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 <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 thread0_storage thread0_st __aligned(32);
 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 #ifdef INVARIANTS
 121 FEATURE(invariants, "Kernel compiled with INVARIANTS, may affect performance");
 122 #endif
 123
 124 /*
 125  * This ensures that there is at least one entry so that the sysinit_set
 126  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
 127  * executed.
 128  */
 129 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
 130
 131 /*
 132  * The sysinit table itself.  Items are checked off as the are run.
 133  * If we want to register new sysinit types, add them to newsysinit.
 134  */
 135 SET_DECLARE(sysinit_set, struct sysinit);
 136 struct sysinit **sysinit, **sysinit_end;
 137 struct sysinit **newsysinit, **newsysinit_end;
 138
 139 EVENTHANDLER_LIST_DECLARE(process_init);
 140 EVENTHANDLER_LIST_DECLARE(thread_init);
 141 EVENTHANDLER_LIST_DECLARE(process_ctor);
 142 EVENTHANDLER_LIST_DECLARE(thread_ctor);
 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 {
 379
 380         panic("null_fetch_syscall_args");
 381 }
 382
 383 static void
 384 null_set_syscall_retval(struct thread *td __unused, int error __unused)
 385 {
 386
 387         panic("null_set_syscall_retval");
 388 }
 389
 390 struct sysentvec null_sysvec = {
 391         .sv_size        = 0,
 392         .sv_table       = NULL,
 393         .sv_mask        = 0,
 394         .sv_errsize     = 0,
 395         .sv_errtbl      = NULL,
 396         .sv_transtrap   = NULL,
 397         .sv_fixup       = NULL,
 398         .sv_sendsig     = NULL,
 399         .sv_sigcode     = NULL,
 400         .sv_szsigcode   = NULL,
 401         .sv_name        = "null",
 402         .sv_coredump    = NULL,
 403         .sv_imgact_try  = NULL,
 404         .sv_minsigstksz = 0,
 405         .sv_pagesize    = PAGE_SIZE,
 406         .sv_minuser     = VM_MIN_ADDRESS,
 407         .sv_maxuser     = VM_MAXUSER_ADDRESS,
 408         .sv_usrstack    = USRSTACK,
 409         .sv_psstrings   = PS_STRINGS,
 410         .sv_stackprot   = VM_PROT_ALL,
 411         .sv_copyout_strings     = NULL,
 412         .sv_setregs     = NULL,
 413         .sv_fixlimit    = NULL,
 414         .sv_maxssiz     = NULL,
 415         .sv_flags       = 0,
 416         .sv_set_syscall_retval = null_set_syscall_retval,
 417         .sv_fetch_syscall_args = null_fetch_syscall_args,
 418         .sv_syscallnames = NULL,
 419         .sv_schedtail   = NULL,
 420         .sv_thread_detach = NULL,
 421         .sv_trap        = NULL,
 422 };
 423
 424 /*
 425  ***************************************************************************
 426  ****
 427  **** The two following SYSINIT's are proc0 specific glue code.  I am not
 428  **** convinced that they can not be safely combined, but their order of
 429  **** operation has been maintained as the same as the original init_main.c
 430  **** for right now.
 431  ****
 432  **** These probably belong in init_proc.c or kern_proc.c, since they
 433  **** deal with proc0 (the fork template process).
 434  ****
 435  ***************************************************************************
 436  */
 437 /* ARGSUSED*/
 438 static void
 439 proc0_init(void *dummy __unused)
 440 {
 441         struct proc *p;
 442         struct thread *td;
 443         struct ucred *newcred;
 444         vm_paddr_t pageablemem;
 445         int i;
 446
 447         GIANT_REQUIRED;
 448         p = &proc0;
 449         td = &thread0;
 450
 451         /*
 452          * Initialize magic number and osrel.
 453          */
 454         p->p_magic = P_MAGIC;
 455         p->p_osrel = osreldate;
 456
 457         /*
 458          * Initialize thread and process structures.
 459          */
 460         procinit();     /* set up proc zone */
 461         threadinit();   /* set up UMA zones */
 462
 463         /*
 464          * Initialise scheduler resources.
 465          * Add scheduler specific parts to proc, thread as needed.
 466          */
 467         schedinit();    /* scheduler gets its house in order */
 468
 469         /*
 470          * Create process 0 (the swapper).
 471          */
 472         LIST_INSERT_HEAD(&allproc, p, p_list);
 473         LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
 474         mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
 475         p->p_pgrp = &pgrp0;
 476         LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
 477         LIST_INIT(&pgrp0.pg_members);
 478         LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
 479
 480         pgrp0.pg_session = &session0;
 481         mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
 482         refcount_init(&session0.s_count, 1);
 483         session0.s_leader = p;
 484
 485         p->p_sysent = &null_sysvec;
 486         p->p_flag = P_SYSTEM | P_INMEM | P_KPROC;
 487         p->p_flag2 = 0;
 488         p->p_state = PRS_NORMAL;
 489         p->p_klist = knlist_alloc(&p->p_mtx);
 490         STAILQ_INIT(&p->p_ktr);
 491         p->p_nice = NZERO;
 492         /* pid_max cannot be greater than PID_MAX */
 493         td->td_tid = PID_MAX + 1;
 494         LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
 495         td->td_state = TDS_RUNNING;
 496         td->td_pri_class = PRI_TIMESHARE;
 497         td->td_user_pri = PUSER;
 498         td->td_base_user_pri = PUSER;
 499         td->td_lend_user_pri = PRI_MAX;
 500         td->td_priority = PVM;
 501         td->td_base_pri = PVM;
 502         td->td_oncpu = 0;
 503         td->td_flags = TDF_INMEM;
 504         td->td_pflags = TDP_KTHREAD;
 505         td->td_cpuset = cpuset_thread0();
 506         vm_domain_policy_init(&td->td_vm_dom_policy);
 507         vm_domain_policy_set(&td->td_vm_dom_policy, VM_POLICY_NONE, -1);
 508         vm_domain_policy_init(&p->p_vm_dom_policy);
 509         vm_domain_policy_set(&p->p_vm_dom_policy, VM_POLICY_NONE, -1);
 510         prison0_init();
 511         p->p_peers = 0;
 512         p->p_leader = p;
 513         p->p_reaper = p;
 514         LIST_INIT(&p->p_reaplist);
 515
 516         strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
 517         strncpy(td->td_name, "swapper", sizeof (td->td_name));
 518
 519         callout_init_mtx(&p->p_itcallout, &p->p_mtx, 0);
 520         callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
 521         callout_init(&td->td_slpcallout, 1);
 522
 523         /* Create credentials. */
 524         newcred = crget();
 525         newcred->cr_ngroups = 1;        /* group 0 */
 526         newcred->cr_uidinfo = uifind(0);
 527         newcred->cr_ruidinfo = uifind(0);
 528         newcred->cr_prison = &prison0;
 529         newcred->cr_loginclass = loginclass_find("default");
 530         proc_set_cred_init(p, newcred);
 531 #ifdef AUDIT
 532         audit_cred_kproc0(newcred);
 533 #endif
 534 #ifdef MAC
 535         mac_cred_create_swapper(newcred);
 536 #endif
 537         /* Create sigacts. */
 538         p->p_sigacts = sigacts_alloc();
 539
 540         /* Initialize signal state for process 0. */
 541         siginit(&proc0);
 542
 543         /* Create the file descriptor table. */
 544         p->p_fd = fdinit(NULL, false);
 545         p->p_fdtol = NULL;
 546
 547         /* Create the limits structures. */
 548         p->p_limit = lim_alloc();
 549         for (i = 0; i < RLIM_NLIMITS; i++)
 550                 p->p_limit->pl_rlimit[i].rlim_cur =
 551                     p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
 552         p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
 553             p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
 554         p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
 555             p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
 556         p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
 557         p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
 558         p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
 559         p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
 560         /* Cast to avoid overflow on i386/PAE. */
 561         pageablemem = ptoa((vm_paddr_t)vm_cnt.v_free_count);
 562         p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
 563             p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
 564         p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
 565         p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
 566         p->p_cpulimit = RLIM_INFINITY;
 567
 568         PROC_LOCK(p);
 569         thread_cow_get_proc(td, p);
 570         PROC_UNLOCK(p);
 571
 572         /* Initialize resource accounting structures. */
 573         racct_create(&p->p_racct);
 574
 575         p->p_stats = pstats_alloc();
 576
 577         /* Allocate a prototype map so we have something to fork. */
 578         p->p_vmspace = &vmspace0;
 579         vmspace0.vm_refcnt = 1;
 580         pmap_pinit0(vmspace_pmap(&vmspace0));
 581
 582         /*
 583          * proc0 is not expected to enter usermode, so there is no special
 584          * handling for sv_minuser here, like is done for exec_new_vmspace().
 585          */
 586         vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
 587             p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
 588
 589         /*
 590          * Call the init and ctor for the new thread and proc.  We wait
 591          * to do this until all other structures are fairly sane.
 592          */
 593         EVENTHANDLER_DIRECT_INVOKE(process_init, p);
 594         EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
 595         EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
 596         EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
 597
 598         /*
 599          * Charge root for one process.
 600          */
 601         (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
 602         PROC_LOCK(p);
 603         racct_add_force(p, RACCT_NPROC, 1);
 604         PROC_UNLOCK(p);
 605 }
 606 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
 607
 608 /* ARGSUSED*/
 609 static void
 610 proc0_post(void *dummy __unused)
 611 {
 612         struct timespec ts;
 613         struct proc *p;
 614         struct rusage ru;
 615         struct thread *td;
 616
 617         /*
 618          * Now we can look at the time, having had a chance to verify the
 619          * time from the filesystem.  Pretend that proc0 started now.
 620          */
 621         sx_slock(&allproc_lock);
 622         FOREACH_PROC_IN_SYSTEM(p) {
 623                 microuptime(&p->p_stats->p_start);
 624                 PROC_STATLOCK(p);
 625                 rufetch(p, &ru);        /* Clears thread stats */
 626                 PROC_STATUNLOCK(p);
 627                 p->p_rux.rux_runtime = 0;
 628                 p->p_rux.rux_uticks = 0;
 629                 p->p_rux.rux_sticks = 0;
 630                 p->p_rux.rux_iticks = 0;
 631                 FOREACH_THREAD_IN_PROC(p, td) {
 632                         td->td_runtime = 0;
 633                 }
 634         }
 635         sx_sunlock(&allproc_lock);
 636         PCPU_SET(switchtime, cpu_ticks());
 637         PCPU_SET(switchticks, ticks);
 638
 639         /*
 640          * Give the ``random'' number generator a thump.
 641          */
 642         nanotime(&ts);
 643         srandom(ts.tv_sec ^ ts.tv_nsec);
 644 }
 645 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
 646
 647 static void
 648 random_init(void *dummy __unused)
 649 {
 650
 651         /*
 652          * After CPU has been started we have some randomness on most
 653          * platforms via get_cyclecount().  For platforms that don't
 654          * we will reseed random(9) in proc0_post() as well.
 655          */
 656         srandom(get_cyclecount());
 657 }
 658 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
 659
 660 /*
 661  ***************************************************************************
 662  ****
 663  **** The following SYSINIT's and glue code should be moved to the
 664  **** respective files on a per subsystem basis.
 665  ****
 666  ***************************************************************************
 667  */
 668
 669
 670 /*
 671  ***************************************************************************
 672  ****
 673  **** The following code probably belongs in another file, like
 674  **** kern/init_init.c.
 675  ****
 676  ***************************************************************************
 677  */
 678
 679 /*
 680  * List of paths to try when searching for "init".
 681  */
 682 static char init_path[MAXPATHLEN] =
 683 #ifdef  INIT_PATH
 684     __XSTRING(INIT_PATH);
 685 #else
 686     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
 687 #endif
 688 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
 689         "Path used to search the init process");
 690
 691 /*
 692  * Shutdown timeout of init(8).
 693  * Unused within kernel, but used to control init(8), hence do not remove.
 694  */
 695 #ifndef INIT_SHUTDOWN_TIMEOUT
 696 #define INIT_SHUTDOWN_TIMEOUT 120
 697 #endif
 698 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
 699 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
 700         CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
 701         "Unused within kernel, but used to control init(8)");
 702
 703 /*
 704  * Start the initial user process; try exec'ing each pathname in init_path.
 705  * The program is invoked with one argument containing the boot flags.
 706  */
 707 static void
 708 start_init(void *dummy)
 709 {
 710         vm_offset_t addr;
 711         struct execve_args args;
 712         int options, error;
 713         char *var, *path, *next, *s;
 714         char *ucp, **uap, *arg0, *arg1;
 715         struct thread *td;
 716         struct proc *p;
 717
 718         mtx_lock(&Giant);
 719
 720         GIANT_REQUIRED;
 721
 722         td = curthread;
 723         p = td->td_proc;
 724
 725         vfs_mountroot();
 726
 727         /* Wipe GELI passphrase from the environment. */
 728         kern_unsetenv("kern.geom.eli.passphrase");
 729
 730         /*
 731          * Need just enough stack to hold the faked-up "execve()" arguments.
 732          */
 733         addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
 734         if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE, 0,
 735             VMFS_NO_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
 736                 panic("init: couldn't allocate argument space");
 737         p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
 738         p->p_vmspace->vm_ssize = 1;
 739
 740         if ((var = kern_getenv("init_path")) != NULL) {
 741                 strlcpy(init_path, var, sizeof(init_path));
 742                 freeenv(var);
 743         }
 744
 745         for (path = init_path; *path != '\0'; path = next) {
 746                 while (*path == ':')
 747                         path++;
 748                 if (*path == '\0')
 749                         break;
 750                 for (next = path; *next != '\0' && *next != ':'; next++)
 751                         /* nothing */ ;
 752                 if (bootverbose)
 753                         printf("start_init: trying %.*s\n", (int)(next - path),
 754                             path);
 755
 756                 /*
 757                  * Move out the boot flag argument.
 758                  */
 759                 options = 0;
 760                 ucp = (char *)p->p_sysent->sv_usrstack;
 761                 (void)subyte(--ucp, 0);         /* trailing zero */
 762                 if (boothowto & RB_SINGLE) {
 763                         (void)subyte(--ucp, 's');
 764                         options = 1;
 765                 }
 766 #ifdef notyet
 767                 if (boothowto & RB_FASTBOOT) {
 768                         (void)subyte(--ucp, 'f');
 769                         options = 1;
 770                 }
 771 #endif
 772
 773 #ifdef BOOTCDROM
 774                 (void)subyte(--ucp, 'C');
 775                 options = 1;
 776 #endif
 777
 778                 if (options == 0)
 779                         (void)subyte(--ucp, '-');
 780                 (void)subyte(--ucp, '-');               /* leading hyphen */
 781                 arg1 = ucp;
 782
 783                 /*
 784                  * Move out the file name (also arg 0).
 785                  */
 786                 (void)subyte(--ucp, 0);
 787                 for (s = next - 1; s >= path; s--)
 788                         (void)subyte(--ucp, *s);
 789                 arg0 = ucp;
 790
 791                 /*
 792                  * Move out the arg pointers.
 793                  */
 794                 uap = (char **)rounddown2((intptr_t)ucp, sizeof(intptr_t));
 795                 (void)suword((caddr_t)--uap, (long)0);  /* terminator */
 796                 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
 797                 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
 798
 799                 /*
 800                  * Point at the arguments.
 801                  */
 802                 args.fname = arg0;
 803                 args.argv = uap;
 804                 args.envv = NULL;
 805
 806                 /*
 807                  * Now try to exec the program.  If can't for any reason
 808                  * other than it doesn't exist, complain.
 809                  *
 810                  * Otherwise, return via fork_trampoline() all the way
 811                  * to user mode as init!
 812                  */
 813                 if ((error = sys_execve(td, &args)) == 0) {
 814                         mtx_unlock(&Giant);
 815                         return;
 816                 }
 817                 if (error != ENOENT)
 818                         printf("exec %.*s: error %d\n", (int)(next - path),
 819                             path, error);
 820         }
 821         printf("init: not found in path %s\n", init_path);
 822         panic("no init");
 823 }
 824
 825 /*
 826  * Like kproc_create(), but runs in it's own address space.
 827  * We do this early to reserve pid 1.
 828  *
 829  * Note special case - do not make it runnable yet.  Other work
 830  * in progress will change this more.
 831  */
 832 static void
 833 create_init(const void *udata __unused)
 834 {
 835         struct fork_req fr;
 836         struct ucred *newcred, *oldcred;
 837         struct thread *td;
 838         int error;
 839
 840         bzero(&fr, sizeof(fr));
 841         fr.fr_flags = RFFDG | RFPROC | RFSTOPPED;
 842         fr.fr_procp = &initproc;
 843         error = fork1(&thread0, &fr);
 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_fork_kthread_handler(FIRST_THREAD_IN_PROC(initproc),
 870             start_init, NULL);
 871 }
 872 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
 873
 874 /*
 875  * Make it runnable now.
 876  */
 877 static void
 878 kick_init(const void *udata __unused)
 879 {
 880         struct thread *td;
 881
 882         td = FIRST_THREAD_IN_PROC(initproc);
 883         thread_lock(td);
 884         TD_SET_CAN_RUN(td);
 885         sched_add(td, SRQ_BORING);
 886         thread_unlock(td);
 887 }
 888 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_MIDDLE, kick_init, NULL);