xref: /freebsd/sys/kern/kern_exec.c (revision 42c159fe388a3765f69860c84183700af37aca8a)
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/malloc.h>
45 #include <sys/proc.h>
46 #include <sys/pioctl.h>
47 #include <sys/namei.h>
48 #include <sys/sysent.h>
49 #include <sys/shm.h>
50 #include <sys/sysctl.h>
51 #include <sys/user.h>
52 #include <sys/vnode.h>
53 
54 #include <vm/vm.h>
55 #include <vm/vm_param.h>
56 #include <vm/pmap.h>
57 #include <vm/vm_page.h>
58 #include <vm/vm_map.h>
59 #include <vm/vm_kern.h>
60 #include <vm/vm_extern.h>
61 #include <vm/vm_object.h>
62 #include <vm/vm_pager.h>
63 
64 #include <machine/reg.h>
65 
66 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
67 
68 static MALLOC_DEFINE(M_ATEXEC, "atexec", "atexec callback");
69 
70 /*
71  * callout list for things to do at exec time
72  */
73 struct execlist {
74 	execlist_fn function;
75 	TAILQ_ENTRY(execlist) next;
76 };
77 
78 TAILQ_HEAD(exec_list_head, execlist);
79 static struct exec_list_head exec_list = TAILQ_HEAD_INITIALIZER(exec_list);
80 
81 static register_t *exec_copyout_strings(struct image_params *);
82 
83 /* XXX This should be vm_size_t. */
84 static u_long ps_strings = PS_STRINGS;
85 SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
86 
87 /* XXX This should be vm_size_t. */
88 static u_long usrstack = USRSTACK;
89 SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
90 
91 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
92 SYSCTL_ULONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
93     &ps_arg_cache_limit, 0, "");
94 
95 int ps_argsopen = 1;
96 SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
97 
98 /*
99  * Each of the items is a pointer to a `const struct execsw', hence the
100  * double pointer here.
101  */
102 static const struct execsw **execsw;
103 
104 #ifndef _SYS_SYSPROTO_H_
105 struct execve_args {
106         char    *fname;
107         char    **argv;
108         char    **envv;
109 };
110 #endif
111 
112 /*
113  * execve() system call.
114  *
115  * MPSAFE
116  */
117 int
118 execve(td, uap)
119 	struct thread *td;
120 	register struct execve_args *uap;
121 {
122 	struct proc *p = td->td_proc;
123 	struct nameidata nd, *ndp;
124 	struct ucred *newcred, *oldcred;
125 	register_t *stack_base;
126 	int error, len, i;
127 	struct image_params image_params, *imgp;
128 	struct vattr attr;
129 	int (*img_first)(struct image_params *);
130 	struct pargs *pa;
131 
132 	imgp = &image_params;
133 
134 	/*
135 	 * Lock the process and set the P_INEXEC flag to indicate that
136 	 * it should be left alone until we're done here.  This is
137 	 * necessary to avoid race conditions - e.g. in ptrace() -
138 	 * that might allow a local user to illicitly obtain elevated
139 	 * privileges.
140 	 */
141 	mtx_lock(&Giant);
142 	PROC_LOCK(p);
143 	KASSERT((p->p_flag & P_INEXEC) == 0,
144 	    ("%s(): process already has P_INEXEC flag", __func__));
145 	p->p_flag |= P_INEXEC;
146 	PROC_UNLOCK(p);
147 
148 /* XXXKSE */
149 /* !!!!!!!! we need abort all the other threads of this process before we */
150 /* proceed beyond his point! */
151 
152 	/*
153 	 * Initialize part of the common data
154 	 */
155 	imgp->proc = p;
156 	imgp->uap = uap;
157 	imgp->attr = &attr;
158 	imgp->argc = imgp->envc = 0;
159 	imgp->argv0 = NULL;
160 	imgp->entry_addr = 0;
161 	imgp->vmspace_destroyed = 0;
162 	imgp->interpreted = 0;
163 	imgp->interpreter_name[0] = '\0';
164 	imgp->auxargs = NULL;
165 	imgp->vp = NULL;
166 	imgp->firstpage = NULL;
167 	imgp->ps_strings = 0;
168 	imgp->auxarg_size = 0;
169 
170 	/*
171 	 * Allocate temporary demand zeroed space for argument and
172 	 *	environment strings
173 	 */
174 	imgp->stringbase = (char *)kmem_alloc_wait(exec_map, ARG_MAX + PAGE_SIZE);
175 	if (imgp->stringbase == NULL) {
176 		error = ENOMEM;
177 		goto exec_fail;
178 	}
179 	imgp->stringp = imgp->stringbase;
180 	imgp->stringspace = ARG_MAX;
181 	imgp->image_header = imgp->stringbase + ARG_MAX;
182 
183 	/*
184 	 * Translate the file name. namei() returns a vnode pointer
185 	 *	in ni_vp amoung other things.
186 	 */
187 	ndp = &nd;
188 	NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
189 	    UIO_USERSPACE, uap->fname, td);
190 
191 interpret:
192 
193 	error = namei(ndp);
194 	if (error) {
195 		kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
196 			ARG_MAX + PAGE_SIZE);
197 		goto exec_fail;
198 	}
199 
200 	imgp->vp = ndp->ni_vp;
201 	imgp->fname = uap->fname;
202 
203 	/*
204 	 * Check file permissions (also 'opens' file)
205 	 */
206 	error = exec_check_permissions(imgp);
207 	if (error) {
208 		VOP_UNLOCK(imgp->vp, 0, td);
209 		goto exec_fail_dealloc;
210 	}
211 
212 	error = exec_map_first_page(imgp);
213 	VOP_UNLOCK(imgp->vp, 0, td);
214 	if (error)
215 		goto exec_fail_dealloc;
216 
217 	/*
218 	 *	If the current process has a special image activator it
219 	 *	wants to try first, call it.   For example, emulating shell
220 	 *	scripts differently.
221 	 */
222 	error = -1;
223 	if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
224 		error = img_first(imgp);
225 
226 	/*
227 	 *	Loop through the list of image activators, calling each one.
228 	 *	An activator returns -1 if there is no match, 0 on success,
229 	 *	and an error otherwise.
230 	 */
231 	for (i = 0; error == -1 && execsw[i]; ++i) {
232 		if (execsw[i]->ex_imgact == NULL ||
233 		    execsw[i]->ex_imgact == img_first) {
234 			continue;
235 		}
236 		error = (*execsw[i]->ex_imgact)(imgp);
237 	}
238 
239 	if (error) {
240 		if (error == -1)
241 			error = ENOEXEC;
242 		goto exec_fail_dealloc;
243 	}
244 
245 	/*
246 	 * Special interpreter operation, cleanup and loop up to try to
247 	 * activate the interpreter.
248 	 */
249 	if (imgp->interpreted) {
250 		exec_unmap_first_page(imgp);
251 		/* free name buffer and old vnode */
252 		NDFREE(ndp, NDF_ONLY_PNBUF);
253 		vrele(ndp->ni_vp);
254 		/* set new name to that of the interpreter */
255 		NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
256 		    UIO_SYSSPACE, imgp->interpreter_name, td);
257 		goto interpret;
258 	}
259 
260 	/*
261 	 * Copy out strings (args and env) and initialize stack base
262 	 */
263 	stack_base = exec_copyout_strings(imgp);
264 	p->p_vmspace->vm_minsaddr = (char *)stack_base;
265 
266 	/*
267 	 * If custom stack fixup routine present for this process
268 	 * let it do the stack setup.
269 	 * Else stuff argument count as first item on stack
270 	 */
271 	if (p->p_sysent->sv_fixup)
272 		(*p->p_sysent->sv_fixup)(&stack_base, imgp);
273 	else
274 		suword(--stack_base, imgp->argc);
275 
276 	/*
277 	 * For security and other reasons, the file descriptor table cannot
278 	 * be shared after an exec.
279 	 */
280 	FILEDESC_LOCK(p->p_fd);
281 	if (p->p_fd->fd_refcnt > 1) {
282 		struct filedesc *tmp;
283 
284 		tmp = fdcopy(td);
285 		FILEDESC_UNLOCK(p->p_fd);
286 		fdfree(td);
287 		p->p_fd = tmp;
288 	} else
289 		FILEDESC_UNLOCK(p->p_fd);
290 
291 	/*
292 	 * For security and other reasons, signal handlers cannot
293 	 * be shared after an exec. The new process gets a copy of the old
294 	 * handlers. In execsigs(), the new process will have its signals
295 	 * reset.
296 	 */
297 	if (p->p_procsig->ps_refcnt > 1) {
298 		struct procsig *newprocsig;
299 
300 		MALLOC(newprocsig, struct procsig *, sizeof(struct procsig),
301 		       M_SUBPROC, M_WAITOK);
302 		bcopy(p->p_procsig, newprocsig, sizeof(*newprocsig));
303 		p->p_procsig->ps_refcnt--;
304 		p->p_procsig = newprocsig;
305 		p->p_procsig->ps_refcnt = 1;
306 		if (p->p_sigacts == &p->p_uarea->u_sigacts)
307 			panic("shared procsig but private sigacts?");
308 
309 		p->p_uarea->u_sigacts = *p->p_sigacts;
310 		p->p_sigacts = &p->p_uarea->u_sigacts;
311 	}
312 	/* Stop profiling */
313 	stopprofclock(p);
314 
315 	/* close files on exec */
316 	fdcloseexec(td);
317 
318 	/* reset caught signals */
319 	execsigs(p);
320 
321 	/* name this process - nameiexec(p, ndp) */
322 	len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
323 	bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
324 	p->p_comm[len] = 0;
325 
326 	/*
327 	 * mark as execed, wakeup the process that vforked (if any) and tell
328 	 * it that it now has its own resources back
329 	 */
330 	PROC_LOCK(p);
331 	p->p_flag |= P_EXEC;
332 	if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
333 		p->p_flag &= ~P_PPWAIT;
334 		wakeup((caddr_t)p->p_pptr);
335 	}
336 
337 	/*
338 	 * Implement image setuid/setgid.
339 	 *
340 	 * Don't honor setuid/setgid if the filesystem prohibits it or if
341 	 * the process is being traced.
342 	 */
343 	oldcred = p->p_ucred;
344 	newcred = NULL;
345 	if ((((attr.va_mode & VSUID) && oldcred->cr_uid != attr.va_uid) ||
346 	     ((attr.va_mode & VSGID) && oldcred->cr_gid != attr.va_gid)) &&
347 	    (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
348 	    (p->p_flag & P_TRACED) == 0) {
349 		PROC_UNLOCK(p);
350 		/*
351 		 * Turn off syscall tracing for set-id programs, except for
352 		 * root.  Record any set-id flags first to make sure that
353 		 * we do not regain any tracing during a possible block.
354 		 */
355 		setsugid(p);
356 		if (p->p_tracep && suser_xxx(oldcred, NULL, PRISON_ROOT)) {
357 			struct vnode *vtmp;
358 
359 			if ((vtmp = p->p_tracep) != NULL) {
360 				p->p_tracep = NULL;
361 				p->p_traceflag = 0;
362 				vrele(vtmp);
363 			}
364 		}
365 		/*
366 		 * Set the new credentials.
367 		 */
368 		newcred = crdup(oldcred);
369 		if (attr.va_mode & VSUID)
370 			change_euid(newcred, attr.va_uid);
371 		if (attr.va_mode & VSGID)
372 			change_egid(newcred, attr.va_gid);
373 		setugidsafety(td);
374 	} else {
375 		if (oldcred->cr_uid == oldcred->cr_ruid &&
376 		    oldcred->cr_gid == oldcred->cr_rgid)
377 			p->p_flag &= ~P_SUGID;
378 		PROC_UNLOCK(p);
379 	}
380 
381 	/*
382 	 * Implement correct POSIX saved-id behavior.
383 	 *
384 	 * XXX: It's not clear that the existing behavior is
385 	 * POSIX-compliant.  A number of sources indicate that the saved
386 	 * uid/gid should only be updated if the new ruid is not equal to
387 	 * the old ruid, or the new euid is not equal to the old euid and
388 	 * the new euid is not equal to the old ruid.  The FreeBSD code
389 	 * always updates the saved uid/gid.  Also, this code uses the new
390 	 * (replaced) euid and egid as the source, which may or may not be
391 	 * the right ones to use.
392 	 */
393 	if (newcred == NULL) {
394 		if (oldcred->cr_svuid != oldcred->cr_uid ||
395 		    oldcred->cr_svgid != oldcred->cr_gid) {
396 			newcred = crdup(oldcred);
397 			change_svuid(newcred, newcred->cr_uid);
398 			change_svgid(newcred, newcred->cr_gid);
399 		}
400 	} else {
401 		change_svuid(newcred, newcred->cr_uid);
402 		change_svgid(newcred, newcred->cr_gid);
403 	}
404 
405 	if (newcred != NULL) {
406 		PROC_LOCK(p);
407 		p->p_ucred = newcred;
408 		PROC_UNLOCK(p);
409 		crfree(oldcred);
410 	}
411 
412 	/*
413 	 * Store the vp for use in procfs
414 	 */
415 	if (p->p_textvp)		/* release old reference */
416 		vrele(p->p_textvp);
417 	VREF(ndp->ni_vp);
418 	p->p_textvp = ndp->ni_vp;
419 
420 	/*
421 	 * Notify others that we exec'd, and clear the P_INEXEC flag
422 	 * as we're now a bona fide freshly-execed process.
423 	 */
424 	PROC_LOCK(p);
425 	KNOTE(&p->p_klist, NOTE_EXEC);
426 	p->p_flag &= ~P_INEXEC;
427 
428 	/*
429 	 * If tracing the process, trap to debugger so breakpoints
430 	 * can be set before the program executes.
431 	 */
432 	_STOPEVENT(p, S_EXEC, 0);
433 
434 	if (p->p_flag & P_TRACED)
435 		psignal(p, SIGTRAP);
436 
437 	/* clear "fork but no exec" flag, as we _are_ execing */
438 	p->p_acflag &= ~AFORK;
439 
440 	/* Free any previous argument cache */
441 	pa = p->p_args;
442 	p->p_args = NULL;
443 	PROC_UNLOCK(p);
444 	if (pa != NULL && --pa->ar_ref == 0)
445 		FREE(pa, M_PARGS);
446 
447 	/* Set values passed into the program in registers. */
448 	setregs(td, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
449 	    imgp->ps_strings);
450 
451 	/* Cache arguments if they fit inside our allowance */
452 	i = imgp->endargs - imgp->stringbase;
453 	if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
454 		MALLOC(pa, struct pargs *, sizeof(struct pargs) + i,
455 		    M_PARGS, M_WAITOK);
456 		pa->ar_ref = 1;
457 		pa->ar_length = i;
458 		bcopy(imgp->stringbase, pa->ar_args, i);
459 		PROC_LOCK(p);
460 		p->p_args = pa;
461 		PROC_UNLOCK(p);
462 	}
463 
464 exec_fail_dealloc:
465 
466 	/*
467 	 * free various allocated resources
468 	 */
469 	if (imgp->firstpage)
470 		exec_unmap_first_page(imgp);
471 
472 	if (imgp->stringbase != NULL)
473 		kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
474 			ARG_MAX + PAGE_SIZE);
475 
476 	if (imgp->vp) {
477 		NDFREE(ndp, NDF_ONLY_PNBUF);
478 		vrele(imgp->vp);
479 	}
480 
481 	if (error == 0)
482 		goto done2;
483 
484 exec_fail:
485 	/* we're done here, clear P_INEXEC */
486 	PROC_LOCK(p);
487 	p->p_flag &= ~P_INEXEC;
488 	PROC_UNLOCK(p);
489 
490 	if (imgp->vmspace_destroyed) {
491 		/* sorry, no more process anymore. exit gracefully */
492 		exit1(td, W_EXITCODE(0, SIGABRT));
493 		/* NOT REACHED */
494 		error = 0;
495 	}
496 done2:
497 	mtx_unlock(&Giant);
498 	return (error);
499 }
500 
501 int
502 exec_map_first_page(imgp)
503 	struct image_params *imgp;
504 {
505 	int rv, i;
506 	int initial_pagein;
507 	vm_page_t ma[VM_INITIAL_PAGEIN];
508 	vm_object_t object;
509 
510 	GIANT_REQUIRED;
511 
512 	if (imgp->firstpage) {
513 		exec_unmap_first_page(imgp);
514 	}
515 
516 	VOP_GETVOBJECT(imgp->vp, &object);
517 
518 	ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
519 
520 	if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
521 		initial_pagein = VM_INITIAL_PAGEIN;
522 		if (initial_pagein > object->size)
523 			initial_pagein = object->size;
524 		for (i = 1; i < initial_pagein; i++) {
525 			if ((ma[i] = vm_page_lookup(object, i)) != NULL) {
526 				if ((ma[i]->flags & PG_BUSY) || ma[i]->busy)
527 					break;
528 				if (ma[i]->valid)
529 					break;
530 				vm_page_busy(ma[i]);
531 			} else {
532 				ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
533 				if (ma[i] == NULL)
534 					break;
535 			}
536 		}
537 		initial_pagein = i;
538 
539 		rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
540 		ma[0] = vm_page_lookup(object, 0);
541 
542 		if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) {
543 			if (ma[0]) {
544 				vm_page_protect(ma[0], VM_PROT_NONE);
545 				vm_page_free(ma[0]);
546 			}
547 			return EIO;
548 		}
549 	}
550 
551 	vm_page_wire(ma[0]);
552 	vm_page_wakeup(ma[0]);
553 
554 	pmap_qenter((vm_offset_t)imgp->image_header, ma, 1);
555 	imgp->firstpage = ma[0];
556 
557 	return 0;
558 }
559 
560 void
561 exec_unmap_first_page(imgp)
562 	struct image_params *imgp;
563 {
564 	GIANT_REQUIRED;
565 
566 	if (imgp->firstpage) {
567 		pmap_qremove((vm_offset_t)imgp->image_header, 1);
568 		vm_page_unwire(imgp->firstpage, 1);
569 		imgp->firstpage = NULL;
570 	}
571 }
572 
573 /*
574  * Destroy old address space, and allocate a new stack
575  *	The new stack is only SGROWSIZ large because it is grown
576  *	automatically in trap.c.
577  */
578 int
579 exec_new_vmspace(imgp)
580 	struct image_params *imgp;
581 {
582 	int error;
583 	struct execlist *ep;
584 	struct vmspace *vmspace = imgp->proc->p_vmspace;
585 	vm_offset_t stack_addr = USRSTACK - maxssiz;
586 	vm_map_t map = &vmspace->vm_map;
587 
588 	GIANT_REQUIRED;
589 
590 	imgp->vmspace_destroyed = 1;
591 
592 	/*
593 	 * Perform functions registered with at_exec().
594 	 */
595 	TAILQ_FOREACH(ep, &exec_list, next)
596 		(*ep->function)(imgp->proc);
597 
598 	/*
599 	 * Blow away entire process VM, if address space not shared,
600 	 * otherwise, create a new VM space so that other threads are
601 	 * not disrupted
602 	 */
603 	if (vmspace->vm_refcnt == 1) {
604 		if (vmspace->vm_shm)
605 			shmexit(imgp->proc);
606 		pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
607 		vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
608 	} else {
609 		vmspace_exec(imgp->proc);
610 		vmspace = imgp->proc->p_vmspace;
611 		map = &vmspace->vm_map;
612 	}
613 
614 	/* Allocate a new stack */
615 	error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
616 	    VM_PROT_ALL, VM_PROT_ALL, 0);
617 	if (error)
618 		return (error);
619 
620 #ifdef __ia64__
621 	{
622 		/*
623 		 * Allocate backing store. We really need something
624 		 * similar to vm_map_stack which can allow the backing
625 		 * store to grow upwards. This will do for now.
626 		 */
627 		vm_offset_t bsaddr;
628 		bsaddr = USRSTACK - 2*maxssiz;
629 		error = vm_map_find(&vmspace->vm_map, 0, 0, &bsaddr,
630 				    4*PAGE_SIZE, 0,
631 				    VM_PROT_ALL, VM_PROT_ALL, 0);
632 		FIRST_THREAD_IN_PROC(imgp->proc)->td_md.md_bspstore = bsaddr;
633 	}
634 #endif
635 
636 	/* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
637 	 * VM_STACK case, but they are still used to monitor the size of the
638 	 * process stack so we can check the stack rlimit.
639 	 */
640 	vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
641 	vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
642 
643 	return(0);
644 }
645 
646 /*
647  * Copy out argument and environment strings from the old process
648  *	address space into the temporary string buffer.
649  */
650 int
651 exec_extract_strings(imgp)
652 	struct image_params *imgp;
653 {
654 	char	**argv, **envv;
655 	char	*argp, *envp;
656 	int	error;
657 	size_t	length;
658 
659 	/*
660 	 * extract arguments first
661 	 */
662 
663 	argv = imgp->uap->argv;
664 
665 	if (argv) {
666 		argp = (caddr_t) (intptr_t) fuword(argv);
667 		if (argp == (caddr_t) -1)
668 			return (EFAULT);
669 		if (argp)
670 			argv++;
671 		if (imgp->argv0)
672 			argp = imgp->argv0;
673 		if (argp) {
674 			do {
675 				if (argp == (caddr_t) -1)
676 					return (EFAULT);
677 				if ((error = copyinstr(argp, imgp->stringp,
678 				    imgp->stringspace, &length))) {
679 					if (error == ENAMETOOLONG)
680 						return(E2BIG);
681 					return (error);
682 				}
683 				imgp->stringspace -= length;
684 				imgp->stringp += length;
685 				imgp->argc++;
686 			} while ((argp = (caddr_t) (intptr_t) fuword(argv++)));
687 		}
688 	}
689 
690 	imgp->endargs = imgp->stringp;
691 
692 	/*
693 	 * extract environment strings
694 	 */
695 
696 	envv = imgp->uap->envv;
697 
698 	if (envv) {
699 		while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
700 			if (envp == (caddr_t) -1)
701 				return (EFAULT);
702 			if ((error = copyinstr(envp, imgp->stringp,
703 			    imgp->stringspace, &length))) {
704 				if (error == ENAMETOOLONG)
705 					return(E2BIG);
706 				return (error);
707 			}
708 			imgp->stringspace -= length;
709 			imgp->stringp += length;
710 			imgp->envc++;
711 		}
712 	}
713 
714 	return (0);
715 }
716 
717 /*
718  * Copy strings out to the new process address space, constructing
719  *	new arg and env vector tables. Return a pointer to the base
720  *	so that it can be used as the initial stack pointer.
721  */
722 register_t *
723 exec_copyout_strings(imgp)
724 	struct image_params *imgp;
725 {
726 	int argc, envc;
727 	char **vectp;
728 	char *stringp, *destp;
729 	register_t *stack_base;
730 	struct ps_strings *arginfo;
731 	int szsigcode;
732 
733 	/*
734 	 * Calculate string base and vector table pointers.
735 	 * Also deal with signal trampoline code for this exec type.
736 	 */
737 	arginfo = (struct ps_strings *)PS_STRINGS;
738 	szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
739 	destp =	(caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
740 		roundup((ARG_MAX - imgp->stringspace), sizeof(char *));
741 
742 	/*
743 	 * install sigcode
744 	 */
745 	if (szsigcode)
746 		copyout(imgp->proc->p_sysent->sv_sigcode,
747 			((caddr_t)arginfo - szsigcode), szsigcode);
748 
749 	/*
750 	 * If we have a valid auxargs ptr, prepare some room
751 	 * on the stack.
752 	 */
753 	if (imgp->auxargs) {
754 		/*
755 		 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
756 		 * lower compatibility.
757 		 */
758 		imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size
759 			: (AT_COUNT * 2);
760 		/*
761 		 * The '+ 2' is for the null pointers at the end of each of
762 		 * the arg and env vector sets,and imgp->auxarg_size is room
763 		 * for argument of Runtime loader.
764 		 */
765 		vectp = (char **) (destp - (imgp->argc + imgp->envc + 2 +
766 				       imgp->auxarg_size) * sizeof(char *));
767 
768 	} else
769 		/*
770 		 * The '+ 2' is for the null pointers at the end of each of
771 		 * the arg and env vector sets
772 		 */
773 		vectp = (char **)
774 			(destp - (imgp->argc + imgp->envc + 2) * sizeof(char *));
775 
776 	/*
777 	 * vectp also becomes our initial stack base
778 	 */
779 	stack_base = (register_t *)vectp;
780 
781 	stringp = imgp->stringbase;
782 	argc = imgp->argc;
783 	envc = imgp->envc;
784 
785 	/*
786 	 * Copy out strings - arguments and environment.
787 	 */
788 	copyout(stringp, destp, ARG_MAX - imgp->stringspace);
789 
790 	/*
791 	 * Fill in "ps_strings" struct for ps, w, etc.
792 	 */
793 	suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
794 	suword(&arginfo->ps_nargvstr, argc);
795 
796 	/*
797 	 * Fill in argument portion of vector table.
798 	 */
799 	for (; argc > 0; --argc) {
800 		suword(vectp++, (long)(intptr_t)destp);
801 		while (*stringp++ != 0)
802 			destp++;
803 		destp++;
804 	}
805 
806 	/* a null vector table pointer separates the argp's from the envp's */
807 	suword(vectp++, 0);
808 
809 	suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
810 	suword(&arginfo->ps_nenvstr, envc);
811 
812 	/*
813 	 * Fill in environment portion of vector table.
814 	 */
815 	for (; envc > 0; --envc) {
816 		suword(vectp++, (long)(intptr_t)destp);
817 		while (*stringp++ != 0)
818 			destp++;
819 		destp++;
820 	}
821 
822 	/* end of vector table is a null pointer */
823 	suword(vectp, 0);
824 
825 	return (stack_base);
826 }
827 
828 /*
829  * Check permissions of file to execute.
830  *	Called with imgp->vp locked.
831  *	Return 0 for success or error code on failure.
832  */
833 int
834 exec_check_permissions(imgp)
835 	struct image_params *imgp;
836 {
837 	struct vnode *vp = imgp->vp;
838 	struct vattr *attr = imgp->attr;
839 	struct thread *td;
840 	int error;
841 
842 	td = curthread;			/* XXXKSE */
843 	/* Get file attributes */
844 	error = VOP_GETATTR(vp, attr, td->td_ucred, td);
845 	if (error)
846 		return (error);
847 
848 	/*
849 	 * 1) Check if file execution is disabled for the filesystem that this
850 	 *	file resides on.
851 	 * 2) Insure that at least one execute bit is on - otherwise root
852 	 *	will always succeed, and we don't want to happen unless the
853 	 *	file really is executable.
854 	 * 3) Insure that the file is a regular file.
855 	 */
856 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
857 	    ((attr->va_mode & 0111) == 0) ||
858 	    (attr->va_type != VREG))
859 		return (EACCES);
860 
861 	/*
862 	 * Zero length files can't be exec'd
863 	 */
864 	if (attr->va_size == 0)
865 		return (ENOEXEC);
866 
867 	/*
868 	 *  Check for execute permission to file based on current credentials.
869 	 */
870 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
871 	if (error)
872 		return (error);
873 
874 	/*
875 	 * Check number of open-for-writes on the file and deny execution
876 	 * if there are any.
877 	 */
878 	if (vp->v_writecount)
879 		return (ETXTBSY);
880 
881 	/*
882 	 * Call filesystem specific open routine (which does nothing in the
883 	 * general case).
884 	 */
885 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td);
886 	return (error);
887 }
888 
889 /*
890  * Exec handler registration
891  */
892 int
893 exec_register(execsw_arg)
894 	const struct execsw *execsw_arg;
895 {
896 	const struct execsw **es, **xs, **newexecsw;
897 	int count = 2;	/* New slot and trailing NULL */
898 
899 	if (execsw)
900 		for (es = execsw; *es; es++)
901 			count++;
902 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
903 	if (newexecsw == NULL)
904 		return ENOMEM;
905 	xs = newexecsw;
906 	if (execsw)
907 		for (es = execsw; *es; es++)
908 			*xs++ = *es;
909 	*xs++ = execsw_arg;
910 	*xs = NULL;
911 	if (execsw)
912 		free(execsw, M_TEMP);
913 	execsw = newexecsw;
914 	return 0;
915 }
916 
917 int
918 exec_unregister(execsw_arg)
919 	const struct execsw *execsw_arg;
920 {
921 	const struct execsw **es, **xs, **newexecsw;
922 	int count = 1;
923 
924 	if (execsw == NULL)
925 		panic("unregister with no handlers left?\n");
926 
927 	for (es = execsw; *es; es++) {
928 		if (*es == execsw_arg)
929 			break;
930 	}
931 	if (*es == NULL)
932 		return ENOENT;
933 	for (es = execsw; *es; es++)
934 		if (*es != execsw_arg)
935 			count++;
936 	newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
937 	if (newexecsw == NULL)
938 		return ENOMEM;
939 	xs = newexecsw;
940 	for (es = execsw; *es; es++)
941 		if (*es != execsw_arg)
942 			*xs++ = *es;
943 	*xs = NULL;
944 	if (execsw)
945 		free(execsw, M_TEMP);
946 	execsw = newexecsw;
947 	return 0;
948 }
949 
950 int
951 at_exec(function)
952 	execlist_fn function;
953 {
954 	struct execlist *ep;
955 
956 #ifdef INVARIANTS
957 	/* Be noisy if the programmer has lost track of things */
958 	if (rm_at_exec(function))
959 		printf("WARNING: exec callout entry (%p) already present\n",
960 		    function);
961 #endif
962 	ep = malloc(sizeof(*ep), M_ATEXEC, M_NOWAIT);
963 	if (ep == NULL)
964 		return (ENOMEM);
965 	ep->function = function;
966 	TAILQ_INSERT_TAIL(&exec_list, ep, next);
967 	return (0);
968 }
969 
970 /*
971  * Scan the exec callout list for the given item and remove it.
972  * Returns the number of items removed (0 or 1)
973  */
974 int
975 rm_at_exec(function)
976 	execlist_fn function;
977 {
978 	struct execlist *ep;
979 
980 	TAILQ_FOREACH(ep, &exec_list, next) {
981 		if (ep->function == function) {
982 			TAILQ_REMOVE(&exec_list, ep, next);
983 			free(ep, M_ATEXEC);
984 			return(1);
985 		}
986 	}
987 	return (0);
988 }
989 
990