kern_exec.c revision 1.73 1 /* $NetBSD: kern_exec.c,v 1.73 1995/12/09 04:11:00 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 /*
63 * check exec:
64 * given an "executable" described in the exec package's namei info,
65 * see what we can do with it.
66 *
67 * ON ENTRY:
68 * exec package with appropriate namei info
69 * proc pointer of exec'ing proc
70 * NO SELF-LOCKED VNODES
71 *
72 * ON EXIT:
73 * error: nothing held, etc. exec header still allocated.
74 * ok: filled exec package, one locked vnode.
75 *
76 * EXEC SWITCH ENTRY:
77 * Locked vnode to check, exec package, proc.
78 *
79 * EXEC SWITCH EXIT:
80 * ok: return 0, filled exec package, one locked vnode.
81 * error: destructive:
82 * everything deallocated execept exec header.
83 * non-descructive:
84 * error code, locked vnode, exec header unmodified
85 */
86 int
87 check_exec(p, epp)
88 struct proc *p;
89 struct exec_package *epp;
90 {
91 int error, i;
92 struct vnode *vp;
93 char *cp, *ep, *name;
94 struct nameidata *ndp;
95 int resid;
96
97 ndp = epp->ep_ndp;
98 ndp->ni_cnd.cn_nameiop = LOOKUP;
99 ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF | SAVENAME;
100 /* first get the vnode */
101 if (error = namei(ndp))
102 return error;
103 epp->ep_vp = vp = ndp->ni_vp;
104
105 /* check for regular file */
106 if (vp->v_type != VREG) {
107 error = EACCES;
108 goto bad1;
109 }
110
111 /* get attributes */
112 if (error = VOP_GETATTR(vp, epp->ep_vap, p->p_ucred, p))
113 goto bad1;
114
115 /* Check mount point */
116 if (vp->v_mount->mnt_flag & MNT_NOEXEC) {
117 error = EACCES;
118 goto bad1;
119 }
120 if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_flag & P_TRACED))
121 epp->ep_vap->va_mode &= ~(VSUID | VSGID);
122
123 /* check access. for root we have to see if any exec bit on */
124 if (error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p))
125 goto bad1;
126 if ((epp->ep_vap->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0) {
127 error = EACCES;
128 goto bad1;
129 }
130
131 /* try to open it */
132 if (error = VOP_OPEN(vp, FREAD, p->p_ucred, p))
133 goto bad1;
134
135 /* now we have the file, get the exec header */
136 if (error = vn_rdwr(UIO_READ, vp, epp->ep_hdr, epp->ep_hdrlen, 0,
137 UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid, p))
138 goto bad2;
139 epp->ep_hdrvalid = epp->ep_hdrlen - resid;
140
141 /*
142 * set up the vmcmds for creation of the process
143 * address space
144 */
145 error = ENOEXEC;
146 for (i = 0; i < nexecs && error != 0; i++) {
147 int newerror;
148
149 if (execsw[i].es_check == NULL)
150 continue;
151
152 newerror = (*execsw[i].es_check)(p, epp);
153 /* make sure the first "interesting" error code is saved. */
154 if (!newerror || error == ENOEXEC)
155 error = newerror;
156 if (epp->ep_flags & EXEC_DESTR && error != 0)
157 return error;
158 }
159 if (!error) {
160 /* check that entry point is sane */
161 if (epp->ep_entry > VM_MAXUSER_ADDRESS)
162 error = ENOEXEC;
163
164 /* check limits */
165 if ((epp->ep_tsize > MAXTSIZ) ||
166 (epp->ep_dsize > p->p_rlimit[RLIMIT_DATA].rlim_cur))
167 error = ENOMEM;
168
169 if (!error)
170 return (0);
171 }
172
173 /*
174 * free any vmspace-creation commands,
175 * and release their references
176 */
177 kill_vmcmds(&epp->ep_vmcmds);
178
179 bad2:
180 /*
181 * unlock and close the vnode, restore the old one, free the
182 * pathname buf, and punt.
183 */
184 VOP_UNLOCK(vp);
185 vn_close(vp, FREAD, p->p_ucred, p);
186 FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
187 return error;
188
189 bad1:
190 /*
191 * free the namei pathname buffer, and put the vnode
192 * (which we don't yet have open).
193 */
194 FREE(ndp->ni_cnd.cn_pnbuf, M_NAMEI);
195 vput(vp);
196 return error;
197 }
198
199 /*
200 * exec system call
201 */
202 /* ARGSUSED */
203 sys_execve(p, v, retval)
204 register struct proc *p;
205 void *v;
206 register_t *retval;
207 {
208 register struct sys_execve_args /* {
209 syscallarg(char *) path;
210 syscallarg(char * *) argp;
211 syscallarg(char * *) envp;
212 } */ *uap = v;
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;
220 long argc, envc;
221 size_t len;
222 char *stack;
223 struct ps_strings arginfo;
224 struct vmspace *vm = p->p_vmspace;
225 char **tmpfap;
226 int szsigcode;
227 extern struct emul emul_netbsd;
228
229 /*
230 * figure out the maximum size of an exec header, if necessary.
231 * XXX should be able to keep LKM code from modifying exec switch
232 * when we're still using it, but...
233 */
234 if (exec_maxhdrsz == 0) {
235 for (i = 0; i < nexecs; i++)
236 if (execsw[i].es_check != NULL
237 && execsw[i].es_hdrsz > exec_maxhdrsz)
238 exec_maxhdrsz = execsw[i].es_hdrsz;
239 }
240
241 /* init the namei data to point the file user's program name */
242 NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
243
244 /*
245 * initialize the fields of the exec package.
246 */
247 pack.ep_name = SCARG(uap, path);
248 MALLOC(pack.ep_hdr, void *, exec_maxhdrsz, M_EXEC, M_WAITOK);
249 pack.ep_hdrlen = exec_maxhdrsz;
250 pack.ep_hdrvalid = 0;
251 pack.ep_ndp = &nid;
252 pack.ep_emul_arg = NULL;
253 pack.ep_vmcmds.evs_cnt = 0;
254 pack.ep_vmcmds.evs_used = 0;
255 pack.ep_vap = &attr;
256 pack.ep_emul = &emul_netbsd;
257 pack.ep_flags = 0;
258
259 /* see if we can run it. */
260 if (error = check_exec(p, &pack))
261 goto freehdr;
262
263 /* XXX -- THE FOLLOWING SECTION NEEDS MAJOR CLEANUP */
264
265 /* allocate an argument buffer */
266 argp = (char *) kmem_alloc_wait(exec_map, NCARGS);
267 #ifdef DIAGNOSTIC
268 if (argp == (vm_offset_t) 0)
269 panic("execve: argp == NULL");
270 #endif
271 dp = argp;
272 argc = 0;
273
274 /* copy the fake args list, if there's one, freeing it as we go */
275 if (pack.ep_flags & EXEC_HASARGL) {
276 tmpfap = pack.ep_fa;
277 while (*tmpfap != NULL) {
278 char *cp;
279
280 cp = *tmpfap;
281 while (*cp)
282 *dp++ = *cp++;
283 *dp++;
284
285 FREE(*tmpfap, M_EXEC);
286 tmpfap++; argc++;
287 }
288 FREE(pack.ep_fa, M_EXEC);
289 pack.ep_flags &= ~EXEC_HASARGL;
290 }
291
292 /* Now get argv & environment */
293 if (!(cpp = SCARG(uap, argp))) {
294 error = EINVAL;
295 goto bad;
296 }
297
298 if (pack.ep_flags & EXEC_SKIPARG)
299 cpp++;
300
301 while (1) {
302 len = argp + ARG_MAX - dp;
303 if (error = copyin(cpp, &sp, sizeof(sp)))
304 goto bad;
305 if (!sp)
306 break;
307 if (error = copyinstr(sp, dp, len, &len)) {
308 if (error == ENAMETOOLONG)
309 error = E2BIG;
310 goto bad;
311 }
312 dp += len;
313 cpp++;
314 argc++;
315 }
316
317 envc = 0;
318 if (cpp = SCARG(uap, envp)) { /* environment need not be there */
319 while (1) {
320 len = argp + ARG_MAX - dp;
321 if (error = copyin(cpp, &sp, sizeof(sp)))
322 goto bad;
323 if (!sp)
324 break;
325 if (error = copyinstr(sp, dp, len, &len)) {
326 if (error == ENAMETOOLONG)
327 error = E2BIG;
328 goto bad;
329 }
330 dp += len;
331 cpp++;
332 envc++;
333 }
334 }
335
336 dp = (char *) ALIGN(dp);
337
338 szsigcode = pack.ep_emul->e_esigcode - pack.ep_emul->e_sigcode;
339
340 /* Now check if args & environ fit into new stack */
341 len = ((argc + envc + 2 + pack.ep_emul->e_arglen) * sizeof(char *) +
342 sizeof(long) + dp + STACKGAPLEN + szsigcode +
343 sizeof(struct ps_strings)) - argp;
344
345 len = ALIGN(len); /* make the stack "safely" aligned */
346
347 if (len > pack.ep_ssize) { /* in effect, compare to initial limit */
348 error = ENOMEM;
349 goto bad;
350 }
351
352 /* adjust "active stack depth" for process VSZ */
353 pack.ep_ssize = len; /* maybe should go elsewhere, but... */
354
355 /* Unmap old program */
356 #ifdef sparc
357 kill_user_windows(p); /* before stack addresses go away */
358 #endif
359 /* Kill shared memory and unmap old program */
360 #ifdef SYSVSHM
361 if (vm->vm_shm)
362 shmexit(p);
363 #endif
364 vm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
365 VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
366
367 /* Now map address space */
368 vm->vm_taddr = (char *) pack.ep_taddr;
369 vm->vm_tsize = btoc(pack.ep_tsize);
370 vm->vm_daddr = (char *) pack.ep_daddr;
371 vm->vm_dsize = btoc(pack.ep_dsize);
372 vm->vm_ssize = btoc(pack.ep_ssize);
373 vm->vm_maxsaddr = (char *) pack.ep_maxsaddr;
374
375 /* create the new process's VM space by running the vmcmds */
376 #ifdef DIAGNOSTIC
377 if (pack.ep_vmcmds.evs_used == 0)
378 panic("execve: no vmcmds");
379 #endif
380 for (i = 0; i < pack.ep_vmcmds.evs_used && !error; i++) {
381 struct exec_vmcmd *vcp;
382
383 vcp = &pack.ep_vmcmds.evs_cmds[i];
384 error = (*vcp->ev_proc)(p, vcp);
385 }
386
387 /* free the vmspace-creation commands, and release their references */
388 kill_vmcmds(&pack.ep_vmcmds);
389
390 /* if an error happened, deallocate and punt */
391 if (error)
392 goto exec_abort;
393
394 /* remember information about the process */
395 arginfo.ps_nargvstr = argc;
396 arginfo.ps_nenvstr = envc;
397
398 stack = (char *) (USRSTACK - len);
399 /* Now copy argc, args & environ to new stack */
400 if (!(*pack.ep_emul->e_copyargs)(&pack, &arginfo, stack, argp))
401 goto exec_abort;
402
403 /* copy out the process's ps_strings structure */
404 if (copyout(&arginfo, (char *) PS_STRINGS, sizeof(arginfo)))
405 goto exec_abort;
406
407 /* copy out the process's signal trapoline code */
408 if (szsigcode && copyout((char *) pack.ep_emul->e_sigcode,
409 ((char *) PS_STRINGS) - szsigcode,
410 szsigcode))
411 goto exec_abort;
412
413 fdcloseexec(p); /* handle close on exec */
414 execsigs(p); /* reset catched signals */
415
416 /* set command name & other accounting info */
417 len = min(nid.ni_cnd.cn_namelen, MAXCOMLEN);
418 bcopy(nid.ni_cnd.cn_nameptr, p->p_comm, len);
419 p->p_comm[len] = 0;
420 p->p_acflag &= ~AFORK;
421
422 /* record proc's vnode, for use by procfs and others */
423 if (p->p_textvp)
424 vrele(p->p_textvp);
425 VREF(pack.ep_vp);
426 p->p_textvp = pack.ep_vp;
427
428 p->p_flag |= P_EXEC;
429 if (p->p_flag & P_PPWAIT) {
430 p->p_flag &= ~P_PPWAIT;
431 wakeup((caddr_t) p->p_pptr);
432 }
433
434 /*
435 * deal with set[ug]id.
436 * MNT_NOEXEC and P_TRACED have already been used to disable s[ug]id.
437 */
438 p->p_flag &= ~P_SUGID;
439 if (((attr.va_mode & VSUID) != 0 &&
440 p->p_ucred->cr_uid != attr.va_uid)
441 || (attr.va_mode & VSGID) != 0 &&
442 p->p_ucred->cr_gid != attr.va_gid) {
443 p->p_ucred = crcopy(cred);
444 #ifdef KTRACE
445 /*
446 * If process is being ktraced, turn off - unless
447 * root set it.
448 */
449 if (p->p_tracep && !(p->p_traceflag & KTRFAC_ROOT)) {
450 vrele(p->p_tracep);
451 p->p_tracep = NULL;
452 p->p_traceflag = 0;
453 }
454 #endif
455 if (attr.va_mode & VSUID)
456 p->p_ucred->cr_uid = attr.va_uid;
457 if (attr.va_mode & VSGID)
458 p->p_ucred->cr_gid = attr.va_gid;
459 p->p_flag |= P_SUGID;
460 }
461 p->p_cred->p_svuid = p->p_ucred->cr_uid;
462 p->p_cred->p_svgid = p->p_ucred->cr_gid;
463
464 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
465
466 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
467 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
468 vput(pack.ep_vp);
469
470 /* setup new registers and do misc. setup. */
471 (*pack.ep_emul->e_setregs)(p, &pack, (u_long) stack, retval);
472
473 if (p->p_flag & P_TRACED)
474 psignal(p, SIGTRAP);
475
476 p->p_emul = pack.ep_emul;
477 FREE(pack.ep_hdr, M_EXEC);
478
479 #ifdef KTRACE
480 if (KTRPOINT(p, KTR_EMUL))
481 ktremul(p->p_tracep, p->p_emul->e_name);
482 #endif
483 return 0;
484
485 bad:
486 /* free the vmspace-creation commands, and release their references */
487 kill_vmcmds(&pack.ep_vmcmds);
488 /* kill any opened file descriptor, if necessary */
489 if (pack.ep_flags & EXEC_HASFD) {
490 pack.ep_flags &= ~EXEC_HASFD;
491 (void) fdrelease(p, pack.ep_fd);
492 }
493 /* close and put the exec'd file */
494 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
495 vput(pack.ep_vp);
496 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
497 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
498
499 freehdr:
500 FREE(pack.ep_hdr, M_EXEC);
501 return error;
502
503 exec_abort:
504 /*
505 * the old process doesn't exist anymore. exit gracefully.
506 * get rid of the (new) address space we have created, if any, get rid
507 * of our namei data and vnode, and exit noting failure
508 */
509 vm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
510 VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
511 if (pack.ep_emul_arg)
512 FREE(pack.ep_emul_arg, M_TEMP);
513 FREE(nid.ni_cnd.cn_pnbuf, M_NAMEI);
514 VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
515 vput(pack.ep_vp);
516 kmem_free_wakeup(exec_map, (vm_offset_t) argp, NCARGS);
517 FREE(pack.ep_hdr, M_EXEC);
518 exit1(p, W_EXITCODE(0, SIGABRT));
519 exit1(p, -1);
520
521 /* NOTREACHED */
522 return 0;
523 }
524
525
526 void *
527 copyargs(pack, arginfo, stack, argp)
528 struct exec_package *pack;
529 struct ps_strings *arginfo;
530 void *stack;
531 void *argp;
532 {
533 char **cpp = stack;
534 char *dp, *sp;
535 size_t len;
536 void *nullp = NULL;
537 int argc = arginfo->ps_nargvstr;
538 int envc = arginfo->ps_nenvstr;
539
540 if (copyout(&argc, cpp++, sizeof(argc)))
541 return NULL;
542
543 dp = (char *) (cpp + argc + envc + 2 + pack->ep_emul->e_arglen);
544 sp = argp;
545
546 /* XXX don't copy them out, remap them! */
547 arginfo->ps_argvstr = cpp; /* remember location of argv for later */
548
549 for (; --argc >= 0; sp += len, dp += len)
550 if (copyout(&dp, cpp++, sizeof(dp)) ||
551 copyoutstr(sp, dp, ARG_MAX, &len))
552 return NULL;
553
554 if (copyout(&nullp, cpp++, sizeof(nullp)))
555 return NULL;
556
557 arginfo->ps_envstr = cpp; /* remember location of envp for later */
558
559 for (; --envc >= 0; sp += len, dp += len)
560 if (copyout(&dp, cpp++, sizeof(dp)) ||
561 copyoutstr(sp, dp, ARG_MAX, &len))
562 return NULL;
563
564 if (copyout(&nullp, cpp++, sizeof(nullp)))
565 return NULL;
566
567 return cpp;
568 }
569