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