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 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); 146 147 static int save_ac; 148 static char **save_av; 149 150 /* 151 * Kernel startup for virtual kernels - standard main() 152 */ 153 int main(int ac, char **av) { 154 char *memImageFile = NULL; 155 char *netifFile[VKNETIF_MAX]; 156 char *diskFile[VKDISK_MAX]; 157 char *cdFile[VKDISK_MAX]; 158 char *suffix; 159 char *endp; 160 char *tmp; 161 char *tok; 162 int netifFileNum = 0; 163 int diskFileNum = 0; 164 int cdFileNum = 0; 165 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */ 166 int c; 167 int i; 168 int j; 169 int n; 170 int isq; 171 int pos; 172 int eflag; 173 int dflag = 0; /* disable vmm */ 174 int real_vkernel_enable; 175 int supports_sse; 176 size_t vsize; 177 size_t kenv_size; 178 size_t kenv_size2; 179 pid_t pid; 180 int status; 181 struct sigaction sa; 182 183 /* 184 * Currently a bad hack but rtld-elf needs LD_SHAREDLIB_BASE to 185 * be set to force it to mmap() shared libraries into low memory, 186 * so our module loader can link against the related symbols. 187 */ 188 if (getenv("LD_SHAREDLIB_BASE") == NULL) { 189 setenv("LD_SHAREDLIB_BASE", "0x10000000", 1); 190 execv(av[0], av); 191 fprintf(stderr, "Must run %s with full path\n", av[0]); 192 exit(1); 193 } 194 195 while ((pid = fork()) != 0) { 196 /* Ignore signals */ 197 bzero(&sa, sizeof(sa)); 198 sigemptyset(&sa.sa_mask); 199 sa.sa_handler = SIG_IGN; 200 sigaction(SIGINT, &sa, NULL); 201 sigaction(SIGQUIT, &sa, NULL); 202 sigaction(SIGHUP, &sa, NULL); 203 204 /* 205 * Wait for child to terminate, exit if 206 * someone stole our child. 207 */ 208 while (waitpid(pid, &status, 0) != pid) { 209 if (errno == ECHILD) 210 exit(1); 211 } 212 if (WEXITSTATUS(status) != EX_VKERNEL_REBOOT) 213 return 0; 214 } 215 216 /* 217 * Starting for real 218 */ 219 save_ac = ac; 220 save_av = av; 221 eflag = 0; 222 pos = 0; 223 kenv_size = 0; 224 /* 225 * Process options 226 */ 227 kernel_mem_readonly = 1; 228 optcpus = 2; 229 vkernel_b_arg = 0; 230 vkernel_B_arg = 0; 231 lwp_cpu_lock = LCL_NONE; 232 233 real_vkernel_enable = 0; 234 vsize = sizeof(real_vkernel_enable); 235 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0); 236 237 if (real_vkernel_enable == 0) { 238 errx(1, "vm.vkernel_enable is 0, must be set " 239 "to 1 to execute a vkernel!"); 240 } 241 242 real_ncpus = 1; 243 vsize = sizeof(real_ncpus); 244 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0); 245 246 if (ac < 2) 247 usage_help(false); 248 249 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:Ud")) != -1) { 250 switch(c) { 251 case 'd': 252 dflag = 1; 253 break; 254 case 'e': 255 /* 256 * name=value:name=value:name=value... 257 * name="value"... 258 * 259 * Allow values to be quoted but note that shells 260 * may remove the quotes, so using this feature 261 * to embed colons may require a backslash. 262 */ 263 n = strlen(optarg); 264 isq = 0; 265 266 if (eflag == 0) { 267 kenv_size = n + 2; 268 kern_envp = malloc(kenv_size); 269 if (kern_envp == NULL) 270 errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size); 271 } else { 272 kenv_size2 = kenv_size + n + 1; 273 pos = kenv_size - 1; 274 if ((tmp = realloc(kern_envp, kenv_size2)) == NULL) 275 errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2); 276 kern_envp = tmp; 277 kenv_size = kenv_size2; 278 } 279 280 for (i = 0, j = pos; i < n; ++i) { 281 if (optarg[i] == '"') 282 isq ^= 1; 283 else if (optarg[i] == '\'') 284 isq ^= 2; 285 else if (isq == 0 && optarg[i] == ':') 286 kern_envp[j++] = 0; 287 else 288 kern_envp[j++] = optarg[i]; 289 } 290 kern_envp[j++] = 0; 291 kern_envp[j++] = 0; 292 eflag++; 293 break; 294 case 's': 295 boothowto |= RB_SINGLE; 296 break; 297 case 'v': 298 bootverbose = 1; 299 break; 300 case 'i': 301 memImageFile = optarg; 302 break; 303 case 'I': 304 if (netifFileNum < VKNETIF_MAX) 305 netifFile[netifFileNum++] = strdup(optarg); 306 break; 307 case 'r': 308 if (bootOnDisk < 0) 309 bootOnDisk = 1; 310 if (diskFileNum + cdFileNum < VKDISK_MAX) 311 diskFile[diskFileNum++] = strdup(optarg); 312 break; 313 case 'c': 314 if (bootOnDisk < 0) 315 bootOnDisk = 0; 316 if (diskFileNum + cdFileNum < VKDISK_MAX) 317 cdFile[cdFileNum++] = strdup(optarg); 318 break; 319 case 'm': 320 Maxmem_bytes = strtoull(optarg, &suffix, 0); 321 if (suffix) { 322 switch(*suffix) { 323 case 'g': 324 case 'G': 325 Maxmem_bytes <<= 30; 326 break; 327 case 'm': 328 case 'M': 329 Maxmem_bytes <<= 20; 330 break; 331 case 'k': 332 case 'K': 333 Maxmem_bytes <<= 10; 334 break; 335 default: 336 Maxmem_bytes = 0; 337 usage_err("Bad maxmem option"); 338 /* NOT REACHED */ 339 break; 340 } 341 } 342 break; 343 case 'l': 344 next_cpu = -1; 345 if (strncmp("map", optarg, 3) == 0) { 346 lwp_cpu_lock = LCL_PER_CPU; 347 if (optarg[3] == ',') { 348 next_cpu = strtol(optarg+4, &endp, 0); 349 if (*endp != '\0') 350 usage_err("Bad target CPU number at '%s'", endp); 351 } else { 352 next_cpu = 0; 353 } 354 if (next_cpu < 0 || next_cpu > real_ncpus - 1) 355 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1); 356 } else if (strncmp("any", optarg, 3) == 0) { 357 lwp_cpu_lock = LCL_NONE; 358 } else { 359 lwp_cpu_lock = LCL_SINGLE_CPU; 360 next_cpu = strtol(optarg, &endp, 0); 361 if (*endp != '\0') 362 usage_err("Bad target CPU number at '%s'", endp); 363 if (next_cpu < 0 || next_cpu > real_ncpus - 1) 364 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1); 365 } 366 break; 367 case 'n': 368 /* 369 * This value is set up by mp_start(), don't just 370 * set ncpus here. 371 */ 372 tok = strtok(optarg, ":"); 373 optcpus = strtol(tok, NULL, 0); 374 if (optcpus < 1 || optcpus > MAXCPU) 375 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU); 376 377 /* :lbits argument */ 378 tok = strtok(NULL, ":"); 379 if (tok != NULL) { 380 vkernel_b_arg = strtol(tok, NULL, 0); 381 382 /* :cbits argument */ 383 tok = strtok(NULL, ":"); 384 if (tok != NULL) { 385 vkernel_B_arg = strtol(tok, NULL, 0); 386 } 387 388 } 389 break; 390 case 'p': 391 pid_file = optarg; 392 break; 393 case 'U': 394 kernel_mem_readonly = 0; 395 break; 396 case 'h': 397 usage_help(true); 398 break; 399 default: 400 usage_help(false); 401 } 402 } 403 404 /* 405 * Check VMM presence 406 */ 407 vsize = sizeof(vmm_enabled); 408 sysctlbyname("hw.vmm.enable", &vmm_enabled, &vsize, NULL, 0); 409 vmm_enabled = (vmm_enabled && !dflag); 410 411 writepid(); 412 cpu_disable_intr(); 413 if (vmm_enabled) { 414 /* use a MAP_ANON directly */ 415 init_kern_memory_vmm(); 416 } else { 417 init_sys_memory(memImageFile); 418 init_kern_memory(); 419 } 420 init_globaldata(); 421 init_vkernel(); 422 setrealcpu(); 423 init_kqueue(); 424 425 vmm_guest = 1; 426 427 /* 428 * Check TSC 429 */ 430 vsize = sizeof(tsc_present); 431 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0); 432 vsize = sizeof(tsc_invariant); 433 sysctlbyname("hw.tsc_invariant", &tsc_invariant, &vsize, NULL, 0); 434 vsize = sizeof(tsc_mpsync); 435 sysctlbyname("hw.tsc_mpsync", &tsc_mpsync, &vsize, NULL, 0); 436 vsize = sizeof(tsc_frequency); 437 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0); 438 if (tsc_present) 439 cpu_feature |= CPUID_TSC; 440 441 /* 442 * Check SSE 443 */ 444 vsize = sizeof(supports_sse); 445 supports_sse = 0; 446 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0); 447 init_fpu(supports_sse); 448 if (supports_sse) 449 cpu_feature |= CPUID_SSE | CPUID_FXSR; 450 451 /* 452 * We boot from the first installed disk. 453 */ 454 if (bootOnDisk == 1) { 455 init_disk(diskFile, diskFileNum, VKD_DISK); 456 init_disk(cdFile, cdFileNum, VKD_CD); 457 } else { 458 init_disk(cdFile, cdFileNum, VKD_CD); 459 init_disk(diskFile, diskFileNum, VKD_DISK); 460 } 461 462 init_netif(netifFile, netifFileNum); 463 init_exceptions(); 464 mi_startup(); 465 /* NOT REACHED */ 466 exit(EX_SOFTWARE); 467 } 468 469 /* 470 * Initialize system memory. This is the virtual kernel's 'RAM'. 471 */ 472 static 473 void 474 init_sys_memory(char *imageFile) 475 { 476 struct stat st; 477 int i; 478 int fd; 479 480 /* 481 * Figure out the system memory image size. If an image file was 482 * specified and -m was not specified, use the image file's size. 483 */ 484 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0) 485 Maxmem_bytes = (vm_paddr_t)st.st_size; 486 if ((imageFile == NULL || stat(imageFile, &st) < 0) && 487 Maxmem_bytes == 0) { 488 errx(1, "Cannot create new memory file %s unless " 489 "system memory size is specified with -m", 490 imageFile); 491 /* NOT REACHED */ 492 } 493 494 /* 495 * Maxmem must be known at this time 496 */ 497 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) { 498 errx(1, "Bad maxmem specification: 64MB minimum, " 499 "multiples of %dMB only", 500 SEG_SIZE / 1024 / 1024); 501 /* NOT REACHED */ 502 } 503 504 /* 505 * Generate an image file name if necessary, then open/create the 506 * file exclusively locked. Do not allow multiple virtual kernels 507 * to use the same image file. 508 * 509 * Don't iterate through a million files if we do not have write 510 * access to the directory, stop if our open() failed on a 511 * non-existant file. Otherwise opens can fail for any number 512 */ 513 if (imageFile == NULL) { 514 for (i = 0; i < 1000000; ++i) { 515 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i); 516 fd = open(imageFile, 517 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644); 518 if (fd < 0 && stat(imageFile, &st) == 0) { 519 free(imageFile); 520 continue; 521 } 522 break; 523 } 524 } else { 525 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644); 526 } 527 fprintf(stderr, "Using memory file: %s\n", imageFile); 528 if (fd < 0 || fstat(fd, &st) < 0) { 529 err(1, "Unable to open/create %s", imageFile); 530 /* NOT REACHED */ 531 } 532 533 /* 534 * Truncate or extend the file as necessary. Clean out the contents 535 * of the file, we want it to be full of holes so we don't waste 536 * time reading in data from an old file that we no longer care 537 * about. 538 */ 539 ftruncate(fd, 0); 540 ftruncate(fd, Maxmem_bytes); 541 542 MemImageFd = fd; 543 Maxmem = Maxmem_bytes >> PAGE_SHIFT; 544 physmem = Maxmem; 545 } 546 547 /* 548 * Initialize kernel memory. This reserves kernel virtual memory by using 549 * MAP_VPAGETABLE 550 */ 551 552 static 553 void 554 init_kern_memory(void) 555 { 556 void *base; 557 int i; 558 void *firstfree; 559 560 /* 561 * Memory map our kernel virtual memory space. Note that the 562 * kernel image itself is not made part of this memory for the 563 * moment. 564 * 565 * The memory map must be segment-aligned so we can properly 566 * offset KernelPTD. 567 * 568 * If the system kernel has a different MAXDSIZ, it might not 569 * be possible to map kernel memory in its prefered location. 570 * Try a number of different locations. 571 */ 572 573 base = mmap((void*)KERNEL_KVA_START, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE, 574 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE|MAP_FIXED|MAP_TRYFIXED, 575 MemImageFd, (off_t)KERNEL_KVA_START); 576 577 if (base == MAP_FAILED) { 578 err(1, "Unable to mmap() kernel virtual memory!"); 579 /* NOT REACHED */ 580 } 581 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC); 582 KvaStart = (vm_offset_t)base; 583 KvaSize = KERNEL_KVA_SIZE; 584 KvaEnd = KvaStart + KvaSize; 585 586 /* cannot use kprintf yet */ 587 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd); 588 589 /* MAP_FILE? */ 590 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE, 591 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED, 592 MemImageFd, 0); 593 if (dmap_min_address == MAP_FAILED) { 594 err(1, "Unable to mmap() kernel DMAP region!"); 595 /* NOT REACHED */ 596 } 597 598 /* 599 * Bootstrap the kernel_pmap 600 */ 601 firstfree = NULL; 602 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base); 603 604 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP, 605 0 | VPTE_RW | VPTE_V); 606 607 /* 608 * phys_avail[] represents unallocated physical memory. MI code 609 * will use phys_avail[] to create the vm_page array. 610 */ 611 phys_avail[0] = (vm_paddr_t)firstfree; 612 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK; 613 phys_avail[1] = Maxmem_bytes; 614 615 #if JGV 616 /* 617 * (virtual_start, virtual_end) represent unallocated kernel virtual 618 * memory. MI code will create kernel_map using these parameters. 619 */ 620 virtual_start = KvaStart + (long)firstfree; 621 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK; 622 virtual_end = KvaStart + KERNEL_KVA_SIZE; 623 #endif 624 625 /* 626 * pmap_growkernel() will set the correct value. 627 */ 628 kernel_vm_end = 0; 629 630 /* 631 * Allocate space for process 0's UAREA. 632 */ 633 proc0paddr = (void *)virtual_start; 634 for (i = 0; i < UPAGES; ++i) { 635 pmap_kenter_quick(virtual_start, phys_avail[0]); 636 virtual_start += PAGE_SIZE; 637 phys_avail[0] += PAGE_SIZE; 638 } 639 640 /* 641 * crashdumpmap 642 */ 643 crashdumpmap = virtual_start; 644 virtual_start += MAXDUMPPGS * PAGE_SIZE; 645 646 /* 647 * msgbufp maps the system message buffer 648 */ 649 assert((MSGBUF_SIZE & PAGE_MASK) == 0); 650 msgbufp = (void *)virtual_start; 651 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) { 652 pmap_kenter_quick(virtual_start, phys_avail[0]); 653 virtual_start += PAGE_SIZE; 654 phys_avail[0] += PAGE_SIZE; 655 } 656 msgbufinit(msgbufp, MSGBUF_SIZE); 657 658 /* 659 * used by kern_memio for /dev/mem access 660 */ 661 ptvmmap = (caddr_t)virtual_start; 662 virtual_start += PAGE_SIZE; 663 } 664 665 static 666 void 667 init_kern_memory_vmm(void) 668 { 669 int i; 670 void *firstfree; 671 struct vmm_guest_options options; 672 void *dmap_address; 673 674 KvaStart = (vm_offset_t)KERNEL_KVA_START; 675 KvaSize = KERNEL_KVA_SIZE; 676 KvaEnd = KvaStart + KvaSize; 677 678 Maxmem = Maxmem_bytes >> PAGE_SHIFT; 679 physmem = Maxmem; 680 681 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) { 682 errx(1, "Bad maxmem specification: 64MB minimum, " 683 "multiples of %dMB only", 684 SEG_SIZE / 1024 / 1024); 685 /* NOT REACHED */ 686 } 687 688 /* Call the vmspace_create to allocate the internal 689 * vkernel structures. Won't do anything else (no new 690 * vmspace) 691 */ 692 if (vmspace_create(NULL, 0, NULL) < 0) 693 panic("vmspace_create() failed"); 694 695 696 /* 697 * MAP_ANON the region of the VKERNEL phyisical memory 698 * (known as GPA - Guest Physical Address 699 */ 700 dmap_address = mmap(NULL, Maxmem_bytes, PROT_READ|PROT_WRITE|PROT_EXEC, 701 MAP_ANON|MAP_SHARED, -1, 0); 702 if (dmap_address == MAP_FAILED) { 703 err(1, "Unable to mmap() RAM region!"); 704 /* NOT REACHED */ 705 } 706 707 /* Alloc a new stack in the lowmem */ 708 vkernel_stack = mmap(NULL, KERNEL_STACK_SIZE, 709 PROT_READ|PROT_WRITE|PROT_EXEC, 710 MAP_ANON, -1, 0); 711 if (vkernel_stack == MAP_FAILED) { 712 err(1, "Unable to allocate stack\n"); 713 } 714 715 /* 716 * Bootstrap the kernel_pmap 717 */ 718 firstfree = dmap_address; 719 dmap_min_address = NULL; /* VIRT == PHYS in the first 512G */ 720 pmap_bootstrap((vm_paddr_t *)&firstfree, (uint64_t)KvaStart); 721 722 /* 723 * Enter VMM mode 724 */ 725 options.guest_cr3 = (register_t) KPML4phys; 726 options.new_stack = (uint64_t) vkernel_stack + KERNEL_STACK_SIZE; 727 options.master = 1; 728 if (vmm_guest_ctl(VMM_GUEST_RUN, &options)) { 729 err(1, "Unable to enter VMM mode."); 730 } 731 732 /* 733 * phys_avail[] represents unallocated physical memory. MI code 734 * will use phys_avail[] to create the vm_page array. 735 */ 736 phys_avail[0] = (vm_paddr_t)firstfree; 737 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK; 738 phys_avail[1] = (vm_paddr_t)dmap_address + Maxmem_bytes; 739 740 /* 741 * pmap_growkernel() will set the correct value. 742 */ 743 kernel_vm_end = 0; 744 745 /* 746 * Allocate space for process 0's UAREA. 747 */ 748 proc0paddr = (void *)virtual_start; 749 for (i = 0; i < UPAGES; ++i) { 750 pmap_kenter_quick(virtual_start, phys_avail[0]); 751 virtual_start += PAGE_SIZE; 752 phys_avail[0] += PAGE_SIZE; 753 } 754 755 /* 756 * crashdumpmap 757 */ 758 crashdumpmap = virtual_start; 759 virtual_start += MAXDUMPPGS * PAGE_SIZE; 760 761 /* 762 * msgbufp maps the system message buffer 763 */ 764 assert((MSGBUF_SIZE & PAGE_MASK) == 0); 765 msgbufp = (void *)virtual_start; 766 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) { 767 768 pmap_kenter_quick(virtual_start, phys_avail[0]); 769 virtual_start += PAGE_SIZE; 770 phys_avail[0] += PAGE_SIZE; 771 } 772 773 msgbufinit(msgbufp, MSGBUF_SIZE); 774 775 /* 776 * used by kern_memio for /dev/mem access 777 */ 778 ptvmmap = (caddr_t)virtual_start; 779 virtual_start += PAGE_SIZE; 780 781 printf("vmm: Hardware pagetable enabled for guest\n"); 782 } 783 784 785 /* 786 * Map the per-cpu globaldata for cpu #0. Allocate the space using 787 * virtual_start and phys_avail[0] 788 */ 789 static 790 void 791 init_globaldata(void) 792 { 793 int i; 794 vm_paddr_t pa; 795 vm_offset_t va; 796 797 /* 798 * Reserve enough KVA to cover possible cpus. This is a considerable 799 * amount of KVA since the privatespace structure includes two 800 * whole page table mappings. 801 */ 802 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK; 803 CPU_prvspace = (void *)virtual_start; 804 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU; 805 806 /* 807 * Allocate enough physical memory to cover the mdglobaldata 808 * portion of the space and the idle stack and map the pages 809 * into KVA. For cpu #0 only. 810 */ 811 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) { 812 pa = phys_avail[0]; 813 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i; 814 pmap_kenter_quick(va, pa); 815 phys_avail[0] += PAGE_SIZE; 816 } 817 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) { 818 pa = phys_avail[0]; 819 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i; 820 pmap_kenter_quick(va, pa); 821 phys_avail[0] += PAGE_SIZE; 822 } 823 824 /* 825 * Setup the %gs for cpu #0. The mycpu macro works after this 826 * point. Note that %fs is used by pthreads. 827 */ 828 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace)); 829 } 830 831 832 /* 833 * Initialize pool tokens and other necessary locks 834 */ 835 static void 836 init_locks(void) 837 { 838 839 /* 840 * Get the initial mplock with a count of 1 for the BSP. 841 * This uses a LOGICAL cpu ID, ie BSP == 0. 842 */ 843 cpu_get_initial_mplock(); 844 845 /* our token pool needs to work early */ 846 lwkt_token_pool_init(); 847 848 } 849 850 851 /* 852 * Initialize very low level systems including thread0, proc0, etc. 853 */ 854 static 855 void 856 init_vkernel(void) 857 { 858 struct mdglobaldata *gd; 859 860 gd = &CPU_prvspace[0].mdglobaldata; 861 bzero(gd, sizeof(*gd)); 862 863 gd->mi.gd_curthread = &thread0; 864 thread0.td_gd = &gd->mi; 865 ncpus = 1; 866 ncpus2 = 1; /* rounded down power of 2 */ 867 ncpus_fit = 1; /* rounded up power of 2 */ 868 /* ncpus2_mask and ncpus_fit_mask are 0 */ 869 init_param1(); 870 gd->mi.gd_prvspace = &CPU_prvspace[0]; 871 mi_gdinit(&gd->mi, 0); 872 cpu_gdinit(gd, 0); 873 mi_proc0init(&gd->mi, proc0paddr); 874 lwp0.lwp_md.md_regs = &proc0_tf; 875 876 init_locks(); 877 cninit(); 878 rand_initialize(); 879 #if 0 /* #ifdef DDB */ 880 kdb_init(); 881 if (boothowto & RB_KDB) 882 Debugger("Boot flags requested debugger"); 883 #endif 884 identcpu(); 885 #if 0 886 initializecpu(); /* Initialize CPU registers */ 887 #endif 888 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE); 889 890 #if 0 891 /* 892 * Map the message buffer 893 */ 894 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE) 895 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off); 896 msgbufinit(msgbufp, MSGBUF_SIZE); 897 #endif 898 #if 0 899 thread0.td_pcb_cr3 ... MMU 900 lwp0.lwp_md.md_regs = &proc0_tf; 901 #endif 902 } 903 904 /* 905 * Filesystem image paths for the virtual kernel are optional. 906 * If specified they each should point to a disk image, 907 * the first of which will become the root disk. 908 * 909 * The virtual kernel caches data from our 'disk' just like a normal kernel, 910 * so we do not really want the real kernel to cache the data too. Use 911 * O_DIRECT to remove the duplication. 912 */ 913 static 914 void 915 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type) 916 { 917 char *serno; 918 int i; 919 920 if (diskFileNum == 0) 921 return; 922 923 for(i=0; i < diskFileNum; i++){ 924 char *fname; 925 fname = diskExp[i]; 926 927 if (fname == NULL) { 928 warnx("Invalid argument to '-r'"); 929 continue; 930 } 931 /* 932 * Check for a serial number for the virtual disk 933 * passed from the command line. 934 */ 935 serno = fname; 936 strsep(&serno, ":"); 937 938 if (DiskNum < VKDISK_MAX) { 939 struct stat st; 940 struct vkdisk_info* info = NULL; 941 int fd; 942 size_t l = 0; 943 944 if (type == VKD_DISK) 945 fd = open(fname, O_RDWR|O_DIRECT, 0644); 946 else 947 fd = open(fname, O_RDONLY|O_DIRECT, 0644); 948 if (fd < 0 || fstat(fd, &st) < 0) { 949 err(1, "Unable to open/create %s", fname); 950 /* NOT REACHED */ 951 } 952 if (S_ISREG(st.st_mode)) { 953 if (flock(fd, LOCK_EX|LOCK_NB) < 0) { 954 errx(1, "Disk image %s is already " 955 "in use\n", fname); 956 /* NOT REACHED */ 957 } 958 } 959 960 info = &DiskInfo[DiskNum]; 961 l = strlen(fname); 962 963 info->unit = i; 964 info->fd = fd; 965 info->type = type; 966 memcpy(info->fname, fname, l); 967 info->serno = NULL; 968 if (serno) { 969 if ((info->serno = malloc(SERNOLEN)) != NULL) 970 strlcpy(info->serno, serno, SERNOLEN); 971 else 972 warnx("Couldn't allocate memory for the operation"); 973 } 974 975 if (DiskNum == 0) { 976 if (type == VKD_CD) { 977 rootdevnames[0] = "cd9660:vcd0a"; 978 } else if (type == VKD_DISK) { 979 rootdevnames[0] = "ufs:vkd0s0a"; 980 rootdevnames[1] = "ufs:vkd0s1a"; 981 } 982 } 983 984 DiskNum++; 985 } else { 986 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname); 987 continue; 988 } 989 } 990 } 991 992 static 993 int 994 netif_set_tapflags(int tap_unit, int f, int s) 995 { 996 struct ifreq ifr; 997 int flags; 998 999 bzero(&ifr, sizeof(ifr)); 1000 1001 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit); 1002 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) { 1003 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit); 1004 return -1; 1005 } 1006 1007 /* 1008 * Adjust if_flags 1009 * 1010 * If the flags are already set/cleared, then we return 1011 * immediately to avoid extra syscalls 1012 */ 1013 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16); 1014 if (f < 0) { 1015 /* Turn off flags */ 1016 f = -f; 1017 if ((flags & f) == 0) 1018 return 0; 1019 flags &= ~f; 1020 } else { 1021 /* Turn on flags */ 1022 if (flags & f) 1023 return 0; 1024 flags |= f; 1025 } 1026 1027 /* 1028 * Fix up ifreq.ifr_name, since it may be trashed 1029 * in previous ioctl(SIOCGIFFLAGS) 1030 */ 1031 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit); 1032 1033 ifr.ifr_flags = flags & 0xffff; 1034 ifr.ifr_flagshigh = flags >> 16; 1035 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) { 1036 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit); 1037 return -1; 1038 } 1039 return 0; 1040 } 1041 1042 static 1043 int 1044 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s) 1045 { 1046 struct ifaliasreq ifra; 1047 struct sockaddr_in *in; 1048 1049 bzero(&ifra, sizeof(ifra)); 1050 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit); 1051 1052 /* Setup address */ 1053 in = (struct sockaddr_in *)&ifra.ifra_addr; 1054 in->sin_family = AF_INET; 1055 in->sin_len = sizeof(*in); 1056 in->sin_addr.s_addr = addr; 1057 1058 if (mask != 0) { 1059 /* Setup netmask */ 1060 in = (struct sockaddr_in *)&ifra.ifra_mask; 1061 in->sin_len = sizeof(*in); 1062 in->sin_addr.s_addr = mask; 1063 } 1064 1065 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) { 1066 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit); 1067 return -1; 1068 } 1069 return 0; 1070 } 1071 1072 static 1073 int 1074 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s) 1075 { 1076 struct ifbreq ifbr; 1077 struct ifdrv ifd; 1078 1079 bzero(&ifbr, sizeof(ifbr)); 1080 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname), 1081 "tap%d", tap_unit); 1082 1083 bzero(&ifd, sizeof(ifd)); 1084 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name)); 1085 ifd.ifd_cmd = BRDGADD; 1086 ifd.ifd_len = sizeof(ifbr); 1087 ifd.ifd_data = &ifbr; 1088 1089 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) { 1090 /* 1091 * 'errno == EEXIST' means that the tap(4) is already 1092 * a member of the bridge(4) 1093 */ 1094 if (errno != EEXIST) { 1095 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge); 1096 return -1; 1097 } 1098 } 1099 return 0; 1100 } 1101 1102 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK) 1103 1104 /* 1105 * Locate the first unused tap(4) device file if auto mode is requested, 1106 * or open the user supplied device file, and bring up the corresponding 1107 * tap(4) interface. 1108 * 1109 * NOTE: Only tap(4) device file is supported currently 1110 */ 1111 static 1112 int 1113 netif_open_tap(const char *netif, int *tap_unit, int s) 1114 { 1115 char tap_dev[MAXPATHLEN]; 1116 int tap_fd, failed; 1117 struct stat st; 1118 char *dname; 1119 1120 *tap_unit = -1; 1121 1122 if (strcmp(netif, "auto") == 0) { 1123 /* 1124 * Find first unused tap(4) device file 1125 */ 1126 tap_fd = open("/dev/tap", TAPDEV_OFLAGS); 1127 if (tap_fd < 0) { 1128 warnc(errno, "Unable to find a free tap(4)"); 1129 return -1; 1130 } 1131 } else { 1132 /* 1133 * User supplied tap(4) device file or unix socket. 1134 */ 1135 if (netif[0] == '/') /* Absolute path */ 1136 strlcpy(tap_dev, netif, sizeof(tap_dev)); 1137 else 1138 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif); 1139 1140 tap_fd = open(tap_dev, TAPDEV_OFLAGS); 1141 1142 /* 1143 * If we cannot open normally try to connect to it. 1144 */ 1145 if (tap_fd < 0) 1146 tap_fd = unix_connect(tap_dev); 1147 1148 if (tap_fd < 0) { 1149 warn("Unable to open %s", tap_dev); 1150 return -1; 1151 } 1152 } 1153 1154 /* 1155 * Check whether the device file is a tap(4) 1156 */ 1157 if (fstat(tap_fd, &st) < 0) { 1158 failed = 1; 1159 } else if (S_ISCHR(st.st_mode)) { 1160 dname = fdevname(tap_fd); 1161 if (dname) 1162 dname = strstr(dname, "tap"); 1163 if (dname) { 1164 /* 1165 * Bring up the corresponding tap(4) interface 1166 */ 1167 *tap_unit = strtol(dname + 3, NULL, 10); 1168 printf("TAP UNIT %d\n", *tap_unit); 1169 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0) 1170 failed = 0; 1171 else 1172 failed = 1; 1173 } else { 1174 failed = 1; 1175 } 1176 } else if (S_ISSOCK(st.st_mode)) { 1177 /* 1178 * Special socket connection (typically to vknet). We 1179 * do not have to do anything. 1180 */ 1181 failed = 0; 1182 } else { 1183 failed = 1; 1184 } 1185 1186 if (failed) { 1187 warnx("%s is not a tap(4) device or socket", tap_dev); 1188 close(tap_fd); 1189 tap_fd = -1; 1190 *tap_unit = -1; 1191 } 1192 return tap_fd; 1193 } 1194 1195 static int 1196 unix_connect(const char *path) 1197 { 1198 struct sockaddr_un sunx; 1199 int len; 1200 int net_fd; 1201 int sndbuf = 262144; 1202 struct stat st; 1203 1204 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path); 1205 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]); 1206 ++len; /* include nul */ 1207 sunx.sun_family = AF_UNIX; 1208 sunx.sun_len = len; 1209 1210 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0); 1211 if (net_fd < 0) 1212 return(-1); 1213 if (connect(net_fd, (void *)&sunx, len) < 0) { 1214 close(net_fd); 1215 return(-1); 1216 } 1217 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf)); 1218 if (fstat(net_fd, &st) == 0) 1219 printf("Network socket buffer: %d bytes\n", st.st_blksize); 1220 fcntl(net_fd, F_SETFL, O_NONBLOCK); 1221 return(net_fd); 1222 } 1223 1224 #undef TAPDEV_MAJOR 1225 #undef TAPDEV_MINOR 1226 #undef TAPDEV_OFLAGS 1227 1228 /* 1229 * Following syntax is supported, 1230 * 1) x.x.x.x tap(4)'s address is x.x.x.x 1231 * 1232 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x 1233 * tap(4)'s netmask len is z 1234 * 1235 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x 1236 * pseudo netif's address is y.y.y.y 1237 * 1238 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x 1239 * pseudo netif's address is y.y.y.y 1240 * tap(4) and pseudo netif's netmask len are z 1241 * 1242 * 5) bridgeX tap(4) will be added to bridgeX 1243 * 1244 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX 1245 * pseudo netif's address is y.y.y.y 1246 * 1247 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX 1248 * pseudo netif's address is y.y.y.y 1249 * pseudo netif's netmask len is z 1250 */ 1251 static 1252 int 1253 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s) 1254 { 1255 in_addr_t tap_addr, netmask, netif_addr; 1256 int next_netif_addr; 1257 char *tok, *masklen_str, *ifbridge; 1258 1259 *addr = 0; 1260 *mask = 0; 1261 1262 tok = strtok(NULL, ":/"); 1263 if (tok == NULL) { 1264 /* 1265 * Nothing special, simply use tap(4) as backend 1266 */ 1267 return 0; 1268 } 1269 1270 if (inet_pton(AF_INET, tok, &tap_addr) > 0) { 1271 /* 1272 * tap(4)'s address is supplied 1273 */ 1274 ifbridge = NULL; 1275 1276 /* 1277 * If there is next token, then it may be pseudo 1278 * netif's address or netmask len for tap(4) 1279 */ 1280 next_netif_addr = 0; 1281 } else { 1282 /* 1283 * Not tap(4)'s address, assume it as a bridge(4) 1284 * iface name 1285 */ 1286 tap_addr = 0; 1287 ifbridge = tok; 1288 1289 /* 1290 * If there is next token, then it must be pseudo 1291 * netif's address 1292 */ 1293 next_netif_addr = 1; 1294 } 1295 1296 netmask = netif_addr = 0; 1297 1298 tok = strtok(NULL, ":/"); 1299 if (tok == NULL) 1300 goto back; 1301 1302 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) { 1303 if (next_netif_addr) { 1304 warnx("Invalid pseudo netif address: %s", tok); 1305 return -1; 1306 } 1307 netif_addr = 0; 1308 1309 /* 1310 * Current token is not address, then it must be netmask len 1311 */ 1312 masklen_str = tok; 1313 } else { 1314 /* 1315 * Current token is pseudo netif address, if there is next token 1316 * it must be netmask len 1317 */ 1318 masklen_str = strtok(NULL, "/"); 1319 } 1320 1321 /* Calculate netmask */ 1322 if (masklen_str != NULL) { 1323 u_long masklen; 1324 1325 masklen = strtoul(masklen_str, NULL, 10); 1326 if (masklen < 32 && masklen > 0) { 1327 netmask = htonl(~((1LL << (32 - masklen)) - 1) 1328 & 0xffffffff); 1329 } else { 1330 warnx("Invalid netmask len: %lu", masklen); 1331 return -1; 1332 } 1333 } 1334 1335 /* Make sure there is no more token left */ 1336 if (strtok(NULL, ":/") != NULL) { 1337 warnx("Invalid argument to '-I'"); 1338 return -1; 1339 } 1340 1341 back: 1342 if (tap_unit < 0) { 1343 /* Do nothing */ 1344 } else if (ifbridge == NULL) { 1345 /* Set tap(4) address/netmask */ 1346 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0) 1347 return -1; 1348 } else { 1349 /* Tie tap(4) to bridge(4) */ 1350 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0) 1351 return -1; 1352 } 1353 1354 *addr = netif_addr; 1355 *mask = netmask; 1356 return 0; 1357 } 1358 1359 /* 1360 * NetifInfo[] will be filled for pseudo netif initialization. 1361 * NetifNum will be bumped to reflect the number of valid entries 1362 * in NetifInfo[]. 1363 */ 1364 static 1365 void 1366 init_netif(char *netifExp[], int netifExpNum) 1367 { 1368 int i, s; 1369 char *tmp; 1370 1371 if (netifExpNum == 0) 1372 return; 1373 1374 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */ 1375 if (s < 0) 1376 return; 1377 1378 for (i = 0; i < netifExpNum; ++i) { 1379 struct vknetif_info *info; 1380 in_addr_t netif_addr, netif_mask; 1381 int tap_fd, tap_unit; 1382 char *netif; 1383 1384 /* Extract MAC address if there is one */ 1385 tmp = netifExp[i]; 1386 strsep(&tmp, "="); 1387 1388 netif = strtok(netifExp[i], ":"); 1389 if (netif == NULL) { 1390 warnx("Invalid argument to '-I'"); 1391 continue; 1392 } 1393 1394 /* 1395 * Open tap(4) device file and bring up the 1396 * corresponding interface 1397 */ 1398 tap_fd = netif_open_tap(netif, &tap_unit, s); 1399 if (tap_fd < 0) 1400 continue; 1401 1402 /* 1403 * Initialize tap(4) and get address/netmask 1404 * for pseudo netif 1405 * 1406 * NB: Rest part of netifExp[i] is passed 1407 * to netif_init_tap() implicitly. 1408 */ 1409 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) { 1410 /* 1411 * NB: Closing tap(4) device file will bring 1412 * down the corresponding interface 1413 */ 1414 close(tap_fd); 1415 continue; 1416 } 1417 1418 info = &NetifInfo[NetifNum]; 1419 bzero(info, sizeof(*info)); 1420 info->tap_fd = tap_fd; 1421 info->tap_unit = tap_unit; 1422 info->netif_addr = netif_addr; 1423 info->netif_mask = netif_mask; 1424 /* 1425 * If tmp isn't NULL it means a MAC could have been 1426 * specified so attempt to convert it. 1427 * Setting enaddr to NULL will tell vke_attach() we 1428 * need a pseudo-random MAC address. 1429 */ 1430 if (tmp != NULL) { 1431 if ((info->enaddr = malloc(ETHER_ADDR_LEN)) == NULL) 1432 warnx("Couldn't allocate memory for the operation"); 1433 else { 1434 if ((kether_aton(tmp, info->enaddr)) == NULL) { 1435 free(info->enaddr); 1436 info->enaddr = NULL; 1437 } 1438 } 1439 } 1440 1441 NetifNum++; 1442 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */ 1443 break; 1444 } 1445 close(s); 1446 } 1447 1448 /* 1449 * Create the pid file and leave it open and locked while the vkernel is 1450 * running. This allows a script to use /usr/bin/lockf to probe whether 1451 * a vkernel is still running (so as not to accidently kill an unrelated 1452 * process from a stale pid file). 1453 */ 1454 static 1455 void 1456 writepid(void) 1457 { 1458 char buf[32]; 1459 int fd; 1460 1461 if (pid_file != NULL) { 1462 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid()); 1463 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666); 1464 if (fd < 0) { 1465 if (errno == EWOULDBLOCK) { 1466 perror("Failed to lock pidfile, " 1467 "vkernel already running"); 1468 } else { 1469 perror("Failed to create pidfile"); 1470 } 1471 exit(EX_SOFTWARE); 1472 } 1473 ftruncate(fd, 0); 1474 write(fd, buf, strlen(buf)); 1475 /* leave the file open to maintain the lock */ 1476 } 1477 } 1478 1479 static 1480 void 1481 cleanpid( void ) 1482 { 1483 if (pid_file != NULL) { 1484 if (unlink(pid_file) < 0) 1485 perror("Warning: couldn't remove pidfile"); 1486 } 1487 } 1488 1489 static 1490 void 1491 usage_err(const char *ctl, ...) 1492 { 1493 va_list va; 1494 1495 va_start(va, ctl); 1496 vfprintf(stderr, ctl, va); 1497 va_end(va); 1498 fprintf(stderr, "\n"); 1499 exit(EX_USAGE); 1500 } 1501 1502 static 1503 void 1504 usage_help(_Bool help) 1505 { 1506 fprintf(stderr, "Usage: %s [-hsUvd] [-c file] [-e name=value:name=value:...]\n" 1507 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n" 1508 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n" 1509 "\t[-p file] [-r file]\n", save_av[0]); 1510 1511 if (help) 1512 fprintf(stderr, "\nArguments:\n" 1513 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n" 1514 "\t-e\tSpecify an environment to be used by the kernel.\n" 1515 "\t-h\tThis list of options.\n" 1516 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n" 1517 "\t-I\tCreate a virtual network device.\n" 1518 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n" 1519 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n" 1520 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n" 1521 "\t \t- numcpus - number of cpus\n" 1522 "\t \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n" 1523 "\t \t the logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n" 1524 "\t \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n" 1525 "\t \t the core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n" 1526 "\t-p\tSpecify a file in which to store the process ID.\n" 1527 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n" 1528 "\t-s\tBoot into single-user mode.\n" 1529 "\t-U\tEnable writing to kernel memory and module loading.\n" 1530 "\t-v\tTurn on verbose booting.\n"); 1531 1532 exit(EX_USAGE); 1533 } 1534 1535 void 1536 cpu_reset(void) 1537 { 1538 kprintf("cpu reset, rebooting vkernel\n"); 1539 closefrom(3); 1540 cleanpid(); 1541 exit(EX_VKERNEL_REBOOT); 1542 1543 } 1544 1545 void 1546 cpu_halt(void) 1547 { 1548 kprintf("cpu halt, exiting vkernel\n"); 1549 cleanpid(); 1550 exit(EX_OK); 1551 } 1552 1553 void 1554 setrealcpu(void) 1555 { 1556 switch(lwp_cpu_lock) { 1557 case LCL_PER_CPU: 1558 if (bootverbose) 1559 kprintf("Locking CPU%d to real cpu %d\n", 1560 mycpuid, next_cpu); 1561 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu)); 1562 next_cpu++; 1563 if (next_cpu >= real_ncpus) 1564 next_cpu = 0; 1565 break; 1566 case LCL_SINGLE_CPU: 1567 if (bootverbose) 1568 kprintf("Locking CPU%d to real cpu %d\n", 1569 mycpuid, next_cpu); 1570 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu)); 1571 break; 1572 default: 1573 /* do not map virtual cpus to real cpus */ 1574 break; 1575 } 1576 } 1577 1578 /* 1579 * Allocate and free memory for module loading. The loaded module 1580 * has to be placed somewhere near the current kernel binary load 1581 * point or the relocations will not work. 1582 * 1583 * I'm not sure why this isn't working. 1584 */ 1585 int 1586 vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes) 1587 { 1588 #if 1 1589 size_t xtra; 1590 xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK; 1591 *basep = (vm_offset_t)sbrk(xtra + bytes) + xtra; 1592 bzero((void *)*basep, bytes); 1593 #else 1594 *basep = (vm_offset_t)mmap((void *)0x000000000, bytes, 1595 PROT_READ|PROT_WRITE|PROT_EXEC, 1596 MAP_ANON|MAP_SHARED, -1, 0); 1597 if ((void *)*basep == MAP_FAILED) 1598 return ENOMEM; 1599 #endif 1600 return 0; 1601 } 1602 1603 void 1604 vkernel_module_memory_free(vm_offset_t base, size_t bytes) 1605 { 1606 #if 0 1607 #if 0 1608 munmap((void *)base, bytes); 1609 #endif 1610 #endif 1611 } 1612