1 /* $NetBSD: init_main.c,v 1.194 2001/07/18 00:51:54 matt 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 "fs_nfs.h" 45 #include "opt_nfsserver.h" 46 #include "opt_sysv.h" 47 #include "opt_maxuprc.h" 48 #include "opt_multiprocessor.h" 49 #include "opt_new_pipe.h" 50 #include "opt_syscall_debug.h" 51 52 #include "rnd.h" 53 54 #include <sys/param.h> 55 #include <sys/acct.h> 56 #include <sys/filedesc.h> 57 #include <sys/file.h> 58 #include <sys/errno.h> 59 #include <sys/callout.h> 60 #include <sys/kernel.h> 61 #include <sys/mount.h> 62 #include <sys/map.h> 63 #include <sys/proc.h> 64 #include <sys/kthread.h> 65 #include <sys/resourcevar.h> 66 #include <sys/signalvar.h> 67 #include <sys/systm.h> 68 #include <sys/vnode.h> 69 #include <sys/tty.h> 70 #include <sys/conf.h> 71 #include <sys/disklabel.h> 72 #include <sys/buf.h> 73 #include <sys/device.h> 74 #include <sys/exec.h> 75 #include <sys/socketvar.h> 76 #include <sys/protosw.h> 77 #include <sys/reboot.h> 78 #include <sys/user.h> 79 #include <sys/sysctl.h> 80 #ifdef SYSVSHM 81 #include <sys/shm.h> 82 #endif 83 #ifdef SYSVSEM 84 #include <sys/sem.h> 85 #endif 86 #ifdef SYSVMSG 87 #include <sys/msg.h> 88 #endif 89 #include <sys/domain.h> 90 #include <sys/mbuf.h> 91 #include <sys/namei.h> 92 #if NRND > 0 93 #include <sys/rnd.h> 94 #endif 95 #ifdef NEW_PIPE 96 #include <sys/pipe.h> 97 #endif 98 99 #include <sys/syscall.h> 100 #include <sys/syscallargs.h> 101 102 #include <ufs/ufs/quota.h> 103 104 #include <miscfs/genfs/genfs.h> 105 #include <miscfs/syncfs/syncfs.h> 106 107 #include <machine/cpu.h> 108 109 #include <uvm/uvm.h> 110 111 #include <net/if.h> 112 #include <net/raw_cb.h> 113 114 const char copyright[] = 115 "Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001\n" 116 " The NetBSD Foundation, Inc. All rights reserved.\n" 117 "Copyright (c) 1982, 1986, 1989, 1991, 1993\n" 118 " The Regents of the University of California. All rights reserved.\n" 119 "\n"; 120 121 /* Components of the first process -- never freed. */ 122 struct session session0; 123 struct pgrp pgrp0; 124 struct proc proc0; 125 struct pcred cred0; 126 struct filedesc0 filedesc0; 127 struct cwdinfo cwdi0; 128 struct plimit limit0; 129 struct vmspace vmspace0; 130 struct sigacts sigacts0; 131 #ifndef curproc 132 struct proc *curproc = &proc0; 133 #endif 134 struct proc *initproc; 135 136 int cmask = CMASK; 137 extern struct user *proc0paddr; 138 139 struct vnode *rootvp, *swapdev_vp; 140 int boothowto; 141 int cold = 1; /* still working on startup */ 142 struct timeval boottime; 143 144 __volatile int start_init_exec; /* semaphore for start_init() */ 145 146 static void check_console(struct proc *p); 147 static void start_init(void *); 148 void main(void); 149 150 extern const struct emul emul_netbsd; /* defined in kern_exec.c */ 151 152 /* 153 * System startup; initialize the world, create process 0, mount root 154 * filesystem, and fork to create init and pagedaemon. Most of the 155 * hard work is done in the lower-level initialization routines including 156 * startup(), which does memory initialization and autoconfiguration. 157 */ 158 void 159 main(void) 160 { 161 struct proc *p; 162 struct pdevinit *pdev; 163 int i, s, error; 164 rlim_t lim; 165 extern struct pdevinit pdevinit[]; 166 extern void schedcpu(void *); 167 extern void disk_init(void); 168 #if defined(NFSSERVER) || defined(NFS) 169 extern void nfs_init(void); 170 #endif 171 #ifdef NVNODE_IMPLICIT 172 int usevnodes; 173 #endif 174 175 /* 176 * Initialize the current process pointer (curproc) before 177 * any possible traps/probes to simplify trap processing. 178 */ 179 p = &proc0; 180 curproc = p; 181 p->p_cpu = curcpu(); 182 /* 183 * Attempt to find console and initialize 184 * in case of early panic or other messages. 185 */ 186 consinit(); 187 printf("%s", copyright); 188 189 KERNEL_LOCK_INIT(); 190 191 uvm_init(); 192 193 /* Do machine-dependent initialization. */ 194 cpu_startup(); 195 196 /* Initialize callouts. */ 197 callout_startup(); 198 199 /* 200 * Initialize mbuf's. Do this now because we might attempt to 201 * allocate mbufs or mbuf clusters during autoconfiguration. 202 */ 203 mbinit(); 204 205 /* Initialize sockets. */ 206 soinit(); 207 208 /* 209 * The following 3 things must be done before autoconfiguration. 210 */ 211 disk_init(); /* initialize disk list */ 212 tty_init(); /* initialize tty list */ 213 #if NRND > 0 214 rnd_init(); /* initialize RNG */ 215 #endif 216 217 /* Initialize the sysctl subsystem. */ 218 sysctl_init(); 219 220 /* 221 * Initialize process and pgrp structures. 222 */ 223 procinit(); 224 225 /* 226 * Create process 0 (the swapper). 227 */ 228 s = proclist_lock_write(); 229 LIST_INSERT_HEAD(&allproc, p, p_list); 230 LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash); 231 proclist_unlock_write(s); 232 233 p->p_pgrp = &pgrp0; 234 LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); 235 LIST_INIT(&pgrp0.pg_members); 236 LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); 237 238 pgrp0.pg_session = &session0; 239 session0.s_count = 1; 240 session0.s_sid = p->p_pid; 241 session0.s_leader = p; 242 243 /* 244 * Set P_NOCLDWAIT so that kernel threads are reparented to 245 * init(8) when they exit. init(8) can easily wait them out 246 * for us. 247 */ 248 p->p_flag = P_INMEM | P_SYSTEM | P_NOCLDWAIT; 249 p->p_stat = SONPROC; 250 p->p_nice = NZERO; 251 p->p_emul = &emul_netbsd; 252 #ifdef __HAVE_SYSCALL_INTERN 253 (*p->p_emul->e_syscall_intern)(p); 254 #endif 255 strncpy(p->p_comm, "swapper", MAXCOMLEN); 256 257 callout_init(&p->p_realit_ch); 258 callout_init(&p->p_tsleep_ch); 259 260 /* Create credentials. */ 261 cred0.p_refcnt = 1; 262 p->p_cred = &cred0; 263 p->p_ucred = crget(); 264 p->p_ucred->cr_ngroups = 1; /* group 0 */ 265 266 /* Create the file descriptor table. */ 267 finit(); 268 p->p_fd = &filedesc0.fd_fd; 269 fdinit1(&filedesc0); 270 271 /* Create the CWD info. */ 272 p->p_cwdi = &cwdi0; 273 cwdi0.cwdi_cmask = cmask; 274 cwdi0.cwdi_refcnt = 1; 275 276 /* Create the limits structures. */ 277 p->p_limit = &limit0; 278 for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++) 279 limit0.pl_rlimit[i].rlim_cur = 280 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY; 281 282 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 283 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = 284 maxfiles < NOFILE ? maxfiles : NOFILE; 285 286 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 287 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = 288 maxproc < MAXUPRC ? maxproc : MAXUPRC; 289 290 lim = ptoa(uvmexp.free); 291 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim; 292 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim; 293 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 294 limit0.pl_corename = defcorename; 295 limit0.p_refcnt = 1; 296 297 /* 298 * Initialize proc0's vmspace, which uses the kernel pmap. 299 * All kernel processes (which never have user space mappings) 300 * share proc0's vmspace, and thus, the kernel pmap. 301 */ 302 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 303 trunc_page(VM_MAX_ADDRESS), TRUE); 304 p->p_vmspace = &vmspace0; 305 306 p->p_addr = proc0paddr; /* XXX */ 307 308 /* 309 * We continue to place resource usage info in the 310 * user struct so they're pageable. 311 */ 312 p->p_stats = &p->p_addr->u_stats; 313 314 /* 315 * Charge root for one process. 316 */ 317 (void)chgproccnt(0, 1); 318 319 rqinit(); 320 321 /* Configure virtual memory system, set vm rlimits. */ 322 uvm_init_limits(p); 323 324 /* Initialize the file systems. */ 325 #if defined(NFSSERVER) || defined(NFS) 326 nfs_init(); /* initialize server/shared data */ 327 #endif 328 #ifdef NVNODE_IMPLICIT 329 /* 330 * If maximum number of vnodes in namei vnode cache is not explicitly 331 * defined in kernel config, adjust the number such as we use roughly 332 * 0.5% of memory for vnode cache (but not less than NVNODE vnodes). 333 */ 334 usevnodes = (ptoa((unsigned)physmem) / 200) / sizeof(struct vnode); 335 if (usevnodes > desiredvnodes) 336 desiredvnodes = usevnodes; 337 #endif 338 vfsinit(); 339 340 /* Configure the system hardware. This will enable interrupts. */ 341 configure(); 342 343 ubc_init(); /* must be after autoconfig */ 344 345 /* Lock the kernel on behalf of proc0. */ 346 KERNEL_PROC_LOCK(p); 347 348 #ifdef SYSVSHM 349 /* Initialize System V style shared memory. */ 350 shminit(); 351 #endif 352 353 #ifdef SYSVSEM 354 /* Initialize System V style semaphores. */ 355 seminit(); 356 #endif 357 358 #ifdef SYSVMSG 359 /* Initialize System V style message queues. */ 360 msginit(); 361 #endif 362 363 /* Attach pseudo-devices. */ 364 for (pdev = pdevinit; pdev->pdev_attach != NULL; pdev++) 365 (*pdev->pdev_attach)(pdev->pdev_count); 366 367 /* 368 * Initialize protocols. Block reception of incoming packets 369 * until everything is ready. 370 */ 371 s = splnet(); 372 ifinit(); 373 domaininit(); 374 splx(s); 375 376 #ifdef GPROF 377 /* Initialize kernel profiling. */ 378 kmstartup(); 379 #endif 380 381 /* Initialize system accouting. */ 382 acct_init(); 383 384 /* 385 * Initialize signal-related data structures, and signal state 386 * for proc0. 387 */ 388 signal_init(); 389 p->p_sigacts = &sigacts0; 390 siginit(p); 391 392 /* Kick off timeout driven events by calling first time. */ 393 schedcpu(NULL); 394 395 /* 396 * Create process 1 (init(8)). We do this now, as Unix has 397 * historically had init be process 1, and changing this would 398 * probably upset a lot of people. 399 * 400 * Note that process 1 won't immediately exec init(8), but will 401 * wait for us to inform it that the root file system has been 402 * mounted. 403 */ 404 if (fork1(p, 0, SIGCHLD, NULL, 0, start_init, NULL, NULL, &initproc)) 405 panic("fork init"); 406 407 /* 408 * Create any kernel threads who's creation was deferred because 409 * initproc had not yet been created. 410 */ 411 kthread_run_deferred_queue(); 412 413 /* 414 * Now that device driver threads have been created, wait for 415 * them to finish any deferred autoconfiguration. Note we don't 416 * need to lock this semaphore, since we haven't booted any 417 * secondary processors, yet. 418 */ 419 while (config_pending) 420 (void) tsleep((void *)&config_pending, PWAIT, "cfpend", 0); 421 422 /* 423 * Now that autoconfiguration has completed, we can determine 424 * the root and dump devices. 425 */ 426 cpu_rootconf(); 427 cpu_dumpconf(); 428 429 /* Mount the root file system. */ 430 do { 431 domountroothook(); 432 if ((error = vfs_mountroot())) { 433 printf("cannot mount root, error = %d\n", error); 434 boothowto |= RB_ASKNAME; 435 setroot(root_device, 436 (rootdev != NODEV) ? DISKPART(rootdev) : 0); 437 } 438 } while (error != 0); 439 mountroothook_destroy(); 440 441 mountlist.cqh_first->mnt_flag |= MNT_ROOTFS; 442 mountlist.cqh_first->mnt_op->vfs_refcount++; 443 444 /* 445 * Get the vnode for '/'. Set filedesc0.fd_fd.fd_cdir to 446 * reference it. 447 */ 448 if (VFS_ROOT(mountlist.cqh_first, &rootvnode)) 449 panic("cannot find root vnode"); 450 cwdi0.cwdi_cdir = rootvnode; 451 VREF(cwdi0.cwdi_cdir); 452 VOP_UNLOCK(rootvnode, 0); 453 cwdi0.cwdi_rdir = NULL; 454 455 /* 456 * Now that root is mounted, we can fixup initproc's CWD 457 * info. All other processes are kthreads, which merely 458 * share proc0's CWD info. 459 */ 460 initproc->p_cwdi->cwdi_cdir = rootvnode; 461 VREF(initproc->p_cwdi->cwdi_cdir); 462 initproc->p_cwdi->cwdi_rdir = NULL; 463 464 /* 465 * Now can look at time, having had a chance to verify the time 466 * from the file system. Reset p->p_rtime as it may have been 467 * munched in mi_switch() after the time got set. 468 */ 469 proclist_lock_read(); 470 s = splsched(); 471 for (p = LIST_FIRST(&allproc); p != NULL; 472 p = LIST_NEXT(p, p_list)) { 473 p->p_stats->p_start = mono_time = boottime = time; 474 if (p->p_cpu != NULL) 475 p->p_cpu->ci_schedstate.spc_runtime = time; 476 p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0; 477 } 478 splx(s); 479 proclist_unlock_read(); 480 481 /* Create the pageout daemon kernel thread. */ 482 uvm_swap_init(); 483 if (kthread_create1(uvm_pageout, NULL, NULL, "pagedaemon")) 484 panic("fork pagedaemon"); 485 486 /* Create the process reaper kernel thread. */ 487 if (kthread_create1(reaper, NULL, NULL, "reaper")) 488 panic("fork reaper"); 489 490 /* Create the filesystem syncer kernel thread. */ 491 if (kthread_create1(sched_sync, NULL, NULL, "ioflush")) 492 panic("fork syncer"); 493 494 /* Create the aiodone daemon kernel thread. */ 495 if (kthread_create1(uvm_aiodone_daemon, NULL, NULL, "aiodoned")) 496 panic("fork aiodoned"); 497 498 #if defined(MULTIPROCESSOR) 499 /* Boot the secondary processors. */ 500 cpu_boot_secondary_processors(); 501 #endif 502 503 /* Initialize exec structures */ 504 exec_init(1); 505 506 #ifdef NEW_PIPE 507 /* Initialize pipe structures */ 508 pipe_init(); 509 #endif 510 511 /* 512 * Okay, now we can let init(8) exec! It's off to userland! 513 */ 514 start_init_exec = 1; 515 wakeup((void *)&start_init_exec); 516 517 /* The scheduler is an infinite loop. */ 518 uvm_scheduler(); 519 /* NOTREACHED */ 520 } 521 522 static void 523 check_console(struct proc *p) 524 { 525 struct nameidata nd; 526 int error; 527 528 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/dev/console", p); 529 error = namei(&nd); 530 if (error == 0) 531 vrele(nd.ni_vp); 532 else if (error == ENOENT) 533 printf("warning: no /dev/console\n"); 534 else 535 printf("warning: lookup /dev/console: error %d\n", error); 536 } 537 538 /* 539 * List of paths to try when searching for "init". 540 */ 541 static const char *initpaths[] = { 542 "/sbin/init", 543 "/sbin/oinit", 544 "/sbin/init.bak", 545 NULL, 546 }; 547 548 /* 549 * Start the initial user process; try exec'ing each pathname in "initpaths". 550 * The program is invoked with one argument containing the boot flags. 551 */ 552 static void 553 start_init(void *arg) 554 { 555 struct proc *p = arg; 556 vaddr_t addr; 557 struct sys_execve_args /* { 558 syscallarg(const char *) path; 559 syscallarg(char * const *) argp; 560 syscallarg(char * const *) envp; 561 } */ args; 562 int options, i, error; 563 register_t retval[2]; 564 char flags[4], *flagsp; 565 const char **pathp, *path, *slash; 566 char *ucp, **uap, *arg0, *arg1 = NULL; 567 568 /* 569 * Now in process 1. 570 */ 571 strncpy(p->p_comm, "init", MAXCOMLEN); 572 573 /* 574 * Wait for main() to tell us that it's safe to exec. 575 */ 576 while (start_init_exec == 0) 577 (void) tsleep((void *)&start_init_exec, PWAIT, "initexec", 0); 578 579 /* 580 * This is not the right way to do this. We really should 581 * hand-craft a descriptor onto /dev/console to hand to init, 582 * but that's a _lot_ more work, and the benefit from this easy 583 * hack makes up for the "good is the enemy of the best" effect. 584 */ 585 check_console(p); 586 587 /* 588 * Need just enough stack to hold the faked-up "execve()" arguments. 589 */ 590 addr = USRSTACK - PAGE_SIZE; 591 if (uvm_map(&p->p_vmspace->vm_map, &addr, PAGE_SIZE, 592 NULL, UVM_UNKNOWN_OFFSET, 0, 593 UVM_MAPFLAG(UVM_PROT_ALL, UVM_PROT_ALL, UVM_INH_COPY, 594 UVM_ADV_NORMAL, 595 UVM_FLAG_FIXED|UVM_FLAG_OVERLAY|UVM_FLAG_COPYONW)) != 0) 596 panic("init: couldn't allocate argument space"); 597 p->p_vmspace->vm_maxsaddr = (caddr_t)addr; 598 599 for (pathp = &initpaths[0]; (path = *pathp) != NULL; pathp++) { 600 ucp = (char *)(addr + PAGE_SIZE); 601 602 /* 603 * Construct the boot flag argument. 604 */ 605 flagsp = flags; 606 *flagsp++ = '-'; 607 options = 0; 608 609 if (boothowto & RB_SINGLE) { 610 *flagsp++ = 's'; 611 options = 1; 612 } 613 #ifdef notyet 614 if (boothowto & RB_FASTBOOT) { 615 *flagsp++ = 'f'; 616 options = 1; 617 } 618 #endif 619 620 /* 621 * Move out the flags (arg 1), if necessary. 622 */ 623 if (options != 0) { 624 *flagsp++ = '\0'; 625 i = flagsp - flags; 626 #ifdef DEBUG 627 printf("init: copying out flags `%s' %d\n", flags, i); 628 #endif 629 (void)copyout((caddr_t)flags, (caddr_t)(ucp -= i), i); 630 arg1 = ucp; 631 } 632 633 /* 634 * Move out the file name (also arg 0). 635 */ 636 i = strlen(path) + 1; 637 #ifdef DEBUG 638 printf("init: copying out path `%s' %d\n", path, i); 639 #endif 640 (void)copyout((caddr_t)path, (caddr_t)(ucp -= i), i); 641 arg0 = ucp; 642 643 /* 644 * Move out the arg pointers. 645 */ 646 uap = (char **)((long)ucp & ~ALIGNBYTES); 647 (void)suword((caddr_t)--uap, 0); /* terminator */ 648 if (options != 0) 649 (void)suword((caddr_t)--uap, (long)arg1); 650 slash = strrchr(path, '/'); 651 if (slash) 652 (void)suword((caddr_t)--uap, 653 (long)arg0 + (slash + 1 - path)); 654 else 655 (void)suword((caddr_t)--uap, (long)arg0); 656 657 /* 658 * Point at the arguments. 659 */ 660 SCARG(&args, path) = arg0; 661 SCARG(&args, argp) = uap; 662 SCARG(&args, envp) = NULL; 663 664 /* 665 * Now try to exec the program. If can't for any reason 666 * other than it doesn't exist, complain. 667 */ 668 error = sys_execve(p, &args, retval); 669 if (error == 0 || error == EJUSTRETURN) { 670 KERNEL_PROC_UNLOCK(p); 671 return; 672 } 673 if (error != ENOENT) 674 printf("exec %s: error %d\n", path, error); 675 } 676 printf("init: not found\n"); 677 panic("no init"); 678 } 679