blob: 173d2484f6e38a63cea1857376a7c8d270b2ca68 [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001/* SPDX-License-Identifier: GPL-2.0 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002#ifndef _LINUX_MM_H
3#define _LINUX_MM_H
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/errno.h>
6
7#ifdef __KERNEL__
8
Sasha Levin309381fea2014-01-23 15:52:54 -08009#include <linux/mmdebug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/gfp.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050011#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/list.h>
13#include <linux/mmzone.h>
14#include <linux/rbtree.h>
Konstantin Khlebnikov83aeead2011-12-08 14:33:54 -080015#include <linux/atomic.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070016#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070017#include <linux/mm_types.h>
Yinghai Lu08677212010-02-10 01:20:20 -080018#include <linux/range.h>
Chris Metcalfc6f6b592010-05-24 14:32:53 -070019#include <linux/pfn.h>
Dan Williams3565fce2016-01-15 16:56:55 -080020#include <linux/percpu-refcount.h>
Andrea Arcangelie9da73d2011-01-13 15:46:32 -080021#include <linux/bit_spinlock.h>
Dave Chinnerb0d40c92011-07-08 14:14:42 +100022#include <linux/shrinker.h>
Cyrill Gorcunov9c599022014-10-09 15:27:29 -070023#include <linux/resource.h>
Joonsoo Kime30825f2014-12-12 16:55:49 -080024#include <linux/page_ext.h>
Jan Kara8025e5d2015-07-13 11:55:44 -030025#include <linux/err.h>
Joonsoo Kimfe896d12016-03-17 14:19:26 -070026#include <linux/page_ref.h>
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -070027#include <linux/memremap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028
29struct mempolicy;
30struct anon_vma;
Michel Lespinassebf181b92012-10-08 16:31:39 -070031struct anon_vma_chain;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040032struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040033struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040034struct writeback_control;
Tejun Heo682aa8e2015-05-28 14:50:53 -040035struct bdi_writeback;
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
Michal Hocko597b7302017-03-31 15:11:47 -070037void init_mm_internals(void);
38
Jiang Liufccc9982013-07-03 15:04:23 -070039#ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
Linus Torvalds1da177e2005-04-16 15:20:36 -070040extern unsigned long max_mapnr;
Jiang Liufccc9982013-07-03 15:04:23 -070041
42static inline void set_max_mapnr(unsigned long limit)
43{
44 max_mapnr = limit;
45}
46#else
47static inline void set_max_mapnr(unsigned long limit) { }
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#endif
49
Jan Beulich44813742009-09-21 17:03:05 -070050extern unsigned long totalram_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -070051extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070052extern int page_cluster;
53
54#ifdef CONFIG_SYSCTL
55extern int sysctl_legacy_va_layout;
56#else
57#define sysctl_legacy_va_layout 0
58#endif
59
Daniel Cashmand07e2252016-01-14 15:19:53 -080060#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61extern const int mmap_rnd_bits_min;
62extern const int mmap_rnd_bits_max;
63extern int mmap_rnd_bits __read_mostly;
64#endif
65#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66extern const int mmap_rnd_compat_bits_min;
67extern const int mmap_rnd_compat_bits_max;
68extern int mmap_rnd_compat_bits __read_mostly;
69#endif
70
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <asm/page.h>
72#include <asm/pgtable.h>
73#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Tang Chen79442ed2013-11-12 15:07:59 -080075#ifndef __pa_symbol
76#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
77#endif
78
Ard Biesheuvel1dff8082016-04-18 18:04:57 +020079#ifndef page_to_virt
80#define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
81#endif
82
Laura Abbott568c5fe2017-01-10 13:35:42 -080083#ifndef lm_alias
84#define lm_alias(x) __va(__pa_symbol(x))
85#endif
86
Dominik Dingel593befa2014-10-23 12:07:44 +020087/*
88 * To prevent common memory management code establishing
89 * a zero page mapping on a read fault.
90 * This macro should be defined within <asm/pgtable.h>.
91 * s390 does this to prevent multiplexing of hardware bits
92 * related to the physical page in case of virtualization.
93 */
94#ifndef mm_forbids_zeropage
95#define mm_forbids_zeropage(X) (0)
96#endif
97
Andrey Ryabininea606cf2016-03-17 14:18:48 -070098/*
Pavel Tatashina4a3ede2017-11-15 17:36:31 -080099 * On some architectures it is expensive to call memset() for small sizes.
100 * Those architectures should provide their own implementation of "struct page"
101 * zeroing by defining this macro in <asm/pgtable.h>.
102 */
103#ifndef mm_zero_struct_page
104#define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
105#endif
106
107/*
Andrey Ryabininea606cf2016-03-17 14:18:48 -0700108 * Default maximum number of active map areas, this limits the number of vmas
109 * per mm struct. Users can overwrite this number by sysctl but there is a
110 * problem.
111 *
112 * When a program's coredump is generated as ELF format, a section is created
113 * per a vma. In ELF, the number of sections is represented in unsigned short.
114 * This means the number of sections should be smaller than 65535 at coredump.
115 * Because the kernel adds some informative sections to a image of program at
116 * generating coredump, we need some margin. The number of extra sections is
117 * 1-3 now and depends on arch. We use "5" as safe margin, here.
118 *
119 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
120 * not a hard limit any more. Although some userspace tools can be surprised by
121 * that.
122 */
123#define MAPCOUNT_ELF_CORE_MARGIN (5)
124#define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
125
126extern int sysctl_max_map_count;
127
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -0700128extern unsigned long sysctl_user_reserve_kbytes;
Andrew Shewmaker4eeab4f2013-04-29 15:08:11 -0700129extern unsigned long sysctl_admin_reserve_kbytes;
Andrew Shewmakerc9b1d092013-04-29 15:08:10 -0700130
Jerome Marchand49f0ce52014-01-21 15:49:14 -0800131extern int sysctl_overcommit_memory;
132extern int sysctl_overcommit_ratio;
133extern unsigned long sysctl_overcommit_kbytes;
134
135extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
136 size_t *, loff_t *);
137extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
138 size_t *, loff_t *);
139
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
141
Andrea Righi27ac7922008-07-23 21:28:13 -0700142/* to align the pointer to the (next) page boundary */
143#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
144
Andrew Morton0fa73b82013-07-03 15:02:11 -0700145/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
zijun_hu1061b0d22016-10-07 17:02:04 -0700146#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
Andrew Morton0fa73b82013-07-03 15:02:11 -0700147
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148/*
149 * Linux kernel virtual memory manager primitives.
150 * The idea being to have a "virtual" mm in the same way
151 * we have a virtual fs - giving a cleaner interface to the
152 * mm details, and allowing different kinds of memory mappings
153 * (from shared memory to executable loading to arbitrary
154 * mmap() functions).
155 */
156
Christoph Lameterc43692e2006-12-06 20:32:48 -0800157extern struct kmem_cache *vm_area_cachep;
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +0000160extern struct rb_root nommu_region_tree;
161extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162
163extern unsigned int kobjsize(const void *objp);
164#endif
165
166/*
Hugh Dickins605d9282008-08-16 11:07:21 +0100167 * vm_flags in vm_area_struct, see mm_types.h.
Kirill A. Shutemovbcf66912016-03-17 14:18:53 -0700168 * When changing, update also include/trace/events/mmflags.h
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700170#define VM_NONE 0x00000000
171
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172#define VM_READ 0x00000001 /* currently active flags */
173#define VM_WRITE 0x00000002
174#define VM_EXEC 0x00000004
175#define VM_SHARED 0x00000008
176
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -0700177/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
179#define VM_MAYWRITE 0x00000020
180#define VM_MAYEXEC 0x00000040
181#define VM_MAYSHARE 0x00000080
182
183#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
Andrea Arcangeli16ba6f82015-09-04 15:46:17 -0700184#define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
Linus Torvalds6aab3412005-11-28 14:34:23 -0800185#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
Andrea Arcangeli16ba6f82015-09-04 15:46:17 -0700187#define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189#define VM_LOCKED 0x00002000
190#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
191
192 /* Used by sys_madvise() */
193#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
194#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
195
196#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
197#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Eric B Munsonde60f5f2015-11-05 18:51:36 -0800198#define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -0700200#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
Jan Karab6fb2932017-11-01 16:36:41 +0100202#define VM_SYNC 0x00800000 /* Synchronous page faults */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700203#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
Rik van Rield2cd9ed2017-09-06 16:25:15 -0700204#define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
Konstantin Khlebnikov0103bd12012-10-08 16:28:59 -0700205#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700207#ifdef CONFIG_MEM_SOFT_DIRTY
208# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
209#else
210# define VM_SOFTDIRTY 0
211#endif
212
Jared Hulbertb379d792008-04-28 02:12:58 -0700213#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700214#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
215#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700216#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700217
Dave Hansen63c17fb2016-02-12 13:02:08 -0800218#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
219#define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
220#define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
221#define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
222#define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
Rik van Rieldf3735c2017-09-06 16:25:11 -0700223#define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
Dave Hansen63c17fb2016-02-12 13:02:08 -0800224#define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
225#define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
226#define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
227#define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
Rik van Rieldf3735c2017-09-06 16:25:11 -0700228#define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
Dave Hansen63c17fb2016-02-12 13:02:08 -0800229#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
230
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700231#if defined(CONFIG_X86)
232# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
Dave Hansen8f62c882016-02-12 13:02:10 -0800233#if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
234# define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
235# define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
236# define VM_PKEY_BIT1 VM_HIGH_ARCH_1
237# define VM_PKEY_BIT2 VM_HIGH_ARCH_2
238# define VM_PKEY_BIT3 VM_HIGH_ARCH_3
239#endif
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700240#elif defined(CONFIG_PPC)
241# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
242#elif defined(CONFIG_PARISC)
243# define VM_GROWSUP VM_ARCH_1
James Hogan9ca52ed2012-10-16 10:16:14 +0100244#elif defined(CONFIG_METAG)
245# define VM_GROWSUP VM_ARCH_1
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700246#elif defined(CONFIG_IA64)
247# define VM_GROWSUP VM_ARCH_1
248#elif !defined(CONFIG_MMU)
249# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
250#endif
251
Rik van Rieldf3735c2017-09-06 16:25:11 -0700252#if defined(CONFIG_X86_INTEL_MPX)
Qiaowei Ren4aae7e42014-11-14 07:18:25 -0800253/* MPX specific bounds table or bounds directory */
Kirill A. Shutemovfa87b912017-10-03 16:14:24 -0700254# define VM_MPX VM_HIGH_ARCH_4
Rik van Rieldf3735c2017-09-06 16:25:11 -0700255#else
256# define VM_MPX VM_NONE
Qiaowei Ren4aae7e42014-11-14 07:18:25 -0800257#endif
258
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700259#ifndef VM_GROWSUP
260# define VM_GROWSUP VM_NONE
261#endif
262
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700263/* Bits set in the VMA until the stack is in its final location */
264#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
265
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
267#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
268#endif
269
270#ifdef CONFIG_STACK_GROWSUP
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800271#define VM_STACK VM_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272#else
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800273#define VM_STACK VM_GROWSDOWN
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274#endif
275
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800276#define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
277
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278/*
Andrea Arcangeli78f11a22011-04-27 15:26:45 -0700279 * Special vmas that are non-mergable, non-mlock()able.
280 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
Nick Pigginb291f002008-10-18 20:26:44 -0700281 */
Vlastimil Babka9050d7e2014-03-03 15:38:27 -0800282#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
Nick Pigginb291f002008-10-18 20:26:44 -0700283
Alex Thorltona0715cc2014-04-07 15:37:10 -0700284/* This mask defines which mm->def_flags a process can inherit its parent */
285#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
286
Eric B Munsonde60f5f2015-11-05 18:51:36 -0800287/* This mask is used to clear all the VMA flags used by mlock */
288#define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
289
Nick Pigginb291f002008-10-18 20:26:44 -0700290/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 * mapping from the currently active vm_flags protection bits (the
292 * low four bits) to a page protection mask..
293 */
294extern pgprot_t protection_map[16];
295
Nick Piggind0217ac2007-07-19 01:47:03 -0700296#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
Kirill A. Shutemov9b4bdd22015-02-10 14:09:51 -0800297#define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
298#define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
299#define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
300#define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
301#define FAULT_FLAG_TRIED 0x20 /* Second try */
302#define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
Dave Hansen1b2ee122016-02-12 13:02:21 -0800303#define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
Dave Hansend61172b2016-02-12 13:02:24 -0800304#define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
Nick Piggind0217ac2007-07-19 01:47:03 -0700305
Ross Zwisler282a8e02017-02-22 15:39:50 -0800306#define FAULT_FLAG_TRACE \
307 { FAULT_FLAG_WRITE, "WRITE" }, \
308 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
309 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
310 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
311 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
312 { FAULT_FLAG_TRIED, "TRIED" }, \
313 { FAULT_FLAG_USER, "USER" }, \
314 { FAULT_FLAG_REMOTE, "REMOTE" }, \
315 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
316
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800317/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700318 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700319 * ->fault function. The vma's ->fault is responsible for returning a bitmask
320 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700321 *
Michal Hockoc20cd452016-01-14 15:20:12 -0800322 * MM layer fills up gfp_mask for page allocations but fault handler might
323 * alter it if its implementation requires a different allocation context.
324 *
Kirill A. Shutemov9b4bdd22015-02-10 14:09:51 -0800325 * pgoff should be used in favour of virtual_address, if possible.
Nick Piggin54cb8822007-07-19 01:46:59 -0700326 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700327struct vm_fault {
Jan Kara82b0f8c2016-12-14 15:06:58 -0800328 struct vm_area_struct *vma; /* Target VMA */
Nick Piggind0217ac2007-07-19 01:47:03 -0700329 unsigned int flags; /* FAULT_FLAG_xxx flags */
Michal Hockoc20cd452016-01-14 15:20:12 -0800330 gfp_t gfp_mask; /* gfp mask to be used for allocations */
Nick Piggind0217ac2007-07-19 01:47:03 -0700331 pgoff_t pgoff; /* Logical page offset based on vma */
Jan Kara82b0f8c2016-12-14 15:06:58 -0800332 unsigned long address; /* Faulting virtual address */
Jan Kara82b0f8c2016-12-14 15:06:58 -0800333 pmd_t *pmd; /* Pointer to pmd entry matching
Jan Kara29943022016-12-14 15:07:16 -0800334 * the 'address' */
Dave Jianga2d58162017-02-24 14:56:59 -0800335 pud_t *pud; /* Pointer to pud entry matching
336 * the 'address'
337 */
Jan Kara29943022016-12-14 15:07:16 -0800338 pte_t orig_pte; /* Value of PTE at the time of fault */
Nick Piggind0217ac2007-07-19 01:47:03 -0700339
Jan Kara39170482016-12-14 15:07:18 -0800340 struct page *cow_page; /* Page handler may use for COW fault */
341 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
Nick Piggind0217ac2007-07-19 01:47:03 -0700342 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700343 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700344 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700345 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700346 */
Jan Kara82b0f8c2016-12-14 15:06:58 -0800347 /* These three entries are valid only while holding ptl lock */
Kirill A. Shutemovbae473a2016-07-26 15:25:20 -0700348 pte_t *pte; /* Pointer to pte entry matching
349 * the 'address'. NULL if the page
350 * table hasn't been allocated.
351 */
352 spinlock_t *ptl; /* Page table lock.
353 * Protects pte page table if 'pte'
354 * is not NULL, otherwise pmd.
355 */
Kirill A. Shutemov7267ec002016-07-26 15:25:23 -0700356 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
357 * vm_ops->map_pages() calls
358 * alloc_set_pte() from atomic context.
359 * do_fault_around() pre-allocates
360 * page table to avoid allocation from
361 * atomic context.
362 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700363};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
Dave Jiangc791ace2017-02-24 14:57:08 -0800365/* page entry size for vm->huge_fault() */
366enum page_entry_size {
367 PE_SIZE_PTE = 0,
368 PE_SIZE_PMD,
369 PE_SIZE_PUD,
370};
371
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372/*
373 * These are the virtual MM functions - opening of an area, closing and
374 * unmapping it (needed to keep files on disk up-to-date etc), pointer
375 * to the functions called when a no-page or a wp-page exception occurs.
376 */
377struct vm_operations_struct {
378 void (*open)(struct vm_area_struct * area);
379 void (*close)(struct vm_area_struct * area);
Dan Williams31383c62017-11-29 16:10:28 -0800380 int (*split)(struct vm_area_struct * area, unsigned long addr);
Oleg Nesterov5477e702015-09-04 15:48:04 -0700381 int (*mremap)(struct vm_area_struct * area);
Dave Jiang11bac802017-02-24 14:56:41 -0800382 int (*fault)(struct vm_fault *vmf);
Dave Jiangc791ace2017-02-24 14:57:08 -0800383 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
Jan Kara82b0f8c2016-12-14 15:06:58 -0800384 void (*map_pages)(struct vm_fault *vmf,
Kirill A. Shutemovbae473a2016-07-26 15:25:20 -0700385 pgoff_t start_pgoff, pgoff_t end_pgoff);
David Howells9637a5e2006-06-23 02:03:43 -0700386
387 /* notification that a previously read-only page is about to become
388 * writable, if an error is returned it will cause a SIGBUS */
Dave Jiang11bac802017-02-24 14:56:41 -0800389 int (*page_mkwrite)(struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700390
Boaz Harroshdd906182015-04-15 16:15:11 -0700391 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
Dave Jiang11bac802017-02-24 14:56:41 -0800392 int (*pfn_mkwrite)(struct vm_fault *vmf);
Boaz Harroshdd906182015-04-15 16:15:11 -0700393
Rik van Riel28b2ee22008-07-23 21:27:05 -0700394 /* called by access_process_vm when get_user_pages() fails, typically
395 * for use by special VMAs that can switch between memory and hardware
396 */
397 int (*access)(struct vm_area_struct *vma, unsigned long addr,
398 void *buf, int len, int write);
Andy Lutomirski78d683e2014-05-19 15:58:32 -0700399
400 /* Called by the /proc/PID/maps code to ask the vma whether it
401 * has a special name. Returning non-NULL will also cause this
402 * vma to be dumped unconditionally. */
403 const char *(*name)(struct vm_area_struct *vma);
404
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700406 /*
407 * set_policy() op must add a reference to any non-NULL @new mempolicy
408 * to hold the policy upon return. Caller should pass NULL @new to
409 * remove a policy and fall back to surrounding context--i.e. do not
410 * install a MPOL_DEFAULT policy, nor the task or system default
411 * mempolicy.
412 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700414
415 /*
416 * get_policy() op must add reference [mpol_get()] to any policy at
417 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
418 * in mm/mempolicy.c will do this automatically.
419 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
420 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
421 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
422 * must return NULL--i.e., do not "fallback" to task or system default
423 * policy.
424 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
426 unsigned long addr);
427#endif
David Vrabel667a0a02014-12-18 14:48:15 +0000428 /*
429 * Called by vm_normal_page() for special PTEs to find the
430 * page for @addr. This is useful if the default behavior
431 * (using pte_page()) would not find the correct page.
432 */
433 struct page *(*find_special_page)(struct vm_area_struct *vma,
434 unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435};
436
437struct mmu_gather;
438struct inode;
439
Andrew Morton349aef02006-01-08 01:04:36 -0800440#define page_private(page) ((page)->private)
441#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700442
Dan Williams5c7fb562016-01-15 16:56:52 -0800443#if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
444static inline int pmd_devmap(pmd_t pmd)
445{
446 return 0;
447}
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -0800448static inline int pud_devmap(pud_t pud)
449{
450 return 0;
451}
Kirill A. Shutemovb59f65f2017-03-16 18:26:53 +0300452static inline int pgd_devmap(pgd_t pgd)
453{
454 return 0;
455}
Dan Williams5c7fb562016-01-15 16:56:52 -0800456#endif
457
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458/*
459 * FIXME: take this include out, include page-flags.h in
460 * files which need it (119 of them)
461 */
462#include <linux/page-flags.h>
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800463#include <linux/huge_mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
465/*
466 * Methods to modify the page usage count.
467 *
468 * What counts for a page usage:
469 * - cache mapping (page->mapping)
470 * - private data (page->private)
471 * - page mapped in a task's page tables, each mapping
472 * is counted separately
473 *
474 * Also, many kernel routines increase the page count before a critical
475 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 */
477
478/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700479 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800481static inline int put_page_testzero(struct page *page)
482{
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700483 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
484 return page_ref_dec_and_test(page);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800485}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
487/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800488 * Try to grab a ref unless the page has a refcount of zero, return false if
489 * that is the case.
Alexey Kardashevskiy8e0861fa2013-08-28 18:37:42 +1000490 * This can be called when MMU is off so it must not access
491 * any of the virtual mappings.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800493static inline int get_page_unless_zero(struct page *page)
494{
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700495 return page_ref_add_unless(page, 1, 0);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800496}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800498extern int page_is_ram(unsigned long pfn);
Dan Williams124fe202015-08-10 23:07:05 -0400499
500enum {
501 REGION_INTERSECTS,
502 REGION_DISJOINT,
503 REGION_MIXED,
504};
505
Toshi Kani1c29f252016-01-26 21:57:28 +0100506int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
507 unsigned long desc);
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800508
Christoph Lameter48667e72008-02-04 22:28:31 -0800509/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800510struct page *vmalloc_to_page(const void *addr);
511unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800512
Paul Mundt0738c4b2008-03-12 16:51:31 +0900513/*
514 * Determine if an address is within the vmalloc range
515 *
516 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
517 * is no special casing required.
518 */
Yaowei Baibb00a782016-05-19 17:11:29 -0700519static inline bool is_vmalloc_addr(const void *x)
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800520{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900521#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800522 unsigned long addr = (unsigned long)x;
523
524 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900525#else
Yaowei Baibb00a782016-05-19 17:11:29 -0700526 return false;
David Howells8ca3ed82008-02-23 15:23:37 -0800527#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900528}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700529#ifdef CONFIG_MMU
530extern int is_vmalloc_or_module_addr(const void *x);
531#else
David Howells934831d2009-09-24 12:33:32 +0100532static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700533{
534 return 0;
535}
536#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800537
Michal Hockoa7c3e902017-05-08 15:57:09 -0700538extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
539static inline void *kvmalloc(size_t size, gfp_t flags)
540{
541 return kvmalloc_node(size, flags, NUMA_NO_NODE);
542}
543static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
544{
545 return kvmalloc_node(size, flags | __GFP_ZERO, node);
546}
547static inline void *kvzalloc(size_t size, gfp_t flags)
548{
549 return kvmalloc(size, flags | __GFP_ZERO);
550}
551
Michal Hocko752ade62017-05-08 15:57:27 -0700552static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
553{
554 if (size != 0 && n > SIZE_MAX / size)
555 return NULL;
556
557 return kvmalloc(n * size, flags);
558}
559
Al Viro39f1f782014-05-06 14:02:53 -0400560extern void kvfree(const void *addr);
561
Kirill A. Shutemov53f92632016-01-15 16:53:42 -0800562static inline atomic_t *compound_mapcount_ptr(struct page *page)
563{
564 return &page[1].compound_mapcount;
565}
566
567static inline int compound_mapcount(struct page *page)
568{
Andrea Arcangeli5f527c22016-05-20 16:58:24 -0700569 VM_BUG_ON_PAGE(!PageCompound(page), page);
Kirill A. Shutemov53f92632016-01-15 16:53:42 -0800570 page = compound_head(page);
571 return atomic_read(compound_mapcount_ptr(page)) + 1;
572}
573
Joonsoo Kimccaafd72015-02-10 14:09:35 -0800574/*
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700575 * The atomic page->_mapcount, starts from -1: so that transitions
576 * both from it and to it can be tracked, using atomic_inc_and_test
577 * and atomic_add_negative(-1).
578 */
Mel Gorman22b751c2013-02-22 16:34:59 -0800579static inline void page_mapcount_reset(struct page *page)
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700580{
581 atomic_set(&(page)->_mapcount, -1);
582}
583
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800584int __page_mapcount(struct page *page);
585
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700586static inline int page_mapcount(struct page *page)
587{
Wang, Yalin1d148e22015-02-11 15:24:48 -0800588 VM_BUG_ON_PAGE(PageSlab(page), page);
Kirill A. Shutemov53f92632016-01-15 16:53:42 -0800589
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800590 if (unlikely(PageCompound(page)))
591 return __page_mapcount(page);
592 return atomic_read(&page->_mapcount) + 1;
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700593}
594
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800595#ifdef CONFIG_TRANSPARENT_HUGEPAGE
596int total_mapcount(struct page *page);
Andrea Arcangeli6d0a07e2016-05-12 15:42:25 -0700597int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800598#else
599static inline int total_mapcount(struct page *page)
600{
601 return page_mapcount(page);
602}
Andrea Arcangeli6d0a07e2016-05-12 15:42:25 -0700603static inline int page_trans_huge_mapcount(struct page *page,
604 int *total_mapcount)
605{
606 int mapcount = page_mapcount(page);
607 if (total_mapcount)
608 *total_mapcount = mapcount;
609 return mapcount;
610}
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800611#endif
612
Christoph Lameterb49af682007-05-06 14:49:41 -0700613static inline struct page *virt_to_head_page(const void *x)
614{
615 struct page *page = virt_to_page(x);
Joonsoo Kimccaafd72015-02-10 14:09:35 -0800616
Kirill A. Shutemov1d798ca2015-11-06 16:29:54 -0800617 return compound_head(page);
Christoph Lameterb49af682007-05-06 14:49:41 -0700618}
619
Kirill A. Shutemovddc58f22016-01-15 16:52:56 -0800620void __put_page(struct page *page);
621
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700622void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800624void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800625
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800627 * Compound pages have a destructor function. Provide a
628 * prototype for that function and accessor functions.
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800629 * These are _only_ valid on the head of a compound page.
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800630 */
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800631typedef void compound_page_dtor(struct page *);
632
633/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
634enum compound_dtor_id {
635 NULL_COMPOUND_DTOR,
636 COMPOUND_PAGE_DTOR,
637#ifdef CONFIG_HUGETLB_PAGE
638 HUGETLB_PAGE_DTOR,
639#endif
Kirill A. Shutemov9a982252016-01-15 16:54:17 -0800640#ifdef CONFIG_TRANSPARENT_HUGEPAGE
641 TRANSHUGE_PAGE_DTOR,
642#endif
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800643 NR_COMPOUND_DTORS,
644};
645extern compound_page_dtor * const compound_page_dtors[];
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800646
647static inline void set_compound_page_dtor(struct page *page,
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800648 enum compound_dtor_id compound_dtor)
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800649{
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800650 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
651 page[1].compound_dtor = compound_dtor;
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800652}
653
654static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
655{
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800656 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
657 return compound_page_dtors[page[1].compound_dtor];
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800658}
659
Kirill A. Shutemovd00181b2015-11-06 16:29:57 -0800660static inline unsigned int compound_order(struct page *page)
Christoph Lameterd85f3382007-05-06 14:49:39 -0700661{
Christoph Lameter6d777952007-05-06 14:49:40 -0700662 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700663 return 0;
Kirill A. Shutemove4b294c2015-02-11 15:24:46 -0800664 return page[1].compound_order;
Christoph Lameterd85f3382007-05-06 14:49:39 -0700665}
666
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800667static inline void set_compound_order(struct page *page, unsigned int order)
Christoph Lameterd85f3382007-05-06 14:49:39 -0700668{
Kirill A. Shutemove4b294c2015-02-11 15:24:46 -0800669 page[1].compound_order = order;
Christoph Lameterd85f3382007-05-06 14:49:39 -0700670}
671
Kirill A. Shutemov9a982252016-01-15 16:54:17 -0800672void free_compound_page(struct page *page);
673
Michal Simek3dece372011-01-21 08:49:56 +0100674#ifdef CONFIG_MMU
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800675/*
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800676 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
677 * servicing faults for write access. In the normal case, do always want
678 * pte_mkwrite. But get_user_pages can cause write faults for mappings
679 * that do not have writing enabled, when used by access_process_vm.
680 */
681static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
682{
683 if (likely(vma->vm_flags & VM_WRITE))
684 pte = pte_mkwrite(pte);
685 return pte;
686}
Kirill A. Shutemov8c6e50b2014-04-07 15:37:18 -0700687
Jan Kara82b0f8c2016-12-14 15:06:58 -0800688int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
Kirill A. Shutemov7267ec002016-07-26 15:25:23 -0700689 struct page *page);
Jan Kara9118c0c2016-12-14 15:07:21 -0800690int finish_fault(struct vm_fault *vmf);
Jan Kara66a61972016-12-14 15:07:39 -0800691int finish_mkwrite_fault(struct vm_fault *vmf);
Michal Simek3dece372011-01-21 08:49:56 +0100692#endif
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800693
694/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 * Multiple processes may "see" the same page. E.g. for untouched
696 * mappings of /dev/null, all processes see the same page full of
697 * zeroes, and text pages of executables and shared libraries have
698 * only one copy in memory, at most, normally.
699 *
700 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700701 * page_count() == 0 means the page is free. page->lru is then used for
702 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700703 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700705 * Pages are allocated by the slab allocator in order to provide memory
706 * to kmalloc and kmem_cache_alloc. In this case, the management of the
707 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
708 * unless a particular usage is carefully commented. (the responsibility of
709 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700711 * A page may be used by anyone else who does a __get_free_page().
712 * In this case, page_count still tracks the references, and should only
713 * be used through the normal accessor functions. The top bits of page->flags
714 * and page->virtual store page management information, but all other fields
715 * are unused and could be used privately, carefully. The management of this
716 * page is the responsibility of the one who allocated it, and those who have
717 * subsequently been given references to it.
718 *
719 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 * managed by the Linux memory manager: I/O, buffers, swapping etc.
721 * The following discussion applies only to them.
722 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700723 * A pagecache page contains an opaque `private' member, which belongs to the
724 * page's address_space. Usually, this is the address of a circular list of
725 * the page's disk buffers. PG_private must be set to tell the VM to call
726 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700728 * A page may belong to an inode's memory mapping. In this case, page->mapping
729 * is the pointer to the inode, and page->index is the file offset of the page,
Kirill A. Shutemovea1754a2016-04-01 15:29:48 +0300730 * in units of PAGE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700732 * If pagecache pages are not associated with an inode, they are said to be
733 * anonymous pages. These may become associated with the swapcache, and in that
734 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700736 * In either case (swapcache or inode backed), the pagecache itself holds one
737 * reference to the page. Setting PG_private should also increment the
738 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700740 * The pagecache pages are stored in a per-mapping radix tree, which is
741 * rooted at mapping->page_tree, and indexed by offset.
742 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
743 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700745 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 * - inode pages may need to be read from disk,
747 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700748 * to be written back to the inode on disk,
749 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
750 * modified may need to be swapped out to swap space and (later) to be read
751 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 */
753
754/*
755 * The zone field is never updated after free_area_init_core()
756 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700758
Peter Zijlstra90572892013-10-07 11:29:20 +0100759/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100760#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700761#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
762#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
Peter Zijlstra90572892013-10-07 11:29:20 +0100763#define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700764
765/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300766 * Define the bit shifts to access each section. For non-existent
Dave Hansen348f8b62005-06-23 00:07:40 -0700767 * sections we define the shift as 0; that plus a 0 mask ensures
768 * the compiler will optimise away reference to them.
769 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700770#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
771#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
772#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Peter Zijlstra90572892013-10-07 11:29:20 +0100773#define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700774
Will Deaconbce54bb2010-10-26 14:21:37 -0700775/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
776#ifdef NODE_NOT_IN_PAGE_FLAGS
Christoph Lameter89689ae2006-12-06 20:31:45 -0800777#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800778#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
779 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700780#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800781#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800782#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
783 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700784#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800785
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800786#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700787
Christoph Lameter9223b4192008-04-28 02:12:48 -0700788#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
789#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700790#endif
791
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700792#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
793#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
794#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Srikar Dronamraju834a9642014-04-07 15:37:57 -0700795#define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800796#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700797
Ian Campbell33dd4e02011-07-25 17:11:51 -0700798static inline enum zone_type page_zonenum(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799{
Dave Hansen348f8b62005-06-23 00:07:40 -0700800 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
Dan Williams260ae3f2016-01-15 16:56:17 -0800803#ifdef CONFIG_ZONE_DEVICE
804static inline bool is_zone_device_page(const struct page *page)
805{
806 return page_zonenum(page) == ZONE_DEVICE;
807}
808#else
809static inline bool is_zone_device_page(const struct page *page)
810{
811 return false;
812}
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700813#endif
Jérôme Glisse5042db42017-09-08 16:11:43 -0700814
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700815#if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700816void put_zone_device_private_or_public_page(struct page *page);
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700817DECLARE_STATIC_KEY_FALSE(device_private_key);
818#define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
819static inline bool is_device_private_page(const struct page *page);
820static inline bool is_device_public_page(const struct page *page);
821#else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700822static inline void put_zone_device_private_or_public_page(struct page *page)
Jérôme Glisse5042db42017-09-08 16:11:43 -0700823{
Jérôme Glisse5042db42017-09-08 16:11:43 -0700824}
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700825#define IS_HMM_ENABLED 0
826static inline bool is_device_private_page(const struct page *page)
827{
828 return false;
829}
830static inline bool is_device_public_page(const struct page *page)
831{
832 return false;
833}
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700834#endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
Dan Williams260ae3f2016-01-15 16:56:17 -0800835
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700836
Dan Williams3565fce2016-01-15 16:56:55 -0800837static inline void get_page(struct page *page)
838{
839 page = compound_head(page);
840 /*
841 * Getting a normal page or the head of a compound page
Joonsoo Kim0139aa72016-05-19 17:10:49 -0700842 * requires to already have an elevated page->_refcount.
Dan Williams3565fce2016-01-15 16:56:55 -0800843 */
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700844 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
845 page_ref_inc(page);
Dan Williams3565fce2016-01-15 16:56:55 -0800846}
847
848static inline void put_page(struct page *page)
849{
850 page = compound_head(page);
851
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700852 /*
853 * For private device pages we need to catch refcount transition from
854 * 2 to 1, when refcount reach one it means the private device page is
855 * free and we need to inform the device driver through callback. See
856 * include/linux/memremap.h and HMM for details.
857 */
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700858 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
859 unlikely(is_device_public_page(page)))) {
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700860 put_zone_device_private_or_public_page(page);
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700861 return;
862 }
863
Dan Williams3565fce2016-01-15 16:56:55 -0800864 if (put_page_testzero(page))
865 __put_page(page);
Dan Williams3565fce2016-01-15 16:56:55 -0800866}
867
Cody P Schafer9127ab42013-02-22 16:35:21 -0800868#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
869#define SECTION_IN_PAGE_FLAGS
870#endif
871
Christoph Lameter89689ae2006-12-06 20:31:45 -0800872/*
Vlastimil Babka7a8010c2013-09-11 14:22:35 -0700873 * The identification function is mainly used by the buddy allocator for
874 * determining if two pages could be buddies. We are not really identifying
875 * the zone since we could be using the section number id if we do not have
876 * node id available in page flags.
877 * We only guarantee that it will return the same value for two combinable
878 * pages in a zone.
Christoph Lameter89689ae2006-12-06 20:31:45 -0800879 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700880static inline int page_zone_id(struct page *page)
881{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800882 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883}
884
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800885static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700886{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700887#ifdef CONFIG_NUMA
888 return zone->node;
889#else
890 return 0;
891#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700892}
893
Christoph Lameter89689ae2006-12-06 20:31:45 -0800894#ifdef NODE_NOT_IN_PAGE_FLAGS
Ian Campbell33dd4e02011-07-25 17:11:51 -0700895extern int page_to_nid(const struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800896#else
Ian Campbell33dd4e02011-07-25 17:11:51 -0700897static inline int page_to_nid(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700898{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800899 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700900}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800901#endif
902
Mel Gorman57e0a032012-11-12 09:06:20 +0000903#ifdef CONFIG_NUMA_BALANCING
Peter Zijlstra90572892013-10-07 11:29:20 +0100904static inline int cpu_pid_to_cpupid(int cpu, int pid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000905{
Peter Zijlstra90572892013-10-07 11:29:20 +0100906 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
Mel Gorman57e0a032012-11-12 09:06:20 +0000907}
908
Peter Zijlstra90572892013-10-07 11:29:20 +0100909static inline int cpupid_to_pid(int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000910{
Peter Zijlstra90572892013-10-07 11:29:20 +0100911 return cpupid & LAST__PID_MASK;
Mel Gorman57e0a032012-11-12 09:06:20 +0000912}
Mel Gormanb7958542013-10-07 11:29:07 +0100913
Peter Zijlstra90572892013-10-07 11:29:20 +0100914static inline int cpupid_to_cpu(int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000915{
Peter Zijlstra90572892013-10-07 11:29:20 +0100916 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
Mel Gormanb7958542013-10-07 11:29:07 +0100917}
918
Peter Zijlstra90572892013-10-07 11:29:20 +0100919static inline int cpupid_to_nid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100920{
Peter Zijlstra90572892013-10-07 11:29:20 +0100921 return cpu_to_node(cpupid_to_cpu(cpupid));
Mel Gormanb7958542013-10-07 11:29:07 +0100922}
923
Peter Zijlstra90572892013-10-07 11:29:20 +0100924static inline bool cpupid_pid_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100925{
Peter Zijlstra90572892013-10-07 11:29:20 +0100926 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100927}
928
Peter Zijlstra90572892013-10-07 11:29:20 +0100929static inline bool cpupid_cpu_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100930{
Peter Zijlstra90572892013-10-07 11:29:20 +0100931 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100932}
933
Peter Zijlstra8c8a7432013-10-07 11:29:21 +0100934static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
935{
936 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
937}
938
939#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
Peter Zijlstra90572892013-10-07 11:29:20 +0100940#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
941static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100942{
Liu Ping Fan1ae71d02014-03-03 15:38:39 -0800943 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100944}
Peter Zijlstra90572892013-10-07 11:29:20 +0100945
946static inline int page_cpupid_last(struct page *page)
Mel Gormanb7958542013-10-07 11:29:07 +0100947{
Peter Zijlstra90572892013-10-07 11:29:20 +0100948 return page->_last_cpupid;
949}
950static inline void page_cpupid_reset_last(struct page *page)
951{
Liu Ping Fan1ae71d02014-03-03 15:38:39 -0800952 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
Mel Gorman57e0a032012-11-12 09:06:20 +0000953}
954#else
Peter Zijlstra90572892013-10-07 11:29:20 +0100955static inline int page_cpupid_last(struct page *page)
Peter Zijlstra75980e92013-02-22 16:34:32 -0800956{
Peter Zijlstra90572892013-10-07 11:29:20 +0100957 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
Peter Zijlstra75980e92013-02-22 16:34:32 -0800958}
959
Peter Zijlstra90572892013-10-07 11:29:20 +0100960extern int page_cpupid_xchg_last(struct page *page, int cpupid);
Peter Zijlstra75980e92013-02-22 16:34:32 -0800961
Peter Zijlstra90572892013-10-07 11:29:20 +0100962static inline void page_cpupid_reset_last(struct page *page)
Peter Zijlstra75980e92013-02-22 16:34:32 -0800963{
Mel Gorman09940a42016-05-19 17:13:53 -0700964 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
Peter Zijlstra75980e92013-02-22 16:34:32 -0800965}
Peter Zijlstra90572892013-10-07 11:29:20 +0100966#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
967#else /* !CONFIG_NUMA_BALANCING */
968static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000969{
Peter Zijlstra90572892013-10-07 11:29:20 +0100970 return page_to_nid(page); /* XXX */
Mel Gorman57e0a032012-11-12 09:06:20 +0000971}
972
Peter Zijlstra90572892013-10-07 11:29:20 +0100973static inline int page_cpupid_last(struct page *page)
Mel Gorman57e0a032012-11-12 09:06:20 +0000974{
Peter Zijlstra90572892013-10-07 11:29:20 +0100975 return page_to_nid(page); /* XXX */
Mel Gorman57e0a032012-11-12 09:06:20 +0000976}
977
Peter Zijlstra90572892013-10-07 11:29:20 +0100978static inline int cpupid_to_nid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100979{
980 return -1;
981}
982
Peter Zijlstra90572892013-10-07 11:29:20 +0100983static inline int cpupid_to_pid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100984{
985 return -1;
986}
987
Peter Zijlstra90572892013-10-07 11:29:20 +0100988static inline int cpupid_to_cpu(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100989{
990 return -1;
991}
992
Peter Zijlstra90572892013-10-07 11:29:20 +0100993static inline int cpu_pid_to_cpupid(int nid, int pid)
994{
995 return -1;
996}
997
998static inline bool cpupid_pid_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100999{
1000 return 1;
1001}
1002
Peter Zijlstra90572892013-10-07 11:29:20 +01001003static inline void page_cpupid_reset_last(struct page *page)
Mel Gorman57e0a032012-11-12 09:06:20 +00001004{
1005}
Peter Zijlstra8c8a7432013-10-07 11:29:21 +01001006
1007static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1008{
1009 return false;
1010}
Peter Zijlstra90572892013-10-07 11:29:20 +01001011#endif /* CONFIG_NUMA_BALANCING */
Mel Gorman57e0a032012-11-12 09:06:20 +00001012
Ian Campbell33dd4e02011-07-25 17:11:51 -07001013static inline struct zone *page_zone(const struct page *page)
Christoph Lameter89689ae2006-12-06 20:31:45 -08001014{
1015 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1016}
1017
Mel Gorman75ef7182016-07-28 15:45:24 -07001018static inline pg_data_t *page_pgdat(const struct page *page)
1019{
1020 return NODE_DATA(page_to_nid(page));
1021}
1022
Cody P Schafer9127ab42013-02-22 16:35:21 -08001023#ifdef SECTION_IN_PAGE_FLAGS
Daniel Kiperbf4e8902011-05-24 17:12:32 -07001024static inline void set_page_section(struct page *page, unsigned long section)
1025{
1026 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1027 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1028}
1029
Ian Campbellaa462ab2011-08-17 17:40:33 +01001030static inline unsigned long page_to_section(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001031{
1032 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1033}
Christoph Lameter308c05e2008-04-28 02:12:43 -07001034#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001035
Christoph Lameter2f1b6242006-09-25 23:31:13 -07001036static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037{
Dave Hansen348f8b62005-06-23 00:07:40 -07001038 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1039 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1040}
Christoph Lameter2f1b6242006-09-25 23:31:13 -07001041
Dave Hansen348f8b62005-06-23 00:07:40 -07001042static inline void set_page_node(struct page *page, unsigned long node)
1043{
1044 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1045 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1046}
Christoph Lameter89689ae2006-12-06 20:31:45 -08001047
Christoph Lameter2f1b6242006-09-25 23:31:13 -07001048static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001049 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -07001050{
1051 set_page_zone(page, zone);
1052 set_page_node(page, node);
Cody P Schafer9127ab42013-02-22 16:35:21 -08001053#ifdef SECTION_IN_PAGE_FLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001054 set_page_section(page, pfn_to_section_nr(pfn));
Daniel Kiperbf4e8902011-05-24 17:12:32 -07001055#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056}
1057
Greg Thelen0610c252015-10-01 15:37:02 -07001058#ifdef CONFIG_MEMCG
1059static inline struct mem_cgroup *page_memcg(struct page *page)
1060{
1061 return page->mem_cgroup;
1062}
Johannes Weiner55779ec2016-07-28 15:45:10 -07001063static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1064{
1065 WARN_ON_ONCE(!rcu_read_lock_held());
1066 return READ_ONCE(page->mem_cgroup);
1067}
Greg Thelen0610c252015-10-01 15:37:02 -07001068#else
1069static inline struct mem_cgroup *page_memcg(struct page *page)
1070{
1071 return NULL;
1072}
Johannes Weiner55779ec2016-07-28 15:45:10 -07001073static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1074{
1075 WARN_ON_ONCE(!rcu_read_lock_held());
1076 return NULL;
1077}
Greg Thelen0610c252015-10-01 15:37:02 -07001078#endif
1079
Christoph Lameterf6ac2352006-06-30 01:55:32 -07001080/*
1081 * Some inline functions in vmstat.h depend on page_zone()
1082 */
1083#include <linux/vmstat.h>
1084
Ian Campbell33dd4e02011-07-25 17:11:51 -07001085static __always_inline void *lowmem_page_address(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086{
Ard Biesheuvel1dff8082016-04-18 18:04:57 +02001087 return page_to_virt(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088}
1089
1090#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1091#define HASHED_PAGE_VIRTUAL
1092#endif
1093
1094#if defined(WANT_PAGE_VIRTUAL)
Geert Uytterhoevenf92f4552014-01-21 15:48:47 -08001095static inline void *page_address(const struct page *page)
1096{
1097 return page->virtual;
1098}
1099static inline void set_page_address(struct page *page, void *address)
1100{
1101 page->virtual = address;
1102}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103#define page_address_init() do { } while(0)
1104#endif
1105
1106#if defined(HASHED_PAGE_VIRTUAL)
Ian Campbellf9918792011-08-17 13:45:09 +01001107void *page_address(const struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108void set_page_address(struct page *page, void *virtual);
1109void page_address_init(void);
1110#endif
1111
1112#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1113#define page_address(page) lowmem_page_address(page)
1114#define set_page_address(page, address) do { } while(0)
1115#define page_address_init() do { } while(0)
1116#endif
1117
Kirill A. Shutemove39155e2015-04-15 16:14:53 -07001118extern void *page_rmapping(struct page *page);
1119extern struct anon_vma *page_anon_vma(struct page *page);
Shaohua Li98003392013-02-22 16:34:35 -08001120extern struct address_space *page_mapping(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
Mel Gormanf981c592012-07-31 16:44:47 -07001122extern struct address_space *__page_file_mapping(struct page *);
1123
1124static inline
1125struct address_space *page_file_mapping(struct page *page)
1126{
1127 if (unlikely(PageSwapCache(page)))
1128 return __page_file_mapping(page);
1129
1130 return page->mapping;
1131}
1132
Huang Yingf6ab1f72016-10-07 17:00:21 -07001133extern pgoff_t __page_file_index(struct page *page);
1134
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135/*
1136 * Return the pagecache index of the passed page. Regular pagecache pages
Huang Yingf6ab1f72016-10-07 17:00:21 -07001137 * use ->index whereas swapcache pages use swp_offset(->private)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 */
1139static inline pgoff_t page_index(struct page *page)
1140{
1141 if (unlikely(PageSwapCache(page)))
Huang Yingf6ab1f72016-10-07 17:00:21 -07001142 return __page_file_index(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 return page->index;
1144}
1145
Andrew Morton1aa8aea2016-05-19 17:12:00 -07001146bool page_mapped(struct page *page);
Minchan Kimbda807d2016-07-26 15:23:05 -07001147struct address_space *page_mapping(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
1149/*
Michal Hocko2f064f32015-08-21 14:11:51 -07001150 * Return true only if the page has been allocated with
1151 * ALLOC_NO_WATERMARKS and the low watermark was not
1152 * met implying that the system is under some pressure.
1153 */
1154static inline bool page_is_pfmemalloc(struct page *page)
1155{
1156 /*
1157 * Page index cannot be this large so this must be
1158 * a pfmemalloc page.
1159 */
1160 return page->index == -1UL;
1161}
1162
1163/*
1164 * Only to be called by the page allocator on a freshly allocated
1165 * page.
1166 */
1167static inline void set_page_pfmemalloc(struct page *page)
1168{
1169 page->index = -1UL;
1170}
1171
1172static inline void clear_page_pfmemalloc(struct page *page)
1173{
1174 page->index = 0;
1175}
1176
1177/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 * Different kinds of faults, as returned by handle_mm_fault().
1179 * Used to decide whether a process gets delivered SIGBUS or
1180 * just gets major/minor fault counters bumped up.
1181 */
Nick Piggind0217ac2007-07-19 01:47:03 -07001182
Nick Piggin83c54072007-07-19 01:47:05 -07001183#define VM_FAULT_OOM 0x0001
1184#define VM_FAULT_SIGBUS 0x0002
1185#define VM_FAULT_MAJOR 0x0004
1186#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001187#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1188#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
Linus Torvalds33692f22015-01-29 10:51:32 -08001189#define VM_FAULT_SIGSEGV 0x0040
Nick Pigginf33ea7f2005-08-03 20:24:01 +10001190
Nick Piggin83c54072007-07-19 01:47:05 -07001191#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1192#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Michel Lespinassed065bd82010-10-26 14:21:57 -07001193#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
Kirill A. Shutemovc0292552013-09-12 15:14:05 -07001194#define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
Jan Karab1aa8122016-12-14 15:07:24 -08001195#define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
Jan Karacaa51d22017-11-01 16:36:42 +01001196#define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1197 * and needs fsync() to complete (for
1198 * synchronous page faults in DAX) */
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001199
Linus Torvalds33692f22015-01-29 10:51:32 -08001200#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1201 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1202 VM_FAULT_FALLBACK)
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001203
Ross Zwisler282a8e02017-02-22 15:39:50 -08001204#define VM_FAULT_RESULT_TRACE \
1205 { VM_FAULT_OOM, "OOM" }, \
1206 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1207 { VM_FAULT_MAJOR, "MAJOR" }, \
1208 { VM_FAULT_WRITE, "WRITE" }, \
1209 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1210 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1211 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1212 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1213 { VM_FAULT_LOCKED, "LOCKED" }, \
1214 { VM_FAULT_RETRY, "RETRY" }, \
1215 { VM_FAULT_FALLBACK, "FALLBACK" }, \
Jan Karacaa51d22017-11-01 16:36:42 +01001216 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1217 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
Ross Zwisler282a8e02017-02-22 15:39:50 -08001218
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001219/* Encode hstate index for a hwpoisoned large page */
1220#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1221#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
Nick Piggind0217ac2007-07-19 01:47:03 -07001222
Nick Piggin1c0fe6e2009-01-06 14:38:59 -08001223/*
1224 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1225 */
1226extern void pagefault_out_of_memory(void);
1227
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1229
David Rientjesddd588b2011-03-22 16:30:46 -07001230/*
David Rientjes7bf02ea2011-05-24 17:11:16 -07001231 * Flags passed to show_mem() and show_free_areas() to suppress output in
David Rientjesddd588b2011-03-22 16:30:46 -07001232 * various contexts.
1233 */
David Rientjes4b59e6c2013-04-29 15:06:11 -07001234#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
David Rientjesddd588b2011-03-22 16:30:46 -07001235
Michal Hocko9af744d2017-02-22 15:46:16 -08001236extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237
Wang Xiaoqiang7f43add2016-01-15 16:57:22 -08001238extern bool can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239extern int user_shm_lock(size_t, struct user_struct *);
1240extern void user_shm_unlock(size_t, struct user_struct *);
1241
1242/*
1243 * Parameter block passed down to zap_pte_range in exceptional cases.
1244 */
1245struct zap_details {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 struct address_space *check_mapping; /* Check page->mapping if set */
1247 pgoff_t first_index; /* Lowest page->index to unmap */
1248 pgoff_t last_index; /* Highest page->index to unmap */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249};
1250
Jérôme Glissedf6ad692017-09-08 16:12:24 -07001251struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1252 pte_t pte, bool with_public_device);
1253#define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1254
Gerald Schaefer28093f92016-04-28 16:18:35 -07001255struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1256 pmd_t pmd);
Nick Piggin7e675132008-04-28 02:13:00 -07001257
Jack Steinerc627f9c2008-07-29 22:33:53 -07001258int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1259 unsigned long size);
Al Viro14f5ff52012-03-05 13:38:09 -05001260void zap_page_range(struct vm_area_struct *vma, unsigned long address,
Kirill A. Shutemovecf13852017-02-22 15:46:37 -08001261 unsigned long size);
Linus Torvalds4f74d2c2012-05-06 13:54:06 -07001262void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1263 unsigned long start, unsigned long end);
Matt Mackalle6473092008-02-04 22:29:01 -08001264
1265/**
1266 * mm_walk - callbacks for walk_page_range
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001267 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1268 * this handler should only handle pud_trans_huge() puds.
1269 * the pmd_entry or pte_entry callbacks will be used for
1270 * regular PUDs.
Matt Mackalle6473092008-02-04 22:29:01 -08001271 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
Dave Hansen03319322011-03-22 16:32:56 -07001272 * this handler is required to be able to handle
1273 * pmd_trans_huge() pmds. They may simply choose to
1274 * split_huge_page() instead of handling it explicitly.
Matt Mackalle6473092008-02-04 22:29:01 -08001275 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1276 * @pte_hole: if set, called for each hole at all levels
Naoya Horiguchi5dc37642009-12-14 18:00:01 -08001277 * @hugetlb_entry: if set, called for each hugetlb entry
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001278 * @test_walk: caller specific callback function to determine whether
James Morsef7e23552016-10-07 16:58:36 -07001279 * we walk over the current vma or not. Returning 0
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001280 * value means "do page table walk over the current vma,"
1281 * and a negative one means "abort current page table walk
James Morsef7e23552016-10-07 16:58:36 -07001282 * right now." 1 means "skip the current vma."
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001283 * @mm: mm_struct representing the target process of page table walk
1284 * @vma: vma currently walked (NULL if walking outside vmas)
1285 * @private: private data for callbacks' usage
Matt Mackalle6473092008-02-04 22:29:01 -08001286 *
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001287 * (see the comment on walk_page_range() for more details)
Matt Mackalle6473092008-02-04 22:29:01 -08001288 */
1289struct mm_walk {
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001290 int (*pud_entry)(pud_t *pud, unsigned long addr,
1291 unsigned long next, struct mm_walk *walk);
Andrew Morton0f157a52013-05-07 16:18:10 -07001292 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1293 unsigned long next, struct mm_walk *walk);
1294 int (*pte_entry)(pte_t *pte, unsigned long addr,
1295 unsigned long next, struct mm_walk *walk);
1296 int (*pte_hole)(unsigned long addr, unsigned long next,
1297 struct mm_walk *walk);
1298 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1299 unsigned long addr, unsigned long next,
1300 struct mm_walk *walk);
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001301 int (*test_walk)(unsigned long addr, unsigned long next,
1302 struct mm_walk *walk);
Dave Hansen21650092008-06-12 15:21:47 -07001303 struct mm_struct *mm;
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001304 struct vm_area_struct *vma;
Dave Hansen21650092008-06-12 15:21:47 -07001305 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -08001306};
1307
Dave Hansen21650092008-06-12 15:21:47 -07001308int walk_page_range(unsigned long addr, unsigned long end,
1309 struct mm_walk *walk);
Naoya Horiguchi900fc5f2015-02-11 15:27:40 -08001310int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -07001311void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -07001312 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1314 struct vm_area_struct *vma);
Ross Zwisler09796392017-01-10 16:57:21 -08001315int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
Jérôme Glissea4d1a882017-08-31 17:17:26 -04001316 unsigned long *start, unsigned long *end,
Ross Zwisler09796392017-01-10 16:57:21 -08001317 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
Johannes Weiner3b6748e2009-06-16 15:32:35 -07001318int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1319 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -08001320int follow_phys(struct vm_area_struct *vma, unsigned long address,
1321 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -07001322int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1323 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324
Kirill A. Shutemov7caef262013-09-12 15:13:56 -07001325extern void truncate_pagecache(struct inode *inode, loff_t new);
Christoph Hellwig2c27c652010-06-04 11:30:04 +02001326extern void truncate_setsize(struct inode *inode, loff_t newsize);
Jan Kara90a80202014-10-01 21:49:18 -04001327void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
Hugh Dickins623e3db2012-03-28 14:42:40 -07001328void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
Nick Piggin750b4982009-09-16 11:50:12 +02001329int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +02001330int generic_error_remove_page(struct address_space *mapping, struct page *page);
Wu Fengguang83f78662009-09-16 11:50:13 +02001331int invalidate_inode_page(struct page *page);
1332
David Howells7ee1dd32006-01-06 00:11:44 -08001333#ifdef CONFIG_MMU
Kirill A. Shutemovdcddffd2016-07-26 15:25:18 -07001334extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1335 unsigned int flags);
Peter Zijlstra5c723ba2011-07-27 12:17:11 +02001336extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
Dominik Dingel4a9e1cd2016-01-15 16:57:04 -08001337 unsigned long address, unsigned int fault_flags,
1338 bool *unlocked);
Matthew Wilcox977fbdc2018-01-31 16:17:36 -08001339void unmap_mapping_pages(struct address_space *mapping,
1340 pgoff_t start, pgoff_t nr, bool even_cows);
1341void unmap_mapping_range(struct address_space *mapping,
1342 loff_t const holebegin, loff_t const holelen, int even_cows);
David Howells7ee1dd32006-01-06 00:11:44 -08001343#else
Kirill A. Shutemovdcddffd2016-07-26 15:25:18 -07001344static inline int handle_mm_fault(struct vm_area_struct *vma,
1345 unsigned long address, unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -08001346{
1347 /* should never happen if there's no MMU */
1348 BUG();
1349 return VM_FAULT_SIGBUS;
1350}
Peter Zijlstra5c723ba2011-07-27 12:17:11 +02001351static inline int fixup_user_fault(struct task_struct *tsk,
1352 struct mm_struct *mm, unsigned long address,
Dominik Dingel4a9e1cd2016-01-15 16:57:04 -08001353 unsigned int fault_flags, bool *unlocked)
Peter Zijlstra5c723ba2011-07-27 12:17:11 +02001354{
1355 /* should never happen if there's no MMU */
1356 BUG();
1357 return -EFAULT;
1358}
Matthew Wilcox977fbdc2018-01-31 16:17:36 -08001359static inline void unmap_mapping_pages(struct address_space *mapping,
1360 pgoff_t start, pgoff_t nr, bool even_cows) { }
1361static inline void unmap_mapping_range(struct address_space *mapping,
1362 loff_t const holebegin, loff_t const holelen, int even_cows) { }
David Howells7ee1dd32006-01-06 00:11:44 -08001363#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +10001364
Matthew Wilcox977fbdc2018-01-31 16:17:36 -08001365static inline void unmap_shared_mapping_range(struct address_space *mapping,
1366 loff_t const holebegin, loff_t const holelen)
1367{
1368 unmap_mapping_range(mapping, holebegin, holelen, 0);
1369}
1370
1371extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1372 void *buf, int len, unsigned int gup_flags);
Stephen Wilson5ddd36b2011-03-13 15:49:20 -04001373extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +01001374 void *buf, int len, unsigned int gup_flags);
Eric W. Biederman84d77d32016-11-22 12:06:50 -06001375extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1376 unsigned long addr, void *buf, int len, unsigned int gup_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377
Dave Hansen1e987792016-02-12 13:01:54 -08001378long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1379 unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +01001380 unsigned int gup_flags, struct page **pages,
Lorenzo Stoakes5b56d492016-12-14 15:06:52 -08001381 struct vm_area_struct **vmas, int *locked);
Ingo Molnarc12d2da2016-04-04 10:24:58 +02001382long get_user_pages(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes768ae302016-10-13 01:20:16 +01001383 unsigned int gup_flags, struct page **pages,
Dave Hansencde70142016-02-12 13:01:55 -08001384 struct vm_area_struct **vmas);
Ingo Molnarc12d2da2016-04-04 10:24:58 +02001385long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes3b913172016-10-13 01:20:14 +01001386 unsigned int gup_flags, struct page **pages, int *locked);
Ingo Molnarc12d2da2016-04-04 10:24:58 +02001387long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakesc1641542016-10-13 01:20:13 +01001388 struct page **pages, unsigned int gup_flags);
Dan Williams2bb6d282017-11-29 16:10:35 -08001389#ifdef CONFIG_FS_DAX
1390long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1391 unsigned int gup_flags, struct page **pages,
1392 struct vm_area_struct **vmas);
1393#else
1394static inline long get_user_pages_longterm(unsigned long start,
1395 unsigned long nr_pages, unsigned int gup_flags,
1396 struct page **pages, struct vm_area_struct **vmas)
1397{
1398 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1399}
1400#endif /* CONFIG_FS_DAX */
1401
Nick Piggind2bf6be2009-06-16 15:31:39 -07001402int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1403 struct page **pages);
Jan Kara8025e5d2015-07-13 11:55:44 -03001404
1405/* Container for pinned pfns / pages */
1406struct frame_vector {
1407 unsigned int nr_allocated; /* Number of frames we have space for */
1408 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1409 bool got_ref; /* Did we pin pages by getting page ref? */
1410 bool is_pfns; /* Does array contain pages or pfns? */
1411 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1412 * pfns_vector_pages() or pfns_vector_pfns()
1413 * for access */
1414};
1415
1416struct frame_vector *frame_vector_create(unsigned int nr_frames);
1417void frame_vector_destroy(struct frame_vector *vec);
1418int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
Lorenzo Stoakes7f23b352016-10-13 01:20:15 +01001419 unsigned int gup_flags, struct frame_vector *vec);
Jan Kara8025e5d2015-07-13 11:55:44 -03001420void put_vaddr_frames(struct frame_vector *vec);
1421int frame_vector_to_pages(struct frame_vector *vec);
1422void frame_vector_to_pfns(struct frame_vector *vec);
1423
1424static inline unsigned int frame_vector_count(struct frame_vector *vec)
1425{
1426 return vec->nr_frames;
1427}
1428
1429static inline struct page **frame_vector_pages(struct frame_vector *vec)
1430{
1431 if (vec->is_pfns) {
1432 int err = frame_vector_to_pages(vec);
1433
1434 if (err)
1435 return ERR_PTR(err);
1436 }
1437 return (struct page **)(vec->ptrs);
1438}
1439
1440static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1441{
1442 if (!vec->is_pfns)
1443 frame_vector_to_pfns(vec);
1444 return (unsigned long *)(vec->ptrs);
1445}
1446
Mel Gorman18022c52012-07-31 16:44:51 -07001447struct kvec;
1448int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1449 struct page **pages);
1450int get_kernel_page(unsigned long start, int write, struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07001451struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452
David Howellscf9a2ae2006-08-29 19:05:54 +01001453extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
Lukas Czernerd47992f2013-05-21 23:17:23 -04001454extern void do_invalidatepage(struct page *page, unsigned int offset,
1455 unsigned int length);
David Howellscf9a2ae2006-08-29 19:05:54 +01001456
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -08001458int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459int redirty_page_for_writepage(struct writeback_control *wbc,
1460 struct page *page);
Johannes Weiner62cccb82016-03-15 14:57:22 -07001461void account_page_dirtied(struct page *page, struct address_space *mapping);
Greg Thelenc4843a72015-05-22 17:13:16 -04001462void account_page_cleaned(struct page *page, struct address_space *mapping,
Johannes Weiner62cccb82016-03-15 14:57:22 -07001463 struct bdi_writeback *wb);
Harvey Harrisonb3c97522008-02-13 15:03:15 -08001464int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465int set_page_dirty_lock(struct page *page);
Jan Kara736304f2017-11-15 17:37:11 -08001466void __cancel_dirty_page(struct page *page);
1467static inline void cancel_dirty_page(struct page *page)
1468{
1469 /* Avoid atomic ops, locking, etc. when not actually needed. */
1470 if (PageDirty(page))
1471 __cancel_dirty_page(page);
1472}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473int clear_page_dirty_for_io(struct page *page);
Konstantin Khlebnikovb9ea2512015-04-14 15:45:27 -07001474
William Robertsa9090252014-02-11 10:11:59 -08001475int get_cmdline(struct task_struct *task, char *buffer, int buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476
Oleg Nesterovb5330622015-09-08 14:58:28 -07001477static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1478{
1479 return !vma->vm_ops;
1480}
1481
Mike Rapoportb0506e42017-02-22 15:43:28 -08001482#ifdef CONFIG_SHMEM
1483/*
1484 * The vma_is_shmem is not inline because it is used only by slow
1485 * paths in userfault.
1486 */
1487bool vma_is_shmem(struct vm_area_struct *vma);
1488#else
1489static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1490#endif
1491
Andy Lutomirskid17af502016-09-30 10:58:58 -07001492int vma_is_stack_for_current(struct vm_area_struct *vma);
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07001493
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001494extern unsigned long move_page_tables(struct vm_area_struct *vma,
1495 unsigned long old_addr, struct vm_area_struct *new_vma,
Michel Lespinasse38a76012012-10-08 16:31:50 -07001496 unsigned long new_addr, unsigned long len,
1497 bool need_rmap_locks);
Peter Zijlstra7da4d642012-11-19 03:14:23 +01001498extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1499 unsigned long end, pgprot_t newprot,
Mel Gorman4b10e7d2012-10-25 14:16:32 +02001500 int dirty_accountable, int prot_numa);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001501extern int mprotect_fixup(struct vm_area_struct *vma,
1502 struct vm_area_struct **pprev, unsigned long start,
1503 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
Nick Piggin21cc1992008-07-25 19:45:22 -07001505/*
Peter Zijlstra465a4542009-06-15 12:31:37 +02001506 * doesn't attempt to fault and will return short.
1507 */
1508int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1509 struct page **pages);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001510/*
1511 * per-process(per-mm_struct) statistics.
1512 */
Matt Fleming172703b2011-05-24 17:12:36 -07001513static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1514{
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001515 long val = atomic_long_read(&mm->rss_stat.count[member]);
1516
1517#ifdef SPLIT_RSS_COUNTING
1518 /*
1519 * counter is updated in asynchronous manner and may go to minus.
1520 * But it's never be expected number for users.
1521 */
1522 if (val < 0)
1523 val = 0;
Matt Fleming172703b2011-05-24 17:12:36 -07001524#endif
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001525 return (unsigned long)val;
1526}
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001527
1528static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1529{
1530 atomic_long_add(value, &mm->rss_stat.count[member]);
1531}
1532
1533static inline void inc_mm_counter(struct mm_struct *mm, int member)
1534{
1535 atomic_long_inc(&mm->rss_stat.count[member]);
1536}
1537
1538static inline void dec_mm_counter(struct mm_struct *mm, int member)
1539{
1540 atomic_long_dec(&mm->rss_stat.count[member]);
1541}
1542
Jerome Marchandeca56ff2016-01-14 15:19:26 -08001543/* Optimized variant when page is already known not to be PageAnon */
1544static inline int mm_counter_file(struct page *page)
1545{
1546 if (PageSwapBacked(page))
1547 return MM_SHMEMPAGES;
1548 return MM_FILEPAGES;
1549}
1550
1551static inline int mm_counter(struct page *page)
1552{
1553 if (PageAnon(page))
1554 return MM_ANONPAGES;
1555 return mm_counter_file(page);
1556}
1557
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001558static inline unsigned long get_mm_rss(struct mm_struct *mm)
1559{
1560 return get_mm_counter(mm, MM_FILEPAGES) +
Jerome Marchandeca56ff2016-01-14 15:19:26 -08001561 get_mm_counter(mm, MM_ANONPAGES) +
1562 get_mm_counter(mm, MM_SHMEMPAGES);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001563}
1564
1565static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1566{
1567 return max(mm->hiwater_rss, get_mm_rss(mm));
1568}
1569
1570static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1571{
1572 return max(mm->hiwater_vm, mm->total_vm);
1573}
1574
1575static inline void update_hiwater_rss(struct mm_struct *mm)
1576{
1577 unsigned long _rss = get_mm_rss(mm);
1578
1579 if ((mm)->hiwater_rss < _rss)
1580 (mm)->hiwater_rss = _rss;
1581}
1582
1583static inline void update_hiwater_vm(struct mm_struct *mm)
1584{
1585 if (mm->hiwater_vm < mm->total_vm)
1586 mm->hiwater_vm = mm->total_vm;
1587}
1588
Petr Cermak695f0552015-02-12 15:01:00 -08001589static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1590{
1591 mm->hiwater_rss = get_mm_rss(mm);
1592}
1593
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001594static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1595 struct mm_struct *mm)
1596{
1597 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1598
1599 if (*maxrss < hiwater_rss)
1600 *maxrss = hiwater_rss;
1601}
1602
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001603#if defined(SPLIT_RSS_COUNTING)
David Rientjes05af2e12012-03-21 16:34:13 -07001604void sync_mm_rss(struct mm_struct *mm);
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001605#else
David Rientjes05af2e12012-03-21 16:34:13 -07001606static inline void sync_mm_rss(struct mm_struct *mm)
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001607{
1608}
1609#endif
Peter Zijlstra465a4542009-06-15 12:31:37 +02001610
Dan Williams3565fce2016-01-15 16:56:55 -08001611#ifndef __HAVE_ARCH_PTE_DEVMAP
1612static inline int pte_devmap(pte_t pte)
1613{
1614 return 0;
1615}
1616#endif
1617
Andrea Arcangeli6d2329f2016-10-07 17:01:22 -07001618int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
Peter Zijlstrad08b3852006-09-25 23:30:57 -07001619
Namhyung Kim25ca1d62010-10-26 14:21:59 -07001620extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1621 spinlock_t **ptl);
1622static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1623 spinlock_t **ptl)
1624{
1625 pte_t *ptep;
1626 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1627 return ptep;
1628}
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -08001629
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001630#ifdef __PAGETABLE_P4D_FOLDED
1631static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
Nick Piggin5f22df02007-05-06 14:49:02 -07001632 unsigned long address)
1633{
1634 return 0;
1635}
1636#else
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001637int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1638#endif
1639
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001640#if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001641static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1642 unsigned long address)
1643{
1644 return 0;
1645}
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001646static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1647static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1648
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001649#else
1650int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001651
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001652static inline void mm_inc_nr_puds(struct mm_struct *mm)
1653{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001654 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001655}
1656
1657static inline void mm_dec_nr_puds(struct mm_struct *mm)
1658{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001659 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001660}
Nick Piggin5f22df02007-05-06 14:49:02 -07001661#endif
1662
Kirill A. Shutemov2d2f5112015-02-12 14:59:59 -08001663#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
Nick Piggin5f22df02007-05-06 14:49:02 -07001664static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1665 unsigned long address)
1666{
1667 return 0;
1668}
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001669
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001670static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1671static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1672
Nick Piggin5f22df02007-05-06 14:49:02 -07001673#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001674int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001675
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001676static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1677{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001678 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001679}
1680
1681static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1682{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001683 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001684}
Nick Piggin5f22df02007-05-06 14:49:02 -07001685#endif
1686
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001687#ifdef CONFIG_MMU
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001688static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001689{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001690 atomic_long_set(&mm->pgtables_bytes, 0);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001691}
1692
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001693static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001694{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001695 return atomic_long_read(&mm->pgtables_bytes);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001696}
1697
1698static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1699{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001700 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001701}
1702
1703static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1704{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001705 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001706}
1707#else
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001708
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001709static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1710static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001711{
1712 return 0;
1713}
1714
1715static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1716static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1717#endif
1718
Kirill A. Shutemov3ed3a4f2016-03-17 14:19:11 -07001719int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
Hugh Dickins1bb36302005-10-29 18:16:22 -07001720int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1721
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722/*
1723 * The following ifdef needed to get the 4level-fixup.h header to work.
1724 * Remove it when 4level-fixup.h has been removed.
1725 */
Hugh Dickins1bb36302005-10-29 18:16:22 -07001726#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Kirill A. Shutemov505a60e2017-03-09 17:24:03 +03001727
1728#ifndef __ARCH_HAS_5LEVEL_HACK
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001729static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1730 unsigned long address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731{
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001732 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1733 NULL : p4d_offset(pgd, address);
1734}
1735
1736static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1737 unsigned long address)
1738{
1739 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1740 NULL : pud_offset(p4d, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741}
Kirill A. Shutemov505a60e2017-03-09 17:24:03 +03001742#endif /* !__ARCH_HAS_5LEVEL_HACK */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
1744static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1745{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001746 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1747 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748}
Hugh Dickins1bb36302005-10-29 18:16:22 -07001749#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1750
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001751#if USE_SPLIT_PTE_PTLOCKS
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001752#if ALLOC_SPLIT_PTLOCKS
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001753void __init ptlock_cache_init(void);
Peter Zijlstra539edb52013-11-14 14:31:52 -08001754extern bool ptlock_alloc(struct page *page);
1755extern void ptlock_free(struct page *page);
1756
1757static inline spinlock_t *ptlock_ptr(struct page *page)
1758{
1759 return page->ptl;
1760}
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001761#else /* ALLOC_SPLIT_PTLOCKS */
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001762static inline void ptlock_cache_init(void)
1763{
1764}
1765
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001766static inline bool ptlock_alloc(struct page *page)
1767{
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001768 return true;
1769}
Peter Zijlstra539edb52013-11-14 14:31:52 -08001770
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001771static inline void ptlock_free(struct page *page)
1772{
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001773}
1774
1775static inline spinlock_t *ptlock_ptr(struct page *page)
1776{
Peter Zijlstra539edb52013-11-14 14:31:52 -08001777 return &page->ptl;
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001778}
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001779#endif /* ALLOC_SPLIT_PTLOCKS */
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001780
1781static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1782{
1783 return ptlock_ptr(pmd_page(*pmd));
1784}
1785
1786static inline bool ptlock_init(struct page *page)
1787{
1788 /*
1789 * prep_new_page() initialize page->private (and therefore page->ptl)
1790 * with 0. Make sure nobody took it in use in between.
1791 *
1792 * It can happen if arch try to use slab for page table allocation:
Kirill A. Shutemov1d798ca2015-11-06 16:29:54 -08001793 * slab code uses page->slab_cache, which share storage with page->ptl.
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001794 */
Sasha Levin309381fea2014-01-23 15:52:54 -08001795 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001796 if (!ptlock_alloc(page))
1797 return false;
1798 spin_lock_init(ptlock_ptr(page));
1799 return true;
1800}
1801
1802/* Reset page->mapping so free_pages_check won't complain. */
1803static inline void pte_lock_deinit(struct page *page)
1804{
1805 page->mapping = NULL;
1806 ptlock_free(page);
1807}
1808
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001809#else /* !USE_SPLIT_PTE_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001810/*
1811 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1812 */
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001813static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1814{
1815 return &mm->page_table_lock;
1816}
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001817static inline void ptlock_cache_init(void) {}
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001818static inline bool ptlock_init(struct page *page) { return true; }
1819static inline void pte_lock_deinit(struct page *page) {}
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001820#endif /* USE_SPLIT_PTE_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001821
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001822static inline void pgtable_init(void)
1823{
1824 ptlock_cache_init();
1825 pgtable_cache_init();
1826}
1827
Kirill A. Shutemov390f44e2013-11-14 14:31:20 -08001828static inline bool pgtable_page_ctor(struct page *page)
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001829{
Vladimir Davydov706874e2015-11-05 18:49:27 -08001830 if (!ptlock_init(page))
1831 return false;
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001832 inc_zone_page_state(page, NR_PAGETABLE);
Vladimir Davydov706874e2015-11-05 18:49:27 -08001833 return true;
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001834}
1835
1836static inline void pgtable_page_dtor(struct page *page)
1837{
1838 pte_lock_deinit(page);
1839 dec_zone_page_state(page, NR_PAGETABLE);
1840}
1841
Hugh Dickinsc74df322005-10-29 18:16:23 -07001842#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1843({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001844 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001845 pte_t *__pte = pte_offset_map(pmd, address); \
1846 *(ptlp) = __ptl; \
1847 spin_lock(__ptl); \
1848 __pte; \
1849})
1850
1851#define pte_unmap_unlock(pte, ptl) do { \
1852 spin_unlock(ptl); \
1853 pte_unmap(pte); \
1854} while (0)
1855
Kirill A. Shutemov3ed3a4f2016-03-17 14:19:11 -07001856#define pte_alloc(mm, pmd, address) \
1857 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1858
1859#define pte_alloc_map(mm, pmd, address) \
1860 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
Hugh Dickins1bb36302005-10-29 18:16:22 -07001861
Hugh Dickinsc74df322005-10-29 18:16:23 -07001862#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
Kirill A. Shutemov3ed3a4f2016-03-17 14:19:11 -07001863 (pte_alloc(mm, pmd, address) ? \
1864 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
Hugh Dickinsc74df322005-10-29 18:16:23 -07001865
Hugh Dickins1bb36302005-10-29 18:16:22 -07001866#define pte_alloc_kernel(pmd, address) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001867 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
Hugh Dickins1bb36302005-10-29 18:16:22 -07001868 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001870#if USE_SPLIT_PMD_PTLOCKS
1871
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001872static struct page *pmd_to_page(pmd_t *pmd)
1873{
1874 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1875 return virt_to_page((void *)((unsigned long) pmd & mask));
1876}
1877
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001878static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1879{
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001880 return ptlock_ptr(pmd_to_page(pmd));
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001881}
1882
1883static inline bool pgtable_pmd_page_ctor(struct page *page)
1884{
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001885#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1886 page->pmd_huge_pte = NULL;
1887#endif
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001888 return ptlock_init(page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001889}
1890
1891static inline void pgtable_pmd_page_dtor(struct page *page)
1892{
1893#ifdef CONFIG_TRANSPARENT_HUGEPAGE
Sasha Levin309381fea2014-01-23 15:52:54 -08001894 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001895#endif
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001896 ptlock_free(page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001897}
1898
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001899#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001900
1901#else
1902
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001903static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1904{
1905 return &mm->page_table_lock;
1906}
1907
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001908static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1909static inline void pgtable_pmd_page_dtor(struct page *page) {}
1910
Kirill A. Shutemovc389a252013-11-14 14:30:59 -08001911#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001912
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001913#endif
1914
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001915static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1916{
1917 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1918 spin_lock(ptl);
1919 return ptl;
1920}
1921
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001922/*
1923 * No scalability reason to split PUD locks yet, but follow the same pattern
1924 * as the PMD locks to make it easier if we decide to. The VM should not be
1925 * considered ready to switch to split PUD locks yet; there may be places
1926 * which need to be converted from page_table_lock.
1927 */
1928static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1929{
1930 return &mm->page_table_lock;
1931}
Nicholas Piggin62906022016-12-25 13:00:30 +10001932
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001933static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1934{
1935 spinlock_t *ptl = pud_lockptr(mm, pud);
1936
1937 spin_lock(ptl);
1938 return ptl;
1939}
1940
1941extern void __init pagecache_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001943extern void free_area_init_node(int nid, unsigned long * zones_size,
1944 unsigned long zone_start_pfn, unsigned long *zholes_size);
David Howells49a7f042012-03-28 18:30:03 +01001945extern void free_initmem(void);
1946
Jiang Liu69afade2013-04-29 15:06:21 -07001947/*
1948 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1949 * into the buddy system. The freed pages will be poisoned with pattern
Jiang Liudbe67df2013-07-03 15:02:51 -07001950 * "poison" if it's within range [0, UCHAR_MAX].
Jiang Liu69afade2013-04-29 15:06:21 -07001951 * Return pages freed into the buddy system.
1952 */
Jiang Liu11199692013-07-03 15:02:48 -07001953extern unsigned long free_reserved_area(void *start, void *end,
Jiang Liu69afade2013-04-29 15:06:21 -07001954 int poison, char *s);
Jiang Liuc3d5f5f2013-07-03 15:03:14 -07001955
Jiang Liucfa11e02013-04-29 15:07:00 -07001956#ifdef CONFIG_HIGHMEM
1957/*
1958 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1959 * and totalram_pages.
1960 */
1961extern void free_highmem_page(struct page *page);
1962#endif
Jiang Liu69afade2013-04-29 15:06:21 -07001963
Jiang Liuc3d5f5f2013-07-03 15:03:14 -07001964extern void adjust_managed_page_count(struct page *page, long count);
Jiang Liu7ee3d4e2013-07-03 15:03:41 -07001965extern void mem_init_print_info(const char *str);
Jiang Liu69afade2013-04-29 15:06:21 -07001966
Stefan Bader4b50bcc2016-05-20 16:58:38 -07001967extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
Nathan Zimmer92923ca2015-06-30 14:56:48 -07001968
Jiang Liu69afade2013-04-29 15:06:21 -07001969/* Free the reserved page into the buddy system, so it gets managed. */
1970static inline void __free_reserved_page(struct page *page)
1971{
1972 ClearPageReserved(page);
1973 init_page_count(page);
1974 __free_page(page);
1975}
1976
1977static inline void free_reserved_page(struct page *page)
1978{
1979 __free_reserved_page(page);
1980 adjust_managed_page_count(page, 1);
1981}
1982
1983static inline void mark_page_reserved(struct page *page)
1984{
1985 SetPageReserved(page);
1986 adjust_managed_page_count(page, -1);
1987}
1988
1989/*
1990 * Default method to free all the __init memory into the buddy system.
Jiang Liudbe67df2013-07-03 15:02:51 -07001991 * The freed pages will be poisoned with pattern "poison" if it's within
1992 * range [0, UCHAR_MAX].
1993 * Return pages freed into the buddy system.
Jiang Liu69afade2013-04-29 15:06:21 -07001994 */
1995static inline unsigned long free_initmem_default(int poison)
1996{
1997 extern char __init_begin[], __init_end[];
1998
Jiang Liu11199692013-07-03 15:02:48 -07001999 return free_reserved_area(&__init_begin, &__init_end,
Jiang Liu69afade2013-04-29 15:06:21 -07002000 poison, "unused kernel");
2001}
2002
Jiang Liu7ee3d4e2013-07-03 15:03:41 -07002003static inline unsigned long get_num_physpages(void)
2004{
2005 int nid;
2006 unsigned long phys_pages = 0;
2007
2008 for_each_online_node(nid)
2009 phys_pages += node_present_pages(nid);
2010
2011 return phys_pages;
2012}
2013
Tejun Heo0ee332c2011-12-08 10:22:09 -08002014#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gormanc7132162006-09-27 01:49:43 -07002015/*
Tejun Heo0ee332c2011-12-08 10:22:09 -08002016 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
Mel Gormanc7132162006-09-27 01:49:43 -07002017 * zones, allocate the backing mem_map and account for memory holes in a more
2018 * architecture independent manner. This is a substitute for creating the
2019 * zone_sizes[] and zholes_size[] arrays and passing them to
2020 * free_area_init_node()
2021 *
2022 * An architecture is expected to register range of page frames backed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08002023 * physical memory with memblock_add[_node]() before calling
Mel Gormanc7132162006-09-27 01:49:43 -07002024 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2025 * usage, an architecture is expected to do something like
2026 *
2027 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2028 * max_highmem_pfn};
2029 * for_each_valid_physical_page_range()
Tejun Heo0ee332c2011-12-08 10:22:09 -08002030 * memblock_add_node(base, size, nid)
Mel Gormanc7132162006-09-27 01:49:43 -07002031 * free_area_init_nodes(max_zone_pfns);
2032 *
Tejun Heo0ee332c2011-12-08 10:22:09 -08002033 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2034 * registered physical page range. Similarly
2035 * sparse_memory_present_with_active_regions() calls memory_present() for
2036 * each range when SPARSEMEM is enabled.
Mel Gormanc7132162006-09-27 01:49:43 -07002037 *
2038 * See mm/page_alloc.c for more information on each function exposed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08002039 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
Mel Gormanc7132162006-09-27 01:49:43 -07002040 */
2041extern void free_area_init_nodes(unsigned long *max_zone_pfn);
Tejun Heo1e019792011-07-12 09:45:34 +02002042unsigned long node_map_pfn_alignment(void);
Yinghai Lu32996252009-12-15 17:59:02 -08002043unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2044 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07002045extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2046 unsigned long end_pfn);
2047extern void get_pfn_range_for_nid(unsigned int nid,
2048 unsigned long *start_pfn, unsigned long *end_pfn);
2049extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07002050extern void free_bootmem_with_active_regions(int nid,
2051 unsigned long max_low_pfn);
2052extern void sparse_memory_present_with_active_regions(int nid);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002053
Tejun Heo0ee332c2011-12-08 10:22:09 -08002054#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
Tejun Heo5dfe8662011-07-14 09:46:10 +02002055
Tejun Heo0ee332c2011-12-08 10:22:09 -08002056#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002057 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
Mel Gorman8a942fd2015-06-30 14:56:55 -07002058static inline int __early_pfn_to_nid(unsigned long pfn,
2059 struct mminit_pfnnid_cache *state)
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002060{
2061 return 0;
2062}
2063#else
2064/* please see mm/page_alloc.c */
2065extern int __meminit early_pfn_to_nid(unsigned long pfn);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002066/* there is a per-arch backend function. */
Mel Gorman8a942fd2015-06-30 14:56:55 -07002067extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2068 struct mminit_pfnnid_cache *state);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002069#endif
2070
Pavel Tatashina4a3ede2017-11-15 17:36:31 -08002071#ifdef CONFIG_HAVE_MEMBLOCK
2072void zero_resv_unavail(void);
2073#else
2074static inline void zero_resv_unavail(void) {}
2075#endif
2076
Mel Gorman0e0b8642006-09-27 01:49:56 -07002077extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08002078extern void memmap_init_zone(unsigned long, int, unsigned long,
2079 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07002080extern void setup_per_zone_wmarks(void);
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -07002081extern int __meminit init_per_zone_wmark_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00002083extern void __init mmap_init(void);
Michal Hocko9af744d2017-02-22 15:46:16 -08002084extern void show_mem(unsigned int flags, nodemask_t *nodemask);
Igor Redkod02bd272016-03-17 14:19:05 -07002085extern long si_mem_available(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086extern void si_meminfo(struct sysinfo * val);
2087extern void si_meminfo_node(struct sysinfo *val, int nid);
Srikar Dronamrajuf6f34b42016-10-07 16:59:15 -07002088#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2089extern unsigned long arch_reserved_kernel_pages(void);
2090#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091
Michal Hockoa8e99252017-02-22 15:46:10 -08002092extern __printf(3, 4)
2093void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
Dave Hansena238ab52011-05-24 17:12:16 -07002094
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07002095extern void setup_per_cpu_pageset(void);
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07002096
Shaohua Li112067f2009-09-21 17:01:16 -07002097extern void zone_pcp_update(struct zone *zone);
Jiang Liu340175b2012-07-31 16:43:32 -07002098extern void zone_pcp_reset(struct zone *zone);
Shaohua Li112067f2009-09-21 17:01:16 -07002099
Paul Szabo75f7ad82013-02-22 16:34:42 -08002100/* page_alloc.c */
2101extern int min_free_kbytes;
Johannes Weiner795ae7a2016-03-17 14:19:14 -07002102extern int watermark_scale_factor;
Paul Szabo75f7ad82013-02-22 16:34:42 -08002103
David Howells8feae132009-01-08 12:04:47 +00002104/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07002105extern atomic_long_t mmap_pages_allocated;
David Howells7e660872010-01-15 17:01:39 -08002106extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
David Howells8feae132009-01-08 12:04:47 +00002107
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002108/* interval_tree.c */
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002109void vma_interval_tree_insert(struct vm_area_struct *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002110 struct rb_root_cached *root);
Michel Lespinasse9826a512012-10-08 16:31:35 -07002111void vma_interval_tree_insert_after(struct vm_area_struct *node,
2112 struct vm_area_struct *prev,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002113 struct rb_root_cached *root);
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002114void vma_interval_tree_remove(struct vm_area_struct *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002115 struct rb_root_cached *root);
2116struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002117 unsigned long start, unsigned long last);
2118struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2119 unsigned long start, unsigned long last);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002121#define vma_interval_tree_foreach(vma, root, start, last) \
2122 for (vma = vma_interval_tree_iter_first(root, start, last); \
2123 vma; vma = vma_interval_tree_iter_next(vma, start, last))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124
Michel Lespinassebf181b92012-10-08 16:31:39 -07002125void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002126 struct rb_root_cached *root);
Michel Lespinassebf181b92012-10-08 16:31:39 -07002127void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002128 struct rb_root_cached *root);
2129struct anon_vma_chain *
2130anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2131 unsigned long start, unsigned long last);
Michel Lespinassebf181b92012-10-08 16:31:39 -07002132struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2133 struct anon_vma_chain *node, unsigned long start, unsigned long last);
Michel Lespinasseed8ea812012-10-08 16:31:45 -07002134#ifdef CONFIG_DEBUG_VM_RB
2135void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2136#endif
Michel Lespinassebf181b92012-10-08 16:31:39 -07002137
2138#define anon_vma_interval_tree_foreach(avc, root, start, last) \
2139 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2140 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2141
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07002143extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Andrea Arcangelie86f15e2016-10-07 17:01:28 -07002144extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2145 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2146 struct vm_area_struct *expand);
2147static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2148 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2149{
2150 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2151}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152extern struct vm_area_struct *vma_merge(struct mm_struct *,
2153 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2154 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
Andrea Arcangeli19a809a2015-09-04 15:46:24 -07002155 struct mempolicy *, struct vm_userfaultfd_ctx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
David Rientjesdef5efe2017-02-24 14:58:47 -08002157extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2158 unsigned long addr, int new_below);
2159extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2160 unsigned long addr, int new_below);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2162extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2163 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07002164extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
Michel Lespinasse38a76012012-10-08 16:31:50 -07002166 unsigned long addr, unsigned long len, pgoff_t pgoff,
2167 bool *need_rmap_locks);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002168extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07002169
Cyrill Gorcunov9c599022014-10-09 15:27:29 -07002170static inline int check_data_rlimit(unsigned long rlim,
2171 unsigned long new,
2172 unsigned long start,
2173 unsigned long end_data,
2174 unsigned long start_data)
2175{
2176 if (rlim < RLIM_INFINITY) {
2177 if (((new - start) + (end_data - start_data)) > rlim)
2178 return -ENOSPC;
2179 }
2180
2181 return 0;
2182}
2183
Andrea Arcangeli7906d002008-07-28 15:46:26 -07002184extern int mm_take_all_locks(struct mm_struct *mm);
2185extern void mm_drop_all_locks(struct mm_struct *mm);
2186
Jiri Slaby38646012011-05-26 16:25:46 -07002187extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2188extern struct file *get_mm_exe_file(struct mm_struct *mm);
Mateusz Guzikcd81a9172016-08-23 16:20:38 +02002189extern struct file *get_task_exe_file(struct task_struct *task);
Matt Helsley925d1c42008-04-29 01:01:36 -07002190
Konstantin Khlebnikov84638332016-01-14 15:22:07 -08002191extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2192extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2193
Dmitry Safonov2eefd872016-09-05 16:33:05 +03002194extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2195 const struct vm_special_mapping *sm);
Stefani Seibold3935ed62014-03-17 23:22:02 +01002196extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2197 unsigned long addr, unsigned long len,
Andy Lutomirskia62c34b2014-05-19 15:58:33 -07002198 unsigned long flags,
2199 const struct vm_special_mapping *spec);
2200/* This is an obsolete alternative to _install_special_mapping. */
Roland McGrathfa5dc222007-02-08 14:20:41 -08002201extern int install_special_mapping(struct mm_struct *mm,
2202 unsigned long addr, unsigned long len,
2203 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204
2205extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2206
Miklos Szeredi0165ab42007-07-15 23:38:26 -07002207extern unsigned long mmap_region(struct file *file, unsigned long addr,
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002208 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2209 struct list_head *uf);
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002210extern unsigned long do_mmap(struct file *file, unsigned long addr,
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08002211 unsigned long len, unsigned long prot, unsigned long flags,
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002212 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2213 struct list_head *uf);
2214extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2215 struct list_head *uf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002217static inline unsigned long
2218do_mmap_pgoff(struct file *file, unsigned long addr,
2219 unsigned long len, unsigned long prot, unsigned long flags,
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002220 unsigned long pgoff, unsigned long *populate,
2221 struct list_head *uf)
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002222{
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002223 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002224}
2225
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08002226#ifdef CONFIG_MMU
2227extern int __mm_populate(unsigned long addr, unsigned long len,
2228 int ignore_errors);
2229static inline void mm_populate(unsigned long addr, unsigned long len)
2230{
2231 /* Ignore errors */
2232 (void) __mm_populate(addr, len, 1);
2233}
2234#else
2235static inline void mm_populate(unsigned long addr, unsigned long len) {}
2236#endif
2237
Linus Torvaldse4eb1ff2012-04-20 15:35:40 -07002238/* These take the mm semaphore themselves */
Linus Torvalds5d22fc22016-05-27 15:57:31 -07002239extern int __must_check vm_brk(unsigned long, unsigned long);
Denys Vlasenko16e72e92017-02-22 15:45:16 -08002240extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
Al Virobfce2812012-04-20 21:57:04 -04002241extern int vm_munmap(unsigned long, size_t);
Michal Hocko9fbeb5a2016-05-23 16:25:30 -07002242extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
Linus Torvalds6be5ceb2012-04-20 17:13:58 -07002243 unsigned long, unsigned long,
2244 unsigned long, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245
Michel Lespinassedb4fbfb2012-12-11 16:01:49 -08002246struct vm_unmapped_area_info {
2247#define VM_UNMAPPED_AREA_TOPDOWN 1
2248 unsigned long flags;
2249 unsigned long length;
2250 unsigned long low_limit;
2251 unsigned long high_limit;
2252 unsigned long align_mask;
2253 unsigned long align_offset;
2254};
2255
2256extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2257extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2258
2259/*
2260 * Search for an unmapped address range.
2261 *
2262 * We are looking for a range that:
2263 * - does not intersect with any VMA;
2264 * - is contained within the [low_limit, high_limit) interval;
2265 * - is at least the desired size.
2266 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2267 */
2268static inline unsigned long
2269vm_unmapped_area(struct vm_unmapped_area_info *info)
2270{
Borislav Petkovcdd78752015-04-15 16:14:47 -07002271 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
Michel Lespinassedb4fbfb2012-12-11 16:01:49 -08002272 return unmapped_area_topdown(info);
Borislav Petkovcdd78752015-04-15 16:14:47 -07002273 else
2274 return unmapped_area(info);
Michel Lespinassedb4fbfb2012-12-11 16:01:49 -08002275}
2276
Hugh Dickins85821aa2011-07-25 17:12:23 -07002277/* truncate.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08002279extern void truncate_inode_pages_range(struct address_space *,
2280 loff_t lstart, loff_t lend);
Johannes Weiner91b0abe2014-04-03 14:47:49 -07002281extern void truncate_inode_pages_final(struct address_space *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282
2283/* generic vm_area_ops exported for stackable file systems */
Dave Jiang11bac802017-02-24 14:56:41 -08002284extern int filemap_fault(struct vm_fault *vmf);
Jan Kara82b0f8c2016-12-14 15:06:58 -08002285extern void filemap_map_pages(struct vm_fault *vmf,
Kirill A. Shutemovbae473a2016-07-26 15:25:20 -07002286 pgoff_t start_pgoff, pgoff_t end_pgoff);
Dave Jiang11bac802017-02-24 14:56:41 -08002287extern int filemap_page_mkwrite(struct vm_fault *vmf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288
2289/* mm/page-writeback.c */
Jeff Layton2b69c822017-07-05 15:26:48 -04002290int __must_check write_one_page(struct page *page);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08002291void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292
2293/* readahead.c */
2294#define VM_MAX_READAHEAD 128 /* kbytes */
2295#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08002298 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07002299
2300void page_cache_sync_readahead(struct address_space *mapping,
2301 struct file_ra_state *ra,
2302 struct file *filp,
2303 pgoff_t offset,
2304 unsigned long size);
2305
2306void page_cache_async_readahead(struct address_space *mapping,
2307 struct file_ra_state *ra,
2308 struct file *filp,
2309 struct page *pg,
2310 pgoff_t offset,
2311 unsigned long size);
2312
Hugh Dickins1be71072017-06-19 04:03:24 -07002313extern unsigned long stack_guard_gap;
Michal Hockod05f3162011-05-24 17:11:44 -07002314/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07002315extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Michal Hockod05f3162011-05-24 17:11:44 -07002316
2317/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2318extern int expand_downwards(struct vm_area_struct *vma,
2319 unsigned long address);
Luck, Tony8ca3eb082010-08-24 11:44:18 -07002320#if VM_GROWSUP
Hugh Dickins46dea3d2005-10-29 18:16:20 -07002321extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Luck, Tony8ca3eb082010-08-24 11:44:18 -07002322#else
Linus Torvaldsfee7e492015-01-06 13:00:05 -08002323 #define expand_upwards(vma, address) (0)
Matthew Wilcox9ab88512005-11-18 16:16:42 -05002324#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325
2326/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2327extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2328extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2329 struct vm_area_struct **pprev);
2330
2331/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2332 NULL if none. Assume start_addr < end_addr. */
2333static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2334{
2335 struct vm_area_struct * vma = find_vma(mm,start_addr);
2336
2337 if (vma && end_addr <= vma->vm_start)
2338 vma = NULL;
2339 return vma;
2340}
2341
Hugh Dickins1be71072017-06-19 04:03:24 -07002342static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2343{
2344 unsigned long vm_start = vma->vm_start;
2345
2346 if (vma->vm_flags & VM_GROWSDOWN) {
2347 vm_start -= stack_guard_gap;
2348 if (vm_start > vma->vm_start)
2349 vm_start = 0;
2350 }
2351 return vm_start;
2352}
2353
2354static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2355{
2356 unsigned long vm_end = vma->vm_end;
2357
2358 if (vma->vm_flags & VM_GROWSUP) {
2359 vm_end += stack_guard_gap;
2360 if (vm_end < vma->vm_end)
2361 vm_end = -PAGE_SIZE;
2362 }
2363 return vm_end;
2364}
2365
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366static inline unsigned long vma_pages(struct vm_area_struct *vma)
2367{
2368 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2369}
2370
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002371/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2372static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2373 unsigned long vm_start, unsigned long vm_end)
2374{
2375 struct vm_area_struct *vma = find_vma(mm, vm_start);
2376
2377 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2378 vma = NULL;
2379
2380 return vma;
2381}
2382
David Howellsbad849b2010-08-26 16:00:34 +01002383#ifdef CONFIG_MMU
Hugh Dickins804af2c2006-07-26 21:39:49 +01002384pgprot_t vm_get_page_prot(unsigned long vm_flags);
Peter Feiner64e455072014-10-13 15:55:46 -07002385void vma_set_page_prot(struct vm_area_struct *vma);
David Howellsbad849b2010-08-26 16:00:34 +01002386#else
2387static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2388{
2389 return __pgprot(0);
2390}
Peter Feiner64e455072014-10-13 15:55:46 -07002391static inline void vma_set_page_prot(struct vm_area_struct *vma)
2392{
2393 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2394}
David Howellsbad849b2010-08-26 16:00:34 +01002395#endif
2396
Aneesh Kumar K.V58772312013-12-06 00:08:22 +05302397#ifdef CONFIG_NUMA_BALANCING
Mel Gorman4b10e7d2012-10-25 14:16:32 +02002398unsigned long change_prot_numa(struct vm_area_struct *vma,
Lee Schermerhornb24f53a2012-10-25 14:16:32 +02002399 unsigned long start, unsigned long end);
2400#endif
2401
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07002402struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07002403int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2404 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08002405int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08002406int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2407 unsigned long pfn);
Andy Lutomirski1745cbc2015-12-29 20:12:20 -08002408int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2409 unsigned long pfn, pgprot_t pgprot);
Nick Piggin423bad602008-04-28 02:13:01 -07002410int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
Dan Williams01c8f1c2016-01-15 16:56:40 -08002411 pfn_t pfn);
Ross Zwislerb2770da62017-09-06 16:18:35 -07002412int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2413 pfn_t pfn);
Linus Torvaldsb4cbb192013-04-16 13:45:37 -07002414int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2415
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416
Michel Lespinasse240aade2013-02-22 16:35:56 -08002417struct page *follow_page_mask(struct vm_area_struct *vma,
2418 unsigned long address, unsigned int foll_flags,
2419 unsigned int *page_mask);
2420
2421static inline struct page *follow_page(struct vm_area_struct *vma,
2422 unsigned long address, unsigned int foll_flags)
2423{
2424 unsigned int unused_page_mask;
2425 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2426}
2427
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07002428#define FOLL_WRITE 0x01 /* check pte is writable */
2429#define FOLL_TOUCH 0x02 /* mark page accessed */
2430#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a62009-09-21 17:03:26 -07002431#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07002432#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
Gleb Natapov318b2752011-03-22 16:30:51 -07002433#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2434 * and return without waiting upon it */
Kirill A. Shutemov84d33df2015-04-14 15:44:37 -07002435#define FOLL_POPULATE 0x40 /* fault in page */
Andrea Arcangeli500d65d2011-01-13 15:46:55 -08002436#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
Huang Ying69ebb832011-01-30 11:15:48 +08002437#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
Andrea Arcangeli0b9d7052012-10-05 21:36:27 +02002438#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
Hugh Dickins5117b3b2013-02-22 16:36:07 -08002439#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
Andres Lagar-Cavilla234b2392014-09-17 10:51:48 -07002440#define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
Eric B Munsonde60f5f2015-11-05 18:51:36 -08002441#define FOLL_MLOCK 0x1000 /* lock present pages */
Dave Hansen1e987792016-02-12 13:01:54 -08002442#define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
Linus Torvalds19be0ea2016-10-13 13:07:36 -07002443#define FOLL_COW 0x4000 /* internal GUP flag */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444
James Morse9a291a72017-06-02 14:46:46 -07002445static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2446{
2447 if (vm_fault & VM_FAULT_OOM)
2448 return -ENOMEM;
2449 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2450 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2451 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2452 return -EFAULT;
2453 return 0;
2454}
2455
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08002456typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07002457 void *data);
2458extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2459 unsigned long size, pte_fn_t fn, void *data);
2460
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461
Laura Abbott8823b1d2016-03-15 14:56:27 -07002462#ifdef CONFIG_PAGE_POISONING
2463extern bool page_poisoning_enabled(void);
2464extern void kernel_poison_pages(struct page *page, int numpages, int enable);
Laura Abbott1414c7f2016-03-15 14:56:30 -07002465extern bool page_is_poisoned(struct page *page);
Laura Abbott8823b1d2016-03-15 14:56:27 -07002466#else
2467static inline bool page_poisoning_enabled(void) { return false; }
2468static inline void kernel_poison_pages(struct page *page, int numpages,
2469 int enable) { }
Laura Abbott1414c7f2016-03-15 14:56:30 -07002470static inline bool page_is_poisoned(struct page *page) { return false; }
Laura Abbott8823b1d2016-03-15 14:56:27 -07002471#endif
2472
Ingo Molnar12d6f212008-01-30 13:33:58 +01002473#ifdef CONFIG_DEBUG_PAGEALLOC
Joonsoo Kim031bc572014-12-12 16:55:52 -08002474extern bool _debug_pagealloc_enabled;
2475extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2476
2477static inline bool debug_pagealloc_enabled(void)
2478{
2479 return _debug_pagealloc_enabled;
2480}
2481
2482static inline void
2483kernel_map_pages(struct page *page, int numpages, int enable)
2484{
2485 if (!debug_pagealloc_enabled())
2486 return;
2487
2488 __kernel_map_pages(page, numpages, enable);
2489}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01002490#ifdef CONFIG_HIBERNATION
2491extern bool kernel_page_present(struct page *page);
Joonsoo Kim40b44132016-03-15 14:54:21 -07002492#endif /* CONFIG_HIBERNATION */
2493#else /* CONFIG_DEBUG_PAGEALLOC */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002494static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07002495kernel_map_pages(struct page *page, int numpages, int enable) {}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01002496#ifdef CONFIG_HIBERNATION
2497static inline bool kernel_page_present(struct page *page) { return true; }
Joonsoo Kim40b44132016-03-15 14:54:21 -07002498#endif /* CONFIG_HIBERNATION */
2499static inline bool debug_pagealloc_enabled(void)
2500{
2501 return false;
2502}
2503#endif /* CONFIG_DEBUG_PAGEALLOC */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002504
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002505#ifdef __HAVE_ARCH_GATE_AREA
Stephen Wilson31db58b2011-03-13 15:49:15 -04002506extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002507extern int in_gate_area_no_mm(unsigned long addr);
2508extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509#else
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002510static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2511{
2512 return NULL;
2513}
2514static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2515static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2516{
2517 return 0;
2518}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002519#endif /* __HAVE_ARCH_GATE_AREA */
2520
Michal Hocko44a70ade2016-07-28 15:44:43 -07002521extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2522
Josh Triplett146732c2013-04-29 15:07:22 -07002523#ifdef CONFIG_SYSCTL
2524extern int sysctl_drop_caches;
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07002525int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08002526 void __user *, size_t *, loff_t *);
Josh Triplett146732c2013-04-29 15:07:22 -07002527#endif
2528
Vladimir Davydovcb731d62015-02-12 14:58:54 -08002529void drop_slab(void);
2530void drop_slab_node(int nid);
Andrew Morton9d0243b2006-01-08 01:00:39 -08002531
Luke Yang7a9166e2006-02-20 18:28:07 -08002532#ifndef CONFIG_MMU
2533#define randomize_va_space 0
2534#else
Andi Kleena62eaf12006-02-16 23:41:58 +01002535extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08002536#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01002537
Sam Ravnborg045e72a2007-07-26 10:41:13 -07002538const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01002539void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07002540
Yinghai Lu9bdac912010-02-10 01:20:22 -08002541void sparse_mem_maps_populate_node(struct page **map_map,
2542 unsigned long pnum_begin,
2543 unsigned long pnum_end,
2544 unsigned long map_count,
2545 int nodeid);
2546
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07002547struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07002548pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03002549p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2550pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
Andy Whitcroft29c71112007-10-16 01:24:14 -07002551pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2552pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07002553void *vmemmap_alloc_block(unsigned long size, int node);
Dan Williams4b94ffd2016-01-15 16:56:22 -08002554struct vmem_altmap;
2555void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2556 struct vmem_altmap *altmap);
2557static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2558{
2559 return __vmemmap_alloc_block_buf(size, node, NULL);
2560}
2561
Christoph Lameter8f6aac42007-10-16 01:24:13 -07002562void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Johannes Weiner0aad8182013-04-29 15:07:50 -07002563int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2564 int node);
2565int vmemmap_populate(unsigned long start, unsigned long end, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07002566void vmemmap_populate_print_last(void);
Tang Chen01975182013-02-22 16:33:08 -08002567#ifdef CONFIG_MEMORY_HOTPLUG
Johannes Weiner0aad8182013-04-29 15:07:50 -07002568void vmemmap_free(unsigned long start, unsigned long end);
Tang Chen01975182013-02-22 16:33:08 -08002569#endif
Yasuaki Ishimatsu46723bf2013-02-22 16:33:00 -08002570void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
Baoquan He15670bf2017-10-28 09:30:38 +08002571 unsigned long nr_pages);
Andi Kleen6a460792009-09-16 11:50:15 +02002572
Andi Kleen82ba0112009-12-16 12:19:57 +01002573enum mf_flags {
2574 MF_COUNT_INCREASED = 1 << 0,
Tony Luck7329bbe2011-12-13 09:27:58 -08002575 MF_ACTION_REQUIRED = 1 << 1,
Tony Luck6751ed62012-07-11 10:20:47 -07002576 MF_MUST_KILL = 1 << 2,
Naveen N. Raocf870c72013-07-10 14:57:01 +05302577 MF_SOFT_OFFLINE = 1 << 3,
Andi Kleen82ba0112009-12-16 12:19:57 +01002578};
Eric W. Biederman83b57532017-07-09 18:14:01 -05002579extern int memory_failure(unsigned long pfn, int flags);
2580extern void memory_failure_queue(unsigned long pfn, int flags);
Wu Fengguang847ce402009-12-16 12:19:58 +01002581extern int unpoison_memory(unsigned long pfn);
Naoya Horiguchiead07f62015-06-24 16:56:48 -07002582extern int get_hwpoison_page(struct page *page);
Naoya Horiguchi4e41a302016-01-15 16:54:07 -08002583#define put_hwpoison_page(page) put_page(page)
Andi Kleen6a460792009-09-16 11:50:15 +02002584extern int sysctl_memory_failure_early_kill;
2585extern int sysctl_memory_failure_recovery;
Andi Kleenfacb6012009-12-16 12:20:00 +01002586extern void shake_page(struct page *p, int access);
Xishi Qiu293c07e2013-02-22 16:34:02 -08002587extern atomic_long_t num_poisoned_pages;
Andi Kleenfacb6012009-12-16 12:20:00 +01002588extern int soft_offline_page(struct page *page, int flags);
Andi Kleen6a460792009-09-16 11:50:15 +02002589
Xie XiuQicc637b12015-06-24 16:57:30 -07002590
2591/*
2592 * Error handlers for various types of pages.
2593 */
Xie XiuQicc3e2af2015-06-24 16:57:33 -07002594enum mf_result {
Xie XiuQicc637b12015-06-24 16:57:30 -07002595 MF_IGNORED, /* Error: cannot be handled */
2596 MF_FAILED, /* Error: handling failed */
2597 MF_DELAYED, /* Will be handled later */
2598 MF_RECOVERED, /* Successfully recovered */
2599};
2600
2601enum mf_action_page_type {
2602 MF_MSG_KERNEL,
2603 MF_MSG_KERNEL_HIGH_ORDER,
2604 MF_MSG_SLAB,
2605 MF_MSG_DIFFERENT_COMPOUND,
2606 MF_MSG_POISONED_HUGE,
2607 MF_MSG_HUGE,
2608 MF_MSG_FREE_HUGE,
2609 MF_MSG_UNMAP_FAILED,
2610 MF_MSG_DIRTY_SWAPCACHE,
2611 MF_MSG_CLEAN_SWAPCACHE,
2612 MF_MSG_DIRTY_MLOCKED_LRU,
2613 MF_MSG_CLEAN_MLOCKED_LRU,
2614 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2615 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2616 MF_MSG_DIRTY_LRU,
2617 MF_MSG_CLEAN_LRU,
2618 MF_MSG_TRUNCATED_LRU,
2619 MF_MSG_BUDDY,
2620 MF_MSG_BUDDY_2ND,
2621 MF_MSG_UNKNOWN,
2622};
2623
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08002624#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2625extern void clear_huge_page(struct page *page,
Huang Yingc79b57e2017-09-06 16:25:04 -07002626 unsigned long addr_hint,
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08002627 unsigned int pages_per_huge_page);
2628extern void copy_user_huge_page(struct page *dst, struct page *src,
2629 unsigned long addr, struct vm_area_struct *vma,
2630 unsigned int pages_per_huge_page);
Mike Kravetzfa4d75c2017-02-22 15:42:49 -08002631extern long copy_huge_page_from_user(struct page *dst_page,
2632 const void __user *usr_src,
Mike Kravetz810a56b2017-02-22 15:42:58 -08002633 unsigned int pages_per_huge_page,
2634 bool allow_pagefault);
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08002635#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2636
Joonsoo Kime30825f2014-12-12 16:55:49 -08002637extern struct page_ext_operations debug_guardpage_ops;
Joonsoo Kime30825f2014-12-12 16:55:49 -08002638
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002639#ifdef CONFIG_DEBUG_PAGEALLOC
2640extern unsigned int _debug_guardpage_minorder;
Joonsoo Kime30825f2014-12-12 16:55:49 -08002641extern bool _debug_guardpage_enabled;
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002642
2643static inline unsigned int debug_guardpage_minorder(void)
2644{
2645 return _debug_guardpage_minorder;
2646}
2647
Joonsoo Kime30825f2014-12-12 16:55:49 -08002648static inline bool debug_guardpage_enabled(void)
2649{
2650 return _debug_guardpage_enabled;
2651}
2652
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002653static inline bool page_is_guard(struct page *page)
2654{
Joonsoo Kime30825f2014-12-12 16:55:49 -08002655 struct page_ext *page_ext;
2656
2657 if (!debug_guardpage_enabled())
2658 return false;
2659
2660 page_ext = lookup_page_ext(page);
Yang Shi0bb2fd132016-05-20 16:58:59 -07002661 if (unlikely(!page_ext))
2662 return false;
2663
Joonsoo Kime30825f2014-12-12 16:55:49 -08002664 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002665}
2666#else
2667static inline unsigned int debug_guardpage_minorder(void) { return 0; }
Joonsoo Kime30825f2014-12-12 16:55:49 -08002668static inline bool debug_guardpage_enabled(void) { return false; }
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002669static inline bool page_is_guard(struct page *page) { return false; }
2670#endif /* CONFIG_DEBUG_PAGEALLOC */
2671
Cody P Schaferf9872ca2013-04-29 15:08:01 -07002672#if MAX_NUMNODES > 1
2673void __init setup_nr_node_ids(void);
2674#else
2675static inline void setup_nr_node_ids(void) {}
2676#endif
2677
Linus Torvalds1da177e2005-04-16 15:20:36 -07002678#endif /* __KERNEL__ */
2679#endif /* _LINUX_MM_H */