sys/platform/vkernel/platform/init.c

   1 /*
   2  * Copyright (c) 2006 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  * $DragonFly: src/sys/platform/vkernel/platform/init.c,v 1.37 2007/05/27 18:37:25 dillon Exp $
  35  */
  36
  37 #include <sys/types.h>
  38 #include <sys/systm.h>
  39 #include <sys/kernel.h>
  40 #include <sys/stat.h>
  41 #include <sys/mman.h>
  42 #include <sys/cons.h>
  43 #include <sys/random.h>
  44 #include <sys/vkernel.h>
  45 #include <sys/tls.h>
  46 #include <sys/reboot.h>
  47 #include <sys/proc.h>
  48 #include <sys/msgbuf.h>
  49 #include <sys/vmspace.h>
  50 #include <sys/socket.h>
  51 #include <sys/sockio.h>
  52 #include <vm/vm_page.h>
  53
  54 #include <machine/globaldata.h>
  55 #include <machine/tls.h>
  56 #include <machine/md_var.h>
  57 #include <machine/vmparam.h>
  58
  59 #include <net/if.h>
  60 #include <net/if_arp.h>
  61 #include <net/ethernet.h>
  62 #include <net/bridge/if_bridgevar.h>
  63 #include <netinet/in.h>
  64 #include <arpa/inet.h>
  65
  66 #include <stdio.h>
  67 #include <stdlib.h>
  68 #include <unistd.h>
  69 #include <fcntl.h>
  70 #include <string.h>
  71 #include <err.h>
  72 #include <errno.h>
  73 #include <assert.h>
  74
  75 vm_paddr_t phys_avail[16];
  76 vm_paddr_t Maxmem;
  77 vm_paddr_t Maxmem_bytes;
  78 int MemImageFd = -1;
  79 struct vkdisk_info DiskInfo[VKDISK_MAX];
  80 int DiskNum;
  81 struct vknetif_info NetifInfo[VKNETIF_MAX];
  82 int NetifNum;
  83 vm_offset_t KvaStart;
  84 vm_offset_t KvaEnd;
  85 vm_offset_t KvaSize;
  86 vm_offset_t virtual_start;
  87 vm_offset_t virtual_end;
  88 vm_offset_t kernel_vm_end;
  89 vm_offset_t crashdumpmap;
  90 vm_offset_t clean_sva;
  91 vm_offset_t clean_eva;
  92 struct msgbuf *msgbufp;
  93 caddr_t ptvmmap;
  94 vpte_t  *KernelPTD;
  95 vpte_t  *KernelPTA;     /* Warning: Offset for direct VA translation */
  96 u_int cpu_feature;      /* XXX */
  97 u_int tsc_present;      /* XXX */
  98
  99 struct privatespace *CPU_prvspace;
 100
 101 static struct trapframe proc0_tf;
 102 static void *proc0paddr;
 103
 104 static void init_sys_memory(char *imageFile);
 105 static void init_kern_memory(void);
 106 static void init_globaldata(void);
 107 static void init_vkernel(void);
 108 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
 109 static void init_netif(char *netifExp[], int netifFileNum);
 110 static void usage(const char *ctl);
 111
 112 /*
 113  * Kernel startup for virtual kernels - standard main()
 114  */
 115 int
 116 main(int ac, char **av)
 117 {
 118         char *memImageFile = NULL;
 119         char *netifFile[VKNETIF_MAX];
 120         char *diskFile[VKDISK_MAX];
 121         char *cdFile[VKDISK_MAX];
 122         char *suffix;
 123         int netifFileNum = 0;
 124         int diskFileNum = 0;
 125         int cdFileNum = 0;
 126         int c;
 127         int i;
 128         int n;
 129
 130         /*
 131          * Process options
 132          */
 133         kernel_mem_readonly = 1;
 134
 135         while ((c = getopt(ac, av, "c:svm:r:e:i:I:U")) != -1) {
 136                 switch(c) {
 137                 case 'e':
 138                         /*
 139                          * name=value:name=value:name=value...
 140                          */
 141                         n = strlen(optarg);
 142                         kern_envp = malloc(n + 2);
 143                         for (i = 0; i < n; ++i) {
 144                                 if (optarg[i] == ':')
 145                                         kern_envp[i] = 0;
 146                                 else
 147                                         kern_envp[i] = optarg[i];
 148                         }
 149                         kern_envp[i++] = 0;
 150                         kern_envp[i++] = 0;
 151                         break;
 152                 case 's':
 153                         boothowto |= RB_SINGLE;
 154                         break;
 155                 case 'v':
 156                         bootverbose = 1;
 157                         break;
 158                 case 'i':
 159                         memImageFile = optarg;
 160                         break;
 161                 case 'I':
 162                         if (netifFileNum < VKNETIF_MAX)
 163                                 netifFile[netifFileNum++] = optarg;
 164                         break;
 165                 case 'r':
 166                         if (diskFileNum + cdFileNum < VKDISK_MAX)
 167                                 diskFile[diskFileNum++] = optarg;
 168                         break;
 169                 case 'c':
 170                         if (diskFileNum + cdFileNum < VKDISK_MAX)
 171                                 cdFile[cdFileNum++] = optarg;
 172                         break;
 173                 case 'm':
 174                         Maxmem_bytes = strtoull(optarg, &suffix, 0);
 175                         if (suffix) {
 176                                 switch(*suffix) {
 177                                 case 'g':
 178                                 case 'G':
 179                                         Maxmem_bytes <<= 30;
 180                                         break;
 181                                 case 'm':
 182                                 case 'M':
 183                                         Maxmem_bytes <<= 20;
 184                                         break;
 185                                 case 'k':
 186                                 case 'K':
 187                                         Maxmem_bytes <<= 10;
 188                                         break;
 189                                 default:
 190                                         Maxmem_bytes = 0;
 191                                         usage("Bad maxmem option");
 192                                         /* NOT REACHED */
 193                                         break;
 194                                 }
 195                         }
 196                         break;
 197                 case 'U':
 198                         kernel_mem_readonly = 0;
 199                         break;
 200                 }
 201         }
 202
 203         cpu_disable_intr();
 204         init_sys_memory(memImageFile);
 205         init_kern_memory();
 206         init_globaldata();
 207         init_vkernel();
 208         init_kqueue();
 209         init_disk(diskFile, diskFileNum, VKD_DISK);
 210         init_disk(cdFile, cdFileNum, VKD_CD);
 211         init_netif(netifFile, netifFileNum);
 212         init_exceptions();
 213         mi_startup();
 214         /* NOT REACHED */
 215         exit(1);
 216 }
 217
 218 /*
 219  * Initialize system memory.  This is the virtual kernel's 'RAM'.
 220  */
 221 static
 222 void
 223 init_sys_memory(char *imageFile)
 224 {
 225         struct stat st;
 226         int i;
 227         int fd;
 228
 229         /*
 230          * Figure out the system memory image size.  If an image file was
 231          * specified and -m was not specified, use the image file's size.
 232          */
 233
 234         if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
 235                 Maxmem_bytes = (vm_paddr_t)st.st_size;
 236         if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
 237             Maxmem_bytes == 0) {
 238                 err(1, "Cannot create new memory file %s unless "
 239                        "system memory size is specified with -m",
 240                        imageFile);
 241                 /* NOT REACHED */
 242         }
 243
 244         /*
 245          * Maxmem must be known at this time
 246          */
 247         if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
 248                 err(1, "Bad maxmem specification: 32MB minimum, "
 249                        "multiples of %dMB only",
 250                        SEG_SIZE / 1024 / 1024);
 251                 /* NOT REACHED */
 252         }
 253
 254         /*
 255          * Generate an image file name if necessary, then open/create the
 256          * file exclusively locked.  Do not allow multiple virtual kernels
 257          * to use the same image file.
 258          */
 259         if (imageFile == NULL) {
 260                 for (i = 0; i < 1000000; ++i) {
 261                         asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
 262                         fd = open(imageFile,
 263                                   O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
 264                         if (fd < 0 && errno == EWOULDBLOCK) {
 265                                 free(imageFile);
 266                                 continue;
 267                         }
 268                         break;
 269                 }
 270         } else {
 271                 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
 272         }
 273         printf("Using memory file: %s\n", imageFile);
 274         if (fd < 0 || fstat(fd, &st) < 0) {
 275                 err(1, "Unable to open/create %s", imageFile);
 276                 /* NOT REACHED */
 277         }
 278
 279         /*
 280          * Truncate or extend the file as necessary.
 281          */
 282         if (st.st_size > Maxmem_bytes) {
 283                 ftruncate(fd, Maxmem_bytes);
 284         } else if (st.st_size < Maxmem_bytes) {
 285                 char *zmem;
 286                 off_t off = st.st_size & ~SEG_MASK;
 287
 288                 kprintf("%s: Reserving blocks for memory image\n", imageFile);
 289                 zmem = malloc(SEG_SIZE);
 290                 bzero(zmem, SEG_SIZE);
 291                 lseek(fd, off, SEEK_SET);
 292                 while (off < Maxmem_bytes) {
 293                         if (write(fd, zmem, SEG_SIZE) != SEG_SIZE) {
 294                                 err(1, "Unable to reserve blocks for memory image");
 295                                 /* NOT REACHED */
 296                         }
 297                         off += SEG_SIZE;
 298                 }
 299                 if (fsync(fd) < 0)
 300                         err(1, "Unable to reserve blocks for memory image");
 301                 free(zmem);
 302         }
 303         MemImageFd = fd;
 304         Maxmem = Maxmem_bytes >> PAGE_SHIFT;
 305 }
 306
 307 /*
 308  * Initialize kernel memory.  This reserves kernel virtual memory by using
 309  * MAP_VPAGETABLE
 310  */
 311
 312 static
 313 void
 314 init_kern_memory(void)
 315 {
 316         void *base;
 317         void *try;
 318         char *zero;
 319         char dummy;
 320         char *topofstack = &dummy;
 321         vpte_t pte;
 322         int i;
 323
 324         /*
 325          * Memory map our kernel virtual memory space.  Note that the
 326          * kernel image itself is not made part of this memory for the
 327          * moment.
 328          *
 329          * The memory map must be segment-aligned so we can properly
 330          * offset KernelPTD.
 331          *
 332          * If the system kernel has a different MAXDSIZ, it might not
 333          * be possible to map kernel memory in its prefered location.
 334          * Try a number of different locations.
 335          */
 336         try = (void *)0x40000000;
 337         base = NULL;
 338         while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
 339                 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
 340                             MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
 341                             MemImageFd, 0);
 342                 if (base == try)
 343                         break;
 344                 if (base != MAP_FAILED)
 345                         munmap(base, KERNEL_KVA_SIZE);
 346                 try = (char *)try + 0x10000000;
 347         }
 348         if (base != try) {
 349                 err(1, "Unable to mmap() kernel virtual memory!");
 350                 /* NOT REACHED */
 351         }
 352         madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
 353         KvaStart = (vm_offset_t)base;
 354         KvaSize = KERNEL_KVA_SIZE;
 355         KvaEnd = KvaStart + KvaSize;
 356         printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
 357
 358         /*
 359          * Create a top-level page table self-mapping itself.
 360          *
 361          * Initialize the page directory at physical page index 0 to point
 362          * to an array of page table pages starting at physical page index 1
 363          */
 364         lseek(MemImageFd, 0L, 0);
 365         for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
 366                 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
 367                 write(MemImageFd, &pte, sizeof(pte));
 368         }
 369
 370         /*
 371          * Initialize the PTEs in the page table pages required to map the
 372          * page table itself.  This includes mapping the page directory page
 373          * at the base so we go one more loop then normal.
 374          */
 375         lseek(MemImageFd, PAGE_SIZE, 0);
 376         for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
 377                 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
 378                 write(MemImageFd, &pte, sizeof(pte));
 379         }
 380
 381         /*
 382          * Initialize remaining PTEs to 0.  We may be reusing a memory image
 383          * file.  This is approximately a megabyte.
 384          */
 385         i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
 386         zero = malloc(PAGE_SIZE);
 387         bzero(zero, PAGE_SIZE);
 388         while (i) {
 389                 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
 390                 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
 391         }
 392         free(zero);
 393
 394         /*
 395          * Enable the page table and calculate pointers to our self-map
 396          * for easy kernel page table manipulation.
 397          *
 398          * KernelPTA must be offset so we can do direct VA translations
 399          */
 400         mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
 401                  0 | VPTE_R | VPTE_W | VPTE_V);
 402         KernelPTD = (vpte_t *)base;                       /* pg directory */
 403         KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
 404         KernelPTA -= KvaStart >> PAGE_SHIFT;
 405
 406         /*
 407          * phys_avail[] represents unallocated physical memory.  MI code
 408          * will use phys_avail[] to create the vm_page array.
 409          */
 410         phys_avail[0] = PAGE_SIZE +
 411                         KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
 412         phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
 413         phys_avail[1] = Maxmem_bytes;
 414
 415         /*
 416          * (virtual_start, virtual_end) represent unallocated kernel virtual
 417          * memory.  MI code will create kernel_map using these parameters.
 418          */
 419         virtual_start = KvaStart + PAGE_SIZE +
 420                         KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
 421         virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
 422         virtual_end = KvaStart + KERNEL_KVA_SIZE;
 423
 424         /*
 425          * kernel_vm_end could be set to virtual_end but we want some
 426          * indication of how much of the kernel_map we've used, so
 427          * set it low and let pmap_growkernel increase it even though we
 428          * don't need to create any new page table pages.
 429          */
 430         kernel_vm_end = virtual_start;
 431
 432         /*
 433          * Allocate space for process 0's UAREA.
 434          */
 435         proc0paddr = (void *)virtual_start;
 436         for (i = 0; i < UPAGES; ++i) {
 437                 pmap_kenter_quick(virtual_start, phys_avail[0]);
 438                 virtual_start += PAGE_SIZE;
 439                 phys_avail[0] += PAGE_SIZE;
 440         }
 441
 442         /*
 443          * crashdumpmap
 444          */
 445         crashdumpmap = virtual_start;
 446         virtual_start += MAXDUMPPGS * PAGE_SIZE;
 447
 448         /*
 449          * msgbufp maps the system message buffer
 450          */
 451         assert((MSGBUF_SIZE & PAGE_MASK) == 0);
 452         msgbufp = (void *)virtual_start;
 453         for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
 454                 pmap_kenter_quick(virtual_start, phys_avail[0]);
 455                 virtual_start += PAGE_SIZE;
 456                 phys_avail[0] += PAGE_SIZE;
 457         }
 458         msgbufinit(msgbufp, MSGBUF_SIZE);
 459
 460         /*
 461          * used by kern_memio for /dev/mem access
 462          */
 463         ptvmmap = (caddr_t)virtual_start;
 464         virtual_start += PAGE_SIZE;
 465
 466         /*
 467          * Bootstrap the kernel_pmap
 468          */
 469         pmap_bootstrap();
 470 }
 471
 472 /*
 473  * Map the per-cpu globaldata for cpu #0.  Allocate the space using
 474  * virtual_start and phys_avail[0]
 475  */
 476 static
 477 void
 478 init_globaldata(void)
 479 {
 480         int i;
 481         vm_paddr_t pa;
 482         vm_offset_t va;
 483
 484         /*
 485          * Reserve enough KVA to cover possible cpus.  This is a considerable
 486          * amount of KVA since the privatespace structure includes two
 487          * whole page table mappings.
 488          */
 489         virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
 490         CPU_prvspace = (void *)virtual_start;
 491         virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
 492
 493         /*
 494          * Allocate enough physical memory to cover the mdglobaldata
 495          * portion of the space and the idle stack and map the pages
 496          * into KVA.  For cpu #0 only.
 497          */
 498         for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
 499                 pa = phys_avail[0];
 500                 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
 501                 pmap_kenter_quick(va, pa);
 502                 phys_avail[0] += PAGE_SIZE;
 503         }
 504         for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
 505                 pa = phys_avail[0];
 506                 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
 507                 pmap_kenter_quick(va, pa);
 508                 phys_avail[0] += PAGE_SIZE;
 509         }
 510
 511         /*
 512          * Setup the %gs for cpu #0.  The mycpu macro works after this
 513          * point.
 514          */
 515         tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
 516 }
 517
 518 /*
 519  * Initialize very low level systems including thread0, proc0, etc.
 520  */
 521 static
 522 void
 523 init_vkernel(void)
 524 {
 525         struct mdglobaldata *gd;
 526
 527         gd = &CPU_prvspace[0].mdglobaldata;
 528         bzero(gd, sizeof(*gd));
 529
 530         gd->mi.gd_curthread = &thread0;
 531         thread0.td_gd = &gd->mi;
 532         ncpus = 1;
 533         ncpus2 = 1;     /* rounded down power of 2 */
 534         ncpus_fit = 1;  /* rounded up power of 2 */
 535         /* ncpus2_mask and ncpus_fit_mask are 0 */
 536         init_param1();
 537         gd->mi.gd_prvspace = &CPU_prvspace[0];
 538         mi_gdinit(&gd->mi, 0);
 539         cpu_gdinit(gd, 0);
 540         mi_proc0init(&gd->mi, proc0paddr);
 541         lwp0.lwp_md.md_regs = &proc0_tf;
 542
 543         /*init_locks();*/
 544         cninit();
 545         rand_initialize();
 546 #if 0   /* #ifdef DDB */
 547         kdb_init();
 548         if (boothowto & RB_KDB)
 549                 Debugger("Boot flags requested debugger");
 550 #endif
 551 #if 0
 552         initializecpu();        /* Initialize CPU registers */
 553 #endif
 554         init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
 555
 556 #if 0
 557         /*
 558          * Map the message buffer
 559          */
 560         for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
 561                 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
 562         msgbufinit(msgbufp, MSGBUF_SIZE);
 563 #endif
 564 #if 0
 565         thread0.td_pcb_cr3 ... MMU
 566         lwp0.lwp_md.md_regs = &proc0_tf;
 567 #endif
 568 }
 569
 570 /*
 571  * Filesystem image paths for the virtual kernel are optional.
 572  * If specified they each should point to a disk image,
 573  * the first of which will become the root disk.
 574  *
 575  * The virtual kernel caches data from our 'disk' just like a normal kernel,
 576  * so we do not really want the real kernel to cache the data too.  Use
 577  * O_DIRECT to remove the duplication.
 578  */
 579 static
 580 void
 581 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
 582 {
 583         int i;
 584
 585         if (diskFileNum == 0)
 586                 return;
 587
 588         for(i=0; i < diskFileNum; i++){
 589                 char *fname;
 590                 fname = diskExp[i];
 591
 592                 if (fname == NULL) {
 593                         warnx("Invalid argument to '-r'");
 594                         continue;
 595                 }
 596
 597                 if (DiskNum < VKDISK_MAX) {
 598                         struct stat st;
 599                         struct vkdisk_info* info = NULL;
 600                         int fd;
 601                         size_t l = 0;
 602
 603                         if (type == VKD_DISK)
 604                             fd = open(fname, O_RDWR|O_DIRECT, 0644);
 605                         else
 606                             fd = open(fname, O_RDONLY|O_DIRECT, 0644);
 607                         if (fd < 0 || fstat(fd, &st) < 0) {
 608                                 err(1, "Unable to open/create %s", fname);
 609                                 /* NOT REACHED */
 610                         }
 611
 612                         info = &DiskInfo[DiskNum];
 613                         l = strlen(fname);
 614
 615                         info->unit = i;
 616                         info->fd = fd;
 617                         info->type = type;
 618                         memcpy(info->fname, fname, l);
 619
 620                         if (i == 0) {
 621                                 if (type == VKD_CD)
 622                                     rootdevnames[0] = "cd9660:vcd0a";
 623                                 else if (type == VKD_DISK)
 624                                     rootdevnames[0] = "ufs:vkd0s0a";
 625                         }
 626
 627                         DiskNum++;
 628                 } else {
 629                         warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
 630                         continue;
 631                 }
 632         }
 633 }
 634
 635 static
 636 int
 637 netif_set_tapflags(int tap_unit, int f, int s)
 638 {
 639         struct ifreq ifr;
 640         int flags;
 641
 642         bzero(&ifr, sizeof(ifr));
 643
 644         snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
 645         if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
 646                 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
 647                 return -1;
 648         }
 649
 650         /*
 651          * Adjust if_flags
 652          *
 653          * If the flags are already set/cleared, then we return
 654          * immediately to avoid extra syscalls
 655          */
 656         flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
 657         if (f < 0) {
 658                 /* Turn off flags */
 659                 f = -f;
 660                 if ((flags & f) == 0)
 661                         return 0;
 662                 flags &= ~f;
 663         } else {
 664                 /* Turn on flags */
 665                 if (flags & f)
 666                         return 0;
 667                 flags |= f;
 668         }
 669
 670         /*
 671          * Fix up ifreq.ifr_name, since it may be trashed
 672          * in previous ioctl(SIOCGIFFLAGS)
 673          */
 674         snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
 675
 676         ifr.ifr_flags = flags & 0xffff;
 677         ifr.ifr_flagshigh = flags >> 16;
 678         if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
 679                 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
 680                 return -1;
 681         }
 682         return 0;
 683 }
 684
 685 static
 686 int
 687 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
 688 {
 689         struct ifaliasreq ifra;
 690         struct sockaddr_in *in;
 691
 692         bzero(&ifra, sizeof(ifra));
 693         snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
 694
 695         /* Setup address */
 696         in = (struct sockaddr_in *)&ifra.ifra_addr;
 697         in->sin_family = AF_INET;
 698         in->sin_len = sizeof(*in);
 699         in->sin_addr.s_addr = addr;
 700
 701         if (mask != 0) {
 702                 /* Setup netmask */
 703                 in = (struct sockaddr_in *)&ifra.ifra_mask;
 704                 in->sin_len = sizeof(*in);
 705                 in->sin_addr.s_addr = mask;
 706         }
 707
 708         if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
 709                 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
 710                 return -1;
 711         }
 712         return 0;
 713 }
 714
 715 static
 716 int
 717 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
 718 {
 719         struct ifbreq ifbr;
 720         struct ifdrv ifd;
 721
 722         bzero(&ifbr, sizeof(ifbr));
 723         snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
 724                  "tap%d", tap_unit);
 725
 726         bzero(&ifd, sizeof(ifd));
 727         strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
 728         ifd.ifd_cmd = BRDGADD;
 729         ifd.ifd_len = sizeof(ifbr);
 730         ifd.ifd_data = &ifbr;
 731
 732         if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
 733                 /*
 734                  * 'errno == EEXIST' means that the tap(4) is already
 735                  * a member of the bridge(4)
 736                  */
 737                 if (errno != EEXIST) {
 738                         warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
 739                         return -1;
 740                 }
 741         }
 742         return 0;
 743 }
 744
 745 #define TAPDEV_OFLAGS   (O_RDWR | O_NONBLOCK)
 746
 747 /* XXX major()/minor() can't be used in vkernel */
 748 #define TAPDEV_MAJOR(x) ((int)(((u_int)(x) >> 8) & 0xff))
 749 #define TAPDEV_MINOR(x) ((int)((x) & 0xffff00ff))
 750
 751 #ifndef TAP_CDEV_MAJOR
 752 #define TAP_CDEV_MAJOR  149
 753 #endif
 754
 755 /*
 756  * Locate the first unused tap(4) device file if auto mode is requested,
 757  * or open the user supplied device file, and bring up the corresponding
 758  * tap(4) interface.
 759  *
 760  * NOTE: Only tap(4) device file is supported currently
 761  */
 762 static
 763 int
 764 netif_open_tap(const char *netif, int *tap_unit, int s)
 765 {
 766         char tap_dev[MAXPATHLEN];
 767         int tap_fd, failed;
 768         struct stat st;
 769
 770         *tap_unit = -1;
 771
 772         if (strcmp(netif, "auto") == 0) {
 773                 int i;
 774
 775                 /*
 776                  * Find first unused tap(4) device file
 777                  */
 778                 for (i = 0; ; ++i) {
 779                         snprintf(tap_dev, sizeof(tap_dev), "/dev/tap%d", i);
 780                         tap_fd = open(tap_dev, TAPDEV_OFLAGS);
 781                         if (tap_fd >= 0 || errno == ENOENT)
 782                                 break;
 783                 }
 784                 if (tap_fd < 0) {
 785                         warnx("Unable to find a free tap(4)");
 786                         return -1;
 787                 }
 788         } else {
 789                 /*
 790                  * User supplied tap(4) device file
 791                  */
 792                 if (netif[0] == '/')    /* Absolute path */
 793                         strlcpy(tap_dev, netif, sizeof(tap_dev));
 794                 else
 795                         snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
 796
 797                 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
 798                 if (tap_fd < 0) {
 799                         warn("Unable to open %s", tap_dev);
 800                         return -1;
 801                 }
 802         }
 803
 804         /*
 805          * Check whether the device file is a tap(4)
 806          */
 807         failed = 1;
 808         if (fstat(tap_fd, &st) == 0 && S_ISCHR(st.st_mode) &&
 809             TAPDEV_MAJOR(st.st_rdev) == TAP_CDEV_MAJOR) {
 810                 *tap_unit = TAPDEV_MINOR(st.st_rdev);
 811
 812                 /*
 813                  * Bring up the corresponding tap(4) interface
 814                  */
 815                 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
 816                         failed = 0;
 817         } else {
 818                 warnx("%s is not a tap(4) device", tap_dev);
 819         }
 820
 821         if (failed) {
 822                 close(tap_fd);
 823                 tap_fd = -1;
 824                 *tap_unit = -1;
 825         }
 826         return tap_fd;
 827 }
 828
 829 #undef TAPDEV_MAJOR
 830 #undef TAPDEV_MINOR
 831 #undef TAPDEV_OFLAGS
 832
 833 /*
 834  * Following syntax is supported,
 835  * 1) x.x.x.x             tap(4)'s address is x.x.x.x
 836  *
 837  * 2) x.x.x.x/z           tap(4)'s address is x.x.x.x
 838  *                        tap(4)'s netmask len is z
 839  *
 840  * 3) x.x.x.x:y.y.y.y     tap(4)'s address is x.x.x.x
 841  *                        pseudo netif's address is y.y.y.y
 842  *
 843  * 4) x.x.x.x:y.y.y.y/z   tap(4)'s address is x.x.x.x
 844  *                        pseudo netif's address is y.y.y.y
 845  *                        tap(4) and pseudo netif's netmask len are z
 846  *
 847  * 5) bridgeX             tap(4) will be added to bridgeX
 848  *
 849  * 6) bridgeX:y.y.y.y     tap(4) will be added to bridgeX
 850  *                        pseudo netif's address is y.y.y.y
 851  *
 852  * 7) bridgeX:y.y.y.y/z   tap(4) will be added to bridgeX
 853  *                        pseudo netif's address is y.y.y.y
 854  *                        pseudo netif's netmask len is z
 855  */
 856 static
 857 int
 858 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
 859 {
 860         in_addr_t tap_addr, netmask, netif_addr;
 861         int next_netif_addr;
 862         char *tok, *masklen_str, *ifbridge;
 863
 864         *addr = 0;
 865         *mask = 0;
 866
 867         tok = strtok(NULL, ":/");
 868         if (tok == NULL) {
 869                 /*
 870                  * Nothing special, simply use tap(4) as backend
 871                  */
 872                 return 0;
 873         }
 874
 875         if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
 876                 /*
 877                  * tap(4)'s address is supplied
 878                  */
 879                 ifbridge = NULL;
 880
 881                 /*
 882                  * If there is next token, then it may be pseudo
 883                  * netif's address or netmask len for tap(4)
 884                  */
 885                 next_netif_addr = 0;
 886         } else {
 887                 /*
 888                  * Not tap(4)'s address, assume it as a bridge(4)
 889                  * iface name
 890                  */
 891                 tap_addr = 0;
 892                 ifbridge = tok;
 893
 894                 /*
 895                  * If there is next token, then it must be pseudo
 896                  * netif's address
 897                  */
 898                 next_netif_addr = 1;
 899         }
 900
 901         netmask = netif_addr = 0;
 902
 903         tok = strtok(NULL, ":/");
 904         if (tok == NULL)
 905                 goto back;
 906
 907         if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
 908                 if (next_netif_addr) {
 909                         warnx("Invalid pseudo netif address: %s", tok);
 910                         return -1;
 911                 }
 912                 netif_addr = 0;
 913
 914                 /*
 915                  * Current token is not address, then it must be netmask len
 916                  */
 917                 masklen_str = tok;
 918         } else {
 919                 /*
 920                  * Current token is pseudo netif address, if there is next token
 921                  * it must be netmask len
 922                  */
 923                 masklen_str = strtok(NULL, "/");
 924         }
 925
 926         /* Calculate netmask */
 927         if (masklen_str != NULL) {
 928                 u_long masklen;
 929
 930                 masklen = strtoul(masklen_str, NULL, 10);
 931                 if (masklen < 32 && masklen > 0) {
 932                         netmask = htonl(~((1LL << (32 - masklen)) - 1)
 933                                         & 0xffffffff);
 934                 } else {
 935                         warnx("Invalid netmask len: %lu", masklen);
 936                         return -1;
 937                 }
 938         }
 939
 940         /* Make sure there is no more token left */
 941         if (strtok(NULL, ":/") != NULL) {
 942                 warnx("Invalid argument to '-I'");
 943                 return -1;
 944         }
 945
 946 back:
 947         if (ifbridge == NULL) {
 948                 /* Set tap(4) address/netmask */
 949                 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
 950                         return -1;
 951         } else {
 952                 /* Tie tap(4) to bridge(4) */
 953                 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
 954                         return -1;
 955         }
 956
 957         *addr = netif_addr;
 958         *mask = netmask;
 959         return 0;
 960 }
 961
 962 /*
 963  * NetifInfo[] will be filled for pseudo netif initialization.
 964  * NetifNum will be bumped to reflect the number of valid entries
 965  * in NetifInfo[].
 966  */
 967 static
 968 void
 969 init_netif(char *netifExp[], int netifExpNum)
 970 {
 971         int i, s;
 972
 973         if (netifExpNum == 0)
 974                 return;
 975
 976         s = socket(AF_INET, SOCK_DGRAM, 0);     /* for ioctl(SIOC) */
 977         if (s < 0)
 978                 return;
 979
 980         for (i = 0; i < netifExpNum; ++i) {
 981                 struct vknetif_info *info;
 982                 in_addr_t netif_addr, netif_mask;
 983                 int tap_fd, tap_unit;
 984                 char *netif;
 985
 986                 netif = strtok(netifExp[i], ":");
 987                 if (netif == NULL) {
 988                         warnx("Invalid argument to '-I'");
 989                         continue;
 990                 }
 991
 992                 /*
 993                  * Open tap(4) device file and bring up the
 994                  * corresponding interface
 995                  */
 996                 tap_fd = netif_open_tap(netif, &tap_unit, s);
 997                 if (tap_fd < 0)
 998                         continue;
 999
1000                 /*
1001                  * Initialize tap(4) and get address/netmask
1002                  * for pseudo netif
1003                  *
1004                  * NB: Rest part of netifExp[i] is passed
1005                  *     to netif_init_tap() implicitly.
1006                  */
1007                 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1008                         /*
1009                          * NB: Closing tap(4) device file will bring
1010                          *     down the corresponding interface
1011                          */
1012                         close(tap_fd);
1013                         continue;
1014                 }
1015
1016                 info = &NetifInfo[NetifNum];
1017                 info->tap_fd = tap_fd;
1018                 info->tap_unit = tap_unit;
1019                 info->netif_addr = netif_addr;
1020                 info->netif_mask = netif_mask;
1021
1022                 NetifNum++;
1023                 if (NetifNum >= VKNETIF_MAX)    /* XXX will this happen? */
1024                         break;
1025         }
1026         close(s);
1027 }
1028
1029 static
1030 void
1031 usage(const char *ctl)
1032 {
1033
1034 }
1035
1036 void
1037 cpu_reset(void)
1038 {
1039         kprintf("cpu reset\n");
1040         exit(0);
1041 }
1042
1043 void
1044 cpu_halt(void)
1045 {
1046         kprintf("cpu halt\n");
1047         for (;;)
1048                 __asm__ __volatile("hlt");
1049 }