| File: | src/usr.sbin/vmd/vmm.c |
| Warning: | line 747, column 2 Value stored to 'ret' is never read |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
| 1 | /* $OpenBSD: vmm.c,v 1.103 2022/01/04 15:25:05 claudio Exp $ */ |
| 2 | |
| 3 | /* |
| 4 | * Copyright (c) 2015 Mike Larkin <mlarkin@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/ioctl.h> |
| 21 | #include <sys/queue.h> |
| 22 | #include <sys/wait.h> |
| 23 | #include <sys/uio.h> |
| 24 | #include <sys/socket.h> |
| 25 | #include <sys/time.h> |
| 26 | #include <sys/mman.h> |
| 27 | |
| 28 | #include <dev/ic/i8253reg.h> |
| 29 | #include <dev/isa/isareg.h> |
| 30 | #include <dev/pci/pcireg.h> |
| 31 | |
| 32 | #include <machine/psl.h> |
| 33 | #include <machine/specialreg.h> |
| 34 | #include <machine/vmmvar.h> |
| 35 | |
| 36 | #include <net/if.h> |
| 37 | |
| 38 | #include <errno(*__errno()).h> |
| 39 | #include <event.h> |
| 40 | #include <fcntl.h> |
| 41 | #include <imsg.h> |
| 42 | #include <limits.h> |
| 43 | #include <poll.h> |
| 44 | #include <pthread.h> |
| 45 | #include <stddef.h> |
| 46 | #include <stdio.h> |
| 47 | #include <stdlib.h> |
| 48 | #include <string.h> |
| 49 | #include <unistd.h> |
| 50 | #include <util.h> |
| 51 | |
| 52 | #include "vmd.h" |
| 53 | #include "vmm.h" |
| 54 | |
| 55 | void vmm_sighdlr(int, short, void *); |
| 56 | int vmm_start_vm(struct imsg *, uint32_t *, pid_t *); |
| 57 | int vmm_dispatch_parent(int, struct privsep_proc *, struct imsg *); |
| 58 | void vmm_run(struct privsep *, struct privsep_proc *, void *); |
| 59 | void vmm_dispatch_vm(int, short, void *); |
| 60 | int terminate_vm(struct vm_terminate_params *); |
| 61 | int get_info_vm(struct privsep *, struct imsg *, int); |
| 62 | int opentap(char *); |
| 63 | |
| 64 | extern struct vmd *env; |
| 65 | |
| 66 | static struct privsep_proc procs[] = { |
| 67 | { "parent", PROC_PARENT, vmm_dispatch_parent }, |
| 68 | }; |
| 69 | |
| 70 | void |
| 71 | vmm(struct privsep *ps, struct privsep_proc *p) |
| 72 | { |
| 73 | proc_run(ps, p, procs, nitems(procs)(sizeof((procs)) / sizeof((procs)[0])), vmm_run, NULL((void*)0)); |
| 74 | } |
| 75 | |
| 76 | void |
| 77 | vmm_run(struct privsep *ps, struct privsep_proc *p, void *arg) |
| 78 | { |
| 79 | if (config_init(ps->ps_env) == -1) |
| 80 | fatal("failed to initialize configuration"); |
| 81 | |
| 82 | signal_del(&ps->ps_evsigchld)event_del(&ps->ps_evsigchld); |
| 83 | signal_set(&ps->ps_evsigchld, SIGCHLD, vmm_sighdlr, ps)event_set(&ps->ps_evsigchld, 20, 0x08|0x10, vmm_sighdlr , ps); |
| 84 | signal_add(&ps->ps_evsigchld, NULL)event_add(&ps->ps_evsigchld, ((void*)0)); |
| 85 | |
| 86 | /* |
| 87 | * pledge in the vmm process: |
| 88 | * stdio - for malloc and basic I/O including events. |
| 89 | * vmm - for the vmm ioctls and operations. |
| 90 | * proc - for forking and maitaining vms. |
| 91 | * send - for sending send/recv fds to vm proc. |
| 92 | * recvfd - for disks, interfaces and other fds. |
| 93 | */ |
| 94 | if (pledge("stdio vmm sendfd recvfd proc", NULL((void*)0)) == -1) |
| 95 | fatal("pledge"); |
| 96 | |
| 97 | /* Get and terminate all running VMs */ |
| 98 | get_info_vm(ps, NULL((void*)0), 1); |
| 99 | } |
| 100 | |
| 101 | int |
| 102 | vmm_dispatch_parent(int fd, struct privsep_proc *p, struct imsg *imsg) |
| 103 | { |
| 104 | struct privsep *ps = p->p_ps; |
| 105 | int res = 0, cmd = 0, verbose; |
| 106 | struct vmd_vm *vm = NULL((void*)0); |
| 107 | struct vm_terminate_params vtp; |
| 108 | struct vmop_id vid; |
| 109 | struct vmop_result vmr; |
| 110 | struct vmop_create_params vmc; |
| 111 | struct vmop_addr_result var; |
| 112 | uint32_t id = 0, peerid = imsg->hdr.peerid; |
| 113 | pid_t pid = 0; |
| 114 | unsigned int mode, flags; |
| 115 | |
| 116 | switch (imsg->hdr.type) { |
| 117 | case IMSG_VMDOP_START_VM_REQUEST: |
| 118 | res = config_getvm(ps, imsg); |
| 119 | if (res == -1) { |
| 120 | res = errno(*__errno()); |
| 121 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 122 | } |
| 123 | break; |
| 124 | case IMSG_VMDOP_START_VM_CDROM: |
| 125 | res = config_getcdrom(ps, imsg); |
| 126 | if (res == -1) { |
| 127 | res = errno(*__errno()); |
| 128 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 129 | } |
| 130 | break; |
| 131 | case IMSG_VMDOP_START_VM_DISK: |
| 132 | res = config_getdisk(ps, imsg); |
| 133 | if (res == -1) { |
| 134 | res = errno(*__errno()); |
| 135 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 136 | } |
| 137 | break; |
| 138 | case IMSG_VMDOP_START_VM_IF: |
| 139 | res = config_getif(ps, imsg); |
| 140 | if (res == -1) { |
| 141 | res = errno(*__errno()); |
| 142 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 143 | } |
| 144 | break; |
| 145 | case IMSG_VMDOP_START_VM_END: |
| 146 | res = vmm_start_vm(imsg, &id, &pid); |
| 147 | /* Check if the ID can be mapped correctly */ |
| 148 | if ((id = vm_id2vmid(id, NULL((void*)0))) == 0) |
| 149 | res = ENOENT2; |
| 150 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 151 | break; |
| 152 | case IMSG_VMDOP_TERMINATE_VM_REQUEST: |
| 153 | IMSG_SIZE_CHECK(imsg, &vid)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&vid)) fatalx("bad length imsg received (%s)", "&vid" ); } while (0); |
| 154 | memcpy(&vid, imsg->data, sizeof(vid)); |
| 155 | id = vid.vid_id; |
| 156 | flags = vid.vid_flags; |
| 157 | |
| 158 | DPRINTF("%s: recv'ed TERMINATE_VM for %d", __func__, id)do {} while(0); |
| 159 | |
| 160 | cmd = IMSG_VMDOP_TERMINATE_VM_RESPONSE; |
| 161 | |
| 162 | if (id == 0) { |
| 163 | res = ENOENT2; |
| 164 | } else if ((vm = vm_getbyvmid(id)) != NULL((void*)0)) { |
| 165 | if (flags & VMOP_FORCE0x01) { |
| 166 | vtp.vtp_vm_id = vm_vmid2id(vm->vm_vmid, vm); |
| 167 | vm->vm_state |= VM_STATE_SHUTDOWN0x04; |
| 168 | (void)terminate_vm(&vtp); |
| 169 | res = 0; |
| 170 | } else if (!(vm->vm_state & VM_STATE_SHUTDOWN0x04)) { |
| 171 | log_debug("%s: sending shutdown request" |
| 172 | " to vm %d", __func__, id); |
| 173 | |
| 174 | /* |
| 175 | * Request reboot but mark the VM as shutting |
| 176 | * down. This way we can terminate the VM after |
| 177 | * the triple fault instead of reboot and |
| 178 | * avoid being stuck in the ACPI-less powerdown |
| 179 | * ("press any key to reboot") of the VM. |
| 180 | */ |
| 181 | vm->vm_state |= VM_STATE_SHUTDOWN0x04; |
| 182 | if (imsg_compose_event(&vm->vm_iev, |
| 183 | IMSG_VMDOP_VM_REBOOT, |
| 184 | 0, 0, -1, NULL((void*)0), 0) == -1) |
| 185 | res = errno(*__errno()); |
| 186 | else |
| 187 | res = 0; |
| 188 | } else { |
| 189 | /* |
| 190 | * VM is currently being shutdown. |
| 191 | * Check to see if the VM process is still |
| 192 | * active. If not, return VMD_VM_STOP_INVALID. |
| 193 | */ |
| 194 | if (vm_vmid2id(vm->vm_vmid, vm) == 0) { |
| 195 | log_debug("%s: no vm running anymore", |
| 196 | __func__); |
| 197 | res = VMD_VM_STOP_INVALID1004; |
| 198 | } |
| 199 | } |
| 200 | } else { |
| 201 | /* VM doesn't exist, cannot stop vm */ |
| 202 | log_debug("%s: cannot stop vm that is not running", |
| 203 | __func__); |
| 204 | res = VMD_VM_STOP_INVALID1004; |
| 205 | } |
| 206 | break; |
| 207 | case IMSG_VMDOP_GET_INFO_VM_REQUEST: |
| 208 | res = get_info_vm(ps, imsg, 0); |
| 209 | cmd = IMSG_VMDOP_GET_INFO_VM_END_DATA; |
| 210 | break; |
| 211 | case IMSG_VMDOP_CONFIG: |
| 212 | config_getconfig(env, imsg); |
| 213 | break; |
| 214 | case IMSG_CTL_RESET: |
| 215 | IMSG_SIZE_CHECK(imsg, &mode)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&mode)) fatalx("bad length imsg received (%s)", "&mode" ); } while (0); |
| 216 | memcpy(&mode, imsg->data, sizeof(mode)); |
| 217 | |
| 218 | if (mode & CONFIG_VMS0x01) { |
| 219 | /* Terminate and remove all VMs */ |
| 220 | vmm_shutdown(); |
| 221 | mode &= ~CONFIG_VMS0x01; |
| 222 | } |
| 223 | |
| 224 | config_getreset(env, imsg); |
| 225 | break; |
| 226 | case IMSG_CTL_VERBOSE: |
| 227 | IMSG_SIZE_CHECK(imsg, &verbose)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&verbose)) fatalx("bad length imsg received (%s)", "&verbose" ); } while (0); |
| 228 | memcpy(&verbose, imsg->data, sizeof(verbose)); |
| 229 | log_setverbose(verbose); |
| 230 | |
| 231 | /* Forward message to each VM process */ |
| 232 | TAILQ_FOREACH(vm, env->vmd_vms, vm_entry)for((vm) = ((env->vmd_vms)->tqh_first); (vm) != ((void* )0); (vm) = ((vm)->vm_entry.tqe_next)) { |
| 233 | imsg_compose_event(&vm->vm_iev, |
| 234 | imsg->hdr.type, imsg->hdr.peerid, imsg->hdr.pid, |
| 235 | -1, &verbose, sizeof(verbose)); |
| 236 | } |
| 237 | break; |
| 238 | case IMSG_VMDOP_PAUSE_VM: |
| 239 | IMSG_SIZE_CHECK(imsg, &vid)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&vid)) fatalx("bad length imsg received (%s)", "&vid" ); } while (0); |
| 240 | memcpy(&vid, imsg->data, sizeof(vid)); |
| 241 | id = vid.vid_id; |
| 242 | if ((vm = vm_getbyvmid(id)) == NULL((void*)0)) { |
| 243 | res = ENOENT2; |
| 244 | cmd = IMSG_VMDOP_PAUSE_VM_RESPONSE; |
| 245 | break; |
| 246 | } |
| 247 | imsg_compose_event(&vm->vm_iev, |
| 248 | imsg->hdr.type, imsg->hdr.peerid, imsg->hdr.pid, |
| 249 | imsg->fd, &vid, sizeof(vid)); |
| 250 | break; |
| 251 | case IMSG_VMDOP_UNPAUSE_VM: |
| 252 | IMSG_SIZE_CHECK(imsg, &vid)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&vid)) fatalx("bad length imsg received (%s)", "&vid" ); } while (0); |
| 253 | memcpy(&vid, imsg->data, sizeof(vid)); |
| 254 | id = vid.vid_id; |
| 255 | if ((vm = vm_getbyvmid(id)) == NULL((void*)0)) { |
| 256 | res = ENOENT2; |
| 257 | cmd = IMSG_VMDOP_UNPAUSE_VM_RESPONSE; |
| 258 | break; |
| 259 | } |
| 260 | imsg_compose_event(&vm->vm_iev, |
| 261 | imsg->hdr.type, imsg->hdr.peerid, imsg->hdr.pid, |
| 262 | imsg->fd, &vid, sizeof(vid)); |
| 263 | break; |
| 264 | case IMSG_VMDOP_SEND_VM_REQUEST: |
| 265 | IMSG_SIZE_CHECK(imsg, &vid)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&vid)) fatalx("bad length imsg received (%s)", "&vid" ); } while (0); |
| 266 | memcpy(&vid, imsg->data, sizeof(vid)); |
| 267 | id = vid.vid_id; |
| 268 | if ((vm = vm_getbyvmid(id)) == NULL((void*)0)) { |
| 269 | res = ENOENT2; |
| 270 | close(imsg->fd); |
| 271 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 272 | break; |
| 273 | } |
| 274 | imsg_compose_event(&vm->vm_iev, |
| 275 | imsg->hdr.type, imsg->hdr.peerid, imsg->hdr.pid, |
| 276 | imsg->fd, &vid, sizeof(vid)); |
| 277 | break; |
| 278 | case IMSG_VMDOP_RECEIVE_VM_REQUEST: |
| 279 | IMSG_SIZE_CHECK(imsg, &vmc)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&vmc)) fatalx("bad length imsg received (%s)", "&vmc" ); } while (0); |
| 280 | memcpy(&vmc, imsg->data, sizeof(vmc)); |
| 281 | if (vm_register(ps, &vmc, &vm, |
| 282 | imsg->hdr.peerid, vmc.vmc_owner.uid) != 0) { |
| 283 | res = errno(*__errno()); |
| 284 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 285 | break; |
| 286 | } |
| 287 | vm->vm_tty = imsg->fd; |
| 288 | vm->vm_state |= VM_STATE_RECEIVED0x08; |
| 289 | vm->vm_state |= VM_STATE_PAUSED0x10; |
| 290 | break; |
| 291 | case IMSG_VMDOP_RECEIVE_VM_END: |
| 292 | if ((vm = vm_getbyvmid(imsg->hdr.peerid)) == NULL((void*)0)) { |
| 293 | res = ENOENT2; |
| 294 | close(imsg->fd); |
| 295 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 296 | break; |
| 297 | } |
| 298 | vm->vm_receive_fd = imsg->fd; |
| 299 | res = vmm_start_vm(imsg, &id, &pid); |
| 300 | /* Check if the ID can be mapped correctly */ |
| 301 | if ((id = vm_id2vmid(id, NULL((void*)0))) == 0) |
| 302 | res = ENOENT2; |
| 303 | cmd = IMSG_VMDOP_START_VM_RESPONSE; |
| 304 | break; |
| 305 | case IMSG_VMDOP_PRIV_GET_ADDR_RESPONSE: |
| 306 | IMSG_SIZE_CHECK(imsg, &var)do { if (((imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof (*&var)) fatalx("bad length imsg received (%s)", "&var" ); } while (0); |
| 307 | memcpy(&var, imsg->data, sizeof(var)); |
| 308 | if ((vm = vm_getbyvmid(var.var_vmid)) == NULL((void*)0)) { |
| 309 | res = ENOENT2; |
| 310 | break; |
| 311 | } |
| 312 | /* Forward hardware address details to the guest vm */ |
| 313 | imsg_compose_event(&vm->vm_iev, |
| 314 | imsg->hdr.type, imsg->hdr.peerid, imsg->hdr.pid, |
| 315 | imsg->fd, &var, sizeof(var)); |
| 316 | break; |
| 317 | default: |
| 318 | return (-1); |
| 319 | } |
| 320 | |
| 321 | switch (cmd) { |
| 322 | case 0: |
| 323 | break; |
| 324 | case IMSG_VMDOP_START_VM_RESPONSE: |
| 325 | if (res != 0) { |
| 326 | /* Remove local reference if it exists */ |
| 327 | if ((vm = vm_getbyvmid(imsg->hdr.peerid)) != NULL((void*)0)) { |
| 328 | log_debug("%s: removing vm, START_VM_RESPONSE", |
| 329 | __func__); |
| 330 | vm_remove(vm, __func__); |
| 331 | } |
| 332 | } |
| 333 | if (id == 0) |
| 334 | id = imsg->hdr.peerid; |
| 335 | /* FALLTHROUGH */ |
| 336 | case IMSG_VMDOP_PAUSE_VM_RESPONSE: |
| 337 | case IMSG_VMDOP_UNPAUSE_VM_RESPONSE: |
| 338 | case IMSG_VMDOP_TERMINATE_VM_RESPONSE: |
| 339 | memset(&vmr, 0, sizeof(vmr)); |
| 340 | vmr.vmr_result = res; |
| 341 | vmr.vmr_id = id; |
| 342 | vmr.vmr_pid = pid; |
| 343 | if (proc_compose_imsg(ps, PROC_PARENT, -1, cmd, |
| 344 | peerid, -1, &vmr, sizeof(vmr)) == -1) |
| 345 | return (-1); |
| 346 | break; |
| 347 | default: |
| 348 | if (proc_compose_imsg(ps, PROC_PARENT, -1, cmd, |
| 349 | peerid, -1, &res, sizeof(res)) == -1) |
| 350 | return (-1); |
| 351 | break; |
| 352 | } |
| 353 | |
| 354 | return (0); |
| 355 | } |
| 356 | |
| 357 | void |
| 358 | vmm_sighdlr(int sig, short event, void *arg) |
| 359 | { |
| 360 | struct privsep *ps = arg; |
| 361 | int status, ret = 0; |
| 362 | uint32_t vmid; |
| 363 | pid_t pid; |
| 364 | struct vmop_result vmr; |
| 365 | struct vmd_vm *vm; |
| 366 | struct vm_terminate_params vtp; |
| 367 | |
| 368 | log_debug("%s: handling signal %d", __func__, sig); |
| 369 | switch (sig) { |
| 370 | case SIGCHLD20: |
| 371 | do { |
| 372 | pid = waitpid(-1, &status, WNOHANG1); |
| 373 | if (pid <= 0) |
| 374 | continue; |
| 375 | |
| 376 | if (WIFEXITED(status)(((status) & 0177) == 0) || WIFSIGNALED(status)(((status) & 0177) != 0177 && ((status) & 0177 ) != 0)) { |
| 377 | vm = vm_getbypid(pid); |
| 378 | if (vm == NULL((void*)0)) { |
| 379 | /* |
| 380 | * If the VM is gone already, it |
| 381 | * got terminated via a |
| 382 | * IMSG_VMDOP_TERMINATE_VM_REQUEST. |
| 383 | */ |
| 384 | continue; |
| 385 | } |
| 386 | |
| 387 | if (WIFEXITED(status)(((status) & 0177) == 0)) |
| 388 | ret = WEXITSTATUS(status)(int)(((unsigned)(status) >> 8) & 0xff); |
| 389 | |
| 390 | /* Don't reboot on pending shutdown */ |
| 391 | if (ret == EAGAIN35 && |
| 392 | (vm->vm_state & VM_STATE_SHUTDOWN0x04)) |
| 393 | ret = 0; |
| 394 | |
| 395 | vmid = vm->vm_params.vmc_params.vcp_id; |
| 396 | vtp.vtp_vm_id = vmid; |
| 397 | |
| 398 | if (terminate_vm(&vtp) == 0) |
| 399 | log_debug("%s: terminated vm %s" |
| 400 | " (id %d)", __func__, |
| 401 | vm->vm_params.vmc_params.vcp_name, |
| 402 | vm->vm_vmid); |
| 403 | |
| 404 | memset(&vmr, 0, sizeof(vmr)); |
| 405 | vmr.vmr_result = ret; |
| 406 | vmr.vmr_id = vm_id2vmid(vmid, vm); |
| 407 | if (proc_compose_imsg(ps, PROC_PARENT, |
| 408 | -1, IMSG_VMDOP_TERMINATE_VM_EVENT, |
| 409 | vm->vm_peerid, -1, |
| 410 | &vmr, sizeof(vmr)) == -1) |
| 411 | log_warnx("could not signal " |
| 412 | "termination of VM %u to " |
| 413 | "parent", vm->vm_vmid); |
| 414 | |
| 415 | vm_remove(vm, __func__); |
| 416 | } else |
| 417 | fatalx("unexpected cause of SIGCHLD"); |
| 418 | } while (pid > 0 || (pid == -1 && errno(*__errno()) == EINTR4)); |
| 419 | break; |
| 420 | default: |
| 421 | fatalx("unexpected signal"); |
| 422 | } |
| 423 | } |
| 424 | |
| 425 | /* |
| 426 | * vmm_shutdown |
| 427 | * |
| 428 | * Terminate VMs on shutdown to avoid "zombie VM" processes. |
| 429 | */ |
| 430 | void |
| 431 | vmm_shutdown(void) |
| 432 | { |
| 433 | struct vm_terminate_params vtp; |
| 434 | struct vmd_vm *vm, *vm_next; |
| 435 | |
| 436 | TAILQ_FOREACH_SAFE(vm, env->vmd_vms, vm_entry, vm_next)for ((vm) = ((env->vmd_vms)->tqh_first); (vm) != ((void *)0) && ((vm_next) = ((vm)->vm_entry.tqe_next), 1) ; (vm) = (vm_next)) { |
| 437 | vtp.vtp_vm_id = vm_vmid2id(vm->vm_vmid, vm); |
| 438 | |
| 439 | /* XXX suspend or request graceful shutdown */ |
| 440 | (void)terminate_vm(&vtp); |
| 441 | vm_remove(vm, __func__); |
| 442 | } |
| 443 | } |
| 444 | |
| 445 | /* |
| 446 | * vmm_pipe |
| 447 | * |
| 448 | * Create a new imsg control channel between vmm parent and a VM |
| 449 | * (can be called on both sides). |
| 450 | */ |
| 451 | int |
| 452 | vmm_pipe(struct vmd_vm *vm, int fd, void (*cb)(int, short, void *)) |
| 453 | { |
| 454 | struct imsgev *iev = &vm->vm_iev; |
| 455 | |
| 456 | if (fcntl(fd, F_SETFL4, O_NONBLOCK0x0004) == -1) { |
| 457 | log_warn("failed to set nonblocking mode on vm pipe"); |
| 458 | return (-1); |
| 459 | } |
| 460 | |
| 461 | imsg_init(&iev->ibuf, fd); |
| 462 | iev->handler = cb; |
| 463 | iev->data = vm; |
| 464 | imsg_event_add(iev); |
| 465 | |
| 466 | return (0); |
| 467 | } |
| 468 | |
| 469 | /* |
| 470 | * vmm_dispatch_vm |
| 471 | * |
| 472 | * imsg callback for messages that are received from a VM child process. |
| 473 | */ |
| 474 | void |
| 475 | vmm_dispatch_vm(int fd, short event, void *arg) |
| 476 | { |
| 477 | struct vmd_vm *vm = arg; |
| 478 | struct vmop_result vmr; |
| 479 | struct imsgev *iev = &vm->vm_iev; |
| 480 | struct imsgbuf *ibuf = &iev->ibuf; |
| 481 | struct imsg imsg; |
| 482 | ssize_t n; |
| 483 | unsigned int i; |
| 484 | |
| 485 | if (event & EV_READ0x02) { |
| 486 | if ((n = imsg_read(ibuf)) == -1 && errno(*__errno()) != EAGAIN35) |
| 487 | fatal("%s: imsg_read", __func__); |
| 488 | if (n == 0) { |
| 489 | /* This pipe is dead, so remove the event handler */ |
| 490 | event_del(&iev->ev); |
| 491 | return; |
| 492 | } |
| 493 | } |
| 494 | |
| 495 | if (event & EV_WRITE0x04) { |
| 496 | if ((n = msgbuf_write(&ibuf->w)) == -1 && errno(*__errno()) != EAGAIN35) |
| 497 | fatal("%s: msgbuf_write fd %d", __func__, ibuf->fd); |
| 498 | if (n == 0) { |
| 499 | /* This pipe is dead, so remove the event handler */ |
| 500 | event_del(&iev->ev); |
| 501 | return; |
| 502 | } |
| 503 | } |
| 504 | |
| 505 | for (;;) { |
| 506 | if ((n = imsg_get(ibuf, &imsg)) == -1) |
| 507 | fatal("%s: imsg_get", __func__); |
| 508 | if (n == 0) |
| 509 | break; |
| 510 | |
| 511 | DPRINTF("%s: got imsg %d from %s",do {} while(0) |
| 512 | __func__, imsg.hdr.type,do {} while(0) |
| 513 | vm->vm_params.vmc_params.vcp_name)do {} while(0); |
| 514 | |
| 515 | switch (imsg.hdr.type) { |
| 516 | case IMSG_VMDOP_VM_SHUTDOWN: |
| 517 | vm->vm_state |= VM_STATE_SHUTDOWN0x04; |
| 518 | break; |
| 519 | case IMSG_VMDOP_VM_REBOOT: |
| 520 | vm->vm_state &= ~VM_STATE_SHUTDOWN0x04; |
| 521 | break; |
| 522 | case IMSG_VMDOP_SEND_VM_RESPONSE: |
| 523 | IMSG_SIZE_CHECK(&imsg, &vmr)do { if (((&imsg)->hdr.len - sizeof(struct imsg_hdr)) < sizeof(*&vmr)) fatalx("bad length imsg received (%s)", "&vmr" ); } while (0); |
| 524 | case IMSG_VMDOP_PAUSE_VM_RESPONSE: |
| 525 | case IMSG_VMDOP_UNPAUSE_VM_RESPONSE: |
| 526 | for (i = 0; i < nitems(procs)(sizeof((procs)) / sizeof((procs)[0])); i++) { |
| 527 | if (procs[i].p_id == PROC_PARENT) { |
| 528 | proc_forward_imsg(procs[i].p_ps, |
| 529 | &imsg, PROC_PARENT, -1); |
| 530 | break; |
| 531 | } |
| 532 | } |
| 533 | break; |
| 534 | |
| 535 | default: |
| 536 | fatalx("%s: got invalid imsg %d from %s", |
| 537 | __func__, imsg.hdr.type, |
| 538 | vm->vm_params.vmc_params.vcp_name); |
| 539 | } |
| 540 | imsg_free(&imsg); |
| 541 | } |
| 542 | imsg_event_add(iev); |
| 543 | } |
| 544 | |
| 545 | /* |
| 546 | * terminate_vm |
| 547 | * |
| 548 | * Requests vmm(4) to terminate the VM whose ID is provided in the |
| 549 | * supplied vm_terminate_params structure (vtp->vtp_vm_id) |
| 550 | * |
| 551 | * Parameters |
| 552 | * vtp: vm_terminate_params struct containing the ID of the VM to terminate |
| 553 | * |
| 554 | * Return values: |
| 555 | * 0: success |
| 556 | * !0: ioctl to vmm(4) failed (eg, ENOENT if the supplied VM is not valid) |
| 557 | */ |
| 558 | int |
| 559 | terminate_vm(struct vm_terminate_params *vtp) |
| 560 | { |
| 561 | if (ioctl(env->vmd_fd, VMM_IOC_TERM((unsigned long)0x80000000 | ((sizeof(struct vm_terminate_params ) & 0x1fff) << 16) | ((('V')) << 8) | ((4))), vtp) == -1) |
| 562 | return (errno(*__errno())); |
| 563 | |
| 564 | return (0); |
| 565 | } |
| 566 | |
| 567 | /* |
| 568 | * opentap |
| 569 | * |
| 570 | * Opens the next available tap device, up to MAX_TAP. |
| 571 | * |
| 572 | * Parameters |
| 573 | * ifname: a buffer of at least IF_NAMESIZE bytes. |
| 574 | * |
| 575 | * Returns a file descriptor to the tap node opened, or -1 if no tap |
| 576 | * devices were available. |
| 577 | */ |
| 578 | int |
| 579 | opentap(char *ifname) |
| 580 | { |
| 581 | int i, fd; |
| 582 | char path[PATH_MAX1024]; |
| 583 | |
| 584 | for (i = 0; i < MAX_TAP256; i++) { |
| 585 | snprintf(path, PATH_MAX1024, "/dev/tap%d", i); |
| 586 | fd = open(path, O_RDWR0x0002 | O_NONBLOCK0x0004); |
| 587 | if (fd != -1) { |
| 588 | snprintf(ifname, IF_NAMESIZE16, "tap%d", i); |
| 589 | return (fd); |
| 590 | } |
| 591 | } |
| 592 | strlcpy(ifname, "tap", IF_NAMESIZE16); |
| 593 | |
| 594 | return (-1); |
| 595 | } |
| 596 | |
| 597 | /* |
| 598 | * vmm_start_vm |
| 599 | * |
| 600 | * Prepares and forks a new VM process. |
| 601 | * |
| 602 | * Parameters: |
| 603 | * imsg: The VM data structure that is including the VM create parameters. |
| 604 | * id: Returns the VM id as reported by the kernel and obtained from the VM. |
| 605 | * pid: Returns the VM pid to the parent. |
| 606 | * |
| 607 | * Return values: |
| 608 | * 0: success |
| 609 | * !0: failure - typically an errno indicating the source of the failure |
| 610 | */ |
| 611 | int |
| 612 | vmm_start_vm(struct imsg *imsg, uint32_t *id, pid_t *pid) |
| 613 | { |
| 614 | struct vm_create_params *vcp; |
| 615 | struct vmd_vm *vm; |
| 616 | int ret = EINVAL22; |
| 617 | int fds[2]; |
| 618 | size_t i, j; |
| 619 | |
| 620 | if ((vm = vm_getbyvmid(imsg->hdr.peerid)) == NULL((void*)0)) { |
| 621 | log_warnx("%s: can't find vm", __func__); |
| 622 | ret = ENOENT2; |
| 623 | goto err; |
| 624 | } |
| 625 | vcp = &vm->vm_params.vmc_params; |
| 626 | |
| 627 | if (!(vm->vm_state & VM_STATE_RECEIVED0x08)) { |
| 628 | if ((vm->vm_tty = imsg->fd) == -1) { |
| 629 | log_warnx("%s: can't get tty", __func__); |
| 630 | goto err; |
| 631 | } |
| 632 | } |
| 633 | |
| 634 | if (socketpair(AF_UNIX1, SOCK_STREAM1, PF_UNSPEC0, fds) == -1) |
| 635 | fatal("socketpair"); |
| 636 | |
| 637 | /* Start child vmd for this VM (fork, chroot, drop privs) */ |
| 638 | ret = fork(); |
| 639 | |
| 640 | /* Start child failed? - cleanup and leave */ |
| 641 | if (ret == -1) { |
| 642 | log_warnx("%s: start child failed", __func__); |
| 643 | ret = EIO5; |
| 644 | goto err; |
| 645 | } |
| 646 | |
| 647 | if (ret > 0) { |
| 648 | /* Parent */ |
| 649 | vm->vm_pid = ret; |
| 650 | close(fds[1]); |
| 651 | |
| 652 | for (i = 0 ; i < vcp->vcp_ndisks; i++) { |
| 653 | for (j = 0; j < VM_MAX_BASE_PER_DISK4; j++) { |
| 654 | if (vm->vm_disks[i][j] != -1) |
| 655 | close(vm->vm_disks[i][j]); |
| 656 | vm->vm_disks[i][j] = -1; |
| 657 | } |
| 658 | } |
| 659 | for (i = 0 ; i < vcp->vcp_nnics; i++) { |
| 660 | close(vm->vm_ifs[i].vif_fd); |
| 661 | vm->vm_ifs[i].vif_fd = -1; |
| 662 | } |
| 663 | if (vm->vm_kernel != -1) { |
| 664 | close(vm->vm_kernel); |
| 665 | vm->vm_kernel = -1; |
| 666 | } |
| 667 | if (vm->vm_cdrom != -1) { |
| 668 | close(vm->vm_cdrom); |
| 669 | vm->vm_cdrom = -1; |
| 670 | } |
| 671 | if (vm->vm_tty != -1) { |
| 672 | close(vm->vm_tty); |
| 673 | vm->vm_tty = -1; |
| 674 | } |
| 675 | |
| 676 | /* Read back the kernel-generated vm id from the child */ |
| 677 | if (read(fds[0], &vcp->vcp_id, sizeof(vcp->vcp_id)) != |
| 678 | sizeof(vcp->vcp_id)) |
| 679 | fatal("read vcp id"); |
| 680 | |
| 681 | if (vcp->vcp_id == 0) |
| 682 | goto err; |
| 683 | |
| 684 | *id = vcp->vcp_id; |
| 685 | *pid = vm->vm_pid; |
| 686 | |
| 687 | if (vmm_pipe(vm, fds[0], vmm_dispatch_vm) == -1) |
| 688 | fatal("setup vm pipe"); |
| 689 | |
| 690 | return (0); |
| 691 | } else { |
| 692 | /* Child */ |
| 693 | close(fds[0]); |
| 694 | close(PROC_PARENT_SOCK_FILENO3); |
| 695 | |
| 696 | ret = start_vm(vm, fds[1]); |
| 697 | |
| 698 | _exit(ret); |
| 699 | } |
| 700 | |
| 701 | return (0); |
| 702 | |
| 703 | err: |
| 704 | vm_remove(vm, __func__); |
| 705 | |
| 706 | return (ret); |
| 707 | } |
| 708 | |
| 709 | /* |
| 710 | * get_info_vm |
| 711 | * |
| 712 | * Returns a list of VMs known to vmm(4). |
| 713 | * |
| 714 | * Parameters: |
| 715 | * ps: the privsep context. |
| 716 | * imsg: the received imsg including the peer id. |
| 717 | * terminate: terminate the listed vm. |
| 718 | * |
| 719 | * Return values: |
| 720 | * 0: success |
| 721 | * !0: failure (eg, ENOMEM, EIO or another error code from vmm(4) ioctl) |
| 722 | */ |
| 723 | int |
| 724 | get_info_vm(struct privsep *ps, struct imsg *imsg, int terminate) |
| 725 | { |
| 726 | int ret; |
| 727 | size_t ct, i; |
| 728 | struct vm_info_params vip; |
| 729 | struct vm_info_result *info; |
| 730 | struct vm_terminate_params vtp; |
| 731 | struct vmop_info_result vir; |
| 732 | |
| 733 | /* |
| 734 | * We issue the VMM_IOC_INFO ioctl twice, once with an input |
| 735 | * buffer size of 0, which results in vmm(4) returning the |
| 736 | * number of bytes required back to us in vip.vip_size, |
| 737 | * and then we call it again after malloc'ing the required |
| 738 | * number of bytes. |
| 739 | * |
| 740 | * It is possible that we could fail a second time (e.g. if |
| 741 | * another VM was created in the instant between the two |
| 742 | * ioctls, but in that case the caller can just try again |
| 743 | * as vmm(4) will return a zero-sized list in that case. |
| 744 | */ |
| 745 | vip.vip_size = 0; |
| 746 | info = NULL((void*)0); |
| 747 | ret = 0; |
Value stored to 'ret' is never read | |
| 748 | memset(&vir, 0, sizeof(vir)); |
| 749 | |
| 750 | /* First ioctl to see how many bytes needed (vip.vip_size) */ |
| 751 | if (ioctl(env->vmd_fd, VMM_IOC_INFO(((unsigned long)0x80000000|(unsigned long)0x40000000) | ((sizeof (struct vm_info_params) & 0x1fff) << 16) | ((('V')) << 8) | ((3))), &vip) == -1) |
| 752 | return (errno(*__errno())); |
| 753 | |
| 754 | if (vip.vip_info_ct != 0) |
| 755 | return (EIO5); |
| 756 | |
| 757 | info = malloc(vip.vip_size); |
| 758 | if (info == NULL((void*)0)) |
| 759 | return (ENOMEM12); |
| 760 | |
| 761 | /* Second ioctl to get the actual list */ |
| 762 | vip.vip_info = info; |
| 763 | if (ioctl(env->vmd_fd, VMM_IOC_INFO(((unsigned long)0x80000000|(unsigned long)0x40000000) | ((sizeof (struct vm_info_params) & 0x1fff) << 16) | ((('V')) << 8) | ((3))), &vip) == -1) { |
| 764 | ret = errno(*__errno()); |
| 765 | free(info); |
| 766 | return (ret); |
| 767 | } |
| 768 | |
| 769 | /* Return info */ |
| 770 | ct = vip.vip_size / sizeof(struct vm_info_result); |
| 771 | for (i = 0; i < ct; i++) { |
| 772 | if (terminate) { |
| 773 | vtp.vtp_vm_id = info[i].vir_id; |
| 774 | if ((ret = terminate_vm(&vtp)) != 0) |
| 775 | return (ret); |
| 776 | log_debug("%s: terminated vm %s (id %d)", __func__, |
| 777 | info[i].vir_name, info[i].vir_id); |
| 778 | continue; |
| 779 | } |
| 780 | memcpy(&vir.vir_info, &info[i], sizeof(vir.vir_info)); |
| 781 | vir.vir_info.vir_id = vm_id2vmid(info[i].vir_id, NULL((void*)0)); |
| 782 | if (proc_compose_imsg(ps, PROC_PARENT, -1, |
| 783 | IMSG_VMDOP_GET_INFO_VM_DATA, imsg->hdr.peerid, -1, |
| 784 | &vir, sizeof(vir)) == -1) |
| 785 | return (EIO5); |
| 786 | } |
| 787 | free(info); |
| 788 | |
| 789 | return (0); |
| 790 | } |