1 /*- 2 * Copyright (c) 1990 The Regents of the University of California. 3 * Copyright (c) 2008 The DragonFly Project. 4 * All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * William Jolitz. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * from: @(#)autoconf.c 7.1 (Berkeley) 5/9/91 34 * $FreeBSD: src/sys/i386/i386/autoconf.c,v 1.146.2.2 2001/06/07 06:05:58 dd Exp $ 35 */ 36 37 /* 38 * Setup the system to run on the current machine. 39 * 40 * Configure() is called at boot time and initializes the vba 41 * device tables and the memory controller monitoring. Available 42 * devices are determined (from possibilities mentioned in ioconf.c), 43 * and the drivers are initialized. 44 */ 45 #include "opt_bootp.h" 46 #include "opt_ffs.h" 47 #include "opt_cd9660.h" 48 #include "opt_nfs.h" 49 #include "opt_nfsroot.h" 50 #include "opt_rootdevname.h" 51 52 #include "use_isa.h" 53 54 #include <sys/param.h> 55 #include <sys/systm.h> 56 #include <sys/bootmaj.h> 57 #include <sys/bus.h> 58 #include <sys/buf.h> 59 #include <sys/conf.h> 60 #include <sys/diskslice.h> 61 #include <sys/reboot.h> 62 #include <sys/kernel.h> 63 #include <sys/malloc.h> 64 #include <sys/mount.h> 65 #include <sys/cons.h> 66 #include <sys/thread.h> 67 #include <sys/device.h> 68 #include <sys/machintr.h> 69 70 #include <vm/vm_kern.h> 71 #include <vm/vm_extern.h> 72 #include <vm/vm_pager.h> 73 74 #if 0 75 #include <machine/pcb.h> 76 #include <machine/pcb_ext.h> 77 #endif 78 #include <machine/smp.h> 79 #include <machine/globaldata.h> 80 #include <machine/md_var.h> 81 82 #if NISA > 0 83 #include <bus/isa/isavar.h> 84 85 device_t isa_bus_device = NULL; 86 #endif 87 88 static void cpu_startup (void *); 89 static void configure_first (void *); 90 static void configure (void *); 91 static void configure_final (void *); 92 93 #if defined(FFS) && defined(FFS_ROOT) 94 static void setroot (void); 95 #endif 96 97 #if defined(NFS) && defined(NFS_ROOT) 98 #if !defined(BOOTP_NFSROOT) 99 static void pxe_setup_nfsdiskless(void); 100 #endif 101 #endif 102 103 SYSINIT(cpu, SI_BOOT2_START_CPU, SI_ORDER_FIRST, cpu_startup, NULL); 104 SYSINIT(configure1, SI_SUB_CONFIGURE, SI_ORDER_FIRST, configure_first, NULL); 105 /* SI_ORDER_SECOND is hookable */ 106 SYSINIT(configure2, SI_SUB_CONFIGURE, SI_ORDER_THIRD, configure, NULL); 107 /* SI_ORDER_MIDDLE is hookable */ 108 SYSINIT(configure3, SI_SUB_CONFIGURE, SI_ORDER_ANY, configure_final, NULL); 109 110 cdev_t rootdev = NULL; 111 cdev_t dumpdev = NULL; 112 113 /* 114 * nfsroot.iosize may be set in loader.conf, 32768 is recommended to 115 * be able to max-out a GigE link if the server supports it. Many servers 116 * do not so the default is 8192. 117 * 118 * nfsroot.rahead defaults to something reasonable, can be overridden. 119 */ 120 static int nfsroot_iosize = 8192; 121 TUNABLE_INT("nfsroot.iosize", &nfsroot_iosize); 122 static int nfsroot_rahead = 4; 123 TUNABLE_INT("nfsroot.rahead", &nfsroot_rahead); 124 125 /* 126 * 127 */ 128 static void 129 cpu_startup(void *dummy) 130 { 131 vm_offset_t buffer_sva; 132 vm_offset_t buffer_eva; 133 vm_offset_t pager_sva; 134 vm_offset_t pager_eva; 135 136 kprintf("%s", version); 137 kprintf("real memory = %ju (%juK bytes)\n", 138 (uintmax_t)ptoa(Maxmem), (uintmax_t)(ptoa(Maxmem) / 1024)); 139 140 if (nbuf == 0) { 141 int factor = 4 * BKVASIZE / 1024; 142 int kbytes = Maxmem * (PAGE_SIZE / 1024); 143 144 nbuf = 50; 145 if (kbytes > 4096) 146 nbuf += min((kbytes - 4096) / factor, 65536 / factor); 147 if (kbytes > 65536) 148 nbuf += (kbytes - 65536) * 2 / (factor * 5); 149 if (maxbcache && nbuf > maxbcache / BKVASIZE) 150 nbuf = maxbcache / BKVASIZE; 151 } 152 if (nbuf > (virtual_end - virtual_start) / (BKVASIZE * 2)) { 153 nbuf = (virtual_end - virtual_start) / (BKVASIZE * 2); 154 kprintf("Warning: nbufs capped at %ld\n", nbuf); 155 } 156 157 nswbuf = lmax(lmin(nbuf / 4, 256), 16); 158 #ifdef NSWBUF_MIN 159 if (nswbuf < NSWBUF_MIN) 160 nswbuf = NSWBUF_MIN; 161 #endif 162 163 /* 164 * Allocate memory for the buffer cache 165 */ 166 buf = (void *)kmem_alloc(&kernel_map, nbuf * sizeof(struct buf)); 167 swbuf = (void *)kmem_alloc(&kernel_map, nswbuf * sizeof(struct buf)); 168 169 170 #ifdef DIRECTIO 171 ffs_rawread_setup(); 172 #endif 173 kmem_suballoc(&kernel_map, &clean_map, &clean_sva, &clean_eva, 174 (nbuf*BKVASIZE*2) + (nswbuf*MAXPHYS) + pager_map_size); 175 kmem_suballoc(&clean_map, &buffer_map, &buffer_sva, &buffer_eva, 176 (nbuf*BKVASIZE*2)); 177 buffer_map.system_map = 1; 178 kmem_suballoc(&clean_map, &pager_map, &pager_sva, &pager_eva, 179 (nswbuf*MAXPHYS) + pager_map_size); 180 pager_map.system_map = 1; 181 kprintf("avail memory = %lu (%luK bytes)\n", ptoa(vmstats.v_free_count), 182 ptoa(vmstats.v_free_count) / 1024); 183 mp_start(); 184 mp_announce(); 185 cpu_setregs(); 186 } 187 188 /* 189 * Determine i/o configuration for a machine. 190 */ 191 static void 192 configure_first(void *dummy) 193 { 194 } 195 196 static void 197 configure(void *dummy) 198 { 199 /* 200 * Final interrupt support acviation, then enable hardware interrupts. 201 */ 202 MachIntrABI.finalize(); 203 cpu_enable_intr(); 204 205 /* 206 * This will configure all devices, generally starting with the 207 * nexus (i386/i386/nexus.c). The nexus ISA code explicitly 208 * dummies up the attach in order to delay legacy initialization 209 * until after all other busses/subsystems have had a chance 210 * at those resources. 211 */ 212 root_bus_configure(); 213 214 #if NISA > 0 215 /* 216 * Explicitly probe and attach ISA last. The isa bus saves 217 * it's device node at attach time for us here. 218 */ 219 if (isa_bus_device) 220 isa_probe_children(isa_bus_device); 221 #endif 222 223 /* 224 * Allow lowering of the ipl to the lowest kernel level if we 225 * panic (or call tsleep() before clearing `cold'). No level is 226 * completely safe (since a panic may occur in a critical region 227 * at splhigh()), but we want at least bio interrupts to work. 228 */ 229 safepri = TDPRI_KERN_USER; 230 } 231 232 static void 233 configure_final(void *dummy) 234 { 235 cninit_finish(); 236 237 if (bootverbose) 238 kprintf("Device configuration finished.\n"); 239 } 240 241 #ifdef BOOTP 242 void bootpc_init(void); 243 #endif 244 /* 245 * Do legacy root filesystem discovery. 246 */ 247 void 248 cpu_rootconf(void) 249 { 250 #ifdef BOOTP 251 bootpc_init(); 252 #endif 253 #if defined(NFS) && defined(NFS_ROOT) 254 #if !defined(BOOTP_NFSROOT) 255 pxe_setup_nfsdiskless(); 256 if (nfs_diskless_valid) 257 #endif 258 rootdevnames[0] = "nfs:"; 259 #endif 260 #if defined(FFS) && defined(FFS_ROOT) 261 if (!rootdevnames[0]) 262 setroot(); 263 #endif 264 } 265 SYSINIT(cpu_rootconf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, cpu_rootconf, NULL); 266 267 u_long bootdev = 0; /* not a cdev_t - encoding is different */ 268 269 #if defined(FFS) && defined(FFS_ROOT) 270 271 /* 272 * The boot code uses old block device major numbers to pass bootdev to 273 * us. We have to translate these to character device majors because 274 * we don't have block devices any more. 275 */ 276 static int 277 boot_translate_majdev(int bmajor) 278 { 279 static int conv[] = { BOOTMAJOR_CONVARY }; 280 281 if (bmajor >= 0 && bmajor < NELEM(conv)) 282 return(conv[bmajor]); 283 return(-1); 284 } 285 286 /* 287 * Attempt to find the device from which we were booted. 288 * If we can do so, and not instructed not to do so, 289 * set rootdevs[] and rootdevnames[] to correspond to the 290 * boot device(s). 291 * 292 * This code survives in order to allow the system to be 293 * booted from legacy environments that do not correctly 294 * populate the kernel environment. There are significant 295 * restrictions on the bootability of the system in this 296 * situation; it can only be mounting root from a 'da' 297 * 'wd' or 'fd' device, and the root filesystem must be ufs. 298 */ 299 static void 300 setroot(void) 301 { 302 int majdev, mindev, unit, slice, part; 303 cdev_t newrootdev, dev; 304 char partname[2]; 305 char *sname; 306 307 if ((bootdev & B_MAGICMASK) != B_DEVMAGIC) { 308 kprintf("no B_DEVMAGIC (bootdev=%#lx)\n", bootdev); 309 return; 310 } 311 majdev = boot_translate_majdev(B_TYPE(bootdev)); 312 if (bootverbose) { 313 kprintf("bootdev: %08lx type=%ld unit=%ld " 314 "slice=%ld part=%ld major=%d\n", 315 bootdev, B_TYPE(bootdev), B_UNIT(bootdev), 316 B_SLICE(bootdev), B_PARTITION(bootdev), majdev); 317 } 318 dev = udev2dev(makeudev(majdev, 0), 0); 319 if (!dev_is_good(dev)) 320 return; 321 unit = B_UNIT(bootdev); 322 slice = B_SLICE(bootdev); 323 if (slice == WHOLE_DISK_SLICE) 324 slice = COMPATIBILITY_SLICE; 325 if (slice < 0 || slice >= MAX_SLICES) { 326 kprintf("bad slice\n"); 327 return; 328 } 329 330 part = B_PARTITION(bootdev); 331 mindev = dkmakeminor(unit, slice, part); 332 newrootdev = udev2dev(makeudev(majdev, mindev), 0); 333 if (!dev_is_good(newrootdev)) 334 return; 335 sname = dsname(newrootdev, unit, slice, part, partname); 336 rootdevnames[0] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK); 337 ksprintf(rootdevnames[0], "ufs:%s%s", sname, partname); 338 339 /* 340 * For properly dangerously dedicated disks (ones with a historical 341 * bogus partition table), the boot blocks will give slice = 4, but 342 * the kernel will only provide the compatibility slice since it 343 * knows that slice 4 is not a real slice. Arrange to try mounting 344 * the compatibility slice as root if mounting the slice passed by 345 * the boot blocks fails. This handles the dangerously dedicated 346 * case and perhaps others. 347 */ 348 if (slice == COMPATIBILITY_SLICE) 349 return; 350 slice = COMPATIBILITY_SLICE; 351 sname = dsname(newrootdev, unit, slice, part, partname); 352 rootdevnames[1] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK); 353 ksprintf(rootdevnames[1], "ufs:%s%s", sname, partname); 354 } 355 #endif 356 357 #if defined(NFS) && defined(NFS_ROOT) 358 #if !defined(BOOTP_NFSROOT) 359 360 #include <sys/socket.h> 361 #include <net/if.h> 362 #include <net/if_dl.h> 363 #include <net/if_types.h> 364 #include <net/if_var.h> 365 #include <net/ethernet.h> 366 #include <netinet/in.h> 367 #include <vfs/nfs/rpcv2.h> 368 #include <vfs/nfs/nfsproto.h> 369 #include <vfs/nfs/nfs.h> 370 #include <vfs/nfs/nfsdiskless.h> 371 372 extern struct nfs_diskless nfs_diskless; 373 374 /* 375 * Convert a kenv variable to a sockaddr. If the kenv variable does not 376 * exist the sockaddr will remain zerod out (callers typically just check 377 * sin_len). A network address of 0.0.0.0 is equivalent to failure. 378 */ 379 static int 380 inaddr_to_sockaddr(char *ev, struct sockaddr_in *sa) 381 { 382 u_int32_t a[4]; 383 char *cp; 384 385 bzero(sa, sizeof(*sa)); 386 387 if ((cp = kgetenv(ev)) == NULL) 388 return(1); 389 if (ksscanf(cp, "%d.%d.%d.%d", &a[0], &a[1], &a[2], &a[3]) != 4) 390 return(1); 391 if (a[0] == 0 && a[1] == 0 && a[2] == 0 && a[3] == 0) 392 return(1); 393 /* XXX is this ordering correct? */ 394 sa->sin_addr.s_addr = (a[3] << 24) + (a[2] << 16) + (a[1] << 8) + a[0]; 395 sa->sin_len = sizeof(*sa); 396 sa->sin_family = AF_INET; 397 return(0); 398 } 399 400 static int 401 hwaddr_to_sockaddr(char *ev, struct sockaddr_dl *sa) 402 { 403 char *cp; 404 u_int32_t a[6]; 405 406 bzero(sa, sizeof(*sa)); 407 sa->sdl_len = sizeof(*sa); 408 sa->sdl_family = AF_LINK; 409 sa->sdl_type = IFT_ETHER; 410 sa->sdl_alen = ETHER_ADDR_LEN; 411 if ((cp = kgetenv(ev)) == NULL) 412 return(1); 413 if (ksscanf(cp, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], &a[4], &a[5]) != 6) 414 return(1); 415 sa->sdl_data[0] = a[0]; 416 sa->sdl_data[1] = a[1]; 417 sa->sdl_data[2] = a[2]; 418 sa->sdl_data[3] = a[3]; 419 sa->sdl_data[4] = a[4]; 420 sa->sdl_data[5] = a[5]; 421 return(0); 422 } 423 424 static int 425 decode_nfshandle(char *ev, u_char *fh) 426 { 427 u_char *cp; 428 int len, val; 429 430 if (((cp = kgetenv(ev)) == NULL) || (strlen(cp) < 2) || (*cp != 'X')) 431 return(0); 432 len = 0; 433 cp++; 434 for (;;) { 435 if (*cp == 'X') 436 return(len); 437 if ((ksscanf(cp, "%2x", &val) != 1) || (val > 0xff)) 438 return(0); 439 *(fh++) = val; 440 len++; 441 cp += 2; 442 if (len > NFSX_V2FH) 443 return(0); 444 } 445 } 446 447 /* 448 * Populate the essential fields in the nfsv3_diskless structure. 449 * 450 * The loader is expected to export the following environment variables: 451 * 452 * boot.netif.ip IP address on boot interface 453 * boot.netif.netmask netmask on boot interface 454 * boot.netif.gateway default gateway (optional) 455 * boot.netif.hwaddr hardware address of boot interface 456 * boot.netif.name name of boot interface (instead of hw addr) 457 * boot.nfsroot.server IP address of root filesystem server 458 * boot.nfsroot.path path of the root filesystem on server 459 * boot.nfsroot.nfshandle NFS handle for root filesystem on server 460 */ 461 static void 462 pxe_setup_nfsdiskless(void) 463 { 464 struct nfs_diskless *nd = &nfs_diskless; 465 struct ifnet *ifp; 466 struct ifaddr *ifa; 467 struct sockaddr_dl *sdl, ourdl; 468 struct sockaddr_in myaddr, netmask; 469 char *cp; 470 471 /* set up interface */ 472 if (inaddr_to_sockaddr("boot.netif.ip", &myaddr)) 473 return; 474 if (inaddr_to_sockaddr("boot.netif.netmask", &netmask)) { 475 kprintf("PXE: no netmask\n"); 476 return; 477 } 478 bcopy(&myaddr, &nd->myif.ifra_addr, sizeof(myaddr)); 479 bcopy(&myaddr, &nd->myif.ifra_broadaddr, sizeof(myaddr)); 480 ((struct sockaddr_in *) &nd->myif.ifra_broadaddr)->sin_addr.s_addr = 481 myaddr.sin_addr.s_addr | ~ netmask.sin_addr.s_addr; 482 bcopy(&netmask, &nd->myif.ifra_mask, sizeof(netmask)); 483 484 if ((cp = kgetenv("boot.netif.name")) != NULL) { 485 ifnet_lock(); 486 ifp = ifunit(cp); 487 if (ifp) { 488 strlcpy(nd->myif.ifra_name, ifp->if_xname, 489 sizeof(nd->myif.ifra_name)); 490 ifnet_unlock(); 491 goto match_done; 492 } 493 ifnet_unlock(); 494 kprintf("PXE: cannot find interface %s\n", cp); 495 return; 496 } 497 498 if (hwaddr_to_sockaddr("boot.netif.hwaddr", &ourdl)) { 499 kprintf("PXE: no hardware address\n"); 500 return; 501 } 502 ifa = NULL; 503 ifnet_lock(); 504 TAILQ_FOREACH(ifp, &ifnetlist, if_link) { 505 struct ifaddr_container *ifac; 506 507 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 508 ifa = ifac->ifa; 509 510 if ((ifa->ifa_addr->sa_family == AF_LINK) && 511 (sdl = ((struct sockaddr_dl *)ifa->ifa_addr))) { 512 if ((sdl->sdl_type == ourdl.sdl_type) && 513 (sdl->sdl_alen == ourdl.sdl_alen) && 514 !bcmp(sdl->sdl_data + sdl->sdl_nlen, 515 ourdl.sdl_data + ourdl.sdl_nlen, 516 sdl->sdl_alen)) { 517 strlcpy(nd->myif.ifra_name, 518 ifp->if_xname, 519 sizeof(nd->myif.ifra_name)); 520 ifnet_unlock(); 521 goto match_done; 522 } 523 } 524 } 525 } 526 ifnet_unlock(); 527 kprintf("PXE: no interface\n"); 528 return; /* no matching interface */ 529 match_done: 530 /* set up gateway */ 531 inaddr_to_sockaddr("boot.netif.gateway", &nd->mygateway); 532 533 /* XXX set up swap? */ 534 535 /* set up root mount */ 536 nd->root_args.rsize = nfsroot_iosize; 537 nd->root_args.wsize = nfsroot_iosize; 538 nd->root_args.sotype = SOCK_STREAM; 539 nd->root_args.readahead = nfsroot_rahead; 540 nd->root_args.flags = NFSMNT_WSIZE | NFSMNT_RSIZE | NFSMNT_RESVPORT | 541 NFSMNT_READAHEAD; 542 if (inaddr_to_sockaddr("boot.nfsroot.server", &nd->root_saddr)) { 543 kprintf("PXE: no server\n"); 544 return; 545 } 546 nd->root_saddr.sin_port = htons(NFS_PORT); 547 548 /* 549 * A tftp-only loader may pass NFS path information without a 550 * root handle. Generate a warning but continue configuring. 551 */ 552 if (decode_nfshandle("boot.nfsroot.nfshandle", &nd->root_fh[0]) == 0) { 553 kprintf("PXE: Warning, no NFS handle passed from loader\n"); 554 } 555 if ((cp = kgetenv("boot.nfsroot.path")) != NULL) 556 strncpy(nd->root_hostnam, cp, MNAMELEN - 1); 557 558 nfs_diskless_valid = 1; 559 } 560 561 #endif 562 #endif 563