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 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include "opt_hwpmc_hooks.h" 31 #include "opt_kdtrace.h" 32 #include "opt_ktrace.h" 33 #include "opt_vm.h" 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/eventhandler.h> 38 #include <sys/lock.h> 39 #include <sys/mutex.h> 40 #include <sys/sysproto.h> 41 #include <sys/signalvar.h> 42 #include <sys/kernel.h> 43 #include <sys/mount.h> 44 #include <sys/filedesc.h> 45 #include <sys/fcntl.h> 46 #include <sys/acct.h> 47 #include <sys/exec.h> 48 #include <sys/imgact.h> 49 #include <sys/imgact_elf.h> 50 #include <sys/wait.h> 51 #include <sys/malloc.h> 52 #include <sys/priv.h> 53 #include <sys/proc.h> 54 #include <sys/pioctl.h> 55 #include <sys/namei.h> 56 #include <sys/resourcevar.h> 57 #include <sys/sdt.h> 58 #include <sys/sf_buf.h> 59 #include <sys/syscallsubr.h> 60 #include <sys/sysent.h> 61 #include <sys/shm.h> 62 #include <sys/sysctl.h> 63 #include <sys/vnode.h> 64 #include <sys/stat.h> 65 #ifdef KTRACE 66 #include <sys/ktrace.h> 67 #endif 68 69 #include <vm/vm.h> 70 #include <vm/vm_param.h> 71 #include <vm/pmap.h> 72 #include <vm/vm_page.h> 73 #include <vm/vm_map.h> 74 #include <vm/vm_kern.h> 75 #include <vm/vm_extern.h> 76 #include <vm/vm_object.h> 77 #include <vm/vm_pager.h> 78 79 #ifdef HWPMC_HOOKS 80 #include <sys/pmckern.h> 81 #endif 82 83 #include <machine/reg.h> 84 85 #include <security/audit/audit.h> 86 #include <security/mac/mac_framework.h> 87 88 #ifdef KDTRACE_HOOKS 89 #include <sys/dtrace_bsd.h> 90 dtrace_execexit_func_t dtrace_fasttrap_exec; 91 #endif 92 93 SDT_PROVIDER_DECLARE(proc); 94 SDT_PROBE_DEFINE(proc, kernel, , exec); 95 SDT_PROBE_ARGTYPE(proc, kernel, , exec, 0, "char *"); 96 SDT_PROBE_DEFINE(proc, kernel, , exec_failure); 97 SDT_PROBE_ARGTYPE(proc, kernel, , exec_failure, 0, "int"); 98 SDT_PROBE_DEFINE(proc, kernel, , exec_success); 99 SDT_PROBE_ARGTYPE(proc, kernel, , exec_success, 0, "char *"); 100 101 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments"); 102 103 static int sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS); 104 static int sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS); 105 static int sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS); 106 static int do_execve(struct thread *td, struct image_args *args, 107 struct mac *mac_p); 108 109 /* XXX This should be vm_size_t. */ 110 SYSCTL_PROC(_kern, KERN_PS_STRINGS, ps_strings, CTLTYPE_ULONG|CTLFLAG_RD, 111 NULL, 0, sysctl_kern_ps_strings, "LU", ""); 112 113 /* XXX This should be vm_size_t. */ 114 SYSCTL_PROC(_kern, KERN_USRSTACK, usrstack, CTLTYPE_ULONG|CTLFLAG_RD, 115 NULL, 0, sysctl_kern_usrstack, "LU", ""); 116 117 SYSCTL_PROC(_kern, OID_AUTO, stackprot, CTLTYPE_INT|CTLFLAG_RD, 118 NULL, 0, sysctl_kern_stackprot, "I", ""); 119 120 u_long ps_arg_cache_limit = PAGE_SIZE / 16; 121 SYSCTL_ULONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW, 122 &ps_arg_cache_limit, 0, ""); 123 124 static int map_at_zero = 0; 125 TUNABLE_INT("security.bsd.map_at_zero", &map_at_zero); 126 SYSCTL_INT(_security_bsd, OID_AUTO, map_at_zero, CTLFLAG_RW, &map_at_zero, 0, 127 "Permit processes to map an object at virtual address 0."); 128 129 static int 130 sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS) 131 { 132 struct proc *p; 133 int error; 134 135 p = curproc; 136 #ifdef SCTL_MASK32 137 if (req->flags & SCTL_MASK32) { 138 unsigned int val; 139 val = (unsigned int)p->p_sysent->sv_psstrings; 140 error = SYSCTL_OUT(req, &val, sizeof(val)); 141 } else 142 #endif 143 error = SYSCTL_OUT(req, &p->p_sysent->sv_psstrings, 144 sizeof(p->p_sysent->sv_psstrings)); 145 return error; 146 } 147 148 static int 149 sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS) 150 { 151 struct proc *p; 152 int error; 153 154 p = curproc; 155 #ifdef SCTL_MASK32 156 if (req->flags & SCTL_MASK32) { 157 unsigned int val; 158 val = (unsigned int)p->p_sysent->sv_usrstack; 159 error = SYSCTL_OUT(req, &val, sizeof(val)); 160 } else 161 #endif 162 error = SYSCTL_OUT(req, &p->p_sysent->sv_usrstack, 163 sizeof(p->p_sysent->sv_usrstack)); 164 return error; 165 } 166 167 static int 168 sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS) 169 { 170 struct proc *p; 171 172 p = curproc; 173 return (SYSCTL_OUT(req, &p->p_sysent->sv_stackprot, 174 sizeof(p->p_sysent->sv_stackprot))); 175 } 176 177 /* 178 * Each of the items is a pointer to a `const struct execsw', hence the 179 * double pointer here. 180 */ 181 static const struct execsw **execsw; 182 183 #ifndef _SYS_SYSPROTO_H_ 184 struct execve_args { 185 char *fname; 186 char **argv; 187 char **envv; 188 }; 189 #endif 190 191 int 192 execve(td, uap) 193 struct thread *td; 194 struct execve_args /* { 195 char *fname; 196 char **argv; 197 char **envv; 198 } */ *uap; 199 { 200 int error; 201 struct image_args args; 202 203 error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE, 204 uap->argv, uap->envv); 205 if (error == 0) 206 error = kern_execve(td, &args, NULL); 207 return (error); 208 } 209 210 #ifndef _SYS_SYSPROTO_H_ 211 struct fexecve_args { 212 int fd; 213 char **argv; 214 char **envv; 215 } 216 #endif 217 int 218 fexecve(struct thread *td, struct fexecve_args *uap) 219 { 220 int error; 221 struct image_args args; 222 223 error = exec_copyin_args(&args, NULL, UIO_SYSSPACE, 224 uap->argv, uap->envv); 225 if (error == 0) { 226 args.fd = uap->fd; 227 error = kern_execve(td, &args, NULL); 228 } 229 return (error); 230 } 231 232 #ifndef _SYS_SYSPROTO_H_ 233 struct __mac_execve_args { 234 char *fname; 235 char **argv; 236 char **envv; 237 struct mac *mac_p; 238 }; 239 #endif 240 241 int 242 __mac_execve(td, uap) 243 struct thread *td; 244 struct __mac_execve_args /* { 245 char *fname; 246 char **argv; 247 char **envv; 248 struct mac *mac_p; 249 } */ *uap; 250 { 251 #ifdef MAC 252 int error; 253 struct image_args args; 254 255 error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE, 256 uap->argv, uap->envv); 257 if (error == 0) 258 error = kern_execve(td, &args, uap->mac_p); 259 return (error); 260 #else 261 return (ENOSYS); 262 #endif 263 } 264 265 /* 266 * XXX: kern_execve has the astonishing property of not always returning to 267 * the caller. If sufficiently bad things happen during the call to 268 * do_execve(), it can end up calling exit1(); as a result, callers must 269 * avoid doing anything which they might need to undo (e.g., allocating 270 * memory). 271 */ 272 int 273 kern_execve(td, args, mac_p) 274 struct thread *td; 275 struct image_args *args; 276 struct mac *mac_p; 277 { 278 struct proc *p = td->td_proc; 279 int error; 280 281 AUDIT_ARG_ARGV(args->begin_argv, args->argc, 282 args->begin_envv - args->begin_argv); 283 AUDIT_ARG_ENVV(args->begin_envv, args->envc, 284 args->endp - args->begin_envv); 285 if (p->p_flag & P_HADTHREADS) { 286 PROC_LOCK(p); 287 if (thread_single(SINGLE_BOUNDARY)) { 288 PROC_UNLOCK(p); 289 exec_free_args(args); 290 return (ERESTART); /* Try again later. */ 291 } 292 PROC_UNLOCK(p); 293 } 294 295 error = do_execve(td, args, mac_p); 296 297 if (p->p_flag & P_HADTHREADS) { 298 PROC_LOCK(p); 299 /* 300 * If success, we upgrade to SINGLE_EXIT state to 301 * force other threads to suicide. 302 */ 303 if (error == 0) 304 thread_single(SINGLE_EXIT); 305 else 306 thread_single_end(); 307 PROC_UNLOCK(p); 308 } 309 310 return (error); 311 } 312 313 /* 314 * In-kernel implementation of execve(). All arguments are assumed to be 315 * userspace pointers from the passed thread. 316 */ 317 static int 318 do_execve(td, args, mac_p) 319 struct thread *td; 320 struct image_args *args; 321 struct mac *mac_p; 322 { 323 struct proc *p = td->td_proc; 324 struct nameidata nd; 325 struct ucred *newcred = NULL, *oldcred; 326 struct uidinfo *euip; 327 register_t *stack_base; 328 int error, i; 329 struct image_params image_params, *imgp; 330 struct vattr attr; 331 int (*img_first)(struct image_params *); 332 struct pargs *oldargs = NULL, *newargs = NULL; 333 struct sigacts *oldsigacts, *newsigacts; 334 #ifdef KTRACE 335 struct vnode *tracevp = NULL; 336 struct ucred *tracecred = NULL; 337 #endif 338 struct vnode *textvp = NULL, *binvp = NULL; 339 int credential_changing; 340 int vfslocked; 341 int textset; 342 #ifdef MAC 343 struct label *interpvplabel = NULL; 344 int will_transition; 345 #endif 346 #ifdef HWPMC_HOOKS 347 struct pmckern_procexec pe; 348 #endif 349 static const char fexecv_proc_title[] = "(fexecv)"; 350 351 vfslocked = 0; 352 imgp = &image_params; 353 354 /* 355 * Lock the process and set the P_INEXEC flag to indicate that 356 * it should be left alone until we're done here. This is 357 * necessary to avoid race conditions - e.g. in ptrace() - 358 * that might allow a local user to illicitly obtain elevated 359 * privileges. 360 */ 361 PROC_LOCK(p); 362 KASSERT((p->p_flag & P_INEXEC) == 0, 363 ("%s(): process already has P_INEXEC flag", __func__)); 364 p->p_flag |= P_INEXEC; 365 PROC_UNLOCK(p); 366 367 /* 368 * Initialize part of the common data 369 */ 370 imgp->proc = p; 371 imgp->execlabel = NULL; 372 imgp->attr = &attr; 373 imgp->entry_addr = 0; 374 imgp->reloc_base = 0; 375 imgp->vmspace_destroyed = 0; 376 imgp->interpreted = 0; 377 imgp->opened = 0; 378 imgp->interpreter_name = NULL; 379 imgp->auxargs = NULL; 380 imgp->vp = NULL; 381 imgp->object = NULL; 382 imgp->firstpage = NULL; 383 imgp->ps_strings = 0; 384 imgp->auxarg_size = 0; 385 imgp->args = args; 386 imgp->execpath = imgp->freepath = NULL; 387 imgp->execpathp = 0; 388 imgp->canary = 0; 389 imgp->canarylen = 0; 390 imgp->pagesizes = 0; 391 imgp->pagesizeslen = 0; 392 393 #ifdef MAC 394 error = mac_execve_enter(imgp, mac_p); 395 if (error) 396 goto exec_fail; 397 #endif 398 399 imgp->image_header = NULL; 400 401 /* 402 * Translate the file name. namei() returns a vnode pointer 403 * in ni_vp amoung other things. 404 * 405 * XXXAUDIT: It would be desirable to also audit the name of the 406 * interpreter if this is an interpreted binary. 407 */ 408 if (args->fname != NULL) { 409 NDINIT(&nd, LOOKUP, ISOPEN | LOCKLEAF | FOLLOW | SAVENAME 410 | MPSAFE | AUDITVNODE1, UIO_SYSSPACE, args->fname, td); 411 } 412 413 SDT_PROBE(proc, kernel, , exec, args->fname, 0, 0, 0, 0 ); 414 415 interpret: 416 if (args->fname != NULL) { 417 error = namei(&nd); 418 if (error) 419 goto exec_fail; 420 421 vfslocked = NDHASGIANT(&nd); 422 binvp = nd.ni_vp; 423 imgp->vp = binvp; 424 } else { 425 AUDIT_ARG_FD(args->fd); 426 error = fgetvp(td, args->fd, &binvp); 427 if (error) 428 goto exec_fail; 429 vfslocked = VFS_LOCK_GIANT(binvp->v_mount); 430 vn_lock(binvp, LK_EXCLUSIVE | LK_RETRY); 431 AUDIT_ARG_VNODE1(binvp); 432 imgp->vp = binvp; 433 } 434 435 /* 436 * Check file permissions (also 'opens' file) 437 */ 438 error = exec_check_permissions(imgp); 439 if (error) 440 goto exec_fail_dealloc; 441 442 imgp->object = imgp->vp->v_object; 443 if (imgp->object != NULL) 444 vm_object_reference(imgp->object); 445 446 /* 447 * Set VV_TEXT now so no one can write to the executable while we're 448 * activating it. 449 * 450 * Remember if this was set before and unset it in case this is not 451 * actually an executable image. 452 */ 453 textset = imgp->vp->v_vflag & VV_TEXT; 454 imgp->vp->v_vflag |= VV_TEXT; 455 456 error = exec_map_first_page(imgp); 457 if (error) 458 goto exec_fail_dealloc; 459 460 imgp->proc->p_osrel = 0; 461 /* 462 * If the current process has a special image activator it 463 * wants to try first, call it. For example, emulating shell 464 * scripts differently. 465 */ 466 error = -1; 467 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL) 468 error = img_first(imgp); 469 470 /* 471 * Loop through the list of image activators, calling each one. 472 * An activator returns -1 if there is no match, 0 on success, 473 * and an error otherwise. 474 */ 475 for (i = 0; error == -1 && execsw[i]; ++i) { 476 if (execsw[i]->ex_imgact == NULL || 477 execsw[i]->ex_imgact == img_first) { 478 continue; 479 } 480 error = (*execsw[i]->ex_imgact)(imgp); 481 } 482 483 if (error) { 484 if (error == -1) { 485 if (textset == 0) 486 imgp->vp->v_vflag &= ~VV_TEXT; 487 error = ENOEXEC; 488 } 489 goto exec_fail_dealloc; 490 } 491 492 /* 493 * Special interpreter operation, cleanup and loop up to try to 494 * activate the interpreter. 495 */ 496 if (imgp->interpreted) { 497 exec_unmap_first_page(imgp); 498 /* 499 * VV_TEXT needs to be unset for scripts. There is a short 500 * period before we determine that something is a script where 501 * VV_TEXT will be set. The vnode lock is held over this 502 * entire period so nothing should illegitimately be blocked. 503 */ 504 imgp->vp->v_vflag &= ~VV_TEXT; 505 /* free name buffer and old vnode */ 506 if (args->fname != NULL) 507 NDFREE(&nd, NDF_ONLY_PNBUF); 508 #ifdef MAC 509 mac_execve_interpreter_enter(binvp, &interpvplabel); 510 #endif 511 if (imgp->opened) { 512 VOP_CLOSE(binvp, FREAD, td->td_ucred, td); 513 imgp->opened = 0; 514 } 515 vput(binvp); 516 vm_object_deallocate(imgp->object); 517 imgp->object = NULL; 518 VFS_UNLOCK_GIANT(vfslocked); 519 vfslocked = 0; 520 /* set new name to that of the interpreter */ 521 NDINIT(&nd, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME | MPSAFE, 522 UIO_SYSSPACE, imgp->interpreter_name, td); 523 args->fname = imgp->interpreter_name; 524 goto interpret; 525 } 526 527 /* 528 * NB: We unlock the vnode here because it is believed that none 529 * of the sv_copyout_strings/sv_fixup operations require the vnode. 530 */ 531 VOP_UNLOCK(imgp->vp, 0); 532 533 /* 534 * Do the best to calculate the full path to the image file. 535 */ 536 if (imgp->auxargs != NULL && 537 ((args->fname != NULL && args->fname[0] == '/') || 538 vn_fullpath(td, imgp->vp, &imgp->execpath, &imgp->freepath) != 0)) 539 imgp->execpath = args->fname; 540 541 /* 542 * Copy out strings (args and env) and initialize stack base 543 */ 544 if (p->p_sysent->sv_copyout_strings) 545 stack_base = (*p->p_sysent->sv_copyout_strings)(imgp); 546 else 547 stack_base = exec_copyout_strings(imgp); 548 549 /* 550 * If custom stack fixup routine present for this process 551 * let it do the stack setup. 552 * Else stuff argument count as first item on stack 553 */ 554 if (p->p_sysent->sv_fixup != NULL) 555 (*p->p_sysent->sv_fixup)(&stack_base, imgp); 556 else 557 suword(--stack_base, imgp->args->argc); 558 559 /* 560 * For security and other reasons, the file descriptor table cannot 561 * be shared after an exec. 562 */ 563 fdunshare(p, td); 564 565 /* 566 * Malloc things before we need locks. 567 */ 568 newcred = crget(); 569 euip = uifind(attr.va_uid); 570 i = imgp->args->begin_envv - imgp->args->begin_argv; 571 /* Cache arguments if they fit inside our allowance */ 572 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) { 573 newargs = pargs_alloc(i); 574 bcopy(imgp->args->begin_argv, newargs->ar_args, i); 575 } 576 577 /* close files on exec */ 578 fdcloseexec(td); 579 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY); 580 581 /* Get a reference to the vnode prior to locking the proc */ 582 VREF(binvp); 583 584 /* 585 * For security and other reasons, signal handlers cannot 586 * be shared after an exec. The new process gets a copy of the old 587 * handlers. In execsigs(), the new process will have its signals 588 * reset. 589 */ 590 PROC_LOCK(p); 591 oldcred = crcopysafe(p, newcred); 592 if (sigacts_shared(p->p_sigacts)) { 593 oldsigacts = p->p_sigacts; 594 PROC_UNLOCK(p); 595 newsigacts = sigacts_alloc(); 596 sigacts_copy(newsigacts, oldsigacts); 597 PROC_LOCK(p); 598 p->p_sigacts = newsigacts; 599 } else 600 oldsigacts = NULL; 601 602 /* Stop profiling */ 603 stopprofclock(p); 604 605 /* reset caught signals */ 606 execsigs(p); 607 608 /* name this process - nameiexec(p, ndp) */ 609 bzero(p->p_comm, sizeof(p->p_comm)); 610 if (args->fname) 611 bcopy(nd.ni_cnd.cn_nameptr, p->p_comm, 612 min(nd.ni_cnd.cn_namelen, MAXCOMLEN)); 613 else if (vn_commname(binvp, p->p_comm, sizeof(p->p_comm)) != 0) 614 bcopy(fexecv_proc_title, p->p_comm, sizeof(fexecv_proc_title)); 615 bcopy(p->p_comm, td->td_name, sizeof(td->td_name)); 616 617 /* 618 * mark as execed, wakeup the process that vforked (if any) and tell 619 * it that it now has its own resources back 620 */ 621 p->p_flag |= P_EXEC; 622 if (p->p_pptr && (p->p_flag & P_PPWAIT)) { 623 p->p_flag &= ~P_PPWAIT; 624 cv_broadcast(&p->p_pwait); 625 } 626 627 /* 628 * Implement image setuid/setgid. 629 * 630 * Don't honor setuid/setgid if the filesystem prohibits it or if 631 * the process is being traced. 632 * 633 * XXXMAC: For the time being, use NOSUID to also prohibit 634 * transitions on the file system. 635 */ 636 credential_changing = 0; 637 credential_changing |= (attr.va_mode & S_ISUID) && oldcred->cr_uid != 638 attr.va_uid; 639 credential_changing |= (attr.va_mode & S_ISGID) && oldcred->cr_gid != 640 attr.va_gid; 641 #ifdef MAC 642 will_transition = mac_vnode_execve_will_transition(oldcred, imgp->vp, 643 interpvplabel, imgp); 644 credential_changing |= will_transition; 645 #endif 646 647 if (credential_changing && 648 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 && 649 (p->p_flag & P_TRACED) == 0) { 650 /* 651 * Turn off syscall tracing for set-id programs, except for 652 * root. Record any set-id flags first to make sure that 653 * we do not regain any tracing during a possible block. 654 */ 655 setsugid(p); 656 657 #ifdef KTRACE 658 if (p->p_tracevp != NULL && 659 priv_check_cred(oldcred, PRIV_DEBUG_DIFFCRED, 0)) { 660 mtx_lock(&ktrace_mtx); 661 p->p_traceflag = 0; 662 tracevp = p->p_tracevp; 663 p->p_tracevp = NULL; 664 tracecred = p->p_tracecred; 665 p->p_tracecred = NULL; 666 mtx_unlock(&ktrace_mtx); 667 } 668 #endif 669 /* 670 * Close any file descriptors 0..2 that reference procfs, 671 * then make sure file descriptors 0..2 are in use. 672 * 673 * setugidsafety() may call closef() and then pfind() 674 * which may grab the process lock. 675 * fdcheckstd() may call falloc() which may block to 676 * allocate memory, so temporarily drop the process lock. 677 */ 678 PROC_UNLOCK(p); 679 VOP_UNLOCK(imgp->vp, 0); 680 setugidsafety(td); 681 error = fdcheckstd(td); 682 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY); 683 if (error != 0) 684 goto done1; 685 PROC_LOCK(p); 686 /* 687 * Set the new credentials. 688 */ 689 if (attr.va_mode & S_ISUID) 690 change_euid(newcred, euip); 691 if (attr.va_mode & S_ISGID) 692 change_egid(newcred, attr.va_gid); 693 #ifdef MAC 694 if (will_transition) { 695 mac_vnode_execve_transition(oldcred, newcred, imgp->vp, 696 interpvplabel, imgp); 697 } 698 #endif 699 /* 700 * Implement correct POSIX saved-id behavior. 701 * 702 * XXXMAC: Note that the current logic will save the 703 * uid and gid if a MAC domain transition occurs, even 704 * though maybe it shouldn't. 705 */ 706 change_svuid(newcred, newcred->cr_uid); 707 change_svgid(newcred, newcred->cr_gid); 708 p->p_ucred = newcred; 709 newcred = NULL; 710 } else { 711 if (oldcred->cr_uid == oldcred->cr_ruid && 712 oldcred->cr_gid == oldcred->cr_rgid) 713 p->p_flag &= ~P_SUGID; 714 /* 715 * Implement correct POSIX saved-id behavior. 716 * 717 * XXX: It's not clear that the existing behavior is 718 * POSIX-compliant. A number of sources indicate that the 719 * saved uid/gid should only be updated if the new ruid is 720 * not equal to the old ruid, or the new euid is not equal 721 * to the old euid and the new euid is not equal to the old 722 * ruid. The FreeBSD code always updates the saved uid/gid. 723 * Also, this code uses the new (replaced) euid and egid as 724 * the source, which may or may not be the right ones to use. 725 */ 726 if (oldcred->cr_svuid != oldcred->cr_uid || 727 oldcred->cr_svgid != oldcred->cr_gid) { 728 change_svuid(newcred, newcred->cr_uid); 729 change_svgid(newcred, newcred->cr_gid); 730 p->p_ucred = newcred; 731 newcred = NULL; 732 } 733 } 734 735 /* 736 * Store the vp for use in procfs. This vnode was referenced prior 737 * to locking the proc lock. 738 */ 739 textvp = p->p_textvp; 740 p->p_textvp = binvp; 741 742 #ifdef KDTRACE_HOOKS 743 /* 744 * Tell the DTrace fasttrap provider about the exec if it 745 * has declared an interest. 746 */ 747 if (dtrace_fasttrap_exec) 748 dtrace_fasttrap_exec(p); 749 #endif 750 751 /* 752 * Notify others that we exec'd, and clear the P_INEXEC flag 753 * as we're now a bona fide freshly-execed process. 754 */ 755 KNOTE_LOCKED(&p->p_klist, NOTE_EXEC); 756 p->p_flag &= ~P_INEXEC; 757 758 /* 759 * If tracing the process, trap to the debugger so that 760 * breakpoints can be set before the program executes. We 761 * have to use tdsignal() to deliver the signal to the current 762 * thread since any other threads in this process will exit if 763 * execve() succeeds. 764 */ 765 if (p->p_flag & P_TRACED) 766 tdsignal(td, SIGTRAP); 767 768 /* clear "fork but no exec" flag, as we _are_ execing */ 769 p->p_acflag &= ~AFORK; 770 771 /* 772 * Free any previous argument cache and replace it with 773 * the new argument cache, if any. 774 */ 775 oldargs = p->p_args; 776 p->p_args = newargs; 777 newargs = NULL; 778 779 #ifdef HWPMC_HOOKS 780 /* 781 * Check if system-wide sampling is in effect or if the 782 * current process is using PMCs. If so, do exec() time 783 * processing. This processing needs to happen AFTER the 784 * P_INEXEC flag is cleared. 785 * 786 * The proc lock needs to be released before taking the PMC 787 * SX. 788 */ 789 if (PMC_SYSTEM_SAMPLING_ACTIVE() || PMC_PROC_IS_USING_PMCS(p)) { 790 PROC_UNLOCK(p); 791 VOP_UNLOCK(imgp->vp, 0); 792 pe.pm_credentialschanged = credential_changing; 793 pe.pm_entryaddr = imgp->entry_addr; 794 795 PMC_CALL_HOOK_X(td, PMC_FN_PROCESS_EXEC, (void *) &pe); 796 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY); 797 } else 798 PROC_UNLOCK(p); 799 #else /* !HWPMC_HOOKS */ 800 PROC_UNLOCK(p); 801 #endif 802 803 /* Set values passed into the program in registers. */ 804 if (p->p_sysent->sv_setregs) 805 (*p->p_sysent->sv_setregs)(td, imgp, 806 (u_long)(uintptr_t)stack_base); 807 else 808 exec_setregs(td, imgp, (u_long)(uintptr_t)stack_base); 809 810 vfs_mark_atime(imgp->vp, td->td_ucred); 811 812 SDT_PROBE(proc, kernel, , exec_success, args->fname, 0, 0, 0, 0); 813 814 done1: 815 /* 816 * Free any resources malloc'd earlier that we didn't use. 817 */ 818 uifree(euip); 819 if (newcred == NULL) 820 crfree(oldcred); 821 else 822 crfree(newcred); 823 VOP_UNLOCK(imgp->vp, 0); 824 825 /* 826 * Handle deferred decrement of ref counts. 827 */ 828 if (textvp != NULL) { 829 int tvfslocked; 830 831 tvfslocked = VFS_LOCK_GIANT(textvp->v_mount); 832 vrele(textvp); 833 VFS_UNLOCK_GIANT(tvfslocked); 834 } 835 if (binvp && error != 0) 836 vrele(binvp); 837 #ifdef KTRACE 838 if (tracevp != NULL) { 839 int tvfslocked; 840 841 tvfslocked = VFS_LOCK_GIANT(tracevp->v_mount); 842 vrele(tracevp); 843 VFS_UNLOCK_GIANT(tvfslocked); 844 } 845 if (tracecred != NULL) 846 crfree(tracecred); 847 #endif 848 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY); 849 pargs_drop(oldargs); 850 pargs_drop(newargs); 851 if (oldsigacts != NULL) 852 sigacts_free(oldsigacts); 853 854 exec_fail_dealloc: 855 856 /* 857 * free various allocated resources 858 */ 859 if (imgp->firstpage != NULL) 860 exec_unmap_first_page(imgp); 861 862 if (imgp->vp != NULL) { 863 if (args->fname) 864 NDFREE(&nd, NDF_ONLY_PNBUF); 865 if (imgp->opened) 866 VOP_CLOSE(imgp->vp, FREAD, td->td_ucred, td); 867 vput(imgp->vp); 868 } 869 870 if (imgp->object != NULL) 871 vm_object_deallocate(imgp->object); 872 873 free(imgp->freepath, M_TEMP); 874 875 if (error == 0) { 876 PROC_LOCK(p); 877 td->td_dbgflags |= TDB_EXEC; 878 PROC_UNLOCK(p); 879 880 /* 881 * Stop the process here if its stop event mask has 882 * the S_EXEC bit set. 883 */ 884 STOPEVENT(p, S_EXEC, 0); 885 goto done2; 886 } 887 888 exec_fail: 889 /* we're done here, clear P_INEXEC */ 890 PROC_LOCK(p); 891 p->p_flag &= ~P_INEXEC; 892 PROC_UNLOCK(p); 893 894 SDT_PROBE(proc, kernel, , exec_failure, error, 0, 0, 0, 0); 895 896 done2: 897 #ifdef MAC 898 mac_execve_exit(imgp); 899 mac_execve_interpreter_exit(interpvplabel); 900 #endif 901 VFS_UNLOCK_GIANT(vfslocked); 902 exec_free_args(args); 903 904 if (error && imgp->vmspace_destroyed) { 905 /* sorry, no more process anymore. exit gracefully */ 906 exit1(td, W_EXITCODE(0, SIGABRT)); 907 /* NOT REACHED */ 908 } 909 return (error); 910 } 911 912 int 913 exec_map_first_page(imgp) 914 struct image_params *imgp; 915 { 916 int rv, i; 917 int initial_pagein; 918 vm_page_t ma[VM_INITIAL_PAGEIN]; 919 vm_object_t object; 920 921 if (imgp->firstpage != NULL) 922 exec_unmap_first_page(imgp); 923 924 object = imgp->vp->v_object; 925 if (object == NULL) 926 return (EACCES); 927 VM_OBJECT_LOCK(object); 928 #if VM_NRESERVLEVEL > 0 929 if ((object->flags & OBJ_COLORED) == 0) { 930 object->flags |= OBJ_COLORED; 931 object->pg_color = 0; 932 } 933 #endif 934 ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY); 935 if (ma[0]->valid != VM_PAGE_BITS_ALL) { 936 initial_pagein = VM_INITIAL_PAGEIN; 937 if (initial_pagein > object->size) 938 initial_pagein = object->size; 939 for (i = 1; i < initial_pagein; i++) { 940 if ((ma[i] = vm_page_next(ma[i - 1])) != NULL) { 941 if (ma[i]->valid) 942 break; 943 if ((ma[i]->oflags & VPO_BUSY) || ma[i]->busy) 944 break; 945 vm_page_busy(ma[i]); 946 } else { 947 ma[i] = vm_page_alloc(object, i, 948 VM_ALLOC_NORMAL | VM_ALLOC_IFNOTCACHED); 949 if (ma[i] == NULL) 950 break; 951 } 952 } 953 initial_pagein = i; 954 rv = vm_pager_get_pages(object, ma, initial_pagein, 0); 955 ma[0] = vm_page_lookup(object, 0); 956 if ((rv != VM_PAGER_OK) || (ma[0] == NULL)) { 957 if (ma[0] != NULL) { 958 vm_page_lock(ma[0]); 959 vm_page_free(ma[0]); 960 vm_page_unlock(ma[0]); 961 } 962 VM_OBJECT_UNLOCK(object); 963 return (EIO); 964 } 965 } 966 vm_page_lock(ma[0]); 967 vm_page_hold(ma[0]); 968 vm_page_unlock(ma[0]); 969 vm_page_wakeup(ma[0]); 970 VM_OBJECT_UNLOCK(object); 971 972 imgp->firstpage = sf_buf_alloc(ma[0], 0); 973 imgp->image_header = (char *)sf_buf_kva(imgp->firstpage); 974 975 return (0); 976 } 977 978 void 979 exec_unmap_first_page(imgp) 980 struct image_params *imgp; 981 { 982 vm_page_t m; 983 984 if (imgp->firstpage != NULL) { 985 m = sf_buf_page(imgp->firstpage); 986 sf_buf_free(imgp->firstpage); 987 imgp->firstpage = NULL; 988 vm_page_lock(m); 989 vm_page_unhold(m); 990 vm_page_unlock(m); 991 } 992 } 993 994 /* 995 * Destroy old address space, and allocate a new stack 996 * The new stack is only SGROWSIZ large because it is grown 997 * automatically in trap.c. 998 */ 999 int 1000 exec_new_vmspace(imgp, sv) 1001 struct image_params *imgp; 1002 struct sysentvec *sv; 1003 { 1004 int error; 1005 struct proc *p = imgp->proc; 1006 struct vmspace *vmspace = p->p_vmspace; 1007 vm_offset_t sv_minuser, stack_addr; 1008 vm_map_t map; 1009 u_long ssiz; 1010 1011 imgp->vmspace_destroyed = 1; 1012 imgp->sysent = sv; 1013 1014 /* May be called with Giant held */ 1015 EVENTHANDLER_INVOKE(process_exec, p, imgp); 1016 1017 /* 1018 * Blow away entire process VM, if address space not shared, 1019 * otherwise, create a new VM space so that other threads are 1020 * not disrupted 1021 */ 1022 map = &vmspace->vm_map; 1023 if (map_at_zero) 1024 sv_minuser = sv->sv_minuser; 1025 else 1026 sv_minuser = MAX(sv->sv_minuser, PAGE_SIZE); 1027 if (vmspace->vm_refcnt == 1 && vm_map_min(map) == sv_minuser && 1028 vm_map_max(map) == sv->sv_maxuser) { 1029 shmexit(vmspace); 1030 pmap_remove_pages(vmspace_pmap(vmspace)); 1031 vm_map_remove(map, vm_map_min(map), vm_map_max(map)); 1032 } else { 1033 error = vmspace_exec(p, sv_minuser, sv->sv_maxuser); 1034 if (error) 1035 return (error); 1036 vmspace = p->p_vmspace; 1037 map = &vmspace->vm_map; 1038 } 1039 1040 /* Allocate a new stack */ 1041 if (sv->sv_maxssiz != NULL) 1042 ssiz = *sv->sv_maxssiz; 1043 else 1044 ssiz = maxssiz; 1045 stack_addr = sv->sv_usrstack - ssiz; 1046 error = vm_map_stack(map, stack_addr, (vm_size_t)ssiz, 1047 sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_DOWN); 1048 if (error) 1049 return (error); 1050 1051 #ifdef __ia64__ 1052 /* Allocate a new register stack */ 1053 stack_addr = IA64_BACKINGSTORE; 1054 error = vm_map_stack(map, stack_addr, (vm_size_t)ssiz, 1055 sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_UP); 1056 if (error) 1057 return (error); 1058 #endif 1059 1060 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the 1061 * VM_STACK case, but they are still used to monitor the size of the 1062 * process stack so we can check the stack rlimit. 1063 */ 1064 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT; 1065 vmspace->vm_maxsaddr = (char *)sv->sv_usrstack - ssiz; 1066 1067 return (0); 1068 } 1069 1070 /* 1071 * Copy out argument and environment strings from the old process address 1072 * space into the temporary string buffer. 1073 */ 1074 int 1075 exec_copyin_args(struct image_args *args, char *fname, 1076 enum uio_seg segflg, char **argv, char **envv) 1077 { 1078 char *argp, *envp; 1079 int error; 1080 size_t length; 1081 1082 bzero(args, sizeof(*args)); 1083 if (argv == NULL) 1084 return (EFAULT); 1085 1086 /* 1087 * Allocate demand-paged memory for the file name, argument, and 1088 * environment strings. 1089 */ 1090 error = exec_alloc_args(args); 1091 if (error != 0) 1092 return (error); 1093 1094 /* 1095 * Copy the file name. 1096 */ 1097 if (fname != NULL) { 1098 args->fname = args->buf; 1099 error = (segflg == UIO_SYSSPACE) ? 1100 copystr(fname, args->fname, PATH_MAX, &length) : 1101 copyinstr(fname, args->fname, PATH_MAX, &length); 1102 if (error != 0) 1103 goto err_exit; 1104 } else 1105 length = 0; 1106 1107 args->begin_argv = args->buf + length; 1108 args->endp = args->begin_argv; 1109 args->stringspace = ARG_MAX; 1110 1111 /* 1112 * extract arguments first 1113 */ 1114 while ((argp = (caddr_t) (intptr_t) fuword(argv++))) { 1115 if (argp == (caddr_t) -1) { 1116 error = EFAULT; 1117 goto err_exit; 1118 } 1119 if ((error = copyinstr(argp, args->endp, 1120 args->stringspace, &length))) { 1121 if (error == ENAMETOOLONG) 1122 error = E2BIG; 1123 goto err_exit; 1124 } 1125 args->stringspace -= length; 1126 args->endp += length; 1127 args->argc++; 1128 } 1129 1130 args->begin_envv = args->endp; 1131 1132 /* 1133 * extract environment strings 1134 */ 1135 if (envv) { 1136 while ((envp = (caddr_t)(intptr_t)fuword(envv++))) { 1137 if (envp == (caddr_t)-1) { 1138 error = EFAULT; 1139 goto err_exit; 1140 } 1141 if ((error = copyinstr(envp, args->endp, 1142 args->stringspace, &length))) { 1143 if (error == ENAMETOOLONG) 1144 error = E2BIG; 1145 goto err_exit; 1146 } 1147 args->stringspace -= length; 1148 args->endp += length; 1149 args->envc++; 1150 } 1151 } 1152 1153 return (0); 1154 1155 err_exit: 1156 exec_free_args(args); 1157 return (error); 1158 } 1159 1160 /* 1161 * Allocate temporary demand-paged, zero-filled memory for the file name, 1162 * argument, and environment strings. Returns zero if the allocation succeeds 1163 * and ENOMEM otherwise. 1164 */ 1165 int 1166 exec_alloc_args(struct image_args *args) 1167 { 1168 1169 args->buf = (char *)kmem_alloc_wait(exec_map, PATH_MAX + ARG_MAX); 1170 return (args->buf != NULL ? 0 : ENOMEM); 1171 } 1172 1173 void 1174 exec_free_args(struct image_args *args) 1175 { 1176 1177 if (args->buf != NULL) { 1178 kmem_free_wakeup(exec_map, (vm_offset_t)args->buf, 1179 PATH_MAX + ARG_MAX); 1180 args->buf = NULL; 1181 } 1182 if (args->fname_buf != NULL) { 1183 free(args->fname_buf, M_TEMP); 1184 args->fname_buf = NULL; 1185 } 1186 } 1187 1188 /* 1189 * Copy strings out to the new process address space, constructing new arg 1190 * and env vector tables. Return a pointer to the base so that it can be used 1191 * as the initial stack pointer. 1192 */ 1193 register_t * 1194 exec_copyout_strings(imgp) 1195 struct image_params *imgp; 1196 { 1197 int argc, envc; 1198 char **vectp; 1199 char *stringp, *destp; 1200 register_t *stack_base; 1201 struct ps_strings *arginfo; 1202 struct proc *p; 1203 size_t execpath_len; 1204 int szsigcode, szps; 1205 char canary[sizeof(long) * 8]; 1206 1207 szps = sizeof(pagesizes[0]) * MAXPAGESIZES; 1208 /* 1209 * Calculate string base and vector table pointers. 1210 * Also deal with signal trampoline code for this exec type. 1211 */ 1212 if (imgp->execpath != NULL && imgp->auxargs != NULL) 1213 execpath_len = strlen(imgp->execpath) + 1; 1214 else 1215 execpath_len = 0; 1216 p = imgp->proc; 1217 szsigcode = 0; 1218 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; 1219 if (p->p_sysent->sv_szsigcode != NULL) 1220 szsigcode = *(p->p_sysent->sv_szsigcode); 1221 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - 1222 roundup(execpath_len, sizeof(char *)) - 1223 roundup(sizeof(canary), sizeof(char *)) - 1224 roundup(szps, sizeof(char *)) - 1225 roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *)); 1226 1227 /* 1228 * install sigcode 1229 */ 1230 if (szsigcode) 1231 copyout(p->p_sysent->sv_sigcode, ((caddr_t)arginfo - 1232 szsigcode), szsigcode); 1233 1234 /* 1235 * Copy the image path for the rtld. 1236 */ 1237 if (execpath_len != 0) { 1238 imgp->execpathp = (uintptr_t)arginfo - szsigcode - execpath_len; 1239 copyout(imgp->execpath, (void *)imgp->execpathp, 1240 execpath_len); 1241 } 1242 1243 /* 1244 * Prepare the canary for SSP. 1245 */ 1246 arc4rand(canary, sizeof(canary), 0); 1247 imgp->canary = (uintptr_t)arginfo - szsigcode - execpath_len - 1248 sizeof(canary); 1249 copyout(canary, (void *)imgp->canary, sizeof(canary)); 1250 imgp->canarylen = sizeof(canary); 1251 1252 /* 1253 * Prepare the pagesizes array. 1254 */ 1255 imgp->pagesizes = (uintptr_t)arginfo - szsigcode - execpath_len - 1256 roundup(sizeof(canary), sizeof(char *)) - szps; 1257 copyout(pagesizes, (void *)imgp->pagesizes, szps); 1258 imgp->pagesizeslen = szps; 1259 1260 /* 1261 * If we have a valid auxargs ptr, prepare some room 1262 * on the stack. 1263 */ 1264 if (imgp->auxargs) { 1265 /* 1266 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for 1267 * lower compatibility. 1268 */ 1269 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size : 1270 (AT_COUNT * 2); 1271 /* 1272 * The '+ 2' is for the null pointers at the end of each of 1273 * the arg and env vector sets,and imgp->auxarg_size is room 1274 * for argument of Runtime loader. 1275 */ 1276 vectp = (char **)(destp - (imgp->args->argc + 1277 imgp->args->envc + 2 + imgp->auxarg_size) 1278 * sizeof(char *)); 1279 } else { 1280 /* 1281 * The '+ 2' is for the null pointers at the end of each of 1282 * the arg and env vector sets 1283 */ 1284 vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2) * 1285 sizeof(char *)); 1286 } 1287 1288 /* 1289 * vectp also becomes our initial stack base 1290 */ 1291 stack_base = (register_t *)vectp; 1292 1293 stringp = imgp->args->begin_argv; 1294 argc = imgp->args->argc; 1295 envc = imgp->args->envc; 1296 1297 /* 1298 * Copy out strings - arguments and environment. 1299 */ 1300 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace); 1301 1302 /* 1303 * Fill in "ps_strings" struct for ps, w, etc. 1304 */ 1305 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp); 1306 suword32(&arginfo->ps_nargvstr, argc); 1307 1308 /* 1309 * Fill in argument portion of vector table. 1310 */ 1311 for (; argc > 0; --argc) { 1312 suword(vectp++, (long)(intptr_t)destp); 1313 while (*stringp++ != 0) 1314 destp++; 1315 destp++; 1316 } 1317 1318 /* a null vector table pointer separates the argp's from the envp's */ 1319 suword(vectp++, 0); 1320 1321 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp); 1322 suword32(&arginfo->ps_nenvstr, envc); 1323 1324 /* 1325 * Fill in environment portion of vector table. 1326 */ 1327 for (; envc > 0; --envc) { 1328 suword(vectp++, (long)(intptr_t)destp); 1329 while (*stringp++ != 0) 1330 destp++; 1331 destp++; 1332 } 1333 1334 /* end of vector table is a null pointer */ 1335 suword(vectp, 0); 1336 1337 return (stack_base); 1338 } 1339 1340 /* 1341 * Check permissions of file to execute. 1342 * Called with imgp->vp locked. 1343 * Return 0 for success or error code on failure. 1344 */ 1345 int 1346 exec_check_permissions(imgp) 1347 struct image_params *imgp; 1348 { 1349 struct vnode *vp = imgp->vp; 1350 struct vattr *attr = imgp->attr; 1351 struct thread *td; 1352 int error; 1353 1354 td = curthread; 1355 1356 /* Get file attributes */ 1357 error = VOP_GETATTR(vp, attr, td->td_ucred); 1358 if (error) 1359 return (error); 1360 1361 #ifdef MAC 1362 error = mac_vnode_check_exec(td->td_ucred, imgp->vp, imgp); 1363 if (error) 1364 return (error); 1365 #endif 1366 1367 /* 1368 * 1) Check if file execution is disabled for the filesystem that this 1369 * file resides on. 1370 * 2) Insure that at least one execute bit is on - otherwise root 1371 * will always succeed, and we don't want to happen unless the 1372 * file really is executable. 1373 * 3) Insure that the file is a regular file. 1374 */ 1375 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 1376 ((attr->va_mode & 0111) == 0) || 1377 (attr->va_type != VREG)) 1378 return (EACCES); 1379 1380 /* 1381 * Zero length files can't be exec'd 1382 */ 1383 if (attr->va_size == 0) 1384 return (ENOEXEC); 1385 1386 /* 1387 * Check for execute permission to file based on current credentials. 1388 */ 1389 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 1390 if (error) 1391 return (error); 1392 1393 /* 1394 * Check number of open-for-writes on the file and deny execution 1395 * if there are any. 1396 */ 1397 if (vp->v_writecount) 1398 return (ETXTBSY); 1399 1400 /* 1401 * Call filesystem specific open routine (which does nothing in the 1402 * general case). 1403 */ 1404 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 1405 if (error == 0) 1406 imgp->opened = 1; 1407 return (error); 1408 } 1409 1410 /* 1411 * Exec handler registration 1412 */ 1413 int 1414 exec_register(execsw_arg) 1415 const struct execsw *execsw_arg; 1416 { 1417 const struct execsw **es, **xs, **newexecsw; 1418 int count = 2; /* New slot and trailing NULL */ 1419 1420 if (execsw) 1421 for (es = execsw; *es; es++) 1422 count++; 1423 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK); 1424 if (newexecsw == NULL) 1425 return (ENOMEM); 1426 xs = newexecsw; 1427 if (execsw) 1428 for (es = execsw; *es; es++) 1429 *xs++ = *es; 1430 *xs++ = execsw_arg; 1431 *xs = NULL; 1432 if (execsw) 1433 free(execsw, M_TEMP); 1434 execsw = newexecsw; 1435 return (0); 1436 } 1437 1438 int 1439 exec_unregister(execsw_arg) 1440 const struct execsw *execsw_arg; 1441 { 1442 const struct execsw **es, **xs, **newexecsw; 1443 int count = 1; 1444 1445 if (execsw == NULL) 1446 panic("unregister with no handlers left?\n"); 1447 1448 for (es = execsw; *es; es++) { 1449 if (*es == execsw_arg) 1450 break; 1451 } 1452 if (*es == NULL) 1453 return (ENOENT); 1454 for (es = execsw; *es; es++) 1455 if (*es != execsw_arg) 1456 count++; 1457 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK); 1458 if (newexecsw == NULL) 1459 return (ENOMEM); 1460 xs = newexecsw; 1461 for (es = execsw; *es; es++) 1462 if (*es != execsw_arg) 1463 *xs++ = *es; 1464 *xs = NULL; 1465 if (execsw) 1466 free(execsw, M_TEMP); 1467 execsw = newexecsw; 1468 return (0); 1469 } 1470