xref: /freebsd/sys/kern/kern_exec.c (revision 77a0943ded95b9e6438f7db70c4a28e4d93946d4)
1 /*
2  * Copyright (c) 1993, David Greenman
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD$
27  */
28 
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysproto.h>
32 #include <sys/signalvar.h>
33 #include <sys/kernel.h>
34 #include <sys/mount.h>
35 #include <sys/mutex.h>
36 #include <sys/filedesc.h>
37 #include <sys/fcntl.h>
38 #include <sys/acct.h>
39 #include <sys/exec.h>
40 #include <sys/imgact.h>
41 #include <sys/imgact_elf.h>
42 #include <sys/wait.h>
43 #include <sys/proc.h>
44 #include <sys/pioctl.h>
45 #include <sys/malloc.h>
46 #include <sys/namei.h>
47 #include <sys/sysent.h>
48 #include <sys/shm.h>
49 #include <sys/sysctl.h>
50 #include <sys/vnode.h>
51 
52 #include <vm/vm.h>
53 #include <vm/vm_param.h>
54 #include <sys/lock.h>
55 #include <vm/pmap.h>
56 #include <vm/vm_page.h>
57 #include <vm/vm_map.h>
58 #include <vm/vm_kern.h>
59 #include <vm/vm_extern.h>
60 #include <vm/vm_object.h>
61 #include <vm/vm_pager.h>
62 
63 #include <machine/reg.h>
64 
65 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
66 
67 static register_t *exec_copyout_strings __P((struct image_params *));
68 
69 /* XXX This should be vm_size_t. */
70 static u_long ps_strings = PS_STRINGS;
71 SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
72 
73 /* XXX This should be vm_size_t. */
74 static u_long usrstack = USRSTACK;
75 SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
76 
77 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
78 SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
79     &ps_arg_cache_limit, 0, "");
80 
81 int ps_argsopen = 1;
82 SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
83 
84 /*
85  * Each of the items is a pointer to a `const struct execsw', hence the
86  * double pointer here.
87  */
88 static const struct execsw **execsw;
89 
90 #ifndef _SYS_SYSPROTO_H_
91 struct execve_args {
92         char    *fname;
93         char    **argv;
94         char    **envv;
95 };
96 #endif
97 
98 /*
99  * execve() system call.
100  */
101 int
102 execve(p, uap)
103 	struct proc *p;
104 	register struct execve_args *uap;
105 {
106 	struct nameidata nd, *ndp;
107 	register_t *stack_base;
108 	int error, len, i;
109 	struct image_params image_params, *imgp;
110 	struct vattr attr;
111 	int (*img_first) __P((struct image_params *));
112 
113 	imgp = &image_params;
114 
115 	/*
116 	 * Initialize part of the common data
117 	 */
118 	imgp->proc = p;
119 	imgp->uap = uap;
120 	imgp->attr = &attr;
121 	imgp->argc = imgp->envc = 0;
122 	imgp->argv0 = NULL;
123 	imgp->entry_addr = 0;
124 	imgp->vmspace_destroyed = 0;
125 	imgp->interpreted = 0;
126 	imgp->interpreter_name[0] = '\0';
127 	imgp->auxargs = NULL;
128 	imgp->vp = NULL;
129 	imgp->firstpage = NULL;
130 	imgp->ps_strings = 0;
131 	imgp->auxarg_size = 0;
132 
133 	/*
134 	 * Allocate temporary demand zeroed space for argument and
135 	 *	environment strings
136 	 */
137 	imgp->stringbase = (char *)kmem_alloc_wait(exec_map, ARG_MAX + PAGE_SIZE);
138 	if (imgp->stringbase == NULL) {
139 		error = ENOMEM;
140 		goto exec_fail;
141 	}
142 	imgp->stringp = imgp->stringbase;
143 	imgp->stringspace = ARG_MAX;
144 	imgp->image_header = imgp->stringbase + ARG_MAX;
145 
146 	/*
147 	 * Translate the file name. namei() returns a vnode pointer
148 	 *	in ni_vp amoung other things.
149 	 */
150 	ndp = &nd;
151 	NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
152 	    UIO_USERSPACE, uap->fname, p);
153 
154 interpret:
155 
156 	error = namei(ndp);
157 	if (error) {
158 		kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
159 			ARG_MAX + PAGE_SIZE);
160 		goto exec_fail;
161 	}
162 
163 	imgp->vp = ndp->ni_vp;
164 	imgp->fname = uap->fname;
165 
166 	/*
167 	 * Check file permissions (also 'opens' file)
168 	 */
169 	error = exec_check_permissions(imgp);
170 	if (error) {
171 		VOP_UNLOCK(imgp->vp, 0, p);
172 		goto exec_fail_dealloc;
173 	}
174 
175 	error = exec_map_first_page(imgp);
176 	VOP_UNLOCK(imgp->vp, 0, p);
177 	if (error)
178 		goto exec_fail_dealloc;
179 
180 	/*
181 	 *	If the current process has a special image activator it
182 	 *	wants to try first, call it.   For example, emulating shell
183 	 *	scripts differently.
184 	 */
185 	error = -1;
186 	if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
187 		error = img_first(imgp);
188 
189 	/*
190 	 *	Loop through the list of image activators, calling each one.
191 	 *	An activator returns -1 if there is no match, 0 on success,
192 	 *	and an error otherwise.
193 	 */
194 	for (i = 0; error == -1 && execsw[i]; ++i) {
195 		if (execsw[i]->ex_imgact == NULL ||
196 		    execsw[i]->ex_imgact == img_first) {
197 			continue;
198 		}
199 		error = (*execsw[i]->ex_imgact)(imgp);
200 	}
201 
202 	if (error) {
203 		if (error == -1)
204 			error = ENOEXEC;
205 		goto exec_fail_dealloc;
206 	}
207 
208 	/*
209 	 * Special interpreter operation, cleanup and loop up to try to
210 	 * activate the interpreter.
211 	 */
212 	if (imgp->interpreted) {
213 		exec_unmap_first_page(imgp);
214 		/* free name buffer and old vnode */
215 		NDFREE(ndp, NDF_ONLY_PNBUF);
216 		vrele(ndp->ni_vp);
217 		/* set new name to that of the interpreter */
218 		NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
219 		    UIO_SYSSPACE, imgp->interpreter_name, p);
220 		goto interpret;
221 	}
222 
223 	/*
224 	 * Copy out strings (args and env) and initialize stack base
225 	 */
226 	stack_base = exec_copyout_strings(imgp);
227 	p->p_vmspace->vm_minsaddr = (char *)stack_base;
228 
229 	/*
230 	 * If custom stack fixup routine present for this process
231 	 * let it do the stack setup.
232 	 * Else stuff argument count as first item on stack
233 	 */
234 	if (p->p_sysent->sv_fixup)
235 		(*p->p_sysent->sv_fixup)(&stack_base, imgp);
236 	else
237 		suword(--stack_base, imgp->argc);
238 
239 	/*
240 	 * For security and other reasons, the file descriptor table cannot
241 	 * be shared after an exec.
242 	 */
243 	if (p->p_fd->fd_refcnt > 1) {
244 		struct filedesc *tmp;
245 
246 		tmp = fdcopy(p);
247 		fdfree(p);
248 		p->p_fd = tmp;
249 	}
250 
251 	/* Stop profiling */
252 	stopprofclock(p);
253 
254 	/* close files on exec */
255 	fdcloseexec(p);
256 
257 	/* reset caught signals */
258 	execsigs(p);
259 
260 	/* name this process - nameiexec(p, ndp) */
261 	len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
262 	bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
263 	p->p_comm[len] = 0;
264 
265 	/*
266 	 * mark as execed, wakeup the process that vforked (if any) and tell
267 	 * it that it now has its own resources back
268 	 */
269 	p->p_flag |= P_EXEC;
270 	if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
271 		p->p_flag &= ~P_PPWAIT;
272 		wakeup((caddr_t)p->p_pptr);
273 	}
274 
275 	/*
276 	 * Implement image setuid/setgid.
277 	 *
278 	 * Don't honor setuid/setgid if the filesystem prohibits it or if
279 	 * the process is being traced.
280 	 */
281 	if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
282 	     ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
283 	    (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
284 	    (p->p_flag & P_TRACED) == 0) {
285 		/*
286 		 * Turn off syscall tracing for set-id programs, except for
287 		 * root.
288 		 */
289 		if (p->p_tracep && suser(p)) {
290 			p->p_traceflag = 0;
291 			vrele(p->p_tracep);
292 			p->p_tracep = NULL;
293 		}
294 		/*
295 		 * Set the new credentials.
296 		 */
297 		p->p_ucred = crcopy(p->p_ucred);
298 		if (attr.va_mode & VSUID)
299 			change_euid(p, attr.va_uid);
300 		if (attr.va_mode & VSGID)
301 			p->p_ucred->cr_gid = attr.va_gid;
302 		setsugid(p);
303 		setugidsafety(p);
304 	} else {
305 		if (p->p_ucred->cr_uid == p->p_cred->p_ruid &&
306 		    p->p_ucred->cr_gid == p->p_cred->p_rgid)
307 			p->p_flag &= ~P_SUGID;
308 	}
309 
310 	/*
311 	 * Implement correct POSIX saved-id behavior.
312 	 */
313 	p->p_cred->p_svuid = p->p_ucred->cr_uid;
314 	p->p_cred->p_svgid = p->p_ucred->cr_gid;
315 
316 	/*
317 	 * Store the vp for use in procfs
318 	 */
319 	if (p->p_textvp)		/* release old reference */
320 		vrele(p->p_textvp);
321 	VREF(ndp->ni_vp);
322 	p->p_textvp = ndp->ni_vp;
323 
324 	/*
325 	 * notify others that we exec'd
326 	 */
327 	KNOTE(&p->p_klist, NOTE_EXEC);
328 
329 	/*
330 	 * If tracing the process, trap to debugger so breakpoints
331 	 * 	can be set before the program executes.
332 	 */
333 	STOPEVENT(p, S_EXEC, 0);
334 
335 	if (p->p_flag & P_TRACED)
336 		psignal(p, SIGTRAP);
337 
338 	/* clear "fork but no exec" flag, as we _are_ execing */
339 	p->p_acflag &= ~AFORK;
340 
341 	/* Set values passed into the program in registers. */
342 	setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
343 	    imgp->ps_strings);
344 
345 	/* Free any previous argument cache */
346 	if (p->p_args && --p->p_args->ar_ref == 0)
347 		FREE(p->p_args, M_PARGS);
348 	p->p_args = NULL;
349 
350 	/* Cache arguments if they fit inside our allowance */
351 	i = imgp->endargs - imgp->stringbase;
352 	if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
353 		MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
354 		    M_PARGS, M_WAITOK);
355 		p->p_args->ar_ref = 1;
356 		p->p_args->ar_length = i;
357 		bcopy(imgp->stringbase, p->p_args->ar_args, i);
358 	}
359 
360 exec_fail_dealloc:
361 
362 	/*
363 	 * free various allocated resources
364 	 */
365 	if (imgp->firstpage)
366 		exec_unmap_first_page(imgp);
367 
368 	if (imgp->stringbase != NULL)
369 		kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
370 			ARG_MAX + PAGE_SIZE);
371 
372 	if (imgp->vp) {
373 		NDFREE(ndp, NDF_ONLY_PNBUF);
374 		vrele(imgp->vp);
375 	}
376 
377 	if (error == 0)
378 		return (0);
379 
380 exec_fail:
381 	if (imgp->vmspace_destroyed) {
382 		/* sorry, no more process anymore. exit gracefully */
383 		exit1(p, W_EXITCODE(0, SIGABRT));
384 		/* NOT REACHED */
385 		return(0);
386 	} else {
387 		return(error);
388 	}
389 }
390 
391 int
392 exec_map_first_page(imgp)
393 	struct image_params *imgp;
394 {
395 	int s, rv, i;
396 	int initial_pagein;
397 	vm_page_t ma[VM_INITIAL_PAGEIN];
398 	vm_object_t object;
399 
400 
401 	if (imgp->firstpage) {
402 		exec_unmap_first_page(imgp);
403 	}
404 
405 	VOP_GETVOBJECT(imgp->vp, &object);
406 	s = splvm();
407 
408 	ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
409 
410 	if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
411 		initial_pagein = VM_INITIAL_PAGEIN;
412 		if (initial_pagein > object->size)
413 			initial_pagein = object->size;
414 		for (i = 1; i < initial_pagein; i++) {
415 			if ((ma[i] = vm_page_lookup(object, i)) != NULL) {
416 				if ((ma[i]->flags & PG_BUSY) || ma[i]->busy)
417 					break;
418 				if (ma[i]->valid)
419 					break;
420 				vm_page_busy(ma[i]);
421 			} else {
422 				ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
423 				if (ma[i] == NULL)
424 					break;
425 			}
426 		}
427 		initial_pagein = i;
428 
429 		rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
430 		ma[0] = vm_page_lookup(object, 0);
431 
432 		if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) {
433 			if (ma[0]) {
434 				vm_page_protect(ma[0], VM_PROT_NONE);
435 				vm_page_free(ma[0]);
436 			}
437 			splx(s);
438 			return EIO;
439 		}
440 	}
441 
442 	vm_page_wire(ma[0]);
443 	vm_page_wakeup(ma[0]);
444 	splx(s);
445 
446 	pmap_kenter((vm_offset_t) imgp->image_header, VM_PAGE_TO_PHYS(ma[0]));
447 	imgp->firstpage = ma[0];
448 
449 	return 0;
450 }
451 
452 void
453 exec_unmap_first_page(imgp)
454 	struct image_params *imgp;
455 {
456 	if (imgp->firstpage) {
457 		pmap_kremove((vm_offset_t) imgp->image_header);
458 		vm_page_unwire(imgp->firstpage, 1);
459 		imgp->firstpage = NULL;
460 	}
461 }
462 
463 /*
464  * Destroy old address space, and allocate a new stack
465  *	The new stack is only SGROWSIZ large because it is grown
466  *	automatically in trap.c.
467  */
468 int
469 exec_new_vmspace(imgp)
470 	struct image_params *imgp;
471 {
472 	int error;
473 	struct vmspace *vmspace = imgp->proc->p_vmspace;
474 	caddr_t	stack_addr = (caddr_t) (USRSTACK - MAXSSIZ);
475 	vm_map_t map = &vmspace->vm_map;
476 
477 	imgp->vmspace_destroyed = 1;
478 
479 	/*
480 	 * Blow away entire process VM, if address space not shared,
481 	 * otherwise, create a new VM space so that other threads are
482 	 * not disrupted
483 	 */
484 	if (vmspace->vm_refcnt == 1) {
485 		if (vmspace->vm_shm)
486 			shmexit(imgp->proc);
487 		pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
488 		vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
489 	} else {
490 		vmspace_exec(imgp->proc);
491 		vmspace = imgp->proc->p_vmspace;
492 		map = &vmspace->vm_map;
493 	}
494 
495 	/* Allocate a new stack */
496 	error = vm_map_stack (&vmspace->vm_map, (vm_offset_t)stack_addr,
497 			      (vm_size_t)MAXSSIZ, VM_PROT_ALL, VM_PROT_ALL, 0);
498 	if (error)
499 		return (error);
500 
501 #ifdef __ia64__
502 	{
503 		/*
504 		 * Allocate backing store. We really need something
505 		 * similar to vm_map_stack which can allow the backing
506 		 * store to grow upwards. This will do for now.
507 		 */
508 		vm_offset_t bsaddr;
509 		bsaddr = USRSTACK - 2*MAXSSIZ;
510 		error = vm_map_find(&vmspace->vm_map, 0, 0, &bsaddr,
511 				    4*PAGE_SIZE, 0,
512 				    VM_PROT_ALL, VM_PROT_ALL, 0);
513 		imgp->proc->p_md.md_bspstore = bsaddr;
514 	}
515 #endif
516 
517 	/* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
518 	 * VM_STACK case, but they are still used to monitor the size of the
519 	 * process stack so we can check the stack rlimit.
520 	 */
521 	vmspace->vm_ssize = SGROWSIZ >> PAGE_SHIFT;
522 	vmspace->vm_maxsaddr = (char *)USRSTACK - MAXSSIZ;
523 
524 	return(0);
525 }
526 
527 /*
528  * Copy out argument and environment strings from the old process
529  *	address space into the temporary string buffer.
530  */
531 int
532 exec_extract_strings(imgp)
533 	struct image_params *imgp;
534 {
535 	char	**argv, **envv;
536 	char	*argp, *envp;
537 	int	error;
538 	size_t	length;
539 
540 	/*
541 	 * extract arguments first
542 	 */
543 
544 	argv = imgp->uap->argv;
545 
546 	if (argv) {
547 		argp = (caddr_t) (intptr_t) fuword(argv);
548 		if (argp == (caddr_t) -1)
549 			return (EFAULT);
550 		if (argp)
551 			argv++;
552 		if (imgp->argv0)
553 			argp = imgp->argv0;
554 		if (argp) {
555 			do {
556 				if (argp == (caddr_t) -1)
557 					return (EFAULT);
558 				if ((error = copyinstr(argp, imgp->stringp,
559 				    imgp->stringspace, &length))) {
560 					if (error == ENAMETOOLONG)
561 						return(E2BIG);
562 					return (error);
563 				}
564 				imgp->stringspace -= length;
565 				imgp->stringp += length;
566 				imgp->argc++;
567 			} while ((argp = (caddr_t) (intptr_t) fuword(argv++)));
568 		}
569 	}
570 
571 	imgp->endargs = imgp->stringp;
572 
573 	/*
574 	 * extract environment strings
575 	 */
576 
577 	envv = imgp->uap->envv;
578 
579 	if (envv) {
580 		while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
581 			if (envp == (caddr_t) -1)
582 				return (EFAULT);
583 			if ((error = copyinstr(envp, imgp->stringp,
584 			    imgp->stringspace, &length))) {
585 				if (error == ENAMETOOLONG)
586 					return(E2BIG);
587 				return (error);
588 			}
589 			imgp->stringspace -= length;
590 			imgp->stringp += length;
591 			imgp->envc++;
592 		}
593 	}
594 
595 	return (0);
596 }
597 
598 /*
599  * Copy strings out to the new process address space, constructing
600  *	new arg and env vector tables. Return a pointer to the base
601  *	so that it can be used as the initial stack pointer.
602  */
603 register_t *
604 exec_copyout_strings(imgp)
605 	struct image_params *imgp;
606 {
607 	int argc, envc;
608 	char **vectp;
609 	char *stringp, *destp;
610 	register_t *stack_base;
611 	struct ps_strings *arginfo;
612 	int szsigcode;
613 
614 	/*
615 	 * Calculate string base and vector table pointers.
616 	 * Also deal with signal trampoline code for this exec type.
617 	 */
618 	arginfo = (struct ps_strings *)PS_STRINGS;
619 	szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
620 	destp =	(caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
621 		roundup((ARG_MAX - imgp->stringspace), sizeof(char *));
622 
623 	/*
624 	 * install sigcode
625 	 */
626 	if (szsigcode)
627 		copyout(imgp->proc->p_sysent->sv_sigcode,
628 			((caddr_t)arginfo - szsigcode), szsigcode);
629 
630 	/*
631 	 * If we have a valid auxargs ptr, prepare some room
632 	 * on the stack.
633 	 */
634 	if (imgp->auxargs) {
635 		/*
636 		 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
637 		 * lower compatibility.
638 		 */
639 		imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size
640 			: (AT_COUNT * 2);
641 		/*
642 		 * The '+ 2' is for the null pointers at the end of each of
643 		 * the arg and env vector sets,and imgp->auxarg_size is room
644 		 * for argument of Runtime loader.
645 		 */
646 		vectp = (char **) (destp - (imgp->argc + imgp->envc + 2 +
647 				       imgp->auxarg_size) * sizeof(char *));
648 
649 	} else
650 		/*
651 		 * The '+ 2' is for the null pointers at the end of each of
652 		 * the arg and env vector sets
653 		 */
654 		vectp = (char **)
655 			(destp - (imgp->argc + imgp->envc + 2) * sizeof(char *));
656 
657 	/*
658 	 * vectp also becomes our initial stack base
659 	 */
660 	stack_base = (register_t *)vectp;
661 
662 	stringp = imgp->stringbase;
663 	argc = imgp->argc;
664 	envc = imgp->envc;
665 
666 	/*
667 	 * Copy out strings - arguments and environment.
668 	 */
669 	copyout(stringp, destp, ARG_MAX - imgp->stringspace);
670 
671 	/*
672 	 * Fill in "ps_strings" struct for ps, w, etc.
673 	 */
674 	suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
675 	suword(&arginfo->ps_nargvstr, argc);
676 
677 	/*
678 	 * Fill in argument portion of vector table.
679 	 */
680 	for (; argc > 0; --argc) {
681 		suword(vectp++, (long)(intptr_t)destp);
682 		while (*stringp++ != 0)
683 			destp++;
684 		destp++;
685 	}
686 
687 	/* a null vector table pointer seperates the argp's from the envp's */
688 	suword(vectp++, 0);
689 
690 	suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
691 	suword(&arginfo->ps_nenvstr, envc);
692 
693 	/*
694 	 * Fill in environment portion of vector table.
695 	 */
696 	for (; envc > 0; --envc) {
697 		suword(vectp++, (long)(intptr_t)destp);
698 		while (*stringp++ != 0)
699 			destp++;
700 		destp++;
701 	}
702 
703 	/* end of vector table is a null pointer */
704 	suword(vectp, 0);
705 
706 	return (stack_base);
707 }
708 
709 /*
710  * Check permissions of file to execute.
711  *	Return 0 for success or error code on failure.
712  */
713 int
714 exec_check_permissions(imgp)
715 	struct image_params *imgp;
716 {
717 	struct proc *p = imgp->proc;
718 	struct vnode *vp = imgp->vp;
719 	struct vattr *attr = imgp->attr;
720 	int error;
721 
722 	/* Get file attributes */
723 	error = VOP_GETATTR(vp, attr, p->p_ucred, p);
724 	if (error)
725 		return (error);
726 
727 	/*
728 	 * 1) Check if file execution is disabled for the filesystem that this
729 	 *	file resides on.
730 	 * 2) Insure that at least one execute bit is on - otherwise root
731 	 *	will always succeed, and we don't want to happen unless the
732 	 *	file really is executable.
733 	 * 3) Insure that the file is a regular file.
734 	 */
735 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
736 	    ((attr->va_mode & 0111) == 0) ||
737 	    (attr->va_type != VREG)) {
738 		return (EACCES);
739 	}
740 
741 	/*
742 	 * Zero length files can't be exec'd
743 	 */
744 	if (attr->va_size == 0)
745 		return (ENOEXEC);
746 
747 	/*
748 	 *  Check for execute permission to file based on current credentials.
749 	 */
750 	error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
751 	if (error)
752 		return (error);
753 
754 	/*
755 	 * Check number of open-for-writes on the file and deny execution
756 	 * if there are any.
757 	 */
758 	if (vp->v_writecount)
759 		return (ETXTBSY);
760 
761 	/*
762 	 * Call filesystem specific open routine (which does nothing in the
763 	 * general case).
764 	 */
765 	error = VOP_OPEN(vp, FREAD, p->p_ucred, p);
766 	if (error)
767 		return (error);
768 
769 	return (0);
770 }
771 
772 /*
773  * Exec handler registration
774  */
775 int
776 exec_register(execsw_arg)
777 	const struct execsw *execsw_arg;
778 {
779 	const struct execsw **es, **xs, **newexecsw;
780 	int count = 2;	/* New slot and trailing NULL */
781 
782 	if (execsw)
783 		for (es = execsw; *es; es++)
784 			count++;
785 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
786 	if (newexecsw == NULL)
787 		return ENOMEM;
788 	xs = newexecsw;
789 	if (execsw)
790 		for (es = execsw; *es; es++)
791 			*xs++ = *es;
792 	*xs++ = execsw_arg;
793 	*xs = NULL;
794 	if (execsw)
795 		free(execsw, M_TEMP);
796 	execsw = newexecsw;
797 	return 0;
798 }
799 
800 int
801 exec_unregister(execsw_arg)
802 	const struct execsw *execsw_arg;
803 {
804 	const struct execsw **es, **xs, **newexecsw;
805 	int count = 1;
806 
807 	if (execsw == NULL)
808 		panic("unregister with no handlers left?\n");
809 
810 	for (es = execsw; *es; es++) {
811 		if (*es == execsw_arg)
812 			break;
813 	}
814 	if (*es == NULL)
815 		return ENOENT;
816 	for (es = execsw; *es; es++)
817 		if (*es != execsw_arg)
818 			count++;
819 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
820 	if (newexecsw == NULL)
821 		return ENOMEM;
822 	xs = newexecsw;
823 	for (es = execsw; *es; es++)
824 		if (*es != execsw_arg)
825 			*xs++ = *es;
826 	*xs = NULL;
827 	if (execsw)
828 		free(execsw, M_TEMP);
829 	execsw = newexecsw;
830 	return 0;
831 }
832