blob: a17b147c61e74853e51268c3eaf144aee8c37006 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
4#include <linux/sched.h>
5#include <linux/errno.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -08006#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07007
8#ifdef __KERNEL__
9
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/gfp.h>
11#include <linux/list.h>
12#include <linux/mmzone.h>
13#include <linux/rbtree.h>
14#include <linux/prio_tree.h>
15#include <linux/fs.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080016#include <linux/mutex.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070017#include <linux/debug_locks.h>
Peter Zijlstrad08b3852006-09-25 23:30:57 -070018#include <linux/backing-dev.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070019#include <linux/mm_types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020
21struct mempolicy;
22struct anon_vma;
23
24#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
25extern unsigned long max_mapnr;
26#endif
27
28extern unsigned long num_physpages;
29extern void * high_memory;
30extern unsigned long vmalloc_earlyreserve;
31extern int page_cluster;
32
33#ifdef CONFIG_SYSCTL
34extern int sysctl_legacy_va_layout;
35#else
36#define sysctl_legacy_va_layout 0
37#endif
38
39#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
45/*
46 * Linux kernel virtual memory manager primitives.
47 * The idea being to have a "virtual" mm in the same way
48 * we have a virtual fs - giving a cleaner interface to the
49 * mm details, and allowing different kinds of memory mappings
50 * (from shared memory to executable loading to arbitrary
51 * mmap() functions).
52 */
53
54/*
55 * This struct defines a memory VMM memory area. There is one of these
56 * per VM-area/task. A VM area is any part of the process virtual memory
57 * space that has a special rule for the page-fault handlers (ie a shared
58 * library, the executable area etc).
59 */
60struct vm_area_struct {
61 struct mm_struct * vm_mm; /* The address space we belong to. */
62 unsigned long vm_start; /* Our start address within vm_mm. */
63 unsigned long vm_end; /* The first byte after our end address
64 within vm_mm. */
65
66 /* linked list of VM areas per task, sorted by address */
67 struct vm_area_struct *vm_next;
68
69 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
70 unsigned long vm_flags; /* Flags, listed below. */
71
72 struct rb_node vm_rb;
73
74 /*
75 * For areas with an address space and backing store,
76 * linkage into the address_space->i_mmap prio tree, or
77 * linkage to the list of like vmas hanging off its node, or
78 * linkage of vma in the address_space->i_mmap_nonlinear list.
79 */
80 union {
81 struct {
82 struct list_head list;
83 void *parent; /* aligns with prio_tree_node parent */
84 struct vm_area_struct *head;
85 } vm_set;
86
87 struct raw_prio_tree_node prio_tree_node;
88 } shared;
89
90 /*
91 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
92 * list, after a COW of one of the file pages. A MAP_SHARED vma
93 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
94 * or brk vma (with NULL file) can only be in an anon_vma list.
95 */
96 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
97 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
98
99 /* Function pointers to deal with this struct. */
100 struct vm_operations_struct * vm_ops;
101
102 /* Information about our backing store: */
103 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
104 units, *not* PAGE_CACHE_SIZE */
105 struct file * vm_file; /* File we map to (can be NULL). */
106 void * vm_private_data; /* was vm_pte (shared mem) */
107 unsigned long vm_truncate_count;/* truncate_count or restart_addr */
108
109#ifndef CONFIG_MMU
110 atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
111#endif
112#ifdef CONFIG_NUMA
113 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
114#endif
115};
116
Christoph Lameterc43692e2006-12-06 20:32:48 -0800117extern struct kmem_cache *vm_area_cachep;
118
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119/*
120 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
121 * disabled, then there's a single shared list of VMAs maintained by the
122 * system, and mm's subscribe to these individually
123 */
124struct vm_list_struct {
125 struct vm_list_struct *next;
126 struct vm_area_struct *vma;
127};
128
129#ifndef CONFIG_MMU
130extern struct rb_root nommu_vma_tree;
131extern struct rw_semaphore nommu_vma_sem;
132
133extern unsigned int kobjsize(const void *objp);
134#endif
135
136/*
137 * vm_flags..
138 */
139#define VM_READ 0x00000001 /* currently active flags */
140#define VM_WRITE 0x00000002
141#define VM_EXEC 0x00000004
142#define VM_SHARED 0x00000008
143
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -0700144/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
146#define VM_MAYWRITE 0x00000020
147#define VM_MAYEXEC 0x00000040
148#define VM_MAYSHARE 0x00000080
149
150#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
151#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -0800152#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
154
155#define VM_EXECUTABLE 0x00001000
156#define VM_LOCKED 0x00002000
157#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
158
159 /* Used by sys_madvise() */
160#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
161#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
162
163#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
164#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800165#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
167#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
168#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
169#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800170#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
172#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
173#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
174#endif
175
176#ifdef CONFIG_STACK_GROWSUP
177#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
178#else
179#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
180#endif
181
182#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
183#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
184#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
185#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
186#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
187
188/*
189 * mapping from the currently active vm_flags protection bits (the
190 * low four bits) to a page protection mask..
191 */
192extern pgprot_t protection_map[16];
193
194
195/*
196 * These are the virtual MM functions - opening of an area, closing and
197 * unmapping it (needed to keep files on disk up-to-date etc), pointer
198 * to the functions called when a no-page or a wp-page exception occurs.
199 */
200struct vm_operations_struct {
201 void (*open)(struct vm_area_struct * area);
202 void (*close)(struct vm_area_struct * area);
203 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
Jes Sorensenf4b81802006-09-27 01:50:10 -0700204 unsigned long (*nopfn)(struct vm_area_struct * area, unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
David Howells9637a5e2006-06-23 02:03:43 -0700206
207 /* notification that a previously read-only page is about to become
208 * writable, if an error is returned it will cause a SIGBUS */
209 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210#ifdef CONFIG_NUMA
211 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
212 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
213 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700214 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
215 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216#endif
217};
218
219struct mmu_gather;
220struct inode;
221
Andrew Morton349aef02006-01-08 01:04:36 -0800222#define page_private(page) ((page)->private)
223#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700224
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225/*
226 * FIXME: take this include out, include page-flags.h in
227 * files which need it (119 of them)
228 */
229#include <linux/page-flags.h>
230
Nick Piggin725d7042006-09-25 23:30:55 -0700231#ifdef CONFIG_DEBUG_VM
232#define VM_BUG_ON(cond) BUG_ON(cond)
233#else
234#define VM_BUG_ON(condition) do { } while(0)
235#endif
236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237/*
238 * Methods to modify the page usage count.
239 *
240 * What counts for a page usage:
241 * - cache mapping (page->mapping)
242 * - private data (page->private)
243 * - page mapped in a task's page tables, each mapping
244 * is counted separately
245 *
246 * Also, many kernel routines increase the page count before a critical
247 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 */
249
250/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700251 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800253static inline int put_page_testzero(struct page *page)
254{
Nick Piggin725d7042006-09-25 23:30:55 -0700255 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800256 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800257}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
259/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800260 * Try to grab a ref unless the page has a refcount of zero, return false if
261 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800263static inline int get_page_unless_zero(struct page *page)
264{
Nick Piggin725d7042006-09-25 23:30:55 -0700265 VM_BUG_ON(PageCompound(page));
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800266 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800267}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700269static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270{
Nick Piggin617d2212006-03-22 00:08:43 -0800271 if (unlikely(PageCompound(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700272 page = (struct page *)page_private(page);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800273 return atomic_read(&page->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274}
275
276static inline void get_page(struct page *page)
277{
278 if (unlikely(PageCompound(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700279 page = (struct page *)page_private(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700280 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 atomic_inc(&page->_count);
282}
283
Nick Piggin7835e982006-03-22 00:08:40 -0800284/*
285 * Setup the page count before being freed into the page allocator for
286 * the first time (boot or memory hotplug)
287 */
288static inline void init_page_count(struct page *page)
289{
290 atomic_set(&page->_count, 1);
291}
292
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700294void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800296void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800297
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800299 * Compound pages have a destructor function. Provide a
300 * prototype for that function and accessor functions.
301 * These are _only_ valid on the head of a PG_compound page.
302 */
303typedef void compound_page_dtor(struct page *);
304
305static inline void set_compound_page_dtor(struct page *page,
306 compound_page_dtor *dtor)
307{
308 page[1].lru.next = (void *)dtor;
309}
310
311static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
312{
313 return (compound_page_dtor *)page[1].lru.next;
314}
315
316/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 * Multiple processes may "see" the same page. E.g. for untouched
318 * mappings of /dev/null, all processes see the same page full of
319 * zeroes, and text pages of executables and shared libraries have
320 * only one copy in memory, at most, normally.
321 *
322 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700323 * page_count() == 0 means the page is free. page->lru is then used for
324 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700325 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700327 * Pages are allocated by the slab allocator in order to provide memory
328 * to kmalloc and kmem_cache_alloc. In this case, the management of the
329 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
330 * unless a particular usage is carefully commented. (the responsibility of
331 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700333 * A page may be used by anyone else who does a __get_free_page().
334 * In this case, page_count still tracks the references, and should only
335 * be used through the normal accessor functions. The top bits of page->flags
336 * and page->virtual store page management information, but all other fields
337 * are unused and could be used privately, carefully. The management of this
338 * page is the responsibility of the one who allocated it, and those who have
339 * subsequently been given references to it.
340 *
341 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 * managed by the Linux memory manager: I/O, buffers, swapping etc.
343 * The following discussion applies only to them.
344 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700345 * A pagecache page contains an opaque `private' member, which belongs to the
346 * page's address_space. Usually, this is the address of a circular list of
347 * the page's disk buffers. PG_private must be set to tell the VM to call
348 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700350 * A page may belong to an inode's memory mapping. In this case, page->mapping
351 * is the pointer to the inode, and page->index is the file offset of the page,
352 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700354 * If pagecache pages are not associated with an inode, they are said to be
355 * anonymous pages. These may become associated with the swapcache, and in that
356 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700358 * In either case (swapcache or inode backed), the pagecache itself holds one
359 * reference to the page. Setting PG_private should also increment the
360 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700362 * The pagecache pages are stored in a per-mapping radix tree, which is
363 * rooted at mapping->page_tree, and indexed by offset.
364 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
365 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700367 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 * - inode pages may need to be read from disk,
369 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700370 * to be written back to the inode on disk,
371 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
372 * modified may need to be swapped out to swap space and (later) to be read
373 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 */
375
376/*
377 * The zone field is never updated after free_area_init_core()
378 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700380
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700381
382/*
383 * page->flags layout:
384 *
385 * There are three possibilities for how page->flags get
386 * laid out. The first is for the normal case, without
387 * sparsemem. The second is for sparsemem when there is
388 * plenty of space for node and section. The last is when
389 * we have run out of space and have to fall back to an
390 * alternate (slower) way of determining the node.
391 *
392 * No sparsemem: | NODE | ZONE | ... | FLAGS |
393 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
394 * no space for node: | SECTION | ZONE | ... | FLAGS |
395 */
396#ifdef CONFIG_SPARSEMEM
397#define SECTIONS_WIDTH SECTIONS_SHIFT
398#else
399#define SECTIONS_WIDTH 0
400#endif
401
402#define ZONES_WIDTH ZONES_SHIFT
403
404#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
405#define NODES_WIDTH NODES_SHIFT
406#else
407#define NODES_WIDTH 0
408#endif
409
410/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100411#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700412#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
413#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
414
415/*
416 * We are going to use the flags for the page to node mapping if its in
417 * there. This includes the case where there is no node, so it is implicit.
418 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800419#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
420#define NODE_NOT_IN_PAGE_FLAGS
421#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700422
423#ifndef PFN_SECTION_SHIFT
424#define PFN_SECTION_SHIFT 0
425#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700426
427/*
428 * Define the bit shifts to access each section. For non-existant
429 * sections we define the shift as 0; that plus a 0 mask ensures
430 * the compiler will optimise away reference to them.
431 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700432#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
433#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
434#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700435
Christoph Lameter89689ae2006-12-06 20:31:45 -0800436/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
437#ifdef NODE_NOT_IN_PAGEFLAGS
438#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700439#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800440#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700441#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800442
443#if ZONES_WIDTH > 0
444#define ZONEID_PGSHIFT ZONES_PGSHIFT
445#else
446#define ZONEID_PGSHIFT NODES_PGOFF
447#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700448
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700449#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
450#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
Dave Hansen348f8b62005-06-23 00:07:40 -0700451#endif
452
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700453#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
454#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
455#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800456#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700457
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700458static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459{
Dave Hansen348f8b62005-06-23 00:07:40 -0700460 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462
Christoph Lameter89689ae2006-12-06 20:31:45 -0800463/*
464 * The identification function is only used by the buddy allocator for
465 * determining if two pages could be buddies. We are not really
466 * identifying a zone since we could be using a the section number
467 * id if we have not node id available in page flags.
468 * We guarantee only that it will return the same value for two
469 * combinable pages in a zone.
470 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700471static inline int page_zone_id(struct page *page)
472{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800473 BUILD_BUG_ON(ZONEID_PGSHIFT == 0 && ZONEID_MASK);
474 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475}
476
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800477static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700478{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700479#ifdef CONFIG_NUMA
480 return zone->node;
481#else
482 return 0;
483#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700484}
485
Christoph Lameter89689ae2006-12-06 20:31:45 -0800486#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800487extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800488#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800489static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700490{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800491 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700492}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800493#endif
494
495static inline struct zone *page_zone(struct page *page)
496{
497 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
498}
499
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700500static inline unsigned long page_to_section(struct page *page)
501{
502 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
503}
504
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700505static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506{
Dave Hansen348f8b62005-06-23 00:07:40 -0700507 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
508 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
509}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700510
Dave Hansen348f8b62005-06-23 00:07:40 -0700511static inline void set_page_node(struct page *page, unsigned long node)
512{
513 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
514 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
515}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800516
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700517static inline void set_page_section(struct page *page, unsigned long section)
518{
519 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
520 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
521}
Dave Hansen348f8b62005-06-23 00:07:40 -0700522
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700523static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700524 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700525{
526 set_page_zone(page, zone);
527 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700528 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529}
530
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700531/*
532 * Some inline functions in vmstat.h depend on page_zone()
533 */
534#include <linux/vmstat.h>
535
Ingo Molnar652050a2006-01-14 13:21:30 -0800536static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
538 return __va(page_to_pfn(page) << PAGE_SHIFT);
539}
540
541#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
542#define HASHED_PAGE_VIRTUAL
543#endif
544
545#if defined(WANT_PAGE_VIRTUAL)
546#define page_address(page) ((page)->virtual)
547#define set_page_address(page, address) \
548 do { \
549 (page)->virtual = (address); \
550 } while(0)
551#define page_address_init() do { } while(0)
552#endif
553
554#if defined(HASHED_PAGE_VIRTUAL)
555void *page_address(struct page *page);
556void set_page_address(struct page *page, void *virtual);
557void page_address_init(void);
558#endif
559
560#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
561#define page_address(page) lowmem_page_address(page)
562#define set_page_address(page, address) do { } while(0)
563#define page_address_init() do { } while(0)
564#endif
565
566/*
567 * On an anonymous page mapped into a user virtual memory area,
568 * page->mapping points to its anon_vma, not to a struct address_space;
569 * with the PAGE_MAPPING_ANON bit set to distinguish it.
570 *
571 * Please note that, confusingly, "page_mapping" refers to the inode
572 * address_space which maps the page from disk; whereas "page_mapped"
573 * refers to user virtual address space into which the page is mapped.
574 */
575#define PAGE_MAPPING_ANON 1
576
577extern struct address_space swapper_space;
578static inline struct address_space *page_mapping(struct page *page)
579{
580 struct address_space *mapping = page->mapping;
581
582 if (unlikely(PageSwapCache(page)))
583 mapping = &swapper_space;
584 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
585 mapping = NULL;
586 return mapping;
587}
588
589static inline int PageAnon(struct page *page)
590{
591 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
592}
593
594/*
595 * Return the pagecache index of the passed page. Regular pagecache pages
596 * use ->index whereas swapcache pages use ->private
597 */
598static inline pgoff_t page_index(struct page *page)
599{
600 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700601 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 return page->index;
603}
604
605/*
606 * The atomic page->_mapcount, like _count, starts from -1:
607 * so that transitions both from it and to it can be tracked,
608 * using atomic_inc_and_test and atomic_add_negative(-1).
609 */
610static inline void reset_page_mapcount(struct page *page)
611{
612 atomic_set(&(page)->_mapcount, -1);
613}
614
615static inline int page_mapcount(struct page *page)
616{
617 return atomic_read(&(page)->_mapcount) + 1;
618}
619
620/*
621 * Return true if this page is mapped into pagetables.
622 */
623static inline int page_mapped(struct page *page)
624{
625 return atomic_read(&(page)->_mapcount) >= 0;
626}
627
628/*
629 * Error return values for the *_nopage functions
630 */
631#define NOPAGE_SIGBUS (NULL)
632#define NOPAGE_OOM ((struct page *) (-1))
Benjamin Herrenschmidt7f7bbbe2006-10-06 00:43:53 -0700633#define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
635/*
Jes Sorensenf4b81802006-09-27 01:50:10 -0700636 * Error return values for the *_nopfn functions
637 */
638#define NOPFN_SIGBUS ((unsigned long) -1)
639#define NOPFN_OOM ((unsigned long) -2)
640
641/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 * Different kinds of faults, as returned by handle_mm_fault().
643 * Used to decide whether a process gets delivered SIGBUS or
644 * just gets major/minor fault counters bumped up.
645 */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000646#define VM_FAULT_OOM 0x00
647#define VM_FAULT_SIGBUS 0x01
648#define VM_FAULT_MINOR 0x02
649#define VM_FAULT_MAJOR 0x03
650
651/*
652 * Special case for get_user_pages.
653 * Must be in a distinct bit from the above VM_FAULT_ flags.
654 */
655#define VM_FAULT_WRITE 0x10
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
657#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
658
659extern void show_free_areas(void);
660
661#ifdef CONFIG_SHMEM
662struct page *shmem_nopage(struct vm_area_struct *vma,
663 unsigned long address, int *type);
664int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
665struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
666 unsigned long addr);
667int shmem_lock(struct file *file, int lock, struct user_struct *user);
668#else
669#define shmem_nopage filemap_nopage
Russell King03b00eb2006-01-06 00:10:52 -0800670
671static inline int shmem_lock(struct file *file, int lock,
672 struct user_struct *user)
673{
674 return 0;
675}
676
677static inline int shmem_set_policy(struct vm_area_struct *vma,
678 struct mempolicy *new)
679{
680 return 0;
681}
682
683static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
684 unsigned long addr)
685{
686 return NULL;
687}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688#endif
689struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
David Howellsb0e15192006-01-06 00:11:42 -0800690extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691
692int shmem_zero_setup(struct vm_area_struct *);
693
David Howellsb0e15192006-01-06 00:11:42 -0800694#ifndef CONFIG_MMU
695extern unsigned long shmem_get_unmapped_area(struct file *file,
696 unsigned long addr,
697 unsigned long len,
698 unsigned long pgoff,
699 unsigned long flags);
700#endif
701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702static inline int can_do_mlock(void)
703{
704 if (capable(CAP_IPC_LOCK))
705 return 1;
706 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
707 return 1;
708 return 0;
709}
710extern int user_shm_lock(size_t, struct user_struct *);
711extern void user_shm_unlock(size_t, struct user_struct *);
712
713/*
714 * Parameter block passed down to zap_pte_range in exceptional cases.
715 */
716struct zap_details {
717 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
718 struct address_space *check_mapping; /* Check page->mapping if set */
719 pgoff_t first_index; /* Lowest page->index to unmap */
720 pgoff_t last_index; /* Highest page->index to unmap */
721 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 unsigned long truncate_count; /* Compare vm_truncate_count */
723};
724
Linus Torvalds6aab3412005-11-28 14:34:23 -0800725struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700726unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700728unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 struct vm_area_struct *start_vma, unsigned long start_addr,
730 unsigned long end_addr, unsigned long *nr_accounted,
731 struct zap_details *);
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700732void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
733 unsigned long end, unsigned long floor, unsigned long ceiling);
734void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
Hugh Dickinse0da3822005-04-19 13:29:15 -0700735 unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
737 struct vm_area_struct *vma);
738int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
739 unsigned long size, pgprot_t prot);
740void unmap_mapping_range(struct address_space *mapping,
741 loff_t const holebegin, loff_t const holelen, int even_cows);
742
743static inline void unmap_shared_mapping_range(struct address_space *mapping,
744 loff_t const holebegin, loff_t const holelen)
745{
746 unmap_mapping_range(mapping, holebegin, holelen, 0);
747}
748
749extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800750extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
752extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000753
David Howells7ee1dd32006-01-06 00:11:44 -0800754#ifdef CONFIG_MMU
755extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
756 unsigned long address, int write_access);
757
758static inline int handle_mm_fault(struct mm_struct *mm,
759 struct vm_area_struct *vma, unsigned long address,
760 int write_access)
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000761{
David Howells7ee1dd32006-01-06 00:11:44 -0800762 return __handle_mm_fault(mm, vma, address, write_access) &
763 (~VM_FAULT_WRITE);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000764}
David Howells7ee1dd32006-01-06 00:11:44 -0800765#else
766static inline int handle_mm_fault(struct mm_struct *mm,
767 struct vm_area_struct *vma, unsigned long address,
768 int write_access)
769{
770 /* should never happen if there's no MMU */
771 BUG();
772 return VM_FAULT_SIGBUS;
773}
774#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000775
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776extern int make_pages_present(unsigned long addr, unsigned long end);
777extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
778void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
779
780int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
781 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
Nick Pigginb5810032005-10-29 18:16:12 -0700782void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
David Howellscf9a2ae2006-08-29 19:05:54 +0100784extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
785extern void do_invalidatepage(struct page *page, unsigned long offset);
786
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787int __set_page_dirty_nobuffers(struct page *page);
788int redirty_page_for_writepage(struct writeback_control *wbc,
789 struct page *page);
790int FASTCALL(set_page_dirty(struct page *page));
791int set_page_dirty_lock(struct page *page);
792int clear_page_dirty_for_io(struct page *page);
793
794extern unsigned long do_mremap(unsigned long addr,
795 unsigned long old_len, unsigned long new_len,
796 unsigned long flags, unsigned long new_addr);
797
798/*
799 * Prototype to add a shrinker callback for ageable caches.
800 *
801 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
802 * scan `nr_to_scan' objects, attempting to free them.
803 *
Domen Puncer845d3432005-05-05 16:16:14 -0700804 * The callback must return the number of objects which remain in the cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 *
Domen Puncer845d3432005-05-05 16:16:14 -0700806 * The callback will be passed nr_to_scan == 0 when the VM is querying the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 * cache size, so a fastpath for that case is appropriate.
808 */
Al Viro6daa0e22005-10-21 03:18:50 -0400809typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811/*
812 * Add an aging callback. The int is the number of 'seeks' it takes
813 * to recreate one of the objects that these functions age.
814 */
815
816#define DEFAULT_SEEKS 2
817struct shrinker;
818extern struct shrinker *set_shrinker(int, shrinker_t);
819extern void remove_shrinker(struct shrinker *shrinker);
820
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700821/*
822 * Some shared mappigns will want the pages marked read-only
823 * to track write events. If so, we'll downgrade vm_page_prot
824 * to the private version (using protection_map[] without the
825 * VM_SHARED bit).
826 */
827static inline int vma_wants_writenotify(struct vm_area_struct *vma)
828{
829 unsigned int vm_flags = vma->vm_flags;
830
831 /* If it was private or non-writable, the write bit is already clear */
832 if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED)))
833 return 0;
834
835 /* The backer wishes to know when pages are first written to? */
836 if (vma->vm_ops && vma->vm_ops->page_mkwrite)
837 return 1;
838
839 /* The open routine did something to the protections already? */
840 if (pgprot_val(vma->vm_page_prot) !=
841 pgprot_val(protection_map[vm_flags &
842 (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]))
843 return 0;
844
845 /* Specialty mapping? */
846 if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE))
847 return 0;
848
849 /* Can the mapping track the dirty pages? */
850 return vma->vm_file && vma->vm_file->f_mapping &&
851 mapping_cap_account_dirty(vma->vm_file->f_mapping);
852}
853
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800854extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
855
Hugh Dickins1bb36302005-10-29 18:16:22 -0700856int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
857int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
858int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
859int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
860
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861/*
862 * The following ifdef needed to get the 4level-fixup.h header to work.
863 * Remove it when 4level-fixup.h has been removed.
864 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700865#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
867{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700868 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
869 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870}
871
872static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
873{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700874 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
875 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700877#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
878
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700879#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
880/*
881 * We tuck a spinlock to guard each pagetable page into its struct page,
882 * at page->private, with BUILD_BUG_ON to make sure that this will not
883 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
884 * When freeing, reset page->mapping so free_pages_check won't complain.
885 */
Andrew Morton349aef02006-01-08 01:04:36 -0800886#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700887#define pte_lock_init(_page) do { \
888 spin_lock_init(__pte_lockptr(_page)); \
889} while (0)
890#define pte_lock_deinit(page) ((page)->mapping = NULL)
891#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
892#else
893/*
894 * We use mm->page_table_lock to guard all pagetable pages of the mm.
895 */
896#define pte_lock_init(page) do {} while (0)
897#define pte_lock_deinit(page) do {} while (0)
898#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
899#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
900
Hugh Dickinsc74df322005-10-29 18:16:23 -0700901#define pte_offset_map_lock(mm, pmd, address, ptlp) \
902({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700903 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700904 pte_t *__pte = pte_offset_map(pmd, address); \
905 *(ptlp) = __ptl; \
906 spin_lock(__ptl); \
907 __pte; \
908})
909
910#define pte_unmap_unlock(pte, ptl) do { \
911 spin_unlock(ptl); \
912 pte_unmap(pte); \
913} while (0)
914
Hugh Dickins1bb36302005-10-29 18:16:22 -0700915#define pte_alloc_map(mm, pmd, address) \
916 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
917 NULL: pte_offset_map(pmd, address))
918
Hugh Dickinsc74df322005-10-29 18:16:23 -0700919#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
920 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
921 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
922
Hugh Dickins1bb36302005-10-29 18:16:22 -0700923#define pte_alloc_kernel(pmd, address) \
924 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
925 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926
927extern void free_area_init(unsigned long * zones_size);
928extern void free_area_init_node(int nid, pg_data_t *pgdat,
929 unsigned long * zones_size, unsigned long zone_start_pfn,
930 unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -0700931#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
932/*
933 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
934 * zones, allocate the backing mem_map and account for memory holes in a more
935 * architecture independent manner. This is a substitute for creating the
936 * zone_sizes[] and zholes_size[] arrays and passing them to
937 * free_area_init_node()
938 *
939 * An architecture is expected to register range of page frames backed by
940 * physical memory with add_active_range() before calling
941 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
942 * usage, an architecture is expected to do something like
943 *
944 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
945 * max_highmem_pfn};
946 * for_each_valid_physical_page_range()
947 * add_active_range(node_id, start_pfn, end_pfn)
948 * free_area_init_nodes(max_zone_pfns);
949 *
950 * If the architecture guarantees that there are no holes in the ranges
951 * registered with add_active_range(), free_bootmem_active_regions()
952 * will call free_bootmem_node() for each registered physical page range.
953 * Similarly sparse_memory_present_with_active_regions() calls
954 * memory_present() for each range when SPARSEMEM is enabled.
955 *
956 * See mm/page_alloc.c for more information on each function exposed by
957 * CONFIG_ARCH_POPULATES_NODE_MAP
958 */
959extern void free_area_init_nodes(unsigned long *max_zone_pfn);
960extern void add_active_range(unsigned int nid, unsigned long start_pfn,
961 unsigned long end_pfn);
962extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
963 unsigned long new_end_pfn);
Mel Gormanfb014392006-09-27 01:49:59 -0700964extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
965 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -0700966extern void remove_all_active_ranges(void);
967extern unsigned long absent_pages_in_range(unsigned long start_pfn,
968 unsigned long end_pfn);
969extern void get_pfn_range_for_nid(unsigned int nid,
970 unsigned long *start_pfn, unsigned long *end_pfn);
971extern unsigned long find_min_pfn_with_active_regions(void);
972extern unsigned long find_max_pfn_with_active_regions(void);
973extern void free_bootmem_with_active_regions(int nid,
974 unsigned long max_low_pfn);
975extern void sparse_memory_present_with_active_regions(int nid);
976#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
977extern int early_pfn_to_nid(unsigned long pfn);
978#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
979#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
Mel Gorman0e0b8642006-09-27 01:49:56 -0700980extern void set_dma_reserve(unsigned long new_dma_reserve);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
Dave Hansen3947be12005-10-29 18:16:54 -0700982extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983extern void mem_init(void);
984extern void show_mem(void);
985extern void si_meminfo(struct sysinfo * val);
986extern void si_meminfo_node(struct sysinfo *val, int nid);
987
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700988#ifdef CONFIG_NUMA
989extern void setup_per_cpu_pageset(void);
990#else
991static inline void setup_per_cpu_pageset(void) {}
992#endif
993
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994/* prio_tree.c */
995void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
996void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
997void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
998struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
999 struct prio_tree_iter *iter);
1000
1001#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1002 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1003 (vma = vma_prio_tree_next(vma, iter)); )
1004
1005static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1006 struct list_head *list)
1007{
1008 vma->shared.vm_set.parent = NULL;
1009 list_add_tail(&vma->shared.vm_set.list, list);
1010}
1011
1012/* mmap.c */
1013extern int __vm_enough_memory(long pages, int cap_sys_admin);
1014extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1015 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1016extern struct vm_area_struct *vma_merge(struct mm_struct *,
1017 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1018 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1019 struct mempolicy *);
1020extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1021extern int split_vma(struct mm_struct *,
1022 struct vm_area_struct *, unsigned long addr, int new_below);
1023extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1024extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1025 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001026extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1028 unsigned long addr, unsigned long len, pgoff_t pgoff);
1029extern void exit_mmap(struct mm_struct *);
akpm@osdl.org119f6572005-05-01 08:58:35 -07001030extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031
1032extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1033
1034extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1035 unsigned long len, unsigned long prot,
1036 unsigned long flag, unsigned long pgoff);
1037
1038static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1039 unsigned long len, unsigned long prot,
1040 unsigned long flag, unsigned long offset)
1041{
1042 unsigned long ret = -EINVAL;
1043 if ((offset + PAGE_ALIGN(len)) < offset)
1044 goto out;
1045 if (!(offset & ~PAGE_MASK))
1046 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1047out:
1048 return ret;
1049}
1050
1051extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1052
1053extern unsigned long do_brk(unsigned long, unsigned long);
1054
1055/* filemap.c */
1056extern unsigned long page_unuse(struct page *);
1057extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001058extern void truncate_inode_pages_range(struct address_space *,
1059 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060
1061/* generic vm_area_ops exported for stackable file systems */
1062extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
1063extern int filemap_populate(struct vm_area_struct *, unsigned long,
1064 unsigned long, pgprot_t, unsigned long, int);
1065
1066/* mm/page-writeback.c */
1067int write_one_page(struct page *page, int wait);
1068
1069/* readahead.c */
1070#define VM_MAX_READAHEAD 128 /* kbytes */
1071#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1072#define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1073 * turning readahead off */
1074
1075int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001076 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001078 pgoff_t offset, unsigned long nr_to_read);
1079unsigned long page_cache_readahead(struct address_space *mapping,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 struct file_ra_state *ra,
1081 struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001082 pgoff_t offset,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 unsigned long size);
1084void handle_ra_miss(struct address_space *mapping,
1085 struct file_ra_state *ra, pgoff_t offset);
1086unsigned long max_sane_readahead(unsigned long nr);
1087
1088/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001089extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001090#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001091extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001092#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093
1094/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1095extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1096extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1097 struct vm_area_struct **pprev);
1098
1099/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1100 NULL if none. Assume start_addr < end_addr. */
1101static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1102{
1103 struct vm_area_struct * vma = find_vma(mm,start_addr);
1104
1105 if (vma && end_addr <= vma->vm_start)
1106 vma = NULL;
1107 return vma;
1108}
1109
1110static inline unsigned long vma_pages(struct vm_area_struct *vma)
1111{
1112 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1113}
1114
Hugh Dickins804af2c2006-07-26 21:39:49 +01001115pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001116struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1117struct page *vmalloc_to_page(void *addr);
1118unsigned long vmalloc_to_pfn(void *addr);
1119int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1120 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001121int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122
Linus Torvalds6aab3412005-11-28 14:34:23 -08001123struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001124 unsigned int foll_flags);
1125#define FOLL_WRITE 0x01 /* check pte is writable */
1126#define FOLL_TOUCH 0x02 /* mark page accessed */
1127#define FOLL_GET 0x04 /* do get_page on page */
1128#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129
1130#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001131void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001133static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 unsigned long flags, struct file *file, long pages)
1135{
1136}
1137#endif /* CONFIG_PROC_FS */
1138
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139#ifndef CONFIG_DEBUG_PAGEALLOC
1140static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001141kernel_map_pages(struct page *page, int numpages, int enable) {}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142#endif
1143
1144extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1145#ifdef __HAVE_ARCH_GATE_AREA
1146int in_gate_area_no_task(unsigned long addr);
1147int in_gate_area(struct task_struct *task, unsigned long addr);
1148#else
1149int in_gate_area_no_task(unsigned long addr);
1150#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1151#endif /* __HAVE_ARCH_GATE_AREA */
1152
Andrew Morton9d0243b2006-01-08 01:00:39 -08001153int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1154 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001155unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001156 unsigned long lru_pages);
1157void drop_pagecache(void);
1158void drop_slab(void);
1159
Luke Yang7a9166e2006-02-20 18:28:07 -08001160#ifndef CONFIG_MMU
1161#define randomize_va_space 0
1162#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001163extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001164#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001165
David Howellsf269fdd2006-09-27 01:50:23 -07001166__attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma);
Ingo Molnare6e54942006-06-27 02:53:50 -07001167
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168#endif /* __KERNEL__ */
1169#endif /* _LINUX_MM_H */