1 /*- 2 * Copyright (c) 1995 Terrence R. Lambert 3 * All rights reserved. 4 * 5 * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. All advertising materials mentioning features or use of this software 22 * must display the following acknowledgement: 23 * This product includes software developed by the University of 24 * California, Berkeley and its contributors. 25 * 4. Neither the name of the University nor the names of its contributors 26 * may be used to endorse or promote products derived from this software 27 * without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39 * SUCH DAMAGE. 40 * 41 * @(#)init_main.c 8.9 (Berkeley) 1/21/94 42 */ 43 44 #include <sys/cdefs.h> 45 __FBSDID("$FreeBSD$"); 46 47 #include "opt_ddb.h" 48 #include "opt_init_path.h" 49 #include "opt_verbose_sysinit.h" 50 51 #include <sys/param.h> 52 #include <sys/kernel.h> 53 #include <sys/exec.h> 54 #include <sys/file.h> 55 #include <sys/filedesc.h> 56 #include <sys/jail.h> 57 #include <sys/ktr.h> 58 #include <sys/lock.h> 59 #include <sys/loginclass.h> 60 #include <sys/mount.h> 61 #include <sys/mutex.h> 62 #include <sys/syscallsubr.h> 63 #include <sys/sysctl.h> 64 #include <sys/proc.h> 65 #include <sys/racct.h> 66 #include <sys/resourcevar.h> 67 #include <sys/systm.h> 68 #include <sys/signalvar.h> 69 #include <sys/vnode.h> 70 #include <sys/sysent.h> 71 #include <sys/reboot.h> 72 #include <sys/sched.h> 73 #include <sys/sx.h> 74 #include <sys/sysproto.h> 75 #include <sys/vmmeter.h> 76 #include <sys/unistd.h> 77 #include <sys/malloc.h> 78 #include <sys/conf.h> 79 #include <sys/cpuset.h> 80 81 #include <machine/cpu.h> 82 83 #include <security/audit/audit.h> 84 #include <security/mac/mac_framework.h> 85 86 #include <vm/vm.h> 87 #include <vm/vm_param.h> 88 #include <vm/pmap.h> 89 #include <vm/vm_map.h> 90 #include <vm/vm_domain.h> 91 #include <sys/copyright.h> 92 93 #include <ddb/ddb.h> 94 #include <ddb/db_sym.h> 95 96 void mi_startup(void); /* Should be elsewhere */ 97 98 /* Components of the first process -- never freed. */ 99 static struct session session0; 100 static struct pgrp pgrp0; 101 struct proc proc0; 102 struct thread thread0 __aligned(16); 103 struct vmspace vmspace0; 104 struct proc *initproc; 105 106 #ifndef BOOTHOWTO 107 #define BOOTHOWTO 0 108 #endif 109 int boothowto = BOOTHOWTO; /* initialized so that it can be patched */ 110 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, 111 "Boot control flags, passed from loader"); 112 113 #ifndef BOOTVERBOSE 114 #define BOOTVERBOSE 0 115 #endif 116 int bootverbose = BOOTVERBOSE; 117 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0, 118 "Control the output of verbose kernel messages"); 119 120 /* Want to avoid defining INVARIANTS if not already defined */ 121 #ifdef INVARIANTS 122 static int invariants = 1; 123 #else 124 static int invariants = 0; 125 #endif 126 SYSCTL_INT(_debug, OID_AUTO, invariants, CTLFLAG_RD, &invariants, 0, 127 "Kernel compiled with invariants"); 128 129 /* 130 * This ensures that there is at least one entry so that the sysinit_set 131 * symbol is not undefined. A sybsystem ID of SI_SUB_DUMMY is never 132 * executed. 133 */ 134 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL); 135 136 /* 137 * The sysinit table itself. Items are checked off as the are run. 138 * If we want to register new sysinit types, add them to newsysinit. 139 */ 140 SET_DECLARE(sysinit_set, struct sysinit); 141 struct sysinit **sysinit, **sysinit_end; 142 struct sysinit **newsysinit, **newsysinit_end; 143 144 /* 145 * Merge a new sysinit set into the current set, reallocating it if 146 * necessary. This can only be called after malloc is running. 147 */ 148 void 149 sysinit_add(struct sysinit **set, struct sysinit **set_end) 150 { 151 struct sysinit **newset; 152 struct sysinit **sipp; 153 struct sysinit **xipp; 154 int count; 155 156 count = set_end - set; 157 if (newsysinit) 158 count += newsysinit_end - newsysinit; 159 else 160 count += sysinit_end - sysinit; 161 newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT); 162 if (newset == NULL) 163 panic("cannot malloc for sysinit"); 164 xipp = newset; 165 if (newsysinit) 166 for (sipp = newsysinit; sipp < newsysinit_end; sipp++) 167 *xipp++ = *sipp; 168 else 169 for (sipp = sysinit; sipp < sysinit_end; sipp++) 170 *xipp++ = *sipp; 171 for (sipp = set; sipp < set_end; sipp++) 172 *xipp++ = *sipp; 173 if (newsysinit) 174 free(newsysinit, M_TEMP); 175 newsysinit = newset; 176 newsysinit_end = newset + count; 177 } 178 179 #if defined (DDB) && defined(VERBOSE_SYSINIT) 180 static const char * 181 symbol_name(vm_offset_t va, db_strategy_t strategy) 182 { 183 const char *name; 184 c_db_sym_t sym; 185 db_expr_t offset; 186 187 if (va == 0) 188 return (NULL); 189 sym = db_search_symbol(va, strategy, &offset); 190 if (offset != 0) 191 return (NULL); 192 db_symbol_values(sym, &name, NULL); 193 return (name); 194 } 195 #endif 196 197 /* 198 * System startup; initialize the world, create process 0, mount root 199 * filesystem, and fork to create init and pagedaemon. Most of the 200 * hard work is done in the lower-level initialization routines including 201 * startup(), which does memory initialization and autoconfiguration. 202 * 203 * This allows simple addition of new kernel subsystems that require 204 * boot time initialization. It also allows substitution of subsystem 205 * (for instance, a scheduler, kernel profiler, or VM system) by object 206 * module. Finally, it allows for optional "kernel threads". 207 */ 208 void 209 mi_startup(void) 210 { 211 212 register struct sysinit **sipp; /* system initialization*/ 213 register struct sysinit **xipp; /* interior loop of sort*/ 214 register struct sysinit *save; /* bubble*/ 215 216 #if defined(VERBOSE_SYSINIT) 217 int last; 218 int verbose; 219 #endif 220 221 if (boothowto & RB_VERBOSE) 222 bootverbose++; 223 224 if (sysinit == NULL) { 225 sysinit = SET_BEGIN(sysinit_set); 226 sysinit_end = SET_LIMIT(sysinit_set); 227 } 228 229 restart: 230 /* 231 * Perform a bubble sort of the system initialization objects by 232 * their subsystem (primary key) and order (secondary key). 233 */ 234 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 235 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) { 236 if ((*sipp)->subsystem < (*xipp)->subsystem || 237 ((*sipp)->subsystem == (*xipp)->subsystem && 238 (*sipp)->order <= (*xipp)->order)) 239 continue; /* skip*/ 240 save = *sipp; 241 *sipp = *xipp; 242 *xipp = save; 243 } 244 } 245 246 #if defined(VERBOSE_SYSINIT) 247 last = SI_SUB_COPYRIGHT; 248 verbose = 0; 249 #if !defined(DDB) 250 printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n"); 251 #endif 252 #endif 253 254 /* 255 * Traverse the (now) ordered list of system initialization tasks. 256 * Perform each task, and continue on to the next task. 257 */ 258 for (sipp = sysinit; sipp < sysinit_end; sipp++) { 259 260 if ((*sipp)->subsystem == SI_SUB_DUMMY) 261 continue; /* skip dummy task(s)*/ 262 263 if ((*sipp)->subsystem == SI_SUB_DONE) 264 continue; 265 266 #if defined(VERBOSE_SYSINIT) 267 if ((*sipp)->subsystem > last) { 268 verbose = 1; 269 last = (*sipp)->subsystem; 270 printf("subsystem %x\n", last); 271 } 272 if (verbose) { 273 #if defined(DDB) 274 const char *func, *data; 275 276 func = symbol_name((vm_offset_t)(*sipp)->func, 277 DB_STGY_PROC); 278 data = symbol_name((vm_offset_t)(*sipp)->udata, 279 DB_STGY_ANY); 280 if (func != NULL && data != NULL) 281 printf(" %s(&%s)... ", func, data); 282 else if (func != NULL) 283 printf(" %s(%p)... ", func, (*sipp)->udata); 284 else 285 #endif 286 printf(" %p(%p)... ", (*sipp)->func, 287 (*sipp)->udata); 288 } 289 #endif 290 291 /* Call function */ 292 (*((*sipp)->func))((*sipp)->udata); 293 294 #if defined(VERBOSE_SYSINIT) 295 if (verbose) 296 printf("done.\n"); 297 #endif 298 299 /* Check off the one we're just done */ 300 (*sipp)->subsystem = SI_SUB_DONE; 301 302 /* Check if we've installed more sysinit items via KLD */ 303 if (newsysinit != NULL) { 304 if (sysinit != SET_BEGIN(sysinit_set)) 305 free(sysinit, M_TEMP); 306 sysinit = newsysinit; 307 sysinit_end = newsysinit_end; 308 newsysinit = NULL; 309 newsysinit_end = NULL; 310 goto restart; 311 } 312 } 313 314 mtx_assert(&Giant, MA_OWNED | MA_NOTRECURSED); 315 mtx_unlock(&Giant); 316 317 /* 318 * Now hand over this thread to swapper. 319 */ 320 swapper(); 321 /* NOTREACHED*/ 322 } 323 324 325 /* 326 *************************************************************************** 327 **** 328 **** The following SYSINIT's belong elsewhere, but have not yet 329 **** been moved. 330 **** 331 *************************************************************************** 332 */ 333 static void 334 print_caddr_t(void *data) 335 { 336 printf("%s", (char *)data); 337 } 338 339 static void 340 print_version(void *data __unused) 341 { 342 int len; 343 344 /* Strip a trailing newline from version. */ 345 len = strlen(version); 346 while (len > 0 && version[len - 1] == '\n') 347 len--; 348 printf("%.*s %s\n", len, version, machine); 349 printf("%s\n", compiler_version); 350 } 351 352 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, 353 copyright); 354 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t, 355 trademark); 356 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL); 357 358 #ifdef WITNESS 359 static char wit_warn[] = 360 "WARNING: WITNESS option enabled, expect reduced performance.\n"; 361 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1, 362 print_caddr_t, wit_warn); 363 SYSINIT(witwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 1, 364 print_caddr_t, wit_warn); 365 #endif 366 367 #ifdef DIAGNOSTIC 368 static char diag_warn[] = 369 "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n"; 370 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2, 371 print_caddr_t, diag_warn); 372 SYSINIT(diagwarn2, SI_SUB_LAST, SI_ORDER_THIRD + 2, 373 print_caddr_t, diag_warn); 374 #endif 375 376 static int 377 null_fetch_syscall_args(struct thread *td __unused, 378 struct syscall_args *sa __unused) 379 { 380 381 panic("null_fetch_syscall_args"); 382 } 383 384 static void 385 null_set_syscall_retval(struct thread *td __unused, int error __unused) 386 { 387 388 panic("null_set_syscall_retval"); 389 } 390 391 struct sysentvec null_sysvec = { 392 .sv_size = 0, 393 .sv_table = NULL, 394 .sv_mask = 0, 395 .sv_sigsize = 0, 396 .sv_sigtbl = NULL, 397 .sv_errsize = 0, 398 .sv_errtbl = NULL, 399 .sv_transtrap = NULL, 400 .sv_fixup = NULL, 401 .sv_sendsig = NULL, 402 .sv_sigcode = NULL, 403 .sv_szsigcode = NULL, 404 .sv_prepsyscall = NULL, 405 .sv_name = "null", 406 .sv_coredump = NULL, 407 .sv_imgact_try = NULL, 408 .sv_minsigstksz = 0, 409 .sv_pagesize = PAGE_SIZE, 410 .sv_minuser = VM_MIN_ADDRESS, 411 .sv_maxuser = VM_MAXUSER_ADDRESS, 412 .sv_usrstack = USRSTACK, 413 .sv_psstrings = PS_STRINGS, 414 .sv_stackprot = VM_PROT_ALL, 415 .sv_copyout_strings = NULL, 416 .sv_setregs = NULL, 417 .sv_fixlimit = NULL, 418 .sv_maxssiz = NULL, 419 .sv_flags = 0, 420 .sv_set_syscall_retval = null_set_syscall_retval, 421 .sv_fetch_syscall_args = null_fetch_syscall_args, 422 .sv_syscallnames = NULL, 423 .sv_schedtail = NULL, 424 .sv_thread_detach = NULL, 425 }; 426 427 /* 428 *************************************************************************** 429 **** 430 **** The two following SYSINIT's are proc0 specific glue code. I am not 431 **** convinced that they can not be safely combined, but their order of 432 **** operation has been maintained as the same as the original init_main.c 433 **** for right now. 434 **** 435 **** These probably belong in init_proc.c or kern_proc.c, since they 436 **** deal with proc0 (the fork template process). 437 **** 438 *************************************************************************** 439 */ 440 /* ARGSUSED*/ 441 static void 442 proc0_init(void *dummy __unused) 443 { 444 struct proc *p; 445 struct thread *td; 446 struct ucred *newcred; 447 vm_paddr_t pageablemem; 448 int i; 449 450 GIANT_REQUIRED; 451 p = &proc0; 452 td = &thread0; 453 454 /* 455 * Initialize magic number and osrel. 456 */ 457 p->p_magic = P_MAGIC; 458 p->p_osrel = osreldate; 459 460 /* 461 * Initialize thread and process structures. 462 */ 463 procinit(); /* set up proc zone */ 464 threadinit(); /* set up UMA zones */ 465 466 /* 467 * Initialise scheduler resources. 468 * Add scheduler specific parts to proc, thread as needed. 469 */ 470 schedinit(); /* scheduler gets its house in order */ 471 472 /* 473 * Create process 0 (the swapper). 474 */ 475 LIST_INSERT_HEAD(&allproc, p, p_list); 476 LIST_INSERT_HEAD(PIDHASH(0), p, p_hash); 477 mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); 478 p->p_pgrp = &pgrp0; 479 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 480 LIST_INIT(&pgrp0.pg_members); 481 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 482 483 pgrp0.pg_session = &session0; 484 mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF); 485 refcount_init(&session0.s_count, 1); 486 session0.s_leader = p; 487 488 p->p_sysent = &null_sysvec; 489 p->p_flag = P_SYSTEM | P_INMEM; 490 p->p_flag2 = 0; 491 p->p_state = PRS_NORMAL; 492 knlist_init_mtx(&p->p_klist, &p->p_mtx); 493 STAILQ_INIT(&p->p_ktr); 494 p->p_nice = NZERO; 495 /* pid_max cannot be greater than PID_MAX */ 496 td->td_tid = PID_MAX + 1; 497 LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash); 498 td->td_state = TDS_RUNNING; 499 td->td_pri_class = PRI_TIMESHARE; 500 td->td_user_pri = PUSER; 501 td->td_base_user_pri = PUSER; 502 td->td_lend_user_pri = PRI_MAX; 503 td->td_priority = PVM; 504 td->td_base_pri = PVM; 505 td->td_oncpu = 0; 506 td->td_flags = TDF_INMEM; 507 td->td_pflags = TDP_KTHREAD; 508 td->td_cpuset = cpuset_thread0(); 509 vm_domain_policy_init(&td->td_vm_dom_policy); 510 vm_domain_policy_set(&td->td_vm_dom_policy, VM_POLICY_NONE, -1); 511 vm_domain_policy_init(&p->p_vm_dom_policy); 512 vm_domain_policy_set(&p->p_vm_dom_policy, VM_POLICY_NONE, -1); 513 prison0_init(); 514 p->p_peers = 0; 515 p->p_leader = p; 516 p->p_reaper = p; 517 LIST_INIT(&p->p_reaplist); 518 519 strncpy(p->p_comm, "kernel", sizeof (p->p_comm)); 520 strncpy(td->td_name, "swapper", sizeof (td->td_name)); 521 522 callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0); 523 callout_init_mtx(&p->p_limco, &p->p_mtx, 0); 524 callout_init(&td->td_slpcallout, 1); 525 526 /* Create credentials. */ 527 newcred = crget(); 528 newcred->cr_ngroups = 1; /* group 0 */ 529 newcred->cr_uidinfo = uifind(0); 530 newcred->cr_ruidinfo = uifind(0); 531 newcred->cr_prison = &prison0; 532 newcred->cr_loginclass = loginclass_find("default"); 533 proc_set_cred_init(p, newcred); 534 #ifdef AUDIT 535 audit_cred_kproc0(newcred); 536 #endif 537 #ifdef MAC 538 mac_cred_create_swapper(newcred); 539 #endif 540 /* Create sigacts. */ 541 p->p_sigacts = sigacts_alloc(); 542 543 /* Initialize signal state for process 0. */ 544 siginit(&proc0); 545 546 /* Create the file descriptor table. */ 547 p->p_fd = fdinit(NULL, false); 548 p->p_fdtol = NULL; 549 550 /* Create the limits structures. */ 551 p->p_limit = lim_alloc(); 552 for (i = 0; i < RLIM_NLIMITS; i++) 553 p->p_limit->pl_rlimit[i].rlim_cur = 554 p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY; 555 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur = 556 p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 557 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur = 558 p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 559 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz; 560 p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz; 561 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz; 562 p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz; 563 /* Cast to avoid overflow on i386/PAE. */ 564 pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count); 565 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur = 566 p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem; 567 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3; 568 p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem; 569 p->p_cpulimit = RLIM_INFINITY; 570 571 PROC_LOCK(p); 572 thread_cow_get_proc(td, p); 573 PROC_UNLOCK(p); 574 575 /* Initialize resource accounting structures. */ 576 racct_create(&p->p_racct); 577 578 p->p_stats = pstats_alloc(); 579 580 /* Allocate a prototype map so we have something to fork. */ 581 p->p_vmspace = &vmspace0; 582 vmspace0.vm_refcnt = 1; 583 pmap_pinit0(vmspace_pmap(&vmspace0)); 584 585 /* 586 * proc0 is not expected to enter usermode, so there is no special 587 * handling for sv_minuser here, like is done for exec_new_vmspace(). 588 */ 589 vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0), 590 p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser); 591 592 /* 593 * Call the init and ctor for the new thread and proc. We wait 594 * to do this until all other structures are fairly sane. 595 */ 596 EVENTHANDLER_INVOKE(process_init, p); 597 EVENTHANDLER_INVOKE(thread_init, td); 598 EVENTHANDLER_INVOKE(process_ctor, p); 599 EVENTHANDLER_INVOKE(thread_ctor, td); 600 601 /* 602 * Charge root for one process. 603 */ 604 (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0); 605 PROC_LOCK(p); 606 racct_add_force(p, RACCT_NPROC, 1); 607 PROC_UNLOCK(p); 608 } 609 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL); 610 611 /* ARGSUSED*/ 612 static void 613 proc0_post(void *dummy __unused) 614 { 615 struct timespec ts; 616 struct proc *p; 617 struct rusage ru; 618 struct thread *td; 619 620 /* 621 * Now we can look at the time, having had a chance to verify the 622 * time from the filesystem. Pretend that proc0 started now. 623 */ 624 sx_slock(&allproc_lock); 625 FOREACH_PROC_IN_SYSTEM(p) { 626 microuptime(&p->p_stats->p_start); 627 PROC_STATLOCK(p); 628 rufetch(p, &ru); /* Clears thread stats */ 629 PROC_STATUNLOCK(p); 630 p->p_rux.rux_runtime = 0; 631 p->p_rux.rux_uticks = 0; 632 p->p_rux.rux_sticks = 0; 633 p->p_rux.rux_iticks = 0; 634 FOREACH_THREAD_IN_PROC(p, td) { 635 td->td_runtime = 0; 636 } 637 } 638 sx_sunlock(&allproc_lock); 639 PCPU_SET(switchtime, cpu_ticks()); 640 PCPU_SET(switchticks, ticks); 641 642 /* 643 * Give the ``random'' number generator a thump. 644 */ 645 nanotime(&ts); 646 srandom(ts.tv_sec ^ ts.tv_nsec); 647 } 648 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL); 649 650 static void 651 random_init(void *dummy __unused) 652 { 653 654 /* 655 * After CPU has been started we have some randomness on most 656 * platforms via get_cyclecount(). For platforms that don't 657 * we will reseed random(9) in proc0_post() as well. 658 */ 659 srandom(get_cyclecount()); 660 } 661 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL); 662 663 /* 664 *************************************************************************** 665 **** 666 **** The following SYSINIT's and glue code should be moved to the 667 **** respective files on a per subsystem basis. 668 **** 669 *************************************************************************** 670 */ 671 672 673 /* 674 *************************************************************************** 675 **** 676 **** The following code probably belongs in another file, like 677 **** kern/init_init.c. 678 **** 679 *************************************************************************** 680 */ 681 682 /* 683 * List of paths to try when searching for "init". 684 */ 685 static char init_path[MAXPATHLEN] = 686 #ifdef INIT_PATH 687 __XSTRING(INIT_PATH); 688 #else 689 "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init"; 690 #endif 691 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, 692 "Path used to search the init process"); 693 694 /* 695 * Shutdown timeout of init(8). 696 * Unused within kernel, but used to control init(8), hence do not remove. 697 */ 698 #ifndef INIT_SHUTDOWN_TIMEOUT 699 #define INIT_SHUTDOWN_TIMEOUT 120 700 #endif 701 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT; 702 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout, 703 CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). " 704 "Unused within kernel, but used to control init(8)"); 705 706 /* 707 * Start the initial user process; try exec'ing each pathname in init_path. 708 * The program is invoked with one argument containing the boot flags. 709 */ 710 static void 711 start_init(void *dummy) 712 { 713 vm_offset_t addr; 714 struct execve_args args; 715 int options, error; 716 char *var, *path, *next, *s; 717 char *ucp, **uap, *arg0, *arg1; 718 struct thread *td; 719 struct proc *p; 720 721 mtx_lock(&Giant); 722 723 GIANT_REQUIRED; 724 725 td = curthread; 726 p = td->td_proc; 727 728 vfs_mountroot(); 729 730 /* Wipe GELI passphrase from the environment. */ 731 kern_unsetenv("kern.geom.eli.passphrase"); 732 733 /* 734 * Need just enough stack to hold the faked-up "execve()" arguments. 735 */ 736 addr = p->p_sysent->sv_usrstack - PAGE_SIZE; 737 if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0, 738 VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0) 739 panic("init: couldn't allocate argument space"); 740 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 741 p->p_vmspace->vm_ssize = 1; 742 743 if ((var = kern_getenv("init_path")) != NULL) { 744 strlcpy(init_path, var, sizeof(init_path)); 745 freeenv(var); 746 } 747 748 for (path = init_path; *path != '\0'; path = next) { 749 while (*path == ':') 750 path++; 751 if (*path == '\0') 752 break; 753 for (next = path; *next != '\0' && *next != ':'; next++) 754 /* nothing */ ; 755 if (bootverbose) 756 printf("start_init: trying %.*s\n", (int)(next - path), 757 path); 758 759 /* 760 * Move out the boot flag argument. 761 */ 762 options = 0; 763 ucp = (char *)p->p_sysent->sv_usrstack; 764 (void)subyte(--ucp, 0); /* trailing zero */ 765 if (boothowto & RB_SINGLE) { 766 (void)subyte(--ucp, 's'); 767 options = 1; 768 } 769 #ifdef notyet 770 if (boothowto & RB_FASTBOOT) { 771 (void)subyte(--ucp, 'f'); 772 options = 1; 773 } 774 #endif 775 776 #ifdef BOOTCDROM 777 (void)subyte(--ucp, 'C'); 778 options = 1; 779 #endif 780 781 if (options == 0) 782 (void)subyte(--ucp, '-'); 783 (void)subyte(--ucp, '-'); /* leading hyphen */ 784 arg1 = ucp; 785 786 /* 787 * Move out the file name (also arg 0). 788 */ 789 (void)subyte(--ucp, 0); 790 for (s = next - 1; s >= path; s--) 791 (void)subyte(--ucp, *s); 792 arg0 = ucp; 793 794 /* 795 * Move out the arg pointers. 796 */ 797 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1)); 798 (void)suword((caddr_t)--uap, (long)0); /* terminator */ 799 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1); 800 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0); 801 802 /* 803 * Point at the arguments. 804 */ 805 args.fname = arg0; 806 args.argv = uap; 807 args.envv = NULL; 808 809 /* 810 * Now try to exec the program. If can't for any reason 811 * other than it doesn't exist, complain. 812 * 813 * Otherwise, return via fork_trampoline() all the way 814 * to user mode as init! 815 */ 816 if ((error = sys_execve(td, &args)) == 0) { 817 mtx_unlock(&Giant); 818 return; 819 } 820 if (error != ENOENT) 821 printf("exec %.*s: error %d\n", (int)(next - path), 822 path, error); 823 } 824 printf("init: not found in path %s\n", init_path); 825 panic("no init"); 826 } 827 828 /* 829 * Like kproc_create(), but runs in it's own address space. 830 * We do this early to reserve pid 1. 831 * 832 * Note special case - do not make it runnable yet. Other work 833 * in progress will change this more. 834 */ 835 static void 836 create_init(const void *udata __unused) 837 { 838 struct ucred *newcred, *oldcred; 839 struct thread *td; 840 int error; 841 842 error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc, 843 NULL, 0, NULL); 844 if (error) 845 panic("cannot fork init: %d\n", error); 846 KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1")); 847 /* divorce init's credentials from the kernel's */ 848 newcred = crget(); 849 sx_xlock(&proctree_lock); 850 PROC_LOCK(initproc); 851 initproc->p_flag |= P_SYSTEM | P_INMEM; 852 initproc->p_treeflag |= P_TREE_REAPER; 853 LIST_INSERT_HEAD(&initproc->p_reaplist, &proc0, p_reapsibling); 854 oldcred = initproc->p_ucred; 855 crcopy(newcred, oldcred); 856 #ifdef MAC 857 mac_cred_create_init(newcred); 858 #endif 859 #ifdef AUDIT 860 audit_cred_proc1(newcred); 861 #endif 862 proc_set_cred(initproc, newcred); 863 td = FIRST_THREAD_IN_PROC(initproc); 864 crfree(td->td_ucred); 865 td->td_ucred = crhold(initproc->p_ucred); 866 PROC_UNLOCK(initproc); 867 sx_xunlock(&proctree_lock); 868 crfree(oldcred); 869 cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL); 870 } 871 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL); 872 873 /* 874 * Make it runnable now. 875 */ 876 static void 877 kick_init(const void *udata __unused) 878 { 879 struct thread *td; 880 881 td = FIRST_THREAD_IN_PROC(initproc); 882 thread_lock(td); 883 TD_SET_CAN_RUN(td); 884 sched_add(td, SRQ_BORING); 885 thread_unlock(td); 886 } 887 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL); 888