blob: 9670f71d7be961bbe56e9c1701f475bc4be7981c [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
Linus Torvalds1da177e2005-04-16 15:20:36 -07004#include <linux/errno.h>
5
6#ifdef __KERNEL__
7
Linus Torvalds1da177e2005-04-16 15:20:36 -07008#include <linux/gfp.h>
9#include <linux/list.h>
10#include <linux/mmzone.h>
11#include <linux/rbtree.h>
12#include <linux/prio_tree.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070013#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070014#include <linux/mm_types.h>
Yinghai Lu08677212010-02-10 01:20:20 -080015#include <linux/range.h>
Chris Metcalfc6f6b592010-05-24 14:32:53 -070016#include <linux/pfn.h>
Andrea Arcangelie9da73d2011-01-13 15:46:32 -080017#include <linux/bit_spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018
19struct mempolicy;
20struct anon_vma;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040021struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040022struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040023struct writeback_control;
Linus Torvalds1da177e2005-04-16 15:20:36 -070024
25#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
26extern unsigned long max_mapnr;
27#endif
28
29extern unsigned long num_physpages;
Jan Beulich44813742009-09-21 17:03:05 -070030extern unsigned long totalram_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -070031extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070032extern int page_cluster;
33
34#ifdef CONFIG_SYSCTL
35extern int sysctl_legacy_va_layout;
36#else
37#define sysctl_legacy_va_layout 0
38#endif
39
40#include <asm/page.h>
41#include <asm/pgtable.h>
42#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
45
Andrea Righi27ac7922008-07-23 21:28:13 -070046/* to align the pointer to the (next) page boundary */
47#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
48
Linus Torvalds1da177e2005-04-16 15:20:36 -070049/*
50 * Linux kernel virtual memory manager primitives.
51 * The idea being to have a "virtual" mm in the same way
52 * we have a virtual fs - giving a cleaner interface to the
53 * mm details, and allowing different kinds of memory mappings
54 * (from shared memory to executable loading to arbitrary
55 * mmap() functions).
56 */
57
Christoph Lameterc43692e2006-12-06 20:32:48 -080058extern struct kmem_cache *vm_area_cachep;
59
Linus Torvalds1da177e2005-04-16 15:20:36 -070060#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +000061extern struct rb_root nommu_region_tree;
62extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64extern unsigned int kobjsize(const void *objp);
65#endif
66
67/*
Hugh Dickins605d9282008-08-16 11:07:21 +010068 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 */
70#define VM_READ 0x00000001 /* currently active flags */
71#define VM_WRITE 0x00000002
72#define VM_EXEC 0x00000004
73#define VM_SHARED 0x00000008
74
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070075/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
77#define VM_MAYWRITE 0x00000020
78#define VM_MAYEXEC 0x00000040
79#define VM_MAYSHARE 0x00000080
80
81#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
Luck, Tony8ca3eb02010-08-24 11:44:18 -070082#if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#define VM_GROWSUP 0x00000200
Luck, Tony8ca3eb02010-08-24 11:44:18 -070084#else
85#define VM_GROWSUP 0x00000000
Andrea Arcangelia664b2d2011-01-13 15:47:17 -080086#define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
Luck, Tony8ca3eb02010-08-24 11:44:18 -070087#endif
Linus Torvalds6aab3412005-11-28 14:34:23 -080088#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070089#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
90
91#define VM_EXECUTABLE 0x00001000
92#define VM_LOCKED 0x00002000
93#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
94
95 /* Used by sys_madvise() */
96#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
98
99#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800101#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -0700103#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
105#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
Johannes Weinerf2d6bfe2011-01-13 15:47:01 -0800106#ifndef CONFIG_TRANSPARENT_HUGEPAGE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Johannes Weinerf2d6bfe2011-01-13 15:47:01 -0800108#else
109#define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
110#endif
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700111#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800112#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Nick Piggind0217ac2007-07-19 01:47:03 -0700114#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700115#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000116#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700117#define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700118#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700119
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700120/* Bits set in the VMA until the stack is in its final location */
121#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
122
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
124#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
125#endif
126
127#ifdef CONFIG_STACK_GROWSUP
128#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
129#else
130#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
131#endif
132
133#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
134#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
135#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
136#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
137#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
138
139/*
Andrea Arcangeli78f11a22011-04-27 15:26:45 -0700140 * Special vmas that are non-mergable, non-mlock()able.
141 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
Nick Pigginb291f002008-10-18 20:26:44 -0700142 */
143#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
144
145/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 * mapping from the currently active vm_flags protection bits (the
147 * low four bits) to a page protection mask..
148 */
149extern pgprot_t protection_map[16];
150
Nick Piggind0217ac2007-07-19 01:47:03 -0700151#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
152#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700153#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Michel Lespinassed065bd82010-10-26 14:21:57 -0700154#define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
Gleb Natapov318b2752011-03-22 16:30:51 -0700155#define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
KOSAKI Motohiro37b23e02011-05-24 17:11:30 -0700156#define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
Nick Piggind0217ac2007-07-19 01:47:03 -0700157
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800158/*
159 * This interface is used by x86 PAT code to identify a pfn mapping that is
160 * linear over entire vma. This is to optimize PAT code that deals with
161 * marking the physical region with a particular prot. This is not for generic
162 * mm use. Note also that this check will not work if the pfn mapping is
163 * linear for a vma starting at physical address 0. In which case PAT code
164 * falls back to slow path of reserving physical range page by page.
165 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800166static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
167{
KOSAKI Motohiroca16d142011-05-26 19:16:19 +0900168 return !!(vma->vm_flags & VM_PFN_AT_MMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800169}
170
171static inline int is_pfn_mapping(struct vm_area_struct *vma)
172{
KOSAKI Motohiroca16d142011-05-26 19:16:19 +0900173 return !!(vma->vm_flags & VM_PFNMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800174}
Nick Piggind0217ac2007-07-19 01:47:03 -0700175
Nick Piggin54cb8822007-07-19 01:46:59 -0700176/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700177 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700178 * ->fault function. The vma's ->fault is responsible for returning a bitmask
179 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700180 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700181 * pgoff should be used in favour of virtual_address, if possible. If pgoff
182 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
183 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700184 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700185struct vm_fault {
186 unsigned int flags; /* FAULT_FLAG_xxx flags */
187 pgoff_t pgoff; /* Logical page offset based on vma */
188 void __user *virtual_address; /* Faulting virtual address */
189
190 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700191 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700192 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700193 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700194 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700195};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196
197/*
198 * These are the virtual MM functions - opening of an area, closing and
199 * unmapping it (needed to keep files on disk up-to-date etc), pointer
200 * to the functions called when a no-page or a wp-page exception occurs.
201 */
202struct vm_operations_struct {
203 void (*open)(struct vm_area_struct * area);
204 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700205 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700206
207 /* notification that a previously read-only page is about to become
208 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700209 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700210
211 /* called by access_process_vm when get_user_pages() fails, typically
212 * for use by special VMAs that can switch between memory and hardware
213 */
214 int (*access)(struct vm_area_struct *vma, unsigned long addr,
215 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700217 /*
218 * set_policy() op must add a reference to any non-NULL @new mempolicy
219 * to hold the policy upon return. Caller should pass NULL @new to
220 * remove a policy and fall back to surrounding context--i.e. do not
221 * install a MPOL_DEFAULT policy, nor the task or system default
222 * mempolicy.
223 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700225
226 /*
227 * get_policy() op must add reference [mpol_get()] to any policy at
228 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
229 * in mm/mempolicy.c will do this automatically.
230 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
231 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
232 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
233 * must return NULL--i.e., do not "fallback" to task or system default
234 * policy.
235 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
237 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700238 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
239 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240#endif
241};
242
243struct mmu_gather;
244struct inode;
245
Andrew Morton349aef02006-01-08 01:04:36 -0800246#define page_private(page) ((page)->private)
247#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249/*
250 * FIXME: take this include out, include page-flags.h in
251 * files which need it (119 of them)
252 */
253#include <linux/page-flags.h>
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800254#include <linux/huge_mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256/*
257 * Methods to modify the page usage count.
258 *
259 * What counts for a page usage:
260 * - cache mapping (page->mapping)
261 * - private data (page->private)
262 * - page mapped in a task's page tables, each mapping
263 * is counted separately
264 *
265 * Also, many kernel routines increase the page count before a critical
266 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 */
268
269/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700270 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800272static inline int put_page_testzero(struct page *page)
273{
Nick Piggin725d7042006-09-25 23:30:55 -0700274 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800275 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800276}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277
278/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800279 * Try to grab a ref unless the page has a refcount of zero, return false if
280 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800282static inline int get_page_unless_zero(struct page *page)
283{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800284 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800285}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800287extern int page_is_ram(unsigned long pfn);
288
Christoph Lameter48667e72008-02-04 22:28:31 -0800289/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800290struct page *vmalloc_to_page(const void *addr);
291unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800292
Paul Mundt0738c4b2008-03-12 16:51:31 +0900293/*
294 * Determine if an address is within the vmalloc range
295 *
296 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
297 * is no special casing required.
298 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800299static inline int is_vmalloc_addr(const void *x)
300{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900301#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800302 unsigned long addr = (unsigned long)x;
303
304 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900305#else
306 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800307#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900308}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700309#ifdef CONFIG_MMU
310extern int is_vmalloc_or_module_addr(const void *x);
311#else
David Howells934831d2009-09-24 12:33:32 +0100312static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700313{
314 return 0;
315}
316#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800317
Andrea Arcangelie9da73d2011-01-13 15:46:32 -0800318static inline void compound_lock(struct page *page)
319{
320#ifdef CONFIG_TRANSPARENT_HUGEPAGE
321 bit_spin_lock(PG_compound_lock, &page->flags);
322#endif
323}
324
325static inline void compound_unlock(struct page *page)
326{
327#ifdef CONFIG_TRANSPARENT_HUGEPAGE
328 bit_spin_unlock(PG_compound_lock, &page->flags);
329#endif
330}
331
332static inline unsigned long compound_lock_irqsave(struct page *page)
333{
334 unsigned long uninitialized_var(flags);
335#ifdef CONFIG_TRANSPARENT_HUGEPAGE
336 local_irq_save(flags);
337 compound_lock(page);
338#endif
339 return flags;
340}
341
342static inline void compound_unlock_irqrestore(struct page *page,
343 unsigned long flags)
344{
345#ifdef CONFIG_TRANSPARENT_HUGEPAGE
346 compound_unlock(page);
347 local_irq_restore(flags);
348#endif
349}
350
Christoph Lameterd85f3382007-05-06 14:49:39 -0700351static inline struct page *compound_head(struct page *page)
352{
Christoph Lameter6d777952007-05-06 14:49:40 -0700353 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700354 return page->first_page;
355 return page;
356}
357
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700358static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700360 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361}
362
363static inline void get_page(struct page *page)
364{
Andrea Arcangeli91807062011-01-13 15:46:32 -0800365 /*
366 * Getting a normal page or the head of a compound page
367 * requires to already have an elevated page->_count. Only if
368 * we're getting a tail page, the elevated page->_count is
369 * required only in the head page, so for tail pages the
370 * bugcheck only verifies that the page->_count isn't
371 * negative.
372 */
373 VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 atomic_inc(&page->_count);
Andrea Arcangeli91807062011-01-13 15:46:32 -0800375 /*
376 * Getting a tail page will elevate both the head and tail
377 * page->_count(s).
378 */
379 if (unlikely(PageTail(page))) {
380 /*
381 * This is safe only because
382 * __split_huge_page_refcount can't run under
383 * get_page().
384 */
385 VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
386 atomic_inc(&page->first_page->_count);
387 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388}
389
Christoph Lameterb49af682007-05-06 14:49:41 -0700390static inline struct page *virt_to_head_page(const void *x)
391{
392 struct page *page = virt_to_page(x);
393 return compound_head(page);
394}
395
Nick Piggin7835e982006-03-22 00:08:40 -0800396/*
397 * Setup the page count before being freed into the page allocator for
398 * the first time (boot or memory hotplug)
399 */
400static inline void init_page_count(struct page *page)
401{
402 atomic_set(&page->_count, 1);
403}
404
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800405/*
406 * PageBuddy() indicate that the page is free and in the buddy system
407 * (see mm/page_alloc.c).
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100408 *
409 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
410 * -2 so that an underflow of the page_mapcount() won't be mistaken
411 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
412 * efficiently by most CPU architectures.
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800413 */
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100414#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
415
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800416static inline int PageBuddy(struct page *page)
417{
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100418 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800419}
420
421static inline void __SetPageBuddy(struct page *page)
422{
423 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100424 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800425}
426
427static inline void __ClearPageBuddy(struct page *page)
428{
429 VM_BUG_ON(!PageBuddy(page));
430 atomic_set(&page->_mapcount, -1);
431}
432
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700434void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800436void split_page(struct page *page, unsigned int order);
Mel Gorman748446b2010-05-24 14:32:27 -0700437int split_free_page(struct page *page);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800438
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800440 * Compound pages have a destructor function. Provide a
441 * prototype for that function and accessor functions.
442 * These are _only_ valid on the head of a PG_compound page.
443 */
444typedef void compound_page_dtor(struct page *);
445
446static inline void set_compound_page_dtor(struct page *page,
447 compound_page_dtor *dtor)
448{
449 page[1].lru.next = (void *)dtor;
450}
451
452static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
453{
454 return (compound_page_dtor *)page[1].lru.next;
455}
456
Christoph Lameterd85f3382007-05-06 14:49:39 -0700457static inline int compound_order(struct page *page)
458{
Christoph Lameter6d777952007-05-06 14:49:40 -0700459 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700460 return 0;
461 return (unsigned long)page[1].lru.prev;
462}
463
Andrea Arcangeli37c2ac72011-01-13 15:47:16 -0800464static inline int compound_trans_order(struct page *page)
465{
466 int order;
467 unsigned long flags;
468
469 if (!PageHead(page))
470 return 0;
471
472 flags = compound_lock_irqsave(page);
473 order = compound_order(page);
474 compound_unlock_irqrestore(page, flags);
475 return order;
476}
477
Christoph Lameterd85f3382007-05-06 14:49:39 -0700478static inline void set_compound_order(struct page *page, unsigned long order)
479{
480 page[1].lru.prev = (void *)order;
481}
482
Michal Simek3dece372011-01-21 08:49:56 +0100483#ifdef CONFIG_MMU
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800484/*
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800485 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
486 * servicing faults for write access. In the normal case, do always want
487 * pte_mkwrite. But get_user_pages can cause write faults for mappings
488 * that do not have writing enabled, when used by access_process_vm.
489 */
490static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
491{
492 if (likely(vma->vm_flags & VM_WRITE))
493 pte = pte_mkwrite(pte);
494 return pte;
495}
Michal Simek3dece372011-01-21 08:49:56 +0100496#endif
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800497
498/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 * Multiple processes may "see" the same page. E.g. for untouched
500 * mappings of /dev/null, all processes see the same page full of
501 * zeroes, and text pages of executables and shared libraries have
502 * only one copy in memory, at most, normally.
503 *
504 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700505 * page_count() == 0 means the page is free. page->lru is then used for
506 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700507 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700509 * Pages are allocated by the slab allocator in order to provide memory
510 * to kmalloc and kmem_cache_alloc. In this case, the management of the
511 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
512 * unless a particular usage is carefully commented. (the responsibility of
513 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700515 * A page may be used by anyone else who does a __get_free_page().
516 * In this case, page_count still tracks the references, and should only
517 * be used through the normal accessor functions. The top bits of page->flags
518 * and page->virtual store page management information, but all other fields
519 * are unused and could be used privately, carefully. The management of this
520 * page is the responsibility of the one who allocated it, and those who have
521 * subsequently been given references to it.
522 *
523 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 * managed by the Linux memory manager: I/O, buffers, swapping etc.
525 * The following discussion applies only to them.
526 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700527 * A pagecache page contains an opaque `private' member, which belongs to the
528 * page's address_space. Usually, this is the address of a circular list of
529 * the page's disk buffers. PG_private must be set to tell the VM to call
530 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700532 * A page may belong to an inode's memory mapping. In this case, page->mapping
533 * is the pointer to the inode, and page->index is the file offset of the page,
534 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700536 * If pagecache pages are not associated with an inode, they are said to be
537 * anonymous pages. These may become associated with the swapcache, and in that
538 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700540 * In either case (swapcache or inode backed), the pagecache itself holds one
541 * reference to the page. Setting PG_private should also increment the
542 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700544 * The pagecache pages are stored in a per-mapping radix tree, which is
545 * rooted at mapping->page_tree, and indexed by offset.
546 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
547 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700549 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 * - inode pages may need to be read from disk,
551 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700552 * to be written back to the inode on disk,
553 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
554 * modified may need to be swapped out to swap space and (later) to be read
555 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 */
557
558/*
559 * The zone field is never updated after free_area_init_core()
560 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700562
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700563
564/*
565 * page->flags layout:
566 *
567 * There are three possibilities for how page->flags get
568 * laid out. The first is for the normal case, without
569 * sparsemem. The second is for sparsemem when there is
570 * plenty of space for node and section. The last is when
571 * we have run out of space and have to fall back to an
572 * alternate (slower) way of determining the node.
573 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700574 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
575 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
576 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700577 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700578#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700579#define SECTIONS_WIDTH SECTIONS_SHIFT
580#else
581#define SECTIONS_WIDTH 0
582#endif
583
584#define ZONES_WIDTH ZONES_SHIFT
585
Christoph Lameter9223b412008-04-28 02:12:48 -0700586#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700587#define NODES_WIDTH NODES_SHIFT
588#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700589#ifdef CONFIG_SPARSEMEM_VMEMMAP
590#error "Vmemmap: No space for nodes field in page flags"
591#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700592#define NODES_WIDTH 0
593#endif
594
595/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100596#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700597#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
598#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
599
600/*
601 * We are going to use the flags for the page to node mapping if its in
602 * there. This includes the case where there is no node, so it is implicit.
603 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800604#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
605#define NODE_NOT_IN_PAGE_FLAGS
606#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700607
Dave Hansen348f8b62005-06-23 00:07:40 -0700608/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300609 * Define the bit shifts to access each section. For non-existent
Dave Hansen348f8b62005-06-23 00:07:40 -0700610 * sections we define the shift as 0; that plus a 0 mask ensures
611 * the compiler will optimise away reference to them.
612 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700613#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
614#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
615#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700616
Will Deaconbce54bb2010-10-26 14:21:37 -0700617/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
618#ifdef NODE_NOT_IN_PAGE_FLAGS
Christoph Lameter89689ae2006-12-06 20:31:45 -0800619#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800620#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
621 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700622#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800623#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800624#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
625 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700626#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800627
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800628#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700629
Christoph Lameter9223b412008-04-28 02:12:48 -0700630#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
631#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700632#endif
633
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700634#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
635#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
636#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800637#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700638
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700639static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640{
Dave Hansen348f8b62005-06-23 00:07:40 -0700641 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643
Christoph Lameter89689ae2006-12-06 20:31:45 -0800644/*
645 * The identification function is only used by the buddy allocator for
646 * determining if two pages could be buddies. We are not really
647 * identifying a zone since we could be using a the section number
648 * id if we have not node id available in page flags.
649 * We guarantee only that it will return the same value for two
650 * combinable pages in a zone.
651 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700652static inline int page_zone_id(struct page *page)
653{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800654 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655}
656
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800657static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700658{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700659#ifdef CONFIG_NUMA
660 return zone->node;
661#else
662 return 0;
663#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700664}
665
Christoph Lameter89689ae2006-12-06 20:31:45 -0800666#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800667extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800668#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800669static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700670{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800671 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700672}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800673#endif
674
675static inline struct zone *page_zone(struct page *page)
676{
677 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
678}
679
Christoph Lameter308c05e2008-04-28 02:12:43 -0700680#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700681static inline void set_page_section(struct page *page, unsigned long section)
682{
683 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
684 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
685}
686
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700687static inline unsigned long page_to_section(struct page *page)
688{
689 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
690}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700691#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700692
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700693static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694{
Dave Hansen348f8b62005-06-23 00:07:40 -0700695 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
696 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
697}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700698
Dave Hansen348f8b62005-06-23 00:07:40 -0700699static inline void set_page_node(struct page *page, unsigned long node)
700{
701 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
702 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
703}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800704
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700705static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700706 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700707{
708 set_page_zone(page, zone);
709 set_page_node(page, node);
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700710#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700711 set_page_section(page, pfn_to_section_nr(pfn));
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700712#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713}
714
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700715/*
716 * Some inline functions in vmstat.h depend on page_zone()
717 */
718#include <linux/vmstat.h>
719
Ingo Molnar652050a2006-01-14 13:21:30 -0800720static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721{
Chris Metcalfc6f6b592010-05-24 14:32:53 -0700722 return __va(PFN_PHYS(page_to_pfn(page)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723}
724
725#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
726#define HASHED_PAGE_VIRTUAL
727#endif
728
729#if defined(WANT_PAGE_VIRTUAL)
730#define page_address(page) ((page)->virtual)
731#define set_page_address(page, address) \
732 do { \
733 (page)->virtual = (address); \
734 } while(0)
735#define page_address_init() do { } while(0)
736#endif
737
738#if defined(HASHED_PAGE_VIRTUAL)
739void *page_address(struct page *page);
740void set_page_address(struct page *page, void *virtual);
741void page_address_init(void);
742#endif
743
744#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
745#define page_address(page) lowmem_page_address(page)
746#define set_page_address(page, address) do { } while(0)
747#define page_address_init() do { } while(0)
748#endif
749
750/*
751 * On an anonymous page mapped into a user virtual memory area,
752 * page->mapping points to its anon_vma, not to a struct address_space;
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800753 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
754 *
755 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
756 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
757 * and then page->mapping points, not to an anon_vma, but to a private
758 * structure which KSM associates with that merged page. See ksm.h.
759 *
760 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 *
762 * Please note that, confusingly, "page_mapping" refers to the inode
763 * address_space which maps the page from disk; whereas "page_mapped"
764 * refers to user virtual address space into which the page is mapped.
765 */
766#define PAGE_MAPPING_ANON 1
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800767#define PAGE_MAPPING_KSM 2
768#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
770extern struct address_space swapper_space;
771static inline struct address_space *page_mapping(struct page *page)
772{
773 struct address_space *mapping = page->mapping;
774
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700775 VM_BUG_ON(PageSlab(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 if (unlikely(PageSwapCache(page)))
777 mapping = &swapper_space;
Steven Rostedte20e8772011-01-13 15:46:17 -0800778 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 mapping = NULL;
780 return mapping;
781}
782
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800783/* Neutral page->mapping pointer to address_space or anon_vma or other */
784static inline void *page_rmapping(struct page *page)
785{
786 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
787}
788
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789static inline int PageAnon(struct page *page)
790{
791 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
792}
793
794/*
795 * Return the pagecache index of the passed page. Regular pagecache pages
796 * use ->index whereas swapcache pages use ->private
797 */
798static inline pgoff_t page_index(struct page *page)
799{
800 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700801 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 return page->index;
803}
804
805/*
806 * The atomic page->_mapcount, like _count, starts from -1:
807 * so that transitions both from it and to it can be tracked,
808 * using atomic_inc_and_test and atomic_add_negative(-1).
809 */
810static inline void reset_page_mapcount(struct page *page)
811{
812 atomic_set(&(page)->_mapcount, -1);
813}
814
815static inline int page_mapcount(struct page *page)
816{
817 return atomic_read(&(page)->_mapcount) + 1;
818}
819
820/*
821 * Return true if this page is mapped into pagetables.
822 */
823static inline int page_mapped(struct page *page)
824{
825 return atomic_read(&(page)->_mapcount) >= 0;
826}
827
828/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 * Different kinds of faults, as returned by handle_mm_fault().
830 * Used to decide whether a process gets delivered SIGBUS or
831 * just gets major/minor fault counters bumped up.
832 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700833
Nick Piggin83c54072007-07-19 01:47:05 -0700834#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700835
Nick Piggin83c54072007-07-19 01:47:05 -0700836#define VM_FAULT_OOM 0x0001
837#define VM_FAULT_SIGBUS 0x0002
838#define VM_FAULT_MAJOR 0x0004
839#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleenaa50d3a2010-10-06 21:45:00 +0200840#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
841#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000842
Nick Piggin83c54072007-07-19 01:47:05 -0700843#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
844#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Michel Lespinassed065bd82010-10-26 14:21:57 -0700845#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
Andi Kleenaa50d3a2010-10-06 21:45:00 +0200847#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
848
849#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
850 VM_FAULT_HWPOISON_LARGE)
851
852/* Encode hstate index for a hwpoisoned large page */
853#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
854#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
Nick Piggind0217ac2007-07-19 01:47:03 -0700855
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800856/*
857 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
858 */
859extern void pagefault_out_of_memory(void);
860
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
862
David Rientjesddd588b2011-03-22 16:30:46 -0700863/*
David Rientjes7bf02ea2011-05-24 17:11:16 -0700864 * Flags passed to show_mem() and show_free_areas() to suppress output in
David Rientjesddd588b2011-03-22 16:30:46 -0700865 * various contexts.
866 */
867#define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
868
David Rientjes7bf02ea2011-05-24 17:11:16 -0700869extern void show_free_areas(unsigned int flags);
870extern bool skip_free_areas_node(unsigned int flags, int nid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871
Hugh Dickins3f96b792009-09-21 17:03:37 -0700872int shmem_lock(struct file *file, int lock, struct user_struct *user);
Sergei Trofimovich168f5ac2009-06-16 15:33:02 -0700873struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874int shmem_zero_setup(struct vm_area_struct *);
875
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400876extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877extern int user_shm_lock(size_t, struct user_struct *);
878extern void user_shm_unlock(size_t, struct user_struct *);
879
880/*
881 * Parameter block passed down to zap_pte_range in exceptional cases.
882 */
883struct zap_details {
884 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
885 struct address_space *check_mapping; /* Check page->mapping if set */
886 pgoff_t first_index; /* Lowest page->index to unmap */
887 pgoff_t last_index; /* Highest page->index to unmap */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888};
889
Nick Piggin7e675132008-04-28 02:13:00 -0700890struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
891 pte_t pte);
892
Jack Steinerc627f9c2008-07-29 22:33:53 -0700893int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
894 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700895unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 unsigned long size, struct zap_details *);
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700897unsigned long unmap_vmas(struct mmu_gather *tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 struct vm_area_struct *start_vma, unsigned long start_addr,
899 unsigned long end_addr, unsigned long *nr_accounted,
900 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800901
902/**
903 * mm_walk - callbacks for walk_page_range
904 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
905 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
906 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
Dave Hansen03319322011-03-22 16:32:56 -0700907 * this handler is required to be able to handle
908 * pmd_trans_huge() pmds. They may simply choose to
909 * split_huge_page() instead of handling it explicitly.
Matt Mackalle6473092008-02-04 22:29:01 -0800910 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
911 * @pte_hole: if set, called for each hole at all levels
Naoya Horiguchi5dc37642009-12-14 18:00:01 -0800912 * @hugetlb_entry: if set, called for each hugetlb entry
Matt Mackalle6473092008-02-04 22:29:01 -0800913 *
914 * (see walk_page_range for more details)
915 */
916struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700917 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
918 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
919 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
920 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
921 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
Naoya Horiguchi116354d2010-04-06 14:35:04 -0700922 int (*hugetlb_entry)(pte_t *, unsigned long,
923 unsigned long, unsigned long, struct mm_walk *);
Dave Hansen21650092008-06-12 15:21:47 -0700924 struct mm_struct *mm;
925 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800926};
927
Dave Hansen21650092008-06-12 15:21:47 -0700928int walk_page_range(unsigned long addr, unsigned long end,
929 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700930void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700931 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
933 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934void unmap_mapping_range(struct address_space *mapping,
935 loff_t const holebegin, loff_t const holelen, int even_cows);
Johannes Weiner3b6748e2009-06-16 15:32:35 -0700936int follow_pfn(struct vm_area_struct *vma, unsigned long address,
937 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800938int follow_phys(struct vm_area_struct *vma, unsigned long address,
939 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700940int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
941 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942
943static inline void unmap_shared_mapping_range(struct address_space *mapping,
944 loff_t const holebegin, loff_t const holelen)
945{
946 unmap_mapping_range(mapping, holebegin, holelen, 0);
947}
948
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000949extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
Christoph Hellwig2c27c652010-06-04 11:30:04 +0200950extern void truncate_setsize(struct inode *inode, loff_t newsize);
npiggin@suse.de25d9e2d2009-08-21 02:35:05 +1000951extern int vmtruncate(struct inode *inode, loff_t offset);
952extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000953
Nick Piggin750b4982009-09-16 11:50:12 +0200954int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +0200955int generic_error_remove_page(struct address_space *mapping, struct page *page);
Nick Piggin750b4982009-09-16 11:50:12 +0200956
Wu Fengguang83f78662009-09-16 11:50:13 +0200957int invalidate_inode_page(struct page *page);
958
David Howells7ee1dd32006-01-06 00:11:44 -0800959#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700960extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700961 unsigned long address, unsigned int flags);
David Howells7ee1dd32006-01-06 00:11:44 -0800962#else
963static inline int handle_mm_fault(struct mm_struct *mm,
964 struct vm_area_struct *vma, unsigned long address,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700965 unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -0800966{
967 /* should never happen if there's no MMU */
968 BUG();
969 return VM_FAULT_SIGBUS;
970}
971#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000972
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973extern int make_pages_present(unsigned long addr, unsigned long end);
974extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Stephen Wilson5ddd36b2011-03-13 15:49:20 -0400975extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
976 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
Huang Ying0014bd92011-01-30 11:15:47 +0800978int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
979 unsigned long start, int len, unsigned int foll_flags,
980 struct page **pages, struct vm_area_struct **vmas,
981 int *nonblocking);
Nick Piggind2bf6be2009-06-16 15:31:39 -0700982int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
Peter Zijlstra9d737772009-06-25 11:58:55 +0200983 unsigned long start, int nr_pages, int write, int force,
Nick Piggind2bf6be2009-06-16 15:31:39 -0700984 struct page **pages, struct vm_area_struct **vmas);
985int get_user_pages_fast(unsigned long start, int nr_pages, int write,
986 struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -0700987struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988
David Howellscf9a2ae2006-08-29 19:05:54 +0100989extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
990extern void do_invalidatepage(struct page *page, unsigned long offset);
991
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800993int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994int redirty_page_for_writepage(struct writeback_control *wbc,
995 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -0700996void account_page_dirtied(struct page *page, struct address_space *mapping);
Michael Rubinf629d1c2010-10-26 14:21:33 -0700997void account_page_writeback(struct page *page);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800998int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999int set_page_dirty_lock(struct page *page);
1000int clear_page_dirty_for_io(struct page *page);
1001
Stefan Bader39aa3cb2010-08-31 15:52:27 +02001002/* Is the vma a continuation of the stack vma above it? */
Mikulas Patockaa09a79f2011-05-09 13:01:09 +02001003static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
Stefan Bader39aa3cb2010-08-31 15:52:27 +02001004{
1005 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1006}
1007
Mikulas Patockaa09a79f2011-05-09 13:01:09 +02001008static inline int stack_guard_page_start(struct vm_area_struct *vma,
1009 unsigned long addr)
1010{
1011 return (vma->vm_flags & VM_GROWSDOWN) &&
1012 (vma->vm_start == addr) &&
1013 !vma_growsdown(vma->vm_prev, addr);
1014}
1015
1016/* Is the vma a continuation of the stack vma below it? */
1017static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1018{
1019 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1020}
1021
1022static inline int stack_guard_page_end(struct vm_area_struct *vma,
1023 unsigned long addr)
1024{
1025 return (vma->vm_flags & VM_GROWSUP) &&
1026 (vma->vm_end == addr) &&
1027 !vma_growsup(vma->vm_next, addr);
1028}
1029
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001030extern unsigned long move_page_tables(struct vm_area_struct *vma,
1031 unsigned long old_addr, struct vm_area_struct *new_vma,
1032 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033extern unsigned long do_mremap(unsigned long addr,
1034 unsigned long old_len, unsigned long new_len,
1035 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001036extern int mprotect_fixup(struct vm_area_struct *vma,
1037 struct vm_area_struct **pprev, unsigned long start,
1038 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
Nick Piggin21cc1992008-07-25 19:45:22 -07001040/*
Peter Zijlstra465a4542009-06-15 12:31:37 +02001041 * doesn't attempt to fault and will return short.
1042 */
1043int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1044 struct page **pages);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001045/*
1046 * per-process(per-mm_struct) statistics.
1047 */
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001048static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
1049{
1050 atomic_long_set(&mm->rss_stat.count[member], value);
1051}
1052
Matt Fleming172703b2011-05-24 17:12:36 -07001053#if defined(SPLIT_RSS_COUNTING)
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -08001054unsigned long get_mm_counter(struct mm_struct *mm, int member);
Matt Fleming172703b2011-05-24 17:12:36 -07001055#else
1056static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1057{
1058 return atomic_long_read(&mm->rss_stat.count[member]);
1059}
1060#endif
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001061
1062static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1063{
1064 atomic_long_add(value, &mm->rss_stat.count[member]);
1065}
1066
1067static inline void inc_mm_counter(struct mm_struct *mm, int member)
1068{
1069 atomic_long_inc(&mm->rss_stat.count[member]);
1070}
1071
1072static inline void dec_mm_counter(struct mm_struct *mm, int member)
1073{
1074 atomic_long_dec(&mm->rss_stat.count[member]);
1075}
1076
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001077static inline unsigned long get_mm_rss(struct mm_struct *mm)
1078{
1079 return get_mm_counter(mm, MM_FILEPAGES) +
1080 get_mm_counter(mm, MM_ANONPAGES);
1081}
1082
1083static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1084{
1085 return max(mm->hiwater_rss, get_mm_rss(mm));
1086}
1087
1088static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1089{
1090 return max(mm->hiwater_vm, mm->total_vm);
1091}
1092
1093static inline void update_hiwater_rss(struct mm_struct *mm)
1094{
1095 unsigned long _rss = get_mm_rss(mm);
1096
1097 if ((mm)->hiwater_rss < _rss)
1098 (mm)->hiwater_rss = _rss;
1099}
1100
1101static inline void update_hiwater_vm(struct mm_struct *mm)
1102{
1103 if (mm->hiwater_vm < mm->total_vm)
1104 mm->hiwater_vm = mm->total_vm;
1105}
1106
1107static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1108 struct mm_struct *mm)
1109{
1110 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1111
1112 if (*maxrss < hiwater_rss)
1113 *maxrss = hiwater_rss;
1114}
1115
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001116#if defined(SPLIT_RSS_COUNTING)
KAMEZAWA Hiroyuki34e55232010-03-05 13:41:40 -08001117void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001118#else
1119static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
1120{
1121}
1122#endif
Peter Zijlstra465a4542009-06-15 12:31:37 +02001123
1124/*
Ying Hana09ed5e2011-05-24 17:12:26 -07001125 * This struct is used to pass information from page reclaim to the shrinkers.
1126 * We consolidate the values for easier extention later.
1127 */
1128struct shrink_control {
Ying Hana09ed5e2011-05-24 17:12:26 -07001129 gfp_t gfp_mask;
Ying Han1495f232011-05-24 17:12:27 -07001130
1131 /* How many slab objects shrinker() should scan and try to reclaim */
1132 unsigned long nr_to_scan;
Ying Hana09ed5e2011-05-24 17:12:26 -07001133};
1134
1135/*
Rusty Russell8e1f9362007-07-17 04:03:17 -07001136 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 *
Ying Han1495f232011-05-24 17:12:27 -07001138 * 'sc' is passed shrink_control which includes a count 'nr_to_scan'
1139 * and a 'gfpmask'. It should look through the least-recently-used
1140 * 'nr_to_scan' entries and attempt to free them up. It should return
1141 * the number of objects which remain in the cache. If it returns -1, it means
1142 * it cannot do any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 *
Rusty Russell8e1f9362007-07-17 04:03:17 -07001144 * The 'gfpmask' refers to the allocation we are currently trying to
1145 * fulfil.
1146 *
1147 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
1148 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 */
Rusty Russell8e1f9362007-07-17 04:03:17 -07001150struct shrinker {
Ying Han1495f232011-05-24 17:12:27 -07001151 int (*shrink)(struct shrinker *, struct shrink_control *sc);
Rusty Russell8e1f9362007-07-17 04:03:17 -07001152 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
Rusty Russell8e1f9362007-07-17 04:03:17 -07001154 /* These are for internal use */
1155 struct list_head list;
1156 long nr; /* objs pending delete */
1157};
1158#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1159extern void register_shrinker(struct shrinker *);
1160extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161
Alexey Dobriyan4e950f62007-07-30 02:36:13 +04001162int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -07001163
Namhyung Kim25ca1d62010-10-26 14:21:59 -07001164extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1165 spinlock_t **ptl);
1166static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1167 spinlock_t **ptl)
1168{
1169 pte_t *ptep;
1170 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1171 return ptep;
1172}
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -08001173
Nick Piggin5f22df02007-05-06 14:49:02 -07001174#ifdef __PAGETABLE_PUD_FOLDED
1175static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1176 unsigned long address)
1177{
1178 return 0;
1179}
1180#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001181int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001182#endif
1183
1184#ifdef __PAGETABLE_PMD_FOLDED
1185static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1186 unsigned long address)
1187{
1188 return 0;
1189}
1190#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001191int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001192#endif
1193
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001194int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1195 pmd_t *pmd, unsigned long address);
Hugh Dickins1bb36302005-10-29 18:16:22 -07001196int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1197
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198/*
1199 * The following ifdef needed to get the 4level-fixup.h header to work.
1200 * Remove it when 4level-fixup.h has been removed.
1201 */
Hugh Dickins1bb36302005-10-29 18:16:22 -07001202#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1204{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001205 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1206 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207}
1208
1209static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1210{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001211 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1212 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213}
Hugh Dickins1bb36302005-10-29 18:16:22 -07001214#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1215
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001216#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001217/*
1218 * We tuck a spinlock to guard each pagetable page into its struct page,
1219 * at page->private, with BUILD_BUG_ON to make sure that this will not
1220 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1221 * When freeing, reset page->mapping so free_pages_check won't complain.
1222 */
Andrew Morton349aef02006-01-08 01:04:36 -08001223#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001224#define pte_lock_init(_page) do { \
1225 spin_lock_init(__pte_lockptr(_page)); \
1226} while (0)
1227#define pte_lock_deinit(page) ((page)->mapping = NULL)
1228#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001229#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001230/*
1231 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1232 */
1233#define pte_lock_init(page) do {} while (0)
1234#define pte_lock_deinit(page) do {} while (0)
1235#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -07001236#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001237
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001238static inline void pgtable_page_ctor(struct page *page)
1239{
1240 pte_lock_init(page);
1241 inc_zone_page_state(page, NR_PAGETABLE);
1242}
1243
1244static inline void pgtable_page_dtor(struct page *page)
1245{
1246 pte_lock_deinit(page);
1247 dec_zone_page_state(page, NR_PAGETABLE);
1248}
1249
Hugh Dickinsc74df322005-10-29 18:16:23 -07001250#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1251({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001252 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001253 pte_t *__pte = pte_offset_map(pmd, address); \
1254 *(ptlp) = __ptl; \
1255 spin_lock(__ptl); \
1256 __pte; \
1257})
1258
1259#define pte_unmap_unlock(pte, ptl) do { \
1260 spin_unlock(ptl); \
1261 pte_unmap(pte); \
1262} while (0)
1263
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001264#define pte_alloc_map(mm, vma, pmd, address) \
1265 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1266 pmd, address))? \
1267 NULL: pte_offset_map(pmd, address))
Hugh Dickins1bb36302005-10-29 18:16:22 -07001268
Hugh Dickinsc74df322005-10-29 18:16:23 -07001269#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001270 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1271 pmd, address))? \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001272 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1273
Hugh Dickins1bb36302005-10-29 18:16:22 -07001274#define pte_alloc_kernel(pmd, address) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001275 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
Hugh Dickins1bb36302005-10-29 18:16:22 -07001276 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277
1278extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001279extern void free_area_init_node(int nid, unsigned long * zones_size,
1280 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -07001281#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1282/*
1283 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1284 * zones, allocate the backing mem_map and account for memory holes in a more
1285 * architecture independent manner. This is a substitute for creating the
1286 * zone_sizes[] and zholes_size[] arrays and passing them to
1287 * free_area_init_node()
1288 *
1289 * An architecture is expected to register range of page frames backed by
1290 * physical memory with add_active_range() before calling
1291 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1292 * usage, an architecture is expected to do something like
1293 *
1294 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1295 * max_highmem_pfn};
1296 * for_each_valid_physical_page_range()
1297 * add_active_range(node_id, start_pfn, end_pfn)
1298 * free_area_init_nodes(max_zone_pfns);
1299 *
1300 * If the architecture guarantees that there are no holes in the ranges
1301 * registered with add_active_range(), free_bootmem_active_regions()
1302 * will call free_bootmem_node() for each registered physical page range.
1303 * Similarly sparse_memory_present_with_active_regions() calls
1304 * memory_present() for each range when SPARSEMEM is enabled.
1305 *
1306 * See mm/page_alloc.c for more information on each function exposed by
1307 * CONFIG_ARCH_POPULATES_NODE_MAP
1308 */
1309extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1310extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1311 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001312extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1313 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001314extern void remove_all_active_ranges(void);
Yinghai Lu32996252009-12-15 17:59:02 -08001315void sort_node_map(void);
1316unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1317 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001318extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1319 unsigned long end_pfn);
1320extern void get_pfn_range_for_nid(unsigned int nid,
1321 unsigned long *start_pfn, unsigned long *end_pfn);
1322extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001323extern void free_bootmem_with_active_regions(int nid,
1324 unsigned long max_low_pfn);
Yinghai Lu08677212010-02-10 01:20:20 -08001325int add_from_early_node_map(struct range *range, int az,
1326 int nr_range, int nid);
Yinghai Luedbe7d22010-08-25 13:39:16 -07001327u64 __init find_memory_core_early(int nid, u64 size, u64 align,
1328 u64 goal, u64 limit);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001329typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001330extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001331extern void sparse_memory_present_with_active_regions(int nid);
Mel Gormanc7132162006-09-27 01:49:43 -07001332#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001333
1334#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1335 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1336static inline int __early_pfn_to_nid(unsigned long pfn)
1337{
1338 return 0;
1339}
1340#else
1341/* please see mm/page_alloc.c */
1342extern int __meminit early_pfn_to_nid(unsigned long pfn);
1343#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1344/* there is a per-arch backend function. */
1345extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1346#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1347#endif
1348
Mel Gorman0e0b8642006-09-27 01:49:56 -07001349extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001350extern void memmap_init_zone(unsigned long, int, unsigned long,
1351 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07001352extern void setup_per_zone_wmarks(void);
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -07001353extern int __meminit init_per_zone_wmark_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001355extern void __init mmap_init(void);
David Rientjesb2b755b2011-03-24 15:18:15 -07001356extern void show_mem(unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357extern void si_meminfo(struct sysinfo * val);
1358extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001359extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
Dave Hansena238ab52011-05-24 17:12:16 -07001361extern void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1362
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001363extern void setup_per_cpu_pageset(void);
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001364
Shaohua Li112067f2009-09-21 17:01:16 -07001365extern void zone_pcp_update(struct zone *zone);
1366
David Howells8feae132009-01-08 12:04:47 +00001367/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001368extern atomic_long_t mmap_pages_allocated;
David Howells7e660872010-01-15 17:01:39 -08001369extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
David Howells8feae132009-01-08 12:04:47 +00001370
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371/* prio_tree.c */
1372void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1373void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1374void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1375struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1376 struct prio_tree_iter *iter);
1377
1378#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1379 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1380 (vma = vma_prio_tree_next(vma, iter)); )
1381
1382static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1383 struct list_head *list)
1384{
1385 vma->shared.vm_set.parent = NULL;
1386 list_add_tail(&vma->shared.vm_set.list, list);
1387}
1388
1389/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001390extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Rik van Riel5beb4932010-03-05 13:42:07 -08001391extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1393extern struct vm_area_struct *vma_merge(struct mm_struct *,
1394 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1395 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1396 struct mempolicy *);
1397extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1398extern int split_vma(struct mm_struct *,
1399 struct vm_area_struct *, unsigned long addr, int new_below);
1400extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1401extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1402 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001403extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1405 unsigned long addr, unsigned long len, pgoff_t pgoff);
1406extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001407
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001408extern int mm_take_all_locks(struct mm_struct *mm);
1409extern void mm_drop_all_locks(struct mm_struct *mm);
1410
Matt Helsley925d1c42008-04-29 01:01:36 -07001411/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1412extern void added_exe_file_vma(struct mm_struct *mm);
1413extern void removed_exe_file_vma(struct mm_struct *mm);
Jiri Slaby38646012011-05-26 16:25:46 -07001414extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1415extern struct file *get_mm_exe_file(struct mm_struct *mm);
Matt Helsley925d1c42008-04-29 01:01:36 -07001416
akpm@osdl.org119f6572005-05-01 08:58:35 -07001417extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001418extern int install_special_mapping(struct mm_struct *mm,
1419 unsigned long addr, unsigned long len,
1420 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
1422extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1423
1424extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1425 unsigned long len, unsigned long prot,
1426 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001427extern unsigned long mmap_region(struct file *file, unsigned long addr,
1428 unsigned long len, unsigned long flags,
KOSAKI Motohiroca16d142011-05-26 19:16:19 +09001429 vm_flags_t vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430
1431static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1432 unsigned long len, unsigned long prot,
1433 unsigned long flag, unsigned long offset)
1434{
1435 unsigned long ret = -EINVAL;
1436 if ((offset + PAGE_ALIGN(len)) < offset)
1437 goto out;
1438 if (!(offset & ~PAGE_MASK))
1439 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1440out:
1441 return ret;
1442}
1443
1444extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1445
1446extern unsigned long do_brk(unsigned long, unsigned long);
1447
1448/* filemap.c */
1449extern unsigned long page_unuse(struct page *);
1450extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001451extern void truncate_inode_pages_range(struct address_space *,
1452 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453
1454/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001455extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
1457/* mm/page-writeback.c */
1458int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001459void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460
1461/* readahead.c */
1462#define VM_MAX_READAHEAD 128 /* kbytes */
1463#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001466 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001467
1468void page_cache_sync_readahead(struct address_space *mapping,
1469 struct file_ra_state *ra,
1470 struct file *filp,
1471 pgoff_t offset,
1472 unsigned long size);
1473
1474void page_cache_async_readahead(struct address_space *mapping,
1475 struct file_ra_state *ra,
1476 struct file *filp,
1477 struct page *pg,
1478 pgoff_t offset,
1479 unsigned long size);
1480
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481unsigned long max_sane_readahead(unsigned long nr);
Wu Fengguangd30a1102009-06-16 15:31:30 -07001482unsigned long ra_submit(struct file_ra_state *ra,
1483 struct address_space *mapping,
1484 struct file *filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485
Michal Hockod05f3162011-05-24 17:11:44 -07001486/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001487extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Michal Hockod05f3162011-05-24 17:11:44 -07001488
1489/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1490extern int expand_downwards(struct vm_area_struct *vma,
1491 unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001492#if VM_GROWSUP
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001493extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001494#else
1495 #define expand_upwards(vma, address) do { } while (0)
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001496#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497
1498/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1499extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1500extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1501 struct vm_area_struct **pprev);
1502
1503/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1504 NULL if none. Assume start_addr < end_addr. */
1505static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1506{
1507 struct vm_area_struct * vma = find_vma(mm,start_addr);
1508
1509 if (vma && end_addr <= vma->vm_start)
1510 vma = NULL;
1511 return vma;
1512}
1513
1514static inline unsigned long vma_pages(struct vm_area_struct *vma)
1515{
1516 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1517}
1518
David Howellsbad849b2010-08-26 16:00:34 +01001519#ifdef CONFIG_MMU
Hugh Dickins804af2c2006-07-26 21:39:49 +01001520pgprot_t vm_get_page_prot(unsigned long vm_flags);
David Howellsbad849b2010-08-26 16:00:34 +01001521#else
1522static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1523{
1524 return __pgprot(0);
1525}
1526#endif
1527
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001528struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001529int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1530 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001531int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001532int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1533 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001534int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1535 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
Linus Torvalds6aab3412005-11-28 14:34:23 -08001537struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001538 unsigned int foll_flags);
1539#define FOLL_WRITE 0x01 /* check pte is writable */
1540#define FOLL_TOUCH 0x02 /* mark page accessed */
1541#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a602009-09-21 17:03:26 -07001542#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07001543#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
Gleb Natapov318b2752011-03-22 16:30:51 -07001544#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1545 * and return without waiting upon it */
Michel Lespinasse110d74a2011-01-13 15:46:11 -08001546#define FOLL_MLOCK 0x40 /* mark page as mlocked */
Andrea Arcangeli500d65d2011-01-13 15:46:55 -08001547#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
Huang Ying69ebb832011-01-30 11:15:48 +08001548#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001550typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001551 void *data);
1552extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1553 unsigned long size, pte_fn_t fn, void *data);
1554
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001556void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001558static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 unsigned long flags, struct file *file, long pages)
1560{
1561}
1562#endif /* CONFIG_PROC_FS */
1563
Ingo Molnar12d6f212008-01-30 13:33:58 +01001564#ifdef CONFIG_DEBUG_PAGEALLOC
1565extern int debug_pagealloc_enabled;
1566
1567extern void kernel_map_pages(struct page *page, int numpages, int enable);
1568
1569static inline void enable_debug_pagealloc(void)
1570{
1571 debug_pagealloc_enabled = 1;
1572}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001573#ifdef CONFIG_HIBERNATION
1574extern bool kernel_page_present(struct page *page);
1575#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001576#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001578kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001579static inline void enable_debug_pagealloc(void)
1580{
1581}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001582#ifdef CONFIG_HIBERNATION
1583static inline bool kernel_page_present(struct page *page) { return true; }
1584#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585#endif
1586
Stephen Wilson31db58b2011-03-13 15:49:15 -04001587extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588#ifdef __HAVE_ARCH_GATE_AREA
Stephen Wilsoncae5d392011-03-13 15:49:17 -04001589int in_gate_area_no_mm(unsigned long addr);
Stephen Wilson83b964b2011-03-13 15:49:16 -04001590int in_gate_area(struct mm_struct *mm, unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591#else
Stephen Wilsoncae5d392011-03-13 15:49:17 -04001592int in_gate_area_no_mm(unsigned long addr);
1593#define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594#endif /* __HAVE_ARCH_GATE_AREA */
1595
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07001596int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001597 void __user *, size_t *, loff_t *);
Ying Hana09ed5e2011-05-24 17:12:26 -07001598unsigned long shrink_slab(struct shrink_control *shrink,
Ying Han1495f232011-05-24 17:12:27 -07001599 unsigned long nr_pages_scanned,
1600 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001601
Luke Yang7a9166e2006-02-20 18:28:07 -08001602#ifndef CONFIG_MMU
1603#define randomize_va_space 0
1604#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001605extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001606#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001607
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001608const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001609void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001610
Yinghai Lu9bdac912010-02-10 01:20:22 -08001611void sparse_mem_maps_populate_node(struct page **map_map,
1612 unsigned long pnum_begin,
1613 unsigned long pnum_end,
1614 unsigned long map_count,
1615 int nodeid);
1616
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001617struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001618pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1619pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1620pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1621pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001622void *vmemmap_alloc_block(unsigned long size, int node);
Yinghai Lu9bdac912010-02-10 01:20:22 -08001623void *vmemmap_alloc_block_buf(unsigned long size, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001624void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001625int vmemmap_populate_basepages(struct page *start_page,
1626 unsigned long pages, int node);
1627int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001628void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001629
Andi Kleen6a460792009-09-16 11:50:15 +02001630
Andi Kleen82ba0112009-12-16 12:19:57 +01001631enum mf_flags {
1632 MF_COUNT_INCREASED = 1 << 0,
1633};
Andi Kleen6a460792009-09-16 11:50:15 +02001634extern void memory_failure(unsigned long pfn, int trapno);
Andi Kleen82ba0112009-12-16 12:19:57 +01001635extern int __memory_failure(unsigned long pfn, int trapno, int flags);
Wu Fengguang847ce402009-12-16 12:19:58 +01001636extern int unpoison_memory(unsigned long pfn);
Andi Kleen6a460792009-09-16 11:50:15 +02001637extern int sysctl_memory_failure_early_kill;
1638extern int sysctl_memory_failure_recovery;
Andi Kleenfacb6012009-12-16 12:20:00 +01001639extern void shake_page(struct page *p, int access);
Andi Kleen6a460792009-09-16 11:50:15 +02001640extern atomic_long_t mce_bad_pages;
Andi Kleenfacb6012009-12-16 12:20:00 +01001641extern int soft_offline_page(struct page *page, int flags);
Andi Kleen6a460792009-09-16 11:50:15 +02001642
Wu Fengguang718a3822010-03-10 15:20:43 -08001643extern void dump_page(struct page *page);
1644
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08001645#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1646extern void clear_huge_page(struct page *page,
1647 unsigned long addr,
1648 unsigned int pages_per_huge_page);
1649extern void copy_user_huge_page(struct page *dst, struct page *src,
1650 unsigned long addr, struct vm_area_struct *vma,
1651 unsigned int pages_per_huge_page);
1652#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654#endif /* __KERNEL__ */
1655#endif /* _LINUX_MM_H */