xref: /openbsd-src/sys/uvm/uvm_map.h (revision 50b7afb2c2c0993b0894d4e34bf857cb13ed9c80)
1 /*	$OpenBSD: uvm_map.h,v 1.51 2014/07/11 16:35:40 jsg Exp $	*/
2 /*	$NetBSD: uvm_map.h,v 1.24 2001/02/18 21:19:08 chs Exp $	*/
3 
4 /*
5  * Copyright (c) 2011 Ariane van der Steldt <ariane@openbsd.org>
6  *
7  * Permission to use, copy, modify, and distribute this software for any
8  * purpose with or without fee is hereby granted, provided that the above
9  * copyright notice and this permission notice appear in all copies.
10  *
11  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18  *
19  *
20  * Copyright (c) 1997 Charles D. Cranor and Washington University.
21  * Copyright (c) 1991, 1993, The Regents of the University of California.
22  *
23  * All rights reserved.
24  *
25  * This code is derived from software contributed to Berkeley by
26  * The Mach Operating System project at Carnegie-Mellon University.
27  *
28  * Redistribution and use in source and binary forms, with or without
29  * modification, are permitted provided that the following conditions
30  * are met:
31  * 1. Redistributions of source code must retain the above copyright
32  *    notice, this list of conditions and the following disclaimer.
33  * 2. Redistributions in binary form must reproduce the above copyright
34  *    notice, this list of conditions and the following disclaimer in the
35  *    documentation and/or other materials provided with the distribution.
36  * 3. Neither the name of the University nor the names of its contributors
37  *    may be used to endorse or promote products derived from this software
38  *    without specific prior written permission.
39  *
40  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
41  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
42  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
43  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
44  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
45  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
46  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
47  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
48  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
49  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
50  * SUCH DAMAGE.
51  *
52  *	@(#)vm_map.h    8.3 (Berkeley) 3/15/94
53  * from: Id: uvm_map.h,v 1.1.2.3 1998/02/07 01:16:55 chs Exp
54  *
55  *
56  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
57  * All rights reserved.
58  *
59  * Permission to use, copy, modify and distribute this software and
60  * its documentation is hereby granted, provided that both the copyright
61  * notice and this permission notice appear in all copies of the
62  * software, derivative works or modified versions, and any portions
63  * thereof, and that both notices appear in supporting documentation.
64  *
65  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
66  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
67  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
68  *
69  * Carnegie Mellon requests users of this software to return to
70  *
71  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
72  *  School of Computer Science
73  *  Carnegie Mellon University
74  *  Pittsburgh PA 15213-3890
75  *
76  * any improvements or extensions that they make and grant Carnegie the
77  * rights to redistribute these changes.
78  */
79 
80 #ifndef _UVM_UVM_MAP_H_
81 #define _UVM_UVM_MAP_H_
82 
83 #include <sys/rwlock.h>
84 
85 #ifdef _KERNEL
86 
87 /*
88  * Internal functions.
89  *
90  * Required by clipping macros.
91  */
92 void			 uvm_map_clip_end(struct vm_map*, struct vm_map_entry*,
93 			    vaddr_t);
94 void			 uvm_map_clip_start(struct vm_map*,
95 			    struct vm_map_entry*, vaddr_t);
96 
97 /*
98  * UVM_MAP_CLIP_START: ensure that the entry begins at or after
99  * the starting address, if it doesn't we split the entry.
100  *
101  * => map must be locked by caller
102  */
103 
104 #define UVM_MAP_CLIP_START(_map, _entry, _addr)				\
105 	do {								\
106 		KASSERT((_entry)->end + (_entry)->fspace > (_addr));	\
107 		if ((_entry)->start < (_addr))				\
108 			uvm_map_clip_start((_map), (_entry), (_addr));	\
109 	} while (0)
110 
111 /*
112  * UVM_MAP_CLIP_END: ensure that the entry ends at or before
113  *      the ending address, if it does't we split the entry.
114  *
115  * => map must be locked by caller
116  */
117 
118 #define UVM_MAP_CLIP_END(_map, _entry, _addr)				\
119 	do {								\
120 		KASSERT((_entry)->start < (_addr));			\
121 		if ((_entry)->end > (_addr))				\
122 			uvm_map_clip_end((_map), (_entry), (_addr));	\
123 	} while (0)
124 
125 /*
126  * extract flags
127  */
128 #define UVM_EXTRACT_FIXPROT	0x8	/* set prot to maxprot as we go */
129 
130 #endif /* _KERNEL */
131 
132 #include <uvm/uvm_anon.h>
133 
134 /*
135  * types defined:
136  *
137  *	vm_map_t		the high-level address map data structure.
138  *	vm_map_entry_t		an entry in an address map.
139  *	vm_map_version_t	a timestamp of a map, for use with vm_map_lookup
140  */
141 
142 /*
143  * Objects which live in maps may be either VM objects, or another map
144  * (called a "sharing map") which denotes read-write sharing with other maps.
145  *
146  * XXXCDC: private pager data goes here now
147  */
148 
149 union vm_map_object {
150 	struct uvm_object	*uvm_obj;	/* UVM OBJECT */
151 	struct vm_map		*sub_map;	/* belongs to another map */
152 };
153 
154 /*
155  * Address map entries consist of start and end addresses,
156  * a VM object (or sharing map) and offset into that object,
157  * and user-exported inheritance and protection information.
158  * Also included is control information for virtual copy operations.
159  */
160 struct vm_map_entry {
161 	union {
162 		RB_ENTRY(vm_map_entry)	addr_entry; /* address tree */
163 	} daddrs;
164 
165 	union {
166 		RB_ENTRY(vm_map_entry)	rbtree;	/* Link freespace tree. */
167 		TAILQ_ENTRY(vm_map_entry) tailq;/* Link freespace queue. */
168 		TAILQ_ENTRY(vm_map_entry) deadq;/* dead entry queue */
169 	} dfree;
170 
171 #define uvm_map_entry_start_copy start
172 	vaddr_t			start;		/* start address */
173 	vaddr_t			end;		/* end address */
174 
175 	vsize_t			guard;		/* bytes in guard */
176 	vsize_t			fspace;		/* free space */
177 
178 	union vm_map_object	object;		/* object I point to */
179 	voff_t			offset;		/* offset into object */
180 	struct vm_aref		aref;		/* anonymous overlay */
181 
182 	int			etype;		/* entry type */
183 
184 	vm_prot_t		protection;	/* protection code */
185 	vm_prot_t		max_protection;	/* maximum protection */
186 	vm_inherit_t		inheritance;	/* inheritance */
187 
188 	int			wired_count;	/* can be paged if == 0 */
189 	int			advice;		/* madvise advice */
190 #define uvm_map_entry_stop_copy flags
191 	u_int8_t		flags;		/* flags */
192 
193 #define UVM_MAP_STATIC		0x01		/* static map entry */
194 #define UVM_MAP_KMEM		0x02		/* from kmem entry pool */
195 
196 	vsize_t			fspace_augment;	/* max(fspace) in subtree */
197 };
198 
199 #define	VM_MAPENT_ISWIRED(entry)	((entry)->wired_count != 0)
200 
201 TAILQ_HEAD(uvm_map_deadq, vm_map_entry);	/* dead entry queue */
202 RB_HEAD(uvm_map_addr, vm_map_entry);
203 RB_PROTOTYPE(uvm_map_addr, vm_map_entry, daddrs.addr_entry,
204     uvm_mapentry_addrcmp);
205 
206 /*
207  *	A Map is a rbtree of map entries, kept sorted by address.
208  *	In addition, free space entries are also kept in a rbtree,
209  *	indexed by free size.
210  *
211  *
212  *
213  *	LOCKING PROTOCOL NOTES:
214  *	-----------------------
215  *
216  *	VM map locking is a little complicated.  There are both shared
217  *	and exclusive locks on maps.  However, it is sometimes required
218  *	to downgrade an exclusive lock to a shared lock, and upgrade to
219  *	an exclusive lock again (to perform error recovery).  However,
220  *	another thread *must not* queue itself to receive an exclusive
221  *	lock while before we upgrade back to exclusive, otherwise the
222  *	error recovery becomes extremely difficult, if not impossible.
223  *
224  *	In order to prevent this scenario, we introduce the notion of
225  *	a `busy' map.  A `busy' map is read-locked, but other threads
226  *	attempting to write-lock wait for this flag to clear before
227  *	entering the lock manager.  A map may only be marked busy
228  *	when the map is write-locked (and then the map must be downgraded
229  *	to read-locked), and may only be marked unbusy by the thread
230  *	which marked it busy (holding *either* a read-lock or a
231  *	write-lock, the latter being gained by an upgrade).
232  *
233  *	Access to the map `flags' member is controlled by the `flags_lock'
234  *	simple lock.  Note that some flags are static (set once at map
235  *	creation time, and never changed), and thus require no locking
236  *	to check those flags.  All flags which are r/w must be set or
237  *	cleared while the `flags_lock' is asserted.  Additional locking
238  *	requirements are:
239  *
240  *		VM_MAP_PAGEABLE		r/o static flag; no locking required
241  *
242  *		VM_MAP_INTRSAFE		r/o static flag; no locking required
243  *
244  *		VM_MAP_WIREFUTURE	r/w; may only be set or cleared when
245  *					map is write-locked.  may be tested
246  *					without asserting `flags_lock'.
247  *
248  *		VM_MAP_BUSY		r/w; may only be set when map is
249  *					write-locked, may only be cleared by
250  *					thread which set it, map read-locked
251  *					or write-locked.  must be tested
252  *					while `flags_lock' is asserted.
253  *
254  *		VM_MAP_WANTLOCK		r/w; may only be set when the map
255  *					is busy, and thread is attempting
256  *					to write-lock.  must be tested
257  *					while `flags_lock' is asserted.
258  *
259  *		VM_MAP_GUARDPAGES	r/o; must be specified at map
260  *					initialization time.
261  *					If set, guards will appear between
262  *					automatic allocations.
263  *					No locking required.
264  *
265  *		VM_MAP_ISVMSPACE	r/o; set by uvmspace_alloc.
266  *					Signifies that this map is a vmspace.
267  *					(The implementation treats all maps
268  *					without this bit as kernel maps.)
269  *					No locking required.
270  *
271  *
272  * All automatic allocations (uvm_map without MAP_FIXED) will allocate
273  * from vm_map.free.
274  * If that allocation fails:
275  * - vmspace maps will spill over into vm_map.bfree,
276  * - all other maps will call uvm_map_kmem_grow() to increase the arena.
277  *
278  * vmspace maps have their data, brk() and stack arenas automatically
279  * updated when uvm_map() is invoked without MAP_FIXED.
280  * The spill over arena (vm_map.bfree) will contain the space in the brk()
281  * and stack ranges.
282  * Kernel maps never have a bfree arena and this tree will always be empty.
283  *
284  *
285  * read_locks and write_locks are used in lock debugging code.
286  */
287 struct vm_map {
288 	struct pmap *		pmap;		/* Physical map */
289 	struct rwlock		lock;		/* Lock for map data */
290 
291 	struct uvm_map_addr	addr;		/* Entry tree, by addr */
292 
293 	vsize_t			size;		/* virtual size */
294 	int			ref_count;	/* Reference count */
295 	int			flags;		/* flags */
296 	unsigned int		timestamp;	/* Version number */
297 
298 	vaddr_t			min_offset;	/* First address in map. */
299 	vaddr_t			max_offset;	/* Last address in map. */
300 
301 	/*
302 	 * Allocation overflow regions.
303 	 */
304 	vaddr_t			b_start;	/* Start for brk() alloc. */
305 	vaddr_t			b_end;		/* End for brk() alloc. */
306 	vaddr_t			s_start;	/* Start for stack alloc. */
307 	vaddr_t			s_end;		/* End for stack alloc. */
308 
309 	/*
310 	 * Special address selectors.
311 	 *
312 	 * The uaddr_exe mapping is used if:
313 	 * - protX is selected
314 	 * - the pointer is not NULL
315 	 *
316 	 * If uaddr_exe is not used, the other mappings are checked in
317 	 * order of appearance.
318 	 * If a hint is given, the selection will only be used if the hint
319 	 * falls in the range described by the mapping.
320 	 *
321 	 * The states are pointers because:
322 	 * - they may not all be in use
323 	 * - the struct size for different schemes is variable
324 	 *
325 	 * The uaddr_brk_stack selector will select addresses that are in
326 	 * the brk/stack area of the map.
327 	 */
328 	struct uvm_addr_state	*uaddr_exe;	/* Executable selector. */
329 	struct uvm_addr_state	*uaddr_any[4];	/* More selectors. */
330 	struct uvm_addr_state	*uaddr_brk_stack; /* Brk/stack selector. */
331 };
332 
333 /* vm_map flags */
334 #define	VM_MAP_PAGEABLE		0x01		/* ro: entries are pageable */
335 #define	VM_MAP_INTRSAFE		0x02		/* ro: interrupt safe map */
336 #define	VM_MAP_WIREFUTURE	0x04		/* rw: wire future mappings */
337 #define	VM_MAP_BUSY		0x08		/* rw: map is busy */
338 #define	VM_MAP_WANTLOCK		0x10		/* rw: want to write-lock */
339 #define VM_MAP_GUARDPAGES	0x20		/* rw: add guard pgs to map */
340 #define VM_MAP_ISVMSPACE	0x40		/* ro: map is a vmspace */
341 
342 /* XXX: number of kernel maps and entries to statically allocate */
343 
344 #if !defined(MAX_KMAPENT)
345 #define	MAX_KMAPENT	1024	/* Sufficient to make it to the scheduler. */
346 #endif	/* !defined MAX_KMAPENT */
347 
348 #ifdef _KERNEL
349 #define	vm_map_modflags(map, set, clear)				\
350 do {									\
351 	(map)->flags = ((map)->flags | (set)) & ~(clear);		\
352 } while (0)
353 #endif /* _KERNEL */
354 
355 /*
356  *	Interrupt-safe maps must also be kept on a special list,
357  *	to assist uvm_fault() in avoiding locking problems.
358  */
359 struct vm_map_intrsafe {
360 	struct vm_map	vmi_map;
361 	LIST_ENTRY(vm_map_intrsafe) vmi_list;
362 };
363 
364 /*
365  * globals:
366  */
367 
368 #ifdef _KERNEL
369 
370 extern vaddr_t	uvm_maxkaddr;
371 
372 /*
373  * protos: the following prototypes define the interface to vm_map
374  */
375 
376 void		uvm_map_deallocate(vm_map_t);
377 
378 int		uvm_map_clean(vm_map_t, vaddr_t, vaddr_t, int);
379 vm_map_t	uvm_map_create(pmap_t, vaddr_t, vaddr_t, int);
380 int		uvm_map_extract(struct vm_map*, vaddr_t, vsize_t, vaddr_t*,
381 		    int);
382 vaddr_t		uvm_map_pie(vaddr_t);
383 vaddr_t		uvm_map_hint(struct vmspace *, vm_prot_t);
384 int		uvm_map_inherit(vm_map_t, vaddr_t, vaddr_t, vm_inherit_t);
385 int		uvm_map_advice(vm_map_t, vaddr_t, vaddr_t, int);
386 void		uvm_map_init(void);
387 boolean_t	uvm_map_lookup_entry(vm_map_t, vaddr_t, vm_map_entry_t *);
388 int		uvm_map_replace(vm_map_t, vaddr_t, vaddr_t,
389 		    vm_map_entry_t, int);
390 int		uvm_map_reserve(vm_map_t, vsize_t, vaddr_t, vsize_t,
391 		    vaddr_t *);
392 void		uvm_map_setup(vm_map_t, vaddr_t, vaddr_t, int);
393 int		uvm_map_submap(vm_map_t, vaddr_t, vaddr_t, vm_map_t);
394 void		uvm_unmap(vm_map_t, vaddr_t, vaddr_t);
395 void		uvm_map_set_uaddr(struct vm_map*, struct uvm_addr_state**,
396 		    struct uvm_addr_state*);
397 int		uvm_map_mquery(struct vm_map*, vaddr_t*, vsize_t, voff_t, int);
398 
399 void		uvm_unmap_detach(struct uvm_map_deadq*, int);
400 void		uvm_unmap_remove(struct vm_map*, vaddr_t, vaddr_t,
401 		    struct uvm_map_deadq*, boolean_t, boolean_t);
402 
403 #endif /* _KERNEL */
404 
405 /*
406  * VM map locking operations:
407  *
408  *	These operations perform locking on the data portion of the
409  *	map.
410  *
411  *	vm_map_lock_try: try to lock a map, failing if it is already locked.
412  *
413  *	vm_map_lock: acquire an exclusive (write) lock on a map.
414  *
415  *	vm_map_lock_read: acquire a shared (read) lock on a map.
416  *
417  *	vm_map_unlock: release an exclusive lock on a map.
418  *
419  *	vm_map_unlock_read: release a shared lock on a map.
420  *
421  *	vm_map_downgrade: downgrade an exclusive lock to a shared lock.
422  *
423  *	vm_map_upgrade: upgrade a shared lock to an exclusive lock.
424  *
425  *	vm_map_busy: mark a map as busy.
426  *
427  *	vm_map_unbusy: clear busy status on a map.
428  *
429  */
430 
431 #ifdef _KERNEL
432 /*
433  * XXX: clean up later
434  * Half the kernel seems to depend on them being included here.
435  */
436 #include <sys/time.h>
437 #include <sys/systm.h>  /* for panic() */
438 
439 boolean_t	vm_map_lock_try_ln(struct vm_map*, char*, int);
440 void		vm_map_lock_ln(struct vm_map*, char*, int);
441 void		vm_map_lock_read_ln(struct vm_map*, char*, int);
442 void		vm_map_unlock_ln(struct vm_map*, char*, int);
443 void		vm_map_unlock_read_ln(struct vm_map*, char*, int);
444 void		vm_map_downgrade_ln(struct vm_map*, char*, int);
445 void		vm_map_upgrade_ln(struct vm_map*, char*, int);
446 void		vm_map_busy_ln(struct vm_map*, char*, int);
447 void		vm_map_unbusy_ln(struct vm_map*, char*, int);
448 
449 #ifdef DIAGNOSTIC
450 #define vm_map_lock_try(map)	vm_map_lock_try_ln(map, __FILE__, __LINE__)
451 #define vm_map_lock(map)	vm_map_lock_ln(map, __FILE__, __LINE__)
452 #define vm_map_lock_read(map)	vm_map_lock_read_ln(map, __FILE__, __LINE__)
453 #define vm_map_unlock(map)	vm_map_unlock_ln(map, __FILE__, __LINE__)
454 #define vm_map_unlock_read(map)	vm_map_unlock_read_ln(map, __FILE__, __LINE__)
455 #define vm_map_downgrade(map)	vm_map_downgrade_ln(map, __FILE__, __LINE__)
456 #define vm_map_upgrade(map)	vm_map_upgrade_ln(map, __FILE__, __LINE__)
457 #define vm_map_busy(map)	vm_map_busy_ln(map, __FILE__, __LINE__)
458 #define vm_map_unbusy(map)	vm_map_unbusy_ln(map, __FILE__, __LINE__)
459 #else
460 #define vm_map_lock_try(map)	vm_map_lock_try_ln(map, NULL, 0)
461 #define vm_map_lock(map)	vm_map_lock_ln(map, NULL, 0)
462 #define vm_map_lock_read(map)	vm_map_lock_read_ln(map, NULL, 0)
463 #define vm_map_unlock(map)	vm_map_unlock_ln(map, NULL, 0)
464 #define vm_map_unlock_read(map)	vm_map_unlock_read_ln(map, NULL, 0)
465 #define vm_map_downgrade(map)	vm_map_downgrade_ln(map, NULL, 0)
466 #define vm_map_upgrade(map)	vm_map_upgrade_ln(map, NULL, 0)
467 #define vm_map_busy(map)	vm_map_busy_ln(map, NULL, 0)
468 #define vm_map_unbusy(map)	vm_map_unbusy_ln(map, NULL, 0)
469 #endif
470 
471 #endif /* _KERNEL */
472 
473 /*
474  *	Functions implemented as macros
475  */
476 #define		vm_map_min(map)		((map)->min_offset)
477 #define		vm_map_max(map)		((map)->max_offset)
478 #define		vm_map_pmap(map)	((map)->pmap)
479 
480 #endif /* _UVM_UVM_MAP_H_ */
481