Create 'k' versions of the kernel malloc API.
[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 $
84204577 27 * $DragonFly: src/sys/kern/kern_exec.c,v 1.43 2006/09/03 18:29:16 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>
70#include <machine/reg.h>
71
5fd012e0
MD
72#include <sys/thread2.h>
73
984263bc
MD
74MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
75
402ed7e1 76static register_t *exec_copyout_strings (struct image_params *);
984263bc
MD
77
78/* XXX This should be vm_size_t. */
79static u_long ps_strings = PS_STRINGS;
80SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
81
82/* XXX This should be vm_size_t. */
83static u_long usrstack = USRSTACK;
84SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
85
86u_long ps_arg_cache_limit = PAGE_SIZE / 16;
87SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
88 &ps_arg_cache_limit, 0, "");
89
90int ps_argsopen = 1;
91SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
92
2bd9d75c
DRJ
93void print_execve_args(struct image_args *args);
94int debug_execve_args = 0;
95SYSCTL_INT(_kern, OID_AUTO, debug_execve_args, CTLFLAG_RW, &debug_execve_args,
96 0, "");
97
17a7ca1f
MD
98/*
99 * stackgap_random specifies if the stackgap should have a random size added
100 * to it. It must be a power of 2. If non-zero, the stack gap will be
0ced1954 101 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
17a7ca1f
MD
102 */
103static int stackgap_random = 1024;
104static int
105sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS)
106{
107 int error, new_val;
108 new_val = stackgap_random;
109 error = sysctl_handle_int(oidp, &new_val, 0, req);
110 if (error != 0 || req->newptr == NULL)
111 return (error);
112 if ((new_val < 0) || (new_val > 16 * PAGE_SIZE) || ! powerof2(new_val))
113 return (EINVAL);
114 stackgap_random = new_val;
115
116 return(0);
117}
118
119SYSCTL_PROC(_kern, OID_AUTO, stackgap_random, CTLFLAG_RW|CTLTYPE_UINT,
120 0, 0, sysctl_kern_stackgap, "IU", "Max random stack gap (power of 2)");
121
2bd9d75c
DRJ
122void
123print_execve_args(struct image_args *args)
124{
125 char *cp;
126 int ndx;
127
128 cp = args->begin_argv;
129 for (ndx = 0; ndx < args->argc; ndx++) {
130 printf("\targv[%d]: %s\n", ndx, cp);
131 while (*cp++ != '\0');
132 }
133 for (ndx = 0; ndx < args->envc; ndx++) {
134 printf("\tenvv[%d]: %s\n", ndx, cp);
135 while (*cp++ != '\0');
136 }
137}
138
984263bc
MD
139/*
140 * Each of the items is a pointer to a `const struct execsw', hence the
141 * double pointer here.
142 */
143static const struct execsw **execsw;
144
984263bc 145int
fad57d0e 146kern_execve(struct nlookupdata *nd, struct image_args *args)
984263bc 147{
dadab5e9
MD
148 struct thread *td = curthread;
149 struct proc *p = td->td_proc;
984263bc
MD
150 register_t *stack_base;
151 int error, len, i;
152 struct image_params image_params, *imgp;
153 struct vattr attr;
402ed7e1 154 int (*img_first) (struct image_params *);
984263bc 155
2bd9d75c
DRJ
156 if (debug_execve_args) {
157 printf("%s()\n", __func__);
158 print_execve_args(args);
159 }
160
dadab5e9 161 KKASSERT(p);
984263bc
MD
162 imgp = &image_params;
163
164 /*
165 * Lock the process and set the P_INEXEC flag to indicate that
166 * it should be left alone until we're done here. This is
167 * necessary to avoid race conditions - e.g. in ptrace() -
168 * that might allow a local user to illicitly obtain elevated
169 * privileges.
170 */
171 p->p_flag |= P_INEXEC;
172
173 /*
174 * Initialize part of the common data
175 */
176 imgp->proc = p;
2bd9d75c 177 imgp->args = args;
984263bc 178 imgp->attr = &attr;
984263bc 179 imgp->entry_addr = 0;
29802dbb 180 imgp->resident = 0;
984263bc
MD
181 imgp->vmspace_destroyed = 0;
182 imgp->interpreted = 0;
2abfe22f 183 imgp->interpreter_name[0] = 0;
984263bc
MD
184 imgp->auxargs = NULL;
185 imgp->vp = NULL;
186 imgp->firstpage = NULL;
187 imgp->ps_strings = 0;
18f40545 188 imgp->image_header = NULL;
2bd9d75c
DRJ
189
190interpret:
984263bc
MD
191
192 /*
fad57d0e
MD
193 * Translate the file name to a vnode. Unlock the cache entry to
194 * improve parallelism for programs exec'd in parallel.
984263bc 195 */
fad57d0e
MD
196 if ((error = nlookup(nd)) != 0)
197 goto exec_fail;
198 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
199 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
200 nd->nl_flags &= ~NLC_NCPISLOCKED;
201 cache_unlock(nd->nl_ncp);
202 if (error)
984263bc 203 goto exec_fail;
984263bc
MD
204
205 /*
206 * Check file permissions (also 'opens' file)
207 */
208 error = exec_check_permissions(imgp);
209 if (error) {
a11aaa81 210 vn_unlock(imgp->vp);
984263bc
MD
211 goto exec_fail_dealloc;
212 }
213
214 error = exec_map_first_page(imgp);
a11aaa81 215 vn_unlock(imgp->vp);
984263bc
MD
216 if (error)
217 goto exec_fail_dealloc;
218
2bd9d75c
DRJ
219 if (debug_execve_args && imgp->interpreted) {
220 printf(" target is interpreted -- recursive pass\n");
221 printf(" interpreter: %s\n", imgp->interpreter_name);
222 print_execve_args(args);
223 }
224
984263bc
MD
225 /*
226 * If the current process has a special image activator it
227 * wants to try first, call it. For example, emulating shell
228 * scripts differently.
229 */
230 error = -1;
231 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
232 error = img_first(imgp);
233
29802dbb
MD
234 /*
235 * If the vnode has a registered vmspace, exec the vmspace
236 */
237 if (error == -1 && imgp->vp->v_resident) {
238 error = exec_resident_imgact(imgp);
239 }
240
984263bc
MD
241 /*
242 * Loop through the list of image activators, calling each one.
243 * An activator returns -1 if there is no match, 0 on success,
244 * and an error otherwise.
245 */
246 for (i = 0; error == -1 && execsw[i]; ++i) {
247 if (execsw[i]->ex_imgact == NULL ||
248 execsw[i]->ex_imgact == img_first) {
249 continue;
250 }
251 error = (*execsw[i]->ex_imgact)(imgp);
252 }
253
254 if (error) {
255 if (error == -1)
256 error = ENOEXEC;
257 goto exec_fail_dealloc;
258 }
259
260 /*
261 * Special interpreter operation, cleanup and loop up to try to
262 * activate the interpreter.
263 */
264 if (imgp->interpreted) {
265 exec_unmap_first_page(imgp);
fad57d0e
MD
266 nlookup_done(nd);
267 vrele(imgp->vp);
268 imgp->vp = NULL;
269 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
270 NLC_FOLLOW);
271 if (error)
272 goto exec_fail;
984263bc
MD
273 goto interpret;
274 }
275
276 /*
277 * Copy out strings (args and env) and initialize stack base
278 */
279 stack_base = exec_copyout_strings(imgp);
280 p->p_vmspace->vm_minsaddr = (char *)stack_base;
281
282 /*
283 * If custom stack fixup routine present for this process
29802dbb
MD
284 * let it do the stack setup. If we are running a resident
285 * image there is no auxinfo or other image activator context
286 * so don't try to add fixups to the stack.
287 *
984263bc
MD
288 * Else stuff argument count as first item on stack
289 */
29802dbb 290 if (p->p_sysent->sv_fixup && imgp->resident == 0)
984263bc
MD
291 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
292 else
2bd9d75c 293 suword(--stack_base, imgp->args->argc);
984263bc
MD
294
295 /*
296 * For security and other reasons, the file descriptor table cannot
297 * be shared after an exec.
298 */
299 if (p->p_fd->fd_refcnt > 1) {
300 struct filedesc *tmp;
301
302 tmp = fdcopy(p);
303 fdfree(p);
304 p->p_fd = tmp;
305 }
306
307 /*
308 * For security and other reasons, signal handlers cannot
309 * be shared after an exec. The new proces gets a copy of the old
310 * handlers. In execsigs(), the new process will have its signals
311 * reset.
312 */
313 if (p->p_procsig->ps_refcnt > 1) {
314 struct procsig *newprocsig;
315
316 MALLOC(newprocsig, struct procsig *, sizeof(struct procsig),
317 M_SUBPROC, M_WAITOK);
318 bcopy(p->p_procsig, newprocsig, sizeof(*newprocsig));
319 p->p_procsig->ps_refcnt--;
320 p->p_procsig = newprocsig;
321 p->p_procsig->ps_refcnt = 1;
322 if (p->p_sigacts == &p->p_addr->u_sigacts)
323 panic("shared procsig but private sigacts?");
324
325 p->p_addr->u_sigacts = *p->p_sigacts;
326 p->p_sigacts = &p->p_addr->u_sigacts;
327 }
328
329 /* Stop profiling */
330 stopprofclock(p);
331
332 /* close files on exec */
333 fdcloseexec(p);
334
335 /* reset caught signals */
336 execsigs(p);
337
338 /* name this process - nameiexec(p, ndp) */
fad57d0e
MD
339 len = min(nd->nl_ncp->nc_nlen, MAXCOMLEN);
340 bcopy(nd->nl_ncp->nc_name, p->p_comm, len);
984263bc 341 p->p_comm[len] = 0;
2966ed3e 342 bcopy(p->p_comm, p->p_lwp.lwp_thread->td_comm, MAXCOMLEN+1);
984263bc
MD
343
344 /*
345 * mark as execed, wakeup the process that vforked (if any) and tell
346 * it that it now has its own resources back
347 */
348 p->p_flag |= P_EXEC;
349 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
350 p->p_flag &= ~P_PPWAIT;
351 wakeup((caddr_t)p->p_pptr);
352 }
353
354 /*
355 * Implement image setuid/setgid.
356 *
357 * Don't honor setuid/setgid if the filesystem prohibits it or if
358 * the process is being traced.
359 */
360 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
361 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
362 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
363 (p->p_flag & P_TRACED) == 0) {
364 /*
365 * Turn off syscall tracing for set-id programs, except for
366 * root. Record any set-id flags first to make sure that
367 * we do not regain any tracing during a possible block.
368 */
41c20dac 369 setsugid();
29f58392
MD
370 if (p->p_tracenode && suser(td) != 0) {
371 ktrdestroy(&p->p_tracenode);
372 p->p_traceflag = 0;
984263bc
MD
373 }
374 /* Close any file descriptors 0..2 that reference procfs */
375 setugidsafety(p);
376 /* Make sure file descriptors 0..2 are in use. */
377 error = fdcheckstd(p);
378 if (error != 0)
379 goto exec_fail_dealloc;
380 /*
381 * Set the new credentials.
382 */
e9a372eb 383 cratom(&p->p_ucred);
984263bc 384 if (attr.va_mode & VSUID)
41c20dac 385 change_euid(attr.va_uid);
984263bc
MD
386 if (attr.va_mode & VSGID)
387 p->p_ucred->cr_gid = attr.va_gid;
98a7f915
MD
388
389 /*
390 * Clear local varsym variables
391 */
392 varsymset_clean(&p->p_varsymset);
984263bc 393 } else {
41c20dac
MD
394 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
395 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
984263bc
MD
396 p->p_flag &= ~P_SUGID;
397 }
398
399 /*
400 * Implement correct POSIX saved-id behavior.
401 */
115ccd83
MD
402 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
403 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
404 cratom(&p->p_ucred);
405 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
406 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
407 }
984263bc
MD
408
409 /*
410 * Store the vp for use in procfs
411 */
412 if (p->p_textvp) /* release old reference */
413 vrele(p->p_textvp);
fad57d0e
MD
414 p->p_textvp = imgp->vp;
415 vref(p->p_textvp);
984263bc
MD
416
417 /*
418 * Notify others that we exec'd, and clear the P_INEXEC flag
419 * as we're now a bona fide freshly-execed process.
420 */
421 KNOTE(&p->p_klist, NOTE_EXEC);
422 p->p_flag &= ~P_INEXEC;
423
424 /*
425 * If tracing the process, trap to debugger so breakpoints
426 * can be set before the program executes.
427 */
428 STOPEVENT(p, S_EXEC, 0);
429
430 if (p->p_flag & P_TRACED)
84204577 431 ksignal(p, SIGTRAP);
984263bc
MD
432
433 /* clear "fork but no exec" flag, as we _are_ execing */
434 p->p_acflag &= ~AFORK;
435
436 /* Set values passed into the program in registers. */
437 setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
438 imgp->ps_strings);
439
440 /* Free any previous argument cache */
441 if (p->p_args && --p->p_args->ar_ref == 0)
442 FREE(p->p_args, M_PARGS);
443 p->p_args = NULL;
444
445 /* Cache arguments if they fit inside our allowance */
2bd9d75c 446 i = imgp->args->begin_envv - imgp->args->begin_argv;
984263bc
MD
447 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
448 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
449 M_PARGS, M_WAITOK);
450 p->p_args->ar_ref = 1;
451 p->p_args->ar_length = i;
2bd9d75c 452 bcopy(imgp->args->begin_argv, p->p_args->ar_args, i);
984263bc
MD
453 }
454
455exec_fail_dealloc:
456
457 /*
458 * free various allocated resources
459 */
460 if (imgp->firstpage)
461 exec_unmap_first_page(imgp);
462
984263bc 463 if (imgp->vp) {
984263bc 464 vrele(imgp->vp);
fad57d0e 465 imgp->vp = NULL;
984263bc
MD
466 }
467
a6f89c72
DR
468 if (error == 0) {
469 ++mycpu->gd_cnt.v_exec;
984263bc 470 return (0);
a6f89c72 471 }
984263bc
MD
472
473exec_fail:
474 /* we're done here, clear P_INEXEC */
475 p->p_flag &= ~P_INEXEC;
476 if (imgp->vmspace_destroyed) {
477 /* sorry, no more process anymore. exit gracefully */
41c20dac 478 exit1(W_EXITCODE(0, SIGABRT));
984263bc
MD
479 /* NOT REACHED */
480 return(0);
481 } else {
482 return(error);
483 }
484}
485
2bd9d75c
DRJ
486/*
487 * execve() system call.
488 */
489int
753fd850 490sys_execve(struct execve_args *uap)
2bd9d75c 491{
fad57d0e 492 struct nlookupdata nd;
2bd9d75c
DRJ
493 struct image_args args;
494 int error;
495
fad57d0e
MD
496 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
497 if (error == 0) {
498 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
499 uap->argv, uap->envv);
500 }
2abfe22f
MD
501 if (error == 0)
502 error = kern_execve(&nd, &args);
fad57d0e 503 nlookup_done(&nd);
2bd9d75c
DRJ
504 exec_free_args(&args);
505
506 /*
507 * The syscall result is returned in registers to the new program.
508 * Linux will register %edx as an atexit function and we must be
509 * sure to set it to 0. XXX
510 */
511 if (error == 0)
512 uap->sysmsg_result64 = 0;
513
514 return (error);
515}
516
984263bc 517int
41c20dac 518exec_map_first_page(struct image_params *imgp)
984263bc 519{
5fd012e0 520 int rv, i;
984263bc
MD
521 int initial_pagein;
522 vm_page_t ma[VM_INITIAL_PAGEIN];
06ecca5a 523 vm_page_t m;
984263bc
MD
524 vm_object_t object;
525
18f40545 526 if (imgp->firstpage)
984263bc 527 exec_unmap_first_page(imgp);
984263bc 528
fad57d0e 529 /*
7540ab49 530 * The file has to be mappable.
fad57d0e 531 */
7540ab49 532 if ((object = imgp->vp->v_object) == NULL)
fad57d0e 533 return (EIO);
984263bc 534
06ecca5a
MD
535 /*
536 * We shouldn't need protection for vm_page_grab() but we certainly
537 * need it for the lookup loop below (lookup/busy race), since
538 * an interrupt can unbusy and free the page before our busy check.
539 */
5fd012e0 540 crit_enter();
06ecca5a 541 m = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
984263bc 542
06ecca5a
MD
543 if ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
544 ma[0] = m;
984263bc
MD
545 initial_pagein = VM_INITIAL_PAGEIN;
546 if (initial_pagein > object->size)
547 initial_pagein = object->size;
548 for (i = 1; i < initial_pagein; i++) {
06ecca5a
MD
549 if ((m = vm_page_lookup(object, i)) != NULL) {
550 if ((m->flags & PG_BUSY) || m->busy)
984263bc 551 break;
06ecca5a 552 if (m->valid)
984263bc 553 break;
06ecca5a 554 vm_page_busy(m);
984263bc 555 } else {
06ecca5a
MD
556 m = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
557 if (m == NULL)
984263bc
MD
558 break;
559 }
06ecca5a 560 ma[i] = m;
984263bc
MD
561 }
562 initial_pagein = i;
563
06ecca5a
MD
564 /*
565 * get_pages unbusies all the requested pages except the
566 * primary page (at index 0 in this case).
567 */
984263bc 568 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
06ecca5a 569 m = vm_page_lookup(object, 0);
984263bc 570
06ecca5a
MD
571 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
572 if (m) {
573 vm_page_protect(m, VM_PROT_NONE);
574 vm_page_free(m);
984263bc 575 }
5fd012e0 576 crit_exit();
984263bc
MD
577 return EIO;
578 }
579 }
06ecca5a
MD
580 vm_page_hold(m);
581 vm_page_wakeup(m); /* unbusy the page */
5fd012e0 582 crit_exit();
984263bc 583
4f1640d6 584 imgp->firstpage = sf_buf_alloc(m, SFB_CPUPRIVATE);
18f40545 585 imgp->image_header = (void *)sf_buf_kva(imgp->firstpage);
984263bc
MD
586
587 return 0;
588}
589
590void
77153250 591exec_unmap_first_page(struct image_params *imgp)
984263bc 592{
18f40545
MD
593 vm_page_t m;
594
5fd012e0 595 crit_enter();
18f40545
MD
596 if (imgp->firstpage != NULL) {
597 m = sf_buf_page(imgp->firstpage);
598 sf_buf_free(imgp->firstpage);
984263bc 599 imgp->firstpage = NULL;
18f40545 600 imgp->image_header = NULL;
e4ba7818 601 vm_page_unhold(m);
984263bc 602 }
5fd012e0 603 crit_exit();
984263bc
MD
604}
605
606/*
607 * Destroy old address space, and allocate a new stack
608 * The new stack is only SGROWSIZ large because it is grown
609 * automatically in trap.c.
610 */
611int
29802dbb 612exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
984263bc
MD
613{
614 int error;
615 struct vmspace *vmspace = imgp->proc->p_vmspace;
616 vm_offset_t stack_addr = USRSTACK - maxssiz;
29802dbb 617 vm_map_t map;
984263bc
MD
618
619 imgp->vmspace_destroyed = 1;
620
621 /*
622 * Prevent a pending AIO from modifying the new address space.
623 */
624 aio_proc_rundown(imgp->proc);
625
626 /*
627 * Blow away entire process VM, if address space not shared,
628 * otherwise, create a new VM space so that other threads are
29802dbb
MD
629 * not disrupted. If we are execing a resident vmspace we
630 * create a duplicate of it and remap the stack.
239b4df9
MD
631 *
632 * The exitingcnt test is not strictly necessary but has been
633 * included for code sanity (to make the code more deterministic).
984263bc 634 */
29802dbb
MD
635 map = &vmspace->vm_map;
636 if (vmcopy) {
637 vmspace_exec(imgp->proc, vmcopy);
638 vmspace = imgp->proc->p_vmspace;
639 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
640 map = &vmspace->vm_map;
239b4df9 641 } else if (vmspace->vm_refcnt == 1 && vmspace->vm_exitingcnt == 0) {
fef0fdf2 642 shmexit(vmspace);
239b4df9 643 if (vmspace->vm_upcalls)
4170cc8b 644 upc_release(vmspace, &imgp->proc->p_lwp);
984263bc
MD
645 pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
646 vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
647 } else {
29802dbb 648 vmspace_exec(imgp->proc, NULL);
984263bc
MD
649 vmspace = imgp->proc->p_vmspace;
650 map = &vmspace->vm_map;
651 }
652
653 /* Allocate a new stack */
654 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
655 VM_PROT_ALL, VM_PROT_ALL, 0);
656 if (error)
657 return (error);
658
659 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
660 * VM_STACK case, but they are still used to monitor the size of the
661 * process stack so we can check the stack rlimit.
662 */
663 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
664 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
665
666 return(0);
667}
668
669/*
670 * Copy out argument and environment strings from the old process
671 * address space into the temporary string buffer.
672 */
673int
2bd9d75c 674exec_copyin_args(struct image_args *args, char *fname,
2abfe22f 675 enum exec_path_segflg segflg, char **argv, char **envv)
984263bc 676{
984263bc 677 char *argp, *envp;
2bd9d75c 678 int error = 0;
984263bc
MD
679 size_t length;
680
2abfe22f 681 bzero(args, sizeof(*args));
2bd9d75c
DRJ
682 args->buf = (char *) kmem_alloc_wait(exec_map, PATH_MAX + ARG_MAX);
683 if (args->buf == NULL)
684 return (ENOMEM);
685 args->begin_argv = args->buf;
686 args->endp = args->begin_argv;
687 args->space = ARG_MAX;
2bd9d75c
DRJ
688
689 args->fname = args->buf + ARG_MAX;
690
984263bc 691 /*
2bd9d75c 692 * Copy the file name.
984263bc 693 */
2bd9d75c
DRJ
694 if (segflg == PATH_SYSSPACE) {
695 error = copystr(fname, args->fname, PATH_MAX, &length);
696 } else if (segflg == PATH_USERSPACE) {
697 error = copyinstr(fname, args->fname, PATH_MAX, &length);
698 }
984263bc 699
2bd9d75c 700 /*
66be6566
MD
701 * Extract argument strings. argv may not be NULL. The argv
702 * array is terminated by a NULL entry. We special-case the
703 * situation where argv[0] is NULL by passing { filename, NULL }
704 * to the new program to guarentee that the interpreter knows what
705 * file to open in case we exec an interpreted file. Note that
706 * a NULL argv[0] terminates the argv[] array.
707 *
708 * XXX the special-casing of argv[0] is historical and needs to be
709 * revisited.
2bd9d75c 710 */
66be6566
MD
711 if (argv == NULL)
712 error = EFAULT;
713 if (error == 0) {
714 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
715 if (argp == (caddr_t)-1) {
2bd9d75c 716 error = EFAULT;
66be6566 717 break;
2bd9d75c
DRJ
718 }
719 error = copyinstr(argp, args->endp,
2abfe22f 720 args->space, &length);
66be6566
MD
721 if (error) {
722 if (error == ENAMETOOLONG)
723 error = E2BIG;
724 break;
725 }
2bd9d75c
DRJ
726 args->space -= length;
727 args->endp += length;
728 args->argc++;
984263bc 729 }
66be6566
MD
730 if (args->argc == 0 && error == 0) {
731 length = strlen(args->fname) + 1;
732 if (length > args->space) {
733 error = E2BIG;
734 } else {
735 bcopy(args->fname, args->endp, length);
736 args->space -= length;
737 args->endp += length;
738 args->argc++;
739 }
740 }
984263bc
MD
741 }
742
2bd9d75c 743 args->begin_envv = args->endp;
984263bc
MD
744
745 /*
66be6566 746 * extract environment strings. envv may be NULL.
984263bc 747 */
2abfe22f 748 if (envv && error == 0) {
984263bc 749 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
2bd9d75c
DRJ
750 if (envp == (caddr_t) -1) {
751 error = EFAULT;
66be6566 752 break;
984263bc 753 }
2bd9d75c
DRJ
754 error = copyinstr(envp, args->endp, args->space,
755 &length);
66be6566
MD
756 if (error) {
757 if (error == ENAMETOOLONG)
758 error = E2BIG;
759 break;
760 }
2bd9d75c
DRJ
761 args->space -= length;
762 args->endp += length;
763 args->envc++;
984263bc
MD
764 }
765 }
2bd9d75c
DRJ
766 return (error);
767}
768
769void
770exec_free_args(struct image_args *args)
771{
2abfe22f
MD
772 if (args->buf) {
773 kmem_free_wakeup(exec_map,
774 (vm_offset_t)args->buf, PATH_MAX + ARG_MAX);
775 args->buf = NULL;
776 }
984263bc
MD
777}
778
779/*
780 * Copy strings out to the new process address space, constructing
781 * new arg and env vector tables. Return a pointer to the base
782 * so that it can be used as the initial stack pointer.
783 */
784register_t *
2bd9d75c 785exec_copyout_strings(struct image_params *imgp)
984263bc 786{
17a7ca1f 787 int argc, envc, sgap;
984263bc
MD
788 char **vectp;
789 char *stringp, *destp;
790 register_t *stack_base;
791 struct ps_strings *arginfo;
792 int szsigcode;
793
794 /*
795 * Calculate string base and vector table pointers.
796 * Also deal with signal trampoline code for this exec type.
797 */
798 arginfo = (struct ps_strings *)PS_STRINGS;
799 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
17a7ca1f 800 if (stackgap_random != 0)
0ced1954 801 sgap = ALIGN(karc4random() & (stackgap_random - 1));
17a7ca1f
MD
802 else
803 sgap = 0;
804 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
2bd9d75c 805 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
984263bc
MD
806
807 /*
808 * install sigcode
809 */
810 if (szsigcode)
811 copyout(imgp->proc->p_sysent->sv_sigcode,
2bd9d75c 812 ((caddr_t)arginfo - szsigcode), szsigcode);
984263bc
MD
813
814 /*
815 * If we have a valid auxargs ptr, prepare some room
816 * on the stack.
7ec1918b 817 *
984263bc
MD
818 * The '+ 2' is for the null pointers at the end of each of the
819 * arg and env vector sets, and 'AT_COUNT*2' is room for the
820 * ELF Auxargs data.
821 */
7ec1918b
MD
822 if (imgp->auxargs) {
823 vectp = (char **)(destp - (imgp->args->argc +
824 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
825 } else {
2bd9d75c 826 vectp = (char **)(destp - (imgp->args->argc +
7ec1918b
MD
827 imgp->args->envc + 2) * sizeof(char*));
828 }
829
984263bc 830 /*
31bc42c8
MD
831 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
832 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
984263bc 833 */
984263bc
MD
834
835 /*
836 * vectp also becomes our initial stack base
837 */
838 stack_base = (register_t *)vectp;
839
2bd9d75c
DRJ
840 stringp = imgp->args->begin_argv;
841 argc = imgp->args->argc;
842 envc = imgp->args->envc;
984263bc
MD
843
844 /*
845 * Copy out strings - arguments and environment.
846 */
2bd9d75c 847 copyout(stringp, destp, ARG_MAX - imgp->args->space);
984263bc
MD
848
849 /*
850 * Fill in "ps_strings" struct for ps, w, etc.
851 */
852 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
853 suword(&arginfo->ps_nargvstr, argc);
854
855 /*
856 * Fill in argument portion of vector table.
857 */
858 for (; argc > 0; --argc) {
859 suword(vectp++, (long)(intptr_t)destp);
860 while (*stringp++ != 0)
861 destp++;
862 destp++;
863 }
864
7a44d1cb 865 /* a null vector table pointer separates the argp's from the envp's */
984263bc
MD
866 suword(vectp++, 0);
867
868 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
869 suword(&arginfo->ps_nenvstr, envc);
870
871 /*
872 * Fill in environment portion of vector table.
873 */
874 for (; envc > 0; --envc) {
875 suword(vectp++, (long)(intptr_t)destp);
876 while (*stringp++ != 0)
877 destp++;
878 destp++;
879 }
880
881 /* end of vector table is a null pointer */
882 suword(vectp, 0);
883
884 return (stack_base);
885}
886
887/*
888 * Check permissions of file to execute.
889 * Return 0 for success or error code on failure.
890 */
891int
77153250 892exec_check_permissions(struct image_params *imgp)
984263bc
MD
893{
894 struct proc *p = imgp->proc;
895 struct vnode *vp = imgp->vp;
896 struct vattr *attr = imgp->attr;
897 int error;
898
899 /* Get file attributes */
87de5057 900 error = VOP_GETATTR(vp, attr);
984263bc
MD
901 if (error)
902 return (error);
903
904 /*
905 * 1) Check if file execution is disabled for the filesystem that this
906 * file resides on.
907 * 2) Insure that at least one execute bit is on - otherwise root
908 * will always succeed, and we don't want to happen unless the
909 * file really is executable.
910 * 3) Insure that the file is a regular file.
911 */
912 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
913 ((attr->va_mode & 0111) == 0) ||
914 (attr->va_type != VREG)) {
915 return (EACCES);
916 }
917
918 /*
919 * Zero length files can't be exec'd
920 */
921 if (attr->va_size == 0)
922 return (ENOEXEC);
923
924 /*
925 * Check for execute permission to file based on current credentials.
926 */
87de5057 927 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
984263bc
MD
928 if (error)
929 return (error);
930
931 /*
932 * Check number of open-for-writes on the file and deny execution
933 * if there are any.
934 */
935 if (vp->v_writecount)
936 return (ETXTBSY);
937
938 /*
7540ab49
MD
939 * Call filesystem specific open routine, which allows us to read,
940 * write, and mmap the file. Without the VOP_OPEN we can only
941 * stat the file.
984263bc 942 */
87de5057 943 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
984263bc
MD
944 if (error)
945 return (error);
946
947 return (0);
948}
949
950/*
951 * Exec handler registration
952 */
953int
77153250 954exec_register(const struct execsw *execsw_arg)
984263bc
MD
955{
956 const struct execsw **es, **xs, **newexecsw;
957 int count = 2; /* New slot and trailing NULL */
958
959 if (execsw)
960 for (es = execsw; *es; es++)
961 count++;
962 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
963 if (newexecsw == NULL)
964 return ENOMEM;
965 xs = newexecsw;
966 if (execsw)
967 for (es = execsw; *es; es++)
968 *xs++ = *es;
969 *xs++ = execsw_arg;
970 *xs = NULL;
971 if (execsw)
972 free(execsw, M_TEMP);
973 execsw = newexecsw;
974 return 0;
975}
976
977int
77153250 978exec_unregister(const struct execsw *execsw_arg)
984263bc
MD
979{
980 const struct execsw **es, **xs, **newexecsw;
981 int count = 1;
982
983 if (execsw == NULL)
ae7e74d4 984 panic("unregister with no handlers left?");
984263bc
MD
985
986 for (es = execsw; *es; es++) {
987 if (*es == execsw_arg)
988 break;
989 }
990 if (*es == NULL)
991 return ENOENT;
992 for (es = execsw; *es; es++)
993 if (*es != execsw_arg)
994 count++;
995 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
996 if (newexecsw == NULL)
997 return ENOMEM;
998 xs = newexecsw;
999 for (es = execsw; *es; es++)
1000 if (*es != execsw_arg)
1001 *xs++ = *es;
1002 *xs = NULL;
1003 if (execsw)
1004 free(execsw, M_TEMP);
1005 execsw = newexecsw;
1006 return 0;
1007}