play viking, swing axe, remove deadwood..
[dragonfly.git] / sys / kern / kern_exec.c
CommitLineData
984263bc
MD
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 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD: src/sys/kern/kern_exec.c,v 1.107.2.15 2002/07/30 15:40:46 nectar Exp $
115ccd83 27 * $DragonFly: src/sys/kern/kern_exec.c,v 1.9 2003/08/08 21:47:49 dillon Exp $
984263bc
MD
28 */
29
30#include <sys/param.h>
31#include <sys/systm.h>
32#include <sys/sysproto.h>
33#include <sys/kernel.h>
34#include <sys/mount.h>
35#include <sys/filedesc.h>
36#include <sys/fcntl.h>
37#include <sys/acct.h>
38#include <sys/exec.h>
39#include <sys/imgact.h>
40#include <sys/imgact_elf.h>
41#include <sys/wait.h>
42#include <sys/malloc.h>
43#include <sys/proc.h>
44#include <sys/signalvar.h>
45#include <sys/pioctl.h>
46#include <sys/namei.h>
47#include <sys/sysent.h>
48#include <sys/shm.h>
49#include <sys/sysctl.h>
50#include <sys/vnode.h>
51#include <sys/aio.h>
52
53#include <vm/vm.h>
54#include <vm/vm_param.h>
55#include <sys/lock.h>
56#include <vm/pmap.h>
57#include <vm/vm_page.h>
58#include <vm/vm_map.h>
59#include <vm/vm_kern.h>
60#include <vm/vm_extern.h>
61#include <vm/vm_object.h>
62#include <vm/vm_pager.h>
63
64#include <sys/user.h>
65#include <machine/reg.h>
66
67MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
68
69static register_t *exec_copyout_strings __P((struct image_params *));
70
71/* XXX This should be vm_size_t. */
72static u_long ps_strings = PS_STRINGS;
73SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
74
75/* XXX This should be vm_size_t. */
76static u_long usrstack = USRSTACK;
77SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
78
79u_long ps_arg_cache_limit = PAGE_SIZE / 16;
80SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
81 &ps_arg_cache_limit, 0, "");
82
83int ps_argsopen = 1;
84SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
85
86/*
87 * Each of the items is a pointer to a `const struct execsw', hence the
88 * double pointer here.
89 */
90static const struct execsw **execsw;
91
984263bc
MD
92/*
93 * execve() system call.
94 */
95int
41c20dac 96execve(struct execve_args *uap)
984263bc 97{
dadab5e9
MD
98 struct thread *td = curthread;
99 struct proc *p = td->td_proc;
984263bc
MD
100 struct nameidata nd, *ndp;
101 register_t *stack_base;
102 int error, len, i;
103 struct image_params image_params, *imgp;
104 struct vattr attr;
105 int (*img_first) __P((struct image_params *));
106
dadab5e9 107 KKASSERT(p);
984263bc
MD
108 imgp = &image_params;
109
110 /*
111 * Lock the process and set the P_INEXEC flag to indicate that
112 * it should be left alone until we're done here. This is
113 * necessary to avoid race conditions - e.g. in ptrace() -
114 * that might allow a local user to illicitly obtain elevated
115 * privileges.
116 */
117 p->p_flag |= P_INEXEC;
118
119 /*
120 * Initialize part of the common data
121 */
122 imgp->proc = p;
123 imgp->uap = uap;
124 imgp->attr = &attr;
125 imgp->argc = imgp->envc = 0;
126 imgp->argv0 = NULL;
127 imgp->entry_addr = 0;
128 imgp->vmspace_destroyed = 0;
129 imgp->interpreted = 0;
130 imgp->interpreter_name[0] = '\0';
131 imgp->auxargs = NULL;
132 imgp->vp = NULL;
133 imgp->firstpage = NULL;
134 imgp->ps_strings = 0;
135
136 /*
137 * Allocate temporary demand zeroed space for argument and
138 * environment strings
139 */
140 imgp->stringbase = (char *)kmem_alloc_wait(exec_map, ARG_MAX + PAGE_SIZE);
141 if (imgp->stringbase == NULL) {
142 error = ENOMEM;
143 goto exec_fail;
144 }
145 imgp->stringp = imgp->stringbase;
146 imgp->stringspace = ARG_MAX;
147 imgp->image_header = imgp->stringbase + ARG_MAX;
148
149 /*
150 * Translate the file name. namei() returns a vnode pointer
151 * in ni_vp amoung other things.
152 */
153 ndp = &nd;
154 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
dadab5e9 155 UIO_USERSPACE, uap->fname, td);
984263bc
MD
156
157interpret:
158
159 error = namei(ndp);
160 if (error) {
161 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
162 ARG_MAX + PAGE_SIZE);
163 goto exec_fail;
164 }
165
166 imgp->vp = ndp->ni_vp;
167 imgp->fname = uap->fname;
168
169 /*
170 * Check file permissions (also 'opens' file)
171 */
172 error = exec_check_permissions(imgp);
173 if (error) {
dadab5e9 174 VOP_UNLOCK(imgp->vp, 0, td);
984263bc
MD
175 goto exec_fail_dealloc;
176 }
177
178 error = exec_map_first_page(imgp);
dadab5e9 179 VOP_UNLOCK(imgp->vp, 0, td);
984263bc
MD
180 if (error)
181 goto exec_fail_dealloc;
182
183 /*
184 * If the current process has a special image activator it
185 * wants to try first, call it. For example, emulating shell
186 * scripts differently.
187 */
188 error = -1;
189 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
190 error = img_first(imgp);
191
192 /*
193 * Loop through the list of image activators, calling each one.
194 * An activator returns -1 if there is no match, 0 on success,
195 * and an error otherwise.
196 */
197 for (i = 0; error == -1 && execsw[i]; ++i) {
198 if (execsw[i]->ex_imgact == NULL ||
199 execsw[i]->ex_imgact == img_first) {
200 continue;
201 }
202 error = (*execsw[i]->ex_imgact)(imgp);
203 }
204
205 if (error) {
206 if (error == -1)
207 error = ENOEXEC;
208 goto exec_fail_dealloc;
209 }
210
211 /*
212 * Special interpreter operation, cleanup and loop up to try to
213 * activate the interpreter.
214 */
215 if (imgp->interpreted) {
216 exec_unmap_first_page(imgp);
217 /* free name buffer and old vnode */
218 NDFREE(ndp, NDF_ONLY_PNBUF);
219 vrele(ndp->ni_vp);
220 /* set new name to that of the interpreter */
221 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
dadab5e9 222 UIO_SYSSPACE, imgp->interpreter_name, td);
984263bc
MD
223 goto interpret;
224 }
225
226 /*
227 * Copy out strings (args and env) and initialize stack base
228 */
229 stack_base = exec_copyout_strings(imgp);
230 p->p_vmspace->vm_minsaddr = (char *)stack_base;
231
232 /*
233 * If custom stack fixup routine present for this process
234 * let it do the stack setup.
235 * Else stuff argument count as first item on stack
236 */
237 if (p->p_sysent->sv_fixup)
238 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
239 else
240 suword(--stack_base, imgp->argc);
241
242 /*
243 * For security and other reasons, the file descriptor table cannot
244 * be shared after an exec.
245 */
246 if (p->p_fd->fd_refcnt > 1) {
247 struct filedesc *tmp;
248
249 tmp = fdcopy(p);
250 fdfree(p);
251 p->p_fd = tmp;
252 }
253
254 /*
255 * For security and other reasons, signal handlers cannot
256 * be shared after an exec. The new proces gets a copy of the old
257 * handlers. In execsigs(), the new process will have its signals
258 * reset.
259 */
260 if (p->p_procsig->ps_refcnt > 1) {
261 struct procsig *newprocsig;
262
263 MALLOC(newprocsig, struct procsig *, sizeof(struct procsig),
264 M_SUBPROC, M_WAITOK);
265 bcopy(p->p_procsig, newprocsig, sizeof(*newprocsig));
266 p->p_procsig->ps_refcnt--;
267 p->p_procsig = newprocsig;
268 p->p_procsig->ps_refcnt = 1;
269 if (p->p_sigacts == &p->p_addr->u_sigacts)
270 panic("shared procsig but private sigacts?");
271
272 p->p_addr->u_sigacts = *p->p_sigacts;
273 p->p_sigacts = &p->p_addr->u_sigacts;
274 }
275
276 /* Stop profiling */
277 stopprofclock(p);
278
279 /* close files on exec */
280 fdcloseexec(p);
281
282 /* reset caught signals */
283 execsigs(p);
284
285 /* name this process - nameiexec(p, ndp) */
286 len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
287 bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
288 p->p_comm[len] = 0;
289
290 /*
291 * mark as execed, wakeup the process that vforked (if any) and tell
292 * it that it now has its own resources back
293 */
294 p->p_flag |= P_EXEC;
295 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
296 p->p_flag &= ~P_PPWAIT;
297 wakeup((caddr_t)p->p_pptr);
298 }
299
300 /*
301 * Implement image setuid/setgid.
302 *
303 * Don't honor setuid/setgid if the filesystem prohibits it or if
304 * the process is being traced.
305 */
306 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
307 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
308 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
309 (p->p_flag & P_TRACED) == 0) {
310 /*
311 * Turn off syscall tracing for set-id programs, except for
312 * root. Record any set-id flags first to make sure that
313 * we do not regain any tracing during a possible block.
314 */
41c20dac 315 setsugid();
dadab5e9 316 if (p->p_tracep && suser(td)) {
984263bc
MD
317 struct vnode *vtmp;
318
319 if ((vtmp = p->p_tracep) != NULL) {
320 p->p_tracep = NULL;
321 p->p_traceflag = 0;
322 vrele(vtmp);
323 }
324 }
325 /* Close any file descriptors 0..2 that reference procfs */
326 setugidsafety(p);
327 /* Make sure file descriptors 0..2 are in use. */
328 error = fdcheckstd(p);
329 if (error != 0)
330 goto exec_fail_dealloc;
331 /*
332 * Set the new credentials.
333 */
e9a372eb 334 cratom(&p->p_ucred);
984263bc 335 if (attr.va_mode & VSUID)
41c20dac 336 change_euid(attr.va_uid);
984263bc
MD
337 if (attr.va_mode & VSGID)
338 p->p_ucred->cr_gid = attr.va_gid;
339 } else {
41c20dac
MD
340 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
341 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
984263bc
MD
342 p->p_flag &= ~P_SUGID;
343 }
344
345 /*
346 * Implement correct POSIX saved-id behavior.
347 */
115ccd83
MD
348 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
349 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
350 cratom(&p->p_ucred);
351 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
352 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
353 }
984263bc
MD
354
355 /*
356 * Store the vp for use in procfs
357 */
358 if (p->p_textvp) /* release old reference */
359 vrele(p->p_textvp);
360 VREF(ndp->ni_vp);
361 p->p_textvp = ndp->ni_vp;
362
363 /*
364 * Notify others that we exec'd, and clear the P_INEXEC flag
365 * as we're now a bona fide freshly-execed process.
366 */
367 KNOTE(&p->p_klist, NOTE_EXEC);
368 p->p_flag &= ~P_INEXEC;
369
370 /*
371 * If tracing the process, trap to debugger so breakpoints
372 * can be set before the program executes.
373 */
374 STOPEVENT(p, S_EXEC, 0);
375
376 if (p->p_flag & P_TRACED)
377 psignal(p, SIGTRAP);
378
379 /* clear "fork but no exec" flag, as we _are_ execing */
380 p->p_acflag &= ~AFORK;
381
382 /* Set values passed into the program in registers. */
383 setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
384 imgp->ps_strings);
385
386 /* Free any previous argument cache */
387 if (p->p_args && --p->p_args->ar_ref == 0)
388 FREE(p->p_args, M_PARGS);
389 p->p_args = NULL;
390
391 /* Cache arguments if they fit inside our allowance */
392 i = imgp->endargs - imgp->stringbase;
393 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
394 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
395 M_PARGS, M_WAITOK);
396 p->p_args->ar_ref = 1;
397 p->p_args->ar_length = i;
398 bcopy(imgp->stringbase, p->p_args->ar_args, i);
399 }
400
401exec_fail_dealloc:
402
403 /*
404 * free various allocated resources
405 */
406 if (imgp->firstpage)
407 exec_unmap_first_page(imgp);
408
409 if (imgp->stringbase != NULL)
410 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
411 ARG_MAX + PAGE_SIZE);
412
413 if (imgp->vp) {
414 NDFREE(ndp, NDF_ONLY_PNBUF);
415 vrele(imgp->vp);
416 }
417
418 if (error == 0)
419 return (0);
420
421exec_fail:
422 /* we're done here, clear P_INEXEC */
423 p->p_flag &= ~P_INEXEC;
424 if (imgp->vmspace_destroyed) {
425 /* sorry, no more process anymore. exit gracefully */
41c20dac 426 exit1(W_EXITCODE(0, SIGABRT));
984263bc
MD
427 /* NOT REACHED */
428 return(0);
429 } else {
430 return(error);
431 }
432}
433
434int
41c20dac 435exec_map_first_page(struct image_params *imgp)
984263bc
MD
436{
437 int s, rv, i;
438 int initial_pagein;
439 vm_page_t ma[VM_INITIAL_PAGEIN];
440 vm_object_t object;
441
442
443 if (imgp->firstpage) {
444 exec_unmap_first_page(imgp);
445 }
446
447 VOP_GETVOBJECT(imgp->vp, &object);
448 s = splvm();
449
450 ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
451
452 if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
453 initial_pagein = VM_INITIAL_PAGEIN;
454 if (initial_pagein > object->size)
455 initial_pagein = object->size;
456 for (i = 1; i < initial_pagein; i++) {
457 if ((ma[i] = vm_page_lookup(object, i)) != NULL) {
458 if ((ma[i]->flags & PG_BUSY) || ma[i]->busy)
459 break;
460 if (ma[i]->valid)
461 break;
462 vm_page_busy(ma[i]);
463 } else {
464 ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
465 if (ma[i] == NULL)
466 break;
467 }
468 }
469 initial_pagein = i;
470
471 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
472 ma[0] = vm_page_lookup(object, 0);
473
474 if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) {
475 if (ma[0]) {
476 vm_page_protect(ma[0], VM_PROT_NONE);
477 vm_page_free(ma[0]);
478 }
479 splx(s);
480 return EIO;
481 }
482 }
483
484 vm_page_wire(ma[0]);
485 vm_page_wakeup(ma[0]);
486 splx(s);
487
488 pmap_kenter((vm_offset_t) imgp->image_header, VM_PAGE_TO_PHYS(ma[0]));
489 imgp->firstpage = ma[0];
490
491 return 0;
492}
493
494void
495exec_unmap_first_page(imgp)
496 struct image_params *imgp;
497{
498 if (imgp->firstpage) {
499 pmap_kremove((vm_offset_t) imgp->image_header);
500 vm_page_unwire(imgp->firstpage, 1);
501 imgp->firstpage = NULL;
502 }
503}
504
505/*
506 * Destroy old address space, and allocate a new stack
507 * The new stack is only SGROWSIZ large because it is grown
508 * automatically in trap.c.
509 */
510int
511exec_new_vmspace(imgp)
512 struct image_params *imgp;
513{
514 int error;
515 struct vmspace *vmspace = imgp->proc->p_vmspace;
516 vm_offset_t stack_addr = USRSTACK - maxssiz;
517 vm_map_t map = &vmspace->vm_map;
518
519 imgp->vmspace_destroyed = 1;
520
521 /*
522 * Prevent a pending AIO from modifying the new address space.
523 */
524 aio_proc_rundown(imgp->proc);
525
526 /*
527 * Blow away entire process VM, if address space not shared,
528 * otherwise, create a new VM space so that other threads are
529 * not disrupted
530 */
531 if (vmspace->vm_refcnt == 1) {
fef0fdf2 532 shmexit(vmspace);
984263bc
MD
533 pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
534 vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
535 } else {
536 vmspace_exec(imgp->proc);
537 vmspace = imgp->proc->p_vmspace;
538 map = &vmspace->vm_map;
539 }
540
541 /* Allocate a new stack */
542 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
543 VM_PROT_ALL, VM_PROT_ALL, 0);
544 if (error)
545 return (error);
546
547 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
548 * VM_STACK case, but they are still used to monitor the size of the
549 * process stack so we can check the stack rlimit.
550 */
551 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
552 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
553
554 return(0);
555}
556
557/*
558 * Copy out argument and environment strings from the old process
559 * address space into the temporary string buffer.
560 */
561int
562exec_extract_strings(imgp)
563 struct image_params *imgp;
564{
565 char **argv, **envv;
566 char *argp, *envp;
567 int error;
568 size_t length;
569
570 /*
571 * extract arguments first
572 */
573
574 argv = imgp->uap->argv;
575
576 if (argv) {
577 argp = (caddr_t) (intptr_t) fuword(argv);
578 if (argp == (caddr_t) -1)
579 return (EFAULT);
580 if (argp)
581 argv++;
582 if (imgp->argv0)
583 argp = imgp->argv0;
584 if (argp) {
585 do {
586 if (argp == (caddr_t) -1)
587 return (EFAULT);
588 if ((error = copyinstr(argp, imgp->stringp,
589 imgp->stringspace, &length))) {
590 if (error == ENAMETOOLONG)
591 return(E2BIG);
592 return (error);
593 }
594 imgp->stringspace -= length;
595 imgp->stringp += length;
596 imgp->argc++;
597 } while ((argp = (caddr_t) (intptr_t) fuword(argv++)));
598 }
599 }
600
601 imgp->endargs = imgp->stringp;
602
603 /*
604 * extract environment strings
605 */
606
607 envv = imgp->uap->envv;
608
609 if (envv) {
610 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
611 if (envp == (caddr_t) -1)
612 return (EFAULT);
613 if ((error = copyinstr(envp, imgp->stringp,
614 imgp->stringspace, &length))) {
615 if (error == ENAMETOOLONG)
616 return(E2BIG);
617 return (error);
618 }
619 imgp->stringspace -= length;
620 imgp->stringp += length;
621 imgp->envc++;
622 }
623 }
624
625 return (0);
626}
627
628/*
629 * Copy strings out to the new process address space, constructing
630 * new arg and env vector tables. Return a pointer to the base
631 * so that it can be used as the initial stack pointer.
632 */
633register_t *
634exec_copyout_strings(imgp)
635 struct image_params *imgp;
636{
637 int argc, envc;
638 char **vectp;
639 char *stringp, *destp;
640 register_t *stack_base;
641 struct ps_strings *arginfo;
642 int szsigcode;
643
644 /*
645 * Calculate string base and vector table pointers.
646 * Also deal with signal trampoline code for this exec type.
647 */
648 arginfo = (struct ps_strings *)PS_STRINGS;
649 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
650 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
651 roundup((ARG_MAX - imgp->stringspace), sizeof(char *));
652
653 /*
654 * install sigcode
655 */
656 if (szsigcode)
657 copyout(imgp->proc->p_sysent->sv_sigcode,
658 ((caddr_t)arginfo - szsigcode), szsigcode);
659
660 /*
661 * If we have a valid auxargs ptr, prepare some room
662 * on the stack.
663 */
664 if (imgp->auxargs)
665 /*
666 * The '+ 2' is for the null pointers at the end of each of the
667 * arg and env vector sets, and 'AT_COUNT*2' is room for the
668 * ELF Auxargs data.
669 */
670 vectp = (char **)(destp - (imgp->argc + imgp->envc + 2 +
671 AT_COUNT*2) * sizeof(char*));
672 else
673 /*
674 * The '+ 2' is for the null pointers at the end of each of the
675 * arg and env vector sets
676 */
677 vectp = (char **)
678 (destp - (imgp->argc + imgp->envc + 2) * sizeof(char*));
679
680 /*
681 * vectp also becomes our initial stack base
682 */
683 stack_base = (register_t *)vectp;
684
685 stringp = imgp->stringbase;
686 argc = imgp->argc;
687 envc = imgp->envc;
688
689 /*
690 * Copy out strings - arguments and environment.
691 */
692 copyout(stringp, destp, ARG_MAX - imgp->stringspace);
693
694 /*
695 * Fill in "ps_strings" struct for ps, w, etc.
696 */
697 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
698 suword(&arginfo->ps_nargvstr, argc);
699
700 /*
701 * Fill in argument portion of vector table.
702 */
703 for (; argc > 0; --argc) {
704 suword(vectp++, (long)(intptr_t)destp);
705 while (*stringp++ != 0)
706 destp++;
707 destp++;
708 }
709
710 /* a null vector table pointer seperates the argp's from the envp's */
711 suword(vectp++, 0);
712
713 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
714 suword(&arginfo->ps_nenvstr, envc);
715
716 /*
717 * Fill in environment portion of vector table.
718 */
719 for (; envc > 0; --envc) {
720 suword(vectp++, (long)(intptr_t)destp);
721 while (*stringp++ != 0)
722 destp++;
723 destp++;
724 }
725
726 /* end of vector table is a null pointer */
727 suword(vectp, 0);
728
729 return (stack_base);
730}
731
732/*
733 * Check permissions of file to execute.
734 * Return 0 for success or error code on failure.
735 */
736int
737exec_check_permissions(imgp)
738 struct image_params *imgp;
739{
740 struct proc *p = imgp->proc;
741 struct vnode *vp = imgp->vp;
742 struct vattr *attr = imgp->attr;
dadab5e9 743 struct thread *td = p->p_thread;
984263bc
MD
744 int error;
745
746 /* Get file attributes */
3b568787 747 error = VOP_GETATTR(vp, attr, td);
984263bc
MD
748 if (error)
749 return (error);
750
751 /*
752 * 1) Check if file execution is disabled for the filesystem that this
753 * file resides on.
754 * 2) Insure that at least one execute bit is on - otherwise root
755 * will always succeed, and we don't want to happen unless the
756 * file really is executable.
757 * 3) Insure that the file is a regular file.
758 */
759 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
760 ((attr->va_mode & 0111) == 0) ||
761 (attr->va_type != VREG)) {
762 return (EACCES);
763 }
764
765 /*
766 * Zero length files can't be exec'd
767 */
768 if (attr->va_size == 0)
769 return (ENOEXEC);
770
771 /*
772 * Check for execute permission to file based on current credentials.
773 */
dadab5e9 774 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
984263bc
MD
775 if (error)
776 return (error);
777
778 /*
779 * Check number of open-for-writes on the file and deny execution
780 * if there are any.
781 */
782 if (vp->v_writecount)
783 return (ETXTBSY);
784
785 /*
786 * Call filesystem specific open routine (which does nothing in the
787 * general case).
788 */
dadab5e9 789 error = VOP_OPEN(vp, FREAD, p->p_ucred, td);
984263bc
MD
790 if (error)
791 return (error);
792
793 return (0);
794}
795
796/*
797 * Exec handler registration
798 */
799int
800exec_register(execsw_arg)
801 const struct execsw *execsw_arg;
802{
803 const struct execsw **es, **xs, **newexecsw;
804 int count = 2; /* New slot and trailing NULL */
805
806 if (execsw)
807 for (es = execsw; *es; es++)
808 count++;
809 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
810 if (newexecsw == NULL)
811 return ENOMEM;
812 xs = newexecsw;
813 if (execsw)
814 for (es = execsw; *es; es++)
815 *xs++ = *es;
816 *xs++ = execsw_arg;
817 *xs = NULL;
818 if (execsw)
819 free(execsw, M_TEMP);
820 execsw = newexecsw;
821 return 0;
822}
823
824int
825exec_unregister(execsw_arg)
826 const struct execsw *execsw_arg;
827{
828 const struct execsw **es, **xs, **newexecsw;
829 int count = 1;
830
831 if (execsw == NULL)
832 panic("unregister with no handlers left?\n");
833
834 for (es = execsw; *es; es++) {
835 if (*es == execsw_arg)
836 break;
837 }
838 if (*es == NULL)
839 return ENOENT;
840 for (es = execsw; *es; es++)
841 if (*es != execsw_arg)
842 count++;
843 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
844 if (newexecsw == NULL)
845 return ENOMEM;
846 xs = newexecsw;
847 for (es = execsw; *es; es++)
848 if (*es != execsw_arg)
849 *xs++ = *es;
850 *xs = NULL;
851 if (execsw)
852 free(execsw, M_TEMP);
853 execsw = newexecsw;
854 return 0;
855}