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
173 VM_SUBSYS_LIMIT /* end of list */
177 * Address map entries consist of start and end addresses,
178 * a VM object (or sharing map) and offset into that object,
179 * and user-exported inheritance and protection information.
180 * Also included is control information for virtual copy operations.
182 * When used with MAP_STACK, avail_ssize is used to determine the
183 * limits of stack growth.
185 * When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the
186 * page directory index.
188 struct vm_map_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 MAPENT_FREELIST(ent) (ent)->rb_entry.rbe_left
206 #define MAP_ENTRY_NOSYNC 0x0001
207 #define MAP_ENTRY_STACK 0x0002
208 #define MAP_ENTRY_COW 0x0004
209 #define MAP_ENTRY_NEEDS_COPY 0x0008
210 #define MAP_ENTRY_NOFAULT 0x0010
211 #define MAP_ENTRY_USER_WIRED 0x0020
213 #define MAP_ENTRY_BEHAV_NORMAL 0x0000 /* default behavior */
214 #define MAP_ENTRY_BEHAV_SEQUENTIAL 0x0040 /* expect sequential access */
215 #define MAP_ENTRY_BEHAV_RANDOM 0x0080 /* expect random access */
216 #define MAP_ENTRY_BEHAV_RESERVED 0x00C0 /* future use */
218 #define MAP_ENTRY_BEHAV_MASK 0x00C0
220 #define MAP_ENTRY_IN_TRANSITION 0x0100 /* entry being changed */
221 #define MAP_ENTRY_NEEDS_WAKEUP 0x0200 /* waiter's in transition */
222 #define MAP_ENTRY_NOCOREDUMP 0x0400 /* don't include in a core */
223 #define MAP_ENTRY_KSTACK 0x0800 /* guarded kernel stack */
226 * flags for vm_map_[un]clip_range()
228 #define MAP_CLIP_NO_HOLES 0x0001
231 * This reserve count for vm_map_entry_reserve() should cover all nominal
232 * single-insertion operations, including any necessary clipping.
234 #define MAP_RESERVE_COUNT 4
235 #define MAP_RESERVE_SLOP 64
236 #define MAP_RESERVE_HYST (MAP_RESERVE_SLOP - MAP_RESERVE_SLOP / 8)
239 * vm_map_lookup wflags
241 #define FW_WIRED 0x0001
242 #define FW_DIDCOW 0x0002
244 static __inline u_char
245 vm_map_entry_behavior(struct vm_map_entry *entry)
247 return entry->eflags & MAP_ENTRY_BEHAV_MASK;
251 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
253 entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
254 (behavior & MAP_ENTRY_BEHAV_MASK);
258 * VA interlock for map (VPAGETABLE / vkernel support)
260 struct vm_map_ilock {
261 struct vm_map_ilock *next;
264 vm_offset_t ran_end; /* non-inclusive */
267 #define ILOCK_WAITING 0x00000001
270 * Hinting mechanism used by vm_map_findspace() to figure out where to start
271 * an iteration looking for a hole big enough for the requested allocation.
272 * This can be important in situations where large amounts of kernel memory
273 * are being managed. For example, if the system is managing tens of
274 * thousands of processes or threads.
276 * If a hint is present it guarantees that no compatible hole exists prior
277 * to the (start) address. The (start) address itself is not necessarily
280 #define VM_MAP_FFCOUNT 4
281 #define VM_MAP_FFMASK (VM_MAP_FFCOUNT - 1)
283 struct vm_map_freehint {
289 typedef struct vm_map_freehint vm_map_freehint_t;
292 * Maps are doubly-linked lists of map entries, kept sorted by address.
293 * A single hint is provided to start searches again from the last
294 * successful search, insertion, or removal.
296 * NOTE: The lock structure cannot be the first element of vm_map
297 * because this can result in a running lockup between two or more
298 * system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
299 * and free tsleep/waking up 'map' and the underlying lockmgr also
300 * sleeping and waking up on 'map'. The lockup occurs when the map fills
301 * up. The 'exec' map, for example.
303 * NOTE: The vm_map structure can be hard-locked with the lockmgr lock
304 * or soft-serialized with the token, or both.
306 RB_HEAD(vm_map_rb_tree, vm_map_entry);
309 struct lock lock; /* Lock for map data */
310 struct vm_map_rb_tree rb_root; /* Organize map entries */
311 vm_offset_t min_addr; /* min address */
312 vm_offset_t max_addr; /* max address */
313 int nentries; /* Number of entries */
314 unsigned int timestamp; /* Version number */
315 vm_size_t size; /* virtual size */
316 u_char system_map; /* Am I a system map? */
317 u_char freehint_newindex;
320 vm_flags_t flags; /* flags for this vm_map */
321 vm_map_freehint_t freehint[VM_MAP_FFCOUNT];
322 struct pmap *pmap; /* Physical map */
323 u_int president_cache; /* Remember president count */
324 u_int president_ticks; /* Save ticks for cache */
325 struct vm_map_ilock *ilock_base;/* interlocks */
326 struct spinlock ilock_spin; /* interlocks (spinlock for) */
327 struct lwkt_token token; /* Soft serializer */
328 vm_offset_t pgout_offset; /* for RLIMIT_RSS scans */
334 #define MAP_WIREFUTURE 0x0001 /* wire all future pages */
337 * Shareable process virtual address space.
339 * Refd pointers from vmresident, proc
342 struct vm_map vm_map; /* VM address map */
343 struct pmap vm_pmap; /* private physical map */
345 caddr_t vm_shm; /* SYS5 shared memory private data XXX */
346 /* we copy from vm_startcopy to the end of the structure on fork */
347 #define vm_startcopy vm_rssize
348 segsz_t vm_rssize; /* current resident set size in pages */
349 segsz_t vm_swrss; /* resident set size before last swap */
350 segsz_t vm_tsize; /* text size (bytes) */
351 segsz_t vm_dsize; /* data size (bytes) */
352 segsz_t vm_ssize; /* stack size (bytes) */
353 caddr_t vm_taddr; /* user virtual address of text XXX */
354 caddr_t vm_daddr; /* user virtual address of data XXX */
355 caddr_t vm_maxsaddr; /* user VA at max stack growth */
356 caddr_t vm_minsaddr; /* user VA at max stack growth */
357 #define vm_endcopy vm_unused01
361 u_int vm_holdcnt; /* temporary hold count and exit sequencing */
362 u_int vm_refcnt; /* normal ref count */
365 #define VM_REF_DELETED 0x80000000U
367 #define VMSPACE_EXIT1 0x0001 /* partial exit */
368 #define VMSPACE_EXIT2 0x0002 /* full exit */
370 #define VMSPACE_HOLDEXIT 0x80000000
373 * Resident executable holding structure. A user program can take a snapshot
374 * of just its VM address space (typically done just after dynamic link
375 * libraries have completed loading) and register it as a resident
376 * executable associated with the program binary's vnode, which is also
377 * locked into memory. Future execs of the vnode will start with a copy
378 * of the resident vmspace instead of running the binary from scratch,
379 * avoiding both the kernel ELF loader *AND* all shared library mapping and
380 * relocation code, and will call a different entry point (the stack pointer
381 * is reset to the top of the stack) supplied when the vmspace was registered.
384 struct vnode *vr_vnode; /* associated vnode */
385 TAILQ_ENTRY(vmresident) vr_link; /* linked list of res sts */
386 struct vmspace *vr_vmspace; /* vmspace to fork */
387 intptr_t vr_entry_addr; /* registered entry point */
388 struct sysentvec *vr_sysent; /* system call vects */
389 int vr_id; /* registration id */
390 int vr_refs; /* temporary refs */
395 * Macros: vm_map_lock, etc.
397 * Perform locking on the data portion of a map. Note that
398 * these macros mimic procedure calls returning void. The
399 * semicolon is supplied by the user of these macros, not
400 * by the macros themselves. The macros can safely be used
401 * as unbraced elements in a higher level statement.
404 #define ASSERT_VM_MAP_LOCKED(map) KKASSERT(lockowned(&(map)->lock))
407 /* #define MAP_LOCK_DIAGNOSTIC 1 */
408 #ifdef MAP_LOCK_DIAGNOSTIC
409 #define vm_map_lock(map) \
411 kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
412 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
413 panic("vm_map_lock: failed to get lock"); \
415 (map)->timestamp++; \
418 #define vm_map_lock(map) \
420 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
421 panic("vm_map_lock: failed to get lock"); \
423 (map)->timestamp++; \
427 #define vm_map_lock(map) \
429 lockmgr(&(map)->lock, LK_EXCLUSIVE); \
430 (map)->timestamp++; \
432 #endif /* DIAGNOSTIC */
434 #if defined(MAP_LOCK_DIAGNOSTIC)
435 #define vm_map_unlock(map) \
437 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
438 lockmgr(&(map)->lock, LK_RELEASE); \
440 #define vm_map_lock_read(map) \
442 kprintf ("locking map LK_SHARED: 0x%x\n", map); \
443 lockmgr(&(map)->lock, LK_SHARED); \
445 #define vm_map_unlock_read(map) \
447 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
448 lockmgr(&(map)->lock, LK_RELEASE); \
451 #define vm_map_unlock(map) \
452 lockmgr(&(map)->lock, LK_RELEASE)
453 #define vm_map_lock_read(map) \
454 lockmgr(&(map)->lock, LK_SHARED)
455 #define vm_map_unlock_read(map) \
456 lockmgr(&(map)->lock, LK_RELEASE)
459 #define vm_map_lock_read_try(map) \
460 lockmgr(&(map)->lock, LK_SHARED | LK_NOWAIT)
462 static __inline__ int
463 vm_map_lock_read_to(vm_map_t map)
467 #if defined(MAP_LOCK_DIAGNOSTIC)
468 kprintf ("locking map LK_SHARED: 0x%x\n", map);
470 error = lockmgr(&(map)->lock, LK_SHARED | LK_TIMELOCK);
474 static __inline__ int
475 vm_map_lock_upgrade(vm_map_t map) {
477 #if defined(MAP_LOCK_DIAGNOSTIC)
478 kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
480 error = lockmgr(&map->lock, LK_EXCLUPGRADE);
486 #if defined(MAP_LOCK_DIAGNOSTIC)
487 #define vm_map_lock_downgrade(map) \
489 kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
490 lockmgr(&(map)->lock, LK_DOWNGRADE); \
493 #define vm_map_lock_downgrade(map) \
494 lockmgr(&(map)->lock, LK_DOWNGRADE)
500 * Functions implemented as macros
502 #define vm_map_min(map) ((map)->min_addr)
503 #define vm_map_max(map) ((map)->max_addr)
504 #define vm_map_pmap(map) ((map)->pmap)
509 static __inline struct pmap *
510 vmspace_pmap(struct vmspace *vmspace)
512 return &vmspace->vm_pmap;
516 * Caller must hold the vmspace->vm_map.token
519 vmspace_resident_count(struct vmspace *vmspace)
521 return pmap_resident_count(vmspace_pmap(vmspace));
525 * Calculates the proportional RSS and returning the
526 * accrued result. This is a loose value for statistics/display
527 * purposes only and will only be updated if we can acquire
528 * a non-blocking map lock.
530 * (used by userland or the kernel)
532 static __inline u_int
533 vmspace_president_count(struct vmspace *vmspace)
535 vm_map_t map = &vmspace->vm_map;
541 if (map->president_ticks == ticks / hz || vm_map_lock_read_try(map))
542 return(map->president_cache);
545 RB_FOREACH(cur, vm_map_rb_tree, &map->rb_root) {
546 switch(cur->maptype) {
547 case VM_MAPTYPE_NORMAL:
548 case VM_MAPTYPE_VPAGETABLE:
549 if ((object = cur->object.vm_object) == NULL)
551 if (object->type != OBJT_DEFAULT &&
552 object->type != OBJT_SWAP) {
558 * synchronize non-zero case, contents of field
559 * can change at any time due to pmap ops.
561 if ((n = object->agg_pv_list_count) != 0) {
565 count += object->resident_page_count / n;
574 map->president_cache = count;
575 map->president_ticks = ticks / hz;
576 vm_map_unlock_read(map);
583 * Number of kernel maps and entries to statically allocate, required
584 * during boot to bootstrap the VM system.
587 #define MAX_MAPENT (SMP_MAXCPU * 32 + 1024)
590 * Copy-on-write flags for vm_map operations
592 #define MAP_UNUSED_01 0x0001
593 #define MAP_COPY_ON_WRITE 0x0002
594 #define MAP_NOFAULT 0x0004
595 #define MAP_PREFAULT 0x0008
596 #define MAP_PREFAULT_PARTIAL 0x0010
597 #define MAP_DISABLE_SYNCER 0x0020
598 #define MAP_IS_STACK 0x0040
599 #define MAP_IS_KSTACK 0x0080
600 #define MAP_DISABLE_COREDUMP 0x0100
601 #define MAP_PREFAULT_MADVISE 0x0200 /* from (user) madvise request */
602 #define MAP_PREFAULT_RELOCK 0x0200
605 * vm_fault option flags
607 #define VM_FAULT_NORMAL 0x00 /* Nothing special */
608 #define VM_FAULT_CHANGE_WIRING 0x01 /* Change the wiring as appropriate */
609 #define VM_FAULT_USER_WIRE 0x02 /* Likewise, but for user purposes */
610 #define VM_FAULT_BURST 0x04 /* Burst fault can be done */
611 #define VM_FAULT_DIRTY 0x08 /* Dirty the page */
612 #define VM_FAULT_UNSWAP 0x10 /* Remove backing store from the page */
613 #define VM_FAULT_BURST_QUICK 0x20 /* Special case shared vm_object */
614 #define VM_FAULT_WIRE_MASK (VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
615 #define VM_FAULT_USERMODE 0x40
619 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t,
620 vm_prot_t, boolean_t);
623 void vm_map_entry_allocate_object(vm_map_entry_t);
624 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
625 int vm_map_entry_reserve(int);
626 int vm_map_entry_kreserve(int);
627 void vm_map_entry_release(int);
628 void vm_map_entry_krelease(int);
629 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
630 int vm_map_find (vm_map_t, void *, void *,
631 vm_ooffset_t, vm_offset_t *, vm_size_t,
632 vm_size_t, boolean_t,
633 vm_maptype_t, vm_subsys_t id,
634 vm_prot_t, vm_prot_t, int);
635 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_size_t,
637 vm_offset_t vm_map_hint(struct proc *, vm_offset_t, vm_prot_t);
638 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
639 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t, pmap_t);
640 int vm_map_insert (vm_map_t, int *, void *, void *,
641 vm_ooffset_t, vm_offset_t, vm_offset_t,
642 vm_maptype_t, vm_subsys_t id,
643 vm_prot_t, vm_prot_t, int);
644 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t,
645 vm_map_entry_t *, vm_object_t *,
646 vm_pindex_t *, vm_prot_t *, int *);
647 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
648 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
649 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
650 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
651 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
652 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
653 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
654 void vm_map_startup (void);
655 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
656 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int, off_t);
657 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
658 void vm_init2 (void);
659 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
660 int vm_map_stack (vm_map_t, vm_offset_t *, vm_size_t, int,
661 vm_prot_t, vm_prot_t, int);
662 int vm_map_growstack (vm_map_t map, vm_offset_t addr);
663 vm_offset_t vmspace_swap_count (struct vmspace *vmspace);
664 vm_offset_t vmspace_anonymous_count (struct vmspace *vmspace);
665 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
666 void vm_map_transition_wait(vm_map_t map, int relock);
668 void vm_map_interlock(vm_map_t map, struct vm_map_ilock *ilock,
669 vm_offset_t ran_beg, vm_offset_t ran_end);
670 void vm_map_deinterlock(vm_map_t map, struct vm_map_ilock *ilock);
673 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
674 int vkernel_module_memory_alloc(vm_offset_t *, size_t);
675 void vkernel_module_memory_free(vm_offset_t, size_t);
679 #endif /* _VM_VM_MAP_H_ */