blob: 24c395694f4d3fbbc935287018e8dce157215afb [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015
16struct mempolicy;
17struct anon_vma;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040018struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040019struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040020struct writeback_control;
Markus Metzger1cb81b12009-04-24 09:51:43 +020021struct rlimit;
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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
110#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
111#endif
112
113#ifdef CONFIG_STACK_GROWSUP
114#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
115#else
116#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
117#endif
118
119#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
120#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
121#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
122#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
123#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
124
125/*
Nick Pigginb291f002008-10-18 20:26:44 -0700126 * special vmas that are non-mergable, non-mlock()able
127 */
128#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
129
130/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 * mapping from the currently active vm_flags protection bits (the
132 * low four bits) to a page protection mask..
133 */
134extern pgprot_t protection_map[16];
135
Nick Piggind0217ac2007-07-19 01:47:03 -0700136#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
137#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700138#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Nick Piggind0217ac2007-07-19 01:47:03 -0700139
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800140/*
141 * This interface is used by x86 PAT code to identify a pfn mapping that is
142 * linear over entire vma. This is to optimize PAT code that deals with
143 * marking the physical region with a particular prot. This is not for generic
144 * mm use. Note also that this check will not work if the pfn mapping is
145 * linear for a vma starting at physical address 0. In which case PAT code
146 * falls back to slow path of reserving physical range page by page.
147 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800148static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
149{
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700150 return (vma->vm_flags & VM_PFN_AT_MMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800151}
152
153static inline int is_pfn_mapping(struct vm_area_struct *vma)
154{
155 return (vma->vm_flags & VM_PFNMAP);
156}
Nick Piggind0217ac2007-07-19 01:47:03 -0700157
Nick Piggin54cb8822007-07-19 01:46:59 -0700158/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700159 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700160 * ->fault function. The vma's ->fault is responsible for returning a bitmask
161 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700162 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700163 * pgoff should be used in favour of virtual_address, if possible. If pgoff
164 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
165 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700166 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700167struct vm_fault {
168 unsigned int flags; /* FAULT_FLAG_xxx flags */
169 pgoff_t pgoff; /* Logical page offset based on vma */
170 void __user *virtual_address; /* Faulting virtual address */
171
172 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700173 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700174 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700175 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700176 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700177};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178
179/*
180 * These are the virtual MM functions - opening of an area, closing and
181 * unmapping it (needed to keep files on disk up-to-date etc), pointer
182 * to the functions called when a no-page or a wp-page exception occurs.
183 */
184struct vm_operations_struct {
185 void (*open)(struct vm_area_struct * area);
186 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700187 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700188
189 /* notification that a previously read-only page is about to become
190 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700191 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700192
193 /* called by access_process_vm when get_user_pages() fails, typically
194 * for use by special VMAs that can switch between memory and hardware
195 */
196 int (*access)(struct vm_area_struct *vma, unsigned long addr,
197 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700199 /*
200 * set_policy() op must add a reference to any non-NULL @new mempolicy
201 * to hold the policy upon return. Caller should pass NULL @new to
202 * remove a policy and fall back to surrounding context--i.e. do not
203 * install a MPOL_DEFAULT policy, nor the task or system default
204 * mempolicy.
205 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700207
208 /*
209 * get_policy() op must add reference [mpol_get()] to any policy at
210 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
211 * in mm/mempolicy.c will do this automatically.
212 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
213 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
214 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
215 * must return NULL--i.e., do not "fallback" to task or system default
216 * policy.
217 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
219 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700220 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
221 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222#endif
223};
224
225struct mmu_gather;
226struct inode;
227
Andrew Morton349aef02006-01-08 01:04:36 -0800228#define page_private(page) ((page)->private)
229#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700230
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231/*
232 * FIXME: take this include out, include page-flags.h in
233 * files which need it (119 of them)
234 */
235#include <linux/page-flags.h>
236
237/*
238 * Methods to modify the page usage count.
239 *
240 * What counts for a page usage:
241 * - cache mapping (page->mapping)
242 * - private data (page->private)
243 * - page mapped in a task's page tables, each mapping
244 * is counted separately
245 *
246 * Also, many kernel routines increase the page count before a critical
247 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 */
249
250/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700251 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800253static inline int put_page_testzero(struct page *page)
254{
Nick Piggin725d7042006-09-25 23:30:55 -0700255 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800256 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800257}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
259/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800260 * Try to grab a ref unless the page has a refcount of zero, return false if
261 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800263static inline int get_page_unless_zero(struct page *page)
264{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800265 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800266}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
Christoph Lameter48667e72008-02-04 22:28:31 -0800268/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800269struct page *vmalloc_to_page(const void *addr);
270unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800271
Paul Mundt0738c4b2008-03-12 16:51:31 +0900272/*
273 * Determine if an address is within the vmalloc range
274 *
275 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
276 * is no special casing required.
277 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800278static inline int is_vmalloc_addr(const void *x)
279{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900280#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800281 unsigned long addr = (unsigned long)x;
282
283 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900284#else
285 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800286#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900287}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700288#ifdef CONFIG_MMU
289extern int is_vmalloc_or_module_addr(const void *x);
290#else
David Howells934831d2009-09-24 12:33:32 +0100291static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700292{
293 return 0;
294}
295#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800296
Christoph Lameterd85f3382007-05-06 14:49:39 -0700297static inline struct page *compound_head(struct page *page)
298{
Christoph Lameter6d777952007-05-06 14:49:40 -0700299 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700300 return page->first_page;
301 return page;
302}
303
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700304static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700306 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307}
308
309static inline void get_page(struct page *page)
310{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700311 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700312 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 atomic_inc(&page->_count);
314}
315
Christoph Lameterb49af682007-05-06 14:49:41 -0700316static inline struct page *virt_to_head_page(const void *x)
317{
318 struct page *page = virt_to_page(x);
319 return compound_head(page);
320}
321
Nick Piggin7835e982006-03-22 00:08:40 -0800322/*
323 * Setup the page count before being freed into the page allocator for
324 * the first time (boot or memory hotplug)
325 */
326static inline void init_page_count(struct page *page)
327{
328 atomic_set(&page->_count, 1);
329}
330
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700332void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800334void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800337 * Compound pages have a destructor function. Provide a
338 * prototype for that function and accessor functions.
339 * These are _only_ valid on the head of a PG_compound page.
340 */
341typedef void compound_page_dtor(struct page *);
342
343static inline void set_compound_page_dtor(struct page *page,
344 compound_page_dtor *dtor)
345{
346 page[1].lru.next = (void *)dtor;
347}
348
349static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
350{
351 return (compound_page_dtor *)page[1].lru.next;
352}
353
Christoph Lameterd85f3382007-05-06 14:49:39 -0700354static inline int compound_order(struct page *page)
355{
Christoph Lameter6d777952007-05-06 14:49:40 -0700356 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700357 return 0;
358 return (unsigned long)page[1].lru.prev;
359}
360
361static inline void set_compound_order(struct page *page, unsigned long order)
362{
363 page[1].lru.prev = (void *)order;
364}
365
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800366/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 * Multiple processes may "see" the same page. E.g. for untouched
368 * mappings of /dev/null, all processes see the same page full of
369 * zeroes, and text pages of executables and shared libraries have
370 * only one copy in memory, at most, normally.
371 *
372 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700373 * page_count() == 0 means the page is free. page->lru is then used for
374 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700375 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700377 * Pages are allocated by the slab allocator in order to provide memory
378 * to kmalloc and kmem_cache_alloc. In this case, the management of the
379 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
380 * unless a particular usage is carefully commented. (the responsibility of
381 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700383 * A page may be used by anyone else who does a __get_free_page().
384 * In this case, page_count still tracks the references, and should only
385 * be used through the normal accessor functions. The top bits of page->flags
386 * and page->virtual store page management information, but all other fields
387 * are unused and could be used privately, carefully. The management of this
388 * page is the responsibility of the one who allocated it, and those who have
389 * subsequently been given references to it.
390 *
391 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 * managed by the Linux memory manager: I/O, buffers, swapping etc.
393 * The following discussion applies only to them.
394 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700395 * A pagecache page contains an opaque `private' member, which belongs to the
396 * page's address_space. Usually, this is the address of a circular list of
397 * the page's disk buffers. PG_private must be set to tell the VM to call
398 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700400 * A page may belong to an inode's memory mapping. In this case, page->mapping
401 * is the pointer to the inode, and page->index is the file offset of the page,
402 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700404 * If pagecache pages are not associated with an inode, they are said to be
405 * anonymous pages. These may become associated with the swapcache, and in that
406 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700408 * In either case (swapcache or inode backed), the pagecache itself holds one
409 * reference to the page. Setting PG_private should also increment the
410 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700412 * The pagecache pages are stored in a per-mapping radix tree, which is
413 * rooted at mapping->page_tree, and indexed by offset.
414 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
415 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700417 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 * - inode pages may need to be read from disk,
419 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700420 * to be written back to the inode on disk,
421 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
422 * modified may need to be swapped out to swap space and (later) to be read
423 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 */
425
426/*
427 * The zone field is never updated after free_area_init_core()
428 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700430
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700431
432/*
433 * page->flags layout:
434 *
435 * There are three possibilities for how page->flags get
436 * laid out. The first is for the normal case, without
437 * sparsemem. The second is for sparsemem when there is
438 * plenty of space for node and section. The last is when
439 * we have run out of space and have to fall back to an
440 * alternate (slower) way of determining the node.
441 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700442 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
443 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
444 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700445 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700446#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700447#define SECTIONS_WIDTH SECTIONS_SHIFT
448#else
449#define SECTIONS_WIDTH 0
450#endif
451
452#define ZONES_WIDTH ZONES_SHIFT
453
Christoph Lameter9223b412008-04-28 02:12:48 -0700454#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700455#define NODES_WIDTH NODES_SHIFT
456#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700457#ifdef CONFIG_SPARSEMEM_VMEMMAP
458#error "Vmemmap: No space for nodes field in page flags"
459#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700460#define NODES_WIDTH 0
461#endif
462
463/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100464#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700465#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
466#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
467
468/*
469 * We are going to use the flags for the page to node mapping if its in
470 * there. This includes the case where there is no node, so it is implicit.
471 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800472#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
473#define NODE_NOT_IN_PAGE_FLAGS
474#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700475
476#ifndef PFN_SECTION_SHIFT
477#define PFN_SECTION_SHIFT 0
478#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700479
480/*
481 * Define the bit shifts to access each section. For non-existant
482 * sections we define the shift as 0; that plus a 0 mask ensures
483 * the compiler will optimise away reference to them.
484 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700485#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
486#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
487#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700488
Christoph Lameter89689ae2006-12-06 20:31:45 -0800489/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
490#ifdef NODE_NOT_IN_PAGEFLAGS
491#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800492#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
493 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700494#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800495#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800496#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
497 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700498#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800499
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800500#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700501
Christoph Lameter9223b412008-04-28 02:12:48 -0700502#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
503#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700504#endif
505
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700506#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
507#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
508#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800509#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700510
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700511static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
Dave Hansen348f8b62005-06-23 00:07:40 -0700513 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Christoph Lameter89689ae2006-12-06 20:31:45 -0800516/*
517 * The identification function is only used by the buddy allocator for
518 * determining if two pages could be buddies. We are not really
519 * identifying a zone since we could be using a the section number
520 * id if we have not node id available in page flags.
521 * We guarantee only that it will return the same value for two
522 * combinable pages in a zone.
523 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700524static inline int page_zone_id(struct page *page)
525{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800526 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527}
528
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800529static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700530{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700531#ifdef CONFIG_NUMA
532 return zone->node;
533#else
534 return 0;
535#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700536}
537
Christoph Lameter89689ae2006-12-06 20:31:45 -0800538#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800539extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800540#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800541static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700542{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800543 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700544}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800545#endif
546
547static inline struct zone *page_zone(struct page *page)
548{
549 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
550}
551
Christoph Lameter308c05e2008-04-28 02:12:43 -0700552#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700553static inline unsigned long page_to_section(struct page *page)
554{
555 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
556}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700557#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700558
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700559static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560{
Dave Hansen348f8b62005-06-23 00:07:40 -0700561 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
562 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
563}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700564
Dave Hansen348f8b62005-06-23 00:07:40 -0700565static inline void set_page_node(struct page *page, unsigned long node)
566{
567 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
568 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
569}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800570
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700571static inline void set_page_section(struct page *page, unsigned long section)
572{
573 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
574 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
575}
Dave Hansen348f8b62005-06-23 00:07:40 -0700576
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700577static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700578 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700579{
580 set_page_zone(page, zone);
581 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700582 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583}
584
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700585/*
586 * Some inline functions in vmstat.h depend on page_zone()
587 */
588#include <linux/vmstat.h>
589
Ingo Molnar652050a2006-01-14 13:21:30 -0800590static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591{
592 return __va(page_to_pfn(page) << PAGE_SHIFT);
593}
594
595#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
596#define HASHED_PAGE_VIRTUAL
597#endif
598
599#if defined(WANT_PAGE_VIRTUAL)
600#define page_address(page) ((page)->virtual)
601#define set_page_address(page, address) \
602 do { \
603 (page)->virtual = (address); \
604 } while(0)
605#define page_address_init() do { } while(0)
606#endif
607
608#if defined(HASHED_PAGE_VIRTUAL)
609void *page_address(struct page *page);
610void set_page_address(struct page *page, void *virtual);
611void page_address_init(void);
612#endif
613
614#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
615#define page_address(page) lowmem_page_address(page)
616#define set_page_address(page, address) do { } while(0)
617#define page_address_init() do { } while(0)
618#endif
619
620/*
621 * On an anonymous page mapped into a user virtual memory area,
622 * page->mapping points to its anon_vma, not to a struct address_space;
623 * with the PAGE_MAPPING_ANON bit set to distinguish it.
624 *
625 * Please note that, confusingly, "page_mapping" refers to the inode
626 * address_space which maps the page from disk; whereas "page_mapped"
627 * refers to user virtual address space into which the page is mapped.
628 */
629#define PAGE_MAPPING_ANON 1
630
631extern struct address_space swapper_space;
632static inline struct address_space *page_mapping(struct page *page)
633{
634 struct address_space *mapping = page->mapping;
635
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700636 VM_BUG_ON(PageSlab(page));
Andrew Morton726b8012008-04-28 02:12:44 -0700637#ifdef CONFIG_SWAP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 if (unlikely(PageSwapCache(page)))
639 mapping = &swapper_space;
Andrew Morton726b8012008-04-28 02:12:44 -0700640 else
641#endif
642 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 mapping = NULL;
644 return mapping;
645}
646
647static inline int PageAnon(struct page *page)
648{
649 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
650}
651
652/*
653 * Return the pagecache index of the passed page. Regular pagecache pages
654 * use ->index whereas swapcache pages use ->private
655 */
656static inline pgoff_t page_index(struct page *page)
657{
658 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700659 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 return page->index;
661}
662
663/*
664 * The atomic page->_mapcount, like _count, starts from -1:
665 * so that transitions both from it and to it can be tracked,
666 * using atomic_inc_and_test and atomic_add_negative(-1).
667 */
668static inline void reset_page_mapcount(struct page *page)
669{
670 atomic_set(&(page)->_mapcount, -1);
671}
672
673static inline int page_mapcount(struct page *page)
674{
675 return atomic_read(&(page)->_mapcount) + 1;
676}
677
678/*
679 * Return true if this page is mapped into pagetables.
680 */
681static inline int page_mapped(struct page *page)
682{
683 return atomic_read(&(page)->_mapcount) >= 0;
684}
685
686/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 * Different kinds of faults, as returned by handle_mm_fault().
688 * Used to decide whether a process gets delivered SIGBUS or
689 * just gets major/minor fault counters bumped up.
690 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700691
Nick Piggin83c54072007-07-19 01:47:05 -0700692#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700693
Nick Piggin83c54072007-07-19 01:47:05 -0700694#define VM_FAULT_OOM 0x0001
695#define VM_FAULT_SIGBUS 0x0002
696#define VM_FAULT_MAJOR 0x0004
697#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleend1737fd2009-09-16 11:50:06 +0200698#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000699
Nick Piggin83c54072007-07-19 01:47:05 -0700700#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
701#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
Andi Kleend1737fd2009-09-16 11:50:06 +0200703#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON)
Nick Piggind0217ac2007-07-19 01:47:03 -0700704
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800705/*
706 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
707 */
708extern void pagefault_out_of_memory(void);
709
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
711
712extern void show_free_areas(void);
713
Hugh Dickins3f96b792009-09-21 17:03:37 -0700714int shmem_lock(struct file *file, int lock, struct user_struct *user);
Sergei Trofimovich168f5ac2009-06-16 15:33:02 -0700715struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716int shmem_zero_setup(struct vm_area_struct *);
717
David Howellsb0e15192006-01-06 00:11:42 -0800718#ifndef CONFIG_MMU
719extern unsigned long shmem_get_unmapped_area(struct file *file,
720 unsigned long addr,
721 unsigned long len,
722 unsigned long pgoff,
723 unsigned long flags);
724#endif
725
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400726extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727extern int user_shm_lock(size_t, struct user_struct *);
728extern void user_shm_unlock(size_t, struct user_struct *);
729
730/*
731 * Parameter block passed down to zap_pte_range in exceptional cases.
732 */
733struct zap_details {
734 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
735 struct address_space *check_mapping; /* Check page->mapping if set */
736 pgoff_t first_index; /* Lowest page->index to unmap */
737 pgoff_t last_index; /* Highest page->index to unmap */
738 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 unsigned long truncate_count; /* Compare vm_truncate_count */
740};
741
Nick Piggin7e675132008-04-28 02:13:00 -0700742struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
743 pte_t pte);
744
Jack Steinerc627f9c2008-07-29 22:33:53 -0700745int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
746 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700747unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700749unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 struct vm_area_struct *start_vma, unsigned long start_addr,
751 unsigned long end_addr, unsigned long *nr_accounted,
752 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800753
754/**
755 * mm_walk - callbacks for walk_page_range
756 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
757 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
758 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
759 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
760 * @pte_hole: if set, called for each hole at all levels
761 *
762 * (see walk_page_range for more details)
763 */
764struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700765 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
766 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
767 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
768 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
769 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
770 struct mm_struct *mm;
771 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800772};
773
Dave Hansen21650092008-06-12 15:21:47 -0700774int walk_page_range(unsigned long addr, unsigned long end,
775 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700776void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700777 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
779 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780void unmap_mapping_range(struct address_space *mapping,
781 loff_t const holebegin, loff_t const holelen, int even_cows);
Johannes Weiner3b6748e2009-06-16 15:32:35 -0700782int follow_pfn(struct vm_area_struct *vma, unsigned long address,
783 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800784int follow_phys(struct vm_area_struct *vma, unsigned long address,
785 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700786int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
787 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789static inline void unmap_shared_mapping_range(struct address_space *mapping,
790 loff_t const holebegin, loff_t const holelen)
791{
792 unmap_mapping_range(mapping, holebegin, holelen, 0);
793}
794
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000795extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
796extern int vmtruncate(struct inode *inode, loff_t offset);
797extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000798
Nick Piggin750b4982009-09-16 11:50:12 +0200799int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +0200800int generic_error_remove_page(struct address_space *mapping, struct page *page);
Nick Piggin750b4982009-09-16 11:50:12 +0200801
Wu Fengguang83f78662009-09-16 11:50:13 +0200802int invalidate_inode_page(struct page *page);
803
David Howells7ee1dd32006-01-06 00:11:44 -0800804#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700805extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700806 unsigned long address, unsigned int flags);
David Howells7ee1dd32006-01-06 00:11:44 -0800807#else
808static inline int handle_mm_fault(struct mm_struct *mm,
809 struct vm_area_struct *vma, unsigned long address,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700810 unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -0800811{
812 /* should never happen if there's no MMU */
813 BUG();
814 return VM_FAULT_SIGBUS;
815}
816#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000817
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818extern int make_pages_present(unsigned long addr, unsigned long end);
819extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
Nick Piggind2bf6be2009-06-16 15:31:39 -0700821int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
Peter Zijlstra9d737772009-06-25 11:58:55 +0200822 unsigned long start, int nr_pages, int write, int force,
Nick Piggind2bf6be2009-06-16 15:31:39 -0700823 struct page **pages, struct vm_area_struct **vmas);
824int get_user_pages_fast(unsigned long start, int nr_pages, int write,
825 struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -0700826struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
David Howellscf9a2ae2006-08-29 19:05:54 +0100828extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
829extern void do_invalidatepage(struct page *page, unsigned long offset);
830
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800832int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833int redirty_page_for_writepage(struct writeback_control *wbc,
834 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -0700835void account_page_dirtied(struct page *page, struct address_space *mapping);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800836int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837int set_page_dirty_lock(struct page *page);
838int clear_page_dirty_for_io(struct page *page);
839
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700840extern unsigned long move_page_tables(struct vm_area_struct *vma,
841 unsigned long old_addr, struct vm_area_struct *new_vma,
842 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843extern unsigned long do_mremap(unsigned long addr,
844 unsigned long old_len, unsigned long new_len,
845 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700846extern int mprotect_fixup(struct vm_area_struct *vma,
847 struct vm_area_struct **pprev, unsigned long start,
848 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Nick Piggin21cc1992008-07-25 19:45:22 -0700850/*
Peter Zijlstra465a4542009-06-15 12:31:37 +0200851 * doesn't attempt to fault and will return short.
852 */
853int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
854 struct page **pages);
855
856/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700857 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700859 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
860 * look through the least-recently-used 'nr_to_scan' entries and
861 * attempt to free them up. It should return the number of objects
862 * which remain in the cache. If it returns -1, it means it cannot do
863 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700865 * The 'gfpmask' refers to the allocation we are currently trying to
866 * fulfil.
867 *
868 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
869 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 */
Rusty Russell8e1f9362007-07-17 04:03:17 -0700871struct shrinker {
872 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
873 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
Rusty Russell8e1f9362007-07-17 04:03:17 -0700875 /* These are for internal use */
876 struct list_head list;
877 long nr; /* objs pending delete */
878};
879#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
880extern void register_shrinker(struct shrinker *);
881extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882
Alexey Dobriyan4e950f62007-07-30 02:36:13 +0400883int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700884
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800885extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800886
Nick Piggin5f22df02007-05-06 14:49:02 -0700887#ifdef __PAGETABLE_PUD_FOLDED
888static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
889 unsigned long address)
890{
891 return 0;
892}
893#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700894int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700895#endif
896
897#ifdef __PAGETABLE_PMD_FOLDED
898static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
899 unsigned long address)
900{
901 return 0;
902}
903#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700904int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700905#endif
906
Hugh Dickins1bb36302005-10-29 18:16:22 -0700907int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
908int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
909
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910/*
911 * The following ifdef needed to get the 4level-fixup.h header to work.
912 * Remove it when 4level-fixup.h has been removed.
913 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700914#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
916{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700917 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
918 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919}
920
921static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
922{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700923 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
924 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700926#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
927
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700928#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700929/*
930 * We tuck a spinlock to guard each pagetable page into its struct page,
931 * at page->private, with BUILD_BUG_ON to make sure that this will not
932 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
933 * When freeing, reset page->mapping so free_pages_check won't complain.
934 */
Andrew Morton349aef02006-01-08 01:04:36 -0800935#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700936#define pte_lock_init(_page) do { \
937 spin_lock_init(__pte_lockptr(_page)); \
938} while (0)
939#define pte_lock_deinit(page) ((page)->mapping = NULL)
940#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700941#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700942/*
943 * We use mm->page_table_lock to guard all pagetable pages of the mm.
944 */
945#define pte_lock_init(page) do {} while (0)
946#define pte_lock_deinit(page) do {} while (0)
947#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700948#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700949
Martin Schwidefsky2f569af2008-02-08 04:22:04 -0800950static inline void pgtable_page_ctor(struct page *page)
951{
952 pte_lock_init(page);
953 inc_zone_page_state(page, NR_PAGETABLE);
954}
955
956static inline void pgtable_page_dtor(struct page *page)
957{
958 pte_lock_deinit(page);
959 dec_zone_page_state(page, NR_PAGETABLE);
960}
961
Hugh Dickinsc74df322005-10-29 18:16:23 -0700962#define pte_offset_map_lock(mm, pmd, address, ptlp) \
963({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700964 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700965 pte_t *__pte = pte_offset_map(pmd, address); \
966 *(ptlp) = __ptl; \
967 spin_lock(__ptl); \
968 __pte; \
969})
970
971#define pte_unmap_unlock(pte, ptl) do { \
972 spin_unlock(ptl); \
973 pte_unmap(pte); \
974} while (0)
975
Hugh Dickins1bb36302005-10-29 18:16:22 -0700976#define pte_alloc_map(mm, pmd, address) \
977 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
978 NULL: pte_offset_map(pmd, address))
979
Hugh Dickinsc74df322005-10-29 18:16:23 -0700980#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
981 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
982 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
983
Hugh Dickins1bb36302005-10-29 18:16:22 -0700984#define pte_alloc_kernel(pmd, address) \
985 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
986 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
988extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -0700989extern void free_area_init_node(int nid, unsigned long * zones_size,
990 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -0700991#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
992/*
993 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
994 * zones, allocate the backing mem_map and account for memory holes in a more
995 * architecture independent manner. This is a substitute for creating the
996 * zone_sizes[] and zholes_size[] arrays and passing them to
997 * free_area_init_node()
998 *
999 * An architecture is expected to register range of page frames backed by
1000 * physical memory with add_active_range() before calling
1001 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1002 * usage, an architecture is expected to do something like
1003 *
1004 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1005 * max_highmem_pfn};
1006 * for_each_valid_physical_page_range()
1007 * add_active_range(node_id, start_pfn, end_pfn)
1008 * free_area_init_nodes(max_zone_pfns);
1009 *
1010 * If the architecture guarantees that there are no holes in the ranges
1011 * registered with add_active_range(), free_bootmem_active_regions()
1012 * will call free_bootmem_node() for each registered physical page range.
1013 * Similarly sparse_memory_present_with_active_regions() calls
1014 * memory_present() for each range when SPARSEMEM is enabled.
1015 *
1016 * See mm/page_alloc.c for more information on each function exposed by
1017 * CONFIG_ARCH_POPULATES_NODE_MAP
1018 */
1019extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1020extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1021 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001022extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1023 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001024extern void remove_all_active_ranges(void);
1025extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1026 unsigned long end_pfn);
1027extern void get_pfn_range_for_nid(unsigned int nid,
1028 unsigned long *start_pfn, unsigned long *end_pfn);
1029extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001030extern void free_bootmem_with_active_regions(int nid,
1031 unsigned long max_low_pfn);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001032typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001033extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001034extern void sparse_memory_present_with_active_regions(int nid);
Mel Gormanc7132162006-09-27 01:49:43 -07001035#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001036
1037#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1038 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1039static inline int __early_pfn_to_nid(unsigned long pfn)
1040{
1041 return 0;
1042}
1043#else
1044/* please see mm/page_alloc.c */
1045extern int __meminit early_pfn_to_nid(unsigned long pfn);
1046#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1047/* there is a per-arch backend function. */
1048extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1049#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1050#endif
1051
Mel Gorman0e0b8642006-09-27 01:49:56 -07001052extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001053extern void memmap_init_zone(unsigned long, int, unsigned long,
1054 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07001055extern void setup_per_zone_wmarks(void);
Minchan Kim96cb4df2009-06-16 15:32:49 -07001056extern void calculate_zone_inactive_ratio(struct zone *zone);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001058extern void __init mmap_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059extern void show_mem(void);
1060extern void si_meminfo(struct sysinfo * val);
1061extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001062extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001064#ifdef CONFIG_NUMA
1065extern void setup_per_cpu_pageset(void);
1066#else
1067static inline void setup_per_cpu_pageset(void) {}
1068#endif
1069
Shaohua Li112067f2009-09-21 17:01:16 -07001070extern void zone_pcp_update(struct zone *zone);
1071
David Howells8feae132009-01-08 12:04:47 +00001072/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001073extern atomic_long_t mmap_pages_allocated;
David Howells8feae132009-01-08 12:04:47 +00001074
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075/* prio_tree.c */
1076void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1077void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1078void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1079struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1080 struct prio_tree_iter *iter);
1081
1082#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1083 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1084 (vma = vma_prio_tree_next(vma, iter)); )
1085
1086static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1087 struct list_head *list)
1088{
1089 vma->shared.vm_set.parent = NULL;
1090 list_add_tail(&vma->shared.vm_set.list, list);
1091}
1092
1093/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001094extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1096 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1097extern struct vm_area_struct *vma_merge(struct mm_struct *,
1098 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1099 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1100 struct mempolicy *);
1101extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1102extern int split_vma(struct mm_struct *,
1103 struct vm_area_struct *, unsigned long addr, int new_below);
1104extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1105extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1106 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001107extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1109 unsigned long addr, unsigned long len, pgoff_t pgoff);
1110extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001111
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001112extern int mm_take_all_locks(struct mm_struct *mm);
1113extern void mm_drop_all_locks(struct mm_struct *mm);
1114
Matt Helsley925d1c42008-04-29 01:01:36 -07001115#ifdef CONFIG_PROC_FS
1116/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1117extern void added_exe_file_vma(struct mm_struct *mm);
1118extern void removed_exe_file_vma(struct mm_struct *mm);
1119#else
1120static inline void added_exe_file_vma(struct mm_struct *mm)
1121{}
1122
1123static inline void removed_exe_file_vma(struct mm_struct *mm)
1124{}
1125#endif /* CONFIG_PROC_FS */
1126
akpm@osdl.org119f6572005-05-01 08:58:35 -07001127extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001128extern int install_special_mapping(struct mm_struct *mm,
1129 unsigned long addr, unsigned long len,
1130 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131
1132extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1133
1134extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1135 unsigned long len, unsigned long prot,
1136 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001137extern unsigned long mmap_region(struct file *file, unsigned long addr,
1138 unsigned long len, unsigned long flags,
Mel Gorman5a6fe122009-02-10 14:02:27 +00001139 unsigned int vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140
1141static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1142 unsigned long len, unsigned long prot,
1143 unsigned long flag, unsigned long offset)
1144{
1145 unsigned long ret = -EINVAL;
1146 if ((offset + PAGE_ALIGN(len)) < offset)
1147 goto out;
1148 if (!(offset & ~PAGE_MASK))
1149 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1150out:
1151 return ret;
1152}
1153
1154extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1155
1156extern unsigned long do_brk(unsigned long, unsigned long);
1157
1158/* filemap.c */
1159extern unsigned long page_unuse(struct page *);
1160extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001161extern void truncate_inode_pages_range(struct address_space *,
1162 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
1164/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001165extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
1167/* mm/page-writeback.c */
1168int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001169void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170
1171/* readahead.c */
1172#define VM_MAX_READAHEAD 128 /* kbytes */
1173#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001176 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001177
1178void page_cache_sync_readahead(struct address_space *mapping,
1179 struct file_ra_state *ra,
1180 struct file *filp,
1181 pgoff_t offset,
1182 unsigned long size);
1183
1184void page_cache_async_readahead(struct address_space *mapping,
1185 struct file_ra_state *ra,
1186 struct file *filp,
1187 struct page *pg,
1188 pgoff_t offset,
1189 unsigned long size);
1190
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191unsigned long max_sane_readahead(unsigned long nr);
Wu Fengguangd30a1102009-06-16 15:31:30 -07001192unsigned long ra_submit(struct file_ra_state *ra,
1193 struct address_space *mapping,
1194 struct file *filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195
1196/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001197extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001198#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001199extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001200#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001201extern int expand_stack_downwards(struct vm_area_struct *vma,
1202 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
1204/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1205extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1206extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1207 struct vm_area_struct **pprev);
1208
1209/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1210 NULL if none. Assume start_addr < end_addr. */
1211static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1212{
1213 struct vm_area_struct * vma = find_vma(mm,start_addr);
1214
1215 if (vma && end_addr <= vma->vm_start)
1216 vma = NULL;
1217 return vma;
1218}
1219
1220static inline unsigned long vma_pages(struct vm_area_struct *vma)
1221{
1222 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1223}
1224
Hugh Dickins804af2c2006-07-26 21:39:49 +01001225pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001226struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001227int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1228 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001229int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001230int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1231 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001232int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1233 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
Linus Torvalds6aab3412005-11-28 14:34:23 -08001235struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001236 unsigned int foll_flags);
1237#define FOLL_WRITE 0x01 /* check pte is writable */
1238#define FOLL_TOUCH 0x02 /* mark page accessed */
1239#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a602009-09-21 17:03:26 -07001240#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07001241#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001243typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001244 void *data);
1245extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1246 unsigned long size, pte_fn_t fn, void *data);
1247
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001249void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001251static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 unsigned long flags, struct file *file, long pages)
1253{
1254}
1255#endif /* CONFIG_PROC_FS */
1256
Ingo Molnar12d6f212008-01-30 13:33:58 +01001257#ifdef CONFIG_DEBUG_PAGEALLOC
1258extern int debug_pagealloc_enabled;
1259
1260extern void kernel_map_pages(struct page *page, int numpages, int enable);
1261
1262static inline void enable_debug_pagealloc(void)
1263{
1264 debug_pagealloc_enabled = 1;
1265}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001266#ifdef CONFIG_HIBERNATION
1267extern bool kernel_page_present(struct page *page);
1268#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001269#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001271kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001272static inline void enable_debug_pagealloc(void)
1273{
1274}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001275#ifdef CONFIG_HIBERNATION
1276static inline bool kernel_page_present(struct page *page) { return true; }
1277#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278#endif
1279
1280extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1281#ifdef __HAVE_ARCH_GATE_AREA
1282int in_gate_area_no_task(unsigned long addr);
1283int in_gate_area(struct task_struct *task, unsigned long addr);
1284#else
1285int in_gate_area_no_task(unsigned long addr);
1286#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1287#endif /* __HAVE_ARCH_GATE_AREA */
1288
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07001289int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001290 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001291unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001292 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001293
Luke Yang7a9166e2006-02-20 18:28:07 -08001294#ifndef CONFIG_MMU
1295#define randomize_va_space 0
1296#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001297extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001298#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001299
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001300const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001301void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001302
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001303struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001304pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1305pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1306pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1307pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001308void *vmemmap_alloc_block(unsigned long size, int node);
1309void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001310int vmemmap_populate_basepages(struct page *start_page,
1311 unsigned long pages, int node);
1312int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001313void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001314
Markus Metzger1cb81b12009-04-24 09:51:43 +02001315extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim,
1316 size_t size);
1317extern void refund_locked_memory(struct mm_struct *mm, size_t size);
Andi Kleen6a460792009-09-16 11:50:15 +02001318
1319extern void memory_failure(unsigned long pfn, int trapno);
1320extern int __memory_failure(unsigned long pfn, int trapno, int ref);
1321extern int sysctl_memory_failure_early_kill;
1322extern int sysctl_memory_failure_recovery;
1323extern atomic_long_t mce_bad_pages;
1324
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325#endif /* __KERNEL__ */
1326#endif /* _LINUX_MM_H */