sys/platform/pc64/x86_64/autoconf.c

   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/conf.h>
  59 #include <sys/diskslice.h>
  60 #include <sys/reboot.h>
  61 #include <sys/kernel.h>
  62 #include <sys/malloc.h>
  63 #include <sys/mount.h>
  64 #include <sys/cons.h>
  65 #include <sys/thread.h>
  66 #include <sys/device.h>
  67 #include <sys/machintr.h>
  68
  69 #include <machine/bootinfo.h>
  70 #include <machine/md_var.h>
  71 #include <machine/smp.h>
  72 #include <machine_base/icu/icu.h>
  73
  74 #include <machine/pcb.h>
  75 #include <machine/pcb_ext.h>
  76 #include <machine/globaldata.h>
  77
  78 #if NISA > 0
  79 #include <bus/isa/isavar.h>
  80
  81 device_t isa_bus_device = NULL;
  82 #endif
  83
  84 static void     configure_first (void *);
  85 static void     configure (void *);
  86 static void     configure_final (void *);
  87
  88 #if defined(FFS) && defined(FFS_ROOT)
  89 static void     setroot (void);
  90 #endif
  91
  92 #if defined(NFS) && defined(NFS_ROOT)
  93 #if !defined(BOOTP_NFSROOT)
  94 static void     pxe_setup_nfsdiskless(void);
  95 #endif
  96 #endif
  97
  98 SYSINIT(configure1, SI_SUB_CONFIGURE, SI_ORDER_FIRST, configure_first, NULL);
  99 /* SI_ORDER_SECOND is hookable */
 100 SYSINIT(configure2, SI_SUB_CONFIGURE, SI_ORDER_THIRD, configure, NULL);
 101 /* SI_ORDER_MIDDLE is hookable */
 102 SYSINIT(configure3, SI_SUB_CONFIGURE, SI_ORDER_ANY, configure_final, NULL);
 103
 104 cdev_t  rootdev = NULL;
 105 cdev_t  dumpdev = NULL;
 106
 107 /*
 108  * nfsroot.iosize may be set in loader.conf, 32768 is recommended to
 109  * be able to max-out a GigE link if the server supports it.  Many servers
 110  * do not so the default is 8192.
 111  *
 112  * nfsroot.rahead defaults to something reasonable, can be overridden.
 113  */
 114 static int nfsroot_iosize = 8192;
 115 TUNABLE_INT("nfsroot.iosize", &nfsroot_iosize);
 116 static int nfsroot_rahead = 4;
 117 TUNABLE_INT("nfsroot.rahead", &nfsroot_rahead);
 118
 119 /*
 120  * Determine i/o configuration for a machine.
 121  */
 122 static void
 123 configure_first(void *dummy)
 124 {
 125 }
 126
 127 static void
 128 configure(void *dummy)
 129 {
 130         /*
 131          * This will configure all devices, generally starting with the
 132          * nexus (i386/i386/nexus.c).  The nexus ISA code explicitly
 133          * dummies up the attach in order to delay legacy initialization
 134          * until after all other busses/subsystems have had a chance
 135          * at those resources.
 136          */
 137         root_bus_configure();
 138
 139 #if NISA > 0
 140         /*
 141          * Explicitly probe and attach ISA last.  The isa bus saves
 142          * it's device node at attach time for us here.
 143          */
 144         if (isa_bus_device)
 145                 isa_probe_children(isa_bus_device);
 146 #endif
 147
 148         /*
 149          * Allow lowering of the ipl to the lowest kernel level if we
 150          * panic (or call tsleep() before clearing `cold').  No level is
 151          * completely safe (since a panic may occur in a critical region
 152          * at splhigh()), but we want at least bio interrupts to work.
 153          */
 154         safepri = TDPRI_KERN_USER;
 155 }
 156
 157 static void
 158 configure_final(void *dummy)
 159 {
 160         cninit_finish();
 161
 162         if (bootverbose) {
 163 #if 0 /* JG */
 164                 /*
 165                  * Print out the BIOS's idea of the disk geometries.
 166                  */
 167                 int i;
 168                 kprintf("BIOS Geometries:\n");
 169                 for (i = 0; i < N_BIOS_GEOM; i++) {
 170                         unsigned long bios_geom;
 171                         int max_cylinder, max_head, max_sector;
 172
 173                         bios_geom = bootinfo.bi_bios_geom[i];
 174
 175                         /*
 176                          * XXX the bootstrap punts a 1200K floppy geometry
 177                          * when the get-disk-geometry interrupt fails.  Skip
 178                          * drives that have this geometry.
 179                          */
 180                         if (bios_geom == 0x4f010f)
 181                                 continue;
 182
 183                         kprintf(" %x:%08lx ", i, bios_geom);
 184                         max_cylinder = bios_geom >> 16;
 185                         max_head = (bios_geom >> 8) & 0xff;
 186                         max_sector = bios_geom & 0xff;
 187                         kprintf(
 188                 "0..%d=%d cylinders, 0..%d=%d heads, 1..%d=%d sectors\n",
 189                                max_cylinder, max_cylinder + 1,
 190                                max_head, max_head + 1,
 191                                max_sector, max_sector);
 192                 }
 193                 kprintf(" %d accounted for\n", bootinfo.bi_n_bios_used);
 194
 195                 kprintf("Device configuration finished.\n");
 196 #endif
 197         }
 198 }
 199
 200 #ifdef BOOTP
 201 void bootpc_init(void);
 202 #endif
 203 /*
 204  * Do legacy root filesystem discovery.
 205  */
 206 void
 207 cpu_rootconf(void)
 208 {
 209 #ifdef BOOTP
 210         bootpc_init();
 211 #endif
 212 #if defined(NFS) && defined(NFS_ROOT)
 213 #if !defined(BOOTP_NFSROOT)
 214         pxe_setup_nfsdiskless();
 215         if (nfs_diskless_valid)
 216 #endif
 217                 rootdevnames[0] = "nfs:";
 218 #endif
 219 #if defined(FFS) && defined(FFS_ROOT)
 220         if (!rootdevnames[0])
 221                 setroot();
 222 #endif
 223 }
 224 SYSINIT(cpu_rootconf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, cpu_rootconf, NULL);
 225
 226 u_long  bootdev = 0;            /* not a cdev_t - encoding is different */
 227
 228 #if defined(FFS) && defined(FFS_ROOT)
 229 #define FDMAJOR         2
 230 #define FDUNITSHIFT     6
 231
 232 /*
 233  * The boot code uses old block device major numbers to pass bootdev to
 234  * us.  We have to translate these to character device majors because
 235  * we don't have block devices any more.
 236  */
 237 static int
 238 boot_translate_majdev(int bmajor)
 239 {
 240         static int conv[] = { BOOTMAJOR_CONVARY };
 241
 242         if (bmajor >= 0 && bmajor < NELEM(conv))
 243                 return(conv[bmajor]);
 244         return(-1);
 245 }
 246
 247 /*
 248  * Attempt to find the device from which we were booted.
 249  * If we can do so, and not instructed not to do so,
 250  * set rootdevs[] and rootdevnames[] to correspond to the
 251  * boot device(s).
 252  *
 253  * This code survives in order to allow the system to be
 254  * booted from legacy environments that do not correctly
 255  * populate the kernel environment. There are significant
 256  * restrictions on the bootability of the system in this
 257  * situation; it can only be mounting root from a 'da'
 258  * 'wd' or 'fd' device, and the root filesystem must be ufs.
 259  */
 260 static void
 261 setroot(void)
 262 {
 263         int majdev, mindev, unit, slice, part;
 264         cdev_t newrootdev, dev;
 265         char partname[2];
 266         char *sname;
 267
 268         if ((bootdev & B_MAGICMASK) != B_DEVMAGIC) {
 269                 kprintf("no B_DEVMAGIC (bootdev=%#lx)\n", bootdev);
 270                 return;
 271         }
 272         majdev = boot_translate_majdev(B_TYPE(bootdev));
 273         if (bootverbose) {
 274                 kprintf("bootdev: %08lx type=%ld unit=%ld "
 275                         "slice=%ld part=%ld major=%d\n",
 276                         bootdev, B_TYPE(bootdev), B_UNIT(bootdev),
 277                         B_SLICE(bootdev), B_PARTITION(bootdev), majdev);
 278         }
 279         dev = udev2dev(makeudev(majdev, 0), 0);
 280         if (!dev_is_good(dev))
 281                 return;
 282         unit = B_UNIT(bootdev);
 283         slice = B_SLICE(bootdev);
 284         if (slice == WHOLE_DISK_SLICE)
 285                 slice = COMPATIBILITY_SLICE;
 286         if (slice < 0 || slice >= MAX_SLICES) {
 287                 kprintf("bad slice\n");
 288                 return;
 289         }
 290
 291         part = B_PARTITION(bootdev);
 292         mindev = dkmakeminor(unit, slice, part);
 293         newrootdev = udev2dev(makeudev(majdev, mindev), 0);
 294         if (!dev_is_good(newrootdev))
 295                 return;
 296         sname = dsname(newrootdev, unit, slice, part, partname);
 297         rootdevnames[0] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
 298         ksprintf(rootdevnames[0], "ufs:%s%s", sname, partname);
 299
 300         /*
 301          * For properly dangerously dedicated disks (ones with a historical
 302          * bogus partition table), the boot blocks will give slice = 4, but
 303          * the kernel will only provide the compatibility slice since it
 304          * knows that slice 4 is not a real slice.  Arrange to try mounting
 305          * the compatibility slice as root if mounting the slice passed by
 306          * the boot blocks fails.  This handles the dangerously dedicated
 307          * case and perhaps others.
 308          */
 309         if (slice == COMPATIBILITY_SLICE)
 310                 return;
 311         slice = COMPATIBILITY_SLICE;
 312         sname = dsname(newrootdev, unit, slice, part, partname);
 313         rootdevnames[1] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
 314         ksprintf(rootdevnames[1], "ufs:%s%s", sname, partname);
 315 }
 316 #endif
 317
 318 #if defined(NFS) && defined(NFS_ROOT)
 319 #if !defined(BOOTP_NFSROOT)
 320
 321 #include <sys/socket.h>
 322 #include <net/if.h>
 323 #include <net/if_dl.h>
 324 #include <net/if_types.h>
 325 #include <net/if_var.h>
 326 #include <net/ethernet.h>
 327 #include <netinet/in.h>
 328 #include <vfs/nfs/rpcv2.h>
 329 #include <vfs/nfs/nfsproto.h>
 330 #include <vfs/nfs/nfs.h>
 331 #include <vfs/nfs/nfsdiskless.h>
 332
 333 extern struct nfs_diskless      nfs_diskless;
 334
 335 /*
 336  * Convert a kenv variable to a sockaddr.  If the kenv variable does not
 337  * exist the sockaddr will remain zerod out (callers typically just check
 338  * sin_len).  A network address of 0.0.0.0 is equivalent to failure.
 339  */
 340 static int
 341 inaddr_to_sockaddr(char *ev, struct sockaddr_in *sa)
 342 {
 343         u_int32_t       a[4];
 344         char            *cp;
 345
 346         bzero(sa, sizeof(*sa));
 347
 348         if ((cp = kgetenv(ev)) == NULL)
 349                 return(1);
 350         if (ksscanf(cp, "%d.%d.%d.%d", &a[0], &a[1], &a[2], &a[3]) != 4)
 351                 return(1);
 352         if (a[0] == 0 && a[1] == 0 && a[2] == 0 && a[3] == 0)
 353                 return(1);
 354         /* XXX is this ordering correct? */
 355         sa->sin_addr.s_addr = (a[3] << 24) + (a[2] << 16) + (a[1] << 8) + a[0];
 356         sa->sin_len = sizeof(*sa);
 357         sa->sin_family = AF_INET;
 358         return(0);
 359 }
 360
 361 static int
 362 hwaddr_to_sockaddr(char *ev, struct sockaddr_dl *sa)
 363 {
 364         char            *cp;
 365         u_int32_t       a[6];
 366
 367         bzero(sa, sizeof(*sa));
 368         sa->sdl_len = sizeof(*sa);
 369         sa->sdl_family = AF_LINK;
 370         sa->sdl_type = IFT_ETHER;
 371         sa->sdl_alen = ETHER_ADDR_LEN;
 372         if ((cp = kgetenv(ev)) == NULL)
 373                 return(1);
 374         if (ksscanf(cp, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], &a[4], &a[5]) != 6)
 375                 return(1);
 376         sa->sdl_data[0] = a[0];
 377         sa->sdl_data[1] = a[1];
 378         sa->sdl_data[2] = a[2];
 379         sa->sdl_data[3] = a[3];
 380         sa->sdl_data[4] = a[4];
 381         sa->sdl_data[5] = a[5];
 382         return(0);
 383 }
 384
 385 static int
 386 decode_nfshandle(char *ev, u_char *fh)
 387 {
 388         u_char  *cp;
 389         int     len, val;
 390
 391         if (((cp = kgetenv(ev)) == NULL) || (strlen(cp) < 2) || (*cp != 'X'))
 392                 return(0);
 393         len = 0;
 394         cp++;
 395         for (;;) {
 396                 if (*cp == 'X')
 397                         return(len);
 398                 if ((ksscanf(cp, "%2x", &val) != 1) || (val > 0xff))
 399                         return(0);
 400                 *(fh++) = val;
 401                 len++;
 402                 cp += 2;
 403                 if (len > NFSX_V2FH)
 404                     return(0);
 405         }
 406 }
 407
 408 /*
 409  * Populate the essential fields in the nfsv3_diskless structure.
 410  *
 411  * The loader is expected to export the following environment variables:
 412  *
 413  * boot.netif.ip                IP address on boot interface
 414  * boot.netif.netmask           netmask on boot interface
 415  * boot.netif.gateway           default gateway (optional)
 416  * boot.netif.hwaddr            hardware address of boot interface
 417  * boot.nfsroot.server          IP address of root filesystem server
 418  * boot.nfsroot.path            path of the root filesystem on server
 419  * boot.nfsroot.nfshandle       NFS handle for root filesystem on server
 420  */
 421 static void
 422 pxe_setup_nfsdiskless(void)
 423 {
 424         struct nfs_diskless     *nd = &nfs_diskless;
 425         struct ifnet            *ifp;
 426         struct sockaddr_dl      *sdl, ourdl;
 427         struct sockaddr_in      myaddr, netmask;
 428         char                    *cp;
 429
 430         /* set up interface */
 431         if (inaddr_to_sockaddr("boot.netif.ip", &myaddr))
 432                 return;
 433         if (inaddr_to_sockaddr("boot.netif.netmask", &netmask)) {
 434                 kprintf("PXE: no netmask\n");
 435                 return;
 436         }
 437         bcopy(&myaddr, &nd->myif.ifra_addr, sizeof(myaddr));
 438         bcopy(&myaddr, &nd->myif.ifra_broadaddr, sizeof(myaddr));
 439         ((struct sockaddr_in *) &nd->myif.ifra_broadaddr)->sin_addr.s_addr =
 440                 myaddr.sin_addr.s_addr | ~ netmask.sin_addr.s_addr;
 441         bcopy(&netmask, &nd->myif.ifra_mask, sizeof(netmask));
 442
 443         if (hwaddr_to_sockaddr("boot.netif.hwaddr", &ourdl)) {
 444                 kprintf("PXE: no hardware address\n");
 445                 return;
 446         }
 447         ifnet_lock();
 448         TAILQ_FOREACH(ifp, &ifnetlist, if_link) {
 449                 struct ifaddr_container *ifac;
 450
 451                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 452                         struct ifaddr *ifa = ifac->ifa;
 453
 454                         if ((ifa->ifa_addr->sa_family == AF_LINK) &&
 455                             (sdl = ((struct sockaddr_dl *)ifa->ifa_addr))) {
 456                                 if ((sdl->sdl_type == ourdl.sdl_type) &&
 457                                     (sdl->sdl_alen == ourdl.sdl_alen) &&
 458                                     !bcmp(sdl->sdl_data + sdl->sdl_nlen,
 459                                           ourdl.sdl_data + ourdl.sdl_nlen,
 460                                           sdl->sdl_alen)) {
 461                                         strlcpy(nd->myif.ifra_name,
 462                                             ifp->if_xname,
 463                                             sizeof(nd->myif.ifra_name));
 464                                         ifnet_unlock();
 465                                         goto match_done;
 466                                 }
 467                         }
 468                 }
 469         }
 470         ifnet_unlock();
 471         kprintf("PXE: no interface\n");
 472         return; /* no matching interface */
 473 match_done:
 474         /* set up gateway */
 475         inaddr_to_sockaddr("boot.netif.gateway", &nd->mygateway);
 476
 477         /* XXX set up swap? */
 478
 479         /* set up root mount */
 480         nd->root_args.rsize = nfsroot_iosize;
 481         nd->root_args.wsize = nfsroot_iosize;
 482         nd->root_args.sotype = SOCK_STREAM;
 483         nd->root_args.readahead = nfsroot_rahead;
 484         nd->root_args.flags = NFSMNT_WSIZE | NFSMNT_RSIZE | NFSMNT_RESVPORT |
 485                               NFSMNT_READAHEAD;
 486         if (inaddr_to_sockaddr("boot.nfsroot.server", &nd->root_saddr)) {
 487                 kprintf("PXE: no server\n");
 488                 return;
 489         }
 490         nd->root_saddr.sin_port = htons(NFS_PORT);
 491
 492         /*
 493          * A tftp-only loader may pass NFS path information without a
 494          * root handle.  Generate a warning but continue configuring.
 495          */
 496         if (decode_nfshandle("boot.nfsroot.nfshandle", &nd->root_fh[0]) == 0) {
 497                 kprintf("PXE: Warning, no NFS handle passed from loader\n");
 498         }
 499         if ((cp = kgetenv("boot.nfsroot.path")) != NULL)
 500                 strncpy(nd->root_hostnam, cp, MNAMELEN - 1);
 501
 502         nfs_diskless_valid = 1;
 503 }
 504
 505 #endif
 506 #endif