kern_sharedpage.c revision 79480
1550Srgrimes/* 2550Srgrimes * Copyright (c) 1993, David Greenman 3550Srgrimes * All rights reserved. 4550Srgrimes * 5550Srgrimes * Redistribution and use in source and binary forms, with or without 6550Srgrimes * modification, are permitted provided that the following conditions 7550Srgrimes * are met: 8550Srgrimes * 1. Redistributions of source code must retain the above copyright 9550Srgrimes * notice, this list of conditions and the following disclaimer. 10550Srgrimes * 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: head/sys/kern/kern_exec.c 79480 2001-07-09 19:01:42Z guido $ 27 */ 28 29#include <sys/param.h> 30#include <sys/systm.h> 31#include <sys/lock.h> 32#include <sys/mutex.h> 33#include <sys/sysproto.h> 34#include <sys/signalvar.h> 35#include <sys/kernel.h> 36#include <sys/mount.h> 37#include <sys/filedesc.h> 38#include <sys/fcntl.h> 39#include <sys/acct.h> 40#include <sys/exec.h> 41#include <sys/imgact.h> 42#include <sys/imgact_elf.h> 43#include <sys/wait.h> 44#include <sys/malloc.h> 45#include <sys/proc.h> 46#include <sys/pioctl.h> 47#include <sys/namei.h> 48#include <sys/sysent.h> 49#include <sys/shm.h> 50#include <sys/sysctl.h> 51#include <sys/user.h> 52#include <sys/vnode.h> 53 54#include <vm/vm.h> 55#include <vm/vm_param.h> 56#include <vm/pmap.h> 57#include <vm/vm_page.h> 58#include <vm/vm_map.h> 59#include <vm/vm_kern.h> 60#include <vm/vm_extern.h> 61#include <vm/vm_object.h> 62#include <vm/vm_pager.h> 63 64#include <machine/reg.h> 65 66MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments"); 67 68static register_t *exec_copyout_strings __P((struct image_params *)); 69 70/* XXX This should be vm_size_t. */ 71static u_long ps_strings = PS_STRINGS; 72SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, ""); 73 74/* XXX This should be vm_size_t. */ 75static u_long usrstack = USRSTACK; 76SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, ""); 77 78u_long ps_arg_cache_limit = PAGE_SIZE / 16; 79SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW, 80 &ps_arg_cache_limit, 0, ""); 81 82int ps_argsopen = 1; 83SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, ""); 84 85/* 86 * Each of the items is a pointer to a `const struct execsw', hence the 87 * double pointer here. 88 */ 89static const struct execsw **execsw; 90 91#ifndef _SYS_SYSPROTO_H_ 92struct execve_args { 93 char *fname; 94 char **argv; 95 char **envv; 96}; 97#endif 98 99/* 100 * execve() system call. 101 */ 102int 103execve(p, uap) 104 struct proc *p; 105 register struct execve_args *uap; 106{ 107 struct nameidata nd, *ndp; 108 struct ucred *newcred, *oldcred; 109 register_t *stack_base; 110 int error, len, i; 111 struct image_params image_params, *imgp; 112 struct vattr attr; 113 int (*img_first) __P((struct image_params *)); 114 struct pargs *pa; 115 116 imgp = &image_params; 117 118 /* 119 * Initialize part of the common data 120 */ 121 imgp->proc = p; 122 imgp->uap = uap; 123 imgp->attr = &attr; 124 imgp->argc = imgp->envc = 0; 125 imgp->argv0 = NULL; 126 imgp->entry_addr = 0; 127 imgp->vmspace_destroyed = 0; 128 imgp->interpreted = 0; 129 imgp->interpreter_name[0] = '\0'; 130 imgp->auxargs = NULL; 131 imgp->vp = NULL; 132 imgp->firstpage = NULL; 133 imgp->ps_strings = 0; 134 imgp->auxarg_size = 0; 135 136 /* 137 * Allocate temporary demand zeroed space for argument and 138 * environment strings 139 */ 140 imgp->stringbase = (char *)kmem_alloc_wait(exec_map, ARG_MAX + PAGE_SIZE); 141 if (imgp->stringbase == NULL) { 142 error = ENOMEM; 143 goto exec_fail; 144 } 145 imgp->stringp = imgp->stringbase; 146 imgp->stringspace = ARG_MAX; 147 imgp->image_header = imgp->stringbase + ARG_MAX; 148 149 /* 150 * Translate the file name. namei() returns a vnode pointer 151 * in ni_vp amoung other things. 152 */ 153 ndp = &nd; 154 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME, 155 UIO_USERSPACE, uap->fname, p); 156 157interpret: 158 159 error = namei(ndp); 160 if (error) { 161 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase, 162 ARG_MAX + PAGE_SIZE); 163 goto exec_fail; 164 } 165 166 imgp->vp = ndp->ni_vp; 167 imgp->fname = uap->fname; 168 169 /* 170 * Check file permissions (also 'opens' file) 171 */ 172 error = exec_check_permissions(imgp); 173 if (error) { 174 VOP_UNLOCK(imgp->vp, 0, p); 175 goto exec_fail_dealloc; 176 } 177 178 error = exec_map_first_page(imgp); 179 VOP_UNLOCK(imgp->vp, 0, p); 180 if (error) 181 goto exec_fail_dealloc; 182 183 /* 184 * If the current process has a special image activator it 185 * wants to try first, call it. For example, emulating shell 186 * scripts differently. 187 */ 188 error = -1; 189 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL) 190 error = img_first(imgp); 191 192 /* 193 * Loop through the list of image activators, calling each one. 194 * An activator returns -1 if there is no match, 0 on success, 195 * and an error otherwise. 196 */ 197 for (i = 0; error == -1 && execsw[i]; ++i) { 198 if (execsw[i]->ex_imgact == NULL || 199 execsw[i]->ex_imgact == img_first) { 200 continue; 201 } 202 error = (*execsw[i]->ex_imgact)(imgp); 203 } 204 205 if (error) { 206 if (error == -1) 207 error = ENOEXEC; 208 goto exec_fail_dealloc; 209 } 210 211 /* 212 * Special interpreter operation, cleanup and loop up to try to 213 * activate the interpreter. 214 */ 215 if (imgp->interpreted) { 216 exec_unmap_first_page(imgp); 217 /* free name buffer and old vnode */ 218 NDFREE(ndp, NDF_ONLY_PNBUF); 219 vrele(ndp->ni_vp); 220 /* set new name to that of the interpreter */ 221 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME, 222 UIO_SYSSPACE, imgp->interpreter_name, p); 223 goto interpret; 224 } 225 226 /* 227 * Copy out strings (args and env) and initialize stack base 228 */ 229 stack_base = exec_copyout_strings(imgp); 230 p->p_vmspace->vm_minsaddr = (char *)stack_base; 231 232 /* 233 * If custom stack fixup routine present for this process 234 * let it do the stack setup. 235 * Else stuff argument count as first item on stack 236 */ 237 if (p->p_sysent->sv_fixup) 238 (*p->p_sysent->sv_fixup)(&stack_base, imgp); 239 else 240 suword(--stack_base, imgp->argc); 241 242 /* 243 * For security and other reasons, the file descriptor table cannot 244 * be shared after an exec. 245 */ 246 if (p->p_fd->fd_refcnt > 1) { 247 struct filedesc *tmp; 248 249 tmp = fdcopy(p); 250 fdfree(p); 251 p->p_fd = tmp; 252 } 253 254 /* 255 * For security and other reasons, signal handlers cannot 256 * be shared after an exec. The new proces gets a copy of the old 257 * handlers. In execsigs(), the new process wll have its signals 258 * reset. 259 */ 260 if (p->p_procsig->ps_refcnt > 1) { 261 struct procsig *newprocsig; 262 263 MALLOC(newprocsig, struct procsig *, sizeof(struct procsig), 264 M_SUBPROC, M_WAITOK); 265 bcopy(p->p_procsig, newprocsig, sizeof(*newprocsig)); 266 p->p_procsig->ps_refcnt--; 267 p->p_procsig = newprocsig; 268 p->p_procsig->ps_refcnt = 1; 269 if (p->p_sigacts == &p->p_addr->u_sigacts) 270 panic("shared procsig but private sigacts?\n"); 271 272 p->p_addr->u_sigacts = *p->p_sigacts; 273 p->p_sigacts = &p->p_addr->u_sigacts; 274 } 275 /* Stop profiling */ 276 stopprofclock(p); 277 278 /* close files on exec */ 279 fdcloseexec(p); 280 281 /* reset caught signals */ 282 execsigs(p); 283 284 /* name this process - nameiexec(p, ndp) */ 285 len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN); 286 bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len); 287 p->p_comm[len] = 0; 288 289 /* 290 * mark as execed, wakeup the process that vforked (if any) and tell 291 * it that it now has its own resources back 292 */ 293 PROC_LOCK(p); 294 p->p_flag |= P_EXEC; 295 if (p->p_pptr && (p->p_flag & P_PPWAIT)) { 296 p->p_flag &= ~P_PPWAIT; 297 wakeup((caddr_t)p->p_pptr); 298 } 299 300 /* 301 * XXX: Note, the whole execve() is incredibly racey right now 302 * with regards to debugging and privilege/credential management. 303 * In particular, it's possible to race during exec() to attach 304 * debugging to a process that will gain privilege. 305 * 306 * This MUST be fixed prior to any release. 307 */ 308 309 /* 310 * Implement image setuid/setgid. 311 * 312 * Don't honor setuid/setgid if the filesystem prohibits it or if 313 * the process is being traced. 314 */ 315 oldcred = p->p_ucred; 316 newcred = NULL; 317 if ((((attr.va_mode & VSUID) && oldcred->cr_uid != attr.va_uid) || 318 ((attr.va_mode & VSGID) && oldcred->cr_gid != attr.va_gid)) && 319 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 && 320 (p->p_flag & P_TRACED) == 0) { 321 PROC_UNLOCK(p); 322 /* 323 * Turn off syscall tracing for set-id programs, except for 324 * root. Record any set-id flags first to make sure that 325 * we do not regain any tracing during a possible block. 326 */ 327 setsugid(p); 328 if (p->p_tracep && suser_xxx(oldcred, NULL, PRISON_ROOT)) { 329 p->p_traceflag = 0; 330 vrele(p->p_tracep); 331 p->p_tracep = NULL; 332 } 333 /* 334 * Set the new credentials. 335 */ 336 newcred = crdup(oldcred); 337 if (attr.va_mode & VSUID) 338 change_euid(newcred, attr.va_uid); 339 if (attr.va_mode & VSGID) 340 change_egid(newcred, attr.va_gid); 341 setugidsafety(p); 342 } else { 343 if (oldcred->cr_uid == oldcred->cr_ruid && 344 oldcred->cr_gid == oldcred->cr_rgid) 345 p->p_flag &= ~P_SUGID; 346 PROC_UNLOCK(p); 347 } 348 349 /* 350 * Implement correct POSIX saved-id behavior. 351 * 352 * XXX: It's not clear that the existing behavior is 353 * POSIX-compliant. A number of sourses indicate that the saved 354 * uid/gid should only be updated if the new ruid is not equal to 355 * the old ruid, or the new euid is not equal to the old euid and 356 * the new euid is not equal to the old ruid. The FreeBSD code 357 * always updates the saved uid/gid. Also, this code uses the new 358 * (replaced) euid and egid as the source, which may or may not be 359 * the right ones to use. 360 */ 361 if (newcred == NULL) { 362 if (oldcred->cr_svuid != oldcred->cr_uid || 363 oldcred->cr_svgid != oldcred->cr_gid) { 364 newcred = crdup(oldcred); 365 change_svuid(newcred, newcred->cr_uid); 366 change_svgid(newcred, newcred->cr_gid); 367 } 368 } else { 369 change_svuid(newcred, newcred->cr_uid); 370 change_svgid(newcred, newcred->cr_gid); 371 } 372 373 if (newcred != NULL) { 374 PROC_LOCK(p); 375 p->p_ucred = newcred; 376 PROC_UNLOCK(p); 377 crfree(oldcred); 378 } 379 380 /* 381 * Store the vp for use in procfs 382 */ 383 if (p->p_textvp) /* release old reference */ 384 vrele(p->p_textvp); 385 VREF(ndp->ni_vp); 386 p->p_textvp = ndp->ni_vp; 387 388 /* 389 * notify others that we exec'd 390 */ 391 PROC_LOCK(p); 392 KNOTE(&p->p_klist, NOTE_EXEC); 393 394 /* 395 * If tracing the process, trap to debugger so breakpoints 396 * can be set before the program executes. 397 */ 398 _STOPEVENT(p, S_EXEC, 0); 399 400 if (p->p_flag & P_TRACED) 401 psignal(p, SIGTRAP); 402 403 /* clear "fork but no exec" flag, as we _are_ execing */ 404 p->p_acflag &= ~AFORK; 405 406 /* Set values passed into the program in registers. */ 407 setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base, 408 imgp->ps_strings); 409 410 /* Free any previous argument cache */ 411 pa = p->p_args; 412 p->p_args = NULL; 413 PROC_UNLOCK(p); 414 if (pa != NULL && --pa->ar_ref == 0) 415 FREE(pa, M_PARGS); 416 417 /* Cache arguments if they fit inside our allowance */ 418 i = imgp->endargs - imgp->stringbase; 419 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) { 420 MALLOC(pa, struct pargs *, sizeof(struct pargs) + i, 421 M_PARGS, M_WAITOK); 422 pa->ar_ref = 1; 423 pa->ar_length = i; 424 bcopy(imgp->stringbase, pa->ar_args, i); 425 PROC_LOCK(p); 426 p->p_args = pa; 427 PROC_UNLOCK(p); 428 } 429 430exec_fail_dealloc: 431 432 /* 433 * free various allocated resources 434 */ 435 if (imgp->firstpage) 436 exec_unmap_first_page(imgp); 437 438 if (imgp->stringbase != NULL) 439 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase, 440 ARG_MAX + PAGE_SIZE); 441 442 if (imgp->vp) { 443 NDFREE(ndp, NDF_ONLY_PNBUF); 444 vrele(imgp->vp); 445 } 446 447 if (error == 0) 448 return (0); 449 450exec_fail: 451 if (imgp->vmspace_destroyed) { 452 /* sorry, no more process anymore. exit gracefully */ 453 exit1(p, W_EXITCODE(0, SIGABRT)); 454 /* NOT REACHED */ 455 return(0); 456 } else { 457 return(error); 458 } 459} 460 461int 462exec_map_first_page(imgp) 463 struct image_params *imgp; 464{ 465 int rv, i; 466 int initial_pagein; 467 vm_page_t ma[VM_INITIAL_PAGEIN]; 468 vm_object_t object; 469 470 GIANT_REQUIRED; 471 472 if (imgp->firstpage) { 473 exec_unmap_first_page(imgp); 474 } 475 476 VOP_GETVOBJECT(imgp->vp, &object); 477 478 ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY); 479 480 if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) { 481 initial_pagein = VM_INITIAL_PAGEIN; 482 if (initial_pagein > object->size) 483 initial_pagein = object->size; 484 for (i = 1; i < initial_pagein; i++) { 485 if ((ma[i] = vm_page_lookup(object, i)) != NULL) { 486 if ((ma[i]->flags & PG_BUSY) || ma[i]->busy) 487 break; 488 if (ma[i]->valid) 489 break; 490 vm_page_busy(ma[i]); 491 } else { 492 ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL); 493 if (ma[i] == NULL) 494 break; 495 } 496 } 497 initial_pagein = i; 498 499 rv = vm_pager_get_pages(object, ma, initial_pagein, 0); 500 ma[0] = vm_page_lookup(object, 0); 501 502 if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) { 503 if (ma[0]) { 504 vm_page_protect(ma[0], VM_PROT_NONE); 505 vm_page_free(ma[0]); 506 } 507 return EIO; 508 } 509 } 510 511 vm_page_wire(ma[0]); 512 vm_page_wakeup(ma[0]); 513 514 pmap_kenter((vm_offset_t) imgp->image_header, VM_PAGE_TO_PHYS(ma[0])); 515 imgp->firstpage = ma[0]; 516 517 return 0; 518} 519 520void 521exec_unmap_first_page(imgp) 522 struct image_params *imgp; 523{ 524 GIANT_REQUIRED; 525 526 if (imgp->firstpage) { 527 pmap_kremove((vm_offset_t) imgp->image_header); 528 vm_page_unwire(imgp->firstpage, 1); 529 imgp->firstpage = NULL; 530 } 531} 532 533/* 534 * Destroy old address space, and allocate a new stack 535 * The new stack is only SGROWSIZ large because it is grown 536 * automatically in trap.c. 537 */ 538int 539exec_new_vmspace(imgp) 540 struct image_params *imgp; 541{ 542 int error; 543 struct vmspace *vmspace = imgp->proc->p_vmspace; 544 caddr_t stack_addr = (caddr_t) (USRSTACK - MAXSSIZ); 545 vm_map_t map = &vmspace->vm_map; 546 547 GIANT_REQUIRED; 548 549 imgp->vmspace_destroyed = 1; 550 551 /* 552 * Blow away entire process VM, if address space not shared, 553 * otherwise, create a new VM space so that other threads are 554 * not disrupted 555 */ 556 if (vmspace->vm_refcnt == 1) { 557 if (vmspace->vm_shm) 558 shmexit(imgp->proc); 559 pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS); 560 vm_map_remove(map, 0, VM_MAXUSER_ADDRESS); 561 } else { 562 vmspace_exec(imgp->proc); 563 vmspace = imgp->proc->p_vmspace; 564 map = &vmspace->vm_map; 565 } 566 567 /* Allocate a new stack */ 568 error = vm_map_stack (&vmspace->vm_map, (vm_offset_t)stack_addr, 569 (vm_size_t)MAXSSIZ, VM_PROT_ALL, VM_PROT_ALL, 0); 570 if (error) 571 return (error); 572 573#ifdef __ia64__ 574 { 575 /* 576 * Allocate backing store. We really need something 577 * similar to vm_map_stack which can allow the backing 578 * store to grow upwards. This will do for now. 579 */ 580 vm_offset_t bsaddr; 581 bsaddr = USRSTACK - 2*MAXSSIZ; 582 error = vm_map_find(&vmspace->vm_map, 0, 0, &bsaddr, 583 4*PAGE_SIZE, 0, 584 VM_PROT_ALL, VM_PROT_ALL, 0); 585 imgp->proc->p_md.md_bspstore = bsaddr; 586 } 587#endif 588 589 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the 590 * VM_STACK case, but they are still used to monitor the size of the 591 * process stack so we can check the stack rlimit. 592 */ 593 vmspace->vm_ssize = SGROWSIZ >> PAGE_SHIFT; 594 vmspace->vm_maxsaddr = (char *)USRSTACK - MAXSSIZ; 595 596 return(0); 597} 598 599/* 600 * Copy out argument and environment strings from the old process 601 * address space into the temporary string buffer. 602 */ 603int 604exec_extract_strings(imgp) 605 struct image_params *imgp; 606{ 607 char **argv, **envv; 608 char *argp, *envp; 609 int error; 610 size_t length; 611 612 /* 613 * extract arguments first 614 */ 615 616 argv = imgp->uap->argv; 617 618 if (argv) { 619 argp = (caddr_t) (intptr_t) fuword(argv); 620 if (argp == (caddr_t) -1) 621 return (EFAULT); 622 if (argp) 623 argv++; 624 if (imgp->argv0) 625 argp = imgp->argv0; 626 if (argp) { 627 do { 628 if (argp == (caddr_t) -1) 629 return (EFAULT); 630 if ((error = copyinstr(argp, imgp->stringp, 631 imgp->stringspace, &length))) { 632 if (error == ENAMETOOLONG) 633 return(E2BIG); 634 return (error); 635 } 636 imgp->stringspace -= length; 637 imgp->stringp += length; 638 imgp->argc++; 639 } while ((argp = (caddr_t) (intptr_t) fuword(argv++))); 640 } 641 } 642 643 imgp->endargs = imgp->stringp; 644 645 /* 646 * extract environment strings 647 */ 648 649 envv = imgp->uap->envv; 650 651 if (envv) { 652 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) { 653 if (envp == (caddr_t) -1) 654 return (EFAULT); 655 if ((error = copyinstr(envp, imgp->stringp, 656 imgp->stringspace, &length))) { 657 if (error == ENAMETOOLONG) 658 return(E2BIG); 659 return (error); 660 } 661 imgp->stringspace -= length; 662 imgp->stringp += length; 663 imgp->envc++; 664 } 665 } 666 667 return (0); 668} 669 670/* 671 * Copy strings out to the new process address space, constructing 672 * new arg and env vector tables. Return a pointer to the base 673 * so that it can be used as the initial stack pointer. 674 */ 675register_t * 676exec_copyout_strings(imgp) 677 struct image_params *imgp; 678{ 679 int argc, envc; 680 char **vectp; 681 char *stringp, *destp; 682 register_t *stack_base; 683 struct ps_strings *arginfo; 684 int szsigcode; 685 686 /* 687 * Calculate string base and vector table pointers. 688 * Also deal with signal trampoline code for this exec type. 689 */ 690 arginfo = (struct ps_strings *)PS_STRINGS; 691 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode); 692 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - 693 roundup((ARG_MAX - imgp->stringspace), sizeof(char *)); 694 695 /* 696 * install sigcode 697 */ 698 if (szsigcode) 699 copyout(imgp->proc->p_sysent->sv_sigcode, 700 ((caddr_t)arginfo - szsigcode), szsigcode); 701 702 /* 703 * If we have a valid auxargs ptr, prepare some room 704 * on the stack. 705 */ 706 if (imgp->auxargs) { 707 /* 708 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for 709 * lower compatibility. 710 */ 711 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size 712 : (AT_COUNT * 2); 713 /* 714 * The '+ 2' is for the null pointers at the end of each of 715 * the arg and env vector sets,and imgp->auxarg_size is room 716 * for argument of Runtime loader. 717 */ 718 vectp = (char **) (destp - (imgp->argc + imgp->envc + 2 + 719 imgp->auxarg_size) * sizeof(char *)); 720 721 } else 722 /* 723 * The '+ 2' is for the null pointers at the end of each of 724 * the arg and env vector sets 725 */ 726 vectp = (char **) 727 (destp - (imgp->argc + imgp->envc + 2) * sizeof(char *)); 728 729 /* 730 * vectp also becomes our initial stack base 731 */ 732 stack_base = (register_t *)vectp; 733 734 stringp = imgp->stringbase; 735 argc = imgp->argc; 736 envc = imgp->envc; 737 738 /* 739 * Copy out strings - arguments and environment. 740 */ 741 copyout(stringp, destp, ARG_MAX - imgp->stringspace); 742 743 /* 744 * Fill in "ps_strings" struct for ps, w, etc. 745 */ 746 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp); 747 suword(&arginfo->ps_nargvstr, argc); 748 749 /* 750 * Fill in argument portion of vector table. 751 */ 752 for (; argc > 0; --argc) { 753 suword(vectp++, (long)(intptr_t)destp); 754 while (*stringp++ != 0) 755 destp++; 756 destp++; 757 } 758 759 /* a null vector table pointer separates the argp's from the envp's */ 760 suword(vectp++, 0); 761 762 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp); 763 suword(&arginfo->ps_nenvstr, envc); 764 765 /* 766 * Fill in environment portion of vector table. 767 */ 768 for (; envc > 0; --envc) { 769 suword(vectp++, (long)(intptr_t)destp); 770 while (*stringp++ != 0) 771 destp++; 772 destp++; 773 } 774 775 /* end of vector table is a null pointer */ 776 suword(vectp, 0); 777 778 return (stack_base); 779} 780 781/* 782 * Check permissions of file to execute. 783 * Called with imgp->vp locked. 784 * Return 0 for success or error code on failure. 785 */ 786int 787exec_check_permissions(imgp) 788 struct image_params *imgp; 789{ 790 struct proc *p = imgp->proc; 791 struct vnode *vp = imgp->vp; 792 struct vattr *attr = imgp->attr; 793 int error; 794 795 /* Get file attributes */ 796 error = VOP_GETATTR(vp, attr, p->p_ucred, p); 797 if (error) 798 return (error); 799 800 /* 801 * 1) Check if file execution is disabled for the filesystem that this 802 * file resides on. 803 * 2) Insure that at least one execute bit is on - otherwise root 804 * will always succeed, and we don't want to happen unless the 805 * file really is executable. 806 * 3) Insure that the file is a regular file. 807 */ 808 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 809 ((attr->va_mode & 0111) == 0) || 810 (attr->va_type != VREG)) { 811 return (EACCES); 812 } 813 814 /* 815 * Zero length files can't be exec'd 816 */ 817 if (attr->va_size == 0) 818 return (ENOEXEC); 819 820 /* 821 * Check for execute permission to file based on current credentials. 822 */ 823 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 824 if (error) 825 return (error); 826 827 /* 828 * Check number of open-for-writes on the file and deny execution 829 * if there are any. 830 */ 831 if (vp->v_writecount) 832 return (ETXTBSY); 833 834 /* 835 * Call filesystem specific open routine (which does nothing in the 836 * general case). 837 */ 838 error = VOP_OPEN(vp, FREAD, p->p_ucred, p); 839 if (error) 840 return (error); 841 842 return (0); 843} 844 845/* 846 * Exec handler registration 847 */ 848int 849exec_register(execsw_arg) 850 const struct execsw *execsw_arg; 851{ 852 const struct execsw **es, **xs, **newexecsw; 853 int count = 2; /* New slot and trailing NULL */ 854 855 if (execsw) 856 for (es = execsw; *es; es++) 857 count++; 858 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK); 859 if (newexecsw == NULL) 860 return ENOMEM; 861 xs = newexecsw; 862 if (execsw) 863 for (es = execsw; *es; es++) 864 *xs++ = *es; 865 *xs++ = execsw_arg; 866 *xs = NULL; 867 if (execsw) 868 free(execsw, M_TEMP); 869 execsw = newexecsw; 870 return 0; 871} 872 873int 874exec_unregister(execsw_arg) 875 const struct execsw *execsw_arg; 876{ 877 const struct execsw **es, **xs, **newexecsw; 878 int count = 1; 879 880 if (execsw == NULL) 881 panic("unregister with no handlers left?\n"); 882 883 for (es = execsw; *es; es++) { 884 if (*es == execsw_arg) 885 break; 886 } 887 if (*es == NULL) 888 return ENOENT; 889 for (es = execsw; *es; es++) 890 if (*es != execsw_arg) 891 count++; 892 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK); 893 if (newexecsw == NULL) 894 return ENOMEM; 895 xs = newexecsw; 896 for (es = execsw; *es; es++) 897 if (*es != execsw_arg) 898 *xs++ = *es; 899 *xs = NULL; 900 if (execsw) 901 free(execsw, M_TEMP); 902 execsw = newexecsw; 903 return 0; 904} 905