xref: /freebsd/sys/kern/kern_exec.c (revision afe61c15161c324a7af299a9b8457aba5afc92db)
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_execve.c,v 1.20 1994/03/26 12:24:27 davidg Exp $
32  */
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/signalvar.h>
37 #include <sys/resourcevar.h>
38 #include <sys/kernel.h>
39 #include <sys/mount.h>
40 #include <sys/file.h>
41 #include <sys/acct.h>
42 #include <sys/exec.h>
43 #include <sys/imgact.h>
44 #include <sys/stat.h>
45 #include <sys/wait.h>
46 #include <sys/mman.h>
47 #include <sys/malloc.h>
48 #include <sys/syslog.h>
49 
50 #include <vm/vm.h>
51 #include <vm/vm_kern.h>
52 
53 #include <machine/reg.h>
54 
55 int exec_extract_strings __P((struct image_params *));
56 int *exec_copyout_strings __P((struct image_params *));
57 
58 /*
59  * execsw_set is constructed for us by the linker.  Each of the items
60  * is a pointer to a `const struct execsw', hence the double pointer here.
61  */
62 extern const struct linker_set execsw_set;
63 const struct execsw **execsw = (const struct execsw **)&execsw_set.ls_items[0];
64 
65 /*
66  * execve() system call.
67  */
68 int
69 execve(p, uap, retval)
70 	struct proc *p;
71 	register struct execve_args *uap;
72 	int *retval;
73 {
74 	struct nameidata nd, *ndp;
75 	char *stringbase, *stringp;
76 	int *stack_base;
77 	int error, resid, len, i;
78 	struct image_params image_params, *iparams;
79 	struct vnode *vnodep;
80 	struct vattr attr;
81 	char *image_header;
82 
83 	iparams = &image_params;
84 	bzero((caddr_t)iparams, sizeof(struct image_params));
85 	image_header = (char *)0;
86 
87 	/*
88 	 * Initialize a few constants in the common area
89 	 */
90 	iparams->proc = p;
91 	iparams->uap = uap;
92 	iparams->attr = &attr;
93 
94 	/*
95 	 * Allocate temporary demand zeroed space for argument and
96 	 *	environment strings
97 	 */
98 	error = vm_allocate(kernel_map, (vm_offset_t *)&iparams->stringbase,
99 			    ARG_MAX, TRUE);
100 	if (error) {
101 		log(LOG_WARNING, "execve: failed to allocate string space\n");
102 		return (error);
103 	}
104 
105 	if (!iparams->stringbase) {
106 		error = ENOMEM;
107 		goto exec_fail;
108 	}
109 	iparams->stringp = iparams->stringbase;
110 	iparams->stringspace = ARG_MAX;
111 
112 	/*
113 	 * Translate the file name. namei() returns a vnode pointer
114 	 *	in ni_vp amoung other things.
115 	 */
116 	ndp = &nd;
117 	ndp->ni_cnd.cn_nameiop = LOOKUP;
118 	ndp->ni_cnd.cn_flags = LOCKLEAF | FOLLOW | SAVENAME;
119 	ndp->ni_cnd.cn_proc = curproc;
120 	ndp->ni_cnd.cn_cred = curproc->p_cred->pc_ucred;
121 	ndp->ni_segflg = UIO_USERSPACE;
122 	ndp->ni_dirp = uap->fname;
123 
124 interpret:
125 
126 	error = namei(ndp);
127 	if (error) {
128 		vm_deallocate(kernel_map, (vm_offset_t)iparams->stringbase,
129 			      ARG_MAX);
130 		goto exec_fail;
131 	}
132 
133 	iparams->vnodep = vnodep = ndp->ni_vp;
134 
135 	if (vnodep == NULL) {
136 		error = ENOEXEC;
137 		goto exec_fail_dealloc;
138 	}
139 
140 	/*
141 	 * Check file permissions (also 'opens' file)
142 	 */
143 	error = exec_check_permissions(iparams);
144 	if (error)
145 		goto exec_fail_dealloc;
146 
147 	/*
148 	 * Map the image header (first page) of the file into
149 	 *	kernel address space
150 	 */
151 	error = vm_mmap(kernel_map,			/* map */
152 			(vm_offset_t *)&image_header,	/* address */
153 			PAGE_SIZE,			/* size */
154 			VM_PROT_READ, 			/* protection */
155 			VM_PROT_READ, 			/* max protection */
156 			0,	 			/* flags */
157 			(caddr_t)vnodep,		/* vnode */
158 			0);				/* offset */
159 	if (error) {
160 		uprintf("mmap failed: %d\n",error);
161 		goto exec_fail_dealloc;
162 	}
163 	iparams->image_header = image_header;
164 
165 	/*
166 	 * Loop through list of image activators, calling each one.
167 	 *	If there is no match, the activator returns -1. If there
168 	 *	is a match, but there was an error during the activation,
169 	 *	the error is returned. Otherwise 0 means success. If the
170 	 *	image is interpreted, loop back up and try activating
171 	 *	the interpreter.
172 	 */
173 	for (i = 0; execsw[i]; ++i) {
174 		if (execsw[i]->ex_imgact)
175 			error = (*execsw[i]->ex_imgact)(iparams);
176 		else
177 			continue;
178 
179 		if (error == -1)
180 			continue;
181 		if (error)
182 			goto exec_fail_dealloc;
183 		if (iparams->interpreted) {
184 			/* free old vnode and name buffer */
185 			vput(ndp->ni_vp);
186 			FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
187 			if (vm_deallocate(kernel_map,
188 					  (vm_offset_t)image_header, PAGE_SIZE))
189 				panic("execve: header dealloc failed (1)");
190 
191 			/* set new name to that of the interpreter */
192 			ndp->ni_segflg = UIO_SYSSPACE;
193 			ndp->ni_dirp = iparams->interpreter_name;
194 			ndp->ni_cnd.cn_nameiop = LOOKUP;
195 			ndp->ni_cnd.cn_flags = LOCKLEAF | FOLLOW | SAVENAME;
196 			ndp->ni_cnd.cn_proc = curproc;
197 			ndp->ni_cnd.cn_cred = curproc->p_cred->pc_ucred;
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(iparams);
212 	p->p_vmspace->vm_minsaddr = (char *)stack_base;
213 
214 	/*
215 	 * Stuff argument count as first item on stack
216 	 */
217 	*(--stack_base) = iparams->argc;
218 
219 	/* close files on exec */
220 	fdcloseexec(p);
221 
222 	/* reset caught signals */
223 	execsigs(p);
224 
225 	/* name this process - nameiexec(p, ndp) */
226 	len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
227 	bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
228 	p->p_comm[len] = 0;
229 
230 	/*
231 	 * mark as executable, wakeup any process that was vforked and tell
232 	 * it that it now has it's own resources back
233 	 */
234 	p->p_flag |= P_EXEC;
235 	if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
236 		p->p_flag &= ~P_PPWAIT;
237 		wakeup((caddr_t)p->p_pptr);
238 	}
239 
240 	/* implement set userid/groupid */
241 	p->p_flag &= ~P_SUGID;
242 
243 	/*
244 	 * Turn off kernel tracing for set-id programs, except for
245 	 * root.
246 	 */
247 	if (p->p_tracep && (attr.va_mode & (VSUID | VSGID)) &&
248 	    suser(p->p_ucred, &p->p_acflag)) {
249 		p->p_traceflag = 0;
250 		vrele(p->p_tracep);
251 		p->p_tracep = 0;
252 	}
253 	if ((attr.va_mode & VSUID) && (p->p_flag & P_TRACED) == 0) {
254 		p->p_ucred = crcopy(p->p_ucred);
255 		p->p_ucred->cr_uid = attr.va_uid;
256 		p->p_flag |= P_SUGID;
257 	}
258 	if ((attr.va_mode & VSGID) && (p->p_flag & P_TRACED) == 0) {
259 		p->p_ucred = crcopy(p->p_ucred);
260 		p->p_ucred->cr_groups[0] = attr.va_gid;
261 		p->p_flag |= P_SUGID;
262 	}
263 
264 	/*
265 	 * Implement correct POSIX saved uid behavior.
266 	 */
267 	p->p_cred->p_svuid = p->p_ucred->cr_uid;
268 	p->p_cred->p_svgid = p->p_ucred->cr_gid;
269 
270 	/* mark vnode pure text */
271  	ndp->ni_vp->v_flag |= VTEXT;
272 
273 	/*
274 	 * If tracing the process, trap to debugger so breakpoints
275 	 * 	can be set before the program executes.
276 	 */
277 	if (p->p_flag & P_TRACED)
278 		psignal(p, SIGTRAP);
279 
280 	/* clear "fork but no exec" flag, as we _are_ execing */
281 	p->p_acflag &= ~AFORK;
282 
283 	/* Set entry address */
284 	setregs(p, iparams->entry_addr, stack_base);
285 
286 	/*
287 	 * free various allocated resources
288 	 */
289 	if (vm_deallocate(kernel_map, (vm_offset_t)iparams->stringbase, ARG_MAX))
290 		panic("execve: string buffer dealloc failed (1)");
291 	if (vm_deallocate(kernel_map, (vm_offset_t)image_header, PAGE_SIZE))
292 		panic("execve: header dealloc failed (2)");
293 	vput(ndp->ni_vp);
294 	FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
295 
296 	return (0);
297 
298 exec_fail_dealloc:
299 	if (iparams->stringbase && iparams->stringbase != (char *)-1)
300 		if (vm_deallocate(kernel_map, (vm_offset_t)iparams->stringbase,
301 				  ARG_MAX))
302 			panic("execve: string buffer dealloc failed (2)");
303 	if (iparams->image_header && iparams->image_header != (char *)-1)
304 		if (vm_deallocate(kernel_map,
305 				  (vm_offset_t)iparams->image_header, PAGE_SIZE))
306 			panic("execve: header dealloc failed (3)");
307 	vput(ndp->ni_vp);
308 	FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
309 
310 exec_fail:
311 	if (iparams->vmspace_destroyed) {
312 		/* sorry, no more process anymore. exit gracefully */
313 #if 0	/* XXX */
314 		vm_deallocate(&vs->vm_map, USRSTACK - MAXSSIZ, MAXSSIZ);
315 #endif
316 		exit1(p, W_EXITCODE(0, SIGABRT));
317 		/* NOT REACHED */
318 		return(0);
319 	} else {
320 		return(error);
321 	}
322 }
323 
324 /*
325  * Destroy old address space, and allocate a new stack
326  *	The new stack is only SGROWSIZ large because it is grown
327  *	automatically in trap.c.
328  */
329 int
330 exec_new_vmspace(iparams)
331 	struct image_params *iparams;
332 {
333 	int error;
334 	struct vmspace *vmspace = iparams->proc->p_vmspace;
335 	caddr_t	stack_addr = (caddr_t) (USRSTACK - SGROWSIZ);
336 
337 	iparams->vmspace_destroyed = 1;
338 
339 	/* Blow away entire process VM */
340 	vm_deallocate(&vmspace->vm_map, 0, USRSTACK);
341 
342 	/* Allocate a new stack */
343 	error = vm_allocate(&vmspace->vm_map, (vm_offset_t *)&stack_addr,
344 			    SGROWSIZ, FALSE);
345 	if (error)
346 		return(error);
347 
348 	vmspace->vm_ssize = SGROWSIZ >> PAGE_SHIFT;
349 
350 	/* Initialize maximum stack address */
351 	vmspace->vm_maxsaddr = (char *)USRSTACK - MAXSSIZ;
352 
353 	return(0);
354 }
355 
356 /*
357  * Copy out argument and environment strings from the old process
358  *	address space into the temporary string buffer.
359  */
360 int
361 exec_extract_strings(iparams)
362 	struct image_params *iparams;
363 {
364 	char	**argv, **envv;
365 	char	*argp, *envp;
366 	int	length;
367 
368 	/*
369 	 * extract arguments first
370 	 */
371 
372 	argv = iparams->uap->argv;
373 
374 	if (argv)
375 		while (argp = (caddr_t) fuword(argv++)) {
376 			if (argp == (caddr_t) -1)
377 				return (EFAULT);
378 			if (copyinstr(argp, iparams->stringp, iparams->stringspace,
379 				&length) == ENAMETOOLONG)
380 					return(E2BIG);
381 			iparams->stringspace -= length;
382 			iparams->stringp += length;
383 			iparams->argc++;
384 		}
385 
386 	/*
387 	 * extract environment strings
388 	 */
389 
390 	envv = iparams->uap->envv;
391 
392 	if (envv)
393 		while (envp = (caddr_t) fuword(envv++)) {
394 			if (envp == (caddr_t) -1)
395 				return (EFAULT);
396 			if (copyinstr(envp, iparams->stringp, iparams->stringspace,
397 				&length) == ENAMETOOLONG)
398 					return(E2BIG);
399 			iparams->stringspace -= length;
400 			iparams->stringp += length;
401 			iparams->envc++;
402 		}
403 
404 	return (0);
405 }
406 
407 /*
408  * Copy strings out to the new process address space, constructing
409  *	new arg and env vector tables. Return a pointer to the base
410  *	so that it can be used as the initial stack pointer.
411  */
412 int *
413 exec_copyout_strings(iparams)
414 	struct image_params *iparams;
415 {
416 	int argc, envc;
417 	char **vectp;
418 	char *stringp, *destp;
419 	int *stack_base;
420 	int vect_table_size, string_table_size;
421 
422 	/*
423 	 * Calculate string base and vector table pointers.
424 	 */
425 	destp = (caddr_t) ((caddr_t)USRSTACK -
426 		roundup((ARG_MAX - iparams->stringspace), sizeof(char *)));
427 	/*
428 	 * The '+ 2' is for the null pointers at the end of each of the
429 	 *	arg and	env vector sets
430 	 */
431 	vectp = (char **) (destp -
432 		(iparams->argc + iparams->envc + 2) * sizeof(char *));
433 
434 	/*
435 	 * vectp also becomes our initial stack base
436 	 */
437 	stack_base = (int *)vectp;
438 
439 	stringp = iparams->stringbase;
440 	argc = iparams->argc;
441 	envc = iparams->envc;
442 
443 	for (; argc > 0; --argc) {
444 		*(vectp++) = destp;
445 		while (*destp++ = *stringp++);
446 	}
447 
448 	/* a null vector table pointer seperates the argp's from the envp's */
449 	*(vectp++) = NULL;
450 
451 	for (; envc > 0; --envc) {
452 		*(vectp++) = destp;
453 		while (*destp++ = *stringp++);
454 	}
455 
456 	/* end of vector table is a null pointer */
457 	*vectp = NULL;
458 
459 	return (stack_base);
460 }
461 
462 /*
463  * Check permissions of file to execute.
464  *	Return 0 for success or error code on failure.
465  */
466 int
467 exec_check_permissions(iparams)
468 	struct image_params *iparams;
469 {
470 	struct proc *p = iparams->proc;
471 	struct vnode *vnodep = iparams->vnodep;
472 	struct vattr *attr = iparams->attr;
473 	int error;
474 
475 	/*
476 	 * Check number of open-for-writes on the file and deny execution
477 	 *	if there are any.
478 	 */
479 	if (vnodep->v_writecount) {
480 		return (ETXTBSY);
481 	}
482 
483 	/* Get file attributes */
484 	error = VOP_GETATTR(vnodep, attr, p->p_ucred, p);
485 	if (error)
486 		return (error);
487 
488 	/*
489 	 * 1) Check if file execution is disabled for the filesystem that this
490 	 *	file resides on.
491 	 * 2) Insure that at least one execute bit is on - otherwise root
492 	 *	will always succeed, and we don't want to happen unless the
493 	 *	file really is executable.
494 	 * 3) Insure that the file is a regular file.
495 	 */
496 	if ((vnodep->v_mount->mnt_flag & MNT_NOEXEC) ||
497 	    ((attr->va_mode & 0111) == 0) ||
498 	    (attr->va_type != VREG)) {
499 		return (EACCES);
500 	}
501 
502 	/*
503 	 * Zero length files can't be exec'd
504 	 */
505 	if (attr->va_size == 0)
506 		return (ENOEXEC);
507 
508 	/*
509 	 * Disable setuid/setgid if the filesystem prohibits it or if
510 	 *	the process is being traced.
511 	 */
512         if ((vnodep->v_mount->mnt_flag & MNT_NOSUID) || (p->p_flag & P_TRACED))
513 		attr->va_mode &= ~(VSUID | VSGID);
514 
515 	/*
516 	 *  Check for execute permission to file based on current credentials.
517 	 *	Then call filesystem specific open routine (which does nothing
518 	 *	in the general case).
519 	 */
520 	error = VOP_ACCESS(vnodep, VEXEC, p->p_ucred, p);
521 	if (error)
522 		return (error);
523 
524 	error = VOP_OPEN(vnodep, FREAD, p->p_ucred, p);
525 	if (error)
526 		return (error);
527 
528 	return (0);
529 }
530