xref: /dragonfly/sys/vm/vm_map.h (revision 40f79625)
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  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vm_map.h	8.9 (Berkeley) 5/17/95
37  *
38  *
39  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
40  * All rights reserved.
41  *
42  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
43  *
44  * Permission to use, copy, modify and distribute this software and
45  * its documentation is hereby granted, provided that both the copyright
46  * notice and this permission notice appear in all copies of the
47  * software, derivative works or modified versions, and any portions
48  * thereof, and that both notices appear in supporting documentation.
49  *
50  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
51  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
52  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
53  *
54  * Carnegie Mellon requests users of this software to return to
55  *
56  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
57  *  School of Computer Science
58  *  Carnegie Mellon University
59  *  Pittsburgh PA 15213-3890
60  *
61  * any improvements or extensions that they make and grant Carnegie the
62  * rights to redistribute these changes.
63  *
64  * $FreeBSD: src/sys/vm/vm_map.h,v 1.54.2.5 2003/01/13 22:51:17 dillon Exp $
65  * $DragonFly: src/sys/vm/vm_map.h,v 1.18 2006/03/15 07:58:37 dillon Exp $
66  */
67 
68 /*
69  *	Virtual memory map module definitions.
70  */
71 
72 #ifndef	_VM_MAP_
73 #define	_VM_MAP_
74 
75 #include <sys/tree.h>
76 struct vm_map_rb_tree;
77 RB_PROTOTYPE(vm_map_rb_tree, vm_map_entry, rb_entry, rb_vm_map_compare);
78 
79 /*
80  *	Types defined:
81  *
82  *	vm_map_t		the high-level address map data structure.
83  *	vm_map_entry_t		an entry in an address map.
84  */
85 
86 typedef u_int vm_eflags_t;
87 
88 /*
89  *	Objects which live in maps may be either VM objects, or
90  *	another map (called a "sharing map") which denotes read-write
91  *	sharing with other maps.
92  */
93 
94 union vm_map_object {
95 	struct vm_object *vm_object;	/* object object */
96 	struct vm_map *sub_map;		/* belongs to another map */
97 };
98 
99 /*
100  *	Address map entries consist of start and end addresses,
101  *	a VM object (or sharing map) and offset into that object,
102  *	and user-exported inheritance and protection information.
103  *	Also included is control information for virtual copy operations.
104  */
105 struct vm_map_entry {
106 	struct vm_map_entry *prev;	/* previous entry */
107 	struct vm_map_entry *next;	/* next entry */
108 	RB_ENTRY(vm_map_entry) rb_entry;
109 	vm_offset_t start;		/* start address */
110 	vm_offset_t end;		/* end address */
111 	vm_offset_t avail_ssize;	/* amt can grow if this is a stack */
112 	union vm_map_object object;	/* object I point to */
113 	vm_ooffset_t offset;		/* offset into object */
114 	vm_eflags_t eflags;		/* map entry flags */
115 	/* Only in task maps: */
116 	vm_prot_t protection;		/* protection code */
117 	vm_prot_t max_protection;	/* maximum protection */
118 	vm_inherit_t inheritance;	/* inheritance */
119 	int wired_count;		/* can be paged if = 0 */
120 	vm_pindex_t lastr;		/* last read */
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	/* waiter's in transition */
139 #define MAP_ENTRY_NOCOREDUMP		0x0400	/* don't include in a core */
140 
141 /*
142  * flags for vm_map_[un]clip_range()
143  */
144 #define MAP_CLIP_NO_HOLES		0x0001
145 
146 /*
147  * This reserve count for vm_map_entry_reserve() should cover all nominal
148  * single-insertion operations, including any necessary clipping.
149  */
150 #define MAP_RESERVE_COUNT	4
151 #define MAP_RESERVE_SLOP	32
152 
153 static __inline u_char
154 vm_map_entry_behavior(struct vm_map_entry *entry)
155 {
156 	return entry->eflags & MAP_ENTRY_BEHAV_MASK;
157 }
158 
159 static __inline void
160 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
161 {
162 	entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
163 		(behavior & MAP_ENTRY_BEHAV_MASK);
164 }
165 
166 /*
167  *	Maps are doubly-linked lists of map entries, kept sorted
168  *	by address.  A single hint is provided to start
169  *	searches again from the last successful search,
170  *	insertion, or removal.
171  *
172  *	Note: the lock structure cannot be the first element of vm_map
173  *	because this can result in a running lockup between two or more
174  *	system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
175  *	and free tsleep/waking up 'map' and the underlying lockmgr also
176  *	sleeping and waking up on 'map'.  The lockup occurs when the map fills
177  *	up.  The 'exec' map, for example.
178  */
179 struct vm_map {
180 	struct vm_map_entry header;	/* List of entries */
181 	RB_HEAD(vm_map_rb_tree, vm_map_entry) rb_root;
182 	struct lock lock;		/* Lock for map data */
183 	int nentries;			/* Number of entries */
184 	vm_size_t size;			/* virtual size */
185 	u_char system_map;		/* Am I a system map? */
186 	u_char infork;			/* Am I in fork processing? */
187 	vm_map_entry_t hint;		/* hint for quick lookups */
188 	unsigned int timestamp;		/* Version number */
189 	vm_map_entry_t first_free;	/* First free space hint */
190 	struct pmap *pmap;		/* Physical map */
191 #define	min_offset		header.start
192 #define max_offset		header.end
193 };
194 
195 /*
196  * Registered upcall
197  */
198 struct upcall;
199 
200 struct vmupcall {
201 	struct vmupcall	*vu_next;
202 	void		*vu_func;	/* user upcall function */
203 	void		*vu_data;	/* user data */
204 	void		*vu_ctx;	/* user context function */
205 	struct lwp	*vu_lwp;	/* process that registered upcall */
206 	int		vu_id;		/* upcall identifier */
207 	int		vu_pending;	/* upcall request pending */
208 };
209 
210 /*
211  * Shareable process virtual address space.
212  * May eventually be merged with vm_map.
213  * Several fields are temporary (text, data stuff).
214  */
215 struct vmspace {
216 	struct vm_map vm_map;	/* VM address map */
217 	struct pmap vm_pmap;	/* private physical map */
218 	int vm_refcnt;		/* number of references */
219 	caddr_t vm_shm;		/* SYS5 shared memory private data XXX */
220 /* we copy from vm_startcopy to the end of the structure on fork */
221 #define vm_startcopy vm_rssize
222 	segsz_t vm_rssize;	/* current resident set size in pages */
223 	segsz_t vm_swrss;	/* resident set size before last swap */
224 	segsz_t vm_tsize;	/* text size (pages) XXX */
225 	segsz_t vm_dsize;	/* data size (pages) XXX */
226 	segsz_t vm_ssize;	/* stack size (pages) */
227 	caddr_t vm_taddr;	/* user virtual address of text XXX */
228 	caddr_t vm_daddr;	/* user virtual address of data XXX */
229 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
230 	caddr_t vm_minsaddr;	/* user VA at max stack growth */
231 #define vm_endcopy	vm_exitingcnt
232 	int	vm_exitingcnt;	/* several procsses zombied in exit1 */
233 	int	vm_upccount;	/* number of registered upcalls */
234 	int	vm_pagesupply;
235 	struct vmupcall *vm_upcalls; /* registered upcalls */
236 };
237 
238 /*
239  * Resident executable holding structure.  A user program can take a snapshot
240  * of just its VM address space (typically done just after dynamic link
241  * libraries have completed loading) and register it as a resident
242  * executable associated with the program binary's vnode, which is also
243  * locked into memory.  Future execs of the vnode will start with a copy
244  * of the resident vmspace instead of running the binary from scratch,
245  * avoiding both the kernel ELF loader *AND* all shared library mapping and
246  * relocation code, and will call a different entry point (the stack pointer
247  * is reset to the top of the stack) supplied when the vmspace was registered.
248  */
249 struct vmresident {
250 	struct vnode	*vr_vnode;		/* associated vnode */
251 	TAILQ_ENTRY(vmresident) vr_link;	/* linked list of res sts */
252 	struct vmspace	*vr_vmspace;		/* vmspace to fork */
253 	intptr_t	vr_entry_addr;		/* registered entry point */
254 	struct sysentvec *vr_sysent;		/* system call vects */
255 	int		vr_id;			/* registration id */
256 };
257 
258 #ifdef _KERNEL
259 /*
260  *	Macros:		vm_map_lock, etc.
261  *	Function:
262  *		Perform locking on the data portion of a map.  Note that
263  *		these macros mimic procedure calls returning void.  The
264  *		semicolon is supplied by the user of these macros, not
265  *		by the macros themselves.  The macros can safely be used
266  *		as unbraced elements in a higher level statement.
267  */
268 
269 #ifdef DIAGNOSTIC
270 /* #define MAP_LOCK_DIAGNOSTIC 1 */
271 #ifdef MAP_LOCK_DIAGNOSTIC
272 #define	vm_map_lock(map) \
273 	do { \
274 		printf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
275 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE, NULL, curthread) != 0) { \
276 			panic("vm_map_lock: failed to get lock"); \
277 		} \
278 		(map)->timestamp++; \
279 	} while(0)
280 #else
281 #define	vm_map_lock(map) \
282 	do { \
283 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE, NULL, curthread) != 0) { \
284 			panic("vm_map_lock: failed to get lock"); \
285 		} \
286 		(map)->timestamp++; \
287 	} while(0)
288 #endif
289 #else
290 #define	vm_map_lock(map) \
291 	do { \
292 		lockmgr(&(map)->lock, LK_EXCLUSIVE, NULL, curthread); \
293 		(map)->timestamp++; \
294 	} while(0)
295 #endif /* DIAGNOSTIC */
296 
297 #if defined(MAP_LOCK_DIAGNOSTIC)
298 #define	vm_map_unlock(map) \
299 	do { \
300 		printf ("locking map LK_RELEASE: 0x%x\n", map); \
301 		lockmgr(&(map)->lock, LK_RELEASE, NULL, curthread); \
302 	} while (0)
303 #define	vm_map_lock_read(map) \
304 	do { \
305 		printf ("locking map LK_SHARED: 0x%x\n", map); \
306 		lockmgr(&(map)->lock, LK_SHARED, NULL, curthread); \
307 	} while (0)
308 #define	vm_map_unlock_read(map) \
309 	do { \
310 		printf ("locking map LK_RELEASE: 0x%x\n", map); \
311 		lockmgr(&(map)->lock, LK_RELEASE, NULL, curthread); \
312 	} while (0)
313 #else
314 #define	vm_map_unlock(map) \
315 	lockmgr(&(map)->lock, LK_RELEASE, NULL, curthread)
316 #define	vm_map_lock_read(map) \
317 	lockmgr(&(map)->lock, LK_SHARED, NULL, curthread)
318 #define	vm_map_unlock_read(map) \
319 	lockmgr(&(map)->lock, LK_RELEASE, NULL, curthread)
320 #endif
321 
322 static __inline__ int
323 _vm_map_lock_upgrade(vm_map_t map, struct thread *td) {
324 	int error;
325 #if defined(MAP_LOCK_DIAGNOSTIC)
326 	printf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
327 #endif
328 	error = lockmgr(&map->lock, LK_EXCLUPGRADE, NULL, td);
329 	if (error == 0)
330 		map->timestamp++;
331 	return error;
332 }
333 
334 #define vm_map_lock_upgrade(map) _vm_map_lock_upgrade(map, curthread)
335 
336 #if defined(MAP_LOCK_DIAGNOSTIC)
337 #define vm_map_lock_downgrade(map) \
338 	do { \
339 		printf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
340 		lockmgr(&(map)->lock, LK_DOWNGRADE, NULL, curthread); \
341 	} while (0)
342 #else
343 #define vm_map_lock_downgrade(map) \
344 	lockmgr(&(map)->lock, LK_DOWNGRADE, NULL, curthread)
345 #endif
346 
347 #endif /* _KERNEL */
348 
349 /*
350  *	Functions implemented as macros
351  */
352 #define		vm_map_min(map)		((map)->min_offset)
353 #define		vm_map_max(map)		((map)->max_offset)
354 #define		vm_map_pmap(map)	((map)->pmap)
355 
356 static __inline struct pmap *
357 vmspace_pmap(struct vmspace *vmspace)
358 {
359 	return &vmspace->vm_pmap;
360 }
361 
362 static __inline long
363 vmspace_resident_count(struct vmspace *vmspace)
364 {
365 	return pmap_resident_count(vmspace_pmap(vmspace));
366 }
367 
368 /*
369  * Number of kernel maps and entries to statically allocate, required
370  * during boot to bootstrap the VM system.
371  */
372 #define MAX_KMAP	10
373 #define	MAX_MAPENT	256
374 
375 /*
376  * Copy-on-write flags for vm_map operations
377  */
378 #define MAP_UNUSED_01		0x0001
379 #define MAP_COPY_ON_WRITE	0x0002
380 #define MAP_NOFAULT		0x0004
381 #define MAP_PREFAULT		0x0008
382 #define MAP_PREFAULT_PARTIAL	0x0010
383 #define MAP_DISABLE_SYNCER	0x0020
384 #define MAP_DISABLE_COREDUMP	0x0100
385 #define MAP_PREFAULT_MADVISE	0x0200	/* from (user) madvise request */
386 
387 /*
388  * vm_fault option flags
389  */
390 #define VM_FAULT_NORMAL 0		/* Nothing special */
391 #define VM_FAULT_CHANGE_WIRING 1	/* Change the wiring as appropriate */
392 #define VM_FAULT_USER_WIRE 2		/* Likewise, but for user purposes */
393 #define VM_FAULT_WIRE_MASK (VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
394 #define	VM_FAULT_HOLD 4			/* Hold the page */
395 #define VM_FAULT_DIRTY 8		/* Dirty the page */
396 
397 #ifdef _KERNEL
398 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t);
399 struct pmap;
400 struct globaldata;
401 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
402 int vm_map_entry_reserve(int);
403 int vm_map_entry_kreserve(int);
404 void vm_map_entry_release(int);
405 void vm_map_entry_krelease(int);
406 vm_map_t vm_map_create (struct pmap *, vm_offset_t, vm_offset_t);
407 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
408 int vm_map_find (vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *, vm_size_t, boolean_t, vm_prot_t, vm_prot_t, int);
409 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_offset_t, vm_offset_t *);
410 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
411 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t);
412 int vm_map_insert (vm_map_t, int *, vm_object_t, vm_ooffset_t, vm_offset_t, vm_offset_t, vm_prot_t, vm_prot_t, int);
413 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
414     vm_pindex_t *, vm_prot_t *, boolean_t *);
415 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
416 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
417 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
418 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
419 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
420 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
421 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
422 void vm_map_startup (void);
423 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
424 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int);
425 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
426 void vm_init2 (void);
427 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
428 void vm_freeze_copyopts (vm_object_t, vm_pindex_t, vm_pindex_t);
429 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, vm_prot_t, vm_prot_t, int);
430 int vm_map_growstack (struct proc *p, vm_offset_t addr);
431 int vmspace_swap_count (struct vmspace *vmspace);
432 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
433 
434 #endif
435 #endif				/* _VM_MAP_ */
436