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 56 #include <machine/cpu.h> 57 #include <machine/globaldata.h> 58 #include <machine/tls.h> 59 #include <machine/md_var.h> 60 #include <machine/vmparam.h> 61 #include <cpu/specialreg.h> 62 63 #include <net/if.h> 64 #include <net/if_arp.h> 65 #include <net/ethernet.h> 66 #include <net/bridge/if_bridgevar.h> 67 #include <netinet/in.h> 68 #include <arpa/inet.h> 69 70 #include <stdio.h> 71 #include <stdlib.h> 72 #include <stdarg.h> 73 #include <stdbool.h> 74 #include <unistd.h> 75 #include <fcntl.h> 76 #include <string.h> 77 #include <err.h> 78 #include <errno.h> 79 #include <assert.h> 80 #include <sysexits.h> 81 82 vm_paddr_t phys_avail[16]; 83 vm_paddr_t Maxmem; 84 vm_paddr_t Maxmem_bytes; 85 long physmem; 86 int MemImageFd = -1; 87 struct vkdisk_info DiskInfo[VKDISK_MAX]; 88 int DiskNum; 89 struct vknetif_info NetifInfo[VKNETIF_MAX]; 90 int NetifNum; 91 char *pid_file; 92 vm_offset_t KvaStart; 93 vm_offset_t KvaEnd; 94 vm_offset_t KvaSize; 95 vm_offset_t virtual_start; 96 vm_offset_t virtual_end; 97 vm_offset_t virtual2_start; 98 vm_offset_t virtual2_end; 99 vm_offset_t kernel_vm_end; 100 vm_offset_t crashdumpmap; 101 vm_offset_t clean_sva; 102 vm_offset_t clean_eva; 103 struct msgbuf *msgbufp; 104 caddr_t ptvmmap; 105 vpte_t *KernelPTD; 106 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */ 107 void *dmap_min_address; 108 u_int cpu_feature; /* XXX */ 109 int tsc_present; 110 int64_t tsc_frequency; 111 int optcpus; /* number of cpus - see mp_start() */ 112 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */ 113 int real_ncpus; /* number of real CPUs */ 114 int next_cpu; /* next real CPU to lock a virtual CPU to */ 115 int vkernel_b_arg; /* -b argument - no of logical CPU bits - only SMP */ 116 int vkernel_B_arg; /* -B argument - no of core bits - only SMP */ 117 118 struct privatespace *CPU_prvspace; 119 120 static struct trapframe proc0_tf; 121 static void *proc0paddr; 122 123 static void init_sys_memory(char *imageFile); 124 static void init_kern_memory(void); 125 static void init_globaldata(void); 126 static void init_vkernel(void); 127 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type); 128 static void init_netif(char *netifExp[], int netifFileNum); 129 static void writepid(void); 130 static void cleanpid(void); 131 static int unix_connect(const char *path); 132 static void usage_err(const char *ctl, ...); 133 static void usage_help(_Bool); 134 static void init_locks(void); 135 136 static int save_ac; 137 static char **save_av; 138 139 /* 140 * Kernel startup for virtual kernels - standard main() 141 */ 142 int 143 main(int ac, char **av) 144 { 145 char *memImageFile = NULL; 146 char *netifFile[VKNETIF_MAX]; 147 char *diskFile[VKDISK_MAX]; 148 char *cdFile[VKDISK_MAX]; 149 char *suffix; 150 char *endp; 151 char *tmp; 152 char *tok; 153 int netifFileNum = 0; 154 int diskFileNum = 0; 155 int cdFileNum = 0; 156 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */ 157 int c; 158 int i; 159 int j; 160 int n; 161 int isq; 162 int pos; 163 int eflag; 164 int real_vkernel_enable; 165 int supports_sse; 166 size_t vsize; 167 size_t kenv_size; 168 size_t kenv_size2; 169 170 save_ac = ac; 171 save_av = av; 172 eflag = 0; 173 pos = 0; 174 kenv_size = 0; 175 176 /* 177 * Process options 178 */ 179 kernel_mem_readonly = 1; 180 optcpus = 2; 181 vkernel_b_arg = 0; 182 vkernel_B_arg = 0; 183 lwp_cpu_lock = LCL_NONE; 184 185 real_vkernel_enable = 0; 186 vsize = sizeof(real_vkernel_enable); 187 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0); 188 189 if (real_vkernel_enable == 0) { 190 errx(1, "vm.vkernel_enable is 0, must be set " 191 "to 1 to execute a vkernel!"); 192 } 193 194 real_ncpus = 1; 195 vsize = sizeof(real_ncpus); 196 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0); 197 198 if (ac < 2) 199 usage_help(false); 200 201 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:Ub:B:")) != -1) { 202 switch(c) { 203 case 'e': 204 /* 205 * name=value:name=value:name=value... 206 * name="value"... 207 * 208 * Allow values to be quoted but note that shells 209 * may remove the quotes, so using this feature 210 * to embed colons may require a backslash. 211 */ 212 n = strlen(optarg); 213 isq = 0; 214 215 if (eflag == 0) { 216 kenv_size = n + 2; 217 kern_envp = malloc(kenv_size); 218 if (kern_envp == NULL) 219 errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size); 220 } else { 221 kenv_size2 = kenv_size + n + 1; 222 pos = kenv_size - 1; 223 if ((tmp = realloc(kern_envp, kenv_size2)) == NULL) 224 errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2); 225 kern_envp = tmp; 226 kenv_size = kenv_size2; 227 } 228 229 for (i = 0, j = pos; i < n; ++i) { 230 if (optarg[i] == '"') 231 isq ^= 1; 232 else if (optarg[i] == '\'') 233 isq ^= 2; 234 else if (isq == 0 && optarg[i] == ':') 235 kern_envp[j++] = 0; 236 else 237 kern_envp[j++] = optarg[i]; 238 } 239 kern_envp[j++] = 0; 240 kern_envp[j++] = 0; 241 eflag++; 242 break; 243 case 's': 244 boothowto |= RB_SINGLE; 245 break; 246 case 'v': 247 bootverbose = 1; 248 break; 249 case 'i': 250 memImageFile = optarg; 251 break; 252 case 'I': 253 if (netifFileNum < VKNETIF_MAX) 254 netifFile[netifFileNum++] = strdup(optarg); 255 break; 256 case 'r': 257 if (bootOnDisk < 0) 258 bootOnDisk = 1; 259 if (diskFileNum + cdFileNum < VKDISK_MAX) 260 diskFile[diskFileNum++] = strdup(optarg); 261 break; 262 case 'c': 263 if (bootOnDisk < 0) 264 bootOnDisk = 0; 265 if (diskFileNum + cdFileNum < VKDISK_MAX) 266 cdFile[cdFileNum++] = strdup(optarg); 267 break; 268 case 'm': 269 Maxmem_bytes = strtoull(optarg, &suffix, 0); 270 if (suffix) { 271 switch(*suffix) { 272 case 'g': 273 case 'G': 274 Maxmem_bytes <<= 30; 275 break; 276 case 'm': 277 case 'M': 278 Maxmem_bytes <<= 20; 279 break; 280 case 'k': 281 case 'K': 282 Maxmem_bytes <<= 10; 283 break; 284 default: 285 Maxmem_bytes = 0; 286 usage_err("Bad maxmem option"); 287 /* NOT REACHED */ 288 break; 289 } 290 } 291 break; 292 case 'l': 293 next_cpu = -1; 294 if (strncmp("map", optarg, 3) == 0) { 295 lwp_cpu_lock = LCL_PER_CPU; 296 if (optarg[3] == ',') { 297 next_cpu = strtol(optarg+4, &endp, 0); 298 if (*endp != '\0') 299 usage_err("Bad target CPU number at '%s'", endp); 300 } else { 301 next_cpu = 0; 302 } 303 if (next_cpu < 0 || next_cpu > real_ncpus - 1) 304 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1); 305 } else if (strncmp("any", optarg, 3) == 0) { 306 lwp_cpu_lock = LCL_NONE; 307 } else { 308 lwp_cpu_lock = LCL_SINGLE_CPU; 309 next_cpu = strtol(optarg, &endp, 0); 310 if (*endp != '\0') 311 usage_err("Bad target CPU number at '%s'", endp); 312 if (next_cpu < 0 || next_cpu > real_ncpus - 1) 313 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1); 314 } 315 break; 316 case 'n': 317 /* 318 * This value is set up by mp_start(), don't just 319 * set ncpus here. 320 */ 321 tok = strtok(optarg, ":"); 322 optcpus = strtol(tok, NULL, 0); 323 if (optcpus < 1 || optcpus > MAXCPU) 324 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU); 325 326 /* :lbits argument */ 327 tok = strtok(NULL, ":"); 328 if (tok != NULL) { 329 vkernel_b_arg = strtol(tok, NULL, 0); 330 331 /* :cbits argument */ 332 tok = strtok(NULL, ":"); 333 if (tok != NULL) { 334 vkernel_B_arg = strtol(tok, NULL, 0); 335 } 336 337 } 338 break; 339 case 'p': 340 pid_file = optarg; 341 break; 342 case 'U': 343 kernel_mem_readonly = 0; 344 break; 345 case 'h': 346 usage_help(true); 347 break; 348 default: 349 usage_help(false); 350 } 351 } 352 353 writepid(); 354 cpu_disable_intr(); 355 init_sys_memory(memImageFile); 356 init_kern_memory(); 357 init_globaldata(); 358 init_vkernel(); 359 setrealcpu(); 360 init_kqueue(); 361 362 vmm_guest = 1; 363 364 /* 365 * Check TSC 366 */ 367 vsize = sizeof(tsc_present); 368 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0); 369 vsize = sizeof(tsc_frequency); 370 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0); 371 if (tsc_present) 372 cpu_feature |= CPUID_TSC; 373 374 /* 375 * Check SSE 376 */ 377 vsize = sizeof(supports_sse); 378 supports_sse = 0; 379 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0); 380 init_fpu(supports_sse); 381 if (supports_sse) 382 cpu_feature |= CPUID_SSE | CPUID_FXSR; 383 384 /* 385 * We boot from the first installed disk. 386 */ 387 if (bootOnDisk == 1) { 388 init_disk(diskFile, diskFileNum, VKD_DISK); 389 init_disk(cdFile, cdFileNum, VKD_CD); 390 } else { 391 init_disk(cdFile, cdFileNum, VKD_CD); 392 init_disk(diskFile, diskFileNum, VKD_DISK); 393 } 394 init_netif(netifFile, netifFileNum); 395 init_exceptions(); 396 mi_startup(); 397 /* NOT REACHED */ 398 exit(EX_SOFTWARE); 399 } 400 401 /* 402 * Initialize system memory. This is the virtual kernel's 'RAM'. 403 */ 404 static 405 void 406 init_sys_memory(char *imageFile) 407 { 408 struct stat st; 409 int i; 410 int fd; 411 412 /* 413 * Figure out the system memory image size. If an image file was 414 * specified and -m was not specified, use the image file's size. 415 */ 416 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0) 417 Maxmem_bytes = (vm_paddr_t)st.st_size; 418 if ((imageFile == NULL || stat(imageFile, &st) < 0) && 419 Maxmem_bytes == 0) { 420 errx(1, "Cannot create new memory file %s unless " 421 "system memory size is specified with -m", 422 imageFile); 423 /* NOT REACHED */ 424 } 425 426 /* 427 * Maxmem must be known at this time 428 */ 429 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) { 430 errx(1, "Bad maxmem specification: 64MB minimum, " 431 "multiples of %dMB only", 432 SEG_SIZE / 1024 / 1024); 433 /* NOT REACHED */ 434 } 435 436 /* 437 * Generate an image file name if necessary, then open/create the 438 * file exclusively locked. Do not allow multiple virtual kernels 439 * to use the same image file. 440 * 441 * Don't iterate through a million files if we do not have write 442 * access to the directory, stop if our open() failed on a 443 * non-existant file. Otherwise opens can fail for any number 444 */ 445 if (imageFile == NULL) { 446 for (i = 0; i < 1000000; ++i) { 447 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i); 448 fd = open(imageFile, 449 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644); 450 if (fd < 0 && stat(imageFile, &st) == 0) { 451 free(imageFile); 452 continue; 453 } 454 break; 455 } 456 } else { 457 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644); 458 } 459 fprintf(stderr, "Using memory file: %s\n", imageFile); 460 if (fd < 0 || fstat(fd, &st) < 0) { 461 err(1, "Unable to open/create %s", imageFile); 462 /* NOT REACHED */ 463 } 464 465 /* 466 * Truncate or extend the file as necessary. Clean out the contents 467 * of the file, we want it to be full of holes so we don't waste 468 * time reading in data from an old file that we no longer care 469 * about. 470 */ 471 ftruncate(fd, 0); 472 ftruncate(fd, Maxmem_bytes); 473 474 MemImageFd = fd; 475 Maxmem = Maxmem_bytes >> PAGE_SHIFT; 476 physmem = Maxmem; 477 } 478 479 /* 480 * Initialize kernel memory. This reserves kernel virtual memory by using 481 * MAP_VPAGETABLE 482 */ 483 484 static 485 void 486 init_kern_memory(void) 487 { 488 void *base; 489 void *try; 490 char dummy; 491 char *topofstack = &dummy; 492 int i; 493 void *firstfree; 494 495 /* 496 * Memory map our kernel virtual memory space. Note that the 497 * kernel image itself is not made part of this memory for the 498 * moment. 499 * 500 * The memory map must be segment-aligned so we can properly 501 * offset KernelPTD. 502 * 503 * If the system kernel has a different MAXDSIZ, it might not 504 * be possible to map kernel memory in its prefered location. 505 * Try a number of different locations. 506 */ 507 try = (void *)(512UL << 30); 508 base = NULL; 509 while ((char *)try + KERNEL_KVA_SIZE < topofstack) { 510 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE, 511 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE, 512 MemImageFd, (off_t)try); 513 if (base == try) 514 break; 515 if (base != MAP_FAILED) 516 munmap(base, KERNEL_KVA_SIZE); 517 try = (char *)try + (512UL << 30); 518 } 519 if (base != try) { 520 err(1, "Unable to mmap() kernel virtual memory!"); 521 /* NOT REACHED */ 522 } 523 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC); 524 KvaStart = (vm_offset_t)base; 525 KvaSize = KERNEL_KVA_SIZE; 526 KvaEnd = KvaStart + KvaSize; 527 528 /* cannot use kprintf yet */ 529 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd); 530 531 /* MAP_FILE? */ 532 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE, 533 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED, 534 MemImageFd, 0); 535 if (dmap_min_address == MAP_FAILED) { 536 err(1, "Unable to mmap() kernel DMAP region!"); 537 /* NOT REACHED */ 538 } 539 540 firstfree = NULL; 541 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base); 542 543 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP, 544 0 | VPTE_R | VPTE_W | VPTE_V); 545 546 /* 547 * phys_avail[] represents unallocated physical memory. MI code 548 * will use phys_avail[] to create the vm_page array. 549 */ 550 phys_avail[0] = (vm_paddr_t)firstfree; 551 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK; 552 phys_avail[1] = Maxmem_bytes; 553 554 #if JGV 555 /* 556 * (virtual_start, virtual_end) represent unallocated kernel virtual 557 * memory. MI code will create kernel_map using these parameters. 558 */ 559 virtual_start = KvaStart + (long)firstfree; 560 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK; 561 virtual_end = KvaStart + KERNEL_KVA_SIZE; 562 #endif 563 564 /* 565 * pmap_growkernel() will set the correct value. 566 */ 567 kernel_vm_end = 0; 568 569 /* 570 * Allocate space for process 0's UAREA. 571 */ 572 proc0paddr = (void *)virtual_start; 573 for (i = 0; i < UPAGES; ++i) { 574 pmap_kenter_quick(virtual_start, phys_avail[0]); 575 virtual_start += PAGE_SIZE; 576 phys_avail[0] += PAGE_SIZE; 577 } 578 579 /* 580 * crashdumpmap 581 */ 582 crashdumpmap = virtual_start; 583 virtual_start += MAXDUMPPGS * PAGE_SIZE; 584 585 /* 586 * msgbufp maps the system message buffer 587 */ 588 assert((MSGBUF_SIZE & PAGE_MASK) == 0); 589 msgbufp = (void *)virtual_start; 590 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) { 591 pmap_kenter_quick(virtual_start, phys_avail[0]); 592 virtual_start += PAGE_SIZE; 593 phys_avail[0] += PAGE_SIZE; 594 } 595 msgbufinit(msgbufp, MSGBUF_SIZE); 596 597 /* 598 * used by kern_memio for /dev/mem access 599 */ 600 ptvmmap = (caddr_t)virtual_start; 601 virtual_start += PAGE_SIZE; 602 603 /* 604 * Bootstrap the kernel_pmap 605 */ 606 #if JGV 607 pmap_bootstrap(); 608 #endif 609 } 610 611 /* 612 * Map the per-cpu globaldata for cpu #0. Allocate the space using 613 * virtual_start and phys_avail[0] 614 */ 615 static 616 void 617 init_globaldata(void) 618 { 619 int i; 620 vm_paddr_t pa; 621 vm_offset_t va; 622 623 /* 624 * Reserve enough KVA to cover possible cpus. This is a considerable 625 * amount of KVA since the privatespace structure includes two 626 * whole page table mappings. 627 */ 628 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK; 629 CPU_prvspace = (void *)virtual_start; 630 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU; 631 632 /* 633 * Allocate enough physical memory to cover the mdglobaldata 634 * portion of the space and the idle stack and map the pages 635 * into KVA. For cpu #0 only. 636 */ 637 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) { 638 pa = phys_avail[0]; 639 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i; 640 pmap_kenter_quick(va, pa); 641 phys_avail[0] += PAGE_SIZE; 642 } 643 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) { 644 pa = phys_avail[0]; 645 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i; 646 pmap_kenter_quick(va, pa); 647 phys_avail[0] += PAGE_SIZE; 648 } 649 650 /* 651 * Setup the %gs for cpu #0. The mycpu macro works after this 652 * point. Note that %fs is used by pthreads. 653 */ 654 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace)); 655 } 656 657 658 /* 659 * Initialize pool tokens and other necessary locks 660 */ 661 static void 662 init_locks(void) 663 { 664 665 /* 666 * Get the initial mplock with a count of 1 for the BSP. 667 * This uses a LOGICAL cpu ID, ie BSP == 0. 668 */ 669 cpu_get_initial_mplock(); 670 671 /* our token pool needs to work early */ 672 lwkt_token_pool_init(); 673 674 } 675 676 677 /* 678 * Initialize very low level systems including thread0, proc0, etc. 679 */ 680 static 681 void 682 init_vkernel(void) 683 { 684 struct mdglobaldata *gd; 685 686 gd = &CPU_prvspace[0].mdglobaldata; 687 bzero(gd, sizeof(*gd)); 688 689 gd->mi.gd_curthread = &thread0; 690 thread0.td_gd = &gd->mi; 691 ncpus = 1; 692 ncpus2 = 1; /* rounded down power of 2 */ 693 ncpus_fit = 1; /* rounded up power of 2 */ 694 /* ncpus2_mask and ncpus_fit_mask are 0 */ 695 init_param1(); 696 gd->mi.gd_prvspace = &CPU_prvspace[0]; 697 mi_gdinit(&gd->mi, 0); 698 cpu_gdinit(gd, 0); 699 mi_proc0init(&gd->mi, proc0paddr); 700 lwp0.lwp_md.md_regs = &proc0_tf; 701 702 init_locks(); 703 cninit(); 704 rand_initialize(); 705 #if 0 /* #ifdef DDB */ 706 kdb_init(); 707 if (boothowto & RB_KDB) 708 Debugger("Boot flags requested debugger"); 709 #endif 710 identcpu(); 711 #if 0 712 initializecpu(); /* Initialize CPU registers */ 713 #endif 714 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE); 715 716 #if 0 717 /* 718 * Map the message buffer 719 */ 720 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE) 721 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off); 722 msgbufinit(msgbufp, MSGBUF_SIZE); 723 #endif 724 #if 0 725 thread0.td_pcb_cr3 ... MMU 726 lwp0.lwp_md.md_regs = &proc0_tf; 727 #endif 728 } 729 730 /* 731 * Filesystem image paths for the virtual kernel are optional. 732 * If specified they each should point to a disk image, 733 * the first of which will become the root disk. 734 * 735 * The virtual kernel caches data from our 'disk' just like a normal kernel, 736 * so we do not really want the real kernel to cache the data too. Use 737 * O_DIRECT to remove the duplication. 738 */ 739 static 740 void 741 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type) 742 { 743 int i; 744 745 if (diskFileNum == 0) 746 return; 747 748 for(i=0; i < diskFileNum; i++){ 749 char *fname; 750 fname = diskExp[i]; 751 752 if (fname == NULL) { 753 warnx("Invalid argument to '-r'"); 754 continue; 755 } 756 757 if (DiskNum < VKDISK_MAX) { 758 struct stat st; 759 struct vkdisk_info* info = NULL; 760 int fd; 761 size_t l = 0; 762 763 if (type == VKD_DISK) 764 fd = open(fname, O_RDWR|O_DIRECT, 0644); 765 else 766 fd = open(fname, O_RDONLY|O_DIRECT, 0644); 767 if (fd < 0 || fstat(fd, &st) < 0) { 768 err(1, "Unable to open/create %s", fname); 769 /* NOT REACHED */ 770 } 771 if (S_ISREG(st.st_mode)) { 772 if (flock(fd, LOCK_EX|LOCK_NB) < 0) { 773 errx(1, "Disk image %s is already " 774 "in use\n", fname); 775 /* NOT REACHED */ 776 } 777 } 778 779 info = &DiskInfo[DiskNum]; 780 l = strlen(fname); 781 782 info->unit = i; 783 info->fd = fd; 784 info->type = type; 785 memcpy(info->fname, fname, l); 786 787 if (DiskNum == 0) { 788 if (type == VKD_CD) { 789 rootdevnames[0] = "cd9660:vcd0a"; 790 } else if (type == VKD_DISK) { 791 rootdevnames[0] = "ufs:vkd0s0a"; 792 rootdevnames[1] = "ufs:vkd0s1a"; 793 } 794 } 795 796 DiskNum++; 797 } else { 798 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname); 799 continue; 800 } 801 } 802 } 803 804 static 805 int 806 netif_set_tapflags(int tap_unit, int f, int s) 807 { 808 struct ifreq ifr; 809 int flags; 810 811 bzero(&ifr, sizeof(ifr)); 812 813 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit); 814 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) { 815 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit); 816 return -1; 817 } 818 819 /* 820 * Adjust if_flags 821 * 822 * If the flags are already set/cleared, then we return 823 * immediately to avoid extra syscalls 824 */ 825 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16); 826 if (f < 0) { 827 /* Turn off flags */ 828 f = -f; 829 if ((flags & f) == 0) 830 return 0; 831 flags &= ~f; 832 } else { 833 /* Turn on flags */ 834 if (flags & f) 835 return 0; 836 flags |= f; 837 } 838 839 /* 840 * Fix up ifreq.ifr_name, since it may be trashed 841 * in previous ioctl(SIOCGIFFLAGS) 842 */ 843 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit); 844 845 ifr.ifr_flags = flags & 0xffff; 846 ifr.ifr_flagshigh = flags >> 16; 847 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) { 848 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit); 849 return -1; 850 } 851 return 0; 852 } 853 854 static 855 int 856 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s) 857 { 858 struct ifaliasreq ifra; 859 struct sockaddr_in *in; 860 861 bzero(&ifra, sizeof(ifra)); 862 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit); 863 864 /* Setup address */ 865 in = (struct sockaddr_in *)&ifra.ifra_addr; 866 in->sin_family = AF_INET; 867 in->sin_len = sizeof(*in); 868 in->sin_addr.s_addr = addr; 869 870 if (mask != 0) { 871 /* Setup netmask */ 872 in = (struct sockaddr_in *)&ifra.ifra_mask; 873 in->sin_len = sizeof(*in); 874 in->sin_addr.s_addr = mask; 875 } 876 877 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) { 878 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit); 879 return -1; 880 } 881 return 0; 882 } 883 884 static 885 int 886 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s) 887 { 888 struct ifbreq ifbr; 889 struct ifdrv ifd; 890 891 bzero(&ifbr, sizeof(ifbr)); 892 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname), 893 "tap%d", tap_unit); 894 895 bzero(&ifd, sizeof(ifd)); 896 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name)); 897 ifd.ifd_cmd = BRDGADD; 898 ifd.ifd_len = sizeof(ifbr); 899 ifd.ifd_data = &ifbr; 900 901 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) { 902 /* 903 * 'errno == EEXIST' means that the tap(4) is already 904 * a member of the bridge(4) 905 */ 906 if (errno != EEXIST) { 907 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge); 908 return -1; 909 } 910 } 911 return 0; 912 } 913 914 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK) 915 916 /* 917 * Locate the first unused tap(4) device file if auto mode is requested, 918 * or open the user supplied device file, and bring up the corresponding 919 * tap(4) interface. 920 * 921 * NOTE: Only tap(4) device file is supported currently 922 */ 923 static 924 int 925 netif_open_tap(const char *netif, int *tap_unit, int s) 926 { 927 char tap_dev[MAXPATHLEN]; 928 int tap_fd, failed; 929 struct stat st; 930 char *dname; 931 932 *tap_unit = -1; 933 934 if (strcmp(netif, "auto") == 0) { 935 /* 936 * Find first unused tap(4) device file 937 */ 938 tap_fd = open("/dev/tap", TAPDEV_OFLAGS); 939 if (tap_fd < 0) { 940 warnc(errno, "Unable to find a free tap(4)"); 941 return -1; 942 } 943 } else { 944 /* 945 * User supplied tap(4) device file or unix socket. 946 */ 947 if (netif[0] == '/') /* Absolute path */ 948 strlcpy(tap_dev, netif, sizeof(tap_dev)); 949 else 950 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif); 951 952 tap_fd = open(tap_dev, TAPDEV_OFLAGS); 953 954 /* 955 * If we cannot open normally try to connect to it. 956 */ 957 if (tap_fd < 0) 958 tap_fd = unix_connect(tap_dev); 959 960 if (tap_fd < 0) { 961 warn("Unable to open %s", tap_dev); 962 return -1; 963 } 964 } 965 966 /* 967 * Check whether the device file is a tap(4) 968 */ 969 if (fstat(tap_fd, &st) < 0) { 970 failed = 1; 971 } else if (S_ISCHR(st.st_mode)) { 972 dname = fdevname(tap_fd); 973 if (dname) 974 dname = strstr(dname, "tap"); 975 if (dname) { 976 /* 977 * Bring up the corresponding tap(4) interface 978 */ 979 *tap_unit = strtol(dname + 3, NULL, 10); 980 printf("TAP UNIT %d\n", *tap_unit); 981 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0) 982 failed = 0; 983 else 984 failed = 1; 985 } else { 986 failed = 1; 987 } 988 } else if (S_ISSOCK(st.st_mode)) { 989 /* 990 * Special socket connection (typically to vknet). We 991 * do not have to do anything. 992 */ 993 failed = 0; 994 } else { 995 failed = 1; 996 } 997 998 if (failed) { 999 warnx("%s is not a tap(4) device or socket", tap_dev); 1000 close(tap_fd); 1001 tap_fd = -1; 1002 *tap_unit = -1; 1003 } 1004 return tap_fd; 1005 } 1006 1007 static int 1008 unix_connect(const char *path) 1009 { 1010 struct sockaddr_un sunx; 1011 int len; 1012 int net_fd; 1013 int sndbuf = 262144; 1014 struct stat st; 1015 1016 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path); 1017 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]); 1018 ++len; /* include nul */ 1019 sunx.sun_family = AF_UNIX; 1020 sunx.sun_len = len; 1021 1022 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0); 1023 if (net_fd < 0) 1024 return(-1); 1025 if (connect(net_fd, (void *)&sunx, len) < 0) { 1026 close(net_fd); 1027 return(-1); 1028 } 1029 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf)); 1030 if (fstat(net_fd, &st) == 0) 1031 printf("Network socket buffer: %d bytes\n", st.st_blksize); 1032 fcntl(net_fd, F_SETFL, O_NONBLOCK); 1033 return(net_fd); 1034 } 1035 1036 #undef TAPDEV_MAJOR 1037 #undef TAPDEV_MINOR 1038 #undef TAPDEV_OFLAGS 1039 1040 /* 1041 * Following syntax is supported, 1042 * 1) x.x.x.x tap(4)'s address is x.x.x.x 1043 * 1044 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x 1045 * tap(4)'s netmask len is z 1046 * 1047 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x 1048 * pseudo netif's address is y.y.y.y 1049 * 1050 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x 1051 * pseudo netif's address is y.y.y.y 1052 * tap(4) and pseudo netif's netmask len are z 1053 * 1054 * 5) bridgeX tap(4) will be added to bridgeX 1055 * 1056 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX 1057 * pseudo netif's address is y.y.y.y 1058 * 1059 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX 1060 * pseudo netif's address is y.y.y.y 1061 * pseudo netif's netmask len is z 1062 */ 1063 static 1064 int 1065 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s) 1066 { 1067 in_addr_t tap_addr, netmask, netif_addr; 1068 int next_netif_addr; 1069 char *tok, *masklen_str, *ifbridge; 1070 1071 *addr = 0; 1072 *mask = 0; 1073 1074 tok = strtok(NULL, ":/"); 1075 if (tok == NULL) { 1076 /* 1077 * Nothing special, simply use tap(4) as backend 1078 */ 1079 return 0; 1080 } 1081 1082 if (inet_pton(AF_INET, tok, &tap_addr) > 0) { 1083 /* 1084 * tap(4)'s address is supplied 1085 */ 1086 ifbridge = NULL; 1087 1088 /* 1089 * If there is next token, then it may be pseudo 1090 * netif's address or netmask len for tap(4) 1091 */ 1092 next_netif_addr = 0; 1093 } else { 1094 /* 1095 * Not tap(4)'s address, assume it as a bridge(4) 1096 * iface name 1097 */ 1098 tap_addr = 0; 1099 ifbridge = tok; 1100 1101 /* 1102 * If there is next token, then it must be pseudo 1103 * netif's address 1104 */ 1105 next_netif_addr = 1; 1106 } 1107 1108 netmask = netif_addr = 0; 1109 1110 tok = strtok(NULL, ":/"); 1111 if (tok == NULL) 1112 goto back; 1113 1114 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) { 1115 if (next_netif_addr) { 1116 warnx("Invalid pseudo netif address: %s", tok); 1117 return -1; 1118 } 1119 netif_addr = 0; 1120 1121 /* 1122 * Current token is not address, then it must be netmask len 1123 */ 1124 masklen_str = tok; 1125 } else { 1126 /* 1127 * Current token is pseudo netif address, if there is next token 1128 * it must be netmask len 1129 */ 1130 masklen_str = strtok(NULL, "/"); 1131 } 1132 1133 /* Calculate netmask */ 1134 if (masklen_str != NULL) { 1135 u_long masklen; 1136 1137 masklen = strtoul(masklen_str, NULL, 10); 1138 if (masklen < 32 && masklen > 0) { 1139 netmask = htonl(~((1LL << (32 - masklen)) - 1) 1140 & 0xffffffff); 1141 } else { 1142 warnx("Invalid netmask len: %lu", masklen); 1143 return -1; 1144 } 1145 } 1146 1147 /* Make sure there is no more token left */ 1148 if (strtok(NULL, ":/") != NULL) { 1149 warnx("Invalid argument to '-I'"); 1150 return -1; 1151 } 1152 1153 back: 1154 if (tap_unit < 0) { 1155 /* Do nothing */ 1156 } else if (ifbridge == NULL) { 1157 /* Set tap(4) address/netmask */ 1158 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0) 1159 return -1; 1160 } else { 1161 /* Tie tap(4) to bridge(4) */ 1162 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0) 1163 return -1; 1164 } 1165 1166 *addr = netif_addr; 1167 *mask = netmask; 1168 return 0; 1169 } 1170 1171 /* 1172 * NetifInfo[] will be filled for pseudo netif initialization. 1173 * NetifNum will be bumped to reflect the number of valid entries 1174 * in NetifInfo[]. 1175 */ 1176 static 1177 void 1178 init_netif(char *netifExp[], int netifExpNum) 1179 { 1180 int i, s; 1181 1182 if (netifExpNum == 0) 1183 return; 1184 1185 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */ 1186 if (s < 0) 1187 return; 1188 1189 for (i = 0; i < netifExpNum; ++i) { 1190 struct vknetif_info *info; 1191 in_addr_t netif_addr, netif_mask; 1192 int tap_fd, tap_unit; 1193 char *netif; 1194 1195 netif = strtok(netifExp[i], ":"); 1196 if (netif == NULL) { 1197 warnx("Invalid argument to '-I'"); 1198 continue; 1199 } 1200 1201 /* 1202 * Open tap(4) device file and bring up the 1203 * corresponding interface 1204 */ 1205 tap_fd = netif_open_tap(netif, &tap_unit, s); 1206 if (tap_fd < 0) 1207 continue; 1208 1209 /* 1210 * Initialize tap(4) and get address/netmask 1211 * for pseudo netif 1212 * 1213 * NB: Rest part of netifExp[i] is passed 1214 * to netif_init_tap() implicitly. 1215 */ 1216 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) { 1217 /* 1218 * NB: Closing tap(4) device file will bring 1219 * down the corresponding interface 1220 */ 1221 close(tap_fd); 1222 continue; 1223 } 1224 1225 info = &NetifInfo[NetifNum]; 1226 info->tap_fd = tap_fd; 1227 info->tap_unit = tap_unit; 1228 info->netif_addr = netif_addr; 1229 info->netif_mask = netif_mask; 1230 1231 NetifNum++; 1232 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */ 1233 break; 1234 } 1235 close(s); 1236 } 1237 1238 /* 1239 * Create the pid file and leave it open and locked while the vkernel is 1240 * running. This allows a script to use /usr/bin/lockf to probe whether 1241 * a vkernel is still running (so as not to accidently kill an unrelated 1242 * process from a stale pid file). 1243 */ 1244 static 1245 void 1246 writepid(void) 1247 { 1248 char buf[32]; 1249 int fd; 1250 1251 if (pid_file != NULL) { 1252 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid()); 1253 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666); 1254 if (fd < 0) { 1255 if (errno == EWOULDBLOCK) { 1256 perror("Failed to lock pidfile, " 1257 "vkernel already running"); 1258 } else { 1259 perror("Failed to create pidfile"); 1260 } 1261 exit(EX_SOFTWARE); 1262 } 1263 ftruncate(fd, 0); 1264 write(fd, buf, strlen(buf)); 1265 /* leave the file open to maintain the lock */ 1266 } 1267 } 1268 1269 static 1270 void 1271 cleanpid( void ) 1272 { 1273 if (pid_file != NULL) { 1274 if (unlink(pid_file) < 0) 1275 perror("Warning: couldn't remove pidfile"); 1276 } 1277 } 1278 1279 static 1280 void 1281 usage_err(const char *ctl, ...) 1282 { 1283 va_list va; 1284 1285 va_start(va, ctl); 1286 vfprintf(stderr, ctl, va); 1287 va_end(va); 1288 fprintf(stderr, "\n"); 1289 exit(EX_USAGE); 1290 } 1291 1292 static 1293 void 1294 usage_help(_Bool help) 1295 { 1296 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n" 1297 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n" 1298 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n" 1299 "\t[-p file] [-r file]\n", save_av[0]); 1300 1301 if (help) 1302 fprintf(stderr, "\nArguments:\n" 1303 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n" 1304 "\t-e\tSpecify an environment to be used by the kernel.\n" 1305 "\t-h\tThis list of options.\n" 1306 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n" 1307 "\t-I\tCreate a virtual network device.\n" 1308 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n" 1309 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n" 1310 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n" 1311 "\t \t- numcpus - number of cpus\n" 1312 "\t \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n" 1313 "\t \t the logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n" 1314 "\t \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n" 1315 "\t \t the core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n" 1316 "\t-p\tSpecify a file in which to store the process ID.\n" 1317 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n" 1318 "\t-s\tBoot into single-user mode.\n" 1319 "\t-U\tEnable writing to kernel memory and module loading.\n" 1320 "\t-v\tTurn on verbose booting.\n"); 1321 1322 exit(EX_USAGE); 1323 } 1324 1325 void 1326 cpu_reset(void) 1327 { 1328 kprintf("cpu reset, rebooting vkernel\n"); 1329 closefrom(3); 1330 cleanpid(); 1331 execv(save_av[0], save_av); 1332 } 1333 1334 void 1335 cpu_halt(void) 1336 { 1337 kprintf("cpu halt, exiting vkernel\n"); 1338 cleanpid(); 1339 exit(EX_OK); 1340 } 1341 1342 void 1343 setrealcpu(void) 1344 { 1345 switch(lwp_cpu_lock) { 1346 case LCL_PER_CPU: 1347 if (bootverbose) 1348 kprintf("Locking CPU%d to real cpu %d\n", 1349 mycpuid, next_cpu); 1350 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu)); 1351 next_cpu++; 1352 if (next_cpu >= real_ncpus) 1353 next_cpu = 0; 1354 break; 1355 case LCL_SINGLE_CPU: 1356 if (bootverbose) 1357 kprintf("Locking CPU%d to real cpu %d\n", 1358 mycpuid, next_cpu); 1359 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu)); 1360 break; 1361 default: 1362 /* do not map virtual cpus to real cpus */ 1363 break; 1364 } 1365 } 1366 1367 /* 1368 * Allocate and free memory for module loading. The loaded module 1369 * has to be placed somewhere near the current kernel binary load 1370 * point or the relocations will not work. 1371 * 1372 * I'm not sure why this isn't working. 1373 */ 1374 int 1375 vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes) 1376 { 1377 kprintf("module loading for vkernel64's not currently supported\n"); 1378 *basep = 0; 1379 return ENOMEM; 1380 #if 0 1381 #if 1 1382 size_t xtra; 1383 xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK; 1384 *basep = (vm_offset_t)sbrk(xtra + bytes) + xtra; 1385 bzero((void *)*basep, bytes); 1386 #else 1387 *basep = (vm_offset_t)mmap((void *)0x000000000, bytes, 1388 PROT_READ|PROT_WRITE|PROT_EXEC, 1389 MAP_ANON|MAP_SHARED, -1, 0); 1390 if ((void *)*basep == MAP_FAILED) 1391 return ENOMEM; 1392 #endif 1393 kprintf("basep %p %p %zd\n", 1394 (void *)vkernel_module_memory_alloc, (void *)*basep, bytes); 1395 return 0; 1396 #endif 1397 } 1398 1399 void 1400 vkernel_module_memory_free(vm_offset_t base, size_t bytes) 1401 { 1402 #if 0 1403 #if 0 1404 munmap((void *)base, bytes); 1405 #endif 1406 #endif 1407 } 1408