2 * Copyright (c) 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 * Copyright (c) 2003-2017 The DragonFly Project. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * The Mach Operating System project at Carnegie-Mellon University.
9 * This code is derived from software contributed to The DragonFly Project
10 * by Matthew Dillon <dillon@backplane.com>
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. 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.
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
36 * @(#)vm_map.h 8.9 (Berkeley) 5/17/95
39 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
40 * All rights reserved.
42 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
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.
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.
54 * Carnegie Mellon requests users of this software to return to
56 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
57 * School of Computer Science
58 * Carnegie Mellon University
59 * Pittsburgh PA 15213-3890
61 * any improvements or extensions that they make and grant Carnegie the
62 * rights to redistribute these changes.
64 * $FreeBSD: src/sys/vm/vm_map.h,v 1.54.2.5 2003/01/13 22:51:17 dillon Exp $
68 * Virtual memory map module definitions.
75 #include <sys/types.h>
78 #ifndef _SYS_KERNEL_H_
79 #include <sys/kernel.h> /* ticks */
88 #ifndef _SYS_VKERNEL_H_
89 #include <sys/vkernel.h>
94 #ifndef _MACHINE_PMAP_H_
95 #include <machine/pmap.h>
97 #ifndef _VM_VM_OBJECT_H_
98 #include <vm/vm_object.h>
101 #include <sys/_null.h>
104 struct vm_map_rb_tree
;
105 RB_PROTOTYPE(vm_map_rb_tree
, vm_map_entry
, rb_entry
, rb_vm_map_compare
);
110 * vm_map_t the high-level address map data structure.
111 * vm_map_entry_t an entry in an address map.
114 typedef u_int vm_flags_t
;
115 typedef u_int vm_eflags_t
;
118 * A vm_map_entry may reference an object, a submap, a uksmap, or a
119 * direct user-kernel shared map.
121 union vm_map_object
{
122 struct vm_object
*vm_object
; /* object object */
123 struct vm_map
*sub_map
; /* belongs to another map */
124 int (*uksmap
)(struct cdev
*dev
, vm_page_t fake
);
125 void *map_object
; /* generic */
129 vm_offset_t avail_ssize
; /* amt can grow if this is a stack */
130 vpte_t master_pde
; /* virtual page table root */
136 * vm_map_entry identifiers, used as a debugging aid
171 VM_SUBSYS_LIMIT
/* end of list */
175 * Address map entries consist of start and end addresses,
176 * a VM object (or sharing map) and offset into that object,
177 * and user-exported inheritance and protection information.
178 * Also included is control information for virtual copy operations.
180 * When used with MAP_STACK, avail_ssize is used to determine the
181 * limits of stack growth.
183 * When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the
184 * page directory index.
186 struct vm_map_entry
{
187 struct vm_map_entry
*prev
; /* previous entry */
188 struct vm_map_entry
*next
; /* next entry */
189 RB_ENTRY(vm_map_entry
) rb_entry
;
190 vm_offset_t start
; /* start address */
191 vm_offset_t end
; /* end address */
192 union vm_map_aux aux
; /* auxillary data */
193 union vm_map_object object
; /* object I point to */
194 vm_ooffset_t offset
; /* offset into object */
195 vm_eflags_t eflags
; /* map entry flags */
196 vm_maptype_t maptype
; /* type of VM mapping */
197 vm_prot_t protection
; /* protection code */
198 vm_prot_t max_protection
; /* maximum protection */
199 vm_inherit_t inheritance
; /* inheritance */
200 int wired_count
; /* can be paged if = 0 */
201 vm_subsys_t id
; /* subsystem id */
204 #define MAP_ENTRY_NOSYNC 0x0001
205 #define MAP_ENTRY_STACK 0x0002
206 #define MAP_ENTRY_COW 0x0004
207 #define MAP_ENTRY_NEEDS_COPY 0x0008
208 #define MAP_ENTRY_NOFAULT 0x0010
209 #define MAP_ENTRY_USER_WIRED 0x0020
211 #define MAP_ENTRY_BEHAV_NORMAL 0x0000 /* default behavior */
212 #define MAP_ENTRY_BEHAV_SEQUENTIAL 0x0040 /* expect sequential access */
213 #define MAP_ENTRY_BEHAV_RANDOM 0x0080 /* expect random access */
214 #define MAP_ENTRY_BEHAV_RESERVED 0x00C0 /* future use */
216 #define MAP_ENTRY_BEHAV_MASK 0x00C0
218 #define MAP_ENTRY_IN_TRANSITION 0x0100 /* entry being changed */
219 #define MAP_ENTRY_NEEDS_WAKEUP 0x0200 /* waiter's in transition */
220 #define MAP_ENTRY_NOCOREDUMP 0x0400 /* don't include in a core */
221 #define MAP_ENTRY_KSTACK 0x0800 /* guarded kernel stack */
224 * flags for vm_map_[un]clip_range()
226 #define MAP_CLIP_NO_HOLES 0x0001
229 * This reserve count for vm_map_entry_reserve() should cover all nominal
230 * single-insertion operations, including any necessary clipping.
232 #define MAP_RESERVE_COUNT 4
233 #define MAP_RESERVE_SLOP 64
234 #define MAP_RESERVE_HYST (MAP_RESERVE_SLOP - MAP_RESERVE_SLOP / 8)
237 * vm_map_lookup wflags
239 #define FW_WIRED 0x0001
240 #define FW_DIDCOW 0x0002
242 static __inline u_char
243 vm_map_entry_behavior(struct vm_map_entry
*entry
)
245 return entry
->eflags
& MAP_ENTRY_BEHAV_MASK
;
249 vm_map_entry_set_behavior(struct vm_map_entry
*entry
, u_char behavior
)
251 entry
->eflags
= (entry
->eflags
& ~MAP_ENTRY_BEHAV_MASK
) |
252 (behavior
& MAP_ENTRY_BEHAV_MASK
);
256 * VA interlock for map (VPAGETABLE / vkernel support)
258 struct vm_map_ilock
{
259 struct vm_map_ilock
*next
;
262 vm_offset_t ran_end
; /* non-inclusive */
265 #define ILOCK_WAITING 0x00000001
268 * Hinting mechanism used by vm_map_findspace() to figure out where to start
269 * an iteration looking for a hole big enough for the requested allocation.
270 * This can be important in situations where large amounts of kernel memory
271 * are being managed. For example, if the system is managing tens of
272 * thousands of processes or threads.
274 * If a hint is present it guarantees that no compatible hole exists prior
275 * to the (start) address. The (start) address itself is not necessarily
278 #define VM_MAP_FFCOUNT 4
279 #define VM_MAP_FFMASK (VM_MAP_FFCOUNT - 1)
281 struct vm_map_freehint
{
287 typedef struct vm_map_freehint vm_map_freehint_t
;
290 * Maps are doubly-linked lists of map entries, kept sorted by address.
291 * A single hint is provided to start searches again from the last
292 * successful search, insertion, or removal.
294 * NOTE: The lock structure cannot be the first element of vm_map
295 * because this can result in a running lockup between two or more
296 * system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
297 * and free tsleep/waking up 'map' and the underlying lockmgr also
298 * sleeping and waking up on 'map'. The lockup occurs when the map fills
299 * up. The 'exec' map, for example.
301 * NOTE: The vm_map structure can be hard-locked with the lockmgr lock
302 * or soft-serialized with the token, or both.
305 struct vm_map_entry header
; /* List of entries */
306 RB_HEAD(vm_map_rb_tree
, vm_map_entry
) rb_root
;
307 struct lock lock
; /* Lock for map data */
308 int nentries
; /* Number of entries */
309 unsigned int timestamp
; /* Version number */
310 vm_size_t size
; /* virtual size */
311 u_char system_map
; /* Am I a system map? */
312 u_char freehint_newindex
;
315 vm_flags_t flags
; /* flags for this vm_map */
316 vm_map_freehint_t freehint
[VM_MAP_FFCOUNT
];
317 struct pmap
*pmap
; /* Physical map */
318 u_int president_cache
; /* Remember president count */
319 u_int president_ticks
; /* Save ticks for cache */
320 struct vm_map_ilock
*ilock_base
;/* interlocks */
321 struct spinlock ilock_spin
; /* interlocks (spinlock for) */
322 struct lwkt_token token
; /* Soft serializer */
323 vm_offset_t pgout_offset
; /* for RLIMIT_RSS scans */
324 #define min_offset header.start
325 #define max_offset header.end
331 #define MAP_WIREFUTURE 0x0001 /* wire all future pages */
334 * Shareable process virtual address space.
336 * Refd pointers from vmresident, proc
339 struct vm_map vm_map
; /* VM address map */
340 struct pmap vm_pmap
; /* private physical map */
342 caddr_t vm_shm
; /* SYS5 shared memory private data XXX */
343 /* we copy from vm_startcopy to the end of the structure on fork */
344 #define vm_startcopy vm_rssize
345 segsz_t vm_rssize
; /* current resident set size in pages */
346 segsz_t vm_swrss
; /* resident set size before last swap */
347 segsz_t vm_tsize
; /* text size (pages) XXX */
348 segsz_t vm_dsize
; /* data size (pages) XXX */
349 segsz_t vm_ssize
; /* stack size (pages) */
350 caddr_t vm_taddr
; /* user virtual address of text XXX */
351 caddr_t vm_daddr
; /* user virtual address of data XXX */
352 caddr_t vm_maxsaddr
; /* user VA at max stack growth */
353 caddr_t vm_minsaddr
; /* user VA at max stack growth */
354 #define vm_endcopy vm_unused01
358 u_int vm_holdcnt
; /* temporary hold count and exit sequencing */
359 u_int vm_refcnt
; /* normal ref count */
362 #define VM_REF_DELETED 0x80000000U
364 #define VMSPACE_EXIT1 0x0001 /* partial exit */
365 #define VMSPACE_EXIT2 0x0002 /* full exit */
367 #define VMSPACE_HOLDEXIT 0x80000000
370 * Resident executable holding structure. A user program can take a snapshot
371 * of just its VM address space (typically done just after dynamic link
372 * libraries have completed loading) and register it as a resident
373 * executable associated with the program binary's vnode, which is also
374 * locked into memory. Future execs of the vnode will start with a copy
375 * of the resident vmspace instead of running the binary from scratch,
376 * avoiding both the kernel ELF loader *AND* all shared library mapping and
377 * relocation code, and will call a different entry point (the stack pointer
378 * is reset to the top of the stack) supplied when the vmspace was registered.
381 struct vnode
*vr_vnode
; /* associated vnode */
382 TAILQ_ENTRY(vmresident
) vr_link
; /* linked list of res sts */
383 struct vmspace
*vr_vmspace
; /* vmspace to fork */
384 intptr_t vr_entry_addr
; /* registered entry point */
385 struct sysentvec
*vr_sysent
; /* system call vects */
386 int vr_id
; /* registration id */
387 int vr_refs
; /* temporary refs */
392 * Macros: vm_map_lock, etc.
394 * Perform locking on the data portion of a map. Note that
395 * these macros mimic procedure calls returning void. The
396 * semicolon is supplied by the user of these macros, not
397 * by the macros themselves. The macros can safely be used
398 * as unbraced elements in a higher level statement.
401 #define ASSERT_VM_MAP_LOCKED(map) KKASSERT(lockowned(&(map)->lock))
404 /* #define MAP_LOCK_DIAGNOSTIC 1 */
405 #ifdef MAP_LOCK_DIAGNOSTIC
406 #define vm_map_lock(map) \
408 kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
409 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
410 panic("vm_map_lock: failed to get lock"); \
412 (map)->timestamp++; \
415 #define vm_map_lock(map) \
417 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
418 panic("vm_map_lock: failed to get lock"); \
420 (map)->timestamp++; \
424 #define vm_map_lock(map) \
426 lockmgr(&(map)->lock, LK_EXCLUSIVE); \
427 (map)->timestamp++; \
429 #endif /* DIAGNOSTIC */
431 #if defined(MAP_LOCK_DIAGNOSTIC)
432 #define vm_map_unlock(map) \
434 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
435 lockmgr(&(map)->lock, LK_RELEASE); \
437 #define vm_map_lock_read(map) \
439 kprintf ("locking map LK_SHARED: 0x%x\n", map); \
440 lockmgr(&(map)->lock, LK_SHARED); \
442 #define vm_map_unlock_read(map) \
444 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
445 lockmgr(&(map)->lock, LK_RELEASE); \
448 #define vm_map_unlock(map) \
449 lockmgr(&(map)->lock, LK_RELEASE)
450 #define vm_map_lock_read(map) \
451 lockmgr(&(map)->lock, LK_SHARED)
452 #define vm_map_unlock_read(map) \
453 lockmgr(&(map)->lock, LK_RELEASE)
456 #define vm_map_lock_read_try(map) \
457 lockmgr(&(map)->lock, LK_SHARED | LK_NOWAIT)
459 static __inline__
int
460 vm_map_lock_read_to(vm_map_t map
)
464 #if defined(MAP_LOCK_DIAGNOSTIC)
465 kprintf ("locking map LK_SHARED: 0x%x\n", map
);
467 error
= lockmgr(&(map
)->lock
, LK_SHARED
| LK_TIMELOCK
);
471 static __inline__
int
472 vm_map_lock_upgrade(vm_map_t map
) {
474 #if defined(MAP_LOCK_DIAGNOSTIC)
475 kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map
);
477 error
= lockmgr(&map
->lock
, LK_EXCLUPGRADE
);
483 #if defined(MAP_LOCK_DIAGNOSTIC)
484 #define vm_map_lock_downgrade(map) \
486 kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
487 lockmgr(&(map)->lock, LK_DOWNGRADE); \
490 #define vm_map_lock_downgrade(map) \
491 lockmgr(&(map)->lock, LK_DOWNGRADE)
497 * Functions implemented as macros
499 #define vm_map_min(map) ((map)->min_offset)
500 #define vm_map_max(map) ((map)->max_offset)
501 #define vm_map_pmap(map) ((map)->pmap)
506 static __inline
struct pmap
*
507 vmspace_pmap(struct vmspace
*vmspace
)
509 return &vmspace
->vm_pmap
;
513 * Caller must hold the vmspace->vm_map.token
516 vmspace_resident_count(struct vmspace
*vmspace
)
518 return pmap_resident_count(vmspace_pmap(vmspace
));
522 * Calculates the proportional RSS and returning the
523 * accrued result. This is a loose value for statistics/display
524 * purposes only and will only be updated if we can acquire
525 * a non-blocking map lock.
527 * (used by userland or the kernel)
529 static __inline u_int
530 vmspace_president_count(struct vmspace
*vmspace
)
532 vm_map_t map
= &vmspace
->vm_map
;
538 if (map
->president_ticks
== ticks
/ hz
|| vm_map_lock_read_try(map
))
539 return(map
->president_cache
);
542 for (cur
= map
->header
.next
; cur
!= &map
->header
; cur
= cur
->next
) {
543 switch(cur
->maptype
) {
544 case VM_MAPTYPE_NORMAL
:
545 case VM_MAPTYPE_VPAGETABLE
:
546 if ((object
= cur
->object
.vm_object
) == NULL
)
548 if (object
->type
!= OBJT_DEFAULT
&&
549 object
->type
!= OBJT_SWAP
) {
555 * synchronize non-zero case, contents of field
556 * can change at any time due to pmap ops.
558 if ((n
= object
->agg_pv_list_count
) != 0) {
562 count
+= object
->resident_page_count
/ n
;
571 map
->president_cache
= count
;
572 map
->president_ticks
= ticks
/ hz
;
573 vm_map_unlock_read(map
);
580 * Number of kernel maps and entries to statically allocate, required
581 * during boot to bootstrap the VM system.
584 #define MAX_MAPENT (SMP_MAXCPU * 32 + 1024)
587 * Copy-on-write flags for vm_map operations
589 #define MAP_UNUSED_01 0x0001
590 #define MAP_COPY_ON_WRITE 0x0002
591 #define MAP_NOFAULT 0x0004
592 #define MAP_PREFAULT 0x0008
593 #define MAP_PREFAULT_PARTIAL 0x0010
594 #define MAP_DISABLE_SYNCER 0x0020
595 #define MAP_IS_STACK 0x0040
596 #define MAP_IS_KSTACK 0x0080
597 #define MAP_DISABLE_COREDUMP 0x0100
598 #define MAP_PREFAULT_MADVISE 0x0200 /* from (user) madvise request */
599 #define MAP_PREFAULT_RELOCK 0x0200
602 * vm_fault option flags
604 #define VM_FAULT_NORMAL 0x00 /* Nothing special */
605 #define VM_FAULT_CHANGE_WIRING 0x01 /* Change the wiring as appropriate */
606 #define VM_FAULT_USER_WIRE 0x02 /* Likewise, but for user purposes */
607 #define VM_FAULT_BURST 0x04 /* Burst fault can be done */
608 #define VM_FAULT_DIRTY 0x08 /* Dirty the page */
609 #define VM_FAULT_UNSWAP 0x10 /* Remove backing store from the page */
610 #define VM_FAULT_BURST_QUICK 0x20 /* Special case shared vm_object */
611 #define VM_FAULT_WIRE_MASK (VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
612 #define VM_FAULT_USERMODE 0x40
616 boolean_t
vm_map_check_protection (vm_map_t
, vm_offset_t
, vm_offset_t
,
617 vm_prot_t
, boolean_t
);
620 void vm_map_entry_allocate_object(vm_map_entry_t
);
621 void vm_map_entry_reserve_cpu_init(struct globaldata
*gd
);
622 int vm_map_entry_reserve(int);
623 int vm_map_entry_kreserve(int);
624 void vm_map_entry_release(int);
625 void vm_map_entry_krelease(int);
626 int vm_map_delete (vm_map_t
, vm_offset_t
, vm_offset_t
, int *);
627 int vm_map_find (vm_map_t
, void *, void *,
628 vm_ooffset_t
, vm_offset_t
*, vm_size_t
,
629 vm_size_t
, boolean_t
,
630 vm_maptype_t
, vm_subsys_t id
,
631 vm_prot_t
, vm_prot_t
, int);
632 int vm_map_findspace (vm_map_t
, vm_offset_t
, vm_size_t
, vm_size_t
,
634 vm_offset_t
vm_map_hint(struct proc
*, vm_offset_t
, vm_prot_t
);
635 int vm_map_inherit (vm_map_t
, vm_offset_t
, vm_offset_t
, vm_inherit_t
);
636 void vm_map_init (struct vm_map
*, vm_offset_t
, vm_offset_t
, pmap_t
);
637 int vm_map_insert (vm_map_t
, int *, void *, void *,
638 vm_ooffset_t
, vm_offset_t
, vm_offset_t
,
639 vm_maptype_t
, vm_subsys_t id
,
640 vm_prot_t
, vm_prot_t
, int);
641 int vm_map_lookup (vm_map_t
*, vm_offset_t
, vm_prot_t
,
642 vm_map_entry_t
*, vm_object_t
*,
643 vm_pindex_t
*, vm_prot_t
*, int *);
644 void vm_map_lookup_done (vm_map_t
, vm_map_entry_t
, int);
645 boolean_t
vm_map_lookup_entry (vm_map_t
, vm_offset_t
, vm_map_entry_t
*);
646 int vm_map_wire (vm_map_t
, vm_offset_t
, vm_offset_t
, int);
647 int vm_map_unwire (vm_map_t
, vm_offset_t
, vm_offset_t
, boolean_t
);
648 int vm_map_clean (vm_map_t
, vm_offset_t
, vm_offset_t
, boolean_t
, boolean_t
);
649 int vm_map_protect (vm_map_t
, vm_offset_t
, vm_offset_t
, vm_prot_t
, boolean_t
);
650 int vm_map_remove (vm_map_t
, vm_offset_t
, vm_offset_t
);
651 void vm_map_startup (void);
652 int vm_map_submap (vm_map_t
, vm_offset_t
, vm_offset_t
, vm_map_t
);
653 int vm_map_madvise (vm_map_t
, vm_offset_t
, vm_offset_t
, int, off_t
);
654 void vm_map_simplify_entry (vm_map_t
, vm_map_entry_t
, int *);
655 void vm_init2 (void);
656 int vm_uiomove (vm_map_t
, vm_object_t
, off_t
, int, vm_offset_t
, int *);
657 int vm_map_stack (vm_map_t
, vm_offset_t
, vm_size_t
, int,
658 vm_prot_t
, vm_prot_t
, int);
659 int vm_map_growstack (vm_map_t map
, vm_offset_t addr
);
660 vm_offset_t
vmspace_swap_count (struct vmspace
*vmspace
);
661 vm_offset_t
vmspace_anonymous_count (struct vmspace
*vmspace
);
662 void vm_map_set_wired_quick(vm_map_t map
, vm_offset_t addr
, vm_size_t size
, int *);
663 void vm_map_transition_wait(vm_map_t map
, int relock
);
665 void vm_map_interlock(vm_map_t map
, struct vm_map_ilock
*ilock
,
666 vm_offset_t ran_beg
, vm_offset_t ran_end
);
667 void vm_map_deinterlock(vm_map_t map
, struct vm_map_ilock
*ilock
);
670 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
671 int vkernel_module_memory_alloc(vm_offset_t
*, size_t);
672 void vkernel_module_memory_free(vm_offset_t
, size_t);
676 #endif /* _VM_VM_MAP_H_ */