xref: /netbsd-src/sys/rump/net/lib/libvirtif/if_virt.c (revision eaac9e3d28859567590b7fef63a96c293f76dbf9)
1 /*	$NetBSD: if_virt.c,v 1.36 2013/07/04 11:46:51 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 1.36 2013/07/04 11:46:51 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 "if_virt.h"
57 #include "rumpcomp_user.h"
58 
59 /*
60  * Virtual interface.  Uses hypercalls to shovel packets back
61  * and forth.  The exact method for shoveling depends on the
62  * hypercall implementation.
63  */
64 
65 static int	virtif_init(struct ifnet *);
66 static int	virtif_ioctl(struct ifnet *, u_long, void *);
67 static void	virtif_start(struct ifnet *);
68 static void	virtif_stop(struct ifnet *, int);
69 
70 struct virtif_sc {
71 	struct ethercom sc_ec;
72 	struct virtif_user *sc_viu;
73 	bool sc_dying;
74 	struct lwp *sc_l_snd, *sc_l_rcv;
75 	kmutex_t sc_mtx;
76 	kcondvar_t sc_cv;
77 };
78 
79 static void virtif_receiver(void *);
80 static void virtif_sender(void *);
81 static int  virtif_clone(struct if_clone *, int);
82 static int  virtif_unclone(struct ifnet *);
83 
84 struct if_clone VIF_CLONER =
85     IF_CLONE_INITIALIZER(VIF_NAME, virtif_clone, virtif_unclone);
86 
87 static int
88 virtif_clone(struct if_clone *ifc, int num)
89 {
90 	struct virtif_sc *sc;
91 	struct virtif_user *viu;
92 	struct ifnet *ifp;
93 	uint8_t enaddr[ETHER_ADDR_LEN] = { 0xb2, 0x0a, 0x00, 0x0b, 0x0e, 0x01 };
94 	int error = 0;
95 
96 	if (num >= 0x100)
97 		return E2BIG;
98 
99 	if ((error = VIFHYPER_CREATE(num, &viu)) != 0)
100 		return error;
101 
102 	enaddr[2] = cprng_fast32() & 0xff;
103 	enaddr[5] = num;
104 
105 	sc = kmem_zalloc(sizeof(*sc), KM_SLEEP);
106 	sc->sc_dying = false;
107 	sc->sc_viu = viu;
108 
109 	mutex_init(&sc->sc_mtx, MUTEX_DEFAULT, IPL_NONE);
110 	cv_init(&sc->sc_cv, VIF_NAME "snd");
111 	ifp = &sc->sc_ec.ec_if;
112 	sprintf(ifp->if_xname, "%s%d", VIF_NAME, num);
113 	ifp->if_softc = sc;
114 
115 	if (rump_threads) {
116 		if ((error = kthread_create(PRI_NONE, KTHREAD_MUSTJOIN, NULL,
117 		    virtif_receiver, ifp, &sc->sc_l_rcv, VIF_NAME "ifr")) != 0)
118 			goto out;
119 
120 		if ((error = kthread_create(PRI_NONE,
121 		    KTHREAD_MUSTJOIN | KTHREAD_MPSAFE, NULL,
122 		    virtif_sender, ifp, &sc->sc_l_snd, VIF_NAME "ifs")) != 0)
123 			goto out;
124 	} else {
125 		printf("WARNING: threads not enabled, receive NOT working\n");
126 	}
127 
128 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
129 	ifp->if_init = virtif_init;
130 	ifp->if_ioctl = virtif_ioctl;
131 	ifp->if_start = virtif_start;
132 	ifp->if_stop = virtif_stop;
133 	IFQ_SET_READY(&ifp->if_snd);
134 
135 	if_attach(ifp);
136 	ether_ifattach(ifp, enaddr);
137 
138  out:
139 	if (error) {
140 		virtif_unclone(ifp);
141 	}
142 
143 	return error;
144 }
145 
146 static int
147 virtif_unclone(struct ifnet *ifp)
148 {
149 	struct virtif_sc *sc = ifp->if_softc;
150 
151 	mutex_enter(&sc->sc_mtx);
152 	if (sc->sc_dying) {
153 		mutex_exit(&sc->sc_mtx);
154 		return EINPROGRESS;
155 	}
156 	sc->sc_dying = true;
157 	cv_broadcast(&sc->sc_cv);
158 	mutex_exit(&sc->sc_mtx);
159 
160 	VIFHYPER_DYING(sc->sc_viu);
161 
162 	virtif_stop(ifp, 1);
163 	if_down(ifp);
164 
165 	if (sc->sc_l_snd) {
166 		kthread_join(sc->sc_l_snd);
167 		sc->sc_l_snd = NULL;
168 	}
169 	if (sc->sc_l_rcv) {
170 		kthread_join(sc->sc_l_rcv);
171 		sc->sc_l_rcv = NULL;
172 	}
173 
174 	VIFHYPER_DESTROY(sc->sc_viu);
175 
176 	mutex_destroy(&sc->sc_mtx);
177 	cv_destroy(&sc->sc_cv);
178 	kmem_free(sc, sizeof(*sc));
179 
180 	ether_ifdetach(ifp);
181 	if_detach(ifp);
182 
183 	return 0;
184 }
185 
186 static int
187 virtif_init(struct ifnet *ifp)
188 {
189 	struct virtif_sc *sc = ifp->if_softc;
190 
191 	ifp->if_flags |= IFF_RUNNING;
192 
193 	mutex_enter(&sc->sc_mtx);
194 	cv_broadcast(&sc->sc_cv);
195 	mutex_exit(&sc->sc_mtx);
196 
197 	return 0;
198 }
199 
200 static int
201 virtif_ioctl(struct ifnet *ifp, u_long cmd, void *data)
202 {
203 	int s, rv;
204 
205 	s = splnet();
206 	rv = ether_ioctl(ifp, cmd, data);
207 	if (rv == ENETRESET)
208 		rv = 0;
209 	splx(s);
210 
211 	return rv;
212 }
213 
214 static void
215 virtif_start(struct ifnet *ifp)
216 {
217 	struct virtif_sc *sc = ifp->if_softc;
218 
219 	mutex_enter(&sc->sc_mtx);
220 	ifp->if_flags |= IFF_OACTIVE;
221 	cv_broadcast(&sc->sc_cv);
222 	mutex_exit(&sc->sc_mtx);
223 }
224 
225 static void
226 virtif_stop(struct ifnet *ifp, int disable)
227 {
228 	struct virtif_sc *sc = ifp->if_softc;
229 
230 	ifp->if_flags &= ~IFF_RUNNING;
231 
232 	mutex_enter(&sc->sc_mtx);
233 	cv_broadcast(&sc->sc_cv);
234 	mutex_exit(&sc->sc_mtx);
235 }
236 
237 #define POLLTIMO_MS 1
238 static void
239 virtif_receiver(void *arg)
240 {
241 	struct ifnet *ifp = arg;
242 	struct virtif_sc *sc = ifp->if_softc;
243 	struct mbuf *m;
244 	size_t plen = ETHER_MAX_LEN_JUMBO+1;
245 	size_t n;
246 	int error;
247 
248 	for (;;) {
249 		m = m_gethdr(M_WAIT, MT_DATA);
250 		MEXTMALLOC(m, plen, M_WAIT);
251 
252  again:
253 		if (sc->sc_dying) {
254 			m_freem(m);
255 			break;
256 		}
257 
258 		error = VIFHYPER_RECV(sc->sc_viu,
259 		    mtod(m, void *), plen, &n);
260 		if (error) {
261 			printf("%s: read hypercall failed %d. host if down?\n",
262 			    ifp->if_xname, error);
263 			mutex_enter(&sc->sc_mtx);
264 			/* could check if need go, done soon anyway */
265 			cv_timedwait(&sc->sc_cv, &sc->sc_mtx, hz);
266 			mutex_exit(&sc->sc_mtx);
267 			goto again;
268 		}
269 
270 		/* tap sometimes returns EOF.  don't sweat it and plow on */
271 		if (__predict_false(n == 0))
272 			goto again;
273 
274 		/* discard if we're not up */
275 		if ((ifp->if_flags & IFF_RUNNING) == 0)
276 			goto again;
277 
278 		m->m_len = m->m_pkthdr.len = n;
279 		m->m_pkthdr.rcvif = ifp;
280 		bpf_mtap(ifp, m);
281 		ether_input(ifp, m);
282 	}
283 
284 	kthread_exit(0);
285 }
286 
287 /* lazy bum stetson-harrison magic value */
288 #define LB_SH 32
289 static void
290 virtif_sender(void *arg)
291 {
292 	struct ifnet *ifp = arg;
293 	struct virtif_sc *sc = ifp->if_softc;
294 	struct mbuf *m, *m0;
295 	struct iovec io[LB_SH];
296 	int i;
297 
298 	mutex_enter(&sc->sc_mtx);
299 	KERNEL_LOCK(1, NULL);
300 	while (!sc->sc_dying) {
301 		if (!(ifp->if_flags & IFF_RUNNING)) {
302 			cv_wait(&sc->sc_cv, &sc->sc_mtx);
303 			continue;
304 		}
305 		IF_DEQUEUE(&ifp->if_snd, m0);
306 		if (!m0) {
307 			ifp->if_flags &= ~IFF_OACTIVE;
308 			cv_wait(&sc->sc_cv, &sc->sc_mtx);
309 			continue;
310 		}
311 		mutex_exit(&sc->sc_mtx);
312 
313 		m = m0;
314 		for (i = 0; i < LB_SH && m; i++) {
315 			io[i].iov_base = mtod(m, void *);
316 			io[i].iov_len = m->m_len;
317 			m = m->m_next;
318 		}
319 		if (i == LB_SH)
320 			panic("lazy bum");
321 		bpf_mtap(ifp, m0);
322 
323 		VIFHYPER_SEND(sc->sc_viu, io, i);
324 
325 		m_freem(m0);
326 		mutex_enter(&sc->sc_mtx);
327 	}
328 	KERNEL_UNLOCK_LAST(curlwp);
329 
330 	mutex_exit(&sc->sc_mtx);
331 
332 	kthread_exit(0);
333 }
334