blob: 8981cc882ed2eb0d2dfdebfbc5355e8c1cc2eb16 [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
Sasha Levin309381fea2014-01-23 15:52:54 -08008#include <linux/mmdebug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/gfp.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050010#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/list.h>
12#include <linux/mmzone.h>
13#include <linux/rbtree.h>
Konstantin Khlebnikov83aeead2011-12-08 14:33:54 -080014#include <linux/atomic.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070015#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070016#include <linux/mm_types.h>
Yinghai Lu08677212010-02-10 01:20:20 -080017#include <linux/range.h>
Chris Metcalfc6f6b592010-05-24 14:32:53 -070018#include <linux/pfn.h>
Andrea Arcangelie9da73d2011-01-13 15:46:32 -080019#include <linux/bit_spinlock.h>
Dave Chinnerb0d40c92011-07-08 14:14:42 +100020#include <linux/shrinker.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
22struct mempolicy;
23struct anon_vma;
Michel Lespinassebf181b92012-10-08 16:31:39 -070024struct anon_vma_chain;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040025struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040026struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040027struct writeback_control;
Linus Torvalds1da177e2005-04-16 15:20:36 -070028
Jiang Liufccc9982013-07-03 15:04:23 -070029#ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
Linus Torvalds1da177e2005-04-16 15:20:36 -070030extern unsigned long max_mapnr;
Jiang Liufccc9982013-07-03 15:04:23 -070031
32static inline void set_max_mapnr(unsigned long limit)
33{
34 max_mapnr = limit;
35}
36#else
37static inline void set_max_mapnr(unsigned long limit) { }
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#endif
39
Jan Beulich44813742009-09-21 17:03:05 -070040extern unsigned long totalram_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -070041extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070042extern int page_cluster;
43
44#ifdef CONFIG_SYSCTL
45extern int sysctl_legacy_va_layout;
46#else
47#define sysctl_legacy_va_layout 0
48#endif
49
50#include <asm/page.h>
51#include <asm/pgtable.h>
52#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
Tang Chen79442ed2013-11-12 15:07:59 -080054#ifndef __pa_symbol
55#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
56#endif
57
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -070058extern unsigned long sysctl_user_reserve_kbytes;
Andrew Shewmaker4eeab4f2013-04-29 15:08:11 -070059extern unsigned long sysctl_admin_reserve_kbytes;
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -070060
Jerome Marchand49f0ce52014-01-21 15:49:14 -080061extern int sysctl_overcommit_memory;
62extern int sysctl_overcommit_ratio;
63extern unsigned long sysctl_overcommit_kbytes;
64
65extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
66 size_t *, loff_t *);
67extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
68 size_t *, loff_t *);
69
Linus Torvalds1da177e2005-04-16 15:20:36 -070070#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
71
Andrea Righi27ac7922008-07-23 21:28:13 -070072/* to align the pointer to the (next) page boundary */
73#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
74
Andrew Morton0fa73b82013-07-03 15:02:11 -070075/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
76#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
77
Linus Torvalds1da177e2005-04-16 15:20:36 -070078/*
79 * Linux kernel virtual memory manager primitives.
80 * The idea being to have a "virtual" mm in the same way
81 * we have a virtual fs - giving a cleaner interface to the
82 * mm details, and allowing different kinds of memory mappings
83 * (from shared memory to executable loading to arbitrary
84 * mmap() functions).
85 */
86
Christoph Lameterc43692e2006-12-06 20:32:48 -080087extern struct kmem_cache *vm_area_cachep;
88
Linus Torvalds1da177e2005-04-16 15:20:36 -070089#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +000090extern struct rb_root nommu_region_tree;
91extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93extern unsigned int kobjsize(const void *objp);
94#endif
95
96/*
Hugh Dickins605d9282008-08-16 11:07:21 +010097 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070098 */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -070099#define VM_NONE 0x00000000
100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#define VM_READ 0x00000001 /* currently active flags */
102#define VM_WRITE 0x00000002
103#define VM_EXEC 0x00000004
104#define VM_SHARED 0x00000008
105
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -0700106/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
108#define VM_MAYWRITE 0x00000020
109#define VM_MAYEXEC 0x00000040
110#define VM_MAYSHARE 0x00000080
111
112#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
Linus Torvalds6aab3412005-11-28 14:34:23 -0800113#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116#define VM_LOCKED 0x00002000
117#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
118
119 /* Used by sys_madvise() */
120#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
121#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
122
123#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
124#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -0700126#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
128#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700129#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
Konstantin Khlebnikov0103bd12012-10-08 16:28:59 -0700130#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700132#ifdef CONFIG_MEM_SOFT_DIRTY
133# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
134#else
135# define VM_SOFTDIRTY 0
136#endif
137
Jared Hulbertb379d792008-04-28 02:12:58 -0700138#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700139#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
140#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700141#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700142
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700143#if defined(CONFIG_X86)
144# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
145#elif defined(CONFIG_PPC)
146# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
147#elif defined(CONFIG_PARISC)
148# define VM_GROWSUP VM_ARCH_1
James Hogan9ca52ed2012-10-16 10:16:14 +0100149#elif defined(CONFIG_METAG)
150# define VM_GROWSUP VM_ARCH_1
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700151#elif defined(CONFIG_IA64)
152# define VM_GROWSUP VM_ARCH_1
153#elif !defined(CONFIG_MMU)
154# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
155#endif
156
157#ifndef VM_GROWSUP
158# define VM_GROWSUP VM_NONE
159#endif
160
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700161/* Bits set in the VMA until the stack is in its final location */
162#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
163
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
165#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
166#endif
167
168#ifdef CONFIG_STACK_GROWSUP
169#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
170#else
171#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
172#endif
173
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174/*
Andrea Arcangeli78f11a22011-04-27 15:26:45 -0700175 * Special vmas that are non-mergable, non-mlock()able.
176 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
Nick Pigginb291f002008-10-18 20:26:44 -0700177 */
Vlastimil Babka9050d7e2014-03-03 15:38:27 -0800178#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
Nick Pigginb291f002008-10-18 20:26:44 -0700179
Alex Thorltona0715cc2014-04-07 15:37:10 -0700180/* This mask defines which mm->def_flags a process can inherit its parent */
181#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
182
Nick Pigginb291f002008-10-18 20:26:44 -0700183/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 * mapping from the currently active vm_flags protection bits (the
185 * low four bits) to a page protection mask..
186 */
187extern pgprot_t protection_map[16];
188
Nick Piggind0217ac2007-07-19 01:47:03 -0700189#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
190#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700191#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Michel Lespinassed065bd82010-10-26 14:21:57 -0700192#define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
Gleb Natapov318b2752011-03-22 16:30:51 -0700193#define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
KOSAKI Motohiro37b23e02011-05-24 17:11:30 -0700194#define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
Shaohua Li45cac652012-10-08 16:32:19 -0700195#define FAULT_FLAG_TRIED 0x40 /* second try */
Johannes Weiner759496b2013-09-12 15:13:39 -0700196#define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
Nick Piggind0217ac2007-07-19 01:47:03 -0700197
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800198/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700199 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700200 * ->fault function. The vma's ->fault is responsible for returning a bitmask
201 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700202 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700203 * pgoff should be used in favour of virtual_address, if possible. If pgoff
Konstantin Khlebnikov0b173bc2012-10-08 16:28:46 -0700204 * is used, one may implement ->remap_pages to get nonlinear mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700205 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700206struct vm_fault {
207 unsigned int flags; /* FAULT_FLAG_xxx flags */
208 pgoff_t pgoff; /* Logical page offset based on vma */
209 void __user *virtual_address; /* Faulting virtual address */
210
211 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700212 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700213 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700214 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700215 */
Kirill A. Shutemov8c6e50b2014-04-07 15:37:18 -0700216 /* for ->map_pages() only */
217 pgoff_t max_pgoff; /* map pages for offset from pgoff till
218 * max_pgoff inclusive */
219 pte_t *pte; /* pte entry associated with ->pgoff */
Nick Piggin54cb8822007-07-19 01:46:59 -0700220};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221
222/*
223 * These are the virtual MM functions - opening of an area, closing and
224 * unmapping it (needed to keep files on disk up-to-date etc), pointer
225 * to the functions called when a no-page or a wp-page exception occurs.
226 */
227struct vm_operations_struct {
228 void (*open)(struct vm_area_struct * area);
229 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700230 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
Kirill A. Shutemov8c6e50b2014-04-07 15:37:18 -0700231 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700232
233 /* notification that a previously read-only page is about to become
234 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700235 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700236
237 /* called by access_process_vm when get_user_pages() fails, typically
238 * for use by special VMAs that can switch between memory and hardware
239 */
240 int (*access)(struct vm_area_struct *vma, unsigned long addr,
241 void *buf, int len, int write);
Andy Lutomirski78d683e2014-05-19 15:58:32 -0700242
243 /* Called by the /proc/PID/maps code to ask the vma whether it
244 * has a special name. Returning non-NULL will also cause this
245 * vma to be dumped unconditionally. */
246 const char *(*name)(struct vm_area_struct *vma);
247
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700249 /*
250 * set_policy() op must add a reference to any non-NULL @new mempolicy
251 * to hold the policy upon return. Caller should pass NULL @new to
252 * remove a policy and fall back to surrounding context--i.e. do not
253 * install a MPOL_DEFAULT policy, nor the task or system default
254 * mempolicy.
255 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700257
258 /*
259 * get_policy() op must add reference [mpol_get()] to any policy at
260 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
261 * in mm/mempolicy.c will do this automatically.
262 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
263 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
264 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
265 * must return NULL--i.e., do not "fallback" to task or system default
266 * policy.
267 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
269 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700270 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
271 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272#endif
Konstantin Khlebnikov0b173bc2012-10-08 16:28:46 -0700273 /* called by sys_remap_file_pages() to populate non-linear mapping */
274 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
275 unsigned long size, pgoff_t pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276};
277
278struct mmu_gather;
279struct inode;
280
Andrew Morton349aef02006-01-08 01:04:36 -0800281#define page_private(page) ((page)->private)
282#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700283
Minchan Kimb12c4ad2012-10-08 16:32:08 -0700284/* It's valid only if the page is free path or free_list */
285static inline void set_freepage_migratetype(struct page *page, int migratetype)
286{
Minchan Kim95e34412012-10-08 16:32:11 -0700287 page->index = migratetype;
Minchan Kimb12c4ad2012-10-08 16:32:08 -0700288}
289
290/* It's valid only if the page is free path or free_list */
291static inline int get_freepage_migratetype(struct page *page)
292{
Minchan Kim95e34412012-10-08 16:32:11 -0700293 return page->index;
Minchan Kimb12c4ad2012-10-08 16:32:08 -0700294}
295
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296/*
297 * FIXME: take this include out, include page-flags.h in
298 * files which need it (119 of them)
299 */
300#include <linux/page-flags.h>
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800301#include <linux/huge_mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
303/*
304 * Methods to modify the page usage count.
305 *
306 * What counts for a page usage:
307 * - cache mapping (page->mapping)
308 * - private data (page->private)
309 * - page mapped in a task's page tables, each mapping
310 * is counted separately
311 *
312 * Also, many kernel routines increase the page count before a critical
313 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 */
315
316/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700317 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800319static inline int put_page_testzero(struct page *page)
320{
Sasha Levin309381fea2014-01-23 15:52:54 -0800321 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800322 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800323}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
325/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800326 * Try to grab a ref unless the page has a refcount of zero, return false if
327 * that is the case.
Alexey Kardashevskiy8e0861f2013-08-28 18:37:42 +1000328 * This can be called when MMU is off so it must not access
329 * any of the virtual mappings.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800331static inline int get_page_unless_zero(struct page *page)
332{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800333 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800334}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335
Alexey Kardashevskiy8e0861f2013-08-28 18:37:42 +1000336/*
337 * Try to drop a ref unless the page has a refcount of one, return false if
338 * that is the case.
339 * This is to make sure that the refcount won't become zero after this drop.
340 * This can be called when MMU is off so it must not access
341 * any of the virtual mappings.
342 */
343static inline int put_page_unless_one(struct page *page)
344{
345 return atomic_add_unless(&page->_count, -1, 1);
346}
347
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800348extern int page_is_ram(unsigned long pfn);
349
Christoph Lameter48667e72008-02-04 22:28:31 -0800350/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800351struct page *vmalloc_to_page(const void *addr);
352unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800353
Paul Mundt0738c4b2008-03-12 16:51:31 +0900354/*
355 * Determine if an address is within the vmalloc range
356 *
357 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
358 * is no special casing required.
359 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800360static inline int is_vmalloc_addr(const void *x)
361{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900362#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800363 unsigned long addr = (unsigned long)x;
364
365 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900366#else
367 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800368#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900369}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700370#ifdef CONFIG_MMU
371extern int is_vmalloc_or_module_addr(const void *x);
372#else
David Howells934831d2009-09-24 12:33:32 +0100373static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700374{
375 return 0;
376}
377#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800378
Al Viro39f1f782014-05-06 14:02:53 -0400379extern void kvfree(const void *addr);
380
Andrea Arcangelie9da73d2011-01-13 15:46:32 -0800381static inline void compound_lock(struct page *page)
382{
383#ifdef CONFIG_TRANSPARENT_HUGEPAGE
Sasha Levin309381fea2014-01-23 15:52:54 -0800384 VM_BUG_ON_PAGE(PageSlab(page), page);
Andrea Arcangelie9da73d2011-01-13 15:46:32 -0800385 bit_spin_lock(PG_compound_lock, &page->flags);
386#endif
387}
388
389static inline void compound_unlock(struct page *page)
390{
391#ifdef CONFIG_TRANSPARENT_HUGEPAGE
Sasha Levin309381fea2014-01-23 15:52:54 -0800392 VM_BUG_ON_PAGE(PageSlab(page), page);
Andrea Arcangelie9da73d2011-01-13 15:46:32 -0800393 bit_spin_unlock(PG_compound_lock, &page->flags);
394#endif
395}
396
397static inline unsigned long compound_lock_irqsave(struct page *page)
398{
399 unsigned long uninitialized_var(flags);
400#ifdef CONFIG_TRANSPARENT_HUGEPAGE
401 local_irq_save(flags);
402 compound_lock(page);
403#endif
404 return flags;
405}
406
407static inline void compound_unlock_irqrestore(struct page *page,
408 unsigned long flags)
409{
410#ifdef CONFIG_TRANSPARENT_HUGEPAGE
411 compound_unlock(page);
412 local_irq_restore(flags);
413#endif
414}
415
Jianyu Zhand2ee40e2014-06-04 16:08:02 -0700416static inline struct page *compound_head_by_tail(struct page *tail)
417{
418 struct page *head = tail->first_page;
419
420 /*
421 * page->first_page may be a dangling pointer to an old
422 * compound page, so recheck that it is still a tail
423 * page before returning.
424 */
425 smp_rmb();
426 if (likely(PageTail(tail)))
427 return head;
428 return tail;
429}
430
Christoph Lameterd85f3382007-05-06 14:49:39 -0700431static inline struct page *compound_head(struct page *page)
432{
Jianyu Zhand2ee40e2014-06-04 16:08:02 -0700433 if (unlikely(PageTail(page)))
434 return compound_head_by_tail(page);
Christoph Lameterd85f3382007-05-06 14:49:39 -0700435 return page;
436}
437
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700438/*
439 * The atomic page->_mapcount, starts from -1: so that transitions
440 * both from it and to it can be tracked, using atomic_inc_and_test
441 * and atomic_add_negative(-1).
442 */
Mel Gorman22b751c2013-02-22 16:34:59 -0800443static inline void page_mapcount_reset(struct page *page)
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700444{
445 atomic_set(&(page)->_mapcount, -1);
446}
447
448static inline int page_mapcount(struct page *page)
449{
450 return atomic_read(&(page)->_mapcount) + 1;
451}
452
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700453static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700455 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456}
457
Andrea Arcangeli44518d22014-01-21 15:48:54 -0800458#ifdef CONFIG_HUGETLB_PAGE
459extern int PageHeadHuge(struct page *page_head);
460#else /* CONFIG_HUGETLB_PAGE */
461static inline int PageHeadHuge(struct page *page_head)
462{
463 return 0;
464}
465#endif /* CONFIG_HUGETLB_PAGE */
466
467static inline bool __compound_tail_refcounted(struct page *page)
468{
469 return !PageSlab(page) && !PageHeadHuge(page);
470}
471
472/*
473 * This takes a head page as parameter and tells if the
474 * tail page reference counting can be skipped.
475 *
476 * For this to be safe, PageSlab and PageHeadHuge must remain true on
477 * any given page where they return true here, until all tail pins
478 * have been released.
479 */
480static inline bool compound_tail_refcounted(struct page *page)
481{
Sasha Levin309381fea2014-01-23 15:52:54 -0800482 VM_BUG_ON_PAGE(!PageHead(page), page);
Andrea Arcangeli44518d22014-01-21 15:48:54 -0800483 return __compound_tail_refcounted(page);
484}
485
Andrea Arcangelib35a35b2011-11-02 13:37:36 -0700486static inline void get_huge_page_tail(struct page *page)
487{
488 /*
Oleg Nesterov5eaf1a92014-01-21 15:49:04 -0800489 * __split_huge_page_refcount() cannot run from under us.
Andrea Arcangelib35a35b2011-11-02 13:37:36 -0700490 */
Sasha Levin309381fea2014-01-23 15:52:54 -0800491 VM_BUG_ON_PAGE(!PageTail(page), page);
492 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
493 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
Oleg Nesterov5eaf1a92014-01-21 15:49:04 -0800494 if (compound_tail_refcounted(page->first_page))
Andrea Arcangeli44518d22014-01-21 15:48:54 -0800495 atomic_inc(&page->_mapcount);
Andrea Arcangelib35a35b2011-11-02 13:37:36 -0700496}
497
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700498extern bool __get_page_tail(struct page *page);
499
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500static inline void get_page(struct page *page)
501{
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700502 if (unlikely(PageTail(page)))
503 if (likely(__get_page_tail(page)))
504 return;
Andrea Arcangeli91807062011-01-13 15:46:32 -0800505 /*
506 * Getting a normal page or the head of a compound page
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700507 * requires to already have an elevated page->_count.
Andrea Arcangeli91807062011-01-13 15:46:32 -0800508 */
Sasha Levin309381fea2014-01-23 15:52:54 -0800509 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 atomic_inc(&page->_count);
511}
512
Christoph Lameterb49af682007-05-06 14:49:41 -0700513static inline struct page *virt_to_head_page(const void *x)
514{
515 struct page *page = virt_to_page(x);
516 return compound_head(page);
517}
518
Nick Piggin7835e982006-03-22 00:08:40 -0800519/*
520 * Setup the page count before being freed into the page allocator for
521 * the first time (boot or memory hotplug)
522 */
523static inline void init_page_count(struct page *page)
524{
525 atomic_set(&page->_count, 1);
526}
527
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800528/*
529 * PageBuddy() indicate that the page is free and in the buddy system
530 * (see mm/page_alloc.c).
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100531 *
532 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
533 * -2 so that an underflow of the page_mapcount() won't be mistaken
534 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
535 * efficiently by most CPU architectures.
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800536 */
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100537#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
538
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800539static inline int PageBuddy(struct page *page)
540{
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100541 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800542}
543
544static inline void __SetPageBuddy(struct page *page)
545{
Sasha Levin309381fea2014-01-23 15:52:54 -0800546 VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
Andrea Arcangelief2b4b92011-03-18 00:16:35 +0100547 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800548}
549
550static inline void __ClearPageBuddy(struct page *page)
551{
Sasha Levin309381fea2014-01-23 15:52:54 -0800552 VM_BUG_ON_PAGE(!PageBuddy(page), page);
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800553 atomic_set(&page->_mapcount, -1);
554}
555
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700557void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800559void split_page(struct page *page, unsigned int order);
Mel Gorman748446b2010-05-24 14:32:27 -0700560int split_free_page(struct page *page);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800561
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800563 * Compound pages have a destructor function. Provide a
564 * prototype for that function and accessor functions.
565 * These are _only_ valid on the head of a PG_compound page.
566 */
567typedef void compound_page_dtor(struct page *);
568
569static inline void set_compound_page_dtor(struct page *page,
570 compound_page_dtor *dtor)
571{
572 page[1].lru.next = (void *)dtor;
573}
574
575static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
576{
577 return (compound_page_dtor *)page[1].lru.next;
578}
579
Christoph Lameterd85f3382007-05-06 14:49:39 -0700580static inline int compound_order(struct page *page)
581{
Christoph Lameter6d777952007-05-06 14:49:40 -0700582 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700583 return 0;
584 return (unsigned long)page[1].lru.prev;
585}
586
587static inline void set_compound_order(struct page *page, unsigned long order)
588{
589 page[1].lru.prev = (void *)order;
590}
591
Michal Simek3dece372011-01-21 08:49:56 +0100592#ifdef CONFIG_MMU
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800593/*
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800594 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
595 * servicing faults for write access. In the normal case, do always want
596 * pte_mkwrite. But get_user_pages can cause write faults for mappings
597 * that do not have writing enabled, when used by access_process_vm.
598 */
599static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
600{
601 if (likely(vma->vm_flags & VM_WRITE))
602 pte = pte_mkwrite(pte);
603 return pte;
604}
Kirill A. Shutemov8c6e50b2014-04-07 15:37:18 -0700605
606void do_set_pte(struct vm_area_struct *vma, unsigned long address,
607 struct page *page, pte_t *pte, bool write, bool anon);
Michal Simek3dece372011-01-21 08:49:56 +0100608#endif
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800609
610/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 * Multiple processes may "see" the same page. E.g. for untouched
612 * mappings of /dev/null, all processes see the same page full of
613 * zeroes, and text pages of executables and shared libraries have
614 * only one copy in memory, at most, normally.
615 *
616 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700617 * page_count() == 0 means the page is free. page->lru is then used for
618 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700619 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700621 * Pages are allocated by the slab allocator in order to provide memory
622 * to kmalloc and kmem_cache_alloc. In this case, the management of the
623 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
624 * unless a particular usage is carefully commented. (the responsibility of
625 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700627 * A page may be used by anyone else who does a __get_free_page().
628 * In this case, page_count still tracks the references, and should only
629 * be used through the normal accessor functions. The top bits of page->flags
630 * and page->virtual store page management information, but all other fields
631 * are unused and could be used privately, carefully. The management of this
632 * page is the responsibility of the one who allocated it, and those who have
633 * subsequently been given references to it.
634 *
635 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 * managed by the Linux memory manager: I/O, buffers, swapping etc.
637 * The following discussion applies only to them.
638 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700639 * A pagecache page contains an opaque `private' member, which belongs to the
640 * page's address_space. Usually, this is the address of a circular list of
641 * the page's disk buffers. PG_private must be set to tell the VM to call
642 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700644 * A page may belong to an inode's memory mapping. In this case, page->mapping
645 * is the pointer to the inode, and page->index is the file offset of the page,
646 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700648 * If pagecache pages are not associated with an inode, they are said to be
649 * anonymous pages. These may become associated with the swapcache, and in that
650 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700652 * In either case (swapcache or inode backed), the pagecache itself holds one
653 * reference to the page. Setting PG_private should also increment the
654 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700656 * The pagecache pages are stored in a per-mapping radix tree, which is
657 * rooted at mapping->page_tree, and indexed by offset.
658 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
659 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700661 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 * - inode pages may need to be read from disk,
663 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700664 * to be written back to the inode on disk,
665 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
666 * modified may need to be swapped out to swap space and (later) to be read
667 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 */
669
670/*
671 * The zone field is never updated after free_area_init_core()
672 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700674
Peter Zijlstra90572892013-10-07 11:29:20 +0100675/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100676#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700677#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
678#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
Peter Zijlstra90572892013-10-07 11:29:20 +0100679#define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700680
681/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300682 * Define the bit shifts to access each section. For non-existent
Dave Hansen348f8b62005-06-23 00:07:40 -0700683 * sections we define the shift as 0; that plus a 0 mask ensures
684 * the compiler will optimise away reference to them.
685 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700686#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
687#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
688#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Peter Zijlstra90572892013-10-07 11:29:20 +0100689#define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700690
Will Deaconbce54bb2010-10-26 14:21:37 -0700691/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
692#ifdef NODE_NOT_IN_PAGE_FLAGS
Christoph Lameter89689ae2006-12-06 20:31:45 -0800693#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800694#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
695 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700696#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800697#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800698#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
699 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700700#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800701
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800702#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700703
Christoph Lameter9223b412008-04-28 02:12:48 -0700704#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
705#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700706#endif
707
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700708#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
709#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
710#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Srikar Dronamraju834a9642014-04-07 15:37:57 -0700711#define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800712#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700713
Ian Campbell33dd4e02011-07-25 17:11:51 -0700714static inline enum zone_type page_zonenum(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715{
Dave Hansen348f8b62005-06-23 00:07:40 -0700716 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718
Cody P Schafer9127ab42013-02-22 16:35:21 -0800719#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
720#define SECTION_IN_PAGE_FLAGS
721#endif
722
Christoph Lameter89689ae2006-12-06 20:31:45 -0800723/*
Vlastimil Babka7a8010c2013-09-11 14:22:35 -0700724 * The identification function is mainly used by the buddy allocator for
725 * determining if two pages could be buddies. We are not really identifying
726 * the zone since we could be using the section number id if we do not have
727 * node id available in page flags.
728 * We only guarantee that it will return the same value for two combinable
729 * pages in a zone.
Christoph Lameter89689ae2006-12-06 20:31:45 -0800730 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700731static inline int page_zone_id(struct page *page)
732{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800733 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734}
735
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800736static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700737{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700738#ifdef CONFIG_NUMA
739 return zone->node;
740#else
741 return 0;
742#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700743}
744
Christoph Lameter89689ae2006-12-06 20:31:45 -0800745#ifdef NODE_NOT_IN_PAGE_FLAGS
Ian Campbell33dd4e02011-07-25 17:11:51 -0700746extern int page_to_nid(const struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800747#else
Ian Campbell33dd4e02011-07-25 17:11:51 -0700748static inline int page_to_nid(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700749{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800750 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700751}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800752#endif
753
Mel Gorman57e0a032012-11-12 09:06:20 +0000754#ifdef CONFIG_NUMA_BALANCING
Peter Zijlstra90572892013-10-07 11:29:20 +0100755static inline int cpu_pid_to_cpupid(int cpu, int pid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000756{
Peter Zijlstra90572892013-10-07 11:29:20 +0100757 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
Mel Gorman57e0a032012-11-12 09:06:20 +0000758}
759
Peter Zijlstra90572892013-10-07 11:29:20 +0100760static inline int cpupid_to_pid(int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000761{
Peter Zijlstra90572892013-10-07 11:29:20 +0100762 return cpupid & LAST__PID_MASK;
Mel Gorman57e0a032012-11-12 09:06:20 +0000763}
Mel Gormanb7958542013-10-07 11:29:07 +0100764
Peter Zijlstra90572892013-10-07 11:29:20 +0100765static inline int cpupid_to_cpu(int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000766{
Peter Zijlstra90572892013-10-07 11:29:20 +0100767 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
Mel Gormanb7958542013-10-07 11:29:07 +0100768}
769
Peter Zijlstra90572892013-10-07 11:29:20 +0100770static inline int cpupid_to_nid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100771{
Peter Zijlstra90572892013-10-07 11:29:20 +0100772 return cpu_to_node(cpupid_to_cpu(cpupid));
Mel Gormanb7958542013-10-07 11:29:07 +0100773}
774
Peter Zijlstra90572892013-10-07 11:29:20 +0100775static inline bool cpupid_pid_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100776{
Peter Zijlstra90572892013-10-07 11:29:20 +0100777 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100778}
779
Peter Zijlstra90572892013-10-07 11:29:20 +0100780static inline bool cpupid_cpu_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100781{
Peter Zijlstra90572892013-10-07 11:29:20 +0100782 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100783}
784
Peter Zijlstra8c8a7432013-10-07 11:29:21 +0100785static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
786{
787 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
788}
789
790#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
Peter Zijlstra90572892013-10-07 11:29:20 +0100791#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
792static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100793{
Liu Ping Fan1ae71d02014-03-03 15:38:39 -0800794 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100795}
Peter Zijlstra90572892013-10-07 11:29:20 +0100796
797static inline int page_cpupid_last(struct page *page)
Mel Gormanb7958542013-10-07 11:29:07 +0100798{
Peter Zijlstra90572892013-10-07 11:29:20 +0100799 return page->_last_cpupid;
800}
801static inline void page_cpupid_reset_last(struct page *page)
802{
Liu Ping Fan1ae71d02014-03-03 15:38:39 -0800803 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
Mel Gorman57e0a032012-11-12 09:06:20 +0000804}
805#else
Peter Zijlstra90572892013-10-07 11:29:20 +0100806static inline int page_cpupid_last(struct page *page)
Peter Zijlstra75980e92013-02-22 16:34:32 -0800807{
Peter Zijlstra90572892013-10-07 11:29:20 +0100808 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
Peter Zijlstra75980e92013-02-22 16:34:32 -0800809}
810
Peter Zijlstra90572892013-10-07 11:29:20 +0100811extern int page_cpupid_xchg_last(struct page *page, int cpupid);
Peter Zijlstra75980e92013-02-22 16:34:32 -0800812
Peter Zijlstra90572892013-10-07 11:29:20 +0100813static inline void page_cpupid_reset_last(struct page *page)
Peter Zijlstra75980e92013-02-22 16:34:32 -0800814{
Peter Zijlstra90572892013-10-07 11:29:20 +0100815 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
Mel Gorman4468b8f2013-02-22 16:34:46 -0800816
Peter Zijlstra90572892013-10-07 11:29:20 +0100817 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
818 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
Peter Zijlstra75980e92013-02-22 16:34:32 -0800819}
Peter Zijlstra90572892013-10-07 11:29:20 +0100820#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
821#else /* !CONFIG_NUMA_BALANCING */
822static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000823{
Peter Zijlstra90572892013-10-07 11:29:20 +0100824 return page_to_nid(page); /* XXX */
Mel Gorman57e0a032012-11-12 09:06:20 +0000825}
826
Peter Zijlstra90572892013-10-07 11:29:20 +0100827static inline int page_cpupid_last(struct page *page)
Mel Gorman57e0a032012-11-12 09:06:20 +0000828{
Peter Zijlstra90572892013-10-07 11:29:20 +0100829 return page_to_nid(page); /* XXX */
Mel Gorman57e0a032012-11-12 09:06:20 +0000830}
831
Peter Zijlstra90572892013-10-07 11:29:20 +0100832static inline int cpupid_to_nid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100833{
834 return -1;
835}
836
Peter Zijlstra90572892013-10-07 11:29:20 +0100837static inline int cpupid_to_pid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100838{
839 return -1;
840}
841
Peter Zijlstra90572892013-10-07 11:29:20 +0100842static inline int cpupid_to_cpu(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100843{
844 return -1;
845}
846
Peter Zijlstra90572892013-10-07 11:29:20 +0100847static inline int cpu_pid_to_cpupid(int nid, int pid)
848{
849 return -1;
850}
851
852static inline bool cpupid_pid_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100853{
854 return 1;
855}
856
Peter Zijlstra90572892013-10-07 11:29:20 +0100857static inline void page_cpupid_reset_last(struct page *page)
Mel Gorman57e0a032012-11-12 09:06:20 +0000858{
859}
Peter Zijlstra8c8a7432013-10-07 11:29:21 +0100860
861static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
862{
863 return false;
864}
Peter Zijlstra90572892013-10-07 11:29:20 +0100865#endif /* CONFIG_NUMA_BALANCING */
Mel Gorman57e0a032012-11-12 09:06:20 +0000866
Ian Campbell33dd4e02011-07-25 17:11:51 -0700867static inline struct zone *page_zone(const struct page *page)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800868{
869 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
870}
871
Cody P Schafer9127ab42013-02-22 16:35:21 -0800872#ifdef SECTION_IN_PAGE_FLAGS
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700873static inline void set_page_section(struct page *page, unsigned long section)
874{
875 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
876 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
877}
878
Ian Campbellaa462ab2011-08-17 17:40:33 +0100879static inline unsigned long page_to_section(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700880{
881 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
882}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700883#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700884
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700885static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886{
Dave Hansen348f8b62005-06-23 00:07:40 -0700887 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
888 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
889}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700890
Dave Hansen348f8b62005-06-23 00:07:40 -0700891static inline void set_page_node(struct page *page, unsigned long node)
892{
893 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
894 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
895}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800896
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700897static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700898 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700899{
900 set_page_zone(page, zone);
901 set_page_node(page, node);
Cody P Schafer9127ab42013-02-22 16:35:21 -0800902#ifdef SECTION_IN_PAGE_FLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700903 set_page_section(page, pfn_to_section_nr(pfn));
Daniel Kiperbf4e8902011-05-24 17:12:32 -0700904#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905}
906
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700907/*
908 * Some inline functions in vmstat.h depend on page_zone()
909 */
910#include <linux/vmstat.h>
911
Ian Campbell33dd4e02011-07-25 17:11:51 -0700912static __always_inline void *lowmem_page_address(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913{
Ian Campbellaa462ab2011-08-17 17:40:33 +0100914 return __va(PFN_PHYS(page_to_pfn(page)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916
917#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
918#define HASHED_PAGE_VIRTUAL
919#endif
920
921#if defined(WANT_PAGE_VIRTUAL)
Geert Uytterhoevenf92f4552014-01-21 15:48:47 -0800922static inline void *page_address(const struct page *page)
923{
924 return page->virtual;
925}
926static inline void set_page_address(struct page *page, void *address)
927{
928 page->virtual = address;
929}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930#define page_address_init() do { } while(0)
931#endif
932
933#if defined(HASHED_PAGE_VIRTUAL)
Ian Campbellf9918792011-08-17 13:45:09 +0100934void *page_address(const struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935void set_page_address(struct page *page, void *virtual);
936void page_address_init(void);
937#endif
938
939#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
940#define page_address(page) lowmem_page_address(page)
941#define set_page_address(page, address) do { } while(0)
942#define page_address_init() do { } while(0)
943#endif
944
945/*
946 * On an anonymous page mapped into a user virtual memory area,
947 * page->mapping points to its anon_vma, not to a struct address_space;
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800948 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
949 *
950 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
951 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
952 * and then page->mapping points, not to an anon_vma, but to a private
953 * structure which KSM associates with that merged page. See ksm.h.
954 *
955 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 *
957 * Please note that, confusingly, "page_mapping" refers to the inode
958 * address_space which maps the page from disk; whereas "page_mapped"
959 * refers to user virtual address space into which the page is mapped.
960 */
961#define PAGE_MAPPING_ANON 1
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800962#define PAGE_MAPPING_KSM 2
963#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964
Shaohua Li98003392013-02-22 16:34:35 -0800965extern struct address_space *page_mapping(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -0800967/* Neutral page->mapping pointer to address_space or anon_vma or other */
968static inline void *page_rmapping(struct page *page)
969{
970 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
971}
972
Mel Gormanf981c592012-07-31 16:44:47 -0700973extern struct address_space *__page_file_mapping(struct page *);
974
975static inline
976struct address_space *page_file_mapping(struct page *page)
977{
978 if (unlikely(PageSwapCache(page)))
979 return __page_file_mapping(page);
980
981 return page->mapping;
982}
983
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984static inline int PageAnon(struct page *page)
985{
986 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
987}
988
989/*
990 * Return the pagecache index of the passed page. Regular pagecache pages
991 * use ->index whereas swapcache pages use ->private
992 */
993static inline pgoff_t page_index(struct page *page)
994{
995 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700996 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 return page->index;
998}
999
Mel Gormanf981c592012-07-31 16:44:47 -07001000extern pgoff_t __page_file_index(struct page *page);
1001
1002/*
1003 * Return the file index of the page. Regular pagecache pages use ->index
1004 * whereas swapcache pages use swp_offset(->private)
1005 */
1006static inline pgoff_t page_file_index(struct page *page)
1007{
1008 if (unlikely(PageSwapCache(page)))
1009 return __page_file_index(page);
1010
1011 return page->index;
1012}
1013
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 * Return true if this page is mapped into pagetables.
1016 */
1017static inline int page_mapped(struct page *page)
1018{
1019 return atomic_read(&(page)->_mapcount) >= 0;
1020}
1021
1022/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 * Different kinds of faults, as returned by handle_mm_fault().
1024 * Used to decide whether a process gets delivered SIGBUS or
1025 * just gets major/minor fault counters bumped up.
1026 */
Nick Piggind0217ac2007-07-19 01:47:03 -07001027
Nick Piggin83c54072007-07-19 01:47:05 -07001028#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -07001029
Nick Piggin83c54072007-07-19 01:47:05 -07001030#define VM_FAULT_OOM 0x0001
1031#define VM_FAULT_SIGBUS 0x0002
1032#define VM_FAULT_MAJOR 0x0004
1033#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001034#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1035#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
Nick Pigginf33ea7f2005-08-03 20:24:01 +10001036
Nick Piggin83c54072007-07-19 01:47:05 -07001037#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1038#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Michel Lespinassed065bd82010-10-26 14:21:57 -07001039#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
Kirill A. Shutemovc0292552013-09-12 15:14:05 -07001040#define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001042#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1043
1044#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
Kirill A. Shutemovc0292552013-09-12 15:14:05 -07001045 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001046
1047/* Encode hstate index for a hwpoisoned large page */
1048#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1049#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
Nick Piggind0217ac2007-07-19 01:47:03 -07001050
Nick Piggin1c0fe6e2009-01-06 14:38:59 -08001051/*
1052 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1053 */
1054extern void pagefault_out_of_memory(void);
1055
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1057
David Rientjesddd588b2011-03-22 16:30:46 -07001058/*
David Rientjes7bf02ea2011-05-24 17:11:16 -07001059 * Flags passed to show_mem() and show_free_areas() to suppress output in
David Rientjesddd588b2011-03-22 16:30:46 -07001060 * various contexts.
1061 */
David Rientjes4b59e6c2013-04-29 15:06:11 -07001062#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
David Rientjesddd588b2011-03-22 16:30:46 -07001063
David Rientjes7bf02ea2011-05-24 17:11:16 -07001064extern void show_free_areas(unsigned int flags);
1065extern bool skip_free_areas_node(unsigned int flags, int nid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067int shmem_zero_setup(struct vm_area_struct *);
Johannes Weiner0cd61442014-04-03 14:47:46 -07001068#ifdef CONFIG_SHMEM
1069bool shmem_mapping(struct address_space *mapping);
1070#else
1071static inline bool shmem_mapping(struct address_space *mapping)
1072{
1073 return false;
1074}
1075#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +04001077extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078extern int user_shm_lock(size_t, struct user_struct *);
1079extern void user_shm_unlock(size_t, struct user_struct *);
1080
1081/*
1082 * Parameter block passed down to zap_pte_range in exceptional cases.
1083 */
1084struct zap_details {
1085 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
1086 struct address_space *check_mapping; /* Check page->mapping if set */
1087 pgoff_t first_index; /* Lowest page->index to unmap */
1088 pgoff_t last_index; /* Highest page->index to unmap */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089};
1090
Nick Piggin7e675132008-04-28 02:13:00 -07001091struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1092 pte_t pte);
1093
Jack Steinerc627f9c2008-07-29 22:33:53 -07001094int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1095 unsigned long size);
Al Viro14f5ff52012-03-05 13:38:09 -05001096void zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 unsigned long size, struct zap_details *);
Linus Torvalds4f74d2c2012-05-06 13:54:06 -07001098void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1099 unsigned long start, unsigned long end);
Matt Mackalle6473092008-02-04 22:29:01 -08001100
1101/**
1102 * mm_walk - callbacks for walk_page_range
1103 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1104 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1105 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
Dave Hansen03319322011-03-22 16:32:56 -07001106 * this handler is required to be able to handle
1107 * pmd_trans_huge() pmds. They may simply choose to
1108 * split_huge_page() instead of handling it explicitly.
Matt Mackalle6473092008-02-04 22:29:01 -08001109 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1110 * @pte_hole: if set, called for each hole at all levels
Naoya Horiguchi5dc37642009-12-14 18:00:01 -08001111 * @hugetlb_entry: if set, called for each hugetlb entry
KOSAKI Motohiroc27fe4c82011-07-25 17:12:10 -07001112 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1113 * is used.
Matt Mackalle6473092008-02-04 22:29:01 -08001114 *
1115 * (see walk_page_range for more details)
1116 */
1117struct mm_walk {
Andrew Morton0f157a52013-05-07 16:18:10 -07001118 int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1119 unsigned long next, struct mm_walk *walk);
1120 int (*pud_entry)(pud_t *pud, unsigned long addr,
1121 unsigned long next, struct mm_walk *walk);
1122 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1123 unsigned long next, struct mm_walk *walk);
1124 int (*pte_entry)(pte_t *pte, unsigned long addr,
1125 unsigned long next, struct mm_walk *walk);
1126 int (*pte_hole)(unsigned long addr, unsigned long next,
1127 struct mm_walk *walk);
1128 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1129 unsigned long addr, unsigned long next,
1130 struct mm_walk *walk);
Dave Hansen21650092008-06-12 15:21:47 -07001131 struct mm_struct *mm;
1132 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -08001133};
1134
Dave Hansen21650092008-06-12 15:21:47 -07001135int walk_page_range(unsigned long addr, unsigned long end,
1136 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -07001137void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -07001138 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1140 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141void unmap_mapping_range(struct address_space *mapping,
1142 loff_t const holebegin, loff_t const holelen, int even_cows);
Johannes Weiner3b6748e2009-06-16 15:32:35 -07001143int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1144 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -08001145int follow_phys(struct vm_area_struct *vma, unsigned long address,
1146 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -07001147int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1148 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
1150static inline void unmap_shared_mapping_range(struct address_space *mapping,
1151 loff_t const holebegin, loff_t const holelen)
1152{
1153 unmap_mapping_range(mapping, holebegin, holelen, 0);
1154}
1155
Kirill A. Shutemov7caef262013-09-12 15:13:56 -07001156extern void truncate_pagecache(struct inode *inode, loff_t new);
Christoph Hellwig2c27c652010-06-04 11:30:04 +02001157extern void truncate_setsize(struct inode *inode, loff_t newsize);
Hugh Dickins623e3db2012-03-28 14:42:40 -07001158void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
Nick Piggin750b4982009-09-16 11:50:12 +02001159int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +02001160int generic_error_remove_page(struct address_space *mapping, struct page *page);
Wu Fengguang83f78662009-09-16 11:50:13 +02001161int invalidate_inode_page(struct page *page);
1162
David Howells7ee1dd32006-01-06 00:11:44 -08001163#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -07001164extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
Linus Torvaldsd06063c2009-04-10 09:01:23 -07001165 unsigned long address, unsigned int flags);
Peter Zijlstra5c723ba2011-07-27 12:17:11 +02001166extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1167 unsigned long address, unsigned int fault_flags);
David Howells7ee1dd32006-01-06 00:11:44 -08001168#else
1169static inline int handle_mm_fault(struct mm_struct *mm,
1170 struct vm_area_struct *vma, unsigned long address,
Linus Torvaldsd06063c2009-04-10 09:01:23 -07001171 unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -08001172{
1173 /* should never happen if there's no MMU */
1174 BUG();
1175 return VM_FAULT_SIGBUS;
1176}
Peter Zijlstra5c723ba2011-07-27 12:17:11 +02001177static inline int fixup_user_fault(struct task_struct *tsk,
1178 struct mm_struct *mm, unsigned long address,
1179 unsigned int fault_flags)
1180{
1181 /* should never happen if there's no MMU */
1182 BUG();
1183 return -EFAULT;
1184}
David Howells7ee1dd32006-01-06 00:11:44 -08001185#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +10001186
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Stephen Wilson5ddd36b2011-03-13 15:49:20 -04001188extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1189 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190
Michel Lespinasse28a35712013-02-22 16:35:55 -08001191long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1192 unsigned long start, unsigned long nr_pages,
1193 unsigned int foll_flags, struct page **pages,
1194 struct vm_area_struct **vmas, int *nonblocking);
1195long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1196 unsigned long start, unsigned long nr_pages,
1197 int write, int force, struct page **pages,
1198 struct vm_area_struct **vmas);
Nick Piggind2bf6be2009-06-16 15:31:39 -07001199int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1200 struct page **pages);
Mel Gorman18022c52012-07-31 16:44:51 -07001201struct kvec;
1202int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1203 struct page **pages);
1204int get_kernel_page(unsigned long start, int write, struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07001205struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206
David Howellscf9a2ae2006-08-29 19:05:54 +01001207extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
Lukas Czernerd47992f2013-05-21 23:17:23 -04001208extern void do_invalidatepage(struct page *page, unsigned int offset,
1209 unsigned int length);
David Howellscf9a2ae2006-08-29 19:05:54 +01001210
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -08001212int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213int redirty_page_for_writepage(struct writeback_control *wbc,
1214 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -07001215void account_page_dirtied(struct page *page, struct address_space *mapping);
Michael Rubinf629d1c2010-10-26 14:21:33 -07001216void account_page_writeback(struct page *page);
Harvey Harrisonb3c97522008-02-13 15:03:15 -08001217int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218int set_page_dirty_lock(struct page *page);
1219int clear_page_dirty_for_io(struct page *page);
William Robertsa9090252014-02-11 10:11:59 -08001220int get_cmdline(struct task_struct *task, char *buffer, int buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221
Stefan Bader39aa3cb2010-08-31 15:52:27 +02001222/* Is the vma a continuation of the stack vma above it? */
Mikulas Patockaa09a79f2011-05-09 13:01:09 +02001223static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
Stefan Bader39aa3cb2010-08-31 15:52:27 +02001224{
1225 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1226}
1227
Mikulas Patockaa09a79f2011-05-09 13:01:09 +02001228static inline int stack_guard_page_start(struct vm_area_struct *vma,
1229 unsigned long addr)
1230{
1231 return (vma->vm_flags & VM_GROWSDOWN) &&
1232 (vma->vm_start == addr) &&
1233 !vma_growsdown(vma->vm_prev, addr);
1234}
1235
1236/* Is the vma a continuation of the stack vma below it? */
1237static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1238{
1239 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1240}
1241
1242static inline int stack_guard_page_end(struct vm_area_struct *vma,
1243 unsigned long addr)
1244{
1245 return (vma->vm_flags & VM_GROWSUP) &&
1246 (vma->vm_end == addr) &&
1247 !vma_growsup(vma->vm_next, addr);
1248}
1249
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07001250extern pid_t
1251vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1252
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001253extern unsigned long move_page_tables(struct vm_area_struct *vma,
1254 unsigned long old_addr, struct vm_area_struct *new_vma,
Michel Lespinasse38a76012012-10-08 16:31:50 -07001255 unsigned long new_addr, unsigned long len,
1256 bool need_rmap_locks);
Peter Zijlstra7da4d642012-11-19 03:14:23 +01001257extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1258 unsigned long end, pgprot_t newprot,
Mel Gorman4b10e7d2012-10-25 14:16:32 +02001259 int dirty_accountable, int prot_numa);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001260extern int mprotect_fixup(struct vm_area_struct *vma,
1261 struct vm_area_struct **pprev, unsigned long start,
1262 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263
Nick Piggin21cc1992008-07-25 19:45:22 -07001264/*
Peter Zijlstra465a4542009-06-15 12:31:37 +02001265 * doesn't attempt to fault and will return short.
1266 */
1267int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1268 struct page **pages);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001269/*
1270 * per-process(per-mm_struct) statistics.
1271 */
Matt Fleming172703b2011-05-24 17:12:36 -07001272static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1273{
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001274 long val = atomic_long_read(&mm->rss_stat.count[member]);
1275
1276#ifdef SPLIT_RSS_COUNTING
1277 /*
1278 * counter is updated in asynchronous manner and may go to minus.
1279 * But it's never be expected number for users.
1280 */
1281 if (val < 0)
1282 val = 0;
Matt Fleming172703b2011-05-24 17:12:36 -07001283#endif
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001284 return (unsigned long)val;
1285}
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001286
1287static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1288{
1289 atomic_long_add(value, &mm->rss_stat.count[member]);
1290}
1291
1292static inline void inc_mm_counter(struct mm_struct *mm, int member)
1293{
1294 atomic_long_inc(&mm->rss_stat.count[member]);
1295}
1296
1297static inline void dec_mm_counter(struct mm_struct *mm, int member)
1298{
1299 atomic_long_dec(&mm->rss_stat.count[member]);
1300}
1301
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001302static inline unsigned long get_mm_rss(struct mm_struct *mm)
1303{
1304 return get_mm_counter(mm, MM_FILEPAGES) +
1305 get_mm_counter(mm, MM_ANONPAGES);
1306}
1307
1308static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1309{
1310 return max(mm->hiwater_rss, get_mm_rss(mm));
1311}
1312
1313static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1314{
1315 return max(mm->hiwater_vm, mm->total_vm);
1316}
1317
1318static inline void update_hiwater_rss(struct mm_struct *mm)
1319{
1320 unsigned long _rss = get_mm_rss(mm);
1321
1322 if ((mm)->hiwater_rss < _rss)
1323 (mm)->hiwater_rss = _rss;
1324}
1325
1326static inline void update_hiwater_vm(struct mm_struct *mm)
1327{
1328 if (mm->hiwater_vm < mm->total_vm)
1329 mm->hiwater_vm = mm->total_vm;
1330}
1331
1332static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1333 struct mm_struct *mm)
1334{
1335 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1336
1337 if (*maxrss < hiwater_rss)
1338 *maxrss = hiwater_rss;
1339}
1340
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001341#if defined(SPLIT_RSS_COUNTING)
David Rientjes05af2e12012-03-21 16:34:13 -07001342void sync_mm_rss(struct mm_struct *mm);
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001343#else
David Rientjes05af2e12012-03-21 16:34:13 -07001344static inline void sync_mm_rss(struct mm_struct *mm)
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001345{
1346}
1347#endif
Peter Zijlstra465a4542009-06-15 12:31:37 +02001348
Alexey Dobriyan4e950f62007-07-30 02:36:13 +04001349int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -07001350
Namhyung Kim25ca1d62010-10-26 14:21:59 -07001351extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1352 spinlock_t **ptl);
1353static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1354 spinlock_t **ptl)
1355{
1356 pte_t *ptep;
1357 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1358 return ptep;
1359}
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -08001360
Nick Piggin5f22df02007-05-06 14:49:02 -07001361#ifdef __PAGETABLE_PUD_FOLDED
1362static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1363 unsigned long address)
1364{
1365 return 0;
1366}
1367#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001368int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001369#endif
1370
1371#ifdef __PAGETABLE_PMD_FOLDED
1372static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1373 unsigned long address)
1374{
1375 return 0;
1376}
1377#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001378int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -07001379#endif
1380
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001381int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1382 pmd_t *pmd, unsigned long address);
Hugh Dickins1bb36302005-10-29 18:16:22 -07001383int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1384
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385/*
1386 * The following ifdef needed to get the 4level-fixup.h header to work.
1387 * Remove it when 4level-fixup.h has been removed.
1388 */
Hugh Dickins1bb36302005-10-29 18:16:22 -07001389#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1391{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001392 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1393 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394}
1395
1396static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1397{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001398 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1399 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400}
Hugh Dickins1bb36302005-10-29 18:16:22 -07001401#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1402
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001403#if USE_SPLIT_PTE_PTLOCKS
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001404#if ALLOC_SPLIT_PTLOCKS
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001405void __init ptlock_cache_init(void);
Peter Zijlstra539edb52013-11-14 14:31:52 -08001406extern bool ptlock_alloc(struct page *page);
1407extern void ptlock_free(struct page *page);
1408
1409static inline spinlock_t *ptlock_ptr(struct page *page)
1410{
1411 return page->ptl;
1412}
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001413#else /* ALLOC_SPLIT_PTLOCKS */
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001414static inline void ptlock_cache_init(void)
1415{
1416}
1417
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001418static inline bool ptlock_alloc(struct page *page)
1419{
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001420 return true;
1421}
Peter Zijlstra539edb52013-11-14 14:31:52 -08001422
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001423static inline void ptlock_free(struct page *page)
1424{
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001425}
1426
1427static inline spinlock_t *ptlock_ptr(struct page *page)
1428{
Peter Zijlstra539edb52013-11-14 14:31:52 -08001429 return &page->ptl;
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001430}
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001431#endif /* ALLOC_SPLIT_PTLOCKS */
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001432
1433static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1434{
1435 return ptlock_ptr(pmd_page(*pmd));
1436}
1437
1438static inline bool ptlock_init(struct page *page)
1439{
1440 /*
1441 * prep_new_page() initialize page->private (and therefore page->ptl)
1442 * with 0. Make sure nobody took it in use in between.
1443 *
1444 * It can happen if arch try to use slab for page table allocation:
1445 * slab code uses page->slab_cache and page->first_page (for tail
1446 * pages), which share storage with page->ptl.
1447 */
Sasha Levin309381fea2014-01-23 15:52:54 -08001448 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001449 if (!ptlock_alloc(page))
1450 return false;
1451 spin_lock_init(ptlock_ptr(page));
1452 return true;
1453}
1454
1455/* Reset page->mapping so free_pages_check won't complain. */
1456static inline void pte_lock_deinit(struct page *page)
1457{
1458 page->mapping = NULL;
1459 ptlock_free(page);
1460}
1461
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001462#else /* !USE_SPLIT_PTE_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001463/*
1464 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1465 */
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001466static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1467{
1468 return &mm->page_table_lock;
1469}
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001470static inline void ptlock_cache_init(void) {}
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001471static inline bool ptlock_init(struct page *page) { return true; }
1472static inline void pte_lock_deinit(struct page *page) {}
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001473#endif /* USE_SPLIT_PTE_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001474
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001475static inline void pgtable_init(void)
1476{
1477 ptlock_cache_init();
1478 pgtable_cache_init();
1479}
1480
Kirill A. Shutemov390f44e2013-11-14 14:31:20 -08001481static inline bool pgtable_page_ctor(struct page *page)
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001482{
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001483 inc_zone_page_state(page, NR_PAGETABLE);
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001484 return ptlock_init(page);
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001485}
1486
1487static inline void pgtable_page_dtor(struct page *page)
1488{
1489 pte_lock_deinit(page);
1490 dec_zone_page_state(page, NR_PAGETABLE);
1491}
1492
Hugh Dickinsc74df322005-10-29 18:16:23 -07001493#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1494({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001495 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001496 pte_t *__pte = pte_offset_map(pmd, address); \
1497 *(ptlp) = __ptl; \
1498 spin_lock(__ptl); \
1499 __pte; \
1500})
1501
1502#define pte_unmap_unlock(pte, ptl) do { \
1503 spin_unlock(ptl); \
1504 pte_unmap(pte); \
1505} while (0)
1506
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001507#define pte_alloc_map(mm, vma, pmd, address) \
1508 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1509 pmd, address))? \
1510 NULL: pte_offset_map(pmd, address))
Hugh Dickins1bb36302005-10-29 18:16:22 -07001511
Hugh Dickinsc74df322005-10-29 18:16:23 -07001512#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001513 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1514 pmd, address))? \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001515 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1516
Hugh Dickins1bb36302005-10-29 18:16:22 -07001517#define pte_alloc_kernel(pmd, address) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001518 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
Hugh Dickins1bb36302005-10-29 18:16:22 -07001519 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001521#if USE_SPLIT_PMD_PTLOCKS
1522
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001523static struct page *pmd_to_page(pmd_t *pmd)
1524{
1525 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1526 return virt_to_page((void *)((unsigned long) pmd & mask));
1527}
1528
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001529static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1530{
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001531 return ptlock_ptr(pmd_to_page(pmd));
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001532}
1533
1534static inline bool pgtable_pmd_page_ctor(struct page *page)
1535{
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001536#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1537 page->pmd_huge_pte = NULL;
1538#endif
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001539 return ptlock_init(page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001540}
1541
1542static inline void pgtable_pmd_page_dtor(struct page *page)
1543{
1544#ifdef CONFIG_TRANSPARENT_HUGEPAGE
Sasha Levin309381fea2014-01-23 15:52:54 -08001545 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001546#endif
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001547 ptlock_free(page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001548}
1549
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001550#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001551
1552#else
1553
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001554static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1555{
1556 return &mm->page_table_lock;
1557}
1558
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001559static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1560static inline void pgtable_pmd_page_dtor(struct page *page) {}
1561
Kirill A. Shutemovc389a252013-11-14 14:30:59 -08001562#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001563
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001564#endif
1565
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001566static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1567{
1568 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1569 spin_lock(ptl);
1570 return ptl;
1571}
1572
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001574extern void free_area_init_node(int nid, unsigned long * zones_size,
1575 unsigned long zone_start_pfn, unsigned long *zholes_size);
David Howells49a7f042012-03-28 18:30:03 +01001576extern void free_initmem(void);
1577
Jiang Liu69afade2013-04-29 15:06:21 -07001578/*
1579 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1580 * into the buddy system. The freed pages will be poisoned with pattern
Jiang Liudbe67df2013-07-03 15:02:51 -07001581 * "poison" if it's within range [0, UCHAR_MAX].
Jiang Liu69afade2013-04-29 15:06:21 -07001582 * Return pages freed into the buddy system.
1583 */
Jiang Liu11199692013-07-03 15:02:48 -07001584extern unsigned long free_reserved_area(void *start, void *end,
Jiang Liu69afade2013-04-29 15:06:21 -07001585 int poison, char *s);
Jiang Liuc3d5f5f2013-07-03 15:03:14 -07001586
Jiang Liucfa11e02013-04-29 15:07:00 -07001587#ifdef CONFIG_HIGHMEM
1588/*
1589 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1590 * and totalram_pages.
1591 */
1592extern void free_highmem_page(struct page *page);
1593#endif
Jiang Liu69afade2013-04-29 15:06:21 -07001594
Jiang Liuc3d5f5f2013-07-03 15:03:14 -07001595extern void adjust_managed_page_count(struct page *page, long count);
Jiang Liu7ee3d4e2013-07-03 15:03:41 -07001596extern void mem_init_print_info(const char *str);
Jiang Liu69afade2013-04-29 15:06:21 -07001597
1598/* Free the reserved page into the buddy system, so it gets managed. */
1599static inline void __free_reserved_page(struct page *page)
1600{
1601 ClearPageReserved(page);
1602 init_page_count(page);
1603 __free_page(page);
1604}
1605
1606static inline void free_reserved_page(struct page *page)
1607{
1608 __free_reserved_page(page);
1609 adjust_managed_page_count(page, 1);
1610}
1611
1612static inline void mark_page_reserved(struct page *page)
1613{
1614 SetPageReserved(page);
1615 adjust_managed_page_count(page, -1);
1616}
1617
1618/*
1619 * Default method to free all the __init memory into the buddy system.
Jiang Liudbe67df2013-07-03 15:02:51 -07001620 * The freed pages will be poisoned with pattern "poison" if it's within
1621 * range [0, UCHAR_MAX].
1622 * Return pages freed into the buddy system.
Jiang Liu69afade2013-04-29 15:06:21 -07001623 */
1624static inline unsigned long free_initmem_default(int poison)
1625{
1626 extern char __init_begin[], __init_end[];
1627
Jiang Liu11199692013-07-03 15:02:48 -07001628 return free_reserved_area(&__init_begin, &__init_end,
Jiang Liu69afade2013-04-29 15:06:21 -07001629 poison, "unused kernel");
1630}
1631
Jiang Liu7ee3d4e2013-07-03 15:03:41 -07001632static inline unsigned long get_num_physpages(void)
1633{
1634 int nid;
1635 unsigned long phys_pages = 0;
1636
1637 for_each_online_node(nid)
1638 phys_pages += node_present_pages(nid);
1639
1640 return phys_pages;
1641}
1642
Tejun Heo0ee332c2011-12-08 10:22:09 -08001643#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gormanc7132162006-09-27 01:49:43 -07001644/*
Tejun Heo0ee332c2011-12-08 10:22:09 -08001645 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
Mel Gormanc7132162006-09-27 01:49:43 -07001646 * zones, allocate the backing mem_map and account for memory holes in a more
1647 * architecture independent manner. This is a substitute for creating the
1648 * zone_sizes[] and zholes_size[] arrays and passing them to
1649 * free_area_init_node()
1650 *
1651 * An architecture is expected to register range of page frames backed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08001652 * physical memory with memblock_add[_node]() before calling
Mel Gormanc7132162006-09-27 01:49:43 -07001653 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1654 * usage, an architecture is expected to do something like
1655 *
1656 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1657 * max_highmem_pfn};
1658 * for_each_valid_physical_page_range()
Tejun Heo0ee332c2011-12-08 10:22:09 -08001659 * memblock_add_node(base, size, nid)
Mel Gormanc7132162006-09-27 01:49:43 -07001660 * free_area_init_nodes(max_zone_pfns);
1661 *
Tejun Heo0ee332c2011-12-08 10:22:09 -08001662 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1663 * registered physical page range. Similarly
1664 * sparse_memory_present_with_active_regions() calls memory_present() for
1665 * each range when SPARSEMEM is enabled.
Mel Gormanc7132162006-09-27 01:49:43 -07001666 *
1667 * See mm/page_alloc.c for more information on each function exposed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08001668 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
Mel Gormanc7132162006-09-27 01:49:43 -07001669 */
1670extern void free_area_init_nodes(unsigned long *max_zone_pfn);
Tejun Heo1e019792011-07-12 09:45:34 +02001671unsigned long node_map_pfn_alignment(void);
Yinghai Lu32996252009-12-15 17:59:02 -08001672unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1673 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001674extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1675 unsigned long end_pfn);
1676extern void get_pfn_range_for_nid(unsigned int nid,
1677 unsigned long *start_pfn, unsigned long *end_pfn);
1678extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001679extern void free_bootmem_with_active_regions(int nid,
1680 unsigned long max_low_pfn);
1681extern void sparse_memory_present_with_active_regions(int nid);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001682
Tejun Heo0ee332c2011-12-08 10:22:09 -08001683#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
Tejun Heo5dfe8662011-07-14 09:46:10 +02001684
Tejun Heo0ee332c2011-12-08 10:22:09 -08001685#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001686 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1687static inline int __early_pfn_to_nid(unsigned long pfn)
1688{
1689 return 0;
1690}
1691#else
1692/* please see mm/page_alloc.c */
1693extern int __meminit early_pfn_to_nid(unsigned long pfn);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001694/* there is a per-arch backend function. */
1695extern int __meminit __early_pfn_to_nid(unsigned long pfn);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001696#endif
1697
Mel Gorman0e0b8642006-09-27 01:49:56 -07001698extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001699extern void memmap_init_zone(unsigned long, int, unsigned long,
1700 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07001701extern void setup_per_zone_wmarks(void);
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -07001702extern int __meminit init_per_zone_wmark_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001704extern void __init mmap_init(void);
David Rientjesb2b755b2011-03-24 15:18:15 -07001705extern void show_mem(unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706extern void si_meminfo(struct sysinfo * val);
1707extern void si_meminfo_node(struct sysinfo *val, int nid);
1708
Joe Perches3ee9a4f2011-10-31 17:08:35 -07001709extern __printf(3, 4)
1710void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
Dave Hansena238ab52011-05-24 17:12:16 -07001711
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001712extern void setup_per_cpu_pageset(void);
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001713
Shaohua Li112067f2009-09-21 17:01:16 -07001714extern void zone_pcp_update(struct zone *zone);
Jiang Liu340175b2012-07-31 16:43:32 -07001715extern void zone_pcp_reset(struct zone *zone);
Shaohua Li112067f2009-09-21 17:01:16 -07001716
Paul Szabo75f7ad82013-02-22 16:34:42 -08001717/* page_alloc.c */
1718extern int min_free_kbytes;
1719
David Howells8feae132009-01-08 12:04:47 +00001720/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001721extern atomic_long_t mmap_pages_allocated;
David Howells7e660872010-01-15 17:01:39 -08001722extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
David Howells8feae132009-01-08 12:04:47 +00001723
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07001724/* interval_tree.c */
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07001725void vma_interval_tree_insert(struct vm_area_struct *node,
1726 struct rb_root *root);
Michel Lespinasse9826a512012-10-08 16:31:35 -07001727void vma_interval_tree_insert_after(struct vm_area_struct *node,
1728 struct vm_area_struct *prev,
1729 struct rb_root *root);
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07001730void vma_interval_tree_remove(struct vm_area_struct *node,
1731 struct rb_root *root);
1732struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1733 unsigned long start, unsigned long last);
1734struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1735 unsigned long start, unsigned long last);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07001737#define vma_interval_tree_foreach(vma, root, start, last) \
1738 for (vma = vma_interval_tree_iter_first(root, start, last); \
1739 vma; vma = vma_interval_tree_iter_next(vma, start, last))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
1741static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1742 struct list_head *list)
1743{
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07001744 list_add_tail(&vma->shared.nonlinear, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745}
1746
Michel Lespinassebf181b92012-10-08 16:31:39 -07001747void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1748 struct rb_root *root);
1749void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1750 struct rb_root *root);
1751struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1752 struct rb_root *root, unsigned long start, unsigned long last);
1753struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1754 struct anon_vma_chain *node, unsigned long start, unsigned long last);
Michel Lespinasseed8ea812012-10-08 16:31:45 -07001755#ifdef CONFIG_DEBUG_VM_RB
1756void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1757#endif
Michel Lespinassebf181b92012-10-08 16:31:39 -07001758
1759#define anon_vma_interval_tree_foreach(avc, root, start, last) \
1760 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1761 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1762
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001764extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Rik van Riel5beb4932010-03-05 13:42:07 -08001765extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1767extern struct vm_area_struct *vma_merge(struct mm_struct *,
1768 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1769 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1770 struct mempolicy *);
1771extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1772extern int split_vma(struct mm_struct *,
1773 struct vm_area_struct *, unsigned long addr, int new_below);
1774extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1775extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1776 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001777extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
Michel Lespinasse38a76012012-10-08 16:31:50 -07001779 unsigned long addr, unsigned long len, pgoff_t pgoff,
1780 bool *need_rmap_locks);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001782
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001783extern int mm_take_all_locks(struct mm_struct *mm);
1784extern void mm_drop_all_locks(struct mm_struct *mm);
1785
Jiri Slaby38646012011-05-26 16:25:46 -07001786extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1787extern struct file *get_mm_exe_file(struct mm_struct *mm);
Matt Helsley925d1c42008-04-29 01:01:36 -07001788
akpm@osdl.org119f6572005-05-01 08:58:35 -07001789extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Stefani Seibold3935ed62014-03-17 23:22:02 +01001790extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1791 unsigned long addr, unsigned long len,
Andy Lutomirskia62c34b2014-05-19 15:58:33 -07001792 unsigned long flags,
1793 const struct vm_special_mapping *spec);
1794/* This is an obsolete alternative to _install_special_mapping. */
Roland McGrathfa5dc222007-02-08 14:20:41 -08001795extern int install_special_mapping(struct mm_struct *mm,
1796 unsigned long addr, unsigned long len,
1797 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798
1799extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1800
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001801extern unsigned long mmap_region(struct file *file, unsigned long addr,
Michel Lespinassec22c0d62013-02-22 16:32:43 -08001802 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08001803extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1804 unsigned long len, unsigned long prot, unsigned long flags,
Michel Lespinasse41badc12013-02-22 16:32:47 -08001805 unsigned long pgoff, unsigned long *populate);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1807
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08001808#ifdef CONFIG_MMU
1809extern int __mm_populate(unsigned long addr, unsigned long len,
1810 int ignore_errors);
1811static inline void mm_populate(unsigned long addr, unsigned long len)
1812{
1813 /* Ignore errors */
1814 (void) __mm_populate(addr, len, 1);
1815}
1816#else
1817static inline void mm_populate(unsigned long addr, unsigned long len) {}
1818#endif
1819
Linus Torvaldse4eb1ff2012-04-20 15:35:40 -07001820/* These take the mm semaphore themselves */
1821extern unsigned long vm_brk(unsigned long, unsigned long);
Al Virobfce2812012-04-20 21:57:04 -04001822extern int vm_munmap(unsigned long, size_t);
Linus Torvalds6be5ceb2012-04-20 17:13:58 -07001823extern unsigned long vm_mmap(struct file *, unsigned long,
1824 unsigned long, unsigned long,
1825 unsigned long, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826
Michel Lespinassedb4fbfb2012-12-11 16:01:49 -08001827struct vm_unmapped_area_info {
1828#define VM_UNMAPPED_AREA_TOPDOWN 1
1829 unsigned long flags;
1830 unsigned long length;
1831 unsigned long low_limit;
1832 unsigned long high_limit;
1833 unsigned long align_mask;
1834 unsigned long align_offset;
1835};
1836
1837extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1838extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1839
1840/*
1841 * Search for an unmapped address range.
1842 *
1843 * We are looking for a range that:
1844 * - does not intersect with any VMA;
1845 * - is contained within the [low_limit, high_limit) interval;
1846 * - is at least the desired size.
1847 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1848 */
1849static inline unsigned long
1850vm_unmapped_area(struct vm_unmapped_area_info *info)
1851{
1852 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1853 return unmapped_area(info);
1854 else
1855 return unmapped_area_topdown(info);
1856}
1857
Hugh Dickins85821aa2011-07-25 17:12:23 -07001858/* truncate.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001860extern void truncate_inode_pages_range(struct address_space *,
1861 loff_t lstart, loff_t lend);
Johannes Weiner91b0abe2014-04-03 14:47:49 -07001862extern void truncate_inode_pages_final(struct address_space *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863
1864/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001865extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Kirill A. Shutemovf1820362014-04-07 15:37:19 -07001866extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
Jan Kara4fcf1c62012-06-12 16:20:29 +02001867extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868
1869/* mm/page-writeback.c */
1870int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001871void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872
1873/* readahead.c */
1874#define VM_MAX_READAHEAD 128 /* kbytes */
1875#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001878 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001879
1880void page_cache_sync_readahead(struct address_space *mapping,
1881 struct file_ra_state *ra,
1882 struct file *filp,
1883 pgoff_t offset,
1884 unsigned long size);
1885
1886void page_cache_async_readahead(struct address_space *mapping,
1887 struct file_ra_state *ra,
1888 struct file *filp,
1889 struct page *pg,
1890 pgoff_t offset,
1891 unsigned long size);
1892
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893unsigned long max_sane_readahead(unsigned long nr);
1894
Michal Hockod05f3162011-05-24 17:11:44 -07001895/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001896extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Michal Hockod05f3162011-05-24 17:11:44 -07001897
1898/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1899extern int expand_downwards(struct vm_area_struct *vma,
1900 unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001901#if VM_GROWSUP
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001902extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Luck, Tony8ca3eb02010-08-24 11:44:18 -07001903#else
1904 #define expand_upwards(vma, address) do { } while (0)
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001905#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906
1907/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1908extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1909extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1910 struct vm_area_struct **pprev);
1911
1912/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1913 NULL if none. Assume start_addr < end_addr. */
1914static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1915{
1916 struct vm_area_struct * vma = find_vma(mm,start_addr);
1917
1918 if (vma && end_addr <= vma->vm_start)
1919 vma = NULL;
1920 return vma;
1921}
1922
1923static inline unsigned long vma_pages(struct vm_area_struct *vma)
1924{
1925 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1926}
1927
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001928/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1929static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1930 unsigned long vm_start, unsigned long vm_end)
1931{
1932 struct vm_area_struct *vma = find_vma(mm, vm_start);
1933
1934 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1935 vma = NULL;
1936
1937 return vma;
1938}
1939
David Howellsbad849b2010-08-26 16:00:34 +01001940#ifdef CONFIG_MMU
Hugh Dickins804af2c2006-07-26 21:39:49 +01001941pgprot_t vm_get_page_prot(unsigned long vm_flags);
David Howellsbad849b2010-08-26 16:00:34 +01001942#else
1943static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1944{
1945 return __pgprot(0);
1946}
1947#endif
1948
Aneesh Kumar K.V58772312013-12-06 00:08:22 +05301949#ifdef CONFIG_NUMA_BALANCING
Mel Gorman4b10e7d2012-10-25 14:16:32 +02001950unsigned long change_prot_numa(struct vm_area_struct *vma,
Lee Schermerhornb24f53a2012-10-25 14:16:32 +02001951 unsigned long start, unsigned long end);
1952#endif
1953
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001954struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001955int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1956 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001957int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001958int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1959 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001960int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1961 unsigned long pfn);
Linus Torvaldsb4cbb192013-04-16 13:45:37 -07001962int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1963
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964
Michel Lespinasse240aade2013-02-22 16:35:56 -08001965struct page *follow_page_mask(struct vm_area_struct *vma,
1966 unsigned long address, unsigned int foll_flags,
1967 unsigned int *page_mask);
1968
1969static inline struct page *follow_page(struct vm_area_struct *vma,
1970 unsigned long address, unsigned int foll_flags)
1971{
1972 unsigned int unused_page_mask;
1973 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1974}
1975
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001976#define FOLL_WRITE 0x01 /* check pte is writable */
1977#define FOLL_TOUCH 0x02 /* mark page accessed */
1978#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a602009-09-21 17:03:26 -07001979#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07001980#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
Gleb Natapov318b2752011-03-22 16:30:51 -07001981#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1982 * and return without waiting upon it */
Michel Lespinasse110d74a2011-01-13 15:46:11 -08001983#define FOLL_MLOCK 0x40 /* mark page as mlocked */
Andrea Arcangeli500d65d2011-01-13 15:46:55 -08001984#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
Huang Ying69ebb832011-01-30 11:15:48 +08001985#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
Andrea Arcangeli0b9d7052012-10-05 21:36:27 +02001986#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
Hugh Dickins5117b3b2013-02-22 16:36:07 -08001987#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001989typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001990 void *data);
1991extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1992 unsigned long size, pte_fn_t fn, void *data);
1993
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001995void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001997static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 unsigned long flags, struct file *file, long pages)
1999{
Huang Shijie44de9d02012-07-31 16:41:49 -07002000 mm->total_vm += pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001}
2002#endif /* CONFIG_PROC_FS */
2003
Ingo Molnar12d6f212008-01-30 13:33:58 +01002004#ifdef CONFIG_DEBUG_PAGEALLOC
Ingo Molnar12d6f212008-01-30 13:33:58 +01002005extern void kernel_map_pages(struct page *page, int numpages, int enable);
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01002006#ifdef CONFIG_HIBERNATION
2007extern bool kernel_page_present(struct page *page);
2008#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01002009#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07002011kernel_map_pages(struct page *page, int numpages, int enable) {}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01002012#ifdef CONFIG_HIBERNATION
2013static inline bool kernel_page_present(struct page *page) { return true; }
2014#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015#endif
2016
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002017#ifdef __HAVE_ARCH_GATE_AREA
Stephen Wilson31db58b2011-03-13 15:49:15 -04002018extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002019extern int in_gate_area_no_mm(unsigned long addr);
2020extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021#else
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002022static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2023{
2024 return NULL;
2025}
2026static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2027static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2028{
2029 return 0;
2030}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031#endif /* __HAVE_ARCH_GATE_AREA */
2032
Josh Triplett146732c2013-04-29 15:07:22 -07002033#ifdef CONFIG_SYSCTL
2034extern int sysctl_drop_caches;
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07002035int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08002036 void __user *, size_t *, loff_t *);
Josh Triplett146732c2013-04-29 15:07:22 -07002037#endif
2038
Ying Hana09ed5e2011-05-24 17:12:26 -07002039unsigned long shrink_slab(struct shrink_control *shrink,
Ying Han1495f232011-05-24 17:12:27 -07002040 unsigned long nr_pages_scanned,
2041 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08002042
Luke Yang7a9166e2006-02-20 18:28:07 -08002043#ifndef CONFIG_MMU
2044#define randomize_va_space 0
2045#else
Andi Kleena62eaf12006-02-16 23:41:58 +01002046extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08002047#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01002048
Sam Ravnborg045e72a2007-07-26 10:41:13 -07002049const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01002050void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07002051
Yinghai Lu9bdac912010-02-10 01:20:22 -08002052void sparse_mem_maps_populate_node(struct page **map_map,
2053 unsigned long pnum_begin,
2054 unsigned long pnum_end,
2055 unsigned long map_count,
2056 int nodeid);
2057
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07002058struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07002059pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2060pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2061pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2062pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07002063void *vmemmap_alloc_block(unsigned long size, int node);
Yinghai Lu9bdac912010-02-10 01:20:22 -08002064void *vmemmap_alloc_block_buf(unsigned long size, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07002065void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Johannes Weiner0aad8182013-04-29 15:07:50 -07002066int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2067 int node);
2068int vmemmap_populate(unsigned long start, unsigned long end, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07002069void vmemmap_populate_print_last(void);
Tang Chen01975182013-02-22 16:33:08 -08002070#ifdef CONFIG_MEMORY_HOTPLUG
Johannes Weiner0aad8182013-04-29 15:07:50 -07002071void vmemmap_free(unsigned long start, unsigned long end);
Tang Chen01975182013-02-22 16:33:08 -08002072#endif
Yasuaki Ishimatsu46723bf2013-02-22 16:33:00 -08002073void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2074 unsigned long size);
Andi Kleen6a460792009-09-16 11:50:15 +02002075
Andi Kleen82ba0112009-12-16 12:19:57 +01002076enum mf_flags {
2077 MF_COUNT_INCREASED = 1 << 0,
Tony Luck7329bbe2011-12-13 09:27:58 -08002078 MF_ACTION_REQUIRED = 1 << 1,
Tony Luck6751ed62012-07-11 10:20:47 -07002079 MF_MUST_KILL = 1 << 2,
Naveen N. Raocf870c72013-07-10 14:57:01 +05302080 MF_SOFT_OFFLINE = 1 << 3,
Andi Kleen82ba0112009-12-16 12:19:57 +01002081};
Tony Luckcd42f4a2011-12-15 10:48:12 -08002082extern int memory_failure(unsigned long pfn, int trapno, int flags);
Huang Yingea8f5fb2011-07-13 13:14:27 +08002083extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
Wu Fengguang847ce402009-12-16 12:19:58 +01002084extern int unpoison_memory(unsigned long pfn);
Andi Kleen6a460792009-09-16 11:50:15 +02002085extern int sysctl_memory_failure_early_kill;
2086extern int sysctl_memory_failure_recovery;
Andi Kleenfacb6012009-12-16 12:20:00 +01002087extern void shake_page(struct page *p, int access);
Xishi Qiu293c07e2013-02-22 16:34:02 -08002088extern atomic_long_t num_poisoned_pages;
Andi Kleenfacb6012009-12-16 12:20:00 +01002089extern int soft_offline_page(struct page *page, int flags);
Andi Kleen6a460792009-09-16 11:50:15 +02002090
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08002091#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2092extern void clear_huge_page(struct page *page,
2093 unsigned long addr,
2094 unsigned int pages_per_huge_page);
2095extern void copy_user_huge_page(struct page *dst, struct page *src,
2096 unsigned long addr, struct vm_area_struct *vma,
2097 unsigned int pages_per_huge_page);
2098#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2099
Stanislaw Gruszkac0a32fc52012-01-10 15:07:28 -08002100#ifdef CONFIG_DEBUG_PAGEALLOC
2101extern unsigned int _debug_guardpage_minorder;
2102
2103static inline unsigned int debug_guardpage_minorder(void)
2104{
2105 return _debug_guardpage_minorder;
2106}
2107
2108static inline bool page_is_guard(struct page *page)
2109{
2110 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2111}
2112#else
2113static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2114static inline bool page_is_guard(struct page *page) { return false; }
2115#endif /* CONFIG_DEBUG_PAGEALLOC */
2116
Cody P Schaferf9872ca2013-04-29 15:08:01 -07002117#if MAX_NUMNODES > 1
2118void __init setup_nr_node_ids(void);
2119#else
2120static inline void setup_nr_node_ids(void) {}
2121#endif
2122
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123#endif /* __KERNEL__ */
2124#endif /* _LINUX_MM_H */