ng_eiface.c revision 141195 
1/*-
2 *
3 * Copyright (c) 1999-2001, Vitaly V Belekhov
4 * 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 unmodified, this list of conditions, and the following
11 *    disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 *
28 * $FreeBSD: head/sys/netgraph/ng_eiface.c 141195 2005-02-03 12:50:10Z ru $
29 */
30
31#include <sys/param.h>
32#include <sys/systm.h>
33#include <sys/errno.h>
34#include <sys/kernel.h>
35#include <sys/lock.h>
36#include <sys/malloc.h>
37#include <sys/mbuf.h>
38#include <sys/mutex.h>
39#include <sys/errno.h>
40#include <sys/sockio.h>
41#include <sys/socket.h>
42#include <sys/syslog.h>
43
44#include <net/if.h>
45#include <net/if_dl.h>
46#include <net/if_types.h>
47#include <net/netisr.h>
48
49#include <netgraph/ng_message.h>
50#include <netgraph/netgraph.h>
51#include <netgraph/ng_parse.h>
52#include <netgraph/ng_eiface.h>
53
54#include <net/bpf.h>
55#include <net/ethernet.h>
56#include <net/if_arp.h>
57
58static const struct ng_cmdlist ng_eiface_cmdlist[] = {
59	{
60	  NGM_EIFACE_COOKIE,
61	  NGM_EIFACE_GET_IFNAME,
62	  "getifname",
63	  NULL,
64	  &ng_parse_string_type
65	},
66	{
67	  NGM_EIFACE_COOKIE,
68	  NGM_EIFACE_SET,
69	  "set",
70	  &ng_parse_enaddr_type,
71	  NULL
72	},
73	{ 0 }
74};
75
76/* Node private data */
77struct ng_eiface_private {
78	struct arpcom	arpcom;		/* per-interface network data */
79	struct ifnet	*ifp;		/* This interface */
80	int		unit;		/* Interface unit number */
81	node_p		node;		/* Our netgraph node */
82	hook_p		ether;		/* Hook for ethernet stream */
83};
84typedef struct ng_eiface_private *priv_p;
85
86/* Interface methods */
87static void	ng_eiface_init(void *xsc);
88static void	ng_eiface_start(struct ifnet *ifp);
89static int	ng_eiface_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data);
90#ifdef DEBUG
91static void	ng_eiface_print_ioctl(struct ifnet *ifp, int cmd, caddr_t data);
92#endif
93
94/* Netgraph methods */
95static ng_constructor_t	ng_eiface_constructor;
96static ng_rcvmsg_t	ng_eiface_rcvmsg;
97static ng_shutdown_t	ng_eiface_rmnode;
98static ng_newhook_t	ng_eiface_newhook;
99static ng_rcvdata_t	ng_eiface_rcvdata;
100static ng_connect_t	ng_eiface_connect;
101static ng_disconnect_t	ng_eiface_disconnect;
102
103/* Node type descriptor */
104static struct ng_type typestruct = {
105	.version =	NG_ABI_VERSION,
106	.name =		NG_EIFACE_NODE_TYPE,
107	.constructor =	ng_eiface_constructor,
108	.rcvmsg =	ng_eiface_rcvmsg,
109	.shutdown =	ng_eiface_rmnode,
110	.newhook =	ng_eiface_newhook,
111	.connect =	ng_eiface_connect,
112	.rcvdata =	ng_eiface_rcvdata,
113	.disconnect =	ng_eiface_disconnect,
114	.cmdlist =	ng_eiface_cmdlist
115};
116NETGRAPH_INIT(eiface, &typestruct);
117
118/* We keep a bitmap indicating which unit numbers are free.
119   One means the unit number is free, zero means it's taken. */
120static int	*ng_eiface_units = NULL;
121static int	ng_eiface_units_len = 0;
122static int	ng_units_in_use = 0;
123
124#define UNITS_BITSPERWORD	(sizeof(*ng_eiface_units) * NBBY)
125
126static struct mtx	ng_eiface_mtx;
127MTX_SYSINIT(ng_eiface, &ng_eiface_mtx, "ng_eiface", MTX_DEF);
128
129/************************************************************************
130			HELPER STUFF
131 ************************************************************************/
132/*
133 * Find the first free unit number for a new interface.
134 * Increase the size of the unit bitmap as necessary.
135 */
136static __inline int
137ng_eiface_get_unit(int *unit)
138{
139	int index, bit;
140
141	mtx_lock(&ng_eiface_mtx);
142	for (index = 0; index < ng_eiface_units_len
143	    && ng_eiface_units[index] == 0; index++);
144	if (index == ng_eiface_units_len) {		/* extend array */
145		int i, *newarray, newlen;
146
147		newlen = (2 * ng_eiface_units_len) + 4;
148		MALLOC(newarray, int *, newlen * sizeof(*ng_eiface_units),
149		    M_NETGRAPH, M_NOWAIT);
150		if (newarray == NULL) {
151			mtx_unlock(&ng_eiface_mtx);
152			return (ENOMEM);
153		}
154		bcopy(ng_eiface_units, newarray,
155		    ng_eiface_units_len * sizeof(*ng_eiface_units));
156		for (i = ng_eiface_units_len; i < newlen; i++)
157			newarray[i] = ~0;
158		if (ng_eiface_units != NULL)
159			FREE(ng_eiface_units, M_NETGRAPH);
160		ng_eiface_units = newarray;
161		ng_eiface_units_len = newlen;
162	}
163	bit = ffs(ng_eiface_units[index]) - 1;
164	KASSERT(bit >= 0 && bit <= UNITS_BITSPERWORD - 1,
165	    ("%s: word=%d bit=%d", __func__, ng_eiface_units[index], bit));
166	ng_eiface_units[index] &= ~(1 << bit);
167	*unit = (index * UNITS_BITSPERWORD) + bit;
168	ng_units_in_use++;
169	mtx_unlock(&ng_eiface_mtx);
170	return (0);
171}
172
173/*
174 * Free a no longer needed unit number.
175 */
176static __inline void
177ng_eiface_free_unit(int unit)
178{
179	int index, bit;
180
181	index = unit / UNITS_BITSPERWORD;
182	bit = unit % UNITS_BITSPERWORD;
183	mtx_lock(&ng_eiface_mtx);
184	KASSERT(index < ng_eiface_units_len,
185	    ("%s: unit=%d len=%d", __func__, unit, ng_eiface_units_len));
186	KASSERT((ng_eiface_units[index] & (1 << bit)) == 0,
187	    ("%s: unit=%d is free", __func__, unit));
188	ng_eiface_units[index] |= (1 << bit);
189	/*
190	 * XXX We could think about reducing the size of ng_eiface_units[]
191	 * XXX here if the last portion is all ones
192	 * XXX At least free it if no more units.
193	 * Needed if we are to eventually be able to unload.
194	 */
195	ng_units_in_use--;
196	if (ng_units_in_use == 0) { /* XXX make SMP safe */
197		FREE(ng_eiface_units, M_NETGRAPH);
198		ng_eiface_units_len = 0;
199		ng_eiface_units = NULL;
200	}
201	mtx_unlock(&ng_eiface_mtx);
202}
203
204/************************************************************************
205			INTERFACE STUFF
206 ************************************************************************/
207
208/*
209 * Process an ioctl for the virtual interface
210 */
211static int
212ng_eiface_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
213{
214	struct ifreq *const ifr = (struct ifreq *)data;
215	int s, error = 0;
216
217#ifdef DEBUG
218	ng_eiface_print_ioctl(ifp, command, data);
219#endif
220	s = splimp();
221	switch (command) {
222
223	/* These two are mostly handled at a higher layer */
224	case SIOCSIFADDR:
225		error = ether_ioctl(ifp, command, data);
226		break;
227	case SIOCGIFADDR:
228		break;
229
230	/* Set flags */
231	case SIOCSIFFLAGS:
232		/*
233		 * If the interface is marked up and stopped, then start it.
234		 * If it is marked down and running, then stop it.
235		 */
236		if (ifr->ifr_flags & IFF_UP) {
237			if (!(ifp->if_flags & IFF_RUNNING)) {
238				ifp->if_flags &= ~(IFF_OACTIVE);
239				ifp->if_flags |= IFF_RUNNING;
240			}
241		} else {
242			if (ifp->if_flags & IFF_RUNNING)
243				ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
244		}
245		break;
246
247	/* Set the interface MTU */
248	case SIOCSIFMTU:
249		if (ifr->ifr_mtu > NG_EIFACE_MTU_MAX ||
250		    ifr->ifr_mtu < NG_EIFACE_MTU_MIN)
251			error = EINVAL;
252		else
253			ifp->if_mtu = ifr->ifr_mtu;
254		break;
255
256	/* Stuff that's not supported */
257	case SIOCADDMULTI:
258	case SIOCDELMULTI:
259		error = 0;
260		break;
261	case SIOCSIFPHYS:
262		error = EOPNOTSUPP;
263		break;
264
265	default:
266		error = EINVAL;
267		break;
268	}
269	splx(s);
270	return (error);
271}
272
273static void
274ng_eiface_init(void *xsc)
275{
276	priv_p sc = xsc;
277	struct ifnet *ifp = sc->ifp;
278	int s;
279
280	s = splimp();
281
282	ifp->if_flags |= IFF_RUNNING;
283	ifp->if_flags &= ~IFF_OACTIVE;
284
285	splx(s);
286}
287
288/*
289 * We simply relay the packet to the "ether" hook, if it is connected.
290 * We have been through the netgraph locking and are guaranteed to
291 * be the only code running in this node at this time.
292 */
293static void
294ng_eiface_start2(node_p node, hook_p hook, void *arg1, int arg2)
295{
296	struct ifnet *ifp = arg1;
297	const priv_p priv = (priv_p)ifp->if_softc;
298	int len, error = 0;
299	struct mbuf *m;
300
301	/* Check interface flags */
302	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
303		return;
304
305	/* Don't do anything if output is active */
306	if (ifp->if_flags & IFF_OACTIVE)
307		return;
308
309	ifp->if_flags |= IFF_OACTIVE;
310
311	/*
312	 * Grab a packet to transmit.
313	 */
314	IF_DEQUEUE(&ifp->if_snd, m);
315
316	/* If there's nothing to send, return. */
317	if (m == NULL) {
318		ifp->if_flags &= ~IFF_OACTIVE;
319		return;
320	}
321
322	/*
323	 * Berkeley packet filter.
324	 * Pass packet to bpf if there is a listener.
325	 * XXX is this safe? locking?
326	 */
327	BPF_MTAP(ifp, m);
328
329	/* Copy length before the mbuf gets invalidated */
330	len = m->m_pkthdr.len;
331
332	/*
333	 * Send packet; if hook is not connected, mbuf will get
334	 * freed.
335	 */
336	NG_SEND_DATA_ONLY(error, priv->ether, m);
337
338	/* Update stats */
339	if (error == 0) {
340		ifp->if_obytes += len;
341		ifp->if_opackets++;
342	}
343
344	ifp->if_flags &= ~IFF_OACTIVE;
345
346	return;
347}
348
349/*
350 * This routine is called to deliver a packet out the interface.
351 * We simply queue the netgraph version to be called when netgraph locking
352 * allows it to happen.
353 * Until we know what the rest of the networking code is doing for
354 * locking, we don't know how we will interact with it.
355 * Take comfort from the fact that the ifnet struct is part of our
356 * private info and can't go away while we are queued.
357 * [Though we don't know it is still there now....]
358 * it is possible we don't gain anything from this because
359 * we would like to get the mbuf and queue it as data
360 * somehow, but we can't and if we did would we solve anything?
361 */
362static void
363ng_eiface_start(struct ifnet *ifp)
364{
365
366	const priv_p priv = (priv_p)ifp->if_softc;
367
368	ng_send_fn(priv->node, NULL, &ng_eiface_start2, ifp, 0);
369}
370
371#ifdef DEBUG
372/*
373 * Display an ioctl to the virtual interface
374 */
375
376static void
377ng_eiface_print_ioctl(struct ifnet *ifp, int command, caddr_t data)
378{
379	char *str;
380
381	switch (command & IOC_DIRMASK) {
382	case IOC_VOID:
383		str = "IO";
384		break;
385	case IOC_OUT:
386		str = "IOR";
387		break;
388	case IOC_IN:
389		str = "IOW";
390		break;
391	case IOC_INOUT:
392		str = "IORW";
393		break;
394	default:
395		str = "IO??";
396	}
397	log(LOG_DEBUG, "%s: %s('%c', %d, char[%d])\n",
398	    ifp->if_xname,
399	    str,
400	    IOCGROUP(command),
401	    command & 0xff,
402	    IOCPARM_LEN(command));
403}
404#endif /* DEBUG */
405
406/************************************************************************
407			NETGRAPH NODE STUFF
408 ************************************************************************/
409
410/*
411 * Constructor for a node
412 */
413static int
414ng_eiface_constructor(node_p node)
415{
416	struct ifnet *ifp;
417	priv_p priv;
418	int error = 0;
419
420	/* Allocate node and interface private structures */
421	MALLOC(priv, priv_p, sizeof(*priv), M_NETGRAPH, M_WAITOK);
422	if (priv == NULL)
423		return (ENOMEM);
424	bzero(priv, sizeof(*priv));
425
426	ifp = &(priv->arpcom.ac_if);
427
428	/* Link them together */
429	ifp->if_softc = priv;
430	priv->ifp = ifp;
431
432	/* Get an interface unit number */
433	if ((error = ng_eiface_get_unit(&priv->unit)) != 0) {
434		FREE(priv, M_NETGRAPH);
435		return (error);
436	}
437
438	/* Link together node and private info */
439	NG_NODE_SET_PRIVATE(node, priv);
440	priv->node = node;
441
442	/* Initialize interface structure */
443	if_initname(ifp, NG_EIFACE_EIFACE_NAME, priv->unit);
444	ifp->if_init = ng_eiface_init;
445	ifp->if_output = ether_output;
446	ifp->if_start = ng_eiface_start;
447	ifp->if_ioctl = ng_eiface_ioctl;
448	ifp->if_watchdog = NULL;
449	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
450	ifp->if_flags = (IFF_SIMPLEX | IFF_BROADCAST | IFF_MULTICAST);
451
452#if 0
453	/* Give this node name */
454	bzero(ifname, sizeof(ifname));
455	sprintf(ifname, "if%s", ifp->if_xname);
456	(void)ng_name_node(node, ifname);
457#endif
458
459	/* Attach the interface */
460	ether_ifattach(ifp, priv->arpcom.ac_enaddr);
461
462	/* Done */
463	return (0);
464}
465
466/*
467 * Give our ok for a hook to be added
468 */
469static int
470ng_eiface_newhook(node_p node, hook_p hook, const char *name)
471{
472	priv_p priv = NG_NODE_PRIVATE(node);
473
474	if (strcmp(name, NG_EIFACE_HOOK_ETHER))
475		return (EPFNOSUPPORT);
476	if (priv->ether != NULL)
477		return (EISCONN);
478	priv->ether = hook;
479	NG_HOOK_SET_PRIVATE(hook, &priv->ether);
480
481	return (0);
482}
483
484/*
485 * Receive a control message
486 */
487static int
488ng_eiface_rcvmsg(node_p node, item_p item, hook_p lasthook)
489{
490	const priv_p priv = NG_NODE_PRIVATE(node);
491	struct ifnet *const ifp = priv->ifp;
492	struct ng_mesg *resp = NULL;
493	int error = 0;
494	struct ng_mesg *msg;
495
496	NGI_GET_MSG(item, msg);
497	switch (msg->header.typecookie) {
498	case NGM_EIFACE_COOKIE:
499		switch (msg->header.cmd) {
500
501		case NGM_EIFACE_SET:
502		    {
503			struct ether_addr *eaddr;
504			struct ifaddr *ifa;
505			struct sockaddr_dl *sdl;
506
507			if (msg->header.arglen != sizeof(struct ether_addr)) {
508				error = EINVAL;
509				break;
510			}
511			eaddr = (struct ether_addr *)(msg->data);
512			bcopy(eaddr, priv->arpcom.ac_enaddr, ETHER_ADDR_LEN);
513
514			/* And put it in the ifaddr list */
515			TAILQ_FOREACH(ifa, &(ifp->if_addrhead), ifa_link) {
516				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
517				if (sdl->sdl_type == IFT_ETHER) {
518					bcopy((IFP2AC(ifp))->ac_enaddr,
519						LLADDR(sdl), ifp->if_addrlen);
520					break;
521				}
522			}
523			break;
524		    }
525
526		case NGM_EIFACE_GET_IFNAME:
527			NG_MKRESPONSE(resp, msg, IFNAMSIZ, M_NOWAIT);
528			if (resp == NULL) {
529				error = ENOMEM;
530				break;
531			}
532			strlcpy(resp->data, ifp->if_xname, IFNAMSIZ);
533			break;
534
535		case NGM_EIFACE_GET_IFADDRS:
536		    {
537			struct ifaddr *ifa;
538			caddr_t ptr;
539			int buflen;
540
541#define SA_SIZE(s)	((s)->sa_len<sizeof(*(s))? sizeof(*(s)):(s)->sa_len)
542
543			/* Determine size of response and allocate it */
544			buflen = 0;
545			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
546				buflen += SA_SIZE(ifa->ifa_addr);
547			NG_MKRESPONSE(resp, msg, buflen, M_NOWAIT);
548			if (resp == NULL) {
549				error = ENOMEM;
550				break;
551			}
552
553			/* Add addresses */
554			ptr = resp->data;
555			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
556				const int len = SA_SIZE(ifa->ifa_addr);
557
558				if (buflen < len) {
559					log(LOG_ERR, "%s: len changed?\n",
560					    ifp->if_xname);
561					break;
562				}
563				bcopy(ifa->ifa_addr, ptr, len);
564				ptr += len;
565				buflen -= len;
566			}
567			break;
568#undef SA_SIZE
569		    }
570
571		default:
572			error = EINVAL;
573			break;
574		} /* end of inner switch() */
575		break;
576	case NGM_FLOW_COOKIE:
577		switch (msg->header.cmd) {
578		case NGM_LINK_IS_UP:
579			ifp->if_flags |= IFF_RUNNING;
580			break;
581		case NGM_LINK_IS_DOWN:
582			ifp->if_flags &= ~IFF_RUNNING;
583			break;
584		default:
585			break;
586		}
587		break;
588	default:
589		error = EINVAL;
590		break;
591	}
592	NG_RESPOND_MSG(error, node, item, resp);
593	NG_FREE_MSG(msg);
594	return (error);
595}
596
597/*
598 * Receive data from a hook. Pass the packet to the ether_input routine.
599 */
600static int
601ng_eiface_rcvdata(hook_p hook, item_p item)
602{
603	const priv_p priv = NG_NODE_PRIVATE(NG_HOOK_NODE(hook));
604	struct ifnet *const ifp = priv->ifp;
605	struct mbuf *m;
606
607	NGI_GET_M(item, m);
608	NG_FREE_ITEM(item);
609
610	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
611	    (IFF_UP | IFF_RUNNING)) {
612		NG_FREE_M(m);
613		return (ENETDOWN);
614	}
615
616	/* Note receiving interface */
617	m->m_pkthdr.rcvif = ifp;
618
619	/* Update interface stats */
620	ifp->if_ipackets++;
621
622	(*ifp->if_input)(ifp, m);
623
624	/* Done */
625	return (0);
626}
627
628/*
629 * Shutdown processing.
630 */
631static int
632ng_eiface_rmnode(node_p node)
633{
634	const priv_p priv = NG_NODE_PRIVATE(node);
635	struct ifnet *const ifp = priv->ifp;
636
637	ether_ifdetach(ifp);
638	ng_eiface_free_unit(priv->unit);
639	FREE(priv, M_NETGRAPH);
640	NG_NODE_SET_PRIVATE(node, NULL);
641	NG_NODE_UNREF(node);
642	return (0);
643}
644
645
646/*
647 * This is called once we've already connected a new hook to the other node.
648 * It gives us a chance to balk at the last minute.
649 */
650static int
651ng_eiface_connect(hook_p hook)
652{
653	/* be really amiable and just say "YUP that's OK by me! " */
654	return (0);
655}
656
657/*
658 * Hook disconnection
659 */
660static int
661ng_eiface_disconnect(hook_p hook)
662{
663	const priv_p priv = NG_NODE_PRIVATE(NG_HOOK_NODE(hook));
664
665	priv->ether = NULL;
666	return (0);
667}
668