blob: c7b1d617dff6851b121539610981a491a4d0d415 [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 */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700202#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
Rik van Rield2cd9ed2017-09-06 16:25:15 -0700203#define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
Konstantin Khlebnikov0103bd12012-10-08 16:28:59 -0700204#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700206#ifdef CONFIG_MEM_SOFT_DIRTY
207# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
208#else
209# define VM_SOFTDIRTY 0
210#endif
211
Jared Hulbertb379d792008-04-28 02:12:58 -0700212#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700213#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
214#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700215#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
Nick Piggind00806b2007-07-19 01:46:57 -0700216
Dave Hansen63c17fb2016-02-12 13:02:08 -0800217#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
218#define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
219#define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
220#define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
221#define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
Rik van Rieldf3735c2017-09-06 16:25:11 -0700222#define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
Dave Hansen63c17fb2016-02-12 13:02:08 -0800223#define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
224#define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
225#define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
226#define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
Rik van Rieldf3735c2017-09-06 16:25:11 -0700227#define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
Dave Hansen63c17fb2016-02-12 13:02:08 -0800228#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
229
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700230#if defined(CONFIG_X86)
231# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
Dave Hansen8f62c882016-02-12 13:02:10 -0800232#if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
233# define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
234# define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
235# define VM_PKEY_BIT1 VM_HIGH_ARCH_1
236# define VM_PKEY_BIT2 VM_HIGH_ARCH_2
237# define VM_PKEY_BIT3 VM_HIGH_ARCH_3
238#endif
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700239#elif defined(CONFIG_PPC)
240# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
241#elif defined(CONFIG_PARISC)
242# define VM_GROWSUP VM_ARCH_1
James Hogan9ca52ed2012-10-16 10:16:14 +0100243#elif defined(CONFIG_METAG)
244# define VM_GROWSUP VM_ARCH_1
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700245#elif defined(CONFIG_IA64)
246# define VM_GROWSUP VM_ARCH_1
247#elif !defined(CONFIG_MMU)
248# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
249#endif
250
Rik van Rieldf3735c2017-09-06 16:25:11 -0700251#if defined(CONFIG_X86_INTEL_MPX)
Qiaowei Ren4aae7e42014-11-14 07:18:25 -0800252/* MPX specific bounds table or bounds directory */
Kirill A. Shutemovfa87b912017-10-03 16:14:24 -0700253# define VM_MPX VM_HIGH_ARCH_4
Rik van Rieldf3735c2017-09-06 16:25:11 -0700254#else
255# define VM_MPX VM_NONE
Qiaowei Ren4aae7e42014-11-14 07:18:25 -0800256#endif
257
Konstantin Khlebnikovcc2383e2012-10-08 16:28:37 -0700258#ifndef VM_GROWSUP
259# define VM_GROWSUP VM_NONE
260#endif
261
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700262/* Bits set in the VMA until the stack is in its final location */
263#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
264
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
266#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
267#endif
268
269#ifdef CONFIG_STACK_GROWSUP
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800270#define VM_STACK VM_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271#else
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800272#define VM_STACK VM_GROWSDOWN
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273#endif
274
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800275#define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277/*
Andrea Arcangeli78f11a22011-04-27 15:26:45 -0700278 * Special vmas that are non-mergable, non-mlock()able.
279 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
Nick Pigginb291f002008-10-18 20:26:44 -0700280 */
Vlastimil Babka9050d7e2014-03-03 15:38:27 -0800281#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
Nick Pigginb291f002008-10-18 20:26:44 -0700282
Alex Thorltona0715cc2014-04-07 15:37:10 -0700283/* This mask defines which mm->def_flags a process can inherit its parent */
284#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
285
Eric B Munsonde60f5f2015-11-05 18:51:36 -0800286/* This mask is used to clear all the VMA flags used by mlock */
287#define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
288
Nick Pigginb291f002008-10-18 20:26:44 -0700289/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 * mapping from the currently active vm_flags protection bits (the
291 * low four bits) to a page protection mask..
292 */
293extern pgprot_t protection_map[16];
294
Nick Piggind0217ac042007-07-19 01:47:03 -0700295#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
Kirill A. Shutemov9b4bdd22015-02-10 14:09:51 -0800296#define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
297#define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
298#define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
299#define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
300#define FAULT_FLAG_TRIED 0x20 /* Second try */
301#define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
Dave Hansen1b2ee122016-02-12 13:02:21 -0800302#define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
Dave Hansend61172b2016-02-12 13:02:24 -0800303#define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
Nick Piggind0217ac042007-07-19 01:47:03 -0700304
Ross Zwisler282a8e02017-02-22 15:39:50 -0800305#define FAULT_FLAG_TRACE \
306 { FAULT_FLAG_WRITE, "WRITE" }, \
307 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
308 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
309 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
310 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
311 { FAULT_FLAG_TRIED, "TRIED" }, \
312 { FAULT_FLAG_USER, "USER" }, \
313 { FAULT_FLAG_REMOTE, "REMOTE" }, \
314 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
315
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800316/*
Nick Piggind0217ac042007-07-19 01:47:03 -0700317 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700318 * ->fault function. The vma's ->fault is responsible for returning a bitmask
319 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700320 *
Michal Hockoc20cd452016-01-14 15:20:12 -0800321 * MM layer fills up gfp_mask for page allocations but fault handler might
322 * alter it if its implementation requires a different allocation context.
323 *
Kirill A. Shutemov9b4bdd22015-02-10 14:09:51 -0800324 * pgoff should be used in favour of virtual_address, if possible.
Nick Piggin54cb8822007-07-19 01:46:59 -0700325 */
Nick Piggind0217ac042007-07-19 01:47:03 -0700326struct vm_fault {
Jan Kara82b0f8c2016-12-14 15:06:58 -0800327 struct vm_area_struct *vma; /* Target VMA */
Nick Piggind0217ac042007-07-19 01:47:03 -0700328 unsigned int flags; /* FAULT_FLAG_xxx flags */
Michal Hockoc20cd452016-01-14 15:20:12 -0800329 gfp_t gfp_mask; /* gfp mask to be used for allocations */
Nick Piggind0217ac042007-07-19 01:47:03 -0700330 pgoff_t pgoff; /* Logical page offset based on vma */
Jan Kara82b0f8c2016-12-14 15:06:58 -0800331 unsigned long address; /* Faulting virtual address */
Jan Kara82b0f8c2016-12-14 15:06:58 -0800332 pmd_t *pmd; /* Pointer to pmd entry matching
Jan Kara29943022016-12-14 15:07:16 -0800333 * the 'address' */
Dave Jianga2d58162017-02-24 14:56:59 -0800334 pud_t *pud; /* Pointer to pud entry matching
335 * the 'address'
336 */
Jan Kara29943022016-12-14 15:07:16 -0800337 pte_t orig_pte; /* Value of PTE at the time of fault */
Nick Piggind0217ac042007-07-19 01:47:03 -0700338
Jan Kara39170482016-12-14 15:07:18 -0800339 struct page *cow_page; /* Page handler may use for COW fault */
340 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
Nick Piggind0217ac042007-07-19 01:47:03 -0700341 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700342 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac042007-07-19 01:47:03 -0700343 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700344 * VM_FAULT_ERROR).
Nick Piggind0217ac042007-07-19 01:47:03 -0700345 */
Jan Kara82b0f8c2016-12-14 15:06:58 -0800346 /* These three entries are valid only while holding ptl lock */
Kirill A. Shutemovbae473a2016-07-26 15:25:20 -0700347 pte_t *pte; /* Pointer to pte entry matching
348 * the 'address'. NULL if the page
349 * table hasn't been allocated.
350 */
351 spinlock_t *ptl; /* Page table lock.
352 * Protects pte page table if 'pte'
353 * is not NULL, otherwise pmd.
354 */
Kirill A. Shutemov7267ec002016-07-26 15:25:23 -0700355 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
356 * vm_ops->map_pages() calls
357 * alloc_set_pte() from atomic context.
358 * do_fault_around() pre-allocates
359 * page table to avoid allocation from
360 * atomic context.
361 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700362};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363
Dave Jiangc791ace2017-02-24 14:57:08 -0800364/* page entry size for vm->huge_fault() */
365enum page_entry_size {
366 PE_SIZE_PTE = 0,
367 PE_SIZE_PMD,
368 PE_SIZE_PUD,
369};
370
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371/*
372 * These are the virtual MM functions - opening of an area, closing and
373 * unmapping it (needed to keep files on disk up-to-date etc), pointer
374 * to the functions called when a no-page or a wp-page exception occurs.
375 */
376struct vm_operations_struct {
377 void (*open)(struct vm_area_struct * area);
378 void (*close)(struct vm_area_struct * area);
Oleg Nesterov5477e702015-09-04 15:48:04 -0700379 int (*mremap)(struct vm_area_struct * area);
Dave Jiang11bac802017-02-24 14:56:41 -0800380 int (*fault)(struct vm_fault *vmf);
Dave Jiangc791ace2017-02-24 14:57:08 -0800381 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
Jan Kara82b0f8c2016-12-14 15:06:58 -0800382 void (*map_pages)(struct vm_fault *vmf,
Kirill A. Shutemovbae473a2016-07-26 15:25:20 -0700383 pgoff_t start_pgoff, pgoff_t end_pgoff);
David Howells9637a5e2006-06-23 02:03:43 -0700384
385 /* notification that a previously read-only page is about to become
386 * writable, if an error is returned it will cause a SIGBUS */
Dave Jiang11bac802017-02-24 14:56:41 -0800387 int (*page_mkwrite)(struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700388
Boaz Harroshdd906182015-04-15 16:15:11 -0700389 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
Dave Jiang11bac802017-02-24 14:56:41 -0800390 int (*pfn_mkwrite)(struct vm_fault *vmf);
Boaz Harroshdd906182015-04-15 16:15:11 -0700391
Rik van Riel28b2ee22008-07-23 21:27:05 -0700392 /* called by access_process_vm when get_user_pages() fails, typically
393 * for use by special VMAs that can switch between memory and hardware
394 */
395 int (*access)(struct vm_area_struct *vma, unsigned long addr,
396 void *buf, int len, int write);
Andy Lutomirski78d683e2014-05-19 15:58:32 -0700397
398 /* Called by the /proc/PID/maps code to ask the vma whether it
399 * has a special name. Returning non-NULL will also cause this
400 * vma to be dumped unconditionally. */
401 const char *(*name)(struct vm_area_struct *vma);
402
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700404 /*
405 * set_policy() op must add a reference to any non-NULL @new mempolicy
406 * to hold the policy upon return. Caller should pass NULL @new to
407 * remove a policy and fall back to surrounding context--i.e. do not
408 * install a MPOL_DEFAULT policy, nor the task or system default
409 * mempolicy.
410 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700412
413 /*
414 * get_policy() op must add reference [mpol_get()] to any policy at
415 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
416 * in mm/mempolicy.c will do this automatically.
417 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
418 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
419 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
420 * must return NULL--i.e., do not "fallback" to task or system default
421 * policy.
422 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
424 unsigned long addr);
425#endif
David Vrabel667a0a02014-12-18 14:48:15 +0000426 /*
427 * Called by vm_normal_page() for special PTEs to find the
428 * page for @addr. This is useful if the default behavior
429 * (using pte_page()) would not find the correct page.
430 */
431 struct page *(*find_special_page)(struct vm_area_struct *vma,
432 unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433};
434
435struct mmu_gather;
436struct inode;
437
Andrew Morton349aef02006-01-08 01:04:36 -0800438#define page_private(page) ((page)->private)
439#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700440
Dan Williams5c7fb562016-01-15 16:56:52 -0800441#if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
442static inline int pmd_devmap(pmd_t pmd)
443{
444 return 0;
445}
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -0800446static inline int pud_devmap(pud_t pud)
447{
448 return 0;
449}
Kirill A. Shutemovb59f65f2017-03-16 18:26:53 +0300450static inline int pgd_devmap(pgd_t pgd)
451{
452 return 0;
453}
Dan Williams5c7fb562016-01-15 16:56:52 -0800454#endif
455
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456/*
457 * FIXME: take this include out, include page-flags.h in
458 * files which need it (119 of them)
459 */
460#include <linux/page-flags.h>
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800461#include <linux/huge_mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462
463/*
464 * Methods to modify the page usage count.
465 *
466 * What counts for a page usage:
467 * - cache mapping (page->mapping)
468 * - private data (page->private)
469 * - page mapped in a task's page tables, each mapping
470 * is counted separately
471 *
472 * Also, many kernel routines increase the page count before a critical
473 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 */
475
476/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700477 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800479static inline int put_page_testzero(struct page *page)
480{
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700481 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
482 return page_ref_dec_and_test(page);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800483}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800486 * Try to grab a ref unless the page has a refcount of zero, return false if
487 * that is the case.
Alexey Kardashevskiy8e0861fa2013-08-28 18:37:42 +1000488 * This can be called when MMU is off so it must not access
489 * any of the virtual mappings.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800491static inline int get_page_unless_zero(struct page *page)
492{
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700493 return page_ref_add_unless(page, 1, 0);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800494}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800496extern int page_is_ram(unsigned long pfn);
Dan Williams124fe202015-08-10 23:07:05 -0400497
498enum {
499 REGION_INTERSECTS,
500 REGION_DISJOINT,
501 REGION_MIXED,
502};
503
Toshi Kani1c29f252016-01-26 21:57:28 +0100504int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
505 unsigned long desc);
Wu Fengguang53df8fd2010-01-27 11:06:39 +0800506
Christoph Lameter48667e72008-02-04 22:28:31 -0800507/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800508struct page *vmalloc_to_page(const void *addr);
509unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800510
Paul Mundt0738c4b2008-03-12 16:51:31 +0900511/*
512 * Determine if an address is within the vmalloc range
513 *
514 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
515 * is no special casing required.
516 */
Yaowei Baibb00a782016-05-19 17:11:29 -0700517static inline bool is_vmalloc_addr(const void *x)
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800518{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900519#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800520 unsigned long addr = (unsigned long)x;
521
522 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900523#else
Yaowei Baibb00a782016-05-19 17:11:29 -0700524 return false;
David Howells8ca3ed82008-02-23 15:23:37 -0800525#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900526}
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700527#ifdef CONFIG_MMU
528extern int is_vmalloc_or_module_addr(const void *x);
529#else
David Howells934831d2009-09-24 12:33:32 +0100530static inline int is_vmalloc_or_module_addr(const void *x)
KAMEZAWA Hiroyuki81ac3ad2009-09-22 16:45:49 -0700531{
532 return 0;
533}
534#endif
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800535
Michal Hockoa7c3e902017-05-08 15:57:09 -0700536extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
537static inline void *kvmalloc(size_t size, gfp_t flags)
538{
539 return kvmalloc_node(size, flags, NUMA_NO_NODE);
540}
541static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
542{
543 return kvmalloc_node(size, flags | __GFP_ZERO, node);
544}
545static inline void *kvzalloc(size_t size, gfp_t flags)
546{
547 return kvmalloc(size, flags | __GFP_ZERO);
548}
549
Michal Hocko752ade62017-05-08 15:57:27 -0700550static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
551{
552 if (size != 0 && n > SIZE_MAX / size)
553 return NULL;
554
555 return kvmalloc(n * size, flags);
556}
557
Al Viro39f1f782014-05-06 14:02:53 -0400558extern void kvfree(const void *addr);
559
Kirill A. Shutemov53f92632016-01-15 16:53:42 -0800560static inline atomic_t *compound_mapcount_ptr(struct page *page)
561{
562 return &page[1].compound_mapcount;
563}
564
565static inline int compound_mapcount(struct page *page)
566{
Andrea Arcangeli5f527c22016-05-20 16:58:24 -0700567 VM_BUG_ON_PAGE(!PageCompound(page), page);
Kirill A. Shutemov53f92632016-01-15 16:53:42 -0800568 page = compound_head(page);
569 return atomic_read(compound_mapcount_ptr(page)) + 1;
570}
571
Joonsoo Kimccaafd72015-02-10 14:09:35 -0800572/*
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700573 * The atomic page->_mapcount, starts from -1: so that transitions
574 * both from it and to it can be tracked, using atomic_inc_and_test
575 * and atomic_add_negative(-1).
576 */
Mel Gorman22b751c2013-02-22 16:34:59 -0800577static inline void page_mapcount_reset(struct page *page)
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700578{
579 atomic_set(&(page)->_mapcount, -1);
580}
581
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800582int __page_mapcount(struct page *page);
583
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700584static inline int page_mapcount(struct page *page)
585{
Wang, Yalin1d148e22015-02-11 15:24:48 -0800586 VM_BUG_ON_PAGE(PageSlab(page), page);
Kirill A. Shutemov53f92632016-01-15 16:53:42 -0800587
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800588 if (unlikely(PageCompound(page)))
589 return __page_mapcount(page);
590 return atomic_read(&page->_mapcount) + 1;
Andrea Arcangeli70b50f92011-11-02 13:36:59 -0700591}
592
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800593#ifdef CONFIG_TRANSPARENT_HUGEPAGE
594int total_mapcount(struct page *page);
Andrea Arcangeli6d0a07e2016-05-12 15:42:25 -0700595int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800596#else
597static inline int total_mapcount(struct page *page)
598{
599 return page_mapcount(page);
600}
Andrea Arcangeli6d0a07e2016-05-12 15:42:25 -0700601static inline int page_trans_huge_mapcount(struct page *page,
602 int *total_mapcount)
603{
604 int mapcount = page_mapcount(page);
605 if (total_mapcount)
606 *total_mapcount = mapcount;
607 return mapcount;
608}
Kirill A. Shutemovb20ce5e2016-01-15 16:54:37 -0800609#endif
610
Christoph Lameterb49af682007-05-06 14:49:41 -0700611static inline struct page *virt_to_head_page(const void *x)
612{
613 struct page *page = virt_to_page(x);
Joonsoo Kimccaafd72015-02-10 14:09:35 -0800614
Kirill A. Shutemov1d798ca2015-11-06 16:29:54 -0800615 return compound_head(page);
Christoph Lameterb49af682007-05-06 14:49:41 -0700616}
617
Kirill A. Shutemovddc58f22016-01-15 16:52:56 -0800618void __put_page(struct page *page);
619
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700620void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800622void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800623
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800625 * Compound pages have a destructor function. Provide a
626 * prototype for that function and accessor functions.
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800627 * These are _only_ valid on the head of a compound page.
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800628 */
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800629typedef void compound_page_dtor(struct page *);
630
631/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
632enum compound_dtor_id {
633 NULL_COMPOUND_DTOR,
634 COMPOUND_PAGE_DTOR,
635#ifdef CONFIG_HUGETLB_PAGE
636 HUGETLB_PAGE_DTOR,
637#endif
Kirill A. Shutemov9a982252016-01-15 16:54:17 -0800638#ifdef CONFIG_TRANSPARENT_HUGEPAGE
639 TRANSHUGE_PAGE_DTOR,
640#endif
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800641 NR_COMPOUND_DTORS,
642};
643extern compound_page_dtor * const compound_page_dtors[];
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800644
645static inline void set_compound_page_dtor(struct page *page,
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800646 enum compound_dtor_id compound_dtor)
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800647{
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800648 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
649 page[1].compound_dtor = compound_dtor;
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800650}
651
652static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
653{
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800654 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
655 return compound_page_dtors[page[1].compound_dtor];
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800656}
657
Kirill A. Shutemovd00181b2015-11-06 16:29:57 -0800658static inline unsigned int compound_order(struct page *page)
Christoph Lameterd85f3382007-05-06 14:49:39 -0700659{
Christoph Lameter6d777952007-05-06 14:49:40 -0700660 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700661 return 0;
Kirill A. Shutemove4b294c2015-02-11 15:24:46 -0800662 return page[1].compound_order;
Christoph Lameterd85f3382007-05-06 14:49:39 -0700663}
664
Kirill A. Shutemovf1e61552015-11-06 16:29:50 -0800665static inline void set_compound_order(struct page *page, unsigned int order)
Christoph Lameterd85f3382007-05-06 14:49:39 -0700666{
Kirill A. Shutemove4b294c2015-02-11 15:24:46 -0800667 page[1].compound_order = order;
Christoph Lameterd85f3382007-05-06 14:49:39 -0700668}
669
Kirill A. Shutemov9a982252016-01-15 16:54:17 -0800670void free_compound_page(struct page *page);
671
Michal Simek3dece372011-01-21 08:49:56 +0100672#ifdef CONFIG_MMU
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800673/*
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800674 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
675 * servicing faults for write access. In the normal case, do always want
676 * pte_mkwrite. But get_user_pages can cause write faults for mappings
677 * that do not have writing enabled, when used by access_process_vm.
678 */
679static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
680{
681 if (likely(vma->vm_flags & VM_WRITE))
682 pte = pte_mkwrite(pte);
683 return pte;
684}
Kirill A. Shutemov8c6e50b2014-04-07 15:37:18 -0700685
Jan Kara82b0f8c2016-12-14 15:06:58 -0800686int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
Kirill A. Shutemov7267ec002016-07-26 15:25:23 -0700687 struct page *page);
Jan Kara9118c0c2016-12-14 15:07:21 -0800688int finish_fault(struct vm_fault *vmf);
Jan Kara66a61972016-12-14 15:07:39 -0800689int finish_mkwrite_fault(struct vm_fault *vmf);
Michal Simek3dece372011-01-21 08:49:56 +0100690#endif
Andrea Arcangeli14fd4032011-01-13 15:46:37 -0800691
692/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 * Multiple processes may "see" the same page. E.g. for untouched
694 * mappings of /dev/null, all processes see the same page full of
695 * zeroes, and text pages of executables and shared libraries have
696 * only one copy in memory, at most, normally.
697 *
698 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700699 * page_count() == 0 means the page is free. page->lru is then used for
700 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700701 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700703 * Pages are allocated by the slab allocator in order to provide memory
704 * to kmalloc and kmem_cache_alloc. In this case, the management of the
705 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
706 * unless a particular usage is carefully commented. (the responsibility of
707 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700709 * A page may be used by anyone else who does a __get_free_page().
710 * In this case, page_count still tracks the references, and should only
711 * be used through the normal accessor functions. The top bits of page->flags
712 * and page->virtual store page management information, but all other fields
713 * are unused and could be used privately, carefully. The management of this
714 * page is the responsibility of the one who allocated it, and those who have
715 * subsequently been given references to it.
716 *
717 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 * managed by the Linux memory manager: I/O, buffers, swapping etc.
719 * The following discussion applies only to them.
720 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700721 * A pagecache page contains an opaque `private' member, which belongs to the
722 * page's address_space. Usually, this is the address of a circular list of
723 * the page's disk buffers. PG_private must be set to tell the VM to call
724 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700726 * A page may belong to an inode's memory mapping. In this case, page->mapping
727 * is the pointer to the inode, and page->index is the file offset of the page,
Kirill A. Shutemovea1754a2016-04-01 15:29:48 +0300728 * in units of PAGE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700730 * If pagecache pages are not associated with an inode, they are said to be
731 * anonymous pages. These may become associated with the swapcache, and in that
732 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700734 * In either case (swapcache or inode backed), the pagecache itself holds one
735 * reference to the page. Setting PG_private should also increment the
736 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700738 * The pagecache pages are stored in a per-mapping radix tree, which is
739 * rooted at mapping->page_tree, and indexed by offset.
740 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
741 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700743 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 * - inode pages may need to be read from disk,
745 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700746 * to be written back to the inode on disk,
747 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
748 * modified may need to be swapped out to swap space and (later) to be read
749 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 */
751
752/*
753 * The zone field is never updated after free_area_init_core()
754 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700756
Peter Zijlstra90572892013-10-07 11:29:20 +0100757/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100758#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700759#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
760#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
Peter Zijlstra90572892013-10-07 11:29:20 +0100761#define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700762
763/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300764 * Define the bit shifts to access each section. For non-existent
Dave Hansen348f8b62005-06-23 00:07:40 -0700765 * sections we define the shift as 0; that plus a 0 mask ensures
766 * the compiler will optimise away reference to them.
767 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700768#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
769#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
770#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Peter Zijlstra90572892013-10-07 11:29:20 +0100771#define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700772
Will Deaconbce54bb2010-10-26 14:21:37 -0700773/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
774#ifdef NODE_NOT_IN_PAGE_FLAGS
Christoph Lameter89689ae2006-12-06 20:31:45 -0800775#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800776#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
777 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700778#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800779#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800780#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
781 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700782#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800783
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800784#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700785
Christoph Lameter9223b4192008-04-28 02:12:48 -0700786#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
787#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700788#endif
789
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700790#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
791#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
792#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Srikar Dronamraju834a9642014-04-07 15:37:57 -0700793#define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800794#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700795
Ian Campbell33dd4e02011-07-25 17:11:51 -0700796static inline enum zone_type page_zonenum(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797{
Dave Hansen348f8b62005-06-23 00:07:40 -0700798 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
Dan Williams260ae3f2016-01-15 16:56:17 -0800801#ifdef CONFIG_ZONE_DEVICE
802static inline bool is_zone_device_page(const struct page *page)
803{
804 return page_zonenum(page) == ZONE_DEVICE;
805}
806#else
807static inline bool is_zone_device_page(const struct page *page)
808{
809 return false;
810}
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700811#endif
Jérôme Glisse5042db42017-09-08 16:11:43 -0700812
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700813#if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700814void put_zone_device_private_or_public_page(struct page *page);
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700815DECLARE_STATIC_KEY_FALSE(device_private_key);
816#define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
817static inline bool is_device_private_page(const struct page *page);
818static inline bool is_device_public_page(const struct page *page);
819#else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700820static inline void put_zone_device_private_or_public_page(struct page *page)
Jérôme Glisse5042db42017-09-08 16:11:43 -0700821{
Jérôme Glisse5042db42017-09-08 16:11:43 -0700822}
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700823#define IS_HMM_ENABLED 0
824static inline bool is_device_private_page(const struct page *page)
825{
826 return false;
827}
828static inline bool is_device_public_page(const struct page *page)
829{
830 return false;
831}
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700832#endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
Dan Williams260ae3f2016-01-15 16:56:17 -0800833
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700834
Dan Williams3565fce2016-01-15 16:56:55 -0800835static inline void get_page(struct page *page)
836{
837 page = compound_head(page);
838 /*
839 * Getting a normal page or the head of a compound page
Joonsoo Kim0139aa72016-05-19 17:10:49 -0700840 * requires to already have an elevated page->_refcount.
Dan Williams3565fce2016-01-15 16:56:55 -0800841 */
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700842 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
843 page_ref_inc(page);
Dan Williams3565fce2016-01-15 16:56:55 -0800844}
845
846static inline void put_page(struct page *page)
847{
848 page = compound_head(page);
849
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700850 /*
851 * For private device pages we need to catch refcount transition from
852 * 2 to 1, when refcount reach one it means the private device page is
853 * free and we need to inform the device driver through callback. See
854 * include/linux/memremap.h and HMM for details.
855 */
Jérôme Glisse6b368cd2017-09-08 16:12:32 -0700856 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
857 unlikely(is_device_public_page(page)))) {
Jérôme Glissedf6ad692017-09-08 16:12:24 -0700858 put_zone_device_private_or_public_page(page);
Jérôme Glisse7b2d55d22017-09-08 16:11:46 -0700859 return;
860 }
861
Dan Williams3565fce2016-01-15 16:56:55 -0800862 if (put_page_testzero(page))
863 __put_page(page);
Dan Williams3565fce2016-01-15 16:56:55 -0800864}
865
Cody P Schafer9127ab42013-02-22 16:35:21 -0800866#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
867#define SECTION_IN_PAGE_FLAGS
868#endif
869
Christoph Lameter89689ae2006-12-06 20:31:45 -0800870/*
Vlastimil Babka7a8010c2013-09-11 14:22:35 -0700871 * The identification function is mainly used by the buddy allocator for
872 * determining if two pages could be buddies. We are not really identifying
873 * the zone since we could be using the section number id if we do not have
874 * node id available in page flags.
875 * We only guarantee that it will return the same value for two combinable
876 * pages in a zone.
Christoph Lameter89689ae2006-12-06 20:31:45 -0800877 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700878static inline int page_zone_id(struct page *page)
879{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800880 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881}
882
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800883static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700884{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700885#ifdef CONFIG_NUMA
886 return zone->node;
887#else
888 return 0;
889#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700890}
891
Christoph Lameter89689ae2006-12-06 20:31:45 -0800892#ifdef NODE_NOT_IN_PAGE_FLAGS
Ian Campbell33dd4e02011-07-25 17:11:51 -0700893extern int page_to_nid(const struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800894#else
Ian Campbell33dd4e02011-07-25 17:11:51 -0700895static inline int page_to_nid(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700896{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800897 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700898}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800899#endif
900
Mel Gorman57e0a032012-11-12 09:06:20 +0000901#ifdef CONFIG_NUMA_BALANCING
Peter Zijlstra90572892013-10-07 11:29:20 +0100902static inline int cpu_pid_to_cpupid(int cpu, int pid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000903{
Peter Zijlstra90572892013-10-07 11:29:20 +0100904 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
Mel Gorman57e0a032012-11-12 09:06:20 +0000905}
906
Peter Zijlstra90572892013-10-07 11:29:20 +0100907static inline int cpupid_to_pid(int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000908{
Peter Zijlstra90572892013-10-07 11:29:20 +0100909 return cpupid & LAST__PID_MASK;
Mel Gorman57e0a032012-11-12 09:06:20 +0000910}
Mel Gormanb7958542013-10-07 11:29:07 +0100911
Peter Zijlstra90572892013-10-07 11:29:20 +0100912static inline int cpupid_to_cpu(int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000913{
Peter Zijlstra90572892013-10-07 11:29:20 +0100914 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
Mel Gormanb7958542013-10-07 11:29:07 +0100915}
916
Peter Zijlstra90572892013-10-07 11:29:20 +0100917static inline int cpupid_to_nid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100918{
Peter Zijlstra90572892013-10-07 11:29:20 +0100919 return cpu_to_node(cpupid_to_cpu(cpupid));
Mel Gormanb7958542013-10-07 11:29:07 +0100920}
921
Peter Zijlstra90572892013-10-07 11:29:20 +0100922static inline bool cpupid_pid_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100923{
Peter Zijlstra90572892013-10-07 11:29:20 +0100924 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100925}
926
Peter Zijlstra90572892013-10-07 11:29:20 +0100927static inline bool cpupid_cpu_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100928{
Peter Zijlstra90572892013-10-07 11:29:20 +0100929 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100930}
931
Peter Zijlstra8c8a7432013-10-07 11:29:21 +0100932static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
933{
934 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
935}
936
937#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
Peter Zijlstra90572892013-10-07 11:29:20 +0100938#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
939static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100940{
Liu Ping Fan1ae71d02014-03-03 15:38:39 -0800941 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
Mel Gormanb7958542013-10-07 11:29:07 +0100942}
Peter Zijlstra90572892013-10-07 11:29:20 +0100943
944static inline int page_cpupid_last(struct page *page)
Mel Gormanb7958542013-10-07 11:29:07 +0100945{
Peter Zijlstra90572892013-10-07 11:29:20 +0100946 return page->_last_cpupid;
947}
948static inline void page_cpupid_reset_last(struct page *page)
949{
Liu Ping Fan1ae71d02014-03-03 15:38:39 -0800950 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
Mel Gorman57e0a032012-11-12 09:06:20 +0000951}
952#else
Peter Zijlstra90572892013-10-07 11:29:20 +0100953static inline int page_cpupid_last(struct page *page)
Peter Zijlstra75980e92013-02-22 16:34:32 -0800954{
Peter Zijlstra90572892013-10-07 11:29:20 +0100955 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
Peter Zijlstra75980e92013-02-22 16:34:32 -0800956}
957
Peter Zijlstra90572892013-10-07 11:29:20 +0100958extern int page_cpupid_xchg_last(struct page *page, int cpupid);
Peter Zijlstra75980e92013-02-22 16:34:32 -0800959
Peter Zijlstra90572892013-10-07 11:29:20 +0100960static inline void page_cpupid_reset_last(struct page *page)
Peter Zijlstra75980e92013-02-22 16:34:32 -0800961{
Mel Gorman09940a42016-05-19 17:13:53 -0700962 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
Peter Zijlstra75980e92013-02-22 16:34:32 -0800963}
Peter Zijlstra90572892013-10-07 11:29:20 +0100964#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
965#else /* !CONFIG_NUMA_BALANCING */
966static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
Mel Gorman57e0a032012-11-12 09:06:20 +0000967{
Peter Zijlstra90572892013-10-07 11:29:20 +0100968 return page_to_nid(page); /* XXX */
Mel Gorman57e0a032012-11-12 09:06:20 +0000969}
970
Peter Zijlstra90572892013-10-07 11:29:20 +0100971static inline int page_cpupid_last(struct page *page)
Mel Gorman57e0a032012-11-12 09:06:20 +0000972{
Peter Zijlstra90572892013-10-07 11:29:20 +0100973 return page_to_nid(page); /* XXX */
Mel Gorman57e0a032012-11-12 09:06:20 +0000974}
975
Peter Zijlstra90572892013-10-07 11:29:20 +0100976static inline int cpupid_to_nid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100977{
978 return -1;
979}
980
Peter Zijlstra90572892013-10-07 11:29:20 +0100981static inline int cpupid_to_pid(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100982{
983 return -1;
984}
985
Peter Zijlstra90572892013-10-07 11:29:20 +0100986static inline int cpupid_to_cpu(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100987{
988 return -1;
989}
990
Peter Zijlstra90572892013-10-07 11:29:20 +0100991static inline int cpu_pid_to_cpupid(int nid, int pid)
992{
993 return -1;
994}
995
996static inline bool cpupid_pid_unset(int cpupid)
Mel Gormanb7958542013-10-07 11:29:07 +0100997{
998 return 1;
999}
1000
Peter Zijlstra90572892013-10-07 11:29:20 +01001001static inline void page_cpupid_reset_last(struct page *page)
Mel Gorman57e0a032012-11-12 09:06:20 +00001002{
1003}
Peter Zijlstra8c8a7432013-10-07 11:29:21 +01001004
1005static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1006{
1007 return false;
1008}
Peter Zijlstra90572892013-10-07 11:29:20 +01001009#endif /* CONFIG_NUMA_BALANCING */
Mel Gorman57e0a032012-11-12 09:06:20 +00001010
Ian Campbell33dd4e02011-07-25 17:11:51 -07001011static inline struct zone *page_zone(const struct page *page)
Christoph Lameter89689ae2006-12-06 20:31:45 -08001012{
1013 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1014}
1015
Mel Gorman75ef7182016-07-28 15:45:24 -07001016static inline pg_data_t *page_pgdat(const struct page *page)
1017{
1018 return NODE_DATA(page_to_nid(page));
1019}
1020
Cody P Schafer9127ab42013-02-22 16:35:21 -08001021#ifdef SECTION_IN_PAGE_FLAGS
Daniel Kiperbf4e8902011-05-24 17:12:32 -07001022static inline void set_page_section(struct page *page, unsigned long section)
1023{
1024 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1025 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1026}
1027
Ian Campbellaa462ab2011-08-17 17:40:33 +01001028static inline unsigned long page_to_section(const struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001029{
1030 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1031}
Christoph Lameter308c05e2008-04-28 02:12:43 -07001032#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001033
Christoph Lameter2f1b6242006-09-25 23:31:13 -07001034static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035{
Dave Hansen348f8b62005-06-23 00:07:40 -07001036 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1037 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1038}
Christoph Lameter2f1b6242006-09-25 23:31:13 -07001039
Dave Hansen348f8b62005-06-23 00:07:40 -07001040static inline void set_page_node(struct page *page, unsigned long node)
1041{
1042 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1043 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1044}
Christoph Lameter89689ae2006-12-06 20:31:45 -08001045
Christoph Lameter2f1b6242006-09-25 23:31:13 -07001046static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001047 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -07001048{
1049 set_page_zone(page, zone);
1050 set_page_node(page, node);
Cody P Schafer9127ab42013-02-22 16:35:21 -08001051#ifdef SECTION_IN_PAGE_FLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001052 set_page_section(page, pfn_to_section_nr(pfn));
Daniel Kiperbf4e8902011-05-24 17:12:32 -07001053#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054}
1055
Greg Thelen0610c252015-10-01 15:37:02 -07001056#ifdef CONFIG_MEMCG
1057static inline struct mem_cgroup *page_memcg(struct page *page)
1058{
1059 return page->mem_cgroup;
1060}
Johannes Weiner55779ec2016-07-28 15:45:10 -07001061static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1062{
1063 WARN_ON_ONCE(!rcu_read_lock_held());
1064 return READ_ONCE(page->mem_cgroup);
1065}
Greg Thelen0610c252015-10-01 15:37:02 -07001066#else
1067static inline struct mem_cgroup *page_memcg(struct page *page)
1068{
1069 return NULL;
1070}
Johannes Weiner55779ec2016-07-28 15:45:10 -07001071static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1072{
1073 WARN_ON_ONCE(!rcu_read_lock_held());
1074 return NULL;
1075}
Greg Thelen0610c252015-10-01 15:37:02 -07001076#endif
1077
Christoph Lameterf6ac2352006-06-30 01:55:32 -07001078/*
1079 * Some inline functions in vmstat.h depend on page_zone()
1080 */
1081#include <linux/vmstat.h>
1082
Ian Campbell33dd4e02011-07-25 17:11:51 -07001083static __always_inline void *lowmem_page_address(const struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084{
Ard Biesheuvel1dff8082016-04-18 18:04:57 +02001085 return page_to_virt(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086}
1087
1088#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1089#define HASHED_PAGE_VIRTUAL
1090#endif
1091
1092#if defined(WANT_PAGE_VIRTUAL)
Geert Uytterhoevenf92f4552014-01-21 15:48:47 -08001093static inline void *page_address(const struct page *page)
1094{
1095 return page->virtual;
1096}
1097static inline void set_page_address(struct page *page, void *address)
1098{
1099 page->virtual = address;
1100}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101#define page_address_init() do { } while(0)
1102#endif
1103
1104#if defined(HASHED_PAGE_VIRTUAL)
Ian Campbellf9918792011-08-17 13:45:09 +01001105void *page_address(const struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106void set_page_address(struct page *page, void *virtual);
1107void page_address_init(void);
1108#endif
1109
1110#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1111#define page_address(page) lowmem_page_address(page)
1112#define set_page_address(page, address) do { } while(0)
1113#define page_address_init() do { } while(0)
1114#endif
1115
Kirill A. Shutemove39155e2015-04-15 16:14:53 -07001116extern void *page_rmapping(struct page *page);
1117extern struct anon_vma *page_anon_vma(struct page *page);
Shaohua Li98003392013-02-22 16:34:35 -08001118extern struct address_space *page_mapping(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119
Mel Gormanf981c592012-07-31 16:44:47 -07001120extern struct address_space *__page_file_mapping(struct page *);
1121
1122static inline
1123struct address_space *page_file_mapping(struct page *page)
1124{
1125 if (unlikely(PageSwapCache(page)))
1126 return __page_file_mapping(page);
1127
1128 return page->mapping;
1129}
1130
Huang Yingf6ab1f72016-10-07 17:00:21 -07001131extern pgoff_t __page_file_index(struct page *page);
1132
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133/*
1134 * Return the pagecache index of the passed page. Regular pagecache pages
Huang Yingf6ab1f72016-10-07 17:00:21 -07001135 * use ->index whereas swapcache pages use swp_offset(->private)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 */
1137static inline pgoff_t page_index(struct page *page)
1138{
1139 if (unlikely(PageSwapCache(page)))
Huang Yingf6ab1f72016-10-07 17:00:21 -07001140 return __page_file_index(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 return page->index;
1142}
1143
Andrew Morton1aa8aea2016-05-19 17:12:00 -07001144bool page_mapped(struct page *page);
Minchan Kimbda807d2016-07-26 15:23:05 -07001145struct address_space *page_mapping(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
1147/*
Michal Hocko2f064f32015-08-21 14:11:51 -07001148 * Return true only if the page has been allocated with
1149 * ALLOC_NO_WATERMARKS and the low watermark was not
1150 * met implying that the system is under some pressure.
1151 */
1152static inline bool page_is_pfmemalloc(struct page *page)
1153{
1154 /*
1155 * Page index cannot be this large so this must be
1156 * a pfmemalloc page.
1157 */
1158 return page->index == -1UL;
1159}
1160
1161/*
1162 * Only to be called by the page allocator on a freshly allocated
1163 * page.
1164 */
1165static inline void set_page_pfmemalloc(struct page *page)
1166{
1167 page->index = -1UL;
1168}
1169
1170static inline void clear_page_pfmemalloc(struct page *page)
1171{
1172 page->index = 0;
1173}
1174
1175/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 * Different kinds of faults, as returned by handle_mm_fault().
1177 * Used to decide whether a process gets delivered SIGBUS or
1178 * just gets major/minor fault counters bumped up.
1179 */
Nick Piggind0217ac042007-07-19 01:47:03 -07001180
Nick Piggin83c54072007-07-19 01:47:05 -07001181#define VM_FAULT_OOM 0x0001
1182#define VM_FAULT_SIGBUS 0x0002
1183#define VM_FAULT_MAJOR 0x0004
1184#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001185#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1186#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
Linus Torvalds33692f22015-01-29 10:51:32 -08001187#define VM_FAULT_SIGSEGV 0x0040
Nick Pigginf33ea7f2005-08-03 20:24:01 +10001188
Nick Piggin83c54072007-07-19 01:47:05 -07001189#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1190#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Michel Lespinassed065bd82010-10-26 14:21:57 -07001191#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
Kirill A. Shutemovc0292552013-09-12 15:14:05 -07001192#define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
Jan Karab1aa8122016-12-14 15:07:24 -08001193#define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001195#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1196
Linus Torvalds33692f22015-01-29 10:51:32 -08001197#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1198 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1199 VM_FAULT_FALLBACK)
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001200
Ross Zwisler282a8e02017-02-22 15:39:50 -08001201#define VM_FAULT_RESULT_TRACE \
1202 { VM_FAULT_OOM, "OOM" }, \
1203 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1204 { VM_FAULT_MAJOR, "MAJOR" }, \
1205 { VM_FAULT_WRITE, "WRITE" }, \
1206 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1207 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1208 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1209 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1210 { VM_FAULT_LOCKED, "LOCKED" }, \
1211 { VM_FAULT_RETRY, "RETRY" }, \
1212 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1213 { VM_FAULT_DONE_COW, "DONE_COW" }
1214
Andi Kleenaa50d3a2010-10-06 21:45:00 +02001215/* Encode hstate index for a hwpoisoned large page */
1216#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1217#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
Nick Piggind0217ac042007-07-19 01:47:03 -07001218
Nick Piggin1c0fe6e2009-01-06 14:38:59 -08001219/*
1220 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1221 */
1222extern void pagefault_out_of_memory(void);
1223
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1225
David Rientjesddd588b2011-03-22 16:30:46 -07001226/*
David Rientjes7bf02ea2011-05-24 17:11:16 -07001227 * Flags passed to show_mem() and show_free_areas() to suppress output in
David Rientjesddd588b2011-03-22 16:30:46 -07001228 * various contexts.
1229 */
David Rientjes4b59e6c2013-04-29 15:06:11 -07001230#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
David Rientjesddd588b2011-03-22 16:30:46 -07001231
Michal Hocko9af744d2017-02-22 15:46:16 -08001232extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233
Wang Xiaoqiang7f43add2016-01-15 16:57:22 -08001234extern bool can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235extern int user_shm_lock(size_t, struct user_struct *);
1236extern void user_shm_unlock(size_t, struct user_struct *);
1237
1238/*
1239 * Parameter block passed down to zap_pte_range in exceptional cases.
1240 */
1241struct zap_details {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 struct address_space *check_mapping; /* Check page->mapping if set */
1243 pgoff_t first_index; /* Lowest page->index to unmap */
1244 pgoff_t last_index; /* Highest page->index to unmap */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245};
1246
Jérôme Glissedf6ad692017-09-08 16:12:24 -07001247struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1248 pte_t pte, bool with_public_device);
1249#define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1250
Gerald Schaefer28093f92016-04-28 16:18:35 -07001251struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1252 pmd_t pmd);
Nick Piggin7e675132008-04-28 02:13:00 -07001253
Jack Steinerc627f9c2008-07-29 22:33:53 -07001254int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1255 unsigned long size);
Al Viro14f5ff52012-03-05 13:38:09 -05001256void zap_page_range(struct vm_area_struct *vma, unsigned long address,
Kirill A. Shutemovecf13852017-02-22 15:46:37 -08001257 unsigned long size);
Linus Torvalds4f74d2c2012-05-06 13:54:06 -07001258void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1259 unsigned long start, unsigned long end);
Matt Mackalle64730922008-02-04 22:29:01 -08001260
1261/**
1262 * mm_walk - callbacks for walk_page_range
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001263 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1264 * this handler should only handle pud_trans_huge() puds.
1265 * the pmd_entry or pte_entry callbacks will be used for
1266 * regular PUDs.
Matt Mackalle64730922008-02-04 22:29:01 -08001267 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
Dave Hansen03319322011-03-22 16:32:56 -07001268 * this handler is required to be able to handle
1269 * pmd_trans_huge() pmds. They may simply choose to
1270 * split_huge_page() instead of handling it explicitly.
Matt Mackalle64730922008-02-04 22:29:01 -08001271 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1272 * @pte_hole: if set, called for each hole at all levels
Naoya Horiguchi5dc37642009-12-14 18:00:01 -08001273 * @hugetlb_entry: if set, called for each hugetlb entry
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001274 * @test_walk: caller specific callback function to determine whether
James Morsef7e23552016-10-07 16:58:36 -07001275 * we walk over the current vma or not. Returning 0
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001276 * value means "do page table walk over the current vma,"
1277 * and a negative one means "abort current page table walk
James Morsef7e23552016-10-07 16:58:36 -07001278 * right now." 1 means "skip the current vma."
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001279 * @mm: mm_struct representing the target process of page table walk
1280 * @vma: vma currently walked (NULL if walking outside vmas)
1281 * @private: private data for callbacks' usage
Matt Mackalle64730922008-02-04 22:29:01 -08001282 *
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001283 * (see the comment on walk_page_range() for more details)
Matt Mackalle64730922008-02-04 22:29:01 -08001284 */
1285struct mm_walk {
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001286 int (*pud_entry)(pud_t *pud, unsigned long addr,
1287 unsigned long next, struct mm_walk *walk);
Andrew Morton0f157a52013-05-07 16:18:10 -07001288 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1289 unsigned long next, struct mm_walk *walk);
1290 int (*pte_entry)(pte_t *pte, unsigned long addr,
1291 unsigned long next, struct mm_walk *walk);
1292 int (*pte_hole)(unsigned long addr, unsigned long next,
1293 struct mm_walk *walk);
1294 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1295 unsigned long addr, unsigned long next,
1296 struct mm_walk *walk);
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001297 int (*test_walk)(unsigned long addr, unsigned long next,
1298 struct mm_walk *walk);
Dave Hansen21650092008-06-12 15:21:47 -07001299 struct mm_struct *mm;
Naoya Horiguchifafaa422015-02-11 15:27:37 -08001300 struct vm_area_struct *vma;
Dave Hansen21650092008-06-12 15:21:47 -07001301 void *private;
Matt Mackalle64730922008-02-04 22:29:01 -08001302};
1303
Dave Hansen21650092008-06-12 15:21:47 -07001304int walk_page_range(unsigned long addr, unsigned long end,
1305 struct mm_walk *walk);
Naoya Horiguchi900fc5f2015-02-11 15:27:40 -08001306int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -07001307void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -07001308 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1310 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311void unmap_mapping_range(struct address_space *mapping,
1312 loff_t const holebegin, loff_t const holelen, int even_cows);
Ross Zwisler09796392017-01-10 16:57:21 -08001313int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
Jérôme Glissea4d1a882017-08-31 17:17:26 -04001314 unsigned long *start, unsigned long *end,
Ross Zwisler09796392017-01-10 16:57:21 -08001315 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
Johannes Weiner3b6748e2009-06-16 15:32:35 -07001316int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1317 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -08001318int follow_phys(struct vm_area_struct *vma, unsigned long address,
1319 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -07001320int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1321 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322
1323static inline void unmap_shared_mapping_range(struct address_space *mapping,
1324 loff_t const holebegin, loff_t const holelen)
1325{
1326 unmap_mapping_range(mapping, holebegin, holelen, 0);
1327}
1328
Kirill A. Shutemov7caef262013-09-12 15:13:56 -07001329extern void truncate_pagecache(struct inode *inode, loff_t new);
Christoph Hellwig2c27c652010-06-04 11:30:04 +02001330extern void truncate_setsize(struct inode *inode, loff_t newsize);
Jan Kara90a80202014-10-01 21:49:18 -04001331void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
Hugh Dickins623e3db2012-03-28 14:42:40 -07001332void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
Nick Piggin750b4982009-09-16 11:50:12 +02001333int truncate_inode_page(struct address_space *mapping, struct page *page);
Andi Kleen25718732009-09-16 11:50:13 +02001334int generic_error_remove_page(struct address_space *mapping, struct page *page);
Wu Fengguang83f78662009-09-16 11:50:13 +02001335int invalidate_inode_page(struct page *page);
1336
David Howells7ee1dd32006-01-06 00:11:44 -08001337#ifdef CONFIG_MMU
Kirill A. Shutemovdcddffd2016-07-26 15:25:18 -07001338extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1339 unsigned int flags);
Peter Zijlstra5c723ba2011-07-27 12:17:11 +02001340extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
Dominik Dingel4a9e1cd2016-01-15 16:57:04 -08001341 unsigned long address, unsigned int fault_flags,
1342 bool *unlocked);
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}
David Howells7ee1dd32006-01-06 00:11:44 -08001359#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +10001360
Lorenzo Stoakesf307ab62016-10-13 01:20:20 +01001361extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1362 unsigned int gup_flags);
Stephen Wilson5ddd36b2011-03-13 15:49:20 -04001363extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +01001364 void *buf, int len, unsigned int gup_flags);
Eric W. Biederman84d77d32016-11-22 12:06:50 -06001365extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1366 unsigned long addr, void *buf, int len, unsigned int gup_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
Dave Hansen1e987792016-02-12 13:01:54 -08001368long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1369 unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +01001370 unsigned int gup_flags, struct page **pages,
Lorenzo Stoakes5b56d492016-12-14 15:06:52 -08001371 struct vm_area_struct **vmas, int *locked);
Ingo Molnarc12d2da2016-04-04 10:24:58 +02001372long get_user_pages(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes768ae302016-10-13 01:20:16 +01001373 unsigned int gup_flags, struct page **pages,
Dave Hansencde70142016-02-12 13:01:55 -08001374 struct vm_area_struct **vmas);
Ingo Molnarc12d2da2016-04-04 10:24:58 +02001375long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakes3b913172016-10-13 01:20:14 +01001376 unsigned int gup_flags, struct page **pages, int *locked);
Ingo Molnarc12d2da2016-04-04 10:24:58 +02001377long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
Lorenzo Stoakesc1641542016-10-13 01:20:13 +01001378 struct page **pages, unsigned int gup_flags);
Nick Piggind2bf6be2009-06-16 15:31:39 -07001379int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1380 struct page **pages);
Jan Kara8025e5d2015-07-13 11:55:44 -03001381
1382/* Container for pinned pfns / pages */
1383struct frame_vector {
1384 unsigned int nr_allocated; /* Number of frames we have space for */
1385 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1386 bool got_ref; /* Did we pin pages by getting page ref? */
1387 bool is_pfns; /* Does array contain pages or pfns? */
1388 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1389 * pfns_vector_pages() or pfns_vector_pfns()
1390 * for access */
1391};
1392
1393struct frame_vector *frame_vector_create(unsigned int nr_frames);
1394void frame_vector_destroy(struct frame_vector *vec);
1395int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
Lorenzo Stoakes7f23b352016-10-13 01:20:15 +01001396 unsigned int gup_flags, struct frame_vector *vec);
Jan Kara8025e5d2015-07-13 11:55:44 -03001397void put_vaddr_frames(struct frame_vector *vec);
1398int frame_vector_to_pages(struct frame_vector *vec);
1399void frame_vector_to_pfns(struct frame_vector *vec);
1400
1401static inline unsigned int frame_vector_count(struct frame_vector *vec)
1402{
1403 return vec->nr_frames;
1404}
1405
1406static inline struct page **frame_vector_pages(struct frame_vector *vec)
1407{
1408 if (vec->is_pfns) {
1409 int err = frame_vector_to_pages(vec);
1410
1411 if (err)
1412 return ERR_PTR(err);
1413 }
1414 return (struct page **)(vec->ptrs);
1415}
1416
1417static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1418{
1419 if (!vec->is_pfns)
1420 frame_vector_to_pfns(vec);
1421 return (unsigned long *)(vec->ptrs);
1422}
1423
Mel Gorman18022c52012-07-31 16:44:51 -07001424struct kvec;
1425int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1426 struct page **pages);
1427int get_kernel_page(unsigned long start, int write, struct page **pages);
Hugh Dickinsf3e8fcc2009-09-21 17:03:25 -07001428struct page *get_dump_page(unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
David Howellscf9a2ae2006-08-29 19:05:54 +01001430extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
Lukas Czernerd47992f2013-05-21 23:17:23 -04001431extern void do_invalidatepage(struct page *page, unsigned int offset,
1432 unsigned int length);
David Howellscf9a2ae2006-08-29 19:05:54 +01001433
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -08001435int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436int redirty_page_for_writepage(struct writeback_control *wbc,
1437 struct page *page);
Johannes Weiner62cccb82016-03-15 14:57:22 -07001438void account_page_dirtied(struct page *page, struct address_space *mapping);
Greg Thelenc4843a72015-05-22 17:13:16 -04001439void account_page_cleaned(struct page *page, struct address_space *mapping,
Johannes Weiner62cccb82016-03-15 14:57:22 -07001440 struct bdi_writeback *wb);
Harvey Harrisonb3c97522008-02-13 15:03:15 -08001441int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442int set_page_dirty_lock(struct page *page);
Jan Kara736304f2017-11-15 17:37:11 -08001443void __cancel_dirty_page(struct page *page);
1444static inline void cancel_dirty_page(struct page *page)
1445{
1446 /* Avoid atomic ops, locking, etc. when not actually needed. */
1447 if (PageDirty(page))
1448 __cancel_dirty_page(page);
1449}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450int clear_page_dirty_for_io(struct page *page);
Konstantin Khlebnikovb9ea2512015-04-14 15:45:27 -07001451
William Robertsa9090252014-02-11 10:11:59 -08001452int get_cmdline(struct task_struct *task, char *buffer, int buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453
Oleg Nesterovb5330622015-09-08 14:58:28 -07001454static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1455{
1456 return !vma->vm_ops;
1457}
1458
Mike Rapoportb0506e42017-02-22 15:43:28 -08001459#ifdef CONFIG_SHMEM
1460/*
1461 * The vma_is_shmem is not inline because it is used only by slow
1462 * paths in userfault.
1463 */
1464bool vma_is_shmem(struct vm_area_struct *vma);
1465#else
1466static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1467#endif
1468
Andy Lutomirskid17af502016-09-30 10:58:58 -07001469int vma_is_stack_for_current(struct vm_area_struct *vma);
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07001470
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001471extern unsigned long move_page_tables(struct vm_area_struct *vma,
1472 unsigned long old_addr, struct vm_area_struct *new_vma,
Michel Lespinasse38a76012012-10-08 16:31:50 -07001473 unsigned long new_addr, unsigned long len,
1474 bool need_rmap_locks);
Peter Zijlstra7da4d642012-11-19 03:14:23 +01001475extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1476 unsigned long end, pgprot_t newprot,
Mel Gorman4b10e7d2012-10-25 14:16:32 +02001477 int dirty_accountable, int prot_numa);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001478extern int mprotect_fixup(struct vm_area_struct *vma,
1479 struct vm_area_struct **pprev, unsigned long start,
1480 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
Nick Piggin21cc1992008-07-25 19:45:22 -07001482/*
Peter Zijlstra465a4542009-06-15 12:31:37 +02001483 * doesn't attempt to fault and will return short.
1484 */
1485int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1486 struct page **pages);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001487/*
1488 * per-process(per-mm_struct) statistics.
1489 */
Matt Fleming172703b2011-05-24 17:12:36 -07001490static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1491{
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001492 long val = atomic_long_read(&mm->rss_stat.count[member]);
1493
1494#ifdef SPLIT_RSS_COUNTING
1495 /*
1496 * counter is updated in asynchronous manner and may go to minus.
1497 * But it's never be expected number for users.
1498 */
1499 if (val < 0)
1500 val = 0;
Matt Fleming172703b2011-05-24 17:12:36 -07001501#endif
Konstantin Khlebnikov69c97822012-03-21 16:33:49 -07001502 return (unsigned long)val;
1503}
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001504
1505static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1506{
1507 atomic_long_add(value, &mm->rss_stat.count[member]);
1508}
1509
1510static inline void inc_mm_counter(struct mm_struct *mm, int member)
1511{
1512 atomic_long_inc(&mm->rss_stat.count[member]);
1513}
1514
1515static inline void dec_mm_counter(struct mm_struct *mm, int member)
1516{
1517 atomic_long_dec(&mm->rss_stat.count[member]);
1518}
1519
Jerome Marchandeca56ff2016-01-14 15:19:26 -08001520/* Optimized variant when page is already known not to be PageAnon */
1521static inline int mm_counter_file(struct page *page)
1522{
1523 if (PageSwapBacked(page))
1524 return MM_SHMEMPAGES;
1525 return MM_FILEPAGES;
1526}
1527
1528static inline int mm_counter(struct page *page)
1529{
1530 if (PageAnon(page))
1531 return MM_ANONPAGES;
1532 return mm_counter_file(page);
1533}
1534
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001535static inline unsigned long get_mm_rss(struct mm_struct *mm)
1536{
1537 return get_mm_counter(mm, MM_FILEPAGES) +
Jerome Marchandeca56ff2016-01-14 15:19:26 -08001538 get_mm_counter(mm, MM_ANONPAGES) +
1539 get_mm_counter(mm, MM_SHMEMPAGES);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001540}
1541
1542static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1543{
1544 return max(mm->hiwater_rss, get_mm_rss(mm));
1545}
1546
1547static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1548{
1549 return max(mm->hiwater_vm, mm->total_vm);
1550}
1551
1552static inline void update_hiwater_rss(struct mm_struct *mm)
1553{
1554 unsigned long _rss = get_mm_rss(mm);
1555
1556 if ((mm)->hiwater_rss < _rss)
1557 (mm)->hiwater_rss = _rss;
1558}
1559
1560static inline void update_hiwater_vm(struct mm_struct *mm)
1561{
1562 if (mm->hiwater_vm < mm->total_vm)
1563 mm->hiwater_vm = mm->total_vm;
1564}
1565
Petr Cermak695f0552015-02-12 15:01:00 -08001566static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1567{
1568 mm->hiwater_rss = get_mm_rss(mm);
1569}
1570
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001571static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1572 struct mm_struct *mm)
1573{
1574 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1575
1576 if (*maxrss < hiwater_rss)
1577 *maxrss = hiwater_rss;
1578}
1579
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001580#if defined(SPLIT_RSS_COUNTING)
David Rientjes05af2e12012-03-21 16:34:13 -07001581void sync_mm_rss(struct mm_struct *mm);
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001582#else
David Rientjes05af2e12012-03-21 16:34:13 -07001583static inline void sync_mm_rss(struct mm_struct *mm)
KAMEZAWA Hiroyuki53bddb42010-03-10 15:20:38 -08001584{
1585}
1586#endif
Peter Zijlstra465a4542009-06-15 12:31:37 +02001587
Dan Williams3565fce2016-01-15 16:56:55 -08001588#ifndef __HAVE_ARCH_PTE_DEVMAP
1589static inline int pte_devmap(pte_t pte)
1590{
1591 return 0;
1592}
1593#endif
1594
Andrea Arcangeli6d2329f2016-10-07 17:01:22 -07001595int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
Peter Zijlstrad08b3852006-09-25 23:30:57 -07001596
Namhyung Kim25ca1d62010-10-26 14:21:59 -07001597extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1598 spinlock_t **ptl);
1599static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1600 spinlock_t **ptl)
1601{
1602 pte_t *ptep;
1603 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1604 return ptep;
1605}
Linus Torvaldsc9cfcddf2005-11-29 14:03:14 -08001606
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001607#ifdef __PAGETABLE_P4D_FOLDED
1608static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
Nick Piggin5f22df02007-05-06 14:49:02 -07001609 unsigned long address)
1610{
1611 return 0;
1612}
1613#else
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001614int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1615#endif
1616
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001617#if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001618static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1619 unsigned long address)
1620{
1621 return 0;
1622}
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001623static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1624static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1625
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001626#else
1627int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001628
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001629static inline void mm_inc_nr_puds(struct mm_struct *mm)
1630{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001631 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001632}
1633
1634static inline void mm_dec_nr_puds(struct mm_struct *mm)
1635{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001636 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
Kirill A. Shutemovb4e98d92017-11-15 17:35:33 -08001637}
Nick Piggin5f22df02007-05-06 14:49:02 -07001638#endif
1639
Kirill A. Shutemov2d2f5112015-02-12 14:59:59 -08001640#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
Nick Piggin5f22df02007-05-06 14:49:02 -07001641static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1642 unsigned long address)
1643{
1644 return 0;
1645}
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001646
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001647static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1648static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1649
Nick Piggin5f22df02007-05-06 14:49:02 -07001650#else
Hugh Dickins1bb36302005-10-29 18:16:22 -07001651int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001652
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001653static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1654{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001655 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001656}
1657
1658static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1659{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001660 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
Kirill A. Shutemovdc6c9a32015-02-11 15:26:50 -08001661}
Nick Piggin5f22df02007-05-06 14:49:02 -07001662#endif
1663
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001664#ifdef CONFIG_MMU
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001665static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001666{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001667 atomic_long_set(&mm->pgtables_bytes, 0);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001668}
1669
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001670static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001671{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001672 return atomic_long_read(&mm->pgtables_bytes);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001673}
1674
1675static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1676{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001677 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001678}
1679
1680static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1681{
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001682 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001683}
1684#else
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001685
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001686static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1687static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
Kirill A. Shutemovc4812902017-11-15 17:35:37 -08001688{
1689 return 0;
1690}
1691
1692static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1693static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1694#endif
1695
Kirill A. Shutemov3ed3a4f2016-03-17 14:19:11 -07001696int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
Hugh Dickins1bb36302005-10-29 18:16:22 -07001697int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1698
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699/*
1700 * The following ifdef needed to get the 4level-fixup.h header to work.
1701 * Remove it when 4level-fixup.h has been removed.
1702 */
Hugh Dickins1bb36302005-10-29 18:16:22 -07001703#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Kirill A. Shutemov505a60e2017-03-09 17:24:03 +03001704
1705#ifndef __ARCH_HAS_5LEVEL_HACK
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001706static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1707 unsigned long address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708{
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03001709 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1710 NULL : p4d_offset(pgd, address);
1711}
1712
1713static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1714 unsigned long address)
1715{
1716 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1717 NULL : pud_offset(p4d, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718}
Kirill A. Shutemov505a60e2017-03-09 17:24:03 +03001719#endif /* !__ARCH_HAS_5LEVEL_HACK */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720
1721static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1722{
Hugh Dickins1bb36302005-10-29 18:16:22 -07001723 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1724 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725}
Hugh Dickins1bb36302005-10-29 18:16:22 -07001726#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1727
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001728#if USE_SPLIT_PTE_PTLOCKS
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001729#if ALLOC_SPLIT_PTLOCKS
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001730void __init ptlock_cache_init(void);
Peter Zijlstra539edb52013-11-14 14:31:52 -08001731extern bool ptlock_alloc(struct page *page);
1732extern void ptlock_free(struct page *page);
1733
1734static inline spinlock_t *ptlock_ptr(struct page *page)
1735{
1736 return page->ptl;
1737}
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001738#else /* ALLOC_SPLIT_PTLOCKS */
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001739static inline void ptlock_cache_init(void)
1740{
1741}
1742
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001743static inline bool ptlock_alloc(struct page *page)
1744{
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001745 return true;
1746}
Peter Zijlstra539edb52013-11-14 14:31:52 -08001747
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001748static inline void ptlock_free(struct page *page)
1749{
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001750}
1751
1752static inline spinlock_t *ptlock_ptr(struct page *page)
1753{
Peter Zijlstra539edb52013-11-14 14:31:52 -08001754 return &page->ptl;
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001755}
Kirill A. Shutemov597d7952013-12-20 13:35:58 +02001756#endif /* ALLOC_SPLIT_PTLOCKS */
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001757
1758static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1759{
1760 return ptlock_ptr(pmd_page(*pmd));
1761}
1762
1763static inline bool ptlock_init(struct page *page)
1764{
1765 /*
1766 * prep_new_page() initialize page->private (and therefore page->ptl)
1767 * with 0. Make sure nobody took it in use in between.
1768 *
1769 * It can happen if arch try to use slab for page table allocation:
Kirill A. Shutemov1d798ca2015-11-06 16:29:54 -08001770 * slab code uses page->slab_cache, which share storage with page->ptl.
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001771 */
Sasha Levin309381fea2014-01-23 15:52:54 -08001772 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001773 if (!ptlock_alloc(page))
1774 return false;
1775 spin_lock_init(ptlock_ptr(page));
1776 return true;
1777}
1778
1779/* Reset page->mapping so free_pages_check won't complain. */
1780static inline void pte_lock_deinit(struct page *page)
1781{
1782 page->mapping = NULL;
1783 ptlock_free(page);
1784}
1785
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001786#else /* !USE_SPLIT_PTE_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001787/*
1788 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1789 */
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001790static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1791{
1792 return &mm->page_table_lock;
1793}
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001794static inline void ptlock_cache_init(void) {}
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001795static inline bool ptlock_init(struct page *page) { return true; }
1796static inline void pte_lock_deinit(struct page *page) {}
Kirill A. Shutemov57c1ffc2013-11-14 14:30:45 -08001797#endif /* USE_SPLIT_PTE_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001798
Kirill A. Shutemovb35f1812014-01-21 15:49:07 -08001799static inline void pgtable_init(void)
1800{
1801 ptlock_cache_init();
1802 pgtable_cache_init();
1803}
1804
Kirill A. Shutemov390f44e2013-11-14 14:31:20 -08001805static inline bool pgtable_page_ctor(struct page *page)
Martin Schwidefsky2f569afd2008-02-08 04:22:04 -08001806{
Vladimir Davydov706874e2015-11-05 18:49:27 -08001807 if (!ptlock_init(page))
1808 return false;
Martin Schwidefsky2f569afd2008-02-08 04:22:04 -08001809 inc_zone_page_state(page, NR_PAGETABLE);
Vladimir Davydov706874e2015-11-05 18:49:27 -08001810 return true;
Martin Schwidefsky2f569afd2008-02-08 04:22:04 -08001811}
1812
1813static inline void pgtable_page_dtor(struct page *page)
1814{
1815 pte_lock_deinit(page);
1816 dec_zone_page_state(page, NR_PAGETABLE);
1817}
1818
Hugh Dickinsc74df322005-10-29 18:16:23 -07001819#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1820({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -07001821 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -07001822 pte_t *__pte = pte_offset_map(pmd, address); \
1823 *(ptlp) = __ptl; \
1824 spin_lock(__ptl); \
1825 __pte; \
1826})
1827
1828#define pte_unmap_unlock(pte, ptl) do { \
1829 spin_unlock(ptl); \
1830 pte_unmap(pte); \
1831} while (0)
1832
Kirill A. Shutemov3ed3a4f2016-03-17 14:19:11 -07001833#define pte_alloc(mm, pmd, address) \
1834 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1835
1836#define pte_alloc_map(mm, pmd, address) \
1837 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
Hugh Dickins1bb36302005-10-29 18:16:22 -07001838
Hugh Dickinsc74df322005-10-29 18:16:23 -07001839#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
Kirill A. Shutemov3ed3a4f2016-03-17 14:19:11 -07001840 (pte_alloc(mm, pmd, address) ? \
1841 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
Hugh Dickinsc74df322005-10-29 18:16:23 -07001842
Hugh Dickins1bb36302005-10-29 18:16:22 -07001843#define pte_alloc_kernel(pmd, address) \
Andrea Arcangeli8ac1f832011-01-13 15:46:43 -08001844 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
Hugh Dickins1bb36302005-10-29 18:16:22 -07001845 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001847#if USE_SPLIT_PMD_PTLOCKS
1848
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001849static struct page *pmd_to_page(pmd_t *pmd)
1850{
1851 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1852 return virt_to_page((void *)((unsigned long) pmd & mask));
1853}
1854
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001855static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1856{
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001857 return ptlock_ptr(pmd_to_page(pmd));
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001858}
1859
1860static inline bool pgtable_pmd_page_ctor(struct page *page)
1861{
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001862#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1863 page->pmd_huge_pte = NULL;
1864#endif
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001865 return ptlock_init(page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001866}
1867
1868static inline void pgtable_pmd_page_dtor(struct page *page)
1869{
1870#ifdef CONFIG_TRANSPARENT_HUGEPAGE
Sasha Levin309381fea2014-01-23 15:52:54 -08001871 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001872#endif
Kirill A. Shutemov49076ec2013-11-14 14:31:51 -08001873 ptlock_free(page);
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001874}
1875
Martin Schwidefsky634391a2014-02-13 13:53:33 +01001876#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001877
1878#else
1879
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001880static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1881{
1882 return &mm->page_table_lock;
1883}
1884
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001885static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1886static inline void pgtable_pmd_page_dtor(struct page *page) {}
1887
Kirill A. Shutemovc389a252013-11-14 14:30:59 -08001888#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001889
Kirill A. Shutemove009bb32013-11-14 14:31:07 -08001890#endif
1891
Kirill A. Shutemov9a86cb72013-11-14 14:30:51 -08001892static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1893{
1894 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1895 spin_lock(ptl);
1896 return ptl;
1897}
1898
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001899/*
1900 * No scalability reason to split PUD locks yet, but follow the same pattern
1901 * as the PMD locks to make it easier if we decide to. The VM should not be
1902 * considered ready to switch to split PUD locks yet; there may be places
1903 * which need to be converted from page_table_lock.
1904 */
1905static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1906{
1907 return &mm->page_table_lock;
1908}
Nicholas Piggin62906022016-12-25 13:00:30 +10001909
Matthew Wilcoxa00cc7d2017-02-24 14:57:02 -08001910static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1911{
1912 spinlock_t *ptl = pud_lockptr(mm, pud);
1913
1914 spin_lock(ptl);
1915 return ptl;
1916}
1917
1918extern void __init pagecache_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -07001920extern void free_area_init_node(int nid, unsigned long * zones_size,
1921 unsigned long zone_start_pfn, unsigned long *zholes_size);
David Howells49a7f042012-03-28 18:30:03 +01001922extern void free_initmem(void);
1923
Jiang Liu69afade2013-04-29 15:06:21 -07001924/*
1925 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1926 * into the buddy system. The freed pages will be poisoned with pattern
Jiang Liudbe67df2013-07-03 15:02:51 -07001927 * "poison" if it's within range [0, UCHAR_MAX].
Jiang Liu69afade2013-04-29 15:06:21 -07001928 * Return pages freed into the buddy system.
1929 */
Jiang Liu11199692013-07-03 15:02:48 -07001930extern unsigned long free_reserved_area(void *start, void *end,
Jiang Liu69afade2013-04-29 15:06:21 -07001931 int poison, char *s);
Jiang Liuc3d5f5f2013-07-03 15:03:14 -07001932
Jiang Liucfa11e02013-04-29 15:07:00 -07001933#ifdef CONFIG_HIGHMEM
1934/*
1935 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1936 * and totalram_pages.
1937 */
1938extern void free_highmem_page(struct page *page);
1939#endif
Jiang Liu69afade2013-04-29 15:06:21 -07001940
Jiang Liuc3d5f5f2013-07-03 15:03:14 -07001941extern void adjust_managed_page_count(struct page *page, long count);
Jiang Liu7ee3d4e2013-07-03 15:03:41 -07001942extern void mem_init_print_info(const char *str);
Jiang Liu69afade2013-04-29 15:06:21 -07001943
Stefan Bader4b50bcc2016-05-20 16:58:38 -07001944extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
Nathan Zimmer92923ca2015-06-30 14:56:48 -07001945
Jiang Liu69afade2013-04-29 15:06:21 -07001946/* Free the reserved page into the buddy system, so it gets managed. */
1947static inline void __free_reserved_page(struct page *page)
1948{
1949 ClearPageReserved(page);
1950 init_page_count(page);
1951 __free_page(page);
1952}
1953
1954static inline void free_reserved_page(struct page *page)
1955{
1956 __free_reserved_page(page);
1957 adjust_managed_page_count(page, 1);
1958}
1959
1960static inline void mark_page_reserved(struct page *page)
1961{
1962 SetPageReserved(page);
1963 adjust_managed_page_count(page, -1);
1964}
1965
1966/*
1967 * Default method to free all the __init memory into the buddy system.
Jiang Liudbe67df2013-07-03 15:02:51 -07001968 * The freed pages will be poisoned with pattern "poison" if it's within
1969 * range [0, UCHAR_MAX].
1970 * Return pages freed into the buddy system.
Jiang Liu69afade2013-04-29 15:06:21 -07001971 */
1972static inline unsigned long free_initmem_default(int poison)
1973{
1974 extern char __init_begin[], __init_end[];
1975
Jiang Liu11199692013-07-03 15:02:48 -07001976 return free_reserved_area(&__init_begin, &__init_end,
Jiang Liu69afade2013-04-29 15:06:21 -07001977 poison, "unused kernel");
1978}
1979
Jiang Liu7ee3d4e2013-07-03 15:03:41 -07001980static inline unsigned long get_num_physpages(void)
1981{
1982 int nid;
1983 unsigned long phys_pages = 0;
1984
1985 for_each_online_node(nid)
1986 phys_pages += node_present_pages(nid);
1987
1988 return phys_pages;
1989}
1990
Tejun Heo0ee332c2011-12-08 10:22:09 -08001991#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gormanc7132162006-09-27 01:49:43 -07001992/*
Tejun Heo0ee332c2011-12-08 10:22:09 -08001993 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
Mel Gormanc7132162006-09-27 01:49:43 -07001994 * zones, allocate the backing mem_map and account for memory holes in a more
1995 * architecture independent manner. This is a substitute for creating the
1996 * zone_sizes[] and zholes_size[] arrays and passing them to
1997 * free_area_init_node()
1998 *
1999 * An architecture is expected to register range of page frames backed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08002000 * physical memory with memblock_add[_node]() before calling
Mel Gormanc7132162006-09-27 01:49:43 -07002001 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2002 * usage, an architecture is expected to do something like
2003 *
2004 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2005 * max_highmem_pfn};
2006 * for_each_valid_physical_page_range()
Tejun Heo0ee332c2011-12-08 10:22:09 -08002007 * memblock_add_node(base, size, nid)
Mel Gormanc7132162006-09-27 01:49:43 -07002008 * free_area_init_nodes(max_zone_pfns);
2009 *
Tejun Heo0ee332c2011-12-08 10:22:09 -08002010 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2011 * registered physical page range. Similarly
2012 * sparse_memory_present_with_active_regions() calls memory_present() for
2013 * each range when SPARSEMEM is enabled.
Mel Gormanc7132162006-09-27 01:49:43 -07002014 *
2015 * See mm/page_alloc.c for more information on each function exposed by
Tejun Heo0ee332c2011-12-08 10:22:09 -08002016 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
Mel Gormanc7132162006-09-27 01:49:43 -07002017 */
2018extern void free_area_init_nodes(unsigned long *max_zone_pfn);
Tejun Heo1e019792011-07-12 09:45:34 +02002019unsigned long node_map_pfn_alignment(void);
Yinghai Lu32996252009-12-15 17:59:02 -08002020unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2021 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07002022extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2023 unsigned long end_pfn);
2024extern void get_pfn_range_for_nid(unsigned int nid,
2025 unsigned long *start_pfn, unsigned long *end_pfn);
2026extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07002027extern void free_bootmem_with_active_regions(int nid,
2028 unsigned long max_low_pfn);
2029extern void sparse_memory_present_with_active_regions(int nid);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002030
Tejun Heo0ee332c2011-12-08 10:22:09 -08002031#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
Tejun Heo5dfe8662011-07-14 09:46:10 +02002032
Tejun Heo0ee332c2011-12-08 10:22:09 -08002033#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002034 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
Mel Gorman8a942fd2015-06-30 14:56:55 -07002035static inline int __early_pfn_to_nid(unsigned long pfn,
2036 struct mminit_pfnnid_cache *state)
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002037{
2038 return 0;
2039}
2040#else
2041/* please see mm/page_alloc.c */
2042extern int __meminit early_pfn_to_nid(unsigned long pfn);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002043/* there is a per-arch backend function. */
Mel Gorman8a942fd2015-06-30 14:56:55 -07002044extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2045 struct mminit_pfnnid_cache *state);
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08002046#endif
2047
Pavel Tatashina4a3ede2017-11-15 17:36:31 -08002048#ifdef CONFIG_HAVE_MEMBLOCK
2049void zero_resv_unavail(void);
2050#else
2051static inline void zero_resv_unavail(void) {}
2052#endif
2053
Mel Gorman0e0b8642006-09-27 01:49:56 -07002054extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08002055extern void memmap_init_zone(unsigned long, int, unsigned long,
2056 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07002057extern void setup_per_zone_wmarks(void);
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -07002058extern int __meminit init_per_zone_wmark_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00002060extern void __init mmap_init(void);
Michal Hocko9af744d2017-02-22 15:46:16 -08002061extern void show_mem(unsigned int flags, nodemask_t *nodemask);
Igor Redkod02bd272016-03-17 14:19:05 -07002062extern long si_mem_available(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063extern void si_meminfo(struct sysinfo * val);
2064extern void si_meminfo_node(struct sysinfo *val, int nid);
Srikar Dronamrajuf6f34b42016-10-07 16:59:15 -07002065#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2066extern unsigned long arch_reserved_kernel_pages(void);
2067#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068
Michal Hockoa8e99252017-02-22 15:46:10 -08002069extern __printf(3, 4)
2070void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
Dave Hansena238ab52011-05-24 17:12:16 -07002071
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07002072extern void setup_per_cpu_pageset(void);
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07002073
Shaohua Li112067f2009-09-21 17:01:16 -07002074extern void zone_pcp_update(struct zone *zone);
Jiang Liu340175b2012-07-31 16:43:32 -07002075extern void zone_pcp_reset(struct zone *zone);
Shaohua Li112067f2009-09-21 17:01:16 -07002076
Paul Szabo75f7ad82013-02-22 16:34:42 -08002077/* page_alloc.c */
2078extern int min_free_kbytes;
Johannes Weiner795ae7a2016-03-17 14:19:14 -07002079extern int watermark_scale_factor;
Paul Szabo75f7ad82013-02-22 16:34:42 -08002080
David Howells8feae132009-01-08 12:04:47 +00002081/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07002082extern atomic_long_t mmap_pages_allocated;
David Howells7e660872010-01-15 17:01:39 -08002083extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
David Howells8feae132009-01-08 12:04:47 +00002084
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002085/* interval_tree.c */
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002086void vma_interval_tree_insert(struct vm_area_struct *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002087 struct rb_root_cached *root);
Michel Lespinasse9826a512012-10-08 16:31:35 -07002088void vma_interval_tree_insert_after(struct vm_area_struct *node,
2089 struct vm_area_struct *prev,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002090 struct rb_root_cached *root);
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002091void vma_interval_tree_remove(struct vm_area_struct *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002092 struct rb_root_cached *root);
2093struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002094 unsigned long start, unsigned long last);
2095struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2096 unsigned long start, unsigned long last);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
Michel Lespinasse6b2dbba2012-10-08 16:31:25 -07002098#define vma_interval_tree_foreach(vma, root, start, last) \
2099 for (vma = vma_interval_tree_iter_first(root, start, last); \
2100 vma; vma = vma_interval_tree_iter_next(vma, start, last))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101
Michel Lespinassebf181b92012-10-08 16:31:39 -07002102void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002103 struct rb_root_cached *root);
Michel Lespinassebf181b92012-10-08 16:31:39 -07002104void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
Davidlohr Buesof808c132017-09-08 16:15:08 -07002105 struct rb_root_cached *root);
2106struct anon_vma_chain *
2107anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2108 unsigned long start, unsigned long last);
Michel Lespinassebf181b92012-10-08 16:31:39 -07002109struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2110 struct anon_vma_chain *node, unsigned long start, unsigned long last);
Michel Lespinasseed8ea812012-10-08 16:31:45 -07002111#ifdef CONFIG_DEBUG_VM_RB
2112void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2113#endif
Michel Lespinassebf181b92012-10-08 16:31:39 -07002114
2115#define anon_vma_interval_tree_foreach(avc, root, start, last) \
2116 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2117 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2118
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07002120extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Andrea Arcangelie86f15e2016-10-07 17:01:28 -07002121extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2122 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2123 struct vm_area_struct *expand);
2124static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2125 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2126{
2127 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2128}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129extern struct vm_area_struct *vma_merge(struct mm_struct *,
2130 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2131 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
Andrea Arcangeli19a809a2015-09-04 15:46:24 -07002132 struct mempolicy *, struct vm_userfaultfd_ctx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
David Rientjesdef5efe2017-02-24 14:58:47 -08002134extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2135 unsigned long addr, int new_below);
2136extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2137 unsigned long addr, int new_below);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2139extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2140 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07002141extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
Michel Lespinasse38a76012012-10-08 16:31:50 -07002143 unsigned long addr, unsigned long len, pgoff_t pgoff,
2144 bool *need_rmap_locks);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07002146
Cyrill Gorcunov9c599022014-10-09 15:27:29 -07002147static inline int check_data_rlimit(unsigned long rlim,
2148 unsigned long new,
2149 unsigned long start,
2150 unsigned long end_data,
2151 unsigned long start_data)
2152{
2153 if (rlim < RLIM_INFINITY) {
2154 if (((new - start) + (end_data - start_data)) > rlim)
2155 return -ENOSPC;
2156 }
2157
2158 return 0;
2159}
2160
Andrea Arcangeli7906d002008-07-28 15:46:26 -07002161extern int mm_take_all_locks(struct mm_struct *mm);
2162extern void mm_drop_all_locks(struct mm_struct *mm);
2163
Jiri Slaby38646012011-05-26 16:25:46 -07002164extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2165extern struct file *get_mm_exe_file(struct mm_struct *mm);
Mateusz Guzikcd81a9172016-08-23 16:20:38 +02002166extern struct file *get_task_exe_file(struct task_struct *task);
Matt Helsley925d1c42008-04-29 01:01:36 -07002167
Konstantin Khlebnikov84638332016-01-14 15:22:07 -08002168extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2169extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2170
Dmitry Safonov2eefd872016-09-05 16:33:05 +03002171extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2172 const struct vm_special_mapping *sm);
Stefani Seibold3935ed62014-03-17 23:22:02 +01002173extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2174 unsigned long addr, unsigned long len,
Andy Lutomirskia62c34b2014-05-19 15:58:33 -07002175 unsigned long flags,
2176 const struct vm_special_mapping *spec);
2177/* This is an obsolete alternative to _install_special_mapping. */
Roland McGrathfa5dc222007-02-08 14:20:41 -08002178extern int install_special_mapping(struct mm_struct *mm,
2179 unsigned long addr, unsigned long len,
2180 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181
2182extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2183
Miklos Szeredi0165ab42007-07-15 23:38:26 -07002184extern unsigned long mmap_region(struct file *file, unsigned long addr,
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002185 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2186 struct list_head *uf);
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002187extern unsigned long do_mmap(struct file *file, unsigned long addr,
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08002188 unsigned long len, unsigned long prot, unsigned long flags,
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002189 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2190 struct list_head *uf);
2191extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2192 struct list_head *uf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002194static inline unsigned long
2195do_mmap_pgoff(struct file *file, unsigned long addr,
2196 unsigned long len, unsigned long prot, unsigned long flags,
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002197 unsigned long pgoff, unsigned long *populate,
2198 struct list_head *uf)
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002199{
Mike Rapoport897ab3e2017-02-24 14:58:22 -08002200 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
Oleg Nesterov1fcfd8d2015-09-09 15:39:29 -07002201}
2202
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08002203#ifdef CONFIG_MMU
2204extern int __mm_populate(unsigned long addr, unsigned long len,
2205 int ignore_errors);
2206static inline void mm_populate(unsigned long addr, unsigned long len)
2207{
2208 /* Ignore errors */
2209 (void) __mm_populate(addr, len, 1);
2210}
2211#else
2212static inline void mm_populate(unsigned long addr, unsigned long len) {}
2213#endif
2214
Linus Torvaldse4eb1ff2012-04-20 15:35:40 -07002215/* These take the mm semaphore themselves */
Linus Torvalds5d22fc22016-05-27 15:57:31 -07002216extern int __must_check vm_brk(unsigned long, unsigned long);
Denys Vlasenko16e72e92017-02-22 15:45:16 -08002217extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
Al Virobfce2812012-04-20 21:57:04 -04002218extern int vm_munmap(unsigned long, size_t);
Michal Hocko9fbeb5a2016-05-23 16:25:30 -07002219extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
Linus Torvalds6be5ceb2012-04-20 17:13:58 -07002220 unsigned long, unsigned long,
2221 unsigned long, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222
Michel Lespinassedb4fbfb2012-12-11 16:01:49 -08002223struct vm_unmapped_area_info {
2224#define VM_UNMAPPED_AREA_TOPDOWN 1
2225 unsigned long flags;
2226 unsigned long length;
2227 unsigned long low_limit;
2228 unsigned long high_limit;
2229 unsigned long align_mask;
2230 unsigned long align_offset;
2231};
2232
2233extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2234extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2235
2236/*
2237 * Search for an unmapped address range.
2238 *
2239 * We are looking for a range that:
2240 * - does not intersect with any VMA;
2241 * - is contained within the [low_limit, high_limit) interval;
2242 * - is at least the desired size.
2243 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2244 */
2245static inline unsigned long
2246vm_unmapped_area(struct vm_unmapped_area_info *info)
2247{
Borislav Petkovcdd78752015-04-15 16:14:47 -07002248 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
Michel Lespinassedb4fbfb2012-12-11 16:01:49 -08002249 return unmapped_area_topdown(info);
Borislav Petkovcdd78752015-04-15 16:14:47 -07002250 else
2251 return unmapped_area(info);
Michel Lespinassedb4fbfb2012-12-11 16:01:49 -08002252}
2253
Hugh Dickins85821aa2011-07-25 17:12:23 -07002254/* truncate.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08002256extern void truncate_inode_pages_range(struct address_space *,
2257 loff_t lstart, loff_t lend);
Johannes Weiner91b0abe2014-04-03 14:47:49 -07002258extern void truncate_inode_pages_final(struct address_space *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259
2260/* generic vm_area_ops exported for stackable file systems */
Dave Jiang11bac802017-02-24 14:56:41 -08002261extern int filemap_fault(struct vm_fault *vmf);
Jan Kara82b0f8c2016-12-14 15:06:58 -08002262extern void filemap_map_pages(struct vm_fault *vmf,
Kirill A. Shutemovbae473a2016-07-26 15:25:20 -07002263 pgoff_t start_pgoff, pgoff_t end_pgoff);
Dave Jiang11bac802017-02-24 14:56:41 -08002264extern int filemap_page_mkwrite(struct vm_fault *vmf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265
2266/* mm/page-writeback.c */
Jeff Layton2b69c822017-07-05 15:26:48 -04002267int __must_check write_one_page(struct page *page);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08002268void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002269
2270/* readahead.c */
2271#define VM_MAX_READAHEAD 128 /* kbytes */
2272#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002273
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08002275 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07002276
2277void page_cache_sync_readahead(struct address_space *mapping,
2278 struct file_ra_state *ra,
2279 struct file *filp,
2280 pgoff_t offset,
2281 unsigned long size);
2282
2283void page_cache_async_readahead(struct address_space *mapping,
2284 struct file_ra_state *ra,
2285 struct file *filp,
2286 struct page *pg,
2287 pgoff_t offset,
2288 unsigned long size);
2289
Hugh Dickins1be71072017-06-19 04:03:24 -07002290extern unsigned long stack_guard_gap;
Michal Hockod05f3162011-05-24 17:11:44 -07002291/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07002292extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Michal Hockod05f3162011-05-24 17:11:44 -07002293
2294/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2295extern int expand_downwards(struct vm_area_struct *vma,
2296 unsigned long address);
Luck, Tony8ca3eb082010-08-24 11:44:18 -07002297#if VM_GROWSUP
Hugh Dickins46dea3d2005-10-29 18:16:20 -07002298extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Luck, Tony8ca3eb082010-08-24 11:44:18 -07002299#else
Linus Torvaldsfee7e492015-01-06 13:00:05 -08002300 #define expand_upwards(vma, address) (0)
Matthew Wilcox9ab88512005-11-18 16:16:42 -05002301#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302
2303/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2304extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2305extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2306 struct vm_area_struct **pprev);
2307
2308/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2309 NULL if none. Assume start_addr < end_addr. */
2310static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2311{
2312 struct vm_area_struct * vma = find_vma(mm,start_addr);
2313
2314 if (vma && end_addr <= vma->vm_start)
2315 vma = NULL;
2316 return vma;
2317}
2318
Hugh Dickins1be71072017-06-19 04:03:24 -07002319static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2320{
2321 unsigned long vm_start = vma->vm_start;
2322
2323 if (vma->vm_flags & VM_GROWSDOWN) {
2324 vm_start -= stack_guard_gap;
2325 if (vm_start > vma->vm_start)
2326 vm_start = 0;
2327 }
2328 return vm_start;
2329}
2330
2331static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2332{
2333 unsigned long vm_end = vma->vm_end;
2334
2335 if (vma->vm_flags & VM_GROWSUP) {
2336 vm_end += stack_guard_gap;
2337 if (vm_end < vma->vm_end)
2338 vm_end = -PAGE_SIZE;
2339 }
2340 return vm_end;
2341}
2342
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343static inline unsigned long vma_pages(struct vm_area_struct *vma)
2344{
2345 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2346}
2347
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002348/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2349static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2350 unsigned long vm_start, unsigned long vm_end)
2351{
2352 struct vm_area_struct *vma = find_vma(mm, vm_start);
2353
2354 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2355 vma = NULL;
2356
2357 return vma;
2358}
2359
David Howellsbad849b2010-08-26 16:00:34 +01002360#ifdef CONFIG_MMU
Hugh Dickins804af2c2006-07-26 21:39:49 +01002361pgprot_t vm_get_page_prot(unsigned long vm_flags);
Peter Feiner64e455072014-10-13 15:55:46 -07002362void vma_set_page_prot(struct vm_area_struct *vma);
David Howellsbad849b2010-08-26 16:00:34 +01002363#else
2364static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2365{
2366 return __pgprot(0);
2367}
Peter Feiner64e455072014-10-13 15:55:46 -07002368static inline void vma_set_page_prot(struct vm_area_struct *vma)
2369{
2370 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2371}
David Howellsbad849b2010-08-26 16:00:34 +01002372#endif
2373
Aneesh Kumar K.V58772312013-12-06 00:08:22 +05302374#ifdef CONFIG_NUMA_BALANCING
Mel Gorman4b10e7d2012-10-25 14:16:32 +02002375unsigned long change_prot_numa(struct vm_area_struct *vma,
Lee Schermerhornb24f53a2012-10-25 14:16:32 +02002376 unsigned long start, unsigned long end);
2377#endif
2378
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07002379struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07002380int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2381 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08002382int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08002383int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2384 unsigned long pfn);
Andy Lutomirski1745cbc2015-12-29 20:12:20 -08002385int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2386 unsigned long pfn, pgprot_t pgprot);
Nick Piggin423bad602008-04-28 02:13:01 -07002387int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
Dan Williams01c8f1c2016-01-15 16:56:40 -08002388 pfn_t pfn);
Ross Zwislerb2770da62017-09-06 16:18:35 -07002389int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2390 pfn_t pfn);
Linus Torvaldsb4cbb192013-04-16 13:45:37 -07002391int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2392
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393
Michel Lespinasse240aade2013-02-22 16:35:56 -08002394struct page *follow_page_mask(struct vm_area_struct *vma,
2395 unsigned long address, unsigned int foll_flags,
2396 unsigned int *page_mask);
2397
2398static inline struct page *follow_page(struct vm_area_struct *vma,
2399 unsigned long address, unsigned int foll_flags)
2400{
2401 unsigned int unused_page_mask;
2402 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2403}
2404
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07002405#define FOLL_WRITE 0x01 /* check pte is writable */
2406#define FOLL_TOUCH 0x02 /* mark page accessed */
2407#define FOLL_GET 0x04 /* do get_page on page */
Hugh Dickins8e4b9a62009-09-21 17:03:26 -07002408#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
Hugh Dickins58fa8792009-09-21 17:03:31 -07002409#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
Gleb Natapov318b2752011-03-22 16:30:51 -07002410#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2411 * and return without waiting upon it */
Kirill A. Shutemov84d33df2015-04-14 15:44:37 -07002412#define FOLL_POPULATE 0x40 /* fault in page */
Andrea Arcangeli500d65d2011-01-13 15:46:55 -08002413#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
Huang Ying69ebb832011-01-30 11:15:48 +08002414#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
Andrea Arcangeli0b9d7052012-10-05 21:36:27 +02002415#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
Hugh Dickins5117b3b2013-02-22 16:36:07 -08002416#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
Andres Lagar-Cavilla234b2392014-09-17 10:51:48 -07002417#define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
Eric B Munsonde60f5f2015-11-05 18:51:36 -08002418#define FOLL_MLOCK 0x1000 /* lock present pages */
Dave Hansen1e987792016-02-12 13:01:54 -08002419#define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
Linus Torvalds19be0ea2016-10-13 13:07:36 -07002420#define FOLL_COW 0x4000 /* internal GUP flag */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421
James Morse9a291a72017-06-02 14:46:46 -07002422static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2423{
2424 if (vm_fault & VM_FAULT_OOM)
2425 return -ENOMEM;
2426 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2427 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2428 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2429 return -EFAULT;
2430 return 0;
2431}
2432
Martin Schwidefsky2f569afd2008-02-08 04:22:04 -08002433typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07002434 void *data);
2435extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2436 unsigned long size, pte_fn_t fn, void *data);
2437
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438
Laura Abbott8823b1d2016-03-15 14:56:27 -07002439#ifdef CONFIG_PAGE_POISONING
2440extern bool page_poisoning_enabled(void);
2441extern void kernel_poison_pages(struct page *page, int numpages, int enable);
Laura Abbott1414c7f2016-03-15 14:56:30 -07002442extern bool page_is_poisoned(struct page *page);
Laura Abbott8823b1d2016-03-15 14:56:27 -07002443#else
2444static inline bool page_poisoning_enabled(void) { return false; }
2445static inline void kernel_poison_pages(struct page *page, int numpages,
2446 int enable) { }
Laura Abbott1414c7f2016-03-15 14:56:30 -07002447static inline bool page_is_poisoned(struct page *page) { return false; }
Laura Abbott8823b1d2016-03-15 14:56:27 -07002448#endif
2449
Ingo Molnar12d6f212008-01-30 13:33:58 +01002450#ifdef CONFIG_DEBUG_PAGEALLOC
Joonsoo Kim031bc572014-12-12 16:55:52 -08002451extern bool _debug_pagealloc_enabled;
2452extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2453
2454static inline bool debug_pagealloc_enabled(void)
2455{
2456 return _debug_pagealloc_enabled;
2457}
2458
2459static inline void
2460kernel_map_pages(struct page *page, int numpages, int enable)
2461{
2462 if (!debug_pagealloc_enabled())
2463 return;
2464
2465 __kernel_map_pages(page, numpages, enable);
2466}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01002467#ifdef CONFIG_HIBERNATION
2468extern bool kernel_page_present(struct page *page);
Joonsoo Kim40b44132016-03-15 14:54:21 -07002469#endif /* CONFIG_HIBERNATION */
2470#else /* CONFIG_DEBUG_PAGEALLOC */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07002472kernel_map_pages(struct page *page, int numpages, int enable) {}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01002473#ifdef CONFIG_HIBERNATION
2474static inline bool kernel_page_present(struct page *page) { return true; }
Joonsoo Kim40b44132016-03-15 14:54:21 -07002475#endif /* CONFIG_HIBERNATION */
2476static inline bool debug_pagealloc_enabled(void)
2477{
2478 return false;
2479}
2480#endif /* CONFIG_DEBUG_PAGEALLOC */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002481
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002482#ifdef __HAVE_ARCH_GATE_AREA
Stephen Wilson31db58b2011-03-13 15:49:15 -04002483extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002484extern int in_gate_area_no_mm(unsigned long addr);
2485extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002486#else
Andy Lutomirskia6c19df2014-08-08 14:23:40 -07002487static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2488{
2489 return NULL;
2490}
2491static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2492static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2493{
2494 return 0;
2495}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496#endif /* __HAVE_ARCH_GATE_AREA */
2497
Michal Hocko44a70ade2016-07-28 15:44:43 -07002498extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2499
Josh Triplett146732c2013-04-29 15:07:22 -07002500#ifdef CONFIG_SYSCTL
2501extern int sysctl_drop_caches;
Alexey Dobriyan8d65af72009-09-23 15:57:19 -07002502int drop_caches_sysctl_handler(struct ctl_table *, int,
Andrew Morton9d0243b2006-01-08 01:00:39 -08002503 void __user *, size_t *, loff_t *);
Josh Triplett146732c2013-04-29 15:07:22 -07002504#endif
2505
Vladimir Davydovcb731d62015-02-12 14:58:54 -08002506void drop_slab(void);
2507void drop_slab_node(int nid);
Andrew Morton9d0243b2006-01-08 01:00:39 -08002508
Luke Yang7a9166e2006-02-20 18:28:07 -08002509#ifndef CONFIG_MMU
2510#define randomize_va_space 0
2511#else
Andi Kleena62eaf12006-02-16 23:41:58 +01002512extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08002513#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01002514
Sam Ravnborg045e72a2007-07-26 10:41:13 -07002515const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01002516void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07002517
Yinghai Lu9bdac912010-02-10 01:20:22 -08002518void sparse_mem_maps_populate_node(struct page **map_map,
2519 unsigned long pnum_begin,
2520 unsigned long pnum_end,
2521 unsigned long map_count,
2522 int nodeid);
2523
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07002524struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07002525pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
Kirill A. Shutemovc2febaf2017-03-09 17:24:07 +03002526p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2527pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
Andy Whitcroft29c71112007-10-16 01:24:14 -07002528pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2529pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07002530void *vmemmap_alloc_block(unsigned long size, int node);
Dan Williams4b94ffd2016-01-15 16:56:22 -08002531struct vmem_altmap;
2532void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2533 struct vmem_altmap *altmap);
2534static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2535{
2536 return __vmemmap_alloc_block_buf(size, node, NULL);
2537}
2538
Christoph Lameter8f6aac42007-10-16 01:24:13 -07002539void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Johannes Weiner0aad8182013-04-29 15:07:50 -07002540int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2541 int node);
2542int vmemmap_populate(unsigned long start, unsigned long end, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07002543void vmemmap_populate_print_last(void);
Tang Chen01975182013-02-22 16:33:08 -08002544#ifdef CONFIG_MEMORY_HOTPLUG
Johannes Weiner0aad8182013-04-29 15:07:50 -07002545void vmemmap_free(unsigned long start, unsigned long end);
Tang Chen01975182013-02-22 16:33:08 -08002546#endif
Yasuaki Ishimatsu46723bf2013-02-22 16:33:00 -08002547void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
Baoquan He15670bf2017-10-28 09:30:38 +08002548 unsigned long nr_pages);
Andi Kleen6a460792009-09-16 11:50:15 +02002549
Andi Kleen82ba0112009-12-16 12:19:57 +01002550enum mf_flags {
2551 MF_COUNT_INCREASED = 1 << 0,
Tony Luck7329bbe2011-12-13 09:27:58 -08002552 MF_ACTION_REQUIRED = 1 << 1,
Tony Luck6751ed62012-07-11 10:20:47 -07002553 MF_MUST_KILL = 1 << 2,
Naveen N. Raocf870c72013-07-10 14:57:01 +05302554 MF_SOFT_OFFLINE = 1 << 3,
Andi Kleen82ba0112009-12-16 12:19:57 +01002555};
Tony Luckcd42f4a2011-12-15 10:48:12 -08002556extern int memory_failure(unsigned long pfn, int trapno, int flags);
Huang Yingea8f5fb2011-07-13 13:14:27 +08002557extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
Wu Fengguang847ce402009-12-16 12:19:58 +01002558extern int unpoison_memory(unsigned long pfn);
Naoya Horiguchiead07f62015-06-24 16:56:48 -07002559extern int get_hwpoison_page(struct page *page);
Naoya Horiguchi4e41a302016-01-15 16:54:07 -08002560#define put_hwpoison_page(page) put_page(page)
Andi Kleen6a460792009-09-16 11:50:15 +02002561extern int sysctl_memory_failure_early_kill;
2562extern int sysctl_memory_failure_recovery;
Andi Kleenfacb6012009-12-16 12:20:00 +01002563extern void shake_page(struct page *p, int access);
Xishi Qiu293c07e2013-02-22 16:34:02 -08002564extern atomic_long_t num_poisoned_pages;
Andi Kleenfacb6012009-12-16 12:20:00 +01002565extern int soft_offline_page(struct page *page, int flags);
Andi Kleen6a460792009-09-16 11:50:15 +02002566
Xie XiuQicc637b12015-06-24 16:57:30 -07002567
2568/*
2569 * Error handlers for various types of pages.
2570 */
Xie XiuQicc3e2af2015-06-24 16:57:33 -07002571enum mf_result {
Xie XiuQicc637b12015-06-24 16:57:30 -07002572 MF_IGNORED, /* Error: cannot be handled */
2573 MF_FAILED, /* Error: handling failed */
2574 MF_DELAYED, /* Will be handled later */
2575 MF_RECOVERED, /* Successfully recovered */
2576};
2577
2578enum mf_action_page_type {
2579 MF_MSG_KERNEL,
2580 MF_MSG_KERNEL_HIGH_ORDER,
2581 MF_MSG_SLAB,
2582 MF_MSG_DIFFERENT_COMPOUND,
2583 MF_MSG_POISONED_HUGE,
2584 MF_MSG_HUGE,
2585 MF_MSG_FREE_HUGE,
2586 MF_MSG_UNMAP_FAILED,
2587 MF_MSG_DIRTY_SWAPCACHE,
2588 MF_MSG_CLEAN_SWAPCACHE,
2589 MF_MSG_DIRTY_MLOCKED_LRU,
2590 MF_MSG_CLEAN_MLOCKED_LRU,
2591 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2592 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2593 MF_MSG_DIRTY_LRU,
2594 MF_MSG_CLEAN_LRU,
2595 MF_MSG_TRUNCATED_LRU,
2596 MF_MSG_BUDDY,
2597 MF_MSG_BUDDY_2ND,
2598 MF_MSG_UNKNOWN,
2599};
2600
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08002601#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2602extern void clear_huge_page(struct page *page,
Huang Yingc79b57e2017-09-06 16:25:04 -07002603 unsigned long addr_hint,
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08002604 unsigned int pages_per_huge_page);
2605extern void copy_user_huge_page(struct page *dst, struct page *src,
2606 unsigned long addr, struct vm_area_struct *vma,
2607 unsigned int pages_per_huge_page);
Mike Kravetzfa4d75c2017-02-22 15:42:49 -08002608extern long copy_huge_page_from_user(struct page *dst_page,
2609 const void __user *usr_src,
Mike Kravetz810a56b2017-02-22 15:42:58 -08002610 unsigned int pages_per_huge_page,
2611 bool allow_pagefault);
Andrea Arcangeli47ad8472011-01-13 15:46:47 -08002612#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2613
Joonsoo Kime30825f2014-12-12 16:55:49 -08002614extern struct page_ext_operations debug_guardpage_ops;
Joonsoo Kime30825f2014-12-12 16:55:49 -08002615
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002616#ifdef CONFIG_DEBUG_PAGEALLOC
2617extern unsigned int _debug_guardpage_minorder;
Joonsoo Kime30825f2014-12-12 16:55:49 -08002618extern bool _debug_guardpage_enabled;
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002619
2620static inline unsigned int debug_guardpage_minorder(void)
2621{
2622 return _debug_guardpage_minorder;
2623}
2624
Joonsoo Kime30825f2014-12-12 16:55:49 -08002625static inline bool debug_guardpage_enabled(void)
2626{
2627 return _debug_guardpage_enabled;
2628}
2629
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002630static inline bool page_is_guard(struct page *page)
2631{
Joonsoo Kime30825f2014-12-12 16:55:49 -08002632 struct page_ext *page_ext;
2633
2634 if (!debug_guardpage_enabled())
2635 return false;
2636
2637 page_ext = lookup_page_ext(page);
Yang Shi0bb2fd132016-05-20 16:58:59 -07002638 if (unlikely(!page_ext))
2639 return false;
2640
Joonsoo Kime30825f2014-12-12 16:55:49 -08002641 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002642}
2643#else
2644static inline unsigned int debug_guardpage_minorder(void) { return 0; }
Joonsoo Kime30825f2014-12-12 16:55:49 -08002645static inline bool debug_guardpage_enabled(void) { return false; }
Stanislaw Gruszkac0a32fc2012-01-10 15:07:28 -08002646static inline bool page_is_guard(struct page *page) { return false; }
2647#endif /* CONFIG_DEBUG_PAGEALLOC */
2648
Cody P Schaferf9872ca2013-04-29 15:08:01 -07002649#if MAX_NUMNODES > 1
2650void __init setup_nr_node_ids(void);
2651#else
2652static inline void setup_nr_node_ids(void) {}
2653#endif
2654
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655#endif /* __KERNEL__ */
2656#endif /* _LINUX_MM_H */