xref: /freebsd/sys/vm/vm_map.h (revision 6b2c1e49da284f28ec7b52f7c031474087e37104)
1 /*-
2  * SPDX-License-Identifier: (BSD-3-Clause AND MIT-CMU)
3  *
4  * Copyright (c) 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * The Mach Operating System project at Carnegie-Mellon University.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vm_map.h	8.9 (Berkeley) 5/17/95
35  *
36  *
37  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
38  * All rights reserved.
39  *
40  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
41  *
42  * Permission to use, copy, modify and distribute this software and
43  * its documentation is hereby granted, provided that both the copyright
44  * notice and this permission notice appear in all copies of the
45  * software, derivative works or modified versions, and any portions
46  * thereof, and that both notices appear in supporting documentation.
47  *
48  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
49  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
50  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
51  *
52  * Carnegie Mellon requests users of this software to return to
53  *
54  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
55  *  School of Computer Science
56  *  Carnegie Mellon University
57  *  Pittsburgh PA 15213-3890
58  *
59  * any improvements or extensions that they make and grant Carnegie the
60  * rights to redistribute these changes.
61  *
62  * $FreeBSD$
63  */
64 
65 /*
66  *	Virtual memory map module definitions.
67  */
68 #ifndef	_VM_MAP_
69 #define	_VM_MAP_
70 
71 #include <sys/lock.h>
72 #include <sys/sx.h>
73 #include <sys/_mutex.h>
74 
75 /*
76  *	Types defined:
77  *
78  *	vm_map_t		the high-level address map data structure.
79  *	vm_map_entry_t		an entry in an address map.
80  */
81 
82 typedef u_char vm_flags_t;
83 typedef u_int vm_eflags_t;
84 
85 /*
86  *	Objects which live in maps may be either VM objects, or
87  *	another map (called a "sharing map") which denotes read-write
88  *	sharing with other maps.
89  */
90 union vm_map_object {
91 	struct vm_object *vm_object;	/* object object */
92 	struct vm_map *sub_map;		/* belongs to another map */
93 };
94 
95 /*
96  *	Address map entries consist of start and end addresses,
97  *	a VM object (or sharing map) and offset into that object,
98  *	and user-exported inheritance and protection information.
99  *	Also included is control information for virtual copy operations.
100  */
101 struct vm_map_entry {
102 	struct vm_map_entry *prev;	/* previous entry */
103 	struct vm_map_entry *next;	/* next entry */
104 	struct vm_map_entry *left;	/* left child in binary search tree */
105 	struct vm_map_entry *right;	/* right child in binary search tree */
106 	vm_offset_t start;		/* start address */
107 	vm_offset_t end;		/* end address */
108 	vm_offset_t next_read;		/* vaddr of the next sequential read */
109 	vm_size_t max_free;		/* max free space in subtree */
110 	union vm_map_object object;	/* object I point to */
111 	vm_ooffset_t offset;		/* offset into object */
112 	vm_eflags_t eflags;		/* map entry flags */
113 	vm_prot_t protection;		/* protection code */
114 	vm_prot_t max_protection;	/* maximum protection */
115 	vm_inherit_t inheritance;	/* inheritance */
116 	uint8_t read_ahead;		/* pages in the read-ahead window */
117 	int wired_count;		/* can be paged if = 0 */
118 	struct ucred *cred;		/* tmp storage for creator ref */
119 	struct thread *wiring_thread;
120 };
121 
122 #define	MAP_ENTRY_NOSYNC		0x00000001
123 #define	MAP_ENTRY_IS_SUB_MAP		0x00000002
124 #define	MAP_ENTRY_COW			0x00000004
125 #define	MAP_ENTRY_NEEDS_COPY		0x00000008
126 #define	MAP_ENTRY_NOFAULT		0x00000010
127 #define	MAP_ENTRY_USER_WIRED		0x00000020
128 
129 #define	MAP_ENTRY_BEHAV_NORMAL		0x00000000	/* default behavior */
130 #define	MAP_ENTRY_BEHAV_SEQUENTIAL	0x00000040	/* expect sequential
131 							   access */
132 #define	MAP_ENTRY_BEHAV_RANDOM		0x00000080	/* expect random
133 							   access */
134 #define	MAP_ENTRY_BEHAV_RESERVED	0x000000c0	/* future use */
135 #define	MAP_ENTRY_BEHAV_MASK		0x000000c0
136 #define	MAP_ENTRY_IN_TRANSITION		0x00000100	/* entry being
137 							   changed */
138 #define	MAP_ENTRY_NEEDS_WAKEUP		0x00000200	/* waiters in
139 							   transition */
140 #define	MAP_ENTRY_NOCOREDUMP		0x00000400	/* don't include in
141 							   a core */
142 #define	MAP_ENTRY_VN_EXEC		0x00000800	/* text vnode mapping */
143 #define	MAP_ENTRY_GROWS_DOWN		0x00001000	/* top-down stacks */
144 #define	MAP_ENTRY_GROWS_UP		0x00002000	/* bottom-up stacks */
145 
146 #define	MAP_ENTRY_WIRE_SKIPPED		0x00004000
147 #define	MAP_ENTRY_WRITECNT		0x00008000	/* tracked writeable
148 							   mapping */
149 #define	MAP_ENTRY_GUARD			0x00010000
150 #define	MAP_ENTRY_STACK_GAP_DN		0x00020000
151 #define	MAP_ENTRY_STACK_GAP_UP		0x00040000
152 #define	MAP_ENTRY_HEADER		0x00080000
153 
154 #ifdef	_KERNEL
155 static __inline u_char
156 vm_map_entry_behavior(vm_map_entry_t entry)
157 {
158 	return (entry->eflags & MAP_ENTRY_BEHAV_MASK);
159 }
160 
161 static __inline int
162 vm_map_entry_user_wired_count(vm_map_entry_t entry)
163 {
164 	if (entry->eflags & MAP_ENTRY_USER_WIRED)
165 		return (1);
166 	return (0);
167 }
168 
169 static __inline int
170 vm_map_entry_system_wired_count(vm_map_entry_t entry)
171 {
172 	return (entry->wired_count - vm_map_entry_user_wired_count(entry));
173 }
174 #endif	/* _KERNEL */
175 
176 /*
177  *	A map is a set of map entries.  These map entries are
178  *	organized both as a binary search tree and as a doubly-linked
179  *	list.  Both structures are ordered based upon the start and
180  *	end addresses contained within each map entry.
181  *
182  *	Sleator and Tarjan's top-down splay algorithm is employed to
183  *	control height imbalance in the binary search tree.
184  *
185  *	The map's min offset value is stored in map->header.end, and
186  *	its max offset value is stored in map->header.start.  These
187  *	values act as sentinels for any forward or backward address
188  *	scan of the list.  The map header has a special value for the
189  *	eflags field, MAP_ENTRY_HEADER, that is set initially, is
190  *	never changed, and prevents an eflags match of the header
191  *	with any other map entry.
192  *
193  *	List of locks
194  *	(c)	const until freed
195  */
196 struct vm_map {
197 	struct vm_map_entry header;	/* List of entries */
198 	struct sx lock;			/* Lock for map data */
199 	struct mtx system_mtx;
200 	int nentries;			/* Number of entries */
201 	vm_size_t size;			/* virtual size */
202 	u_int timestamp;		/* Version number */
203 	u_char needs_wakeup;
204 	u_char system_map;		/* (c) Am I a system map? */
205 	vm_flags_t flags;		/* flags for this vm_map */
206 	vm_map_entry_t root;		/* Root of a binary search tree */
207 	pmap_t pmap;			/* (c) Physical map */
208 	vm_offset_t anon_loc;
209 	int busy;
210 #ifdef DIAGNOSTIC
211 	int nupdates;
212 #endif
213 };
214 
215 /*
216  * vm_flags_t values
217  */
218 #define MAP_WIREFUTURE		0x01	/* wire all future pages */
219 #define	MAP_BUSY_WAKEUP		0x02
220 #define	MAP_IS_SUB_MAP		0x04	/* has parent */
221 #define	MAP_ASLR		0x08	/* enabled ASLR */
222 #define	MAP_ASLR_IGNSTART	0x10
223 
224 #ifdef	_KERNEL
225 #if defined(KLD_MODULE) && !defined(KLD_TIED)
226 #define	vm_map_max(map)		vm_map_max_KBI((map))
227 #define	vm_map_min(map)		vm_map_min_KBI((map))
228 #define	vm_map_pmap(map)	vm_map_pmap_KBI((map))
229 #else
230 static __inline vm_offset_t
231 vm_map_max(const struct vm_map *map)
232 {
233 
234 	return (map->header.start);
235 }
236 
237 static __inline vm_offset_t
238 vm_map_min(const struct vm_map *map)
239 {
240 
241 	return (map->header.end);
242 }
243 
244 static __inline pmap_t
245 vm_map_pmap(vm_map_t map)
246 {
247 	return (map->pmap);
248 }
249 
250 static __inline void
251 vm_map_modflags(vm_map_t map, vm_flags_t set, vm_flags_t clear)
252 {
253 	map->flags = (map->flags | set) & ~clear;
254 }
255 #endif	/* KLD_MODULE */
256 #endif	/* _KERNEL */
257 
258 /*
259  * Shareable process virtual address space.
260  *
261  * List of locks
262  *	(c)	const until freed
263  */
264 struct vmspace {
265 	struct vm_map vm_map;	/* VM address map */
266 	struct shmmap_state *vm_shm;	/* SYS5 shared memory private data XXX */
267 	segsz_t vm_swrss;	/* resident set size before last swap */
268 	segsz_t vm_tsize;	/* text size (pages) XXX */
269 	segsz_t vm_dsize;	/* data size (pages) XXX */
270 	segsz_t vm_ssize;	/* stack size (pages) */
271 	caddr_t vm_taddr;	/* (c) user virtual address of text */
272 	caddr_t vm_daddr;	/* (c) user virtual address of data */
273 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
274 	volatile int vm_refcnt;	/* number of references */
275 	/*
276 	 * Keep the PMAP last, so that CPU-specific variations of that
277 	 * structure on a single architecture don't result in offset
278 	 * variations of the machine-independent fields in the vmspace.
279 	 */
280 	struct pmap vm_pmap;	/* private physical map */
281 };
282 
283 #ifdef	_KERNEL
284 static __inline pmap_t
285 vmspace_pmap(struct vmspace *vmspace)
286 {
287 	return &vmspace->vm_pmap;
288 }
289 #endif	/* _KERNEL */
290 
291 #ifdef	_KERNEL
292 /*
293  *	Macros:		vm_map_lock, etc.
294  *	Function:
295  *		Perform locking on the data portion of a map.  Note that
296  *		these macros mimic procedure calls returning void.  The
297  *		semicolon is supplied by the user of these macros, not
298  *		by the macros themselves.  The macros can safely be used
299  *		as unbraced elements in a higher level statement.
300  */
301 
302 void _vm_map_lock(vm_map_t map, const char *file, int line);
303 void _vm_map_unlock(vm_map_t map, const char *file, int line);
304 int _vm_map_unlock_and_wait(vm_map_t map, int timo, const char *file, int line);
305 void _vm_map_lock_read(vm_map_t map, const char *file, int line);
306 void _vm_map_unlock_read(vm_map_t map, const char *file, int line);
307 int _vm_map_trylock(vm_map_t map, const char *file, int line);
308 int _vm_map_trylock_read(vm_map_t map, const char *file, int line);
309 int _vm_map_lock_upgrade(vm_map_t map, const char *file, int line);
310 void _vm_map_lock_downgrade(vm_map_t map, const char *file, int line);
311 int vm_map_locked(vm_map_t map);
312 void vm_map_wakeup(vm_map_t map);
313 void vm_map_busy(vm_map_t map);
314 void vm_map_unbusy(vm_map_t map);
315 void vm_map_wait_busy(vm_map_t map);
316 vm_offset_t vm_map_max_KBI(const struct vm_map *map);
317 vm_offset_t vm_map_min_KBI(const struct vm_map *map);
318 pmap_t vm_map_pmap_KBI(vm_map_t map);
319 
320 #define	vm_map_lock(map)	_vm_map_lock(map, LOCK_FILE, LOCK_LINE)
321 #define	vm_map_unlock(map)	_vm_map_unlock(map, LOCK_FILE, LOCK_LINE)
322 #define	vm_map_unlock_and_wait(map, timo)	\
323 			_vm_map_unlock_and_wait(map, timo, LOCK_FILE, LOCK_LINE)
324 #define	vm_map_lock_read(map)	_vm_map_lock_read(map, LOCK_FILE, LOCK_LINE)
325 #define	vm_map_unlock_read(map)	_vm_map_unlock_read(map, LOCK_FILE, LOCK_LINE)
326 #define	vm_map_trylock(map)	_vm_map_trylock(map, LOCK_FILE, LOCK_LINE)
327 #define	vm_map_trylock_read(map)	\
328 			_vm_map_trylock_read(map, LOCK_FILE, LOCK_LINE)
329 #define	vm_map_lock_upgrade(map)	\
330 			_vm_map_lock_upgrade(map, LOCK_FILE, LOCK_LINE)
331 #define	vm_map_lock_downgrade(map)	\
332 			_vm_map_lock_downgrade(map, LOCK_FILE, LOCK_LINE)
333 
334 long vmspace_resident_count(struct vmspace *vmspace);
335 #endif	/* _KERNEL */
336 
337 
338 /* XXX: number of kernel maps to statically allocate */
339 #define MAX_KMAP	10
340 
341 /*
342  * Copy-on-write flags for vm_map operations
343  */
344 #define	MAP_INHERIT_SHARE	0x00000001
345 #define	MAP_COPY_ON_WRITE	0x00000002
346 #define	MAP_NOFAULT		0x00000004
347 #define	MAP_PREFAULT		0x00000008
348 #define	MAP_PREFAULT_PARTIAL	0x00000010
349 #define	MAP_DISABLE_SYNCER	0x00000020
350 #define	MAP_CHECK_EXCL		0x00000040
351 #define	MAP_CREATE_GUARD	0x00000080
352 #define	MAP_DISABLE_COREDUMP	0x00000100
353 #define	MAP_PREFAULT_MADVISE	0x00000200    /* from (user) madvise request */
354 #define	MAP_WRITECOUNT		0x00000400
355 #define	MAP_REMAP		0x00000800
356 #define	MAP_STACK_GROWS_DOWN	0x00001000
357 #define	MAP_STACK_GROWS_UP	0x00002000
358 #define	MAP_ACC_CHARGED		0x00004000
359 #define	MAP_ACC_NO_CHARGE	0x00008000
360 #define	MAP_CREATE_STACK_GAP_UP	0x00010000
361 #define	MAP_CREATE_STACK_GAP_DN	0x00020000
362 #define	MAP_VN_EXEC		0x00040000
363 
364 /*
365  * vm_fault option flags
366  */
367 #define	VM_FAULT_NORMAL	0		/* Nothing special */
368 #define	VM_FAULT_WIRE	1		/* Wire the mapped page */
369 #define	VM_FAULT_DIRTY	2		/* Dirty the page; use w/VM_PROT_COPY */
370 
371 /*
372  * Initially, mappings are slightly sequential.  The maximum window size must
373  * account for the map entry's "read_ahead" field being defined as an uint8_t.
374  */
375 #define	VM_FAULT_READ_AHEAD_MIN		7
376 #define	VM_FAULT_READ_AHEAD_INIT	15
377 #define	VM_FAULT_READ_AHEAD_MAX		min(atop(MAXPHYS) - 1, UINT8_MAX)
378 
379 /*
380  * The following "find_space" options are supported by vm_map_find().
381  *
382  * For VMFS_ALIGNED_SPACE, the desired alignment is specified to
383  * the macro argument as log base 2 of the desired alignment.
384  */
385 #define	VMFS_NO_SPACE		0	/* don't find; use the given range */
386 #define	VMFS_ANY_SPACE		1	/* find a range with any alignment */
387 #define	VMFS_OPTIMAL_SPACE	2	/* find a range with optimal alignment*/
388 #define	VMFS_SUPER_SPACE	3	/* find a superpage-aligned range */
389 #define	VMFS_ALIGNED_SPACE(x)	((x) << 8) /* find a range with fixed alignment */
390 
391 /*
392  * vm_map_wire and vm_map_unwire option flags
393  */
394 #define VM_MAP_WIRE_SYSTEM	0	/* wiring in a kernel map */
395 #define VM_MAP_WIRE_USER	1	/* wiring in a user map */
396 
397 #define VM_MAP_WIRE_NOHOLES	0	/* region must not have holes */
398 #define VM_MAP_WIRE_HOLESOK	2	/* region may have holes */
399 
400 #define VM_MAP_WIRE_WRITE	4	/* Validate writable. */
401 
402 #ifdef _KERNEL
403 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t);
404 vm_map_t vm_map_create(pmap_t, vm_offset_t, vm_offset_t);
405 int vm_map_delete(vm_map_t, vm_offset_t, vm_offset_t);
406 int vm_map_find(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *, vm_size_t,
407     vm_offset_t, int, vm_prot_t, vm_prot_t, int);
408 int vm_map_find_min(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *,
409     vm_size_t, vm_offset_t, vm_offset_t, int, vm_prot_t, vm_prot_t, int);
410 int vm_map_fixed(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t, vm_size_t,
411     vm_prot_t, vm_prot_t, int);
412 vm_offset_t vm_map_findspace(vm_map_t, vm_offset_t, vm_size_t);
413 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
414 void vm_map_init(vm_map_t, pmap_t, vm_offset_t, vm_offset_t);
415 int vm_map_insert (vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t, vm_offset_t, vm_prot_t, vm_prot_t, int);
416 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
417     vm_pindex_t *, vm_prot_t *, boolean_t *);
418 int vm_map_lookup_locked(vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
419     vm_pindex_t *, vm_prot_t *, boolean_t *);
420 void vm_map_lookup_done (vm_map_t, vm_map_entry_t);
421 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
422 
423 static inline vm_map_entry_t
424 vm_map_entry_first(vm_map_t map)
425 {
426 
427 	return (map->header.next);
428 }
429 
430 static inline vm_map_entry_t
431 vm_map_entry_succ(vm_map_entry_t entry)
432 {
433 
434 	return (entry->next);
435 }
436 
437 #define VM_MAP_ENTRY_FOREACH(it, map)		\
438 	for ((it) = vm_map_entry_first(map);	\
439 	    (it) != &(map)->header;		\
440 	    (it) = vm_map_entry_succ(it))
441 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
442 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
443 void vm_map_try_merge_entries(vm_map_t map, vm_map_entry_t prev,
444     vm_map_entry_t entry);
445 void vm_map_startup (void);
446 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
447 int vm_map_sync(vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
448 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int);
449 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, vm_prot_t, vm_prot_t, int);
450 int vm_map_unwire(vm_map_t map, vm_offset_t start, vm_offset_t end,
451     int flags);
452 int vm_map_wire(vm_map_t map, vm_offset_t start, vm_offset_t end, int flags);
453 int vm_map_wire_locked(vm_map_t map, vm_offset_t start, vm_offset_t end,
454     int flags);
455 long vmspace_swap_count(struct vmspace *vmspace);
456 void vm_map_entry_set_vnode_text(vm_map_entry_t entry, bool add);
457 #endif				/* _KERNEL */
458 #endif				/* _VM_MAP_ */
459