xref: /freebsd/sys/vm/vm_map.h (revision 74ca7bf1d4c7173d5575ba168bc4b5f6d181ff5a)
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 adj_free;		/* amount of adjacent free space */
110 	vm_size_t max_free;		/* max free space in subtree */
111 	union vm_map_object object;	/* object I point to */
112 	vm_ooffset_t offset;		/* offset into object */
113 	vm_eflags_t eflags;		/* map entry flags */
114 	vm_prot_t protection;		/* protection code */
115 	vm_prot_t max_protection;	/* maximum protection */
116 	vm_inherit_t inheritance;	/* inheritance */
117 	uint8_t read_ahead;		/* pages in the read-ahead window */
118 	int wired_count;		/* can be paged if = 0 */
119 	struct ucred *cred;		/* tmp storage for creator ref */
120 	struct thread *wiring_thread;
121 };
122 
123 #define MAP_ENTRY_NOSYNC		0x0001
124 #define MAP_ENTRY_IS_SUB_MAP		0x0002
125 #define MAP_ENTRY_COW			0x0004
126 #define MAP_ENTRY_NEEDS_COPY		0x0008
127 #define MAP_ENTRY_NOFAULT		0x0010
128 #define MAP_ENTRY_USER_WIRED		0x0020
129 
130 #define MAP_ENTRY_BEHAV_NORMAL		0x0000	/* default behavior */
131 #define MAP_ENTRY_BEHAV_SEQUENTIAL	0x0040	/* expect sequential access */
132 #define MAP_ENTRY_BEHAV_RANDOM		0x0080	/* expect random access */
133 #define MAP_ENTRY_BEHAV_RESERVED	0x00C0	/* future use */
134 
135 #define MAP_ENTRY_BEHAV_MASK		0x00C0
136 
137 #define MAP_ENTRY_IN_TRANSITION		0x0100	/* entry being changed */
138 #define MAP_ENTRY_NEEDS_WAKEUP		0x0200	/* waiters in transition */
139 #define MAP_ENTRY_NOCOREDUMP		0x0400	/* don't include in a core */
140 
141 #define	MAP_ENTRY_GROWS_DOWN		0x1000	/* Top-down stacks */
142 #define	MAP_ENTRY_GROWS_UP		0x2000	/* Bottom-up stacks */
143 
144 #define	MAP_ENTRY_WIRE_SKIPPED		0x4000
145 #define	MAP_ENTRY_VN_WRITECNT		0x8000	/* writeable vnode mapping */
146 #define	MAP_ENTRY_GUARD			0x10000
147 #define	MAP_ENTRY_STACK_GAP_DN		0x20000
148 #define	MAP_ENTRY_STACK_GAP_UP		0x40000
149 
150 #ifdef	_KERNEL
151 static __inline u_char
152 vm_map_entry_behavior(vm_map_entry_t entry)
153 {
154 	return (entry->eflags & MAP_ENTRY_BEHAV_MASK);
155 }
156 
157 static __inline int
158 vm_map_entry_user_wired_count(vm_map_entry_t entry)
159 {
160 	if (entry->eflags & MAP_ENTRY_USER_WIRED)
161 		return (1);
162 	return (0);
163 }
164 
165 static __inline int
166 vm_map_entry_system_wired_count(vm_map_entry_t entry)
167 {
168 	return (entry->wired_count - vm_map_entry_user_wired_count(entry));
169 }
170 #endif	/* _KERNEL */
171 
172 /*
173  *	A map is a set of map entries.  These map entries are
174  *	organized both as a binary search tree and as a doubly-linked
175  *	list.  Both structures are ordered based upon the start and
176  *	end addresses contained within each map entry.  The list
177  *	header has max start value and min end value to act as
178  *	sentinels for sequential search of the doubly-linked list.
179  *	Sleator and Tarjan's top-down splay algorithm is employed to
180  *	control height imbalance in the binary search tree.
181  *
182  * List of locks
183  *	(c)	const until freed
184  */
185 struct vm_map {
186 	struct vm_map_entry header;	/* List of entries */
187 #define	min_offset	header.end	/* (c) */
188 #define	max_offset	header.start	/* (c) */
189 	struct sx lock;			/* Lock for map data */
190 	struct mtx system_mtx;
191 	int nentries;			/* Number of entries */
192 	vm_size_t size;			/* virtual size */
193 	u_int timestamp;		/* Version number */
194 	u_char needs_wakeup;
195 	u_char system_map;		/* (c) Am I a system map? */
196 	vm_flags_t flags;		/* flags for this vm_map */
197 	vm_map_entry_t root;		/* Root of a binary search tree */
198 	pmap_t pmap;			/* (c) Physical map */
199 	int busy;
200 };
201 
202 /*
203  * vm_flags_t values
204  */
205 #define MAP_WIREFUTURE		0x01	/* wire all future pages */
206 #define	MAP_BUSY_WAKEUP		0x02
207 
208 #ifdef	_KERNEL
209 #ifdef KLD_MODULE
210 #define	vm_map_max(map)		vm_map_max_KBI((map))
211 #define	vm_map_min(map)		vm_map_min_KBI((map))
212 #define	vm_map_pmap(map)	vm_map_pmap_KBI((map))
213 #else
214 static __inline vm_offset_t
215 vm_map_max(const struct vm_map *map)
216 {
217 	return (map->max_offset);
218 }
219 
220 static __inline vm_offset_t
221 vm_map_min(const struct vm_map *map)
222 {
223 	return (map->min_offset);
224 }
225 
226 static __inline pmap_t
227 vm_map_pmap(vm_map_t map)
228 {
229 	return (map->pmap);
230 }
231 
232 static __inline void
233 vm_map_modflags(vm_map_t map, vm_flags_t set, vm_flags_t clear)
234 {
235 	map->flags = (map->flags | set) & ~clear;
236 }
237 #endif	/* KLD_MODULE */
238 #endif	/* _KERNEL */
239 
240 /*
241  * Shareable process virtual address space.
242  *
243  * List of locks
244  *	(c)	const until freed
245  */
246 struct vmspace {
247 	struct vm_map vm_map;	/* VM address map */
248 	struct shmmap_state *vm_shm;	/* SYS5 shared memory private data XXX */
249 	segsz_t vm_swrss;	/* resident set size before last swap */
250 	segsz_t vm_tsize;	/* text size (pages) XXX */
251 	segsz_t vm_dsize;	/* data size (pages) XXX */
252 	segsz_t vm_ssize;	/* stack size (pages) */
253 	caddr_t vm_taddr;	/* (c) user virtual address of text */
254 	caddr_t vm_daddr;	/* (c) user virtual address of data */
255 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
256 	volatile int vm_refcnt;	/* number of references */
257 	/*
258 	 * Keep the PMAP last, so that CPU-specific variations of that
259 	 * structure on a single architecture don't result in offset
260 	 * variations of the machine-independent fields in the vmspace.
261 	 */
262 	struct pmap vm_pmap;	/* private physical map */
263 };
264 
265 #ifdef	_KERNEL
266 static __inline pmap_t
267 vmspace_pmap(struct vmspace *vmspace)
268 {
269 	return &vmspace->vm_pmap;
270 }
271 #endif	/* _KERNEL */
272 
273 #ifdef	_KERNEL
274 /*
275  *	Macros:		vm_map_lock, etc.
276  *	Function:
277  *		Perform locking on the data portion of a map.  Note that
278  *		these macros mimic procedure calls returning void.  The
279  *		semicolon is supplied by the user of these macros, not
280  *		by the macros themselves.  The macros can safely be used
281  *		as unbraced elements in a higher level statement.
282  */
283 
284 void _vm_map_lock(vm_map_t map, const char *file, int line);
285 void _vm_map_unlock(vm_map_t map, const char *file, int line);
286 int _vm_map_unlock_and_wait(vm_map_t map, int timo, const char *file, int line);
287 void _vm_map_lock_read(vm_map_t map, const char *file, int line);
288 void _vm_map_unlock_read(vm_map_t map, const char *file, int line);
289 int _vm_map_trylock(vm_map_t map, const char *file, int line);
290 int _vm_map_trylock_read(vm_map_t map, const char *file, int line);
291 int _vm_map_lock_upgrade(vm_map_t map, const char *file, int line);
292 void _vm_map_lock_downgrade(vm_map_t map, const char *file, int line);
293 int vm_map_locked(vm_map_t map);
294 void vm_map_wakeup(vm_map_t map);
295 void vm_map_busy(vm_map_t map);
296 void vm_map_unbusy(vm_map_t map);
297 void vm_map_wait_busy(vm_map_t map);
298 vm_offset_t vm_map_max_KBI(const struct vm_map *map);
299 vm_offset_t vm_map_min_KBI(const struct vm_map *map);
300 pmap_t vm_map_pmap_KBI(vm_map_t map);
301 
302 #define	vm_map_lock(map)	_vm_map_lock(map, LOCK_FILE, LOCK_LINE)
303 #define	vm_map_unlock(map)	_vm_map_unlock(map, LOCK_FILE, LOCK_LINE)
304 #define	vm_map_unlock_and_wait(map, timo)	\
305 			_vm_map_unlock_and_wait(map, timo, LOCK_FILE, LOCK_LINE)
306 #define	vm_map_lock_read(map)	_vm_map_lock_read(map, LOCK_FILE, LOCK_LINE)
307 #define	vm_map_unlock_read(map)	_vm_map_unlock_read(map, LOCK_FILE, LOCK_LINE)
308 #define	vm_map_trylock(map)	_vm_map_trylock(map, LOCK_FILE, LOCK_LINE)
309 #define	vm_map_trylock_read(map)	\
310 			_vm_map_trylock_read(map, LOCK_FILE, LOCK_LINE)
311 #define	vm_map_lock_upgrade(map)	\
312 			_vm_map_lock_upgrade(map, LOCK_FILE, LOCK_LINE)
313 #define	vm_map_lock_downgrade(map)	\
314 			_vm_map_lock_downgrade(map, LOCK_FILE, LOCK_LINE)
315 
316 long vmspace_resident_count(struct vmspace *vmspace);
317 #endif	/* _KERNEL */
318 
319 
320 /* XXX: number of kernel maps to statically allocate */
321 #define MAX_KMAP	10
322 
323 /*
324  * Copy-on-write flags for vm_map operations
325  */
326 #define MAP_INHERIT_SHARE	0x0001
327 #define MAP_COPY_ON_WRITE	0x0002
328 #define MAP_NOFAULT		0x0004
329 #define MAP_PREFAULT		0x0008
330 #define MAP_PREFAULT_PARTIAL	0x0010
331 #define MAP_DISABLE_SYNCER	0x0020
332 #define	MAP_CHECK_EXCL		0x0040
333 #define	MAP_CREATE_GUARD	0x0080
334 #define MAP_DISABLE_COREDUMP	0x0100
335 #define MAP_PREFAULT_MADVISE	0x0200	/* from (user) madvise request */
336 #define	MAP_VN_WRITECOUNT	0x0400
337 #define	MAP_STACK_GROWS_DOWN	0x1000
338 #define	MAP_STACK_GROWS_UP	0x2000
339 #define	MAP_ACC_CHARGED		0x4000
340 #define	MAP_ACC_NO_CHARGE	0x8000
341 #define	MAP_CREATE_STACK_GAP_UP	0x10000
342 #define	MAP_CREATE_STACK_GAP_DN	0x20000
343 
344 /*
345  * vm_fault option flags
346  */
347 #define	VM_FAULT_NORMAL	0		/* Nothing special */
348 #define	VM_FAULT_WIRE	1		/* Wire the mapped page */
349 #define	VM_FAULT_DIRTY	2		/* Dirty the page; use w/VM_PROT_COPY */
350 
351 /*
352  * Initially, mappings are slightly sequential.  The maximum window size must
353  * account for the map entry's "read_ahead" field being defined as an uint8_t.
354  */
355 #define	VM_FAULT_READ_AHEAD_MIN		7
356 #define	VM_FAULT_READ_AHEAD_INIT	15
357 #define	VM_FAULT_READ_AHEAD_MAX		min(atop(MAXPHYS) - 1, UINT8_MAX)
358 
359 /*
360  * The following "find_space" options are supported by vm_map_find().
361  *
362  * For VMFS_ALIGNED_SPACE, the desired alignment is specified to
363  * the macro argument as log base 2 of the desired alignment.
364  */
365 #define	VMFS_NO_SPACE		0	/* don't find; use the given range */
366 #define	VMFS_ANY_SPACE		1	/* find a range with any alignment */
367 #define	VMFS_OPTIMAL_SPACE	2	/* find a range with optimal alignment*/
368 #define	VMFS_SUPER_SPACE	3	/* find a superpage-aligned range */
369 #define	VMFS_ALIGNED_SPACE(x)	((x) << 8) /* find a range with fixed alignment */
370 
371 /*
372  * vm_map_wire and vm_map_unwire option flags
373  */
374 #define VM_MAP_WIRE_SYSTEM	0	/* wiring in a kernel map */
375 #define VM_MAP_WIRE_USER	1	/* wiring in a user map */
376 
377 #define VM_MAP_WIRE_NOHOLES	0	/* region must not have holes */
378 #define VM_MAP_WIRE_HOLESOK	2	/* region may have holes */
379 
380 #define VM_MAP_WIRE_WRITE	4	/* Validate writable. */
381 
382 #ifdef _KERNEL
383 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t);
384 vm_map_t vm_map_create(pmap_t, vm_offset_t, vm_offset_t);
385 int vm_map_delete(vm_map_t, vm_offset_t, vm_offset_t);
386 int vm_map_find(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *, vm_size_t,
387     vm_offset_t, int, vm_prot_t, vm_prot_t, int);
388 int vm_map_find_min(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *,
389     vm_size_t, vm_offset_t, vm_offset_t, int, vm_prot_t, vm_prot_t, int);
390 int vm_map_fixed(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t, vm_size_t,
391     vm_prot_t, vm_prot_t, int);
392 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_offset_t *);
393 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
394 void vm_map_init(vm_map_t, pmap_t, vm_offset_t, vm_offset_t);
395 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);
396 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
397     vm_pindex_t *, vm_prot_t *, boolean_t *);
398 int vm_map_lookup_locked(vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
399     vm_pindex_t *, vm_prot_t *, boolean_t *);
400 void vm_map_lookup_done (vm_map_t, vm_map_entry_t);
401 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
402 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
403 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
404 void vm_map_simplify_entry(vm_map_t map, vm_map_entry_t entry);
405 void vm_map_startup (void);
406 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
407 int vm_map_sync(vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
408 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int);
409 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, vm_prot_t, vm_prot_t, int);
410 int vm_map_unwire(vm_map_t map, vm_offset_t start, vm_offset_t end,
411     int flags);
412 int vm_map_wire(vm_map_t map, vm_offset_t start, vm_offset_t end,
413     int flags);
414 long vmspace_swap_count(struct vmspace *vmspace);
415 #endif				/* _KERNEL */
416 #endif				/* _VM_MAP_ */
417