blob: 679300c050f5720f07597d5ee4ac3ec3e564f19f [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>
9#include <linux/list.h>
10#include <linux/mmzone.h>
11#include <linux/rbtree.h>
12#include <linux/prio_tree.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070013#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070014#include <linux/mm_types.h>
Yinghai Lu08677212010-02-10 01:20:20 -080015#include <linux/range.h>
Chris Metcalfc6f6b592010-05-24 14:32:53 -070016#include <linux/pfn.h>
Andrea Arcangelie9da73d2011-01-13 15:46:32 -080017#include <linux/bit_spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018
19struct mempolicy;
20struct anon_vma;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040021struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040022struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040023struct writeback_control;
Linus Torvalds1da177e2005-04-16 15:20:36 -070024
25#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
26extern unsigned long max_mapnr;
27#endif
28
29extern unsigned long num_physpages;
Jan Beulich44813742009-09-21 17:03:05 -070030extern unsigned long totalram_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -070031extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070032extern int page_cluster;
33
34#ifdef CONFIG_SYSCTL
35extern int sysctl_legacy_va_layout;
36#else
37#define sysctl_legacy_va_layout 0
38#endif
39
40#include <asm/page.h>
41#include <asm/pgtable.h>
42#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
45
Andrea Righi27ac7922008-07-23 21:28:13 -070046/* to align the pointer to the (next) page boundary */
47#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
48
Linus Torvalds1da177e2005-04-16 15:20:36 -070049/*
50 * Linux kernel virtual memory manager primitives.
51 * The idea being to have a "virtual" mm in the same way
52 * we have a virtual fs - giving a cleaner interface to the
53 * mm details, and allowing different kinds of memory mappings
54 * (from shared memory to executable loading to arbitrary
55 * mmap() functions).
56 */
57
Christoph Lameterc43692e2006-12-06 20:32:48 -080058extern struct kmem_cache *vm_area_cachep;
59
Linus Torvalds1da177e2005-04-16 15:20:36 -070060#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +000061extern struct rb_root nommu_region_tree;
62extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64extern unsigned int kobjsize(const void *objp);
65#endif
66
67/*
Hugh Dickins605d9282008-08-16 11:07:21 +010068 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 */
70#define VM_READ 0x00000001 /* currently active flags */
71#define VM_WRITE 0x00000002
72#define VM_EXEC 0x00000004
73#define VM_SHARED 0x00000008
74
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070075/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
77#define VM_MAYWRITE 0x00000020
78#define VM_MAYEXEC 0x00000040
79#define VM_MAYSHARE 0x00000080
80
81#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
Luck, Tony8ca3eb02010-08-24 11:44:18 -070082#if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#define VM_GROWSUP 0x00000200
Luck, Tony8ca3eb02010-08-24 11:44:18 -070084#else
85#define VM_GROWSUP 0x00000000
Andrea Arcangelia664b2d2011-01-13 15:47:17 -080086#define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
Luck, Tony8ca3eb02010-08-24 11:44:18 -070087#endif
Linus Torvalds6aab3412005-11-28 14:34:23 -080088#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070089#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
90
91#define VM_EXECUTABLE 0x00001000
92#define VM_LOCKED 0x00002000
93#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
94
95 /* Used by sys_madvise() */
96#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
98
99#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800101#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -0700103#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
105#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
Johannes Weinerf2d6bfe2011-01-13 15:47:01 -0800106#ifndef CONFIG_TRANSPARENT_HUGEPAGE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Johannes Weinerf2d6bfe2011-01-13 15:47:01 -0800108#else
109#define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
110#endif
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700111#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800112#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Nick Piggind0217ac2007-07-19 01:47:03 -0700114#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700115#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000116#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700117#define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700118#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700119
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700120/* Bits set in the VMA until the stack is in its final location */
121#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
122
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
124#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
125#endif
126
127#ifdef CONFIG_STACK_GROWSUP
128#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
129#else
130#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
131#endif
132
133#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
134#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
135#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
136#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
137#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
138
139/*
Nick Pigginb291f002008-10-18 20:26:44 -0700140 * special vmas that are non-mergable, non-mlock()able
141 */
142#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
143
144/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 * mapping from the currently active vm_flags protection bits (the
146 * low four bits) to a page protection mask..
147 */
148extern pgprot_t protection_map[16];
149
Nick Piggind0217ac2007-07-19 01:47:03 -0700150#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
151#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700152#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Michel Lespinassed065bd82010-10-26 14:21:57 -0700153#define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
Nick Piggind0217ac2007-07-19 01:47:03 -0700154
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800155/*
156 * This interface is used by x86 PAT code to identify a pfn mapping that is
157 * linear over entire vma. This is to optimize PAT code that deals with
158 * marking the physical region with a particular prot. This is not for generic
159 * mm use. Note also that this check will not work if the pfn mapping is
160 * linear for a vma starting at physical address 0. In which case PAT code
161 * falls back to slow path of reserving physical range page by page.
162 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800163static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
164{
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700165 return (vma->vm_flags & VM_PFN_AT_MMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800166}
167
168static inline int is_pfn_mapping(struct vm_area_struct *vma)
169{
170 return (vma->vm_flags & VM_PFNMAP);
171}
Nick Piggind0217ac2007-07-19 01:47:03 -0700172
Nick Piggin54cb8822007-07-19 01:46:59 -0700173/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700174 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700175 * ->fault function. The vma's ->fault is responsible for returning a bitmask
176 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700177 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700178 * pgoff should be used in favour of virtual_address, if possible. If pgoff
179 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
180 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700181 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700182struct vm_fault {
183 unsigned int flags; /* FAULT_FLAG_xxx flags */
184 pgoff_t pgoff; /* Logical page offset based on vma */
185 void __user *virtual_address; /* Faulting virtual address */
186
187 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700188 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700189 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700190 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700191 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700192};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193
194/*
195 * These are the virtual MM functions - opening of an area, closing and
196 * unmapping it (needed to keep files on disk up-to-date etc), pointer
197 * to the functions called when a no-page or a wp-page exception occurs.
198 */
199struct vm_operations_struct {
200 void (*open)(struct vm_area_struct * area);
201 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700202 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700203
204 /* notification that a previously read-only page is about to become
205 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700206 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700207
208 /* called by access_process_vm when get_user_pages() fails, typically
209 * for use by special VMAs that can switch between memory and hardware
210 */
211 int (*access)(struct vm_area_struct *vma, unsigned long addr,
212 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700214 /*
215 * set_policy() op must add a reference to any non-NULL @new mempolicy
216 * to hold the policy upon return. Caller should pass NULL @new to
217 * remove a policy and fall back to surrounding context--i.e. do not
218 * install a MPOL_DEFAULT policy, nor the task or system default
219 * mempolicy.
220 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700222
223 /*
224 * get_policy() op must add reference [mpol_get()] to any policy at
225 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
226 * in mm/mempolicy.c will do this automatically.
227 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
228 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
229 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
230 * must return NULL--i.e., do not "fallback" to task or system default
231 * policy.
232 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
234 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700235 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
236 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237#endif
238};
239
240struct mmu_gather;
241struct inode;
242
Andrew Morton349aef02006-01-08 01:04:36 -0800243#define page_private(page) ((page)->private)
244#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700245
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246/*
247 * FIXME: take this include out, include page-flags.h in
248 * files which need it (119 of them)
249 */
250#include <linux/page-flags.h>
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800251#include <linux/huge_mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
253/*
254 * Methods to modify the page usage count.
255 *
256 * What counts for a page usage:
257 * - cache mapping (page->mapping)
258 * - private data (page->private)
259 * - page mapped in a task's page tables, each mapping
260 * is counted separately
261 *
262 * Also, many kernel routines increase the page count before a critical
263 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 */
265
266/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700267 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800269static inline int put_page_testzero(struct page *page)
270{
Nick Piggin725d7042006-09-25 23:30:55 -0700271 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800272 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800273}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
275/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800276 * Try to grab a ref unless the page has a refcount of zero, return false if
277 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800279static inline int get_page_unless_zero(struct page *page)
280{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800281 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800282}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800284extern int page_is_ram(unsigned long pfn);
285
Christoph Lameter48667e72008-02-04 22:28:31 -0800286/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800287struct page *vmalloc_to_page(const void *addr);
288unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800289
Paul Mundt0738c4b2008-03-12 16:51:31 +0900290/*
291 * Determine if an address is within the vmalloc range
292 *
293 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
294 * is no special casing required.
295 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800296static inline int is_vmalloc_addr(const void *x)
297{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900298#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800299 unsigned long addr = (unsigned long)x;
300
301 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900302#else
303 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800304#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900305}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700306#ifdef CONFIG_MMU
307extern int is_vmalloc_or_module_addr(const void *x);
308#else
David Howells934831d2009-09-24 12:33:32 +0100309static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700310{
311 return 0;
312}
313#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800314
Andrea Arcangelie9da73d2011-01-13 15:46:32 -0800315static inline void compound_lock(struct page *page)
316{
317#ifdef CONFIG_TRANSPARENT_HUGEPAGE
318 bit_spin_lock(PG_compound_lock, &page->flags);
319#endif
320}
321
322static inline void compound_unlock(struct page *page)
323{
324#ifdef CONFIG_TRANSPARENT_HUGEPAGE
325 bit_spin_unlock(PG_compound_lock, &page->flags);
326#endif
327}
328
329static inline unsigned long compound_lock_irqsave(struct page *page)
330{
331 unsigned long uninitialized_var(flags);
332#ifdef CONFIG_TRANSPARENT_HUGEPAGE
333 local_irq_save(flags);
334 compound_lock(page);
335#endif
336 return flags;
337}
338
339static inline void compound_unlock_irqrestore(struct page *page,
340 unsigned long flags)
341{
342#ifdef CONFIG_TRANSPARENT_HUGEPAGE
343 compound_unlock(page);
344 local_irq_restore(flags);
345#endif
346}
347
Christoph Lameterd85f3382007-05-06 14:49:39 -0700348static inline struct page *compound_head(struct page *page)
349{
Christoph Lameter6d777952007-05-06 14:49:40 -0700350 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700351 return page->first_page;
352 return page;
353}
354
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700355static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700357 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358}
359
360static inline void get_page(struct page *page)
361{
Andrea Arcangeli91807062011-01-13 15:46:32 -0800362 /*
363 * Getting a normal page or the head of a compound page
364 * requires to already have an elevated page->_count. Only if
365 * we're getting a tail page, the elevated page->_count is
366 * required only in the head page, so for tail pages the
367 * bugcheck only verifies that the page->_count isn't
368 * negative.
369 */
370 VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 atomic_inc(&page->_count);
Andrea Arcangeli91807062011-01-13 15:46:32 -0800372 /*
373 * Getting a tail page will elevate both the head and tail
374 * page->_count(s).
375 */
376 if (unlikely(PageTail(page))) {
377 /*
378 * This is safe only because
379 * __split_huge_page_refcount can't run under
380 * get_page().
381 */
382 VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
383 atomic_inc(&page->first_page->_count);
384 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385}
386
Christoph Lameterb49af682007-05-06 14:49:41 -0700387static inline struct page *virt_to_head_page(const void *x)
388{
389 struct page *page = virt_to_page(x);
390 return compound_head(page);
391}
392
Nick Piggin7835e982006-03-22 00:08:40 -0800393/*
394 * Setup the page count before being freed into the page allocator for
395 * the first time (boot or memory hotplug)
396 */
397static inline void init_page_count(struct page *page)
398{
399 atomic_set(&page->_count, 1);
400}
401
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800402/*
403 * PageBuddy() indicate that the page is free and in the buddy system
404 * (see mm/page_alloc.c).
405 */
406static inline int PageBuddy(struct page *page)
407{
408 return atomic_read(&page->_mapcount) == -2;
409}
410
411static inline void __SetPageBuddy(struct page *page)
412{
413 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
414 atomic_set(&page->_mapcount, -2);
415}
416
417static inline void __ClearPageBuddy(struct page *page)
418{
419 VM_BUG_ON(!PageBuddy(page));
420 atomic_set(&page->_mapcount, -1);
421}
422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700424void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800426void split_page(struct page *page, unsigned int order);
Mel Gorman748446b2010-05-24 14:32:27 -0700427int split_free_page(struct page *page);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800428
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800430 * Compound pages have a destructor function. Provide a
431 * prototype for that function and accessor functions.
432 * These are _only_ valid on the head of a PG_compound page.
433 */
434typedef void compound_page_dtor(struct page *);
435
436static inline void set_compound_page_dtor(struct page *page,
437 compound_page_dtor *dtor)
438{
439 page[1].lru.next = (void *)dtor;
440}
441
442static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
443{
444 return (compound_page_dtor *)page[1].lru.next;
445}
446
Christoph Lameterd85f3382007-05-06 14:49:39 -0700447static inline int compound_order(struct page *page)
448{
Christoph Lameter6d777952007-05-06 14:49:40 -0700449 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700450 return 0;
451 return (unsigned long)page[1].lru.prev;
452}
453
Andrea Arcangeli37c2ac72011-01-13 15:47:16 -0800454static inline int compound_trans_order(struct page *page)
455{
456 int order;
457 unsigned long flags;
458
459 if (!PageHead(page))
460 return 0;
461
462 flags = compound_lock_irqsave(page);
463 order = compound_order(page);
464 compound_unlock_irqrestore(page, flags);
465 return order;
466}
467
Christoph Lameterd85f3382007-05-06 14:49:39 -0700468static inline void set_compound_order(struct page *page, unsigned long order)
469{
470 page[1].lru.prev = (void *)order;
471}
472
Michal Simek3dece372011-01-21 08:49:56 +0100473#ifdef CONFIG_MMU
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800474/*
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800475 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
476 * servicing faults for write access. In the normal case, do always want
477 * pte_mkwrite. But get_user_pages can cause write faults for mappings
478 * that do not have writing enabled, when used by access_process_vm.
479 */
480static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
481{
482 if (likely(vma->vm_flags & VM_WRITE))
483 pte = pte_mkwrite(pte);
484 return pte;
485}
Michal Simek3dece372011-01-21 08:49:56 +0100486#endif
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800487
488/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 * Multiple processes may "see" the same page. E.g. for untouched
490 * mappings of /dev/null, all processes see the same page full of
491 * zeroes, and text pages of executables and shared libraries have
492 * only one copy in memory, at most, normally.
493 *
494 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700495 * page_count() == 0 means the page is free. page->lru is then used for
496 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700497 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700499 * Pages are allocated by the slab allocator in order to provide memory
500 * to kmalloc and kmem_cache_alloc. In this case, the management of the
501 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
502 * unless a particular usage is carefully commented. (the responsibility of
503 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700505 * A page may be used by anyone else who does a __get_free_page().
506 * In this case, page_count still tracks the references, and should only
507 * be used through the normal accessor functions. The top bits of page->flags
508 * and page->virtual store page management information, but all other fields
509 * are unused and could be used privately, carefully. The management of this
510 * page is the responsibility of the one who allocated it, and those who have
511 * subsequently been given references to it.
512 *
513 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 * managed by the Linux memory manager: I/O, buffers, swapping etc.
515 * The following discussion applies only to them.
516 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700517 * A pagecache page contains an opaque `private' member, which belongs to the
518 * page's address_space. Usually, this is the address of a circular list of
519 * the page's disk buffers. PG_private must be set to tell the VM to call
520 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700522 * A page may belong to an inode's memory mapping. In this case, page->mapping
523 * is the pointer to the inode, and page->index is the file offset of the page,
524 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700526 * If pagecache pages are not associated with an inode, they are said to be
527 * anonymous pages. These may become associated with the swapcache, and in that
528 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700530 * In either case (swapcache or inode backed), the pagecache itself holds one
531 * reference to the page. Setting PG_private should also increment the
532 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700534 * The pagecache pages are stored in a per-mapping radix tree, which is
535 * rooted at mapping->page_tree, and indexed by offset.
536 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
537 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700539 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 * - inode pages may need to be read from disk,
541 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700542 * to be written back to the inode on disk,
543 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
544 * modified may need to be swapped out to swap space and (later) to be read
545 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 */
547
548/*
549 * The zone field is never updated after free_area_init_core()
550 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700552
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700553
554/*
555 * page->flags layout:
556 *
557 * There are three possibilities for how page->flags get
558 * laid out. The first is for the normal case, without
559 * sparsemem. The second is for sparsemem when there is
560 * plenty of space for node and section. The last is when
561 * we have run out of space and have to fall back to an
562 * alternate (slower) way of determining the node.
563 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700564 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
565 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
566 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700567 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700568#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700569#define SECTIONS_WIDTH SECTIONS_SHIFT
570#else
571#define SECTIONS_WIDTH 0
572#endif
573
574#define ZONES_WIDTH ZONES_SHIFT
575
Christoph Lameter9223b412008-04-28 02:12:48 -0700576#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700577#define NODES_WIDTH NODES_SHIFT
578#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700579#ifdef CONFIG_SPARSEMEM_VMEMMAP
580#error "Vmemmap: No space for nodes field in page flags"
581#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700582#define NODES_WIDTH 0
583#endif
584
585/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100586#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700587#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
588#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
589
590/*
591 * We are going to use the flags for the page to node mapping if its in
592 * there. This includes the case where there is no node, so it is implicit.
593 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800594#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
595#define NODE_NOT_IN_PAGE_FLAGS
596#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700597
598#ifndef PFN_SECTION_SHIFT
599#define PFN_SECTION_SHIFT 0
600#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700601
602/*
603 * Define the bit shifts to access each section. For non-existant
604 * sections we define the shift as 0; that plus a 0 mask ensures
605 * the compiler will optimise away reference to them.
606 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700607#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
608#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
609#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700610
Will Deaconbce54bb2010-10-26 14:21:37 -0700611/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
612#ifdef NODE_NOT_IN_PAGE_FLAGS
Christoph Lameter89689ae2006-12-06 20:31:45 -0800613#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800614#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
615 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700616#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800617#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800618#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
619 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700620#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800621
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800622#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700623
Christoph Lameter9223b412008-04-28 02:12:48 -0700624#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
625#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700626#endif
627
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700628#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
629#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
630#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800631#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700632
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700633static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634{
Dave Hansen348f8b62005-06-23 00:07:40 -0700635 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637
Christoph Lameter89689ae2006-12-06 20:31:45 -0800638/*
639 * The identification function is only used by the buddy allocator for
640 * determining if two pages could be buddies. We are not really
641 * identifying a zone since we could be using a the section number
642 * id if we have not node id available in page flags.
643 * We guarantee only that it will return the same value for two
644 * combinable pages in a zone.
645 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700646static inline int page_zone_id(struct page *page)
647{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800648 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649}
650
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800651static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700652{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700653#ifdef CONFIG_NUMA
654 return zone->node;
655#else
656 return 0;
657#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700658}
659
Christoph Lameter89689ae2006-12-06 20:31:45 -0800660#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800661extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800662#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800663static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700664{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800665 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700666}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800667#endif
668
669static inline struct zone *page_zone(struct page *page)
670{
671 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
672}
673
Christoph Lameter308c05e2008-04-28 02:12:43 -0700674#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700675static inline unsigned long page_to_section(struct page *page)
676{
677 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
678}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700679#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700680
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700681static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682{
Dave Hansen348f8b62005-06-23 00:07:40 -0700683 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
684 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
685}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700686
Dave Hansen348f8b62005-06-23 00:07:40 -0700687static inline void set_page_node(struct page *page, unsigned long node)
688{
689 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
690 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
691}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800692
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700693static inline void set_page_section(struct page *page, unsigned long section)
694{
695 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
696 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
697}
Dave Hansen348f8b62005-06-23 00:07:40 -0700698
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700699static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700700 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700701{
702 set_page_zone(page, zone);
703 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700704 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705}
706
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700707/*
708 * Some inline functions in vmstat.h depend on page_zone()
709 */
710#include <linux/vmstat.h>
711
Ingo Molnar652050a2006-01-14 13:21:30 -0800712static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713{
Chris Metcalfc6f6b592010-05-24 14:32:53 -0700714 return __va(PFN_PHYS(page_to_pfn(page)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715}
716
717#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
718#define HASHED_PAGE_VIRTUAL
719#endif
720
721#if defined(WANT_PAGE_VIRTUAL)
722#define page_address(page) ((page)->virtual)
723#define set_page_address(page, address) \
724 do { \
725 (page)->virtual = (address); \
726 } while(0)
727#define page_address_init() do { } while(0)
728#endif
729
730#if defined(HASHED_PAGE_VIRTUAL)
731void *page_address(struct page *page);
732void set_page_address(struct page *page, void *virtual);
733void page_address_init(void);
734#endif
735
736#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
737#define page_address(page) lowmem_page_address(page)
738#define set_page_address(page, address) do { } while(0)
739#define page_address_init() do { } while(0)
740#endif
741
742/*
743 * On an anonymous page mapped into a user virtual memory area,
744 * page->mapping points to its anon_vma, not to a struct address_space;
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800745 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
746 *
747 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
748 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
749 * and then page->mapping points, not to an anon_vma, but to a private
750 * structure which KSM associates with that merged page. See ksm.h.
751 *
752 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 *
754 * Please note that, confusingly, "page_mapping" refers to the inode
755 * address_space which maps the page from disk; whereas "page_mapped"
756 * refers to user virtual address space into which the page is mapped.
757 */
758#define PAGE_MAPPING_ANON 1
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800759#define PAGE_MAPPING_KSM 2
760#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
762extern struct address_space swapper_space;
763static inline struct address_space *page_mapping(struct page *page)
764{
765 struct address_space *mapping = page->mapping;
766
Christoph Lameterb5fab142007-07-17 04:03:33 -0700767 VM_BUG_ON(PageSlab(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 if (unlikely(PageSwapCache(page)))
769 mapping = &swapper_space;
Steven Rostedte20e8772011-01-13 15:46:17 -0800770 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 mapping = NULL;
772 return mapping;
773}
774
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800775/* Neutral page->mapping pointer to address_space or anon_vma or other */
776static inline void *page_rmapping(struct page *page)
777{
778 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
779}
780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781static inline int PageAnon(struct page *page)
782{
783 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
784}
785
786/*
787 * Return the pagecache index of the passed page. Regular pagecache pages
788 * use ->index whereas swapcache pages use ->private
789 */
790static inline pgoff_t page_index(struct page *page)
791{
792 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700793 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 return page->index;
795}
796
797/*
798 * The atomic page->_mapcount, like _count, starts from -1:
799 * so that transitions both from it and to it can be tracked,
800 * using atomic_inc_and_test and atomic_add_negative(-1).
801 */
802static inline void reset_page_mapcount(struct page *page)
803{
804 atomic_set(&(page)->_mapcount, -1);
805}
806
807static inline int page_mapcount(struct page *page)
808{
809 return atomic_read(&(page)->_mapcount) + 1;
810}
811
812/*
813 * Return true if this page is mapped into pagetables.
814 */
815static inline int page_mapped(struct page *page)
816{
817 return atomic_read(&(page)->_mapcount) >= 0;
818}
819
820/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 * Different kinds of faults, as returned by handle_mm_fault().
822 * Used to decide whether a process gets delivered SIGBUS or
823 * just gets major/minor fault counters bumped up.
824 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700825
Nick Piggin83c54072007-07-19 01:47:05 -0700826#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700827
Nick Piggin83c54072007-07-19 01:47:05 -0700828#define VM_FAULT_OOM 0x0001
829#define VM_FAULT_SIGBUS 0x0002
830#define VM_FAULT_MAJOR 0x0004
831#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleenaa50d3a2010-10-06 21:45:00 +0200832#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
833#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000834
Nick Piggin83c54072007-07-19 01:47:05 -0700835#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
836#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Michel Lespinassed065bd82010-10-26 14:21:57 -0700837#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
Andi Kleenaa50d3a2010-10-06 21:45:00 +0200839#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
840
841#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
842 VM_FAULT_HWPOISON_LARGE)
843
844/* Encode hstate index for a hwpoisoned large page */
845#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
846#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
Nick Piggind0217ac2007-07-19 01:47:03 -0700847
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800848/*
849 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
850 */
851extern void pagefault_out_of_memory(void);
852
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
854
855extern void show_free_areas(void);
856
Hugh Dickins3f96b792009-09-21 17:03:37 -0700857int shmem_lock(struct file *file, int lock, struct user_struct *user);
Sergei Trofimovich168f5ac2009-06-16 15:33:02 -0700858struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859int shmem_zero_setup(struct vm_area_struct *);
860
David Howellsb0e15192006-01-06 00:11:42 -0800861#ifndef CONFIG_MMU
862extern unsigned long shmem_get_unmapped_area(struct file *file,
863 unsigned long addr,
864 unsigned long len,
865 unsigned long pgoff,
866 unsigned long flags);
867#endif
868
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400869extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870extern int user_shm_lock(size_t, struct user_struct *);
871extern void user_shm_unlock(size_t, struct user_struct *);
872
873/*
874 * Parameter block passed down to zap_pte_range in exceptional cases.
875 */
876struct zap_details {
877 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
878 struct address_space *check_mapping; /* Check page->mapping if set */
879 pgoff_t first_index; /* Lowest page->index to unmap */
880 pgoff_t last_index; /* Highest page->index to unmap */
881 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 unsigned long truncate_count; /* Compare vm_truncate_count */
883};
884
Nick Piggin7e675132008-04-28 02:13:00 -0700885struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
886 pte_t pte);
887
Jack Steinerc627f9c2008-07-29 22:33:53 -0700888int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
889 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700890unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700892unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 struct vm_area_struct *start_vma, unsigned long start_addr,
894 unsigned long end_addr, unsigned long *nr_accounted,
895 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800896
897/**
898 * mm_walk - callbacks for walk_page_range
899 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
900 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
901 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
902 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
903 * @pte_hole: if set, called for each hole at all levels
Naoya Horiguchi5dc37642009-12-14 18:00:01 -0800904 * @hugetlb_entry: if set, called for each hugetlb entry
Matt Mackalle6473092008-02-04 22:29:01 -0800905 *
906 * (see walk_page_range for more details)
907 */
908struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700909 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
910 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
911 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
912 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
913 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
Naoya Horiguchi116354d2010-04-06 14:35:04 -0700914 int (*hugetlb_entry)(pte_t *, unsigned long,
915 unsigned long, unsigned long, struct mm_walk *);
Dave Hansen21650092008-06-12 15:21:47 -0700916 struct mm_struct *mm;
917 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800918};
919
Dave Hansen21650092008-06-12 15:21:47 -0700920int walk_page_range(unsigned long addr, unsigned long end,
921 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700922void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700923 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
925 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926void unmap_mapping_range(struct address_space *mapping,
927 loff_t const holebegin, loff_t const holelen, int even_cows);
Johannes Weiner3b6748e2009-06-16 15:32:35 -0700928int follow_pfn(struct vm_area_struct *vma, unsigned long address,
929 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800930int follow_phys(struct vm_area_struct *vma, unsigned long address,
931 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700932int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
933 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934
935static inline void unmap_shared_mapping_range(struct address_space *mapping,
936 loff_t const holebegin, loff_t const holelen)
937{
938 unmap_mapping_range(mapping, holebegin, holelen, 0);
939}
940
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000941extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
Christoph Hellwig2c27c652010-06-04 11:30:04 +0200942extern void truncate_setsize(struct inode *inode, loff_t newsize);
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000943extern int vmtruncate(struct inode *inode, loff_t offset);
944extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000945
Nick Piggin750b4982009-09-16 11:50:12 +0200946int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +0200947int generic_error_remove_page(struct address_space *mapping, struct page *page);
Nick Piggin750b4982009-09-16 11:50:12 +0200948
Wu Fengguang83f78662009-09-16 11:50:13 +0200949int invalidate_inode_page(struct page *page);
950
David Howells7ee1dd32006-01-06 00:11:44 -0800951#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700952extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700953 unsigned long address, unsigned int flags);
David Howells7ee1dd32006-01-06 00:11:44 -0800954#else
955static inline int handle_mm_fault(struct mm_struct *mm,
956 struct vm_area_struct *vma, unsigned long address,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700957 unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -0800958{
959 /* should never happen if there's no MMU */
960 BUG();
961 return VM_FAULT_SIGBUS;
962}
963#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000964
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965extern int make_pages_present(unsigned long addr, unsigned long end);
966extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967
Nick Piggind2bf6be2009-06-16 15:31:39 -0700968int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
Peter Zijlstra9d737772009-06-25 11:58:55 +0200969 unsigned long start, int nr_pages, int write, int force,
Nick Piggind2bf6be2009-06-16 15:31:39 -0700970 struct page **pages, struct vm_area_struct **vmas);
971int get_user_pages_fast(unsigned long start, int nr_pages, int write,
972 struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -0700973struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974
David Howellscf9a2ae2006-08-29 19:05:54 +0100975extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
976extern void do_invalidatepage(struct page *page, unsigned long offset);
977
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800979int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980int redirty_page_for_writepage(struct writeback_control *wbc,
981 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -0700982void account_page_dirtied(struct page *page, struct address_space *mapping);
Michael Rubinf629d1c2010-10-26 14:21:33 -0700983void account_page_writeback(struct page *page);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800984int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985int set_page_dirty_lock(struct page *page);
986int clear_page_dirty_for_io(struct page *page);
987
Stefan Bader39aa3cb2010-08-31 15:52:27 +0200988/* Is the vma a continuation of the stack vma above it? */
989static inline int vma_stack_continue(struct vm_area_struct *vma, unsigned long addr)
990{
991 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
992}
993
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700994extern unsigned long move_page_tables(struct vm_area_struct *vma,
995 unsigned long old_addr, struct vm_area_struct *new_vma,
996 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997extern unsigned long do_mremap(unsigned long addr,
998 unsigned long old_len, unsigned long new_len,
999 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001000extern int mprotect_fixup(struct vm_area_struct *vma,
1001 struct vm_area_struct **pprev, unsigned long start,
1002 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Nick Piggin21cc1992008-07-25 19:45:22 -07001004/*
Peter Zijlstra465a4542009-06-15 12:31:37 +02001005 * doesn't attempt to fault and will return short.
1006 */
1007int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1008 struct page **pages);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001009/*
1010 * per-process(per-mm_struct) statistics.
1011 */
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -08001012#if defined(SPLIT_RSS_COUNTING)
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001013/*
1014 * The mm counters are not protected by its page_table_lock,
1015 * so must be incremented atomically.
1016 */
1017static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
1018{
1019 atomic_long_set(&mm->rss_stat.count[member], value);
1020}
1021
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -08001022unsigned long get_mm_counter(struct mm_struct *mm, int member);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001023
1024static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1025{
1026 atomic_long_add(value, &mm->rss_stat.count[member]);
1027}
1028
1029static inline void inc_mm_counter(struct mm_struct *mm, int member)
1030{
1031 atomic_long_inc(&mm->rss_stat.count[member]);
1032}
1033
1034static inline void dec_mm_counter(struct mm_struct *mm, int member)
1035{
1036 atomic_long_dec(&mm->rss_stat.count[member]);
1037}
1038
1039#else /* !USE_SPLIT_PTLOCKS */
1040/*
1041 * The mm counters are protected by its page_table_lock,
1042 * so can be incremented directly.
1043 */
1044static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
1045{
1046 mm->rss_stat.count[member] = value;
1047}
1048
1049static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1050{
1051 return mm->rss_stat.count[member];
1052}
1053
1054static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1055{
1056 mm->rss_stat.count[member] += value;
1057}
1058
1059static inline void inc_mm_counter(struct mm_struct *mm, int member)
1060{
1061 mm->rss_stat.count[member]++;
1062}
1063
1064static inline void dec_mm_counter(struct mm_struct *mm, int member)
1065{
1066 mm->rss_stat.count[member]--;
1067}
1068
1069#endif /* !USE_SPLIT_PTLOCKS */
1070
1071static inline unsigned long get_mm_rss(struct mm_struct *mm)
1072{
1073 return get_mm_counter(mm, MM_FILEPAGES) +
1074 get_mm_counter(mm, MM_ANONPAGES);
1075}
1076
1077static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1078{
1079 return max(mm->hiwater_rss, get_mm_rss(mm));
1080}
1081
1082static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1083{
1084 return max(mm->hiwater_vm, mm->total_vm);
1085}
1086
1087static inline void update_hiwater_rss(struct mm_struct *mm)
1088{
1089 unsigned long _rss = get_mm_rss(mm);
1090
1091 if ((mm)->hiwater_rss < _rss)
1092 (mm)->hiwater_rss = _rss;
1093}
1094
1095static inline void update_hiwater_vm(struct mm_struct *mm)
1096{
1097 if (mm->hiwater_vm < mm->total_vm)
1098 mm->hiwater_vm = mm->total_vm;
1099}
1100
1101static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1102 struct mm_struct *mm)
1103{
1104 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1105
1106 if (*maxrss < hiwater_rss)
1107 *maxrss = hiwater_rss;
1108}
1109
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001110#if defined(SPLIT_RSS_COUNTING)
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -08001111void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001112#else
1113static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
1114{
1115}
1116#endif
Peter Zijlstra465a4542009-06-15 12:31:37 +02001117
1118/*
Rusty Russell8e1f9362007-07-17 04:03:17 -07001119 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 *
Rusty Russell8e1f9362007-07-17 04:03:17 -07001121 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
1122 * look through the least-recently-used 'nr_to_scan' entries and
1123 * attempt to free them up. It should return the number of objects
1124 * which remain in the cache. If it returns -1, it means it cannot do
1125 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 *
Rusty Russell8e1f9362007-07-17 04:03:17 -07001127 * The 'gfpmask' refers to the allocation we are currently trying to
1128 * fulfil.
1129 *
1130 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
1131 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 */
Rusty Russell8e1f9362007-07-17 04:03:17 -07001133struct shrinker {
Dave Chinner7f8275d2010-07-19 14:56:17 +10001134 int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask);
Rusty Russell8e1f9362007-07-17 04:03:17 -07001135 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
Rusty Russell8e1f9362007-07-17 04:03:17 -07001137 /* These are for internal use */
1138 struct list_head list;
1139 long nr; /* objs pending delete */
1140};
1141#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1142extern void register_shrinker(struct shrinker *);
1143extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
Alexey Dobriyan4e950f62007-07-30 02:36:13 +04001145int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -07001146
Namhyung Kim25ca1d62010-10-26 14:21:59 -07001147extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1148 spinlock_t **ptl);
1149static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1150 spinlock_t **ptl)
1151{
1152 pte_t *ptep;
1153 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1154 return ptep;
1155}
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -08001156
Nick Piggin5f22df02007-05-06 14:49:02 -07001157#ifdef __PAGETABLE_PUD_FOLDED
1158static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1159 unsigned long address)
1160{
1161 return 0;
1162}
1163#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001164int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001165#endif
1166
1167#ifdef __PAGETABLE_PMD_FOLDED
1168static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1169 unsigned long address)
1170{
1171 return 0;
1172}
1173#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001174int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001175#endif
1176
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001177int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1178 pmd_t *pmd, unsigned long address);
Hugh Dickins1bb36302005-10-29 18:16:22 -07001179int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1180
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181/*
1182 * The following ifdef needed to get the 4level-fixup.h header to work.
1183 * Remove it when 4level-fixup.h has been removed.
1184 */
Hugh Dickins1bb36302005-10-29 18:16:22 -07001185#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1187{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001188 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1189 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190}
1191
1192static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1193{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001194 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1195 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196}
Hugh Dickins1bb36302005-10-29 18:16:22 -07001197#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1198
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001199#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001200/*
1201 * We tuck a spinlock to guard each pagetable page into its struct page,
1202 * at page->private, with BUILD_BUG_ON to make sure that this will not
1203 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1204 * When freeing, reset page->mapping so free_pages_check won't complain.
1205 */
Andrew Morton349aef02006-01-08 01:04:36 -08001206#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001207#define pte_lock_init(_page) do { \
1208 spin_lock_init(__pte_lockptr(_page)); \
1209} while (0)
1210#define pte_lock_deinit(page) ((page)->mapping = NULL)
1211#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001212#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001213/*
1214 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1215 */
1216#define pte_lock_init(page) do {} while (0)
1217#define pte_lock_deinit(page) do {} while (0)
1218#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001219#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001220
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001221static inline void pgtable_page_ctor(struct page *page)
1222{
1223 pte_lock_init(page);
1224 inc_zone_page_state(page, NR_PAGETABLE);
1225}
1226
1227static inline void pgtable_page_dtor(struct page *page)
1228{
1229 pte_lock_deinit(page);
1230 dec_zone_page_state(page, NR_PAGETABLE);
1231}
1232
Hugh Dickinsc74df322005-10-29 18:16:23 -07001233#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1234({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001235 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001236 pte_t *__pte = pte_offset_map(pmd, address); \
1237 *(ptlp) = __ptl; \
1238 spin_lock(__ptl); \
1239 __pte; \
1240})
1241
1242#define pte_unmap_unlock(pte, ptl) do { \
1243 spin_unlock(ptl); \
1244 pte_unmap(pte); \
1245} while (0)
1246
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001247#define pte_alloc_map(mm, vma, pmd, address) \
1248 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1249 pmd, address))? \
1250 NULL: pte_offset_map(pmd, address))
Hugh Dickins1bb36302005-10-29 18:16:22 -07001251
Hugh Dickinsc74df322005-10-29 18:16:23 -07001252#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001253 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1254 pmd, address))? \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001255 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1256
Hugh Dickins1bb36302005-10-29 18:16:22 -07001257#define pte_alloc_kernel(pmd, address) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001258 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
Hugh Dickins1bb36302005-10-29 18:16:22 -07001259 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260
1261extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001262extern void free_area_init_node(int nid, unsigned long * zones_size,
1263 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -07001264#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1265/*
1266 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1267 * zones, allocate the backing mem_map and account for memory holes in a more
1268 * architecture independent manner. This is a substitute for creating the
1269 * zone_sizes[] and zholes_size[] arrays and passing them to
1270 * free_area_init_node()
1271 *
1272 * An architecture is expected to register range of page frames backed by
1273 * physical memory with add_active_range() before calling
1274 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1275 * usage, an architecture is expected to do something like
1276 *
1277 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1278 * max_highmem_pfn};
1279 * for_each_valid_physical_page_range()
1280 * add_active_range(node_id, start_pfn, end_pfn)
1281 * free_area_init_nodes(max_zone_pfns);
1282 *
1283 * If the architecture guarantees that there are no holes in the ranges
1284 * registered with add_active_range(), free_bootmem_active_regions()
1285 * will call free_bootmem_node() for each registered physical page range.
1286 * Similarly sparse_memory_present_with_active_regions() calls
1287 * memory_present() for each range when SPARSEMEM is enabled.
1288 *
1289 * See mm/page_alloc.c for more information on each function exposed by
1290 * CONFIG_ARCH_POPULATES_NODE_MAP
1291 */
1292extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1293extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1294 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001295extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1296 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001297extern void remove_all_active_ranges(void);
Yinghai Lu32996252009-12-15 17:59:02 -08001298void sort_node_map(void);
1299unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1300 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001301extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1302 unsigned long end_pfn);
1303extern void get_pfn_range_for_nid(unsigned int nid,
1304 unsigned long *start_pfn, unsigned long *end_pfn);
1305extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001306extern void free_bootmem_with_active_regions(int nid,
1307 unsigned long max_low_pfn);
Yinghai Lu08677212010-02-10 01:20:20 -08001308int add_from_early_node_map(struct range *range, int az,
1309 int nr_range, int nid);
Yinghai Luedbe7d22010-08-25 13:39:16 -07001310u64 __init find_memory_core_early(int nid, u64 size, u64 align,
1311 u64 goal, u64 limit);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001312typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001313extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001314extern void sparse_memory_present_with_active_regions(int nid);
Mel Gormanc7132162006-09-27 01:49:43 -07001315#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001316
1317#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1318 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1319static inline int __early_pfn_to_nid(unsigned long pfn)
1320{
1321 return 0;
1322}
1323#else
1324/* please see mm/page_alloc.c */
1325extern int __meminit early_pfn_to_nid(unsigned long pfn);
1326#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1327/* there is a per-arch backend function. */
1328extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1329#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1330#endif
1331
Mel Gorman0e0b8642006-09-27 01:49:56 -07001332extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa02007-01-10 23:15:30 -08001333extern void memmap_init_zone(unsigned long, int, unsigned long,
1334 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07001335extern void setup_per_zone_wmarks(void);
Minchan Kim96cb4df2009-06-16 15:32:49 -07001336extern void calculate_zone_inactive_ratio(struct zone *zone);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001338extern void __init mmap_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339extern void show_mem(void);
1340extern void si_meminfo(struct sysinfo * val);
1341extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001342extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001344extern void setup_per_cpu_pageset(void);
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001345
Shaohua Li112067f2009-09-21 17:01:16 -07001346extern void zone_pcp_update(struct zone *zone);
1347
David Howells8feae132009-01-08 12:04:47 +00001348/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001349extern atomic_long_t mmap_pages_allocated;
David Howells7e660872010-01-15 17:01:39 -08001350extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
David Howells8feae132009-01-08 12:04:47 +00001351
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352/* prio_tree.c */
1353void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1354void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1355void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1356struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1357 struct prio_tree_iter *iter);
1358
1359#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1360 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1361 (vma = vma_prio_tree_next(vma, iter)); )
1362
1363static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1364 struct list_head *list)
1365{
1366 vma->shared.vm_set.parent = NULL;
1367 list_add_tail(&vma->shared.vm_set.list, list);
1368}
1369
1370/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001371extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Rik van Riel5beb4932010-03-05 13:42:07 -08001372extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1374extern struct vm_area_struct *vma_merge(struct mm_struct *,
1375 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1376 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1377 struct mempolicy *);
1378extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1379extern int split_vma(struct mm_struct *,
1380 struct vm_area_struct *, unsigned long addr, int new_below);
1381extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1382extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1383 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001384extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1386 unsigned long addr, unsigned long len, pgoff_t pgoff);
1387extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001388
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001389extern int mm_take_all_locks(struct mm_struct *mm);
1390extern void mm_drop_all_locks(struct mm_struct *mm);
1391
Matt Helsley925d1c42008-04-29 01:01:36 -07001392#ifdef CONFIG_PROC_FS
1393/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1394extern void added_exe_file_vma(struct mm_struct *mm);
1395extern void removed_exe_file_vma(struct mm_struct *mm);
1396#else
1397static inline void added_exe_file_vma(struct mm_struct *mm)
1398{}
1399
1400static inline void removed_exe_file_vma(struct mm_struct *mm)
1401{}
1402#endif /* CONFIG_PROC_FS */
1403
akpm@osdl.org119f6572005-05-01 08:58:35 -07001404extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001405extern int install_special_mapping(struct mm_struct *mm,
1406 unsigned long addr, unsigned long len,
1407 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408
1409extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1410
1411extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1412 unsigned long len, unsigned long prot,
1413 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001414extern unsigned long mmap_region(struct file *file, unsigned long addr,
1415 unsigned long len, unsigned long flags,
Mel Gorman5a6fe122009-02-10 14:02:27 +00001416 unsigned int vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417
1418static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1419 unsigned long len, unsigned long prot,
1420 unsigned long flag, unsigned long offset)
1421{
1422 unsigned long ret = -EINVAL;
1423 if ((offset + PAGE_ALIGN(len)) < offset)
1424 goto out;
1425 if (!(offset & ~PAGE_MASK))
1426 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1427out:
1428 return ret;
1429}
1430
1431extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1432
1433extern unsigned long do_brk(unsigned long, unsigned long);
1434
1435/* filemap.c */
1436extern unsigned long page_unuse(struct page *);
1437extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001438extern void truncate_inode_pages_range(struct address_space *,
1439 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440
1441/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001442extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443
1444/* mm/page-writeback.c */
1445int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001446void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447
1448/* readahead.c */
1449#define VM_MAX_READAHEAD 128 /* kbytes */
1450#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001453 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001454
1455void page_cache_sync_readahead(struct address_space *mapping,
1456 struct file_ra_state *ra,
1457 struct file *filp,
1458 pgoff_t offset,
1459 unsigned long size);
1460
1461void page_cache_async_readahead(struct address_space *mapping,
1462 struct file_ra_state *ra,
1463 struct file *filp,
1464 struct page *pg,
1465 pgoff_t offset,
1466 unsigned long size);
1467
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468unsigned long max_sane_readahead(unsigned long nr);
Wu Fengguangd30a1102009-06-16 15:31:30 -07001469unsigned long ra_submit(struct file_ra_state *ra,
1470 struct address_space *mapping,
1471 struct file *filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
1473/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001474extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001475#if VM_GROWSUP
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001476extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001477#else
1478 #define expand_upwards(vma, address) do { } while (0)
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001479#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001480extern int expand_stack_downwards(struct vm_area_struct *vma,
1481 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482
1483/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1484extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1485extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1486 struct vm_area_struct **pprev);
1487
1488/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1489 NULL if none. Assume start_addr < end_addr. */
1490static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1491{
1492 struct vm_area_struct * vma = find_vma(mm,start_addr);
1493
1494 if (vma && end_addr <= vma->vm_start)
1495 vma = NULL;
1496 return vma;
1497}
1498
1499static inline unsigned long vma_pages(struct vm_area_struct *vma)
1500{
1501 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1502}
1503
David Howellsbad849b2010-08-26 16:00:34 +01001504#ifdef CONFIG_MMU
Hugh Dickins804af2c2006-07-26 21:39:49 +01001505pgprot_t vm_get_page_prot(unsigned long vm_flags);
David Howellsbad849b2010-08-26 16:00:34 +01001506#else
1507static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1508{
1509 return __pgprot(0);
1510}
1511#endif
1512
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001513struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001514int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1515 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001516int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001517int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1518 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001519int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1520 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521
Linus Torvalds6aab3412005-11-28 14:34:23 -08001522struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001523 unsigned int foll_flags);
1524#define FOLL_WRITE 0x01 /* check pte is writable */
1525#define FOLL_TOUCH 0x02 /* mark page accessed */
1526#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a62009-09-21 17:03:26 -07001527#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07001528#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
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 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001532typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001533 void *data);
1534extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1535 unsigned long size, pte_fn_t fn, void *data);
1536
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001538void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001540static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 unsigned long flags, struct file *file, long pages)
1542{
1543}
1544#endif /* CONFIG_PROC_FS */
1545
Ingo Molnar12d6f212008-01-30 13:33:58 +01001546#ifdef CONFIG_DEBUG_PAGEALLOC
1547extern int debug_pagealloc_enabled;
1548
1549extern void kernel_map_pages(struct page *page, int numpages, int enable);
1550
1551static inline void enable_debug_pagealloc(void)
1552{
1553 debug_pagealloc_enabled = 1;
1554}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001555#ifdef CONFIG_HIBERNATION
1556extern bool kernel_page_present(struct page *page);
1557#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001558#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001560kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001561static inline void enable_debug_pagealloc(void)
1562{
1563}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001564#ifdef CONFIG_HIBERNATION
1565static inline bool kernel_page_present(struct page *page) { return true; }
1566#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567#endif
1568
1569extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1570#ifdef __HAVE_ARCH_GATE_AREA
1571int in_gate_area_no_task(unsigned long addr);
1572int in_gate_area(struct task_struct *task, unsigned long addr);
1573#else
1574int in_gate_area_no_task(unsigned long addr);
1575#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1576#endif /* __HAVE_ARCH_GATE_AREA */
1577
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07001578int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001579 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001580unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001581 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001582
Luke Yang7a9166e2006-02-20 18:28:07 -08001583#ifndef CONFIG_MMU
1584#define randomize_va_space 0
1585#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001586extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001587#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001588
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001589const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001590void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001591
Yinghai Lu9bdac912010-02-10 01:20:22 -08001592void sparse_mem_maps_populate_node(struct page **map_map,
1593 unsigned long pnum_begin,
1594 unsigned long pnum_end,
1595 unsigned long map_count,
1596 int nodeid);
1597
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001598struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001599pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1600pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1601pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1602pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001603void *vmemmap_alloc_block(unsigned long size, int node);
Yinghai Lu9bdac912010-02-10 01:20:22 -08001604void *vmemmap_alloc_block_buf(unsigned long size, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001605void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001606int vmemmap_populate_basepages(struct page *start_page,
1607 unsigned long pages, int node);
1608int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001609void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001610
Andi Kleen6a460792009-09-16 11:50:15 +02001611
Andi Kleen82ba0112009-12-16 12:19:57 +01001612enum mf_flags {
1613 MF_COUNT_INCREASED = 1 << 0,
1614};
Andi Kleen6a460792009-09-16 11:50:15 +02001615extern void memory_failure(unsigned long pfn, int trapno);
Andi Kleen82ba0112009-12-16 12:19:57 +01001616extern int __memory_failure(unsigned long pfn, int trapno, int flags);
Wu Fengguang847ce402009-12-16 12:19:58 +01001617extern int unpoison_memory(unsigned long pfn);
Andi Kleen6a460792009-09-16 11:50:15 +02001618extern int sysctl_memory_failure_early_kill;
1619extern int sysctl_memory_failure_recovery;
Andi Kleenfacb6012009-12-16 12:20:00 +01001620extern void shake_page(struct page *p, int access);
Andi Kleen6a460792009-09-16 11:50:15 +02001621extern atomic_long_t mce_bad_pages;
Andi Kleenfacb6012009-12-16 12:20:00 +01001622extern int soft_offline_page(struct page *page, int flags);
Huang Yingbf998152010-05-31 14:28:19 +08001623#ifdef CONFIG_MEMORY_FAILURE
1624int is_hwpoison_address(unsigned long addr);
1625#else
1626static inline int is_hwpoison_address(unsigned long addr)
1627{
1628 return 0;
1629}
1630#endif
Andi Kleen6a460792009-09-16 11:50:15 +02001631
Wu Fengguang718a3822010-03-10 15:20:43 -08001632extern void dump_page(struct page *page);
1633
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08001634#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1635extern void clear_huge_page(struct page *page,
1636 unsigned long addr,
1637 unsigned int pages_per_huge_page);
1638extern void copy_user_huge_page(struct page *dst, struct page *src,
1639 unsigned long addr, struct vm_area_struct *vma,
1640 unsigned int pages_per_huge_page);
1641#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1642
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643#endif /* __KERNEL__ */
1644#endif /* _LINUX_MM_H */