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