Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
at nocache-cleanup 1863 lines 51 kB view raw
1/* SPDX-License-Identifier: GPL-2.0+ */ 2/* 3 * vma_internal.h 4 * 5 * Header providing userland wrappers and shims for the functionality provided 6 * by mm/vma_internal.h. 7 * 8 * We make the header guard the same as mm/vma_internal.h, so if this shim 9 * header is included, it precludes the inclusion of the kernel one. 10 */ 11 12#ifndef __MM_VMA_INTERNAL_H 13#define __MM_VMA_INTERNAL_H 14 15#define __private 16#define __bitwise 17#define __randomize_layout 18 19#define CONFIG_MMU 20#define CONFIG_PER_VMA_LOCK 21 22#include <stdlib.h> 23 24#include <linux/atomic.h> 25#include <linux/list.h> 26#include <linux/maple_tree.h> 27#include <linux/mm.h> 28#include <linux/rbtree.h> 29#include <linux/refcount.h> 30#include <linux/slab.h> 31 32extern unsigned long stack_guard_gap; 33#ifdef CONFIG_MMU 34extern unsigned long mmap_min_addr; 35extern unsigned long dac_mmap_min_addr; 36#else 37#define mmap_min_addr 0UL 38#define dac_mmap_min_addr 0UL 39#endif 40 41#define VM_WARN_ON(_expr) (WARN_ON(_expr)) 42#define VM_WARN_ON_ONCE(_expr) (WARN_ON_ONCE(_expr)) 43#define VM_WARN_ON_VMG(_expr, _vmg) (WARN_ON(_expr)) 44#define VM_BUG_ON(_expr) (BUG_ON(_expr)) 45#define VM_BUG_ON_VMA(_expr, _vma) (BUG_ON(_expr)) 46 47#define MMF_HAS_MDWE 28 48 49/* 50 * vm_flags in vm_area_struct, see mm_types.h. 51 * When changing, update also include/trace/events/mmflags.h 52 */ 53 54#define VM_NONE 0x00000000 55 56/** 57 * typedef vma_flag_t - specifies an individual VMA flag by bit number. 58 * 59 * This value is made type safe by sparse to avoid passing invalid flag values 60 * around. 61 */ 62typedef int __bitwise vma_flag_t; 63 64#define DECLARE_VMA_BIT(name, bitnum) \ 65 VMA_ ## name ## _BIT = ((__force vma_flag_t)bitnum) 66#define DECLARE_VMA_BIT_ALIAS(name, aliased) \ 67 VMA_ ## name ## _BIT = VMA_ ## aliased ## _BIT 68enum { 69 DECLARE_VMA_BIT(READ, 0), 70 DECLARE_VMA_BIT(WRITE, 1), 71 DECLARE_VMA_BIT(EXEC, 2), 72 DECLARE_VMA_BIT(SHARED, 3), 73 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ 74 DECLARE_VMA_BIT(MAYREAD, 4), /* limits for mprotect() etc. */ 75 DECLARE_VMA_BIT(MAYWRITE, 5), 76 DECLARE_VMA_BIT(MAYEXEC, 6), 77 DECLARE_VMA_BIT(MAYSHARE, 7), 78 DECLARE_VMA_BIT(GROWSDOWN, 8), /* general info on the segment */ 79#ifdef CONFIG_MMU 80 DECLARE_VMA_BIT(UFFD_MISSING, 9),/* missing pages tracking */ 81#else 82 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */ 83 DECLARE_VMA_BIT(MAYOVERLAY, 9), 84#endif /* CONFIG_MMU */ 85 /* Page-ranges managed without "struct page", just pure PFN */ 86 DECLARE_VMA_BIT(PFNMAP, 10), 87 DECLARE_VMA_BIT(MAYBE_GUARD, 11), 88 DECLARE_VMA_BIT(UFFD_WP, 12), /* wrprotect pages tracking */ 89 DECLARE_VMA_BIT(LOCKED, 13), 90 DECLARE_VMA_BIT(IO, 14), /* Memory mapped I/O or similar */ 91 DECLARE_VMA_BIT(SEQ_READ, 15), /* App will access data sequentially */ 92 DECLARE_VMA_BIT(RAND_READ, 16), /* App will not benefit from clustered reads */ 93 DECLARE_VMA_BIT(DONTCOPY, 17), /* Do not copy this vma on fork */ 94 DECLARE_VMA_BIT(DONTEXPAND, 18),/* Cannot expand with mremap() */ 95 DECLARE_VMA_BIT(LOCKONFAULT, 19),/* Lock pages covered when faulted in */ 96 DECLARE_VMA_BIT(ACCOUNT, 20), /* Is a VM accounted object */ 97 DECLARE_VMA_BIT(NORESERVE, 21), /* should the VM suppress accounting */ 98 DECLARE_VMA_BIT(HUGETLB, 22), /* Huge TLB Page VM */ 99 DECLARE_VMA_BIT(SYNC, 23), /* Synchronous page faults */ 100 DECLARE_VMA_BIT(ARCH_1, 24), /* Architecture-specific flag */ 101 DECLARE_VMA_BIT(WIPEONFORK, 25),/* Wipe VMA contents in child. */ 102 DECLARE_VMA_BIT(DONTDUMP, 26), /* Do not include in the core dump */ 103 DECLARE_VMA_BIT(SOFTDIRTY, 27), /* NOT soft dirty clean area */ 104 DECLARE_VMA_BIT(MIXEDMAP, 28), /* Can contain struct page and pure PFN pages */ 105 DECLARE_VMA_BIT(HUGEPAGE, 29), /* MADV_HUGEPAGE marked this vma */ 106 DECLARE_VMA_BIT(NOHUGEPAGE, 30),/* MADV_NOHUGEPAGE marked this vma */ 107 DECLARE_VMA_BIT(MERGEABLE, 31), /* KSM may merge identical pages */ 108 /* These bits are reused, we define specific uses below. */ 109 DECLARE_VMA_BIT(HIGH_ARCH_0, 32), 110 DECLARE_VMA_BIT(HIGH_ARCH_1, 33), 111 DECLARE_VMA_BIT(HIGH_ARCH_2, 34), 112 DECLARE_VMA_BIT(HIGH_ARCH_3, 35), 113 DECLARE_VMA_BIT(HIGH_ARCH_4, 36), 114 DECLARE_VMA_BIT(HIGH_ARCH_5, 37), 115 DECLARE_VMA_BIT(HIGH_ARCH_6, 38), 116 /* 117 * This flag is used to connect VFIO to arch specific KVM code. It 118 * indicates that the memory under this VMA is safe for use with any 119 * non-cachable memory type inside KVM. Some VFIO devices, on some 120 * platforms, are thought to be unsafe and can cause machine crashes 121 * if KVM does not lock down the memory type. 122 */ 123 DECLARE_VMA_BIT(ALLOW_ANY_UNCACHED, 39), 124#ifdef CONFIG_PPC32 125 DECLARE_VMA_BIT_ALIAS(DROPPABLE, ARCH_1), 126#else 127 DECLARE_VMA_BIT(DROPPABLE, 40), 128#endif 129 DECLARE_VMA_BIT(UFFD_MINOR, 41), 130 DECLARE_VMA_BIT(SEALED, 42), 131 /* Flags that reuse flags above. */ 132 DECLARE_VMA_BIT_ALIAS(PKEY_BIT0, HIGH_ARCH_0), 133 DECLARE_VMA_BIT_ALIAS(PKEY_BIT1, HIGH_ARCH_1), 134 DECLARE_VMA_BIT_ALIAS(PKEY_BIT2, HIGH_ARCH_2), 135 DECLARE_VMA_BIT_ALIAS(PKEY_BIT3, HIGH_ARCH_3), 136 DECLARE_VMA_BIT_ALIAS(PKEY_BIT4, HIGH_ARCH_4), 137#if defined(CONFIG_X86_USER_SHADOW_STACK) 138 /* 139 * VM_SHADOW_STACK should not be set with VM_SHARED because of lack of 140 * support core mm. 141 * 142 * These VMAs will get a single end guard page. This helps userspace 143 * protect itself from attacks. A single page is enough for current 144 * shadow stack archs (x86). See the comments near alloc_shstk() in 145 * arch/x86/kernel/shstk.c for more details on the guard size. 146 */ 147 DECLARE_VMA_BIT_ALIAS(SHADOW_STACK, HIGH_ARCH_5), 148#elif defined(CONFIG_ARM64_GCS) 149 /* 150 * arm64's Guarded Control Stack implements similar functionality and 151 * has similar constraints to shadow stacks. 152 */ 153 DECLARE_VMA_BIT_ALIAS(SHADOW_STACK, HIGH_ARCH_6), 154#endif 155 DECLARE_VMA_BIT_ALIAS(SAO, ARCH_1), /* Strong Access Ordering (powerpc) */ 156 DECLARE_VMA_BIT_ALIAS(GROWSUP, ARCH_1), /* parisc */ 157 DECLARE_VMA_BIT_ALIAS(SPARC_ADI, ARCH_1), /* sparc64 */ 158 DECLARE_VMA_BIT_ALIAS(ARM64_BTI, ARCH_1), /* arm64 */ 159 DECLARE_VMA_BIT_ALIAS(ARCH_CLEAR, ARCH_1), /* sparc64, arm64 */ 160 DECLARE_VMA_BIT_ALIAS(MAPPED_COPY, ARCH_1), /* !CONFIG_MMU */ 161 DECLARE_VMA_BIT_ALIAS(MTE, HIGH_ARCH_4), /* arm64 */ 162 DECLARE_VMA_BIT_ALIAS(MTE_ALLOWED, HIGH_ARCH_5),/* arm64 */ 163#ifdef CONFIG_STACK_GROWSUP 164 DECLARE_VMA_BIT_ALIAS(STACK, GROWSUP), 165 DECLARE_VMA_BIT_ALIAS(STACK_EARLY, GROWSDOWN), 166#else 167 DECLARE_VMA_BIT_ALIAS(STACK, GROWSDOWN), 168#endif 169}; 170 171#define INIT_VM_FLAG(name) BIT((__force int) VMA_ ## name ## _BIT) 172#define VM_READ INIT_VM_FLAG(READ) 173#define VM_WRITE INIT_VM_FLAG(WRITE) 174#define VM_EXEC INIT_VM_FLAG(EXEC) 175#define VM_SHARED INIT_VM_FLAG(SHARED) 176#define VM_MAYREAD INIT_VM_FLAG(MAYREAD) 177#define VM_MAYWRITE INIT_VM_FLAG(MAYWRITE) 178#define VM_MAYEXEC INIT_VM_FLAG(MAYEXEC) 179#define VM_MAYSHARE INIT_VM_FLAG(MAYSHARE) 180#define VM_GROWSDOWN INIT_VM_FLAG(GROWSDOWN) 181#ifdef CONFIG_MMU 182#define VM_UFFD_MISSING INIT_VM_FLAG(UFFD_MISSING) 183#else 184#define VM_UFFD_MISSING VM_NONE 185#define VM_MAYOVERLAY INIT_VM_FLAG(MAYOVERLAY) 186#endif 187#define VM_PFNMAP INIT_VM_FLAG(PFNMAP) 188#define VM_MAYBE_GUARD INIT_VM_FLAG(MAYBE_GUARD) 189#define VM_UFFD_WP INIT_VM_FLAG(UFFD_WP) 190#define VM_LOCKED INIT_VM_FLAG(LOCKED) 191#define VM_IO INIT_VM_FLAG(IO) 192#define VM_SEQ_READ INIT_VM_FLAG(SEQ_READ) 193#define VM_RAND_READ INIT_VM_FLAG(RAND_READ) 194#define VM_DONTCOPY INIT_VM_FLAG(DONTCOPY) 195#define VM_DONTEXPAND INIT_VM_FLAG(DONTEXPAND) 196#define VM_LOCKONFAULT INIT_VM_FLAG(LOCKONFAULT) 197#define VM_ACCOUNT INIT_VM_FLAG(ACCOUNT) 198#define VM_NORESERVE INIT_VM_FLAG(NORESERVE) 199#define VM_HUGETLB INIT_VM_FLAG(HUGETLB) 200#define VM_SYNC INIT_VM_FLAG(SYNC) 201#define VM_ARCH_1 INIT_VM_FLAG(ARCH_1) 202#define VM_WIPEONFORK INIT_VM_FLAG(WIPEONFORK) 203#define VM_DONTDUMP INIT_VM_FLAG(DONTDUMP) 204#ifdef CONFIG_MEM_SOFT_DIRTY 205#define VM_SOFTDIRTY INIT_VM_FLAG(SOFTDIRTY) 206#else 207#define VM_SOFTDIRTY VM_NONE 208#endif 209#define VM_MIXEDMAP INIT_VM_FLAG(MIXEDMAP) 210#define VM_HUGEPAGE INIT_VM_FLAG(HUGEPAGE) 211#define VM_NOHUGEPAGE INIT_VM_FLAG(NOHUGEPAGE) 212#define VM_MERGEABLE INIT_VM_FLAG(MERGEABLE) 213#define VM_STACK INIT_VM_FLAG(STACK) 214#ifdef CONFIG_STACK_GROWS_UP 215#define VM_STACK_EARLY INIT_VM_FLAG(STACK_EARLY) 216#else 217#define VM_STACK_EARLY VM_NONE 218#endif 219#ifdef CONFIG_ARCH_HAS_PKEYS 220#define VM_PKEY_SHIFT ((__force int)VMA_HIGH_ARCH_0_BIT) 221/* Despite the naming, these are FLAGS not bits. */ 222#define VM_PKEY_BIT0 INIT_VM_FLAG(PKEY_BIT0) 223#define VM_PKEY_BIT1 INIT_VM_FLAG(PKEY_BIT1) 224#define VM_PKEY_BIT2 INIT_VM_FLAG(PKEY_BIT2) 225#if CONFIG_ARCH_PKEY_BITS > 3 226#define VM_PKEY_BIT3 INIT_VM_FLAG(PKEY_BIT3) 227#else 228#define VM_PKEY_BIT3 VM_NONE 229#endif /* CONFIG_ARCH_PKEY_BITS > 3 */ 230#if CONFIG_ARCH_PKEY_BITS > 4 231#define VM_PKEY_BIT4 INIT_VM_FLAG(PKEY_BIT4) 232#else 233#define VM_PKEY_BIT4 VM_NONE 234#endif /* CONFIG_ARCH_PKEY_BITS > 4 */ 235#endif /* CONFIG_ARCH_HAS_PKEYS */ 236#if defined(CONFIG_X86_USER_SHADOW_STACK) || defined(CONFIG_ARM64_GCS) 237#define VM_SHADOW_STACK INIT_VM_FLAG(SHADOW_STACK) 238#else 239#define VM_SHADOW_STACK VM_NONE 240#endif 241#if defined(CONFIG_PPC64) 242#define VM_SAO INIT_VM_FLAG(SAO) 243#elif defined(CONFIG_PARISC) 244#define VM_GROWSUP INIT_VM_FLAG(GROWSUP) 245#elif defined(CONFIG_SPARC64) 246#define VM_SPARC_ADI INIT_VM_FLAG(SPARC_ADI) 247#define VM_ARCH_CLEAR INIT_VM_FLAG(ARCH_CLEAR) 248#elif defined(CONFIG_ARM64) 249#define VM_ARM64_BTI INIT_VM_FLAG(ARM64_BTI) 250#define VM_ARCH_CLEAR INIT_VM_FLAG(ARCH_CLEAR) 251#elif !defined(CONFIG_MMU) 252#define VM_MAPPED_COPY INIT_VM_FLAG(MAPPED_COPY) 253#endif 254#ifndef VM_GROWSUP 255#define VM_GROWSUP VM_NONE 256#endif 257#ifdef CONFIG_ARM64_MTE 258#define VM_MTE INIT_VM_FLAG(MTE) 259#define VM_MTE_ALLOWED INIT_VM_FLAG(MTE_ALLOWED) 260#else 261#define VM_MTE VM_NONE 262#define VM_MTE_ALLOWED VM_NONE 263#endif 264#ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR 265#define VM_UFFD_MINOR INIT_VM_FLAG(UFFD_MINOR) 266#else 267#define VM_UFFD_MINOR VM_NONE 268#endif 269#ifdef CONFIG_64BIT 270#define VM_ALLOW_ANY_UNCACHED INIT_VM_FLAG(ALLOW_ANY_UNCACHED) 271#define VM_SEALED INIT_VM_FLAG(SEALED) 272#else 273#define VM_ALLOW_ANY_UNCACHED VM_NONE 274#define VM_SEALED VM_NONE 275#endif 276#if defined(CONFIG_64BIT) || defined(CONFIG_PPC32) 277#define VM_DROPPABLE INIT_VM_FLAG(DROPPABLE) 278#else 279#define VM_DROPPABLE VM_NONE 280#endif 281 282/* Bits set in the VMA until the stack is in its final location */ 283#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY) 284 285#define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0) 286 287/* Common data flag combinations */ 288#define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \ 289 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) 290#define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \ 291 VM_MAYWRITE | VM_MAYEXEC) 292#define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \ 293 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) 294 295#ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */ 296#define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC 297#endif 298 299#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 300#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 301#endif 302 303#define VM_STARTGAP_FLAGS (VM_GROWSDOWN | VM_SHADOW_STACK) 304 305#define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 306 307/* VMA basic access permission flags */ 308#define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC) 309 310/* 311 * Special vmas that are non-mergable, non-mlock()able. 312 */ 313#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) 314 315#define DEFAULT_MAP_WINDOW ((1UL << 47) - PAGE_SIZE) 316#define TASK_SIZE_LOW DEFAULT_MAP_WINDOW 317#define TASK_SIZE_MAX DEFAULT_MAP_WINDOW 318#define STACK_TOP TASK_SIZE_LOW 319#define STACK_TOP_MAX TASK_SIZE_MAX 320 321/* This mask represents all the VMA flag bits used by mlock */ 322#define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT) 323 324#define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0) 325 326#define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \ 327 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC) 328 329#define RLIMIT_STACK 3 /* max stack size */ 330#define RLIMIT_MEMLOCK 8 /* max locked-in-memory address space */ 331 332#define CAP_IPC_LOCK 14 333 334/* 335 * Flags which should be 'sticky' on merge - that is, flags which, when one VMA 336 * possesses it but the other does not, the merged VMA should nonetheless have 337 * applied to it: 338 * 339 * VM_SOFTDIRTY - if a VMA is marked soft-dirty, that is has not had its 340 * references cleared via /proc/$pid/clear_refs, any merged VMA 341 * should be considered soft-dirty also as it operates at a VMA 342 * granularity. 343 */ 344#define VM_STICKY (VM_SOFTDIRTY | VM_MAYBE_GUARD) 345 346/* 347 * VMA flags we ignore for the purposes of merge, i.e. one VMA possessing one 348 * of these flags and the other not does not preclude a merge. 349 * 350 * VM_STICKY - When merging VMAs, VMA flags must match, unless they are 351 * 'sticky'. If any sticky flags exist in either VMA, we simply 352 * set all of them on the merged VMA. 353 */ 354#define VM_IGNORE_MERGE VM_STICKY 355 356/* 357 * Flags which should result in page tables being copied on fork. These are 358 * flags which indicate that the VMA maps page tables which cannot be 359 * reconsistuted upon page fault, so necessitate page table copying upon 360 * 361 * VM_PFNMAP / VM_MIXEDMAP - These contain kernel-mapped data which cannot be 362 * reasonably reconstructed on page fault. 363 * 364 * VM_UFFD_WP - Encodes metadata about an installed uffd 365 * write protect handler, which cannot be 366 * reconstructed on page fault. 367 * 368 * We always copy pgtables when dst_vma has uffd-wp 369 * enabled even if it's file-backed 370 * (e.g. shmem). Because when uffd-wp is enabled, 371 * pgtable contains uffd-wp protection information, 372 * that's something we can't retrieve from page cache, 373 * and skip copying will lose those info. 374 * 375 * VM_MAYBE_GUARD - Could contain page guard region markers which 376 * by design are a property of the page tables 377 * only and thus cannot be reconstructed on page 378 * fault. 379 */ 380#define VM_COPY_ON_FORK (VM_PFNMAP | VM_MIXEDMAP | VM_UFFD_WP | VM_MAYBE_GUARD) 381 382#define FIRST_USER_ADDRESS 0UL 383#define USER_PGTABLES_CEILING 0UL 384 385#define vma_policy(vma) NULL 386 387#define down_write_nest_lock(sem, nest_lock) 388 389#define pgprot_val(x) ((x).pgprot) 390#define __pgprot(x) ((pgprot_t) { (x) } ) 391 392#define for_each_vma(__vmi, __vma) \ 393 while (((__vma) = vma_next(&(__vmi))) != NULL) 394 395/* The MM code likes to work with exclusive end addresses */ 396#define for_each_vma_range(__vmi, __vma, __end) \ 397 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL) 398 399#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 400 401#define PHYS_PFN(x) ((unsigned long)((x) >> PAGE_SHIFT)) 402 403#define test_and_set_bit(nr, addr) __test_and_set_bit(nr, addr) 404#define test_and_clear_bit(nr, addr) __test_and_clear_bit(nr, addr) 405 406#define TASK_SIZE ((1ul << 47)-PAGE_SIZE) 407 408#define AS_MM_ALL_LOCKS 2 409 410/* We hardcode this for now. */ 411#define sysctl_max_map_count 0x1000000UL 412 413#define pgoff_t unsigned long 414typedef unsigned long pgprotval_t; 415typedef struct pgprot { pgprotval_t pgprot; } pgprot_t; 416typedef unsigned long vm_flags_t; 417typedef __bitwise unsigned int vm_fault_t; 418 419/* 420 * The shared stubs do not implement this, it amounts to an fprintf(STDERR,...) 421 * either way :) 422 */ 423#define pr_warn_once pr_err 424 425#define data_race(expr) expr 426 427#define ASSERT_EXCLUSIVE_WRITER(x) 428 429#define pgtable_supports_soft_dirty() 1 430 431/** 432 * swap - swap values of @a and @b 433 * @a: first value 434 * @b: second value 435 */ 436#define swap(a, b) \ 437 do { typeof(a) __tmp = (a); (a) = (b); (b) = __tmp; } while (0) 438 439struct kref { 440 refcount_t refcount; 441}; 442 443/* 444 * Define the task command name length as enum, then it can be visible to 445 * BPF programs. 446 */ 447enum { 448 TASK_COMM_LEN = 16, 449}; 450 451/* 452 * Flags for bug emulation. 453 * 454 * These occupy the top three bytes. 455 */ 456enum { 457 READ_IMPLIES_EXEC = 0x0400000, 458}; 459 460struct task_struct { 461 char comm[TASK_COMM_LEN]; 462 pid_t pid; 463 struct mm_struct *mm; 464 465 /* Used for emulating ABI behavior of previous Linux versions: */ 466 unsigned int personality; 467}; 468 469struct task_struct *get_current(void); 470#define current get_current() 471 472struct anon_vma { 473 struct anon_vma *root; 474 struct rb_root_cached rb_root; 475 476 /* Test fields. */ 477 bool was_cloned; 478 bool was_unlinked; 479}; 480 481struct anon_vma_chain { 482 struct anon_vma *anon_vma; 483 struct list_head same_vma; 484}; 485 486struct anon_vma_name { 487 struct kref kref; 488 /* The name needs to be at the end because it is dynamically sized. */ 489 char name[]; 490}; 491 492struct vma_iterator { 493 struct ma_state mas; 494}; 495 496#define VMA_ITERATOR(name, __mm, __addr) \ 497 struct vma_iterator name = { \ 498 .mas = { \ 499 .tree = &(__mm)->mm_mt, \ 500 .index = __addr, \ 501 .node = NULL, \ 502 .status = ma_start, \ 503 }, \ 504 } 505 506struct address_space { 507 struct rb_root_cached i_mmap; 508 unsigned long flags; 509 atomic_t i_mmap_writable; 510}; 511 512struct vm_userfaultfd_ctx {}; 513struct mempolicy {}; 514struct mmu_gather {}; 515struct mutex {}; 516#define DEFINE_MUTEX(mutexname) \ 517 struct mutex mutexname = {} 518 519#define DECLARE_BITMAP(name, bits) \ 520 unsigned long name[BITS_TO_LONGS(bits)] 521 522#define NUM_MM_FLAG_BITS (64) 523typedef struct { 524 __private DECLARE_BITMAP(__mm_flags, NUM_MM_FLAG_BITS); 525} mm_flags_t; 526 527/* 528 * Opaque type representing current VMA (vm_area_struct) flag state. Must be 529 * accessed via vma_flags_xxx() helper functions. 530 */ 531#define NUM_VMA_FLAG_BITS BITS_PER_LONG 532typedef struct { 533 DECLARE_BITMAP(__vma_flags, NUM_VMA_FLAG_BITS); 534} __private vma_flags_t; 535 536struct mm_struct { 537 struct maple_tree mm_mt; 538 int map_count; /* number of VMAs */ 539 unsigned long total_vm; /* Total pages mapped */ 540 unsigned long locked_vm; /* Pages that have PG_mlocked set */ 541 unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ 542 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ 543 unsigned long stack_vm; /* VM_STACK */ 544 545 unsigned long def_flags; 546 547 mm_flags_t flags; /* Must use mm_flags_* helpers to access */ 548}; 549 550struct vm_area_struct; 551 552 553/* What action should be taken after an .mmap_prepare call is complete? */ 554enum mmap_action_type { 555 MMAP_NOTHING, /* Mapping is complete, no further action. */ 556 MMAP_REMAP_PFN, /* Remap PFN range. */ 557 MMAP_IO_REMAP_PFN, /* I/O remap PFN range. */ 558}; 559 560/* 561 * Describes an action an mmap_prepare hook can instruct to be taken to complete 562 * the mapping of a VMA. Specified in vm_area_desc. 563 */ 564struct mmap_action { 565 union { 566 /* Remap range. */ 567 struct { 568 unsigned long start; 569 unsigned long start_pfn; 570 unsigned long size; 571 pgprot_t pgprot; 572 } remap; 573 }; 574 enum mmap_action_type type; 575 576 /* 577 * If specified, this hook is invoked after the selected action has been 578 * successfully completed. Note that the VMA write lock still held. 579 * 580 * The absolute minimum ought to be done here. 581 * 582 * Returns 0 on success, or an error code. 583 */ 584 int (*success_hook)(const struct vm_area_struct *vma); 585 586 /* 587 * If specified, this hook is invoked when an error occurred when 588 * attempting the selection action. 589 * 590 * The hook can return an error code in order to filter the error, but 591 * it is not valid to clear the error here. 592 */ 593 int (*error_hook)(int err); 594 595 /* 596 * This should be set in rare instances where the operation required 597 * that the rmap should not be able to access the VMA until 598 * completely set up. 599 */ 600 bool hide_from_rmap_until_complete :1; 601}; 602 603/* 604 * Describes a VMA that is about to be mmap()'ed. Drivers may choose to 605 * manipulate mutable fields which will cause those fields to be updated in the 606 * resultant VMA. 607 * 608 * Helper functions are not required for manipulating any field. 609 */ 610struct vm_area_desc { 611 /* Immutable state. */ 612 const struct mm_struct *const mm; 613 struct file *const file; /* May vary from vm_file in stacked callers. */ 614 unsigned long start; 615 unsigned long end; 616 617 /* Mutable fields. Populated with initial state. */ 618 pgoff_t pgoff; 619 struct file *vm_file; 620 union { 621 vm_flags_t vm_flags; 622 vma_flags_t vma_flags; 623 }; 624 pgprot_t page_prot; 625 626 /* Write-only fields. */ 627 const struct vm_operations_struct *vm_ops; 628 void *private_data; 629 630 /* Take further action? */ 631 struct mmap_action action; 632}; 633 634struct file_operations { 635 int (*mmap)(struct file *, struct vm_area_struct *); 636 int (*mmap_prepare)(struct vm_area_desc *); 637}; 638 639struct file { 640 struct address_space *f_mapping; 641 const struct file_operations *f_op; 642}; 643 644#define VMA_LOCK_OFFSET 0x40000000 645 646typedef struct { unsigned long v; } freeptr_t; 647 648struct vm_area_struct { 649 /* The first cache line has the info for VMA tree walking. */ 650 651 union { 652 struct { 653 /* VMA covers [vm_start; vm_end) addresses within mm */ 654 unsigned long vm_start; 655 unsigned long vm_end; 656 }; 657 freeptr_t vm_freeptr; /* Pointer used by SLAB_TYPESAFE_BY_RCU */ 658 }; 659 660 struct mm_struct *vm_mm; /* The address space we belong to. */ 661 pgprot_t vm_page_prot; /* Access permissions of this VMA. */ 662 663 /* 664 * Flags, see mm.h. 665 * To modify use vm_flags_{init|reset|set|clear|mod} functions. 666 */ 667 union { 668 const vm_flags_t vm_flags; 669 vma_flags_t flags; 670 }; 671 672#ifdef CONFIG_PER_VMA_LOCK 673 /* 674 * Can only be written (using WRITE_ONCE()) while holding both: 675 * - mmap_lock (in write mode) 676 * - vm_refcnt bit at VMA_LOCK_OFFSET is set 677 * Can be read reliably while holding one of: 678 * - mmap_lock (in read or write mode) 679 * - vm_refcnt bit at VMA_LOCK_OFFSET is set or vm_refcnt > 1 680 * Can be read unreliably (using READ_ONCE()) for pessimistic bailout 681 * while holding nothing (except RCU to keep the VMA struct allocated). 682 * 683 * This sequence counter is explicitly allowed to overflow; sequence 684 * counter reuse can only lead to occasional unnecessary use of the 685 * slowpath. 686 */ 687 unsigned int vm_lock_seq; 688#endif 689 690 /* 691 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma 692 * list, after a COW of one of the file pages. A MAP_SHARED vma 693 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack 694 * or brk vma (with NULL file) can only be in an anon_vma list. 695 */ 696 struct list_head anon_vma_chain; /* Serialized by mmap_lock & 697 * page_table_lock */ 698 struct anon_vma *anon_vma; /* Serialized by page_table_lock */ 699 700 /* Function pointers to deal with this struct. */ 701 const struct vm_operations_struct *vm_ops; 702 703 /* Information about our backing store: */ 704 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE 705 units */ 706 struct file * vm_file; /* File we map to (can be NULL). */ 707 void * vm_private_data; /* was vm_pte (shared mem) */ 708 709#ifdef CONFIG_SWAP 710 atomic_long_t swap_readahead_info; 711#endif 712#ifndef CONFIG_MMU 713 struct vm_region *vm_region; /* NOMMU mapping region */ 714#endif 715#ifdef CONFIG_NUMA 716 struct mempolicy *vm_policy; /* NUMA policy for the VMA */ 717#endif 718#ifdef CONFIG_NUMA_BALANCING 719 struct vma_numab_state *numab_state; /* NUMA Balancing state */ 720#endif 721#ifdef CONFIG_PER_VMA_LOCK 722 /* Unstable RCU readers are allowed to read this. */ 723 refcount_t vm_refcnt; 724#endif 725 /* 726 * For areas with an address space and backing store, 727 * linkage into the address_space->i_mmap interval tree. 728 * 729 */ 730 struct { 731 struct rb_node rb; 732 unsigned long rb_subtree_last; 733 } shared; 734#ifdef CONFIG_ANON_VMA_NAME 735 /* 736 * For private and shared anonymous mappings, a pointer to a null 737 * terminated string containing the name given to the vma, or NULL if 738 * unnamed. Serialized by mmap_lock. Use anon_vma_name to access. 739 */ 740 struct anon_vma_name *anon_name; 741#endif 742 struct vm_userfaultfd_ctx vm_userfaultfd_ctx; 743} __randomize_layout; 744 745struct vm_fault {}; 746 747struct vm_operations_struct { 748 void (*open)(struct vm_area_struct * area); 749 /** 750 * @close: Called when the VMA is being removed from the MM. 751 * Context: User context. May sleep. Caller holds mmap_lock. 752 */ 753 void (*close)(struct vm_area_struct * area); 754 /* Called any time before splitting to check if it's allowed */ 755 int (*may_split)(struct vm_area_struct *area, unsigned long addr); 756 int (*mremap)(struct vm_area_struct *area); 757 /* 758 * Called by mprotect() to make driver-specific permission 759 * checks before mprotect() is finalised. The VMA must not 760 * be modified. Returns 0 if mprotect() can proceed. 761 */ 762 int (*mprotect)(struct vm_area_struct *vma, unsigned long start, 763 unsigned long end, unsigned long newflags); 764 vm_fault_t (*fault)(struct vm_fault *vmf); 765 vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order); 766 vm_fault_t (*map_pages)(struct vm_fault *vmf, 767 pgoff_t start_pgoff, pgoff_t end_pgoff); 768 unsigned long (*pagesize)(struct vm_area_struct * area); 769 770 /* notification that a previously read-only page is about to become 771 * writable, if an error is returned it will cause a SIGBUS */ 772 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf); 773 774 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ 775 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf); 776 777 /* called by access_process_vm when get_user_pages() fails, typically 778 * for use by special VMAs. See also generic_access_phys() for a generic 779 * implementation useful for any iomem mapping. 780 */ 781 int (*access)(struct vm_area_struct *vma, unsigned long addr, 782 void *buf, int len, int write); 783 784 /* Called by the /proc/PID/maps code to ask the vma whether it 785 * has a special name. Returning non-NULL will also cause this 786 * vma to be dumped unconditionally. */ 787 const char *(*name)(struct vm_area_struct *vma); 788 789#ifdef CONFIG_NUMA 790 /* 791 * set_policy() op must add a reference to any non-NULL @new mempolicy 792 * to hold the policy upon return. Caller should pass NULL @new to 793 * remove a policy and fall back to surrounding context--i.e. do not 794 * install a MPOL_DEFAULT policy, nor the task or system default 795 * mempolicy. 796 */ 797 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 798 799 /* 800 * get_policy() op must add reference [mpol_get()] to any policy at 801 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 802 * in mm/mempolicy.c will do this automatically. 803 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 804 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock. 805 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 806 * must return NULL--i.e., do not "fallback" to task or system default 807 * policy. 808 */ 809 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 810 unsigned long addr, pgoff_t *ilx); 811#endif 812#ifdef CONFIG_FIND_NORMAL_PAGE 813 /* 814 * Called by vm_normal_page() for special PTEs in @vma at @addr. This 815 * allows for returning a "normal" page from vm_normal_page() even 816 * though the PTE indicates that the "struct page" either does not exist 817 * or should not be touched: "special". 818 * 819 * Do not add new users: this really only works when a "normal" page 820 * was mapped, but then the PTE got changed to something weird (+ 821 * marked special) that would not make pte_pfn() identify the originally 822 * inserted page. 823 */ 824 struct page *(*find_normal_page)(struct vm_area_struct *vma, 825 unsigned long addr); 826#endif /* CONFIG_FIND_NORMAL_PAGE */ 827}; 828 829struct vm_unmapped_area_info { 830#define VM_UNMAPPED_AREA_TOPDOWN 1 831 unsigned long flags; 832 unsigned long length; 833 unsigned long low_limit; 834 unsigned long high_limit; 835 unsigned long align_mask; 836 unsigned long align_offset; 837 unsigned long start_gap; 838}; 839 840struct pagetable_move_control { 841 struct vm_area_struct *old; /* Source VMA. */ 842 struct vm_area_struct *new; /* Destination VMA. */ 843 unsigned long old_addr; /* Address from which the move begins. */ 844 unsigned long old_end; /* Exclusive address at which old range ends. */ 845 unsigned long new_addr; /* Address to move page tables to. */ 846 unsigned long len_in; /* Bytes to remap specified by user. */ 847 848 bool need_rmap_locks; /* Do rmap locks need to be taken? */ 849 bool for_stack; /* Is this an early temp stack being moved? */ 850}; 851 852#define PAGETABLE_MOVE(name, old_, new_, old_addr_, new_addr_, len_) \ 853 struct pagetable_move_control name = { \ 854 .old = old_, \ 855 .new = new_, \ 856 .old_addr = old_addr_, \ 857 .old_end = (old_addr_) + (len_), \ 858 .new_addr = new_addr_, \ 859 .len_in = len_, \ 860 } 861 862static inline void vma_iter_invalidate(struct vma_iterator *vmi) 863{ 864 mas_pause(&vmi->mas); 865} 866 867static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot) 868{ 869 return __pgprot(pgprot_val(oldprot) | pgprot_val(newprot)); 870} 871 872static inline pgprot_t vm_get_page_prot(vm_flags_t vm_flags) 873{ 874 return __pgprot(vm_flags); 875} 876 877static inline bool is_shared_maywrite(vm_flags_t vm_flags) 878{ 879 return (vm_flags & (VM_SHARED | VM_MAYWRITE)) == 880 (VM_SHARED | VM_MAYWRITE); 881} 882 883static inline bool vma_is_shared_maywrite(struct vm_area_struct *vma) 884{ 885 return is_shared_maywrite(vma->vm_flags); 886} 887 888static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi) 889{ 890 /* 891 * Uses mas_find() to get the first VMA when the iterator starts. 892 * Calling mas_next() could skip the first entry. 893 */ 894 return mas_find(&vmi->mas, ULONG_MAX); 895} 896 897/* 898 * WARNING: to avoid racing with vma_mark_attached()/vma_mark_detached(), these 899 * assertions should be made either under mmap_write_lock or when the object 900 * has been isolated under mmap_write_lock, ensuring no competing writers. 901 */ 902static inline void vma_assert_attached(struct vm_area_struct *vma) 903{ 904 WARN_ON_ONCE(!refcount_read(&vma->vm_refcnt)); 905} 906 907static inline void vma_assert_detached(struct vm_area_struct *vma) 908{ 909 WARN_ON_ONCE(refcount_read(&vma->vm_refcnt)); 910} 911 912static inline void vma_assert_write_locked(struct vm_area_struct *); 913static inline void vma_mark_attached(struct vm_area_struct *vma) 914{ 915 vma_assert_write_locked(vma); 916 vma_assert_detached(vma); 917 refcount_set_release(&vma->vm_refcnt, 1); 918} 919 920static inline void vma_mark_detached(struct vm_area_struct *vma) 921{ 922 vma_assert_write_locked(vma); 923 vma_assert_attached(vma); 924 /* We are the only writer, so no need to use vma_refcount_put(). */ 925 if (unlikely(!refcount_dec_and_test(&vma->vm_refcnt))) { 926 /* 927 * Reader must have temporarily raised vm_refcnt but it will 928 * drop it without using the vma since vma is write-locked. 929 */ 930 } 931} 932 933extern const struct vm_operations_struct vma_dummy_vm_ops; 934 935extern unsigned long rlimit(unsigned int limit); 936 937static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm) 938{ 939 memset(vma, 0, sizeof(*vma)); 940 vma->vm_mm = mm; 941 vma->vm_ops = &vma_dummy_vm_ops; 942 INIT_LIST_HEAD(&vma->anon_vma_chain); 943 vma->vm_lock_seq = UINT_MAX; 944} 945 946/* 947 * These are defined in vma.h, but sadly vm_stat_account() is referenced by 948 * kernel/fork.c, so we have to these broadly available there, and temporarily 949 * define them here to resolve the dependency cycle. 950 */ 951 952#define is_exec_mapping(flags) \ 953 ((flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC) 954 955#define is_stack_mapping(flags) \ 956 (((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK)) 957 958#define is_data_mapping(flags) \ 959 ((flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE) 960 961static inline void vm_stat_account(struct mm_struct *mm, vm_flags_t flags, 962 long npages) 963{ 964 WRITE_ONCE(mm->total_vm, READ_ONCE(mm->total_vm)+npages); 965 966 if (is_exec_mapping(flags)) 967 mm->exec_vm += npages; 968 else if (is_stack_mapping(flags)) 969 mm->stack_vm += npages; 970 else if (is_data_mapping(flags)) 971 mm->data_vm += npages; 972} 973 974#undef is_exec_mapping 975#undef is_stack_mapping 976#undef is_data_mapping 977 978/* Currently stubbed but we may later wish to un-stub. */ 979static inline void vm_acct_memory(long pages); 980static inline void vm_unacct_memory(long pages) 981{ 982 vm_acct_memory(-pages); 983} 984 985static inline void mapping_allow_writable(struct address_space *mapping) 986{ 987 atomic_inc(&mapping->i_mmap_writable); 988} 989 990static inline void vma_set_range(struct vm_area_struct *vma, 991 unsigned long start, unsigned long end, 992 pgoff_t pgoff) 993{ 994 vma->vm_start = start; 995 vma->vm_end = end; 996 vma->vm_pgoff = pgoff; 997} 998 999static inline 1000struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max) 1001{ 1002 return mas_find(&vmi->mas, max - 1); 1003} 1004 1005static inline int vma_iter_clear_gfp(struct vma_iterator *vmi, 1006 unsigned long start, unsigned long end, gfp_t gfp) 1007{ 1008 __mas_set_range(&vmi->mas, start, end - 1); 1009 mas_store_gfp(&vmi->mas, NULL, gfp); 1010 if (unlikely(mas_is_err(&vmi->mas))) 1011 return -ENOMEM; 1012 1013 return 0; 1014} 1015 1016static inline void mmap_assert_locked(struct mm_struct *); 1017static inline struct vm_area_struct *find_vma_intersection(struct mm_struct *mm, 1018 unsigned long start_addr, 1019 unsigned long end_addr) 1020{ 1021 unsigned long index = start_addr; 1022 1023 mmap_assert_locked(mm); 1024 return mt_find(&mm->mm_mt, &index, end_addr - 1); 1025} 1026 1027static inline 1028struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr) 1029{ 1030 return mtree_load(&mm->mm_mt, addr); 1031} 1032 1033static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi) 1034{ 1035 return mas_prev(&vmi->mas, 0); 1036} 1037 1038static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr) 1039{ 1040 mas_set(&vmi->mas, addr); 1041} 1042 1043static inline bool vma_is_anonymous(struct vm_area_struct *vma) 1044{ 1045 return !vma->vm_ops; 1046} 1047 1048/* Defined in vma.h, so temporarily define here to avoid circular dependency. */ 1049#define vma_iter_load(vmi) \ 1050 mas_walk(&(vmi)->mas) 1051 1052static inline struct vm_area_struct * 1053find_vma_prev(struct mm_struct *mm, unsigned long addr, 1054 struct vm_area_struct **pprev) 1055{ 1056 struct vm_area_struct *vma; 1057 VMA_ITERATOR(vmi, mm, addr); 1058 1059 vma = vma_iter_load(&vmi); 1060 *pprev = vma_prev(&vmi); 1061 if (!vma) 1062 vma = vma_next(&vmi); 1063 return vma; 1064} 1065 1066#undef vma_iter_load 1067 1068static inline void vma_iter_init(struct vma_iterator *vmi, 1069 struct mm_struct *mm, unsigned long addr) 1070{ 1071 mas_init(&vmi->mas, &mm->mm_mt, addr); 1072} 1073 1074/* Stubbed functions. */ 1075 1076static inline struct anon_vma_name *anon_vma_name(struct vm_area_struct *vma) 1077{ 1078 return NULL; 1079} 1080 1081static inline bool is_mergeable_vm_userfaultfd_ctx(struct vm_area_struct *vma, 1082 struct vm_userfaultfd_ctx vm_ctx) 1083{ 1084 return true; 1085} 1086 1087static inline bool anon_vma_name_eq(struct anon_vma_name *anon_name1, 1088 struct anon_vma_name *anon_name2) 1089{ 1090 return true; 1091} 1092 1093static inline void might_sleep(void) 1094{ 1095} 1096 1097static inline unsigned long vma_pages(struct vm_area_struct *vma) 1098{ 1099 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1100} 1101 1102static inline void fput(struct file *file) 1103{ 1104} 1105 1106static inline void mpol_put(struct mempolicy *pol) 1107{ 1108} 1109 1110static inline void lru_add_drain(void) 1111{ 1112} 1113 1114static inline void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm) 1115{ 1116} 1117 1118static inline void update_hiwater_rss(struct mm_struct *mm) 1119{ 1120} 1121 1122static inline void update_hiwater_vm(struct mm_struct *mm) 1123{ 1124} 1125 1126static inline void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas, 1127 struct vm_area_struct *vma, unsigned long start_addr, 1128 unsigned long end_addr, unsigned long tree_end) 1129{ 1130} 1131 1132static inline void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas, 1133 struct vm_area_struct *vma, unsigned long floor, 1134 unsigned long ceiling, bool mm_wr_locked) 1135{ 1136} 1137 1138static inline void mapping_unmap_writable(struct address_space *mapping) 1139{ 1140} 1141 1142static inline void flush_dcache_mmap_lock(struct address_space *mapping) 1143{ 1144} 1145 1146static inline void tlb_finish_mmu(struct mmu_gather *tlb) 1147{ 1148} 1149 1150static inline struct file *get_file(struct file *f) 1151{ 1152 return f; 1153} 1154 1155static inline int vma_dup_policy(struct vm_area_struct *src, struct vm_area_struct *dst) 1156{ 1157 return 0; 1158} 1159 1160static inline int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src) 1161{ 1162 /* For testing purposes. We indicate that an anon_vma has been cloned. */ 1163 if (src->anon_vma != NULL) { 1164 dst->anon_vma = src->anon_vma; 1165 dst->anon_vma->was_cloned = true; 1166 } 1167 1168 return 0; 1169} 1170 1171static inline void vma_start_write(struct vm_area_struct *vma) 1172{ 1173 /* Used to indicate to tests that a write operation has begun. */ 1174 vma->vm_lock_seq++; 1175} 1176 1177static inline __must_check 1178int vma_start_write_killable(struct vm_area_struct *vma) 1179{ 1180 /* Used to indicate to tests that a write operation has begun. */ 1181 vma->vm_lock_seq++; 1182 return 0; 1183} 1184 1185static inline void vma_adjust_trans_huge(struct vm_area_struct *vma, 1186 unsigned long start, 1187 unsigned long end, 1188 struct vm_area_struct *next) 1189{ 1190} 1191 1192static inline void hugetlb_split(struct vm_area_struct *, unsigned long) {} 1193 1194static inline void vma_iter_free(struct vma_iterator *vmi) 1195{ 1196 mas_destroy(&vmi->mas); 1197} 1198 1199static inline 1200struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi) 1201{ 1202 return mas_next_range(&vmi->mas, ULONG_MAX); 1203} 1204 1205static inline void vm_acct_memory(long pages) 1206{ 1207} 1208 1209static inline void vma_interval_tree_insert(struct vm_area_struct *vma, 1210 struct rb_root_cached *rb) 1211{ 1212} 1213 1214static inline void vma_interval_tree_remove(struct vm_area_struct *vma, 1215 struct rb_root_cached *rb) 1216{ 1217} 1218 1219static inline void flush_dcache_mmap_unlock(struct address_space *mapping) 1220{ 1221} 1222 1223static inline void anon_vma_interval_tree_insert(struct anon_vma_chain *avc, 1224 struct rb_root_cached *rb) 1225{ 1226} 1227 1228static inline void anon_vma_interval_tree_remove(struct anon_vma_chain *avc, 1229 struct rb_root_cached *rb) 1230{ 1231} 1232 1233static inline void uprobe_mmap(struct vm_area_struct *vma) 1234{ 1235} 1236 1237static inline void uprobe_munmap(struct vm_area_struct *vma, 1238 unsigned long start, unsigned long end) 1239{ 1240} 1241 1242static inline void i_mmap_lock_write(struct address_space *mapping) 1243{ 1244} 1245 1246static inline void anon_vma_lock_write(struct anon_vma *anon_vma) 1247{ 1248} 1249 1250static inline void vma_assert_write_locked(struct vm_area_struct *vma) 1251{ 1252} 1253 1254static inline void unlink_anon_vmas(struct vm_area_struct *vma) 1255{ 1256 /* For testing purposes, indicate that the anon_vma was unlinked. */ 1257 vma->anon_vma->was_unlinked = true; 1258} 1259 1260static inline void anon_vma_unlock_write(struct anon_vma *anon_vma) 1261{ 1262} 1263 1264static inline void i_mmap_unlock_write(struct address_space *mapping) 1265{ 1266} 1267 1268static inline void anon_vma_merge(struct vm_area_struct *vma, 1269 struct vm_area_struct *next) 1270{ 1271} 1272 1273static inline int userfaultfd_unmap_prep(struct vm_area_struct *vma, 1274 unsigned long start, 1275 unsigned long end, 1276 struct list_head *unmaps) 1277{ 1278 return 0; 1279} 1280 1281static inline void mmap_write_downgrade(struct mm_struct *mm) 1282{ 1283} 1284 1285static inline void mmap_read_unlock(struct mm_struct *mm) 1286{ 1287} 1288 1289static inline void mmap_write_unlock(struct mm_struct *mm) 1290{ 1291} 1292 1293static inline int mmap_write_lock_killable(struct mm_struct *mm) 1294{ 1295 return 0; 1296} 1297 1298static inline bool can_modify_mm(struct mm_struct *mm, 1299 unsigned long start, 1300 unsigned long end) 1301{ 1302 return true; 1303} 1304 1305static inline void arch_unmap(struct mm_struct *mm, 1306 unsigned long start, 1307 unsigned long end) 1308{ 1309} 1310 1311static inline void mmap_assert_locked(struct mm_struct *mm) 1312{ 1313} 1314 1315static inline bool mpol_equal(struct mempolicy *a, struct mempolicy *b) 1316{ 1317 return true; 1318} 1319 1320static inline void khugepaged_enter_vma(struct vm_area_struct *vma, 1321 vm_flags_t vm_flags) 1322{ 1323} 1324 1325static inline bool mapping_can_writeback(struct address_space *mapping) 1326{ 1327 return true; 1328} 1329 1330static inline bool is_vm_hugetlb_page(struct vm_area_struct *vma) 1331{ 1332 return false; 1333} 1334 1335static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma) 1336{ 1337 return false; 1338} 1339 1340static inline bool userfaultfd_wp(struct vm_area_struct *vma) 1341{ 1342 return false; 1343} 1344 1345static inline void mmap_assert_write_locked(struct mm_struct *mm) 1346{ 1347} 1348 1349static inline void mutex_lock(struct mutex *lock) 1350{ 1351} 1352 1353static inline void mutex_unlock(struct mutex *lock) 1354{ 1355} 1356 1357static inline bool mutex_is_locked(struct mutex *lock) 1358{ 1359 return true; 1360} 1361 1362static inline bool signal_pending(void *p) 1363{ 1364 return false; 1365} 1366 1367static inline bool is_file_hugepages(struct file *file) 1368{ 1369 return false; 1370} 1371 1372static inline int security_vm_enough_memory_mm(struct mm_struct *mm, long pages) 1373{ 1374 return 0; 1375} 1376 1377static inline bool may_expand_vm(struct mm_struct *mm, vm_flags_t flags, 1378 unsigned long npages) 1379{ 1380 return true; 1381} 1382 1383static inline int shmem_zero_setup(struct vm_area_struct *vma) 1384{ 1385 return 0; 1386} 1387 1388static inline void vma_set_anonymous(struct vm_area_struct *vma) 1389{ 1390 vma->vm_ops = NULL; 1391} 1392 1393static inline void ksm_add_vma(struct vm_area_struct *vma) 1394{ 1395} 1396 1397static inline void perf_event_mmap(struct vm_area_struct *vma) 1398{ 1399} 1400 1401static inline bool vma_is_dax(struct vm_area_struct *vma) 1402{ 1403 return false; 1404} 1405 1406static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) 1407{ 1408 return NULL; 1409} 1410 1411bool vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot); 1412 1413/* Update vma->vm_page_prot to reflect vma->vm_flags. */ 1414static inline void vma_set_page_prot(struct vm_area_struct *vma) 1415{ 1416 vm_flags_t vm_flags = vma->vm_flags; 1417 pgprot_t vm_page_prot; 1418 1419 /* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */ 1420 vm_page_prot = pgprot_modify(vma->vm_page_prot, vm_get_page_prot(vm_flags)); 1421 1422 if (vma_wants_writenotify(vma, vm_page_prot)) { 1423 vm_flags &= ~VM_SHARED; 1424 /* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */ 1425 vm_page_prot = pgprot_modify(vm_page_prot, vm_get_page_prot(vm_flags)); 1426 } 1427 /* remove_protection_ptes reads vma->vm_page_prot without mmap_lock */ 1428 WRITE_ONCE(vma->vm_page_prot, vm_page_prot); 1429} 1430 1431static inline bool arch_validate_flags(vm_flags_t flags) 1432{ 1433 return true; 1434} 1435 1436static inline void vma_close(struct vm_area_struct *vma) 1437{ 1438} 1439 1440static inline int mmap_file(struct file *file, struct vm_area_struct *vma) 1441{ 1442 return 0; 1443} 1444 1445static inline unsigned long stack_guard_start_gap(struct vm_area_struct *vma) 1446{ 1447 if (vma->vm_flags & VM_GROWSDOWN) 1448 return stack_guard_gap; 1449 1450 /* See reasoning around the VM_SHADOW_STACK definition */ 1451 if (vma->vm_flags & VM_SHADOW_STACK) 1452 return PAGE_SIZE; 1453 1454 return 0; 1455} 1456 1457static inline unsigned long vm_start_gap(struct vm_area_struct *vma) 1458{ 1459 unsigned long gap = stack_guard_start_gap(vma); 1460 unsigned long vm_start = vma->vm_start; 1461 1462 vm_start -= gap; 1463 if (vm_start > vma->vm_start) 1464 vm_start = 0; 1465 return vm_start; 1466} 1467 1468static inline unsigned long vm_end_gap(struct vm_area_struct *vma) 1469{ 1470 unsigned long vm_end = vma->vm_end; 1471 1472 if (vma->vm_flags & VM_GROWSUP) { 1473 vm_end += stack_guard_gap; 1474 if (vm_end < vma->vm_end) 1475 vm_end = -PAGE_SIZE; 1476 } 1477 return vm_end; 1478} 1479 1480static inline int is_hugepage_only_range(struct mm_struct *mm, 1481 unsigned long addr, unsigned long len) 1482{ 1483 return 0; 1484} 1485 1486static inline bool vma_is_accessible(struct vm_area_struct *vma) 1487{ 1488 return vma->vm_flags & VM_ACCESS_FLAGS; 1489} 1490 1491static inline bool capable(int cap) 1492{ 1493 return true; 1494} 1495 1496static inline bool mlock_future_ok(const struct mm_struct *mm, 1497 vm_flags_t vm_flags, unsigned long bytes) 1498{ 1499 unsigned long locked_pages, limit_pages; 1500 1501 if (!(vm_flags & VM_LOCKED) || capable(CAP_IPC_LOCK)) 1502 return true; 1503 1504 locked_pages = bytes >> PAGE_SHIFT; 1505 locked_pages += mm->locked_vm; 1506 1507 limit_pages = rlimit(RLIMIT_MEMLOCK); 1508 limit_pages >>= PAGE_SHIFT; 1509 1510 return locked_pages <= limit_pages; 1511} 1512 1513static inline int __anon_vma_prepare(struct vm_area_struct *vma) 1514{ 1515 struct anon_vma *anon_vma = calloc(1, sizeof(struct anon_vma)); 1516 1517 if (!anon_vma) 1518 return -ENOMEM; 1519 1520 anon_vma->root = anon_vma; 1521 vma->anon_vma = anon_vma; 1522 1523 return 0; 1524} 1525 1526static inline int anon_vma_prepare(struct vm_area_struct *vma) 1527{ 1528 if (likely(vma->anon_vma)) 1529 return 0; 1530 1531 return __anon_vma_prepare(vma); 1532} 1533 1534static inline void userfaultfd_unmap_complete(struct mm_struct *mm, 1535 struct list_head *uf) 1536{ 1537} 1538 1539#define ACCESS_PRIVATE(p, member) ((p)->member) 1540 1541#define bitmap_size(nbits) (ALIGN(nbits, BITS_PER_LONG) / BITS_PER_BYTE) 1542 1543static __always_inline void bitmap_zero(unsigned long *dst, unsigned int nbits) 1544{ 1545 unsigned int len = bitmap_size(nbits); 1546 1547 if (small_const_nbits(nbits)) 1548 *dst = 0; 1549 else 1550 memset(dst, 0, len); 1551} 1552 1553static inline bool mm_flags_test(int flag, const struct mm_struct *mm) 1554{ 1555 return test_bit(flag, ACCESS_PRIVATE(&mm->flags, __mm_flags)); 1556} 1557 1558/* Clears all bits in the VMA flags bitmap, non-atomically. */ 1559static inline void vma_flags_clear_all(vma_flags_t *flags) 1560{ 1561 bitmap_zero(ACCESS_PRIVATE(flags, __vma_flags), NUM_VMA_FLAG_BITS); 1562} 1563 1564/* 1565 * Copy value to the first system word of VMA flags, non-atomically. 1566 * 1567 * IMPORTANT: This does not overwrite bytes past the first system word. The 1568 * caller must account for this. 1569 */ 1570static inline void vma_flags_overwrite_word(vma_flags_t *flags, unsigned long value) 1571{ 1572 *ACCESS_PRIVATE(flags, __vma_flags) = value; 1573} 1574 1575/* 1576 * Copy value to the first system word of VMA flags ONCE, non-atomically. 1577 * 1578 * IMPORTANT: This does not overwrite bytes past the first system word. The 1579 * caller must account for this. 1580 */ 1581static inline void vma_flags_overwrite_word_once(vma_flags_t *flags, unsigned long value) 1582{ 1583 unsigned long *bitmap = ACCESS_PRIVATE(flags, __vma_flags); 1584 1585 WRITE_ONCE(*bitmap, value); 1586} 1587 1588/* Update the first system word of VMA flags setting bits, non-atomically. */ 1589static inline void vma_flags_set_word(vma_flags_t *flags, unsigned long value) 1590{ 1591 unsigned long *bitmap = ACCESS_PRIVATE(flags, __vma_flags); 1592 1593 *bitmap |= value; 1594} 1595 1596/* Update the first system word of VMA flags clearing bits, non-atomically. */ 1597static inline void vma_flags_clear_word(vma_flags_t *flags, unsigned long value) 1598{ 1599 unsigned long *bitmap = ACCESS_PRIVATE(flags, __vma_flags); 1600 1601 *bitmap &= ~value; 1602} 1603 1604 1605/* Use when VMA is not part of the VMA tree and needs no locking */ 1606static inline void vm_flags_init(struct vm_area_struct *vma, 1607 vm_flags_t flags) 1608{ 1609 vma_flags_clear_all(&vma->flags); 1610 vma_flags_overwrite_word(&vma->flags, flags); 1611} 1612 1613/* 1614 * Use when VMA is part of the VMA tree and modifications need coordination 1615 * Note: vm_flags_reset and vm_flags_reset_once do not lock the vma and 1616 * it should be locked explicitly beforehand. 1617 */ 1618static inline void vm_flags_reset(struct vm_area_struct *vma, 1619 vm_flags_t flags) 1620{ 1621 vma_assert_write_locked(vma); 1622 vm_flags_init(vma, flags); 1623} 1624 1625static inline void vm_flags_reset_once(struct vm_area_struct *vma, 1626 vm_flags_t flags) 1627{ 1628 vma_assert_write_locked(vma); 1629 /* 1630 * The user should only be interested in avoiding reordering of 1631 * assignment to the first word. 1632 */ 1633 vma_flags_clear_all(&vma->flags); 1634 vma_flags_overwrite_word_once(&vma->flags, flags); 1635} 1636 1637static inline void vm_flags_set(struct vm_area_struct *vma, 1638 vm_flags_t flags) 1639{ 1640 vma_start_write(vma); 1641 vma_flags_set_word(&vma->flags, flags); 1642} 1643 1644static inline void vm_flags_clear(struct vm_area_struct *vma, 1645 vm_flags_t flags) 1646{ 1647 vma_start_write(vma); 1648 vma_flags_clear_word(&vma->flags, flags); 1649} 1650 1651/* 1652 * Denies creating a writable executable mapping or gaining executable permissions. 1653 * 1654 * This denies the following: 1655 * 1656 * a) mmap(PROT_WRITE | PROT_EXEC) 1657 * 1658 * b) mmap(PROT_WRITE) 1659 * mprotect(PROT_EXEC) 1660 * 1661 * c) mmap(PROT_WRITE) 1662 * mprotect(PROT_READ) 1663 * mprotect(PROT_EXEC) 1664 * 1665 * But allows the following: 1666 * 1667 * d) mmap(PROT_READ | PROT_EXEC) 1668 * mmap(PROT_READ | PROT_EXEC | PROT_BTI) 1669 * 1670 * This is only applicable if the user has set the Memory-Deny-Write-Execute 1671 * (MDWE) protection mask for the current process. 1672 * 1673 * @old specifies the VMA flags the VMA originally possessed, and @new the ones 1674 * we propose to set. 1675 * 1676 * Return: false if proposed change is OK, true if not ok and should be denied. 1677 */ 1678static inline bool map_deny_write_exec(unsigned long old, unsigned long new) 1679{ 1680 /* If MDWE is disabled, we have nothing to deny. */ 1681 if (mm_flags_test(MMF_HAS_MDWE, current->mm)) 1682 return false; 1683 1684 /* If the new VMA is not executable, we have nothing to deny. */ 1685 if (!(new & VM_EXEC)) 1686 return false; 1687 1688 /* Under MDWE we do not accept newly writably executable VMAs... */ 1689 if (new & VM_WRITE) 1690 return true; 1691 1692 /* ...nor previously non-executable VMAs becoming executable. */ 1693 if (!(old & VM_EXEC)) 1694 return true; 1695 1696 return false; 1697} 1698 1699static inline int mapping_map_writable(struct address_space *mapping) 1700{ 1701 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ? 1702 0 : -EPERM; 1703} 1704 1705static inline unsigned long move_page_tables(struct pagetable_move_control *pmc) 1706{ 1707 return 0; 1708} 1709 1710static inline void free_pgd_range(struct mmu_gather *tlb, 1711 unsigned long addr, unsigned long end, 1712 unsigned long floor, unsigned long ceiling) 1713{ 1714} 1715 1716static inline int ksm_execve(struct mm_struct *mm) 1717{ 1718 return 0; 1719} 1720 1721static inline void ksm_exit(struct mm_struct *mm) 1722{ 1723} 1724 1725static inline void vma_lock_init(struct vm_area_struct *vma, bool reset_refcnt) 1726{ 1727 if (reset_refcnt) 1728 refcount_set(&vma->vm_refcnt, 0); 1729} 1730 1731static inline void vma_numab_state_init(struct vm_area_struct *vma) 1732{ 1733} 1734 1735static inline void vma_numab_state_free(struct vm_area_struct *vma) 1736{ 1737} 1738 1739static inline void dup_anon_vma_name(struct vm_area_struct *orig_vma, 1740 struct vm_area_struct *new_vma) 1741{ 1742} 1743 1744static inline void free_anon_vma_name(struct vm_area_struct *vma) 1745{ 1746} 1747 1748/* Declared in vma.h. */ 1749static inline void set_vma_from_desc(struct vm_area_struct *vma, 1750 struct vm_area_desc *desc); 1751 1752static inline void mmap_action_prepare(struct mmap_action *action, 1753 struct vm_area_desc *desc) 1754{ 1755} 1756 1757static inline int mmap_action_complete(struct mmap_action *action, 1758 struct vm_area_struct *vma) 1759{ 1760 return 0; 1761} 1762 1763static inline int __compat_vma_mmap(const struct file_operations *f_op, 1764 struct file *file, struct vm_area_struct *vma) 1765{ 1766 struct vm_area_desc desc = { 1767 .mm = vma->vm_mm, 1768 .file = file, 1769 .start = vma->vm_start, 1770 .end = vma->vm_end, 1771 1772 .pgoff = vma->vm_pgoff, 1773 .vm_file = vma->vm_file, 1774 .vm_flags = vma->vm_flags, 1775 .page_prot = vma->vm_page_prot, 1776 1777 .action.type = MMAP_NOTHING, /* Default */ 1778 }; 1779 int err; 1780 1781 err = f_op->mmap_prepare(&desc); 1782 if (err) 1783 return err; 1784 1785 mmap_action_prepare(&desc.action, &desc); 1786 set_vma_from_desc(vma, &desc); 1787 return mmap_action_complete(&desc.action, vma); 1788} 1789 1790static inline int compat_vma_mmap(struct file *file, 1791 struct vm_area_struct *vma) 1792{ 1793 return __compat_vma_mmap(file->f_op, file, vma); 1794} 1795 1796/* Did the driver provide valid mmap hook configuration? */ 1797static inline bool can_mmap_file(struct file *file) 1798{ 1799 bool has_mmap = file->f_op->mmap; 1800 bool has_mmap_prepare = file->f_op->mmap_prepare; 1801 1802 /* Hooks are mutually exclusive. */ 1803 if (WARN_ON_ONCE(has_mmap && has_mmap_prepare)) 1804 return false; 1805 if (!has_mmap && !has_mmap_prepare) 1806 return false; 1807 1808 return true; 1809} 1810 1811static inline int vfs_mmap(struct file *file, struct vm_area_struct *vma) 1812{ 1813 if (file->f_op->mmap_prepare) 1814 return compat_vma_mmap(file, vma); 1815 1816 return file->f_op->mmap(file, vma); 1817} 1818 1819static inline int vfs_mmap_prepare(struct file *file, struct vm_area_desc *desc) 1820{ 1821 return file->f_op->mmap_prepare(desc); 1822} 1823 1824static inline void fixup_hugetlb_reservations(struct vm_area_struct *vma) 1825{ 1826} 1827 1828static inline void vma_set_file(struct vm_area_struct *vma, struct file *file) 1829{ 1830 /* Changing an anonymous vma with this is illegal */ 1831 get_file(file); 1832 swap(vma->vm_file, file); 1833 fput(file); 1834} 1835 1836static inline bool shmem_file(struct file *file) 1837{ 1838 return false; 1839} 1840 1841static inline vm_flags_t ksm_vma_flags(const struct mm_struct *mm, 1842 const struct file *file, vm_flags_t vm_flags) 1843{ 1844 return vm_flags; 1845} 1846 1847static inline void remap_pfn_range_prepare(struct vm_area_desc *desc, unsigned long pfn) 1848{ 1849} 1850 1851static inline int remap_pfn_range_complete(struct vm_area_struct *vma, unsigned long addr, 1852 unsigned long pfn, unsigned long size, pgprot_t pgprot) 1853{ 1854 return 0; 1855} 1856 1857static inline int do_munmap(struct mm_struct *, unsigned long, size_t, 1858 struct list_head *uf) 1859{ 1860 return 0; 1861} 1862 1863#endif /* __MM_VMA_INTERNAL_H */