1 /* $NetBSD: init_main.c,v 1.200 2002/05/27 13:46:45 itojun Exp $ */ 2 3 /* 4 * Copyright (c) 1995 Christopher G. Demetriou. All rights reserved. 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 8.16 (Berkeley) 5/14/95 42 */ 43 44 #include <sys/cdefs.h> 45 __KERNEL_RCSID(0, "$NetBSD: init_main.c,v 1.200 2002/05/27 13:46:45 itojun Exp $"); 46 47 #include "fs_nfs.h" 48 #include "opt_nfsserver.h" 49 #include "opt_sysv.h" 50 #include "opt_maxuprc.h" 51 #include "opt_multiprocessor.h" 52 #include "opt_pipe.h" 53 #include "opt_syscall_debug.h" 54 55 #include "rnd.h" 56 57 #include <sys/param.h> 58 #include <sys/acct.h> 59 #include <sys/filedesc.h> 60 #include <sys/file.h> 61 #include <sys/errno.h> 62 #include <sys/callout.h> 63 #include <sys/kernel.h> 64 #include <sys/mount.h> 65 #include <sys/map.h> 66 #include <sys/proc.h> 67 #include <sys/kthread.h> 68 #include <sys/resourcevar.h> 69 #include <sys/signalvar.h> 70 #include <sys/systm.h> 71 #include <sys/vnode.h> 72 #include <sys/tty.h> 73 #include <sys/conf.h> 74 #include <sys/disklabel.h> 75 #include <sys/buf.h> 76 #include <sys/device.h> 77 #include <sys/disk.h> 78 #include <sys/exec.h> 79 #include <sys/socketvar.h> 80 #include <sys/protosw.h> 81 #include <sys/reboot.h> 82 #include <sys/user.h> 83 #include <sys/sysctl.h> 84 #ifdef SYSVSHM 85 #include <sys/shm.h> 86 #endif 87 #ifdef SYSVSEM 88 #include <sys/sem.h> 89 #endif 90 #ifdef SYSVMSG 91 #include <sys/msg.h> 92 #endif 93 #include <sys/domain.h> 94 #include <sys/mbuf.h> 95 #include <sys/namei.h> 96 #if NRND > 0 97 #include <sys/rnd.h> 98 #endif 99 #ifndef PIPE_SOCKETPAIR 100 #include <sys/pipe.h> 101 #endif 102 103 #include <sys/syscall.h> 104 #include <sys/syscallargs.h> 105 106 #include <ufs/ufs/quota.h> 107 108 #include <miscfs/genfs/genfs.h> 109 #include <miscfs/syncfs/syncfs.h> 110 111 #include <machine/cpu.h> 112 113 #include <uvm/uvm.h> 114 115 #include <net/if.h> 116 #include <net/raw_cb.h> 117 118 const char copyright[] = 119 "Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002\n" 120 " The NetBSD Foundation, Inc. All rights reserved.\n" 121 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n" 122 " The Regents of the University of California. All rights reserved.\n" 123 "\n"; 124 125 /* Components of the first process -- never freed. */ 126 struct session session0; 127 struct pgrp pgrp0; 128 struct proc proc0; 129 struct pcred cred0; 130 struct filedesc0 filedesc0; 131 struct cwdinfo cwdi0; 132 struct plimit limit0; 133 struct vmspace vmspace0; 134 struct sigacts sigacts0; 135 #ifndef curproc 136 struct proc *curproc = &proc0; 137 #endif 138 struct proc *initproc; 139 140 int cmask = CMASK; 141 extern struct user *proc0paddr; 142 143 struct vnode *rootvp, *swapdev_vp; 144 int boothowto; 145 int cold = 1; /* still working on startup */ 146 struct timeval boottime; 147 148 __volatile int start_init_exec; /* semaphore for start_init() */ 149 150 static void check_console(struct proc *p); 151 static void start_init(void *); 152 void main(void); 153 154 extern const struct emul emul_netbsd; /* defined in kern_exec.c */ 155 156 /* 157 * System startup; initialize the world, create process 0, mount root 158 * filesystem, and fork to create init and pagedaemon. Most of the 159 * hard work is done in the lower-level initialization routines including 160 * startup(), which does memory initialization and autoconfiguration. 161 */ 162 void 163 main(void) 164 { 165 struct proc *p; 166 struct pdevinit *pdev; 167 int i, s, error; 168 rlim_t lim; 169 extern struct pdevinit pdevinit[]; 170 extern void schedcpu(void *); 171 #if defined(NFSSERVER) || defined(NFS) 172 extern void nfs_init(void); 173 #endif 174 #ifdef NVNODE_IMPLICIT 175 int usevnodes; 176 #endif 177 178 /* 179 * Initialize the current process pointer (curproc) before 180 * any possible traps/probes to simplify trap processing. 181 */ 182 p = &proc0; 183 curproc = p; 184 p->p_cpu = curcpu(); 185 /* 186 * Attempt to find console and initialize 187 * in case of early panic or other messages. 188 */ 189 consinit(); 190 printf("%s", copyright); 191 192 KERNEL_LOCK_INIT(); 193 194 uvm_init(); 195 196 /* Do machine-dependent initialization. */ 197 cpu_startup(); 198 199 /* Initialize callouts. */ 200 callout_startup(); 201 202 /* 203 * Initialize mbuf's. Do this now because we might attempt to 204 * allocate mbufs or mbuf clusters during autoconfiguration. 205 */ 206 mbinit(); 207 208 /* Initialize sockets. */ 209 soinit(); 210 211 /* 212 * The following 3 things must be done before autoconfiguration. 213 */ 214 disk_init(); /* initialize disk list */ 215 tty_init(); /* initialize tty list */ 216 #if NRND > 0 217 rnd_init(); /* initialize RNG */ 218 #endif 219 220 /* Initialize the sysctl subsystem. */ 221 sysctl_init(); 222 223 /* 224 * Initialize process and pgrp structures. 225 */ 226 procinit(); 227 228 /* 229 * Create process 0 (the swapper). 230 */ 231 s = proclist_lock_write(); 232 LIST_INSERT_HEAD(&allproc, p, p_list); 233 LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash); 234 proclist_unlock_write(s); 235 236 p->p_pgrp = &pgrp0; 237 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 238 LIST_INIT(&pgrp0.pg_members); 239 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 240 241 pgrp0.pg_session = &session0; 242 session0.s_count = 1; 243 session0.s_sid = p->p_pid; 244 session0.s_leader = p; 245 246 /* 247 * Set P_NOCLDWAIT so that kernel threads are reparented to 248 * init(8) when they exit. init(8) can easily wait them out 249 * for us. 250 */ 251 p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT; 252 p->p_stat = SONPROC; 253 p->p_nice = NZERO; 254 p->p_emul = &emul_netbsd; 255 #ifdef __HAVE_SYSCALL_INTERN 256 (*p->p_emul->e_syscall_intern)(p); 257 #endif 258 strncpy(p->p_comm, "swapper", MAXCOMLEN); 259 260 callout_init(&p->p_realit_ch); 261 callout_init(&p->p_tsleep_ch); 262 263 /* Create credentials. */ 264 cred0.p_refcnt = 1; 265 p->p_cred = &cred0; 266 p->p_ucred = crget(); 267 p->p_ucred->cr_ngroups = 1; /* group 0 */ 268 269 /* Create the file descriptor table. */ 270 finit(); 271 p->p_fd = &filedesc0.fd_fd; 272 fdinit1(&filedesc0); 273 274 /* Create the CWD info. */ 275 p->p_cwdi = &cwdi0; 276 cwdi0.cwdi_cmask = cmask; 277 cwdi0.cwdi_refcnt = 1; 278 279 /* Create the limits structures. */ 280 p->p_limit = &limit0; 281 for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++) 282 limit0.pl_rlimit[i].rlim_cur = 283 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY; 284 285 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 286 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = 287 maxfiles < NOFILE ? maxfiles : NOFILE; 288 289 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 290 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = 291 maxproc < MAXUPRC ? maxproc : MAXUPRC; 292 293 lim = ptoa(uvmexp.free); 294 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim; 295 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim; 296 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 297 limit0.pl_corename = defcorename; 298 limit0.p_refcnt = 1; 299 300 /* 301 * Initialize proc0's vmspace, which uses the kernel pmap. 302 * All kernel processes (which never have user space mappings) 303 * share proc0's vmspace, and thus, the kernel pmap. 304 */ 305 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 306 trunc_page(VM_MAX_ADDRESS)); 307 p->p_vmspace = &vmspace0; 308 309 p->p_addr = proc0paddr; /* XXX */ 310 311 /* 312 * We continue to place resource usage info in the 313 * user struct so they're pageable. 314 */ 315 p->p_stats = &p->p_addr->u_stats; 316 317 /* 318 * Charge root for one process. 319 */ 320 (void)chgproccnt(0, 1); 321 322 rqinit(); 323 324 /* Configure virtual memory system, set vm rlimits. */ 325 uvm_init_limits(p); 326 327 /* Initialize the file systems. */ 328 #if defined(NFSSERVER) || defined(NFS) 329 nfs_init(); /* initialize server/shared data */ 330 #endif 331 #ifdef NVNODE_IMPLICIT 332 /* 333 * If maximum number of vnodes in namei vnode cache is not explicitly 334 * defined in kernel config, adjust the number such as we use roughly 335 * 0.5% of memory for vnode cache (but not less than NVNODE vnodes). 336 */ 337 usevnodes = (ptoa((unsigned)physmem) / 200) / sizeof(struct vnode); 338 if (usevnodes > desiredvnodes) 339 desiredvnodes = usevnodes; 340 #endif 341 vfsinit(); 342 343 /* Configure the system hardware. This will enable interrupts. */ 344 configure(); 345 346 ubc_init(); /* must be after autoconfig */ 347 348 /* Lock the kernel on behalf of proc0. */ 349 KERNEL_PROC_LOCK(p); 350 351 #ifdef SYSVSHM 352 /* Initialize System V style shared memory. */ 353 shminit(); 354 #endif 355 356 #ifdef SYSVSEM 357 /* Initialize System V style semaphores. */ 358 seminit(); 359 #endif 360 361 #ifdef SYSVMSG 362 /* Initialize System V style message queues. */ 363 msginit(); 364 #endif 365 366 /* Attach pseudo-devices. */ 367 for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++) 368 (*pdev->pdev_attach)(pdev->pdev_count); 369 370 /* 371 * Initialize protocols. Block reception of incoming packets 372 * until everything is ready. 373 */ 374 s = splnet(); 375 ifinit(); 376 domaininit(); 377 if_attachdomain(); 378 splx(s); 379 380 #ifdef GPROF 381 /* Initialize kernel profiling. */ 382 kmstartup(); 383 #endif 384 385 /* Initialize system accouting. */ 386 acct_init(); 387 388 /* 389 * Initialize signal-related data structures, and signal state 390 * for proc0. 391 */ 392 signal_init(); 393 p->p_sigacts = &sigacts0; 394 siginit(p); 395 396 /* Kick off timeout driven events by calling first time. */ 397 schedcpu(NULL); 398 399 /* 400 * Create process 1 (init(8)). We do this now, as Unix has 401 * historically had init be process 1, and changing this would 402 * probably upset a lot of people. 403 * 404 * Note that process 1 won't immediately exec init(8), but will 405 * wait for us to inform it that the root file system has been 406 * mounted. 407 */ 408 if (fork1(p, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc)) 409 panic("fork init"); 410 411 /* 412 * Create any kernel threads who's creation was deferred because 413 * initproc had not yet been created. 414 */ 415 kthread_run_deferred_queue(); 416 417 /* 418 * Now that device driver threads have been created, wait for 419 * them to finish any deferred autoconfiguration. Note we don't 420 * need to lock this semaphore, since we haven't booted any 421 * secondary processors, yet. 422 */ 423 while (config_pending) 424 (void) tsleep((void *)&config_pending, PWAIT, "cfpend", 0); 425 426 /* 427 * Now that autoconfiguration has completed, we can determine 428 * the root and dump devices. 429 */ 430 cpu_rootconf(); 431 cpu_dumpconf(); 432 433 /* Mount the root file system. */ 434 do { 435 domountroothook(); 436 if ((error = vfs_mountroot())) { 437 printf("cannot mount root, error = %d\n", error); 438 boothowto |= RB_ASKNAME; 439 setroot(root_device, 440 (rootdev != NODEV) ? DISKPART(rootdev) : 0); 441 } 442 } while (error != 0); 443 mountroothook_destroy(); 444 445 mountlist.cqh_first->mnt_flag |= MNT_ROOTFS; 446 mountlist.cqh_first->mnt_op->vfs_refcount++; 447 448 /* 449 * Get the vnode for '/'. Set filedesc0.fd_fd.fd_cdir to 450 * reference it. 451 */ 452 if (VFS_ROOT(mountlist.cqh_first, &rootvnode)) 453 panic("cannot find root vnode"); 454 cwdi0.cwdi_cdir = rootvnode; 455 VREF(cwdi0.cwdi_cdir); 456 VOP_UNLOCK(rootvnode, 0); 457 cwdi0.cwdi_rdir = NULL; 458 459 /* 460 * Now that root is mounted, we can fixup initproc's CWD 461 * info. All other processes are kthreads, which merely 462 * share proc0's CWD info. 463 */ 464 initproc->p_cwdi->cwdi_cdir = rootvnode; 465 VREF(initproc->p_cwdi->cwdi_cdir); 466 initproc->p_cwdi->cwdi_rdir = NULL; 467 468 /* 469 * Now can look at time, having had a chance to verify the time 470 * from the file system. Reset p->p_rtime as it may have been 471 * munched in mi_switch() after the time got set. 472 */ 473 proclist_lock_read(); 474 s = splsched(); 475 for (p = LIST_FIRST(&allproc); p != NULL; 476 p = LIST_NEXT(p, p_list)) { 477 p->p_stats->p_start = mono_time = boottime = time; 478 if (p->p_cpu != NULL) 479 p->p_cpu->ci_schedstate.spc_runtime = time; 480 p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0; 481 } 482 splx(s); 483 proclist_unlock_read(); 484 485 /* Create the pageout daemon kernel thread. */ 486 uvm_swap_init(); 487 if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon")) 488 panic("fork pagedaemon"); 489 490 /* Create the process reaper kernel thread. */ 491 if (kthread_create1(reaper, NULL, NULL, "reaper")) 492 panic("fork reaper"); 493 494 /* Create the filesystem syncer kernel thread. */ 495 if (kthread_create1(sched_sync, NULL, NULL, "ioflush")) 496 panic("fork syncer"); 497 498 /* Create the aiodone daemon kernel thread. */ 499 if (kthread_create1(uvm_aiodone_daemon, NULL, NULL, "aiodoned")) 500 panic("fork aiodoned"); 501 502 #if defined(MULTIPROCESSOR) 503 /* Boot the secondary processors. */ 504 cpu_boot_secondary_processors(); 505 #endif 506 507 /* Initialize exec structures */ 508 exec_init(1); 509 510 #ifndef PIPE_SOCKETPAIR 511 /* Initialize pipe structures */ 512 pipe_init(); 513 #endif 514 515 /* 516 * Okay, now we can let init(8) exec! It's off to userland! 517 */ 518 start_init_exec = 1; 519 wakeup((void *)&start_init_exec); 520 521 /* The scheduler is an infinite loop. */ 522 uvm_scheduler(); 523 /* NOTREACHED */ 524 } 525 526 static void 527 check_console(struct proc *p) 528 { 529 struct nameidata nd; 530 int error; 531 532 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p); 533 error = namei(&nd); 534 if (error == 0) 535 vrele(nd.ni_vp); 536 else if (error == ENOENT) 537 printf("warning: no /dev/console\n"); 538 else 539 printf("warning: lookup /dev/console: error %d\n", error); 540 } 541 542 /* 543 * List of paths to try when searching for "init". 544 */ 545 static const char *initpaths[] = { 546 "/sbin/init", 547 "/sbin/oinit", 548 "/sbin/init.bak", 549 NULL, 550 }; 551 552 /* 553 * Start the initial user process; try exec'ing each pathname in "initpaths". 554 * The program is invoked with one argument containing the boot flags. 555 */ 556 static void 557 start_init(void *arg) 558 { 559 struct proc *p = arg; 560 vaddr_t addr; 561 struct sys_execve_args /* { 562 syscallarg(const char *) path; 563 syscallarg(char * const *) argp; 564 syscallarg(char * const *) envp; 565 } */ args; 566 int options, i, error; 567 register_t retval[2]; 568 char flags[4], *flagsp; 569 const char **pathp, *path, *slash; 570 char *ucp, **uap, *arg0, *arg1 = NULL; 571 572 /* 573 * Now in process 1. 574 */ 575 strncpy(p->p_comm, "init", MAXCOMLEN); 576 577 /* 578 * Wait for main() to tell us that it's safe to exec. 579 */ 580 while (start_init_exec == 0) 581 (void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0); 582 583 /* 584 * This is not the right way to do this. We really should 585 * hand-craft a descriptor onto /dev/console to hand to init, 586 * but that's a _lot_ more work, and the benefit from this easy 587 * hack makes up for the "good is the enemy of the best" effect. 588 */ 589 check_console(p); 590 591 /* 592 * Need just enough stack to hold the faked-up "execve()" arguments. 593 */ 594 addr = USRSTACK - PAGE_SIZE; 595 if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE, 596 NULL, UVM_UNKNOWN_OFFSET, 0, 597 UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY, 598 UVM_ADV_NORMAL, 599 UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0) 600 panic("init: couldn't allocate argument space"); 601 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 602 603 for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) { 604 ucp = (char *)(addr + PAGE_SIZE); 605 606 /* 607 * Construct the boot flag argument. 608 */ 609 flagsp = flags; 610 *flagsp++ = '-'; 611 options = 0; 612 613 if (boothowto & RB_SINGLE) { 614 *flagsp++ = 's'; 615 options = 1; 616 } 617 #ifdef notyet 618 if (boothowto & RB_FASTBOOT) { 619 *flagsp++ = 'f'; 620 options = 1; 621 } 622 #endif 623 624 /* 625 * Move out the flags (arg 1), if necessary. 626 */ 627 if (options != 0) { 628 *flagsp++ = '\0'; 629 i = flagsp - flags; 630 #ifdef DEBUG 631 printf("init: copying out flags `%s' %d\n", flags, i); 632 #endif 633 (void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i); 634 arg1 = ucp; 635 } 636 637 /* 638 * Move out the file name (also arg 0). 639 */ 640 i = strlen(path) + 1; 641 #ifdef DEBUG 642 printf("init: copying out path `%s' %d\n", path, i); 643 #endif 644 (void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i); 645 arg0 = ucp; 646 647 /* 648 * Move out the arg pointers. 649 */ 650 uap = (char **)((long)ucp & ~ALIGNBYTES); 651 (void)suword((caddr_t)--uap, 0); /* terminator */ 652 if (options != 0) 653 (void)suword((caddr_t)--uap, (long)arg1); 654 slash = strrchr(path, '/'); 655 if (slash) 656 (void)suword((caddr_t)--uap, 657 (long)arg0 + (slash + 1 - path)); 658 else 659 (void)suword((caddr_t)--uap, (long)arg0); 660 661 /* 662 * Point at the arguments. 663 */ 664 SCARG(&args, path) = arg0; 665 SCARG(&args, argp) = uap; 666 SCARG(&args, envp) = NULL; 667 668 /* 669 * Now try to exec the program. If can't for any reason 670 * other than it doesn't exist, complain. 671 */ 672 error = sys_execve(p, &args, retval); 673 if (error == 0 || error == EJUSTRETURN) { 674 KERNEL_PROC_UNLOCK(p); 675 return; 676 } 677 if (error != ENOENT) 678 printf("exec %s: error %d\n", path, error); 679 } 680 printf("init: not found\n"); 681 panic("no init"); 682 } 683