Commit 29f175d125f0f3a9503af8a5596f93d714cceb08
Committed by
Linus Torvalds
1 parent
55f67141a8
Exists in
master
and in
13 other branches
mm/readahead.c: inline ra_submit
Commit f9acc8c7b35a ("readahead: sanify file_ra_state names") left ra_submit with a single function call. Move ra_submit to internal.h and inline it to save some stack. Thanks to Andrew Morton for commenting different versions. Signed-off-by: Fabian Frederick <fabf@skynet.be> Suggested-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Showing 3 changed files with 18 additions and 21 deletions Inline Diff
include/linux/mm.h
1 | #ifndef _LINUX_MM_H | 1 | #ifndef _LINUX_MM_H |
2 | #define _LINUX_MM_H | 2 | #define _LINUX_MM_H |
3 | 3 | ||
4 | #include <linux/errno.h> | 4 | #include <linux/errno.h> |
5 | 5 | ||
6 | #ifdef __KERNEL__ | 6 | #ifdef __KERNEL__ |
7 | 7 | ||
8 | #include <linux/mmdebug.h> | 8 | #include <linux/mmdebug.h> |
9 | #include <linux/gfp.h> | 9 | #include <linux/gfp.h> |
10 | #include <linux/bug.h> | 10 | #include <linux/bug.h> |
11 | #include <linux/list.h> | 11 | #include <linux/list.h> |
12 | #include <linux/mmzone.h> | 12 | #include <linux/mmzone.h> |
13 | #include <linux/rbtree.h> | 13 | #include <linux/rbtree.h> |
14 | #include <linux/atomic.h> | 14 | #include <linux/atomic.h> |
15 | #include <linux/debug_locks.h> | 15 | #include <linux/debug_locks.h> |
16 | #include <linux/mm_types.h> | 16 | #include <linux/mm_types.h> |
17 | #include <linux/range.h> | 17 | #include <linux/range.h> |
18 | #include <linux/pfn.h> | 18 | #include <linux/pfn.h> |
19 | #include <linux/bit_spinlock.h> | 19 | #include <linux/bit_spinlock.h> |
20 | #include <linux/shrinker.h> | 20 | #include <linux/shrinker.h> |
21 | 21 | ||
22 | struct mempolicy; | 22 | struct mempolicy; |
23 | struct anon_vma; | 23 | struct anon_vma; |
24 | struct anon_vma_chain; | 24 | struct anon_vma_chain; |
25 | struct file_ra_state; | 25 | struct file_ra_state; |
26 | struct user_struct; | 26 | struct user_struct; |
27 | struct writeback_control; | 27 | struct writeback_control; |
28 | 28 | ||
29 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ | 29 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ |
30 | extern unsigned long max_mapnr; | 30 | extern unsigned long max_mapnr; |
31 | 31 | ||
32 | static inline void set_max_mapnr(unsigned long limit) | 32 | static inline void set_max_mapnr(unsigned long limit) |
33 | { | 33 | { |
34 | max_mapnr = limit; | 34 | max_mapnr = limit; |
35 | } | 35 | } |
36 | #else | 36 | #else |
37 | static inline void set_max_mapnr(unsigned long limit) { } | 37 | static inline void set_max_mapnr(unsigned long limit) { } |
38 | #endif | 38 | #endif |
39 | 39 | ||
40 | extern unsigned long totalram_pages; | 40 | extern unsigned long totalram_pages; |
41 | extern void * high_memory; | 41 | extern void * high_memory; |
42 | extern int page_cluster; | 42 | extern int page_cluster; |
43 | 43 | ||
44 | #ifdef CONFIG_SYSCTL | 44 | #ifdef CONFIG_SYSCTL |
45 | extern int sysctl_legacy_va_layout; | 45 | extern int sysctl_legacy_va_layout; |
46 | #else | 46 | #else |
47 | #define sysctl_legacy_va_layout 0 | 47 | #define sysctl_legacy_va_layout 0 |
48 | #endif | 48 | #endif |
49 | 49 | ||
50 | #include <asm/page.h> | 50 | #include <asm/page.h> |
51 | #include <asm/pgtable.h> | 51 | #include <asm/pgtable.h> |
52 | #include <asm/processor.h> | 52 | #include <asm/processor.h> |
53 | 53 | ||
54 | #ifndef __pa_symbol | 54 | #ifndef __pa_symbol |
55 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) | 55 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) |
56 | #endif | 56 | #endif |
57 | 57 | ||
58 | extern unsigned long sysctl_user_reserve_kbytes; | 58 | extern unsigned long sysctl_user_reserve_kbytes; |
59 | extern unsigned long sysctl_admin_reserve_kbytes; | 59 | extern unsigned long sysctl_admin_reserve_kbytes; |
60 | 60 | ||
61 | extern int sysctl_overcommit_memory; | 61 | extern int sysctl_overcommit_memory; |
62 | extern int sysctl_overcommit_ratio; | 62 | extern int sysctl_overcommit_ratio; |
63 | extern unsigned long sysctl_overcommit_kbytes; | 63 | extern unsigned long sysctl_overcommit_kbytes; |
64 | 64 | ||
65 | extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, | 65 | extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, |
66 | size_t *, loff_t *); | 66 | size_t *, loff_t *); |
67 | extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, | 67 | extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, |
68 | size_t *, loff_t *); | 68 | size_t *, loff_t *); |
69 | 69 | ||
70 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) | 70 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
71 | 71 | ||
72 | /* to align the pointer to the (next) page boundary */ | 72 | /* to align the pointer to the (next) page boundary */ |
73 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | 73 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) |
74 | 74 | ||
75 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ | 75 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ |
76 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE) | 76 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE) |
77 | 77 | ||
78 | /* | 78 | /* |
79 | * Linux kernel virtual memory manager primitives. | 79 | * Linux kernel virtual memory manager primitives. |
80 | * The idea being to have a "virtual" mm in the same way | 80 | * The idea being to have a "virtual" mm in the same way |
81 | * we have a virtual fs - giving a cleaner interface to the | 81 | * we have a virtual fs - giving a cleaner interface to the |
82 | * mm details, and allowing different kinds of memory mappings | 82 | * mm details, and allowing different kinds of memory mappings |
83 | * (from shared memory to executable loading to arbitrary | 83 | * (from shared memory to executable loading to arbitrary |
84 | * mmap() functions). | 84 | * mmap() functions). |
85 | */ | 85 | */ |
86 | 86 | ||
87 | extern struct kmem_cache *vm_area_cachep; | 87 | extern struct kmem_cache *vm_area_cachep; |
88 | 88 | ||
89 | #ifndef CONFIG_MMU | 89 | #ifndef CONFIG_MMU |
90 | extern struct rb_root nommu_region_tree; | 90 | extern struct rb_root nommu_region_tree; |
91 | extern struct rw_semaphore nommu_region_sem; | 91 | extern struct rw_semaphore nommu_region_sem; |
92 | 92 | ||
93 | extern unsigned int kobjsize(const void *objp); | 93 | extern unsigned int kobjsize(const void *objp); |
94 | #endif | 94 | #endif |
95 | 95 | ||
96 | /* | 96 | /* |
97 | * vm_flags in vm_area_struct, see mm_types.h. | 97 | * vm_flags in vm_area_struct, see mm_types.h. |
98 | */ | 98 | */ |
99 | #define VM_NONE 0x00000000 | 99 | #define VM_NONE 0x00000000 |
100 | 100 | ||
101 | #define VM_READ 0x00000001 /* currently active flags */ | 101 | #define VM_READ 0x00000001 /* currently active flags */ |
102 | #define VM_WRITE 0x00000002 | 102 | #define VM_WRITE 0x00000002 |
103 | #define VM_EXEC 0x00000004 | 103 | #define VM_EXEC 0x00000004 |
104 | #define VM_SHARED 0x00000008 | 104 | #define VM_SHARED 0x00000008 |
105 | 105 | ||
106 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ | 106 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
107 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ | 107 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
108 | #define VM_MAYWRITE 0x00000020 | 108 | #define VM_MAYWRITE 0x00000020 |
109 | #define VM_MAYEXEC 0x00000040 | 109 | #define VM_MAYEXEC 0x00000040 |
110 | #define VM_MAYSHARE 0x00000080 | 110 | #define VM_MAYSHARE 0x00000080 |
111 | 111 | ||
112 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | 112 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ |
113 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ | 113 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
114 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ | 114 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
115 | 115 | ||
116 | #define VM_LOCKED 0x00002000 | 116 | #define VM_LOCKED 0x00002000 |
117 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | 117 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ |
118 | 118 | ||
119 | /* Used by sys_madvise() */ | 119 | /* Used by sys_madvise() */ |
120 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | 120 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ |
121 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | 121 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ |
122 | 122 | ||
123 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | 123 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ |
124 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | 124 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ |
125 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ | 125 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
126 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ | 126 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
127 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ | 127 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
128 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ | 128 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ |
129 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ | 129 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ |
130 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ | 130 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ |
131 | 131 | ||
132 | #ifdef CONFIG_MEM_SOFT_DIRTY | 132 | #ifdef CONFIG_MEM_SOFT_DIRTY |
133 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ | 133 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ |
134 | #else | 134 | #else |
135 | # define VM_SOFTDIRTY 0 | 135 | # define VM_SOFTDIRTY 0 |
136 | #endif | 136 | #endif |
137 | 137 | ||
138 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ | 138 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
139 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ | 139 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ |
140 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ | 140 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ |
141 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ | 141 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
142 | 142 | ||
143 | #if defined(CONFIG_X86) | 143 | #if defined(CONFIG_X86) |
144 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ | 144 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ |
145 | #elif defined(CONFIG_PPC) | 145 | #elif defined(CONFIG_PPC) |
146 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ | 146 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ |
147 | #elif defined(CONFIG_PARISC) | 147 | #elif defined(CONFIG_PARISC) |
148 | # define VM_GROWSUP VM_ARCH_1 | 148 | # define VM_GROWSUP VM_ARCH_1 |
149 | #elif defined(CONFIG_METAG) | 149 | #elif defined(CONFIG_METAG) |
150 | # define VM_GROWSUP VM_ARCH_1 | 150 | # define VM_GROWSUP VM_ARCH_1 |
151 | #elif defined(CONFIG_IA64) | 151 | #elif defined(CONFIG_IA64) |
152 | # define VM_GROWSUP VM_ARCH_1 | 152 | # define VM_GROWSUP VM_ARCH_1 |
153 | #elif !defined(CONFIG_MMU) | 153 | #elif !defined(CONFIG_MMU) |
154 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ | 154 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ |
155 | #endif | 155 | #endif |
156 | 156 | ||
157 | #ifndef VM_GROWSUP | 157 | #ifndef VM_GROWSUP |
158 | # define VM_GROWSUP VM_NONE | 158 | # define VM_GROWSUP VM_NONE |
159 | #endif | 159 | #endif |
160 | 160 | ||
161 | /* Bits set in the VMA until the stack is in its final location */ | 161 | /* Bits set in the VMA until the stack is in its final location */ |
162 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | 162 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) |
163 | 163 | ||
164 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ | 164 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
165 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | 165 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS |
166 | #endif | 166 | #endif |
167 | 167 | ||
168 | #ifdef CONFIG_STACK_GROWSUP | 168 | #ifdef CONFIG_STACK_GROWSUP |
169 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | 169 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
170 | #else | 170 | #else |
171 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | 171 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
172 | #endif | 172 | #endif |
173 | 173 | ||
174 | /* | 174 | /* |
175 | * Special vmas that are non-mergable, non-mlock()able. | 175 | * Special vmas that are non-mergable, non-mlock()able. |
176 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. | 176 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. |
177 | */ | 177 | */ |
178 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) | 178 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) |
179 | 179 | ||
180 | /* This mask defines which mm->def_flags a process can inherit its parent */ | 180 | /* This mask defines which mm->def_flags a process can inherit its parent */ |
181 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE | 181 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE |
182 | 182 | ||
183 | /* | 183 | /* |
184 | * mapping from the currently active vm_flags protection bits (the | 184 | * mapping from the currently active vm_flags protection bits (the |
185 | * low four bits) to a page protection mask.. | 185 | * low four bits) to a page protection mask.. |
186 | */ | 186 | */ |
187 | extern pgprot_t protection_map[16]; | 187 | extern pgprot_t protection_map[16]; |
188 | 188 | ||
189 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ | 189 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ |
190 | #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ | 190 | #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ |
191 | #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ | 191 | #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ |
192 | #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ | 192 | #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ |
193 | #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ | 193 | #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ |
194 | #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */ | 194 | #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */ |
195 | #define FAULT_FLAG_TRIED 0x40 /* second try */ | 195 | #define FAULT_FLAG_TRIED 0x40 /* second try */ |
196 | #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */ | 196 | #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */ |
197 | 197 | ||
198 | /* | 198 | /* |
199 | * vm_fault is filled by the the pagefault handler and passed to the vma's | 199 | * vm_fault is filled by the the pagefault handler and passed to the vma's |
200 | * ->fault function. The vma's ->fault is responsible for returning a bitmask | 200 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
201 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | 201 | * of VM_FAULT_xxx flags that give details about how the fault was handled. |
202 | * | 202 | * |
203 | * pgoff should be used in favour of virtual_address, if possible. If pgoff | 203 | * pgoff should be used in favour of virtual_address, if possible. If pgoff |
204 | * is used, one may implement ->remap_pages to get nonlinear mapping support. | 204 | * is used, one may implement ->remap_pages to get nonlinear mapping support. |
205 | */ | 205 | */ |
206 | struct vm_fault { | 206 | struct vm_fault { |
207 | unsigned int flags; /* FAULT_FLAG_xxx flags */ | 207 | unsigned int flags; /* FAULT_FLAG_xxx flags */ |
208 | pgoff_t pgoff; /* Logical page offset based on vma */ | 208 | pgoff_t pgoff; /* Logical page offset based on vma */ |
209 | void __user *virtual_address; /* Faulting virtual address */ | 209 | void __user *virtual_address; /* Faulting virtual address */ |
210 | 210 | ||
211 | struct page *page; /* ->fault handlers should return a | 211 | struct page *page; /* ->fault handlers should return a |
212 | * page here, unless VM_FAULT_NOPAGE | 212 | * page here, unless VM_FAULT_NOPAGE |
213 | * is set (which is also implied by | 213 | * is set (which is also implied by |
214 | * VM_FAULT_ERROR). | 214 | * VM_FAULT_ERROR). |
215 | */ | 215 | */ |
216 | /* for ->map_pages() only */ | 216 | /* for ->map_pages() only */ |
217 | pgoff_t max_pgoff; /* map pages for offset from pgoff till | 217 | pgoff_t max_pgoff; /* map pages for offset from pgoff till |
218 | * max_pgoff inclusive */ | 218 | * max_pgoff inclusive */ |
219 | pte_t *pte; /* pte entry associated with ->pgoff */ | 219 | pte_t *pte; /* pte entry associated with ->pgoff */ |
220 | }; | 220 | }; |
221 | 221 | ||
222 | /* | 222 | /* |
223 | * These are the virtual MM functions - opening of an area, closing and | 223 | * These are the virtual MM functions - opening of an area, closing and |
224 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | 224 | * unmapping it (needed to keep files on disk up-to-date etc), pointer |
225 | * to the functions called when a no-page or a wp-page exception occurs. | 225 | * to the functions called when a no-page or a wp-page exception occurs. |
226 | */ | 226 | */ |
227 | struct vm_operations_struct { | 227 | struct vm_operations_struct { |
228 | void (*open)(struct vm_area_struct * area); | 228 | void (*open)(struct vm_area_struct * area); |
229 | void (*close)(struct vm_area_struct * area); | 229 | void (*close)(struct vm_area_struct * area); |
230 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); | 230 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); |
231 | void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf); | 231 | void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf); |
232 | 232 | ||
233 | /* notification that a previously read-only page is about to become | 233 | /* notification that a previously read-only page is about to become |
234 | * writable, if an error is returned it will cause a SIGBUS */ | 234 | * writable, if an error is returned it will cause a SIGBUS */ |
235 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); | 235 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); |
236 | 236 | ||
237 | /* called by access_process_vm when get_user_pages() fails, typically | 237 | /* called by access_process_vm when get_user_pages() fails, typically |
238 | * for use by special VMAs that can switch between memory and hardware | 238 | * for use by special VMAs that can switch between memory and hardware |
239 | */ | 239 | */ |
240 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | 240 | int (*access)(struct vm_area_struct *vma, unsigned long addr, |
241 | void *buf, int len, int write); | 241 | void *buf, int len, int write); |
242 | #ifdef CONFIG_NUMA | 242 | #ifdef CONFIG_NUMA |
243 | /* | 243 | /* |
244 | * set_policy() op must add a reference to any non-NULL @new mempolicy | 244 | * set_policy() op must add a reference to any non-NULL @new mempolicy |
245 | * to hold the policy upon return. Caller should pass NULL @new to | 245 | * to hold the policy upon return. Caller should pass NULL @new to |
246 | * remove a policy and fall back to surrounding context--i.e. do not | 246 | * remove a policy and fall back to surrounding context--i.e. do not |
247 | * install a MPOL_DEFAULT policy, nor the task or system default | 247 | * install a MPOL_DEFAULT policy, nor the task or system default |
248 | * mempolicy. | 248 | * mempolicy. |
249 | */ | 249 | */ |
250 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); | 250 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
251 | 251 | ||
252 | /* | 252 | /* |
253 | * get_policy() op must add reference [mpol_get()] to any policy at | 253 | * get_policy() op must add reference [mpol_get()] to any policy at |
254 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | 254 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure |
255 | * in mm/mempolicy.c will do this automatically. | 255 | * in mm/mempolicy.c will do this automatically. |
256 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | 256 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not |
257 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. | 257 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. |
258 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op | 258 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op |
259 | * must return NULL--i.e., do not "fallback" to task or system default | 259 | * must return NULL--i.e., do not "fallback" to task or system default |
260 | * policy. | 260 | * policy. |
261 | */ | 261 | */ |
262 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, | 262 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
263 | unsigned long addr); | 263 | unsigned long addr); |
264 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, | 264 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, |
265 | const nodemask_t *to, unsigned long flags); | 265 | const nodemask_t *to, unsigned long flags); |
266 | #endif | 266 | #endif |
267 | /* called by sys_remap_file_pages() to populate non-linear mapping */ | 267 | /* called by sys_remap_file_pages() to populate non-linear mapping */ |
268 | int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr, | 268 | int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr, |
269 | unsigned long size, pgoff_t pgoff); | 269 | unsigned long size, pgoff_t pgoff); |
270 | }; | 270 | }; |
271 | 271 | ||
272 | struct mmu_gather; | 272 | struct mmu_gather; |
273 | struct inode; | 273 | struct inode; |
274 | 274 | ||
275 | #define page_private(page) ((page)->private) | 275 | #define page_private(page) ((page)->private) |
276 | #define set_page_private(page, v) ((page)->private = (v)) | 276 | #define set_page_private(page, v) ((page)->private = (v)) |
277 | 277 | ||
278 | /* It's valid only if the page is free path or free_list */ | 278 | /* It's valid only if the page is free path or free_list */ |
279 | static inline void set_freepage_migratetype(struct page *page, int migratetype) | 279 | static inline void set_freepage_migratetype(struct page *page, int migratetype) |
280 | { | 280 | { |
281 | page->index = migratetype; | 281 | page->index = migratetype; |
282 | } | 282 | } |
283 | 283 | ||
284 | /* It's valid only if the page is free path or free_list */ | 284 | /* It's valid only if the page is free path or free_list */ |
285 | static inline int get_freepage_migratetype(struct page *page) | 285 | static inline int get_freepage_migratetype(struct page *page) |
286 | { | 286 | { |
287 | return page->index; | 287 | return page->index; |
288 | } | 288 | } |
289 | 289 | ||
290 | /* | 290 | /* |
291 | * FIXME: take this include out, include page-flags.h in | 291 | * FIXME: take this include out, include page-flags.h in |
292 | * files which need it (119 of them) | 292 | * files which need it (119 of them) |
293 | */ | 293 | */ |
294 | #include <linux/page-flags.h> | 294 | #include <linux/page-flags.h> |
295 | #include <linux/huge_mm.h> | 295 | #include <linux/huge_mm.h> |
296 | 296 | ||
297 | /* | 297 | /* |
298 | * Methods to modify the page usage count. | 298 | * Methods to modify the page usage count. |
299 | * | 299 | * |
300 | * What counts for a page usage: | 300 | * What counts for a page usage: |
301 | * - cache mapping (page->mapping) | 301 | * - cache mapping (page->mapping) |
302 | * - private data (page->private) | 302 | * - private data (page->private) |
303 | * - page mapped in a task's page tables, each mapping | 303 | * - page mapped in a task's page tables, each mapping |
304 | * is counted separately | 304 | * is counted separately |
305 | * | 305 | * |
306 | * Also, many kernel routines increase the page count before a critical | 306 | * Also, many kernel routines increase the page count before a critical |
307 | * routine so they can be sure the page doesn't go away from under them. | 307 | * routine so they can be sure the page doesn't go away from under them. |
308 | */ | 308 | */ |
309 | 309 | ||
310 | /* | 310 | /* |
311 | * Drop a ref, return true if the refcount fell to zero (the page has no users) | 311 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
312 | */ | 312 | */ |
313 | static inline int put_page_testzero(struct page *page) | 313 | static inline int put_page_testzero(struct page *page) |
314 | { | 314 | { |
315 | VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page); | 315 | VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page); |
316 | return atomic_dec_and_test(&page->_count); | 316 | return atomic_dec_and_test(&page->_count); |
317 | } | 317 | } |
318 | 318 | ||
319 | /* | 319 | /* |
320 | * Try to grab a ref unless the page has a refcount of zero, return false if | 320 | * Try to grab a ref unless the page has a refcount of zero, return false if |
321 | * that is the case. | 321 | * that is the case. |
322 | * This can be called when MMU is off so it must not access | 322 | * This can be called when MMU is off so it must not access |
323 | * any of the virtual mappings. | 323 | * any of the virtual mappings. |
324 | */ | 324 | */ |
325 | static inline int get_page_unless_zero(struct page *page) | 325 | static inline int get_page_unless_zero(struct page *page) |
326 | { | 326 | { |
327 | return atomic_inc_not_zero(&page->_count); | 327 | return atomic_inc_not_zero(&page->_count); |
328 | } | 328 | } |
329 | 329 | ||
330 | /* | 330 | /* |
331 | * Try to drop a ref unless the page has a refcount of one, return false if | 331 | * Try to drop a ref unless the page has a refcount of one, return false if |
332 | * that is the case. | 332 | * that is the case. |
333 | * This is to make sure that the refcount won't become zero after this drop. | 333 | * This is to make sure that the refcount won't become zero after this drop. |
334 | * This can be called when MMU is off so it must not access | 334 | * This can be called when MMU is off so it must not access |
335 | * any of the virtual mappings. | 335 | * any of the virtual mappings. |
336 | */ | 336 | */ |
337 | static inline int put_page_unless_one(struct page *page) | 337 | static inline int put_page_unless_one(struct page *page) |
338 | { | 338 | { |
339 | return atomic_add_unless(&page->_count, -1, 1); | 339 | return atomic_add_unless(&page->_count, -1, 1); |
340 | } | 340 | } |
341 | 341 | ||
342 | extern int page_is_ram(unsigned long pfn); | 342 | extern int page_is_ram(unsigned long pfn); |
343 | 343 | ||
344 | /* Support for virtually mapped pages */ | 344 | /* Support for virtually mapped pages */ |
345 | struct page *vmalloc_to_page(const void *addr); | 345 | struct page *vmalloc_to_page(const void *addr); |
346 | unsigned long vmalloc_to_pfn(const void *addr); | 346 | unsigned long vmalloc_to_pfn(const void *addr); |
347 | 347 | ||
348 | /* | 348 | /* |
349 | * Determine if an address is within the vmalloc range | 349 | * Determine if an address is within the vmalloc range |
350 | * | 350 | * |
351 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | 351 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there |
352 | * is no special casing required. | 352 | * is no special casing required. |
353 | */ | 353 | */ |
354 | static inline int is_vmalloc_addr(const void *x) | 354 | static inline int is_vmalloc_addr(const void *x) |
355 | { | 355 | { |
356 | #ifdef CONFIG_MMU | 356 | #ifdef CONFIG_MMU |
357 | unsigned long addr = (unsigned long)x; | 357 | unsigned long addr = (unsigned long)x; |
358 | 358 | ||
359 | return addr >= VMALLOC_START && addr < VMALLOC_END; | 359 | return addr >= VMALLOC_START && addr < VMALLOC_END; |
360 | #else | 360 | #else |
361 | return 0; | 361 | return 0; |
362 | #endif | 362 | #endif |
363 | } | 363 | } |
364 | #ifdef CONFIG_MMU | 364 | #ifdef CONFIG_MMU |
365 | extern int is_vmalloc_or_module_addr(const void *x); | 365 | extern int is_vmalloc_or_module_addr(const void *x); |
366 | #else | 366 | #else |
367 | static inline int is_vmalloc_or_module_addr(const void *x) | 367 | static inline int is_vmalloc_or_module_addr(const void *x) |
368 | { | 368 | { |
369 | return 0; | 369 | return 0; |
370 | } | 370 | } |
371 | #endif | 371 | #endif |
372 | 372 | ||
373 | static inline void compound_lock(struct page *page) | 373 | static inline void compound_lock(struct page *page) |
374 | { | 374 | { |
375 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | 375 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
376 | VM_BUG_ON_PAGE(PageSlab(page), page); | 376 | VM_BUG_ON_PAGE(PageSlab(page), page); |
377 | bit_spin_lock(PG_compound_lock, &page->flags); | 377 | bit_spin_lock(PG_compound_lock, &page->flags); |
378 | #endif | 378 | #endif |
379 | } | 379 | } |
380 | 380 | ||
381 | static inline void compound_unlock(struct page *page) | 381 | static inline void compound_unlock(struct page *page) |
382 | { | 382 | { |
383 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | 383 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
384 | VM_BUG_ON_PAGE(PageSlab(page), page); | 384 | VM_BUG_ON_PAGE(PageSlab(page), page); |
385 | bit_spin_unlock(PG_compound_lock, &page->flags); | 385 | bit_spin_unlock(PG_compound_lock, &page->flags); |
386 | #endif | 386 | #endif |
387 | } | 387 | } |
388 | 388 | ||
389 | static inline unsigned long compound_lock_irqsave(struct page *page) | 389 | static inline unsigned long compound_lock_irqsave(struct page *page) |
390 | { | 390 | { |
391 | unsigned long uninitialized_var(flags); | 391 | unsigned long uninitialized_var(flags); |
392 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | 392 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
393 | local_irq_save(flags); | 393 | local_irq_save(flags); |
394 | compound_lock(page); | 394 | compound_lock(page); |
395 | #endif | 395 | #endif |
396 | return flags; | 396 | return flags; |
397 | } | 397 | } |
398 | 398 | ||
399 | static inline void compound_unlock_irqrestore(struct page *page, | 399 | static inline void compound_unlock_irqrestore(struct page *page, |
400 | unsigned long flags) | 400 | unsigned long flags) |
401 | { | 401 | { |
402 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | 402 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
403 | compound_unlock(page); | 403 | compound_unlock(page); |
404 | local_irq_restore(flags); | 404 | local_irq_restore(flags); |
405 | #endif | 405 | #endif |
406 | } | 406 | } |
407 | 407 | ||
408 | static inline struct page *compound_head(struct page *page) | 408 | static inline struct page *compound_head(struct page *page) |
409 | { | 409 | { |
410 | if (unlikely(PageTail(page))) { | 410 | if (unlikely(PageTail(page))) { |
411 | struct page *head = page->first_page; | 411 | struct page *head = page->first_page; |
412 | 412 | ||
413 | /* | 413 | /* |
414 | * page->first_page may be a dangling pointer to an old | 414 | * page->first_page may be a dangling pointer to an old |
415 | * compound page, so recheck that it is still a tail | 415 | * compound page, so recheck that it is still a tail |
416 | * page before returning. | 416 | * page before returning. |
417 | */ | 417 | */ |
418 | smp_rmb(); | 418 | smp_rmb(); |
419 | if (likely(PageTail(page))) | 419 | if (likely(PageTail(page))) |
420 | return head; | 420 | return head; |
421 | } | 421 | } |
422 | return page; | 422 | return page; |
423 | } | 423 | } |
424 | 424 | ||
425 | /* | 425 | /* |
426 | * The atomic page->_mapcount, starts from -1: so that transitions | 426 | * The atomic page->_mapcount, starts from -1: so that transitions |
427 | * both from it and to it can be tracked, using atomic_inc_and_test | 427 | * both from it and to it can be tracked, using atomic_inc_and_test |
428 | * and atomic_add_negative(-1). | 428 | * and atomic_add_negative(-1). |
429 | */ | 429 | */ |
430 | static inline void page_mapcount_reset(struct page *page) | 430 | static inline void page_mapcount_reset(struct page *page) |
431 | { | 431 | { |
432 | atomic_set(&(page)->_mapcount, -1); | 432 | atomic_set(&(page)->_mapcount, -1); |
433 | } | 433 | } |
434 | 434 | ||
435 | static inline int page_mapcount(struct page *page) | 435 | static inline int page_mapcount(struct page *page) |
436 | { | 436 | { |
437 | return atomic_read(&(page)->_mapcount) + 1; | 437 | return atomic_read(&(page)->_mapcount) + 1; |
438 | } | 438 | } |
439 | 439 | ||
440 | static inline int page_count(struct page *page) | 440 | static inline int page_count(struct page *page) |
441 | { | 441 | { |
442 | return atomic_read(&compound_head(page)->_count); | 442 | return atomic_read(&compound_head(page)->_count); |
443 | } | 443 | } |
444 | 444 | ||
445 | #ifdef CONFIG_HUGETLB_PAGE | 445 | #ifdef CONFIG_HUGETLB_PAGE |
446 | extern int PageHeadHuge(struct page *page_head); | 446 | extern int PageHeadHuge(struct page *page_head); |
447 | #else /* CONFIG_HUGETLB_PAGE */ | 447 | #else /* CONFIG_HUGETLB_PAGE */ |
448 | static inline int PageHeadHuge(struct page *page_head) | 448 | static inline int PageHeadHuge(struct page *page_head) |
449 | { | 449 | { |
450 | return 0; | 450 | return 0; |
451 | } | 451 | } |
452 | #endif /* CONFIG_HUGETLB_PAGE */ | 452 | #endif /* CONFIG_HUGETLB_PAGE */ |
453 | 453 | ||
454 | static inline bool __compound_tail_refcounted(struct page *page) | 454 | static inline bool __compound_tail_refcounted(struct page *page) |
455 | { | 455 | { |
456 | return !PageSlab(page) && !PageHeadHuge(page); | 456 | return !PageSlab(page) && !PageHeadHuge(page); |
457 | } | 457 | } |
458 | 458 | ||
459 | /* | 459 | /* |
460 | * This takes a head page as parameter and tells if the | 460 | * This takes a head page as parameter and tells if the |
461 | * tail page reference counting can be skipped. | 461 | * tail page reference counting can be skipped. |
462 | * | 462 | * |
463 | * For this to be safe, PageSlab and PageHeadHuge must remain true on | 463 | * For this to be safe, PageSlab and PageHeadHuge must remain true on |
464 | * any given page where they return true here, until all tail pins | 464 | * any given page where they return true here, until all tail pins |
465 | * have been released. | 465 | * have been released. |
466 | */ | 466 | */ |
467 | static inline bool compound_tail_refcounted(struct page *page) | 467 | static inline bool compound_tail_refcounted(struct page *page) |
468 | { | 468 | { |
469 | VM_BUG_ON_PAGE(!PageHead(page), page); | 469 | VM_BUG_ON_PAGE(!PageHead(page), page); |
470 | return __compound_tail_refcounted(page); | 470 | return __compound_tail_refcounted(page); |
471 | } | 471 | } |
472 | 472 | ||
473 | static inline void get_huge_page_tail(struct page *page) | 473 | static inline void get_huge_page_tail(struct page *page) |
474 | { | 474 | { |
475 | /* | 475 | /* |
476 | * __split_huge_page_refcount() cannot run from under us. | 476 | * __split_huge_page_refcount() cannot run from under us. |
477 | */ | 477 | */ |
478 | VM_BUG_ON_PAGE(!PageTail(page), page); | 478 | VM_BUG_ON_PAGE(!PageTail(page), page); |
479 | VM_BUG_ON_PAGE(page_mapcount(page) < 0, page); | 479 | VM_BUG_ON_PAGE(page_mapcount(page) < 0, page); |
480 | VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page); | 480 | VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page); |
481 | if (compound_tail_refcounted(page->first_page)) | 481 | if (compound_tail_refcounted(page->first_page)) |
482 | atomic_inc(&page->_mapcount); | 482 | atomic_inc(&page->_mapcount); |
483 | } | 483 | } |
484 | 484 | ||
485 | extern bool __get_page_tail(struct page *page); | 485 | extern bool __get_page_tail(struct page *page); |
486 | 486 | ||
487 | static inline void get_page(struct page *page) | 487 | static inline void get_page(struct page *page) |
488 | { | 488 | { |
489 | if (unlikely(PageTail(page))) | 489 | if (unlikely(PageTail(page))) |
490 | if (likely(__get_page_tail(page))) | 490 | if (likely(__get_page_tail(page))) |
491 | return; | 491 | return; |
492 | /* | 492 | /* |
493 | * Getting a normal page or the head of a compound page | 493 | * Getting a normal page or the head of a compound page |
494 | * requires to already have an elevated page->_count. | 494 | * requires to already have an elevated page->_count. |
495 | */ | 495 | */ |
496 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); | 496 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); |
497 | atomic_inc(&page->_count); | 497 | atomic_inc(&page->_count); |
498 | } | 498 | } |
499 | 499 | ||
500 | static inline struct page *virt_to_head_page(const void *x) | 500 | static inline struct page *virt_to_head_page(const void *x) |
501 | { | 501 | { |
502 | struct page *page = virt_to_page(x); | 502 | struct page *page = virt_to_page(x); |
503 | return compound_head(page); | 503 | return compound_head(page); |
504 | } | 504 | } |
505 | 505 | ||
506 | /* | 506 | /* |
507 | * Setup the page count before being freed into the page allocator for | 507 | * Setup the page count before being freed into the page allocator for |
508 | * the first time (boot or memory hotplug) | 508 | * the first time (boot or memory hotplug) |
509 | */ | 509 | */ |
510 | static inline void init_page_count(struct page *page) | 510 | static inline void init_page_count(struct page *page) |
511 | { | 511 | { |
512 | atomic_set(&page->_count, 1); | 512 | atomic_set(&page->_count, 1); |
513 | } | 513 | } |
514 | 514 | ||
515 | /* | 515 | /* |
516 | * PageBuddy() indicate that the page is free and in the buddy system | 516 | * PageBuddy() indicate that the page is free and in the buddy system |
517 | * (see mm/page_alloc.c). | 517 | * (see mm/page_alloc.c). |
518 | * | 518 | * |
519 | * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to | 519 | * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to |
520 | * -2 so that an underflow of the page_mapcount() won't be mistaken | 520 | * -2 so that an underflow of the page_mapcount() won't be mistaken |
521 | * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very | 521 | * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very |
522 | * efficiently by most CPU architectures. | 522 | * efficiently by most CPU architectures. |
523 | */ | 523 | */ |
524 | #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) | 524 | #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) |
525 | 525 | ||
526 | static inline int PageBuddy(struct page *page) | 526 | static inline int PageBuddy(struct page *page) |
527 | { | 527 | { |
528 | return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; | 528 | return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; |
529 | } | 529 | } |
530 | 530 | ||
531 | static inline void __SetPageBuddy(struct page *page) | 531 | static inline void __SetPageBuddy(struct page *page) |
532 | { | 532 | { |
533 | VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page); | 533 | VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page); |
534 | atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); | 534 | atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); |
535 | } | 535 | } |
536 | 536 | ||
537 | static inline void __ClearPageBuddy(struct page *page) | 537 | static inline void __ClearPageBuddy(struct page *page) |
538 | { | 538 | { |
539 | VM_BUG_ON_PAGE(!PageBuddy(page), page); | 539 | VM_BUG_ON_PAGE(!PageBuddy(page), page); |
540 | atomic_set(&page->_mapcount, -1); | 540 | atomic_set(&page->_mapcount, -1); |
541 | } | 541 | } |
542 | 542 | ||
543 | void put_page(struct page *page); | 543 | void put_page(struct page *page); |
544 | void put_pages_list(struct list_head *pages); | 544 | void put_pages_list(struct list_head *pages); |
545 | 545 | ||
546 | void split_page(struct page *page, unsigned int order); | 546 | void split_page(struct page *page, unsigned int order); |
547 | int split_free_page(struct page *page); | 547 | int split_free_page(struct page *page); |
548 | 548 | ||
549 | /* | 549 | /* |
550 | * Compound pages have a destructor function. Provide a | 550 | * Compound pages have a destructor function. Provide a |
551 | * prototype for that function and accessor functions. | 551 | * prototype for that function and accessor functions. |
552 | * These are _only_ valid on the head of a PG_compound page. | 552 | * These are _only_ valid on the head of a PG_compound page. |
553 | */ | 553 | */ |
554 | typedef void compound_page_dtor(struct page *); | 554 | typedef void compound_page_dtor(struct page *); |
555 | 555 | ||
556 | static inline void set_compound_page_dtor(struct page *page, | 556 | static inline void set_compound_page_dtor(struct page *page, |
557 | compound_page_dtor *dtor) | 557 | compound_page_dtor *dtor) |
558 | { | 558 | { |
559 | page[1].lru.next = (void *)dtor; | 559 | page[1].lru.next = (void *)dtor; |
560 | } | 560 | } |
561 | 561 | ||
562 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | 562 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) |
563 | { | 563 | { |
564 | return (compound_page_dtor *)page[1].lru.next; | 564 | return (compound_page_dtor *)page[1].lru.next; |
565 | } | 565 | } |
566 | 566 | ||
567 | static inline int compound_order(struct page *page) | 567 | static inline int compound_order(struct page *page) |
568 | { | 568 | { |
569 | if (!PageHead(page)) | 569 | if (!PageHead(page)) |
570 | return 0; | 570 | return 0; |
571 | return (unsigned long)page[1].lru.prev; | 571 | return (unsigned long)page[1].lru.prev; |
572 | } | 572 | } |
573 | 573 | ||
574 | static inline void set_compound_order(struct page *page, unsigned long order) | 574 | static inline void set_compound_order(struct page *page, unsigned long order) |
575 | { | 575 | { |
576 | page[1].lru.prev = (void *)order; | 576 | page[1].lru.prev = (void *)order; |
577 | } | 577 | } |
578 | 578 | ||
579 | #ifdef CONFIG_MMU | 579 | #ifdef CONFIG_MMU |
580 | /* | 580 | /* |
581 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | 581 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when |
582 | * servicing faults for write access. In the normal case, do always want | 582 | * servicing faults for write access. In the normal case, do always want |
583 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | 583 | * pte_mkwrite. But get_user_pages can cause write faults for mappings |
584 | * that do not have writing enabled, when used by access_process_vm. | 584 | * that do not have writing enabled, when used by access_process_vm. |
585 | */ | 585 | */ |
586 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | 586 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) |
587 | { | 587 | { |
588 | if (likely(vma->vm_flags & VM_WRITE)) | 588 | if (likely(vma->vm_flags & VM_WRITE)) |
589 | pte = pte_mkwrite(pte); | 589 | pte = pte_mkwrite(pte); |
590 | return pte; | 590 | return pte; |
591 | } | 591 | } |
592 | 592 | ||
593 | void do_set_pte(struct vm_area_struct *vma, unsigned long address, | 593 | void do_set_pte(struct vm_area_struct *vma, unsigned long address, |
594 | struct page *page, pte_t *pte, bool write, bool anon); | 594 | struct page *page, pte_t *pte, bool write, bool anon); |
595 | #endif | 595 | #endif |
596 | 596 | ||
597 | /* | 597 | /* |
598 | * Multiple processes may "see" the same page. E.g. for untouched | 598 | * Multiple processes may "see" the same page. E.g. for untouched |
599 | * mappings of /dev/null, all processes see the same page full of | 599 | * mappings of /dev/null, all processes see the same page full of |
600 | * zeroes, and text pages of executables and shared libraries have | 600 | * zeroes, and text pages of executables and shared libraries have |
601 | * only one copy in memory, at most, normally. | 601 | * only one copy in memory, at most, normally. |
602 | * | 602 | * |
603 | * For the non-reserved pages, page_count(page) denotes a reference count. | 603 | * For the non-reserved pages, page_count(page) denotes a reference count. |
604 | * page_count() == 0 means the page is free. page->lru is then used for | 604 | * page_count() == 0 means the page is free. page->lru is then used for |
605 | * freelist management in the buddy allocator. | 605 | * freelist management in the buddy allocator. |
606 | * page_count() > 0 means the page has been allocated. | 606 | * page_count() > 0 means the page has been allocated. |
607 | * | 607 | * |
608 | * Pages are allocated by the slab allocator in order to provide memory | 608 | * Pages are allocated by the slab allocator in order to provide memory |
609 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | 609 | * to kmalloc and kmem_cache_alloc. In this case, the management of the |
610 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | 610 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c |
611 | * unless a particular usage is carefully commented. (the responsibility of | 611 | * unless a particular usage is carefully commented. (the responsibility of |
612 | * freeing the kmalloc memory is the caller's, of course). | 612 | * freeing the kmalloc memory is the caller's, of course). |
613 | * | 613 | * |
614 | * A page may be used by anyone else who does a __get_free_page(). | 614 | * A page may be used by anyone else who does a __get_free_page(). |
615 | * In this case, page_count still tracks the references, and should only | 615 | * In this case, page_count still tracks the references, and should only |
616 | * be used through the normal accessor functions. The top bits of page->flags | 616 | * be used through the normal accessor functions. The top bits of page->flags |
617 | * and page->virtual store page management information, but all other fields | 617 | * and page->virtual store page management information, but all other fields |
618 | * are unused and could be used privately, carefully. The management of this | 618 | * are unused and could be used privately, carefully. The management of this |
619 | * page is the responsibility of the one who allocated it, and those who have | 619 | * page is the responsibility of the one who allocated it, and those who have |
620 | * subsequently been given references to it. | 620 | * subsequently been given references to it. |
621 | * | 621 | * |
622 | * The other pages (we may call them "pagecache pages") are completely | 622 | * The other pages (we may call them "pagecache pages") are completely |
623 | * managed by the Linux memory manager: I/O, buffers, swapping etc. | 623 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
624 | * The following discussion applies only to them. | 624 | * The following discussion applies only to them. |
625 | * | 625 | * |
626 | * A pagecache page contains an opaque `private' member, which belongs to the | 626 | * A pagecache page contains an opaque `private' member, which belongs to the |
627 | * page's address_space. Usually, this is the address of a circular list of | 627 | * page's address_space. Usually, this is the address of a circular list of |
628 | * the page's disk buffers. PG_private must be set to tell the VM to call | 628 | * the page's disk buffers. PG_private must be set to tell the VM to call |
629 | * into the filesystem to release these pages. | 629 | * into the filesystem to release these pages. |
630 | * | 630 | * |
631 | * A page may belong to an inode's memory mapping. In this case, page->mapping | 631 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
632 | * is the pointer to the inode, and page->index is the file offset of the page, | 632 | * is the pointer to the inode, and page->index is the file offset of the page, |
633 | * in units of PAGE_CACHE_SIZE. | 633 | * in units of PAGE_CACHE_SIZE. |
634 | * | 634 | * |
635 | * If pagecache pages are not associated with an inode, they are said to be | 635 | * If pagecache pages are not associated with an inode, they are said to be |
636 | * anonymous pages. These may become associated with the swapcache, and in that | 636 | * anonymous pages. These may become associated with the swapcache, and in that |
637 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | 637 | * case PG_swapcache is set, and page->private is an offset into the swapcache. |
638 | * | 638 | * |
639 | * In either case (swapcache or inode backed), the pagecache itself holds one | 639 | * In either case (swapcache or inode backed), the pagecache itself holds one |
640 | * reference to the page. Setting PG_private should also increment the | 640 | * reference to the page. Setting PG_private should also increment the |
641 | * refcount. The each user mapping also has a reference to the page. | 641 | * refcount. The each user mapping also has a reference to the page. |
642 | * | 642 | * |
643 | * The pagecache pages are stored in a per-mapping radix tree, which is | 643 | * The pagecache pages are stored in a per-mapping radix tree, which is |
644 | * rooted at mapping->page_tree, and indexed by offset. | 644 | * rooted at mapping->page_tree, and indexed by offset. |
645 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space | 645 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
646 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | 646 | * lists, we instead now tag pages as dirty/writeback in the radix tree. |
647 | * | 647 | * |
648 | * All pagecache pages may be subject to I/O: | 648 | * All pagecache pages may be subject to I/O: |
649 | * - inode pages may need to be read from disk, | 649 | * - inode pages may need to be read from disk, |
650 | * - inode pages which have been modified and are MAP_SHARED may need | 650 | * - inode pages which have been modified and are MAP_SHARED may need |
651 | * to be written back to the inode on disk, | 651 | * to be written back to the inode on disk, |
652 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | 652 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been |
653 | * modified may need to be swapped out to swap space and (later) to be read | 653 | * modified may need to be swapped out to swap space and (later) to be read |
654 | * back into memory. | 654 | * back into memory. |
655 | */ | 655 | */ |
656 | 656 | ||
657 | /* | 657 | /* |
658 | * The zone field is never updated after free_area_init_core() | 658 | * The zone field is never updated after free_area_init_core() |
659 | * sets it, so none of the operations on it need to be atomic. | 659 | * sets it, so none of the operations on it need to be atomic. |
660 | */ | 660 | */ |
661 | 661 | ||
662 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ | 662 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ |
663 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) | 663 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
664 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) | 664 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
665 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | 665 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) |
666 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) | 666 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) |
667 | 667 | ||
668 | /* | 668 | /* |
669 | * Define the bit shifts to access each section. For non-existent | 669 | * Define the bit shifts to access each section. For non-existent |
670 | * sections we define the shift as 0; that plus a 0 mask ensures | 670 | * sections we define the shift as 0; that plus a 0 mask ensures |
671 | * the compiler will optimise away reference to them. | 671 | * the compiler will optimise away reference to them. |
672 | */ | 672 | */ |
673 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) | 673 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
674 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | 674 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) |
675 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | 675 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) |
676 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) | 676 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) |
677 | 677 | ||
678 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ | 678 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ |
679 | #ifdef NODE_NOT_IN_PAGE_FLAGS | 679 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
680 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) | 680 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
681 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ | 681 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
682 | SECTIONS_PGOFF : ZONES_PGOFF) | 682 | SECTIONS_PGOFF : ZONES_PGOFF) |
683 | #else | 683 | #else |
684 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) | 684 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
685 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ | 685 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
686 | NODES_PGOFF : ZONES_PGOFF) | 686 | NODES_PGOFF : ZONES_PGOFF) |
687 | #endif | 687 | #endif |
688 | 688 | ||
689 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) | 689 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
690 | 690 | ||
691 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | 691 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
692 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | 692 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
693 | #endif | 693 | #endif |
694 | 694 | ||
695 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) | 695 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
696 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | 696 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) |
697 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | 697 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) |
698 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_WIDTH) - 1) | 698 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_WIDTH) - 1) |
699 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) | 699 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
700 | 700 | ||
701 | static inline enum zone_type page_zonenum(const struct page *page) | 701 | static inline enum zone_type page_zonenum(const struct page *page) |
702 | { | 702 | { |
703 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; | 703 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
704 | } | 704 | } |
705 | 705 | ||
706 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) | 706 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
707 | #define SECTION_IN_PAGE_FLAGS | 707 | #define SECTION_IN_PAGE_FLAGS |
708 | #endif | 708 | #endif |
709 | 709 | ||
710 | /* | 710 | /* |
711 | * The identification function is mainly used by the buddy allocator for | 711 | * The identification function is mainly used by the buddy allocator for |
712 | * determining if two pages could be buddies. We are not really identifying | 712 | * determining if two pages could be buddies. We are not really identifying |
713 | * the zone since we could be using the section number id if we do not have | 713 | * the zone since we could be using the section number id if we do not have |
714 | * node id available in page flags. | 714 | * node id available in page flags. |
715 | * We only guarantee that it will return the same value for two combinable | 715 | * We only guarantee that it will return the same value for two combinable |
716 | * pages in a zone. | 716 | * pages in a zone. |
717 | */ | 717 | */ |
718 | static inline int page_zone_id(struct page *page) | 718 | static inline int page_zone_id(struct page *page) |
719 | { | 719 | { |
720 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; | 720 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
721 | } | 721 | } |
722 | 722 | ||
723 | static inline int zone_to_nid(struct zone *zone) | 723 | static inline int zone_to_nid(struct zone *zone) |
724 | { | 724 | { |
725 | #ifdef CONFIG_NUMA | 725 | #ifdef CONFIG_NUMA |
726 | return zone->node; | 726 | return zone->node; |
727 | #else | 727 | #else |
728 | return 0; | 728 | return 0; |
729 | #endif | 729 | #endif |
730 | } | 730 | } |
731 | 731 | ||
732 | #ifdef NODE_NOT_IN_PAGE_FLAGS | 732 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
733 | extern int page_to_nid(const struct page *page); | 733 | extern int page_to_nid(const struct page *page); |
734 | #else | 734 | #else |
735 | static inline int page_to_nid(const struct page *page) | 735 | static inline int page_to_nid(const struct page *page) |
736 | { | 736 | { |
737 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; | 737 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; |
738 | } | 738 | } |
739 | #endif | 739 | #endif |
740 | 740 | ||
741 | #ifdef CONFIG_NUMA_BALANCING | 741 | #ifdef CONFIG_NUMA_BALANCING |
742 | static inline int cpu_pid_to_cpupid(int cpu, int pid) | 742 | static inline int cpu_pid_to_cpupid(int cpu, int pid) |
743 | { | 743 | { |
744 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); | 744 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); |
745 | } | 745 | } |
746 | 746 | ||
747 | static inline int cpupid_to_pid(int cpupid) | 747 | static inline int cpupid_to_pid(int cpupid) |
748 | { | 748 | { |
749 | return cpupid & LAST__PID_MASK; | 749 | return cpupid & LAST__PID_MASK; |
750 | } | 750 | } |
751 | 751 | ||
752 | static inline int cpupid_to_cpu(int cpupid) | 752 | static inline int cpupid_to_cpu(int cpupid) |
753 | { | 753 | { |
754 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; | 754 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; |
755 | } | 755 | } |
756 | 756 | ||
757 | static inline int cpupid_to_nid(int cpupid) | 757 | static inline int cpupid_to_nid(int cpupid) |
758 | { | 758 | { |
759 | return cpu_to_node(cpupid_to_cpu(cpupid)); | 759 | return cpu_to_node(cpupid_to_cpu(cpupid)); |
760 | } | 760 | } |
761 | 761 | ||
762 | static inline bool cpupid_pid_unset(int cpupid) | 762 | static inline bool cpupid_pid_unset(int cpupid) |
763 | { | 763 | { |
764 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); | 764 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); |
765 | } | 765 | } |
766 | 766 | ||
767 | static inline bool cpupid_cpu_unset(int cpupid) | 767 | static inline bool cpupid_cpu_unset(int cpupid) |
768 | { | 768 | { |
769 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); | 769 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); |
770 | } | 770 | } |
771 | 771 | ||
772 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) | 772 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) |
773 | { | 773 | { |
774 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); | 774 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); |
775 | } | 775 | } |
776 | 776 | ||
777 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) | 777 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) |
778 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS | 778 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
779 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | 779 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) |
780 | { | 780 | { |
781 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); | 781 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); |
782 | } | 782 | } |
783 | 783 | ||
784 | static inline int page_cpupid_last(struct page *page) | 784 | static inline int page_cpupid_last(struct page *page) |
785 | { | 785 | { |
786 | return page->_last_cpupid; | 786 | return page->_last_cpupid; |
787 | } | 787 | } |
788 | static inline void page_cpupid_reset_last(struct page *page) | 788 | static inline void page_cpupid_reset_last(struct page *page) |
789 | { | 789 | { |
790 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; | 790 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; |
791 | } | 791 | } |
792 | #else | 792 | #else |
793 | static inline int page_cpupid_last(struct page *page) | 793 | static inline int page_cpupid_last(struct page *page) |
794 | { | 794 | { |
795 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; | 795 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; |
796 | } | 796 | } |
797 | 797 | ||
798 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); | 798 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); |
799 | 799 | ||
800 | static inline void page_cpupid_reset_last(struct page *page) | 800 | static inline void page_cpupid_reset_last(struct page *page) |
801 | { | 801 | { |
802 | int cpupid = (1 << LAST_CPUPID_SHIFT) - 1; | 802 | int cpupid = (1 << LAST_CPUPID_SHIFT) - 1; |
803 | 803 | ||
804 | page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT); | 804 | page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT); |
805 | page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT; | 805 | page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT; |
806 | } | 806 | } |
807 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ | 807 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ |
808 | #else /* !CONFIG_NUMA_BALANCING */ | 808 | #else /* !CONFIG_NUMA_BALANCING */ |
809 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | 809 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) |
810 | { | 810 | { |
811 | return page_to_nid(page); /* XXX */ | 811 | return page_to_nid(page); /* XXX */ |
812 | } | 812 | } |
813 | 813 | ||
814 | static inline int page_cpupid_last(struct page *page) | 814 | static inline int page_cpupid_last(struct page *page) |
815 | { | 815 | { |
816 | return page_to_nid(page); /* XXX */ | 816 | return page_to_nid(page); /* XXX */ |
817 | } | 817 | } |
818 | 818 | ||
819 | static inline int cpupid_to_nid(int cpupid) | 819 | static inline int cpupid_to_nid(int cpupid) |
820 | { | 820 | { |
821 | return -1; | 821 | return -1; |
822 | } | 822 | } |
823 | 823 | ||
824 | static inline int cpupid_to_pid(int cpupid) | 824 | static inline int cpupid_to_pid(int cpupid) |
825 | { | 825 | { |
826 | return -1; | 826 | return -1; |
827 | } | 827 | } |
828 | 828 | ||
829 | static inline int cpupid_to_cpu(int cpupid) | 829 | static inline int cpupid_to_cpu(int cpupid) |
830 | { | 830 | { |
831 | return -1; | 831 | return -1; |
832 | } | 832 | } |
833 | 833 | ||
834 | static inline int cpu_pid_to_cpupid(int nid, int pid) | 834 | static inline int cpu_pid_to_cpupid(int nid, int pid) |
835 | { | 835 | { |
836 | return -1; | 836 | return -1; |
837 | } | 837 | } |
838 | 838 | ||
839 | static inline bool cpupid_pid_unset(int cpupid) | 839 | static inline bool cpupid_pid_unset(int cpupid) |
840 | { | 840 | { |
841 | return 1; | 841 | return 1; |
842 | } | 842 | } |
843 | 843 | ||
844 | static inline void page_cpupid_reset_last(struct page *page) | 844 | static inline void page_cpupid_reset_last(struct page *page) |
845 | { | 845 | { |
846 | } | 846 | } |
847 | 847 | ||
848 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) | 848 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) |
849 | { | 849 | { |
850 | return false; | 850 | return false; |
851 | } | 851 | } |
852 | #endif /* CONFIG_NUMA_BALANCING */ | 852 | #endif /* CONFIG_NUMA_BALANCING */ |
853 | 853 | ||
854 | static inline struct zone *page_zone(const struct page *page) | 854 | static inline struct zone *page_zone(const struct page *page) |
855 | { | 855 | { |
856 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | 856 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; |
857 | } | 857 | } |
858 | 858 | ||
859 | #ifdef SECTION_IN_PAGE_FLAGS | 859 | #ifdef SECTION_IN_PAGE_FLAGS |
860 | static inline void set_page_section(struct page *page, unsigned long section) | 860 | static inline void set_page_section(struct page *page, unsigned long section) |
861 | { | 861 | { |
862 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | 862 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); |
863 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | 863 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; |
864 | } | 864 | } |
865 | 865 | ||
866 | static inline unsigned long page_to_section(const struct page *page) | 866 | static inline unsigned long page_to_section(const struct page *page) |
867 | { | 867 | { |
868 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | 868 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; |
869 | } | 869 | } |
870 | #endif | 870 | #endif |
871 | 871 | ||
872 | static inline void set_page_zone(struct page *page, enum zone_type zone) | 872 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
873 | { | 873 | { |
874 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | 874 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); |
875 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | 875 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; |
876 | } | 876 | } |
877 | 877 | ||
878 | static inline void set_page_node(struct page *page, unsigned long node) | 878 | static inline void set_page_node(struct page *page, unsigned long node) |
879 | { | 879 | { |
880 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | 880 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); |
881 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | 881 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; |
882 | } | 882 | } |
883 | 883 | ||
884 | static inline void set_page_links(struct page *page, enum zone_type zone, | 884 | static inline void set_page_links(struct page *page, enum zone_type zone, |
885 | unsigned long node, unsigned long pfn) | 885 | unsigned long node, unsigned long pfn) |
886 | { | 886 | { |
887 | set_page_zone(page, zone); | 887 | set_page_zone(page, zone); |
888 | set_page_node(page, node); | 888 | set_page_node(page, node); |
889 | #ifdef SECTION_IN_PAGE_FLAGS | 889 | #ifdef SECTION_IN_PAGE_FLAGS |
890 | set_page_section(page, pfn_to_section_nr(pfn)); | 890 | set_page_section(page, pfn_to_section_nr(pfn)); |
891 | #endif | 891 | #endif |
892 | } | 892 | } |
893 | 893 | ||
894 | /* | 894 | /* |
895 | * Some inline functions in vmstat.h depend on page_zone() | 895 | * Some inline functions in vmstat.h depend on page_zone() |
896 | */ | 896 | */ |
897 | #include <linux/vmstat.h> | 897 | #include <linux/vmstat.h> |
898 | 898 | ||
899 | static __always_inline void *lowmem_page_address(const struct page *page) | 899 | static __always_inline void *lowmem_page_address(const struct page *page) |
900 | { | 900 | { |
901 | return __va(PFN_PHYS(page_to_pfn(page))); | 901 | return __va(PFN_PHYS(page_to_pfn(page))); |
902 | } | 902 | } |
903 | 903 | ||
904 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | 904 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) |
905 | #define HASHED_PAGE_VIRTUAL | 905 | #define HASHED_PAGE_VIRTUAL |
906 | #endif | 906 | #endif |
907 | 907 | ||
908 | #if defined(WANT_PAGE_VIRTUAL) | 908 | #if defined(WANT_PAGE_VIRTUAL) |
909 | static inline void *page_address(const struct page *page) | 909 | static inline void *page_address(const struct page *page) |
910 | { | 910 | { |
911 | return page->virtual; | 911 | return page->virtual; |
912 | } | 912 | } |
913 | static inline void set_page_address(struct page *page, void *address) | 913 | static inline void set_page_address(struct page *page, void *address) |
914 | { | 914 | { |
915 | page->virtual = address; | 915 | page->virtual = address; |
916 | } | 916 | } |
917 | #define page_address_init() do { } while(0) | 917 | #define page_address_init() do { } while(0) |
918 | #endif | 918 | #endif |
919 | 919 | ||
920 | #if defined(HASHED_PAGE_VIRTUAL) | 920 | #if defined(HASHED_PAGE_VIRTUAL) |
921 | void *page_address(const struct page *page); | 921 | void *page_address(const struct page *page); |
922 | void set_page_address(struct page *page, void *virtual); | 922 | void set_page_address(struct page *page, void *virtual); |
923 | void page_address_init(void); | 923 | void page_address_init(void); |
924 | #endif | 924 | #endif |
925 | 925 | ||
926 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | 926 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) |
927 | #define page_address(page) lowmem_page_address(page) | 927 | #define page_address(page) lowmem_page_address(page) |
928 | #define set_page_address(page, address) do { } while(0) | 928 | #define set_page_address(page, address) do { } while(0) |
929 | #define page_address_init() do { } while(0) | 929 | #define page_address_init() do { } while(0) |
930 | #endif | 930 | #endif |
931 | 931 | ||
932 | /* | 932 | /* |
933 | * On an anonymous page mapped into a user virtual memory area, | 933 | * On an anonymous page mapped into a user virtual memory area, |
934 | * page->mapping points to its anon_vma, not to a struct address_space; | 934 | * page->mapping points to its anon_vma, not to a struct address_space; |
935 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. | 935 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. |
936 | * | 936 | * |
937 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, | 937 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, |
938 | * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; | 938 | * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; |
939 | * and then page->mapping points, not to an anon_vma, but to a private | 939 | * and then page->mapping points, not to an anon_vma, but to a private |
940 | * structure which KSM associates with that merged page. See ksm.h. | 940 | * structure which KSM associates with that merged page. See ksm.h. |
941 | * | 941 | * |
942 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. | 942 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. |
943 | * | 943 | * |
944 | * Please note that, confusingly, "page_mapping" refers to the inode | 944 | * Please note that, confusingly, "page_mapping" refers to the inode |
945 | * address_space which maps the page from disk; whereas "page_mapped" | 945 | * address_space which maps the page from disk; whereas "page_mapped" |
946 | * refers to user virtual address space into which the page is mapped. | 946 | * refers to user virtual address space into which the page is mapped. |
947 | */ | 947 | */ |
948 | #define PAGE_MAPPING_ANON 1 | 948 | #define PAGE_MAPPING_ANON 1 |
949 | #define PAGE_MAPPING_KSM 2 | 949 | #define PAGE_MAPPING_KSM 2 |
950 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) | 950 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) |
951 | 951 | ||
952 | extern struct address_space *page_mapping(struct page *page); | 952 | extern struct address_space *page_mapping(struct page *page); |
953 | 953 | ||
954 | /* Neutral page->mapping pointer to address_space or anon_vma or other */ | 954 | /* Neutral page->mapping pointer to address_space or anon_vma or other */ |
955 | static inline void *page_rmapping(struct page *page) | 955 | static inline void *page_rmapping(struct page *page) |
956 | { | 956 | { |
957 | return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); | 957 | return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); |
958 | } | 958 | } |
959 | 959 | ||
960 | extern struct address_space *__page_file_mapping(struct page *); | 960 | extern struct address_space *__page_file_mapping(struct page *); |
961 | 961 | ||
962 | static inline | 962 | static inline |
963 | struct address_space *page_file_mapping(struct page *page) | 963 | struct address_space *page_file_mapping(struct page *page) |
964 | { | 964 | { |
965 | if (unlikely(PageSwapCache(page))) | 965 | if (unlikely(PageSwapCache(page))) |
966 | return __page_file_mapping(page); | 966 | return __page_file_mapping(page); |
967 | 967 | ||
968 | return page->mapping; | 968 | return page->mapping; |
969 | } | 969 | } |
970 | 970 | ||
971 | static inline int PageAnon(struct page *page) | 971 | static inline int PageAnon(struct page *page) |
972 | { | 972 | { |
973 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; | 973 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; |
974 | } | 974 | } |
975 | 975 | ||
976 | /* | 976 | /* |
977 | * Return the pagecache index of the passed page. Regular pagecache pages | 977 | * Return the pagecache index of the passed page. Regular pagecache pages |
978 | * use ->index whereas swapcache pages use ->private | 978 | * use ->index whereas swapcache pages use ->private |
979 | */ | 979 | */ |
980 | static inline pgoff_t page_index(struct page *page) | 980 | static inline pgoff_t page_index(struct page *page) |
981 | { | 981 | { |
982 | if (unlikely(PageSwapCache(page))) | 982 | if (unlikely(PageSwapCache(page))) |
983 | return page_private(page); | 983 | return page_private(page); |
984 | return page->index; | 984 | return page->index; |
985 | } | 985 | } |
986 | 986 | ||
987 | extern pgoff_t __page_file_index(struct page *page); | 987 | extern pgoff_t __page_file_index(struct page *page); |
988 | 988 | ||
989 | /* | 989 | /* |
990 | * Return the file index of the page. Regular pagecache pages use ->index | 990 | * Return the file index of the page. Regular pagecache pages use ->index |
991 | * whereas swapcache pages use swp_offset(->private) | 991 | * whereas swapcache pages use swp_offset(->private) |
992 | */ | 992 | */ |
993 | static inline pgoff_t page_file_index(struct page *page) | 993 | static inline pgoff_t page_file_index(struct page *page) |
994 | { | 994 | { |
995 | if (unlikely(PageSwapCache(page))) | 995 | if (unlikely(PageSwapCache(page))) |
996 | return __page_file_index(page); | 996 | return __page_file_index(page); |
997 | 997 | ||
998 | return page->index; | 998 | return page->index; |
999 | } | 999 | } |
1000 | 1000 | ||
1001 | /* | 1001 | /* |
1002 | * Return true if this page is mapped into pagetables. | 1002 | * Return true if this page is mapped into pagetables. |
1003 | */ | 1003 | */ |
1004 | static inline int page_mapped(struct page *page) | 1004 | static inline int page_mapped(struct page *page) |
1005 | { | 1005 | { |
1006 | return atomic_read(&(page)->_mapcount) >= 0; | 1006 | return atomic_read(&(page)->_mapcount) >= 0; |
1007 | } | 1007 | } |
1008 | 1008 | ||
1009 | /* | 1009 | /* |
1010 | * Different kinds of faults, as returned by handle_mm_fault(). | 1010 | * Different kinds of faults, as returned by handle_mm_fault(). |
1011 | * Used to decide whether a process gets delivered SIGBUS or | 1011 | * Used to decide whether a process gets delivered SIGBUS or |
1012 | * just gets major/minor fault counters bumped up. | 1012 | * just gets major/minor fault counters bumped up. |
1013 | */ | 1013 | */ |
1014 | 1014 | ||
1015 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ | 1015 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ |
1016 | 1016 | ||
1017 | #define VM_FAULT_OOM 0x0001 | 1017 | #define VM_FAULT_OOM 0x0001 |
1018 | #define VM_FAULT_SIGBUS 0x0002 | 1018 | #define VM_FAULT_SIGBUS 0x0002 |
1019 | #define VM_FAULT_MAJOR 0x0004 | 1019 | #define VM_FAULT_MAJOR 0x0004 |
1020 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ | 1020 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ |
1021 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ | 1021 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ |
1022 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ | 1022 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ |
1023 | 1023 | ||
1024 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ | 1024 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ |
1025 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ | 1025 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ |
1026 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ | 1026 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ |
1027 | #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */ | 1027 | #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */ |
1028 | 1028 | ||
1029 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ | 1029 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ |
1030 | 1030 | ||
1031 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \ | 1031 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \ |
1032 | VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE) | 1032 | VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE) |
1033 | 1033 | ||
1034 | /* Encode hstate index for a hwpoisoned large page */ | 1034 | /* Encode hstate index for a hwpoisoned large page */ |
1035 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) | 1035 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) |
1036 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) | 1036 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) |
1037 | 1037 | ||
1038 | /* | 1038 | /* |
1039 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | 1039 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. |
1040 | */ | 1040 | */ |
1041 | extern void pagefault_out_of_memory(void); | 1041 | extern void pagefault_out_of_memory(void); |
1042 | 1042 | ||
1043 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) | 1043 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
1044 | 1044 | ||
1045 | /* | 1045 | /* |
1046 | * Flags passed to show_mem() and show_free_areas() to suppress output in | 1046 | * Flags passed to show_mem() and show_free_areas() to suppress output in |
1047 | * various contexts. | 1047 | * various contexts. |
1048 | */ | 1048 | */ |
1049 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ | 1049 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ |
1050 | 1050 | ||
1051 | extern void show_free_areas(unsigned int flags); | 1051 | extern void show_free_areas(unsigned int flags); |
1052 | extern bool skip_free_areas_node(unsigned int flags, int nid); | 1052 | extern bool skip_free_areas_node(unsigned int flags, int nid); |
1053 | 1053 | ||
1054 | int shmem_zero_setup(struct vm_area_struct *); | 1054 | int shmem_zero_setup(struct vm_area_struct *); |
1055 | #ifdef CONFIG_SHMEM | 1055 | #ifdef CONFIG_SHMEM |
1056 | bool shmem_mapping(struct address_space *mapping); | 1056 | bool shmem_mapping(struct address_space *mapping); |
1057 | #else | 1057 | #else |
1058 | static inline bool shmem_mapping(struct address_space *mapping) | 1058 | static inline bool shmem_mapping(struct address_space *mapping) |
1059 | { | 1059 | { |
1060 | return false; | 1060 | return false; |
1061 | } | 1061 | } |
1062 | #endif | 1062 | #endif |
1063 | 1063 | ||
1064 | extern int can_do_mlock(void); | 1064 | extern int can_do_mlock(void); |
1065 | extern int user_shm_lock(size_t, struct user_struct *); | 1065 | extern int user_shm_lock(size_t, struct user_struct *); |
1066 | extern void user_shm_unlock(size_t, struct user_struct *); | 1066 | extern void user_shm_unlock(size_t, struct user_struct *); |
1067 | 1067 | ||
1068 | /* | 1068 | /* |
1069 | * Parameter block passed down to zap_pte_range in exceptional cases. | 1069 | * Parameter block passed down to zap_pte_range in exceptional cases. |
1070 | */ | 1070 | */ |
1071 | struct zap_details { | 1071 | struct zap_details { |
1072 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ | 1072 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ |
1073 | struct address_space *check_mapping; /* Check page->mapping if set */ | 1073 | struct address_space *check_mapping; /* Check page->mapping if set */ |
1074 | pgoff_t first_index; /* Lowest page->index to unmap */ | 1074 | pgoff_t first_index; /* Lowest page->index to unmap */ |
1075 | pgoff_t last_index; /* Highest page->index to unmap */ | 1075 | pgoff_t last_index; /* Highest page->index to unmap */ |
1076 | }; | 1076 | }; |
1077 | 1077 | ||
1078 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, | 1078 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
1079 | pte_t pte); | 1079 | pte_t pte); |
1080 | 1080 | ||
1081 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, | 1081 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
1082 | unsigned long size); | 1082 | unsigned long size); |
1083 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, | 1083 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, |
1084 | unsigned long size, struct zap_details *); | 1084 | unsigned long size, struct zap_details *); |
1085 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, | 1085 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
1086 | unsigned long start, unsigned long end); | 1086 | unsigned long start, unsigned long end); |
1087 | 1087 | ||
1088 | /** | 1088 | /** |
1089 | * mm_walk - callbacks for walk_page_range | 1089 | * mm_walk - callbacks for walk_page_range |
1090 | * @pgd_entry: if set, called for each non-empty PGD (top-level) entry | 1090 | * @pgd_entry: if set, called for each non-empty PGD (top-level) entry |
1091 | * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry | 1091 | * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry |
1092 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry | 1092 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry |
1093 | * this handler is required to be able to handle | 1093 | * this handler is required to be able to handle |
1094 | * pmd_trans_huge() pmds. They may simply choose to | 1094 | * pmd_trans_huge() pmds. They may simply choose to |
1095 | * split_huge_page() instead of handling it explicitly. | 1095 | * split_huge_page() instead of handling it explicitly. |
1096 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry | 1096 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry |
1097 | * @pte_hole: if set, called for each hole at all levels | 1097 | * @pte_hole: if set, called for each hole at all levels |
1098 | * @hugetlb_entry: if set, called for each hugetlb entry | 1098 | * @hugetlb_entry: if set, called for each hugetlb entry |
1099 | * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry | 1099 | * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry |
1100 | * is used. | 1100 | * is used. |
1101 | * | 1101 | * |
1102 | * (see walk_page_range for more details) | 1102 | * (see walk_page_range for more details) |
1103 | */ | 1103 | */ |
1104 | struct mm_walk { | 1104 | struct mm_walk { |
1105 | int (*pgd_entry)(pgd_t *pgd, unsigned long addr, | 1105 | int (*pgd_entry)(pgd_t *pgd, unsigned long addr, |
1106 | unsigned long next, struct mm_walk *walk); | 1106 | unsigned long next, struct mm_walk *walk); |
1107 | int (*pud_entry)(pud_t *pud, unsigned long addr, | 1107 | int (*pud_entry)(pud_t *pud, unsigned long addr, |
1108 | unsigned long next, struct mm_walk *walk); | 1108 | unsigned long next, struct mm_walk *walk); |
1109 | int (*pmd_entry)(pmd_t *pmd, unsigned long addr, | 1109 | int (*pmd_entry)(pmd_t *pmd, unsigned long addr, |
1110 | unsigned long next, struct mm_walk *walk); | 1110 | unsigned long next, struct mm_walk *walk); |
1111 | int (*pte_entry)(pte_t *pte, unsigned long addr, | 1111 | int (*pte_entry)(pte_t *pte, unsigned long addr, |
1112 | unsigned long next, struct mm_walk *walk); | 1112 | unsigned long next, struct mm_walk *walk); |
1113 | int (*pte_hole)(unsigned long addr, unsigned long next, | 1113 | int (*pte_hole)(unsigned long addr, unsigned long next, |
1114 | struct mm_walk *walk); | 1114 | struct mm_walk *walk); |
1115 | int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, | 1115 | int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, |
1116 | unsigned long addr, unsigned long next, | 1116 | unsigned long addr, unsigned long next, |
1117 | struct mm_walk *walk); | 1117 | struct mm_walk *walk); |
1118 | struct mm_struct *mm; | 1118 | struct mm_struct *mm; |
1119 | void *private; | 1119 | void *private; |
1120 | }; | 1120 | }; |
1121 | 1121 | ||
1122 | int walk_page_range(unsigned long addr, unsigned long end, | 1122 | int walk_page_range(unsigned long addr, unsigned long end, |
1123 | struct mm_walk *walk); | 1123 | struct mm_walk *walk); |
1124 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, | 1124 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
1125 | unsigned long end, unsigned long floor, unsigned long ceiling); | 1125 | unsigned long end, unsigned long floor, unsigned long ceiling); |
1126 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, | 1126 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
1127 | struct vm_area_struct *vma); | 1127 | struct vm_area_struct *vma); |
1128 | void unmap_mapping_range(struct address_space *mapping, | 1128 | void unmap_mapping_range(struct address_space *mapping, |
1129 | loff_t const holebegin, loff_t const holelen, int even_cows); | 1129 | loff_t const holebegin, loff_t const holelen, int even_cows); |
1130 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, | 1130 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
1131 | unsigned long *pfn); | 1131 | unsigned long *pfn); |
1132 | int follow_phys(struct vm_area_struct *vma, unsigned long address, | 1132 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
1133 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | 1133 | unsigned int flags, unsigned long *prot, resource_size_t *phys); |
1134 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, | 1134 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
1135 | void *buf, int len, int write); | 1135 | void *buf, int len, int write); |
1136 | 1136 | ||
1137 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | 1137 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
1138 | loff_t const holebegin, loff_t const holelen) | 1138 | loff_t const holebegin, loff_t const holelen) |
1139 | { | 1139 | { |
1140 | unmap_mapping_range(mapping, holebegin, holelen, 0); | 1140 | unmap_mapping_range(mapping, holebegin, holelen, 0); |
1141 | } | 1141 | } |
1142 | 1142 | ||
1143 | extern void truncate_pagecache(struct inode *inode, loff_t new); | 1143 | extern void truncate_pagecache(struct inode *inode, loff_t new); |
1144 | extern void truncate_setsize(struct inode *inode, loff_t newsize); | 1144 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
1145 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); | 1145 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); |
1146 | int truncate_inode_page(struct address_space *mapping, struct page *page); | 1146 | int truncate_inode_page(struct address_space *mapping, struct page *page); |
1147 | int generic_error_remove_page(struct address_space *mapping, struct page *page); | 1147 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
1148 | int invalidate_inode_page(struct page *page); | 1148 | int invalidate_inode_page(struct page *page); |
1149 | 1149 | ||
1150 | #ifdef CONFIG_MMU | 1150 | #ifdef CONFIG_MMU |
1151 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, | 1151 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, |
1152 | unsigned long address, unsigned int flags); | 1152 | unsigned long address, unsigned int flags); |
1153 | extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, | 1153 | extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, |
1154 | unsigned long address, unsigned int fault_flags); | 1154 | unsigned long address, unsigned int fault_flags); |
1155 | #else | 1155 | #else |
1156 | static inline int handle_mm_fault(struct mm_struct *mm, | 1156 | static inline int handle_mm_fault(struct mm_struct *mm, |
1157 | struct vm_area_struct *vma, unsigned long address, | 1157 | struct vm_area_struct *vma, unsigned long address, |
1158 | unsigned int flags) | 1158 | unsigned int flags) |
1159 | { | 1159 | { |
1160 | /* should never happen if there's no MMU */ | 1160 | /* should never happen if there's no MMU */ |
1161 | BUG(); | 1161 | BUG(); |
1162 | return VM_FAULT_SIGBUS; | 1162 | return VM_FAULT_SIGBUS; |
1163 | } | 1163 | } |
1164 | static inline int fixup_user_fault(struct task_struct *tsk, | 1164 | static inline int fixup_user_fault(struct task_struct *tsk, |
1165 | struct mm_struct *mm, unsigned long address, | 1165 | struct mm_struct *mm, unsigned long address, |
1166 | unsigned int fault_flags) | 1166 | unsigned int fault_flags) |
1167 | { | 1167 | { |
1168 | /* should never happen if there's no MMU */ | 1168 | /* should never happen if there's no MMU */ |
1169 | BUG(); | 1169 | BUG(); |
1170 | return -EFAULT; | 1170 | return -EFAULT; |
1171 | } | 1171 | } |
1172 | #endif | 1172 | #endif |
1173 | 1173 | ||
1174 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); | 1174 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); |
1175 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, | 1175 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
1176 | void *buf, int len, int write); | 1176 | void *buf, int len, int write); |
1177 | 1177 | ||
1178 | long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | 1178 | long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
1179 | unsigned long start, unsigned long nr_pages, | 1179 | unsigned long start, unsigned long nr_pages, |
1180 | unsigned int foll_flags, struct page **pages, | 1180 | unsigned int foll_flags, struct page **pages, |
1181 | struct vm_area_struct **vmas, int *nonblocking); | 1181 | struct vm_area_struct **vmas, int *nonblocking); |
1182 | long get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | 1182 | long get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
1183 | unsigned long start, unsigned long nr_pages, | 1183 | unsigned long start, unsigned long nr_pages, |
1184 | int write, int force, struct page **pages, | 1184 | int write, int force, struct page **pages, |
1185 | struct vm_area_struct **vmas); | 1185 | struct vm_area_struct **vmas); |
1186 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, | 1186 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, |
1187 | struct page **pages); | 1187 | struct page **pages); |
1188 | struct kvec; | 1188 | struct kvec; |
1189 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, | 1189 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, |
1190 | struct page **pages); | 1190 | struct page **pages); |
1191 | int get_kernel_page(unsigned long start, int write, struct page **pages); | 1191 | int get_kernel_page(unsigned long start, int write, struct page **pages); |
1192 | struct page *get_dump_page(unsigned long addr); | 1192 | struct page *get_dump_page(unsigned long addr); |
1193 | 1193 | ||
1194 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); | 1194 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
1195 | extern void do_invalidatepage(struct page *page, unsigned int offset, | 1195 | extern void do_invalidatepage(struct page *page, unsigned int offset, |
1196 | unsigned int length); | 1196 | unsigned int length); |
1197 | 1197 | ||
1198 | int __set_page_dirty_nobuffers(struct page *page); | 1198 | int __set_page_dirty_nobuffers(struct page *page); |
1199 | int __set_page_dirty_no_writeback(struct page *page); | 1199 | int __set_page_dirty_no_writeback(struct page *page); |
1200 | int redirty_page_for_writepage(struct writeback_control *wbc, | 1200 | int redirty_page_for_writepage(struct writeback_control *wbc, |
1201 | struct page *page); | 1201 | struct page *page); |
1202 | void account_page_dirtied(struct page *page, struct address_space *mapping); | 1202 | void account_page_dirtied(struct page *page, struct address_space *mapping); |
1203 | void account_page_writeback(struct page *page); | 1203 | void account_page_writeback(struct page *page); |
1204 | int set_page_dirty(struct page *page); | 1204 | int set_page_dirty(struct page *page); |
1205 | int set_page_dirty_lock(struct page *page); | 1205 | int set_page_dirty_lock(struct page *page); |
1206 | int clear_page_dirty_for_io(struct page *page); | 1206 | int clear_page_dirty_for_io(struct page *page); |
1207 | 1207 | ||
1208 | /* Is the vma a continuation of the stack vma above it? */ | 1208 | /* Is the vma a continuation of the stack vma above it? */ |
1209 | static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) | 1209 | static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) |
1210 | { | 1210 | { |
1211 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); | 1211 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); |
1212 | } | 1212 | } |
1213 | 1213 | ||
1214 | static inline int stack_guard_page_start(struct vm_area_struct *vma, | 1214 | static inline int stack_guard_page_start(struct vm_area_struct *vma, |
1215 | unsigned long addr) | 1215 | unsigned long addr) |
1216 | { | 1216 | { |
1217 | return (vma->vm_flags & VM_GROWSDOWN) && | 1217 | return (vma->vm_flags & VM_GROWSDOWN) && |
1218 | (vma->vm_start == addr) && | 1218 | (vma->vm_start == addr) && |
1219 | !vma_growsdown(vma->vm_prev, addr); | 1219 | !vma_growsdown(vma->vm_prev, addr); |
1220 | } | 1220 | } |
1221 | 1221 | ||
1222 | /* Is the vma a continuation of the stack vma below it? */ | 1222 | /* Is the vma a continuation of the stack vma below it? */ |
1223 | static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) | 1223 | static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) |
1224 | { | 1224 | { |
1225 | return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); | 1225 | return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); |
1226 | } | 1226 | } |
1227 | 1227 | ||
1228 | static inline int stack_guard_page_end(struct vm_area_struct *vma, | 1228 | static inline int stack_guard_page_end(struct vm_area_struct *vma, |
1229 | unsigned long addr) | 1229 | unsigned long addr) |
1230 | { | 1230 | { |
1231 | return (vma->vm_flags & VM_GROWSUP) && | 1231 | return (vma->vm_flags & VM_GROWSUP) && |
1232 | (vma->vm_end == addr) && | 1232 | (vma->vm_end == addr) && |
1233 | !vma_growsup(vma->vm_next, addr); | 1233 | !vma_growsup(vma->vm_next, addr); |
1234 | } | 1234 | } |
1235 | 1235 | ||
1236 | extern pid_t | 1236 | extern pid_t |
1237 | vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group); | 1237 | vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group); |
1238 | 1238 | ||
1239 | extern unsigned long move_page_tables(struct vm_area_struct *vma, | 1239 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
1240 | unsigned long old_addr, struct vm_area_struct *new_vma, | 1240 | unsigned long old_addr, struct vm_area_struct *new_vma, |
1241 | unsigned long new_addr, unsigned long len, | 1241 | unsigned long new_addr, unsigned long len, |
1242 | bool need_rmap_locks); | 1242 | bool need_rmap_locks); |
1243 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, | 1243 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, |
1244 | unsigned long end, pgprot_t newprot, | 1244 | unsigned long end, pgprot_t newprot, |
1245 | int dirty_accountable, int prot_numa); | 1245 | int dirty_accountable, int prot_numa); |
1246 | extern int mprotect_fixup(struct vm_area_struct *vma, | 1246 | extern int mprotect_fixup(struct vm_area_struct *vma, |
1247 | struct vm_area_struct **pprev, unsigned long start, | 1247 | struct vm_area_struct **pprev, unsigned long start, |
1248 | unsigned long end, unsigned long newflags); | 1248 | unsigned long end, unsigned long newflags); |
1249 | 1249 | ||
1250 | /* | 1250 | /* |
1251 | * doesn't attempt to fault and will return short. | 1251 | * doesn't attempt to fault and will return short. |
1252 | */ | 1252 | */ |
1253 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, | 1253 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, |
1254 | struct page **pages); | 1254 | struct page **pages); |
1255 | /* | 1255 | /* |
1256 | * per-process(per-mm_struct) statistics. | 1256 | * per-process(per-mm_struct) statistics. |
1257 | */ | 1257 | */ |
1258 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) | 1258 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) |
1259 | { | 1259 | { |
1260 | long val = atomic_long_read(&mm->rss_stat.count[member]); | 1260 | long val = atomic_long_read(&mm->rss_stat.count[member]); |
1261 | 1261 | ||
1262 | #ifdef SPLIT_RSS_COUNTING | 1262 | #ifdef SPLIT_RSS_COUNTING |
1263 | /* | 1263 | /* |
1264 | * counter is updated in asynchronous manner and may go to minus. | 1264 | * counter is updated in asynchronous manner and may go to minus. |
1265 | * But it's never be expected number for users. | 1265 | * But it's never be expected number for users. |
1266 | */ | 1266 | */ |
1267 | if (val < 0) | 1267 | if (val < 0) |
1268 | val = 0; | 1268 | val = 0; |
1269 | #endif | 1269 | #endif |
1270 | return (unsigned long)val; | 1270 | return (unsigned long)val; |
1271 | } | 1271 | } |
1272 | 1272 | ||
1273 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) | 1273 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) |
1274 | { | 1274 | { |
1275 | atomic_long_add(value, &mm->rss_stat.count[member]); | 1275 | atomic_long_add(value, &mm->rss_stat.count[member]); |
1276 | } | 1276 | } |
1277 | 1277 | ||
1278 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | 1278 | static inline void inc_mm_counter(struct mm_struct *mm, int member) |
1279 | { | 1279 | { |
1280 | atomic_long_inc(&mm->rss_stat.count[member]); | 1280 | atomic_long_inc(&mm->rss_stat.count[member]); |
1281 | } | 1281 | } |
1282 | 1282 | ||
1283 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | 1283 | static inline void dec_mm_counter(struct mm_struct *mm, int member) |
1284 | { | 1284 | { |
1285 | atomic_long_dec(&mm->rss_stat.count[member]); | 1285 | atomic_long_dec(&mm->rss_stat.count[member]); |
1286 | } | 1286 | } |
1287 | 1287 | ||
1288 | static inline unsigned long get_mm_rss(struct mm_struct *mm) | 1288 | static inline unsigned long get_mm_rss(struct mm_struct *mm) |
1289 | { | 1289 | { |
1290 | return get_mm_counter(mm, MM_FILEPAGES) + | 1290 | return get_mm_counter(mm, MM_FILEPAGES) + |
1291 | get_mm_counter(mm, MM_ANONPAGES); | 1291 | get_mm_counter(mm, MM_ANONPAGES); |
1292 | } | 1292 | } |
1293 | 1293 | ||
1294 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | 1294 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) |
1295 | { | 1295 | { |
1296 | return max(mm->hiwater_rss, get_mm_rss(mm)); | 1296 | return max(mm->hiwater_rss, get_mm_rss(mm)); |
1297 | } | 1297 | } |
1298 | 1298 | ||
1299 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | 1299 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) |
1300 | { | 1300 | { |
1301 | return max(mm->hiwater_vm, mm->total_vm); | 1301 | return max(mm->hiwater_vm, mm->total_vm); |
1302 | } | 1302 | } |
1303 | 1303 | ||
1304 | static inline void update_hiwater_rss(struct mm_struct *mm) | 1304 | static inline void update_hiwater_rss(struct mm_struct *mm) |
1305 | { | 1305 | { |
1306 | unsigned long _rss = get_mm_rss(mm); | 1306 | unsigned long _rss = get_mm_rss(mm); |
1307 | 1307 | ||
1308 | if ((mm)->hiwater_rss < _rss) | 1308 | if ((mm)->hiwater_rss < _rss) |
1309 | (mm)->hiwater_rss = _rss; | 1309 | (mm)->hiwater_rss = _rss; |
1310 | } | 1310 | } |
1311 | 1311 | ||
1312 | static inline void update_hiwater_vm(struct mm_struct *mm) | 1312 | static inline void update_hiwater_vm(struct mm_struct *mm) |
1313 | { | 1313 | { |
1314 | if (mm->hiwater_vm < mm->total_vm) | 1314 | if (mm->hiwater_vm < mm->total_vm) |
1315 | mm->hiwater_vm = mm->total_vm; | 1315 | mm->hiwater_vm = mm->total_vm; |
1316 | } | 1316 | } |
1317 | 1317 | ||
1318 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, | 1318 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, |
1319 | struct mm_struct *mm) | 1319 | struct mm_struct *mm) |
1320 | { | 1320 | { |
1321 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | 1321 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); |
1322 | 1322 | ||
1323 | if (*maxrss < hiwater_rss) | 1323 | if (*maxrss < hiwater_rss) |
1324 | *maxrss = hiwater_rss; | 1324 | *maxrss = hiwater_rss; |
1325 | } | 1325 | } |
1326 | 1326 | ||
1327 | #if defined(SPLIT_RSS_COUNTING) | 1327 | #if defined(SPLIT_RSS_COUNTING) |
1328 | void sync_mm_rss(struct mm_struct *mm); | 1328 | void sync_mm_rss(struct mm_struct *mm); |
1329 | #else | 1329 | #else |
1330 | static inline void sync_mm_rss(struct mm_struct *mm) | 1330 | static inline void sync_mm_rss(struct mm_struct *mm) |
1331 | { | 1331 | { |
1332 | } | 1332 | } |
1333 | #endif | 1333 | #endif |
1334 | 1334 | ||
1335 | int vma_wants_writenotify(struct vm_area_struct *vma); | 1335 | int vma_wants_writenotify(struct vm_area_struct *vma); |
1336 | 1336 | ||
1337 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, | 1337 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
1338 | spinlock_t **ptl); | 1338 | spinlock_t **ptl); |
1339 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | 1339 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, |
1340 | spinlock_t **ptl) | 1340 | spinlock_t **ptl) |
1341 | { | 1341 | { |
1342 | pte_t *ptep; | 1342 | pte_t *ptep; |
1343 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | 1343 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); |
1344 | return ptep; | 1344 | return ptep; |
1345 | } | 1345 | } |
1346 | 1346 | ||
1347 | #ifdef __PAGETABLE_PUD_FOLDED | 1347 | #ifdef __PAGETABLE_PUD_FOLDED |
1348 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, | 1348 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, |
1349 | unsigned long address) | 1349 | unsigned long address) |
1350 | { | 1350 | { |
1351 | return 0; | 1351 | return 0; |
1352 | } | 1352 | } |
1353 | #else | 1353 | #else |
1354 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); | 1354 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
1355 | #endif | 1355 | #endif |
1356 | 1356 | ||
1357 | #ifdef __PAGETABLE_PMD_FOLDED | 1357 | #ifdef __PAGETABLE_PMD_FOLDED |
1358 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, | 1358 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
1359 | unsigned long address) | 1359 | unsigned long address) |
1360 | { | 1360 | { |
1361 | return 0; | 1361 | return 0; |
1362 | } | 1362 | } |
1363 | #else | 1363 | #else |
1364 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); | 1364 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
1365 | #endif | 1365 | #endif |
1366 | 1366 | ||
1367 | int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, | 1367 | int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, |
1368 | pmd_t *pmd, unsigned long address); | 1368 | pmd_t *pmd, unsigned long address); |
1369 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); | 1369 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); |
1370 | 1370 | ||
1371 | /* | 1371 | /* |
1372 | * The following ifdef needed to get the 4level-fixup.h header to work. | 1372 | * The following ifdef needed to get the 4level-fixup.h header to work. |
1373 | * Remove it when 4level-fixup.h has been removed. | 1373 | * Remove it when 4level-fixup.h has been removed. |
1374 | */ | 1374 | */ |
1375 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) | 1375 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
1376 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) | 1376 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
1377 | { | 1377 | { |
1378 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? | 1378 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? |
1379 | NULL: pud_offset(pgd, address); | 1379 | NULL: pud_offset(pgd, address); |
1380 | } | 1380 | } |
1381 | 1381 | ||
1382 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | 1382 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
1383 | { | 1383 | { |
1384 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? | 1384 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
1385 | NULL: pmd_offset(pud, address); | 1385 | NULL: pmd_offset(pud, address); |
1386 | } | 1386 | } |
1387 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ | 1387 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
1388 | 1388 | ||
1389 | #if USE_SPLIT_PTE_PTLOCKS | 1389 | #if USE_SPLIT_PTE_PTLOCKS |
1390 | #if ALLOC_SPLIT_PTLOCKS | 1390 | #if ALLOC_SPLIT_PTLOCKS |
1391 | void __init ptlock_cache_init(void); | 1391 | void __init ptlock_cache_init(void); |
1392 | extern bool ptlock_alloc(struct page *page); | 1392 | extern bool ptlock_alloc(struct page *page); |
1393 | extern void ptlock_free(struct page *page); | 1393 | extern void ptlock_free(struct page *page); |
1394 | 1394 | ||
1395 | static inline spinlock_t *ptlock_ptr(struct page *page) | 1395 | static inline spinlock_t *ptlock_ptr(struct page *page) |
1396 | { | 1396 | { |
1397 | return page->ptl; | 1397 | return page->ptl; |
1398 | } | 1398 | } |
1399 | #else /* ALLOC_SPLIT_PTLOCKS */ | 1399 | #else /* ALLOC_SPLIT_PTLOCKS */ |
1400 | static inline void ptlock_cache_init(void) | 1400 | static inline void ptlock_cache_init(void) |
1401 | { | 1401 | { |
1402 | } | 1402 | } |
1403 | 1403 | ||
1404 | static inline bool ptlock_alloc(struct page *page) | 1404 | static inline bool ptlock_alloc(struct page *page) |
1405 | { | 1405 | { |
1406 | return true; | 1406 | return true; |
1407 | } | 1407 | } |
1408 | 1408 | ||
1409 | static inline void ptlock_free(struct page *page) | 1409 | static inline void ptlock_free(struct page *page) |
1410 | { | 1410 | { |
1411 | } | 1411 | } |
1412 | 1412 | ||
1413 | static inline spinlock_t *ptlock_ptr(struct page *page) | 1413 | static inline spinlock_t *ptlock_ptr(struct page *page) |
1414 | { | 1414 | { |
1415 | return &page->ptl; | 1415 | return &page->ptl; |
1416 | } | 1416 | } |
1417 | #endif /* ALLOC_SPLIT_PTLOCKS */ | 1417 | #endif /* ALLOC_SPLIT_PTLOCKS */ |
1418 | 1418 | ||
1419 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | 1419 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
1420 | { | 1420 | { |
1421 | return ptlock_ptr(pmd_page(*pmd)); | 1421 | return ptlock_ptr(pmd_page(*pmd)); |
1422 | } | 1422 | } |
1423 | 1423 | ||
1424 | static inline bool ptlock_init(struct page *page) | 1424 | static inline bool ptlock_init(struct page *page) |
1425 | { | 1425 | { |
1426 | /* | 1426 | /* |
1427 | * prep_new_page() initialize page->private (and therefore page->ptl) | 1427 | * prep_new_page() initialize page->private (and therefore page->ptl) |
1428 | * with 0. Make sure nobody took it in use in between. | 1428 | * with 0. Make sure nobody took it in use in between. |
1429 | * | 1429 | * |
1430 | * It can happen if arch try to use slab for page table allocation: | 1430 | * It can happen if arch try to use slab for page table allocation: |
1431 | * slab code uses page->slab_cache and page->first_page (for tail | 1431 | * slab code uses page->slab_cache and page->first_page (for tail |
1432 | * pages), which share storage with page->ptl. | 1432 | * pages), which share storage with page->ptl. |
1433 | */ | 1433 | */ |
1434 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); | 1434 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); |
1435 | if (!ptlock_alloc(page)) | 1435 | if (!ptlock_alloc(page)) |
1436 | return false; | 1436 | return false; |
1437 | spin_lock_init(ptlock_ptr(page)); | 1437 | spin_lock_init(ptlock_ptr(page)); |
1438 | return true; | 1438 | return true; |
1439 | } | 1439 | } |
1440 | 1440 | ||
1441 | /* Reset page->mapping so free_pages_check won't complain. */ | 1441 | /* Reset page->mapping so free_pages_check won't complain. */ |
1442 | static inline void pte_lock_deinit(struct page *page) | 1442 | static inline void pte_lock_deinit(struct page *page) |
1443 | { | 1443 | { |
1444 | page->mapping = NULL; | 1444 | page->mapping = NULL; |
1445 | ptlock_free(page); | 1445 | ptlock_free(page); |
1446 | } | 1446 | } |
1447 | 1447 | ||
1448 | #else /* !USE_SPLIT_PTE_PTLOCKS */ | 1448 | #else /* !USE_SPLIT_PTE_PTLOCKS */ |
1449 | /* | 1449 | /* |
1450 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | 1450 | * We use mm->page_table_lock to guard all pagetable pages of the mm. |
1451 | */ | 1451 | */ |
1452 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | 1452 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
1453 | { | 1453 | { |
1454 | return &mm->page_table_lock; | 1454 | return &mm->page_table_lock; |
1455 | } | 1455 | } |
1456 | static inline void ptlock_cache_init(void) {} | 1456 | static inline void ptlock_cache_init(void) {} |
1457 | static inline bool ptlock_init(struct page *page) { return true; } | 1457 | static inline bool ptlock_init(struct page *page) { return true; } |
1458 | static inline void pte_lock_deinit(struct page *page) {} | 1458 | static inline void pte_lock_deinit(struct page *page) {} |
1459 | #endif /* USE_SPLIT_PTE_PTLOCKS */ | 1459 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
1460 | 1460 | ||
1461 | static inline void pgtable_init(void) | 1461 | static inline void pgtable_init(void) |
1462 | { | 1462 | { |
1463 | ptlock_cache_init(); | 1463 | ptlock_cache_init(); |
1464 | pgtable_cache_init(); | 1464 | pgtable_cache_init(); |
1465 | } | 1465 | } |
1466 | 1466 | ||
1467 | static inline bool pgtable_page_ctor(struct page *page) | 1467 | static inline bool pgtable_page_ctor(struct page *page) |
1468 | { | 1468 | { |
1469 | inc_zone_page_state(page, NR_PAGETABLE); | 1469 | inc_zone_page_state(page, NR_PAGETABLE); |
1470 | return ptlock_init(page); | 1470 | return ptlock_init(page); |
1471 | } | 1471 | } |
1472 | 1472 | ||
1473 | static inline void pgtable_page_dtor(struct page *page) | 1473 | static inline void pgtable_page_dtor(struct page *page) |
1474 | { | 1474 | { |
1475 | pte_lock_deinit(page); | 1475 | pte_lock_deinit(page); |
1476 | dec_zone_page_state(page, NR_PAGETABLE); | 1476 | dec_zone_page_state(page, NR_PAGETABLE); |
1477 | } | 1477 | } |
1478 | 1478 | ||
1479 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ | 1479 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
1480 | ({ \ | 1480 | ({ \ |
1481 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ | 1481 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
1482 | pte_t *__pte = pte_offset_map(pmd, address); \ | 1482 | pte_t *__pte = pte_offset_map(pmd, address); \ |
1483 | *(ptlp) = __ptl; \ | 1483 | *(ptlp) = __ptl; \ |
1484 | spin_lock(__ptl); \ | 1484 | spin_lock(__ptl); \ |
1485 | __pte; \ | 1485 | __pte; \ |
1486 | }) | 1486 | }) |
1487 | 1487 | ||
1488 | #define pte_unmap_unlock(pte, ptl) do { \ | 1488 | #define pte_unmap_unlock(pte, ptl) do { \ |
1489 | spin_unlock(ptl); \ | 1489 | spin_unlock(ptl); \ |
1490 | pte_unmap(pte); \ | 1490 | pte_unmap(pte); \ |
1491 | } while (0) | 1491 | } while (0) |
1492 | 1492 | ||
1493 | #define pte_alloc_map(mm, vma, pmd, address) \ | 1493 | #define pte_alloc_map(mm, vma, pmd, address) \ |
1494 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ | 1494 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ |
1495 | pmd, address))? \ | 1495 | pmd, address))? \ |
1496 | NULL: pte_offset_map(pmd, address)) | 1496 | NULL: pte_offset_map(pmd, address)) |
1497 | 1497 | ||
1498 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ | 1498 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
1499 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ | 1499 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ |
1500 | pmd, address))? \ | 1500 | pmd, address))? \ |
1501 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) | 1501 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) |
1502 | 1502 | ||
1503 | #define pte_alloc_kernel(pmd, address) \ | 1503 | #define pte_alloc_kernel(pmd, address) \ |
1504 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ | 1504 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ |
1505 | NULL: pte_offset_kernel(pmd, address)) | 1505 | NULL: pte_offset_kernel(pmd, address)) |
1506 | 1506 | ||
1507 | #if USE_SPLIT_PMD_PTLOCKS | 1507 | #if USE_SPLIT_PMD_PTLOCKS |
1508 | 1508 | ||
1509 | static struct page *pmd_to_page(pmd_t *pmd) | 1509 | static struct page *pmd_to_page(pmd_t *pmd) |
1510 | { | 1510 | { |
1511 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); | 1511 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); |
1512 | return virt_to_page((void *)((unsigned long) pmd & mask)); | 1512 | return virt_to_page((void *)((unsigned long) pmd & mask)); |
1513 | } | 1513 | } |
1514 | 1514 | ||
1515 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) | 1515 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
1516 | { | 1516 | { |
1517 | return ptlock_ptr(pmd_to_page(pmd)); | 1517 | return ptlock_ptr(pmd_to_page(pmd)); |
1518 | } | 1518 | } |
1519 | 1519 | ||
1520 | static inline bool pgtable_pmd_page_ctor(struct page *page) | 1520 | static inline bool pgtable_pmd_page_ctor(struct page *page) |
1521 | { | 1521 | { |
1522 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | 1522 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
1523 | page->pmd_huge_pte = NULL; | 1523 | page->pmd_huge_pte = NULL; |
1524 | #endif | 1524 | #endif |
1525 | return ptlock_init(page); | 1525 | return ptlock_init(page); |
1526 | } | 1526 | } |
1527 | 1527 | ||
1528 | static inline void pgtable_pmd_page_dtor(struct page *page) | 1528 | static inline void pgtable_pmd_page_dtor(struct page *page) |
1529 | { | 1529 | { |
1530 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | 1530 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
1531 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); | 1531 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); |
1532 | #endif | 1532 | #endif |
1533 | ptlock_free(page); | 1533 | ptlock_free(page); |
1534 | } | 1534 | } |
1535 | 1535 | ||
1536 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) | 1536 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) |
1537 | 1537 | ||
1538 | #else | 1538 | #else |
1539 | 1539 | ||
1540 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) | 1540 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
1541 | { | 1541 | { |
1542 | return &mm->page_table_lock; | 1542 | return &mm->page_table_lock; |
1543 | } | 1543 | } |
1544 | 1544 | ||
1545 | static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } | 1545 | static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } |
1546 | static inline void pgtable_pmd_page_dtor(struct page *page) {} | 1546 | static inline void pgtable_pmd_page_dtor(struct page *page) {} |
1547 | 1547 | ||
1548 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) | 1548 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) |
1549 | 1549 | ||
1550 | #endif | 1550 | #endif |
1551 | 1551 | ||
1552 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) | 1552 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) |
1553 | { | 1553 | { |
1554 | spinlock_t *ptl = pmd_lockptr(mm, pmd); | 1554 | spinlock_t *ptl = pmd_lockptr(mm, pmd); |
1555 | spin_lock(ptl); | 1555 | spin_lock(ptl); |
1556 | return ptl; | 1556 | return ptl; |
1557 | } | 1557 | } |
1558 | 1558 | ||
1559 | extern void free_area_init(unsigned long * zones_size); | 1559 | extern void free_area_init(unsigned long * zones_size); |
1560 | extern void free_area_init_node(int nid, unsigned long * zones_size, | 1560 | extern void free_area_init_node(int nid, unsigned long * zones_size, |
1561 | unsigned long zone_start_pfn, unsigned long *zholes_size); | 1561 | unsigned long zone_start_pfn, unsigned long *zholes_size); |
1562 | extern void free_initmem(void); | 1562 | extern void free_initmem(void); |
1563 | 1563 | ||
1564 | /* | 1564 | /* |
1565 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) | 1565 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) |
1566 | * into the buddy system. The freed pages will be poisoned with pattern | 1566 | * into the buddy system. The freed pages will be poisoned with pattern |
1567 | * "poison" if it's within range [0, UCHAR_MAX]. | 1567 | * "poison" if it's within range [0, UCHAR_MAX]. |
1568 | * Return pages freed into the buddy system. | 1568 | * Return pages freed into the buddy system. |
1569 | */ | 1569 | */ |
1570 | extern unsigned long free_reserved_area(void *start, void *end, | 1570 | extern unsigned long free_reserved_area(void *start, void *end, |
1571 | int poison, char *s); | 1571 | int poison, char *s); |
1572 | 1572 | ||
1573 | #ifdef CONFIG_HIGHMEM | 1573 | #ifdef CONFIG_HIGHMEM |
1574 | /* | 1574 | /* |
1575 | * Free a highmem page into the buddy system, adjusting totalhigh_pages | 1575 | * Free a highmem page into the buddy system, adjusting totalhigh_pages |
1576 | * and totalram_pages. | 1576 | * and totalram_pages. |
1577 | */ | 1577 | */ |
1578 | extern void free_highmem_page(struct page *page); | 1578 | extern void free_highmem_page(struct page *page); |
1579 | #endif | 1579 | #endif |
1580 | 1580 | ||
1581 | extern void adjust_managed_page_count(struct page *page, long count); | 1581 | extern void adjust_managed_page_count(struct page *page, long count); |
1582 | extern void mem_init_print_info(const char *str); | 1582 | extern void mem_init_print_info(const char *str); |
1583 | 1583 | ||
1584 | /* Free the reserved page into the buddy system, so it gets managed. */ | 1584 | /* Free the reserved page into the buddy system, so it gets managed. */ |
1585 | static inline void __free_reserved_page(struct page *page) | 1585 | static inline void __free_reserved_page(struct page *page) |
1586 | { | 1586 | { |
1587 | ClearPageReserved(page); | 1587 | ClearPageReserved(page); |
1588 | init_page_count(page); | 1588 | init_page_count(page); |
1589 | __free_page(page); | 1589 | __free_page(page); |
1590 | } | 1590 | } |
1591 | 1591 | ||
1592 | static inline void free_reserved_page(struct page *page) | 1592 | static inline void free_reserved_page(struct page *page) |
1593 | { | 1593 | { |
1594 | __free_reserved_page(page); | 1594 | __free_reserved_page(page); |
1595 | adjust_managed_page_count(page, 1); | 1595 | adjust_managed_page_count(page, 1); |
1596 | } | 1596 | } |
1597 | 1597 | ||
1598 | static inline void mark_page_reserved(struct page *page) | 1598 | static inline void mark_page_reserved(struct page *page) |
1599 | { | 1599 | { |
1600 | SetPageReserved(page); | 1600 | SetPageReserved(page); |
1601 | adjust_managed_page_count(page, -1); | 1601 | adjust_managed_page_count(page, -1); |
1602 | } | 1602 | } |
1603 | 1603 | ||
1604 | /* | 1604 | /* |
1605 | * Default method to free all the __init memory into the buddy system. | 1605 | * Default method to free all the __init memory into the buddy system. |
1606 | * The freed pages will be poisoned with pattern "poison" if it's within | 1606 | * The freed pages will be poisoned with pattern "poison" if it's within |
1607 | * range [0, UCHAR_MAX]. | 1607 | * range [0, UCHAR_MAX]. |
1608 | * Return pages freed into the buddy system. | 1608 | * Return pages freed into the buddy system. |
1609 | */ | 1609 | */ |
1610 | static inline unsigned long free_initmem_default(int poison) | 1610 | static inline unsigned long free_initmem_default(int poison) |
1611 | { | 1611 | { |
1612 | extern char __init_begin[], __init_end[]; | 1612 | extern char __init_begin[], __init_end[]; |
1613 | 1613 | ||
1614 | return free_reserved_area(&__init_begin, &__init_end, | 1614 | return free_reserved_area(&__init_begin, &__init_end, |
1615 | poison, "unused kernel"); | 1615 | poison, "unused kernel"); |
1616 | } | 1616 | } |
1617 | 1617 | ||
1618 | static inline unsigned long get_num_physpages(void) | 1618 | static inline unsigned long get_num_physpages(void) |
1619 | { | 1619 | { |
1620 | int nid; | 1620 | int nid; |
1621 | unsigned long phys_pages = 0; | 1621 | unsigned long phys_pages = 0; |
1622 | 1622 | ||
1623 | for_each_online_node(nid) | 1623 | for_each_online_node(nid) |
1624 | phys_pages += node_present_pages(nid); | 1624 | phys_pages += node_present_pages(nid); |
1625 | 1625 | ||
1626 | return phys_pages; | 1626 | return phys_pages; |
1627 | } | 1627 | } |
1628 | 1628 | ||
1629 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP | 1629 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1630 | /* | 1630 | /* |
1631 | * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its | 1631 | * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its |
1632 | * zones, allocate the backing mem_map and account for memory holes in a more | 1632 | * zones, allocate the backing mem_map and account for memory holes in a more |
1633 | * architecture independent manner. This is a substitute for creating the | 1633 | * architecture independent manner. This is a substitute for creating the |
1634 | * zone_sizes[] and zholes_size[] arrays and passing them to | 1634 | * zone_sizes[] and zholes_size[] arrays and passing them to |
1635 | * free_area_init_node() | 1635 | * free_area_init_node() |
1636 | * | 1636 | * |
1637 | * An architecture is expected to register range of page frames backed by | 1637 | * An architecture is expected to register range of page frames backed by |
1638 | * physical memory with memblock_add[_node]() before calling | 1638 | * physical memory with memblock_add[_node]() before calling |
1639 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic | 1639 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic |
1640 | * usage, an architecture is expected to do something like | 1640 | * usage, an architecture is expected to do something like |
1641 | * | 1641 | * |
1642 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | 1642 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, |
1643 | * max_highmem_pfn}; | 1643 | * max_highmem_pfn}; |
1644 | * for_each_valid_physical_page_range() | 1644 | * for_each_valid_physical_page_range() |
1645 | * memblock_add_node(base, size, nid) | 1645 | * memblock_add_node(base, size, nid) |
1646 | * free_area_init_nodes(max_zone_pfns); | 1646 | * free_area_init_nodes(max_zone_pfns); |
1647 | * | 1647 | * |
1648 | * free_bootmem_with_active_regions() calls free_bootmem_node() for each | 1648 | * free_bootmem_with_active_regions() calls free_bootmem_node() for each |
1649 | * registered physical page range. Similarly | 1649 | * registered physical page range. Similarly |
1650 | * sparse_memory_present_with_active_regions() calls memory_present() for | 1650 | * sparse_memory_present_with_active_regions() calls memory_present() for |
1651 | * each range when SPARSEMEM is enabled. | 1651 | * each range when SPARSEMEM is enabled. |
1652 | * | 1652 | * |
1653 | * See mm/page_alloc.c for more information on each function exposed by | 1653 | * See mm/page_alloc.c for more information on each function exposed by |
1654 | * CONFIG_HAVE_MEMBLOCK_NODE_MAP. | 1654 | * CONFIG_HAVE_MEMBLOCK_NODE_MAP. |
1655 | */ | 1655 | */ |
1656 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | 1656 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); |
1657 | unsigned long node_map_pfn_alignment(void); | 1657 | unsigned long node_map_pfn_alignment(void); |
1658 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, | 1658 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, |
1659 | unsigned long end_pfn); | 1659 | unsigned long end_pfn); |
1660 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, | 1660 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
1661 | unsigned long end_pfn); | 1661 | unsigned long end_pfn); |
1662 | extern void get_pfn_range_for_nid(unsigned int nid, | 1662 | extern void get_pfn_range_for_nid(unsigned int nid, |
1663 | unsigned long *start_pfn, unsigned long *end_pfn); | 1663 | unsigned long *start_pfn, unsigned long *end_pfn); |
1664 | extern unsigned long find_min_pfn_with_active_regions(void); | 1664 | extern unsigned long find_min_pfn_with_active_regions(void); |
1665 | extern void free_bootmem_with_active_regions(int nid, | 1665 | extern void free_bootmem_with_active_regions(int nid, |
1666 | unsigned long max_low_pfn); | 1666 | unsigned long max_low_pfn); |
1667 | extern void sparse_memory_present_with_active_regions(int nid); | 1667 | extern void sparse_memory_present_with_active_regions(int nid); |
1668 | 1668 | ||
1669 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | 1669 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ |
1670 | 1670 | ||
1671 | #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ | 1671 | #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ |
1672 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) | 1672 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) |
1673 | static inline int __early_pfn_to_nid(unsigned long pfn) | 1673 | static inline int __early_pfn_to_nid(unsigned long pfn) |
1674 | { | 1674 | { |
1675 | return 0; | 1675 | return 0; |
1676 | } | 1676 | } |
1677 | #else | 1677 | #else |
1678 | /* please see mm/page_alloc.c */ | 1678 | /* please see mm/page_alloc.c */ |
1679 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | 1679 | extern int __meminit early_pfn_to_nid(unsigned long pfn); |
1680 | /* there is a per-arch backend function. */ | 1680 | /* there is a per-arch backend function. */ |
1681 | extern int __meminit __early_pfn_to_nid(unsigned long pfn); | 1681 | extern int __meminit __early_pfn_to_nid(unsigned long pfn); |
1682 | #endif | 1682 | #endif |
1683 | 1683 | ||
1684 | extern void set_dma_reserve(unsigned long new_dma_reserve); | 1684 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
1685 | extern void memmap_init_zone(unsigned long, int, unsigned long, | 1685 | extern void memmap_init_zone(unsigned long, int, unsigned long, |
1686 | unsigned long, enum memmap_context); | 1686 | unsigned long, enum memmap_context); |
1687 | extern void setup_per_zone_wmarks(void); | 1687 | extern void setup_per_zone_wmarks(void); |
1688 | extern int __meminit init_per_zone_wmark_min(void); | 1688 | extern int __meminit init_per_zone_wmark_min(void); |
1689 | extern void mem_init(void); | 1689 | extern void mem_init(void); |
1690 | extern void __init mmap_init(void); | 1690 | extern void __init mmap_init(void); |
1691 | extern void show_mem(unsigned int flags); | 1691 | extern void show_mem(unsigned int flags); |
1692 | extern void si_meminfo(struct sysinfo * val); | 1692 | extern void si_meminfo(struct sysinfo * val); |
1693 | extern void si_meminfo_node(struct sysinfo *val, int nid); | 1693 | extern void si_meminfo_node(struct sysinfo *val, int nid); |
1694 | 1694 | ||
1695 | extern __printf(3, 4) | 1695 | extern __printf(3, 4) |
1696 | void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...); | 1696 | void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...); |
1697 | 1697 | ||
1698 | extern void setup_per_cpu_pageset(void); | 1698 | extern void setup_per_cpu_pageset(void); |
1699 | 1699 | ||
1700 | extern void zone_pcp_update(struct zone *zone); | 1700 | extern void zone_pcp_update(struct zone *zone); |
1701 | extern void zone_pcp_reset(struct zone *zone); | 1701 | extern void zone_pcp_reset(struct zone *zone); |
1702 | 1702 | ||
1703 | /* page_alloc.c */ | 1703 | /* page_alloc.c */ |
1704 | extern int min_free_kbytes; | 1704 | extern int min_free_kbytes; |
1705 | 1705 | ||
1706 | /* nommu.c */ | 1706 | /* nommu.c */ |
1707 | extern atomic_long_t mmap_pages_allocated; | 1707 | extern atomic_long_t mmap_pages_allocated; |
1708 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); | 1708 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
1709 | 1709 | ||
1710 | /* interval_tree.c */ | 1710 | /* interval_tree.c */ |
1711 | void vma_interval_tree_insert(struct vm_area_struct *node, | 1711 | void vma_interval_tree_insert(struct vm_area_struct *node, |
1712 | struct rb_root *root); | 1712 | struct rb_root *root); |
1713 | void vma_interval_tree_insert_after(struct vm_area_struct *node, | 1713 | void vma_interval_tree_insert_after(struct vm_area_struct *node, |
1714 | struct vm_area_struct *prev, | 1714 | struct vm_area_struct *prev, |
1715 | struct rb_root *root); | 1715 | struct rb_root *root); |
1716 | void vma_interval_tree_remove(struct vm_area_struct *node, | 1716 | void vma_interval_tree_remove(struct vm_area_struct *node, |
1717 | struct rb_root *root); | 1717 | struct rb_root *root); |
1718 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root, | 1718 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root, |
1719 | unsigned long start, unsigned long last); | 1719 | unsigned long start, unsigned long last); |
1720 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, | 1720 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, |
1721 | unsigned long start, unsigned long last); | 1721 | unsigned long start, unsigned long last); |
1722 | 1722 | ||
1723 | #define vma_interval_tree_foreach(vma, root, start, last) \ | 1723 | #define vma_interval_tree_foreach(vma, root, start, last) \ |
1724 | for (vma = vma_interval_tree_iter_first(root, start, last); \ | 1724 | for (vma = vma_interval_tree_iter_first(root, start, last); \ |
1725 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) | 1725 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) |
1726 | 1726 | ||
1727 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, | 1727 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, |
1728 | struct list_head *list) | 1728 | struct list_head *list) |
1729 | { | 1729 | { |
1730 | list_add_tail(&vma->shared.nonlinear, list); | 1730 | list_add_tail(&vma->shared.nonlinear, list); |
1731 | } | 1731 | } |
1732 | 1732 | ||
1733 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, | 1733 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, |
1734 | struct rb_root *root); | 1734 | struct rb_root *root); |
1735 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, | 1735 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, |
1736 | struct rb_root *root); | 1736 | struct rb_root *root); |
1737 | struct anon_vma_chain *anon_vma_interval_tree_iter_first( | 1737 | struct anon_vma_chain *anon_vma_interval_tree_iter_first( |
1738 | struct rb_root *root, unsigned long start, unsigned long last); | 1738 | struct rb_root *root, unsigned long start, unsigned long last); |
1739 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( | 1739 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( |
1740 | struct anon_vma_chain *node, unsigned long start, unsigned long last); | 1740 | struct anon_vma_chain *node, unsigned long start, unsigned long last); |
1741 | #ifdef CONFIG_DEBUG_VM_RB | 1741 | #ifdef CONFIG_DEBUG_VM_RB |
1742 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); | 1742 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); |
1743 | #endif | 1743 | #endif |
1744 | 1744 | ||
1745 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ | 1745 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ |
1746 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ | 1746 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ |
1747 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) | 1747 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) |
1748 | 1748 | ||
1749 | /* mmap.c */ | 1749 | /* mmap.c */ |
1750 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); | 1750 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
1751 | extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, | 1751 | extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, |
1752 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); | 1752 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); |
1753 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | 1753 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
1754 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | 1754 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, |
1755 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | 1755 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, |
1756 | struct mempolicy *); | 1756 | struct mempolicy *); |
1757 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | 1757 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
1758 | extern int split_vma(struct mm_struct *, | 1758 | extern int split_vma(struct mm_struct *, |
1759 | struct vm_area_struct *, unsigned long addr, int new_below); | 1759 | struct vm_area_struct *, unsigned long addr, int new_below); |
1760 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | 1760 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
1761 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | 1761 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, |
1762 | struct rb_node **, struct rb_node *); | 1762 | struct rb_node **, struct rb_node *); |
1763 | extern void unlink_file_vma(struct vm_area_struct *); | 1763 | extern void unlink_file_vma(struct vm_area_struct *); |
1764 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, | 1764 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
1765 | unsigned long addr, unsigned long len, pgoff_t pgoff, | 1765 | unsigned long addr, unsigned long len, pgoff_t pgoff, |
1766 | bool *need_rmap_locks); | 1766 | bool *need_rmap_locks); |
1767 | extern void exit_mmap(struct mm_struct *); | 1767 | extern void exit_mmap(struct mm_struct *); |
1768 | 1768 | ||
1769 | extern int mm_take_all_locks(struct mm_struct *mm); | 1769 | extern int mm_take_all_locks(struct mm_struct *mm); |
1770 | extern void mm_drop_all_locks(struct mm_struct *mm); | 1770 | extern void mm_drop_all_locks(struct mm_struct *mm); |
1771 | 1771 | ||
1772 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); | 1772 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); |
1773 | extern struct file *get_mm_exe_file(struct mm_struct *mm); | 1773 | extern struct file *get_mm_exe_file(struct mm_struct *mm); |
1774 | 1774 | ||
1775 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); | 1775 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
1776 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, | 1776 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, |
1777 | unsigned long addr, unsigned long len, | 1777 | unsigned long addr, unsigned long len, |
1778 | unsigned long flags, struct page **pages); | 1778 | unsigned long flags, struct page **pages); |
1779 | extern int install_special_mapping(struct mm_struct *mm, | 1779 | extern int install_special_mapping(struct mm_struct *mm, |
1780 | unsigned long addr, unsigned long len, | 1780 | unsigned long addr, unsigned long len, |
1781 | unsigned long flags, struct page **pages); | 1781 | unsigned long flags, struct page **pages); |
1782 | 1782 | ||
1783 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | 1783 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
1784 | 1784 | ||
1785 | extern unsigned long mmap_region(struct file *file, unsigned long addr, | 1785 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
1786 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff); | 1786 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff); |
1787 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, | 1787 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, |
1788 | unsigned long len, unsigned long prot, unsigned long flags, | 1788 | unsigned long len, unsigned long prot, unsigned long flags, |
1789 | unsigned long pgoff, unsigned long *populate); | 1789 | unsigned long pgoff, unsigned long *populate); |
1790 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | 1790 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); |
1791 | 1791 | ||
1792 | #ifdef CONFIG_MMU | 1792 | #ifdef CONFIG_MMU |
1793 | extern int __mm_populate(unsigned long addr, unsigned long len, | 1793 | extern int __mm_populate(unsigned long addr, unsigned long len, |
1794 | int ignore_errors); | 1794 | int ignore_errors); |
1795 | static inline void mm_populate(unsigned long addr, unsigned long len) | 1795 | static inline void mm_populate(unsigned long addr, unsigned long len) |
1796 | { | 1796 | { |
1797 | /* Ignore errors */ | 1797 | /* Ignore errors */ |
1798 | (void) __mm_populate(addr, len, 1); | 1798 | (void) __mm_populate(addr, len, 1); |
1799 | } | 1799 | } |
1800 | #else | 1800 | #else |
1801 | static inline void mm_populate(unsigned long addr, unsigned long len) {} | 1801 | static inline void mm_populate(unsigned long addr, unsigned long len) {} |
1802 | #endif | 1802 | #endif |
1803 | 1803 | ||
1804 | /* These take the mm semaphore themselves */ | 1804 | /* These take the mm semaphore themselves */ |
1805 | extern unsigned long vm_brk(unsigned long, unsigned long); | 1805 | extern unsigned long vm_brk(unsigned long, unsigned long); |
1806 | extern int vm_munmap(unsigned long, size_t); | 1806 | extern int vm_munmap(unsigned long, size_t); |
1807 | extern unsigned long vm_mmap(struct file *, unsigned long, | 1807 | extern unsigned long vm_mmap(struct file *, unsigned long, |
1808 | unsigned long, unsigned long, | 1808 | unsigned long, unsigned long, |
1809 | unsigned long, unsigned long); | 1809 | unsigned long, unsigned long); |
1810 | 1810 | ||
1811 | struct vm_unmapped_area_info { | 1811 | struct vm_unmapped_area_info { |
1812 | #define VM_UNMAPPED_AREA_TOPDOWN 1 | 1812 | #define VM_UNMAPPED_AREA_TOPDOWN 1 |
1813 | unsigned long flags; | 1813 | unsigned long flags; |
1814 | unsigned long length; | 1814 | unsigned long length; |
1815 | unsigned long low_limit; | 1815 | unsigned long low_limit; |
1816 | unsigned long high_limit; | 1816 | unsigned long high_limit; |
1817 | unsigned long align_mask; | 1817 | unsigned long align_mask; |
1818 | unsigned long align_offset; | 1818 | unsigned long align_offset; |
1819 | }; | 1819 | }; |
1820 | 1820 | ||
1821 | extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); | 1821 | extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); |
1822 | extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); | 1822 | extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); |
1823 | 1823 | ||
1824 | /* | 1824 | /* |
1825 | * Search for an unmapped address range. | 1825 | * Search for an unmapped address range. |
1826 | * | 1826 | * |
1827 | * We are looking for a range that: | 1827 | * We are looking for a range that: |
1828 | * - does not intersect with any VMA; | 1828 | * - does not intersect with any VMA; |
1829 | * - is contained within the [low_limit, high_limit) interval; | 1829 | * - is contained within the [low_limit, high_limit) interval; |
1830 | * - is at least the desired size. | 1830 | * - is at least the desired size. |
1831 | * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) | 1831 | * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) |
1832 | */ | 1832 | */ |
1833 | static inline unsigned long | 1833 | static inline unsigned long |
1834 | vm_unmapped_area(struct vm_unmapped_area_info *info) | 1834 | vm_unmapped_area(struct vm_unmapped_area_info *info) |
1835 | { | 1835 | { |
1836 | if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN)) | 1836 | if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN)) |
1837 | return unmapped_area(info); | 1837 | return unmapped_area(info); |
1838 | else | 1838 | else |
1839 | return unmapped_area_topdown(info); | 1839 | return unmapped_area_topdown(info); |
1840 | } | 1840 | } |
1841 | 1841 | ||
1842 | /* truncate.c */ | 1842 | /* truncate.c */ |
1843 | extern void truncate_inode_pages(struct address_space *, loff_t); | 1843 | extern void truncate_inode_pages(struct address_space *, loff_t); |
1844 | extern void truncate_inode_pages_range(struct address_space *, | 1844 | extern void truncate_inode_pages_range(struct address_space *, |
1845 | loff_t lstart, loff_t lend); | 1845 | loff_t lstart, loff_t lend); |
1846 | extern void truncate_inode_pages_final(struct address_space *); | 1846 | extern void truncate_inode_pages_final(struct address_space *); |
1847 | 1847 | ||
1848 | /* generic vm_area_ops exported for stackable file systems */ | 1848 | /* generic vm_area_ops exported for stackable file systems */ |
1849 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); | 1849 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); |
1850 | extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf); | 1850 | extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf); |
1851 | extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); | 1851 | extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); |
1852 | 1852 | ||
1853 | /* mm/page-writeback.c */ | 1853 | /* mm/page-writeback.c */ |
1854 | int write_one_page(struct page *page, int wait); | 1854 | int write_one_page(struct page *page, int wait); |
1855 | void task_dirty_inc(struct task_struct *tsk); | 1855 | void task_dirty_inc(struct task_struct *tsk); |
1856 | 1856 | ||
1857 | /* readahead.c */ | 1857 | /* readahead.c */ |
1858 | #define VM_MAX_READAHEAD 128 /* kbytes */ | 1858 | #define VM_MAX_READAHEAD 128 /* kbytes */ |
1859 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | 1859 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ |
1860 | 1860 | ||
1861 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, | 1861 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
1862 | pgoff_t offset, unsigned long nr_to_read); | 1862 | pgoff_t offset, unsigned long nr_to_read); |
1863 | 1863 | ||
1864 | void page_cache_sync_readahead(struct address_space *mapping, | 1864 | void page_cache_sync_readahead(struct address_space *mapping, |
1865 | struct file_ra_state *ra, | 1865 | struct file_ra_state *ra, |
1866 | struct file *filp, | 1866 | struct file *filp, |
1867 | pgoff_t offset, | 1867 | pgoff_t offset, |
1868 | unsigned long size); | 1868 | unsigned long size); |
1869 | 1869 | ||
1870 | void page_cache_async_readahead(struct address_space *mapping, | 1870 | void page_cache_async_readahead(struct address_space *mapping, |
1871 | struct file_ra_state *ra, | 1871 | struct file_ra_state *ra, |
1872 | struct file *filp, | 1872 | struct file *filp, |
1873 | struct page *pg, | 1873 | struct page *pg, |
1874 | pgoff_t offset, | 1874 | pgoff_t offset, |
1875 | unsigned long size); | 1875 | unsigned long size); |
1876 | 1876 | ||
1877 | unsigned long max_sane_readahead(unsigned long nr); | 1877 | unsigned long max_sane_readahead(unsigned long nr); |
1878 | unsigned long ra_submit(struct file_ra_state *ra, | ||
1879 | struct address_space *mapping, | ||
1880 | struct file *filp); | ||
1881 | 1878 | ||
1882 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ | 1879 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ |
1883 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); | 1880 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
1884 | 1881 | ||
1885 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ | 1882 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ |
1886 | extern int expand_downwards(struct vm_area_struct *vma, | 1883 | extern int expand_downwards(struct vm_area_struct *vma, |
1887 | unsigned long address); | 1884 | unsigned long address); |
1888 | #if VM_GROWSUP | 1885 | #if VM_GROWSUP |
1889 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); | 1886 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
1890 | #else | 1887 | #else |
1891 | #define expand_upwards(vma, address) do { } while (0) | 1888 | #define expand_upwards(vma, address) do { } while (0) |
1892 | #endif | 1889 | #endif |
1893 | 1890 | ||
1894 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | 1891 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ |
1895 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | 1892 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); |
1896 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | 1893 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, |
1897 | struct vm_area_struct **pprev); | 1894 | struct vm_area_struct **pprev); |
1898 | 1895 | ||
1899 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | 1896 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, |
1900 | NULL if none. Assume start_addr < end_addr. */ | 1897 | NULL if none. Assume start_addr < end_addr. */ |
1901 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | 1898 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) |
1902 | { | 1899 | { |
1903 | struct vm_area_struct * vma = find_vma(mm,start_addr); | 1900 | struct vm_area_struct * vma = find_vma(mm,start_addr); |
1904 | 1901 | ||
1905 | if (vma && end_addr <= vma->vm_start) | 1902 | if (vma && end_addr <= vma->vm_start) |
1906 | vma = NULL; | 1903 | vma = NULL; |
1907 | return vma; | 1904 | return vma; |
1908 | } | 1905 | } |
1909 | 1906 | ||
1910 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | 1907 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
1911 | { | 1908 | { |
1912 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | 1909 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
1913 | } | 1910 | } |
1914 | 1911 | ||
1915 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ | 1912 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ |
1916 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, | 1913 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, |
1917 | unsigned long vm_start, unsigned long vm_end) | 1914 | unsigned long vm_start, unsigned long vm_end) |
1918 | { | 1915 | { |
1919 | struct vm_area_struct *vma = find_vma(mm, vm_start); | 1916 | struct vm_area_struct *vma = find_vma(mm, vm_start); |
1920 | 1917 | ||
1921 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) | 1918 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) |
1922 | vma = NULL; | 1919 | vma = NULL; |
1923 | 1920 | ||
1924 | return vma; | 1921 | return vma; |
1925 | } | 1922 | } |
1926 | 1923 | ||
1927 | #ifdef CONFIG_MMU | 1924 | #ifdef CONFIG_MMU |
1928 | pgprot_t vm_get_page_prot(unsigned long vm_flags); | 1925 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
1929 | #else | 1926 | #else |
1930 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | 1927 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) |
1931 | { | 1928 | { |
1932 | return __pgprot(0); | 1929 | return __pgprot(0); |
1933 | } | 1930 | } |
1934 | #endif | 1931 | #endif |
1935 | 1932 | ||
1936 | #ifdef CONFIG_NUMA_BALANCING | 1933 | #ifdef CONFIG_NUMA_BALANCING |
1937 | unsigned long change_prot_numa(struct vm_area_struct *vma, | 1934 | unsigned long change_prot_numa(struct vm_area_struct *vma, |
1938 | unsigned long start, unsigned long end); | 1935 | unsigned long start, unsigned long end); |
1939 | #endif | 1936 | #endif |
1940 | 1937 | ||
1941 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); | 1938 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
1942 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, | 1939 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
1943 | unsigned long pfn, unsigned long size, pgprot_t); | 1940 | unsigned long pfn, unsigned long size, pgprot_t); |
1944 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); | 1941 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
1945 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, | 1942 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
1946 | unsigned long pfn); | 1943 | unsigned long pfn); |
1947 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, | 1944 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
1948 | unsigned long pfn); | 1945 | unsigned long pfn); |
1949 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); | 1946 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); |
1950 | 1947 | ||
1951 | 1948 | ||
1952 | struct page *follow_page_mask(struct vm_area_struct *vma, | 1949 | struct page *follow_page_mask(struct vm_area_struct *vma, |
1953 | unsigned long address, unsigned int foll_flags, | 1950 | unsigned long address, unsigned int foll_flags, |
1954 | unsigned int *page_mask); | 1951 | unsigned int *page_mask); |
1955 | 1952 | ||
1956 | static inline struct page *follow_page(struct vm_area_struct *vma, | 1953 | static inline struct page *follow_page(struct vm_area_struct *vma, |
1957 | unsigned long address, unsigned int foll_flags) | 1954 | unsigned long address, unsigned int foll_flags) |
1958 | { | 1955 | { |
1959 | unsigned int unused_page_mask; | 1956 | unsigned int unused_page_mask; |
1960 | return follow_page_mask(vma, address, foll_flags, &unused_page_mask); | 1957 | return follow_page_mask(vma, address, foll_flags, &unused_page_mask); |
1961 | } | 1958 | } |
1962 | 1959 | ||
1963 | #define FOLL_WRITE 0x01 /* check pte is writable */ | 1960 | #define FOLL_WRITE 0x01 /* check pte is writable */ |
1964 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | 1961 | #define FOLL_TOUCH 0x02 /* mark page accessed */ |
1965 | #define FOLL_GET 0x04 /* do get_page on page */ | 1962 | #define FOLL_GET 0x04 /* do get_page on page */ |
1966 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ | 1963 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ |
1967 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ | 1964 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ |
1968 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO | 1965 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO |
1969 | * and return without waiting upon it */ | 1966 | * and return without waiting upon it */ |
1970 | #define FOLL_MLOCK 0x40 /* mark page as mlocked */ | 1967 | #define FOLL_MLOCK 0x40 /* mark page as mlocked */ |
1971 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ | 1968 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ |
1972 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ | 1969 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ |
1973 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ | 1970 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ |
1974 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ | 1971 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ |
1975 | 1972 | ||
1976 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, | 1973 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, |
1977 | void *data); | 1974 | void *data); |
1978 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | 1975 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, |
1979 | unsigned long size, pte_fn_t fn, void *data); | 1976 | unsigned long size, pte_fn_t fn, void *data); |
1980 | 1977 | ||
1981 | #ifdef CONFIG_PROC_FS | 1978 | #ifdef CONFIG_PROC_FS |
1982 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); | 1979 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); |
1983 | #else | 1980 | #else |
1984 | static inline void vm_stat_account(struct mm_struct *mm, | 1981 | static inline void vm_stat_account(struct mm_struct *mm, |
1985 | unsigned long flags, struct file *file, long pages) | 1982 | unsigned long flags, struct file *file, long pages) |
1986 | { | 1983 | { |
1987 | mm->total_vm += pages; | 1984 | mm->total_vm += pages; |
1988 | } | 1985 | } |
1989 | #endif /* CONFIG_PROC_FS */ | 1986 | #endif /* CONFIG_PROC_FS */ |
1990 | 1987 | ||
1991 | #ifdef CONFIG_DEBUG_PAGEALLOC | 1988 | #ifdef CONFIG_DEBUG_PAGEALLOC |
1992 | extern void kernel_map_pages(struct page *page, int numpages, int enable); | 1989 | extern void kernel_map_pages(struct page *page, int numpages, int enable); |
1993 | #ifdef CONFIG_HIBERNATION | 1990 | #ifdef CONFIG_HIBERNATION |
1994 | extern bool kernel_page_present(struct page *page); | 1991 | extern bool kernel_page_present(struct page *page); |
1995 | #endif /* CONFIG_HIBERNATION */ | 1992 | #endif /* CONFIG_HIBERNATION */ |
1996 | #else | 1993 | #else |
1997 | static inline void | 1994 | static inline void |
1998 | kernel_map_pages(struct page *page, int numpages, int enable) {} | 1995 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
1999 | #ifdef CONFIG_HIBERNATION | 1996 | #ifdef CONFIG_HIBERNATION |
2000 | static inline bool kernel_page_present(struct page *page) { return true; } | 1997 | static inline bool kernel_page_present(struct page *page) { return true; } |
2001 | #endif /* CONFIG_HIBERNATION */ | 1998 | #endif /* CONFIG_HIBERNATION */ |
2002 | #endif | 1999 | #endif |
2003 | 2000 | ||
2004 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); | 2001 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
2005 | #ifdef __HAVE_ARCH_GATE_AREA | 2002 | #ifdef __HAVE_ARCH_GATE_AREA |
2006 | int in_gate_area_no_mm(unsigned long addr); | 2003 | int in_gate_area_no_mm(unsigned long addr); |
2007 | int in_gate_area(struct mm_struct *mm, unsigned long addr); | 2004 | int in_gate_area(struct mm_struct *mm, unsigned long addr); |
2008 | #else | 2005 | #else |
2009 | int in_gate_area_no_mm(unsigned long addr); | 2006 | int in_gate_area_no_mm(unsigned long addr); |
2010 | #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);}) | 2007 | #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);}) |
2011 | #endif /* __HAVE_ARCH_GATE_AREA */ | 2008 | #endif /* __HAVE_ARCH_GATE_AREA */ |
2012 | 2009 | ||
2013 | #ifdef CONFIG_SYSCTL | 2010 | #ifdef CONFIG_SYSCTL |
2014 | extern int sysctl_drop_caches; | 2011 | extern int sysctl_drop_caches; |
2015 | int drop_caches_sysctl_handler(struct ctl_table *, int, | 2012 | int drop_caches_sysctl_handler(struct ctl_table *, int, |
2016 | void __user *, size_t *, loff_t *); | 2013 | void __user *, size_t *, loff_t *); |
2017 | #endif | 2014 | #endif |
2018 | 2015 | ||
2019 | unsigned long shrink_slab(struct shrink_control *shrink, | 2016 | unsigned long shrink_slab(struct shrink_control *shrink, |
2020 | unsigned long nr_pages_scanned, | 2017 | unsigned long nr_pages_scanned, |
2021 | unsigned long lru_pages); | 2018 | unsigned long lru_pages); |
2022 | 2019 | ||
2023 | #ifndef CONFIG_MMU | 2020 | #ifndef CONFIG_MMU |
2024 | #define randomize_va_space 0 | 2021 | #define randomize_va_space 0 |
2025 | #else | 2022 | #else |
2026 | extern int randomize_va_space; | 2023 | extern int randomize_va_space; |
2027 | #endif | 2024 | #endif |
2028 | 2025 | ||
2029 | const char * arch_vma_name(struct vm_area_struct *vma); | 2026 | const char * arch_vma_name(struct vm_area_struct *vma); |
2030 | void print_vma_addr(char *prefix, unsigned long rip); | 2027 | void print_vma_addr(char *prefix, unsigned long rip); |
2031 | 2028 | ||
2032 | void sparse_mem_maps_populate_node(struct page **map_map, | 2029 | void sparse_mem_maps_populate_node(struct page **map_map, |
2033 | unsigned long pnum_begin, | 2030 | unsigned long pnum_begin, |
2034 | unsigned long pnum_end, | 2031 | unsigned long pnum_end, |
2035 | unsigned long map_count, | 2032 | unsigned long map_count, |
2036 | int nodeid); | 2033 | int nodeid); |
2037 | 2034 | ||
2038 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); | 2035 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); |
2039 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); | 2036 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
2040 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); | 2037 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); |
2041 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); | 2038 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); |
2042 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); | 2039 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); |
2043 | void *vmemmap_alloc_block(unsigned long size, int node); | 2040 | void *vmemmap_alloc_block(unsigned long size, int node); |
2044 | void *vmemmap_alloc_block_buf(unsigned long size, int node); | 2041 | void *vmemmap_alloc_block_buf(unsigned long size, int node); |
2045 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); | 2042 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
2046 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, | 2043 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, |
2047 | int node); | 2044 | int node); |
2048 | int vmemmap_populate(unsigned long start, unsigned long end, int node); | 2045 | int vmemmap_populate(unsigned long start, unsigned long end, int node); |
2049 | void vmemmap_populate_print_last(void); | 2046 | void vmemmap_populate_print_last(void); |
2050 | #ifdef CONFIG_MEMORY_HOTPLUG | 2047 | #ifdef CONFIG_MEMORY_HOTPLUG |
2051 | void vmemmap_free(unsigned long start, unsigned long end); | 2048 | void vmemmap_free(unsigned long start, unsigned long end); |
2052 | #endif | 2049 | #endif |
2053 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, | 2050 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, |
2054 | unsigned long size); | 2051 | unsigned long size); |
2055 | 2052 | ||
2056 | enum mf_flags { | 2053 | enum mf_flags { |
2057 | MF_COUNT_INCREASED = 1 << 0, | 2054 | MF_COUNT_INCREASED = 1 << 0, |
2058 | MF_ACTION_REQUIRED = 1 << 1, | 2055 | MF_ACTION_REQUIRED = 1 << 1, |
2059 | MF_MUST_KILL = 1 << 2, | 2056 | MF_MUST_KILL = 1 << 2, |
2060 | MF_SOFT_OFFLINE = 1 << 3, | 2057 | MF_SOFT_OFFLINE = 1 << 3, |
2061 | }; | 2058 | }; |
2062 | extern int memory_failure(unsigned long pfn, int trapno, int flags); | 2059 | extern int memory_failure(unsigned long pfn, int trapno, int flags); |
2063 | extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); | 2060 | extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); |
2064 | extern int unpoison_memory(unsigned long pfn); | 2061 | extern int unpoison_memory(unsigned long pfn); |
2065 | extern int sysctl_memory_failure_early_kill; | 2062 | extern int sysctl_memory_failure_early_kill; |
2066 | extern int sysctl_memory_failure_recovery; | 2063 | extern int sysctl_memory_failure_recovery; |
2067 | extern void shake_page(struct page *p, int access); | 2064 | extern void shake_page(struct page *p, int access); |
2068 | extern atomic_long_t num_poisoned_pages; | 2065 | extern atomic_long_t num_poisoned_pages; |
2069 | extern int soft_offline_page(struct page *page, int flags); | 2066 | extern int soft_offline_page(struct page *page, int flags); |
2070 | 2067 | ||
2071 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) | 2068 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
2072 | extern void clear_huge_page(struct page *page, | 2069 | extern void clear_huge_page(struct page *page, |
2073 | unsigned long addr, | 2070 | unsigned long addr, |
2074 | unsigned int pages_per_huge_page); | 2071 | unsigned int pages_per_huge_page); |
2075 | extern void copy_user_huge_page(struct page *dst, struct page *src, | 2072 | extern void copy_user_huge_page(struct page *dst, struct page *src, |
2076 | unsigned long addr, struct vm_area_struct *vma, | 2073 | unsigned long addr, struct vm_area_struct *vma, |
2077 | unsigned int pages_per_huge_page); | 2074 | unsigned int pages_per_huge_page); |
2078 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ | 2075 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ |
2079 | 2076 | ||
2080 | #ifdef CONFIG_DEBUG_PAGEALLOC | 2077 | #ifdef CONFIG_DEBUG_PAGEALLOC |
2081 | extern unsigned int _debug_guardpage_minorder; | 2078 | extern unsigned int _debug_guardpage_minorder; |
2082 | 2079 | ||
2083 | static inline unsigned int debug_guardpage_minorder(void) | 2080 | static inline unsigned int debug_guardpage_minorder(void) |
2084 | { | 2081 | { |
2085 | return _debug_guardpage_minorder; | 2082 | return _debug_guardpage_minorder; |
2086 | } | 2083 | } |
2087 | 2084 | ||
2088 | static inline bool page_is_guard(struct page *page) | 2085 | static inline bool page_is_guard(struct page *page) |
2089 | { | 2086 | { |
2090 | return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags); | 2087 | return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags); |
2091 | } | 2088 | } |
2092 | #else | 2089 | #else |
2093 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } | 2090 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } |
2094 | static inline bool page_is_guard(struct page *page) { return false; } | 2091 | static inline bool page_is_guard(struct page *page) { return false; } |
2095 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | 2092 | #endif /* CONFIG_DEBUG_PAGEALLOC */ |
2096 | 2093 | ||
2097 | #if MAX_NUMNODES > 1 | 2094 | #if MAX_NUMNODES > 1 |
2098 | void __init setup_nr_node_ids(void); | 2095 | void __init setup_nr_node_ids(void); |
2099 | #else | 2096 | #else |
2100 | static inline void setup_nr_node_ids(void) {} | 2097 | static inline void setup_nr_node_ids(void) {} |
2101 | #endif | 2098 | #endif |
2102 | 2099 | ||
2103 | #endif /* __KERNEL__ */ | 2100 | #endif /* __KERNEL__ */ |
2104 | #endif /* _LINUX_MM_H */ | 2101 | #endif /* _LINUX_MM_H */ |
2105 | 2102 |
mm/internal.h
1 | /* internal.h: mm/ internal definitions | 1 | /* internal.h: mm/ internal definitions |
2 | * | 2 | * |
3 | * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. | 3 | * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. |
4 | * Written by David Howells (dhowells@redhat.com) | 4 | * Written by David Howells (dhowells@redhat.com) |
5 | * | 5 | * |
6 | * This program is free software; you can redistribute it and/or | 6 | * This program is free software; you can redistribute it and/or |
7 | * modify it under the terms of the GNU General Public License | 7 | * modify it under the terms of the GNU General Public License |
8 | * as published by the Free Software Foundation; either version | 8 | * as published by the Free Software Foundation; either version |
9 | * 2 of the License, or (at your option) any later version. | 9 | * 2 of the License, or (at your option) any later version. |
10 | */ | 10 | */ |
11 | #ifndef __MM_INTERNAL_H | 11 | #ifndef __MM_INTERNAL_H |
12 | #define __MM_INTERNAL_H | 12 | #define __MM_INTERNAL_H |
13 | 13 | ||
14 | #include <linux/fs.h> | ||
14 | #include <linux/mm.h> | 15 | #include <linux/mm.h> |
15 | 16 | ||
16 | void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, | 17 | void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
17 | unsigned long floor, unsigned long ceiling); | 18 | unsigned long floor, unsigned long ceiling); |
18 | 19 | ||
19 | static inline void set_page_count(struct page *page, int v) | 20 | static inline void set_page_count(struct page *page, int v) |
20 | { | 21 | { |
21 | atomic_set(&page->_count, v); | 22 | atomic_set(&page->_count, v); |
23 | } | ||
24 | |||
25 | extern int __do_page_cache_readahead(struct address_space *mapping, | ||
26 | struct file *filp, pgoff_t offset, unsigned long nr_to_read, | ||
27 | unsigned long lookahead_size); | ||
28 | |||
29 | /* | ||
30 | * Submit IO for the read-ahead request in file_ra_state. | ||
31 | */ | ||
32 | static inline unsigned long ra_submit(struct file_ra_state *ra, | ||
33 | struct address_space *mapping, struct file *filp) | ||
34 | { | ||
35 | return __do_page_cache_readahead(mapping, filp, | ||
36 | ra->start, ra->size, ra->async_size); | ||
22 | } | 37 | } |
23 | 38 | ||
24 | /* | 39 | /* |
25 | * Turn a non-refcounted page (->_count == 0) into refcounted with | 40 | * Turn a non-refcounted page (->_count == 0) into refcounted with |
26 | * a count of one. | 41 | * a count of one. |
27 | */ | 42 | */ |
28 | static inline void set_page_refcounted(struct page *page) | 43 | static inline void set_page_refcounted(struct page *page) |
29 | { | 44 | { |
30 | VM_BUG_ON_PAGE(PageTail(page), page); | 45 | VM_BUG_ON_PAGE(PageTail(page), page); |
31 | VM_BUG_ON_PAGE(atomic_read(&page->_count), page); | 46 | VM_BUG_ON_PAGE(atomic_read(&page->_count), page); |
32 | set_page_count(page, 1); | 47 | set_page_count(page, 1); |
33 | } | 48 | } |
34 | 49 | ||
35 | static inline void __get_page_tail_foll(struct page *page, | 50 | static inline void __get_page_tail_foll(struct page *page, |
36 | bool get_page_head) | 51 | bool get_page_head) |
37 | { | 52 | { |
38 | /* | 53 | /* |
39 | * If we're getting a tail page, the elevated page->_count is | 54 | * If we're getting a tail page, the elevated page->_count is |
40 | * required only in the head page and we will elevate the head | 55 | * required only in the head page and we will elevate the head |
41 | * page->_count and tail page->_mapcount. | 56 | * page->_count and tail page->_mapcount. |
42 | * | 57 | * |
43 | * We elevate page_tail->_mapcount for tail pages to force | 58 | * We elevate page_tail->_mapcount for tail pages to force |
44 | * page_tail->_count to be zero at all times to avoid getting | 59 | * page_tail->_count to be zero at all times to avoid getting |
45 | * false positives from get_page_unless_zero() with | 60 | * false positives from get_page_unless_zero() with |
46 | * speculative page access (like in | 61 | * speculative page access (like in |
47 | * page_cache_get_speculative()) on tail pages. | 62 | * page_cache_get_speculative()) on tail pages. |
48 | */ | 63 | */ |
49 | VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page); | 64 | VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page); |
50 | if (get_page_head) | 65 | if (get_page_head) |
51 | atomic_inc(&page->first_page->_count); | 66 | atomic_inc(&page->first_page->_count); |
52 | get_huge_page_tail(page); | 67 | get_huge_page_tail(page); |
53 | } | 68 | } |
54 | 69 | ||
55 | /* | 70 | /* |
56 | * This is meant to be called as the FOLL_GET operation of | 71 | * This is meant to be called as the FOLL_GET operation of |
57 | * follow_page() and it must be called while holding the proper PT | 72 | * follow_page() and it must be called while holding the proper PT |
58 | * lock while the pte (or pmd_trans_huge) is still mapping the page. | 73 | * lock while the pte (or pmd_trans_huge) is still mapping the page. |
59 | */ | 74 | */ |
60 | static inline void get_page_foll(struct page *page) | 75 | static inline void get_page_foll(struct page *page) |
61 | { | 76 | { |
62 | if (unlikely(PageTail(page))) | 77 | if (unlikely(PageTail(page))) |
63 | /* | 78 | /* |
64 | * This is safe only because | 79 | * This is safe only because |
65 | * __split_huge_page_refcount() can't run under | 80 | * __split_huge_page_refcount() can't run under |
66 | * get_page_foll() because we hold the proper PT lock. | 81 | * get_page_foll() because we hold the proper PT lock. |
67 | */ | 82 | */ |
68 | __get_page_tail_foll(page, true); | 83 | __get_page_tail_foll(page, true); |
69 | else { | 84 | else { |
70 | /* | 85 | /* |
71 | * Getting a normal page or the head of a compound page | 86 | * Getting a normal page or the head of a compound page |
72 | * requires to already have an elevated page->_count. | 87 | * requires to already have an elevated page->_count. |
73 | */ | 88 | */ |
74 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); | 89 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); |
75 | atomic_inc(&page->_count); | 90 | atomic_inc(&page->_count); |
76 | } | 91 | } |
77 | } | 92 | } |
78 | 93 | ||
79 | extern unsigned long highest_memmap_pfn; | 94 | extern unsigned long highest_memmap_pfn; |
80 | 95 | ||
81 | /* | 96 | /* |
82 | * in mm/vmscan.c: | 97 | * in mm/vmscan.c: |
83 | */ | 98 | */ |
84 | extern int isolate_lru_page(struct page *page); | 99 | extern int isolate_lru_page(struct page *page); |
85 | extern void putback_lru_page(struct page *page); | 100 | extern void putback_lru_page(struct page *page); |
86 | extern bool zone_reclaimable(struct zone *zone); | 101 | extern bool zone_reclaimable(struct zone *zone); |
87 | 102 | ||
88 | /* | 103 | /* |
89 | * in mm/rmap.c: | 104 | * in mm/rmap.c: |
90 | */ | 105 | */ |
91 | extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); | 106 | extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); |
92 | 107 | ||
93 | /* | 108 | /* |
94 | * in mm/page_alloc.c | 109 | * in mm/page_alloc.c |
95 | */ | 110 | */ |
96 | extern void __free_pages_bootmem(struct page *page, unsigned int order); | 111 | extern void __free_pages_bootmem(struct page *page, unsigned int order); |
97 | extern void prep_compound_page(struct page *page, unsigned long order); | 112 | extern void prep_compound_page(struct page *page, unsigned long order); |
98 | #ifdef CONFIG_MEMORY_FAILURE | 113 | #ifdef CONFIG_MEMORY_FAILURE |
99 | extern bool is_free_buddy_page(struct page *page); | 114 | extern bool is_free_buddy_page(struct page *page); |
100 | #endif | 115 | #endif |
101 | extern int user_min_free_kbytes; | 116 | extern int user_min_free_kbytes; |
102 | 117 | ||
103 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA | 118 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
104 | 119 | ||
105 | /* | 120 | /* |
106 | * in mm/compaction.c | 121 | * in mm/compaction.c |
107 | */ | 122 | */ |
108 | /* | 123 | /* |
109 | * compact_control is used to track pages being migrated and the free pages | 124 | * compact_control is used to track pages being migrated and the free pages |
110 | * they are being migrated to during memory compaction. The free_pfn starts | 125 | * they are being migrated to during memory compaction. The free_pfn starts |
111 | * at the end of a zone and migrate_pfn begins at the start. Movable pages | 126 | * at the end of a zone and migrate_pfn begins at the start. Movable pages |
112 | * are moved to the end of a zone during a compaction run and the run | 127 | * are moved to the end of a zone during a compaction run and the run |
113 | * completes when free_pfn <= migrate_pfn | 128 | * completes when free_pfn <= migrate_pfn |
114 | */ | 129 | */ |
115 | struct compact_control { | 130 | struct compact_control { |
116 | struct list_head freepages; /* List of free pages to migrate to */ | 131 | struct list_head freepages; /* List of free pages to migrate to */ |
117 | struct list_head migratepages; /* List of pages being migrated */ | 132 | struct list_head migratepages; /* List of pages being migrated */ |
118 | unsigned long nr_freepages; /* Number of isolated free pages */ | 133 | unsigned long nr_freepages; /* Number of isolated free pages */ |
119 | unsigned long nr_migratepages; /* Number of pages to migrate */ | 134 | unsigned long nr_migratepages; /* Number of pages to migrate */ |
120 | unsigned long free_pfn; /* isolate_freepages search base */ | 135 | unsigned long free_pfn; /* isolate_freepages search base */ |
121 | unsigned long migrate_pfn; /* isolate_migratepages search base */ | 136 | unsigned long migrate_pfn; /* isolate_migratepages search base */ |
122 | bool sync; /* Synchronous migration */ | 137 | bool sync; /* Synchronous migration */ |
123 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ | 138 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
124 | bool finished_update_free; /* True when the zone cached pfns are | 139 | bool finished_update_free; /* True when the zone cached pfns are |
125 | * no longer being updated | 140 | * no longer being updated |
126 | */ | 141 | */ |
127 | bool finished_update_migrate; | 142 | bool finished_update_migrate; |
128 | 143 | ||
129 | int order; /* order a direct compactor needs */ | 144 | int order; /* order a direct compactor needs */ |
130 | int migratetype; /* MOVABLE, RECLAIMABLE etc */ | 145 | int migratetype; /* MOVABLE, RECLAIMABLE etc */ |
131 | struct zone *zone; | 146 | struct zone *zone; |
132 | bool contended; /* True if a lock was contended */ | 147 | bool contended; /* True if a lock was contended */ |
133 | }; | 148 | }; |
134 | 149 | ||
135 | unsigned long | 150 | unsigned long |
136 | isolate_freepages_range(struct compact_control *cc, | 151 | isolate_freepages_range(struct compact_control *cc, |
137 | unsigned long start_pfn, unsigned long end_pfn); | 152 | unsigned long start_pfn, unsigned long end_pfn); |
138 | unsigned long | 153 | unsigned long |
139 | isolate_migratepages_range(struct zone *zone, struct compact_control *cc, | 154 | isolate_migratepages_range(struct zone *zone, struct compact_control *cc, |
140 | unsigned long low_pfn, unsigned long end_pfn, bool unevictable); | 155 | unsigned long low_pfn, unsigned long end_pfn, bool unevictable); |
141 | 156 | ||
142 | #endif | 157 | #endif |
143 | 158 | ||
144 | /* | 159 | /* |
145 | * This function returns the order of a free page in the buddy system. In | 160 | * This function returns the order of a free page in the buddy system. In |
146 | * general, page_zone(page)->lock must be held by the caller to prevent the | 161 | * general, page_zone(page)->lock must be held by the caller to prevent the |
147 | * page from being allocated in parallel and returning garbage as the order. | 162 | * page from being allocated in parallel and returning garbage as the order. |
148 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the | 163 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the |
149 | * page cannot be allocated or merged in parallel. | 164 | * page cannot be allocated or merged in parallel. |
150 | */ | 165 | */ |
151 | static inline unsigned long page_order(struct page *page) | 166 | static inline unsigned long page_order(struct page *page) |
152 | { | 167 | { |
153 | /* PageBuddy() must be checked by the caller */ | 168 | /* PageBuddy() must be checked by the caller */ |
154 | return page_private(page); | 169 | return page_private(page); |
155 | } | 170 | } |
156 | 171 | ||
157 | /* mm/util.c */ | 172 | /* mm/util.c */ |
158 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, | 173 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, |
159 | struct vm_area_struct *prev, struct rb_node *rb_parent); | 174 | struct vm_area_struct *prev, struct rb_node *rb_parent); |
160 | 175 | ||
161 | #ifdef CONFIG_MMU | 176 | #ifdef CONFIG_MMU |
162 | extern long __mlock_vma_pages_range(struct vm_area_struct *vma, | 177 | extern long __mlock_vma_pages_range(struct vm_area_struct *vma, |
163 | unsigned long start, unsigned long end, int *nonblocking); | 178 | unsigned long start, unsigned long end, int *nonblocking); |
164 | extern void munlock_vma_pages_range(struct vm_area_struct *vma, | 179 | extern void munlock_vma_pages_range(struct vm_area_struct *vma, |
165 | unsigned long start, unsigned long end); | 180 | unsigned long start, unsigned long end); |
166 | static inline void munlock_vma_pages_all(struct vm_area_struct *vma) | 181 | static inline void munlock_vma_pages_all(struct vm_area_struct *vma) |
167 | { | 182 | { |
168 | munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end); | 183 | munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end); |
169 | } | 184 | } |
170 | 185 | ||
171 | /* | 186 | /* |
172 | * Called only in fault path, to determine if a new page is being | 187 | * Called only in fault path, to determine if a new page is being |
173 | * mapped into a LOCKED vma. If it is, mark page as mlocked. | 188 | * mapped into a LOCKED vma. If it is, mark page as mlocked. |
174 | */ | 189 | */ |
175 | static inline int mlocked_vma_newpage(struct vm_area_struct *vma, | 190 | static inline int mlocked_vma_newpage(struct vm_area_struct *vma, |
176 | struct page *page) | 191 | struct page *page) |
177 | { | 192 | { |
178 | VM_BUG_ON_PAGE(PageLRU(page), page); | 193 | VM_BUG_ON_PAGE(PageLRU(page), page); |
179 | 194 | ||
180 | if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED)) | 195 | if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED)) |
181 | return 0; | 196 | return 0; |
182 | 197 | ||
183 | if (!TestSetPageMlocked(page)) { | 198 | if (!TestSetPageMlocked(page)) { |
184 | mod_zone_page_state(page_zone(page), NR_MLOCK, | 199 | mod_zone_page_state(page_zone(page), NR_MLOCK, |
185 | hpage_nr_pages(page)); | 200 | hpage_nr_pages(page)); |
186 | count_vm_event(UNEVICTABLE_PGMLOCKED); | 201 | count_vm_event(UNEVICTABLE_PGMLOCKED); |
187 | } | 202 | } |
188 | return 1; | 203 | return 1; |
189 | } | 204 | } |
190 | 205 | ||
191 | /* | 206 | /* |
192 | * must be called with vma's mmap_sem held for read or write, and page locked. | 207 | * must be called with vma's mmap_sem held for read or write, and page locked. |
193 | */ | 208 | */ |
194 | extern void mlock_vma_page(struct page *page); | 209 | extern void mlock_vma_page(struct page *page); |
195 | extern unsigned int munlock_vma_page(struct page *page); | 210 | extern unsigned int munlock_vma_page(struct page *page); |
196 | 211 | ||
197 | /* | 212 | /* |
198 | * Clear the page's PageMlocked(). This can be useful in a situation where | 213 | * Clear the page's PageMlocked(). This can be useful in a situation where |
199 | * we want to unconditionally remove a page from the pagecache -- e.g., | 214 | * we want to unconditionally remove a page from the pagecache -- e.g., |
200 | * on truncation or freeing. | 215 | * on truncation or freeing. |
201 | * | 216 | * |
202 | * It is legal to call this function for any page, mlocked or not. | 217 | * It is legal to call this function for any page, mlocked or not. |
203 | * If called for a page that is still mapped by mlocked vmas, all we do | 218 | * If called for a page that is still mapped by mlocked vmas, all we do |
204 | * is revert to lazy LRU behaviour -- semantics are not broken. | 219 | * is revert to lazy LRU behaviour -- semantics are not broken. |
205 | */ | 220 | */ |
206 | extern void clear_page_mlock(struct page *page); | 221 | extern void clear_page_mlock(struct page *page); |
207 | 222 | ||
208 | /* | 223 | /* |
209 | * mlock_migrate_page - called only from migrate_page_copy() to | 224 | * mlock_migrate_page - called only from migrate_page_copy() to |
210 | * migrate the Mlocked page flag; update statistics. | 225 | * migrate the Mlocked page flag; update statistics. |
211 | */ | 226 | */ |
212 | static inline void mlock_migrate_page(struct page *newpage, struct page *page) | 227 | static inline void mlock_migrate_page(struct page *newpage, struct page *page) |
213 | { | 228 | { |
214 | if (TestClearPageMlocked(page)) { | 229 | if (TestClearPageMlocked(page)) { |
215 | unsigned long flags; | 230 | unsigned long flags; |
216 | int nr_pages = hpage_nr_pages(page); | 231 | int nr_pages = hpage_nr_pages(page); |
217 | 232 | ||
218 | local_irq_save(flags); | 233 | local_irq_save(flags); |
219 | __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages); | 234 | __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages); |
220 | SetPageMlocked(newpage); | 235 | SetPageMlocked(newpage); |
221 | __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages); | 236 | __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages); |
222 | local_irq_restore(flags); | 237 | local_irq_restore(flags); |
223 | } | 238 | } |
224 | } | 239 | } |
225 | 240 | ||
226 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); | 241 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
227 | 242 | ||
228 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | 243 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
229 | extern unsigned long vma_address(struct page *page, | 244 | extern unsigned long vma_address(struct page *page, |
230 | struct vm_area_struct *vma); | 245 | struct vm_area_struct *vma); |
231 | #endif | 246 | #endif |
232 | #else /* !CONFIG_MMU */ | 247 | #else /* !CONFIG_MMU */ |
233 | static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p) | 248 | static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p) |
234 | { | 249 | { |
235 | return 0; | 250 | return 0; |
236 | } | 251 | } |
237 | static inline void clear_page_mlock(struct page *page) { } | 252 | static inline void clear_page_mlock(struct page *page) { } |
238 | static inline void mlock_vma_page(struct page *page) { } | 253 | static inline void mlock_vma_page(struct page *page) { } |
239 | static inline void mlock_migrate_page(struct page *new, struct page *old) { } | 254 | static inline void mlock_migrate_page(struct page *new, struct page *old) { } |
240 | 255 | ||
241 | #endif /* !CONFIG_MMU */ | 256 | #endif /* !CONFIG_MMU */ |
242 | 257 | ||
243 | /* | 258 | /* |
244 | * Return the mem_map entry representing the 'offset' subpage within | 259 | * Return the mem_map entry representing the 'offset' subpage within |
245 | * the maximally aligned gigantic page 'base'. Handle any discontiguity | 260 | * the maximally aligned gigantic page 'base'. Handle any discontiguity |
246 | * in the mem_map at MAX_ORDER_NR_PAGES boundaries. | 261 | * in the mem_map at MAX_ORDER_NR_PAGES boundaries. |
247 | */ | 262 | */ |
248 | static inline struct page *mem_map_offset(struct page *base, int offset) | 263 | static inline struct page *mem_map_offset(struct page *base, int offset) |
249 | { | 264 | { |
250 | if (unlikely(offset >= MAX_ORDER_NR_PAGES)) | 265 | if (unlikely(offset >= MAX_ORDER_NR_PAGES)) |
251 | return pfn_to_page(page_to_pfn(base) + offset); | 266 | return pfn_to_page(page_to_pfn(base) + offset); |
252 | return base + offset; | 267 | return base + offset; |
253 | } | 268 | } |
254 | 269 | ||
255 | /* | 270 | /* |
256 | * Iterator over all subpages within the maximally aligned gigantic | 271 | * Iterator over all subpages within the maximally aligned gigantic |
257 | * page 'base'. Handle any discontiguity in the mem_map. | 272 | * page 'base'. Handle any discontiguity in the mem_map. |
258 | */ | 273 | */ |
259 | static inline struct page *mem_map_next(struct page *iter, | 274 | static inline struct page *mem_map_next(struct page *iter, |
260 | struct page *base, int offset) | 275 | struct page *base, int offset) |
261 | { | 276 | { |
262 | if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { | 277 | if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { |
263 | unsigned long pfn = page_to_pfn(base) + offset; | 278 | unsigned long pfn = page_to_pfn(base) + offset; |
264 | if (!pfn_valid(pfn)) | 279 | if (!pfn_valid(pfn)) |
265 | return NULL; | 280 | return NULL; |
266 | return pfn_to_page(pfn); | 281 | return pfn_to_page(pfn); |
267 | } | 282 | } |
268 | return iter + 1; | 283 | return iter + 1; |
269 | } | 284 | } |
270 | 285 | ||
271 | /* | 286 | /* |
272 | * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node, | 287 | * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node, |
273 | * so all functions starting at paging_init should be marked __init | 288 | * so all functions starting at paging_init should be marked __init |
274 | * in those cases. SPARSEMEM, however, allows for memory hotplug, | 289 | * in those cases. SPARSEMEM, however, allows for memory hotplug, |
275 | * and alloc_bootmem_node is not used. | 290 | * and alloc_bootmem_node is not used. |
276 | */ | 291 | */ |
277 | #ifdef CONFIG_SPARSEMEM | 292 | #ifdef CONFIG_SPARSEMEM |
278 | #define __paginginit __meminit | 293 | #define __paginginit __meminit |
279 | #else | 294 | #else |
280 | #define __paginginit __init | 295 | #define __paginginit __init |
281 | #endif | 296 | #endif |
282 | 297 | ||
283 | /* Memory initialisation debug and verification */ | 298 | /* Memory initialisation debug and verification */ |
284 | enum mminit_level { | 299 | enum mminit_level { |
285 | MMINIT_WARNING, | 300 | MMINIT_WARNING, |
286 | MMINIT_VERIFY, | 301 | MMINIT_VERIFY, |
287 | MMINIT_TRACE | 302 | MMINIT_TRACE |
288 | }; | 303 | }; |
289 | 304 | ||
290 | #ifdef CONFIG_DEBUG_MEMORY_INIT | 305 | #ifdef CONFIG_DEBUG_MEMORY_INIT |
291 | 306 | ||
292 | extern int mminit_loglevel; | 307 | extern int mminit_loglevel; |
293 | 308 | ||
294 | #define mminit_dprintk(level, prefix, fmt, arg...) \ | 309 | #define mminit_dprintk(level, prefix, fmt, arg...) \ |
295 | do { \ | 310 | do { \ |
296 | if (level < mminit_loglevel) { \ | 311 | if (level < mminit_loglevel) { \ |
297 | printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \ | 312 | printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \ |
298 | printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \ | 313 | printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \ |
299 | } \ | 314 | } \ |
300 | } while (0) | 315 | } while (0) |
301 | 316 | ||
302 | extern void mminit_verify_pageflags_layout(void); | 317 | extern void mminit_verify_pageflags_layout(void); |
303 | extern void mminit_verify_page_links(struct page *page, | 318 | extern void mminit_verify_page_links(struct page *page, |
304 | enum zone_type zone, unsigned long nid, unsigned long pfn); | 319 | enum zone_type zone, unsigned long nid, unsigned long pfn); |
305 | extern void mminit_verify_zonelist(void); | 320 | extern void mminit_verify_zonelist(void); |
306 | 321 | ||
307 | #else | 322 | #else |
308 | 323 | ||
309 | static inline void mminit_dprintk(enum mminit_level level, | 324 | static inline void mminit_dprintk(enum mminit_level level, |
310 | const char *prefix, const char *fmt, ...) | 325 | const char *prefix, const char *fmt, ...) |
311 | { | 326 | { |
312 | } | 327 | } |
313 | 328 | ||
314 | static inline void mminit_verify_pageflags_layout(void) | 329 | static inline void mminit_verify_pageflags_layout(void) |
315 | { | 330 | { |
316 | } | 331 | } |
317 | 332 | ||
318 | static inline void mminit_verify_page_links(struct page *page, | 333 | static inline void mminit_verify_page_links(struct page *page, |
319 | enum zone_type zone, unsigned long nid, unsigned long pfn) | 334 | enum zone_type zone, unsigned long nid, unsigned long pfn) |
320 | { | 335 | { |
321 | } | 336 | } |
322 | 337 | ||
323 | static inline void mminit_verify_zonelist(void) | 338 | static inline void mminit_verify_zonelist(void) |
324 | { | 339 | { |
325 | } | 340 | } |
326 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ | 341 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
327 | 342 | ||
328 | /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */ | 343 | /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */ |
329 | #if defined(CONFIG_SPARSEMEM) | 344 | #if defined(CONFIG_SPARSEMEM) |
330 | extern void mminit_validate_memmodel_limits(unsigned long *start_pfn, | 345 | extern void mminit_validate_memmodel_limits(unsigned long *start_pfn, |
331 | unsigned long *end_pfn); | 346 | unsigned long *end_pfn); |
332 | #else | 347 | #else |
333 | static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn, | 348 | static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn, |
334 | unsigned long *end_pfn) | 349 | unsigned long *end_pfn) |
335 | { | 350 | { |
336 | } | 351 | } |
337 | #endif /* CONFIG_SPARSEMEM */ | 352 | #endif /* CONFIG_SPARSEMEM */ |
338 | 353 | ||
339 | #define ZONE_RECLAIM_NOSCAN -2 | 354 | #define ZONE_RECLAIM_NOSCAN -2 |
340 | #define ZONE_RECLAIM_FULL -1 | 355 | #define ZONE_RECLAIM_FULL -1 |
341 | #define ZONE_RECLAIM_SOME 0 | 356 | #define ZONE_RECLAIM_SOME 0 |
342 | #define ZONE_RECLAIM_SUCCESS 1 | 357 | #define ZONE_RECLAIM_SUCCESS 1 |
343 | 358 | ||
344 | extern int hwpoison_filter(struct page *p); | 359 | extern int hwpoison_filter(struct page *p); |
345 | 360 | ||
346 | extern u32 hwpoison_filter_dev_major; | 361 | extern u32 hwpoison_filter_dev_major; |
347 | extern u32 hwpoison_filter_dev_minor; | 362 | extern u32 hwpoison_filter_dev_minor; |
348 | extern u64 hwpoison_filter_flags_mask; | 363 | extern u64 hwpoison_filter_flags_mask; |
349 | extern u64 hwpoison_filter_flags_value; | 364 | extern u64 hwpoison_filter_flags_value; |
350 | extern u64 hwpoison_filter_memcg; | 365 | extern u64 hwpoison_filter_memcg; |
351 | extern u32 hwpoison_filter_enable; | 366 | extern u32 hwpoison_filter_enable; |
352 | 367 | ||
353 | extern unsigned long vm_mmap_pgoff(struct file *, unsigned long, | 368 | extern unsigned long vm_mmap_pgoff(struct file *, unsigned long, |
354 | unsigned long, unsigned long, | 369 | unsigned long, unsigned long, |
355 | unsigned long, unsigned long); | 370 | unsigned long, unsigned long); |
356 | 371 | ||
357 | extern void set_pageblock_order(void); | 372 | extern void set_pageblock_order(void); |
358 | unsigned long reclaim_clean_pages_from_list(struct zone *zone, | 373 | unsigned long reclaim_clean_pages_from_list(struct zone *zone, |
359 | struct list_head *page_list); | 374 | struct list_head *page_list); |
360 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ | 375 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
361 | #define ALLOC_WMARK_MIN WMARK_MIN | 376 | #define ALLOC_WMARK_MIN WMARK_MIN |
362 | #define ALLOC_WMARK_LOW WMARK_LOW | 377 | #define ALLOC_WMARK_LOW WMARK_LOW |
363 | #define ALLOC_WMARK_HIGH WMARK_HIGH | 378 | #define ALLOC_WMARK_HIGH WMARK_HIGH |
364 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ | 379 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ |
365 | 380 | ||
366 | /* Mask to get the watermark bits */ | 381 | /* Mask to get the watermark bits */ |
367 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) | 382 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) |
368 | 383 | ||
369 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ | 384 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ |
370 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ | 385 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ |
371 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ | 386 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ |
372 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ | 387 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ |
373 | #define ALLOC_FAIR 0x100 /* fair zone allocation */ | 388 | #define ALLOC_FAIR 0x100 /* fair zone allocation */ |
374 | 389 | ||
375 | #endif /* __MM_INTERNAL_H */ | 390 | #endif /* __MM_INTERNAL_H */ |
376 | 391 |
mm/readahead.c
1 | /* | 1 | /* |
2 | * mm/readahead.c - address_space-level file readahead. | 2 | * mm/readahead.c - address_space-level file readahead. |
3 | * | 3 | * |
4 | * Copyright (C) 2002, Linus Torvalds | 4 | * Copyright (C) 2002, Linus Torvalds |
5 | * | 5 | * |
6 | * 09Apr2002 Andrew Morton | 6 | * 09Apr2002 Andrew Morton |
7 | * Initial version. | 7 | * Initial version. |
8 | */ | 8 | */ |
9 | 9 | ||
10 | #include <linux/kernel.h> | 10 | #include <linux/kernel.h> |
11 | #include <linux/fs.h> | ||
12 | #include <linux/gfp.h> | 11 | #include <linux/gfp.h> |
13 | #include <linux/mm.h> | ||
14 | #include <linux/export.h> | 12 | #include <linux/export.h> |
15 | #include <linux/blkdev.h> | 13 | #include <linux/blkdev.h> |
16 | #include <linux/backing-dev.h> | 14 | #include <linux/backing-dev.h> |
17 | #include <linux/task_io_accounting_ops.h> | 15 | #include <linux/task_io_accounting_ops.h> |
18 | #include <linux/pagevec.h> | 16 | #include <linux/pagevec.h> |
19 | #include <linux/pagemap.h> | 17 | #include <linux/pagemap.h> |
20 | #include <linux/syscalls.h> | 18 | #include <linux/syscalls.h> |
21 | #include <linux/file.h> | 19 | #include <linux/file.h> |
22 | 20 | ||
21 | #include "internal.h" | ||
22 | |||
23 | /* | 23 | /* |
24 | * Initialise a struct file's readahead state. Assumes that the caller has | 24 | * Initialise a struct file's readahead state. Assumes that the caller has |
25 | * memset *ra to zero. | 25 | * memset *ra to zero. |
26 | */ | 26 | */ |
27 | void | 27 | void |
28 | file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping) | 28 | file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping) |
29 | { | 29 | { |
30 | ra->ra_pages = mapping->backing_dev_info->ra_pages; | 30 | ra->ra_pages = mapping->backing_dev_info->ra_pages; |
31 | ra->prev_pos = -1; | 31 | ra->prev_pos = -1; |
32 | } | 32 | } |
33 | EXPORT_SYMBOL_GPL(file_ra_state_init); | 33 | EXPORT_SYMBOL_GPL(file_ra_state_init); |
34 | 34 | ||
35 | #define list_to_page(head) (list_entry((head)->prev, struct page, lru)) | 35 | #define list_to_page(head) (list_entry((head)->prev, struct page, lru)) |
36 | 36 | ||
37 | /* | 37 | /* |
38 | * see if a page needs releasing upon read_cache_pages() failure | 38 | * see if a page needs releasing upon read_cache_pages() failure |
39 | * - the caller of read_cache_pages() may have set PG_private or PG_fscache | 39 | * - the caller of read_cache_pages() may have set PG_private or PG_fscache |
40 | * before calling, such as the NFS fs marking pages that are cached locally | 40 | * before calling, such as the NFS fs marking pages that are cached locally |
41 | * on disk, thus we need to give the fs a chance to clean up in the event of | 41 | * on disk, thus we need to give the fs a chance to clean up in the event of |
42 | * an error | 42 | * an error |
43 | */ | 43 | */ |
44 | static void read_cache_pages_invalidate_page(struct address_space *mapping, | 44 | static void read_cache_pages_invalidate_page(struct address_space *mapping, |
45 | struct page *page) | 45 | struct page *page) |
46 | { | 46 | { |
47 | if (page_has_private(page)) { | 47 | if (page_has_private(page)) { |
48 | if (!trylock_page(page)) | 48 | if (!trylock_page(page)) |
49 | BUG(); | 49 | BUG(); |
50 | page->mapping = mapping; | 50 | page->mapping = mapping; |
51 | do_invalidatepage(page, 0, PAGE_CACHE_SIZE); | 51 | do_invalidatepage(page, 0, PAGE_CACHE_SIZE); |
52 | page->mapping = NULL; | 52 | page->mapping = NULL; |
53 | unlock_page(page); | 53 | unlock_page(page); |
54 | } | 54 | } |
55 | page_cache_release(page); | 55 | page_cache_release(page); |
56 | } | 56 | } |
57 | 57 | ||
58 | /* | 58 | /* |
59 | * release a list of pages, invalidating them first if need be | 59 | * release a list of pages, invalidating them first if need be |
60 | */ | 60 | */ |
61 | static void read_cache_pages_invalidate_pages(struct address_space *mapping, | 61 | static void read_cache_pages_invalidate_pages(struct address_space *mapping, |
62 | struct list_head *pages) | 62 | struct list_head *pages) |
63 | { | 63 | { |
64 | struct page *victim; | 64 | struct page *victim; |
65 | 65 | ||
66 | while (!list_empty(pages)) { | 66 | while (!list_empty(pages)) { |
67 | victim = list_to_page(pages); | 67 | victim = list_to_page(pages); |
68 | list_del(&victim->lru); | 68 | list_del(&victim->lru); |
69 | read_cache_pages_invalidate_page(mapping, victim); | 69 | read_cache_pages_invalidate_page(mapping, victim); |
70 | } | 70 | } |
71 | } | 71 | } |
72 | 72 | ||
73 | /** | 73 | /** |
74 | * read_cache_pages - populate an address space with some pages & start reads against them | 74 | * read_cache_pages - populate an address space with some pages & start reads against them |
75 | * @mapping: the address_space | 75 | * @mapping: the address_space |
76 | * @pages: The address of a list_head which contains the target pages. These | 76 | * @pages: The address of a list_head which contains the target pages. These |
77 | * pages have their ->index populated and are otherwise uninitialised. | 77 | * pages have their ->index populated and are otherwise uninitialised. |
78 | * @filler: callback routine for filling a single page. | 78 | * @filler: callback routine for filling a single page. |
79 | * @data: private data for the callback routine. | 79 | * @data: private data for the callback routine. |
80 | * | 80 | * |
81 | * Hides the details of the LRU cache etc from the filesystems. | 81 | * Hides the details of the LRU cache etc from the filesystems. |
82 | */ | 82 | */ |
83 | int read_cache_pages(struct address_space *mapping, struct list_head *pages, | 83 | int read_cache_pages(struct address_space *mapping, struct list_head *pages, |
84 | int (*filler)(void *, struct page *), void *data) | 84 | int (*filler)(void *, struct page *), void *data) |
85 | { | 85 | { |
86 | struct page *page; | 86 | struct page *page; |
87 | int ret = 0; | 87 | int ret = 0; |
88 | 88 | ||
89 | while (!list_empty(pages)) { | 89 | while (!list_empty(pages)) { |
90 | page = list_to_page(pages); | 90 | page = list_to_page(pages); |
91 | list_del(&page->lru); | 91 | list_del(&page->lru); |
92 | if (add_to_page_cache_lru(page, mapping, | 92 | if (add_to_page_cache_lru(page, mapping, |
93 | page->index, GFP_KERNEL)) { | 93 | page->index, GFP_KERNEL)) { |
94 | read_cache_pages_invalidate_page(mapping, page); | 94 | read_cache_pages_invalidate_page(mapping, page); |
95 | continue; | 95 | continue; |
96 | } | 96 | } |
97 | page_cache_release(page); | 97 | page_cache_release(page); |
98 | 98 | ||
99 | ret = filler(data, page); | 99 | ret = filler(data, page); |
100 | if (unlikely(ret)) { | 100 | if (unlikely(ret)) { |
101 | read_cache_pages_invalidate_pages(mapping, pages); | 101 | read_cache_pages_invalidate_pages(mapping, pages); |
102 | break; | 102 | break; |
103 | } | 103 | } |
104 | task_io_account_read(PAGE_CACHE_SIZE); | 104 | task_io_account_read(PAGE_CACHE_SIZE); |
105 | } | 105 | } |
106 | return ret; | 106 | return ret; |
107 | } | 107 | } |
108 | 108 | ||
109 | EXPORT_SYMBOL(read_cache_pages); | 109 | EXPORT_SYMBOL(read_cache_pages); |
110 | 110 | ||
111 | static int read_pages(struct address_space *mapping, struct file *filp, | 111 | static int read_pages(struct address_space *mapping, struct file *filp, |
112 | struct list_head *pages, unsigned nr_pages) | 112 | struct list_head *pages, unsigned nr_pages) |
113 | { | 113 | { |
114 | struct blk_plug plug; | 114 | struct blk_plug plug; |
115 | unsigned page_idx; | 115 | unsigned page_idx; |
116 | int ret; | 116 | int ret; |
117 | 117 | ||
118 | blk_start_plug(&plug); | 118 | blk_start_plug(&plug); |
119 | 119 | ||
120 | if (mapping->a_ops->readpages) { | 120 | if (mapping->a_ops->readpages) { |
121 | ret = mapping->a_ops->readpages(filp, mapping, pages, nr_pages); | 121 | ret = mapping->a_ops->readpages(filp, mapping, pages, nr_pages); |
122 | /* Clean up the remaining pages */ | 122 | /* Clean up the remaining pages */ |
123 | put_pages_list(pages); | 123 | put_pages_list(pages); |
124 | goto out; | 124 | goto out; |
125 | } | 125 | } |
126 | 126 | ||
127 | for (page_idx = 0; page_idx < nr_pages; page_idx++) { | 127 | for (page_idx = 0; page_idx < nr_pages; page_idx++) { |
128 | struct page *page = list_to_page(pages); | 128 | struct page *page = list_to_page(pages); |
129 | list_del(&page->lru); | 129 | list_del(&page->lru); |
130 | if (!add_to_page_cache_lru(page, mapping, | 130 | if (!add_to_page_cache_lru(page, mapping, |
131 | page->index, GFP_KERNEL)) { | 131 | page->index, GFP_KERNEL)) { |
132 | mapping->a_ops->readpage(filp, page); | 132 | mapping->a_ops->readpage(filp, page); |
133 | } | 133 | } |
134 | page_cache_release(page); | 134 | page_cache_release(page); |
135 | } | 135 | } |
136 | ret = 0; | 136 | ret = 0; |
137 | 137 | ||
138 | out: | 138 | out: |
139 | blk_finish_plug(&plug); | 139 | blk_finish_plug(&plug); |
140 | 140 | ||
141 | return ret; | 141 | return ret; |
142 | } | 142 | } |
143 | 143 | ||
144 | /* | 144 | /* |
145 | * __do_page_cache_readahead() actually reads a chunk of disk. It allocates all | 145 | * __do_page_cache_readahead() actually reads a chunk of disk. It allocates all |
146 | * the pages first, then submits them all for I/O. This avoids the very bad | 146 | * the pages first, then submits them all for I/O. This avoids the very bad |
147 | * behaviour which would occur if page allocations are causing VM writeback. | 147 | * behaviour which would occur if page allocations are causing VM writeback. |
148 | * We really don't want to intermingle reads and writes like that. | 148 | * We really don't want to intermingle reads and writes like that. |
149 | * | 149 | * |
150 | * Returns the number of pages requested, or the maximum amount of I/O allowed. | 150 | * Returns the number of pages requested, or the maximum amount of I/O allowed. |
151 | */ | 151 | */ |
152 | static int | 152 | int __do_page_cache_readahead(struct address_space *mapping, struct file *filp, |
153 | __do_page_cache_readahead(struct address_space *mapping, struct file *filp, | ||
154 | pgoff_t offset, unsigned long nr_to_read, | 153 | pgoff_t offset, unsigned long nr_to_read, |
155 | unsigned long lookahead_size) | 154 | unsigned long lookahead_size) |
156 | { | 155 | { |
157 | struct inode *inode = mapping->host; | 156 | struct inode *inode = mapping->host; |
158 | struct page *page; | 157 | struct page *page; |
159 | unsigned long end_index; /* The last page we want to read */ | 158 | unsigned long end_index; /* The last page we want to read */ |
160 | LIST_HEAD(page_pool); | 159 | LIST_HEAD(page_pool); |
161 | int page_idx; | 160 | int page_idx; |
162 | int ret = 0; | 161 | int ret = 0; |
163 | loff_t isize = i_size_read(inode); | 162 | loff_t isize = i_size_read(inode); |
164 | 163 | ||
165 | if (isize == 0) | 164 | if (isize == 0) |
166 | goto out; | 165 | goto out; |
167 | 166 | ||
168 | end_index = ((isize - 1) >> PAGE_CACHE_SHIFT); | 167 | end_index = ((isize - 1) >> PAGE_CACHE_SHIFT); |
169 | 168 | ||
170 | /* | 169 | /* |
171 | * Preallocate as many pages as we will need. | 170 | * Preallocate as many pages as we will need. |
172 | */ | 171 | */ |
173 | for (page_idx = 0; page_idx < nr_to_read; page_idx++) { | 172 | for (page_idx = 0; page_idx < nr_to_read; page_idx++) { |
174 | pgoff_t page_offset = offset + page_idx; | 173 | pgoff_t page_offset = offset + page_idx; |
175 | 174 | ||
176 | if (page_offset > end_index) | 175 | if (page_offset > end_index) |
177 | break; | 176 | break; |
178 | 177 | ||
179 | rcu_read_lock(); | 178 | rcu_read_lock(); |
180 | page = radix_tree_lookup(&mapping->page_tree, page_offset); | 179 | page = radix_tree_lookup(&mapping->page_tree, page_offset); |
181 | rcu_read_unlock(); | 180 | rcu_read_unlock(); |
182 | if (page && !radix_tree_exceptional_entry(page)) | 181 | if (page && !radix_tree_exceptional_entry(page)) |
183 | continue; | 182 | continue; |
184 | 183 | ||
185 | page = page_cache_alloc_readahead(mapping); | 184 | page = page_cache_alloc_readahead(mapping); |
186 | if (!page) | 185 | if (!page) |
187 | break; | 186 | break; |
188 | page->index = page_offset; | 187 | page->index = page_offset; |
189 | list_add(&page->lru, &page_pool); | 188 | list_add(&page->lru, &page_pool); |
190 | if (page_idx == nr_to_read - lookahead_size) | 189 | if (page_idx == nr_to_read - lookahead_size) |
191 | SetPageReadahead(page); | 190 | SetPageReadahead(page); |
192 | ret++; | 191 | ret++; |
193 | } | 192 | } |
194 | 193 | ||
195 | /* | 194 | /* |
196 | * Now start the IO. We ignore I/O errors - if the page is not | 195 | * Now start the IO. We ignore I/O errors - if the page is not |
197 | * uptodate then the caller will launch readpage again, and | 196 | * uptodate then the caller will launch readpage again, and |
198 | * will then handle the error. | 197 | * will then handle the error. |
199 | */ | 198 | */ |
200 | if (ret) | 199 | if (ret) |
201 | read_pages(mapping, filp, &page_pool, ret); | 200 | read_pages(mapping, filp, &page_pool, ret); |
202 | BUG_ON(!list_empty(&page_pool)); | 201 | BUG_ON(!list_empty(&page_pool)); |
203 | out: | 202 | out: |
204 | return ret; | 203 | return ret; |
205 | } | 204 | } |
206 | 205 | ||
207 | /* | 206 | /* |
208 | * Chunk the readahead into 2 megabyte units, so that we don't pin too much | 207 | * Chunk the readahead into 2 megabyte units, so that we don't pin too much |
209 | * memory at once. | 208 | * memory at once. |
210 | */ | 209 | */ |
211 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, | 210 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
212 | pgoff_t offset, unsigned long nr_to_read) | 211 | pgoff_t offset, unsigned long nr_to_read) |
213 | { | 212 | { |
214 | if (unlikely(!mapping->a_ops->readpage && !mapping->a_ops->readpages)) | 213 | if (unlikely(!mapping->a_ops->readpage && !mapping->a_ops->readpages)) |
215 | return -EINVAL; | 214 | return -EINVAL; |
216 | 215 | ||
217 | nr_to_read = max_sane_readahead(nr_to_read); | 216 | nr_to_read = max_sane_readahead(nr_to_read); |
218 | while (nr_to_read) { | 217 | while (nr_to_read) { |
219 | int err; | 218 | int err; |
220 | 219 | ||
221 | unsigned long this_chunk = (2 * 1024 * 1024) / PAGE_CACHE_SIZE; | 220 | unsigned long this_chunk = (2 * 1024 * 1024) / PAGE_CACHE_SIZE; |
222 | 221 | ||
223 | if (this_chunk > nr_to_read) | 222 | if (this_chunk > nr_to_read) |
224 | this_chunk = nr_to_read; | 223 | this_chunk = nr_to_read; |
225 | err = __do_page_cache_readahead(mapping, filp, | 224 | err = __do_page_cache_readahead(mapping, filp, |
226 | offset, this_chunk, 0); | 225 | offset, this_chunk, 0); |
227 | if (err < 0) | 226 | if (err < 0) |
228 | return err; | 227 | return err; |
229 | 228 | ||
230 | offset += this_chunk; | 229 | offset += this_chunk; |
231 | nr_to_read -= this_chunk; | 230 | nr_to_read -= this_chunk; |
232 | } | 231 | } |
233 | return 0; | 232 | return 0; |
234 | } | 233 | } |
235 | 234 | ||
236 | #define MAX_READAHEAD ((512*4096)/PAGE_CACHE_SIZE) | 235 | #define MAX_READAHEAD ((512*4096)/PAGE_CACHE_SIZE) |
237 | /* | 236 | /* |
238 | * Given a desired number of PAGE_CACHE_SIZE readahead pages, return a | 237 | * Given a desired number of PAGE_CACHE_SIZE readahead pages, return a |
239 | * sensible upper limit. | 238 | * sensible upper limit. |
240 | */ | 239 | */ |
241 | unsigned long max_sane_readahead(unsigned long nr) | 240 | unsigned long max_sane_readahead(unsigned long nr) |
242 | { | 241 | { |
243 | return min(nr, MAX_READAHEAD); | 242 | return min(nr, MAX_READAHEAD); |
244 | } | ||
245 | |||
246 | /* | ||
247 | * Submit IO for the read-ahead request in file_ra_state. | ||
248 | */ | ||
249 | unsigned long ra_submit(struct file_ra_state *ra, | ||
250 | struct address_space *mapping, struct file *filp) | ||
251 | { | ||
252 | int actual; | ||
253 | |||
254 | actual = __do_page_cache_readahead(mapping, filp, | ||
255 | ra->start, ra->size, ra->async_size); | ||
256 | |||
257 | return actual; | ||
258 | } | 243 | } |
259 | 244 | ||
260 | /* | 245 | /* |
261 | * Set the initial window size, round to next power of 2 and square | 246 | * Set the initial window size, round to next power of 2 and square |
262 | * for small size, x 4 for medium, and x 2 for large | 247 | * for small size, x 4 for medium, and x 2 for large |
263 | * for 128k (32 page) max ra | 248 | * for 128k (32 page) max ra |
264 | * 1-8 page = 32k initial, > 8 page = 128k initial | 249 | * 1-8 page = 32k initial, > 8 page = 128k initial |
265 | */ | 250 | */ |
266 | static unsigned long get_init_ra_size(unsigned long size, unsigned long max) | 251 | static unsigned long get_init_ra_size(unsigned long size, unsigned long max) |
267 | { | 252 | { |
268 | unsigned long newsize = roundup_pow_of_two(size); | 253 | unsigned long newsize = roundup_pow_of_two(size); |
269 | 254 | ||
270 | if (newsize <= max / 32) | 255 | if (newsize <= max / 32) |
271 | newsize = newsize * 4; | 256 | newsize = newsize * 4; |
272 | else if (newsize <= max / 4) | 257 | else if (newsize <= max / 4) |
273 | newsize = newsize * 2; | 258 | newsize = newsize * 2; |
274 | else | 259 | else |
275 | newsize = max; | 260 | newsize = max; |
276 | 261 | ||
277 | return newsize; | 262 | return newsize; |
278 | } | 263 | } |
279 | 264 | ||
280 | /* | 265 | /* |
281 | * Get the previous window size, ramp it up, and | 266 | * Get the previous window size, ramp it up, and |
282 | * return it as the new window size. | 267 | * return it as the new window size. |
283 | */ | 268 | */ |
284 | static unsigned long get_next_ra_size(struct file_ra_state *ra, | 269 | static unsigned long get_next_ra_size(struct file_ra_state *ra, |
285 | unsigned long max) | 270 | unsigned long max) |
286 | { | 271 | { |
287 | unsigned long cur = ra->size; | 272 | unsigned long cur = ra->size; |
288 | unsigned long newsize; | 273 | unsigned long newsize; |
289 | 274 | ||
290 | if (cur < max / 16) | 275 | if (cur < max / 16) |
291 | newsize = 4 * cur; | 276 | newsize = 4 * cur; |
292 | else | 277 | else |
293 | newsize = 2 * cur; | 278 | newsize = 2 * cur; |
294 | 279 | ||
295 | return min(newsize, max); | 280 | return min(newsize, max); |
296 | } | 281 | } |
297 | 282 | ||
298 | /* | 283 | /* |
299 | * On-demand readahead design. | 284 | * On-demand readahead design. |
300 | * | 285 | * |
301 | * The fields in struct file_ra_state represent the most-recently-executed | 286 | * The fields in struct file_ra_state represent the most-recently-executed |
302 | * readahead attempt: | 287 | * readahead attempt: |
303 | * | 288 | * |
304 | * |<----- async_size ---------| | 289 | * |<----- async_size ---------| |
305 | * |------------------- size -------------------->| | 290 | * |------------------- size -------------------->| |
306 | * |==================#===========================| | 291 | * |==================#===========================| |
307 | * ^start ^page marked with PG_readahead | 292 | * ^start ^page marked with PG_readahead |
308 | * | 293 | * |
309 | * To overlap application thinking time and disk I/O time, we do | 294 | * To overlap application thinking time and disk I/O time, we do |
310 | * `readahead pipelining': Do not wait until the application consumed all | 295 | * `readahead pipelining': Do not wait until the application consumed all |
311 | * readahead pages and stalled on the missing page at readahead_index; | 296 | * readahead pages and stalled on the missing page at readahead_index; |
312 | * Instead, submit an asynchronous readahead I/O as soon as there are | 297 | * Instead, submit an asynchronous readahead I/O as soon as there are |
313 | * only async_size pages left in the readahead window. Normally async_size | 298 | * only async_size pages left in the readahead window. Normally async_size |
314 | * will be equal to size, for maximum pipelining. | 299 | * will be equal to size, for maximum pipelining. |
315 | * | 300 | * |
316 | * In interleaved sequential reads, concurrent streams on the same fd can | 301 | * In interleaved sequential reads, concurrent streams on the same fd can |
317 | * be invalidating each other's readahead state. So we flag the new readahead | 302 | * be invalidating each other's readahead state. So we flag the new readahead |
318 | * page at (start+size-async_size) with PG_readahead, and use it as readahead | 303 | * page at (start+size-async_size) with PG_readahead, and use it as readahead |
319 | * indicator. The flag won't be set on already cached pages, to avoid the | 304 | * indicator. The flag won't be set on already cached pages, to avoid the |
320 | * readahead-for-nothing fuss, saving pointless page cache lookups. | 305 | * readahead-for-nothing fuss, saving pointless page cache lookups. |
321 | * | 306 | * |
322 | * prev_pos tracks the last visited byte in the _previous_ read request. | 307 | * prev_pos tracks the last visited byte in the _previous_ read request. |
323 | * It should be maintained by the caller, and will be used for detecting | 308 | * It should be maintained by the caller, and will be used for detecting |
324 | * small random reads. Note that the readahead algorithm checks loosely | 309 | * small random reads. Note that the readahead algorithm checks loosely |
325 | * for sequential patterns. Hence interleaved reads might be served as | 310 | * for sequential patterns. Hence interleaved reads might be served as |
326 | * sequential ones. | 311 | * sequential ones. |
327 | * | 312 | * |
328 | * There is a special-case: if the first page which the application tries to | 313 | * There is a special-case: if the first page which the application tries to |
329 | * read happens to be the first page of the file, it is assumed that a linear | 314 | * read happens to be the first page of the file, it is assumed that a linear |
330 | * read is about to happen and the window is immediately set to the initial size | 315 | * read is about to happen and the window is immediately set to the initial size |
331 | * based on I/O request size and the max_readahead. | 316 | * based on I/O request size and the max_readahead. |
332 | * | 317 | * |
333 | * The code ramps up the readahead size aggressively at first, but slow down as | 318 | * The code ramps up the readahead size aggressively at first, but slow down as |
334 | * it approaches max_readhead. | 319 | * it approaches max_readhead. |
335 | */ | 320 | */ |
336 | 321 | ||
337 | /* | 322 | /* |
338 | * Count contiguously cached pages from @offset-1 to @offset-@max, | 323 | * Count contiguously cached pages from @offset-1 to @offset-@max, |
339 | * this count is a conservative estimation of | 324 | * this count is a conservative estimation of |
340 | * - length of the sequential read sequence, or | 325 | * - length of the sequential read sequence, or |
341 | * - thrashing threshold in memory tight systems | 326 | * - thrashing threshold in memory tight systems |
342 | */ | 327 | */ |
343 | static pgoff_t count_history_pages(struct address_space *mapping, | 328 | static pgoff_t count_history_pages(struct address_space *mapping, |
344 | struct file_ra_state *ra, | 329 | struct file_ra_state *ra, |
345 | pgoff_t offset, unsigned long max) | 330 | pgoff_t offset, unsigned long max) |
346 | { | 331 | { |
347 | pgoff_t head; | 332 | pgoff_t head; |
348 | 333 | ||
349 | rcu_read_lock(); | 334 | rcu_read_lock(); |
350 | head = page_cache_prev_hole(mapping, offset - 1, max); | 335 | head = page_cache_prev_hole(mapping, offset - 1, max); |
351 | rcu_read_unlock(); | 336 | rcu_read_unlock(); |
352 | 337 | ||
353 | return offset - 1 - head; | 338 | return offset - 1 - head; |
354 | } | 339 | } |
355 | 340 | ||
356 | /* | 341 | /* |
357 | * page cache context based read-ahead | 342 | * page cache context based read-ahead |
358 | */ | 343 | */ |
359 | static int try_context_readahead(struct address_space *mapping, | 344 | static int try_context_readahead(struct address_space *mapping, |
360 | struct file_ra_state *ra, | 345 | struct file_ra_state *ra, |
361 | pgoff_t offset, | 346 | pgoff_t offset, |
362 | unsigned long req_size, | 347 | unsigned long req_size, |
363 | unsigned long max) | 348 | unsigned long max) |
364 | { | 349 | { |
365 | pgoff_t size; | 350 | pgoff_t size; |
366 | 351 | ||
367 | size = count_history_pages(mapping, ra, offset, max); | 352 | size = count_history_pages(mapping, ra, offset, max); |
368 | 353 | ||
369 | /* | 354 | /* |
370 | * not enough history pages: | 355 | * not enough history pages: |
371 | * it could be a random read | 356 | * it could be a random read |
372 | */ | 357 | */ |
373 | if (size <= req_size) | 358 | if (size <= req_size) |
374 | return 0; | 359 | return 0; |
375 | 360 | ||
376 | /* | 361 | /* |
377 | * starts from beginning of file: | 362 | * starts from beginning of file: |
378 | * it is a strong indication of long-run stream (or whole-file-read) | 363 | * it is a strong indication of long-run stream (or whole-file-read) |
379 | */ | 364 | */ |
380 | if (size >= offset) | 365 | if (size >= offset) |
381 | size *= 2; | 366 | size *= 2; |
382 | 367 | ||
383 | ra->start = offset; | 368 | ra->start = offset; |
384 | ra->size = min(size + req_size, max); | 369 | ra->size = min(size + req_size, max); |
385 | ra->async_size = 1; | 370 | ra->async_size = 1; |
386 | 371 | ||
387 | return 1; | 372 | return 1; |
388 | } | 373 | } |
389 | 374 | ||
390 | /* | 375 | /* |
391 | * A minimal readahead algorithm for trivial sequential/random reads. | 376 | * A minimal readahead algorithm for trivial sequential/random reads. |
392 | */ | 377 | */ |
393 | static unsigned long | 378 | static unsigned long |
394 | ondemand_readahead(struct address_space *mapping, | 379 | ondemand_readahead(struct address_space *mapping, |
395 | struct file_ra_state *ra, struct file *filp, | 380 | struct file_ra_state *ra, struct file *filp, |
396 | bool hit_readahead_marker, pgoff_t offset, | 381 | bool hit_readahead_marker, pgoff_t offset, |
397 | unsigned long req_size) | 382 | unsigned long req_size) |
398 | { | 383 | { |
399 | unsigned long max = max_sane_readahead(ra->ra_pages); | 384 | unsigned long max = max_sane_readahead(ra->ra_pages); |
400 | pgoff_t prev_offset; | 385 | pgoff_t prev_offset; |
401 | 386 | ||
402 | /* | 387 | /* |
403 | * start of file | 388 | * start of file |
404 | */ | 389 | */ |
405 | if (!offset) | 390 | if (!offset) |
406 | goto initial_readahead; | 391 | goto initial_readahead; |
407 | 392 | ||
408 | /* | 393 | /* |
409 | * It's the expected callback offset, assume sequential access. | 394 | * It's the expected callback offset, assume sequential access. |
410 | * Ramp up sizes, and push forward the readahead window. | 395 | * Ramp up sizes, and push forward the readahead window. |
411 | */ | 396 | */ |
412 | if ((offset == (ra->start + ra->size - ra->async_size) || | 397 | if ((offset == (ra->start + ra->size - ra->async_size) || |
413 | offset == (ra->start + ra->size))) { | 398 | offset == (ra->start + ra->size))) { |
414 | ra->start += ra->size; | 399 | ra->start += ra->size; |
415 | ra->size = get_next_ra_size(ra, max); | 400 | ra->size = get_next_ra_size(ra, max); |
416 | ra->async_size = ra->size; | 401 | ra->async_size = ra->size; |
417 | goto readit; | 402 | goto readit; |
418 | } | 403 | } |
419 | 404 | ||
420 | /* | 405 | /* |
421 | * Hit a marked page without valid readahead state. | 406 | * Hit a marked page without valid readahead state. |
422 | * E.g. interleaved reads. | 407 | * E.g. interleaved reads. |
423 | * Query the pagecache for async_size, which normally equals to | 408 | * Query the pagecache for async_size, which normally equals to |
424 | * readahead size. Ramp it up and use it as the new readahead size. | 409 | * readahead size. Ramp it up and use it as the new readahead size. |
425 | */ | 410 | */ |
426 | if (hit_readahead_marker) { | 411 | if (hit_readahead_marker) { |
427 | pgoff_t start; | 412 | pgoff_t start; |
428 | 413 | ||
429 | rcu_read_lock(); | 414 | rcu_read_lock(); |
430 | start = page_cache_next_hole(mapping, offset + 1, max); | 415 | start = page_cache_next_hole(mapping, offset + 1, max); |
431 | rcu_read_unlock(); | 416 | rcu_read_unlock(); |
432 | 417 | ||
433 | if (!start || start - offset > max) | 418 | if (!start || start - offset > max) |
434 | return 0; | 419 | return 0; |
435 | 420 | ||
436 | ra->start = start; | 421 | ra->start = start; |
437 | ra->size = start - offset; /* old async_size */ | 422 | ra->size = start - offset; /* old async_size */ |
438 | ra->size += req_size; | 423 | ra->size += req_size; |
439 | ra->size = get_next_ra_size(ra, max); | 424 | ra->size = get_next_ra_size(ra, max); |
440 | ra->async_size = ra->size; | 425 | ra->async_size = ra->size; |
441 | goto readit; | 426 | goto readit; |
442 | } | 427 | } |
443 | 428 | ||
444 | /* | 429 | /* |
445 | * oversize read | 430 | * oversize read |
446 | */ | 431 | */ |
447 | if (req_size > max) | 432 | if (req_size > max) |
448 | goto initial_readahead; | 433 | goto initial_readahead; |
449 | 434 | ||
450 | /* | 435 | /* |
451 | * sequential cache miss | 436 | * sequential cache miss |
452 | * trivial case: (offset - prev_offset) == 1 | 437 | * trivial case: (offset - prev_offset) == 1 |
453 | * unaligned reads: (offset - prev_offset) == 0 | 438 | * unaligned reads: (offset - prev_offset) == 0 |
454 | */ | 439 | */ |
455 | prev_offset = (unsigned long long)ra->prev_pos >> PAGE_CACHE_SHIFT; | 440 | prev_offset = (unsigned long long)ra->prev_pos >> PAGE_CACHE_SHIFT; |
456 | if (offset - prev_offset <= 1UL) | 441 | if (offset - prev_offset <= 1UL) |
457 | goto initial_readahead; | 442 | goto initial_readahead; |
458 | 443 | ||
459 | /* | 444 | /* |
460 | * Query the page cache and look for the traces(cached history pages) | 445 | * Query the page cache and look for the traces(cached history pages) |
461 | * that a sequential stream would leave behind. | 446 | * that a sequential stream would leave behind. |
462 | */ | 447 | */ |
463 | if (try_context_readahead(mapping, ra, offset, req_size, max)) | 448 | if (try_context_readahead(mapping, ra, offset, req_size, max)) |
464 | goto readit; | 449 | goto readit; |
465 | 450 | ||
466 | /* | 451 | /* |
467 | * standalone, small random read | 452 | * standalone, small random read |
468 | * Read as is, and do not pollute the readahead state. | 453 | * Read as is, and do not pollute the readahead state. |
469 | */ | 454 | */ |
470 | return __do_page_cache_readahead(mapping, filp, offset, req_size, 0); | 455 | return __do_page_cache_readahead(mapping, filp, offset, req_size, 0); |
471 | 456 | ||
472 | initial_readahead: | 457 | initial_readahead: |
473 | ra->start = offset; | 458 | ra->start = offset; |
474 | ra->size = get_init_ra_size(req_size, max); | 459 | ra->size = get_init_ra_size(req_size, max); |
475 | ra->async_size = ra->size > req_size ? ra->size - req_size : ra->size; | 460 | ra->async_size = ra->size > req_size ? ra->size - req_size : ra->size; |
476 | 461 | ||
477 | readit: | 462 | readit: |
478 | /* | 463 | /* |
479 | * Will this read hit the readahead marker made by itself? | 464 | * Will this read hit the readahead marker made by itself? |
480 | * If so, trigger the readahead marker hit now, and merge | 465 | * If so, trigger the readahead marker hit now, and merge |
481 | * the resulted next readahead window into the current one. | 466 | * the resulted next readahead window into the current one. |
482 | */ | 467 | */ |
483 | if (offset == ra->start && ra->size == ra->async_size) { | 468 | if (offset == ra->start && ra->size == ra->async_size) { |
484 | ra->async_size = get_next_ra_size(ra, max); | 469 | ra->async_size = get_next_ra_size(ra, max); |
485 | ra->size += ra->async_size; | 470 | ra->size += ra->async_size; |
486 | } | 471 | } |
487 | 472 | ||
488 | return ra_submit(ra, mapping, filp); | 473 | return ra_submit(ra, mapping, filp); |
489 | } | 474 | } |
490 | 475 | ||
491 | /** | 476 | /** |
492 | * page_cache_sync_readahead - generic file readahead | 477 | * page_cache_sync_readahead - generic file readahead |
493 | * @mapping: address_space which holds the pagecache and I/O vectors | 478 | * @mapping: address_space which holds the pagecache and I/O vectors |
494 | * @ra: file_ra_state which holds the readahead state | 479 | * @ra: file_ra_state which holds the readahead state |
495 | * @filp: passed on to ->readpage() and ->readpages() | 480 | * @filp: passed on to ->readpage() and ->readpages() |
496 | * @offset: start offset into @mapping, in pagecache page-sized units | 481 | * @offset: start offset into @mapping, in pagecache page-sized units |
497 | * @req_size: hint: total size of the read which the caller is performing in | 482 | * @req_size: hint: total size of the read which the caller is performing in |
498 | * pagecache pages | 483 | * pagecache pages |
499 | * | 484 | * |
500 | * page_cache_sync_readahead() should be called when a cache miss happened: | 485 | * page_cache_sync_readahead() should be called when a cache miss happened: |
501 | * it will submit the read. The readahead logic may decide to piggyback more | 486 | * it will submit the read. The readahead logic may decide to piggyback more |
502 | * pages onto the read request if access patterns suggest it will improve | 487 | * pages onto the read request if access patterns suggest it will improve |
503 | * performance. | 488 | * performance. |
504 | */ | 489 | */ |
505 | void page_cache_sync_readahead(struct address_space *mapping, | 490 | void page_cache_sync_readahead(struct address_space *mapping, |
506 | struct file_ra_state *ra, struct file *filp, | 491 | struct file_ra_state *ra, struct file *filp, |
507 | pgoff_t offset, unsigned long req_size) | 492 | pgoff_t offset, unsigned long req_size) |
508 | { | 493 | { |
509 | /* no read-ahead */ | 494 | /* no read-ahead */ |
510 | if (!ra->ra_pages) | 495 | if (!ra->ra_pages) |
511 | return; | 496 | return; |
512 | 497 | ||
513 | /* be dumb */ | 498 | /* be dumb */ |
514 | if (filp && (filp->f_mode & FMODE_RANDOM)) { | 499 | if (filp && (filp->f_mode & FMODE_RANDOM)) { |
515 | force_page_cache_readahead(mapping, filp, offset, req_size); | 500 | force_page_cache_readahead(mapping, filp, offset, req_size); |
516 | return; | 501 | return; |
517 | } | 502 | } |
518 | 503 | ||
519 | /* do read-ahead */ | 504 | /* do read-ahead */ |
520 | ondemand_readahead(mapping, ra, filp, false, offset, req_size); | 505 | ondemand_readahead(mapping, ra, filp, false, offset, req_size); |
521 | } | 506 | } |
522 | EXPORT_SYMBOL_GPL(page_cache_sync_readahead); | 507 | EXPORT_SYMBOL_GPL(page_cache_sync_readahead); |
523 | 508 | ||
524 | /** | 509 | /** |
525 | * page_cache_async_readahead - file readahead for marked pages | 510 | * page_cache_async_readahead - file readahead for marked pages |
526 | * @mapping: address_space which holds the pagecache and I/O vectors | 511 | * @mapping: address_space which holds the pagecache and I/O vectors |
527 | * @ra: file_ra_state which holds the readahead state | 512 | * @ra: file_ra_state which holds the readahead state |
528 | * @filp: passed on to ->readpage() and ->readpages() | 513 | * @filp: passed on to ->readpage() and ->readpages() |
529 | * @page: the page at @offset which has the PG_readahead flag set | 514 | * @page: the page at @offset which has the PG_readahead flag set |
530 | * @offset: start offset into @mapping, in pagecache page-sized units | 515 | * @offset: start offset into @mapping, in pagecache page-sized units |
531 | * @req_size: hint: total size of the read which the caller is performing in | 516 | * @req_size: hint: total size of the read which the caller is performing in |
532 | * pagecache pages | 517 | * pagecache pages |
533 | * | 518 | * |
534 | * page_cache_async_readahead() should be called when a page is used which | 519 | * page_cache_async_readahead() should be called when a page is used which |
535 | * has the PG_readahead flag; this is a marker to suggest that the application | 520 | * has the PG_readahead flag; this is a marker to suggest that the application |
536 | * has used up enough of the readahead window that we should start pulling in | 521 | * has used up enough of the readahead window that we should start pulling in |
537 | * more pages. | 522 | * more pages. |
538 | */ | 523 | */ |
539 | void | 524 | void |
540 | page_cache_async_readahead(struct address_space *mapping, | 525 | page_cache_async_readahead(struct address_space *mapping, |
541 | struct file_ra_state *ra, struct file *filp, | 526 | struct file_ra_state *ra, struct file *filp, |
542 | struct page *page, pgoff_t offset, | 527 | struct page *page, pgoff_t offset, |
543 | unsigned long req_size) | 528 | unsigned long req_size) |
544 | { | 529 | { |
545 | /* no read-ahead */ | 530 | /* no read-ahead */ |
546 | if (!ra->ra_pages) | 531 | if (!ra->ra_pages) |
547 | return; | 532 | return; |
548 | 533 | ||
549 | /* | 534 | /* |
550 | * Same bit is used for PG_readahead and PG_reclaim. | 535 | * Same bit is used for PG_readahead and PG_reclaim. |
551 | */ | 536 | */ |
552 | if (PageWriteback(page)) | 537 | if (PageWriteback(page)) |
553 | return; | 538 | return; |
554 | 539 | ||
555 | ClearPageReadahead(page); | 540 | ClearPageReadahead(page); |
556 | 541 | ||
557 | /* | 542 | /* |
558 | * Defer asynchronous read-ahead on IO congestion. | 543 | * Defer asynchronous read-ahead on IO congestion. |
559 | */ | 544 | */ |
560 | if (bdi_read_congested(mapping->backing_dev_info)) | 545 | if (bdi_read_congested(mapping->backing_dev_info)) |
561 | return; | 546 | return; |
562 | 547 | ||
563 | /* do read-ahead */ | 548 | /* do read-ahead */ |
564 | ondemand_readahead(mapping, ra, filp, true, offset, req_size); | 549 | ondemand_readahead(mapping, ra, filp, true, offset, req_size); |
565 | } | 550 | } |
566 | EXPORT_SYMBOL_GPL(page_cache_async_readahead); | 551 | EXPORT_SYMBOL_GPL(page_cache_async_readahead); |
567 | 552 | ||
568 | static ssize_t | 553 | static ssize_t |
569 | do_readahead(struct address_space *mapping, struct file *filp, | 554 | do_readahead(struct address_space *mapping, struct file *filp, |
570 | pgoff_t index, unsigned long nr) | 555 | pgoff_t index, unsigned long nr) |
571 | { | 556 | { |
572 | if (!mapping || !mapping->a_ops) | 557 | if (!mapping || !mapping->a_ops) |
573 | return -EINVAL; | 558 | return -EINVAL; |
574 | 559 | ||
575 | return force_page_cache_readahead(mapping, filp, index, nr); | 560 | return force_page_cache_readahead(mapping, filp, index, nr); |
576 | } | 561 | } |
577 | 562 | ||
578 | SYSCALL_DEFINE3(readahead, int, fd, loff_t, offset, size_t, count) | 563 | SYSCALL_DEFINE3(readahead, int, fd, loff_t, offset, size_t, count) |
579 | { | 564 | { |
580 | ssize_t ret; | 565 | ssize_t ret; |
581 | struct fd f; | 566 | struct fd f; |
582 | 567 | ||
583 | ret = -EBADF; | 568 | ret = -EBADF; |
584 | f = fdget(fd); | 569 | f = fdget(fd); |
585 | if (f.file) { | 570 | if (f.file) { |
586 | if (f.file->f_mode & FMODE_READ) { | 571 | if (f.file->f_mode & FMODE_READ) { |
587 | struct address_space *mapping = f.file->f_mapping; | 572 | struct address_space *mapping = f.file->f_mapping; |
588 | pgoff_t start = offset >> PAGE_CACHE_SHIFT; | 573 | pgoff_t start = offset >> PAGE_CACHE_SHIFT; |
589 | pgoff_t end = (offset + count - 1) >> PAGE_CACHE_SHIFT; | 574 | pgoff_t end = (offset + count - 1) >> PAGE_CACHE_SHIFT; |
590 | unsigned long len = end - start + 1; | 575 | unsigned long len = end - start + 1; |
591 | ret = do_readahead(mapping, f.file, start, len); | 576 | ret = do_readahead(mapping, f.file, start, len); |
592 | } | 577 | } |
593 | fdput(f); | 578 | fdput(f); |
594 | } | 579 | } |
595 | return ret; | 580 | return ret; |