| 1 | /* |
| 2 | * Copyright (c) 2007 The DragonFly Project. All rights reserved. |
| 3 | * |
| 4 | * This code is derived from software contributed to The DragonFly Project |
| 5 | * by Sepherosa Ziehau <sepherosa@gmail.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 | * $DragonFly: src/sys/dev/virtual/net/if_vke.c,v 1.10 2008/05/27 23:44:46 dillon Exp $ |
| 35 | */ |
| 36 | |
| 37 | #include <sys/param.h> |
| 38 | #include <sys/endian.h> |
| 39 | #include <sys/kernel.h> |
| 40 | #include <sys/malloc.h> |
| 41 | #include <sys/proc.h> |
| 42 | #include <sys/serialize.h> |
| 43 | #include <sys/socket.h> |
| 44 | #include <sys/sockio.h> |
| 45 | #include <sys/sysctl.h> |
| 46 | |
| 47 | #include <machine/md_var.h> |
| 48 | #include <machine/cothread.h> |
| 49 | |
| 50 | #include <net/ethernet.h> |
| 51 | #include <net/if.h> |
| 52 | #include <net/bpf.h> |
| 53 | #include <net/if_arp.h> |
| 54 | #include <net/ifq_var.h> |
| 55 | |
| 56 | #include <netinet/in_var.h> |
| 57 | |
| 58 | #include <sys/stat.h> |
| 59 | #include <net/tap/if_tap.h> |
| 60 | #include <err.h> |
| 61 | #include <errno.h> |
| 62 | #include <stdio.h> |
| 63 | #include <string.h> |
| 64 | #include <unistd.h> |
| 65 | #include <fcntl.h> |
| 66 | |
| 67 | #define VKE_DEVNAME "vke" |
| 68 | |
| 69 | #define VKE_CHUNK 8 /* number of mbufs to queue before interrupting */ |
| 70 | |
| 71 | #define NETFIFOSIZE 256 |
| 72 | #define NETFIFOMASK (NETFIFOSIZE -1) |
| 73 | #define NETFIFOINDEX(u) ((u) & NETFIFOMASK) |
| 74 | |
| 75 | #define VKE_COTD_RUN 0 |
| 76 | #define VKE_COTD_EXIT 1 |
| 77 | #define VKE_COTD_DEAD 2 |
| 78 | |
| 79 | struct vke_fifo { |
| 80 | struct mbuf *array[NETFIFOSIZE]; |
| 81 | int rindex; |
| 82 | int windex; |
| 83 | }; |
| 84 | typedef struct vke_fifo *fifo_t; |
| 85 | |
| 86 | struct vke_softc { |
| 87 | struct arpcom arpcom; |
| 88 | int sc_fd; |
| 89 | int sc_unit; |
| 90 | |
| 91 | cothread_t cotd_tx; |
| 92 | cothread_t cotd_rx; |
| 93 | |
| 94 | int cotd_tx_exit; |
| 95 | int cotd_rx_exit; |
| 96 | |
| 97 | void *sc_txbuf; |
| 98 | int sc_txbuf_len; |
| 99 | |
| 100 | fifo_t sc_txfifo; |
| 101 | fifo_t sc_txfifo_done; |
| 102 | fifo_t sc_rxfifo; |
| 103 | |
| 104 | struct sysctl_ctx_list sc_sysctl_ctx; |
| 105 | struct sysctl_oid *sc_sysctl_tree; |
| 106 | |
| 107 | int sc_tap_unit; /* unit of backend tap(4) */ |
| 108 | in_addr_t sc_addr; /* address */ |
| 109 | in_addr_t sc_mask; /* netmask */ |
| 110 | }; |
| 111 | |
| 112 | static void vke_start(struct ifnet *); |
| 113 | static void vke_init(void *); |
| 114 | static int vke_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *); |
| 115 | |
| 116 | static int vke_attach(const struct vknetif_info *, int); |
| 117 | static int vke_stop(struct vke_softc *); |
| 118 | static int vke_init_addr(struct ifnet *, in_addr_t, in_addr_t); |
| 119 | static void vke_tx_intr(cothread_t cotd); |
| 120 | static void vke_tx_thread(cothread_t cotd); |
| 121 | static void vke_rx_intr(cothread_t cotd); |
| 122 | static void vke_rx_thread(cothread_t cotd); |
| 123 | |
| 124 | static int vke_txfifo_enqueue(struct vke_softc *sc, struct mbuf *m); |
| 125 | static struct mbuf *vke_txfifo_dequeue(struct vke_softc *sc); |
| 126 | |
| 127 | static int vke_txfifo_done_enqueue(struct vke_softc *sc, struct mbuf *m); |
| 128 | static struct mbuf * vke_txfifo_done_dequeue(struct vke_softc *sc, struct mbuf *nm); |
| 129 | |
| 130 | static struct mbuf *vke_rxfifo_dequeue(struct vke_softc *sc, struct mbuf *nm); |
| 131 | static struct mbuf *vke_rxfifo_sniff(struct vke_softc *sc); |
| 132 | |
| 133 | static void |
| 134 | vke_sysinit(void *arg __unused) |
| 135 | { |
| 136 | int i, unit; |
| 137 | |
| 138 | KASSERT(NetifNum <= VKNETIF_MAX, ("too many netifs: %d\n", NetifNum)); |
| 139 | |
| 140 | unit = 0; |
| 141 | for (i = 0; i < NetifNum; ++i) { |
| 142 | if (vke_attach(&NetifInfo[i], unit) == 0) |
| 143 | ++unit; |
| 144 | } |
| 145 | } |
| 146 | SYSINIT(vke, SI_SUB_DRIVERS, SI_ORDER_MIDDLE, vke_sysinit, NULL); |
| 147 | |
| 148 | /* |
| 149 | * vke_txfifo_done_enqueue() - Add an mbuf to the transmit done fifo. Since |
| 150 | * the cothread cannot free transmit mbufs after processing we put them on |
| 151 | * the done fifo so the kernel can free them. |
| 152 | */ |
| 153 | static int |
| 154 | vke_txfifo_done_enqueue(struct vke_softc *sc, struct mbuf *m) |
| 155 | { |
| 156 | fifo_t fifo = sc->sc_txfifo_done; |
| 157 | |
| 158 | if (NETFIFOINDEX(fifo->windex + 1) == NETFIFOINDEX(fifo->rindex)) |
| 159 | return (-1); |
| 160 | |
| 161 | fifo->array[NETFIFOINDEX(fifo->windex)] = m; |
| 162 | cpu_sfence(); |
| 163 | ++fifo->windex; |
| 164 | return (0); |
| 165 | } |
| 166 | |
| 167 | /* |
| 168 | * vke_txfifo_done_dequeue() - Remove an mbuf from the transmit done fifo. |
| 169 | */ |
| 170 | static struct mbuf * |
| 171 | vke_txfifo_done_dequeue(struct vke_softc *sc, struct mbuf *nm) |
| 172 | { |
| 173 | fifo_t fifo = sc->sc_txfifo_done; |
| 174 | struct mbuf *m; |
| 175 | |
| 176 | if (NETFIFOINDEX(fifo->rindex) == NETFIFOINDEX(fifo->windex)) |
| 177 | return (NULL); |
| 178 | |
| 179 | m = fifo->array[NETFIFOINDEX(fifo->rindex)]; |
| 180 | fifo->array[NETFIFOINDEX(fifo->rindex)] = nm; |
| 181 | cpu_lfence(); |
| 182 | ++fifo->rindex; |
| 183 | return (m); |
| 184 | } |
| 185 | |
| 186 | /* |
| 187 | * vke_txfifo_enqueue() - Add an mbuf to the transmit fifo. Wake up the |
| 188 | * cothread via cothread_signal(). |
| 189 | */ |
| 190 | static int |
| 191 | vke_txfifo_enqueue(struct vke_softc *sc, struct mbuf *m) |
| 192 | { |
| 193 | fifo_t fifo = sc->sc_txfifo; |
| 194 | cothread_t cotd = sc->cotd_tx; |
| 195 | |
| 196 | if (NETFIFOINDEX(fifo->windex + 1) == NETFIFOINDEX(fifo->rindex)) |
| 197 | return (-1); |
| 198 | |
| 199 | fifo->array[NETFIFOINDEX(fifo->windex)] = m; |
| 200 | cpu_sfence(); |
| 201 | cothread_signal(cotd); |
| 202 | ++fifo->windex; |
| 203 | |
| 204 | return (0); |
| 205 | } |
| 206 | |
| 207 | /* |
| 208 | * vke_txfifo_dequeue() - Return next mbuf on the transmit fifo if one |
| 209 | * exists. |
| 210 | */ |
| 211 | static struct mbuf * |
| 212 | vke_txfifo_dequeue(struct vke_softc *sc) |
| 213 | { |
| 214 | fifo_t fifo = sc->sc_txfifo; |
| 215 | struct mbuf *m; |
| 216 | |
| 217 | if (NETFIFOINDEX(fifo->rindex) == NETFIFOINDEX(fifo->windex)) |
| 218 | return (NULL); |
| 219 | |
| 220 | m = fifo->array[NETFIFOINDEX(fifo->rindex)]; |
| 221 | fifo->array[NETFIFOINDEX(fifo->rindex)] = NULL; |
| 222 | |
| 223 | cpu_lfence(); |
| 224 | ++fifo->rindex; |
| 225 | return (m); |
| 226 | } |
| 227 | |
| 228 | /* |
| 229 | * vke_rxfifo_dequeue() - Return next mbuf on the receice fifo if one |
| 230 | * exists replacing it with newm which should point to a newly allocated |
| 231 | * mbuf. |
| 232 | */ |
| 233 | static struct mbuf * |
| 234 | vke_rxfifo_dequeue(struct vke_softc *sc, struct mbuf *newm) |
| 235 | { |
| 236 | fifo_t fifo = sc->sc_rxfifo; |
| 237 | struct mbuf *m; |
| 238 | |
| 239 | if (NETFIFOINDEX(fifo->rindex) == NETFIFOINDEX(fifo->windex)) |
| 240 | return (NULL); |
| 241 | |
| 242 | m = fifo->array[NETFIFOINDEX(fifo->rindex)]; |
| 243 | fifo->array[NETFIFOINDEX(fifo->rindex)] = newm; |
| 244 | cpu_lfence(); |
| 245 | ++fifo->rindex; |
| 246 | return (m); |
| 247 | } |
| 248 | |
| 249 | /* |
| 250 | * Return the next mbuf if available but do NOT remove it from the FIFO. |
| 251 | */ |
| 252 | static struct mbuf * |
| 253 | vke_rxfifo_sniff(struct vke_softc *sc) |
| 254 | { |
| 255 | fifo_t fifo = sc->sc_rxfifo; |
| 256 | struct mbuf *m; |
| 257 | |
| 258 | if (NETFIFOINDEX(fifo->rindex) == NETFIFOINDEX(fifo->windex)) |
| 259 | return (NULL); |
| 260 | |
| 261 | m = fifo->array[NETFIFOINDEX(fifo->rindex)]; |
| 262 | cpu_lfence(); |
| 263 | return (m); |
| 264 | } |
| 265 | |
| 266 | static void |
| 267 | vke_init(void *xsc) |
| 268 | { |
| 269 | struct vke_softc *sc = xsc; |
| 270 | struct ifnet *ifp = &sc->arpcom.ac_if; |
| 271 | int i; |
| 272 | |
| 273 | ASSERT_SERIALIZED(ifp->if_serializer); |
| 274 | |
| 275 | |
| 276 | vke_stop(sc); |
| 277 | |
| 278 | ifp->if_flags |= IFF_RUNNING; |
| 279 | ifp->if_flags &= ~IFF_OACTIVE; |
| 280 | |
| 281 | if (sc->sc_addr != 0) { |
| 282 | in_addr_t addr, mask; |
| 283 | |
| 284 | addr = sc->sc_addr; |
| 285 | mask = sc->sc_mask; |
| 286 | |
| 287 | /* |
| 288 | * Make sure vkernel assigned |
| 289 | * address will not be added |
| 290 | * again. |
| 291 | */ |
| 292 | sc->sc_addr = 0; |
| 293 | sc->sc_mask = 0; |
| 294 | |
| 295 | vke_init_addr(ifp, addr, mask); |
| 296 | } |
| 297 | |
| 298 | sc->sc_txfifo = kmalloc(sizeof(*sc->sc_txfifo), M_DEVBUF, M_WAITOK); |
| 299 | sc->sc_txfifo_done = kmalloc(sizeof(*sc->sc_txfifo_done), M_DEVBUF, M_WAITOK); |
| 300 | |
| 301 | sc->sc_rxfifo = kmalloc(sizeof(*sc->sc_rxfifo), M_DEVBUF, M_WAITOK); |
| 302 | for (i = 0; i < NETFIFOSIZE; i++) { |
| 303 | sc->sc_rxfifo->array[i] = m_getcl(MB_WAIT, MT_DATA, M_PKTHDR); |
| 304 | sc->sc_txfifo->array[i] = NULL; |
| 305 | sc->sc_txfifo_done->array[i] = NULL; |
| 306 | } |
| 307 | |
| 308 | sc->cotd_tx_exit = sc->cotd_rx_exit = VKE_COTD_RUN; |
| 309 | sc->cotd_tx = cothread_create(vke_tx_thread, vke_tx_intr, sc, "vke_tx"); |
| 310 | sc->cotd_rx = cothread_create(vke_rx_thread, vke_rx_intr, sc, "vke_rx"); |
| 311 | } |
| 312 | |
| 313 | static void |
| 314 | vke_start(struct ifnet *ifp) |
| 315 | { |
| 316 | struct vke_softc *sc = ifp->if_softc; |
| 317 | struct mbuf *m; |
| 318 | cothread_t cotd = sc->cotd_tx; |
| 319 | int count; |
| 320 | |
| 321 | ASSERT_SERIALIZED(ifp->if_serializer); |
| 322 | |
| 323 | if ((ifp->if_flags & (IFF_RUNNING | IFF_OACTIVE)) != IFF_RUNNING) |
| 324 | return; |
| 325 | |
| 326 | cothread_lock(cotd, 0); |
| 327 | count = 0; |
| 328 | |
| 329 | while ((m = ifq_dequeue(&ifp->if_snd, NULL)) != NULL) { |
| 330 | if (vke_txfifo_enqueue(sc, m) != -1) { |
| 331 | if (count++ == VKE_CHUNK) { |
| 332 | cothread_signal(cotd); |
| 333 | count = 0; |
| 334 | } |
| 335 | } else { |
| 336 | m_freem(m); |
| 337 | } |
| 338 | } |
| 339 | |
| 340 | if (count) { |
| 341 | cothread_signal(cotd); |
| 342 | } |
| 343 | |
| 344 | cothread_unlock(cotd, 0); |
| 345 | } |
| 346 | |
| 347 | static int |
| 348 | vke_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr) |
| 349 | { |
| 350 | struct vke_softc *sc = ifp->if_softc; |
| 351 | int error = 0; |
| 352 | |
| 353 | ASSERT_SERIALIZED(ifp->if_serializer); |
| 354 | |
| 355 | switch (cmd) { |
| 356 | case SIOCSIFFLAGS: |
| 357 | if (ifp->if_flags & IFF_UP) { |
| 358 | if ((ifp->if_flags & IFF_RUNNING) == 0) |
| 359 | vke_init(sc); |
| 360 | } else { |
| 361 | if (ifp->if_flags & IFF_RUNNING) |
| 362 | vke_stop(sc); |
| 363 | } |
| 364 | break; |
| 365 | case SIOCGIFMEDIA: |
| 366 | case SIOCSIFMEDIA: |
| 367 | error = EOPNOTSUPP; |
| 368 | /* TODO */ |
| 369 | break; |
| 370 | case SIOCGIFSTATUS: { |
| 371 | struct ifstat *ifs = (struct ifstat *)data; |
| 372 | int len; |
| 373 | |
| 374 | len = strlen(ifs->ascii); |
| 375 | if (len < sizeof(ifs->ascii)) { |
| 376 | ksnprintf(ifs->ascii + len, sizeof(ifs->ascii) - len, |
| 377 | "\tBacked by tap%d\n", sc->sc_tap_unit); |
| 378 | } |
| 379 | break; |
| 380 | } |
| 381 | case SIOCSIFADDR: |
| 382 | if (((struct ifaddr *)data)->ifa_addr->sa_family == AF_INET) { |
| 383 | /* |
| 384 | * If we are explicitly requested to change address, |
| 385 | * we should invalidate address/netmask passed in |
| 386 | * from vkernel command line. |
| 387 | */ |
| 388 | sc->sc_addr = 0; |
| 389 | sc->sc_mask = 0; |
| 390 | } |
| 391 | /* FALL THROUGH */ |
| 392 | default: |
| 393 | error = ether_ioctl(ifp, cmd, data); |
| 394 | break; |
| 395 | } |
| 396 | return error; |
| 397 | } |
| 398 | |
| 399 | static int |
| 400 | vke_stop(struct vke_softc *sc) |
| 401 | { |
| 402 | struct ifnet *ifp = &sc->arpcom.ac_if; |
| 403 | int i; |
| 404 | |
| 405 | ASSERT_SERIALIZED(ifp->if_serializer); |
| 406 | |
| 407 | ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE); |
| 408 | |
| 409 | if (sc) { |
| 410 | if (sc->cotd_tx) { |
| 411 | cothread_lock(sc->cotd_tx, 0); |
| 412 | if (sc->cotd_tx_exit == VKE_COTD_RUN) |
| 413 | sc->cotd_tx_exit = VKE_COTD_EXIT; |
| 414 | cothread_signal(sc->cotd_tx); |
| 415 | cothread_unlock(sc->cotd_tx, 0); |
| 416 | cothread_delete(&sc->cotd_tx); |
| 417 | } |
| 418 | if (sc->cotd_rx) { |
| 419 | cothread_lock(sc->cotd_rx, 0); |
| 420 | if (sc->cotd_rx_exit == VKE_COTD_RUN) |
| 421 | sc->cotd_rx_exit = VKE_COTD_EXIT; |
| 422 | cothread_signal(sc->cotd_rx); |
| 423 | cothread_unlock(sc->cotd_rx, 0); |
| 424 | cothread_delete(&sc->cotd_rx); |
| 425 | } |
| 426 | |
| 427 | for (i = 0; i < NETFIFOSIZE; i++) { |
| 428 | if (sc->sc_rxfifo && sc->sc_rxfifo->array[i]) { |
| 429 | m_freem(sc->sc_rxfifo->array[i]); |
| 430 | sc->sc_rxfifo->array[i] = NULL; |
| 431 | } |
| 432 | if (sc->sc_txfifo && sc->sc_txfifo->array[i]) { |
| 433 | m_freem(sc->sc_txfifo->array[i]); |
| 434 | sc->sc_txfifo->array[i] = NULL; |
| 435 | } |
| 436 | if (sc->sc_txfifo_done && sc->sc_txfifo_done->array[i]) { |
| 437 | m_freem(sc->sc_txfifo_done->array[i]); |
| 438 | sc->sc_txfifo_done->array[i] = NULL; |
| 439 | } |
| 440 | } |
| 441 | |
| 442 | if (sc->sc_txfifo) { |
| 443 | kfree(sc->sc_txfifo, M_DEVBUF); |
| 444 | sc->sc_txfifo = NULL; |
| 445 | } |
| 446 | |
| 447 | if (sc->sc_txfifo_done) { |
| 448 | kfree(sc->sc_txfifo_done, M_DEVBUF); |
| 449 | sc->sc_txfifo_done = NULL; |
| 450 | } |
| 451 | |
| 452 | if (sc->sc_rxfifo) { |
| 453 | kfree(sc->sc_rxfifo, M_DEVBUF); |
| 454 | sc->sc_rxfifo = NULL; |
| 455 | } |
| 456 | } |
| 457 | |
| 458 | |
| 459 | return 0; |
| 460 | } |
| 461 | |
| 462 | /* |
| 463 | * vke_rx_intr() is the interrupt function for the receive cothread. |
| 464 | */ |
| 465 | static void |
| 466 | vke_rx_intr(cothread_t cotd) |
| 467 | { |
| 468 | struct mbuf *m; |
| 469 | struct mbuf *nm; |
| 470 | struct vke_softc *sc = cotd->arg; |
| 471 | struct ifnet *ifp = &sc->arpcom.ac_if; |
| 472 | static int count = 0; |
| 473 | |
| 474 | ifnet_serialize_all(ifp); |
| 475 | cothread_lock(cotd, 0); |
| 476 | |
| 477 | if (sc->cotd_rx_exit != VKE_COTD_RUN) { |
| 478 | cothread_unlock(cotd, 0); |
| 479 | ifnet_deserialize_all(ifp); |
| 480 | return; |
| 481 | } |
| 482 | |
| 483 | while ((m = vke_rxfifo_sniff(sc)) != NULL) { |
| 484 | nm = m_getcl(MB_DONTWAIT, MT_DATA, M_PKTHDR); |
| 485 | if (nm) { |
| 486 | vke_rxfifo_dequeue(sc, nm); |
| 487 | ifp->if_input(ifp, m); |
| 488 | if (count++ == VKE_CHUNK) { |
| 489 | cothread_signal(cotd); |
| 490 | count = 0; |
| 491 | } |
| 492 | } else { |
| 493 | vke_rxfifo_dequeue(sc, m); |
| 494 | } |
| 495 | } |
| 496 | |
| 497 | if (count) |
| 498 | cothread_signal(cotd); |
| 499 | |
| 500 | cothread_unlock(cotd, 0); |
| 501 | ifnet_deserialize_all(ifp); |
| 502 | } |
| 503 | |
| 504 | /* |
| 505 | * vke_tx_intr() is the interrupt function for the transmit cothread. |
| 506 | * Calls vke_start() to handle processing transmit mbufs. |
| 507 | */ |
| 508 | static void |
| 509 | vke_tx_intr(cothread_t cotd) |
| 510 | { |
| 511 | struct vke_softc *sc = cotd->arg; |
| 512 | struct ifnet *ifp = &sc->arpcom.ac_if; |
| 513 | struct mbuf *m; |
| 514 | |
| 515 | ifnet_serialize_all(ifp); |
| 516 | cothread_lock(cotd, 0); |
| 517 | |
| 518 | if (sc->cotd_tx_exit != VKE_COTD_RUN) { |
| 519 | cothread_unlock(cotd, 0); |
| 520 | ifnet_deserialize_all(ifp); |
| 521 | return; |
| 522 | } |
| 523 | |
| 524 | if ((ifp->if_flags & IFF_RUNNING) == 0) |
| 525 | ifp->if_start(ifp); |
| 526 | |
| 527 | /* Free TX mbufs that have been processed */ |
| 528 | while ((m = vke_txfifo_done_dequeue(sc, NULL)) != NULL) { |
| 529 | m_freem(m); |
| 530 | } |
| 531 | |
| 532 | cothread_unlock(cotd, 0); |
| 533 | |
| 534 | ifnet_deserialize_all(ifp); |
| 535 | } |
| 536 | |
| 537 | /* |
| 538 | * vke_rx_thread() is the body of the receive cothread. |
| 539 | */ |
| 540 | static void |
| 541 | vke_rx_thread(cothread_t cotd) |
| 542 | { |
| 543 | struct mbuf *m; |
| 544 | struct vke_softc *sc = cotd->arg; |
| 545 | struct ifnet *ifp = &sc->arpcom.ac_if; |
| 546 | int count; |
| 547 | fifo_t fifo = sc->sc_rxfifo; |
| 548 | fd_set fdset; |
| 549 | struct timeval tv; |
| 550 | |
| 551 | /* Select timeout cannot be infinite since we need to check for |
| 552 | * the exit flag sc->cotd_rx_exit. |
| 553 | */ |
| 554 | tv.tv_sec = 0; |
| 555 | tv.tv_usec = 500000; |
| 556 | |
| 557 | FD_ZERO(&fdset); |
| 558 | |
| 559 | cothread_lock(cotd, 1); |
| 560 | |
| 561 | for (;;) { |
| 562 | int n; |
| 563 | count = 0; |
| 564 | while (sc->cotd_rx_exit == VKE_COTD_RUN) { |
| 565 | /* Wait for the RX FIFO to drain */ |
| 566 | while (NETFIFOINDEX(fifo->windex + 1) == |
| 567 | NETFIFOINDEX(fifo->rindex)) { |
| 568 | usleep(20000); |
| 569 | } |
| 570 | |
| 571 | if ((m = fifo->array[NETFIFOINDEX(fifo->windex)]) != |
| 572 | NULL) { |
| 573 | cothread_unlock(cotd, 1); |
| 574 | n = read(sc->sc_fd, mtod(m, void *), MCLBYTES); |
| 575 | cothread_lock(cotd, 1); |
| 576 | if (n <= 0) |
| 577 | break; |
| 578 | ifp->if_ipackets++; |
| 579 | m->m_pkthdr.rcvif = ifp; |
| 580 | m->m_pkthdr.len = m->m_len = n; |
| 581 | ++fifo->windex; |
| 582 | cpu_sfence(); |
| 583 | if (count++ == VKE_CHUNK) { |
| 584 | cothread_intr(cotd); |
| 585 | count = 0; |
| 586 | } |
| 587 | } |
| 588 | } |
| 589 | |
| 590 | if (count) { |
| 591 | cothread_intr(cotd); |
| 592 | } |
| 593 | |
| 594 | if (sc->cotd_rx_exit != VKE_COTD_RUN) |
| 595 | break; |
| 596 | |
| 597 | cothread_unlock(cotd, 1); |
| 598 | |
| 599 | /* Set up data for select() call */ |
| 600 | FD_SET(sc->sc_fd, &fdset); |
| 601 | |
| 602 | if (select(sc->sc_fd + 1, &fdset, NULL, NULL, &tv) == -1) |
| 603 | kprintf(VKE_DEVNAME "%d: select failed for TAP device\n", sc->sc_unit); |
| 604 | |
| 605 | cothread_lock(cotd, 1); |
| 606 | } |
| 607 | |
| 608 | sc->cotd_rx_exit = VKE_COTD_DEAD; |
| 609 | cothread_unlock(cotd, 1); |
| 610 | } |
| 611 | |
| 612 | /* |
| 613 | * vke_tx_thread() is the body of the transmit cothread. |
| 614 | */ |
| 615 | static void |
| 616 | vke_tx_thread(cothread_t cotd) |
| 617 | { |
| 618 | struct mbuf *m; |
| 619 | struct vke_softc *sc = cotd->arg; |
| 620 | struct ifnet *ifp = &sc->arpcom.ac_if; |
| 621 | int count = 0; |
| 622 | |
| 623 | cothread_lock(cotd, 1); |
| 624 | |
| 625 | while (sc->cotd_tx_exit == VKE_COTD_RUN) { |
| 626 | /* Write outgoing packets to the TAP interface */ |
| 627 | while ((m = vke_txfifo_dequeue(sc)) != NULL) { |
| 628 | if (m->m_pkthdr.len <= MCLBYTES) { |
| 629 | m_copydata(m, 0, m->m_pkthdr.len, sc->sc_txbuf); |
| 630 | sc->sc_txbuf_len = m->m_pkthdr.len; |
| 631 | cothread_unlock(cotd, 1); |
| 632 | |
| 633 | if (write(sc->sc_fd, sc->sc_txbuf, sc->sc_txbuf_len) < 0) { |
| 634 | cothread_lock(cotd, 1); |
| 635 | ifp->if_oerrors++; |
| 636 | } else { |
| 637 | cothread_lock(cotd, 1); |
| 638 | vke_txfifo_done_enqueue(sc, m); |
| 639 | ifp->if_opackets++; |
| 640 | if (count++ == VKE_CHUNK) { |
| 641 | cothread_intr(cotd); |
| 642 | count = 0; |
| 643 | } |
| 644 | } |
| 645 | } |
| 646 | } |
| 647 | |
| 648 | if (count) { |
| 649 | cothread_intr(cotd); |
| 650 | } |
| 651 | |
| 652 | cothread_wait(cotd); /* interlocks cothread lock */ |
| 653 | } |
| 654 | /* NOT REACHED */ |
| 655 | sc->cotd_tx_exit = VKE_COTD_DEAD; |
| 656 | cothread_unlock(cotd, 1); |
| 657 | } |
| 658 | |
| 659 | static int |
| 660 | vke_attach(const struct vknetif_info *info, int unit) |
| 661 | { |
| 662 | struct vke_softc *sc; |
| 663 | struct ifnet *ifp; |
| 664 | struct tapinfo tapinfo; |
| 665 | uint8_t enaddr[ETHER_ADDR_LEN]; |
| 666 | int fd; |
| 667 | |
| 668 | KKASSERT(info->tap_fd >= 0); |
| 669 | fd = info->tap_fd; |
| 670 | |
| 671 | /* |
| 672 | * This is only a TAP device if tap_unit is non-zero. If |
| 673 | * connecting to a virtual socket we generate a unique MAC. |
| 674 | */ |
| 675 | if (info->tap_unit >= 0) { |
| 676 | if (ioctl(fd, TAPGIFINFO, &tapinfo) < 0) { |
| 677 | kprintf(VKE_DEVNAME "%d: ioctl(TAPGIFINFO) " |
| 678 | "failed: %s\n", unit, strerror(errno)); |
| 679 | return ENXIO; |
| 680 | } |
| 681 | |
| 682 | if (ioctl(fd, SIOCGIFADDR, enaddr) < 0) { |
| 683 | kprintf(VKE_DEVNAME "%d: ioctl(SIOCGIFADDR) " |
| 684 | "failed: %s\n", unit, strerror(errno)); |
| 685 | return ENXIO; |
| 686 | } |
| 687 | } else { |
| 688 | int fd = open("/dev/urandom", O_RDONLY); |
| 689 | if (fd >= 0) { |
| 690 | read(fd, enaddr + 2, 4); |
| 691 | close(fd); |
| 692 | } |
| 693 | enaddr[4] = (int)getpid() >> 8; |
| 694 | enaddr[5] = (int)getpid() & 255; |
| 695 | |
| 696 | } |
| 697 | enaddr[1] += 1; |
| 698 | |
| 699 | sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO); |
| 700 | |
| 701 | sc->sc_txbuf = kmalloc(MCLBYTES, M_DEVBUF, M_WAITOK); |
| 702 | sc->sc_fd = fd; |
| 703 | sc->sc_unit = unit; |
| 704 | sc->sc_tap_unit = info->tap_unit; |
| 705 | sc->sc_addr = info->netif_addr; |
| 706 | sc->sc_mask = info->netif_mask; |
| 707 | |
| 708 | ifp = &sc->arpcom.ac_if; |
| 709 | if_initname(ifp, VKE_DEVNAME, sc->sc_unit); |
| 710 | |
| 711 | /* NB: after if_initname() */ |
| 712 | sysctl_ctx_init(&sc->sc_sysctl_ctx); |
| 713 | sc->sc_sysctl_tree = SYSCTL_ADD_NODE(&sc->sc_sysctl_ctx, |
| 714 | SYSCTL_STATIC_CHILDREN(_hw), |
| 715 | OID_AUTO, ifp->if_xname, |
| 716 | CTLFLAG_RD, 0, ""); |
| 717 | if (sc->sc_sysctl_tree == NULL) { |
| 718 | kprintf(VKE_DEVNAME "%d: can't add sysctl node\n", unit); |
| 719 | } else { |
| 720 | SYSCTL_ADD_INT(&sc->sc_sysctl_ctx, |
| 721 | SYSCTL_CHILDREN(sc->sc_sysctl_tree), |
| 722 | OID_AUTO, "tap_unit", |
| 723 | CTLFLAG_RD, &sc->sc_tap_unit, 0, |
| 724 | "Backend tap(4) unit"); |
| 725 | } |
| 726 | |
| 727 | ifp->if_softc = sc; |
| 728 | ifp->if_ioctl = vke_ioctl; |
| 729 | ifp->if_start = vke_start; |
| 730 | ifp->if_init = vke_init; |
| 731 | ifp->if_mtu = tapinfo.mtu; |
| 732 | ifp->if_baudrate = tapinfo.baudrate; |
| 733 | ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; |
| 734 | ifq_set_maxlen(&ifp->if_snd, IFQ_MAXLEN); |
| 735 | ifq_set_ready(&ifp->if_snd); |
| 736 | |
| 737 | /* TODO: if_media */ |
| 738 | |
| 739 | ether_ifattach(ifp, enaddr, NULL); |
| 740 | |
| 741 | if (bootverbose && sc->sc_addr != 0) { |
| 742 | if_printf(ifp, "pre-configured " |
| 743 | "address 0x%08x, netmask 0x%08x\n", |
| 744 | ntohl(sc->sc_addr), ntohl(sc->sc_mask)); |
| 745 | } |
| 746 | |
| 747 | return 0; |
| 748 | } |
| 749 | |
| 750 | static int |
| 751 | vke_init_addr(struct ifnet *ifp, in_addr_t addr, in_addr_t mask) |
| 752 | { |
| 753 | struct ifaliasreq ifra; |
| 754 | struct sockaddr_in *sin; |
| 755 | int ret; |
| 756 | |
| 757 | ASSERT_SERIALIZED(ifp->if_serializer); |
| 758 | |
| 759 | if (bootverbose) { |
| 760 | if_printf(ifp, "add pre-configured " |
| 761 | "address 0x%08x, netmask 0x%08x\n", |
| 762 | ntohl(addr), ntohl(mask)); |
| 763 | } |
| 764 | |
| 765 | bzero(&ifra, sizeof(ifra)); |
| 766 | |
| 767 | /* NB: no need to set ifaliasreq.ifra_name */ |
| 768 | |
| 769 | sin = (struct sockaddr_in *)&ifra.ifra_addr; |
| 770 | sin->sin_family = AF_INET; |
| 771 | sin->sin_len = sizeof(*sin); |
| 772 | sin->sin_addr.s_addr = addr; |
| 773 | |
| 774 | if (mask != 0) { |
| 775 | sin = (struct sockaddr_in *)&ifra.ifra_mask; |
| 776 | sin->sin_len = sizeof(*sin); |
| 777 | sin->sin_addr.s_addr = mask; |
| 778 | } |
| 779 | |
| 780 | /* |
| 781 | * Temporarily release serializer, in_control() will hold |
| 782 | * it again before calling ifnet.if_ioctl(). |
| 783 | */ |
| 784 | ifnet_deserialize_all(ifp); |
| 785 | ret = in_control(NULL, SIOCAIFADDR, (caddr_t)&ifra, ifp, NULL); |
| 786 | ifnet_serialize_all(ifp); |
| 787 | |
| 788 | return ret; |
| 789 | } |