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 * vm_map_entry identifiers, used as a debugging aid
170 VM_SUBSYS_LIMIT /* end of list */
174 * Address map entries consist of start and end addresses,
175 * a VM object (or sharing map) and offset into that object,
176 * and user-exported inheritance and protection information.
177 * Also included is control information for virtual copy operations.
179 * When used with MAP_STACK, avail_ssize is used to determine the
180 * limits of stack growth.
182 * When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the
183 * page directory index.
185 struct vm_map_entry {
186 struct vm_map_entry *prev; /* previous entry */
187 struct vm_map_entry *next; /* next entry */
188 RB_ENTRY(vm_map_entry) rb_entry;
189 vm_offset_t start; /* start address */
190 vm_offset_t end; /* end address */
191 union vm_map_aux aux; /* auxillary data */
192 union vm_map_object object; /* object I point to */
193 vm_ooffset_t offset; /* offset into object */
194 vm_eflags_t eflags; /* map entry flags */
195 vm_maptype_t maptype; /* type of VM mapping */
196 vm_prot_t protection; /* protection code */
197 vm_prot_t max_protection; /* maximum protection */
198 vm_inherit_t inheritance; /* inheritance */
199 int wired_count; /* can be paged if = 0 */
200 vm_subsys_t id; /* subsystem id */
203 #define MAP_ENTRY_NOSYNC 0x0001
204 #define MAP_ENTRY_STACK 0x0002
205 #define MAP_ENTRY_COW 0x0004
206 #define MAP_ENTRY_NEEDS_COPY 0x0008
207 #define MAP_ENTRY_NOFAULT 0x0010
208 #define MAP_ENTRY_USER_WIRED 0x0020
210 #define MAP_ENTRY_BEHAV_NORMAL 0x0000 /* default behavior */
211 #define MAP_ENTRY_BEHAV_SEQUENTIAL 0x0040 /* expect sequential access */
212 #define MAP_ENTRY_BEHAV_RANDOM 0x0080 /* expect random access */
213 #define MAP_ENTRY_BEHAV_RESERVED 0x00C0 /* future use */
215 #define MAP_ENTRY_BEHAV_MASK 0x00C0
217 #define MAP_ENTRY_IN_TRANSITION 0x0100 /* entry being changed */
218 #define MAP_ENTRY_NEEDS_WAKEUP 0x0200 /* waiter's in transition */
219 #define MAP_ENTRY_NOCOREDUMP 0x0400 /* don't include in a core */
220 #define MAP_ENTRY_KSTACK 0x0800 /* guarded kernel stack */
223 * flags for vm_map_[un]clip_range()
225 #define MAP_CLIP_NO_HOLES 0x0001
228 * This reserve count for vm_map_entry_reserve() should cover all nominal
229 * single-insertion operations, including any necessary clipping.
231 #define MAP_RESERVE_COUNT 4
232 #define MAP_RESERVE_SLOP 64
233 #define MAP_RESERVE_HYST (MAP_RESERVE_SLOP - MAP_RESERVE_SLOP / 8)
235 static __inline u_char
236 vm_map_entry_behavior(struct vm_map_entry *entry)
238 return entry->eflags & MAP_ENTRY_BEHAV_MASK;
242 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
244 entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
245 (behavior & MAP_ENTRY_BEHAV_MASK);
249 * VA interlock for map (VPAGETABLE / vkernel support)
251 struct vm_map_ilock {
252 struct vm_map_ilock *next;
255 vm_offset_t ran_end; /* non-inclusive */
258 #define ILOCK_WAITING 0x00000001
261 * Hinting mechanism used by vm_map_findspace() to figure out where to start
262 * an iteration looking for a hole big enough for the requested allocation.
263 * This can be important in situations where large amounts of kernel memory
264 * are being managed. For example, if the system is managing tens of
265 * thousands of processes or threads.
267 * If a hint is present it guarantees that no compatible hole exists prior
268 * to the (start) address. The (start) address itself is not necessarily
271 #define VM_MAP_FFCOUNT 4
272 #define VM_MAP_FFMASK (VM_MAP_FFCOUNT - 1)
274 struct vm_map_freehint {
280 typedef struct vm_map_freehint vm_map_freehint_t;
283 * Maps are doubly-linked lists of map entries, kept sorted by address.
284 * A single hint is provided to start searches again from the last
285 * successful search, insertion, or removal.
287 * NOTE: The lock structure cannot be the first element of vm_map
288 * because this can result in a running lockup between two or more
289 * system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
290 * and free tsleep/waking up 'map' and the underlying lockmgr also
291 * sleeping and waking up on 'map'. The lockup occurs when the map fills
292 * up. The 'exec' map, for example.
294 * NOTE: The vm_map structure can be hard-locked with the lockmgr lock
295 * or soft-serialized with the token, or both.
298 struct vm_map_entry header; /* List of entries */
299 RB_HEAD(vm_map_rb_tree, vm_map_entry) rb_root;
300 struct lock lock; /* Lock for map data */
301 int nentries; /* Number of entries */
302 unsigned int timestamp; /* Version number */
303 vm_size_t size; /* virtual size */
304 u_char system_map; /* Am I a system map? */
305 u_char freehint_newindex;
308 vm_flags_t flags; /* flags for this vm_map */
309 vm_map_freehint_t freehint[VM_MAP_FFCOUNT];
310 struct pmap *pmap; /* Physical map */
311 u_int president_cache; /* Remember president count */
312 u_int president_ticks; /* Save ticks for cache */
313 struct vm_map_ilock *ilock_base;/* interlocks */
314 struct spinlock ilock_spin; /* interlocks (spinlock for) */
315 struct lwkt_token token; /* Soft serializer */
316 vm_offset_t pgout_offset; /* for RLIMIT_RSS scans */
317 #define min_offset header.start
318 #define max_offset header.end
324 #define MAP_WIREFUTURE 0x0001 /* wire all future pages */
327 * Shareable process virtual address space.
329 * Refd pointers from vmresident, proc
332 struct vm_map vm_map; /* VM address map */
333 struct pmap vm_pmap; /* private physical map */
335 caddr_t vm_shm; /* SYS5 shared memory private data XXX */
336 /* we copy from vm_startcopy to the end of the structure on fork */
337 #define vm_startcopy vm_rssize
338 segsz_t vm_rssize; /* current resident set size in pages */
339 segsz_t vm_swrss; /* resident set size before last swap */
340 segsz_t vm_tsize; /* text size (pages) XXX */
341 segsz_t vm_dsize; /* data size (pages) XXX */
342 segsz_t vm_ssize; /* stack size (pages) */
343 caddr_t vm_taddr; /* user virtual address of text XXX */
344 caddr_t vm_daddr; /* user virtual address of data XXX */
345 caddr_t vm_maxsaddr; /* user VA at max stack growth */
346 caddr_t vm_minsaddr; /* user VA at max stack growth */
347 #define vm_endcopy vm_unused01
351 u_int vm_holdcnt; /* temporary hold count and exit sequencing */
352 u_int vm_refcnt; /* normal ref count */
355 #define VM_REF_DELETED 0x80000000U
357 #define VMSPACE_EXIT1 0x0001 /* partial exit */
358 #define VMSPACE_EXIT2 0x0002 /* full exit */
360 #define VMSPACE_HOLDEXIT 0x80000000
363 * Resident executable holding structure. A user program can take a snapshot
364 * of just its VM address space (typically done just after dynamic link
365 * libraries have completed loading) and register it as a resident
366 * executable associated with the program binary's vnode, which is also
367 * locked into memory. Future execs of the vnode will start with a copy
368 * of the resident vmspace instead of running the binary from scratch,
369 * avoiding both the kernel ELF loader *AND* all shared library mapping and
370 * relocation code, and will call a different entry point (the stack pointer
371 * is reset to the top of the stack) supplied when the vmspace was registered.
374 struct vnode *vr_vnode; /* associated vnode */
375 TAILQ_ENTRY(vmresident) vr_link; /* linked list of res sts */
376 struct vmspace *vr_vmspace; /* vmspace to fork */
377 intptr_t vr_entry_addr; /* registered entry point */
378 struct sysentvec *vr_sysent; /* system call vects */
379 int vr_id; /* registration id */
380 int vr_refs; /* temporary refs */
385 * Macros: vm_map_lock, etc.
387 * Perform locking on the data portion of a map. Note that
388 * these macros mimic procedure calls returning void. The
389 * semicolon is supplied by the user of these macros, not
390 * by the macros themselves. The macros can safely be used
391 * as unbraced elements in a higher level statement.
394 #define ASSERT_VM_MAP_LOCKED(map) KKASSERT(lockowned(&(map)->lock))
397 /* #define MAP_LOCK_DIAGNOSTIC 1 */
398 #ifdef MAP_LOCK_DIAGNOSTIC
399 #define vm_map_lock(map) \
401 kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
402 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
403 panic("vm_map_lock: failed to get lock"); \
405 (map)->timestamp++; \
408 #define vm_map_lock(map) \
410 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
411 panic("vm_map_lock: failed to get lock"); \
413 (map)->timestamp++; \
417 #define vm_map_lock(map) \
419 lockmgr(&(map)->lock, LK_EXCLUSIVE); \
420 (map)->timestamp++; \
422 #endif /* DIAGNOSTIC */
424 #if defined(MAP_LOCK_DIAGNOSTIC)
425 #define vm_map_unlock(map) \
427 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
428 lockmgr(&(map)->lock, LK_RELEASE); \
430 #define vm_map_lock_read(map) \
432 kprintf ("locking map LK_SHARED: 0x%x\n", map); \
433 lockmgr(&(map)->lock, LK_SHARED); \
435 #define vm_map_unlock_read(map) \
437 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
438 lockmgr(&(map)->lock, LK_RELEASE); \
441 #define vm_map_unlock(map) \
442 lockmgr(&(map)->lock, LK_RELEASE)
443 #define vm_map_lock_read(map) \
444 lockmgr(&(map)->lock, LK_SHARED)
445 #define vm_map_unlock_read(map) \
446 lockmgr(&(map)->lock, LK_RELEASE)
449 #define vm_map_lock_read_try(map) \
450 lockmgr(&(map)->lock, LK_SHARED | LK_NOWAIT)
452 static __inline__ int
453 vm_map_lock_read_to(vm_map_t map)
457 #if defined(MAP_LOCK_DIAGNOSTIC)
458 kprintf ("locking map LK_SHARED: 0x%x\n", map);
460 error = lockmgr(&(map)->lock, LK_SHARED | LK_TIMELOCK);
464 static __inline__ int
465 vm_map_lock_upgrade(vm_map_t map) {
467 #if defined(MAP_LOCK_DIAGNOSTIC)
468 kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
470 error = lockmgr(&map->lock, LK_EXCLUPGRADE);
476 #if defined(MAP_LOCK_DIAGNOSTIC)
477 #define vm_map_lock_downgrade(map) \
479 kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
480 lockmgr(&(map)->lock, LK_DOWNGRADE); \
483 #define vm_map_lock_downgrade(map) \
484 lockmgr(&(map)->lock, LK_DOWNGRADE)
490 * Functions implemented as macros
492 #define vm_map_min(map) ((map)->min_offset)
493 #define vm_map_max(map) ((map)->max_offset)
494 #define vm_map_pmap(map) ((map)->pmap)
499 static __inline struct pmap *
500 vmspace_pmap(struct vmspace *vmspace)
502 return &vmspace->vm_pmap;
506 * Caller must hold the vmspace->vm_map.token
509 vmspace_resident_count(struct vmspace *vmspace)
511 return pmap_resident_count(vmspace_pmap(vmspace));
515 * Calculates the proportional RSS and returning the
516 * accrued result. This is a loose value for statistics/display
517 * purposes only and will only be updated if we can acquire
518 * a non-blocking map lock.
520 * (used by userland or the kernel)
522 static __inline u_int
523 vmspace_president_count(struct vmspace *vmspace)
525 vm_map_t map = &vmspace->vm_map;
531 if (map->president_ticks == ticks / hz || vm_map_lock_read_try(map))
532 return(map->president_cache);
535 for (cur = map->header.next; cur != &map->header; cur = cur->next) {
536 switch(cur->maptype) {
537 case VM_MAPTYPE_NORMAL:
538 case VM_MAPTYPE_VPAGETABLE:
539 if ((object = cur->object.vm_object) == NULL)
541 if (object->type != OBJT_DEFAULT &&
542 object->type != OBJT_SWAP) {
548 * synchronize non-zero case, contents of field
549 * can change at any time due to pmap ops.
551 if ((n = object->agg_pv_list_count) != 0) {
555 count += object->resident_page_count / n;
564 map->president_cache = count;
565 map->president_ticks = ticks / hz;
566 vm_map_unlock_read(map);
573 * Number of kernel maps and entries to statically allocate, required
574 * during boot to bootstrap the VM system.
577 #define MAX_MAPENT (SMP_MAXCPU * 32 + 1024)
580 * Copy-on-write flags for vm_map operations
582 #define MAP_UNUSED_01 0x0001
583 #define MAP_COPY_ON_WRITE 0x0002
584 #define MAP_NOFAULT 0x0004
585 #define MAP_PREFAULT 0x0008
586 #define MAP_PREFAULT_PARTIAL 0x0010
587 #define MAP_DISABLE_SYNCER 0x0020
588 #define MAP_IS_STACK 0x0040
589 #define MAP_IS_KSTACK 0x0080
590 #define MAP_DISABLE_COREDUMP 0x0100
591 #define MAP_PREFAULT_MADVISE 0x0200 /* from (user) madvise request */
592 #define MAP_PREFAULT_RELOCK 0x0200
595 * vm_fault option flags
597 #define VM_FAULT_NORMAL 0x00 /* Nothing special */
598 #define VM_FAULT_CHANGE_WIRING 0x01 /* Change the wiring as appropriate */
599 #define VM_FAULT_USER_WIRE 0x02 /* Likewise, but for user purposes */
600 #define VM_FAULT_BURST 0x04 /* Burst fault can be done */
601 #define VM_FAULT_DIRTY 0x08 /* Dirty the page */
602 #define VM_FAULT_UNSWAP 0x10 /* Remove backing store from the page */
603 #define VM_FAULT_BURST_QUICK 0x20 /* Special case shared vm_object */
604 #define VM_FAULT_WIRE_MASK (VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
605 #define VM_FAULT_USERMODE 0x40
609 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t,
610 vm_prot_t, boolean_t);
613 void vm_map_entry_allocate_object(vm_map_entry_t);
614 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
615 int vm_map_entry_reserve(int);
616 int vm_map_entry_kreserve(int);
617 void vm_map_entry_release(int);
618 void vm_map_entry_krelease(int);
619 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
620 int vm_map_find (vm_map_t, void *, void *,
621 vm_ooffset_t, vm_offset_t *, vm_size_t,
622 vm_size_t, boolean_t,
623 vm_maptype_t, vm_subsys_t id,
624 vm_prot_t, vm_prot_t, int);
625 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_size_t,
627 vm_offset_t vm_map_hint(struct proc *, vm_offset_t, vm_prot_t);
628 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
629 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t, pmap_t);
630 int vm_map_insert (vm_map_t, int *, void *, void *,
631 vm_ooffset_t, vm_offset_t, vm_offset_t,
632 vm_maptype_t, vm_subsys_t id,
633 vm_prot_t, vm_prot_t, int);
634 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t, vm_map_entry_t *, vm_object_t *,
635 vm_pindex_t *, vm_prot_t *, boolean_t *);
636 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
637 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
638 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
639 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
640 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
641 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
642 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
643 void vm_map_startup (void);
644 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
645 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int, off_t);
646 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
647 void vm_init2 (void);
648 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
649 int vm_map_stack (vm_map_t, vm_offset_t, vm_size_t, int,
650 vm_prot_t, vm_prot_t, int);
651 int vm_map_growstack (vm_map_t map, vm_offset_t addr);
652 vm_offset_t vmspace_swap_count (struct vmspace *vmspace);
653 vm_offset_t vmspace_anonymous_count (struct vmspace *vmspace);
654 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
655 void vm_map_transition_wait(vm_map_t map);
657 void vm_map_interlock(vm_map_t map, struct vm_map_ilock *ilock,
658 vm_offset_t ran_beg, vm_offset_t ran_end);
659 void vm_map_deinterlock(vm_map_t map, struct vm_map_ilock *ilock);
662 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
663 int vkernel_module_memory_alloc(vm_offset_t *, size_t);
664 void vkernel_module_memory_free(vm_offset_t, size_t);
668 #endif /* _VM_VM_MAP_H_ */