xref: /dflybsd-src/sys/vm/vm_map.h (revision 744c01d0dc2aa1481a40e5b0988d15691602f5c9)
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 1.29 2007/01/11 10:15:21 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 _SYS_VKERNEL_H_
85 #include <sys/vkernel.h>
86 #endif
87 #ifndef _VM_VM_H_
88 #include <vm/vm.h>
89 #endif
90 #ifndef _MACHINE_PMAP_H_
91 #include <machine/pmap.h>
92 #endif
93 
94 struct vm_map_rb_tree;
95 RB_PROTOTYPE(vm_map_rb_tree, vm_map_entry, rb_entry, rb_vm_map_compare);
96 
97 /*
98  *	Types defined:
99  *
100  *	vm_map_t		the high-level address map data structure.
101  *	vm_map_entry_t		an entry in an address map.
102  */
103 
104 typedef u_int vm_eflags_t;
105 
106 /*
107  *	Objects which live in maps may be either VM objects, or
108  *	another map (called a "sharing map") which denotes read-write
109  *	sharing with other maps.
110  */
111 union vm_map_object {
112 	struct vm_object *vm_object;	/* object object */
113 	struct vm_map *sub_map;		/* belongs to another map */
114 };
115 
116 union vm_map_aux {
117 	vm_offset_t avail_ssize;	/* amt can grow if this is a stack */
118 	vpte_t master_pde;		/* virtual page table root */
119 };
120 
121 /*
122  *	Address map entries consist of start and end addresses,
123  *	a VM object (or sharing map) and offset into that object,
124  *	and user-exported inheritance and protection information.
125  *	Also included is control information for virtual copy operations.
126  *
127  *	When used with MAP_STACK, avail_ssize is used to determine the
128  *	limits of stack growth.
129  *
130  *	When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the
131  *	page directory index.
132  */
133 struct vm_map_entry {
134 	struct vm_map_entry *prev;	/* previous entry */
135 	struct vm_map_entry *next;	/* next entry */
136 	RB_ENTRY(vm_map_entry) rb_entry;
137 	vm_offset_t start;		/* start address */
138 	vm_offset_t end;		/* end address */
139 	union vm_map_aux aux;		/* auxillary data */
140 	union vm_map_object object;	/* object I point to */
141 	vm_ooffset_t offset;		/* offset into object */
142 	vm_eflags_t eflags;		/* map entry flags */
143 	vm_maptype_t maptype;		/* type of VM mapping */
144 	vm_prot_t protection;		/* protection code */
145 	vm_prot_t max_protection;	/* maximum protection */
146 	vm_inherit_t inheritance;	/* inheritance */
147 	int wired_count;		/* can be paged if = 0 */
148 	vm_pindex_t lastr;		/* last read */
149 };
150 
151 #define MAP_ENTRY_NOSYNC		0x0001
152 #define MAP_ENTRY_UNUSED0002		0x0002
153 #define MAP_ENTRY_COW			0x0004
154 #define MAP_ENTRY_NEEDS_COPY		0x0008
155 #define MAP_ENTRY_NOFAULT		0x0010
156 #define MAP_ENTRY_USER_WIRED		0x0020
157 
158 #define MAP_ENTRY_BEHAV_NORMAL		0x0000	/* default behavior */
159 #define MAP_ENTRY_BEHAV_SEQUENTIAL	0x0040	/* expect sequential access */
160 #define MAP_ENTRY_BEHAV_RANDOM		0x0080	/* expect random access */
161 #define MAP_ENTRY_BEHAV_RESERVED	0x00C0	/* future use */
162 
163 #define MAP_ENTRY_BEHAV_MASK		0x00C0
164 
165 #define MAP_ENTRY_IN_TRANSITION		0x0100	/* entry being changed */
166 #define MAP_ENTRY_NEEDS_WAKEUP		0x0200	/* waiter's in transition */
167 #define MAP_ENTRY_NOCOREDUMP		0x0400	/* don't include in a core */
168 
169 /*
170  * flags for vm_map_[un]clip_range()
171  */
172 #define MAP_CLIP_NO_HOLES		0x0001
173 
174 /*
175  * This reserve count for vm_map_entry_reserve() should cover all nominal
176  * single-insertion operations, including any necessary clipping.
177  */
178 #define MAP_RESERVE_COUNT	4
179 #define MAP_RESERVE_SLOP	32
180 
181 static __inline u_char
182 vm_map_entry_behavior(struct vm_map_entry *entry)
183 {
184 	return entry->eflags & MAP_ENTRY_BEHAV_MASK;
185 }
186 
187 static __inline void
188 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
189 {
190 	entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
191 		(behavior & MAP_ENTRY_BEHAV_MASK);
192 }
193 
194 /*
195  *	Maps are doubly-linked lists of map entries, kept sorted
196  *	by address.  A single hint is provided to start
197  *	searches again from the last successful search,
198  *	insertion, or removal.
199  *
200  *	Note: the lock structure cannot be the first element of vm_map
201  *	because this can result in a running lockup between two or more
202  *	system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
203  *	and free tsleep/waking up 'map' and the underlying lockmgr also
204  *	sleeping and waking up on 'map'.  The lockup occurs when the map fills
205  *	up.  The 'exec' map, for example.
206  */
207 struct vm_map {
208 	struct vm_map_entry header;	/* List of entries */
209 	RB_HEAD(vm_map_rb_tree, vm_map_entry) rb_root;
210 	struct lock lock;		/* Lock for map data */
211 	int nentries;			/* Number of entries */
212 	vm_size_t size;			/* virtual size */
213 	u_char system_map;		/* Am I a system map? */
214 	u_char infork;			/* Am I in fork processing? */
215 	vm_map_entry_t hint;		/* hint for quick lookups */
216 	unsigned int timestamp;		/* Version number */
217 	vm_map_entry_t first_free;	/* First free space hint */
218 	struct pmap *pmap;		/* Physical map */
219 #define	min_offset		header.start
220 #define max_offset		header.end
221 };
222 
223 /*
224  * Registered upcall
225  */
226 struct upcall;
227 
228 struct vmupcall {
229 	struct vmupcall	*vu_next;
230 	void		*vu_func;	/* user upcall function */
231 	void		*vu_data;	/* user data */
232 	void		*vu_ctx;	/* user context function */
233 	struct lwp	*vu_lwp;	/* process that registered upcall */
234 	int		vu_id;		/* upcall identifier */
235 	int		vu_pending;	/* upcall request pending */
236 };
237 
238 /*
239  * Shareable process virtual address space.
240  * May eventually be merged with vm_map.
241  * Several fields are temporary (text, data stuff).
242  */
243 struct vmspace {
244 	struct vm_map vm_map;	/* VM address map */
245 	struct pmap vm_pmap;	/* private physical map */
246 	int vm_refcnt;		/* number of references */
247 	caddr_t vm_shm;		/* SYS5 shared memory private data XXX */
248 /* we copy from vm_startcopy to the end of the structure on fork */
249 #define vm_startcopy vm_rssize
250 	segsz_t vm_rssize;	/* current resident set size in pages */
251 	segsz_t vm_swrss;	/* resident set size before last swap */
252 	segsz_t vm_tsize;	/* text size (pages) XXX */
253 	segsz_t vm_dsize;	/* data size (pages) XXX */
254 	segsz_t vm_ssize;	/* stack size (pages) */
255 	caddr_t vm_taddr;	/* user virtual address of text XXX */
256 	caddr_t vm_daddr;	/* user virtual address of data XXX */
257 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
258 	caddr_t vm_minsaddr;	/* user VA at max stack growth */
259 #define vm_endcopy	vm_exitingcnt
260 	int	vm_exitingcnt;	/* several procsses zombied in exit1 */
261 	int	vm_upccount;	/* number of registered upcalls */
262 	int	vm_pagesupply;
263 	struct vmupcall *vm_upcalls; /* registered upcalls */
264 };
265 
266 /*
267  * Resident executable holding structure.  A user program can take a snapshot
268  * of just its VM address space (typically done just after dynamic link
269  * libraries have completed loading) and register it as a resident
270  * executable associated with the program binary's vnode, which is also
271  * locked into memory.  Future execs of the vnode will start with a copy
272  * of the resident vmspace instead of running the binary from scratch,
273  * avoiding both the kernel ELF loader *AND* all shared library mapping and
274  * relocation code, and will call a different entry point (the stack pointer
275  * is reset to the top of the stack) supplied when the vmspace was registered.
276  */
277 struct vmresident {
278 	struct vnode	*vr_vnode;		/* associated vnode */
279 	TAILQ_ENTRY(vmresident) vr_link;	/* linked list of res sts */
280 	struct vmspace	*vr_vmspace;		/* vmspace to fork */
281 	intptr_t	vr_entry_addr;		/* registered entry point */
282 	struct sysentvec *vr_sysent;		/* system call vects */
283 	int		vr_id;			/* registration id */
284 };
285 
286 #ifdef _KERNEL
287 /*
288  *	Macros:		vm_map_lock, etc.
289  *	Function:
290  *		Perform locking on the data portion of a map.  Note that
291  *		these macros mimic procedure calls returning void.  The
292  *		semicolon is supplied by the user of these macros, not
293  *		by the macros themselves.  The macros can safely be used
294  *		as unbraced elements in a higher level statement.
295  */
296 
297 #ifdef DIAGNOSTIC
298 /* #define MAP_LOCK_DIAGNOSTIC 1 */
299 #ifdef MAP_LOCK_DIAGNOSTIC
300 #define	vm_map_lock(map) \
301 	do { \
302 		kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
303 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
304 			panic("vm_map_lock: failed to get lock"); \
305 		} \
306 		(map)->timestamp++; \
307 	} while(0)
308 #else
309 #define	vm_map_lock(map) \
310 	do { \
311 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
312 			panic("vm_map_lock: failed to get lock"); \
313 		} \
314 		(map)->timestamp++; \
315 	} while(0)
316 #endif
317 #else
318 #define	vm_map_lock(map) \
319 	do { \
320 		lockmgr(&(map)->lock, LK_EXCLUSIVE); \
321 		(map)->timestamp++; \
322 	} while(0)
323 #endif /* DIAGNOSTIC */
324 
325 #if defined(MAP_LOCK_DIAGNOSTIC)
326 #define	vm_map_unlock(map) \
327 	do { \
328 		kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
329 		lockmgr(&(map)->lock, LK_RELEASE); \
330 	} while (0)
331 #define	vm_map_lock_read(map) \
332 	do { \
333 		kprintf ("locking map LK_SHARED: 0x%x\n", map); \
334 		lockmgr(&(map)->lock, LK_SHARED); \
335 	} while (0)
336 #define	vm_map_unlock_read(map) \
337 	do { \
338 		kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
339 		lockmgr(&(map)->lock, LK_RELEASE); \
340 	} while (0)
341 #else
342 #define	vm_map_unlock(map) \
343 	lockmgr(&(map)->lock, LK_RELEASE)
344 #define	vm_map_lock_read(map) \
345 	lockmgr(&(map)->lock, LK_SHARED)
346 #define	vm_map_unlock_read(map) \
347 	lockmgr(&(map)->lock, LK_RELEASE)
348 #endif
349 
350 static __inline__ int
351 vm_map_lock_upgrade(vm_map_t map) {
352 	int error;
353 #if defined(MAP_LOCK_DIAGNOSTIC)
354 	kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
355 #endif
356 	error = lockmgr(&map->lock, LK_EXCLUPGRADE);
357 	if (error == 0)
358 		map->timestamp++;
359 	return error;
360 }
361 
362 #if defined(MAP_LOCK_DIAGNOSTIC)
363 #define vm_map_lock_downgrade(map) \
364 	do { \
365 		kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
366 		lockmgr(&(map)->lock, LK_DOWNGRADE); \
367 	} while (0)
368 #else
369 #define vm_map_lock_downgrade(map) \
370 	lockmgr(&(map)->lock, LK_DOWNGRADE)
371 #endif
372 
373 #endif /* _KERNEL */
374 
375 /*
376  *	Functions implemented as macros
377  */
378 #define		vm_map_min(map)		((map)->min_offset)
379 #define		vm_map_max(map)		((map)->max_offset)
380 #define		vm_map_pmap(map)	((map)->pmap)
381 
382 static __inline struct pmap *
383 vmspace_pmap(struct vmspace *vmspace)
384 {
385 	return &vmspace->vm_pmap;
386 }
387 
388 static __inline long
389 vmspace_resident_count(struct vmspace *vmspace)
390 {
391 	return pmap_resident_count(vmspace_pmap(vmspace));
392 }
393 
394 /*
395  * Number of kernel maps and entries to statically allocate, required
396  * during boot to bootstrap the VM system.
397  */
398 #define MAX_KMAP	10
399 #define	MAX_MAPENT	256
400 
401 /*
402  * Copy-on-write flags for vm_map operations
403  */
404 #define MAP_UNUSED_01		0x0001
405 #define MAP_COPY_ON_WRITE	0x0002
406 #define MAP_NOFAULT		0x0004
407 #define MAP_PREFAULT		0x0008
408 #define MAP_PREFAULT_PARTIAL	0x0010
409 #define MAP_DISABLE_SYNCER	0x0020
410 #define MAP_DISABLE_COREDUMP	0x0100
411 #define MAP_PREFAULT_MADVISE	0x0200	/* from (user) madvise request */
412 
413 /*
414  * vm_fault option flags
415  */
416 #define VM_FAULT_NORMAL		0x00	/* Nothing special */
417 #define VM_FAULT_CHANGE_WIRING	0x01	/* Change the wiring as appropriate */
418 #define VM_FAULT_USER_WIRE	0x02	/* Likewise, but for user purposes */
419 #define VM_FAULT_DIRTY		0x08	/* Dirty the page */
420 #define VM_FAULT_WIRE_MASK	(VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
421 
422 #ifdef _KERNEL
423 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t);
424 struct pmap;
425 struct globaldata;
426 void vm_map_entry_allocate_object(vm_map_entry_t);
427 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
428 int vm_map_entry_reserve(int);
429 int vm_map_entry_kreserve(int);
430 void vm_map_entry_release(int);
431 void vm_map_entry_krelease(int);
432 vm_map_t vm_map_create (vm_map_t, struct pmap *, vm_offset_t, vm_offset_t);
433 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
434 int vm_map_find (vm_map_t, vm_object_t, vm_ooffset_t,
435 		 vm_offset_t *, vm_size_t,
436 		 boolean_t,
437 		 vm_maptype_t,
438 		 vm_prot_t, vm_prot_t,
439 		 int);
440 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_offset_t, vm_offset_t *);
441 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
442 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t, pmap_t);
443 int vm_map_insert (vm_map_t, int *, vm_object_t, vm_ooffset_t,
444 		   vm_offset_t, vm_offset_t,
445 		   vm_maptype_t,
446 		   vm_prot_t, vm_prot_t,
447 		   int);
448 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
449     vm_pindex_t *, vm_prot_t *, boolean_t *);
450 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
451 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
452 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
453 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
454 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
455 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
456 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
457 void vm_map_startup (void);
458 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
459 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int, off_t);
460 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
461 void vm_init2 (void);
462 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
463 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, vm_prot_t, vm_prot_t, int);
464 int vm_map_growstack (struct proc *p, vm_offset_t addr);
465 int vmspace_swap_count (struct vmspace *vmspace);
466 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
467 
468 #endif
469 #endif				/* _VM_VM_MAP_H_ */
470