Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
1da177e4 LT |
2 | #ifndef _LINUX_MM_H |
3 | #define _LINUX_MM_H | |
4 | ||
1da177e4 LT |
5 | #include <linux/errno.h> |
6 | ||
7 | #ifdef __KERNEL__ | |
8 | ||
309381fe | 9 | #include <linux/mmdebug.h> |
1da177e4 | 10 | #include <linux/gfp.h> |
187f1882 | 11 | #include <linux/bug.h> |
1da177e4 LT |
12 | #include <linux/list.h> |
13 | #include <linux/mmzone.h> | |
14 | #include <linux/rbtree.h> | |
83aeeada | 15 | #include <linux/atomic.h> |
9a11b49a | 16 | #include <linux/debug_locks.h> |
5b99cd0e | 17 | #include <linux/mm_types.h> |
08677214 | 18 | #include <linux/range.h> |
c6f6b596 | 19 | #include <linux/pfn.h> |
3565fce3 | 20 | #include <linux/percpu-refcount.h> |
e9da73d6 | 21 | #include <linux/bit_spinlock.h> |
b0d40c92 | 22 | #include <linux/shrinker.h> |
9c599024 | 23 | #include <linux/resource.h> |
e30825f1 | 24 | #include <linux/page_ext.h> |
8025e5dd | 25 | #include <linux/err.h> |
fe896d18 | 26 | #include <linux/page_ref.h> |
7b2d55d2 | 27 | #include <linux/memremap.h> |
3b3b1a29 | 28 | #include <linux/overflow.h> |
b5420237 | 29 | #include <linux/sizes.h> |
1da177e4 LT |
30 | |
31 | struct mempolicy; | |
32 | struct anon_vma; | |
bf181b9f | 33 | struct anon_vma_chain; |
4e950f6f | 34 | struct file_ra_state; |
e8edc6e0 | 35 | struct user_struct; |
4e950f6f | 36 | struct writeback_control; |
682aa8e1 | 37 | struct bdi_writeback; |
1da177e4 | 38 | |
597b7305 MH |
39 | void init_mm_internals(void); |
40 | ||
fccc9987 | 41 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ |
1da177e4 | 42 | extern unsigned long max_mapnr; |
fccc9987 JL |
43 | |
44 | static inline void set_max_mapnr(unsigned long limit) | |
45 | { | |
46 | max_mapnr = limit; | |
47 | } | |
48 | #else | |
49 | static inline void set_max_mapnr(unsigned long limit) { } | |
1da177e4 LT |
50 | #endif |
51 | ||
ca79b0c2 AK |
52 | extern atomic_long_t _totalram_pages; |
53 | static inline unsigned long totalram_pages(void) | |
54 | { | |
55 | return (unsigned long)atomic_long_read(&_totalram_pages); | |
56 | } | |
57 | ||
58 | static inline void totalram_pages_inc(void) | |
59 | { | |
60 | atomic_long_inc(&_totalram_pages); | |
61 | } | |
62 | ||
63 | static inline void totalram_pages_dec(void) | |
64 | { | |
65 | atomic_long_dec(&_totalram_pages); | |
66 | } | |
67 | ||
68 | static inline void totalram_pages_add(long count) | |
69 | { | |
70 | atomic_long_add(count, &_totalram_pages); | |
71 | } | |
72 | ||
73 | static inline void totalram_pages_set(long val) | |
74 | { | |
75 | atomic_long_set(&_totalram_pages, val); | |
76 | } | |
77 | ||
1da177e4 | 78 | extern void * high_memory; |
1da177e4 LT |
79 | extern int page_cluster; |
80 | ||
81 | #ifdef CONFIG_SYSCTL | |
82 | extern int sysctl_legacy_va_layout; | |
83 | #else | |
84 | #define sysctl_legacy_va_layout 0 | |
85 | #endif | |
86 | ||
d07e2259 DC |
87 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS |
88 | extern const int mmap_rnd_bits_min; | |
89 | extern const int mmap_rnd_bits_max; | |
90 | extern int mmap_rnd_bits __read_mostly; | |
91 | #endif | |
92 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS | |
93 | extern const int mmap_rnd_compat_bits_min; | |
94 | extern const int mmap_rnd_compat_bits_max; | |
95 | extern int mmap_rnd_compat_bits __read_mostly; | |
96 | #endif | |
97 | ||
1da177e4 LT |
98 | #include <asm/page.h> |
99 | #include <asm/pgtable.h> | |
100 | #include <asm/processor.h> | |
1da177e4 | 101 | |
d9344522 AK |
102 | /* |
103 | * Architectures that support memory tagging (assigning tags to memory regions, | |
104 | * embedding these tags into addresses that point to these memory regions, and | |
105 | * checking that the memory and the pointer tags match on memory accesses) | |
106 | * redefine this macro to strip tags from pointers. | |
107 | * It's defined as noop for arcitectures that don't support memory tagging. | |
108 | */ | |
109 | #ifndef untagged_addr | |
110 | #define untagged_addr(addr) (addr) | |
111 | #endif | |
112 | ||
79442ed1 TC |
113 | #ifndef __pa_symbol |
114 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) | |
115 | #endif | |
116 | ||
1dff8083 AB |
117 | #ifndef page_to_virt |
118 | #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x))) | |
119 | #endif | |
120 | ||
568c5fe5 LA |
121 | #ifndef lm_alias |
122 | #define lm_alias(x) __va(__pa_symbol(x)) | |
123 | #endif | |
124 | ||
593befa6 DD |
125 | /* |
126 | * To prevent common memory management code establishing | |
127 | * a zero page mapping on a read fault. | |
128 | * This macro should be defined within <asm/pgtable.h>. | |
129 | * s390 does this to prevent multiplexing of hardware bits | |
130 | * related to the physical page in case of virtualization. | |
131 | */ | |
132 | #ifndef mm_forbids_zeropage | |
133 | #define mm_forbids_zeropage(X) (0) | |
134 | #endif | |
135 | ||
a4a3ede2 PT |
136 | /* |
137 | * On some architectures it is expensive to call memset() for small sizes. | |
5470dea4 AD |
138 | * If an architecture decides to implement their own version of |
139 | * mm_zero_struct_page they should wrap the defines below in a #ifndef and | |
140 | * define their own version of this macro in <asm/pgtable.h> | |
a4a3ede2 | 141 | */ |
5470dea4 AD |
142 | #if BITS_PER_LONG == 64 |
143 | /* This function must be updated when the size of struct page grows above 80 | |
144 | * or reduces below 56. The idea that compiler optimizes out switch() | |
145 | * statement, and only leaves move/store instructions. Also the compiler can | |
146 | * combine write statments if they are both assignments and can be reordered, | |
147 | * this can result in several of the writes here being dropped. | |
148 | */ | |
149 | #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp) | |
150 | static inline void __mm_zero_struct_page(struct page *page) | |
151 | { | |
152 | unsigned long *_pp = (void *)page; | |
153 | ||
154 | /* Check that struct page is either 56, 64, 72, or 80 bytes */ | |
155 | BUILD_BUG_ON(sizeof(struct page) & 7); | |
156 | BUILD_BUG_ON(sizeof(struct page) < 56); | |
157 | BUILD_BUG_ON(sizeof(struct page) > 80); | |
158 | ||
159 | switch (sizeof(struct page)) { | |
160 | case 80: | |
161 | _pp[9] = 0; /* fallthrough */ | |
162 | case 72: | |
163 | _pp[8] = 0; /* fallthrough */ | |
164 | case 64: | |
165 | _pp[7] = 0; /* fallthrough */ | |
166 | case 56: | |
167 | _pp[6] = 0; | |
168 | _pp[5] = 0; | |
169 | _pp[4] = 0; | |
170 | _pp[3] = 0; | |
171 | _pp[2] = 0; | |
172 | _pp[1] = 0; | |
173 | _pp[0] = 0; | |
174 | } | |
175 | } | |
176 | #else | |
a4a3ede2 PT |
177 | #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page))) |
178 | #endif | |
179 | ||
ea606cf5 AR |
180 | /* |
181 | * Default maximum number of active map areas, this limits the number of vmas | |
182 | * per mm struct. Users can overwrite this number by sysctl but there is a | |
183 | * problem. | |
184 | * | |
185 | * When a program's coredump is generated as ELF format, a section is created | |
186 | * per a vma. In ELF, the number of sections is represented in unsigned short. | |
187 | * This means the number of sections should be smaller than 65535 at coredump. | |
188 | * Because the kernel adds some informative sections to a image of program at | |
189 | * generating coredump, we need some margin. The number of extra sections is | |
190 | * 1-3 now and depends on arch. We use "5" as safe margin, here. | |
191 | * | |
192 | * ELF extended numbering allows more than 65535 sections, so 16-bit bound is | |
193 | * not a hard limit any more. Although some userspace tools can be surprised by | |
194 | * that. | |
195 | */ | |
196 | #define MAPCOUNT_ELF_CORE_MARGIN (5) | |
197 | #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) | |
198 | ||
199 | extern int sysctl_max_map_count; | |
200 | ||
c9b1d098 | 201 | extern unsigned long sysctl_user_reserve_kbytes; |
4eeab4f5 | 202 | extern unsigned long sysctl_admin_reserve_kbytes; |
c9b1d098 | 203 | |
49f0ce5f JM |
204 | extern int sysctl_overcommit_memory; |
205 | extern int sysctl_overcommit_ratio; | |
206 | extern unsigned long sysctl_overcommit_kbytes; | |
207 | ||
208 | extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, | |
209 | size_t *, loff_t *); | |
210 | extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, | |
211 | size_t *, loff_t *); | |
212 | ||
1da177e4 LT |
213 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
214 | ||
27ac792c AR |
215 | /* to align the pointer to the (next) page boundary */ |
216 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | |
217 | ||
0fa73b86 | 218 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ |
1061b0d2 | 219 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE) |
0fa73b86 | 220 | |
f86196ea NB |
221 | #define lru_to_page(head) (list_entry((head)->prev, struct page, lru)) |
222 | ||
1da177e4 LT |
223 | /* |
224 | * Linux kernel virtual memory manager primitives. | |
225 | * The idea being to have a "virtual" mm in the same way | |
226 | * we have a virtual fs - giving a cleaner interface to the | |
227 | * mm details, and allowing different kinds of memory mappings | |
228 | * (from shared memory to executable loading to arbitrary | |
229 | * mmap() functions). | |
230 | */ | |
231 | ||
490fc053 | 232 | struct vm_area_struct *vm_area_alloc(struct mm_struct *); |
3928d4f5 LT |
233 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *); |
234 | void vm_area_free(struct vm_area_struct *); | |
c43692e8 | 235 | |
1da177e4 | 236 | #ifndef CONFIG_MMU |
8feae131 DH |
237 | extern struct rb_root nommu_region_tree; |
238 | extern struct rw_semaphore nommu_region_sem; | |
1da177e4 LT |
239 | |
240 | extern unsigned int kobjsize(const void *objp); | |
241 | #endif | |
242 | ||
243 | /* | |
605d9288 | 244 | * vm_flags in vm_area_struct, see mm_types.h. |
bcf66917 | 245 | * When changing, update also include/trace/events/mmflags.h |
1da177e4 | 246 | */ |
cc2383ec KK |
247 | #define VM_NONE 0x00000000 |
248 | ||
1da177e4 LT |
249 | #define VM_READ 0x00000001 /* currently active flags */ |
250 | #define VM_WRITE 0x00000002 | |
251 | #define VM_EXEC 0x00000004 | |
252 | #define VM_SHARED 0x00000008 | |
253 | ||
7e2cff42 | 254 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
1da177e4 LT |
255 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
256 | #define VM_MAYWRITE 0x00000020 | |
257 | #define VM_MAYEXEC 0x00000040 | |
258 | #define VM_MAYSHARE 0x00000080 | |
259 | ||
260 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
16ba6f81 | 261 | #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */ |
6aab341e | 262 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
1da177e4 | 263 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
16ba6f81 | 264 | #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */ |
1da177e4 | 265 | |
1da177e4 LT |
266 | #define VM_LOCKED 0x00002000 |
267 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | |
268 | ||
269 | /* Used by sys_madvise() */ | |
270 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | |
271 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | |
272 | ||
273 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | |
274 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | |
de60f5f1 | 275 | #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */ |
1da177e4 | 276 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
cdfd4325 | 277 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
1da177e4 | 278 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
b6fb293f | 279 | #define VM_SYNC 0x00800000 /* Synchronous page faults */ |
cc2383ec | 280 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ |
d2cd9ede | 281 | #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */ |
0103bd16 | 282 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ |
d00806b1 | 283 | |
d9104d1c CG |
284 | #ifdef CONFIG_MEM_SOFT_DIRTY |
285 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ | |
286 | #else | |
287 | # define VM_SOFTDIRTY 0 | |
288 | #endif | |
289 | ||
b379d790 | 290 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
cc2383ec KK |
291 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ |
292 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ | |
f8af4da3 | 293 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
1da177e4 | 294 | |
63c17fb8 DH |
295 | #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS |
296 | #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */ | |
297 | #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */ | |
298 | #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */ | |
299 | #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */ | |
df3735c5 | 300 | #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */ |
63c17fb8 DH |
301 | #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0) |
302 | #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1) | |
303 | #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2) | |
304 | #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3) | |
df3735c5 | 305 | #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4) |
63c17fb8 DH |
306 | #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */ |
307 | ||
5212213a | 308 | #ifdef CONFIG_ARCH_HAS_PKEYS |
8f62c883 DH |
309 | # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0 |
310 | # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */ | |
2c9e0a6f | 311 | # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */ |
8f62c883 DH |
312 | # define VM_PKEY_BIT2 VM_HIGH_ARCH_2 |
313 | # define VM_PKEY_BIT3 VM_HIGH_ARCH_3 | |
2c9e0a6f RP |
314 | #ifdef CONFIG_PPC |
315 | # define VM_PKEY_BIT4 VM_HIGH_ARCH_4 | |
316 | #else | |
317 | # define VM_PKEY_BIT4 0 | |
8f62c883 | 318 | #endif |
5212213a RP |
319 | #endif /* CONFIG_ARCH_HAS_PKEYS */ |
320 | ||
321 | #if defined(CONFIG_X86) | |
322 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ | |
cc2383ec KK |
323 | #elif defined(CONFIG_PPC) |
324 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ | |
325 | #elif defined(CONFIG_PARISC) | |
326 | # define VM_GROWSUP VM_ARCH_1 | |
327 | #elif defined(CONFIG_IA64) | |
328 | # define VM_GROWSUP VM_ARCH_1 | |
74a04967 KA |
329 | #elif defined(CONFIG_SPARC64) |
330 | # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */ | |
331 | # define VM_ARCH_CLEAR VM_SPARC_ADI | |
cc2383ec KK |
332 | #elif !defined(CONFIG_MMU) |
333 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ | |
334 | #endif | |
335 | ||
df3735c5 | 336 | #if defined(CONFIG_X86_INTEL_MPX) |
4aae7e43 | 337 | /* MPX specific bounds table or bounds directory */ |
fa87b91c | 338 | # define VM_MPX VM_HIGH_ARCH_4 |
df3735c5 RR |
339 | #else |
340 | # define VM_MPX VM_NONE | |
4aae7e43 QR |
341 | #endif |
342 | ||
cc2383ec KK |
343 | #ifndef VM_GROWSUP |
344 | # define VM_GROWSUP VM_NONE | |
345 | #endif | |
346 | ||
a8bef8ff MG |
347 | /* Bits set in the VMA until the stack is in its final location */ |
348 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | |
349 | ||
1da177e4 LT |
350 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
351 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
352 | #endif | |
353 | ||
354 | #ifdef CONFIG_STACK_GROWSUP | |
30bdbb78 | 355 | #define VM_STACK VM_GROWSUP |
1da177e4 | 356 | #else |
30bdbb78 | 357 | #define VM_STACK VM_GROWSDOWN |
1da177e4 LT |
358 | #endif |
359 | ||
30bdbb78 KK |
360 | #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
361 | ||
b291f000 | 362 | /* |
78f11a25 AA |
363 | * Special vmas that are non-mergable, non-mlock()able. |
364 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. | |
b291f000 | 365 | */ |
9050d7eb | 366 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) |
b291f000 | 367 | |
a0715cc2 AT |
368 | /* This mask defines which mm->def_flags a process can inherit its parent */ |
369 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE | |
370 | ||
de60f5f1 EM |
371 | /* This mask is used to clear all the VMA flags used by mlock */ |
372 | #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT)) | |
373 | ||
2c2d57b5 KA |
374 | /* Arch-specific flags to clear when updating VM flags on protection change */ |
375 | #ifndef VM_ARCH_CLEAR | |
376 | # define VM_ARCH_CLEAR VM_NONE | |
377 | #endif | |
378 | #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR) | |
379 | ||
1da177e4 LT |
380 | /* |
381 | * mapping from the currently active vm_flags protection bits (the | |
382 | * low four bits) to a page protection mask.. | |
383 | */ | |
384 | extern pgprot_t protection_map[16]; | |
385 | ||
d0217ac0 | 386 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ |
9b4bdd2f KS |
387 | #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */ |
388 | #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */ | |
389 | #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */ | |
390 | #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */ | |
391 | #define FAULT_FLAG_TRIED 0x20 /* Second try */ | |
392 | #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */ | |
1b2ee126 | 393 | #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */ |
d61172b4 | 394 | #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */ |
d0217ac0 | 395 | |
282a8e03 RZ |
396 | #define FAULT_FLAG_TRACE \ |
397 | { FAULT_FLAG_WRITE, "WRITE" }, \ | |
398 | { FAULT_FLAG_MKWRITE, "MKWRITE" }, \ | |
399 | { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \ | |
400 | { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \ | |
401 | { FAULT_FLAG_KILLABLE, "KILLABLE" }, \ | |
402 | { FAULT_FLAG_TRIED, "TRIED" }, \ | |
403 | { FAULT_FLAG_USER, "USER" }, \ | |
404 | { FAULT_FLAG_REMOTE, "REMOTE" }, \ | |
405 | { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" } | |
406 | ||
54cb8821 | 407 | /* |
d0217ac0 | 408 | * vm_fault is filled by the the pagefault handler and passed to the vma's |
83c54070 NP |
409 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
410 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
54cb8821 | 411 | * |
c20cd45e MH |
412 | * MM layer fills up gfp_mask for page allocations but fault handler might |
413 | * alter it if its implementation requires a different allocation context. | |
414 | * | |
9b4bdd2f | 415 | * pgoff should be used in favour of virtual_address, if possible. |
54cb8821 | 416 | */ |
d0217ac0 | 417 | struct vm_fault { |
82b0f8c3 | 418 | struct vm_area_struct *vma; /* Target VMA */ |
d0217ac0 | 419 | unsigned int flags; /* FAULT_FLAG_xxx flags */ |
c20cd45e | 420 | gfp_t gfp_mask; /* gfp mask to be used for allocations */ |
d0217ac0 | 421 | pgoff_t pgoff; /* Logical page offset based on vma */ |
82b0f8c3 | 422 | unsigned long address; /* Faulting virtual address */ |
82b0f8c3 | 423 | pmd_t *pmd; /* Pointer to pmd entry matching |
2994302b | 424 | * the 'address' */ |
a2d58167 DJ |
425 | pud_t *pud; /* Pointer to pud entry matching |
426 | * the 'address' | |
427 | */ | |
2994302b | 428 | pte_t orig_pte; /* Value of PTE at the time of fault */ |
d0217ac0 | 429 | |
3917048d JK |
430 | struct page *cow_page; /* Page handler may use for COW fault */ |
431 | struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */ | |
d0217ac0 | 432 | struct page *page; /* ->fault handlers should return a |
83c54070 | 433 | * page here, unless VM_FAULT_NOPAGE |
d0217ac0 | 434 | * is set (which is also implied by |
83c54070 | 435 | * VM_FAULT_ERROR). |
d0217ac0 | 436 | */ |
82b0f8c3 | 437 | /* These three entries are valid only while holding ptl lock */ |
bae473a4 KS |
438 | pte_t *pte; /* Pointer to pte entry matching |
439 | * the 'address'. NULL if the page | |
440 | * table hasn't been allocated. | |
441 | */ | |
442 | spinlock_t *ptl; /* Page table lock. | |
443 | * Protects pte page table if 'pte' | |
444 | * is not NULL, otherwise pmd. | |
445 | */ | |
7267ec00 KS |
446 | pgtable_t prealloc_pte; /* Pre-allocated pte page table. |
447 | * vm_ops->map_pages() calls | |
448 | * alloc_set_pte() from atomic context. | |
449 | * do_fault_around() pre-allocates | |
450 | * page table to avoid allocation from | |
451 | * atomic context. | |
452 | */ | |
54cb8821 | 453 | }; |
1da177e4 | 454 | |
c791ace1 DJ |
455 | /* page entry size for vm->huge_fault() */ |
456 | enum page_entry_size { | |
457 | PE_SIZE_PTE = 0, | |
458 | PE_SIZE_PMD, | |
459 | PE_SIZE_PUD, | |
460 | }; | |
461 | ||
1da177e4 LT |
462 | /* |
463 | * These are the virtual MM functions - opening of an area, closing and | |
464 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
27d036e3 | 465 | * to the functions called when a no-page or a wp-page exception occurs. |
1da177e4 LT |
466 | */ |
467 | struct vm_operations_struct { | |
468 | void (*open)(struct vm_area_struct * area); | |
469 | void (*close)(struct vm_area_struct * area); | |
31383c68 | 470 | int (*split)(struct vm_area_struct * area, unsigned long addr); |
5477e70a | 471 | int (*mremap)(struct vm_area_struct * area); |
1c8f4220 SJ |
472 | vm_fault_t (*fault)(struct vm_fault *vmf); |
473 | vm_fault_t (*huge_fault)(struct vm_fault *vmf, | |
474 | enum page_entry_size pe_size); | |
82b0f8c3 | 475 | void (*map_pages)(struct vm_fault *vmf, |
bae473a4 | 476 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
05ea8860 | 477 | unsigned long (*pagesize)(struct vm_area_struct * area); |
9637a5ef DH |
478 | |
479 | /* notification that a previously read-only page is about to become | |
480 | * writable, if an error is returned it will cause a SIGBUS */ | |
1c8f4220 | 481 | vm_fault_t (*page_mkwrite)(struct vm_fault *vmf); |
28b2ee20 | 482 | |
dd906184 | 483 | /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ |
1c8f4220 | 484 | vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf); |
dd906184 | 485 | |
28b2ee20 RR |
486 | /* called by access_process_vm when get_user_pages() fails, typically |
487 | * for use by special VMAs that can switch between memory and hardware | |
488 | */ | |
489 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | |
490 | void *buf, int len, int write); | |
78d683e8 AL |
491 | |
492 | /* Called by the /proc/PID/maps code to ask the vma whether it | |
493 | * has a special name. Returning non-NULL will also cause this | |
494 | * vma to be dumped unconditionally. */ | |
495 | const char *(*name)(struct vm_area_struct *vma); | |
496 | ||
1da177e4 | 497 | #ifdef CONFIG_NUMA |
a6020ed7 LS |
498 | /* |
499 | * set_policy() op must add a reference to any non-NULL @new mempolicy | |
500 | * to hold the policy upon return. Caller should pass NULL @new to | |
501 | * remove a policy and fall back to surrounding context--i.e. do not | |
502 | * install a MPOL_DEFAULT policy, nor the task or system default | |
503 | * mempolicy. | |
504 | */ | |
1da177e4 | 505 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
a6020ed7 LS |
506 | |
507 | /* | |
508 | * get_policy() op must add reference [mpol_get()] to any policy at | |
509 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | |
510 | * in mm/mempolicy.c will do this automatically. | |
511 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | |
512 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. | |
513 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op | |
514 | * must return NULL--i.e., do not "fallback" to task or system default | |
515 | * policy. | |
516 | */ | |
1da177e4 LT |
517 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
518 | unsigned long addr); | |
519 | #endif | |
667a0a06 DV |
520 | /* |
521 | * Called by vm_normal_page() for special PTEs to find the | |
522 | * page for @addr. This is useful if the default behavior | |
523 | * (using pte_page()) would not find the correct page. | |
524 | */ | |
525 | struct page *(*find_special_page)(struct vm_area_struct *vma, | |
526 | unsigned long addr); | |
1da177e4 LT |
527 | }; |
528 | ||
027232da KS |
529 | static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm) |
530 | { | |
bfd40eaf KS |
531 | static const struct vm_operations_struct dummy_vm_ops = {}; |
532 | ||
a670468f | 533 | memset(vma, 0, sizeof(*vma)); |
027232da | 534 | vma->vm_mm = mm; |
bfd40eaf | 535 | vma->vm_ops = &dummy_vm_ops; |
027232da KS |
536 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
537 | } | |
538 | ||
bfd40eaf KS |
539 | static inline void vma_set_anonymous(struct vm_area_struct *vma) |
540 | { | |
541 | vma->vm_ops = NULL; | |
542 | } | |
543 | ||
8b11ec1b LT |
544 | /* flush_tlb_range() takes a vma, not a mm, and can care about flags */ |
545 | #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) } | |
546 | ||
1da177e4 LT |
547 | struct mmu_gather; |
548 | struct inode; | |
549 | ||
5c7fb56e DW |
550 | #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE) |
551 | static inline int pmd_devmap(pmd_t pmd) | |
552 | { | |
553 | return 0; | |
554 | } | |
a00cc7d9 MW |
555 | static inline int pud_devmap(pud_t pud) |
556 | { | |
557 | return 0; | |
558 | } | |
b59f65fa KS |
559 | static inline int pgd_devmap(pgd_t pgd) |
560 | { | |
561 | return 0; | |
562 | } | |
5c7fb56e DW |
563 | #endif |
564 | ||
1da177e4 LT |
565 | /* |
566 | * FIXME: take this include out, include page-flags.h in | |
567 | * files which need it (119 of them) | |
568 | */ | |
569 | #include <linux/page-flags.h> | |
71e3aac0 | 570 | #include <linux/huge_mm.h> |
1da177e4 LT |
571 | |
572 | /* | |
573 | * Methods to modify the page usage count. | |
574 | * | |
575 | * What counts for a page usage: | |
576 | * - cache mapping (page->mapping) | |
577 | * - private data (page->private) | |
578 | * - page mapped in a task's page tables, each mapping | |
579 | * is counted separately | |
580 | * | |
581 | * Also, many kernel routines increase the page count before a critical | |
582 | * routine so they can be sure the page doesn't go away from under them. | |
1da177e4 LT |
583 | */ |
584 | ||
585 | /* | |
da6052f7 | 586 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
1da177e4 | 587 | */ |
7c8ee9a8 NP |
588 | static inline int put_page_testzero(struct page *page) |
589 | { | |
fe896d18 JK |
590 | VM_BUG_ON_PAGE(page_ref_count(page) == 0, page); |
591 | return page_ref_dec_and_test(page); | |
7c8ee9a8 | 592 | } |
1da177e4 LT |
593 | |
594 | /* | |
7c8ee9a8 NP |
595 | * Try to grab a ref unless the page has a refcount of zero, return false if |
596 | * that is the case. | |
8e0861fa AK |
597 | * This can be called when MMU is off so it must not access |
598 | * any of the virtual mappings. | |
1da177e4 | 599 | */ |
7c8ee9a8 NP |
600 | static inline int get_page_unless_zero(struct page *page) |
601 | { | |
fe896d18 | 602 | return page_ref_add_unless(page, 1, 0); |
7c8ee9a8 | 603 | } |
1da177e4 | 604 | |
53df8fdc | 605 | extern int page_is_ram(unsigned long pfn); |
124fe20d DW |
606 | |
607 | enum { | |
608 | REGION_INTERSECTS, | |
609 | REGION_DISJOINT, | |
610 | REGION_MIXED, | |
611 | }; | |
612 | ||
1c29f25b TK |
613 | int region_intersects(resource_size_t offset, size_t size, unsigned long flags, |
614 | unsigned long desc); | |
53df8fdc | 615 | |
48667e7a | 616 | /* Support for virtually mapped pages */ |
b3bdda02 CL |
617 | struct page *vmalloc_to_page(const void *addr); |
618 | unsigned long vmalloc_to_pfn(const void *addr); | |
48667e7a | 619 | |
0738c4bb PM |
620 | /* |
621 | * Determine if an address is within the vmalloc range | |
622 | * | |
623 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | |
624 | * is no special casing required. | |
625 | */ | |
bb00a789 | 626 | static inline bool is_vmalloc_addr(const void *x) |
9e2779fa | 627 | { |
0738c4bb | 628 | #ifdef CONFIG_MMU |
9e2779fa CL |
629 | unsigned long addr = (unsigned long)x; |
630 | ||
631 | return addr >= VMALLOC_START && addr < VMALLOC_END; | |
0738c4bb | 632 | #else |
bb00a789 | 633 | return false; |
8ca3ed87 | 634 | #endif |
0738c4bb | 635 | } |
9bd3bb67 AK |
636 | |
637 | #ifndef is_ioremap_addr | |
638 | #define is_ioremap_addr(x) is_vmalloc_addr(x) | |
639 | #endif | |
640 | ||
81ac3ad9 KH |
641 | #ifdef CONFIG_MMU |
642 | extern int is_vmalloc_or_module_addr(const void *x); | |
643 | #else | |
934831d0 | 644 | static inline int is_vmalloc_or_module_addr(const void *x) |
81ac3ad9 KH |
645 | { |
646 | return 0; | |
647 | } | |
648 | #endif | |
9e2779fa | 649 | |
a7c3e901 MH |
650 | extern void *kvmalloc_node(size_t size, gfp_t flags, int node); |
651 | static inline void *kvmalloc(size_t size, gfp_t flags) | |
652 | { | |
653 | return kvmalloc_node(size, flags, NUMA_NO_NODE); | |
654 | } | |
655 | static inline void *kvzalloc_node(size_t size, gfp_t flags, int node) | |
656 | { | |
657 | return kvmalloc_node(size, flags | __GFP_ZERO, node); | |
658 | } | |
659 | static inline void *kvzalloc(size_t size, gfp_t flags) | |
660 | { | |
661 | return kvmalloc(size, flags | __GFP_ZERO); | |
662 | } | |
663 | ||
752ade68 MH |
664 | static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags) |
665 | { | |
3b3b1a29 KC |
666 | size_t bytes; |
667 | ||
668 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
752ade68 MH |
669 | return NULL; |
670 | ||
3b3b1a29 | 671 | return kvmalloc(bytes, flags); |
752ade68 MH |
672 | } |
673 | ||
1c542f38 KC |
674 | static inline void *kvcalloc(size_t n, size_t size, gfp_t flags) |
675 | { | |
676 | return kvmalloc_array(n, size, flags | __GFP_ZERO); | |
677 | } | |
678 | ||
39f1f78d AV |
679 | extern void kvfree(const void *addr); |
680 | ||
53f9263b KS |
681 | static inline atomic_t *compound_mapcount_ptr(struct page *page) |
682 | { | |
683 | return &page[1].compound_mapcount; | |
684 | } | |
685 | ||
686 | static inline int compound_mapcount(struct page *page) | |
687 | { | |
5f527c2b | 688 | VM_BUG_ON_PAGE(!PageCompound(page), page); |
53f9263b KS |
689 | page = compound_head(page); |
690 | return atomic_read(compound_mapcount_ptr(page)) + 1; | |
691 | } | |
692 | ||
70b50f94 AA |
693 | /* |
694 | * The atomic page->_mapcount, starts from -1: so that transitions | |
695 | * both from it and to it can be tracked, using atomic_inc_and_test | |
696 | * and atomic_add_negative(-1). | |
697 | */ | |
22b751c3 | 698 | static inline void page_mapcount_reset(struct page *page) |
70b50f94 AA |
699 | { |
700 | atomic_set(&(page)->_mapcount, -1); | |
701 | } | |
702 | ||
b20ce5e0 KS |
703 | int __page_mapcount(struct page *page); |
704 | ||
70b50f94 AA |
705 | static inline int page_mapcount(struct page *page) |
706 | { | |
1d148e21 | 707 | VM_BUG_ON_PAGE(PageSlab(page), page); |
53f9263b | 708 | |
b20ce5e0 KS |
709 | if (unlikely(PageCompound(page))) |
710 | return __page_mapcount(page); | |
711 | return atomic_read(&page->_mapcount) + 1; | |
712 | } | |
713 | ||
714 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
715 | int total_mapcount(struct page *page); | |
6d0a07ed | 716 | int page_trans_huge_mapcount(struct page *page, int *total_mapcount); |
b20ce5e0 KS |
717 | #else |
718 | static inline int total_mapcount(struct page *page) | |
719 | { | |
720 | return page_mapcount(page); | |
70b50f94 | 721 | } |
6d0a07ed AA |
722 | static inline int page_trans_huge_mapcount(struct page *page, |
723 | int *total_mapcount) | |
724 | { | |
725 | int mapcount = page_mapcount(page); | |
726 | if (total_mapcount) | |
727 | *total_mapcount = mapcount; | |
728 | return mapcount; | |
729 | } | |
b20ce5e0 | 730 | #endif |
70b50f94 | 731 | |
b49af68f CL |
732 | static inline struct page *virt_to_head_page(const void *x) |
733 | { | |
734 | struct page *page = virt_to_page(x); | |
ccaafd7f | 735 | |
1d798ca3 | 736 | return compound_head(page); |
b49af68f CL |
737 | } |
738 | ||
ddc58f27 KS |
739 | void __put_page(struct page *page); |
740 | ||
1d7ea732 | 741 | void put_pages_list(struct list_head *pages); |
1da177e4 | 742 | |
8dfcc9ba | 743 | void split_page(struct page *page, unsigned int order); |
8dfcc9ba | 744 | |
33f2ef89 AW |
745 | /* |
746 | * Compound pages have a destructor function. Provide a | |
747 | * prototype for that function and accessor functions. | |
f1e61557 | 748 | * These are _only_ valid on the head of a compound page. |
33f2ef89 | 749 | */ |
f1e61557 KS |
750 | typedef void compound_page_dtor(struct page *); |
751 | ||
752 | /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */ | |
753 | enum compound_dtor_id { | |
754 | NULL_COMPOUND_DTOR, | |
755 | COMPOUND_PAGE_DTOR, | |
756 | #ifdef CONFIG_HUGETLB_PAGE | |
757 | HUGETLB_PAGE_DTOR, | |
9a982250 KS |
758 | #endif |
759 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
760 | TRANSHUGE_PAGE_DTOR, | |
f1e61557 KS |
761 | #endif |
762 | NR_COMPOUND_DTORS, | |
763 | }; | |
764 | extern compound_page_dtor * const compound_page_dtors[]; | |
33f2ef89 AW |
765 | |
766 | static inline void set_compound_page_dtor(struct page *page, | |
f1e61557 | 767 | enum compound_dtor_id compound_dtor) |
33f2ef89 | 768 | { |
f1e61557 KS |
769 | VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page); |
770 | page[1].compound_dtor = compound_dtor; | |
33f2ef89 AW |
771 | } |
772 | ||
773 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | |
774 | { | |
f1e61557 KS |
775 | VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page); |
776 | return compound_page_dtors[page[1].compound_dtor]; | |
33f2ef89 AW |
777 | } |
778 | ||
d00181b9 | 779 | static inline unsigned int compound_order(struct page *page) |
d85f3385 | 780 | { |
6d777953 | 781 | if (!PageHead(page)) |
d85f3385 | 782 | return 0; |
e4b294c2 | 783 | return page[1].compound_order; |
d85f3385 CL |
784 | } |
785 | ||
f1e61557 | 786 | static inline void set_compound_order(struct page *page, unsigned int order) |
d85f3385 | 787 | { |
e4b294c2 | 788 | page[1].compound_order = order; |
d85f3385 CL |
789 | } |
790 | ||
9a982250 KS |
791 | void free_compound_page(struct page *page); |
792 | ||
3dece370 | 793 | #ifdef CONFIG_MMU |
14fd403f AA |
794 | /* |
795 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | |
796 | * servicing faults for write access. In the normal case, do always want | |
797 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | |
798 | * that do not have writing enabled, when used by access_process_vm. | |
799 | */ | |
800 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | |
801 | { | |
802 | if (likely(vma->vm_flags & VM_WRITE)) | |
803 | pte = pte_mkwrite(pte); | |
804 | return pte; | |
805 | } | |
8c6e50b0 | 806 | |
2b740303 | 807 | vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg, |
7267ec00 | 808 | struct page *page); |
2b740303 SJ |
809 | vm_fault_t finish_fault(struct vm_fault *vmf); |
810 | vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf); | |
3dece370 | 811 | #endif |
14fd403f | 812 | |
1da177e4 LT |
813 | /* |
814 | * Multiple processes may "see" the same page. E.g. for untouched | |
815 | * mappings of /dev/null, all processes see the same page full of | |
816 | * zeroes, and text pages of executables and shared libraries have | |
817 | * only one copy in memory, at most, normally. | |
818 | * | |
819 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
7e871b6c PBG |
820 | * page_count() == 0 means the page is free. page->lru is then used for |
821 | * freelist management in the buddy allocator. | |
da6052f7 | 822 | * page_count() > 0 means the page has been allocated. |
1da177e4 | 823 | * |
da6052f7 NP |
824 | * Pages are allocated by the slab allocator in order to provide memory |
825 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
826 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
827 | * unless a particular usage is carefully commented. (the responsibility of | |
828 | * freeing the kmalloc memory is the caller's, of course). | |
1da177e4 | 829 | * |
da6052f7 NP |
830 | * A page may be used by anyone else who does a __get_free_page(). |
831 | * In this case, page_count still tracks the references, and should only | |
832 | * be used through the normal accessor functions. The top bits of page->flags | |
833 | * and page->virtual store page management information, but all other fields | |
834 | * are unused and could be used privately, carefully. The management of this | |
835 | * page is the responsibility of the one who allocated it, and those who have | |
836 | * subsequently been given references to it. | |
837 | * | |
838 | * The other pages (we may call them "pagecache pages") are completely | |
1da177e4 LT |
839 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
840 | * The following discussion applies only to them. | |
841 | * | |
da6052f7 NP |
842 | * A pagecache page contains an opaque `private' member, which belongs to the |
843 | * page's address_space. Usually, this is the address of a circular list of | |
844 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
845 | * into the filesystem to release these pages. | |
1da177e4 | 846 | * |
da6052f7 NP |
847 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
848 | * is the pointer to the inode, and page->index is the file offset of the page, | |
ea1754a0 | 849 | * in units of PAGE_SIZE. |
1da177e4 | 850 | * |
da6052f7 NP |
851 | * If pagecache pages are not associated with an inode, they are said to be |
852 | * anonymous pages. These may become associated with the swapcache, and in that | |
853 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
1da177e4 | 854 | * |
da6052f7 NP |
855 | * In either case (swapcache or inode backed), the pagecache itself holds one |
856 | * reference to the page. Setting PG_private should also increment the | |
857 | * refcount. The each user mapping also has a reference to the page. | |
1da177e4 | 858 | * |
da6052f7 | 859 | * The pagecache pages are stored in a per-mapping radix tree, which is |
b93b0163 | 860 | * rooted at mapping->i_pages, and indexed by offset. |
da6052f7 NP |
861 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
862 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
1da177e4 | 863 | * |
da6052f7 | 864 | * All pagecache pages may be subject to I/O: |
1da177e4 LT |
865 | * - inode pages may need to be read from disk, |
866 | * - inode pages which have been modified and are MAP_SHARED may need | |
da6052f7 NP |
867 | * to be written back to the inode on disk, |
868 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
869 | * modified may need to be swapped out to swap space and (later) to be read | |
870 | * back into memory. | |
1da177e4 LT |
871 | */ |
872 | ||
873 | /* | |
874 | * The zone field is never updated after free_area_init_core() | |
875 | * sets it, so none of the operations on it need to be atomic. | |
1da177e4 | 876 | */ |
348f8b6c | 877 | |
90572890 | 878 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ |
07808b74 | 879 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
d41dee36 AW |
880 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
881 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
90572890 | 882 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) |
2813b9c0 | 883 | #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH) |
d41dee36 | 884 | |
348f8b6c | 885 | /* |
25985edc | 886 | * Define the bit shifts to access each section. For non-existent |
348f8b6c DH |
887 | * sections we define the shift as 0; that plus a 0 mask ensures |
888 | * the compiler will optimise away reference to them. | |
889 | */ | |
d41dee36 AW |
890 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
891 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
892 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
90572890 | 893 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) |
2813b9c0 | 894 | #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0)) |
348f8b6c | 895 | |
bce54bbf WD |
896 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ |
897 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
89689ae7 | 898 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
899 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
900 | SECTIONS_PGOFF : ZONES_PGOFF) | |
d41dee36 | 901 | #else |
89689ae7 | 902 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
903 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
904 | NODES_PGOFF : ZONES_PGOFF) | |
89689ae7 CL |
905 | #endif |
906 | ||
bd8029b6 | 907 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
348f8b6c | 908 | |
9223b419 CL |
909 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
910 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
348f8b6c DH |
911 | #endif |
912 | ||
d41dee36 AW |
913 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
914 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
915 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
834a964a | 916 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1) |
2813b9c0 | 917 | #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1) |
89689ae7 | 918 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
348f8b6c | 919 | |
33dd4e0e | 920 | static inline enum zone_type page_zonenum(const struct page *page) |
1da177e4 | 921 | { |
348f8b6c | 922 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
1da177e4 | 923 | } |
1da177e4 | 924 | |
260ae3f7 DW |
925 | #ifdef CONFIG_ZONE_DEVICE |
926 | static inline bool is_zone_device_page(const struct page *page) | |
927 | { | |
928 | return page_zonenum(page) == ZONE_DEVICE; | |
929 | } | |
966cf44f AD |
930 | extern void memmap_init_zone_device(struct zone *, unsigned long, |
931 | unsigned long, struct dev_pagemap *); | |
260ae3f7 DW |
932 | #else |
933 | static inline bool is_zone_device_page(const struct page *page) | |
934 | { | |
935 | return false; | |
936 | } | |
7b2d55d2 | 937 | #endif |
5042db43 | 938 | |
e7638488 DW |
939 | #ifdef CONFIG_DEV_PAGEMAP_OPS |
940 | void dev_pagemap_get_ops(void); | |
941 | void dev_pagemap_put_ops(void); | |
942 | void __put_devmap_managed_page(struct page *page); | |
943 | DECLARE_STATIC_KEY_FALSE(devmap_managed_key); | |
944 | static inline bool put_devmap_managed_page(struct page *page) | |
945 | { | |
946 | if (!static_branch_unlikely(&devmap_managed_key)) | |
947 | return false; | |
948 | if (!is_zone_device_page(page)) | |
949 | return false; | |
950 | switch (page->pgmap->type) { | |
951 | case MEMORY_DEVICE_PRIVATE: | |
952 | case MEMORY_DEVICE_PUBLIC: | |
953 | case MEMORY_DEVICE_FS_DAX: | |
954 | __put_devmap_managed_page(page); | |
955 | return true; | |
956 | default: | |
957 | break; | |
958 | } | |
959 | return false; | |
960 | } | |
961 | ||
962 | static inline bool is_device_private_page(const struct page *page) | |
963 | { | |
964 | return is_zone_device_page(page) && | |
965 | page->pgmap->type == MEMORY_DEVICE_PRIVATE; | |
966 | } | |
967 | ||
968 | static inline bool is_device_public_page(const struct page *page) | |
969 | { | |
970 | return is_zone_device_page(page) && | |
971 | page->pgmap->type == MEMORY_DEVICE_PUBLIC; | |
972 | } | |
973 | ||
52916982 LG |
974 | #ifdef CONFIG_PCI_P2PDMA |
975 | static inline bool is_pci_p2pdma_page(const struct page *page) | |
976 | { | |
977 | return is_zone_device_page(page) && | |
978 | page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA; | |
979 | } | |
980 | #else /* CONFIG_PCI_P2PDMA */ | |
981 | static inline bool is_pci_p2pdma_page(const struct page *page) | |
982 | { | |
983 | return false; | |
984 | } | |
985 | #endif /* CONFIG_PCI_P2PDMA */ | |
986 | ||
e7638488 DW |
987 | #else /* CONFIG_DEV_PAGEMAP_OPS */ |
988 | static inline void dev_pagemap_get_ops(void) | |
5042db43 | 989 | { |
5042db43 | 990 | } |
e7638488 DW |
991 | |
992 | static inline void dev_pagemap_put_ops(void) | |
993 | { | |
994 | } | |
995 | ||
996 | static inline bool put_devmap_managed_page(struct page *page) | |
997 | { | |
998 | return false; | |
999 | } | |
1000 | ||
6b368cd4 JG |
1001 | static inline bool is_device_private_page(const struct page *page) |
1002 | { | |
1003 | return false; | |
1004 | } | |
e7638488 | 1005 | |
6b368cd4 JG |
1006 | static inline bool is_device_public_page(const struct page *page) |
1007 | { | |
1008 | return false; | |
1009 | } | |
52916982 LG |
1010 | |
1011 | static inline bool is_pci_p2pdma_page(const struct page *page) | |
1012 | { | |
1013 | return false; | |
1014 | } | |
e7638488 | 1015 | #endif /* CONFIG_DEV_PAGEMAP_OPS */ |
7b2d55d2 | 1016 | |
f958d7b5 LT |
1017 | /* 127: arbitrary random number, small enough to assemble well */ |
1018 | #define page_ref_zero_or_close_to_overflow(page) \ | |
1019 | ((unsigned int) page_ref_count(page) + 127u <= 127u) | |
1020 | ||
3565fce3 DW |
1021 | static inline void get_page(struct page *page) |
1022 | { | |
1023 | page = compound_head(page); | |
1024 | /* | |
1025 | * Getting a normal page or the head of a compound page | |
0139aa7b | 1026 | * requires to already have an elevated page->_refcount. |
3565fce3 | 1027 | */ |
f958d7b5 | 1028 | VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page); |
fe896d18 | 1029 | page_ref_inc(page); |
3565fce3 DW |
1030 | } |
1031 | ||
88b1a17d LT |
1032 | static inline __must_check bool try_get_page(struct page *page) |
1033 | { | |
1034 | page = compound_head(page); | |
1035 | if (WARN_ON_ONCE(page_ref_count(page) <= 0)) | |
1036 | return false; | |
fe896d18 | 1037 | page_ref_inc(page); |
88b1a17d | 1038 | return true; |
3565fce3 DW |
1039 | } |
1040 | ||
1041 | static inline void put_page(struct page *page) | |
1042 | { | |
1043 | page = compound_head(page); | |
1044 | ||
7b2d55d2 | 1045 | /* |
e7638488 DW |
1046 | * For devmap managed pages we need to catch refcount transition from |
1047 | * 2 to 1, when refcount reach one it means the page is free and we | |
1048 | * need to inform the device driver through callback. See | |
7b2d55d2 JG |
1049 | * include/linux/memremap.h and HMM for details. |
1050 | */ | |
e7638488 | 1051 | if (put_devmap_managed_page(page)) |
7b2d55d2 | 1052 | return; |
7b2d55d2 | 1053 | |
3565fce3 DW |
1054 | if (put_page_testzero(page)) |
1055 | __put_page(page); | |
3565fce3 DW |
1056 | } |
1057 | ||
fc1d8e7c JH |
1058 | /** |
1059 | * put_user_page() - release a gup-pinned page | |
1060 | * @page: pointer to page to be released | |
1061 | * | |
1062 | * Pages that were pinned via get_user_pages*() must be released via | |
1063 | * either put_user_page(), or one of the put_user_pages*() routines | |
1064 | * below. This is so that eventually, pages that are pinned via | |
1065 | * get_user_pages*() can be separately tracked and uniquely handled. In | |
1066 | * particular, interactions with RDMA and filesystems need special | |
1067 | * handling. | |
1068 | * | |
1069 | * put_user_page() and put_page() are not interchangeable, despite this early | |
1070 | * implementation that makes them look the same. put_user_page() calls must | |
1071 | * be perfectly matched up with get_user_page() calls. | |
1072 | */ | |
1073 | static inline void put_user_page(struct page *page) | |
1074 | { | |
1075 | put_page(page); | |
1076 | } | |
1077 | ||
1078 | void put_user_pages_dirty(struct page **pages, unsigned long npages); | |
1079 | void put_user_pages_dirty_lock(struct page **pages, unsigned long npages); | |
1080 | void put_user_pages(struct page **pages, unsigned long npages); | |
1081 | ||
9127ab4f CS |
1082 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
1083 | #define SECTION_IN_PAGE_FLAGS | |
1084 | #endif | |
1085 | ||
89689ae7 | 1086 | /* |
7a8010cd VB |
1087 | * The identification function is mainly used by the buddy allocator for |
1088 | * determining if two pages could be buddies. We are not really identifying | |
1089 | * the zone since we could be using the section number id if we do not have | |
1090 | * node id available in page flags. | |
1091 | * We only guarantee that it will return the same value for two combinable | |
1092 | * pages in a zone. | |
89689ae7 | 1093 | */ |
cb2b95e1 AW |
1094 | static inline int page_zone_id(struct page *page) |
1095 | { | |
89689ae7 | 1096 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
348f8b6c DH |
1097 | } |
1098 | ||
89689ae7 | 1099 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
33dd4e0e | 1100 | extern int page_to_nid(const struct page *page); |
89689ae7 | 1101 | #else |
33dd4e0e | 1102 | static inline int page_to_nid(const struct page *page) |
d41dee36 | 1103 | { |
f165b378 PT |
1104 | struct page *p = (struct page *)page; |
1105 | ||
1106 | return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK; | |
d41dee36 | 1107 | } |
89689ae7 CL |
1108 | #endif |
1109 | ||
57e0a030 | 1110 | #ifdef CONFIG_NUMA_BALANCING |
90572890 | 1111 | static inline int cpu_pid_to_cpupid(int cpu, int pid) |
57e0a030 | 1112 | { |
90572890 | 1113 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); |
57e0a030 MG |
1114 | } |
1115 | ||
90572890 | 1116 | static inline int cpupid_to_pid(int cpupid) |
57e0a030 | 1117 | { |
90572890 | 1118 | return cpupid & LAST__PID_MASK; |
57e0a030 | 1119 | } |
b795854b | 1120 | |
90572890 | 1121 | static inline int cpupid_to_cpu(int cpupid) |
b795854b | 1122 | { |
90572890 | 1123 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; |
b795854b MG |
1124 | } |
1125 | ||
90572890 | 1126 | static inline int cpupid_to_nid(int cpupid) |
b795854b | 1127 | { |
90572890 | 1128 | return cpu_to_node(cpupid_to_cpu(cpupid)); |
b795854b MG |
1129 | } |
1130 | ||
90572890 | 1131 | static inline bool cpupid_pid_unset(int cpupid) |
57e0a030 | 1132 | { |
90572890 | 1133 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); |
b795854b MG |
1134 | } |
1135 | ||
90572890 | 1136 | static inline bool cpupid_cpu_unset(int cpupid) |
b795854b | 1137 | { |
90572890 | 1138 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); |
b795854b MG |
1139 | } |
1140 | ||
8c8a743c PZ |
1141 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) |
1142 | { | |
1143 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); | |
1144 | } | |
1145 | ||
1146 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) | |
90572890 PZ |
1147 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
1148 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
b795854b | 1149 | { |
1ae71d03 | 1150 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); |
b795854b | 1151 | } |
90572890 PZ |
1152 | |
1153 | static inline int page_cpupid_last(struct page *page) | |
1154 | { | |
1155 | return page->_last_cpupid; | |
1156 | } | |
1157 | static inline void page_cpupid_reset_last(struct page *page) | |
b795854b | 1158 | { |
1ae71d03 | 1159 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; |
57e0a030 MG |
1160 | } |
1161 | #else | |
90572890 | 1162 | static inline int page_cpupid_last(struct page *page) |
75980e97 | 1163 | { |
90572890 | 1164 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; |
75980e97 PZ |
1165 | } |
1166 | ||
90572890 | 1167 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); |
75980e97 | 1168 | |
90572890 | 1169 | static inline void page_cpupid_reset_last(struct page *page) |
75980e97 | 1170 | { |
09940a4f | 1171 | page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT; |
75980e97 | 1172 | } |
90572890 PZ |
1173 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ |
1174 | #else /* !CONFIG_NUMA_BALANCING */ | |
1175 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
57e0a030 | 1176 | { |
90572890 | 1177 | return page_to_nid(page); /* XXX */ |
57e0a030 MG |
1178 | } |
1179 | ||
90572890 | 1180 | static inline int page_cpupid_last(struct page *page) |
57e0a030 | 1181 | { |
90572890 | 1182 | return page_to_nid(page); /* XXX */ |
57e0a030 MG |
1183 | } |
1184 | ||
90572890 | 1185 | static inline int cpupid_to_nid(int cpupid) |
b795854b MG |
1186 | { |
1187 | return -1; | |
1188 | } | |
1189 | ||
90572890 | 1190 | static inline int cpupid_to_pid(int cpupid) |
b795854b MG |
1191 | { |
1192 | return -1; | |
1193 | } | |
1194 | ||
90572890 | 1195 | static inline int cpupid_to_cpu(int cpupid) |
b795854b MG |
1196 | { |
1197 | return -1; | |
1198 | } | |
1199 | ||
90572890 PZ |
1200 | static inline int cpu_pid_to_cpupid(int nid, int pid) |
1201 | { | |
1202 | return -1; | |
1203 | } | |
1204 | ||
1205 | static inline bool cpupid_pid_unset(int cpupid) | |
b795854b MG |
1206 | { |
1207 | return 1; | |
1208 | } | |
1209 | ||
90572890 | 1210 | static inline void page_cpupid_reset_last(struct page *page) |
57e0a030 MG |
1211 | { |
1212 | } | |
8c8a743c PZ |
1213 | |
1214 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) | |
1215 | { | |
1216 | return false; | |
1217 | } | |
90572890 | 1218 | #endif /* CONFIG_NUMA_BALANCING */ |
57e0a030 | 1219 | |
2813b9c0 AK |
1220 | #ifdef CONFIG_KASAN_SW_TAGS |
1221 | static inline u8 page_kasan_tag(const struct page *page) | |
1222 | { | |
1223 | return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK; | |
1224 | } | |
1225 | ||
1226 | static inline void page_kasan_tag_set(struct page *page, u8 tag) | |
1227 | { | |
1228 | page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT); | |
1229 | page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT; | |
1230 | } | |
1231 | ||
1232 | static inline void page_kasan_tag_reset(struct page *page) | |
1233 | { | |
1234 | page_kasan_tag_set(page, 0xff); | |
1235 | } | |
1236 | #else | |
1237 | static inline u8 page_kasan_tag(const struct page *page) | |
1238 | { | |
1239 | return 0xff; | |
1240 | } | |
1241 | ||
1242 | static inline void page_kasan_tag_set(struct page *page, u8 tag) { } | |
1243 | static inline void page_kasan_tag_reset(struct page *page) { } | |
1244 | #endif | |
1245 | ||
33dd4e0e | 1246 | static inline struct zone *page_zone(const struct page *page) |
89689ae7 CL |
1247 | { |
1248 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
1249 | } | |
1250 | ||
75ef7184 MG |
1251 | static inline pg_data_t *page_pgdat(const struct page *page) |
1252 | { | |
1253 | return NODE_DATA(page_to_nid(page)); | |
1254 | } | |
1255 | ||
9127ab4f | 1256 | #ifdef SECTION_IN_PAGE_FLAGS |
bf4e8902 DK |
1257 | static inline void set_page_section(struct page *page, unsigned long section) |
1258 | { | |
1259 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
1260 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
1261 | } | |
1262 | ||
aa462abe | 1263 | static inline unsigned long page_to_section(const struct page *page) |
d41dee36 AW |
1264 | { |
1265 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
1266 | } | |
308c05e3 | 1267 | #endif |
d41dee36 | 1268 | |
2f1b6248 | 1269 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
348f8b6c DH |
1270 | { |
1271 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
1272 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
1273 | } | |
2f1b6248 | 1274 | |
348f8b6c DH |
1275 | static inline void set_page_node(struct page *page, unsigned long node) |
1276 | { | |
1277 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
1278 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
1da177e4 | 1279 | } |
89689ae7 | 1280 | |
2f1b6248 | 1281 | static inline void set_page_links(struct page *page, enum zone_type zone, |
d41dee36 | 1282 | unsigned long node, unsigned long pfn) |
1da177e4 | 1283 | { |
348f8b6c DH |
1284 | set_page_zone(page, zone); |
1285 | set_page_node(page, node); | |
9127ab4f | 1286 | #ifdef SECTION_IN_PAGE_FLAGS |
d41dee36 | 1287 | set_page_section(page, pfn_to_section_nr(pfn)); |
bf4e8902 | 1288 | #endif |
1da177e4 LT |
1289 | } |
1290 | ||
0610c25d GT |
1291 | #ifdef CONFIG_MEMCG |
1292 | static inline struct mem_cgroup *page_memcg(struct page *page) | |
1293 | { | |
1294 | return page->mem_cgroup; | |
1295 | } | |
55779ec7 JW |
1296 | static inline struct mem_cgroup *page_memcg_rcu(struct page *page) |
1297 | { | |
1298 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
1299 | return READ_ONCE(page->mem_cgroup); | |
1300 | } | |
0610c25d GT |
1301 | #else |
1302 | static inline struct mem_cgroup *page_memcg(struct page *page) | |
1303 | { | |
1304 | return NULL; | |
1305 | } | |
55779ec7 JW |
1306 | static inline struct mem_cgroup *page_memcg_rcu(struct page *page) |
1307 | { | |
1308 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
1309 | return NULL; | |
1310 | } | |
0610c25d GT |
1311 | #endif |
1312 | ||
f6ac2354 CL |
1313 | /* |
1314 | * Some inline functions in vmstat.h depend on page_zone() | |
1315 | */ | |
1316 | #include <linux/vmstat.h> | |
1317 | ||
33dd4e0e | 1318 | static __always_inline void *lowmem_page_address(const struct page *page) |
1da177e4 | 1319 | { |
1dff8083 | 1320 | return page_to_virt(page); |
1da177e4 LT |
1321 | } |
1322 | ||
1323 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
1324 | #define HASHED_PAGE_VIRTUAL | |
1325 | #endif | |
1326 | ||
1327 | #if defined(WANT_PAGE_VIRTUAL) | |
f92f455f GU |
1328 | static inline void *page_address(const struct page *page) |
1329 | { | |
1330 | return page->virtual; | |
1331 | } | |
1332 | static inline void set_page_address(struct page *page, void *address) | |
1333 | { | |
1334 | page->virtual = address; | |
1335 | } | |
1da177e4 LT |
1336 | #define page_address_init() do { } while(0) |
1337 | #endif | |
1338 | ||
1339 | #if defined(HASHED_PAGE_VIRTUAL) | |
f9918794 | 1340 | void *page_address(const struct page *page); |
1da177e4 LT |
1341 | void set_page_address(struct page *page, void *virtual); |
1342 | void page_address_init(void); | |
1343 | #endif | |
1344 | ||
1345 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
1346 | #define page_address(page) lowmem_page_address(page) | |
1347 | #define set_page_address(page, address) do { } while(0) | |
1348 | #define page_address_init() do { } while(0) | |
1349 | #endif | |
1350 | ||
e39155ea KS |
1351 | extern void *page_rmapping(struct page *page); |
1352 | extern struct anon_vma *page_anon_vma(struct page *page); | |
9800339b | 1353 | extern struct address_space *page_mapping(struct page *page); |
1da177e4 | 1354 | |
f981c595 MG |
1355 | extern struct address_space *__page_file_mapping(struct page *); |
1356 | ||
1357 | static inline | |
1358 | struct address_space *page_file_mapping(struct page *page) | |
1359 | { | |
1360 | if (unlikely(PageSwapCache(page))) | |
1361 | return __page_file_mapping(page); | |
1362 | ||
1363 | return page->mapping; | |
1364 | } | |
1365 | ||
f6ab1f7f HY |
1366 | extern pgoff_t __page_file_index(struct page *page); |
1367 | ||
1da177e4 LT |
1368 | /* |
1369 | * Return the pagecache index of the passed page. Regular pagecache pages | |
f6ab1f7f | 1370 | * use ->index whereas swapcache pages use swp_offset(->private) |
1da177e4 LT |
1371 | */ |
1372 | static inline pgoff_t page_index(struct page *page) | |
1373 | { | |
1374 | if (unlikely(PageSwapCache(page))) | |
f6ab1f7f | 1375 | return __page_file_index(page); |
1da177e4 LT |
1376 | return page->index; |
1377 | } | |
1378 | ||
1aa8aea5 | 1379 | bool page_mapped(struct page *page); |
bda807d4 | 1380 | struct address_space *page_mapping(struct page *page); |
cb9f753a | 1381 | struct address_space *page_mapping_file(struct page *page); |
1da177e4 | 1382 | |
2f064f34 MH |
1383 | /* |
1384 | * Return true only if the page has been allocated with | |
1385 | * ALLOC_NO_WATERMARKS and the low watermark was not | |
1386 | * met implying that the system is under some pressure. | |
1387 | */ | |
1388 | static inline bool page_is_pfmemalloc(struct page *page) | |
1389 | { | |
1390 | /* | |
1391 | * Page index cannot be this large so this must be | |
1392 | * a pfmemalloc page. | |
1393 | */ | |
1394 | return page->index == -1UL; | |
1395 | } | |
1396 | ||
1397 | /* | |
1398 | * Only to be called by the page allocator on a freshly allocated | |
1399 | * page. | |
1400 | */ | |
1401 | static inline void set_page_pfmemalloc(struct page *page) | |
1402 | { | |
1403 | page->index = -1UL; | |
1404 | } | |
1405 | ||
1406 | static inline void clear_page_pfmemalloc(struct page *page) | |
1407 | { | |
1408 | page->index = 0; | |
1409 | } | |
1410 | ||
1c0fe6e3 NP |
1411 | /* |
1412 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | |
1413 | */ | |
1414 | extern void pagefault_out_of_memory(void); | |
1415 | ||
1da177e4 LT |
1416 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
1417 | ||
ddd588b5 | 1418 | /* |
7bf02ea2 | 1419 | * Flags passed to show_mem() and show_free_areas() to suppress output in |
ddd588b5 DR |
1420 | * various contexts. |
1421 | */ | |
4b59e6c4 | 1422 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ |
ddd588b5 | 1423 | |
9af744d7 | 1424 | extern void show_free_areas(unsigned int flags, nodemask_t *nodemask); |
1da177e4 | 1425 | |
7f43add4 | 1426 | extern bool can_do_mlock(void); |
1da177e4 LT |
1427 | extern int user_shm_lock(size_t, struct user_struct *); |
1428 | extern void user_shm_unlock(size_t, struct user_struct *); | |
1429 | ||
1430 | /* | |
1431 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
1432 | */ | |
1433 | struct zap_details { | |
1da177e4 LT |
1434 | struct address_space *check_mapping; /* Check page->mapping if set */ |
1435 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
1436 | pgoff_t last_index; /* Highest page->index to unmap */ | |
1da177e4 LT |
1437 | }; |
1438 | ||
df6ad698 JG |
1439 | struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
1440 | pte_t pte, bool with_public_device); | |
1441 | #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false) | |
1442 | ||
28093f9f GS |
1443 | struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr, |
1444 | pmd_t pmd); | |
7e675137 | 1445 | |
27d036e3 LR |
1446 | void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
1447 | unsigned long size); | |
14f5ff5d | 1448 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, |
27d036e3 | 1449 | unsigned long size); |
4f74d2c8 LT |
1450 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
1451 | unsigned long start, unsigned long end); | |
e6473092 MM |
1452 | |
1453 | /** | |
1454 | * mm_walk - callbacks for walk_page_range | |
a00cc7d9 MW |
1455 | * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry |
1456 | * this handler should only handle pud_trans_huge() puds. | |
1457 | * the pmd_entry or pte_entry callbacks will be used for | |
1458 | * regular PUDs. | |
e6473092 | 1459 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry |
03319327 DH |
1460 | * this handler is required to be able to handle |
1461 | * pmd_trans_huge() pmds. They may simply choose to | |
1462 | * split_huge_page() instead of handling it explicitly. | |
e6473092 MM |
1463 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry |
1464 | * @pte_hole: if set, called for each hole at all levels | |
5dc37642 | 1465 | * @hugetlb_entry: if set, called for each hugetlb entry |
fafaa426 | 1466 | * @test_walk: caller specific callback function to determine whether |
f7e2355f | 1467 | * we walk over the current vma or not. Returning 0 |
fafaa426 NH |
1468 | * value means "do page table walk over the current vma," |
1469 | * and a negative one means "abort current page table walk | |
f7e2355f | 1470 | * right now." 1 means "skip the current vma." |
fafaa426 NH |
1471 | * @mm: mm_struct representing the target process of page table walk |
1472 | * @vma: vma currently walked (NULL if walking outside vmas) | |
1473 | * @private: private data for callbacks' usage | |
e6473092 | 1474 | * |
fafaa426 | 1475 | * (see the comment on walk_page_range() for more details) |
e6473092 MM |
1476 | */ |
1477 | struct mm_walk { | |
a00cc7d9 MW |
1478 | int (*pud_entry)(pud_t *pud, unsigned long addr, |
1479 | unsigned long next, struct mm_walk *walk); | |
0f157a5b AM |
1480 | int (*pmd_entry)(pmd_t *pmd, unsigned long addr, |
1481 | unsigned long next, struct mm_walk *walk); | |
1482 | int (*pte_entry)(pte_t *pte, unsigned long addr, | |
1483 | unsigned long next, struct mm_walk *walk); | |
1484 | int (*pte_hole)(unsigned long addr, unsigned long next, | |
1485 | struct mm_walk *walk); | |
1486 | int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, | |
1487 | unsigned long addr, unsigned long next, | |
1488 | struct mm_walk *walk); | |
fafaa426 NH |
1489 | int (*test_walk)(unsigned long addr, unsigned long next, |
1490 | struct mm_walk *walk); | |
2165009b | 1491 | struct mm_struct *mm; |
fafaa426 | 1492 | struct vm_area_struct *vma; |
2165009b | 1493 | void *private; |
e6473092 MM |
1494 | }; |
1495 | ||
ac46d4f3 JG |
1496 | struct mmu_notifier_range; |
1497 | ||
2165009b DH |
1498 | int walk_page_range(unsigned long addr, unsigned long end, |
1499 | struct mm_walk *walk); | |
900fc5f1 | 1500 | int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk); |
42b77728 | 1501 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
3bf5ee95 | 1502 | unsigned long end, unsigned long floor, unsigned long ceiling); |
1da177e4 LT |
1503 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
1504 | struct vm_area_struct *vma); | |
09796395 | 1505 | int follow_pte_pmd(struct mm_struct *mm, unsigned long address, |
ac46d4f3 JG |
1506 | struct mmu_notifier_range *range, |
1507 | pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp); | |
3b6748e2 JW |
1508 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
1509 | unsigned long *pfn); | |
d87fe660 | 1510 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
1511 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | |
28b2ee20 RR |
1512 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
1513 | void *buf, int len, int write); | |
1da177e4 | 1514 | |
7caef267 | 1515 | extern void truncate_pagecache(struct inode *inode, loff_t new); |
2c27c65e | 1516 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
90a80202 | 1517 | void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); |
623e3db9 | 1518 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); |
750b4987 | 1519 | int truncate_inode_page(struct address_space *mapping, struct page *page); |
25718736 | 1520 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
83f78668 WF |
1521 | int invalidate_inode_page(struct page *page); |
1522 | ||
7ee1dd3f | 1523 | #ifdef CONFIG_MMU |
2b740303 SJ |
1524 | extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
1525 | unsigned long address, unsigned int flags); | |
5c723ba5 | 1526 | extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, |
4a9e1cda DD |
1527 | unsigned long address, unsigned int fault_flags, |
1528 | bool *unlocked); | |
977fbdcd MW |
1529 | void unmap_mapping_pages(struct address_space *mapping, |
1530 | pgoff_t start, pgoff_t nr, bool even_cows); | |
1531 | void unmap_mapping_range(struct address_space *mapping, | |
1532 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
7ee1dd3f | 1533 | #else |
2b740303 | 1534 | static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
dcddffd4 | 1535 | unsigned long address, unsigned int flags) |
7ee1dd3f DH |
1536 | { |
1537 | /* should never happen if there's no MMU */ | |
1538 | BUG(); | |
1539 | return VM_FAULT_SIGBUS; | |
1540 | } | |
5c723ba5 PZ |
1541 | static inline int fixup_user_fault(struct task_struct *tsk, |
1542 | struct mm_struct *mm, unsigned long address, | |
4a9e1cda | 1543 | unsigned int fault_flags, bool *unlocked) |
5c723ba5 PZ |
1544 | { |
1545 | /* should never happen if there's no MMU */ | |
1546 | BUG(); | |
1547 | return -EFAULT; | |
1548 | } | |
977fbdcd MW |
1549 | static inline void unmap_mapping_pages(struct address_space *mapping, |
1550 | pgoff_t start, pgoff_t nr, bool even_cows) { } | |
1551 | static inline void unmap_mapping_range(struct address_space *mapping, | |
1552 | loff_t const holebegin, loff_t const holelen, int even_cows) { } | |
7ee1dd3f | 1553 | #endif |
f33ea7f4 | 1554 | |
977fbdcd MW |
1555 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
1556 | loff_t const holebegin, loff_t const holelen) | |
1557 | { | |
1558 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
1559 | } | |
1560 | ||
1561 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, | |
1562 | void *buf, int len, unsigned int gup_flags); | |
5ddd36b9 | 1563 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
6347e8d5 | 1564 | void *buf, int len, unsigned int gup_flags); |
84d77d3f EB |
1565 | extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm, |
1566 | unsigned long addr, void *buf, int len, unsigned int gup_flags); | |
1da177e4 | 1567 | |
1e987790 DH |
1568 | long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm, |
1569 | unsigned long start, unsigned long nr_pages, | |
9beae1ea | 1570 | unsigned int gup_flags, struct page **pages, |
5b56d49f | 1571 | struct vm_area_struct **vmas, int *locked); |
c12d2da5 | 1572 | long get_user_pages(unsigned long start, unsigned long nr_pages, |
768ae309 | 1573 | unsigned int gup_flags, struct page **pages, |
cde70140 | 1574 | struct vm_area_struct **vmas); |
c12d2da5 | 1575 | long get_user_pages_locked(unsigned long start, unsigned long nr_pages, |
3b913179 | 1576 | unsigned int gup_flags, struct page **pages, int *locked); |
c12d2da5 | 1577 | long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages, |
c164154f | 1578 | struct page **pages, unsigned int gup_flags); |
9a4e9f3b | 1579 | |
73b0140b IW |
1580 | int get_user_pages_fast(unsigned long start, int nr_pages, |
1581 | unsigned int gup_flags, struct page **pages); | |
8025e5dd JK |
1582 | |
1583 | /* Container for pinned pfns / pages */ | |
1584 | struct frame_vector { | |
1585 | unsigned int nr_allocated; /* Number of frames we have space for */ | |
1586 | unsigned int nr_frames; /* Number of frames stored in ptrs array */ | |
1587 | bool got_ref; /* Did we pin pages by getting page ref? */ | |
1588 | bool is_pfns; /* Does array contain pages or pfns? */ | |
1589 | void *ptrs[0]; /* Array of pinned pfns / pages. Use | |
1590 | * pfns_vector_pages() or pfns_vector_pfns() | |
1591 | * for access */ | |
1592 | }; | |
1593 | ||
1594 | struct frame_vector *frame_vector_create(unsigned int nr_frames); | |
1595 | void frame_vector_destroy(struct frame_vector *vec); | |
1596 | int get_vaddr_frames(unsigned long start, unsigned int nr_pfns, | |
7f23b350 | 1597 | unsigned int gup_flags, struct frame_vector *vec); |
8025e5dd JK |
1598 | void put_vaddr_frames(struct frame_vector *vec); |
1599 | int frame_vector_to_pages(struct frame_vector *vec); | |
1600 | void frame_vector_to_pfns(struct frame_vector *vec); | |
1601 | ||
1602 | static inline unsigned int frame_vector_count(struct frame_vector *vec) | |
1603 | { | |
1604 | return vec->nr_frames; | |
1605 | } | |
1606 | ||
1607 | static inline struct page **frame_vector_pages(struct frame_vector *vec) | |
1608 | { | |
1609 | if (vec->is_pfns) { | |
1610 | int err = frame_vector_to_pages(vec); | |
1611 | ||
1612 | if (err) | |
1613 | return ERR_PTR(err); | |
1614 | } | |
1615 | return (struct page **)(vec->ptrs); | |
1616 | } | |
1617 | ||
1618 | static inline unsigned long *frame_vector_pfns(struct frame_vector *vec) | |
1619 | { | |
1620 | if (!vec->is_pfns) | |
1621 | frame_vector_to_pfns(vec); | |
1622 | return (unsigned long *)(vec->ptrs); | |
1623 | } | |
1624 | ||
18022c5d MG |
1625 | struct kvec; |
1626 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, | |
1627 | struct page **pages); | |
1628 | int get_kernel_page(unsigned long start, int write, struct page **pages); | |
f3e8fccd | 1629 | struct page *get_dump_page(unsigned long addr); |
1da177e4 | 1630 | |
cf9a2ae8 | 1631 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
d47992f8 LC |
1632 | extern void do_invalidatepage(struct page *page, unsigned int offset, |
1633 | unsigned int length); | |
cf9a2ae8 | 1634 | |
f82b3764 | 1635 | void __set_page_dirty(struct page *, struct address_space *, int warn); |
1da177e4 | 1636 | int __set_page_dirty_nobuffers(struct page *page); |
76719325 | 1637 | int __set_page_dirty_no_writeback(struct page *page); |
1da177e4 LT |
1638 | int redirty_page_for_writepage(struct writeback_control *wbc, |
1639 | struct page *page); | |
62cccb8c | 1640 | void account_page_dirtied(struct page *page, struct address_space *mapping); |
c4843a75 | 1641 | void account_page_cleaned(struct page *page, struct address_space *mapping, |
62cccb8c | 1642 | struct bdi_writeback *wb); |
b3c97528 | 1643 | int set_page_dirty(struct page *page); |
1da177e4 | 1644 | int set_page_dirty_lock(struct page *page); |
736304f3 JK |
1645 | void __cancel_dirty_page(struct page *page); |
1646 | static inline void cancel_dirty_page(struct page *page) | |
1647 | { | |
1648 | /* Avoid atomic ops, locking, etc. when not actually needed. */ | |
1649 | if (PageDirty(page)) | |
1650 | __cancel_dirty_page(page); | |
1651 | } | |
1da177e4 | 1652 | int clear_page_dirty_for_io(struct page *page); |
b9ea2515 | 1653 | |
a9090253 | 1654 | int get_cmdline(struct task_struct *task, char *buffer, int buflen); |
1da177e4 | 1655 | |
b5330628 ON |
1656 | static inline bool vma_is_anonymous(struct vm_area_struct *vma) |
1657 | { | |
1658 | return !vma->vm_ops; | |
1659 | } | |
1660 | ||
b0506e48 MR |
1661 | #ifdef CONFIG_SHMEM |
1662 | /* | |
1663 | * The vma_is_shmem is not inline because it is used only by slow | |
1664 | * paths in userfault. | |
1665 | */ | |
1666 | bool vma_is_shmem(struct vm_area_struct *vma); | |
1667 | #else | |
1668 | static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; } | |
1669 | #endif | |
1670 | ||
d17af505 | 1671 | int vma_is_stack_for_current(struct vm_area_struct *vma); |
b7643757 | 1672 | |
b6a2fea3 OW |
1673 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
1674 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
38a76013 ML |
1675 | unsigned long new_addr, unsigned long len, |
1676 | bool need_rmap_locks); | |
7da4d641 PZ |
1677 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, |
1678 | unsigned long end, pgprot_t newprot, | |
4b10e7d5 | 1679 | int dirty_accountable, int prot_numa); |
b6a2fea3 OW |
1680 | extern int mprotect_fixup(struct vm_area_struct *vma, |
1681 | struct vm_area_struct **pprev, unsigned long start, | |
1682 | unsigned long end, unsigned long newflags); | |
1da177e4 | 1683 | |
465a454f PZ |
1684 | /* |
1685 | * doesn't attempt to fault and will return short. | |
1686 | */ | |
1687 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1688 | struct page **pages); | |
d559db08 KH |
1689 | /* |
1690 | * per-process(per-mm_struct) statistics. | |
1691 | */ | |
d559db08 KH |
1692 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) |
1693 | { | |
69c97823 KK |
1694 | long val = atomic_long_read(&mm->rss_stat.count[member]); |
1695 | ||
1696 | #ifdef SPLIT_RSS_COUNTING | |
1697 | /* | |
1698 | * counter is updated in asynchronous manner and may go to minus. | |
1699 | * But it's never be expected number for users. | |
1700 | */ | |
1701 | if (val < 0) | |
1702 | val = 0; | |
172703b0 | 1703 | #endif |
69c97823 KK |
1704 | return (unsigned long)val; |
1705 | } | |
d559db08 KH |
1706 | |
1707 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) | |
1708 | { | |
172703b0 | 1709 | atomic_long_add(value, &mm->rss_stat.count[member]); |
d559db08 KH |
1710 | } |
1711 | ||
1712 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
1713 | { | |
172703b0 | 1714 | atomic_long_inc(&mm->rss_stat.count[member]); |
d559db08 KH |
1715 | } |
1716 | ||
1717 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
1718 | { | |
172703b0 | 1719 | atomic_long_dec(&mm->rss_stat.count[member]); |
d559db08 KH |
1720 | } |
1721 | ||
eca56ff9 JM |
1722 | /* Optimized variant when page is already known not to be PageAnon */ |
1723 | static inline int mm_counter_file(struct page *page) | |
1724 | { | |
1725 | if (PageSwapBacked(page)) | |
1726 | return MM_SHMEMPAGES; | |
1727 | return MM_FILEPAGES; | |
1728 | } | |
1729 | ||
1730 | static inline int mm_counter(struct page *page) | |
1731 | { | |
1732 | if (PageAnon(page)) | |
1733 | return MM_ANONPAGES; | |
1734 | return mm_counter_file(page); | |
1735 | } | |
1736 | ||
d559db08 KH |
1737 | static inline unsigned long get_mm_rss(struct mm_struct *mm) |
1738 | { | |
1739 | return get_mm_counter(mm, MM_FILEPAGES) + | |
eca56ff9 JM |
1740 | get_mm_counter(mm, MM_ANONPAGES) + |
1741 | get_mm_counter(mm, MM_SHMEMPAGES); | |
d559db08 KH |
1742 | } |
1743 | ||
1744 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | |
1745 | { | |
1746 | return max(mm->hiwater_rss, get_mm_rss(mm)); | |
1747 | } | |
1748 | ||
1749 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | |
1750 | { | |
1751 | return max(mm->hiwater_vm, mm->total_vm); | |
1752 | } | |
1753 | ||
1754 | static inline void update_hiwater_rss(struct mm_struct *mm) | |
1755 | { | |
1756 | unsigned long _rss = get_mm_rss(mm); | |
1757 | ||
1758 | if ((mm)->hiwater_rss < _rss) | |
1759 | (mm)->hiwater_rss = _rss; | |
1760 | } | |
1761 | ||
1762 | static inline void update_hiwater_vm(struct mm_struct *mm) | |
1763 | { | |
1764 | if (mm->hiwater_vm < mm->total_vm) | |
1765 | mm->hiwater_vm = mm->total_vm; | |
1766 | } | |
1767 | ||
695f0559 PC |
1768 | static inline void reset_mm_hiwater_rss(struct mm_struct *mm) |
1769 | { | |
1770 | mm->hiwater_rss = get_mm_rss(mm); | |
1771 | } | |
1772 | ||
d559db08 KH |
1773 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, |
1774 | struct mm_struct *mm) | |
1775 | { | |
1776 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | |
1777 | ||
1778 | if (*maxrss < hiwater_rss) | |
1779 | *maxrss = hiwater_rss; | |
1780 | } | |
1781 | ||
53bddb4e | 1782 | #if defined(SPLIT_RSS_COUNTING) |
05af2e10 | 1783 | void sync_mm_rss(struct mm_struct *mm); |
53bddb4e | 1784 | #else |
05af2e10 | 1785 | static inline void sync_mm_rss(struct mm_struct *mm) |
53bddb4e KH |
1786 | { |
1787 | } | |
1788 | #endif | |
465a454f | 1789 | |
3565fce3 DW |
1790 | #ifndef __HAVE_ARCH_PTE_DEVMAP |
1791 | static inline int pte_devmap(pte_t pte) | |
1792 | { | |
1793 | return 0; | |
1794 | } | |
1795 | #endif | |
1796 | ||
6d2329f8 | 1797 | int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot); |
d08b3851 | 1798 | |
25ca1d6c NK |
1799 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
1800 | spinlock_t **ptl); | |
1801 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1802 | spinlock_t **ptl) | |
1803 | { | |
1804 | pte_t *ptep; | |
1805 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | |
1806 | return ptep; | |
1807 | } | |
c9cfcddf | 1808 | |
c2febafc KS |
1809 | #ifdef __PAGETABLE_P4D_FOLDED |
1810 | static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, | |
1811 | unsigned long address) | |
1812 | { | |
1813 | return 0; | |
1814 | } | |
1815 | #else | |
1816 | int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); | |
1817 | #endif | |
1818 | ||
b4e98d9a | 1819 | #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU) |
c2febafc | 1820 | static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, |
5f22df00 NP |
1821 | unsigned long address) |
1822 | { | |
1823 | return 0; | |
1824 | } | |
b4e98d9a KS |
1825 | static inline void mm_inc_nr_puds(struct mm_struct *mm) {} |
1826 | static inline void mm_dec_nr_puds(struct mm_struct *mm) {} | |
1827 | ||
5f22df00 | 1828 | #else |
c2febafc | 1829 | int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address); |
b4e98d9a | 1830 | |
b4e98d9a KS |
1831 | static inline void mm_inc_nr_puds(struct mm_struct *mm) |
1832 | { | |
6d212db1 MS |
1833 | if (mm_pud_folded(mm)) |
1834 | return; | |
af5b0f6a | 1835 | atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
b4e98d9a KS |
1836 | } |
1837 | ||
1838 | static inline void mm_dec_nr_puds(struct mm_struct *mm) | |
1839 | { | |
6d212db1 MS |
1840 | if (mm_pud_folded(mm)) |
1841 | return; | |
af5b0f6a | 1842 | atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
b4e98d9a | 1843 | } |
5f22df00 NP |
1844 | #endif |
1845 | ||
2d2f5119 | 1846 | #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU) |
5f22df00 NP |
1847 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
1848 | unsigned long address) | |
1849 | { | |
1850 | return 0; | |
1851 | } | |
dc6c9a35 | 1852 | |
dc6c9a35 KS |
1853 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) {} |
1854 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) {} | |
1855 | ||
5f22df00 | 1856 | #else |
1bb3630e | 1857 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
dc6c9a35 | 1858 | |
dc6c9a35 KS |
1859 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) |
1860 | { | |
6d212db1 MS |
1861 | if (mm_pmd_folded(mm)) |
1862 | return; | |
af5b0f6a | 1863 | atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
dc6c9a35 KS |
1864 | } |
1865 | ||
1866 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) | |
1867 | { | |
6d212db1 MS |
1868 | if (mm_pmd_folded(mm)) |
1869 | return; | |
af5b0f6a | 1870 | atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
dc6c9a35 | 1871 | } |
5f22df00 NP |
1872 | #endif |
1873 | ||
c4812909 | 1874 | #ifdef CONFIG_MMU |
af5b0f6a | 1875 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) |
c4812909 | 1876 | { |
af5b0f6a | 1877 | atomic_long_set(&mm->pgtables_bytes, 0); |
c4812909 KS |
1878 | } |
1879 | ||
af5b0f6a | 1880 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) |
c4812909 | 1881 | { |
af5b0f6a | 1882 | return atomic_long_read(&mm->pgtables_bytes); |
c4812909 KS |
1883 | } |
1884 | ||
1885 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) | |
1886 | { | |
af5b0f6a | 1887 | atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
c4812909 KS |
1888 | } |
1889 | ||
1890 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) | |
1891 | { | |
af5b0f6a | 1892 | atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
c4812909 KS |
1893 | } |
1894 | #else | |
c4812909 | 1895 | |
af5b0f6a KS |
1896 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {} |
1897 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) | |
c4812909 KS |
1898 | { |
1899 | return 0; | |
1900 | } | |
1901 | ||
1902 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) {} | |
1903 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) {} | |
1904 | #endif | |
1905 | ||
4cf58924 JFG |
1906 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd); |
1907 | int __pte_alloc_kernel(pmd_t *pmd); | |
1bb3630e | 1908 | |
1da177e4 LT |
1909 | /* |
1910 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
1911 | * Remove it when 4level-fixup.h has been removed. | |
1912 | */ | |
1bb3630e | 1913 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
505a60e2 KS |
1914 | |
1915 | #ifndef __ARCH_HAS_5LEVEL_HACK | |
c2febafc KS |
1916 | static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd, |
1917 | unsigned long address) | |
1918 | { | |
1919 | return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ? | |
1920 | NULL : p4d_offset(pgd, address); | |
1921 | } | |
1922 | ||
1923 | static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d, | |
1924 | unsigned long address) | |
1da177e4 | 1925 | { |
c2febafc KS |
1926 | return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ? |
1927 | NULL : pud_offset(p4d, address); | |
1da177e4 | 1928 | } |
505a60e2 | 1929 | #endif /* !__ARCH_HAS_5LEVEL_HACK */ |
1da177e4 LT |
1930 | |
1931 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
1932 | { | |
1bb3630e HD |
1933 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
1934 | NULL: pmd_offset(pud, address); | |
1da177e4 | 1935 | } |
1bb3630e HD |
1936 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
1937 | ||
57c1ffce | 1938 | #if USE_SPLIT_PTE_PTLOCKS |
597d795a | 1939 | #if ALLOC_SPLIT_PTLOCKS |
b35f1819 | 1940 | void __init ptlock_cache_init(void); |
539edb58 PZ |
1941 | extern bool ptlock_alloc(struct page *page); |
1942 | extern void ptlock_free(struct page *page); | |
1943 | ||
1944 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
1945 | { | |
1946 | return page->ptl; | |
1947 | } | |
597d795a | 1948 | #else /* ALLOC_SPLIT_PTLOCKS */ |
b35f1819 KS |
1949 | static inline void ptlock_cache_init(void) |
1950 | { | |
1951 | } | |
1952 | ||
49076ec2 KS |
1953 | static inline bool ptlock_alloc(struct page *page) |
1954 | { | |
49076ec2 KS |
1955 | return true; |
1956 | } | |
539edb58 | 1957 | |
49076ec2 KS |
1958 | static inline void ptlock_free(struct page *page) |
1959 | { | |
49076ec2 KS |
1960 | } |
1961 | ||
1962 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
1963 | { | |
539edb58 | 1964 | return &page->ptl; |
49076ec2 | 1965 | } |
597d795a | 1966 | #endif /* ALLOC_SPLIT_PTLOCKS */ |
49076ec2 KS |
1967 | |
1968 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1969 | { | |
1970 | return ptlock_ptr(pmd_page(*pmd)); | |
1971 | } | |
1972 | ||
1973 | static inline bool ptlock_init(struct page *page) | |
1974 | { | |
1975 | /* | |
1976 | * prep_new_page() initialize page->private (and therefore page->ptl) | |
1977 | * with 0. Make sure nobody took it in use in between. | |
1978 | * | |
1979 | * It can happen if arch try to use slab for page table allocation: | |
1d798ca3 | 1980 | * slab code uses page->slab_cache, which share storage with page->ptl. |
49076ec2 | 1981 | */ |
309381fe | 1982 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); |
49076ec2 KS |
1983 | if (!ptlock_alloc(page)) |
1984 | return false; | |
1985 | spin_lock_init(ptlock_ptr(page)); | |
1986 | return true; | |
1987 | } | |
1988 | ||
57c1ffce | 1989 | #else /* !USE_SPLIT_PTE_PTLOCKS */ |
4c21e2f2 HD |
1990 | /* |
1991 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
1992 | */ | |
49076ec2 KS |
1993 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
1994 | { | |
1995 | return &mm->page_table_lock; | |
1996 | } | |
b35f1819 | 1997 | static inline void ptlock_cache_init(void) {} |
49076ec2 | 1998 | static inline bool ptlock_init(struct page *page) { return true; } |
9e247bab | 1999 | static inline void ptlock_free(struct page *page) {} |
57c1ffce | 2000 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
4c21e2f2 | 2001 | |
b35f1819 KS |
2002 | static inline void pgtable_init(void) |
2003 | { | |
2004 | ptlock_cache_init(); | |
2005 | pgtable_cache_init(); | |
2006 | } | |
2007 | ||
390f44e2 | 2008 | static inline bool pgtable_page_ctor(struct page *page) |
2f569afd | 2009 | { |
706874e9 VD |
2010 | if (!ptlock_init(page)) |
2011 | return false; | |
1d40a5ea | 2012 | __SetPageTable(page); |
2f569afd | 2013 | inc_zone_page_state(page, NR_PAGETABLE); |
706874e9 | 2014 | return true; |
2f569afd MS |
2015 | } |
2016 | ||
2017 | static inline void pgtable_page_dtor(struct page *page) | |
2018 | { | |
9e247bab | 2019 | ptlock_free(page); |
1d40a5ea | 2020 | __ClearPageTable(page); |
2f569afd MS |
2021 | dec_zone_page_state(page, NR_PAGETABLE); |
2022 | } | |
2023 | ||
c74df32c HD |
2024 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
2025 | ({ \ | |
4c21e2f2 | 2026 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
c74df32c HD |
2027 | pte_t *__pte = pte_offset_map(pmd, address); \ |
2028 | *(ptlp) = __ptl; \ | |
2029 | spin_lock(__ptl); \ | |
2030 | __pte; \ | |
2031 | }) | |
2032 | ||
2033 | #define pte_unmap_unlock(pte, ptl) do { \ | |
2034 | spin_unlock(ptl); \ | |
2035 | pte_unmap(pte); \ | |
2036 | } while (0) | |
2037 | ||
4cf58924 | 2038 | #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd)) |
3ed3a4f0 KS |
2039 | |
2040 | #define pte_alloc_map(mm, pmd, address) \ | |
4cf58924 | 2041 | (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address)) |
1bb3630e | 2042 | |
c74df32c | 2043 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
4cf58924 | 2044 | (pte_alloc(mm, pmd) ? \ |
3ed3a4f0 | 2045 | NULL : pte_offset_map_lock(mm, pmd, address, ptlp)) |
c74df32c | 2046 | |
1bb3630e | 2047 | #define pte_alloc_kernel(pmd, address) \ |
4cf58924 | 2048 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \ |
1bb3630e | 2049 | NULL: pte_offset_kernel(pmd, address)) |
1da177e4 | 2050 | |
e009bb30 KS |
2051 | #if USE_SPLIT_PMD_PTLOCKS |
2052 | ||
634391ac MS |
2053 | static struct page *pmd_to_page(pmd_t *pmd) |
2054 | { | |
2055 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); | |
2056 | return virt_to_page((void *)((unsigned long) pmd & mask)); | |
2057 | } | |
2058 | ||
e009bb30 KS |
2059 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2060 | { | |
634391ac | 2061 | return ptlock_ptr(pmd_to_page(pmd)); |
e009bb30 KS |
2062 | } |
2063 | ||
2064 | static inline bool pgtable_pmd_page_ctor(struct page *page) | |
2065 | { | |
e009bb30 KS |
2066 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
2067 | page->pmd_huge_pte = NULL; | |
2068 | #endif | |
49076ec2 | 2069 | return ptlock_init(page); |
e009bb30 KS |
2070 | } |
2071 | ||
2072 | static inline void pgtable_pmd_page_dtor(struct page *page) | |
2073 | { | |
2074 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
309381fe | 2075 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); |
e009bb30 | 2076 | #endif |
49076ec2 | 2077 | ptlock_free(page); |
e009bb30 KS |
2078 | } |
2079 | ||
634391ac | 2080 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) |
e009bb30 KS |
2081 | |
2082 | #else | |
2083 | ||
9a86cb7b KS |
2084 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2085 | { | |
2086 | return &mm->page_table_lock; | |
2087 | } | |
2088 | ||
e009bb30 KS |
2089 | static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } |
2090 | static inline void pgtable_pmd_page_dtor(struct page *page) {} | |
2091 | ||
c389a250 | 2092 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) |
9a86cb7b | 2093 | |
e009bb30 KS |
2094 | #endif |
2095 | ||
9a86cb7b KS |
2096 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) |
2097 | { | |
2098 | spinlock_t *ptl = pmd_lockptr(mm, pmd); | |
2099 | spin_lock(ptl); | |
2100 | return ptl; | |
2101 | } | |
2102 | ||
a00cc7d9 MW |
2103 | /* |
2104 | * No scalability reason to split PUD locks yet, but follow the same pattern | |
2105 | * as the PMD locks to make it easier if we decide to. The VM should not be | |
2106 | * considered ready to switch to split PUD locks yet; there may be places | |
2107 | * which need to be converted from page_table_lock. | |
2108 | */ | |
2109 | static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud) | |
2110 | { | |
2111 | return &mm->page_table_lock; | |
2112 | } | |
2113 | ||
2114 | static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud) | |
2115 | { | |
2116 | spinlock_t *ptl = pud_lockptr(mm, pud); | |
2117 | ||
2118 | spin_lock(ptl); | |
2119 | return ptl; | |
2120 | } | |
62906027 | 2121 | |
a00cc7d9 | 2122 | extern void __init pagecache_init(void); |
1da177e4 | 2123 | extern void free_area_init(unsigned long * zones_size); |
03e85f9d | 2124 | extern void __init free_area_init_node(int nid, unsigned long * zones_size, |
9109fb7b | 2125 | unsigned long zone_start_pfn, unsigned long *zholes_size); |
49a7f04a DH |
2126 | extern void free_initmem(void); |
2127 | ||
69afade7 JL |
2128 | /* |
2129 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) | |
2130 | * into the buddy system. The freed pages will be poisoned with pattern | |
dbe67df4 | 2131 | * "poison" if it's within range [0, UCHAR_MAX]. |
69afade7 JL |
2132 | * Return pages freed into the buddy system. |
2133 | */ | |
11199692 | 2134 | extern unsigned long free_reserved_area(void *start, void *end, |
e5cb113f | 2135 | int poison, const char *s); |
c3d5f5f0 | 2136 | |
cfa11e08 JL |
2137 | #ifdef CONFIG_HIGHMEM |
2138 | /* | |
2139 | * Free a highmem page into the buddy system, adjusting totalhigh_pages | |
2140 | * and totalram_pages. | |
2141 | */ | |
2142 | extern void free_highmem_page(struct page *page); | |
2143 | #endif | |
69afade7 | 2144 | |
c3d5f5f0 | 2145 | extern void adjust_managed_page_count(struct page *page, long count); |
7ee3d4e8 | 2146 | extern void mem_init_print_info(const char *str); |
69afade7 | 2147 | |
4b50bcc7 | 2148 | extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end); |
92923ca3 | 2149 | |
69afade7 JL |
2150 | /* Free the reserved page into the buddy system, so it gets managed. */ |
2151 | static inline void __free_reserved_page(struct page *page) | |
2152 | { | |
2153 | ClearPageReserved(page); | |
2154 | init_page_count(page); | |
2155 | __free_page(page); | |
2156 | } | |
2157 | ||
2158 | static inline void free_reserved_page(struct page *page) | |
2159 | { | |
2160 | __free_reserved_page(page); | |
2161 | adjust_managed_page_count(page, 1); | |
2162 | } | |
2163 | ||
2164 | static inline void mark_page_reserved(struct page *page) | |
2165 | { | |
2166 | SetPageReserved(page); | |
2167 | adjust_managed_page_count(page, -1); | |
2168 | } | |
2169 | ||
2170 | /* | |
2171 | * Default method to free all the __init memory into the buddy system. | |
dbe67df4 JL |
2172 | * The freed pages will be poisoned with pattern "poison" if it's within |
2173 | * range [0, UCHAR_MAX]. | |
2174 | * Return pages freed into the buddy system. | |
69afade7 JL |
2175 | */ |
2176 | static inline unsigned long free_initmem_default(int poison) | |
2177 | { | |
2178 | extern char __init_begin[], __init_end[]; | |
2179 | ||
11199692 | 2180 | return free_reserved_area(&__init_begin, &__init_end, |
69afade7 JL |
2181 | poison, "unused kernel"); |
2182 | } | |
2183 | ||
7ee3d4e8 JL |
2184 | static inline unsigned long get_num_physpages(void) |
2185 | { | |
2186 | int nid; | |
2187 | unsigned long phys_pages = 0; | |
2188 | ||
2189 | for_each_online_node(nid) | |
2190 | phys_pages += node_present_pages(nid); | |
2191 | ||
2192 | return phys_pages; | |
2193 | } | |
2194 | ||
0ee332c1 | 2195 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
c713216d | 2196 | /* |
0ee332c1 | 2197 | * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its |
c713216d MG |
2198 | * zones, allocate the backing mem_map and account for memory holes in a more |
2199 | * architecture independent manner. This is a substitute for creating the | |
2200 | * zone_sizes[] and zholes_size[] arrays and passing them to | |
2201 | * free_area_init_node() | |
2202 | * | |
2203 | * An architecture is expected to register range of page frames backed by | |
0ee332c1 | 2204 | * physical memory with memblock_add[_node]() before calling |
c713216d MG |
2205 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic |
2206 | * usage, an architecture is expected to do something like | |
2207 | * | |
2208 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
2209 | * max_highmem_pfn}; | |
2210 | * for_each_valid_physical_page_range() | |
0ee332c1 | 2211 | * memblock_add_node(base, size, nid) |
c713216d MG |
2212 | * free_area_init_nodes(max_zone_pfns); |
2213 | * | |
0ee332c1 TH |
2214 | * free_bootmem_with_active_regions() calls free_bootmem_node() for each |
2215 | * registered physical page range. Similarly | |
2216 | * sparse_memory_present_with_active_regions() calls memory_present() for | |
2217 | * each range when SPARSEMEM is enabled. | |
c713216d MG |
2218 | * |
2219 | * See mm/page_alloc.c for more information on each function exposed by | |
0ee332c1 | 2220 | * CONFIG_HAVE_MEMBLOCK_NODE_MAP. |
c713216d MG |
2221 | */ |
2222 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | |
1e01979c | 2223 | unsigned long node_map_pfn_alignment(void); |
32996250 YL |
2224 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, |
2225 | unsigned long end_pfn); | |
c713216d MG |
2226 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
2227 | unsigned long end_pfn); | |
2228 | extern void get_pfn_range_for_nid(unsigned int nid, | |
2229 | unsigned long *start_pfn, unsigned long *end_pfn); | |
2230 | extern unsigned long find_min_pfn_with_active_regions(void); | |
c713216d MG |
2231 | extern void free_bootmem_with_active_regions(int nid, |
2232 | unsigned long max_low_pfn); | |
2233 | extern void sparse_memory_present_with_active_regions(int nid); | |
f2dbcfa7 | 2234 | |
0ee332c1 | 2235 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ |
f2dbcfa7 | 2236 | |
0ee332c1 | 2237 | #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ |
f2dbcfa7 | 2238 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) |
8a942fde MG |
2239 | static inline int __early_pfn_to_nid(unsigned long pfn, |
2240 | struct mminit_pfnnid_cache *state) | |
f2dbcfa7 KH |
2241 | { |
2242 | return 0; | |
2243 | } | |
2244 | #else | |
2245 | /* please see mm/page_alloc.c */ | |
2246 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | |
f2dbcfa7 | 2247 | /* there is a per-arch backend function. */ |
8a942fde MG |
2248 | extern int __meminit __early_pfn_to_nid(unsigned long pfn, |
2249 | struct mminit_pfnnid_cache *state); | |
f2dbcfa7 KH |
2250 | #endif |
2251 | ||
aca52c39 | 2252 | #if !defined(CONFIG_FLAT_NODE_MEM_MAP) |
a4a3ede2 PT |
2253 | void zero_resv_unavail(void); |
2254 | #else | |
2255 | static inline void zero_resv_unavail(void) {} | |
2256 | #endif | |
2257 | ||
0e0b864e | 2258 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
a99583e7 CH |
2259 | extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long, |
2260 | enum memmap_context, struct vmem_altmap *); | |
bc75d33f | 2261 | extern void setup_per_zone_wmarks(void); |
1b79acc9 | 2262 | extern int __meminit init_per_zone_wmark_min(void); |
1da177e4 | 2263 | extern void mem_init(void); |
8feae131 | 2264 | extern void __init mmap_init(void); |
9af744d7 | 2265 | extern void show_mem(unsigned int flags, nodemask_t *nodemask); |
d02bd27b | 2266 | extern long si_mem_available(void); |
1da177e4 LT |
2267 | extern void si_meminfo(struct sysinfo * val); |
2268 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
f6f34b43 SD |
2269 | #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES |
2270 | extern unsigned long arch_reserved_kernel_pages(void); | |
2271 | #endif | |
1da177e4 | 2272 | |
a8e99259 MH |
2273 | extern __printf(3, 4) |
2274 | void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...); | |
a238ab5b | 2275 | |
e7c8d5c9 | 2276 | extern void setup_per_cpu_pageset(void); |
e7c8d5c9 | 2277 | |
112067f0 | 2278 | extern void zone_pcp_update(struct zone *zone); |
340175b7 | 2279 | extern void zone_pcp_reset(struct zone *zone); |
112067f0 | 2280 | |
75f7ad8e PS |
2281 | /* page_alloc.c */ |
2282 | extern int min_free_kbytes; | |
1c30844d | 2283 | extern int watermark_boost_factor; |
795ae7a0 | 2284 | extern int watermark_scale_factor; |
75f7ad8e | 2285 | |
8feae131 | 2286 | /* nommu.c */ |
33e5d769 | 2287 | extern atomic_long_t mmap_pages_allocated; |
7e660872 | 2288 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
8feae131 | 2289 | |
6b2dbba8 | 2290 | /* interval_tree.c */ |
6b2dbba8 | 2291 | void vma_interval_tree_insert(struct vm_area_struct *node, |
f808c13f | 2292 | struct rb_root_cached *root); |
9826a516 ML |
2293 | void vma_interval_tree_insert_after(struct vm_area_struct *node, |
2294 | struct vm_area_struct *prev, | |
f808c13f | 2295 | struct rb_root_cached *root); |
6b2dbba8 | 2296 | void vma_interval_tree_remove(struct vm_area_struct *node, |
f808c13f DB |
2297 | struct rb_root_cached *root); |
2298 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root, | |
6b2dbba8 ML |
2299 | unsigned long start, unsigned long last); |
2300 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, | |
2301 | unsigned long start, unsigned long last); | |
2302 | ||
2303 | #define vma_interval_tree_foreach(vma, root, start, last) \ | |
2304 | for (vma = vma_interval_tree_iter_first(root, start, last); \ | |
2305 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) | |
1da177e4 | 2306 | |
bf181b9f | 2307 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, |
f808c13f | 2308 | struct rb_root_cached *root); |
bf181b9f | 2309 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, |
f808c13f DB |
2310 | struct rb_root_cached *root); |
2311 | struct anon_vma_chain * | |
2312 | anon_vma_interval_tree_iter_first(struct rb_root_cached *root, | |
2313 | unsigned long start, unsigned long last); | |
bf181b9f ML |
2314 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( |
2315 | struct anon_vma_chain *node, unsigned long start, unsigned long last); | |
ed8ea815 ML |
2316 | #ifdef CONFIG_DEBUG_VM_RB |
2317 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); | |
2318 | #endif | |
bf181b9f ML |
2319 | |
2320 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ | |
2321 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ | |
2322 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) | |
2323 | ||
1da177e4 | 2324 | /* mmap.c */ |
34b4e4aa | 2325 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
e86f15ee AA |
2326 | extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start, |
2327 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert, | |
2328 | struct vm_area_struct *expand); | |
2329 | static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
2330 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert) | |
2331 | { | |
2332 | return __vma_adjust(vma, start, end, pgoff, insert, NULL); | |
2333 | } | |
1da177e4 LT |
2334 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
2335 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
2336 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
19a809af | 2337 | struct mempolicy *, struct vm_userfaultfd_ctx); |
1da177e4 | 2338 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
def5efe0 DR |
2339 | extern int __split_vma(struct mm_struct *, struct vm_area_struct *, |
2340 | unsigned long addr, int new_below); | |
2341 | extern int split_vma(struct mm_struct *, struct vm_area_struct *, | |
2342 | unsigned long addr, int new_below); | |
1da177e4 LT |
2343 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
2344 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
2345 | struct rb_node **, struct rb_node *); | |
a8fb5618 | 2346 | extern void unlink_file_vma(struct vm_area_struct *); |
1da177e4 | 2347 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
38a76013 ML |
2348 | unsigned long addr, unsigned long len, pgoff_t pgoff, |
2349 | bool *need_rmap_locks); | |
1da177e4 | 2350 | extern void exit_mmap(struct mm_struct *); |
925d1c40 | 2351 | |
9c599024 CG |
2352 | static inline int check_data_rlimit(unsigned long rlim, |
2353 | unsigned long new, | |
2354 | unsigned long start, | |
2355 | unsigned long end_data, | |
2356 | unsigned long start_data) | |
2357 | { | |
2358 | if (rlim < RLIM_INFINITY) { | |
2359 | if (((new - start) + (end_data - start_data)) > rlim) | |
2360 | return -ENOSPC; | |
2361 | } | |
2362 | ||
2363 | return 0; | |
2364 | } | |
2365 | ||
7906d00c AA |
2366 | extern int mm_take_all_locks(struct mm_struct *mm); |
2367 | extern void mm_drop_all_locks(struct mm_struct *mm); | |
2368 | ||
38646013 JS |
2369 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); |
2370 | extern struct file *get_mm_exe_file(struct mm_struct *mm); | |
cd81a917 | 2371 | extern struct file *get_task_exe_file(struct task_struct *task); |
925d1c40 | 2372 | |
84638335 KK |
2373 | extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages); |
2374 | extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages); | |
2375 | ||
2eefd878 DS |
2376 | extern bool vma_is_special_mapping(const struct vm_area_struct *vma, |
2377 | const struct vm_special_mapping *sm); | |
3935ed6a SS |
2378 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, |
2379 | unsigned long addr, unsigned long len, | |
a62c34bd AL |
2380 | unsigned long flags, |
2381 | const struct vm_special_mapping *spec); | |
2382 | /* This is an obsolete alternative to _install_special_mapping. */ | |
fa5dc22f RM |
2383 | extern int install_special_mapping(struct mm_struct *mm, |
2384 | unsigned long addr, unsigned long len, | |
2385 | unsigned long flags, struct page **pages); | |
1da177e4 LT |
2386 | |
2387 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
2388 | ||
0165ab44 | 2389 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
897ab3e0 MR |
2390 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff, |
2391 | struct list_head *uf); | |
1fcfd8db | 2392 | extern unsigned long do_mmap(struct file *file, unsigned long addr, |
bebeb3d6 | 2393 | unsigned long len, unsigned long prot, unsigned long flags, |
897ab3e0 MR |
2394 | vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate, |
2395 | struct list_head *uf); | |
85a06835 YS |
2396 | extern int __do_munmap(struct mm_struct *, unsigned long, size_t, |
2397 | struct list_head *uf, bool downgrade); | |
897ab3e0 MR |
2398 | extern int do_munmap(struct mm_struct *, unsigned long, size_t, |
2399 | struct list_head *uf); | |
1da177e4 | 2400 | |
1fcfd8db ON |
2401 | static inline unsigned long |
2402 | do_mmap_pgoff(struct file *file, unsigned long addr, | |
2403 | unsigned long len, unsigned long prot, unsigned long flags, | |
897ab3e0 MR |
2404 | unsigned long pgoff, unsigned long *populate, |
2405 | struct list_head *uf) | |
1fcfd8db | 2406 | { |
897ab3e0 | 2407 | return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf); |
1fcfd8db ON |
2408 | } |
2409 | ||
bebeb3d6 ML |
2410 | #ifdef CONFIG_MMU |
2411 | extern int __mm_populate(unsigned long addr, unsigned long len, | |
2412 | int ignore_errors); | |
2413 | static inline void mm_populate(unsigned long addr, unsigned long len) | |
2414 | { | |
2415 | /* Ignore errors */ | |
2416 | (void) __mm_populate(addr, len, 1); | |
2417 | } | |
2418 | #else | |
2419 | static inline void mm_populate(unsigned long addr, unsigned long len) {} | |
2420 | #endif | |
2421 | ||
e4eb1ff6 | 2422 | /* These take the mm semaphore themselves */ |
5d22fc25 | 2423 | extern int __must_check vm_brk(unsigned long, unsigned long); |
16e72e9b | 2424 | extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long); |
bfce281c | 2425 | extern int vm_munmap(unsigned long, size_t); |
9fbeb5ab | 2426 | extern unsigned long __must_check vm_mmap(struct file *, unsigned long, |
6be5ceb0 LT |
2427 | unsigned long, unsigned long, |
2428 | unsigned long, unsigned long); | |
1da177e4 | 2429 | |
db4fbfb9 ML |
2430 | struct vm_unmapped_area_info { |
2431 | #define VM_UNMAPPED_AREA_TOPDOWN 1 | |
2432 | unsigned long flags; | |
2433 | unsigned long length; | |
2434 | unsigned long low_limit; | |
2435 | unsigned long high_limit; | |
2436 | unsigned long align_mask; | |
2437 | unsigned long align_offset; | |
2438 | }; | |
2439 | ||
2440 | extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); | |
2441 | extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); | |
2442 | ||
2443 | /* | |
2444 | * Search for an unmapped address range. | |
2445 | * | |
2446 | * We are looking for a range that: | |
2447 | * - does not intersect with any VMA; | |
2448 | * - is contained within the [low_limit, high_limit) interval; | |
2449 | * - is at least the desired size. | |
2450 | * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) | |
2451 | */ | |
2452 | static inline unsigned long | |
2453 | vm_unmapped_area(struct vm_unmapped_area_info *info) | |
2454 | { | |
cdd7875e | 2455 | if (info->flags & VM_UNMAPPED_AREA_TOPDOWN) |
db4fbfb9 | 2456 | return unmapped_area_topdown(info); |
cdd7875e BP |
2457 | else |
2458 | return unmapped_area(info); | |
db4fbfb9 ML |
2459 | } |
2460 | ||
85821aab | 2461 | /* truncate.c */ |
1da177e4 | 2462 | extern void truncate_inode_pages(struct address_space *, loff_t); |
d7339071 HR |
2463 | extern void truncate_inode_pages_range(struct address_space *, |
2464 | loff_t lstart, loff_t lend); | |
91b0abe3 | 2465 | extern void truncate_inode_pages_final(struct address_space *); |
1da177e4 LT |
2466 | |
2467 | /* generic vm_area_ops exported for stackable file systems */ | |
2bcd6454 | 2468 | extern vm_fault_t filemap_fault(struct vm_fault *vmf); |
82b0f8c3 | 2469 | extern void filemap_map_pages(struct vm_fault *vmf, |
bae473a4 | 2470 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
2bcd6454 | 2471 | extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf); |
1da177e4 LT |
2472 | |
2473 | /* mm/page-writeback.c */ | |
2b69c828 | 2474 | int __must_check write_one_page(struct page *page); |
1cf6e7d8 | 2475 | void task_dirty_inc(struct task_struct *tsk); |
1da177e4 LT |
2476 | |
2477 | /* readahead.c */ | |
b5420237 | 2478 | #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE) |
1da177e4 | 2479 | |
1da177e4 | 2480 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
7361f4d8 | 2481 | pgoff_t offset, unsigned long nr_to_read); |
cf914a7d RR |
2482 | |
2483 | void page_cache_sync_readahead(struct address_space *mapping, | |
2484 | struct file_ra_state *ra, | |
2485 | struct file *filp, | |
2486 | pgoff_t offset, | |
2487 | unsigned long size); | |
2488 | ||
2489 | void page_cache_async_readahead(struct address_space *mapping, | |
2490 | struct file_ra_state *ra, | |
2491 | struct file *filp, | |
2492 | struct page *pg, | |
2493 | pgoff_t offset, | |
2494 | unsigned long size); | |
2495 | ||
1be7107f | 2496 | extern unsigned long stack_guard_gap; |
d05f3169 | 2497 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ |
46dea3d0 | 2498 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
d05f3169 MH |
2499 | |
2500 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ | |
2501 | extern int expand_downwards(struct vm_area_struct *vma, | |
2502 | unsigned long address); | |
8ca3eb08 | 2503 | #if VM_GROWSUP |
46dea3d0 | 2504 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
8ca3eb08 | 2505 | #else |
fee7e49d | 2506 | #define expand_upwards(vma, address) (0) |
9ab88515 | 2507 | #endif |
1da177e4 LT |
2508 | |
2509 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
2510 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
2511 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
2512 | struct vm_area_struct **pprev); | |
2513 | ||
2514 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
2515 | NULL if none. Assume start_addr < end_addr. */ | |
2516 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
2517 | { | |
2518 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
2519 | ||
2520 | if (vma && end_addr <= vma->vm_start) | |
2521 | vma = NULL; | |
2522 | return vma; | |
2523 | } | |
2524 | ||
1be7107f HD |
2525 | static inline unsigned long vm_start_gap(struct vm_area_struct *vma) |
2526 | { | |
2527 | unsigned long vm_start = vma->vm_start; | |
2528 | ||
2529 | if (vma->vm_flags & VM_GROWSDOWN) { | |
2530 | vm_start -= stack_guard_gap; | |
2531 | if (vm_start > vma->vm_start) | |
2532 | vm_start = 0; | |
2533 | } | |
2534 | return vm_start; | |
2535 | } | |
2536 | ||
2537 | static inline unsigned long vm_end_gap(struct vm_area_struct *vma) | |
2538 | { | |
2539 | unsigned long vm_end = vma->vm_end; | |
2540 | ||
2541 | if (vma->vm_flags & VM_GROWSUP) { | |
2542 | vm_end += stack_guard_gap; | |
2543 | if (vm_end < vma->vm_end) | |
2544 | vm_end = -PAGE_SIZE; | |
2545 | } | |
2546 | return vm_end; | |
2547 | } | |
2548 | ||
1da177e4 LT |
2549 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
2550 | { | |
2551 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
2552 | } | |
2553 | ||
640708a2 PE |
2554 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ |
2555 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, | |
2556 | unsigned long vm_start, unsigned long vm_end) | |
2557 | { | |
2558 | struct vm_area_struct *vma = find_vma(mm, vm_start); | |
2559 | ||
2560 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) | |
2561 | vma = NULL; | |
2562 | ||
2563 | return vma; | |
2564 | } | |
2565 | ||
017b1660 MK |
2566 | static inline bool range_in_vma(struct vm_area_struct *vma, |
2567 | unsigned long start, unsigned long end) | |
2568 | { | |
2569 | return (vma && vma->vm_start <= start && end <= vma->vm_end); | |
2570 | } | |
2571 | ||
bad849b3 | 2572 | #ifdef CONFIG_MMU |
804af2cf | 2573 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
64e45507 | 2574 | void vma_set_page_prot(struct vm_area_struct *vma); |
bad849b3 DH |
2575 | #else |
2576 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | |
2577 | { | |
2578 | return __pgprot(0); | |
2579 | } | |
64e45507 PF |
2580 | static inline void vma_set_page_prot(struct vm_area_struct *vma) |
2581 | { | |
2582 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); | |
2583 | } | |
bad849b3 DH |
2584 | #endif |
2585 | ||
5877231f | 2586 | #ifdef CONFIG_NUMA_BALANCING |
4b10e7d5 | 2587 | unsigned long change_prot_numa(struct vm_area_struct *vma, |
b24f53a0 LS |
2588 | unsigned long start, unsigned long end); |
2589 | #endif | |
2590 | ||
deceb6cd | 2591 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
deceb6cd HD |
2592 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
2593 | unsigned long pfn, unsigned long size, pgprot_t); | |
a145dd41 | 2594 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
a667d745 SJ |
2595 | int vm_map_pages(struct vm_area_struct *vma, struct page **pages, |
2596 | unsigned long num); | |
2597 | int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages, | |
2598 | unsigned long num); | |
ae2b01f3 | 2599 | vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
e0dc0d8f | 2600 | unsigned long pfn); |
f5e6d1d5 MW |
2601 | vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr, |
2602 | unsigned long pfn, pgprot_t pgprot); | |
5d747637 | 2603 | vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
01c8f1c4 | 2604 | pfn_t pfn); |
ab77dab4 SJ |
2605 | vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma, |
2606 | unsigned long addr, pfn_t pfn); | |
b4cbb197 LT |
2607 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); |
2608 | ||
1c8f4220 SJ |
2609 | static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma, |
2610 | unsigned long addr, struct page *page) | |
2611 | { | |
2612 | int err = vm_insert_page(vma, addr, page); | |
2613 | ||
2614 | if (err == -ENOMEM) | |
2615 | return VM_FAULT_OOM; | |
2616 | if (err < 0 && err != -EBUSY) | |
2617 | return VM_FAULT_SIGBUS; | |
2618 | ||
2619 | return VM_FAULT_NOPAGE; | |
2620 | } | |
2621 | ||
d97baf94 SJ |
2622 | static inline vm_fault_t vmf_error(int err) |
2623 | { | |
2624 | if (err == -ENOMEM) | |
2625 | return VM_FAULT_OOM; | |
2626 | return VM_FAULT_SIGBUS; | |
2627 | } | |
2628 | ||
df06b37f KB |
2629 | struct page *follow_page(struct vm_area_struct *vma, unsigned long address, |
2630 | unsigned int foll_flags); | |
240aadee | 2631 | |
deceb6cd HD |
2632 | #define FOLL_WRITE 0x01 /* check pte is writable */ |
2633 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
2634 | #define FOLL_GET 0x04 /* do get_page on page */ | |
8e4b9a60 | 2635 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ |
58fa879e | 2636 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ |
318b275f GN |
2637 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO |
2638 | * and return without waiting upon it */ | |
84d33df2 | 2639 | #define FOLL_POPULATE 0x40 /* fault in page */ |
500d65d4 | 2640 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ |
69ebb83e | 2641 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ |
0b9d7052 | 2642 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ |
5117b3b8 | 2643 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ |
234b239b | 2644 | #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */ |
de60f5f1 | 2645 | #define FOLL_MLOCK 0x1000 /* lock present pages */ |
1e987790 | 2646 | #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */ |
19be0eaf | 2647 | #define FOLL_COW 0x4000 /* internal GUP flag */ |
7f7ccc2c | 2648 | #define FOLL_ANON 0x8000 /* don't do file mappings */ |
932f4a63 IW |
2649 | #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */ |
2650 | ||
2651 | /* | |
2652 | * NOTE on FOLL_LONGTERM: | |
2653 | * | |
2654 | * FOLL_LONGTERM indicates that the page will be held for an indefinite time | |
2655 | * period _often_ under userspace control. This is contrasted with | |
2656 | * iov_iter_get_pages() where usages which are transient. | |
2657 | * | |
2658 | * FIXME: For pages which are part of a filesystem, mappings are subject to the | |
2659 | * lifetime enforced by the filesystem and we need guarantees that longterm | |
2660 | * users like RDMA and V4L2 only establish mappings which coordinate usage with | |
2661 | * the filesystem. Ideas for this coordination include revoking the longterm | |
2662 | * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was | |
2663 | * added after the problem with filesystems was found FS DAX VMAs are | |
2664 | * specifically failed. Filesystem pages are still subject to bugs and use of | |
2665 | * FOLL_LONGTERM should be avoided on those pages. | |
2666 | * | |
2667 | * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call. | |
2668 | * Currently only get_user_pages() and get_user_pages_fast() support this flag | |
2669 | * and calls to get_user_pages_[un]locked are specifically not allowed. This | |
2670 | * is due to an incompatibility with the FS DAX check and | |
2671 | * FAULT_FLAG_ALLOW_RETRY | |
2672 | * | |
2673 | * In the CMA case: longterm pins in a CMA region would unnecessarily fragment | |
2674 | * that region. And so CMA attempts to migrate the page before pinning when | |
2675 | * FOLL_LONGTERM is specified. | |
2676 | */ | |
1da177e4 | 2677 | |
2b740303 | 2678 | static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags) |
9a291a7c JM |
2679 | { |
2680 | if (vm_fault & VM_FAULT_OOM) | |
2681 | return -ENOMEM; | |
2682 | if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) | |
2683 | return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT; | |
2684 | if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV)) | |
2685 | return -EFAULT; | |
2686 | return 0; | |
2687 | } | |
2688 | ||
2f569afd | 2689 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, |
aee16b3c JF |
2690 | void *data); |
2691 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | |
2692 | unsigned long size, pte_fn_t fn, void *data); | |
2693 | ||
1da177e4 | 2694 | |
8823b1db LA |
2695 | #ifdef CONFIG_PAGE_POISONING |
2696 | extern bool page_poisoning_enabled(void); | |
2697 | extern void kernel_poison_pages(struct page *page, int numpages, int enable); | |
2698 | #else | |
2699 | static inline bool page_poisoning_enabled(void) { return false; } | |
2700 | static inline void kernel_poison_pages(struct page *page, int numpages, | |
2701 | int enable) { } | |
2702 | #endif | |
2703 | ||
96a2b03f VB |
2704 | #ifdef CONFIG_DEBUG_PAGEALLOC_ENABLE_DEFAULT |
2705 | DECLARE_STATIC_KEY_TRUE(_debug_pagealloc_enabled); | |
2706 | #else | |
2707 | DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled); | |
2708 | #endif | |
031bc574 JK |
2709 | |
2710 | static inline bool debug_pagealloc_enabled(void) | |
2711 | { | |
96a2b03f VB |
2712 | if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC)) |
2713 | return false; | |
2714 | ||
2715 | return static_branch_unlikely(&_debug_pagealloc_enabled); | |
031bc574 JK |
2716 | } |
2717 | ||
d6332692 RE |
2718 | #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP) |
2719 | extern void __kernel_map_pages(struct page *page, int numpages, int enable); | |
2720 | ||
031bc574 JK |
2721 | static inline void |
2722 | kernel_map_pages(struct page *page, int numpages, int enable) | |
2723 | { | |
031bc574 JK |
2724 | __kernel_map_pages(page, numpages, enable); |
2725 | } | |
8a235efa RW |
2726 | #ifdef CONFIG_HIBERNATION |
2727 | extern bool kernel_page_present(struct page *page); | |
40b44137 | 2728 | #endif /* CONFIG_HIBERNATION */ |
d6332692 | 2729 | #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */ |
1da177e4 | 2730 | static inline void |
9858db50 | 2731 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
8a235efa RW |
2732 | #ifdef CONFIG_HIBERNATION |
2733 | static inline bool kernel_page_present(struct page *page) { return true; } | |
40b44137 | 2734 | #endif /* CONFIG_HIBERNATION */ |
d6332692 | 2735 | #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */ |
1da177e4 | 2736 | |
a6c19dfe | 2737 | #ifdef __HAVE_ARCH_GATE_AREA |
31db58b3 | 2738 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
a6c19dfe AL |
2739 | extern int in_gate_area_no_mm(unsigned long addr); |
2740 | extern int in_gate_area(struct mm_struct *mm, unsigned long addr); | |
1da177e4 | 2741 | #else |
a6c19dfe AL |
2742 | static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) |
2743 | { | |
2744 | return NULL; | |
2745 | } | |
2746 | static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } | |
2747 | static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) | |
2748 | { | |
2749 | return 0; | |
2750 | } | |
1da177e4 LT |
2751 | #endif /* __HAVE_ARCH_GATE_AREA */ |
2752 | ||
44a70ade MH |
2753 | extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm); |
2754 | ||
146732ce JT |
2755 | #ifdef CONFIG_SYSCTL |
2756 | extern int sysctl_drop_caches; | |
8d65af78 | 2757 | int drop_caches_sysctl_handler(struct ctl_table *, int, |
9d0243bc | 2758 | void __user *, size_t *, loff_t *); |
146732ce JT |
2759 | #endif |
2760 | ||
cb731d6c VD |
2761 | void drop_slab(void); |
2762 | void drop_slab_node(int nid); | |
9d0243bc | 2763 | |
7a9166e3 LY |
2764 | #ifndef CONFIG_MMU |
2765 | #define randomize_va_space 0 | |
2766 | #else | |
a62eaf15 | 2767 | extern int randomize_va_space; |
7a9166e3 | 2768 | #endif |
a62eaf15 | 2769 | |
045e72ac | 2770 | const char * arch_vma_name(struct vm_area_struct *vma); |
03252919 | 2771 | void print_vma_addr(char *prefix, unsigned long rip); |
e6e5494c | 2772 | |
35fd1eb1 | 2773 | void *sparse_buffer_alloc(unsigned long size); |
7b73d978 CH |
2774 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid, |
2775 | struct vmem_altmap *altmap); | |
29c71111 | 2776 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
c2febafc KS |
2777 | p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node); |
2778 | pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node); | |
29c71111 AW |
2779 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); |
2780 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); | |
8f6aac41 | 2781 | void *vmemmap_alloc_block(unsigned long size, int node); |
4b94ffdc | 2782 | struct vmem_altmap; |
a8fc357b CH |
2783 | void *vmemmap_alloc_block_buf(unsigned long size, int node); |
2784 | void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap); | |
8f6aac41 | 2785 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
0aad818b JW |
2786 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, |
2787 | int node); | |
7b73d978 CH |
2788 | int vmemmap_populate(unsigned long start, unsigned long end, int node, |
2789 | struct vmem_altmap *altmap); | |
c2b91e2e | 2790 | void vmemmap_populate_print_last(void); |
0197518c | 2791 | #ifdef CONFIG_MEMORY_HOTPLUG |
24b6d416 CH |
2792 | void vmemmap_free(unsigned long start, unsigned long end, |
2793 | struct vmem_altmap *altmap); | |
0197518c | 2794 | #endif |
46723bfa | 2795 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, |
15670bfe | 2796 | unsigned long nr_pages); |
6a46079c | 2797 | |
82ba011b AK |
2798 | enum mf_flags { |
2799 | MF_COUNT_INCREASED = 1 << 0, | |
7329bbeb | 2800 | MF_ACTION_REQUIRED = 1 << 1, |
6751ed65 | 2801 | MF_MUST_KILL = 1 << 2, |
cf870c70 | 2802 | MF_SOFT_OFFLINE = 1 << 3, |
82ba011b | 2803 | }; |
83b57531 EB |
2804 | extern int memory_failure(unsigned long pfn, int flags); |
2805 | extern void memory_failure_queue(unsigned long pfn, int flags); | |
847ce401 | 2806 | extern int unpoison_memory(unsigned long pfn); |
ead07f6a | 2807 | extern int get_hwpoison_page(struct page *page); |
4e41a30c | 2808 | #define put_hwpoison_page(page) put_page(page) |
6a46079c AK |
2809 | extern int sysctl_memory_failure_early_kill; |
2810 | extern int sysctl_memory_failure_recovery; | |
facb6011 | 2811 | extern void shake_page(struct page *p, int access); |
5844a486 | 2812 | extern atomic_long_t num_poisoned_pages __read_mostly; |
facb6011 | 2813 | extern int soft_offline_page(struct page *page, int flags); |
6a46079c | 2814 | |
cc637b17 XX |
2815 | |
2816 | /* | |
2817 | * Error handlers for various types of pages. | |
2818 | */ | |
cc3e2af4 | 2819 | enum mf_result { |
cc637b17 XX |
2820 | MF_IGNORED, /* Error: cannot be handled */ |
2821 | MF_FAILED, /* Error: handling failed */ | |
2822 | MF_DELAYED, /* Will be handled later */ | |
2823 | MF_RECOVERED, /* Successfully recovered */ | |
2824 | }; | |
2825 | ||
2826 | enum mf_action_page_type { | |
2827 | MF_MSG_KERNEL, | |
2828 | MF_MSG_KERNEL_HIGH_ORDER, | |
2829 | MF_MSG_SLAB, | |
2830 | MF_MSG_DIFFERENT_COMPOUND, | |
2831 | MF_MSG_POISONED_HUGE, | |
2832 | MF_MSG_HUGE, | |
2833 | MF_MSG_FREE_HUGE, | |
31286a84 | 2834 | MF_MSG_NON_PMD_HUGE, |
cc637b17 XX |
2835 | MF_MSG_UNMAP_FAILED, |
2836 | MF_MSG_DIRTY_SWAPCACHE, | |
2837 | MF_MSG_CLEAN_SWAPCACHE, | |
2838 | MF_MSG_DIRTY_MLOCKED_LRU, | |
2839 | MF_MSG_CLEAN_MLOCKED_LRU, | |
2840 | MF_MSG_DIRTY_UNEVICTABLE_LRU, | |
2841 | MF_MSG_CLEAN_UNEVICTABLE_LRU, | |
2842 | MF_MSG_DIRTY_LRU, | |
2843 | MF_MSG_CLEAN_LRU, | |
2844 | MF_MSG_TRUNCATED_LRU, | |
2845 | MF_MSG_BUDDY, | |
2846 | MF_MSG_BUDDY_2ND, | |
6100e34b | 2847 | MF_MSG_DAX, |
cc637b17 XX |
2848 | MF_MSG_UNKNOWN, |
2849 | }; | |
2850 | ||
47ad8475 AA |
2851 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
2852 | extern void clear_huge_page(struct page *page, | |
c79b57e4 | 2853 | unsigned long addr_hint, |
47ad8475 AA |
2854 | unsigned int pages_per_huge_page); |
2855 | extern void copy_user_huge_page(struct page *dst, struct page *src, | |
c9f4cd71 HY |
2856 | unsigned long addr_hint, |
2857 | struct vm_area_struct *vma, | |
47ad8475 | 2858 | unsigned int pages_per_huge_page); |
fa4d75c1 MK |
2859 | extern long copy_huge_page_from_user(struct page *dst_page, |
2860 | const void __user *usr_src, | |
810a56b9 MK |
2861 | unsigned int pages_per_huge_page, |
2862 | bool allow_pagefault); | |
47ad8475 AA |
2863 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ |
2864 | ||
e30825f1 | 2865 | extern struct page_ext_operations debug_guardpage_ops; |
e30825f1 | 2866 | |
c0a32fc5 SG |
2867 | #ifdef CONFIG_DEBUG_PAGEALLOC |
2868 | extern unsigned int _debug_guardpage_minorder; | |
96a2b03f | 2869 | DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled); |
c0a32fc5 SG |
2870 | |
2871 | static inline unsigned int debug_guardpage_minorder(void) | |
2872 | { | |
2873 | return _debug_guardpage_minorder; | |
2874 | } | |
2875 | ||
e30825f1 JK |
2876 | static inline bool debug_guardpage_enabled(void) |
2877 | { | |
96a2b03f | 2878 | return static_branch_unlikely(&_debug_guardpage_enabled); |
e30825f1 JK |
2879 | } |
2880 | ||
c0a32fc5 SG |
2881 | static inline bool page_is_guard(struct page *page) |
2882 | { | |
e30825f1 JK |
2883 | struct page_ext *page_ext; |
2884 | ||
2885 | if (!debug_guardpage_enabled()) | |
2886 | return false; | |
2887 | ||
2888 | page_ext = lookup_page_ext(page); | |
0bb2fd13 YS |
2889 | if (unlikely(!page_ext)) |
2890 | return false; | |
2891 | ||
e30825f1 | 2892 | return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags); |
c0a32fc5 SG |
2893 | } |
2894 | #else | |
2895 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } | |
e30825f1 | 2896 | static inline bool debug_guardpage_enabled(void) { return false; } |
c0a32fc5 SG |
2897 | static inline bool page_is_guard(struct page *page) { return false; } |
2898 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | |
2899 | ||
f9872caf CS |
2900 | #if MAX_NUMNODES > 1 |
2901 | void __init setup_nr_node_ids(void); | |
2902 | #else | |
2903 | static inline void setup_nr_node_ids(void) {} | |
2904 | #endif | |
2905 | ||
1da177e4 LT |
2906 | #endif /* __KERNEL__ */ |
2907 | #endif /* _LINUX_MM_H */ |