1 /* $NetBSD: if_virt.c,v 1.22 2010/12/01 15:13:24 pooka Exp $ */ 2 3 /* 4 * Copyright (c) 2008 Antti Kantee. All Rights Reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS 16 * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 17 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 18 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 21 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28 #include <sys/cdefs.h> 29 __KERNEL_RCSID(0, "$NetBSD: if_virt.c,v 1.22 2010/12/01 15:13:24 pooka Exp $"); 30 31 #include <sys/param.h> 32 #include <sys/condvar.h> 33 #include <sys/fcntl.h> 34 #include <sys/kernel.h> 35 #include <sys/kmem.h> 36 #include <sys/kthread.h> 37 #include <sys/mutex.h> 38 #include <sys/poll.h> 39 #include <sys/sockio.h> 40 #include <sys/socketvar.h> 41 42 #include <net/bpf.h> 43 #include <net/if.h> 44 #include <net/if_ether.h> 45 #include <net/if_tap.h> 46 47 #include <netinet/in.h> 48 #include <netinet/in_var.h> 49 50 #include <rump/rump.h> 51 #include <rump/rumpuser.h> 52 53 #include "rump_private.h" 54 #include "rump_net_private.h" 55 56 /* 57 * Virtual interface for userspace purposes. Uses tap(4) to 58 * interface with the kernel and just simply shovels data 59 * to/from /dev/tap. 60 */ 61 62 #define VIRTIF_BASE "virt" 63 64 static int virtif_init(struct ifnet *); 65 static int virtif_ioctl(struct ifnet *, u_long, void *); 66 static void virtif_start(struct ifnet *); 67 static void virtif_stop(struct ifnet *, int); 68 69 struct virtif_sc { 70 struct ethercom sc_ec; 71 int sc_tapfd; 72 bool sc_dying; 73 struct lwp *sc_l_snd, *sc_l_rcv; 74 kmutex_t sc_mtx; 75 kcondvar_t sc_cv; 76 }; 77 78 static void virtif_receiver(void *); 79 static void virtif_sender(void *); 80 static int virtif_clone(struct if_clone *, int); 81 static int virtif_unclone(struct ifnet *); 82 83 struct if_clone virtif_cloner = 84 IF_CLONE_INITIALIZER(VIRTIF_BASE, virtif_clone, virtif_unclone); 85 86 int 87 rump_virtif_create(int num) 88 { 89 struct virtif_sc *sc; 90 struct ifnet *ifp; 91 uint8_t enaddr[ETHER_ADDR_LEN] = { 0xb2, 0x0a, 0x00, 0x0b, 0x0e, 0x01 }; 92 char tapdev[16]; 93 int fd, error = 0; 94 95 if (num >= 0x100) 96 return E2BIG; 97 98 snprintf(tapdev, sizeof(tapdev), "/dev/tap%d", num); 99 fd = rumpuser_open(tapdev, O_RDWR, &error); 100 if (fd == -1) { 101 printf("virtif_create: can't open /dev/tap%d: %d\n", 102 num, error); 103 return error; 104 } 105 enaddr[2] = arc4random() & 0xff; 106 enaddr[5] = num; 107 108 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP); 109 sc->sc_dying = false; 110 sc->sc_tapfd = fd; 111 112 mutex_init(&sc->sc_mtx, MUTEX_DEFAULT, IPL_NONE); 113 cv_init(&sc->sc_cv, "virtsnd"); 114 ifp = &sc->sc_ec.ec_if; 115 sprintf(ifp->if_xname, "%s%d", VIRTIF_BASE, num); 116 ifp->if_softc = sc; 117 118 if (rump_threads) { 119 if ((error = kthread_create(PRI_NONE, KTHREAD_JOINABLE, NULL, 120 virtif_receiver, ifp, &sc->sc_l_rcv, "virtifr")) != 0) 121 goto out; 122 123 if ((error = kthread_create(PRI_NONE, 124 KTHREAD_JOINABLE | KTHREAD_MPSAFE, NULL, 125 virtif_sender, ifp, &sc->sc_l_snd, "virtifs")) != 0) 126 goto out; 127 } else { 128 printf("WARNING: threads not enabled, receive NOT working\n"); 129 } 130 131 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 132 ifp->if_init = virtif_init; 133 ifp->if_ioctl = virtif_ioctl; 134 ifp->if_start = virtif_start; 135 ifp->if_stop = virtif_stop; 136 IFQ_SET_READY(&ifp->if_snd); 137 138 if_attach(ifp); 139 ether_ifattach(ifp, enaddr); 140 141 out: 142 if (error) { 143 virtif_unclone(ifp); 144 } 145 146 return error; 147 } 148 149 static int 150 virtif_clone(struct if_clone *ifc, int unit) 151 { 152 153 return rump_virtif_create(unit); 154 } 155 156 static int 157 virtif_unclone(struct ifnet *ifp) 158 { 159 struct virtif_sc *sc = ifp->if_softc; 160 161 mutex_enter(&sc->sc_mtx); 162 if (sc->sc_dying) { 163 mutex_exit(&sc->sc_mtx); 164 return EINPROGRESS; 165 } 166 sc->sc_dying = true; 167 cv_broadcast(&sc->sc_cv); 168 mutex_exit(&sc->sc_mtx); 169 170 virtif_stop(ifp, 1); 171 if_down(ifp); 172 173 if (sc->sc_l_snd) { 174 kthread_join(sc->sc_l_snd); 175 sc->sc_l_snd = NULL; 176 } 177 if (sc->sc_l_rcv) { 178 kthread_join(sc->sc_l_rcv); 179 sc->sc_l_rcv = NULL; 180 } 181 182 rumpuser_close(sc->sc_tapfd, NULL); 183 184 mutex_destroy(&sc->sc_mtx); 185 cv_destroy(&sc->sc_cv); 186 kmem_free(sc, sizeof(*sc)); 187 188 ether_ifdetach(ifp); 189 if_detach(ifp); 190 191 return 0; 192 } 193 194 static int 195 virtif_init(struct ifnet *ifp) 196 { 197 struct virtif_sc *sc = ifp->if_softc; 198 199 ifp->if_flags |= IFF_RUNNING; 200 201 mutex_enter(&sc->sc_mtx); 202 cv_broadcast(&sc->sc_cv); 203 mutex_exit(&sc->sc_mtx); 204 205 return 0; 206 } 207 208 static int 209 virtif_ioctl(struct ifnet *ifp, u_long cmd, void *data) 210 { 211 int s, rv; 212 213 s = splnet(); 214 rv = ether_ioctl(ifp, cmd, data); 215 if (rv == ENETRESET) 216 rv = 0; 217 splx(s); 218 219 return rv; 220 } 221 222 /* just send everything in-context */ 223 static void 224 virtif_start(struct ifnet *ifp) 225 { 226 struct virtif_sc *sc = ifp->if_softc; 227 228 mutex_enter(&sc->sc_mtx); 229 ifp->if_flags |= IFF_OACTIVE; 230 cv_broadcast(&sc->sc_cv); 231 mutex_exit(&sc->sc_mtx); 232 } 233 234 static void 235 virtif_stop(struct ifnet *ifp, int disable) 236 { 237 struct virtif_sc *sc = ifp->if_softc; 238 239 ifp->if_flags &= ~IFF_RUNNING; 240 241 mutex_enter(&sc->sc_mtx); 242 cv_broadcast(&sc->sc_cv); 243 mutex_exit(&sc->sc_mtx); 244 } 245 246 #define POLLTIMO_MS 1 247 static void 248 virtif_receiver(void *arg) 249 { 250 struct ifnet *ifp = arg; 251 struct virtif_sc *sc = ifp->if_softc; 252 struct mbuf *m; 253 size_t plen = ETHER_MAX_LEN_JUMBO+1; 254 struct pollfd pfd; 255 ssize_t n; 256 int error, rv; 257 258 pfd.fd = sc->sc_tapfd; 259 pfd.events = POLLIN; 260 261 for (;;) { 262 m = m_gethdr(M_WAIT, MT_DATA); 263 MEXTMALLOC(m, plen, M_WAIT); 264 265 again: 266 /* poll, but periodically check if we should die */ 267 rv = rumpuser_poll(&pfd, 1, POLLTIMO_MS, &error); 268 if (sc->sc_dying) { 269 m_freem(m); 270 break; 271 } 272 if (rv == 0) 273 goto again; 274 275 n = rumpuser_read(sc->sc_tapfd, mtod(m, void *), plen, &error); 276 KASSERT(n < ETHER_MAX_LEN_JUMBO); 277 if (__predict_false(n < 0)) { 278 if (n == -1 && error == EAGAIN) { 279 goto again; 280 } 281 282 printf("%s: read from /dev/tap failed. host if down?\n", 283 ifp->if_xname); 284 mutex_enter(&sc->sc_mtx); 285 /* could check if need go, done soon anyway */ 286 cv_timedwait(&sc->sc_cv, &sc->sc_mtx, hz); 287 mutex_exit(&sc->sc_mtx); 288 goto again; 289 } 290 291 /* tap sometimes returns EOF. don't sweat it and plow on */ 292 if (__predict_false(n == 0)) 293 goto again; 294 295 /* discard if we're not up */ 296 if ((ifp->if_flags & IFF_RUNNING) == 0) 297 goto again; 298 299 m->m_len = m->m_pkthdr.len = n; 300 m->m_pkthdr.rcvif = ifp; 301 bpf_mtap(ifp, m); 302 ether_input(ifp, m); 303 } 304 305 kthread_exit(0); 306 } 307 308 /* lazy bum stetson-harrison magic value */ 309 #define LB_SH 32 310 static void 311 virtif_sender(void *arg) 312 { 313 struct ifnet *ifp = arg; 314 struct virtif_sc *sc = ifp->if_softc; 315 struct mbuf *m, *m0; 316 struct rumpuser_iovec io[LB_SH]; 317 int i, error; 318 319 mutex_enter(&sc->sc_mtx); 320 KERNEL_LOCK(1, NULL); 321 while (!sc->sc_dying) { 322 if (!ifp->if_flags & IFF_RUNNING) { 323 cv_wait(&sc->sc_cv, &sc->sc_mtx); 324 continue; 325 } 326 IF_DEQUEUE(&ifp->if_snd, m0); 327 if (!m0) { 328 ifp->if_flags &= ~IFF_OACTIVE; 329 cv_wait(&sc->sc_cv, &sc->sc_mtx); 330 continue; 331 } 332 mutex_exit(&sc->sc_mtx); 333 334 m = m0; 335 for (i = 0; i < LB_SH && m; i++) { 336 io[i].iov_base = mtod(m, void *); 337 io[i].iov_len = m->m_len; 338 m = m->m_next; 339 } 340 if (i == LB_SH) 341 panic("lazy bum"); 342 bpf_mtap(ifp, m0); 343 KERNEL_UNLOCK_LAST(curlwp); 344 345 rumpuser_writev(sc->sc_tapfd, io, i, &error); 346 347 KERNEL_LOCK(1, NULL); 348 m_freem(m0); 349 mutex_enter(&sc->sc_mtx); 350 } 351 KERNEL_UNLOCK_LAST(curlwp); 352 353 mutex_exit(&sc->sc_mtx); 354 355 kthread_exit(0); 356 } 357 358 /* 359 * dummyif is a nada-interface. 360 * As it requires nothing external, it can be used for testing 361 * interface configuration. 362 */ 363 static int dummyif_init(struct ifnet *); 364 static void dummyif_start(struct ifnet *); 365 366 void 367 rump_dummyif_create() 368 { 369 struct ifnet *ifp; 370 struct ethercom *ec; 371 uint8_t enaddr[ETHER_ADDR_LEN] = { 0xb2, 0x0a, 0x00, 0x0b, 0x0e, 0x01 }; 372 373 enaddr[2] = arc4random() & 0xff; 374 enaddr[5] = arc4random() & 0xff; 375 376 ec = kmem_zalloc(sizeof(*ec), KM_SLEEP); 377 378 ifp = &ec->ec_if; 379 strlcpy(ifp->if_xname, "dummy0", sizeof(ifp->if_xname)); 380 ifp->if_softc = ifp; 381 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 382 ifp->if_init = dummyif_init; 383 ifp->if_ioctl = virtif_ioctl; 384 ifp->if_start = dummyif_start; 385 386 if_attach(ifp); 387 ether_ifattach(ifp, enaddr); 388 } 389 390 static int 391 dummyif_init(struct ifnet *ifp) 392 { 393 394 ifp->if_flags |= IFF_RUNNING; 395 return 0; 396 } 397 398 static void 399 dummyif_start(struct ifnet *ifp) 400 { 401 402 } 403