kern_exec.c revision 1.65 1 /* $NetBSD: kern_exec.c,v 1.65 1995/04/07 22:33:23 fvdl 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 #endif
65
66 /*
67 * check exec:
68 * given an "executable" described in the exec package's namei info,
69 * see what we can do with it.
70 *
71 * ON ENTRY:
72 * exec package with appropriate namei info
73 * proc pointer of exec'ing proc
74 * NO SELF-LOCKED VNODES
75 *
76 * ON EXIT:
77 * error: nothing held, etc. exec header still allocated.
78 * ok: filled exec package, one locked vnode.
79 *
80 * EXEC SWITCH ENTRY:
81 * Locked vnode to check, exec package, proc.
82 *
83 * EXEC SWITCH EXIT:
84 * ok: return 0, filled exec package, one locked vnode.
85 * error: destructive:
86 * everything deallocated execept exec header.
87 * non-descructive:
88 * error code, locked vnode, exec header unmodified
89 */
90 int
91 check_exec(p, epp)
92 struct proc *p;
93 struct exec_package *epp;
94 {
95 int error, i;
96 struct vnode *vp;
97 char *cp, *ep, *name;
98 struct nameidata *ndp;
99 int resid;
100
101 ndp = epp->ep_ndp;
102 ndp->ni_cnd.cn_nameiop = LOOKUP;
103 ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF | SAVENAME;
104 /* first get the vnode */
105 if (error = namei(ndp))
106 return error;
107 epp->ep_vp = vp = ndp->ni_vp;
108
109 /* check for regular file */
110 if (vp->v_type != VREG) {
111 error = EACCES;
112 goto bad1;
113 }
114
115 /* get attributes */
116 if (error = VOP_GETATTR(vp, epp->ep_vap, p->p_ucred, p))
117 goto bad1;
118
119 /* Check mount point */
120 if (vp->v_mount->mnt_flag & MNT_NOEXEC) {
121 error = EACCES;
122 goto bad1;
123 }
124 if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_flag & P_TRACED))
125 epp->ep_vap->va_mode &= ~(VSUID | VSGID);
126
127 /* check access. for root we have to see if any exec bit on */
128 if (error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p))
129 goto bad1;
130 if ((epp->ep_vap->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0) {
131 error = EACCES;
132 goto bad1;
133 }
134
135 /* try to open it */
136 if (error = VOP_OPEN(vp, FREAD, p->p_ucred, p))
137 goto bad1;
138
139 /* now we have the file, get the exec header */
140 if (error = vn_rdwr(UIO_READ, vp, epp->ep_hdr, epp->ep_hdrlen, 0,
141 UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid, p))
142 goto bad2;
143 epp->ep_hdrvalid = epp->ep_hdrlen - resid;
144
145 /*
146 * set up the vmcmds for creation of the process
147 * address space
148 */
149 error = ENOEXEC;
150 for (i = 0; i < nexecs && error != 0; i++) {
151 if (execsw[i].es_check != NULL)
152 error = (*execsw[i].es_check)(p, epp);
153 if (epp->ep_flags & EXEC_DESTR && error != 0)
154 return error;
155 }
156 if (!error) {
157 /* check that entry point is sane */
158 if (epp->ep_entry > VM_MAXUSER_ADDRESS)
159 error = ENOEXEC;
160
161 /* check limits */
162 if ((epp->ep_tsize > MAXTSIZ) ||
163 (epp->ep_dsize > p->p_rlimit[RLIMIT_DATA].rlim_cur))
164 error = ENOMEM;
165
166 if (!error)
167 return (0);
168 }
169
170 /*
171 * free any vmspace-creation commands,
172 * and release their references
173 */
174 kill_vmcmds(&epp->ep_vmcmds);
175
176 bad2:
177 /*
178 * unlock and close the vnode, restore the old one, free the
179 * pathname buf, and punt.
180 */
181 VOP_UNLOCK(vp);
182 vn_close(vp, FREAD, p->p_ucred, p);
183 FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
184 return error;
185
186 bad1:
187 /*
188 * free the namei pathname buffer, and put the vnode
189 * (which we don't yet have open).
190 */
191 FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
192 vput(vp);
193 return error;
194 }
195
196 /*
197 * exec system call
198 */
199 /* ARGSUSED */
200 execve(p, uap, retval)
201 register struct proc *p;
202 register struct execve_args /* {
203 syscallarg(char *) path;
204 syscallarg(char * *) argp;
205 syscallarg(char * *) envp;
206 } */ *uap;
207 register_t *retval;
208 {
209 int error, i;
210 struct exec_package pack;
211 struct nameidata nid;
212 struct vattr attr;
213 struct ucred *cred = p->p_ucred;
214 char *argp;
215 char **cpp, *dp, *sp, *np;
216 long argc, envc;
217 size_t len;
218 char *stack;
219 struct ps_strings arginfo;
220 struct vmspace *vm = p->p_vmspace;
221 char **tmpfap;
222 int szsigcode;
223
224 /*
225 * figure out the maximum size of an exec header, if necessary.
226 * XXX should be able to keep LKM code from modifying exec switch
227 * when we're still using it, but...
228 */
229 if (exec_maxhdrsz == 0) {
230 for (i = 0; i < nexecs; i++)
231 if (execsw[i].es_check != NULL
232 && execsw[i].es_hdrsz > exec_maxhdrsz)
233 exec_maxhdrsz = execsw[i].es_hdrsz;
234 }
235
236 /* init the namei data to point the file user's program name */
237 NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
238
239 /*
240 * initialize the fields of the exec package.
241 */
242 pack.ep_name = SCARG(uap, path);
243 MALLOC(pack.ep_hdr, void *, exec_maxhdrsz, M_EXEC, M_WAITOK);
244 pack.ep_hdrlen = exec_maxhdrsz;
245 pack.ep_hdrvalid = 0;
246 pack.ep_ndp = &nid;
247 pack.ep_setup = NULL; /* assume no setup function */
248 pack.ep_setup_arg = NULL;
249 pack.ep_setup_arglen = 0;
250 pack.ep_vmcmds.evs_cnt = 0;
251 pack.ep_vmcmds.evs_used = 0;
252 pack.ep_vap = &attr;
253 pack.ep_emul = EMUL_NETBSD;
254 pack.ep_flags = 0;
255 #ifdef COPY_SIGCODE
256 pack.ep_sigcode = sigcode;
257 pack.ep_esigcode = esigcode;
258 #endif
259
260 /* see if we can run it. */
261 if (error = check_exec(p, &pack))
262 goto freehdr;
263
264 /* XXX -- THE FOLLOWING SECTION NEEDS MAJOR CLEANUP */
265
266 /* allocate an argument buffer */
267 argp = (char *) kmem_alloc_wait(exec_map, NCARGS);
268 #ifdef DIAGNOSTIC
269 if (argp == (vm_offset_t) 0)
270 panic("execve: argp == NULL");
271 #endif
272 dp = argp;
273 argc = 0;
274
275 /* copy the fake args list, if there's one, freeing it as we go */
276 if (pack.ep_flags & EXEC_HASARGL) {
277 tmpfap = pack.ep_fa;
278 while (*tmpfap != NULL) {
279 char *cp;
280
281 cp = *tmpfap;
282 while (*cp)
283 *dp++ = *cp++;
284 *dp++;
285
286 FREE(*tmpfap, M_EXEC);
287 tmpfap++; argc++;
288 }
289 FREE(pack.ep_fa, M_EXEC);
290 pack.ep_flags &= ~EXEC_HASARGL;
291 }
292
293 /* Now get argv & environment */
294 if (!(cpp = SCARG(uap, argp))) {
295 error = EINVAL;
296 goto bad;
297 }
298
299 if (pack.ep_flags & EXEC_SKIPARG)
300 cpp++;
301
302 while (1) {
303 len = argp + ARG_MAX - dp;
304 if (error = copyin(cpp, &sp, sizeof(sp)))
305 goto bad;
306 if (!sp)
307 break;
308 if (error = copyinstr(sp, dp, len, &len)) {
309 if (error == ENAMETOOLONG)
310 error = E2BIG;
311 goto bad;
312 }
313 dp += len;
314 cpp++;
315 argc++;
316 }
317
318 envc = 0;
319 if (cpp = SCARG(uap, envp)) { /* environment need not be there */
320 while (1) {
321 len = argp + ARG_MAX - dp;
322 if (error = copyin(cpp, &sp, sizeof(sp)))
323 goto bad;
324 if (!sp)
325 break;
326 if (error = copyinstr(sp, dp, len, &len)) {
327 if (error == ENAMETOOLONG)
328 error = E2BIG;
329 goto bad;
330 }
331 dp += len;
332 cpp++;
333 envc++;
334 }
335 }
336
337 dp = (char *) ALIGN(dp);
338
339 #ifdef COPY_SIGCODE
340 szsigcode = pack.ep_esigcode - pack.ep_sigcode;
341 #else
342 szsigcode = 0;
343 #endif
344
345 /* Now check if args & environ fit into new stack */
346 len = ((argc + envc + 2 + pack.ep_setup_arglen) * sizeof(char *) +
347 sizeof(long) + dp + STACKGAPLEN + szsigcode +
348 sizeof(struct ps_strings)) - argp;
349 #ifdef COMPAT_LINUX
350 /* XXXX need this for envp and argv on stack, and sigcode */
351 if (pack.ep_emul == EMUL_LINUX)
352 len += 2 * sizeof (char *);
353 #endif
354 len = ALIGN(len); /* make the stack "safely" aligned */
355
356 if (len > pack.ep_ssize) { /* in effect, compare to initial limit */
357 error = ENOMEM;
358 goto bad;
359 }
360
361 /* adjust "active stack depth" for process VSZ */
362 pack.ep_ssize = len; /* maybe should go elsewhere, but... */
363
364 /* Unmap old program */
365 #ifdef sparc
366 kill_user_windows(p); /* before stack addresses go away */
367 #endif
368 /* Kill shared memory and unmap old program */
369 #ifdef SYSVSHM
370 if (vm->vm_shm)
371 shmexit(p);
372 #endif
373 vm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
374 VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
375
376 /* Now map address space */
377 vm->vm_taddr = (char *) pack.ep_taddr;
378 vm->vm_tsize = btoc(pack.ep_tsize);
379 vm->vm_daddr = (char *) pack.ep_daddr;
380 vm->vm_dsize = btoc(pack.ep_dsize);
381 vm->vm_ssize = btoc(pack.ep_ssize);
382 vm->vm_maxsaddr = (char *) pack.ep_maxsaddr;
383
384 /* create the new process's VM space by running the vmcmds */
385 #ifdef DIAGNOSTIC
386 if (pack.ep_vmcmds.evs_used == 0)
387 panic("execve: no vmcmds");
388 #endif
389 for (i = 0; i < pack.ep_vmcmds.evs_used && !error; i++) {
390 struct exec_vmcmd *vcp;
391
392 vcp = &pack.ep_vmcmds.evs_cmds[i];
393 error = (*vcp->ev_proc)(p, vcp);
394 }
395
396 /* free the vmspace-creation commands, and release their references */
397 kill_vmcmds(&pack.ep_vmcmds);
398
399 /* if an error happened, deallocate and punt */
400 if (error)
401 goto exec_abort;
402
403 /* remember information about the process */
404 arginfo.ps_nargvstr = argc;
405 arginfo.ps_nenvstr = envc;
406
407 /* Now copy argc, args & environ to new stack */
408 stack = (char *) (USRSTACK - len);
409 cpp = (char **) stack;
410
411 if (copyout(&argc, cpp++, sizeof(argc)))
412 goto exec_abort;
413 #ifdef COMPAT_LINUX
414 /* XXXX Linux puts argv and envp on stack too, store argv now */
415 if (pack.ep_emul == EMUL_LINUX) {
416 char **argv_loc = cpp + 2, **stk = (char **) stack;
417
418 if (copyout(&argv_loc, &stk[1], sizeof (argv_loc)))
419 goto exec_abort;
420 /* leave room for envp and argv */
421 cpp += 2;
422 }
423 #endif
424 dp = (char *) (cpp + argc + envc + 2 + pack.ep_setup_arglen);
425 sp = argp;
426 np = 0;
427
428 /* XXX don't copy them out, remap them! */
429 arginfo.ps_argvstr = dp; /* remember location of argv for later */
430 for (; --argc >= 0; sp += len, dp += len) {
431 if (copyout(&dp, cpp++, sizeof(dp)) ||
432 copyoutstr(sp, dp, ARG_MAX, &len))
433 goto exec_abort;
434 }
435 if (copyout(&np, cpp++, sizeof(np)))
436 goto exec_abort;
437
438 #ifdef COMPAT_LINUX
439 /* XXXX Linux puts argv and envp on stack too, store envp now */
440 if (pack.ep_emul == EMUL_LINUX) {
441 char **envp_loc = cpp, **stk = (char **) stack;
442
443 if (copyout(&envp_loc, &stk[2], sizeof (envp_loc)))
444 goto exec_abort;
445 }
446 #endif
447
448 arginfo.ps_envstr = dp; /* remember location of envp for later */
449 for (; --envc >= 0; sp += len, dp += len) {
450 if (copyout(&dp, cpp++, sizeof(dp)) ||
451 copyoutstr(sp, dp, ARG_MAX, &len))
452 goto exec_abort;
453 }
454 if (copyout(&np, cpp++, sizeof(np)))
455 goto exec_abort;
456
457 if (pack.ep_setup != NULL)
458 (*pack.ep_setup)(EXEC_SETUP_ADDARGS, p, &pack, cpp);
459
460 /* copy out the process's ps_strings structure */
461 if (copyout(&arginfo, (char *) PS_STRINGS, sizeof(arginfo)))
462 goto exec_abort;
463
464 #ifdef COPY_SIGCODE
465 /* copy out the process's signal trapoline code */
466 if (copyout((char *) pack.ep_sigcode, ((char *) PS_STRINGS) - szsigcode,
467 szsigcode)) {
468 goto exec_abort;
469 }
470 #endif
471
472 fdcloseexec(p); /* handle close on exec */
473 execsigs(p); /* reset catched signals */
474
475 /* set command name & other accounting info */
476 len = min(nid.ni_cnd.cn_namelen, MAXCOMLEN);
477 bcopy(nid.ni_cnd.cn_nameptr, p->p_comm, len);
478 p->p_comm[len] = 0;
479 p->p_acflag &= ~AFORK;
480
481 /* record proc's vnode, for use by procfs and others */
482 if (p->p_textvp)
483 vrele(p->p_textvp);
484 VREF(pack.ep_vp);
485 p->p_textvp = pack.ep_vp;
486
487 p->p_flag |= P_EXEC;
488 if (p->p_flag & P_PPWAIT) {
489 p->p_flag &= ~P_PPWAIT;
490 wakeup((caddr_t) p->p_pptr);
491 }
492
493 /*
494 * deal with set[ug]id.
495 * MNT_NOEXEC and P_TRACED have already been used to disable s[ug]id.
496 */
497 p->p_flag &= ~P_SUGID;
498 if (((attr.va_mode & VSUID) != 0 &&
499 p->p_ucred->cr_uid != attr.va_uid)
500 || (attr.va_mode & VSGID) != 0 &&
501 p->p_ucred->cr_gid != attr.va_gid) {
502 p->p_ucred = crcopy(cred);
503 #ifdef KTRACE
504 /*
505 * If process is being ktraced, turn off - unless
506 * root set it.
507 */
508 if (p->p_tracep && !(p->p_traceflag & KTRFAC_ROOT)) {
509 vrele(p->p_tracep);
510 p->p_tracep = NULL;
511 p->p_traceflag = 0;
512 }
513 #endif
514 if (attr.va_mode & VSUID)
515 p->p_ucred->cr_uid = attr.va_uid;
516 if (attr.va_mode & VSGID)
517 p->p_ucred->cr_gid = attr.va_gid;
518 p->p_flag |= P_SUGID;
519 }
520 p->p_cred->p_svuid = p->p_ucred->cr_uid;
521 p->p_cred->p_svgid = p->p_ucred->cr_gid;
522
523 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
524
525 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
526 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
527 vput(pack.ep_vp);
528
529 /* setup new registers and do misc. setup. */
530 setregs(p, pack.ep_entry, (u_long) stack, retval);
531 if (pack.ep_setup != NULL)
532 (*pack.ep_setup)(EXEC_SETUP_FINISH, p, &pack, NULL);
533
534 if (p->p_flag & P_TRACED)
535 psignal(p, SIGTRAP);
536
537 p->p_emul = pack.ep_emul;
538 FREE(pack.ep_hdr, M_EXEC);
539 return 0;
540
541 bad:
542 if (pack.ep_setup != NULL)
543 (*pack.ep_setup)(EXEC_SETUP_CLEANUP, p, &pack, dp);
544 /* free the vmspace-creation commands, and release their references */
545 kill_vmcmds(&pack.ep_vmcmds);
546 /* kill any opened file descriptor, if necessary */
547 if (pack.ep_flags & EXEC_HASFD) {
548 pack.ep_flags &= ~EXEC_HASFD;
549 (void) fdclose(p, pack.ep_fd);
550 }
551 /* close and put the exec'd file */
552 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
553 vput(pack.ep_vp);
554 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
555 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
556
557 freehdr:
558 FREE(pack.ep_hdr, M_EXEC);
559 return error;
560
561 exec_abort:
562 /*
563 * the old process doesn't exist anymore. exit gracefully.
564 * get rid of the (new) address space we have created, if any, get rid
565 * of our namei data and vnode, and exit noting failure
566 */
567 vm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
568 VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
569 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
570 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
571 vput(pack.ep_vp);
572 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
573 FREE(pack.ep_hdr, M_EXEC);
574 exit1(p, W_EXITCODE(0, SIGABRT));
575 exit1(p, -1);
576
577 /* NOTREACHED */
578 return 0;
579 }
580