blob: 5a6068ff5556f5fcd62566988133aa25d7591de0 [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
117/*
118 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
119 * disabled, then there's a single shared list of VMAs maintained by the
120 * system, and mm's subscribe to these individually
121 */
122struct vm_list_struct {
123 struct vm_list_struct *next;
124 struct vm_area_struct *vma;
125};
126
127#ifndef CONFIG_MMU
128extern struct rb_root nommu_vma_tree;
129extern struct rw_semaphore nommu_vma_sem;
130
131extern unsigned int kobjsize(const void *objp);
132#endif
133
134/*
135 * vm_flags..
136 */
137#define VM_READ 0x00000001 /* currently active flags */
138#define VM_WRITE 0x00000002
139#define VM_EXEC 0x00000004
140#define VM_SHARED 0x00000008
141
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -0700142/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
144#define VM_MAYWRITE 0x00000020
145#define VM_MAYEXEC 0x00000040
146#define VM_MAYSHARE 0x00000080
147
148#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
149#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -0800150#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
152
153#define VM_EXECUTABLE 0x00001000
154#define VM_LOCKED 0x00002000
155#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
156
157 /* Used by sys_madvise() */
158#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
159#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
160
161#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
162#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800163#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
165#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
166#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
167#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800168#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169
170#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
171#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
172#endif
173
174#ifdef CONFIG_STACK_GROWSUP
175#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
176#else
177#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
178#endif
179
180#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
181#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
182#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
183#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
184#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
185
186/*
187 * mapping from the currently active vm_flags protection bits (the
188 * low four bits) to a page protection mask..
189 */
190extern pgprot_t protection_map[16];
191
192
193/*
194 * These are the virtual MM functions - opening of an area, closing and
195 * unmapping it (needed to keep files on disk up-to-date etc), pointer
196 * to the functions called when a no-page or a wp-page exception occurs.
197 */
198struct vm_operations_struct {
199 void (*open)(struct vm_area_struct * area);
200 void (*close)(struct vm_area_struct * area);
201 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
Jes Sorensenf4b81802006-09-27 01:50:10 -0700202 unsigned long (*nopfn)(struct vm_area_struct * area, unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 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 -0700204
205 /* notification that a previously read-only page is about to become
206 * writable, if an error is returned it will cause a SIGBUS */
207 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208#ifdef CONFIG_NUMA
209 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
210 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
211 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700212 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
213 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214#endif
215};
216
217struct mmu_gather;
218struct inode;
219
Andrew Morton349aef02006-01-08 01:04:36 -0800220#define page_private(page) ((page)->private)
221#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700222
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223/*
224 * FIXME: take this include out, include page-flags.h in
225 * files which need it (119 of them)
226 */
227#include <linux/page-flags.h>
228
Nick Piggin725d7042006-09-25 23:30:55 -0700229#ifdef CONFIG_DEBUG_VM
230#define VM_BUG_ON(cond) BUG_ON(cond)
231#else
232#define VM_BUG_ON(condition) do { } while(0)
233#endif
234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235/*
236 * Methods to modify the page usage count.
237 *
238 * What counts for a page usage:
239 * - cache mapping (page->mapping)
240 * - private data (page->private)
241 * - page mapped in a task's page tables, each mapping
242 * is counted separately
243 *
244 * Also, many kernel routines increase the page count before a critical
245 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 */
247
248/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700249 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800251static inline int put_page_testzero(struct page *page)
252{
Nick Piggin725d7042006-09-25 23:30:55 -0700253 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800254 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800255}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256
257/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800258 * Try to grab a ref unless the page has a refcount of zero, return false if
259 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800261static inline int get_page_unless_zero(struct page *page)
262{
Nick Piggin725d7042006-09-25 23:30:55 -0700263 VM_BUG_ON(PageCompound(page));
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800264 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800265}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700267static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268{
Nick Piggin617d2212006-03-22 00:08:43 -0800269 if (unlikely(PageCompound(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700270 page = (struct page *)page_private(page);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800271 return atomic_read(&page->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272}
273
274static inline void get_page(struct page *page)
275{
276 if (unlikely(PageCompound(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700277 page = (struct page *)page_private(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700278 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 atomic_inc(&page->_count);
280}
281
Nick Piggin7835e982006-03-22 00:08:40 -0800282/*
283 * Setup the page count before being freed into the page allocator for
284 * the first time (boot or memory hotplug)
285 */
286static inline void init_page_count(struct page *page)
287{
288 atomic_set(&page->_count, 1);
289}
290
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700292void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800294void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800295
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296/*
297 * Multiple processes may "see" the same page. E.g. for untouched
298 * mappings of /dev/null, all processes see the same page full of
299 * zeroes, and text pages of executables and shared libraries have
300 * only one copy in memory, at most, normally.
301 *
302 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700303 * page_count() == 0 means the page is free. page->lru is then used for
304 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700305 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700307 * Pages are allocated by the slab allocator in order to provide memory
308 * to kmalloc and kmem_cache_alloc. In this case, the management of the
309 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
310 * unless a particular usage is carefully commented. (the responsibility of
311 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700313 * A page may be used by anyone else who does a __get_free_page().
314 * In this case, page_count still tracks the references, and should only
315 * be used through the normal accessor functions. The top bits of page->flags
316 * and page->virtual store page management information, but all other fields
317 * are unused and could be used privately, carefully. The management of this
318 * page is the responsibility of the one who allocated it, and those who have
319 * subsequently been given references to it.
320 *
321 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 * managed by the Linux memory manager: I/O, buffers, swapping etc.
323 * The following discussion applies only to them.
324 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700325 * A pagecache page contains an opaque `private' member, which belongs to the
326 * page's address_space. Usually, this is the address of a circular list of
327 * the page's disk buffers. PG_private must be set to tell the VM to call
328 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700330 * A page may belong to an inode's memory mapping. In this case, page->mapping
331 * is the pointer to the inode, and page->index is the file offset of the page,
332 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700334 * If pagecache pages are not associated with an inode, they are said to be
335 * anonymous pages. These may become associated with the swapcache, and in that
336 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700338 * In either case (swapcache or inode backed), the pagecache itself holds one
339 * reference to the page. Setting PG_private should also increment the
340 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700342 * The pagecache pages are stored in a per-mapping radix tree, which is
343 * rooted at mapping->page_tree, and indexed by offset.
344 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
345 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700347 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 * - inode pages may need to be read from disk,
349 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700350 * to be written back to the inode on disk,
351 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
352 * modified may need to be swapped out to swap space and (later) to be read
353 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 */
355
356/*
357 * The zone field is never updated after free_area_init_core()
358 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700360
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700361
362/*
363 * page->flags layout:
364 *
365 * There are three possibilities for how page->flags get
366 * laid out. The first is for the normal case, without
367 * sparsemem. The second is for sparsemem when there is
368 * plenty of space for node and section. The last is when
369 * we have run out of space and have to fall back to an
370 * alternate (slower) way of determining the node.
371 *
372 * No sparsemem: | NODE | ZONE | ... | FLAGS |
373 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
374 * no space for node: | SECTION | ZONE | ... | FLAGS |
375 */
376#ifdef CONFIG_SPARSEMEM
377#define SECTIONS_WIDTH SECTIONS_SHIFT
378#else
379#define SECTIONS_WIDTH 0
380#endif
381
382#define ZONES_WIDTH ZONES_SHIFT
383
384#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
385#define NODES_WIDTH NODES_SHIFT
386#else
387#define NODES_WIDTH 0
388#endif
389
390/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100391#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700392#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
393#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
394
395/*
396 * We are going to use the flags for the page to node mapping if its in
397 * there. This includes the case where there is no node, so it is implicit.
398 */
399#define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
400
401#ifndef PFN_SECTION_SHIFT
402#define PFN_SECTION_SHIFT 0
403#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700404
405/*
406 * Define the bit shifts to access each section. For non-existant
407 * sections we define the shift as 0; that plus a 0 mask ensures
408 * the compiler will optimise away reference to them.
409 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700410#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
411#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
412#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700413
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700414/* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
415#if FLAGS_HAS_NODE
Dave Hansen348f8b62005-06-23 00:07:40 -0700416#define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700417#else
418#define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
419#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700420#define ZONETABLE_PGSHIFT ZONES_PGSHIFT
421
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700422#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
423#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
Dave Hansen348f8b62005-06-23 00:07:40 -0700424#endif
425
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700426#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
427#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
428#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700429#define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
430
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700431static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432{
Dave Hansen348f8b62005-06-23 00:07:40 -0700433 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
436struct zone;
437extern struct zone *zone_table[];
438
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700439static inline int page_zone_id(struct page *page)
440{
441 return (page->flags >> ZONETABLE_PGSHIFT) & ZONETABLE_MASK;
442}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443static inline struct zone *page_zone(struct page *page)
444{
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700445 return zone_table[page_zone_id(page)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446}
447
Christoph Lameter89fa3022006-09-25 23:31:55 -0700448static inline unsigned long zone_to_nid(struct zone *zone)
449{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700450#ifdef CONFIG_NUMA
451 return zone->node;
452#else
453 return 0;
454#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700455}
456
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700457static inline unsigned long page_to_nid(struct page *page)
458{
459 if (FLAGS_HAS_NODE)
460 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
461 else
Christoph Lameter89fa3022006-09-25 23:31:55 -0700462 return zone_to_nid(page_zone(page));
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700463}
464static inline unsigned long page_to_section(struct page *page)
465{
466 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
467}
468
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700469static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470{
Dave Hansen348f8b62005-06-23 00:07:40 -0700471 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
472 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
473}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700474
Dave Hansen348f8b62005-06-23 00:07:40 -0700475static inline void set_page_node(struct page *page, unsigned long node)
476{
477 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
478 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
479}
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700480static inline void set_page_section(struct page *page, unsigned long section)
481{
482 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
483 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
484}
Dave Hansen348f8b62005-06-23 00:07:40 -0700485
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700486static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700487 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700488{
489 set_page_zone(page, zone);
490 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700491 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492}
493
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700494/*
495 * Some inline functions in vmstat.h depend on page_zone()
496 */
497#include <linux/vmstat.h>
498
Ingo Molnar652050a2006-01-14 13:21:30 -0800499static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500{
501 return __va(page_to_pfn(page) << PAGE_SHIFT);
502}
503
504#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
505#define HASHED_PAGE_VIRTUAL
506#endif
507
508#if defined(WANT_PAGE_VIRTUAL)
509#define page_address(page) ((page)->virtual)
510#define set_page_address(page, address) \
511 do { \
512 (page)->virtual = (address); \
513 } while(0)
514#define page_address_init() do { } while(0)
515#endif
516
517#if defined(HASHED_PAGE_VIRTUAL)
518void *page_address(struct page *page);
519void set_page_address(struct page *page, void *virtual);
520void page_address_init(void);
521#endif
522
523#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
524#define page_address(page) lowmem_page_address(page)
525#define set_page_address(page, address) do { } while(0)
526#define page_address_init() do { } while(0)
527#endif
528
529/*
530 * On an anonymous page mapped into a user virtual memory area,
531 * page->mapping points to its anon_vma, not to a struct address_space;
532 * with the PAGE_MAPPING_ANON bit set to distinguish it.
533 *
534 * Please note that, confusingly, "page_mapping" refers to the inode
535 * address_space which maps the page from disk; whereas "page_mapped"
536 * refers to user virtual address space into which the page is mapped.
537 */
538#define PAGE_MAPPING_ANON 1
539
540extern struct address_space swapper_space;
541static inline struct address_space *page_mapping(struct page *page)
542{
543 struct address_space *mapping = page->mapping;
544
545 if (unlikely(PageSwapCache(page)))
546 mapping = &swapper_space;
547 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
548 mapping = NULL;
549 return mapping;
550}
551
552static inline int PageAnon(struct page *page)
553{
554 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
555}
556
557/*
558 * Return the pagecache index of the passed page. Regular pagecache pages
559 * use ->index whereas swapcache pages use ->private
560 */
561static inline pgoff_t page_index(struct page *page)
562{
563 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700564 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 return page->index;
566}
567
568/*
569 * The atomic page->_mapcount, like _count, starts from -1:
570 * so that transitions both from it and to it can be tracked,
571 * using atomic_inc_and_test and atomic_add_negative(-1).
572 */
573static inline void reset_page_mapcount(struct page *page)
574{
575 atomic_set(&(page)->_mapcount, -1);
576}
577
578static inline int page_mapcount(struct page *page)
579{
580 return atomic_read(&(page)->_mapcount) + 1;
581}
582
583/*
584 * Return true if this page is mapped into pagetables.
585 */
586static inline int page_mapped(struct page *page)
587{
588 return atomic_read(&(page)->_mapcount) >= 0;
589}
590
591/*
592 * Error return values for the *_nopage functions
593 */
594#define NOPAGE_SIGBUS (NULL)
595#define NOPAGE_OOM ((struct page *) (-1))
Benjamin Herrenschmidt7f7bbbe2006-10-06 00:43:53 -0700596#define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597
598/*
Jes Sorensenf4b81802006-09-27 01:50:10 -0700599 * Error return values for the *_nopfn functions
600 */
601#define NOPFN_SIGBUS ((unsigned long) -1)
602#define NOPFN_OOM ((unsigned long) -2)
603
604/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 * Different kinds of faults, as returned by handle_mm_fault().
606 * Used to decide whether a process gets delivered SIGBUS or
607 * just gets major/minor fault counters bumped up.
608 */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000609#define VM_FAULT_OOM 0x00
610#define VM_FAULT_SIGBUS 0x01
611#define VM_FAULT_MINOR 0x02
612#define VM_FAULT_MAJOR 0x03
613
614/*
615 * Special case for get_user_pages.
616 * Must be in a distinct bit from the above VM_FAULT_ flags.
617 */
618#define VM_FAULT_WRITE 0x10
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619
620#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
621
622extern void show_free_areas(void);
623
624#ifdef CONFIG_SHMEM
625struct page *shmem_nopage(struct vm_area_struct *vma,
626 unsigned long address, int *type);
627int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
628struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
629 unsigned long addr);
630int shmem_lock(struct file *file, int lock, struct user_struct *user);
631#else
632#define shmem_nopage filemap_nopage
Russell King03b00eb2006-01-06 00:10:52 -0800633
634static inline int shmem_lock(struct file *file, int lock,
635 struct user_struct *user)
636{
637 return 0;
638}
639
640static inline int shmem_set_policy(struct vm_area_struct *vma,
641 struct mempolicy *new)
642{
643 return 0;
644}
645
646static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
647 unsigned long addr)
648{
649 return NULL;
650}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651#endif
652struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
David Howellsb0e15192006-01-06 00:11:42 -0800653extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
655int shmem_zero_setup(struct vm_area_struct *);
656
David Howellsb0e15192006-01-06 00:11:42 -0800657#ifndef CONFIG_MMU
658extern unsigned long shmem_get_unmapped_area(struct file *file,
659 unsigned long addr,
660 unsigned long len,
661 unsigned long pgoff,
662 unsigned long flags);
663#endif
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665static inline int can_do_mlock(void)
666{
667 if (capable(CAP_IPC_LOCK))
668 return 1;
669 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
670 return 1;
671 return 0;
672}
673extern int user_shm_lock(size_t, struct user_struct *);
674extern void user_shm_unlock(size_t, struct user_struct *);
675
676/*
677 * Parameter block passed down to zap_pte_range in exceptional cases.
678 */
679struct zap_details {
680 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
681 struct address_space *check_mapping; /* Check page->mapping if set */
682 pgoff_t first_index; /* Lowest page->index to unmap */
683 pgoff_t last_index; /* Highest page->index to unmap */
684 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 unsigned long truncate_count; /* Compare vm_truncate_count */
686};
687
Linus Torvalds6aab3412005-11-28 14:34:23 -0800688struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700689unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700691unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 struct vm_area_struct *start_vma, unsigned long start_addr,
693 unsigned long end_addr, unsigned long *nr_accounted,
694 struct zap_details *);
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700695void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
696 unsigned long end, unsigned long floor, unsigned long ceiling);
697void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
Hugh Dickinse0da3822005-04-19 13:29:15 -0700698 unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
700 struct vm_area_struct *vma);
701int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
702 unsigned long size, pgprot_t prot);
703void unmap_mapping_range(struct address_space *mapping,
704 loff_t const holebegin, loff_t const holelen, int even_cows);
705
706static inline void unmap_shared_mapping_range(struct address_space *mapping,
707 loff_t const holebegin, loff_t const holelen)
708{
709 unmap_mapping_range(mapping, holebegin, holelen, 0);
710}
711
712extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800713extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
715extern 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 +1000716
David Howells7ee1dd32006-01-06 00:11:44 -0800717#ifdef CONFIG_MMU
718extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
719 unsigned long address, int write_access);
720
721static inline int handle_mm_fault(struct mm_struct *mm,
722 struct vm_area_struct *vma, unsigned long address,
723 int write_access)
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000724{
David Howells7ee1dd32006-01-06 00:11:44 -0800725 return __handle_mm_fault(mm, vma, address, write_access) &
726 (~VM_FAULT_WRITE);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000727}
David Howells7ee1dd32006-01-06 00:11:44 -0800728#else
729static inline int handle_mm_fault(struct mm_struct *mm,
730 struct vm_area_struct *vma, unsigned long address,
731 int write_access)
732{
733 /* should never happen if there's no MMU */
734 BUG();
735 return VM_FAULT_SIGBUS;
736}
737#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000738
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739extern int make_pages_present(unsigned long addr, unsigned long end);
740extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
741void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
742
743int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
744 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
Nick Pigginb5810032005-10-29 18:16:12 -0700745void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
David Howellscf9a2ae2006-08-29 19:05:54 +0100747extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
748extern void do_invalidatepage(struct page *page, unsigned long offset);
749
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750int __set_page_dirty_nobuffers(struct page *page);
751int redirty_page_for_writepage(struct writeback_control *wbc,
752 struct page *page);
753int FASTCALL(set_page_dirty(struct page *page));
754int set_page_dirty_lock(struct page *page);
755int clear_page_dirty_for_io(struct page *page);
756
757extern unsigned long do_mremap(unsigned long addr,
758 unsigned long old_len, unsigned long new_len,
759 unsigned long flags, unsigned long new_addr);
760
761/*
762 * Prototype to add a shrinker callback for ageable caches.
763 *
764 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
765 * scan `nr_to_scan' objects, attempting to free them.
766 *
Domen Puncer845d3432005-05-05 16:16:14 -0700767 * The callback must return the number of objects which remain in the cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 *
Domen Puncer845d3432005-05-05 16:16:14 -0700769 * The callback will be passed nr_to_scan == 0 when the VM is querying the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 * cache size, so a fastpath for that case is appropriate.
771 */
Al Viro6daa0e22005-10-21 03:18:50 -0400772typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774/*
775 * Add an aging callback. The int is the number of 'seeks' it takes
776 * to recreate one of the objects that these functions age.
777 */
778
779#define DEFAULT_SEEKS 2
780struct shrinker;
781extern struct shrinker *set_shrinker(int, shrinker_t);
782extern void remove_shrinker(struct shrinker *shrinker);
783
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700784/*
785 * Some shared mappigns will want the pages marked read-only
786 * to track write events. If so, we'll downgrade vm_page_prot
787 * to the private version (using protection_map[] without the
788 * VM_SHARED bit).
789 */
790static inline int vma_wants_writenotify(struct vm_area_struct *vma)
791{
792 unsigned int vm_flags = vma->vm_flags;
793
794 /* If it was private or non-writable, the write bit is already clear */
795 if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED)))
796 return 0;
797
798 /* The backer wishes to know when pages are first written to? */
799 if (vma->vm_ops && vma->vm_ops->page_mkwrite)
800 return 1;
801
802 /* The open routine did something to the protections already? */
803 if (pgprot_val(vma->vm_page_prot) !=
804 pgprot_val(protection_map[vm_flags &
805 (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]))
806 return 0;
807
808 /* Specialty mapping? */
809 if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE))
810 return 0;
811
812 /* Can the mapping track the dirty pages? */
813 return vma->vm_file && vma->vm_file->f_mapping &&
814 mapping_cap_account_dirty(vma->vm_file->f_mapping);
815}
816
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800817extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
818
Hugh Dickins1bb36302005-10-29 18:16:22 -0700819int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
820int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
821int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
822int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
823
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824/*
825 * The following ifdef needed to get the 4level-fixup.h header to work.
826 * Remove it when 4level-fixup.h has been removed.
827 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700828#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
830{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700831 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
832 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833}
834
835static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
836{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700837 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
838 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700840#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
841
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700842#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
843/*
844 * We tuck a spinlock to guard each pagetable page into its struct page,
845 * at page->private, with BUILD_BUG_ON to make sure that this will not
846 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
847 * When freeing, reset page->mapping so free_pages_check won't complain.
848 */
Andrew Morton349aef02006-01-08 01:04:36 -0800849#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700850#define pte_lock_init(_page) do { \
851 spin_lock_init(__pte_lockptr(_page)); \
852} while (0)
853#define pte_lock_deinit(page) ((page)->mapping = NULL)
854#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
855#else
856/*
857 * We use mm->page_table_lock to guard all pagetable pages of the mm.
858 */
859#define pte_lock_init(page) do {} while (0)
860#define pte_lock_deinit(page) do {} while (0)
861#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
862#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
863
Hugh Dickinsc74df322005-10-29 18:16:23 -0700864#define pte_offset_map_lock(mm, pmd, address, ptlp) \
865({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700866 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700867 pte_t *__pte = pte_offset_map(pmd, address); \
868 *(ptlp) = __ptl; \
869 spin_lock(__ptl); \
870 __pte; \
871})
872
873#define pte_unmap_unlock(pte, ptl) do { \
874 spin_unlock(ptl); \
875 pte_unmap(pte); \
876} while (0)
877
Hugh Dickins1bb36302005-10-29 18:16:22 -0700878#define pte_alloc_map(mm, pmd, address) \
879 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
880 NULL: pte_offset_map(pmd, address))
881
Hugh Dickinsc74df322005-10-29 18:16:23 -0700882#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
883 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
884 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
885
Hugh Dickins1bb36302005-10-29 18:16:22 -0700886#define pte_alloc_kernel(pmd, address) \
887 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
888 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
890extern void free_area_init(unsigned long * zones_size);
891extern void free_area_init_node(int nid, pg_data_t *pgdat,
892 unsigned long * zones_size, unsigned long zone_start_pfn,
893 unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -0700894#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
895/*
896 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
897 * zones, allocate the backing mem_map and account for memory holes in a more
898 * architecture independent manner. This is a substitute for creating the
899 * zone_sizes[] and zholes_size[] arrays and passing them to
900 * free_area_init_node()
901 *
902 * An architecture is expected to register range of page frames backed by
903 * physical memory with add_active_range() before calling
904 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
905 * usage, an architecture is expected to do something like
906 *
907 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
908 * max_highmem_pfn};
909 * for_each_valid_physical_page_range()
910 * add_active_range(node_id, start_pfn, end_pfn)
911 * free_area_init_nodes(max_zone_pfns);
912 *
913 * If the architecture guarantees that there are no holes in the ranges
914 * registered with add_active_range(), free_bootmem_active_regions()
915 * will call free_bootmem_node() for each registered physical page range.
916 * Similarly sparse_memory_present_with_active_regions() calls
917 * memory_present() for each range when SPARSEMEM is enabled.
918 *
919 * See mm/page_alloc.c for more information on each function exposed by
920 * CONFIG_ARCH_POPULATES_NODE_MAP
921 */
922extern void free_area_init_nodes(unsigned long *max_zone_pfn);
923extern void add_active_range(unsigned int nid, unsigned long start_pfn,
924 unsigned long end_pfn);
925extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
926 unsigned long new_end_pfn);
Mel Gormanfb014392006-09-27 01:49:59 -0700927extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
928 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -0700929extern void remove_all_active_ranges(void);
930extern unsigned long absent_pages_in_range(unsigned long start_pfn,
931 unsigned long end_pfn);
932extern void get_pfn_range_for_nid(unsigned int nid,
933 unsigned long *start_pfn, unsigned long *end_pfn);
934extern unsigned long find_min_pfn_with_active_regions(void);
935extern unsigned long find_max_pfn_with_active_regions(void);
936extern void free_bootmem_with_active_regions(int nid,
937 unsigned long max_low_pfn);
938extern void sparse_memory_present_with_active_regions(int nid);
939#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
940extern int early_pfn_to_nid(unsigned long pfn);
941#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
942#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
Mel Gorman0e0b8642006-09-27 01:49:56 -0700943extern void set_dma_reserve(unsigned long new_dma_reserve);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
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);
Keith Manntheyf28c5ed2006-09-30 23:27:04 -0700950extern void zonetable_add(struct zone *zone, int nid, enum zone_type zid,
951 unsigned long pfn, unsigned long size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700953#ifdef CONFIG_NUMA
954extern void setup_per_cpu_pageset(void);
955#else
956static inline void setup_per_cpu_pageset(void) {}
957#endif
958
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959/* prio_tree.c */
960void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
961void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
962void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
963struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
964 struct prio_tree_iter *iter);
965
966#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
967 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
968 (vma = vma_prio_tree_next(vma, iter)); )
969
970static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
971 struct list_head *list)
972{
973 vma->shared.vm_set.parent = NULL;
974 list_add_tail(&vma->shared.vm_set.list, list);
975}
976
977/* mmap.c */
978extern int __vm_enough_memory(long pages, int cap_sys_admin);
979extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
980 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
981extern struct vm_area_struct *vma_merge(struct mm_struct *,
982 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
983 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
984 struct mempolicy *);
985extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
986extern int split_vma(struct mm_struct *,
987 struct vm_area_struct *, unsigned long addr, int new_below);
988extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
989extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
990 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -0700991extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
993 unsigned long addr, unsigned long len, pgoff_t pgoff);
994extern void exit_mmap(struct mm_struct *);
akpm@osdl.org119f6572005-05-01 08:58:35 -0700995extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996
997extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
998
999extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1000 unsigned long len, unsigned long prot,
1001 unsigned long flag, unsigned long pgoff);
1002
1003static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1004 unsigned long len, unsigned long prot,
1005 unsigned long flag, unsigned long offset)
1006{
1007 unsigned long ret = -EINVAL;
1008 if ((offset + PAGE_ALIGN(len)) < offset)
1009 goto out;
1010 if (!(offset & ~PAGE_MASK))
1011 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1012out:
1013 return ret;
1014}
1015
1016extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1017
1018extern unsigned long do_brk(unsigned long, unsigned long);
1019
1020/* filemap.c */
1021extern unsigned long page_unuse(struct page *);
1022extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001023extern void truncate_inode_pages_range(struct address_space *,
1024 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025
1026/* generic vm_area_ops exported for stackable file systems */
1027extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
1028extern int filemap_populate(struct vm_area_struct *, unsigned long,
1029 unsigned long, pgprot_t, unsigned long, int);
1030
1031/* mm/page-writeback.c */
1032int write_one_page(struct page *page, int wait);
1033
1034/* readahead.c */
1035#define VM_MAX_READAHEAD 128 /* kbytes */
1036#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1037#define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1038 * turning readahead off */
1039
1040int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001041 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001043 pgoff_t offset, unsigned long nr_to_read);
1044unsigned long page_cache_readahead(struct address_space *mapping,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 struct file_ra_state *ra,
1046 struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001047 pgoff_t offset,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 unsigned long size);
1049void handle_ra_miss(struct address_space *mapping,
1050 struct file_ra_state *ra, pgoff_t offset);
1051unsigned long max_sane_readahead(unsigned long nr);
1052
1053/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001054extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001055#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001056extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001057#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
1059/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1060extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1061extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1062 struct vm_area_struct **pprev);
1063
1064/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1065 NULL if none. Assume start_addr < end_addr. */
1066static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1067{
1068 struct vm_area_struct * vma = find_vma(mm,start_addr);
1069
1070 if (vma && end_addr <= vma->vm_start)
1071 vma = NULL;
1072 return vma;
1073}
1074
1075static inline unsigned long vma_pages(struct vm_area_struct *vma)
1076{
1077 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1078}
1079
Hugh Dickins804af2c2006-07-26 21:39:49 +01001080pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001081struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1082struct page *vmalloc_to_page(void *addr);
1083unsigned long vmalloc_to_pfn(void *addr);
1084int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1085 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001086int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087
Linus Torvalds6aab3412005-11-28 14:34:23 -08001088struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001089 unsigned int foll_flags);
1090#define FOLL_WRITE 0x01 /* check pte is writable */
1091#define FOLL_TOUCH 0x02 /* mark page accessed */
1092#define FOLL_GET 0x04 /* do get_page on page */
1093#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
1095#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001096void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001098static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 unsigned long flags, struct file *file, long pages)
1100{
1101}
1102#endif /* CONFIG_PROC_FS */
1103
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104#ifndef CONFIG_DEBUG_PAGEALLOC
1105static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001106kernel_map_pages(struct page *page, int numpages, int enable) {}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107#endif
1108
1109extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1110#ifdef __HAVE_ARCH_GATE_AREA
1111int in_gate_area_no_task(unsigned long addr);
1112int in_gate_area(struct task_struct *task, unsigned long addr);
1113#else
1114int in_gate_area_no_task(unsigned long addr);
1115#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1116#endif /* __HAVE_ARCH_GATE_AREA */
1117
Andrea Arcangeli79befd02005-04-16 15:24:05 -07001118/* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1119#define OOM_DISABLE -17
1120
Andrew Morton9d0243b2006-01-08 01:00:39 -08001121int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1122 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001123unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001124 unsigned long lru_pages);
1125void drop_pagecache(void);
1126void drop_slab(void);
1127
Luke Yang7a9166e2006-02-20 18:28:07 -08001128#ifndef CONFIG_MMU
1129#define randomize_va_space 0
1130#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001131extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001132#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001133
David Howellsf269fdd2006-09-27 01:50:23 -07001134__attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma);
Ingo Molnare6e54942006-06-27 02:53:50 -07001135
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136#endif /* __KERNEL__ */
1137#endif /* _LINUX_MM_H */