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 * $Id: kern_exec.c,v 1.85 1998/08/24 08:39:38 dfr Exp $ 27 */ 28 29 #include <sys/param.h> 30 #include <sys/systm.h> 31 #include <sys/sysproto.h> 32 #include <sys/signalvar.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> 41 #include <sys/wait.h> 42 #include <sys/proc.h> 43 #include <sys/pioctl.h> 44 #include <sys/namei.h> 45 #include <sys/sysent.h> 46 #include <sys/shm.h> 47 #include <sys/sysctl.h> 48 #include <sys/vnode.h> 49 #include <sys/buf.h> 50 51 #include <vm/vm.h> 52 #include <vm/vm_param.h> 53 #include <vm/vm_prot.h> 54 #include <sys/lock.h> 55 #include <vm/pmap.h> 56 #include <vm/vm_page.h> 57 #include <vm/vm_map.h> 58 #include <vm/vm_kern.h> 59 #include <vm/vm_extern.h> 60 #include <vm/vm_object.h> 61 #include <vm/vm_zone.h> 62 #include <vm/vm_pager.h> 63 64 #include <machine/reg.h> 65 66 static long *exec_copyout_strings __P((struct image_params *)); 67 68 static struct ps_strings *ps_strings = PS_STRINGS; 69 SYSCTL_INTPTR(_kern, KERN_PS_STRINGS, ps_strings, 0, &ps_strings, 0, ""); 70 71 static caddr_t usrstack = (caddr_t)USRSTACK; 72 SYSCTL_INTPTR(_kern, KERN_USRSTACK, usrstack, 0, &usrstack, 0, ""); 73 /* 74 * execsw_set is constructed for us by the linker. Each of the items 75 * is a pointer to a `const struct execsw', hence the double pointer here. 76 */ 77 static const struct execsw **execsw = 78 (const struct execsw **)&execsw_set.ls_items[0]; 79 80 #ifndef _SYS_SYSPROTO_H_ 81 struct execve_args { 82 char *fname; 83 char **argv; 84 char **envv; 85 }; 86 #endif 87 88 /* 89 * execve() system call. 90 */ 91 int 92 execve(p, uap) 93 struct proc *p; 94 register struct execve_args *uap; 95 { 96 struct nameidata nd, *ndp; 97 long *stack_base; 98 int error, len, i; 99 struct image_params image_params, *imgp; 100 struct vattr attr; 101 102 imgp = &image_params; 103 104 /* 105 * Initialize part of the common data 106 */ 107 imgp->proc = p; 108 imgp->uap = uap; 109 imgp->attr = &attr; 110 imgp->argc = imgp->envc = 0; 111 imgp->argv0 = NULL; 112 imgp->entry_addr = 0; 113 imgp->vmspace_destroyed = 0; 114 imgp->interpreted = 0; 115 imgp->interpreter_name[0] = '\0'; 116 imgp->auxargs = NULL; 117 imgp->vp = NULL; 118 imgp->firstpage = NULL; 119 120 /* 121 * Allocate temporary demand zeroed space for argument and 122 * environment strings 123 */ 124 imgp->stringbase = (char *)kmem_alloc_wait(exec_map, ARG_MAX + PAGE_SIZE); 125 if (imgp->stringbase == NULL) { 126 error = ENOMEM; 127 goto exec_fail; 128 } 129 imgp->stringp = imgp->stringbase; 130 imgp->stringspace = ARG_MAX; 131 imgp->image_header = imgp->stringbase + ARG_MAX; 132 133 /* 134 * Translate the file name. namei() returns a vnode pointer 135 * in ni_vp amoung other things. 136 */ 137 ndp = &nd; 138 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME, 139 UIO_USERSPACE, uap->fname, p); 140 141 interpret: 142 143 error = namei(ndp); 144 if (error) { 145 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase, 146 ARG_MAX + PAGE_SIZE); 147 goto exec_fail; 148 } 149 150 imgp->vp = ndp->ni_vp; 151 152 /* 153 * Check file permissions (also 'opens' file) 154 */ 155 error = exec_check_permissions(imgp); 156 if (error) { 157 VOP_UNLOCK(imgp->vp, 0, p); 158 goto exec_fail_dealloc; 159 } 160 161 error = exec_map_first_page(imgp); 162 VOP_UNLOCK(imgp->vp, 0, p); 163 if (error) 164 goto exec_fail_dealloc; 165 166 /* 167 * Loop through list of image activators, calling each one. 168 * If there is no match, the activator returns -1. If there 169 * is a match, but there was an error during the activation, 170 * the error is returned. Otherwise 0 means success. If the 171 * image is interpreted, loop back up and try activating 172 * the interpreter. 173 */ 174 for (i = 0; execsw[i]; ++i) { 175 if (execsw[i]->ex_imgact) 176 error = (*execsw[i]->ex_imgact)(imgp); 177 else 178 continue; 179 if (error == -1) 180 continue; 181 if (error) 182 goto exec_fail_dealloc; 183 if (imgp->interpreted) { 184 exec_unmap_first_page(imgp); 185 /* free old vnode and name buffer */ 186 vrele(ndp->ni_vp); 187 zfree(namei_zone, ndp->ni_cnd.cn_pnbuf); 188 /* set new name to that of the interpreter */ 189 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME, 190 UIO_SYSSPACE, imgp->interpreter_name, p); 191 goto interpret; 192 } 193 break; 194 } 195 /* If we made it through all the activators and none matched, exit. */ 196 if (error == -1) { 197 error = ENOEXEC; 198 goto exec_fail_dealloc; 199 } 200 201 /* 202 * Copy out strings (args and env) and initialize stack base 203 */ 204 stack_base = exec_copyout_strings(imgp); 205 p->p_vmspace->vm_minsaddr = (char *)stack_base; 206 207 /* 208 * If custom stack fixup routine present for this process 209 * let it do the stack setup. 210 * Else stuff argument count as first item on stack 211 */ 212 if (p->p_sysent->sv_fixup) 213 (*p->p_sysent->sv_fixup)(&stack_base, imgp); 214 else 215 suword(--stack_base, imgp->argc); 216 217 /* 218 * For security and other reasons, the file descriptor table cannot 219 * be shared after an exec. 220 */ 221 if (p->p_fd->fd_refcnt > 1) { 222 struct filedesc *tmp; 223 224 tmp = fdcopy(p); 225 fdfree(p); 226 p->p_fd = tmp; 227 } 228 229 /* close files on exec */ 230 fdcloseexec(p); 231 232 /* reset caught signals */ 233 execsigs(p); 234 235 /* name this process - nameiexec(p, ndp) */ 236 len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN); 237 bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len); 238 p->p_comm[len] = 0; 239 240 /* 241 * mark as execed, wakeup the process that vforked (if any) and tell 242 * it that it now has its own resources back 243 */ 244 p->p_flag |= P_EXEC; 245 if (p->p_pptr && (p->p_flag & P_PPWAIT)) { 246 p->p_flag &= ~P_PPWAIT; 247 wakeup((caddr_t)p->p_pptr); 248 } 249 250 /* 251 * Implement image setuid/setgid. 252 * 253 * Don't honor setuid/setgid if the filesystem prohibits it or if 254 * the process is being traced. 255 */ 256 if ((attr.va_mode & VSUID && p->p_ucred->cr_uid != attr.va_uid || 257 attr.va_mode & VSGID && p->p_ucred->cr_gid != attr.va_gid) && 258 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 && 259 (p->p_flag & P_TRACED) == 0) { 260 /* 261 * Turn off syscall tracing for set-id programs, except for 262 * root. 263 */ 264 if (p->p_tracep && suser(p->p_ucred, &p->p_acflag)) { 265 p->p_traceflag = 0; 266 vrele(p->p_tracep); 267 p->p_tracep = NULL; 268 } 269 /* 270 * Set the new credentials. 271 */ 272 p->p_ucred = crcopy(p->p_ucred); 273 if (attr.va_mode & VSUID) 274 p->p_ucred->cr_uid = attr.va_uid; 275 if (attr.va_mode & VSGID) 276 p->p_ucred->cr_gid = attr.va_gid; 277 setsugid(p); 278 } else { 279 if (p->p_ucred->cr_uid == p->p_cred->p_ruid && 280 p->p_ucred->cr_gid == p->p_cred->p_rgid) 281 p->p_flag &= ~P_SUGID; 282 } 283 284 /* 285 * Implement correct POSIX saved-id behavior. 286 */ 287 p->p_cred->p_svuid = p->p_ucred->cr_uid; 288 p->p_cred->p_svgid = p->p_ucred->cr_gid; 289 290 /* 291 * Store the vp for use in procfs 292 */ 293 if (p->p_textvp) /* release old reference */ 294 vrele(p->p_textvp); 295 VREF(ndp->ni_vp); 296 p->p_textvp = ndp->ni_vp; 297 298 /* 299 * If tracing the process, trap to debugger so breakpoints 300 * can be set before the program executes. 301 */ 302 STOPEVENT(p, S_EXEC, 0); 303 304 if (p->p_flag & P_TRACED) 305 psignal(p, SIGTRAP); 306 307 /* clear "fork but no exec" flag, as we _are_ execing */ 308 p->p_acflag &= ~AFORK; 309 310 /* Set entry address */ 311 setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base); 312 313 exec_fail_dealloc: 314 315 /* 316 * free various allocated resources 317 */ 318 if (imgp->firstpage) 319 exec_unmap_first_page(imgp); 320 321 if (imgp->stringbase != NULL) 322 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase, 323 ARG_MAX + PAGE_SIZE); 324 325 if (ndp->ni_vp) { 326 vrele(ndp->ni_vp); 327 zfree(namei_zone, ndp->ni_cnd.cn_pnbuf); 328 } 329 330 if (error == 0) 331 return (0); 332 333 exec_fail: 334 if (imgp->vmspace_destroyed) { 335 /* sorry, no more process anymore. exit gracefully */ 336 exit1(p, W_EXITCODE(0, SIGABRT)); 337 /* NOT REACHED */ 338 return(0); 339 } else { 340 return(error); 341 } 342 } 343 344 int 345 exec_map_first_page(imgp) 346 struct image_params *imgp; 347 { 348 int s, rv, i; 349 int initial_pagein; 350 vm_page_t ma[VM_INITIAL_PAGEIN]; 351 vm_object_t object; 352 353 354 if (imgp->firstpage) { 355 exec_unmap_first_page(imgp); 356 } 357 358 object = imgp->vp->v_object; 359 s = splvm(); 360 361 ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY); 362 363 if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) { 364 initial_pagein = VM_INITIAL_PAGEIN; 365 if (initial_pagein > object->size) 366 initial_pagein = object->size; 367 for (i = 1; i < initial_pagein; i++) { 368 if (ma[i] = vm_page_lookup(object, i)) { 369 if ((ma[i]->flags & PG_BUSY) || ma[i]->busy) 370 break; 371 if (ma[i]->valid) 372 break; 373 vm_page_busy(ma[i]); 374 } else { 375 ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL); 376 if (ma[i] == NULL) 377 break; 378 } 379 } 380 initial_pagein = i; 381 382 rv = vm_pager_get_pages(object, ma, initial_pagein, 0); 383 ma[0] = vm_page_lookup(object, 0); 384 385 if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) { 386 if (ma[0]) { 387 vm_page_protect(ma[0], VM_PROT_NONE); 388 vm_page_free(ma[0]); 389 } 390 splx(s); 391 return EIO; 392 } 393 } 394 395 vm_page_wire(ma[0]); 396 vm_page_wakeup(ma[0]); 397 splx(s); 398 399 pmap_kenter((vm_offset_t) imgp->image_header, VM_PAGE_TO_PHYS(ma[0])); 400 imgp->firstpage = ma[0]; 401 402 return 0; 403 } 404 405 void 406 exec_unmap_first_page(imgp) 407 struct image_params *imgp; 408 { 409 if (imgp->firstpage) { 410 pmap_kremove((vm_offset_t) imgp->image_header); 411 vm_page_unwire(imgp->firstpage); 412 imgp->firstpage = NULL; 413 } 414 } 415 416 /* 417 * Destroy old address space, and allocate a new stack 418 * The new stack is only SGROWSIZ large because it is grown 419 * automatically in trap.c. 420 */ 421 int 422 exec_new_vmspace(imgp) 423 struct image_params *imgp; 424 { 425 int error; 426 struct vmspace *vmspace = imgp->proc->p_vmspace; 427 caddr_t stack_addr = (caddr_t) (USRSTACK - SGROWSIZ); 428 vm_map_t map = &vmspace->vm_map; 429 430 imgp->vmspace_destroyed = 1; 431 432 /* 433 * Blow away entire process VM, if address space not shared, 434 * otherwise, create a new VM space so that other threads are 435 * not disrupted 436 */ 437 if (vmspace->vm_refcnt == 1) { 438 if (vmspace->vm_shm) 439 shmexit(imgp->proc); 440 pmap_remove_pages(&vmspace->vm_pmap, 0, USRSTACK); 441 vm_map_remove(map, 0, USRSTACK); 442 } else { 443 vmspace_exec(imgp->proc); 444 vmspace = imgp->proc->p_vmspace; 445 map = &vmspace->vm_map; 446 } 447 448 /* Allocate a new stack */ 449 error = vm_map_insert(&vmspace->vm_map, NULL, 0, 450 (vm_offset_t) stack_addr, (vm_offset_t) USRSTACK, 451 VM_PROT_ALL, VM_PROT_ALL, 0); 452 if (error) 453 return (error); 454 455 vmspace->vm_ssize = SGROWSIZ >> PAGE_SHIFT; 456 457 /* Initialize maximum stack address */ 458 vmspace->vm_maxsaddr = (char *)USRSTACK - MAXSSIZ; 459 460 return(0); 461 } 462 463 /* 464 * Copy out argument and environment strings from the old process 465 * address space into the temporary string buffer. 466 */ 467 int 468 exec_extract_strings(imgp) 469 struct image_params *imgp; 470 { 471 char **argv, **envv; 472 char *argp, *envp; 473 int error; 474 size_t length; 475 476 /* 477 * extract arguments first 478 */ 479 480 argv = imgp->uap->argv; 481 482 if (argv) { 483 argp = (caddr_t) (intptr_t) fuword(argv); 484 if (argp == (caddr_t) -1) 485 return (EFAULT); 486 if (argp) 487 argv++; 488 if (imgp->argv0) 489 argp = imgp->argv0; 490 if (argp) { 491 do { 492 if (argp == (caddr_t) -1) 493 return (EFAULT); 494 if ((error = copyinstr(argp, imgp->stringp, 495 imgp->stringspace, &length))) { 496 if (error == ENAMETOOLONG) 497 return(E2BIG); 498 return (error); 499 } 500 imgp->stringspace -= length; 501 imgp->stringp += length; 502 imgp->argc++; 503 } while ((argp = (caddr_t) (intptr_t) fuword(argv++))); 504 } 505 } 506 507 /* 508 * extract environment strings 509 */ 510 511 envv = imgp->uap->envv; 512 513 if (envv) { 514 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) { 515 if (envp == (caddr_t) -1) 516 return (EFAULT); 517 if ((error = copyinstr(envp, imgp->stringp, 518 imgp->stringspace, &length))) { 519 if (error == ENAMETOOLONG) 520 return(E2BIG); 521 return (error); 522 } 523 imgp->stringspace -= length; 524 imgp->stringp += length; 525 imgp->envc++; 526 } 527 } 528 529 return (0); 530 } 531 532 /* 533 * Copy strings out to the new process address space, constructing 534 * new arg and env vector tables. Return a pointer to the base 535 * so that it can be used as the initial stack pointer. 536 */ 537 long * 538 exec_copyout_strings(imgp) 539 struct image_params *imgp; 540 { 541 int argc, envc; 542 char **vectp; 543 char *stringp, *destp; 544 long *stack_base; 545 struct ps_strings *arginfo; 546 int szsigcode; 547 548 /* 549 * Calculate string base and vector table pointers. 550 * Also deal with signal trampoline code for this exec type. 551 */ 552 arginfo = PS_STRINGS; 553 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode); 554 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - 555 roundup((ARG_MAX - imgp->stringspace), sizeof(char *)); 556 557 /* 558 * install sigcode 559 */ 560 if (szsigcode) 561 copyout(imgp->proc->p_sysent->sv_sigcode, 562 ((caddr_t)arginfo - szsigcode), szsigcode); 563 564 /* 565 * If we have a valid auxargs ptr, prepare some room 566 * on the stack. 567 */ 568 if (imgp->auxargs) 569 /* 570 * The '+ 2' is for the null pointers at the end of each of the 571 * arg and env vector sets, and 'AT_COUNT*2' is room for the 572 * ELF Auxargs data. 573 */ 574 vectp = (char **)(destp - (imgp->argc + imgp->envc + 2 + 575 AT_COUNT*2) * sizeof(char*)); 576 else 577 /* 578 * The '+ 2' is for the null pointers at the end of each of the 579 * arg and env vector sets 580 */ 581 vectp = (char **) 582 (destp - (imgp->argc + imgp->envc + 2) * sizeof(char*)); 583 584 /* 585 * vectp also becomes our initial stack base 586 */ 587 stack_base = (long *)vectp; 588 589 stringp = imgp->stringbase; 590 argc = imgp->argc; 591 envc = imgp->envc; 592 593 /* 594 * Copy out strings - arguments and environment. 595 */ 596 copyout(stringp, destp, ARG_MAX - imgp->stringspace); 597 598 /* 599 * Fill in "ps_strings" struct for ps, w, etc. 600 */ 601 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp); 602 suword(&arginfo->ps_nargvstr, argc); 603 604 /* 605 * Fill in argument portion of vector table. 606 */ 607 for (; argc > 0; --argc) { 608 suword(vectp++, (long)(intptr_t)destp); 609 while (*stringp++ != 0) 610 destp++; 611 destp++; 612 } 613 614 /* a null vector table pointer seperates the argp's from the envp's */ 615 suword(vectp++, 0); 616 617 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp); 618 suword(&arginfo->ps_nenvstr, envc); 619 620 /* 621 * Fill in environment portion of vector table. 622 */ 623 for (; envc > 0; --envc) { 624 suword(vectp++, (long)(intptr_t)destp); 625 while (*stringp++ != 0) 626 destp++; 627 destp++; 628 } 629 630 /* end of vector table is a null pointer */ 631 suword(vectp, 0); 632 633 return (stack_base); 634 } 635 636 /* 637 * Check permissions of file to execute. 638 * Return 0 for success or error code on failure. 639 */ 640 int 641 exec_check_permissions(imgp) 642 struct image_params *imgp; 643 { 644 struct proc *p = imgp->proc; 645 struct vnode *vp = imgp->vp; 646 struct vattr *attr = imgp->attr; 647 int error; 648 649 /* Get file attributes */ 650 error = VOP_GETATTR(vp, attr, p->p_ucred, p); 651 if (error) 652 return (error); 653 654 /* 655 * 1) Check if file execution is disabled for the filesystem that this 656 * file resides on. 657 * 2) Insure that at least one execute bit is on - otherwise root 658 * will always succeed, and we don't want to happen unless the 659 * file really is executable. 660 * 3) Insure that the file is a regular file. 661 */ 662 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 663 ((attr->va_mode & 0111) == 0) || 664 (attr->va_type != VREG)) { 665 return (EACCES); 666 } 667 668 /* 669 * Zero length files can't be exec'd 670 */ 671 if (attr->va_size == 0) 672 return (ENOEXEC); 673 674 /* 675 * Check for execute permission to file based on current credentials. 676 */ 677 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 678 if (error) 679 return (error); 680 681 /* 682 * Check number of open-for-writes on the file and deny execution 683 * if there are any. 684 */ 685 if (vp->v_writecount) 686 return (ETXTBSY); 687 688 /* 689 * Call filesystem specific open routine (which does nothing in the 690 * general case). 691 */ 692 error = VOP_OPEN(vp, FREAD, p->p_ucred, p); 693 if (error) 694 return (error); 695 696 return (0); 697 } 698