1 /* $NetBSD: kern_exec.c,v 1.465 2019/05/09 20:50:14 kamil Exp $ */ 2 3 /*- 4 * Copyright (c) 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 17 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 18 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 19 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 20 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 21 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 22 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 23 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 24 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 25 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 26 * POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 /*- 30 * Copyright (C) 1993, 1994, 1996 Christopher G. Demetriou 31 * Copyright (C) 1992 Wolfgang Solfrank. 32 * Copyright (C) 1992 TooLs GmbH. 33 * All rights reserved. 34 * 35 * Redistribution and use in source and binary forms, with or without 36 * modification, are permitted provided that the following conditions 37 * are met: 38 * 1. Redistributions of source code must retain the above copyright 39 * notice, this list of conditions and the following disclaimer. 40 * 2. Redistributions in binary form must reproduce the above copyright 41 * notice, this list of conditions and the following disclaimer in the 42 * documentation and/or other materials provided with the distribution. 43 * 3. All advertising materials mentioning features or use of this software 44 * must display the following acknowledgement: 45 * This product includes software developed by TooLs GmbH. 46 * 4. The name of TooLs GmbH may not be used to endorse or promote products 47 * derived from this software without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR 50 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 51 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 52 * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 53 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 54 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 55 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 56 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR 57 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 58 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 59 */ 60 61 #include <sys/cdefs.h> 62 __KERNEL_RCSID(0, "$NetBSD: kern_exec.c,v 1.465 2019/05/09 20:50:14 kamil Exp $"); 63 64 #include "opt_exec.h" 65 #include "opt_execfmt.h" 66 #include "opt_ktrace.h" 67 #include "opt_modular.h" 68 #include "opt_syscall_debug.h" 69 #include "veriexec.h" 70 #include "opt_pax.h" 71 72 #include <sys/param.h> 73 #include <sys/systm.h> 74 #include <sys/filedesc.h> 75 #include <sys/kernel.h> 76 #include <sys/proc.h> 77 #include <sys/mount.h> 78 #include <sys/kmem.h> 79 #include <sys/namei.h> 80 #include <sys/vnode.h> 81 #include <sys/file.h> 82 #include <sys/filedesc.h> 83 #include <sys/acct.h> 84 #include <sys/atomic.h> 85 #include <sys/exec.h> 86 #include <sys/ktrace.h> 87 #include <sys/uidinfo.h> 88 #include <sys/wait.h> 89 #include <sys/mman.h> 90 #include <sys/ras.h> 91 #include <sys/signalvar.h> 92 #include <sys/stat.h> 93 #include <sys/syscall.h> 94 #include <sys/kauth.h> 95 #include <sys/lwpctl.h> 96 #include <sys/pax.h> 97 #include <sys/cpu.h> 98 #include <sys/module.h> 99 #include <sys/syscallvar.h> 100 #include <sys/syscallargs.h> 101 #if NVERIEXEC > 0 102 #include <sys/verified_exec.h> 103 #endif /* NVERIEXEC > 0 */ 104 #include <sys/sdt.h> 105 #include <sys/spawn.h> 106 #include <sys/prot.h> 107 #include <sys/cprng.h> 108 109 #include <uvm/uvm_extern.h> 110 111 #include <machine/reg.h> 112 113 #include <compat/common/compat_util.h> 114 115 #ifndef MD_TOPDOWN_INIT 116 #ifdef __USE_TOPDOWN_VM 117 #define MD_TOPDOWN_INIT(epp) (epp)->ep_flags |= EXEC_TOPDOWN_VM 118 #else 119 #define MD_TOPDOWN_INIT(epp) 120 #endif 121 #endif 122 123 struct execve_data; 124 125 extern int user_va0_disable; 126 127 static size_t calcargs(struct execve_data * restrict, const size_t); 128 static size_t calcstack(struct execve_data * restrict, const size_t); 129 static int copyoutargs(struct execve_data * restrict, struct lwp *, 130 char * const); 131 static int copyoutpsstrs(struct execve_data * restrict, struct proc *); 132 static int copyinargs(struct execve_data * restrict, char * const *, 133 char * const *, execve_fetch_element_t, char **); 134 static int copyinargstrs(struct execve_data * restrict, char * const *, 135 execve_fetch_element_t, char **, size_t *, void (*)(const void *, size_t)); 136 static int exec_sigcode_map(struct proc *, const struct emul *); 137 138 #if defined(DEBUG) && !defined(DEBUG_EXEC) 139 #define DEBUG_EXEC 140 #endif 141 #ifdef DEBUG_EXEC 142 #define DPRINTF(a) printf a 143 #define COPYPRINTF(s, a, b) printf("%s, %d: copyout%s @%p %zu\n", __func__, \ 144 __LINE__, (s), (a), (b)) 145 static void dump_vmcmds(const struct exec_package * const, size_t, int); 146 #define DUMPVMCMDS(p, x, e) do { dump_vmcmds((p), (x), (e)); } while (0) 147 #else 148 #define DPRINTF(a) 149 #define COPYPRINTF(s, a, b) 150 #define DUMPVMCMDS(p, x, e) do {} while (0) 151 #endif /* DEBUG_EXEC */ 152 153 /* 154 * DTrace SDT provider definitions 155 */ 156 SDT_PROVIDER_DECLARE(proc); 157 SDT_PROBE_DEFINE1(proc, kernel, , exec, "char *"); 158 SDT_PROBE_DEFINE1(proc, kernel, , exec__success, "char *"); 159 SDT_PROBE_DEFINE1(proc, kernel, , exec__failure, "int"); 160 161 /* 162 * Exec function switch: 163 * 164 * Note that each makecmds function is responsible for loading the 165 * exec package with the necessary functions for any exec-type-specific 166 * handling. 167 * 168 * Functions for specific exec types should be defined in their own 169 * header file. 170 */ 171 static const struct execsw **execsw = NULL; 172 static int nexecs; 173 174 u_int exec_maxhdrsz; /* must not be static - used by netbsd32 */ 175 176 /* list of dynamically loaded execsw entries */ 177 static LIST_HEAD(execlist_head, exec_entry) ex_head = 178 LIST_HEAD_INITIALIZER(ex_head); 179 struct exec_entry { 180 LIST_ENTRY(exec_entry) ex_list; 181 SLIST_ENTRY(exec_entry) ex_slist; 182 const struct execsw *ex_sw; 183 }; 184 185 #ifndef __HAVE_SYSCALL_INTERN 186 void syscall(void); 187 #endif 188 189 /* NetBSD autoloadable syscalls */ 190 #ifdef MODULAR 191 #include <kern/syscalls_autoload.c> 192 #endif 193 194 /* NetBSD emul struct */ 195 struct emul emul_netbsd = { 196 .e_name = "netbsd", 197 #ifdef EMUL_NATIVEROOT 198 .e_path = EMUL_NATIVEROOT, 199 #else 200 .e_path = NULL, 201 #endif 202 #ifndef __HAVE_MINIMAL_EMUL 203 .e_flags = EMUL_HAS_SYS___syscall, 204 .e_errno = NULL, 205 .e_nosys = SYS_syscall, 206 .e_nsysent = SYS_NSYSENT, 207 #endif 208 #ifdef MODULAR 209 .e_sc_autoload = netbsd_syscalls_autoload, 210 #endif 211 .e_sysent = sysent, 212 .e_nomodbits = sysent_nomodbits, 213 #ifdef SYSCALL_DEBUG 214 .e_syscallnames = syscallnames, 215 #else 216 .e_syscallnames = NULL, 217 #endif 218 .e_sendsig = sendsig, 219 .e_trapsignal = trapsignal, 220 .e_sigcode = NULL, 221 .e_esigcode = NULL, 222 .e_sigobject = NULL, 223 .e_setregs = setregs, 224 .e_proc_exec = NULL, 225 .e_proc_fork = NULL, 226 .e_proc_exit = NULL, 227 .e_lwp_fork = NULL, 228 .e_lwp_exit = NULL, 229 #ifdef __HAVE_SYSCALL_INTERN 230 .e_syscall_intern = syscall_intern, 231 #else 232 .e_syscall = syscall, 233 #endif 234 .e_sysctlovly = NULL, 235 .e_vm_default_addr = uvm_default_mapaddr, 236 .e_usertrap = NULL, 237 .e_ucsize = sizeof(ucontext_t), 238 .e_startlwp = startlwp 239 }; 240 241 /* 242 * Exec lock. Used to control access to execsw[] structures. 243 * This must not be static so that netbsd32 can access it, too. 244 */ 245 krwlock_t exec_lock; 246 247 static kmutex_t sigobject_lock; 248 249 /* 250 * Data used between a loadvm and execve part of an "exec" operation 251 */ 252 struct execve_data { 253 struct exec_package ed_pack; 254 struct pathbuf *ed_pathbuf; 255 struct vattr ed_attr; 256 struct ps_strings ed_arginfo; 257 char *ed_argp; 258 const char *ed_pathstring; 259 char *ed_resolvedpathbuf; 260 size_t ed_ps_strings_sz; 261 int ed_szsigcode; 262 size_t ed_argslen; 263 long ed_argc; 264 long ed_envc; 265 }; 266 267 /* 268 * data passed from parent lwp to child during a posix_spawn() 269 */ 270 struct spawn_exec_data { 271 struct execve_data sed_exec; 272 struct posix_spawn_file_actions 273 *sed_actions; 274 struct posix_spawnattr *sed_attrs; 275 struct proc *sed_parent; 276 kcondvar_t sed_cv_child_ready; 277 kmutex_t sed_mtx_child; 278 int sed_error; 279 volatile uint32_t sed_refcnt; 280 }; 281 282 static struct vm_map *exec_map; 283 static struct pool exec_pool; 284 285 static void * 286 exec_pool_alloc(struct pool *pp, int flags) 287 { 288 289 return (void *)uvm_km_alloc(exec_map, NCARGS, 0, 290 UVM_KMF_PAGEABLE | UVM_KMF_WAITVA); 291 } 292 293 static void 294 exec_pool_free(struct pool *pp, void *addr) 295 { 296 297 uvm_km_free(exec_map, (vaddr_t)addr, NCARGS, UVM_KMF_PAGEABLE); 298 } 299 300 static struct pool_allocator exec_palloc = { 301 .pa_alloc = exec_pool_alloc, 302 .pa_free = exec_pool_free, 303 .pa_pagesz = NCARGS 304 }; 305 306 /* 307 * check exec: 308 * given an "executable" described in the exec package's namei info, 309 * see what we can do with it. 310 * 311 * ON ENTRY: 312 * exec package with appropriate namei info 313 * lwp pointer of exec'ing lwp 314 * NO SELF-LOCKED VNODES 315 * 316 * ON EXIT: 317 * error: nothing held, etc. exec header still allocated. 318 * ok: filled exec package, executable's vnode (unlocked). 319 * 320 * EXEC SWITCH ENTRY: 321 * Locked vnode to check, exec package, proc. 322 * 323 * EXEC SWITCH EXIT: 324 * ok: return 0, filled exec package, executable's vnode (unlocked). 325 * error: destructive: 326 * everything deallocated execept exec header. 327 * non-destructive: 328 * error code, executable's vnode (unlocked), 329 * exec header unmodified. 330 */ 331 int 332 /*ARGSUSED*/ 333 check_exec(struct lwp *l, struct exec_package *epp, struct pathbuf *pb) 334 { 335 int error, i; 336 struct vnode *vp; 337 struct nameidata nd; 338 size_t resid; 339 340 #if 1 341 // grab the absolute pathbuf here before namei() trashes it. 342 pathbuf_copystring(pb, epp->ep_resolvedname, PATH_MAX); 343 #endif 344 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 345 346 /* first get the vnode */ 347 if ((error = namei(&nd)) != 0) 348 return error; 349 epp->ep_vp = vp = nd.ni_vp; 350 #if 0 351 /* 352 * XXX: can't use nd.ni_pnbuf, because although pb contains an 353 * absolute path, nd.ni_pnbuf does not if the path contains symlinks. 354 */ 355 /* normally this can't fail */ 356 error = copystr(nd.ni_pnbuf, epp->ep_resolvedname, PATH_MAX, NULL); 357 KASSERT(error == 0); 358 #endif 359 360 #ifdef DIAGNOSTIC 361 /* paranoia (take this out once namei stuff stabilizes) */ 362 memset(nd.ni_pnbuf, '~', PATH_MAX); 363 #endif 364 365 /* check access and type */ 366 if (vp->v_type != VREG) { 367 error = EACCES; 368 goto bad1; 369 } 370 if ((error = VOP_ACCESS(vp, VEXEC, l->l_cred)) != 0) 371 goto bad1; 372 373 /* get attributes */ 374 if ((error = VOP_GETATTR(vp, epp->ep_vap, l->l_cred)) != 0) 375 goto bad1; 376 377 /* Check mount point */ 378 if (vp->v_mount->mnt_flag & MNT_NOEXEC) { 379 error = EACCES; 380 goto bad1; 381 } 382 if (vp->v_mount->mnt_flag & MNT_NOSUID) 383 epp->ep_vap->va_mode &= ~(S_ISUID | S_ISGID); 384 385 /* try to open it */ 386 if ((error = VOP_OPEN(vp, FREAD, l->l_cred)) != 0) 387 goto bad1; 388 389 /* unlock vp, since we need it unlocked from here on out. */ 390 VOP_UNLOCK(vp); 391 392 #if NVERIEXEC > 0 393 error = veriexec_verify(l, vp, epp->ep_resolvedname, 394 epp->ep_flags & EXEC_INDIR ? VERIEXEC_INDIRECT : VERIEXEC_DIRECT, 395 NULL); 396 if (error) 397 goto bad2; 398 #endif /* NVERIEXEC > 0 */ 399 400 #ifdef PAX_SEGVGUARD 401 error = pax_segvguard(l, vp, epp->ep_resolvedname, false); 402 if (error) 403 goto bad2; 404 #endif /* PAX_SEGVGUARD */ 405 406 /* now we have the file, get the exec header */ 407 error = vn_rdwr(UIO_READ, vp, epp->ep_hdr, epp->ep_hdrlen, 0, 408 UIO_SYSSPACE, 0, l->l_cred, &resid, NULL); 409 if (error) 410 goto bad2; 411 epp->ep_hdrvalid = epp->ep_hdrlen - resid; 412 413 /* 414 * Set up default address space limits. Can be overridden 415 * by individual exec packages. 416 */ 417 epp->ep_vm_minaddr = exec_vm_minaddr(VM_MIN_ADDRESS); 418 epp->ep_vm_maxaddr = VM_MAXUSER_ADDRESS; 419 420 /* 421 * set up the vmcmds for creation of the process 422 * address space 423 */ 424 error = ENOEXEC; 425 for (i = 0; i < nexecs; i++) { 426 int newerror; 427 428 epp->ep_esch = execsw[i]; 429 newerror = (*execsw[i]->es_makecmds)(l, epp); 430 431 if (!newerror) { 432 /* Seems ok: check that entry point is not too high */ 433 if (epp->ep_entry >= epp->ep_vm_maxaddr) { 434 #ifdef DIAGNOSTIC 435 printf("%s: rejecting %p due to " 436 "too high entry address (>= %p)\n", 437 __func__, (void *)epp->ep_entry, 438 (void *)epp->ep_vm_maxaddr); 439 #endif 440 error = ENOEXEC; 441 break; 442 } 443 /* Seems ok: check that entry point is not too low */ 444 if (epp->ep_entry < epp->ep_vm_minaddr) { 445 #ifdef DIAGNOSTIC 446 printf("%s: rejecting %p due to " 447 "too low entry address (< %p)\n", 448 __func__, (void *)epp->ep_entry, 449 (void *)epp->ep_vm_minaddr); 450 #endif 451 error = ENOEXEC; 452 break; 453 } 454 455 /* check limits */ 456 if ((epp->ep_tsize > MAXTSIZ) || 457 (epp->ep_dsize > (u_quad_t)l->l_proc->p_rlimit 458 [RLIMIT_DATA].rlim_cur)) { 459 #ifdef DIAGNOSTIC 460 printf("%s: rejecting due to " 461 "limits (t=%llu > %llu || d=%llu > %llu)\n", 462 __func__, 463 (unsigned long long)epp->ep_tsize, 464 (unsigned long long)MAXTSIZ, 465 (unsigned long long)epp->ep_dsize, 466 (unsigned long long) 467 l->l_proc->p_rlimit[RLIMIT_DATA].rlim_cur); 468 #endif 469 error = ENOMEM; 470 break; 471 } 472 return 0; 473 } 474 475 /* 476 * Reset all the fields that may have been modified by the 477 * loader. 478 */ 479 KASSERT(epp->ep_emul_arg == NULL); 480 if (epp->ep_emul_root != NULL) { 481 vrele(epp->ep_emul_root); 482 epp->ep_emul_root = NULL; 483 } 484 if (epp->ep_interp != NULL) { 485 vrele(epp->ep_interp); 486 epp->ep_interp = NULL; 487 } 488 epp->ep_pax_flags = 0; 489 490 /* make sure the first "interesting" error code is saved. */ 491 if (error == ENOEXEC) 492 error = newerror; 493 494 if (epp->ep_flags & EXEC_DESTR) 495 /* Error from "#!" code, tidied up by recursive call */ 496 return error; 497 } 498 499 /* not found, error */ 500 501 /* 502 * free any vmspace-creation commands, 503 * and release their references 504 */ 505 kill_vmcmds(&epp->ep_vmcmds); 506 507 bad2: 508 /* 509 * close and release the vnode, restore the old one, free the 510 * pathname buf, and punt. 511 */ 512 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 513 VOP_CLOSE(vp, FREAD, l->l_cred); 514 vput(vp); 515 return error; 516 517 bad1: 518 /* 519 * free the namei pathname buffer, and put the vnode 520 * (which we don't yet have open). 521 */ 522 vput(vp); /* was still locked */ 523 return error; 524 } 525 526 #ifdef __MACHINE_STACK_GROWS_UP 527 #define STACK_PTHREADSPACE NBPG 528 #else 529 #define STACK_PTHREADSPACE 0 530 #endif 531 532 static int 533 execve_fetch_element(char * const *array, size_t index, char **value) 534 { 535 return copyin(array + index, value, sizeof(*value)); 536 } 537 538 /* 539 * exec system call 540 */ 541 int 542 sys_execve(struct lwp *l, const struct sys_execve_args *uap, register_t *retval) 543 { 544 /* { 545 syscallarg(const char *) path; 546 syscallarg(char * const *) argp; 547 syscallarg(char * const *) envp; 548 } */ 549 550 return execve1(l, SCARG(uap, path), SCARG(uap, argp), 551 SCARG(uap, envp), execve_fetch_element); 552 } 553 554 int 555 sys_fexecve(struct lwp *l, const struct sys_fexecve_args *uap, 556 register_t *retval) 557 { 558 /* { 559 syscallarg(int) fd; 560 syscallarg(char * const *) argp; 561 syscallarg(char * const *) envp; 562 } */ 563 564 return ENOSYS; 565 } 566 567 /* 568 * Load modules to try and execute an image that we do not understand. 569 * If no execsw entries are present, we load those likely to be needed 570 * in order to run native images only. Otherwise, we autoload all 571 * possible modules that could let us run the binary. XXX lame 572 */ 573 static void 574 exec_autoload(void) 575 { 576 #ifdef MODULAR 577 static const char * const native[] = { 578 "exec_elf32", 579 "exec_elf64", 580 "exec_script", 581 NULL 582 }; 583 static const char * const compat[] = { 584 "exec_elf32", 585 "exec_elf64", 586 "exec_script", 587 "exec_aout", 588 "exec_coff", 589 "exec_ecoff", 590 "compat_aoutm68k", 591 "compat_netbsd32", 592 "compat_sunos", 593 "compat_sunos32", 594 "compat_ultrix", 595 NULL 596 }; 597 char const * const *list; 598 int i; 599 600 list = (nexecs == 0 ? native : compat); 601 for (i = 0; list[i] != NULL; i++) { 602 if (module_autoload(list[i], MODULE_CLASS_EXEC) != 0) { 603 continue; 604 } 605 yield(); 606 } 607 #endif 608 } 609 610 int 611 exec_makepathbuf(struct lwp *l, const char *upath, enum uio_seg seg, 612 struct pathbuf **pbp, size_t *offs) 613 { 614 char *path, *bp; 615 size_t len, tlen; 616 int error; 617 struct cwdinfo *cwdi; 618 619 path = PNBUF_GET(); 620 if (seg == UIO_SYSSPACE) { 621 error = copystr(upath, path, MAXPATHLEN, &len); 622 } else { 623 error = copyinstr(upath, path, MAXPATHLEN, &len); 624 } 625 if (error) { 626 PNBUF_PUT(path); 627 DPRINTF(("%s: copyin path @%p %d\n", __func__, upath, error)); 628 return error; 629 } 630 631 if (path[0] == '/') { 632 if (offs) 633 *offs = 0; 634 goto out; 635 } 636 637 len++; 638 if (len + 1 >= MAXPATHLEN) 639 goto out; 640 bp = path + MAXPATHLEN - len; 641 memmove(bp, path, len); 642 *(--bp) = '/'; 643 644 cwdi = l->l_proc->p_cwdi; 645 rw_enter(&cwdi->cwdi_lock, RW_READER); 646 error = getcwd_common(cwdi->cwdi_cdir, NULL, &bp, path, MAXPATHLEN / 2, 647 GETCWD_CHECK_ACCESS, l); 648 rw_exit(&cwdi->cwdi_lock); 649 650 if (error) { 651 DPRINTF(("%s: getcwd_common path %s %d\n", __func__, path, 652 error)); 653 goto out; 654 } 655 tlen = path + MAXPATHLEN - bp; 656 657 memmove(path, bp, tlen); 658 path[tlen] = '\0'; 659 if (offs) 660 *offs = tlen - len; 661 out: 662 *pbp = pathbuf_assimilate(path); 663 return 0; 664 } 665 666 vaddr_t 667 exec_vm_minaddr(vaddr_t va_min) 668 { 669 /* 670 * Increase va_min if we don't want NULL to be mappable by the 671 * process. 672 */ 673 #define VM_MIN_GUARD PAGE_SIZE 674 if (user_va0_disable && (va_min < VM_MIN_GUARD)) 675 return VM_MIN_GUARD; 676 return va_min; 677 } 678 679 static int 680 execve_loadvm(struct lwp *l, const char *path, char * const *args, 681 char * const *envs, execve_fetch_element_t fetch_element, 682 struct execve_data * restrict data) 683 { 684 struct exec_package * const epp = &data->ed_pack; 685 int error; 686 struct proc *p; 687 char *dp; 688 u_int modgen; 689 size_t offs = 0; // XXX: GCC 690 691 KASSERT(data != NULL); 692 693 p = l->l_proc; 694 modgen = 0; 695 696 SDT_PROBE(proc, kernel, , exec, path, 0, 0, 0, 0); 697 698 /* 699 * Check if we have exceeded our number of processes limit. 700 * This is so that we handle the case where a root daemon 701 * forked, ran setuid to become the desired user and is trying 702 * to exec. The obvious place to do the reference counting check 703 * is setuid(), but we don't do the reference counting check there 704 * like other OS's do because then all the programs that use setuid() 705 * must be modified to check the return code of setuid() and exit(). 706 * It is dangerous to make setuid() fail, because it fails open and 707 * the program will continue to run as root. If we make it succeed 708 * and return an error code, again we are not enforcing the limit. 709 * The best place to enforce the limit is here, when the process tries 710 * to execute a new image, because eventually the process will need 711 * to call exec in order to do something useful. 712 */ 713 retry: 714 if (p->p_flag & PK_SUGID) { 715 if (kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT, 716 p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS), 717 &p->p_rlimit[RLIMIT_NPROC], 718 KAUTH_ARG(RLIMIT_NPROC)) != 0 && 719 chgproccnt(kauth_cred_getuid(l->l_cred), 0) > 720 p->p_rlimit[RLIMIT_NPROC].rlim_cur) 721 return EAGAIN; 722 } 723 724 /* 725 * Drain existing references and forbid new ones. The process 726 * should be left alone until we're done here. This is necessary 727 * to avoid race conditions - e.g. in ptrace() - that might allow 728 * a local user to illicitly obtain elevated privileges. 729 */ 730 rw_enter(&p->p_reflock, RW_WRITER); 731 732 /* 733 * Init the namei data to point the file user's program name. 734 * This is done here rather than in check_exec(), so that it's 735 * possible to override this settings if any of makecmd/probe 736 * functions call check_exec() recursively - for example, 737 * see exec_script_makecmds(). 738 */ 739 if ((error = exec_makepathbuf(l, path, UIO_USERSPACE, 740 &data->ed_pathbuf, &offs)) != 0) 741 goto clrflg; 742 data->ed_pathstring = pathbuf_stringcopy_get(data->ed_pathbuf); 743 data->ed_resolvedpathbuf = PNBUF_GET(); 744 745 /* 746 * initialize the fields of the exec package. 747 */ 748 epp->ep_kname = data->ed_pathstring + offs; 749 epp->ep_resolvedname = data->ed_resolvedpathbuf; 750 epp->ep_hdr = kmem_alloc(exec_maxhdrsz, KM_SLEEP); 751 epp->ep_hdrlen = exec_maxhdrsz; 752 epp->ep_hdrvalid = 0; 753 epp->ep_emul_arg = NULL; 754 epp->ep_emul_arg_free = NULL; 755 memset(&epp->ep_vmcmds, 0, sizeof(epp->ep_vmcmds)); 756 epp->ep_vap = &data->ed_attr; 757 epp->ep_flags = (p->p_flag & PK_32) ? EXEC_FROM32 : 0; 758 MD_TOPDOWN_INIT(epp); 759 epp->ep_emul_root = NULL; 760 epp->ep_interp = NULL; 761 epp->ep_esch = NULL; 762 epp->ep_pax_flags = 0; 763 memset(epp->ep_machine_arch, 0, sizeof(epp->ep_machine_arch)); 764 765 rw_enter(&exec_lock, RW_READER); 766 767 /* see if we can run it. */ 768 if ((error = check_exec(l, epp, data->ed_pathbuf)) != 0) { 769 if (error != ENOENT && error != EACCES && error != ENOEXEC) { 770 DPRINTF(("%s: check exec failed for %s, error %d\n", 771 __func__, epp->ep_kname, error)); 772 } 773 goto freehdr; 774 } 775 776 /* allocate an argument buffer */ 777 data->ed_argp = pool_get(&exec_pool, PR_WAITOK); 778 KASSERT(data->ed_argp != NULL); 779 dp = data->ed_argp; 780 781 if ((error = copyinargs(data, args, envs, fetch_element, &dp)) != 0) { 782 goto bad; 783 } 784 785 /* 786 * Calculate the new stack size. 787 */ 788 789 #ifdef __MACHINE_STACK_GROWS_UP 790 /* 791 * copyargs() fills argc/argv/envp from the lower address even on 792 * __MACHINE_STACK_GROWS_UP machines. Reserve a few words just below the SP 793 * so that _rtld() use it. 794 */ 795 #define RTLD_GAP 32 796 #else 797 #define RTLD_GAP 0 798 #endif 799 800 const size_t argenvstrlen = (char *)ALIGN(dp) - data->ed_argp; 801 802 data->ed_argslen = calcargs(data, argenvstrlen); 803 804 const size_t len = calcstack(data, pax_aslr_stack_gap(epp) + RTLD_GAP); 805 806 if (len > epp->ep_ssize) { 807 /* in effect, compare to initial limit */ 808 DPRINTF(("%s: stack limit exceeded %zu\n", __func__, len)); 809 error = ENOMEM; 810 goto bad; 811 } 812 /* adjust "active stack depth" for process VSZ */ 813 epp->ep_ssize = len; 814 815 return 0; 816 817 bad: 818 /* free the vmspace-creation commands, and release their references */ 819 kill_vmcmds(&epp->ep_vmcmds); 820 /* kill any opened file descriptor, if necessary */ 821 if (epp->ep_flags & EXEC_HASFD) { 822 epp->ep_flags &= ~EXEC_HASFD; 823 fd_close(epp->ep_fd); 824 } 825 /* close and put the exec'd file */ 826 vn_lock(epp->ep_vp, LK_EXCLUSIVE | LK_RETRY); 827 VOP_CLOSE(epp->ep_vp, FREAD, l->l_cred); 828 vput(epp->ep_vp); 829 pool_put(&exec_pool, data->ed_argp); 830 831 freehdr: 832 kmem_free(epp->ep_hdr, epp->ep_hdrlen); 833 if (epp->ep_emul_root != NULL) 834 vrele(epp->ep_emul_root); 835 if (epp->ep_interp != NULL) 836 vrele(epp->ep_interp); 837 838 rw_exit(&exec_lock); 839 840 pathbuf_stringcopy_put(data->ed_pathbuf, data->ed_pathstring); 841 pathbuf_destroy(data->ed_pathbuf); 842 PNBUF_PUT(data->ed_resolvedpathbuf); 843 844 clrflg: 845 rw_exit(&p->p_reflock); 846 847 if (modgen != module_gen && error == ENOEXEC) { 848 modgen = module_gen; 849 exec_autoload(); 850 goto retry; 851 } 852 853 SDT_PROBE(proc, kernel, , exec__failure, error, 0, 0, 0, 0); 854 return error; 855 } 856 857 static int 858 execve_dovmcmds(struct lwp *l, struct execve_data * restrict data) 859 { 860 struct exec_package * const epp = &data->ed_pack; 861 struct proc *p = l->l_proc; 862 struct exec_vmcmd *base_vcp; 863 int error = 0; 864 size_t i; 865 866 /* record proc's vnode, for use by procfs and others */ 867 if (p->p_textvp) 868 vrele(p->p_textvp); 869 vref(epp->ep_vp); 870 p->p_textvp = epp->ep_vp; 871 872 /* create the new process's VM space by running the vmcmds */ 873 KASSERTMSG(epp->ep_vmcmds.evs_used != 0, "%s: no vmcmds", __func__); 874 875 #ifdef TRACE_EXEC 876 DUMPVMCMDS(epp, 0, 0); 877 #endif 878 879 base_vcp = NULL; 880 881 for (i = 0; i < epp->ep_vmcmds.evs_used && !error; i++) { 882 struct exec_vmcmd *vcp; 883 884 vcp = &epp->ep_vmcmds.evs_cmds[i]; 885 if (vcp->ev_flags & VMCMD_RELATIVE) { 886 KASSERTMSG(base_vcp != NULL, 887 "%s: relative vmcmd with no base", __func__); 888 KASSERTMSG((vcp->ev_flags & VMCMD_BASE) == 0, 889 "%s: illegal base & relative vmcmd", __func__); 890 vcp->ev_addr += base_vcp->ev_addr; 891 } 892 error = (*vcp->ev_proc)(l, vcp); 893 if (error) 894 DUMPVMCMDS(epp, i, error); 895 if (vcp->ev_flags & VMCMD_BASE) 896 base_vcp = vcp; 897 } 898 899 /* free the vmspace-creation commands, and release their references */ 900 kill_vmcmds(&epp->ep_vmcmds); 901 902 vn_lock(epp->ep_vp, LK_EXCLUSIVE | LK_RETRY); 903 VOP_CLOSE(epp->ep_vp, FREAD, l->l_cred); 904 vput(epp->ep_vp); 905 906 /* if an error happened, deallocate and punt */ 907 if (error != 0) { 908 DPRINTF(("%s: vmcmd %zu failed: %d\n", __func__, i - 1, error)); 909 } 910 return error; 911 } 912 913 static void 914 execve_free_data(struct execve_data *data) 915 { 916 struct exec_package * const epp = &data->ed_pack; 917 918 /* free the vmspace-creation commands, and release their references */ 919 kill_vmcmds(&epp->ep_vmcmds); 920 /* kill any opened file descriptor, if necessary */ 921 if (epp->ep_flags & EXEC_HASFD) { 922 epp->ep_flags &= ~EXEC_HASFD; 923 fd_close(epp->ep_fd); 924 } 925 926 /* close and put the exec'd file */ 927 vn_lock(epp->ep_vp, LK_EXCLUSIVE | LK_RETRY); 928 VOP_CLOSE(epp->ep_vp, FREAD, curlwp->l_cred); 929 vput(epp->ep_vp); 930 pool_put(&exec_pool, data->ed_argp); 931 932 kmem_free(epp->ep_hdr, epp->ep_hdrlen); 933 if (epp->ep_emul_root != NULL) 934 vrele(epp->ep_emul_root); 935 if (epp->ep_interp != NULL) 936 vrele(epp->ep_interp); 937 938 pathbuf_stringcopy_put(data->ed_pathbuf, data->ed_pathstring); 939 pathbuf_destroy(data->ed_pathbuf); 940 PNBUF_PUT(data->ed_resolvedpathbuf); 941 } 942 943 static void 944 pathexec(struct proc *p, const char *resolvedname) 945 { 946 KASSERT(resolvedname[0] == '/'); 947 948 /* set command name & other accounting info */ 949 strlcpy(p->p_comm, strrchr(resolvedname, '/') + 1, sizeof(p->p_comm)); 950 951 kmem_strfree(p->p_path); 952 p->p_path = kmem_strdupsize(resolvedname, NULL, KM_SLEEP); 953 } 954 955 /* XXX elsewhere */ 956 static int 957 credexec(struct lwp *l, struct vattr *attr) 958 { 959 struct proc *p = l->l_proc; 960 int error; 961 962 /* 963 * Deal with set[ug]id. MNT_NOSUID has already been used to disable 964 * s[ug]id. It's OK to check for PSL_TRACED here as we have blocked 965 * out additional references on the process for the moment. 966 */ 967 if ((p->p_slflag & PSL_TRACED) == 0 && 968 969 (((attr->va_mode & S_ISUID) != 0 && 970 kauth_cred_geteuid(l->l_cred) != attr->va_uid) || 971 972 ((attr->va_mode & S_ISGID) != 0 && 973 kauth_cred_getegid(l->l_cred) != attr->va_gid))) { 974 /* 975 * Mark the process as SUGID before we do 976 * anything that might block. 977 */ 978 proc_crmod_enter(); 979 proc_crmod_leave(NULL, NULL, true); 980 981 /* Make sure file descriptors 0..2 are in use. */ 982 if ((error = fd_checkstd()) != 0) { 983 DPRINTF(("%s: fdcheckstd failed %d\n", 984 __func__, error)); 985 return error; 986 } 987 988 /* 989 * Copy the credential so other references don't see our 990 * changes. 991 */ 992 l->l_cred = kauth_cred_copy(l->l_cred); 993 #ifdef KTRACE 994 /* 995 * If the persistent trace flag isn't set, turn off. 996 */ 997 if (p->p_tracep) { 998 mutex_enter(&ktrace_lock); 999 if (!(p->p_traceflag & KTRFAC_PERSISTENT)) 1000 ktrderef(p); 1001 mutex_exit(&ktrace_lock); 1002 } 1003 #endif 1004 if (attr->va_mode & S_ISUID) 1005 kauth_cred_seteuid(l->l_cred, attr->va_uid); 1006 if (attr->va_mode & S_ISGID) 1007 kauth_cred_setegid(l->l_cred, attr->va_gid); 1008 } else { 1009 if (kauth_cred_geteuid(l->l_cred) == 1010 kauth_cred_getuid(l->l_cred) && 1011 kauth_cred_getegid(l->l_cred) == 1012 kauth_cred_getgid(l->l_cred)) 1013 p->p_flag &= ~PK_SUGID; 1014 } 1015 1016 /* 1017 * Copy the credential so other references don't see our changes. 1018 * Test to see if this is necessary first, since in the common case 1019 * we won't need a private reference. 1020 */ 1021 if (kauth_cred_geteuid(l->l_cred) != kauth_cred_getsvuid(l->l_cred) || 1022 kauth_cred_getegid(l->l_cred) != kauth_cred_getsvgid(l->l_cred)) { 1023 l->l_cred = kauth_cred_copy(l->l_cred); 1024 kauth_cred_setsvuid(l->l_cred, kauth_cred_geteuid(l->l_cred)); 1025 kauth_cred_setsvgid(l->l_cred, kauth_cred_getegid(l->l_cred)); 1026 } 1027 1028 /* Update the master credentials. */ 1029 if (l->l_cred != p->p_cred) { 1030 kauth_cred_t ocred; 1031 1032 kauth_cred_hold(l->l_cred); 1033 mutex_enter(p->p_lock); 1034 ocred = p->p_cred; 1035 p->p_cred = l->l_cred; 1036 mutex_exit(p->p_lock); 1037 kauth_cred_free(ocred); 1038 } 1039 1040 return 0; 1041 } 1042 1043 static void 1044 emulexec(struct lwp *l, struct exec_package *epp) 1045 { 1046 struct proc *p = l->l_proc; 1047 1048 /* The emulation root will usually have been found when we looked 1049 * for the elf interpreter (or similar), if not look now. */ 1050 if (epp->ep_esch->es_emul->e_path != NULL && 1051 epp->ep_emul_root == NULL) 1052 emul_find_root(l, epp); 1053 1054 /* Any old emulation root got removed by fdcloseexec */ 1055 rw_enter(&p->p_cwdi->cwdi_lock, RW_WRITER); 1056 p->p_cwdi->cwdi_edir = epp->ep_emul_root; 1057 rw_exit(&p->p_cwdi->cwdi_lock); 1058 epp->ep_emul_root = NULL; 1059 if (epp->ep_interp != NULL) 1060 vrele(epp->ep_interp); 1061 1062 /* 1063 * Call emulation specific exec hook. This can setup per-process 1064 * p->p_emuldata or do any other per-process stuff an emulation needs. 1065 * 1066 * If we are executing process of different emulation than the 1067 * original forked process, call e_proc_exit() of the old emulation 1068 * first, then e_proc_exec() of new emulation. If the emulation is 1069 * same, the exec hook code should deallocate any old emulation 1070 * resources held previously by this process. 1071 */ 1072 if (p->p_emul && p->p_emul->e_proc_exit 1073 && p->p_emul != epp->ep_esch->es_emul) 1074 (*p->p_emul->e_proc_exit)(p); 1075 1076 /* 1077 * This is now LWP 1. 1078 */ 1079 /* XXX elsewhere */ 1080 mutex_enter(p->p_lock); 1081 p->p_nlwpid = 1; 1082 l->l_lid = 1; 1083 mutex_exit(p->p_lock); 1084 1085 /* 1086 * Call exec hook. Emulation code may NOT store reference to anything 1087 * from &pack. 1088 */ 1089 if (epp->ep_esch->es_emul->e_proc_exec) 1090 (*epp->ep_esch->es_emul->e_proc_exec)(p, epp); 1091 1092 /* update p_emul, the old value is no longer needed */ 1093 p->p_emul = epp->ep_esch->es_emul; 1094 1095 /* ...and the same for p_execsw */ 1096 p->p_execsw = epp->ep_esch; 1097 1098 #ifdef __HAVE_SYSCALL_INTERN 1099 (*p->p_emul->e_syscall_intern)(p); 1100 #endif 1101 ktremul(); 1102 } 1103 1104 static int 1105 execve_runproc(struct lwp *l, struct execve_data * restrict data, 1106 bool no_local_exec_lock, bool is_spawn) 1107 { 1108 struct exec_package * const epp = &data->ed_pack; 1109 int error = 0; 1110 struct proc *p; 1111 1112 /* 1113 * In case of a posix_spawn operation, the child doing the exec 1114 * might not hold the reader lock on exec_lock, but the parent 1115 * will do this instead. 1116 */ 1117 KASSERT(no_local_exec_lock || rw_lock_held(&exec_lock)); 1118 KASSERT(!no_local_exec_lock || is_spawn); 1119 KASSERT(data != NULL); 1120 1121 p = l->l_proc; 1122 1123 /* Get rid of other LWPs. */ 1124 if (p->p_nlwps > 1) { 1125 mutex_enter(p->p_lock); 1126 exit_lwps(l); 1127 mutex_exit(p->p_lock); 1128 } 1129 KDASSERT(p->p_nlwps == 1); 1130 1131 /* Destroy any lwpctl info. */ 1132 if (p->p_lwpctl != NULL) 1133 lwp_ctl_exit(); 1134 1135 /* Remove POSIX timers */ 1136 timers_free(p, TIMERS_POSIX); 1137 1138 /* Set the PaX flags. */ 1139 pax_set_flags(epp, p); 1140 1141 /* 1142 * Do whatever is necessary to prepare the address space 1143 * for remapping. Note that this might replace the current 1144 * vmspace with another! 1145 */ 1146 if (is_spawn) 1147 uvmspace_spawn(l, epp->ep_vm_minaddr, 1148 epp->ep_vm_maxaddr, 1149 epp->ep_flags & EXEC_TOPDOWN_VM); 1150 else 1151 uvmspace_exec(l, epp->ep_vm_minaddr, 1152 epp->ep_vm_maxaddr, 1153 epp->ep_flags & EXEC_TOPDOWN_VM); 1154 1155 struct vmspace *vm; 1156 vm = p->p_vmspace; 1157 vm->vm_taddr = (void *)epp->ep_taddr; 1158 vm->vm_tsize = btoc(epp->ep_tsize); 1159 vm->vm_daddr = (void*)epp->ep_daddr; 1160 vm->vm_dsize = btoc(epp->ep_dsize); 1161 vm->vm_ssize = btoc(epp->ep_ssize); 1162 vm->vm_issize = 0; 1163 vm->vm_maxsaddr = (void *)epp->ep_maxsaddr; 1164 vm->vm_minsaddr = (void *)epp->ep_minsaddr; 1165 1166 pax_aslr_init_vm(l, vm, epp); 1167 1168 /* Now map address space. */ 1169 error = execve_dovmcmds(l, data); 1170 if (error != 0) 1171 goto exec_abort; 1172 1173 pathexec(p, epp->ep_resolvedname); 1174 1175 char * const newstack = STACK_GROW(vm->vm_minsaddr, epp->ep_ssize); 1176 1177 error = copyoutargs(data, l, newstack); 1178 if (error != 0) 1179 goto exec_abort; 1180 1181 cwdexec(p); 1182 fd_closeexec(); /* handle close on exec */ 1183 1184 if (__predict_false(ktrace_on)) 1185 fd_ktrexecfd(); 1186 1187 execsigs(p); /* reset caught signals */ 1188 1189 mutex_enter(p->p_lock); 1190 l->l_ctxlink = NULL; /* reset ucontext link */ 1191 p->p_acflag &= ~AFORK; 1192 p->p_flag |= PK_EXEC; 1193 mutex_exit(p->p_lock); 1194 1195 /* 1196 * Stop profiling. 1197 */ 1198 if ((p->p_stflag & PST_PROFIL) != 0) { 1199 mutex_spin_enter(&p->p_stmutex); 1200 stopprofclock(p); 1201 mutex_spin_exit(&p->p_stmutex); 1202 } 1203 1204 /* 1205 * It's OK to test PL_PPWAIT unlocked here, as other LWPs have 1206 * exited and exec()/exit() are the only places it will be cleared. 1207 */ 1208 if ((p->p_lflag & PL_PPWAIT) != 0) { 1209 mutex_enter(proc_lock); 1210 l->l_lwpctl = NULL; /* was on loan from blocked parent */ 1211 p->p_lflag &= ~PL_PPWAIT; 1212 cv_broadcast(&p->p_pptr->p_waitcv); 1213 mutex_exit(proc_lock); 1214 } 1215 1216 error = credexec(l, &data->ed_attr); 1217 if (error) 1218 goto exec_abort; 1219 1220 #if defined(__HAVE_RAS) 1221 /* 1222 * Remove all RASs from the address space. 1223 */ 1224 ras_purgeall(); 1225 #endif 1226 1227 doexechooks(p); 1228 1229 /* 1230 * Set initial SP at the top of the stack. 1231 * 1232 * Note that on machines where stack grows up (e.g. hppa), SP points to 1233 * the end of arg/env strings. Userland guesses the address of argc 1234 * via ps_strings::ps_argvstr. 1235 */ 1236 1237 /* Setup new registers and do misc. setup. */ 1238 (*epp->ep_esch->es_emul->e_setregs)(l, epp, (vaddr_t)newstack); 1239 if (epp->ep_esch->es_setregs) 1240 (*epp->ep_esch->es_setregs)(l, epp, (vaddr_t)newstack); 1241 1242 /* Provide a consistent LWP private setting */ 1243 (void)lwp_setprivate(l, NULL); 1244 1245 /* Discard all PCU state; need to start fresh */ 1246 pcu_discard_all(l); 1247 1248 /* map the process's signal trampoline code */ 1249 if ((error = exec_sigcode_map(p, epp->ep_esch->es_emul)) != 0) { 1250 DPRINTF(("%s: map sigcode failed %d\n", __func__, error)); 1251 goto exec_abort; 1252 } 1253 1254 pool_put(&exec_pool, data->ed_argp); 1255 1256 /* notify others that we exec'd */ 1257 KNOTE(&p->p_klist, NOTE_EXEC); 1258 1259 kmem_free(epp->ep_hdr, epp->ep_hdrlen); 1260 1261 SDT_PROBE(proc, kernel, , exec__success, epp->ep_kname, 0, 0, 0, 0); 1262 1263 emulexec(l, epp); 1264 1265 /* Allow new references from the debugger/procfs. */ 1266 rw_exit(&p->p_reflock); 1267 if (!no_local_exec_lock) 1268 rw_exit(&exec_lock); 1269 1270 mutex_enter(proc_lock); 1271 1272 if (p->p_slflag & PSL_TRACED) { 1273 mutex_enter(p->p_lock); 1274 eventswitch(TRAP_EXEC); 1275 mutex_enter(proc_lock); 1276 } 1277 1278 if (p->p_sflag & PS_STOPEXEC) { 1279 ksiginfoq_t kq; 1280 1281 KERNEL_UNLOCK_ALL(l, &l->l_biglocks); 1282 p->p_pptr->p_nstopchild++; 1283 p->p_waited = 0; 1284 mutex_enter(p->p_lock); 1285 ksiginfo_queue_init(&kq); 1286 sigclearall(p, &contsigmask, &kq); 1287 lwp_lock(l); 1288 l->l_stat = LSSTOP; 1289 p->p_stat = SSTOP; 1290 p->p_nrlwps--; 1291 lwp_unlock(l); 1292 mutex_exit(p->p_lock); 1293 mutex_exit(proc_lock); 1294 lwp_lock(l); 1295 mi_switch(l); 1296 ksiginfo_queue_drain(&kq); 1297 KERNEL_LOCK(l->l_biglocks, l); 1298 } else { 1299 mutex_exit(proc_lock); 1300 } 1301 1302 pathbuf_stringcopy_put(data->ed_pathbuf, data->ed_pathstring); 1303 pathbuf_destroy(data->ed_pathbuf); 1304 PNBUF_PUT(data->ed_resolvedpathbuf); 1305 #ifdef TRACE_EXEC 1306 DPRINTF(("%s finished\n", __func__)); 1307 #endif 1308 return EJUSTRETURN; 1309 1310 exec_abort: 1311 SDT_PROBE(proc, kernel, , exec__failure, error, 0, 0, 0, 0); 1312 rw_exit(&p->p_reflock); 1313 if (!no_local_exec_lock) 1314 rw_exit(&exec_lock); 1315 1316 pathbuf_stringcopy_put(data->ed_pathbuf, data->ed_pathstring); 1317 pathbuf_destroy(data->ed_pathbuf); 1318 PNBUF_PUT(data->ed_resolvedpathbuf); 1319 1320 /* 1321 * the old process doesn't exist anymore. exit gracefully. 1322 * get rid of the (new) address space we have created, if any, get rid 1323 * of our namei data and vnode, and exit noting failure 1324 */ 1325 uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS, 1326 VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS); 1327 1328 exec_free_emul_arg(epp); 1329 pool_put(&exec_pool, data->ed_argp); 1330 kmem_free(epp->ep_hdr, epp->ep_hdrlen); 1331 if (epp->ep_emul_root != NULL) 1332 vrele(epp->ep_emul_root); 1333 if (epp->ep_interp != NULL) 1334 vrele(epp->ep_interp); 1335 1336 /* Acquire the sched-state mutex (exit1() will release it). */ 1337 if (!is_spawn) { 1338 mutex_enter(p->p_lock); 1339 exit1(l, error, SIGABRT); 1340 } 1341 1342 return error; 1343 } 1344 1345 int 1346 execve1(struct lwp *l, const char *path, char * const *args, 1347 char * const *envs, execve_fetch_element_t fetch_element) 1348 { 1349 struct execve_data data; 1350 int error; 1351 1352 error = execve_loadvm(l, path, args, envs, fetch_element, &data); 1353 if (error) 1354 return error; 1355 error = execve_runproc(l, &data, false, false); 1356 return error; 1357 } 1358 1359 static size_t 1360 fromptrsz(const struct exec_package *epp) 1361 { 1362 return (epp->ep_flags & EXEC_FROM32) ? sizeof(int) : sizeof(char *); 1363 } 1364 1365 static size_t 1366 ptrsz(const struct exec_package *epp) 1367 { 1368 return (epp->ep_flags & EXEC_32) ? sizeof(int) : sizeof(char *); 1369 } 1370 1371 static size_t 1372 calcargs(struct execve_data * restrict data, const size_t argenvstrlen) 1373 { 1374 struct exec_package * const epp = &data->ed_pack; 1375 1376 const size_t nargenvptrs = 1377 1 + /* long argc */ 1378 data->ed_argc + /* char *argv[] */ 1379 1 + /* \0 */ 1380 data->ed_envc + /* char *env[] */ 1381 1; /* \0 */ 1382 1383 return (nargenvptrs * ptrsz(epp)) /* pointers */ 1384 + argenvstrlen /* strings */ 1385 + epp->ep_esch->es_arglen; /* auxinfo */ 1386 } 1387 1388 static size_t 1389 calcstack(struct execve_data * restrict data, const size_t gaplen) 1390 { 1391 struct exec_package * const epp = &data->ed_pack; 1392 1393 data->ed_szsigcode = epp->ep_esch->es_emul->e_esigcode - 1394 epp->ep_esch->es_emul->e_sigcode; 1395 1396 data->ed_ps_strings_sz = (epp->ep_flags & EXEC_32) ? 1397 sizeof(struct ps_strings32) : sizeof(struct ps_strings); 1398 1399 const size_t sigcode_psstr_sz = 1400 data->ed_szsigcode + /* sigcode */ 1401 data->ed_ps_strings_sz + /* ps_strings */ 1402 STACK_PTHREADSPACE; /* pthread space */ 1403 1404 const size_t stacklen = 1405 data->ed_argslen + 1406 gaplen + 1407 sigcode_psstr_sz; 1408 1409 /* make the stack "safely" aligned */ 1410 return STACK_LEN_ALIGN(stacklen, STACK_ALIGNBYTES); 1411 } 1412 1413 static int 1414 copyoutargs(struct execve_data * restrict data, struct lwp *l, 1415 char * const newstack) 1416 { 1417 struct exec_package * const epp = &data->ed_pack; 1418 struct proc *p = l->l_proc; 1419 int error; 1420 1421 memset(&data->ed_arginfo, 0, sizeof(data->ed_arginfo)); 1422 1423 /* remember information about the process */ 1424 data->ed_arginfo.ps_nargvstr = data->ed_argc; 1425 data->ed_arginfo.ps_nenvstr = data->ed_envc; 1426 1427 /* 1428 * Allocate the stack address passed to the newly execve()'ed process. 1429 * 1430 * The new stack address will be set to the SP (stack pointer) register 1431 * in setregs(). 1432 */ 1433 1434 char *newargs = STACK_ALLOC( 1435 STACK_SHRINK(newstack, data->ed_argslen), data->ed_argslen); 1436 1437 error = (*epp->ep_esch->es_copyargs)(l, epp, 1438 &data->ed_arginfo, &newargs, data->ed_argp); 1439 1440 if (error) { 1441 DPRINTF(("%s: copyargs failed %d\n", __func__, error)); 1442 return error; 1443 } 1444 1445 error = copyoutpsstrs(data, p); 1446 if (error != 0) 1447 return error; 1448 1449 return 0; 1450 } 1451 1452 static int 1453 copyoutpsstrs(struct execve_data * restrict data, struct proc *p) 1454 { 1455 struct exec_package * const epp = &data->ed_pack; 1456 struct ps_strings32 arginfo32; 1457 void *aip; 1458 int error; 1459 1460 /* fill process ps_strings info */ 1461 p->p_psstrp = (vaddr_t)STACK_ALLOC(STACK_GROW(epp->ep_minsaddr, 1462 STACK_PTHREADSPACE), data->ed_ps_strings_sz); 1463 1464 if (epp->ep_flags & EXEC_32) { 1465 aip = &arginfo32; 1466 arginfo32.ps_argvstr = (vaddr_t)data->ed_arginfo.ps_argvstr; 1467 arginfo32.ps_nargvstr = data->ed_arginfo.ps_nargvstr; 1468 arginfo32.ps_envstr = (vaddr_t)data->ed_arginfo.ps_envstr; 1469 arginfo32.ps_nenvstr = data->ed_arginfo.ps_nenvstr; 1470 } else 1471 aip = &data->ed_arginfo; 1472 1473 /* copy out the process's ps_strings structure */ 1474 if ((error = copyout(aip, (void *)p->p_psstrp, data->ed_ps_strings_sz)) 1475 != 0) { 1476 DPRINTF(("%s: ps_strings copyout %p->%p size %zu failed\n", 1477 __func__, aip, (void *)p->p_psstrp, data->ed_ps_strings_sz)); 1478 return error; 1479 } 1480 1481 return 0; 1482 } 1483 1484 static int 1485 copyinargs(struct execve_data * restrict data, char * const *args, 1486 char * const *envs, execve_fetch_element_t fetch_element, char **dpp) 1487 { 1488 struct exec_package * const epp = &data->ed_pack; 1489 char *dp; 1490 size_t i; 1491 int error; 1492 1493 dp = *dpp; 1494 1495 data->ed_argc = 0; 1496 1497 /* copy the fake args list, if there's one, freeing it as we go */ 1498 if (epp->ep_flags & EXEC_HASARGL) { 1499 struct exec_fakearg *fa = epp->ep_fa; 1500 1501 while (fa->fa_arg != NULL) { 1502 const size_t maxlen = ARG_MAX - (dp - data->ed_argp); 1503 size_t len; 1504 1505 len = strlcpy(dp, fa->fa_arg, maxlen); 1506 /* Count NUL into len. */ 1507 if (len < maxlen) 1508 len++; 1509 else { 1510 while (fa->fa_arg != NULL) { 1511 kmem_free(fa->fa_arg, fa->fa_len); 1512 fa++; 1513 } 1514 kmem_free(epp->ep_fa, epp->ep_fa_len); 1515 epp->ep_flags &= ~EXEC_HASARGL; 1516 return E2BIG; 1517 } 1518 ktrexecarg(fa->fa_arg, len - 1); 1519 dp += len; 1520 1521 kmem_free(fa->fa_arg, fa->fa_len); 1522 fa++; 1523 data->ed_argc++; 1524 } 1525 kmem_free(epp->ep_fa, epp->ep_fa_len); 1526 epp->ep_flags &= ~EXEC_HASARGL; 1527 } 1528 1529 /* 1530 * Read and count argument strings from user. 1531 */ 1532 1533 if (args == NULL) { 1534 DPRINTF(("%s: null args\n", __func__)); 1535 return EINVAL; 1536 } 1537 if (epp->ep_flags & EXEC_SKIPARG) 1538 args = (const void *)((const char *)args + fromptrsz(epp)); 1539 i = 0; 1540 error = copyinargstrs(data, args, fetch_element, &dp, &i, ktr_execarg); 1541 if (error != 0) { 1542 DPRINTF(("%s: copyin arg %d\n", __func__, error)); 1543 return error; 1544 } 1545 data->ed_argc += i; 1546 1547 /* 1548 * Read and count environment strings from user. 1549 */ 1550 1551 data->ed_envc = 0; 1552 /* environment need not be there */ 1553 if (envs == NULL) 1554 goto done; 1555 i = 0; 1556 error = copyinargstrs(data, envs, fetch_element, &dp, &i, ktr_execenv); 1557 if (error != 0) { 1558 DPRINTF(("%s: copyin env %d\n", __func__, error)); 1559 return error; 1560 } 1561 data->ed_envc += i; 1562 1563 done: 1564 *dpp = dp; 1565 1566 return 0; 1567 } 1568 1569 static int 1570 copyinargstrs(struct execve_data * restrict data, char * const *strs, 1571 execve_fetch_element_t fetch_element, char **dpp, size_t *ip, 1572 void (*ktr)(const void *, size_t)) 1573 { 1574 char *dp, *sp; 1575 size_t i; 1576 int error; 1577 1578 dp = *dpp; 1579 1580 i = 0; 1581 while (1) { 1582 const size_t maxlen = ARG_MAX - (dp - data->ed_argp); 1583 size_t len; 1584 1585 if ((error = (*fetch_element)(strs, i, &sp)) != 0) { 1586 return error; 1587 } 1588 if (!sp) 1589 break; 1590 if ((error = copyinstr(sp, dp, maxlen, &len)) != 0) { 1591 if (error == ENAMETOOLONG) 1592 error = E2BIG; 1593 return error; 1594 } 1595 if (__predict_false(ktrace_on)) 1596 (*ktr)(dp, len - 1); 1597 dp += len; 1598 i++; 1599 } 1600 1601 *dpp = dp; 1602 *ip = i; 1603 1604 return 0; 1605 } 1606 1607 /* 1608 * Copy argv and env strings from kernel buffer (argp) to the new stack. 1609 * Those strings are located just after auxinfo. 1610 */ 1611 int 1612 copyargs(struct lwp *l, struct exec_package *pack, struct ps_strings *arginfo, 1613 char **stackp, void *argp) 1614 { 1615 char **cpp, *dp, *sp; 1616 size_t len; 1617 void *nullp; 1618 long argc, envc; 1619 int error; 1620 1621 cpp = (char **)*stackp; 1622 nullp = NULL; 1623 argc = arginfo->ps_nargvstr; 1624 envc = arginfo->ps_nenvstr; 1625 1626 /* argc on stack is long */ 1627 CTASSERT(sizeof(*cpp) == sizeof(argc)); 1628 1629 dp = (char *)(cpp + 1630 1 + /* long argc */ 1631 argc + /* char *argv[] */ 1632 1 + /* \0 */ 1633 envc + /* char *env[] */ 1634 1) + /* \0 */ 1635 pack->ep_esch->es_arglen; /* auxinfo */ 1636 sp = argp; 1637 1638 if ((error = copyout(&argc, cpp++, sizeof(argc))) != 0) { 1639 COPYPRINTF("", cpp - 1, sizeof(argc)); 1640 return error; 1641 } 1642 1643 /* XXX don't copy them out, remap them! */ 1644 arginfo->ps_argvstr = cpp; /* remember location of argv for later */ 1645 1646 for (; --argc >= 0; sp += len, dp += len) { 1647 if ((error = copyout(&dp, cpp++, sizeof(dp))) != 0) { 1648 COPYPRINTF("", cpp - 1, sizeof(dp)); 1649 return error; 1650 } 1651 if ((error = copyoutstr(sp, dp, ARG_MAX, &len)) != 0) { 1652 COPYPRINTF("str", dp, (size_t)ARG_MAX); 1653 return error; 1654 } 1655 } 1656 1657 if ((error = copyout(&nullp, cpp++, sizeof(nullp))) != 0) { 1658 COPYPRINTF("", cpp - 1, sizeof(nullp)); 1659 return error; 1660 } 1661 1662 arginfo->ps_envstr = cpp; /* remember location of envp for later */ 1663 1664 for (; --envc >= 0; sp += len, dp += len) { 1665 if ((error = copyout(&dp, cpp++, sizeof(dp))) != 0) { 1666 COPYPRINTF("", cpp - 1, sizeof(dp)); 1667 return error; 1668 } 1669 if ((error = copyoutstr(sp, dp, ARG_MAX, &len)) != 0) { 1670 COPYPRINTF("str", dp, (size_t)ARG_MAX); 1671 return error; 1672 } 1673 1674 } 1675 1676 if ((error = copyout(&nullp, cpp++, sizeof(nullp))) != 0) { 1677 COPYPRINTF("", cpp - 1, sizeof(nullp)); 1678 return error; 1679 } 1680 1681 *stackp = (char *)cpp; 1682 return 0; 1683 } 1684 1685 1686 /* 1687 * Add execsw[] entries. 1688 */ 1689 int 1690 exec_add(struct execsw *esp, int count) 1691 { 1692 struct exec_entry *it; 1693 int i; 1694 1695 if (count == 0) { 1696 return 0; 1697 } 1698 1699 /* Check for duplicates. */ 1700 rw_enter(&exec_lock, RW_WRITER); 1701 for (i = 0; i < count; i++) { 1702 LIST_FOREACH(it, &ex_head, ex_list) { 1703 /* assume unique (makecmds, probe_func, emulation) */ 1704 if (it->ex_sw->es_makecmds == esp[i].es_makecmds && 1705 it->ex_sw->u.elf_probe_func == 1706 esp[i].u.elf_probe_func && 1707 it->ex_sw->es_emul == esp[i].es_emul) { 1708 rw_exit(&exec_lock); 1709 return EEXIST; 1710 } 1711 } 1712 } 1713 1714 /* Allocate new entries. */ 1715 for (i = 0; i < count; i++) { 1716 it = kmem_alloc(sizeof(*it), KM_SLEEP); 1717 it->ex_sw = &esp[i]; 1718 LIST_INSERT_HEAD(&ex_head, it, ex_list); 1719 } 1720 1721 /* update execsw[] */ 1722 exec_init(0); 1723 rw_exit(&exec_lock); 1724 return 0; 1725 } 1726 1727 /* 1728 * Remove execsw[] entry. 1729 */ 1730 int 1731 exec_remove(struct execsw *esp, int count) 1732 { 1733 struct exec_entry *it, *next; 1734 int i; 1735 const struct proclist_desc *pd; 1736 proc_t *p; 1737 1738 if (count == 0) { 1739 return 0; 1740 } 1741 1742 /* Abort if any are busy. */ 1743 rw_enter(&exec_lock, RW_WRITER); 1744 for (i = 0; i < count; i++) { 1745 mutex_enter(proc_lock); 1746 for (pd = proclists; pd->pd_list != NULL; pd++) { 1747 PROCLIST_FOREACH(p, pd->pd_list) { 1748 if (p->p_execsw == &esp[i]) { 1749 mutex_exit(proc_lock); 1750 rw_exit(&exec_lock); 1751 return EBUSY; 1752 } 1753 } 1754 } 1755 mutex_exit(proc_lock); 1756 } 1757 1758 /* None are busy, so remove them all. */ 1759 for (i = 0; i < count; i++) { 1760 for (it = LIST_FIRST(&ex_head); it != NULL; it = next) { 1761 next = LIST_NEXT(it, ex_list); 1762 if (it->ex_sw == &esp[i]) { 1763 LIST_REMOVE(it, ex_list); 1764 kmem_free(it, sizeof(*it)); 1765 break; 1766 } 1767 } 1768 } 1769 1770 /* update execsw[] */ 1771 exec_init(0); 1772 rw_exit(&exec_lock); 1773 return 0; 1774 } 1775 1776 /* 1777 * Initialize exec structures. If init_boot is true, also does necessary 1778 * one-time initialization (it's called from main() that way). 1779 * Once system is multiuser, this should be called with exec_lock held, 1780 * i.e. via exec_{add|remove}(). 1781 */ 1782 int 1783 exec_init(int init_boot) 1784 { 1785 const struct execsw **sw; 1786 struct exec_entry *ex; 1787 SLIST_HEAD(,exec_entry) first; 1788 SLIST_HEAD(,exec_entry) any; 1789 SLIST_HEAD(,exec_entry) last; 1790 int i, sz; 1791 1792 if (init_boot) { 1793 /* do one-time initializations */ 1794 vaddr_t vmin = 0, vmax; 1795 1796 rw_init(&exec_lock); 1797 mutex_init(&sigobject_lock, MUTEX_DEFAULT, IPL_NONE); 1798 exec_map = uvm_km_suballoc(kernel_map, &vmin, &vmax, 1799 maxexec*NCARGS, VM_MAP_PAGEABLE, false, NULL); 1800 pool_init(&exec_pool, NCARGS, 0, 0, PR_NOALIGN|PR_NOTOUCH, 1801 "execargs", &exec_palloc, IPL_NONE); 1802 pool_sethardlimit(&exec_pool, maxexec, "should not happen", 0); 1803 } else { 1804 KASSERT(rw_write_held(&exec_lock)); 1805 } 1806 1807 /* Sort each entry onto the appropriate queue. */ 1808 SLIST_INIT(&first); 1809 SLIST_INIT(&any); 1810 SLIST_INIT(&last); 1811 sz = 0; 1812 LIST_FOREACH(ex, &ex_head, ex_list) { 1813 switch(ex->ex_sw->es_prio) { 1814 case EXECSW_PRIO_FIRST: 1815 SLIST_INSERT_HEAD(&first, ex, ex_slist); 1816 break; 1817 case EXECSW_PRIO_ANY: 1818 SLIST_INSERT_HEAD(&any, ex, ex_slist); 1819 break; 1820 case EXECSW_PRIO_LAST: 1821 SLIST_INSERT_HEAD(&last, ex, ex_slist); 1822 break; 1823 default: 1824 panic("%s", __func__); 1825 break; 1826 } 1827 sz++; 1828 } 1829 1830 /* 1831 * Create new execsw[]. Ensure we do not try a zero-sized 1832 * allocation. 1833 */ 1834 sw = kmem_alloc(sz * sizeof(struct execsw *) + 1, KM_SLEEP); 1835 i = 0; 1836 SLIST_FOREACH(ex, &first, ex_slist) { 1837 sw[i++] = ex->ex_sw; 1838 } 1839 SLIST_FOREACH(ex, &any, ex_slist) { 1840 sw[i++] = ex->ex_sw; 1841 } 1842 SLIST_FOREACH(ex, &last, ex_slist) { 1843 sw[i++] = ex->ex_sw; 1844 } 1845 1846 /* Replace old execsw[] and free used memory. */ 1847 if (execsw != NULL) { 1848 kmem_free(__UNCONST(execsw), 1849 nexecs * sizeof(struct execsw *) + 1); 1850 } 1851 execsw = sw; 1852 nexecs = sz; 1853 1854 /* Figure out the maximum size of an exec header. */ 1855 exec_maxhdrsz = sizeof(int); 1856 for (i = 0; i < nexecs; i++) { 1857 if (execsw[i]->es_hdrsz > exec_maxhdrsz) 1858 exec_maxhdrsz = execsw[i]->es_hdrsz; 1859 } 1860 1861 return 0; 1862 } 1863 1864 static int 1865 exec_sigcode_map(struct proc *p, const struct emul *e) 1866 { 1867 vaddr_t va; 1868 vsize_t sz; 1869 int error; 1870 struct uvm_object *uobj; 1871 1872 sz = (vaddr_t)e->e_esigcode - (vaddr_t)e->e_sigcode; 1873 1874 if (e->e_sigobject == NULL || sz == 0) { 1875 return 0; 1876 } 1877 1878 /* 1879 * If we don't have a sigobject for this emulation, create one. 1880 * 1881 * sigobject is an anonymous memory object (just like SYSV shared 1882 * memory) that we keep a permanent reference to and that we map 1883 * in all processes that need this sigcode. The creation is simple, 1884 * we create an object, add a permanent reference to it, map it in 1885 * kernel space, copy out the sigcode to it and unmap it. 1886 * We map it with PROT_READ|PROT_EXEC into the process just 1887 * the way sys_mmap() would map it. 1888 */ 1889 1890 uobj = *e->e_sigobject; 1891 if (uobj == NULL) { 1892 mutex_enter(&sigobject_lock); 1893 if ((uobj = *e->e_sigobject) == NULL) { 1894 uobj = uao_create(sz, 0); 1895 (*uobj->pgops->pgo_reference)(uobj); 1896 va = vm_map_min(kernel_map); 1897 if ((error = uvm_map(kernel_map, &va, round_page(sz), 1898 uobj, 0, 0, 1899 UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, 1900 UVM_INH_SHARE, UVM_ADV_RANDOM, 0)))) { 1901 printf("kernel mapping failed %d\n", error); 1902 (*uobj->pgops->pgo_detach)(uobj); 1903 mutex_exit(&sigobject_lock); 1904 return error; 1905 } 1906 memcpy((void *)va, e->e_sigcode, sz); 1907 #ifdef PMAP_NEED_PROCWR 1908 pmap_procwr(&proc0, va, sz); 1909 #endif 1910 uvm_unmap(kernel_map, va, va + round_page(sz)); 1911 *e->e_sigobject = uobj; 1912 } 1913 mutex_exit(&sigobject_lock); 1914 } 1915 1916 /* Just a hint to uvm_map where to put it. */ 1917 va = e->e_vm_default_addr(p, (vaddr_t)p->p_vmspace->vm_daddr, 1918 round_page(sz), p->p_vmspace->vm_map.flags & VM_MAP_TOPDOWN); 1919 1920 #ifdef __alpha__ 1921 /* 1922 * Tru64 puts /sbin/loader at the end of user virtual memory, 1923 * which causes the above calculation to put the sigcode at 1924 * an invalid address. Put it just below the text instead. 1925 */ 1926 if (va == (vaddr_t)vm_map_max(&p->p_vmspace->vm_map)) { 1927 va = (vaddr_t)p->p_vmspace->vm_taddr - round_page(sz); 1928 } 1929 #endif 1930 1931 (*uobj->pgops->pgo_reference)(uobj); 1932 error = uvm_map(&p->p_vmspace->vm_map, &va, round_page(sz), 1933 uobj, 0, 0, 1934 UVM_MAPFLAG(UVM_PROT_RX, UVM_PROT_RX, UVM_INH_SHARE, 1935 UVM_ADV_RANDOM, 0)); 1936 if (error) { 1937 DPRINTF(("%s, %d: map %p " 1938 "uvm_map %#"PRIxVSIZE"@%#"PRIxVADDR" failed %d\n", 1939 __func__, __LINE__, &p->p_vmspace->vm_map, round_page(sz), 1940 va, error)); 1941 (*uobj->pgops->pgo_detach)(uobj); 1942 return error; 1943 } 1944 p->p_sigctx.ps_sigcode = (void *)va; 1945 return 0; 1946 } 1947 1948 /* 1949 * Release a refcount on spawn_exec_data and destroy memory, if this 1950 * was the last one. 1951 */ 1952 static void 1953 spawn_exec_data_release(struct spawn_exec_data *data) 1954 { 1955 if (atomic_dec_32_nv(&data->sed_refcnt) != 0) 1956 return; 1957 1958 cv_destroy(&data->sed_cv_child_ready); 1959 mutex_destroy(&data->sed_mtx_child); 1960 1961 if (data->sed_actions) 1962 posix_spawn_fa_free(data->sed_actions, 1963 data->sed_actions->len); 1964 if (data->sed_attrs) 1965 kmem_free(data->sed_attrs, 1966 sizeof(*data->sed_attrs)); 1967 kmem_free(data, sizeof(*data)); 1968 } 1969 1970 /* 1971 * A child lwp of a posix_spawn operation starts here and ends up in 1972 * cpu_spawn_return, dealing with all filedescriptor and scheduler 1973 * manipulations in between. 1974 * The parent waits for the child, as it is not clear whether the child 1975 * will be able to acquire its own exec_lock. If it can, the parent can 1976 * be released early and continue running in parallel. If not (or if the 1977 * magic debug flag is passed in the scheduler attribute struct), the 1978 * child rides on the parent's exec lock until it is ready to return to 1979 * to userland - and only then releases the parent. This method loses 1980 * concurrency, but improves error reporting. 1981 */ 1982 static void 1983 spawn_return(void *arg) 1984 { 1985 struct spawn_exec_data *spawn_data = arg; 1986 struct lwp *l = curlwp; 1987 int error, newfd; 1988 int ostat; 1989 size_t i; 1990 const struct posix_spawn_file_actions_entry *fae; 1991 pid_t ppid; 1992 register_t retval; 1993 bool have_reflock; 1994 bool parent_is_waiting = true; 1995 1996 /* 1997 * Check if we can release parent early. 1998 * We either need to have no sed_attrs, or sed_attrs does not 1999 * have POSIX_SPAWN_RETURNERROR or one of the flags, that require 2000 * safe access to the parent proc (passed in sed_parent). 2001 * We then try to get the exec_lock, and only if that works, we can 2002 * release the parent here already. 2003 */ 2004 ppid = spawn_data->sed_parent->p_pid; 2005 if ((!spawn_data->sed_attrs 2006 || (spawn_data->sed_attrs->sa_flags 2007 & (POSIX_SPAWN_RETURNERROR|POSIX_SPAWN_SETPGROUP)) == 0) 2008 && rw_tryenter(&exec_lock, RW_READER)) { 2009 parent_is_waiting = false; 2010 mutex_enter(&spawn_data->sed_mtx_child); 2011 cv_signal(&spawn_data->sed_cv_child_ready); 2012 mutex_exit(&spawn_data->sed_mtx_child); 2013 } 2014 2015 /* don't allow debugger access yet */ 2016 rw_enter(&l->l_proc->p_reflock, RW_WRITER); 2017 have_reflock = true; 2018 2019 error = 0; 2020 /* handle posix_spawn_file_actions */ 2021 if (spawn_data->sed_actions != NULL) { 2022 for (i = 0; i < spawn_data->sed_actions->len; i++) { 2023 fae = &spawn_data->sed_actions->fae[i]; 2024 switch (fae->fae_action) { 2025 case FAE_OPEN: 2026 if (fd_getfile(fae->fae_fildes) != NULL) { 2027 error = fd_close(fae->fae_fildes); 2028 if (error) 2029 break; 2030 } 2031 error = fd_open(fae->fae_path, fae->fae_oflag, 2032 fae->fae_mode, &newfd); 2033 if (error) 2034 break; 2035 if (newfd != fae->fae_fildes) { 2036 error = dodup(l, newfd, 2037 fae->fae_fildes, 0, &retval); 2038 if (fd_getfile(newfd) != NULL) 2039 fd_close(newfd); 2040 } 2041 break; 2042 case FAE_DUP2: 2043 error = dodup(l, fae->fae_fildes, 2044 fae->fae_newfildes, 0, &retval); 2045 break; 2046 case FAE_CLOSE: 2047 if (fd_getfile(fae->fae_fildes) == NULL) { 2048 error = EBADF; 2049 break; 2050 } 2051 error = fd_close(fae->fae_fildes); 2052 break; 2053 } 2054 if (error) 2055 goto report_error; 2056 } 2057 } 2058 2059 /* handle posix_spawnattr */ 2060 if (spawn_data->sed_attrs != NULL) { 2061 struct sigaction sigact; 2062 sigact._sa_u._sa_handler = SIG_DFL; 2063 sigact.sa_flags = 0; 2064 2065 /* 2066 * set state to SSTOP so that this proc can be found by pid. 2067 * see proc_enterprp, do_sched_setparam below 2068 */ 2069 mutex_enter(proc_lock); 2070 /* 2071 * p_stat should be SACTIVE, so we need to adjust the 2072 * parent's p_nstopchild here. For safety, just make 2073 * we're on the good side of SDEAD before we adjust. 2074 */ 2075 ostat = l->l_proc->p_stat; 2076 KASSERT(ostat < SSTOP); 2077 l->l_proc->p_stat = SSTOP; 2078 l->l_proc->p_waited = 0; 2079 l->l_proc->p_pptr->p_nstopchild++; 2080 mutex_exit(proc_lock); 2081 2082 /* Set process group */ 2083 if (spawn_data->sed_attrs->sa_flags & POSIX_SPAWN_SETPGROUP) { 2084 pid_t mypid = l->l_proc->p_pid, 2085 pgrp = spawn_data->sed_attrs->sa_pgroup; 2086 2087 if (pgrp == 0) 2088 pgrp = mypid; 2089 2090 error = proc_enterpgrp(spawn_data->sed_parent, 2091 mypid, pgrp, false); 2092 if (error) 2093 goto report_error_stopped; 2094 } 2095 2096 /* Set scheduler policy */ 2097 if (spawn_data->sed_attrs->sa_flags & POSIX_SPAWN_SETSCHEDULER) 2098 error = do_sched_setparam(l->l_proc->p_pid, 0, 2099 spawn_data->sed_attrs->sa_schedpolicy, 2100 &spawn_data->sed_attrs->sa_schedparam); 2101 else if (spawn_data->sed_attrs->sa_flags 2102 & POSIX_SPAWN_SETSCHEDPARAM) { 2103 error = do_sched_setparam(ppid, 0, 2104 SCHED_NONE, &spawn_data->sed_attrs->sa_schedparam); 2105 } 2106 if (error) 2107 goto report_error_stopped; 2108 2109 /* Reset user ID's */ 2110 if (spawn_data->sed_attrs->sa_flags & POSIX_SPAWN_RESETIDS) { 2111 error = do_setresuid(l, -1, 2112 kauth_cred_getgid(l->l_cred), -1, 2113 ID_E_EQ_R | ID_E_EQ_S); 2114 if (error) 2115 goto report_error_stopped; 2116 error = do_setresuid(l, -1, 2117 kauth_cred_getuid(l->l_cred), -1, 2118 ID_E_EQ_R | ID_E_EQ_S); 2119 if (error) 2120 goto report_error_stopped; 2121 } 2122 2123 /* Set signal masks/defaults */ 2124 if (spawn_data->sed_attrs->sa_flags & POSIX_SPAWN_SETSIGMASK) { 2125 mutex_enter(l->l_proc->p_lock); 2126 error = sigprocmask1(l, SIG_SETMASK, 2127 &spawn_data->sed_attrs->sa_sigmask, NULL); 2128 mutex_exit(l->l_proc->p_lock); 2129 if (error) 2130 goto report_error_stopped; 2131 } 2132 2133 if (spawn_data->sed_attrs->sa_flags & POSIX_SPAWN_SETSIGDEF) { 2134 /* 2135 * The following sigaction call is using a sigaction 2136 * version 0 trampoline which is in the compatibility 2137 * code only. This is not a problem because for SIG_DFL 2138 * and SIG_IGN, the trampolines are now ignored. If they 2139 * were not, this would be a problem because we are 2140 * holding the exec_lock, and the compat code needs 2141 * to do the same in order to replace the trampoline 2142 * code of the process. 2143 */ 2144 for (i = 1; i <= NSIG; i++) { 2145 if (sigismember( 2146 &spawn_data->sed_attrs->sa_sigdefault, i)) 2147 sigaction1(l, i, &sigact, NULL, NULL, 2148 0); 2149 } 2150 } 2151 mutex_enter(proc_lock); 2152 l->l_proc->p_stat = ostat; 2153 l->l_proc->p_pptr->p_nstopchild--; 2154 mutex_exit(proc_lock); 2155 } 2156 2157 /* now do the real exec */ 2158 error = execve_runproc(l, &spawn_data->sed_exec, parent_is_waiting, 2159 true); 2160 have_reflock = false; 2161 if (error == EJUSTRETURN) 2162 error = 0; 2163 else if (error) 2164 goto report_error; 2165 2166 if (parent_is_waiting) { 2167 mutex_enter(&spawn_data->sed_mtx_child); 2168 cv_signal(&spawn_data->sed_cv_child_ready); 2169 mutex_exit(&spawn_data->sed_mtx_child); 2170 } 2171 2172 /* release our refcount on the data */ 2173 spawn_exec_data_release(spawn_data); 2174 2175 /* and finally: leave to userland for the first time */ 2176 cpu_spawn_return(l); 2177 2178 /* NOTREACHED */ 2179 return; 2180 2181 report_error_stopped: 2182 mutex_enter(proc_lock); 2183 l->l_proc->p_stat = ostat; 2184 l->l_proc->p_pptr->p_nstopchild--; 2185 mutex_exit(proc_lock); 2186 report_error: 2187 if (have_reflock) { 2188 /* 2189 * We have not passed through execve_runproc(), 2190 * which would have released the p_reflock and also 2191 * taken ownership of the sed_exec part of spawn_data, 2192 * so release/free both here. 2193 */ 2194 rw_exit(&l->l_proc->p_reflock); 2195 execve_free_data(&spawn_data->sed_exec); 2196 } 2197 2198 if (parent_is_waiting) { 2199 /* pass error to parent */ 2200 mutex_enter(&spawn_data->sed_mtx_child); 2201 spawn_data->sed_error = error; 2202 cv_signal(&spawn_data->sed_cv_child_ready); 2203 mutex_exit(&spawn_data->sed_mtx_child); 2204 } else { 2205 rw_exit(&exec_lock); 2206 } 2207 2208 /* release our refcount on the data */ 2209 spawn_exec_data_release(spawn_data); 2210 2211 /* done, exit */ 2212 mutex_enter(l->l_proc->p_lock); 2213 /* 2214 * Posix explicitly asks for an exit code of 127 if we report 2215 * errors from the child process - so, unfortunately, there 2216 * is no way to report a more exact error code. 2217 * A NetBSD specific workaround is POSIX_SPAWN_RETURNERROR as 2218 * flag bit in the attrp argument to posix_spawn(2), see above. 2219 */ 2220 exit1(l, 127, 0); 2221 } 2222 2223 void 2224 posix_spawn_fa_free(struct posix_spawn_file_actions *fa, size_t len) 2225 { 2226 2227 for (size_t i = 0; i < len; i++) { 2228 struct posix_spawn_file_actions_entry *fae = &fa->fae[i]; 2229 if (fae->fae_action != FAE_OPEN) 2230 continue; 2231 kmem_strfree(fae->fae_path); 2232 } 2233 if (fa->len > 0) 2234 kmem_free(fa->fae, sizeof(*fa->fae) * fa->len); 2235 kmem_free(fa, sizeof(*fa)); 2236 } 2237 2238 static int 2239 posix_spawn_fa_alloc(struct posix_spawn_file_actions **fap, 2240 const struct posix_spawn_file_actions *ufa, rlim_t lim) 2241 { 2242 struct posix_spawn_file_actions *fa; 2243 struct posix_spawn_file_actions_entry *fae; 2244 char *pbuf = NULL; 2245 int error; 2246 size_t i = 0; 2247 2248 fa = kmem_alloc(sizeof(*fa), KM_SLEEP); 2249 error = copyin(ufa, fa, sizeof(*fa)); 2250 if (error || fa->len == 0) { 2251 kmem_free(fa, sizeof(*fa)); 2252 return error; /* 0 if not an error, and len == 0 */ 2253 } 2254 2255 if (fa->len > lim) { 2256 kmem_free(fa, sizeof(*fa)); 2257 return EINVAL; 2258 } 2259 2260 fa->size = fa->len; 2261 size_t fal = fa->len * sizeof(*fae); 2262 fae = fa->fae; 2263 fa->fae = kmem_alloc(fal, KM_SLEEP); 2264 error = copyin(fae, fa->fae, fal); 2265 if (error) 2266 goto out; 2267 2268 pbuf = PNBUF_GET(); 2269 for (; i < fa->len; i++) { 2270 fae = &fa->fae[i]; 2271 if (fae->fae_action != FAE_OPEN) 2272 continue; 2273 error = copyinstr(fae->fae_path, pbuf, MAXPATHLEN, &fal); 2274 if (error) 2275 goto out; 2276 fae->fae_path = kmem_alloc(fal, KM_SLEEP); 2277 memcpy(fae->fae_path, pbuf, fal); 2278 } 2279 PNBUF_PUT(pbuf); 2280 2281 *fap = fa; 2282 return 0; 2283 out: 2284 if (pbuf) 2285 PNBUF_PUT(pbuf); 2286 posix_spawn_fa_free(fa, i); 2287 return error; 2288 } 2289 2290 int 2291 check_posix_spawn(struct lwp *l1) 2292 { 2293 int error, tnprocs, count; 2294 uid_t uid; 2295 struct proc *p1; 2296 2297 p1 = l1->l_proc; 2298 uid = kauth_cred_getuid(l1->l_cred); 2299 tnprocs = atomic_inc_uint_nv(&nprocs); 2300 2301 /* 2302 * Although process entries are dynamically created, we still keep 2303 * a global limit on the maximum number we will create. 2304 */ 2305 if (__predict_false(tnprocs >= maxproc)) 2306 error = -1; 2307 else 2308 error = kauth_authorize_process(l1->l_cred, 2309 KAUTH_PROCESS_FORK, p1, KAUTH_ARG(tnprocs), NULL, NULL); 2310 2311 if (error) { 2312 atomic_dec_uint(&nprocs); 2313 return EAGAIN; 2314 } 2315 2316 /* 2317 * Enforce limits. 2318 */ 2319 count = chgproccnt(uid, 1); 2320 if (kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_RLIMIT, 2321 p1, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS), 2322 &p1->p_rlimit[RLIMIT_NPROC], KAUTH_ARG(RLIMIT_NPROC)) != 0 && 2323 __predict_false(count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) { 2324 (void)chgproccnt(uid, -1); 2325 atomic_dec_uint(&nprocs); 2326 return EAGAIN; 2327 } 2328 2329 return 0; 2330 } 2331 2332 int 2333 do_posix_spawn(struct lwp *l1, pid_t *pid_res, bool *child_ok, const char *path, 2334 struct posix_spawn_file_actions *fa, 2335 struct posix_spawnattr *sa, 2336 char *const *argv, char *const *envp, 2337 execve_fetch_element_t fetch) 2338 { 2339 2340 struct proc *p1, *p2; 2341 struct lwp *l2; 2342 int error; 2343 struct spawn_exec_data *spawn_data; 2344 vaddr_t uaddr; 2345 pid_t pid; 2346 bool have_exec_lock = false; 2347 2348 p1 = l1->l_proc; 2349 2350 /* Allocate and init spawn_data */ 2351 spawn_data = kmem_zalloc(sizeof(*spawn_data), KM_SLEEP); 2352 spawn_data->sed_refcnt = 1; /* only parent so far */ 2353 cv_init(&spawn_data->sed_cv_child_ready, "pspawn"); 2354 mutex_init(&spawn_data->sed_mtx_child, MUTEX_DEFAULT, IPL_NONE); 2355 mutex_enter(&spawn_data->sed_mtx_child); 2356 2357 /* 2358 * Do the first part of the exec now, collect state 2359 * in spawn_data. 2360 */ 2361 error = execve_loadvm(l1, path, argv, 2362 envp, fetch, &spawn_data->sed_exec); 2363 if (error == EJUSTRETURN) 2364 error = 0; 2365 else if (error) 2366 goto error_exit; 2367 2368 have_exec_lock = true; 2369 2370 /* 2371 * Allocate virtual address space for the U-area now, while it 2372 * is still easy to abort the fork operation if we're out of 2373 * kernel virtual address space. 2374 */ 2375 uaddr = uvm_uarea_alloc(); 2376 if (__predict_false(uaddr == 0)) { 2377 error = ENOMEM; 2378 goto error_exit; 2379 } 2380 2381 /* 2382 * Allocate new proc. Borrow proc0 vmspace for it, we will 2383 * replace it with its own before returning to userland 2384 * in the child. 2385 * This is a point of no return, we will have to go through 2386 * the child proc to properly clean it up past this point. 2387 */ 2388 p2 = proc_alloc(); 2389 pid = p2->p_pid; 2390 2391 /* 2392 * Make a proc table entry for the new process. 2393 * Start by zeroing the section of proc that is zero-initialized, 2394 * then copy the section that is copied directly from the parent. 2395 */ 2396 memset(&p2->p_startzero, 0, 2397 (unsigned) ((char *)&p2->p_endzero - (char *)&p2->p_startzero)); 2398 memcpy(&p2->p_startcopy, &p1->p_startcopy, 2399 (unsigned) ((char *)&p2->p_endcopy - (char *)&p2->p_startcopy)); 2400 p2->p_vmspace = proc0.p_vmspace; 2401 2402 TAILQ_INIT(&p2->p_sigpend.sp_info); 2403 2404 LIST_INIT(&p2->p_lwps); 2405 LIST_INIT(&p2->p_sigwaiters); 2406 2407 /* 2408 * Duplicate sub-structures as needed. 2409 * Increase reference counts on shared objects. 2410 * Inherit flags we want to keep. The flags related to SIGCHLD 2411 * handling are important in order to keep a consistent behaviour 2412 * for the child after the fork. If we are a 32-bit process, the 2413 * child will be too. 2414 */ 2415 p2->p_flag = 2416 p1->p_flag & (PK_SUGID | PK_NOCLDWAIT | PK_CLDSIGIGN | PK_32); 2417 p2->p_emul = p1->p_emul; 2418 p2->p_execsw = p1->p_execsw; 2419 2420 mutex_init(&p2->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 2421 mutex_init(&p2->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 2422 rw_init(&p2->p_reflock); 2423 cv_init(&p2->p_waitcv, "wait"); 2424 cv_init(&p2->p_lwpcv, "lwpwait"); 2425 2426 p2->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 2427 2428 kauth_proc_fork(p1, p2); 2429 2430 p2->p_raslist = NULL; 2431 p2->p_fd = fd_copy(); 2432 2433 /* XXX racy */ 2434 p2->p_mqueue_cnt = p1->p_mqueue_cnt; 2435 2436 p2->p_cwdi = cwdinit(); 2437 2438 /* 2439 * Note: p_limit (rlimit stuff) is copy-on-write, so normally 2440 * we just need increase pl_refcnt. 2441 */ 2442 if (!p1->p_limit->pl_writeable) { 2443 lim_addref(p1->p_limit); 2444 p2->p_limit = p1->p_limit; 2445 } else { 2446 p2->p_limit = lim_copy(p1->p_limit); 2447 } 2448 2449 p2->p_lflag = 0; 2450 p2->p_sflag = 0; 2451 p2->p_slflag = 0; 2452 p2->p_pptr = p1; 2453 p2->p_ppid = p1->p_pid; 2454 LIST_INIT(&p2->p_children); 2455 2456 p2->p_aio = NULL; 2457 2458 #ifdef KTRACE 2459 /* 2460 * Copy traceflag and tracefile if enabled. 2461 * If not inherited, these were zeroed above. 2462 */ 2463 if (p1->p_traceflag & KTRFAC_INHERIT) { 2464 mutex_enter(&ktrace_lock); 2465 p2->p_traceflag = p1->p_traceflag; 2466 if ((p2->p_tracep = p1->p_tracep) != NULL) 2467 ktradref(p2); 2468 mutex_exit(&ktrace_lock); 2469 } 2470 #endif 2471 2472 /* 2473 * Create signal actions for the child process. 2474 */ 2475 p2->p_sigacts = sigactsinit(p1, 0); 2476 mutex_enter(p1->p_lock); 2477 p2->p_sflag |= 2478 (p1->p_sflag & (PS_STOPFORK | PS_STOPEXEC | PS_NOCLDSTOP)); 2479 sched_proc_fork(p1, p2); 2480 mutex_exit(p1->p_lock); 2481 2482 p2->p_stflag = p1->p_stflag; 2483 2484 /* 2485 * p_stats. 2486 * Copy parts of p_stats, and zero out the rest. 2487 */ 2488 p2->p_stats = pstatscopy(p1->p_stats); 2489 2490 /* copy over machdep flags to the new proc */ 2491 cpu_proc_fork(p1, p2); 2492 2493 /* 2494 * Prepare remaining parts of spawn data 2495 */ 2496 spawn_data->sed_actions = fa; 2497 spawn_data->sed_attrs = sa; 2498 2499 spawn_data->sed_parent = p1; 2500 2501 /* create LWP */ 2502 lwp_create(l1, p2, uaddr, 0, NULL, 0, spawn_return, spawn_data, 2503 &l2, l1->l_class, &l1->l_sigmask, &l1->l_sigstk); 2504 l2->l_ctxlink = NULL; /* reset ucontext link */ 2505 2506 /* 2507 * Copy the credential so other references don't see our changes. 2508 * Test to see if this is necessary first, since in the common case 2509 * we won't need a private reference. 2510 */ 2511 if (kauth_cred_geteuid(l2->l_cred) != kauth_cred_getsvuid(l2->l_cred) || 2512 kauth_cred_getegid(l2->l_cred) != kauth_cred_getsvgid(l2->l_cred)) { 2513 l2->l_cred = kauth_cred_copy(l2->l_cred); 2514 kauth_cred_setsvuid(l2->l_cred, kauth_cred_geteuid(l2->l_cred)); 2515 kauth_cred_setsvgid(l2->l_cred, kauth_cred_getegid(l2->l_cred)); 2516 } 2517 2518 /* Update the master credentials. */ 2519 if (l2->l_cred != p2->p_cred) { 2520 kauth_cred_t ocred; 2521 2522 kauth_cred_hold(l2->l_cred); 2523 mutex_enter(p2->p_lock); 2524 ocred = p2->p_cred; 2525 p2->p_cred = l2->l_cred; 2526 mutex_exit(p2->p_lock); 2527 kauth_cred_free(ocred); 2528 } 2529 2530 *child_ok = true; 2531 spawn_data->sed_refcnt = 2; /* child gets it as well */ 2532 #if 0 2533 l2->l_nopreempt = 1; /* start it non-preemptable */ 2534 #endif 2535 2536 /* 2537 * It's now safe for the scheduler and other processes to see the 2538 * child process. 2539 */ 2540 mutex_enter(proc_lock); 2541 2542 if (p1->p_session->s_ttyvp != NULL && p1->p_lflag & PL_CONTROLT) 2543 p2->p_lflag |= PL_CONTROLT; 2544 2545 LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling); 2546 p2->p_exitsig = SIGCHLD; /* signal for parent on exit */ 2547 2548 LIST_INSERT_AFTER(p1, p2, p_pglist); 2549 LIST_INSERT_HEAD(&allproc, p2, p_list); 2550 2551 p2->p_trace_enabled = trace_is_enabled(p2); 2552 #ifdef __HAVE_SYSCALL_INTERN 2553 (*p2->p_emul->e_syscall_intern)(p2); 2554 #endif 2555 2556 /* 2557 * Make child runnable, set start time, and add to run queue except 2558 * if the parent requested the child to start in SSTOP state. 2559 */ 2560 mutex_enter(p2->p_lock); 2561 2562 getmicrotime(&p2->p_stats->p_start); 2563 2564 lwp_lock(l2); 2565 KASSERT(p2->p_nrlwps == 1); 2566 p2->p_nrlwps = 1; 2567 p2->p_stat = SACTIVE; 2568 l2->l_stat = LSRUN; 2569 sched_enqueue(l2, false); 2570 lwp_unlock(l2); 2571 2572 mutex_exit(p2->p_lock); 2573 mutex_exit(proc_lock); 2574 2575 cv_wait(&spawn_data->sed_cv_child_ready, &spawn_data->sed_mtx_child); 2576 error = spawn_data->sed_error; 2577 mutex_exit(&spawn_data->sed_mtx_child); 2578 spawn_exec_data_release(spawn_data); 2579 2580 rw_exit(&p1->p_reflock); 2581 rw_exit(&exec_lock); 2582 have_exec_lock = false; 2583 2584 *pid_res = pid; 2585 return error; 2586 2587 error_exit: 2588 if (have_exec_lock) { 2589 execve_free_data(&spawn_data->sed_exec); 2590 rw_exit(&p1->p_reflock); 2591 rw_exit(&exec_lock); 2592 } 2593 mutex_exit(&spawn_data->sed_mtx_child); 2594 spawn_exec_data_release(spawn_data); 2595 2596 return error; 2597 } 2598 2599 int 2600 sys_posix_spawn(struct lwp *l1, const struct sys_posix_spawn_args *uap, 2601 register_t *retval) 2602 { 2603 /* { 2604 syscallarg(pid_t *) pid; 2605 syscallarg(const char *) path; 2606 syscallarg(const struct posix_spawn_file_actions *) file_actions; 2607 syscallarg(const struct posix_spawnattr *) attrp; 2608 syscallarg(char *const *) argv; 2609 syscallarg(char *const *) envp; 2610 } */ 2611 2612 int error; 2613 struct posix_spawn_file_actions *fa = NULL; 2614 struct posix_spawnattr *sa = NULL; 2615 pid_t pid; 2616 bool child_ok = false; 2617 rlim_t max_fileactions; 2618 proc_t *p = l1->l_proc; 2619 2620 error = check_posix_spawn(l1); 2621 if (error) { 2622 *retval = error; 2623 return 0; 2624 } 2625 2626 /* copy in file_actions struct */ 2627 if (SCARG(uap, file_actions) != NULL) { 2628 max_fileactions = 2 * uimin(p->p_rlimit[RLIMIT_NOFILE].rlim_cur, 2629 maxfiles); 2630 error = posix_spawn_fa_alloc(&fa, SCARG(uap, file_actions), 2631 max_fileactions); 2632 if (error) 2633 goto error_exit; 2634 } 2635 2636 /* copyin posix_spawnattr struct */ 2637 if (SCARG(uap, attrp) != NULL) { 2638 sa = kmem_alloc(sizeof(*sa), KM_SLEEP); 2639 error = copyin(SCARG(uap, attrp), sa, sizeof(*sa)); 2640 if (error) 2641 goto error_exit; 2642 } 2643 2644 /* 2645 * Do the spawn 2646 */ 2647 error = do_posix_spawn(l1, &pid, &child_ok, SCARG(uap, path), fa, sa, 2648 SCARG(uap, argv), SCARG(uap, envp), execve_fetch_element); 2649 if (error) 2650 goto error_exit; 2651 2652 if (error == 0 && SCARG(uap, pid) != NULL) 2653 error = copyout(&pid, SCARG(uap, pid), sizeof(pid)); 2654 2655 *retval = error; 2656 return 0; 2657 2658 error_exit: 2659 if (!child_ok) { 2660 (void)chgproccnt(kauth_cred_getuid(l1->l_cred), -1); 2661 atomic_dec_uint(&nprocs); 2662 2663 if (sa) 2664 kmem_free(sa, sizeof(*sa)); 2665 if (fa) 2666 posix_spawn_fa_free(fa, fa->len); 2667 } 2668 2669 *retval = error; 2670 return 0; 2671 } 2672 2673 void 2674 exec_free_emul_arg(struct exec_package *epp) 2675 { 2676 if (epp->ep_emul_arg_free != NULL) { 2677 KASSERT(epp->ep_emul_arg != NULL); 2678 (*epp->ep_emul_arg_free)(epp->ep_emul_arg); 2679 epp->ep_emul_arg_free = NULL; 2680 epp->ep_emul_arg = NULL; 2681 } else { 2682 KASSERT(epp->ep_emul_arg == NULL); 2683 } 2684 } 2685 2686 #ifdef DEBUG_EXEC 2687 static void 2688 dump_vmcmds(const struct exec_package * const epp, size_t x, int error) 2689 { 2690 struct exec_vmcmd *vp = &epp->ep_vmcmds.evs_cmds[0]; 2691 size_t j; 2692 2693 if (error == 0) 2694 DPRINTF(("vmcmds %u\n", epp->ep_vmcmds.evs_used)); 2695 else 2696 DPRINTF(("vmcmds %zu/%u, error %d\n", x, 2697 epp->ep_vmcmds.evs_used, error)); 2698 2699 for (j = 0; j < epp->ep_vmcmds.evs_used; j++) { 2700 DPRINTF(("vmcmd[%zu] = vmcmd_map_%s %#" 2701 PRIxVADDR"/%#"PRIxVSIZE" fd@%#" 2702 PRIxVSIZE" prot=0%o flags=%d\n", j, 2703 vp[j].ev_proc == vmcmd_map_pagedvn ? 2704 "pagedvn" : 2705 vp[j].ev_proc == vmcmd_map_readvn ? 2706 "readvn" : 2707 vp[j].ev_proc == vmcmd_map_zero ? 2708 "zero" : "*unknown*", 2709 vp[j].ev_addr, vp[j].ev_len, 2710 vp[j].ev_offset, vp[j].ev_prot, 2711 vp[j].ev_flags)); 2712 if (error != 0 && j == x) 2713 DPRINTF((" ^--- failed\n")); 2714 } 2715 } 2716 #endif 2717