blob: d8738a464b94a71822f191bdd45bcac29b01814c [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
Linus Torvalds1da177e2005-04-16 15:20:36 -07004#include <linux/errno.h>
5
6#ifdef __KERNEL__
7
Linus Torvalds1da177e2005-04-16 15:20:36 -07008#include <linux/gfp.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -05009#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/list.h>
11#include <linux/mmzone.h>
12#include <linux/rbtree.h>
13#include <linux/prio_tree.h>
Konstantin Khlebnikov83aeead2011-12-08 14:33:54 -080014#include <linux/atomic.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070015#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070016#include <linux/mm_types.h>
Yinghai Lu08677212010-02-10 01:20:20 -080017#include <linux/range.h>
Chris Metcalfc6f6b592010-05-24 14:32:53 -070018#include <linux/pfn.h>
Andrea Arcangelie9da73d2011-01-13 15:46:32 -080019#include <linux/bit_spinlock.h>
Dave Chinnerb0d40c92011-07-08 14:14:42 +100020#include <linux/shrinker.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
22struct mempolicy;
23struct anon_vma;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040024struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040025struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040026struct writeback_control;
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
28#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
29extern unsigned long max_mapnr;
30#endif
31
32extern unsigned long num_physpages;
Jan Beulich44813742009-09-21 17:03:05 -070033extern unsigned long totalram_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -070034extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070035extern int page_cluster;
36
37#ifdef CONFIG_SYSCTL
38extern int sysctl_legacy_va_layout;
39#else
40#define sysctl_legacy_va_layout 0
41#endif
42
43#include <asm/page.h>
44#include <asm/pgtable.h>
45#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
48
Andrea Righi27ac7922008-07-23 21:28:13 -070049/* to align the pointer to the (next) page boundary */
50#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
51
Linus Torvalds1da177e2005-04-16 15:20:36 -070052/*
53 * Linux kernel virtual memory manager primitives.
54 * The idea being to have a "virtual" mm in the same way
55 * we have a virtual fs - giving a cleaner interface to the
56 * mm details, and allowing different kinds of memory mappings
57 * (from shared memory to executable loading to arbitrary
58 * mmap() functions).
59 */
60
Christoph Lameterc43692e2006-12-06 20:32:48 -080061extern struct kmem_cache *vm_area_cachep;
62
Linus Torvalds1da177e2005-04-16 15:20:36 -070063#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +000064extern struct rb_root nommu_region_tree;
65extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
67extern unsigned int kobjsize(const void *objp);
68#endif
69
70/*
Hugh Dickins605d9282008-08-16 11:07:21 +010071 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 */
73#define VM_READ 0x00000001 /* currently active flags */
74#define VM_WRITE 0x00000002
75#define VM_EXEC 0x00000004
76#define VM_SHARED 0x00000008
77
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070078/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
80#define VM_MAYWRITE 0x00000020
81#define VM_MAYEXEC 0x00000040
82#define VM_MAYSHARE 0x00000080
83
84#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
Luck, Tony8ca3eb02010-08-24 11:44:18 -070085#if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
Linus Torvalds1da177e2005-04-16 15:20:36 -070086#define VM_GROWSUP 0x00000200
Luck, Tony8ca3eb02010-08-24 11:44:18 -070087#else
88#define VM_GROWSUP 0x00000000
Andrea Arcangelia664b2d2011-01-13 15:47:17 -080089#define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
Luck, Tony8ca3eb02010-08-24 11:44:18 -070090#endif
Linus Torvalds6aab3412005-11-28 14:34:23 -080091#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070092#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
93
94#define VM_EXECUTABLE 0x00001000
95#define VM_LOCKED 0x00002000
96#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
97
98 /* Used by sys_madvise() */
99#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
100#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
101
102#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
103#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800104#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -0700106#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
108#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
Johannes Weinerf2d6bfe2011-01-13 15:47:01 -0800109#ifndef CONFIG_TRANSPARENT_HUGEPAGE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Johannes Weinerf2d6bfe2011-01-13 15:47:01 -0800111#else
112#define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
113#endif
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700114#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Jason Baronaccb61f2012-03-23 15:02:51 -0700115#define VM_NODUMP 0x04000000 /* Do not include in the core dump */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Nick Piggind0217ac2007-07-19 01:47:03 -0700117#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700118#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000119#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700120#define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700121#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700122
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700123/* Bits set in the VMA until the stack is in its final location */
124#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
125
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
127#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
128#endif
129
130#ifdef CONFIG_STACK_GROWSUP
131#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
132#else
133#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
134#endif
135
136#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
137#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
138#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
139#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
140#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
141
142/*
Andrea Arcangeli78f11a22011-04-27 15:26:45 -0700143 * Special vmas that are non-mergable, non-mlock()able.
144 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
Nick Pigginb291f002008-10-18 20:26:44 -0700145 */
146#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
147
148/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 * mapping from the currently active vm_flags protection bits (the
150 * low four bits) to a page protection mask..
151 */
152extern pgprot_t protection_map[16];
153
Nick Piggind0217ac2007-07-19 01:47:03 -0700154#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
155#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700156#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Michel Lespinassed065bd82010-10-26 14:21:57 -0700157#define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
Gleb Natapov318b2752011-03-22 16:30:51 -0700158#define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
KOSAKI Motohiro37b23e02011-05-24 17:11:30 -0700159#define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
Nick Piggind0217ac2007-07-19 01:47:03 -0700160
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800161/*
162 * This interface is used by x86 PAT code to identify a pfn mapping that is
163 * linear over entire vma. This is to optimize PAT code that deals with
164 * marking the physical region with a particular prot. This is not for generic
165 * mm use. Note also that this check will not work if the pfn mapping is
166 * linear for a vma starting at physical address 0. In which case PAT code
167 * falls back to slow path of reserving physical range page by page.
168 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800169static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
170{
KOSAKI Motohiroca16d142011-05-26 19:16:19 +0900171 return !!(vma->vm_flags & VM_PFN_AT_MMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800172}
173
174static inline int is_pfn_mapping(struct vm_area_struct *vma)
175{
KOSAKI Motohiroca16d142011-05-26 19:16:19 +0900176 return !!(vma->vm_flags & VM_PFNMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800177}
Nick Piggind0217ac2007-07-19 01:47:03 -0700178
Nick Piggin54cb8822007-07-19 01:46:59 -0700179/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700180 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700181 * ->fault function. The vma's ->fault is responsible for returning a bitmask
182 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700183 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700184 * pgoff should be used in favour of virtual_address, if possible. If pgoff
185 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
186 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700187 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700188struct vm_fault {
189 unsigned int flags; /* FAULT_FLAG_xxx flags */
190 pgoff_t pgoff; /* Logical page offset based on vma */
191 void __user *virtual_address; /* Faulting virtual address */
192
193 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700194 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700195 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700196 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700197 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700198};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
200/*
201 * These are the virtual MM functions - opening of an area, closing and
202 * unmapping it (needed to keep files on disk up-to-date etc), pointer
203 * to the functions called when a no-page or a wp-page exception occurs.
204 */
205struct vm_operations_struct {
206 void (*open)(struct vm_area_struct * area);
207 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700208 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700209
210 /* notification that a previously read-only page is about to become
211 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700212 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700213
214 /* called by access_process_vm when get_user_pages() fails, typically
215 * for use by special VMAs that can switch between memory and hardware
216 */
217 int (*access)(struct vm_area_struct *vma, unsigned long addr,
218 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700220 /*
221 * set_policy() op must add a reference to any non-NULL @new mempolicy
222 * to hold the policy upon return. Caller should pass NULL @new to
223 * remove a policy and fall back to surrounding context--i.e. do not
224 * install a MPOL_DEFAULT policy, nor the task or system default
225 * mempolicy.
226 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700228
229 /*
230 * get_policy() op must add reference [mpol_get()] to any policy at
231 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
232 * in mm/mempolicy.c will do this automatically.
233 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
234 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
235 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
236 * must return NULL--i.e., do not "fallback" to task or system default
237 * policy.
238 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
240 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700241 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
242 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243#endif
244};
245
246struct mmu_gather;
247struct inode;
248
Andrew Morton349aef02006-01-08 01:04:36 -0800249#define page_private(page) ((page)->private)
250#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700251
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252/*
253 * FIXME: take this include out, include page-flags.h in
254 * files which need it (119 of them)
255 */
256#include <linux/page-flags.h>
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800257#include <linux/huge_mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
259/*
260 * Methods to modify the page usage count.
261 *
262 * What counts for a page usage:
263 * - cache mapping (page->mapping)
264 * - private data (page->private)
265 * - page mapped in a task's page tables, each mapping
266 * is counted separately
267 *
268 * Also, many kernel routines increase the page count before a critical
269 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 */
271
272/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700273 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800275static inline int put_page_testzero(struct page *page)
276{
Nick Piggin725d7042006-09-25 23:30:55 -0700277 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800278 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800279}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280
281/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800282 * Try to grab a ref unless the page has a refcount of zero, return false if
283 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800285static inline int get_page_unless_zero(struct page *page)
286{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800287 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800288}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800290extern int page_is_ram(unsigned long pfn);
291
Christoph Lameter48667e72008-02-04 22:28:31 -0800292/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800293struct page *vmalloc_to_page(const void *addr);
294unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800295
Paul Mundt0738c4b2008-03-12 16:51:31 +0900296/*
297 * Determine if an address is within the vmalloc range
298 *
299 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
300 * is no special casing required.
301 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800302static inline int is_vmalloc_addr(const void *x)
303{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900304#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800305 unsigned long addr = (unsigned long)x;
306
307 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900308#else
309 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800310#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900311}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700312#ifdef CONFIG_MMU
313extern int is_vmalloc_or_module_addr(const void *x);
314#else
David Howells934831d2009-09-24 12:33:32 +0100315static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700316{
317 return 0;
318}
319#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800320
Andrea Arcangelie9da73d2011-01-13 15:46:32 -0800321static inline void compound_lock(struct page *page)
322{
323#ifdef CONFIG_TRANSPARENT_HUGEPAGE
324 bit_spin_lock(PG_compound_lock, &page->flags);
325#endif
326}
327
328static inline void compound_unlock(struct page *page)
329{
330#ifdef CONFIG_TRANSPARENT_HUGEPAGE
331 bit_spin_unlock(PG_compound_lock, &page->flags);
332#endif
333}
334
335static inline unsigned long compound_lock_irqsave(struct page *page)
336{
337 unsigned long uninitialized_var(flags);
338#ifdef CONFIG_TRANSPARENT_HUGEPAGE
339 local_irq_save(flags);
340 compound_lock(page);
341#endif
342 return flags;
343}
344
345static inline void compound_unlock_irqrestore(struct page *page,
346 unsigned long flags)
347{
348#ifdef CONFIG_TRANSPARENT_HUGEPAGE
349 compound_unlock(page);
350 local_irq_restore(flags);
351#endif
352}
353
Christoph Lameterd85f3382007-05-06 14:49:39 -0700354static inline struct page *compound_head(struct page *page)
355{
Christoph Lameter6d777952007-05-06 14:49:40 -0700356 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700357 return page->first_page;
358 return page;
359}
360
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700361/*
362 * The atomic page->_mapcount, starts from -1: so that transitions
363 * both from it and to it can be tracked, using atomic_inc_and_test
364 * and atomic_add_negative(-1).
365 */
366static inline void reset_page_mapcount(struct page *page)
367{
368 atomic_set(&(page)->_mapcount, -1);
369}
370
371static inline int page_mapcount(struct page *page)
372{
373 return atomic_read(&(page)->_mapcount) + 1;
374}
375
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700376static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700378 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379}
380
Andrea Arcangelib35a35b2011-11-02 13:37:36 -0700381static inline void get_huge_page_tail(struct page *page)
382{
383 /*
384 * __split_huge_page_refcount() cannot run
385 * from under us.
386 */
387 VM_BUG_ON(page_mapcount(page) < 0);
388 VM_BUG_ON(atomic_read(&page->_count) != 0);
389 atomic_inc(&page->_mapcount);
390}
391
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700392extern bool __get_page_tail(struct page *page);
393
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394static inline void get_page(struct page *page)
395{
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700396 if (unlikely(PageTail(page)))
397 if (likely(__get_page_tail(page)))
398 return;
Andrea Arcangeli91807062011-01-13 15:46:32 -0800399 /*
400 * Getting a normal page or the head of a compound page
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700401 * requires to already have an elevated page->_count.
Andrea Arcangeli91807062011-01-13 15:46:32 -0800402 */
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700403 VM_BUG_ON(atomic_read(&page->_count) <= 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 atomic_inc(&page->_count);
405}
406
Christoph Lameterb49af682007-05-06 14:49:41 -0700407static inline struct page *virt_to_head_page(const void *x)
408{
409 struct page *page = virt_to_page(x);
410 return compound_head(page);
411}
412
Nick Piggin7835e982006-03-22 00:08:40 -0800413/*
414 * Setup the page count before being freed into the page allocator for
415 * the first time (boot or memory hotplug)
416 */
417static inline void init_page_count(struct page *page)
418{
419 atomic_set(&page->_count, 1);
420}
421
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800422/*
423 * PageBuddy() indicate that the page is free and in the buddy system
424 * (see mm/page_alloc.c).
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100425 *
426 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
427 * -2 so that an underflow of the page_mapcount() won't be mistaken
428 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
429 * efficiently by most CPU architectures.
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800430 */
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100431#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
432
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800433static inline int PageBuddy(struct page *page)
434{
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100435 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800436}
437
438static inline void __SetPageBuddy(struct page *page)
439{
440 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100441 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800442}
443
444static inline void __ClearPageBuddy(struct page *page)
445{
446 VM_BUG_ON(!PageBuddy(page));
447 atomic_set(&page->_mapcount, -1);
448}
449
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700451void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800453void split_page(struct page *page, unsigned int order);
Mel Gorman748446b2010-05-24 14:32:27 -0700454int split_free_page(struct page *page);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800455
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800457 * Compound pages have a destructor function. Provide a
458 * prototype for that function and accessor functions.
459 * These are _only_ valid on the head of a PG_compound page.
460 */
461typedef void compound_page_dtor(struct page *);
462
463static inline void set_compound_page_dtor(struct page *page,
464 compound_page_dtor *dtor)
465{
466 page[1].lru.next = (void *)dtor;
467}
468
469static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
470{
471 return (compound_page_dtor *)page[1].lru.next;
472}
473
Christoph Lameterd85f3382007-05-06 14:49:39 -0700474static inline int compound_order(struct page *page)
475{
Christoph Lameter6d777952007-05-06 14:49:40 -0700476 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700477 return 0;
478 return (unsigned long)page[1].lru.prev;
479}
480
Andrea Arcangeli37c2ac72011-01-13 15:47:16 -0800481static inline int compound_trans_order(struct page *page)
482{
483 int order;
484 unsigned long flags;
485
486 if (!PageHead(page))
487 return 0;
488
489 flags = compound_lock_irqsave(page);
490 order = compound_order(page);
491 compound_unlock_irqrestore(page, flags);
492 return order;
493}
494
Christoph Lameterd85f3382007-05-06 14:49:39 -0700495static inline void set_compound_order(struct page *page, unsigned long order)
496{
497 page[1].lru.prev = (void *)order;
498}
499
Michal Simek3dece372011-01-21 08:49:56 +0100500#ifdef CONFIG_MMU
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800501/*
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800502 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
503 * servicing faults for write access. In the normal case, do always want
504 * pte_mkwrite. But get_user_pages can cause write faults for mappings
505 * that do not have writing enabled, when used by access_process_vm.
506 */
507static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
508{
509 if (likely(vma->vm_flags & VM_WRITE))
510 pte = pte_mkwrite(pte);
511 return pte;
512}
Michal Simek3dece372011-01-21 08:49:56 +0100513#endif
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800514
515/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 * Multiple processes may "see" the same page. E.g. for untouched
517 * mappings of /dev/null, all processes see the same page full of
518 * zeroes, and text pages of executables and shared libraries have
519 * only one copy in memory, at most, normally.
520 *
521 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700522 * page_count() == 0 means the page is free. page->lru is then used for
523 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700524 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700526 * Pages are allocated by the slab allocator in order to provide memory
527 * to kmalloc and kmem_cache_alloc. In this case, the management of the
528 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
529 * unless a particular usage is carefully commented. (the responsibility of
530 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700532 * A page may be used by anyone else who does a __get_free_page().
533 * In this case, page_count still tracks the references, and should only
534 * be used through the normal accessor functions. The top bits of page->flags
535 * and page->virtual store page management information, but all other fields
536 * are unused and could be used privately, carefully. The management of this
537 * page is the responsibility of the one who allocated it, and those who have
538 * subsequently been given references to it.
539 *
540 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 * managed by the Linux memory manager: I/O, buffers, swapping etc.
542 * The following discussion applies only to them.
543 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700544 * A pagecache page contains an opaque `private' member, which belongs to the
545 * page's address_space. Usually, this is the address of a circular list of
546 * the page's disk buffers. PG_private must be set to tell the VM to call
547 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700549 * A page may belong to an inode's memory mapping. In this case, page->mapping
550 * is the pointer to the inode, and page->index is the file offset of the page,
551 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700553 * If pagecache pages are not associated with an inode, they are said to be
554 * anonymous pages. These may become associated with the swapcache, and in that
555 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700557 * In either case (swapcache or inode backed), the pagecache itself holds one
558 * reference to the page. Setting PG_private should also increment the
559 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700561 * The pagecache pages are stored in a per-mapping radix tree, which is
562 * rooted at mapping->page_tree, and indexed by offset.
563 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
564 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700566 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 * - inode pages may need to be read from disk,
568 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700569 * to be written back to the inode on disk,
570 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
571 * modified may need to be swapped out to swap space and (later) to be read
572 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 */
574
575/*
576 * The zone field is never updated after free_area_init_core()
577 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700579
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700580
581/*
582 * page->flags layout:
583 *
584 * There are three possibilities for how page->flags get
585 * laid out. The first is for the normal case, without
586 * sparsemem. The second is for sparsemem when there is
587 * plenty of space for node and section. The last is when
588 * we have run out of space and have to fall back to an
589 * alternate (slower) way of determining the node.
590 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700591 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
592 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
593 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700594 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700595#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700596#define SECTIONS_WIDTH SECTIONS_SHIFT
597#else
598#define SECTIONS_WIDTH 0
599#endif
600
601#define ZONES_WIDTH ZONES_SHIFT
602
Christoph Lameter9223b412008-04-28 02:12:48 -0700603#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700604#define NODES_WIDTH NODES_SHIFT
605#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700606#ifdef CONFIG_SPARSEMEM_VMEMMAP
607#error "Vmemmap: No space for nodes field in page flags"
608#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700609#define NODES_WIDTH 0
610#endif
611
612/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100613#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700614#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
615#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
616
617/*
618 * We are going to use the flags for the page to node mapping if its in
619 * there. This includes the case where there is no node, so it is implicit.
620 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800621#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
622#define NODE_NOT_IN_PAGE_FLAGS
623#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700624
Dave Hansen348f8b62005-06-23 00:07:40 -0700625/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300626 * Define the bit shifts to access each section. For non-existent
Dave Hansen348f8b62005-06-23 00:07:40 -0700627 * sections we define the shift as 0; that plus a 0 mask ensures
628 * the compiler will optimise away reference to them.
629 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700630#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
631#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
632#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700633
Will Deaconbce54bb2010-10-26 14:21:37 -0700634/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
635#ifdef NODE_NOT_IN_PAGE_FLAGS
Christoph Lameter89689ae2006-12-06 20:31:45 -0800636#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800637#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
638 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700639#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800640#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800641#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
642 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700643#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800644
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800645#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700646
Christoph Lameter9223b412008-04-28 02:12:48 -0700647#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
648#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700649#endif
650
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700651#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
652#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
653#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800654#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700655
Ian Campbell33dd4e02011-07-25 17:11:51 -0700656static inline enum zone_type page_zonenum(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657{
Dave Hansen348f8b62005-06-23 00:07:40 -0700658 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
Christoph Lameter89689ae2006-12-06 20:31:45 -0800661/*
662 * The identification function is only used by the buddy allocator for
663 * determining if two pages could be buddies. We are not really
664 * identifying a zone since we could be using a the section number
665 * id if we have not node id available in page flags.
666 * We guarantee only that it will return the same value for two
667 * combinable pages in a zone.
668 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700669static inline int page_zone_id(struct page *page)
670{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800671 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672}
673
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800674static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700675{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700676#ifdef CONFIG_NUMA
677 return zone->node;
678#else
679 return 0;
680#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700681}
682
Christoph Lameter89689ae2006-12-06 20:31:45 -0800683#ifdef NODE_NOT_IN_PAGE_FLAGS
Ian Campbell33dd4e02011-07-25 17:11:51 -0700684extern int page_to_nid(const struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800685#else
Ian Campbell33dd4e02011-07-25 17:11:51 -0700686static inline int page_to_nid(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700687{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800688 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700689}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800690#endif
691
Ian Campbell33dd4e02011-07-25 17:11:51 -0700692static inline struct zone *page_zone(const struct page *page)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800693{
694 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
695}
696
Christoph Lameter308c05e2008-04-28 02:12:43 -0700697#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700698static inline void set_page_section(struct page *page, unsigned long section)
699{
700 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
701 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
702}
703
Ian Campbellaa462ab2011-08-17 17:40:33 +0100704static inline unsigned long page_to_section(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700705{
706 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
707}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700708#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700709
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700710static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711{
Dave Hansen348f8b62005-06-23 00:07:40 -0700712 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
713 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
714}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700715
Dave Hansen348f8b62005-06-23 00:07:40 -0700716static inline void set_page_node(struct page *page, unsigned long node)
717{
718 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
719 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
720}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800721
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700722static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700723 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700724{
725 set_page_zone(page, zone);
726 set_page_node(page, node);
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700727#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700728 set_page_section(page, pfn_to_section_nr(pfn));
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700729#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730}
731
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700732/*
733 * Some inline functions in vmstat.h depend on page_zone()
734 */
735#include <linux/vmstat.h>
736
Ian Campbell33dd4e02011-07-25 17:11:51 -0700737static __always_inline void *lowmem_page_address(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738{
Ian Campbellaa462ab2011-08-17 17:40:33 +0100739 return __va(PFN_PHYS(page_to_pfn(page)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740}
741
742#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
743#define HASHED_PAGE_VIRTUAL
744#endif
745
746#if defined(WANT_PAGE_VIRTUAL)
747#define page_address(page) ((page)->virtual)
748#define set_page_address(page, address) \
749 do { \
750 (page)->virtual = (address); \
751 } while(0)
752#define page_address_init() do { } while(0)
753#endif
754
755#if defined(HASHED_PAGE_VIRTUAL)
Ian Campbellf9918792011-08-17 13:45:09 +0100756void *page_address(const struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757void set_page_address(struct page *page, void *virtual);
758void page_address_init(void);
759#endif
760
761#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
762#define page_address(page) lowmem_page_address(page)
763#define set_page_address(page, address) do { } while(0)
764#define page_address_init() do { } while(0)
765#endif
766
767/*
768 * On an anonymous page mapped into a user virtual memory area,
769 * page->mapping points to its anon_vma, not to a struct address_space;
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800770 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
771 *
772 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
773 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
774 * and then page->mapping points, not to an anon_vma, but to a private
775 * structure which KSM associates with that merged page. See ksm.h.
776 *
777 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 *
779 * Please note that, confusingly, "page_mapping" refers to the inode
780 * address_space which maps the page from disk; whereas "page_mapped"
781 * refers to user virtual address space into which the page is mapped.
782 */
783#define PAGE_MAPPING_ANON 1
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800784#define PAGE_MAPPING_KSM 2
785#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
787extern struct address_space swapper_space;
788static inline struct address_space *page_mapping(struct page *page)
789{
790 struct address_space *mapping = page->mapping;
791
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700792 VM_BUG_ON(PageSlab(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 if (unlikely(PageSwapCache(page)))
794 mapping = &swapper_space;
Steven Rostedte20e8772011-01-13 15:46:17 -0800795 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 mapping = NULL;
797 return mapping;
798}
799
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800800/* Neutral page->mapping pointer to address_space or anon_vma or other */
801static inline void *page_rmapping(struct page *page)
802{
803 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
804}
805
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806static inline int PageAnon(struct page *page)
807{
808 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
809}
810
811/*
812 * Return the pagecache index of the passed page. Regular pagecache pages
813 * use ->index whereas swapcache pages use ->private
814 */
815static inline pgoff_t page_index(struct page *page)
816{
817 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700818 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 return page->index;
820}
821
822/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 * Return true if this page is mapped into pagetables.
824 */
825static inline int page_mapped(struct page *page)
826{
827 return atomic_read(&(page)->_mapcount) >= 0;
828}
829
830/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 * Different kinds of faults, as returned by handle_mm_fault().
832 * Used to decide whether a process gets delivered SIGBUS or
833 * just gets major/minor fault counters bumped up.
834 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700835
Nick Piggin83c54072007-07-19 01:47:05 -0700836#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700837
Nick Piggin83c54072007-07-19 01:47:05 -0700838#define VM_FAULT_OOM 0x0001
839#define VM_FAULT_SIGBUS 0x0002
840#define VM_FAULT_MAJOR 0x0004
841#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleenaa50d3a2010-10-06 21:45:00 +0200842#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
843#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000844
Nick Piggin83c54072007-07-19 01:47:05 -0700845#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
846#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Michel Lespinassed065bd82010-10-26 14:21:57 -0700847#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848
Andi Kleenaa50d3a2010-10-06 21:45:00 +0200849#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
850
851#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
852 VM_FAULT_HWPOISON_LARGE)
853
854/* Encode hstate index for a hwpoisoned large page */
855#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
856#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
Nick Piggind0217ac2007-07-19 01:47:03 -0700857
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800858/*
859 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
860 */
861extern void pagefault_out_of_memory(void);
862
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
864
David Rientjesddd588b2011-03-22 16:30:46 -0700865/*
David Rientjes7bf02ea2011-05-24 17:11:16 -0700866 * Flags passed to show_mem() and show_free_areas() to suppress output in
David Rientjesddd588b2011-03-22 16:30:46 -0700867 * various contexts.
868 */
869#define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
870
David Rientjes7bf02ea2011-05-24 17:11:16 -0700871extern void show_free_areas(unsigned int flags);
872extern bool skip_free_areas_node(unsigned int flags, int nid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873
Hugh Dickins3f96b792009-09-21 17:03:37 -0700874int shmem_lock(struct file *file, int lock, struct user_struct *user);
Sergei Trofimovich168f5ac2009-06-16 15:33:02 -0700875struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876int shmem_zero_setup(struct vm_area_struct *);
877
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400878extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879extern int user_shm_lock(size_t, struct user_struct *);
880extern void user_shm_unlock(size_t, struct user_struct *);
881
882/*
883 * Parameter block passed down to zap_pte_range in exceptional cases.
884 */
885struct zap_details {
886 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
887 struct address_space *check_mapping; /* Check page->mapping if set */
888 pgoff_t first_index; /* Lowest page->index to unmap */
889 pgoff_t last_index; /* Highest page->index to unmap */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890};
891
Nick Piggin7e675132008-04-28 02:13:00 -0700892struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
893 pte_t pte);
894
Jack Steinerc627f9c2008-07-29 22:33:53 -0700895int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
896 unsigned long size);
Al Viro14f5ff52012-03-05 13:38:09 -0500897void zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 unsigned long size, struct zap_details *);
Al Viro6e8bb012012-03-05 13:41:15 -0500899void unmap_vmas(struct mmu_gather *tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 struct vm_area_struct *start_vma, unsigned long start_addr,
901 unsigned long end_addr, unsigned long *nr_accounted,
902 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800903
904/**
905 * mm_walk - callbacks for walk_page_range
906 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
907 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
908 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
Dave Hansen03319322011-03-22 16:32:56 -0700909 * this handler is required to be able to handle
910 * pmd_trans_huge() pmds. They may simply choose to
911 * split_huge_page() instead of handling it explicitly.
Matt Mackalle6473092008-02-04 22:29:01 -0800912 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
913 * @pte_hole: if set, called for each hole at all levels
Naoya Horiguchi5dc37642009-12-14 18:00:01 -0800914 * @hugetlb_entry: if set, called for each hugetlb entry
KOSAKI Motohiroc27fe4c82011-07-25 17:12:10 -0700915 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
916 * is used.
Matt Mackalle6473092008-02-04 22:29:01 -0800917 *
918 * (see walk_page_range for more details)
919 */
920struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700921 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
922 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
923 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
924 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
925 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
Naoya Horiguchi116354d2010-04-06 14:35:04 -0700926 int (*hugetlb_entry)(pte_t *, unsigned long,
927 unsigned long, unsigned long, struct mm_walk *);
Dave Hansen21650092008-06-12 15:21:47 -0700928 struct mm_struct *mm;
929 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800930};
931
Dave Hansen21650092008-06-12 15:21:47 -0700932int walk_page_range(unsigned long addr, unsigned long end,
933 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700934void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700935 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
937 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938void unmap_mapping_range(struct address_space *mapping,
939 loff_t const holebegin, loff_t const holelen, int even_cows);
Johannes Weiner3b6748e2009-06-16 15:32:35 -0700940int follow_pfn(struct vm_area_struct *vma, unsigned long address,
941 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800942int follow_phys(struct vm_area_struct *vma, unsigned long address,
943 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700944int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
945 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946
947static inline void unmap_shared_mapping_range(struct address_space *mapping,
948 loff_t const holebegin, loff_t const holelen)
949{
950 unmap_mapping_range(mapping, holebegin, holelen, 0);
951}
952
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000953extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
Christoph Hellwig2c27c652010-06-04 11:30:04 +0200954extern void truncate_setsize(struct inode *inode, loff_t newsize);
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000955extern int vmtruncate(struct inode *inode, loff_t offset);
956extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
Hugh Dickins623e3db2012-03-28 14:42:40 -0700957void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
Nick Piggin750b4982009-09-16 11:50:12 +0200958int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +0200959int generic_error_remove_page(struct address_space *mapping, struct page *page);
Nick Piggin750b4982009-09-16 11:50:12 +0200960
Wu Fengguang83f78662009-09-16 11:50:13 +0200961int invalidate_inode_page(struct page *page);
962
David Howells7ee1dd32006-01-06 00:11:44 -0800963#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700964extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700965 unsigned long address, unsigned int flags);
Peter Zijlstra5c723ba2011-07-27 12:17:11 +0200966extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
967 unsigned long address, unsigned int fault_flags);
David Howells7ee1dd32006-01-06 00:11:44 -0800968#else
969static inline int handle_mm_fault(struct mm_struct *mm,
970 struct vm_area_struct *vma, unsigned long address,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700971 unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -0800972{
973 /* should never happen if there's no MMU */
974 BUG();
975 return VM_FAULT_SIGBUS;
976}
Peter Zijlstra5c723ba2011-07-27 12:17:11 +0200977static inline int fixup_user_fault(struct task_struct *tsk,
978 struct mm_struct *mm, unsigned long address,
979 unsigned int fault_flags)
980{
981 /* should never happen if there's no MMU */
982 BUG();
983 return -EFAULT;
984}
David Howells7ee1dd32006-01-06 00:11:44 -0800985#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000986
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987extern int make_pages_present(unsigned long addr, unsigned long end);
988extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Stephen Wilson5ddd36b2011-03-13 15:49:20 -0400989extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
990 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991
Huang Ying0014bd92011-01-30 11:15:47 +0800992int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
993 unsigned long start, int len, unsigned int foll_flags,
994 struct page **pages, struct vm_area_struct **vmas,
995 int *nonblocking);
Nick Piggind2bf6be2009-06-16 15:31:39 -0700996int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
Peter Zijlstra9d737772009-06-25 11:58:55 +0200997 unsigned long start, int nr_pages, int write, int force,
Nick Piggind2bf6be2009-06-16 15:31:39 -0700998 struct page **pages, struct vm_area_struct **vmas);
999int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1000 struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07001001struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002
David Howellscf9a2ae2006-08-29 19:05:54 +01001003extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1004extern void do_invalidatepage(struct page *page, unsigned long offset);
1005
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -08001007int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008int redirty_page_for_writepage(struct writeback_control *wbc,
1009 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -07001010void account_page_dirtied(struct page *page, struct address_space *mapping);
Michael Rubinf629d1c2010-10-26 14:21:33 -07001011void account_page_writeback(struct page *page);
Harvey Harrisonb3c97522008-02-13 15:03:15 -08001012int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013int set_page_dirty_lock(struct page *page);
1014int clear_page_dirty_for_io(struct page *page);
1015
Stefan Bader39aa3cb2010-08-31 15:52:27 +02001016/* Is the vma a continuation of the stack vma above it? */
Mikulas Patockaa09a79f2011-05-09 13:01:09 +02001017static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
Stefan Bader39aa3cb2010-08-31 15:52:27 +02001018{
1019 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1020}
1021
Mikulas Patockaa09a79f2011-05-09 13:01:09 +02001022static inline int stack_guard_page_start(struct vm_area_struct *vma,
1023 unsigned long addr)
1024{
1025 return (vma->vm_flags & VM_GROWSDOWN) &&
1026 (vma->vm_start == addr) &&
1027 !vma_growsdown(vma->vm_prev, addr);
1028}
1029
1030/* Is the vma a continuation of the stack vma below it? */
1031static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1032{
1033 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1034}
1035
1036static inline int stack_guard_page_end(struct vm_area_struct *vma,
1037 unsigned long addr)
1038{
1039 return (vma->vm_flags & VM_GROWSUP) &&
1040 (vma->vm_end == addr) &&
1041 !vma_growsup(vma->vm_next, addr);
1042}
1043
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07001044extern pid_t
1045vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1046
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001047extern unsigned long move_page_tables(struct vm_area_struct *vma,
1048 unsigned long old_addr, struct vm_area_struct *new_vma,
1049 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050extern unsigned long do_mremap(unsigned long addr,
1051 unsigned long old_len, unsigned long new_len,
1052 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001053extern int mprotect_fixup(struct vm_area_struct *vma,
1054 struct vm_area_struct **pprev, unsigned long start,
1055 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
Nick Piggin21cc1992008-07-25 19:45:22 -07001057/*
Peter Zijlstra465a4542009-06-15 12:31:37 +02001058 * doesn't attempt to fault and will return short.
1059 */
1060int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1061 struct page **pages);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001062/*
1063 * per-process(per-mm_struct) statistics.
1064 */
Matt Fleming172703b2011-05-24 17:12:36 -07001065static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1066{
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001067 long val = atomic_long_read(&mm->rss_stat.count[member]);
1068
1069#ifdef SPLIT_RSS_COUNTING
1070 /*
1071 * counter is updated in asynchronous manner and may go to minus.
1072 * But it's never be expected number for users.
1073 */
1074 if (val < 0)
1075 val = 0;
Matt Fleming172703b2011-05-24 17:12:36 -07001076#endif
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001077 return (unsigned long)val;
1078}
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001079
1080static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1081{
1082 atomic_long_add(value, &mm->rss_stat.count[member]);
1083}
1084
1085static inline void inc_mm_counter(struct mm_struct *mm, int member)
1086{
1087 atomic_long_inc(&mm->rss_stat.count[member]);
1088}
1089
1090static inline void dec_mm_counter(struct mm_struct *mm, int member)
1091{
1092 atomic_long_dec(&mm->rss_stat.count[member]);
1093}
1094
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001095static inline unsigned long get_mm_rss(struct mm_struct *mm)
1096{
1097 return get_mm_counter(mm, MM_FILEPAGES) +
1098 get_mm_counter(mm, MM_ANONPAGES);
1099}
1100
1101static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1102{
1103 return max(mm->hiwater_rss, get_mm_rss(mm));
1104}
1105
1106static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1107{
1108 return max(mm->hiwater_vm, mm->total_vm);
1109}
1110
1111static inline void update_hiwater_rss(struct mm_struct *mm)
1112{
1113 unsigned long _rss = get_mm_rss(mm);
1114
1115 if ((mm)->hiwater_rss < _rss)
1116 (mm)->hiwater_rss = _rss;
1117}
1118
1119static inline void update_hiwater_vm(struct mm_struct *mm)
1120{
1121 if (mm->hiwater_vm < mm->total_vm)
1122 mm->hiwater_vm = mm->total_vm;
1123}
1124
1125static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1126 struct mm_struct *mm)
1127{
1128 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1129
1130 if (*maxrss < hiwater_rss)
1131 *maxrss = hiwater_rss;
1132}
1133
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001134#if defined(SPLIT_RSS_COUNTING)
David Rientjes05af2e12012-03-21 16:34:13 -07001135void sync_mm_rss(struct mm_struct *mm);
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001136#else
David Rientjes05af2e12012-03-21 16:34:13 -07001137static inline void sync_mm_rss(struct mm_struct *mm)
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001138{
1139}
1140#endif
Peter Zijlstra465a4542009-06-15 12:31:37 +02001141
Alexey Dobriyan4e950f62007-07-30 02:36:13 +04001142int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -07001143
Namhyung Kim25ca1d62010-10-26 14:21:59 -07001144extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1145 spinlock_t **ptl);
1146static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1147 spinlock_t **ptl)
1148{
1149 pte_t *ptep;
1150 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1151 return ptep;
1152}
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -08001153
Nick Piggin5f22df02007-05-06 14:49:02 -07001154#ifdef __PAGETABLE_PUD_FOLDED
1155static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1156 unsigned long address)
1157{
1158 return 0;
1159}
1160#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001161int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001162#endif
1163
1164#ifdef __PAGETABLE_PMD_FOLDED
1165static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1166 unsigned long address)
1167{
1168 return 0;
1169}
1170#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001171int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001172#endif
1173
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001174int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1175 pmd_t *pmd, unsigned long address);
Hugh Dickins1bb36302005-10-29 18:16:22 -07001176int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1177
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178/*
1179 * The following ifdef needed to get the 4level-fixup.h header to work.
1180 * Remove it when 4level-fixup.h has been removed.
1181 */
Hugh Dickins1bb36302005-10-29 18:16:22 -07001182#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1184{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001185 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1186 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187}
1188
1189static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1190{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001191 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1192 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193}
Hugh Dickins1bb36302005-10-29 18:16:22 -07001194#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1195
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001196#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001197/*
1198 * We tuck a spinlock to guard each pagetable page into its struct page,
1199 * at page->private, with BUILD_BUG_ON to make sure that this will not
1200 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1201 * When freeing, reset page->mapping so free_pages_check won't complain.
1202 */
Andrew Morton349aef02006-01-08 01:04:36 -08001203#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001204#define pte_lock_init(_page) do { \
1205 spin_lock_init(__pte_lockptr(_page)); \
1206} while (0)
1207#define pte_lock_deinit(page) ((page)->mapping = NULL)
1208#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001209#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001210/*
1211 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1212 */
1213#define pte_lock_init(page) do {} while (0)
1214#define pte_lock_deinit(page) do {} while (0)
1215#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001216#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001217
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001218static inline void pgtable_page_ctor(struct page *page)
1219{
1220 pte_lock_init(page);
1221 inc_zone_page_state(page, NR_PAGETABLE);
1222}
1223
1224static inline void pgtable_page_dtor(struct page *page)
1225{
1226 pte_lock_deinit(page);
1227 dec_zone_page_state(page, NR_PAGETABLE);
1228}
1229
Hugh Dickinsc74df322005-10-29 18:16:23 -07001230#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1231({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001232 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001233 pte_t *__pte = pte_offset_map(pmd, address); \
1234 *(ptlp) = __ptl; \
1235 spin_lock(__ptl); \
1236 __pte; \
1237})
1238
1239#define pte_unmap_unlock(pte, ptl) do { \
1240 spin_unlock(ptl); \
1241 pte_unmap(pte); \
1242} while (0)
1243
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001244#define pte_alloc_map(mm, vma, pmd, address) \
1245 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1246 pmd, address))? \
1247 NULL: pte_offset_map(pmd, address))
Hugh Dickins1bb36302005-10-29 18:16:22 -07001248
Hugh Dickinsc74df322005-10-29 18:16:23 -07001249#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001250 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1251 pmd, address))? \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001252 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1253
Hugh Dickins1bb36302005-10-29 18:16:22 -07001254#define pte_alloc_kernel(pmd, address) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001255 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
Hugh Dickins1bb36302005-10-29 18:16:22 -07001256 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257
1258extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001259extern void free_area_init_node(int nid, unsigned long * zones_size,
1260 unsigned long zone_start_pfn, unsigned long *zholes_size);
David Howells49a7f042012-03-28 18:30:03 +01001261extern void free_initmem(void);
1262
Tejun Heo0ee332c2011-12-08 10:22:09 -08001263#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gormanc7132162006-09-27 01:49:43 -07001264/*
Tejun Heo0ee332c2011-12-08 10:22:09 -08001265 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
Mel Gormanc7132162006-09-27 01:49:43 -07001266 * zones, allocate the backing mem_map and account for memory holes in a more
1267 * architecture independent manner. This is a substitute for creating the
1268 * zone_sizes[] and zholes_size[] arrays and passing them to
1269 * free_area_init_node()
1270 *
1271 * An architecture is expected to register range of page frames backed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08001272 * physical memory with memblock_add[_node]() before calling
Mel Gormanc7132162006-09-27 01:49:43 -07001273 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1274 * usage, an architecture is expected to do something like
1275 *
1276 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1277 * max_highmem_pfn};
1278 * for_each_valid_physical_page_range()
Tejun Heo0ee332c2011-12-08 10:22:09 -08001279 * memblock_add_node(base, size, nid)
Mel Gormanc7132162006-09-27 01:49:43 -07001280 * free_area_init_nodes(max_zone_pfns);
1281 *
Tejun Heo0ee332c2011-12-08 10:22:09 -08001282 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1283 * registered physical page range. Similarly
1284 * sparse_memory_present_with_active_regions() calls memory_present() for
1285 * each range when SPARSEMEM is enabled.
Mel Gormanc7132162006-09-27 01:49:43 -07001286 *
1287 * See mm/page_alloc.c for more information on each function exposed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08001288 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
Mel Gormanc7132162006-09-27 01:49:43 -07001289 */
1290extern void free_area_init_nodes(unsigned long *max_zone_pfn);
Tejun Heo1e019792011-07-12 09:45:34 +02001291unsigned long node_map_pfn_alignment(void);
Yinghai Lu32996252009-12-15 17:59:02 -08001292unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1293 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001294extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1295 unsigned long end_pfn);
1296extern void get_pfn_range_for_nid(unsigned int nid,
1297 unsigned long *start_pfn, unsigned long *end_pfn);
1298extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001299extern void free_bootmem_with_active_regions(int nid,
1300 unsigned long max_low_pfn);
1301extern void sparse_memory_present_with_active_regions(int nid);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001302
Tejun Heo0ee332c2011-12-08 10:22:09 -08001303#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
Tejun Heo5dfe8662011-07-14 09:46:10 +02001304
Tejun Heo0ee332c2011-12-08 10:22:09 -08001305#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001306 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1307static inline int __early_pfn_to_nid(unsigned long pfn)
1308{
1309 return 0;
1310}
1311#else
1312/* please see mm/page_alloc.c */
1313extern int __meminit early_pfn_to_nid(unsigned long pfn);
1314#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1315/* there is a per-arch backend function. */
1316extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1317#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1318#endif
1319
Mel Gorman0e0b8642006-09-27 01:49:56 -07001320extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001321extern void memmap_init_zone(unsigned long, int, unsigned long,
1322 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07001323extern void setup_per_zone_wmarks(void);
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -07001324extern int __meminit init_per_zone_wmark_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001326extern void __init mmap_init(void);
David Rientjesb2b755b2011-03-24 15:18:15 -07001327extern void show_mem(unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328extern void si_meminfo(struct sysinfo * val);
1329extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001330extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331
Joe Perches3ee9a4f2011-10-31 17:08:35 -07001332extern __printf(3, 4)
1333void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
Dave Hansena238ab52011-05-24 17:12:16 -07001334
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001335extern void setup_per_cpu_pageset(void);
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001336
Shaohua Li112067f2009-09-21 17:01:16 -07001337extern void zone_pcp_update(struct zone *zone);
1338
David Howells8feae132009-01-08 12:04:47 +00001339/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001340extern atomic_long_t mmap_pages_allocated;
David Howells7e660872010-01-15 17:01:39 -08001341extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
David Howells8feae132009-01-08 12:04:47 +00001342
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343/* prio_tree.c */
1344void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1345void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1346void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1347struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1348 struct prio_tree_iter *iter);
1349
1350#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1351 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1352 (vma = vma_prio_tree_next(vma, iter)); )
1353
1354static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1355 struct list_head *list)
1356{
1357 vma->shared.vm_set.parent = NULL;
1358 list_add_tail(&vma->shared.vm_set.list, list);
1359}
1360
1361/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001362extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Rik van Riel5beb4932010-03-05 13:42:07 -08001363extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1365extern struct vm_area_struct *vma_merge(struct mm_struct *,
1366 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1367 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1368 struct mempolicy *);
1369extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1370extern int split_vma(struct mm_struct *,
1371 struct vm_area_struct *, unsigned long addr, int new_below);
1372extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1373extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1374 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001375extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1377 unsigned long addr, unsigned long len, pgoff_t pgoff);
1378extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001379
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001380extern int mm_take_all_locks(struct mm_struct *mm);
1381extern void mm_drop_all_locks(struct mm_struct *mm);
1382
Matt Helsley925d1c42008-04-29 01:01:36 -07001383/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1384extern void added_exe_file_vma(struct mm_struct *mm);
1385extern void removed_exe_file_vma(struct mm_struct *mm);
Jiri Slaby38646012011-05-26 16:25:46 -07001386extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1387extern struct file *get_mm_exe_file(struct mm_struct *mm);
Matt Helsley925d1c42008-04-29 01:01:36 -07001388
akpm@osdl.org119f6572005-05-01 08:58:35 -07001389extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001390extern int install_special_mapping(struct mm_struct *mm,
1391 unsigned long addr, unsigned long len,
1392 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
1394extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1395
1396extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1397 unsigned long len, unsigned long prot,
1398 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001399extern unsigned long mmap_region(struct file *file, unsigned long addr,
1400 unsigned long len, unsigned long flags,
KOSAKI Motohiroca16d142011-05-26 19:16:19 +09001401 vm_flags_t vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
1403static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1404 unsigned long len, unsigned long prot,
1405 unsigned long flag, unsigned long offset)
1406{
1407 unsigned long ret = -EINVAL;
1408 if ((offset + PAGE_ALIGN(len)) < offset)
1409 goto out;
1410 if (!(offset & ~PAGE_MASK))
1411 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1412out:
1413 return ret;
1414}
1415
1416extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1417
1418extern unsigned long do_brk(unsigned long, unsigned long);
1419
Hugh Dickins85821aa2011-07-25 17:12:23 -07001420/* truncate.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001422extern void truncate_inode_pages_range(struct address_space *,
1423 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424
1425/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001426extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427
1428/* mm/page-writeback.c */
1429int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001430void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
1432/* readahead.c */
1433#define VM_MAX_READAHEAD 128 /* kbytes */
1434#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001437 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001438
1439void page_cache_sync_readahead(struct address_space *mapping,
1440 struct file_ra_state *ra,
1441 struct file *filp,
1442 pgoff_t offset,
1443 unsigned long size);
1444
1445void page_cache_async_readahead(struct address_space *mapping,
1446 struct file_ra_state *ra,
1447 struct file *filp,
1448 struct page *pg,
1449 pgoff_t offset,
1450 unsigned long size);
1451
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452unsigned long max_sane_readahead(unsigned long nr);
Wu Fengguangd30a1102009-06-16 15:31:30 -07001453unsigned long ra_submit(struct file_ra_state *ra,
1454 struct address_space *mapping,
1455 struct file *filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
Michal Hockod05f3162011-05-24 17:11:44 -07001457/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001458extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Michal Hockod05f3162011-05-24 17:11:44 -07001459
1460/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1461extern int expand_downwards(struct vm_area_struct *vma,
1462 unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001463#if VM_GROWSUP
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001464extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001465#else
1466 #define expand_upwards(vma, address) do { } while (0)
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001467#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
1469/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1470extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1471extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1472 struct vm_area_struct **pprev);
1473
1474/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1475 NULL if none. Assume start_addr < end_addr. */
1476static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1477{
1478 struct vm_area_struct * vma = find_vma(mm,start_addr);
1479
1480 if (vma && end_addr <= vma->vm_start)
1481 vma = NULL;
1482 return vma;
1483}
1484
1485static inline unsigned long vma_pages(struct vm_area_struct *vma)
1486{
1487 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1488}
1489
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001490/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1491static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1492 unsigned long vm_start, unsigned long vm_end)
1493{
1494 struct vm_area_struct *vma = find_vma(mm, vm_start);
1495
1496 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1497 vma = NULL;
1498
1499 return vma;
1500}
1501
David Howellsbad849b2010-08-26 16:00:34 +01001502#ifdef CONFIG_MMU
Hugh Dickins804af2c2006-07-26 21:39:49 +01001503pgprot_t vm_get_page_prot(unsigned long vm_flags);
David Howellsbad849b2010-08-26 16:00:34 +01001504#else
1505static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1506{
1507 return __pgprot(0);
1508}
1509#endif
1510
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001511struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001512int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1513 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001514int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001515int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1516 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001517int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1518 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
Linus Torvalds6aab3412005-11-28 14:34:23 -08001520struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001521 unsigned int foll_flags);
1522#define FOLL_WRITE 0x01 /* check pte is writable */
1523#define FOLL_TOUCH 0x02 /* mark page accessed */
1524#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a602009-09-21 17:03:26 -07001525#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07001526#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
Gleb Natapov318b2752011-03-22 16:30:51 -07001527#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1528 * and return without waiting upon it */
Michel Lespinasse110d74a2011-01-13 15:46:11 -08001529#define FOLL_MLOCK 0x40 /* mark page as mlocked */
Andrea Arcangeli500d65d2011-01-13 15:46:55 -08001530#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
Huang Ying69ebb832011-01-30 11:15:48 +08001531#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001533typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001534 void *data);
1535extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1536 unsigned long size, pte_fn_t fn, void *data);
1537
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001539void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001541static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 unsigned long flags, struct file *file, long pages)
1543{
1544}
1545#endif /* CONFIG_PROC_FS */
1546
Ingo Molnar12d6f212008-01-30 13:33:58 +01001547#ifdef CONFIG_DEBUG_PAGEALLOC
Ingo Molnar12d6f212008-01-30 13:33:58 +01001548extern void kernel_map_pages(struct page *page, int numpages, int enable);
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001549#ifdef CONFIG_HIBERNATION
1550extern bool kernel_page_present(struct page *page);
1551#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001552#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001554kernel_map_pages(struct page *page, int numpages, int enable) {}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001555#ifdef CONFIG_HIBERNATION
1556static inline bool kernel_page_present(struct page *page) { return true; }
1557#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558#endif
1559
Stephen Wilson31db58b2011-03-13 15:49:15 -04001560extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561#ifdef __HAVE_ARCH_GATE_AREA
Stephen Wilsoncae5d392011-03-13 15:49:17 -04001562int in_gate_area_no_mm(unsigned long addr);
Stephen Wilson83b964b2011-03-13 15:49:16 -04001563int in_gate_area(struct mm_struct *mm, unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564#else
Stephen Wilsoncae5d392011-03-13 15:49:17 -04001565int in_gate_area_no_mm(unsigned long addr);
1566#define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567#endif /* __HAVE_ARCH_GATE_AREA */
1568
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07001569int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001570 void __user *, size_t *, loff_t *);
Ying Hana09ed5e2011-05-24 17:12:26 -07001571unsigned long shrink_slab(struct shrink_control *shrink,
Ying Han1495f232011-05-24 17:12:27 -07001572 unsigned long nr_pages_scanned,
1573 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001574
Luke Yang7a9166e2006-02-20 18:28:07 -08001575#ifndef CONFIG_MMU
1576#define randomize_va_space 0
1577#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001578extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001579#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001580
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001581const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001582void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001583
Yinghai Lu9bdac912010-02-10 01:20:22 -08001584void sparse_mem_maps_populate_node(struct page **map_map,
1585 unsigned long pnum_begin,
1586 unsigned long pnum_end,
1587 unsigned long map_count,
1588 int nodeid);
1589
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001590struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001591pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1592pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1593pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1594pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001595void *vmemmap_alloc_block(unsigned long size, int node);
Yinghai Lu9bdac912010-02-10 01:20:22 -08001596void *vmemmap_alloc_block_buf(unsigned long size, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001597void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001598int vmemmap_populate_basepages(struct page *start_page,
1599 unsigned long pages, int node);
1600int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001601void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001602
Andi Kleen6a460792009-09-16 11:50:15 +02001603
Andi Kleen82ba0112009-12-16 12:19:57 +01001604enum mf_flags {
1605 MF_COUNT_INCREASED = 1 << 0,
Tony Luck7329bbe2011-12-13 09:27:58 -08001606 MF_ACTION_REQUIRED = 1 << 1,
Andi Kleen82ba0112009-12-16 12:19:57 +01001607};
Tony Luckcd42f4a2011-12-15 10:48:12 -08001608extern int memory_failure(unsigned long pfn, int trapno, int flags);
Huang Yingea8f5fb2011-07-13 13:14:27 +08001609extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
Wu Fengguang847ce402009-12-16 12:19:58 +01001610extern int unpoison_memory(unsigned long pfn);
Andi Kleen6a460792009-09-16 11:50:15 +02001611extern int sysctl_memory_failure_early_kill;
1612extern int sysctl_memory_failure_recovery;
Andi Kleenfacb6012009-12-16 12:20:00 +01001613extern void shake_page(struct page *p, int access);
Andi Kleen6a460792009-09-16 11:50:15 +02001614extern atomic_long_t mce_bad_pages;
Andi Kleenfacb6012009-12-16 12:20:00 +01001615extern int soft_offline_page(struct page *page, int flags);
Andi Kleen6a460792009-09-16 11:50:15 +02001616
Wu Fengguang718a3822010-03-10 15:20:43 -08001617extern void dump_page(struct page *page);
1618
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08001619#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1620extern void clear_huge_page(struct page *page,
1621 unsigned long addr,
1622 unsigned int pages_per_huge_page);
1623extern void copy_user_huge_page(struct page *dst, struct page *src,
1624 unsigned long addr, struct vm_area_struct *vma,
1625 unsigned int pages_per_huge_page);
1626#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1627
Stanislaw Gruszkac0a32fc52012-01-10 15:07:28 -08001628#ifdef CONFIG_DEBUG_PAGEALLOC
1629extern unsigned int _debug_guardpage_minorder;
1630
1631static inline unsigned int debug_guardpage_minorder(void)
1632{
1633 return _debug_guardpage_minorder;
1634}
1635
1636static inline bool page_is_guard(struct page *page)
1637{
1638 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1639}
1640#else
1641static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1642static inline bool page_is_guard(struct page *page) { return false; }
1643#endif /* CONFIG_DEBUG_PAGEALLOC */
1644
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645#endif /* __KERNEL__ */
1646#endif /* _LINUX_MM_H */