xref: /dragonfly/sys/vm/vm_map.h (revision fe76c4fb)
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 1.22 2006/05/20 02:42:15 dillon Exp $
66  */
67 
68 /*
69  *	Virtual memory map module definitions.
70  */
71 
72 #ifndef	_VM_VM_MAP_H_
73 #define	_VM_VM_MAP_H_
74 
75 #ifndef _SYS_TYPES_H_
76 #include <sys/types.h>
77 #endif
78 #ifndef _SYS_TREE_H_
79 #include <sys/tree.h>
80 #endif
81 #ifndef _SYS_LOCK_H_
82 #include <sys/lock.h>
83 #endif
84 #ifndef _VM_VM_H_
85 #include <vm/vm.h>
86 #endif
87 #ifndef _MACHINE_PMAP_H_
88 #include <machine/pmap.h>
89 #endif
90 
91 struct vm_map_rb_tree;
92 RB_PROTOTYPE(vm_map_rb_tree, vm_map_entry, rb_entry, rb_vm_map_compare);
93 
94 /*
95  *	Types defined:
96  *
97  *	vm_map_t		the high-level address map data structure.
98  *	vm_map_entry_t		an entry in an address map.
99  */
100 
101 typedef u_int vm_eflags_t;
102 
103 /*
104  *	Objects which live in maps may be either VM objects, or
105  *	another map (called a "sharing map") which denotes read-write
106  *	sharing with other maps.
107  */
108 
109 union vm_map_object {
110 	struct vm_object *vm_object;	/* object object */
111 	struct vm_map *sub_map;		/* belongs to another map */
112 };
113 
114 /*
115  *	Address map entries consist of start and end addresses,
116  *	a VM object (or sharing map) and offset into that object,
117  *	and user-exported inheritance and protection information.
118  *	Also included is control information for virtual copy operations.
119  */
120 struct vm_map_entry {
121 	struct vm_map_entry *prev;	/* previous entry */
122 	struct vm_map_entry *next;	/* next entry */
123 	RB_ENTRY(vm_map_entry) rb_entry;
124 	vm_offset_t start;		/* start address */
125 	vm_offset_t end;		/* end address */
126 	vm_offset_t avail_ssize;	/* amt can grow if this is a stack */
127 	union vm_map_object object;	/* object I point to */
128 	vm_ooffset_t offset;		/* offset into object */
129 	vm_eflags_t eflags;		/* map entry flags */
130 	/* Only in task maps: */
131 	vm_prot_t protection;		/* protection code */
132 	vm_prot_t max_protection;	/* maximum protection */
133 	vm_inherit_t inheritance;	/* inheritance */
134 	int wired_count;		/* can be paged if = 0 */
135 	vm_pindex_t lastr;		/* last read */
136 };
137 
138 #define MAP_ENTRY_NOSYNC		0x0001
139 #define MAP_ENTRY_IS_SUB_MAP		0x0002
140 #define MAP_ENTRY_COW			0x0004
141 #define MAP_ENTRY_NEEDS_COPY		0x0008
142 #define MAP_ENTRY_NOFAULT		0x0010
143 #define MAP_ENTRY_USER_WIRED		0x0020
144 
145 #define MAP_ENTRY_BEHAV_NORMAL		0x0000	/* default behavior */
146 #define MAP_ENTRY_BEHAV_SEQUENTIAL	0x0040	/* expect sequential access */
147 #define MAP_ENTRY_BEHAV_RANDOM		0x0080	/* expect random access */
148 #define MAP_ENTRY_BEHAV_RESERVED	0x00C0	/* future use */
149 
150 #define MAP_ENTRY_BEHAV_MASK		0x00C0
151 
152 #define MAP_ENTRY_IN_TRANSITION		0x0100	/* entry being changed */
153 #define MAP_ENTRY_NEEDS_WAKEUP		0x0200	/* waiter's in transition */
154 #define MAP_ENTRY_NOCOREDUMP		0x0400	/* don't include in a core */
155 
156 /*
157  * flags for vm_map_[un]clip_range()
158  */
159 #define MAP_CLIP_NO_HOLES		0x0001
160 
161 /*
162  * This reserve count for vm_map_entry_reserve() should cover all nominal
163  * single-insertion operations, including any necessary clipping.
164  */
165 #define MAP_RESERVE_COUNT	4
166 #define MAP_RESERVE_SLOP	32
167 
168 static __inline u_char
169 vm_map_entry_behavior(struct vm_map_entry *entry)
170 {
171 	return entry->eflags & MAP_ENTRY_BEHAV_MASK;
172 }
173 
174 static __inline void
175 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
176 {
177 	entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
178 		(behavior & MAP_ENTRY_BEHAV_MASK);
179 }
180 
181 /*
182  *	Maps are doubly-linked lists of map entries, kept sorted
183  *	by address.  A single hint is provided to start
184  *	searches again from the last successful search,
185  *	insertion, or removal.
186  *
187  *	Note: the lock structure cannot be the first element of vm_map
188  *	because this can result in a running lockup between two or more
189  *	system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
190  *	and free tsleep/waking up 'map' and the underlying lockmgr also
191  *	sleeping and waking up on 'map'.  The lockup occurs when the map fills
192  *	up.  The 'exec' map, for example.
193  */
194 struct vm_map {
195 	struct vm_map_entry header;	/* List of entries */
196 	RB_HEAD(vm_map_rb_tree, vm_map_entry) rb_root;
197 	struct lock lock;		/* Lock for map data */
198 	int nentries;			/* Number of entries */
199 	vm_size_t size;			/* virtual size */
200 	u_char system_map;		/* Am I a system map? */
201 	u_char infork;			/* Am I in fork processing? */
202 	vm_map_entry_t hint;		/* hint for quick lookups */
203 	unsigned int timestamp;		/* Version number */
204 	vm_map_entry_t first_free;	/* First free space hint */
205 	struct pmap *pmap;		/* Physical map */
206 #define	min_offset		header.start
207 #define max_offset		header.end
208 };
209 
210 /*
211  * Registered upcall
212  */
213 struct upcall;
214 
215 struct vmupcall {
216 	struct vmupcall	*vu_next;
217 	void		*vu_func;	/* user upcall function */
218 	void		*vu_data;	/* user data */
219 	void		*vu_ctx;	/* user context function */
220 	struct lwp	*vu_lwp;	/* process that registered upcall */
221 	int		vu_id;		/* upcall identifier */
222 	int		vu_pending;	/* upcall request pending */
223 };
224 
225 /*
226  * Shareable process virtual address space.
227  * May eventually be merged with vm_map.
228  * Several fields are temporary (text, data stuff).
229  */
230 struct vmspace {
231 	struct vm_map vm_map;	/* VM address map */
232 	struct pmap vm_pmap;	/* private physical map */
233 	int vm_refcnt;		/* number of references */
234 	caddr_t vm_shm;		/* SYS5 shared memory private data XXX */
235 /* we copy from vm_startcopy to the end of the structure on fork */
236 #define vm_startcopy vm_rssize
237 	segsz_t vm_rssize;	/* current resident set size in pages */
238 	segsz_t vm_swrss;	/* resident set size before last swap */
239 	segsz_t vm_tsize;	/* text size (pages) XXX */
240 	segsz_t vm_dsize;	/* data size (pages) XXX */
241 	segsz_t vm_ssize;	/* stack size (pages) */
242 	caddr_t vm_taddr;	/* user virtual address of text XXX */
243 	caddr_t vm_daddr;	/* user virtual address of data XXX */
244 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
245 	caddr_t vm_minsaddr;	/* user VA at max stack growth */
246 #define vm_endcopy	vm_exitingcnt
247 	int	vm_exitingcnt;	/* several procsses zombied in exit1 */
248 	int	vm_upccount;	/* number of registered upcalls */
249 	int	vm_pagesupply;
250 	struct vmupcall *vm_upcalls; /* registered upcalls */
251 };
252 
253 /*
254  * Resident executable holding structure.  A user program can take a snapshot
255  * of just its VM address space (typically done just after dynamic link
256  * libraries have completed loading) and register it as a resident
257  * executable associated with the program binary's vnode, which is also
258  * locked into memory.  Future execs of the vnode will start with a copy
259  * of the resident vmspace instead of running the binary from scratch,
260  * avoiding both the kernel ELF loader *AND* all shared library mapping and
261  * relocation code, and will call a different entry point (the stack pointer
262  * is reset to the top of the stack) supplied when the vmspace was registered.
263  */
264 struct vmresident {
265 	struct vnode	*vr_vnode;		/* associated vnode */
266 	TAILQ_ENTRY(vmresident) vr_link;	/* linked list of res sts */
267 	struct vmspace	*vr_vmspace;		/* vmspace to fork */
268 	intptr_t	vr_entry_addr;		/* registered entry point */
269 	struct sysentvec *vr_sysent;		/* system call vects */
270 	int		vr_id;			/* registration id */
271 };
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 #ifdef DIAGNOSTIC
285 /* #define MAP_LOCK_DIAGNOSTIC 1 */
286 #ifdef MAP_LOCK_DIAGNOSTIC
287 #define	vm_map_lock(map) \
288 	do { \
289 		printf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
290 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
291 			panic("vm_map_lock: failed to get lock"); \
292 		} \
293 		(map)->timestamp++; \
294 	} while(0)
295 #else
296 #define	vm_map_lock(map) \
297 	do { \
298 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
299 			panic("vm_map_lock: failed to get lock"); \
300 		} \
301 		(map)->timestamp++; \
302 	} while(0)
303 #endif
304 #else
305 #define	vm_map_lock(map) \
306 	do { \
307 		lockmgr(&(map)->lock, LK_EXCLUSIVE); \
308 		(map)->timestamp++; \
309 	} while(0)
310 #endif /* DIAGNOSTIC */
311 
312 #if defined(MAP_LOCK_DIAGNOSTIC)
313 #define	vm_map_unlock(map) \
314 	do { \
315 		printf ("locking map LK_RELEASE: 0x%x\n", map); \
316 		lockmgr(&(map)->lock, LK_RELEASE); \
317 	} while (0)
318 #define	vm_map_lock_read(map) \
319 	do { \
320 		printf ("locking map LK_SHARED: 0x%x\n", map); \
321 		lockmgr(&(map)->lock, LK_SHARED); \
322 	} while (0)
323 #define	vm_map_unlock_read(map) \
324 	do { \
325 		printf ("locking map LK_RELEASE: 0x%x\n", map); \
326 		lockmgr(&(map)->lock, LK_RELEASE); \
327 	} while (0)
328 #else
329 #define	vm_map_unlock(map) \
330 	lockmgr(&(map)->lock, LK_RELEASE)
331 #define	vm_map_lock_read(map) \
332 	lockmgr(&(map)->lock, LK_SHARED)
333 #define	vm_map_unlock_read(map) \
334 	lockmgr(&(map)->lock, LK_RELEASE)
335 #endif
336 
337 static __inline__ int
338 vm_map_lock_upgrade(vm_map_t map) {
339 	int error;
340 #if defined(MAP_LOCK_DIAGNOSTIC)
341 	printf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
342 #endif
343 	error = lockmgr(&map->lock, LK_EXCLUPGRADE);
344 	if (error == 0)
345 		map->timestamp++;
346 	return error;
347 }
348 
349 #if defined(MAP_LOCK_DIAGNOSTIC)
350 #define vm_map_lock_downgrade(map) \
351 	do { \
352 		printf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
353 		lockmgr(&(map)->lock, LK_DOWNGRADE); \
354 	} while (0)
355 #else
356 #define vm_map_lock_downgrade(map) \
357 	lockmgr(&(map)->lock, LK_DOWNGRADE)
358 #endif
359 
360 #endif /* _KERNEL */
361 
362 /*
363  *	Functions implemented as macros
364  */
365 #define		vm_map_min(map)		((map)->min_offset)
366 #define		vm_map_max(map)		((map)->max_offset)
367 #define		vm_map_pmap(map)	((map)->pmap)
368 
369 static __inline struct pmap *
370 vmspace_pmap(struct vmspace *vmspace)
371 {
372 	return &vmspace->vm_pmap;
373 }
374 
375 static __inline long
376 vmspace_resident_count(struct vmspace *vmspace)
377 {
378 	return pmap_resident_count(vmspace_pmap(vmspace));
379 }
380 
381 /*
382  * Number of kernel maps and entries to statically allocate, required
383  * during boot to bootstrap the VM system.
384  */
385 #define MAX_KMAP	10
386 #define	MAX_MAPENT	256
387 
388 /*
389  * Copy-on-write flags for vm_map operations
390  */
391 #define MAP_UNUSED_01		0x0001
392 #define MAP_COPY_ON_WRITE	0x0002
393 #define MAP_NOFAULT		0x0004
394 #define MAP_PREFAULT		0x0008
395 #define MAP_PREFAULT_PARTIAL	0x0010
396 #define MAP_DISABLE_SYNCER	0x0020
397 #define MAP_DISABLE_COREDUMP	0x0100
398 #define MAP_PREFAULT_MADVISE	0x0200	/* from (user) madvise request */
399 
400 /*
401  * vm_fault option flags
402  */
403 #define VM_FAULT_NORMAL 0		/* Nothing special */
404 #define VM_FAULT_CHANGE_WIRING 1	/* Change the wiring as appropriate */
405 #define VM_FAULT_USER_WIRE 2		/* Likewise, but for user purposes */
406 #define VM_FAULT_WIRE_MASK (VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
407 #define	VM_FAULT_HOLD 4			/* Hold the page */
408 #define VM_FAULT_DIRTY 8		/* Dirty the page */
409 
410 #ifdef _KERNEL
411 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t);
412 struct pmap;
413 struct globaldata;
414 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
415 int vm_map_entry_reserve(int);
416 int vm_map_entry_kreserve(int);
417 void vm_map_entry_release(int);
418 void vm_map_entry_krelease(int);
419 vm_map_t vm_map_create (struct pmap *, vm_offset_t, vm_offset_t);
420 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
421 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);
422 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_offset_t, vm_offset_t *);
423 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
424 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t);
425 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);
426 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
427     vm_pindex_t *, vm_prot_t *, boolean_t *);
428 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
429 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
430 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
431 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
432 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
433 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
434 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
435 void vm_map_startup (void);
436 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
437 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int);
438 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
439 void vm_init2 (void);
440 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
441 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, vm_prot_t, vm_prot_t, int);
442 int vm_map_growstack (struct proc *p, vm_offset_t addr);
443 int vmspace_swap_count (struct vmspace *vmspace);
444 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
445 
446 #endif
447 #endif				/* _VM_VM_MAP_H_ */
448