if_virt.c revision 1.35 1 /* $NetBSD: if_virt.c,v 1.35 2013/07/03 20:17:07 pooka Exp $ */
2
3 /*
4 * Copyright (c) 2008, 2013 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 1.35 2013/07/03 20:17:07 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 #include <sys/cprng.h>
42
43 #include <net/bpf.h>
44 #include <net/if.h>
45 #include <net/if_ether.h>
46 #include <net/if_tap.h>
47
48 #include <netinet/in.h>
49 #include <netinet/in_var.h>
50
51 #include <rump/rump.h>
52
53 #include "rump_private.h"
54 #include "rump_net_private.h"
55
56 #include "rumpcomp_user.h"
57
58 /*
59 * Virtual interface. Uses hypercalls to shovel packets back
60 * and forth. The exact method for shoveling depends on the
61 * hypercall implementation.
62 */
63
64 #ifndef VIRTIF_BASE
65 #define VIRTIF_BASE "virt"
66 #endif
67
68 static int virtif_init(struct ifnet *);
69 static int virtif_ioctl(struct ifnet *, u_long, void *);
70 static void virtif_start(struct ifnet *);
71 static void virtif_stop(struct ifnet *, int);
72
73 struct virtif_sc {
74 struct ethercom sc_ec;
75 struct virtif_user *sc_viu;
76 bool sc_dying;
77 struct lwp *sc_l_snd, *sc_l_rcv;
78 kmutex_t sc_mtx;
79 kcondvar_t sc_cv;
80 };
81
82 static void virtif_receiver(void *);
83 static void virtif_sender(void *);
84 static int virtif_clone(struct if_clone *, int);
85 static int virtif_unclone(struct ifnet *);
86
87 struct if_clone virtif_cloner =
88 IF_CLONE_INITIALIZER(VIRTIF_BASE, virtif_clone, virtif_unclone);
89
90 static int
91 virtif_clone(struct if_clone *ifc, int num)
92 {
93 struct virtif_sc *sc;
94 struct virtif_user *viu;
95 struct ifnet *ifp;
96 uint8_t enaddr[ETHER_ADDR_LEN] = { 0xb2, 0x0a, 0x00, 0x0b, 0x0e, 0x01 };
97 int error = 0;
98
99 if (num >= 0x100)
100 return E2BIG;
101
102 if ((error = rumpcomp_virtif_create(num, &viu)) != 0)
103 return error;
104
105 enaddr[2] = cprng_fast32() & 0xff;
106 enaddr[5] = num;
107
108 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP);
109 sc->sc_dying = false;
110 sc->sc_viu = viu;
111
112 mutex_init(&sc->sc_mtx, MUTEX_DEFAULT, IPL_NONE);
113 cv_init(&sc->sc_cv, VIRTIF_BASE "snd");
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_MUSTJOIN, NULL,
120 virtif_receiver, ifp, &sc->sc_l_rcv,
121 VIRTIF_BASE "ifr")) != 0)
122 goto out;
123
124 if ((error = kthread_create(PRI_NONE,
125 KTHREAD_MUSTJOIN | KTHREAD_MPSAFE, NULL,
126 virtif_sender, ifp, &sc->sc_l_snd, VIRTIF_BASE "ifs")) != 0)
127 goto out;
128 } else {
129 printf("WARNING: threads not enabled, receive NOT working\n");
130 }
131
132 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
133 ifp->if_init = virtif_init;
134 ifp->if_ioctl = virtif_ioctl;
135 ifp->if_start = virtif_start;
136 ifp->if_stop = virtif_stop;
137 IFQ_SET_READY(&ifp->if_snd);
138
139 if_attach(ifp);
140 ether_ifattach(ifp, enaddr);
141
142 out:
143 if (error) {
144 virtif_unclone(ifp);
145 }
146
147 return error;
148 }
149
150 static int
151 virtif_unclone(struct ifnet *ifp)
152 {
153 struct virtif_sc *sc = ifp->if_softc;
154
155 mutex_enter(&sc->sc_mtx);
156 if (sc->sc_dying) {
157 mutex_exit(&sc->sc_mtx);
158 return EINPROGRESS;
159 }
160 sc->sc_dying = true;
161 cv_broadcast(&sc->sc_cv);
162 mutex_exit(&sc->sc_mtx);
163
164 rumpcomp_virtif_dying(sc->sc_viu);
165
166 virtif_stop(ifp, 1);
167 if_down(ifp);
168
169 if (sc->sc_l_snd) {
170 kthread_join(sc->sc_l_snd);
171 sc->sc_l_snd = NULL;
172 }
173 if (sc->sc_l_rcv) {
174 kthread_join(sc->sc_l_rcv);
175 sc->sc_l_rcv = NULL;
176 }
177
178 rumpcomp_virtif_destroy(sc->sc_viu);
179
180 mutex_destroy(&sc->sc_mtx);
181 cv_destroy(&sc->sc_cv);
182 kmem_free(sc, sizeof(*sc));
183
184 ether_ifdetach(ifp);
185 if_detach(ifp);
186
187 return 0;
188 }
189
190 static int
191 virtif_init(struct ifnet *ifp)
192 {
193 struct virtif_sc *sc = ifp->if_softc;
194
195 ifp->if_flags |= IFF_RUNNING;
196
197 mutex_enter(&sc->sc_mtx);
198 cv_broadcast(&sc->sc_cv);
199 mutex_exit(&sc->sc_mtx);
200
201 return 0;
202 }
203
204 static int
205 virtif_ioctl(struct ifnet *ifp, u_long cmd, void *data)
206 {
207 int s, rv;
208
209 s = splnet();
210 rv = ether_ioctl(ifp, cmd, data);
211 if (rv == ENETRESET)
212 rv = 0;
213 splx(s);
214
215 return rv;
216 }
217
218 static void
219 virtif_start(struct ifnet *ifp)
220 {
221 struct virtif_sc *sc = ifp->if_softc;
222
223 mutex_enter(&sc->sc_mtx);
224 ifp->if_flags |= IFF_OACTIVE;
225 cv_broadcast(&sc->sc_cv);
226 mutex_exit(&sc->sc_mtx);
227 }
228
229 static void
230 virtif_stop(struct ifnet *ifp, int disable)
231 {
232 struct virtif_sc *sc = ifp->if_softc;
233
234 ifp->if_flags &= ~IFF_RUNNING;
235
236 mutex_enter(&sc->sc_mtx);
237 cv_broadcast(&sc->sc_cv);
238 mutex_exit(&sc->sc_mtx);
239 }
240
241 #define POLLTIMO_MS 1
242 static void
243 virtif_receiver(void *arg)
244 {
245 struct ifnet *ifp = arg;
246 struct virtif_sc *sc = ifp->if_softc;
247 struct mbuf *m;
248 size_t plen = ETHER_MAX_LEN_JUMBO+1;
249 size_t n;
250 int error;
251
252 for (;;) {
253 m = m_gethdr(M_WAIT, MT_DATA);
254 MEXTMALLOC(m, plen, M_WAIT);
255
256 again:
257 if (sc->sc_dying) {
258 m_freem(m);
259 break;
260 }
261
262 error = rumpcomp_virtif_recv(sc->sc_viu,
263 mtod(m, void *), plen, &n);
264 if (error) {
265 printf("%s: read hypercall failed %d. host if down?\n",
266 ifp->if_xname, error);
267 mutex_enter(&sc->sc_mtx);
268 /* could check if need go, done soon anyway */
269 cv_timedwait(&sc->sc_cv, &sc->sc_mtx, hz);
270 mutex_exit(&sc->sc_mtx);
271 goto again;
272 }
273
274 /* tap sometimes returns EOF. don't sweat it and plow on */
275 if (__predict_false(n == 0))
276 goto again;
277
278 /* discard if we're not up */
279 if ((ifp->if_flags & IFF_RUNNING) == 0)
280 goto again;
281
282 m->m_len = m->m_pkthdr.len = n;
283 m->m_pkthdr.rcvif = ifp;
284 bpf_mtap(ifp, m);
285 ether_input(ifp, m);
286 }
287
288 kthread_exit(0);
289 }
290
291 /* lazy bum stetson-harrison magic value */
292 #define LB_SH 32
293 static void
294 virtif_sender(void *arg)
295 {
296 struct ifnet *ifp = arg;
297 struct virtif_sc *sc = ifp->if_softc;
298 struct mbuf *m, *m0;
299 struct iovec io[LB_SH];
300 int i;
301
302 mutex_enter(&sc->sc_mtx);
303 KERNEL_LOCK(1, NULL);
304 while (!sc->sc_dying) {
305 if (!(ifp->if_flags & IFF_RUNNING)) {
306 cv_wait(&sc->sc_cv, &sc->sc_mtx);
307 continue;
308 }
309 IF_DEQUEUE(&ifp->if_snd, m0);
310 if (!m0) {
311 ifp->if_flags &= ~IFF_OACTIVE;
312 cv_wait(&sc->sc_cv, &sc->sc_mtx);
313 continue;
314 }
315 mutex_exit(&sc->sc_mtx);
316
317 m = m0;
318 for (i = 0; i < LB_SH && m; i++) {
319 io[i].iov_base = mtod(m, void *);
320 io[i].iov_len = m->m_len;
321 m = m->m_next;
322 }
323 if (i == LB_SH)
324 panic("lazy bum");
325 bpf_mtap(ifp, m0);
326
327 rumpcomp_virtif_send(sc->sc_viu, io, i);
328
329 m_freem(m0);
330 mutex_enter(&sc->sc_mtx);
331 }
332 KERNEL_UNLOCK_LAST(curlwp);
333
334 mutex_exit(&sc->sc_mtx);
335
336 kthread_exit(0);
337 }
338