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 35 #include <sys/types.h> 36 #include <sys/systm.h> 37 #include <sys/kernel.h> 38 #include <sys/stat.h> 39 #include <sys/mman.h> 40 #include <sys/cons.h> 41 #include <sys/random.h> 42 #include <sys/vkernel.h> 43 #include <sys/tls.h> 44 #include <sys/reboot.h> 45 #include <sys/proc.h> 46 #include <sys/msgbuf.h> 47 #include <sys/vmspace.h> 48 #include <sys/socket.h> 49 #include <sys/sockio.h> 50 #include <sys/sysctl.h> 51 #include <sys/un.h> 52 #include <vm/vm_page.h> 53 #include <vm/vm_map.h> 54 #include <sys/mplock2.h> 55 #include <sys/wait.h> 56 #include <sys/vmm.h> 57 58 #include <machine/cpu.h> 59 #include <machine/globaldata.h> 60 #include <machine/tls.h> 61 #include <machine/md_var.h> 62 #include <machine/vmparam.h> 63 #include <cpu/specialreg.h> 64 65 #include <net/if.h> 66 #include <net/if_arp.h> 67 #include <net/ethernet.h> 68 #include <net/bridge/if_bridgevar.h> 69 #include <netinet/in.h> 70 #include <arpa/inet.h> 71 #include <net/if_var.h> 72 73 #include <stdio.h> 74 #include <stdlib.h> 75 #include <stdarg.h> 76 #include <stdbool.h> 77 #include <unistd.h> 78 #include <fcntl.h> 79 #include <string.h> 80 #include <err.h> 81 #include <errno.h> 82 #include <assert.h> 83 #include <sysexits.h> 84 85 #define EX_VKERNEL_REBOOT 32 86 87 vm_paddr_t phys_avail[16]; 88 vm_paddr_t Maxmem; 89 vm_paddr_t Maxmem_bytes; 90 long physmem; 91 int MemImageFd = -1; 92 struct vkdisk_info DiskInfo[VKDISK_MAX]; 93 int DiskNum; 94 struct vknetif_info NetifInfo[VKNETIF_MAX]; 95 int NetifNum; 96 char *pid_file; 97 vm_offset_t KvaStart; 98 vm_offset_t KvaEnd; 99 vm_offset_t KvaSize; 100 vm_offset_t virtual_start; 101 vm_offset_t virtual_end; 102 vm_offset_t virtual2_start; 103 vm_offset_t virtual2_end; 104 vm_offset_t kernel_vm_end; 105 vm_offset_t crashdumpmap; 106 vm_offset_t clean_sva; 107 vm_offset_t clean_eva; 108 struct msgbuf *msgbufp; 109 caddr_t ptvmmap; 110 vpte_t *KernelPTD; 111 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */ 112 void *dmap_min_address; 113 void *vkernel_stack; 114 u_int cpu_feature; /* XXX */ 115 int tsc_present; 116 int tsc_invariant; 117 int tsc_mpsync; 118 int64_t tsc_frequency; 119 int optcpus; /* number of cpus - see mp_start() */ 120 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */ 121 int real_ncpus; /* number of real CPUs */ 122 int next_cpu; /* next real CPU to lock a virtual CPU to */ 123 int vkernel_b_arg; /* no of logical CPU bits - only SMP */ 124 int vkernel_B_arg; /* no of core bits - only SMP */ 125 int vmm_enabled; /* VMM HW assisted enable */ 126 struct privatespace *CPU_prvspace; 127 128 extern uint64_t KPML4phys; /* phys addr of kernel level 4 */ 129 130 static struct trapframe proc0_tf; 131 static void *proc0paddr; 132 133 static void init_sys_memory(char *imageFile); 134 static void init_kern_memory(void); 135 static void init_kern_memory_vmm(void); 136 static void init_globaldata(void); 137 static void init_vkernel(void); 138 static void init_disk(char **diskExp, int *diskFlags, int diskFileNum, enum vkdisk_type type); 139 static void init_netif(char *netifExp[], int netifFileNum); 140 static void writepid(void); 141 static void cleanpid(void); 142 static int unix_connect(const char *path); 143 static void usage_err(const char *ctl, ...); 144 static void usage_help(_Bool); 145 static void init_locks(void); 146 static void handle_term(int); 147 148 pid_t childpid; 149 150 static int save_ac; 151 static char **save_av; 152 153 /* 154 * Kernel startup for virtual kernels - standard main() 155 */ 156 int 157 main(int ac, char **av) 158 { 159 char *memImageFile = NULL; 160 char *netifFile[VKNETIF_MAX]; 161 char *diskFile[VKDISK_MAX]; 162 char *cdFile[VKDISK_MAX]; 163 char *suffix; 164 char *endp; 165 char *tmp; 166 char *tok; 167 int diskFlags[VKDISK_MAX]; 168 int netifFileNum = 0; 169 int diskFileNum = 0; 170 int cdFileNum = 0; 171 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */ 172 int c; 173 int i; 174 int j; 175 int n; 176 int isq; 177 int pos; 178 int eflag; 179 int dflag = 0; /* disable vmm */ 180 int real_vkernel_enable; 181 int supports_sse; 182 uint32_t mxcsr_mask; 183 size_t vsize; 184 size_t msize; 185 size_t kenv_size; 186 size_t kenv_size2; 187 int status; 188 struct sigaction sa; 189 190 /* 191 * Currently a bad hack but rtld-elf needs LD_SHAREDLIB_BASE to 192 * be set to force it to mmap() shared libraries into low memory, 193 * so our module loader can link against the related symbols. 194 */ 195 if (getenv("LD_SHAREDLIB_BASE") == NULL) { 196 setenv("LD_SHAREDLIB_BASE", "0x10000000", 1); 197 execv(av[0], av); 198 fprintf(stderr, "Must run %s with full path\n", av[0]); 199 exit(1); 200 } 201 202 while ((childpid = fork()) != 0) { 203 /* Ignore signals */ 204 bzero(&sa, sizeof(sa)); 205 sigemptyset(&sa.sa_mask); 206 sa.sa_handler = SIG_IGN; 207 sigaction(SIGINT, &sa, NULL); 208 sigaction(SIGQUIT, &sa, NULL); 209 sigaction(SIGHUP, &sa, NULL); 210 211 /* 212 * Forward SIGTERM to the child so that 213 * the shutdown process initiates correctly. 214 */ 215 sa.sa_handler = handle_term; 216 sigaction(SIGTERM, &sa, NULL); 217 218 /* 219 * Wait for child to terminate, exit if 220 * someone stole our child. 221 */ 222 while (waitpid(childpid, &status, 0) != childpid) { 223 if (errno == ECHILD) 224 exit(1); 225 } 226 if (WEXITSTATUS(status) != EX_VKERNEL_REBOOT) 227 return 0; 228 } 229 230 /* 231 * Starting for real 232 */ 233 save_ac = ac; 234 save_av = av; 235 eflag = 0; 236 pos = 0; 237 kenv_size = 0; 238 /* 239 * Process options 240 */ 241 kernel_mem_readonly = 1; 242 optcpus = 2; 243 vkernel_b_arg = 0; 244 vkernel_B_arg = 0; 245 lwp_cpu_lock = LCL_NONE; 246 247 real_vkernel_enable = 0; 248 vsize = sizeof(real_vkernel_enable); 249 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0); 250 251 if (real_vkernel_enable == 0) { 252 errx(1, "vm.vkernel_enable is 0, must be set " 253 "to 1 to execute a vkernel!"); 254 } 255 256 real_ncpus = 1; 257 vsize = sizeof(real_ncpus); 258 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0); 259 260 if (ac < 2) 261 usage_help(false); 262 263 while ((c = getopt(ac, av, "c:hsvl:m:n:r:R:e:i:p:I:Ud")) != -1) { 264 switch(c) { 265 case 'd': 266 dflag = 1; 267 break; 268 case 'e': 269 /* 270 * name=value:name=value:name=value... 271 * name="value"... 272 * 273 * Allow values to be quoted but note that shells 274 * may remove the quotes, so using this feature 275 * to embed colons may require a backslash. 276 */ 277 n = strlen(optarg); 278 isq = 0; 279 280 if (eflag == 0) { 281 kenv_size = n + 2; 282 kern_envp = malloc(kenv_size); 283 if (kern_envp == NULL) 284 errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size); 285 } else { 286 kenv_size2 = kenv_size + n + 1; 287 pos = kenv_size - 1; 288 if ((tmp = realloc(kern_envp, kenv_size2)) == NULL) 289 errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2); 290 kern_envp = tmp; 291 kenv_size = kenv_size2; 292 } 293 294 for (i = 0, j = pos; i < n; ++i) { 295 if (optarg[i] == '"') 296 isq ^= 1; 297 else if (optarg[i] == '\'') 298 isq ^= 2; 299 else if (isq == 0 && optarg[i] == ':') 300 kern_envp[j++] = 0; 301 else 302 kern_envp[j++] = optarg[i]; 303 } 304 kern_envp[j++] = 0; 305 kern_envp[j++] = 0; 306 eflag++; 307 break; 308 case 's': 309 boothowto |= RB_SINGLE; 310 break; 311 case 'v': 312 bootverbose = 1; 313 break; 314 case 'i': 315 memImageFile = optarg; 316 break; 317 case 'I': 318 if (netifFileNum < VKNETIF_MAX) 319 netifFile[netifFileNum++] = strdup(optarg); 320 break; 321 case 'r': 322 case 'R': 323 if (bootOnDisk < 0) 324 bootOnDisk = 1; 325 if (diskFileNum + cdFileNum < VKDISK_MAX) { 326 diskFile[diskFileNum] = strdup(optarg); 327 diskFlags[diskFileNum] = (c == 'R'); 328 ++diskFileNum; 329 } 330 break; 331 case 'c': 332 if (bootOnDisk < 0) 333 bootOnDisk = 0; 334 if (diskFileNum + cdFileNum < VKDISK_MAX) 335 cdFile[cdFileNum++] = strdup(optarg); 336 break; 337 case 'm': 338 Maxmem_bytes = strtoull(optarg, &suffix, 0); 339 if (suffix) { 340 switch(*suffix) { 341 case 'g': 342 case 'G': 343 Maxmem_bytes <<= 30; 344 break; 345 case 'm': 346 case 'M': 347 Maxmem_bytes <<= 20; 348 break; 349 case 'k': 350 case 'K': 351 Maxmem_bytes <<= 10; 352 break; 353 default: 354 Maxmem_bytes = 0; 355 usage_err("Bad maxmem option"); 356 /* NOT REACHED */ 357 break; 358 } 359 } 360 break; 361 case 'l': 362 next_cpu = -1; 363 if (strncmp("map", optarg, 3) == 0) { 364 lwp_cpu_lock = LCL_PER_CPU; 365 if (optarg[3] == ',') { 366 next_cpu = strtol(optarg+4, &endp, 0); 367 if (*endp != '\0') 368 usage_err("Bad target CPU number at '%s'", endp); 369 } else { 370 next_cpu = 0; 371 } 372 if (next_cpu < 0 || next_cpu > real_ncpus - 1) 373 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1); 374 } else if (strncmp("any", optarg, 3) == 0) { 375 lwp_cpu_lock = LCL_NONE; 376 } else { 377 lwp_cpu_lock = LCL_SINGLE_CPU; 378 next_cpu = strtol(optarg, &endp, 0); 379 if (*endp != '\0') 380 usage_err("Bad target CPU number at '%s'", endp); 381 if (next_cpu < 0 || next_cpu > real_ncpus - 1) 382 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1); 383 } 384 break; 385 case 'n': 386 /* 387 * This value is set up by mp_start(), don't just 388 * set ncpus here. 389 */ 390 tok = strtok(optarg, ":"); 391 optcpus = strtol(tok, NULL, 0); 392 if (optcpus < 1 || optcpus > MAXCPU) 393 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU); 394 395 /* :lbits argument */ 396 tok = strtok(NULL, ":"); 397 if (tok != NULL) { 398 vkernel_b_arg = strtol(tok, NULL, 0); 399 400 /* :cbits argument */ 401 tok = strtok(NULL, ":"); 402 if (tok != NULL) { 403 vkernel_B_arg = strtol(tok, NULL, 0); 404 } 405 406 } 407 break; 408 case 'p': 409 pid_file = optarg; 410 break; 411 case 'U': 412 kernel_mem_readonly = 0; 413 break; 414 case 'h': 415 usage_help(true); 416 break; 417 default: 418 usage_help(false); 419 } 420 } 421 422 /* 423 * Check VMM presence 424 */ 425 vsize = sizeof(vmm_enabled); 426 sysctlbyname("hw.vmm.enable", &vmm_enabled, &vsize, NULL, 0); 427 vmm_enabled = (vmm_enabled && !dflag); 428 429 writepid(); 430 cpu_disable_intr(); 431 if (vmm_enabled) { 432 /* use a MAP_ANON directly */ 433 init_kern_memory_vmm(); 434 } else { 435 init_sys_memory(memImageFile); 436 init_kern_memory(); 437 } 438 init_globaldata(); 439 init_vkernel(); 440 setrealcpu(); 441 init_kqueue(); 442 443 vmm_guest = VMM_GUEST_VKERNEL; 444 445 /* 446 * Check TSC 447 */ 448 vsize = sizeof(tsc_present); 449 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0); 450 vsize = sizeof(tsc_invariant); 451 sysctlbyname("hw.tsc_invariant", &tsc_invariant, &vsize, NULL, 0); 452 vsize = sizeof(tsc_mpsync); 453 sysctlbyname("hw.tsc_mpsync", &tsc_mpsync, &vsize, NULL, 0); 454 vsize = sizeof(tsc_frequency); 455 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0); 456 if (tsc_present) 457 cpu_feature |= CPUID_TSC; 458 459 /* 460 * Check SSE 461 */ 462 vsize = sizeof(supports_sse); 463 supports_sse = 0; 464 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0); 465 sysctlbyname("hw.mxcsr_mask", &mxcsr_mask, &msize, NULL, 0); 466 init_fpu(supports_sse); 467 if (supports_sse) 468 cpu_feature |= CPUID_SSE | CPUID_FXSR; 469 470 /* 471 * We boot from the first installed disk. 472 */ 473 if (bootOnDisk == 1) { 474 init_disk(diskFile, diskFlags, diskFileNum, VKD_DISK); 475 init_disk(cdFile, NULL, cdFileNum, VKD_CD); 476 } else { 477 init_disk(cdFile, NULL, cdFileNum, VKD_CD); 478 init_disk(diskFile, diskFlags, diskFileNum, VKD_DISK); 479 } 480 481 init_netif(netifFile, netifFileNum); 482 init_exceptions(); 483 mi_startup(); 484 /* NOT REACHED */ 485 exit(EX_SOFTWARE); 486 } 487 488 /* SIGTERM handler */ 489 static 490 void 491 handle_term(int sig) 492 { 493 kill(childpid, sig); 494 } 495 496 /* 497 * Initialize system memory. This is the virtual kernel's 'RAM'. 498 */ 499 static 500 void 501 init_sys_memory(char *imageFile) 502 { 503 struct stat st; 504 int i; 505 int fd; 506 507 /* 508 * Figure out the system memory image size. If an image file was 509 * specified and -m was not specified, use the image file's size. 510 */ 511 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0) 512 Maxmem_bytes = (vm_paddr_t)st.st_size; 513 if ((imageFile == NULL || stat(imageFile, &st) < 0) && 514 Maxmem_bytes == 0) { 515 errx(1, "Cannot create new memory file %s unless " 516 "system memory size is specified with -m", 517 imageFile); 518 /* NOT REACHED */ 519 } 520 521 /* 522 * Maxmem must be known at this time 523 */ 524 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) { 525 errx(1, "Bad maxmem specification: 64MB minimum, " 526 "multiples of %dMB only", 527 SEG_SIZE / 1024 / 1024); 528 /* NOT REACHED */ 529 } 530 531 /* 532 * Generate an image file name if necessary, then open/create the 533 * file exclusively locked. Do not allow multiple virtual kernels 534 * to use the same image file. 535 * 536 * Don't iterate through a million files if we do not have write 537 * access to the directory, stop if our open() failed on a 538 * non-existant file. Otherwise opens can fail for any number 539 */ 540 if (imageFile == NULL) { 541 for (i = 0; i < 1000000; ++i) { 542 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i); 543 fd = open(imageFile, 544 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644); 545 if (fd < 0 && stat(imageFile, &st) == 0) { 546 free(imageFile); 547 continue; 548 } 549 break; 550 } 551 } else { 552 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644); 553 } 554 fprintf(stderr, "Using memory file: %s\n", imageFile); 555 if (fd < 0 || fstat(fd, &st) < 0) { 556 err(1, "Unable to open/create %s", imageFile); 557 /* NOT REACHED */ 558 } 559 560 /* 561 * Truncate or extend the file as necessary. Clean out the contents 562 * of the file, we want it to be full of holes so we don't waste 563 * time reading in data from an old file that we no longer care 564 * about. 565 */ 566 ftruncate(fd, 0); 567 ftruncate(fd, Maxmem_bytes); 568 569 MemImageFd = fd; 570 Maxmem = Maxmem_bytes >> PAGE_SHIFT; 571 physmem = Maxmem; 572 } 573 574 /* 575 * Initialize kernel memory. This reserves kernel virtual memory by using 576 * MAP_VPAGETABLE 577 */ 578 579 static 580 void 581 init_kern_memory(void) 582 { 583 void *base; 584 int i; 585 void *firstfree; 586 587 /* 588 * Memory map our kernel virtual memory space. Note that the 589 * kernel image itself is not made part of this memory for the 590 * moment. 591 * 592 * The memory map must be segment-aligned so we can properly 593 * offset KernelPTD. 594 * 595 * If the system kernel has a different MAXDSIZ, it might not 596 * be possible to map kernel memory in its prefered location. 597 * Try a number of different locations. 598 */ 599 600 base = mmap((void*)KERNEL_KVA_START, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE, 601 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE|MAP_FIXED|MAP_TRYFIXED, 602 MemImageFd, (off_t)KERNEL_KVA_START); 603 604 if (base == MAP_FAILED) { 605 err(1, "Unable to mmap() kernel virtual memory!"); 606 /* NOT REACHED */ 607 } 608 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC); 609 KvaStart = (vm_offset_t)base; 610 KvaSize = KERNEL_KVA_SIZE; 611 KvaEnd = KvaStart + KvaSize; 612 613 /* cannot use kprintf yet */ 614 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd); 615 616 /* MAP_FILE? */ 617 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE, 618 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED, 619 MemImageFd, 0); 620 if (dmap_min_address == MAP_FAILED) { 621 err(1, "Unable to mmap() kernel DMAP region!"); 622 /* NOT REACHED */ 623 } 624 625 /* 626 * Bootstrap the kernel_pmap 627 */ 628 firstfree = NULL; 629 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base); 630 631 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP, 632 0 | VPTE_RW | VPTE_V); 633 634 /* 635 * phys_avail[] represents unallocated physical memory. MI code 636 * will use phys_avail[] to create the vm_page array. 637 */ 638 phys_avail[0] = (vm_paddr_t)firstfree; 639 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK; 640 phys_avail[1] = Maxmem_bytes; 641 642 #if JGV 643 /* 644 * (virtual_start, virtual_end) represent unallocated kernel virtual 645 * memory. MI code will create kernel_map using these parameters. 646 */ 647 virtual_start = KvaStart + (long)firstfree; 648 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK; 649 virtual_end = KvaStart + KERNEL_KVA_SIZE; 650 #endif 651 652 /* 653 * pmap_growkernel() will set the correct value. 654 */ 655 kernel_vm_end = 0; 656 657 /* 658 * Allocate space for process 0's UAREA. 659 */ 660 proc0paddr = (void *)virtual_start; 661 for (i = 0; i < UPAGES; ++i) { 662 pmap_kenter_quick(virtual_start, phys_avail[0]); 663 virtual_start += PAGE_SIZE; 664 phys_avail[0] += PAGE_SIZE; 665 } 666 667 /* 668 * crashdumpmap 669 */ 670 crashdumpmap = virtual_start; 671 virtual_start += MAXDUMPPGS * PAGE_SIZE; 672 673 /* 674 * msgbufp maps the system message buffer 675 */ 676 assert((MSGBUF_SIZE & PAGE_MASK) == 0); 677 msgbufp = (void *)virtual_start; 678 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) { 679 pmap_kenter_quick(virtual_start, phys_avail[0]); 680 virtual_start += PAGE_SIZE; 681 phys_avail[0] += PAGE_SIZE; 682 } 683 msgbufinit(msgbufp, MSGBUF_SIZE); 684 685 /* 686 * used by kern_memio for /dev/mem access 687 */ 688 ptvmmap = (caddr_t)virtual_start; 689 virtual_start += PAGE_SIZE; 690 } 691 692 static 693 void 694 init_kern_memory_vmm(void) 695 { 696 int i; 697 void *firstfree; 698 struct vmm_guest_options options; 699 void *dmap_address; 700 701 KvaStart = (vm_offset_t)KERNEL_KVA_START; 702 KvaSize = KERNEL_KVA_SIZE; 703 KvaEnd = KvaStart + KvaSize; 704 705 Maxmem = Maxmem_bytes >> PAGE_SHIFT; 706 physmem = Maxmem; 707 708 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) { 709 errx(1, "Bad maxmem specification: 64MB minimum, " 710 "multiples of %dMB only", 711 SEG_SIZE / 1024 / 1024); 712 /* NOT REACHED */ 713 } 714 715 /* Call the vmspace_create to allocate the internal 716 * vkernel structures. Won't do anything else (no new 717 * vmspace) 718 */ 719 if (vmspace_create(NULL, 0, NULL) < 0) 720 panic("vmspace_create() failed"); 721 722 723 /* 724 * MAP_ANON the region of the VKERNEL phyisical memory 725 * (known as GPA - Guest Physical Address 726 */ 727 dmap_address = mmap(NULL, Maxmem_bytes, PROT_READ|PROT_WRITE|PROT_EXEC, 728 MAP_ANON|MAP_SHARED, -1, 0); 729 if (dmap_address == MAP_FAILED) { 730 err(1, "Unable to mmap() RAM region!"); 731 /* NOT REACHED */ 732 } 733 /* bzero(dmap_address, Maxmem_bytes); */ 734 735 /* Alloc a new stack in the lowmem */ 736 vkernel_stack = mmap(NULL, KERNEL_STACK_SIZE, 737 PROT_READ|PROT_WRITE|PROT_EXEC, 738 MAP_ANON, -1, 0); 739 if (vkernel_stack == MAP_FAILED) { 740 err(1, "Unable to allocate stack\n"); 741 } 742 743 /* 744 * Bootstrap the kernel_pmap 745 */ 746 firstfree = dmap_address; 747 dmap_min_address = NULL; /* VIRT == PHYS in the first 512G */ 748 pmap_bootstrap((vm_paddr_t *)&firstfree, (uint64_t)KvaStart); 749 750 /* 751 * Enter VMM mode 752 */ 753 options.guest_cr3 = (register_t) KPML4phys; 754 options.new_stack = (uint64_t) vkernel_stack + KERNEL_STACK_SIZE; 755 options.master = 1; 756 if (vmm_guest_ctl(VMM_GUEST_RUN, &options)) { 757 err(1, "Unable to enter VMM mode."); 758 } 759 760 /* 761 * phys_avail[] represents unallocated physical memory. MI code 762 * will use phys_avail[] to create the vm_page array. 763 */ 764 phys_avail[0] = (vm_paddr_t)firstfree; 765 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK; 766 phys_avail[1] = (vm_paddr_t)dmap_address + Maxmem_bytes; 767 768 /* 769 * pmap_growkernel() will set the correct value. 770 */ 771 kernel_vm_end = 0; 772 773 /* 774 * Allocate space for process 0's UAREA. 775 */ 776 proc0paddr = (void *)virtual_start; 777 for (i = 0; i < UPAGES; ++i) { 778 pmap_kenter_quick(virtual_start, phys_avail[0]); 779 virtual_start += PAGE_SIZE; 780 phys_avail[0] += PAGE_SIZE; 781 } 782 783 /* 784 * crashdumpmap 785 */ 786 crashdumpmap = virtual_start; 787 virtual_start += MAXDUMPPGS * PAGE_SIZE; 788 789 /* 790 * msgbufp maps the system message buffer 791 */ 792 assert((MSGBUF_SIZE & PAGE_MASK) == 0); 793 msgbufp = (void *)virtual_start; 794 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) { 795 796 pmap_kenter_quick(virtual_start, phys_avail[0]); 797 virtual_start += PAGE_SIZE; 798 phys_avail[0] += PAGE_SIZE; 799 } 800 801 msgbufinit(msgbufp, MSGBUF_SIZE); 802 803 /* 804 * used by kern_memio for /dev/mem access 805 */ 806 ptvmmap = (caddr_t)virtual_start; 807 virtual_start += PAGE_SIZE; 808 809 printf("vmm: Hardware pagetable enabled for guest\n"); 810 } 811 812 813 /* 814 * Map the per-cpu globaldata for cpu #0. Allocate the space using 815 * virtual_start and phys_avail[0] 816 */ 817 static 818 void 819 init_globaldata(void) 820 { 821 int i; 822 vm_paddr_t pa; 823 vm_offset_t va; 824 825 /* 826 * Reserve enough KVA to cover possible cpus. This is a considerable 827 * amount of KVA since the privatespace structure includes two 828 * whole page table mappings. 829 */ 830 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK; 831 CPU_prvspace = (void *)virtual_start; 832 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU; 833 834 /* 835 * Allocate enough physical memory to cover the mdglobaldata 836 * portion of the space and the idle stack and map the pages 837 * into KVA. For cpu #0 only. 838 */ 839 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) { 840 pa = phys_avail[0]; 841 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i; 842 pmap_kenter_quick(va, pa); 843 phys_avail[0] += PAGE_SIZE; 844 } 845 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) { 846 pa = phys_avail[0]; 847 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i; 848 pmap_kenter_quick(va, pa); 849 phys_avail[0] += PAGE_SIZE; 850 } 851 852 /* 853 * Setup the %gs for cpu #0. The mycpu macro works after this 854 * point. Note that %fs is used by pthreads. 855 */ 856 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace)); 857 } 858 859 860 /* 861 * Initialize pool tokens and other necessary locks 862 */ 863 static void 864 init_locks(void) 865 { 866 867 /* 868 * Get the initial mplock with a count of 1 for the BSP. 869 * This uses a LOGICAL cpu ID, ie BSP == 0. 870 */ 871 cpu_get_initial_mplock(); 872 873 /* our token pool needs to work early */ 874 lwkt_token_pool_init(); 875 876 } 877 878 879 /* 880 * Initialize very low level systems including thread0, proc0, etc. 881 */ 882 static 883 void 884 init_vkernel(void) 885 { 886 struct mdglobaldata *gd; 887 888 gd = &CPU_prvspace[0].mdglobaldata; 889 bzero(gd, sizeof(*gd)); 890 891 gd->mi.gd_curthread = &thread0; 892 thread0.td_gd = &gd->mi; 893 ncpus = 1; 894 ncpus2 = 1; /* rounded down power of 2 */ 895 ncpus_fit = 1; /* rounded up power of 2 */ 896 /* ncpus2_mask and ncpus_fit_mask are 0 */ 897 init_param1(); 898 gd->mi.gd_prvspace = &CPU_prvspace[0]; 899 mi_gdinit(&gd->mi, 0); 900 cpu_gdinit(gd, 0); 901 mi_proc0init(&gd->mi, proc0paddr); 902 lwp0.lwp_md.md_regs = &proc0_tf; 903 904 init_locks(); 905 cninit(); 906 rand_initialize(); 907 #if 0 /* #ifdef DDB */ 908 kdb_init(); 909 if (boothowto & RB_KDB) 910 Debugger("Boot flags requested debugger"); 911 #endif 912 identcpu(); 913 #if 0 914 initializecpu(); /* Initialize CPU registers */ 915 #endif 916 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE); 917 918 #if 0 919 /* 920 * Map the message buffer 921 */ 922 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE) 923 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off); 924 msgbufinit(msgbufp, MSGBUF_SIZE); 925 #endif 926 #if 0 927 thread0.td_pcb_cr3 ... MMU 928 lwp0.lwp_md.md_regs = &proc0_tf; 929 #endif 930 } 931 932 /* 933 * Filesystem image paths for the virtual kernel are optional. 934 * If specified they each should point to a disk image, 935 * the first of which will become the root disk. 936 * 937 * The virtual kernel caches data from our 'disk' just like a normal kernel, 938 * so we do not really want the real kernel to cache the data too. Use 939 * O_DIRECT to remove the duplication. 940 */ 941 static 942 void 943 init_disk(char **diskExp, int *diskFlags, int diskFileNum, enum vkdisk_type type) 944 { 945 char *serno; 946 int i; 947 948 if (diskFileNum == 0) 949 return; 950 951 for (i=0; i < diskFileNum; i++){ 952 char *fname; 953 fname = diskExp[i]; 954 955 if (fname == NULL) { 956 warnx("Invalid argument to '-r'"); 957 continue; 958 } 959 /* 960 * Check for a serial number for the virtual disk 961 * passed from the command line. 962 */ 963 serno = fname; 964 strsep(&serno, ":"); 965 966 if (DiskNum < VKDISK_MAX) { 967 struct stat st; 968 struct vkdisk_info *info = NULL; 969 int fd; 970 size_t l = 0; 971 972 if (type == VKD_DISK) 973 fd = open(fname, O_RDWR|O_DIRECT, 0644); 974 else 975 fd = open(fname, O_RDONLY|O_DIRECT, 0644); 976 if (fd < 0 || fstat(fd, &st) < 0) { 977 err(1, "Unable to open/create %s", fname); 978 /* NOT REACHED */ 979 } 980 if (S_ISREG(st.st_mode) && (diskFlags[i] & 1) == 0) { 981 if (flock(fd, LOCK_EX|LOCK_NB) < 0) { 982 errx(1, "Disk image %s is already " 983 "in use\n", fname); 984 /* NOT REACHED */ 985 } 986 } 987 988 info = &DiskInfo[DiskNum]; 989 l = strlen(fname); 990 991 info->unit = i; 992 info->fd = fd; 993 info->type = type; 994 info->flags = diskFlags[i]; 995 memcpy(info->fname, fname, l); 996 info->serno = NULL; 997 if (serno) { 998 if ((info->serno = malloc(SERNOLEN)) != NULL) 999 strlcpy(info->serno, serno, SERNOLEN); 1000 else 1001 warnx("Couldn't allocate memory for the operation"); 1002 } 1003 1004 if (DiskNum == 0) { 1005 if (type == VKD_CD) { 1006 rootdevnames[0] = "cd9660:vcd0"; 1007 } else if (type == VKD_DISK) { 1008 rootdevnames[0] = "ufs:vkd0s0a"; 1009 rootdevnames[1] = "ufs:vkd0s1a"; 1010 } 1011 } 1012 1013 DiskNum++; 1014 } else { 1015 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname); 1016 continue; 1017 } 1018 } 1019 } 1020 1021 static 1022 int 1023 netif_set_tapflags(int tap_unit, int f, int s) 1024 { 1025 struct ifreq ifr; 1026 int flags; 1027 1028 bzero(&ifr, sizeof(ifr)); 1029 1030 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit); 1031 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) { 1032 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit); 1033 return -1; 1034 } 1035 1036 /* 1037 * Adjust if_flags 1038 * 1039 * If the flags are already set/cleared, then we return 1040 * immediately to avoid extra syscalls 1041 */ 1042 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16); 1043 if (f < 0) { 1044 /* Turn off flags */ 1045 f = -f; 1046 if ((flags & f) == 0) 1047 return 0; 1048 flags &= ~f; 1049 } else { 1050 /* Turn on flags */ 1051 if (flags & f) 1052 return 0; 1053 flags |= f; 1054 } 1055 1056 /* 1057 * Fix up ifreq.ifr_name, since it may be trashed 1058 * in previous ioctl(SIOCGIFFLAGS) 1059 */ 1060 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit); 1061 1062 ifr.ifr_flags = flags & 0xffff; 1063 ifr.ifr_flagshigh = flags >> 16; 1064 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) { 1065 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit); 1066 return -1; 1067 } 1068 return 0; 1069 } 1070 1071 static 1072 int 1073 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s) 1074 { 1075 struct ifaliasreq ifra; 1076 struct sockaddr_in *in; 1077 1078 bzero(&ifra, sizeof(ifra)); 1079 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit); 1080 1081 /* Setup address */ 1082 in = (struct sockaddr_in *)&ifra.ifra_addr; 1083 in->sin_family = AF_INET; 1084 in->sin_len = sizeof(*in); 1085 in->sin_addr.s_addr = addr; 1086 1087 if (mask != 0) { 1088 /* Setup netmask */ 1089 in = (struct sockaddr_in *)&ifra.ifra_mask; 1090 in->sin_len = sizeof(*in); 1091 in->sin_addr.s_addr = mask; 1092 } 1093 1094 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) { 1095 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit); 1096 return -1; 1097 } 1098 return 0; 1099 } 1100 1101 static 1102 int 1103 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s) 1104 { 1105 struct ifbreq ifbr; 1106 struct ifdrv ifd; 1107 1108 bzero(&ifbr, sizeof(ifbr)); 1109 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname), 1110 "tap%d", tap_unit); 1111 1112 bzero(&ifd, sizeof(ifd)); 1113 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name)); 1114 ifd.ifd_cmd = BRDGADD; 1115 ifd.ifd_len = sizeof(ifbr); 1116 ifd.ifd_data = &ifbr; 1117 1118 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) { 1119 /* 1120 * 'errno == EEXIST' means that the tap(4) is already 1121 * a member of the bridge(4) 1122 */ 1123 if (errno != EEXIST) { 1124 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge); 1125 return -1; 1126 } 1127 } 1128 return 0; 1129 } 1130 1131 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK) 1132 1133 /* 1134 * Locate the first unused tap(4) device file if auto mode is requested, 1135 * or open the user supplied device file, and bring up the corresponding 1136 * tap(4) interface. 1137 * 1138 * NOTE: Only tap(4) device file is supported currently 1139 */ 1140 static 1141 int 1142 netif_open_tap(const char *netif, int *tap_unit, int s) 1143 { 1144 char tap_dev[MAXPATHLEN]; 1145 int tap_fd, failed; 1146 struct stat st; 1147 char *dname; 1148 1149 *tap_unit = -1; 1150 1151 if (strcmp(netif, "auto") == 0) { 1152 /* 1153 * Find first unused tap(4) device file 1154 */ 1155 tap_fd = open("/dev/tap", TAPDEV_OFLAGS); 1156 if (tap_fd < 0) { 1157 warnc(errno, "Unable to find a free tap(4)"); 1158 return -1; 1159 } 1160 } else { 1161 /* 1162 * User supplied tap(4) device file or unix socket. 1163 */ 1164 if (netif[0] == '/') /* Absolute path */ 1165 strlcpy(tap_dev, netif, sizeof(tap_dev)); 1166 else 1167 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif); 1168 1169 tap_fd = open(tap_dev, TAPDEV_OFLAGS); 1170 1171 /* 1172 * If we cannot open normally try to connect to it. 1173 */ 1174 if (tap_fd < 0) 1175 tap_fd = unix_connect(tap_dev); 1176 1177 if (tap_fd < 0) { 1178 warn("Unable to open %s", tap_dev); 1179 return -1; 1180 } 1181 } 1182 1183 /* 1184 * Check whether the device file is a tap(4) 1185 */ 1186 if (fstat(tap_fd, &st) < 0) { 1187 failed = 1; 1188 } else if (S_ISCHR(st.st_mode)) { 1189 dname = fdevname(tap_fd); 1190 if (dname) 1191 dname = strstr(dname, "tap"); 1192 if (dname) { 1193 /* 1194 * Bring up the corresponding tap(4) interface 1195 */ 1196 *tap_unit = strtol(dname + 3, NULL, 10); 1197 printf("TAP UNIT %d\n", *tap_unit); 1198 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0) 1199 failed = 0; 1200 else 1201 failed = 1; 1202 } else { 1203 failed = 1; 1204 } 1205 } else if (S_ISSOCK(st.st_mode)) { 1206 /* 1207 * Special socket connection (typically to vknet). We 1208 * do not have to do anything. 1209 */ 1210 failed = 0; 1211 } else { 1212 failed = 1; 1213 } 1214 1215 if (failed) { 1216 warnx("%s is not a tap(4) device or socket", tap_dev); 1217 close(tap_fd); 1218 tap_fd = -1; 1219 *tap_unit = -1; 1220 } 1221 return tap_fd; 1222 } 1223 1224 static int 1225 unix_connect(const char *path) 1226 { 1227 struct sockaddr_un sunx; 1228 int len; 1229 int net_fd; 1230 int sndbuf = 262144; 1231 struct stat st; 1232 1233 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path); 1234 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]); 1235 ++len; /* include nul */ 1236 sunx.sun_family = AF_UNIX; 1237 sunx.sun_len = len; 1238 1239 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0); 1240 if (net_fd < 0) 1241 return(-1); 1242 if (connect(net_fd, (void *)&sunx, len) < 0) { 1243 close(net_fd); 1244 return(-1); 1245 } 1246 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf)); 1247 if (fstat(net_fd, &st) == 0) 1248 printf("Network socket buffer: %d bytes\n", st.st_blksize); 1249 fcntl(net_fd, F_SETFL, O_NONBLOCK); 1250 return(net_fd); 1251 } 1252 1253 #undef TAPDEV_MAJOR 1254 #undef TAPDEV_MINOR 1255 #undef TAPDEV_OFLAGS 1256 1257 /* 1258 * Following syntax is supported, 1259 * 1) x.x.x.x tap(4)'s address is x.x.x.x 1260 * 1261 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x 1262 * tap(4)'s netmask len is z 1263 * 1264 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x 1265 * pseudo netif's address is y.y.y.y 1266 * 1267 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x 1268 * pseudo netif's address is y.y.y.y 1269 * tap(4) and pseudo netif's netmask len are z 1270 * 1271 * 5) bridgeX tap(4) will be added to bridgeX 1272 * 1273 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX 1274 * pseudo netif's address is y.y.y.y 1275 * 1276 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX 1277 * pseudo netif's address is y.y.y.y 1278 * pseudo netif's netmask len is z 1279 */ 1280 static 1281 int 1282 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s) 1283 { 1284 in_addr_t tap_addr, netmask, netif_addr; 1285 int next_netif_addr; 1286 char *tok, *masklen_str, *ifbridge; 1287 1288 *addr = 0; 1289 *mask = 0; 1290 1291 tok = strtok(NULL, ":/"); 1292 if (tok == NULL) { 1293 /* 1294 * Nothing special, simply use tap(4) as backend 1295 */ 1296 return 0; 1297 } 1298 1299 if (inet_pton(AF_INET, tok, &tap_addr) > 0) { 1300 /* 1301 * tap(4)'s address is supplied 1302 */ 1303 ifbridge = NULL; 1304 1305 /* 1306 * If there is next token, then it may be pseudo 1307 * netif's address or netmask len for tap(4) 1308 */ 1309 next_netif_addr = 0; 1310 } else { 1311 /* 1312 * Not tap(4)'s address, assume it as a bridge(4) 1313 * iface name 1314 */ 1315 tap_addr = 0; 1316 ifbridge = tok; 1317 1318 /* 1319 * If there is next token, then it must be pseudo 1320 * netif's address 1321 */ 1322 next_netif_addr = 1; 1323 } 1324 1325 netmask = netif_addr = 0; 1326 1327 tok = strtok(NULL, ":/"); 1328 if (tok == NULL) 1329 goto back; 1330 1331 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) { 1332 if (next_netif_addr) { 1333 warnx("Invalid pseudo netif address: %s", tok); 1334 return -1; 1335 } 1336 netif_addr = 0; 1337 1338 /* 1339 * Current token is not address, then it must be netmask len 1340 */ 1341 masklen_str = tok; 1342 } else { 1343 /* 1344 * Current token is pseudo netif address, if there is next token 1345 * it must be netmask len 1346 */ 1347 masklen_str = strtok(NULL, "/"); 1348 } 1349 1350 /* Calculate netmask */ 1351 if (masklen_str != NULL) { 1352 u_long masklen; 1353 1354 masklen = strtoul(masklen_str, NULL, 10); 1355 if (masklen < 32 && masklen > 0) { 1356 netmask = htonl(~((1LL << (32 - masklen)) - 1) 1357 & 0xffffffff); 1358 } else { 1359 warnx("Invalid netmask len: %lu", masklen); 1360 return -1; 1361 } 1362 } 1363 1364 /* Make sure there is no more token left */ 1365 if (strtok(NULL, ":/") != NULL) { 1366 warnx("Invalid argument to '-I'"); 1367 return -1; 1368 } 1369 1370 back: 1371 if (tap_unit < 0) { 1372 /* Do nothing */ 1373 } else if (ifbridge == NULL) { 1374 /* Set tap(4) address/netmask */ 1375 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0) 1376 return -1; 1377 } else { 1378 /* Tie tap(4) to bridge(4) */ 1379 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0) 1380 return -1; 1381 } 1382 1383 *addr = netif_addr; 1384 *mask = netmask; 1385 return 0; 1386 } 1387 1388 /* 1389 * NetifInfo[] will be filled for pseudo netif initialization. 1390 * NetifNum will be bumped to reflect the number of valid entries 1391 * in NetifInfo[]. 1392 */ 1393 static 1394 void 1395 init_netif(char *netifExp[], int netifExpNum) 1396 { 1397 int i, s; 1398 char *tmp; 1399 1400 if (netifExpNum == 0) 1401 return; 1402 1403 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */ 1404 if (s < 0) 1405 return; 1406 1407 for (i = 0; i < netifExpNum; ++i) { 1408 struct vknetif_info *info; 1409 in_addr_t netif_addr, netif_mask; 1410 int tap_fd, tap_unit; 1411 char *netif; 1412 1413 /* Extract MAC address if there is one */ 1414 tmp = netifExp[i]; 1415 strsep(&tmp, "="); 1416 1417 netif = strtok(netifExp[i], ":"); 1418 if (netif == NULL) { 1419 warnx("Invalid argument to '-I'"); 1420 continue; 1421 } 1422 1423 /* 1424 * Open tap(4) device file and bring up the 1425 * corresponding interface 1426 */ 1427 tap_fd = netif_open_tap(netif, &tap_unit, s); 1428 if (tap_fd < 0) 1429 continue; 1430 1431 /* 1432 * Initialize tap(4) and get address/netmask 1433 * for pseudo netif 1434 * 1435 * NB: Rest part of netifExp[i] is passed 1436 * to netif_init_tap() implicitly. 1437 */ 1438 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) { 1439 /* 1440 * NB: Closing tap(4) device file will bring 1441 * down the corresponding interface 1442 */ 1443 close(tap_fd); 1444 continue; 1445 } 1446 1447 info = &NetifInfo[NetifNum]; 1448 bzero(info, sizeof(*info)); 1449 info->tap_fd = tap_fd; 1450 info->tap_unit = tap_unit; 1451 info->netif_addr = netif_addr; 1452 info->netif_mask = netif_mask; 1453 /* 1454 * If tmp isn't NULL it means a MAC could have been 1455 * specified so attempt to convert it. 1456 * Setting enaddr to NULL will tell vke_attach() we 1457 * need a pseudo-random MAC address. 1458 */ 1459 if (tmp != NULL) { 1460 if ((info->enaddr = malloc(ETHER_ADDR_LEN)) == NULL) 1461 warnx("Couldn't allocate memory for the operation"); 1462 else { 1463 if ((kether_aton(tmp, info->enaddr)) == NULL) { 1464 free(info->enaddr); 1465 info->enaddr = NULL; 1466 } 1467 } 1468 } 1469 1470 NetifNum++; 1471 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */ 1472 break; 1473 } 1474 close(s); 1475 } 1476 1477 /* 1478 * Create the pid file and leave it open and locked while the vkernel is 1479 * running. This allows a script to use /usr/bin/lockf to probe whether 1480 * a vkernel is still running (so as not to accidently kill an unrelated 1481 * process from a stale pid file). 1482 */ 1483 static 1484 void 1485 writepid(void) 1486 { 1487 char buf[32]; 1488 int fd; 1489 1490 if (pid_file != NULL) { 1491 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid()); 1492 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666); 1493 if (fd < 0) { 1494 if (errno == EWOULDBLOCK) { 1495 perror("Failed to lock pidfile, " 1496 "vkernel already running"); 1497 } else { 1498 perror("Failed to create pidfile"); 1499 } 1500 exit(EX_SOFTWARE); 1501 } 1502 ftruncate(fd, 0); 1503 write(fd, buf, strlen(buf)); 1504 /* leave the file open to maintain the lock */ 1505 } 1506 } 1507 1508 static 1509 void 1510 cleanpid( void ) 1511 { 1512 if (pid_file != NULL) { 1513 if (unlink(pid_file) < 0) 1514 perror("Warning: couldn't remove pidfile"); 1515 } 1516 } 1517 1518 static 1519 void 1520 usage_err(const char *ctl, ...) 1521 { 1522 va_list va; 1523 1524 va_start(va, ctl); 1525 vfprintf(stderr, ctl, va); 1526 va_end(va); 1527 fprintf(stderr, "\n"); 1528 exit(EX_USAGE); 1529 } 1530 1531 static 1532 void 1533 usage_help(_Bool help) 1534 { 1535 fprintf(stderr, "Usage: %s [-hsUvd] [-c file] [-e name=value:name=value:...]\n" 1536 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n" 1537 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n" 1538 "\t[-p file] [-r file]\n", save_av[0]); 1539 1540 if (help) 1541 fprintf(stderr, "\nArguments:\n" 1542 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n" 1543 "\t-e\tSpecify an environment to be used by the kernel.\n" 1544 "\t-h\tThis list of options.\n" 1545 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n" 1546 "\t-I\tCreate a virtual network device.\n" 1547 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n" 1548 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n" 1549 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n" 1550 "\t\t\tnumcpus - number of cpus\n" 1551 "\t\t\tlbits - specify the number of bits within APICID(=CPUID)\n" 1552 "\t\t\t needed for representing the logical ID.\n" 1553 "\t\t\t Controls the number of threads/core:\n" 1554 "\t\t\t (0 bits - 1 thread, 1 bit - 2 threads).\n" 1555 "\t\t\tcbits - specify the number of bits within APICID(=CPUID)\n" 1556 "\t\t\t needed for representing the core ID.\n" 1557 "\t\t\t Controls the number of cores/package:\n" 1558 "\t\t\t (0 bits - 1 core, 1 bit - 2 cores).\n" 1559 "\t-p\tSpecify a file in which to store the process ID.\n" 1560 "\t-r\tSpecify a R/W disk image file, iterates vkd0..n\n" 1561 "\t-R\tSpecify a COW disk image file, iterates vkd0..n\n" 1562 "\t-s\tBoot into single-user mode.\n" 1563 "\t-U\tEnable writing to kernel memory and module loading.\n" 1564 "\t-v\tTurn on verbose booting.\n"); 1565 1566 exit(EX_USAGE); 1567 } 1568 1569 void 1570 cpu_smp_stopped(void) 1571 { 1572 } 1573 1574 void 1575 cpu_reset(void) 1576 { 1577 kprintf("cpu reset, rebooting vkernel\n"); 1578 closefrom(3); 1579 cleanpid(); 1580 exit(EX_VKERNEL_REBOOT); 1581 } 1582 1583 void 1584 cpu_halt(void) 1585 { 1586 kprintf("cpu halt, exiting vkernel\n"); 1587 cleanpid(); 1588 exit(EX_OK); 1589 } 1590 1591 void 1592 setrealcpu(void) 1593 { 1594 switch(lwp_cpu_lock) { 1595 case LCL_PER_CPU: 1596 if (bootverbose) 1597 kprintf("Locking CPU%d to real cpu %d\n", 1598 mycpuid, next_cpu); 1599 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu)); 1600 next_cpu++; 1601 if (next_cpu >= real_ncpus) 1602 next_cpu = 0; 1603 break; 1604 case LCL_SINGLE_CPU: 1605 if (bootverbose) 1606 kprintf("Locking CPU%d to real cpu %d\n", 1607 mycpuid, next_cpu); 1608 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu)); 1609 break; 1610 default: 1611 /* do not map virtual cpus to real cpus */ 1612 break; 1613 } 1614 } 1615 1616 /* 1617 * Allocate and free memory for module loading. The loaded module 1618 * has to be placed somewhere near the current kernel binary load 1619 * point or the relocations will not work. 1620 * 1621 * I'm not sure why this isn't working. 1622 */ 1623 int 1624 vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes) 1625 { 1626 #if 1 1627 size_t xtra; 1628 xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK; 1629 *basep = (vm_offset_t)sbrk(xtra + bytes) + xtra; 1630 bzero((void *)*basep, bytes); 1631 #else 1632 *basep = (vm_offset_t)mmap((void *)0x000000000, bytes, 1633 PROT_READ|PROT_WRITE|PROT_EXEC, 1634 MAP_ANON|MAP_SHARED, -1, 0); 1635 if ((void *)*basep == MAP_FAILED) 1636 return ENOMEM; 1637 #endif 1638 return 0; 1639 } 1640 1641 void 1642 vkernel_module_memory_free(vm_offset_t base, size_t bytes) 1643 { 1644 #if 0 1645 #if 0 1646 munmap((void *)base, bytes); 1647 #endif 1648 #endif 1649 } 1650