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 * NBUFCALCSIZE / 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 / NBUFCALCSIZE) 150 nbuf = maxbcache / NBUFCALCSIZE; 151 } 152 if (nbuf > (virtual_end - virtual_start) / (MAXBSIZE * 2)) { 153 nbuf = (virtual_end - virtual_start) / (MAXBSIZE * 2); 154 kprintf("Warning: nbufs capped at %ld\n", nbuf); 155 } 156 157 nswbuf_mem = lmax(lmin(nbuf / 32, 32), 4); 158 #ifdef NSWBUF_MIN 159 if (nswbuf_mem < NSWBUF_MIN) 160 nswbuf_mem = NSWBUF_MIN; 161 #endif 162 nswbuf_kva = lmax(lmin(nbuf / 4, 256), 16); 163 #ifdef NSWBUF_MIN 164 if (nswbuf_kva < NSWBUF_MIN) 165 nswbuf_kva = NSWBUF_MIN; 166 #endif 167 168 /* 169 * Allocate memory for the buffer cache 170 */ 171 buf = (void *)kmem_alloc(&kernel_map, 172 nbuf * sizeof(struct buf), 173 VM_SUBSYS_BUF); 174 swbuf_mem = (void *)kmem_alloc(&kernel_map, 175 nswbuf_mem * sizeof(struct buf), 176 VM_SUBSYS_BUF); 177 swbuf_kva = (void *)kmem_alloc(&kernel_map, 178 nswbuf_kva * sizeof(struct buf), 179 VM_SUBSYS_BUF); 180 181 kmem_suballoc(&kernel_map, &clean_map, &clean_sva, &clean_eva, 182 (nbuf * MAXBSIZE * 2) + 183 (nswbuf_mem + nswbuf_kva) *MAXPHYS + 184 pager_map_size); 185 kmem_suballoc(&clean_map, &buffer_map, &buffer_sva, &buffer_eva, 186 (nbuf * MAXBSIZE * 2)); 187 buffer_map.system_map = 1; 188 kmem_suballoc(&clean_map, &pager_map, &pager_sva, &pager_eva, 189 (nswbuf_mem + nswbuf_kva) *MAXPHYS + 190 pager_map_size); 191 pager_map.system_map = 1; 192 kprintf("avail memory = %lu (%luK bytes)\n", ptoa(vmstats.v_free_count), 193 ptoa(vmstats.v_free_count) / 1024); 194 mp_start(); 195 mp_announce(); 196 cpu_setregs(); 197 } 198 199 /* 200 * Determine i/o configuration for a machine. 201 */ 202 static void 203 configure_first(void *dummy) 204 { 205 } 206 207 static void 208 configure(void *dummy) 209 { 210 /* 211 * Final interrupt support acviation, then enable hardware interrupts. 212 */ 213 MachIntrABI.finalize(); 214 cpu_enable_intr(); 215 216 /* 217 * This will configure all devices, generally starting with the 218 * nexus (i386/i386/nexus.c). The nexus ISA code explicitly 219 * dummies up the attach in order to delay legacy initialization 220 * until after all other busses/subsystems have had a chance 221 * at those resources. 222 */ 223 root_bus_configure(); 224 225 #if NISA > 0 226 /* 227 * Explicitly probe and attach ISA last. The isa bus saves 228 * it's device node at attach time for us here. 229 */ 230 if (isa_bus_device) 231 isa_probe_children(isa_bus_device); 232 #endif 233 234 /* 235 * Allow lowering of the ipl to the lowest kernel level if we 236 * panic (or call tsleep() before clearing `cold'). No level is 237 * completely safe (since a panic may occur in a critical region 238 * at splhigh()), but we want at least bio interrupts to work. 239 */ 240 safepri = TDPRI_KERN_USER; 241 } 242 243 static void 244 configure_final(void *dummy) 245 { 246 cninit_finish(); 247 248 if (bootverbose) 249 kprintf("Device configuration finished.\n"); 250 } 251 252 #ifdef BOOTP 253 void bootpc_init(void); 254 #endif 255 /* 256 * Do legacy root filesystem discovery. 257 */ 258 void 259 cpu_rootconf(void) 260 { 261 #ifdef BOOTP 262 bootpc_init(); 263 #endif 264 #if defined(NFS) && defined(NFS_ROOT) 265 #if !defined(BOOTP_NFSROOT) 266 pxe_setup_nfsdiskless(); 267 if (nfs_diskless_valid) 268 #endif 269 rootdevnames[0] = "nfs:"; 270 #endif 271 #if defined(FFS) && defined(FFS_ROOT) 272 if (!rootdevnames[0]) 273 setroot(); 274 #endif 275 } 276 SYSINIT(cpu_rootconf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, cpu_rootconf, NULL); 277 278 u_long bootdev = 0; /* not a cdev_t - encoding is different */ 279 280 #if defined(FFS) && defined(FFS_ROOT) 281 282 /* 283 * The boot code uses old block device major numbers to pass bootdev to 284 * us. We have to translate these to character device majors because 285 * we don't have block devices any more. 286 */ 287 static int 288 boot_translate_majdev(int bmajor) 289 { 290 static int conv[] = { BOOTMAJOR_CONVARY }; 291 292 if (bmajor >= 0 && bmajor < NELEM(conv)) 293 return(conv[bmajor]); 294 return(-1); 295 } 296 297 /* 298 * Attempt to find the device from which we were booted. 299 * If we can do so, and not instructed not to do so, 300 * set rootdevs[] and rootdevnames[] to correspond to the 301 * boot device(s). 302 * 303 * This code survives in order to allow the system to be 304 * booted from legacy environments that do not correctly 305 * populate the kernel environment. There are significant 306 * restrictions on the bootability of the system in this 307 * situation; it can only be mounting root from a 'da' 308 * 'wd' or 'fd' device, and the root filesystem must be ufs. 309 */ 310 static void 311 setroot(void) 312 { 313 int majdev, mindev, unit, slice, part; 314 cdev_t newrootdev, dev; 315 char partname[2]; 316 char *sname; 317 318 if ((bootdev & B_MAGICMASK) != B_DEVMAGIC) { 319 kprintf("no B_DEVMAGIC (bootdev=%#lx)\n", bootdev); 320 return; 321 } 322 majdev = boot_translate_majdev(B_TYPE(bootdev)); 323 if (bootverbose) { 324 kprintf("bootdev: %08lx type=%ld unit=%ld " 325 "slice=%ld part=%ld major=%d\n", 326 bootdev, B_TYPE(bootdev), B_UNIT(bootdev), 327 B_SLICE(bootdev), B_PARTITION(bootdev), majdev); 328 } 329 dev = udev2dev(makeudev(majdev, 0), 0); 330 if (!dev_is_good(dev)) 331 return; 332 unit = B_UNIT(bootdev); 333 slice = B_SLICE(bootdev); 334 if (slice == WHOLE_DISK_SLICE) 335 slice = COMPATIBILITY_SLICE; 336 if (slice < 0 || slice >= MAX_SLICES) { 337 kprintf("bad slice\n"); 338 return; 339 } 340 341 part = B_PARTITION(bootdev); 342 mindev = dkmakeminor(unit, slice, part); 343 newrootdev = udev2dev(makeudev(majdev, mindev), 0); 344 if (!dev_is_good(newrootdev)) 345 return; 346 sname = dsname(newrootdev, unit, slice, part, partname); 347 rootdevnames[0] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK); 348 ksprintf(rootdevnames[0], "ufs:%s%s", sname, partname); 349 350 /* 351 * For properly dangerously dedicated disks (ones with a historical 352 * bogus partition table), the boot blocks will give slice = 4, but 353 * the kernel will only provide the compatibility slice since it 354 * knows that slice 4 is not a real slice. Arrange to try mounting 355 * the compatibility slice as root if mounting the slice passed by 356 * the boot blocks fails. This handles the dangerously dedicated 357 * case and perhaps others. 358 */ 359 if (slice == COMPATIBILITY_SLICE) 360 return; 361 slice = COMPATIBILITY_SLICE; 362 sname = dsname(newrootdev, unit, slice, part, partname); 363 rootdevnames[1] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK); 364 ksprintf(rootdevnames[1], "ufs:%s%s", sname, partname); 365 } 366 #endif 367 368 #if defined(NFS) && defined(NFS_ROOT) 369 #if !defined(BOOTP_NFSROOT) 370 371 #include <sys/socket.h> 372 #include <net/if.h> 373 #include <net/if_dl.h> 374 #include <net/if_types.h> 375 #include <net/if_var.h> 376 #include <net/ethernet.h> 377 #include <netinet/in.h> 378 #include <vfs/nfs/rpcv2.h> 379 #include <vfs/nfs/nfsproto.h> 380 #include <vfs/nfs/nfs.h> 381 #include <vfs/nfs/nfsdiskless.h> 382 383 extern struct nfs_diskless nfs_diskless; 384 385 /* 386 * Convert a kenv variable to a sockaddr. If the kenv variable does not 387 * exist the sockaddr will remain zerod out (callers typically just check 388 * sin_len). A network address of 0.0.0.0 is equivalent to failure. 389 */ 390 static int 391 inaddr_to_sockaddr(char *ev, struct sockaddr_in *sa) 392 { 393 u_int32_t a[4]; 394 char *cp; 395 396 bzero(sa, sizeof(*sa)); 397 398 if ((cp = kgetenv(ev)) == NULL) 399 return(1); 400 if (ksscanf(cp, "%d.%d.%d.%d", &a[0], &a[1], &a[2], &a[3]) != 4) 401 return(1); 402 if (a[0] == 0 && a[1] == 0 && a[2] == 0 && a[3] == 0) 403 return(1); 404 /* XXX is this ordering correct? */ 405 sa->sin_addr.s_addr = (a[3] << 24) + (a[2] << 16) + (a[1] << 8) + a[0]; 406 sa->sin_len = sizeof(*sa); 407 sa->sin_family = AF_INET; 408 return(0); 409 } 410 411 static int 412 hwaddr_to_sockaddr(char *ev, struct sockaddr_dl *sa) 413 { 414 char *cp; 415 u_int32_t a[6]; 416 417 bzero(sa, sizeof(*sa)); 418 sa->sdl_len = sizeof(*sa); 419 sa->sdl_family = AF_LINK; 420 sa->sdl_type = IFT_ETHER; 421 sa->sdl_alen = ETHER_ADDR_LEN; 422 if ((cp = kgetenv(ev)) == NULL) 423 return(1); 424 if (ksscanf(cp, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], &a[4], &a[5]) != 6) 425 return(1); 426 sa->sdl_data[0] = a[0]; 427 sa->sdl_data[1] = a[1]; 428 sa->sdl_data[2] = a[2]; 429 sa->sdl_data[3] = a[3]; 430 sa->sdl_data[4] = a[4]; 431 sa->sdl_data[5] = a[5]; 432 return(0); 433 } 434 435 static int 436 decode_nfshandle(char *ev, u_char *fh) 437 { 438 u_char *cp; 439 int len, val; 440 441 if (((cp = kgetenv(ev)) == NULL) || (strlen(cp) < 2) || (*cp != 'X')) 442 return(0); 443 len = 0; 444 cp++; 445 for (;;) { 446 if (*cp == 'X') 447 return(len); 448 if ((ksscanf(cp, "%2x", &val) != 1) || (val > 0xff)) 449 return(0); 450 *(fh++) = val; 451 len++; 452 cp += 2; 453 if (len > NFSX_V2FH) 454 return(0); 455 } 456 } 457 458 /* 459 * Populate the essential fields in the nfsv3_diskless structure. 460 * 461 * The loader is expected to export the following environment variables: 462 * 463 * boot.netif.ip IP address on boot interface 464 * boot.netif.netmask netmask on boot interface 465 * boot.netif.gateway default gateway (optional) 466 * boot.netif.hwaddr hardware address of boot interface 467 * boot.netif.name name of boot interface (instead of hw addr) 468 * boot.nfsroot.server IP address of root filesystem server 469 * boot.nfsroot.path path of the root filesystem on server 470 * boot.nfsroot.nfshandle NFS handle for root filesystem on server 471 */ 472 static void 473 pxe_setup_nfsdiskless(void) 474 { 475 struct nfs_diskless *nd = &nfs_diskless; 476 struct ifnet *ifp; 477 struct ifaddr *ifa; 478 struct sockaddr_dl *sdl, ourdl; 479 struct sockaddr_in myaddr, netmask; 480 char *cp; 481 482 /* set up interface */ 483 if (inaddr_to_sockaddr("boot.netif.ip", &myaddr)) 484 return; 485 if (inaddr_to_sockaddr("boot.netif.netmask", &netmask)) { 486 kprintf("PXE: no netmask\n"); 487 return; 488 } 489 bcopy(&myaddr, &nd->myif.ifra_addr, sizeof(myaddr)); 490 bcopy(&myaddr, &nd->myif.ifra_broadaddr, sizeof(myaddr)); 491 ((struct sockaddr_in *) &nd->myif.ifra_broadaddr)->sin_addr.s_addr = 492 myaddr.sin_addr.s_addr | ~ netmask.sin_addr.s_addr; 493 bcopy(&netmask, &nd->myif.ifra_mask, sizeof(netmask)); 494 495 if ((cp = kgetenv("boot.netif.name")) != NULL) { 496 ifnet_lock(); 497 ifp = ifunit(cp); 498 if (ifp) { 499 strlcpy(nd->myif.ifra_name, ifp->if_xname, 500 sizeof(nd->myif.ifra_name)); 501 ifnet_unlock(); 502 goto match_done; 503 } 504 ifnet_unlock(); 505 kprintf("PXE: cannot find interface %s\n", cp); 506 return; 507 } 508 509 if (hwaddr_to_sockaddr("boot.netif.hwaddr", &ourdl)) { 510 kprintf("PXE: no hardware address\n"); 511 return; 512 } 513 ifa = NULL; 514 ifnet_lock(); 515 TAILQ_FOREACH(ifp, &ifnetlist, if_link) { 516 struct ifaddr_container *ifac; 517 518 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 519 ifa = ifac->ifa; 520 521 if ((ifa->ifa_addr->sa_family == AF_LINK) && 522 (sdl = ((struct sockaddr_dl *)ifa->ifa_addr))) { 523 if ((sdl->sdl_type == ourdl.sdl_type) && 524 (sdl->sdl_alen == ourdl.sdl_alen) && 525 !bcmp(sdl->sdl_data + sdl->sdl_nlen, 526 ourdl.sdl_data + ourdl.sdl_nlen, 527 sdl->sdl_alen)) { 528 strlcpy(nd->myif.ifra_name, 529 ifp->if_xname, 530 sizeof(nd->myif.ifra_name)); 531 ifnet_unlock(); 532 goto match_done; 533 } 534 } 535 } 536 } 537 ifnet_unlock(); 538 kprintf("PXE: no interface\n"); 539 return; /* no matching interface */ 540 match_done: 541 /* set up gateway */ 542 inaddr_to_sockaddr("boot.netif.gateway", &nd->mygateway); 543 544 /* XXX set up swap? */ 545 546 /* set up root mount */ 547 nd->root_args.rsize = nfsroot_iosize; 548 nd->root_args.wsize = nfsroot_iosize; 549 nd->root_args.sotype = SOCK_STREAM; 550 nd->root_args.readahead = nfsroot_rahead; 551 nd->root_args.flags = NFSMNT_WSIZE | NFSMNT_RSIZE | NFSMNT_RESVPORT | 552 NFSMNT_READAHEAD; 553 if (inaddr_to_sockaddr("boot.nfsroot.server", &nd->root_saddr)) { 554 kprintf("PXE: no server\n"); 555 return; 556 } 557 nd->root_saddr.sin_port = htons(NFS_PORT); 558 559 /* 560 * A tftp-only loader may pass NFS path information without a 561 * root handle. Generate a warning but continue configuring. 562 */ 563 if (decode_nfshandle("boot.nfsroot.nfshandle", &nd->root_fh[0]) == 0) { 564 kprintf("PXE: Warning, no NFS handle passed from loader\n"); 565 } 566 if ((cp = kgetenv("boot.nfsroot.path")) != NULL) 567 strncpy(nd->root_hostnam, cp, MNAMELEN - 1); 568 569 nfs_diskless_valid = 1; 570 } 571 572 #endif 573 #endif 574