6 * Copyright (c) 2001 Berkeley Software Design, Inc.
7 * Copyright (c) 2000, 2001
8 * Bill Paul <wpaul@osd.bsdi.com>. All rights reserved.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by Bill Paul.
21 * 4. Neither the name of the author nor the names of any co-contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
29 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
35 * THE POSSIBILITY OF SUCH DAMAGE.
40 * Copyright (c) 1996-1999 Whistle Communications, Inc.
41 * All rights reserved.
43 * Subject to the following obligations and disclaimer of warranty, use and
44 * redistribution of this software, in source or object code forms, with or
45 * without modifications are expressly permitted by Whistle Communications;
46 * provided, however, that:
47 * 1. Any and all reproductions of the source or object code must include the
48 * copyright notice above and the following disclaimer of warranties; and
49 * 2. No rights are granted, in any manner or form, to use Whistle
50 * Communications, Inc. trademarks, including the mark "WHISTLE
51 * COMMUNICATIONS" on advertising, endorsements, or otherwise except as
52 * such appears in the above copyright notice or in the software.
54 * THIS SOFTWARE IS BEING PROVIDED BY WHISTLE COMMUNICATIONS "AS IS", AND
55 * TO THE MAXIMUM EXTENT PERMITTED BY LAW, WHISTLE COMMUNICATIONS MAKES NO
56 * REPRESENTATIONS OR WARRANTIES, EXPRESS OR IMPLIED, REGARDING THIS SOFTWARE,
57 * INCLUDING WITHOUT LIMITATION, ANY AND ALL IMPLIED WARRANTIES OF
58 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT.
59 * WHISTLE COMMUNICATIONS DOES NOT WARRANT, GUARANTEE, OR MAKE ANY
60 * REPRESENTATIONS REGARDING THE USE OF, OR THE RESULTS OF THE USE OF THIS
61 * SOFTWARE IN TERMS OF ITS CORRECTNESS, ACCURACY, RELIABILITY OR OTHERWISE.
62 * IN NO EVENT SHALL WHISTLE COMMUNICATIONS BE LIABLE FOR ANY DAMAGES
63 * RESULTING FROM OR ARISING OUT OF ANY USE OF THIS SOFTWARE, INCLUDING
64 * WITHOUT LIMITATION, ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY,
65 * PUNITIVE, OR CONSEQUENTIAL DAMAGES, PROCUREMENT OF SUBSTITUTE GOODS OR
66 * SERVICES, LOSS OF USE, DATA OR PROFITS, HOWEVER CAUSED AND UNDER ANY
67 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
68 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
69 * THIS SOFTWARE, EVEN IF WHISTLE COMMUNICATIONS IS ADVISED OF THE POSSIBILITY
72 * Author: Archie Cobbs <archie@freebsd.org>
74 * $Whistle: ng_fec.c,v 1.33 1999/11/01 09:24:51 julian Exp $
78 * This module implements ethernet channel bonding using the Cisco
79 * Fast EtherChannel mechanism. Two or four ports may be combined
80 * into a single aggregate interface.
82 * Interfaces are named fec0, fec1, etc. New nodes take the
83 * first available interface name.
85 * This node also includes Berkeley packet filter support.
87 * Note that this node doesn't need to connect to any other
88 * netgraph nodes in order to do its work.
91 #include <sys/param.h>
92 #include <sys/systm.h>
93 #include <sys/errno.h>
94 #include <sys/kernel.h>
95 #include <sys/malloc.h>
97 #include <sys/errno.h>
98 #include <sys/sockio.h>
99 #include <sys/socket.h>
100 #include <sys/syslog.h>
101 #include <sys/libkern.h>
102 #include <sys/queue.h>
105 #include <net/if_dl.h>
106 #include <net/if_types.h>
107 #include <net/if_media.h>
109 #include <net/ethernet.h>
111 #include "opt_inet.h"
112 #include "opt_inet6.h"
114 #include <netinet/in.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/ip.h>
121 #include <netinet/ip6.h>
124 #include <netgraph/ng_message.h>
125 #include <netgraph/netgraph.h>
126 #include <netgraph/ng_parse.h>
127 #include <netgraph/ng_fec.h>
130 * We need a way to stash a pointer to our netgraph node in the
131 * ifnet structure so that receive handling works. As far as I can
132 * tell, although there is an AF_NETGRAPH address family, it's only
133 * used to identify sockaddr_ng structures: there is no netgraph address
134 * family domain. This means the AF_NETGRAPH entry in ifp->if_afdata
135 * should be unused, so we can use to hold our node context.
137 #define IFP2NG(ifp) ((ifp)->if_afdata[AF_NETGRAPH])
140 * Current fast etherchannel implementations use either 2 or 4
141 * ports, so for now we limit the maximum bundle size to 4 interfaces.
143 #define FEC_BUNDLESIZ 4
145 struct ng_fec_portlist {
146 struct ifnet *fec_if;
147 void (*fec_if_input) (struct ifnet *,
151 struct ether_addr fec_mac;
152 TAILQ_ENTRY(ng_fec_portlist) fec_list;
155 struct ng_fec_bundle {
156 TAILQ_HEAD(,ng_fec_portlist) ng_fec_ports;
159 int (*fec_if_output) (struct ifnet *,
165 #define FEC_BTYPE_MAC 0x01
166 #define FEC_BTYPE_INET 0x02
167 #define FEC_BTYPE_INET6 0x03
169 /* Node private data */
170 struct ng_fec_private {
172 struct ifmedia ifmedia;
174 int if_error; /* XXX */
175 int unit; /* Interface unit number */
176 node_p node; /* Our netgraph node */
177 struct ng_fec_bundle fec_bundle;/* Aggregate bundle */
178 struct callout_handle fec_ch; /* callout handle for ticker */
180 typedef struct ng_fec_private *priv_p;
182 /* Interface methods */
183 static void ng_fec_input(struct ifnet *, struct mbuf *);
184 static void ng_fec_start(struct ifnet *ifp);
185 static int ng_fec_choose_port(struct ng_fec_bundle *b,
186 struct mbuf *m, struct ifnet **ifp);
187 static int ng_fec_setport(struct ifnet *ifp, u_long cmd, caddr_t data);
188 static void ng_fec_init(void *arg);
189 static void ng_fec_stop(struct ifnet *ifp);
190 static int ng_fec_ifmedia_upd(struct ifnet *ifp);
191 static void ng_fec_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr);
192 static int ng_fec_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data);
193 static int ng_fec_output(struct ifnet *ifp, struct mbuf *m0,
194 struct sockaddr *dst, struct rtentry *rt0);
195 static void ng_fec_tick(void *arg);
196 static int ng_fec_addport(struct ng_fec_private *priv, char *iface);
197 static int ng_fec_delport(struct ng_fec_private *priv, char *iface);
200 static void ng_fec_print_ioctl(struct ifnet *ifp, int cmd, caddr_t data);
203 /* Netgraph methods */
204 static int ng_fec_mod_event(module_t, int, void *);
205 static ng_constructor_t ng_fec_constructor;
206 static ng_rcvmsg_t ng_fec_rcvmsg;
207 static ng_shutdown_t ng_fec_shutdown;
209 /* List of commands and how to convert arguments to/from ASCII */
210 static const struct ng_cmdlist ng_fec_cmds[] = {
215 &ng_parse_string_type,
222 &ng_parse_string_type,
227 NGM_FEC_SET_MODE_MAC,
234 NGM_FEC_SET_MODE_INET,
242 /* Node type descriptor */
243 static struct ng_type typestruct = {
244 .version = NG_ABI_VERSION,
245 .name = NG_FEC_NODE_TYPE,
246 .mod_event = ng_fec_mod_event,
247 .constructor = ng_fec_constructor,
248 .rcvmsg = ng_fec_rcvmsg,
249 .shutdown = ng_fec_shutdown,
250 .cmdlist = ng_fec_cmds,
252 NETGRAPH_INIT(fec, &typestruct);
254 /* We keep a bitmap indicating which unit numbers are free.
255 One means the unit number is free, zero means it's taken. */
256 static int *ng_fec_units = NULL;
257 static int ng_fec_units_len = 0;
258 static int ng_units_in_use = 0;
260 #define UNITS_BITSPERWORD (sizeof(*ng_fec_units) * NBBY)
262 static struct mtx ng_fec_mtx;
265 * Find the first free unit number for a new interface.
266 * Increase the size of the unit bitmap as necessary.
269 ng_fec_get_unit(int *unit)
273 mtx_lock(&ng_fec_mtx);
274 for (index = 0; index < ng_fec_units_len
275 && ng_fec_units[index] == 0; index++);
276 if (index == ng_fec_units_len) { /* extend array */
277 int i, *newarray, newlen;
279 newlen = (2 * ng_fec_units_len) + 4;
280 MALLOC(newarray, int *, newlen * sizeof(*ng_fec_units),
281 M_NETGRAPH, M_NOWAIT);
282 if (newarray == NULL) {
283 mtx_unlock(&ng_fec_mtx);
286 bcopy(ng_fec_units, newarray,
287 ng_fec_units_len * sizeof(*ng_fec_units));
288 for (i = ng_fec_units_len; i < newlen; i++)
290 if (ng_fec_units != NULL)
291 FREE(ng_fec_units, M_NETGRAPH);
292 ng_fec_units = newarray;
293 ng_fec_units_len = newlen;
295 bit = ffs(ng_fec_units[index]) - 1;
296 KASSERT(bit >= 0 && bit <= UNITS_BITSPERWORD - 1,
297 ("%s: word=%d bit=%d", __func__, ng_fec_units[index], bit));
298 ng_fec_units[index] &= ~(1 << bit);
299 *unit = (index * UNITS_BITSPERWORD) + bit;
301 mtx_unlock(&ng_fec_mtx);
306 * Free a no longer needed unit number.
309 ng_fec_free_unit(int unit)
313 index = unit / UNITS_BITSPERWORD;
314 bit = unit % UNITS_BITSPERWORD;
315 mtx_lock(&ng_fec_mtx);
316 KASSERT(index < ng_fec_units_len,
317 ("%s: unit=%d len=%d", __func__, unit, ng_fec_units_len));
318 KASSERT((ng_fec_units[index] & (1 << bit)) == 0,
319 ("%s: unit=%d is free", __func__, unit));
320 ng_fec_units[index] |= (1 << bit);
322 * XXX We could think about reducing the size of ng_fec_units[]
323 * XXX here if the last portion is all ones
324 * XXX At least free it if no more units
325 * Needed if we are to eventually be able to unload.
328 if (ng_units_in_use == 0) { /* XXX make SMP safe */
329 FREE(ng_fec_units, M_NETGRAPH);
330 ng_fec_units_len = 0;
333 mtx_unlock(&ng_fec_mtx);
336 /************************************************************************
338 ************************************************************************/
341 ng_fec_addport(struct ng_fec_private *priv, char *iface)
343 struct ng_fec_bundle *b;
344 struct ifnet *ifp, *bifp;
345 struct ng_fec_portlist *p, *new;
347 if (priv == NULL || iface == NULL)
350 b = &priv->fec_bundle;
353 /* Only allow reconfiguration if not running. */
354 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
355 printf("fec%d: can't add new iface; bundle is running\n",
360 /* Find the interface */
361 bifp = ifunit(iface);
363 printf("fec%d: tried to add iface %s, which "
364 "doesn't seem to exist\n", priv->unit, iface);
368 /* See if we have room in the bundle */
369 if (b->fec_ifcnt == FEC_BUNDLESIZ) {
370 printf("fec%d: can't add new iface; bundle is full\n",
375 /* See if the interface is already in the bundle */
376 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
377 if (p->fec_if == bifp) {
378 printf("fec%d: iface %s is already in this "
379 "bundle\n", priv->unit, iface);
385 * All interfaces must use the same output vector. Once the
386 * user attaches an interface of one type, make all subsequent
387 * interfaces have the same output vector.
389 if (b->fec_if_output != NULL) {
390 if (b->fec_if_output != bifp->if_output) {
391 printf("fec%d: iface %s is not the same type "
392 "as the other interface(s) already in "
393 "the bundle\n", priv->unit, iface);
398 /* Allocate new list entry. */
399 MALLOC(new, struct ng_fec_portlist *,
400 sizeof(struct ng_fec_portlist), M_NETGRAPH, M_NOWAIT);
404 IF_AFDATA_LOCK(bifp);
405 IFP2NG(bifp) = priv->node;
406 IF_AFDATA_UNLOCK(bifp);
409 * If this is the first interface added to the bundle,
410 * use its MAC address for the virtual interface (and,
411 * by extension, all the other ports in the bundle).
413 if (b->fec_ifcnt == 0)
414 if_setlladdr(ifp, IF_LLADDR(bifp), ETHER_ADDR_LEN);
416 b->fec_btype = FEC_BTYPE_MAC;
417 new->fec_idx = b->fec_ifcnt;
420 /* Save the real MAC address. */
421 bcopy(IF_LLADDR(bifp),
422 (char *)&new->fec_mac, ETHER_ADDR_LEN);
424 /* Set up phony MAC address. */
425 if_setlladdr(bifp, IF_LLADDR(ifp), ETHER_ADDR_LEN);
427 /* Save original input vector */
428 new->fec_if_input = bifp->if_input;
430 /* Override it with our own */
431 bifp->if_input = ng_fec_input;
433 /* Save output vector too. */
434 if (b->fec_if_output == NULL)
435 b->fec_if_output = bifp->if_output;
437 /* Add to the queue */
439 new->fec_ifstat = -1;
440 TAILQ_INSERT_TAIL(&b->ng_fec_ports, new, fec_list);
446 ng_fec_delport(struct ng_fec_private *priv, char *iface)
448 struct ng_fec_bundle *b;
449 struct ifnet *ifp, *bifp;
450 struct ng_fec_portlist *p;
452 if (priv == NULL || iface == NULL)
455 b = &priv->fec_bundle;
458 /* Only allow reconfiguration if not running. */
459 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
460 printf("fec%d: can't remove iface; bundle is running\n",
465 /* Find the interface */
466 bifp = ifunit(iface);
468 printf("fec%d: tried to remove iface %s, which "
469 "doesn't seem to exist\n", priv->unit, iface);
473 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
474 if (p->fec_if == bifp)
479 printf("fec%d: tried to remove iface %s which "
480 "is not in our bundle\n", priv->unit, iface);
485 bifp->if_flags &= ~IFF_UP;
486 (*bifp->if_ioctl)(bifp, SIOCSIFFLAGS, NULL);
488 /* Restore MAC address. */
489 if_setlladdr(bifp, (u_char *)&p->fec_mac, ETHER_ADDR_LEN);
491 /* Restore input vector */
492 bifp->if_input = p->fec_if_input;
494 /* Remove our node context pointer. */
495 IF_AFDATA_LOCK(bifp);
497 IF_AFDATA_UNLOCK(bifp);
500 TAILQ_REMOVE(&b->ng_fec_ports, p, fec_list);
504 if (b->fec_ifcnt == 0)
505 b->fec_if_output = NULL;
511 * Pass an ioctl command down to all the underyling interfaces in a
512 * bundle. Used for setting multicast filters and flags.
516 ng_fec_setport(struct ifnet *ifp, u_long command, caddr_t data)
518 struct ng_fec_private *priv;
519 struct ng_fec_bundle *b;
521 struct ng_fec_portlist *p;
523 priv = ifp->if_softc;
524 b = &priv->fec_bundle;
526 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
529 (*oifp->if_ioctl)(oifp, command, data);
536 ng_fec_init(void *arg)
538 struct ng_fec_private *priv;
539 struct ng_fec_bundle *b;
540 struct ifnet *ifp, *bifp;
541 struct ng_fec_portlist *p;
545 b = &priv->fec_bundle;
547 if (b->fec_ifcnt != 2 && b->fec_ifcnt != FEC_BUNDLESIZ) {
548 printf("fec%d: invalid bundle "
549 "size: %d\n", priv->unit,
556 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
558 bifp->if_flags |= IFF_UP;
559 (*bifp->if_ioctl)(bifp, SIOCSIFFLAGS, NULL);
560 /* mark iface as up and let the monitor check it */
564 ifp->if_drv_flags &= ~(IFF_DRV_OACTIVE);
565 ifp->if_drv_flags |= IFF_DRV_RUNNING;
567 priv->fec_ch = timeout(ng_fec_tick, priv, hz);
573 ng_fec_stop(struct ifnet *ifp)
575 struct ng_fec_private *priv;
576 struct ng_fec_bundle *b;
578 struct ng_fec_portlist *p;
580 priv = ifp->if_softc;
581 b = &priv->fec_bundle;
583 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
585 bifp->if_flags &= ~IFF_UP;
586 (*bifp->if_ioctl)(bifp, SIOCSIFFLAGS, NULL);
589 untimeout(ng_fec_tick, priv, priv->fec_ch);
591 ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
597 ng_fec_tick(void *arg)
599 struct ng_fec_private *priv;
600 struct ng_fec_bundle *b;
601 struct ifmediareq ifmr;
603 struct ng_fec_portlist *p;
607 b = &priv->fec_bundle;
609 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
610 bzero((char *)&ifmr, sizeof(ifmr));
612 error = (*ifp->if_ioctl)(ifp, SIOCGIFMEDIA, (caddr_t)&ifmr);
614 printf("fec%d: failed to check status "
615 "of link %s\n", priv->unit, ifp->if_xname);
619 if (ifmr.ifm_status & IFM_AVALID) {
620 if (ifmr.ifm_status & IFM_ACTIVE) {
621 if (p->fec_ifstat == -1 ||
622 p->fec_ifstat == 0) {
624 printf("fec%d: port %s in bundle "
625 "is up\n", priv->unit,
629 if (p->fec_ifstat == -1 ||
630 p->fec_ifstat == 1) {
632 printf("fec%d: port %s in bundle "
633 "is down\n", priv->unit,
641 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
642 priv->fec_ch = timeout(ng_fec_tick, priv, hz);
648 ng_fec_ifmedia_upd(struct ifnet *ifp)
653 static void ng_fec_ifmedia_sts(struct ifnet *ifp,
654 struct ifmediareq *ifmr)
656 struct ng_fec_private *priv;
657 struct ng_fec_bundle *b;
658 struct ng_fec_portlist *p;
660 priv = ifp->if_softc;
661 b = &priv->fec_bundle;
663 ifmr->ifm_status = IFM_AVALID;
664 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
665 if (p->fec_ifstat == 1) {
666 ifmr->ifm_status |= IFM_ACTIVE;
675 * Process an ioctl for the virtual interface
678 ng_fec_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
680 struct ifreq *const ifr = (struct ifreq *) data;
682 struct ng_fec_private *priv;
683 struct ng_fec_bundle *b;
685 priv = ifp->if_softc;
686 b = &priv->fec_bundle;
689 ng_fec_print_ioctl(ifp, command, data);
694 /* These two are mostly handled at a higher layer */
697 error = ether_ioctl(ifp, command, data);
701 if (ifr->ifr_mtu >= NG_FEC_MTU_MIN &&
702 ifr->ifr_mtu <= NG_FEC_MTU_MAX) {
703 struct ng_fec_portlist *p;
706 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
708 error = (*bifp->if_ioctl)(bifp, SIOCSIFMTU,
714 ifp->if_mtu = ifr->ifr_mtu;
722 * If the interface is marked up and stopped, then start it.
723 * If it is marked down and running, then stop it.
725 if (ifr->ifr_flags & IFF_UP) {
726 if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
728 if (b->fec_ifcnt != 2 &&
729 b->fec_ifcnt != FEC_BUNDLESIZ) {
730 printf("fec%d: invalid bundle "
731 "size: %d\n", priv->unit,
739 * Bubble down changes in promisc mode to
740 * underlying interfaces.
742 if ((ifp->if_flags & IFF_PROMISC) !=
743 (priv->if_flags & IFF_PROMISC)) {
744 ng_fec_setport(ifp, command, data);
745 priv->if_flags = ifp->if_flags;
748 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
755 ng_fec_setport(ifp, command, data);
760 error = ifmedia_ioctl(ifp, ifr, &priv->ifmedia, command);
762 /* Stuff that's not supported */
776 * This routine spies on mbufs received by underlying network device
777 * drivers. When we add an interface to our bundle, we override its
778 * if_input routine with a pointer to ng_fec_input(). This means we
779 * get to look at all the device's packets before sending them to the
780 * real ether_input() for processing by the stack. Once we verify the
781 * packet comes from an interface that's been aggregated into
782 * our bundle, we fix up the rcvif pointer and increment our
783 * packet counters so that it looks like the frames are actually
787 ng_fec_input(struct ifnet *ifp, struct mbuf *m0)
789 struct ng_node *node;
790 struct ng_fec_private *priv;
791 struct ng_fec_bundle *b;
793 struct ng_fec_portlist *p;
796 if (ifp == NULL || m0 == NULL)
801 /* Sanity check part II */
805 priv = NG_NODE_PRIVATE(node);
806 b = &priv->fec_bundle;
809 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
810 if (p->fec_if == m0->m_pkthdr.rcvif)
814 /* Wasn't meant for us; leave this frame alone. */
819 * Check for a BPF tap on the underlying interface. This
820 * is mainly a debugging aid: it allows tcpdump-ing of an
821 * individual interface in a bundle to work, which it
822 * otherwise would not. BPF tapping of our own aggregate
823 * interface will occur once we call ether_input().
825 BPF_MTAP(m0->m_pkthdr.rcvif, m0);
827 /* Convince the system that this is our frame. */
828 m0->m_pkthdr.rcvif = bifp;
831 * Count bytes on an individual interface in a bundle.
832 * The bytes will also be added to the aggregate interface
833 * once we call ether_input().
835 ifp->if_ibytes += m0->m_pkthdr.len;
838 (*bifp->if_input)(bifp, m0);
844 * Take a quick peek at the packet and see if it's ok for us to use
845 * the inet or inet6 hash methods on it, if they're enabled. We do
846 * this by setting flags in the mbuf header. Once we've made up our
847 * mind what to do, we pass the frame to output vector for further
852 ng_fec_output(struct ifnet *ifp, struct mbuf *m,
853 struct sockaddr *dst, struct rtentry *rt0)
855 const priv_p priv = (priv_p) ifp->if_softc;
856 struct ng_fec_bundle *b;
859 /* Check interface flags */
860 if (!((ifp->if_flags & IFF_UP) &&
861 (ifp->if_drv_flags & IFF_DRV_RUNNING))) {
866 b = &priv->fec_bundle;
868 switch (b->fec_btype) {
870 m->m_flags |= M_FEC_MAC;
875 * We can't use the INET address port selection
876 * scheme if this isn't an INET packet.
878 if (dst->sa_family == AF_INET)
879 m->m_flags |= M_FEC_INET;
881 else if (dst->sa_family == AF_INET6)
882 m->m_flags |= M_FEC_INET6;
886 if_printf(ifp, "can't do inet aggregation of non "
889 m->m_flags |= M_FEC_MAC;
894 if_printf(ifp, "bogus hash type: %d\n",
902 * Pass the frame to the output vector for all the protocol
903 * handling. This will put the ethernet header on the packet
907 error = (*b->fec_if_output)(ifp, m, dst, rt0);
908 if (priv->if_error && !error)
909 error = priv->if_error;
915 * Apply a hash to the source and destination addresses in the packet
916 * in order to select an interface. Also check link status and handle
917 * dead links accordingly.
921 ng_fec_choose_port(struct ng_fec_bundle *b,
922 struct mbuf *m, struct ifnet **ifp)
924 struct ether_header *eh;
933 struct ng_fec_portlist *p;
937 * If there are only two ports, mask off all but the
938 * last bit for XORing. If there are 4, mask off all
939 * but the last 2 bits.
941 mask = b->fec_ifcnt == 2 ? 0x1 : 0x3;
942 eh = mtod(m, struct ether_header *);
944 ip = (struct ip *)(mtod(m, char *) +
945 sizeof(struct ether_header));
947 ip6 = (struct ip6_hdr *)(mtod(m, char *) +
948 sizeof(struct ether_header));
953 * The fg_fec_output() routine is supposed to leave a
954 * flag for us in the mbuf that tells us what hash to
955 * use, but sometimes a new mbuf is prepended to the
956 * chain, so we have to search every mbuf in the chain
961 if (m0->m_flags & (M_FEC_MAC|M_FEC_INET|M_FEC_INET6))
968 switch (m0->m_flags & (M_FEC_MAC|M_FEC_INET|M_FEC_INET6)) {
970 port = (eh->ether_dhost[5] ^
971 eh->ether_shost[5]) & mask;
975 port = (ntohl(ip->ip_dst.s_addr) ^
976 ntohl(ip->ip_src.s_addr)) & mask;
980 port = (ip6->ip6_dst.s6_addr[15] ^
981 ip6->ip6_dst.s6_addr[15]) & mask;
990 TAILQ_FOREACH(p, &b->ng_fec_ports, fec_list) {
991 if (port == p->fec_idx)
996 * Now that we've chosen a port, make sure it's
997 * alive. If it's not alive, cycle through the bundle
998 * looking for a port that is alive. If we don't find
999 * any, return an error.
1001 if (p->fec_ifstat != 1) {
1002 struct ng_fec_portlist *n = NULL;
1004 n = TAILQ_NEXT(p, fec_list);
1006 n = TAILQ_FIRST(&b->ng_fec_ports);
1008 if (n->fec_ifstat == 1)
1010 n = TAILQ_NEXT(n, fec_list);
1012 n = TAILQ_FIRST(&b->ng_fec_ports);
1025 * Now that the packet has been run through ether_output(), yank it
1026 * off our own send queue and stick it on the queue for the appropriate
1027 * underlying physical interface. Note that if the interface's send
1028 * queue is full, we save an error status in our private netgraph
1029 * space which will eventually be handed up to ng_fec_output(), which
1030 * will return it to the rest of the IP stack. We need to do this
1031 * in order to duplicate the effect of ether_output() returning ENOBUFS
1032 * when it detects that an interface's send queue is full. There's no
1033 * other way to signal the error status from here since the if_start()
1034 * routine is spec'ed to return void.
1036 * Once the frame is queued, we call ether_output_frame() to initiate
1040 ng_fec_start(struct ifnet *ifp)
1042 struct ng_fec_private *priv;
1043 struct ng_fec_bundle *b;
1044 struct ifnet *oifp = NULL;
1048 priv = ifp->if_softc;
1049 b = &priv->fec_bundle;
1051 IF_DEQUEUE(&ifp->if_snd, m0);
1057 /* Queue up packet on the proper port. */
1058 error = ng_fec_choose_port(b, m0, &oifp);
1062 priv->if_error = ENOBUFS;
1067 priv->if_error = IF_HANDOFF(&oifp->if_snd, m0, oifp) ? 0 : ENOBUFS;
1074 * Display an ioctl to the virtual interface
1078 ng_fec_print_ioctl(struct ifnet *ifp, int command, caddr_t data)
1082 switch (command & IOC_DIRMASK) {
1098 log(LOG_DEBUG, "%s: %s('%c', %d, char[%d])\n",
1103 IOCPARM_LEN(command));
1107 /************************************************************************
1109 ************************************************************************/
1112 * Constructor for a node
1115 ng_fec_constructor(node_p node)
1117 char ifname[NG_FEC_FEC_NAME_MAX + 1];
1120 const uint8_t eaddr[ETHER_ADDR_LEN] = {0, 0, 0, 0, 0, 0};
1121 struct ng_fec_bundle *b;
1124 /* Allocate node and interface private structures */
1125 MALLOC(priv, priv_p, sizeof(*priv), M_NETGRAPH, M_NOWAIT | M_ZERO);
1129 ifp = priv->ifp = if_alloc(IFT_ETHER);
1131 FREE(priv, M_NETGRAPH);
1134 b = &priv->fec_bundle;
1136 /* Link them together */
1137 ifp->if_softc = priv;
1139 /* Get an interface unit number */
1140 if ((error = ng_fec_get_unit(&priv->unit)) != 0) {
1142 FREE(priv, M_NETGRAPH);
1146 /* Link together node and private info */
1147 NG_NODE_SET_PRIVATE(node, priv);
1150 /* Initialize interface structure */
1151 if_initname(ifp, NG_FEC_FEC_NAME, priv->unit);
1152 ifp->if_start = ng_fec_start;
1153 ifp->if_ioctl = ng_fec_ioctl;
1154 ifp->if_init = ng_fec_init;
1155 ifp->if_watchdog = NULL;
1156 ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
1157 ifp->if_mtu = NG_FEC_MTU_DEFAULT;
1158 ifp->if_flags = (IFF_SIMPLEX|IFF_BROADCAST|IFF_MULTICAST);
1159 ifp->if_addrlen = 0; /* XXX */
1160 ifp->if_hdrlen = 0; /* XXX */
1161 ifp->if_baudrate = 100000000; /* XXX */
1162 TAILQ_INIT(&ifp->if_addrhead); /* XXX useless - done in if_attach */
1164 /* Give this node the same name as the interface (if possible) */
1165 bzero(ifname, sizeof(ifname));
1166 strlcpy(ifname, ifp->if_xname, sizeof(ifname));
1167 if (ng_name_node(node, ifname) != 0)
1168 log(LOG_WARNING, "%s: can't acquire netgraph name\n", ifname);
1170 /* Attach the interface */
1171 ether_ifattach(ifp, eaddr);
1172 callout_handle_init(&priv->fec_ch);
1174 /* Override output method with our own */
1175 ifp->if_output = ng_fec_output;
1177 TAILQ_INIT(&b->ng_fec_ports);
1180 ifmedia_init(&priv->ifmedia, 0,
1181 ng_fec_ifmedia_upd, ng_fec_ifmedia_sts);
1182 ifmedia_add(&priv->ifmedia, IFM_ETHER|IFM_NONE, 0, NULL);
1183 ifmedia_set(&priv->ifmedia, IFM_ETHER|IFM_NONE);
1190 * Receive a control message
1193 ng_fec_rcvmsg(node_p node, item_p item, hook_p lasthook)
1195 const priv_p priv = NG_NODE_PRIVATE(node);
1196 struct ng_fec_bundle *b;
1197 struct ng_mesg *resp = NULL;
1198 struct ng_mesg *msg;
1202 NGI_GET_MSG(item, msg);
1203 b = &priv->fec_bundle;
1205 switch (msg->header.typecookie) {
1206 case NGM_FEC_COOKIE:
1207 switch (msg->header.cmd) {
1208 case NGM_FEC_ADD_IFACE:
1210 error = ng_fec_addport(priv, ifname);
1212 case NGM_FEC_DEL_IFACE:
1214 error = ng_fec_delport(priv, ifname);
1216 case NGM_FEC_SET_MODE_MAC:
1217 b->fec_btype = FEC_BTYPE_MAC;
1220 case NGM_FEC_SET_MODE_INET:
1221 b->fec_btype = FEC_BTYPE_INET;
1224 case NGM_FEC_SET_MODE_INET6:
1225 b->fec_btype = FEC_BTYPE_INET6;
1238 NG_RESPOND_MSG(error, node, item, resp);
1244 * Shutdown and remove the node and its associated interface.
1247 ng_fec_shutdown(node_p node)
1249 const priv_p priv = NG_NODE_PRIVATE(node);
1250 struct ng_fec_bundle *b;
1251 struct ng_fec_portlist *p;
1253 b = &priv->fec_bundle;
1254 ng_fec_stop(priv->ifp);
1256 while (!TAILQ_EMPTY(&b->ng_fec_ports)) {
1257 p = TAILQ_FIRST(&b->ng_fec_ports);
1258 ng_fec_delport(priv, p->fec_if->if_xname);
1261 ether_ifdetach(priv->ifp);
1262 if_free_type(priv->ifp, IFT_ETHER);
1263 ifmedia_removeall(&priv->ifmedia);
1264 ng_fec_free_unit(priv->unit);
1265 FREE(priv, M_NETGRAPH);
1266 NG_NODE_SET_PRIVATE(node, NULL);
1267 NG_NODE_UNREF(node);
1272 * Handle loading and unloading for this node type.
1275 ng_fec_mod_event(module_t mod, int event, void *data)
1281 mtx_init(&ng_fec_mtx, "ng_fec", NULL, MTX_DEF);
1284 mtx_destroy(&ng_fec_mtx);