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/refcount.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 struct kref {
139 refcount_t refcount;
140 };
141
142 /*
143 * Define the task command name length as enum, then it can be visible to
144 * BPF programs.
145 */
146 enum {
147 TASK_COMM_LEN = 16,
148 };
149
150 /*
151 * Flags for bug emulation.
152 *
153 * These occupy the top three bytes.
154 */
155 enum {
156 READ_IMPLIES_EXEC = 0x0400000,
157 };
158
159 struct task_struct {
160 char comm[TASK_COMM_LEN];
161 pid_t pid;
162 struct mm_struct *mm;
163
164 /* Used for emulating ABI behavior of previous Linux versions: */
165 unsigned int personality;
166 };
167
168 struct task_struct *get_current(void);
169 #define current get_current()
170
171 struct anon_vma {
172 struct anon_vma *root;
173 struct rb_root_cached rb_root;
174
175 /* Test fields. */
176 bool was_cloned;
177 bool was_unlinked;
178 };
179
180 struct anon_vma_chain {
181 struct anon_vma *anon_vma;
182 struct list_head same_vma;
183 };
184
185 struct anon_vma_name {
186 struct kref kref;
187 /* The name needs to be at the end because it is dynamically sized. */
188 char name[];
189 };
190
191 struct vma_iterator {
192 struct ma_state mas;
193 };
194
195 #define VMA_ITERATOR(name, __mm, __addr) \
196 struct vma_iterator name = { \
197 .mas = { \
198 .tree = &(__mm)->mm_mt, \
199 .index = __addr, \
200 .node = NULL, \
201 .status = ma_start, \
202 }, \
203 }
204
205 struct address_space {
206 struct rb_root_cached i_mmap;
207 unsigned long flags;
208 atomic_t i_mmap_writable;
209 };
210
211 struct vm_userfaultfd_ctx {};
212 struct mempolicy {};
213 struct mmu_gather {};
214 struct mutex {};
215 #define DEFINE_MUTEX(mutexname) \
216 struct mutex mutexname = {}
217
218 struct mm_struct {
219 struct maple_tree mm_mt;
220 int map_count; /* number of VMAs */
221 unsigned long total_vm; /* Total pages mapped */
222 unsigned long locked_vm; /* Pages that have PG_mlocked set */
223 unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
224 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
225 unsigned long stack_vm; /* VM_STACK */
226
227 unsigned long def_flags;
228
229 unsigned long flags; /* Must use atomic bitops to access */
230 };
231
232 struct file {
233 struct address_space *f_mapping;
234 };
235
236 #define VMA_LOCK_OFFSET 0x40000000
237
238 struct vm_area_struct {
239 /* The first cache line has the info for VMA tree walking. */
240
241 union {
242 struct {
243 /* VMA covers [vm_start; vm_end) addresses within mm */
244 unsigned long vm_start;
245 unsigned long vm_end;
246 };
247 #ifdef CONFIG_PER_VMA_LOCK
248 struct rcu_head vm_rcu; /* Used for deferred freeing. */
249 #endif
250 };
251
252 struct mm_struct *vm_mm; /* The address space we belong to. */
253 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
254
255 /*
256 * Flags, see mm.h.
257 * To modify use vm_flags_{init|reset|set|clear|mod} functions.
258 */
259 union {
260 const vm_flags_t vm_flags;
261 vm_flags_t __private __vm_flags;
262 };
263
264 #ifdef CONFIG_PER_VMA_LOCK
265 /*
266 * Can only be written (using WRITE_ONCE()) while holding both:
267 * - mmap_lock (in write mode)
268 * - vm_refcnt bit at VMA_LOCK_OFFSET is set
269 * Can be read reliably while holding one of:
270 * - mmap_lock (in read or write mode)
271 * - vm_refcnt bit at VMA_LOCK_OFFSET is set or vm_refcnt > 1
272 * Can be read unreliably (using READ_ONCE()) for pessimistic bailout
273 * while holding nothing (except RCU to keep the VMA struct allocated).
274 *
275 * This sequence counter is explicitly allowed to overflow; sequence
276 * counter reuse can only lead to occasional unnecessary use of the
277 * slowpath.
278 */
279 unsigned int vm_lock_seq;
280 #endif
281
282 /*
283 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
284 * list, after a COW of one of the file pages. A MAP_SHARED vma
285 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
286 * or brk vma (with NULL file) can only be in an anon_vma list.
287 */
288 struct list_head anon_vma_chain; /* Serialized by mmap_lock &
289 * page_table_lock */
290 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
291
292 /* Function pointers to deal with this struct. */
293 const struct vm_operations_struct *vm_ops;
294
295 /* Information about our backing store: */
296 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
297 units */
298 struct file * vm_file; /* File we map to (can be NULL). */
299 void * vm_private_data; /* was vm_pte (shared mem) */
300
301 #ifdef CONFIG_SWAP
302 atomic_long_t swap_readahead_info;
303 #endif
304 #ifndef CONFIG_MMU
305 struct vm_region *vm_region; /* NOMMU mapping region */
306 #endif
307 #ifdef CONFIG_NUMA
308 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
309 #endif
310 #ifdef CONFIG_NUMA_BALANCING
311 struct vma_numab_state *numab_state; /* NUMA Balancing state */
312 #endif
313 #ifdef CONFIG_PER_VMA_LOCK
314 /* Unstable RCU readers are allowed to read this. */
315 refcount_t vm_refcnt;
316 #endif
317 /*
318 * For areas with an address space and backing store,
319 * linkage into the address_space->i_mmap interval tree.
320 *
321 */
322 struct {
323 struct rb_node rb;
324 unsigned long rb_subtree_last;
325 } shared;
326 #ifdef CONFIG_ANON_VMA_NAME
327 /*
328 * For private and shared anonymous mappings, a pointer to a null
329 * terminated string containing the name given to the vma, or NULL if
330 * unnamed. Serialized by mmap_lock. Use anon_vma_name to access.
331 */
332 struct anon_vma_name *anon_name;
333 #endif
334 struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
335 } __randomize_layout;
336
337 struct vm_fault {};
338
339 struct vm_operations_struct {
340 void (*open)(struct vm_area_struct * area);
341 /**
342 * @close: Called when the VMA is being removed from the MM.
343 * Context: User context. May sleep. Caller holds mmap_lock.
344 */
345 void (*close)(struct vm_area_struct * area);
346 /* Called any time before splitting to check if it's allowed */
347 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
348 int (*mremap)(struct vm_area_struct *area);
349 /*
350 * Called by mprotect() to make driver-specific permission
351 * checks before mprotect() is finalised. The VMA must not
352 * be modified. Returns 0 if mprotect() can proceed.
353 */
354 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
355 unsigned long end, unsigned long newflags);
356 vm_fault_t (*fault)(struct vm_fault *vmf);
357 vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
358 vm_fault_t (*map_pages)(struct vm_fault *vmf,
359 pgoff_t start_pgoff, pgoff_t end_pgoff);
360 unsigned long (*pagesize)(struct vm_area_struct * area);
361
362 /* notification that a previously read-only page is about to become
363 * writable, if an error is returned it will cause a SIGBUS */
364 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
365
366 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
367 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
368
369 /* called by access_process_vm when get_user_pages() fails, typically
370 * for use by special VMAs. See also generic_access_phys() for a generic
371 * implementation useful for any iomem mapping.
372 */
373 int (*access)(struct vm_area_struct *vma, unsigned long addr,
374 void *buf, int len, int write);
375
376 /* Called by the /proc/PID/maps code to ask the vma whether it
377 * has a special name. Returning non-NULL will also cause this
378 * vma to be dumped unconditionally. */
379 const char *(*name)(struct vm_area_struct *vma);
380
381 #ifdef CONFIG_NUMA
382 /*
383 * set_policy() op must add a reference to any non-NULL @new mempolicy
384 * to hold the policy upon return. Caller should pass NULL @new to
385 * remove a policy and fall back to surrounding context--i.e. do not
386 * install a MPOL_DEFAULT policy, nor the task or system default
387 * mempolicy.
388 */
389 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
390
391 /*
392 * get_policy() op must add reference [mpol_get()] to any policy at
393 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
394 * in mm/mempolicy.c will do this automatically.
395 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
396 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
397 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
398 * must return NULL--i.e., do not "fallback" to task or system default
399 * policy.
400 */
401 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
402 unsigned long addr, pgoff_t *ilx);
403 #endif
404 /*
405 * Called by vm_normal_page() for special PTEs to find the
406 * page for @addr. This is useful if the default behavior
407 * (using pte_page()) would not find the correct page.
408 */
409 struct page *(*find_special_page)(struct vm_area_struct *vma,
410 unsigned long addr);
411 };
412
413 struct vm_unmapped_area_info {
414 #define VM_UNMAPPED_AREA_TOPDOWN 1
415 unsigned long flags;
416 unsigned long length;
417 unsigned long low_limit;
418 unsigned long high_limit;
419 unsigned long align_mask;
420 unsigned long align_offset;
421 unsigned long start_gap;
422 };
423
vma_iter_invalidate(struct vma_iterator * vmi)424 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
425 {
426 mas_pause(&vmi->mas);
427 }
428
pgprot_modify(pgprot_t oldprot,pgprot_t newprot)429 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
430 {
431 return __pgprot(pgprot_val(oldprot) | pgprot_val(newprot));
432 }
433
vm_get_page_prot(unsigned long vm_flags)434 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
435 {
436 return __pgprot(vm_flags);
437 }
438
is_shared_maywrite(vm_flags_t vm_flags)439 static inline bool is_shared_maywrite(vm_flags_t vm_flags)
440 {
441 return (vm_flags & (VM_SHARED | VM_MAYWRITE)) ==
442 (VM_SHARED | VM_MAYWRITE);
443 }
444
vma_is_shared_maywrite(struct vm_area_struct * vma)445 static inline bool vma_is_shared_maywrite(struct vm_area_struct *vma)
446 {
447 return is_shared_maywrite(vma->vm_flags);
448 }
449
vma_next(struct vma_iterator * vmi)450 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
451 {
452 /*
453 * Uses mas_find() to get the first VMA when the iterator starts.
454 * Calling mas_next() could skip the first entry.
455 */
456 return mas_find(&vmi->mas, ULONG_MAX);
457 }
458
459 /*
460 * WARNING: to avoid racing with vma_mark_attached()/vma_mark_detached(), these
461 * assertions should be made either under mmap_write_lock or when the object
462 * has been isolated under mmap_write_lock, ensuring no competing writers.
463 */
vma_assert_attached(struct vm_area_struct * vma)464 static inline void vma_assert_attached(struct vm_area_struct *vma)
465 {
466 WARN_ON_ONCE(!refcount_read(&vma->vm_refcnt));
467 }
468
vma_assert_detached(struct vm_area_struct * vma)469 static inline void vma_assert_detached(struct vm_area_struct *vma)
470 {
471 WARN_ON_ONCE(refcount_read(&vma->vm_refcnt));
472 }
473
474 static inline void vma_assert_write_locked(struct vm_area_struct *);
vma_mark_attached(struct vm_area_struct * vma)475 static inline void vma_mark_attached(struct vm_area_struct *vma)
476 {
477 vma_assert_write_locked(vma);
478 vma_assert_detached(vma);
479 refcount_set_release(&vma->vm_refcnt, 1);
480 }
481
vma_mark_detached(struct vm_area_struct * vma)482 static inline void vma_mark_detached(struct vm_area_struct *vma)
483 {
484 vma_assert_write_locked(vma);
485 vma_assert_attached(vma);
486 /* We are the only writer, so no need to use vma_refcount_put(). */
487 if (unlikely(!refcount_dec_and_test(&vma->vm_refcnt))) {
488 /*
489 * Reader must have temporarily raised vm_refcnt but it will
490 * drop it without using the vma since vma is write-locked.
491 */
492 }
493 }
494
495 extern const struct vm_operations_struct vma_dummy_vm_ops;
496
497 extern unsigned long rlimit(unsigned int limit);
498
vma_init(struct vm_area_struct * vma,struct mm_struct * mm)499 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
500 {
501 memset(vma, 0, sizeof(*vma));
502 vma->vm_mm = mm;
503 vma->vm_ops = &vma_dummy_vm_ops;
504 INIT_LIST_HEAD(&vma->anon_vma_chain);
505 vma->vm_lock_seq = UINT_MAX;
506 }
507
vm_area_alloc(struct mm_struct * mm)508 static inline struct vm_area_struct *vm_area_alloc(struct mm_struct *mm)
509 {
510 struct vm_area_struct *vma = calloc(1, sizeof(struct vm_area_struct));
511
512 if (!vma)
513 return NULL;
514
515 vma_init(vma, mm);
516
517 return vma;
518 }
519
vm_area_dup(struct vm_area_struct * orig)520 static inline struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
521 {
522 struct vm_area_struct *new = calloc(1, sizeof(struct vm_area_struct));
523
524 if (!new)
525 return NULL;
526
527 memcpy(new, orig, sizeof(*new));
528 refcount_set(&new->vm_refcnt, 0);
529 new->vm_lock_seq = UINT_MAX;
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
vm_area_free(struct vm_area_struct * vma)699 static inline void vm_area_free(struct vm_area_struct *vma)
700 {
701 free(vma);
702 }
703
lru_add_drain(void)704 static inline void lru_add_drain(void)
705 {
706 }
707
tlb_gather_mmu(struct mmu_gather *,struct mm_struct *)708 static inline void tlb_gather_mmu(struct mmu_gather *, struct mm_struct *)
709 {
710 }
711
update_hiwater_rss(struct mm_struct *)712 static inline void update_hiwater_rss(struct mm_struct *)
713 {
714 }
715
update_hiwater_vm(struct mm_struct *)716 static inline void update_hiwater_vm(struct mm_struct *)
717 {
718 }
719
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)720 static inline void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
721 struct vm_area_struct *vma, unsigned long start_addr,
722 unsigned long end_addr, unsigned long tree_end,
723 bool mm_wr_locked)
724 {
725 (void)tlb;
726 (void)mas;
727 (void)vma;
728 (void)start_addr;
729 (void)end_addr;
730 (void)tree_end;
731 (void)mm_wr_locked;
732 }
733
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)734 static inline void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
735 struct vm_area_struct *vma, unsigned long floor,
736 unsigned long ceiling, bool mm_wr_locked)
737 {
738 (void)tlb;
739 (void)mas;
740 (void)vma;
741 (void)floor;
742 (void)ceiling;
743 (void)mm_wr_locked;
744 }
745
mapping_unmap_writable(struct address_space *)746 static inline void mapping_unmap_writable(struct address_space *)
747 {
748 }
749
flush_dcache_mmap_lock(struct address_space *)750 static inline void flush_dcache_mmap_lock(struct address_space *)
751 {
752 }
753
tlb_finish_mmu(struct mmu_gather *)754 static inline void tlb_finish_mmu(struct mmu_gather *)
755 {
756 }
757
get_file(struct file * f)758 static inline struct file *get_file(struct file *f)
759 {
760 return f;
761 }
762
vma_dup_policy(struct vm_area_struct *,struct vm_area_struct *)763 static inline int vma_dup_policy(struct vm_area_struct *, struct vm_area_struct *)
764 {
765 return 0;
766 }
767
anon_vma_clone(struct vm_area_struct * dst,struct vm_area_struct * src)768 static inline int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
769 {
770 /* For testing purposes. We indicate that an anon_vma has been cloned. */
771 if (src->anon_vma != NULL) {
772 dst->anon_vma = src->anon_vma;
773 dst->anon_vma->was_cloned = true;
774 }
775
776 return 0;
777 }
778
vma_start_write(struct vm_area_struct * vma)779 static inline void vma_start_write(struct vm_area_struct *vma)
780 {
781 /* Used to indicate to tests that a write operation has begun. */
782 vma->vm_lock_seq++;
783 }
784
vma_adjust_trans_huge(struct vm_area_struct * vma,unsigned long start,unsigned long end,struct vm_area_struct * next)785 static inline void vma_adjust_trans_huge(struct vm_area_struct *vma,
786 unsigned long start,
787 unsigned long end,
788 struct vm_area_struct *next)
789 {
790 (void)vma;
791 (void)start;
792 (void)end;
793 (void)next;
794 }
795
vma_iter_free(struct vma_iterator * vmi)796 static inline void vma_iter_free(struct vma_iterator *vmi)
797 {
798 mas_destroy(&vmi->mas);
799 }
800
801 static inline
vma_iter_next_range(struct vma_iterator * vmi)802 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
803 {
804 return mas_next_range(&vmi->mas, ULONG_MAX);
805 }
806
vm_acct_memory(long pages)807 static inline void vm_acct_memory(long pages)
808 {
809 }
810
vma_interval_tree_insert(struct vm_area_struct *,struct rb_root_cached *)811 static inline void vma_interval_tree_insert(struct vm_area_struct *,
812 struct rb_root_cached *)
813 {
814 }
815
vma_interval_tree_remove(struct vm_area_struct *,struct rb_root_cached *)816 static inline void vma_interval_tree_remove(struct vm_area_struct *,
817 struct rb_root_cached *)
818 {
819 }
820
flush_dcache_mmap_unlock(struct address_space *)821 static inline void flush_dcache_mmap_unlock(struct address_space *)
822 {
823 }
824
anon_vma_interval_tree_insert(struct anon_vma_chain *,struct rb_root_cached *)825 static inline void anon_vma_interval_tree_insert(struct anon_vma_chain*,
826 struct rb_root_cached *)
827 {
828 }
829
anon_vma_interval_tree_remove(struct anon_vma_chain *,struct rb_root_cached *)830 static inline void anon_vma_interval_tree_remove(struct anon_vma_chain*,
831 struct rb_root_cached *)
832 {
833 }
834
uprobe_mmap(struct vm_area_struct *)835 static inline void uprobe_mmap(struct vm_area_struct *)
836 {
837 }
838
uprobe_munmap(struct vm_area_struct * vma,unsigned long start,unsigned long end)839 static inline void uprobe_munmap(struct vm_area_struct *vma,
840 unsigned long start, unsigned long end)
841 {
842 (void)vma;
843 (void)start;
844 (void)end;
845 }
846
i_mmap_lock_write(struct address_space *)847 static inline void i_mmap_lock_write(struct address_space *)
848 {
849 }
850
anon_vma_lock_write(struct anon_vma *)851 static inline void anon_vma_lock_write(struct anon_vma *)
852 {
853 }
854
vma_assert_write_locked(struct vm_area_struct *)855 static inline void vma_assert_write_locked(struct vm_area_struct *)
856 {
857 }
858
unlink_anon_vmas(struct vm_area_struct * vma)859 static inline void unlink_anon_vmas(struct vm_area_struct *vma)
860 {
861 /* For testing purposes, indicate that the anon_vma was unlinked. */
862 vma->anon_vma->was_unlinked = true;
863 }
864
anon_vma_unlock_write(struct anon_vma *)865 static inline void anon_vma_unlock_write(struct anon_vma *)
866 {
867 }
868
i_mmap_unlock_write(struct address_space *)869 static inline void i_mmap_unlock_write(struct address_space *)
870 {
871 }
872
anon_vma_merge(struct vm_area_struct *,struct vm_area_struct *)873 static inline void anon_vma_merge(struct vm_area_struct *,
874 struct vm_area_struct *)
875 {
876 }
877
userfaultfd_unmap_prep(struct vm_area_struct * vma,unsigned long start,unsigned long end,struct list_head * unmaps)878 static inline int userfaultfd_unmap_prep(struct vm_area_struct *vma,
879 unsigned long start,
880 unsigned long end,
881 struct list_head *unmaps)
882 {
883 (void)vma;
884 (void)start;
885 (void)end;
886 (void)unmaps;
887
888 return 0;
889 }
890
mmap_write_downgrade(struct mm_struct *)891 static inline void mmap_write_downgrade(struct mm_struct *)
892 {
893 }
894
mmap_read_unlock(struct mm_struct *)895 static inline void mmap_read_unlock(struct mm_struct *)
896 {
897 }
898
mmap_write_unlock(struct mm_struct *)899 static inline void mmap_write_unlock(struct mm_struct *)
900 {
901 }
902
mmap_write_lock_killable(struct mm_struct *)903 static inline int mmap_write_lock_killable(struct mm_struct *)
904 {
905 return 0;
906 }
907
can_modify_mm(struct mm_struct * mm,unsigned long start,unsigned long end)908 static inline bool can_modify_mm(struct mm_struct *mm,
909 unsigned long start,
910 unsigned long end)
911 {
912 (void)mm;
913 (void)start;
914 (void)end;
915
916 return true;
917 }
918
arch_unmap(struct mm_struct * mm,unsigned long start,unsigned long end)919 static inline void arch_unmap(struct mm_struct *mm,
920 unsigned long start,
921 unsigned long end)
922 {
923 (void)mm;
924 (void)start;
925 (void)end;
926 }
927
mmap_assert_locked(struct mm_struct *)928 static inline void mmap_assert_locked(struct mm_struct *)
929 {
930 }
931
mpol_equal(struct mempolicy *,struct mempolicy *)932 static inline bool mpol_equal(struct mempolicy *, struct mempolicy *)
933 {
934 return true;
935 }
936
khugepaged_enter_vma(struct vm_area_struct * vma,unsigned long vm_flags)937 static inline void khugepaged_enter_vma(struct vm_area_struct *vma,
938 unsigned long vm_flags)
939 {
940 (void)vma;
941 (void)vm_flags;
942 }
943
mapping_can_writeback(struct address_space *)944 static inline bool mapping_can_writeback(struct address_space *)
945 {
946 return true;
947 }
948
is_vm_hugetlb_page(struct vm_area_struct *)949 static inline bool is_vm_hugetlb_page(struct vm_area_struct *)
950 {
951 return false;
952 }
953
vma_soft_dirty_enabled(struct vm_area_struct *)954 static inline bool vma_soft_dirty_enabled(struct vm_area_struct *)
955 {
956 return false;
957 }
958
userfaultfd_wp(struct vm_area_struct *)959 static inline bool userfaultfd_wp(struct vm_area_struct *)
960 {
961 return false;
962 }
963
mmap_assert_write_locked(struct mm_struct *)964 static inline void mmap_assert_write_locked(struct mm_struct *)
965 {
966 }
967
mutex_lock(struct mutex *)968 static inline void mutex_lock(struct mutex *)
969 {
970 }
971
mutex_unlock(struct mutex *)972 static inline void mutex_unlock(struct mutex *)
973 {
974 }
975
mutex_is_locked(struct mutex *)976 static inline bool mutex_is_locked(struct mutex *)
977 {
978 return true;
979 }
980
signal_pending(void *)981 static inline bool signal_pending(void *)
982 {
983 return false;
984 }
985
is_file_hugepages(struct file *)986 static inline bool is_file_hugepages(struct file *)
987 {
988 return false;
989 }
990
security_vm_enough_memory_mm(struct mm_struct *,long)991 static inline int security_vm_enough_memory_mm(struct mm_struct *, long)
992 {
993 return 0;
994 }
995
may_expand_vm(struct mm_struct *,vm_flags_t,unsigned long)996 static inline bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long)
997 {
998 return true;
999 }
1000
vm_flags_init(struct vm_area_struct * vma,vm_flags_t flags)1001 static inline void vm_flags_init(struct vm_area_struct *vma,
1002 vm_flags_t flags)
1003 {
1004 vma->__vm_flags = flags;
1005 }
1006
vm_flags_set(struct vm_area_struct * vma,vm_flags_t flags)1007 static inline void vm_flags_set(struct vm_area_struct *vma,
1008 vm_flags_t flags)
1009 {
1010 vma_start_write(vma);
1011 vma->__vm_flags |= flags;
1012 }
1013
vm_flags_clear(struct vm_area_struct * vma,vm_flags_t flags)1014 static inline void vm_flags_clear(struct vm_area_struct *vma,
1015 vm_flags_t flags)
1016 {
1017 vma_start_write(vma);
1018 vma->__vm_flags &= ~flags;
1019 }
1020
call_mmap(struct file *,struct vm_area_struct *)1021 static inline int call_mmap(struct file *, struct vm_area_struct *)
1022 {
1023 return 0;
1024 }
1025
shmem_zero_setup(struct vm_area_struct *)1026 static inline int shmem_zero_setup(struct vm_area_struct *)
1027 {
1028 return 0;
1029 }
1030
vma_set_anonymous(struct vm_area_struct * vma)1031 static inline void vma_set_anonymous(struct vm_area_struct *vma)
1032 {
1033 vma->vm_ops = NULL;
1034 }
1035
ksm_add_vma(struct vm_area_struct *)1036 static inline void ksm_add_vma(struct vm_area_struct *)
1037 {
1038 }
1039
perf_event_mmap(struct vm_area_struct *)1040 static inline void perf_event_mmap(struct vm_area_struct *)
1041 {
1042 }
1043
vma_is_dax(struct vm_area_struct *)1044 static inline bool vma_is_dax(struct vm_area_struct *)
1045 {
1046 return false;
1047 }
1048
get_gate_vma(struct mm_struct *)1049 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *)
1050 {
1051 return NULL;
1052 }
1053
1054 bool vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1055
1056 /* Update vma->vm_page_prot to reflect vma->vm_flags. */
vma_set_page_prot(struct vm_area_struct * vma)1057 static inline void vma_set_page_prot(struct vm_area_struct *vma)
1058 {
1059 unsigned long vm_flags = vma->vm_flags;
1060 pgprot_t vm_page_prot;
1061
1062 /* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1063 vm_page_prot = pgprot_modify(vma->vm_page_prot, vm_get_page_prot(vm_flags));
1064
1065 if (vma_wants_writenotify(vma, vm_page_prot)) {
1066 vm_flags &= ~VM_SHARED;
1067 /* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1068 vm_page_prot = pgprot_modify(vm_page_prot, vm_get_page_prot(vm_flags));
1069 }
1070 /* remove_protection_ptes reads vma->vm_page_prot without mmap_lock */
1071 WRITE_ONCE(vma->vm_page_prot, vm_page_prot);
1072 }
1073
arch_validate_flags(unsigned long)1074 static inline bool arch_validate_flags(unsigned long)
1075 {
1076 return true;
1077 }
1078
vma_close(struct vm_area_struct *)1079 static inline void vma_close(struct vm_area_struct *)
1080 {
1081 }
1082
mmap_file(struct file *,struct vm_area_struct *)1083 static inline int mmap_file(struct file *, struct vm_area_struct *)
1084 {
1085 return 0;
1086 }
1087
stack_guard_start_gap(struct vm_area_struct * vma)1088 static inline unsigned long stack_guard_start_gap(struct vm_area_struct *vma)
1089 {
1090 if (vma->vm_flags & VM_GROWSDOWN)
1091 return stack_guard_gap;
1092
1093 /* See reasoning around the VM_SHADOW_STACK definition */
1094 if (vma->vm_flags & VM_SHADOW_STACK)
1095 return PAGE_SIZE;
1096
1097 return 0;
1098 }
1099
vm_start_gap(struct vm_area_struct * vma)1100 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
1101 {
1102 unsigned long gap = stack_guard_start_gap(vma);
1103 unsigned long vm_start = vma->vm_start;
1104
1105 vm_start -= gap;
1106 if (vm_start > vma->vm_start)
1107 vm_start = 0;
1108 return vm_start;
1109 }
1110
vm_end_gap(struct vm_area_struct * vma)1111 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
1112 {
1113 unsigned long vm_end = vma->vm_end;
1114
1115 if (vma->vm_flags & VM_GROWSUP) {
1116 vm_end += stack_guard_gap;
1117 if (vm_end < vma->vm_end)
1118 vm_end = -PAGE_SIZE;
1119 }
1120 return vm_end;
1121 }
1122
is_hugepage_only_range(struct mm_struct * mm,unsigned long addr,unsigned long len)1123 static inline int is_hugepage_only_range(struct mm_struct *mm,
1124 unsigned long addr, unsigned long len)
1125 {
1126 return 0;
1127 }
1128
vma_is_accessible(struct vm_area_struct * vma)1129 static inline bool vma_is_accessible(struct vm_area_struct *vma)
1130 {
1131 return vma->vm_flags & VM_ACCESS_FLAGS;
1132 }
1133
capable(int cap)1134 static inline bool capable(int cap)
1135 {
1136 return true;
1137 }
1138
mlock_future_ok(struct mm_struct * mm,unsigned long flags,unsigned long bytes)1139 static inline bool mlock_future_ok(struct mm_struct *mm, unsigned long flags,
1140 unsigned long bytes)
1141 {
1142 unsigned long locked_pages, limit_pages;
1143
1144 if (!(flags & VM_LOCKED) || capable(CAP_IPC_LOCK))
1145 return true;
1146
1147 locked_pages = bytes >> PAGE_SHIFT;
1148 locked_pages += mm->locked_vm;
1149
1150 limit_pages = rlimit(RLIMIT_MEMLOCK);
1151 limit_pages >>= PAGE_SHIFT;
1152
1153 return locked_pages <= limit_pages;
1154 }
1155
__anon_vma_prepare(struct vm_area_struct * vma)1156 static inline int __anon_vma_prepare(struct vm_area_struct *vma)
1157 {
1158 struct anon_vma *anon_vma = calloc(1, sizeof(struct anon_vma));
1159
1160 if (!anon_vma)
1161 return -ENOMEM;
1162
1163 anon_vma->root = anon_vma;
1164 vma->anon_vma = anon_vma;
1165
1166 return 0;
1167 }
1168
anon_vma_prepare(struct vm_area_struct * vma)1169 static inline int anon_vma_prepare(struct vm_area_struct *vma)
1170 {
1171 if (likely(vma->anon_vma))
1172 return 0;
1173
1174 return __anon_vma_prepare(vma);
1175 }
1176
userfaultfd_unmap_complete(struct mm_struct * mm,struct list_head * uf)1177 static inline void userfaultfd_unmap_complete(struct mm_struct *mm,
1178 struct list_head *uf)
1179 {
1180 }
1181
1182 /*
1183 * Denies creating a writable executable mapping or gaining executable permissions.
1184 *
1185 * This denies the following:
1186 *
1187 * a) mmap(PROT_WRITE | PROT_EXEC)
1188 *
1189 * b) mmap(PROT_WRITE)
1190 * mprotect(PROT_EXEC)
1191 *
1192 * c) mmap(PROT_WRITE)
1193 * mprotect(PROT_READ)
1194 * mprotect(PROT_EXEC)
1195 *
1196 * But allows the following:
1197 *
1198 * d) mmap(PROT_READ | PROT_EXEC)
1199 * mmap(PROT_READ | PROT_EXEC | PROT_BTI)
1200 *
1201 * This is only applicable if the user has set the Memory-Deny-Write-Execute
1202 * (MDWE) protection mask for the current process.
1203 *
1204 * @old specifies the VMA flags the VMA originally possessed, and @new the ones
1205 * we propose to set.
1206 *
1207 * Return: false if proposed change is OK, true if not ok and should be denied.
1208 */
map_deny_write_exec(unsigned long old,unsigned long new)1209 static inline bool map_deny_write_exec(unsigned long old, unsigned long new)
1210 {
1211 /* If MDWE is disabled, we have nothing to deny. */
1212 if (!test_bit(MMF_HAS_MDWE, ¤t->mm->flags))
1213 return false;
1214
1215 /* If the new VMA is not executable, we have nothing to deny. */
1216 if (!(new & VM_EXEC))
1217 return false;
1218
1219 /* Under MDWE we do not accept newly writably executable VMAs... */
1220 if (new & VM_WRITE)
1221 return true;
1222
1223 /* ...nor previously non-executable VMAs becoming executable. */
1224 if (!(old & VM_EXEC))
1225 return true;
1226
1227 return false;
1228 }
1229
mapping_map_writable(struct address_space * mapping)1230 static inline int mapping_map_writable(struct address_space *mapping)
1231 {
1232 int c = atomic_read(&mapping->i_mmap_writable);
1233
1234 /* Derived from the raw_atomic_inc_unless_negative() implementation. */
1235 do {
1236 if (c < 0)
1237 return -EPERM;
1238 } while (!__sync_bool_compare_and_swap(&mapping->i_mmap_writable, c, c+1));
1239
1240 return 0;
1241 }
1242
1243 #endif /* __MM_VMA_INTERNAL_H */
1244