sys/platform/vkernel64/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/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, nbuf * sizeof(struct buf));
 172         swbuf_mem = (void *)kmem_alloc(&kernel_map, nswbuf_mem * sizeof(struct buf));
 173         swbuf_kva = (void *)kmem_alloc(&kernel_map, nswbuf_kva * sizeof(struct buf));
 174
 175 #ifdef DIRECTIO
 176         ffs_rawread_setup();
 177 #endif
 178         kmem_suballoc(&kernel_map, &clean_map, &clean_sva, &clean_eva,
 179                       (nbuf * MAXBSIZE * 2) +
 180                       (nswbuf_mem + nswbuf_kva) *MAXPHYS +
 181                       pager_map_size);
 182         kmem_suballoc(&clean_map, &buffer_map, &buffer_sva, &buffer_eva,
 183                       (nbuf * MAXBSIZE * 2));
 184         buffer_map.system_map = 1;
 185         kmem_suballoc(&clean_map, &pager_map, &pager_sva, &pager_eva,
 186                       (nswbuf_mem + nswbuf_kva) *MAXPHYS +
 187                       pager_map_size);
 188         pager_map.system_map = 1;
 189         kprintf("avail memory = %lu (%luK bytes)\n", ptoa(vmstats.v_free_count),
 190                 ptoa(vmstats.v_free_count) / 1024);
 191         mp_start();
 192         mp_announce();
 193         cpu_setregs();
 194 }
 195
 196 /*
 197  * Determine i/o configuration for a machine.
 198  */
 199 static void
 200 configure_first(void *dummy)
 201 {
 202 }
 203
 204 static void
 205 configure(void *dummy)
 206 {
 207         /*
 208          * Final interrupt support acviation, then enable hardware interrupts.
 209          */
 210         MachIntrABI.finalize();
 211         cpu_enable_intr();
 212
 213         /*
 214          * This will configure all devices, generally starting with the
 215          * nexus (i386/i386/nexus.c).  The nexus ISA code explicitly
 216          * dummies up the attach in order to delay legacy initialization
 217          * until after all other busses/subsystems have had a chance
 218          * at those resources.
 219          */
 220         root_bus_configure();
 221
 222 #if NISA > 0
 223         /*
 224          * Explicitly probe and attach ISA last.  The isa bus saves
 225          * it's device node at attach time for us here.
 226          */
 227         if (isa_bus_device)
 228                 isa_probe_children(isa_bus_device);
 229 #endif
 230
 231         /*
 232          * Allow lowering of the ipl to the lowest kernel level if we
 233          * panic (or call tsleep() before clearing `cold').  No level is
 234          * completely safe (since a panic may occur in a critical region
 235          * at splhigh()), but we want at least bio interrupts to work.
 236          */
 237         safepri = TDPRI_KERN_USER;
 238 }
 239
 240 static void
 241 configure_final(void *dummy)
 242 {
 243         cninit_finish();
 244
 245         if (bootverbose)
 246                 kprintf("Device configuration finished.\n");
 247 }
 248
 249 #ifdef BOOTP
 250 void bootpc_init(void);
 251 #endif
 252 /*
 253  * Do legacy root filesystem discovery.
 254  */
 255 void
 256 cpu_rootconf(void)
 257 {
 258 #ifdef BOOTP
 259         bootpc_init();
 260 #endif
 261 #if defined(NFS) && defined(NFS_ROOT)
 262 #if !defined(BOOTP_NFSROOT)
 263         pxe_setup_nfsdiskless();
 264         if (nfs_diskless_valid)
 265 #endif
 266                 rootdevnames[0] = "nfs:";
 267 #endif
 268 #if defined(FFS) && defined(FFS_ROOT)
 269         if (!rootdevnames[0])
 270                 setroot();
 271 #endif
 272 }
 273 SYSINIT(cpu_rootconf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, cpu_rootconf, NULL);
 274
 275 u_long  bootdev = 0;            /* not a cdev_t - encoding is different */
 276
 277 #if defined(FFS) && defined(FFS_ROOT)
 278
 279 /*
 280  * The boot code uses old block device major numbers to pass bootdev to
 281  * us.  We have to translate these to character device majors because
 282  * we don't have block devices any more.
 283  */
 284 static int
 285 boot_translate_majdev(int bmajor)
 286 {
 287         static int conv[] = { BOOTMAJOR_CONVARY };
 288
 289         if (bmajor >= 0 && bmajor < NELEM(conv))
 290                 return(conv[bmajor]);
 291         return(-1);
 292 }
 293
 294 /*
 295  * Attempt to find the device from which we were booted.
 296  * If we can do so, and not instructed not to do so,
 297  * set rootdevs[] and rootdevnames[] to correspond to the
 298  * boot device(s).
 299  *
 300  * This code survives in order to allow the system to be
 301  * booted from legacy environments that do not correctly
 302  * populate the kernel environment. There are significant
 303  * restrictions on the bootability of the system in this
 304  * situation; it can only be mounting root from a 'da'
 305  * 'wd' or 'fd' device, and the root filesystem must be ufs.
 306  */
 307 static void
 308 setroot(void)
 309 {
 310         int majdev, mindev, unit, slice, part;
 311         cdev_t newrootdev, dev;
 312         char partname[2];
 313         char *sname;
 314
 315         if ((bootdev & B_MAGICMASK) != B_DEVMAGIC) {
 316                 kprintf("no B_DEVMAGIC (bootdev=%#lx)\n", bootdev);
 317                 return;
 318         }
 319         majdev = boot_translate_majdev(B_TYPE(bootdev));
 320         if (bootverbose) {
 321                 kprintf("bootdev: %08lx type=%ld unit=%ld "
 322                         "slice=%ld part=%ld major=%d\n",
 323                         bootdev, B_TYPE(bootdev), B_UNIT(bootdev),
 324                         B_SLICE(bootdev), B_PARTITION(bootdev), majdev);
 325         }
 326         dev = udev2dev(makeudev(majdev, 0), 0);
 327         if (!dev_is_good(dev))
 328                 return;
 329         unit = B_UNIT(bootdev);
 330         slice = B_SLICE(bootdev);
 331         if (slice == WHOLE_DISK_SLICE)
 332                 slice = COMPATIBILITY_SLICE;
 333         if (slice < 0 || slice >= MAX_SLICES) {
 334                 kprintf("bad slice\n");
 335                 return;
 336         }
 337
 338         part = B_PARTITION(bootdev);
 339         mindev = dkmakeminor(unit, slice, part);
 340         newrootdev = udev2dev(makeudev(majdev, mindev), 0);
 341         if (!dev_is_good(newrootdev))
 342                 return;
 343         sname = dsname(newrootdev, unit, slice, part, partname);
 344         rootdevnames[0] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
 345         ksprintf(rootdevnames[0], "ufs:%s%s", sname, partname);
 346
 347         /*
 348          * For properly dangerously dedicated disks (ones with a historical
 349          * bogus partition table), the boot blocks will give slice = 4, but
 350          * the kernel will only provide the compatibility slice since it
 351          * knows that slice 4 is not a real slice.  Arrange to try mounting
 352          * the compatibility slice as root if mounting the slice passed by
 353          * the boot blocks fails.  This handles the dangerously dedicated
 354          * case and perhaps others.
 355          */
 356         if (slice == COMPATIBILITY_SLICE)
 357                 return;
 358         slice = COMPATIBILITY_SLICE;
 359         sname = dsname(newrootdev, unit, slice, part, partname);
 360         rootdevnames[1] = kmalloc(strlen(sname) + 6, M_DEVBUF, M_WAITOK);
 361         ksprintf(rootdevnames[1], "ufs:%s%s", sname, partname);
 362 }
 363 #endif
 364
 365 #if defined(NFS) && defined(NFS_ROOT)
 366 #if !defined(BOOTP_NFSROOT)
 367
 368 #include <sys/socket.h>
 369 #include <net/if.h>
 370 #include <net/if_dl.h>
 371 #include <net/if_types.h>
 372 #include <net/if_var.h>
 373 #include <net/ethernet.h>
 374 #include <netinet/in.h>
 375 #include <vfs/nfs/rpcv2.h>
 376 #include <vfs/nfs/nfsproto.h>
 377 #include <vfs/nfs/nfs.h>
 378 #include <vfs/nfs/nfsdiskless.h>
 379
 380 extern struct nfs_diskless      nfs_diskless;
 381
 382 /*
 383  * Convert a kenv variable to a sockaddr.  If the kenv variable does not
 384  * exist the sockaddr will remain zerod out (callers typically just check
 385  * sin_len).  A network address of 0.0.0.0 is equivalent to failure.
 386  */
 387 static int
 388 inaddr_to_sockaddr(char *ev, struct sockaddr_in *sa)
 389 {
 390         u_int32_t       a[4];
 391         char            *cp;
 392
 393         bzero(sa, sizeof(*sa));
 394
 395         if ((cp = kgetenv(ev)) == NULL)
 396                 return(1);
 397         if (ksscanf(cp, "%d.%d.%d.%d", &a[0], &a[1], &a[2], &a[3]) != 4)
 398                 return(1);
 399         if (a[0] == 0 && a[1] == 0 && a[2] == 0 && a[3] == 0)
 400                 return(1);
 401         /* XXX is this ordering correct? */
 402         sa->sin_addr.s_addr = (a[3] << 24) + (a[2] << 16) + (a[1] << 8) + a[0];
 403         sa->sin_len = sizeof(*sa);
 404         sa->sin_family = AF_INET;
 405         return(0);
 406 }
 407
 408 static int
 409 hwaddr_to_sockaddr(char *ev, struct sockaddr_dl *sa)
 410 {
 411         char            *cp;
 412         u_int32_t       a[6];
 413
 414         bzero(sa, sizeof(*sa));
 415         sa->sdl_len = sizeof(*sa);
 416         sa->sdl_family = AF_LINK;
 417         sa->sdl_type = IFT_ETHER;
 418         sa->sdl_alen = ETHER_ADDR_LEN;
 419         if ((cp = kgetenv(ev)) == NULL)
 420                 return(1);
 421         if (ksscanf(cp, "%x:%x:%x:%x:%x:%x", &a[0], &a[1], &a[2], &a[3], &a[4], &a[5]) != 6)
 422                 return(1);
 423         sa->sdl_data[0] = a[0];
 424         sa->sdl_data[1] = a[1];
 425         sa->sdl_data[2] = a[2];
 426         sa->sdl_data[3] = a[3];
 427         sa->sdl_data[4] = a[4];
 428         sa->sdl_data[5] = a[5];
 429         return(0);
 430 }
 431
 432 static int
 433 decode_nfshandle(char *ev, u_char *fh)
 434 {
 435         u_char  *cp;
 436         int     len, val;
 437
 438         if (((cp = kgetenv(ev)) == NULL) || (strlen(cp) < 2) || (*cp != 'X'))
 439                 return(0);
 440         len = 0;
 441         cp++;
 442         for (;;) {
 443                 if (*cp == 'X')
 444                         return(len);
 445                 if ((ksscanf(cp, "%2x", &val) != 1) || (val > 0xff))
 446                         return(0);
 447                 *(fh++) = val;
 448                 len++;
 449                 cp += 2;
 450                 if (len > NFSX_V2FH)
 451                     return(0);
 452         }
 453 }
 454
 455 /*
 456  * Populate the essential fields in the nfsv3_diskless structure.
 457  *
 458  * The loader is expected to export the following environment variables:
 459  *
 460  * boot.netif.ip                IP address on boot interface
 461  * boot.netif.netmask           netmask on boot interface
 462  * boot.netif.gateway           default gateway (optional)
 463  * boot.netif.hwaddr            hardware address of boot interface
 464  * boot.netif.name              name of boot interface (instead of hw addr)
 465  * boot.nfsroot.server          IP address of root filesystem server
 466  * boot.nfsroot.path            path of the root filesystem on server
 467  * boot.nfsroot.nfshandle       NFS handle for root filesystem on server
 468  */
 469 static void
 470 pxe_setup_nfsdiskless(void)
 471 {
 472         struct nfs_diskless     *nd = &nfs_diskless;
 473         struct ifnet            *ifp;
 474         struct ifaddr           *ifa;
 475         struct sockaddr_dl      *sdl, ourdl;
 476         struct sockaddr_in      myaddr, netmask;
 477         char                    *cp;
 478
 479         /* set up interface */
 480         if (inaddr_to_sockaddr("boot.netif.ip", &myaddr))
 481                 return;
 482         if (inaddr_to_sockaddr("boot.netif.netmask", &netmask)) {
 483                 kprintf("PXE: no netmask\n");
 484                 return;
 485         }
 486         bcopy(&myaddr, &nd->myif.ifra_addr, sizeof(myaddr));
 487         bcopy(&myaddr, &nd->myif.ifra_broadaddr, sizeof(myaddr));
 488         ((struct sockaddr_in *) &nd->myif.ifra_broadaddr)->sin_addr.s_addr =
 489                 myaddr.sin_addr.s_addr | ~ netmask.sin_addr.s_addr;
 490         bcopy(&netmask, &nd->myif.ifra_mask, sizeof(netmask));
 491
 492         if ((cp = kgetenv("boot.netif.name")) != NULL) {
 493                 ifnet_lock();
 494                 ifp = ifunit(cp);
 495                 if (ifp) {
 496                         strlcpy(nd->myif.ifra_name, ifp->if_xname,
 497                             sizeof(nd->myif.ifra_name));
 498                         ifnet_unlock();
 499                         goto match_done;
 500                 }
 501                 ifnet_unlock();
 502                 kprintf("PXE: cannot find interface %s\n", cp);
 503                 return;
 504         }
 505
 506         if (hwaddr_to_sockaddr("boot.netif.hwaddr", &ourdl)) {
 507                 kprintf("PXE: no hardware address\n");
 508                 return;
 509         }
 510         ifa = NULL;
 511         ifnet_lock();
 512         TAILQ_FOREACH(ifp, &ifnetlist, if_link) {
 513                 struct ifaddr_container *ifac;
 514
 515                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 516                         ifa = ifac->ifa;
 517
 518                         if ((ifa->ifa_addr->sa_family == AF_LINK) &&
 519                             (sdl = ((struct sockaddr_dl *)ifa->ifa_addr))) {
 520                                 if ((sdl->sdl_type == ourdl.sdl_type) &&
 521                                     (sdl->sdl_alen == ourdl.sdl_alen) &&
 522                                     !bcmp(sdl->sdl_data + sdl->sdl_nlen,
 523                                           ourdl.sdl_data + ourdl.sdl_nlen,
 524                                           sdl->sdl_alen)) {
 525                                         strlcpy(nd->myif.ifra_name,
 526                                             ifp->if_xname,
 527                                             sizeof(nd->myif.ifra_name));
 528                                         ifnet_unlock();
 529                                         goto match_done;
 530                                 }
 531                         }
 532                 }
 533         }
 534         ifnet_unlock();
 535         kprintf("PXE: no interface\n");
 536         return; /* no matching interface */
 537 match_done:
 538         /* set up gateway */
 539         inaddr_to_sockaddr("boot.netif.gateway", &nd->mygateway);
 540
 541         /* XXX set up swap? */
 542
 543         /* set up root mount */
 544         nd->root_args.rsize = nfsroot_iosize;
 545         nd->root_args.wsize = nfsroot_iosize;
 546         nd->root_args.sotype = SOCK_STREAM;
 547         nd->root_args.readahead = nfsroot_rahead;
 548         nd->root_args.flags = NFSMNT_WSIZE | NFSMNT_RSIZE | NFSMNT_RESVPORT |
 549                               NFSMNT_READAHEAD;
 550         if (inaddr_to_sockaddr("boot.nfsroot.server", &nd->root_saddr)) {
 551                 kprintf("PXE: no server\n");
 552                 return;
 553         }
 554         nd->root_saddr.sin_port = htons(NFS_PORT);
 555
 556         /*
 557          * A tftp-only loader may pass NFS path information without a
 558          * root handle.  Generate a warning but continue configuring.
 559          */
 560         if (decode_nfshandle("boot.nfsroot.nfshandle", &nd->root_fh[0]) == 0) {
 561                 kprintf("PXE: Warning, no NFS handle passed from loader\n");
 562         }
 563         if ((cp = kgetenv("boot.nfsroot.path")) != NULL)
 564                 strncpy(nd->root_hostnam, cp, MNAMELEN - 1);
 565
 566         nfs_diskless_valid = 1;
 567 }
 568
 569 #endif
 570 #endif