blob: 963f908af9d055d09ab0d83b7d66db933f9150b8 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016
17struct mempolicy;
18struct anon_vma;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040019struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040020struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040021struct writeback_control;
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
23#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
24extern unsigned long max_mapnr;
25#endif
26
27extern unsigned long num_physpages;
Jan Beulich44813742009-09-21 17:03:05 -070028extern unsigned long totalram_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -070029extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070030extern int page_cluster;
31
32#ifdef CONFIG_SYSCTL
33extern int sysctl_legacy_va_layout;
34#else
35#define sysctl_legacy_va_layout 0
36#endif
37
38#include <asm/page.h>
39#include <asm/pgtable.h>
40#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
43
Andrea Righi27ac7922008-07-23 21:28:13 -070044/* to align the pointer to the (next) page boundary */
45#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
46
Linus Torvalds1da177e2005-04-16 15:20:36 -070047/*
48 * Linux kernel virtual memory manager primitives.
49 * The idea being to have a "virtual" mm in the same way
50 * we have a virtual fs - giving a cleaner interface to the
51 * mm details, and allowing different kinds of memory mappings
52 * (from shared memory to executable loading to arbitrary
53 * mmap() functions).
54 */
55
Christoph Lameterc43692e2006-12-06 20:32:48 -080056extern struct kmem_cache *vm_area_cachep;
57
Linus Torvalds1da177e2005-04-16 15:20:36 -070058#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +000059extern struct rb_root nommu_region_tree;
60extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
62extern unsigned int kobjsize(const void *objp);
63#endif
64
65/*
Hugh Dickins605d9282008-08-16 11:07:21 +010066 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 */
68#define VM_READ 0x00000001 /* currently active flags */
69#define VM_WRITE 0x00000002
70#define VM_EXEC 0x00000004
71#define VM_SHARED 0x00000008
72
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070073/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070074#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
75#define VM_MAYWRITE 0x00000020
76#define VM_MAYEXEC 0x00000040
77#define VM_MAYSHARE 0x00000080
78
79#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
80#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -080081#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
83
84#define VM_EXECUTABLE 0x00001000
85#define VM_LOCKED 0x00002000
86#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
87
88 /* Used by sys_madvise() */
89#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
90#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
91
92#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
93#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -080094#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -070096#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
98#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
99#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700100#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800101#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102
Nick Piggind0217ac2007-07-19 01:47:03 -0700103#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700104#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000105#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700106#define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700107#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700108
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700109/* Bits set in the VMA until the stack is in its final location */
110#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
111
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
113#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
114#endif
115
116#ifdef CONFIG_STACK_GROWSUP
117#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
118#else
119#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
120#endif
121
122#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
123#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
124#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
125#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
126#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
127
128/*
Nick Pigginb291f002008-10-18 20:26:44 -0700129 * special vmas that are non-mergable, non-mlock()able
130 */
131#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
132
133/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 * mapping from the currently active vm_flags protection bits (the
135 * low four bits) to a page protection mask..
136 */
137extern pgprot_t protection_map[16];
138
Nick Piggind0217ac2007-07-19 01:47:03 -0700139#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
140#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700141#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Nick Piggind0217ac2007-07-19 01:47:03 -0700142
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800143/*
144 * This interface is used by x86 PAT code to identify a pfn mapping that is
145 * linear over entire vma. This is to optimize PAT code that deals with
146 * marking the physical region with a particular prot. This is not for generic
147 * mm use. Note also that this check will not work if the pfn mapping is
148 * linear for a vma starting at physical address 0. In which case PAT code
149 * falls back to slow path of reserving physical range page by page.
150 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800151static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
152{
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700153 return (vma->vm_flags & VM_PFN_AT_MMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800154}
155
156static inline int is_pfn_mapping(struct vm_area_struct *vma)
157{
158 return (vma->vm_flags & VM_PFNMAP);
159}
Nick Piggind0217ac2007-07-19 01:47:03 -0700160
Nick Piggin54cb8822007-07-19 01:46:59 -0700161/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700162 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700163 * ->fault function. The vma's ->fault is responsible for returning a bitmask
164 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700165 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700166 * pgoff should be used in favour of virtual_address, if possible. If pgoff
167 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
168 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700169 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700170struct vm_fault {
171 unsigned int flags; /* FAULT_FLAG_xxx flags */
172 pgoff_t pgoff; /* Logical page offset based on vma */
173 void __user *virtual_address; /* Faulting virtual address */
174
175 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700176 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700177 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700178 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700179 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700180};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181
182/*
183 * These are the virtual MM functions - opening of an area, closing and
184 * unmapping it (needed to keep files on disk up-to-date etc), pointer
185 * to the functions called when a no-page or a wp-page exception occurs.
186 */
187struct vm_operations_struct {
188 void (*open)(struct vm_area_struct * area);
189 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700190 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700191
192 /* notification that a previously read-only page is about to become
193 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700194 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700195
196 /* called by access_process_vm when get_user_pages() fails, typically
197 * for use by special VMAs that can switch between memory and hardware
198 */
199 int (*access)(struct vm_area_struct *vma, unsigned long addr,
200 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700202 /*
203 * set_policy() op must add a reference to any non-NULL @new mempolicy
204 * to hold the policy upon return. Caller should pass NULL @new to
205 * remove a policy and fall back to surrounding context--i.e. do not
206 * install a MPOL_DEFAULT policy, nor the task or system default
207 * mempolicy.
208 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700210
211 /*
212 * get_policy() op must add reference [mpol_get()] to any policy at
213 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
214 * in mm/mempolicy.c will do this automatically.
215 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
216 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
217 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
218 * must return NULL--i.e., do not "fallback" to task or system default
219 * policy.
220 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
222 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700223 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
224 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225#endif
226};
227
228struct mmu_gather;
229struct inode;
230
Andrew Morton349aef02006-01-08 01:04:36 -0800231#define page_private(page) ((page)->private)
232#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700233
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234/*
235 * FIXME: take this include out, include page-flags.h in
236 * files which need it (119 of them)
237 */
238#include <linux/page-flags.h>
239
240/*
241 * Methods to modify the page usage count.
242 *
243 * What counts for a page usage:
244 * - cache mapping (page->mapping)
245 * - private data (page->private)
246 * - page mapped in a task's page tables, each mapping
247 * is counted separately
248 *
249 * Also, many kernel routines increase the page count before a critical
250 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 */
252
253/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700254 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800256static inline int put_page_testzero(struct page *page)
257{
Nick Piggin725d7042006-09-25 23:30:55 -0700258 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800259 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800260}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
262/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800263 * Try to grab a ref unless the page has a refcount of zero, return false if
264 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800266static inline int get_page_unless_zero(struct page *page)
267{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800268 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800269}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800271extern int page_is_ram(unsigned long pfn);
272
Christoph Lameter48667e72008-02-04 22:28:31 -0800273/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800274struct page *vmalloc_to_page(const void *addr);
275unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800276
Paul Mundt0738c4b2008-03-12 16:51:31 +0900277/*
278 * Determine if an address is within the vmalloc range
279 *
280 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
281 * is no special casing required.
282 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800283static inline int is_vmalloc_addr(const void *x)
284{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900285#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800286 unsigned long addr = (unsigned long)x;
287
288 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900289#else
290 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800291#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900292}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700293#ifdef CONFIG_MMU
294extern int is_vmalloc_or_module_addr(const void *x);
295#else
David Howells934831d2009-09-24 12:33:32 +0100296static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700297{
298 return 0;
299}
300#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800301
Christoph Lameterd85f3382007-05-06 14:49:39 -0700302static inline struct page *compound_head(struct page *page)
303{
Christoph Lameter6d777952007-05-06 14:49:40 -0700304 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700305 return page->first_page;
306 return page;
307}
308
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700309static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700311 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312}
313
314static inline void get_page(struct page *page)
315{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700316 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700317 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 atomic_inc(&page->_count);
319}
320
Christoph Lameterb49af682007-05-06 14:49:41 -0700321static inline struct page *virt_to_head_page(const void *x)
322{
323 struct page *page = virt_to_page(x);
324 return compound_head(page);
325}
326
Nick Piggin7835e982006-03-22 00:08:40 -0800327/*
328 * Setup the page count before being freed into the page allocator for
329 * the first time (boot or memory hotplug)
330 */
331static inline void init_page_count(struct page *page)
332{
333 atomic_set(&page->_count, 1);
334}
335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700337void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800339void split_page(struct page *page, unsigned int order);
Mel Gorman748446b2010-05-24 14:32:27 -0700340int split_free_page(struct page *page);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800341
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800343 * Compound pages have a destructor function. Provide a
344 * prototype for that function and accessor functions.
345 * These are _only_ valid on the head of a PG_compound page.
346 */
347typedef void compound_page_dtor(struct page *);
348
349static inline void set_compound_page_dtor(struct page *page,
350 compound_page_dtor *dtor)
351{
352 page[1].lru.next = (void *)dtor;
353}
354
355static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
356{
357 return (compound_page_dtor *)page[1].lru.next;
358}
359
Christoph Lameterd85f3382007-05-06 14:49:39 -0700360static inline int compound_order(struct page *page)
361{
Christoph Lameter6d777952007-05-06 14:49:40 -0700362 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700363 return 0;
364 return (unsigned long)page[1].lru.prev;
365}
366
367static inline void set_compound_order(struct page *page, unsigned long order)
368{
369 page[1].lru.prev = (void *)order;
370}
371
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800372/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 * Multiple processes may "see" the same page. E.g. for untouched
374 * mappings of /dev/null, all processes see the same page full of
375 * zeroes, and text pages of executables and shared libraries have
376 * only one copy in memory, at most, normally.
377 *
378 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700379 * page_count() == 0 means the page is free. page->lru is then used for
380 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700381 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700383 * Pages are allocated by the slab allocator in order to provide memory
384 * to kmalloc and kmem_cache_alloc. In this case, the management of the
385 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
386 * unless a particular usage is carefully commented. (the responsibility of
387 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700389 * A page may be used by anyone else who does a __get_free_page().
390 * In this case, page_count still tracks the references, and should only
391 * be used through the normal accessor functions. The top bits of page->flags
392 * and page->virtual store page management information, but all other fields
393 * are unused and could be used privately, carefully. The management of this
394 * page is the responsibility of the one who allocated it, and those who have
395 * subsequently been given references to it.
396 *
397 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 * managed by the Linux memory manager: I/O, buffers, swapping etc.
399 * The following discussion applies only to them.
400 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700401 * A pagecache page contains an opaque `private' member, which belongs to the
402 * page's address_space. Usually, this is the address of a circular list of
403 * the page's disk buffers. PG_private must be set to tell the VM to call
404 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700406 * A page may belong to an inode's memory mapping. In this case, page->mapping
407 * is the pointer to the inode, and page->index is the file offset of the page,
408 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700410 * If pagecache pages are not associated with an inode, they are said to be
411 * anonymous pages. These may become associated with the swapcache, and in that
412 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700414 * In either case (swapcache or inode backed), the pagecache itself holds one
415 * reference to the page. Setting PG_private should also increment the
416 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700418 * The pagecache pages are stored in a per-mapping radix tree, which is
419 * rooted at mapping->page_tree, and indexed by offset.
420 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
421 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700423 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 * - inode pages may need to be read from disk,
425 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700426 * to be written back to the inode on disk,
427 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
428 * modified may need to be swapped out to swap space and (later) to be read
429 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 */
431
432/*
433 * The zone field is never updated after free_area_init_core()
434 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700436
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700437
438/*
439 * page->flags layout:
440 *
441 * There are three possibilities for how page->flags get
442 * laid out. The first is for the normal case, without
443 * sparsemem. The second is for sparsemem when there is
444 * plenty of space for node and section. The last is when
445 * we have run out of space and have to fall back to an
446 * alternate (slower) way of determining the node.
447 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700448 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
449 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
450 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700451 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700452#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700453#define SECTIONS_WIDTH SECTIONS_SHIFT
454#else
455#define SECTIONS_WIDTH 0
456#endif
457
458#define ZONES_WIDTH ZONES_SHIFT
459
Christoph Lameter9223b412008-04-28 02:12:48 -0700460#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700461#define NODES_WIDTH NODES_SHIFT
462#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700463#ifdef CONFIG_SPARSEMEM_VMEMMAP
464#error "Vmemmap: No space for nodes field in page flags"
465#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700466#define NODES_WIDTH 0
467#endif
468
469/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100470#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700471#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
472#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
473
474/*
475 * We are going to use the flags for the page to node mapping if its in
476 * there. This includes the case where there is no node, so it is implicit.
477 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800478#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
479#define NODE_NOT_IN_PAGE_FLAGS
480#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700481
482#ifndef PFN_SECTION_SHIFT
483#define PFN_SECTION_SHIFT 0
484#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700485
486/*
487 * Define the bit shifts to access each section. For non-existant
488 * sections we define the shift as 0; that plus a 0 mask ensures
489 * the compiler will optimise away reference to them.
490 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700491#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
492#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
493#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700494
Christoph Lameter89689ae2006-12-06 20:31:45 -0800495/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
496#ifdef NODE_NOT_IN_PAGEFLAGS
497#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800498#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
499 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700500#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800501#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800502#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
503 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700504#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800505
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800506#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700507
Christoph Lameter9223b412008-04-28 02:12:48 -0700508#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
509#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700510#endif
511
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700512#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
513#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
514#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800515#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700516
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700517static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518{
Dave Hansen348f8b62005-06-23 00:07:40 -0700519 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
Christoph Lameter89689ae2006-12-06 20:31:45 -0800522/*
523 * The identification function is only used by the buddy allocator for
524 * determining if two pages could be buddies. We are not really
525 * identifying a zone since we could be using a the section number
526 * id if we have not node id available in page flags.
527 * We guarantee only that it will return the same value for two
528 * combinable pages in a zone.
529 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700530static inline int page_zone_id(struct page *page)
531{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800532 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533}
534
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800535static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700536{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700537#ifdef CONFIG_NUMA
538 return zone->node;
539#else
540 return 0;
541#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700542}
543
Christoph Lameter89689ae2006-12-06 20:31:45 -0800544#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800545extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800546#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800547static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700548{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800549 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700550}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800551#endif
552
553static inline struct zone *page_zone(struct page *page)
554{
555 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
556}
557
Christoph Lameter308c05e2008-04-28 02:12:43 -0700558#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700559static inline unsigned long page_to_section(struct page *page)
560{
561 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
562}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700563#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700564
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700565static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566{
Dave Hansen348f8b62005-06-23 00:07:40 -0700567 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
568 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
569}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700570
Dave Hansen348f8b62005-06-23 00:07:40 -0700571static inline void set_page_node(struct page *page, unsigned long node)
572{
573 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
574 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
575}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800576
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700577static inline void set_page_section(struct page *page, unsigned long section)
578{
579 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
580 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
581}
Dave Hansen348f8b62005-06-23 00:07:40 -0700582
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700583static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700584 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700585{
586 set_page_zone(page, zone);
587 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700588 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589}
590
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700591/*
592 * Some inline functions in vmstat.h depend on page_zone()
593 */
594#include <linux/vmstat.h>
595
Ingo Molnar652050a2006-01-14 13:21:30 -0800596static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597{
598 return __va(page_to_pfn(page) << PAGE_SHIFT);
599}
600
601#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
602#define HASHED_PAGE_VIRTUAL
603#endif
604
605#if defined(WANT_PAGE_VIRTUAL)
606#define page_address(page) ((page)->virtual)
607#define set_page_address(page, address) \
608 do { \
609 (page)->virtual = (address); \
610 } while(0)
611#define page_address_init() do { } while(0)
612#endif
613
614#if defined(HASHED_PAGE_VIRTUAL)
615void *page_address(struct page *page);
616void set_page_address(struct page *page, void *virtual);
617void page_address_init(void);
618#endif
619
620#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
621#define page_address(page) lowmem_page_address(page)
622#define set_page_address(page, address) do { } while(0)
623#define page_address_init() do { } while(0)
624#endif
625
626/*
627 * On an anonymous page mapped into a user virtual memory area,
628 * page->mapping points to its anon_vma, not to a struct address_space;
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800629 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
630 *
631 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
632 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
633 * and then page->mapping points, not to an anon_vma, but to a private
634 * structure which KSM associates with that merged page. See ksm.h.
635 *
636 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 *
638 * Please note that, confusingly, "page_mapping" refers to the inode
639 * address_space which maps the page from disk; whereas "page_mapped"
640 * refers to user virtual address space into which the page is mapped.
641 */
642#define PAGE_MAPPING_ANON 1
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800643#define PAGE_MAPPING_KSM 2
644#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
646extern struct address_space swapper_space;
647static inline struct address_space *page_mapping(struct page *page)
648{
649 struct address_space *mapping = page->mapping;
650
Christoph Lameterb5fab142007-07-17 04:03:33 -0700651 VM_BUG_ON(PageSlab(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 if (unlikely(PageSwapCache(page)))
653 mapping = &swapper_space;
Huang Shijief096e592009-12-14 17:59:51 -0800654 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 mapping = NULL;
656 return mapping;
657}
658
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800659/* Neutral page->mapping pointer to address_space or anon_vma or other */
660static inline void *page_rmapping(struct page *page)
661{
662 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
663}
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665static inline int PageAnon(struct page *page)
666{
667 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
668}
669
670/*
671 * Return the pagecache index of the passed page. Regular pagecache pages
672 * use ->index whereas swapcache pages use ->private
673 */
674static inline pgoff_t page_index(struct page *page)
675{
676 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700677 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 return page->index;
679}
680
681/*
682 * The atomic page->_mapcount, like _count, starts from -1:
683 * so that transitions both from it and to it can be tracked,
684 * using atomic_inc_and_test and atomic_add_negative(-1).
685 */
686static inline void reset_page_mapcount(struct page *page)
687{
688 atomic_set(&(page)->_mapcount, -1);
689}
690
691static inline int page_mapcount(struct page *page)
692{
693 return atomic_read(&(page)->_mapcount) + 1;
694}
695
696/*
697 * Return true if this page is mapped into pagetables.
698 */
699static inline int page_mapped(struct page *page)
700{
701 return atomic_read(&(page)->_mapcount) >= 0;
702}
703
704/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 * Different kinds of faults, as returned by handle_mm_fault().
706 * Used to decide whether a process gets delivered SIGBUS or
707 * just gets major/minor fault counters bumped up.
708 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700709
Nick Piggin83c54072007-07-19 01:47:05 -0700710#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700711
Nick Piggin83c54072007-07-19 01:47:05 -0700712#define VM_FAULT_OOM 0x0001
713#define VM_FAULT_SIGBUS 0x0002
714#define VM_FAULT_MAJOR 0x0004
715#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleend1737fd2009-09-16 11:50:06 +0200716#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000717
Nick Piggin83c54072007-07-19 01:47:05 -0700718#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
719#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720
Andi Kleend1737fd2009-09-16 11:50:06 +0200721#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON)
Nick Piggind0217ac2007-07-19 01:47:03 -0700722
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800723/*
724 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
725 */
726extern void pagefault_out_of_memory(void);
727
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
729
730extern void show_free_areas(void);
731
Hugh Dickins3f96b792009-09-21 17:03:37 -0700732int shmem_lock(struct file *file, int lock, struct user_struct *user);
Sergei Trofimovich168f5ac2009-06-16 15:33:02 -0700733struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734int shmem_zero_setup(struct vm_area_struct *);
735
David Howellsb0e15192006-01-06 00:11:42 -0800736#ifndef CONFIG_MMU
737extern unsigned long shmem_get_unmapped_area(struct file *file,
738 unsigned long addr,
739 unsigned long len,
740 unsigned long pgoff,
741 unsigned long flags);
742#endif
743
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400744extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745extern int user_shm_lock(size_t, struct user_struct *);
746extern void user_shm_unlock(size_t, struct user_struct *);
747
748/*
749 * Parameter block passed down to zap_pte_range in exceptional cases.
750 */
751struct zap_details {
752 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
753 struct address_space *check_mapping; /* Check page->mapping if set */
754 pgoff_t first_index; /* Lowest page->index to unmap */
755 pgoff_t last_index; /* Highest page->index to unmap */
756 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 unsigned long truncate_count; /* Compare vm_truncate_count */
758};
759
Nick Piggin7e675132008-04-28 02:13:00 -0700760struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
761 pte_t pte);
762
Jack Steinerc627f9c2008-07-29 22:33:53 -0700763int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
764 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700765unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700767unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 struct vm_area_struct *start_vma, unsigned long start_addr,
769 unsigned long end_addr, unsigned long *nr_accounted,
770 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800771
772/**
773 * mm_walk - callbacks for walk_page_range
774 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
775 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
776 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
777 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
778 * @pte_hole: if set, called for each hole at all levels
Naoya Horiguchi5dc37642009-12-14 18:00:01 -0800779 * @hugetlb_entry: if set, called for each hugetlb entry
Matt Mackalle6473092008-02-04 22:29:01 -0800780 *
781 * (see walk_page_range for more details)
782 */
783struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700784 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
785 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
786 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
787 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
788 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
Naoya Horiguchi116354d2010-04-06 14:35:04 -0700789 int (*hugetlb_entry)(pte_t *, unsigned long,
790 unsigned long, unsigned long, struct mm_walk *);
Dave Hansen21650092008-06-12 15:21:47 -0700791 struct mm_struct *mm;
792 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800793};
794
Dave Hansen21650092008-06-12 15:21:47 -0700795int walk_page_range(unsigned long addr, unsigned long end,
796 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700797void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700798 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
800 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801void unmap_mapping_range(struct address_space *mapping,
802 loff_t const holebegin, loff_t const holelen, int even_cows);
Johannes Weiner3b6748e2009-06-16 15:32:35 -0700803int follow_pfn(struct vm_area_struct *vma, unsigned long address,
804 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800805int follow_phys(struct vm_area_struct *vma, unsigned long address,
806 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700807int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
808 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
810static inline void unmap_shared_mapping_range(struct address_space *mapping,
811 loff_t const holebegin, loff_t const holelen)
812{
813 unmap_mapping_range(mapping, holebegin, holelen, 0);
814}
815
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000816extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
817extern int vmtruncate(struct inode *inode, loff_t offset);
818extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000819
Nick Piggin750b4982009-09-16 11:50:12 +0200820int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +0200821int generic_error_remove_page(struct address_space *mapping, struct page *page);
Nick Piggin750b4982009-09-16 11:50:12 +0200822
Wu Fengguang83f78662009-09-16 11:50:13 +0200823int invalidate_inode_page(struct page *page);
824
David Howells7ee1dd32006-01-06 00:11:44 -0800825#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700826extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700827 unsigned long address, unsigned int flags);
David Howells7ee1dd32006-01-06 00:11:44 -0800828#else
829static inline int handle_mm_fault(struct mm_struct *mm,
830 struct vm_area_struct *vma, unsigned long address,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700831 unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -0800832{
833 /* should never happen if there's no MMU */
834 BUG();
835 return VM_FAULT_SIGBUS;
836}
837#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000838
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839extern int make_pages_present(unsigned long addr, unsigned long end);
840extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
Nick Piggind2bf6be2009-06-16 15:31:39 -0700842int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
Peter Zijlstra9d737772009-06-25 11:58:55 +0200843 unsigned long start, int nr_pages, int write, int force,
Nick Piggind2bf6be2009-06-16 15:31:39 -0700844 struct page **pages, struct vm_area_struct **vmas);
845int get_user_pages_fast(unsigned long start, int nr_pages, int write,
846 struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -0700847struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848
David Howellscf9a2ae2006-08-29 19:05:54 +0100849extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
850extern void do_invalidatepage(struct page *page, unsigned long offset);
851
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800853int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854int redirty_page_for_writepage(struct writeback_control *wbc,
855 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -0700856void account_page_dirtied(struct page *page, struct address_space *mapping);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800857int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858int set_page_dirty_lock(struct page *page);
859int clear_page_dirty_for_io(struct page *page);
860
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700861extern unsigned long move_page_tables(struct vm_area_struct *vma,
862 unsigned long old_addr, struct vm_area_struct *new_vma,
863 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864extern unsigned long do_mremap(unsigned long addr,
865 unsigned long old_len, unsigned long new_len,
866 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700867extern int mprotect_fixup(struct vm_area_struct *vma,
868 struct vm_area_struct **pprev, unsigned long start,
869 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
Nick Piggin21cc1992008-07-25 19:45:22 -0700871/*
Peter Zijlstra465a4542009-06-15 12:31:37 +0200872 * doesn't attempt to fault and will return short.
873 */
874int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
875 struct page **pages);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -0800876/*
877 * per-process(per-mm_struct) statistics.
878 */
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -0800879#if defined(SPLIT_RSS_COUNTING)
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -0800880/*
881 * The mm counters are not protected by its page_table_lock,
882 * so must be incremented atomically.
883 */
884static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
885{
886 atomic_long_set(&mm->rss_stat.count[member], value);
887}
888
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -0800889unsigned long get_mm_counter(struct mm_struct *mm, int member);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -0800890
891static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
892{
893 atomic_long_add(value, &mm->rss_stat.count[member]);
894}
895
896static inline void inc_mm_counter(struct mm_struct *mm, int member)
897{
898 atomic_long_inc(&mm->rss_stat.count[member]);
899}
900
901static inline void dec_mm_counter(struct mm_struct *mm, int member)
902{
903 atomic_long_dec(&mm->rss_stat.count[member]);
904}
905
906#else /* !USE_SPLIT_PTLOCKS */
907/*
908 * The mm counters are protected by its page_table_lock,
909 * so can be incremented directly.
910 */
911static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
912{
913 mm->rss_stat.count[member] = value;
914}
915
916static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
917{
918 return mm->rss_stat.count[member];
919}
920
921static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
922{
923 mm->rss_stat.count[member] += value;
924}
925
926static inline void inc_mm_counter(struct mm_struct *mm, int member)
927{
928 mm->rss_stat.count[member]++;
929}
930
931static inline void dec_mm_counter(struct mm_struct *mm, int member)
932{
933 mm->rss_stat.count[member]--;
934}
935
936#endif /* !USE_SPLIT_PTLOCKS */
937
938static inline unsigned long get_mm_rss(struct mm_struct *mm)
939{
940 return get_mm_counter(mm, MM_FILEPAGES) +
941 get_mm_counter(mm, MM_ANONPAGES);
942}
943
944static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
945{
946 return max(mm->hiwater_rss, get_mm_rss(mm));
947}
948
949static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
950{
951 return max(mm->hiwater_vm, mm->total_vm);
952}
953
954static inline void update_hiwater_rss(struct mm_struct *mm)
955{
956 unsigned long _rss = get_mm_rss(mm);
957
958 if ((mm)->hiwater_rss < _rss)
959 (mm)->hiwater_rss = _rss;
960}
961
962static inline void update_hiwater_vm(struct mm_struct *mm)
963{
964 if (mm->hiwater_vm < mm->total_vm)
965 mm->hiwater_vm = mm->total_vm;
966}
967
968static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
969 struct mm_struct *mm)
970{
971 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
972
973 if (*maxrss < hiwater_rss)
974 *maxrss = hiwater_rss;
975}
976
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -0800977#if defined(SPLIT_RSS_COUNTING)
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -0800978void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -0800979#else
980static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
981{
982}
983#endif
Peter Zijlstra465a4542009-06-15 12:31:37 +0200984
985/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700986 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700988 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
989 * look through the least-recently-used 'nr_to_scan' entries and
990 * attempt to free them up. It should return the number of objects
991 * which remain in the cache. If it returns -1, it means it cannot do
992 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700994 * The 'gfpmask' refers to the allocation we are currently trying to
995 * fulfil.
996 *
997 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
998 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 */
Rusty Russell8e1f9362007-07-17 04:03:17 -07001000struct shrinker {
1001 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
1002 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Rusty Russell8e1f9362007-07-17 04:03:17 -07001004 /* These are for internal use */
1005 struct list_head list;
1006 long nr; /* objs pending delete */
1007};
1008#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1009extern void register_shrinker(struct shrinker *);
1010extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
Alexey Dobriyan4e950f62007-07-30 02:36:13 +04001012int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -07001013
Harvey Harrisonb3c97522008-02-13 15:03:15 -08001014extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -08001015
Nick Piggin5f22df02007-05-06 14:49:02 -07001016#ifdef __PAGETABLE_PUD_FOLDED
1017static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1018 unsigned long address)
1019{
1020 return 0;
1021}
1022#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001023int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001024#endif
1025
1026#ifdef __PAGETABLE_PMD_FOLDED
1027static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1028 unsigned long address)
1029{
1030 return 0;
1031}
1032#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001033int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001034#endif
1035
Hugh Dickins1bb36302005-10-29 18:16:22 -07001036int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1037int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039/*
1040 * The following ifdef needed to get the 4level-fixup.h header to work.
1041 * Remove it when 4level-fixup.h has been removed.
1042 */
Hugh Dickins1bb36302005-10-29 18:16:22 -07001043#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1045{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001046 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1047 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048}
1049
1050static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1051{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001052 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1053 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054}
Hugh Dickins1bb36302005-10-29 18:16:22 -07001055#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1056
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001057#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001058/*
1059 * We tuck a spinlock to guard each pagetable page into its struct page,
1060 * at page->private, with BUILD_BUG_ON to make sure that this will not
1061 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1062 * When freeing, reset page->mapping so free_pages_check won't complain.
1063 */
Andrew Morton349aef02006-01-08 01:04:36 -08001064#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001065#define pte_lock_init(_page) do { \
1066 spin_lock_init(__pte_lockptr(_page)); \
1067} while (0)
1068#define pte_lock_deinit(page) ((page)->mapping = NULL)
1069#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001070#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001071/*
1072 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1073 */
1074#define pte_lock_init(page) do {} while (0)
1075#define pte_lock_deinit(page) do {} while (0)
1076#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001077#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001078
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001079static inline void pgtable_page_ctor(struct page *page)
1080{
1081 pte_lock_init(page);
1082 inc_zone_page_state(page, NR_PAGETABLE);
1083}
1084
1085static inline void pgtable_page_dtor(struct page *page)
1086{
1087 pte_lock_deinit(page);
1088 dec_zone_page_state(page, NR_PAGETABLE);
1089}
1090
Hugh Dickinsc74df322005-10-29 18:16:23 -07001091#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1092({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001093 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001094 pte_t *__pte = pte_offset_map(pmd, address); \
1095 *(ptlp) = __ptl; \
1096 spin_lock(__ptl); \
1097 __pte; \
1098})
1099
1100#define pte_unmap_unlock(pte, ptl) do { \
1101 spin_unlock(ptl); \
1102 pte_unmap(pte); \
1103} while (0)
1104
Hugh Dickins1bb36302005-10-29 18:16:22 -07001105#define pte_alloc_map(mm, pmd, address) \
1106 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1107 NULL: pte_offset_map(pmd, address))
1108
Hugh Dickinsc74df322005-10-29 18:16:23 -07001109#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1110 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1111 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1112
Hugh Dickins1bb36302005-10-29 18:16:22 -07001113#define pte_alloc_kernel(pmd, address) \
1114 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1115 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116
1117extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001118extern void free_area_init_node(int nid, unsigned long * zones_size,
1119 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -07001120#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1121/*
1122 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1123 * zones, allocate the backing mem_map and account for memory holes in a more
1124 * architecture independent manner. This is a substitute for creating the
1125 * zone_sizes[] and zholes_size[] arrays and passing them to
1126 * free_area_init_node()
1127 *
1128 * An architecture is expected to register range of page frames backed by
1129 * physical memory with add_active_range() before calling
1130 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1131 * usage, an architecture is expected to do something like
1132 *
1133 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1134 * max_highmem_pfn};
1135 * for_each_valid_physical_page_range()
1136 * add_active_range(node_id, start_pfn, end_pfn)
1137 * free_area_init_nodes(max_zone_pfns);
1138 *
1139 * If the architecture guarantees that there are no holes in the ranges
1140 * registered with add_active_range(), free_bootmem_active_regions()
1141 * will call free_bootmem_node() for each registered physical page range.
1142 * Similarly sparse_memory_present_with_active_regions() calls
1143 * memory_present() for each range when SPARSEMEM is enabled.
1144 *
1145 * See mm/page_alloc.c for more information on each function exposed by
1146 * CONFIG_ARCH_POPULATES_NODE_MAP
1147 */
1148extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1149extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1150 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001151extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1152 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001153extern void remove_all_active_ranges(void);
Yinghai Lu32996252009-12-15 17:59:02 -08001154void sort_node_map(void);
1155unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1156 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001157extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1158 unsigned long end_pfn);
1159extern void get_pfn_range_for_nid(unsigned int nid,
1160 unsigned long *start_pfn, unsigned long *end_pfn);
1161extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001162extern void free_bootmem_with_active_regions(int nid,
1163 unsigned long max_low_pfn);
Yinghai Lu08677212010-02-10 01:20:20 -08001164int add_from_early_node_map(struct range *range, int az,
1165 int nr_range, int nid);
1166void *__alloc_memory_core_early(int nodeid, u64 size, u64 align,
1167 u64 goal, u64 limit);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001168typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001169extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001170extern void sparse_memory_present_with_active_regions(int nid);
Mel Gormanc7132162006-09-27 01:49:43 -07001171#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001172
1173#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1174 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1175static inline int __early_pfn_to_nid(unsigned long pfn)
1176{
1177 return 0;
1178}
1179#else
1180/* please see mm/page_alloc.c */
1181extern int __meminit early_pfn_to_nid(unsigned long pfn);
1182#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1183/* there is a per-arch backend function. */
1184extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1185#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1186#endif
1187
Mel Gorman0e0b8642006-09-27 01:49:56 -07001188extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa02007-01-10 23:15:30 -08001189extern void memmap_init_zone(unsigned long, int, unsigned long,
1190 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07001191extern void setup_per_zone_wmarks(void);
Minchan Kim96cb4df2009-06-16 15:32:49 -07001192extern void calculate_zone_inactive_ratio(struct zone *zone);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001194extern void __init mmap_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195extern void show_mem(void);
1196extern void si_meminfo(struct sysinfo * val);
1197extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001198extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001200extern void setup_per_cpu_pageset(void);
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001201
Shaohua Li112067f2009-09-21 17:01:16 -07001202extern void zone_pcp_update(struct zone *zone);
1203
David Howells8feae132009-01-08 12:04:47 +00001204/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001205extern atomic_long_t mmap_pages_allocated;
David Howells7e660872010-01-15 17:01:39 -08001206extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
David Howells8feae132009-01-08 12:04:47 +00001207
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208/* prio_tree.c */
1209void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1210void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1211void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1212struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1213 struct prio_tree_iter *iter);
1214
1215#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1216 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1217 (vma = vma_prio_tree_next(vma, iter)); )
1218
1219static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1220 struct list_head *list)
1221{
1222 vma->shared.vm_set.parent = NULL;
1223 list_add_tail(&vma->shared.vm_set.list, list);
1224}
1225
1226/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001227extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Rik van Riel5beb4932010-03-05 13:42:07 -08001228extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1230extern struct vm_area_struct *vma_merge(struct mm_struct *,
1231 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1232 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1233 struct mempolicy *);
1234extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1235extern int split_vma(struct mm_struct *,
1236 struct vm_area_struct *, unsigned long addr, int new_below);
1237extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1238extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1239 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001240extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1242 unsigned long addr, unsigned long len, pgoff_t pgoff);
1243extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001244
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001245extern int mm_take_all_locks(struct mm_struct *mm);
1246extern void mm_drop_all_locks(struct mm_struct *mm);
1247
Matt Helsley925d1c42008-04-29 01:01:36 -07001248#ifdef CONFIG_PROC_FS
1249/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1250extern void added_exe_file_vma(struct mm_struct *mm);
1251extern void removed_exe_file_vma(struct mm_struct *mm);
1252#else
1253static inline void added_exe_file_vma(struct mm_struct *mm)
1254{}
1255
1256static inline void removed_exe_file_vma(struct mm_struct *mm)
1257{}
1258#endif /* CONFIG_PROC_FS */
1259
akpm@osdl.org119f6572005-05-01 08:58:35 -07001260extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001261extern int install_special_mapping(struct mm_struct *mm,
1262 unsigned long addr, unsigned long len,
1263 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264
1265extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1266
1267extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1268 unsigned long len, unsigned long prot,
1269 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001270extern unsigned long mmap_region(struct file *file, unsigned long addr,
1271 unsigned long len, unsigned long flags,
Mel Gorman5a6fe122009-02-10 14:02:27 +00001272 unsigned int vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273
1274static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1275 unsigned long len, unsigned long prot,
1276 unsigned long flag, unsigned long offset)
1277{
1278 unsigned long ret = -EINVAL;
1279 if ((offset + PAGE_ALIGN(len)) < offset)
1280 goto out;
1281 if (!(offset & ~PAGE_MASK))
1282 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1283out:
1284 return ret;
1285}
1286
1287extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1288
1289extern unsigned long do_brk(unsigned long, unsigned long);
1290
1291/* filemap.c */
1292extern unsigned long page_unuse(struct page *);
1293extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001294extern void truncate_inode_pages_range(struct address_space *,
1295 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296
1297/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001298extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
1300/* mm/page-writeback.c */
1301int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001302void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303
1304/* readahead.c */
1305#define VM_MAX_READAHEAD 128 /* kbytes */
1306#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001309 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001310
1311void page_cache_sync_readahead(struct address_space *mapping,
1312 struct file_ra_state *ra,
1313 struct file *filp,
1314 pgoff_t offset,
1315 unsigned long size);
1316
1317void page_cache_async_readahead(struct address_space *mapping,
1318 struct file_ra_state *ra,
1319 struct file *filp,
1320 struct page *pg,
1321 pgoff_t offset,
1322 unsigned long size);
1323
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324unsigned long max_sane_readahead(unsigned long nr);
Wu Fengguangd30a1102009-06-16 15:31:30 -07001325unsigned long ra_submit(struct file_ra_state *ra,
1326 struct address_space *mapping,
1327 struct file *filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328
1329/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001330extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001331#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001332extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001333#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001334extern int expand_stack_downwards(struct vm_area_struct *vma,
1335 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
1337/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1338extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1339extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1340 struct vm_area_struct **pprev);
1341
1342/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1343 NULL if none. Assume start_addr < end_addr. */
1344static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1345{
1346 struct vm_area_struct * vma = find_vma(mm,start_addr);
1347
1348 if (vma && end_addr <= vma->vm_start)
1349 vma = NULL;
1350 return vma;
1351}
1352
1353static inline unsigned long vma_pages(struct vm_area_struct *vma)
1354{
1355 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1356}
1357
Hugh Dickins804af2c2006-07-26 21:39:49 +01001358pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001359struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001360int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1361 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001362int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001363int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1364 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001365int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1366 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
Linus Torvalds6aab3412005-11-28 14:34:23 -08001368struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001369 unsigned int foll_flags);
1370#define FOLL_WRITE 0x01 /* check pte is writable */
1371#define FOLL_TOUCH 0x02 /* mark page accessed */
1372#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a62009-09-21 17:03:26 -07001373#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07001374#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001376typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001377 void *data);
1378extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1379 unsigned long size, pte_fn_t fn, void *data);
1380
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001382void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001384static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 unsigned long flags, struct file *file, long pages)
1386{
1387}
1388#endif /* CONFIG_PROC_FS */
1389
Ingo Molnar12d6f212008-01-30 13:33:58 +01001390#ifdef CONFIG_DEBUG_PAGEALLOC
1391extern int debug_pagealloc_enabled;
1392
1393extern void kernel_map_pages(struct page *page, int numpages, int enable);
1394
1395static inline void enable_debug_pagealloc(void)
1396{
1397 debug_pagealloc_enabled = 1;
1398}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001399#ifdef CONFIG_HIBERNATION
1400extern bool kernel_page_present(struct page *page);
1401#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001402#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001404kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001405static inline void enable_debug_pagealloc(void)
1406{
1407}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001408#ifdef CONFIG_HIBERNATION
1409static inline bool kernel_page_present(struct page *page) { return true; }
1410#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411#endif
1412
1413extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1414#ifdef __HAVE_ARCH_GATE_AREA
1415int in_gate_area_no_task(unsigned long addr);
1416int in_gate_area(struct task_struct *task, unsigned long addr);
1417#else
1418int in_gate_area_no_task(unsigned long addr);
1419#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1420#endif /* __HAVE_ARCH_GATE_AREA */
1421
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07001422int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001423 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001424unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001425 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001426
Luke Yang7a9166e2006-02-20 18:28:07 -08001427#ifndef CONFIG_MMU
1428#define randomize_va_space 0
1429#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001430extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001431#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001432
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001433const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001434void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001435
Yinghai Lu9bdac912010-02-10 01:20:22 -08001436void sparse_mem_maps_populate_node(struct page **map_map,
1437 unsigned long pnum_begin,
1438 unsigned long pnum_end,
1439 unsigned long map_count,
1440 int nodeid);
1441
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001442struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001443pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1444pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1445pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1446pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001447void *vmemmap_alloc_block(unsigned long size, int node);
Yinghai Lu9bdac912010-02-10 01:20:22 -08001448void *vmemmap_alloc_block_buf(unsigned long size, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001449void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001450int vmemmap_populate_basepages(struct page *start_page,
1451 unsigned long pages, int node);
1452int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001453void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001454
Andi Kleen6a460792009-09-16 11:50:15 +02001455
Andi Kleen82ba0112009-12-16 12:19:57 +01001456enum mf_flags {
1457 MF_COUNT_INCREASED = 1 << 0,
1458};
Andi Kleen6a460792009-09-16 11:50:15 +02001459extern void memory_failure(unsigned long pfn, int trapno);
Andi Kleen82ba0112009-12-16 12:19:57 +01001460extern int __memory_failure(unsigned long pfn, int trapno, int flags);
Wu Fengguang847ce402009-12-16 12:19:58 +01001461extern int unpoison_memory(unsigned long pfn);
Andi Kleen6a460792009-09-16 11:50:15 +02001462extern int sysctl_memory_failure_early_kill;
1463extern int sysctl_memory_failure_recovery;
Andi Kleenfacb6012009-12-16 12:20:00 +01001464extern void shake_page(struct page *p, int access);
Andi Kleen6a460792009-09-16 11:50:15 +02001465extern atomic_long_t mce_bad_pages;
Andi Kleenfacb6012009-12-16 12:20:00 +01001466extern int soft_offline_page(struct page *page, int flags);
Andi Kleen6a460792009-09-16 11:50:15 +02001467
Wu Fengguang718a3822010-03-10 15:20:43 -08001468extern void dump_page(struct page *page);
1469
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470#endif /* __KERNEL__ */
1471#endif /* _LINUX_MM_H */