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> |
9740ca4e | 18 | #include <linux/mmap_lock.h> |
08677214 | 19 | #include <linux/range.h> |
c6f6b596 | 20 | #include <linux/pfn.h> |
3565fce3 | 21 | #include <linux/percpu-refcount.h> |
e9da73d6 | 22 | #include <linux/bit_spinlock.h> |
b0d40c92 | 23 | #include <linux/shrinker.h> |
9c599024 | 24 | #include <linux/resource.h> |
e30825f1 | 25 | #include <linux/page_ext.h> |
8025e5dd | 26 | #include <linux/err.h> |
41901567 | 27 | #include <linux/page-flags.h> |
fe896d18 | 28 | #include <linux/page_ref.h> |
7b2d55d2 | 29 | #include <linux/memremap.h> |
3b3b1a29 | 30 | #include <linux/overflow.h> |
b5420237 | 31 | #include <linux/sizes.h> |
7969f226 | 32 | #include <linux/sched.h> |
65fddcfc | 33 | #include <linux/pgtable.h> |
34303244 | 34 | #include <linux/kasan.h> |
1da177e4 LT |
35 | |
36 | struct mempolicy; | |
37 | struct anon_vma; | |
bf181b9f | 38 | struct anon_vma_chain; |
4e950f6f | 39 | struct file_ra_state; |
e8edc6e0 | 40 | struct user_struct; |
4e950f6f | 41 | struct writeback_control; |
682aa8e1 | 42 | struct bdi_writeback; |
bce617ed | 43 | struct pt_regs; |
1da177e4 | 44 | |
5ef64cc8 LT |
45 | extern int sysctl_page_lock_unfairness; |
46 | ||
597b7305 MH |
47 | void init_mm_internals(void); |
48 | ||
fccc9987 | 49 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ |
1da177e4 | 50 | extern unsigned long max_mapnr; |
fccc9987 JL |
51 | |
52 | static inline void set_max_mapnr(unsigned long limit) | |
53 | { | |
54 | max_mapnr = limit; | |
55 | } | |
56 | #else | |
57 | static inline void set_max_mapnr(unsigned long limit) { } | |
1da177e4 LT |
58 | #endif |
59 | ||
ca79b0c2 AK |
60 | extern atomic_long_t _totalram_pages; |
61 | static inline unsigned long totalram_pages(void) | |
62 | { | |
63 | return (unsigned long)atomic_long_read(&_totalram_pages); | |
64 | } | |
65 | ||
66 | static inline void totalram_pages_inc(void) | |
67 | { | |
68 | atomic_long_inc(&_totalram_pages); | |
69 | } | |
70 | ||
71 | static inline void totalram_pages_dec(void) | |
72 | { | |
73 | atomic_long_dec(&_totalram_pages); | |
74 | } | |
75 | ||
76 | static inline void totalram_pages_add(long count) | |
77 | { | |
78 | atomic_long_add(count, &_totalram_pages); | |
79 | } | |
80 | ||
1da177e4 | 81 | extern void * high_memory; |
1da177e4 LT |
82 | extern int page_cluster; |
83 | ||
84 | #ifdef CONFIG_SYSCTL | |
85 | extern int sysctl_legacy_va_layout; | |
86 | #else | |
87 | #define sysctl_legacy_va_layout 0 | |
88 | #endif | |
89 | ||
d07e2259 DC |
90 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS |
91 | extern const int mmap_rnd_bits_min; | |
92 | extern const int mmap_rnd_bits_max; | |
93 | extern int mmap_rnd_bits __read_mostly; | |
94 | #endif | |
95 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS | |
96 | extern const int mmap_rnd_compat_bits_min; | |
97 | extern const int mmap_rnd_compat_bits_max; | |
98 | extern int mmap_rnd_compat_bits __read_mostly; | |
99 | #endif | |
100 | ||
1da177e4 | 101 | #include <asm/page.h> |
1da177e4 | 102 | #include <asm/processor.h> |
1da177e4 | 103 | |
d9344522 AK |
104 | /* |
105 | * Architectures that support memory tagging (assigning tags to memory regions, | |
106 | * embedding these tags into addresses that point to these memory regions, and | |
107 | * checking that the memory and the pointer tags match on memory accesses) | |
108 | * redefine this macro to strip tags from pointers. | |
f0953a1b | 109 | * It's defined as noop for architectures that don't support memory tagging. |
d9344522 AK |
110 | */ |
111 | #ifndef untagged_addr | |
112 | #define untagged_addr(addr) (addr) | |
113 | #endif | |
114 | ||
79442ed1 TC |
115 | #ifndef __pa_symbol |
116 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) | |
117 | #endif | |
118 | ||
1dff8083 AB |
119 | #ifndef page_to_virt |
120 | #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x))) | |
121 | #endif | |
122 | ||
568c5fe5 LA |
123 | #ifndef lm_alias |
124 | #define lm_alias(x) __va(__pa_symbol(x)) | |
125 | #endif | |
126 | ||
5caf9682 ST |
127 | /* |
128 | * With CONFIG_CFI_CLANG, the compiler replaces function addresses in | |
129 | * instrumented C code with jump table addresses. Architectures that | |
130 | * support CFI can define this macro to return the actual function address | |
131 | * when needed. | |
132 | */ | |
133 | #ifndef function_nocfi | |
134 | #define function_nocfi(x) (x) | |
135 | #endif | |
136 | ||
593befa6 DD |
137 | /* |
138 | * To prevent common memory management code establishing | |
139 | * a zero page mapping on a read fault. | |
140 | * This macro should be defined within <asm/pgtable.h>. | |
141 | * s390 does this to prevent multiplexing of hardware bits | |
142 | * related to the physical page in case of virtualization. | |
143 | */ | |
144 | #ifndef mm_forbids_zeropage | |
145 | #define mm_forbids_zeropage(X) (0) | |
146 | #endif | |
147 | ||
a4a3ede2 PT |
148 | /* |
149 | * On some architectures it is expensive to call memset() for small sizes. | |
5470dea4 AD |
150 | * If an architecture decides to implement their own version of |
151 | * mm_zero_struct_page they should wrap the defines below in a #ifndef and | |
152 | * define their own version of this macro in <asm/pgtable.h> | |
a4a3ede2 | 153 | */ |
5470dea4 AD |
154 | #if BITS_PER_LONG == 64 |
155 | /* This function must be updated when the size of struct page grows above 80 | |
156 | * or reduces below 56. The idea that compiler optimizes out switch() | |
157 | * statement, and only leaves move/store instructions. Also the compiler can | |
158 | * combine write statments if they are both assignments and can be reordered, | |
159 | * this can result in several of the writes here being dropped. | |
160 | */ | |
161 | #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp) | |
162 | static inline void __mm_zero_struct_page(struct page *page) | |
163 | { | |
164 | unsigned long *_pp = (void *)page; | |
165 | ||
166 | /* Check that struct page is either 56, 64, 72, or 80 bytes */ | |
167 | BUILD_BUG_ON(sizeof(struct page) & 7); | |
168 | BUILD_BUG_ON(sizeof(struct page) < 56); | |
169 | BUILD_BUG_ON(sizeof(struct page) > 80); | |
170 | ||
171 | switch (sizeof(struct page)) { | |
172 | case 80: | |
df561f66 GS |
173 | _pp[9] = 0; |
174 | fallthrough; | |
5470dea4 | 175 | case 72: |
df561f66 GS |
176 | _pp[8] = 0; |
177 | fallthrough; | |
5470dea4 | 178 | case 64: |
df561f66 GS |
179 | _pp[7] = 0; |
180 | fallthrough; | |
5470dea4 AD |
181 | case 56: |
182 | _pp[6] = 0; | |
183 | _pp[5] = 0; | |
184 | _pp[4] = 0; | |
185 | _pp[3] = 0; | |
186 | _pp[2] = 0; | |
187 | _pp[1] = 0; | |
188 | _pp[0] = 0; | |
189 | } | |
190 | } | |
191 | #else | |
a4a3ede2 PT |
192 | #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page))) |
193 | #endif | |
194 | ||
ea606cf5 AR |
195 | /* |
196 | * Default maximum number of active map areas, this limits the number of vmas | |
197 | * per mm struct. Users can overwrite this number by sysctl but there is a | |
198 | * problem. | |
199 | * | |
200 | * When a program's coredump is generated as ELF format, a section is created | |
201 | * per a vma. In ELF, the number of sections is represented in unsigned short. | |
202 | * This means the number of sections should be smaller than 65535 at coredump. | |
203 | * Because the kernel adds some informative sections to a image of program at | |
204 | * generating coredump, we need some margin. The number of extra sections is | |
205 | * 1-3 now and depends on arch. We use "5" as safe margin, here. | |
206 | * | |
207 | * ELF extended numbering allows more than 65535 sections, so 16-bit bound is | |
208 | * not a hard limit any more. Although some userspace tools can be surprised by | |
209 | * that. | |
210 | */ | |
211 | #define MAPCOUNT_ELF_CORE_MARGIN (5) | |
212 | #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) | |
213 | ||
214 | extern int sysctl_max_map_count; | |
215 | ||
c9b1d098 | 216 | extern unsigned long sysctl_user_reserve_kbytes; |
4eeab4f5 | 217 | extern unsigned long sysctl_admin_reserve_kbytes; |
c9b1d098 | 218 | |
49f0ce5f JM |
219 | extern int sysctl_overcommit_memory; |
220 | extern int sysctl_overcommit_ratio; | |
221 | extern unsigned long sysctl_overcommit_kbytes; | |
222 | ||
32927393 CH |
223 | int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *, |
224 | loff_t *); | |
225 | int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *, | |
226 | loff_t *); | |
56f3547b FT |
227 | int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *, |
228 | loff_t *); | |
6d87d0ec SJ |
229 | /* |
230 | * Any attempt to mark this function as static leads to build failure | |
231 | * when CONFIG_DEBUG_INFO_BTF is enabled because __add_to_page_cache_locked() | |
232 | * is referred to by BPF code. This must be visible for error injection. | |
233 | */ | |
234 | int __add_to_page_cache_locked(struct page *page, struct address_space *mapping, | |
235 | pgoff_t index, gfp_t gfp, void **shadowp); | |
49f0ce5f | 236 | |
1da177e4 LT |
237 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
238 | ||
27ac792c AR |
239 | /* to align the pointer to the (next) page boundary */ |
240 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | |
241 | ||
0fa73b86 | 242 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ |
1061b0d2 | 243 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE) |
0fa73b86 | 244 | |
f86196ea NB |
245 | #define lru_to_page(head) (list_entry((head)->prev, struct page, lru)) |
246 | ||
1da177e4 LT |
247 | /* |
248 | * Linux kernel virtual memory manager primitives. | |
249 | * The idea being to have a "virtual" mm in the same way | |
250 | * we have a virtual fs - giving a cleaner interface to the | |
251 | * mm details, and allowing different kinds of memory mappings | |
252 | * (from shared memory to executable loading to arbitrary | |
253 | * mmap() functions). | |
254 | */ | |
255 | ||
490fc053 | 256 | struct vm_area_struct *vm_area_alloc(struct mm_struct *); |
3928d4f5 LT |
257 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *); |
258 | void vm_area_free(struct vm_area_struct *); | |
c43692e8 | 259 | |
1da177e4 | 260 | #ifndef CONFIG_MMU |
8feae131 DH |
261 | extern struct rb_root nommu_region_tree; |
262 | extern struct rw_semaphore nommu_region_sem; | |
1da177e4 LT |
263 | |
264 | extern unsigned int kobjsize(const void *objp); | |
265 | #endif | |
266 | ||
267 | /* | |
605d9288 | 268 | * vm_flags in vm_area_struct, see mm_types.h. |
bcf66917 | 269 | * When changing, update also include/trace/events/mmflags.h |
1da177e4 | 270 | */ |
cc2383ec KK |
271 | #define VM_NONE 0x00000000 |
272 | ||
1da177e4 LT |
273 | #define VM_READ 0x00000001 /* currently active flags */ |
274 | #define VM_WRITE 0x00000002 | |
275 | #define VM_EXEC 0x00000004 | |
276 | #define VM_SHARED 0x00000008 | |
277 | ||
7e2cff42 | 278 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
1da177e4 LT |
279 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
280 | #define VM_MAYWRITE 0x00000020 | |
281 | #define VM_MAYEXEC 0x00000040 | |
282 | #define VM_MAYSHARE 0x00000080 | |
283 | ||
284 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
16ba6f81 | 285 | #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */ |
6aab341e | 286 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
1da177e4 | 287 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
16ba6f81 | 288 | #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */ |
1da177e4 | 289 | |
1da177e4 LT |
290 | #define VM_LOCKED 0x00002000 |
291 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | |
292 | ||
293 | /* Used by sys_madvise() */ | |
294 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | |
295 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | |
296 | ||
297 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | |
298 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | |
de60f5f1 | 299 | #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */ |
1da177e4 | 300 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
cdfd4325 | 301 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
1da177e4 | 302 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
b6fb293f | 303 | #define VM_SYNC 0x00800000 /* Synchronous page faults */ |
cc2383ec | 304 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ |
d2cd9ede | 305 | #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */ |
0103bd16 | 306 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ |
d00806b1 | 307 | |
d9104d1c CG |
308 | #ifdef CONFIG_MEM_SOFT_DIRTY |
309 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ | |
310 | #else | |
311 | # define VM_SOFTDIRTY 0 | |
312 | #endif | |
313 | ||
b379d790 | 314 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
cc2383ec KK |
315 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ |
316 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ | |
f8af4da3 | 317 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
1da177e4 | 318 | |
63c17fb8 DH |
319 | #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS |
320 | #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */ | |
321 | #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */ | |
322 | #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */ | |
323 | #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */ | |
df3735c5 | 324 | #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */ |
63c17fb8 DH |
325 | #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0) |
326 | #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1) | |
327 | #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2) | |
328 | #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3) | |
df3735c5 | 329 | #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4) |
63c17fb8 DH |
330 | #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */ |
331 | ||
5212213a | 332 | #ifdef CONFIG_ARCH_HAS_PKEYS |
8f62c883 DH |
333 | # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0 |
334 | # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */ | |
2c9e0a6f | 335 | # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */ |
8f62c883 DH |
336 | # define VM_PKEY_BIT2 VM_HIGH_ARCH_2 |
337 | # define VM_PKEY_BIT3 VM_HIGH_ARCH_3 | |
2c9e0a6f RP |
338 | #ifdef CONFIG_PPC |
339 | # define VM_PKEY_BIT4 VM_HIGH_ARCH_4 | |
340 | #else | |
341 | # define VM_PKEY_BIT4 0 | |
8f62c883 | 342 | #endif |
5212213a RP |
343 | #endif /* CONFIG_ARCH_HAS_PKEYS */ |
344 | ||
345 | #if defined(CONFIG_X86) | |
346 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ | |
12564485 SA |
347 | #elif defined(CONFIG_PPC) |
348 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ | |
cc2383ec KK |
349 | #elif defined(CONFIG_PARISC) |
350 | # define VM_GROWSUP VM_ARCH_1 | |
351 | #elif defined(CONFIG_IA64) | |
352 | # define VM_GROWSUP VM_ARCH_1 | |
74a04967 KA |
353 | #elif defined(CONFIG_SPARC64) |
354 | # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */ | |
355 | # define VM_ARCH_CLEAR VM_SPARC_ADI | |
8ef8f360 DM |
356 | #elif defined(CONFIG_ARM64) |
357 | # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */ | |
358 | # define VM_ARCH_CLEAR VM_ARM64_BTI | |
cc2383ec KK |
359 | #elif !defined(CONFIG_MMU) |
360 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ | |
361 | #endif | |
362 | ||
9f341931 CM |
363 | #if defined(CONFIG_ARM64_MTE) |
364 | # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */ | |
365 | # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */ | |
366 | #else | |
367 | # define VM_MTE VM_NONE | |
368 | # define VM_MTE_ALLOWED VM_NONE | |
369 | #endif | |
370 | ||
cc2383ec KK |
371 | #ifndef VM_GROWSUP |
372 | # define VM_GROWSUP VM_NONE | |
373 | #endif | |
374 | ||
7677f7fd AR |
375 | #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR |
376 | # define VM_UFFD_MINOR_BIT 37 | |
377 | # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */ | |
378 | #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */ | |
379 | # define VM_UFFD_MINOR VM_NONE | |
380 | #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */ | |
381 | ||
a8bef8ff MG |
382 | /* Bits set in the VMA until the stack is in its final location */ |
383 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | |
384 | ||
c62da0c3 AK |
385 | #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0) |
386 | ||
387 | /* Common data flag combinations */ | |
388 | #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \ | |
389 | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) | |
390 | #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \ | |
391 | VM_MAYWRITE | VM_MAYEXEC) | |
392 | #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \ | |
393 | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) | |
394 | ||
395 | #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */ | |
396 | #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC | |
397 | #endif | |
398 | ||
1da177e4 LT |
399 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
400 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
401 | #endif | |
402 | ||
403 | #ifdef CONFIG_STACK_GROWSUP | |
30bdbb78 | 404 | #define VM_STACK VM_GROWSUP |
1da177e4 | 405 | #else |
30bdbb78 | 406 | #define VM_STACK VM_GROWSDOWN |
1da177e4 LT |
407 | #endif |
408 | ||
30bdbb78 KK |
409 | #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
410 | ||
6cb4d9a2 AK |
411 | /* VMA basic access permission flags */ |
412 | #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC) | |
413 | ||
414 | ||
b291f000 | 415 | /* |
78f11a25 | 416 | * Special vmas that are non-mergable, non-mlock()able. |
b291f000 | 417 | */ |
9050d7eb | 418 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) |
b291f000 | 419 | |
b4443772 AK |
420 | /* This mask prevents VMA from being scanned with khugepaged */ |
421 | #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB) | |
422 | ||
a0715cc2 AT |
423 | /* This mask defines which mm->def_flags a process can inherit its parent */ |
424 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE | |
425 | ||
de60f5f1 EM |
426 | /* This mask is used to clear all the VMA flags used by mlock */ |
427 | #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT)) | |
428 | ||
2c2d57b5 KA |
429 | /* Arch-specific flags to clear when updating VM flags on protection change */ |
430 | #ifndef VM_ARCH_CLEAR | |
431 | # define VM_ARCH_CLEAR VM_NONE | |
432 | #endif | |
433 | #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR) | |
434 | ||
1da177e4 LT |
435 | /* |
436 | * mapping from the currently active vm_flags protection bits (the | |
437 | * low four bits) to a page protection mask.. | |
438 | */ | |
439 | extern pgprot_t protection_map[16]; | |
440 | ||
c270a7ee | 441 | /** |
da2f5eb3 | 442 | * enum fault_flag - Fault flag definitions. |
c270a7ee PX |
443 | * @FAULT_FLAG_WRITE: Fault was a write fault. |
444 | * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE. | |
445 | * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked. | |
c1e8d7c6 | 446 | * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying. |
c270a7ee PX |
447 | * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region. |
448 | * @FAULT_FLAG_TRIED: The fault has been tried once. | |
449 | * @FAULT_FLAG_USER: The fault originated in userspace. | |
450 | * @FAULT_FLAG_REMOTE: The fault is not for current task/mm. | |
451 | * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch. | |
452 | * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals. | |
4064b982 PX |
453 | * |
454 | * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify | |
455 | * whether we would allow page faults to retry by specifying these two | |
456 | * fault flags correctly. Currently there can be three legal combinations: | |
457 | * | |
458 | * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and | |
459 | * this is the first try | |
460 | * | |
461 | * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and | |
462 | * we've already tried at least once | |
463 | * | |
464 | * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry | |
465 | * | |
466 | * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never | |
467 | * be used. Note that page faults can be allowed to retry for multiple times, | |
468 | * in which case we'll have an initial fault with flags (a) then later on | |
469 | * continuous faults with flags (b). We should always try to detect pending | |
470 | * signals before a retry to make sure the continuous page faults can still be | |
471 | * interrupted if necessary. | |
c270a7ee | 472 | */ |
da2f5eb3 MWO |
473 | enum fault_flag { |
474 | FAULT_FLAG_WRITE = 1 << 0, | |
475 | FAULT_FLAG_MKWRITE = 1 << 1, | |
476 | FAULT_FLAG_ALLOW_RETRY = 1 << 2, | |
477 | FAULT_FLAG_RETRY_NOWAIT = 1 << 3, | |
478 | FAULT_FLAG_KILLABLE = 1 << 4, | |
479 | FAULT_FLAG_TRIED = 1 << 5, | |
480 | FAULT_FLAG_USER = 1 << 6, | |
481 | FAULT_FLAG_REMOTE = 1 << 7, | |
482 | FAULT_FLAG_INSTRUCTION = 1 << 8, | |
483 | FAULT_FLAG_INTERRUPTIBLE = 1 << 9, | |
484 | }; | |
d0217ac0 | 485 | |
dde16072 PX |
486 | /* |
487 | * The default fault flags that should be used by most of the | |
488 | * arch-specific page fault handlers. | |
489 | */ | |
490 | #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \ | |
c270a7ee PX |
491 | FAULT_FLAG_KILLABLE | \ |
492 | FAULT_FLAG_INTERRUPTIBLE) | |
dde16072 | 493 | |
4064b982 PX |
494 | /** |
495 | * fault_flag_allow_retry_first - check ALLOW_RETRY the first time | |
78f4841e | 496 | * @flags: Fault flags. |
4064b982 PX |
497 | * |
498 | * This is mostly used for places where we want to try to avoid taking | |
c1e8d7c6 | 499 | * the mmap_lock for too long a time when waiting for another condition |
4064b982 | 500 | * to change, in which case we can try to be polite to release the |
c1e8d7c6 ML |
501 | * mmap_lock in the first round to avoid potential starvation of other |
502 | * processes that would also want the mmap_lock. | |
4064b982 PX |
503 | * |
504 | * Return: true if the page fault allows retry and this is the first | |
505 | * attempt of the fault handling; false otherwise. | |
506 | */ | |
da2f5eb3 | 507 | static inline bool fault_flag_allow_retry_first(enum fault_flag flags) |
4064b982 PX |
508 | { |
509 | return (flags & FAULT_FLAG_ALLOW_RETRY) && | |
510 | (!(flags & FAULT_FLAG_TRIED)); | |
511 | } | |
512 | ||
282a8e03 RZ |
513 | #define FAULT_FLAG_TRACE \ |
514 | { FAULT_FLAG_WRITE, "WRITE" }, \ | |
515 | { FAULT_FLAG_MKWRITE, "MKWRITE" }, \ | |
516 | { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \ | |
517 | { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \ | |
518 | { FAULT_FLAG_KILLABLE, "KILLABLE" }, \ | |
519 | { FAULT_FLAG_TRIED, "TRIED" }, \ | |
520 | { FAULT_FLAG_USER, "USER" }, \ | |
521 | { FAULT_FLAG_REMOTE, "REMOTE" }, \ | |
c270a7ee PX |
522 | { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \ |
523 | { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" } | |
282a8e03 | 524 | |
54cb8821 | 525 | /* |
11192337 | 526 | * vm_fault is filled by the pagefault handler and passed to the vma's |
83c54070 NP |
527 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
528 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
54cb8821 | 529 | * |
c20cd45e MH |
530 | * MM layer fills up gfp_mask for page allocations but fault handler might |
531 | * alter it if its implementation requires a different allocation context. | |
532 | * | |
9b4bdd2f | 533 | * pgoff should be used in favour of virtual_address, if possible. |
54cb8821 | 534 | */ |
d0217ac0 | 535 | struct vm_fault { |
5857c920 | 536 | const struct { |
742d3372 WD |
537 | struct vm_area_struct *vma; /* Target VMA */ |
538 | gfp_t gfp_mask; /* gfp mask to be used for allocations */ | |
539 | pgoff_t pgoff; /* Logical page offset based on vma */ | |
540 | unsigned long address; /* Faulting virtual address */ | |
541 | }; | |
da2f5eb3 | 542 | enum fault_flag flags; /* FAULT_FLAG_xxx flags |
742d3372 | 543 | * XXX: should really be 'const' */ |
82b0f8c3 | 544 | pmd_t *pmd; /* Pointer to pmd entry matching |
2994302b | 545 | * the 'address' */ |
a2d58167 DJ |
546 | pud_t *pud; /* Pointer to pud entry matching |
547 | * the 'address' | |
548 | */ | |
2994302b | 549 | pte_t orig_pte; /* Value of PTE at the time of fault */ |
d0217ac0 | 550 | |
3917048d | 551 | struct page *cow_page; /* Page handler may use for COW fault */ |
d0217ac0 | 552 | struct page *page; /* ->fault handlers should return a |
83c54070 | 553 | * page here, unless VM_FAULT_NOPAGE |
d0217ac0 | 554 | * is set (which is also implied by |
83c54070 | 555 | * VM_FAULT_ERROR). |
d0217ac0 | 556 | */ |
82b0f8c3 | 557 | /* These three entries are valid only while holding ptl lock */ |
bae473a4 KS |
558 | pte_t *pte; /* Pointer to pte entry matching |
559 | * the 'address'. NULL if the page | |
560 | * table hasn't been allocated. | |
561 | */ | |
562 | spinlock_t *ptl; /* Page table lock. | |
563 | * Protects pte page table if 'pte' | |
564 | * is not NULL, otherwise pmd. | |
565 | */ | |
7267ec00 | 566 | pgtable_t prealloc_pte; /* Pre-allocated pte page table. |
f9ce0be7 KS |
567 | * vm_ops->map_pages() sets up a page |
568 | * table from atomic context. | |
7267ec00 KS |
569 | * do_fault_around() pre-allocates |
570 | * page table to avoid allocation from | |
571 | * atomic context. | |
572 | */ | |
54cb8821 | 573 | }; |
1da177e4 | 574 | |
c791ace1 DJ |
575 | /* page entry size for vm->huge_fault() */ |
576 | enum page_entry_size { | |
577 | PE_SIZE_PTE = 0, | |
578 | PE_SIZE_PMD, | |
579 | PE_SIZE_PUD, | |
580 | }; | |
581 | ||
1da177e4 LT |
582 | /* |
583 | * These are the virtual MM functions - opening of an area, closing and | |
584 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
27d036e3 | 585 | * to the functions called when a no-page or a wp-page exception occurs. |
1da177e4 LT |
586 | */ |
587 | struct vm_operations_struct { | |
588 | void (*open)(struct vm_area_struct * area); | |
589 | void (*close)(struct vm_area_struct * area); | |
dd3b614f DS |
590 | /* Called any time before splitting to check if it's allowed */ |
591 | int (*may_split)(struct vm_area_struct *area, unsigned long addr); | |
14d07113 | 592 | int (*mremap)(struct vm_area_struct *area); |
95bb7c42 SC |
593 | /* |
594 | * Called by mprotect() to make driver-specific permission | |
595 | * checks before mprotect() is finalised. The VMA must not | |
596 | * be modified. Returns 0 if eprotect() can proceed. | |
597 | */ | |
598 | int (*mprotect)(struct vm_area_struct *vma, unsigned long start, | |
599 | unsigned long end, unsigned long newflags); | |
1c8f4220 SJ |
600 | vm_fault_t (*fault)(struct vm_fault *vmf); |
601 | vm_fault_t (*huge_fault)(struct vm_fault *vmf, | |
602 | enum page_entry_size pe_size); | |
f9ce0be7 | 603 | vm_fault_t (*map_pages)(struct vm_fault *vmf, |
bae473a4 | 604 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
05ea8860 | 605 | unsigned long (*pagesize)(struct vm_area_struct * area); |
9637a5ef DH |
606 | |
607 | /* notification that a previously read-only page is about to become | |
608 | * writable, if an error is returned it will cause a SIGBUS */ | |
1c8f4220 | 609 | vm_fault_t (*page_mkwrite)(struct vm_fault *vmf); |
28b2ee20 | 610 | |
dd906184 | 611 | /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ |
1c8f4220 | 612 | vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf); |
dd906184 | 613 | |
28b2ee20 | 614 | /* called by access_process_vm when get_user_pages() fails, typically |
96667f8a DV |
615 | * for use by special VMAs. See also generic_access_phys() for a generic |
616 | * implementation useful for any iomem mapping. | |
28b2ee20 RR |
617 | */ |
618 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | |
619 | void *buf, int len, int write); | |
78d683e8 AL |
620 | |
621 | /* Called by the /proc/PID/maps code to ask the vma whether it | |
622 | * has a special name. Returning non-NULL will also cause this | |
623 | * vma to be dumped unconditionally. */ | |
624 | const char *(*name)(struct vm_area_struct *vma); | |
625 | ||
1da177e4 | 626 | #ifdef CONFIG_NUMA |
a6020ed7 LS |
627 | /* |
628 | * set_policy() op must add a reference to any non-NULL @new mempolicy | |
629 | * to hold the policy upon return. Caller should pass NULL @new to | |
630 | * remove a policy and fall back to surrounding context--i.e. do not | |
631 | * install a MPOL_DEFAULT policy, nor the task or system default | |
632 | * mempolicy. | |
633 | */ | |
1da177e4 | 634 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
a6020ed7 LS |
635 | |
636 | /* | |
637 | * get_policy() op must add reference [mpol_get()] to any policy at | |
638 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | |
639 | * in mm/mempolicy.c will do this automatically. | |
640 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | |
c1e8d7c6 | 641 | * marked as MPOL_SHARED. vma policies are protected by the mmap_lock. |
a6020ed7 LS |
642 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op |
643 | * must return NULL--i.e., do not "fallback" to task or system default | |
644 | * policy. | |
645 | */ | |
1da177e4 LT |
646 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
647 | unsigned long addr); | |
648 | #endif | |
667a0a06 DV |
649 | /* |
650 | * Called by vm_normal_page() for special PTEs to find the | |
651 | * page for @addr. This is useful if the default behavior | |
652 | * (using pte_page()) would not find the correct page. | |
653 | */ | |
654 | struct page *(*find_special_page)(struct vm_area_struct *vma, | |
655 | unsigned long addr); | |
1da177e4 LT |
656 | }; |
657 | ||
027232da KS |
658 | static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm) |
659 | { | |
bfd40eaf KS |
660 | static const struct vm_operations_struct dummy_vm_ops = {}; |
661 | ||
a670468f | 662 | memset(vma, 0, sizeof(*vma)); |
027232da | 663 | vma->vm_mm = mm; |
bfd40eaf | 664 | vma->vm_ops = &dummy_vm_ops; |
027232da KS |
665 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
666 | } | |
667 | ||
bfd40eaf KS |
668 | static inline void vma_set_anonymous(struct vm_area_struct *vma) |
669 | { | |
670 | vma->vm_ops = NULL; | |
671 | } | |
672 | ||
43675e6f YS |
673 | static inline bool vma_is_anonymous(struct vm_area_struct *vma) |
674 | { | |
675 | return !vma->vm_ops; | |
676 | } | |
677 | ||
222100ee AK |
678 | static inline bool vma_is_temporary_stack(struct vm_area_struct *vma) |
679 | { | |
680 | int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP); | |
681 | ||
682 | if (!maybe_stack) | |
683 | return false; | |
684 | ||
685 | if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) == | |
686 | VM_STACK_INCOMPLETE_SETUP) | |
687 | return true; | |
688 | ||
689 | return false; | |
690 | } | |
691 | ||
7969f226 AK |
692 | static inline bool vma_is_foreign(struct vm_area_struct *vma) |
693 | { | |
694 | if (!current->mm) | |
695 | return true; | |
696 | ||
697 | if (current->mm != vma->vm_mm) | |
698 | return true; | |
699 | ||
700 | return false; | |
701 | } | |
3122e80e AK |
702 | |
703 | static inline bool vma_is_accessible(struct vm_area_struct *vma) | |
704 | { | |
6cb4d9a2 | 705 | return vma->vm_flags & VM_ACCESS_FLAGS; |
3122e80e AK |
706 | } |
707 | ||
43675e6f YS |
708 | #ifdef CONFIG_SHMEM |
709 | /* | |
710 | * The vma_is_shmem is not inline because it is used only by slow | |
711 | * paths in userfault. | |
712 | */ | |
713 | bool vma_is_shmem(struct vm_area_struct *vma); | |
714 | #else | |
715 | static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; } | |
716 | #endif | |
717 | ||
718 | int vma_is_stack_for_current(struct vm_area_struct *vma); | |
719 | ||
8b11ec1b LT |
720 | /* flush_tlb_range() takes a vma, not a mm, and can care about flags */ |
721 | #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) } | |
722 | ||
1da177e4 LT |
723 | struct mmu_gather; |
724 | struct inode; | |
725 | ||
71e3aac0 | 726 | #include <linux/huge_mm.h> |
1da177e4 LT |
727 | |
728 | /* | |
729 | * Methods to modify the page usage count. | |
730 | * | |
731 | * What counts for a page usage: | |
732 | * - cache mapping (page->mapping) | |
733 | * - private data (page->private) | |
734 | * - page mapped in a task's page tables, each mapping | |
735 | * is counted separately | |
736 | * | |
737 | * Also, many kernel routines increase the page count before a critical | |
738 | * routine so they can be sure the page doesn't go away from under them. | |
1da177e4 LT |
739 | */ |
740 | ||
741 | /* | |
da6052f7 | 742 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
1da177e4 | 743 | */ |
7c8ee9a8 NP |
744 | static inline int put_page_testzero(struct page *page) |
745 | { | |
fe896d18 JK |
746 | VM_BUG_ON_PAGE(page_ref_count(page) == 0, page); |
747 | return page_ref_dec_and_test(page); | |
7c8ee9a8 | 748 | } |
1da177e4 LT |
749 | |
750 | /* | |
7c8ee9a8 NP |
751 | * Try to grab a ref unless the page has a refcount of zero, return false if |
752 | * that is the case. | |
8e0861fa AK |
753 | * This can be called when MMU is off so it must not access |
754 | * any of the virtual mappings. | |
1da177e4 | 755 | */ |
7c8ee9a8 NP |
756 | static inline int get_page_unless_zero(struct page *page) |
757 | { | |
fe896d18 | 758 | return page_ref_add_unless(page, 1, 0); |
7c8ee9a8 | 759 | } |
1da177e4 | 760 | |
53df8fdc | 761 | extern int page_is_ram(unsigned long pfn); |
124fe20d DW |
762 | |
763 | enum { | |
764 | REGION_INTERSECTS, | |
765 | REGION_DISJOINT, | |
766 | REGION_MIXED, | |
767 | }; | |
768 | ||
1c29f25b TK |
769 | int region_intersects(resource_size_t offset, size_t size, unsigned long flags, |
770 | unsigned long desc); | |
53df8fdc | 771 | |
48667e7a | 772 | /* Support for virtually mapped pages */ |
b3bdda02 CL |
773 | struct page *vmalloc_to_page(const void *addr); |
774 | unsigned long vmalloc_to_pfn(const void *addr); | |
48667e7a | 775 | |
0738c4bb PM |
776 | /* |
777 | * Determine if an address is within the vmalloc range | |
778 | * | |
779 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | |
780 | * is no special casing required. | |
781 | */ | |
9bd3bb67 AK |
782 | |
783 | #ifndef is_ioremap_addr | |
784 | #define is_ioremap_addr(x) is_vmalloc_addr(x) | |
785 | #endif | |
786 | ||
81ac3ad9 | 787 | #ifdef CONFIG_MMU |
186525bd | 788 | extern bool is_vmalloc_addr(const void *x); |
81ac3ad9 KH |
789 | extern int is_vmalloc_or_module_addr(const void *x); |
790 | #else | |
186525bd IM |
791 | static inline bool is_vmalloc_addr(const void *x) |
792 | { | |
793 | return false; | |
794 | } | |
934831d0 | 795 | static inline int is_vmalloc_or_module_addr(const void *x) |
81ac3ad9 KH |
796 | { |
797 | return 0; | |
798 | } | |
799 | #endif | |
9e2779fa | 800 | |
a7c3e901 MH |
801 | extern void *kvmalloc_node(size_t size, gfp_t flags, int node); |
802 | static inline void *kvmalloc(size_t size, gfp_t flags) | |
803 | { | |
804 | return kvmalloc_node(size, flags, NUMA_NO_NODE); | |
805 | } | |
806 | static inline void *kvzalloc_node(size_t size, gfp_t flags, int node) | |
807 | { | |
808 | return kvmalloc_node(size, flags | __GFP_ZERO, node); | |
809 | } | |
810 | static inline void *kvzalloc(size_t size, gfp_t flags) | |
811 | { | |
812 | return kvmalloc(size, flags | __GFP_ZERO); | |
813 | } | |
814 | ||
752ade68 MH |
815 | static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags) |
816 | { | |
3b3b1a29 KC |
817 | size_t bytes; |
818 | ||
819 | if (unlikely(check_mul_overflow(n, size, &bytes))) | |
752ade68 MH |
820 | return NULL; |
821 | ||
3b3b1a29 | 822 | return kvmalloc(bytes, flags); |
752ade68 MH |
823 | } |
824 | ||
1c542f38 KC |
825 | static inline void *kvcalloc(size_t n, size_t size, gfp_t flags) |
826 | { | |
827 | return kvmalloc_array(n, size, flags | __GFP_ZERO); | |
828 | } | |
829 | ||
39f1f78d | 830 | extern void kvfree(const void *addr); |
d4eaa283 | 831 | extern void kvfree_sensitive(const void *addr, size_t len); |
39f1f78d | 832 | |
bac3cf4d | 833 | static inline int head_compound_mapcount(struct page *head) |
6dc5ea16 JH |
834 | { |
835 | return atomic_read(compound_mapcount_ptr(head)) + 1; | |
836 | } | |
837 | ||
6988f31d KK |
838 | /* |
839 | * Mapcount of compound page as a whole, does not include mapped sub-pages. | |
840 | * | |
841 | * Must be called only for compound pages or any their tail sub-pages. | |
842 | */ | |
53f9263b KS |
843 | static inline int compound_mapcount(struct page *page) |
844 | { | |
5f527c2b | 845 | VM_BUG_ON_PAGE(!PageCompound(page), page); |
53f9263b | 846 | page = compound_head(page); |
bac3cf4d | 847 | return head_compound_mapcount(page); |
53f9263b KS |
848 | } |
849 | ||
70b50f94 AA |
850 | /* |
851 | * The atomic page->_mapcount, starts from -1: so that transitions | |
852 | * both from it and to it can be tracked, using atomic_inc_and_test | |
853 | * and atomic_add_negative(-1). | |
854 | */ | |
22b751c3 | 855 | static inline void page_mapcount_reset(struct page *page) |
70b50f94 AA |
856 | { |
857 | atomic_set(&(page)->_mapcount, -1); | |
858 | } | |
859 | ||
b20ce5e0 KS |
860 | int __page_mapcount(struct page *page); |
861 | ||
6988f31d KK |
862 | /* |
863 | * Mapcount of 0-order page; when compound sub-page, includes | |
864 | * compound_mapcount(). | |
865 | * | |
866 | * Result is undefined for pages which cannot be mapped into userspace. | |
867 | * For example SLAB or special types of pages. See function page_has_type(). | |
868 | * They use this place in struct page differently. | |
869 | */ | |
70b50f94 AA |
870 | static inline int page_mapcount(struct page *page) |
871 | { | |
b20ce5e0 KS |
872 | if (unlikely(PageCompound(page))) |
873 | return __page_mapcount(page); | |
874 | return atomic_read(&page->_mapcount) + 1; | |
875 | } | |
876 | ||
877 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
878 | int total_mapcount(struct page *page); | |
6d0a07ed | 879 | int page_trans_huge_mapcount(struct page *page, int *total_mapcount); |
b20ce5e0 KS |
880 | #else |
881 | static inline int total_mapcount(struct page *page) | |
882 | { | |
883 | return page_mapcount(page); | |
70b50f94 | 884 | } |
6d0a07ed AA |
885 | static inline int page_trans_huge_mapcount(struct page *page, |
886 | int *total_mapcount) | |
887 | { | |
888 | int mapcount = page_mapcount(page); | |
889 | if (total_mapcount) | |
890 | *total_mapcount = mapcount; | |
891 | return mapcount; | |
892 | } | |
b20ce5e0 | 893 | #endif |
70b50f94 | 894 | |
b49af68f CL |
895 | static inline struct page *virt_to_head_page(const void *x) |
896 | { | |
897 | struct page *page = virt_to_page(x); | |
ccaafd7f | 898 | |
1d798ca3 | 899 | return compound_head(page); |
b49af68f CL |
900 | } |
901 | ||
ddc58f27 KS |
902 | void __put_page(struct page *page); |
903 | ||
1d7ea732 | 904 | void put_pages_list(struct list_head *pages); |
1da177e4 | 905 | |
8dfcc9ba | 906 | void split_page(struct page *page, unsigned int order); |
8dfcc9ba | 907 | |
33f2ef89 AW |
908 | /* |
909 | * Compound pages have a destructor function. Provide a | |
910 | * prototype for that function and accessor functions. | |
f1e61557 | 911 | * These are _only_ valid on the head of a compound page. |
33f2ef89 | 912 | */ |
f1e61557 KS |
913 | typedef void compound_page_dtor(struct page *); |
914 | ||
915 | /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */ | |
916 | enum compound_dtor_id { | |
917 | NULL_COMPOUND_DTOR, | |
918 | COMPOUND_PAGE_DTOR, | |
919 | #ifdef CONFIG_HUGETLB_PAGE | |
920 | HUGETLB_PAGE_DTOR, | |
9a982250 KS |
921 | #endif |
922 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
923 | TRANSHUGE_PAGE_DTOR, | |
f1e61557 KS |
924 | #endif |
925 | NR_COMPOUND_DTORS, | |
926 | }; | |
ae70eddd | 927 | extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS]; |
33f2ef89 AW |
928 | |
929 | static inline void set_compound_page_dtor(struct page *page, | |
f1e61557 | 930 | enum compound_dtor_id compound_dtor) |
33f2ef89 | 931 | { |
f1e61557 KS |
932 | VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page); |
933 | page[1].compound_dtor = compound_dtor; | |
33f2ef89 AW |
934 | } |
935 | ||
ff45fc3c | 936 | static inline void destroy_compound_page(struct page *page) |
33f2ef89 | 937 | { |
f1e61557 | 938 | VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page); |
ff45fc3c | 939 | compound_page_dtors[page[1].compound_dtor](page); |
33f2ef89 AW |
940 | } |
941 | ||
d00181b9 | 942 | static inline unsigned int compound_order(struct page *page) |
d85f3385 | 943 | { |
6d777953 | 944 | if (!PageHead(page)) |
d85f3385 | 945 | return 0; |
e4b294c2 | 946 | return page[1].compound_order; |
d85f3385 CL |
947 | } |
948 | ||
47e29d32 JH |
949 | static inline bool hpage_pincount_available(struct page *page) |
950 | { | |
951 | /* | |
952 | * Can the page->hpage_pinned_refcount field be used? That field is in | |
953 | * the 3rd page of the compound page, so the smallest (2-page) compound | |
954 | * pages cannot support it. | |
955 | */ | |
956 | page = compound_head(page); | |
957 | return PageCompound(page) && compound_order(page) > 1; | |
958 | } | |
959 | ||
bac3cf4d | 960 | static inline int head_compound_pincount(struct page *head) |
6dc5ea16 JH |
961 | { |
962 | return atomic_read(compound_pincount_ptr(head)); | |
963 | } | |
964 | ||
47e29d32 JH |
965 | static inline int compound_pincount(struct page *page) |
966 | { | |
967 | VM_BUG_ON_PAGE(!hpage_pincount_available(page), page); | |
968 | page = compound_head(page); | |
bac3cf4d | 969 | return head_compound_pincount(page); |
47e29d32 JH |
970 | } |
971 | ||
f1e61557 | 972 | static inline void set_compound_order(struct page *page, unsigned int order) |
d85f3385 | 973 | { |
e4b294c2 | 974 | page[1].compound_order = order; |
1378a5ee | 975 | page[1].compound_nr = 1U << order; |
d85f3385 CL |
976 | } |
977 | ||
d8c6546b MWO |
978 | /* Returns the number of pages in this potentially compound page. */ |
979 | static inline unsigned long compound_nr(struct page *page) | |
980 | { | |
1378a5ee MWO |
981 | if (!PageHead(page)) |
982 | return 1; | |
983 | return page[1].compound_nr; | |
d8c6546b MWO |
984 | } |
985 | ||
a50b854e MWO |
986 | /* Returns the number of bytes in this potentially compound page. */ |
987 | static inline unsigned long page_size(struct page *page) | |
988 | { | |
989 | return PAGE_SIZE << compound_order(page); | |
990 | } | |
991 | ||
94ad9338 MWO |
992 | /* Returns the number of bits needed for the number of bytes in a page */ |
993 | static inline unsigned int page_shift(struct page *page) | |
994 | { | |
995 | return PAGE_SHIFT + compound_order(page); | |
996 | } | |
997 | ||
9a982250 KS |
998 | void free_compound_page(struct page *page); |
999 | ||
3dece370 | 1000 | #ifdef CONFIG_MMU |
14fd403f AA |
1001 | /* |
1002 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | |
1003 | * servicing faults for write access. In the normal case, do always want | |
1004 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | |
1005 | * that do not have writing enabled, when used by access_process_vm. | |
1006 | */ | |
1007 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | |
1008 | { | |
1009 | if (likely(vma->vm_flags & VM_WRITE)) | |
1010 | pte = pte_mkwrite(pte); | |
1011 | return pte; | |
1012 | } | |
8c6e50b0 | 1013 | |
f9ce0be7 | 1014 | vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page); |
9d3af4b4 | 1015 | void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr); |
f9ce0be7 | 1016 | |
2b740303 SJ |
1017 | vm_fault_t finish_fault(struct vm_fault *vmf); |
1018 | vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf); | |
3dece370 | 1019 | #endif |
14fd403f | 1020 | |
1da177e4 LT |
1021 | /* |
1022 | * Multiple processes may "see" the same page. E.g. for untouched | |
1023 | * mappings of /dev/null, all processes see the same page full of | |
1024 | * zeroes, and text pages of executables and shared libraries have | |
1025 | * only one copy in memory, at most, normally. | |
1026 | * | |
1027 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
7e871b6c PBG |
1028 | * page_count() == 0 means the page is free. page->lru is then used for |
1029 | * freelist management in the buddy allocator. | |
da6052f7 | 1030 | * page_count() > 0 means the page has been allocated. |
1da177e4 | 1031 | * |
da6052f7 NP |
1032 | * Pages are allocated by the slab allocator in order to provide memory |
1033 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
1034 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
1035 | * unless a particular usage is carefully commented. (the responsibility of | |
1036 | * freeing the kmalloc memory is the caller's, of course). | |
1da177e4 | 1037 | * |
da6052f7 NP |
1038 | * A page may be used by anyone else who does a __get_free_page(). |
1039 | * In this case, page_count still tracks the references, and should only | |
1040 | * be used through the normal accessor functions. The top bits of page->flags | |
1041 | * and page->virtual store page management information, but all other fields | |
1042 | * are unused and could be used privately, carefully. The management of this | |
1043 | * page is the responsibility of the one who allocated it, and those who have | |
1044 | * subsequently been given references to it. | |
1045 | * | |
1046 | * The other pages (we may call them "pagecache pages") are completely | |
1da177e4 LT |
1047 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
1048 | * The following discussion applies only to them. | |
1049 | * | |
da6052f7 NP |
1050 | * A pagecache page contains an opaque `private' member, which belongs to the |
1051 | * page's address_space. Usually, this is the address of a circular list of | |
1052 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
1053 | * into the filesystem to release these pages. | |
1da177e4 | 1054 | * |
da6052f7 NP |
1055 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
1056 | * is the pointer to the inode, and page->index is the file offset of the page, | |
ea1754a0 | 1057 | * in units of PAGE_SIZE. |
1da177e4 | 1058 | * |
da6052f7 NP |
1059 | * If pagecache pages are not associated with an inode, they are said to be |
1060 | * anonymous pages. These may become associated with the swapcache, and in that | |
1061 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
1da177e4 | 1062 | * |
da6052f7 NP |
1063 | * In either case (swapcache or inode backed), the pagecache itself holds one |
1064 | * reference to the page. Setting PG_private should also increment the | |
1065 | * refcount. The each user mapping also has a reference to the page. | |
1da177e4 | 1066 | * |
da6052f7 | 1067 | * The pagecache pages are stored in a per-mapping radix tree, which is |
b93b0163 | 1068 | * rooted at mapping->i_pages, and indexed by offset. |
da6052f7 NP |
1069 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
1070 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
1da177e4 | 1071 | * |
da6052f7 | 1072 | * All pagecache pages may be subject to I/O: |
1da177e4 LT |
1073 | * - inode pages may need to be read from disk, |
1074 | * - inode pages which have been modified and are MAP_SHARED may need | |
da6052f7 NP |
1075 | * to be written back to the inode on disk, |
1076 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
1077 | * modified may need to be swapped out to swap space and (later) to be read | |
1078 | * back into memory. | |
1da177e4 LT |
1079 | */ |
1080 | ||
1081 | /* | |
1082 | * The zone field is never updated after free_area_init_core() | |
1083 | * sets it, so none of the operations on it need to be atomic. | |
1da177e4 | 1084 | */ |
348f8b6c | 1085 | |
90572890 | 1086 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ |
07808b74 | 1087 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
d41dee36 AW |
1088 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
1089 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
90572890 | 1090 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) |
2813b9c0 | 1091 | #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH) |
d41dee36 | 1092 | |
348f8b6c | 1093 | /* |
25985edc | 1094 | * Define the bit shifts to access each section. For non-existent |
348f8b6c DH |
1095 | * sections we define the shift as 0; that plus a 0 mask ensures |
1096 | * the compiler will optimise away reference to them. | |
1097 | */ | |
d41dee36 AW |
1098 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
1099 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
1100 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
90572890 | 1101 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) |
2813b9c0 | 1102 | #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0)) |
348f8b6c | 1103 | |
bce54bbf WD |
1104 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ |
1105 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
89689ae7 | 1106 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
1107 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
1108 | SECTIONS_PGOFF : ZONES_PGOFF) | |
d41dee36 | 1109 | #else |
89689ae7 | 1110 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
1111 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
1112 | NODES_PGOFF : ZONES_PGOFF) | |
89689ae7 CL |
1113 | #endif |
1114 | ||
bd8029b6 | 1115 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
348f8b6c | 1116 | |
d41dee36 AW |
1117 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
1118 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
1119 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
834a964a | 1120 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1) |
2813b9c0 | 1121 | #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1) |
89689ae7 | 1122 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
348f8b6c | 1123 | |
33dd4e0e | 1124 | static inline enum zone_type page_zonenum(const struct page *page) |
1da177e4 | 1125 | { |
c403f6a3 | 1126 | ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT); |
348f8b6c | 1127 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
1da177e4 | 1128 | } |
1da177e4 | 1129 | |
260ae3f7 DW |
1130 | #ifdef CONFIG_ZONE_DEVICE |
1131 | static inline bool is_zone_device_page(const struct page *page) | |
1132 | { | |
1133 | return page_zonenum(page) == ZONE_DEVICE; | |
1134 | } | |
966cf44f AD |
1135 | extern void memmap_init_zone_device(struct zone *, unsigned long, |
1136 | unsigned long, struct dev_pagemap *); | |
260ae3f7 DW |
1137 | #else |
1138 | static inline bool is_zone_device_page(const struct page *page) | |
1139 | { | |
1140 | return false; | |
1141 | } | |
7b2d55d2 | 1142 | #endif |
5042db43 | 1143 | |
8e3560d9 PT |
1144 | static inline bool is_zone_movable_page(const struct page *page) |
1145 | { | |
1146 | return page_zonenum(page) == ZONE_MOVABLE; | |
1147 | } | |
1148 | ||
e7638488 | 1149 | #ifdef CONFIG_DEV_PAGEMAP_OPS |
07d80269 | 1150 | void free_devmap_managed_page(struct page *page); |
e7638488 | 1151 | DECLARE_STATIC_KEY_FALSE(devmap_managed_key); |
07d80269 JH |
1152 | |
1153 | static inline bool page_is_devmap_managed(struct page *page) | |
e7638488 DW |
1154 | { |
1155 | if (!static_branch_unlikely(&devmap_managed_key)) | |
1156 | return false; | |
1157 | if (!is_zone_device_page(page)) | |
1158 | return false; | |
1159 | switch (page->pgmap->type) { | |
1160 | case MEMORY_DEVICE_PRIVATE: | |
e7638488 | 1161 | case MEMORY_DEVICE_FS_DAX: |
e7638488 DW |
1162 | return true; |
1163 | default: | |
1164 | break; | |
1165 | } | |
1166 | return false; | |
1167 | } | |
1168 | ||
07d80269 JH |
1169 | void put_devmap_managed_page(struct page *page); |
1170 | ||
e7638488 | 1171 | #else /* CONFIG_DEV_PAGEMAP_OPS */ |
07d80269 | 1172 | static inline bool page_is_devmap_managed(struct page *page) |
e7638488 DW |
1173 | { |
1174 | return false; | |
1175 | } | |
07d80269 JH |
1176 | |
1177 | static inline void put_devmap_managed_page(struct page *page) | |
1178 | { | |
1179 | } | |
7588adf8 | 1180 | #endif /* CONFIG_DEV_PAGEMAP_OPS */ |
e7638488 | 1181 | |
6b368cd4 JG |
1182 | static inline bool is_device_private_page(const struct page *page) |
1183 | { | |
7588adf8 RM |
1184 | return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) && |
1185 | IS_ENABLED(CONFIG_DEVICE_PRIVATE) && | |
1186 | is_zone_device_page(page) && | |
1187 | page->pgmap->type == MEMORY_DEVICE_PRIVATE; | |
6b368cd4 | 1188 | } |
e7638488 | 1189 | |
52916982 LG |
1190 | static inline bool is_pci_p2pdma_page(const struct page *page) |
1191 | { | |
7588adf8 RM |
1192 | return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) && |
1193 | IS_ENABLED(CONFIG_PCI_P2PDMA) && | |
1194 | is_zone_device_page(page) && | |
1195 | page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA; | |
52916982 | 1196 | } |
7b2d55d2 | 1197 | |
f958d7b5 LT |
1198 | /* 127: arbitrary random number, small enough to assemble well */ |
1199 | #define page_ref_zero_or_close_to_overflow(page) \ | |
1200 | ((unsigned int) page_ref_count(page) + 127u <= 127u) | |
1201 | ||
3565fce3 DW |
1202 | static inline void get_page(struct page *page) |
1203 | { | |
1204 | page = compound_head(page); | |
1205 | /* | |
1206 | * Getting a normal page or the head of a compound page | |
0139aa7b | 1207 | * requires to already have an elevated page->_refcount. |
3565fce3 | 1208 | */ |
f958d7b5 | 1209 | VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page); |
fe896d18 | 1210 | page_ref_inc(page); |
3565fce3 DW |
1211 | } |
1212 | ||
3faa52c0 | 1213 | bool __must_check try_grab_page(struct page *page, unsigned int flags); |
0fa5bc40 JM |
1214 | __maybe_unused struct page *try_grab_compound_head(struct page *page, int refs, |
1215 | unsigned int flags); | |
1216 | ||
3faa52c0 | 1217 | |
88b1a17d LT |
1218 | static inline __must_check bool try_get_page(struct page *page) |
1219 | { | |
1220 | page = compound_head(page); | |
1221 | if (WARN_ON_ONCE(page_ref_count(page) <= 0)) | |
1222 | return false; | |
fe896d18 | 1223 | page_ref_inc(page); |
88b1a17d | 1224 | return true; |
3565fce3 DW |
1225 | } |
1226 | ||
1227 | static inline void put_page(struct page *page) | |
1228 | { | |
1229 | page = compound_head(page); | |
1230 | ||
7b2d55d2 | 1231 | /* |
e7638488 DW |
1232 | * For devmap managed pages we need to catch refcount transition from |
1233 | * 2 to 1, when refcount reach one it means the page is free and we | |
1234 | * need to inform the device driver through callback. See | |
7b2d55d2 JG |
1235 | * include/linux/memremap.h and HMM for details. |
1236 | */ | |
07d80269 JH |
1237 | if (page_is_devmap_managed(page)) { |
1238 | put_devmap_managed_page(page); | |
7b2d55d2 | 1239 | return; |
07d80269 | 1240 | } |
7b2d55d2 | 1241 | |
3565fce3 DW |
1242 | if (put_page_testzero(page)) |
1243 | __put_page(page); | |
3565fce3 DW |
1244 | } |
1245 | ||
3faa52c0 JH |
1246 | /* |
1247 | * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload | |
1248 | * the page's refcount so that two separate items are tracked: the original page | |
1249 | * reference count, and also a new count of how many pin_user_pages() calls were | |
1250 | * made against the page. ("gup-pinned" is another term for the latter). | |
1251 | * | |
1252 | * With this scheme, pin_user_pages() becomes special: such pages are marked as | |
1253 | * distinct from normal pages. As such, the unpin_user_page() call (and its | |
1254 | * variants) must be used in order to release gup-pinned pages. | |
1255 | * | |
1256 | * Choice of value: | |
1257 | * | |
1258 | * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference | |
1259 | * counts with respect to pin_user_pages() and unpin_user_page() becomes | |
1260 | * simpler, due to the fact that adding an even power of two to the page | |
1261 | * refcount has the effect of using only the upper N bits, for the code that | |
1262 | * counts up using the bias value. This means that the lower bits are left for | |
1263 | * the exclusive use of the original code that increments and decrements by one | |
1264 | * (or at least, by much smaller values than the bias value). | |
fc1d8e7c | 1265 | * |
3faa52c0 JH |
1266 | * Of course, once the lower bits overflow into the upper bits (and this is |
1267 | * OK, because subtraction recovers the original values), then visual inspection | |
1268 | * no longer suffices to directly view the separate counts. However, for normal | |
1269 | * applications that don't have huge page reference counts, this won't be an | |
1270 | * issue. | |
fc1d8e7c | 1271 | * |
3faa52c0 JH |
1272 | * Locking: the lockless algorithm described in page_cache_get_speculative() |
1273 | * and page_cache_gup_pin_speculative() provides safe operation for | |
1274 | * get_user_pages and page_mkclean and other calls that race to set up page | |
1275 | * table entries. | |
fc1d8e7c | 1276 | */ |
3faa52c0 | 1277 | #define GUP_PIN_COUNTING_BIAS (1U << 10) |
fc1d8e7c | 1278 | |
3faa52c0 | 1279 | void unpin_user_page(struct page *page); |
f1f6a7dd JH |
1280 | void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages, |
1281 | bool make_dirty); | |
458a4f78 JM |
1282 | void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages, |
1283 | bool make_dirty); | |
f1f6a7dd | 1284 | void unpin_user_pages(struct page **pages, unsigned long npages); |
fc1d8e7c | 1285 | |
3faa52c0 | 1286 | /** |
136dfc99 MWO |
1287 | * page_maybe_dma_pinned - Report if a page is pinned for DMA. |
1288 | * @page: The page. | |
3faa52c0 JH |
1289 | * |
1290 | * This function checks if a page has been pinned via a call to | |
136dfc99 | 1291 | * a function in the pin_user_pages() family. |
3faa52c0 JH |
1292 | * |
1293 | * For non-huge pages, the return value is partially fuzzy: false is not fuzzy, | |
1294 | * because it means "definitely not pinned for DMA", but true means "probably | |
1295 | * pinned for DMA, but possibly a false positive due to having at least | |
1296 | * GUP_PIN_COUNTING_BIAS worth of normal page references". | |
1297 | * | |
1298 | * False positives are OK, because: a) it's unlikely for a page to get that many | |
1299 | * refcounts, and b) all the callers of this routine are expected to be able to | |
1300 | * deal gracefully with a false positive. | |
1301 | * | |
47e29d32 JH |
1302 | * For huge pages, the result will be exactly correct. That's because we have |
1303 | * more tracking data available: the 3rd struct page in the compound page is | |
1304 | * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS | |
1305 | * scheme). | |
1306 | * | |
72ef5e52 | 1307 | * For more information, please see Documentation/core-api/pin_user_pages.rst. |
3faa52c0 | 1308 | * |
136dfc99 MWO |
1309 | * Return: True, if it is likely that the page has been "dma-pinned". |
1310 | * False, if the page is definitely not dma-pinned. | |
3faa52c0 JH |
1311 | */ |
1312 | static inline bool page_maybe_dma_pinned(struct page *page) | |
1313 | { | |
47e29d32 JH |
1314 | if (hpage_pincount_available(page)) |
1315 | return compound_pincount(page) > 0; | |
1316 | ||
3faa52c0 JH |
1317 | /* |
1318 | * page_ref_count() is signed. If that refcount overflows, then | |
1319 | * page_ref_count() returns a negative value, and callers will avoid | |
1320 | * further incrementing the refcount. | |
1321 | * | |
1322 | * Here, for that overflow case, use the signed bit to count a little | |
1323 | * bit higher via unsigned math, and thus still get an accurate result. | |
1324 | */ | |
1325 | return ((unsigned int)page_ref_count(compound_head(page))) >= | |
1326 | GUP_PIN_COUNTING_BIAS; | |
1327 | } | |
1328 | ||
97a7e473 PX |
1329 | static inline bool is_cow_mapping(vm_flags_t flags) |
1330 | { | |
1331 | return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE; | |
1332 | } | |
1333 | ||
1334 | /* | |
1335 | * This should most likely only be called during fork() to see whether we | |
1336 | * should break the cow immediately for a page on the src mm. | |
1337 | */ | |
1338 | static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma, | |
1339 | struct page *page) | |
1340 | { | |
1341 | if (!is_cow_mapping(vma->vm_flags)) | |
1342 | return false; | |
1343 | ||
a458b76a | 1344 | if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags)) |
97a7e473 PX |
1345 | return false; |
1346 | ||
1347 | return page_maybe_dma_pinned(page); | |
1348 | } | |
1349 | ||
9127ab4f CS |
1350 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
1351 | #define SECTION_IN_PAGE_FLAGS | |
1352 | #endif | |
1353 | ||
89689ae7 | 1354 | /* |
7a8010cd VB |
1355 | * The identification function is mainly used by the buddy allocator for |
1356 | * determining if two pages could be buddies. We are not really identifying | |
1357 | * the zone since we could be using the section number id if we do not have | |
1358 | * node id available in page flags. | |
1359 | * We only guarantee that it will return the same value for two combinable | |
1360 | * pages in a zone. | |
89689ae7 | 1361 | */ |
cb2b95e1 AW |
1362 | static inline int page_zone_id(struct page *page) |
1363 | { | |
89689ae7 | 1364 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
348f8b6c DH |
1365 | } |
1366 | ||
89689ae7 | 1367 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
33dd4e0e | 1368 | extern int page_to_nid(const struct page *page); |
89689ae7 | 1369 | #else |
33dd4e0e | 1370 | static inline int page_to_nid(const struct page *page) |
d41dee36 | 1371 | { |
f165b378 PT |
1372 | struct page *p = (struct page *)page; |
1373 | ||
1374 | return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK; | |
d41dee36 | 1375 | } |
89689ae7 CL |
1376 | #endif |
1377 | ||
57e0a030 | 1378 | #ifdef CONFIG_NUMA_BALANCING |
90572890 | 1379 | static inline int cpu_pid_to_cpupid(int cpu, int pid) |
57e0a030 | 1380 | { |
90572890 | 1381 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); |
57e0a030 MG |
1382 | } |
1383 | ||
90572890 | 1384 | static inline int cpupid_to_pid(int cpupid) |
57e0a030 | 1385 | { |
90572890 | 1386 | return cpupid & LAST__PID_MASK; |
57e0a030 | 1387 | } |
b795854b | 1388 | |
90572890 | 1389 | static inline int cpupid_to_cpu(int cpupid) |
b795854b | 1390 | { |
90572890 | 1391 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; |
b795854b MG |
1392 | } |
1393 | ||
90572890 | 1394 | static inline int cpupid_to_nid(int cpupid) |
b795854b | 1395 | { |
90572890 | 1396 | return cpu_to_node(cpupid_to_cpu(cpupid)); |
b795854b MG |
1397 | } |
1398 | ||
90572890 | 1399 | static inline bool cpupid_pid_unset(int cpupid) |
57e0a030 | 1400 | { |
90572890 | 1401 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); |
b795854b MG |
1402 | } |
1403 | ||
90572890 | 1404 | static inline bool cpupid_cpu_unset(int cpupid) |
b795854b | 1405 | { |
90572890 | 1406 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); |
b795854b MG |
1407 | } |
1408 | ||
8c8a743c PZ |
1409 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) |
1410 | { | |
1411 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); | |
1412 | } | |
1413 | ||
1414 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) | |
90572890 PZ |
1415 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
1416 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
b795854b | 1417 | { |
1ae71d03 | 1418 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); |
b795854b | 1419 | } |
90572890 PZ |
1420 | |
1421 | static inline int page_cpupid_last(struct page *page) | |
1422 | { | |
1423 | return page->_last_cpupid; | |
1424 | } | |
1425 | static inline void page_cpupid_reset_last(struct page *page) | |
b795854b | 1426 | { |
1ae71d03 | 1427 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; |
57e0a030 MG |
1428 | } |
1429 | #else | |
90572890 | 1430 | static inline int page_cpupid_last(struct page *page) |
75980e97 | 1431 | { |
90572890 | 1432 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; |
75980e97 PZ |
1433 | } |
1434 | ||
90572890 | 1435 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); |
75980e97 | 1436 | |
90572890 | 1437 | static inline void page_cpupid_reset_last(struct page *page) |
75980e97 | 1438 | { |
09940a4f | 1439 | page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT; |
75980e97 | 1440 | } |
90572890 PZ |
1441 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ |
1442 | #else /* !CONFIG_NUMA_BALANCING */ | |
1443 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
57e0a030 | 1444 | { |
90572890 | 1445 | return page_to_nid(page); /* XXX */ |
57e0a030 MG |
1446 | } |
1447 | ||
90572890 | 1448 | static inline int page_cpupid_last(struct page *page) |
57e0a030 | 1449 | { |
90572890 | 1450 | return page_to_nid(page); /* XXX */ |
57e0a030 MG |
1451 | } |
1452 | ||
90572890 | 1453 | static inline int cpupid_to_nid(int cpupid) |
b795854b MG |
1454 | { |
1455 | return -1; | |
1456 | } | |
1457 | ||
90572890 | 1458 | static inline int cpupid_to_pid(int cpupid) |
b795854b MG |
1459 | { |
1460 | return -1; | |
1461 | } | |
1462 | ||
90572890 | 1463 | static inline int cpupid_to_cpu(int cpupid) |
b795854b MG |
1464 | { |
1465 | return -1; | |
1466 | } | |
1467 | ||
90572890 PZ |
1468 | static inline int cpu_pid_to_cpupid(int nid, int pid) |
1469 | { | |
1470 | return -1; | |
1471 | } | |
1472 | ||
1473 | static inline bool cpupid_pid_unset(int cpupid) | |
b795854b | 1474 | { |
2b787449 | 1475 | return true; |
b795854b MG |
1476 | } |
1477 | ||
90572890 | 1478 | static inline void page_cpupid_reset_last(struct page *page) |
57e0a030 MG |
1479 | { |
1480 | } | |
8c8a743c PZ |
1481 | |
1482 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) | |
1483 | { | |
1484 | return false; | |
1485 | } | |
90572890 | 1486 | #endif /* CONFIG_NUMA_BALANCING */ |
57e0a030 | 1487 | |
2e903b91 | 1488 | #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS) |
34303244 | 1489 | |
cf10bd4c AK |
1490 | /* |
1491 | * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid | |
1492 | * setting tags for all pages to native kernel tag value 0xff, as the default | |
1493 | * value 0x00 maps to 0xff. | |
1494 | */ | |
1495 | ||
2813b9c0 AK |
1496 | static inline u8 page_kasan_tag(const struct page *page) |
1497 | { | |
cf10bd4c AK |
1498 | u8 tag = 0xff; |
1499 | ||
1500 | if (kasan_enabled()) { | |
1501 | tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK; | |
1502 | tag ^= 0xff; | |
1503 | } | |
1504 | ||
1505 | return tag; | |
2813b9c0 AK |
1506 | } |
1507 | ||
1508 | static inline void page_kasan_tag_set(struct page *page, u8 tag) | |
1509 | { | |
34303244 | 1510 | if (kasan_enabled()) { |
cf10bd4c | 1511 | tag ^= 0xff; |
34303244 AK |
1512 | page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT); |
1513 | page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT; | |
1514 | } | |
2813b9c0 AK |
1515 | } |
1516 | ||
1517 | static inline void page_kasan_tag_reset(struct page *page) | |
1518 | { | |
34303244 AK |
1519 | if (kasan_enabled()) |
1520 | page_kasan_tag_set(page, 0xff); | |
2813b9c0 | 1521 | } |
34303244 AK |
1522 | |
1523 | #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */ | |
1524 | ||
2813b9c0 AK |
1525 | static inline u8 page_kasan_tag(const struct page *page) |
1526 | { | |
1527 | return 0xff; | |
1528 | } | |
1529 | ||
1530 | static inline void page_kasan_tag_set(struct page *page, u8 tag) { } | |
1531 | static inline void page_kasan_tag_reset(struct page *page) { } | |
34303244 AK |
1532 | |
1533 | #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */ | |
2813b9c0 | 1534 | |
33dd4e0e | 1535 | static inline struct zone *page_zone(const struct page *page) |
89689ae7 CL |
1536 | { |
1537 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
1538 | } | |
1539 | ||
75ef7184 MG |
1540 | static inline pg_data_t *page_pgdat(const struct page *page) |
1541 | { | |
1542 | return NODE_DATA(page_to_nid(page)); | |
1543 | } | |
1544 | ||
9127ab4f | 1545 | #ifdef SECTION_IN_PAGE_FLAGS |
bf4e8902 DK |
1546 | static inline void set_page_section(struct page *page, unsigned long section) |
1547 | { | |
1548 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
1549 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
1550 | } | |
1551 | ||
aa462abe | 1552 | static inline unsigned long page_to_section(const struct page *page) |
d41dee36 AW |
1553 | { |
1554 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
1555 | } | |
308c05e3 | 1556 | #endif |
d41dee36 | 1557 | |
8e3560d9 PT |
1558 | /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */ |
1559 | #ifdef CONFIG_MIGRATION | |
1560 | static inline bool is_pinnable_page(struct page *page) | |
1561 | { | |
9afaf30f PT |
1562 | return !(is_zone_movable_page(page) || is_migrate_cma_page(page)) || |
1563 | is_zero_pfn(page_to_pfn(page)); | |
8e3560d9 PT |
1564 | } |
1565 | #else | |
1566 | static inline bool is_pinnable_page(struct page *page) | |
1567 | { | |
1568 | return true; | |
1569 | } | |
1570 | #endif | |
1571 | ||
2f1b6248 | 1572 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
348f8b6c DH |
1573 | { |
1574 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
1575 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
1576 | } | |
2f1b6248 | 1577 | |
348f8b6c DH |
1578 | static inline void set_page_node(struct page *page, unsigned long node) |
1579 | { | |
1580 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
1581 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
1da177e4 | 1582 | } |
89689ae7 | 1583 | |
2f1b6248 | 1584 | static inline void set_page_links(struct page *page, enum zone_type zone, |
d41dee36 | 1585 | unsigned long node, unsigned long pfn) |
1da177e4 | 1586 | { |
348f8b6c DH |
1587 | set_page_zone(page, zone); |
1588 | set_page_node(page, node); | |
9127ab4f | 1589 | #ifdef SECTION_IN_PAGE_FLAGS |
d41dee36 | 1590 | set_page_section(page, pfn_to_section_nr(pfn)); |
bf4e8902 | 1591 | #endif |
1da177e4 LT |
1592 | } |
1593 | ||
f6ac2354 CL |
1594 | /* |
1595 | * Some inline functions in vmstat.h depend on page_zone() | |
1596 | */ | |
1597 | #include <linux/vmstat.h> | |
1598 | ||
33dd4e0e | 1599 | static __always_inline void *lowmem_page_address(const struct page *page) |
1da177e4 | 1600 | { |
1dff8083 | 1601 | return page_to_virt(page); |
1da177e4 LT |
1602 | } |
1603 | ||
1604 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
1605 | #define HASHED_PAGE_VIRTUAL | |
1606 | #endif | |
1607 | ||
1608 | #if defined(WANT_PAGE_VIRTUAL) | |
f92f455f GU |
1609 | static inline void *page_address(const struct page *page) |
1610 | { | |
1611 | return page->virtual; | |
1612 | } | |
1613 | static inline void set_page_address(struct page *page, void *address) | |
1614 | { | |
1615 | page->virtual = address; | |
1616 | } | |
1da177e4 LT |
1617 | #define page_address_init() do { } while(0) |
1618 | #endif | |
1619 | ||
1620 | #if defined(HASHED_PAGE_VIRTUAL) | |
f9918794 | 1621 | void *page_address(const struct page *page); |
1da177e4 LT |
1622 | void set_page_address(struct page *page, void *virtual); |
1623 | void page_address_init(void); | |
1624 | #endif | |
1625 | ||
1626 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
1627 | #define page_address(page) lowmem_page_address(page) | |
1628 | #define set_page_address(page, address) do { } while(0) | |
1629 | #define page_address_init() do { } while(0) | |
1630 | #endif | |
1631 | ||
e39155ea KS |
1632 | extern void *page_rmapping(struct page *page); |
1633 | extern struct anon_vma *page_anon_vma(struct page *page); | |
9800339b | 1634 | extern struct address_space *page_mapping(struct page *page); |
1da177e4 | 1635 | |
f981c595 MG |
1636 | extern struct address_space *__page_file_mapping(struct page *); |
1637 | ||
1638 | static inline | |
1639 | struct address_space *page_file_mapping(struct page *page) | |
1640 | { | |
1641 | if (unlikely(PageSwapCache(page))) | |
1642 | return __page_file_mapping(page); | |
1643 | ||
1644 | return page->mapping; | |
1645 | } | |
1646 | ||
f6ab1f7f HY |
1647 | extern pgoff_t __page_file_index(struct page *page); |
1648 | ||
1da177e4 LT |
1649 | /* |
1650 | * Return the pagecache index of the passed page. Regular pagecache pages | |
f6ab1f7f | 1651 | * use ->index whereas swapcache pages use swp_offset(->private) |
1da177e4 LT |
1652 | */ |
1653 | static inline pgoff_t page_index(struct page *page) | |
1654 | { | |
1655 | if (unlikely(PageSwapCache(page))) | |
f6ab1f7f | 1656 | return __page_file_index(page); |
1da177e4 LT |
1657 | return page->index; |
1658 | } | |
1659 | ||
1aa8aea5 | 1660 | bool page_mapped(struct page *page); |
bda807d4 | 1661 | struct address_space *page_mapping(struct page *page); |
1da177e4 | 1662 | |
2f064f34 MH |
1663 | /* |
1664 | * Return true only if the page has been allocated with | |
1665 | * ALLOC_NO_WATERMARKS and the low watermark was not | |
1666 | * met implying that the system is under some pressure. | |
1667 | */ | |
1d7bab6a | 1668 | static inline bool page_is_pfmemalloc(const struct page *page) |
2f064f34 MH |
1669 | { |
1670 | /* | |
1671 | * Page index cannot be this large so this must be | |
1672 | * a pfmemalloc page. | |
1673 | */ | |
1674 | return page->index == -1UL; | |
1675 | } | |
1676 | ||
1677 | /* | |
1678 | * Only to be called by the page allocator on a freshly allocated | |
1679 | * page. | |
1680 | */ | |
1681 | static inline void set_page_pfmemalloc(struct page *page) | |
1682 | { | |
1683 | page->index = -1UL; | |
1684 | } | |
1685 | ||
1686 | static inline void clear_page_pfmemalloc(struct page *page) | |
1687 | { | |
1688 | page->index = 0; | |
1689 | } | |
1690 | ||
1c0fe6e3 NP |
1691 | /* |
1692 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | |
1693 | */ | |
1694 | extern void pagefault_out_of_memory(void); | |
1695 | ||
1da177e4 | 1696 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
ee6c400f | 1697 | #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1)) |
1da177e4 | 1698 | |
ddd588b5 | 1699 | /* |
7bf02ea2 | 1700 | * Flags passed to show_mem() and show_free_areas() to suppress output in |
ddd588b5 DR |
1701 | * various contexts. |
1702 | */ | |
4b59e6c4 | 1703 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ |
ddd588b5 | 1704 | |
9af744d7 | 1705 | extern void show_free_areas(unsigned int flags, nodemask_t *nodemask); |
1da177e4 | 1706 | |
710ec38b | 1707 | #ifdef CONFIG_MMU |
7f43add4 | 1708 | extern bool can_do_mlock(void); |
710ec38b AB |
1709 | #else |
1710 | static inline bool can_do_mlock(void) { return false; } | |
1711 | #endif | |
1da177e4 LT |
1712 | extern int user_shm_lock(size_t, struct user_struct *); |
1713 | extern void user_shm_unlock(size_t, struct user_struct *); | |
1714 | ||
1715 | /* | |
1716 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
1717 | */ | |
1718 | struct zap_details { | |
1da177e4 LT |
1719 | struct address_space *check_mapping; /* Check page->mapping if set */ |
1720 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
1721 | pgoff_t last_index; /* Highest page->index to unmap */ | |
22061a1f | 1722 | struct page *single_page; /* Locked page to be unmapped */ |
1da177e4 LT |
1723 | }; |
1724 | ||
25b2995a CH |
1725 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
1726 | pte_t pte); | |
28093f9f GS |
1727 | struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr, |
1728 | pmd_t pmd); | |
7e675137 | 1729 | |
27d036e3 LR |
1730 | void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
1731 | unsigned long size); | |
14f5ff5d | 1732 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, |
27d036e3 | 1733 | unsigned long size); |
4f74d2c8 LT |
1734 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
1735 | unsigned long start, unsigned long end); | |
e6473092 | 1736 | |
ac46d4f3 JG |
1737 | struct mmu_notifier_range; |
1738 | ||
42b77728 | 1739 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
3bf5ee95 | 1740 | unsigned long end, unsigned long floor, unsigned long ceiling); |
c78f4636 PX |
1741 | int |
1742 | copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma); | |
9fd6dad1 PB |
1743 | int follow_invalidate_pte(struct mm_struct *mm, unsigned long address, |
1744 | struct mmu_notifier_range *range, pte_t **ptepp, | |
1745 | pmd_t **pmdpp, spinlock_t **ptlp); | |
ff5c19ed | 1746 | int follow_pte(struct mm_struct *mm, unsigned long address, |
9fd6dad1 | 1747 | pte_t **ptepp, spinlock_t **ptlp); |
3b6748e2 JW |
1748 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
1749 | unsigned long *pfn); | |
d87fe660 | 1750 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
1751 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | |
28b2ee20 RR |
1752 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
1753 | void *buf, int len, int write); | |
1da177e4 | 1754 | |
7caef267 | 1755 | extern void truncate_pagecache(struct inode *inode, loff_t new); |
2c27c65e | 1756 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
90a80202 | 1757 | void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); |
623e3db9 | 1758 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); |
750b4987 | 1759 | int truncate_inode_page(struct address_space *mapping, struct page *page); |
25718736 | 1760 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
83f78668 WF |
1761 | int invalidate_inode_page(struct page *page); |
1762 | ||
7ee1dd3f | 1763 | #ifdef CONFIG_MMU |
2b740303 | 1764 | extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
bce617ed PX |
1765 | unsigned long address, unsigned int flags, |
1766 | struct pt_regs *regs); | |
64019a2e | 1767 | extern int fixup_user_fault(struct mm_struct *mm, |
4a9e1cda DD |
1768 | unsigned long address, unsigned int fault_flags, |
1769 | bool *unlocked); | |
22061a1f | 1770 | void unmap_mapping_page(struct page *page); |
977fbdcd MW |
1771 | void unmap_mapping_pages(struct address_space *mapping, |
1772 | pgoff_t start, pgoff_t nr, bool even_cows); | |
1773 | void unmap_mapping_range(struct address_space *mapping, | |
1774 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
7ee1dd3f | 1775 | #else |
2b740303 | 1776 | static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
bce617ed PX |
1777 | unsigned long address, unsigned int flags, |
1778 | struct pt_regs *regs) | |
7ee1dd3f DH |
1779 | { |
1780 | /* should never happen if there's no MMU */ | |
1781 | BUG(); | |
1782 | return VM_FAULT_SIGBUS; | |
1783 | } | |
64019a2e | 1784 | static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address, |
4a9e1cda | 1785 | unsigned int fault_flags, bool *unlocked) |
5c723ba5 PZ |
1786 | { |
1787 | /* should never happen if there's no MMU */ | |
1788 | BUG(); | |
1789 | return -EFAULT; | |
1790 | } | |
22061a1f | 1791 | static inline void unmap_mapping_page(struct page *page) { } |
977fbdcd MW |
1792 | static inline void unmap_mapping_pages(struct address_space *mapping, |
1793 | pgoff_t start, pgoff_t nr, bool even_cows) { } | |
1794 | static inline void unmap_mapping_range(struct address_space *mapping, | |
1795 | loff_t const holebegin, loff_t const holelen, int even_cows) { } | |
7ee1dd3f | 1796 | #endif |
f33ea7f4 | 1797 | |
977fbdcd MW |
1798 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
1799 | loff_t const holebegin, loff_t const holelen) | |
1800 | { | |
1801 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
1802 | } | |
1803 | ||
1804 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, | |
1805 | void *buf, int len, unsigned int gup_flags); | |
5ddd36b9 | 1806 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
6347e8d5 | 1807 | void *buf, int len, unsigned int gup_flags); |
d3f5ffca JH |
1808 | extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr, |
1809 | void *buf, int len, unsigned int gup_flags); | |
1da177e4 | 1810 | |
64019a2e | 1811 | long get_user_pages_remote(struct mm_struct *mm, |
1e987790 | 1812 | unsigned long start, unsigned long nr_pages, |
9beae1ea | 1813 | unsigned int gup_flags, struct page **pages, |
5b56d49f | 1814 | struct vm_area_struct **vmas, int *locked); |
64019a2e | 1815 | long pin_user_pages_remote(struct mm_struct *mm, |
eddb1c22 JH |
1816 | unsigned long start, unsigned long nr_pages, |
1817 | unsigned int gup_flags, struct page **pages, | |
1818 | struct vm_area_struct **vmas, int *locked); | |
c12d2da5 | 1819 | long get_user_pages(unsigned long start, unsigned long nr_pages, |
768ae309 | 1820 | unsigned int gup_flags, struct page **pages, |
cde70140 | 1821 | struct vm_area_struct **vmas); |
eddb1c22 JH |
1822 | long pin_user_pages(unsigned long start, unsigned long nr_pages, |
1823 | unsigned int gup_flags, struct page **pages, | |
1824 | struct vm_area_struct **vmas); | |
c12d2da5 | 1825 | long get_user_pages_locked(unsigned long start, unsigned long nr_pages, |
3b913179 | 1826 | unsigned int gup_flags, struct page **pages, int *locked); |
420c2091 JH |
1827 | long pin_user_pages_locked(unsigned long start, unsigned long nr_pages, |
1828 | unsigned int gup_flags, struct page **pages, int *locked); | |
c12d2da5 | 1829 | long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages, |
c164154f | 1830 | struct page **pages, unsigned int gup_flags); |
91429023 JH |
1831 | long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages, |
1832 | struct page **pages, unsigned int gup_flags); | |
9a4e9f3b | 1833 | |
73b0140b IW |
1834 | int get_user_pages_fast(unsigned long start, int nr_pages, |
1835 | unsigned int gup_flags, struct page **pages); | |
eddb1c22 JH |
1836 | int pin_user_pages_fast(unsigned long start, int nr_pages, |
1837 | unsigned int gup_flags, struct page **pages); | |
8025e5dd | 1838 | |
79eb597c DJ |
1839 | int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc); |
1840 | int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc, | |
1841 | struct task_struct *task, bool bypass_rlim); | |
1842 | ||
18022c5d MG |
1843 | struct kvec; |
1844 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, | |
1845 | struct page **pages); | |
1846 | int get_kernel_page(unsigned long start, int write, struct page **pages); | |
f3e8fccd | 1847 | struct page *get_dump_page(unsigned long addr); |
1da177e4 | 1848 | |
cf9a2ae8 | 1849 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
d47992f8 LC |
1850 | extern void do_invalidatepage(struct page *page, unsigned int offset, |
1851 | unsigned int length); | |
cf9a2ae8 | 1852 | |
1da177e4 LT |
1853 | int redirty_page_for_writepage(struct writeback_control *wbc, |
1854 | struct page *page); | |
c4843a75 | 1855 | void account_page_cleaned(struct page *page, struct address_space *mapping, |
62cccb8c | 1856 | struct bdi_writeback *wb); |
b3c97528 | 1857 | int set_page_dirty(struct page *page); |
1da177e4 | 1858 | int set_page_dirty_lock(struct page *page); |
736304f3 JK |
1859 | void __cancel_dirty_page(struct page *page); |
1860 | static inline void cancel_dirty_page(struct page *page) | |
1861 | { | |
1862 | /* Avoid atomic ops, locking, etc. when not actually needed. */ | |
1863 | if (PageDirty(page)) | |
1864 | __cancel_dirty_page(page); | |
1865 | } | |
1da177e4 | 1866 | int clear_page_dirty_for_io(struct page *page); |
b9ea2515 | 1867 | |
a9090253 | 1868 | int get_cmdline(struct task_struct *task, char *buffer, int buflen); |
1da177e4 | 1869 | |
b6a2fea3 OW |
1870 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
1871 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
38a76013 ML |
1872 | unsigned long new_addr, unsigned long len, |
1873 | bool need_rmap_locks); | |
58705444 PX |
1874 | |
1875 | /* | |
1876 | * Flags used by change_protection(). For now we make it a bitmap so | |
1877 | * that we can pass in multiple flags just like parameters. However | |
1878 | * for now all the callers are only use one of the flags at the same | |
1879 | * time. | |
1880 | */ | |
1881 | /* Whether we should allow dirty bit accounting */ | |
1882 | #define MM_CP_DIRTY_ACCT (1UL << 0) | |
1883 | /* Whether this protection change is for NUMA hints */ | |
1884 | #define MM_CP_PROT_NUMA (1UL << 1) | |
292924b2 PX |
1885 | /* Whether this change is for write protecting */ |
1886 | #define MM_CP_UFFD_WP (1UL << 2) /* do wp */ | |
1887 | #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */ | |
1888 | #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \ | |
1889 | MM_CP_UFFD_WP_RESOLVE) | |
58705444 | 1890 | |
7da4d641 PZ |
1891 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, |
1892 | unsigned long end, pgprot_t newprot, | |
58705444 | 1893 | unsigned long cp_flags); |
b6a2fea3 OW |
1894 | extern int mprotect_fixup(struct vm_area_struct *vma, |
1895 | struct vm_area_struct **pprev, unsigned long start, | |
1896 | unsigned long end, unsigned long newflags); | |
1da177e4 | 1897 | |
465a454f PZ |
1898 | /* |
1899 | * doesn't attempt to fault and will return short. | |
1900 | */ | |
dadbb612 SJ |
1901 | int get_user_pages_fast_only(unsigned long start, int nr_pages, |
1902 | unsigned int gup_flags, struct page **pages); | |
104acc32 JH |
1903 | int pin_user_pages_fast_only(unsigned long start, int nr_pages, |
1904 | unsigned int gup_flags, struct page **pages); | |
dadbb612 SJ |
1905 | |
1906 | static inline bool get_user_page_fast_only(unsigned long addr, | |
1907 | unsigned int gup_flags, struct page **pagep) | |
1908 | { | |
1909 | return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1; | |
1910 | } | |
d559db08 KH |
1911 | /* |
1912 | * per-process(per-mm_struct) statistics. | |
1913 | */ | |
d559db08 KH |
1914 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) |
1915 | { | |
69c97823 KK |
1916 | long val = atomic_long_read(&mm->rss_stat.count[member]); |
1917 | ||
1918 | #ifdef SPLIT_RSS_COUNTING | |
1919 | /* | |
1920 | * counter is updated in asynchronous manner and may go to minus. | |
1921 | * But it's never be expected number for users. | |
1922 | */ | |
1923 | if (val < 0) | |
1924 | val = 0; | |
172703b0 | 1925 | #endif |
69c97823 KK |
1926 | return (unsigned long)val; |
1927 | } | |
d559db08 | 1928 | |
e4dcad20 | 1929 | void mm_trace_rss_stat(struct mm_struct *mm, int member, long count); |
b3d1411b | 1930 | |
d559db08 KH |
1931 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) |
1932 | { | |
b3d1411b JFG |
1933 | long count = atomic_long_add_return(value, &mm->rss_stat.count[member]); |
1934 | ||
e4dcad20 | 1935 | mm_trace_rss_stat(mm, member, count); |
d559db08 KH |
1936 | } |
1937 | ||
1938 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
1939 | { | |
b3d1411b JFG |
1940 | long count = atomic_long_inc_return(&mm->rss_stat.count[member]); |
1941 | ||
e4dcad20 | 1942 | mm_trace_rss_stat(mm, member, count); |
d559db08 KH |
1943 | } |
1944 | ||
1945 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
1946 | { | |
b3d1411b JFG |
1947 | long count = atomic_long_dec_return(&mm->rss_stat.count[member]); |
1948 | ||
e4dcad20 | 1949 | mm_trace_rss_stat(mm, member, count); |
d559db08 KH |
1950 | } |
1951 | ||
eca56ff9 JM |
1952 | /* Optimized variant when page is already known not to be PageAnon */ |
1953 | static inline int mm_counter_file(struct page *page) | |
1954 | { | |
1955 | if (PageSwapBacked(page)) | |
1956 | return MM_SHMEMPAGES; | |
1957 | return MM_FILEPAGES; | |
1958 | } | |
1959 | ||
1960 | static inline int mm_counter(struct page *page) | |
1961 | { | |
1962 | if (PageAnon(page)) | |
1963 | return MM_ANONPAGES; | |
1964 | return mm_counter_file(page); | |
1965 | } | |
1966 | ||
d559db08 KH |
1967 | static inline unsigned long get_mm_rss(struct mm_struct *mm) |
1968 | { | |
1969 | return get_mm_counter(mm, MM_FILEPAGES) + | |
eca56ff9 JM |
1970 | get_mm_counter(mm, MM_ANONPAGES) + |
1971 | get_mm_counter(mm, MM_SHMEMPAGES); | |
d559db08 KH |
1972 | } |
1973 | ||
1974 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | |
1975 | { | |
1976 | return max(mm->hiwater_rss, get_mm_rss(mm)); | |
1977 | } | |
1978 | ||
1979 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | |
1980 | { | |
1981 | return max(mm->hiwater_vm, mm->total_vm); | |
1982 | } | |
1983 | ||
1984 | static inline void update_hiwater_rss(struct mm_struct *mm) | |
1985 | { | |
1986 | unsigned long _rss = get_mm_rss(mm); | |
1987 | ||
1988 | if ((mm)->hiwater_rss < _rss) | |
1989 | (mm)->hiwater_rss = _rss; | |
1990 | } | |
1991 | ||
1992 | static inline void update_hiwater_vm(struct mm_struct *mm) | |
1993 | { | |
1994 | if (mm->hiwater_vm < mm->total_vm) | |
1995 | mm->hiwater_vm = mm->total_vm; | |
1996 | } | |
1997 | ||
695f0559 PC |
1998 | static inline void reset_mm_hiwater_rss(struct mm_struct *mm) |
1999 | { | |
2000 | mm->hiwater_rss = get_mm_rss(mm); | |
2001 | } | |
2002 | ||
d559db08 KH |
2003 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, |
2004 | struct mm_struct *mm) | |
2005 | { | |
2006 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | |
2007 | ||
2008 | if (*maxrss < hiwater_rss) | |
2009 | *maxrss = hiwater_rss; | |
2010 | } | |
2011 | ||
53bddb4e | 2012 | #if defined(SPLIT_RSS_COUNTING) |
05af2e10 | 2013 | void sync_mm_rss(struct mm_struct *mm); |
53bddb4e | 2014 | #else |
05af2e10 | 2015 | static inline void sync_mm_rss(struct mm_struct *mm) |
53bddb4e KH |
2016 | { |
2017 | } | |
2018 | #endif | |
465a454f | 2019 | |
78e7c5af AK |
2020 | #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL |
2021 | static inline int pte_special(pte_t pte) | |
2022 | { | |
2023 | return 0; | |
2024 | } | |
2025 | ||
2026 | static inline pte_t pte_mkspecial(pte_t pte) | |
2027 | { | |
2028 | return pte; | |
2029 | } | |
2030 | #endif | |
2031 | ||
17596731 | 2032 | #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP |
3565fce3 DW |
2033 | static inline int pte_devmap(pte_t pte) |
2034 | { | |
2035 | return 0; | |
2036 | } | |
2037 | #endif | |
2038 | ||
6d2329f8 | 2039 | int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot); |
d08b3851 | 2040 | |
25ca1d6c NK |
2041 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
2042 | spinlock_t **ptl); | |
2043 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
2044 | spinlock_t **ptl) | |
2045 | { | |
2046 | pte_t *ptep; | |
2047 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | |
2048 | return ptep; | |
2049 | } | |
c9cfcddf | 2050 | |
c2febafc KS |
2051 | #ifdef __PAGETABLE_P4D_FOLDED |
2052 | static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, | |
2053 | unsigned long address) | |
2054 | { | |
2055 | return 0; | |
2056 | } | |
2057 | #else | |
2058 | int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); | |
2059 | #endif | |
2060 | ||
b4e98d9a | 2061 | #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU) |
c2febafc | 2062 | static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, |
5f22df00 NP |
2063 | unsigned long address) |
2064 | { | |
2065 | return 0; | |
2066 | } | |
b4e98d9a KS |
2067 | static inline void mm_inc_nr_puds(struct mm_struct *mm) {} |
2068 | static inline void mm_dec_nr_puds(struct mm_struct *mm) {} | |
2069 | ||
5f22df00 | 2070 | #else |
c2febafc | 2071 | int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address); |
b4e98d9a | 2072 | |
b4e98d9a KS |
2073 | static inline void mm_inc_nr_puds(struct mm_struct *mm) |
2074 | { | |
6d212db1 MS |
2075 | if (mm_pud_folded(mm)) |
2076 | return; | |
af5b0f6a | 2077 | atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
b4e98d9a KS |
2078 | } |
2079 | ||
2080 | static inline void mm_dec_nr_puds(struct mm_struct *mm) | |
2081 | { | |
6d212db1 MS |
2082 | if (mm_pud_folded(mm)) |
2083 | return; | |
af5b0f6a | 2084 | atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
b4e98d9a | 2085 | } |
5f22df00 NP |
2086 | #endif |
2087 | ||
2d2f5119 | 2088 | #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU) |
5f22df00 NP |
2089 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
2090 | unsigned long address) | |
2091 | { | |
2092 | return 0; | |
2093 | } | |
dc6c9a35 | 2094 | |
dc6c9a35 KS |
2095 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) {} |
2096 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) {} | |
2097 | ||
5f22df00 | 2098 | #else |
1bb3630e | 2099 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
dc6c9a35 | 2100 | |
dc6c9a35 KS |
2101 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) |
2102 | { | |
6d212db1 MS |
2103 | if (mm_pmd_folded(mm)) |
2104 | return; | |
af5b0f6a | 2105 | atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
dc6c9a35 KS |
2106 | } |
2107 | ||
2108 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) | |
2109 | { | |
6d212db1 MS |
2110 | if (mm_pmd_folded(mm)) |
2111 | return; | |
af5b0f6a | 2112 | atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
dc6c9a35 | 2113 | } |
5f22df00 NP |
2114 | #endif |
2115 | ||
c4812909 | 2116 | #ifdef CONFIG_MMU |
af5b0f6a | 2117 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) |
c4812909 | 2118 | { |
af5b0f6a | 2119 | atomic_long_set(&mm->pgtables_bytes, 0); |
c4812909 KS |
2120 | } |
2121 | ||
af5b0f6a | 2122 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) |
c4812909 | 2123 | { |
af5b0f6a | 2124 | return atomic_long_read(&mm->pgtables_bytes); |
c4812909 KS |
2125 | } |
2126 | ||
2127 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) | |
2128 | { | |
af5b0f6a | 2129 | atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
c4812909 KS |
2130 | } |
2131 | ||
2132 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) | |
2133 | { | |
af5b0f6a | 2134 | atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
c4812909 KS |
2135 | } |
2136 | #else | |
c4812909 | 2137 | |
af5b0f6a KS |
2138 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {} |
2139 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) | |
c4812909 KS |
2140 | { |
2141 | return 0; | |
2142 | } | |
2143 | ||
2144 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) {} | |
2145 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) {} | |
2146 | #endif | |
2147 | ||
4cf58924 JFG |
2148 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd); |
2149 | int __pte_alloc_kernel(pmd_t *pmd); | |
1bb3630e | 2150 | |
f949286c MR |
2151 | #if defined(CONFIG_MMU) |
2152 | ||
c2febafc KS |
2153 | static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd, |
2154 | unsigned long address) | |
2155 | { | |
2156 | return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ? | |
2157 | NULL : p4d_offset(pgd, address); | |
2158 | } | |
2159 | ||
2160 | static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d, | |
2161 | unsigned long address) | |
1da177e4 | 2162 | { |
c2febafc KS |
2163 | return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ? |
2164 | NULL : pud_offset(p4d, address); | |
1da177e4 | 2165 | } |
d8626138 | 2166 | |
1da177e4 LT |
2167 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
2168 | { | |
1bb3630e HD |
2169 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
2170 | NULL: pmd_offset(pud, address); | |
1da177e4 | 2171 | } |
f949286c | 2172 | #endif /* CONFIG_MMU */ |
1bb3630e | 2173 | |
57c1ffce | 2174 | #if USE_SPLIT_PTE_PTLOCKS |
597d795a | 2175 | #if ALLOC_SPLIT_PTLOCKS |
b35f1819 | 2176 | void __init ptlock_cache_init(void); |
539edb58 PZ |
2177 | extern bool ptlock_alloc(struct page *page); |
2178 | extern void ptlock_free(struct page *page); | |
2179 | ||
2180 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
2181 | { | |
2182 | return page->ptl; | |
2183 | } | |
597d795a | 2184 | #else /* ALLOC_SPLIT_PTLOCKS */ |
b35f1819 KS |
2185 | static inline void ptlock_cache_init(void) |
2186 | { | |
2187 | } | |
2188 | ||
49076ec2 KS |
2189 | static inline bool ptlock_alloc(struct page *page) |
2190 | { | |
49076ec2 KS |
2191 | return true; |
2192 | } | |
539edb58 | 2193 | |
49076ec2 KS |
2194 | static inline void ptlock_free(struct page *page) |
2195 | { | |
49076ec2 KS |
2196 | } |
2197 | ||
2198 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
2199 | { | |
539edb58 | 2200 | return &page->ptl; |
49076ec2 | 2201 | } |
597d795a | 2202 | #endif /* ALLOC_SPLIT_PTLOCKS */ |
49076ec2 KS |
2203 | |
2204 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
2205 | { | |
2206 | return ptlock_ptr(pmd_page(*pmd)); | |
2207 | } | |
2208 | ||
2209 | static inline bool ptlock_init(struct page *page) | |
2210 | { | |
2211 | /* | |
2212 | * prep_new_page() initialize page->private (and therefore page->ptl) | |
2213 | * with 0. Make sure nobody took it in use in between. | |
2214 | * | |
2215 | * It can happen if arch try to use slab for page table allocation: | |
1d798ca3 | 2216 | * slab code uses page->slab_cache, which share storage with page->ptl. |
49076ec2 | 2217 | */ |
309381fe | 2218 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); |
49076ec2 KS |
2219 | if (!ptlock_alloc(page)) |
2220 | return false; | |
2221 | spin_lock_init(ptlock_ptr(page)); | |
2222 | return true; | |
2223 | } | |
2224 | ||
57c1ffce | 2225 | #else /* !USE_SPLIT_PTE_PTLOCKS */ |
4c21e2f2 HD |
2226 | /* |
2227 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
2228 | */ | |
49076ec2 KS |
2229 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2230 | { | |
2231 | return &mm->page_table_lock; | |
2232 | } | |
b35f1819 | 2233 | static inline void ptlock_cache_init(void) {} |
49076ec2 | 2234 | static inline bool ptlock_init(struct page *page) { return true; } |
9e247bab | 2235 | static inline void ptlock_free(struct page *page) {} |
57c1ffce | 2236 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
4c21e2f2 | 2237 | |
b35f1819 KS |
2238 | static inline void pgtable_init(void) |
2239 | { | |
2240 | ptlock_cache_init(); | |
2241 | pgtable_cache_init(); | |
2242 | } | |
2243 | ||
b4ed71f5 | 2244 | static inline bool pgtable_pte_page_ctor(struct page *page) |
2f569afd | 2245 | { |
706874e9 VD |
2246 | if (!ptlock_init(page)) |
2247 | return false; | |
1d40a5ea | 2248 | __SetPageTable(page); |
f0c0c115 | 2249 | inc_lruvec_page_state(page, NR_PAGETABLE); |
706874e9 | 2250 | return true; |
2f569afd MS |
2251 | } |
2252 | ||
b4ed71f5 | 2253 | static inline void pgtable_pte_page_dtor(struct page *page) |
2f569afd | 2254 | { |
9e247bab | 2255 | ptlock_free(page); |
1d40a5ea | 2256 | __ClearPageTable(page); |
f0c0c115 | 2257 | dec_lruvec_page_state(page, NR_PAGETABLE); |
2f569afd MS |
2258 | } |
2259 | ||
c74df32c HD |
2260 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
2261 | ({ \ | |
4c21e2f2 | 2262 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
c74df32c HD |
2263 | pte_t *__pte = pte_offset_map(pmd, address); \ |
2264 | *(ptlp) = __ptl; \ | |
2265 | spin_lock(__ptl); \ | |
2266 | __pte; \ | |
2267 | }) | |
2268 | ||
2269 | #define pte_unmap_unlock(pte, ptl) do { \ | |
2270 | spin_unlock(ptl); \ | |
2271 | pte_unmap(pte); \ | |
2272 | } while (0) | |
2273 | ||
4cf58924 | 2274 | #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd)) |
3ed3a4f0 KS |
2275 | |
2276 | #define pte_alloc_map(mm, pmd, address) \ | |
4cf58924 | 2277 | (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address)) |
1bb3630e | 2278 | |
c74df32c | 2279 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
4cf58924 | 2280 | (pte_alloc(mm, pmd) ? \ |
3ed3a4f0 | 2281 | NULL : pte_offset_map_lock(mm, pmd, address, ptlp)) |
c74df32c | 2282 | |
1bb3630e | 2283 | #define pte_alloc_kernel(pmd, address) \ |
4cf58924 | 2284 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \ |
1bb3630e | 2285 | NULL: pte_offset_kernel(pmd, address)) |
1da177e4 | 2286 | |
e009bb30 KS |
2287 | #if USE_SPLIT_PMD_PTLOCKS |
2288 | ||
634391ac MS |
2289 | static struct page *pmd_to_page(pmd_t *pmd) |
2290 | { | |
2291 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); | |
2292 | return virt_to_page((void *)((unsigned long) pmd & mask)); | |
2293 | } | |
2294 | ||
e009bb30 KS |
2295 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2296 | { | |
634391ac | 2297 | return ptlock_ptr(pmd_to_page(pmd)); |
e009bb30 KS |
2298 | } |
2299 | ||
b2b29d6d | 2300 | static inline bool pmd_ptlock_init(struct page *page) |
e009bb30 | 2301 | { |
e009bb30 KS |
2302 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
2303 | page->pmd_huge_pte = NULL; | |
2304 | #endif | |
49076ec2 | 2305 | return ptlock_init(page); |
e009bb30 KS |
2306 | } |
2307 | ||
b2b29d6d | 2308 | static inline void pmd_ptlock_free(struct page *page) |
e009bb30 KS |
2309 | { |
2310 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
309381fe | 2311 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); |
e009bb30 | 2312 | #endif |
49076ec2 | 2313 | ptlock_free(page); |
e009bb30 KS |
2314 | } |
2315 | ||
634391ac | 2316 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) |
e009bb30 KS |
2317 | |
2318 | #else | |
2319 | ||
9a86cb7b KS |
2320 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2321 | { | |
2322 | return &mm->page_table_lock; | |
2323 | } | |
2324 | ||
b2b29d6d MW |
2325 | static inline bool pmd_ptlock_init(struct page *page) { return true; } |
2326 | static inline void pmd_ptlock_free(struct page *page) {} | |
e009bb30 | 2327 | |
c389a250 | 2328 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) |
9a86cb7b | 2329 | |
e009bb30 KS |
2330 | #endif |
2331 | ||
9a86cb7b KS |
2332 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) |
2333 | { | |
2334 | spinlock_t *ptl = pmd_lockptr(mm, pmd); | |
2335 | spin_lock(ptl); | |
2336 | return ptl; | |
2337 | } | |
2338 | ||
b2b29d6d MW |
2339 | static inline bool pgtable_pmd_page_ctor(struct page *page) |
2340 | { | |
2341 | if (!pmd_ptlock_init(page)) | |
2342 | return false; | |
2343 | __SetPageTable(page); | |
f0c0c115 | 2344 | inc_lruvec_page_state(page, NR_PAGETABLE); |
b2b29d6d MW |
2345 | return true; |
2346 | } | |
2347 | ||
2348 | static inline void pgtable_pmd_page_dtor(struct page *page) | |
2349 | { | |
2350 | pmd_ptlock_free(page); | |
2351 | __ClearPageTable(page); | |
f0c0c115 | 2352 | dec_lruvec_page_state(page, NR_PAGETABLE); |
b2b29d6d MW |
2353 | } |
2354 | ||
a00cc7d9 MW |
2355 | /* |
2356 | * No scalability reason to split PUD locks yet, but follow the same pattern | |
2357 | * as the PMD locks to make it easier if we decide to. The VM should not be | |
2358 | * considered ready to switch to split PUD locks yet; there may be places | |
2359 | * which need to be converted from page_table_lock. | |
2360 | */ | |
2361 | static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud) | |
2362 | { | |
2363 | return &mm->page_table_lock; | |
2364 | } | |
2365 | ||
2366 | static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud) | |
2367 | { | |
2368 | spinlock_t *ptl = pud_lockptr(mm, pud); | |
2369 | ||
2370 | spin_lock(ptl); | |
2371 | return ptl; | |
2372 | } | |
62906027 | 2373 | |
a00cc7d9 | 2374 | extern void __init pagecache_init(void); |
bc9331a1 | 2375 | extern void __init free_area_init_memoryless_node(int nid); |
49a7f04a DH |
2376 | extern void free_initmem(void); |
2377 | ||
69afade7 JL |
2378 | /* |
2379 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) | |
2380 | * into the buddy system. The freed pages will be poisoned with pattern | |
dbe67df4 | 2381 | * "poison" if it's within range [0, UCHAR_MAX]. |
69afade7 JL |
2382 | * Return pages freed into the buddy system. |
2383 | */ | |
11199692 | 2384 | extern unsigned long free_reserved_area(void *start, void *end, |
e5cb113f | 2385 | int poison, const char *s); |
c3d5f5f0 | 2386 | |
c3d5f5f0 | 2387 | extern void adjust_managed_page_count(struct page *page, long count); |
1f9d03c5 | 2388 | extern void mem_init_print_info(void); |
69afade7 | 2389 | |
4b50bcc7 | 2390 | extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end); |
92923ca3 | 2391 | |
69afade7 | 2392 | /* Free the reserved page into the buddy system, so it gets managed. */ |
a0cd7a7c | 2393 | static inline void free_reserved_page(struct page *page) |
69afade7 JL |
2394 | { |
2395 | ClearPageReserved(page); | |
2396 | init_page_count(page); | |
2397 | __free_page(page); | |
69afade7 JL |
2398 | adjust_managed_page_count(page, 1); |
2399 | } | |
a0cd7a7c | 2400 | #define free_highmem_page(page) free_reserved_page(page) |
69afade7 JL |
2401 | |
2402 | static inline void mark_page_reserved(struct page *page) | |
2403 | { | |
2404 | SetPageReserved(page); | |
2405 | adjust_managed_page_count(page, -1); | |
2406 | } | |
2407 | ||
2408 | /* | |
2409 | * Default method to free all the __init memory into the buddy system. | |
dbe67df4 JL |
2410 | * The freed pages will be poisoned with pattern "poison" if it's within |
2411 | * range [0, UCHAR_MAX]. | |
2412 | * Return pages freed into the buddy system. | |
69afade7 JL |
2413 | */ |
2414 | static inline unsigned long free_initmem_default(int poison) | |
2415 | { | |
2416 | extern char __init_begin[], __init_end[]; | |
2417 | ||
11199692 | 2418 | return free_reserved_area(&__init_begin, &__init_end, |
c5a54c70 | 2419 | poison, "unused kernel image (initmem)"); |
69afade7 JL |
2420 | } |
2421 | ||
7ee3d4e8 JL |
2422 | static inline unsigned long get_num_physpages(void) |
2423 | { | |
2424 | int nid; | |
2425 | unsigned long phys_pages = 0; | |
2426 | ||
2427 | for_each_online_node(nid) | |
2428 | phys_pages += node_present_pages(nid); | |
2429 | ||
2430 | return phys_pages; | |
2431 | } | |
2432 | ||
c713216d | 2433 | /* |
3f08a302 | 2434 | * Using memblock node mappings, an architecture may initialise its |
bc9331a1 MR |
2435 | * zones, allocate the backing mem_map and account for memory holes in an |
2436 | * architecture independent manner. | |
c713216d MG |
2437 | * |
2438 | * An architecture is expected to register range of page frames backed by | |
0ee332c1 | 2439 | * physical memory with memblock_add[_node]() before calling |
9691a071 | 2440 | * free_area_init() passing in the PFN each zone ends at. At a basic |
c713216d MG |
2441 | * usage, an architecture is expected to do something like |
2442 | * | |
2443 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
2444 | * max_highmem_pfn}; | |
2445 | * for_each_valid_physical_page_range() | |
0ee332c1 | 2446 | * memblock_add_node(base, size, nid) |
9691a071 | 2447 | * free_area_init(max_zone_pfns); |
c713216d | 2448 | */ |
9691a071 | 2449 | void free_area_init(unsigned long *max_zone_pfn); |
1e01979c | 2450 | unsigned long node_map_pfn_alignment(void); |
32996250 YL |
2451 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, |
2452 | unsigned long end_pfn); | |
c713216d MG |
2453 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
2454 | unsigned long end_pfn); | |
2455 | extern void get_pfn_range_for_nid(unsigned int nid, | |
2456 | unsigned long *start_pfn, unsigned long *end_pfn); | |
2457 | extern unsigned long find_min_pfn_with_active_regions(void); | |
f2dbcfa7 | 2458 | |
3f08a302 | 2459 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
6f24fbd3 | 2460 | static inline int early_pfn_to_nid(unsigned long pfn) |
f2dbcfa7 KH |
2461 | { |
2462 | return 0; | |
2463 | } | |
2464 | #else | |
2465 | /* please see mm/page_alloc.c */ | |
2466 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | |
f2dbcfa7 KH |
2467 | #endif |
2468 | ||
0e0b864e | 2469 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
ab28cb6e | 2470 | extern void memmap_init_range(unsigned long, int, unsigned long, |
dc2da7b4 BH |
2471 | unsigned long, unsigned long, enum meminit_context, |
2472 | struct vmem_altmap *, int migratetype); | |
bc75d33f | 2473 | extern void setup_per_zone_wmarks(void); |
1b79acc9 | 2474 | extern int __meminit init_per_zone_wmark_min(void); |
1da177e4 | 2475 | extern void mem_init(void); |
8feae131 | 2476 | extern void __init mmap_init(void); |
9af744d7 | 2477 | extern void show_mem(unsigned int flags, nodemask_t *nodemask); |
d02bd27b | 2478 | extern long si_mem_available(void); |
1da177e4 LT |
2479 | extern void si_meminfo(struct sysinfo * val); |
2480 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
f6f34b43 SD |
2481 | #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES |
2482 | extern unsigned long arch_reserved_kernel_pages(void); | |
2483 | #endif | |
1da177e4 | 2484 | |
a8e99259 MH |
2485 | extern __printf(3, 4) |
2486 | void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...); | |
a238ab5b | 2487 | |
e7c8d5c9 | 2488 | extern void setup_per_cpu_pageset(void); |
e7c8d5c9 | 2489 | |
75f7ad8e PS |
2490 | /* page_alloc.c */ |
2491 | extern int min_free_kbytes; | |
1c30844d | 2492 | extern int watermark_boost_factor; |
795ae7a0 | 2493 | extern int watermark_scale_factor; |
51930df5 | 2494 | extern bool arch_has_descending_max_zone_pfns(void); |
75f7ad8e | 2495 | |
8feae131 | 2496 | /* nommu.c */ |
33e5d769 | 2497 | extern atomic_long_t mmap_pages_allocated; |
7e660872 | 2498 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
8feae131 | 2499 | |
6b2dbba8 | 2500 | /* interval_tree.c */ |
6b2dbba8 | 2501 | void vma_interval_tree_insert(struct vm_area_struct *node, |
f808c13f | 2502 | struct rb_root_cached *root); |
9826a516 ML |
2503 | void vma_interval_tree_insert_after(struct vm_area_struct *node, |
2504 | struct vm_area_struct *prev, | |
f808c13f | 2505 | struct rb_root_cached *root); |
6b2dbba8 | 2506 | void vma_interval_tree_remove(struct vm_area_struct *node, |
f808c13f DB |
2507 | struct rb_root_cached *root); |
2508 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root, | |
6b2dbba8 ML |
2509 | unsigned long start, unsigned long last); |
2510 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, | |
2511 | unsigned long start, unsigned long last); | |
2512 | ||
2513 | #define vma_interval_tree_foreach(vma, root, start, last) \ | |
2514 | for (vma = vma_interval_tree_iter_first(root, start, last); \ | |
2515 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) | |
1da177e4 | 2516 | |
bf181b9f | 2517 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, |
f808c13f | 2518 | struct rb_root_cached *root); |
bf181b9f | 2519 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, |
f808c13f DB |
2520 | struct rb_root_cached *root); |
2521 | struct anon_vma_chain * | |
2522 | anon_vma_interval_tree_iter_first(struct rb_root_cached *root, | |
2523 | unsigned long start, unsigned long last); | |
bf181b9f ML |
2524 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( |
2525 | struct anon_vma_chain *node, unsigned long start, unsigned long last); | |
ed8ea815 ML |
2526 | #ifdef CONFIG_DEBUG_VM_RB |
2527 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); | |
2528 | #endif | |
bf181b9f ML |
2529 | |
2530 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ | |
2531 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ | |
2532 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) | |
2533 | ||
1da177e4 | 2534 | /* mmap.c */ |
34b4e4aa | 2535 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
e86f15ee AA |
2536 | extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start, |
2537 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert, | |
2538 | struct vm_area_struct *expand); | |
2539 | static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
2540 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert) | |
2541 | { | |
2542 | return __vma_adjust(vma, start, end, pgoff, insert, NULL); | |
2543 | } | |
1da177e4 LT |
2544 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
2545 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
2546 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
19a809af | 2547 | struct mempolicy *, struct vm_userfaultfd_ctx); |
1da177e4 | 2548 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
def5efe0 DR |
2549 | extern int __split_vma(struct mm_struct *, struct vm_area_struct *, |
2550 | unsigned long addr, int new_below); | |
2551 | extern int split_vma(struct mm_struct *, struct vm_area_struct *, | |
2552 | unsigned long addr, int new_below); | |
1da177e4 LT |
2553 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
2554 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
2555 | struct rb_node **, struct rb_node *); | |
a8fb5618 | 2556 | extern void unlink_file_vma(struct vm_area_struct *); |
1da177e4 | 2557 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
38a76013 ML |
2558 | unsigned long addr, unsigned long len, pgoff_t pgoff, |
2559 | bool *need_rmap_locks); | |
1da177e4 | 2560 | extern void exit_mmap(struct mm_struct *); |
925d1c40 | 2561 | |
9c599024 CG |
2562 | static inline int check_data_rlimit(unsigned long rlim, |
2563 | unsigned long new, | |
2564 | unsigned long start, | |
2565 | unsigned long end_data, | |
2566 | unsigned long start_data) | |
2567 | { | |
2568 | if (rlim < RLIM_INFINITY) { | |
2569 | if (((new - start) + (end_data - start_data)) > rlim) | |
2570 | return -ENOSPC; | |
2571 | } | |
2572 | ||
2573 | return 0; | |
2574 | } | |
2575 | ||
7906d00c AA |
2576 | extern int mm_take_all_locks(struct mm_struct *mm); |
2577 | extern void mm_drop_all_locks(struct mm_struct *mm); | |
2578 | ||
38646013 JS |
2579 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); |
2580 | extern struct file *get_mm_exe_file(struct mm_struct *mm); | |
cd81a917 | 2581 | extern struct file *get_task_exe_file(struct task_struct *task); |
925d1c40 | 2582 | |
84638335 KK |
2583 | extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages); |
2584 | extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages); | |
2585 | ||
2eefd878 DS |
2586 | extern bool vma_is_special_mapping(const struct vm_area_struct *vma, |
2587 | const struct vm_special_mapping *sm); | |
3935ed6a SS |
2588 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, |
2589 | unsigned long addr, unsigned long len, | |
a62c34bd AL |
2590 | unsigned long flags, |
2591 | const struct vm_special_mapping *spec); | |
2592 | /* This is an obsolete alternative to _install_special_mapping. */ | |
fa5dc22f RM |
2593 | extern int install_special_mapping(struct mm_struct *mm, |
2594 | unsigned long addr, unsigned long len, | |
2595 | unsigned long flags, struct page **pages); | |
1da177e4 | 2596 | |
649775be AG |
2597 | unsigned long randomize_stack_top(unsigned long stack_top); |
2598 | ||
1da177e4 LT |
2599 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
2600 | ||
0165ab44 | 2601 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
897ab3e0 MR |
2602 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff, |
2603 | struct list_head *uf); | |
1fcfd8db | 2604 | extern unsigned long do_mmap(struct file *file, unsigned long addr, |
bebeb3d6 | 2605 | unsigned long len, unsigned long prot, unsigned long flags, |
45e55300 | 2606 | unsigned long pgoff, unsigned long *populate, struct list_head *uf); |
85a06835 YS |
2607 | extern int __do_munmap(struct mm_struct *, unsigned long, size_t, |
2608 | struct list_head *uf, bool downgrade); | |
897ab3e0 MR |
2609 | extern int do_munmap(struct mm_struct *, unsigned long, size_t, |
2610 | struct list_head *uf); | |
0726b01e | 2611 | extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior); |
1da177e4 | 2612 | |
bebeb3d6 ML |
2613 | #ifdef CONFIG_MMU |
2614 | extern int __mm_populate(unsigned long addr, unsigned long len, | |
2615 | int ignore_errors); | |
2616 | static inline void mm_populate(unsigned long addr, unsigned long len) | |
2617 | { | |
2618 | /* Ignore errors */ | |
2619 | (void) __mm_populate(addr, len, 1); | |
2620 | } | |
2621 | #else | |
2622 | static inline void mm_populate(unsigned long addr, unsigned long len) {} | |
2623 | #endif | |
2624 | ||
e4eb1ff6 | 2625 | /* These take the mm semaphore themselves */ |
5d22fc25 | 2626 | extern int __must_check vm_brk(unsigned long, unsigned long); |
16e72e9b | 2627 | extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long); |
bfce281c | 2628 | extern int vm_munmap(unsigned long, size_t); |
9fbeb5ab | 2629 | extern unsigned long __must_check vm_mmap(struct file *, unsigned long, |
6be5ceb0 LT |
2630 | unsigned long, unsigned long, |
2631 | unsigned long, unsigned long); | |
1da177e4 | 2632 | |
db4fbfb9 ML |
2633 | struct vm_unmapped_area_info { |
2634 | #define VM_UNMAPPED_AREA_TOPDOWN 1 | |
2635 | unsigned long flags; | |
2636 | unsigned long length; | |
2637 | unsigned long low_limit; | |
2638 | unsigned long high_limit; | |
2639 | unsigned long align_mask; | |
2640 | unsigned long align_offset; | |
2641 | }; | |
2642 | ||
baceaf1c | 2643 | extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info); |
db4fbfb9 | 2644 | |
85821aab | 2645 | /* truncate.c */ |
1da177e4 | 2646 | extern void truncate_inode_pages(struct address_space *, loff_t); |
d7339071 HR |
2647 | extern void truncate_inode_pages_range(struct address_space *, |
2648 | loff_t lstart, loff_t lend); | |
91b0abe3 | 2649 | extern void truncate_inode_pages_final(struct address_space *); |
1da177e4 LT |
2650 | |
2651 | /* generic vm_area_ops exported for stackable file systems */ | |
2bcd6454 | 2652 | extern vm_fault_t filemap_fault(struct vm_fault *vmf); |
f9ce0be7 | 2653 | extern vm_fault_t filemap_map_pages(struct vm_fault *vmf, |
bae473a4 | 2654 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
2bcd6454 | 2655 | extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf); |
1da177e4 LT |
2656 | |
2657 | /* mm/page-writeback.c */ | |
2b69c828 | 2658 | int __must_check write_one_page(struct page *page); |
1cf6e7d8 | 2659 | void task_dirty_inc(struct task_struct *tsk); |
1da177e4 | 2660 | |
1be7107f | 2661 | extern unsigned long stack_guard_gap; |
d05f3169 | 2662 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ |
46dea3d0 | 2663 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
d05f3169 | 2664 | |
11192337 | 2665 | /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */ |
d05f3169 MH |
2666 | extern int expand_downwards(struct vm_area_struct *vma, |
2667 | unsigned long address); | |
8ca3eb08 | 2668 | #if VM_GROWSUP |
46dea3d0 | 2669 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
8ca3eb08 | 2670 | #else |
fee7e49d | 2671 | #define expand_upwards(vma, address) (0) |
9ab88515 | 2672 | #endif |
1da177e4 LT |
2673 | |
2674 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
2675 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
2676 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
2677 | struct vm_area_struct **pprev); | |
2678 | ||
ce6d42f2 LH |
2679 | /** |
2680 | * find_vma_intersection() - Look up the first VMA which intersects the interval | |
2681 | * @mm: The process address space. | |
2682 | * @start_addr: The inclusive start user address. | |
2683 | * @end_addr: The exclusive end user address. | |
2684 | * | |
2685 | * Returns: The first VMA within the provided range, %NULL otherwise. Assumes | |
2686 | * start_addr < end_addr. | |
2687 | */ | |
2688 | static inline | |
2689 | struct vm_area_struct *find_vma_intersection(struct mm_struct *mm, | |
2690 | unsigned long start_addr, | |
2691 | unsigned long end_addr) | |
1da177e4 | 2692 | { |
ce6d42f2 | 2693 | struct vm_area_struct *vma = find_vma(mm, start_addr); |
1da177e4 LT |
2694 | |
2695 | if (vma && end_addr <= vma->vm_start) | |
2696 | vma = NULL; | |
2697 | return vma; | |
2698 | } | |
2699 | ||
ce6d42f2 LH |
2700 | /** |
2701 | * vma_lookup() - Find a VMA at a specific address | |
2702 | * @mm: The process address space. | |
2703 | * @addr: The user address. | |
2704 | * | |
2705 | * Return: The vm_area_struct at the given address, %NULL otherwise. | |
2706 | */ | |
2707 | static inline | |
2708 | struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr) | |
2709 | { | |
2710 | struct vm_area_struct *vma = find_vma(mm, addr); | |
2711 | ||
2712 | if (vma && addr < vma->vm_start) | |
2713 | vma = NULL; | |
2714 | ||
2715 | return vma; | |
2716 | } | |
2717 | ||
1be7107f HD |
2718 | static inline unsigned long vm_start_gap(struct vm_area_struct *vma) |
2719 | { | |
2720 | unsigned long vm_start = vma->vm_start; | |
2721 | ||
2722 | if (vma->vm_flags & VM_GROWSDOWN) { | |
2723 | vm_start -= stack_guard_gap; | |
2724 | if (vm_start > vma->vm_start) | |
2725 | vm_start = 0; | |
2726 | } | |
2727 | return vm_start; | |
2728 | } | |
2729 | ||
2730 | static inline unsigned long vm_end_gap(struct vm_area_struct *vma) | |
2731 | { | |
2732 | unsigned long vm_end = vma->vm_end; | |
2733 | ||
2734 | if (vma->vm_flags & VM_GROWSUP) { | |
2735 | vm_end += stack_guard_gap; | |
2736 | if (vm_end < vma->vm_end) | |
2737 | vm_end = -PAGE_SIZE; | |
2738 | } | |
2739 | return vm_end; | |
2740 | } | |
2741 | ||
1da177e4 LT |
2742 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
2743 | { | |
2744 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
2745 | } | |
2746 | ||
640708a2 PE |
2747 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ |
2748 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, | |
2749 | unsigned long vm_start, unsigned long vm_end) | |
2750 | { | |
2751 | struct vm_area_struct *vma = find_vma(mm, vm_start); | |
2752 | ||
2753 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) | |
2754 | vma = NULL; | |
2755 | ||
2756 | return vma; | |
2757 | } | |
2758 | ||
017b1660 MK |
2759 | static inline bool range_in_vma(struct vm_area_struct *vma, |
2760 | unsigned long start, unsigned long end) | |
2761 | { | |
2762 | return (vma && vma->vm_start <= start && end <= vma->vm_end); | |
2763 | } | |
2764 | ||
bad849b3 | 2765 | #ifdef CONFIG_MMU |
804af2cf | 2766 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
64e45507 | 2767 | void vma_set_page_prot(struct vm_area_struct *vma); |
bad849b3 DH |
2768 | #else |
2769 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | |
2770 | { | |
2771 | return __pgprot(0); | |
2772 | } | |
64e45507 PF |
2773 | static inline void vma_set_page_prot(struct vm_area_struct *vma) |
2774 | { | |
2775 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); | |
2776 | } | |
bad849b3 DH |
2777 | #endif |
2778 | ||
295992fb CK |
2779 | void vma_set_file(struct vm_area_struct *vma, struct file *file); |
2780 | ||
5877231f | 2781 | #ifdef CONFIG_NUMA_BALANCING |
4b10e7d5 | 2782 | unsigned long change_prot_numa(struct vm_area_struct *vma, |
b24f53a0 LS |
2783 | unsigned long start, unsigned long end); |
2784 | #endif | |
2785 | ||
deceb6cd | 2786 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
deceb6cd HD |
2787 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
2788 | unsigned long pfn, unsigned long size, pgprot_t); | |
74ffa5a3 CH |
2789 | int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr, |
2790 | unsigned long pfn, unsigned long size, pgprot_t prot); | |
a145dd41 | 2791 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
8cd3984d AR |
2792 | int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr, |
2793 | struct page **pages, unsigned long *num); | |
a667d745 SJ |
2794 | int vm_map_pages(struct vm_area_struct *vma, struct page **pages, |
2795 | unsigned long num); | |
2796 | int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages, | |
2797 | unsigned long num); | |
ae2b01f3 | 2798 | vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
e0dc0d8f | 2799 | unsigned long pfn); |
f5e6d1d5 MW |
2800 | vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr, |
2801 | unsigned long pfn, pgprot_t pgprot); | |
5d747637 | 2802 | vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
01c8f1c4 | 2803 | pfn_t pfn); |
574c5b3d TH |
2804 | vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr, |
2805 | pfn_t pfn, pgprot_t pgprot); | |
ab77dab4 SJ |
2806 | vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma, |
2807 | unsigned long addr, pfn_t pfn); | |
b4cbb197 LT |
2808 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); |
2809 | ||
1c8f4220 SJ |
2810 | static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma, |
2811 | unsigned long addr, struct page *page) | |
2812 | { | |
2813 | int err = vm_insert_page(vma, addr, page); | |
2814 | ||
2815 | if (err == -ENOMEM) | |
2816 | return VM_FAULT_OOM; | |
2817 | if (err < 0 && err != -EBUSY) | |
2818 | return VM_FAULT_SIGBUS; | |
2819 | ||
2820 | return VM_FAULT_NOPAGE; | |
2821 | } | |
2822 | ||
f8f6ae5d JG |
2823 | #ifndef io_remap_pfn_range |
2824 | static inline int io_remap_pfn_range(struct vm_area_struct *vma, | |
2825 | unsigned long addr, unsigned long pfn, | |
2826 | unsigned long size, pgprot_t prot) | |
2827 | { | |
2828 | return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot)); | |
2829 | } | |
2830 | #endif | |
2831 | ||
d97baf94 SJ |
2832 | static inline vm_fault_t vmf_error(int err) |
2833 | { | |
2834 | if (err == -ENOMEM) | |
2835 | return VM_FAULT_OOM; | |
2836 | return VM_FAULT_SIGBUS; | |
2837 | } | |
2838 | ||
df06b37f KB |
2839 | struct page *follow_page(struct vm_area_struct *vma, unsigned long address, |
2840 | unsigned int foll_flags); | |
240aadee | 2841 | |
deceb6cd HD |
2842 | #define FOLL_WRITE 0x01 /* check pte is writable */ |
2843 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
2844 | #define FOLL_GET 0x04 /* do get_page on page */ | |
8e4b9a60 | 2845 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ |
58fa879e | 2846 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ |
318b275f GN |
2847 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO |
2848 | * and return without waiting upon it */ | |
84d33df2 | 2849 | #define FOLL_POPULATE 0x40 /* fault in page */ |
69ebb83e | 2850 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ |
0b9d7052 | 2851 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ |
5117b3b8 | 2852 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ |
234b239b | 2853 | #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */ |
de60f5f1 | 2854 | #define FOLL_MLOCK 0x1000 /* lock present pages */ |
1e987790 | 2855 | #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */ |
19be0eaf | 2856 | #define FOLL_COW 0x4000 /* internal GUP flag */ |
7f7ccc2c | 2857 | #define FOLL_ANON 0x8000 /* don't do file mappings */ |
932f4a63 | 2858 | #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */ |
bfe7b00d | 2859 | #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */ |
f1f6a7dd | 2860 | #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */ |
376a34ef | 2861 | #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */ |
932f4a63 IW |
2862 | |
2863 | /* | |
eddb1c22 JH |
2864 | * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each |
2865 | * other. Here is what they mean, and how to use them: | |
932f4a63 IW |
2866 | * |
2867 | * FOLL_LONGTERM indicates that the page will be held for an indefinite time | |
eddb1c22 JH |
2868 | * period _often_ under userspace control. This is in contrast to |
2869 | * iov_iter_get_pages(), whose usages are transient. | |
932f4a63 IW |
2870 | * |
2871 | * FIXME: For pages which are part of a filesystem, mappings are subject to the | |
2872 | * lifetime enforced by the filesystem and we need guarantees that longterm | |
2873 | * users like RDMA and V4L2 only establish mappings which coordinate usage with | |
2874 | * the filesystem. Ideas for this coordination include revoking the longterm | |
2875 | * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was | |
2876 | * added after the problem with filesystems was found FS DAX VMAs are | |
2877 | * specifically failed. Filesystem pages are still subject to bugs and use of | |
2878 | * FOLL_LONGTERM should be avoided on those pages. | |
2879 | * | |
2880 | * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call. | |
2881 | * Currently only get_user_pages() and get_user_pages_fast() support this flag | |
2882 | * and calls to get_user_pages_[un]locked are specifically not allowed. This | |
2883 | * is due to an incompatibility with the FS DAX check and | |
eddb1c22 | 2884 | * FAULT_FLAG_ALLOW_RETRY. |
932f4a63 | 2885 | * |
eddb1c22 JH |
2886 | * In the CMA case: long term pins in a CMA region would unnecessarily fragment |
2887 | * that region. And so, CMA attempts to migrate the page before pinning, when | |
932f4a63 | 2888 | * FOLL_LONGTERM is specified. |
eddb1c22 JH |
2889 | * |
2890 | * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount, | |
2891 | * but an additional pin counting system) will be invoked. This is intended for | |
2892 | * anything that gets a page reference and then touches page data (for example, | |
2893 | * Direct IO). This lets the filesystem know that some non-file-system entity is | |
2894 | * potentially changing the pages' data. In contrast to FOLL_GET (whose pages | |
2895 | * are released via put_page()), FOLL_PIN pages must be released, ultimately, by | |
f1f6a7dd | 2896 | * a call to unpin_user_page(). |
eddb1c22 JH |
2897 | * |
2898 | * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different | |
2899 | * and separate refcounting mechanisms, however, and that means that each has | |
2900 | * its own acquire and release mechanisms: | |
2901 | * | |
2902 | * FOLL_GET: get_user_pages*() to acquire, and put_page() to release. | |
2903 | * | |
f1f6a7dd | 2904 | * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release. |
eddb1c22 JH |
2905 | * |
2906 | * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call. | |
2907 | * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based | |
2908 | * calls applied to them, and that's perfectly OK. This is a constraint on the | |
2909 | * callers, not on the pages.) | |
2910 | * | |
2911 | * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never | |
2912 | * directly by the caller. That's in order to help avoid mismatches when | |
2913 | * releasing pages: get_user_pages*() pages must be released via put_page(), | |
f1f6a7dd | 2914 | * while pin_user_pages*() pages must be released via unpin_user_page(). |
eddb1c22 | 2915 | * |
72ef5e52 | 2916 | * Please see Documentation/core-api/pin_user_pages.rst for more information. |
932f4a63 | 2917 | */ |
1da177e4 | 2918 | |
2b740303 | 2919 | static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags) |
9a291a7c JM |
2920 | { |
2921 | if (vm_fault & VM_FAULT_OOM) | |
2922 | return -ENOMEM; | |
2923 | if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) | |
2924 | return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT; | |
2925 | if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV)) | |
2926 | return -EFAULT; | |
2927 | return 0; | |
2928 | } | |
2929 | ||
8b1e0f81 | 2930 | typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data); |
aee16b3c JF |
2931 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, |
2932 | unsigned long size, pte_fn_t fn, void *data); | |
be1db475 DA |
2933 | extern int apply_to_existing_page_range(struct mm_struct *mm, |
2934 | unsigned long address, unsigned long size, | |
2935 | pte_fn_t fn, void *data); | |
aee16b3c | 2936 | |
04013513 | 2937 | extern void init_mem_debugging_and_hardening(void); |
8823b1db | 2938 | #ifdef CONFIG_PAGE_POISONING |
8db26a3d VB |
2939 | extern void __kernel_poison_pages(struct page *page, int numpages); |
2940 | extern void __kernel_unpoison_pages(struct page *page, int numpages); | |
2941 | extern bool _page_poisoning_enabled_early; | |
2942 | DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled); | |
2943 | static inline bool page_poisoning_enabled(void) | |
2944 | { | |
2945 | return _page_poisoning_enabled_early; | |
2946 | } | |
2947 | /* | |
2948 | * For use in fast paths after init_mem_debugging() has run, or when a | |
2949 | * false negative result is not harmful when called too early. | |
2950 | */ | |
2951 | static inline bool page_poisoning_enabled_static(void) | |
2952 | { | |
2953 | return static_branch_unlikely(&_page_poisoning_enabled); | |
2954 | } | |
2955 | static inline void kernel_poison_pages(struct page *page, int numpages) | |
2956 | { | |
2957 | if (page_poisoning_enabled_static()) | |
2958 | __kernel_poison_pages(page, numpages); | |
2959 | } | |
2960 | static inline void kernel_unpoison_pages(struct page *page, int numpages) | |
2961 | { | |
2962 | if (page_poisoning_enabled_static()) | |
2963 | __kernel_unpoison_pages(page, numpages); | |
2964 | } | |
8823b1db LA |
2965 | #else |
2966 | static inline bool page_poisoning_enabled(void) { return false; } | |
8db26a3d | 2967 | static inline bool page_poisoning_enabled_static(void) { return false; } |
03b6c9a3 | 2968 | static inline void __kernel_poison_pages(struct page *page, int nunmpages) { } |
8db26a3d VB |
2969 | static inline void kernel_poison_pages(struct page *page, int numpages) { } |
2970 | static inline void kernel_unpoison_pages(struct page *page, int numpages) { } | |
8823b1db LA |
2971 | #endif |
2972 | ||
51cba1eb | 2973 | DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc); |
6471384a AP |
2974 | static inline bool want_init_on_alloc(gfp_t flags) |
2975 | { | |
51cba1eb KC |
2976 | if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, |
2977 | &init_on_alloc)) | |
6471384a AP |
2978 | return true; |
2979 | return flags & __GFP_ZERO; | |
2980 | } | |
2981 | ||
51cba1eb | 2982 | DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free); |
6471384a AP |
2983 | static inline bool want_init_on_free(void) |
2984 | { | |
51cba1eb KC |
2985 | return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON, |
2986 | &init_on_free); | |
6471384a AP |
2987 | } |
2988 | ||
8e57f8ac VB |
2989 | extern bool _debug_pagealloc_enabled_early; |
2990 | DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled); | |
031bc574 JK |
2991 | |
2992 | static inline bool debug_pagealloc_enabled(void) | |
8e57f8ac VB |
2993 | { |
2994 | return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) && | |
2995 | _debug_pagealloc_enabled_early; | |
2996 | } | |
2997 | ||
2998 | /* | |
2999 | * For use in fast paths after init_debug_pagealloc() has run, or when a | |
3000 | * false negative result is not harmful when called too early. | |
3001 | */ | |
3002 | static inline bool debug_pagealloc_enabled_static(void) | |
031bc574 | 3003 | { |
96a2b03f VB |
3004 | if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC)) |
3005 | return false; | |
3006 | ||
3007 | return static_branch_unlikely(&_debug_pagealloc_enabled); | |
031bc574 JK |
3008 | } |
3009 | ||
5d6ad668 | 3010 | #ifdef CONFIG_DEBUG_PAGEALLOC |
c87cbc1f | 3011 | /* |
5d6ad668 MR |
3012 | * To support DEBUG_PAGEALLOC architecture must ensure that |
3013 | * __kernel_map_pages() never fails | |
c87cbc1f | 3014 | */ |
d6332692 RE |
3015 | extern void __kernel_map_pages(struct page *page, int numpages, int enable); |
3016 | ||
77bc7fd6 MR |
3017 | static inline void debug_pagealloc_map_pages(struct page *page, int numpages) |
3018 | { | |
3019 | if (debug_pagealloc_enabled_static()) | |
3020 | __kernel_map_pages(page, numpages, 1); | |
3021 | } | |
3022 | ||
3023 | static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) | |
3024 | { | |
3025 | if (debug_pagealloc_enabled_static()) | |
3026 | __kernel_map_pages(page, numpages, 0); | |
3027 | } | |
5d6ad668 | 3028 | #else /* CONFIG_DEBUG_PAGEALLOC */ |
77bc7fd6 MR |
3029 | static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {} |
3030 | static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {} | |
5d6ad668 | 3031 | #endif /* CONFIG_DEBUG_PAGEALLOC */ |
1da177e4 | 3032 | |
a6c19dfe | 3033 | #ifdef __HAVE_ARCH_GATE_AREA |
31db58b3 | 3034 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
a6c19dfe AL |
3035 | extern int in_gate_area_no_mm(unsigned long addr); |
3036 | extern int in_gate_area(struct mm_struct *mm, unsigned long addr); | |
1da177e4 | 3037 | #else |
a6c19dfe AL |
3038 | static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) |
3039 | { | |
3040 | return NULL; | |
3041 | } | |
3042 | static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } | |
3043 | static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) | |
3044 | { | |
3045 | return 0; | |
3046 | } | |
1da177e4 LT |
3047 | #endif /* __HAVE_ARCH_GATE_AREA */ |
3048 | ||
44a70ade MH |
3049 | extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm); |
3050 | ||
146732ce JT |
3051 | #ifdef CONFIG_SYSCTL |
3052 | extern int sysctl_drop_caches; | |
32927393 CH |
3053 | int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *, |
3054 | loff_t *); | |
146732ce JT |
3055 | #endif |
3056 | ||
cb731d6c VD |
3057 | void drop_slab(void); |
3058 | void drop_slab_node(int nid); | |
9d0243bc | 3059 | |
7a9166e3 LY |
3060 | #ifndef CONFIG_MMU |
3061 | #define randomize_va_space 0 | |
3062 | #else | |
a62eaf15 | 3063 | extern int randomize_va_space; |
7a9166e3 | 3064 | #endif |
a62eaf15 | 3065 | |
045e72ac | 3066 | const char * arch_vma_name(struct vm_area_struct *vma); |
89165b8b | 3067 | #ifdef CONFIG_MMU |
03252919 | 3068 | void print_vma_addr(char *prefix, unsigned long rip); |
89165b8b CH |
3069 | #else |
3070 | static inline void print_vma_addr(char *prefix, unsigned long rip) | |
3071 | { | |
3072 | } | |
3073 | #endif | |
e6e5494c | 3074 | |
35fd1eb1 | 3075 | void *sparse_buffer_alloc(unsigned long size); |
e9c0a3f0 DW |
3076 | struct page * __populate_section_memmap(unsigned long pfn, |
3077 | unsigned long nr_pages, int nid, struct vmem_altmap *altmap); | |
29c71111 | 3078 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
c2febafc KS |
3079 | p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node); |
3080 | pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node); | |
29c71111 | 3081 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); |
1d9cfee7 AK |
3082 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node, |
3083 | struct vmem_altmap *altmap); | |
8f6aac41 | 3084 | void *vmemmap_alloc_block(unsigned long size, int node); |
4b94ffdc | 3085 | struct vmem_altmap; |
56993b4e AK |
3086 | void *vmemmap_alloc_block_buf(unsigned long size, int node, |
3087 | struct vmem_altmap *altmap); | |
8f6aac41 | 3088 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
0aad818b | 3089 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, |
1d9cfee7 | 3090 | int node, struct vmem_altmap *altmap); |
7b73d978 CH |
3091 | int vmemmap_populate(unsigned long start, unsigned long end, int node, |
3092 | struct vmem_altmap *altmap); | |
c2b91e2e | 3093 | void vmemmap_populate_print_last(void); |
0197518c | 3094 | #ifdef CONFIG_MEMORY_HOTPLUG |
24b6d416 CH |
3095 | void vmemmap_free(unsigned long start, unsigned long end, |
3096 | struct vmem_altmap *altmap); | |
0197518c | 3097 | #endif |
46723bfa | 3098 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, |
15670bfe | 3099 | unsigned long nr_pages); |
6a46079c | 3100 | |
82ba011b AK |
3101 | enum mf_flags { |
3102 | MF_COUNT_INCREASED = 1 << 0, | |
7329bbeb | 3103 | MF_ACTION_REQUIRED = 1 << 1, |
6751ed65 | 3104 | MF_MUST_KILL = 1 << 2, |
cf870c70 | 3105 | MF_SOFT_OFFLINE = 1 << 3, |
82ba011b | 3106 | }; |
83b57531 EB |
3107 | extern int memory_failure(unsigned long pfn, int flags); |
3108 | extern void memory_failure_queue(unsigned long pfn, int flags); | |
06202231 | 3109 | extern void memory_failure_queue_kick(int cpu); |
847ce401 | 3110 | extern int unpoison_memory(unsigned long pfn); |
6a46079c AK |
3111 | extern int sysctl_memory_failure_early_kill; |
3112 | extern int sysctl_memory_failure_recovery; | |
facb6011 | 3113 | extern void shake_page(struct page *p, int access); |
5844a486 | 3114 | extern atomic_long_t num_poisoned_pages __read_mostly; |
feec24a6 | 3115 | extern int soft_offline_page(unsigned long pfn, int flags); |
6a46079c | 3116 | |
cc637b17 XX |
3117 | |
3118 | /* | |
3119 | * Error handlers for various types of pages. | |
3120 | */ | |
cc3e2af4 | 3121 | enum mf_result { |
cc637b17 XX |
3122 | MF_IGNORED, /* Error: cannot be handled */ |
3123 | MF_FAILED, /* Error: handling failed */ | |
3124 | MF_DELAYED, /* Will be handled later */ | |
3125 | MF_RECOVERED, /* Successfully recovered */ | |
3126 | }; | |
3127 | ||
3128 | enum mf_action_page_type { | |
3129 | MF_MSG_KERNEL, | |
3130 | MF_MSG_KERNEL_HIGH_ORDER, | |
3131 | MF_MSG_SLAB, | |
3132 | MF_MSG_DIFFERENT_COMPOUND, | |
3133 | MF_MSG_POISONED_HUGE, | |
3134 | MF_MSG_HUGE, | |
3135 | MF_MSG_FREE_HUGE, | |
31286a84 | 3136 | MF_MSG_NON_PMD_HUGE, |
cc637b17 XX |
3137 | MF_MSG_UNMAP_FAILED, |
3138 | MF_MSG_DIRTY_SWAPCACHE, | |
3139 | MF_MSG_CLEAN_SWAPCACHE, | |
3140 | MF_MSG_DIRTY_MLOCKED_LRU, | |
3141 | MF_MSG_CLEAN_MLOCKED_LRU, | |
3142 | MF_MSG_DIRTY_UNEVICTABLE_LRU, | |
3143 | MF_MSG_CLEAN_UNEVICTABLE_LRU, | |
3144 | MF_MSG_DIRTY_LRU, | |
3145 | MF_MSG_CLEAN_LRU, | |
3146 | MF_MSG_TRUNCATED_LRU, | |
3147 | MF_MSG_BUDDY, | |
3148 | MF_MSG_BUDDY_2ND, | |
6100e34b | 3149 | MF_MSG_DAX, |
5d1fd5dc | 3150 | MF_MSG_UNSPLIT_THP, |
cc637b17 XX |
3151 | MF_MSG_UNKNOWN, |
3152 | }; | |
3153 | ||
47ad8475 AA |
3154 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
3155 | extern void clear_huge_page(struct page *page, | |
c79b57e4 | 3156 | unsigned long addr_hint, |
47ad8475 AA |
3157 | unsigned int pages_per_huge_page); |
3158 | extern void copy_user_huge_page(struct page *dst, struct page *src, | |
c9f4cd71 HY |
3159 | unsigned long addr_hint, |
3160 | struct vm_area_struct *vma, | |
47ad8475 | 3161 | unsigned int pages_per_huge_page); |
fa4d75c1 MK |
3162 | extern long copy_huge_page_from_user(struct page *dst_page, |
3163 | const void __user *usr_src, | |
810a56b9 MK |
3164 | unsigned int pages_per_huge_page, |
3165 | bool allow_pagefault); | |
2484ca9b THV |
3166 | |
3167 | /** | |
3168 | * vma_is_special_huge - Are transhuge page-table entries considered special? | |
3169 | * @vma: Pointer to the struct vm_area_struct to consider | |
3170 | * | |
3171 | * Whether transhuge page-table entries are considered "special" following | |
3172 | * the definition in vm_normal_page(). | |
3173 | * | |
3174 | * Return: true if transhuge page-table entries should be considered special, | |
3175 | * false otherwise. | |
3176 | */ | |
3177 | static inline bool vma_is_special_huge(const struct vm_area_struct *vma) | |
3178 | { | |
3179 | return vma_is_dax(vma) || (vma->vm_file && | |
3180 | (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))); | |
3181 | } | |
3182 | ||
47ad8475 AA |
3183 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ |
3184 | ||
c0a32fc5 SG |
3185 | #ifdef CONFIG_DEBUG_PAGEALLOC |
3186 | extern unsigned int _debug_guardpage_minorder; | |
96a2b03f | 3187 | DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled); |
c0a32fc5 SG |
3188 | |
3189 | static inline unsigned int debug_guardpage_minorder(void) | |
3190 | { | |
3191 | return _debug_guardpage_minorder; | |
3192 | } | |
3193 | ||
e30825f1 JK |
3194 | static inline bool debug_guardpage_enabled(void) |
3195 | { | |
96a2b03f | 3196 | return static_branch_unlikely(&_debug_guardpage_enabled); |
e30825f1 JK |
3197 | } |
3198 | ||
c0a32fc5 SG |
3199 | static inline bool page_is_guard(struct page *page) |
3200 | { | |
e30825f1 JK |
3201 | if (!debug_guardpage_enabled()) |
3202 | return false; | |
3203 | ||
3972f6bb | 3204 | return PageGuard(page); |
c0a32fc5 SG |
3205 | } |
3206 | #else | |
3207 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } | |
e30825f1 | 3208 | static inline bool debug_guardpage_enabled(void) { return false; } |
c0a32fc5 SG |
3209 | static inline bool page_is_guard(struct page *page) { return false; } |
3210 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | |
3211 | ||
f9872caf CS |
3212 | #if MAX_NUMNODES > 1 |
3213 | void __init setup_nr_node_ids(void); | |
3214 | #else | |
3215 | static inline void setup_nr_node_ids(void) {} | |
3216 | #endif | |
3217 | ||
010c164a SL |
3218 | extern int memcmp_pages(struct page *page1, struct page *page2); |
3219 | ||
3220 | static inline int pages_identical(struct page *page1, struct page *page2) | |
3221 | { | |
3222 | return !memcmp_pages(page1, page2); | |
3223 | } | |
3224 | ||
c5acad84 TH |
3225 | #ifdef CONFIG_MAPPING_DIRTY_HELPERS |
3226 | unsigned long clean_record_shared_mapping_range(struct address_space *mapping, | |
3227 | pgoff_t first_index, pgoff_t nr, | |
3228 | pgoff_t bitmap_pgoff, | |
3229 | unsigned long *bitmap, | |
3230 | pgoff_t *start, | |
3231 | pgoff_t *end); | |
3232 | ||
3233 | unsigned long wp_shared_mapping_range(struct address_space *mapping, | |
3234 | pgoff_t first_index, pgoff_t nr); | |
3235 | #endif | |
3236 | ||
2374c09b CH |
3237 | extern int sysctl_nr_trim_pages; |
3238 | ||
5bb1bb35 | 3239 | #ifdef CONFIG_PRINTK |
8e7f37f2 | 3240 | void mem_dump_obj(void *object); |
5bb1bb35 PM |
3241 | #else |
3242 | static inline void mem_dump_obj(void *object) {} | |
3243 | #endif | |
8e7f37f2 | 3244 | |
22247efd PX |
3245 | /** |
3246 | * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it | |
3247 | * @seals: the seals to check | |
3248 | * @vma: the vma to operate on | |
3249 | * | |
3250 | * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on | |
3251 | * the vma flags. Return 0 if check pass, or <0 for errors. | |
3252 | */ | |
3253 | static inline int seal_check_future_write(int seals, struct vm_area_struct *vma) | |
3254 | { | |
3255 | if (seals & F_SEAL_FUTURE_WRITE) { | |
3256 | /* | |
3257 | * New PROT_WRITE and MAP_SHARED mmaps are not allowed when | |
3258 | * "future write" seal active. | |
3259 | */ | |
3260 | if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE)) | |
3261 | return -EPERM; | |
3262 | ||
3263 | /* | |
3264 | * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as | |
3265 | * MAP_SHARED and read-only, take care to not allow mprotect to | |
3266 | * revert protections on such mappings. Do this only for shared | |
3267 | * mappings. For private mappings, don't need to mask | |
3268 | * VM_MAYWRITE as we still want them to be COW-writable. | |
3269 | */ | |
3270 | if (vma->vm_flags & VM_SHARED) | |
3271 | vma->vm_flags &= ~(VM_MAYWRITE); | |
3272 | } | |
3273 | ||
3274 | return 0; | |
3275 | } | |
3276 | ||
1da177e4 LT |
3277 | #endif /* __KERNEL__ */ |
3278 | #endif /* _LINUX_MM_H */ |