blob: 1897ca223eca99c402523994c04e64287b769cbd [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;
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
22#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
23extern unsigned long max_mapnr;
24#endif
25
26extern unsigned long num_physpages;
27extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070028extern int page_cluster;
29
30#ifdef CONFIG_SYSCTL
31extern int sysctl_legacy_va_layout;
32#else
33#define sysctl_legacy_va_layout 0
34#endif
35
James Morris42d78962007-12-26 15:12:37 +110036extern unsigned long mmap_min_addr;
37
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#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
44/*
45 * Linux kernel virtual memory manager primitives.
46 * The idea being to have a "virtual" mm in the same way
47 * we have a virtual fs - giving a cleaner interface to the
48 * mm details, and allowing different kinds of memory mappings
49 * (from shared memory to executable loading to arbitrary
50 * mmap() functions).
51 */
52
Christoph Lameterc43692e2006-12-06 20:32:48 -080053extern struct kmem_cache *vm_area_cachep;
54
Linus Torvalds1da177e2005-04-16 15:20:36 -070055/*
56 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
57 * disabled, then there's a single shared list of VMAs maintained by the
58 * system, and mm's subscribe to these individually
59 */
60struct vm_list_struct {
61 struct vm_list_struct *next;
62 struct vm_area_struct *vma;
63};
64
65#ifndef CONFIG_MMU
66extern struct rb_root nommu_vma_tree;
67extern struct rw_semaphore nommu_vma_sem;
68
69extern unsigned int kobjsize(const void *objp);
70#endif
71
72/*
73 * vm_flags..
74 */
75#define VM_READ 0x00000001 /* currently active flags */
76#define VM_WRITE 0x00000002
77#define VM_EXEC 0x00000004
78#define VM_SHARED 0x00000008
79
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070080/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
82#define VM_MAYWRITE 0x00000020
83#define VM_MAYEXEC 0x00000040
84#define VM_MAYSHARE 0x00000080
85
86#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
87#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -080088#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070089#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
90
91#define VM_EXECUTABLE 0x00001000
92#define VM_LOCKED 0x00002000
93#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
94
95 /* Used by sys_madvise() */
96#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
98
99#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800101#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
103#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
104#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
105#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800106#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800107#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108
Nick Piggind0217ac2007-07-19 01:47:03 -0700109#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700110
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
112#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
113#endif
114
115#ifdef CONFIG_STACK_GROWSUP
116#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
117#else
118#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
119#endif
120
121#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
122#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
123#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
124#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
125#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
126
127/*
128 * mapping from the currently active vm_flags protection bits (the
129 * low four bits) to a page protection mask..
130 */
131extern pgprot_t protection_map[16];
132
Nick Piggind0217ac2007-07-19 01:47:03 -0700133#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
134#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
135
136
Nick Piggin54cb8822007-07-19 01:46:59 -0700137/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700138 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700139 * ->fault function. The vma's ->fault is responsible for returning a bitmask
140 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700141 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700142 * pgoff should be used in favour of virtual_address, if possible. If pgoff
143 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
144 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700145 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700146struct vm_fault {
147 unsigned int flags; /* FAULT_FLAG_xxx flags */
148 pgoff_t pgoff; /* Logical page offset based on vma */
149 void __user *virtual_address; /* Faulting virtual address */
150
151 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700152 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700153 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700154 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700155 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700156};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
158/*
159 * These are the virtual MM functions - opening of an area, closing and
160 * unmapping it (needed to keep files on disk up-to-date etc), pointer
161 * to the functions called when a no-page or a wp-page exception occurs.
162 */
163struct vm_operations_struct {
164 void (*open)(struct vm_area_struct * area);
165 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700166 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
Nick Piggin54cb8822007-07-19 01:46:59 -0700167 struct page *(*nopage)(struct vm_area_struct *area,
168 unsigned long address, int *type);
169 unsigned long (*nopfn)(struct vm_area_struct *area,
170 unsigned long address);
David Howells9637a5e2006-06-23 02:03:43 -0700171
172 /* notification that a previously read-only page is about to become
173 * writable, if an error is returned it will cause a SIGBUS */
174 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175#ifdef CONFIG_NUMA
176 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
177 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
178 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700179 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
180 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181#endif
182};
183
184struct mmu_gather;
185struct inode;
186
Andrew Morton349aef02006-01-08 01:04:36 -0800187#define page_private(page) ((page)->private)
188#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190/*
191 * FIXME: take this include out, include page-flags.h in
192 * files which need it (119 of them)
193 */
194#include <linux/page-flags.h>
195
Nick Piggin725d7042006-09-25 23:30:55 -0700196#ifdef CONFIG_DEBUG_VM
197#define VM_BUG_ON(cond) BUG_ON(cond)
198#else
199#define VM_BUG_ON(condition) do { } while(0)
200#endif
201
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202/*
203 * Methods to modify the page usage count.
204 *
205 * What counts for a page usage:
206 * - cache mapping (page->mapping)
207 * - private data (page->private)
208 * - page mapped in a task's page tables, each mapping
209 * is counted separately
210 *
211 * Also, many kernel routines increase the page count before a critical
212 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 */
214
215/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700216 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800218static inline int put_page_testzero(struct page *page)
219{
Nick Piggin725d7042006-09-25 23:30:55 -0700220 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800221 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800222}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
224/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800225 * Try to grab a ref unless the page has a refcount of zero, return false if
226 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800228static inline int get_page_unless_zero(struct page *page)
229{
Nick Piggin725d7042006-09-25 23:30:55 -0700230 VM_BUG_ON(PageCompound(page));
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800231 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800232}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
Christoph Lameterd85f3382007-05-06 14:49:39 -0700234static inline struct page *compound_head(struct page *page)
235{
Christoph Lameter6d777952007-05-06 14:49:40 -0700236 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700237 return page->first_page;
238 return page;
239}
240
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700241static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700243 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244}
245
246static inline void get_page(struct page *page)
247{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700248 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700249 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 atomic_inc(&page->_count);
251}
252
Christoph Lameterb49af682007-05-06 14:49:41 -0700253static inline struct page *virt_to_head_page(const void *x)
254{
255 struct page *page = virt_to_page(x);
256 return compound_head(page);
257}
258
Nick Piggin7835e982006-03-22 00:08:40 -0800259/*
260 * Setup the page count before being freed into the page allocator for
261 * the first time (boot or memory hotplug)
262 */
263static inline void init_page_count(struct page *page)
264{
265 atomic_set(&page->_count, 1);
266}
267
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700269void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800271void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800274 * Compound pages have a destructor function. Provide a
275 * prototype for that function and accessor functions.
276 * These are _only_ valid on the head of a PG_compound page.
277 */
278typedef void compound_page_dtor(struct page *);
279
280static inline void set_compound_page_dtor(struct page *page,
281 compound_page_dtor *dtor)
282{
283 page[1].lru.next = (void *)dtor;
284}
285
286static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
287{
288 return (compound_page_dtor *)page[1].lru.next;
289}
290
Christoph Lameterd85f3382007-05-06 14:49:39 -0700291static inline int compound_order(struct page *page)
292{
Christoph Lameter6d777952007-05-06 14:49:40 -0700293 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700294 return 0;
295 return (unsigned long)page[1].lru.prev;
296}
297
298static inline void set_compound_order(struct page *page, unsigned long order)
299{
300 page[1].lru.prev = (void *)order;
301}
302
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800303/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 * Multiple processes may "see" the same page. E.g. for untouched
305 * mappings of /dev/null, all processes see the same page full of
306 * zeroes, and text pages of executables and shared libraries have
307 * only one copy in memory, at most, normally.
308 *
309 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700310 * page_count() == 0 means the page is free. page->lru is then used for
311 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700312 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700314 * Pages are allocated by the slab allocator in order to provide memory
315 * to kmalloc and kmem_cache_alloc. In this case, the management of the
316 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
317 * unless a particular usage is carefully commented. (the responsibility of
318 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700320 * A page may be used by anyone else who does a __get_free_page().
321 * In this case, page_count still tracks the references, and should only
322 * be used through the normal accessor functions. The top bits of page->flags
323 * and page->virtual store page management information, but all other fields
324 * are unused and could be used privately, carefully. The management of this
325 * page is the responsibility of the one who allocated it, and those who have
326 * subsequently been given references to it.
327 *
328 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 * managed by the Linux memory manager: I/O, buffers, swapping etc.
330 * The following discussion applies only to them.
331 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700332 * A pagecache page contains an opaque `private' member, which belongs to the
333 * page's address_space. Usually, this is the address of a circular list of
334 * the page's disk buffers. PG_private must be set to tell the VM to call
335 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700337 * A page may belong to an inode's memory mapping. In this case, page->mapping
338 * is the pointer to the inode, and page->index is the file offset of the page,
339 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700341 * If pagecache pages are not associated with an inode, they are said to be
342 * anonymous pages. These may become associated with the swapcache, and in that
343 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700345 * In either case (swapcache or inode backed), the pagecache itself holds one
346 * reference to the page. Setting PG_private should also increment the
347 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700349 * The pagecache pages are stored in a per-mapping radix tree, which is
350 * rooted at mapping->page_tree, and indexed by offset.
351 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
352 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700354 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 * - inode pages may need to be read from disk,
356 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700357 * to be written back to the inode on disk,
358 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
359 * modified may need to be swapped out to swap space and (later) to be read
360 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 */
362
363/*
364 * The zone field is never updated after free_area_init_core()
365 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700367
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700368
369/*
370 * page->flags layout:
371 *
372 * There are three possibilities for how page->flags get
373 * laid out. The first is for the normal case, without
374 * sparsemem. The second is for sparsemem when there is
375 * plenty of space for node and section. The last is when
376 * we have run out of space and have to fall back to an
377 * alternate (slower) way of determining the node.
378 *
379 * No sparsemem: | NODE | ZONE | ... | FLAGS |
380 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
381 * no space for node: | SECTION | ZONE | ... | FLAGS |
382 */
383#ifdef CONFIG_SPARSEMEM
384#define SECTIONS_WIDTH SECTIONS_SHIFT
385#else
386#define SECTIONS_WIDTH 0
387#endif
388
389#define ZONES_WIDTH ZONES_SHIFT
390
391#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
392#define NODES_WIDTH NODES_SHIFT
393#else
394#define NODES_WIDTH 0
395#endif
396
397/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100398#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700399#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
400#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
401
402/*
403 * We are going to use the flags for the page to node mapping if its in
404 * there. This includes the case where there is no node, so it is implicit.
405 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800406#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
407#define NODE_NOT_IN_PAGE_FLAGS
408#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700409
410#ifndef PFN_SECTION_SHIFT
411#define PFN_SECTION_SHIFT 0
412#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700413
414/*
415 * Define the bit shifts to access each section. For non-existant
416 * sections we define the shift as 0; that plus a 0 mask ensures
417 * the compiler will optimise away reference to them.
418 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700419#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
420#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
421#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700422
Christoph Lameter89689ae2006-12-06 20:31:45 -0800423/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
424#ifdef NODE_NOT_IN_PAGEFLAGS
425#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800426#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
427 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700428#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800429#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800430#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
431 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700432#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800433
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800434#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700435
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700436#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
437#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
Dave Hansen348f8b62005-06-23 00:07:40 -0700438#endif
439
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700440#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
441#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
442#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800443#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700444
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700445static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446{
Dave Hansen348f8b62005-06-23 00:07:40 -0700447 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
Christoph Lameter89689ae2006-12-06 20:31:45 -0800450/*
451 * The identification function is only used by the buddy allocator for
452 * determining if two pages could be buddies. We are not really
453 * identifying a zone since we could be using a the section number
454 * id if we have not node id available in page flags.
455 * We guarantee only that it will return the same value for two
456 * combinable pages in a zone.
457 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700458static inline int page_zone_id(struct page *page)
459{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800460 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461}
462
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800463static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700464{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700465#ifdef CONFIG_NUMA
466 return zone->node;
467#else
468 return 0;
469#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700470}
471
Christoph Lameter89689ae2006-12-06 20:31:45 -0800472#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800473extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800474#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800475static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700476{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800477 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700478}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800479#endif
480
481static inline struct zone *page_zone(struct page *page)
482{
483 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
484}
485
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700486static inline unsigned long page_to_section(struct page *page)
487{
488 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
489}
490
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700491static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
Dave Hansen348f8b62005-06-23 00:07:40 -0700493 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
494 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
495}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700496
Dave Hansen348f8b62005-06-23 00:07:40 -0700497static inline void set_page_node(struct page *page, unsigned long node)
498{
499 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
500 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
501}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800502
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700503static inline void set_page_section(struct page *page, unsigned long section)
504{
505 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
506 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
507}
Dave Hansen348f8b62005-06-23 00:07:40 -0700508
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700509static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700510 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700511{
512 set_page_zone(page, zone);
513 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700514 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515}
516
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700517/*
Eric Paris7cd94142007-11-26 18:47:40 -0500518 * If a hint addr is less than mmap_min_addr change hint to be as
519 * low as possible but still greater than mmap_min_addr
520 */
521static inline unsigned long round_hint_to_min(unsigned long hint)
522{
523#ifdef CONFIG_SECURITY
524 hint &= PAGE_MASK;
525 if (((void *)hint != NULL) &&
526 (hint < mmap_min_addr))
527 return PAGE_ALIGN(mmap_min_addr);
528#endif
529 return hint;
530}
531
532/*
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700533 * Some inline functions in vmstat.h depend on page_zone()
534 */
535#include <linux/vmstat.h>
536
Ingo Molnar652050a2006-01-14 13:21:30 -0800537static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538{
539 return __va(page_to_pfn(page) << PAGE_SHIFT);
540}
541
542#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
543#define HASHED_PAGE_VIRTUAL
544#endif
545
546#if defined(WANT_PAGE_VIRTUAL)
547#define page_address(page) ((page)->virtual)
548#define set_page_address(page, address) \
549 do { \
550 (page)->virtual = (address); \
551 } while(0)
552#define page_address_init() do { } while(0)
553#endif
554
555#if defined(HASHED_PAGE_VIRTUAL)
556void *page_address(struct page *page);
557void set_page_address(struct page *page, void *virtual);
558void page_address_init(void);
559#endif
560
561#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
562#define page_address(page) lowmem_page_address(page)
563#define set_page_address(page, address) do { } while(0)
564#define page_address_init() do { } while(0)
565#endif
566
567/*
568 * On an anonymous page mapped into a user virtual memory area,
569 * page->mapping points to its anon_vma, not to a struct address_space;
570 * with the PAGE_MAPPING_ANON bit set to distinguish it.
571 *
572 * Please note that, confusingly, "page_mapping" refers to the inode
573 * address_space which maps the page from disk; whereas "page_mapped"
574 * refers to user virtual address space into which the page is mapped.
575 */
576#define PAGE_MAPPING_ANON 1
577
578extern struct address_space swapper_space;
579static inline struct address_space *page_mapping(struct page *page)
580{
581 struct address_space *mapping = page->mapping;
582
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700583 VM_BUG_ON(PageSlab(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 if (unlikely(PageSwapCache(page)))
585 mapping = &swapper_space;
586 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
587 mapping = NULL;
588 return mapping;
589}
590
591static inline int PageAnon(struct page *page)
592{
593 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
594}
595
596/*
597 * Return the pagecache index of the passed page. Regular pagecache pages
598 * use ->index whereas swapcache pages use ->private
599 */
600static inline pgoff_t page_index(struct page *page)
601{
602 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700603 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 return page->index;
605}
606
607/*
608 * The atomic page->_mapcount, like _count, starts from -1:
609 * so that transitions both from it and to it can be tracked,
610 * using atomic_inc_and_test and atomic_add_negative(-1).
611 */
612static inline void reset_page_mapcount(struct page *page)
613{
614 atomic_set(&(page)->_mapcount, -1);
615}
616
617static inline int page_mapcount(struct page *page)
618{
619 return atomic_read(&(page)->_mapcount) + 1;
620}
621
622/*
623 * Return true if this page is mapped into pagetables.
624 */
625static inline int page_mapped(struct page *page)
626{
627 return atomic_read(&(page)->_mapcount) >= 0;
628}
629
630/*
631 * Error return values for the *_nopage functions
632 */
633#define NOPAGE_SIGBUS (NULL)
634#define NOPAGE_OOM ((struct page *) (-1))
635
636/*
Jes Sorensenf4b81802006-09-27 01:50:10 -0700637 * Error return values for the *_nopfn functions
638 */
639#define NOPFN_SIGBUS ((unsigned long) -1)
640#define NOPFN_OOM ((unsigned long) -2)
Benjamin Herrenschmidt22cd25e2007-02-12 00:51:38 -0800641#define NOPFN_REFAULT ((unsigned long) -3)
Jes Sorensenf4b81802006-09-27 01:50:10 -0700642
643/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 * Different kinds of faults, as returned by handle_mm_fault().
645 * Used to decide whether a process gets delivered SIGBUS or
646 * just gets major/minor fault counters bumped up.
647 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700648
Nick Piggin83c54072007-07-19 01:47:05 -0700649#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700650
Nick Piggin83c54072007-07-19 01:47:05 -0700651#define VM_FAULT_OOM 0x0001
652#define VM_FAULT_SIGBUS 0x0002
653#define VM_FAULT_MAJOR 0x0004
654#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000655
Nick Piggin83c54072007-07-19 01:47:05 -0700656#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
657#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
Nick Piggin83c54072007-07-19 01:47:05 -0700659#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
Nick Piggind0217ac2007-07-19 01:47:03 -0700660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
662
663extern void show_free_areas(void);
664
665#ifdef CONFIG_SHMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666int shmem_lock(struct file *file, int lock, struct user_struct *user);
667#else
Russell King03b00eb2006-01-06 00:10:52 -0800668static inline int shmem_lock(struct file *file, int lock,
669 struct user_struct *user)
670{
671 return 0;
672}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673#endif
674struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
675
676int shmem_zero_setup(struct vm_area_struct *);
677
David Howellsb0e15192006-01-06 00:11:42 -0800678#ifndef CONFIG_MMU
679extern unsigned long shmem_get_unmapped_area(struct file *file,
680 unsigned long addr,
681 unsigned long len,
682 unsigned long pgoff,
683 unsigned long flags);
684#endif
685
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400686extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687extern int user_shm_lock(size_t, struct user_struct *);
688extern void user_shm_unlock(size_t, struct user_struct *);
689
690/*
691 * Parameter block passed down to zap_pte_range in exceptional cases.
692 */
693struct zap_details {
694 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
695 struct address_space *check_mapping; /* Check page->mapping if set */
696 pgoff_t first_index; /* Lowest page->index to unmap */
697 pgoff_t last_index; /* Highest page->index to unmap */
698 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 unsigned long truncate_count; /* Compare vm_truncate_count */
700};
701
Linus Torvalds6aab3412005-11-28 14:34:23 -0800702struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700703unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700705unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 struct vm_area_struct *start_vma, unsigned long start_addr,
707 unsigned long end_addr, unsigned long *nr_accounted,
708 struct zap_details *);
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700709void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
710 unsigned long end, unsigned long floor, unsigned long ceiling);
711void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
Hugh Dickinse0da3822005-04-19 13:29:15 -0700712 unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
714 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715void unmap_mapping_range(struct address_space *mapping,
716 loff_t const holebegin, loff_t const holelen, int even_cows);
717
718static inline void unmap_shared_mapping_range(struct address_space *mapping,
719 loff_t const holebegin, loff_t const holelen)
720{
721 unmap_mapping_range(mapping, holebegin, holelen, 0);
722}
723
724extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800725extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000726
David Howells7ee1dd32006-01-06 00:11:44 -0800727#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700728extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
David Howells7ee1dd32006-01-06 00:11:44 -0800729 unsigned long address, int write_access);
David Howells7ee1dd32006-01-06 00:11:44 -0800730#else
731static inline int handle_mm_fault(struct mm_struct *mm,
732 struct vm_area_struct *vma, unsigned long address,
733 int write_access)
734{
735 /* should never happen if there's no MMU */
736 BUG();
737 return VM_FAULT_SIGBUS;
738}
739#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000740
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741extern int make_pages_present(unsigned long addr, unsigned long end);
742extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
744int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
745 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
Nick Pigginb5810032005-10-29 18:16:12 -0700746void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
David Howellscf9a2ae2006-08-29 19:05:54 +0100748extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
749extern void do_invalidatepage(struct page *page, unsigned long offset);
750
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800752int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753int redirty_page_for_writepage(struct writeback_control *wbc,
754 struct page *page);
755int FASTCALL(set_page_dirty(struct page *page));
756int set_page_dirty_lock(struct page *page);
757int clear_page_dirty_for_io(struct page *page);
758
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700759extern unsigned long move_page_tables(struct vm_area_struct *vma,
760 unsigned long old_addr, struct vm_area_struct *new_vma,
761 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762extern unsigned long do_mremap(unsigned long addr,
763 unsigned long old_len, unsigned long new_len,
764 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700765extern int mprotect_fixup(struct vm_area_struct *vma,
766 struct vm_area_struct **pprev, unsigned long start,
767 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
769/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700770 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700772 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
773 * look through the least-recently-used 'nr_to_scan' entries and
774 * attempt to free them up. It should return the number of objects
775 * which remain in the cache. If it returns -1, it means it cannot do
776 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700778 * The 'gfpmask' refers to the allocation we are currently trying to
779 * fulfil.
780 *
781 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
782 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 */
Rusty Russell8e1f9362007-07-17 04:03:17 -0700784struct shrinker {
785 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
786 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Rusty Russell8e1f9362007-07-17 04:03:17 -0700788 /* These are for internal use */
789 struct list_head list;
790 long nr; /* objs pending delete */
791};
792#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
793extern void register_shrinker(struct shrinker *);
794extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Alexey Dobriyan4e950f62007-07-30 02:36:13 +0400796int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700797
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800798extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
799
Nick Piggin5f22df02007-05-06 14:49:02 -0700800#ifdef __PAGETABLE_PUD_FOLDED
801static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
802 unsigned long address)
803{
804 return 0;
805}
806#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700807int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700808#endif
809
810#ifdef __PAGETABLE_PMD_FOLDED
811static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
812 unsigned long address)
813{
814 return 0;
815}
816#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700817int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700818#endif
819
Hugh Dickins1bb36302005-10-29 18:16:22 -0700820int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
821int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
822
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823/*
824 * The following ifdef needed to get the 4level-fixup.h header to work.
825 * Remove it when 4level-fixup.h has been removed.
826 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700827#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
829{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700830 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
831 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832}
833
834static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
835{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700836 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
837 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700839#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
840
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700841#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
842/*
843 * We tuck a spinlock to guard each pagetable page into its struct page,
844 * at page->private, with BUILD_BUG_ON to make sure that this will not
845 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
846 * When freeing, reset page->mapping so free_pages_check won't complain.
847 */
Andrew Morton349aef02006-01-08 01:04:36 -0800848#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700849#define pte_lock_init(_page) do { \
850 spin_lock_init(__pte_lockptr(_page)); \
851} while (0)
852#define pte_lock_deinit(page) ((page)->mapping = NULL)
853#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
854#else
855/*
856 * We use mm->page_table_lock to guard all pagetable pages of the mm.
857 */
858#define pte_lock_init(page) do {} while (0)
859#define pte_lock_deinit(page) do {} while (0)
860#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
861#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
862
Hugh Dickinsc74df322005-10-29 18:16:23 -0700863#define pte_offset_map_lock(mm, pmd, address, ptlp) \
864({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700865 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700866 pte_t *__pte = pte_offset_map(pmd, address); \
867 *(ptlp) = __ptl; \
868 spin_lock(__ptl); \
869 __pte; \
870})
871
872#define pte_unmap_unlock(pte, ptl) do { \
873 spin_unlock(ptl); \
874 pte_unmap(pte); \
875} while (0)
876
Hugh Dickins1bb36302005-10-29 18:16:22 -0700877#define pte_alloc_map(mm, pmd, address) \
878 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
879 NULL: pte_offset_map(pmd, address))
880
Hugh Dickinsc74df322005-10-29 18:16:23 -0700881#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
882 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
883 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
884
Hugh Dickins1bb36302005-10-29 18:16:22 -0700885#define pte_alloc_kernel(pmd, address) \
886 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
887 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888
889extern void free_area_init(unsigned long * zones_size);
890extern void free_area_init_node(int nid, pg_data_t *pgdat,
891 unsigned long * zones_size, unsigned long zone_start_pfn,
892 unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -0700893#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
894/*
895 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
896 * zones, allocate the backing mem_map and account for memory holes in a more
897 * architecture independent manner. This is a substitute for creating the
898 * zone_sizes[] and zholes_size[] arrays and passing them to
899 * free_area_init_node()
900 *
901 * An architecture is expected to register range of page frames backed by
902 * physical memory with add_active_range() before calling
903 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
904 * usage, an architecture is expected to do something like
905 *
906 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
907 * max_highmem_pfn};
908 * for_each_valid_physical_page_range()
909 * add_active_range(node_id, start_pfn, end_pfn)
910 * free_area_init_nodes(max_zone_pfns);
911 *
912 * If the architecture guarantees that there are no holes in the ranges
913 * registered with add_active_range(), free_bootmem_active_regions()
914 * will call free_bootmem_node() for each registered physical page range.
915 * Similarly sparse_memory_present_with_active_regions() calls
916 * memory_present() for each range when SPARSEMEM is enabled.
917 *
918 * See mm/page_alloc.c for more information on each function exposed by
919 * CONFIG_ARCH_POPULATES_NODE_MAP
920 */
921extern void free_area_init_nodes(unsigned long *max_zone_pfn);
922extern void add_active_range(unsigned int nid, unsigned long start_pfn,
923 unsigned long end_pfn);
924extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
925 unsigned long new_end_pfn);
Mel Gormanfb014392006-09-27 01:49:59 -0700926extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
927 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -0700928extern void remove_all_active_ranges(void);
929extern unsigned long absent_pages_in_range(unsigned long start_pfn,
930 unsigned long end_pfn);
931extern void get_pfn_range_for_nid(unsigned int nid,
932 unsigned long *start_pfn, unsigned long *end_pfn);
933extern unsigned long find_min_pfn_with_active_regions(void);
934extern unsigned long find_max_pfn_with_active_regions(void);
935extern void free_bootmem_with_active_regions(int nid,
936 unsigned long max_low_pfn);
937extern void sparse_memory_present_with_active_regions(int nid);
938#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
939extern int early_pfn_to_nid(unsigned long pfn);
940#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
941#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
Mel Gorman0e0b8642006-09-27 01:49:56 -0700942extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -0800943extern void memmap_init_zone(unsigned long, int, unsigned long,
944 unsigned long, enum memmap_context);
Dave Hansen3947be12005-10-29 18:16:54 -0700945extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946extern void mem_init(void);
947extern void show_mem(void);
948extern void si_meminfo(struct sysinfo * val);
949extern void si_meminfo_node(struct sysinfo *val, int nid);
950
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700951#ifdef CONFIG_NUMA
952extern void setup_per_cpu_pageset(void);
953#else
954static inline void setup_per_cpu_pageset(void) {}
955#endif
956
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957/* prio_tree.c */
958void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
959void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
960void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
961struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
962 struct prio_tree_iter *iter);
963
964#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
965 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
966 (vma = vma_prio_tree_next(vma, iter)); )
967
968static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
969 struct list_head *list)
970{
971 vma->shared.vm_set.parent = NULL;
972 list_add_tail(&vma->shared.vm_set.list, list);
973}
974
975/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -0700976extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
978 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
979extern struct vm_area_struct *vma_merge(struct mm_struct *,
980 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
981 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
982 struct mempolicy *);
983extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
984extern int split_vma(struct mm_struct *,
985 struct vm_area_struct *, unsigned long addr, int new_below);
986extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
987extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
988 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -0700989extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
991 unsigned long addr, unsigned long len, pgoff_t pgoff);
992extern void exit_mmap(struct mm_struct *);
akpm@osdl.org119f6572005-05-01 08:58:35 -0700993extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -0800994extern int install_special_mapping(struct mm_struct *mm,
995 unsigned long addr, unsigned long len,
996 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997
998extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
999
1000extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1001 unsigned long len, unsigned long prot,
1002 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001003extern unsigned long mmap_region(struct file *file, unsigned long addr,
1004 unsigned long len, unsigned long flags,
1005 unsigned int vm_flags, unsigned long pgoff,
1006 int accountable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
1008static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1009 unsigned long len, unsigned long prot,
1010 unsigned long flag, unsigned long offset)
1011{
1012 unsigned long ret = -EINVAL;
1013 if ((offset + PAGE_ALIGN(len)) < offset)
1014 goto out;
1015 if (!(offset & ~PAGE_MASK))
1016 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1017out:
1018 return ret;
1019}
1020
1021extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1022
1023extern unsigned long do_brk(unsigned long, unsigned long);
1024
1025/* filemap.c */
1026extern unsigned long page_unuse(struct page *);
1027extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001028extern void truncate_inode_pages_range(struct address_space *,
1029 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
1031/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001032extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
1034/* mm/page-writeback.c */
1035int write_one_page(struct page *page, int wait);
1036
1037/* readahead.c */
1038#define VM_MAX_READAHEAD 128 /* kbytes */
1039#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040
1041int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001042 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001044 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001045
1046void page_cache_sync_readahead(struct address_space *mapping,
1047 struct file_ra_state *ra,
1048 struct file *filp,
1049 pgoff_t offset,
1050 unsigned long size);
1051
1052void page_cache_async_readahead(struct address_space *mapping,
1053 struct file_ra_state *ra,
1054 struct file *filp,
1055 struct page *pg,
1056 pgoff_t offset,
1057 unsigned long size);
1058
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059unsigned long max_sane_readahead(unsigned long nr);
1060
1061/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001062extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001063#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001064extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001065#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001066extern int expand_stack_downwards(struct vm_area_struct *vma,
1067 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068
1069/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1070extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1071extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1072 struct vm_area_struct **pprev);
1073
1074/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1075 NULL if none. Assume start_addr < end_addr. */
1076static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1077{
1078 struct vm_area_struct * vma = find_vma(mm,start_addr);
1079
1080 if (vma && end_addr <= vma->vm_start)
1081 vma = NULL;
1082 return vma;
1083}
1084
1085static inline unsigned long vma_pages(struct vm_area_struct *vma)
1086{
1087 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1088}
1089
Hugh Dickins804af2c2006-07-26 21:39:49 +01001090pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001091struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1092struct page *vmalloc_to_page(void *addr);
1093unsigned long vmalloc_to_pfn(void *addr);
1094int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1095 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001096int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001097int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1098 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099
Linus Torvalds6aab3412005-11-28 14:34:23 -08001100struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001101 unsigned int foll_flags);
1102#define FOLL_WRITE 0x01 /* check pte is writable */
1103#define FOLL_TOUCH 0x02 /* mark page accessed */
1104#define FOLL_GET 0x04 /* do get_page on page */
1105#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001107typedef int (*pte_fn_t)(pte_t *pte, struct page *pmd_page, unsigned long addr,
1108 void *data);
1109extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1110 unsigned long size, pte_fn_t fn, void *data);
1111
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001113void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001115static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 unsigned long flags, struct file *file, long pages)
1117{
1118}
1119#endif /* CONFIG_PROC_FS */
1120
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121#ifndef CONFIG_DEBUG_PAGEALLOC
1122static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001123kernel_map_pages(struct page *page, int numpages, int enable) {}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124#endif
1125
1126extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1127#ifdef __HAVE_ARCH_GATE_AREA
1128int in_gate_area_no_task(unsigned long addr);
1129int in_gate_area(struct task_struct *task, unsigned long addr);
1130#else
1131int in_gate_area_no_task(unsigned long addr);
1132#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1133#endif /* __HAVE_ARCH_GATE_AREA */
1134
Andrew Morton9d0243b2006-01-08 01:00:39 -08001135int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1136 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001137unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001138 unsigned long lru_pages);
1139void drop_pagecache(void);
1140void drop_slab(void);
1141
Luke Yang7a9166e2006-02-20 18:28:07 -08001142#ifndef CONFIG_MMU
1143#define randomize_va_space 0
1144#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001145extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001146#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001147
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001148const char * arch_vma_name(struct vm_area_struct *vma);
Ingo Molnare6e54942006-06-27 02:53:50 -07001149
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001150struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001151pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1152pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1153pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1154pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001155void *vmemmap_alloc_block(unsigned long size, int node);
1156void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001157int vmemmap_populate_basepages(struct page *start_page,
1158 unsigned long pages, int node);
1159int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001160
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161#endif /* __KERNEL__ */
1162#endif /* _LINUX_MM_H */