xref: /linux/tools/testing/vma/vma_internal.h (revision 10725cd2b09afe512ff60df651b9dd0e11720801)
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_SEQ_READ	0x00008000	/* App will access data sequentially */
60 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
61 #define VM_DONTEXPAND	0x00040000
62 #define VM_LOCKONFAULT	0x00080000
63 #define VM_ACCOUNT	0x00100000
64 #define VM_NORESERVE	0x00200000
65 #define VM_MIXEDMAP	0x10000000
66 #define VM_STACK	VM_GROWSDOWN
67 #define VM_SHADOW_STACK	VM_NONE
68 #define VM_SOFTDIRTY	0
69 #define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
70 #define VM_GROWSUP	VM_NONE
71 
72 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
73 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
74 
75 #ifdef CONFIG_STACK_GROWSUP
76 #define VM_STACK	VM_GROWSUP
77 #define VM_STACK_EARLY	VM_GROWSDOWN
78 #else
79 #define VM_STACK	VM_GROWSDOWN
80 #define VM_STACK_EARLY	0
81 #endif
82 
83 #define DEFAULT_MAP_WINDOW	((1UL << 47) - PAGE_SIZE)
84 #define TASK_SIZE_LOW		DEFAULT_MAP_WINDOW
85 #define TASK_SIZE_MAX		DEFAULT_MAP_WINDOW
86 #define STACK_TOP		TASK_SIZE_LOW
87 #define STACK_TOP_MAX		TASK_SIZE_MAX
88 
89 /* This mask represents all the VMA flag bits used by mlock */
90 #define VM_LOCKED_MASK	(VM_LOCKED | VM_LOCKONFAULT)
91 
92 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
93 
94 #define VM_DATA_FLAGS_TSK_EXEC	(VM_READ | VM_WRITE | TASK_EXEC | \
95 				 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
96 
97 #define VM_DATA_DEFAULT_FLAGS	VM_DATA_FLAGS_TSK_EXEC
98 
99 #define VM_STARTGAP_FLAGS (VM_GROWSDOWN | VM_SHADOW_STACK)
100 
101 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
102 #define VM_STACK_FLAGS	(VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
103 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY)
104 
105 #define RLIMIT_STACK		3	/* max stack size */
106 #define RLIMIT_MEMLOCK		8	/* max locked-in-memory address space */
107 
108 #define CAP_IPC_LOCK         14
109 
110 #ifdef CONFIG_64BIT
111 #define VM_SEALED_BIT	42
112 #define VM_SEALED	BIT(VM_SEALED_BIT)
113 #else
114 #define VM_SEALED	VM_NONE
115 #endif
116 
117 #define FIRST_USER_ADDRESS	0UL
118 #define USER_PGTABLES_CEILING	0UL
119 
120 #define vma_policy(vma) NULL
121 
122 #define down_write_nest_lock(sem, nest_lock)
123 
124 #define pgprot_val(x)		((x).pgprot)
125 #define __pgprot(x)		((pgprot_t) { (x) } )
126 
127 #define for_each_vma(__vmi, __vma)					\
128 	while (((__vma) = vma_next(&(__vmi))) != NULL)
129 
130 /* The MM code likes to work with exclusive end addresses */
131 #define for_each_vma_range(__vmi, __vma, __end)				\
132 	while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
133 
134 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
135 
136 #define PHYS_PFN(x)	((unsigned long)((x) >> PAGE_SHIFT))
137 
138 #define test_and_set_bit(nr, addr) __test_and_set_bit(nr, addr)
139 #define test_and_clear_bit(nr, addr) __test_and_clear_bit(nr, addr)
140 
141 #define TASK_SIZE ((1ul << 47)-PAGE_SIZE)
142 
143 #define AS_MM_ALL_LOCKS 2
144 
145 /* We hardcode this for now. */
146 #define sysctl_max_map_count 0x1000000UL
147 
148 #define pgoff_t unsigned long
149 typedef unsigned long	pgprotval_t;
150 typedef struct pgprot { pgprotval_t pgprot; } pgprot_t;
151 typedef unsigned long vm_flags_t;
152 typedef __bitwise unsigned int vm_fault_t;
153 
154 /*
155  * The shared stubs do not implement this, it amounts to an fprintf(STDERR,...)
156  * either way :)
157  */
158 #define pr_warn_once pr_err
159 
160 #define data_race(expr) expr
161 
162 #define ASSERT_EXCLUSIVE_WRITER(x)
163 
164 /**
165  * swap - swap values of @a and @b
166  * @a: first value
167  * @b: second value
168  */
169 #define swap(a, b) \
170 	do { typeof(a) __tmp = (a); (a) = (b); (b) = __tmp; } while (0)
171 
172 struct kref {
173 	refcount_t refcount;
174 };
175 
176 /*
177  * Define the task command name length as enum, then it can be visible to
178  * BPF programs.
179  */
180 enum {
181 	TASK_COMM_LEN = 16,
182 };
183 
184 /*
185  * Flags for bug emulation.
186  *
187  * These occupy the top three bytes.
188  */
189 enum {
190 	READ_IMPLIES_EXEC =	0x0400000,
191 };
192 
193 struct task_struct {
194 	char comm[TASK_COMM_LEN];
195 	pid_t pid;
196 	struct mm_struct *mm;
197 
198 	/* Used for emulating ABI behavior of previous Linux versions: */
199 	unsigned int			personality;
200 };
201 
202 struct task_struct *get_current(void);
203 #define current get_current()
204 
205 struct anon_vma {
206 	struct anon_vma *root;
207 	struct rb_root_cached rb_root;
208 
209 	/* Test fields. */
210 	bool was_cloned;
211 	bool was_unlinked;
212 };
213 
214 struct anon_vma_chain {
215 	struct anon_vma *anon_vma;
216 	struct list_head same_vma;
217 };
218 
219 struct anon_vma_name {
220 	struct kref kref;
221 	/* The name needs to be at the end because it is dynamically sized. */
222 	char name[];
223 };
224 
225 struct vma_iterator {
226 	struct ma_state mas;
227 };
228 
229 #define VMA_ITERATOR(name, __mm, __addr)				\
230 	struct vma_iterator name = {					\
231 		.mas = {						\
232 			.tree = &(__mm)->mm_mt,				\
233 			.index = __addr,				\
234 			.node = NULL,					\
235 			.status = ma_start,				\
236 		},							\
237 	}
238 
239 struct address_space {
240 	struct rb_root_cached	i_mmap;
241 	unsigned long		flags;
242 	atomic_t		i_mmap_writable;
243 };
244 
245 struct vm_userfaultfd_ctx {};
246 struct mempolicy {};
247 struct mmu_gather {};
248 struct mutex {};
249 #define DEFINE_MUTEX(mutexname) \
250 	struct mutex mutexname = {}
251 
252 struct mm_struct {
253 	struct maple_tree mm_mt;
254 	int map_count;			/* number of VMAs */
255 	unsigned long total_vm;	   /* Total pages mapped */
256 	unsigned long locked_vm;   /* Pages that have PG_mlocked set */
257 	unsigned long data_vm;	   /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
258 	unsigned long exec_vm;	   /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
259 	unsigned long stack_vm;	   /* VM_STACK */
260 
261 	unsigned long def_flags;
262 
263 	unsigned long flags; /* Must use atomic bitops to access */
264 };
265 
266 struct vm_area_struct;
267 
268 /*
269  * Describes a VMA that is about to be mmap()'ed. Drivers may choose to
270  * manipulate mutable fields which will cause those fields to be updated in the
271  * resultant VMA.
272  *
273  * Helper functions are not required for manipulating any field.
274  */
275 struct vm_area_desc {
276 	/* Immutable state. */
277 	struct mm_struct *mm;
278 	unsigned long start;
279 	unsigned long end;
280 
281 	/* Mutable fields. Populated with initial state. */
282 	pgoff_t pgoff;
283 	struct file *file;
284 	vm_flags_t vm_flags;
285 	pgprot_t page_prot;
286 
287 	/* Write-only fields. */
288 	const struct vm_operations_struct *vm_ops;
289 	void *private_data;
290 };
291 
292 struct file_operations {
293 	int (*mmap)(struct file *, struct vm_area_struct *);
294 	int (*mmap_prepare)(struct vm_area_desc *);
295 };
296 
297 struct file {
298 	struct address_space	*f_mapping;
299 	const struct file_operations	*f_op;
300 };
301 
302 #define VMA_LOCK_OFFSET	0x40000000
303 
304 typedef struct { unsigned long v; } freeptr_t;
305 
306 struct vm_area_struct {
307 	/* The first cache line has the info for VMA tree walking. */
308 
309 	union {
310 		struct {
311 			/* VMA covers [vm_start; vm_end) addresses within mm */
312 			unsigned long vm_start;
313 			unsigned long vm_end;
314 		};
315 		freeptr_t vm_freeptr; /* Pointer used by SLAB_TYPESAFE_BY_RCU */
316 	};
317 
318 	struct mm_struct *vm_mm;	/* The address space we belong to. */
319 	pgprot_t vm_page_prot;          /* Access permissions of this VMA. */
320 
321 	/*
322 	 * Flags, see mm.h.
323 	 * To modify use vm_flags_{init|reset|set|clear|mod} functions.
324 	 */
325 	union {
326 		const vm_flags_t vm_flags;
327 		vm_flags_t __private __vm_flags;
328 	};
329 
330 #ifdef CONFIG_PER_VMA_LOCK
331 	/*
332 	 * Can only be written (using WRITE_ONCE()) while holding both:
333 	 *  - mmap_lock (in write mode)
334 	 *  - vm_refcnt bit at VMA_LOCK_OFFSET is set
335 	 * Can be read reliably while holding one of:
336 	 *  - mmap_lock (in read or write mode)
337 	 *  - vm_refcnt bit at VMA_LOCK_OFFSET is set or vm_refcnt > 1
338 	 * Can be read unreliably (using READ_ONCE()) for pessimistic bailout
339 	 * while holding nothing (except RCU to keep the VMA struct allocated).
340 	 *
341 	 * This sequence counter is explicitly allowed to overflow; sequence
342 	 * counter reuse can only lead to occasional unnecessary use of the
343 	 * slowpath.
344 	 */
345 	unsigned int vm_lock_seq;
346 #endif
347 
348 	/*
349 	 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
350 	 * list, after a COW of one of the file pages.	A MAP_SHARED vma
351 	 * can only be in the i_mmap tree.  An anonymous MAP_PRIVATE, stack
352 	 * or brk vma (with NULL file) can only be in an anon_vma list.
353 	 */
354 	struct list_head anon_vma_chain; /* Serialized by mmap_lock &
355 					  * page_table_lock */
356 	struct anon_vma *anon_vma;	/* Serialized by page_table_lock */
357 
358 	/* Function pointers to deal with this struct. */
359 	const struct vm_operations_struct *vm_ops;
360 
361 	/* Information about our backing store: */
362 	unsigned long vm_pgoff;		/* Offset (within vm_file) in PAGE_SIZE
363 					   units */
364 	struct file * vm_file;		/* File we map to (can be NULL). */
365 	void * vm_private_data;		/* was vm_pte (shared mem) */
366 
367 #ifdef CONFIG_SWAP
368 	atomic_long_t swap_readahead_info;
369 #endif
370 #ifndef CONFIG_MMU
371 	struct vm_region *vm_region;	/* NOMMU mapping region */
372 #endif
373 #ifdef CONFIG_NUMA
374 	struct mempolicy *vm_policy;	/* NUMA policy for the VMA */
375 #endif
376 #ifdef CONFIG_NUMA_BALANCING
377 	struct vma_numab_state *numab_state;	/* NUMA Balancing state */
378 #endif
379 #ifdef CONFIG_PER_VMA_LOCK
380 	/* Unstable RCU readers are allowed to read this. */
381 	refcount_t vm_refcnt;
382 #endif
383 	/*
384 	 * For areas with an address space and backing store,
385 	 * linkage into the address_space->i_mmap interval tree.
386 	 *
387 	 */
388 	struct {
389 		struct rb_node rb;
390 		unsigned long rb_subtree_last;
391 	} shared;
392 #ifdef CONFIG_ANON_VMA_NAME
393 	/*
394 	 * For private and shared anonymous mappings, a pointer to a null
395 	 * terminated string containing the name given to the vma, or NULL if
396 	 * unnamed. Serialized by mmap_lock. Use anon_vma_name to access.
397 	 */
398 	struct anon_vma_name *anon_name;
399 #endif
400 	struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
401 } __randomize_layout;
402 
403 struct vm_fault {};
404 
405 struct vm_operations_struct {
406 	void (*open)(struct vm_area_struct * area);
407 	/**
408 	 * @close: Called when the VMA is being removed from the MM.
409 	 * Context: User context.  May sleep.  Caller holds mmap_lock.
410 	 */
411 	void (*close)(struct vm_area_struct * area);
412 	/* Called any time before splitting to check if it's allowed */
413 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
414 	int (*mremap)(struct vm_area_struct *area);
415 	/*
416 	 * Called by mprotect() to make driver-specific permission
417 	 * checks before mprotect() is finalised.   The VMA must not
418 	 * be modified.  Returns 0 if mprotect() can proceed.
419 	 */
420 	int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
421 			unsigned long end, unsigned long newflags);
422 	vm_fault_t (*fault)(struct vm_fault *vmf);
423 	vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
424 	vm_fault_t (*map_pages)(struct vm_fault *vmf,
425 			pgoff_t start_pgoff, pgoff_t end_pgoff);
426 	unsigned long (*pagesize)(struct vm_area_struct * area);
427 
428 	/* notification that a previously read-only page is about to become
429 	 * writable, if an error is returned it will cause a SIGBUS */
430 	vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
431 
432 	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
433 	vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
434 
435 	/* called by access_process_vm when get_user_pages() fails, typically
436 	 * for use by special VMAs. See also generic_access_phys() for a generic
437 	 * implementation useful for any iomem mapping.
438 	 */
439 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
440 		      void *buf, int len, int write);
441 
442 	/* Called by the /proc/PID/maps code to ask the vma whether it
443 	 * has a special name.  Returning non-NULL will also cause this
444 	 * vma to be dumped unconditionally. */
445 	const char *(*name)(struct vm_area_struct *vma);
446 
447 #ifdef CONFIG_NUMA
448 	/*
449 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
450 	 * to hold the policy upon return.  Caller should pass NULL @new to
451 	 * remove a policy and fall back to surrounding context--i.e. do not
452 	 * install a MPOL_DEFAULT policy, nor the task or system default
453 	 * mempolicy.
454 	 */
455 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
456 
457 	/*
458 	 * get_policy() op must add reference [mpol_get()] to any policy at
459 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
460 	 * in mm/mempolicy.c will do this automatically.
461 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
462 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
463 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
464 	 * must return NULL--i.e., do not "fallback" to task or system default
465 	 * policy.
466 	 */
467 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
468 					unsigned long addr, pgoff_t *ilx);
469 #endif
470 #ifdef CONFIG_FIND_NORMAL_PAGE
471 	/*
472 	 * Called by vm_normal_page() for special PTEs in @vma at @addr. This
473 	 * allows for returning a "normal" page from vm_normal_page() even
474 	 * though the PTE indicates that the "struct page" either does not exist
475 	 * or should not be touched: "special".
476 	 *
477 	 * Do not add new users: this really only works when a "normal" page
478 	 * was mapped, but then the PTE got changed to something weird (+
479 	 * marked special) that would not make pte_pfn() identify the originally
480 	 * inserted page.
481 	 */
482 	struct page *(*find_normal_page)(struct vm_area_struct *vma,
483 					 unsigned long addr);
484 #endif /* CONFIG_FIND_NORMAL_PAGE */
485 };
486 
487 struct vm_unmapped_area_info {
488 #define VM_UNMAPPED_AREA_TOPDOWN 1
489 	unsigned long flags;
490 	unsigned long length;
491 	unsigned long low_limit;
492 	unsigned long high_limit;
493 	unsigned long align_mask;
494 	unsigned long align_offset;
495 	unsigned long start_gap;
496 };
497 
498 struct pagetable_move_control {
499 	struct vm_area_struct *old; /* Source VMA. */
500 	struct vm_area_struct *new; /* Destination VMA. */
501 	unsigned long old_addr; /* Address from which the move begins. */
502 	unsigned long old_end; /* Exclusive address at which old range ends. */
503 	unsigned long new_addr; /* Address to move page tables to. */
504 	unsigned long len_in; /* Bytes to remap specified by user. */
505 
506 	bool need_rmap_locks; /* Do rmap locks need to be taken? */
507 	bool for_stack; /* Is this an early temp stack being moved? */
508 };
509 
510 #define PAGETABLE_MOVE(name, old_, new_, old_addr_, new_addr_, len_)	\
511 	struct pagetable_move_control name = {				\
512 		.old = old_,						\
513 		.new = new_,						\
514 		.old_addr = old_addr_,					\
515 		.old_end = (old_addr_) + (len_),			\
516 		.new_addr = new_addr_,					\
517 		.len_in = len_,						\
518 	}
519 
520 struct kmem_cache_args {
521 	/**
522 	 * @align: The required alignment for the objects.
523 	 *
524 	 * %0 means no specific alignment is requested.
525 	 */
526 	unsigned int align;
527 	/**
528 	 * @useroffset: Usercopy region offset.
529 	 *
530 	 * %0 is a valid offset, when @usersize is non-%0
531 	 */
532 	unsigned int useroffset;
533 	/**
534 	 * @usersize: Usercopy region size.
535 	 *
536 	 * %0 means no usercopy region is specified.
537 	 */
538 	unsigned int usersize;
539 	/**
540 	 * @freeptr_offset: Custom offset for the free pointer
541 	 * in &SLAB_TYPESAFE_BY_RCU caches
542 	 *
543 	 * By default &SLAB_TYPESAFE_BY_RCU caches place the free pointer
544 	 * outside of the object. This might cause the object to grow in size.
545 	 * Cache creators that have a reason to avoid this can specify a custom
546 	 * free pointer offset in their struct where the free pointer will be
547 	 * placed.
548 	 *
549 	 * Note that placing the free pointer inside the object requires the
550 	 * caller to ensure that no fields are invalidated that are required to
551 	 * guard against object recycling (See &SLAB_TYPESAFE_BY_RCU for
552 	 * details).
553 	 *
554 	 * Using %0 as a value for @freeptr_offset is valid. If @freeptr_offset
555 	 * is specified, %use_freeptr_offset must be set %true.
556 	 *
557 	 * Note that @ctor currently isn't supported with custom free pointers
558 	 * as a @ctor requires an external free pointer.
559 	 */
560 	unsigned int freeptr_offset;
561 	/**
562 	 * @use_freeptr_offset: Whether a @freeptr_offset is used.
563 	 */
564 	bool use_freeptr_offset;
565 	/**
566 	 * @ctor: A constructor for the objects.
567 	 *
568 	 * The constructor is invoked for each object in a newly allocated slab
569 	 * page. It is the cache user's responsibility to free object in the
570 	 * same state as after calling the constructor, or deal appropriately
571 	 * with any differences between a freshly constructed and a reallocated
572 	 * object.
573 	 *
574 	 * %NULL means no constructor.
575 	 */
576 	void (*ctor)(void *);
577 };
578 
579 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
580 {
581 	mas_pause(&vmi->mas);
582 }
583 
584 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
585 {
586 	return __pgprot(pgprot_val(oldprot) | pgprot_val(newprot));
587 }
588 
589 static inline pgprot_t vm_get_page_prot(vm_flags_t vm_flags)
590 {
591 	return __pgprot(vm_flags);
592 }
593 
594 static inline bool is_shared_maywrite(vm_flags_t vm_flags)
595 {
596 	return (vm_flags & (VM_SHARED | VM_MAYWRITE)) ==
597 		(VM_SHARED | VM_MAYWRITE);
598 }
599 
600 static inline bool vma_is_shared_maywrite(struct vm_area_struct *vma)
601 {
602 	return is_shared_maywrite(vma->vm_flags);
603 }
604 
605 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
606 {
607 	/*
608 	 * Uses mas_find() to get the first VMA when the iterator starts.
609 	 * Calling mas_next() could skip the first entry.
610 	 */
611 	return mas_find(&vmi->mas, ULONG_MAX);
612 }
613 
614 /*
615  * WARNING: to avoid racing with vma_mark_attached()/vma_mark_detached(), these
616  * assertions should be made either under mmap_write_lock or when the object
617  * has been isolated under mmap_write_lock, ensuring no competing writers.
618  */
619 static inline void vma_assert_attached(struct vm_area_struct *vma)
620 {
621 	WARN_ON_ONCE(!refcount_read(&vma->vm_refcnt));
622 }
623 
624 static inline void vma_assert_detached(struct vm_area_struct *vma)
625 {
626 	WARN_ON_ONCE(refcount_read(&vma->vm_refcnt));
627 }
628 
629 static inline void vma_assert_write_locked(struct vm_area_struct *);
630 static inline void vma_mark_attached(struct vm_area_struct *vma)
631 {
632 	vma_assert_write_locked(vma);
633 	vma_assert_detached(vma);
634 	refcount_set_release(&vma->vm_refcnt, 1);
635 }
636 
637 static inline void vma_mark_detached(struct vm_area_struct *vma)
638 {
639 	vma_assert_write_locked(vma);
640 	vma_assert_attached(vma);
641 	/* We are the only writer, so no need to use vma_refcount_put(). */
642 	if (unlikely(!refcount_dec_and_test(&vma->vm_refcnt))) {
643 		/*
644 		 * Reader must have temporarily raised vm_refcnt but it will
645 		 * drop it without using the vma since vma is write-locked.
646 		 */
647 	}
648 }
649 
650 extern const struct vm_operations_struct vma_dummy_vm_ops;
651 
652 extern unsigned long rlimit(unsigned int limit);
653 
654 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
655 {
656 	memset(vma, 0, sizeof(*vma));
657 	vma->vm_mm = mm;
658 	vma->vm_ops = &vma_dummy_vm_ops;
659 	INIT_LIST_HEAD(&vma->anon_vma_chain);
660 	vma->vm_lock_seq = UINT_MAX;
661 }
662 
663 struct kmem_cache {
664 	const char *name;
665 	size_t object_size;
666 	struct kmem_cache_args *args;
667 };
668 
669 static inline struct kmem_cache *__kmem_cache_create(const char *name,
670 						     size_t object_size,
671 						     struct kmem_cache_args *args)
672 {
673 	struct kmem_cache *ret = malloc(sizeof(struct kmem_cache));
674 
675 	ret->name = name;
676 	ret->object_size = object_size;
677 	ret->args = args;
678 
679 	return ret;
680 }
681 
682 #define kmem_cache_create(__name, __object_size, __args, ...)           \
683 	__kmem_cache_create((__name), (__object_size), (__args))
684 
685 static inline void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
686 {
687 	(void)gfpflags;
688 
689 	return calloc(s->object_size, 1);
690 }
691 
692 static inline void kmem_cache_free(struct kmem_cache *s, void *x)
693 {
694 	free(x);
695 }
696 
697 /*
698  * These are defined in vma.h, but sadly vm_stat_account() is referenced by
699  * kernel/fork.c, so we have to these broadly available there, and temporarily
700  * define them here to resolve the dependency cycle.
701  */
702 
703 #define is_exec_mapping(flags) \
704 	((flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC)
705 
706 #define is_stack_mapping(flags) \
707 	(((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK))
708 
709 #define is_data_mapping(flags) \
710 	((flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE)
711 
712 static inline void vm_stat_account(struct mm_struct *mm, vm_flags_t flags,
713 				   long npages)
714 {
715 	WRITE_ONCE(mm->total_vm, READ_ONCE(mm->total_vm)+npages);
716 
717 	if (is_exec_mapping(flags))
718 		mm->exec_vm += npages;
719 	else if (is_stack_mapping(flags))
720 		mm->stack_vm += npages;
721 	else if (is_data_mapping(flags))
722 		mm->data_vm += npages;
723 }
724 
725 #undef is_exec_mapping
726 #undef is_stack_mapping
727 #undef is_data_mapping
728 
729 /* Currently stubbed but we may later wish to un-stub. */
730 static inline void vm_acct_memory(long pages);
731 static inline void vm_unacct_memory(long pages)
732 {
733 	vm_acct_memory(-pages);
734 }
735 
736 static inline void mapping_allow_writable(struct address_space *mapping)
737 {
738 	atomic_inc(&mapping->i_mmap_writable);
739 }
740 
741 static inline void vma_set_range(struct vm_area_struct *vma,
742 				 unsigned long start, unsigned long end,
743 				 pgoff_t pgoff)
744 {
745 	vma->vm_start = start;
746 	vma->vm_end = end;
747 	vma->vm_pgoff = pgoff;
748 }
749 
750 static inline
751 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
752 {
753 	return mas_find(&vmi->mas, max - 1);
754 }
755 
756 static inline int vma_iter_clear_gfp(struct vma_iterator *vmi,
757 			unsigned long start, unsigned long end, gfp_t gfp)
758 {
759 	__mas_set_range(&vmi->mas, start, end - 1);
760 	mas_store_gfp(&vmi->mas, NULL, gfp);
761 	if (unlikely(mas_is_err(&vmi->mas)))
762 		return -ENOMEM;
763 
764 	return 0;
765 }
766 
767 static inline void mmap_assert_locked(struct mm_struct *);
768 static inline struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
769 						unsigned long start_addr,
770 						unsigned long end_addr)
771 {
772 	unsigned long index = start_addr;
773 
774 	mmap_assert_locked(mm);
775 	return mt_find(&mm->mm_mt, &index, end_addr - 1);
776 }
777 
778 static inline
779 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
780 {
781 	return mtree_load(&mm->mm_mt, addr);
782 }
783 
784 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
785 {
786 	return mas_prev(&vmi->mas, 0);
787 }
788 
789 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
790 {
791 	mas_set(&vmi->mas, addr);
792 }
793 
794 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
795 {
796 	return !vma->vm_ops;
797 }
798 
799 /* Defined in vma.h, so temporarily define here to avoid circular dependency. */
800 #define vma_iter_load(vmi) \
801 	mas_walk(&(vmi)->mas)
802 
803 static inline struct vm_area_struct *
804 find_vma_prev(struct mm_struct *mm, unsigned long addr,
805 			struct vm_area_struct **pprev)
806 {
807 	struct vm_area_struct *vma;
808 	VMA_ITERATOR(vmi, mm, addr);
809 
810 	vma = vma_iter_load(&vmi);
811 	*pprev = vma_prev(&vmi);
812 	if (!vma)
813 		vma = vma_next(&vmi);
814 	return vma;
815 }
816 
817 #undef vma_iter_load
818 
819 static inline void vma_iter_init(struct vma_iterator *vmi,
820 		struct mm_struct *mm, unsigned long addr)
821 {
822 	mas_init(&vmi->mas, &mm->mm_mt, addr);
823 }
824 
825 /* Stubbed functions. */
826 
827 static inline struct anon_vma_name *anon_vma_name(struct vm_area_struct *vma)
828 {
829 	return NULL;
830 }
831 
832 static inline bool is_mergeable_vm_userfaultfd_ctx(struct vm_area_struct *vma,
833 					struct vm_userfaultfd_ctx vm_ctx)
834 {
835 	return true;
836 }
837 
838 static inline bool anon_vma_name_eq(struct anon_vma_name *anon_name1,
839 				    struct anon_vma_name *anon_name2)
840 {
841 	return true;
842 }
843 
844 static inline void might_sleep(void)
845 {
846 }
847 
848 static inline unsigned long vma_pages(struct vm_area_struct *vma)
849 {
850 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
851 }
852 
853 static inline void fput(struct file *)
854 {
855 }
856 
857 static inline void mpol_put(struct mempolicy *)
858 {
859 }
860 
861 static inline void lru_add_drain(void)
862 {
863 }
864 
865 static inline void tlb_gather_mmu(struct mmu_gather *, struct mm_struct *)
866 {
867 }
868 
869 static inline void update_hiwater_rss(struct mm_struct *)
870 {
871 }
872 
873 static inline void update_hiwater_vm(struct mm_struct *)
874 {
875 }
876 
877 static inline void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
878 		      struct vm_area_struct *vma, unsigned long start_addr,
879 		      unsigned long end_addr, unsigned long tree_end,
880 		      bool mm_wr_locked)
881 {
882 	(void)tlb;
883 	(void)mas;
884 	(void)vma;
885 	(void)start_addr;
886 	(void)end_addr;
887 	(void)tree_end;
888 	(void)mm_wr_locked;
889 }
890 
891 static inline void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
892 		   struct vm_area_struct *vma, unsigned long floor,
893 		   unsigned long ceiling, bool mm_wr_locked)
894 {
895 	(void)tlb;
896 	(void)mas;
897 	(void)vma;
898 	(void)floor;
899 	(void)ceiling;
900 	(void)mm_wr_locked;
901 }
902 
903 static inline void mapping_unmap_writable(struct address_space *)
904 {
905 }
906 
907 static inline void flush_dcache_mmap_lock(struct address_space *)
908 {
909 }
910 
911 static inline void tlb_finish_mmu(struct mmu_gather *)
912 {
913 }
914 
915 static inline struct file *get_file(struct file *f)
916 {
917 	return f;
918 }
919 
920 static inline int vma_dup_policy(struct vm_area_struct *, struct vm_area_struct *)
921 {
922 	return 0;
923 }
924 
925 static inline int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
926 {
927 	/* For testing purposes. We indicate that an anon_vma has been cloned. */
928 	if (src->anon_vma != NULL) {
929 		dst->anon_vma = src->anon_vma;
930 		dst->anon_vma->was_cloned = true;
931 	}
932 
933 	return 0;
934 }
935 
936 static inline void vma_start_write(struct vm_area_struct *vma)
937 {
938 	/* Used to indicate to tests that a write operation has begun. */
939 	vma->vm_lock_seq++;
940 }
941 
942 static inline void vma_adjust_trans_huge(struct vm_area_struct *vma,
943 					 unsigned long start,
944 					 unsigned long end,
945 					 struct vm_area_struct *next)
946 {
947 	(void)vma;
948 	(void)start;
949 	(void)end;
950 	(void)next;
951 }
952 
953 static inline void hugetlb_split(struct vm_area_struct *, unsigned long) {}
954 
955 static inline void vma_iter_free(struct vma_iterator *vmi)
956 {
957 	mas_destroy(&vmi->mas);
958 }
959 
960 static inline
961 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
962 {
963 	return mas_next_range(&vmi->mas, ULONG_MAX);
964 }
965 
966 static inline void vm_acct_memory(long pages)
967 {
968 }
969 
970 static inline void vma_interval_tree_insert(struct vm_area_struct *,
971 					    struct rb_root_cached *)
972 {
973 }
974 
975 static inline void vma_interval_tree_remove(struct vm_area_struct *,
976 					    struct rb_root_cached *)
977 {
978 }
979 
980 static inline void flush_dcache_mmap_unlock(struct address_space *)
981 {
982 }
983 
984 static inline void anon_vma_interval_tree_insert(struct anon_vma_chain*,
985 						 struct rb_root_cached *)
986 {
987 }
988 
989 static inline void anon_vma_interval_tree_remove(struct anon_vma_chain*,
990 						 struct rb_root_cached *)
991 {
992 }
993 
994 static inline void uprobe_mmap(struct vm_area_struct *)
995 {
996 }
997 
998 static inline void uprobe_munmap(struct vm_area_struct *vma,
999 				 unsigned long start, unsigned long end)
1000 {
1001 	(void)vma;
1002 	(void)start;
1003 	(void)end;
1004 }
1005 
1006 static inline void i_mmap_lock_write(struct address_space *)
1007 {
1008 }
1009 
1010 static inline void anon_vma_lock_write(struct anon_vma *)
1011 {
1012 }
1013 
1014 static inline void vma_assert_write_locked(struct vm_area_struct *)
1015 {
1016 }
1017 
1018 static inline void unlink_anon_vmas(struct vm_area_struct *vma)
1019 {
1020 	/* For testing purposes, indicate that the anon_vma was unlinked. */
1021 	vma->anon_vma->was_unlinked = true;
1022 }
1023 
1024 static inline void anon_vma_unlock_write(struct anon_vma *)
1025 {
1026 }
1027 
1028 static inline void i_mmap_unlock_write(struct address_space *)
1029 {
1030 }
1031 
1032 static inline void anon_vma_merge(struct vm_area_struct *,
1033 				  struct vm_area_struct *)
1034 {
1035 }
1036 
1037 static inline int userfaultfd_unmap_prep(struct vm_area_struct *vma,
1038 					 unsigned long start,
1039 					 unsigned long end,
1040 					 struct list_head *unmaps)
1041 {
1042 	(void)vma;
1043 	(void)start;
1044 	(void)end;
1045 	(void)unmaps;
1046 
1047 	return 0;
1048 }
1049 
1050 static inline void mmap_write_downgrade(struct mm_struct *)
1051 {
1052 }
1053 
1054 static inline void mmap_read_unlock(struct mm_struct *)
1055 {
1056 }
1057 
1058 static inline void mmap_write_unlock(struct mm_struct *)
1059 {
1060 }
1061 
1062 static inline int mmap_write_lock_killable(struct mm_struct *)
1063 {
1064 	return 0;
1065 }
1066 
1067 static inline bool can_modify_mm(struct mm_struct *mm,
1068 				 unsigned long start,
1069 				 unsigned long end)
1070 {
1071 	(void)mm;
1072 	(void)start;
1073 	(void)end;
1074 
1075 	return true;
1076 }
1077 
1078 static inline void arch_unmap(struct mm_struct *mm,
1079 				 unsigned long start,
1080 				 unsigned long end)
1081 {
1082 	(void)mm;
1083 	(void)start;
1084 	(void)end;
1085 }
1086 
1087 static inline void mmap_assert_locked(struct mm_struct *)
1088 {
1089 }
1090 
1091 static inline bool mpol_equal(struct mempolicy *, struct mempolicy *)
1092 {
1093 	return true;
1094 }
1095 
1096 static inline void khugepaged_enter_vma(struct vm_area_struct *vma,
1097 			  vm_flags_t vm_flags)
1098 {
1099 	(void)vma;
1100 	(void)vm_flags;
1101 }
1102 
1103 static inline bool mapping_can_writeback(struct address_space *)
1104 {
1105 	return true;
1106 }
1107 
1108 static inline bool is_vm_hugetlb_page(struct vm_area_struct *)
1109 {
1110 	return false;
1111 }
1112 
1113 static inline bool vma_soft_dirty_enabled(struct vm_area_struct *)
1114 {
1115 	return false;
1116 }
1117 
1118 static inline bool userfaultfd_wp(struct vm_area_struct *)
1119 {
1120 	return false;
1121 }
1122 
1123 static inline void mmap_assert_write_locked(struct mm_struct *)
1124 {
1125 }
1126 
1127 static inline void mutex_lock(struct mutex *)
1128 {
1129 }
1130 
1131 static inline void mutex_unlock(struct mutex *)
1132 {
1133 }
1134 
1135 static inline bool mutex_is_locked(struct mutex *)
1136 {
1137 	return true;
1138 }
1139 
1140 static inline bool signal_pending(void *)
1141 {
1142 	return false;
1143 }
1144 
1145 static inline bool is_file_hugepages(struct file *)
1146 {
1147 	return false;
1148 }
1149 
1150 static inline int security_vm_enough_memory_mm(struct mm_struct *, long)
1151 {
1152 	return 0;
1153 }
1154 
1155 static inline bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long)
1156 {
1157 	return true;
1158 }
1159 
1160 static inline void vm_flags_init(struct vm_area_struct *vma,
1161 				 vm_flags_t flags)
1162 {
1163 	vma->__vm_flags = flags;
1164 }
1165 
1166 static inline void vm_flags_set(struct vm_area_struct *vma,
1167 				vm_flags_t flags)
1168 {
1169 	vma_start_write(vma);
1170 	vma->__vm_flags |= flags;
1171 }
1172 
1173 static inline void vm_flags_clear(struct vm_area_struct *vma,
1174 				  vm_flags_t flags)
1175 {
1176 	vma_start_write(vma);
1177 	vma->__vm_flags &= ~flags;
1178 }
1179 
1180 static inline int shmem_zero_setup(struct vm_area_struct *)
1181 {
1182 	return 0;
1183 }
1184 
1185 static inline void vma_set_anonymous(struct vm_area_struct *vma)
1186 {
1187 	vma->vm_ops = NULL;
1188 }
1189 
1190 static inline void ksm_add_vma(struct vm_area_struct *)
1191 {
1192 }
1193 
1194 static inline void perf_event_mmap(struct vm_area_struct *)
1195 {
1196 }
1197 
1198 static inline bool vma_is_dax(struct vm_area_struct *)
1199 {
1200 	return false;
1201 }
1202 
1203 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *)
1204 {
1205 	return NULL;
1206 }
1207 
1208 bool vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1209 
1210 /* Update vma->vm_page_prot to reflect vma->vm_flags. */
1211 static inline void vma_set_page_prot(struct vm_area_struct *vma)
1212 {
1213 	vm_flags_t vm_flags = vma->vm_flags;
1214 	pgprot_t vm_page_prot;
1215 
1216 	/* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1217 	vm_page_prot = pgprot_modify(vma->vm_page_prot, vm_get_page_prot(vm_flags));
1218 
1219 	if (vma_wants_writenotify(vma, vm_page_prot)) {
1220 		vm_flags &= ~VM_SHARED;
1221 		/* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1222 		vm_page_prot = pgprot_modify(vm_page_prot, vm_get_page_prot(vm_flags));
1223 	}
1224 	/* remove_protection_ptes reads vma->vm_page_prot without mmap_lock */
1225 	WRITE_ONCE(vma->vm_page_prot, vm_page_prot);
1226 }
1227 
1228 static inline bool arch_validate_flags(vm_flags_t)
1229 {
1230 	return true;
1231 }
1232 
1233 static inline void vma_close(struct vm_area_struct *)
1234 {
1235 }
1236 
1237 static inline int mmap_file(struct file *, struct vm_area_struct *)
1238 {
1239 	return 0;
1240 }
1241 
1242 static inline unsigned long stack_guard_start_gap(struct vm_area_struct *vma)
1243 {
1244 	if (vma->vm_flags & VM_GROWSDOWN)
1245 		return stack_guard_gap;
1246 
1247 	/* See reasoning around the VM_SHADOW_STACK definition */
1248 	if (vma->vm_flags & VM_SHADOW_STACK)
1249 		return PAGE_SIZE;
1250 
1251 	return 0;
1252 }
1253 
1254 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
1255 {
1256 	unsigned long gap = stack_guard_start_gap(vma);
1257 	unsigned long vm_start = vma->vm_start;
1258 
1259 	vm_start -= gap;
1260 	if (vm_start > vma->vm_start)
1261 		vm_start = 0;
1262 	return vm_start;
1263 }
1264 
1265 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
1266 {
1267 	unsigned long vm_end = vma->vm_end;
1268 
1269 	if (vma->vm_flags & VM_GROWSUP) {
1270 		vm_end += stack_guard_gap;
1271 		if (vm_end < vma->vm_end)
1272 			vm_end = -PAGE_SIZE;
1273 	}
1274 	return vm_end;
1275 }
1276 
1277 static inline int is_hugepage_only_range(struct mm_struct *mm,
1278 					unsigned long addr, unsigned long len)
1279 {
1280 	return 0;
1281 }
1282 
1283 static inline bool vma_is_accessible(struct vm_area_struct *vma)
1284 {
1285 	return vma->vm_flags & VM_ACCESS_FLAGS;
1286 }
1287 
1288 static inline bool capable(int cap)
1289 {
1290 	return true;
1291 }
1292 
1293 static inline bool mlock_future_ok(struct mm_struct *mm, vm_flags_t vm_flags,
1294 			unsigned long bytes)
1295 {
1296 	unsigned long locked_pages, limit_pages;
1297 
1298 	if (!(vm_flags & VM_LOCKED) || capable(CAP_IPC_LOCK))
1299 		return true;
1300 
1301 	locked_pages = bytes >> PAGE_SHIFT;
1302 	locked_pages += mm->locked_vm;
1303 
1304 	limit_pages = rlimit(RLIMIT_MEMLOCK);
1305 	limit_pages >>= PAGE_SHIFT;
1306 
1307 	return locked_pages <= limit_pages;
1308 }
1309 
1310 static inline int __anon_vma_prepare(struct vm_area_struct *vma)
1311 {
1312 	struct anon_vma *anon_vma = calloc(1, sizeof(struct anon_vma));
1313 
1314 	if (!anon_vma)
1315 		return -ENOMEM;
1316 
1317 	anon_vma->root = anon_vma;
1318 	vma->anon_vma = anon_vma;
1319 
1320 	return 0;
1321 }
1322 
1323 static inline int anon_vma_prepare(struct vm_area_struct *vma)
1324 {
1325 	if (likely(vma->anon_vma))
1326 		return 0;
1327 
1328 	return __anon_vma_prepare(vma);
1329 }
1330 
1331 static inline void userfaultfd_unmap_complete(struct mm_struct *mm,
1332 					      struct list_head *uf)
1333 {
1334 }
1335 
1336 /*
1337  * Denies creating a writable executable mapping or gaining executable permissions.
1338  *
1339  * This denies the following:
1340  *
1341  *     a)      mmap(PROT_WRITE | PROT_EXEC)
1342  *
1343  *     b)      mmap(PROT_WRITE)
1344  *             mprotect(PROT_EXEC)
1345  *
1346  *     c)      mmap(PROT_WRITE)
1347  *             mprotect(PROT_READ)
1348  *             mprotect(PROT_EXEC)
1349  *
1350  * But allows the following:
1351  *
1352  *     d)      mmap(PROT_READ | PROT_EXEC)
1353  *             mmap(PROT_READ | PROT_EXEC | PROT_BTI)
1354  *
1355  * This is only applicable if the user has set the Memory-Deny-Write-Execute
1356  * (MDWE) protection mask for the current process.
1357  *
1358  * @old specifies the VMA flags the VMA originally possessed, and @new the ones
1359  * we propose to set.
1360  *
1361  * Return: false if proposed change is OK, true if not ok and should be denied.
1362  */
1363 static inline bool map_deny_write_exec(unsigned long old, unsigned long new)
1364 {
1365 	/* If MDWE is disabled, we have nothing to deny. */
1366 	if (!test_bit(MMF_HAS_MDWE, &current->mm->flags))
1367 		return false;
1368 
1369 	/* If the new VMA is not executable, we have nothing to deny. */
1370 	if (!(new & VM_EXEC))
1371 		return false;
1372 
1373 	/* Under MDWE we do not accept newly writably executable VMAs... */
1374 	if (new & VM_WRITE)
1375 		return true;
1376 
1377 	/* ...nor previously non-executable VMAs becoming executable. */
1378 	if (!(old & VM_EXEC))
1379 		return true;
1380 
1381 	return false;
1382 }
1383 
1384 static inline int mapping_map_writable(struct address_space *mapping)
1385 {
1386 	int c = atomic_read(&mapping->i_mmap_writable);
1387 
1388 	/* Derived from the raw_atomic_inc_unless_negative() implementation. */
1389 	do {
1390 		if (c < 0)
1391 			return -EPERM;
1392 	} while (!__sync_bool_compare_and_swap(&mapping->i_mmap_writable, c, c+1));
1393 
1394 	return 0;
1395 }
1396 
1397 static inline unsigned long move_page_tables(struct pagetable_move_control *pmc)
1398 {
1399 	(void)pmc;
1400 
1401 	return 0;
1402 }
1403 
1404 static inline void free_pgd_range(struct mmu_gather *tlb,
1405 			unsigned long addr, unsigned long end,
1406 			unsigned long floor, unsigned long ceiling)
1407 {
1408 	(void)tlb;
1409 	(void)addr;
1410 	(void)end;
1411 	(void)floor;
1412 	(void)ceiling;
1413 }
1414 
1415 static inline int ksm_execve(struct mm_struct *mm)
1416 {
1417 	(void)mm;
1418 
1419 	return 0;
1420 }
1421 
1422 static inline void ksm_exit(struct mm_struct *mm)
1423 {
1424 	(void)mm;
1425 }
1426 
1427 static inline void vma_lock_init(struct vm_area_struct *vma, bool reset_refcnt)
1428 {
1429 	(void)vma;
1430 	(void)reset_refcnt;
1431 }
1432 
1433 static inline void vma_numab_state_init(struct vm_area_struct *vma)
1434 {
1435 	(void)vma;
1436 }
1437 
1438 static inline void vma_numab_state_free(struct vm_area_struct *vma)
1439 {
1440 	(void)vma;
1441 }
1442 
1443 static inline void dup_anon_vma_name(struct vm_area_struct *orig_vma,
1444 				     struct vm_area_struct *new_vma)
1445 {
1446 	(void)orig_vma;
1447 	(void)new_vma;
1448 }
1449 
1450 static inline void free_anon_vma_name(struct vm_area_struct *vma)
1451 {
1452 	(void)vma;
1453 }
1454 
1455 /* Declared in vma.h. */
1456 static inline void set_vma_from_desc(struct vm_area_struct *vma,
1457 		struct vm_area_desc *desc);
1458 
1459 static inline struct vm_area_desc *vma_to_desc(struct vm_area_struct *vma,
1460 		struct vm_area_desc *desc);
1461 
1462 static int compat_vma_mmap_prepare(struct file *file,
1463 		struct vm_area_struct *vma)
1464 {
1465 	struct vm_area_desc desc;
1466 	int err;
1467 
1468 	err = file->f_op->mmap_prepare(vma_to_desc(vma, &desc));
1469 	if (err)
1470 		return err;
1471 	set_vma_from_desc(vma, &desc);
1472 
1473 	return 0;
1474 }
1475 
1476 /* Did the driver provide valid mmap hook configuration? */
1477 static inline bool can_mmap_file(struct file *file)
1478 {
1479 	bool has_mmap = file->f_op->mmap;
1480 	bool has_mmap_prepare = file->f_op->mmap_prepare;
1481 
1482 	/* Hooks are mutually exclusive. */
1483 	if (WARN_ON_ONCE(has_mmap && has_mmap_prepare))
1484 		return false;
1485 	if (!has_mmap && !has_mmap_prepare)
1486 		return false;
1487 
1488 	return true;
1489 }
1490 
1491 static inline int vfs_mmap(struct file *file, struct vm_area_struct *vma)
1492 {
1493 	if (file->f_op->mmap_prepare)
1494 		return compat_vma_mmap_prepare(file, vma);
1495 
1496 	return file->f_op->mmap(file, vma);
1497 }
1498 
1499 static inline int vfs_mmap_prepare(struct file *file, struct vm_area_desc *desc)
1500 {
1501 	return file->f_op->mmap_prepare(desc);
1502 }
1503 
1504 static inline void fixup_hugetlb_reservations(struct vm_area_struct *vma)
1505 {
1506 	(void)vma;
1507 }
1508 
1509 static inline void vma_set_file(struct vm_area_struct *vma, struct file *file)
1510 {
1511 	/* Changing an anonymous vma with this is illegal */
1512 	get_file(file);
1513 	swap(vma->vm_file, file);
1514 	fput(file);
1515 }
1516 
1517 static inline bool shmem_file(struct file *)
1518 {
1519 	return false;
1520 }
1521 
1522 static inline vm_flags_t ksm_vma_flags(const struct mm_struct *, const struct file *,
1523 			 vm_flags_t vm_flags)
1524 {
1525 	return vm_flags;
1526 }
1527 
1528 #endif	/* __MM_VMA_INTERNAL_H */
1529