Remove extra semi colin and a few remove redundent checks.
[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 $
77153250 27 * $DragonFly: src/sys/kern/kern_exec.c,v 1.32 2005/04/20 16:37:09 cpressey 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>
2bd9d75c 41#include <sys/kern_syscall.h>
984263bc
MD
42#include <sys/wait.h>
43#include <sys/malloc.h>
44#include <sys/proc.h>
45#include <sys/signalvar.h>
46#include <sys/pioctl.h>
fad57d0e 47#include <sys/nlookup.h>
18f40545 48#include <sys/sfbuf.h>
984263bc
MD
49#include <sys/sysent.h>
50#include <sys/shm.h>
51#include <sys/sysctl.h>
52#include <sys/vnode.h>
a6f89c72 53#include <sys/vmmeter.h>
984263bc
MD
54#include <sys/aio.h>
55
56#include <vm/vm.h>
57#include <vm/vm_param.h>
58#include <sys/lock.h>
59#include <vm/pmap.h>
60#include <vm/vm_page.h>
61#include <vm/vm_map.h>
62#include <vm/vm_kern.h>
63#include <vm/vm_extern.h>
64#include <vm/vm_object.h>
65#include <vm/vm_pager.h>
66
67#include <sys/user.h>
68#include <machine/reg.h>
69
5fd012e0
MD
70#include <sys/thread2.h>
71
984263bc
MD
72MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
73
402ed7e1 74static register_t *exec_copyout_strings (struct image_params *);
984263bc
MD
75
76/* XXX This should be vm_size_t. */
77static u_long ps_strings = PS_STRINGS;
78SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
79
80/* XXX This should be vm_size_t. */
81static u_long usrstack = USRSTACK;
82SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
83
84u_long ps_arg_cache_limit = PAGE_SIZE / 16;
85SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
86 &ps_arg_cache_limit, 0, "");
87
88int ps_argsopen = 1;
89SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
90
2bd9d75c
DRJ
91void print_execve_args(struct image_args *args);
92int debug_execve_args = 0;
93SYSCTL_INT(_kern, OID_AUTO, debug_execve_args, CTLFLAG_RW, &debug_execve_args,
94 0, "");
95
96void
97print_execve_args(struct image_args *args)
98{
99 char *cp;
100 int ndx;
101
102 cp = args->begin_argv;
103 for (ndx = 0; ndx < args->argc; ndx++) {
104 printf("\targv[%d]: %s\n", ndx, cp);
105 while (*cp++ != '\0');
106 }
107 for (ndx = 0; ndx < args->envc; ndx++) {
108 printf("\tenvv[%d]: %s\n", ndx, cp);
109 while (*cp++ != '\0');
110 }
111}
112
984263bc
MD
113/*
114 * Each of the items is a pointer to a `const struct execsw', hence the
115 * double pointer here.
116 */
117static const struct execsw **execsw;
118
984263bc 119int
fad57d0e 120kern_execve(struct nlookupdata *nd, struct image_args *args)
984263bc 121{
dadab5e9
MD
122 struct thread *td = curthread;
123 struct proc *p = td->td_proc;
984263bc
MD
124 register_t *stack_base;
125 int error, len, i;
126 struct image_params image_params, *imgp;
127 struct vattr attr;
402ed7e1 128 int (*img_first) (struct image_params *);
984263bc 129
2bd9d75c
DRJ
130 if (debug_execve_args) {
131 printf("%s()\n", __func__);
132 print_execve_args(args);
133 }
134
dadab5e9 135 KKASSERT(p);
984263bc
MD
136 imgp = &image_params;
137
138 /*
139 * Lock the process and set the P_INEXEC flag to indicate that
140 * it should be left alone until we're done here. This is
141 * necessary to avoid race conditions - e.g. in ptrace() -
142 * that might allow a local user to illicitly obtain elevated
143 * privileges.
144 */
145 p->p_flag |= P_INEXEC;
146
147 /*
148 * Initialize part of the common data
149 */
150 imgp->proc = p;
2bd9d75c 151 imgp->args = args;
984263bc 152 imgp->attr = &attr;
984263bc 153 imgp->entry_addr = 0;
29802dbb 154 imgp->resident = 0;
984263bc
MD
155 imgp->vmspace_destroyed = 0;
156 imgp->interpreted = 0;
2abfe22f 157 imgp->interpreter_name[0] = 0;
984263bc
MD
158 imgp->auxargs = NULL;
159 imgp->vp = NULL;
160 imgp->firstpage = NULL;
161 imgp->ps_strings = 0;
18f40545 162 imgp->image_header = NULL;
2bd9d75c
DRJ
163
164interpret:
984263bc
MD
165
166 /*
fad57d0e
MD
167 * Translate the file name to a vnode. Unlock the cache entry to
168 * improve parallelism for programs exec'd in parallel.
984263bc 169 */
fad57d0e
MD
170 if ((error = nlookup(nd)) != 0)
171 goto exec_fail;
172 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
173 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
174 nd->nl_flags &= ~NLC_NCPISLOCKED;
175 cache_unlock(nd->nl_ncp);
176 if (error)
984263bc 177 goto exec_fail;
984263bc
MD
178
179 /*
180 * Check file permissions (also 'opens' file)
181 */
182 error = exec_check_permissions(imgp);
183 if (error) {
5fd012e0 184 VOP_UNLOCK(imgp->vp, 0, td);
984263bc
MD
185 goto exec_fail_dealloc;
186 }
187
188 error = exec_map_first_page(imgp);
5fd012e0 189 VOP_UNLOCK(imgp->vp, 0, td);
984263bc
MD
190 if (error)
191 goto exec_fail_dealloc;
192
2bd9d75c
DRJ
193 if (debug_execve_args && imgp->interpreted) {
194 printf(" target is interpreted -- recursive pass\n");
195 printf(" interpreter: %s\n", imgp->interpreter_name);
196 print_execve_args(args);
197 }
198
984263bc
MD
199 /*
200 * If the current process has a special image activator it
201 * wants to try first, call it. For example, emulating shell
202 * scripts differently.
203 */
204 error = -1;
205 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
206 error = img_first(imgp);
207
29802dbb
MD
208 /*
209 * If the vnode has a registered vmspace, exec the vmspace
210 */
211 if (error == -1 && imgp->vp->v_resident) {
212 error = exec_resident_imgact(imgp);
213 }
214
984263bc
MD
215 /*
216 * Loop through the list of image activators, calling each one.
217 * An activator returns -1 if there is no match, 0 on success,
218 * and an error otherwise.
219 */
220 for (i = 0; error == -1 && execsw[i]; ++i) {
221 if (execsw[i]->ex_imgact == NULL ||
222 execsw[i]->ex_imgact == img_first) {
223 continue;
224 }
225 error = (*execsw[i]->ex_imgact)(imgp);
226 }
227
228 if (error) {
229 if (error == -1)
230 error = ENOEXEC;
231 goto exec_fail_dealloc;
232 }
233
234 /*
235 * Special interpreter operation, cleanup and loop up to try to
236 * activate the interpreter.
237 */
238 if (imgp->interpreted) {
239 exec_unmap_first_page(imgp);
fad57d0e
MD
240 nlookup_done(nd);
241 vrele(imgp->vp);
242 imgp->vp = NULL;
243 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
244 NLC_FOLLOW);
245 if (error)
246 goto exec_fail;
984263bc
MD
247 goto interpret;
248 }
249
250 /*
251 * Copy out strings (args and env) and initialize stack base
252 */
253 stack_base = exec_copyout_strings(imgp);
254 p->p_vmspace->vm_minsaddr = (char *)stack_base;
255
256 /*
257 * If custom stack fixup routine present for this process
29802dbb
MD
258 * let it do the stack setup. If we are running a resident
259 * image there is no auxinfo or other image activator context
260 * so don't try to add fixups to the stack.
261 *
984263bc
MD
262 * Else stuff argument count as first item on stack
263 */
29802dbb 264 if (p->p_sysent->sv_fixup && imgp->resident == 0)
984263bc
MD
265 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
266 else
2bd9d75c 267 suword(--stack_base, imgp->args->argc);
984263bc
MD
268
269 /*
270 * For security and other reasons, the file descriptor table cannot
271 * be shared after an exec.
272 */
273 if (p->p_fd->fd_refcnt > 1) {
274 struct filedesc *tmp;
275
276 tmp = fdcopy(p);
277 fdfree(p);
278 p->p_fd = tmp;
279 }
280
281 /*
282 * For security and other reasons, signal handlers cannot
283 * be shared after an exec. The new proces gets a copy of the old
284 * handlers. In execsigs(), the new process will have its signals
285 * reset.
286 */
287 if (p->p_procsig->ps_refcnt > 1) {
288 struct procsig *newprocsig;
289
290 MALLOC(newprocsig, struct procsig *, sizeof(struct procsig),
291 M_SUBPROC, M_WAITOK);
292 bcopy(p->p_procsig, newprocsig, sizeof(*newprocsig));
293 p->p_procsig->ps_refcnt--;
294 p->p_procsig = newprocsig;
295 p->p_procsig->ps_refcnt = 1;
296 if (p->p_sigacts == &p->p_addr->u_sigacts)
297 panic("shared procsig but private sigacts?");
298
299 p->p_addr->u_sigacts = *p->p_sigacts;
300 p->p_sigacts = &p->p_addr->u_sigacts;
301 }
302
303 /* Stop profiling */
304 stopprofclock(p);
305
306 /* close files on exec */
307 fdcloseexec(p);
308
309 /* reset caught signals */
310 execsigs(p);
311
312 /* name this process - nameiexec(p, ndp) */
fad57d0e
MD
313 len = min(nd->nl_ncp->nc_nlen, MAXCOMLEN);
314 bcopy(nd->nl_ncp->nc_name, p->p_comm, len);
984263bc
MD
315 p->p_comm[len] = 0;
316
317 /*
318 * mark as execed, wakeup the process that vforked (if any) and tell
319 * it that it now has its own resources back
320 */
321 p->p_flag |= P_EXEC;
322 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
323 p->p_flag &= ~P_PPWAIT;
324 wakeup((caddr_t)p->p_pptr);
325 }
326
327 /*
328 * Implement image setuid/setgid.
329 *
330 * Don't honor setuid/setgid if the filesystem prohibits it or if
331 * the process is being traced.
332 */
333 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
334 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
335 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
336 (p->p_flag & P_TRACED) == 0) {
337 /*
338 * Turn off syscall tracing for set-id programs, except for
339 * root. Record any set-id flags first to make sure that
340 * we do not regain any tracing during a possible block.
341 */
41c20dac 342 setsugid();
dadab5e9 343 if (p->p_tracep && suser(td)) {
984263bc
MD
344 struct vnode *vtmp;
345
346 if ((vtmp = p->p_tracep) != NULL) {
347 p->p_tracep = NULL;
348 p->p_traceflag = 0;
349 vrele(vtmp);
350 }
351 }
352 /* Close any file descriptors 0..2 that reference procfs */
353 setugidsafety(p);
354 /* Make sure file descriptors 0..2 are in use. */
355 error = fdcheckstd(p);
356 if (error != 0)
357 goto exec_fail_dealloc;
358 /*
359 * Set the new credentials.
360 */
e9a372eb 361 cratom(&p->p_ucred);
984263bc 362 if (attr.va_mode & VSUID)
41c20dac 363 change_euid(attr.va_uid);
984263bc
MD
364 if (attr.va_mode & VSGID)
365 p->p_ucred->cr_gid = attr.va_gid;
98a7f915
MD
366
367 /*
368 * Clear local varsym variables
369 */
370 varsymset_clean(&p->p_varsymset);
984263bc 371 } else {
41c20dac
MD
372 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
373 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
984263bc
MD
374 p->p_flag &= ~P_SUGID;
375 }
376
377 /*
378 * Implement correct POSIX saved-id behavior.
379 */
115ccd83
MD
380 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
381 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
382 cratom(&p->p_ucred);
383 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
384 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
385 }
984263bc
MD
386
387 /*
388 * Store the vp for use in procfs
389 */
390 if (p->p_textvp) /* release old reference */
391 vrele(p->p_textvp);
fad57d0e
MD
392 p->p_textvp = imgp->vp;
393 vref(p->p_textvp);
984263bc
MD
394
395 /*
396 * Notify others that we exec'd, and clear the P_INEXEC flag
397 * as we're now a bona fide freshly-execed process.
398 */
399 KNOTE(&p->p_klist, NOTE_EXEC);
400 p->p_flag &= ~P_INEXEC;
401
402 /*
403 * If tracing the process, trap to debugger so breakpoints
404 * can be set before the program executes.
405 */
406 STOPEVENT(p, S_EXEC, 0);
407
408 if (p->p_flag & P_TRACED)
409 psignal(p, SIGTRAP);
410
411 /* clear "fork but no exec" flag, as we _are_ execing */
412 p->p_acflag &= ~AFORK;
413
414 /* Set values passed into the program in registers. */
415 setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
416 imgp->ps_strings);
417
418 /* Free any previous argument cache */
419 if (p->p_args && --p->p_args->ar_ref == 0)
420 FREE(p->p_args, M_PARGS);
421 p->p_args = NULL;
422
423 /* Cache arguments if they fit inside our allowance */
2bd9d75c 424 i = imgp->args->begin_envv - imgp->args->begin_argv;
984263bc
MD
425 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
426 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
427 M_PARGS, M_WAITOK);
428 p->p_args->ar_ref = 1;
429 p->p_args->ar_length = i;
2bd9d75c 430 bcopy(imgp->args->begin_argv, p->p_args->ar_args, i);
984263bc
MD
431 }
432
433exec_fail_dealloc:
434
435 /*
436 * free various allocated resources
437 */
438 if (imgp->firstpage)
439 exec_unmap_first_page(imgp);
440
984263bc 441 if (imgp->vp) {
984263bc 442 vrele(imgp->vp);
fad57d0e 443 imgp->vp = NULL;
984263bc
MD
444 }
445
a6f89c72
DR
446 if (error == 0) {
447 ++mycpu->gd_cnt.v_exec;
984263bc 448 return (0);
a6f89c72 449 }
984263bc
MD
450
451exec_fail:
452 /* we're done here, clear P_INEXEC */
453 p->p_flag &= ~P_INEXEC;
454 if (imgp->vmspace_destroyed) {
455 /* sorry, no more process anymore. exit gracefully */
41c20dac 456 exit1(W_EXITCODE(0, SIGABRT));
984263bc
MD
457 /* NOT REACHED */
458 return(0);
459 } else {
460 return(error);
461 }
462}
463
2bd9d75c
DRJ
464/*
465 * execve() system call.
466 */
467int
468execve(struct execve_args *uap)
469{
fad57d0e 470 struct nlookupdata nd;
2bd9d75c
DRJ
471 struct image_args args;
472 int error;
473
fad57d0e
MD
474 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
475 if (error == 0) {
476 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
477 uap->argv, uap->envv);
478 }
2abfe22f
MD
479 if (error == 0)
480 error = kern_execve(&nd, &args);
fad57d0e 481 nlookup_done(&nd);
2bd9d75c
DRJ
482 exec_free_args(&args);
483
484 /*
485 * The syscall result is returned in registers to the new program.
486 * Linux will register %edx as an atexit function and we must be
487 * sure to set it to 0. XXX
488 */
489 if (error == 0)
490 uap->sysmsg_result64 = 0;
491
492 return (error);
493}
494
984263bc 495int
41c20dac 496exec_map_first_page(struct image_params *imgp)
984263bc 497{
5fd012e0 498 int rv, i;
984263bc
MD
499 int initial_pagein;
500 vm_page_t ma[VM_INITIAL_PAGEIN];
06ecca5a 501 vm_page_t m;
984263bc 502 vm_object_t object;
fad57d0e 503 int error;
984263bc 504
18f40545 505 if (imgp->firstpage)
984263bc 506 exec_unmap_first_page(imgp);
984263bc 507
fad57d0e
MD
508 /*
509 * XXX the callers should really use vn_open so we don't have to
510 * do this junk.
511 */
512 if ((error = VOP_GETVOBJECT(imgp->vp, &object)) != 0) {
513 if (vn_canvmio(imgp->vp) == TRUE) {
514 error = vfs_object_create(imgp->vp, curthread);
515 if (error == 0)
516 error = VOP_GETVOBJECT(imgp->vp, &object);
517 }
518 }
519 if (error)
520 return (EIO);
984263bc 521
06ecca5a
MD
522 /*
523 * We shouldn't need protection for vm_page_grab() but we certainly
524 * need it for the lookup loop below (lookup/busy race), since
525 * an interrupt can unbusy and free the page before our busy check.
526 */
5fd012e0 527 crit_enter();
06ecca5a 528 m = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
984263bc 529
06ecca5a
MD
530 if ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
531 ma[0] = m;
984263bc
MD
532 initial_pagein = VM_INITIAL_PAGEIN;
533 if (initial_pagein > object->size)
534 initial_pagein = object->size;
535 for (i = 1; i < initial_pagein; i++) {
06ecca5a
MD
536 if ((m = vm_page_lookup(object, i)) != NULL) {
537 if ((m->flags & PG_BUSY) || m->busy)
984263bc 538 break;
06ecca5a 539 if (m->valid)
984263bc 540 break;
06ecca5a 541 vm_page_busy(m);
984263bc 542 } else {
06ecca5a
MD
543 m = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
544 if (m == NULL)
984263bc
MD
545 break;
546 }
06ecca5a 547 ma[i] = m;
984263bc
MD
548 }
549 initial_pagein = i;
550
06ecca5a
MD
551 /*
552 * get_pages unbusies all the requested pages except the
553 * primary page (at index 0 in this case).
554 */
984263bc 555 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
06ecca5a 556 m = vm_page_lookup(object, 0);
984263bc 557
06ecca5a
MD
558 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
559 if (m) {
560 vm_page_protect(m, VM_PROT_NONE);
561 vm_page_free(m);
984263bc 562 }
5fd012e0 563 crit_exit();
984263bc
MD
564 return EIO;
565 }
566 }
06ecca5a
MD
567 vm_page_hold(m);
568 vm_page_wakeup(m); /* unbusy the page */
5fd012e0 569 crit_exit();
984263bc 570
4f1640d6 571 imgp->firstpage = sf_buf_alloc(m, SFB_CPUPRIVATE);
18f40545 572 imgp->image_header = (void *)sf_buf_kva(imgp->firstpage);
984263bc
MD
573
574 return 0;
575}
576
577void
77153250 578exec_unmap_first_page(struct image_params *imgp)
984263bc 579{
18f40545
MD
580 vm_page_t m;
581
5fd012e0 582 crit_enter();
18f40545
MD
583 if (imgp->firstpage != NULL) {
584 m = sf_buf_page(imgp->firstpage);
585 sf_buf_free(imgp->firstpage);
984263bc 586 imgp->firstpage = NULL;
18f40545 587 imgp->image_header = NULL;
e4ba7818 588 vm_page_unhold(m);
984263bc 589 }
5fd012e0 590 crit_exit();
984263bc
MD
591}
592
593/*
594 * Destroy old address space, and allocate a new stack
595 * The new stack is only SGROWSIZ large because it is grown
596 * automatically in trap.c.
597 */
598int
29802dbb 599exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
984263bc
MD
600{
601 int error;
602 struct vmspace *vmspace = imgp->proc->p_vmspace;
603 vm_offset_t stack_addr = USRSTACK - maxssiz;
29802dbb 604 vm_map_t map;
984263bc
MD
605
606 imgp->vmspace_destroyed = 1;
607
608 /*
609 * Prevent a pending AIO from modifying the new address space.
610 */
611 aio_proc_rundown(imgp->proc);
612
613 /*
614 * Blow away entire process VM, if address space not shared,
615 * otherwise, create a new VM space so that other threads are
29802dbb
MD
616 * not disrupted. If we are execing a resident vmspace we
617 * create a duplicate of it and remap the stack.
239b4df9
MD
618 *
619 * The exitingcnt test is not strictly necessary but has been
620 * included for code sanity (to make the code more deterministic).
984263bc 621 */
29802dbb
MD
622 map = &vmspace->vm_map;
623 if (vmcopy) {
624 vmspace_exec(imgp->proc, vmcopy);
625 vmspace = imgp->proc->p_vmspace;
626 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
627 map = &vmspace->vm_map;
239b4df9 628 } else if (vmspace->vm_refcnt == 1 && vmspace->vm_exitingcnt == 0) {
fef0fdf2 629 shmexit(vmspace);
239b4df9
MD
630 if (vmspace->vm_upcalls)
631 upc_release(vmspace, imgp->proc);
984263bc
MD
632 pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
633 vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
634 } else {
29802dbb 635 vmspace_exec(imgp->proc, NULL);
984263bc
MD
636 vmspace = imgp->proc->p_vmspace;
637 map = &vmspace->vm_map;
638 }
639
640 /* Allocate a new stack */
641 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
642 VM_PROT_ALL, VM_PROT_ALL, 0);
643 if (error)
644 return (error);
645
646 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
647 * VM_STACK case, but they are still used to monitor the size of the
648 * process stack so we can check the stack rlimit.
649 */
650 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
651 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
652
653 return(0);
654}
655
656/*
657 * Copy out argument and environment strings from the old process
658 * address space into the temporary string buffer.
659 */
660int
2bd9d75c 661exec_copyin_args(struct image_args *args, char *fname,
2abfe22f 662 enum exec_path_segflg segflg, char **argv, char **envv)
984263bc 663{
984263bc 664 char *argp, *envp;
2bd9d75c 665 int error = 0;
984263bc
MD
666 size_t length;
667
2abfe22f 668 bzero(args, sizeof(*args));
2bd9d75c
DRJ
669 args->buf = (char *) kmem_alloc_wait(exec_map, PATH_MAX + ARG_MAX);
670 if (args->buf == NULL)
671 return (ENOMEM);
672 args->begin_argv = args->buf;
673 args->endp = args->begin_argv;
674 args->space = ARG_MAX;
2bd9d75c
DRJ
675
676 args->fname = args->buf + ARG_MAX;
677
984263bc 678 /*
2bd9d75c 679 * Copy the file name.
984263bc 680 */
2bd9d75c
DRJ
681 if (segflg == PATH_SYSSPACE) {
682 error = copystr(fname, args->fname, PATH_MAX, &length);
683 } else if (segflg == PATH_USERSPACE) {
684 error = copyinstr(fname, args->fname, PATH_MAX, &length);
685 }
984263bc 686
2bd9d75c 687 /*
66be6566
MD
688 * Extract argument strings. argv may not be NULL. The argv
689 * array is terminated by a NULL entry. We special-case the
690 * situation where argv[0] is NULL by passing { filename, NULL }
691 * to the new program to guarentee that the interpreter knows what
692 * file to open in case we exec an interpreted file. Note that
693 * a NULL argv[0] terminates the argv[] array.
694 *
695 * XXX the special-casing of argv[0] is historical and needs to be
696 * revisited.
2bd9d75c 697 */
66be6566
MD
698 if (argv == NULL)
699 error = EFAULT;
700 if (error == 0) {
701 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
702 if (argp == (caddr_t)-1) {
2bd9d75c 703 error = EFAULT;
66be6566 704 break;
2bd9d75c
DRJ
705 }
706 error = copyinstr(argp, args->endp,
2abfe22f 707 args->space, &length);
66be6566
MD
708 if (error) {
709 if (error == ENAMETOOLONG)
710 error = E2BIG;
711 break;
712 }
2bd9d75c
DRJ
713 args->space -= length;
714 args->endp += length;
715 args->argc++;
984263bc 716 }
66be6566
MD
717 if (args->argc == 0 && error == 0) {
718 length = strlen(args->fname) + 1;
719 if (length > args->space) {
720 error = E2BIG;
721 } else {
722 bcopy(args->fname, args->endp, length);
723 args->space -= length;
724 args->endp += length;
725 args->argc++;
726 }
727 }
984263bc
MD
728 }
729
2bd9d75c 730 args->begin_envv = args->endp;
984263bc
MD
731
732 /*
66be6566 733 * extract environment strings. envv may be NULL.
984263bc 734 */
2abfe22f 735 if (envv && error == 0) {
984263bc 736 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
2bd9d75c
DRJ
737 if (envp == (caddr_t) -1) {
738 error = EFAULT;
66be6566 739 break;
984263bc 740 }
2bd9d75c
DRJ
741 error = copyinstr(envp, args->endp, args->space,
742 &length);
66be6566
MD
743 if (error) {
744 if (error == ENAMETOOLONG)
745 error = E2BIG;
746 break;
747 }
2bd9d75c
DRJ
748 args->space -= length;
749 args->endp += length;
750 args->envc++;
984263bc
MD
751 }
752 }
2bd9d75c
DRJ
753 return (error);
754}
755
756void
757exec_free_args(struct image_args *args)
758{
2abfe22f
MD
759 if (args->buf) {
760 kmem_free_wakeup(exec_map,
761 (vm_offset_t)args->buf, PATH_MAX + ARG_MAX);
762 args->buf = NULL;
763 }
984263bc
MD
764}
765
766/*
767 * Copy strings out to the new process address space, constructing
768 * new arg and env vector tables. Return a pointer to the base
769 * so that it can be used as the initial stack pointer.
770 */
771register_t *
2bd9d75c 772exec_copyout_strings(struct image_params *imgp)
984263bc
MD
773{
774 int argc, envc;
775 char **vectp;
776 char *stringp, *destp;
777 register_t *stack_base;
778 struct ps_strings *arginfo;
779 int szsigcode;
780
781 /*
782 * Calculate string base and vector table pointers.
783 * Also deal with signal trampoline code for this exec type.
784 */
785 arginfo = (struct ps_strings *)PS_STRINGS;
786 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
787 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
2bd9d75c 788 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
984263bc
MD
789
790 /*
791 * install sigcode
792 */
793 if (szsigcode)
794 copyout(imgp->proc->p_sysent->sv_sigcode,
2bd9d75c 795 ((caddr_t)arginfo - szsigcode), szsigcode);
984263bc
MD
796
797 /*
798 * If we have a valid auxargs ptr, prepare some room
799 * on the stack.
7ec1918b 800 *
984263bc
MD
801 * The '+ 2' is for the null pointers at the end of each of the
802 * arg and env vector sets, and 'AT_COUNT*2' is room for the
803 * ELF Auxargs data.
804 */
7ec1918b
MD
805 if (imgp->auxargs) {
806 vectp = (char **)(destp - (imgp->args->argc +
807 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
808 } else {
2bd9d75c 809 vectp = (char **)(destp - (imgp->args->argc +
7ec1918b
MD
810 imgp->args->envc + 2) * sizeof(char*));
811 }
812
984263bc 813 /*
31bc42c8
MD
814 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
815 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
984263bc 816 */
984263bc
MD
817
818 /*
819 * vectp also becomes our initial stack base
820 */
821 stack_base = (register_t *)vectp;
822
2bd9d75c
DRJ
823 stringp = imgp->args->begin_argv;
824 argc = imgp->args->argc;
825 envc = imgp->args->envc;
984263bc
MD
826
827 /*
828 * Copy out strings - arguments and environment.
829 */
2bd9d75c 830 copyout(stringp, destp, ARG_MAX - imgp->args->space);
984263bc
MD
831
832 /*
833 * Fill in "ps_strings" struct for ps, w, etc.
834 */
835 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
836 suword(&arginfo->ps_nargvstr, argc);
837
838 /*
839 * Fill in argument portion of vector table.
840 */
841 for (; argc > 0; --argc) {
842 suword(vectp++, (long)(intptr_t)destp);
843 while (*stringp++ != 0)
844 destp++;
845 destp++;
846 }
847
7a44d1cb 848 /* a null vector table pointer separates the argp's from the envp's */
984263bc
MD
849 suword(vectp++, 0);
850
851 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
852 suword(&arginfo->ps_nenvstr, envc);
853
854 /*
855 * Fill in environment portion of vector table.
856 */
857 for (; envc > 0; --envc) {
858 suword(vectp++, (long)(intptr_t)destp);
859 while (*stringp++ != 0)
860 destp++;
861 destp++;
862 }
863
864 /* end of vector table is a null pointer */
865 suword(vectp, 0);
866
867 return (stack_base);
868}
869
870/*
871 * Check permissions of file to execute.
872 * Return 0 for success or error code on failure.
873 */
874int
77153250 875exec_check_permissions(struct image_params *imgp)
984263bc
MD
876{
877 struct proc *p = imgp->proc;
878 struct vnode *vp = imgp->vp;
879 struct vattr *attr = imgp->attr;
dadab5e9 880 struct thread *td = p->p_thread;
984263bc
MD
881 int error;
882
883 /* Get file attributes */
3b568787 884 error = VOP_GETATTR(vp, attr, td);
984263bc
MD
885 if (error)
886 return (error);
887
888 /*
889 * 1) Check if file execution is disabled for the filesystem that this
890 * file resides on.
891 * 2) Insure that at least one execute bit is on - otherwise root
892 * will always succeed, and we don't want to happen unless the
893 * file really is executable.
894 * 3) Insure that the file is a regular file.
895 */
896 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
897 ((attr->va_mode & 0111) == 0) ||
898 (attr->va_type != VREG)) {
899 return (EACCES);
900 }
901
902 /*
903 * Zero length files can't be exec'd
904 */
905 if (attr->va_size == 0)
906 return (ENOEXEC);
907
908 /*
909 * Check for execute permission to file based on current credentials.
910 */
dadab5e9 911 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
984263bc
MD
912 if (error)
913 return (error);
914
915 /*
916 * Check number of open-for-writes on the file and deny execution
917 * if there are any.
918 */
919 if (vp->v_writecount)
920 return (ETXTBSY);
921
922 /*
923 * Call filesystem specific open routine (which does nothing in the
924 * general case).
925 */
fad57d0e 926 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL, td);
984263bc
MD
927 if (error)
928 return (error);
929
930 return (0);
931}
932
933/*
934 * Exec handler registration
935 */
936int
77153250 937exec_register(const struct execsw *execsw_arg)
984263bc
MD
938{
939 const struct execsw **es, **xs, **newexecsw;
940 int count = 2; /* New slot and trailing NULL */
941
942 if (execsw)
943 for (es = execsw; *es; es++)
944 count++;
945 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
946 if (newexecsw == NULL)
947 return ENOMEM;
948 xs = newexecsw;
949 if (execsw)
950 for (es = execsw; *es; es++)
951 *xs++ = *es;
952 *xs++ = execsw_arg;
953 *xs = NULL;
954 if (execsw)
955 free(execsw, M_TEMP);
956 execsw = newexecsw;
957 return 0;
958}
959
960int
77153250 961exec_unregister(const struct execsw *execsw_arg)
984263bc
MD
962{
963 const struct execsw **es, **xs, **newexecsw;
964 int count = 1;
965
966 if (execsw == NULL)
ae7e74d4 967 panic("unregister with no handlers left?");
984263bc
MD
968
969 for (es = execsw; *es; es++) {
970 if (*es == execsw_arg)
971 break;
972 }
973 if (*es == NULL)
974 return ENOENT;
975 for (es = execsw; *es; es++)
976 if (*es != execsw_arg)
977 count++;
978 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
979 if (newexecsw == NULL)
980 return ENOMEM;
981 xs = newexecsw;
982 for (es = execsw; *es; es++)
983 if (*es != execsw_arg)
984 *xs++ = *es;
985 *xs = NULL;
986 if (execsw)
987 free(execsw, M_TEMP);
988 execsw = newexecsw;
989 return 0;
990}