vmscan: respect higher order in zone_reclaim()
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / include / linux / mm.h
CommitLineData
1da177e4
LT
1#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
1da177e4
LT
4#include <linux/errno.h>
5
6#ifdef __KERNEL__
7
1da177e4
LT
8#include <linux/gfp.h>
9#include <linux/list.h>
59ea7463 10#include <linux/mmdebug.h>
1da177e4
LT
11#include <linux/mmzone.h>
12#include <linux/rbtree.h>
13#include <linux/prio_tree.h>
9a11b49a 14#include <linux/debug_locks.h>
5b99cd0e 15#include <linux/mm_types.h>
1da177e4
LT
16
17struct mempolicy;
18struct anon_vma;
4e950f6f 19struct file_ra_state;
e8edc6e0 20struct user_struct;
4e950f6f 21struct writeback_control;
1da177e4
LT
22
23#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
24extern unsigned long max_mapnr;
25#endif
26
27extern unsigned long num_physpages;
28extern void * high_memory;
1da177e4
LT
29extern int page_cluster;
30
31#ifdef CONFIG_SYSCTL
32extern int sysctl_legacy_va_layout;
33#else
34#define sysctl_legacy_va_layout 0
35#endif
36
42d7896e
JM
37extern unsigned long mmap_min_addr;
38
1da177e4
LT
39#include <asm/page.h>
40#include <asm/pgtable.h>
41#include <asm/processor.h>
1da177e4 42
1da177e4
LT
43#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
44
27ac792c
AR
45/* to align the pointer to the (next) page boundary */
46#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
47
1da177e4
LT
48/*
49 * Linux kernel virtual memory manager primitives.
50 * The idea being to have a "virtual" mm in the same way
51 * we have a virtual fs - giving a cleaner interface to the
52 * mm details, and allowing different kinds of memory mappings
53 * (from shared memory to executable loading to arbitrary
54 * mmap() functions).
55 */
56
c43692e8
CL
57extern struct kmem_cache *vm_area_cachep;
58
1da177e4 59#ifndef CONFIG_MMU
8feae131
DH
60extern struct rb_root nommu_region_tree;
61extern struct rw_semaphore nommu_region_sem;
1da177e4
LT
62
63extern unsigned int kobjsize(const void *objp);
64#endif
65
66/*
605d9288 67 * vm_flags in vm_area_struct, see mm_types.h.
1da177e4
LT
68 */
69#define VM_READ 0x00000001 /* currently active flags */
70#define VM_WRITE 0x00000002
71#define VM_EXEC 0x00000004
72#define VM_SHARED 0x00000008
73
7e2cff42 74/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
1da177e4
LT
75#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
76#define VM_MAYWRITE 0x00000020
77#define VM_MAYEXEC 0x00000040
78#define VM_MAYSHARE 0x00000080
79
80#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
81#define VM_GROWSUP 0x00000200
6aab341e 82#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
1da177e4
LT
83#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
84
85#define VM_EXECUTABLE 0x00001000
86#define VM_LOCKED 0x00002000
87#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
88
89 /* Used by sys_madvise() */
90#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
91#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
92
93#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
94#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
0b14c179 95#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
1da177e4 96#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
cdfd4325 97#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
1da177e4
LT
98#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
99#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
100#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
895791da 101#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
e5b97dde 102#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
d00806b1 103
d0217ac0 104#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
b379d790 105#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
aba46c50 106#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
895791da 107#define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
1da177e4
LT
108
109#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
110#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
111#endif
112
113#ifdef CONFIG_STACK_GROWSUP
114#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
115#else
116#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
117#endif
118
119#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
120#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
121#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
122#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
123#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
124
b291f000
NP
125/*
126 * special vmas that are non-mergable, non-mlock()able
127 */
128#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
129
1da177e4
LT
130/*
131 * mapping from the currently active vm_flags protection bits (the
132 * low four bits) to a page protection mask..
133 */
134extern pgprot_t protection_map[16];
135
d0217ac0
NP
136#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
137#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
138
6bd9cd50 139/*
140 * This interface is used by x86 PAT code to identify a pfn mapping that is
141 * linear over entire vma. This is to optimize PAT code that deals with
142 * marking the physical region with a particular prot. This is not for generic
143 * mm use. Note also that this check will not work if the pfn mapping is
144 * linear for a vma starting at physical address 0. In which case PAT code
145 * falls back to slow path of reserving physical range page by page.
146 */
3c8bb73a 147static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
148{
895791da 149 return (vma->vm_flags & VM_PFN_AT_MMAP);
3c8bb73a 150}
151
152static inline int is_pfn_mapping(struct vm_area_struct *vma)
153{
154 return (vma->vm_flags & VM_PFNMAP);
155}
d0217ac0 156
54cb8821 157/*
d0217ac0 158 * vm_fault is filled by the the pagefault handler and passed to the vma's
83c54070
NP
159 * ->fault function. The vma's ->fault is responsible for returning a bitmask
160 * of VM_FAULT_xxx flags that give details about how the fault was handled.
54cb8821 161 *
d0217ac0
NP
162 * pgoff should be used in favour of virtual_address, if possible. If pgoff
163 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
164 * mapping support.
54cb8821 165 */
d0217ac0
NP
166struct vm_fault {
167 unsigned int flags; /* FAULT_FLAG_xxx flags */
168 pgoff_t pgoff; /* Logical page offset based on vma */
169 void __user *virtual_address; /* Faulting virtual address */
170
171 struct page *page; /* ->fault handlers should return a
83c54070 172 * page here, unless VM_FAULT_NOPAGE
d0217ac0 173 * is set (which is also implied by
83c54070 174 * VM_FAULT_ERROR).
d0217ac0 175 */
54cb8821 176};
1da177e4
LT
177
178/*
179 * These are the virtual MM functions - opening of an area, closing and
180 * unmapping it (needed to keep files on disk up-to-date etc), pointer
181 * to the functions called when a no-page or a wp-page exception occurs.
182 */
183struct vm_operations_struct {
184 void (*open)(struct vm_area_struct * area);
185 void (*close)(struct vm_area_struct * area);
d0217ac0 186 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
9637a5ef
DH
187
188 /* notification that a previously read-only page is about to become
189 * writable, if an error is returned it will cause a SIGBUS */
190 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
28b2ee20
RR
191
192 /* called by access_process_vm when get_user_pages() fails, typically
193 * for use by special VMAs that can switch between memory and hardware
194 */
195 int (*access)(struct vm_area_struct *vma, unsigned long addr,
196 void *buf, int len, int write);
1da177e4 197#ifdef CONFIG_NUMA
a6020ed7
LS
198 /*
199 * set_policy() op must add a reference to any non-NULL @new mempolicy
200 * to hold the policy upon return. Caller should pass NULL @new to
201 * remove a policy and fall back to surrounding context--i.e. do not
202 * install a MPOL_DEFAULT policy, nor the task or system default
203 * mempolicy.
204 */
1da177e4 205 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
a6020ed7
LS
206
207 /*
208 * get_policy() op must add reference [mpol_get()] to any policy at
209 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
210 * in mm/mempolicy.c will do this automatically.
211 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
212 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
213 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
214 * must return NULL--i.e., do not "fallback" to task or system default
215 * policy.
216 */
1da177e4
LT
217 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
218 unsigned long addr);
7b2259b3
CL
219 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
220 const nodemask_t *to, unsigned long flags);
1da177e4
LT
221#endif
222};
223
224struct mmu_gather;
225struct inode;
226
349aef0b
AM
227#define page_private(page) ((page)->private)
228#define set_page_private(page, v) ((page)->private = (v))
4c21e2f2 229
1da177e4
LT
230/*
231 * FIXME: take this include out, include page-flags.h in
232 * files which need it (119 of them)
233 */
234#include <linux/page-flags.h>
235
236/*
237 * Methods to modify the page usage count.
238 *
239 * What counts for a page usage:
240 * - cache mapping (page->mapping)
241 * - private data (page->private)
242 * - page mapped in a task's page tables, each mapping
243 * is counted separately
244 *
245 * Also, many kernel routines increase the page count before a critical
246 * routine so they can be sure the page doesn't go away from under them.
1da177e4
LT
247 */
248
249/*
da6052f7 250 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1da177e4 251 */
7c8ee9a8
NP
252static inline int put_page_testzero(struct page *page)
253{
725d704e 254 VM_BUG_ON(atomic_read(&page->_count) == 0);
8dc04efb 255 return atomic_dec_and_test(&page->_count);
7c8ee9a8 256}
1da177e4
LT
257
258/*
7c8ee9a8
NP
259 * Try to grab a ref unless the page has a refcount of zero, return false if
260 * that is the case.
1da177e4 261 */
7c8ee9a8
NP
262static inline int get_page_unless_zero(struct page *page)
263{
8dc04efb 264 return atomic_inc_not_zero(&page->_count);
7c8ee9a8 265}
1da177e4 266
48667e7a 267/* Support for virtually mapped pages */
b3bdda02
CL
268struct page *vmalloc_to_page(const void *addr);
269unsigned long vmalloc_to_pfn(const void *addr);
48667e7a 270
0738c4bb
PM
271/*
272 * Determine if an address is within the vmalloc range
273 *
274 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
275 * is no special casing required.
276 */
9e2779fa
CL
277static inline int is_vmalloc_addr(const void *x)
278{
0738c4bb 279#ifdef CONFIG_MMU
9e2779fa
CL
280 unsigned long addr = (unsigned long)x;
281
282 return addr >= VMALLOC_START && addr < VMALLOC_END;
0738c4bb
PM
283#else
284 return 0;
8ca3ed87 285#endif
0738c4bb 286}
9e2779fa 287
d85f3385
CL
288static inline struct page *compound_head(struct page *page)
289{
6d777953 290 if (unlikely(PageTail(page)))
d85f3385
CL
291 return page->first_page;
292 return page;
293}
294
4c21e2f2 295static inline int page_count(struct page *page)
1da177e4 296{
d85f3385 297 return atomic_read(&compound_head(page)->_count);
1da177e4
LT
298}
299
300static inline void get_page(struct page *page)
301{
d85f3385 302 page = compound_head(page);
725d704e 303 VM_BUG_ON(atomic_read(&page->_count) == 0);
1da177e4
LT
304 atomic_inc(&page->_count);
305}
306
b49af68f
CL
307static inline struct page *virt_to_head_page(const void *x)
308{
309 struct page *page = virt_to_page(x);
310 return compound_head(page);
311}
312
7835e98b
NP
313/*
314 * Setup the page count before being freed into the page allocator for
315 * the first time (boot or memory hotplug)
316 */
317static inline void init_page_count(struct page *page)
318{
319 atomic_set(&page->_count, 1);
320}
321
1da177e4 322void put_page(struct page *page);
1d7ea732 323void put_pages_list(struct list_head *pages);
1da177e4 324
8dfcc9ba 325void split_page(struct page *page, unsigned int order);
8dfcc9ba 326
33f2ef89
AW
327/*
328 * Compound pages have a destructor function. Provide a
329 * prototype for that function and accessor functions.
330 * These are _only_ valid on the head of a PG_compound page.
331 */
332typedef void compound_page_dtor(struct page *);
333
334static inline void set_compound_page_dtor(struct page *page,
335 compound_page_dtor *dtor)
336{
337 page[1].lru.next = (void *)dtor;
338}
339
340static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
341{
342 return (compound_page_dtor *)page[1].lru.next;
343}
344
d85f3385
CL
345static inline int compound_order(struct page *page)
346{
6d777953 347 if (!PageHead(page))
d85f3385
CL
348 return 0;
349 return (unsigned long)page[1].lru.prev;
350}
351
352static inline void set_compound_order(struct page *page, unsigned long order)
353{
354 page[1].lru.prev = (void *)order;
355}
356
1da177e4
LT
357/*
358 * Multiple processes may "see" the same page. E.g. for untouched
359 * mappings of /dev/null, all processes see the same page full of
360 * zeroes, and text pages of executables and shared libraries have
361 * only one copy in memory, at most, normally.
362 *
363 * For the non-reserved pages, page_count(page) denotes a reference count.
7e871b6c
PBG
364 * page_count() == 0 means the page is free. page->lru is then used for
365 * freelist management in the buddy allocator.
da6052f7 366 * page_count() > 0 means the page has been allocated.
1da177e4 367 *
da6052f7
NP
368 * Pages are allocated by the slab allocator in order to provide memory
369 * to kmalloc and kmem_cache_alloc. In this case, the management of the
370 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
371 * unless a particular usage is carefully commented. (the responsibility of
372 * freeing the kmalloc memory is the caller's, of course).
1da177e4 373 *
da6052f7
NP
374 * A page may be used by anyone else who does a __get_free_page().
375 * In this case, page_count still tracks the references, and should only
376 * be used through the normal accessor functions. The top bits of page->flags
377 * and page->virtual store page management information, but all other fields
378 * are unused and could be used privately, carefully. The management of this
379 * page is the responsibility of the one who allocated it, and those who have
380 * subsequently been given references to it.
381 *
382 * The other pages (we may call them "pagecache pages") are completely
1da177e4
LT
383 * managed by the Linux memory manager: I/O, buffers, swapping etc.
384 * The following discussion applies only to them.
385 *
da6052f7
NP
386 * A pagecache page contains an opaque `private' member, which belongs to the
387 * page's address_space. Usually, this is the address of a circular list of
388 * the page's disk buffers. PG_private must be set to tell the VM to call
389 * into the filesystem to release these pages.
1da177e4 390 *
da6052f7
NP
391 * A page may belong to an inode's memory mapping. In this case, page->mapping
392 * is the pointer to the inode, and page->index is the file offset of the page,
393 * in units of PAGE_CACHE_SIZE.
1da177e4 394 *
da6052f7
NP
395 * If pagecache pages are not associated with an inode, they are said to be
396 * anonymous pages. These may become associated with the swapcache, and in that
397 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1da177e4 398 *
da6052f7
NP
399 * In either case (swapcache or inode backed), the pagecache itself holds one
400 * reference to the page. Setting PG_private should also increment the
401 * refcount. The each user mapping also has a reference to the page.
1da177e4 402 *
da6052f7
NP
403 * The pagecache pages are stored in a per-mapping radix tree, which is
404 * rooted at mapping->page_tree, and indexed by offset.
405 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
406 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1da177e4 407 *
da6052f7 408 * All pagecache pages may be subject to I/O:
1da177e4
LT
409 * - inode pages may need to be read from disk,
410 * - inode pages which have been modified and are MAP_SHARED may need
da6052f7
NP
411 * to be written back to the inode on disk,
412 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
413 * modified may need to be swapped out to swap space and (later) to be read
414 * back into memory.
1da177e4
LT
415 */
416
417/*
418 * The zone field is never updated after free_area_init_core()
419 * sets it, so none of the operations on it need to be atomic.
1da177e4 420 */
348f8b6c 421
d41dee36
AW
422
423/*
424 * page->flags layout:
425 *
426 * There are three possibilities for how page->flags get
427 * laid out. The first is for the normal case, without
428 * sparsemem. The second is for sparsemem when there is
429 * plenty of space for node and section. The last is when
430 * we have run out of space and have to fall back to an
431 * alternate (slower) way of determining the node.
432 *
308c05e3
CL
433 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
434 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
435 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
d41dee36 436 */
308c05e3 437#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
d41dee36
AW
438#define SECTIONS_WIDTH SECTIONS_SHIFT
439#else
440#define SECTIONS_WIDTH 0
441#endif
442
443#define ZONES_WIDTH ZONES_SHIFT
444
9223b419 445#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
d41dee36
AW
446#define NODES_WIDTH NODES_SHIFT
447#else
308c05e3
CL
448#ifdef CONFIG_SPARSEMEM_VMEMMAP
449#error "Vmemmap: No space for nodes field in page flags"
450#endif
d41dee36
AW
451#define NODES_WIDTH 0
452#endif
453
454/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
07808b74 455#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
d41dee36
AW
456#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
457#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
458
459/*
460 * We are going to use the flags for the page to node mapping if its in
461 * there. This includes the case where there is no node, so it is implicit.
462 */
89689ae7
CL
463#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
464#define NODE_NOT_IN_PAGE_FLAGS
465#endif
d41dee36
AW
466
467#ifndef PFN_SECTION_SHIFT
468#define PFN_SECTION_SHIFT 0
469#endif
348f8b6c
DH
470
471/*
472 * Define the bit shifts to access each section. For non-existant
473 * sections we define the shift as 0; that plus a 0 mask ensures
474 * the compiler will optimise away reference to them.
475 */
d41dee36
AW
476#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
477#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
478#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
348f8b6c 479
89689ae7
CL
480/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
481#ifdef NODE_NOT_IN_PAGEFLAGS
482#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
bd8029b6
AW
483#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
484 SECTIONS_PGOFF : ZONES_PGOFF)
d41dee36 485#else
89689ae7 486#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
bd8029b6
AW
487#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
488 NODES_PGOFF : ZONES_PGOFF)
89689ae7
CL
489#endif
490
bd8029b6 491#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
348f8b6c 492
9223b419
CL
493#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
494#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
348f8b6c
DH
495#endif
496
d41dee36
AW
497#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
498#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
499#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
89689ae7 500#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
348f8b6c 501
2f1b6248 502static inline enum zone_type page_zonenum(struct page *page)
1da177e4 503{
348f8b6c 504 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1da177e4 505}
1da177e4 506
89689ae7
CL
507/*
508 * The identification function is only used by the buddy allocator for
509 * determining if two pages could be buddies. We are not really
510 * identifying a zone since we could be using a the section number
511 * id if we have not node id available in page flags.
512 * We guarantee only that it will return the same value for two
513 * combinable pages in a zone.
514 */
cb2b95e1
AW
515static inline int page_zone_id(struct page *page)
516{
89689ae7 517 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
348f8b6c
DH
518}
519
25ba77c1 520static inline int zone_to_nid(struct zone *zone)
89fa3024 521{
d5f541ed
CL
522#ifdef CONFIG_NUMA
523 return zone->node;
524#else
525 return 0;
526#endif
89fa3024
CL
527}
528
89689ae7 529#ifdef NODE_NOT_IN_PAGE_FLAGS
25ba77c1 530extern int page_to_nid(struct page *page);
89689ae7 531#else
25ba77c1 532static inline int page_to_nid(struct page *page)
d41dee36 533{
89689ae7 534 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
d41dee36 535}
89689ae7
CL
536#endif
537
538static inline struct zone *page_zone(struct page *page)
539{
540 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
541}
542
308c05e3 543#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
d41dee36
AW
544static inline unsigned long page_to_section(struct page *page)
545{
546 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
547}
308c05e3 548#endif
d41dee36 549
2f1b6248 550static inline void set_page_zone(struct page *page, enum zone_type zone)
348f8b6c
DH
551{
552 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
553 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
554}
2f1b6248 555
348f8b6c
DH
556static inline void set_page_node(struct page *page, unsigned long node)
557{
558 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
559 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1da177e4 560}
89689ae7 561
d41dee36
AW
562static inline void set_page_section(struct page *page, unsigned long section)
563{
564 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
565 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
566}
1da177e4 567
2f1b6248 568static inline void set_page_links(struct page *page, enum zone_type zone,
d41dee36 569 unsigned long node, unsigned long pfn)
1da177e4 570{
348f8b6c
DH
571 set_page_zone(page, zone);
572 set_page_node(page, node);
d41dee36 573 set_page_section(page, pfn_to_section_nr(pfn));
1da177e4
LT
574}
575
7cd94146
EP
576/*
577 * If a hint addr is less than mmap_min_addr change hint to be as
578 * low as possible but still greater than mmap_min_addr
579 */
580static inline unsigned long round_hint_to_min(unsigned long hint)
581{
582#ifdef CONFIG_SECURITY
583 hint &= PAGE_MASK;
584 if (((void *)hint != NULL) &&
585 (hint < mmap_min_addr))
586 return PAGE_ALIGN(mmap_min_addr);
587#endif
588 return hint;
589}
590
f6ac2354
CL
591/*
592 * Some inline functions in vmstat.h depend on page_zone()
593 */
594#include <linux/vmstat.h>
595
652050ae 596static __always_inline void *lowmem_page_address(struct page *page)
1da177e4
LT
597{
598 return __va(page_to_pfn(page) << PAGE_SHIFT);
599}
600
601#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
602#define HASHED_PAGE_VIRTUAL
603#endif
604
605#if defined(WANT_PAGE_VIRTUAL)
606#define page_address(page) ((page)->virtual)
607#define set_page_address(page, address) \
608 do { \
609 (page)->virtual = (address); \
610 } while(0)
611#define page_address_init() do { } while(0)
612#endif
613
614#if defined(HASHED_PAGE_VIRTUAL)
615void *page_address(struct page *page);
616void set_page_address(struct page *page, void *virtual);
617void page_address_init(void);
618#endif
619
620#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
621#define page_address(page) lowmem_page_address(page)
622#define set_page_address(page, address) do { } while(0)
623#define page_address_init() do { } while(0)
624#endif
625
626/*
627 * On an anonymous page mapped into a user virtual memory area,
628 * page->mapping points to its anon_vma, not to a struct address_space;
629 * with the PAGE_MAPPING_ANON bit set to distinguish it.
630 *
631 * Please note that, confusingly, "page_mapping" refers to the inode
632 * address_space which maps the page from disk; whereas "page_mapped"
633 * refers to user virtual address space into which the page is mapped.
634 */
635#define PAGE_MAPPING_ANON 1
636
637extern struct address_space swapper_space;
638static inline struct address_space *page_mapping(struct page *page)
639{
640 struct address_space *mapping = page->mapping;
641
b5fab14e 642 VM_BUG_ON(PageSlab(page));
726b8012 643#ifdef CONFIG_SWAP
1da177e4
LT
644 if (unlikely(PageSwapCache(page)))
645 mapping = &swapper_space;
726b8012
AM
646 else
647#endif
648 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
1da177e4
LT
649 mapping = NULL;
650 return mapping;
651}
652
653static inline int PageAnon(struct page *page)
654{
655 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
656}
657
658/*
659 * Return the pagecache index of the passed page. Regular pagecache pages
660 * use ->index whereas swapcache pages use ->private
661 */
662static inline pgoff_t page_index(struct page *page)
663{
664 if (unlikely(PageSwapCache(page)))
4c21e2f2 665 return page_private(page);
1da177e4
LT
666 return page->index;
667}
668
669/*
670 * The atomic page->_mapcount, like _count, starts from -1:
671 * so that transitions both from it and to it can be tracked,
672 * using atomic_inc_and_test and atomic_add_negative(-1).
673 */
674static inline void reset_page_mapcount(struct page *page)
675{
676 atomic_set(&(page)->_mapcount, -1);
677}
678
679static inline int page_mapcount(struct page *page)
680{
681 return atomic_read(&(page)->_mapcount) + 1;
682}
683
684/*
685 * Return true if this page is mapped into pagetables.
686 */
687static inline int page_mapped(struct page *page)
688{
689 return atomic_read(&(page)->_mapcount) >= 0;
690}
691
1da177e4
LT
692/*
693 * Different kinds of faults, as returned by handle_mm_fault().
694 * Used to decide whether a process gets delivered SIGBUS or
695 * just gets major/minor fault counters bumped up.
696 */
d0217ac0 697
83c54070 698#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
d0217ac0 699
83c54070
NP
700#define VM_FAULT_OOM 0x0001
701#define VM_FAULT_SIGBUS 0x0002
702#define VM_FAULT_MAJOR 0x0004
703#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
f33ea7f4 704
83c54070
NP
705#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
706#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1da177e4 707
83c54070 708#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
d0217ac0 709
1c0fe6e3
NP
710/*
711 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
712 */
713extern void pagefault_out_of_memory(void);
714
1da177e4
LT
715#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
716
717extern void show_free_areas(void);
718
719#ifdef CONFIG_SHMEM
89e004ea 720extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
1da177e4 721#else
03b00ebc 722static inline int shmem_lock(struct file *file, int lock,
89e004ea 723 struct user_struct *user)
03b00ebc
RK
724{
725 return 0;
726}
1da177e4
LT
727#endif
728struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
729
730int shmem_zero_setup(struct vm_area_struct *);
731
b0e15190
DH
732#ifndef CONFIG_MMU
733extern unsigned long shmem_get_unmapped_area(struct file *file,
734 unsigned long addr,
735 unsigned long len,
736 unsigned long pgoff,
737 unsigned long flags);
738#endif
739
e8edc6e0 740extern int can_do_mlock(void);
1da177e4
LT
741extern int user_shm_lock(size_t, struct user_struct *);
742extern void user_shm_unlock(size_t, struct user_struct *);
743
744/*
745 * Parameter block passed down to zap_pte_range in exceptional cases.
746 */
747struct zap_details {
748 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
749 struct address_space *check_mapping; /* Check page->mapping if set */
750 pgoff_t first_index; /* Lowest page->index to unmap */
751 pgoff_t last_index; /* Highest page->index to unmap */
752 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
1da177e4
LT
753 unsigned long truncate_count; /* Compare vm_truncate_count */
754};
755
7e675137
NP
756struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
757 pte_t pte);
758
c627f9cc
JS
759int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
760 unsigned long size);
ee39b37b 761unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
1da177e4 762 unsigned long size, struct zap_details *);
508034a3 763unsigned long unmap_vmas(struct mmu_gather **tlb,
1da177e4
LT
764 struct vm_area_struct *start_vma, unsigned long start_addr,
765 unsigned long end_addr, unsigned long *nr_accounted,
766 struct zap_details *);
e6473092
MM
767
768/**
769 * mm_walk - callbacks for walk_page_range
770 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
771 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
772 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
773 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
774 * @pte_hole: if set, called for each hole at all levels
775 *
776 * (see walk_page_range for more details)
777 */
778struct mm_walk {
2165009b
DH
779 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
780 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
781 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
782 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
783 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
784 struct mm_struct *mm;
785 void *private;
e6473092
MM
786};
787
2165009b
DH
788int walk_page_range(unsigned long addr, unsigned long end,
789 struct mm_walk *walk);
42b77728 790void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
3bf5ee95 791 unsigned long end, unsigned long floor, unsigned long ceiling);
1da177e4
LT
792int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
793 struct vm_area_struct *vma);
1da177e4
LT
794void unmap_mapping_range(struct address_space *mapping,
795 loff_t const holebegin, loff_t const holelen, int even_cows);
d87fe660 796int follow_phys(struct vm_area_struct *vma, unsigned long address,
797 unsigned int flags, unsigned long *prot, resource_size_t *phys);
28b2ee20
RR
798int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
799 void *buf, int len, int write);
1da177e4
LT
800
801static inline void unmap_shared_mapping_range(struct address_space *mapping,
802 loff_t const holebegin, loff_t const holelen)
803{
804 unmap_mapping_range(mapping, holebegin, holelen, 0);
805}
806
807extern int vmtruncate(struct inode * inode, loff_t offset);
f6b3ec23 808extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
f33ea7f4 809
7ee1dd3f 810#ifdef CONFIG_MMU
83c54070 811extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
7ee1dd3f 812 unsigned long address, int write_access);
7ee1dd3f
DH
813#else
814static inline int handle_mm_fault(struct mm_struct *mm,
815 struct vm_area_struct *vma, unsigned long address,
816 int write_access)
817{
818 /* should never happen if there's no MMU */
819 BUG();
820 return VM_FAULT_SIGBUS;
821}
822#endif
f33ea7f4 823
1da177e4
LT
824extern int make_pages_present(unsigned long addr, unsigned long end);
825extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1da177e4
LT
826
827int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
828 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
829
cf9a2ae8
DH
830extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
831extern void do_invalidatepage(struct page *page, unsigned long offset);
832
1da177e4 833int __set_page_dirty_nobuffers(struct page *page);
76719325 834int __set_page_dirty_no_writeback(struct page *page);
1da177e4
LT
835int redirty_page_for_writepage(struct writeback_control *wbc,
836 struct page *page);
b3c97528 837int set_page_dirty(struct page *page);
1da177e4
LT
838int set_page_dirty_lock(struct page *page);
839int clear_page_dirty_for_io(struct page *page);
840
b6a2fea3
OW
841extern unsigned long move_page_tables(struct vm_area_struct *vma,
842 unsigned long old_addr, struct vm_area_struct *new_vma,
843 unsigned long new_addr, unsigned long len);
1da177e4
LT
844extern unsigned long do_mremap(unsigned long addr,
845 unsigned long old_len, unsigned long new_len,
846 unsigned long flags, unsigned long new_addr);
b6a2fea3
OW
847extern int mprotect_fixup(struct vm_area_struct *vma,
848 struct vm_area_struct **pprev, unsigned long start,
849 unsigned long end, unsigned long newflags);
1da177e4 850
21cc199b
NP
851/*
852 * get_user_pages_fast provides equivalent functionality to get_user_pages,
853 * operating on current and current->mm (force=0 and doesn't return any vmas).
854 *
855 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
856 * can be made about locking. get_user_pages_fast is to be implemented in a
857 * way that is advantageous (vs get_user_pages()) when the user memory area is
858 * already faulted in and present in ptes. However if the pages have to be
859 * faulted in, it may turn out to be slightly slower).
860 */
861int get_user_pages_fast(unsigned long start, int nr_pages, int write,
862 struct page **pages);
863
1da177e4 864/*
8e1f936b 865 * A callback you can register to apply pressure to ageable caches.
1da177e4 866 *
8e1f936b
RR
867 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
868 * look through the least-recently-used 'nr_to_scan' entries and
869 * attempt to free them up. It should return the number of objects
870 * which remain in the cache. If it returns -1, it means it cannot do
871 * any scanning at this time (eg. there is a risk of deadlock).
1da177e4 872 *
8e1f936b
RR
873 * The 'gfpmask' refers to the allocation we are currently trying to
874 * fulfil.
875 *
876 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
877 * querying the cache size, so a fastpath for that case is appropriate.
1da177e4 878 */
8e1f936b
RR
879struct shrinker {
880 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
881 int seeks; /* seeks to recreate an obj */
1da177e4 882
8e1f936b
RR
883 /* These are for internal use */
884 struct list_head list;
885 long nr; /* objs pending delete */
886};
887#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
888extern void register_shrinker(struct shrinker *);
889extern void unregister_shrinker(struct shrinker *);
1da177e4 890
4e950f6f 891int vma_wants_writenotify(struct vm_area_struct *vma);
d08b3851 892
b3c97528 893extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
c9cfcddf 894
5f22df00
NP
895#ifdef __PAGETABLE_PUD_FOLDED
896static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
897 unsigned long address)
898{
899 return 0;
900}
901#else
1bb3630e 902int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
5f22df00
NP
903#endif
904
905#ifdef __PAGETABLE_PMD_FOLDED
906static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
907 unsigned long address)
908{
909 return 0;
910}
911#else
1bb3630e 912int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
5f22df00
NP
913#endif
914
1bb3630e
HD
915int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
916int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
917
1da177e4
LT
918/*
919 * The following ifdef needed to get the 4level-fixup.h header to work.
920 * Remove it when 4level-fixup.h has been removed.
921 */
1bb3630e 922#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1da177e4
LT
923static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
924{
1bb3630e
HD
925 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
926 NULL: pud_offset(pgd, address);
1da177e4
LT
927}
928
929static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
930{
1bb3630e
HD
931 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
932 NULL: pmd_offset(pud, address);
1da177e4 933}
1bb3630e
HD
934#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
935
f7d0b926 936#if USE_SPLIT_PTLOCKS
4c21e2f2
HD
937/*
938 * We tuck a spinlock to guard each pagetable page into its struct page,
939 * at page->private, with BUILD_BUG_ON to make sure that this will not
940 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
941 * When freeing, reset page->mapping so free_pages_check won't complain.
942 */
349aef0b 943#define __pte_lockptr(page) &((page)->ptl)
4c21e2f2
HD
944#define pte_lock_init(_page) do { \
945 spin_lock_init(__pte_lockptr(_page)); \
946} while (0)
947#define pte_lock_deinit(page) ((page)->mapping = NULL)
948#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
f7d0b926 949#else /* !USE_SPLIT_PTLOCKS */
4c21e2f2
HD
950/*
951 * We use mm->page_table_lock to guard all pagetable pages of the mm.
952 */
953#define pte_lock_init(page) do {} while (0)
954#define pte_lock_deinit(page) do {} while (0)
955#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
f7d0b926 956#endif /* USE_SPLIT_PTLOCKS */
4c21e2f2 957
2f569afd
MS
958static inline void pgtable_page_ctor(struct page *page)
959{
960 pte_lock_init(page);
961 inc_zone_page_state(page, NR_PAGETABLE);
962}
963
964static inline void pgtable_page_dtor(struct page *page)
965{
966 pte_lock_deinit(page);
967 dec_zone_page_state(page, NR_PAGETABLE);
968}
969
c74df32c
HD
970#define pte_offset_map_lock(mm, pmd, address, ptlp) \
971({ \
4c21e2f2 972 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
c74df32c
HD
973 pte_t *__pte = pte_offset_map(pmd, address); \
974 *(ptlp) = __ptl; \
975 spin_lock(__ptl); \
976 __pte; \
977})
978
979#define pte_unmap_unlock(pte, ptl) do { \
980 spin_unlock(ptl); \
981 pte_unmap(pte); \
982} while (0)
983
1bb3630e
HD
984#define pte_alloc_map(mm, pmd, address) \
985 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
986 NULL: pte_offset_map(pmd, address))
987
c74df32c
HD
988#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
989 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
990 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
991
1bb3630e
HD
992#define pte_alloc_kernel(pmd, address) \
993 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
994 NULL: pte_offset_kernel(pmd, address))
1da177e4
LT
995
996extern void free_area_init(unsigned long * zones_size);
9109fb7b
JW
997extern void free_area_init_node(int nid, unsigned long * zones_size,
998 unsigned long zone_start_pfn, unsigned long *zholes_size);
c713216d
MG
999#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1000/*
1001 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1002 * zones, allocate the backing mem_map and account for memory holes in a more
1003 * architecture independent manner. This is a substitute for creating the
1004 * zone_sizes[] and zholes_size[] arrays and passing them to
1005 * free_area_init_node()
1006 *
1007 * An architecture is expected to register range of page frames backed by
1008 * physical memory with add_active_range() before calling
1009 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1010 * usage, an architecture is expected to do something like
1011 *
1012 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1013 * max_highmem_pfn};
1014 * for_each_valid_physical_page_range()
1015 * add_active_range(node_id, start_pfn, end_pfn)
1016 * free_area_init_nodes(max_zone_pfns);
1017 *
1018 * If the architecture guarantees that there are no holes in the ranges
1019 * registered with add_active_range(), free_bootmem_active_regions()
1020 * will call free_bootmem_node() for each registered physical page range.
1021 * Similarly sparse_memory_present_with_active_regions() calls
1022 * memory_present() for each range when SPARSEMEM is enabled.
1023 *
1024 * See mm/page_alloc.c for more information on each function exposed by
1025 * CONFIG_ARCH_POPULATES_NODE_MAP
1026 */
1027extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1028extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1029 unsigned long end_pfn);
cc1050ba
YL
1030extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1031 unsigned long end_pfn);
fb01439c
MG
1032extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1033 unsigned long end_pfn);
c713216d
MG
1034extern void remove_all_active_ranges(void);
1035extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1036 unsigned long end_pfn);
1037extern void get_pfn_range_for_nid(unsigned int nid,
1038 unsigned long *start_pfn, unsigned long *end_pfn);
1039extern unsigned long find_min_pfn_with_active_regions(void);
c713216d
MG
1040extern void free_bootmem_with_active_regions(int nid,
1041 unsigned long max_low_pfn);
d52d53b8 1042typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
b5bc6c0e 1043extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
c713216d 1044extern void sparse_memory_present_with_active_regions(int nid);
c713216d 1045#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
f2dbcfa7
KH
1046
1047#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1048 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1049static inline int __early_pfn_to_nid(unsigned long pfn)
1050{
1051 return 0;
1052}
1053#else
1054/* please see mm/page_alloc.c */
1055extern int __meminit early_pfn_to_nid(unsigned long pfn);
1056#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1057/* there is a per-arch backend function. */
1058extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1059#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1060#endif
1061
0e0b864e 1062extern void set_dma_reserve(unsigned long new_dma_reserve);
a2f3aa02
DH
1063extern void memmap_init_zone(unsigned long, int, unsigned long,
1064 unsigned long, enum memmap_context);
3947be19 1065extern void setup_per_zone_pages_min(void);
1da177e4 1066extern void mem_init(void);
8feae131 1067extern void __init mmap_init(void);
1da177e4
LT
1068extern void show_mem(void);
1069extern void si_meminfo(struct sysinfo * val);
1070extern void si_meminfo_node(struct sysinfo *val, int nid);
3461b0af 1071extern int after_bootmem;
1da177e4 1072
e7c8d5c9
CL
1073#ifdef CONFIG_NUMA
1074extern void setup_per_cpu_pageset(void);
1075#else
1076static inline void setup_per_cpu_pageset(void) {}
1077#endif
1078
8feae131
DH
1079/* nommu.c */
1080extern atomic_t mmap_pages_allocated;
1081
1da177e4
LT
1082/* prio_tree.c */
1083void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1084void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1085void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1086struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1087 struct prio_tree_iter *iter);
1088
1089#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1090 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1091 (vma = vma_prio_tree_next(vma, iter)); )
1092
1093static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1094 struct list_head *list)
1095{
1096 vma->shared.vm_set.parent = NULL;
1097 list_add_tail(&vma->shared.vm_set.list, list);
1098}
1099
1100/* mmap.c */
34b4e4aa 1101extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1da177e4
LT
1102extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1103 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1104extern struct vm_area_struct *vma_merge(struct mm_struct *,
1105 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1106 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1107 struct mempolicy *);
1108extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1109extern int split_vma(struct mm_struct *,
1110 struct vm_area_struct *, unsigned long addr, int new_below);
1111extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1112extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1113 struct rb_node **, struct rb_node *);
a8fb5618 1114extern void unlink_file_vma(struct vm_area_struct *);
1da177e4
LT
1115extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1116 unsigned long addr, unsigned long len, pgoff_t pgoff);
1117extern void exit_mmap(struct mm_struct *);
925d1c40 1118
7906d00c
AA
1119extern int mm_take_all_locks(struct mm_struct *mm);
1120extern void mm_drop_all_locks(struct mm_struct *mm);
1121
925d1c40
MH
1122#ifdef CONFIG_PROC_FS
1123/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1124extern void added_exe_file_vma(struct mm_struct *mm);
1125extern void removed_exe_file_vma(struct mm_struct *mm);
1126#else
1127static inline void added_exe_file_vma(struct mm_struct *mm)
1128{}
1129
1130static inline void removed_exe_file_vma(struct mm_struct *mm)
1131{}
1132#endif /* CONFIG_PROC_FS */
1133
119f657c 1134extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
fa5dc22f
RM
1135extern int install_special_mapping(struct mm_struct *mm,
1136 unsigned long addr, unsigned long len,
1137 unsigned long flags, struct page **pages);
1da177e4
LT
1138
1139extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1140
1141extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1142 unsigned long len, unsigned long prot,
1143 unsigned long flag, unsigned long pgoff);
0165ab44
MS
1144extern unsigned long mmap_region(struct file *file, unsigned long addr,
1145 unsigned long len, unsigned long flags,
5a6fe125 1146 unsigned int vm_flags, unsigned long pgoff);
1da177e4
LT
1147
1148static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1149 unsigned long len, unsigned long prot,
1150 unsigned long flag, unsigned long offset)
1151{
1152 unsigned long ret = -EINVAL;
1153 if ((offset + PAGE_ALIGN(len)) < offset)
1154 goto out;
1155 if (!(offset & ~PAGE_MASK))
1156 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1157out:
1158 return ret;
1159}
1160
1161extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1162
1163extern unsigned long do_brk(unsigned long, unsigned long);
1164
1165/* filemap.c */
1166extern unsigned long page_unuse(struct page *);
1167extern void truncate_inode_pages(struct address_space *, loff_t);
d7339071
HR
1168extern void truncate_inode_pages_range(struct address_space *,
1169 loff_t lstart, loff_t lend);
1da177e4
LT
1170
1171/* generic vm_area_ops exported for stackable file systems */
d0217ac0 1172extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1da177e4
LT
1173
1174/* mm/page-writeback.c */
1175int write_one_page(struct page *page, int wait);
1cf6e7d8 1176void task_dirty_inc(struct task_struct *tsk);
1da177e4
LT
1177
1178/* readahead.c */
1179#define VM_MAX_READAHEAD 128 /* kbytes */
1180#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1da177e4
LT
1181
1182int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
7361f4d8 1183 pgoff_t offset, unsigned long nr_to_read);
1da177e4 1184int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
7361f4d8 1185 pgoff_t offset, unsigned long nr_to_read);
cf914a7d
RR
1186
1187void page_cache_sync_readahead(struct address_space *mapping,
1188 struct file_ra_state *ra,
1189 struct file *filp,
1190 pgoff_t offset,
1191 unsigned long size);
1192
1193void page_cache_async_readahead(struct address_space *mapping,
1194 struct file_ra_state *ra,
1195 struct file *filp,
1196 struct page *pg,
1197 pgoff_t offset,
1198 unsigned long size);
1199
1da177e4
LT
1200unsigned long max_sane_readahead(unsigned long nr);
1201
1202/* Do stack extension */
46dea3d0 1203extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
9ab88515 1204#ifdef CONFIG_IA64
46dea3d0 1205extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
9ab88515 1206#endif
b6a2fea3
OW
1207extern int expand_stack_downwards(struct vm_area_struct *vma,
1208 unsigned long address);
1da177e4
LT
1209
1210/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1211extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1212extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1213 struct vm_area_struct **pprev);
1214
1215/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1216 NULL if none. Assume start_addr < end_addr. */
1217static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1218{
1219 struct vm_area_struct * vma = find_vma(mm,start_addr);
1220
1221 if (vma && end_addr <= vma->vm_start)
1222 vma = NULL;
1223 return vma;
1224}
1225
1226static inline unsigned long vma_pages(struct vm_area_struct *vma)
1227{
1228 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1229}
1230
804af2cf 1231pgprot_t vm_get_page_prot(unsigned long vm_flags);
deceb6cd 1232struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
deceb6cd
HD
1233int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1234 unsigned long pfn, unsigned long size, pgprot_t);
a145dd41 1235int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
e0dc0d8f
NP
1236int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1237 unsigned long pfn);
423bad60
NP
1238int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1239 unsigned long pfn);
deceb6cd 1240
6aab341e 1241struct page *follow_page(struct vm_area_struct *, unsigned long address,
deceb6cd
HD
1242 unsigned int foll_flags);
1243#define FOLL_WRITE 0x01 /* check pte is writable */
1244#define FOLL_TOUCH 0x02 /* mark page accessed */
1245#define FOLL_GET 0x04 /* do get_page on page */
1246#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1da177e4 1247
2f569afd 1248typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
aee16b3c
JF
1249 void *data);
1250extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1251 unsigned long size, pte_fn_t fn, void *data);
1252
1da177e4 1253#ifdef CONFIG_PROC_FS
ab50b8ed 1254void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1da177e4 1255#else
ab50b8ed 1256static inline void vm_stat_account(struct mm_struct *mm,
1da177e4
LT
1257 unsigned long flags, struct file *file, long pages)
1258{
1259}
1260#endif /* CONFIG_PROC_FS */
1261
12d6f21e
IM
1262#ifdef CONFIG_DEBUG_PAGEALLOC
1263extern int debug_pagealloc_enabled;
1264
1265extern void kernel_map_pages(struct page *page, int numpages, int enable);
1266
1267static inline void enable_debug_pagealloc(void)
1268{
1269 debug_pagealloc_enabled = 1;
1270}
8a235efa
RW
1271#ifdef CONFIG_HIBERNATION
1272extern bool kernel_page_present(struct page *page);
1273#endif /* CONFIG_HIBERNATION */
12d6f21e 1274#else
1da177e4 1275static inline void
9858db50 1276kernel_map_pages(struct page *page, int numpages, int enable) {}
12d6f21e
IM
1277static inline void enable_debug_pagealloc(void)
1278{
1279}
8a235efa
RW
1280#ifdef CONFIG_HIBERNATION
1281static inline bool kernel_page_present(struct page *page) { return true; }
1282#endif /* CONFIG_HIBERNATION */
1da177e4
LT
1283#endif
1284
1285extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1286#ifdef __HAVE_ARCH_GATE_AREA
1287int in_gate_area_no_task(unsigned long addr);
1288int in_gate_area(struct task_struct *task, unsigned long addr);
1289#else
1290int in_gate_area_no_task(unsigned long addr);
1291#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1292#endif /* __HAVE_ARCH_GATE_AREA */
1293
9d0243bc
AM
1294int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1295 void __user *, size_t *, loff_t *);
69e05944 1296unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
9d0243bc 1297 unsigned long lru_pages);
9d0243bc 1298
7a9166e3
LY
1299#ifndef CONFIG_MMU
1300#define randomize_va_space 0
1301#else
a62eaf15 1302extern int randomize_va_space;
7a9166e3 1303#endif
a62eaf15 1304
045e72ac 1305const char * arch_vma_name(struct vm_area_struct *vma);
03252919 1306void print_vma_addr(char *prefix, unsigned long rip);
e6e5494c 1307
98f3cfc1 1308struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
29c71111
AW
1309pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1310pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1311pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1312pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
8f6aac41
CL
1313void *vmemmap_alloc_block(unsigned long size, int node);
1314void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
29c71111
AW
1315int vmemmap_populate_basepages(struct page *start_page,
1316 unsigned long pages, int node);
1317int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
c2b91e2e 1318void vmemmap_populate_print_last(void);
8f6aac41 1319
c5dee617
MM
1320extern void *alloc_locked_buffer(size_t size);
1321extern void free_locked_buffer(void *buffer, size_t size);
9f339e70 1322extern void release_locked_buffer(void *buffer, size_t size);
1da177e4
LT
1323#endif /* __KERNEL__ */
1324#endif /* _LINUX_MM_H */