1 /*- 2 * SPDX-License-Identifier: BSD-4-Clause 3 * 4 * Copyright (c) 1995 Terrence R. Lambert 5 * All rights reserved. 6 * 7 * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993 8 * The Regents of the University of California. All rights reserved. 9 * (c) UNIX System Laboratories, Inc. 10 * All or some portions of this file are derived from material licensed 11 * to the University of California by American Telephone and Telegraph 12 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 13 * the permission of UNIX System Laboratories, Inc. 14 * 15 * Redistribution and use in source and binary forms, with or without 16 * modification, are permitted provided that the following conditions 17 * are met: 18 * 1. Redistributions of source code must retain the above copyright 19 * notice, this list of conditions and the following disclaimer. 20 * 2. Redistributions in binary form must reproduce the above copyright 21 * notice, this list of conditions and the following disclaimer in the 22 * documentation and/or other materials provided with the distribution. 23 * 3. All advertising materials mentioning features or use of this software 24 * must display the following acknowledgement: 25 * This product includes software developed by the University of 26 * California, Berkeley and its contributors. 27 * 4. Neither the name of the University nor the names of its contributors 28 * may be used to endorse or promote products derived from this software 29 * without specific prior written permission. 30 * 31 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 32 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 33 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 34 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 35 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 36 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 37 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 38 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 39 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 40 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 41 * SUCH DAMAGE. 42 * 43 * @(#)init_main.c 8.9 (Berkeley) 1/21/94 44 */ 45 46 #include <sys/cdefs.h> 47 __FBSDID("$FreeBSD$"); 48 49 #include "opt_ddb.h" 50 #include "opt_init_path.h" 51 #include "opt_verbose_sysinit.h" 52 53 #include <sys/param.h> 54 #include <sys/kernel.h> 55 #include <sys/epoch.h> 56 #include <sys/exec.h> 57 #include <sys/file.h> 58 #include <sys/filedesc.h> 59 #include <sys/imgact.h> 60 #include <sys/jail.h> 61 #include <sys/ktr.h> 62 #include <sys/lock.h> 63 #include <sys/loginclass.h> 64 #include <sys/mount.h> 65 #include <sys/mutex.h> 66 #include <sys/syscallsubr.h> 67 #include <sys/sysctl.h> 68 #include <sys/proc.h> 69 #include <sys/racct.h> 70 #include <sys/resourcevar.h> 71 #include <sys/systm.h> 72 #include <sys/signalvar.h> 73 #include <sys/vnode.h> 74 #include <sys/sysent.h> 75 #include <sys/reboot.h> 76 #include <sys/sched.h> 77 #include <sys/sx.h> 78 #include <sys/sysproto.h> 79 #include <sys/vmmeter.h> 80 #include <sys/unistd.h> 81 #include <sys/malloc.h> 82 #include <sys/conf.h> 83 #include <sys/cpuset.h> 84 85 #include <machine/cpu.h> 86 87 #include <security/audit/audit.h> 88 #include <security/mac/mac_framework.h> 89 90 #include <vm/vm.h> 91 #include <vm/vm_param.h> 92 #include <vm/vm_extern.h> 93 #include <vm/pmap.h> 94 #include <vm/vm_map.h> 95 #include <sys/copyright.h> 96 97 #include <ddb/ddb.h> 98 #include <ddb/db_sym.h> 99 100 void mi_startup(void); /* Should be elsewhere */ 101 102 /* Components of the first process -- never freed. */ 103 static struct session session0; 104 static struct pgrp pgrp0; 105 struct proc proc0; 106 struct thread0_storage thread0_st __aligned(32); 107 struct vmspace vmspace0; 108 struct proc *initproc; 109 110 #ifndef BOOTHOWTO 111 #define BOOTHOWTO 0 112 #endif 113 int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 114 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, 115 "Boot control flags, passed from loader"); 116 117 #ifndef BOOTVERBOSE 118 #define BOOTVERBOSE 0 119 #endif 120 int bootverbose = BOOTVERBOSE; 121 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 122 "Control the output of verbose kernel messages"); 123 124 #ifdef VERBOSE_SYSINIT 125 /* 126 * We'll use the defined value of VERBOSE_SYSINIT from the kernel config to 127 * dictate the default VERBOSE_SYSINIT behavior. Significant values for this 128 * option and associated tunable are: 129 * - 0, 'compiled in but silent by default' 130 * - 1, 'compiled in but verbose by default' (default) 131 */ 132 int verbose_sysinit = VERBOSE_SYSINIT; 133 TUNABLE_INT("debug.verbose_sysinit", &verbose_sysinit); 134 #endif 135 136 #ifdef INVARIANTS 137 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance"); 138 #endif 139 140 /* 141 * This ensures that there is at least one entry so that the sysinit_set 142 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 143 * executed. 144 */ 145 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 146 147 /* 148 * The sysinit table itself. Items are checked off as the are run. 149 * If we want to register new sysinit types, add them to newsysinit. 150 */ 151 SET_DECLARE(sysinit_set, struct sysinit); 152 struct sysinit **sysinit, **sysinit_end; 153 struct sysinit **newsysinit, **newsysinit_end; 154 155 EVENTHANDLER_LIST_DECLARE(process_init); 156 EVENTHANDLER_LIST_DECLARE(thread_init); 157 EVENTHANDLER_LIST_DECLARE(process_ctor); 158 EVENTHANDLER_LIST_DECLARE(thread_ctor); 159 160 /* 161 * Merge a new sysinit set into the current set, reallocating it if 162 * necessary. This can only be called after malloc is running. 163 */ 164 void 165 sysinit_add(struct sysinit **set, struct sysinit **set_end) 166 { 167 struct sysinit **newset; 168 struct sysinit **sipp; 169 struct sysinit **xipp; 170 int count; 171 172 count = set_end - set; 173 if (newsysinit) 174 count += newsysinit_end - newsysinit; 175 else 176 count += sysinit_end - sysinit; 177 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 178 if (newset == NULL) 179 panic("cannot malloc for sysinit"); 180 xipp = newset; 181 if (newsysinit) 182 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 183 *xipp++ = *sipp; 184 else 185 for (sipp = sysinit; sipp < sysinit_end; sipp++) 186 *xipp++ = *sipp; 187 for (sipp = set; sipp < set_end; sipp++) 188 *xipp++ = *sipp; 189 if (newsysinit) 190 free(newsysinit, M_TEMP); 191 newsysinit = newset; 192 newsysinit_end = newset + count; 193 } 194 195 #if defined (DDB) && defined(VERBOSE_SYSINIT) 196 static const char * 197 symbol_name(vm_offset_t va, db_strategy_t strategy) 198 { 199 const char *name; 200 c_db_sym_t sym; 201 db_expr_t offset; 202 203 if (va == 0) 204 return (NULL); 205 sym = db_search_symbol(va, strategy, &offset); 206 if (offset != 0) 207 return (NULL); 208 db_symbol_values(sym, &name, NULL); 209 return (name); 210 } 211 #endif 212 213 /* 214 * System startup; initialize the world, create process 0, mount root 215 * filesystem, and fork to create init and pagedaemon. Most of the 216 * hard work is done in the lower-level initialization routines including 217 * startup(), which does memory initialization and autoconfiguration. 218 * 219 * This allows simple addition of new kernel subsystems that require 220 * boot time initialization. It also allows substitution of subsystem 221 * (for instance, a scheduler, kernel profiler, or VM system) by object 222 * module. Finally, it allows for optional "kernel threads". 223 */ 224 void 225 mi_startup(void) 226 { 227 228 struct sysinit **sipp; /* system initialization*/ 229 struct sysinit **xipp; /* interior loop of sort*/ 230 struct sysinit *save; /* bubble*/ 231 232 #if defined(VERBOSE_SYSINIT) 233 int last; 234 int verbose; 235 #endif 236 237 TSENTER(); 238 239 if (boothowto & RB_VERBOSE) 240 bootverbose++; 241 242 if (sysinit == NULL) { 243 sysinit = SET_BEGIN(sysinit_set); 244 sysinit_end = SET_LIMIT(sysinit_set); 245 } 246 247 restart: 248 /* 249 * Perform a bubble sort of the system initialization objects by 250 * their subsystem (primary key) and order (secondary key). 251 */ 252 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 253 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 254 if ((*sipp)->subsystem < (*xipp)->subsystem || 255 ((*sipp)->subsystem == (*xipp)->subsystem && 256 (*sipp)->order <= (*xipp)->order)) 257 continue; /* skip*/ 258 save = *sipp; 259 *sipp = *xipp; 260 *xipp = save; 261 } 262 } 263 264 #if defined(VERBOSE_SYSINIT) 265 last = SI_SUB_COPYRIGHT; 266 verbose = 0; 267 #if !defined(DDB) 268 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 269 #endif 270 #endif 271 272 /* 273 * Traverse the (now) ordered list of system initialization tasks. 274 * Perform each task, and continue on to the next task. 275 */ 276 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 277 278 if ((*sipp)->subsystem == SI_SUB_DUMMY) 279 continue; /* skip dummy task(s)*/ 280 281 if ((*sipp)->subsystem == SI_SUB_DONE) 282 continue; 283 284 #if defined(VERBOSE_SYSINIT) 285 if ((*sipp)->subsystem > last && verbose_sysinit != 0) { 286 verbose = 1; 287 last = (*sipp)->subsystem; 288 printf("subsystem %x\n", last); 289 } 290 if (verbose) { 291 #if defined(DDB) 292 const char *func, *data; 293 294 func = symbol_name((vm_offset_t)(*sipp)->func, 295 DB_STGY_PROC); 296 data = symbol_name((vm_offset_t)(*sipp)->udata, 297 DB_STGY_ANY); 298 if (func != NULL && data != NULL) 299 printf(" %s(&%s)... ", func, data); 300 else if (func != NULL) 301 printf(" %s(%p)... ", func, (*sipp)->udata); 302 else 303 #endif 304 printf(" %p(%p)... ", (*sipp)->func, 305 (*sipp)->udata); 306 } 307 #endif 308 309 /* Call function */ 310 (*((*sipp)->func))((*sipp)->udata); 311 312 #if defined(VERBOSE_SYSINIT) 313 if (verbose) 314 printf("done.\n"); 315 #endif 316 317 /* Check off the one we're just done */ 318 (*sipp)->subsystem = SI_SUB_DONE; 319 320 /* Check if we've installed more sysinit items via KLD */ 321 if (newsysinit != NULL) { 322 if (sysinit != SET_BEGIN(sysinit_set)) 323 free(sysinit, M_TEMP); 324 sysinit = newsysinit; 325 sysinit_end = newsysinit_end; 326 newsysinit = NULL; 327 newsysinit_end = NULL; 328 goto restart; 329 } 330 } 331 332 TSEXIT(); /* Here so we don't overlap with start_init. */ 333 334 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 335 mtx_unlock(&Giant); 336 337 /* 338 * Now hand over this thread to swapper. 339 */ 340 swapper(); 341 /* NOTREACHED*/ 342 } 343 344 static void 345 print_caddr_t(void *data) 346 { 347 printf("%s", (char *)data); 348 } 349 350 static void 351 print_version(void *data __unused) 352 { 353 int len; 354 355 /* Strip a trailing newline from version. */ 356 len = strlen(version); 357 while (len > 0 && version[len - 1] == '\n') 358 len--; 359 printf("%.*s %s\n", len, version, machine); 360 printf("%s\n", compiler_version); 361 } 362 363 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 364 copyright); 365 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 366 trademark); 367 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 368 369 #ifdef WITNESS 370 static char wit_warn[] = 371 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 372 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 373 print_caddr_t, wit_warn); 374 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1, 375 print_caddr_t, wit_warn); 376 #endif 377 378 #ifdef DIAGNOSTIC 379 static char diag_warn[] = 380 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 381 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 382 print_caddr_t, diag_warn); 383 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2, 384 print_caddr_t, diag_warn); 385 #endif 386 387 static int 388 null_fetch_syscall_args(struct thread *td __unused) 389 { 390 391 panic("null_fetch_syscall_args"); 392 } 393 394 static void 395 null_set_syscall_retval(struct thread *td __unused, int error __unused) 396 { 397 398 panic("null_set_syscall_retval"); 399 } 400 401 struct sysentvec null_sysvec = { 402 .sv_size = 0, 403 .sv_table = NULL, 404 .sv_mask = 0, 405 .sv_errsize = 0, 406 .sv_errtbl = NULL, 407 .sv_transtrap = NULL, 408 .sv_fixup = NULL, 409 .sv_sendsig = NULL, 410 .sv_sigcode = NULL, 411 .sv_szsigcode = NULL, 412 .sv_name = "null", 413 .sv_coredump = NULL, 414 .sv_imgact_try = NULL, 415 .sv_minsigstksz = 0, 416 .sv_pagesize = PAGE_SIZE, 417 .sv_minuser = VM_MIN_ADDRESS, 418 .sv_maxuser = VM_MAXUSER_ADDRESS, 419 .sv_usrstack = USRSTACK, 420 .sv_psstrings = PS_STRINGS, 421 .sv_stackprot = VM_PROT_ALL, 422 .sv_copyout_strings = NULL, 423 .sv_setregs = NULL, 424 .sv_fixlimit = NULL, 425 .sv_maxssiz = NULL, 426 .sv_flags = 0, 427 .sv_set_syscall_retval = null_set_syscall_retval, 428 .sv_fetch_syscall_args = null_fetch_syscall_args, 429 .sv_syscallnames = NULL, 430 .sv_schedtail = NULL, 431 .sv_thread_detach = NULL, 432 .sv_trap = NULL, 433 }; 434 435 /* 436 * The two following SYSINIT's are proc0 specific glue code. I am not 437 * convinced that they can not be safely combined, but their order of 438 * operation has been maintained as the same as the original init_main.c 439 * for right now. 440 */ 441 /* ARGSUSED*/ 442 static void 443 proc0_init(void *dummy __unused) 444 { 445 struct proc *p; 446 struct thread *td; 447 struct ucred *newcred; 448 struct uidinfo tmpuinfo; 449 struct loginclass tmplc = { 450 .lc_name = "", 451 }; 452 vm_paddr_t pageablemem; 453 int i; 454 455 GIANT_REQUIRED; 456 p = &proc0; 457 td = &thread0; 458 459 /* 460 * Initialize magic number and osrel. 461 */ 462 p->p_magic = P_MAGIC; 463 p->p_osrel = osreldate; 464 465 /* 466 * Initialize thread and process structures. 467 */ 468 procinit(); /* set up proc zone */ 469 threadinit(); /* set up UMA zones */ 470 471 /* 472 * Initialise scheduler resources. 473 * Add scheduler specific parts to proc, thread as needed. 474 */ 475 schedinit(); /* scheduler gets its house in order */ 476 477 /* 478 * Create process 0 (the swapper). 479 */ 480 LIST_INSERT_HEAD(&allproc, p, p_list); 481 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 482 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 483 p->p_pgrp = &pgrp0; 484 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 485 LIST_INIT(&pgrp0.pg_members); 486 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 487 488 pgrp0.pg_session = &session0; 489 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 490 refcount_init(&session0.s_count, 1); 491 session0.s_leader = p; 492 493 p->p_sysent = &null_sysvec; 494 p->p_flag = P_SYSTEM | P_INMEM | P_KPROC; 495 p->p_flag2 = 0; 496 p->p_state = PRS_NORMAL; 497 p->p_klist = knlist_alloc(&p->p_mtx); 498 STAILQ_INIT(&p->p_ktr); 499 p->p_nice = NZERO; 500 /* pid_max cannot be greater than PID_MAX */ 501 td->td_tid = PID_MAX + 1; 502 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash); 503 td->td_state = TDS_RUNNING; 504 td->td_pri_class = PRI_TIMESHARE; 505 td->td_user_pri = PUSER; 506 td->td_base_user_pri = PUSER; 507 td->td_lend_user_pri = PRI_MAX; 508 td->td_priority = PVM; 509 td->td_base_pri = PVM; 510 td->td_oncpu = curcpu; 511 td->td_flags = TDF_INMEM; 512 td->td_pflags = TDP_KTHREAD; 513 td->td_cpuset = cpuset_thread0(); 514 td->td_domain.dr_policy = td->td_cpuset->cs_domain; 515 epoch_thread_init(td); 516 prison0_init(); 517 p->p_peers = 0; 518 p->p_leader = p; 519 p->p_reaper = p; 520 p->p_treeflag |= P_TREE_REAPER; 521 LIST_INIT(&p->p_reaplist); 522 523 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 524 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 525 526 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 527 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 528 callout_init(&td->td_slpcallout, 1); 529 530 /* Create credentials. */ 531 newcred = crget(); 532 newcred->cr_ngroups = 1; /* group 0 */ 533 /* A hack to prevent uifind from tripping over NULL pointers. */ 534 curthread->td_ucred = newcred; 535 tmpuinfo.ui_uid = 1; 536 newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo; 537 newcred->cr_uidinfo = uifind(0); 538 newcred->cr_ruidinfo = uifind(0); 539 newcred->cr_loginclass = &tmplc; 540 newcred->cr_loginclass = loginclass_find("default"); 541 /* End hack. creds get properly set later with thread_cow_get_proc */ 542 curthread->td_ucred = NULL; 543 newcred->cr_prison = &prison0; 544 proc_set_cred_init(p, newcred); 545 #ifdef AUDIT 546 audit_cred_kproc0(newcred); 547 #endif 548 #ifdef MAC 549 mac_cred_create_swapper(newcred); 550 #endif 551 /* Create sigacts. */ 552 p->p_sigacts = sigacts_alloc(); 553 554 /* Initialize signal state for process 0. */ 555 siginit(&proc0); 556 557 /* Create the file descriptor table. */ 558 p->p_fd = fdinit(NULL, false); 559 p->p_fdtol = NULL; 560 561 /* Create the limits structures. */ 562 p->p_limit = lim_alloc(); 563 for (i = 0; i < RLIM_NLIMITS; i++) 564 p->p_limit->pl_rlimit[i].rlim_cur = 565 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 566 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 567 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 568 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 569 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 570 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 571 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 572 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 573 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 574 /* Cast to avoid overflow on i386/PAE. */ 575 pageablemem = ptoa((vm_paddr_t)vm_free_count()); 576 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 577 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 578 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 579 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 580 p->p_cpulimit = RLIM_INFINITY; 581 582 PROC_LOCK(p); 583 thread_cow_get_proc(td, p); 584 PROC_UNLOCK(p); 585 586 /* Initialize resource accounting structures. */ 587 racct_create(&p->p_racct); 588 589 p->p_stats = pstats_alloc(); 590 591 /* Allocate a prototype map so we have something to fork. */ 592 p->p_vmspace = &vmspace0; 593 vmspace0.vm_refcnt = 1; 594 pmap_pinit0(vmspace_pmap(&vmspace0)); 595 596 /* 597 * proc0 is not expected to enter usermode, so there is no special 598 * handling for sv_minuser here, like is done for exec_new_vmspace(). 599 */ 600 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 601 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 602 603 /* 604 * Call the init and ctor for the new thread and proc. We wait 605 * to do this until all other structures are fairly sane. 606 */ 607 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 608 EVENTHANDLER_DIRECT_INVOKE(thread_init, td); 609 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 610 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 611 612 /* 613 * Charge root for one process. 614 */ 615 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 616 PROC_LOCK(p); 617 racct_add_force(p, RACCT_NPROC, 1); 618 PROC_UNLOCK(p); 619 } 620 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 621 622 /* ARGSUSED*/ 623 static void 624 proc0_post(void *dummy __unused) 625 { 626 struct timespec ts; 627 struct proc *p; 628 struct rusage ru; 629 struct thread *td; 630 631 /* 632 * Now we can look at the time, having had a chance to verify the 633 * time from the filesystem. Pretend that proc0 started now. 634 */ 635 sx_slock(&allproc_lock); 636 FOREACH_PROC_IN_SYSTEM(p) { 637 PROC_LOCK(p); 638 if (p->p_state == PRS_NEW) { 639 PROC_UNLOCK(p); 640 continue; 641 } 642 microuptime(&p->p_stats->p_start); 643 PROC_STATLOCK(p); 644 rufetch(p, &ru); /* Clears thread stats */ 645 p->p_rux.rux_runtime = 0; 646 p->p_rux.rux_uticks = 0; 647 p->p_rux.rux_sticks = 0; 648 p->p_rux.rux_iticks = 0; 649 PROC_STATUNLOCK(p); 650 FOREACH_THREAD_IN_PROC(p, td) { 651 td->td_runtime = 0; 652 } 653 PROC_UNLOCK(p); 654 } 655 sx_sunlock(&allproc_lock); 656 PCPU_SET(switchtime, cpu_ticks()); 657 PCPU_SET(switchticks, ticks); 658 659 /* 660 * Give the ``random'' number generator a thump. 661 */ 662 nanotime(&ts); 663 srandom(ts.tv_sec ^ ts.tv_nsec); 664 } 665 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 666 667 static void 668 random_init(void *dummy __unused) 669 { 670 671 /* 672 * After CPU has been started we have some randomness on most 673 * platforms via get_cyclecount(). For platforms that don't 674 * we will reseed random(9) in proc0_post() as well. 675 */ 676 srandom(get_cyclecount()); 677 } 678 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 679 680 /* 681 *************************************************************************** 682 **** 683 **** The following SYSINIT's and glue code should be moved to the 684 **** respective files on a per subsystem basis. 685 **** 686 *************************************************************************** 687 */ 688 689 /* 690 * List of paths to try when searching for "init". 691 */ 692 static char init_path[MAXPATHLEN] = 693 #ifdef INIT_PATH 694 __XSTRING(INIT_PATH); 695 #else 696 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init"; 697 #endif 698 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 699 "Path used to search the init process"); 700 701 /* 702 * Shutdown timeout of init(8). 703 * Unused within kernel, but used to control init(8), hence do not remove. 704 */ 705 #ifndef INIT_SHUTDOWN_TIMEOUT 706 #define INIT_SHUTDOWN_TIMEOUT 120 707 #endif 708 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 709 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 710 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). " 711 "Unused within kernel, but used to control init(8)"); 712 713 /* 714 * Start the initial user process; try exec'ing each pathname in init_path. 715 * The program is invoked with one argument containing the boot flags. 716 */ 717 static void 718 start_init(void *dummy) 719 { 720 struct image_args args; 721 int error; 722 char *var, *path; 723 char *free_init_path, *tmp_init_path; 724 struct thread *td; 725 struct proc *p; 726 struct vmspace *oldvmspace; 727 728 TSENTER(); /* Here so we don't overlap with mi_startup. */ 729 730 td = curthread; 731 p = td->td_proc; 732 733 vfs_mountroot(); 734 735 /* Wipe GELI passphrase from the environment. */ 736 kern_unsetenv("kern.geom.eli.passphrase"); 737 738 if ((var = kern_getenv("init_path")) != NULL) { 739 strlcpy(init_path, var, sizeof(init_path)); 740 freeenv(var); 741 } 742 free_init_path = tmp_init_path = strdup(init_path, M_TEMP); 743 744 while ((path = strsep(&tmp_init_path, ":")) != NULL) { 745 if (bootverbose) 746 printf("start_init: trying %s\n", path); 747 748 memset(&args, 0, sizeof(args)); 749 error = exec_alloc_args(&args); 750 if (error != 0) 751 panic("%s: Can't allocate space for init arguments %d", 752 __func__, error); 753 754 error = exec_args_add_fname(&args, path, UIO_SYSSPACE); 755 if (error != 0) 756 panic("%s: Can't add fname %d", __func__, error); 757 error = exec_args_add_arg(&args, path, UIO_SYSSPACE); 758 if (error != 0) 759 panic("%s: Can't add argv[0] %d", __func__, error); 760 if (boothowto & RB_SINGLE) 761 error = exec_args_add_arg(&args, "-s", UIO_SYSSPACE); 762 if (error != 0) 763 panic("%s: Can't add argv[0] %d", __func__, error); 764 765 /* 766 * Now try to exec the program. If can't for any reason 767 * other than it doesn't exist, complain. 768 * 769 * Otherwise, return via fork_trampoline() all the way 770 * to user mode as init! 771 */ 772 KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0, 773 ("nested execve")); 774 oldvmspace = td->td_proc->p_vmspace; 775 error = kern_execve(td, &args, NULL); 776 KASSERT(error != 0, 777 ("kern_execve returned success, not EJUSTRETURN")); 778 if (error == EJUSTRETURN) { 779 if ((td->td_pflags & TDP_EXECVMSPC) != 0) { 780 KASSERT(p->p_vmspace != oldvmspace, 781 ("oldvmspace still used")); 782 vmspace_free(oldvmspace); 783 td->td_pflags &= ~TDP_EXECVMSPC; 784 } 785 free(free_init_path, M_TEMP); 786 TSEXIT(); 787 return; 788 } 789 if (error != ENOENT) 790 printf("exec %s: error %d\n", path, error); 791 } 792 free(free_init_path, M_TEMP); 793 printf("init: not found in path %s\n", init_path); 794 panic("no init"); 795 } 796 797 /* 798 * Like kproc_create(), but runs in its own address space. 799 * We do this early to reserve pid 1. 800 * 801 * Note special case - do not make it runnable yet. Other work 802 * in progress will change this more. 803 */ 804 static void 805 create_init(const void *udata __unused) 806 { 807 struct fork_req fr; 808 struct ucred *newcred, *oldcred; 809 struct thread *td; 810 int error; 811 812 bzero(&fr, sizeof(fr)); 813 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 814 fr.fr_procp = &initproc; 815 error = fork1(&thread0, &fr); 816 if (error) 817 panic("cannot fork init: %d\n", error); 818 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 819 /* divorce init's credentials from the kernel's */ 820 newcred = crget(); 821 sx_xlock(&proctree_lock); 822 PROC_LOCK(initproc); 823 initproc->p_flag |= P_SYSTEM | P_INMEM; 824 initproc->p_treeflag |= P_TREE_REAPER; 825 oldcred = initproc->p_ucred; 826 crcopy(newcred, oldcred); 827 #ifdef MAC 828 mac_cred_create_init(newcred); 829 #endif 830 #ifdef AUDIT 831 audit_cred_proc1(newcred); 832 #endif 833 proc_set_cred(initproc, newcred); 834 td = FIRST_THREAD_IN_PROC(initproc); 835 crfree(td->td_ucred); 836 td->td_ucred = crhold(initproc->p_ucred); 837 PROC_UNLOCK(initproc); 838 sx_xunlock(&proctree_lock); 839 crfree(oldcred); 840 cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc), 841 start_init, NULL); 842 } 843 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 844 845 /* 846 * Make it runnable now. 847 */ 848 static void 849 kick_init(const void *udata __unused) 850 { 851 struct thread *td; 852 853 td = FIRST_THREAD_IN_PROC(initproc); 854 thread_lock(td); 855 TD_SET_CAN_RUN(td); 856 sched_add(td, SRQ_BORING); 857 thread_unlock(td); 858 } 859 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL); 860