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