xref: /freebsd/sys/vm/vm_mmap.c (revision 3e0f6b97b257a96f7275e4442204263e44b16686)
1 /*
2  * Copyright (c) 1988 University of Utah.
3  * Copyright (c) 1991, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * the Systems Programming Group of the University of Utah Computer
8  * Science Department.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
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.
25  *
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
36  * SUCH DAMAGE.
37  *
38  * from: Utah $Hdr: vm_mmap.c 1.6 91/10/21$
39  *
40  *	@(#)vm_mmap.c	8.4 (Berkeley) 1/12/94
41  * $FreeBSD$
42  */
43 
44 /*
45  * Mapped file (mmap) interface to VM
46  */
47 
48 #include "opt_rlimit.h"
49 
50 #include <sys/param.h>
51 #include <sys/systm.h>
52 #include <sys/sysproto.h>
53 #include <sys/filedesc.h>
54 #include <sys/resourcevar.h>
55 #include <sys/proc.h>
56 #include <sys/vnode.h>
57 #include <sys/file.h>
58 #include <sys/mman.h>
59 #include <sys/conf.h>
60 #include <sys/vmmeter.h>
61 
62 #include <miscfs/specfs/specdev.h>
63 
64 #include <vm/vm.h>
65 #include <vm/vm_param.h>
66 #include <vm/vm_prot.h>
67 #include <vm/vm_inherit.h>
68 #include <sys/lock.h>
69 #include <vm/pmap.h>
70 #include <vm/vm_map.h>
71 #include <vm/vm_object.h>
72 #include <vm/vm_pager.h>
73 #include <vm/vm_pageout.h>
74 #include <vm/vm_extern.h>
75 #include <vm/vm_kern.h>
76 #include <vm/vm_page.h>
77 
78 #ifndef _SYS_SYSPROTO_H_
79 struct sbrk_args {
80 	int incr;
81 };
82 #endif
83 
84 /* ARGSUSED */
85 int
86 sbrk(p, uap, retval)
87 	struct proc *p;
88 	struct sbrk_args *uap;
89 	int *retval;
90 {
91 
92 	/* Not yet implemented */
93 	return (EOPNOTSUPP);
94 }
95 
96 #ifndef _SYS_SYSPROTO_H_
97 struct sstk_args {
98 	int incr;
99 };
100 #endif
101 
102 /* ARGSUSED */
103 int
104 sstk(p, uap, retval)
105 	struct proc *p;
106 	struct sstk_args *uap;
107 	int *retval;
108 {
109 
110 	/* Not yet implemented */
111 	return (EOPNOTSUPP);
112 }
113 
114 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
115 #ifndef _SYS_SYSPROTO_H_
116 struct getpagesize_args {
117 	int dummy;
118 };
119 #endif
120 
121 /* ARGSUSED */
122 int
123 ogetpagesize(p, uap, retval)
124 	struct proc *p;
125 	struct getpagesize_args *uap;
126 	int *retval;
127 {
128 
129 	*retval = PAGE_SIZE;
130 	return (0);
131 }
132 #endif				/* COMPAT_43 || COMPAT_SUNOS */
133 
134 #ifndef _SYS_SYSPROTO_H_
135 struct mmap_args {
136 	caddr_t addr;
137 	size_t len;
138 	int prot;
139 	int flags;
140 	int fd;
141 	long pad;
142 	off_t pos;
143 };
144 #endif
145 
146 int
147 mmap(p, uap, retval)
148 	struct proc *p;
149 	register struct mmap_args *uap;
150 	int *retval;
151 {
152 	register struct filedesc *fdp = p->p_fd;
153 	register struct file *fp;
154 	struct vnode *vp;
155 	vm_offset_t addr;
156 	vm_size_t size, pageoff;
157 	vm_prot_t prot, maxprot;
158 	caddr_t handle;
159 	int flags, error;
160 
161 	prot = uap->prot & VM_PROT_ALL;
162 	flags = uap->flags;
163 	/*
164 	 * Address (if FIXED) must be page aligned. Size is implicitly rounded
165 	 * to a page boundary.
166 	 */
167 	addr = (vm_offset_t) uap->addr;
168 	if (((flags & MAP_FIXED) && (addr & PAGE_MASK)) ||
169 	    (ssize_t) uap->len < 0 || ((flags & MAP_ANON) && uap->fd != -1))
170 		return (EINVAL);
171 
172 	/*
173 	 * Round page if not already disallowed by above test
174 	 * XXX: Is there any point in the MAP_FIXED align requirement above?
175 	 */
176 	size = uap->len;
177 	pageoff = (addr & PAGE_MASK);
178 	addr -= pageoff;
179 	size += pageoff;
180 	size = (vm_size_t) round_page(size);
181 
182 	/*
183 	 * Check for illegal addresses.  Watch out for address wrap... Note
184 	 * that VM_*_ADDRESS are not constants due to casts (argh).
185 	 */
186 	if (flags & MAP_FIXED) {
187 		if (VM_MAXUSER_ADDRESS > 0 && addr + size > VM_MAXUSER_ADDRESS)
188 			return (EINVAL);
189 #ifndef i386
190 		if (VM_MIN_ADDRESS > 0 && addr < VM_MIN_ADDRESS)
191 			return (EINVAL);
192 #endif
193 		if (addr + size < addr)
194 			return (EINVAL);
195 	}
196 	/*
197 	 * XXX if no hint provided for a non-fixed mapping place it after the
198 	 * end of the largest possible heap.
199 	 *
200 	 * There should really be a pmap call to determine a reasonable location.
201 	 */
202 	if (addr == 0 && (flags & MAP_FIXED) == 0)
203 		addr = round_page(p->p_vmspace->vm_daddr + MAXDSIZ);
204 	if (flags & MAP_ANON) {
205 		/*
206 		 * Mapping blank space is trivial.
207 		 */
208 		handle = NULL;
209 		maxprot = VM_PROT_ALL;
210 	} else {
211 		/*
212 		 * Mapping file, get fp for validation. Obtain vnode and make
213 		 * sure it is of appropriate type.
214 		 */
215 		if (((unsigned) uap->fd) >= fdp->fd_nfiles ||
216 		    (fp = fdp->fd_ofiles[uap->fd]) == NULL)
217 			return (EBADF);
218 		if (fp->f_type != DTYPE_VNODE)
219 			return (EINVAL);
220 		vp = (struct vnode *) fp->f_data;
221 		if (vp->v_type != VREG && vp->v_type != VCHR)
222 			return (EINVAL);
223 		/*
224 		 * XXX hack to handle use of /dev/zero to map anon memory (ala
225 		 * SunOS).
226 		 */
227 		if (vp->v_type == VCHR && iszerodev(vp->v_rdev)) {
228 			handle = NULL;
229 			maxprot = VM_PROT_ALL;
230 			flags |= MAP_ANON;
231 		} else {
232 			/*
233 			 * Ensure that file and memory protections are
234 			 * compatible.  Note that we only worry about
235 			 * writability if mapping is shared; in this case,
236 			 * current and max prot are dictated by the open file.
237 			 * XXX use the vnode instead?  Problem is: what
238 			 * credentials do we use for determination? What if
239 			 * proc does a setuid?
240 			 */
241 			maxprot = VM_PROT_EXECUTE;	/* ??? */
242 			if (fp->f_flag & FREAD)
243 				maxprot |= VM_PROT_READ;
244 			else if (prot & PROT_READ)
245 				return (EACCES);
246 			if (flags & MAP_SHARED) {
247 				if (fp->f_flag & FWRITE)
248 					maxprot |= VM_PROT_WRITE;
249 				else if (prot & PROT_WRITE)
250 					return (EACCES);
251 			} else
252 				maxprot |= VM_PROT_WRITE;
253 			handle = (caddr_t) vp;
254 		}
255 	}
256 	error = vm_mmap(&p->p_vmspace->vm_map, &addr, size, prot, maxprot,
257 	    flags, handle, uap->pos);
258 	if (error == 0)
259 		*retval = (int) addr;
260 	return (error);
261 }
262 
263 #ifdef COMPAT_43
264 #ifndef _SYS_SYSPROTO_H_
265 struct ommap_args {
266 	caddr_t addr;
267 	int len;
268 	int prot;
269 	int flags;
270 	int fd;
271 	long pos;
272 };
273 #endif
274 int
275 ommap(p, uap, retval)
276 	struct proc *p;
277 	register struct ommap_args *uap;
278 	int *retval;
279 {
280 	struct mmap_args nargs;
281 	static const char cvtbsdprot[8] = {
282 		0,
283 		PROT_EXEC,
284 		PROT_WRITE,
285 		PROT_EXEC | PROT_WRITE,
286 		PROT_READ,
287 		PROT_EXEC | PROT_READ,
288 		PROT_WRITE | PROT_READ,
289 		PROT_EXEC | PROT_WRITE | PROT_READ,
290 	};
291 
292 #define	OMAP_ANON	0x0002
293 #define	OMAP_COPY	0x0020
294 #define	OMAP_SHARED	0x0010
295 #define	OMAP_FIXED	0x0100
296 #define	OMAP_INHERIT	0x0800
297 
298 	nargs.addr = uap->addr;
299 	nargs.len = uap->len;
300 	nargs.prot = cvtbsdprot[uap->prot & 0x7];
301 	nargs.flags = 0;
302 	if (uap->flags & OMAP_ANON)
303 		nargs.flags |= MAP_ANON;
304 	if (uap->flags & OMAP_COPY)
305 		nargs.flags |= MAP_COPY;
306 	if (uap->flags & OMAP_SHARED)
307 		nargs.flags |= MAP_SHARED;
308 	else
309 		nargs.flags |= MAP_PRIVATE;
310 	if (uap->flags & OMAP_FIXED)
311 		nargs.flags |= MAP_FIXED;
312 	if (uap->flags & OMAP_INHERIT)
313 		nargs.flags |= MAP_INHERIT;
314 	nargs.fd = uap->fd;
315 	nargs.pos = uap->pos;
316 	return (mmap(p, &nargs, retval));
317 }
318 #endif				/* COMPAT_43 */
319 
320 
321 #ifndef _SYS_SYSPROTO_H_
322 struct msync_args {
323 	caddr_t addr;
324 	int len;
325 	int flags;
326 };
327 #endif
328 int
329 msync(p, uap, retval)
330 	struct proc *p;
331 	struct msync_args *uap;
332 	int *retval;
333 {
334 	vm_offset_t addr;
335 	vm_size_t size, pageoff;
336 	int flags;
337 	vm_map_t map;
338 	int rv;
339 
340 	addr = (vm_offset_t) uap->addr;
341 	size = uap->len;
342 	flags = uap->flags;
343 
344 	pageoff = (addr & PAGE_MASK);
345 	addr -= pageoff;
346 	size += pageoff;
347 	size = (vm_size_t) round_page(size);
348 	if (addr + size < addr)
349 		return(EINVAL);
350 
351 	if ((flags & (MS_ASYNC|MS_INVALIDATE)) == (MS_ASYNC|MS_INVALIDATE))
352 		return (EINVAL);
353 
354 	map = &p->p_vmspace->vm_map;
355 
356 	/*
357 	 * XXX Gak!  If size is zero we are supposed to sync "all modified
358 	 * pages with the region containing addr".  Unfortunately, we don't
359 	 * really keep track of individual mmaps so we approximate by flushing
360 	 * the range of the map entry containing addr. This can be incorrect
361 	 * if the region splits or is coalesced with a neighbor.
362 	 */
363 	if (size == 0) {
364 		vm_map_entry_t entry;
365 
366 		vm_map_lock_read(map);
367 		rv = vm_map_lookup_entry(map, addr, &entry);
368 		vm_map_unlock_read(map);
369 		if (rv == FALSE)
370 			return (EINVAL);
371 		addr = entry->start;
372 		size = entry->end - entry->start;
373 	}
374 
375 	/*
376 	 * Clean the pages and interpret the return value.
377 	 */
378 	rv = vm_map_clean(map, addr, addr + size, (flags & MS_ASYNC) == 0,
379 	    (flags & MS_INVALIDATE) != 0);
380 
381 	switch (rv) {
382 	case KERN_SUCCESS:
383 		break;
384 	case KERN_INVALID_ADDRESS:
385 		return (EINVAL);	/* Sun returns ENOMEM? */
386 	case KERN_FAILURE:
387 		return (EIO);
388 	default:
389 		return (EINVAL);
390 	}
391 
392 	return (0);
393 }
394 
395 #ifndef _SYS_SYSPROTO_H_
396 struct munmap_args {
397 	caddr_t addr;
398 	size_t len;
399 };
400 #endif
401 int
402 munmap(p, uap, retval)
403 	register struct proc *p;
404 	register struct munmap_args *uap;
405 	int *retval;
406 {
407 	vm_offset_t addr;
408 	vm_size_t size, pageoff;
409 	vm_map_t map;
410 
411 	addr = (vm_offset_t) uap->addr;
412 	size = uap->len;
413 
414 	pageoff = (addr & PAGE_MASK);
415 	addr -= pageoff;
416 	size += pageoff;
417 	size = (vm_size_t) round_page(size);
418 	if (addr + size < addr)
419 		return(EINVAL);
420 
421 	if (size == 0)
422 		return (0);
423 
424 	/*
425 	 * Check for illegal addresses.  Watch out for address wrap... Note
426 	 * that VM_*_ADDRESS are not constants due to casts (argh).
427 	 */
428 	if (VM_MAXUSER_ADDRESS > 0 && addr + size > VM_MAXUSER_ADDRESS)
429 		return (EINVAL);
430 #ifndef i386
431 	if (VM_MIN_ADDRESS > 0 && addr < VM_MIN_ADDRESS)
432 		return (EINVAL);
433 #endif
434 	if (addr + size < addr)
435 		return (EINVAL);
436 	map = &p->p_vmspace->vm_map;
437 	/*
438 	 * Make sure entire range is allocated.
439 	 */
440 	if (!vm_map_check_protection(map, addr, addr + size, VM_PROT_NONE))
441 		return (EINVAL);
442 	/* returns nothing but KERN_SUCCESS anyway */
443 	(void) vm_map_remove(map, addr, addr + size);
444 	return (0);
445 }
446 
447 void
448 munmapfd(p, fd)
449 	struct proc *p;
450 	int fd;
451 {
452 	/*
453 	 * XXX should unmap any regions mapped to this file
454 	 */
455 	p->p_fd->fd_ofileflags[fd] &= ~UF_MAPPED;
456 }
457 
458 #ifndef _SYS_SYSPROTO_H_
459 struct mprotect_args {
460 	caddr_t addr;
461 	size_t len;
462 	int prot;
463 };
464 #endif
465 int
466 mprotect(p, uap, retval)
467 	struct proc *p;
468 	struct mprotect_args *uap;
469 	int *retval;
470 {
471 	vm_offset_t addr;
472 	vm_size_t size, pageoff;
473 	register vm_prot_t prot;
474 
475 	addr = (vm_offset_t) uap->addr;
476 	size = uap->len;
477 	prot = uap->prot & VM_PROT_ALL;
478 #if defined(VM_PROT_READ_IS_EXEC)
479 	if (prot & VM_PROT_READ)
480 		prot |= VM_PROT_EXECUTE;
481 #endif
482 
483 	pageoff = (addr & PAGE_MASK);
484 	addr -= pageoff;
485 	size += pageoff;
486 	size = (vm_size_t) round_page(size);
487 	if (addr + size < addr)
488 		return(EINVAL);
489 
490 	switch (vm_map_protect(&p->p_vmspace->vm_map, addr, addr + size, prot,
491 		FALSE)) {
492 	case KERN_SUCCESS:
493 		return (0);
494 	case KERN_PROTECTION_FAILURE:
495 		return (EACCES);
496 	}
497 	return (EINVAL);
498 }
499 
500 #ifndef _SYS_SYSPROTO_H_
501 struct minherit_args {
502 	caddr_t addr;
503 	size_t len;
504 	int inherit;
505 };
506 #endif
507 int
508 minherit(p, uap, retval)
509 	struct proc *p;
510 	struct minherit_args *uap;
511 	int *retval;
512 {
513 	vm_offset_t addr;
514 	vm_size_t size, pageoff;
515 	register vm_inherit_t inherit;
516 
517 	addr = (vm_offset_t)uap->addr;
518 	size = uap->len;
519 	inherit = uap->inherit;
520 
521 	pageoff = (addr & PAGE_MASK);
522 	addr -= pageoff;
523 	size += pageoff;
524 	size = (vm_size_t) round_page(size);
525 	if (addr + size < addr)
526 		return(EINVAL);
527 
528 	switch (vm_map_inherit(&p->p_vmspace->vm_map, addr, addr+size,
529 	    inherit)) {
530 	case KERN_SUCCESS:
531 		return (0);
532 	case KERN_PROTECTION_FAILURE:
533 		return (EACCES);
534 	}
535 	return (EINVAL);
536 }
537 
538 #ifndef _SYS_SYSPROTO_H_
539 struct madvise_args {
540 	caddr_t addr;
541 	size_t len;
542 	int behav;
543 };
544 #endif
545 
546 /* ARGSUSED */
547 int
548 madvise(p, uap, retval)
549 	struct proc *p;
550 	struct madvise_args *uap;
551 	int *retval;
552 {
553 	vm_map_t map;
554 	pmap_t pmap;
555 	vm_offset_t start, end;
556 	/*
557 	 * Check for illegal addresses.  Watch out for address wrap... Note
558 	 * that VM_*_ADDRESS are not constants due to casts (argh).
559 	 */
560 	if (VM_MAXUSER_ADDRESS > 0 &&
561 		((vm_offset_t) uap->addr + uap->len) > VM_MAXUSER_ADDRESS)
562 		return (EINVAL);
563 #ifndef i386
564 	if (VM_MIN_ADDRESS > 0 && uap->addr < VM_MIN_ADDRESS)
565 		return (EINVAL);
566 #endif
567 	if (((vm_offset_t) uap->addr + uap->len) < (vm_offset_t) uap->addr)
568 		return (EINVAL);
569 
570 	/*
571 	 * Since this routine is only advisory, we default to conservative
572 	 * behavior.
573 	 */
574 	start = trunc_page((vm_offset_t) uap->addr);
575 	end = round_page((vm_offset_t) uap->addr + uap->len);
576 
577 	map = &p->p_vmspace->vm_map;
578 	pmap = &p->p_vmspace->vm_pmap;
579 
580 	vm_map_madvise(map, pmap, start, end, uap->behav);
581 
582 	return (0);
583 }
584 
585 #ifndef _SYS_SYSPROTO_H_
586 struct mincore_args {
587 	caddr_t addr;
588 	size_t len;
589 	char *vec;
590 };
591 #endif
592 
593 /* ARGSUSED */
594 int
595 mincore(p, uap, retval)
596 	struct proc *p;
597 	struct mincore_args *uap;
598 	int *retval;
599 {
600 	vm_offset_t addr, first_addr;
601 	vm_offset_t end, cend;
602 	pmap_t pmap;
603 	vm_map_t map;
604 	char *vec;
605 	int error;
606 	int vecindex, lastvecindex;
607 	register vm_map_entry_t current;
608 	vm_map_entry_t entry;
609 	int mincoreinfo;
610 
611 	/*
612 	 * Make sure that the addresses presented are valid for user
613 	 * mode.
614 	 */
615 	first_addr = addr = trunc_page((vm_offset_t) uap->addr);
616 	end = addr + (vm_size_t)round_page(uap->len);
617 	if (VM_MAXUSER_ADDRESS > 0 && end > VM_MAXUSER_ADDRESS)
618 		return (EINVAL);
619 	if (end < addr)
620 		return (EINVAL);
621 
622 	/*
623 	 * Address of byte vector
624 	 */
625 	vec = uap->vec;
626 
627 	map = &p->p_vmspace->vm_map;
628 	pmap = &p->p_vmspace->vm_pmap;
629 
630 	vm_map_lock(map);
631 
632 	/*
633 	 * Not needed here
634 	 */
635 #if 0
636 	VM_MAP_RANGE_CHECK(map, addr, end);
637 #endif
638 
639 	if (!vm_map_lookup_entry(map, addr, &entry))
640 		entry = entry->next;
641 
642 	/*
643 	 * Do this on a map entry basis so that if the pages are not
644 	 * in the current processes address space, we can easily look
645 	 * up the pages elsewhere.
646 	 */
647 	lastvecindex = -1;
648 	for(current = entry;
649 		(current != &map->header) && (current->start < end);
650 		current = current->next) {
651 
652 		/*
653 		 * ignore submaps (for now) or null objects
654 		 */
655 		if ((current->eflags & (MAP_ENTRY_IS_A_MAP|MAP_ENTRY_IS_SUB_MAP)) ||
656 			current->object.vm_object == NULL)
657 			continue;
658 
659 		/*
660 		 * limit this scan to the current map entry and the
661 		 * limits for the mincore call
662 		 */
663 		if (addr < current->start)
664 			addr = current->start;
665 		cend = current->end;
666 		if (cend > end)
667 			cend = end;
668 
669 		/*
670 		 * scan this entry one page at a time
671 		 */
672 		while(addr < cend) {
673 			/*
674 			 * Check pmap first, it is likely faster, also
675 			 * it can provide info as to whether we are the
676 			 * one referencing or modifying the page.
677 			 */
678 			mincoreinfo = pmap_mincore(pmap, addr);
679 			if (!mincoreinfo) {
680 				vm_pindex_t pindex;
681 				vm_ooffset_t offset;
682 				vm_page_t m;
683 				/*
684 				 * calculate the page index into the object
685 				 */
686 				offset = current->offset + (addr - current->start);
687 				pindex = OFF_TO_IDX(offset);
688 				m = vm_page_lookup(current->object.vm_object,
689 					pindex);
690 				/*
691 				 * if the page is resident, then gather information about
692 				 * it.
693 				 */
694 				if (m) {
695 					mincoreinfo = MINCORE_INCORE;
696 					if (m->dirty ||
697 						pmap_is_modified(VM_PAGE_TO_PHYS(m)))
698 						mincoreinfo |= MINCORE_MODIFIED_OTHER;
699 					if ((m->flags & PG_REFERENCED) ||
700 						pmap_ts_referenced(VM_PAGE_TO_PHYS(m))) {
701 						m->flags |= PG_REFERENCED;
702 						mincoreinfo |= MINCORE_REFERENCED_OTHER;
703 					}
704 				}
705 			}
706 
707 			/*
708 			 * calculate index into user supplied byte vector
709 			 */
710 			vecindex = OFF_TO_IDX(addr - first_addr);
711 
712 			/*
713 			 * If we have skipped map entries, we need to make sure that
714 			 * the byte vector is zeroed for those skipped entries.
715 			 */
716 			while((lastvecindex + 1) < vecindex) {
717 				error = subyte( vec + lastvecindex, 0);
718 				if (error) {
719 					vm_map_unlock(map);
720 					return (EFAULT);
721 				}
722 				++lastvecindex;
723 			}
724 
725 			/*
726 			 * Pass the page information to the user
727 			 */
728 			error = subyte( vec + vecindex, mincoreinfo);
729 			if (error) {
730 				vm_map_unlock(map);
731 				return (EFAULT);
732 			}
733 			lastvecindex = vecindex;
734 			addr += PAGE_SIZE;
735 		}
736 	}
737 
738 	/*
739 	 * Zero the last entries in the byte vector.
740 	 */
741 	vecindex = OFF_TO_IDX(end - first_addr);
742 	while((lastvecindex + 1) < vecindex) {
743 		error = subyte( vec + lastvecindex, 0);
744 		if (error) {
745 			vm_map_unlock(map);
746 			return (EFAULT);
747 		}
748 		++lastvecindex;
749 	}
750 
751 	vm_map_unlock(map);
752 	return (0);
753 }
754 
755 #ifndef _SYS_SYSPROTO_H_
756 struct mlock_args {
757 	caddr_t addr;
758 	size_t len;
759 };
760 #endif
761 int
762 mlock(p, uap, retval)
763 	struct proc *p;
764 	struct mlock_args *uap;
765 	int *retval;
766 {
767 	vm_offset_t addr;
768 	vm_size_t size, pageoff;
769 	int error;
770 
771 	addr = (vm_offset_t) uap->addr;
772 	size = uap->len;
773 
774 	pageoff = (addr & PAGE_MASK);
775 	addr -= pageoff;
776 	size += pageoff;
777 	size = (vm_size_t) round_page(size);
778 
779 	/* disable wrap around */
780 	if (addr + size < addr)
781 		return (EINVAL);
782 
783 	if (atop(size) + cnt.v_wire_count > vm_page_max_wired)
784 		return (EAGAIN);
785 
786 #ifdef pmap_wired_count
787 	if (size + ptoa(pmap_wired_count(vm_map_pmap(&p->p_vmspace->vm_map))) >
788 	    p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur)
789 		return (EAGAIN);
790 #else
791 	error = suser(p->p_ucred, &p->p_acflag);
792 	if (error)
793 		return (error);
794 #endif
795 
796 	error = vm_map_user_pageable(&p->p_vmspace->vm_map, addr, addr + size, FALSE);
797 	return (error == KERN_SUCCESS ? 0 : ENOMEM);
798 }
799 
800 #ifndef _SYS_SYSPROTO_H_
801 struct munlock_args {
802 	caddr_t addr;
803 	size_t len;
804 };
805 #endif
806 int
807 munlock(p, uap, retval)
808 	struct proc *p;
809 	struct munlock_args *uap;
810 	int *retval;
811 {
812 	vm_offset_t addr;
813 	vm_size_t size, pageoff;
814 	int error;
815 
816 	addr = (vm_offset_t) uap->addr;
817 	size = uap->len;
818 
819 	pageoff = (addr & PAGE_MASK);
820 	addr -= pageoff;
821 	size += pageoff;
822 	size = (vm_size_t) round_page(size);
823 
824 	/* disable wrap around */
825 	if (addr + size < addr)
826 		return (EINVAL);
827 
828 #ifndef pmap_wired_count
829 	error = suser(p->p_ucred, &p->p_acflag);
830 	if (error)
831 		return (error);
832 #endif
833 
834 	error = vm_map_user_pageable(&p->p_vmspace->vm_map, addr, addr + size, TRUE);
835 	return (error == KERN_SUCCESS ? 0 : ENOMEM);
836 }
837 
838 /*
839  * Internal version of mmap.
840  * Currently used by mmap, exec, and sys5 shared memory.
841  * Handle is either a vnode pointer or NULL for MAP_ANON.
842  */
843 int
844 vm_mmap(map, addr, size, prot, maxprot, flags, handle, foff)
845 	register vm_map_t map;
846 	register vm_offset_t *addr;
847 	register vm_size_t size;
848 	vm_prot_t prot, maxprot;
849 	register int flags;
850 	caddr_t handle;		/* XXX should be vp */
851 	vm_ooffset_t foff;
852 {
853 	boolean_t fitit;
854 	vm_object_t object;
855 	struct vnode *vp = NULL;
856 	objtype_t type;
857 	int rv = KERN_SUCCESS;
858 	vm_ooffset_t objsize;
859 	int docow;
860 	struct proc *p = curproc;
861 
862 	if (size == 0)
863 		return (0);
864 
865 	objsize = size = round_page(size);
866 
867 	/*
868 	 * We currently can only deal with page aligned file offsets.
869 	 * The check is here rather than in the syscall because the
870 	 * kernel calls this function internally for other mmaping
871 	 * operations (such as in exec) and non-aligned offsets will
872 	 * cause pmap inconsistencies...so we want to be sure to
873 	 * disallow this in all cases.
874 	 */
875 	if (foff & PAGE_MASK)
876 		return (EINVAL);
877 
878 	if ((flags & MAP_FIXED) == 0) {
879 		fitit = TRUE;
880 		*addr = round_page(*addr);
881 	} else {
882 		if (*addr != trunc_page(*addr))
883 			return (EINVAL);
884 		fitit = FALSE;
885 		(void) vm_map_remove(map, *addr, *addr + size);
886 	}
887 
888 	/*
889 	 * Lookup/allocate object.
890 	 */
891 	if (flags & MAP_ANON) {
892 		type = OBJT_DEFAULT;
893 		/*
894 		 * Unnamed anonymous regions always start at 0.
895 		 */
896 		if (handle == 0)
897 			foff = 0;
898 	} else {
899 		vp = (struct vnode *) handle;
900 		if (vp->v_type == VCHR) {
901 			type = OBJT_DEVICE;
902 			handle = (caddr_t) vp->v_rdev;
903 		} else {
904 			struct vattr vat;
905 			int error;
906 
907 			error = VOP_GETATTR(vp, &vat, p->p_ucred, p);
908 			if (error)
909 				return (error);
910 			objsize = round_page(vat.va_size);
911 			type = OBJT_VNODE;
912 		}
913 	}
914 
915 	if (handle == NULL) {
916 		object = NULL;
917 	} else {
918 		object = vm_pager_allocate(type, handle, OFF_TO_IDX(objsize), prot, foff);
919 		if (object == NULL)
920 			return (type == OBJT_DEVICE ? EINVAL : ENOMEM);
921 	}
922 
923 	/*
924 	 * Force device mappings to be shared.
925 	 */
926 	if (type == OBJT_DEVICE) {
927 		flags &= ~(MAP_PRIVATE|MAP_COPY);
928 		flags |= MAP_SHARED;
929 	}
930 
931 	docow = 0;
932 	if ((flags & (MAP_ANON|MAP_SHARED)) == 0) {
933 		docow = MAP_COPY_ON_WRITE | MAP_COPY_NEEDED;
934 	}
935 
936 #if defined(VM_PROT_READ_IS_EXEC)
937 	if (prot & VM_PROT_READ)
938 		prot |= VM_PROT_EXECUTE;
939 
940 	if (maxprot & VM_PROT_READ)
941 		maxprot |= VM_PROT_EXECUTE;
942 #endif
943 
944 	rv = vm_map_find(map, object, foff, addr, size, fitit,
945 			prot, maxprot, docow);
946 
947 
948 	if (rv != KERN_SUCCESS) {
949 		/*
950 		 * Lose the object reference. Will destroy the
951 		 * object if it's an unnamed anonymous mapping
952 		 * or named anonymous without other references.
953 		 */
954 		vm_object_deallocate(object);
955 		goto out;
956 	}
957 
958 	/*
959 	 * "Pre-fault" resident pages.
960 	 */
961 	if ((type == OBJT_VNODE) && (map->pmap != NULL) && (object != NULL)) {
962 		pmap_object_init_pt(map->pmap, *addr,
963 			object, (vm_pindex_t) OFF_TO_IDX(foff), size, 1);
964 	}
965 
966 	/*
967 	 * Shared memory is also shared with children.
968 	 */
969 	if (flags & (MAP_SHARED|MAP_INHERIT)) {
970 		rv = vm_map_inherit(map, *addr, *addr + size, VM_INHERIT_SHARE);
971 		if (rv != KERN_SUCCESS) {
972 			(void) vm_map_remove(map, *addr, *addr + size);
973 			goto out;
974 		}
975 	}
976 out:
977 	switch (rv) {
978 	case KERN_SUCCESS:
979 		return (0);
980 	case KERN_INVALID_ADDRESS:
981 	case KERN_NO_SPACE:
982 		return (ENOMEM);
983 	case KERN_PROTECTION_FAILURE:
984 		return (EACCES);
985 	default:
986 		return (EINVAL);
987 	}
988 }
989