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/exec.h> 56 #include <sys/file.h> 57 #include <sys/filedesc.h> 58 #include <sys/jail.h> 59 #include <sys/ktr.h> 60 #include <sys/lock.h> 61 #include <sys/loginclass.h> 62 #include <sys/mount.h> 63 #include <sys/mutex.h> 64 #include <sys/syscallsubr.h> 65 #include <sys/sysctl.h> 66 #include <sys/proc.h> 67 #include <sys/racct.h> 68 #include <sys/resourcevar.h> 69 #include <sys/systm.h> 70 #include <sys/signalvar.h> 71 #include <sys/vnode.h> 72 #include <sys/sysent.h> 73 #include <sys/reboot.h> 74 #include <sys/sched.h> 75 #include <sys/sx.h> 76 #include <sys/sysproto.h> 77 #include <sys/vmmeter.h> 78 #include <sys/unistd.h> 79 #include <sys/malloc.h> 80 #include <sys/conf.h> 81 #include <sys/cpuset.h> 82 83 #include <machine/cpu.h> 84 85 #include <security/audit/audit.h> 86 #include <security/mac/mac_framework.h> 87 88 #include <vm/vm.h> 89 #include <vm/vm_param.h> 90 #include <vm/pmap.h> 91 #include <vm/vm_map.h> 92 #include <sys/copyright.h> 93 94 #include <ddb/ddb.h> 95 #include <ddb/db_sym.h> 96 97 void mi_startup(void); /* Should be elsewhere */ 98 99 /* Components of the first process -- never freed. */ 100 static struct session session0; 101 static struct pgrp pgrp0; 102 struct proc proc0; 103 struct thread0_storage thread0_st __aligned(32); 104 struct vmspace vmspace0; 105 struct proc *initproc; 106 107 #ifndef BOOTHOWTO 108 #define BOOTHOWTO 0 109 #endif 110 int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 111 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, 112 "Boot control flags, passed from loader"); 113 114 #ifndef BOOTVERBOSE 115 #define BOOTVERBOSE 0 116 #endif 117 int bootverbose = BOOTVERBOSE; 118 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 119 "Control the output of verbose kernel messages"); 120 121 #ifdef INVARIANTS 122 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance"); 123 #endif 124 125 /* 126 * This ensures that there is at least one entry so that the sysinit_set 127 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 128 * executed. 129 */ 130 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 131 132 /* 133 * The sysinit table itself. Items are checked off as the are run. 134 * If we want to register new sysinit types, add them to newsysinit. 135 */ 136 SET_DECLARE(sysinit_set, struct sysinit); 137 struct sysinit **sysinit, **sysinit_end; 138 struct sysinit **newsysinit, **newsysinit_end; 139 140 EVENTHANDLER_LIST_DECLARE(process_init); 141 EVENTHANDLER_LIST_DECLARE(thread_init); 142 EVENTHANDLER_LIST_DECLARE(process_ctor); 143 EVENTHANDLER_LIST_DECLARE(thread_ctor); 144 145 /* 146 * Merge a new sysinit set into the current set, reallocating it if 147 * necessary. This can only be called after malloc is running. 148 */ 149 void 150 sysinit_add(struct sysinit **set, struct sysinit **set_end) 151 { 152 struct sysinit **newset; 153 struct sysinit **sipp; 154 struct sysinit **xipp; 155 int count; 156 157 count = set_end - set; 158 if (newsysinit) 159 count += newsysinit_end - newsysinit; 160 else 161 count += sysinit_end - sysinit; 162 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 163 if (newset == NULL) 164 panic("cannot malloc for sysinit"); 165 xipp = newset; 166 if (newsysinit) 167 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 168 *xipp++ = *sipp; 169 else 170 for (sipp = sysinit; sipp < sysinit_end; sipp++) 171 *xipp++ = *sipp; 172 for (sipp = set; sipp < set_end; sipp++) 173 *xipp++ = *sipp; 174 if (newsysinit) 175 free(newsysinit, M_TEMP); 176 newsysinit = newset; 177 newsysinit_end = newset + count; 178 } 179 180 #if defined (DDB) && defined(VERBOSE_SYSINIT) 181 static const char * 182 symbol_name(vm_offset_t va, db_strategy_t strategy) 183 { 184 const char *name; 185 c_db_sym_t sym; 186 db_expr_t offset; 187 188 if (va == 0) 189 return (NULL); 190 sym = db_search_symbol(va, strategy, &offset); 191 if (offset != 0) 192 return (NULL); 193 db_symbol_values(sym, &name, NULL); 194 return (name); 195 } 196 #endif 197 198 /* 199 * System startup; initialize the world, create process 0, mount root 200 * filesystem, and fork to create init and pagedaemon. Most of the 201 * hard work is done in the lower-level initialization routines including 202 * startup(), which does memory initialization and autoconfiguration. 203 * 204 * This allows simple addition of new kernel subsystems that require 205 * boot time initialization. It also allows substitution of subsystem 206 * (for instance, a scheduler, kernel profiler, or VM system) by object 207 * module. Finally, it allows for optional "kernel threads". 208 */ 209 void 210 mi_startup(void) 211 { 212 213 struct sysinit **sipp; /* system initialization*/ 214 struct sysinit **xipp; /* interior loop of sort*/ 215 struct sysinit *save; /* bubble*/ 216 217 #if defined(VERBOSE_SYSINIT) 218 int last; 219 int verbose; 220 #endif 221 222 TSENTER(); 223 224 if (boothowto & RB_VERBOSE) 225 bootverbose++; 226 227 if (sysinit == NULL) { 228 sysinit = SET_BEGIN(sysinit_set); 229 sysinit_end = SET_LIMIT(sysinit_set); 230 } 231 232 restart: 233 /* 234 * Perform a bubble sort of the system initialization objects by 235 * their subsystem (primary key) and order (secondary key). 236 */ 237 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 238 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 239 if ((*sipp)->subsystem < (*xipp)->subsystem || 240 ((*sipp)->subsystem == (*xipp)->subsystem && 241 (*sipp)->order <= (*xipp)->order)) 242 continue; /* skip*/ 243 save = *sipp; 244 *sipp = *xipp; 245 *xipp = save; 246 } 247 } 248 249 #if defined(VERBOSE_SYSINIT) 250 last = SI_SUB_COPYRIGHT; 251 verbose = 0; 252 #if !defined(DDB) 253 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 254 #endif 255 #endif 256 257 /* 258 * Traverse the (now) ordered list of system initialization tasks. 259 * Perform each task, and continue on to the next task. 260 */ 261 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 262 263 if ((*sipp)->subsystem == SI_SUB_DUMMY) 264 continue; /* skip dummy task(s)*/ 265 266 if ((*sipp)->subsystem == SI_SUB_DONE) 267 continue; 268 269 #if defined(VERBOSE_SYSINIT) 270 if ((*sipp)->subsystem > last) { 271 verbose = 1; 272 last = (*sipp)->subsystem; 273 printf("subsystem %x\n", last); 274 } 275 if (verbose) { 276 #if defined(DDB) 277 const char *func, *data; 278 279 func = symbol_name((vm_offset_t)(*sipp)->func, 280 DB_STGY_PROC); 281 data = symbol_name((vm_offset_t)(*sipp)->udata, 282 DB_STGY_ANY); 283 if (func != NULL && data != NULL) 284 printf(" %s(&%s)... ", func, data); 285 else if (func != NULL) 286 printf(" %s(%p)... ", func, (*sipp)->udata); 287 else 288 #endif 289 printf(" %p(%p)... ", (*sipp)->func, 290 (*sipp)->udata); 291 } 292 #endif 293 294 /* Call function */ 295 (*((*sipp)->func))((*sipp)->udata); 296 297 #if defined(VERBOSE_SYSINIT) 298 if (verbose) 299 printf("done.\n"); 300 #endif 301 302 /* Check off the one we're just done */ 303 (*sipp)->subsystem = SI_SUB_DONE; 304 305 /* Check if we've installed more sysinit items via KLD */ 306 if (newsysinit != NULL) { 307 if (sysinit != SET_BEGIN(sysinit_set)) 308 free(sysinit, M_TEMP); 309 sysinit = newsysinit; 310 sysinit_end = newsysinit_end; 311 newsysinit = NULL; 312 newsysinit_end = NULL; 313 goto restart; 314 } 315 } 316 317 TSEXIT(); /* Here so we don't overlap with start_init. */ 318 319 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 320 mtx_unlock(&Giant); 321 322 /* 323 * Now hand over this thread to swapper. 324 */ 325 swapper(); 326 /* NOTREACHED*/ 327 } 328 329 static void 330 print_caddr_t(void *data) 331 { 332 printf("%s", (char *)data); 333 } 334 335 static void 336 print_version(void *data __unused) 337 { 338 int len; 339 340 /* Strip a trailing newline from version. */ 341 len = strlen(version); 342 while (len > 0 && version[len - 1] == '\n') 343 len--; 344 printf("%.*s %s\n", len, version, machine); 345 printf("%s\n", compiler_version); 346 } 347 348 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 349 copyright); 350 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 351 trademark); 352 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 353 354 #ifdef WITNESS 355 static char wit_warn[] = 356 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 357 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 358 print_caddr_t, wit_warn); 359 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1, 360 print_caddr_t, wit_warn); 361 #endif 362 363 #ifdef DIAGNOSTIC 364 static char diag_warn[] = 365 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 366 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 367 print_caddr_t, diag_warn); 368 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2, 369 print_caddr_t, diag_warn); 370 #endif 371 372 static int 373 null_fetch_syscall_args(struct thread *td __unused) 374 { 375 376 panic("null_fetch_syscall_args"); 377 } 378 379 static void 380 null_set_syscall_retval(struct thread *td __unused, int error __unused) 381 { 382 383 panic("null_set_syscall_retval"); 384 } 385 386 struct sysentvec null_sysvec = { 387 .sv_size = 0, 388 .sv_table = NULL, 389 .sv_mask = 0, 390 .sv_errsize = 0, 391 .sv_errtbl = NULL, 392 .sv_transtrap = NULL, 393 .sv_fixup = NULL, 394 .sv_sendsig = NULL, 395 .sv_sigcode = NULL, 396 .sv_szsigcode = NULL, 397 .sv_name = "null", 398 .sv_coredump = NULL, 399 .sv_imgact_try = NULL, 400 .sv_minsigstksz = 0, 401 .sv_pagesize = PAGE_SIZE, 402 .sv_minuser = VM_MIN_ADDRESS, 403 .sv_maxuser = VM_MAXUSER_ADDRESS, 404 .sv_usrstack = USRSTACK, 405 .sv_psstrings = PS_STRINGS, 406 .sv_stackprot = VM_PROT_ALL, 407 .sv_copyout_strings = NULL, 408 .sv_setregs = NULL, 409 .sv_fixlimit = NULL, 410 .sv_maxssiz = NULL, 411 .sv_flags = 0, 412 .sv_set_syscall_retval = null_set_syscall_retval, 413 .sv_fetch_syscall_args = null_fetch_syscall_args, 414 .sv_syscallnames = NULL, 415 .sv_schedtail = NULL, 416 .sv_thread_detach = NULL, 417 .sv_trap = NULL, 418 }; 419 420 /* 421 * The two following SYSINIT's are proc0 specific glue code. I am not 422 * convinced that they can not be safely combined, but their order of 423 * operation has been maintained as the same as the original init_main.c 424 * for right now. 425 */ 426 /* ARGSUSED*/ 427 static void 428 proc0_init(void *dummy __unused) 429 { 430 struct proc *p; 431 struct thread *td; 432 struct ucred *newcred; 433 struct uidinfo tmpuinfo; 434 struct loginclass tmplc = { 435 .lc_name = "", 436 }; 437 vm_paddr_t pageablemem; 438 int i; 439 440 GIANT_REQUIRED; 441 p = &proc0; 442 td = &thread0; 443 444 /* 445 * Initialize magic number and osrel. 446 */ 447 p->p_magic = P_MAGIC; 448 p->p_osrel = osreldate; 449 450 /* 451 * Initialize thread and process structures. 452 */ 453 procinit(); /* set up proc zone */ 454 threadinit(); /* set up UMA zones */ 455 456 /* 457 * Initialise scheduler resources. 458 * Add scheduler specific parts to proc, thread as needed. 459 */ 460 schedinit(); /* scheduler gets its house in order */ 461 462 /* 463 * Create process 0 (the swapper). 464 */ 465 LIST_INSERT_HEAD(&allproc, p, p_list); 466 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 467 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 468 p->p_pgrp = &pgrp0; 469 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 470 LIST_INIT(&pgrp0.pg_members); 471 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 472 473 pgrp0.pg_session = &session0; 474 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 475 refcount_init(&session0.s_count, 1); 476 session0.s_leader = p; 477 478 p->p_sysent = &null_sysvec; 479 p->p_flag = P_SYSTEM | P_INMEM | P_KPROC; 480 p->p_flag2 = 0; 481 p->p_state = PRS_NORMAL; 482 p->p_klist = knlist_alloc(&p->p_mtx); 483 STAILQ_INIT(&p->p_ktr); 484 p->p_nice = NZERO; 485 /* pid_max cannot be greater than PID_MAX */ 486 td->td_tid = PID_MAX + 1; 487 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash); 488 td->td_state = TDS_RUNNING; 489 td->td_pri_class = PRI_TIMESHARE; 490 td->td_user_pri = PUSER; 491 td->td_base_user_pri = PUSER; 492 td->td_lend_user_pri = PRI_MAX; 493 td->td_priority = PVM; 494 td->td_base_pri = PVM; 495 td->td_oncpu = curcpu; 496 td->td_flags = TDF_INMEM; 497 td->td_pflags = TDP_KTHREAD; 498 td->td_cpuset = cpuset_thread0(); 499 td->td_domain.dr_policy = td->td_cpuset->cs_domain; 500 prison0_init(); 501 p->p_peers = 0; 502 p->p_leader = p; 503 p->p_reaper = p; 504 LIST_INIT(&p->p_reaplist); 505 506 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 507 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 508 509 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 510 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 511 callout_init(&td->td_slpcallout, 1); 512 513 /* Create credentials. */ 514 newcred = crget(); 515 newcred->cr_ngroups = 1; /* group 0 */ 516 /* A hack to prevent uifind from tripping over NULL pointers. */ 517 curthread->td_ucred = newcred; 518 tmpuinfo.ui_uid = 1; 519 newcred->cr_uidinfo = newcred->cr_ruidinfo = &tmpuinfo; 520 newcred->cr_uidinfo = uifind(0); 521 newcred->cr_ruidinfo = uifind(0); 522 newcred->cr_loginclass = &tmplc; 523 newcred->cr_loginclass = loginclass_find("default"); 524 /* End hack. creds get properly set later with thread_cow_get_proc */ 525 curthread->td_ucred = NULL; 526 newcred->cr_prison = &prison0; 527 proc_set_cred_init(p, newcred); 528 #ifdef AUDIT 529 audit_cred_kproc0(newcred); 530 #endif 531 #ifdef MAC 532 mac_cred_create_swapper(newcred); 533 #endif 534 /* Create sigacts. */ 535 p->p_sigacts = sigacts_alloc(); 536 537 /* Initialize signal state for process 0. */ 538 siginit(&proc0); 539 540 /* Create the file descriptor table. */ 541 p->p_fd = fdinit(NULL, false); 542 p->p_fdtol = NULL; 543 544 /* Create the limits structures. */ 545 p->p_limit = lim_alloc(); 546 for (i = 0; i < RLIM_NLIMITS; i++) 547 p->p_limit->pl_rlimit[i].rlim_cur = 548 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 549 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 550 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 551 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 552 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 553 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 554 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 555 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 556 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 557 /* Cast to avoid overflow on i386/PAE. */ 558 pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count); 559 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 560 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 561 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 562 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 563 p->p_cpulimit = RLIM_INFINITY; 564 565 PROC_LOCK(p); 566 thread_cow_get_proc(td, p); 567 PROC_UNLOCK(p); 568 569 /* Initialize resource accounting structures. */ 570 racct_create(&p->p_racct); 571 572 p->p_stats = pstats_alloc(); 573 574 /* Allocate a prototype map so we have something to fork. */ 575 p->p_vmspace = &vmspace0; 576 vmspace0.vm_refcnt = 1; 577 pmap_pinit0(vmspace_pmap(&vmspace0)); 578 579 /* 580 * proc0 is not expected to enter usermode, so there is no special 581 * handling for sv_minuser here, like is done for exec_new_vmspace(). 582 */ 583 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 584 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 585 586 /* 587 * Call the init and ctor for the new thread and proc. We wait 588 * to do this until all other structures are fairly sane. 589 */ 590 EVENTHANDLER_DIRECT_INVOKE(process_init, p); 591 EVENTHANDLER_DIRECT_INVOKE(thread_init, td); 592 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p); 593 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td); 594 595 /* 596 * Charge root for one process. 597 */ 598 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 599 PROC_LOCK(p); 600 racct_add_force(p, RACCT_NPROC, 1); 601 PROC_UNLOCK(p); 602 } 603 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 604 605 /* ARGSUSED*/ 606 static void 607 proc0_post(void *dummy __unused) 608 { 609 struct timespec ts; 610 struct proc *p; 611 struct rusage ru; 612 struct thread *td; 613 614 /* 615 * Now we can look at the time, having had a chance to verify the 616 * time from the filesystem. Pretend that proc0 started now. 617 */ 618 sx_slock(&allproc_lock); 619 FOREACH_PROC_IN_SYSTEM(p) { 620 microuptime(&p->p_stats->p_start); 621 PROC_STATLOCK(p); 622 rufetch(p, &ru); /* Clears thread stats */ 623 PROC_STATUNLOCK(p); 624 p->p_rux.rux_runtime = 0; 625 p->p_rux.rux_uticks = 0; 626 p->p_rux.rux_sticks = 0; 627 p->p_rux.rux_iticks = 0; 628 FOREACH_THREAD_IN_PROC(p, td) { 629 td->td_runtime = 0; 630 } 631 } 632 sx_sunlock(&allproc_lock); 633 PCPU_SET(switchtime, cpu_ticks()); 634 PCPU_SET(switchticks, ticks); 635 636 /* 637 * Give the ``random'' number generator a thump. 638 */ 639 nanotime(&ts); 640 srandom(ts.tv_sec ^ ts.tv_nsec); 641 } 642 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 643 644 static void 645 random_init(void *dummy __unused) 646 { 647 648 /* 649 * After CPU has been started we have some randomness on most 650 * platforms via get_cyclecount(). For platforms that don't 651 * we will reseed random(9) in proc0_post() as well. 652 */ 653 srandom(get_cyclecount()); 654 } 655 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 656 657 /* 658 *************************************************************************** 659 **** 660 **** The following SYSINIT's and glue code should be moved to the 661 **** respective files on a per subsystem basis. 662 **** 663 *************************************************************************** 664 */ 665 666 /* 667 * List of paths to try when searching for "init". 668 */ 669 static char init_path[MAXPATHLEN] = 670 #ifdef INIT_PATH 671 __XSTRING(INIT_PATH); 672 #else 673 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init"; 674 #endif 675 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 676 "Path used to search the init process"); 677 678 /* 679 * Shutdown timeout of init(8). 680 * Unused within kernel, but used to control init(8), hence do not remove. 681 */ 682 #ifndef INIT_SHUTDOWN_TIMEOUT 683 #define INIT_SHUTDOWN_TIMEOUT 120 684 #endif 685 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 686 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 687 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). " 688 "Unused within kernel, but used to control init(8)"); 689 690 /* 691 * Start the initial user process; try exec'ing each pathname in init_path. 692 * The program is invoked with one argument containing the boot flags. 693 */ 694 static void 695 start_init(void *dummy) 696 { 697 vm_offset_t addr; 698 struct execve_args args; 699 int options, error; 700 char *var, *path, *next, *s; 701 char *ucp, **uap, *arg0, *arg1; 702 struct thread *td; 703 struct proc *p; 704 705 mtx_lock(&Giant); 706 707 GIANT_REQUIRED; 708 709 TSENTER(); /* Here so we don't overlap with mi_startup. */ 710 711 td = curthread; 712 p = td->td_proc; 713 714 vfs_mountroot(); 715 716 /* Wipe GELI passphrase from the environment. */ 717 kern_unsetenv("kern.geom.eli.passphrase"); 718 719 /* 720 * Need just enough stack to hold the faked-up "execve()" arguments. 721 */ 722 addr = p->p_sysent->sv_usrstack - PAGE_SIZE; 723 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0, 724 VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 725 panic("init: couldn't allocate argument space"); 726 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 727 p->p_vmspace->vm_ssize = 1; 728 729 if ((var = kern_getenv("init_path")) != NULL) { 730 strlcpy(init_path, var, sizeof(init_path)); 731 freeenv(var); 732 } 733 734 for (path = init_path; *path != '\0'; path = next) { 735 while (*path == ':') 736 path++; 737 if (*path == '\0') 738 break; 739 for (next = path; *next != '\0' && *next != ':'; next++) 740 /* nothing */ ; 741 if (bootverbose) 742 printf("start_init: trying %.*s\n", (int)(next - path), 743 path); 744 745 /* 746 * Move out the boot flag argument. 747 */ 748 options = 0; 749 ucp = (char *)p->p_sysent->sv_usrstack; 750 (void)subyte(--ucp, 0); /* trailing zero */ 751 if (boothowto & RB_SINGLE) { 752 (void)subyte(--ucp, 's'); 753 options = 1; 754 } 755 #ifdef notyet 756 if (boothowto & RB_FASTBOOT) { 757 (void)subyte(--ucp, 'f'); 758 options = 1; 759 } 760 #endif 761 762 #ifdef BOOTCDROM 763 (void)subyte(--ucp, 'C'); 764 options = 1; 765 #endif 766 767 if (options == 0) 768 (void)subyte(--ucp, '-'); 769 (void)subyte(--ucp, '-'); /* leading hyphen */ 770 arg1 = ucp; 771 772 /* 773 * Move out the file name (also arg 0). 774 */ 775 (void)subyte(--ucp, 0); 776 for (s = next - 1; s >= path; s--) 777 (void)subyte(--ucp, *s); 778 arg0 = ucp; 779 780 /* 781 * Move out the arg pointers. 782 */ 783 uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t)); 784 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 785 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 786 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 787 788 /* 789 * Point at the arguments. 790 */ 791 args.fname = arg0; 792 args.argv = uap; 793 args.envv = NULL; 794 795 /* 796 * Now try to exec the program. If can't for any reason 797 * other than it doesn't exist, complain. 798 * 799 * Otherwise, return via fork_trampoline() all the way 800 * to user mode as init! 801 */ 802 if ((error = sys_execve(td, &args)) == EJUSTRETURN) { 803 mtx_unlock(&Giant); 804 TSEXIT(); 805 return; 806 } 807 if (error != ENOENT) 808 printf("exec %.*s: error %d\n", (int)(next - path), 809 path, error); 810 } 811 printf("init: not found in path %s\n", init_path); 812 panic("no init"); 813 } 814 815 /* 816 * Like kproc_create(), but runs in its own address space. 817 * We do this early to reserve pid 1. 818 * 819 * Note special case - do not make it runnable yet. Other work 820 * in progress will change this more. 821 */ 822 static void 823 create_init(const void *udata __unused) 824 { 825 struct fork_req fr; 826 struct ucred *newcred, *oldcred; 827 struct thread *td; 828 int error; 829 830 bzero(&fr, sizeof(fr)); 831 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 832 fr.fr_procp = &initproc; 833 error = fork1(&thread0, &fr); 834 if (error) 835 panic("cannot fork init: %d\n", error); 836 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 837 /* divorce init's credentials from the kernel's */ 838 newcred = crget(); 839 sx_xlock(&proctree_lock); 840 PROC_LOCK(initproc); 841 initproc->p_flag |= P_SYSTEM | P_INMEM; 842 initproc->p_treeflag |= P_TREE_REAPER; 843 LIST_INSERT_HEAD(&initproc->p_reaplist, &proc0, p_reapsibling); 844 oldcred = initproc->p_ucred; 845 crcopy(newcred, oldcred); 846 #ifdef MAC 847 mac_cred_create_init(newcred); 848 #endif 849 #ifdef AUDIT 850 audit_cred_proc1(newcred); 851 #endif 852 proc_set_cred(initproc, newcred); 853 td = FIRST_THREAD_IN_PROC(initproc); 854 crfree(td->td_ucred); 855 td->td_ucred = crhold(initproc->p_ucred); 856 PROC_UNLOCK(initproc); 857 sx_xunlock(&proctree_lock); 858 crfree(oldcred); 859 cpu_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc), 860 start_init, NULL); 861 } 862 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 863 864 /* 865 * Make it runnable now. 866 */ 867 static void 868 kick_init(const void *udata __unused) 869 { 870 struct thread *td; 871 872 td = FIRST_THREAD_IN_PROC(initproc); 873 thread_lock(td); 874 TD_SET_CAN_RUN(td); 875 sched_add(td, SRQ_BORING); 876 thread_unlock(td); 877 } 878 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL); 879