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