xref: /dragonfly/sys/vm/vm_map.h (revision 81c11cd3)
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_flags_t;
114 typedef u_int vm_eflags_t;
115 
116 /*
117  *	Objects which live in maps may be either VM objects, or
118  *	another map (called a "sharing map") which denotes read-write
119  *	sharing with other maps.
120  */
121 union vm_map_object {
122 	struct vm_object *vm_object;	/* object object */
123 	struct vm_map *sub_map;		/* belongs to another map */
124 };
125 
126 union vm_map_aux {
127 	vm_offset_t avail_ssize;	/* amt can grow if this is a stack */
128 	vpte_t master_pde;		/* virtual page table root */
129 };
130 
131 /*
132  *	Address map entries consist of start and end addresses,
133  *	a VM object (or sharing map) and offset into that object,
134  *	and user-exported inheritance and protection information.
135  *	Also included is control information for virtual copy operations.
136  *
137  *	When used with MAP_STACK, avail_ssize is used to determine the
138  *	limits of stack growth.
139  *
140  *	When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the
141  *	page directory index.
142  */
143 struct vm_map_entry {
144 	struct vm_map_entry *prev;	/* previous entry */
145 	struct vm_map_entry *next;	/* next entry */
146 	RB_ENTRY(vm_map_entry) rb_entry;
147 	vm_offset_t start;		/* start address */
148 	vm_offset_t end;		/* end address */
149 	union vm_map_aux aux;		/* auxillary data */
150 	union vm_map_object object;	/* object I point to */
151 	vm_ooffset_t offset;		/* offset into object */
152 	vm_eflags_t eflags;		/* map entry flags */
153 	vm_maptype_t maptype;		/* type of VM mapping */
154 	vm_prot_t protection;		/* protection code */
155 	vm_prot_t max_protection;	/* maximum protection */
156 	vm_inherit_t inheritance;	/* inheritance */
157 	int wired_count;		/* can be paged if = 0 */
158 };
159 
160 #define MAP_ENTRY_NOSYNC		0x0001
161 #define MAP_ENTRY_STACK			0x0002
162 #define MAP_ENTRY_COW			0x0004
163 #define MAP_ENTRY_NEEDS_COPY		0x0008
164 #define MAP_ENTRY_NOFAULT		0x0010
165 #define MAP_ENTRY_USER_WIRED		0x0020
166 
167 #define MAP_ENTRY_BEHAV_NORMAL		0x0000	/* default behavior */
168 #define MAP_ENTRY_BEHAV_SEQUENTIAL	0x0040	/* expect sequential access */
169 #define MAP_ENTRY_BEHAV_RANDOM		0x0080	/* expect random access */
170 #define MAP_ENTRY_BEHAV_RESERVED	0x00C0	/* future use */
171 
172 #define MAP_ENTRY_BEHAV_MASK		0x00C0
173 
174 #define MAP_ENTRY_IN_TRANSITION		0x0100	/* entry being changed */
175 #define MAP_ENTRY_NEEDS_WAKEUP		0x0200	/* waiter's in transition */
176 #define MAP_ENTRY_NOCOREDUMP		0x0400	/* don't include in a core */
177 #define MAP_ENTRY_KSTACK		0x0800	/* guarded kernel stack */
178 
179 /*
180  * flags for vm_map_[un]clip_range()
181  */
182 #define MAP_CLIP_NO_HOLES		0x0001
183 
184 /*
185  * This reserve count for vm_map_entry_reserve() should cover all nominal
186  * single-insertion operations, including any necessary clipping.
187  */
188 #define MAP_RESERVE_COUNT	4
189 #define MAP_RESERVE_SLOP	32
190 
191 static __inline u_char
192 vm_map_entry_behavior(struct vm_map_entry *entry)
193 {
194 	return entry->eflags & MAP_ENTRY_BEHAV_MASK;
195 }
196 
197 static __inline void
198 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
199 {
200 	entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
201 		(behavior & MAP_ENTRY_BEHAV_MASK);
202 }
203 
204 /*
205  *	Maps are doubly-linked lists of map entries, kept sorted
206  *	by address.  A single hint is provided to start
207  *	searches again from the last successful search,
208  *	insertion, or removal.
209  *
210  *	Note: the lock structure cannot be the first element of vm_map
211  *	because this can result in a running lockup between two or more
212  *	system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
213  *	and free tsleep/waking up 'map' and the underlying lockmgr also
214  *	sleeping and waking up on 'map'.  The lockup occurs when the map fills
215  *	up.  The 'exec' map, for example.
216  */
217 struct vm_map {
218 	struct vm_map_entry header;	/* List of entries */
219 	RB_HEAD(vm_map_rb_tree, vm_map_entry) rb_root;
220 	struct lock lock;		/* Lock for map data */
221 	int nentries;			/* Number of entries */
222 	vm_size_t size;			/* virtual size */
223 	u_char system_map;		/* Am I a system map? */
224 	u_char infork;			/* Am I in fork processing? */
225 	vm_map_entry_t hint;		/* hint for quick lookups */
226 	unsigned int timestamp;		/* Version number */
227 	vm_map_entry_t first_free;	/* First free space hint */
228 	vm_flags_t flags;		/* flags for this vm_map */
229 	struct pmap *pmap;		/* Physical map */
230 #define	min_offset		header.start
231 #define max_offset		header.end
232 };
233 
234 /*
235  * vm_flags_t values
236  */
237 #define MAP_WIREFUTURE		0x01	/* wire all future pages */
238 
239 /*
240  * Registered upcall
241  */
242 struct upcall;
243 
244 struct vmupcall {
245 	struct vmupcall	*vu_next;
246 	void		*vu_func;	/* user upcall function */
247 	void		*vu_data;	/* user data */
248 	void		*vu_ctx;	/* user context function */
249 	struct lwp	*vu_lwp;	/* process that registered upcall */
250 	int		vu_id;		/* upcall identifier */
251 	int		vu_pending;	/* upcall request pending */
252 };
253 
254 /*
255  * Shareable process virtual address space.
256  *
257  * Refd pointers from vmresident, proc
258  */
259 struct vmspace {
260 	struct vm_map vm_map;	/* VM address map */
261 	struct pmap vm_pmap;	/* private physical map */
262 	int vm_unused01;
263 	caddr_t vm_shm;		/* SYS5 shared memory private data XXX */
264 /* we copy from vm_startcopy to the end of the structure on fork */
265 #define vm_startcopy vm_rssize
266 	segsz_t vm_rssize;	/* current resident set size in pages */
267 	segsz_t vm_swrss;	/* resident set size before last swap */
268 	segsz_t vm_tsize;	/* text size (pages) XXX */
269 	segsz_t vm_dsize;	/* data size (pages) XXX */
270 	segsz_t vm_ssize;	/* stack size (pages) */
271 	caddr_t vm_taddr;	/* user virtual address of text XXX */
272 	caddr_t vm_daddr;	/* user virtual address of data XXX */
273 	caddr_t vm_maxsaddr;	/* user VA at max stack growth */
274 	caddr_t vm_minsaddr;	/* user VA at max stack growth */
275 #define vm_endcopy	vm_exitingcnt
276 	int	vm_exitingcnt;	/* several procsses zombied in exit1 */
277 	int	vm_upccount;	/* number of registered upcalls */
278 	int	vm_pagesupply;
279 	struct vmupcall *vm_upcalls;	/* registered upcalls */
280 	struct sysref vm_sysref;	/* sysref, refcnt, etc */
281 };
282 
283 /*
284  * Resident executable holding structure.  A user program can take a snapshot
285  * of just its VM address space (typically done just after dynamic link
286  * libraries have completed loading) and register it as a resident
287  * executable associated with the program binary's vnode, which is also
288  * locked into memory.  Future execs of the vnode will start with a copy
289  * of the resident vmspace instead of running the binary from scratch,
290  * avoiding both the kernel ELF loader *AND* all shared library mapping and
291  * relocation code, and will call a different entry point (the stack pointer
292  * is reset to the top of the stack) supplied when the vmspace was registered.
293  */
294 struct vmresident {
295 	struct vnode	*vr_vnode;		/* associated vnode */
296 	TAILQ_ENTRY(vmresident) vr_link;	/* linked list of res sts */
297 	struct vmspace	*vr_vmspace;		/* vmspace to fork */
298 	intptr_t	vr_entry_addr;		/* registered entry point */
299 	struct sysentvec *vr_sysent;		/* system call vects */
300 	int		vr_id;			/* registration id */
301 	int		vr_refs;		/* temporary refs */
302 };
303 
304 #ifdef _KERNEL
305 /*
306  *	Macros:		vm_map_lock, etc.
307  *	Function:
308  *		Perform locking on the data portion of a map.  Note that
309  *		these macros mimic procedure calls returning void.  The
310  *		semicolon is supplied by the user of these macros, not
311  *		by the macros themselves.  The macros can safely be used
312  *		as unbraced elements in a higher level statement.
313  */
314 
315 #define ASSERT_VM_MAP_LOCKED(map)	KKASSERT(lockowned(&(map)->lock))
316 
317 #ifdef DIAGNOSTIC
318 /* #define MAP_LOCK_DIAGNOSTIC 1 */
319 #ifdef MAP_LOCK_DIAGNOSTIC
320 #define	vm_map_lock(map) \
321 	do { \
322 		kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
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 #else
329 #define	vm_map_lock(map) \
330 	do { \
331 		if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
332 			panic("vm_map_lock: failed to get lock"); \
333 		} \
334 		(map)->timestamp++; \
335 	} while(0)
336 #endif
337 #else
338 #define	vm_map_lock(map) \
339 	do { \
340 		lockmgr(&(map)->lock, LK_EXCLUSIVE); \
341 		(map)->timestamp++; \
342 	} while(0)
343 #endif /* DIAGNOSTIC */
344 
345 #if defined(MAP_LOCK_DIAGNOSTIC)
346 #define	vm_map_unlock(map) \
347 	do { \
348 		kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
349 		lockmgr(&(map)->lock, LK_RELEASE); \
350 	} while (0)
351 #define	vm_map_lock_read(map) \
352 	do { \
353 		kprintf ("locking map LK_SHARED: 0x%x\n", map); \
354 		lockmgr(&(map)->lock, LK_SHARED); \
355 	} while (0)
356 #define	vm_map_unlock_read(map) \
357 	do { \
358 		kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
359 		lockmgr(&(map)->lock, LK_RELEASE); \
360 	} while (0)
361 #else
362 #define	vm_map_unlock(map) \
363 	lockmgr(&(map)->lock, LK_RELEASE)
364 #define	vm_map_lock_read(map) \
365 	lockmgr(&(map)->lock, LK_SHARED)
366 #define	vm_map_unlock_read(map) \
367 	lockmgr(&(map)->lock, LK_RELEASE)
368 #endif
369 
370 static __inline__ int
371 vm_map_lock_upgrade(vm_map_t map) {
372 	int error;
373 #if defined(MAP_LOCK_DIAGNOSTIC)
374 	kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
375 #endif
376 	error = lockmgr(&map->lock, LK_EXCLUPGRADE);
377 	if (error == 0)
378 		map->timestamp++;
379 	return error;
380 }
381 
382 #if defined(MAP_LOCK_DIAGNOSTIC)
383 #define vm_map_lock_downgrade(map) \
384 	do { \
385 		kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
386 		lockmgr(&(map)->lock, LK_DOWNGRADE); \
387 	} while (0)
388 #else
389 #define vm_map_lock_downgrade(map) \
390 	lockmgr(&(map)->lock, LK_DOWNGRADE)
391 #endif
392 
393 #endif /* _KERNEL */
394 
395 /*
396  *	Functions implemented as macros
397  */
398 #define		vm_map_min(map)		((map)->min_offset)
399 #define		vm_map_max(map)		((map)->max_offset)
400 #define		vm_map_pmap(map)	((map)->pmap)
401 
402 /*
403  * Must not block
404  */
405 static __inline struct pmap *
406 vmspace_pmap(struct vmspace *vmspace)
407 {
408 	return &vmspace->vm_pmap;
409 }
410 
411 static __inline long
412 vmspace_resident_count(struct vmspace *vmspace)
413 {
414 	return pmap_resident_count(vmspace_pmap(vmspace));
415 }
416 
417 /* Calculates the proportional RSS and returning the
418  * accrued result.
419  */
420 static __inline u_int
421 vmspace_president_count(struct vmspace *vmspace)
422 {
423 	vm_map_t map = &vmspace->vm_map;
424 	vm_map_entry_t cur;
425 	vm_object_t object;
426 	u_int count = 0;
427 
428 	for (cur = map->header.next; cur != &map->header; cur = cur->next) {
429 		switch(cur->maptype) {
430 		case VM_MAPTYPE_NORMAL:
431 		case VM_MAPTYPE_VPAGETABLE:
432 			if ((object = cur->object.vm_object) == NULL)
433 				break;
434 			if (object->type != OBJT_DEFAULT &&
435 			    object->type != OBJT_SWAP) {
436 				break;
437 			}
438 			if(object->agg_pv_list_count != 0) {
439 					count += (object->resident_page_count / object->agg_pv_list_count);
440 			}
441 			break;
442 		default:
443 			break;
444 		}
445 	}
446 	return(count);
447 }
448 
449 /*
450  * Number of kernel maps and entries to statically allocate, required
451  * during boot to bootstrap the VM system.
452  */
453 #define MAX_KMAP	10
454 #define	MAX_MAPENT	2048	/* required to support up to 64 cpus */
455 
456 /*
457  * Copy-on-write flags for vm_map operations
458  */
459 #define MAP_UNUSED_01		0x0001
460 #define MAP_COPY_ON_WRITE	0x0002
461 #define MAP_NOFAULT		0x0004
462 #define MAP_PREFAULT		0x0008
463 #define MAP_PREFAULT_PARTIAL	0x0010
464 #define MAP_DISABLE_SYNCER	0x0020
465 #define MAP_IS_STACK		0x0040
466 #define MAP_IS_KSTACK		0x0080
467 #define MAP_DISABLE_COREDUMP	0x0100
468 #define MAP_PREFAULT_MADVISE	0x0200	/* from (user) madvise request */
469 
470 /*
471  * vm_fault option flags
472  */
473 #define VM_FAULT_NORMAL		0x00	/* Nothing special */
474 #define VM_FAULT_CHANGE_WIRING	0x01	/* Change the wiring as appropriate */
475 #define VM_FAULT_USER_WIRE	0x02	/* Likewise, but for user purposes */
476 #define VM_FAULT_BURST		0x04	/* Burst fault can be done */
477 #define VM_FAULT_DIRTY		0x08	/* Dirty the page */
478 #define VM_FAULT_UNSWAP		0x10	/* Remove backing store from the page */
479 #define VM_FAULT_WIRE_MASK	(VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
480 
481 #ifdef _KERNEL
482 
483 extern struct sysref_class vmspace_sysref_class;
484 
485 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t,
486 		vm_prot_t, boolean_t);
487 struct pmap;
488 struct globaldata;
489 void vm_map_entry_allocate_object(vm_map_entry_t);
490 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
491 int vm_map_entry_reserve(int);
492 int vm_map_entry_kreserve(int);
493 void vm_map_entry_release(int);
494 void vm_map_entry_krelease(int);
495 vm_map_t vm_map_create (vm_map_t, struct pmap *, vm_offset_t, vm_offset_t);
496 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
497 int vm_map_find (vm_map_t, vm_object_t, vm_ooffset_t,
498 		 vm_offset_t *, vm_size_t, vm_size_t,
499 		 boolean_t, vm_maptype_t,
500 		 vm_prot_t, vm_prot_t,
501 		 int);
502 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_size_t,
503 		      int, vm_offset_t *);
504 vm_offset_t vm_map_hint(struct proc *, vm_offset_t, vm_prot_t);
505 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
506 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t, pmap_t);
507 int vm_map_insert (vm_map_t, int *, vm_object_t, vm_ooffset_t,
508 		   vm_offset_t, vm_offset_t,
509 		   vm_maptype_t,
510 		   vm_prot_t, vm_prot_t,
511 		   int);
512 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
513     vm_pindex_t *, vm_prot_t *, boolean_t *);
514 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
515 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
516 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
517 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
518 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
519 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
520 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
521 void vm_map_startup (void);
522 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
523 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int, off_t);
524 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
525 void vm_init2 (void);
526 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
527 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, int,
528 		  vm_prot_t, vm_prot_t, int);
529 int vm_map_growstack (struct proc *p, vm_offset_t addr);
530 int vmspace_swap_count (struct vmspace *vmspace);
531 int vmspace_anonymous_count (struct vmspace *vmspace);
532 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
533 void vm_map_transition_wait(vm_map_t map);
534 
535 
536 #endif
537 #endif				/* _VM_VM_MAP_H_ */
538