xref: /freebsd/sys/kern/kern_exec.c (revision b3e24f9ce91bb200d621c57bfa92eac473a435f9)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by David Greenman
16  * 4. The name of the developer may not be used to endorse or promote products
17  *    derived from this software without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$Id: kern_exec.c,v 1.26 1995/11/12 06:42:53 bde Exp $
32  */
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/sysproto.h>
37 #include <sys/signalvar.h>
38 #include <sys/kernel.h>
39 #include <sys/mount.h>
40 #include <sys/filedesc.h>
41 #include <sys/fcntl.h>
42 #include <sys/acct.h>
43 #include <sys/exec.h>
44 #include <sys/imgact.h>
45 #include <sys/wait.h>
46 #include <sys/malloc.h>
47 #include <sys/sysent.h>
48 #include <sys/syslog.h>
49 #include <sys/shm.h>
50 
51 #include <vm/vm.h>
52 #include <vm/vm_kern.h>
53 
54 #include <machine/reg.h>
55 
56 int *exec_copyout_strings __P((struct image_params *));
57 
58 static int exec_check_permissions(struct image_params *);
59 
60 /*
61  * execsw_set is constructed for us by the linker.  Each of the items
62  * is a pointer to a `const struct execsw', hence the double pointer here.
63  */
64 const struct execsw **execsw = (const struct execsw **)&execsw_set.ls_items[0];
65 
66 #ifndef _SYS_SYSPROTO_H_
67 struct execve_args {
68         char    *fname;
69         char    **argv;
70         char    **envv;
71 };
72 #endif
73 
74 /*
75  * execve() system call.
76  */
77 int
78 execve(p, uap, retval)
79 	struct proc *p;
80 	register struct execve_args *uap;
81 	int *retval;
82 {
83 	struct nameidata nd, *ndp;
84 	int *stack_base;
85 	int error, len, i;
86 	struct image_params image_params, *imgp;
87 	struct vattr attr;
88 
89 	imgp = &image_params;
90 
91 	/*
92 	 * Initialize part of the common data
93 	 */
94 	imgp->proc = p;
95 	imgp->uap = uap;
96 	imgp->attr = &attr;
97 	imgp->image_header = NULL;
98 	imgp->argc = imgp->envc = 0;
99 	imgp->entry_addr = 0;
100 	imgp->vmspace_destroyed = 0;
101 	imgp->interpreted = 0;
102 	imgp->interpreter_name[0] = '\0';
103 
104 	/*
105 	 * Allocate temporary demand zeroed space for argument and
106 	 *	environment strings
107 	 */
108 	imgp->stringbase = (char *)kmem_alloc_pageable(exec_map, ARG_MAX);
109 	if (imgp->stringbase == NULL) {
110 		error = ENOMEM;
111 		goto exec_fail;
112 	}
113 	imgp->stringp = imgp->stringbase;
114 	imgp->stringspace = ARG_MAX;
115 
116 	/*
117 	 * Translate the file name. namei() returns a vnode pointer
118 	 *	in ni_vp amoung other things.
119 	 */
120 	ndp = &nd;
121 	NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
122 	    UIO_USERSPACE, uap->fname, p);
123 
124 interpret:
125 
126 	error = namei(ndp);
127 	if (error) {
128 		kmem_free(exec_map, (vm_offset_t)imgp->stringbase, ARG_MAX);
129 		goto exec_fail;
130 	}
131 
132 	imgp->vp = ndp->ni_vp;
133 	if (imgp->vp == NULL) {
134 		error = ENOEXEC;
135 		goto exec_fail_dealloc;
136 	}
137 
138 	/*
139 	 * Check file permissions (also 'opens' file)
140 	 */
141 	error = exec_check_permissions(imgp);
142 
143 	/*
144 	 * Lose the lock on the vnode. It's no longer needed, and must not
145 	 * exist for the pagefault paging to work below.
146 	 */
147 	VOP_UNLOCK(imgp->vp);
148 
149 	if (error)
150 		goto exec_fail_dealloc;
151 
152 	/*
153 	 * Map the image header (first page) of the file into
154 	 *	kernel address space
155 	 */
156 	error = vm_mmap(kernel_map,			/* map */
157 			(vm_offset_t *)&imgp->image_header, /* address */
158 			PAGE_SIZE,			/* size */
159 			VM_PROT_READ, 			/* protection */
160 			VM_PROT_READ, 			/* max protection */
161 			0,	 			/* flags */
162 			(caddr_t)imgp->vp,		/* vnode */
163 			0);				/* offset */
164 	if (error) {
165 		uprintf("mmap failed: %d\n",error);
166 		goto exec_fail_dealloc;
167 	}
168 
169 	/*
170 	 * Loop through list of image activators, calling each one.
171 	 *	If there is no match, the activator returns -1. If there
172 	 *	is a match, but there was an error during the activation,
173 	 *	the error is returned. Otherwise 0 means success. If the
174 	 *	image is interpreted, loop back up and try activating
175 	 *	the interpreter.
176 	 */
177 	for (i = 0; execsw[i]; ++i) {
178 		if (execsw[i]->ex_imgact)
179 			error = (*execsw[i]->ex_imgact)(imgp);
180 		else
181 			continue;
182 
183 		if (error == -1)
184 			continue;
185 		if (error)
186 			goto exec_fail_dealloc;
187 		if (imgp->interpreted) {
188 			/* free old vnode and name buffer */
189 			vrele(ndp->ni_vp);
190 			FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
191 			if (vm_map_remove(kernel_map, (vm_offset_t)imgp->image_header,
192 			    (vm_offset_t)imgp->image_header + PAGE_SIZE))
193 				panic("execve: header dealloc failed (1)");
194 
195 			/* set new name to that of the interpreter */
196 			NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
197 			    UIO_SYSSPACE, imgp->interpreter_name, p);
198 			goto interpret;
199 		}
200 		break;
201 	}
202 	/* If we made it through all the activators and none matched, exit. */
203 	if (error == -1) {
204 		error = ENOEXEC;
205 		goto exec_fail_dealloc;
206 	}
207 
208 	/*
209 	 * Copy out strings (args and env) and initialize stack base
210 	 */
211 	stack_base = exec_copyout_strings(imgp);
212 	p->p_vmspace->vm_minsaddr = (char *)stack_base;
213 
214 	/*
215 	 * If custom stack fixup routine present for this process
216 	 * let it do the stack setup.
217 	 * Else stuff argument count as first item on stack
218 	 */
219 	if (p->p_sysent->sv_fixup)
220 		(*p->p_sysent->sv_fixup)(&stack_base, imgp);
221 	else
222 		suword(--stack_base, imgp->argc);
223 
224 	/* close files on exec */
225 	fdcloseexec(p);
226 
227 	/* reset caught signals */
228 	execsigs(p);
229 
230 	/* name this process - nameiexec(p, ndp) */
231 	len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
232 	bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
233 	p->p_comm[len] = 0;
234 
235 	/*
236 	 * mark as executable, wakeup any process that was vforked and tell
237 	 * it that it now has it's own resources back
238 	 */
239 	p->p_flag |= P_EXEC;
240 	if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
241 		p->p_flag &= ~P_PPWAIT;
242 		wakeup((caddr_t)p->p_pptr);
243 	}
244 
245 	/*
246 	 * Implement image setuid/setgid. Disallow if the process is
247 	 * being traced.
248 	 */
249 	if ((attr.va_mode & (VSUID | VSGID)) &&
250 	    (p->p_flag & P_TRACED) == 0) {
251 		/*
252 		 * Turn off syscall tracing for set-id programs, except for
253 		 * root.
254 		 */
255 		if (p->p_tracep && suser(p->p_ucred, &p->p_acflag)) {
256 			p->p_traceflag = 0;
257 			vrele(p->p_tracep);
258 			p->p_tracep = NULL;
259 		}
260 		/*
261 		 * Set the new credentials.
262 		 */
263 		p->p_ucred = crcopy(p->p_ucred);
264 		if (attr.va_mode & VSUID)
265 			p->p_ucred->cr_uid = attr.va_uid;
266 		if (attr.va_mode & VSGID)
267 			p->p_ucred->cr_groups[0] = attr.va_gid;
268 		p->p_flag |= P_SUGID;
269 	} else {
270 		p->p_flag &= ~P_SUGID;
271 	}
272 
273 	/*
274 	 * Implement correct POSIX saved-id behavior.
275 	 */
276 	p->p_cred->p_svuid = p->p_ucred->cr_uid;
277 	p->p_cred->p_svgid = p->p_ucred->cr_gid;
278 
279 	/*
280 	 * Store the vp for use in procfs
281 	 */
282 	if (p->p_textvp)		/* release old reference */
283 		vrele(p->p_textvp);
284 	VREF(ndp->ni_vp);
285 	p->p_textvp = ndp->ni_vp;
286 
287 	/*
288 	 * If tracing the process, trap to debugger so breakpoints
289 	 * 	can be set before the program executes.
290 	 */
291 	if (p->p_flag & P_TRACED)
292 		psignal(p, SIGTRAP);
293 
294 	/* clear "fork but no exec" flag, as we _are_ execing */
295 	p->p_acflag &= ~AFORK;
296 
297 	/* Set entry address */
298 	setregs(p, imgp->entry_addr, (u_long)stack_base);
299 
300 	/*
301 	 * free various allocated resources
302 	 */
303 	kmem_free(exec_map, (vm_offset_t)imgp->stringbase, ARG_MAX);
304 	if (vm_map_remove(kernel_map, (vm_offset_t)imgp->image_header,
305 	    (vm_offset_t)imgp->image_header + PAGE_SIZE))
306 		panic("execve: header dealloc failed (2)");
307 	vrele(ndp->ni_vp);
308 	FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
309 
310 	return (0);
311 
312 exec_fail_dealloc:
313 	if (imgp->stringbase != NULL)
314 		kmem_free(exec_map, (vm_offset_t)imgp->stringbase, ARG_MAX);
315 	if (imgp->image_header && imgp->image_header != (char *)-1)
316 		if (vm_map_remove(kernel_map, (vm_offset_t)imgp->image_header,
317 		    (vm_offset_t)imgp->image_header + PAGE_SIZE))
318 			panic("execve: header dealloc failed (3)");
319 	if (ndp->ni_vp)
320 		vrele(ndp->ni_vp);
321 	FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
322 
323 exec_fail:
324 	if (imgp->vmspace_destroyed) {
325 		/* sorry, no more process anymore. exit gracefully */
326 		exit1(p, W_EXITCODE(0, SIGABRT));
327 		/* NOT REACHED */
328 		return(0);
329 	} else {
330 		return(error);
331 	}
332 }
333 
334 /*
335  * Destroy old address space, and allocate a new stack
336  *	The new stack is only SGROWSIZ large because it is grown
337  *	automatically in trap.c.
338  */
339 int
340 exec_new_vmspace(imgp)
341 	struct image_params *imgp;
342 {
343 	int error;
344 	struct vmspace *vmspace = imgp->proc->p_vmspace;
345 	caddr_t	stack_addr = (caddr_t) (USRSTACK - SGROWSIZ);
346 
347 	imgp->vmspace_destroyed = 1;
348 
349 	/* Blow away entire process VM */
350 #ifdef SYSVSHM
351 	if (vmspace->vm_shm)
352 		shmexit(imgp->proc);
353 #endif
354 	vm_map_remove(&vmspace->vm_map, 0, USRSTACK);
355 
356 	/* Allocate a new stack */
357 	error = vm_map_find(&vmspace->vm_map, NULL, 0, (vm_offset_t *)&stack_addr,
358 	    SGROWSIZ, FALSE);
359 	if (error)
360 		return(error);
361 
362 	vmspace->vm_ssize = SGROWSIZ >> PAGE_SHIFT;
363 
364 	/* Initialize maximum stack address */
365 	vmspace->vm_maxsaddr = (char *)USRSTACK - MAXSSIZ;
366 
367 	return(0);
368 }
369 
370 /*
371  * Copy out argument and environment strings from the old process
372  *	address space into the temporary string buffer.
373  */
374 int
375 exec_extract_strings(imgp)
376 	struct image_params *imgp;
377 {
378 	char	**argv, **envv;
379 	char	*argp, *envp;
380 	int	error, length;
381 
382 	/*
383 	 * extract arguments first
384 	 */
385 
386 	argv = imgp->uap->argv;
387 
388 	if (argv) {
389 		while ((argp = (caddr_t) fuword(argv++))) {
390 			if (argp == (caddr_t) -1)
391 				return (EFAULT);
392 			if ((error = copyinstr(argp, imgp->stringp,
393 			    imgp->stringspace, &length))) {
394 				if (error == ENAMETOOLONG)
395 					return(E2BIG);
396 				return (error);
397 			}
398 			imgp->stringspace -= length;
399 			imgp->stringp += length;
400 			imgp->argc++;
401 		}
402 	}
403 
404 	/*
405 	 * extract environment strings
406 	 */
407 
408 	envv = imgp->uap->envv;
409 
410 	if (envv) {
411 		while ((envp = (caddr_t) fuword(envv++))) {
412 			if (envp == (caddr_t) -1)
413 				return (EFAULT);
414 			if ((error = copyinstr(envp, imgp->stringp,
415 			    imgp->stringspace, &length))) {
416 				if (error == ENAMETOOLONG)
417 					return(E2BIG);
418 				return (error);
419 			}
420 			imgp->stringspace -= length;
421 			imgp->stringp += length;
422 			imgp->envc++;
423 		}
424 	}
425 
426 	return (0);
427 }
428 
429 /*
430  * Copy strings out to the new process address space, constructing
431  *	new arg and env vector tables. Return a pointer to the base
432  *	so that it can be used as the initial stack pointer.
433  */
434 int *
435 exec_copyout_strings(imgp)
436 	struct image_params *imgp;
437 {
438 	int argc, envc;
439 	char **vectp;
440 	char *stringp, *destp;
441 	int *stack_base;
442 	struct ps_strings *arginfo;
443 
444 	/*
445 	 * Calculate string base and vector table pointers.
446 	 */
447 	arginfo = PS_STRINGS;
448 	destp =	(caddr_t)arginfo - roundup((ARG_MAX - imgp->stringspace), sizeof(char *));
449 	/*
450 	 * The '+ 2' is for the null pointers at the end of each of the
451 	 *	arg and	env vector sets
452 	 */
453 	vectp = (char **) (destp -
454 		(imgp->argc + imgp->envc + 2) * sizeof(char *));
455 
456 	/*
457 	 * vectp also becomes our initial stack base
458 	 */
459 	stack_base = (int *)vectp;
460 
461 	stringp = imgp->stringbase;
462 	argc = imgp->argc;
463 	envc = imgp->envc;
464 
465 	/*
466 	 * Copy out strings - arguments and environment.
467 	 */
468 	copyout(stringp, destp, ARG_MAX - imgp->stringspace);
469 
470 	/*
471 	 * Fill in "ps_strings" struct for ps, w, etc.
472 	 */
473 	suword(&arginfo->ps_argvstr, (int)destp);
474 	suword(&arginfo->ps_nargvstr, argc);
475 
476 	/*
477 	 * Fill in argument portion of vector table.
478 	 */
479 	for (; argc > 0; --argc) {
480 		suword(vectp++, (int)destp);
481 		while (*stringp++ != 0)
482 			destp++;
483 		destp++;
484 	}
485 
486 	/* a null vector table pointer seperates the argp's from the envp's */
487 	suword(vectp++, NULL);
488 
489 	suword(&arginfo->ps_envstr, (int)destp);
490 	suword(&arginfo->ps_nenvstr, envc);
491 
492 	/*
493 	 * Fill in environment portion of vector table.
494 	 */
495 	for (; envc > 0; --envc) {
496 		suword(vectp++, (int)destp);
497 		while (*stringp++ != 0)
498 			destp++;
499 		destp++;
500 	}
501 
502 	/* end of vector table is a null pointer */
503 	suword(vectp, NULL);
504 
505 	return (stack_base);
506 }
507 
508 /*
509  * Check permissions of file to execute.
510  *	Return 0 for success or error code on failure.
511  */
512 static int
513 exec_check_permissions(imgp)
514 	struct image_params *imgp;
515 {
516 	struct proc *p = imgp->proc;
517 	struct vnode *vp = imgp->vp;
518 	struct vattr *attr = imgp->attr;
519 	int error;
520 
521 	/*
522 	 * Check number of open-for-writes on the file and deny execution
523 	 *	if there are any.
524 	 */
525 	if (vp->v_writecount) {
526 		return (ETXTBSY);
527 	}
528 
529 	/* Get file attributes */
530 	error = VOP_GETATTR(vp, attr, p->p_ucred, p);
531 	if (error)
532 		return (error);
533 
534 	/*
535 	 * 1) Check if file execution is disabled for the filesystem that this
536 	 *	file resides on.
537 	 * 2) Insure that at least one execute bit is on - otherwise root
538 	 *	will always succeed, and we don't want to happen unless the
539 	 *	file really is executable.
540 	 * 3) Insure that the file is a regular file.
541 	 */
542 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
543 	    ((attr->va_mode & 0111) == 0) ||
544 	    (attr->va_type != VREG)) {
545 		return (EACCES);
546 	}
547 
548 	/*
549 	 * Zero length files can't be exec'd
550 	 */
551 	if (attr->va_size == 0)
552 		return (ENOEXEC);
553 
554 	/*
555 	 * Disable setuid/setgid if the filesystem prohibits it or if
556 	 *	the process is being traced.
557 	 */
558         if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_flag & P_TRACED))
559 		attr->va_mode &= ~(VSUID | VSGID);
560 
561 	/*
562 	 *  Check for execute permission to file based on current credentials.
563 	 *	Then call filesystem specific open routine (which does nothing
564 	 *	in the general case).
565 	 */
566 	error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
567 	if (error)
568 		return (error);
569 
570 	error = VOP_OPEN(vp, FREAD, p->p_ucred, p);
571 	if (error)
572 		return (error);
573 
574 	return (0);
575 }
576