2 * Copyright (c) 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * The Mach Operating System project at Carnegie-Mellon University.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)vm_map.h 8.9 (Berkeley) 5/17/95
35 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
36 * All rights reserved.
38 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
40 * Permission to use, copy, modify and distribute this software and
41 * its documentation is hereby granted, provided that both the copyright
42 * notice and this permission notice appear in all copies of the
43 * software, derivative works or modified versions, and any portions
44 * thereof, and that both notices appear in supporting documentation.
46 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
47 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
48 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
50 * Carnegie Mellon requests users of this software to return to
52 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
53 * School of Computer Science
54 * Carnegie Mellon University
55 * Pittsburgh PA 15213-3890
57 * any improvements or extensions that they make and grant Carnegie the
58 * rights to redistribute these changes.
60 * $FreeBSD: src/sys/vm/vm_map.h,v 1.54.2.5 2003/01/13 22:51:17 dillon Exp $
64 * Virtual memory map module definitions.
71 #include <sys/types.h>
74 #ifndef _SYS_KERNEL_H_
75 #include <sys/kernel.h> /* ticks */
81 #ifndef _SYS_SYSREF_H_
82 #include <sys/sysref.h>
87 #ifndef _SYS_VKERNEL_H_
88 #include <sys/vkernel.h>
93 #ifndef _MACHINE_PMAP_H_
94 #include <machine/pmap.h>
96 #ifndef _VM_VM_OBJECT_H_
97 #include <vm/vm_object.h>
100 #include <sys/_null.h>
103 struct vm_map_rb_tree
;
104 RB_PROTOTYPE(vm_map_rb_tree
, vm_map_entry
, rb_entry
, rb_vm_map_compare
);
109 * vm_map_t the high-level address map data structure.
110 * vm_map_entry_t an entry in an address map.
113 typedef u_int vm_flags_t
;
114 typedef u_int vm_eflags_t
;
117 * A vm_map_entry may reference an object, a submap, a uksmap, or a
118 * direct user-kernel shared map.
120 union vm_map_object
{
121 struct vm_object
*vm_object
; /* object object */
122 struct vm_map
*sub_map
; /* belongs to another map */
123 int (*uksmap
)(struct cdev
*dev
, vm_page_t fake
);
124 void *map_object
; /* generic */
128 vm_offset_t avail_ssize
; /* amt can grow if this is a stack */
129 vpte_t master_pde
; /* virtual page table root */
135 * Address map entries consist of start and end addresses,
136 * a VM object (or sharing map) and offset into that object,
137 * and user-exported inheritance and protection information.
138 * Also included is control information for virtual copy operations.
140 * When used with MAP_STACK, avail_ssize is used to determine the
141 * limits of stack growth.
143 * When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the
144 * page directory index.
146 struct vm_map_entry
{
147 struct vm_map_entry
*prev
; /* previous entry */
148 struct vm_map_entry
*next
; /* next entry */
149 RB_ENTRY(vm_map_entry
) rb_entry
;
150 vm_offset_t start
; /* start address */
151 vm_offset_t end
; /* end address */
152 union vm_map_aux aux
; /* auxillary data */
153 union vm_map_object object
; /* object I point to */
154 vm_ooffset_t offset
; /* offset into object */
155 vm_eflags_t eflags
; /* map entry flags */
156 vm_maptype_t maptype
; /* type of VM mapping */
157 vm_prot_t protection
; /* protection code */
158 vm_prot_t max_protection
; /* maximum protection */
159 vm_inherit_t inheritance
; /* inheritance */
160 int wired_count
; /* can be paged if = 0 */
163 #define MAP_ENTRY_NOSYNC 0x0001
164 #define MAP_ENTRY_STACK 0x0002
165 #define MAP_ENTRY_COW 0x0004
166 #define MAP_ENTRY_NEEDS_COPY 0x0008
167 #define MAP_ENTRY_NOFAULT 0x0010
168 #define MAP_ENTRY_USER_WIRED 0x0020
170 #define MAP_ENTRY_BEHAV_NORMAL 0x0000 /* default behavior */
171 #define MAP_ENTRY_BEHAV_SEQUENTIAL 0x0040 /* expect sequential access */
172 #define MAP_ENTRY_BEHAV_RANDOM 0x0080 /* expect random access */
173 #define MAP_ENTRY_BEHAV_RESERVED 0x00C0 /* future use */
175 #define MAP_ENTRY_BEHAV_MASK 0x00C0
177 #define MAP_ENTRY_IN_TRANSITION 0x0100 /* entry being changed */
178 #define MAP_ENTRY_NEEDS_WAKEUP 0x0200 /* waiter's in transition */
179 #define MAP_ENTRY_NOCOREDUMP 0x0400 /* don't include in a core */
180 #define MAP_ENTRY_KSTACK 0x0800 /* guarded kernel stack */
183 * flags for vm_map_[un]clip_range()
185 #define MAP_CLIP_NO_HOLES 0x0001
188 * This reserve count for vm_map_entry_reserve() should cover all nominal
189 * single-insertion operations, including any necessary clipping.
191 #define MAP_RESERVE_COUNT 4
192 #define MAP_RESERVE_SLOP 32
194 static __inline u_char
195 vm_map_entry_behavior(struct vm_map_entry
*entry
)
197 return entry
->eflags
& MAP_ENTRY_BEHAV_MASK
;
201 vm_map_entry_set_behavior(struct vm_map_entry
*entry
, u_char behavior
)
203 entry
->eflags
= (entry
->eflags
& ~MAP_ENTRY_BEHAV_MASK
) |
204 (behavior
& MAP_ENTRY_BEHAV_MASK
);
208 * Maps are doubly-linked lists of map entries, kept sorted by address.
209 * A single hint is provided to start searches again from the last
210 * successful search, insertion, or removal.
212 * NOTE: The lock structure cannot be the first element of vm_map
213 * because this can result in a running lockup between two or more
214 * system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
215 * and free tsleep/waking up 'map' and the underlying lockmgr also
216 * sleeping and waking up on 'map'. The lockup occurs when the map fills
217 * up. The 'exec' map, for example.
219 * NOTE: The vm_map structure can be hard-locked with the lockmgr lock
220 * or soft-serialized with the token, or both.
223 struct vm_map_entry header
; /* List of entries */
224 RB_HEAD(vm_map_rb_tree
, vm_map_entry
) rb_root
;
225 struct lock lock
; /* Lock for map data */
226 int nentries
; /* Number of entries */
227 vm_size_t size
; /* virtual size */
228 u_char system_map
; /* Am I a system map? */
229 vm_map_entry_t hint
; /* hint for quick lookups */
230 unsigned int timestamp
; /* Version number */
231 vm_map_entry_t first_free
; /* First free space hint */
232 vm_flags_t flags
; /* flags for this vm_map */
233 struct pmap
*pmap
; /* Physical map */
234 u_int president_cache
; /* Remember president count */
235 u_int president_ticks
; /* Save ticks for cache */
236 struct lwkt_token token
; /* Soft serializer */
237 #define min_offset header.start
238 #define max_offset header.end
244 #define MAP_WIREFUTURE 0x0001 /* wire all future pages */
247 * Shareable process virtual address space.
249 * Refd pointers from vmresident, proc
252 struct vm_map vm_map
; /* VM address map */
253 struct pmap vm_pmap
; /* private physical map */
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_unused01
271 u_int vm_holdcnt
; /* temporary hold count and exit sequencing */
272 u_int vm_refcnt
; /* normal ref count */
275 #define VMSPACE_EXIT1 0x0001 /* partial exit */
276 #define VMSPACE_EXIT2 0x0002 /* full exit */
278 #define VMSPACE_HOLDEXIT 0x80000000
281 * Resident executable holding structure. A user program can take a snapshot
282 * of just its VM address space (typically done just after dynamic link
283 * libraries have completed loading) and register it as a resident
284 * executable associated with the program binary's vnode, which is also
285 * locked into memory. Future execs of the vnode will start with a copy
286 * of the resident vmspace instead of running the binary from scratch,
287 * avoiding both the kernel ELF loader *AND* all shared library mapping and
288 * relocation code, and will call a different entry point (the stack pointer
289 * is reset to the top of the stack) supplied when the vmspace was registered.
292 struct vnode
*vr_vnode
; /* associated vnode */
293 TAILQ_ENTRY(vmresident
) vr_link
; /* linked list of res sts */
294 struct vmspace
*vr_vmspace
; /* vmspace to fork */
295 intptr_t vr_entry_addr
; /* registered entry point */
296 struct sysentvec
*vr_sysent
; /* system call vects */
297 int vr_id
; /* registration id */
298 int vr_refs
; /* temporary refs */
303 * Macros: vm_map_lock, etc.
305 * Perform locking on the data portion of a map. Note that
306 * these macros mimic procedure calls returning void. The
307 * semicolon is supplied by the user of these macros, not
308 * by the macros themselves. The macros can safely be used
309 * as unbraced elements in a higher level statement.
312 #define ASSERT_VM_MAP_LOCKED(map) KKASSERT(lockowned(&(map)->lock))
315 /* #define MAP_LOCK_DIAGNOSTIC 1 */
316 #ifdef MAP_LOCK_DIAGNOSTIC
317 #define vm_map_lock(map) \
319 kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
320 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
321 panic("vm_map_lock: failed to get lock"); \
323 (map)->timestamp++; \
326 #define vm_map_lock(map) \
328 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
329 panic("vm_map_lock: failed to get lock"); \
331 (map)->timestamp++; \
335 #define vm_map_lock(map) \
337 lockmgr(&(map)->lock, LK_EXCLUSIVE); \
338 (map)->timestamp++; \
340 #endif /* DIAGNOSTIC */
342 #if defined(MAP_LOCK_DIAGNOSTIC)
343 #define vm_map_unlock(map) \
345 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
346 lockmgr(&(map)->lock, LK_RELEASE); \
348 #define vm_map_lock_read(map) \
350 kprintf ("locking map LK_SHARED: 0x%x\n", map); \
351 lockmgr(&(map)->lock, LK_SHARED); \
353 #define vm_map_unlock_read(map) \
355 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
356 lockmgr(&(map)->lock, LK_RELEASE); \
359 #define vm_map_unlock(map) \
360 lockmgr(&(map)->lock, LK_RELEASE)
361 #define vm_map_lock_read(map) \
362 lockmgr(&(map)->lock, LK_SHARED)
363 #define vm_map_unlock_read(map) \
364 lockmgr(&(map)->lock, LK_RELEASE)
367 #define vm_map_lock_read_try(map) \
368 lockmgr(&(map)->lock, LK_SHARED | LK_NOWAIT)
370 static __inline__
int
371 vm_map_lock_read_to(vm_map_t map
)
375 #if defined(MAP_LOCK_DIAGNOSTIC)
376 kprintf ("locking map LK_SHARED: 0x%x\n", map
);
378 error
= lockmgr(&(map
)->lock
, LK_SHARED
| LK_TIMELOCK
);
382 static __inline__
int
383 vm_map_lock_upgrade(vm_map_t map
) {
385 #if defined(MAP_LOCK_DIAGNOSTIC)
386 kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map
);
388 error
= lockmgr(&map
->lock
, LK_EXCLUPGRADE
);
394 #if defined(MAP_LOCK_DIAGNOSTIC)
395 #define vm_map_lock_downgrade(map) \
397 kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
398 lockmgr(&(map)->lock, LK_DOWNGRADE); \
401 #define vm_map_lock_downgrade(map) \
402 lockmgr(&(map)->lock, LK_DOWNGRADE)
408 * Functions implemented as macros
410 #define vm_map_min(map) ((map)->min_offset)
411 #define vm_map_max(map) ((map)->max_offset)
412 #define vm_map_pmap(map) ((map)->pmap)
417 static __inline
struct pmap
*
418 vmspace_pmap(struct vmspace
*vmspace
)
420 return &vmspace
->vm_pmap
;
424 * Caller must hold the vmspace->vm_map.token
427 vmspace_resident_count(struct vmspace
*vmspace
)
429 return pmap_resident_count(vmspace_pmap(vmspace
));
433 * Calculates the proportional RSS and returning the
434 * accrued result. This is a loose value for statistics/display
435 * purposes only and will only be updated if we can acquire
436 * a non-blocking map lock.
438 * (used by userland or the kernel)
440 static __inline u_int
441 vmspace_president_count(struct vmspace
*vmspace
)
443 vm_map_t map
= &vmspace
->vm_map
;
450 if (map
->president_ticks
== ticks
/ hz
|| vm_map_lock_read_try(map
))
451 return(map
->president_cache
);
454 for (cur
= map
->header
.next
; cur
!= &map
->header
; cur
= cur
->next
) {
455 switch(cur
->maptype
) {
456 case VM_MAPTYPE_NORMAL
:
457 case VM_MAPTYPE_VPAGETABLE
:
458 if ((object
= cur
->object
.vm_object
) == NULL
)
460 if (object
->type
!= OBJT_DEFAULT
&&
461 object
->type
!= OBJT_SWAP
) {
465 * synchronize non-zero case, contents of field
466 * can change at any time due to pmap ops.
468 if ((n
= object
->agg_pv_list_count
) != 0) {
472 count
+= object
->resident_page_count
/ n
;
480 map
->president_cache
= count
;
481 map
->president_ticks
= ticks
/ hz
;
482 vm_map_unlock_read(map
);
489 * Number of kernel maps and entries to statically allocate, required
490 * during boot to bootstrap the VM system.
493 #define MAX_MAPENT (SMP_MAXCPU * 32 + 1024)
496 * Copy-on-write flags for vm_map operations
498 #define MAP_UNUSED_01 0x0001
499 #define MAP_COPY_ON_WRITE 0x0002
500 #define MAP_NOFAULT 0x0004
501 #define MAP_PREFAULT 0x0008
502 #define MAP_PREFAULT_PARTIAL 0x0010
503 #define MAP_DISABLE_SYNCER 0x0020
504 #define MAP_IS_STACK 0x0040
505 #define MAP_IS_KSTACK 0x0080
506 #define MAP_DISABLE_COREDUMP 0x0100
507 #define MAP_PREFAULT_MADVISE 0x0200 /* from (user) madvise request */
508 #define MAP_PREFAULT_RELOCK 0x0200
511 * vm_fault option flags
513 #define VM_FAULT_NORMAL 0x00 /* Nothing special */
514 #define VM_FAULT_CHANGE_WIRING 0x01 /* Change the wiring as appropriate */
515 #define VM_FAULT_USER_WIRE 0x02 /* Likewise, but for user purposes */
516 #define VM_FAULT_BURST 0x04 /* Burst fault can be done */
517 #define VM_FAULT_DIRTY 0x08 /* Dirty the page */
518 #define VM_FAULT_UNSWAP 0x10 /* Remove backing store from the page */
519 #define VM_FAULT_BURST_QUICK 0x20 /* Special case shared vm_object */
520 #define VM_FAULT_WIRE_MASK (VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
524 boolean_t
vm_map_check_protection (vm_map_t
, vm_offset_t
, vm_offset_t
,
525 vm_prot_t
, boolean_t
);
528 void vm_map_entry_allocate_object(vm_map_entry_t
);
529 void vm_map_entry_reserve_cpu_init(struct globaldata
*gd
);
530 int vm_map_entry_reserve(int);
531 int vm_map_entry_kreserve(int);
532 void vm_map_entry_release(int);
533 void vm_map_entry_krelease(int);
534 vm_map_t
vm_map_create (vm_map_t
, struct pmap
*, vm_offset_t
, vm_offset_t
);
535 int vm_map_delete (vm_map_t
, vm_offset_t
, vm_offset_t
, int *);
536 int vm_map_find (vm_map_t
, void *, void *,
537 vm_ooffset_t
, vm_offset_t
*, vm_size_t
,
539 boolean_t
, vm_maptype_t
,
540 vm_prot_t
, vm_prot_t
, int);
541 int vm_map_findspace (vm_map_t
, vm_offset_t
, vm_size_t
, vm_size_t
,
543 vm_offset_t
vm_map_hint(struct proc
*, vm_offset_t
, vm_prot_t
);
544 int vm_map_inherit (vm_map_t
, vm_offset_t
, vm_offset_t
, vm_inherit_t
);
545 void vm_map_init (struct vm_map
*, vm_offset_t
, vm_offset_t
, pmap_t
);
546 int vm_map_insert (vm_map_t
, int *, void *, void *,
547 vm_ooffset_t
, vm_offset_t
, vm_offset_t
,
549 vm_prot_t
, vm_prot_t
, int);
550 int vm_map_lookup (vm_map_t
*, vm_offset_t
, vm_prot_t
, vm_map_entry_t
*, vm_object_t
*,
551 vm_pindex_t
*, vm_prot_t
*, boolean_t
*);
552 void vm_map_lookup_done (vm_map_t
, vm_map_entry_t
, int);
553 boolean_t
vm_map_lookup_entry (vm_map_t
, vm_offset_t
, vm_map_entry_t
*);
554 int vm_map_wire (vm_map_t
, vm_offset_t
, vm_offset_t
, int);
555 int vm_map_unwire (vm_map_t
, vm_offset_t
, vm_offset_t
, boolean_t
);
556 int vm_map_clean (vm_map_t
, vm_offset_t
, vm_offset_t
, boolean_t
, boolean_t
);
557 int vm_map_protect (vm_map_t
, vm_offset_t
, vm_offset_t
, vm_prot_t
, boolean_t
);
558 int vm_map_remove (vm_map_t
, vm_offset_t
, vm_offset_t
);
559 void vm_map_startup (void);
560 int vm_map_submap (vm_map_t
, vm_offset_t
, vm_offset_t
, vm_map_t
);
561 int vm_map_madvise (vm_map_t
, vm_offset_t
, vm_offset_t
, int, off_t
);
562 void vm_map_simplify_entry (vm_map_t
, vm_map_entry_t
, int *);
563 void vm_init2 (void);
564 int vm_uiomove (vm_map_t
, vm_object_t
, off_t
, int, vm_offset_t
, int *);
565 int vm_map_stack (vm_map_t
, vm_offset_t
, vm_size_t
, int,
566 vm_prot_t
, vm_prot_t
, int);
567 int vm_map_growstack (struct proc
*p
, vm_offset_t addr
);
568 int vmspace_swap_count (struct vmspace
*vmspace
);
569 int vmspace_anonymous_count (struct vmspace
*vmspace
);
570 void vm_map_set_wired_quick(vm_map_t map
, vm_offset_t addr
, vm_size_t size
, int *);
571 void vm_map_transition_wait(vm_map_t map
);
573 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
574 int vkernel_module_memory_alloc(vm_offset_t
*, size_t);
575 void vkernel_module_memory_free(vm_offset_t
, size_t);
579 #endif /* _VM_VM_MAP_H_ */