xref: /freebsd-11-stable/sys/vm/vm_map.h (revision 1096127ce8b01c04155a9f10de83ced78f3e2c29)
1 /*-
2  * Copyright (c) 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * The Mach Operating System project at Carnegie-Mellon University.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)vm_map.h	8.9 (Berkeley) 5/17/95
33  *
34  *
35  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
36  * All rights reserved.
37  *
38  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
39  *
40  * Permission to use, copy, modify and distribute this software and
41  * its documentation is hereby granted, provided that both the copyright
42  * notice and this permission notice appear in all copies of the
43  * software, derivative works or modified versions, and any portions
44  * thereof, and that both notices appear in supporting documentation.
45  *
46  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
47  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
48  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
49  *
50  * Carnegie Mellon requests users of this software to return to
51  *
52  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
53  *  School of Computer Science
54  *  Carnegie Mellon University
55  *  Pittsburgh PA 15213-3890
56  *
57  * any improvements or extensions that they make and grant Carnegie the
58  * rights to redistribute these changes.
59  *
60  * $FreeBSD$
61  */
62 
63 /*
64  *	Virtual memory map module definitions.
65  */
66 #ifndef	_VM_MAP_
67 #define	_VM_MAP_
68 
69 #include <sys/lock.h>
70 #include <sys/sx.h>
71 #include <sys/_mutex.h>
72 
73 /*
74  *	Types defined:
75  *
76  *	vm_map_t		the high-level address map data structure.
77  *	vm_map_entry_t		an entry in an address map.
78  */
79 
80 typedef u_char vm_flags_t;
81 typedef u_int vm_eflags_t;
82 
83 /*
84  *	Objects which live in maps may be either VM objects, or
85  *	another map (called a "sharing map") which denotes read-write
86  *	sharing with other maps.
87  */
88 union vm_map_object {
89 	struct vm_object *vm_object;	/* object object */
90 	struct vm_map *sub_map;		/* belongs to another map */
91 };
92 
93 /*
94  *	Address map entries consist of start and end addresses,
95  *	a VM object (or sharing map) and offset into that object,
96  *	and user-exported inheritance and protection information.
97  *	Also included is control information for virtual copy operations.
98  */
99 struct vm_map_entry {
100 	struct vm_map_entry *prev;	/* previous entry */
101 	struct vm_map_entry *next;	/* next entry */
102 	struct vm_map_entry *left;	/* left child in binary search tree */
103 	struct vm_map_entry *right;	/* right child in binary search tree */
104 	vm_offset_t start;		/* start address */
105 	vm_offset_t end;		/* end address */
106 	vm_offset_t pad0;
107 	vm_offset_t next_read;		/* vaddr of the next sequential read */
108 	vm_size_t adj_free;		/* amount of adjacent free space */
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		0x0001
123 #define MAP_ENTRY_IS_SUB_MAP		0x0002
124 #define MAP_ENTRY_COW			0x0004
125 #define MAP_ENTRY_NEEDS_COPY		0x0008
126 #define MAP_ENTRY_NOFAULT		0x0010
127 #define MAP_ENTRY_USER_WIRED		0x0020
128 
129 #define MAP_ENTRY_BEHAV_NORMAL		0x0000	/* default behavior */
130 #define MAP_ENTRY_BEHAV_SEQUENTIAL	0x0040	/* expect sequential access */
131 #define MAP_ENTRY_BEHAV_RANDOM		0x0080	/* expect random access */
132 #define MAP_ENTRY_BEHAV_RESERVED	0x00C0	/* future use */
133 
134 #define MAP_ENTRY_BEHAV_MASK		0x00C0
135 
136 #define MAP_ENTRY_IN_TRANSITION		0x0100	/* entry being changed */
137 #define MAP_ENTRY_NEEDS_WAKEUP		0x0200	/* waiters in transition */
138 #define MAP_ENTRY_NOCOREDUMP		0x0400	/* don't include in a core */
139 
140 #define	MAP_ENTRY_GROWS_DOWN		0x1000	/* Top-down stacks */
141 #define	MAP_ENTRY_GROWS_UP		0x2000	/* Bottom-up stacks */
142 
143 #define	MAP_ENTRY_WIRE_SKIPPED		0x4000
144 #define	MAP_ENTRY_VN_WRITECNT		0x8000	/* writeable vnode mapping */
145 #define	MAP_ENTRY_GUARD			0x10000
146 #define	MAP_ENTRY_STACK_GAP_DN		0x20000
147 #define	MAP_ENTRY_STACK_GAP_UP		0x40000
148 
149 #ifdef	_KERNEL
150 static __inline u_char
vm_map_entry_behavior(vm_map_entry_t entry)151 vm_map_entry_behavior(vm_map_entry_t entry)
152 {
153 	return (entry->eflags & MAP_ENTRY_BEHAV_MASK);
154 }
155 
156 static __inline int
vm_map_entry_user_wired_count(vm_map_entry_t entry)157 vm_map_entry_user_wired_count(vm_map_entry_t entry)
158 {
159 	if (entry->eflags & MAP_ENTRY_USER_WIRED)
160 		return (1);
161 	return (0);
162 }
163 
164 static __inline int
vm_map_entry_system_wired_count(vm_map_entry_t entry)165 vm_map_entry_system_wired_count(vm_map_entry_t entry)
166 {
167 	return (entry->wired_count - vm_map_entry_user_wired_count(entry));
168 }
169 #endif	/* _KERNEL */
170 
171 /*
172  *	A map is a set of map entries.  These map entries are
173  *	organized both as a binary search tree and as a doubly-linked
174  *	list.  Both structures are ordered based upon the start and
175  *	end addresses contained within each map entry.
176  *
177  *	Counterintuitively, the map's min offset value is stored in
178  *	map->header.end, and its max offset value is stored in
179  *	map->header.start.
180  *
181  *	The list header has max start value and min end value to act
182  *	as sentinels for sequential search of the doubly-linked list.
183  *	Sleator and Tarjan's top-down splay algorithm is employed to
184  *	control height imbalance in the binary search tree.
185  *
186  *	List of locks
187  *	(c)	const until freed
188  */
189 struct vm_map {
190 	struct vm_map_entry header;	/* List of entries */
191 /*
192 	map min_offset	header.end	(c)
193 	map max_offset	header.start	(c)
194 */
195 	struct sx lock;			/* Lock for map data */
196 	struct mtx system_mtx;
197 	int nentries;			/* Number of entries */
198 	vm_size_t size;			/* virtual size */
199 	u_int timestamp;		/* Version number */
200 	u_char needs_wakeup;
201 	u_char system_map;		/* (c) Am I a system map? */
202 	vm_flags_t flags;		/* flags for this vm_map */
203 	vm_map_entry_t root;		/* Root of a binary search tree */
204 	pmap_t pmap;			/* (c) Physical map */
205 	int busy;
206 };
207 
208 /*
209  * vm_flags_t values
210  */
211 #define MAP_WIREFUTURE		0x01	/* wire all future pages */
212 #define	MAP_BUSY_WAKEUP		0x02
213 
214 #ifdef	_KERNEL
215 #ifdef KLD_MODULE
216 #define	vm_map_max(map)		vm_map_max_KBI((map))
217 #define	vm_map_min(map)		vm_map_min_KBI((map))
218 #define	vm_map_pmap(map)	vm_map_pmap_KBI((map))
219 #else
220 static __inline vm_offset_t
vm_map_max(const struct vm_map * map)221 vm_map_max(const struct vm_map *map)
222 {
223 
224 	return (map->header.start);
225 }
226 
227 static __inline vm_offset_t
vm_map_min(const struct vm_map * map)228 vm_map_min(const struct vm_map *map)
229 {
230 
231 	return (map->header.end);
232 }
233 
234 static __inline pmap_t
vm_map_pmap(vm_map_t map)235 vm_map_pmap(vm_map_t map)
236 {
237 	return (map->pmap);
238 }
239 
240 static __inline void
vm_map_modflags(vm_map_t map,vm_flags_t set,vm_flags_t clear)241 vm_map_modflags(vm_map_t map, vm_flags_t set, vm_flags_t clear)
242 {
243 	map->flags = (map->flags | set) & ~clear;
244 }
245 #endif	/* KLD_MODULE */
246 #endif	/* _KERNEL */
247 
248 /*
249  * Shareable process virtual address space.
250  *
251  * List of locks
252  *	(c)	const until freed
253  */
254 struct vmspace {
255 	struct vm_map vm_map;	/* VM address map */
256 	struct shmmap_state *vm_shm;	/* SYS5 shared memory private data XXX */
257 	segsz_t vm_swrss;	/* resident set size before last swap */
258 	segsz_t vm_tsize;	/* text size (pages) XXX */
259 	segsz_t vm_dsize;	/* data size (pages) XXX */
260 	segsz_t vm_ssize;	/* stack size (pages) */
261 	caddr_t vm_taddr;	/* (c) user virtual address of text */
262 	caddr_t vm_daddr;	/* (c) user virtual address of data */
263 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
264 	volatile int vm_refcnt;	/* number of references */
265 	/*
266 	 * Keep the PMAP last, so that CPU-specific variations of that
267 	 * structure on a single architecture don't result in offset
268 	 * variations of the machine-independent fields in the vmspace.
269 	 */
270 	struct pmap vm_pmap;	/* private physical map */
271 };
272 
273 #ifdef	_KERNEL
274 static __inline pmap_t
vmspace_pmap(struct vmspace * vmspace)275 vmspace_pmap(struct vmspace *vmspace)
276 {
277 	return &vmspace->vm_pmap;
278 }
279 #endif	/* _KERNEL */
280 
281 #ifdef	_KERNEL
282 /*
283  *	Macros:		vm_map_lock, etc.
284  *	Function:
285  *		Perform locking on the data portion of a map.  Note that
286  *		these macros mimic procedure calls returning void.  The
287  *		semicolon is supplied by the user of these macros, not
288  *		by the macros themselves.  The macros can safely be used
289  *		as unbraced elements in a higher level statement.
290  */
291 
292 void _vm_map_lock(vm_map_t map, const char *file, int line);
293 void _vm_map_unlock(vm_map_t map, const char *file, int line);
294 int _vm_map_unlock_and_wait(vm_map_t map, int timo, const char *file, int line);
295 void _vm_map_lock_read(vm_map_t map, const char *file, int line);
296 void _vm_map_unlock_read(vm_map_t map, const char *file, int line);
297 int _vm_map_trylock(vm_map_t map, const char *file, int line);
298 int _vm_map_trylock_read(vm_map_t map, const char *file, int line);
299 int _vm_map_lock_upgrade(vm_map_t map, const char *file, int line);
300 void _vm_map_lock_downgrade(vm_map_t map, const char *file, int line);
301 int vm_map_locked(vm_map_t map);
302 void vm_map_wakeup(vm_map_t map);
303 void vm_map_busy(vm_map_t map);
304 void vm_map_unbusy(vm_map_t map);
305 void vm_map_wait_busy(vm_map_t map);
306 vm_offset_t vm_map_max_KBI(const struct vm_map *map);
307 vm_offset_t vm_map_min_KBI(const struct vm_map *map);
308 pmap_t vm_map_pmap_KBI(vm_map_t map);
309 
310 #define	vm_map_lock(map)	_vm_map_lock(map, LOCK_FILE, LOCK_LINE)
311 #define	vm_map_unlock(map)	_vm_map_unlock(map, LOCK_FILE, LOCK_LINE)
312 #define	vm_map_unlock_and_wait(map, timo)	\
313 			_vm_map_unlock_and_wait(map, timo, LOCK_FILE, LOCK_LINE)
314 #define	vm_map_lock_read(map)	_vm_map_lock_read(map, LOCK_FILE, LOCK_LINE)
315 #define	vm_map_unlock_read(map)	_vm_map_unlock_read(map, LOCK_FILE, LOCK_LINE)
316 #define	vm_map_trylock(map)	_vm_map_trylock(map, LOCK_FILE, LOCK_LINE)
317 #define	vm_map_trylock_read(map)	\
318 			_vm_map_trylock_read(map, LOCK_FILE, LOCK_LINE)
319 #define	vm_map_lock_upgrade(map)	\
320 			_vm_map_lock_upgrade(map, LOCK_FILE, LOCK_LINE)
321 #define	vm_map_lock_downgrade(map)	\
322 			_vm_map_lock_downgrade(map, LOCK_FILE, LOCK_LINE)
323 
324 long vmspace_resident_count(struct vmspace *vmspace);
325 #endif	/* _KERNEL */
326 
327 
328 /* XXX: number of kernel maps to statically allocate */
329 #define MAX_KMAP	10
330 
331 /*
332  * Copy-on-write flags for vm_map operations
333  */
334 #define MAP_INHERIT_SHARE	0x0001
335 #define MAP_COPY_ON_WRITE	0x0002
336 #define MAP_NOFAULT		0x0004
337 #define MAP_PREFAULT		0x0008
338 #define MAP_PREFAULT_PARTIAL	0x0010
339 #define MAP_DISABLE_SYNCER	0x0020
340 #define	MAP_CHECK_EXCL		0x0040
341 #define	MAP_CREATE_GUARD	0x0080
342 #define MAP_DISABLE_COREDUMP	0x0100
343 #define MAP_PREFAULT_MADVISE	0x0200	/* from (user) madvise request */
344 #define	MAP_VN_WRITECOUNT	0x0400
345 #define	MAP_REMAP		0x0800
346 #define	MAP_STACK_GROWS_DOWN	0x1000
347 #define	MAP_STACK_GROWS_UP	0x2000
348 #define	MAP_ACC_CHARGED		0x4000
349 #define	MAP_ACC_NO_CHARGE	0x8000
350 #define	MAP_CREATE_STACK_GAP_UP	0x10000
351 #define	MAP_CREATE_STACK_GAP_DN	0x20000
352 
353 /*
354  * vm_fault option flags
355  */
356 #define	VM_FAULT_NORMAL	0		/* Nothing special */
357 #define	VM_FAULT_WIRE	1		/* Wire the mapped page */
358 #define	VM_FAULT_DIRTY	2		/* Dirty the page; use w/VM_PROT_COPY */
359 
360 /*
361  * Initially, mappings are slightly sequential.  The maximum window size must
362  * account for the map entry's "read_ahead" field being defined as an uint8_t.
363  */
364 #define	VM_FAULT_READ_AHEAD_MIN		7
365 #define	VM_FAULT_READ_AHEAD_INIT	15
366 #define	VM_FAULT_READ_AHEAD_MAX		min(atop(MAXPHYS) - 1, UINT8_MAX)
367 
368 /*
369  * The following "find_space" options are supported by vm_map_find().
370  *
371  * For VMFS_ALIGNED_SPACE, the desired alignment is specified to
372  * the macro argument as log base 2 of the desired alignment.
373  */
374 #define	VMFS_NO_SPACE		0	/* don't find; use the given range */
375 #define	VMFS_ANY_SPACE		1	/* find a range with any alignment */
376 #define	VMFS_OPTIMAL_SPACE	2	/* find a range with optimal alignment*/
377 #define	VMFS_SUPER_SPACE	3	/* find a superpage-aligned range */
378 #define	VMFS_ALIGNED_SPACE(x)	((x) << 8) /* find a range with fixed alignment */
379 
380 /*
381  * vm_map_wire and vm_map_unwire option flags
382  */
383 #define VM_MAP_WIRE_SYSTEM	0	/* wiring in a kernel map */
384 #define VM_MAP_WIRE_USER	1	/* wiring in a user map */
385 
386 #define VM_MAP_WIRE_NOHOLES	0	/* region must not have holes */
387 #define VM_MAP_WIRE_HOLESOK	2	/* region may have holes */
388 
389 #define VM_MAP_WIRE_WRITE	4	/* Validate writable. */
390 
391 #ifdef _KERNEL
392 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t);
393 vm_map_t vm_map_create(pmap_t, vm_offset_t, vm_offset_t);
394 int vm_map_delete(vm_map_t, vm_offset_t, vm_offset_t);
395 int vm_map_find(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *, vm_size_t,
396     vm_offset_t, int, vm_prot_t, vm_prot_t, int);
397 int vm_map_find_min(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *,
398     vm_size_t, vm_offset_t, vm_offset_t, int, vm_prot_t, vm_prot_t, int);
399 int vm_map_fixed(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t, vm_size_t,
400     vm_prot_t, vm_prot_t, int);
401 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_offset_t *);
402 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
403 void vm_map_init(vm_map_t, pmap_t, vm_offset_t, vm_offset_t);
404 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);
405 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
406     vm_pindex_t *, vm_prot_t *, boolean_t *);
407 int vm_map_lookup_locked(vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
408     vm_pindex_t *, vm_prot_t *, boolean_t *);
409 void vm_map_lookup_done (vm_map_t, vm_map_entry_t);
410 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
411 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
412 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
413 void vm_map_simplify_entry(vm_map_t map, vm_map_entry_t entry);
414 void vm_map_startup (void);
415 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
416 int vm_map_sync(vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
417 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int);
418 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, vm_prot_t, vm_prot_t, int);
419 int vm_map_unwire(vm_map_t map, vm_offset_t start, vm_offset_t end,
420     int flags);
421 int vm_map_wire(vm_map_t map, vm_offset_t start, vm_offset_t end,
422     int flags);
423 long vmspace_swap_count(struct vmspace *vmspace);
424 #endif				/* _KERNEL */
425 #endif				/* _VM_MAP_ */
426