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/list.h>
25 #include <linux/maple_tree.h>
26 #include <linux/mm.h>
27 #include <linux/rbtree.h>
28 #include <linux/rwsem.h>
29
30 extern unsigned long stack_guard_gap;
31 #ifdef CONFIG_MMU
32 extern unsigned long mmap_min_addr;
33 extern unsigned long dac_mmap_min_addr;
34 #else
35 #define mmap_min_addr 0UL
36 #define dac_mmap_min_addr 0UL
37 #endif
38
39 #define VM_WARN_ON(_expr) (WARN_ON(_expr))
40 #define VM_WARN_ON_ONCE(_expr) (WARN_ON_ONCE(_expr))
41 #define VM_WARN_ON_VMG(_expr, _vmg) (WARN_ON(_expr))
42 #define VM_BUG_ON(_expr) (BUG_ON(_expr))
43 #define VM_BUG_ON_VMA(_expr, _vma) (BUG_ON(_expr))
44
45 #define MMF_HAS_MDWE 28
46
47 #define VM_NONE 0x00000000
48 #define VM_READ 0x00000001
49 #define VM_WRITE 0x00000002
50 #define VM_EXEC 0x00000004
51 #define VM_SHARED 0x00000008
52 #define VM_MAYREAD 0x00000010
53 #define VM_MAYWRITE 0x00000020
54 #define VM_MAYEXEC 0x00000040
55 #define VM_GROWSDOWN 0x00000100
56 #define VM_PFNMAP 0x00000400
57 #define VM_LOCKED 0x00002000
58 #define VM_IO 0x00004000
59 #define VM_DONTEXPAND 0x00040000
60 #define VM_LOCKONFAULT 0x00080000
61 #define VM_ACCOUNT 0x00100000
62 #define VM_NORESERVE 0x00200000
63 #define VM_MIXEDMAP 0x10000000
64 #define VM_STACK VM_GROWSDOWN
65 #define VM_SHADOW_STACK VM_NONE
66 #define VM_SOFTDIRTY 0
67 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
68 #define VM_GROWSUP VM_NONE
69
70 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
71 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
72
73 /* This mask represents all the VMA flag bits used by mlock */
74 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
75
76 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
77
78 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
79 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
80
81 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_TSK_EXEC
82
83 #define VM_STARTGAP_FLAGS (VM_GROWSDOWN | VM_SHADOW_STACK)
84
85 #define RLIMIT_STACK 3 /* max stack size */
86 #define RLIMIT_MEMLOCK 8 /* max locked-in-memory address space */
87
88 #define CAP_IPC_LOCK 14
89
90 #ifdef CONFIG_64BIT
91 /* VM is sealed, in vm_flags */
92 #define VM_SEALED _BITUL(63)
93 #endif
94
95 #define FIRST_USER_ADDRESS 0UL
96 #define USER_PGTABLES_CEILING 0UL
97
98 #define vma_policy(vma) NULL
99
100 #define down_write_nest_lock(sem, nest_lock)
101
102 #define pgprot_val(x) ((x).pgprot)
103 #define __pgprot(x) ((pgprot_t) { (x) } )
104
105 #define for_each_vma(__vmi, __vma) \
106 while (((__vma) = vma_next(&(__vmi))) != NULL)
107
108 /* The MM code likes to work with exclusive end addresses */
109 #define for_each_vma_range(__vmi, __vma, __end) \
110 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
111
112 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
113
114 #define PHYS_PFN(x) ((unsigned long)((x) >> PAGE_SHIFT))
115
116 #define test_and_set_bit(nr, addr) __test_and_set_bit(nr, addr)
117 #define test_and_clear_bit(nr, addr) __test_and_clear_bit(nr, addr)
118
119 #define TASK_SIZE ((1ul << 47)-PAGE_SIZE)
120
121 #define AS_MM_ALL_LOCKS 2
122
123 /* We hardcode this for now. */
124 #define sysctl_max_map_count 0x1000000UL
125
126 #define pgoff_t unsigned long
127 typedef unsigned long pgprotval_t;
128 typedef struct pgprot { pgprotval_t pgprot; } pgprot_t;
129 typedef unsigned long vm_flags_t;
130 typedef __bitwise unsigned int vm_fault_t;
131
132 /*
133 * The shared stubs do not implement this, it amounts to an fprintf(STDERR,...)
134 * either way :)
135 */
136 #define pr_warn_once pr_err
137
138 typedef struct refcount_struct {
139 atomic_t refs;
140 } refcount_t;
141
142 struct kref {
143 refcount_t refcount;
144 };
145
146 /*
147 * Define the task command name length as enum, then it can be visible to
148 * BPF programs.
149 */
150 enum {
151 TASK_COMM_LEN = 16,
152 };
153
154 /*
155 * Flags for bug emulation.
156 *
157 * These occupy the top three bytes.
158 */
159 enum {
160 READ_IMPLIES_EXEC = 0x0400000,
161 };
162
163 struct task_struct {
164 char comm[TASK_COMM_LEN];
165 pid_t pid;
166 struct mm_struct *mm;
167
168 /* Used for emulating ABI behavior of previous Linux versions: */
169 unsigned int personality;
170 };
171
172 struct task_struct *get_current(void);
173 #define current get_current()
174
175 struct anon_vma {
176 struct anon_vma *root;
177 struct rb_root_cached rb_root;
178
179 /* Test fields. */
180 bool was_cloned;
181 bool was_unlinked;
182 };
183
184 struct anon_vma_chain {
185 struct anon_vma *anon_vma;
186 struct list_head same_vma;
187 };
188
189 struct anon_vma_name {
190 struct kref kref;
191 /* The name needs to be at the end because it is dynamically sized. */
192 char name[];
193 };
194
195 struct vma_iterator {
196 struct ma_state mas;
197 };
198
199 #define VMA_ITERATOR(name, __mm, __addr) \
200 struct vma_iterator name = { \
201 .mas = { \
202 .tree = &(__mm)->mm_mt, \
203 .index = __addr, \
204 .node = NULL, \
205 .status = ma_start, \
206 }, \
207 }
208
209 struct address_space {
210 struct rb_root_cached i_mmap;
211 unsigned long flags;
212 atomic_t i_mmap_writable;
213 };
214
215 struct vm_userfaultfd_ctx {};
216 struct mempolicy {};
217 struct mmu_gather {};
218 struct mutex {};
219 #define DEFINE_MUTEX(mutexname) \
220 struct mutex mutexname = {}
221
222 struct mm_struct {
223 struct maple_tree mm_mt;
224 int map_count; /* number of VMAs */
225 unsigned long total_vm; /* Total pages mapped */
226 unsigned long locked_vm; /* Pages that have PG_mlocked set */
227 unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
228 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
229 unsigned long stack_vm; /* VM_STACK */
230
231 unsigned long def_flags;
232
233 unsigned long flags; /* Must use atomic bitops to access */
234 };
235
236 struct vma_lock {
237 struct rw_semaphore lock;
238 };
239
240
241 struct file {
242 struct address_space *f_mapping;
243 };
244
245 struct vm_area_struct {
246 /* The first cache line has the info for VMA tree walking. */
247
248 union {
249 struct {
250 /* VMA covers [vm_start; vm_end) addresses within mm */
251 unsigned long vm_start;
252 unsigned long vm_end;
253 };
254 #ifdef CONFIG_PER_VMA_LOCK
255 struct rcu_head vm_rcu; /* Used for deferred freeing. */
256 #endif
257 };
258
259 struct mm_struct *vm_mm; /* The address space we belong to. */
260 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
261
262 /*
263 * Flags, see mm.h.
264 * To modify use vm_flags_{init|reset|set|clear|mod} functions.
265 */
266 union {
267 const vm_flags_t vm_flags;
268 vm_flags_t __private __vm_flags;
269 };
270
271 #ifdef CONFIG_PER_VMA_LOCK
272 /* Flag to indicate areas detached from the mm->mm_mt tree */
273 bool detached;
274
275 /*
276 * Can only be written (using WRITE_ONCE()) while holding both:
277 * - mmap_lock (in write mode)
278 * - vm_lock->lock (in write mode)
279 * Can be read reliably while holding one of:
280 * - mmap_lock (in read or write mode)
281 * - vm_lock->lock (in read or write mode)
282 * Can be read unreliably (using READ_ONCE()) for pessimistic bailout
283 * while holding nothing (except RCU to keep the VMA struct allocated).
284 *
285 * This sequence counter is explicitly allowed to overflow; sequence
286 * counter reuse can only lead to occasional unnecessary use of the
287 * slowpath.
288 */
289 unsigned int vm_lock_seq;
290 struct vma_lock *vm_lock;
291 #endif
292
293 /*
294 * For areas with an address space and backing store,
295 * linkage into the address_space->i_mmap interval tree.
296 *
297 */
298 struct {
299 struct rb_node rb;
300 unsigned long rb_subtree_last;
301 } shared;
302
303 /*
304 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
305 * list, after a COW of one of the file pages. A MAP_SHARED vma
306 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
307 * or brk vma (with NULL file) can only be in an anon_vma list.
308 */
309 struct list_head anon_vma_chain; /* Serialized by mmap_lock &
310 * page_table_lock */
311 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
312
313 /* Function pointers to deal with this struct. */
314 const struct vm_operations_struct *vm_ops;
315
316 /* Information about our backing store: */
317 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
318 units */
319 struct file * vm_file; /* File we map to (can be NULL). */
320 void * vm_private_data; /* was vm_pte (shared mem) */
321
322 #ifdef CONFIG_ANON_VMA_NAME
323 /*
324 * For private and shared anonymous mappings, a pointer to a null
325 * terminated string containing the name given to the vma, or NULL if
326 * unnamed. Serialized by mmap_lock. Use anon_vma_name to access.
327 */
328 struct anon_vma_name *anon_name;
329 #endif
330 #ifdef CONFIG_SWAP
331 atomic_long_t swap_readahead_info;
332 #endif
333 #ifndef CONFIG_MMU
334 struct vm_region *vm_region; /* NOMMU mapping region */
335 #endif
336 #ifdef CONFIG_NUMA
337 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
338 #endif
339 #ifdef CONFIG_NUMA_BALANCING
340 struct vma_numab_state *numab_state; /* NUMA Balancing state */
341 #endif
342 struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
343 } __randomize_layout;
344
345 struct vm_fault {};
346
347 struct vm_operations_struct {
348 void (*open)(struct vm_area_struct * area);
349 /**
350 * @close: Called when the VMA is being removed from the MM.
351 * Context: User context. May sleep. Caller holds mmap_lock.
352 */
353 void (*close)(struct vm_area_struct * area);
354 /* Called any time before splitting to check if it's allowed */
355 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
356 int (*mremap)(struct vm_area_struct *area);
357 /*
358 * Called by mprotect() to make driver-specific permission
359 * checks before mprotect() is finalised. The VMA must not
360 * be modified. Returns 0 if mprotect() can proceed.
361 */
362 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
363 unsigned long end, unsigned long newflags);
364 vm_fault_t (*fault)(struct vm_fault *vmf);
365 vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
366 vm_fault_t (*map_pages)(struct vm_fault *vmf,
367 pgoff_t start_pgoff, pgoff_t end_pgoff);
368 unsigned long (*pagesize)(struct vm_area_struct * area);
369
370 /* notification that a previously read-only page is about to become
371 * writable, if an error is returned it will cause a SIGBUS */
372 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
373
374 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
375 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
376
377 /* called by access_process_vm when get_user_pages() fails, typically
378 * for use by special VMAs. See also generic_access_phys() for a generic
379 * implementation useful for any iomem mapping.
380 */
381 int (*access)(struct vm_area_struct *vma, unsigned long addr,
382 void *buf, int len, int write);
383
384 /* Called by the /proc/PID/maps code to ask the vma whether it
385 * has a special name. Returning non-NULL will also cause this
386 * vma to be dumped unconditionally. */
387 const char *(*name)(struct vm_area_struct *vma);
388
389 #ifdef CONFIG_NUMA
390 /*
391 * set_policy() op must add a reference to any non-NULL @new mempolicy
392 * to hold the policy upon return. Caller should pass NULL @new to
393 * remove a policy and fall back to surrounding context--i.e. do not
394 * install a MPOL_DEFAULT policy, nor the task or system default
395 * mempolicy.
396 */
397 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
398
399 /*
400 * get_policy() op must add reference [mpol_get()] to any policy at
401 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
402 * in mm/mempolicy.c will do this automatically.
403 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
404 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
405 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
406 * must return NULL--i.e., do not "fallback" to task or system default
407 * policy.
408 */
409 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
410 unsigned long addr, pgoff_t *ilx);
411 #endif
412 /*
413 * Called by vm_normal_page() for special PTEs to find the
414 * page for @addr. This is useful if the default behavior
415 * (using pte_page()) would not find the correct page.
416 */
417 struct page *(*find_special_page)(struct vm_area_struct *vma,
418 unsigned long addr);
419 };
420
421 struct vm_unmapped_area_info {
422 #define VM_UNMAPPED_AREA_TOPDOWN 1
423 unsigned long flags;
424 unsigned long length;
425 unsigned long low_limit;
426 unsigned long high_limit;
427 unsigned long align_mask;
428 unsigned long align_offset;
429 unsigned long start_gap;
430 };
431
vma_iter_invalidate(struct vma_iterator * vmi)432 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
433 {
434 mas_pause(&vmi->mas);
435 }
436
pgprot_modify(pgprot_t oldprot,pgprot_t newprot)437 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
438 {
439 return __pgprot(pgprot_val(oldprot) | pgprot_val(newprot));
440 }
441
vm_get_page_prot(unsigned long vm_flags)442 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
443 {
444 return __pgprot(vm_flags);
445 }
446
is_shared_maywrite(vm_flags_t vm_flags)447 static inline bool is_shared_maywrite(vm_flags_t vm_flags)
448 {
449 return (vm_flags & (VM_SHARED | VM_MAYWRITE)) ==
450 (VM_SHARED | VM_MAYWRITE);
451 }
452
vma_is_shared_maywrite(struct vm_area_struct * vma)453 static inline bool vma_is_shared_maywrite(struct vm_area_struct *vma)
454 {
455 return is_shared_maywrite(vma->vm_flags);
456 }
457
vma_next(struct vma_iterator * vmi)458 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
459 {
460 /*
461 * Uses mas_find() to get the first VMA when the iterator starts.
462 * Calling mas_next() could skip the first entry.
463 */
464 return mas_find(&vmi->mas, ULONG_MAX);
465 }
466
vma_lock_alloc(struct vm_area_struct * vma)467 static inline bool vma_lock_alloc(struct vm_area_struct *vma)
468 {
469 vma->vm_lock = calloc(1, sizeof(struct vma_lock));
470
471 if (!vma->vm_lock)
472 return false;
473
474 init_rwsem(&vma->vm_lock->lock);
475 vma->vm_lock_seq = UINT_MAX;
476
477 return true;
478 }
479
480 static inline void vma_assert_write_locked(struct vm_area_struct *);
vma_mark_detached(struct vm_area_struct * vma,bool detached)481 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
482 {
483 /* When detaching vma should be write-locked */
484 if (detached)
485 vma_assert_write_locked(vma);
486 vma->detached = detached;
487 }
488
489 extern const struct vm_operations_struct vma_dummy_vm_ops;
490
491 extern unsigned long rlimit(unsigned int limit);
492
vma_init(struct vm_area_struct * vma,struct mm_struct * mm)493 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
494 {
495 memset(vma, 0, sizeof(*vma));
496 vma->vm_mm = mm;
497 vma->vm_ops = &vma_dummy_vm_ops;
498 INIT_LIST_HEAD(&vma->anon_vma_chain);
499 vma_mark_detached(vma, false);
500 }
501
vm_area_alloc(struct mm_struct * mm)502 static inline struct vm_area_struct *vm_area_alloc(struct mm_struct *mm)
503 {
504 struct vm_area_struct *vma = calloc(1, sizeof(struct vm_area_struct));
505
506 if (!vma)
507 return NULL;
508
509 vma_init(vma, mm);
510 if (!vma_lock_alloc(vma)) {
511 free(vma);
512 return NULL;
513 }
514
515 return vma;
516 }
517
vm_area_dup(struct vm_area_struct * orig)518 static inline struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
519 {
520 struct vm_area_struct *new = calloc(1, sizeof(struct vm_area_struct));
521
522 if (!new)
523 return NULL;
524
525 memcpy(new, orig, sizeof(*new));
526 if (!vma_lock_alloc(new)) {
527 free(new);
528 return NULL;
529 }
530 INIT_LIST_HEAD(&new->anon_vma_chain);
531
532 return new;
533 }
534
535 /*
536 * These are defined in vma.h, but sadly vm_stat_account() is referenced by
537 * kernel/fork.c, so we have to these broadly available there, and temporarily
538 * define them here to resolve the dependency cycle.
539 */
540
541 #define is_exec_mapping(flags) \
542 ((flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC)
543
544 #define is_stack_mapping(flags) \
545 (((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK))
546
547 #define is_data_mapping(flags) \
548 ((flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE)
549
vm_stat_account(struct mm_struct * mm,vm_flags_t flags,long npages)550 static inline void vm_stat_account(struct mm_struct *mm, vm_flags_t flags,
551 long npages)
552 {
553 WRITE_ONCE(mm->total_vm, READ_ONCE(mm->total_vm)+npages);
554
555 if (is_exec_mapping(flags))
556 mm->exec_vm += npages;
557 else if (is_stack_mapping(flags))
558 mm->stack_vm += npages;
559 else if (is_data_mapping(flags))
560 mm->data_vm += npages;
561 }
562
563 #undef is_exec_mapping
564 #undef is_stack_mapping
565 #undef is_data_mapping
566
567 /* Currently stubbed but we may later wish to un-stub. */
568 static inline void vm_acct_memory(long pages);
vm_unacct_memory(long pages)569 static inline void vm_unacct_memory(long pages)
570 {
571 vm_acct_memory(-pages);
572 }
573
mapping_allow_writable(struct address_space * mapping)574 static inline void mapping_allow_writable(struct address_space *mapping)
575 {
576 atomic_inc(&mapping->i_mmap_writable);
577 }
578
vma_set_range(struct vm_area_struct * vma,unsigned long start,unsigned long end,pgoff_t pgoff)579 static inline void vma_set_range(struct vm_area_struct *vma,
580 unsigned long start, unsigned long end,
581 pgoff_t pgoff)
582 {
583 vma->vm_start = start;
584 vma->vm_end = end;
585 vma->vm_pgoff = pgoff;
586 }
587
588 static inline
vma_find(struct vma_iterator * vmi,unsigned long max)589 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
590 {
591 return mas_find(&vmi->mas, max - 1);
592 }
593
vma_iter_clear_gfp(struct vma_iterator * vmi,unsigned long start,unsigned long end,gfp_t gfp)594 static inline int vma_iter_clear_gfp(struct vma_iterator *vmi,
595 unsigned long start, unsigned long end, gfp_t gfp)
596 {
597 __mas_set_range(&vmi->mas, start, end - 1);
598 mas_store_gfp(&vmi->mas, NULL, gfp);
599 if (unlikely(mas_is_err(&vmi->mas)))
600 return -ENOMEM;
601
602 return 0;
603 }
604
605 static inline void mmap_assert_locked(struct mm_struct *);
find_vma_intersection(struct mm_struct * mm,unsigned long start_addr,unsigned long end_addr)606 static inline struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
607 unsigned long start_addr,
608 unsigned long end_addr)
609 {
610 unsigned long index = start_addr;
611
612 mmap_assert_locked(mm);
613 return mt_find(&mm->mm_mt, &index, end_addr - 1);
614 }
615
616 static inline
vma_lookup(struct mm_struct * mm,unsigned long addr)617 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
618 {
619 return mtree_load(&mm->mm_mt, addr);
620 }
621
vma_prev(struct vma_iterator * vmi)622 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
623 {
624 return mas_prev(&vmi->mas, 0);
625 }
626
vma_iter_set(struct vma_iterator * vmi,unsigned long addr)627 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
628 {
629 mas_set(&vmi->mas, addr);
630 }
631
vma_is_anonymous(struct vm_area_struct * vma)632 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
633 {
634 return !vma->vm_ops;
635 }
636
637 /* Defined in vma.h, so temporarily define here to avoid circular dependency. */
638 #define vma_iter_load(vmi) \
639 mas_walk(&(vmi)->mas)
640
641 static inline struct vm_area_struct *
find_vma_prev(struct mm_struct * mm,unsigned long addr,struct vm_area_struct ** pprev)642 find_vma_prev(struct mm_struct *mm, unsigned long addr,
643 struct vm_area_struct **pprev)
644 {
645 struct vm_area_struct *vma;
646 VMA_ITERATOR(vmi, mm, addr);
647
648 vma = vma_iter_load(&vmi);
649 *pprev = vma_prev(&vmi);
650 if (!vma)
651 vma = vma_next(&vmi);
652 return vma;
653 }
654
655 #undef vma_iter_load
656
vma_iter_init(struct vma_iterator * vmi,struct mm_struct * mm,unsigned long addr)657 static inline void vma_iter_init(struct vma_iterator *vmi,
658 struct mm_struct *mm, unsigned long addr)
659 {
660 mas_init(&vmi->mas, &mm->mm_mt, addr);
661 }
662
663 /* Stubbed functions. */
664
anon_vma_name(struct vm_area_struct * vma)665 static inline struct anon_vma_name *anon_vma_name(struct vm_area_struct *vma)
666 {
667 return NULL;
668 }
669
is_mergeable_vm_userfaultfd_ctx(struct vm_area_struct * vma,struct vm_userfaultfd_ctx vm_ctx)670 static inline bool is_mergeable_vm_userfaultfd_ctx(struct vm_area_struct *vma,
671 struct vm_userfaultfd_ctx vm_ctx)
672 {
673 return true;
674 }
675
anon_vma_name_eq(struct anon_vma_name * anon_name1,struct anon_vma_name * anon_name2)676 static inline bool anon_vma_name_eq(struct anon_vma_name *anon_name1,
677 struct anon_vma_name *anon_name2)
678 {
679 return true;
680 }
681
might_sleep(void)682 static inline void might_sleep(void)
683 {
684 }
685
vma_pages(struct vm_area_struct * vma)686 static inline unsigned long vma_pages(struct vm_area_struct *vma)
687 {
688 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
689 }
690
fput(struct file *)691 static inline void fput(struct file *)
692 {
693 }
694
mpol_put(struct mempolicy *)695 static inline void mpol_put(struct mempolicy *)
696 {
697 }
698
vma_lock_free(struct vm_area_struct * vma)699 static inline void vma_lock_free(struct vm_area_struct *vma)
700 {
701 free(vma->vm_lock);
702 }
703
__vm_area_free(struct vm_area_struct * vma)704 static inline void __vm_area_free(struct vm_area_struct *vma)
705 {
706 vma_lock_free(vma);
707 free(vma);
708 }
709
vm_area_free(struct vm_area_struct * vma)710 static inline void vm_area_free(struct vm_area_struct *vma)
711 {
712 __vm_area_free(vma);
713 }
714
lru_add_drain(void)715 static inline void lru_add_drain(void)
716 {
717 }
718
tlb_gather_mmu(struct mmu_gather *,struct mm_struct *)719 static inline void tlb_gather_mmu(struct mmu_gather *, struct mm_struct *)
720 {
721 }
722
update_hiwater_rss(struct mm_struct *)723 static inline void update_hiwater_rss(struct mm_struct *)
724 {
725 }
726
update_hiwater_vm(struct mm_struct *)727 static inline void update_hiwater_vm(struct mm_struct *)
728 {
729 }
730
unmap_vmas(struct mmu_gather * tlb,struct ma_state * mas,struct vm_area_struct * vma,unsigned long start_addr,unsigned long end_addr,unsigned long tree_end,bool mm_wr_locked)731 static inline void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
732 struct vm_area_struct *vma, unsigned long start_addr,
733 unsigned long end_addr, unsigned long tree_end,
734 bool mm_wr_locked)
735 {
736 (void)tlb;
737 (void)mas;
738 (void)vma;
739 (void)start_addr;
740 (void)end_addr;
741 (void)tree_end;
742 (void)mm_wr_locked;
743 }
744
free_pgtables(struct mmu_gather * tlb,struct ma_state * mas,struct vm_area_struct * vma,unsigned long floor,unsigned long ceiling,bool mm_wr_locked)745 static inline void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
746 struct vm_area_struct *vma, unsigned long floor,
747 unsigned long ceiling, bool mm_wr_locked)
748 {
749 (void)tlb;
750 (void)mas;
751 (void)vma;
752 (void)floor;
753 (void)ceiling;
754 (void)mm_wr_locked;
755 }
756
mapping_unmap_writable(struct address_space *)757 static inline void mapping_unmap_writable(struct address_space *)
758 {
759 }
760
flush_dcache_mmap_lock(struct address_space *)761 static inline void flush_dcache_mmap_lock(struct address_space *)
762 {
763 }
764
tlb_finish_mmu(struct mmu_gather *)765 static inline void tlb_finish_mmu(struct mmu_gather *)
766 {
767 }
768
get_file(struct file * f)769 static inline struct file *get_file(struct file *f)
770 {
771 return f;
772 }
773
vma_dup_policy(struct vm_area_struct *,struct vm_area_struct *)774 static inline int vma_dup_policy(struct vm_area_struct *, struct vm_area_struct *)
775 {
776 return 0;
777 }
778
anon_vma_clone(struct vm_area_struct * dst,struct vm_area_struct * src)779 static inline int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
780 {
781 /* For testing purposes. We indicate that an anon_vma has been cloned. */
782 if (src->anon_vma != NULL) {
783 dst->anon_vma = src->anon_vma;
784 dst->anon_vma->was_cloned = true;
785 }
786
787 return 0;
788 }
789
vma_start_write(struct vm_area_struct * vma)790 static inline void vma_start_write(struct vm_area_struct *vma)
791 {
792 /* Used to indicate to tests that a write operation has begun. */
793 vma->vm_lock_seq++;
794 }
795
vma_adjust_trans_huge(struct vm_area_struct * vma,unsigned long start,unsigned long end,long adjust_next)796 static inline void vma_adjust_trans_huge(struct vm_area_struct *vma,
797 unsigned long start,
798 unsigned long end,
799 long adjust_next)
800 {
801 (void)vma;
802 (void)start;
803 (void)end;
804 (void)adjust_next;
805 }
806
vma_iter_free(struct vma_iterator * vmi)807 static inline void vma_iter_free(struct vma_iterator *vmi)
808 {
809 mas_destroy(&vmi->mas);
810 }
811
812 static inline
vma_iter_next_range(struct vma_iterator * vmi)813 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
814 {
815 return mas_next_range(&vmi->mas, ULONG_MAX);
816 }
817
vm_acct_memory(long pages)818 static inline void vm_acct_memory(long pages)
819 {
820 }
821
vma_interval_tree_insert(struct vm_area_struct *,struct rb_root_cached *)822 static inline void vma_interval_tree_insert(struct vm_area_struct *,
823 struct rb_root_cached *)
824 {
825 }
826
vma_interval_tree_remove(struct vm_area_struct *,struct rb_root_cached *)827 static inline void vma_interval_tree_remove(struct vm_area_struct *,
828 struct rb_root_cached *)
829 {
830 }
831
flush_dcache_mmap_unlock(struct address_space *)832 static inline void flush_dcache_mmap_unlock(struct address_space *)
833 {
834 }
835
anon_vma_interval_tree_insert(struct anon_vma_chain *,struct rb_root_cached *)836 static inline void anon_vma_interval_tree_insert(struct anon_vma_chain*,
837 struct rb_root_cached *)
838 {
839 }
840
anon_vma_interval_tree_remove(struct anon_vma_chain *,struct rb_root_cached *)841 static inline void anon_vma_interval_tree_remove(struct anon_vma_chain*,
842 struct rb_root_cached *)
843 {
844 }
845
uprobe_mmap(struct vm_area_struct *)846 static inline void uprobe_mmap(struct vm_area_struct *)
847 {
848 }
849
uprobe_munmap(struct vm_area_struct * vma,unsigned long start,unsigned long end)850 static inline void uprobe_munmap(struct vm_area_struct *vma,
851 unsigned long start, unsigned long end)
852 {
853 (void)vma;
854 (void)start;
855 (void)end;
856 }
857
i_mmap_lock_write(struct address_space *)858 static inline void i_mmap_lock_write(struct address_space *)
859 {
860 }
861
anon_vma_lock_write(struct anon_vma *)862 static inline void anon_vma_lock_write(struct anon_vma *)
863 {
864 }
865
vma_assert_write_locked(struct vm_area_struct *)866 static inline void vma_assert_write_locked(struct vm_area_struct *)
867 {
868 }
869
unlink_anon_vmas(struct vm_area_struct * vma)870 static inline void unlink_anon_vmas(struct vm_area_struct *vma)
871 {
872 /* For testing purposes, indicate that the anon_vma was unlinked. */
873 vma->anon_vma->was_unlinked = true;
874 }
875
anon_vma_unlock_write(struct anon_vma *)876 static inline void anon_vma_unlock_write(struct anon_vma *)
877 {
878 }
879
i_mmap_unlock_write(struct address_space *)880 static inline void i_mmap_unlock_write(struct address_space *)
881 {
882 }
883
anon_vma_merge(struct vm_area_struct *,struct vm_area_struct *)884 static inline void anon_vma_merge(struct vm_area_struct *,
885 struct vm_area_struct *)
886 {
887 }
888
userfaultfd_unmap_prep(struct vm_area_struct * vma,unsigned long start,unsigned long end,struct list_head * unmaps)889 static inline int userfaultfd_unmap_prep(struct vm_area_struct *vma,
890 unsigned long start,
891 unsigned long end,
892 struct list_head *unmaps)
893 {
894 (void)vma;
895 (void)start;
896 (void)end;
897 (void)unmaps;
898
899 return 0;
900 }
901
mmap_write_downgrade(struct mm_struct *)902 static inline void mmap_write_downgrade(struct mm_struct *)
903 {
904 }
905
mmap_read_unlock(struct mm_struct *)906 static inline void mmap_read_unlock(struct mm_struct *)
907 {
908 }
909
mmap_write_unlock(struct mm_struct *)910 static inline void mmap_write_unlock(struct mm_struct *)
911 {
912 }
913
mmap_write_lock_killable(struct mm_struct *)914 static inline int mmap_write_lock_killable(struct mm_struct *)
915 {
916 return 0;
917 }
918
can_modify_mm(struct mm_struct * mm,unsigned long start,unsigned long end)919 static inline bool can_modify_mm(struct mm_struct *mm,
920 unsigned long start,
921 unsigned long end)
922 {
923 (void)mm;
924 (void)start;
925 (void)end;
926
927 return true;
928 }
929
arch_unmap(struct mm_struct * mm,unsigned long start,unsigned long end)930 static inline void arch_unmap(struct mm_struct *mm,
931 unsigned long start,
932 unsigned long end)
933 {
934 (void)mm;
935 (void)start;
936 (void)end;
937 }
938
mmap_assert_locked(struct mm_struct *)939 static inline void mmap_assert_locked(struct mm_struct *)
940 {
941 }
942
mpol_equal(struct mempolicy *,struct mempolicy *)943 static inline bool mpol_equal(struct mempolicy *, struct mempolicy *)
944 {
945 return true;
946 }
947
khugepaged_enter_vma(struct vm_area_struct * vma,unsigned long vm_flags)948 static inline void khugepaged_enter_vma(struct vm_area_struct *vma,
949 unsigned long vm_flags)
950 {
951 (void)vma;
952 (void)vm_flags;
953 }
954
mapping_can_writeback(struct address_space *)955 static inline bool mapping_can_writeback(struct address_space *)
956 {
957 return true;
958 }
959
is_vm_hugetlb_page(struct vm_area_struct *)960 static inline bool is_vm_hugetlb_page(struct vm_area_struct *)
961 {
962 return false;
963 }
964
vma_soft_dirty_enabled(struct vm_area_struct *)965 static inline bool vma_soft_dirty_enabled(struct vm_area_struct *)
966 {
967 return false;
968 }
969
userfaultfd_wp(struct vm_area_struct *)970 static inline bool userfaultfd_wp(struct vm_area_struct *)
971 {
972 return false;
973 }
974
mmap_assert_write_locked(struct mm_struct *)975 static inline void mmap_assert_write_locked(struct mm_struct *)
976 {
977 }
978
mutex_lock(struct mutex *)979 static inline void mutex_lock(struct mutex *)
980 {
981 }
982
mutex_unlock(struct mutex *)983 static inline void mutex_unlock(struct mutex *)
984 {
985 }
986
mutex_is_locked(struct mutex *)987 static inline bool mutex_is_locked(struct mutex *)
988 {
989 return true;
990 }
991
signal_pending(void *)992 static inline bool signal_pending(void *)
993 {
994 return false;
995 }
996
is_file_hugepages(struct file *)997 static inline bool is_file_hugepages(struct file *)
998 {
999 return false;
1000 }
1001
security_vm_enough_memory_mm(struct mm_struct *,long)1002 static inline int security_vm_enough_memory_mm(struct mm_struct *, long)
1003 {
1004 return 0;
1005 }
1006
may_expand_vm(struct mm_struct *,vm_flags_t,unsigned long)1007 static inline bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long)
1008 {
1009 return true;
1010 }
1011
vm_flags_init(struct vm_area_struct * vma,vm_flags_t flags)1012 static inline void vm_flags_init(struct vm_area_struct *vma,
1013 vm_flags_t flags)
1014 {
1015 vma->__vm_flags = flags;
1016 }
1017
vm_flags_set(struct vm_area_struct * vma,vm_flags_t flags)1018 static inline void vm_flags_set(struct vm_area_struct *vma,
1019 vm_flags_t flags)
1020 {
1021 vma_start_write(vma);
1022 vma->__vm_flags |= flags;
1023 }
1024
vm_flags_clear(struct vm_area_struct * vma,vm_flags_t flags)1025 static inline void vm_flags_clear(struct vm_area_struct *vma,
1026 vm_flags_t flags)
1027 {
1028 vma_start_write(vma);
1029 vma->__vm_flags &= ~flags;
1030 }
1031
call_mmap(struct file *,struct vm_area_struct *)1032 static inline int call_mmap(struct file *, struct vm_area_struct *)
1033 {
1034 return 0;
1035 }
1036
shmem_zero_setup(struct vm_area_struct *)1037 static inline int shmem_zero_setup(struct vm_area_struct *)
1038 {
1039 return 0;
1040 }
1041
vma_set_anonymous(struct vm_area_struct * vma)1042 static inline void vma_set_anonymous(struct vm_area_struct *vma)
1043 {
1044 vma->vm_ops = NULL;
1045 }
1046
ksm_add_vma(struct vm_area_struct *)1047 static inline void ksm_add_vma(struct vm_area_struct *)
1048 {
1049 }
1050
perf_event_mmap(struct vm_area_struct *)1051 static inline void perf_event_mmap(struct vm_area_struct *)
1052 {
1053 }
1054
vma_is_dax(struct vm_area_struct *)1055 static inline bool vma_is_dax(struct vm_area_struct *)
1056 {
1057 return false;
1058 }
1059
get_gate_vma(struct mm_struct *)1060 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *)
1061 {
1062 return NULL;
1063 }
1064
1065 bool vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1066
1067 /* Update vma->vm_page_prot to reflect vma->vm_flags. */
vma_set_page_prot(struct vm_area_struct * vma)1068 static inline void vma_set_page_prot(struct vm_area_struct *vma)
1069 {
1070 unsigned long vm_flags = vma->vm_flags;
1071 pgprot_t vm_page_prot;
1072
1073 /* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1074 vm_page_prot = pgprot_modify(vma->vm_page_prot, vm_get_page_prot(vm_flags));
1075
1076 if (vma_wants_writenotify(vma, vm_page_prot)) {
1077 vm_flags &= ~VM_SHARED;
1078 /* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1079 vm_page_prot = pgprot_modify(vm_page_prot, vm_get_page_prot(vm_flags));
1080 }
1081 /* remove_protection_ptes reads vma->vm_page_prot without mmap_lock */
1082 WRITE_ONCE(vma->vm_page_prot, vm_page_prot);
1083 }
1084
arch_validate_flags(unsigned long)1085 static inline bool arch_validate_flags(unsigned long)
1086 {
1087 return true;
1088 }
1089
vma_close(struct vm_area_struct *)1090 static inline void vma_close(struct vm_area_struct *)
1091 {
1092 }
1093
mmap_file(struct file *,struct vm_area_struct *)1094 static inline int mmap_file(struct file *, struct vm_area_struct *)
1095 {
1096 return 0;
1097 }
1098
stack_guard_start_gap(struct vm_area_struct * vma)1099 static inline unsigned long stack_guard_start_gap(struct vm_area_struct *vma)
1100 {
1101 if (vma->vm_flags & VM_GROWSDOWN)
1102 return stack_guard_gap;
1103
1104 /* See reasoning around the VM_SHADOW_STACK definition */
1105 if (vma->vm_flags & VM_SHADOW_STACK)
1106 return PAGE_SIZE;
1107
1108 return 0;
1109 }
1110
vm_start_gap(struct vm_area_struct * vma)1111 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
1112 {
1113 unsigned long gap = stack_guard_start_gap(vma);
1114 unsigned long vm_start = vma->vm_start;
1115
1116 vm_start -= gap;
1117 if (vm_start > vma->vm_start)
1118 vm_start = 0;
1119 return vm_start;
1120 }
1121
vm_end_gap(struct vm_area_struct * vma)1122 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
1123 {
1124 unsigned long vm_end = vma->vm_end;
1125
1126 if (vma->vm_flags & VM_GROWSUP) {
1127 vm_end += stack_guard_gap;
1128 if (vm_end < vma->vm_end)
1129 vm_end = -PAGE_SIZE;
1130 }
1131 return vm_end;
1132 }
1133
is_hugepage_only_range(struct mm_struct * mm,unsigned long addr,unsigned long len)1134 static inline int is_hugepage_only_range(struct mm_struct *mm,
1135 unsigned long addr, unsigned long len)
1136 {
1137 return 0;
1138 }
1139
vma_is_accessible(struct vm_area_struct * vma)1140 static inline bool vma_is_accessible(struct vm_area_struct *vma)
1141 {
1142 return vma->vm_flags & VM_ACCESS_FLAGS;
1143 }
1144
capable(int cap)1145 static inline bool capable(int cap)
1146 {
1147 return true;
1148 }
1149
mlock_future_ok(struct mm_struct * mm,unsigned long flags,unsigned long bytes)1150 static inline bool mlock_future_ok(struct mm_struct *mm, unsigned long flags,
1151 unsigned long bytes)
1152 {
1153 unsigned long locked_pages, limit_pages;
1154
1155 if (!(flags & VM_LOCKED) || capable(CAP_IPC_LOCK))
1156 return true;
1157
1158 locked_pages = bytes >> PAGE_SHIFT;
1159 locked_pages += mm->locked_vm;
1160
1161 limit_pages = rlimit(RLIMIT_MEMLOCK);
1162 limit_pages >>= PAGE_SHIFT;
1163
1164 return locked_pages <= limit_pages;
1165 }
1166
__anon_vma_prepare(struct vm_area_struct * vma)1167 static inline int __anon_vma_prepare(struct vm_area_struct *vma)
1168 {
1169 struct anon_vma *anon_vma = calloc(1, sizeof(struct anon_vma));
1170
1171 if (!anon_vma)
1172 return -ENOMEM;
1173
1174 anon_vma->root = anon_vma;
1175 vma->anon_vma = anon_vma;
1176
1177 return 0;
1178 }
1179
anon_vma_prepare(struct vm_area_struct * vma)1180 static inline int anon_vma_prepare(struct vm_area_struct *vma)
1181 {
1182 if (likely(vma->anon_vma))
1183 return 0;
1184
1185 return __anon_vma_prepare(vma);
1186 }
1187
userfaultfd_unmap_complete(struct mm_struct * mm,struct list_head * uf)1188 static inline void userfaultfd_unmap_complete(struct mm_struct *mm,
1189 struct list_head *uf)
1190 {
1191 }
1192
1193 /*
1194 * Denies creating a writable executable mapping or gaining executable permissions.
1195 *
1196 * This denies the following:
1197 *
1198 * a) mmap(PROT_WRITE | PROT_EXEC)
1199 *
1200 * b) mmap(PROT_WRITE)
1201 * mprotect(PROT_EXEC)
1202 *
1203 * c) mmap(PROT_WRITE)
1204 * mprotect(PROT_READ)
1205 * mprotect(PROT_EXEC)
1206 *
1207 * But allows the following:
1208 *
1209 * d) mmap(PROT_READ | PROT_EXEC)
1210 * mmap(PROT_READ | PROT_EXEC | PROT_BTI)
1211 *
1212 * This is only applicable if the user has set the Memory-Deny-Write-Execute
1213 * (MDWE) protection mask for the current process.
1214 *
1215 * @old specifies the VMA flags the VMA originally possessed, and @new the ones
1216 * we propose to set.
1217 *
1218 * Return: false if proposed change is OK, true if not ok and should be denied.
1219 */
map_deny_write_exec(unsigned long old,unsigned long new)1220 static inline bool map_deny_write_exec(unsigned long old, unsigned long new)
1221 {
1222 /* If MDWE is disabled, we have nothing to deny. */
1223 if (!test_bit(MMF_HAS_MDWE, ¤t->mm->flags))
1224 return false;
1225
1226 /* If the new VMA is not executable, we have nothing to deny. */
1227 if (!(new & VM_EXEC))
1228 return false;
1229
1230 /* Under MDWE we do not accept newly writably executable VMAs... */
1231 if (new & VM_WRITE)
1232 return true;
1233
1234 /* ...nor previously non-executable VMAs becoming executable. */
1235 if (!(old & VM_EXEC))
1236 return true;
1237
1238 return false;
1239 }
1240
mapping_map_writable(struct address_space * mapping)1241 static inline int mapping_map_writable(struct address_space *mapping)
1242 {
1243 int c = atomic_read(&mapping->i_mmap_writable);
1244
1245 /* Derived from the raw_atomic_inc_unless_negative() implementation. */
1246 do {
1247 if (c < 0)
1248 return -EPERM;
1249 } while (!__sync_bool_compare_and_swap(&mapping->i_mmap_writable, c, c+1));
1250
1251 return 0;
1252 }
1253
1254 #endif /* __MM_VMA_INTERNAL_H */
1255