xref: /linux/tools/testing/vma/include/dup.h (revision 90cb921c4d7bf92854344d3e76561f48784c613e)
1 /* SPDX-License-Identifier: GPL-2.0+ */
2 
3 #pragma once
4 
5 /* Forward declarations to avoid header cycle. */
6 struct vm_area_struct;
7 static inline void vma_start_write(struct vm_area_struct *vma);
8 
9 extern const struct vm_operations_struct vma_dummy_vm_ops;
10 extern unsigned long stack_guard_gap;
11 extern const struct vm_operations_struct vma_dummy_vm_ops;
12 extern unsigned long rlimit(unsigned int limit);
13 struct task_struct *get_current(void);
14 
15 #define MMF_HAS_MDWE	28
16 #define current get_current()
17 
18 /*
19  * Define the task command name length as enum, then it can be visible to
20  * BPF programs.
21  */
22 enum {
23 	TASK_COMM_LEN = 16,
24 };
25 
26 /* PARTIALLY implemented types. */
27 struct mm_struct {
28 	struct maple_tree mm_mt;
29 	int map_count;			/* number of VMAs */
30 	unsigned long total_vm;	   /* Total pages mapped */
31 	unsigned long locked_vm;   /* Pages that have PG_mlocked set */
32 	unsigned long data_vm;	   /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
33 	unsigned long exec_vm;	   /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
34 	unsigned long stack_vm;	   /* VM_STACK */
35 
36 	union {
37 		vm_flags_t def_flags;
38 		vma_flags_t def_vma_flags;
39 	};
40 
41 	mm_flags_t flags; /* Must use mm_flags_* helpers to access */
42 };
43 struct address_space {
44 	struct rb_root_cached	i_mmap;
45 	unsigned long		flags;
46 	atomic_t		i_mmap_writable;
47 };
48 struct file_operations {
49 	int (*mmap)(struct file *, struct vm_area_struct *);
50 	int (*mmap_prepare)(struct vm_area_desc *);
51 };
52 struct file {
53 	struct address_space	*f_mapping;
54 	const struct file_operations	*f_op;
55 };
56 struct anon_vma_chain {
57 	struct anon_vma *anon_vma;
58 	struct list_head same_vma;
59 };
60 struct task_struct {
61 	char comm[TASK_COMM_LEN];
62 	pid_t pid;
63 	struct mm_struct *mm;
64 
65 	/* Used for emulating ABI behavior of previous Linux versions: */
66 	unsigned int			personality;
67 };
68 
69 struct kref {
70 	refcount_t refcount;
71 };
72 
73 struct anon_vma_name {
74 	struct kref kref;
75 	/* The name needs to be at the end because it is dynamically sized. */
76 	char name[];
77 };
78 
79 /*
80  * Contains declarations that are DUPLICATED from kernel source in order to
81  * faciliate userland VMA testing.
82  *
83  * These must be kept in sync with kernel source.
84  */
85 
86 #define VMA_LOCK_OFFSET	0x40000000
87 
88 typedef struct { unsigned long v; } freeptr_t;
89 
90 #define VM_NONE		0x00000000
91 
92 typedef int __bitwise vma_flag_t;
93 
94 #define ACCESS_PRIVATE(p, member) ((p)->member)
95 
96 #define DECLARE_VMA_BIT(name, bitnum) \
97 	VMA_ ## name ## _BIT = ((__force vma_flag_t)bitnum)
98 #define DECLARE_VMA_BIT_ALIAS(name, aliased) \
99 	VMA_ ## name ## _BIT = VMA_ ## aliased ## _BIT
100 enum {
101 	DECLARE_VMA_BIT(READ, 0),
102 	DECLARE_VMA_BIT(WRITE, 1),
103 	DECLARE_VMA_BIT(EXEC, 2),
104 	DECLARE_VMA_BIT(SHARED, 3),
105 	/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
106 	DECLARE_VMA_BIT(MAYREAD, 4),	/* limits for mprotect() etc. */
107 	DECLARE_VMA_BIT(MAYWRITE, 5),
108 	DECLARE_VMA_BIT(MAYEXEC, 6),
109 	DECLARE_VMA_BIT(MAYSHARE, 7),
110 	DECLARE_VMA_BIT(GROWSDOWN, 8),	/* general info on the segment */
111 #ifdef CONFIG_MMU
112 	DECLARE_VMA_BIT(UFFD_MISSING, 9),/* missing pages tracking */
113 #else
114 	/* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
115 	DECLARE_VMA_BIT(MAYOVERLAY, 9),
116 #endif /* CONFIG_MMU */
117 	/* Page-ranges managed without "struct page", just pure PFN */
118 	DECLARE_VMA_BIT(PFNMAP, 10),
119 	DECLARE_VMA_BIT(MAYBE_GUARD, 11),
120 	DECLARE_VMA_BIT(UFFD_WP, 12),	/* wrprotect pages tracking */
121 	DECLARE_VMA_BIT(LOCKED, 13),
122 	DECLARE_VMA_BIT(IO, 14),	/* Memory mapped I/O or similar */
123 	DECLARE_VMA_BIT(SEQ_READ, 15),	/* App will access data sequentially */
124 	DECLARE_VMA_BIT(RAND_READ, 16),	/* App will not benefit from clustered reads */
125 	DECLARE_VMA_BIT(DONTCOPY, 17),	/* Do not copy this vma on fork */
126 	DECLARE_VMA_BIT(DONTEXPAND, 18),/* Cannot expand with mremap() */
127 	DECLARE_VMA_BIT(LOCKONFAULT, 19),/* Lock pages covered when faulted in */
128 	DECLARE_VMA_BIT(ACCOUNT, 20),	/* Is a VM accounted object */
129 	DECLARE_VMA_BIT(NORESERVE, 21),	/* should the VM suppress accounting */
130 	DECLARE_VMA_BIT(HUGETLB, 22),	/* Huge TLB Page VM */
131 	DECLARE_VMA_BIT(SYNC, 23),	/* Synchronous page faults */
132 	DECLARE_VMA_BIT(ARCH_1, 24),	/* Architecture-specific flag */
133 	DECLARE_VMA_BIT(WIPEONFORK, 25),/* Wipe VMA contents in child. */
134 	DECLARE_VMA_BIT(DONTDUMP, 26),	/* Do not include in the core dump */
135 	DECLARE_VMA_BIT(SOFTDIRTY, 27),	/* NOT soft dirty clean area */
136 	DECLARE_VMA_BIT(MIXEDMAP, 28),	/* Can contain struct page and pure PFN pages */
137 	DECLARE_VMA_BIT(HUGEPAGE, 29),	/* MADV_HUGEPAGE marked this vma */
138 	DECLARE_VMA_BIT(NOHUGEPAGE, 30),/* MADV_NOHUGEPAGE marked this vma */
139 	DECLARE_VMA_BIT(MERGEABLE, 31),	/* KSM may merge identical pages */
140 	/* These bits are reused, we define specific uses below. */
141 	DECLARE_VMA_BIT(HIGH_ARCH_0, 32),
142 	DECLARE_VMA_BIT(HIGH_ARCH_1, 33),
143 	DECLARE_VMA_BIT(HIGH_ARCH_2, 34),
144 	DECLARE_VMA_BIT(HIGH_ARCH_3, 35),
145 	DECLARE_VMA_BIT(HIGH_ARCH_4, 36),
146 	DECLARE_VMA_BIT(HIGH_ARCH_5, 37),
147 	DECLARE_VMA_BIT(HIGH_ARCH_6, 38),
148 	/*
149 	 * This flag is used to connect VFIO to arch specific KVM code. It
150 	 * indicates that the memory under this VMA is safe for use with any
151 	 * non-cachable memory type inside KVM. Some VFIO devices, on some
152 	 * platforms, are thought to be unsafe and can cause machine crashes
153 	 * if KVM does not lock down the memory type.
154 	 */
155 	DECLARE_VMA_BIT(ALLOW_ANY_UNCACHED, 39),
156 #ifdef CONFIG_PPC32
157 	DECLARE_VMA_BIT_ALIAS(DROPPABLE, ARCH_1),
158 #else
159 	DECLARE_VMA_BIT(DROPPABLE, 40),
160 #endif
161 	DECLARE_VMA_BIT(UFFD_MINOR, 41),
162 	DECLARE_VMA_BIT(SEALED, 42),
163 	/* Flags that reuse flags above. */
164 	DECLARE_VMA_BIT_ALIAS(PKEY_BIT0, HIGH_ARCH_0),
165 	DECLARE_VMA_BIT_ALIAS(PKEY_BIT1, HIGH_ARCH_1),
166 	DECLARE_VMA_BIT_ALIAS(PKEY_BIT2, HIGH_ARCH_2),
167 	DECLARE_VMA_BIT_ALIAS(PKEY_BIT3, HIGH_ARCH_3),
168 	DECLARE_VMA_BIT_ALIAS(PKEY_BIT4, HIGH_ARCH_4),
169 #if defined(CONFIG_X86_USER_SHADOW_STACK)
170 	/*
171 	 * VM_SHADOW_STACK should not be set with VM_SHARED because of lack of
172 	 * support core mm.
173 	 *
174 	 * These VMAs will get a single end guard page. This helps userspace
175 	 * protect itself from attacks. A single page is enough for current
176 	 * shadow stack archs (x86). See the comments near alloc_shstk() in
177 	 * arch/x86/kernel/shstk.c for more details on the guard size.
178 	 */
179 	DECLARE_VMA_BIT_ALIAS(SHADOW_STACK, HIGH_ARCH_5),
180 #elif defined(CONFIG_ARM64_GCS)
181 	/*
182 	 * arm64's Guarded Control Stack implements similar functionality and
183 	 * has similar constraints to shadow stacks.
184 	 */
185 	DECLARE_VMA_BIT_ALIAS(SHADOW_STACK, HIGH_ARCH_6),
186 #endif
187 	DECLARE_VMA_BIT_ALIAS(SAO, ARCH_1),		/* Strong Access Ordering (powerpc) */
188 	DECLARE_VMA_BIT_ALIAS(GROWSUP, ARCH_1),		/* parisc */
189 	DECLARE_VMA_BIT_ALIAS(SPARC_ADI, ARCH_1),	/* sparc64 */
190 	DECLARE_VMA_BIT_ALIAS(ARM64_BTI, ARCH_1),	/* arm64 */
191 	DECLARE_VMA_BIT_ALIAS(ARCH_CLEAR, ARCH_1),	/* sparc64, arm64 */
192 	DECLARE_VMA_BIT_ALIAS(MAPPED_COPY, ARCH_1),	/* !CONFIG_MMU */
193 	DECLARE_VMA_BIT_ALIAS(MTE, HIGH_ARCH_4),	/* arm64 */
194 	DECLARE_VMA_BIT_ALIAS(MTE_ALLOWED, HIGH_ARCH_5),/* arm64 */
195 #ifdef CONFIG_STACK_GROWSUP
196 	DECLARE_VMA_BIT_ALIAS(STACK, GROWSUP),
197 	DECLARE_VMA_BIT_ALIAS(STACK_EARLY, GROWSDOWN),
198 #else
199 	DECLARE_VMA_BIT_ALIAS(STACK, GROWSDOWN),
200 #endif
201 };
202 
203 #define INIT_VM_FLAG(name) BIT((__force int) VMA_ ## name ## _BIT)
204 #define VM_READ		INIT_VM_FLAG(READ)
205 #define VM_WRITE	INIT_VM_FLAG(WRITE)
206 #define VM_EXEC		INIT_VM_FLAG(EXEC)
207 #define VM_SHARED	INIT_VM_FLAG(SHARED)
208 #define VM_MAYREAD	INIT_VM_FLAG(MAYREAD)
209 #define VM_MAYWRITE	INIT_VM_FLAG(MAYWRITE)
210 #define VM_MAYEXEC	INIT_VM_FLAG(MAYEXEC)
211 #define VM_MAYSHARE	INIT_VM_FLAG(MAYSHARE)
212 #define VM_GROWSDOWN	INIT_VM_FLAG(GROWSDOWN)
213 #ifdef CONFIG_MMU
214 #define VM_UFFD_MISSING	INIT_VM_FLAG(UFFD_MISSING)
215 #else
216 #define VM_UFFD_MISSING	VM_NONE
217 #define VM_MAYOVERLAY	INIT_VM_FLAG(MAYOVERLAY)
218 #endif
219 #define VM_PFNMAP	INIT_VM_FLAG(PFNMAP)
220 #define VM_MAYBE_GUARD	INIT_VM_FLAG(MAYBE_GUARD)
221 #define VM_UFFD_WP	INIT_VM_FLAG(UFFD_WP)
222 #define VM_LOCKED	INIT_VM_FLAG(LOCKED)
223 #define VM_IO		INIT_VM_FLAG(IO)
224 #define VM_SEQ_READ	INIT_VM_FLAG(SEQ_READ)
225 #define VM_RAND_READ	INIT_VM_FLAG(RAND_READ)
226 #define VM_DONTCOPY	INIT_VM_FLAG(DONTCOPY)
227 #define VM_DONTEXPAND	INIT_VM_FLAG(DONTEXPAND)
228 #define VM_LOCKONFAULT	INIT_VM_FLAG(LOCKONFAULT)
229 #define VM_ACCOUNT	INIT_VM_FLAG(ACCOUNT)
230 #define VM_NORESERVE	INIT_VM_FLAG(NORESERVE)
231 #define VM_HUGETLB	INIT_VM_FLAG(HUGETLB)
232 #define VM_SYNC		INIT_VM_FLAG(SYNC)
233 #define VM_ARCH_1	INIT_VM_FLAG(ARCH_1)
234 #define VM_WIPEONFORK	INIT_VM_FLAG(WIPEONFORK)
235 #define VM_DONTDUMP	INIT_VM_FLAG(DONTDUMP)
236 #ifdef CONFIG_MEM_SOFT_DIRTY
237 #define VM_SOFTDIRTY	INIT_VM_FLAG(SOFTDIRTY)
238 #else
239 #define VM_SOFTDIRTY	VM_NONE
240 #endif
241 #define VM_MIXEDMAP	INIT_VM_FLAG(MIXEDMAP)
242 #define VM_HUGEPAGE	INIT_VM_FLAG(HUGEPAGE)
243 #define VM_NOHUGEPAGE	INIT_VM_FLAG(NOHUGEPAGE)
244 #define VM_MERGEABLE	INIT_VM_FLAG(MERGEABLE)
245 #define VM_STACK	INIT_VM_FLAG(STACK)
246 #ifdef CONFIG_STACK_GROWS_UP
247 #define VM_STACK_EARLY	INIT_VM_FLAG(STACK_EARLY)
248 #else
249 #define VM_STACK_EARLY	VM_NONE
250 #endif
251 #ifdef CONFIG_ARCH_HAS_PKEYS
252 #define VM_PKEY_SHIFT ((__force int)VMA_HIGH_ARCH_0_BIT)
253 /* Despite the naming, these are FLAGS not bits. */
254 #define VM_PKEY_BIT0 INIT_VM_FLAG(PKEY_BIT0)
255 #define VM_PKEY_BIT1 INIT_VM_FLAG(PKEY_BIT1)
256 #define VM_PKEY_BIT2 INIT_VM_FLAG(PKEY_BIT2)
257 #if CONFIG_ARCH_PKEY_BITS > 3
258 #define VM_PKEY_BIT3 INIT_VM_FLAG(PKEY_BIT3)
259 #else
260 #define VM_PKEY_BIT3  VM_NONE
261 #endif /* CONFIG_ARCH_PKEY_BITS > 3 */
262 #if CONFIG_ARCH_PKEY_BITS > 4
263 #define VM_PKEY_BIT4 INIT_VM_FLAG(PKEY_BIT4)
264 #else
265 #define VM_PKEY_BIT4  VM_NONE
266 #endif /* CONFIG_ARCH_PKEY_BITS > 4 */
267 #endif /* CONFIG_ARCH_HAS_PKEYS */
268 #if defined(CONFIG_X86_USER_SHADOW_STACK) || defined(CONFIG_ARM64_GCS)
269 #define VM_SHADOW_STACK	INIT_VM_FLAG(SHADOW_STACK)
270 #define VMA_STARTGAP_FLAGS mk_vma_flags(VMA_GROWSDOWN_BIT, VMA_SHADOW_STACK_BIT)
271 #else
272 #define VM_SHADOW_STACK	VM_NONE
273 #define VMA_STARTGAP_FLAGS mk_vma_flags(VMA_GROWSDOWN_BIT)
274 #endif
275 #if defined(CONFIG_PPC64)
276 #define VM_SAO		INIT_VM_FLAG(SAO)
277 #elif defined(CONFIG_PARISC)
278 #define VM_GROWSUP	INIT_VM_FLAG(GROWSUP)
279 #elif defined(CONFIG_SPARC64)
280 #define VM_SPARC_ADI	INIT_VM_FLAG(SPARC_ADI)
281 #define VM_ARCH_CLEAR	INIT_VM_FLAG(ARCH_CLEAR)
282 #elif defined(CONFIG_ARM64)
283 #define VM_ARM64_BTI	INIT_VM_FLAG(ARM64_BTI)
284 #define VM_ARCH_CLEAR	INIT_VM_FLAG(ARCH_CLEAR)
285 #elif !defined(CONFIG_MMU)
286 #define VM_MAPPED_COPY	INIT_VM_FLAG(MAPPED_COPY)
287 #endif
288 #ifndef VM_GROWSUP
289 #define VM_GROWSUP	VM_NONE
290 #endif
291 #ifdef CONFIG_ARM64_MTE
292 #define VM_MTE		INIT_VM_FLAG(MTE)
293 #define VM_MTE_ALLOWED	INIT_VM_FLAG(MTE_ALLOWED)
294 #else
295 #define VM_MTE		VM_NONE
296 #define VM_MTE_ALLOWED	VM_NONE
297 #endif
298 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
299 #define VM_UFFD_MINOR	INIT_VM_FLAG(UFFD_MINOR)
300 #else
301 #define VM_UFFD_MINOR	VM_NONE
302 #endif
303 #ifdef CONFIG_64BIT
304 #define VM_ALLOW_ANY_UNCACHED	INIT_VM_FLAG(ALLOW_ANY_UNCACHED)
305 #define VM_SEALED		INIT_VM_FLAG(SEALED)
306 #else
307 #define VM_ALLOW_ANY_UNCACHED	VM_NONE
308 #define VM_SEALED		VM_NONE
309 #endif
310 #if defined(CONFIG_64BIT) || defined(CONFIG_PPC32)
311 #define VM_DROPPABLE		INIT_VM_FLAG(DROPPABLE)
312 #else
313 #define VM_DROPPABLE		VM_NONE
314 #endif
315 
316 /* Bits set in the VMA until the stack is in its final location */
317 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY)
318 
319 #define TASK_EXEC_BIT ((current->personality & READ_IMPLIES_EXEC) ? \
320 		       VM_EXEC_BIT : VM_READ_BIT)
321 
322 /* Common data flag combinations */
323 #define VMA_DATA_FLAGS_TSK_EXEC	mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, \
324 		TASK_EXEC_BIT, VMA_MAYREAD_BIT, VMA_MAYWRITE_BIT,	  \
325 		VMA_MAYEXEC_BIT)
326 #define VMA_DATA_FLAGS_NON_EXEC	mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, \
327 		VMA_MAYREAD_BIT, VMA_MAYWRITE_BIT, VMA_MAYEXEC_BIT)
328 #define VMA_DATA_FLAGS_EXEC	mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, \
329 		VMA_EXEC_BIT, VMA_MAYREAD_BIT, VMA_MAYWRITE_BIT,	  \
330 		VMA_MAYEXEC_BIT)
331 
332 #ifndef VMA_DATA_DEFAULT_FLAGS		/* arch can override this */
333 #define VMA_DATA_DEFAULT_FLAGS  VMA_DATA_FLAGS_EXEC
334 #endif
335 
336 #ifndef VMA_STACK_DEFAULT_FLAGS		/* arch can override this */
337 #define VMA_STACK_DEFAULT_FLAGS VMA_DATA_DEFAULT_FLAGS
338 #endif
339 
340 #define VMA_STACK_FLAGS	append_vma_flags(VMA_STACK_DEFAULT_FLAGS,	\
341 		VMA_STACK_BIT, VMA_ACCOUNT_BIT)
342 /* Temporary until VMA flags conversion complete. */
343 #define VM_STACK_FLAGS vma_flags_to_legacy(VMA_STACK_FLAGS)
344 
345 #define VM_STARTGAP_FLAGS (VM_GROWSDOWN | VM_SHADOW_STACK)
346 
347 /* VMA basic access permission flags */
348 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
349 #define VMA_ACCESS_FLAGS mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, VMA_EXEC_BIT)
350 
351 /*
352  * Special vmas that are non-mergable, non-mlock()able.
353  */
354 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
355 
356 #define VMA_SPECIAL_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_DONTEXPAND_BIT, \
357 				       VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT)
358 
359 #define VMA_REMAP_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_PFNMAP_BIT,	\
360 				     VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT)
361 
362 #define DEFAULT_MAP_WINDOW	((1UL << 47) - PAGE_SIZE)
363 #define TASK_SIZE_LOW		DEFAULT_MAP_WINDOW
364 #define TASK_SIZE_MAX		DEFAULT_MAP_WINDOW
365 #define STACK_TOP		TASK_SIZE_LOW
366 #define STACK_TOP_MAX		TASK_SIZE_MAX
367 
368 /* This mask represents all the VMA flag bits used by mlock */
369 #define VM_LOCKED_MASK	(VM_LOCKED | VM_LOCKONFAULT)
370 
371 #define VMA_LOCKED_MASK	mk_vma_flags(VMA_LOCKED_BIT, VMA_LOCKONFAULT_BIT)
372 
373 #define RLIMIT_STACK		3	/* max stack size */
374 #define RLIMIT_MEMLOCK		8	/* max locked-in-memory address space */
375 
376 #define CAP_IPC_LOCK         14
377 
378 #ifdef CONFIG_MEM_SOFT_DIRTY
379 #define VMA_STICKY_FLAGS mk_vma_flags(VMA_SOFTDIRTY_BIT, VMA_MAYBE_GUARD_BIT)
380 #else
381 #define VMA_STICKY_FLAGS mk_vma_flags(VMA_MAYBE_GUARD_BIT)
382 #endif
383 
384 #define VMA_IGNORE_MERGE_FLAGS VMA_STICKY_FLAGS
385 
386 #define VM_COPY_ON_FORK (VM_PFNMAP | VM_MIXEDMAP | VM_UFFD_WP | VM_MAYBE_GUARD)
387 
388 #define pgprot_val(x)		((x).pgprot)
389 #define __pgprot(x)		((pgprot_t) { (x) } )
390 
391 #define for_each_vma(__vmi, __vma)					\
392 	while (((__vma) = vma_next(&(__vmi))) != NULL)
393 
394 /* The MM code likes to work with exclusive end addresses */
395 #define for_each_vma_range(__vmi, __vma, __end)				\
396 	while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
397 
398 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
399 
400 #define PHYS_PFN(x)	((unsigned long)((x) >> PAGE_SHIFT))
401 
402 #define test_and_set_bit(nr, addr) __test_and_set_bit(nr, addr)
403 #define test_and_clear_bit(nr, addr) __test_and_clear_bit(nr, addr)
404 
405 #define AS_MM_ALL_LOCKS 2
406 
407 #define swap(a, b) \
408 	do { typeof(a) __tmp = (a); (a) = (b); (b) = __tmp; } while (0)
409 
410 /*
411  * Flags for bug emulation.
412  *
413  * These occupy the top three bytes.
414  */
415 enum {
416 	READ_IMPLIES_EXEC =	0x0400000,
417 };
418 
419 struct vma_iterator {
420 	struct ma_state mas;
421 };
422 
423 #define VMA_ITERATOR(name, __mm, __addr)				\
424 	struct vma_iterator name = {					\
425 		.mas = {						\
426 			.tree = &(__mm)->mm_mt,				\
427 			.index = __addr,				\
428 			.node = NULL,					\
429 			.status = ma_start,				\
430 		},							\
431 	}
432 
433 #define DEFINE_MUTEX(mutexname) \
434 	struct mutex mutexname = {}
435 
436 #define DECLARE_BITMAP(name, bits) \
437 	unsigned long name[BITS_TO_LONGS(bits)]
438 
439 #define EMPTY_VMA_FLAGS ((vma_flags_t){ })
440 
441 #define MAPCOUNT_ELF_CORE_MARGIN	(5)
442 #define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
443 
444 static __always_inline bool vma_flags_empty(const vma_flags_t *flags)
445 {
446 	const unsigned long *bitmap = flags->__vma_flags;
447 
448 	return bitmap_empty(bitmap, NUM_VMA_FLAG_BITS);
449 }
450 
451 /* What action should be taken after an .mmap_prepare call is complete? */
452 enum mmap_action_type {
453 	MMAP_NOTHING,		/* Mapping is complete, no further action. */
454 	MMAP_REMAP_PFN,		/* Remap PFN range. */
455 	MMAP_IO_REMAP_PFN,	/* I/O remap PFN range. */
456 };
457 
458 /*
459  * Describes an action an mmap_prepare hook can instruct to be taken to complete
460  * the mapping of a VMA. Specified in vm_area_desc.
461  */
462 struct mmap_action {
463 	union {
464 		/* Remap range. */
465 		struct {
466 			unsigned long start;
467 			unsigned long start_pfn;
468 			unsigned long size;
469 			pgprot_t pgprot;
470 		} remap;
471 	};
472 	enum mmap_action_type type;
473 
474 	/*
475 	 * If specified, this hook is invoked after the selected action has been
476 	 * successfully completed. Note that the VMA write lock still held.
477 	 *
478 	 * The absolute minimum ought to be done here.
479 	 *
480 	 * Returns 0 on success, or an error code.
481 	 */
482 	int (*success_hook)(const struct vm_area_struct *vma);
483 
484 	/*
485 	 * If specified, this hook is invoked when an error occurred when
486 	 * attempting the selection action.
487 	 *
488 	 * The hook can return an error code in order to filter the error, but
489 	 * it is not valid to clear the error here.
490 	 */
491 	int (*error_hook)(int err);
492 
493 	/*
494 	 * This should be set in rare instances where the operation required
495 	 * that the rmap should not be able to access the VMA until
496 	 * completely set up.
497 	 */
498 	bool hide_from_rmap_until_complete :1;
499 };
500 
501 /* Operations which modify VMAs. */
502 enum vma_operation {
503 	VMA_OP_SPLIT,
504 	VMA_OP_MERGE_UNFAULTED,
505 	VMA_OP_REMAP,
506 	VMA_OP_FORK,
507 };
508 
509 /*
510  * Describes a VMA that is about to be mmap()'ed. Drivers may choose to
511  * manipulate mutable fields which will cause those fields to be updated in the
512  * resultant VMA.
513  *
514  * Helper functions are not required for manipulating any field.
515  */
516 struct vm_area_desc {
517 	/* Immutable state. */
518 	const struct mm_struct *const mm;
519 	struct file *const file; /* May vary from vm_file in stacked callers. */
520 	unsigned long start;
521 	unsigned long end;
522 
523 	/* Mutable fields. Populated with initial state. */
524 	pgoff_t pgoff;
525 	struct file *vm_file;
526 	vma_flags_t vma_flags;
527 	pgprot_t page_prot;
528 
529 	/* Write-only fields. */
530 	const struct vm_operations_struct *vm_ops;
531 	void *private_data;
532 
533 	/* Take further action? */
534 	struct mmap_action action;
535 };
536 
537 struct vm_area_struct {
538 	/* The first cache line has the info for VMA tree walking. */
539 
540 	union {
541 		struct {
542 			/* VMA covers [vm_start; vm_end) addresses within mm */
543 			unsigned long vm_start;
544 			unsigned long vm_end;
545 		};
546 		freeptr_t vm_freeptr; /* Pointer used by SLAB_TYPESAFE_BY_RCU */
547 	};
548 
549 	struct mm_struct *vm_mm;	/* The address space we belong to. */
550 	pgprot_t vm_page_prot;          /* Access permissions of this VMA. */
551 
552 	/*
553 	 * Flags, see mm.h.
554 	 * To modify use vm_flags_{init|reset|set|clear|mod} functions.
555 	 */
556 	union {
557 		const vm_flags_t vm_flags;
558 		vma_flags_t flags;
559 	};
560 
561 #ifdef CONFIG_PER_VMA_LOCK
562 	/*
563 	 * Can only be written (using WRITE_ONCE()) while holding both:
564 	 *  - mmap_lock (in write mode)
565 	 *  - vm_refcnt bit at VMA_LOCK_OFFSET is set
566 	 * Can be read reliably while holding one of:
567 	 *  - mmap_lock (in read or write mode)
568 	 *  - vm_refcnt bit at VMA_LOCK_OFFSET is set or vm_refcnt > 1
569 	 * Can be read unreliably (using READ_ONCE()) for pessimistic bailout
570 	 * while holding nothing (except RCU to keep the VMA struct allocated).
571 	 *
572 	 * This sequence counter is explicitly allowed to overflow; sequence
573 	 * counter reuse can only lead to occasional unnecessary use of the
574 	 * slowpath.
575 	 */
576 	unsigned int vm_lock_seq;
577 #endif
578 
579 	/*
580 	 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
581 	 * list, after a COW of one of the file pages.	A MAP_SHARED vma
582 	 * can only be in the i_mmap tree.  An anonymous MAP_PRIVATE, stack
583 	 * or brk vma (with NULL file) can only be in an anon_vma list.
584 	 */
585 	struct list_head anon_vma_chain; /* Serialized by mmap_lock &
586 					  * page_table_lock */
587 	struct anon_vma *anon_vma;	/* Serialized by page_table_lock */
588 
589 	/* Function pointers to deal with this struct. */
590 	const struct vm_operations_struct *vm_ops;
591 
592 	/* Information about our backing store: */
593 	unsigned long vm_pgoff;		/* Offset (within vm_file) in PAGE_SIZE
594 					   units */
595 	struct file * vm_file;		/* File we map to (can be NULL). */
596 	void * vm_private_data;		/* was vm_pte (shared mem) */
597 
598 #ifdef CONFIG_SWAP
599 	atomic_long_t swap_readahead_info;
600 #endif
601 #ifndef CONFIG_MMU
602 	struct vm_region *vm_region;	/* NOMMU mapping region */
603 #endif
604 #ifdef CONFIG_NUMA
605 	struct mempolicy *vm_policy;	/* NUMA policy for the VMA */
606 #endif
607 #ifdef CONFIG_NUMA_BALANCING
608 	struct vma_numab_state *numab_state;	/* NUMA Balancing state */
609 #endif
610 #ifdef CONFIG_PER_VMA_LOCK
611 	/* Unstable RCU readers are allowed to read this. */
612 	refcount_t vm_refcnt;
613 #endif
614 	/*
615 	 * For areas with an address space and backing store,
616 	 * linkage into the address_space->i_mmap interval tree.
617 	 *
618 	 */
619 	struct {
620 		struct rb_node rb;
621 		unsigned long rb_subtree_last;
622 	} shared;
623 #ifdef CONFIG_ANON_VMA_NAME
624 	/*
625 	 * For private and shared anonymous mappings, a pointer to a null
626 	 * terminated string containing the name given to the vma, or NULL if
627 	 * unnamed. Serialized by mmap_lock. Use anon_vma_name to access.
628 	 */
629 	struct anon_vma_name *anon_name;
630 #endif
631 	struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
632 } __randomize_layout;
633 
634 struct vm_operations_struct {
635 	void (*open)(struct vm_area_struct * area);
636 	/**
637 	 * @close: Called when the VMA is being removed from the MM.
638 	 * Context: User context.  May sleep.  Caller holds mmap_lock.
639 	 */
640 	void (*close)(struct vm_area_struct * area);
641 	/* Called any time before splitting to check if it's allowed */
642 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
643 	int (*mremap)(struct vm_area_struct *area);
644 	/*
645 	 * Called by mprotect() to make driver-specific permission
646 	 * checks before mprotect() is finalised.   The VMA must not
647 	 * be modified.  Returns 0 if mprotect() can proceed.
648 	 */
649 	int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
650 			unsigned long end, unsigned long newflags);
651 	vm_fault_t (*fault)(struct vm_fault *vmf);
652 	vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
653 	vm_fault_t (*map_pages)(struct vm_fault *vmf,
654 			pgoff_t start_pgoff, pgoff_t end_pgoff);
655 	unsigned long (*pagesize)(struct vm_area_struct * area);
656 
657 	/* notification that a previously read-only page is about to become
658 	 * writable, if an error is returned it will cause a SIGBUS */
659 	vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
660 
661 	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
662 	vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
663 
664 	/* called by access_process_vm when get_user_pages() fails, typically
665 	 * for use by special VMAs. See also generic_access_phys() for a generic
666 	 * implementation useful for any iomem mapping.
667 	 */
668 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
669 		      void *buf, int len, int write);
670 
671 	/* Called by the /proc/PID/maps code to ask the vma whether it
672 	 * has a special name.  Returning non-NULL will also cause this
673 	 * vma to be dumped unconditionally. */
674 	const char *(*name)(struct vm_area_struct *vma);
675 
676 #ifdef CONFIG_NUMA
677 	/*
678 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
679 	 * to hold the policy upon return.  Caller should pass NULL @new to
680 	 * remove a policy and fall back to surrounding context--i.e. do not
681 	 * install a MPOL_DEFAULT policy, nor the task or system default
682 	 * mempolicy.
683 	 */
684 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
685 
686 	/*
687 	 * get_policy() op must add reference [mpol_get()] to any policy at
688 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
689 	 * in mm/mempolicy.c will do this automatically.
690 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
691 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
692 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
693 	 * must return NULL--i.e., do not "fallback" to task or system default
694 	 * policy.
695 	 */
696 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
697 					unsigned long addr, pgoff_t *ilx);
698 #endif
699 #ifdef CONFIG_FIND_NORMAL_PAGE
700 	/*
701 	 * Called by vm_normal_page() for special PTEs in @vma at @addr. This
702 	 * allows for returning a "normal" page from vm_normal_page() even
703 	 * though the PTE indicates that the "struct page" either does not exist
704 	 * or should not be touched: "special".
705 	 *
706 	 * Do not add new users: this really only works when a "normal" page
707 	 * was mapped, but then the PTE got changed to something weird (+
708 	 * marked special) that would not make pte_pfn() identify the originally
709 	 * inserted page.
710 	 */
711 	struct page *(*find_normal_page)(struct vm_area_struct *vma,
712 					 unsigned long addr);
713 #endif /* CONFIG_FIND_NORMAL_PAGE */
714 };
715 
716 struct vm_unmapped_area_info {
717 #define VM_UNMAPPED_AREA_TOPDOWN 1
718 	unsigned long flags;
719 	unsigned long length;
720 	unsigned long low_limit;
721 	unsigned long high_limit;
722 	unsigned long align_mask;
723 	unsigned long align_offset;
724 	unsigned long start_gap;
725 };
726 
727 struct pagetable_move_control {
728 	struct vm_area_struct *old; /* Source VMA. */
729 	struct vm_area_struct *new; /* Destination VMA. */
730 	unsigned long old_addr; /* Address from which the move begins. */
731 	unsigned long old_end; /* Exclusive address at which old range ends. */
732 	unsigned long new_addr; /* Address to move page tables to. */
733 	unsigned long len_in; /* Bytes to remap specified by user. */
734 
735 	bool need_rmap_locks; /* Do rmap locks need to be taken? */
736 	bool for_stack; /* Is this an early temp stack being moved? */
737 };
738 
739 #define PAGETABLE_MOVE(name, old_, new_, old_addr_, new_addr_, len_)	\
740 	struct pagetable_move_control name = {				\
741 		.old = old_,						\
742 		.new = new_,						\
743 		.old_addr = old_addr_,					\
744 		.old_end = (old_addr_) + (len_),			\
745 		.new_addr = new_addr_,					\
746 		.len_in = len_,						\
747 	}
748 
749 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
750 {
751 	mas_pause(&vmi->mas);
752 }
753 
754 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
755 {
756 	return __pgprot(pgprot_val(oldprot) | pgprot_val(newprot));
757 }
758 
759 static inline pgprot_t vm_get_page_prot(vm_flags_t vm_flags)
760 {
761 	return __pgprot(vm_flags);
762 }
763 
764 static inline bool mm_flags_test(int flag, const struct mm_struct *mm)
765 {
766 	return test_bit(flag, ACCESS_PRIVATE(&mm->flags, __mm_flags));
767 }
768 
769 /*
770  * Copy value to the first system word of VMA flags, non-atomically.
771  *
772  * IMPORTANT: This does not overwrite bytes past the first system word. The
773  * caller must account for this.
774  */
775 static __always_inline void vma_flags_overwrite_word(vma_flags_t *flags,
776 		unsigned long value)
777 {
778 	unsigned long *bitmap = flags->__vma_flags;
779 
780 	bitmap[0] = value;
781 }
782 
783 /*
784  * Copy value to the first system word of VMA flags ONCE, non-atomically.
785  *
786  * IMPORTANT: This does not overwrite bytes past the first system word. The
787  * caller must account for this.
788  */
789 static __always_inline void vma_flags_overwrite_word_once(vma_flags_t *flags,
790 		unsigned long value)
791 {
792 	unsigned long *bitmap = flags->__vma_flags;
793 
794 	WRITE_ONCE(*bitmap, value);
795 }
796 
797 /* Update the first system word of VMA flags setting bits, non-atomically. */
798 static __always_inline void vma_flags_set_word(vma_flags_t *flags,
799 		unsigned long value)
800 {
801 	unsigned long *bitmap = flags->__vma_flags;
802 
803 	*bitmap |= value;
804 }
805 
806 /* Update the first system word of VMA flags clearing bits, non-atomically. */
807 static __always_inline void vma_flags_clear_word(vma_flags_t *flags,
808 		unsigned long value)
809 {
810 	unsigned long *bitmap = flags->__vma_flags;
811 
812 	*bitmap &= ~value;
813 }
814 
815 static __always_inline void vma_flags_clear_all(vma_flags_t *flags)
816 {
817 	bitmap_zero(ACCESS_PRIVATE(flags, __vma_flags), NUM_VMA_FLAG_BITS);
818 }
819 
820 /*
821  * Helper function which converts a vma_flags_t value to a legacy vm_flags_t
822  * value. This is only valid if the input flags value can be expressed in a
823  * system word.
824  *
825  * Will be removed once the conversion to VMA flags is complete.
826  */
827 static __always_inline vm_flags_t vma_flags_to_legacy(vma_flags_t flags)
828 {
829 	return (vm_flags_t)flags.__vma_flags[0];
830 }
831 
832 /*
833  * Helper function which converts a legacy vm_flags_t value to a vma_flags_t
834  * value.
835  *
836  * Will be removed once the conversion to VMA flags is complete.
837  */
838 static __always_inline vma_flags_t legacy_to_vma_flags(vm_flags_t flags)
839 {
840 	vma_flags_t ret = EMPTY_VMA_FLAGS;
841 
842 	vma_flags_overwrite_word(&ret, flags);
843 	return ret;
844 }
845 
846 static __always_inline void vma_flags_set_flag(vma_flags_t *flags,
847 		vma_flag_t bit)
848 {
849 	unsigned long *bitmap = ACCESS_PRIVATE(flags, __vma_flags);
850 
851 	__set_bit((__force int)bit, bitmap);
852 }
853 
854 /* Use when VMA is not part of the VMA tree and needs no locking */
855 static inline void vm_flags_init(struct vm_area_struct *vma,
856 				 vm_flags_t flags)
857 {
858 	vma_flags_clear_all(&vma->flags);
859 	vma_flags_overwrite_word(&vma->flags, flags);
860 }
861 
862 /*
863  * Use when VMA is part of the VMA tree and modifications need coordination
864  * Note: vm_flags_reset and vm_flags_reset_once do not lock the vma and
865  * it should be locked explicitly beforehand.
866  */
867 static inline void vm_flags_reset(struct vm_area_struct *vma,
868 				  vm_flags_t flags)
869 {
870 	vma_assert_write_locked(vma);
871 	vm_flags_init(vma, flags);
872 }
873 
874 static inline void vma_flags_reset_once(struct vm_area_struct *vma,
875 					vma_flags_t *flags)
876 {
877 	const unsigned long word = flags->__vma_flags[0];
878 
879 	/* It is assumed only the first system word must be written once. */
880 	vma_flags_overwrite_word_once(&vma->flags, word);
881 	/* The remainder can be copied normally. */
882 	if (NUM_VMA_FLAG_BITS > BITS_PER_LONG) {
883 		unsigned long *dst = &vma->flags.__vma_flags[1];
884 		const unsigned long *src = &flags->__vma_flags[1];
885 
886 		bitmap_copy(dst, src, NUM_VMA_FLAG_BITS - BITS_PER_LONG);
887 	}
888 }
889 
890 static inline void vm_flags_set(struct vm_area_struct *vma,
891 				vm_flags_t flags)
892 {
893 	vma_start_write(vma);
894 	vma_flags_set_word(&vma->flags, flags);
895 }
896 
897 static inline void vm_flags_clear(struct vm_area_struct *vma,
898 				  vm_flags_t flags)
899 {
900 	vma_start_write(vma);
901 	vma_flags_clear_word(&vma->flags, flags);
902 }
903 
904 static __always_inline vma_flags_t __mk_vma_flags(vma_flags_t flags,
905 		size_t count, const vma_flag_t *bits)
906 {
907 	int i;
908 
909 	for (i = 0; i < count; i++)
910 		vma_flags_set_flag(&flags, bits[i]);
911 	return flags;
912 }
913 
914 #define mk_vma_flags(...) __mk_vma_flags(EMPTY_VMA_FLAGS,			\
915 		COUNT_ARGS(__VA_ARGS__), (const vma_flag_t []){__VA_ARGS__})
916 
917 #define append_vma_flags(flags, ...) __mk_vma_flags(flags,			\
918 		COUNT_ARGS(__VA_ARGS__), (const vma_flag_t []){__VA_ARGS__})
919 
920 static __always_inline int vma_flags_count(const vma_flags_t *flags)
921 {
922 	const unsigned long *bitmap = flags->__vma_flags;
923 
924 	return bitmap_weight(bitmap, NUM_VMA_FLAG_BITS);
925 }
926 
927 static __always_inline bool vma_flags_test(const vma_flags_t *flags,
928 		vma_flag_t bit)
929 {
930 	const unsigned long *bitmap = flags->__vma_flags;
931 
932 	return test_bit((__force int)bit, bitmap);
933 }
934 
935 static __always_inline vma_flags_t vma_flags_and_mask(const vma_flags_t *flags,
936 						      vma_flags_t to_and)
937 {
938 	vma_flags_t dst;
939 	unsigned long *bitmap_dst = dst.__vma_flags;
940 	const unsigned long *bitmap = flags->__vma_flags;
941 	const unsigned long *bitmap_to_and = to_and.__vma_flags;
942 
943 	bitmap_and(bitmap_dst, bitmap, bitmap_to_and, NUM_VMA_FLAG_BITS);
944 	return dst;
945 }
946 
947 #define vma_flags_and(flags, ...)		\
948 	vma_flags_and_mask(flags, mk_vma_flags(__VA_ARGS__))
949 
950 static __always_inline bool vma_flags_test_any_mask(const vma_flags_t *flags,
951 		vma_flags_t to_test)
952 {
953 	const unsigned long *bitmap = flags->__vma_flags;
954 	const unsigned long *bitmap_to_test = to_test.__vma_flags;
955 
956 	return bitmap_intersects(bitmap_to_test, bitmap, NUM_VMA_FLAG_BITS);
957 }
958 
959 #define vma_flags_test_any(flags, ...) \
960 	vma_flags_test_any_mask(flags, mk_vma_flags(__VA_ARGS__))
961 
962 static __always_inline bool vma_flags_test_all_mask(const vma_flags_t *flags,
963 		vma_flags_t to_test)
964 {
965 	const unsigned long *bitmap = flags->__vma_flags;
966 	const unsigned long *bitmap_to_test = to_test.__vma_flags;
967 
968 	return bitmap_subset(bitmap_to_test, bitmap, NUM_VMA_FLAG_BITS);
969 }
970 
971 #define vma_flags_test_all(flags, ...) \
972 	vma_flags_test_all_mask(flags, mk_vma_flags(__VA_ARGS__))
973 
974 static __always_inline bool vma_flags_test_single_mask(const vma_flags_t *flags,
975 						vma_flags_t flagmask)
976 {
977 	VM_WARN_ON_ONCE(vma_flags_count(&flagmask) > 1);
978 
979 	return vma_flags_test_any_mask(flags, flagmask);
980 }
981 
982 static __always_inline void vma_flags_set_mask(vma_flags_t *flags, vma_flags_t to_set)
983 {
984 	unsigned long *bitmap = flags->__vma_flags;
985 	const unsigned long *bitmap_to_set = to_set.__vma_flags;
986 
987 	bitmap_or(bitmap, bitmap, bitmap_to_set, NUM_VMA_FLAG_BITS);
988 }
989 
990 #define vma_flags_set(flags, ...) \
991 	vma_flags_set_mask(flags, mk_vma_flags(__VA_ARGS__))
992 
993 static __always_inline void vma_flags_clear_mask(vma_flags_t *flags, vma_flags_t to_clear)
994 {
995 	unsigned long *bitmap = flags->__vma_flags;
996 	const unsigned long *bitmap_to_clear = to_clear.__vma_flags;
997 
998 	bitmap_andnot(bitmap, bitmap, bitmap_to_clear, NUM_VMA_FLAG_BITS);
999 }
1000 
1001 #define vma_flags_clear(flags, ...) \
1002 	vma_flags_clear_mask(flags, mk_vma_flags(__VA_ARGS__))
1003 
1004 static __always_inline vma_flags_t vma_flags_diff_pair(const vma_flags_t *flags,
1005 		const vma_flags_t *flags_other)
1006 {
1007 	vma_flags_t dst;
1008 	const unsigned long *bitmap_other = flags_other->__vma_flags;
1009 	const unsigned long *bitmap = flags->__vma_flags;
1010 	unsigned long *bitmap_dst = dst.__vma_flags;
1011 
1012 	bitmap_xor(bitmap_dst, bitmap, bitmap_other, NUM_VMA_FLAG_BITS);
1013 	return dst;
1014 }
1015 
1016 static __always_inline bool vma_flags_same_pair(const vma_flags_t *flags,
1017 						const vma_flags_t *flags_other)
1018 {
1019 	const unsigned long *bitmap = flags->__vma_flags;
1020 	const unsigned long *bitmap_other = flags_other->__vma_flags;
1021 
1022 	return bitmap_equal(bitmap, bitmap_other, NUM_VMA_FLAG_BITS);
1023 }
1024 
1025 static __always_inline bool vma_flags_same_mask(const vma_flags_t *flags,
1026 						vma_flags_t flags_other)
1027 {
1028 	const unsigned long *bitmap = flags->__vma_flags;
1029 	const unsigned long *bitmap_other = flags_other.__vma_flags;
1030 
1031 	return bitmap_equal(bitmap, bitmap_other, NUM_VMA_FLAG_BITS);
1032 }
1033 
1034 #define vma_flags_same(flags, ...) \
1035 	vma_flags_same_mask(flags, mk_vma_flags(__VA_ARGS__))
1036 
1037 static __always_inline bool vma_test(const struct vm_area_struct *vma,
1038 		vma_flag_t bit)
1039 {
1040 	return vma_flags_test(&vma->flags, bit);
1041 }
1042 
1043 static __always_inline bool vma_test_any_mask(const struct vm_area_struct *vma,
1044 		vma_flags_t flags)
1045 {
1046 	return vma_flags_test_any_mask(&vma->flags, flags);
1047 }
1048 
1049 #define vma_test_any(vma, ...) \
1050 	vma_test_any_mask(vma, mk_vma_flags(__VA_ARGS__))
1051 
1052 static __always_inline bool vma_test_all_mask(const struct vm_area_struct *vma,
1053 		vma_flags_t flags)
1054 {
1055 	return vma_flags_test_all_mask(&vma->flags, flags);
1056 }
1057 
1058 #define vma_test_all(vma, ...) \
1059 	vma_test_all_mask(vma, mk_vma_flags(__VA_ARGS__))
1060 
1061 static __always_inline bool
1062 vma_test_single_mask(const struct vm_area_struct *vma, vma_flags_t flagmask)
1063 {
1064 	return vma_flags_test_single_mask(&vma->flags, flagmask);
1065 }
1066 
1067 static __always_inline void vma_set_flags_mask(struct vm_area_struct *vma,
1068 		vma_flags_t flags)
1069 {
1070 	vma_flags_set_mask(&vma->flags, flags);
1071 }
1072 
1073 #define vma_set_flags(vma, ...) \
1074 	vma_set_flags_mask(vma, mk_vma_flags(__VA_ARGS__))
1075 
1076 static __always_inline void vma_clear_flags_mask(struct vm_area_struct *vma,
1077 		vma_flags_t flags)
1078 {
1079 	vma_flags_clear_mask(&vma->flags, flags);
1080 }
1081 
1082 #define vma_clear_flags(vma, ...) \
1083 	vma_clear_flags_mask(vma, mk_vma_flags(__VA_ARGS__))
1084 
1085 static __always_inline bool vma_desc_test(const struct vm_area_desc *desc,
1086 		vma_flag_t bit)
1087 {
1088 	return vma_flags_test(&desc->vma_flags, bit);
1089 }
1090 
1091 static __always_inline bool vma_desc_test_any_mask(const struct vm_area_desc *desc,
1092 		vma_flags_t flags)
1093 {
1094 	return vma_flags_test_any_mask(&desc->vma_flags, flags);
1095 }
1096 
1097 #define vma_desc_test_any(desc, ...) \
1098 	vma_desc_test_any_mask(desc, mk_vma_flags(__VA_ARGS__))
1099 
1100 static __always_inline bool vma_desc_test_all_mask(const struct vm_area_desc *desc,
1101 		vma_flags_t flags)
1102 {
1103 	return vma_flags_test_all_mask(&desc->vma_flags, flags);
1104 }
1105 
1106 #define vma_desc_test_all(desc, ...) \
1107 	vma_desc_test_all_mask(desc, mk_vma_flags(__VA_ARGS__))
1108 
1109 static __always_inline void vma_desc_set_flags_mask(struct vm_area_desc *desc,
1110 		vma_flags_t flags)
1111 {
1112 	vma_flags_set_mask(&desc->vma_flags, flags);
1113 }
1114 
1115 #define vma_desc_set_flags(desc, ...) \
1116 	vma_desc_set_flags_mask(desc, mk_vma_flags(__VA_ARGS__))
1117 
1118 static __always_inline void vma_desc_clear_flags_mask(struct vm_area_desc *desc,
1119 		vma_flags_t flags)
1120 {
1121 	vma_flags_clear_mask(&desc->vma_flags, flags);
1122 }
1123 
1124 #define vma_desc_clear_flags(desc, ...) \
1125 	vma_desc_clear_flags_mask(desc, mk_vma_flags(__VA_ARGS__))
1126 
1127 static inline bool is_shared_maywrite(const vma_flags_t *flags)
1128 {
1129 	return vma_flags_test_all(flags, VMA_SHARED_BIT, VMA_MAYWRITE_BIT);
1130 }
1131 
1132 static inline bool vma_is_shared_maywrite(struct vm_area_struct *vma)
1133 {
1134 	return is_shared_maywrite(&vma->flags);
1135 }
1136 
1137 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
1138 {
1139 	/*
1140 	 * Uses mas_find() to get the first VMA when the iterator starts.
1141 	 * Calling mas_next() could skip the first entry.
1142 	 */
1143 	return mas_find(&vmi->mas, ULONG_MAX);
1144 }
1145 
1146 /*
1147  * WARNING: to avoid racing with vma_mark_attached()/vma_mark_detached(), these
1148  * assertions should be made either under mmap_write_lock or when the object
1149  * has been isolated under mmap_write_lock, ensuring no competing writers.
1150  */
1151 static inline void vma_assert_attached(struct vm_area_struct *vma)
1152 {
1153 	WARN_ON_ONCE(!refcount_read(&vma->vm_refcnt));
1154 }
1155 
1156 static inline void vma_assert_detached(struct vm_area_struct *vma)
1157 {
1158 	WARN_ON_ONCE(refcount_read(&vma->vm_refcnt));
1159 }
1160 
1161 static inline void vma_assert_write_locked(struct vm_area_struct *);
1162 static inline void vma_mark_attached(struct vm_area_struct *vma)
1163 {
1164 	vma_assert_write_locked(vma);
1165 	vma_assert_detached(vma);
1166 	refcount_set_release(&vma->vm_refcnt, 1);
1167 }
1168 
1169 static inline void vma_mark_detached(struct vm_area_struct *vma)
1170 {
1171 	vma_assert_write_locked(vma);
1172 	vma_assert_attached(vma);
1173 	/* We are the only writer, so no need to use vma_refcount_put(). */
1174 	if (unlikely(!refcount_dec_and_test(&vma->vm_refcnt))) {
1175 		/*
1176 		 * Reader must have temporarily raised vm_refcnt but it will
1177 		 * drop it without using the vma since vma is write-locked.
1178 		 */
1179 	}
1180 }
1181 
1182 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
1183 {
1184 	memset(vma, 0, sizeof(*vma));
1185 	vma->vm_mm = mm;
1186 	vma->vm_ops = &vma_dummy_vm_ops;
1187 	INIT_LIST_HEAD(&vma->anon_vma_chain);
1188 	vma->vm_lock_seq = UINT_MAX;
1189 }
1190 
1191 /*
1192  * These are defined in vma.h, but sadly vm_stat_account() is referenced by
1193  * kernel/fork.c, so we have to these broadly available there, and temporarily
1194  * define them here to resolve the dependency cycle.
1195  */
1196 #define is_exec_mapping(flags) \
1197 	((flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC)
1198 
1199 #define is_stack_mapping(flags) \
1200 	(((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK))
1201 
1202 #define is_data_mapping(flags) \
1203 	((flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE)
1204 
1205 static inline void vm_stat_account(struct mm_struct *mm, vm_flags_t flags,
1206 				   long npages)
1207 {
1208 	WRITE_ONCE(mm->total_vm, READ_ONCE(mm->total_vm)+npages);
1209 
1210 	if (is_exec_mapping(flags))
1211 		mm->exec_vm += npages;
1212 	else if (is_stack_mapping(flags))
1213 		mm->stack_vm += npages;
1214 	else if (is_data_mapping(flags))
1215 		mm->data_vm += npages;
1216 }
1217 
1218 #undef is_exec_mapping
1219 #undef is_stack_mapping
1220 #undef is_data_mapping
1221 
1222 static inline void vm_unacct_memory(long pages)
1223 {
1224 	vm_acct_memory(-pages);
1225 }
1226 
1227 static inline void mapping_allow_writable(struct address_space *mapping)
1228 {
1229 	atomic_inc(&mapping->i_mmap_writable);
1230 }
1231 
1232 static inline
1233 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
1234 {
1235 	return mas_find(&vmi->mas, max - 1);
1236 }
1237 
1238 static inline int vma_iter_clear_gfp(struct vma_iterator *vmi,
1239 			unsigned long start, unsigned long end, gfp_t gfp)
1240 {
1241 	__mas_set_range(&vmi->mas, start, end - 1);
1242 	mas_store_gfp(&vmi->mas, NULL, gfp);
1243 	if (unlikely(mas_is_err(&vmi->mas)))
1244 		return -ENOMEM;
1245 
1246 	return 0;
1247 }
1248 
1249 static inline void vma_set_anonymous(struct vm_area_struct *vma)
1250 {
1251 	vma->vm_ops = NULL;
1252 }
1253 
1254 /* Declared in vma.h. */
1255 static inline void set_vma_from_desc(struct vm_area_struct *vma,
1256 		struct vm_area_desc *desc);
1257 
1258 static inline int __compat_vma_mmap(const struct file_operations *f_op,
1259 		struct file *file, struct vm_area_struct *vma)
1260 {
1261 	struct vm_area_desc desc = {
1262 		.mm = vma->vm_mm,
1263 		.file = file,
1264 		.start = vma->vm_start,
1265 		.end = vma->vm_end,
1266 
1267 		.pgoff = vma->vm_pgoff,
1268 		.vm_file = vma->vm_file,
1269 		.vma_flags = vma->flags,
1270 		.page_prot = vma->vm_page_prot,
1271 
1272 		.action.type = MMAP_NOTHING, /* Default */
1273 	};
1274 	int err;
1275 
1276 	err = f_op->mmap_prepare(&desc);
1277 	if (err)
1278 		return err;
1279 
1280 	mmap_action_prepare(&desc.action, &desc);
1281 	set_vma_from_desc(vma, &desc);
1282 	return mmap_action_complete(&desc.action, vma);
1283 }
1284 
1285 static inline int compat_vma_mmap(struct file *file,
1286 		struct vm_area_struct *vma)
1287 {
1288 	return __compat_vma_mmap(file->f_op, file, vma);
1289 }
1290 
1291 
1292 static inline void vma_iter_init(struct vma_iterator *vmi,
1293 		struct mm_struct *mm, unsigned long addr)
1294 {
1295 	mas_init(&vmi->mas, &mm->mm_mt, addr);
1296 }
1297 
1298 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1299 {
1300 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1301 }
1302 
1303 static inline void mmap_assert_locked(struct mm_struct *);
1304 static inline struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
1305 						unsigned long start_addr,
1306 						unsigned long end_addr)
1307 {
1308 	unsigned long index = start_addr;
1309 
1310 	mmap_assert_locked(mm);
1311 	return mt_find(&mm->mm_mt, &index, end_addr - 1);
1312 }
1313 
1314 static inline
1315 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
1316 {
1317 	return mtree_load(&mm->mm_mt, addr);
1318 }
1319 
1320 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
1321 {
1322 	return mas_prev(&vmi->mas, 0);
1323 }
1324 
1325 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
1326 {
1327 	mas_set(&vmi->mas, addr);
1328 }
1329 
1330 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1331 {
1332 	return !vma->vm_ops;
1333 }
1334 
1335 /* Defined in vma.h, so temporarily define here to avoid circular dependency. */
1336 #define vma_iter_load(vmi) \
1337 	mas_walk(&(vmi)->mas)
1338 
1339 static inline struct vm_area_struct *
1340 find_vma_prev(struct mm_struct *mm, unsigned long addr,
1341 			struct vm_area_struct **pprev)
1342 {
1343 	struct vm_area_struct *vma;
1344 	VMA_ITERATOR(vmi, mm, addr);
1345 
1346 	vma = vma_iter_load(&vmi);
1347 	*pprev = vma_prev(&vmi);
1348 	if (!vma)
1349 		vma = vma_next(&vmi);
1350 	return vma;
1351 }
1352 
1353 #undef vma_iter_load
1354 
1355 static inline void vma_iter_free(struct vma_iterator *vmi)
1356 {
1357 	mas_destroy(&vmi->mas);
1358 }
1359 
1360 static inline
1361 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
1362 {
1363 	return mas_next_range(&vmi->mas, ULONG_MAX);
1364 }
1365 
1366 bool vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1367 
1368 /* Update vma->vm_page_prot to reflect vma->vm_flags. */
1369 static inline void vma_set_page_prot(struct vm_area_struct *vma)
1370 {
1371 	vm_flags_t vm_flags = vma->vm_flags;
1372 	pgprot_t vm_page_prot;
1373 
1374 	/* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1375 	vm_page_prot = pgprot_modify(vma->vm_page_prot, vm_get_page_prot(vm_flags));
1376 
1377 	if (vma_wants_writenotify(vma, vm_page_prot)) {
1378 		vm_flags &= ~VM_SHARED;
1379 		/* testing: we inline vm_pgprot_modify() to avoid clash with vma.h. */
1380 		vm_page_prot = pgprot_modify(vm_page_prot, vm_get_page_prot(vm_flags));
1381 	}
1382 	/* remove_protection_ptes reads vma->vm_page_prot without mmap_lock */
1383 	WRITE_ONCE(vma->vm_page_prot, vm_page_prot);
1384 }
1385 
1386 static inline unsigned long stack_guard_start_gap(struct vm_area_struct *vma)
1387 {
1388 	if (vma->vm_flags & VM_GROWSDOWN)
1389 		return stack_guard_gap;
1390 
1391 	/* See reasoning around the VM_SHADOW_STACK definition */
1392 	if (vma->vm_flags & VM_SHADOW_STACK)
1393 		return PAGE_SIZE;
1394 
1395 	return 0;
1396 }
1397 
1398 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
1399 {
1400 	unsigned long gap = stack_guard_start_gap(vma);
1401 	unsigned long vm_start = vma->vm_start;
1402 
1403 	vm_start -= gap;
1404 	if (vm_start > vma->vm_start)
1405 		vm_start = 0;
1406 	return vm_start;
1407 }
1408 
1409 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
1410 {
1411 	unsigned long vm_end = vma->vm_end;
1412 
1413 	if (vma->vm_flags & VM_GROWSUP) {
1414 		vm_end += stack_guard_gap;
1415 		if (vm_end < vma->vm_end)
1416 			vm_end = -PAGE_SIZE;
1417 	}
1418 	return vm_end;
1419 }
1420 
1421 static inline bool vma_is_accessible(struct vm_area_struct *vma)
1422 {
1423 	return vma->vm_flags & VM_ACCESS_FLAGS;
1424 }
1425 
1426 static inline bool mlock_future_ok(const struct mm_struct *mm,
1427 		vm_flags_t vm_flags, unsigned long bytes)
1428 {
1429 	unsigned long locked_pages, limit_pages;
1430 
1431 	if (!(vm_flags & VM_LOCKED) || capable(CAP_IPC_LOCK))
1432 		return true;
1433 
1434 	locked_pages = bytes >> PAGE_SHIFT;
1435 	locked_pages += mm->locked_vm;
1436 
1437 	limit_pages = rlimit(RLIMIT_MEMLOCK);
1438 	limit_pages >>= PAGE_SHIFT;
1439 
1440 	return locked_pages <= limit_pages;
1441 }
1442 
1443 static inline int mapping_map_writable(struct address_space *mapping)
1444 {
1445 	return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
1446 		0 : -EPERM;
1447 }
1448 
1449 /* Did the driver provide valid mmap hook configuration? */
1450 static inline bool can_mmap_file(struct file *file)
1451 {
1452 	bool has_mmap = file->f_op->mmap;
1453 	bool has_mmap_prepare = file->f_op->mmap_prepare;
1454 
1455 	/* Hooks are mutually exclusive. */
1456 	if (WARN_ON_ONCE(has_mmap && has_mmap_prepare))
1457 		return false;
1458 	if (!has_mmap && !has_mmap_prepare)
1459 		return false;
1460 
1461 	return true;
1462 }
1463 
1464 static inline int vfs_mmap(struct file *file, struct vm_area_struct *vma)
1465 {
1466 	if (file->f_op->mmap_prepare)
1467 		return compat_vma_mmap(file, vma);
1468 
1469 	return file->f_op->mmap(file, vma);
1470 }
1471 
1472 static inline int vfs_mmap_prepare(struct file *file, struct vm_area_desc *desc)
1473 {
1474 	return file->f_op->mmap_prepare(desc);
1475 }
1476 
1477 static inline void vma_set_file(struct vm_area_struct *vma, struct file *file)
1478 {
1479 	/* Changing an anonymous vma with this is illegal */
1480 	get_file(file);
1481 	swap(vma->vm_file, file);
1482 	fput(file);
1483 }
1484 
1485 extern int sysctl_max_map_count;
1486 static inline int get_sysctl_max_map_count(void)
1487 {
1488 	return READ_ONCE(sysctl_max_map_count);
1489 }
1490 
1491 #ifndef pgtable_supports_soft_dirty
1492 #define pgtable_supports_soft_dirty()	IS_ENABLED(CONFIG_MEM_SOFT_DIRTY)
1493 #endif
1494 
1495 static inline pgprot_t vma_get_page_prot(vma_flags_t vma_flags)
1496 {
1497 	const vm_flags_t vm_flags = vma_flags_to_legacy(vma_flags);
1498 
1499 	return vm_get_page_prot(vm_flags);
1500 }
1501