blob: aaa8b843be28614a39e30de2772fe9e7af6b8815 [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>
Jiri Slaby59ea7462008-06-12 13:56:40 +020010#include <linux/mmdebug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/mmzone.h>
12#include <linux/rbtree.h>
13#include <linux/prio_tree.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070014#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070015#include <linux/mm_types.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;
28extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070029extern int page_cluster;
30
31#ifdef CONFIG_SYSCTL
32extern int sysctl_legacy_va_layout;
33#else
34#define sysctl_legacy_va_layout 0
35#endif
36
James Morris42d78962007-12-26 15:12:37 +110037extern unsigned long mmap_min_addr;
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#include <asm/page.h>
40#include <asm/pgtable.h>
41#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
44
Andrea Righi27ac7922008-07-23 21:28:13 -070045/* to align the pointer to the (next) page boundary */
46#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
47
Linus Torvalds1da177e2005-04-16 15:20:36 -070048/*
49 * Linux kernel virtual memory manager primitives.
50 * The idea being to have a "virtual" mm in the same way
51 * we have a virtual fs - giving a cleaner interface to the
52 * mm details, and allowing different kinds of memory mappings
53 * (from shared memory to executable loading to arbitrary
54 * mmap() functions).
55 */
56
Christoph Lameterc43692e2006-12-06 20:32:48 -080057extern struct kmem_cache *vm_area_cachep;
58
Linus Torvalds1da177e2005-04-16 15:20:36 -070059/*
60 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
61 * disabled, then there's a single shared list of VMAs maintained by the
62 * system, and mm's subscribe to these individually
63 */
64struct vm_list_struct {
65 struct vm_list_struct *next;
66 struct vm_area_struct *vma;
67};
68
69#ifndef CONFIG_MMU
70extern struct rb_root nommu_vma_tree;
71extern struct rw_semaphore nommu_vma_sem;
72
73extern unsigned int kobjsize(const void *objp);
74#endif
75
76/*
Hugh Dickins605d9282008-08-16 11:07:21 +010077 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 */
79#define VM_READ 0x00000001 /* currently active flags */
80#define VM_WRITE 0x00000002
81#define VM_EXEC 0x00000004
82#define VM_SHARED 0x00000008
83
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070084/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070085#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
86#define VM_MAYWRITE 0x00000020
87#define VM_MAYEXEC 0x00000040
88#define VM_MAYSHARE 0x00000080
89
90#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
91#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -080092#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
94
95#define VM_EXECUTABLE 0x00001000
96#define VM_LOCKED 0x00002000
97#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
98
99 /* Used by sys_madvise() */
100#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
101#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
102
103#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
104#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800105#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -0700107#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
109#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
110#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800111#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800112#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Nick Piggind0217ac2007-07-19 01:47:03 -0700114#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700115#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000116#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Nick Piggind00806b2007-07-19 01:46:57 -0700117
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
119#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
120#endif
121
122#ifdef CONFIG_STACK_GROWSUP
123#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
124#else
125#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
126#endif
127
128#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
129#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
130#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
131#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
132#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
133
134/*
Nick Pigginb291f002008-10-18 20:26:44 -0700135 * special vmas that are non-mergable, non-mlock()able
136 */
137#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
138
139/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 * mapping from the currently active vm_flags protection bits (the
141 * low four bits) to a page protection mask..
142 */
143extern pgprot_t protection_map[16];
144
Nick Piggind0217ac2007-07-19 01:47:03 -0700145#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
146#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
147
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800148/*
149 * This interface is used by x86 PAT code to identify a pfn mapping that is
150 * linear over entire vma. This is to optimize PAT code that deals with
151 * marking the physical region with a particular prot. This is not for generic
152 * mm use. Note also that this check will not work if the pfn mapping is
153 * linear for a vma starting at physical address 0. In which case PAT code
154 * falls back to slow path of reserving physical range page by page.
155 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800156static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
157{
158 return ((vma->vm_flags & VM_PFNMAP) && vma->vm_pgoff);
159}
160
161static inline int is_pfn_mapping(struct vm_area_struct *vma)
162{
163 return (vma->vm_flags & VM_PFNMAP);
164}
Nick Piggind0217ac2007-07-19 01:47:03 -0700165
Nick Piggin54cb8822007-07-19 01:46:59 -0700166/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700167 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700168 * ->fault function. The vma's ->fault is responsible for returning a bitmask
169 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700170 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700171 * pgoff should be used in favour of virtual_address, if possible. If pgoff
172 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
173 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700174 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700175struct vm_fault {
176 unsigned int flags; /* FAULT_FLAG_xxx flags */
177 pgoff_t pgoff; /* Logical page offset based on vma */
178 void __user *virtual_address; /* Faulting virtual address */
179
180 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700181 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700182 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700183 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700184 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700185};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
187/*
188 * These are the virtual MM functions - opening of an area, closing and
189 * unmapping it (needed to keep files on disk up-to-date etc), pointer
190 * to the functions called when a no-page or a wp-page exception occurs.
191 */
192struct vm_operations_struct {
193 void (*open)(struct vm_area_struct * area);
194 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700195 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700196
197 /* notification that a previously read-only page is about to become
198 * writable, if an error is returned it will cause a SIGBUS */
199 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700200
201 /* called by access_process_vm when get_user_pages() fails, typically
202 * for use by special VMAs that can switch between memory and hardware
203 */
204 int (*access)(struct vm_area_struct *vma, unsigned long addr,
205 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700207 /*
208 * set_policy() op must add a reference to any non-NULL @new mempolicy
209 * to hold the policy upon return. Caller should pass NULL @new to
210 * remove a policy and fall back to surrounding context--i.e. do not
211 * install a MPOL_DEFAULT policy, nor the task or system default
212 * mempolicy.
213 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700215
216 /*
217 * get_policy() op must add reference [mpol_get()] to any policy at
218 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
219 * in mm/mempolicy.c will do this automatically.
220 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
221 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
222 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
223 * must return NULL--i.e., do not "fallback" to task or system default
224 * policy.
225 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
227 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700228 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
229 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230#endif
231};
232
233struct mmu_gather;
234struct inode;
235
Andrew Morton349aef02006-01-08 01:04:36 -0800236#define page_private(page) ((page)->private)
237#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239/*
240 * FIXME: take this include out, include page-flags.h in
241 * files which need it (119 of them)
242 */
243#include <linux/page-flags.h>
244
245/*
246 * Methods to modify the page usage count.
247 *
248 * What counts for a page usage:
249 * - cache mapping (page->mapping)
250 * - private data (page->private)
251 * - page mapped in a task's page tables, each mapping
252 * is counted separately
253 *
254 * Also, many kernel routines increase the page count before a critical
255 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 */
257
258/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700259 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800261static inline int put_page_testzero(struct page *page)
262{
Nick Piggin725d7042006-09-25 23:30:55 -0700263 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800264 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800265}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
267/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800268 * Try to grab a ref unless the page has a refcount of zero, return false if
269 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800271static inline int get_page_unless_zero(struct page *page)
272{
Christoph Lameteraec2c3e2008-02-04 22:28:35 -0800273 VM_BUG_ON(PageTail(page));
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800274 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800275}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
Christoph Lameter48667e72008-02-04 22:28:31 -0800277/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800278struct page *vmalloc_to_page(const void *addr);
279unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800280
Paul Mundt0738c4b2008-03-12 16:51:31 +0900281/*
282 * Determine if an address is within the vmalloc range
283 *
284 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
285 * is no special casing required.
286 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800287static inline int is_vmalloc_addr(const void *x)
288{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900289#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800290 unsigned long addr = (unsigned long)x;
291
292 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900293#else
294 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800295#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900296}
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800297
Christoph Lameterd85f3382007-05-06 14:49:39 -0700298static inline struct page *compound_head(struct page *page)
299{
Christoph Lameter6d777952007-05-06 14:49:40 -0700300 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700301 return page->first_page;
302 return page;
303}
304
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700305static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700307 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308}
309
310static inline void get_page(struct page *page)
311{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700312 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700313 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 atomic_inc(&page->_count);
315}
316
Christoph Lameterb49af682007-05-06 14:49:41 -0700317static inline struct page *virt_to_head_page(const void *x)
318{
319 struct page *page = virt_to_page(x);
320 return compound_head(page);
321}
322
Nick Piggin7835e982006-03-22 00:08:40 -0800323/*
324 * Setup the page count before being freed into the page allocator for
325 * the first time (boot or memory hotplug)
326 */
327static inline void init_page_count(struct page *page)
328{
329 atomic_set(&page->_count, 1);
330}
331
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700333void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800335void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800336
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800338 * Compound pages have a destructor function. Provide a
339 * prototype for that function and accessor functions.
340 * These are _only_ valid on the head of a PG_compound page.
341 */
342typedef void compound_page_dtor(struct page *);
343
344static inline void set_compound_page_dtor(struct page *page,
345 compound_page_dtor *dtor)
346{
347 page[1].lru.next = (void *)dtor;
348}
349
350static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
351{
352 return (compound_page_dtor *)page[1].lru.next;
353}
354
Christoph Lameterd85f3382007-05-06 14:49:39 -0700355static inline int compound_order(struct page *page)
356{
Christoph Lameter6d777952007-05-06 14:49:40 -0700357 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700358 return 0;
359 return (unsigned long)page[1].lru.prev;
360}
361
362static inline void set_compound_order(struct page *page, unsigned long order)
363{
364 page[1].lru.prev = (void *)order;
365}
366
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800367/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 * Multiple processes may "see" the same page. E.g. for untouched
369 * mappings of /dev/null, all processes see the same page full of
370 * zeroes, and text pages of executables and shared libraries have
371 * only one copy in memory, at most, normally.
372 *
373 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700374 * page_count() == 0 means the page is free. page->lru is then used for
375 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700376 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700378 * Pages are allocated by the slab allocator in order to provide memory
379 * to kmalloc and kmem_cache_alloc. In this case, the management of the
380 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
381 * unless a particular usage is carefully commented. (the responsibility of
382 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700384 * A page may be used by anyone else who does a __get_free_page().
385 * In this case, page_count still tracks the references, and should only
386 * be used through the normal accessor functions. The top bits of page->flags
387 * and page->virtual store page management information, but all other fields
388 * are unused and could be used privately, carefully. The management of this
389 * page is the responsibility of the one who allocated it, and those who have
390 * subsequently been given references to it.
391 *
392 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 * managed by the Linux memory manager: I/O, buffers, swapping etc.
394 * The following discussion applies only to them.
395 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700396 * A pagecache page contains an opaque `private' member, which belongs to the
397 * page's address_space. Usually, this is the address of a circular list of
398 * the page's disk buffers. PG_private must be set to tell the VM to call
399 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700401 * A page may belong to an inode's memory mapping. In this case, page->mapping
402 * is the pointer to the inode, and page->index is the file offset of the page,
403 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700405 * If pagecache pages are not associated with an inode, they are said to be
406 * anonymous pages. These may become associated with the swapcache, and in that
407 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700409 * In either case (swapcache or inode backed), the pagecache itself holds one
410 * reference to the page. Setting PG_private should also increment the
411 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700413 * The pagecache pages are stored in a per-mapping radix tree, which is
414 * rooted at mapping->page_tree, and indexed by offset.
415 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
416 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700418 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 * - inode pages may need to be read from disk,
420 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700421 * to be written back to the inode on disk,
422 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
423 * modified may need to be swapped out to swap space and (later) to be read
424 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 */
426
427/*
428 * The zone field is never updated after free_area_init_core()
429 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700431
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700432
433/*
434 * page->flags layout:
435 *
436 * There are three possibilities for how page->flags get
437 * laid out. The first is for the normal case, without
438 * sparsemem. The second is for sparsemem when there is
439 * plenty of space for node and section. The last is when
440 * we have run out of space and have to fall back to an
441 * alternate (slower) way of determining the node.
442 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700443 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
444 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
445 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700446 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700447#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700448#define SECTIONS_WIDTH SECTIONS_SHIFT
449#else
450#define SECTIONS_WIDTH 0
451#endif
452
453#define ZONES_WIDTH ZONES_SHIFT
454
Christoph Lameter9223b412008-04-28 02:12:48 -0700455#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700456#define NODES_WIDTH NODES_SHIFT
457#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700458#ifdef CONFIG_SPARSEMEM_VMEMMAP
459#error "Vmemmap: No space for nodes field in page flags"
460#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700461#define NODES_WIDTH 0
462#endif
463
464/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100465#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700466#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
467#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
468
469/*
470 * We are going to use the flags for the page to node mapping if its in
471 * there. This includes the case where there is no node, so it is implicit.
472 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800473#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
474#define NODE_NOT_IN_PAGE_FLAGS
475#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700476
477#ifndef PFN_SECTION_SHIFT
478#define PFN_SECTION_SHIFT 0
479#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700480
481/*
482 * Define the bit shifts to access each section. For non-existant
483 * sections we define the shift as 0; that plus a 0 mask ensures
484 * the compiler will optimise away reference to them.
485 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700486#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
487#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
488#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700489
Christoph Lameter89689ae2006-12-06 20:31:45 -0800490/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
491#ifdef NODE_NOT_IN_PAGEFLAGS
492#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800493#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
494 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700495#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800496#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800497#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
498 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700499#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800500
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800501#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700502
Christoph Lameter9223b412008-04-28 02:12:48 -0700503#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
504#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700505#endif
506
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700507#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
508#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
509#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800510#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700511
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700512static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513{
Dave Hansen348f8b62005-06-23 00:07:40 -0700514 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516
Christoph Lameter89689ae2006-12-06 20:31:45 -0800517/*
518 * The identification function is only used by the buddy allocator for
519 * determining if two pages could be buddies. We are not really
520 * identifying a zone since we could be using a the section number
521 * id if we have not node id available in page flags.
522 * We guarantee only that it will return the same value for two
523 * combinable pages in a zone.
524 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700525static inline int page_zone_id(struct page *page)
526{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800527 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528}
529
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800530static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700531{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700532#ifdef CONFIG_NUMA
533 return zone->node;
534#else
535 return 0;
536#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700537}
538
Christoph Lameter89689ae2006-12-06 20:31:45 -0800539#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800540extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800541#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800542static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700543{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800544 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700545}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800546#endif
547
548static inline struct zone *page_zone(struct page *page)
549{
550 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
551}
552
Christoph Lameter308c05e2008-04-28 02:12:43 -0700553#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700554static inline unsigned long page_to_section(struct page *page)
555{
556 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
557}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700558#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700559
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700560static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561{
Dave Hansen348f8b62005-06-23 00:07:40 -0700562 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
563 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
564}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700565
Dave Hansen348f8b62005-06-23 00:07:40 -0700566static inline void set_page_node(struct page *page, unsigned long node)
567{
568 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
569 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
570}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800571
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700572static inline void set_page_section(struct page *page, unsigned long section)
573{
574 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
575 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
576}
Dave Hansen348f8b62005-06-23 00:07:40 -0700577
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700578static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700579 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700580{
581 set_page_zone(page, zone);
582 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700583 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584}
585
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700586/*
Eric Paris7cd94142007-11-26 18:47:40 -0500587 * If a hint addr is less than mmap_min_addr change hint to be as
588 * low as possible but still greater than mmap_min_addr
589 */
590static inline unsigned long round_hint_to_min(unsigned long hint)
591{
592#ifdef CONFIG_SECURITY
593 hint &= PAGE_MASK;
594 if (((void *)hint != NULL) &&
595 (hint < mmap_min_addr))
596 return PAGE_ALIGN(mmap_min_addr);
597#endif
598 return hint;
599}
600
601/*
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700602 * Some inline functions in vmstat.h depend on page_zone()
603 */
604#include <linux/vmstat.h>
605
Ingo Molnar652050a2006-01-14 13:21:30 -0800606static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607{
608 return __va(page_to_pfn(page) << PAGE_SHIFT);
609}
610
611#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
612#define HASHED_PAGE_VIRTUAL
613#endif
614
615#if defined(WANT_PAGE_VIRTUAL)
616#define page_address(page) ((page)->virtual)
617#define set_page_address(page, address) \
618 do { \
619 (page)->virtual = (address); \
620 } while(0)
621#define page_address_init() do { } while(0)
622#endif
623
624#if defined(HASHED_PAGE_VIRTUAL)
625void *page_address(struct page *page);
626void set_page_address(struct page *page, void *virtual);
627void page_address_init(void);
628#endif
629
630#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
631#define page_address(page) lowmem_page_address(page)
632#define set_page_address(page, address) do { } while(0)
633#define page_address_init() do { } while(0)
634#endif
635
636/*
637 * On an anonymous page mapped into a user virtual memory area,
638 * page->mapping points to its anon_vma, not to a struct address_space;
639 * with the PAGE_MAPPING_ANON bit set to distinguish it.
640 *
641 * Please note that, confusingly, "page_mapping" refers to the inode
642 * address_space which maps the page from disk; whereas "page_mapped"
643 * refers to user virtual address space into which the page is mapped.
644 */
645#define PAGE_MAPPING_ANON 1
646
647extern struct address_space swapper_space;
648static inline struct address_space *page_mapping(struct page *page)
649{
650 struct address_space *mapping = page->mapping;
651
Christoph Lameterb5fab142007-07-17 04:03:33 -0700652 VM_BUG_ON(PageSlab(page));
Andrew Morton726b8012008-04-28 02:12:44 -0700653#ifdef CONFIG_SWAP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 if (unlikely(PageSwapCache(page)))
655 mapping = &swapper_space;
Andrew Morton726b8012008-04-28 02:12:44 -0700656 else
657#endif
658 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 mapping = NULL;
660 return mapping;
661}
662
663static inline int PageAnon(struct page *page)
664{
665 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
666}
667
668/*
669 * Return the pagecache index of the passed page. Regular pagecache pages
670 * use ->index whereas swapcache pages use ->private
671 */
672static inline pgoff_t page_index(struct page *page)
673{
674 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700675 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 return page->index;
677}
678
679/*
680 * The atomic page->_mapcount, like _count, starts from -1:
681 * so that transitions both from it and to it can be tracked,
682 * using atomic_inc_and_test and atomic_add_negative(-1).
683 */
684static inline void reset_page_mapcount(struct page *page)
685{
686 atomic_set(&(page)->_mapcount, -1);
687}
688
689static inline int page_mapcount(struct page *page)
690{
691 return atomic_read(&(page)->_mapcount) + 1;
692}
693
694/*
695 * Return true if this page is mapped into pagetables.
696 */
697static inline int page_mapped(struct page *page)
698{
699 return atomic_read(&(page)->_mapcount) >= 0;
700}
701
702/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 * Different kinds of faults, as returned by handle_mm_fault().
704 * Used to decide whether a process gets delivered SIGBUS or
705 * just gets major/minor fault counters bumped up.
706 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700707
Nick Piggin83c54072007-07-19 01:47:05 -0700708#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700709
Nick Piggin83c54072007-07-19 01:47:05 -0700710#define VM_FAULT_OOM 0x0001
711#define VM_FAULT_SIGBUS 0x0002
712#define VM_FAULT_MAJOR 0x0004
713#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000714
Nick Piggin83c54072007-07-19 01:47:05 -0700715#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
716#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
Nick Piggin83c54072007-07-19 01:47:05 -0700718#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
Nick Piggind0217ac2007-07-19 01:47:03 -0700719
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
721
722extern void show_free_areas(void);
723
724#ifdef CONFIG_SHMEM
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700725extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726#else
Russell King03b00eb2006-01-06 00:10:52 -0800727static inline int shmem_lock(struct file *file, int lock,
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700728 struct user_struct *user)
Russell King03b00eb2006-01-06 00:10:52 -0800729{
730 return 0;
731}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732#endif
733struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
734
735int shmem_zero_setup(struct vm_area_struct *);
736
David Howellsb0e15192006-01-06 00:11:42 -0800737#ifndef CONFIG_MMU
738extern unsigned long shmem_get_unmapped_area(struct file *file,
739 unsigned long addr,
740 unsigned long len,
741 unsigned long pgoff,
742 unsigned long flags);
743#endif
744
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400745extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746extern int user_shm_lock(size_t, struct user_struct *);
747extern void user_shm_unlock(size_t, struct user_struct *);
748
749/*
750 * Parameter block passed down to zap_pte_range in exceptional cases.
751 */
752struct zap_details {
753 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
754 struct address_space *check_mapping; /* Check page->mapping if set */
755 pgoff_t first_index; /* Lowest page->index to unmap */
756 pgoff_t last_index; /* Highest page->index to unmap */
757 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 unsigned long truncate_count; /* Compare vm_truncate_count */
759};
760
Nick Piggin7e675132008-04-28 02:13:00 -0700761struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
762 pte_t pte);
763
Jack Steinerc627f9c2008-07-29 22:33:53 -0700764int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
765 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700766unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700768unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 struct vm_area_struct *start_vma, unsigned long start_addr,
770 unsigned long end_addr, unsigned long *nr_accounted,
771 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800772
773/**
774 * mm_walk - callbacks for walk_page_range
775 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
776 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
777 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
778 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
779 * @pte_hole: if set, called for each hole at all levels
780 *
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 *);
789 struct mm_struct *mm;
790 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800791};
792
Dave Hansen21650092008-06-12 15:21:47 -0700793int walk_page_range(unsigned long addr, unsigned long end,
794 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700795void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700796 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
798 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799void unmap_mapping_range(struct address_space *mapping,
800 loff_t const holebegin, loff_t const holelen, int even_cows);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800801int follow_phys(struct vm_area_struct *vma, unsigned long address,
802 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700803int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
804 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805
806static inline void unmap_shared_mapping_range(struct address_space *mapping,
807 loff_t const holebegin, loff_t const holelen)
808{
809 unmap_mapping_range(mapping, holebegin, holelen, 0);
810}
811
812extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800813extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000814
David Howells7ee1dd32006-01-06 00:11:44 -0800815#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700816extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
David Howells7ee1dd32006-01-06 00:11:44 -0800817 unsigned long address, int write_access);
David Howells7ee1dd32006-01-06 00:11:44 -0800818#else
819static inline int handle_mm_fault(struct mm_struct *mm,
820 struct vm_area_struct *vma, unsigned long address,
821 int write_access)
822{
823 /* should never happen if there's no MMU */
824 BUG();
825 return VM_FAULT_SIGBUS;
826}
827#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000828
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829extern int make_pages_present(unsigned long addr, unsigned long end);
830extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
832int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
833 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
834
David Howellscf9a2ae2006-08-29 19:05:54 +0100835extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
836extern void do_invalidatepage(struct page *page, unsigned long offset);
837
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800839int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840int redirty_page_for_writepage(struct writeback_control *wbc,
841 struct page *page);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800842int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843int set_page_dirty_lock(struct page *page);
844int clear_page_dirty_for_io(struct page *page);
845
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700846extern unsigned long move_page_tables(struct vm_area_struct *vma,
847 unsigned long old_addr, struct vm_area_struct *new_vma,
848 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849extern unsigned long do_mremap(unsigned long addr,
850 unsigned long old_len, unsigned long new_len,
851 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700852extern int mprotect_fixup(struct vm_area_struct *vma,
853 struct vm_area_struct **pprev, unsigned long start,
854 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
Nick Piggin21cc1992008-07-25 19:45:22 -0700856/*
857 * get_user_pages_fast provides equivalent functionality to get_user_pages,
858 * operating on current and current->mm (force=0 and doesn't return any vmas).
859 *
860 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
861 * can be made about locking. get_user_pages_fast is to be implemented in a
862 * way that is advantageous (vs get_user_pages()) when the user memory area is
863 * already faulted in and present in ptes. However if the pages have to be
864 * faulted in, it may turn out to be slightly slower).
865 */
866int get_user_pages_fast(unsigned long start, int nr_pages, int write,
867 struct page **pages);
868
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700870 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700872 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
873 * look through the least-recently-used 'nr_to_scan' entries and
874 * attempt to free them up. It should return the number of objects
875 * which remain in the cache. If it returns -1, it means it cannot do
876 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700878 * The 'gfpmask' refers to the allocation we are currently trying to
879 * fulfil.
880 *
881 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
882 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 */
Rusty Russell8e1f9362007-07-17 04:03:17 -0700884struct shrinker {
885 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
886 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
Rusty Russell8e1f9362007-07-17 04:03:17 -0700888 /* These are for internal use */
889 struct list_head list;
890 long nr; /* objs pending delete */
891};
892#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
893extern void register_shrinker(struct shrinker *);
894extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895
Alexey Dobriyan4e950f62007-07-30 02:36:13 +0400896int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700897
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800898extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800899
Nick Piggin5f22df02007-05-06 14:49:02 -0700900#ifdef __PAGETABLE_PUD_FOLDED
901static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
902 unsigned long address)
903{
904 return 0;
905}
906#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700907int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700908#endif
909
910#ifdef __PAGETABLE_PMD_FOLDED
911static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
912 unsigned long address)
913{
914 return 0;
915}
916#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700917int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700918#endif
919
Hugh Dickins1bb36302005-10-29 18:16:22 -0700920int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
921int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
922
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923/*
924 * The following ifdef needed to get the 4level-fixup.h header to work.
925 * Remove it when 4level-fixup.h has been removed.
926 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700927#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
929{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700930 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
931 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932}
933
934static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
935{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700936 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
937 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700939#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
940
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700941#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700942/*
943 * We tuck a spinlock to guard each pagetable page into its struct page,
944 * at page->private, with BUILD_BUG_ON to make sure that this will not
945 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
946 * When freeing, reset page->mapping so free_pages_check won't complain.
947 */
Andrew Morton349aef02006-01-08 01:04:36 -0800948#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700949#define pte_lock_init(_page) do { \
950 spin_lock_init(__pte_lockptr(_page)); \
951} while (0)
952#define pte_lock_deinit(page) ((page)->mapping = NULL)
953#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700954#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700955/*
956 * We use mm->page_table_lock to guard all pagetable pages of the mm.
957 */
958#define pte_lock_init(page) do {} while (0)
959#define pte_lock_deinit(page) do {} while (0)
960#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700961#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700962
Martin Schwidefsky2f569af2008-02-08 04:22:04 -0800963static inline void pgtable_page_ctor(struct page *page)
964{
965 pte_lock_init(page);
966 inc_zone_page_state(page, NR_PAGETABLE);
967}
968
969static inline void pgtable_page_dtor(struct page *page)
970{
971 pte_lock_deinit(page);
972 dec_zone_page_state(page, NR_PAGETABLE);
973}
974
Hugh Dickinsc74df322005-10-29 18:16:23 -0700975#define pte_offset_map_lock(mm, pmd, address, ptlp) \
976({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700977 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700978 pte_t *__pte = pte_offset_map(pmd, address); \
979 *(ptlp) = __ptl; \
980 spin_lock(__ptl); \
981 __pte; \
982})
983
984#define pte_unmap_unlock(pte, ptl) do { \
985 spin_unlock(ptl); \
986 pte_unmap(pte); \
987} while (0)
988
Hugh Dickins1bb36302005-10-29 18:16:22 -0700989#define pte_alloc_map(mm, pmd, address) \
990 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
991 NULL: pte_offset_map(pmd, address))
992
Hugh Dickinsc74df322005-10-29 18:16:23 -0700993#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
994 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
995 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
996
Hugh Dickins1bb36302005-10-29 18:16:22 -0700997#define pte_alloc_kernel(pmd, address) \
998 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
999 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000
1001extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001002extern void free_area_init_node(int nid, unsigned long * zones_size,
1003 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -07001004#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1005/*
1006 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1007 * zones, allocate the backing mem_map and account for memory holes in a more
1008 * architecture independent manner. This is a substitute for creating the
1009 * zone_sizes[] and zholes_size[] arrays and passing them to
1010 * free_area_init_node()
1011 *
1012 * An architecture is expected to register range of page frames backed by
1013 * physical memory with add_active_range() before calling
1014 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1015 * usage, an architecture is expected to do something like
1016 *
1017 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1018 * max_highmem_pfn};
1019 * for_each_valid_physical_page_range()
1020 * add_active_range(node_id, start_pfn, end_pfn)
1021 * free_area_init_nodes(max_zone_pfns);
1022 *
1023 * If the architecture guarantees that there are no holes in the ranges
1024 * registered with add_active_range(), free_bootmem_active_regions()
1025 * will call free_bootmem_node() for each registered physical page range.
1026 * Similarly sparse_memory_present_with_active_regions() calls
1027 * memory_present() for each range when SPARSEMEM is enabled.
1028 *
1029 * See mm/page_alloc.c for more information on each function exposed by
1030 * CONFIG_ARCH_POPULATES_NODE_MAP
1031 */
1032extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1033extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1034 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001035extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1036 unsigned long end_pfn);
Mel Gormanfb014392006-09-27 01:49:59 -07001037extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1038 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001039extern void remove_all_active_ranges(void);
1040extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1041 unsigned long end_pfn);
1042extern void get_pfn_range_for_nid(unsigned int nid,
1043 unsigned long *start_pfn, unsigned long *end_pfn);
1044extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001045extern void free_bootmem_with_active_regions(int nid,
1046 unsigned long max_low_pfn);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001047typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001048extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001049extern void sparse_memory_present_with_active_regions(int nid);
1050#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1051extern int early_pfn_to_nid(unsigned long pfn);
1052#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1053#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
Mel Gorman0e0b8642006-09-27 01:49:56 -07001054extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa02007-01-10 23:15:30 -08001055extern void memmap_init_zone(unsigned long, int, unsigned long,
1056 unsigned long, enum memmap_context);
Dave Hansen3947be12005-10-29 18:16:54 -07001057extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058extern void mem_init(void);
1059extern 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
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070/* prio_tree.c */
1071void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1072void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1073void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1074struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1075 struct prio_tree_iter *iter);
1076
1077#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1078 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1079 (vma = vma_prio_tree_next(vma, iter)); )
1080
1081static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1082 struct list_head *list)
1083{
1084 vma->shared.vm_set.parent = NULL;
1085 list_add_tail(&vma->shared.vm_set.list, list);
1086}
1087
1088/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001089extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1091 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1092extern struct vm_area_struct *vma_merge(struct mm_struct *,
1093 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1094 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1095 struct mempolicy *);
1096extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1097extern int split_vma(struct mm_struct *,
1098 struct vm_area_struct *, unsigned long addr, int new_below);
1099extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1100extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1101 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001102extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1104 unsigned long addr, unsigned long len, pgoff_t pgoff);
1105extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001106
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001107extern int mm_take_all_locks(struct mm_struct *mm);
1108extern void mm_drop_all_locks(struct mm_struct *mm);
1109
Matt Helsley925d1c42008-04-29 01:01:36 -07001110#ifdef CONFIG_PROC_FS
1111/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1112extern void added_exe_file_vma(struct mm_struct *mm);
1113extern void removed_exe_file_vma(struct mm_struct *mm);
1114#else
1115static inline void added_exe_file_vma(struct mm_struct *mm)
1116{}
1117
1118static inline void removed_exe_file_vma(struct mm_struct *mm)
1119{}
1120#endif /* CONFIG_PROC_FS */
1121
akpm@osdl.org119f6572005-05-01 08:58:35 -07001122extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001123extern int install_special_mapping(struct mm_struct *mm,
1124 unsigned long addr, unsigned long len,
1125 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
1127extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1128
1129extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1130 unsigned long len, unsigned long prot,
1131 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001132extern unsigned long mmap_region(struct file *file, unsigned long addr,
1133 unsigned long len, unsigned long flags,
1134 unsigned int vm_flags, unsigned long pgoff,
1135 int accountable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
1137static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1138 unsigned long len, unsigned long prot,
1139 unsigned long flag, unsigned long offset)
1140{
1141 unsigned long ret = -EINVAL;
1142 if ((offset + PAGE_ALIGN(len)) < offset)
1143 goto out;
1144 if (!(offset & ~PAGE_MASK))
1145 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1146out:
1147 return ret;
1148}
1149
1150extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1151
1152extern unsigned long do_brk(unsigned long, unsigned long);
1153
1154/* filemap.c */
1155extern unsigned long page_unuse(struct page *);
1156extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001157extern void truncate_inode_pages_range(struct address_space *,
1158 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
1160/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001161extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
1163/* mm/page-writeback.c */
1164int write_one_page(struct page *page, int wait);
1165
1166/* readahead.c */
1167#define VM_MAX_READAHEAD 128 /* kbytes */
1168#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169
1170int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001171 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001173 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001174
1175void page_cache_sync_readahead(struct address_space *mapping,
1176 struct file_ra_state *ra,
1177 struct file *filp,
1178 pgoff_t offset,
1179 unsigned long size);
1180
1181void page_cache_async_readahead(struct address_space *mapping,
1182 struct file_ra_state *ra,
1183 struct file *filp,
1184 struct page *pg,
1185 pgoff_t offset,
1186 unsigned long size);
1187
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188unsigned long max_sane_readahead(unsigned long nr);
1189
1190/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001191extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001192#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001193extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001194#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001195extern int expand_stack_downwards(struct vm_area_struct *vma,
1196 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197
1198/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1199extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1200extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1201 struct vm_area_struct **pprev);
1202
1203/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1204 NULL if none. Assume start_addr < end_addr. */
1205static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1206{
1207 struct vm_area_struct * vma = find_vma(mm,start_addr);
1208
1209 if (vma && end_addr <= vma->vm_start)
1210 vma = NULL;
1211 return vma;
1212}
1213
1214static inline unsigned long vma_pages(struct vm_area_struct *vma)
1215{
1216 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1217}
1218
Hugh Dickins804af2c2006-07-26 21:39:49 +01001219pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001220struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001221int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1222 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001223int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001224int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1225 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001226int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1227 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228
Linus Torvalds6aab3412005-11-28 14:34:23 -08001229struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001230 unsigned int foll_flags);
1231#define FOLL_WRITE 0x01 /* check pte is writable */
1232#define FOLL_TOUCH 0x02 /* mark page accessed */
1233#define FOLL_GET 0x04 /* do get_page on page */
1234#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001236typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001237 void *data);
1238extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1239 unsigned long size, pte_fn_t fn, void *data);
1240
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001242void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001244static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 unsigned long flags, struct file *file, long pages)
1246{
1247}
1248#endif /* CONFIG_PROC_FS */
1249
Ingo Molnar12d6f212008-01-30 13:33:58 +01001250#ifdef CONFIG_DEBUG_PAGEALLOC
1251extern int debug_pagealloc_enabled;
1252
1253extern void kernel_map_pages(struct page *page, int numpages, int enable);
1254
1255static inline void enable_debug_pagealloc(void)
1256{
1257 debug_pagealloc_enabled = 1;
1258}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001259#ifdef CONFIG_HIBERNATION
1260extern bool kernel_page_present(struct page *page);
1261#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001262#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001264kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001265static inline void enable_debug_pagealloc(void)
1266{
1267}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001268#ifdef CONFIG_HIBERNATION
1269static inline bool kernel_page_present(struct page *page) { return true; }
1270#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271#endif
1272
1273extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1274#ifdef __HAVE_ARCH_GATE_AREA
1275int in_gate_area_no_task(unsigned long addr);
1276int in_gate_area(struct task_struct *task, unsigned long addr);
1277#else
1278int in_gate_area_no_task(unsigned long addr);
1279#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1280#endif /* __HAVE_ARCH_GATE_AREA */
1281
Andrew Morton9d0243b2006-01-08 01:00:39 -08001282int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1283 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001284unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001285 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001286
Luke Yang7a9166e2006-02-20 18:28:07 -08001287#ifndef CONFIG_MMU
1288#define randomize_va_space 0
1289#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001290extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001291#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001292
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001293const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001294void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001295
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001296struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001297pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1298pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1299pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1300pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001301void *vmemmap_alloc_block(unsigned long size, int node);
1302void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001303int vmemmap_populate_basepages(struct page *start_page,
1304 unsigned long pages, int node);
1305int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001306void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001307
Markus Metzgerc5dee612008-12-19 15:17:02 +01001308extern void *alloc_locked_buffer(size_t size);
1309extern void free_locked_buffer(void *buffer, size_t size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310#endif /* __KERNEL__ */
1311#endif /* _LINUX_MM_H */