Add an sc_maxiosize field which the ccd driver now needs.
[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 $
287ebb09 27 * $DragonFly: src/sys/kern/kern_exec.c,v 1.58 2007/06/29 21:54:08 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>
2bd9d75c 41#include <sys/kern_syscall.h>
984263bc
MD
42#include <sys/wait.h>
43#include <sys/malloc.h>
44#include <sys/proc.h>
29f58392 45#include <sys/ktrace.h>
984263bc
MD
46#include <sys/signalvar.h>
47#include <sys/pioctl.h>
fad57d0e 48#include <sys/nlookup.h>
18f40545 49#include <sys/sfbuf.h>
984263bc
MD
50#include <sys/sysent.h>
51#include <sys/shm.h>
52#include <sys/sysctl.h>
53#include <sys/vnode.h>
a6f89c72 54#include <sys/vmmeter.h>
984263bc 55#include <sys/aio.h>
17a7ca1f 56#include <sys/libkern.h>
984263bc
MD
57
58#include <vm/vm.h>
59#include <vm/vm_param.h>
60#include <sys/lock.h>
61#include <vm/pmap.h>
62#include <vm/vm_page.h>
63#include <vm/vm_map.h>
64#include <vm/vm_kern.h>
65#include <vm/vm_extern.h>
66#include <vm/vm_object.h>
67#include <vm/vm_pager.h>
68
69#include <sys/user.h>
527fddf7 70#include <sys/reg.h>
984263bc 71
5fd012e0
MD
72#include <sys/thread2.h>
73
984263bc 74MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
18b3f457 75MALLOC_DEFINE(M_EXECARGS, "exec-args", "Exec arguments");
984263bc 76
402ed7e1 77static register_t *exec_copyout_strings (struct image_params *);
984263bc
MD
78
79/* XXX This should be vm_size_t. */
80static u_long ps_strings = PS_STRINGS;
81SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
82
83/* XXX This should be vm_size_t. */
84static u_long usrstack = USRSTACK;
85SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
86
87u_long ps_arg_cache_limit = PAGE_SIZE / 16;
88SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
89 &ps_arg_cache_limit, 0, "");
90
91int ps_argsopen = 1;
92SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
93
2bd9d75c
DRJ
94void print_execve_args(struct image_args *args);
95int debug_execve_args = 0;
96SYSCTL_INT(_kern, OID_AUTO, debug_execve_args, CTLFLAG_RW, &debug_execve_args,
97 0, "");
98
18b3f457
MD
99/*
100 * Exec arguments object cache
101 */
102static struct objcache *exec_objcache;
103
104static
105void
106exec_objcache_init(void *arg __unused)
107{
108 exec_objcache = objcache_create_mbacked(
109 M_EXECARGS, PATH_MAX + ARG_MAX,
110 16, /* up to this many objects */
111 2, /* minimal magazine capacity */
112 NULL, NULL, NULL);
113}
114SYSINIT(exec_objcache, SI_BOOT2_MACHDEP, SI_ORDER_ANY, exec_objcache_init, 0);
115
17a7ca1f
MD
116/*
117 * stackgap_random specifies if the stackgap should have a random size added
118 * to it. It must be a power of 2. If non-zero, the stack gap will be
0ced1954 119 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
17a7ca1f
MD
120 */
121static int stackgap_random = 1024;
122static int
123sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS)
124{
125 int error, new_val;
126 new_val = stackgap_random;
127 error = sysctl_handle_int(oidp, &new_val, 0, req);
128 if (error != 0 || req->newptr == NULL)
129 return (error);
130 if ((new_val < 0) || (new_val > 16 * PAGE_SIZE) || ! powerof2(new_val))
131 return (EINVAL);
132 stackgap_random = new_val;
133
134 return(0);
135}
136
137SYSCTL_PROC(_kern, OID_AUTO, stackgap_random, CTLFLAG_RW|CTLTYPE_UINT,
138 0, 0, sysctl_kern_stackgap, "IU", "Max random stack gap (power of 2)");
139
2bd9d75c
DRJ
140void
141print_execve_args(struct image_args *args)
142{
143 char *cp;
144 int ndx;
145
146 cp = args->begin_argv;
147 for (ndx = 0; ndx < args->argc; ndx++) {
6ea70f76 148 kprintf("\targv[%d]: %s\n", ndx, cp);
2bd9d75c
DRJ
149 while (*cp++ != '\0');
150 }
151 for (ndx = 0; ndx < args->envc; ndx++) {
6ea70f76 152 kprintf("\tenvv[%d]: %s\n", ndx, cp);
2bd9d75c
DRJ
153 while (*cp++ != '\0');
154 }
155}
156
984263bc
MD
157/*
158 * Each of the items is a pointer to a `const struct execsw', hence the
159 * double pointer here.
160 */
161static const struct execsw **execsw;
162
984263bc 163int
fad57d0e 164kern_execve(struct nlookupdata *nd, struct image_args *args)
984263bc 165{
dadab5e9 166 struct thread *td = curthread;
08f2f1bb 167 struct lwp *lp = td->td_lwp;
dadab5e9 168 struct proc *p = td->td_proc;
984263bc
MD
169 register_t *stack_base;
170 int error, len, i;
171 struct image_params image_params, *imgp;
172 struct vattr attr;
402ed7e1 173 int (*img_first) (struct image_params *);
984263bc 174
287ebb09
MD
175 if (p->p_nthreads != 1) {
176 kprintf("pid %d attempt to exec with multiple LWPs present\n",
177 p->p_pid);
178 return (EINVAL);
179 }
180
2bd9d75c 181 if (debug_execve_args) {
6ea70f76 182 kprintf("%s()\n", __func__);
2bd9d75c
DRJ
183 print_execve_args(args);
184 }
185
dadab5e9 186 KKASSERT(p);
984263bc
MD
187 imgp = &image_params;
188
189 /*
190 * Lock the process and set the P_INEXEC flag to indicate that
191 * it should be left alone until we're done here. This is
192 * necessary to avoid race conditions - e.g. in ptrace() -
193 * that might allow a local user to illicitly obtain elevated
194 * privileges.
195 */
196 p->p_flag |= P_INEXEC;
197
198 /*
199 * Initialize part of the common data
200 */
201 imgp->proc = p;
2bd9d75c 202 imgp->args = args;
984263bc 203 imgp->attr = &attr;
984263bc 204 imgp->entry_addr = 0;
29802dbb 205 imgp->resident = 0;
984263bc
MD
206 imgp->vmspace_destroyed = 0;
207 imgp->interpreted = 0;
2abfe22f 208 imgp->interpreter_name[0] = 0;
984263bc
MD
209 imgp->auxargs = NULL;
210 imgp->vp = NULL;
211 imgp->firstpage = NULL;
212 imgp->ps_strings = 0;
18f40545 213 imgp->image_header = NULL;
2bd9d75c
DRJ
214
215interpret:
984263bc
MD
216
217 /*
fad57d0e
MD
218 * Translate the file name to a vnode. Unlock the cache entry to
219 * improve parallelism for programs exec'd in parallel.
984263bc 220 */
fad57d0e
MD
221 if ((error = nlookup(nd)) != 0)
222 goto exec_fail;
28623bf9 223 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
fad57d0e
MD
224 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
225 nd->nl_flags &= ~NLC_NCPISLOCKED;
28623bf9 226 cache_unlock(&nd->nl_nch);
fad57d0e 227 if (error)
984263bc 228 goto exec_fail;
984263bc
MD
229
230 /*
231 * Check file permissions (also 'opens' file)
232 */
233 error = exec_check_permissions(imgp);
234 if (error) {
a11aaa81 235 vn_unlock(imgp->vp);
984263bc
MD
236 goto exec_fail_dealloc;
237 }
238
239 error = exec_map_first_page(imgp);
a11aaa81 240 vn_unlock(imgp->vp);
984263bc
MD
241 if (error)
242 goto exec_fail_dealloc;
243
2bd9d75c 244 if (debug_execve_args && imgp->interpreted) {
6ea70f76
SW
245 kprintf(" target is interpreted -- recursive pass\n");
246 kprintf(" interpreter: %s\n", imgp->interpreter_name);
2bd9d75c
DRJ
247 print_execve_args(args);
248 }
249
984263bc
MD
250 /*
251 * If the current process has a special image activator it
252 * wants to try first, call it. For example, emulating shell
253 * scripts differently.
254 */
255 error = -1;
256 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
257 error = img_first(imgp);
258
29802dbb
MD
259 /*
260 * If the vnode has a registered vmspace, exec the vmspace
261 */
262 if (error == -1 && imgp->vp->v_resident) {
263 error = exec_resident_imgact(imgp);
264 }
265
984263bc
MD
266 /*
267 * Loop through the list of image activators, calling each one.
268 * An activator returns -1 if there is no match, 0 on success,
269 * and an error otherwise.
270 */
271 for (i = 0; error == -1 && execsw[i]; ++i) {
272 if (execsw[i]->ex_imgact == NULL ||
273 execsw[i]->ex_imgact == img_first) {
274 continue;
275 }
276 error = (*execsw[i]->ex_imgact)(imgp);
277 }
278
279 if (error) {
280 if (error == -1)
281 error = ENOEXEC;
282 goto exec_fail_dealloc;
283 }
284
285 /*
286 * Special interpreter operation, cleanup and loop up to try to
287 * activate the interpreter.
288 */
289 if (imgp->interpreted) {
290 exec_unmap_first_page(imgp);
fad57d0e
MD
291 nlookup_done(nd);
292 vrele(imgp->vp);
293 imgp->vp = NULL;
294 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
295 NLC_FOLLOW);
296 if (error)
297 goto exec_fail;
984263bc
MD
298 goto interpret;
299 }
300
301 /*
302 * Copy out strings (args and env) and initialize stack base
303 */
304 stack_base = exec_copyout_strings(imgp);
305 p->p_vmspace->vm_minsaddr = (char *)stack_base;
306
307 /*
308 * If custom stack fixup routine present for this process
29802dbb
MD
309 * let it do the stack setup. If we are running a resident
310 * image there is no auxinfo or other image activator context
311 * so don't try to add fixups to the stack.
312 *
984263bc
MD
313 * Else stuff argument count as first item on stack
314 */
29802dbb 315 if (p->p_sysent->sv_fixup && imgp->resident == 0)
984263bc
MD
316 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
317 else
2bd9d75c 318 suword(--stack_base, imgp->args->argc);
984263bc
MD
319
320 /*
321 * For security and other reasons, the file descriptor table cannot
322 * be shared after an exec.
323 */
324 if (p->p_fd->fd_refcnt > 1) {
325 struct filedesc *tmp;
326
327 tmp = fdcopy(p);
328 fdfree(p);
329 p->p_fd = tmp;
330 }
331
332 /*
333 * For security and other reasons, signal handlers cannot
334 * be shared after an exec. The new proces gets a copy of the old
335 * handlers. In execsigs(), the new process will have its signals
336 * reset.
337 */
b1b4e5a6
SS
338 if (p->p_sigacts->ps_refcnt > 1) {
339 struct sigacts *newsigacts;
984263bc 340
b1b4e5a6 341 newsigacts = (struct sigacts *)kmalloc(sizeof(*newsigacts),
984263bc 342 M_SUBPROC, M_WAITOK);
b1b4e5a6
SS
343 bcopy(p->p_sigacts, newsigacts, sizeof(*newsigacts));
344 p->p_sigacts->ps_refcnt--;
345 p->p_sigacts = newsigacts;
346 p->p_sigacts->ps_refcnt = 1;
984263bc
MD
347 }
348
0daa37a5
MD
349 /*
350 * For security and other reasons virtual kernels cannot be
351 * inherited by an exec. This also allows a virtual kernel
352 * to fork/exec unrelated applications.
353 */
4a22e893
MD
354 if (p->p_vkernel)
355 vkernel_exit(p);
0daa37a5 356
984263bc
MD
357 /* Stop profiling */
358 stopprofclock(p);
359
360 /* close files on exec */
361 fdcloseexec(p);
362
363 /* reset caught signals */
364 execsigs(p);
365
366 /* name this process - nameiexec(p, ndp) */
28623bf9
MD
367 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
368 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
984263bc 369 p->p_comm[len] = 0;
08f2f1bb 370 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
984263bc
MD
371
372 /*
373 * mark as execed, wakeup the process that vforked (if any) and tell
374 * it that it now has its own resources back
375 */
376 p->p_flag |= P_EXEC;
377 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
378 p->p_flag &= ~P_PPWAIT;
379 wakeup((caddr_t)p->p_pptr);
380 }
381
382 /*
383 * Implement image setuid/setgid.
384 *
385 * Don't honor setuid/setgid if the filesystem prohibits it or if
386 * the process is being traced.
387 */
388 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
389 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
390 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
391 (p->p_flag & P_TRACED) == 0) {
392 /*
393 * Turn off syscall tracing for set-id programs, except for
394 * root. Record any set-id flags first to make sure that
395 * we do not regain any tracing during a possible block.
396 */
41c20dac 397 setsugid();
29f58392
MD
398 if (p->p_tracenode && suser(td) != 0) {
399 ktrdestroy(&p->p_tracenode);
400 p->p_traceflag = 0;
984263bc
MD
401 }
402 /* Close any file descriptors 0..2 that reference procfs */
403 setugidsafety(p);
404 /* Make sure file descriptors 0..2 are in use. */
405 error = fdcheckstd(p);
406 if (error != 0)
407 goto exec_fail_dealloc;
408 /*
409 * Set the new credentials.
410 */
e9a372eb 411 cratom(&p->p_ucred);
984263bc 412 if (attr.va_mode & VSUID)
41c20dac 413 change_euid(attr.va_uid);
984263bc
MD
414 if (attr.va_mode & VSGID)
415 p->p_ucred->cr_gid = attr.va_gid;
98a7f915
MD
416
417 /*
418 * Clear local varsym variables
419 */
420 varsymset_clean(&p->p_varsymset);
984263bc 421 } else {
41c20dac
MD
422 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
423 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
984263bc
MD
424 p->p_flag &= ~P_SUGID;
425 }
426
427 /*
428 * Implement correct POSIX saved-id behavior.
429 */
115ccd83
MD
430 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
431 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
432 cratom(&p->p_ucred);
433 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
434 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
435 }
984263bc
MD
436
437 /*
438 * Store the vp for use in procfs
439 */
440 if (p->p_textvp) /* release old reference */
441 vrele(p->p_textvp);
fad57d0e
MD
442 p->p_textvp = imgp->vp;
443 vref(p->p_textvp);
984263bc
MD
444
445 /*
446 * Notify others that we exec'd, and clear the P_INEXEC flag
447 * as we're now a bona fide freshly-execed process.
448 */
449 KNOTE(&p->p_klist, NOTE_EXEC);
450 p->p_flag &= ~P_INEXEC;
451
452 /*
453 * If tracing the process, trap to debugger so breakpoints
454 * can be set before the program executes.
455 */
456 STOPEVENT(p, S_EXEC, 0);
457
458 if (p->p_flag & P_TRACED)
84204577 459 ksignal(p, SIGTRAP);
984263bc
MD
460
461 /* clear "fork but no exec" flag, as we _are_ execing */
462 p->p_acflag &= ~AFORK;
463
464 /* Set values passed into the program in registers. */
08f2f1bb 465 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
984263bc
MD
466 imgp->ps_strings);
467
468 /* Free any previous argument cache */
469 if (p->p_args && --p->p_args->ar_ref == 0)
470 FREE(p->p_args, M_PARGS);
471 p->p_args = NULL;
472
473 /* Cache arguments if they fit inside our allowance */
2bd9d75c 474 i = imgp->args->begin_envv - imgp->args->begin_argv;
984263bc
MD
475 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
476 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
477 M_PARGS, M_WAITOK);
478 p->p_args->ar_ref = 1;
479 p->p_args->ar_length = i;
2bd9d75c 480 bcopy(imgp->args->begin_argv, p->p_args->ar_args, i);
984263bc
MD
481 }
482
483exec_fail_dealloc:
484
485 /*
486 * free various allocated resources
487 */
488 if (imgp->firstpage)
489 exec_unmap_first_page(imgp);
490
984263bc 491 if (imgp->vp) {
984263bc 492 vrele(imgp->vp);
fad57d0e 493 imgp->vp = NULL;
984263bc
MD
494 }
495
a6f89c72
DR
496 if (error == 0) {
497 ++mycpu->gd_cnt.v_exec;
984263bc 498 return (0);
a6f89c72 499 }
984263bc
MD
500
501exec_fail:
502 /* we're done here, clear P_INEXEC */
503 p->p_flag &= ~P_INEXEC;
504 if (imgp->vmspace_destroyed) {
505 /* sorry, no more process anymore. exit gracefully */
41c20dac 506 exit1(W_EXITCODE(0, SIGABRT));
984263bc
MD
507 /* NOT REACHED */
508 return(0);
509 } else {
510 return(error);
511 }
512}
513
2bd9d75c
DRJ
514/*
515 * execve() system call.
516 */
517int
753fd850 518sys_execve(struct execve_args *uap)
2bd9d75c 519{
fad57d0e 520 struct nlookupdata nd;
2bd9d75c
DRJ
521 struct image_args args;
522 int error;
523
fad57d0e
MD
524 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
525 if (error == 0) {
526 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
527 uap->argv, uap->envv);
528 }
2abfe22f
MD
529 if (error == 0)
530 error = kern_execve(&nd, &args);
fad57d0e 531 nlookup_done(&nd);
2bd9d75c
DRJ
532 exec_free_args(&args);
533
534 /*
535 * The syscall result is returned in registers to the new program.
536 * Linux will register %edx as an atexit function and we must be
537 * sure to set it to 0. XXX
538 */
539 if (error == 0)
540 uap->sysmsg_result64 = 0;
541
542 return (error);
543}
544
984263bc 545int
41c20dac 546exec_map_first_page(struct image_params *imgp)
984263bc 547{
5fd012e0 548 int rv, i;
984263bc
MD
549 int initial_pagein;
550 vm_page_t ma[VM_INITIAL_PAGEIN];
06ecca5a 551 vm_page_t m;
984263bc
MD
552 vm_object_t object;
553
18f40545 554 if (imgp->firstpage)
984263bc 555 exec_unmap_first_page(imgp);
984263bc 556
fad57d0e 557 /*
7540ab49 558 * The file has to be mappable.
fad57d0e 559 */
7540ab49 560 if ((object = imgp->vp->v_object) == NULL)
fad57d0e 561 return (EIO);
984263bc 562
06ecca5a
MD
563 /*
564 * We shouldn't need protection for vm_page_grab() but we certainly
565 * need it for the lookup loop below (lookup/busy race), since
566 * an interrupt can unbusy and free the page before our busy check.
567 */
5fd012e0 568 crit_enter();
06ecca5a 569 m = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
984263bc 570
06ecca5a
MD
571 if ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
572 ma[0] = m;
984263bc
MD
573 initial_pagein = VM_INITIAL_PAGEIN;
574 if (initial_pagein > object->size)
575 initial_pagein = object->size;
576 for (i = 1; i < initial_pagein; i++) {
06ecca5a
MD
577 if ((m = vm_page_lookup(object, i)) != NULL) {
578 if ((m->flags & PG_BUSY) || m->busy)
984263bc 579 break;
06ecca5a 580 if (m->valid)
984263bc 581 break;
06ecca5a 582 vm_page_busy(m);
984263bc 583 } else {
06ecca5a
MD
584 m = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
585 if (m == NULL)
984263bc
MD
586 break;
587 }
06ecca5a 588 ma[i] = m;
984263bc
MD
589 }
590 initial_pagein = i;
591
06ecca5a
MD
592 /*
593 * get_pages unbusies all the requested pages except the
594 * primary page (at index 0 in this case).
595 */
984263bc 596 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
06ecca5a 597 m = vm_page_lookup(object, 0);
984263bc 598
06ecca5a
MD
599 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
600 if (m) {
601 vm_page_protect(m, VM_PROT_NONE);
602 vm_page_free(m);
984263bc 603 }
5fd012e0 604 crit_exit();
984263bc
MD
605 return EIO;
606 }
607 }
06ecca5a
MD
608 vm_page_hold(m);
609 vm_page_wakeup(m); /* unbusy the page */
5fd012e0 610 crit_exit();
984263bc 611
4f1640d6 612 imgp->firstpage = sf_buf_alloc(m, SFB_CPUPRIVATE);
18f40545 613 imgp->image_header = (void *)sf_buf_kva(imgp->firstpage);
984263bc
MD
614
615 return 0;
616}
617
618void
77153250 619exec_unmap_first_page(struct image_params *imgp)
984263bc 620{
18f40545
MD
621 vm_page_t m;
622
5fd012e0 623 crit_enter();
18f40545
MD
624 if (imgp->firstpage != NULL) {
625 m = sf_buf_page(imgp->firstpage);
626 sf_buf_free(imgp->firstpage);
984263bc 627 imgp->firstpage = NULL;
18f40545 628 imgp->image_header = NULL;
e4ba7818 629 vm_page_unhold(m);
984263bc 630 }
5fd012e0 631 crit_exit();
984263bc
MD
632}
633
634/*
635 * Destroy old address space, and allocate a new stack
636 * The new stack is only SGROWSIZ large because it is grown
637 * automatically in trap.c.
638 */
639int
29802dbb 640exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
984263bc
MD
641{
642 int error;
643 struct vmspace *vmspace = imgp->proc->p_vmspace;
644 vm_offset_t stack_addr = USRSTACK - maxssiz;
29802dbb 645 vm_map_t map;
984263bc
MD
646
647 imgp->vmspace_destroyed = 1;
648
70d3d461
SS
649 if (curthread->td_proc->p_nthreads > 1)
650 killlwps(curthread->td_lwp);
08f2f1bb 651
984263bc
MD
652 /*
653 * Prevent a pending AIO from modifying the new address space.
654 */
655 aio_proc_rundown(imgp->proc);
656
657 /*
658 * Blow away entire process VM, if address space not shared,
659 * otherwise, create a new VM space so that other threads are
29802dbb
MD
660 * not disrupted. If we are execing a resident vmspace we
661 * create a duplicate of it and remap the stack.
239b4df9
MD
662 *
663 * The exitingcnt test is not strictly necessary but has been
664 * included for code sanity (to make the code more deterministic).
984263bc 665 */
29802dbb
MD
666 map = &vmspace->vm_map;
667 if (vmcopy) {
668 vmspace_exec(imgp->proc, vmcopy);
669 vmspace = imgp->proc->p_vmspace;
670 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
671 map = &vmspace->vm_map;
e3161323
MD
672 } else if (vmspace->vm_sysref.refcnt == 1 &&
673 vmspace->vm_exitingcnt == 0) {
fef0fdf2 674 shmexit(vmspace);
239b4df9 675 if (vmspace->vm_upcalls)
08f2f1bb 676 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
88181b08
MD
677 pmap_remove_pages(vmspace_pmap(vmspace),
678 0, VM_MAX_USER_ADDRESS);
679 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
984263bc 680 } else {
29802dbb 681 vmspace_exec(imgp->proc, NULL);
984263bc
MD
682 vmspace = imgp->proc->p_vmspace;
683 map = &vmspace->vm_map;
684 }
685
686 /* Allocate a new stack */
687 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
688 VM_PROT_ALL, VM_PROT_ALL, 0);
689 if (error)
690 return (error);
691
692 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
693 * VM_STACK case, but they are still used to monitor the size of the
694 * process stack so we can check the stack rlimit.
695 */
696 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
697 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
698
699 return(0);
700}
701
702/*
703 * Copy out argument and environment strings from the old process
704 * address space into the temporary string buffer.
705 */
706int
2bd9d75c 707exec_copyin_args(struct image_args *args, char *fname,
2abfe22f 708 enum exec_path_segflg segflg, char **argv, char **envv)
984263bc 709{
984263bc 710 char *argp, *envp;
2bd9d75c 711 int error = 0;
984263bc
MD
712 size_t length;
713
2abfe22f 714 bzero(args, sizeof(*args));
18b3f457
MD
715
716 args->buf = objcache_get(exec_objcache, M_WAITOK);
2bd9d75c
DRJ
717 if (args->buf == NULL)
718 return (ENOMEM);
719 args->begin_argv = args->buf;
720 args->endp = args->begin_argv;
721 args->space = ARG_MAX;
2bd9d75c
DRJ
722
723 args->fname = args->buf + ARG_MAX;
724
984263bc 725 /*
2bd9d75c 726 * Copy the file name.
984263bc 727 */
2bd9d75c
DRJ
728 if (segflg == PATH_SYSSPACE) {
729 error = copystr(fname, args->fname, PATH_MAX, &length);
730 } else if (segflg == PATH_USERSPACE) {
731 error = copyinstr(fname, args->fname, PATH_MAX, &length);
732 }
984263bc 733
2bd9d75c 734 /*
66be6566
MD
735 * Extract argument strings. argv may not be NULL. The argv
736 * array is terminated by a NULL entry. We special-case the
737 * situation where argv[0] is NULL by passing { filename, NULL }
738 * to the new program to guarentee that the interpreter knows what
739 * file to open in case we exec an interpreted file. Note that
740 * a NULL argv[0] terminates the argv[] array.
741 *
742 * XXX the special-casing of argv[0] is historical and needs to be
743 * revisited.
2bd9d75c 744 */
66be6566
MD
745 if (argv == NULL)
746 error = EFAULT;
747 if (error == 0) {
748 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
749 if (argp == (caddr_t)-1) {
2bd9d75c 750 error = EFAULT;
66be6566 751 break;
2bd9d75c
DRJ
752 }
753 error = copyinstr(argp, args->endp,
2abfe22f 754 args->space, &length);
66be6566
MD
755 if (error) {
756 if (error == ENAMETOOLONG)
757 error = E2BIG;
758 break;
759 }
2bd9d75c
DRJ
760 args->space -= length;
761 args->endp += length;
762 args->argc++;
984263bc 763 }
66be6566
MD
764 if (args->argc == 0 && error == 0) {
765 length = strlen(args->fname) + 1;
766 if (length > args->space) {
767 error = E2BIG;
768 } else {
769 bcopy(args->fname, args->endp, length);
770 args->space -= length;
771 args->endp += length;
772 args->argc++;
773 }
774 }
984263bc
MD
775 }
776
2bd9d75c 777 args->begin_envv = args->endp;
984263bc
MD
778
779 /*
66be6566 780 * extract environment strings. envv may be NULL.
984263bc 781 */
2abfe22f 782 if (envv && error == 0) {
984263bc 783 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
2bd9d75c
DRJ
784 if (envp == (caddr_t) -1) {
785 error = EFAULT;
66be6566 786 break;
984263bc 787 }
2bd9d75c
DRJ
788 error = copyinstr(envp, args->endp, args->space,
789 &length);
66be6566
MD
790 if (error) {
791 if (error == ENAMETOOLONG)
792 error = E2BIG;
793 break;
794 }
2bd9d75c
DRJ
795 args->space -= length;
796 args->endp += length;
797 args->envc++;
984263bc
MD
798 }
799 }
2bd9d75c
DRJ
800 return (error);
801}
802
803void
804exec_free_args(struct image_args *args)
805{
2abfe22f 806 if (args->buf) {
18b3f457 807 objcache_put(exec_objcache, args->buf);
2abfe22f
MD
808 args->buf = NULL;
809 }
984263bc
MD
810}
811
812/*
813 * Copy strings out to the new process address space, constructing
814 * new arg and env vector tables. Return a pointer to the base
815 * so that it can be used as the initial stack pointer.
816 */
817register_t *
2bd9d75c 818exec_copyout_strings(struct image_params *imgp)
984263bc 819{
17a7ca1f 820 int argc, envc, sgap;
984263bc
MD
821 char **vectp;
822 char *stringp, *destp;
823 register_t *stack_base;
824 struct ps_strings *arginfo;
825 int szsigcode;
826
827 /*
828 * Calculate string base and vector table pointers.
829 * Also deal with signal trampoline code for this exec type.
830 */
831 arginfo = (struct ps_strings *)PS_STRINGS;
832 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
17a7ca1f 833 if (stackgap_random != 0)
0ced1954 834 sgap = ALIGN(karc4random() & (stackgap_random - 1));
17a7ca1f
MD
835 else
836 sgap = 0;
837 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
2bd9d75c 838 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
984263bc
MD
839
840 /*
841 * install sigcode
842 */
843 if (szsigcode)
844 copyout(imgp->proc->p_sysent->sv_sigcode,
2bd9d75c 845 ((caddr_t)arginfo - szsigcode), szsigcode);
984263bc
MD
846
847 /*
848 * If we have a valid auxargs ptr, prepare some room
849 * on the stack.
7ec1918b 850 *
984263bc
MD
851 * The '+ 2' is for the null pointers at the end of each of the
852 * arg and env vector sets, and 'AT_COUNT*2' is room for the
853 * ELF Auxargs data.
854 */
7ec1918b
MD
855 if (imgp->auxargs) {
856 vectp = (char **)(destp - (imgp->args->argc +
857 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
858 } else {
2bd9d75c 859 vectp = (char **)(destp - (imgp->args->argc +
7ec1918b
MD
860 imgp->args->envc + 2) * sizeof(char*));
861 }
862
984263bc 863 /*
31bc42c8
MD
864 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
865 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
984263bc 866 */
984263bc
MD
867
868 /*
869 * vectp also becomes our initial stack base
870 */
871 stack_base = (register_t *)vectp;
872
2bd9d75c
DRJ
873 stringp = imgp->args->begin_argv;
874 argc = imgp->args->argc;
875 envc = imgp->args->envc;
984263bc
MD
876
877 /*
878 * Copy out strings - arguments and environment.
879 */
2bd9d75c 880 copyout(stringp, destp, ARG_MAX - imgp->args->space);
984263bc
MD
881
882 /*
883 * Fill in "ps_strings" struct for ps, w, etc.
884 */
885 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
886 suword(&arginfo->ps_nargvstr, argc);
887
888 /*
889 * Fill in argument portion of vector table.
890 */
891 for (; argc > 0; --argc) {
892 suword(vectp++, (long)(intptr_t)destp);
893 while (*stringp++ != 0)
894 destp++;
895 destp++;
896 }
897
7a44d1cb 898 /* a null vector table pointer separates the argp's from the envp's */
984263bc
MD
899 suword(vectp++, 0);
900
901 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
902 suword(&arginfo->ps_nenvstr, envc);
903
904 /*
905 * Fill in environment portion of vector table.
906 */
907 for (; envc > 0; --envc) {
908 suword(vectp++, (long)(intptr_t)destp);
909 while (*stringp++ != 0)
910 destp++;
911 destp++;
912 }
913
914 /* end of vector table is a null pointer */
915 suword(vectp, 0);
916
917 return (stack_base);
918}
919
920/*
921 * Check permissions of file to execute.
922 * Return 0 for success or error code on failure.
923 */
924int
77153250 925exec_check_permissions(struct image_params *imgp)
984263bc
MD
926{
927 struct proc *p = imgp->proc;
928 struct vnode *vp = imgp->vp;
929 struct vattr *attr = imgp->attr;
930 int error;
931
932 /* Get file attributes */
87de5057 933 error = VOP_GETATTR(vp, attr);
984263bc
MD
934 if (error)
935 return (error);
936
937 /*
938 * 1) Check if file execution is disabled for the filesystem that this
939 * file resides on.
940 * 2) Insure that at least one execute bit is on - otherwise root
941 * will always succeed, and we don't want to happen unless the
942 * file really is executable.
943 * 3) Insure that the file is a regular file.
944 */
945 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
946 ((attr->va_mode & 0111) == 0) ||
947 (attr->va_type != VREG)) {
948 return (EACCES);
949 }
950
951 /*
952 * Zero length files can't be exec'd
953 */
954 if (attr->va_size == 0)
955 return (ENOEXEC);
956
957 /*
958 * Check for execute permission to file based on current credentials.
959 */
87de5057 960 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
984263bc
MD
961 if (error)
962 return (error);
963
964 /*
965 * Check number of open-for-writes on the file and deny execution
966 * if there are any.
967 */
968 if (vp->v_writecount)
969 return (ETXTBSY);
970
971 /*
7540ab49
MD
972 * Call filesystem specific open routine, which allows us to read,
973 * write, and mmap the file. Without the VOP_OPEN we can only
974 * stat the file.
984263bc 975 */
87de5057 976 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
984263bc
MD
977 if (error)
978 return (error);
979
980 return (0);
981}
982
983/*
984 * Exec handler registration
985 */
986int
77153250 987exec_register(const struct execsw *execsw_arg)
984263bc
MD
988{
989 const struct execsw **es, **xs, **newexecsw;
990 int count = 2; /* New slot and trailing NULL */
991
992 if (execsw)
993 for (es = execsw; *es; es++)
994 count++;
efda3bd0 995 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
984263bc
MD
996 if (newexecsw == NULL)
997 return ENOMEM;
998 xs = newexecsw;
999 if (execsw)
1000 for (es = execsw; *es; es++)
1001 *xs++ = *es;
1002 *xs++ = execsw_arg;
1003 *xs = NULL;
1004 if (execsw)
efda3bd0 1005 kfree(execsw, M_TEMP);
984263bc
MD
1006 execsw = newexecsw;
1007 return 0;
1008}
1009
1010int
77153250 1011exec_unregister(const struct execsw *execsw_arg)
984263bc
MD
1012{
1013 const struct execsw **es, **xs, **newexecsw;
1014 int count = 1;
1015
1016 if (execsw == NULL)
ae7e74d4 1017 panic("unregister with no handlers left?");
984263bc
MD
1018
1019 for (es = execsw; *es; es++) {
1020 if (*es == execsw_arg)
1021 break;
1022 }
1023 if (*es == NULL)
1024 return ENOENT;
1025 for (es = execsw; *es; es++)
1026 if (*es != execsw_arg)
1027 count++;
efda3bd0 1028 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
984263bc
MD
1029 if (newexecsw == NULL)
1030 return ENOMEM;
1031 xs = newexecsw;
1032 for (es = execsw; *es; es++)
1033 if (*es != execsw_arg)
1034 *xs++ = *es;
1035 *xs = NULL;
1036 if (execsw)
efda3bd0 1037 kfree(execsw, M_TEMP);
984263bc
MD
1038 execsw = newexecsw;
1039 return 0;
1040}