000				   unsigned long addr, unsigned long len,
2001				   unsigned long flags, struct page **pages);
2002
2003extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2004
2005extern unsigned long mmap_region(struct file *file, unsigned long addr,
2006	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2007extern unsigned long do_mmap(struct file *file, unsigned long addr,
2008	unsigned long len, unsigned long prot, unsigned long flags,
2009	vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2010extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2011
2012static inline unsigned long
2013do_mmap_pgoff(struct file *file, unsigned long addr,
2014	unsigned long len, unsigned long prot, unsigned long flags,
2015	unsigned long pgoff, unsigned long *populate)
2016{
2017	return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2018}
2019
2020#ifdef CONFIG_MMU
2021extern int __mm_populate(unsigned long addr, unsigned long len,
2022			 int ignore_errors);
2023static inline void mm_populate(unsigned long addr, unsigned long len)
2024{
2025	/* Ignore errors */
2026	(void) __mm_populate(addr, len, 1);
2027}
2028#else
2029static inline void mm_populate(unsigned long addr, unsigned long len) {}
2030#endif
2031
2032/* These take the mm semaphore themselves */
2033extern unsigned long vm_brk(unsigned long, unsigned long);
2034extern int vm_munmap(unsigned long, size_t);
2035extern unsigned long vm_mmap(struct file *, unsigned long,
2036        unsigned long, unsigned long,
2037        unsigned long, unsigned long);
2038
2039struct vm_unmapped_area_info {
2040#define VM_UNMAPPED_AREA_TOPDOWN 1
2041	unsigned long flags;
2042	unsigned long length;
2043	unsigned long low_limit;
2044	unsigned long high_limit;
2045	unsigned long align_mask;
2046	unsigned long align_offset;
2047};
2048
2049extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2050extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2051
2052/*
2053 * Search for an unmapped address range.
2054 *
2055 * We are looking for a range that:
2056 * - does not intersect with any VMA;
2057 * - is contained within the [low_limit, high_limit) interval;
2058 * - is at least the desired size.
2059 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2060 */
2061static inline unsigned long
2062vm_unmapped_area(struct vm_unmapped_area_info *info)
2063{
2064	if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2065		return unmapped_area_topdown(info);
2066	else
2067		return unmapped_area(info);
2068}
2069
2070/* truncate.c */
2071extern void truncate_inode_pages(struct address_space *, loff_t);
2072extern void truncate_inode_pages_range(struct address_space *,
2073				       loff_t lstart, loff_t lend);
2074extern void truncate_inode_pages_final(struct address_space *);
2075
2076/* generic vm_area_ops exported for stackable file systems */
2077extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2078extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
 
2079extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2080
2081/* mm/page-writeback.c */
2082int write_one_page(struct page *page, int wait);
2083void task_dirty_inc(struct task_struct *tsk);
2084
2085/* readahead.c */
2086#define VM_MAX_READAHEAD	128	/* kbytes */
2087#define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
2088
2089int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2090			pgoff_t offset, unsigned long nr_to_read);
2091
2092void page_cache_sync_readahead(struct address_space *mapping,
2093			       struct file_ra_state *ra,
2094			       struct file *filp,
2095			       pgoff_t offset,
2096			       unsigned long size);
2097
2098void page_cache_async_readahead(struct address_space *mapping,
2099				struct file_ra_state *ra,
2100				struct file *filp,
2101				struct page *pg,
2102				pgoff_t offset,
2103				unsigned long size);
2104
2105/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2106extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2107
2108/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2109extern int expand_downwards(struct vm_area_struct *vma,
2110		unsigned long address);
2111#if VM_GROWSUP
2112extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2113#else
2114  #define expand_upwards(vma, address) (0)
2115#endif
2116
2117/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2118extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2119extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2120					     struct vm_area_struct **pprev);
2121
2122/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2123   NULL if none.  Assume start_addr < end_addr. */
2124static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2125{
2126	struct vm_area_struct * vma = find_vma(mm,start_addr);
2127
2128	if (vma && end_addr <= vma->vm_start)
2129		vma = NULL;
2130	return vma;
2131}
2132
2133static inline unsigned long vma_pages(struct vm_area_struct *vma)
2134{
2135	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2136}
2137
2138/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2139static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2140				unsigned long vm_start, unsigned long vm_end)
2141{
2142	struct vm_area_struct *vma = find_vma(mm, vm_start);
2143
2144	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2145		vma = NULL;
2146
2147	return vma;
2148}
2149
2150#ifdef CONFIG_MMU
2151pgprot_t vm_get_page_prot(unsigned long vm_flags);
2152void vma_set_page_prot(struct vm_area_struct *vma);
2153#else
2154static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2155{
2156	return __pgprot(0);
2157}
2158static inline void vma_set_page_prot(struct vm_area_struct *vma)
2159{
2160	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2161}
2162#endif
2163
2164#ifdef CONFIG_NUMA_BALANCING
2165unsigned long change_prot_numa(struct vm_area_struct *vma,
2166			unsigned long start, unsigned long end);
2167#endif
2168
2169struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2170int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2171			unsigned long pfn, unsigned long size, pgprot_t);
2172int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2173int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2174			unsigned long pfn);
2175int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2176			unsigned long pfn, pgprot_t pgprot);
2177int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2178			pfn_t pfn);
2179int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2180
2181
2182struct page *follow_page_mask(struct vm_area_struct *vma,
2183			      unsigned long address, unsigned int foll_flags,
2184			      unsigned int *page_mask);
2185
2186static inline struct page *follow_page(struct vm_area_struct *vma,
2187		unsigned long address, unsigned int foll_flags)
2188{
2189	unsigned int unused_page_mask;
2190	return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2191}
2192
2193#define FOLL_WRITE	0x01	/* check pte is writable */
2194#define FOLL_TOUCH	0x02	/* mark page accessed */
2195#define FOLL_GET	0x04	/* do get_page on page */
2196#define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
2197#define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
2198#define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
2199				 * and return without waiting upon it */
2200#define FOLL_POPULATE	0x40	/* fault in page */
2201#define FOLL_SPLIT	0x80	/* don't return transhuge pages, split them */
2202#define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
2203#define FOLL_NUMA	0x200	/* force NUMA hinting page fault */
2204#define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
2205#define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
2206#define FOLL_MLOCK	0x1000	/* lock present pages */
2207#define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
 
2208
2209typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2210			void *data);
2211extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2212			       unsigned long size, pte_fn_t fn, void *data);
2213
2214
2215#ifdef CONFIG_PAGE_POISONING
2216extern bool page_poisoning_enabled(void);
2217extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2218extern bool page_is_poisoned(struct page *page);
2219#else
2220static inline bool page_poisoning_enabled(void) { return false; }
2221static inline void kernel_poison_pages(struct page *page, int numpages,
2222					int enable) { }
2223static inline bool page_is_poisoned(struct page *page) { return false; }
2224#endif
2225
2226#ifdef CONFIG_DEBUG_PAGEALLOC
2227extern bool _debug_pagealloc_enabled;
2228extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2229
2230static inline bool debug_pagealloc_enabled(void)
2231{
2232	return _debug_pagealloc_enabled;
2233}
2234
2235static inline void
2236kernel_map_pages(struct page *page, int numpages, int enable)
2237{
2238	if (!debug_pagealloc_enabled())
2239		return;
2240
2241	__kernel_map_pages(page, numpages, enable);
2242}
2243#ifdef CONFIG_HIBERNATION
2244extern bool kernel_page_present(struct page *page);
2245#endif	/* CONFIG_HIBERNATION */
2246#else	/* CONFIG_DEBUG_PAGEALLOC */
2247static inline void
2248kernel_map_pages(struct page *page, int numpages, int enable) {}
2249#ifdef CONFIG_HIBERNATION
2250static inline bool kernel_page_present(struct page *page) { return true; }
2251#endif	/* CONFIG_HIBERNATION */
2252static inline bool debug_pagealloc_enabled(void)
2253{
2254	return false;
2255}
2256#endif	/* CONFIG_DEBUG_PAGEALLOC */
2257
2258#ifdef __HAVE_ARCH_GATE_AREA
2259extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2260extern int in_gate_area_no_mm(unsigned long addr);
2261extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2262#else
2263static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2264{
2265	return NULL;
2266}
2267static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2268static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2269{
2270	return 0;
2271}
2272#endif	/* __HAVE_ARCH_GATE_AREA */
2273
 
 
2274#ifdef CONFIG_SYSCTL
2275extern int sysctl_drop_caches;
2276int drop_caches_sysctl_handler(struct ctl_table *, int,
2277					void __user *, size_t *, loff_t *);
2278#endif
2279
2280void drop_slab(void);
2281void drop_slab_node(int nid);
2282
2283#ifndef CONFIG_MMU
2284#define randomize_va_space 0
2285#else
2286extern int randomize_va_space;
2287#endif
2288
2289const char * arch_vma_name(struct vm_area_struct *vma);
2290void print_vma_addr(char *prefix, unsigned long rip);
2291
2292void sparse_mem_maps_populate_node(struct page **map_map,
2293				   unsigned long pnum_begin,
2294				   unsigned long pnum_end,
2295				   unsigned long map_count,
2296				   int nodeid);
2297
2298struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2299pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2300pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2301pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2302pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2303void *vmemmap_alloc_block(unsigned long size, int node);
2304struct vmem_altmap;
2305void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2306		struct vmem_altmap *altmap);
2307static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2308{
2309	return __vmemmap_alloc_block_buf(size, node, NULL);
2310}
2311
2312void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2313int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2314			       int node);
2315int vmemmap_populate(unsigned long start, unsigned long end, int node);
2316void vmemmap_populate_print_last(void);
2317#ifdef CONFIG_MEMORY_HOTPLUG
2318void vmemmap_free(unsigned long start, unsigned long end);
2319#endif
2320void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2321				  unsigned long size);
2322
2323enum mf_flags {
2324	MF_COUNT_INCREASED = 1 << 0,
2325	MF_ACTION_REQUIRED = 1 << 1,
2326	MF_MUST_KILL = 1 << 2,
2327	MF_SOFT_OFFLINE = 1 << 3,
2328};
2329extern int memory_failure(unsigned long pfn, int trapno, int flags);
2330extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2331extern int unpoison_memory(unsigned long pfn);
2332extern int get_hwpoison_page(struct page *page);
2333#define put_hwpoison_page(page)	put_page(page)
2334extern int sysctl_memory_failure_early_kill;
2335extern int sysctl_memory_failure_recovery;
2336extern void shake_page(struct page *p, int access);
2337extern atomic_long_t num_poisoned_pages;
2338extern int soft_offline_page(struct page *page, int flags);
2339
2340
2341/*
2342 * Error handlers for various types of pages.
2343 */
2344enum mf_result {
2345	MF_IGNORED,	/* Error: cannot be handled */
2346	MF_FAILED,	/* Error: handling failed */
2347	MF_DELAYED,	/* Will be handled later */
2348	MF_RECOVERED,	/* Successfully recovered */
2349};
2350
2351enum mf_action_page_type {
2352	MF_MSG_KERNEL,
2353	MF_MSG_KERNEL_HIGH_ORDER,
2354	MF_MSG_SLAB,
2355	MF_MSG_DIFFERENT_COMPOUND,
2356	MF_MSG_POISONED_HUGE,
2357	MF_MSG_HUGE,
2358	MF_MSG_FREE_HUGE,
2359	MF_MSG_UNMAP_FAILED,
2360	MF_MSG_DIRTY_SWAPCACHE,
2361	MF_MSG_CLEAN_SWAPCACHE,
2362	MF_MSG_DIRTY_MLOCKED_LRU,
2363	MF_MSG_CLEAN_MLOCKED_LRU,
2364	MF_MSG_DIRTY_UNEVICTABLE_LRU,
2365	MF_MSG_CLEAN_UNEVICTABLE_LRU,
2366	MF_MSG_DIRTY_LRU,
2367	MF_MSG_CLEAN_LRU,
2368	MF_MSG_TRUNCATED_LRU,
2369	MF_MSG_BUDDY,
2370	MF_MSG_BUDDY_2ND,
2371	MF_MSG_UNKNOWN,
2372};
2373
2374#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2375extern void clear_huge_page(struct page *page,
2376			    unsigned long addr,
2377			    unsigned int pages_per_huge_page);
2378extern void copy_user_huge_page(struct page *dst, struct page *src,
2379				unsigned long addr, struct vm_area_struct *vma,
2380				unsigned int pages_per_huge_page);
2381#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2382
2383extern struct page_ext_operations debug_guardpage_ops;
2384extern struct page_ext_operations page_poisoning_ops;
2385
2386#ifdef CONFIG_DEBUG_PAGEALLOC
2387extern unsigned int _debug_guardpage_minorder;
2388extern bool _debug_guardpage_enabled;
2389
2390static inline unsigned int debug_guardpage_minorder(void)
2391{
2392	return _debug_guardpage_minorder;
2393}
2394
2395static inline bool debug_guardpage_enabled(void)
2396{
2397	return _debug_guardpage_enabled;
2398}
2399
2400static inline bool page_is_guard(struct page *page)
2401{
2402	struct page_ext *page_ext;
2403
2404	if (!debug_guardpage_enabled())
2405		return false;
2406
2407	page_ext = lookup_page_ext(page);
 
 
 
2408	return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2409}
2410#else
2411static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2412static inline bool debug_guardpage_enabled(void) { return false; }
2413static inline bool page_is_guard(struct page *page) { return false; }
2414#endif /* CONFIG_DEBUG_PAGEALLOC */
2415
2416#if MAX_NUMNODES > 1
2417void __init setup_nr_node_ids(void);
2418#else
2419static inline void setup_nr_node_ids(void) {}
2420#endif
2421
2422#endif /* __KERNEL__ */
2423#endif /* _LINUX_MM_H */