sys/kern/init_main.c

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