kern_exec.c revision 1.60 1 /* $NetBSD: kern_exec.c,v 1.60 1995/02/04 14:44:48 mycroft Exp $ */
2
3 /*-
4 * Copyright (C) 1993, 1994 Christopher G. Demetriou
5 * Copyright (C) 1992 Wolfgang Solfrank.
6 * Copyright (C) 1992 TooLs GmbH.
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by TooLs GmbH.
20 * 4. The name of TooLs GmbH may not be used to endorse or promote products
21 * derived from this software without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR
24 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
25 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
26 * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
28 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
29 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
30 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
31 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
32 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 */
34
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/filedesc.h>
38 #include <sys/kernel.h>
39 #include <sys/proc.h>
40 #include <sys/mount.h>
41 #include <sys/malloc.h>
42 #include <sys/namei.h>
43 #include <sys/vnode.h>
44 #include <sys/file.h>
45 #include <sys/acct.h>
46 #include <sys/exec.h>
47 #include <sys/ktrace.h>
48 #include <sys/resourcevar.h>
49 #include <sys/wait.h>
50 #include <sys/mman.h>
51 #include <sys/signalvar.h>
52 #include <sys/stat.h>
53
54 #include <sys/syscallargs.h>
55
56 #include <vm/vm.h>
57 #include <vm/vm_kern.h>
58
59 #include <machine/cpu.h>
60 #include <machine/reg.h>
61
62 #ifdef COPY_SIGCODE
63 extern char sigcode[], esigcode[];
64
65 #define szsigcode (esigcode - sigcode)
66 #else
67 #define szsigcode 0
68 #endif
69
70 /*
71 * check exec:
72 * given an "executable" described in the exec package's namei info,
73 * see what we can do with it.
74 *
75 * ON ENTRY:
76 * exec package with appropriate namei info
77 * proc pointer of exec'ing proc
78 * NO SELF-LOCKED VNODES
79 *
80 * ON EXIT:
81 * error: nothing held, etc. exec header still allocated.
82 * ok: filled exec package, one locked vnode.
83 *
84 * EXEC SWITCH ENTRY:
85 * Locked vnode to check, exec package, proc.
86 *
87 * EXEC SWITCH EXIT:
88 * ok: return 0, filled exec package, one locked vnode.
89 * error: destructive:
90 * everything deallocated execept exec header.
91 * non-descructive:
92 * error code, locked vnode, exec header unmodified
93 */
94 int
95 check_exec(p, epp)
96 struct proc *p;
97 struct exec_package *epp;
98 {
99 int error, i;
100 struct vnode *vp;
101 char *cp, *ep, *name;
102 struct nameidata *ndp;
103 int resid;
104
105 ndp = epp->ep_ndp;
106 ndp->ni_cnd.cn_nameiop = LOOKUP;
107 ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF | SAVENAME;
108 /* first get the vnode */
109 if (error = namei(ndp))
110 return error;
111 epp->ep_vp = vp = ndp->ni_vp;
112
113 /* check for regular file */
114 if (vp->v_type != VREG) {
115 error = EACCES;
116 goto bad1;
117 }
118
119 /* get attributes */
120 if (error = VOP_GETATTR(vp, epp->ep_vap, p->p_ucred, p))
121 goto bad1;
122
123 /* Check mount point */
124 if (vp->v_mount->mnt_flag & MNT_NOEXEC) {
125 error = EACCES;
126 goto bad1;
127 }
128 if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_flag & P_TRACED))
129 epp->ep_vap->va_mode &= ~(VSUID | VSGID);
130
131 /* check access. for root we have to see if any exec bit on */
132 if (error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p))
133 goto bad1;
134 if ((epp->ep_vap->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0) {
135 error = EACCES;
136 goto bad1;
137 }
138
139 /* try to open it */
140 if (error = VOP_OPEN(vp, FREAD, p->p_ucred, p))
141 goto bad1;
142
143 /* now we have the file, get the exec header */
144 if (error = vn_rdwr(UIO_READ, vp, epp->ep_hdr, epp->ep_hdrlen, 0,
145 UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid, p))
146 goto bad2;
147 epp->ep_hdrvalid = epp->ep_hdrlen - resid;
148
149 /*
150 * set up the vmcmds for creation of the process
151 * address space
152 */
153 error = ENOEXEC;
154 for (i = 0; i < nexecs && error != 0; i++) {
155 if (execsw[i].es_check != NULL)
156 error = (*execsw[i].es_check)(p, epp);
157 if (epp->ep_flags & EXEC_DESTR && error != 0)
158 return error;
159 }
160 if (!error) {
161 /* check that entry point is sane */
162 if (epp->ep_entry > VM_MAXUSER_ADDRESS)
163 error = ENOEXEC;
164
165 /* check limits */
166 if ((epp->ep_tsize > MAXTSIZ) ||
167 (epp->ep_dsize > p->p_rlimit[RLIMIT_DATA].rlim_cur))
168 error = ENOMEM;
169
170 if (!error)
171 return (0);
172 }
173
174 /*
175 * free any vmspace-creation commands,
176 * and release their references
177 */
178 kill_vmcmds(&epp->ep_vmcmds);
179
180 bad2:
181 /*
182 * unlock and close the vnode, restore the old one, free the
183 * pathname buf, and punt.
184 */
185 VOP_UNLOCK(vp);
186 vn_close(vp, FREAD, p->p_ucred, p);
187 FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
188 return error;
189
190 bad1:
191 /*
192 * free the namei pathname buffer, and put the vnode
193 * (which we don't yet have open).
194 */
195 FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
196 vput(vp);
197 return error;
198 }
199
200 /*
201 * exec system call
202 */
203 /* ARGSUSED */
204 execve(p, uap, retval)
205 register struct proc *p;
206 register struct execve_args /* {
207 syscallarg(char *) path;
208 syscallarg(char * *) argp;
209 syscallarg(char * *) envp;
210 } */ *uap;
211 register_t *retval;
212 {
213 int error, i;
214 struct exec_package pack;
215 struct nameidata nid;
216 struct vattr attr;
217 struct ucred *cred = p->p_ucred;
218 char *argp;
219 char **cpp, *dp, *sp, *np;
220 int argc, envc, len;
221 char *stack;
222 struct ps_strings arginfo;
223 struct vmspace *vm = p->p_vmspace;
224 char **tmpfap;
225
226 /*
227 * figure out the maximum size of an exec header, if necessary.
228 * XXX should be able to keep LKM code from modifying exec switch
229 * when we're still using it, but...
230 */
231 if (exec_maxhdrsz == 0) {
232 for (i = 0; i < nexecs; i++)
233 if (execsw[i].es_check != NULL
234 && execsw[i].es_hdrsz > exec_maxhdrsz)
235 exec_maxhdrsz = execsw[i].es_hdrsz;
236 }
237
238 /* init the namei data to point the file user's program name */
239 NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
240
241 /*
242 * initialize the fields of the exec package.
243 */
244 pack.ep_name = SCARG(uap, path);
245 MALLOC(pack.ep_hdr, void *, exec_maxhdrsz, M_EXEC, M_WAITOK);
246 pack.ep_hdrlen = exec_maxhdrsz;
247 pack.ep_hdrvalid = 0;
248 pack.ep_ndp = &nid;
249 pack.ep_setup = NULL; /* assume no setup function */
250 pack.ep_setup_arg = NULL;
251 pack.ep_setup_arglen = 0;
252 pack.ep_vmcmds.evs_cnt = 0;
253 pack.ep_vmcmds.evs_used = 0;
254 pack.ep_vap = &attr;
255 pack.ep_emul = EMUL_NETBSD;
256 pack.ep_flags = 0;
257
258 /* see if we can run it. */
259 if (error = check_exec(p, &pack))
260 goto freehdr;
261
262 /* XXX -- THE FOLLOWING SECTION NEEDS MAJOR CLEANUP */
263
264 /* allocate an argument buffer */
265 argp = (char *) kmem_alloc_wait(exec_map, NCARGS);
266 #ifdef DIAGNOSTIC
267 if (argp == (vm_offset_t) 0)
268 panic("execve: argp == NULL");
269 #endif
270 dp = argp;
271 argc = 0;
272
273 /* copy the fake args list, if there's one, freeing it as we go */
274 if (pack.ep_flags & EXEC_HASARGL) {
275 tmpfap = pack.ep_fa;
276 while (*tmpfap != NULL) {
277 char *cp;
278
279 cp = *tmpfap;
280 while (*cp)
281 *dp++ = *cp++;
282 *dp++;
283
284 FREE(*tmpfap, M_EXEC);
285 tmpfap++; argc++;
286 }
287 FREE(pack.ep_fa, M_EXEC);
288 pack.ep_flags &= ~EXEC_HASARGL;
289 }
290
291 /* Now get argv & environment */
292 if (!(cpp = SCARG(uap, argp))) {
293 error = EINVAL;
294 goto bad;
295 }
296
297 if (pack.ep_flags & EXEC_SKIPARG)
298 cpp++;
299
300 while (1) {
301 len = argp + ARG_MAX - dp;
302 if (error = copyin(cpp, &sp, sizeof(sp)))
303 goto bad;
304 if (!sp)
305 break;
306 if (error = copyinstr(sp, dp, len, (u_int *) & len)) {
307 if (error == ENAMETOOLONG)
308 error = E2BIG;
309 goto bad;
310 }
311 dp += len;
312 cpp++;
313 argc++;
314 }
315
316 envc = 0;
317 if (cpp = SCARG(uap, envp)) { /* environment need not be there */
318 while (1) {
319 len = argp + ARG_MAX - dp;
320 if (error = copyin(cpp, &sp, sizeof(sp)))
321 goto bad;
322 if (!sp)
323 break;
324 if (error = copyinstr(sp, dp, len, (u_int *) & len)) {
325 if (error == ENAMETOOLONG)
326 error = E2BIG;
327 goto bad;
328 }
329 dp += len;
330 cpp++;
331 envc++;
332 }
333 dp = (char *) ALIGN(dp);
334 }
335
336 /* Now check if args & environ fit into new stack */
337 len = ((argc + envc + 2 + pack.ep_setup_arglen) * sizeof(char *) +
338 sizeof(int) + dp + STACKGAPLEN + szsigcode +
339 sizeof(struct ps_strings)) - argp;
340 len = ALIGN(len); /* make the stack "safely" aligned */
341
342 if (len > pack.ep_ssize) { /* in effect, compare to initial limit */
343 error = ENOMEM;
344 goto bad;
345 }
346
347 /* adjust "active stack depth" for process VSZ */
348 pack.ep_ssize = len; /* maybe should go elsewhere, but... */
349
350 /* Unmap old program */
351 #ifdef sparc
352 kill_user_windows(p); /* before stack addresses go away */
353 #endif
354 /* Kill shared memory and unmap old program */
355 #ifdef SYSVSHM
356 if (vm->vm_shm)
357 shmexit(p);
358 #endif
359 vm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
360 VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
361
362 /* Now map address space */
363 vm->vm_taddr = (char *) pack.ep_taddr;
364 vm->vm_tsize = btoc(pack.ep_tsize);
365 vm->vm_daddr = (char *) pack.ep_daddr;
366 vm->vm_dsize = btoc(pack.ep_dsize);
367 vm->vm_ssize = btoc(pack.ep_ssize);
368 vm->vm_maxsaddr = (char *) pack.ep_maxsaddr;
369
370 /* create the new process's VM space by running the vmcmds */
371 #ifdef DIAGNOSTIC
372 if (pack.ep_vmcmds.evs_used == 0)
373 panic("execve: no vmcmds");
374 #endif
375 for (i = 0; i < pack.ep_vmcmds.evs_used && !error; i++) {
376 struct exec_vmcmd *vcp;
377
378 vcp = &pack.ep_vmcmds.evs_cmds[i];
379 error = (*vcp->ev_proc)(p, vcp);
380 }
381
382 /* free the vmspace-creation commands, and release their references */
383 kill_vmcmds(&pack.ep_vmcmds);
384
385 /* if an error happened, deallocate and punt */
386 if (error)
387 goto exec_abort;
388
389 /* remember information about the process */
390 arginfo.ps_nargvstr = argc;
391 arginfo.ps_nenvstr = envc;
392
393 /* Now copy argc, args & environ to new stack */
394 stack = (char *) (USRSTACK - len);
395 cpp = (char **) stack;
396
397 if (copyout(&argc, cpp++, sizeof(argc)))
398 goto exec_abort;
399 dp = (char *) (cpp + argc + envc + 2 + pack.ep_setup_arglen);
400 sp = argp;
401 np = 0;
402
403 /* XXX don't copy them out, remap them! */
404 arginfo.ps_argvstr = dp; /* remember location of argv for later */
405 for (; --argc >= 0; sp += len, dp += len) {
406 if (copyout(&dp, cpp++, sizeof(dp)) ||
407 copyoutstr(sp, dp, ARG_MAX, &len))
408 goto exec_abort;
409 }
410 if (copyout(&np, cpp++, sizeof(np)))
411 goto exec_abort;
412
413 arginfo.ps_envstr = dp; /* remember location of envp for later */
414 for (; --envc >= 0; sp += len, dp += len) {
415 if (copyout(&dp, cpp++, sizeof(dp)) ||
416 copyoutstr(sp, dp, ARG_MAX, &len))
417 goto exec_abort;
418 }
419 if (copyout(&np, cpp++, sizeof(np)))
420 goto exec_abort;
421
422 if (pack.ep_setup != NULL)
423 (*pack.ep_setup)(EXEC_SETUP_ADDARGS, p, &pack, cpp);
424
425 /* copy out the process's ps_strings structure */
426 if (copyout(&arginfo, (char *) PS_STRINGS, sizeof(arginfo)))
427 goto exec_abort;
428
429 #ifdef COPY_SIGCODE
430 /* copy out the process's signal trapoline code */
431 if (copyout((char *) sigcode, ((char *) PS_STRINGS) - szsigcode,
432 szsigcode)) {
433 goto exec_abort;
434 }
435 #endif
436
437 fdcloseexec(p); /* handle close on exec */
438 execsigs(p); /* reset catched signals */
439
440 /* set command name & other accounting info */
441 len = min(nid.ni_cnd.cn_namelen, MAXCOMLEN);
442 bcopy(nid.ni_cnd.cn_nameptr, p->p_comm, len);
443 p->p_comm[len] = 0;
444 p->p_acflag &= ~AFORK;
445
446 /* record proc's vnode, for use by procfs and others */
447 if (p->p_textvp)
448 vrele(p->p_textvp);
449 VREF(pack.ep_vp);
450 p->p_textvp = pack.ep_vp;
451
452 p->p_flag |= P_EXEC;
453 if (p->p_flag & P_PPWAIT) {
454 p->p_flag &= ~P_PPWAIT;
455 wakeup((caddr_t) p->p_pptr);
456 }
457
458 /*
459 * deal with set[ug]id.
460 * MNT_NOEXEC and P_TRACED have already been used to disable s[ug]id.
461 */
462 p->p_flag &= ~P_SUGID;
463 if (((attr.va_mode & VSUID) != 0 &&
464 p->p_ucred->cr_uid != attr.va_uid)
465 || (attr.va_mode & VSGID) != 0 &&
466 p->p_ucred->cr_gid != attr.va_gid) {
467 p->p_ucred = crcopy(cred);
468 #ifdef KTRACE
469 /*
470 * If process is being ktraced, turn off - unless
471 * root set it.
472 */
473 if (p->p_tracep && !(p->p_traceflag & KTRFAC_ROOT)) {
474 vrele(p->p_tracep);
475 p->p_tracep = NULL;
476 p->p_traceflag = 0;
477 }
478 #endif
479 if (attr.va_mode & VSUID)
480 p->p_ucred->cr_uid = attr.va_uid;
481 if (attr.va_mode & VSGID)
482 p->p_ucred->cr_gid = attr.va_gid;
483 p->p_flag |= P_SUGID;
484 }
485 p->p_cred->p_svuid = p->p_ucred->cr_uid;
486 p->p_cred->p_svgid = p->p_ucred->cr_gid;
487
488 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
489
490 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
491 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
492 vput(pack.ep_vp);
493
494 /* setup new registers and do misc. setup. */
495 setregs(p, pack.ep_entry, (u_long) stack, retval);
496 if (pack.ep_setup != NULL)
497 (*pack.ep_setup)(EXEC_SETUP_FINISH, p, &pack, NULL);
498
499 if (p->p_flag & P_TRACED)
500 psignal(p, SIGTRAP);
501
502 p->p_emul = pack.ep_emul;
503 FREE(pack.ep_hdr, M_EXEC);
504 return 0;
505
506 bad:
507 if (pack.ep_setup != NULL)
508 (*pack.ep_setup)(EXEC_SETUP_CLEANUP, p, &pack, dp);
509 /* free the vmspace-creation commands, and release their references */
510 kill_vmcmds(&pack.ep_vmcmds);
511 /* kill any opened file descriptor, if necessary */
512 if (pack.ep_flags & EXEC_HASFD) {
513 pack.ep_flags &= ~EXEC_HASFD;
514 (void) fdclose(p, pack.ep_fd);
515 }
516 /* close and put the exec'd file */
517 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
518 vput(pack.ep_vp);
519 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
520 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
521
522 freehdr:
523 FREE(pack.ep_hdr, M_EXEC);
524 return error;
525
526 exec_abort:
527 /*
528 * the old process doesn't exist anymore. exit gracefully.
529 * get rid of the (new) address space we have created, if any, get rid
530 * of our namei data and vnode, and exit noting failure
531 */
532 vm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
533 VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
534 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
535 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
536 vput(pack.ep_vp);
537 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
538 FREE(pack.ep_hdr, M_EXEC);
539 exit1(p, W_EXITCODE(0, SIGABRT));
540 exit1(p, -1);
541
542 /* NOTREACHED */
543 return 0;
544 }
545