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 int     boothowto = 0;          /* initialized so that it can be patched */
 105 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0,
 106         "Boot control flags, passed from loader");
 107 int     bootverbose;
 108 SYSCTL_INT(_debug, OID_AUTO, bootverbose, CTLFLAG_RW, &bootverbose, 0,
 109         "Control the output of verbose kernel messages");
 110
 111 /*
 112  * This ensures that there is at least one entry so that the sysinit_set
 113  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
 114  * executed.
 115  */
 116 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL);
 117
 118 /*
 119  * The sysinit table itself.  Items are checked off as the are run.
 120  * If we want to register new sysinit types, add them to newsysinit.
 121  */
 122 SET_DECLARE(sysinit_set, struct sysinit);
 123 struct sysinit **sysinit, **sysinit_end;
 124 struct sysinit **newsysinit, **newsysinit_end;
 125
 126 /*
 127  * Merge a new sysinit set into the current set, reallocating it if
 128  * necessary.  This can only be called after malloc is running.
 129  */
 130 void
 131 sysinit_add(struct sysinit **set, struct sysinit **set_end)
 132 {
 133         struct sysinit **newset;
 134         struct sysinit **sipp;
 135         struct sysinit **xipp;
 136         int count;
 137
 138         count = set_end - set;
 139         if (newsysinit)
 140                 count += newsysinit_end - newsysinit;
 141         else
 142                 count += sysinit_end - sysinit;
 143         newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
 144         if (newset == NULL)
 145                 panic("cannot malloc for sysinit");
 146         xipp = newset;
 147         if (newsysinit)
 148                 for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
 149                         *xipp++ = *sipp;
 150         else
 151                 for (sipp = sysinit; sipp < sysinit_end; sipp++)
 152                         *xipp++ = *sipp;
 153         for (sipp = set; sipp < set_end; sipp++)
 154                 *xipp++ = *sipp;
 155         if (newsysinit)
 156                 free(newsysinit, M_TEMP);
 157         newsysinit = newset;
 158         newsysinit_end = newset + count;
 159 }
 160
 161 /*
 162  * System startup; initialize the world, create process 0, mount root
 163  * filesystem, and fork to create init and pagedaemon.  Most of the
 164  * hard work is done in the lower-level initialization routines including
 165  * startup(), which does memory initialization and autoconfiguration.
 166  *
 167  * This allows simple addition of new kernel subsystems that require
 168  * boot time initialization.  It also allows substitution of subsystem
 169  * (for instance, a scheduler, kernel profiler, or VM system) by object
 170  * module.  Finally, it allows for optional "kernel threads".
 171  */
 172 void
 173 mi_startup(void)
 174 {
 175
 176         register struct sysinit **sipp;         /* system initialization*/
 177         register struct sysinit **xipp;         /* interior loop of sort*/
 178         register struct sysinit *save;          /* bubble*/
 179
 180 #if defined(VERBOSE_SYSINIT)
 181         int last;
 182         int verbose;
 183 #endif
 184
 185         if (boothowto & RB_VERBOSE)
 186                 bootverbose++;
 187
 188         if (sysinit == NULL) {
 189                 sysinit = SET_BEGIN(sysinit_set);
 190                 sysinit_end = SET_LIMIT(sysinit_set);
 191         }
 192
 193 restart:
 194         /*
 195          * Perform a bubble sort of the system initialization objects by
 196          * their subsystem (primary key) and order (secondary key).
 197          */
 198         for (sipp = sysinit; sipp < sysinit_end; sipp++) {
 199                 for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
 200                         if ((*sipp)->subsystem < (*xipp)->subsystem ||
 201                              ((*sipp)->subsystem == (*xipp)->subsystem &&
 202                               (*sipp)->order <= (*xipp)->order))
 203                                 continue;       /* skip*/
 204                         save = *sipp;
 205                         *sipp = *xipp;
 206                         *xipp = save;
 207                 }
 208         }
 209
 210 #if defined(VERBOSE_SYSINIT)
 211         last = SI_SUB_COPYRIGHT;
 212         verbose = 0;
 213 #if !defined(DDB)
 214         printf("VERBOSE_SYSINIT: DDB not enabled, symbol lookups disabled.\n");
 215 #endif
 216 #endif
 217
 218         /*
 219          * Traverse the (now) ordered list of system initialization tasks.
 220          * Perform each task, and continue on to the next task.
 221          *
 222          * The last item on the list is expected to be the scheduler,
 223          * which will not return.
 224          */
 225         for (sipp = sysinit; sipp < sysinit_end; sipp++) {
 226
 227                 if ((*sipp)->subsystem == SI_SUB_DUMMY)
 228                         continue;       /* skip dummy task(s)*/
 229
 230                 if ((*sipp)->subsystem == SI_SUB_DONE)
 231                         continue;
 232
 233 #if defined(VERBOSE_SYSINIT)
 234                 if ((*sipp)->subsystem > last) {
 235                         verbose = 1;
 236                         last = (*sipp)->subsystem;
 237                         printf("subsystem %x\n", last);
 238                 }
 239                 if (verbose) {
 240 #if defined(DDB)
 241                         const char *name;
 242                         c_db_sym_t sym;
 243                         db_expr_t  offset;
 244
 245                         sym = db_search_symbol((vm_offset_t)(*sipp)->func,
 246                             DB_STGY_PROC, &offset);
 247                         db_symbol_values(sym, &name, NULL);
 248                         if (name != NULL)
 249                                 printf("   %s(%p)... ", name, (*sipp)->udata);
 250                         else
 251 #endif
 252                                 printf("   %p(%p)... ", (*sipp)->func,
 253                                     (*sipp)->udata);
 254                 }
 255 #endif
 256
 257                 /* Call function */
 258                 (*((*sipp)->func))((*sipp)->udata);
 259
 260 #if defined(VERBOSE_SYSINIT)
 261                 if (verbose)
 262                         printf("done.\n");
 263 #endif
 264
 265                 /* Check off the one we're just done */
 266                 (*sipp)->subsystem = SI_SUB_DONE;
 267
 268                 /* Check if we've installed more sysinit items via KLD */
 269                 if (newsysinit != NULL) {
 270                         if (sysinit != SET_BEGIN(sysinit_set))
 271                                 free(sysinit, M_TEMP);
 272                         sysinit = newsysinit;
 273                         sysinit_end = newsysinit_end;
 274                         newsysinit = NULL;
 275                         newsysinit_end = NULL;
 276                         goto restart;
 277                 }
 278         }
 279
 280         panic("Shouldn't get here!");
 281         /* NOTREACHED*/
 282 }
 283
 284
 285 /*
 286  ***************************************************************************
 287  ****
 288  **** The following SYSINIT's belong elsewhere, but have not yet
 289  **** been moved.
 290  ****
 291  ***************************************************************************
 292  */
 293 static void
 294 print_caddr_t(void *data)
 295 {
 296         printf("%s", (char *)data);
 297 }
 298
 299 static void
 300 print_version(void *data __unused)
 301 {
 302         int len;
 303
 304         /* Strip a trailing newline from version. */
 305         len = strlen(version);
 306         while (len > 0 && version[len - 1] == '\n')
 307                 len--;
 308         printf("%.*s %s\n", len, version, machine);
 309 }
 310
 311 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t,
 312     copyright);
 313 SYSINIT(trademark, SI_SUB_COPYRIGHT, SI_ORDER_SECOND, print_caddr_t,
 314     trademark);
 315 SYSINIT(version, SI_SUB_COPYRIGHT, SI_ORDER_THIRD, print_version, NULL);
 316
 317 #ifdef WITNESS
 318 static char wit_warn[] =
 319      "WARNING: WITNESS option enabled, expect reduced performance.\n";
 320 SYSINIT(witwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 1,
 321    print_caddr_t, wit_warn);
 322 SYSINIT(witwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 1,
 323    print_caddr_t, wit_warn);
 324 #endif
 325
 326 #ifdef DIAGNOSTIC
 327 static char diag_warn[] =
 328      "WARNING: DIAGNOSTIC option enabled, expect reduced performance.\n";
 329 SYSINIT(diagwarn, SI_SUB_COPYRIGHT, SI_ORDER_THIRD + 2,
 330     print_caddr_t, diag_warn);
 331 SYSINIT(diagwarn2, SI_SUB_RUN_SCHEDULER, SI_ORDER_THIRD + 2,
 332     print_caddr_t, diag_warn);
 333 #endif
 334
 335 static int
 336 null_fetch_syscall_args(struct thread *td __unused,
 337     struct syscall_args *sa __unused)
 338 {
 339
 340         panic("null_fetch_syscall_args");
 341 }
 342
 343 static void
 344 null_set_syscall_retval(struct thread *td __unused, int error __unused)
 345 {
 346
 347         panic("null_set_syscall_retval");
 348 }
 349
 350 struct sysentvec null_sysvec = {
 351         .sv_size        = 0,
 352         .sv_table       = NULL,
 353         .sv_mask        = 0,
 354         .sv_sigsize     = 0,
 355         .sv_sigtbl      = NULL,
 356         .sv_errsize     = 0,
 357         .sv_errtbl      = NULL,
 358         .sv_transtrap   = NULL,
 359         .sv_fixup       = NULL,
 360         .sv_sendsig     = NULL,
 361         .sv_sigcode     = NULL,
 362         .sv_szsigcode   = NULL,
 363         .sv_prepsyscall = NULL,
 364         .sv_name        = "null",
 365         .sv_coredump    = NULL,
 366         .sv_imgact_try  = NULL,
 367         .sv_minsigstksz = 0,
 368         .sv_pagesize    = PAGE_SIZE,
 369         .sv_minuser     = VM_MIN_ADDRESS,
 370         .sv_maxuser     = VM_MAXUSER_ADDRESS,
 371         .sv_usrstack    = USRSTACK,
 372         .sv_psstrings   = PS_STRINGS,
 373         .sv_stackprot   = VM_PROT_ALL,
 374         .sv_copyout_strings     = NULL,
 375         .sv_setregs     = NULL,
 376         .sv_fixlimit    = NULL,
 377         .sv_maxssiz     = NULL,
 378         .sv_flags       = 0,
 379         .sv_set_syscall_retval = null_set_syscall_retval,
 380         .sv_fetch_syscall_args = null_fetch_syscall_args,
 381         .sv_syscallnames = NULL,
 382         .sv_schedtail   = NULL,
 383 };
 384
 385 /*
 386  ***************************************************************************
 387  ****
 388  **** The two following SYSINIT's are proc0 specific glue code.  I am not
 389  **** convinced that they can not be safely combined, but their order of
 390  **** operation has been maintained as the same as the original init_main.c
 391  **** for right now.
 392  ****
 393  **** These probably belong in init_proc.c or kern_proc.c, since they
 394  **** deal with proc0 (the fork template process).
 395  ****
 396  ***************************************************************************
 397  */
 398 /* ARGSUSED*/
 399 static void
 400 proc0_init(void *dummy __unused)
 401 {
 402         struct proc *p;
 403         struct thread *td;
 404         vm_paddr_t pageablemem;
 405         int i;
 406
 407         GIANT_REQUIRED;
 408         p = &proc0;
 409         td = &thread0;
 410
 411         /*
 412          * Initialize magic number and osrel.
 413          */
 414         p->p_magic = P_MAGIC;
 415         p->p_osrel = osreldate;
 416
 417         /*
 418          * Initialize thread and process structures.
 419          */
 420         procinit();     /* set up proc zone */
 421         threadinit();   /* set up UMA zones */
 422
 423         /*
 424          * Initialise scheduler resources.
 425          * Add scheduler specific parts to proc, thread as needed.
 426          */
 427         schedinit();    /* scheduler gets its house in order */
 428         /*
 429          * Initialize sleep queue hash table
 430          */
 431         sleepinit();
 432
 433         /*
 434          * additional VM structures
 435          */
 436         vm_init2();
 437
 438         /*
 439          * Create process 0 (the swapper).
 440          */
 441         LIST_INSERT_HEAD(&allproc, p, p_list);
 442         LIST_INSERT_HEAD(PIDHASH(0), p, p_hash);
 443         mtx_init(&pgrp0.pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
 444         p->p_pgrp = &pgrp0;
 445         LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
 446         LIST_INIT(&pgrp0.pg_members);
 447         LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
 448
 449         pgrp0.pg_session = &session0;
 450         mtx_init(&session0.s_mtx, "session", NULL, MTX_DEF);
 451         refcount_init(&session0.s_count, 1);
 452         session0.s_leader = p;
 453
 454         p->p_sysent = &null_sysvec;
 455         p->p_flag = P_SYSTEM | P_INMEM;
 456         p->p_state = PRS_NORMAL;
 457         knlist_init_mtx(&p->p_klist, &p->p_mtx);
 458         STAILQ_INIT(&p->p_ktr);
 459         p->p_nice = NZERO;
 460         td->td_tid = PID_MAX + 1;
 461         LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
 462         td->td_state = TDS_RUNNING;
 463         td->td_pri_class = PRI_TIMESHARE;
 464         td->td_user_pri = PUSER;
 465         td->td_base_user_pri = PUSER;
 466         td->td_lend_user_pri = PRI_MAX;
 467         td->td_priority = PVM;
 468         td->td_base_pri = PVM;
 469         td->td_oncpu = 0;
 470         td->td_flags = TDF_INMEM;
 471         td->td_pflags = TDP_KTHREAD;
 472         td->td_cpuset = cpuset_thread0();
 473         prison0.pr_cpuset = cpuset_ref(td->td_cpuset);
 474         p->p_peers = 0;
 475         p->p_leader = p;
 476
 477
 478         strncpy(p->p_comm, "kernel", sizeof (p->p_comm));
 479         strncpy(td->td_name, "swapper", sizeof (td->td_name));
 480
 481         callout_init(&p->p_itcallout, CALLOUT_MPSAFE);
 482         callout_init_mtx(&p->p_limco, &p->p_mtx, 0);
 483         callout_init(&td->td_slpcallout, CALLOUT_MPSAFE);
 484
 485         /* Create credentials. */
 486         p->p_ucred = crget();
 487         p->p_ucred->cr_ngroups = 1;     /* group 0 */
 488         p->p_ucred->cr_uidinfo = uifind(0);
 489         p->p_ucred->cr_ruidinfo = uifind(0);
 490         p->p_ucred->cr_prison = &prison0;
 491         p->p_ucred->cr_loginclass = loginclass_find("default");
 492 #ifdef AUDIT
 493         audit_cred_kproc0(p->p_ucred);
 494 #endif
 495 #ifdef MAC
 496         mac_cred_create_swapper(p->p_ucred);
 497 #endif
 498         td->td_ucred = crhold(p->p_ucred);
 499
 500         /* Create sigacts. */
 501         p->p_sigacts = sigacts_alloc();
 502
 503         /* Initialize signal state for process 0. */
 504         siginit(&proc0);
 505
 506         /* Create the file descriptor table. */
 507         p->p_fd = fdinit(NULL);
 508         p->p_fdtol = NULL;
 509
 510         /* Create the limits structures. */
 511         p->p_limit = lim_alloc();
 512         for (i = 0; i < RLIM_NLIMITS; i++)
 513                 p->p_limit->pl_rlimit[i].rlim_cur =
 514                     p->p_limit->pl_rlimit[i].rlim_max = RLIM_INFINITY;
 515         p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur =
 516             p->p_limit->pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
 517         p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_cur =
 518             p->p_limit->pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
 519         p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_cur = dfldsiz;
 520         p->p_limit->pl_rlimit[RLIMIT_DATA].rlim_max = maxdsiz;
 521         p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur = dflssiz;
 522         p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_max = maxssiz;
 523         /* Cast to avoid overflow on i386/PAE. */
 524         pageablemem = ptoa((vm_paddr_t)cnt.v_free_count);
 525         p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_cur =
 526             p->p_limit->pl_rlimit[RLIMIT_RSS].rlim_max = pageablemem;
 527         p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = pageablemem / 3;
 528         p->p_limit->pl_rlimit[RLIMIT_MEMLOCK].rlim_max = pageablemem;
 529         p->p_cpulimit = RLIM_INFINITY;
 530
 531         /* Initialize resource accounting structures. */
 532         racct_create(&p->p_racct);
 533
 534         p->p_stats = pstats_alloc();
 535
 536         /* Allocate a prototype map so we have something to fork. */
 537         pmap_pinit0(vmspace_pmap(&vmspace0));
 538         p->p_vmspace = &vmspace0;
 539         vmspace0.vm_refcnt = 1;
 540
 541         /*
 542          * proc0 is not expected to enter usermode, so there is no special
 543          * handling for sv_minuser here, like is done for exec_new_vmspace().
 544          */
 545         vm_map_init(&vmspace0.vm_map, vmspace_pmap(&vmspace0),
 546             p->p_sysent->sv_minuser, p->p_sysent->sv_maxuser);
 547
 548         /*
 549          * Call the init and ctor for the new thread and proc.  We wait
 550          * to do this until all other structures are fairly sane.
 551          */
 552         EVENTHANDLER_INVOKE(process_init, p);
 553         EVENTHANDLER_INVOKE(thread_init, td);
 554         EVENTHANDLER_INVOKE(process_ctor, p);
 555         EVENTHANDLER_INVOKE(thread_ctor, td);
 556
 557         /*
 558          * Charge root for one process.
 559          */
 560         (void)chgproccnt(p->p_ucred->cr_ruidinfo, 1, 0);
 561         PROC_LOCK(p);
 562         racct_add_force(p, RACCT_NPROC, 1);
 563         PROC_UNLOCK(p);
 564 }
 565 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL);
 566
 567 /* ARGSUSED*/
 568 static void
 569 proc0_post(void *dummy __unused)
 570 {
 571         struct timespec ts;
 572         struct proc *p;
 573         struct rusage ru;
 574         struct thread *td;
 575
 576         /*
 577          * Now we can look at the time, having had a chance to verify the
 578          * time from the filesystem.  Pretend that proc0 started now.
 579          */
 580         sx_slock(&allproc_lock);
 581         FOREACH_PROC_IN_SYSTEM(p) {
 582                 microuptime(&p->p_stats->p_start);
 583                 PROC_SLOCK(p);
 584                 rufetch(p, &ru);        /* Clears thread stats */
 585                 PROC_SUNLOCK(p);
 586                 p->p_rux.rux_runtime = 0;
 587                 p->p_rux.rux_uticks = 0;
 588                 p->p_rux.rux_sticks = 0;
 589                 p->p_rux.rux_iticks = 0;
 590                 FOREACH_THREAD_IN_PROC(p, td) {
 591                         td->td_runtime = 0;
 592                 }
 593         }
 594         sx_sunlock(&allproc_lock);
 595         PCPU_SET(switchtime, cpu_ticks());
 596         PCPU_SET(switchticks, ticks);
 597
 598         /*
 599          * Give the ``random'' number generator a thump.
 600          */
 601         nanotime(&ts);
 602         srandom(ts.tv_sec ^ ts.tv_nsec);
 603 }
 604 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL);
 605
 606 static void
 607 random_init(void *dummy __unused)
 608 {
 609
 610         /*
 611          * After CPU has been started we have some randomness on most
 612          * platforms via get_cyclecount().  For platforms that don't
 613          * we will reseed random(9) in proc0_post() as well.
 614          */
 615         srandom(get_cyclecount());
 616 }
 617 SYSINIT(random, SI_SUB_RANDOM, SI_ORDER_FIRST, random_init, NULL);
 618
 619 /*
 620  ***************************************************************************
 621  ****
 622  **** The following SYSINIT's and glue code should be moved to the
 623  **** respective files on a per subsystem basis.
 624  ****
 625  ***************************************************************************
 626  */
 627
 628
 629 /*
 630  ***************************************************************************
 631  ****
 632  **** The following code probably belongs in another file, like
 633  **** kern/init_init.c.
 634  ****
 635  ***************************************************************************
 636  */
 637
 638 /*
 639  * List of paths to try when searching for "init".
 640  */
 641 static char init_path[MAXPATHLEN] =
 642 #ifdef  INIT_PATH
 643     __XSTRING(INIT_PATH);
 644 #else
 645     "/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init";
 646 #endif
 647 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0,
 648         "Path used to search the init process");
 649
 650 /*
 651  * Shutdown timeout of init(8).
 652  * Unused within kernel, but used to control init(8), hence do not remove.
 653  */
 654 #ifndef INIT_SHUTDOWN_TIMEOUT
 655 #define INIT_SHUTDOWN_TIMEOUT 120
 656 #endif
 657 static int init_shutdown_timeout = INIT_SHUTDOWN_TIMEOUT;
 658 SYSCTL_INT(_kern, OID_AUTO, init_shutdown_timeout,
 659         CTLFLAG_RW, &init_shutdown_timeout, 0, "Shutdown timeout of init(8). "
 660         "Unused within kernel, but used to control init(8)");
 661
 662 /*
 663  * Start the initial user process; try exec'ing each pathname in init_path.
 664  * The program is invoked with one argument containing the boot flags.
 665  */
 666 static void
 667 start_init(void *dummy)
 668 {
 669         vm_offset_t addr;
 670         struct execve_args args;
 671         int options, error;
 672         char *var, *path, *next, *s;
 673         char *ucp, **uap, *arg0, *arg1;
 674         struct thread *td;
 675         struct proc *p;
 676
 677         mtx_lock(&Giant);
 678
 679         GIANT_REQUIRED;
 680
 681         td = curthread;
 682         p = td->td_proc;
 683
 684         vfs_mountroot();
 685
 686         /*
 687          * Need just enough stack to hold the faked-up "execve()" arguments.
 688          */
 689         addr = p->p_sysent->sv_usrstack - PAGE_SIZE;
 690         if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
 691                         FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
 692                 panic("init: couldn't allocate argument space");
 693         p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
 694         p->p_vmspace->vm_ssize = 1;
 695
 696         if ((var = getenv("init_path")) != NULL) {
 697                 strlcpy(init_path, var, sizeof(init_path));
 698                 freeenv(var);
 699         }
 700
 701         for (path = init_path; *path != '\0'; path = next) {
 702                 while (*path == ':')
 703                         path++;
 704                 if (*path == '\0')
 705                         break;
 706                 for (next = path; *next != '\0' && *next != ':'; next++)
 707                         /* nothing */ ;
 708                 if (bootverbose)
 709                         printf("start_init: trying %.*s\n", (int)(next - path),
 710                             path);
 711
 712                 /*
 713                  * Move out the boot flag argument.
 714                  */
 715                 options = 0;
 716                 ucp = (char *)p->p_sysent->sv_usrstack;
 717                 (void)subyte(--ucp, 0);         /* trailing zero */
 718                 if (boothowto & RB_SINGLE) {
 719                         (void)subyte(--ucp, 's');
 720                         options = 1;
 721                 }
 722 #ifdef notyet
 723                 if (boothowto & RB_FASTBOOT) {
 724                         (void)subyte(--ucp, 'f');
 725                         options = 1;
 726                 }
 727 #endif
 728
 729 #ifdef BOOTCDROM
 730                 (void)subyte(--ucp, 'C');
 731                 options = 1;
 732 #endif
 733
 734                 if (options == 0)
 735                         (void)subyte(--ucp, '-');
 736                 (void)subyte(--ucp, '-');               /* leading hyphen */
 737                 arg1 = ucp;
 738
 739                 /*
 740                  * Move out the file name (also arg 0).
 741                  */
 742                 (void)subyte(--ucp, 0);
 743                 for (s = next - 1; s >= path; s--)
 744                         (void)subyte(--ucp, *s);
 745                 arg0 = ucp;
 746
 747                 /*
 748                  * Move out the arg pointers.
 749                  */
 750                 uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
 751                 (void)suword((caddr_t)--uap, (long)0);  /* terminator */
 752                 (void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
 753                 (void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
 754
 755                 /*
 756                  * Point at the arguments.
 757                  */
 758                 args.fname = arg0;
 759                 args.argv = uap;
 760                 args.envv = NULL;
 761
 762                 /*
 763                  * Now try to exec the program.  If can't for any reason
 764                  * other than it doesn't exist, complain.
 765                  *
 766                  * Otherwise, return via fork_trampoline() all the way
 767                  * to user mode as init!
 768                  */
 769                 if ((error = sys_execve(td, &args)) == 0) {
 770                         mtx_unlock(&Giant);
 771                         return;
 772                 }
 773                 if (error != ENOENT)
 774                         printf("exec %.*s: error %d\n", (int)(next - path),
 775                             path, error);
 776         }
 777         printf("init: not found in path %s\n", init_path);
 778         panic("no init");
 779 }
 780
 781 /*
 782  * Like kproc_create(), but runs in it's own address space.
 783  * We do this early to reserve pid 1.
 784  *
 785  * Note special case - do not make it runnable yet.  Other work
 786  * in progress will change this more.
 787  */
 788 static void
 789 create_init(const void *udata __unused)
 790 {
 791         struct ucred *newcred, *oldcred;
 792         int error;
 793
 794         error = fork1(&thread0, RFFDG | RFPROC | RFSTOPPED, 0, &initproc,
 795             NULL, 0);
 796         if (error)
 797                 panic("cannot fork init: %d\n", error);
 798         KASSERT(initproc->p_pid == 1, ("create_init: initproc->p_pid != 1"));
 799         /* divorce init's credentials from the kernel's */
 800         newcred = crget();
 801         PROC_LOCK(initproc);
 802         initproc->p_flag |= P_SYSTEM | P_INMEM;
 803         oldcred = initproc->p_ucred;
 804         crcopy(newcred, oldcred);
 805 #ifdef MAC
 806         mac_cred_create_init(newcred);
 807 #endif
 808 #ifdef AUDIT
 809         audit_cred_proc1(newcred);
 810 #endif
 811         initproc->p_ucred = newcred;
 812         PROC_UNLOCK(initproc);
 813         crfree(oldcred);
 814         cred_update_thread(FIRST_THREAD_IN_PROC(initproc));
 815         cpu_set_fork_handler(FIRST_THREAD_IN_PROC(initproc), start_init, NULL);
 816 }
 817 SYSINIT(init, SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL);
 818
 819 /*
 820  * Make it runnable now.
 821  */
 822 static void
 823 kick_init(const void *udata __unused)
 824 {
 825         struct thread *td;
 826
 827         td = FIRST_THREAD_IN_PROC(initproc);
 828         thread_lock(td);
 829         TD_SET_CAN_RUN(td);
 830         sched_add(td, SRQ_BORING);
 831         thread_unlock(td);
 832 }
 833 SYSINIT(kickinit, SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL);