1 /* $OpenBSD: config.c,v 1.71 2023/04/28 19:46:42 dv Exp $ */ 2 3 /* 4 * Copyright (c) 2015 Reyk Floeter <reyk@openbsd.org> 5 * 6 * Permission to use, copy, modify, and distribute this software for any 7 * purpose with or without fee is hereby granted, provided that the above 8 * copyright notice and this permission notice appear in all copies. 9 * 10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES 11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR 13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 17 */ 18 19 #include <sys/types.h> 20 #include <sys/queue.h> 21 #include <sys/time.h> 22 #include <sys/uio.h> 23 #include <sys/stat.h> 24 #include <sys/socket.h> 25 26 #include <net/if.h> 27 28 #include <stdio.h> 29 #include <stdlib.h> 30 #include <termios.h> 31 #include <unistd.h> 32 #include <limits.h> 33 #include <string.h> 34 #include <fcntl.h> 35 #include <util.h> 36 #include <errno.h> 37 #include <imsg.h> 38 39 #include "proc.h" 40 #include "vmd.h" 41 42 /* Supported bridge types */ 43 const char *vmd_descsw[] = { "bridge", "veb", NULL }; 44 45 static int config_init_localprefix(struct vmd_config *); 46 47 static int 48 config_init_localprefix(struct vmd_config *cfg) 49 { 50 struct sockaddr_in6 *sin6; 51 52 if (host(VMD_DHCP_PREFIX, &cfg->cfg_localprefix) == -1) 53 return (-1); 54 55 /* IPv6 is disabled by default */ 56 cfg->cfg_flags &= ~VMD_CFG_INET6; 57 58 /* Generate random IPv6 prefix only once */ 59 if (cfg->cfg_flags & VMD_CFG_AUTOINET6) 60 return (0); 61 if (host(VMD_ULA_PREFIX, &cfg->cfg_localprefix6) == -1) 62 return (-1); 63 /* Randomize the 56 bits "Global ID" and "Subnet ID" */ 64 sin6 = ss2sin6(&cfg->cfg_localprefix6.ss); 65 arc4random_buf(&sin6->sin6_addr.s6_addr[1], 7); 66 cfg->cfg_flags |= VMD_CFG_AUTOINET6; 67 68 return (0); 69 } 70 71 int 72 config_init(struct vmd *env) 73 { 74 struct privsep *ps = &env->vmd_ps; 75 unsigned int what; 76 77 /* Global configuration */ 78 ps->ps_what[PROC_PARENT] = CONFIG_ALL; 79 ps->ps_what[PROC_VMM] = CONFIG_VMS; 80 81 /* Local prefix */ 82 if (config_init_localprefix(&env->vmd_cfg) == -1) 83 return (-1); 84 85 /* Other configuration */ 86 what = ps->ps_what[privsep_process]; 87 if (what & CONFIG_VMS) { 88 if ((env->vmd_vms = calloc(1, sizeof(*env->vmd_vms))) == NULL) 89 return (-1); 90 if ((env->vmd_known = calloc(1, sizeof(*env->vmd_known))) == NULL) 91 return (-1); 92 TAILQ_INIT(env->vmd_vms); 93 TAILQ_INIT(env->vmd_known); 94 } 95 if (what & CONFIG_SWITCHES) { 96 if ((env->vmd_switches = calloc(1, 97 sizeof(*env->vmd_switches))) == NULL) 98 return (-1); 99 TAILQ_INIT(env->vmd_switches); 100 } 101 102 return (0); 103 } 104 105 void 106 config_purge(struct vmd *env, unsigned int reset) 107 { 108 struct privsep *ps = &env->vmd_ps; 109 struct name2id *n2i; 110 struct vmd_vm *vm; 111 struct vmd_switch *vsw; 112 unsigned int what; 113 114 DPRINTF("%s: %s purging vms and switches", 115 __func__, ps->ps_title[privsep_process]); 116 117 /* Reset global configuration (prefix was verified before) */ 118 config_init_localprefix(&env->vmd_cfg); 119 120 /* Reset other configuration */ 121 what = ps->ps_what[privsep_process] & reset; 122 if (what & CONFIG_VMS && env->vmd_vms != NULL) { 123 while ((vm = TAILQ_FIRST(env->vmd_vms)) != NULL) { 124 vm_remove(vm, __func__); 125 } 126 while ((n2i = TAILQ_FIRST(env->vmd_known)) != NULL) { 127 TAILQ_REMOVE(env->vmd_known, n2i, entry); 128 free(n2i); 129 } 130 env->vmd_nvm = 0; 131 } 132 if (what & CONFIG_SWITCHES && env->vmd_switches != NULL) { 133 while ((vsw = TAILQ_FIRST(env->vmd_switches)) != NULL) 134 switch_remove(vsw); 135 env->vmd_nswitches = 0; 136 } 137 } 138 139 int 140 config_setconfig(struct vmd *env) 141 { 142 struct privsep *ps = &env->vmd_ps; 143 unsigned int id; 144 145 DPRINTF("%s: setting config", __func__); 146 147 for (id = 0; id < PROC_MAX; id++) { 148 if (id == privsep_process) 149 continue; 150 proc_compose(ps, id, IMSG_VMDOP_CONFIG, &env->vmd_cfg, 151 sizeof(env->vmd_cfg)); 152 } 153 154 return (0); 155 } 156 157 int 158 config_getconfig(struct vmd *env, struct imsg *imsg) 159 { 160 struct privsep *ps = &env->vmd_ps; 161 162 log_debug("%s: %s retrieving config", 163 __func__, ps->ps_title[privsep_process]); 164 165 IMSG_SIZE_CHECK(imsg, &env->vmd_cfg); 166 memcpy(&env->vmd_cfg, imsg->data, sizeof(env->vmd_cfg)); 167 168 return (0); 169 } 170 171 int 172 config_setreset(struct vmd *env, unsigned int reset) 173 { 174 struct privsep *ps = &env->vmd_ps; 175 unsigned int id; 176 177 DPRINTF("%s: resetting state", __func__); 178 179 for (id = 0; id < PROC_MAX; id++) { 180 if ((reset & ps->ps_what[id]) == 0 || 181 id == privsep_process) 182 continue; 183 proc_compose(ps, id, IMSG_CTL_RESET, &reset, sizeof(reset)); 184 } 185 186 return (0); 187 } 188 189 int 190 config_getreset(struct vmd *env, struct imsg *imsg) 191 { 192 unsigned int mode; 193 194 IMSG_SIZE_CHECK(imsg, &mode); 195 memcpy(&mode, imsg->data, sizeof(mode)); 196 197 log_debug("%s: %s resetting state", 198 __func__, env->vmd_ps.ps_title[privsep_process]); 199 200 config_purge(env, mode); 201 202 return (0); 203 } 204 205 /* 206 * config_setvm 207 * 208 * Configure a vm, opening any required file descriptors. 209 * 210 * Returns 0 on success, error code on failure. 211 */ 212 int 213 config_setvm(struct privsep *ps, struct vmd_vm *vm, uint32_t peerid, uid_t uid) 214 { 215 int diskfds[VM_MAX_DISKS_PER_VM][VM_MAX_BASE_PER_DISK]; 216 struct vmd_if *vif; 217 struct vmop_create_params *vmc = &vm->vm_params; 218 struct vm_create_params *vcp = &vmc->vmc_params; 219 unsigned int i, j; 220 int fd = -1, cdromfd = -1, kernfd = -1; 221 int *tapfds = NULL; 222 int n = 0, aflags, oflags, ret = -1; 223 char ifname[IF_NAMESIZE], *s; 224 char path[PATH_MAX], base[PATH_MAX]; 225 unsigned int unit; 226 struct timeval tv, rate, since_last; 227 struct vmop_addr_req var; 228 size_t bytes = 0; 229 230 if (vm->vm_state & VM_STATE_RUNNING) { 231 log_warnx("%s: vm is already running", __func__); 232 return (EALREADY); 233 } 234 235 /* 236 * Rate-limit the VM so that it cannot restart in a loop: 237 * if the VM restarts after less than VM_START_RATE_SEC seconds, 238 * we increment the limit counter. After VM_START_RATE_LIMIT 239 * of suchs fast reboots the VM is stopped. 240 */ 241 getmonotime(&tv); 242 if (vm->vm_start_tv.tv_sec) { 243 timersub(&tv, &vm->vm_start_tv, &since_last); 244 245 rate.tv_sec = VM_START_RATE_SEC; 246 rate.tv_usec = 0; 247 if (timercmp(&since_last, &rate, <)) 248 vm->vm_start_limit++; 249 else { 250 /* Reset counter */ 251 vm->vm_start_limit = 0; 252 } 253 254 log_debug("%s: vm %u restarted after %lld.%ld seconds," 255 " limit %d/%d", __func__, vcp->vcp_id, since_last.tv_sec, 256 since_last.tv_usec, vm->vm_start_limit, 257 VM_START_RATE_LIMIT); 258 259 if (vm->vm_start_limit >= VM_START_RATE_LIMIT) { 260 log_warnx("%s: vm %u restarted too quickly", 261 __func__, vcp->vcp_id); 262 return (EPERM); 263 } 264 } 265 vm->vm_start_tv = tv; 266 267 for (i = 0; i < VM_MAX_DISKS_PER_VM; i++) 268 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) 269 diskfds[i][j] = -1; 270 271 tapfds = reallocarray(NULL, vmc->vmc_nnics, sizeof(*tapfds)); 272 if (tapfds == NULL) { 273 ret = errno; 274 log_warn("%s: can't allocate tap fds", __func__); 275 return (ret); 276 } 277 for (i = 0; i < vmc->vmc_nnics; i++) 278 tapfds[i] = -1; 279 280 vm->vm_peerid = peerid; 281 vm->vm_uid = uid; 282 283 /* 284 * From here onward, all failures need cleanup and use goto fail 285 */ 286 if (!(vm->vm_state & VM_STATE_RECEIVED) && vm->vm_kernel == -1) { 287 if (vm->vm_kernel_path != NULL) { 288 /* Open external kernel for child */ 289 kernfd = open(vm->vm_kernel_path, O_RDONLY); 290 if (kernfd == -1) { 291 ret = errno; 292 log_warn("%s: can't open kernel or BIOS " 293 "boot image %s", __func__, 294 vm->vm_kernel_path); 295 goto fail; 296 } 297 } 298 299 /* 300 * Try to open the default BIOS image if no kernel/BIOS has been 301 * specified. The BIOS is an external firmware file that is 302 * typically distributed separately due to an incompatible 303 * license. 304 */ 305 if (kernfd == -1) { 306 if ((kernfd = open(VM_DEFAULT_BIOS, O_RDONLY)) == -1) { 307 log_warn("can't open %s", VM_DEFAULT_BIOS); 308 ret = VMD_BIOS_MISSING; 309 goto fail; 310 } 311 } 312 313 if (vm_checkaccess(kernfd, 314 vmc->vmc_checkaccess & VMOP_CREATE_KERNEL, 315 uid, R_OK) == -1) { 316 log_warnx("vm \"%s\" no read access to kernel " 317 "%s", vcp->vcp_name, vm->vm_kernel_path); 318 ret = EPERM; 319 goto fail; 320 } 321 322 vm->vm_kernel = kernfd; 323 vmc->vmc_kernel = kernfd; 324 } 325 326 /* Open CDROM image for child */ 327 if (strlen(vmc->vmc_cdrom)) { 328 /* Stat cdrom to ensure it is a regular file */ 329 if ((cdromfd = 330 open(vmc->vmc_cdrom, O_RDONLY)) == -1) { 331 log_warn("can't open cdrom %s", vmc->vmc_cdrom); 332 ret = VMD_CDROM_MISSING; 333 goto fail; 334 } 335 336 if (vm_checkaccess(cdromfd, 337 vmc->vmc_checkaccess & VMOP_CREATE_CDROM, 338 uid, R_OK) == -1) { 339 log_warnx("vm \"%s\" no read access to cdrom %s", 340 vcp->vcp_name, vmc->vmc_cdrom); 341 ret = EPERM; 342 goto fail; 343 } 344 } 345 346 /* Open disk images for child */ 347 for (i = 0 ; i < vmc->vmc_ndisks; i++) { 348 if (strlcpy(path, vmc->vmc_disks[i], sizeof(path)) 349 >= sizeof(path)) 350 log_warnx("disk path %s too long", vmc->vmc_disks[i]); 351 memset(vmc->vmc_diskbases, 0, sizeof(vmc->vmc_diskbases)); 352 oflags = O_RDWR|O_EXLOCK|O_NONBLOCK; 353 aflags = R_OK|W_OK; 354 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) { 355 /* Stat disk[i] to ensure it is a regular file */ 356 if ((diskfds[i][j] = open(path, oflags)) == -1) { 357 log_warn("can't open disk %s", 358 vmc->vmc_disks[i]); 359 ret = VMD_DISK_MISSING; 360 goto fail; 361 } 362 363 if (vm_checkaccess(diskfds[i][j], 364 vmc->vmc_checkaccess & VMOP_CREATE_DISK, 365 uid, aflags) == -1) { 366 log_warnx("vm \"%s\" unable to access " 367 "disk %s", vcp->vcp_name, path); 368 errno = EPERM; 369 goto fail; 370 } 371 372 /* 373 * Clear the write and exclusive flags for base images. 374 * All writes should go to the top image, allowing them 375 * to be shared. 376 */ 377 oflags = O_RDONLY|O_NONBLOCK; 378 aflags = R_OK; 379 n = virtio_get_base(diskfds[i][j], base, sizeof(base), 380 vmc->vmc_disktypes[i], path); 381 if (n == 0) 382 break; 383 if (n == -1) { 384 log_warnx("vm \"%s\" unable to read " 385 "base for disk %s", vcp->vcp_name, 386 vmc->vmc_disks[i]); 387 goto fail; 388 } 389 (void)strlcpy(path, base, sizeof(path)); 390 } 391 } 392 393 /* Open network interfaces */ 394 for (i = 0 ; i < vmc->vmc_nnics; i++) { 395 vif = &vm->vm_ifs[i]; 396 397 /* Check if the user has requested a specific tap(4) */ 398 s = vmc->vmc_ifnames[i]; 399 if (*s != '\0' && strcmp("tap", s) != 0) { 400 if (priv_getiftype(s, ifname, &unit) == -1 || 401 strcmp(ifname, "tap") != 0) { 402 log_warnx("%s: invalid tap name %s", 403 __func__, s); 404 ret = EINVAL; 405 goto fail; 406 } 407 } else 408 s = NULL; 409 410 /* 411 * Either open the requested tap(4) device or get 412 * the next available one. 413 */ 414 if (s != NULL) { 415 snprintf(path, PATH_MAX, "/dev/%s", s); 416 tapfds[i] = open(path, O_RDWR | O_NONBLOCK); 417 } else { 418 tapfds[i] = opentap(ifname); 419 s = ifname; 420 } 421 if (tapfds[i] == -1) { 422 log_warnx("%s: can't open tap %s", __func__, s); 423 goto fail; 424 } 425 if ((vif->vif_name = strdup(s)) == NULL) { 426 log_warn("%s: can't save tap %s", __func__, s); 427 goto fail; 428 } 429 430 /* Check if the the interface is attached to a switch */ 431 s = vmc->vmc_ifswitch[i]; 432 if (*s != '\0') { 433 if ((vif->vif_switch = strdup(s)) == NULL) { 434 log_warn("%s: can't save switch %s", 435 __func__, s); 436 goto fail; 437 } 438 } 439 440 /* Check if the the interface is assigned to a group */ 441 s = vmc->vmc_ifgroup[i]; 442 if (*s != '\0') { 443 if ((vif->vif_group = strdup(s)) == NULL) { 444 log_warn("%s: can't save group %s", 445 __func__, s); 446 goto fail; 447 } 448 } 449 450 /* non-default rdomain (requires VMIFF_RDOMAIN below) */ 451 vif->vif_rdomain = vmc->vmc_ifrdomain[i]; 452 453 /* Set the interface status */ 454 vif->vif_flags = 455 vmc->vmc_ifflags[i] & (VMIFF_UP|VMIFF_OPTMASK); 456 } 457 458 /* Open TTY */ 459 if (vm->vm_ttyname[0] == '\0') { 460 if (vm_opentty(vm) == -1) { 461 log_warn("%s: can't open tty %s", __func__, 462 vm->vm_ttyname[0] == '\0' ? "" : vm->vm_ttyname); 463 goto fail; 464 } 465 } 466 if ((fd = dup(vm->vm_tty)) == -1) { 467 log_warn("%s: can't re-open tty %s", __func__, vm->vm_ttyname); 468 goto fail; 469 } 470 471 /* Send VM information */ 472 /* XXX check proc_compose_imsg return values */ 473 if (vm->vm_state & VM_STATE_RECEIVED) 474 proc_compose_imsg(ps, PROC_VMM, -1, 475 IMSG_VMDOP_RECEIVE_VM_REQUEST, vm->vm_vmid, fd, vmc, 476 sizeof(struct vmop_create_params)); 477 else 478 proc_compose_imsg(ps, PROC_VMM, -1, 479 IMSG_VMDOP_START_VM_REQUEST, vm->vm_vmid, vm->vm_kernel, 480 vmc, sizeof(*vmc)); 481 482 if (strlen(vmc->vmc_cdrom)) 483 proc_compose_imsg(ps, PROC_VMM, -1, 484 IMSG_VMDOP_START_VM_CDROM, vm->vm_vmid, cdromfd, 485 NULL, 0); 486 487 for (i = 0; i < vmc->vmc_ndisks; i++) { 488 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) { 489 if (diskfds[i][j] == -1) 490 break; 491 proc_compose_imsg(ps, PROC_VMM, -1, 492 IMSG_VMDOP_START_VM_DISK, vm->vm_vmid, 493 diskfds[i][j], &i, sizeof(i)); 494 } 495 } 496 for (i = 0; i < vmc->vmc_nnics; i++) { 497 proc_compose_imsg(ps, PROC_VMM, -1, 498 IMSG_VMDOP_START_VM_IF, vm->vm_vmid, tapfds[i], 499 &i, sizeof(i)); 500 501 memset(&var, 0, sizeof(var)); 502 var.var_vmid = vm->vm_vmid; 503 var.var_nic_idx = i; 504 proc_compose_imsg(ps, PROC_PRIV, -1, IMSG_VMDOP_PRIV_GET_ADDR, 505 vm->vm_vmid, dup(tapfds[i]), &var, sizeof(var)); 506 } 507 508 if (!(vm->vm_state & VM_STATE_RECEIVED)) 509 proc_compose_imsg(ps, PROC_VMM, -1, 510 IMSG_VMDOP_START_VM_END, vm->vm_vmid, fd, NULL, 0); 511 512 free(tapfds); 513 514 /* Collapse any memranges after the vm was sent to PROC_VMM */ 515 if (vcp->vcp_nmemranges > 0) { 516 for (i = 0; i < vcp->vcp_nmemranges; i++) 517 bytes += vcp->vcp_memranges[i].vmr_size; 518 memset(&vcp->vcp_memranges, 0, sizeof(vcp->vcp_memranges)); 519 vcp->vcp_nmemranges = 0; 520 vcp->vcp_memranges[0].vmr_size = bytes; 521 } 522 vm->vm_state |= VM_STATE_RUNNING; 523 return (0); 524 525 fail: 526 log_warnx("failed to start vm %s", vcp->vcp_name); 527 528 if (vm->vm_kernel != -1) 529 close(kernfd); 530 if (cdromfd != -1) 531 close(cdromfd); 532 for (i = 0; i < vmc->vmc_ndisks; i++) 533 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) 534 if (diskfds[i][j] != -1) 535 close(diskfds[i][j]); 536 if (tapfds != NULL) { 537 for (i = 0; i < vmc->vmc_nnics; i++) 538 close(tapfds[i]); 539 free(tapfds); 540 } 541 542 if (vm->vm_from_config) { 543 vm_stop(vm, 0, __func__); 544 } else { 545 vm_remove(vm, __func__); 546 } 547 548 return (ret); 549 } 550 551 int 552 config_getvm(struct privsep *ps, struct imsg *imsg) 553 { 554 struct vmop_create_params vmc; 555 struct vmd_vm *vm = NULL; 556 557 IMSG_SIZE_CHECK(imsg, &vmc); 558 memcpy(&vmc, imsg->data, sizeof(vmc)); 559 vmc.vmc_kernel = imsg->fd; 560 561 errno = 0; 562 if (vm_register(ps, &vmc, &vm, imsg->hdr.peerid, 0) == -1) 563 goto fail; 564 565 vm->vm_state |= VM_STATE_RUNNING; 566 vm->vm_peerid = (uint32_t)-1; 567 vm->vm_kernel = imsg->fd; 568 return (0); 569 570 fail: 571 if (imsg->fd != -1) { 572 close(imsg->fd); 573 imsg->fd = -1; 574 } 575 576 vm_remove(vm, __func__); 577 if (errno == 0) 578 errno = EINVAL; 579 580 return (-1); 581 } 582 583 int 584 config_getdisk(struct privsep *ps, struct imsg *imsg) 585 { 586 struct vmd_vm *vm; 587 unsigned int n, idx; 588 589 errno = 0; 590 if ((vm = vm_getbyvmid(imsg->hdr.peerid)) == NULL) { 591 errno = ENOENT; 592 return (-1); 593 } 594 595 IMSG_SIZE_CHECK(imsg, &n); 596 memcpy(&n, imsg->data, sizeof(n)); 597 598 if (n >= vm->vm_params.vmc_ndisks || imsg->fd == -1) { 599 log_warnx("invalid disk id"); 600 errno = EINVAL; 601 return (-1); 602 } 603 idx = vm->vm_params.vmc_diskbases[n]++; 604 if (idx >= VM_MAX_BASE_PER_DISK) { 605 log_warnx("too many bases for disk"); 606 errno = EINVAL; 607 return (-1); 608 } 609 vm->vm_disks[n][idx] = imsg->fd; 610 return (0); 611 } 612 613 int 614 config_getif(struct privsep *ps, struct imsg *imsg) 615 { 616 struct vmd_vm *vm; 617 unsigned int n; 618 619 errno = 0; 620 if ((vm = vm_getbyvmid(imsg->hdr.peerid)) == NULL) { 621 errno = ENOENT; 622 return (-1); 623 } 624 625 IMSG_SIZE_CHECK(imsg, &n); 626 memcpy(&n, imsg->data, sizeof(n)); 627 if (n >= vm->vm_params.vmc_nnics || 628 vm->vm_ifs[n].vif_fd != -1 || imsg->fd == -1) { 629 log_warnx("invalid interface id"); 630 goto fail; 631 } 632 vm->vm_ifs[n].vif_fd = imsg->fd; 633 return (0); 634 fail: 635 if (imsg->fd != -1) 636 close(imsg->fd); 637 errno = EINVAL; 638 return (-1); 639 } 640 641 int 642 config_getcdrom(struct privsep *ps, struct imsg *imsg) 643 { 644 struct vmd_vm *vm; 645 646 errno = 0; 647 if ((vm = vm_getbyvmid(imsg->hdr.peerid)) == NULL) { 648 errno = ENOENT; 649 return (-1); 650 } 651 652 if (imsg->fd == -1) { 653 log_warnx("invalid cdrom id"); 654 goto fail; 655 } 656 657 vm->vm_cdrom = imsg->fd; 658 return (0); 659 fail: 660 if (imsg->fd != -1) 661 close(imsg->fd); 662 errno = EINVAL; 663 return (-1); 664 } 665