xref: /freebsd/sys/kern/kern_exec.c (revision b601c69bdbe8755d26570261d7fd4c02ee4eff74)
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/filedesc.h>
36 #include <sys/fcntl.h>
37 #include <sys/acct.h>
38 #include <sys/exec.h>
39 #include <sys/imgact.h>
40 #include <sys/imgact_elf.h>
41 #include <sys/wait.h>
42 #include <sys/proc.h>
43 #include <sys/pioctl.h>
44 #include <sys/malloc.h>
45 #include <sys/namei.h>
46 #include <sys/sysent.h>
47 #include <sys/shm.h>
48 #include <sys/sysctl.h>
49 #include <sys/vnode.h>
50 
51 #include <vm/vm.h>
52 #include <vm/vm_param.h>
53 #include <sys/lock.h>
54 #include <vm/pmap.h>
55 #include <vm/vm_page.h>
56 #include <vm/vm_map.h>
57 #include <vm/vm_kern.h>
58 #include <vm/vm_extern.h>
59 #include <vm/vm_object.h>
60 #include <vm/vm_pager.h>
61 
62 #include <machine/reg.h>
63 
64 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
65 
66 static register_t *exec_copyout_strings __P((struct image_params *));
67 
68 /* XXX This should be vm_size_t. */
69 static u_long ps_strings = PS_STRINGS;
70 SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
71 
72 /* XXX This should be vm_size_t. */
73 static u_long usrstack = USRSTACK;
74 SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
75 
76 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
77 SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
78     &ps_arg_cache_limit, 0, "");
79 
80 int ps_argsopen = 1;
81 SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
82 
83 /*
84  * Each of the items is a pointer to a `const struct execsw', hence the
85  * double pointer here.
86  */
87 static const struct execsw **execsw;
88 
89 #ifndef _SYS_SYSPROTO_H_
90 struct execve_args {
91         char    *fname;
92         char    **argv;
93         char    **envv;
94 };
95 #endif
96 
97 /*
98  * execve() system call.
99  */
100 int
101 execve(p, uap)
102 	struct proc *p;
103 	register struct execve_args *uap;
104 {
105 	struct nameidata nd, *ndp;
106 	register_t *stack_base;
107 	int error, len, i;
108 	struct image_params image_params, *imgp;
109 	struct vattr attr;
110 	int (*img_first) __P((struct image_params *));
111 
112 	imgp = &image_params;
113 
114 	/*
115 	 * Initialize part of the common data
116 	 */
117 	imgp->proc = p;
118 	imgp->uap = uap;
119 	imgp->attr = &attr;
120 	imgp->argc = imgp->envc = 0;
121 	imgp->argv0 = NULL;
122 	imgp->entry_addr = 0;
123 	imgp->vmspace_destroyed = 0;
124 	imgp->interpreted = 0;
125 	imgp->interpreter_name[0] = '\0';
126 	imgp->auxargs = NULL;
127 	imgp->vp = NULL;
128 	imgp->firstpage = NULL;
129 	imgp->ps_strings = 0;
130 
131 	/*
132 	 * Allocate temporary demand zeroed space for argument and
133 	 *	environment strings
134 	 */
135 	imgp->stringbase = (char *)kmem_alloc_wait(exec_map, ARG_MAX + PAGE_SIZE);
136 	if (imgp->stringbase == NULL) {
137 		error = ENOMEM;
138 		goto exec_fail;
139 	}
140 	imgp->stringp = imgp->stringbase;
141 	imgp->stringspace = ARG_MAX;
142 	imgp->image_header = imgp->stringbase + ARG_MAX;
143 
144 	/*
145 	 * Translate the file name. namei() returns a vnode pointer
146 	 *	in ni_vp amoung other things.
147 	 */
148 	ndp = &nd;
149 	NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
150 	    UIO_USERSPACE, uap->fname, p);
151 
152 interpret:
153 
154 	error = namei(ndp);
155 	if (error) {
156 		kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
157 			ARG_MAX + PAGE_SIZE);
158 		goto exec_fail;
159 	}
160 
161 	imgp->vp = ndp->ni_vp;
162 	imgp->fname = uap->fname;
163 
164 	/*
165 	 * Check file permissions (also 'opens' file)
166 	 */
167 	error = exec_check_permissions(imgp);
168 	if (error) {
169 		VOP_UNLOCK(imgp->vp, 0, p);
170 		goto exec_fail_dealloc;
171 	}
172 
173 	error = exec_map_first_page(imgp);
174 	VOP_UNLOCK(imgp->vp, 0, p);
175 	if (error)
176 		goto exec_fail_dealloc;
177 
178 	/*
179 	 *	If the current process has a special image activator it
180 	 *	wants to try first, call it.   For example, emulating shell
181 	 *	scripts differently.
182 	 */
183 	error = -1;
184 	if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
185 		error = img_first(imgp);
186 
187 	/*
188 	 *	Loop through the list of image activators, calling each one.
189 	 *	An activator returns -1 if there is no match, 0 on success,
190 	 *	and an error otherwise.
191 	 */
192 	for (i = 0; error == -1 && execsw[i]; ++i) {
193 		if (execsw[i]->ex_imgact == NULL ||
194 		    execsw[i]->ex_imgact == img_first) {
195 			continue;
196 		}
197 		error = (*execsw[i]->ex_imgact)(imgp);
198 	}
199 
200 	if (error) {
201 		if (error == -1)
202 			error = ENOEXEC;
203 		goto exec_fail_dealloc;
204 	}
205 
206 	/*
207 	 * Special interpreter operation, cleanup and loop up to try to
208 	 * activate the interpreter.
209 	 */
210 	if (imgp->interpreted) {
211 		exec_unmap_first_page(imgp);
212 		/* free name buffer and old vnode */
213 		NDFREE(ndp, NDF_ONLY_PNBUF);
214 		vrele(ndp->ni_vp);
215 		/* set new name to that of the interpreter */
216 		NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
217 		    UIO_SYSSPACE, imgp->interpreter_name, p);
218 		goto interpret;
219 	}
220 
221 	/*
222 	 * Copy out strings (args and env) and initialize stack base
223 	 */
224 	stack_base = exec_copyout_strings(imgp);
225 	p->p_vmspace->vm_minsaddr = (char *)stack_base;
226 
227 	/*
228 	 * If custom stack fixup routine present for this process
229 	 * let it do the stack setup.
230 	 * Else stuff argument count as first item on stack
231 	 */
232 	if (p->p_sysent->sv_fixup)
233 		(*p->p_sysent->sv_fixup)(&stack_base, imgp);
234 	else
235 		suword(--stack_base, imgp->argc);
236 
237 	/*
238 	 * For security and other reasons, the file descriptor table cannot
239 	 * be shared after an exec.
240 	 */
241 	if (p->p_fd->fd_refcnt > 1) {
242 		struct filedesc *tmp;
243 
244 		tmp = fdcopy(p);
245 		fdfree(p);
246 		p->p_fd = tmp;
247 	}
248 
249 	/* Stop profiling */
250 	stopprofclock(p);
251 
252 	/* close files on exec */
253 	fdcloseexec(p);
254 
255 	/* reset caught signals */
256 	execsigs(p);
257 
258 	/* name this process - nameiexec(p, ndp) */
259 	len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
260 	bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
261 	p->p_comm[len] = 0;
262 
263 	/*
264 	 * mark as execed, wakeup the process that vforked (if any) and tell
265 	 * it that it now has its own resources back
266 	 */
267 	p->p_flag |= P_EXEC;
268 	if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
269 		p->p_flag &= ~P_PPWAIT;
270 		wakeup((caddr_t)p->p_pptr);
271 	}
272 
273 	/*
274 	 * Implement image setuid/setgid.
275 	 *
276 	 * Don't honor setuid/setgid if the filesystem prohibits it or if
277 	 * the process is being traced.
278 	 */
279 	if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
280 	     ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
281 	    (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
282 	    (p->p_flag & P_TRACED) == 0) {
283 		/*
284 		 * Turn off syscall tracing for set-id programs, except for
285 		 * root.
286 		 */
287 		if (p->p_tracep && suser(p)) {
288 			p->p_traceflag = 0;
289 			vrele(p->p_tracep);
290 			p->p_tracep = NULL;
291 		}
292 		/*
293 		 * Set the new credentials.
294 		 */
295 		p->p_ucred = crcopy(p->p_ucred);
296 		if (attr.va_mode & VSUID)
297 			p->p_ucred->cr_uid = attr.va_uid;
298 		if (attr.va_mode & VSGID)
299 			p->p_ucred->cr_gid = attr.va_gid;
300 		setsugid(p);
301 		setugidsafety(p);
302 	} else {
303 		if (p->p_ucred->cr_uid == p->p_cred->p_ruid &&
304 		    p->p_ucred->cr_gid == p->p_cred->p_rgid)
305 			p->p_flag &= ~P_SUGID;
306 	}
307 
308 	/*
309 	 * Implement correct POSIX saved-id behavior.
310 	 */
311 	p->p_cred->p_svuid = p->p_ucred->cr_uid;
312 	p->p_cred->p_svgid = p->p_ucred->cr_gid;
313 
314 	/*
315 	 * Store the vp for use in procfs
316 	 */
317 	if (p->p_textvp)		/* release old reference */
318 		vrele(p->p_textvp);
319 	VREF(ndp->ni_vp);
320 	p->p_textvp = ndp->ni_vp;
321 
322 	/*
323 	 * notify others that we exec'd
324 	 */
325 	KNOTE(&p->p_klist, NOTE_EXEC);
326 
327 	/*
328 	 * If tracing the process, trap to debugger so breakpoints
329 	 * 	can be set before the program executes.
330 	 */
331 	STOPEVENT(p, S_EXEC, 0);
332 
333 	if (p->p_flag & P_TRACED)
334 		psignal(p, SIGTRAP);
335 
336 	/* clear "fork but no exec" flag, as we _are_ execing */
337 	p->p_acflag &= ~AFORK;
338 
339 	/* Set values passed into the program in registers. */
340 	setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
341 	    imgp->ps_strings);
342 
343 	/* Free any previous argument cache */
344 	if (p->p_args && --p->p_args->ar_ref == 0)
345 		FREE(p->p_args, M_PARGS);
346 	p->p_args = NULL;
347 
348 	/* Cache arguments if they fit inside our allowance */
349 	i = imgp->endargs - imgp->stringbase;
350 	if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
351 		MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
352 		    M_PARGS, M_WAITOK);
353 		p->p_args->ar_ref = 1;
354 		p->p_args->ar_length = i;
355 		bcopy(imgp->stringbase, p->p_args->ar_args, i);
356 	}
357 
358 exec_fail_dealloc:
359 
360 	/*
361 	 * free various allocated resources
362 	 */
363 	if (imgp->firstpage)
364 		exec_unmap_first_page(imgp);
365 
366 	if (imgp->stringbase != NULL)
367 		kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
368 			ARG_MAX + PAGE_SIZE);
369 
370 	if (imgp->vp) {
371 		NDFREE(ndp, NDF_ONLY_PNBUF);
372 		vrele(imgp->vp);
373 	}
374 
375 	if (error == 0)
376 		return (0);
377 
378 exec_fail:
379 	if (imgp->vmspace_destroyed) {
380 		/* sorry, no more process anymore. exit gracefully */
381 		exit1(p, W_EXITCODE(0, SIGABRT));
382 		/* NOT REACHED */
383 		return(0);
384 	} else {
385 		return(error);
386 	}
387 }
388 
389 int
390 exec_map_first_page(imgp)
391 	struct image_params *imgp;
392 {
393 	int s, rv, i;
394 	int initial_pagein;
395 	vm_page_t ma[VM_INITIAL_PAGEIN];
396 	vm_object_t object;
397 
398 
399 	if (imgp->firstpage) {
400 		exec_unmap_first_page(imgp);
401 	}
402 
403 	object = imgp->vp->v_object;
404 	s = splvm();
405 
406 	ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
407 
408 	if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
409 		initial_pagein = VM_INITIAL_PAGEIN;
410 		if (initial_pagein > object->size)
411 			initial_pagein = object->size;
412 		for (i = 1; i < initial_pagein; i++) {
413 			if ((ma[i] = vm_page_lookup(object, i)) != NULL) {
414 				if ((ma[i]->flags & PG_BUSY) || ma[i]->busy)
415 					break;
416 				if (ma[i]->valid)
417 					break;
418 				vm_page_busy(ma[i]);
419 			} else {
420 				ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
421 				if (ma[i] == NULL)
422 					break;
423 			}
424 		}
425 		initial_pagein = i;
426 
427 		rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
428 		ma[0] = vm_page_lookup(object, 0);
429 
430 		if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) {
431 			if (ma[0]) {
432 				vm_page_protect(ma[0], VM_PROT_NONE);
433 				vm_page_free(ma[0]);
434 			}
435 			splx(s);
436 			return EIO;
437 		}
438 	}
439 
440 	vm_page_wire(ma[0]);
441 	vm_page_wakeup(ma[0]);
442 	splx(s);
443 
444 	pmap_kenter((vm_offset_t) imgp->image_header, VM_PAGE_TO_PHYS(ma[0]));
445 	imgp->firstpage = ma[0];
446 
447 	return 0;
448 }
449 
450 void
451 exec_unmap_first_page(imgp)
452 	struct image_params *imgp;
453 {
454 	if (imgp->firstpage) {
455 		pmap_kremove((vm_offset_t) imgp->image_header);
456 		vm_page_unwire(imgp->firstpage, 1);
457 		imgp->firstpage = NULL;
458 	}
459 }
460 
461 /*
462  * Destroy old address space, and allocate a new stack
463  *	The new stack is only SGROWSIZ large because it is grown
464  *	automatically in trap.c.
465  */
466 int
467 exec_new_vmspace(imgp)
468 	struct image_params *imgp;
469 {
470 	int error;
471 	struct vmspace *vmspace = imgp->proc->p_vmspace;
472 	caddr_t	stack_addr = (caddr_t) (USRSTACK - MAXSSIZ);
473 	vm_map_t map = &vmspace->vm_map;
474 
475 	imgp->vmspace_destroyed = 1;
476 
477 	/*
478 	 * Blow away entire process VM, if address space not shared,
479 	 * otherwise, create a new VM space so that other threads are
480 	 * not disrupted
481 	 */
482 	if (vmspace->vm_refcnt == 1) {
483 		if (vmspace->vm_shm)
484 			shmexit(imgp->proc);
485 		pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
486 		vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
487 	} else {
488 		vmspace_exec(imgp->proc);
489 		vmspace = imgp->proc->p_vmspace;
490 		map = &vmspace->vm_map;
491 	}
492 
493 	/* Allocate a new stack */
494 	error = vm_map_stack (&vmspace->vm_map, (vm_offset_t)stack_addr,
495 			      (vm_size_t)MAXSSIZ, VM_PROT_ALL, VM_PROT_ALL, 0);
496 	if (error)
497 		return (error);
498 
499 	/* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
500 	 * VM_STACK case, but they are still used to monitor the size of the
501 	 * process stack so we can check the stack rlimit.
502 	 */
503 	vmspace->vm_ssize = SGROWSIZ >> PAGE_SHIFT;
504 	vmspace->vm_maxsaddr = (char *)USRSTACK - MAXSSIZ;
505 
506 	return(0);
507 }
508 
509 /*
510  * Copy out argument and environment strings from the old process
511  *	address space into the temporary string buffer.
512  */
513 int
514 exec_extract_strings(imgp)
515 	struct image_params *imgp;
516 {
517 	char	**argv, **envv;
518 	char	*argp, *envp;
519 	int	error;
520 	size_t	length;
521 
522 	/*
523 	 * extract arguments first
524 	 */
525 
526 	argv = imgp->uap->argv;
527 
528 	if (argv) {
529 		argp = (caddr_t) (intptr_t) fuword(argv);
530 		if (argp == (caddr_t) -1)
531 			return (EFAULT);
532 		if (argp)
533 			argv++;
534 		if (imgp->argv0)
535 			argp = imgp->argv0;
536 		if (argp) {
537 			do {
538 				if (argp == (caddr_t) -1)
539 					return (EFAULT);
540 				if ((error = copyinstr(argp, imgp->stringp,
541 				    imgp->stringspace, &length))) {
542 					if (error == ENAMETOOLONG)
543 						return(E2BIG);
544 					return (error);
545 				}
546 				imgp->stringspace -= length;
547 				imgp->stringp += length;
548 				imgp->argc++;
549 			} while ((argp = (caddr_t) (intptr_t) fuword(argv++)));
550 		}
551 	}
552 
553 	imgp->endargs = imgp->stringp;
554 
555 	/*
556 	 * extract environment strings
557 	 */
558 
559 	envv = imgp->uap->envv;
560 
561 	if (envv) {
562 		while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
563 			if (envp == (caddr_t) -1)
564 				return (EFAULT);
565 			if ((error = copyinstr(envp, imgp->stringp,
566 			    imgp->stringspace, &length))) {
567 				if (error == ENAMETOOLONG)
568 					return(E2BIG);
569 				return (error);
570 			}
571 			imgp->stringspace -= length;
572 			imgp->stringp += length;
573 			imgp->envc++;
574 		}
575 	}
576 
577 	return (0);
578 }
579 
580 /*
581  * Copy strings out to the new process address space, constructing
582  *	new arg and env vector tables. Return a pointer to the base
583  *	so that it can be used as the initial stack pointer.
584  */
585 register_t *
586 exec_copyout_strings(imgp)
587 	struct image_params *imgp;
588 {
589 	int argc, envc;
590 	char **vectp;
591 	char *stringp, *destp;
592 	register_t *stack_base;
593 	struct ps_strings *arginfo;
594 	int szsigcode;
595 
596 	/*
597 	 * Calculate string base and vector table pointers.
598 	 * Also deal with signal trampoline code for this exec type.
599 	 */
600 	arginfo = (struct ps_strings *)PS_STRINGS;
601 	szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
602 	destp =	(caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
603 		roundup((ARG_MAX - imgp->stringspace), sizeof(char *));
604 
605 	/*
606 	 * install sigcode
607 	 */
608 	if (szsigcode)
609 		copyout(imgp->proc->p_sysent->sv_sigcode,
610 			((caddr_t)arginfo - szsigcode), szsigcode);
611 
612 	/*
613 	 * If we have a valid auxargs ptr, prepare some room
614 	 * on the stack.
615 	 */
616 	if (imgp->auxargs)
617 	/*
618 	 * The '+ 2' is for the null pointers at the end of each of the
619 	 * arg and env vector sets, and 'AT_COUNT*2' is room for the
620 	 * ELF Auxargs data.
621 	 */
622 		vectp = (char **)(destp - (imgp->argc + imgp->envc + 2 +
623 				  AT_COUNT*2) * sizeof(char*));
624 	else
625 	/*
626 	 * The '+ 2' is for the null pointers at the end of each of the
627 	 * arg and env vector sets
628 	 */
629 		vectp = (char **)
630 			(destp - (imgp->argc + imgp->envc + 2) * sizeof(char*));
631 
632 	/*
633 	 * vectp also becomes our initial stack base
634 	 */
635 	stack_base = (register_t *)vectp;
636 
637 	stringp = imgp->stringbase;
638 	argc = imgp->argc;
639 	envc = imgp->envc;
640 
641 	/*
642 	 * Copy out strings - arguments and environment.
643 	 */
644 	copyout(stringp, destp, ARG_MAX - imgp->stringspace);
645 
646 	/*
647 	 * Fill in "ps_strings" struct for ps, w, etc.
648 	 */
649 	suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
650 	suword(&arginfo->ps_nargvstr, argc);
651 
652 	/*
653 	 * Fill in argument portion of vector table.
654 	 */
655 	for (; argc > 0; --argc) {
656 		suword(vectp++, (long)(intptr_t)destp);
657 		while (*stringp++ != 0)
658 			destp++;
659 		destp++;
660 	}
661 
662 	/* a null vector table pointer seperates the argp's from the envp's */
663 	suword(vectp++, 0);
664 
665 	suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
666 	suword(&arginfo->ps_nenvstr, envc);
667 
668 	/*
669 	 * Fill in environment portion of vector table.
670 	 */
671 	for (; envc > 0; --envc) {
672 		suword(vectp++, (long)(intptr_t)destp);
673 		while (*stringp++ != 0)
674 			destp++;
675 		destp++;
676 	}
677 
678 	/* end of vector table is a null pointer */
679 	suword(vectp, 0);
680 
681 	return (stack_base);
682 }
683 
684 /*
685  * Check permissions of file to execute.
686  *	Return 0 for success or error code on failure.
687  */
688 int
689 exec_check_permissions(imgp)
690 	struct image_params *imgp;
691 {
692 	struct proc *p = imgp->proc;
693 	struct vnode *vp = imgp->vp;
694 	struct vattr *attr = imgp->attr;
695 	int error;
696 
697 	/* Get file attributes */
698 	error = VOP_GETATTR(vp, attr, p->p_ucred, p);
699 	if (error)
700 		return (error);
701 
702 	/*
703 	 * 1) Check if file execution is disabled for the filesystem that this
704 	 *	file resides on.
705 	 * 2) Insure that at least one execute bit is on - otherwise root
706 	 *	will always succeed, and we don't want to happen unless the
707 	 *	file really is executable.
708 	 * 3) Insure that the file is a regular file.
709 	 */
710 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
711 	    ((attr->va_mode & 0111) == 0) ||
712 	    (attr->va_type != VREG)) {
713 		return (EACCES);
714 	}
715 
716 	/*
717 	 * Zero length files can't be exec'd
718 	 */
719 	if (attr->va_size == 0)
720 		return (ENOEXEC);
721 
722 	/*
723 	 *  Check for execute permission to file based on current credentials.
724 	 */
725 	error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
726 	if (error)
727 		return (error);
728 
729 	/*
730 	 * Check number of open-for-writes on the file and deny execution
731 	 * if there are any.
732 	 */
733 	if (vp->v_writecount)
734 		return (ETXTBSY);
735 
736 	/*
737 	 * Call filesystem specific open routine (which does nothing in the
738 	 * general case).
739 	 */
740 	error = VOP_OPEN(vp, FREAD, p->p_ucred, p);
741 	if (error)
742 		return (error);
743 
744 	return (0);
745 }
746 
747 /*
748  * Exec handler registration
749  */
750 int
751 exec_register(execsw_arg)
752 	const struct execsw *execsw_arg;
753 {
754 	const struct execsw **es, **xs, **newexecsw;
755 	int count = 2;	/* New slot and trailing NULL */
756 
757 	if (execsw)
758 		for (es = execsw; *es; es++)
759 			count++;
760 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
761 	if (newexecsw == NULL)
762 		return ENOMEM;
763 	xs = newexecsw;
764 	if (execsw)
765 		for (es = execsw; *es; es++)
766 			*xs++ = *es;
767 	*xs++ = execsw_arg;
768 	*xs = NULL;
769 	if (execsw)
770 		free(execsw, M_TEMP);
771 	execsw = newexecsw;
772 	return 0;
773 }
774 
775 int
776 exec_unregister(execsw_arg)
777 	const struct execsw *execsw_arg;
778 {
779 	const struct execsw **es, **xs, **newexecsw;
780 	int count = 1;
781 
782 	if (execsw == NULL)
783 		panic("unregister with no handlers left?\n");
784 
785 	for (es = execsw; *es; es++) {
786 		if (*es == execsw_arg)
787 			break;
788 	}
789 	if (*es == NULL)
790 		return ENOENT;
791 	for (es = execsw; *es; es++)
792 		if (*es != execsw_arg)
793 			count++;
794 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
795 	if (newexecsw == NULL)
796 		return ENOMEM;
797 	xs = newexecsw;
798 	for (es = execsw; *es; es++)
799 		if (*es != execsw_arg)
800 			*xs++ = *es;
801 	*xs = NULL;
802 	if (execsw)
803 		free(execsw, M_TEMP);
804 	execsw = newexecsw;
805 	return 0;
806 }
807