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