xref: /dragonfly/sys/vm/vm_map.h (revision 8a0bcd56)
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 1.30 2007/04/29 18:25:41 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_SYSREF_H_
82 #include <sys/sysref.h>
83 #endif
84 #ifndef _SYS_LOCK_H_
85 #include <sys/lock.h>
86 #endif
87 #ifndef _SYS_VKERNEL_H_
88 #include <sys/vkernel.h>
89 #endif
90 #ifndef _VM_VM_H_
91 #include <vm/vm.h>
92 #endif
93 #ifndef _MACHINE_PMAP_H_
94 #include <machine/pmap.h>
95 #endif
96 #ifndef _VM_VM_OBJECT_H_
97 #include <vm/vm_object.h>
98 #endif
99 #ifndef _SYS_NULL_H_
100 #include <sys/_null.h>
101 #endif
102 
103 struct vm_map_rb_tree;
104 RB_PROTOTYPE(vm_map_rb_tree, vm_map_entry, rb_entry, rb_vm_map_compare);
105 
106 /*
107  *	Types defined:
108  *
109  *	vm_map_t		the high-level address map data structure.
110  *	vm_map_entry_t		an entry in an address map.
111  */
112 
113 typedef u_int vm_eflags_t;
114 
115 /*
116  *	Objects which live in maps may be either VM objects, or
117  *	another map (called a "sharing map") which denotes read-write
118  *	sharing with other maps.
119  */
120 union vm_map_object {
121 	struct vm_object *vm_object;	/* object object */
122 	struct vm_map *sub_map;		/* belongs to another map */
123 };
124 
125 union vm_map_aux {
126 	vm_offset_t avail_ssize;	/* amt can grow if this is a stack */
127 	vpte_t master_pde;		/* virtual page table root */
128 };
129 
130 /*
131  *	Address map entries consist of start and end addresses,
132  *	a VM object (or sharing map) and offset into that object,
133  *	and user-exported inheritance and protection information.
134  *	Also included is control information for virtual copy operations.
135  *
136  *	When used with MAP_STACK, avail_ssize is used to determine the
137  *	limits of stack growth.
138  *
139  *	When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the
140  *	page directory index.
141  */
142 struct vm_map_entry {
143 	struct vm_map_entry *prev;	/* previous entry */
144 	struct vm_map_entry *next;	/* next entry */
145 	RB_ENTRY(vm_map_entry) rb_entry;
146 	vm_offset_t start;		/* start address */
147 	vm_offset_t end;		/* end address */
148 	union vm_map_aux aux;		/* auxillary data */
149 	union vm_map_object object;	/* object I point to */
150 	vm_ooffset_t offset;		/* offset into object */
151 	vm_eflags_t eflags;		/* map entry flags */
152 	vm_maptype_t maptype;		/* type of VM mapping */
153 	vm_prot_t protection;		/* protection code */
154 	vm_prot_t max_protection;	/* maximum protection */
155 	vm_inherit_t inheritance;	/* inheritance */
156 	int wired_count;		/* can be paged if = 0 */
157 };
158 
159 #define MAP_ENTRY_NOSYNC		0x0001
160 #define MAP_ENTRY_STACK			0x0002
161 #define MAP_ENTRY_COW			0x0004
162 #define MAP_ENTRY_NEEDS_COPY		0x0008
163 #define MAP_ENTRY_NOFAULT		0x0010
164 #define MAP_ENTRY_USER_WIRED		0x0020
165 
166 #define MAP_ENTRY_BEHAV_NORMAL		0x0000	/* default behavior */
167 #define MAP_ENTRY_BEHAV_SEQUENTIAL	0x0040	/* expect sequential access */
168 #define MAP_ENTRY_BEHAV_RANDOM		0x0080	/* expect random access */
169 #define MAP_ENTRY_BEHAV_RESERVED	0x00C0	/* future use */
170 
171 #define MAP_ENTRY_BEHAV_MASK		0x00C0
172 
173 #define MAP_ENTRY_IN_TRANSITION		0x0100	/* entry being changed */
174 #define MAP_ENTRY_NEEDS_WAKEUP		0x0200	/* waiter's in transition */
175 #define MAP_ENTRY_NOCOREDUMP		0x0400	/* don't include in a core */
176 
177 /*
178  * flags for vm_map_[un]clip_range()
179  */
180 #define MAP_CLIP_NO_HOLES		0x0001
181 
182 /*
183  * This reserve count for vm_map_entry_reserve() should cover all nominal
184  * single-insertion operations, including any necessary clipping.
185  */
186 #define MAP_RESERVE_COUNT	4
187 #define MAP_RESERVE_SLOP	32
188 
189 static __inline u_char
190 vm_map_entry_behavior(struct vm_map_entry *entry)
191 {
192 	return entry->eflags & MAP_ENTRY_BEHAV_MASK;
193 }
194 
195 static __inline void
196 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
197 {
198 	entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
199 		(behavior & MAP_ENTRY_BEHAV_MASK);
200 }
201 
202 /*
203  *	Maps are doubly-linked lists of map entries, kept sorted
204  *	by address.  A single hint is provided to start
205  *	searches again from the last successful search,
206  *	insertion, or removal.
207  *
208  *	Note: the lock structure cannot be the first element of vm_map
209  *	because this can result in a running lockup between two or more
210  *	system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
211  *	and free tsleep/waking up 'map' and the underlying lockmgr also
212  *	sleeping and waking up on 'map'.  The lockup occurs when the map fills
213  *	up.  The 'exec' map, for example.
214  */
215 struct vm_map {
216 	struct vm_map_entry header;	/* List of entries */
217 	RB_HEAD(vm_map_rb_tree, vm_map_entry) rb_root;
218 	struct lock lock;		/* Lock for map data */
219 	int nentries;			/* Number of entries */
220 	vm_size_t size;			/* virtual size */
221 	u_char system_map;		/* Am I a system map? */
222 	u_char infork;			/* Am I in fork processing? */
223 	vm_map_entry_t hint;		/* hint for quick lookups */
224 	unsigned int timestamp;		/* Version number */
225 	vm_map_entry_t first_free;	/* First free space hint */
226 	struct pmap *pmap;		/* Physical map */
227 #define	min_offset		header.start
228 #define max_offset		header.end
229 };
230 
231 /*
232  * Registered upcall
233  */
234 struct upcall;
235 
236 struct vmupcall {
237 	struct vmupcall	*vu_next;
238 	void		*vu_func;	/* user upcall function */
239 	void		*vu_data;	/* user data */
240 	void		*vu_ctx;	/* user context function */
241 	struct lwp	*vu_lwp;	/* process that registered upcall */
242 	int		vu_id;		/* upcall identifier */
243 	int		vu_pending;	/* upcall request pending */
244 };
245 
246 /*
247  * Shareable process virtual address space.
248  *
249  * Refd pointers from vmresident, proc
250  */
251 struct vmspace {
252 	struct vm_map vm_map;	/* VM address map */
253 	struct pmap vm_pmap;	/* private physical map */
254 	int vm_unused01;
255 	caddr_t vm_shm;		/* SYS5 shared memory private data XXX */
256 /* we copy from vm_startcopy to the end of the structure on fork */
257 #define vm_startcopy vm_rssize
258 	segsz_t vm_rssize;	/* current resident set size in pages */
259 	segsz_t vm_swrss;	/* resident set size before last swap */
260 	segsz_t vm_tsize;	/* text size (pages) XXX */
261 	segsz_t vm_dsize;	/* data size (pages) XXX */
262 	segsz_t vm_ssize;	/* stack size (pages) */
263 	caddr_t vm_taddr;	/* user virtual address of text XXX */
264 	caddr_t vm_daddr;	/* user virtual address of data XXX */
265 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
266 	caddr_t vm_minsaddr;	/* user VA at max stack growth */
267 #define vm_endcopy	vm_exitingcnt
268 	int	vm_exitingcnt;	/* several procsses zombied in exit1 */
269 	int	vm_upccount;	/* number of registered upcalls */
270 	int	vm_pagesupply;
271 	struct vmupcall *vm_upcalls;	/* registered upcalls */
272 	struct sysref vm_sysref;	/* sysref, refcnt, etc */
273 };
274 
275 /*
276  * Resident executable holding structure.  A user program can take a snapshot
277  * of just its VM address space (typically done just after dynamic link
278  * libraries have completed loading) and register it as a resident
279  * executable associated with the program binary's vnode, which is also
280  * locked into memory.  Future execs of the vnode will start with a copy
281  * of the resident vmspace instead of running the binary from scratch,
282  * avoiding both the kernel ELF loader *AND* all shared library mapping and
283  * relocation code, and will call a different entry point (the stack pointer
284  * is reset to the top of the stack) supplied when the vmspace was registered.
285  */
286 struct vmresident {
287 	struct vnode	*vr_vnode;		/* associated vnode */
288 	TAILQ_ENTRY(vmresident) vr_link;	/* linked list of res sts */
289 	struct vmspace	*vr_vmspace;		/* vmspace to fork */
290 	intptr_t	vr_entry_addr;		/* registered entry point */
291 	struct sysentvec *vr_sysent;		/* system call vects */
292 	int		vr_id;			/* registration id */
293 	int		vr_refs;		/* temporary refs */
294 };
295 
296 #ifdef _KERNEL
297 /*
298  *	Macros:		vm_map_lock, etc.
299  *	Function:
300  *		Perform locking on the data portion of a map.  Note that
301  *		these macros mimic procedure calls returning void.  The
302  *		semicolon is supplied by the user of these macros, not
303  *		by the macros themselves.  The macros can safely be used
304  *		as unbraced elements in a higher level statement.
305  */
306 
307 #define ASSERT_VM_MAP_LOCKED(map)	KKASSERT(lockowned(&(map)->lock))
308 
309 #ifdef DIAGNOSTIC
310 /* #define MAP_LOCK_DIAGNOSTIC 1 */
311 #ifdef MAP_LOCK_DIAGNOSTIC
312 #define	vm_map_lock(map) \
313 	do { \
314 		kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
315 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
316 			panic("vm_map_lock: failed to get lock"); \
317 		} \
318 		(map)->timestamp++; \
319 	} while(0)
320 #else
321 #define	vm_map_lock(map) \
322 	do { \
323 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
324 			panic("vm_map_lock: failed to get lock"); \
325 		} \
326 		(map)->timestamp++; \
327 	} while(0)
328 #endif
329 #else
330 #define	vm_map_lock(map) \
331 	do { \
332 		lockmgr(&(map)->lock, LK_EXCLUSIVE); \
333 		(map)->timestamp++; \
334 	} while(0)
335 #endif /* DIAGNOSTIC */
336 
337 #if defined(MAP_LOCK_DIAGNOSTIC)
338 #define	vm_map_unlock(map) \
339 	do { \
340 		kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
341 		lockmgr(&(map)->lock, LK_RELEASE); \
342 	} while (0)
343 #define	vm_map_lock_read(map) \
344 	do { \
345 		kprintf ("locking map LK_SHARED: 0x%x\n", map); \
346 		lockmgr(&(map)->lock, LK_SHARED); \
347 	} while (0)
348 #define	vm_map_unlock_read(map) \
349 	do { \
350 		kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
351 		lockmgr(&(map)->lock, LK_RELEASE); \
352 	} while (0)
353 #else
354 #define	vm_map_unlock(map) \
355 	lockmgr(&(map)->lock, LK_RELEASE)
356 #define	vm_map_lock_read(map) \
357 	lockmgr(&(map)->lock, LK_SHARED)
358 #define	vm_map_unlock_read(map) \
359 	lockmgr(&(map)->lock, LK_RELEASE)
360 #endif
361 
362 static __inline__ int
363 vm_map_lock_upgrade(vm_map_t map) {
364 	int error;
365 #if defined(MAP_LOCK_DIAGNOSTIC)
366 	kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
367 #endif
368 	error = lockmgr(&map->lock, LK_EXCLUPGRADE);
369 	if (error == 0)
370 		map->timestamp++;
371 	return error;
372 }
373 
374 #if defined(MAP_LOCK_DIAGNOSTIC)
375 #define vm_map_lock_downgrade(map) \
376 	do { \
377 		kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
378 		lockmgr(&(map)->lock, LK_DOWNGRADE); \
379 	} while (0)
380 #else
381 #define vm_map_lock_downgrade(map) \
382 	lockmgr(&(map)->lock, LK_DOWNGRADE)
383 #endif
384 
385 #endif /* _KERNEL */
386 
387 /*
388  *	Functions implemented as macros
389  */
390 #define		vm_map_min(map)		((map)->min_offset)
391 #define		vm_map_max(map)		((map)->max_offset)
392 #define		vm_map_pmap(map)	((map)->pmap)
393 
394 /*
395  * Must not block
396  */
397 static __inline struct pmap *
398 vmspace_pmap(struct vmspace *vmspace)
399 {
400 	return &vmspace->vm_pmap;
401 }
402 
403 static __inline long
404 vmspace_resident_count(struct vmspace *vmspace)
405 {
406 	return pmap_resident_count(vmspace_pmap(vmspace));
407 }
408 
409 /* Calculates the proportional RSS and returning the
410  * accrued result.
411  */
412 static __inline u_int
413 vmspace_president_count(struct vmspace *vmspace)
414 {
415 	vm_map_t map = &vmspace->vm_map;
416 	vm_map_entry_t cur;
417 	vm_object_t object;
418 	u_int count = 0;
419 
420 	for (cur = map->header.next; cur != &map->header; cur = cur->next) {
421 		switch(cur->maptype) {
422 		case VM_MAPTYPE_NORMAL:
423 		case VM_MAPTYPE_VPAGETABLE:
424 			if ((object = cur->object.vm_object) == NULL)
425 				break;
426 			if (object->type != OBJT_DEFAULT &&
427 			    object->type != OBJT_SWAP) {
428 				break;
429 			}
430 			if(object->agg_pv_list_count != 0) {
431 					count += (object->resident_page_count / object->agg_pv_list_count);
432 			}
433 			break;
434 		default:
435 			break;
436 		}
437 	}
438 	return(count);
439 }
440 
441 /*
442  * Number of kernel maps and entries to statically allocate, required
443  * during boot to bootstrap the VM system.
444  */
445 #define MAX_KMAP	10
446 #define	MAX_MAPENT	256
447 
448 /*
449  * Copy-on-write flags for vm_map operations
450  */
451 #define MAP_UNUSED_01		0x0001
452 #define MAP_COPY_ON_WRITE	0x0002
453 #define MAP_NOFAULT		0x0004
454 #define MAP_PREFAULT		0x0008
455 #define MAP_PREFAULT_PARTIAL	0x0010
456 #define MAP_DISABLE_SYNCER	0x0020
457 #define MAP_IS_STACK		0x0040
458 #define MAP_DISABLE_COREDUMP	0x0100
459 #define MAP_PREFAULT_MADVISE	0x0200	/* from (user) madvise request */
460 
461 /*
462  * vm_fault option flags
463  */
464 #define VM_FAULT_NORMAL		0x00	/* Nothing special */
465 #define VM_FAULT_CHANGE_WIRING	0x01	/* Change the wiring as appropriate */
466 #define VM_FAULT_USER_WIRE	0x02	/* Likewise, but for user purposes */
467 #define VM_FAULT_BURST		0x04	/* Burst fault can be done */
468 #define VM_FAULT_DIRTY		0x08	/* Dirty the page */
469 #define VM_FAULT_UNSWAP		0x10	/* Remove backing store from the page */
470 #define VM_FAULT_WIRE_MASK	(VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
471 
472 #ifdef _KERNEL
473 
474 extern struct sysref_class vmspace_sysref_class;
475 
476 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t,
477 		vm_prot_t, boolean_t);
478 struct pmap;
479 struct globaldata;
480 void vm_map_entry_allocate_object(vm_map_entry_t);
481 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
482 int vm_map_entry_reserve(int);
483 int vm_map_entry_kreserve(int);
484 void vm_map_entry_release(int);
485 void vm_map_entry_krelease(int);
486 vm_map_t vm_map_create (vm_map_t, struct pmap *, vm_offset_t, vm_offset_t);
487 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
488 int vm_map_find (vm_map_t, vm_object_t, vm_ooffset_t,
489 		 vm_offset_t *, vm_size_t, vm_size_t,
490 		 boolean_t, vm_maptype_t,
491 		 vm_prot_t, vm_prot_t,
492 		 int);
493 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_size_t,
494 		      int, vm_offset_t *);
495 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
496 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t, pmap_t);
497 int vm_map_insert (vm_map_t, int *, vm_object_t, vm_ooffset_t,
498 		   vm_offset_t, vm_offset_t,
499 		   vm_maptype_t,
500 		   vm_prot_t, vm_prot_t,
501 		   int);
502 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
503     vm_pindex_t *, vm_prot_t *, boolean_t *);
504 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
505 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
506 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
507 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
508 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
509 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
510 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
511 void vm_map_startup (void);
512 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
513 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int, off_t);
514 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
515 void vm_init2 (void);
516 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
517 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, int,
518 		  vm_prot_t, vm_prot_t, int);
519 int vm_map_growstack (struct proc *p, vm_offset_t addr);
520 int vmspace_swap_count (struct vmspace *vmspace);
521 int vmspace_anonymous_count (struct vmspace *vmspace);
522 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
523 
524 #endif
525 #endif				/* _VM_VM_MAP_H_ */
526