xref: /freebsd/sys/kern/kern_exec.c (revision ce4946daa5ce852d28008dac492029500ab2ee95)
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/lock.h>
32 #include <sys/mutex.h>
33 #include <sys/sysproto.h>
34 #include <sys/signalvar.h>
35 #include <sys/kernel.h>
36 #include <sys/mount.h>
37 #include <sys/filedesc.h>
38 #include <sys/fcntl.h>
39 #include <sys/acct.h>
40 #include <sys/exec.h>
41 #include <sys/imgact.h>
42 #include <sys/imgact_elf.h>
43 #include <sys/wait.h>
44 #include <sys/proc.h>
45 #include <sys/pioctl.h>
46 #include <sys/malloc.h>
47 #include <sys/namei.h>
48 #include <sys/sysent.h>
49 #include <sys/shm.h>
50 #include <sys/sysctl.h>
51 #include <sys/vnode.h>
52 
53 #include <vm/vm.h>
54 #include <vm/vm_param.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 	PROC_LOCK(p);
270 	p->p_flag |= P_EXEC;
271 	if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
272 		p->p_flag &= ~P_PPWAIT;
273 		wakeup((caddr_t)p->p_pptr);
274 	}
275 
276 	/*
277 	 * Implement image setuid/setgid.
278 	 *
279 	 * Don't honor setuid/setgid if the filesystem prohibits it or if
280 	 * the process is being traced.
281 	 */
282 	if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
283 	     ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
284 	    (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
285 	    (p->p_flag & P_TRACED) == 0) {
286 		PROC_UNLOCK(p);
287 		/*
288 		 * Turn off syscall tracing for set-id programs, except for
289 		 * root.
290 		 */
291 		if (p->p_tracep && suser(p)) {
292 			p->p_traceflag = 0;
293 			vrele(p->p_tracep);
294 			p->p_tracep = NULL;
295 		}
296 		/*
297 		 * Set the new credentials.
298 		 */
299 		p->p_ucred = crcopy(p->p_ucred);
300 		if (attr.va_mode & VSUID)
301 			change_euid(p, attr.va_uid);
302 		if (attr.va_mode & VSGID)
303 			p->p_ucred->cr_gid = attr.va_gid;
304 		setsugid(p);
305 		setugidsafety(p);
306 	} else {
307 		if (p->p_ucred->cr_uid == p->p_cred->p_ruid &&
308 		    p->p_ucred->cr_gid == p->p_cred->p_rgid)
309 			p->p_flag &= ~P_SUGID;
310 		PROC_UNLOCK(p);
311 	}
312 
313 	/*
314 	 * Implement correct POSIX saved-id behavior.
315 	 */
316 	p->p_cred->p_svuid = p->p_ucred->cr_uid;
317 	p->p_cred->p_svgid = p->p_ucred->cr_gid;
318 
319 	/*
320 	 * Store the vp for use in procfs
321 	 */
322 	if (p->p_textvp)		/* release old reference */
323 		vrele(p->p_textvp);
324 	VREF(ndp->ni_vp);
325 	p->p_textvp = ndp->ni_vp;
326 
327 	/*
328 	 * notify others that we exec'd
329 	 */
330 	PROC_LOCK(p);
331 	KNOTE(&p->p_klist, NOTE_EXEC);
332 
333 	/*
334 	 * If tracing the process, trap to debugger so breakpoints
335 	 * 	can be set before the program executes.
336 	 */
337 	_STOPEVENT(p, S_EXEC, 0);
338 
339 	if (p->p_flag & P_TRACED)
340 		psignal(p, SIGTRAP);
341 
342 	/* clear "fork but no exec" flag, as we _are_ execing */
343 	p->p_acflag &= ~AFORK;
344 
345 	/* Set values passed into the program in registers. */
346 	setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
347 	    imgp->ps_strings);
348 
349 	/* Free any previous argument cache */
350 	if (p->p_args && --p->p_args->ar_ref == 0)
351 		FREE(p->p_args, M_PARGS);
352 	p->p_args = NULL;
353 
354 	/* Cache arguments if they fit inside our allowance */
355 	i = imgp->endargs - imgp->stringbase;
356 	if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
357 		PROC_UNLOCK(p);
358 		MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
359 		    M_PARGS, M_WAITOK);
360 		KASSERT(p->p_args != NULL, ("malloc of p_args failed"));
361 		PROC_LOCK(p);
362 		p->p_args->ar_ref = 1;
363 		p->p_args->ar_length = i;
364 		bcopy(imgp->stringbase, p->p_args->ar_args, i);
365 	}
366 	PROC_UNLOCK(p);
367 
368 exec_fail_dealloc:
369 
370 	/*
371 	 * free various allocated resources
372 	 */
373 	if (imgp->firstpage)
374 		exec_unmap_first_page(imgp);
375 
376 	if (imgp->stringbase != NULL)
377 		kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
378 			ARG_MAX + PAGE_SIZE);
379 
380 	if (imgp->vp) {
381 		NDFREE(ndp, NDF_ONLY_PNBUF);
382 		vrele(imgp->vp);
383 	}
384 
385 	if (error == 0)
386 		return (0);
387 
388 exec_fail:
389 	if (imgp->vmspace_destroyed) {
390 		/* sorry, no more process anymore. exit gracefully */
391 		exit1(p, W_EXITCODE(0, SIGABRT));
392 		/* NOT REACHED */
393 		return(0);
394 	} else {
395 		return(error);
396 	}
397 }
398 
399 int
400 exec_map_first_page(imgp)
401 	struct image_params *imgp;
402 {
403 	int s, rv, i;
404 	int initial_pagein;
405 	vm_page_t ma[VM_INITIAL_PAGEIN];
406 	vm_object_t object;
407 
408 
409 	if (imgp->firstpage) {
410 		exec_unmap_first_page(imgp);
411 	}
412 
413 	VOP_GETVOBJECT(imgp->vp, &object);
414 	s = splvm();
415 
416 	ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
417 
418 	if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
419 		initial_pagein = VM_INITIAL_PAGEIN;
420 		if (initial_pagein > object->size)
421 			initial_pagein = object->size;
422 		for (i = 1; i < initial_pagein; i++) {
423 			if ((ma[i] = vm_page_lookup(object, i)) != NULL) {
424 				if ((ma[i]->flags & PG_BUSY) || ma[i]->busy)
425 					break;
426 				if (ma[i]->valid)
427 					break;
428 				vm_page_busy(ma[i]);
429 			} else {
430 				ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
431 				if (ma[i] == NULL)
432 					break;
433 			}
434 		}
435 		initial_pagein = i;
436 
437 		rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
438 		ma[0] = vm_page_lookup(object, 0);
439 
440 		if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) {
441 			if (ma[0]) {
442 				vm_page_protect(ma[0], VM_PROT_NONE);
443 				vm_page_free(ma[0]);
444 			}
445 			splx(s);
446 			return EIO;
447 		}
448 	}
449 
450 	vm_page_wire(ma[0]);
451 	vm_page_wakeup(ma[0]);
452 	splx(s);
453 
454 	pmap_kenter((vm_offset_t) imgp->image_header, VM_PAGE_TO_PHYS(ma[0]));
455 	imgp->firstpage = ma[0];
456 
457 	return 0;
458 }
459 
460 void
461 exec_unmap_first_page(imgp)
462 	struct image_params *imgp;
463 {
464 	if (imgp->firstpage) {
465 		pmap_kremove((vm_offset_t) imgp->image_header);
466 		vm_page_unwire(imgp->firstpage, 1);
467 		imgp->firstpage = NULL;
468 	}
469 }
470 
471 /*
472  * Destroy old address space, and allocate a new stack
473  *	The new stack is only SGROWSIZ large because it is grown
474  *	automatically in trap.c.
475  */
476 int
477 exec_new_vmspace(imgp)
478 	struct image_params *imgp;
479 {
480 	int error;
481 	struct vmspace *vmspace = imgp->proc->p_vmspace;
482 	caddr_t	stack_addr = (caddr_t) (USRSTACK - MAXSSIZ);
483 	vm_map_t map = &vmspace->vm_map;
484 
485 	imgp->vmspace_destroyed = 1;
486 
487 	/*
488 	 * Blow away entire process VM, if address space not shared,
489 	 * otherwise, create a new VM space so that other threads are
490 	 * not disrupted
491 	 */
492 	if (vmspace->vm_refcnt == 1) {
493 		if (vmspace->vm_shm)
494 			shmexit(imgp->proc);
495 		pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
496 		vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
497 	} else {
498 		vmspace_exec(imgp->proc);
499 		vmspace = imgp->proc->p_vmspace;
500 		map = &vmspace->vm_map;
501 	}
502 
503 	/* Allocate a new stack */
504 	error = vm_map_stack (&vmspace->vm_map, (vm_offset_t)stack_addr,
505 			      (vm_size_t)MAXSSIZ, VM_PROT_ALL, VM_PROT_ALL, 0);
506 	if (error)
507 		return (error);
508 
509 #ifdef __ia64__
510 	{
511 		/*
512 		 * Allocate backing store. We really need something
513 		 * similar to vm_map_stack which can allow the backing
514 		 * store to grow upwards. This will do for now.
515 		 */
516 		vm_offset_t bsaddr;
517 		bsaddr = USRSTACK - 2*MAXSSIZ;
518 		error = vm_map_find(&vmspace->vm_map, 0, 0, &bsaddr,
519 				    4*PAGE_SIZE, 0,
520 				    VM_PROT_ALL, VM_PROT_ALL, 0);
521 		imgp->proc->p_md.md_bspstore = bsaddr;
522 	}
523 #endif
524 
525 	/* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
526 	 * VM_STACK case, but they are still used to monitor the size of the
527 	 * process stack so we can check the stack rlimit.
528 	 */
529 	vmspace->vm_ssize = SGROWSIZ >> PAGE_SHIFT;
530 	vmspace->vm_maxsaddr = (char *)USRSTACK - MAXSSIZ;
531 
532 	return(0);
533 }
534 
535 /*
536  * Copy out argument and environment strings from the old process
537  *	address space into the temporary string buffer.
538  */
539 int
540 exec_extract_strings(imgp)
541 	struct image_params *imgp;
542 {
543 	char	**argv, **envv;
544 	char	*argp, *envp;
545 	int	error;
546 	size_t	length;
547 
548 	/*
549 	 * extract arguments first
550 	 */
551 
552 	argv = imgp->uap->argv;
553 
554 	if (argv) {
555 		argp = (caddr_t) (intptr_t) fuword(argv);
556 		if (argp == (caddr_t) -1)
557 			return (EFAULT);
558 		if (argp)
559 			argv++;
560 		if (imgp->argv0)
561 			argp = imgp->argv0;
562 		if (argp) {
563 			do {
564 				if (argp == (caddr_t) -1)
565 					return (EFAULT);
566 				if ((error = copyinstr(argp, imgp->stringp,
567 				    imgp->stringspace, &length))) {
568 					if (error == ENAMETOOLONG)
569 						return(E2BIG);
570 					return (error);
571 				}
572 				imgp->stringspace -= length;
573 				imgp->stringp += length;
574 				imgp->argc++;
575 			} while ((argp = (caddr_t) (intptr_t) fuword(argv++)));
576 		}
577 	}
578 
579 	imgp->endargs = imgp->stringp;
580 
581 	/*
582 	 * extract environment strings
583 	 */
584 
585 	envv = imgp->uap->envv;
586 
587 	if (envv) {
588 		while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
589 			if (envp == (caddr_t) -1)
590 				return (EFAULT);
591 			if ((error = copyinstr(envp, imgp->stringp,
592 			    imgp->stringspace, &length))) {
593 				if (error == ENAMETOOLONG)
594 					return(E2BIG);
595 				return (error);
596 			}
597 			imgp->stringspace -= length;
598 			imgp->stringp += length;
599 			imgp->envc++;
600 		}
601 	}
602 
603 	return (0);
604 }
605 
606 /*
607  * Copy strings out to the new process address space, constructing
608  *	new arg and env vector tables. Return a pointer to the base
609  *	so that it can be used as the initial stack pointer.
610  */
611 register_t *
612 exec_copyout_strings(imgp)
613 	struct image_params *imgp;
614 {
615 	int argc, envc;
616 	char **vectp;
617 	char *stringp, *destp;
618 	register_t *stack_base;
619 	struct ps_strings *arginfo;
620 	int szsigcode;
621 
622 	/*
623 	 * Calculate string base and vector table pointers.
624 	 * Also deal with signal trampoline code for this exec type.
625 	 */
626 	arginfo = (struct ps_strings *)PS_STRINGS;
627 	szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
628 	destp =	(caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
629 		roundup((ARG_MAX - imgp->stringspace), sizeof(char *));
630 
631 	/*
632 	 * install sigcode
633 	 */
634 	if (szsigcode)
635 		copyout(imgp->proc->p_sysent->sv_sigcode,
636 			((caddr_t)arginfo - szsigcode), szsigcode);
637 
638 	/*
639 	 * If we have a valid auxargs ptr, prepare some room
640 	 * on the stack.
641 	 */
642 	if (imgp->auxargs) {
643 		/*
644 		 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
645 		 * lower compatibility.
646 		 */
647 		imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size
648 			: (AT_COUNT * 2);
649 		/*
650 		 * The '+ 2' is for the null pointers at the end of each of
651 		 * the arg and env vector sets,and imgp->auxarg_size is room
652 		 * for argument of Runtime loader.
653 		 */
654 		vectp = (char **) (destp - (imgp->argc + imgp->envc + 2 +
655 				       imgp->auxarg_size) * sizeof(char *));
656 
657 	} else
658 		/*
659 		 * The '+ 2' is for the null pointers at the end of each of
660 		 * the arg and env vector sets
661 		 */
662 		vectp = (char **)
663 			(destp - (imgp->argc + imgp->envc + 2) * sizeof(char *));
664 
665 	/*
666 	 * vectp also becomes our initial stack base
667 	 */
668 	stack_base = (register_t *)vectp;
669 
670 	stringp = imgp->stringbase;
671 	argc = imgp->argc;
672 	envc = imgp->envc;
673 
674 	/*
675 	 * Copy out strings - arguments and environment.
676 	 */
677 	copyout(stringp, destp, ARG_MAX - imgp->stringspace);
678 
679 	/*
680 	 * Fill in "ps_strings" struct for ps, w, etc.
681 	 */
682 	suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
683 	suword(&arginfo->ps_nargvstr, argc);
684 
685 	/*
686 	 * Fill in argument portion of vector table.
687 	 */
688 	for (; argc > 0; --argc) {
689 		suword(vectp++, (long)(intptr_t)destp);
690 		while (*stringp++ != 0)
691 			destp++;
692 		destp++;
693 	}
694 
695 	/* a null vector table pointer separates the argp's from the envp's */
696 	suword(vectp++, 0);
697 
698 	suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
699 	suword(&arginfo->ps_nenvstr, envc);
700 
701 	/*
702 	 * Fill in environment portion of vector table.
703 	 */
704 	for (; envc > 0; --envc) {
705 		suword(vectp++, (long)(intptr_t)destp);
706 		while (*stringp++ != 0)
707 			destp++;
708 		destp++;
709 	}
710 
711 	/* end of vector table is a null pointer */
712 	suword(vectp, 0);
713 
714 	return (stack_base);
715 }
716 
717 /*
718  * Check permissions of file to execute.
719  *	Called with imgp->vp locked.
720  *	Return 0 for success or error code on failure.
721  */
722 int
723 exec_check_permissions(imgp)
724 	struct image_params *imgp;
725 {
726 	struct proc *p = imgp->proc;
727 	struct vnode *vp = imgp->vp;
728 	struct vattr *attr = imgp->attr;
729 	int error;
730 
731 	/* Get file attributes */
732 	error = VOP_GETATTR(vp, attr, p->p_ucred, p);
733 	if (error)
734 		return (error);
735 
736 	/*
737 	 * 1) Check if file execution is disabled for the filesystem that this
738 	 *	file resides on.
739 	 * 2) Insure that at least one execute bit is on - otherwise root
740 	 *	will always succeed, and we don't want to happen unless the
741 	 *	file really is executable.
742 	 * 3) Insure that the file is a regular file.
743 	 */
744 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
745 	    ((attr->va_mode & 0111) == 0) ||
746 	    (attr->va_type != VREG)) {
747 		return (EACCES);
748 	}
749 
750 	/*
751 	 * Zero length files can't be exec'd
752 	 */
753 	if (attr->va_size == 0)
754 		return (ENOEXEC);
755 
756 	/*
757 	 *  Check for execute permission to file based on current credentials.
758 	 */
759 	error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
760 	if (error)
761 		return (error);
762 
763 	/*
764 	 * Check number of open-for-writes on the file and deny execution
765 	 * if there are any.
766 	 */
767 	if (vp->v_writecount)
768 		return (ETXTBSY);
769 
770 	/*
771 	 * Call filesystem specific open routine (which does nothing in the
772 	 * general case).
773 	 */
774 	error = VOP_OPEN(vp, FREAD, p->p_ucred, p);
775 	if (error)
776 		return (error);
777 
778 	return (0);
779 }
780 
781 /*
782  * Exec handler registration
783  */
784 int
785 exec_register(execsw_arg)
786 	const struct execsw *execsw_arg;
787 {
788 	const struct execsw **es, **xs, **newexecsw;
789 	int count = 2;	/* New slot and trailing NULL */
790 
791 	if (execsw)
792 		for (es = execsw; *es; es++)
793 			count++;
794 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
795 	if (newexecsw == NULL)
796 		return ENOMEM;
797 	xs = newexecsw;
798 	if (execsw)
799 		for (es = execsw; *es; es++)
800 			*xs++ = *es;
801 	*xs++ = execsw_arg;
802 	*xs = NULL;
803 	if (execsw)
804 		free(execsw, M_TEMP);
805 	execsw = newexecsw;
806 	return 0;
807 }
808 
809 int
810 exec_unregister(execsw_arg)
811 	const struct execsw *execsw_arg;
812 {
813 	const struct execsw **es, **xs, **newexecsw;
814 	int count = 1;
815 
816 	if (execsw == NULL)
817 		panic("unregister with no handlers left?\n");
818 
819 	for (es = execsw; *es; es++) {
820 		if (*es == execsw_arg)
821 			break;
822 	}
823 	if (*es == NULL)
824 		return ENOENT;
825 	for (es = execsw; *es; es++)
826 		if (*es != execsw_arg)
827 			count++;
828 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
829 	if (newexecsw == NULL)
830 		return ENOMEM;
831 	xs = newexecsw;
832 	for (es = execsw; *es; es++)
833 		if (*es != execsw_arg)
834 			*xs++ = *es;
835 	*xs = NULL;
836 	if (execsw)
837 		free(execsw, M_TEMP);
838 	execsw = newexecsw;
839 	return 0;
840 }
841