2 * Copyright (c) 1988 University of Utah.
3 * Copyright (c) 1991, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * the Systems Programming Group of the University of Utah Computer
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * from: Utah $Hdr: vm_mmap.c 1.6 91/10/21$
40 * @(#)vm_mmap.c 8.4 (Berkeley) 1/12/94
41 * $FreeBSD: src/sys/vm/vm_mmap.c,v 1.108.2.6 2002/07/02 20:06:19 dillon Exp $
42 * $DragonFly: src/sys/vm/vm_mmap.c,v 1.4 2003/06/25 03:56:13 dillon Exp $
46 * Mapped file (mmap) interface to VM
49 #include "opt_compat.h"
51 #include <sys/param.h>
52 #include <sys/kernel.h>
53 #include <sys/systm.h>
54 #include <sys/sysproto.h>
55 #include <sys/filedesc.h>
57 #include <sys/resource.h>
58 #include <sys/resourcevar.h>
59 #include <sys/vnode.h>
60 #include <sys/fcntl.h>
65 #include <sys/vmmeter.h>
66 #include <sys/sysctl.h>
69 #include <vm/vm_param.h>
72 #include <vm/vm_map.h>
73 #include <vm/vm_object.h>
74 #include <vm/vm_page.h>
75 #include <vm/vm_pager.h>
76 #include <vm/vm_pageout.h>
77 #include <vm/vm_extern.h>
78 #include <vm/vm_page.h>
79 #include <vm/vm_kern.h>
81 #include <sys/file2.h>
83 #ifndef _SYS_SYSPROTO_H_
89 static int max_proc_mmap;
90 SYSCTL_INT(_vm, OID_AUTO, max_proc_mmap, CTLFLAG_RW, &max_proc_mmap, 0, "");
93 * Set the maximum number of vm_map_entry structures per process. Roughly
94 * speaking vm_map_entry structures are tiny, so allowing them to eat 1/100
95 * of our KVM malloc space still results in generous limits. We want a
96 * default that is good enough to prevent the kernel running out of resources
97 * if attacked from compromised user account but generous enough such that
98 * multi-threaded processes are not unduly inconvenienced.
101 static void vmmapentry_rsrc_init __P((void *));
102 SYSINIT(vmmersrc, SI_SUB_KVM_RSRC, SI_ORDER_FIRST, vmmapentry_rsrc_init, NULL)
105 vmmapentry_rsrc_init(dummy)
108 max_proc_mmap = vm_kmem_size / sizeof(struct vm_map_entry);
109 max_proc_mmap /= 100;
114 sbrk(struct sbrk_args *uap)
116 /* Not yet implemented */
121 * sstk_args(int incr)
125 sstk(struct sstk_args *uap)
127 /* Not yet implemented */
131 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
134 * getpagesize_args(int dummy)
138 ogetpagesize(struct getpagesize_args *uap)
140 struct proc *p = curproc;
141 p->p_retval[0] = PAGE_SIZE;
144 #endif /* COMPAT_43 || COMPAT_SUNOS */
148 * mmap_args(void *addr, size_t len, int prot, int flags, int fd,
149 * long pad, off_t pos)
151 * Memory Map (mmap) system call. Note that the file offset
152 * and address are allowed to be NOT page aligned, though if
153 * the MAP_FIXED flag it set, both must have the same remainder
154 * modulo the PAGE_SIZE (POSIX 1003.1b). If the address is not
155 * page-aligned, the actual mapping starts at trunc_page(addr)
156 * and the return value is adjusted up by the page offset.
158 * Generally speaking, only character devices which are themselves
159 * memory-based, such as a video framebuffer, can be mmap'd. Otherwise
160 * there would be no cache coherency between a descriptor and a VM mapping
161 * both to the same character device.
163 * Block devices can be mmap'd no matter what they represent. Cache coherency
164 * is maintained as long as you do not write directly to the underlying
169 mmap(struct mmap_args *uap)
171 struct thread *td = curthread;
172 struct proc *p = td->td_proc;
173 struct filedesc *fdp = p->p_fd;
174 struct file *fp = NULL;
177 vm_size_t size, pageoff;
178 vm_prot_t prot, maxprot;
181 int disablexworkaround;
183 struct vmspace *vms = p->p_vmspace;
188 addr = (vm_offset_t) uap->addr;
190 prot = uap->prot & VM_PROT_ALL;
194 /* make sure mapping fits into numeric range etc */
195 if ((ssize_t) uap->len < 0 ||
196 ((flags & MAP_ANON) && uap->fd != -1))
199 if (flags & MAP_STACK) {
200 if ((uap->fd != -1) ||
201 ((prot & (PROT_READ | PROT_WRITE)) != (PROT_READ | PROT_WRITE)))
208 * Align the file position to a page boundary,
209 * and save its page offset component.
211 pageoff = (pos & PAGE_MASK);
214 /* Adjust size for rounding (on both ends). */
215 size += pageoff; /* low end... */
216 size = (vm_size_t) round_page(size); /* hi end */
219 * Check for illegal addresses. Watch out for address wrap... Note
220 * that VM_*_ADDRESS are not constants due to casts (argh).
222 if (flags & MAP_FIXED) {
224 * The specified address must have the same remainder
225 * as the file offset taken modulo PAGE_SIZE, so it
226 * should be aligned after adjustment by pageoff.
229 if (addr & PAGE_MASK)
231 /* Address range must be all in user VM space. */
232 if (VM_MAXUSER_ADDRESS > 0 && addr + size > VM_MAXUSER_ADDRESS)
235 if (VM_MIN_ADDRESS > 0 && addr < VM_MIN_ADDRESS)
238 if (addr + size < addr)
242 * XXX for non-fixed mappings where no hint is provided or
243 * the hint would fall in the potential heap space,
244 * place it after the end of the largest possible heap.
246 * There should really be a pmap call to determine a reasonable
249 else if (addr == 0 ||
250 (addr >= round_page((vm_offset_t)vms->vm_taddr) &&
251 addr < round_page((vm_offset_t)vms->vm_daddr + maxdsiz)))
252 addr = round_page((vm_offset_t)vms->vm_daddr + maxdsiz);
254 if (flags & MAP_ANON) {
256 * Mapping blank space is trivial.
259 maxprot = VM_PROT_ALL;
263 * Mapping file, get fp for validation. Obtain vnode and make
264 * sure it is of appropriate type.
266 if (((unsigned) uap->fd) >= fdp->fd_nfiles ||
267 (fp = fdp->fd_ofiles[uap->fd]) == NULL)
269 if (fp->f_type != DTYPE_VNODE)
272 * POSIX shared-memory objects are defined to have
273 * kernel persistence, and are not defined to support
274 * read(2)/write(2) -- or even open(2). Thus, we can
275 * use MAP_ASYNC to trade on-disk coherence for speed.
276 * The shm_open(3) library routine turns on the FPOSIXSHM
277 * flag to request this behavior.
279 if (fp->f_flag & FPOSIXSHM)
281 vp = (struct vnode *) fp->f_data;
282 if (vp->v_type != VREG && vp->v_type != VCHR)
284 if (vp->v_type == VREG) {
286 * Get the proper underlying object
288 if (VOP_GETVOBJECT(vp, &obj) != 0)
290 vp = (struct vnode*)obj->handle;
294 * don't let the descriptor disappear on us if we block
299 * XXX hack to handle use of /dev/zero to map anon memory (ala
302 if (vp->v_type == VCHR && iszerodev(vp->v_rdev)) {
304 maxprot = VM_PROT_ALL;
309 * cdevs does not provide private mappings of any kind.
312 * However, for XIG X server to continue to work,
313 * we should allow the superuser to do it anyway.
314 * We only allow it at securelevel < 1.
315 * (Because the XIG X server writes directly to video
316 * memory via /dev/mem, it should never work at any
318 * XXX this will have to go
320 if (securelevel >= 1)
321 disablexworkaround = 1;
323 disablexworkaround = suser(td);
324 if (vp->v_type == VCHR && disablexworkaround &&
325 (flags & (MAP_PRIVATE|MAP_COPY))) {
330 * Ensure that file and memory protections are
331 * compatible. Note that we only worry about
332 * writability if mapping is shared; in this case,
333 * current and max prot are dictated by the open file.
334 * XXX use the vnode instead? Problem is: what
335 * credentials do we use for determination? What if
336 * proc does a setuid?
338 maxprot = VM_PROT_EXECUTE; /* ??? */
339 if (fp->f_flag & FREAD) {
340 maxprot |= VM_PROT_READ;
341 } else if (prot & PROT_READ) {
346 * If we are sharing potential changes (either via
347 * MAP_SHARED or via the implicit sharing of character
348 * device mappings), and we are trying to get write
349 * permission although we opened it without asking
350 * for it, bail out. Check for superuser, only if
351 * we're at securelevel < 1, to allow the XIG X server
352 * to continue to work.
355 if ((flags & MAP_SHARED) != 0 ||
356 (vp->v_type == VCHR && disablexworkaround)) {
357 if ((fp->f_flag & FWRITE) != 0) {
360 VOP_GETATTR(vp, &va, p->p_ucred, td))) {
364 (IMMUTABLE|APPEND)) == 0) {
365 maxprot |= VM_PROT_WRITE;
366 } else if (prot & PROT_WRITE) {
370 } else if ((prot & PROT_WRITE) != 0) {
375 maxprot |= VM_PROT_WRITE;
382 * Do not allow more then a certain number of vm_map_entry structures
383 * per process. Scale with the number of rforks sharing the map
384 * to make the limit reasonable for threads.
387 vms->vm_map.nentries >= max_proc_mmap * vms->vm_refcnt) {
392 error = vm_mmap(&vms->vm_map, &addr, size, prot, maxprot,
395 p->p_retval[0] = (register_t) (addr + pageoff);
404 * ommap_args(caddr_t addr, int len, int prot, int flags, int fd, long pos)
407 ommap(struct ommap_args *uap)
409 struct mmap_args nargs;
410 static const char cvtbsdprot[8] = {
414 PROT_EXEC | PROT_WRITE,
416 PROT_EXEC | PROT_READ,
417 PROT_WRITE | PROT_READ,
418 PROT_EXEC | PROT_WRITE | PROT_READ,
421 #define OMAP_ANON 0x0002
422 #define OMAP_COPY 0x0020
423 #define OMAP_SHARED 0x0010
424 #define OMAP_FIXED 0x0100
425 #define OMAP_INHERIT 0x0800
427 nargs.addr = uap->addr;
428 nargs.len = uap->len;
429 nargs.prot = cvtbsdprot[uap->prot & 0x7];
431 if (uap->flags & OMAP_ANON)
432 nargs.flags |= MAP_ANON;
433 if (uap->flags & OMAP_COPY)
434 nargs.flags |= MAP_COPY;
435 if (uap->flags & OMAP_SHARED)
436 nargs.flags |= MAP_SHARED;
438 nargs.flags |= MAP_PRIVATE;
439 if (uap->flags & OMAP_FIXED)
440 nargs.flags |= MAP_FIXED;
441 if (uap->flags & OMAP_INHERIT)
442 nargs.flags |= MAP_INHERIT;
444 nargs.pos = uap->pos;
445 return (mmap(&nargs));
447 #endif /* COMPAT_43 */
451 * msync_args(void *addr, int len, int flags)
454 msync(struct msync_args *uap)
456 struct proc *p = curproc;
458 vm_size_t size, pageoff;
463 addr = (vm_offset_t) uap->addr;
467 pageoff = (addr & PAGE_MASK);
470 size = (vm_size_t) round_page(size);
471 if (addr + size < addr)
474 if ((flags & (MS_ASYNC|MS_INVALIDATE)) == (MS_ASYNC|MS_INVALIDATE))
477 map = &p->p_vmspace->vm_map;
480 * XXX Gak! If size is zero we are supposed to sync "all modified
481 * pages with the region containing addr". Unfortunately, we don't
482 * really keep track of individual mmaps so we approximate by flushing
483 * the range of the map entry containing addr. This can be incorrect
484 * if the region splits or is coalesced with a neighbor.
487 vm_map_entry_t entry;
489 vm_map_lock_read(map);
490 rv = vm_map_lookup_entry(map, addr, &entry);
491 vm_map_unlock_read(map);
495 size = entry->end - entry->start;
499 * Clean the pages and interpret the return value.
501 rv = vm_map_clean(map, addr, addr + size, (flags & MS_ASYNC) == 0,
502 (flags & MS_INVALIDATE) != 0);
507 case KERN_INVALID_ADDRESS:
508 return (EINVAL); /* Sun returns ENOMEM? */
519 * munmap_args(void *addr, size_t len)
522 munmap(struct munmap_args *uap)
524 struct proc *p = curproc;
526 vm_size_t size, pageoff;
529 addr = (vm_offset_t) uap->addr;
532 pageoff = (addr & PAGE_MASK);
535 size = (vm_size_t) round_page(size);
536 if (addr + size < addr)
543 * Check for illegal addresses. Watch out for address wrap... Note
544 * that VM_*_ADDRESS are not constants due to casts (argh).
546 if (VM_MAXUSER_ADDRESS > 0 && addr + size > VM_MAXUSER_ADDRESS)
549 if (VM_MIN_ADDRESS > 0 && addr < VM_MIN_ADDRESS)
552 map = &p->p_vmspace->vm_map;
554 * Make sure entire range is allocated.
556 if (!vm_map_check_protection(map, addr, addr + size, VM_PROT_NONE))
558 /* returns nothing but KERN_SUCCESS anyway */
559 (void) vm_map_remove(map, addr, addr + size);
570 * XXX should unmap any regions mapped to this file
572 p->p_fd->fd_ofileflags[fd] &= ~UF_MAPPED;
577 * mprotect_args(const void *addr, size_t len, int prot)
580 mprotect(struct mprotect_args *uap)
582 struct proc *p = curproc;
584 vm_size_t size, pageoff;
585 register vm_prot_t prot;
587 addr = (vm_offset_t) uap->addr;
589 prot = uap->prot & VM_PROT_ALL;
590 #if defined(VM_PROT_READ_IS_EXEC)
591 if (prot & VM_PROT_READ)
592 prot |= VM_PROT_EXECUTE;
595 pageoff = (addr & PAGE_MASK);
598 size = (vm_size_t) round_page(size);
599 if (addr + size < addr)
602 switch (vm_map_protect(&p->p_vmspace->vm_map, addr, addr + size, prot,
606 case KERN_PROTECTION_FAILURE:
613 * minherit_args(void *addr, size_t len, int inherit)
616 minherit(struct minherit_args *uap)
618 struct proc *p = curproc;
620 vm_size_t size, pageoff;
621 register vm_inherit_t inherit;
623 addr = (vm_offset_t)uap->addr;
625 inherit = uap->inherit;
627 pageoff = (addr & PAGE_MASK);
630 size = (vm_size_t) round_page(size);
631 if (addr + size < addr)
634 switch (vm_map_inherit(&p->p_vmspace->vm_map, addr, addr+size,
638 case KERN_PROTECTION_FAILURE:
645 * madvise_args(void *addr, size_t len, int behav)
649 madvise(struct madvise_args *uap)
651 struct proc *p = curproc;
652 vm_offset_t start, end;
655 * Check for illegal behavior
657 if (uap->behav < 0 || uap->behav > MADV_CORE)
660 * Check for illegal addresses. Watch out for address wrap... Note
661 * that VM_*_ADDRESS are not constants due to casts (argh).
663 if (VM_MAXUSER_ADDRESS > 0 &&
664 ((vm_offset_t) uap->addr + uap->len) > VM_MAXUSER_ADDRESS)
667 if (VM_MIN_ADDRESS > 0 && uap->addr < VM_MIN_ADDRESS)
670 if (((vm_offset_t) uap->addr + uap->len) < (vm_offset_t) uap->addr)
674 * Since this routine is only advisory, we default to conservative
677 start = trunc_page((vm_offset_t) uap->addr);
678 end = round_page((vm_offset_t) uap->addr + uap->len);
680 if (vm_map_madvise(&p->p_vmspace->vm_map, start, end, uap->behav))
686 * mincore_args(const void *addr, size_t len, char *vec)
690 mincore(struct mincore_args *uap)
692 struct proc *p = curproc;
693 vm_offset_t addr, first_addr;
694 vm_offset_t end, cend;
699 int vecindex, lastvecindex;
700 register vm_map_entry_t current;
701 vm_map_entry_t entry;
703 unsigned int timestamp;
706 * Make sure that the addresses presented are valid for user
709 first_addr = addr = trunc_page((vm_offset_t) uap->addr);
710 end = addr + (vm_size_t)round_page(uap->len);
711 if (VM_MAXUSER_ADDRESS > 0 && end > VM_MAXUSER_ADDRESS)
717 * Address of byte vector
721 map = &p->p_vmspace->vm_map;
722 pmap = vmspace_pmap(p->p_vmspace);
724 vm_map_lock_read(map);
726 timestamp = map->timestamp;
728 if (!vm_map_lookup_entry(map, addr, &entry))
732 * Do this on a map entry basis so that if the pages are not
733 * in the current processes address space, we can easily look
734 * up the pages elsewhere.
738 (current != &map->header) && (current->start < end);
739 current = current->next) {
742 * ignore submaps (for now) or null objects
744 if ((current->eflags & MAP_ENTRY_IS_SUB_MAP) ||
745 current->object.vm_object == NULL)
749 * limit this scan to the current map entry and the
750 * limits for the mincore call
752 if (addr < current->start)
753 addr = current->start;
759 * scan this entry one page at a time
763 * Check pmap first, it is likely faster, also
764 * it can provide info as to whether we are the
765 * one referencing or modifying the page.
767 mincoreinfo = pmap_mincore(pmap, addr);
773 * calculate the page index into the object
775 offset = current->offset + (addr - current->start);
776 pindex = OFF_TO_IDX(offset);
777 m = vm_page_lookup(current->object.vm_object,
780 * if the page is resident, then gather information about
784 mincoreinfo = MINCORE_INCORE;
787 mincoreinfo |= MINCORE_MODIFIED_OTHER;
788 if ((m->flags & PG_REFERENCED) ||
789 pmap_ts_referenced(m)) {
790 vm_page_flag_set(m, PG_REFERENCED);
791 mincoreinfo |= MINCORE_REFERENCED_OTHER;
797 * subyte may page fault. In case it needs to modify
798 * the map, we release the lock.
800 vm_map_unlock_read(map);
803 * calculate index into user supplied byte vector
805 vecindex = OFF_TO_IDX(addr - first_addr);
808 * If we have skipped map entries, we need to make sure that
809 * the byte vector is zeroed for those skipped entries.
811 while((lastvecindex + 1) < vecindex) {
812 error = subyte( vec + lastvecindex, 0);
820 * Pass the page information to the user
822 error = subyte( vec + vecindex, mincoreinfo);
828 * If the map has changed, due to the subyte, the previous
829 * output may be invalid.
831 vm_map_lock_read(map);
832 if (timestamp != map->timestamp)
835 lastvecindex = vecindex;
841 * subyte may page fault. In case it needs to modify
842 * the map, we release the lock.
844 vm_map_unlock_read(map);
847 * Zero the last entries in the byte vector.
849 vecindex = OFF_TO_IDX(end - first_addr);
850 while((lastvecindex + 1) < vecindex) {
851 error = subyte( vec + lastvecindex, 0);
859 * If the map has changed, due to the subyte, the previous
860 * output may be invalid.
862 vm_map_lock_read(map);
863 if (timestamp != map->timestamp)
865 vm_map_unlock_read(map);
871 * mlock_args(const void *addr, size_t len)
874 mlock(struct mlock_args *uap)
877 vm_size_t size, pageoff;
879 struct proc *p = curproc;
881 addr = (vm_offset_t) uap->addr;
884 pageoff = (addr & PAGE_MASK);
887 size = (vm_size_t) round_page(size);
889 /* disable wrap around */
890 if (addr + size < addr)
893 if (atop(size) + cnt.v_wire_count > vm_page_max_wired)
896 #ifdef pmap_wired_count
897 if (size + ptoa(pmap_wired_count(vm_map_pmap(&p->p_vmspace->vm_map))) >
898 p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur)
901 error = suser_cred(p->p_ucred, 0);
906 error = vm_map_user_pageable(&p->p_vmspace->vm_map, addr, addr + size, FALSE);
907 return (error == KERN_SUCCESS ? 0 : ENOMEM);
911 * mlockall_args(int how)
914 mlockall(struct mlockall_args *uap)
920 * mlockall_args(int how)
923 munlockall(struct munlockall_args *uap)
929 * munlock_args(const void *addr, size_t len)
932 munlock(struct munlock_args *uap)
934 struct thread *td = curthread;
935 struct proc *p = td->td_proc;
937 vm_size_t size, pageoff;
940 addr = (vm_offset_t) uap->addr;
943 pageoff = (addr & PAGE_MASK);
946 size = (vm_size_t) round_page(size);
948 /* disable wrap around */
949 if (addr + size < addr)
952 #ifndef pmap_wired_count
958 error = vm_map_user_pageable(&p->p_vmspace->vm_map, addr, addr + size, TRUE);
959 return (error == KERN_SUCCESS ? 0 : ENOMEM);
963 * Internal version of mmap.
964 * Currently used by mmap, exec, and sys5 shared memory.
965 * Handle is either a vnode pointer or NULL for MAP_ANON.
968 vm_mmap(vm_map_t map, vm_offset_t *addr, vm_size_t size, vm_prot_t prot,
969 vm_prot_t maxprot, int flags,
975 struct vnode *vp = NULL;
977 int rv = KERN_SUCCESS;
978 vm_ooffset_t objsize;
980 struct thread *td = curthread; /* XXX */
981 struct proc *p = td->td_proc;
988 objsize = size = round_page(size);
990 if (p->p_vmspace->vm_map.size + size >
991 p->p_rlimit[RLIMIT_VMEM].rlim_cur) {
996 * We currently can only deal with page aligned file offsets.
997 * The check is here rather than in the syscall because the
998 * kernel calls this function internally for other mmaping
999 * operations (such as in exec) and non-aligned offsets will
1000 * cause pmap inconsistencies...so we want to be sure to
1001 * disallow this in all cases.
1003 if (foff & PAGE_MASK)
1006 if ((flags & MAP_FIXED) == 0) {
1008 *addr = round_page(*addr);
1010 if (*addr != trunc_page(*addr))
1013 (void) vm_map_remove(map, *addr, *addr + size);
1017 * Lookup/allocate object.
1019 if (flags & MAP_ANON) {
1020 type = OBJT_DEFAULT;
1022 * Unnamed anonymous regions always start at 0.
1027 vp = (struct vnode *) handle;
1028 if (vp->v_type == VCHR) {
1030 handle = (void *)(intptr_t)vp->v_rdev;
1035 error = VOP_GETATTR(vp, &vat, p->p_ucred, td);
1038 objsize = round_page(vat.va_size);
1041 * if it is a regular file without any references
1042 * we do not need to sync it.
1044 if (vp->v_type == VREG && vat.va_nlink == 0) {
1045 flags |= MAP_NOSYNC;
1050 if (handle == NULL) {
1054 object = vm_pager_allocate(type,
1055 handle, objsize, prot, foff);
1057 return (type == OBJT_DEVICE ? EINVAL : ENOMEM);
1058 docow = MAP_PREFAULT_PARTIAL;
1062 * Force device mappings to be shared.
1064 if (type == OBJT_DEVICE || type == OBJT_PHYS) {
1065 flags &= ~(MAP_PRIVATE|MAP_COPY);
1066 flags |= MAP_SHARED;
1069 if ((flags & (MAP_ANON|MAP_SHARED)) == 0)
1070 docow |= MAP_COPY_ON_WRITE;
1071 if (flags & MAP_NOSYNC)
1072 docow |= MAP_DISABLE_SYNCER;
1073 if (flags & MAP_NOCORE)
1074 docow |= MAP_DISABLE_COREDUMP;
1076 #if defined(VM_PROT_READ_IS_EXEC)
1077 if (prot & VM_PROT_READ)
1078 prot |= VM_PROT_EXECUTE;
1080 if (maxprot & VM_PROT_READ)
1081 maxprot |= VM_PROT_EXECUTE;
1085 *addr = pmap_addr_hint(object, *addr, size);
1088 if (flags & MAP_STACK)
1089 rv = vm_map_stack (map, *addr, size, prot,
1092 rv = vm_map_find(map, object, foff, addr, size, fitit,
1093 prot, maxprot, docow);
1095 if (rv != KERN_SUCCESS) {
1097 * Lose the object reference. Will destroy the
1098 * object if it's an unnamed anonymous mapping
1099 * or named anonymous without other references.
1101 vm_object_deallocate(object);
1106 * Shared memory is also shared with children.
1108 if (flags & (MAP_SHARED|MAP_INHERIT)) {
1109 rv = vm_map_inherit(map, *addr, *addr + size, VM_INHERIT_SHARE);
1110 if (rv != KERN_SUCCESS) {
1111 (void) vm_map_remove(map, *addr, *addr + size);
1119 case KERN_INVALID_ADDRESS:
1122 case KERN_PROTECTION_FAILURE: