1 /* $OpenBSD: config.c,v 1.64 2021/11/10 20:49:04 sthen 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 if (what & CONFIG_USERS) { 102 if ((env->vmd_users = calloc(1, 103 sizeof(*env->vmd_users))) == NULL) 104 return (-1); 105 TAILQ_INIT(env->vmd_users); 106 } 107 108 return (0); 109 } 110 111 void 112 config_purge(struct vmd *env, unsigned int reset) 113 { 114 struct privsep *ps = &env->vmd_ps; 115 struct name2id *n2i; 116 struct vmd_vm *vm; 117 struct vmd_switch *vsw; 118 unsigned int what; 119 120 DPRINTF("%s: %s purging vms and switches", 121 __func__, ps->ps_title[privsep_process]); 122 123 /* Reset global configuration (prefix was verified before) */ 124 config_init_localprefix(&env->vmd_cfg); 125 126 /* Reset other configuration */ 127 what = ps->ps_what[privsep_process] & reset; 128 if (what & CONFIG_VMS && env->vmd_vms != NULL) { 129 while ((vm = TAILQ_FIRST(env->vmd_vms)) != NULL) { 130 vm_remove(vm, __func__); 131 } 132 while ((n2i = TAILQ_FIRST(env->vmd_known)) != NULL) { 133 TAILQ_REMOVE(env->vmd_known, n2i, entry); 134 free(n2i); 135 } 136 env->vmd_nvm = 0; 137 } 138 if (what & CONFIG_SWITCHES && env->vmd_switches != NULL) { 139 while ((vsw = TAILQ_FIRST(env->vmd_switches)) != NULL) 140 switch_remove(vsw); 141 env->vmd_nswitches = 0; 142 } 143 } 144 145 int 146 config_setconfig(struct vmd *env) 147 { 148 struct privsep *ps = &env->vmd_ps; 149 unsigned int id; 150 151 DPRINTF("%s: setting config", __func__); 152 153 for (id = 0; id < PROC_MAX; id++) { 154 if (id == privsep_process) 155 continue; 156 proc_compose(ps, id, IMSG_VMDOP_CONFIG, &env->vmd_cfg, 157 sizeof(env->vmd_cfg)); 158 } 159 160 return (0); 161 } 162 163 int 164 config_getconfig(struct vmd *env, struct imsg *imsg) 165 { 166 struct privsep *ps = &env->vmd_ps; 167 168 log_debug("%s: %s retrieving config", 169 __func__, ps->ps_title[privsep_process]); 170 171 IMSG_SIZE_CHECK(imsg, &env->vmd_cfg); 172 memcpy(&env->vmd_cfg, imsg->data, sizeof(env->vmd_cfg)); 173 174 return (0); 175 } 176 177 int 178 config_setreset(struct vmd *env, unsigned int reset) 179 { 180 struct privsep *ps = &env->vmd_ps; 181 unsigned int id; 182 183 DPRINTF("%s: resetting state", __func__); 184 185 for (id = 0; id < PROC_MAX; id++) { 186 if ((reset & ps->ps_what[id]) == 0 || 187 id == privsep_process) 188 continue; 189 proc_compose(ps, id, IMSG_CTL_RESET, &reset, sizeof(reset)); 190 } 191 192 return (0); 193 } 194 195 int 196 config_getreset(struct vmd *env, struct imsg *imsg) 197 { 198 unsigned int mode; 199 200 IMSG_SIZE_CHECK(imsg, &mode); 201 memcpy(&mode, imsg->data, sizeof(mode)); 202 203 log_debug("%s: %s resetting state", 204 __func__, env->vmd_ps.ps_title[privsep_process]); 205 206 config_purge(env, mode); 207 208 return (0); 209 } 210 211 /* 212 * config_setvm 213 * 214 * Configure a vm, opening any required file descriptors. 215 * 216 * Returns 0 on success, error code on failure. 217 */ 218 int 219 config_setvm(struct privsep *ps, struct vmd_vm *vm, uint32_t peerid, uid_t uid) 220 { 221 int diskfds[VMM_MAX_DISKS_PER_VM][VM_MAX_BASE_PER_DISK]; 222 struct vmd_if *vif; 223 struct vmop_create_params *vmc = &vm->vm_params; 224 struct vm_create_params *vcp = &vmc->vmc_params; 225 unsigned int i, j; 226 int fd = -1, cdromfd = -1, kernfd = -1; 227 int *tapfds = NULL; 228 int n = 0, aflags, oflags, ret = -1; 229 char ifname[IF_NAMESIZE], *s; 230 char path[PATH_MAX], base[PATH_MAX]; 231 unsigned int unit; 232 struct timeval tv, rate, since_last; 233 struct vmop_addr_req var; 234 235 if (vm->vm_state & VM_STATE_RUNNING) { 236 log_warnx("%s: vm is already running", __func__); 237 return (EALREADY); 238 } 239 240 /* increase the user reference counter and check user limits */ 241 if (vm->vm_user != NULL && user_get(vm->vm_user->usr_id.uid) != NULL) { 242 user_inc(vcp, vm->vm_user, 1); 243 if (user_checklimit(vm->vm_user, vcp) == -1) 244 return (EPERM); 245 } 246 247 /* 248 * Rate-limit the VM so that it cannot restart in a loop: 249 * if the VM restarts after less than VM_START_RATE_SEC seconds, 250 * we increment the limit counter. After VM_START_RATE_LIMIT 251 * of suchs fast reboots the VM is stopped. 252 */ 253 getmonotime(&tv); 254 if (vm->vm_start_tv.tv_sec) { 255 timersub(&tv, &vm->vm_start_tv, &since_last); 256 257 rate.tv_sec = VM_START_RATE_SEC; 258 rate.tv_usec = 0; 259 if (timercmp(&since_last, &rate, <)) 260 vm->vm_start_limit++; 261 else { 262 /* Reset counter */ 263 vm->vm_start_limit = 0; 264 } 265 266 log_debug("%s: vm %u restarted after %lld.%ld seconds," 267 " limit %d/%d", __func__, vcp->vcp_id, since_last.tv_sec, 268 since_last.tv_usec, vm->vm_start_limit, 269 VM_START_RATE_LIMIT); 270 271 if (vm->vm_start_limit >= VM_START_RATE_LIMIT) { 272 log_warnx("%s: vm %u restarted too quickly", 273 __func__, vcp->vcp_id); 274 return (EPERM); 275 } 276 } 277 vm->vm_start_tv = tv; 278 279 for (i = 0; i < VMM_MAX_DISKS_PER_VM; i++) 280 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) 281 diskfds[i][j] = -1; 282 283 tapfds = reallocarray(NULL, vcp->vcp_nnics, sizeof(*tapfds)); 284 if (tapfds == NULL) { 285 ret = errno; 286 log_warn("%s: can't allocate tap fds", __func__); 287 return (ret); 288 } 289 for (i = 0; i < vcp->vcp_nnics; i++) 290 tapfds[i] = -1; 291 292 vm->vm_peerid = peerid; 293 vm->vm_uid = uid; 294 295 /* 296 * From here onward, all failures need cleanup and use goto fail 297 */ 298 299 if (!(vm->vm_state & VM_STATE_RECEIVED)) { 300 if (strlen(vcp->vcp_kernel)) { 301 /* Open external kernel for child */ 302 if ((kernfd = open(vcp->vcp_kernel, O_RDONLY)) == -1) { 303 ret = errno; 304 log_warn("%s: can't open kernel or BIOS " 305 "boot image %s", __func__, vcp->vcp_kernel); 306 goto fail; 307 } 308 } 309 310 /* 311 * Try to open the default BIOS image if no kernel/BIOS has been 312 * specified. The BIOS is an external firmware file that is 313 * typically distributed separately due to an incompatible 314 * license. 315 */ 316 if (kernfd == -1 && 317 (kernfd = open(VM_DEFAULT_BIOS, O_RDONLY)) == -1) { 318 log_warn("can't open %s", VM_DEFAULT_BIOS); 319 ret = VMD_BIOS_MISSING; 320 goto fail; 321 } 322 323 if (vm_checkaccess(kernfd, 324 vmc->vmc_checkaccess & VMOP_CREATE_KERNEL, 325 uid, R_OK) == -1) { 326 log_warnx("vm \"%s\" no read access to kernel %s", 327 vcp->vcp_name, vcp->vcp_kernel); 328 ret = EPERM; 329 goto fail; 330 } 331 } 332 333 /* Open CDROM image for child */ 334 if (strlen(vcp->vcp_cdrom)) { 335 /* Stat cdrom to ensure it is a regular file */ 336 if ((cdromfd = 337 open(vcp->vcp_cdrom, O_RDONLY)) == -1) { 338 log_warn("can't open cdrom %s", vcp->vcp_cdrom); 339 ret = VMD_CDROM_MISSING; 340 goto fail; 341 } 342 343 if (vm_checkaccess(cdromfd, 344 vmc->vmc_checkaccess & VMOP_CREATE_CDROM, 345 uid, R_OK) == -1) { 346 log_warnx("vm \"%s\" no read access to cdrom %s", 347 vcp->vcp_name, vcp->vcp_cdrom); 348 ret = EPERM; 349 goto fail; 350 } 351 } 352 353 /* Open disk images for child */ 354 for (i = 0 ; i < vcp->vcp_ndisks; i++) { 355 if (strlcpy(path, vcp->vcp_disks[i], sizeof(path)) 356 >= sizeof(path)) 357 log_warnx("disk path %s too long", vcp->vcp_disks[i]); 358 memset(vmc->vmc_diskbases, 0, sizeof(vmc->vmc_diskbases)); 359 oflags = O_RDWR|O_EXLOCK|O_NONBLOCK; 360 aflags = R_OK|W_OK; 361 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) { 362 /* Stat disk[i] to ensure it is a regular file */ 363 if ((diskfds[i][j] = open(path, oflags)) == -1) { 364 log_warn("can't open disk %s", 365 vcp->vcp_disks[i]); 366 ret = VMD_DISK_MISSING; 367 goto fail; 368 } 369 370 if (vm_checkaccess(diskfds[i][j], 371 vmc->vmc_checkaccess & VMOP_CREATE_DISK, 372 uid, aflags) == -1) { 373 log_warnx("vm \"%s\" unable to access " 374 "disk %s", vcp->vcp_name, path); 375 errno = EPERM; 376 goto fail; 377 } 378 379 /* 380 * Clear the write and exclusive flags for base images. 381 * All writes should go to the top image, allowing them 382 * to be shared. 383 */ 384 oflags = O_RDONLY|O_NONBLOCK; 385 aflags = R_OK; 386 n = virtio_get_base(diskfds[i][j], base, sizeof(base), 387 vmc->vmc_disktypes[i], path); 388 if (n == 0) 389 break; 390 if (n == -1) { 391 log_warnx("vm \"%s\" unable to read " 392 "base for disk %s", vcp->vcp_name, 393 vcp->vcp_disks[i]); 394 goto fail; 395 } 396 (void)strlcpy(path, base, sizeof(path)); 397 } 398 } 399 400 /* Open network interfaces */ 401 for (i = 0 ; i < vcp->vcp_nnics; i++) { 402 vif = &vm->vm_ifs[i]; 403 404 /* Check if the user has requested a specific tap(4) */ 405 s = vmc->vmc_ifnames[i]; 406 if (*s != '\0' && strcmp("tap", s) != 0) { 407 if (priv_getiftype(s, ifname, &unit) == -1 || 408 strcmp(ifname, "tap") != 0) { 409 log_warnx("%s: invalid tap name %s", 410 __func__, s); 411 ret = EINVAL; 412 goto fail; 413 } 414 } else 415 s = NULL; 416 417 /* 418 * Either open the requested tap(4) device or get 419 * the next available one. 420 */ 421 if (s != NULL) { 422 snprintf(path, PATH_MAX, "/dev/%s", s); 423 tapfds[i] = open(path, O_RDWR | O_NONBLOCK); 424 } else { 425 tapfds[i] = opentap(ifname); 426 s = ifname; 427 } 428 if (tapfds[i] == -1) { 429 log_warnx("%s: can't open tap %s", __func__, s); 430 goto fail; 431 } 432 if ((vif->vif_name = strdup(s)) == NULL) { 433 log_warn("%s: can't save tap %s", __func__, s); 434 goto fail; 435 } 436 437 /* Check if the the interface is attached to a switch */ 438 s = vmc->vmc_ifswitch[i]; 439 if (*s != '\0') { 440 if ((vif->vif_switch = strdup(s)) == NULL) { 441 log_warn("%s: can't save switch %s", 442 __func__, s); 443 goto fail; 444 } 445 } 446 447 /* Check if the the interface is assigned to a group */ 448 s = vmc->vmc_ifgroup[i]; 449 if (*s != '\0') { 450 if ((vif->vif_group = strdup(s)) == NULL) { 451 log_warn("%s: can't save group %s", 452 __func__, s); 453 goto fail; 454 } 455 } 456 457 /* non-default rdomain (requires VMIFF_RDOMAIN below) */ 458 vif->vif_rdomain = vmc->vmc_ifrdomain[i]; 459 460 /* Set the interface status */ 461 vif->vif_flags = 462 vmc->vmc_ifflags[i] & (VMIFF_UP|VMIFF_OPTMASK); 463 } 464 465 /* Open TTY */ 466 if (vm->vm_ttyname == NULL) { 467 if (vm_opentty(vm) == -1) { 468 log_warn("%s: can't open tty %s", __func__, 469 vm->vm_ttyname == NULL ? "" : vm->vm_ttyname); 470 goto fail; 471 } 472 } 473 if ((fd = dup(vm->vm_tty)) == -1) { 474 log_warn("%s: can't re-open tty %s", __func__, vm->vm_ttyname); 475 goto fail; 476 } 477 478 /* Send VM information */ 479 /* XXX check proc_compose_imsg return values */ 480 if (vm->vm_state & VM_STATE_RECEIVED) 481 proc_compose_imsg(ps, PROC_VMM, -1, 482 IMSG_VMDOP_RECEIVE_VM_REQUEST, vm->vm_vmid, fd, vmc, 483 sizeof(struct vmop_create_params)); 484 else 485 proc_compose_imsg(ps, PROC_VMM, -1, 486 IMSG_VMDOP_START_VM_REQUEST, vm->vm_vmid, kernfd, 487 vmc, sizeof(*vmc)); 488 489 if (strlen(vcp->vcp_cdrom)) 490 proc_compose_imsg(ps, PROC_VMM, -1, 491 IMSG_VMDOP_START_VM_CDROM, vm->vm_vmid, cdromfd, 492 NULL, 0); 493 494 for (i = 0; i < vcp->vcp_ndisks; i++) { 495 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) { 496 if (diskfds[i][j] == -1) 497 break; 498 proc_compose_imsg(ps, PROC_VMM, -1, 499 IMSG_VMDOP_START_VM_DISK, vm->vm_vmid, 500 diskfds[i][j], &i, sizeof(i)); 501 } 502 } 503 for (i = 0; i < vcp->vcp_nnics; i++) { 504 proc_compose_imsg(ps, PROC_VMM, -1, 505 IMSG_VMDOP_START_VM_IF, vm->vm_vmid, tapfds[i], 506 &i, sizeof(i)); 507 508 memset(&var, 0, sizeof(var)); 509 var.var_vmid = vm->vm_vmid; 510 var.var_nic_idx = i; 511 proc_compose_imsg(ps, PROC_PRIV, -1, IMSG_VMDOP_PRIV_GET_ADDR, 512 vm->vm_vmid, dup(tapfds[i]), &var, sizeof(var)); 513 } 514 515 if (!(vm->vm_state & VM_STATE_RECEIVED)) 516 proc_compose_imsg(ps, PROC_VMM, -1, 517 IMSG_VMDOP_START_VM_END, vm->vm_vmid, fd, NULL, 0); 518 519 free(tapfds); 520 521 vm->vm_state |= VM_STATE_RUNNING; 522 return (0); 523 524 fail: 525 log_warnx("failed to start vm %s", vcp->vcp_name); 526 527 if (kernfd != -1) 528 close(kernfd); 529 if (cdromfd != -1) 530 close(cdromfd); 531 for (i = 0; i < vcp->vcp_ndisks; i++) 532 for (j = 0; j < VM_MAX_BASE_PER_DISK; j++) 533 if (diskfds[i][j] != -1) 534 close(diskfds[i][j]); 535 if (tapfds != NULL) { 536 for (i = 0; i < vcp->vcp_nnics; i++) 537 close(tapfds[i]); 538 free(tapfds); 539 } 540 541 if (vm->vm_from_config) { 542 vm_stop(vm, 0, __func__); 543 } else { 544 vm_remove(vm, __func__); 545 } 546 547 return (ret); 548 } 549 550 int 551 config_getvm(struct privsep *ps, struct imsg *imsg) 552 { 553 struct vmop_create_params vmc; 554 struct vmd_vm *vm; 555 556 IMSG_SIZE_CHECK(imsg, &vmc); 557 memcpy(&vmc, imsg->data, sizeof(vmc)); 558 559 errno = 0; 560 if (vm_register(ps, &vmc, &vm, imsg->hdr.peerid, 0) == -1) 561 goto fail; 562 563 /* If the fd is -1, the kernel will be searched on the disk */ 564 vm->vm_kernel = imsg->fd; 565 vm->vm_state |= VM_STATE_RUNNING; 566 vm->vm_peerid = (uint32_t)-1; 567 568 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_params.vcp_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_params.vcp_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