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 | 5 | #include <linux/errno.h> |
309381fe | 6 | #include <linux/mmdebug.h> |
1da177e4 | 7 | #include <linux/gfp.h> |
187f1882 | 8 | #include <linux/bug.h> |
1da177e4 LT |
9 | #include <linux/list.h> |
10 | #include <linux/mmzone.h> | |
11 | #include <linux/rbtree.h> | |
83aeeada | 12 | #include <linux/atomic.h> |
9a11b49a | 13 | #include <linux/debug_locks.h> |
5b99cd0e | 14 | #include <linux/mm_types.h> |
9740ca4e | 15 | #include <linux/mmap_lock.h> |
08677214 | 16 | #include <linux/range.h> |
c6f6b596 | 17 | #include <linux/pfn.h> |
3565fce3 | 18 | #include <linux/percpu-refcount.h> |
e9da73d6 | 19 | #include <linux/bit_spinlock.h> |
b0d40c92 | 20 | #include <linux/shrinker.h> |
9c599024 | 21 | #include <linux/resource.h> |
e30825f1 | 22 | #include <linux/page_ext.h> |
8025e5dd | 23 | #include <linux/err.h> |
41901567 | 24 | #include <linux/page-flags.h> |
fe896d18 | 25 | #include <linux/page_ref.h> |
3b3b1a29 | 26 | #include <linux/overflow.h> |
b5420237 | 27 | #include <linux/sizes.h> |
7969f226 | 28 | #include <linux/sched.h> |
65fddcfc | 29 | #include <linux/pgtable.h> |
34303244 | 30 | #include <linux/kasan.h> |
f25cbb7a | 31 | #include <linux/memremap.h> |
ef6a22b7 | 32 | #include <linux/slab.h> |
1da177e4 LT |
33 | |
34 | struct mempolicy; | |
35 | struct anon_vma; | |
bf181b9f | 36 | struct anon_vma_chain; |
e8edc6e0 | 37 | struct user_struct; |
bce617ed | 38 | struct pt_regs; |
1da177e4 | 39 | |
5ef64cc8 LT |
40 | extern int sysctl_page_lock_unfairness; |
41 | ||
b7ec1bf3 | 42 | void mm_core_init(void); |
597b7305 MH |
43 | void init_mm_internals(void); |
44 | ||
a9ee6cf5 | 45 | #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */ |
1da177e4 | 46 | extern unsigned long max_mapnr; |
fccc9987 JL |
47 | |
48 | static inline void set_max_mapnr(unsigned long limit) | |
49 | { | |
50 | max_mapnr = limit; | |
51 | } | |
52 | #else | |
53 | static inline void set_max_mapnr(unsigned long limit) { } | |
1da177e4 LT |
54 | #endif |
55 | ||
ca79b0c2 AK |
56 | extern atomic_long_t _totalram_pages; |
57 | static inline unsigned long totalram_pages(void) | |
58 | { | |
59 | return (unsigned long)atomic_long_read(&_totalram_pages); | |
60 | } | |
61 | ||
62 | static inline void totalram_pages_inc(void) | |
63 | { | |
64 | atomic_long_inc(&_totalram_pages); | |
65 | } | |
66 | ||
67 | static inline void totalram_pages_dec(void) | |
68 | { | |
69 | atomic_long_dec(&_totalram_pages); | |
70 | } | |
71 | ||
72 | static inline void totalram_pages_add(long count) | |
73 | { | |
74 | atomic_long_add(count, &_totalram_pages); | |
75 | } | |
76 | ||
1da177e4 | 77 | extern void * high_memory; |
1da177e4 | 78 | extern int page_cluster; |
ea0ffd0c | 79 | extern const int page_cluster_max; |
1da177e4 LT |
80 | |
81 | #ifdef CONFIG_SYSCTL | |
82 | extern int sysctl_legacy_va_layout; | |
83 | #else | |
84 | #define sysctl_legacy_va_layout 0 | |
85 | #endif | |
86 | ||
d07e2259 DC |
87 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS |
88 | extern const int mmap_rnd_bits_min; | |
89 | extern const int mmap_rnd_bits_max; | |
90 | extern int mmap_rnd_bits __read_mostly; | |
91 | #endif | |
92 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS | |
93 | extern const int mmap_rnd_compat_bits_min; | |
94 | extern const int mmap_rnd_compat_bits_max; | |
95 | extern int mmap_rnd_compat_bits __read_mostly; | |
96 | #endif | |
97 | ||
1da177e4 | 98 | #include <asm/page.h> |
1da177e4 | 99 | #include <asm/processor.h> |
1da177e4 | 100 | |
79442ed1 TC |
101 | #ifndef __pa_symbol |
102 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) | |
103 | #endif | |
104 | ||
1dff8083 AB |
105 | #ifndef page_to_virt |
106 | #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x))) | |
107 | #endif | |
108 | ||
568c5fe5 LA |
109 | #ifndef lm_alias |
110 | #define lm_alias(x) __va(__pa_symbol(x)) | |
111 | #endif | |
112 | ||
593befa6 DD |
113 | /* |
114 | * To prevent common memory management code establishing | |
115 | * a zero page mapping on a read fault. | |
116 | * This macro should be defined within <asm/pgtable.h>. | |
117 | * s390 does this to prevent multiplexing of hardware bits | |
118 | * related to the physical page in case of virtualization. | |
119 | */ | |
120 | #ifndef mm_forbids_zeropage | |
121 | #define mm_forbids_zeropage(X) (0) | |
122 | #endif | |
123 | ||
a4a3ede2 PT |
124 | /* |
125 | * On some architectures it is expensive to call memset() for small sizes. | |
5470dea4 AD |
126 | * If an architecture decides to implement their own version of |
127 | * mm_zero_struct_page they should wrap the defines below in a #ifndef and | |
128 | * define their own version of this macro in <asm/pgtable.h> | |
a4a3ede2 | 129 | */ |
5470dea4 | 130 | #if BITS_PER_LONG == 64 |
3770e52f | 131 | /* This function must be updated when the size of struct page grows above 96 |
5470dea4 AD |
132 | * or reduces below 56. The idea that compiler optimizes out switch() |
133 | * statement, and only leaves move/store instructions. Also the compiler can | |
c4ffefd1 | 134 | * combine write statements if they are both assignments and can be reordered, |
5470dea4 AD |
135 | * this can result in several of the writes here being dropped. |
136 | */ | |
137 | #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp) | |
138 | static inline void __mm_zero_struct_page(struct page *page) | |
139 | { | |
140 | unsigned long *_pp = (void *)page; | |
141 | ||
3770e52f | 142 | /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */ |
5470dea4 AD |
143 | BUILD_BUG_ON(sizeof(struct page) & 7); |
144 | BUILD_BUG_ON(sizeof(struct page) < 56); | |
3770e52f | 145 | BUILD_BUG_ON(sizeof(struct page) > 96); |
5470dea4 AD |
146 | |
147 | switch (sizeof(struct page)) { | |
3770e52f AB |
148 | case 96: |
149 | _pp[11] = 0; | |
150 | fallthrough; | |
151 | case 88: | |
152 | _pp[10] = 0; | |
153 | fallthrough; | |
5470dea4 | 154 | case 80: |
df561f66 GS |
155 | _pp[9] = 0; |
156 | fallthrough; | |
5470dea4 | 157 | case 72: |
df561f66 GS |
158 | _pp[8] = 0; |
159 | fallthrough; | |
5470dea4 | 160 | case 64: |
df561f66 GS |
161 | _pp[7] = 0; |
162 | fallthrough; | |
5470dea4 AD |
163 | case 56: |
164 | _pp[6] = 0; | |
165 | _pp[5] = 0; | |
166 | _pp[4] = 0; | |
167 | _pp[3] = 0; | |
168 | _pp[2] = 0; | |
169 | _pp[1] = 0; | |
170 | _pp[0] = 0; | |
171 | } | |
172 | } | |
173 | #else | |
a4a3ede2 PT |
174 | #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page))) |
175 | #endif | |
176 | ||
ea606cf5 AR |
177 | /* |
178 | * Default maximum number of active map areas, this limits the number of vmas | |
179 | * per mm struct. Users can overwrite this number by sysctl but there is a | |
180 | * problem. | |
181 | * | |
182 | * When a program's coredump is generated as ELF format, a section is created | |
183 | * per a vma. In ELF, the number of sections is represented in unsigned short. | |
184 | * This means the number of sections should be smaller than 65535 at coredump. | |
185 | * Because the kernel adds some informative sections to a image of program at | |
186 | * generating coredump, we need some margin. The number of extra sections is | |
187 | * 1-3 now and depends on arch. We use "5" as safe margin, here. | |
188 | * | |
189 | * ELF extended numbering allows more than 65535 sections, so 16-bit bound is | |
190 | * not a hard limit any more. Although some userspace tools can be surprised by | |
191 | * that. | |
192 | */ | |
193 | #define MAPCOUNT_ELF_CORE_MARGIN (5) | |
194 | #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) | |
195 | ||
196 | extern int sysctl_max_map_count; | |
197 | ||
c9b1d098 | 198 | extern unsigned long sysctl_user_reserve_kbytes; |
4eeab4f5 | 199 | extern unsigned long sysctl_admin_reserve_kbytes; |
c9b1d098 | 200 | |
49f0ce5f JM |
201 | extern int sysctl_overcommit_memory; |
202 | extern int sysctl_overcommit_ratio; | |
203 | extern unsigned long sysctl_overcommit_kbytes; | |
204 | ||
32927393 CH |
205 | int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *, |
206 | loff_t *); | |
207 | int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *, | |
208 | loff_t *); | |
56f3547b FT |
209 | int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *, |
210 | loff_t *); | |
49f0ce5f | 211 | |
1cfcee72 | 212 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
1da177e4 | 213 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
659508f9 | 214 | #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio)) |
1cfcee72 MWO |
215 | #else |
216 | #define nth_page(page,n) ((page) + (n)) | |
659508f9 | 217 | #define folio_page_idx(folio, p) ((p) - &(folio)->page) |
1cfcee72 | 218 | #endif |
1da177e4 | 219 | |
27ac792c AR |
220 | /* to align the pointer to the (next) page boundary */ |
221 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | |
222 | ||
335e52c2 DG |
223 | /* to align the pointer to the (prev) page boundary */ |
224 | #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE) | |
225 | ||
0fa73b86 | 226 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ |
1061b0d2 | 227 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE) |
0fa73b86 | 228 | |
f86196ea | 229 | #define lru_to_page(head) (list_entry((head)->prev, struct page, lru)) |
06d20bdb MWO |
230 | static inline struct folio *lru_to_folio(struct list_head *head) |
231 | { | |
232 | return list_entry((head)->prev, struct folio, lru); | |
233 | } | |
f86196ea | 234 | |
5748fbc5 KW |
235 | void setup_initial_init_mm(void *start_code, void *end_code, |
236 | void *end_data, void *brk); | |
237 | ||
1da177e4 LT |
238 | /* |
239 | * Linux kernel virtual memory manager primitives. | |
240 | * The idea being to have a "virtual" mm in the same way | |
241 | * we have a virtual fs - giving a cleaner interface to the | |
242 | * mm details, and allowing different kinds of memory mappings | |
243 | * (from shared memory to executable loading to arbitrary | |
244 | * mmap() functions). | |
245 | */ | |
246 | ||
490fc053 | 247 | struct vm_area_struct *vm_area_alloc(struct mm_struct *); |
3928d4f5 LT |
248 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *); |
249 | void vm_area_free(struct vm_area_struct *); | |
0d2ebf9c SB |
250 | /* Use only if VMA has no other users */ |
251 | void __vm_area_free(struct vm_area_struct *vma); | |
c43692e8 | 252 | |
1da177e4 | 253 | #ifndef CONFIG_MMU |
8feae131 DH |
254 | extern struct rb_root nommu_region_tree; |
255 | extern struct rw_semaphore nommu_region_sem; | |
1da177e4 LT |
256 | |
257 | extern unsigned int kobjsize(const void *objp); | |
258 | #endif | |
259 | ||
260 | /* | |
605d9288 | 261 | * vm_flags in vm_area_struct, see mm_types.h. |
bcf66917 | 262 | * When changing, update also include/trace/events/mmflags.h |
1da177e4 | 263 | */ |
cc2383ec KK |
264 | #define VM_NONE 0x00000000 |
265 | ||
1da177e4 LT |
266 | #define VM_READ 0x00000001 /* currently active flags */ |
267 | #define VM_WRITE 0x00000002 | |
268 | #define VM_EXEC 0x00000004 | |
269 | #define VM_SHARED 0x00000008 | |
270 | ||
7e2cff42 | 271 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
1da177e4 LT |
272 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
273 | #define VM_MAYWRITE 0x00000020 | |
274 | #define VM_MAYEXEC 0x00000040 | |
275 | #define VM_MAYSHARE 0x00000080 | |
276 | ||
277 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
b6b7a8fa | 278 | #ifdef CONFIG_MMU |
16ba6f81 | 279 | #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */ |
b6b7a8fa DH |
280 | #else /* CONFIG_MMU */ |
281 | #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */ | |
282 | #define VM_UFFD_MISSING 0 | |
283 | #endif /* CONFIG_MMU */ | |
6aab341e | 284 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
16ba6f81 | 285 | #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */ |
1da177e4 | 286 | |
1da177e4 LT |
287 | #define VM_LOCKED 0x00002000 |
288 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | |
289 | ||
290 | /* Used by sys_madvise() */ | |
291 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | |
292 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | |
293 | ||
294 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | |
295 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | |
de60f5f1 | 296 | #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */ |
1da177e4 | 297 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
cdfd4325 | 298 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
1da177e4 | 299 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
b6fb293f | 300 | #define VM_SYNC 0x00800000 /* Synchronous page faults */ |
cc2383ec | 301 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ |
d2cd9ede | 302 | #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */ |
0103bd16 | 303 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ |
d00806b1 | 304 | |
d9104d1c CG |
305 | #ifdef CONFIG_MEM_SOFT_DIRTY |
306 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ | |
307 | #else | |
308 | # define VM_SOFTDIRTY 0 | |
309 | #endif | |
310 | ||
b379d790 | 311 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
cc2383ec KK |
312 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ |
313 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ | |
f8af4da3 | 314 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
1da177e4 | 315 | |
63c17fb8 DH |
316 | #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS |
317 | #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */ | |
318 | #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */ | |
319 | #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */ | |
320 | #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */ | |
df3735c5 | 321 | #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */ |
63c17fb8 DH |
322 | #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0) |
323 | #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1) | |
324 | #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2) | |
325 | #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3) | |
df3735c5 | 326 | #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4) |
63c17fb8 DH |
327 | #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */ |
328 | ||
5212213a | 329 | #ifdef CONFIG_ARCH_HAS_PKEYS |
8f62c883 DH |
330 | # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0 |
331 | # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */ | |
2c9e0a6f | 332 | # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */ |
8f62c883 DH |
333 | # define VM_PKEY_BIT2 VM_HIGH_ARCH_2 |
334 | # define VM_PKEY_BIT3 VM_HIGH_ARCH_3 | |
2c9e0a6f RP |
335 | #ifdef CONFIG_PPC |
336 | # define VM_PKEY_BIT4 VM_HIGH_ARCH_4 | |
337 | #else | |
338 | # define VM_PKEY_BIT4 0 | |
8f62c883 | 339 | #endif |
5212213a RP |
340 | #endif /* CONFIG_ARCH_HAS_PKEYS */ |
341 | ||
342 | #if defined(CONFIG_X86) | |
343 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ | |
12564485 SA |
344 | #elif defined(CONFIG_PPC) |
345 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ | |
cc2383ec KK |
346 | #elif defined(CONFIG_PARISC) |
347 | # define VM_GROWSUP VM_ARCH_1 | |
348 | #elif defined(CONFIG_IA64) | |
349 | # define VM_GROWSUP VM_ARCH_1 | |
74a04967 KA |
350 | #elif defined(CONFIG_SPARC64) |
351 | # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */ | |
352 | # define VM_ARCH_CLEAR VM_SPARC_ADI | |
8ef8f360 DM |
353 | #elif defined(CONFIG_ARM64) |
354 | # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */ | |
355 | # define VM_ARCH_CLEAR VM_ARM64_BTI | |
cc2383ec KK |
356 | #elif !defined(CONFIG_MMU) |
357 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ | |
358 | #endif | |
359 | ||
9f341931 CM |
360 | #if defined(CONFIG_ARM64_MTE) |
361 | # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */ | |
362 | # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */ | |
363 | #else | |
364 | # define VM_MTE VM_NONE | |
365 | # define VM_MTE_ALLOWED VM_NONE | |
366 | #endif | |
367 | ||
cc2383ec KK |
368 | #ifndef VM_GROWSUP |
369 | # define VM_GROWSUP VM_NONE | |
370 | #endif | |
371 | ||
7677f7fd AR |
372 | #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR |
373 | # define VM_UFFD_MINOR_BIT 37 | |
374 | # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */ | |
375 | #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */ | |
376 | # define VM_UFFD_MINOR VM_NONE | |
377 | #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */ | |
378 | ||
a8bef8ff MG |
379 | /* Bits set in the VMA until the stack is in its final location */ |
380 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | |
381 | ||
c62da0c3 AK |
382 | #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0) |
383 | ||
384 | /* Common data flag combinations */ | |
385 | #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \ | |
386 | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) | |
387 | #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \ | |
388 | VM_MAYWRITE | VM_MAYEXEC) | |
389 | #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \ | |
390 | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) | |
391 | ||
392 | #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */ | |
393 | #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC | |
394 | #endif | |
395 | ||
1da177e4 LT |
396 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
397 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
398 | #endif | |
399 | ||
400 | #ifdef CONFIG_STACK_GROWSUP | |
30bdbb78 | 401 | #define VM_STACK VM_GROWSUP |
1da177e4 | 402 | #else |
30bdbb78 | 403 | #define VM_STACK VM_GROWSDOWN |
1da177e4 LT |
404 | #endif |
405 | ||
30bdbb78 KK |
406 | #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
407 | ||
6cb4d9a2 AK |
408 | /* VMA basic access permission flags */ |
409 | #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC) | |
410 | ||
411 | ||
b291f000 | 412 | /* |
78f11a25 | 413 | * Special vmas that are non-mergable, non-mlock()able. |
b291f000 | 414 | */ |
9050d7eb | 415 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) |
b291f000 | 416 | |
b4443772 AK |
417 | /* This mask prevents VMA from being scanned with khugepaged */ |
418 | #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB) | |
419 | ||
a0715cc2 AT |
420 | /* This mask defines which mm->def_flags a process can inherit its parent */ |
421 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE | |
422 | ||
e430a95a SB |
423 | /* This mask represents all the VMA flag bits used by mlock */ |
424 | #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT) | |
de60f5f1 | 425 | |
2c2d57b5 KA |
426 | /* Arch-specific flags to clear when updating VM flags on protection change */ |
427 | #ifndef VM_ARCH_CLEAR | |
428 | # define VM_ARCH_CLEAR VM_NONE | |
429 | #endif | |
430 | #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR) | |
431 | ||
1da177e4 LT |
432 | /* |
433 | * mapping from the currently active vm_flags protection bits (the | |
434 | * low four bits) to a page protection mask.. | |
435 | */ | |
1da177e4 | 436 | |
dde16072 PX |
437 | /* |
438 | * The default fault flags that should be used by most of the | |
439 | * arch-specific page fault handlers. | |
440 | */ | |
441 | #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \ | |
c270a7ee PX |
442 | FAULT_FLAG_KILLABLE | \ |
443 | FAULT_FLAG_INTERRUPTIBLE) | |
dde16072 | 444 | |
4064b982 PX |
445 | /** |
446 | * fault_flag_allow_retry_first - check ALLOW_RETRY the first time | |
78f4841e | 447 | * @flags: Fault flags. |
4064b982 PX |
448 | * |
449 | * This is mostly used for places where we want to try to avoid taking | |
c1e8d7c6 | 450 | * the mmap_lock for too long a time when waiting for another condition |
4064b982 | 451 | * to change, in which case we can try to be polite to release the |
c1e8d7c6 ML |
452 | * mmap_lock in the first round to avoid potential starvation of other |
453 | * processes that would also want the mmap_lock. | |
4064b982 PX |
454 | * |
455 | * Return: true if the page fault allows retry and this is the first | |
456 | * attempt of the fault handling; false otherwise. | |
457 | */ | |
da2f5eb3 | 458 | static inline bool fault_flag_allow_retry_first(enum fault_flag flags) |
4064b982 PX |
459 | { |
460 | return (flags & FAULT_FLAG_ALLOW_RETRY) && | |
461 | (!(flags & FAULT_FLAG_TRIED)); | |
462 | } | |
463 | ||
282a8e03 RZ |
464 | #define FAULT_FLAG_TRACE \ |
465 | { FAULT_FLAG_WRITE, "WRITE" }, \ | |
466 | { FAULT_FLAG_MKWRITE, "MKWRITE" }, \ | |
467 | { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \ | |
468 | { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \ | |
469 | { FAULT_FLAG_KILLABLE, "KILLABLE" }, \ | |
470 | { FAULT_FLAG_TRIED, "TRIED" }, \ | |
471 | { FAULT_FLAG_USER, "USER" }, \ | |
472 | { FAULT_FLAG_REMOTE, "REMOTE" }, \ | |
c270a7ee | 473 | { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \ |
55324e46 SB |
474 | { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \ |
475 | { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" } | |
282a8e03 | 476 | |
54cb8821 | 477 | /* |
11192337 | 478 | * vm_fault is filled by the pagefault handler and passed to the vma's |
83c54070 NP |
479 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
480 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
54cb8821 | 481 | * |
c20cd45e MH |
482 | * MM layer fills up gfp_mask for page allocations but fault handler might |
483 | * alter it if its implementation requires a different allocation context. | |
484 | * | |
9b4bdd2f | 485 | * pgoff should be used in favour of virtual_address, if possible. |
54cb8821 | 486 | */ |
d0217ac0 | 487 | struct vm_fault { |
5857c920 | 488 | const struct { |
742d3372 WD |
489 | struct vm_area_struct *vma; /* Target VMA */ |
490 | gfp_t gfp_mask; /* gfp mask to be used for allocations */ | |
491 | pgoff_t pgoff; /* Logical page offset based on vma */ | |
824ddc60 NA |
492 | unsigned long address; /* Faulting virtual address - masked */ |
493 | unsigned long real_address; /* Faulting virtual address - unmasked */ | |
742d3372 | 494 | }; |
da2f5eb3 | 495 | enum fault_flag flags; /* FAULT_FLAG_xxx flags |
742d3372 | 496 | * XXX: should really be 'const' */ |
82b0f8c3 | 497 | pmd_t *pmd; /* Pointer to pmd entry matching |
2994302b | 498 | * the 'address' */ |
a2d58167 DJ |
499 | pud_t *pud; /* Pointer to pud entry matching |
500 | * the 'address' | |
501 | */ | |
5db4f15c YS |
502 | union { |
503 | pte_t orig_pte; /* Value of PTE at the time of fault */ | |
504 | pmd_t orig_pmd; /* Value of PMD at the time of fault, | |
505 | * used by PMD fault only. | |
506 | */ | |
507 | }; | |
d0217ac0 | 508 | |
3917048d | 509 | struct page *cow_page; /* Page handler may use for COW fault */ |
d0217ac0 | 510 | struct page *page; /* ->fault handlers should return a |
83c54070 | 511 | * page here, unless VM_FAULT_NOPAGE |
d0217ac0 | 512 | * is set (which is also implied by |
83c54070 | 513 | * VM_FAULT_ERROR). |
d0217ac0 | 514 | */ |
82b0f8c3 | 515 | /* These three entries are valid only while holding ptl lock */ |
bae473a4 KS |
516 | pte_t *pte; /* Pointer to pte entry matching |
517 | * the 'address'. NULL if the page | |
518 | * table hasn't been allocated. | |
519 | */ | |
520 | spinlock_t *ptl; /* Page table lock. | |
521 | * Protects pte page table if 'pte' | |
522 | * is not NULL, otherwise pmd. | |
523 | */ | |
7267ec00 | 524 | pgtable_t prealloc_pte; /* Pre-allocated pte page table. |
f9ce0be7 KS |
525 | * vm_ops->map_pages() sets up a page |
526 | * table from atomic context. | |
7267ec00 KS |
527 | * do_fault_around() pre-allocates |
528 | * page table to avoid allocation from | |
529 | * atomic context. | |
530 | */ | |
54cb8821 | 531 | }; |
1da177e4 | 532 | |
c791ace1 DJ |
533 | /* page entry size for vm->huge_fault() */ |
534 | enum page_entry_size { | |
535 | PE_SIZE_PTE = 0, | |
536 | PE_SIZE_PMD, | |
537 | PE_SIZE_PUD, | |
538 | }; | |
539 | ||
1da177e4 LT |
540 | /* |
541 | * These are the virtual MM functions - opening of an area, closing and | |
542 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
27d036e3 | 543 | * to the functions called when a no-page or a wp-page exception occurs. |
1da177e4 LT |
544 | */ |
545 | struct vm_operations_struct { | |
546 | void (*open)(struct vm_area_struct * area); | |
cc6dcfee SB |
547 | /** |
548 | * @close: Called when the VMA is being removed from the MM. | |
549 | * Context: User context. May sleep. Caller holds mmap_lock. | |
550 | */ | |
1da177e4 | 551 | void (*close)(struct vm_area_struct * area); |
dd3b614f DS |
552 | /* Called any time before splitting to check if it's allowed */ |
553 | int (*may_split)(struct vm_area_struct *area, unsigned long addr); | |
14d07113 | 554 | int (*mremap)(struct vm_area_struct *area); |
95bb7c42 SC |
555 | /* |
556 | * Called by mprotect() to make driver-specific permission | |
557 | * checks before mprotect() is finalised. The VMA must not | |
3e0ee843 | 558 | * be modified. Returns 0 if mprotect() can proceed. |
95bb7c42 SC |
559 | */ |
560 | int (*mprotect)(struct vm_area_struct *vma, unsigned long start, | |
561 | unsigned long end, unsigned long newflags); | |
1c8f4220 SJ |
562 | vm_fault_t (*fault)(struct vm_fault *vmf); |
563 | vm_fault_t (*huge_fault)(struct vm_fault *vmf, | |
564 | enum page_entry_size pe_size); | |
f9ce0be7 | 565 | vm_fault_t (*map_pages)(struct vm_fault *vmf, |
bae473a4 | 566 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
05ea8860 | 567 | unsigned long (*pagesize)(struct vm_area_struct * area); |
9637a5ef DH |
568 | |
569 | /* notification that a previously read-only page is about to become | |
570 | * writable, if an error is returned it will cause a SIGBUS */ | |
1c8f4220 | 571 | vm_fault_t (*page_mkwrite)(struct vm_fault *vmf); |
28b2ee20 | 572 | |
dd906184 | 573 | /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ |
1c8f4220 | 574 | vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf); |
dd906184 | 575 | |
28b2ee20 | 576 | /* called by access_process_vm when get_user_pages() fails, typically |
96667f8a DV |
577 | * for use by special VMAs. See also generic_access_phys() for a generic |
578 | * implementation useful for any iomem mapping. | |
28b2ee20 RR |
579 | */ |
580 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | |
581 | void *buf, int len, int write); | |
78d683e8 AL |
582 | |
583 | /* Called by the /proc/PID/maps code to ask the vma whether it | |
584 | * has a special name. Returning non-NULL will also cause this | |
585 | * vma to be dumped unconditionally. */ | |
586 | const char *(*name)(struct vm_area_struct *vma); | |
587 | ||
1da177e4 | 588 | #ifdef CONFIG_NUMA |
a6020ed7 LS |
589 | /* |
590 | * set_policy() op must add a reference to any non-NULL @new mempolicy | |
591 | * to hold the policy upon return. Caller should pass NULL @new to | |
592 | * remove a policy and fall back to surrounding context--i.e. do not | |
593 | * install a MPOL_DEFAULT policy, nor the task or system default | |
594 | * mempolicy. | |
595 | */ | |
1da177e4 | 596 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
a6020ed7 LS |
597 | |
598 | /* | |
599 | * get_policy() op must add reference [mpol_get()] to any policy at | |
600 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | |
601 | * in mm/mempolicy.c will do this automatically. | |
602 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | |
c1e8d7c6 | 603 | * marked as MPOL_SHARED. vma policies are protected by the mmap_lock. |
a6020ed7 LS |
604 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op |
605 | * must return NULL--i.e., do not "fallback" to task or system default | |
606 | * policy. | |
607 | */ | |
1da177e4 LT |
608 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
609 | unsigned long addr); | |
610 | #endif | |
667a0a06 DV |
611 | /* |
612 | * Called by vm_normal_page() for special PTEs to find the | |
613 | * page for @addr. This is useful if the default behavior | |
614 | * (using pte_page()) would not find the correct page. | |
615 | */ | |
616 | struct page *(*find_special_page)(struct vm_area_struct *vma, | |
617 | unsigned long addr); | |
1da177e4 LT |
618 | }; |
619 | ||
ef6a22b7 MG |
620 | #ifdef CONFIG_NUMA_BALANCING |
621 | static inline void vma_numab_state_init(struct vm_area_struct *vma) | |
622 | { | |
623 | vma->numab_state = NULL; | |
624 | } | |
625 | static inline void vma_numab_state_free(struct vm_area_struct *vma) | |
626 | { | |
627 | kfree(vma->numab_state); | |
628 | } | |
629 | #else | |
630 | static inline void vma_numab_state_init(struct vm_area_struct *vma) {} | |
631 | static inline void vma_numab_state_free(struct vm_area_struct *vma) {} | |
632 | #endif /* CONFIG_NUMA_BALANCING */ | |
633 | ||
5e31275c | 634 | #ifdef CONFIG_PER_VMA_LOCK |
5e31275c SB |
635 | /* |
636 | * Try to read-lock a vma. The function is allowed to occasionally yield false | |
637 | * locked result to avoid performance overhead, in which case we fall back to | |
638 | * using mmap_lock. The function should never yield false unlocked result. | |
639 | */ | |
640 | static inline bool vma_start_read(struct vm_area_struct *vma) | |
641 | { | |
642 | /* Check before locking. A race might cause false locked result. */ | |
643 | if (vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq)) | |
644 | return false; | |
645 | ||
c7f8f31c | 646 | if (unlikely(down_read_trylock(&vma->vm_lock->lock) == 0)) |
5e31275c SB |
647 | return false; |
648 | ||
649 | /* | |
650 | * Overflow might produce false locked result. | |
651 | * False unlocked result is impossible because we modify and check | |
c7f8f31c | 652 | * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq |
5e31275c SB |
653 | * modification invalidates all existing locks. |
654 | */ | |
655 | if (unlikely(vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))) { | |
c7f8f31c | 656 | up_read(&vma->vm_lock->lock); |
5e31275c SB |
657 | return false; |
658 | } | |
659 | return true; | |
660 | } | |
661 | ||
662 | static inline void vma_end_read(struct vm_area_struct *vma) | |
663 | { | |
664 | rcu_read_lock(); /* keeps vma alive till the end of up_read */ | |
c7f8f31c | 665 | up_read(&vma->vm_lock->lock); |
5e31275c SB |
666 | rcu_read_unlock(); |
667 | } | |
668 | ||
55fd6fcc | 669 | static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq) |
5e31275c | 670 | { |
5e31275c SB |
671 | mmap_assert_write_locked(vma->vm_mm); |
672 | ||
673 | /* | |
674 | * current task is holding mmap_write_lock, both vma->vm_lock_seq and | |
675 | * mm->mm_lock_seq can't be concurrently modified. | |
676 | */ | |
55fd6fcc SB |
677 | *mm_lock_seq = READ_ONCE(vma->vm_mm->mm_lock_seq); |
678 | return (vma->vm_lock_seq == *mm_lock_seq); | |
679 | } | |
680 | ||
681 | static inline void vma_start_write(struct vm_area_struct *vma) | |
682 | { | |
683 | int mm_lock_seq; | |
684 | ||
685 | if (__is_vma_write_locked(vma, &mm_lock_seq)) | |
5e31275c SB |
686 | return; |
687 | ||
c7f8f31c | 688 | down_write(&vma->vm_lock->lock); |
5e31275c | 689 | vma->vm_lock_seq = mm_lock_seq; |
c7f8f31c | 690 | up_write(&vma->vm_lock->lock); |
5e31275c SB |
691 | } |
692 | ||
55fd6fcc SB |
693 | static inline bool vma_try_start_write(struct vm_area_struct *vma) |
694 | { | |
695 | int mm_lock_seq; | |
696 | ||
697 | if (__is_vma_write_locked(vma, &mm_lock_seq)) | |
698 | return true; | |
699 | ||
700 | if (!down_write_trylock(&vma->vm_lock->lock)) | |
701 | return false; | |
702 | ||
703 | vma->vm_lock_seq = mm_lock_seq; | |
704 | up_write(&vma->vm_lock->lock); | |
705 | return true; | |
706 | } | |
707 | ||
5e31275c SB |
708 | static inline void vma_assert_write_locked(struct vm_area_struct *vma) |
709 | { | |
55fd6fcc SB |
710 | int mm_lock_seq; |
711 | ||
712 | VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma); | |
5e31275c SB |
713 | } |
714 | ||
457f67be SB |
715 | static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached) |
716 | { | |
717 | /* When detaching vma should be write-locked */ | |
718 | if (detached) | |
719 | vma_assert_write_locked(vma); | |
720 | vma->detached = detached; | |
721 | } | |
722 | ||
50ee3253 SB |
723 | struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm, |
724 | unsigned long address); | |
725 | ||
5e31275c SB |
726 | #else /* CONFIG_PER_VMA_LOCK */ |
727 | ||
728 | static inline void vma_init_lock(struct vm_area_struct *vma) {} | |
729 | static inline bool vma_start_read(struct vm_area_struct *vma) | |
730 | { return false; } | |
731 | static inline void vma_end_read(struct vm_area_struct *vma) {} | |
732 | static inline void vma_start_write(struct vm_area_struct *vma) {} | |
55fd6fcc SB |
733 | static inline bool vma_try_start_write(struct vm_area_struct *vma) |
734 | { return true; } | |
5e31275c | 735 | static inline void vma_assert_write_locked(struct vm_area_struct *vma) {} |
457f67be SB |
736 | static inline void vma_mark_detached(struct vm_area_struct *vma, |
737 | bool detached) {} | |
5e31275c SB |
738 | |
739 | #endif /* CONFIG_PER_VMA_LOCK */ | |
740 | ||
c7f8f31c SB |
741 | /* |
742 | * WARNING: vma_init does not initialize vma->vm_lock. | |
743 | * Use vm_area_alloc()/vm_area_free() if vma needs locking. | |
744 | */ | |
027232da KS |
745 | static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm) |
746 | { | |
bfd40eaf KS |
747 | static const struct vm_operations_struct dummy_vm_ops = {}; |
748 | ||
a670468f | 749 | memset(vma, 0, sizeof(*vma)); |
027232da | 750 | vma->vm_mm = mm; |
bfd40eaf | 751 | vma->vm_ops = &dummy_vm_ops; |
027232da | 752 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
457f67be | 753 | vma_mark_detached(vma, false); |
ef6a22b7 | 754 | vma_numab_state_init(vma); |
027232da KS |
755 | } |
756 | ||
bc292ab0 SB |
757 | /* Use when VMA is not part of the VMA tree and needs no locking */ |
758 | static inline void vm_flags_init(struct vm_area_struct *vma, | |
759 | vm_flags_t flags) | |
760 | { | |
761 | ACCESS_PRIVATE(vma, __vm_flags) = flags; | |
762 | } | |
763 | ||
764 | /* Use when VMA is part of the VMA tree and modifications need coordination */ | |
765 | static inline void vm_flags_reset(struct vm_area_struct *vma, | |
766 | vm_flags_t flags) | |
767 | { | |
c7322933 | 768 | vma_start_write(vma); |
bc292ab0 SB |
769 | vm_flags_init(vma, flags); |
770 | } | |
771 | ||
601c3c29 SB |
772 | static inline void vm_flags_reset_once(struct vm_area_struct *vma, |
773 | vm_flags_t flags) | |
774 | { | |
c7322933 | 775 | vma_start_write(vma); |
601c3c29 SB |
776 | WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags); |
777 | } | |
778 | ||
bc292ab0 SB |
779 | static inline void vm_flags_set(struct vm_area_struct *vma, |
780 | vm_flags_t flags) | |
781 | { | |
c7322933 | 782 | vma_start_write(vma); |
bc292ab0 SB |
783 | ACCESS_PRIVATE(vma, __vm_flags) |= flags; |
784 | } | |
785 | ||
786 | static inline void vm_flags_clear(struct vm_area_struct *vma, | |
787 | vm_flags_t flags) | |
788 | { | |
c7322933 | 789 | vma_start_write(vma); |
bc292ab0 SB |
790 | ACCESS_PRIVATE(vma, __vm_flags) &= ~flags; |
791 | } | |
792 | ||
68f48381 SB |
793 | /* |
794 | * Use only if VMA is not part of the VMA tree or has no other users and | |
795 | * therefore needs no locking. | |
796 | */ | |
797 | static inline void __vm_flags_mod(struct vm_area_struct *vma, | |
798 | vm_flags_t set, vm_flags_t clear) | |
799 | { | |
800 | vm_flags_init(vma, (vma->vm_flags | set) & ~clear); | |
801 | } | |
802 | ||
bc292ab0 SB |
803 | /* |
804 | * Use only when the order of set/clear operations is unimportant, otherwise | |
805 | * use vm_flags_{set|clear} explicitly. | |
806 | */ | |
807 | static inline void vm_flags_mod(struct vm_area_struct *vma, | |
808 | vm_flags_t set, vm_flags_t clear) | |
809 | { | |
c7322933 | 810 | vma_start_write(vma); |
68f48381 | 811 | __vm_flags_mod(vma, set, clear); |
bc292ab0 SB |
812 | } |
813 | ||
bfd40eaf KS |
814 | static inline void vma_set_anonymous(struct vm_area_struct *vma) |
815 | { | |
816 | vma->vm_ops = NULL; | |
817 | } | |
818 | ||
43675e6f YS |
819 | static inline bool vma_is_anonymous(struct vm_area_struct *vma) |
820 | { | |
821 | return !vma->vm_ops; | |
822 | } | |
823 | ||
222100ee AK |
824 | static inline bool vma_is_temporary_stack(struct vm_area_struct *vma) |
825 | { | |
826 | int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP); | |
827 | ||
828 | if (!maybe_stack) | |
829 | return false; | |
830 | ||
831 | if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) == | |
832 | VM_STACK_INCOMPLETE_SETUP) | |
833 | return true; | |
834 | ||
835 | return false; | |
836 | } | |
837 | ||
7969f226 AK |
838 | static inline bool vma_is_foreign(struct vm_area_struct *vma) |
839 | { | |
840 | if (!current->mm) | |
841 | return true; | |
842 | ||
843 | if (current->mm != vma->vm_mm) | |
844 | return true; | |
845 | ||
846 | return false; | |
847 | } | |
3122e80e AK |
848 | |
849 | static inline bool vma_is_accessible(struct vm_area_struct *vma) | |
850 | { | |
6cb4d9a2 | 851 | return vma->vm_flags & VM_ACCESS_FLAGS; |
3122e80e AK |
852 | } |
853 | ||
f39af059 MWO |
854 | static inline |
855 | struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max) | |
856 | { | |
b62b633e | 857 | return mas_find(&vmi->mas, max - 1); |
f39af059 MWO |
858 | } |
859 | ||
860 | static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi) | |
861 | { | |
862 | /* | |
b62b633e | 863 | * Uses mas_find() to get the first VMA when the iterator starts. |
f39af059 MWO |
864 | * Calling mas_next() could skip the first entry. |
865 | */ | |
b62b633e | 866 | return mas_find(&vmi->mas, ULONG_MAX); |
f39af059 MWO |
867 | } |
868 | ||
869 | static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi) | |
870 | { | |
871 | return mas_prev(&vmi->mas, 0); | |
872 | } | |
873 | ||
874 | static inline unsigned long vma_iter_addr(struct vma_iterator *vmi) | |
875 | { | |
876 | return vmi->mas.index; | |
877 | } | |
878 | ||
b62b633e LH |
879 | static inline unsigned long vma_iter_end(struct vma_iterator *vmi) |
880 | { | |
881 | return vmi->mas.last + 1; | |
882 | } | |
883 | static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi, | |
884 | unsigned long count) | |
885 | { | |
886 | return mas_expected_entries(&vmi->mas, count); | |
887 | } | |
888 | ||
889 | /* Free any unused preallocations */ | |
890 | static inline void vma_iter_free(struct vma_iterator *vmi) | |
891 | { | |
892 | mas_destroy(&vmi->mas); | |
893 | } | |
894 | ||
895 | static inline int vma_iter_bulk_store(struct vma_iterator *vmi, | |
896 | struct vm_area_struct *vma) | |
897 | { | |
898 | vmi->mas.index = vma->vm_start; | |
899 | vmi->mas.last = vma->vm_end - 1; | |
900 | mas_store(&vmi->mas, vma); | |
901 | if (unlikely(mas_is_err(&vmi->mas))) | |
902 | return -ENOMEM; | |
903 | ||
904 | return 0; | |
905 | } | |
906 | ||
907 | static inline void vma_iter_invalidate(struct vma_iterator *vmi) | |
908 | { | |
909 | mas_pause(&vmi->mas); | |
910 | } | |
911 | ||
912 | static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr) | |
913 | { | |
914 | mas_set(&vmi->mas, addr); | |
915 | } | |
916 | ||
f39af059 MWO |
917 | #define for_each_vma(__vmi, __vma) \ |
918 | while (((__vma) = vma_next(&(__vmi))) != NULL) | |
919 | ||
920 | /* The MM code likes to work with exclusive end addresses */ | |
921 | #define for_each_vma_range(__vmi, __vma, __end) \ | |
b62b633e | 922 | while (((__vma) = vma_find(&(__vmi), (__end))) != NULL) |
f39af059 | 923 | |
43675e6f YS |
924 | #ifdef CONFIG_SHMEM |
925 | /* | |
926 | * The vma_is_shmem is not inline because it is used only by slow | |
927 | * paths in userfault. | |
928 | */ | |
929 | bool vma_is_shmem(struct vm_area_struct *vma); | |
d09e8ca6 | 930 | bool vma_is_anon_shmem(struct vm_area_struct *vma); |
43675e6f YS |
931 | #else |
932 | static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; } | |
d09e8ca6 | 933 | static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; } |
43675e6f YS |
934 | #endif |
935 | ||
936 | int vma_is_stack_for_current(struct vm_area_struct *vma); | |
937 | ||
8b11ec1b LT |
938 | /* flush_tlb_range() takes a vma, not a mm, and can care about flags */ |
939 | #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) } | |
940 | ||
1da177e4 LT |
941 | struct mmu_gather; |
942 | struct inode; | |
943 | ||
5eb5cea1 MWO |
944 | /* |
945 | * compound_order() can be called without holding a reference, which means | |
946 | * that niceties like page_folio() don't work. These callers should be | |
947 | * prepared to handle wild return values. For example, PG_head may be | |
948 | * set before _folio_order is initialised, or this may be a tail page. | |
949 | * See compaction.c for some good examples. | |
950 | */ | |
5bf34d7c MWO |
951 | static inline unsigned int compound_order(struct page *page) |
952 | { | |
5eb5cea1 MWO |
953 | struct folio *folio = (struct folio *)page; |
954 | ||
955 | if (!test_bit(PG_head, &folio->flags)) | |
5bf34d7c | 956 | return 0; |
5eb5cea1 | 957 | return folio->_folio_order; |
5bf34d7c MWO |
958 | } |
959 | ||
960 | /** | |
961 | * folio_order - The allocation order of a folio. | |
962 | * @folio: The folio. | |
963 | * | |
964 | * A folio is composed of 2^order pages. See get_order() for the definition | |
965 | * of order. | |
966 | * | |
967 | * Return: The order of the folio. | |
968 | */ | |
969 | static inline unsigned int folio_order(struct folio *folio) | |
970 | { | |
c3a15bff MWO |
971 | if (!folio_test_large(folio)) |
972 | return 0; | |
973 | return folio->_folio_order; | |
5bf34d7c MWO |
974 | } |
975 | ||
71e3aac0 | 976 | #include <linux/huge_mm.h> |
1da177e4 LT |
977 | |
978 | /* | |
979 | * Methods to modify the page usage count. | |
980 | * | |
981 | * What counts for a page usage: | |
982 | * - cache mapping (page->mapping) | |
983 | * - private data (page->private) | |
984 | * - page mapped in a task's page tables, each mapping | |
985 | * is counted separately | |
986 | * | |
987 | * Also, many kernel routines increase the page count before a critical | |
988 | * routine so they can be sure the page doesn't go away from under them. | |
1da177e4 LT |
989 | */ |
990 | ||
991 | /* | |
da6052f7 | 992 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
1da177e4 | 993 | */ |
7c8ee9a8 NP |
994 | static inline int put_page_testzero(struct page *page) |
995 | { | |
fe896d18 JK |
996 | VM_BUG_ON_PAGE(page_ref_count(page) == 0, page); |
997 | return page_ref_dec_and_test(page); | |
7c8ee9a8 | 998 | } |
1da177e4 | 999 | |
b620f633 MWO |
1000 | static inline int folio_put_testzero(struct folio *folio) |
1001 | { | |
1002 | return put_page_testzero(&folio->page); | |
1003 | } | |
1004 | ||
1da177e4 | 1005 | /* |
7c8ee9a8 NP |
1006 | * Try to grab a ref unless the page has a refcount of zero, return false if |
1007 | * that is the case. | |
8e0861fa AK |
1008 | * This can be called when MMU is off so it must not access |
1009 | * any of the virtual mappings. | |
1da177e4 | 1010 | */ |
c2530328 | 1011 | static inline bool get_page_unless_zero(struct page *page) |
7c8ee9a8 | 1012 | { |
fe896d18 | 1013 | return page_ref_add_unless(page, 1, 0); |
7c8ee9a8 | 1014 | } |
1da177e4 | 1015 | |
3c1ea2c7 VMO |
1016 | static inline struct folio *folio_get_nontail_page(struct page *page) |
1017 | { | |
1018 | if (unlikely(!get_page_unless_zero(page))) | |
1019 | return NULL; | |
1020 | return (struct folio *)page; | |
1021 | } | |
1022 | ||
53df8fdc | 1023 | extern int page_is_ram(unsigned long pfn); |
124fe20d DW |
1024 | |
1025 | enum { | |
1026 | REGION_INTERSECTS, | |
1027 | REGION_DISJOINT, | |
1028 | REGION_MIXED, | |
1029 | }; | |
1030 | ||
1c29f25b TK |
1031 | int region_intersects(resource_size_t offset, size_t size, unsigned long flags, |
1032 | unsigned long desc); | |
53df8fdc | 1033 | |
48667e7a | 1034 | /* Support for virtually mapped pages */ |
b3bdda02 CL |
1035 | struct page *vmalloc_to_page(const void *addr); |
1036 | unsigned long vmalloc_to_pfn(const void *addr); | |
48667e7a | 1037 | |
0738c4bb PM |
1038 | /* |
1039 | * Determine if an address is within the vmalloc range | |
1040 | * | |
1041 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | |
1042 | * is no special casing required. | |
1043 | */ | |
9bd3bb67 AK |
1044 | |
1045 | #ifndef is_ioremap_addr | |
1046 | #define is_ioremap_addr(x) is_vmalloc_addr(x) | |
1047 | #endif | |
1048 | ||
81ac3ad9 | 1049 | #ifdef CONFIG_MMU |
186525bd | 1050 | extern bool is_vmalloc_addr(const void *x); |
81ac3ad9 KH |
1051 | extern int is_vmalloc_or_module_addr(const void *x); |
1052 | #else | |
186525bd IM |
1053 | static inline bool is_vmalloc_addr(const void *x) |
1054 | { | |
1055 | return false; | |
1056 | } | |
934831d0 | 1057 | static inline int is_vmalloc_or_module_addr(const void *x) |
81ac3ad9 KH |
1058 | { |
1059 | return 0; | |
1060 | } | |
1061 | #endif | |
9e2779fa | 1062 | |
74e8ee47 MWO |
1063 | /* |
1064 | * How many times the entire folio is mapped as a single unit (eg by a | |
1065 | * PMD or PUD entry). This is probably not what you want, except for | |
cb67f428 HD |
1066 | * debugging purposes - it does not include PTE-mapped sub-pages; look |
1067 | * at folio_mapcount() or page_mapcount() or total_mapcount() instead. | |
74e8ee47 MWO |
1068 | */ |
1069 | static inline int folio_entire_mapcount(struct folio *folio) | |
6dc5ea16 | 1070 | { |
74e8ee47 | 1071 | VM_BUG_ON_FOLIO(!folio_test_large(folio), folio); |
1aa4d03b | 1072 | return atomic_read(&folio->_entire_mapcount) + 1; |
53f9263b KS |
1073 | } |
1074 | ||
70b50f94 AA |
1075 | /* |
1076 | * The atomic page->_mapcount, starts from -1: so that transitions | |
1077 | * both from it and to it can be tracked, using atomic_inc_and_test | |
1078 | * and atomic_add_negative(-1). | |
1079 | */ | |
22b751c3 | 1080 | static inline void page_mapcount_reset(struct page *page) |
70b50f94 AA |
1081 | { |
1082 | atomic_set(&(page)->_mapcount, -1); | |
1083 | } | |
1084 | ||
c97eeb8f MWO |
1085 | /** |
1086 | * page_mapcount() - Number of times this precise page is mapped. | |
1087 | * @page: The page. | |
1088 | * | |
1089 | * The number of times this page is mapped. If this page is part of | |
1090 | * a large folio, it includes the number of times this page is mapped | |
1091 | * as part of that folio. | |
6988f31d | 1092 | * |
c97eeb8f | 1093 | * The result is undefined for pages which cannot be mapped into userspace. |
6988f31d | 1094 | * For example SLAB or special types of pages. See function page_has_type(). |
c97eeb8f | 1095 | * They use this field in struct page differently. |
6988f31d | 1096 | */ |
70b50f94 AA |
1097 | static inline int page_mapcount(struct page *page) |
1098 | { | |
cb67f428 | 1099 | int mapcount = atomic_read(&page->_mapcount) + 1; |
b20ce5e0 | 1100 | |
c97eeb8f MWO |
1101 | if (unlikely(PageCompound(page))) |
1102 | mapcount += folio_entire_mapcount(page_folio(page)); | |
1103 | ||
1104 | return mapcount; | |
b20ce5e0 KS |
1105 | } |
1106 | ||
b14224fb | 1107 | int folio_total_mapcount(struct folio *folio); |
4ba1119c | 1108 | |
cb67f428 HD |
1109 | /** |
1110 | * folio_mapcount() - Calculate the number of mappings of this folio. | |
1111 | * @folio: The folio. | |
1112 | * | |
1113 | * A large folio tracks both how many times the entire folio is mapped, | |
1114 | * and how many times each individual page in the folio is mapped. | |
1115 | * This function calculates the total number of times the folio is | |
1116 | * mapped. | |
1117 | * | |
1118 | * Return: The number of times this folio is mapped. | |
1119 | */ | |
1120 | static inline int folio_mapcount(struct folio *folio) | |
4ba1119c | 1121 | { |
cb67f428 HD |
1122 | if (likely(!folio_test_large(folio))) |
1123 | return atomic_read(&folio->_mapcount) + 1; | |
b14224fb | 1124 | return folio_total_mapcount(folio); |
4ba1119c MWO |
1125 | } |
1126 | ||
b20ce5e0 KS |
1127 | static inline int total_mapcount(struct page *page) |
1128 | { | |
be5ef2d9 HD |
1129 | if (likely(!PageCompound(page))) |
1130 | return atomic_read(&page->_mapcount) + 1; | |
b14224fb | 1131 | return folio_total_mapcount(page_folio(page)); |
be5ef2d9 HD |
1132 | } |
1133 | ||
1134 | static inline bool folio_large_is_mapped(struct folio *folio) | |
1135 | { | |
4b51634c | 1136 | /* |
1aa4d03b | 1137 | * Reading _entire_mapcount below could be omitted if hugetlb |
eec20426 | 1138 | * participated in incrementing nr_pages_mapped when compound mapped. |
4b51634c | 1139 | */ |
eec20426 | 1140 | return atomic_read(&folio->_nr_pages_mapped) > 0 || |
1aa4d03b | 1141 | atomic_read(&folio->_entire_mapcount) >= 0; |
cb67f428 HD |
1142 | } |
1143 | ||
1144 | /** | |
1145 | * folio_mapped - Is this folio mapped into userspace? | |
1146 | * @folio: The folio. | |
1147 | * | |
1148 | * Return: True if any page in this folio is referenced by user page tables. | |
1149 | */ | |
1150 | static inline bool folio_mapped(struct folio *folio) | |
1151 | { | |
be5ef2d9 HD |
1152 | if (likely(!folio_test_large(folio))) |
1153 | return atomic_read(&folio->_mapcount) >= 0; | |
1154 | return folio_large_is_mapped(folio); | |
1155 | } | |
1156 | ||
1157 | /* | |
1158 | * Return true if this page is mapped into pagetables. | |
1159 | * For compound page it returns true if any sub-page of compound page is mapped, | |
1160 | * even if this particular sub-page is not itself mapped by any PTE or PMD. | |
1161 | */ | |
1162 | static inline bool page_mapped(struct page *page) | |
1163 | { | |
1164 | if (likely(!PageCompound(page))) | |
1165 | return atomic_read(&page->_mapcount) >= 0; | |
1166 | return folio_large_is_mapped(page_folio(page)); | |
70b50f94 AA |
1167 | } |
1168 | ||
b49af68f CL |
1169 | static inline struct page *virt_to_head_page(const void *x) |
1170 | { | |
1171 | struct page *page = virt_to_page(x); | |
ccaafd7f | 1172 | |
1d798ca3 | 1173 | return compound_head(page); |
b49af68f CL |
1174 | } |
1175 | ||
7d4203c1 VB |
1176 | static inline struct folio *virt_to_folio(const void *x) |
1177 | { | |
1178 | struct page *page = virt_to_page(x); | |
1179 | ||
1180 | return page_folio(page); | |
1181 | } | |
1182 | ||
8d29c703 | 1183 | void __folio_put(struct folio *folio); |
ddc58f27 | 1184 | |
1d7ea732 | 1185 | void put_pages_list(struct list_head *pages); |
1da177e4 | 1186 | |
8dfcc9ba | 1187 | void split_page(struct page *page, unsigned int order); |
715cbfd6 | 1188 | void folio_copy(struct folio *dst, struct folio *src); |
8dfcc9ba | 1189 | |
a1554c00 ML |
1190 | unsigned long nr_free_buffer_pages(void); |
1191 | ||
33f2ef89 AW |
1192 | /* |
1193 | * Compound pages have a destructor function. Provide a | |
1194 | * prototype for that function and accessor functions. | |
f1e61557 | 1195 | * These are _only_ valid on the head of a compound page. |
33f2ef89 | 1196 | */ |
f1e61557 KS |
1197 | typedef void compound_page_dtor(struct page *); |
1198 | ||
1199 | /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */ | |
1200 | enum compound_dtor_id { | |
1201 | NULL_COMPOUND_DTOR, | |
1202 | COMPOUND_PAGE_DTOR, | |
1203 | #ifdef CONFIG_HUGETLB_PAGE | |
1204 | HUGETLB_PAGE_DTOR, | |
9a982250 KS |
1205 | #endif |
1206 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
1207 | TRANSHUGE_PAGE_DTOR, | |
f1e61557 KS |
1208 | #endif |
1209 | NR_COMPOUND_DTORS, | |
1210 | }; | |
ae70eddd | 1211 | extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS]; |
33f2ef89 AW |
1212 | |
1213 | static inline void set_compound_page_dtor(struct page *page, | |
f1e61557 | 1214 | enum compound_dtor_id compound_dtor) |
33f2ef89 | 1215 | { |
bad6da64 MWO |
1216 | struct folio *folio = (struct folio *)page; |
1217 | ||
f1e61557 | 1218 | VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page); |
bad6da64 MWO |
1219 | VM_BUG_ON_PAGE(!PageHead(page), page); |
1220 | folio->_folio_dtor = compound_dtor; | |
33f2ef89 AW |
1221 | } |
1222 | ||
9fd33058 SK |
1223 | static inline void folio_set_compound_dtor(struct folio *folio, |
1224 | enum compound_dtor_id compound_dtor) | |
1225 | { | |
1226 | VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio); | |
1227 | folio->_folio_dtor = compound_dtor; | |
1228 | } | |
1229 | ||
5375336c | 1230 | void destroy_large_folio(struct folio *folio); |
33f2ef89 | 1231 | |
f1e61557 | 1232 | static inline void set_compound_order(struct page *page, unsigned int order) |
d85f3385 | 1233 | { |
bad6da64 | 1234 | struct folio *folio = (struct folio *)page; |
9fd33058 SK |
1235 | |
1236 | folio->_folio_order = order; | |
1237 | #ifdef CONFIG_64BIT | |
bad6da64 | 1238 | folio->_folio_nr_pages = 1U << order; |
5232c63f | 1239 | #endif |
d8c6546b MWO |
1240 | } |
1241 | ||
a50b854e MWO |
1242 | /* Returns the number of bytes in this potentially compound page. */ |
1243 | static inline unsigned long page_size(struct page *page) | |
1244 | { | |
1245 | return PAGE_SIZE << compound_order(page); | |
1246 | } | |
1247 | ||
94ad9338 MWO |
1248 | /* Returns the number of bits needed for the number of bytes in a page */ |
1249 | static inline unsigned int page_shift(struct page *page) | |
1250 | { | |
1251 | return PAGE_SHIFT + compound_order(page); | |
1252 | } | |
1253 | ||
18788cfa MWO |
1254 | /** |
1255 | * thp_order - Order of a transparent huge page. | |
1256 | * @page: Head page of a transparent huge page. | |
1257 | */ | |
1258 | static inline unsigned int thp_order(struct page *page) | |
1259 | { | |
1260 | VM_BUG_ON_PGFLAGS(PageTail(page), page); | |
1261 | return compound_order(page); | |
1262 | } | |
1263 | ||
18788cfa MWO |
1264 | /** |
1265 | * thp_size - Size of a transparent huge page. | |
1266 | * @page: Head page of a transparent huge page. | |
1267 | * | |
1268 | * Return: Number of bytes in this page. | |
1269 | */ | |
1270 | static inline unsigned long thp_size(struct page *page) | |
1271 | { | |
1272 | return PAGE_SIZE << thp_order(page); | |
1273 | } | |
1274 | ||
9a982250 KS |
1275 | void free_compound_page(struct page *page); |
1276 | ||
3dece370 | 1277 | #ifdef CONFIG_MMU |
14fd403f AA |
1278 | /* |
1279 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | |
1280 | * servicing faults for write access. In the normal case, do always want | |
1281 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | |
1282 | * that do not have writing enabled, when used by access_process_vm. | |
1283 | */ | |
1284 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | |
1285 | { | |
1286 | if (likely(vma->vm_flags & VM_WRITE)) | |
1287 | pte = pte_mkwrite(pte); | |
1288 | return pte; | |
1289 | } | |
8c6e50b0 | 1290 | |
f9ce0be7 | 1291 | vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page); |
9d3af4b4 | 1292 | void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr); |
f9ce0be7 | 1293 | |
2b740303 SJ |
1294 | vm_fault_t finish_fault(struct vm_fault *vmf); |
1295 | vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf); | |
3dece370 | 1296 | #endif |
14fd403f | 1297 | |
1da177e4 LT |
1298 | /* |
1299 | * Multiple processes may "see" the same page. E.g. for untouched | |
1300 | * mappings of /dev/null, all processes see the same page full of | |
1301 | * zeroes, and text pages of executables and shared libraries have | |
1302 | * only one copy in memory, at most, normally. | |
1303 | * | |
1304 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
7e871b6c PBG |
1305 | * page_count() == 0 means the page is free. page->lru is then used for |
1306 | * freelist management in the buddy allocator. | |
da6052f7 | 1307 | * page_count() > 0 means the page has been allocated. |
1da177e4 | 1308 | * |
da6052f7 NP |
1309 | * Pages are allocated by the slab allocator in order to provide memory |
1310 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
1311 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
1312 | * unless a particular usage is carefully commented. (the responsibility of | |
1313 | * freeing the kmalloc memory is the caller's, of course). | |
1da177e4 | 1314 | * |
da6052f7 NP |
1315 | * A page may be used by anyone else who does a __get_free_page(). |
1316 | * In this case, page_count still tracks the references, and should only | |
1317 | * be used through the normal accessor functions. The top bits of page->flags | |
1318 | * and page->virtual store page management information, but all other fields | |
1319 | * are unused and could be used privately, carefully. The management of this | |
1320 | * page is the responsibility of the one who allocated it, and those who have | |
1321 | * subsequently been given references to it. | |
1322 | * | |
1323 | * The other pages (we may call them "pagecache pages") are completely | |
1da177e4 LT |
1324 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
1325 | * The following discussion applies only to them. | |
1326 | * | |
da6052f7 NP |
1327 | * A pagecache page contains an opaque `private' member, which belongs to the |
1328 | * page's address_space. Usually, this is the address of a circular list of | |
1329 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
1330 | * into the filesystem to release these pages. | |
1da177e4 | 1331 | * |
da6052f7 NP |
1332 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
1333 | * is the pointer to the inode, and page->index is the file offset of the page, | |
ea1754a0 | 1334 | * in units of PAGE_SIZE. |
1da177e4 | 1335 | * |
da6052f7 NP |
1336 | * If pagecache pages are not associated with an inode, they are said to be |
1337 | * anonymous pages. These may become associated with the swapcache, and in that | |
1338 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
1da177e4 | 1339 | * |
da6052f7 NP |
1340 | * In either case (swapcache or inode backed), the pagecache itself holds one |
1341 | * reference to the page. Setting PG_private should also increment the | |
1342 | * refcount. The each user mapping also has a reference to the page. | |
1da177e4 | 1343 | * |
da6052f7 | 1344 | * The pagecache pages are stored in a per-mapping radix tree, which is |
b93b0163 | 1345 | * rooted at mapping->i_pages, and indexed by offset. |
da6052f7 NP |
1346 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
1347 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
1da177e4 | 1348 | * |
da6052f7 | 1349 | * All pagecache pages may be subject to I/O: |
1da177e4 LT |
1350 | * - inode pages may need to be read from disk, |
1351 | * - inode pages which have been modified and are MAP_SHARED may need | |
da6052f7 NP |
1352 | * to be written back to the inode on disk, |
1353 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
1354 | * modified may need to be swapped out to swap space and (later) to be read | |
1355 | * back into memory. | |
1da177e4 LT |
1356 | */ |
1357 | ||
27674ef6 | 1358 | #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX) |
e7638488 | 1359 | DECLARE_STATIC_KEY_FALSE(devmap_managed_key); |
07d80269 | 1360 | |
f4f451a1 MS |
1361 | bool __put_devmap_managed_page_refs(struct page *page, int refs); |
1362 | static inline bool put_devmap_managed_page_refs(struct page *page, int refs) | |
e7638488 DW |
1363 | { |
1364 | if (!static_branch_unlikely(&devmap_managed_key)) | |
1365 | return false; | |
1366 | if (!is_zone_device_page(page)) | |
1367 | return false; | |
f4f451a1 | 1368 | return __put_devmap_managed_page_refs(page, refs); |
e7638488 | 1369 | } |
27674ef6 | 1370 | #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */ |
f4f451a1 | 1371 | static inline bool put_devmap_managed_page_refs(struct page *page, int refs) |
e7638488 DW |
1372 | { |
1373 | return false; | |
1374 | } | |
27674ef6 | 1375 | #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */ |
7b2d55d2 | 1376 | |
f4f451a1 MS |
1377 | static inline bool put_devmap_managed_page(struct page *page) |
1378 | { | |
1379 | return put_devmap_managed_page_refs(page, 1); | |
1380 | } | |
1381 | ||
f958d7b5 | 1382 | /* 127: arbitrary random number, small enough to assemble well */ |
86d234cb MWO |
1383 | #define folio_ref_zero_or_close_to_overflow(folio) \ |
1384 | ((unsigned int) folio_ref_count(folio) + 127u <= 127u) | |
1385 | ||
1386 | /** | |
1387 | * folio_get - Increment the reference count on a folio. | |
1388 | * @folio: The folio. | |
1389 | * | |
1390 | * Context: May be called in any context, as long as you know that | |
1391 | * you have a refcount on the folio. If you do not already have one, | |
1392 | * folio_try_get() may be the right interface for you to use. | |
1393 | */ | |
1394 | static inline void folio_get(struct folio *folio) | |
1395 | { | |
1396 | VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio); | |
1397 | folio_ref_inc(folio); | |
1398 | } | |
f958d7b5 | 1399 | |
3565fce3 DW |
1400 | static inline void get_page(struct page *page) |
1401 | { | |
86d234cb | 1402 | folio_get(page_folio(page)); |
3565fce3 DW |
1403 | } |
1404 | ||
cd1adf1b LT |
1405 | static inline __must_check bool try_get_page(struct page *page) |
1406 | { | |
1407 | page = compound_head(page); | |
1408 | if (WARN_ON_ONCE(page_ref_count(page) <= 0)) | |
1409 | return false; | |
1410 | page_ref_inc(page); | |
1411 | return true; | |
1412 | } | |
3565fce3 | 1413 | |
b620f633 MWO |
1414 | /** |
1415 | * folio_put - Decrement the reference count on a folio. | |
1416 | * @folio: The folio. | |
1417 | * | |
1418 | * If the folio's reference count reaches zero, the memory will be | |
1419 | * released back to the page allocator and may be used by another | |
1420 | * allocation immediately. Do not access the memory or the struct folio | |
1421 | * after calling folio_put() unless you can be sure that it wasn't the | |
1422 | * last reference. | |
1423 | * | |
1424 | * Context: May be called in process or interrupt context, but not in NMI | |
1425 | * context. May be called while holding a spinlock. | |
1426 | */ | |
1427 | static inline void folio_put(struct folio *folio) | |
1428 | { | |
1429 | if (folio_put_testzero(folio)) | |
8d29c703 | 1430 | __folio_put(folio); |
b620f633 MWO |
1431 | } |
1432 | ||
3fe7fa58 MWO |
1433 | /** |
1434 | * folio_put_refs - Reduce the reference count on a folio. | |
1435 | * @folio: The folio. | |
1436 | * @refs: The amount to subtract from the folio's reference count. | |
1437 | * | |
1438 | * If the folio's reference count reaches zero, the memory will be | |
1439 | * released back to the page allocator and may be used by another | |
1440 | * allocation immediately. Do not access the memory or the struct folio | |
1441 | * after calling folio_put_refs() unless you can be sure that these weren't | |
1442 | * the last references. | |
1443 | * | |
1444 | * Context: May be called in process or interrupt context, but not in NMI | |
1445 | * context. May be called while holding a spinlock. | |
1446 | */ | |
1447 | static inline void folio_put_refs(struct folio *folio, int refs) | |
1448 | { | |
1449 | if (folio_ref_sub_and_test(folio, refs)) | |
8d29c703 | 1450 | __folio_put(folio); |
3fe7fa58 MWO |
1451 | } |
1452 | ||
0411d6ee SP |
1453 | /* |
1454 | * union release_pages_arg - an array of pages or folios | |
449c7967 | 1455 | * |
0411d6ee | 1456 | * release_pages() releases a simple array of multiple pages, and |
449c7967 LT |
1457 | * accepts various different forms of said page array: either |
1458 | * a regular old boring array of pages, an array of folios, or | |
1459 | * an array of encoded page pointers. | |
1460 | * | |
1461 | * The transparent union syntax for this kind of "any of these | |
1462 | * argument types" is all kinds of ugly, so look away. | |
1463 | */ | |
1464 | typedef union { | |
1465 | struct page **pages; | |
1466 | struct folio **folios; | |
1467 | struct encoded_page **encoded_pages; | |
1468 | } release_pages_arg __attribute__ ((__transparent_union__)); | |
1469 | ||
1470 | void release_pages(release_pages_arg, int nr); | |
e3c4cebf MWO |
1471 | |
1472 | /** | |
1473 | * folios_put - Decrement the reference count on an array of folios. | |
1474 | * @folios: The folios. | |
1475 | * @nr: How many folios there are. | |
1476 | * | |
1477 | * Like folio_put(), but for an array of folios. This is more efficient | |
1478 | * than writing the loop yourself as it will optimise the locks which | |
1479 | * need to be taken if the folios are freed. | |
1480 | * | |
1481 | * Context: May be called in process or interrupt context, but not in NMI | |
1482 | * context. May be called while holding a spinlock. | |
1483 | */ | |
1484 | static inline void folios_put(struct folio **folios, unsigned int nr) | |
1485 | { | |
449c7967 | 1486 | release_pages(folios, nr); |
3fe7fa58 MWO |
1487 | } |
1488 | ||
3565fce3 DW |
1489 | static inline void put_page(struct page *page) |
1490 | { | |
b620f633 | 1491 | struct folio *folio = page_folio(page); |
3565fce3 | 1492 | |
7b2d55d2 | 1493 | /* |
89574945 CH |
1494 | * For some devmap managed pages we need to catch refcount transition |
1495 | * from 2 to 1: | |
7b2d55d2 | 1496 | */ |
89574945 | 1497 | if (put_devmap_managed_page(&folio->page)) |
7b2d55d2 | 1498 | return; |
b620f633 | 1499 | folio_put(folio); |
3565fce3 DW |
1500 | } |
1501 | ||
3faa52c0 JH |
1502 | /* |
1503 | * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload | |
1504 | * the page's refcount so that two separate items are tracked: the original page | |
1505 | * reference count, and also a new count of how many pin_user_pages() calls were | |
1506 | * made against the page. ("gup-pinned" is another term for the latter). | |
1507 | * | |
1508 | * With this scheme, pin_user_pages() becomes special: such pages are marked as | |
1509 | * distinct from normal pages. As such, the unpin_user_page() call (and its | |
1510 | * variants) must be used in order to release gup-pinned pages. | |
1511 | * | |
1512 | * Choice of value: | |
1513 | * | |
1514 | * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference | |
1515 | * counts with respect to pin_user_pages() and unpin_user_page() becomes | |
1516 | * simpler, due to the fact that adding an even power of two to the page | |
1517 | * refcount has the effect of using only the upper N bits, for the code that | |
1518 | * counts up using the bias value. This means that the lower bits are left for | |
1519 | * the exclusive use of the original code that increments and decrements by one | |
1520 | * (or at least, by much smaller values than the bias value). | |
fc1d8e7c | 1521 | * |
3faa52c0 JH |
1522 | * Of course, once the lower bits overflow into the upper bits (and this is |
1523 | * OK, because subtraction recovers the original values), then visual inspection | |
1524 | * no longer suffices to directly view the separate counts. However, for normal | |
1525 | * applications that don't have huge page reference counts, this won't be an | |
1526 | * issue. | |
fc1d8e7c | 1527 | * |
40fcc7fc MWO |
1528 | * Locking: the lockless algorithm described in folio_try_get_rcu() |
1529 | * provides safe operation for get_user_pages(), page_mkclean() and | |
1530 | * other calls that race to set up page table entries. | |
fc1d8e7c | 1531 | */ |
3faa52c0 | 1532 | #define GUP_PIN_COUNTING_BIAS (1U << 10) |
fc1d8e7c | 1533 | |
3faa52c0 | 1534 | void unpin_user_page(struct page *page); |
f1f6a7dd JH |
1535 | void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages, |
1536 | bool make_dirty); | |
458a4f78 JM |
1537 | void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages, |
1538 | bool make_dirty); | |
f1f6a7dd | 1539 | void unpin_user_pages(struct page **pages, unsigned long npages); |
fc1d8e7c | 1540 | |
97a7e473 PX |
1541 | static inline bool is_cow_mapping(vm_flags_t flags) |
1542 | { | |
1543 | return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE; | |
1544 | } | |
1545 | ||
fc4f4be9 DH |
1546 | #ifndef CONFIG_MMU |
1547 | static inline bool is_nommu_shared_mapping(vm_flags_t flags) | |
1548 | { | |
1549 | /* | |
1550 | * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected | |
1551 | * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of | |
1552 | * a file mapping. R/O MAP_PRIVATE mappings might still modify | |
1553 | * underlying memory if ptrace is active, so this is only possible if | |
1554 | * ptrace does not apply. Note that there is no mprotect() to upgrade | |
1555 | * write permissions later. | |
1556 | */ | |
b6b7a8fa | 1557 | return flags & (VM_MAYSHARE | VM_MAYOVERLAY); |
fc4f4be9 DH |
1558 | } |
1559 | #endif | |
1560 | ||
9127ab4f CS |
1561 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
1562 | #define SECTION_IN_PAGE_FLAGS | |
1563 | #endif | |
1564 | ||
89689ae7 | 1565 | /* |
7a8010cd VB |
1566 | * The identification function is mainly used by the buddy allocator for |
1567 | * determining if two pages could be buddies. We are not really identifying | |
1568 | * the zone since we could be using the section number id if we do not have | |
1569 | * node id available in page flags. | |
1570 | * We only guarantee that it will return the same value for two combinable | |
1571 | * pages in a zone. | |
89689ae7 | 1572 | */ |
cb2b95e1 AW |
1573 | static inline int page_zone_id(struct page *page) |
1574 | { | |
89689ae7 | 1575 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
348f8b6c DH |
1576 | } |
1577 | ||
89689ae7 | 1578 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
33dd4e0e | 1579 | extern int page_to_nid(const struct page *page); |
89689ae7 | 1580 | #else |
33dd4e0e | 1581 | static inline int page_to_nid(const struct page *page) |
d41dee36 | 1582 | { |
f165b378 PT |
1583 | struct page *p = (struct page *)page; |
1584 | ||
1585 | return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK; | |
d41dee36 | 1586 | } |
89689ae7 CL |
1587 | #endif |
1588 | ||
874fd90c MWO |
1589 | static inline int folio_nid(const struct folio *folio) |
1590 | { | |
1591 | return page_to_nid(&folio->page); | |
1592 | } | |
1593 | ||
57e0a030 | 1594 | #ifdef CONFIG_NUMA_BALANCING |
33024536 HY |
1595 | /* page access time bits needs to hold at least 4 seconds */ |
1596 | #define PAGE_ACCESS_TIME_MIN_BITS 12 | |
1597 | #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS | |
1598 | #define PAGE_ACCESS_TIME_BUCKETS \ | |
1599 | (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT) | |
1600 | #else | |
1601 | #define PAGE_ACCESS_TIME_BUCKETS 0 | |
1602 | #endif | |
1603 | ||
1604 | #define PAGE_ACCESS_TIME_MASK \ | |
1605 | (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS) | |
1606 | ||
90572890 | 1607 | static inline int cpu_pid_to_cpupid(int cpu, int pid) |
57e0a030 | 1608 | { |
90572890 | 1609 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); |
57e0a030 MG |
1610 | } |
1611 | ||
90572890 | 1612 | static inline int cpupid_to_pid(int cpupid) |
57e0a030 | 1613 | { |
90572890 | 1614 | return cpupid & LAST__PID_MASK; |
57e0a030 | 1615 | } |
b795854b | 1616 | |
90572890 | 1617 | static inline int cpupid_to_cpu(int cpupid) |
b795854b | 1618 | { |
90572890 | 1619 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; |
b795854b MG |
1620 | } |
1621 | ||
90572890 | 1622 | static inline int cpupid_to_nid(int cpupid) |
b795854b | 1623 | { |
90572890 | 1624 | return cpu_to_node(cpupid_to_cpu(cpupid)); |
b795854b MG |
1625 | } |
1626 | ||
90572890 | 1627 | static inline bool cpupid_pid_unset(int cpupid) |
57e0a030 | 1628 | { |
90572890 | 1629 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); |
b795854b MG |
1630 | } |
1631 | ||
90572890 | 1632 | static inline bool cpupid_cpu_unset(int cpupid) |
b795854b | 1633 | { |
90572890 | 1634 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); |
b795854b MG |
1635 | } |
1636 | ||
8c8a743c PZ |
1637 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) |
1638 | { | |
1639 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); | |
1640 | } | |
1641 | ||
1642 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) | |
90572890 PZ |
1643 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
1644 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
b795854b | 1645 | { |
1ae71d03 | 1646 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); |
b795854b | 1647 | } |
90572890 PZ |
1648 | |
1649 | static inline int page_cpupid_last(struct page *page) | |
1650 | { | |
1651 | return page->_last_cpupid; | |
1652 | } | |
1653 | static inline void page_cpupid_reset_last(struct page *page) | |
b795854b | 1654 | { |
1ae71d03 | 1655 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; |
57e0a030 MG |
1656 | } |
1657 | #else | |
90572890 | 1658 | static inline int page_cpupid_last(struct page *page) |
75980e97 | 1659 | { |
90572890 | 1660 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; |
75980e97 PZ |
1661 | } |
1662 | ||
90572890 | 1663 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); |
75980e97 | 1664 | |
90572890 | 1665 | static inline void page_cpupid_reset_last(struct page *page) |
75980e97 | 1666 | { |
09940a4f | 1667 | page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT; |
75980e97 | 1668 | } |
90572890 | 1669 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ |
33024536 HY |
1670 | |
1671 | static inline int xchg_page_access_time(struct page *page, int time) | |
1672 | { | |
1673 | int last_time; | |
1674 | ||
1675 | last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS); | |
1676 | return last_time << PAGE_ACCESS_TIME_BUCKETS; | |
1677 | } | |
fc137c0d R |
1678 | |
1679 | static inline void vma_set_access_pid_bit(struct vm_area_struct *vma) | |
1680 | { | |
1681 | unsigned int pid_bit; | |
1682 | ||
d46031f4 | 1683 | pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG)); |
20f58648 R |
1684 | if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->access_pids[1])) { |
1685 | __set_bit(pid_bit, &vma->numab_state->access_pids[1]); | |
fc137c0d R |
1686 | } |
1687 | } | |
90572890 PZ |
1688 | #else /* !CONFIG_NUMA_BALANCING */ |
1689 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
57e0a030 | 1690 | { |
90572890 | 1691 | return page_to_nid(page); /* XXX */ |
57e0a030 MG |
1692 | } |
1693 | ||
33024536 HY |
1694 | static inline int xchg_page_access_time(struct page *page, int time) |
1695 | { | |
1696 | return 0; | |
1697 | } | |
1698 | ||
90572890 | 1699 | static inline int page_cpupid_last(struct page *page) |
57e0a030 | 1700 | { |
90572890 | 1701 | return page_to_nid(page); /* XXX */ |
57e0a030 MG |
1702 | } |
1703 | ||
90572890 | 1704 | static inline int cpupid_to_nid(int cpupid) |
b795854b MG |
1705 | { |
1706 | return -1; | |
1707 | } | |
1708 | ||
90572890 | 1709 | static inline int cpupid_to_pid(int cpupid) |
b795854b MG |
1710 | { |
1711 | return -1; | |
1712 | } | |
1713 | ||
90572890 | 1714 | static inline int cpupid_to_cpu(int cpupid) |
b795854b MG |
1715 | { |
1716 | return -1; | |
1717 | } | |
1718 | ||
90572890 PZ |
1719 | static inline int cpu_pid_to_cpupid(int nid, int pid) |
1720 | { | |
1721 | return -1; | |
1722 | } | |
1723 | ||
1724 | static inline bool cpupid_pid_unset(int cpupid) | |
b795854b | 1725 | { |
2b787449 | 1726 | return true; |
b795854b MG |
1727 | } |
1728 | ||
90572890 | 1729 | static inline void page_cpupid_reset_last(struct page *page) |
57e0a030 MG |
1730 | { |
1731 | } | |
8c8a743c PZ |
1732 | |
1733 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) | |
1734 | { | |
1735 | return false; | |
1736 | } | |
fc137c0d R |
1737 | |
1738 | static inline void vma_set_access_pid_bit(struct vm_area_struct *vma) | |
1739 | { | |
1740 | } | |
90572890 | 1741 | #endif /* CONFIG_NUMA_BALANCING */ |
57e0a030 | 1742 | |
2e903b91 | 1743 | #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS) |
34303244 | 1744 | |
cf10bd4c AK |
1745 | /* |
1746 | * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid | |
1747 | * setting tags for all pages to native kernel tag value 0xff, as the default | |
1748 | * value 0x00 maps to 0xff. | |
1749 | */ | |
1750 | ||
2813b9c0 AK |
1751 | static inline u8 page_kasan_tag(const struct page *page) |
1752 | { | |
cf10bd4c AK |
1753 | u8 tag = 0xff; |
1754 | ||
1755 | if (kasan_enabled()) { | |
1756 | tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK; | |
1757 | tag ^= 0xff; | |
1758 | } | |
1759 | ||
1760 | return tag; | |
2813b9c0 AK |
1761 | } |
1762 | ||
1763 | static inline void page_kasan_tag_set(struct page *page, u8 tag) | |
1764 | { | |
27fe7339 PC |
1765 | unsigned long old_flags, flags; |
1766 | ||
1767 | if (!kasan_enabled()) | |
1768 | return; | |
1769 | ||
1770 | tag ^= 0xff; | |
1771 | old_flags = READ_ONCE(page->flags); | |
1772 | do { | |
1773 | flags = old_flags; | |
1774 | flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT); | |
1775 | flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT; | |
1776 | } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags))); | |
2813b9c0 AK |
1777 | } |
1778 | ||
1779 | static inline void page_kasan_tag_reset(struct page *page) | |
1780 | { | |
34303244 AK |
1781 | if (kasan_enabled()) |
1782 | page_kasan_tag_set(page, 0xff); | |
2813b9c0 | 1783 | } |
34303244 AK |
1784 | |
1785 | #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */ | |
1786 | ||
2813b9c0 AK |
1787 | static inline u8 page_kasan_tag(const struct page *page) |
1788 | { | |
1789 | return 0xff; | |
1790 | } | |
1791 | ||
1792 | static inline void page_kasan_tag_set(struct page *page, u8 tag) { } | |
1793 | static inline void page_kasan_tag_reset(struct page *page) { } | |
34303244 AK |
1794 | |
1795 | #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */ | |
2813b9c0 | 1796 | |
33dd4e0e | 1797 | static inline struct zone *page_zone(const struct page *page) |
89689ae7 CL |
1798 | { |
1799 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
1800 | } | |
1801 | ||
75ef7184 MG |
1802 | static inline pg_data_t *page_pgdat(const struct page *page) |
1803 | { | |
1804 | return NODE_DATA(page_to_nid(page)); | |
1805 | } | |
1806 | ||
32b8fc48 MWO |
1807 | static inline struct zone *folio_zone(const struct folio *folio) |
1808 | { | |
1809 | return page_zone(&folio->page); | |
1810 | } | |
1811 | ||
1812 | static inline pg_data_t *folio_pgdat(const struct folio *folio) | |
1813 | { | |
1814 | return page_pgdat(&folio->page); | |
1815 | } | |
1816 | ||
9127ab4f | 1817 | #ifdef SECTION_IN_PAGE_FLAGS |
bf4e8902 DK |
1818 | static inline void set_page_section(struct page *page, unsigned long section) |
1819 | { | |
1820 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
1821 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
1822 | } | |
1823 | ||
aa462abe | 1824 | static inline unsigned long page_to_section(const struct page *page) |
d41dee36 AW |
1825 | { |
1826 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
1827 | } | |
308c05e3 | 1828 | #endif |
d41dee36 | 1829 | |
bf6bd276 MWO |
1830 | /** |
1831 | * folio_pfn - Return the Page Frame Number of a folio. | |
1832 | * @folio: The folio. | |
1833 | * | |
1834 | * A folio may contain multiple pages. The pages have consecutive | |
1835 | * Page Frame Numbers. | |
1836 | * | |
1837 | * Return: The Page Frame Number of the first page in the folio. | |
1838 | */ | |
1839 | static inline unsigned long folio_pfn(struct folio *folio) | |
1840 | { | |
1841 | return page_to_pfn(&folio->page); | |
1842 | } | |
1843 | ||
018ee47f YZ |
1844 | static inline struct folio *pfn_folio(unsigned long pfn) |
1845 | { | |
1846 | return page_folio(pfn_to_page(pfn)); | |
1847 | } | |
1848 | ||
0b90ddae MWO |
1849 | /** |
1850 | * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA. | |
1851 | * @folio: The folio. | |
1852 | * | |
1853 | * This function checks if a folio has been pinned via a call to | |
1854 | * a function in the pin_user_pages() family. | |
1855 | * | |
1856 | * For small folios, the return value is partially fuzzy: false is not fuzzy, | |
1857 | * because it means "definitely not pinned for DMA", but true means "probably | |
1858 | * pinned for DMA, but possibly a false positive due to having at least | |
1859 | * GUP_PIN_COUNTING_BIAS worth of normal folio references". | |
1860 | * | |
1861 | * False positives are OK, because: a) it's unlikely for a folio to | |
1862 | * get that many refcounts, and b) all the callers of this routine are | |
1863 | * expected to be able to deal gracefully with a false positive. | |
1864 | * | |
1865 | * For large folios, the result will be exactly correct. That's because | |
94688e8e | 1866 | * we have more tracking data available: the _pincount field is used |
0b90ddae MWO |
1867 | * instead of the GUP_PIN_COUNTING_BIAS scheme. |
1868 | * | |
1869 | * For more information, please see Documentation/core-api/pin_user_pages.rst. | |
1870 | * | |
1871 | * Return: True, if it is likely that the page has been "dma-pinned". | |
1872 | * False, if the page is definitely not dma-pinned. | |
1873 | */ | |
1874 | static inline bool folio_maybe_dma_pinned(struct folio *folio) | |
1875 | { | |
1876 | if (folio_test_large(folio)) | |
94688e8e | 1877 | return atomic_read(&folio->_pincount) > 0; |
0b90ddae MWO |
1878 | |
1879 | /* | |
1880 | * folio_ref_count() is signed. If that refcount overflows, then | |
1881 | * folio_ref_count() returns a negative value, and callers will avoid | |
1882 | * further incrementing the refcount. | |
1883 | * | |
1884 | * Here, for that overflow case, use the sign bit to count a little | |
1885 | * bit higher via unsigned math, and thus still get an accurate result. | |
1886 | */ | |
1887 | return ((unsigned int)folio_ref_count(folio)) >= | |
1888 | GUP_PIN_COUNTING_BIAS; | |
1889 | } | |
1890 | ||
1891 | static inline bool page_maybe_dma_pinned(struct page *page) | |
1892 | { | |
1893 | return folio_maybe_dma_pinned(page_folio(page)); | |
1894 | } | |
1895 | ||
1896 | /* | |
1897 | * This should most likely only be called during fork() to see whether we | |
fb3d824d | 1898 | * should break the cow immediately for an anon page on the src mm. |
623a1ddf DH |
1899 | * |
1900 | * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq. | |
0b90ddae MWO |
1901 | */ |
1902 | static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma, | |
1903 | struct page *page) | |
1904 | { | |
623a1ddf | 1905 | VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1)); |
0b90ddae MWO |
1906 | |
1907 | if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags)) | |
1908 | return false; | |
1909 | ||
1910 | return page_maybe_dma_pinned(page); | |
1911 | } | |
1912 | ||
c8070b78 DH |
1913 | /** |
1914 | * is_zero_page - Query if a page is a zero page | |
1915 | * @page: The page to query | |
1916 | * | |
1917 | * This returns true if @page is one of the permanent zero pages. | |
1918 | */ | |
1919 | static inline bool is_zero_page(const struct page *page) | |
1920 | { | |
1921 | return is_zero_pfn(page_to_pfn(page)); | |
1922 | } | |
1923 | ||
1924 | /** | |
1925 | * is_zero_folio - Query if a folio is a zero page | |
1926 | * @folio: The folio to query | |
1927 | * | |
1928 | * This returns true if @folio is one of the permanent zero pages. | |
1929 | */ | |
1930 | static inline bool is_zero_folio(const struct folio *folio) | |
1931 | { | |
1932 | return is_zero_page(&folio->page); | |
1933 | } | |
1934 | ||
8e3560d9 PT |
1935 | /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */ |
1936 | #ifdef CONFIG_MIGRATION | |
6077c943 | 1937 | static inline bool is_longterm_pinnable_page(struct page *page) |
8e3560d9 | 1938 | { |
1c563432 MK |
1939 | #ifdef CONFIG_CMA |
1940 | int mt = get_pageblock_migratetype(page); | |
1941 | ||
1942 | if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE) | |
1943 | return false; | |
1944 | #endif | |
c8070b78 DH |
1945 | /* The zero page can be "pinned" but gets special handling. */ |
1946 | if (is_zero_page(page)) | |
fcab34b4 AW |
1947 | return true; |
1948 | ||
1949 | /* Coherent device memory must always allow eviction. */ | |
1950 | if (is_device_coherent_page(page)) | |
1951 | return false; | |
1952 | ||
1953 | /* Otherwise, non-movable zone pages can be pinned. */ | |
1954 | return !is_zone_movable_page(page); | |
8e3560d9 PT |
1955 | } |
1956 | #else | |
6077c943 | 1957 | static inline bool is_longterm_pinnable_page(struct page *page) |
8e3560d9 PT |
1958 | { |
1959 | return true; | |
1960 | } | |
1961 | #endif | |
1962 | ||
6077c943 | 1963 | static inline bool folio_is_longterm_pinnable(struct folio *folio) |
536939ff | 1964 | { |
6077c943 | 1965 | return is_longterm_pinnable_page(&folio->page); |
536939ff MWO |
1966 | } |
1967 | ||
2f1b6248 | 1968 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
348f8b6c DH |
1969 | { |
1970 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
1971 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
1972 | } | |
2f1b6248 | 1973 | |
348f8b6c DH |
1974 | static inline void set_page_node(struct page *page, unsigned long node) |
1975 | { | |
1976 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
1977 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
1da177e4 | 1978 | } |
89689ae7 | 1979 | |
2f1b6248 | 1980 | static inline void set_page_links(struct page *page, enum zone_type zone, |
d41dee36 | 1981 | unsigned long node, unsigned long pfn) |
1da177e4 | 1982 | { |
348f8b6c DH |
1983 | set_page_zone(page, zone); |
1984 | set_page_node(page, node); | |
9127ab4f | 1985 | #ifdef SECTION_IN_PAGE_FLAGS |
d41dee36 | 1986 | set_page_section(page, pfn_to_section_nr(pfn)); |
bf4e8902 | 1987 | #endif |
1da177e4 LT |
1988 | } |
1989 | ||
7b230db3 MWO |
1990 | /** |
1991 | * folio_nr_pages - The number of pages in the folio. | |
1992 | * @folio: The folio. | |
1993 | * | |
1994 | * Return: A positive power of two. | |
1995 | */ | |
1996 | static inline long folio_nr_pages(struct folio *folio) | |
1997 | { | |
c3a15bff MWO |
1998 | if (!folio_test_large(folio)) |
1999 | return 1; | |
2000 | #ifdef CONFIG_64BIT | |
2001 | return folio->_folio_nr_pages; | |
2002 | #else | |
2003 | return 1L << folio->_folio_order; | |
2004 | #endif | |
7b230db3 MWO |
2005 | } |
2006 | ||
21a000fe MWO |
2007 | /* |
2008 | * compound_nr() returns the number of pages in this potentially compound | |
2009 | * page. compound_nr() can be called on a tail page, and is defined to | |
2010 | * return 1 in that case. | |
2011 | */ | |
2012 | static inline unsigned long compound_nr(struct page *page) | |
2013 | { | |
2014 | struct folio *folio = (struct folio *)page; | |
2015 | ||
2016 | if (!test_bit(PG_head, &folio->flags)) | |
2017 | return 1; | |
2018 | #ifdef CONFIG_64BIT | |
2019 | return folio->_folio_nr_pages; | |
2020 | #else | |
2021 | return 1L << folio->_folio_order; | |
2022 | #endif | |
2023 | } | |
2024 | ||
2025 | /** | |
2026 | * thp_nr_pages - The number of regular pages in this huge page. | |
2027 | * @page: The head page of a huge page. | |
2028 | */ | |
2029 | static inline int thp_nr_pages(struct page *page) | |
2030 | { | |
2031 | return folio_nr_pages((struct folio *)page); | |
2032 | } | |
2033 | ||
7b230db3 MWO |
2034 | /** |
2035 | * folio_next - Move to the next physical folio. | |
2036 | * @folio: The folio we're currently operating on. | |
2037 | * | |
2038 | * If you have physically contiguous memory which may span more than | |
2039 | * one folio (eg a &struct bio_vec), use this function to move from one | |
2040 | * folio to the next. Do not use it if the memory is only virtually | |
2041 | * contiguous as the folios are almost certainly not adjacent to each | |
2042 | * other. This is the folio equivalent to writing ``page++``. | |
2043 | * | |
2044 | * Context: We assume that the folios are refcounted and/or locked at a | |
2045 | * higher level and do not adjust the reference counts. | |
2046 | * Return: The next struct folio. | |
2047 | */ | |
2048 | static inline struct folio *folio_next(struct folio *folio) | |
2049 | { | |
2050 | return (struct folio *)folio_page(folio, folio_nr_pages(folio)); | |
2051 | } | |
2052 | ||
2053 | /** | |
2054 | * folio_shift - The size of the memory described by this folio. | |
2055 | * @folio: The folio. | |
2056 | * | |
2057 | * A folio represents a number of bytes which is a power-of-two in size. | |
2058 | * This function tells you which power-of-two the folio is. See also | |
2059 | * folio_size() and folio_order(). | |
2060 | * | |
2061 | * Context: The caller should have a reference on the folio to prevent | |
2062 | * it from being split. It is not necessary for the folio to be locked. | |
2063 | * Return: The base-2 logarithm of the size of this folio. | |
2064 | */ | |
2065 | static inline unsigned int folio_shift(struct folio *folio) | |
2066 | { | |
2067 | return PAGE_SHIFT + folio_order(folio); | |
2068 | } | |
2069 | ||
2070 | /** | |
2071 | * folio_size - The number of bytes in a folio. | |
2072 | * @folio: The folio. | |
2073 | * | |
2074 | * Context: The caller should have a reference on the folio to prevent | |
2075 | * it from being split. It is not necessary for the folio to be locked. | |
2076 | * Return: The number of bytes in this folio. | |
2077 | */ | |
2078 | static inline size_t folio_size(struct folio *folio) | |
2079 | { | |
2080 | return PAGE_SIZE << folio_order(folio); | |
2081 | } | |
2082 | ||
fa4e3f5f VMO |
2083 | /** |
2084 | * folio_estimated_sharers - Estimate the number of sharers of a folio. | |
2085 | * @folio: The folio. | |
2086 | * | |
2087 | * folio_estimated_sharers() aims to serve as a function to efficiently | |
2088 | * estimate the number of processes sharing a folio. This is done by | |
2089 | * looking at the precise mapcount of the first subpage in the folio, and | |
2090 | * assuming the other subpages are the same. This may not be true for large | |
2091 | * folios. If you want exact mapcounts for exact calculations, look at | |
2092 | * page_mapcount() or folio_total_mapcount(). | |
2093 | * | |
2094 | * Return: The estimated number of processes sharing a folio. | |
2095 | */ | |
2096 | static inline int folio_estimated_sharers(struct folio *folio) | |
2097 | { | |
2098 | return page_mapcount(folio_page(folio, 0)); | |
2099 | } | |
2100 | ||
b424de33 MWO |
2101 | #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE |
2102 | static inline int arch_make_page_accessible(struct page *page) | |
2103 | { | |
2104 | return 0; | |
2105 | } | |
2106 | #endif | |
2107 | ||
2108 | #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE | |
2109 | static inline int arch_make_folio_accessible(struct folio *folio) | |
2110 | { | |
2111 | int ret; | |
2112 | long i, nr = folio_nr_pages(folio); | |
2113 | ||
2114 | for (i = 0; i < nr; i++) { | |
2115 | ret = arch_make_page_accessible(folio_page(folio, i)); | |
2116 | if (ret) | |
2117 | break; | |
2118 | } | |
2119 | ||
2120 | return ret; | |
2121 | } | |
2122 | #endif | |
2123 | ||
f6ac2354 CL |
2124 | /* |
2125 | * Some inline functions in vmstat.h depend on page_zone() | |
2126 | */ | |
2127 | #include <linux/vmstat.h> | |
2128 | ||
33dd4e0e | 2129 | static __always_inline void *lowmem_page_address(const struct page *page) |
1da177e4 | 2130 | { |
1dff8083 | 2131 | return page_to_virt(page); |
1da177e4 LT |
2132 | } |
2133 | ||
2134 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
2135 | #define HASHED_PAGE_VIRTUAL | |
2136 | #endif | |
2137 | ||
2138 | #if defined(WANT_PAGE_VIRTUAL) | |
f92f455f GU |
2139 | static inline void *page_address(const struct page *page) |
2140 | { | |
2141 | return page->virtual; | |
2142 | } | |
2143 | static inline void set_page_address(struct page *page, void *address) | |
2144 | { | |
2145 | page->virtual = address; | |
2146 | } | |
1da177e4 LT |
2147 | #define page_address_init() do { } while(0) |
2148 | #endif | |
2149 | ||
2150 | #if defined(HASHED_PAGE_VIRTUAL) | |
f9918794 | 2151 | void *page_address(const struct page *page); |
1da177e4 LT |
2152 | void set_page_address(struct page *page, void *virtual); |
2153 | void page_address_init(void); | |
2154 | #endif | |
2155 | ||
2156 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
2157 | #define page_address(page) lowmem_page_address(page) | |
2158 | #define set_page_address(page, address) do { } while(0) | |
2159 | #define page_address_init() do { } while(0) | |
2160 | #endif | |
2161 | ||
7d4203c1 VB |
2162 | static inline void *folio_address(const struct folio *folio) |
2163 | { | |
2164 | return page_address(&folio->page); | |
2165 | } | |
2166 | ||
e39155ea | 2167 | extern void *page_rmapping(struct page *page); |
f6ab1f7f HY |
2168 | extern pgoff_t __page_file_index(struct page *page); |
2169 | ||
1da177e4 LT |
2170 | /* |
2171 | * Return the pagecache index of the passed page. Regular pagecache pages | |
f6ab1f7f | 2172 | * use ->index whereas swapcache pages use swp_offset(->private) |
1da177e4 LT |
2173 | */ |
2174 | static inline pgoff_t page_index(struct page *page) | |
2175 | { | |
2176 | if (unlikely(PageSwapCache(page))) | |
f6ab1f7f | 2177 | return __page_file_index(page); |
1da177e4 LT |
2178 | return page->index; |
2179 | } | |
2180 | ||
2f064f34 MH |
2181 | /* |
2182 | * Return true only if the page has been allocated with | |
2183 | * ALLOC_NO_WATERMARKS and the low watermark was not | |
2184 | * met implying that the system is under some pressure. | |
2185 | */ | |
1d7bab6a | 2186 | static inline bool page_is_pfmemalloc(const struct page *page) |
2f064f34 MH |
2187 | { |
2188 | /* | |
c07aea3e MC |
2189 | * lru.next has bit 1 set if the page is allocated from the |
2190 | * pfmemalloc reserves. Callers may simply overwrite it if | |
2191 | * they do not need to preserve that information. | |
2f064f34 | 2192 | */ |
c07aea3e | 2193 | return (uintptr_t)page->lru.next & BIT(1); |
2f064f34 MH |
2194 | } |
2195 | ||
02d65d6f SK |
2196 | /* |
2197 | * Return true only if the folio has been allocated with | |
2198 | * ALLOC_NO_WATERMARKS and the low watermark was not | |
2199 | * met implying that the system is under some pressure. | |
2200 | */ | |
2201 | static inline bool folio_is_pfmemalloc(const struct folio *folio) | |
2202 | { | |
2203 | /* | |
2204 | * lru.next has bit 1 set if the page is allocated from the | |
2205 | * pfmemalloc reserves. Callers may simply overwrite it if | |
2206 | * they do not need to preserve that information. | |
2207 | */ | |
2208 | return (uintptr_t)folio->lru.next & BIT(1); | |
2209 | } | |
2210 | ||
2f064f34 MH |
2211 | /* |
2212 | * Only to be called by the page allocator on a freshly allocated | |
2213 | * page. | |
2214 | */ | |
2215 | static inline void set_page_pfmemalloc(struct page *page) | |
2216 | { | |
c07aea3e | 2217 | page->lru.next = (void *)BIT(1); |
2f064f34 MH |
2218 | } |
2219 | ||
2220 | static inline void clear_page_pfmemalloc(struct page *page) | |
2221 | { | |
c07aea3e | 2222 | page->lru.next = NULL; |
2f064f34 MH |
2223 | } |
2224 | ||
1c0fe6e3 NP |
2225 | /* |
2226 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | |
2227 | */ | |
2228 | extern void pagefault_out_of_memory(void); | |
2229 | ||
1da177e4 | 2230 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
ee6c400f | 2231 | #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1)) |
7b230db3 | 2232 | #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1)) |
1da177e4 | 2233 | |
ddd588b5 | 2234 | /* |
7bf02ea2 | 2235 | * Flags passed to show_mem() and show_free_areas() to suppress output in |
ddd588b5 DR |
2236 | * various contexts. |
2237 | */ | |
4b59e6c4 | 2238 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ |
ddd588b5 | 2239 | |
974f4367 MH |
2240 | extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx); |
2241 | static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask) | |
2242 | { | |
2243 | __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1); | |
2244 | } | |
1da177e4 | 2245 | |
21b85b09 MK |
2246 | /* |
2247 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
2248 | */ | |
2249 | struct zap_details { | |
2250 | struct folio *single_folio; /* Locked folio to be unmapped */ | |
2251 | bool even_cows; /* Zap COWed private pages too? */ | |
2252 | zap_flags_t zap_flags; /* Extra flags for zapping */ | |
2253 | }; | |
2254 | ||
2255 | /* | |
2256 | * Whether to drop the pte markers, for example, the uffd-wp information for | |
2257 | * file-backed memory. This should only be specified when we will completely | |
2258 | * drop the page in the mm, either by truncation or unmapping of the vma. By | |
2259 | * default, the flag is not set. | |
2260 | */ | |
2261 | #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0)) | |
04ada095 MK |
2262 | /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */ |
2263 | #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1)) | |
21b85b09 | 2264 | |
af7f588d MD |
2265 | #ifdef CONFIG_SCHED_MM_CID |
2266 | void sched_mm_cid_before_execve(struct task_struct *t); | |
2267 | void sched_mm_cid_after_execve(struct task_struct *t); | |
2268 | void sched_mm_cid_fork(struct task_struct *t); | |
2269 | void sched_mm_cid_exit_signals(struct task_struct *t); | |
2270 | static inline int task_mm_cid(struct task_struct *t) | |
2271 | { | |
2272 | return t->mm_cid; | |
2273 | } | |
2274 | #else | |
2275 | static inline void sched_mm_cid_before_execve(struct task_struct *t) { } | |
2276 | static inline void sched_mm_cid_after_execve(struct task_struct *t) { } | |
2277 | static inline void sched_mm_cid_fork(struct task_struct *t) { } | |
2278 | static inline void sched_mm_cid_exit_signals(struct task_struct *t) { } | |
2279 | static inline int task_mm_cid(struct task_struct *t) | |
2280 | { | |
2281 | /* | |
2282 | * Use the processor id as a fall-back when the mm cid feature is | |
2283 | * disabled. This provides functional per-cpu data structure accesses | |
2284 | * in user-space, althrough it won't provide the memory usage benefits. | |
2285 | */ | |
2286 | return raw_smp_processor_id(); | |
2287 | } | |
2288 | #endif | |
2289 | ||
710ec38b | 2290 | #ifdef CONFIG_MMU |
7f43add4 | 2291 | extern bool can_do_mlock(void); |
710ec38b AB |
2292 | #else |
2293 | static inline bool can_do_mlock(void) { return false; } | |
2294 | #endif | |
d7c9e99a AG |
2295 | extern int user_shm_lock(size_t, struct ucounts *); |
2296 | extern void user_shm_unlock(size_t, struct ucounts *); | |
1da177e4 | 2297 | |
318e9342 VMO |
2298 | struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr, |
2299 | pte_t pte); | |
25b2995a CH |
2300 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
2301 | pte_t pte); | |
28093f9f GS |
2302 | struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr, |
2303 | pmd_t pmd); | |
7e675137 | 2304 | |
27d036e3 LR |
2305 | void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
2306 | unsigned long size); | |
21b85b09 MK |
2307 | void zap_page_range_single(struct vm_area_struct *vma, unsigned long address, |
2308 | unsigned long size, struct zap_details *details); | |
e9adcfec MK |
2309 | static inline void zap_vma_pages(struct vm_area_struct *vma) |
2310 | { | |
2311 | zap_page_range_single(vma, vma->vm_start, | |
2312 | vma->vm_end - vma->vm_start, NULL); | |
2313 | } | |
763ecb03 LH |
2314 | void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt, |
2315 | struct vm_area_struct *start_vma, unsigned long start, | |
68f48381 | 2316 | unsigned long end, bool mm_wr_locked); |
e6473092 | 2317 | |
ac46d4f3 JG |
2318 | struct mmu_notifier_range; |
2319 | ||
42b77728 | 2320 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
3bf5ee95 | 2321 | unsigned long end, unsigned long floor, unsigned long ceiling); |
c78f4636 PX |
2322 | int |
2323 | copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma); | |
ff5c19ed | 2324 | int follow_pte(struct mm_struct *mm, unsigned long address, |
9fd6dad1 | 2325 | pte_t **ptepp, spinlock_t **ptlp); |
3b6748e2 JW |
2326 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
2327 | unsigned long *pfn); | |
d87fe660 | 2328 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
2329 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | |
28b2ee20 RR |
2330 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
2331 | void *buf, int len, int write); | |
1da177e4 | 2332 | |
7caef267 | 2333 | extern void truncate_pagecache(struct inode *inode, loff_t new); |
2c27c65e | 2334 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
90a80202 | 2335 | void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); |
623e3db9 | 2336 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); |
25718736 | 2337 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
83f78668 | 2338 | |
7ee1dd3f | 2339 | #ifdef CONFIG_MMU |
2b740303 | 2340 | extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
bce617ed PX |
2341 | unsigned long address, unsigned int flags, |
2342 | struct pt_regs *regs); | |
64019a2e | 2343 | extern int fixup_user_fault(struct mm_struct *mm, |
4a9e1cda DD |
2344 | unsigned long address, unsigned int fault_flags, |
2345 | bool *unlocked); | |
977fbdcd MW |
2346 | void unmap_mapping_pages(struct address_space *mapping, |
2347 | pgoff_t start, pgoff_t nr, bool even_cows); | |
2348 | void unmap_mapping_range(struct address_space *mapping, | |
2349 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
7ee1dd3f | 2350 | #else |
2b740303 | 2351 | static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
bce617ed PX |
2352 | unsigned long address, unsigned int flags, |
2353 | struct pt_regs *regs) | |
7ee1dd3f DH |
2354 | { |
2355 | /* should never happen if there's no MMU */ | |
2356 | BUG(); | |
2357 | return VM_FAULT_SIGBUS; | |
2358 | } | |
64019a2e | 2359 | static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address, |
4a9e1cda | 2360 | unsigned int fault_flags, bool *unlocked) |
5c723ba5 PZ |
2361 | { |
2362 | /* should never happen if there's no MMU */ | |
2363 | BUG(); | |
2364 | return -EFAULT; | |
2365 | } | |
977fbdcd MW |
2366 | static inline void unmap_mapping_pages(struct address_space *mapping, |
2367 | pgoff_t start, pgoff_t nr, bool even_cows) { } | |
2368 | static inline void unmap_mapping_range(struct address_space *mapping, | |
2369 | loff_t const holebegin, loff_t const holelen, int even_cows) { } | |
7ee1dd3f | 2370 | #endif |
f33ea7f4 | 2371 | |
977fbdcd MW |
2372 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
2373 | loff_t const holebegin, loff_t const holelen) | |
2374 | { | |
2375 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
2376 | } | |
2377 | ||
2378 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, | |
2379 | void *buf, int len, unsigned int gup_flags); | |
5ddd36b9 | 2380 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
6347e8d5 | 2381 | void *buf, int len, unsigned int gup_flags); |
d3f5ffca JH |
2382 | extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr, |
2383 | void *buf, int len, unsigned int gup_flags); | |
1da177e4 | 2384 | |
64019a2e | 2385 | long get_user_pages_remote(struct mm_struct *mm, |
1e987790 | 2386 | unsigned long start, unsigned long nr_pages, |
9beae1ea | 2387 | unsigned int gup_flags, struct page **pages, |
5b56d49f | 2388 | struct vm_area_struct **vmas, int *locked); |
64019a2e | 2389 | long pin_user_pages_remote(struct mm_struct *mm, |
eddb1c22 JH |
2390 | unsigned long start, unsigned long nr_pages, |
2391 | unsigned int gup_flags, struct page **pages, | |
2392 | struct vm_area_struct **vmas, int *locked); | |
c12d2da5 | 2393 | long get_user_pages(unsigned long start, unsigned long nr_pages, |
768ae309 | 2394 | unsigned int gup_flags, struct page **pages, |
cde70140 | 2395 | struct vm_area_struct **vmas); |
eddb1c22 JH |
2396 | long pin_user_pages(unsigned long start, unsigned long nr_pages, |
2397 | unsigned int gup_flags, struct page **pages, | |
2398 | struct vm_area_struct **vmas); | |
c12d2da5 | 2399 | long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages, |
c164154f | 2400 | struct page **pages, unsigned int gup_flags); |
91429023 JH |
2401 | long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages, |
2402 | struct page **pages, unsigned int gup_flags); | |
9a4e9f3b | 2403 | |
73b0140b IW |
2404 | int get_user_pages_fast(unsigned long start, int nr_pages, |
2405 | unsigned int gup_flags, struct page **pages); | |
eddb1c22 JH |
2406 | int pin_user_pages_fast(unsigned long start, int nr_pages, |
2407 | unsigned int gup_flags, struct page **pages); | |
8025e5dd | 2408 | |
79eb597c DJ |
2409 | int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc); |
2410 | int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc, | |
2411 | struct task_struct *task, bool bypass_rlim); | |
2412 | ||
18022c5d | 2413 | struct kvec; |
f3e8fccd | 2414 | struct page *get_dump_page(unsigned long addr); |
1da177e4 | 2415 | |
b5e84594 MWO |
2416 | bool folio_mark_dirty(struct folio *folio); |
2417 | bool set_page_dirty(struct page *page); | |
1da177e4 | 2418 | int set_page_dirty_lock(struct page *page); |
b9ea2515 | 2419 | |
a9090253 | 2420 | int get_cmdline(struct task_struct *task, char *buffer, int buflen); |
1da177e4 | 2421 | |
b6a2fea3 OW |
2422 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
2423 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
38a76013 ML |
2424 | unsigned long new_addr, unsigned long len, |
2425 | bool need_rmap_locks); | |
58705444 PX |
2426 | |
2427 | /* | |
2428 | * Flags used by change_protection(). For now we make it a bitmap so | |
2429 | * that we can pass in multiple flags just like parameters. However | |
2430 | * for now all the callers are only use one of the flags at the same | |
2431 | * time. | |
2432 | */ | |
64fe24a3 DH |
2433 | /* |
2434 | * Whether we should manually check if we can map individual PTEs writable, | |
2435 | * because something (e.g., COW, uffd-wp) blocks that from happening for all | |
2436 | * PTEs automatically in a writable mapping. | |
2437 | */ | |
2438 | #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0) | |
58705444 PX |
2439 | /* Whether this protection change is for NUMA hints */ |
2440 | #define MM_CP_PROT_NUMA (1UL << 1) | |
292924b2 PX |
2441 | /* Whether this change is for write protecting */ |
2442 | #define MM_CP_UFFD_WP (1UL << 2) /* do wp */ | |
2443 | #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */ | |
2444 | #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \ | |
2445 | MM_CP_UFFD_WP_RESOLVE) | |
58705444 | 2446 | |
eb309ec8 DH |
2447 | int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot); |
2448 | static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma) | |
2449 | { | |
2450 | /* | |
2451 | * We want to check manually if we can change individual PTEs writable | |
2452 | * if we can't do that automatically for all PTEs in a mapping. For | |
2453 | * private mappings, that's always the case when we have write | |
2454 | * permissions as we properly have to handle COW. | |
2455 | */ | |
2456 | if (vma->vm_flags & VM_SHARED) | |
2457 | return vma_wants_writenotify(vma, vma->vm_page_prot); | |
2458 | return !!(vma->vm_flags & VM_WRITE); | |
2459 | ||
2460 | } | |
6a56ccbc DH |
2461 | bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr, |
2462 | pte_t pte); | |
a79390f5 | 2463 | extern long change_protection(struct mmu_gather *tlb, |
4a18419f | 2464 | struct vm_area_struct *vma, unsigned long start, |
1ef488ed | 2465 | unsigned long end, unsigned long cp_flags); |
2286a691 LH |
2466 | extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb, |
2467 | struct vm_area_struct *vma, struct vm_area_struct **pprev, | |
2468 | unsigned long start, unsigned long end, unsigned long newflags); | |
1da177e4 | 2469 | |
465a454f PZ |
2470 | /* |
2471 | * doesn't attempt to fault and will return short. | |
2472 | */ | |
dadbb612 SJ |
2473 | int get_user_pages_fast_only(unsigned long start, int nr_pages, |
2474 | unsigned int gup_flags, struct page **pages); | |
dadbb612 SJ |
2475 | |
2476 | static inline bool get_user_page_fast_only(unsigned long addr, | |
2477 | unsigned int gup_flags, struct page **pagep) | |
2478 | { | |
2479 | return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1; | |
2480 | } | |
d559db08 KH |
2481 | /* |
2482 | * per-process(per-mm_struct) statistics. | |
2483 | */ | |
d559db08 KH |
2484 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) |
2485 | { | |
f1a79412 | 2486 | return percpu_counter_read_positive(&mm->rss_stat[member]); |
69c97823 | 2487 | } |
d559db08 | 2488 | |
f1a79412 | 2489 | void mm_trace_rss_stat(struct mm_struct *mm, int member); |
b3d1411b | 2490 | |
d559db08 KH |
2491 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) |
2492 | { | |
f1a79412 | 2493 | percpu_counter_add(&mm->rss_stat[member], value); |
b3d1411b | 2494 | |
f1a79412 | 2495 | mm_trace_rss_stat(mm, member); |
d559db08 KH |
2496 | } |
2497 | ||
2498 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
2499 | { | |
f1a79412 | 2500 | percpu_counter_inc(&mm->rss_stat[member]); |
b3d1411b | 2501 | |
f1a79412 | 2502 | mm_trace_rss_stat(mm, member); |
d559db08 KH |
2503 | } |
2504 | ||
2505 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
2506 | { | |
f1a79412 | 2507 | percpu_counter_dec(&mm->rss_stat[member]); |
b3d1411b | 2508 | |
f1a79412 | 2509 | mm_trace_rss_stat(mm, member); |
d559db08 KH |
2510 | } |
2511 | ||
eca56ff9 JM |
2512 | /* Optimized variant when page is already known not to be PageAnon */ |
2513 | static inline int mm_counter_file(struct page *page) | |
2514 | { | |
2515 | if (PageSwapBacked(page)) | |
2516 | return MM_SHMEMPAGES; | |
2517 | return MM_FILEPAGES; | |
2518 | } | |
2519 | ||
2520 | static inline int mm_counter(struct page *page) | |
2521 | { | |
2522 | if (PageAnon(page)) | |
2523 | return MM_ANONPAGES; | |
2524 | return mm_counter_file(page); | |
2525 | } | |
2526 | ||
d559db08 KH |
2527 | static inline unsigned long get_mm_rss(struct mm_struct *mm) |
2528 | { | |
2529 | return get_mm_counter(mm, MM_FILEPAGES) + | |
eca56ff9 JM |
2530 | get_mm_counter(mm, MM_ANONPAGES) + |
2531 | get_mm_counter(mm, MM_SHMEMPAGES); | |
d559db08 KH |
2532 | } |
2533 | ||
2534 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | |
2535 | { | |
2536 | return max(mm->hiwater_rss, get_mm_rss(mm)); | |
2537 | } | |
2538 | ||
2539 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | |
2540 | { | |
2541 | return max(mm->hiwater_vm, mm->total_vm); | |
2542 | } | |
2543 | ||
2544 | static inline void update_hiwater_rss(struct mm_struct *mm) | |
2545 | { | |
2546 | unsigned long _rss = get_mm_rss(mm); | |
2547 | ||
2548 | if ((mm)->hiwater_rss < _rss) | |
2549 | (mm)->hiwater_rss = _rss; | |
2550 | } | |
2551 | ||
2552 | static inline void update_hiwater_vm(struct mm_struct *mm) | |
2553 | { | |
2554 | if (mm->hiwater_vm < mm->total_vm) | |
2555 | mm->hiwater_vm = mm->total_vm; | |
2556 | } | |
2557 | ||
695f0559 PC |
2558 | static inline void reset_mm_hiwater_rss(struct mm_struct *mm) |
2559 | { | |
2560 | mm->hiwater_rss = get_mm_rss(mm); | |
2561 | } | |
2562 | ||
d559db08 KH |
2563 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, |
2564 | struct mm_struct *mm) | |
2565 | { | |
2566 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | |
2567 | ||
2568 | if (*maxrss < hiwater_rss) | |
2569 | *maxrss = hiwater_rss; | |
2570 | } | |
2571 | ||
53bddb4e | 2572 | #if defined(SPLIT_RSS_COUNTING) |
05af2e10 | 2573 | void sync_mm_rss(struct mm_struct *mm); |
53bddb4e | 2574 | #else |
05af2e10 | 2575 | static inline void sync_mm_rss(struct mm_struct *mm) |
53bddb4e KH |
2576 | { |
2577 | } | |
2578 | #endif | |
465a454f | 2579 | |
78e7c5af AK |
2580 | #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL |
2581 | static inline int pte_special(pte_t pte) | |
2582 | { | |
2583 | return 0; | |
2584 | } | |
2585 | ||
2586 | static inline pte_t pte_mkspecial(pte_t pte) | |
2587 | { | |
2588 | return pte; | |
2589 | } | |
2590 | #endif | |
2591 | ||
17596731 | 2592 | #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP |
3565fce3 DW |
2593 | static inline int pte_devmap(pte_t pte) |
2594 | { | |
2595 | return 0; | |
2596 | } | |
2597 | #endif | |
2598 | ||
25ca1d6c NK |
2599 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
2600 | spinlock_t **ptl); | |
2601 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
2602 | spinlock_t **ptl) | |
2603 | { | |
2604 | pte_t *ptep; | |
2605 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | |
2606 | return ptep; | |
2607 | } | |
c9cfcddf | 2608 | |
c2febafc KS |
2609 | #ifdef __PAGETABLE_P4D_FOLDED |
2610 | static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, | |
2611 | unsigned long address) | |
2612 | { | |
2613 | return 0; | |
2614 | } | |
2615 | #else | |
2616 | int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); | |
2617 | #endif | |
2618 | ||
b4e98d9a | 2619 | #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU) |
c2febafc | 2620 | static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, |
5f22df00 NP |
2621 | unsigned long address) |
2622 | { | |
2623 | return 0; | |
2624 | } | |
b4e98d9a KS |
2625 | static inline void mm_inc_nr_puds(struct mm_struct *mm) {} |
2626 | static inline void mm_dec_nr_puds(struct mm_struct *mm) {} | |
2627 | ||
5f22df00 | 2628 | #else |
c2febafc | 2629 | int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address); |
b4e98d9a | 2630 | |
b4e98d9a KS |
2631 | static inline void mm_inc_nr_puds(struct mm_struct *mm) |
2632 | { | |
6d212db1 MS |
2633 | if (mm_pud_folded(mm)) |
2634 | return; | |
af5b0f6a | 2635 | atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
b4e98d9a KS |
2636 | } |
2637 | ||
2638 | static inline void mm_dec_nr_puds(struct mm_struct *mm) | |
2639 | { | |
6d212db1 MS |
2640 | if (mm_pud_folded(mm)) |
2641 | return; | |
af5b0f6a | 2642 | atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
b4e98d9a | 2643 | } |
5f22df00 NP |
2644 | #endif |
2645 | ||
2d2f5119 | 2646 | #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU) |
5f22df00 NP |
2647 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
2648 | unsigned long address) | |
2649 | { | |
2650 | return 0; | |
2651 | } | |
dc6c9a35 | 2652 | |
dc6c9a35 KS |
2653 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) {} |
2654 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) {} | |
2655 | ||
5f22df00 | 2656 | #else |
1bb3630e | 2657 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
dc6c9a35 | 2658 | |
dc6c9a35 KS |
2659 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) |
2660 | { | |
6d212db1 MS |
2661 | if (mm_pmd_folded(mm)) |
2662 | return; | |
af5b0f6a | 2663 | atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
dc6c9a35 KS |
2664 | } |
2665 | ||
2666 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) | |
2667 | { | |
6d212db1 MS |
2668 | if (mm_pmd_folded(mm)) |
2669 | return; | |
af5b0f6a | 2670 | atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
dc6c9a35 | 2671 | } |
5f22df00 NP |
2672 | #endif |
2673 | ||
c4812909 | 2674 | #ifdef CONFIG_MMU |
af5b0f6a | 2675 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) |
c4812909 | 2676 | { |
af5b0f6a | 2677 | atomic_long_set(&mm->pgtables_bytes, 0); |
c4812909 KS |
2678 | } |
2679 | ||
af5b0f6a | 2680 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) |
c4812909 | 2681 | { |
af5b0f6a | 2682 | return atomic_long_read(&mm->pgtables_bytes); |
c4812909 KS |
2683 | } |
2684 | ||
2685 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) | |
2686 | { | |
af5b0f6a | 2687 | atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
c4812909 KS |
2688 | } |
2689 | ||
2690 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) | |
2691 | { | |
af5b0f6a | 2692 | atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
c4812909 KS |
2693 | } |
2694 | #else | |
c4812909 | 2695 | |
af5b0f6a KS |
2696 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {} |
2697 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) | |
c4812909 KS |
2698 | { |
2699 | return 0; | |
2700 | } | |
2701 | ||
2702 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) {} | |
2703 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) {} | |
2704 | #endif | |
2705 | ||
4cf58924 JFG |
2706 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd); |
2707 | int __pte_alloc_kernel(pmd_t *pmd); | |
1bb3630e | 2708 | |
f949286c MR |
2709 | #if defined(CONFIG_MMU) |
2710 | ||
c2febafc KS |
2711 | static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd, |
2712 | unsigned long address) | |
2713 | { | |
2714 | return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ? | |
2715 | NULL : p4d_offset(pgd, address); | |
2716 | } | |
2717 | ||
2718 | static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d, | |
2719 | unsigned long address) | |
1da177e4 | 2720 | { |
c2febafc KS |
2721 | return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ? |
2722 | NULL : pud_offset(p4d, address); | |
1da177e4 | 2723 | } |
d8626138 | 2724 | |
1da177e4 LT |
2725 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
2726 | { | |
1bb3630e HD |
2727 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
2728 | NULL: pmd_offset(pud, address); | |
1da177e4 | 2729 | } |
f949286c | 2730 | #endif /* CONFIG_MMU */ |
1bb3630e | 2731 | |
57c1ffce | 2732 | #if USE_SPLIT_PTE_PTLOCKS |
597d795a | 2733 | #if ALLOC_SPLIT_PTLOCKS |
b35f1819 | 2734 | void __init ptlock_cache_init(void); |
539edb58 PZ |
2735 | extern bool ptlock_alloc(struct page *page); |
2736 | extern void ptlock_free(struct page *page); | |
2737 | ||
2738 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
2739 | { | |
2740 | return page->ptl; | |
2741 | } | |
597d795a | 2742 | #else /* ALLOC_SPLIT_PTLOCKS */ |
b35f1819 KS |
2743 | static inline void ptlock_cache_init(void) |
2744 | { | |
2745 | } | |
2746 | ||
49076ec2 KS |
2747 | static inline bool ptlock_alloc(struct page *page) |
2748 | { | |
49076ec2 KS |
2749 | return true; |
2750 | } | |
539edb58 | 2751 | |
49076ec2 KS |
2752 | static inline void ptlock_free(struct page *page) |
2753 | { | |
49076ec2 KS |
2754 | } |
2755 | ||
2756 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
2757 | { | |
539edb58 | 2758 | return &page->ptl; |
49076ec2 | 2759 | } |
597d795a | 2760 | #endif /* ALLOC_SPLIT_PTLOCKS */ |
49076ec2 KS |
2761 | |
2762 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
2763 | { | |
2764 | return ptlock_ptr(pmd_page(*pmd)); | |
2765 | } | |
2766 | ||
2767 | static inline bool ptlock_init(struct page *page) | |
2768 | { | |
2769 | /* | |
2770 | * prep_new_page() initialize page->private (and therefore page->ptl) | |
2771 | * with 0. Make sure nobody took it in use in between. | |
2772 | * | |
2773 | * It can happen if arch try to use slab for page table allocation: | |
1d798ca3 | 2774 | * slab code uses page->slab_cache, which share storage with page->ptl. |
49076ec2 | 2775 | */ |
309381fe | 2776 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); |
49076ec2 KS |
2777 | if (!ptlock_alloc(page)) |
2778 | return false; | |
2779 | spin_lock_init(ptlock_ptr(page)); | |
2780 | return true; | |
2781 | } | |
2782 | ||
57c1ffce | 2783 | #else /* !USE_SPLIT_PTE_PTLOCKS */ |
4c21e2f2 HD |
2784 | /* |
2785 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
2786 | */ | |
49076ec2 KS |
2787 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2788 | { | |
2789 | return &mm->page_table_lock; | |
2790 | } | |
b35f1819 | 2791 | static inline void ptlock_cache_init(void) {} |
49076ec2 | 2792 | static inline bool ptlock_init(struct page *page) { return true; } |
9e247bab | 2793 | static inline void ptlock_free(struct page *page) {} |
57c1ffce | 2794 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
4c21e2f2 | 2795 | |
b4ed71f5 | 2796 | static inline bool pgtable_pte_page_ctor(struct page *page) |
2f569afd | 2797 | { |
706874e9 VD |
2798 | if (!ptlock_init(page)) |
2799 | return false; | |
1d40a5ea | 2800 | __SetPageTable(page); |
f0c0c115 | 2801 | inc_lruvec_page_state(page, NR_PAGETABLE); |
706874e9 | 2802 | return true; |
2f569afd MS |
2803 | } |
2804 | ||
b4ed71f5 | 2805 | static inline void pgtable_pte_page_dtor(struct page *page) |
2f569afd | 2806 | { |
9e247bab | 2807 | ptlock_free(page); |
1d40a5ea | 2808 | __ClearPageTable(page); |
f0c0c115 | 2809 | dec_lruvec_page_state(page, NR_PAGETABLE); |
2f569afd MS |
2810 | } |
2811 | ||
c74df32c HD |
2812 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
2813 | ({ \ | |
4c21e2f2 | 2814 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
c74df32c HD |
2815 | pte_t *__pte = pte_offset_map(pmd, address); \ |
2816 | *(ptlp) = __ptl; \ | |
2817 | spin_lock(__ptl); \ | |
2818 | __pte; \ | |
2819 | }) | |
2820 | ||
2821 | #define pte_unmap_unlock(pte, ptl) do { \ | |
2822 | spin_unlock(ptl); \ | |
2823 | pte_unmap(pte); \ | |
2824 | } while (0) | |
2825 | ||
4cf58924 | 2826 | #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd)) |
3ed3a4f0 KS |
2827 | |
2828 | #define pte_alloc_map(mm, pmd, address) \ | |
4cf58924 | 2829 | (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address)) |
1bb3630e | 2830 | |
c74df32c | 2831 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
4cf58924 | 2832 | (pte_alloc(mm, pmd) ? \ |
3ed3a4f0 | 2833 | NULL : pte_offset_map_lock(mm, pmd, address, ptlp)) |
c74df32c | 2834 | |
1bb3630e | 2835 | #define pte_alloc_kernel(pmd, address) \ |
4cf58924 | 2836 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \ |
1bb3630e | 2837 | NULL: pte_offset_kernel(pmd, address)) |
1da177e4 | 2838 | |
e009bb30 KS |
2839 | #if USE_SPLIT_PMD_PTLOCKS |
2840 | ||
7e25de77 | 2841 | static inline struct page *pmd_pgtable_page(pmd_t *pmd) |
634391ac MS |
2842 | { |
2843 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); | |
2844 | return virt_to_page((void *)((unsigned long) pmd & mask)); | |
2845 | } | |
2846 | ||
e009bb30 KS |
2847 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2848 | { | |
373dfda2 | 2849 | return ptlock_ptr(pmd_pgtable_page(pmd)); |
e009bb30 KS |
2850 | } |
2851 | ||
b2b29d6d | 2852 | static inline bool pmd_ptlock_init(struct page *page) |
e009bb30 | 2853 | { |
e009bb30 KS |
2854 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
2855 | page->pmd_huge_pte = NULL; | |
2856 | #endif | |
49076ec2 | 2857 | return ptlock_init(page); |
e009bb30 KS |
2858 | } |
2859 | ||
b2b29d6d | 2860 | static inline void pmd_ptlock_free(struct page *page) |
e009bb30 KS |
2861 | { |
2862 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
309381fe | 2863 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); |
e009bb30 | 2864 | #endif |
49076ec2 | 2865 | ptlock_free(page); |
e009bb30 KS |
2866 | } |
2867 | ||
373dfda2 | 2868 | #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte) |
e009bb30 KS |
2869 | |
2870 | #else | |
2871 | ||
9a86cb7b KS |
2872 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
2873 | { | |
2874 | return &mm->page_table_lock; | |
2875 | } | |
2876 | ||
b2b29d6d MW |
2877 | static inline bool pmd_ptlock_init(struct page *page) { return true; } |
2878 | static inline void pmd_ptlock_free(struct page *page) {} | |
e009bb30 | 2879 | |
c389a250 | 2880 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) |
9a86cb7b | 2881 | |
e009bb30 KS |
2882 | #endif |
2883 | ||
9a86cb7b KS |
2884 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) |
2885 | { | |
2886 | spinlock_t *ptl = pmd_lockptr(mm, pmd); | |
2887 | spin_lock(ptl); | |
2888 | return ptl; | |
2889 | } | |
2890 | ||
b2b29d6d MW |
2891 | static inline bool pgtable_pmd_page_ctor(struct page *page) |
2892 | { | |
2893 | if (!pmd_ptlock_init(page)) | |
2894 | return false; | |
2895 | __SetPageTable(page); | |
f0c0c115 | 2896 | inc_lruvec_page_state(page, NR_PAGETABLE); |
b2b29d6d MW |
2897 | return true; |
2898 | } | |
2899 | ||
2900 | static inline void pgtable_pmd_page_dtor(struct page *page) | |
2901 | { | |
2902 | pmd_ptlock_free(page); | |
2903 | __ClearPageTable(page); | |
f0c0c115 | 2904 | dec_lruvec_page_state(page, NR_PAGETABLE); |
b2b29d6d MW |
2905 | } |
2906 | ||
a00cc7d9 MW |
2907 | /* |
2908 | * No scalability reason to split PUD locks yet, but follow the same pattern | |
2909 | * as the PMD locks to make it easier if we decide to. The VM should not be | |
2910 | * considered ready to switch to split PUD locks yet; there may be places | |
2911 | * which need to be converted from page_table_lock. | |
2912 | */ | |
2913 | static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud) | |
2914 | { | |
2915 | return &mm->page_table_lock; | |
2916 | } | |
2917 | ||
2918 | static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud) | |
2919 | { | |
2920 | spinlock_t *ptl = pud_lockptr(mm, pud); | |
2921 | ||
2922 | spin_lock(ptl); | |
2923 | return ptl; | |
2924 | } | |
62906027 | 2925 | |
a00cc7d9 | 2926 | extern void __init pagecache_init(void); |
49a7f04a DH |
2927 | extern void free_initmem(void); |
2928 | ||
69afade7 JL |
2929 | /* |
2930 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) | |
2931 | * into the buddy system. The freed pages will be poisoned with pattern | |
dbe67df4 | 2932 | * "poison" if it's within range [0, UCHAR_MAX]. |
69afade7 JL |
2933 | * Return pages freed into the buddy system. |
2934 | */ | |
11199692 | 2935 | extern unsigned long free_reserved_area(void *start, void *end, |
e5cb113f | 2936 | int poison, const char *s); |
c3d5f5f0 | 2937 | |
c3d5f5f0 | 2938 | extern void adjust_managed_page_count(struct page *page, long count); |
69afade7 | 2939 | |
4b50bcc7 | 2940 | extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end); |
92923ca3 | 2941 | |
69afade7 | 2942 | /* Free the reserved page into the buddy system, so it gets managed. */ |
a0cd7a7c | 2943 | static inline void free_reserved_page(struct page *page) |
69afade7 JL |
2944 | { |
2945 | ClearPageReserved(page); | |
2946 | init_page_count(page); | |
2947 | __free_page(page); | |
69afade7 JL |
2948 | adjust_managed_page_count(page, 1); |
2949 | } | |
a0cd7a7c | 2950 | #define free_highmem_page(page) free_reserved_page(page) |
69afade7 JL |
2951 | |
2952 | static inline void mark_page_reserved(struct page *page) | |
2953 | { | |
2954 | SetPageReserved(page); | |
2955 | adjust_managed_page_count(page, -1); | |
2956 | } | |
2957 | ||
2958 | /* | |
2959 | * Default method to free all the __init memory into the buddy system. | |
dbe67df4 JL |
2960 | * The freed pages will be poisoned with pattern "poison" if it's within |
2961 | * range [0, UCHAR_MAX]. | |
2962 | * Return pages freed into the buddy system. | |
69afade7 JL |
2963 | */ |
2964 | static inline unsigned long free_initmem_default(int poison) | |
2965 | { | |
2966 | extern char __init_begin[], __init_end[]; | |
2967 | ||
11199692 | 2968 | return free_reserved_area(&__init_begin, &__init_end, |
c5a54c70 | 2969 | poison, "unused kernel image (initmem)"); |
69afade7 JL |
2970 | } |
2971 | ||
7ee3d4e8 JL |
2972 | static inline unsigned long get_num_physpages(void) |
2973 | { | |
2974 | int nid; | |
2975 | unsigned long phys_pages = 0; | |
2976 | ||
2977 | for_each_online_node(nid) | |
2978 | phys_pages += node_present_pages(nid); | |
2979 | ||
2980 | return phys_pages; | |
2981 | } | |
2982 | ||
c713216d | 2983 | /* |
3f08a302 | 2984 | * Using memblock node mappings, an architecture may initialise its |
bc9331a1 MR |
2985 | * zones, allocate the backing mem_map and account for memory holes in an |
2986 | * architecture independent manner. | |
c713216d MG |
2987 | * |
2988 | * An architecture is expected to register range of page frames backed by | |
0ee332c1 | 2989 | * physical memory with memblock_add[_node]() before calling |
9691a071 | 2990 | * free_area_init() passing in the PFN each zone ends at. At a basic |
c713216d MG |
2991 | * usage, an architecture is expected to do something like |
2992 | * | |
2993 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
2994 | * max_highmem_pfn}; | |
2995 | * for_each_valid_physical_page_range() | |
952eea9b | 2996 | * memblock_add_node(base, size, nid, MEMBLOCK_NONE) |
9691a071 | 2997 | * free_area_init(max_zone_pfns); |
c713216d | 2998 | */ |
9691a071 | 2999 | void free_area_init(unsigned long *max_zone_pfn); |
1e01979c | 3000 | unsigned long node_map_pfn_alignment(void); |
32996250 YL |
3001 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, |
3002 | unsigned long end_pfn); | |
c713216d MG |
3003 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
3004 | unsigned long end_pfn); | |
3005 | extern void get_pfn_range_for_nid(unsigned int nid, | |
3006 | unsigned long *start_pfn, unsigned long *end_pfn); | |
f2dbcfa7 | 3007 | |
a9ee6cf5 | 3008 | #ifndef CONFIG_NUMA |
6f24fbd3 | 3009 | static inline int early_pfn_to_nid(unsigned long pfn) |
f2dbcfa7 KH |
3010 | { |
3011 | return 0; | |
3012 | } | |
3013 | #else | |
3014 | /* please see mm/page_alloc.c */ | |
3015 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | |
f2dbcfa7 KH |
3016 | #endif |
3017 | ||
0e0b864e | 3018 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
ab28cb6e | 3019 | extern void memmap_init_range(unsigned long, int, unsigned long, |
dc2da7b4 BH |
3020 | unsigned long, unsigned long, enum meminit_context, |
3021 | struct vmem_altmap *, int migratetype); | |
bc75d33f | 3022 | extern void setup_per_zone_wmarks(void); |
bd3400ea | 3023 | extern void calculate_min_free_kbytes(void); |
1b79acc9 | 3024 | extern int __meminit init_per_zone_wmark_min(void); |
1da177e4 | 3025 | extern void mem_init(void); |
8feae131 | 3026 | extern void __init mmap_init(void); |
974f4367 MH |
3027 | |
3028 | extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx); | |
3029 | static inline void show_mem(unsigned int flags, nodemask_t *nodemask) | |
3030 | { | |
3031 | __show_mem(flags, nodemask, MAX_NR_ZONES - 1); | |
3032 | } | |
d02bd27b | 3033 | extern long si_mem_available(void); |
1da177e4 LT |
3034 | extern void si_meminfo(struct sysinfo * val); |
3035 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
f6f34b43 SD |
3036 | #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES |
3037 | extern unsigned long arch_reserved_kernel_pages(void); | |
3038 | #endif | |
1da177e4 | 3039 | |
a8e99259 MH |
3040 | extern __printf(3, 4) |
3041 | void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...); | |
a238ab5b | 3042 | |
e7c8d5c9 | 3043 | extern void setup_per_cpu_pageset(void); |
e7c8d5c9 | 3044 | |
75f7ad8e PS |
3045 | /* page_alloc.c */ |
3046 | extern int min_free_kbytes; | |
1c30844d | 3047 | extern int watermark_boost_factor; |
795ae7a0 | 3048 | extern int watermark_scale_factor; |
75f7ad8e | 3049 | |
8feae131 | 3050 | /* nommu.c */ |
33e5d769 | 3051 | extern atomic_long_t mmap_pages_allocated; |
7e660872 | 3052 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
8feae131 | 3053 | |
6b2dbba8 | 3054 | /* interval_tree.c */ |
6b2dbba8 | 3055 | void vma_interval_tree_insert(struct vm_area_struct *node, |
f808c13f | 3056 | struct rb_root_cached *root); |
9826a516 ML |
3057 | void vma_interval_tree_insert_after(struct vm_area_struct *node, |
3058 | struct vm_area_struct *prev, | |
f808c13f | 3059 | struct rb_root_cached *root); |
6b2dbba8 | 3060 | void vma_interval_tree_remove(struct vm_area_struct *node, |
f808c13f DB |
3061 | struct rb_root_cached *root); |
3062 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root, | |
6b2dbba8 ML |
3063 | unsigned long start, unsigned long last); |
3064 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, | |
3065 | unsigned long start, unsigned long last); | |
3066 | ||
3067 | #define vma_interval_tree_foreach(vma, root, start, last) \ | |
3068 | for (vma = vma_interval_tree_iter_first(root, start, last); \ | |
3069 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) | |
1da177e4 | 3070 | |
bf181b9f | 3071 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, |
f808c13f | 3072 | struct rb_root_cached *root); |
bf181b9f | 3073 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, |
f808c13f DB |
3074 | struct rb_root_cached *root); |
3075 | struct anon_vma_chain * | |
3076 | anon_vma_interval_tree_iter_first(struct rb_root_cached *root, | |
3077 | unsigned long start, unsigned long last); | |
bf181b9f ML |
3078 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( |
3079 | struct anon_vma_chain *node, unsigned long start, unsigned long last); | |
ed8ea815 ML |
3080 | #ifdef CONFIG_DEBUG_VM_RB |
3081 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); | |
3082 | #endif | |
bf181b9f ML |
3083 | |
3084 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ | |
3085 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ | |
3086 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) | |
3087 | ||
1da177e4 | 3088 | /* mmap.c */ |
34b4e4aa | 3089 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
7c9813e8 LH |
3090 | extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma, |
3091 | unsigned long start, unsigned long end, pgoff_t pgoff, | |
3092 | struct vm_area_struct *next); | |
cf51e86d LH |
3093 | extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma, |
3094 | unsigned long start, unsigned long end, pgoff_t pgoff); | |
9760ebff | 3095 | extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi, |
f2ebfe43 LH |
3096 | struct mm_struct *, struct vm_area_struct *prev, unsigned long addr, |
3097 | unsigned long end, unsigned long vm_flags, struct anon_vma *, | |
3098 | struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx, | |
3099 | struct anon_vma_name *); | |
1da177e4 | 3100 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
9760ebff LH |
3101 | extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *, |
3102 | unsigned long addr, int new_below); | |
3103 | extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *, | |
3104 | unsigned long addr, int new_below); | |
1da177e4 | 3105 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
a8fb5618 | 3106 | extern void unlink_file_vma(struct vm_area_struct *); |
1da177e4 | 3107 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
38a76013 ML |
3108 | unsigned long addr, unsigned long len, pgoff_t pgoff, |
3109 | bool *need_rmap_locks); | |
1da177e4 | 3110 | extern void exit_mmap(struct mm_struct *); |
925d1c40 | 3111 | |
9c599024 CG |
3112 | static inline int check_data_rlimit(unsigned long rlim, |
3113 | unsigned long new, | |
3114 | unsigned long start, | |
3115 | unsigned long end_data, | |
3116 | unsigned long start_data) | |
3117 | { | |
3118 | if (rlim < RLIM_INFINITY) { | |
3119 | if (((new - start) + (end_data - start_data)) > rlim) | |
3120 | return -ENOSPC; | |
3121 | } | |
3122 | ||
3123 | return 0; | |
3124 | } | |
3125 | ||
7906d00c AA |
3126 | extern int mm_take_all_locks(struct mm_struct *mm); |
3127 | extern void mm_drop_all_locks(struct mm_struct *mm); | |
3128 | ||
fe69d560 | 3129 | extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); |
35d7bdc8 | 3130 | extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); |
38646013 | 3131 | extern struct file *get_mm_exe_file(struct mm_struct *mm); |
cd81a917 | 3132 | extern struct file *get_task_exe_file(struct task_struct *task); |
925d1c40 | 3133 | |
84638335 KK |
3134 | extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages); |
3135 | extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages); | |
3136 | ||
2eefd878 DS |
3137 | extern bool vma_is_special_mapping(const struct vm_area_struct *vma, |
3138 | const struct vm_special_mapping *sm); | |
3935ed6a SS |
3139 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, |
3140 | unsigned long addr, unsigned long len, | |
a62c34bd AL |
3141 | unsigned long flags, |
3142 | const struct vm_special_mapping *spec); | |
3143 | /* This is an obsolete alternative to _install_special_mapping. */ | |
fa5dc22f RM |
3144 | extern int install_special_mapping(struct mm_struct *mm, |
3145 | unsigned long addr, unsigned long len, | |
3146 | unsigned long flags, struct page **pages); | |
1da177e4 | 3147 | |
649775be | 3148 | unsigned long randomize_stack_top(unsigned long stack_top); |
5ad7dd88 | 3149 | unsigned long randomize_page(unsigned long start, unsigned long range); |
649775be | 3150 | |
1da177e4 LT |
3151 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
3152 | ||
0165ab44 | 3153 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
897ab3e0 MR |
3154 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff, |
3155 | struct list_head *uf); | |
1fcfd8db | 3156 | extern unsigned long do_mmap(struct file *file, unsigned long addr, |
bebeb3d6 | 3157 | unsigned long len, unsigned long prot, unsigned long flags, |
45e55300 | 3158 | unsigned long pgoff, unsigned long *populate, struct list_head *uf); |
183654ce | 3159 | extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm, |
11f9a21a LH |
3160 | unsigned long start, size_t len, struct list_head *uf, |
3161 | bool downgrade); | |
897ab3e0 MR |
3162 | extern int do_munmap(struct mm_struct *, unsigned long, size_t, |
3163 | struct list_head *uf); | |
0726b01e | 3164 | extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior); |
1da177e4 | 3165 | |
bebeb3d6 | 3166 | #ifdef CONFIG_MMU |
27b26701 LH |
3167 | extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma, |
3168 | unsigned long start, unsigned long end, | |
3169 | struct list_head *uf, bool downgrade); | |
bebeb3d6 ML |
3170 | extern int __mm_populate(unsigned long addr, unsigned long len, |
3171 | int ignore_errors); | |
3172 | static inline void mm_populate(unsigned long addr, unsigned long len) | |
3173 | { | |
3174 | /* Ignore errors */ | |
3175 | (void) __mm_populate(addr, len, 1); | |
3176 | } | |
3177 | #else | |
3178 | static inline void mm_populate(unsigned long addr, unsigned long len) {} | |
3179 | #endif | |
3180 | ||
e4eb1ff6 | 3181 | /* These take the mm semaphore themselves */ |
5d22fc25 | 3182 | extern int __must_check vm_brk(unsigned long, unsigned long); |
16e72e9b | 3183 | extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long); |
bfce281c | 3184 | extern int vm_munmap(unsigned long, size_t); |
9fbeb5ab | 3185 | extern unsigned long __must_check vm_mmap(struct file *, unsigned long, |
6be5ceb0 LT |
3186 | unsigned long, unsigned long, |
3187 | unsigned long, unsigned long); | |
1da177e4 | 3188 | |
db4fbfb9 ML |
3189 | struct vm_unmapped_area_info { |
3190 | #define VM_UNMAPPED_AREA_TOPDOWN 1 | |
3191 | unsigned long flags; | |
3192 | unsigned long length; | |
3193 | unsigned long low_limit; | |
3194 | unsigned long high_limit; | |
3195 | unsigned long align_mask; | |
3196 | unsigned long align_offset; | |
3197 | }; | |
3198 | ||
baceaf1c | 3199 | extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info); |
db4fbfb9 | 3200 | |
85821aab | 3201 | /* truncate.c */ |
1da177e4 | 3202 | extern void truncate_inode_pages(struct address_space *, loff_t); |
d7339071 HR |
3203 | extern void truncate_inode_pages_range(struct address_space *, |
3204 | loff_t lstart, loff_t lend); | |
91b0abe3 | 3205 | extern void truncate_inode_pages_final(struct address_space *); |
1da177e4 LT |
3206 | |
3207 | /* generic vm_area_ops exported for stackable file systems */ | |
2bcd6454 | 3208 | extern vm_fault_t filemap_fault(struct vm_fault *vmf); |
f9ce0be7 | 3209 | extern vm_fault_t filemap_map_pages(struct vm_fault *vmf, |
bae473a4 | 3210 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
2bcd6454 | 3211 | extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf); |
1da177e4 | 3212 | |
1be7107f | 3213 | extern unsigned long stack_guard_gap; |
d05f3169 | 3214 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ |
46dea3d0 | 3215 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
d05f3169 | 3216 | |
11192337 | 3217 | /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */ |
d05f3169 MH |
3218 | extern int expand_downwards(struct vm_area_struct *vma, |
3219 | unsigned long address); | |
8ca3eb08 | 3220 | #if VM_GROWSUP |
46dea3d0 | 3221 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
8ca3eb08 | 3222 | #else |
fee7e49d | 3223 | #define expand_upwards(vma, address) (0) |
9ab88515 | 3224 | #endif |
1da177e4 LT |
3225 | |
3226 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
3227 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
3228 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
3229 | struct vm_area_struct **pprev); | |
3230 | ||
abdba2dd LH |
3231 | /* |
3232 | * Look up the first VMA which intersects the interval [start_addr, end_addr) | |
3233 | * NULL if none. Assume start_addr < end_addr. | |
ce6d42f2 | 3234 | */ |
ce6d42f2 | 3235 | struct vm_area_struct *find_vma_intersection(struct mm_struct *mm, |
abdba2dd | 3236 | unsigned long start_addr, unsigned long end_addr); |
1da177e4 | 3237 | |
ce6d42f2 LH |
3238 | /** |
3239 | * vma_lookup() - Find a VMA at a specific address | |
3240 | * @mm: The process address space. | |
3241 | * @addr: The user address. | |
3242 | * | |
3243 | * Return: The vm_area_struct at the given address, %NULL otherwise. | |
3244 | */ | |
3245 | static inline | |
3246 | struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr) | |
3247 | { | |
d7c62295 | 3248 | return mtree_load(&mm->mm_mt, addr); |
ce6d42f2 LH |
3249 | } |
3250 | ||
1be7107f HD |
3251 | static inline unsigned long vm_start_gap(struct vm_area_struct *vma) |
3252 | { | |
3253 | unsigned long vm_start = vma->vm_start; | |
3254 | ||
3255 | if (vma->vm_flags & VM_GROWSDOWN) { | |
3256 | vm_start -= stack_guard_gap; | |
3257 | if (vm_start > vma->vm_start) | |
3258 | vm_start = 0; | |
3259 | } | |
3260 | return vm_start; | |
3261 | } | |
3262 | ||
3263 | static inline unsigned long vm_end_gap(struct vm_area_struct *vma) | |
3264 | { | |
3265 | unsigned long vm_end = vma->vm_end; | |
3266 | ||
3267 | if (vma->vm_flags & VM_GROWSUP) { | |
3268 | vm_end += stack_guard_gap; | |
3269 | if (vm_end < vma->vm_end) | |
3270 | vm_end = -PAGE_SIZE; | |
3271 | } | |
3272 | return vm_end; | |
3273 | } | |
3274 | ||
1da177e4 LT |
3275 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
3276 | { | |
3277 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
3278 | } | |
3279 | ||
640708a2 PE |
3280 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ |
3281 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, | |
3282 | unsigned long vm_start, unsigned long vm_end) | |
3283 | { | |
dc8635b2 | 3284 | struct vm_area_struct *vma = vma_lookup(mm, vm_start); |
640708a2 PE |
3285 | |
3286 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) | |
3287 | vma = NULL; | |
3288 | ||
3289 | return vma; | |
3290 | } | |
3291 | ||
017b1660 MK |
3292 | static inline bool range_in_vma(struct vm_area_struct *vma, |
3293 | unsigned long start, unsigned long end) | |
3294 | { | |
3295 | return (vma && vma->vm_start <= start && end <= vma->vm_end); | |
3296 | } | |
3297 | ||
bad849b3 | 3298 | #ifdef CONFIG_MMU |
804af2cf | 3299 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
64e45507 | 3300 | void vma_set_page_prot(struct vm_area_struct *vma); |
bad849b3 DH |
3301 | #else |
3302 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | |
3303 | { | |
3304 | return __pgprot(0); | |
3305 | } | |
64e45507 PF |
3306 | static inline void vma_set_page_prot(struct vm_area_struct *vma) |
3307 | { | |
3308 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); | |
3309 | } | |
bad849b3 DH |
3310 | #endif |
3311 | ||
295992fb CK |
3312 | void vma_set_file(struct vm_area_struct *vma, struct file *file); |
3313 | ||
5877231f | 3314 | #ifdef CONFIG_NUMA_BALANCING |
4b10e7d5 | 3315 | unsigned long change_prot_numa(struct vm_area_struct *vma, |
b24f53a0 LS |
3316 | unsigned long start, unsigned long end); |
3317 | #endif | |
3318 | ||
deceb6cd | 3319 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
deceb6cd HD |
3320 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
3321 | unsigned long pfn, unsigned long size, pgprot_t); | |
74ffa5a3 CH |
3322 | int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr, |
3323 | unsigned long pfn, unsigned long size, pgprot_t prot); | |
a145dd41 | 3324 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
8cd3984d AR |
3325 | int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr, |
3326 | struct page **pages, unsigned long *num); | |
a667d745 SJ |
3327 | int vm_map_pages(struct vm_area_struct *vma, struct page **pages, |
3328 | unsigned long num); | |
3329 | int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages, | |
3330 | unsigned long num); | |
ae2b01f3 | 3331 | vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
e0dc0d8f | 3332 | unsigned long pfn); |
f5e6d1d5 MW |
3333 | vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr, |
3334 | unsigned long pfn, pgprot_t pgprot); | |
5d747637 | 3335 | vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
01c8f1c4 | 3336 | pfn_t pfn); |
ab77dab4 SJ |
3337 | vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma, |
3338 | unsigned long addr, pfn_t pfn); | |
b4cbb197 LT |
3339 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); |
3340 | ||
1c8f4220 SJ |
3341 | static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma, |
3342 | unsigned long addr, struct page *page) | |
3343 | { | |
3344 | int err = vm_insert_page(vma, addr, page); | |
3345 | ||
3346 | if (err == -ENOMEM) | |
3347 | return VM_FAULT_OOM; | |
3348 | if (err < 0 && err != -EBUSY) | |
3349 | return VM_FAULT_SIGBUS; | |
3350 | ||
3351 | return VM_FAULT_NOPAGE; | |
3352 | } | |
3353 | ||
f8f6ae5d JG |
3354 | #ifndef io_remap_pfn_range |
3355 | static inline int io_remap_pfn_range(struct vm_area_struct *vma, | |
3356 | unsigned long addr, unsigned long pfn, | |
3357 | unsigned long size, pgprot_t prot) | |
3358 | { | |
3359 | return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot)); | |
3360 | } | |
3361 | #endif | |
3362 | ||
d97baf94 SJ |
3363 | static inline vm_fault_t vmf_error(int err) |
3364 | { | |
3365 | if (err == -ENOMEM) | |
3366 | return VM_FAULT_OOM; | |
3367 | return VM_FAULT_SIGBUS; | |
3368 | } | |
3369 | ||
df06b37f KB |
3370 | struct page *follow_page(struct vm_area_struct *vma, unsigned long address, |
3371 | unsigned int foll_flags); | |
240aadee | 3372 | |
2b740303 | 3373 | static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags) |
9a291a7c JM |
3374 | { |
3375 | if (vm_fault & VM_FAULT_OOM) | |
3376 | return -ENOMEM; | |
3377 | if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) | |
3378 | return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT; | |
3379 | if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV)) | |
3380 | return -EFAULT; | |
3381 | return 0; | |
3382 | } | |
3383 | ||
474098ed DH |
3384 | /* |
3385 | * Indicates whether GUP can follow a PROT_NONE mapped page, or whether | |
3386 | * a (NUMA hinting) fault is required. | |
3387 | */ | |
3388 | static inline bool gup_can_follow_protnone(unsigned int flags) | |
3389 | { | |
3390 | /* | |
3391 | * FOLL_FORCE has to be able to make progress even if the VMA is | |
3392 | * inaccessible. Further, FOLL_FORCE access usually does not represent | |
3393 | * application behaviour and we should avoid triggering NUMA hinting | |
3394 | * faults. | |
3395 | */ | |
3396 | return flags & FOLL_FORCE; | |
3397 | } | |
3398 | ||
8b1e0f81 | 3399 | typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data); |
aee16b3c JF |
3400 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, |
3401 | unsigned long size, pte_fn_t fn, void *data); | |
be1db475 DA |
3402 | extern int apply_to_existing_page_range(struct mm_struct *mm, |
3403 | unsigned long address, unsigned long size, | |
3404 | pte_fn_t fn, void *data); | |
aee16b3c | 3405 | |
8823b1db | 3406 | #ifdef CONFIG_PAGE_POISONING |
8db26a3d VB |
3407 | extern void __kernel_poison_pages(struct page *page, int numpages); |
3408 | extern void __kernel_unpoison_pages(struct page *page, int numpages); | |
3409 | extern bool _page_poisoning_enabled_early; | |
3410 | DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled); | |
3411 | static inline bool page_poisoning_enabled(void) | |
3412 | { | |
3413 | return _page_poisoning_enabled_early; | |
3414 | } | |
3415 | /* | |
3416 | * For use in fast paths after init_mem_debugging() has run, or when a | |
3417 | * false negative result is not harmful when called too early. | |
3418 | */ | |
3419 | static inline bool page_poisoning_enabled_static(void) | |
3420 | { | |
3421 | return static_branch_unlikely(&_page_poisoning_enabled); | |
3422 | } | |
3423 | static inline void kernel_poison_pages(struct page *page, int numpages) | |
3424 | { | |
3425 | if (page_poisoning_enabled_static()) | |
3426 | __kernel_poison_pages(page, numpages); | |
3427 | } | |
3428 | static inline void kernel_unpoison_pages(struct page *page, int numpages) | |
3429 | { | |
3430 | if (page_poisoning_enabled_static()) | |
3431 | __kernel_unpoison_pages(page, numpages); | |
3432 | } | |
8823b1db LA |
3433 | #else |
3434 | static inline bool page_poisoning_enabled(void) { return false; } | |
8db26a3d | 3435 | static inline bool page_poisoning_enabled_static(void) { return false; } |
03b6c9a3 | 3436 | static inline void __kernel_poison_pages(struct page *page, int nunmpages) { } |
8db26a3d VB |
3437 | static inline void kernel_poison_pages(struct page *page, int numpages) { } |
3438 | static inline void kernel_unpoison_pages(struct page *page, int numpages) { } | |
8823b1db LA |
3439 | #endif |
3440 | ||
51cba1eb | 3441 | DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc); |
6471384a AP |
3442 | static inline bool want_init_on_alloc(gfp_t flags) |
3443 | { | |
51cba1eb KC |
3444 | if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, |
3445 | &init_on_alloc)) | |
6471384a AP |
3446 | return true; |
3447 | return flags & __GFP_ZERO; | |
3448 | } | |
3449 | ||
51cba1eb | 3450 | DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free); |
6471384a AP |
3451 | static inline bool want_init_on_free(void) |
3452 | { | |
51cba1eb KC |
3453 | return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON, |
3454 | &init_on_free); | |
6471384a AP |
3455 | } |
3456 | ||
8e57f8ac VB |
3457 | extern bool _debug_pagealloc_enabled_early; |
3458 | DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled); | |
031bc574 JK |
3459 | |
3460 | static inline bool debug_pagealloc_enabled(void) | |
8e57f8ac VB |
3461 | { |
3462 | return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) && | |
3463 | _debug_pagealloc_enabled_early; | |
3464 | } | |
3465 | ||
3466 | /* | |
3467 | * For use in fast paths after init_debug_pagealloc() has run, or when a | |
3468 | * false negative result is not harmful when called too early. | |
3469 | */ | |
3470 | static inline bool debug_pagealloc_enabled_static(void) | |
031bc574 | 3471 | { |
96a2b03f VB |
3472 | if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC)) |
3473 | return false; | |
3474 | ||
3475 | return static_branch_unlikely(&_debug_pagealloc_enabled); | |
031bc574 JK |
3476 | } |
3477 | ||
5d6ad668 | 3478 | #ifdef CONFIG_DEBUG_PAGEALLOC |
c87cbc1f | 3479 | /* |
5d6ad668 MR |
3480 | * To support DEBUG_PAGEALLOC architecture must ensure that |
3481 | * __kernel_map_pages() never fails | |
c87cbc1f | 3482 | */ |
d6332692 RE |
3483 | extern void __kernel_map_pages(struct page *page, int numpages, int enable); |
3484 | ||
77bc7fd6 MR |
3485 | static inline void debug_pagealloc_map_pages(struct page *page, int numpages) |
3486 | { | |
3487 | if (debug_pagealloc_enabled_static()) | |
3488 | __kernel_map_pages(page, numpages, 1); | |
3489 | } | |
3490 | ||
3491 | static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) | |
3492 | { | |
3493 | if (debug_pagealloc_enabled_static()) | |
3494 | __kernel_map_pages(page, numpages, 0); | |
3495 | } | |
5d6ad668 | 3496 | #else /* CONFIG_DEBUG_PAGEALLOC */ |
77bc7fd6 MR |
3497 | static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {} |
3498 | static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {} | |
5d6ad668 | 3499 | #endif /* CONFIG_DEBUG_PAGEALLOC */ |
1da177e4 | 3500 | |
a6c19dfe | 3501 | #ifdef __HAVE_ARCH_GATE_AREA |
31db58b3 | 3502 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
a6c19dfe AL |
3503 | extern int in_gate_area_no_mm(unsigned long addr); |
3504 | extern int in_gate_area(struct mm_struct *mm, unsigned long addr); | |
1da177e4 | 3505 | #else |
a6c19dfe AL |
3506 | static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) |
3507 | { | |
3508 | return NULL; | |
3509 | } | |
3510 | static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } | |
3511 | static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) | |
3512 | { | |
3513 | return 0; | |
3514 | } | |
1da177e4 LT |
3515 | #endif /* __HAVE_ARCH_GATE_AREA */ |
3516 | ||
44a70ade MH |
3517 | extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm); |
3518 | ||
146732ce JT |
3519 | #ifdef CONFIG_SYSCTL |
3520 | extern int sysctl_drop_caches; | |
32927393 CH |
3521 | int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *, |
3522 | loff_t *); | |
146732ce JT |
3523 | #endif |
3524 | ||
cb731d6c | 3525 | void drop_slab(void); |
9d0243bc | 3526 | |
7a9166e3 LY |
3527 | #ifndef CONFIG_MMU |
3528 | #define randomize_va_space 0 | |
3529 | #else | |
a62eaf15 | 3530 | extern int randomize_va_space; |
7a9166e3 | 3531 | #endif |
a62eaf15 | 3532 | |
045e72ac | 3533 | const char * arch_vma_name(struct vm_area_struct *vma); |
89165b8b | 3534 | #ifdef CONFIG_MMU |
03252919 | 3535 | void print_vma_addr(char *prefix, unsigned long rip); |
89165b8b CH |
3536 | #else |
3537 | static inline void print_vma_addr(char *prefix, unsigned long rip) | |
3538 | { | |
3539 | } | |
3540 | #endif | |
e6e5494c | 3541 | |
35fd1eb1 | 3542 | void *sparse_buffer_alloc(unsigned long size); |
e9c0a3f0 | 3543 | struct page * __populate_section_memmap(unsigned long pfn, |
e3246d8f JM |
3544 | unsigned long nr_pages, int nid, struct vmem_altmap *altmap, |
3545 | struct dev_pagemap *pgmap); | |
7b09f5af FC |
3546 | void pmd_init(void *addr); |
3547 | void pud_init(void *addr); | |
29c71111 | 3548 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
c2febafc KS |
3549 | p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node); |
3550 | pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node); | |
29c71111 | 3551 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); |
1d9cfee7 | 3552 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node, |
4917f55b | 3553 | struct vmem_altmap *altmap, struct page *reuse); |
8f6aac41 | 3554 | void *vmemmap_alloc_block(unsigned long size, int node); |
4b94ffdc | 3555 | struct vmem_altmap; |
56993b4e AK |
3556 | void *vmemmap_alloc_block_buf(unsigned long size, int node, |
3557 | struct vmem_altmap *altmap); | |
8f6aac41 | 3558 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
2045a3b8 FC |
3559 | void vmemmap_set_pmd(pmd_t *pmd, void *p, int node, |
3560 | unsigned long addr, unsigned long next); | |
3561 | int vmemmap_check_pmd(pmd_t *pmd, int node, | |
3562 | unsigned long addr, unsigned long next); | |
0aad818b | 3563 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, |
1d9cfee7 | 3564 | int node, struct vmem_altmap *altmap); |
2045a3b8 FC |
3565 | int vmemmap_populate_hugepages(unsigned long start, unsigned long end, |
3566 | int node, struct vmem_altmap *altmap); | |
7b73d978 CH |
3567 | int vmemmap_populate(unsigned long start, unsigned long end, int node, |
3568 | struct vmem_altmap *altmap); | |
c2b91e2e | 3569 | void vmemmap_populate_print_last(void); |
0197518c | 3570 | #ifdef CONFIG_MEMORY_HOTPLUG |
24b6d416 CH |
3571 | void vmemmap_free(unsigned long start, unsigned long end, |
3572 | struct vmem_altmap *altmap); | |
0197518c | 3573 | #endif |
87a7ae75 | 3574 | |
0b376f1e | 3575 | #ifdef CONFIG_ARCH_WANT_OPTIMIZE_VMEMMAP |
87a7ae75 AK |
3576 | static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap, |
3577 | struct dev_pagemap *pgmap) | |
3578 | { | |
3579 | return is_power_of_2(sizeof(struct page)) && | |
3580 | pgmap && (pgmap_vmemmap_nr(pgmap) > 1) && !altmap; | |
3581 | } | |
3582 | #else | |
3583 | static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap, | |
3584 | struct dev_pagemap *pgmap) | |
3585 | { | |
3586 | return false; | |
3587 | } | |
3588 | #endif | |
3589 | ||
46723bfa | 3590 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, |
15670bfe | 3591 | unsigned long nr_pages); |
6a46079c | 3592 | |
82ba011b AK |
3593 | enum mf_flags { |
3594 | MF_COUNT_INCREASED = 1 << 0, | |
7329bbeb | 3595 | MF_ACTION_REQUIRED = 1 << 1, |
6751ed65 | 3596 | MF_MUST_KILL = 1 << 2, |
cf870c70 | 3597 | MF_SOFT_OFFLINE = 1 << 3, |
bf181c58 | 3598 | MF_UNPOISON = 1 << 4, |
67f22ba7 | 3599 | MF_SW_SIMULATED = 1 << 5, |
38f6d293 | 3600 | MF_NO_RETRY = 1 << 6, |
82ba011b | 3601 | }; |
c36e2024 SR |
3602 | int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index, |
3603 | unsigned long count, int mf_flags); | |
83b57531 | 3604 | extern int memory_failure(unsigned long pfn, int flags); |
06202231 | 3605 | extern void memory_failure_queue_kick(int cpu); |
847ce401 | 3606 | extern int unpoison_memory(unsigned long pfn); |
d0505e9f | 3607 | extern void shake_page(struct page *p); |
5844a486 | 3608 | extern atomic_long_t num_poisoned_pages __read_mostly; |
feec24a6 | 3609 | extern int soft_offline_page(unsigned long pfn, int flags); |
405ce051 | 3610 | #ifdef CONFIG_MEMORY_FAILURE |
d302c239 | 3611 | extern void memory_failure_queue(unsigned long pfn, int flags); |
e591ef7d NH |
3612 | extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags, |
3613 | bool *migratable_cleared); | |
5033091d NH |
3614 | void num_poisoned_pages_inc(unsigned long pfn); |
3615 | void num_poisoned_pages_sub(unsigned long pfn, long i); | |
4248d008 | 3616 | struct task_struct *task_early_kill(struct task_struct *tsk, int force_early); |
405ce051 | 3617 | #else |
d302c239 TL |
3618 | static inline void memory_failure_queue(unsigned long pfn, int flags) |
3619 | { | |
3620 | } | |
3621 | ||
e591ef7d NH |
3622 | static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags, |
3623 | bool *migratable_cleared) | |
405ce051 NH |
3624 | { |
3625 | return 0; | |
3626 | } | |
d027122d | 3627 | |
a46c9304 | 3628 | static inline void num_poisoned_pages_inc(unsigned long pfn) |
d027122d NH |
3629 | { |
3630 | } | |
5033091d NH |
3631 | |
3632 | static inline void num_poisoned_pages_sub(unsigned long pfn, long i) | |
3633 | { | |
3634 | } | |
3635 | #endif | |
3636 | ||
4248d008 LX |
3637 | #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM) |
3638 | void add_to_kill_ksm(struct task_struct *tsk, struct page *p, | |
3639 | struct vm_area_struct *vma, struct list_head *to_kill, | |
3640 | unsigned long ksm_addr); | |
3641 | #endif | |
3642 | ||
5033091d NH |
3643 | #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG) |
3644 | extern void memblk_nr_poison_inc(unsigned long pfn); | |
3645 | extern void memblk_nr_poison_sub(unsigned long pfn, long i); | |
3646 | #else | |
3647 | static inline void memblk_nr_poison_inc(unsigned long pfn) | |
3648 | { | |
3649 | } | |
3650 | ||
3651 | static inline void memblk_nr_poison_sub(unsigned long pfn, long i) | |
3652 | { | |
3653 | } | |
405ce051 | 3654 | #endif |
6a46079c | 3655 | |
03b122da TL |
3656 | #ifndef arch_memory_failure |
3657 | static inline int arch_memory_failure(unsigned long pfn, int flags) | |
3658 | { | |
3659 | return -ENXIO; | |
3660 | } | |
3661 | #endif | |
3662 | ||
3663 | #ifndef arch_is_platform_page | |
3664 | static inline bool arch_is_platform_page(u64 paddr) | |
3665 | { | |
3666 | return false; | |
3667 | } | |
3668 | #endif | |
cc637b17 XX |
3669 | |
3670 | /* | |
3671 | * Error handlers for various types of pages. | |
3672 | */ | |
cc3e2af4 | 3673 | enum mf_result { |
cc637b17 XX |
3674 | MF_IGNORED, /* Error: cannot be handled */ |
3675 | MF_FAILED, /* Error: handling failed */ | |
3676 | MF_DELAYED, /* Will be handled later */ | |
3677 | MF_RECOVERED, /* Successfully recovered */ | |
3678 | }; | |
3679 | ||
3680 | enum mf_action_page_type { | |
3681 | MF_MSG_KERNEL, | |
3682 | MF_MSG_KERNEL_HIGH_ORDER, | |
3683 | MF_MSG_SLAB, | |
3684 | MF_MSG_DIFFERENT_COMPOUND, | |
cc637b17 XX |
3685 | MF_MSG_HUGE, |
3686 | MF_MSG_FREE_HUGE, | |
3687 | MF_MSG_UNMAP_FAILED, | |
3688 | MF_MSG_DIRTY_SWAPCACHE, | |
3689 | MF_MSG_CLEAN_SWAPCACHE, | |
3690 | MF_MSG_DIRTY_MLOCKED_LRU, | |
3691 | MF_MSG_CLEAN_MLOCKED_LRU, | |
3692 | MF_MSG_DIRTY_UNEVICTABLE_LRU, | |
3693 | MF_MSG_CLEAN_UNEVICTABLE_LRU, | |
3694 | MF_MSG_DIRTY_LRU, | |
3695 | MF_MSG_CLEAN_LRU, | |
3696 | MF_MSG_TRUNCATED_LRU, | |
3697 | MF_MSG_BUDDY, | |
6100e34b | 3698 | MF_MSG_DAX, |
5d1fd5dc | 3699 | MF_MSG_UNSPLIT_THP, |
cc637b17 XX |
3700 | MF_MSG_UNKNOWN, |
3701 | }; | |
3702 | ||
44b8f8bf JY |
3703 | /* |
3704 | * Sysfs entries for memory failure handling statistics. | |
3705 | */ | |
3706 | extern const struct attribute_group memory_failure_attr_group; | |
3707 | ||
47ad8475 AA |
3708 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
3709 | extern void clear_huge_page(struct page *page, | |
c79b57e4 | 3710 | unsigned long addr_hint, |
47ad8475 | 3711 | unsigned int pages_per_huge_page); |
1cb9dc4b LS |
3712 | int copy_user_large_folio(struct folio *dst, struct folio *src, |
3713 | unsigned long addr_hint, | |
3714 | struct vm_area_struct *vma); | |
e87340ca Z |
3715 | long copy_folio_from_user(struct folio *dst_folio, |
3716 | const void __user *usr_src, | |
3717 | bool allow_pagefault); | |
2484ca9b THV |
3718 | |
3719 | /** | |
3720 | * vma_is_special_huge - Are transhuge page-table entries considered special? | |
3721 | * @vma: Pointer to the struct vm_area_struct to consider | |
3722 | * | |
3723 | * Whether transhuge page-table entries are considered "special" following | |
3724 | * the definition in vm_normal_page(). | |
3725 | * | |
3726 | * Return: true if transhuge page-table entries should be considered special, | |
3727 | * false otherwise. | |
3728 | */ | |
3729 | static inline bool vma_is_special_huge(const struct vm_area_struct *vma) | |
3730 | { | |
3731 | return vma_is_dax(vma) || (vma->vm_file && | |
3732 | (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))); | |
3733 | } | |
3734 | ||
47ad8475 AA |
3735 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ |
3736 | ||
c0a32fc5 SG |
3737 | #ifdef CONFIG_DEBUG_PAGEALLOC |
3738 | extern unsigned int _debug_guardpage_minorder; | |
96a2b03f | 3739 | DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled); |
c0a32fc5 SG |
3740 | |
3741 | static inline unsigned int debug_guardpage_minorder(void) | |
3742 | { | |
3743 | return _debug_guardpage_minorder; | |
3744 | } | |
3745 | ||
e30825f1 JK |
3746 | static inline bool debug_guardpage_enabled(void) |
3747 | { | |
96a2b03f | 3748 | return static_branch_unlikely(&_debug_guardpage_enabled); |
e30825f1 JK |
3749 | } |
3750 | ||
c0a32fc5 SG |
3751 | static inline bool page_is_guard(struct page *page) |
3752 | { | |
e30825f1 JK |
3753 | if (!debug_guardpage_enabled()) |
3754 | return false; | |
3755 | ||
3972f6bb | 3756 | return PageGuard(page); |
c0a32fc5 SG |
3757 | } |
3758 | #else | |
3759 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } | |
e30825f1 | 3760 | static inline bool debug_guardpage_enabled(void) { return false; } |
c0a32fc5 SG |
3761 | static inline bool page_is_guard(struct page *page) { return false; } |
3762 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | |
3763 | ||
f9872caf CS |
3764 | #if MAX_NUMNODES > 1 |
3765 | void __init setup_nr_node_ids(void); | |
3766 | #else | |
3767 | static inline void setup_nr_node_ids(void) {} | |
3768 | #endif | |
3769 | ||
010c164a SL |
3770 | extern int memcmp_pages(struct page *page1, struct page *page2); |
3771 | ||
3772 | static inline int pages_identical(struct page *page1, struct page *page2) | |
3773 | { | |
3774 | return !memcmp_pages(page1, page2); | |
3775 | } | |
3776 | ||
c5acad84 TH |
3777 | #ifdef CONFIG_MAPPING_DIRTY_HELPERS |
3778 | unsigned long clean_record_shared_mapping_range(struct address_space *mapping, | |
3779 | pgoff_t first_index, pgoff_t nr, | |
3780 | pgoff_t bitmap_pgoff, | |
3781 | unsigned long *bitmap, | |
3782 | pgoff_t *start, | |
3783 | pgoff_t *end); | |
3784 | ||
3785 | unsigned long wp_shared_mapping_range(struct address_space *mapping, | |
3786 | pgoff_t first_index, pgoff_t nr); | |
3787 | #endif | |
3788 | ||
2374c09b CH |
3789 | extern int sysctl_nr_trim_pages; |
3790 | ||
5bb1bb35 | 3791 | #ifdef CONFIG_PRINTK |
8e7f37f2 | 3792 | void mem_dump_obj(void *object); |
5bb1bb35 PM |
3793 | #else |
3794 | static inline void mem_dump_obj(void *object) {} | |
3795 | #endif | |
8e7f37f2 | 3796 | |
22247efd PX |
3797 | /** |
3798 | * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it | |
3799 | * @seals: the seals to check | |
3800 | * @vma: the vma to operate on | |
3801 | * | |
3802 | * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on | |
3803 | * the vma flags. Return 0 if check pass, or <0 for errors. | |
3804 | */ | |
3805 | static inline int seal_check_future_write(int seals, struct vm_area_struct *vma) | |
3806 | { | |
3807 | if (seals & F_SEAL_FUTURE_WRITE) { | |
3808 | /* | |
3809 | * New PROT_WRITE and MAP_SHARED mmaps are not allowed when | |
3810 | * "future write" seal active. | |
3811 | */ | |
3812 | if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE)) | |
3813 | return -EPERM; | |
3814 | ||
3815 | /* | |
3816 | * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as | |
3817 | * MAP_SHARED and read-only, take care to not allow mprotect to | |
3818 | * revert protections on such mappings. Do this only for shared | |
3819 | * mappings. For private mappings, don't need to mask | |
3820 | * VM_MAYWRITE as we still want them to be COW-writable. | |
3821 | */ | |
3822 | if (vma->vm_flags & VM_SHARED) | |
1c71222e | 3823 | vm_flags_clear(vma, VM_MAYWRITE); |
22247efd PX |
3824 | } |
3825 | ||
3826 | return 0; | |
3827 | } | |
3828 | ||
9a10064f CC |
3829 | #ifdef CONFIG_ANON_VMA_NAME |
3830 | int madvise_set_anon_name(struct mm_struct *mm, unsigned long start, | |
5c26f6ac SB |
3831 | unsigned long len_in, |
3832 | struct anon_vma_name *anon_name); | |
9a10064f CC |
3833 | #else |
3834 | static inline int | |
3835 | madvise_set_anon_name(struct mm_struct *mm, unsigned long start, | |
5c26f6ac | 3836 | unsigned long len_in, struct anon_vma_name *anon_name) { |
9a10064f CC |
3837 | return 0; |
3838 | } | |
3839 | #endif | |
3840 | ||
1da177e4 | 3841 | #endif /* _LINUX_MM_H */ |