2 * Copyright (c) 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 * Copyright (c) 2003-2019 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.
66 * Virtual memory map module definitions. The vm_map houses the pmap
67 * structure which controls the mmu context for a process.
73 #include <sys/types.h>
76 #ifndef _SYS_KERNEL_H_
77 #include <sys/kernel.h> /* ticks */
86 #ifndef _SYS_VKERNEL_H_
87 #include <sys/vkernel.h>
92 #ifndef _MACHINE_PMAP_H_
93 #include <machine/pmap.h>
95 #ifndef _VM_VM_OBJECT_H_
96 #include <vm/vm_object.h>
99 #include <sys/_null.h>
102 struct vm_map_rb_tree;
103 RB_PROTOTYPE(vm_map_rb_tree, vm_map_entry, rb_entry, rb_vm_map_compare);
105 typedef u_int vm_flags_t;
106 typedef u_int vm_eflags_t;
109 * Aux structure depends on map type and/or flags.
112 vm_offset_t avail_ssize; /* amt can grow if this is a stack */
113 vpte_t master_pde; /* virtual page table root */
119 * vm_map_entry identifiers, used as a debugging aid
156 VM_SUBSYS_LIMIT /* end of list */
160 * vm_map backing structure for specifying multiple backings. This
161 * structure is NOT shared across pmaps but may be shared within a pmap.
162 * The offset is cumulatively added from its parent, allowing easy splits
165 struct vm_map_backing {
166 struct vm_map_backing *backing_ba; /* backing store */
169 * A vm_map_entry may reference an object, a submap, a uksmap, or a
170 * direct user-kernel shared map.
173 struct vm_object *object; /* vm_object */
174 struct vm_map *sub_map; /* belongs to another map */
175 int (*uksmap)(struct cdev *dev, vm_page_t fake);
176 void *map_object; /* generic */
179 vm_ooffset_t offset; /* cumulative offset */
180 long refs; /* shared refs */
182 uint32_t backing_count; /* #entries backing us */
185 typedef struct vm_map_backing *vm_map_backing_t;
187 #define VM_MAP_BACK_EXCL_HEUR 0x00000001U
190 * Address map entries consist of start and end addresses, a VM object
191 * (or sharing map) and offset into that object, and user-exported
192 * inheritance and protection information. Also included is control
193 * information for virtual copy operations.
195 * The object information is now encapsulated in a vm_map_backing
196 * structure which contains the backing store chain, if any. This
197 * structure is NOT shared.
199 * When used with MAP_STACK, avail_ssize is used to determine the limits
202 * When used with VM_MAPTYPE_VPAGETABLE, avail_ssize stores the page
205 struct vm_map_entry {
206 RB_ENTRY(vm_map_entry) rb_entry;
207 vm_offset_t start; /* start address */
208 vm_offset_t end; /* end address */
209 union vm_map_aux aux; /* auxillary data */
210 struct vm_map_backing ba; /* backing object chain */
211 vm_eflags_t eflags; /* map entry flags */
212 vm_maptype_t maptype; /* type of VM mapping */
213 vm_prot_t protection; /* protection code */
214 vm_prot_t max_protection; /* maximum protection */
215 vm_inherit_t inheritance; /* inheritance */
216 int wired_count; /* can be paged if = 0 */
217 vm_subsys_t id; /* subsystem id */
220 typedef struct vm_map_entry *vm_map_entry_t;
222 #define MAPENT_FREELIST(ent) (ent)->rb_entry.rbe_left
224 #define MAP_ENTRY_NOSYNC 0x0001
225 #define MAP_ENTRY_STACK 0x0002
226 #define MAP_ENTRY_COW 0x0004
227 #define MAP_ENTRY_NEEDS_COPY 0x0008
228 #define MAP_ENTRY_NOFAULT 0x0010
229 #define MAP_ENTRY_USER_WIRED 0x0020
231 #define MAP_ENTRY_BEHAV_NORMAL 0x0000 /* default behavior */
232 #define MAP_ENTRY_BEHAV_SEQUENTIAL 0x0040 /* expect sequential access */
233 #define MAP_ENTRY_BEHAV_RANDOM 0x0080 /* expect random access */
234 #define MAP_ENTRY_BEHAV_RESERVED 0x00C0 /* future use */
236 #define MAP_ENTRY_BEHAV_MASK 0x00C0
238 #define MAP_ENTRY_IN_TRANSITION 0x0100 /* entry being changed */
239 #define MAP_ENTRY_NEEDS_WAKEUP 0x0200 /* waiter's in transition */
240 #define MAP_ENTRY_NOCOREDUMP 0x0400 /* don't include in a core */
241 #define MAP_ENTRY_KSTACK 0x0800 /* guarded kernel stack */
244 * flags for vm_map_[un]clip_range()
246 #define MAP_CLIP_NO_HOLES 0x0001
249 * This reserve count for vm_map_entry_reserve() should cover all nominal
250 * single-insertion operations, including any necessary clipping.
252 #define MAP_RESERVE_COUNT 4
253 #define MAP_RESERVE_SLOP 64
254 #define MAP_RESERVE_HYST (MAP_RESERVE_SLOP - MAP_RESERVE_SLOP / 8)
257 * vm_map_lookup wflags
259 #define FW_WIRED 0x0001
260 #define FW_DIDCOW 0x0002
262 static __inline u_char
263 vm_map_entry_behavior(struct vm_map_entry *entry)
265 return entry->eflags & MAP_ENTRY_BEHAV_MASK;
269 vm_map_entry_set_behavior(struct vm_map_entry *entry, u_char behavior)
271 entry->eflags = (entry->eflags & ~MAP_ENTRY_BEHAV_MASK) |
272 (behavior & MAP_ENTRY_BEHAV_MASK);
276 * VA interlock for map (VPAGETABLE / vkernel support)
278 struct vm_map_ilock {
279 struct vm_map_ilock *next;
282 vm_offset_t ran_end; /* non-inclusive */
285 #define ILOCK_WAITING 0x00000001
288 * Hinting mechanism used by vm_map_findspace() to figure out where to start
289 * an iteration looking for a hole big enough for the requested allocation.
290 * This can be important in situations where large amounts of kernel memory
291 * are being managed. For example, if the system is managing tens of
292 * thousands of processes or threads.
294 * If a hint is present it guarantees that no compatible hole exists prior
295 * to the (start) address. The (start) address itself is not necessarily
298 #define VM_MAP_FFCOUNT 4
299 #define VM_MAP_FFMASK (VM_MAP_FFCOUNT - 1)
301 struct vm_map_freehint {
307 typedef struct vm_map_freehint vm_map_freehint_t;
310 * Maps are doubly-linked lists of map entries, kept sorted by address.
311 * A single hint is provided to start searches again from the last
312 * successful search, insertion, or removal.
314 * NOTE: The lock structure cannot be the first element of vm_map
315 * because this can result in a running lockup between two or more
316 * system processes trying to kmem_alloc_wait() due to kmem_alloc_wait()
317 * and free tsleep/waking up 'map' and the underlying lockmgr also
318 * sleeping and waking up on 'map'. The lockup occurs when the map fills
319 * up. The 'exec' map, for example.
321 * NOTE: The vm_map structure can be hard-locked with the lockmgr lock
322 * or soft-serialized with the token, or both.
324 RB_HEAD(vm_map_rb_tree, vm_map_entry);
327 struct lock lock; /* Lock for map data */
328 struct vm_map_rb_tree rb_root; /* Organize map entries */
329 vm_offset_t min_addr; /* min address */
330 vm_offset_t max_addr; /* max address */
331 int nentries; /* Number of entries */
332 unsigned int timestamp; /* Version number */
333 vm_size_t size; /* virtual size */
334 u_char system_map; /* Am I a system map? */
335 u_char freehint_newindex;
338 vm_flags_t flags; /* flags for this vm_map */
339 vm_map_freehint_t freehint[VM_MAP_FFCOUNT];
340 struct pmap *pmap; /* Physical map */
341 struct vm_map_ilock *ilock_base;/* interlocks */
342 struct spinlock ilock_spin; /* interlocks (spinlock for) */
343 struct lwkt_token token; /* Soft serializer */
344 vm_offset_t pgout_offset; /* for RLIMIT_RSS scans */
347 typedef struct vm_map *vm_map_t;
352 #define MAP_WIREFUTURE 0x0001 /* wire all future pages */
355 * Shareable process virtual address space.
357 * Refd pointers from vmresident, proc
360 struct vm_map vm_map; /* VM address map */
361 struct pmap vm_pmap; /* private physical map */
363 caddr_t vm_shm; /* SYS5 shared memory private data XXX */
364 /* we copy from vm_startcopy to the end of the structure on fork */
365 #define vm_startcopy vm_rssize
366 segsz_t vm_rssize; /* current resident set size in pages */
367 segsz_t vm_swrss; /* resident set size before last swap */
368 segsz_t vm_tsize; /* text size (bytes) */
369 segsz_t vm_dsize; /* data size (bytes) */
370 segsz_t vm_ssize; /* stack size (bytes) */
371 caddr_t vm_taddr; /* user virtual address of text XXX */
372 caddr_t vm_daddr; /* user virtual address of data XXX */
373 caddr_t vm_maxsaddr; /* user VA at max stack growth */
374 caddr_t vm_minsaddr; /* user VA at max stack growth */
375 #define vm_endcopy vm_unused01
379 u_int vm_holdcnt; /* temporary hold count and exit sequencing */
380 u_int vm_refcnt; /* normal ref count */
383 #define VM_REF_DELETED 0x80000000U
385 #define VMSPACE_EXIT1 0x0001 /* partial exit */
386 #define VMSPACE_EXIT2 0x0002 /* full exit */
388 #define VMSPACE_HOLDEXIT 0x80000000
391 * Resident executable holding structure. A user program can take a snapshot
392 * of just its VM address space (typically done just after dynamic link
393 * libraries have completed loading) and register it as a resident
394 * executable associated with the program binary's vnode, which is also
395 * locked into memory. Future execs of the vnode will start with a copy
396 * of the resident vmspace instead of running the binary from scratch,
397 * avoiding both the kernel ELF loader *AND* all shared library mapping and
398 * relocation code, and will call a different entry point (the stack pointer
399 * is reset to the top of the stack) supplied when the vmspace was registered.
402 struct vnode *vr_vnode; /* associated vnode */
403 TAILQ_ENTRY(vmresident) vr_link; /* linked list of res sts */
404 struct vmspace *vr_vmspace; /* vmspace to fork */
405 intptr_t vr_entry_addr; /* registered entry point */
406 struct sysentvec *vr_sysent; /* system call vects */
407 int vr_id; /* registration id */
408 int vr_refs; /* temporary refs */
413 * Macros: vm_map_lock, etc.
415 * Perform locking on the data portion of a map. Note that
416 * these macros mimic procedure calls returning void. The
417 * semicolon is supplied by the user of these macros, not
418 * by the macros themselves. The macros can safely be used
419 * as unbraced elements in a higher level statement.
422 #define ASSERT_VM_MAP_LOCKED(map) KKASSERT(lockowned(&(map)->lock))
425 /* #define MAP_LOCK_DIAGNOSTIC 1 */
426 #ifdef MAP_LOCK_DIAGNOSTIC
427 #define vm_map_lock(map) \
429 kprintf ("locking map LK_EXCLUSIVE: 0x%x\n", map); \
430 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
431 panic("vm_map_lock: failed to get lock"); \
433 (map)->timestamp++; \
436 #define vm_map_lock(map) \
438 if (lockmgr(&(map)->lock, LK_EXCLUSIVE) != 0) { \
439 panic("vm_map_lock: failed to get lock"); \
441 (map)->timestamp++; \
445 #define vm_map_lock(map) \
447 lockmgr(&(map)->lock, LK_EXCLUSIVE); \
448 (map)->timestamp++; \
450 #endif /* DIAGNOSTIC */
452 #if defined(MAP_LOCK_DIAGNOSTIC)
453 #define vm_map_unlock(map) \
455 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
456 lockmgr(&(map)->lock, LK_RELEASE); \
458 #define vm_map_lock_read(map) \
460 kprintf ("locking map LK_SHARED: 0x%x\n", map); \
461 lockmgr(&(map)->lock, LK_SHARED); \
463 #define vm_map_unlock_read(map) \
465 kprintf ("locking map LK_RELEASE: 0x%x\n", map); \
466 lockmgr(&(map)->lock, LK_RELEASE); \
469 #define vm_map_unlock(map) \
470 lockmgr(&(map)->lock, LK_RELEASE)
471 #define vm_map_lock_read(map) \
472 lockmgr(&(map)->lock, LK_SHARED)
473 #define vm_map_unlock_read(map) \
474 lockmgr(&(map)->lock, LK_RELEASE)
477 #define vm_map_lock_read_try(map) \
478 lockmgr(&(map)->lock, LK_SHARED | LK_NOWAIT)
480 static __inline__ int
481 vm_map_lock_read_to(vm_map_t map)
485 #if defined(MAP_LOCK_DIAGNOSTIC)
486 kprintf ("locking map LK_SHARED: 0x%x\n", map);
488 error = lockmgr(&(map)->lock, LK_SHARED | LK_TIMELOCK);
492 static __inline__ int
493 vm_map_lock_upgrade(vm_map_t map) {
495 #if defined(MAP_LOCK_DIAGNOSTIC)
496 kprintf("locking map LK_EXCLUPGRADE: 0x%x\n", map);
498 error = lockmgr(&map->lock, LK_EXCLUPGRADE);
504 #if defined(MAP_LOCK_DIAGNOSTIC)
505 #define vm_map_lock_downgrade(map) \
507 kprintf ("locking map LK_DOWNGRADE: 0x%x\n", map); \
508 lockmgr(&(map)->lock, LK_DOWNGRADE); \
511 #define vm_map_lock_downgrade(map) \
512 lockmgr(&(map)->lock, LK_DOWNGRADE)
518 * Functions implemented as macros
520 #define vm_map_min(map) ((map)->min_addr)
521 #define vm_map_max(map) ((map)->max_addr)
522 #define vm_map_pmap(map) ((map)->pmap)
527 static __inline struct pmap *
528 vmspace_pmap(struct vmspace *vmspace)
530 return &vmspace->vm_pmap;
534 * Caller must hold the vmspace->vm_map.token
537 vmspace_resident_count(struct vmspace *vmspace)
539 return pmap_resident_count(vmspace_pmap(vmspace));
543 * Number of kernel maps and entries to statically allocate, required
544 * during boot to bootstrap the VM system.
547 #define MAX_MAPENT (SMP_MAXCPU * 32 + 1024)
550 * Copy-on-write flags for vm_map operations
552 #define MAP_UNUSED_01 0x0001
553 #define MAP_COPY_ON_WRITE 0x0002
554 #define MAP_NOFAULT 0x0004
555 #define MAP_PREFAULT 0x0008
556 #define MAP_PREFAULT_PARTIAL 0x0010
557 #define MAP_DISABLE_SYNCER 0x0020
558 #define MAP_IS_STACK 0x0040
559 #define MAP_IS_KSTACK 0x0080
560 #define MAP_DISABLE_COREDUMP 0x0100
561 #define MAP_PREFAULT_MADVISE 0x0200 /* from (user) madvise request */
562 #define MAP_PREFAULT_RELOCK 0x0200
565 * vm_fault option flags
567 #define VM_FAULT_NORMAL 0x00 /* Nothing special */
568 #define VM_FAULT_CHANGE_WIRING 0x01 /* Change the wiring as appropriate */
569 #define VM_FAULT_USER_WIRE 0x02 /* Likewise, but for user purposes */
570 #define VM_FAULT_BURST 0x04 /* Burst fault can be done */
571 #define VM_FAULT_DIRTY 0x08 /* Dirty the page */
572 #define VM_FAULT_UNSWAP 0x10 /* Remove backing store from the page */
573 #define VM_FAULT_BURST_QUICK 0x20 /* Special case shared vm_object */
574 #define VM_FAULT_WIRE_MASK (VM_FAULT_CHANGE_WIRING|VM_FAULT_USER_WIRE)
575 #define VM_FAULT_USERMODE 0x40
579 boolean_t vm_map_check_protection (vm_map_t, vm_offset_t, vm_offset_t,
580 vm_prot_t, boolean_t);
583 void vm_map_entry_allocate_object(vm_map_entry_t);
584 void vm_map_entry_reserve_cpu_init(struct globaldata *gd);
585 int vm_map_entry_reserve(int);
586 int vm_map_entry_kreserve(int);
587 void vm_map_entry_release(int);
588 void vm_map_entry_krelease(int);
589 int vm_map_delete (vm_map_t, vm_offset_t, vm_offset_t, int *);
590 int vm_map_find (vm_map_t, void *, void *,
591 vm_ooffset_t, vm_offset_t *, vm_size_t,
592 vm_size_t, boolean_t,
593 vm_maptype_t, vm_subsys_t id,
594 vm_prot_t, vm_prot_t, int);
595 int vm_map_findspace (vm_map_t, vm_offset_t, vm_size_t, vm_size_t,
597 vm_offset_t vm_map_hint(struct proc *, vm_offset_t, vm_prot_t);
598 int vm_map_inherit (vm_map_t, vm_offset_t, vm_offset_t, vm_inherit_t);
599 void vm_map_init (struct vm_map *, vm_offset_t, vm_offset_t, pmap_t);
600 int vm_map_insert (vm_map_t, int *, void *, void *,
601 vm_ooffset_t, vm_offset_t, vm_offset_t,
602 vm_maptype_t, vm_subsys_t id,
603 vm_prot_t, vm_prot_t, int);
604 int vm_map_lookup (vm_map_t *, vm_offset_t, vm_prot_t,
605 vm_map_entry_t *, struct vm_map_backing **,
606 vm_pindex_t *, vm_prot_t *, int *);
607 void vm_map_lookup_done (vm_map_t, vm_map_entry_t, int);
608 boolean_t vm_map_lookup_entry (vm_map_t, vm_offset_t, vm_map_entry_t *);
609 int vm_map_wire (vm_map_t, vm_offset_t, vm_offset_t, int);
610 int vm_map_unwire (vm_map_t, vm_offset_t, vm_offset_t, boolean_t);
611 int vm_map_clean (vm_map_t, vm_offset_t, vm_offset_t, boolean_t, boolean_t);
612 int vm_map_protect (vm_map_t, vm_offset_t, vm_offset_t, vm_prot_t, boolean_t);
613 int vm_map_remove (vm_map_t, vm_offset_t, vm_offset_t);
614 void vm_map_startup (void);
615 int vm_map_submap (vm_map_t, vm_offset_t, vm_offset_t, vm_map_t);
616 int vm_map_madvise (vm_map_t, vm_offset_t, vm_offset_t, int, off_t);
617 void vm_map_simplify_entry (vm_map_t, vm_map_entry_t, int *);
618 void vm_init2 (void);
619 int vm_uiomove (vm_map_t, vm_object_t, off_t, int, vm_offset_t, int *);
620 int vm_map_stack (vm_map_t, vm_offset_t *, vm_size_t, int,
621 vm_prot_t, vm_prot_t, int);
622 int vm_map_growstack (vm_map_t map, vm_offset_t addr);
623 vm_offset_t vmspace_swap_count (struct vmspace *vmspace);
624 vm_offset_t vmspace_anonymous_count (struct vmspace *vmspace);
625 void vm_map_set_wired_quick(vm_map_t map, vm_offset_t addr, vm_size_t size, int *);
626 void vm_map_transition_wait(vm_map_t map, int relock);
628 void vm_map_interlock(vm_map_t map, struct vm_map_ilock *ilock,
629 vm_offset_t ran_beg, vm_offset_t ran_end);
630 void vm_map_deinterlock(vm_map_t map, struct vm_map_ilock *ilock);
633 #if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
634 int vkernel_module_memory_alloc(vm_offset_t *, size_t);
635 void vkernel_module_memory_free(vm_offset_t, size_t);
639 #endif /* _VM_VM_MAP_H_ */