| 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
| 2 | #ifndef _LINUX_MM_H |
| 3 | #define _LINUX_MM_H |
| 4 | |
| 5 | #include <linux/errno.h> |
| 6 | |
| 7 | #ifdef __KERNEL__ |
| 8 | |
| 9 | #include <linux/mmdebug.h> |
| 10 | #include <linux/gfp.h> |
| 11 | #include <linux/bug.h> |
| 12 | #include <linux/list.h> |
| 13 | #include <linux/mmzone.h> |
| 14 | #include <linux/rbtree.h> |
| 15 | #include <linux/atomic.h> |
| 16 | #include <linux/debug_locks.h> |
| 17 | #include <linux/mm_types.h> |
| 18 | #include <linux/mmap_lock.h> |
| 19 | #include <linux/range.h> |
| 20 | #include <linux/pfn.h> |
| 21 | #include <linux/percpu-refcount.h> |
| 22 | #include <linux/bit_spinlock.h> |
| 23 | #include <linux/shrinker.h> |
| 24 | #include <linux/resource.h> |
| 25 | #include <linux/page_ext.h> |
| 26 | #include <linux/err.h> |
| 27 | #include <linux/page_ref.h> |
| 28 | #include <linux/memremap.h> |
| 29 | #include <linux/overflow.h> |
| 30 | #include <linux/sizes.h> |
| 31 | #include <linux/sched.h> |
| 32 | #include <linux/pgtable.h> |
| 33 | |
| 34 | struct mempolicy; |
| 35 | struct anon_vma; |
| 36 | struct anon_vma_chain; |
| 37 | struct file_ra_state; |
| 38 | struct user_struct; |
| 39 | struct writeback_control; |
| 40 | struct bdi_writeback; |
| 41 | struct pt_regs; |
| 42 | |
| 43 | void init_mm_internals(void); |
| 44 | |
| 45 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ |
| 46 | extern unsigned long max_mapnr; |
| 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) { } |
| 54 | #endif |
| 55 | |
| 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 | |
| 77 | extern void * high_memory; |
| 78 | extern int page_cluster; |
| 79 | |
| 80 | #ifdef CONFIG_SYSCTL |
| 81 | extern int sysctl_legacy_va_layout; |
| 82 | #else |
| 83 | #define sysctl_legacy_va_layout 0 |
| 84 | #endif |
| 85 | |
| 86 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS |
| 87 | extern const int mmap_rnd_bits_min; |
| 88 | extern const int mmap_rnd_bits_max; |
| 89 | extern int mmap_rnd_bits __read_mostly; |
| 90 | #endif |
| 91 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS |
| 92 | extern const int mmap_rnd_compat_bits_min; |
| 93 | extern const int mmap_rnd_compat_bits_max; |
| 94 | extern int mmap_rnd_compat_bits __read_mostly; |
| 95 | #endif |
| 96 | |
| 97 | #include <asm/page.h> |
| 98 | #include <asm/processor.h> |
| 99 | |
| 100 | /* |
| 101 | * Architectures that support memory tagging (assigning tags to memory regions, |
| 102 | * embedding these tags into addresses that point to these memory regions, and |
| 103 | * checking that the memory and the pointer tags match on memory accesses) |
| 104 | * redefine this macro to strip tags from pointers. |
| 105 | * It's defined as noop for arcitectures that don't support memory tagging. |
| 106 | */ |
| 107 | #ifndef untagged_addr |
| 108 | #define untagged_addr(addr) (addr) |
| 109 | #endif |
| 110 | |
| 111 | #ifndef __pa_symbol |
| 112 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) |
| 113 | #endif |
| 114 | |
| 115 | #ifndef page_to_virt |
| 116 | #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x))) |
| 117 | #endif |
| 118 | |
| 119 | #ifndef lm_alias |
| 120 | #define lm_alias(x) __va(__pa_symbol(x)) |
| 121 | #endif |
| 122 | |
| 123 | /* |
| 124 | * To prevent common memory management code establishing |
| 125 | * a zero page mapping on a read fault. |
| 126 | * This macro should be defined within <asm/pgtable.h>. |
| 127 | * s390 does this to prevent multiplexing of hardware bits |
| 128 | * related to the physical page in case of virtualization. |
| 129 | */ |
| 130 | #ifndef mm_forbids_zeropage |
| 131 | #define mm_forbids_zeropage(X) (0) |
| 132 | #endif |
| 133 | |
| 134 | /* |
| 135 | * On some architectures it is expensive to call memset() for small sizes. |
| 136 | * If an architecture decides to implement their own version of |
| 137 | * mm_zero_struct_page they should wrap the defines below in a #ifndef and |
| 138 | * define their own version of this macro in <asm/pgtable.h> |
| 139 | */ |
| 140 | #if BITS_PER_LONG == 64 |
| 141 | /* This function must be updated when the size of struct page grows above 80 |
| 142 | * or reduces below 56. The idea that compiler optimizes out switch() |
| 143 | * statement, and only leaves move/store instructions. Also the compiler can |
| 144 | * combine write statments if they are both assignments and can be reordered, |
| 145 | * this can result in several of the writes here being dropped. |
| 146 | */ |
| 147 | #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp) |
| 148 | static inline void __mm_zero_struct_page(struct page *page) |
| 149 | { |
| 150 | unsigned long *_pp = (void *)page; |
| 151 | |
| 152 | /* Check that struct page is either 56, 64, 72, or 80 bytes */ |
| 153 | BUILD_BUG_ON(sizeof(struct page) & 7); |
| 154 | BUILD_BUG_ON(sizeof(struct page) < 56); |
| 155 | BUILD_BUG_ON(sizeof(struct page) > 80); |
| 156 | |
| 157 | switch (sizeof(struct page)) { |
| 158 | case 80: |
| 159 | _pp[9] = 0; /* fallthrough */ |
| 160 | case 72: |
| 161 | _pp[8] = 0; /* fallthrough */ |
| 162 | case 64: |
| 163 | _pp[7] = 0; /* fallthrough */ |
| 164 | case 56: |
| 165 | _pp[6] = 0; |
| 166 | _pp[5] = 0; |
| 167 | _pp[4] = 0; |
| 168 | _pp[3] = 0; |
| 169 | _pp[2] = 0; |
| 170 | _pp[1] = 0; |
| 171 | _pp[0] = 0; |
| 172 | } |
| 173 | } |
| 174 | #else |
| 175 | #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page))) |
| 176 | #endif |
| 177 | |
| 178 | /* |
| 179 | * Default maximum number of active map areas, this limits the number of vmas |
| 180 | * per mm struct. Users can overwrite this number by sysctl but there is a |
| 181 | * problem. |
| 182 | * |
| 183 | * When a program's coredump is generated as ELF format, a section is created |
| 184 | * per a vma. In ELF, the number of sections is represented in unsigned short. |
| 185 | * This means the number of sections should be smaller than 65535 at coredump. |
| 186 | * Because the kernel adds some informative sections to a image of program at |
| 187 | * generating coredump, we need some margin. The number of extra sections is |
| 188 | * 1-3 now and depends on arch. We use "5" as safe margin, here. |
| 189 | * |
| 190 | * ELF extended numbering allows more than 65535 sections, so 16-bit bound is |
| 191 | * not a hard limit any more. Although some userspace tools can be surprised by |
| 192 | * that. |
| 193 | */ |
| 194 | #define MAPCOUNT_ELF_CORE_MARGIN (5) |
| 195 | #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) |
| 196 | |
| 197 | extern int sysctl_max_map_count; |
| 198 | |
| 199 | extern unsigned long sysctl_user_reserve_kbytes; |
| 200 | extern unsigned long sysctl_admin_reserve_kbytes; |
| 201 | |
| 202 | extern int sysctl_overcommit_memory; |
| 203 | extern int sysctl_overcommit_ratio; |
| 204 | extern unsigned long sysctl_overcommit_kbytes; |
| 205 | |
| 206 | int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *, |
| 207 | loff_t *); |
| 208 | int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *, |
| 209 | loff_t *); |
| 210 | int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *, |
| 211 | loff_t *); |
| 212 | |
| 213 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
| 214 | |
| 215 | /* to align the pointer to the (next) page boundary */ |
| 216 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) |
| 217 | |
| 218 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ |
| 219 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE) |
| 220 | |
| 221 | #define lru_to_page(head) (list_entry((head)->prev, struct page, lru)) |
| 222 | |
| 223 | /* |
| 224 | * Linux kernel virtual memory manager primitives. |
| 225 | * The idea being to have a "virtual" mm in the same way |
| 226 | * we have a virtual fs - giving a cleaner interface to the |
| 227 | * mm details, and allowing different kinds of memory mappings |
| 228 | * (from shared memory to executable loading to arbitrary |
| 229 | * mmap() functions). |
| 230 | */ |
| 231 | |
| 232 | struct vm_area_struct *vm_area_alloc(struct mm_struct *); |
| 233 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *); |
| 234 | void vm_area_free(struct vm_area_struct *); |
| 235 | |
| 236 | #ifndef CONFIG_MMU |
| 237 | extern struct rb_root nommu_region_tree; |
| 238 | extern struct rw_semaphore nommu_region_sem; |
| 239 | |
| 240 | extern unsigned int kobjsize(const void *objp); |
| 241 | #endif |
| 242 | |
| 243 | /* |
| 244 | * vm_flags in vm_area_struct, see mm_types.h. |
| 245 | * When changing, update also include/trace/events/mmflags.h |
| 246 | */ |
| 247 | #define VM_NONE 0x00000000 |
| 248 | |
| 249 | #define VM_READ 0x00000001 /* currently active flags */ |
| 250 | #define VM_WRITE 0x00000002 |
| 251 | #define VM_EXEC 0x00000004 |
| 252 | #define VM_SHARED 0x00000008 |
| 253 | |
| 254 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
| 255 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
| 256 | #define VM_MAYWRITE 0x00000020 |
| 257 | #define VM_MAYEXEC 0x00000040 |
| 258 | #define VM_MAYSHARE 0x00000080 |
| 259 | |
| 260 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ |
| 261 | #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */ |
| 262 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
| 263 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
| 264 | #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */ |
| 265 | |
| 266 | #define VM_LOCKED 0x00002000 |
| 267 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ |
| 268 | |
| 269 | /* Used by sys_madvise() */ |
| 270 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ |
| 271 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ |
| 272 | |
| 273 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ |
| 274 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ |
| 275 | #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */ |
| 276 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
| 277 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
| 278 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
| 279 | #define VM_SYNC 0x00800000 /* Synchronous page faults */ |
| 280 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ |
| 281 | #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */ |
| 282 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ |
| 283 | |
| 284 | #ifdef CONFIG_MEM_SOFT_DIRTY |
| 285 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ |
| 286 | #else |
| 287 | # define VM_SOFTDIRTY 0 |
| 288 | #endif |
| 289 | |
| 290 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
| 291 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ |
| 292 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ |
| 293 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
| 294 | |
| 295 | #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS |
| 296 | #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */ |
| 297 | #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */ |
| 298 | #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */ |
| 299 | #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */ |
| 300 | #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */ |
| 301 | #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0) |
| 302 | #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1) |
| 303 | #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2) |
| 304 | #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3) |
| 305 | #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4) |
| 306 | #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */ |
| 307 | |
| 308 | #ifdef CONFIG_ARCH_HAS_PKEYS |
| 309 | # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0 |
| 310 | # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */ |
| 311 | # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */ |
| 312 | # define VM_PKEY_BIT2 VM_HIGH_ARCH_2 |
| 313 | # define VM_PKEY_BIT3 VM_HIGH_ARCH_3 |
| 314 | #ifdef CONFIG_PPC |
| 315 | # define VM_PKEY_BIT4 VM_HIGH_ARCH_4 |
| 316 | #else |
| 317 | # define VM_PKEY_BIT4 0 |
| 318 | #endif |
| 319 | #endif /* CONFIG_ARCH_HAS_PKEYS */ |
| 320 | |
| 321 | #if defined(CONFIG_X86) |
| 322 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ |
| 323 | #elif defined(CONFIG_PARISC) |
| 324 | # define VM_GROWSUP VM_ARCH_1 |
| 325 | #elif defined(CONFIG_IA64) |
| 326 | # define VM_GROWSUP VM_ARCH_1 |
| 327 | #elif defined(CONFIG_SPARC64) |
| 328 | # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */ |
| 329 | # define VM_ARCH_CLEAR VM_SPARC_ADI |
| 330 | #elif defined(CONFIG_ARM64) |
| 331 | # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */ |
| 332 | # define VM_ARCH_CLEAR VM_ARM64_BTI |
| 333 | #elif !defined(CONFIG_MMU) |
| 334 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ |
| 335 | #endif |
| 336 | |
| 337 | #ifndef VM_GROWSUP |
| 338 | # define VM_GROWSUP VM_NONE |
| 339 | #endif |
| 340 | |
| 341 | /* Bits set in the VMA until the stack is in its final location */ |
| 342 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) |
| 343 | |
| 344 | #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0) |
| 345 | |
| 346 | /* Common data flag combinations */ |
| 347 | #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \ |
| 348 | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) |
| 349 | #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \ |
| 350 | VM_MAYWRITE | VM_MAYEXEC) |
| 351 | #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \ |
| 352 | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) |
| 353 | |
| 354 | #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */ |
| 355 | #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC |
| 356 | #endif |
| 357 | |
| 358 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
| 359 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS |
| 360 | #endif |
| 361 | |
| 362 | #ifdef CONFIG_STACK_GROWSUP |
| 363 | #define VM_STACK VM_GROWSUP |
| 364 | #else |
| 365 | #define VM_STACK VM_GROWSDOWN |
| 366 | #endif |
| 367 | |
| 368 | #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 369 | |
| 370 | /* VMA basic access permission flags */ |
| 371 | #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC) |
| 372 | |
| 373 | |
| 374 | /* |
| 375 | * Special vmas that are non-mergable, non-mlock()able. |
| 376 | */ |
| 377 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) |
| 378 | |
| 379 | /* This mask prevents VMA from being scanned with khugepaged */ |
| 380 | #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB) |
| 381 | |
| 382 | /* This mask defines which mm->def_flags a process can inherit its parent */ |
| 383 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE |
| 384 | |
| 385 | /* This mask is used to clear all the VMA flags used by mlock */ |
| 386 | #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT)) |
| 387 | |
| 388 | /* Arch-specific flags to clear when updating VM flags on protection change */ |
| 389 | #ifndef VM_ARCH_CLEAR |
| 390 | # define VM_ARCH_CLEAR VM_NONE |
| 391 | #endif |
| 392 | #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR) |
| 393 | |
| 394 | /* |
| 395 | * mapping from the currently active vm_flags protection bits (the |
| 396 | * low four bits) to a page protection mask.. |
| 397 | */ |
| 398 | extern pgprot_t protection_map[16]; |
| 399 | |
| 400 | /** |
| 401 | * Fault flag definitions. |
| 402 | * |
| 403 | * @FAULT_FLAG_WRITE: Fault was a write fault. |
| 404 | * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE. |
| 405 | * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked. |
| 406 | * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying. |
| 407 | * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region. |
| 408 | * @FAULT_FLAG_TRIED: The fault has been tried once. |
| 409 | * @FAULT_FLAG_USER: The fault originated in userspace. |
| 410 | * @FAULT_FLAG_REMOTE: The fault is not for current task/mm. |
| 411 | * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch. |
| 412 | * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals. |
| 413 | * |
| 414 | * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify |
| 415 | * whether we would allow page faults to retry by specifying these two |
| 416 | * fault flags correctly. Currently there can be three legal combinations: |
| 417 | * |
| 418 | * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and |
| 419 | * this is the first try |
| 420 | * |
| 421 | * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and |
| 422 | * we've already tried at least once |
| 423 | * |
| 424 | * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry |
| 425 | * |
| 426 | * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never |
| 427 | * be used. Note that page faults can be allowed to retry for multiple times, |
| 428 | * in which case we'll have an initial fault with flags (a) then later on |
| 429 | * continuous faults with flags (b). We should always try to detect pending |
| 430 | * signals before a retry to make sure the continuous page faults can still be |
| 431 | * interrupted if necessary. |
| 432 | */ |
| 433 | #define FAULT_FLAG_WRITE 0x01 |
| 434 | #define FAULT_FLAG_MKWRITE 0x02 |
| 435 | #define FAULT_FLAG_ALLOW_RETRY 0x04 |
| 436 | #define FAULT_FLAG_RETRY_NOWAIT 0x08 |
| 437 | #define FAULT_FLAG_KILLABLE 0x10 |
| 438 | #define FAULT_FLAG_TRIED 0x20 |
| 439 | #define FAULT_FLAG_USER 0x40 |
| 440 | #define FAULT_FLAG_REMOTE 0x80 |
| 441 | #define FAULT_FLAG_INSTRUCTION 0x100 |
| 442 | #define FAULT_FLAG_INTERRUPTIBLE 0x200 |
| 443 | |
| 444 | /* |
| 445 | * The default fault flags that should be used by most of the |
| 446 | * arch-specific page fault handlers. |
| 447 | */ |
| 448 | #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \ |
| 449 | FAULT_FLAG_KILLABLE | \ |
| 450 | FAULT_FLAG_INTERRUPTIBLE) |
| 451 | |
| 452 | /** |
| 453 | * fault_flag_allow_retry_first - check ALLOW_RETRY the first time |
| 454 | * |
| 455 | * This is mostly used for places where we want to try to avoid taking |
| 456 | * the mmap_lock for too long a time when waiting for another condition |
| 457 | * to change, in which case we can try to be polite to release the |
| 458 | * mmap_lock in the first round to avoid potential starvation of other |
| 459 | * processes that would also want the mmap_lock. |
| 460 | * |
| 461 | * Return: true if the page fault allows retry and this is the first |
| 462 | * attempt of the fault handling; false otherwise. |
| 463 | */ |
| 464 | static inline bool fault_flag_allow_retry_first(unsigned int flags) |
| 465 | { |
| 466 | return (flags & FAULT_FLAG_ALLOW_RETRY) && |
| 467 | (!(flags & FAULT_FLAG_TRIED)); |
| 468 | } |
| 469 | |
| 470 | #define FAULT_FLAG_TRACE \ |
| 471 | { FAULT_FLAG_WRITE, "WRITE" }, \ |
| 472 | { FAULT_FLAG_MKWRITE, "MKWRITE" }, \ |
| 473 | { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \ |
| 474 | { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \ |
| 475 | { FAULT_FLAG_KILLABLE, "KILLABLE" }, \ |
| 476 | { FAULT_FLAG_TRIED, "TRIED" }, \ |
| 477 | { FAULT_FLAG_USER, "USER" }, \ |
| 478 | { FAULT_FLAG_REMOTE, "REMOTE" }, \ |
| 479 | { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \ |
| 480 | { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" } |
| 481 | |
| 482 | /* |
| 483 | * vm_fault is filled by the pagefault handler and passed to the vma's |
| 484 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
| 485 | * of VM_FAULT_xxx flags that give details about how the fault was handled. |
| 486 | * |
| 487 | * MM layer fills up gfp_mask for page allocations but fault handler might |
| 488 | * alter it if its implementation requires a different allocation context. |
| 489 | * |
| 490 | * pgoff should be used in favour of virtual_address, if possible. |
| 491 | */ |
| 492 | struct vm_fault { |
| 493 | struct vm_area_struct *vma; /* Target VMA */ |
| 494 | unsigned int flags; /* FAULT_FLAG_xxx flags */ |
| 495 | gfp_t gfp_mask; /* gfp mask to be used for allocations */ |
| 496 | pgoff_t pgoff; /* Logical page offset based on vma */ |
| 497 | unsigned long address; /* Faulting virtual address */ |
| 498 | pmd_t *pmd; /* Pointer to pmd entry matching |
| 499 | * the 'address' */ |
| 500 | pud_t *pud; /* Pointer to pud entry matching |
| 501 | * the 'address' |
| 502 | */ |
| 503 | pte_t orig_pte; /* Value of PTE at the time of fault */ |
| 504 | |
| 505 | struct page *cow_page; /* Page handler may use for COW fault */ |
| 506 | struct page *page; /* ->fault handlers should return a |
| 507 | * page here, unless VM_FAULT_NOPAGE |
| 508 | * is set (which is also implied by |
| 509 | * VM_FAULT_ERROR). |
| 510 | */ |
| 511 | /* These three entries are valid only while holding ptl lock */ |
| 512 | pte_t *pte; /* Pointer to pte entry matching |
| 513 | * the 'address'. NULL if the page |
| 514 | * table hasn't been allocated. |
| 515 | */ |
| 516 | spinlock_t *ptl; /* Page table lock. |
| 517 | * Protects pte page table if 'pte' |
| 518 | * is not NULL, otherwise pmd. |
| 519 | */ |
| 520 | pgtable_t prealloc_pte; /* Pre-allocated pte page table. |
| 521 | * vm_ops->map_pages() calls |
| 522 | * alloc_set_pte() from atomic context. |
| 523 | * do_fault_around() pre-allocates |
| 524 | * page table to avoid allocation from |
| 525 | * atomic context. |
| 526 | */ |
| 527 | }; |
| 528 | |
| 529 | /* page entry size for vm->huge_fault() */ |
| 530 | enum page_entry_size { |
| 531 | PE_SIZE_PTE = 0, |
| 532 | PE_SIZE_PMD, |
| 533 | PE_SIZE_PUD, |
| 534 | }; |
| 535 | |
| 536 | /* |
| 537 | * These are the virtual MM functions - opening of an area, closing and |
| 538 | * unmapping it (needed to keep files on disk up-to-date etc), pointer |
| 539 | * to the functions called when a no-page or a wp-page exception occurs. |
| 540 | */ |
| 541 | struct vm_operations_struct { |
| 542 | void (*open)(struct vm_area_struct * area); |
| 543 | void (*close)(struct vm_area_struct * area); |
| 544 | int (*split)(struct vm_area_struct * area, unsigned long addr); |
| 545 | int (*mremap)(struct vm_area_struct * area); |
| 546 | vm_fault_t (*fault)(struct vm_fault *vmf); |
| 547 | vm_fault_t (*huge_fault)(struct vm_fault *vmf, |
| 548 | enum page_entry_size pe_size); |
| 549 | void (*map_pages)(struct vm_fault *vmf, |
| 550 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
| 551 | unsigned long (*pagesize)(struct vm_area_struct * area); |
| 552 | |
| 553 | /* notification that a previously read-only page is about to become |
| 554 | * writable, if an error is returned it will cause a SIGBUS */ |
| 555 | vm_fault_t (*page_mkwrite)(struct vm_fault *vmf); |
| 556 | |
| 557 | /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ |
| 558 | vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf); |
| 559 | |
| 560 | /* called by access_process_vm when get_user_pages() fails, typically |
| 561 | * for use by special VMAs that can switch between memory and hardware |
| 562 | */ |
| 563 | int (*access)(struct vm_area_struct *vma, unsigned long addr, |
| 564 | void *buf, int len, int write); |
| 565 | |
| 566 | /* Called by the /proc/PID/maps code to ask the vma whether it |
| 567 | * has a special name. Returning non-NULL will also cause this |
| 568 | * vma to be dumped unconditionally. */ |
| 569 | const char *(*name)(struct vm_area_struct *vma); |
| 570 | |
| 571 | #ifdef CONFIG_NUMA |
| 572 | /* |
| 573 | * set_policy() op must add a reference to any non-NULL @new mempolicy |
| 574 | * to hold the policy upon return. Caller should pass NULL @new to |
| 575 | * remove a policy and fall back to surrounding context--i.e. do not |
| 576 | * install a MPOL_DEFAULT policy, nor the task or system default |
| 577 | * mempolicy. |
| 578 | */ |
| 579 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
| 580 | |
| 581 | /* |
| 582 | * get_policy() op must add reference [mpol_get()] to any policy at |
| 583 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure |
| 584 | * in mm/mempolicy.c will do this automatically. |
| 585 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not |
| 586 | * marked as MPOL_SHARED. vma policies are protected by the mmap_lock. |
| 587 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op |
| 588 | * must return NULL--i.e., do not "fallback" to task or system default |
| 589 | * policy. |
| 590 | */ |
| 591 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
| 592 | unsigned long addr); |
| 593 | #endif |
| 594 | /* |
| 595 | * Called by vm_normal_page() for special PTEs to find the |
| 596 | * page for @addr. This is useful if the default behavior |
| 597 | * (using pte_page()) would not find the correct page. |
| 598 | */ |
| 599 | struct page *(*find_special_page)(struct vm_area_struct *vma, |
| 600 | unsigned long addr); |
| 601 | }; |
| 602 | |
| 603 | static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm) |
| 604 | { |
| 605 | static const struct vm_operations_struct dummy_vm_ops = {}; |
| 606 | |
| 607 | memset(vma, 0, sizeof(*vma)); |
| 608 | vma->vm_mm = mm; |
| 609 | vma->vm_ops = &dummy_vm_ops; |
| 610 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
| 611 | } |
| 612 | |
| 613 | static inline void vma_set_anonymous(struct vm_area_struct *vma) |
| 614 | { |
| 615 | vma->vm_ops = NULL; |
| 616 | } |
| 617 | |
| 618 | static inline bool vma_is_anonymous(struct vm_area_struct *vma) |
| 619 | { |
| 620 | return !vma->vm_ops; |
| 621 | } |
| 622 | |
| 623 | static inline bool vma_is_temporary_stack(struct vm_area_struct *vma) |
| 624 | { |
| 625 | int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP); |
| 626 | |
| 627 | if (!maybe_stack) |
| 628 | return false; |
| 629 | |
| 630 | if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) == |
| 631 | VM_STACK_INCOMPLETE_SETUP) |
| 632 | return true; |
| 633 | |
| 634 | return false; |
| 635 | } |
| 636 | |
| 637 | static inline bool vma_is_foreign(struct vm_area_struct *vma) |
| 638 | { |
| 639 | if (!current->mm) |
| 640 | return true; |
| 641 | |
| 642 | if (current->mm != vma->vm_mm) |
| 643 | return true; |
| 644 | |
| 645 | return false; |
| 646 | } |
| 647 | |
| 648 | static inline bool vma_is_accessible(struct vm_area_struct *vma) |
| 649 | { |
| 650 | return vma->vm_flags & VM_ACCESS_FLAGS; |
| 651 | } |
| 652 | |
| 653 | #ifdef CONFIG_SHMEM |
| 654 | /* |
| 655 | * The vma_is_shmem is not inline because it is used only by slow |
| 656 | * paths in userfault. |
| 657 | */ |
| 658 | bool vma_is_shmem(struct vm_area_struct *vma); |
| 659 | #else |
| 660 | static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; } |
| 661 | #endif |
| 662 | |
| 663 | int vma_is_stack_for_current(struct vm_area_struct *vma); |
| 664 | |
| 665 | /* flush_tlb_range() takes a vma, not a mm, and can care about flags */ |
| 666 | #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) } |
| 667 | |
| 668 | struct mmu_gather; |
| 669 | struct inode; |
| 670 | |
| 671 | /* |
| 672 | * FIXME: take this include out, include page-flags.h in |
| 673 | * files which need it (119 of them) |
| 674 | */ |
| 675 | #include <linux/page-flags.h> |
| 676 | #include <linux/huge_mm.h> |
| 677 | |
| 678 | /* |
| 679 | * Methods to modify the page usage count. |
| 680 | * |
| 681 | * What counts for a page usage: |
| 682 | * - cache mapping (page->mapping) |
| 683 | * - private data (page->private) |
| 684 | * - page mapped in a task's page tables, each mapping |
| 685 | * is counted separately |
| 686 | * |
| 687 | * Also, many kernel routines increase the page count before a critical |
| 688 | * routine so they can be sure the page doesn't go away from under them. |
| 689 | */ |
| 690 | |
| 691 | /* |
| 692 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
| 693 | */ |
| 694 | static inline int put_page_testzero(struct page *page) |
| 695 | { |
| 696 | VM_BUG_ON_PAGE(page_ref_count(page) == 0, page); |
| 697 | return page_ref_dec_and_test(page); |
| 698 | } |
| 699 | |
| 700 | /* |
| 701 | * Try to grab a ref unless the page has a refcount of zero, return false if |
| 702 | * that is the case. |
| 703 | * This can be called when MMU is off so it must not access |
| 704 | * any of the virtual mappings. |
| 705 | */ |
| 706 | static inline int get_page_unless_zero(struct page *page) |
| 707 | { |
| 708 | return page_ref_add_unless(page, 1, 0); |
| 709 | } |
| 710 | |
| 711 | extern int page_is_ram(unsigned long pfn); |
| 712 | |
| 713 | enum { |
| 714 | REGION_INTERSECTS, |
| 715 | REGION_DISJOINT, |
| 716 | REGION_MIXED, |
| 717 | }; |
| 718 | |
| 719 | int region_intersects(resource_size_t offset, size_t size, unsigned long flags, |
| 720 | unsigned long desc); |
| 721 | |
| 722 | /* Support for virtually mapped pages */ |
| 723 | struct page *vmalloc_to_page(const void *addr); |
| 724 | unsigned long vmalloc_to_pfn(const void *addr); |
| 725 | |
| 726 | /* |
| 727 | * Determine if an address is within the vmalloc range |
| 728 | * |
| 729 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there |
| 730 | * is no special casing required. |
| 731 | */ |
| 732 | |
| 733 | #ifndef is_ioremap_addr |
| 734 | #define is_ioremap_addr(x) is_vmalloc_addr(x) |
| 735 | #endif |
| 736 | |
| 737 | #ifdef CONFIG_MMU |
| 738 | extern bool is_vmalloc_addr(const void *x); |
| 739 | extern int is_vmalloc_or_module_addr(const void *x); |
| 740 | #else |
| 741 | static inline bool is_vmalloc_addr(const void *x) |
| 742 | { |
| 743 | return false; |
| 744 | } |
| 745 | static inline int is_vmalloc_or_module_addr(const void *x) |
| 746 | { |
| 747 | return 0; |
| 748 | } |
| 749 | #endif |
| 750 | |
| 751 | extern void *kvmalloc_node(size_t size, gfp_t flags, int node); |
| 752 | static inline void *kvmalloc(size_t size, gfp_t flags) |
| 753 | { |
| 754 | return kvmalloc_node(size, flags, NUMA_NO_NODE); |
| 755 | } |
| 756 | static inline void *kvzalloc_node(size_t size, gfp_t flags, int node) |
| 757 | { |
| 758 | return kvmalloc_node(size, flags | __GFP_ZERO, node); |
| 759 | } |
| 760 | static inline void *kvzalloc(size_t size, gfp_t flags) |
| 761 | { |
| 762 | return kvmalloc(size, flags | __GFP_ZERO); |
| 763 | } |
| 764 | |
| 765 | static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags) |
| 766 | { |
| 767 | size_t bytes; |
| 768 | |
| 769 | if (unlikely(check_mul_overflow(n, size, &bytes))) |
| 770 | return NULL; |
| 771 | |
| 772 | return kvmalloc(bytes, flags); |
| 773 | } |
| 774 | |
| 775 | static inline void *kvcalloc(size_t n, size_t size, gfp_t flags) |
| 776 | { |
| 777 | return kvmalloc_array(n, size, flags | __GFP_ZERO); |
| 778 | } |
| 779 | |
| 780 | extern void kvfree(const void *addr); |
| 781 | extern void kvfree_sensitive(const void *addr, size_t len); |
| 782 | |
| 783 | static inline int head_mapcount(struct page *head) |
| 784 | { |
| 785 | return atomic_read(compound_mapcount_ptr(head)) + 1; |
| 786 | } |
| 787 | |
| 788 | /* |
| 789 | * Mapcount of compound page as a whole, does not include mapped sub-pages. |
| 790 | * |
| 791 | * Must be called only for compound pages or any their tail sub-pages. |
| 792 | */ |
| 793 | static inline int compound_mapcount(struct page *page) |
| 794 | { |
| 795 | VM_BUG_ON_PAGE(!PageCompound(page), page); |
| 796 | page = compound_head(page); |
| 797 | return head_mapcount(page); |
| 798 | } |
| 799 | |
| 800 | /* |
| 801 | * The atomic page->_mapcount, starts from -1: so that transitions |
| 802 | * both from it and to it can be tracked, using atomic_inc_and_test |
| 803 | * and atomic_add_negative(-1). |
| 804 | */ |
| 805 | static inline void page_mapcount_reset(struct page *page) |
| 806 | { |
| 807 | atomic_set(&(page)->_mapcount, -1); |
| 808 | } |
| 809 | |
| 810 | int __page_mapcount(struct page *page); |
| 811 | |
| 812 | /* |
| 813 | * Mapcount of 0-order page; when compound sub-page, includes |
| 814 | * compound_mapcount(). |
| 815 | * |
| 816 | * Result is undefined for pages which cannot be mapped into userspace. |
| 817 | * For example SLAB or special types of pages. See function page_has_type(). |
| 818 | * They use this place in struct page differently. |
| 819 | */ |
| 820 | static inline int page_mapcount(struct page *page) |
| 821 | { |
| 822 | if (unlikely(PageCompound(page))) |
| 823 | return __page_mapcount(page); |
| 824 | return atomic_read(&page->_mapcount) + 1; |
| 825 | } |
| 826 | |
| 827 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 828 | int total_mapcount(struct page *page); |
| 829 | int page_trans_huge_mapcount(struct page *page, int *total_mapcount); |
| 830 | #else |
| 831 | static inline int total_mapcount(struct page *page) |
| 832 | { |
| 833 | return page_mapcount(page); |
| 834 | } |
| 835 | static inline int page_trans_huge_mapcount(struct page *page, |
| 836 | int *total_mapcount) |
| 837 | { |
| 838 | int mapcount = page_mapcount(page); |
| 839 | if (total_mapcount) |
| 840 | *total_mapcount = mapcount; |
| 841 | return mapcount; |
| 842 | } |
| 843 | #endif |
| 844 | |
| 845 | static inline struct page *virt_to_head_page(const void *x) |
| 846 | { |
| 847 | struct page *page = virt_to_page(x); |
| 848 | |
| 849 | return compound_head(page); |
| 850 | } |
| 851 | |
| 852 | void __put_page(struct page *page); |
| 853 | |
| 854 | void put_pages_list(struct list_head *pages); |
| 855 | |
| 856 | void split_page(struct page *page, unsigned int order); |
| 857 | |
| 858 | /* |
| 859 | * Compound pages have a destructor function. Provide a |
| 860 | * prototype for that function and accessor functions. |
| 861 | * These are _only_ valid on the head of a compound page. |
| 862 | */ |
| 863 | typedef void compound_page_dtor(struct page *); |
| 864 | |
| 865 | /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */ |
| 866 | enum compound_dtor_id { |
| 867 | NULL_COMPOUND_DTOR, |
| 868 | COMPOUND_PAGE_DTOR, |
| 869 | #ifdef CONFIG_HUGETLB_PAGE |
| 870 | HUGETLB_PAGE_DTOR, |
| 871 | #endif |
| 872 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 873 | TRANSHUGE_PAGE_DTOR, |
| 874 | #endif |
| 875 | NR_COMPOUND_DTORS, |
| 876 | }; |
| 877 | extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS]; |
| 878 | |
| 879 | static inline void set_compound_page_dtor(struct page *page, |
| 880 | enum compound_dtor_id compound_dtor) |
| 881 | { |
| 882 | VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page); |
| 883 | page[1].compound_dtor = compound_dtor; |
| 884 | } |
| 885 | |
| 886 | static inline void destroy_compound_page(struct page *page) |
| 887 | { |
| 888 | VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page); |
| 889 | compound_page_dtors[page[1].compound_dtor](page); |
| 890 | } |
| 891 | |
| 892 | static inline unsigned int compound_order(struct page *page) |
| 893 | { |
| 894 | if (!PageHead(page)) |
| 895 | return 0; |
| 896 | return page[1].compound_order; |
| 897 | } |
| 898 | |
| 899 | static inline bool hpage_pincount_available(struct page *page) |
| 900 | { |
| 901 | /* |
| 902 | * Can the page->hpage_pinned_refcount field be used? That field is in |
| 903 | * the 3rd page of the compound page, so the smallest (2-page) compound |
| 904 | * pages cannot support it. |
| 905 | */ |
| 906 | page = compound_head(page); |
| 907 | return PageCompound(page) && compound_order(page) > 1; |
| 908 | } |
| 909 | |
| 910 | static inline int head_pincount(struct page *head) |
| 911 | { |
| 912 | return atomic_read(compound_pincount_ptr(head)); |
| 913 | } |
| 914 | |
| 915 | static inline int compound_pincount(struct page *page) |
| 916 | { |
| 917 | VM_BUG_ON_PAGE(!hpage_pincount_available(page), page); |
| 918 | page = compound_head(page); |
| 919 | return head_pincount(page); |
| 920 | } |
| 921 | |
| 922 | static inline void set_compound_order(struct page *page, unsigned int order) |
| 923 | { |
| 924 | page[1].compound_order = order; |
| 925 | page[1].compound_nr = 1U << order; |
| 926 | } |
| 927 | |
| 928 | /* Returns the number of pages in this potentially compound page. */ |
| 929 | static inline unsigned long compound_nr(struct page *page) |
| 930 | { |
| 931 | if (!PageHead(page)) |
| 932 | return 1; |
| 933 | return page[1].compound_nr; |
| 934 | } |
| 935 | |
| 936 | /* Returns the number of bytes in this potentially compound page. */ |
| 937 | static inline unsigned long page_size(struct page *page) |
| 938 | { |
| 939 | return PAGE_SIZE << compound_order(page); |
| 940 | } |
| 941 | |
| 942 | /* Returns the number of bits needed for the number of bytes in a page */ |
| 943 | static inline unsigned int page_shift(struct page *page) |
| 944 | { |
| 945 | return PAGE_SHIFT + compound_order(page); |
| 946 | } |
| 947 | |
| 948 | void free_compound_page(struct page *page); |
| 949 | |
| 950 | #ifdef CONFIG_MMU |
| 951 | /* |
| 952 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when |
| 953 | * servicing faults for write access. In the normal case, do always want |
| 954 | * pte_mkwrite. But get_user_pages can cause write faults for mappings |
| 955 | * that do not have writing enabled, when used by access_process_vm. |
| 956 | */ |
| 957 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) |
| 958 | { |
| 959 | if (likely(vma->vm_flags & VM_WRITE)) |
| 960 | pte = pte_mkwrite(pte); |
| 961 | return pte; |
| 962 | } |
| 963 | |
| 964 | vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct page *page); |
| 965 | vm_fault_t finish_fault(struct vm_fault *vmf); |
| 966 | vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf); |
| 967 | #endif |
| 968 | |
| 969 | /* |
| 970 | * Multiple processes may "see" the same page. E.g. for untouched |
| 971 | * mappings of /dev/null, all processes see the same page full of |
| 972 | * zeroes, and text pages of executables and shared libraries have |
| 973 | * only one copy in memory, at most, normally. |
| 974 | * |
| 975 | * For the non-reserved pages, page_count(page) denotes a reference count. |
| 976 | * page_count() == 0 means the page is free. page->lru is then used for |
| 977 | * freelist management in the buddy allocator. |
| 978 | * page_count() > 0 means the page has been allocated. |
| 979 | * |
| 980 | * Pages are allocated by the slab allocator in order to provide memory |
| 981 | * to kmalloc and kmem_cache_alloc. In this case, the management of the |
| 982 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c |
| 983 | * unless a particular usage is carefully commented. (the responsibility of |
| 984 | * freeing the kmalloc memory is the caller's, of course). |
| 985 | * |
| 986 | * A page may be used by anyone else who does a __get_free_page(). |
| 987 | * In this case, page_count still tracks the references, and should only |
| 988 | * be used through the normal accessor functions. The top bits of page->flags |
| 989 | * and page->virtual store page management information, but all other fields |
| 990 | * are unused and could be used privately, carefully. The management of this |
| 991 | * page is the responsibility of the one who allocated it, and those who have |
| 992 | * subsequently been given references to it. |
| 993 | * |
| 994 | * The other pages (we may call them "pagecache pages") are completely |
| 995 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
| 996 | * The following discussion applies only to them. |
| 997 | * |
| 998 | * A pagecache page contains an opaque `private' member, which belongs to the |
| 999 | * page's address_space. Usually, this is the address of a circular list of |
| 1000 | * the page's disk buffers. PG_private must be set to tell the VM to call |
| 1001 | * into the filesystem to release these pages. |
| 1002 | * |
| 1003 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
| 1004 | * is the pointer to the inode, and page->index is the file offset of the page, |
| 1005 | * in units of PAGE_SIZE. |
| 1006 | * |
| 1007 | * If pagecache pages are not associated with an inode, they are said to be |
| 1008 | * anonymous pages. These may become associated with the swapcache, and in that |
| 1009 | * case PG_swapcache is set, and page->private is an offset into the swapcache. |
| 1010 | * |
| 1011 | * In either case (swapcache or inode backed), the pagecache itself holds one |
| 1012 | * reference to the page. Setting PG_private should also increment the |
| 1013 | * refcount. The each user mapping also has a reference to the page. |
| 1014 | * |
| 1015 | * The pagecache pages are stored in a per-mapping radix tree, which is |
| 1016 | * rooted at mapping->i_pages, and indexed by offset. |
| 1017 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
| 1018 | * lists, we instead now tag pages as dirty/writeback in the radix tree. |
| 1019 | * |
| 1020 | * All pagecache pages may be subject to I/O: |
| 1021 | * - inode pages may need to be read from disk, |
| 1022 | * - inode pages which have been modified and are MAP_SHARED may need |
| 1023 | * to be written back to the inode on disk, |
| 1024 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been |
| 1025 | * modified may need to be swapped out to swap space and (later) to be read |
| 1026 | * back into memory. |
| 1027 | */ |
| 1028 | |
| 1029 | /* |
| 1030 | * The zone field is never updated after free_area_init_core() |
| 1031 | * sets it, so none of the operations on it need to be atomic. |
| 1032 | */ |
| 1033 | |
| 1034 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ |
| 1035 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
| 1036 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
| 1037 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) |
| 1038 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) |
| 1039 | #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH) |
| 1040 | |
| 1041 | /* |
| 1042 | * Define the bit shifts to access each section. For non-existent |
| 1043 | * sections we define the shift as 0; that plus a 0 mask ensures |
| 1044 | * the compiler will optimise away reference to them. |
| 1045 | */ |
| 1046 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
| 1047 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) |
| 1048 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) |
| 1049 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) |
| 1050 | #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0)) |
| 1051 | |
| 1052 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ |
| 1053 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
| 1054 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
| 1055 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
| 1056 | SECTIONS_PGOFF : ZONES_PGOFF) |
| 1057 | #else |
| 1058 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
| 1059 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
| 1060 | NODES_PGOFF : ZONES_PGOFF) |
| 1061 | #endif |
| 1062 | |
| 1063 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
| 1064 | |
| 1065 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
| 1066 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) |
| 1067 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) |
| 1068 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1) |
| 1069 | #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1) |
| 1070 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
| 1071 | |
| 1072 | static inline enum zone_type page_zonenum(const struct page *page) |
| 1073 | { |
| 1074 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
| 1075 | } |
| 1076 | |
| 1077 | #ifdef CONFIG_ZONE_DEVICE |
| 1078 | static inline bool is_zone_device_page(const struct page *page) |
| 1079 | { |
| 1080 | return page_zonenum(page) == ZONE_DEVICE; |
| 1081 | } |
| 1082 | extern void memmap_init_zone_device(struct zone *, unsigned long, |
| 1083 | unsigned long, struct dev_pagemap *); |
| 1084 | #else |
| 1085 | static inline bool is_zone_device_page(const struct page *page) |
| 1086 | { |
| 1087 | return false; |
| 1088 | } |
| 1089 | #endif |
| 1090 | |
| 1091 | #ifdef CONFIG_DEV_PAGEMAP_OPS |
| 1092 | void free_devmap_managed_page(struct page *page); |
| 1093 | DECLARE_STATIC_KEY_FALSE(devmap_managed_key); |
| 1094 | |
| 1095 | static inline bool page_is_devmap_managed(struct page *page) |
| 1096 | { |
| 1097 | if (!static_branch_unlikely(&devmap_managed_key)) |
| 1098 | return false; |
| 1099 | if (!is_zone_device_page(page)) |
| 1100 | return false; |
| 1101 | switch (page->pgmap->type) { |
| 1102 | case MEMORY_DEVICE_PRIVATE: |
| 1103 | case MEMORY_DEVICE_FS_DAX: |
| 1104 | return true; |
| 1105 | default: |
| 1106 | break; |
| 1107 | } |
| 1108 | return false; |
| 1109 | } |
| 1110 | |
| 1111 | void put_devmap_managed_page(struct page *page); |
| 1112 | |
| 1113 | #else /* CONFIG_DEV_PAGEMAP_OPS */ |
| 1114 | static inline bool page_is_devmap_managed(struct page *page) |
| 1115 | { |
| 1116 | return false; |
| 1117 | } |
| 1118 | |
| 1119 | static inline void put_devmap_managed_page(struct page *page) |
| 1120 | { |
| 1121 | } |
| 1122 | #endif /* CONFIG_DEV_PAGEMAP_OPS */ |
| 1123 | |
| 1124 | static inline bool is_device_private_page(const struct page *page) |
| 1125 | { |
| 1126 | return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) && |
| 1127 | IS_ENABLED(CONFIG_DEVICE_PRIVATE) && |
| 1128 | is_zone_device_page(page) && |
| 1129 | page->pgmap->type == MEMORY_DEVICE_PRIVATE; |
| 1130 | } |
| 1131 | |
| 1132 | static inline bool is_pci_p2pdma_page(const struct page *page) |
| 1133 | { |
| 1134 | return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) && |
| 1135 | IS_ENABLED(CONFIG_PCI_P2PDMA) && |
| 1136 | is_zone_device_page(page) && |
| 1137 | page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA; |
| 1138 | } |
| 1139 | |
| 1140 | /* 127: arbitrary random number, small enough to assemble well */ |
| 1141 | #define page_ref_zero_or_close_to_overflow(page) \ |
| 1142 | ((unsigned int) page_ref_count(page) + 127u <= 127u) |
| 1143 | |
| 1144 | static inline void get_page(struct page *page) |
| 1145 | { |
| 1146 | page = compound_head(page); |
| 1147 | /* |
| 1148 | * Getting a normal page or the head of a compound page |
| 1149 | * requires to already have an elevated page->_refcount. |
| 1150 | */ |
| 1151 | VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page); |
| 1152 | page_ref_inc(page); |
| 1153 | } |
| 1154 | |
| 1155 | bool __must_check try_grab_page(struct page *page, unsigned int flags); |
| 1156 | |
| 1157 | static inline __must_check bool try_get_page(struct page *page) |
| 1158 | { |
| 1159 | page = compound_head(page); |
| 1160 | if (WARN_ON_ONCE(page_ref_count(page) <= 0)) |
| 1161 | return false; |
| 1162 | page_ref_inc(page); |
| 1163 | return true; |
| 1164 | } |
| 1165 | |
| 1166 | static inline void put_page(struct page *page) |
| 1167 | { |
| 1168 | page = compound_head(page); |
| 1169 | |
| 1170 | /* |
| 1171 | * For devmap managed pages we need to catch refcount transition from |
| 1172 | * 2 to 1, when refcount reach one it means the page is free and we |
| 1173 | * need to inform the device driver through callback. See |
| 1174 | * include/linux/memremap.h and HMM for details. |
| 1175 | */ |
| 1176 | if (page_is_devmap_managed(page)) { |
| 1177 | put_devmap_managed_page(page); |
| 1178 | return; |
| 1179 | } |
| 1180 | |
| 1181 | if (put_page_testzero(page)) |
| 1182 | __put_page(page); |
| 1183 | } |
| 1184 | |
| 1185 | /* |
| 1186 | * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload |
| 1187 | * the page's refcount so that two separate items are tracked: the original page |
| 1188 | * reference count, and also a new count of how many pin_user_pages() calls were |
| 1189 | * made against the page. ("gup-pinned" is another term for the latter). |
| 1190 | * |
| 1191 | * With this scheme, pin_user_pages() becomes special: such pages are marked as |
| 1192 | * distinct from normal pages. As such, the unpin_user_page() call (and its |
| 1193 | * variants) must be used in order to release gup-pinned pages. |
| 1194 | * |
| 1195 | * Choice of value: |
| 1196 | * |
| 1197 | * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference |
| 1198 | * counts with respect to pin_user_pages() and unpin_user_page() becomes |
| 1199 | * simpler, due to the fact that adding an even power of two to the page |
| 1200 | * refcount has the effect of using only the upper N bits, for the code that |
| 1201 | * counts up using the bias value. This means that the lower bits are left for |
| 1202 | * the exclusive use of the original code that increments and decrements by one |
| 1203 | * (or at least, by much smaller values than the bias value). |
| 1204 | * |
| 1205 | * Of course, once the lower bits overflow into the upper bits (and this is |
| 1206 | * OK, because subtraction recovers the original values), then visual inspection |
| 1207 | * no longer suffices to directly view the separate counts. However, for normal |
| 1208 | * applications that don't have huge page reference counts, this won't be an |
| 1209 | * issue. |
| 1210 | * |
| 1211 | * Locking: the lockless algorithm described in page_cache_get_speculative() |
| 1212 | * and page_cache_gup_pin_speculative() provides safe operation for |
| 1213 | * get_user_pages and page_mkclean and other calls that race to set up page |
| 1214 | * table entries. |
| 1215 | */ |
| 1216 | #define GUP_PIN_COUNTING_BIAS (1U << 10) |
| 1217 | |
| 1218 | void unpin_user_page(struct page *page); |
| 1219 | void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages, |
| 1220 | bool make_dirty); |
| 1221 | void unpin_user_pages(struct page **pages, unsigned long npages); |
| 1222 | |
| 1223 | /** |
| 1224 | * page_maybe_dma_pinned() - report if a page is pinned for DMA. |
| 1225 | * |
| 1226 | * This function checks if a page has been pinned via a call to |
| 1227 | * pin_user_pages*(). |
| 1228 | * |
| 1229 | * For non-huge pages, the return value is partially fuzzy: false is not fuzzy, |
| 1230 | * because it means "definitely not pinned for DMA", but true means "probably |
| 1231 | * pinned for DMA, but possibly a false positive due to having at least |
| 1232 | * GUP_PIN_COUNTING_BIAS worth of normal page references". |
| 1233 | * |
| 1234 | * False positives are OK, because: a) it's unlikely for a page to get that many |
| 1235 | * refcounts, and b) all the callers of this routine are expected to be able to |
| 1236 | * deal gracefully with a false positive. |
| 1237 | * |
| 1238 | * For huge pages, the result will be exactly correct. That's because we have |
| 1239 | * more tracking data available: the 3rd struct page in the compound page is |
| 1240 | * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS |
| 1241 | * scheme). |
| 1242 | * |
| 1243 | * For more information, please see Documentation/core-api/pin_user_pages.rst. |
| 1244 | * |
| 1245 | * @page: pointer to page to be queried. |
| 1246 | * @Return: True, if it is likely that the page has been "dma-pinned". |
| 1247 | * False, if the page is definitely not dma-pinned. |
| 1248 | */ |
| 1249 | static inline bool page_maybe_dma_pinned(struct page *page) |
| 1250 | { |
| 1251 | if (hpage_pincount_available(page)) |
| 1252 | return compound_pincount(page) > 0; |
| 1253 | |
| 1254 | /* |
| 1255 | * page_ref_count() is signed. If that refcount overflows, then |
| 1256 | * page_ref_count() returns a negative value, and callers will avoid |
| 1257 | * further incrementing the refcount. |
| 1258 | * |
| 1259 | * Here, for that overflow case, use the signed bit to count a little |
| 1260 | * bit higher via unsigned math, and thus still get an accurate result. |
| 1261 | */ |
| 1262 | return ((unsigned int)page_ref_count(compound_head(page))) >= |
| 1263 | GUP_PIN_COUNTING_BIAS; |
| 1264 | } |
| 1265 | |
| 1266 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
| 1267 | #define SECTION_IN_PAGE_FLAGS |
| 1268 | #endif |
| 1269 | |
| 1270 | /* |
| 1271 | * The identification function is mainly used by the buddy allocator for |
| 1272 | * determining if two pages could be buddies. We are not really identifying |
| 1273 | * the zone since we could be using the section number id if we do not have |
| 1274 | * node id available in page flags. |
| 1275 | * We only guarantee that it will return the same value for two combinable |
| 1276 | * pages in a zone. |
| 1277 | */ |
| 1278 | static inline int page_zone_id(struct page *page) |
| 1279 | { |
| 1280 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
| 1281 | } |
| 1282 | |
| 1283 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
| 1284 | extern int page_to_nid(const struct page *page); |
| 1285 | #else |
| 1286 | static inline int page_to_nid(const struct page *page) |
| 1287 | { |
| 1288 | struct page *p = (struct page *)page; |
| 1289 | |
| 1290 | return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK; |
| 1291 | } |
| 1292 | #endif |
| 1293 | |
| 1294 | #ifdef CONFIG_NUMA_BALANCING |
| 1295 | static inline int cpu_pid_to_cpupid(int cpu, int pid) |
| 1296 | { |
| 1297 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); |
| 1298 | } |
| 1299 | |
| 1300 | static inline int cpupid_to_pid(int cpupid) |
| 1301 | { |
| 1302 | return cpupid & LAST__PID_MASK; |
| 1303 | } |
| 1304 | |
| 1305 | static inline int cpupid_to_cpu(int cpupid) |
| 1306 | { |
| 1307 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; |
| 1308 | } |
| 1309 | |
| 1310 | static inline int cpupid_to_nid(int cpupid) |
| 1311 | { |
| 1312 | return cpu_to_node(cpupid_to_cpu(cpupid)); |
| 1313 | } |
| 1314 | |
| 1315 | static inline bool cpupid_pid_unset(int cpupid) |
| 1316 | { |
| 1317 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); |
| 1318 | } |
| 1319 | |
| 1320 | static inline bool cpupid_cpu_unset(int cpupid) |
| 1321 | { |
| 1322 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); |
| 1323 | } |
| 1324 | |
| 1325 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) |
| 1326 | { |
| 1327 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); |
| 1328 | } |
| 1329 | |
| 1330 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) |
| 1331 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
| 1332 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) |
| 1333 | { |
| 1334 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); |
| 1335 | } |
| 1336 | |
| 1337 | static inline int page_cpupid_last(struct page *page) |
| 1338 | { |
| 1339 | return page->_last_cpupid; |
| 1340 | } |
| 1341 | static inline void page_cpupid_reset_last(struct page *page) |
| 1342 | { |
| 1343 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; |
| 1344 | } |
| 1345 | #else |
| 1346 | static inline int page_cpupid_last(struct page *page) |
| 1347 | { |
| 1348 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; |
| 1349 | } |
| 1350 | |
| 1351 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); |
| 1352 | |
| 1353 | static inline void page_cpupid_reset_last(struct page *page) |
| 1354 | { |
| 1355 | page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT; |
| 1356 | } |
| 1357 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ |
| 1358 | #else /* !CONFIG_NUMA_BALANCING */ |
| 1359 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) |
| 1360 | { |
| 1361 | return page_to_nid(page); /* XXX */ |
| 1362 | } |
| 1363 | |
| 1364 | static inline int page_cpupid_last(struct page *page) |
| 1365 | { |
| 1366 | return page_to_nid(page); /* XXX */ |
| 1367 | } |
| 1368 | |
| 1369 | static inline int cpupid_to_nid(int cpupid) |
| 1370 | { |
| 1371 | return -1; |
| 1372 | } |
| 1373 | |
| 1374 | static inline int cpupid_to_pid(int cpupid) |
| 1375 | { |
| 1376 | return -1; |
| 1377 | } |
| 1378 | |
| 1379 | static inline int cpupid_to_cpu(int cpupid) |
| 1380 | { |
| 1381 | return -1; |
| 1382 | } |
| 1383 | |
| 1384 | static inline int cpu_pid_to_cpupid(int nid, int pid) |
| 1385 | { |
| 1386 | return -1; |
| 1387 | } |
| 1388 | |
| 1389 | static inline bool cpupid_pid_unset(int cpupid) |
| 1390 | { |
| 1391 | return true; |
| 1392 | } |
| 1393 | |
| 1394 | static inline void page_cpupid_reset_last(struct page *page) |
| 1395 | { |
| 1396 | } |
| 1397 | |
| 1398 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) |
| 1399 | { |
| 1400 | return false; |
| 1401 | } |
| 1402 | #endif /* CONFIG_NUMA_BALANCING */ |
| 1403 | |
| 1404 | #ifdef CONFIG_KASAN_SW_TAGS |
| 1405 | static inline u8 page_kasan_tag(const struct page *page) |
| 1406 | { |
| 1407 | return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK; |
| 1408 | } |
| 1409 | |
| 1410 | static inline void page_kasan_tag_set(struct page *page, u8 tag) |
| 1411 | { |
| 1412 | page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT); |
| 1413 | page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT; |
| 1414 | } |
| 1415 | |
| 1416 | static inline void page_kasan_tag_reset(struct page *page) |
| 1417 | { |
| 1418 | page_kasan_tag_set(page, 0xff); |
| 1419 | } |
| 1420 | #else |
| 1421 | static inline u8 page_kasan_tag(const struct page *page) |
| 1422 | { |
| 1423 | return 0xff; |
| 1424 | } |
| 1425 | |
| 1426 | static inline void page_kasan_tag_set(struct page *page, u8 tag) { } |
| 1427 | static inline void page_kasan_tag_reset(struct page *page) { } |
| 1428 | #endif |
| 1429 | |
| 1430 | static inline struct zone *page_zone(const struct page *page) |
| 1431 | { |
| 1432 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; |
| 1433 | } |
| 1434 | |
| 1435 | static inline pg_data_t *page_pgdat(const struct page *page) |
| 1436 | { |
| 1437 | return NODE_DATA(page_to_nid(page)); |
| 1438 | } |
| 1439 | |
| 1440 | #ifdef SECTION_IN_PAGE_FLAGS |
| 1441 | static inline void set_page_section(struct page *page, unsigned long section) |
| 1442 | { |
| 1443 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); |
| 1444 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; |
| 1445 | } |
| 1446 | |
| 1447 | static inline unsigned long page_to_section(const struct page *page) |
| 1448 | { |
| 1449 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; |
| 1450 | } |
| 1451 | #endif |
| 1452 | |
| 1453 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
| 1454 | { |
| 1455 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); |
| 1456 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; |
| 1457 | } |
| 1458 | |
| 1459 | static inline void set_page_node(struct page *page, unsigned long node) |
| 1460 | { |
| 1461 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); |
| 1462 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; |
| 1463 | } |
| 1464 | |
| 1465 | static inline void set_page_links(struct page *page, enum zone_type zone, |
| 1466 | unsigned long node, unsigned long pfn) |
| 1467 | { |
| 1468 | set_page_zone(page, zone); |
| 1469 | set_page_node(page, node); |
| 1470 | #ifdef SECTION_IN_PAGE_FLAGS |
| 1471 | set_page_section(page, pfn_to_section_nr(pfn)); |
| 1472 | #endif |
| 1473 | } |
| 1474 | |
| 1475 | #ifdef CONFIG_MEMCG |
| 1476 | static inline struct mem_cgroup *page_memcg(struct page *page) |
| 1477 | { |
| 1478 | return page->mem_cgroup; |
| 1479 | } |
| 1480 | static inline struct mem_cgroup *page_memcg_rcu(struct page *page) |
| 1481 | { |
| 1482 | WARN_ON_ONCE(!rcu_read_lock_held()); |
| 1483 | return READ_ONCE(page->mem_cgroup); |
| 1484 | } |
| 1485 | #else |
| 1486 | static inline struct mem_cgroup *page_memcg(struct page *page) |
| 1487 | { |
| 1488 | return NULL; |
| 1489 | } |
| 1490 | static inline struct mem_cgroup *page_memcg_rcu(struct page *page) |
| 1491 | { |
| 1492 | WARN_ON_ONCE(!rcu_read_lock_held()); |
| 1493 | return NULL; |
| 1494 | } |
| 1495 | #endif |
| 1496 | |
| 1497 | /* |
| 1498 | * Some inline functions in vmstat.h depend on page_zone() |
| 1499 | */ |
| 1500 | #include <linux/vmstat.h> |
| 1501 | |
| 1502 | static __always_inline void *lowmem_page_address(const struct page *page) |
| 1503 | { |
| 1504 | return page_to_virt(page); |
| 1505 | } |
| 1506 | |
| 1507 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) |
| 1508 | #define HASHED_PAGE_VIRTUAL |
| 1509 | #endif |
| 1510 | |
| 1511 | #if defined(WANT_PAGE_VIRTUAL) |
| 1512 | static inline void *page_address(const struct page *page) |
| 1513 | { |
| 1514 | return page->virtual; |
| 1515 | } |
| 1516 | static inline void set_page_address(struct page *page, void *address) |
| 1517 | { |
| 1518 | page->virtual = address; |
| 1519 | } |
| 1520 | #define page_address_init() do { } while(0) |
| 1521 | #endif |
| 1522 | |
| 1523 | #if defined(HASHED_PAGE_VIRTUAL) |
| 1524 | void *page_address(const struct page *page); |
| 1525 | void set_page_address(struct page *page, void *virtual); |
| 1526 | void page_address_init(void); |
| 1527 | #endif |
| 1528 | |
| 1529 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) |
| 1530 | #define page_address(page) lowmem_page_address(page) |
| 1531 | #define set_page_address(page, address) do { } while(0) |
| 1532 | #define page_address_init() do { } while(0) |
| 1533 | #endif |
| 1534 | |
| 1535 | extern void *page_rmapping(struct page *page); |
| 1536 | extern struct anon_vma *page_anon_vma(struct page *page); |
| 1537 | extern struct address_space *page_mapping(struct page *page); |
| 1538 | |
| 1539 | extern struct address_space *__page_file_mapping(struct page *); |
| 1540 | |
| 1541 | static inline |
| 1542 | struct address_space *page_file_mapping(struct page *page) |
| 1543 | { |
| 1544 | if (unlikely(PageSwapCache(page))) |
| 1545 | return __page_file_mapping(page); |
| 1546 | |
| 1547 | return page->mapping; |
| 1548 | } |
| 1549 | |
| 1550 | extern pgoff_t __page_file_index(struct page *page); |
| 1551 | |
| 1552 | /* |
| 1553 | * Return the pagecache index of the passed page. Regular pagecache pages |
| 1554 | * use ->index whereas swapcache pages use swp_offset(->private) |
| 1555 | */ |
| 1556 | static inline pgoff_t page_index(struct page *page) |
| 1557 | { |
| 1558 | if (unlikely(PageSwapCache(page))) |
| 1559 | return __page_file_index(page); |
| 1560 | return page->index; |
| 1561 | } |
| 1562 | |
| 1563 | bool page_mapped(struct page *page); |
| 1564 | struct address_space *page_mapping(struct page *page); |
| 1565 | struct address_space *page_mapping_file(struct page *page); |
| 1566 | |
| 1567 | /* |
| 1568 | * Return true only if the page has been allocated with |
| 1569 | * ALLOC_NO_WATERMARKS and the low watermark was not |
| 1570 | * met implying that the system is under some pressure. |
| 1571 | */ |
| 1572 | static inline bool page_is_pfmemalloc(struct page *page) |
| 1573 | { |
| 1574 | /* |
| 1575 | * Page index cannot be this large so this must be |
| 1576 | * a pfmemalloc page. |
| 1577 | */ |
| 1578 | return page->index == -1UL; |
| 1579 | } |
| 1580 | |
| 1581 | /* |
| 1582 | * Only to be called by the page allocator on a freshly allocated |
| 1583 | * page. |
| 1584 | */ |
| 1585 | static inline void set_page_pfmemalloc(struct page *page) |
| 1586 | { |
| 1587 | page->index = -1UL; |
| 1588 | } |
| 1589 | |
| 1590 | static inline void clear_page_pfmemalloc(struct page *page) |
| 1591 | { |
| 1592 | page->index = 0; |
| 1593 | } |
| 1594 | |
| 1595 | /* |
| 1596 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. |
| 1597 | */ |
| 1598 | extern void pagefault_out_of_memory(void); |
| 1599 | |
| 1600 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
| 1601 | |
| 1602 | /* |
| 1603 | * Flags passed to show_mem() and show_free_areas() to suppress output in |
| 1604 | * various contexts. |
| 1605 | */ |
| 1606 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ |
| 1607 | |
| 1608 | extern void show_free_areas(unsigned int flags, nodemask_t *nodemask); |
| 1609 | |
| 1610 | #ifdef CONFIG_MMU |
| 1611 | extern bool can_do_mlock(void); |
| 1612 | #else |
| 1613 | static inline bool can_do_mlock(void) { return false; } |
| 1614 | #endif |
| 1615 | extern int user_shm_lock(size_t, struct user_struct *); |
| 1616 | extern void user_shm_unlock(size_t, struct user_struct *); |
| 1617 | |
| 1618 | /* |
| 1619 | * Parameter block passed down to zap_pte_range in exceptional cases. |
| 1620 | */ |
| 1621 | struct zap_details { |
| 1622 | struct address_space *check_mapping; /* Check page->mapping if set */ |
| 1623 | pgoff_t first_index; /* Lowest page->index to unmap */ |
| 1624 | pgoff_t last_index; /* Highest page->index to unmap */ |
| 1625 | }; |
| 1626 | |
| 1627 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
| 1628 | pte_t pte); |
| 1629 | struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr, |
| 1630 | pmd_t pmd); |
| 1631 | |
| 1632 | void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
| 1633 | unsigned long size); |
| 1634 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, |
| 1635 | unsigned long size); |
| 1636 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
| 1637 | unsigned long start, unsigned long end); |
| 1638 | |
| 1639 | struct mmu_notifier_range; |
| 1640 | |
| 1641 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
| 1642 | unsigned long end, unsigned long floor, unsigned long ceiling); |
| 1643 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
| 1644 | struct vm_area_struct *vma); |
| 1645 | int follow_pte_pmd(struct mm_struct *mm, unsigned long address, |
| 1646 | struct mmu_notifier_range *range, |
| 1647 | pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp); |
| 1648 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
| 1649 | unsigned long *pfn); |
| 1650 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
| 1651 | unsigned int flags, unsigned long *prot, resource_size_t *phys); |
| 1652 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
| 1653 | void *buf, int len, int write); |
| 1654 | |
| 1655 | extern void truncate_pagecache(struct inode *inode, loff_t new); |
| 1656 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
| 1657 | void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); |
| 1658 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); |
| 1659 | int truncate_inode_page(struct address_space *mapping, struct page *page); |
| 1660 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
| 1661 | int invalidate_inode_page(struct page *page); |
| 1662 | |
| 1663 | #ifdef CONFIG_MMU |
| 1664 | extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
| 1665 | unsigned long address, unsigned int flags, |
| 1666 | struct pt_regs *regs); |
| 1667 | extern int fixup_user_fault(struct mm_struct *mm, |
| 1668 | unsigned long address, unsigned int fault_flags, |
| 1669 | bool *unlocked); |
| 1670 | void unmap_mapping_pages(struct address_space *mapping, |
| 1671 | pgoff_t start, pgoff_t nr, bool even_cows); |
| 1672 | void unmap_mapping_range(struct address_space *mapping, |
| 1673 | loff_t const holebegin, loff_t const holelen, int even_cows); |
| 1674 | #else |
| 1675 | static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma, |
| 1676 | unsigned long address, unsigned int flags, |
| 1677 | struct pt_regs *regs) |
| 1678 | { |
| 1679 | /* should never happen if there's no MMU */ |
| 1680 | BUG(); |
| 1681 | return VM_FAULT_SIGBUS; |
| 1682 | } |
| 1683 | static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address, |
| 1684 | unsigned int fault_flags, bool *unlocked) |
| 1685 | { |
| 1686 | /* should never happen if there's no MMU */ |
| 1687 | BUG(); |
| 1688 | return -EFAULT; |
| 1689 | } |
| 1690 | static inline void unmap_mapping_pages(struct address_space *mapping, |
| 1691 | pgoff_t start, pgoff_t nr, bool even_cows) { } |
| 1692 | static inline void unmap_mapping_range(struct address_space *mapping, |
| 1693 | loff_t const holebegin, loff_t const holelen, int even_cows) { } |
| 1694 | #endif |
| 1695 | |
| 1696 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
| 1697 | loff_t const holebegin, loff_t const holelen) |
| 1698 | { |
| 1699 | unmap_mapping_range(mapping, holebegin, holelen, 0); |
| 1700 | } |
| 1701 | |
| 1702 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, |
| 1703 | void *buf, int len, unsigned int gup_flags); |
| 1704 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
| 1705 | void *buf, int len, unsigned int gup_flags); |
| 1706 | extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm, |
| 1707 | unsigned long addr, void *buf, int len, unsigned int gup_flags); |
| 1708 | |
| 1709 | long get_user_pages_remote(struct mm_struct *mm, |
| 1710 | unsigned long start, unsigned long nr_pages, |
| 1711 | unsigned int gup_flags, struct page **pages, |
| 1712 | struct vm_area_struct **vmas, int *locked); |
| 1713 | long pin_user_pages_remote(struct mm_struct *mm, |
| 1714 | unsigned long start, unsigned long nr_pages, |
| 1715 | unsigned int gup_flags, struct page **pages, |
| 1716 | struct vm_area_struct **vmas, int *locked); |
| 1717 | long get_user_pages(unsigned long start, unsigned long nr_pages, |
| 1718 | unsigned int gup_flags, struct page **pages, |
| 1719 | struct vm_area_struct **vmas); |
| 1720 | long pin_user_pages(unsigned long start, unsigned long nr_pages, |
| 1721 | unsigned int gup_flags, struct page **pages, |
| 1722 | struct vm_area_struct **vmas); |
| 1723 | long get_user_pages_locked(unsigned long start, unsigned long nr_pages, |
| 1724 | unsigned int gup_flags, struct page **pages, int *locked); |
| 1725 | long pin_user_pages_locked(unsigned long start, unsigned long nr_pages, |
| 1726 | unsigned int gup_flags, struct page **pages, int *locked); |
| 1727 | long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages, |
| 1728 | struct page **pages, unsigned int gup_flags); |
| 1729 | long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages, |
| 1730 | struct page **pages, unsigned int gup_flags); |
| 1731 | |
| 1732 | int get_user_pages_fast(unsigned long start, int nr_pages, |
| 1733 | unsigned int gup_flags, struct page **pages); |
| 1734 | int pin_user_pages_fast(unsigned long start, int nr_pages, |
| 1735 | unsigned int gup_flags, struct page **pages); |
| 1736 | |
| 1737 | int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc); |
| 1738 | int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc, |
| 1739 | struct task_struct *task, bool bypass_rlim); |
| 1740 | |
| 1741 | /* Container for pinned pfns / pages */ |
| 1742 | struct frame_vector { |
| 1743 | unsigned int nr_allocated; /* Number of frames we have space for */ |
| 1744 | unsigned int nr_frames; /* Number of frames stored in ptrs array */ |
| 1745 | bool got_ref; /* Did we pin pages by getting page ref? */ |
| 1746 | bool is_pfns; /* Does array contain pages or pfns? */ |
| 1747 | void *ptrs[]; /* Array of pinned pfns / pages. Use |
| 1748 | * pfns_vector_pages() or pfns_vector_pfns() |
| 1749 | * for access */ |
| 1750 | }; |
| 1751 | |
| 1752 | struct frame_vector *frame_vector_create(unsigned int nr_frames); |
| 1753 | void frame_vector_destroy(struct frame_vector *vec); |
| 1754 | int get_vaddr_frames(unsigned long start, unsigned int nr_pfns, |
| 1755 | unsigned int gup_flags, struct frame_vector *vec); |
| 1756 | void put_vaddr_frames(struct frame_vector *vec); |
| 1757 | int frame_vector_to_pages(struct frame_vector *vec); |
| 1758 | void frame_vector_to_pfns(struct frame_vector *vec); |
| 1759 | |
| 1760 | static inline unsigned int frame_vector_count(struct frame_vector *vec) |
| 1761 | { |
| 1762 | return vec->nr_frames; |
| 1763 | } |
| 1764 | |
| 1765 | static inline struct page **frame_vector_pages(struct frame_vector *vec) |
| 1766 | { |
| 1767 | if (vec->is_pfns) { |
| 1768 | int err = frame_vector_to_pages(vec); |
| 1769 | |
| 1770 | if (err) |
| 1771 | return ERR_PTR(err); |
| 1772 | } |
| 1773 | return (struct page **)(vec->ptrs); |
| 1774 | } |
| 1775 | |
| 1776 | static inline unsigned long *frame_vector_pfns(struct frame_vector *vec) |
| 1777 | { |
| 1778 | if (!vec->is_pfns) |
| 1779 | frame_vector_to_pfns(vec); |
| 1780 | return (unsigned long *)(vec->ptrs); |
| 1781 | } |
| 1782 | |
| 1783 | struct kvec; |
| 1784 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, |
| 1785 | struct page **pages); |
| 1786 | int get_kernel_page(unsigned long start, int write, struct page **pages); |
| 1787 | struct page *get_dump_page(unsigned long addr); |
| 1788 | |
| 1789 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
| 1790 | extern void do_invalidatepage(struct page *page, unsigned int offset, |
| 1791 | unsigned int length); |
| 1792 | |
| 1793 | void __set_page_dirty(struct page *, struct address_space *, int warn); |
| 1794 | int __set_page_dirty_nobuffers(struct page *page); |
| 1795 | int __set_page_dirty_no_writeback(struct page *page); |
| 1796 | int redirty_page_for_writepage(struct writeback_control *wbc, |
| 1797 | struct page *page); |
| 1798 | void account_page_dirtied(struct page *page, struct address_space *mapping); |
| 1799 | void account_page_cleaned(struct page *page, struct address_space *mapping, |
| 1800 | struct bdi_writeback *wb); |
| 1801 | int set_page_dirty(struct page *page); |
| 1802 | int set_page_dirty_lock(struct page *page); |
| 1803 | void __cancel_dirty_page(struct page *page); |
| 1804 | static inline void cancel_dirty_page(struct page *page) |
| 1805 | { |
| 1806 | /* Avoid atomic ops, locking, etc. when not actually needed. */ |
| 1807 | if (PageDirty(page)) |
| 1808 | __cancel_dirty_page(page); |
| 1809 | } |
| 1810 | int clear_page_dirty_for_io(struct page *page); |
| 1811 | |
| 1812 | int get_cmdline(struct task_struct *task, char *buffer, int buflen); |
| 1813 | |
| 1814 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
| 1815 | unsigned long old_addr, struct vm_area_struct *new_vma, |
| 1816 | unsigned long new_addr, unsigned long len, |
| 1817 | bool need_rmap_locks); |
| 1818 | |
| 1819 | /* |
| 1820 | * Flags used by change_protection(). For now we make it a bitmap so |
| 1821 | * that we can pass in multiple flags just like parameters. However |
| 1822 | * for now all the callers are only use one of the flags at the same |
| 1823 | * time. |
| 1824 | */ |
| 1825 | /* Whether we should allow dirty bit accounting */ |
| 1826 | #define MM_CP_DIRTY_ACCT (1UL << 0) |
| 1827 | /* Whether this protection change is for NUMA hints */ |
| 1828 | #define MM_CP_PROT_NUMA (1UL << 1) |
| 1829 | /* Whether this change is for write protecting */ |
| 1830 | #define MM_CP_UFFD_WP (1UL << 2) /* do wp */ |
| 1831 | #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */ |
| 1832 | #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \ |
| 1833 | MM_CP_UFFD_WP_RESOLVE) |
| 1834 | |
| 1835 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, |
| 1836 | unsigned long end, pgprot_t newprot, |
| 1837 | unsigned long cp_flags); |
| 1838 | extern int mprotect_fixup(struct vm_area_struct *vma, |
| 1839 | struct vm_area_struct **pprev, unsigned long start, |
| 1840 | unsigned long end, unsigned long newflags); |
| 1841 | |
| 1842 | /* |
| 1843 | * doesn't attempt to fault and will return short. |
| 1844 | */ |
| 1845 | int get_user_pages_fast_only(unsigned long start, int nr_pages, |
| 1846 | unsigned int gup_flags, struct page **pages); |
| 1847 | int pin_user_pages_fast_only(unsigned long start, int nr_pages, |
| 1848 | unsigned int gup_flags, struct page **pages); |
| 1849 | |
| 1850 | static inline bool get_user_page_fast_only(unsigned long addr, |
| 1851 | unsigned int gup_flags, struct page **pagep) |
| 1852 | { |
| 1853 | return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1; |
| 1854 | } |
| 1855 | /* |
| 1856 | * per-process(per-mm_struct) statistics. |
| 1857 | */ |
| 1858 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) |
| 1859 | { |
| 1860 | long val = atomic_long_read(&mm->rss_stat.count[member]); |
| 1861 | |
| 1862 | #ifdef SPLIT_RSS_COUNTING |
| 1863 | /* |
| 1864 | * counter is updated in asynchronous manner and may go to minus. |
| 1865 | * But it's never be expected number for users. |
| 1866 | */ |
| 1867 | if (val < 0) |
| 1868 | val = 0; |
| 1869 | #endif |
| 1870 | return (unsigned long)val; |
| 1871 | } |
| 1872 | |
| 1873 | void mm_trace_rss_stat(struct mm_struct *mm, int member, long count); |
| 1874 | |
| 1875 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) |
| 1876 | { |
| 1877 | long count = atomic_long_add_return(value, &mm->rss_stat.count[member]); |
| 1878 | |
| 1879 | mm_trace_rss_stat(mm, member, count); |
| 1880 | } |
| 1881 | |
| 1882 | static inline void inc_mm_counter(struct mm_struct *mm, int member) |
| 1883 | { |
| 1884 | long count = atomic_long_inc_return(&mm->rss_stat.count[member]); |
| 1885 | |
| 1886 | mm_trace_rss_stat(mm, member, count); |
| 1887 | } |
| 1888 | |
| 1889 | static inline void dec_mm_counter(struct mm_struct *mm, int member) |
| 1890 | { |
| 1891 | long count = atomic_long_dec_return(&mm->rss_stat.count[member]); |
| 1892 | |
| 1893 | mm_trace_rss_stat(mm, member, count); |
| 1894 | } |
| 1895 | |
| 1896 | /* Optimized variant when page is already known not to be PageAnon */ |
| 1897 | static inline int mm_counter_file(struct page *page) |
| 1898 | { |
| 1899 | if (PageSwapBacked(page)) |
| 1900 | return MM_SHMEMPAGES; |
| 1901 | return MM_FILEPAGES; |
| 1902 | } |
| 1903 | |
| 1904 | static inline int mm_counter(struct page *page) |
| 1905 | { |
| 1906 | if (PageAnon(page)) |
| 1907 | return MM_ANONPAGES; |
| 1908 | return mm_counter_file(page); |
| 1909 | } |
| 1910 | |
| 1911 | static inline unsigned long get_mm_rss(struct mm_struct *mm) |
| 1912 | { |
| 1913 | return get_mm_counter(mm, MM_FILEPAGES) + |
| 1914 | get_mm_counter(mm, MM_ANONPAGES) + |
| 1915 | get_mm_counter(mm, MM_SHMEMPAGES); |
| 1916 | } |
| 1917 | |
| 1918 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) |
| 1919 | { |
| 1920 | return max(mm->hiwater_rss, get_mm_rss(mm)); |
| 1921 | } |
| 1922 | |
| 1923 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) |
| 1924 | { |
| 1925 | return max(mm->hiwater_vm, mm->total_vm); |
| 1926 | } |
| 1927 | |
| 1928 | static inline void update_hiwater_rss(struct mm_struct *mm) |
| 1929 | { |
| 1930 | unsigned long _rss = get_mm_rss(mm); |
| 1931 | |
| 1932 | if ((mm)->hiwater_rss < _rss) |
| 1933 | (mm)->hiwater_rss = _rss; |
| 1934 | } |
| 1935 | |
| 1936 | static inline void update_hiwater_vm(struct mm_struct *mm) |
| 1937 | { |
| 1938 | if (mm->hiwater_vm < mm->total_vm) |
| 1939 | mm->hiwater_vm = mm->total_vm; |
| 1940 | } |
| 1941 | |
| 1942 | static inline void reset_mm_hiwater_rss(struct mm_struct *mm) |
| 1943 | { |
| 1944 | mm->hiwater_rss = get_mm_rss(mm); |
| 1945 | } |
| 1946 | |
| 1947 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, |
| 1948 | struct mm_struct *mm) |
| 1949 | { |
| 1950 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); |
| 1951 | |
| 1952 | if (*maxrss < hiwater_rss) |
| 1953 | *maxrss = hiwater_rss; |
| 1954 | } |
| 1955 | |
| 1956 | #if defined(SPLIT_RSS_COUNTING) |
| 1957 | void sync_mm_rss(struct mm_struct *mm); |
| 1958 | #else |
| 1959 | static inline void sync_mm_rss(struct mm_struct *mm) |
| 1960 | { |
| 1961 | } |
| 1962 | #endif |
| 1963 | |
| 1964 | #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL |
| 1965 | static inline int pte_special(pte_t pte) |
| 1966 | { |
| 1967 | return 0; |
| 1968 | } |
| 1969 | |
| 1970 | static inline pte_t pte_mkspecial(pte_t pte) |
| 1971 | { |
| 1972 | return pte; |
| 1973 | } |
| 1974 | #endif |
| 1975 | |
| 1976 | #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP |
| 1977 | static inline int pte_devmap(pte_t pte) |
| 1978 | { |
| 1979 | return 0; |
| 1980 | } |
| 1981 | #endif |
| 1982 | |
| 1983 | int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot); |
| 1984 | |
| 1985 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
| 1986 | spinlock_t **ptl); |
| 1987 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, |
| 1988 | spinlock_t **ptl) |
| 1989 | { |
| 1990 | pte_t *ptep; |
| 1991 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); |
| 1992 | return ptep; |
| 1993 | } |
| 1994 | |
| 1995 | #ifdef __PAGETABLE_P4D_FOLDED |
| 1996 | static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, |
| 1997 | unsigned long address) |
| 1998 | { |
| 1999 | return 0; |
| 2000 | } |
| 2001 | #else |
| 2002 | int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
| 2003 | #endif |
| 2004 | |
| 2005 | #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU) |
| 2006 | static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, |
| 2007 | unsigned long address) |
| 2008 | { |
| 2009 | return 0; |
| 2010 | } |
| 2011 | static inline void mm_inc_nr_puds(struct mm_struct *mm) {} |
| 2012 | static inline void mm_dec_nr_puds(struct mm_struct *mm) {} |
| 2013 | |
| 2014 | #else |
| 2015 | int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address); |
| 2016 | |
| 2017 | static inline void mm_inc_nr_puds(struct mm_struct *mm) |
| 2018 | { |
| 2019 | if (mm_pud_folded(mm)) |
| 2020 | return; |
| 2021 | atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
| 2022 | } |
| 2023 | |
| 2024 | static inline void mm_dec_nr_puds(struct mm_struct *mm) |
| 2025 | { |
| 2026 | if (mm_pud_folded(mm)) |
| 2027 | return; |
| 2028 | atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); |
| 2029 | } |
| 2030 | #endif |
| 2031 | |
| 2032 | #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU) |
| 2033 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
| 2034 | unsigned long address) |
| 2035 | { |
| 2036 | return 0; |
| 2037 | } |
| 2038 | |
| 2039 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) {} |
| 2040 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) {} |
| 2041 | |
| 2042 | #else |
| 2043 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
| 2044 | |
| 2045 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) |
| 2046 | { |
| 2047 | if (mm_pmd_folded(mm)) |
| 2048 | return; |
| 2049 | atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
| 2050 | } |
| 2051 | |
| 2052 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) |
| 2053 | { |
| 2054 | if (mm_pmd_folded(mm)) |
| 2055 | return; |
| 2056 | atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); |
| 2057 | } |
| 2058 | #endif |
| 2059 | |
| 2060 | #ifdef CONFIG_MMU |
| 2061 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) |
| 2062 | { |
| 2063 | atomic_long_set(&mm->pgtables_bytes, 0); |
| 2064 | } |
| 2065 | |
| 2066 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) |
| 2067 | { |
| 2068 | return atomic_long_read(&mm->pgtables_bytes); |
| 2069 | } |
| 2070 | |
| 2071 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) |
| 2072 | { |
| 2073 | atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
| 2074 | } |
| 2075 | |
| 2076 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) |
| 2077 | { |
| 2078 | atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); |
| 2079 | } |
| 2080 | #else |
| 2081 | |
| 2082 | static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {} |
| 2083 | static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) |
| 2084 | { |
| 2085 | return 0; |
| 2086 | } |
| 2087 | |
| 2088 | static inline void mm_inc_nr_ptes(struct mm_struct *mm) {} |
| 2089 | static inline void mm_dec_nr_ptes(struct mm_struct *mm) {} |
| 2090 | #endif |
| 2091 | |
| 2092 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd); |
| 2093 | int __pte_alloc_kernel(pmd_t *pmd); |
| 2094 | |
| 2095 | #if defined(CONFIG_MMU) |
| 2096 | |
| 2097 | static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd, |
| 2098 | unsigned long address) |
| 2099 | { |
| 2100 | return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ? |
| 2101 | NULL : p4d_offset(pgd, address); |
| 2102 | } |
| 2103 | |
| 2104 | static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d, |
| 2105 | unsigned long address) |
| 2106 | { |
| 2107 | return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ? |
| 2108 | NULL : pud_offset(p4d, address); |
| 2109 | } |
| 2110 | |
| 2111 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
| 2112 | { |
| 2113 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
| 2114 | NULL: pmd_offset(pud, address); |
| 2115 | } |
| 2116 | #endif /* CONFIG_MMU */ |
| 2117 | |
| 2118 | #if USE_SPLIT_PTE_PTLOCKS |
| 2119 | #if ALLOC_SPLIT_PTLOCKS |
| 2120 | void __init ptlock_cache_init(void); |
| 2121 | extern bool ptlock_alloc(struct page *page); |
| 2122 | extern void ptlock_free(struct page *page); |
| 2123 | |
| 2124 | static inline spinlock_t *ptlock_ptr(struct page *page) |
| 2125 | { |
| 2126 | return page->ptl; |
| 2127 | } |
| 2128 | #else /* ALLOC_SPLIT_PTLOCKS */ |
| 2129 | static inline void ptlock_cache_init(void) |
| 2130 | { |
| 2131 | } |
| 2132 | |
| 2133 | static inline bool ptlock_alloc(struct page *page) |
| 2134 | { |
| 2135 | return true; |
| 2136 | } |
| 2137 | |
| 2138 | static inline void ptlock_free(struct page *page) |
| 2139 | { |
| 2140 | } |
| 2141 | |
| 2142 | static inline spinlock_t *ptlock_ptr(struct page *page) |
| 2143 | { |
| 2144 | return &page->ptl; |
| 2145 | } |
| 2146 | #endif /* ALLOC_SPLIT_PTLOCKS */ |
| 2147 | |
| 2148 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 2149 | { |
| 2150 | return ptlock_ptr(pmd_page(*pmd)); |
| 2151 | } |
| 2152 | |
| 2153 | static inline bool ptlock_init(struct page *page) |
| 2154 | { |
| 2155 | /* |
| 2156 | * prep_new_page() initialize page->private (and therefore page->ptl) |
| 2157 | * with 0. Make sure nobody took it in use in between. |
| 2158 | * |
| 2159 | * It can happen if arch try to use slab for page table allocation: |
| 2160 | * slab code uses page->slab_cache, which share storage with page->ptl. |
| 2161 | */ |
| 2162 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); |
| 2163 | if (!ptlock_alloc(page)) |
| 2164 | return false; |
| 2165 | spin_lock_init(ptlock_ptr(page)); |
| 2166 | return true; |
| 2167 | } |
| 2168 | |
| 2169 | #else /* !USE_SPLIT_PTE_PTLOCKS */ |
| 2170 | /* |
| 2171 | * We use mm->page_table_lock to guard all pagetable pages of the mm. |
| 2172 | */ |
| 2173 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 2174 | { |
| 2175 | return &mm->page_table_lock; |
| 2176 | } |
| 2177 | static inline void ptlock_cache_init(void) {} |
| 2178 | static inline bool ptlock_init(struct page *page) { return true; } |
| 2179 | static inline void ptlock_free(struct page *page) {} |
| 2180 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
| 2181 | |
| 2182 | static inline void pgtable_init(void) |
| 2183 | { |
| 2184 | ptlock_cache_init(); |
| 2185 | pgtable_cache_init(); |
| 2186 | } |
| 2187 | |
| 2188 | static inline bool pgtable_pte_page_ctor(struct page *page) |
| 2189 | { |
| 2190 | if (!ptlock_init(page)) |
| 2191 | return false; |
| 2192 | __SetPageTable(page); |
| 2193 | inc_zone_page_state(page, NR_PAGETABLE); |
| 2194 | return true; |
| 2195 | } |
| 2196 | |
| 2197 | static inline void pgtable_pte_page_dtor(struct page *page) |
| 2198 | { |
| 2199 | ptlock_free(page); |
| 2200 | __ClearPageTable(page); |
| 2201 | dec_zone_page_state(page, NR_PAGETABLE); |
| 2202 | } |
| 2203 | |
| 2204 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
| 2205 | ({ \ |
| 2206 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
| 2207 | pte_t *__pte = pte_offset_map(pmd, address); \ |
| 2208 | *(ptlp) = __ptl; \ |
| 2209 | spin_lock(__ptl); \ |
| 2210 | __pte; \ |
| 2211 | }) |
| 2212 | |
| 2213 | #define pte_unmap_unlock(pte, ptl) do { \ |
| 2214 | spin_unlock(ptl); \ |
| 2215 | pte_unmap(pte); \ |
| 2216 | } while (0) |
| 2217 | |
| 2218 | #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd)) |
| 2219 | |
| 2220 | #define pte_alloc_map(mm, pmd, address) \ |
| 2221 | (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address)) |
| 2222 | |
| 2223 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
| 2224 | (pte_alloc(mm, pmd) ? \ |
| 2225 | NULL : pte_offset_map_lock(mm, pmd, address, ptlp)) |
| 2226 | |
| 2227 | #define pte_alloc_kernel(pmd, address) \ |
| 2228 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \ |
| 2229 | NULL: pte_offset_kernel(pmd, address)) |
| 2230 | |
| 2231 | #if USE_SPLIT_PMD_PTLOCKS |
| 2232 | |
| 2233 | static struct page *pmd_to_page(pmd_t *pmd) |
| 2234 | { |
| 2235 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); |
| 2236 | return virt_to_page((void *)((unsigned long) pmd & mask)); |
| 2237 | } |
| 2238 | |
| 2239 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 2240 | { |
| 2241 | return ptlock_ptr(pmd_to_page(pmd)); |
| 2242 | } |
| 2243 | |
| 2244 | static inline bool pgtable_pmd_page_ctor(struct page *page) |
| 2245 | { |
| 2246 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 2247 | page->pmd_huge_pte = NULL; |
| 2248 | #endif |
| 2249 | return ptlock_init(page); |
| 2250 | } |
| 2251 | |
| 2252 | static inline void pgtable_pmd_page_dtor(struct page *page) |
| 2253 | { |
| 2254 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 2255 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); |
| 2256 | #endif |
| 2257 | ptlock_free(page); |
| 2258 | } |
| 2259 | |
| 2260 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) |
| 2261 | |
| 2262 | #else |
| 2263 | |
| 2264 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) |
| 2265 | { |
| 2266 | return &mm->page_table_lock; |
| 2267 | } |
| 2268 | |
| 2269 | static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } |
| 2270 | static inline void pgtable_pmd_page_dtor(struct page *page) {} |
| 2271 | |
| 2272 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) |
| 2273 | |
| 2274 | #endif |
| 2275 | |
| 2276 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) |
| 2277 | { |
| 2278 | spinlock_t *ptl = pmd_lockptr(mm, pmd); |
| 2279 | spin_lock(ptl); |
| 2280 | return ptl; |
| 2281 | } |
| 2282 | |
| 2283 | /* |
| 2284 | * No scalability reason to split PUD locks yet, but follow the same pattern |
| 2285 | * as the PMD locks to make it easier if we decide to. The VM should not be |
| 2286 | * considered ready to switch to split PUD locks yet; there may be places |
| 2287 | * which need to be converted from page_table_lock. |
| 2288 | */ |
| 2289 | static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud) |
| 2290 | { |
| 2291 | return &mm->page_table_lock; |
| 2292 | } |
| 2293 | |
| 2294 | static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud) |
| 2295 | { |
| 2296 | spinlock_t *ptl = pud_lockptr(mm, pud); |
| 2297 | |
| 2298 | spin_lock(ptl); |
| 2299 | return ptl; |
| 2300 | } |
| 2301 | |
| 2302 | extern void __init pagecache_init(void); |
| 2303 | extern void __init free_area_init_memoryless_node(int nid); |
| 2304 | extern void free_initmem(void); |
| 2305 | |
| 2306 | /* |
| 2307 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) |
| 2308 | * into the buddy system. The freed pages will be poisoned with pattern |
| 2309 | * "poison" if it's within range [0, UCHAR_MAX]. |
| 2310 | * Return pages freed into the buddy system. |
| 2311 | */ |
| 2312 | extern unsigned long free_reserved_area(void *start, void *end, |
| 2313 | int poison, const char *s); |
| 2314 | |
| 2315 | #ifdef CONFIG_HIGHMEM |
| 2316 | /* |
| 2317 | * Free a highmem page into the buddy system, adjusting totalhigh_pages |
| 2318 | * and totalram_pages. |
| 2319 | */ |
| 2320 | extern void free_highmem_page(struct page *page); |
| 2321 | #endif |
| 2322 | |
| 2323 | extern void adjust_managed_page_count(struct page *page, long count); |
| 2324 | extern void mem_init_print_info(const char *str); |
| 2325 | |
| 2326 | extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end); |
| 2327 | |
| 2328 | /* Free the reserved page into the buddy system, so it gets managed. */ |
| 2329 | static inline void __free_reserved_page(struct page *page) |
| 2330 | { |
| 2331 | ClearPageReserved(page); |
| 2332 | init_page_count(page); |
| 2333 | __free_page(page); |
| 2334 | } |
| 2335 | |
| 2336 | static inline void free_reserved_page(struct page *page) |
| 2337 | { |
| 2338 | __free_reserved_page(page); |
| 2339 | adjust_managed_page_count(page, 1); |
| 2340 | } |
| 2341 | |
| 2342 | static inline void mark_page_reserved(struct page *page) |
| 2343 | { |
| 2344 | SetPageReserved(page); |
| 2345 | adjust_managed_page_count(page, -1); |
| 2346 | } |
| 2347 | |
| 2348 | /* |
| 2349 | * Default method to free all the __init memory into the buddy system. |
| 2350 | * The freed pages will be poisoned with pattern "poison" if it's within |
| 2351 | * range [0, UCHAR_MAX]. |
| 2352 | * Return pages freed into the buddy system. |
| 2353 | */ |
| 2354 | static inline unsigned long free_initmem_default(int poison) |
| 2355 | { |
| 2356 | extern char __init_begin[], __init_end[]; |
| 2357 | |
| 2358 | return free_reserved_area(&__init_begin, &__init_end, |
| 2359 | poison, "unused kernel"); |
| 2360 | } |
| 2361 | |
| 2362 | static inline unsigned long get_num_physpages(void) |
| 2363 | { |
| 2364 | int nid; |
| 2365 | unsigned long phys_pages = 0; |
| 2366 | |
| 2367 | for_each_online_node(nid) |
| 2368 | phys_pages += node_present_pages(nid); |
| 2369 | |
| 2370 | return phys_pages; |
| 2371 | } |
| 2372 | |
| 2373 | /* |
| 2374 | * Using memblock node mappings, an architecture may initialise its |
| 2375 | * zones, allocate the backing mem_map and account for memory holes in an |
| 2376 | * architecture independent manner. |
| 2377 | * |
| 2378 | * An architecture is expected to register range of page frames backed by |
| 2379 | * physical memory with memblock_add[_node]() before calling |
| 2380 | * free_area_init() passing in the PFN each zone ends at. At a basic |
| 2381 | * usage, an architecture is expected to do something like |
| 2382 | * |
| 2383 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, |
| 2384 | * max_highmem_pfn}; |
| 2385 | * for_each_valid_physical_page_range() |
| 2386 | * memblock_add_node(base, size, nid) |
| 2387 | * free_area_init(max_zone_pfns); |
| 2388 | */ |
| 2389 | void free_area_init(unsigned long *max_zone_pfn); |
| 2390 | unsigned long node_map_pfn_alignment(void); |
| 2391 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, |
| 2392 | unsigned long end_pfn); |
| 2393 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
| 2394 | unsigned long end_pfn); |
| 2395 | extern void get_pfn_range_for_nid(unsigned int nid, |
| 2396 | unsigned long *start_pfn, unsigned long *end_pfn); |
| 2397 | extern unsigned long find_min_pfn_with_active_regions(void); |
| 2398 | |
| 2399 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
| 2400 | static inline int early_pfn_to_nid(unsigned long pfn) |
| 2401 | { |
| 2402 | return 0; |
| 2403 | } |
| 2404 | #else |
| 2405 | /* please see mm/page_alloc.c */ |
| 2406 | extern int __meminit early_pfn_to_nid(unsigned long pfn); |
| 2407 | /* there is a per-arch backend function. */ |
| 2408 | extern int __meminit __early_pfn_to_nid(unsigned long pfn, |
| 2409 | struct mminit_pfnnid_cache *state); |
| 2410 | #endif |
| 2411 | |
| 2412 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
| 2413 | extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long, |
| 2414 | enum memmap_context, struct vmem_altmap *); |
| 2415 | extern void setup_per_zone_wmarks(void); |
| 2416 | extern int __meminit init_per_zone_wmark_min(void); |
| 2417 | extern void mem_init(void); |
| 2418 | extern void __init mmap_init(void); |
| 2419 | extern void show_mem(unsigned int flags, nodemask_t *nodemask); |
| 2420 | extern long si_mem_available(void); |
| 2421 | extern void si_meminfo(struct sysinfo * val); |
| 2422 | extern void si_meminfo_node(struct sysinfo *val, int nid); |
| 2423 | #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES |
| 2424 | extern unsigned long arch_reserved_kernel_pages(void); |
| 2425 | #endif |
| 2426 | |
| 2427 | extern __printf(3, 4) |
| 2428 | void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...); |
| 2429 | |
| 2430 | extern void setup_per_cpu_pageset(void); |
| 2431 | |
| 2432 | /* page_alloc.c */ |
| 2433 | extern int min_free_kbytes; |
| 2434 | extern int watermark_boost_factor; |
| 2435 | extern int watermark_scale_factor; |
| 2436 | extern bool arch_has_descending_max_zone_pfns(void); |
| 2437 | |
| 2438 | /* nommu.c */ |
| 2439 | extern atomic_long_t mmap_pages_allocated; |
| 2440 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
| 2441 | |
| 2442 | /* interval_tree.c */ |
| 2443 | void vma_interval_tree_insert(struct vm_area_struct *node, |
| 2444 | struct rb_root_cached *root); |
| 2445 | void vma_interval_tree_insert_after(struct vm_area_struct *node, |
| 2446 | struct vm_area_struct *prev, |
| 2447 | struct rb_root_cached *root); |
| 2448 | void vma_interval_tree_remove(struct vm_area_struct *node, |
| 2449 | struct rb_root_cached *root); |
| 2450 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root, |
| 2451 | unsigned long start, unsigned long last); |
| 2452 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, |
| 2453 | unsigned long start, unsigned long last); |
| 2454 | |
| 2455 | #define vma_interval_tree_foreach(vma, root, start, last) \ |
| 2456 | for (vma = vma_interval_tree_iter_first(root, start, last); \ |
| 2457 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) |
| 2458 | |
| 2459 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, |
| 2460 | struct rb_root_cached *root); |
| 2461 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, |
| 2462 | struct rb_root_cached *root); |
| 2463 | struct anon_vma_chain * |
| 2464 | anon_vma_interval_tree_iter_first(struct rb_root_cached *root, |
| 2465 | unsigned long start, unsigned long last); |
| 2466 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( |
| 2467 | struct anon_vma_chain *node, unsigned long start, unsigned long last); |
| 2468 | #ifdef CONFIG_DEBUG_VM_RB |
| 2469 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); |
| 2470 | #endif |
| 2471 | |
| 2472 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ |
| 2473 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ |
| 2474 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) |
| 2475 | |
| 2476 | /* mmap.c */ |
| 2477 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
| 2478 | extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start, |
| 2479 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert, |
| 2480 | struct vm_area_struct *expand); |
| 2481 | static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start, |
| 2482 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert) |
| 2483 | { |
| 2484 | return __vma_adjust(vma, start, end, pgoff, insert, NULL); |
| 2485 | } |
| 2486 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
| 2487 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, |
| 2488 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, |
| 2489 | struct mempolicy *, struct vm_userfaultfd_ctx); |
| 2490 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
| 2491 | extern int __split_vma(struct mm_struct *, struct vm_area_struct *, |
| 2492 | unsigned long addr, int new_below); |
| 2493 | extern int split_vma(struct mm_struct *, struct vm_area_struct *, |
| 2494 | unsigned long addr, int new_below); |
| 2495 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
| 2496 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, |
| 2497 | struct rb_node **, struct rb_node *); |
| 2498 | extern void unlink_file_vma(struct vm_area_struct *); |
| 2499 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
| 2500 | unsigned long addr, unsigned long len, pgoff_t pgoff, |
| 2501 | bool *need_rmap_locks); |
| 2502 | extern void exit_mmap(struct mm_struct *); |
| 2503 | |
| 2504 | static inline int check_data_rlimit(unsigned long rlim, |
| 2505 | unsigned long new, |
| 2506 | unsigned long start, |
| 2507 | unsigned long end_data, |
| 2508 | unsigned long start_data) |
| 2509 | { |
| 2510 | if (rlim < RLIM_INFINITY) { |
| 2511 | if (((new - start) + (end_data - start_data)) > rlim) |
| 2512 | return -ENOSPC; |
| 2513 | } |
| 2514 | |
| 2515 | return 0; |
| 2516 | } |
| 2517 | |
| 2518 | extern int mm_take_all_locks(struct mm_struct *mm); |
| 2519 | extern void mm_drop_all_locks(struct mm_struct *mm); |
| 2520 | |
| 2521 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); |
| 2522 | extern struct file *get_mm_exe_file(struct mm_struct *mm); |
| 2523 | extern struct file *get_task_exe_file(struct task_struct *task); |
| 2524 | |
| 2525 | extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages); |
| 2526 | extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages); |
| 2527 | |
| 2528 | extern bool vma_is_special_mapping(const struct vm_area_struct *vma, |
| 2529 | const struct vm_special_mapping *sm); |
| 2530 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, |
| 2531 | unsigned long addr, unsigned long len, |
| 2532 | unsigned long flags, |
| 2533 | const struct vm_special_mapping *spec); |
| 2534 | /* This is an obsolete alternative to _install_special_mapping. */ |
| 2535 | extern int install_special_mapping(struct mm_struct *mm, |
| 2536 | unsigned long addr, unsigned long len, |
| 2537 | unsigned long flags, struct page **pages); |
| 2538 | |
| 2539 | unsigned long randomize_stack_top(unsigned long stack_top); |
| 2540 | |
| 2541 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
| 2542 | |
| 2543 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
| 2544 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff, |
| 2545 | struct list_head *uf); |
| 2546 | extern unsigned long do_mmap(struct file *file, unsigned long addr, |
| 2547 | unsigned long len, unsigned long prot, unsigned long flags, |
| 2548 | unsigned long pgoff, unsigned long *populate, struct list_head *uf); |
| 2549 | extern int __do_munmap(struct mm_struct *, unsigned long, size_t, |
| 2550 | struct list_head *uf, bool downgrade); |
| 2551 | extern int do_munmap(struct mm_struct *, unsigned long, size_t, |
| 2552 | struct list_head *uf); |
| 2553 | extern int do_madvise(unsigned long start, size_t len_in, int behavior); |
| 2554 | |
| 2555 | #ifdef CONFIG_MMU |
| 2556 | extern int __mm_populate(unsigned long addr, unsigned long len, |
| 2557 | int ignore_errors); |
| 2558 | static inline void mm_populate(unsigned long addr, unsigned long len) |
| 2559 | { |
| 2560 | /* Ignore errors */ |
| 2561 | (void) __mm_populate(addr, len, 1); |
| 2562 | } |
| 2563 | #else |
| 2564 | static inline void mm_populate(unsigned long addr, unsigned long len) {} |
| 2565 | #endif |
| 2566 | |
| 2567 | /* These take the mm semaphore themselves */ |
| 2568 | extern int __must_check vm_brk(unsigned long, unsigned long); |
| 2569 | extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long); |
| 2570 | extern int vm_munmap(unsigned long, size_t); |
| 2571 | extern unsigned long __must_check vm_mmap(struct file *, unsigned long, |
| 2572 | unsigned long, unsigned long, |
| 2573 | unsigned long, unsigned long); |
| 2574 | |
| 2575 | struct vm_unmapped_area_info { |
| 2576 | #define VM_UNMAPPED_AREA_TOPDOWN 1 |
| 2577 | unsigned long flags; |
| 2578 | unsigned long length; |
| 2579 | unsigned long low_limit; |
| 2580 | unsigned long high_limit; |
| 2581 | unsigned long align_mask; |
| 2582 | unsigned long align_offset; |
| 2583 | }; |
| 2584 | |
| 2585 | extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info); |
| 2586 | |
| 2587 | /* truncate.c */ |
| 2588 | extern void truncate_inode_pages(struct address_space *, loff_t); |
| 2589 | extern void truncate_inode_pages_range(struct address_space *, |
| 2590 | loff_t lstart, loff_t lend); |
| 2591 | extern void truncate_inode_pages_final(struct address_space *); |
| 2592 | |
| 2593 | /* generic vm_area_ops exported for stackable file systems */ |
| 2594 | extern vm_fault_t filemap_fault(struct vm_fault *vmf); |
| 2595 | extern void filemap_map_pages(struct vm_fault *vmf, |
| 2596 | pgoff_t start_pgoff, pgoff_t end_pgoff); |
| 2597 | extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf); |
| 2598 | |
| 2599 | /* mm/page-writeback.c */ |
| 2600 | int __must_check write_one_page(struct page *page); |
| 2601 | void task_dirty_inc(struct task_struct *tsk); |
| 2602 | |
| 2603 | extern unsigned long stack_guard_gap; |
| 2604 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ |
| 2605 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
| 2606 | |
| 2607 | /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */ |
| 2608 | extern int expand_downwards(struct vm_area_struct *vma, |
| 2609 | unsigned long address); |
| 2610 | #if VM_GROWSUP |
| 2611 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
| 2612 | #else |
| 2613 | #define expand_upwards(vma, address) (0) |
| 2614 | #endif |
| 2615 | |
| 2616 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ |
| 2617 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); |
| 2618 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, |
| 2619 | struct vm_area_struct **pprev); |
| 2620 | |
| 2621 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, |
| 2622 | NULL if none. Assume start_addr < end_addr. */ |
| 2623 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) |
| 2624 | { |
| 2625 | struct vm_area_struct * vma = find_vma(mm,start_addr); |
| 2626 | |
| 2627 | if (vma && end_addr <= vma->vm_start) |
| 2628 | vma = NULL; |
| 2629 | return vma; |
| 2630 | } |
| 2631 | |
| 2632 | static inline unsigned long vm_start_gap(struct vm_area_struct *vma) |
| 2633 | { |
| 2634 | unsigned long vm_start = vma->vm_start; |
| 2635 | |
| 2636 | if (vma->vm_flags & VM_GROWSDOWN) { |
| 2637 | vm_start -= stack_guard_gap; |
| 2638 | if (vm_start > vma->vm_start) |
| 2639 | vm_start = 0; |
| 2640 | } |
| 2641 | return vm_start; |
| 2642 | } |
| 2643 | |
| 2644 | static inline unsigned long vm_end_gap(struct vm_area_struct *vma) |
| 2645 | { |
| 2646 | unsigned long vm_end = vma->vm_end; |
| 2647 | |
| 2648 | if (vma->vm_flags & VM_GROWSUP) { |
| 2649 | vm_end += stack_guard_gap; |
| 2650 | if (vm_end < vma->vm_end) |
| 2651 | vm_end = -PAGE_SIZE; |
| 2652 | } |
| 2653 | return vm_end; |
| 2654 | } |
| 2655 | |
| 2656 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
| 2657 | { |
| 2658 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
| 2659 | } |
| 2660 | |
| 2661 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ |
| 2662 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, |
| 2663 | unsigned long vm_start, unsigned long vm_end) |
| 2664 | { |
| 2665 | struct vm_area_struct *vma = find_vma(mm, vm_start); |
| 2666 | |
| 2667 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) |
| 2668 | vma = NULL; |
| 2669 | |
| 2670 | return vma; |
| 2671 | } |
| 2672 | |
| 2673 | static inline bool range_in_vma(struct vm_area_struct *vma, |
| 2674 | unsigned long start, unsigned long end) |
| 2675 | { |
| 2676 | return (vma && vma->vm_start <= start && end <= vma->vm_end); |
| 2677 | } |
| 2678 | |
| 2679 | #ifdef CONFIG_MMU |
| 2680 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
| 2681 | void vma_set_page_prot(struct vm_area_struct *vma); |
| 2682 | #else |
| 2683 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) |
| 2684 | { |
| 2685 | return __pgprot(0); |
| 2686 | } |
| 2687 | static inline void vma_set_page_prot(struct vm_area_struct *vma) |
| 2688 | { |
| 2689 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
| 2690 | } |
| 2691 | #endif |
| 2692 | |
| 2693 | #ifdef CONFIG_NUMA_BALANCING |
| 2694 | unsigned long change_prot_numa(struct vm_area_struct *vma, |
| 2695 | unsigned long start, unsigned long end); |
| 2696 | #endif |
| 2697 | |
| 2698 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
| 2699 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
| 2700 | unsigned long pfn, unsigned long size, pgprot_t); |
| 2701 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
| 2702 | int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr, |
| 2703 | struct page **pages, unsigned long *num); |
| 2704 | int vm_map_pages(struct vm_area_struct *vma, struct page **pages, |
| 2705 | unsigned long num); |
| 2706 | int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages, |
| 2707 | unsigned long num); |
| 2708 | vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
| 2709 | unsigned long pfn); |
| 2710 | vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr, |
| 2711 | unsigned long pfn, pgprot_t pgprot); |
| 2712 | vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
| 2713 | pfn_t pfn); |
| 2714 | vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr, |
| 2715 | pfn_t pfn, pgprot_t pgprot); |
| 2716 | vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma, |
| 2717 | unsigned long addr, pfn_t pfn); |
| 2718 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); |
| 2719 | |
| 2720 | static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma, |
| 2721 | unsigned long addr, struct page *page) |
| 2722 | { |
| 2723 | int err = vm_insert_page(vma, addr, page); |
| 2724 | |
| 2725 | if (err == -ENOMEM) |
| 2726 | return VM_FAULT_OOM; |
| 2727 | if (err < 0 && err != -EBUSY) |
| 2728 | return VM_FAULT_SIGBUS; |
| 2729 | |
| 2730 | return VM_FAULT_NOPAGE; |
| 2731 | } |
| 2732 | |
| 2733 | static inline vm_fault_t vmf_error(int err) |
| 2734 | { |
| 2735 | if (err == -ENOMEM) |
| 2736 | return VM_FAULT_OOM; |
| 2737 | return VM_FAULT_SIGBUS; |
| 2738 | } |
| 2739 | |
| 2740 | struct page *follow_page(struct vm_area_struct *vma, unsigned long address, |
| 2741 | unsigned int foll_flags); |
| 2742 | |
| 2743 | #define FOLL_WRITE 0x01 /* check pte is writable */ |
| 2744 | #define FOLL_TOUCH 0x02 /* mark page accessed */ |
| 2745 | #define FOLL_GET 0x04 /* do get_page on page */ |
| 2746 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ |
| 2747 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ |
| 2748 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO |
| 2749 | * and return without waiting upon it */ |
| 2750 | #define FOLL_POPULATE 0x40 /* fault in page */ |
| 2751 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ |
| 2752 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ |
| 2753 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ |
| 2754 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ |
| 2755 | #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */ |
| 2756 | #define FOLL_MLOCK 0x1000 /* lock present pages */ |
| 2757 | #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */ |
| 2758 | #define FOLL_COW 0x4000 /* internal GUP flag */ |
| 2759 | #define FOLL_ANON 0x8000 /* don't do file mappings */ |
| 2760 | #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */ |
| 2761 | #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */ |
| 2762 | #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */ |
| 2763 | #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */ |
| 2764 | |
| 2765 | /* |
| 2766 | * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each |
| 2767 | * other. Here is what they mean, and how to use them: |
| 2768 | * |
| 2769 | * FOLL_LONGTERM indicates that the page will be held for an indefinite time |
| 2770 | * period _often_ under userspace control. This is in contrast to |
| 2771 | * iov_iter_get_pages(), whose usages are transient. |
| 2772 | * |
| 2773 | * FIXME: For pages which are part of a filesystem, mappings are subject to the |
| 2774 | * lifetime enforced by the filesystem and we need guarantees that longterm |
| 2775 | * users like RDMA and V4L2 only establish mappings which coordinate usage with |
| 2776 | * the filesystem. Ideas for this coordination include revoking the longterm |
| 2777 | * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was |
| 2778 | * added after the problem with filesystems was found FS DAX VMAs are |
| 2779 | * specifically failed. Filesystem pages are still subject to bugs and use of |
| 2780 | * FOLL_LONGTERM should be avoided on those pages. |
| 2781 | * |
| 2782 | * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call. |
| 2783 | * Currently only get_user_pages() and get_user_pages_fast() support this flag |
| 2784 | * and calls to get_user_pages_[un]locked are specifically not allowed. This |
| 2785 | * is due to an incompatibility with the FS DAX check and |
| 2786 | * FAULT_FLAG_ALLOW_RETRY. |
| 2787 | * |
| 2788 | * In the CMA case: long term pins in a CMA region would unnecessarily fragment |
| 2789 | * that region. And so, CMA attempts to migrate the page before pinning, when |
| 2790 | * FOLL_LONGTERM is specified. |
| 2791 | * |
| 2792 | * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount, |
| 2793 | * but an additional pin counting system) will be invoked. This is intended for |
| 2794 | * anything that gets a page reference and then touches page data (for example, |
| 2795 | * Direct IO). This lets the filesystem know that some non-file-system entity is |
| 2796 | * potentially changing the pages' data. In contrast to FOLL_GET (whose pages |
| 2797 | * are released via put_page()), FOLL_PIN pages must be released, ultimately, by |
| 2798 | * a call to unpin_user_page(). |
| 2799 | * |
| 2800 | * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different |
| 2801 | * and separate refcounting mechanisms, however, and that means that each has |
| 2802 | * its own acquire and release mechanisms: |
| 2803 | * |
| 2804 | * FOLL_GET: get_user_pages*() to acquire, and put_page() to release. |
| 2805 | * |
| 2806 | * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release. |
| 2807 | * |
| 2808 | * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call. |
| 2809 | * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based |
| 2810 | * calls applied to them, and that's perfectly OK. This is a constraint on the |
| 2811 | * callers, not on the pages.) |
| 2812 | * |
| 2813 | * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never |
| 2814 | * directly by the caller. That's in order to help avoid mismatches when |
| 2815 | * releasing pages: get_user_pages*() pages must be released via put_page(), |
| 2816 | * while pin_user_pages*() pages must be released via unpin_user_page(). |
| 2817 | * |
| 2818 | * Please see Documentation/core-api/pin_user_pages.rst for more information. |
| 2819 | */ |
| 2820 | |
| 2821 | static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags) |
| 2822 | { |
| 2823 | if (vm_fault & VM_FAULT_OOM) |
| 2824 | return -ENOMEM; |
| 2825 | if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) |
| 2826 | return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT; |
| 2827 | if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV)) |
| 2828 | return -EFAULT; |
| 2829 | return 0; |
| 2830 | } |
| 2831 | |
| 2832 | typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data); |
| 2833 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, |
| 2834 | unsigned long size, pte_fn_t fn, void *data); |
| 2835 | extern int apply_to_existing_page_range(struct mm_struct *mm, |
| 2836 | unsigned long address, unsigned long size, |
| 2837 | pte_fn_t fn, void *data); |
| 2838 | |
| 2839 | #ifdef CONFIG_PAGE_POISONING |
| 2840 | extern bool page_poisoning_enabled(void); |
| 2841 | extern void kernel_poison_pages(struct page *page, int numpages, int enable); |
| 2842 | #else |
| 2843 | static inline bool page_poisoning_enabled(void) { return false; } |
| 2844 | static inline void kernel_poison_pages(struct page *page, int numpages, |
| 2845 | int enable) { } |
| 2846 | #endif |
| 2847 | |
| 2848 | #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON |
| 2849 | DECLARE_STATIC_KEY_TRUE(init_on_alloc); |
| 2850 | #else |
| 2851 | DECLARE_STATIC_KEY_FALSE(init_on_alloc); |
| 2852 | #endif |
| 2853 | static inline bool want_init_on_alloc(gfp_t flags) |
| 2854 | { |
| 2855 | if (static_branch_unlikely(&init_on_alloc) && |
| 2856 | !page_poisoning_enabled()) |
| 2857 | return true; |
| 2858 | return flags & __GFP_ZERO; |
| 2859 | } |
| 2860 | |
| 2861 | #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON |
| 2862 | DECLARE_STATIC_KEY_TRUE(init_on_free); |
| 2863 | #else |
| 2864 | DECLARE_STATIC_KEY_FALSE(init_on_free); |
| 2865 | #endif |
| 2866 | static inline bool want_init_on_free(void) |
| 2867 | { |
| 2868 | return static_branch_unlikely(&init_on_free) && |
| 2869 | !page_poisoning_enabled(); |
| 2870 | } |
| 2871 | |
| 2872 | #ifdef CONFIG_DEBUG_PAGEALLOC |
| 2873 | extern void init_debug_pagealloc(void); |
| 2874 | #else |
| 2875 | static inline void init_debug_pagealloc(void) {} |
| 2876 | #endif |
| 2877 | extern bool _debug_pagealloc_enabled_early; |
| 2878 | DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled); |
| 2879 | |
| 2880 | static inline bool debug_pagealloc_enabled(void) |
| 2881 | { |
| 2882 | return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) && |
| 2883 | _debug_pagealloc_enabled_early; |
| 2884 | } |
| 2885 | |
| 2886 | /* |
| 2887 | * For use in fast paths after init_debug_pagealloc() has run, or when a |
| 2888 | * false negative result is not harmful when called too early. |
| 2889 | */ |
| 2890 | static inline bool debug_pagealloc_enabled_static(void) |
| 2891 | { |
| 2892 | if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC)) |
| 2893 | return false; |
| 2894 | |
| 2895 | return static_branch_unlikely(&_debug_pagealloc_enabled); |
| 2896 | } |
| 2897 | |
| 2898 | #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP) |
| 2899 | extern void __kernel_map_pages(struct page *page, int numpages, int enable); |
| 2900 | |
| 2901 | /* |
| 2902 | * When called in DEBUG_PAGEALLOC context, the call should most likely be |
| 2903 | * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static() |
| 2904 | */ |
| 2905 | static inline void |
| 2906 | kernel_map_pages(struct page *page, int numpages, int enable) |
| 2907 | { |
| 2908 | __kernel_map_pages(page, numpages, enable); |
| 2909 | } |
| 2910 | #ifdef CONFIG_HIBERNATION |
| 2911 | extern bool kernel_page_present(struct page *page); |
| 2912 | #endif /* CONFIG_HIBERNATION */ |
| 2913 | #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */ |
| 2914 | static inline void |
| 2915 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
| 2916 | #ifdef CONFIG_HIBERNATION |
| 2917 | static inline bool kernel_page_present(struct page *page) { return true; } |
| 2918 | #endif /* CONFIG_HIBERNATION */ |
| 2919 | #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */ |
| 2920 | |
| 2921 | #ifdef __HAVE_ARCH_GATE_AREA |
| 2922 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
| 2923 | extern int in_gate_area_no_mm(unsigned long addr); |
| 2924 | extern int in_gate_area(struct mm_struct *mm, unsigned long addr); |
| 2925 | #else |
| 2926 | static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) |
| 2927 | { |
| 2928 | return NULL; |
| 2929 | } |
| 2930 | static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } |
| 2931 | static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) |
| 2932 | { |
| 2933 | return 0; |
| 2934 | } |
| 2935 | #endif /* __HAVE_ARCH_GATE_AREA */ |
| 2936 | |
| 2937 | extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm); |
| 2938 | |
| 2939 | #ifdef CONFIG_SYSCTL |
| 2940 | extern int sysctl_drop_caches; |
| 2941 | int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *, |
| 2942 | loff_t *); |
| 2943 | #endif |
| 2944 | |
| 2945 | void drop_slab(void); |
| 2946 | void drop_slab_node(int nid); |
| 2947 | |
| 2948 | #ifndef CONFIG_MMU |
| 2949 | #define randomize_va_space 0 |
| 2950 | #else |
| 2951 | extern int randomize_va_space; |
| 2952 | #endif |
| 2953 | |
| 2954 | const char * arch_vma_name(struct vm_area_struct *vma); |
| 2955 | #ifdef CONFIG_MMU |
| 2956 | void print_vma_addr(char *prefix, unsigned long rip); |
| 2957 | #else |
| 2958 | static inline void print_vma_addr(char *prefix, unsigned long rip) |
| 2959 | { |
| 2960 | } |
| 2961 | #endif |
| 2962 | |
| 2963 | void *sparse_buffer_alloc(unsigned long size); |
| 2964 | struct page * __populate_section_memmap(unsigned long pfn, |
| 2965 | unsigned long nr_pages, int nid, struct vmem_altmap *altmap); |
| 2966 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
| 2967 | p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node); |
| 2968 | pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node); |
| 2969 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); |
| 2970 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node, |
| 2971 | struct vmem_altmap *altmap); |
| 2972 | void *vmemmap_alloc_block(unsigned long size, int node); |
| 2973 | struct vmem_altmap; |
| 2974 | void *vmemmap_alloc_block_buf(unsigned long size, int node, |
| 2975 | struct vmem_altmap *altmap); |
| 2976 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
| 2977 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, |
| 2978 | int node, struct vmem_altmap *altmap); |
| 2979 | int vmemmap_populate(unsigned long start, unsigned long end, int node, |
| 2980 | struct vmem_altmap *altmap); |
| 2981 | void vmemmap_populate_print_last(void); |
| 2982 | #ifdef CONFIG_MEMORY_HOTPLUG |
| 2983 | void vmemmap_free(unsigned long start, unsigned long end, |
| 2984 | struct vmem_altmap *altmap); |
| 2985 | #endif |
| 2986 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, |
| 2987 | unsigned long nr_pages); |
| 2988 | |
| 2989 | enum mf_flags { |
| 2990 | MF_COUNT_INCREASED = 1 << 0, |
| 2991 | MF_ACTION_REQUIRED = 1 << 1, |
| 2992 | MF_MUST_KILL = 1 << 2, |
| 2993 | MF_SOFT_OFFLINE = 1 << 3, |
| 2994 | }; |
| 2995 | extern int memory_failure(unsigned long pfn, int flags); |
| 2996 | extern void memory_failure_queue(unsigned long pfn, int flags); |
| 2997 | extern void memory_failure_queue_kick(int cpu); |
| 2998 | extern int unpoison_memory(unsigned long pfn); |
| 2999 | extern int get_hwpoison_page(struct page *page); |
| 3000 | #define put_hwpoison_page(page) put_page(page) |
| 3001 | extern int sysctl_memory_failure_early_kill; |
| 3002 | extern int sysctl_memory_failure_recovery; |
| 3003 | extern void shake_page(struct page *p, int access); |
| 3004 | extern atomic_long_t num_poisoned_pages __read_mostly; |
| 3005 | extern int soft_offline_page(unsigned long pfn, int flags); |
| 3006 | |
| 3007 | |
| 3008 | /* |
| 3009 | * Error handlers for various types of pages. |
| 3010 | */ |
| 3011 | enum mf_result { |
| 3012 | MF_IGNORED, /* Error: cannot be handled */ |
| 3013 | MF_FAILED, /* Error: handling failed */ |
| 3014 | MF_DELAYED, /* Will be handled later */ |
| 3015 | MF_RECOVERED, /* Successfully recovered */ |
| 3016 | }; |
| 3017 | |
| 3018 | enum mf_action_page_type { |
| 3019 | MF_MSG_KERNEL, |
| 3020 | MF_MSG_KERNEL_HIGH_ORDER, |
| 3021 | MF_MSG_SLAB, |
| 3022 | MF_MSG_DIFFERENT_COMPOUND, |
| 3023 | MF_MSG_POISONED_HUGE, |
| 3024 | MF_MSG_HUGE, |
| 3025 | MF_MSG_FREE_HUGE, |
| 3026 | MF_MSG_NON_PMD_HUGE, |
| 3027 | MF_MSG_UNMAP_FAILED, |
| 3028 | MF_MSG_DIRTY_SWAPCACHE, |
| 3029 | MF_MSG_CLEAN_SWAPCACHE, |
| 3030 | MF_MSG_DIRTY_MLOCKED_LRU, |
| 3031 | MF_MSG_CLEAN_MLOCKED_LRU, |
| 3032 | MF_MSG_DIRTY_UNEVICTABLE_LRU, |
| 3033 | MF_MSG_CLEAN_UNEVICTABLE_LRU, |
| 3034 | MF_MSG_DIRTY_LRU, |
| 3035 | MF_MSG_CLEAN_LRU, |
| 3036 | MF_MSG_TRUNCATED_LRU, |
| 3037 | MF_MSG_BUDDY, |
| 3038 | MF_MSG_BUDDY_2ND, |
| 3039 | MF_MSG_DAX, |
| 3040 | MF_MSG_UNKNOWN, |
| 3041 | }; |
| 3042 | |
| 3043 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
| 3044 | extern void clear_huge_page(struct page *page, |
| 3045 | unsigned long addr_hint, |
| 3046 | unsigned int pages_per_huge_page); |
| 3047 | extern void copy_user_huge_page(struct page *dst, struct page *src, |
| 3048 | unsigned long addr_hint, |
| 3049 | struct vm_area_struct *vma, |
| 3050 | unsigned int pages_per_huge_page); |
| 3051 | extern long copy_huge_page_from_user(struct page *dst_page, |
| 3052 | const void __user *usr_src, |
| 3053 | unsigned int pages_per_huge_page, |
| 3054 | bool allow_pagefault); |
| 3055 | |
| 3056 | /** |
| 3057 | * vma_is_special_huge - Are transhuge page-table entries considered special? |
| 3058 | * @vma: Pointer to the struct vm_area_struct to consider |
| 3059 | * |
| 3060 | * Whether transhuge page-table entries are considered "special" following |
| 3061 | * the definition in vm_normal_page(). |
| 3062 | * |
| 3063 | * Return: true if transhuge page-table entries should be considered special, |
| 3064 | * false otherwise. |
| 3065 | */ |
| 3066 | static inline bool vma_is_special_huge(const struct vm_area_struct *vma) |
| 3067 | { |
| 3068 | return vma_is_dax(vma) || (vma->vm_file && |
| 3069 | (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))); |
| 3070 | } |
| 3071 | |
| 3072 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ |
| 3073 | |
| 3074 | #ifdef CONFIG_DEBUG_PAGEALLOC |
| 3075 | extern unsigned int _debug_guardpage_minorder; |
| 3076 | DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled); |
| 3077 | |
| 3078 | static inline unsigned int debug_guardpage_minorder(void) |
| 3079 | { |
| 3080 | return _debug_guardpage_minorder; |
| 3081 | } |
| 3082 | |
| 3083 | static inline bool debug_guardpage_enabled(void) |
| 3084 | { |
| 3085 | return static_branch_unlikely(&_debug_guardpage_enabled); |
| 3086 | } |
| 3087 | |
| 3088 | static inline bool page_is_guard(struct page *page) |
| 3089 | { |
| 3090 | if (!debug_guardpage_enabled()) |
| 3091 | return false; |
| 3092 | |
| 3093 | return PageGuard(page); |
| 3094 | } |
| 3095 | #else |
| 3096 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } |
| 3097 | static inline bool debug_guardpage_enabled(void) { return false; } |
| 3098 | static inline bool page_is_guard(struct page *page) { return false; } |
| 3099 | #endif /* CONFIG_DEBUG_PAGEALLOC */ |
| 3100 | |
| 3101 | #if MAX_NUMNODES > 1 |
| 3102 | void __init setup_nr_node_ids(void); |
| 3103 | #else |
| 3104 | static inline void setup_nr_node_ids(void) {} |
| 3105 | #endif |
| 3106 | |
| 3107 | extern int memcmp_pages(struct page *page1, struct page *page2); |
| 3108 | |
| 3109 | static inline int pages_identical(struct page *page1, struct page *page2) |
| 3110 | { |
| 3111 | return !memcmp_pages(page1, page2); |
| 3112 | } |
| 3113 | |
| 3114 | #ifdef CONFIG_MAPPING_DIRTY_HELPERS |
| 3115 | unsigned long clean_record_shared_mapping_range(struct address_space *mapping, |
| 3116 | pgoff_t first_index, pgoff_t nr, |
| 3117 | pgoff_t bitmap_pgoff, |
| 3118 | unsigned long *bitmap, |
| 3119 | pgoff_t *start, |
| 3120 | pgoff_t *end); |
| 3121 | |
| 3122 | unsigned long wp_shared_mapping_range(struct address_space *mapping, |
| 3123 | pgoff_t first_index, pgoff_t nr); |
| 3124 | #endif |
| 3125 | |
| 3126 | extern int sysctl_nr_trim_pages; |
| 3127 | |
| 3128 | #endif /* __KERNEL__ */ |
| 3129 | #endif /* _LINUX_MM_H */ |