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.
37 * $FreeBSD: src/sys/netgraph/ng_fec.c,v 1.30 2007/05/18 15:05:49 dwmalone Exp $
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 SLIST_HEAD(__mclhd
, ng_fec_mc
) fec_mc_head
;
153 TAILQ_ENTRY(ng_fec_portlist
) fec_list
;
157 struct ifmultiaddr
*mc_ifma
;
158 SLIST_ENTRY(ng_fec_mc
) mc_entries
;
161 struct ng_fec_bundle
{
162 TAILQ_HEAD(,ng_fec_portlist
) ng_fec_ports
;
165 int (*fec_if_output
) (struct ifnet
*,
171 #define FEC_BTYPE_MAC 0x01
172 #define FEC_BTYPE_INET 0x02
173 #define FEC_BTYPE_INET6 0x03
175 /* Node private data */
176 struct ng_fec_private
{
178 struct ifmedia ifmedia
;
180 int if_error
; /* XXX */
181 int unit
; /* Interface unit number */
182 node_p node
; /* Our netgraph node */
183 struct ng_fec_bundle fec_bundle
;/* Aggregate bundle */
184 struct callout_handle fec_ch
; /* callout handle for ticker */
186 typedef struct ng_fec_private
*priv_p
;
188 /* Interface methods */
189 static void ng_fec_input(struct ifnet
*, struct mbuf
*);
190 static void ng_fec_start(struct ifnet
*ifp
);
191 static int ng_fec_choose_port(struct ng_fec_bundle
*b
,
192 struct mbuf
*m
, struct ifnet
**ifp
);
193 static int ng_fec_setport(struct ifnet
*ifp
, u_long cmd
, caddr_t data
);
194 static void ng_fec_init(void *arg
);
195 static void ng_fec_stop(struct ifnet
*ifp
);
196 static int ng_fec_ifmedia_upd(struct ifnet
*ifp
);
197 static void ng_fec_ifmedia_sts(struct ifnet
*ifp
, struct ifmediareq
*ifmr
);
198 static int ng_fec_ioctl(struct ifnet
*ifp
, u_long cmd
, caddr_t data
);
199 static int ng_fec_output(struct ifnet
*ifp
, struct mbuf
*m0
,
200 struct sockaddr
*dst
, struct rtentry
*rt0
);
201 static void ng_fec_tick(void *arg
);
202 static int ng_fec_addport(struct ng_fec_private
*priv
, char *iface
);
203 static int ng_fec_delport(struct ng_fec_private
*priv
, char *iface
);
204 static int ng_fec_ether_cmdmulti(struct ifnet
*trifp
, struct ng_fec_portlist
*p
, int set
);
207 static void ng_fec_print_ioctl(struct ifnet
*ifp
, int cmd
, caddr_t data
);
210 /* Netgraph methods */
211 static int ng_fec_mod_event(module_t
, int, void *);
212 static ng_constructor_t ng_fec_constructor
;
213 static ng_rcvmsg_t ng_fec_rcvmsg
;
214 static ng_shutdown_t ng_fec_shutdown
;
216 /* List of commands and how to convert arguments to/from ASCII */
217 static const struct ng_cmdlist ng_fec_cmds
[] = {
222 &ng_parse_string_type
,
229 &ng_parse_string_type
,
234 NGM_FEC_SET_MODE_MAC
,
241 NGM_FEC_SET_MODE_INET
,
249 /* Node type descriptor */
250 static struct ng_type typestruct
= {
251 .version
= NG_ABI_VERSION
,
252 .name
= NG_FEC_NODE_TYPE
,
253 .mod_event
= ng_fec_mod_event
,
254 .constructor
= ng_fec_constructor
,
255 .rcvmsg
= ng_fec_rcvmsg
,
256 .shutdown
= ng_fec_shutdown
,
257 .cmdlist
= ng_fec_cmds
,
259 NETGRAPH_INIT(fec
, &typestruct
);
261 /* We keep a bitmap indicating which unit numbers are free.
262 One means the unit number is free, zero means it's taken. */
263 static int *ng_fec_units
= NULL
;
264 static int ng_fec_units_len
= 0;
265 static int ng_units_in_use
= 0;
267 #define UNITS_BITSPERWORD (sizeof(*ng_fec_units) * NBBY)
269 static struct mtx ng_fec_mtx
;
272 * Find the first free unit number for a new interface.
273 * Increase the size of the unit bitmap as necessary.
276 ng_fec_get_unit(int *unit
)
280 mtx_lock(&ng_fec_mtx
);
281 for (index
= 0; index
< ng_fec_units_len
282 && ng_fec_units
[index
] == 0; index
++);
283 if (index
== ng_fec_units_len
) { /* extend array */
284 int i
, *newarray
, newlen
;
286 newlen
= (2 * ng_fec_units_len
) + 4;
287 MALLOC(newarray
, int *, newlen
* sizeof(*ng_fec_units
),
288 M_NETGRAPH
, M_NOWAIT
);
289 if (newarray
== NULL
) {
290 mtx_unlock(&ng_fec_mtx
);
293 bcopy(ng_fec_units
, newarray
,
294 ng_fec_units_len
* sizeof(*ng_fec_units
));
295 for (i
= ng_fec_units_len
; i
< newlen
; i
++)
297 if (ng_fec_units
!= NULL
)
298 FREE(ng_fec_units
, M_NETGRAPH
);
299 ng_fec_units
= newarray
;
300 ng_fec_units_len
= newlen
;
302 bit
= ffs(ng_fec_units
[index
]) - 1;
303 KASSERT(bit
>= 0 && bit
<= UNITS_BITSPERWORD
- 1,
304 ("%s: word=%d bit=%d", __func__
, ng_fec_units
[index
], bit
));
305 ng_fec_units
[index
] &= ~(1 << bit
);
306 *unit
= (index
* UNITS_BITSPERWORD
) + bit
;
308 mtx_unlock(&ng_fec_mtx
);
313 * Free a no longer needed unit number.
316 ng_fec_free_unit(int unit
)
320 index
= unit
/ UNITS_BITSPERWORD
;
321 bit
= unit
% UNITS_BITSPERWORD
;
322 mtx_lock(&ng_fec_mtx
);
323 KASSERT(index
< ng_fec_units_len
,
324 ("%s: unit=%d len=%d", __func__
, unit
, ng_fec_units_len
));
325 KASSERT((ng_fec_units
[index
] & (1 << bit
)) == 0,
326 ("%s: unit=%d is free", __func__
, unit
));
327 ng_fec_units
[index
] |= (1 << bit
);
329 * XXX We could think about reducing the size of ng_fec_units[]
330 * XXX here if the last portion is all ones
331 * XXX At least free it if no more units
332 * Needed if we are to eventually be able to unload.
335 if (ng_units_in_use
== 0) { /* XXX make SMP safe */
336 FREE(ng_fec_units
, M_NETGRAPH
);
337 ng_fec_units_len
= 0;
340 mtx_unlock(&ng_fec_mtx
);
343 /************************************************************************
345 ************************************************************************/
348 ng_fec_addport(struct ng_fec_private
*priv
, char *iface
)
350 struct ng_fec_bundle
*b
;
351 struct ifnet
*ifp
, *bifp
;
352 struct ng_fec_portlist
*p
, *new;
354 if (priv
== NULL
|| iface
== NULL
)
357 b
= &priv
->fec_bundle
;
360 /* Only allow reconfiguration if not running. */
361 if (ifp
->if_drv_flags
& IFF_DRV_RUNNING
) {
362 printf("fec%d: can't add new iface; bundle is running\n",
367 /* Find the interface */
368 bifp
= ifunit(iface
);
370 printf("fec%d: tried to add iface %s, which "
371 "doesn't seem to exist\n", priv
->unit
, iface
);
375 /* See if we have room in the bundle */
376 if (b
->fec_ifcnt
== FEC_BUNDLESIZ
) {
377 printf("fec%d: can't add new iface; bundle is full\n",
382 /* See if the interface is already in the bundle */
383 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
384 if (p
->fec_if
== bifp
) {
385 printf("fec%d: iface %s is already in this "
386 "bundle\n", priv
->unit
, iface
);
392 * All interfaces must use the same output vector. Once the
393 * user attaches an interface of one type, make all subsequent
394 * interfaces have the same output vector.
396 if (b
->fec_if_output
!= NULL
) {
397 if (b
->fec_if_output
!= bifp
->if_output
) {
398 printf("fec%d: iface %s is not the same type "
399 "as the other interface(s) already in "
400 "the bundle\n", priv
->unit
, iface
);
405 /* Allocate new list entry. */
406 MALLOC(new, struct ng_fec_portlist
*,
407 sizeof(struct ng_fec_portlist
), M_NETGRAPH
, M_NOWAIT
);
411 IF_AFDATA_LOCK(bifp
);
412 IFP2NG(bifp
) = priv
->node
;
413 IF_AFDATA_UNLOCK(bifp
);
416 * If this is the first interface added to the bundle,
417 * use its MAC address for the virtual interface (and,
418 * by extension, all the other ports in the bundle).
420 if (b
->fec_ifcnt
== 0)
421 if_setlladdr(ifp
, IF_LLADDR(bifp
), ETHER_ADDR_LEN
);
423 b
->fec_btype
= FEC_BTYPE_MAC
;
424 new->fec_idx
= b
->fec_ifcnt
;
427 /* Initialise the list of multicast addresses that we own. */
428 SLIST_INIT(&new->fec_mc_head
);
430 /* Save the real MAC address. */
431 bcopy(IF_LLADDR(bifp
),
432 (char *)&new->fec_mac
, ETHER_ADDR_LEN
);
434 /* Set up phony MAC address. */
435 if_setlladdr(bifp
, IF_LLADDR(ifp
), ETHER_ADDR_LEN
);
437 /* Save original input vector */
438 new->fec_if_input
= bifp
->if_input
;
440 /* Override it with our own */
441 bifp
->if_input
= ng_fec_input
;
443 /* Save output vector too. */
444 if (b
->fec_if_output
== NULL
)
445 b
->fec_if_output
= bifp
->if_output
;
447 /* Add to the queue */
449 new->fec_ifstat
= -1;
450 TAILQ_INSERT_TAIL(&b
->ng_fec_ports
, new, fec_list
);
452 /* Add multicast addresses to this port. */
453 ng_fec_ether_cmdmulti(ifp
, new, 1);
459 ng_fec_delport(struct ng_fec_private
*priv
, char *iface
)
461 struct ng_fec_bundle
*b
;
462 struct ifnet
*ifp
, *bifp
;
463 struct ng_fec_portlist
*p
;
465 if (priv
== NULL
|| iface
== NULL
)
468 b
= &priv
->fec_bundle
;
471 /* Only allow reconfiguration if not running. */
472 if (ifp
->if_drv_flags
& IFF_DRV_RUNNING
) {
473 printf("fec%d: can't remove iface; bundle is running\n",
478 /* Find the interface */
479 bifp
= ifunit(iface
);
481 printf("fec%d: tried to remove iface %s, which "
482 "doesn't seem to exist\n", priv
->unit
, iface
);
486 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
487 if (p
->fec_if
== bifp
)
492 printf("fec%d: tried to remove iface %s which "
493 "is not in our bundle\n", priv
->unit
, iface
);
498 bifp
->if_flags
&= ~IFF_UP
;
499 (*bifp
->if_ioctl
)(bifp
, SIOCSIFFLAGS
, NULL
);
501 /* Restore MAC address. */
502 if_setlladdr(bifp
, (u_char
*)&p
->fec_mac
, ETHER_ADDR_LEN
);
504 /* Restore input vector */
505 bifp
->if_input
= p
->fec_if_input
;
507 /* Remove our node context pointer. */
508 IF_AFDATA_LOCK(bifp
);
510 IF_AFDATA_UNLOCK(bifp
);
513 TAILQ_REMOVE(&b
->ng_fec_ports
, p
, fec_list
);
517 if (b
->fec_ifcnt
== 0)
518 b
->fec_if_output
= NULL
;
524 ng_fec_ether_cmdmulti(struct ifnet
*trifp
, struct ng_fec_portlist
*p
, int set
)
526 struct ifnet
*ifp
= p
->fec_if
;
527 struct ng_fec_mc
*mc
;
528 struct ifmultiaddr
*ifma
, *rifma
= NULL
;
529 struct sockaddr_dl sdl
;
532 bzero((char *)&sdl
, sizeof(sdl
));
533 sdl
.sdl_len
= sizeof(sdl
);
534 sdl
.sdl_family
= AF_LINK
;
535 sdl
.sdl_type
= IFT_ETHER
;
536 sdl
.sdl_alen
= ETHER_ADDR_LEN
;
537 sdl
.sdl_index
= ifp
->if_index
;
540 TAILQ_FOREACH(ifma
, &trifp
->if_multiaddrs
, ifma_link
) {
541 if (ifma
->ifma_addr
->sa_family
!= AF_LINK
)
543 bcopy(LLADDR((struct sockaddr_dl
*)ifma
->ifma_addr
),
544 LLADDR(&sdl
), ETHER_ADDR_LEN
);
546 error
= if_addmulti(ifp
, (struct sockaddr
*)&sdl
, &rifma
);
549 mc
= malloc(sizeof(struct ng_fec_mc
), M_DEVBUF
, M_NOWAIT
);
553 SLIST_INSERT_HEAD(&p
->fec_mc_head
, mc
, mc_entries
);
556 while ((mc
= SLIST_FIRST(&p
->fec_mc_head
)) != NULL
) {
557 SLIST_REMOVE(&p
->fec_mc_head
, mc
, ng_fec_mc
, mc_entries
);
558 if_delmulti_ifma(mc
->mc_ifma
);
566 ng_fec_ether_setmulti(struct ifnet
*ifp
)
568 struct ng_fec_private
*priv
;
569 struct ng_fec_bundle
*b
;
570 struct ng_fec_portlist
*p
;
572 priv
= ifp
->if_softc
;
573 b
= &priv
->fec_bundle
;
575 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
576 /* First, remove any existing filter entries. */
577 ng_fec_ether_cmdmulti(ifp
, p
, 0);
578 /* copy all addresses from the fec interface to the port */
579 ng_fec_ether_cmdmulti(ifp
, p
, 1);
585 * Pass an ioctl command down to all the underyling interfaces in a
586 * bundle. Used for setting flags.
590 ng_fec_setport(struct ifnet
*ifp
, u_long command
, caddr_t data
)
592 struct ng_fec_private
*priv
;
593 struct ng_fec_bundle
*b
;
595 struct ng_fec_portlist
*p
;
597 priv
= ifp
->if_softc
;
598 b
= &priv
->fec_bundle
;
600 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
603 (*oifp
->if_ioctl
)(oifp
, command
, data
);
610 ng_fec_init(void *arg
)
612 struct ng_fec_private
*priv
;
613 struct ng_fec_bundle
*b
;
614 struct ifnet
*ifp
, *bifp
;
615 struct ng_fec_portlist
*p
;
619 b
= &priv
->fec_bundle
;
621 if (b
->fec_ifcnt
!= 2 && b
->fec_ifcnt
!= FEC_BUNDLESIZ
) {
622 printf("fec%d: invalid bundle "
623 "size: %d\n", priv
->unit
,
630 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
632 bifp
->if_flags
|= IFF_UP
;
633 (*bifp
->if_ioctl
)(bifp
, SIOCSIFFLAGS
, NULL
);
634 /* mark iface as up and let the monitor check it */
638 ifp
->if_drv_flags
&= ~(IFF_DRV_OACTIVE
);
639 ifp
->if_drv_flags
|= IFF_DRV_RUNNING
;
641 priv
->fec_ch
= timeout(ng_fec_tick
, priv
, hz
);
647 ng_fec_stop(struct ifnet
*ifp
)
649 struct ng_fec_private
*priv
;
650 struct ng_fec_bundle
*b
;
652 struct ng_fec_portlist
*p
;
654 priv
= ifp
->if_softc
;
655 b
= &priv
->fec_bundle
;
657 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
659 bifp
->if_flags
&= ~IFF_UP
;
660 (*bifp
->if_ioctl
)(bifp
, SIOCSIFFLAGS
, NULL
);
663 untimeout(ng_fec_tick
, priv
, priv
->fec_ch
);
665 ifp
->if_drv_flags
&= ~(IFF_DRV_RUNNING
| IFF_DRV_OACTIVE
);
671 ng_fec_tick(void *arg
)
673 struct ng_fec_private
*priv
;
674 struct ng_fec_bundle
*b
;
675 struct ifmediareq ifmr
;
677 struct ng_fec_portlist
*p
;
681 b
= &priv
->fec_bundle
;
683 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
684 bzero((char *)&ifmr
, sizeof(ifmr
));
686 error
= (*ifp
->if_ioctl
)(ifp
, SIOCGIFMEDIA
, (caddr_t
)&ifmr
);
688 printf("fec%d: failed to check status "
689 "of link %s\n", priv
->unit
, ifp
->if_xname
);
693 if (ifmr
.ifm_status
& IFM_AVALID
) {
694 if (ifmr
.ifm_status
& IFM_ACTIVE
) {
695 if (p
->fec_ifstat
== -1 ||
696 p
->fec_ifstat
== 0) {
698 printf("fec%d: port %s in bundle "
699 "is up\n", priv
->unit
,
703 if (p
->fec_ifstat
== -1 ||
704 p
->fec_ifstat
== 1) {
706 printf("fec%d: port %s in bundle "
707 "is down\n", priv
->unit
,
715 if (ifp
->if_drv_flags
& IFF_DRV_RUNNING
)
716 priv
->fec_ch
= timeout(ng_fec_tick
, priv
, hz
);
722 ng_fec_ifmedia_upd(struct ifnet
*ifp
)
727 static void ng_fec_ifmedia_sts(struct ifnet
*ifp
,
728 struct ifmediareq
*ifmr
)
730 struct ng_fec_private
*priv
;
731 struct ng_fec_bundle
*b
;
732 struct ng_fec_portlist
*p
;
734 priv
= ifp
->if_softc
;
735 b
= &priv
->fec_bundle
;
737 ifmr
->ifm_status
= IFM_AVALID
;
738 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
739 if (p
->fec_ifstat
== 1) {
740 ifmr
->ifm_status
|= IFM_ACTIVE
;
749 * Process an ioctl for the virtual interface
752 ng_fec_ioctl(struct ifnet
*ifp
, u_long command
, caddr_t data
)
754 struct ifreq
*const ifr
= (struct ifreq
*) data
;
756 struct ng_fec_private
*priv
;
757 struct ng_fec_bundle
*b
;
759 priv
= ifp
->if_softc
;
760 b
= &priv
->fec_bundle
;
763 ng_fec_print_ioctl(ifp
, command
, data
);
768 /* These two are mostly handled at a higher layer */
771 error
= ether_ioctl(ifp
, command
, data
);
775 if (ifr
->ifr_mtu
>= NG_FEC_MTU_MIN
&&
776 ifr
->ifr_mtu
<= NG_FEC_MTU_MAX
) {
777 struct ng_fec_portlist
*p
;
780 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
782 error
= (*bifp
->if_ioctl
)(bifp
, SIOCSIFMTU
,
788 ifp
->if_mtu
= ifr
->ifr_mtu
;
796 * If the interface is marked up and stopped, then start it.
797 * If it is marked down and running, then stop it.
799 if (ifr
->ifr_flags
& IFF_UP
) {
800 if (!(ifp
->if_drv_flags
& IFF_DRV_RUNNING
)) {
802 if (b
->fec_ifcnt
!= 2 &&
803 b
->fec_ifcnt
!= FEC_BUNDLESIZ
) {
804 printf("fec%d: invalid bundle "
805 "size: %d\n", priv
->unit
,
813 * Bubble down changes in promisc mode to
814 * underlying interfaces.
816 if ((ifp
->if_flags
& IFF_PROMISC
) !=
817 (priv
->if_flags
& IFF_PROMISC
)) {
818 ng_fec_setport(ifp
, command
, data
);
819 priv
->if_flags
= ifp
->if_flags
;
822 if (ifp
->if_drv_flags
& IFF_DRV_RUNNING
)
829 ng_fec_ether_setmulti(ifp
);
834 error
= ifmedia_ioctl(ifp
, ifr
, &priv
->ifmedia
, command
);
836 /* Stuff that's not supported */
850 * This routine spies on mbufs received by underlying network device
851 * drivers. When we add an interface to our bundle, we override its
852 * if_input routine with a pointer to ng_fec_input(). This means we
853 * get to look at all the device's packets before sending them to the
854 * real ether_input() for processing by the stack. Once we verify the
855 * packet comes from an interface that's been aggregated into
856 * our bundle, we fix up the rcvif pointer and increment our
857 * packet counters so that it looks like the frames are actually
861 ng_fec_input(struct ifnet
*ifp
, struct mbuf
*m0
)
863 struct ng_node
*node
;
864 struct ng_fec_private
*priv
;
865 struct ng_fec_bundle
*b
;
867 struct ng_fec_portlist
*p
;
870 if (ifp
== NULL
|| m0
== NULL
)
875 /* Sanity check part II */
879 priv
= NG_NODE_PRIVATE(node
);
880 b
= &priv
->fec_bundle
;
883 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
884 if (p
->fec_if
== m0
->m_pkthdr
.rcvif
)
888 /* Wasn't meant for us; leave this frame alone. */
893 * Check for a BPF tap on the underlying interface. This
894 * is mainly a debugging aid: it allows tcpdump-ing of an
895 * individual interface in a bundle to work, which it
896 * otherwise would not. BPF tapping of our own aggregate
897 * interface will occur once we call ether_input().
899 BPF_MTAP(m0
->m_pkthdr
.rcvif
, m0
);
901 /* Convince the system that this is our frame. */
902 m0
->m_pkthdr
.rcvif
= bifp
;
905 * Count bytes on an individual interface in a bundle.
906 * The bytes will also be added to the aggregate interface
907 * once we call ether_input().
909 ifp
->if_ibytes
+= m0
->m_pkthdr
.len
;
912 (*bifp
->if_input
)(bifp
, m0
);
918 * Take a quick peek at the packet and see if it's ok for us to use
919 * the inet or inet6 hash methods on it, if they're enabled. We do
920 * this by setting flags in the mbuf header. Once we've made up our
921 * mind what to do, we pass the frame to output vector for further
926 ng_fec_output(struct ifnet
*ifp
, struct mbuf
*m
,
927 struct sockaddr
*dst
, struct rtentry
*rt0
)
929 const priv_p priv
= (priv_p
) ifp
->if_softc
;
930 struct ng_fec_bundle
*b
;
933 /* Check interface flags */
934 if (!((ifp
->if_flags
& IFF_UP
) &&
935 (ifp
->if_drv_flags
& IFF_DRV_RUNNING
))) {
940 b
= &priv
->fec_bundle
;
942 switch (b
->fec_btype
) {
944 m
->m_flags
|= M_FEC_MAC
;
949 * We can't use the INET address port selection
950 * scheme if this isn't an INET packet.
952 if (dst
->sa_family
== AF_INET
)
953 m
->m_flags
|= M_FEC_INET
;
955 else if (dst
->sa_family
== AF_INET6
)
956 m
->m_flags
|= M_FEC_INET6
;
960 if_printf(ifp
, "can't do inet aggregation of non "
963 m
->m_flags
|= M_FEC_MAC
;
968 if_printf(ifp
, "bogus hash type: %d\n",
976 * Pass the frame to the output vector for all the protocol
977 * handling. This will put the ethernet header on the packet
981 error
= (*b
->fec_if_output
)(ifp
, m
, dst
, rt0
);
982 if (priv
->if_error
&& !error
)
983 error
= priv
->if_error
;
989 * Apply a hash to the source and destination addresses in the packet
990 * in order to select an interface. Also check link status and handle
991 * dead links accordingly.
995 ng_fec_choose_port(struct ng_fec_bundle
*b
,
996 struct mbuf
*m
, struct ifnet
**ifp
)
998 struct ether_header
*eh
;
1003 struct ip6_hdr
*ip6
;
1007 struct ng_fec_portlist
*p
;
1011 * If there are only two ports, mask off all but the
1012 * last bit for XORing. If there are 4, mask off all
1013 * but the last 2 bits.
1015 mask
= b
->fec_ifcnt
== 2 ? 0x1 : 0x3;
1016 eh
= mtod(m
, struct ether_header
*);
1018 ip
= (struct ip
*)(mtod(m
, char *) +
1019 sizeof(struct ether_header
));
1021 ip6
= (struct ip6_hdr
*)(mtod(m
, char *) +
1022 sizeof(struct ether_header
));
1027 * The fg_fec_output() routine is supposed to leave a
1028 * flag for us in the mbuf that tells us what hash to
1029 * use, but sometimes a new mbuf is prepended to the
1030 * chain, so we have to search every mbuf in the chain
1031 * to find the flags.
1035 if (m0
->m_flags
& (M_FEC_MAC
|M_FEC_INET
|M_FEC_INET6
))
1042 switch (m0
->m_flags
& (M_FEC_MAC
|M_FEC_INET
|M_FEC_INET6
)) {
1044 port
= (eh
->ether_dhost
[5] ^
1045 eh
->ether_shost
[5]) & mask
;
1049 port
= (ntohl(ip
->ip_dst
.s_addr
) ^
1050 ntohl(ip
->ip_src
.s_addr
)) & mask
;
1054 port
= (ip6
->ip6_dst
.s6_addr
[15] ^
1055 ip6
->ip6_dst
.s6_addr
[15]) & mask
;
1064 TAILQ_FOREACH(p
, &b
->ng_fec_ports
, fec_list
) {
1065 if (port
== p
->fec_idx
)
1070 * Now that we've chosen a port, make sure it's
1071 * alive. If it's not alive, cycle through the bundle
1072 * looking for a port that is alive. If we don't find
1073 * any, return an error.
1075 if (p
->fec_ifstat
!= 1) {
1076 struct ng_fec_portlist
*n
= NULL
;
1078 n
= TAILQ_NEXT(p
, fec_list
);
1080 n
= TAILQ_FIRST(&b
->ng_fec_ports
);
1082 if (n
->fec_ifstat
== 1)
1084 n
= TAILQ_NEXT(n
, fec_list
);
1086 n
= TAILQ_FIRST(&b
->ng_fec_ports
);
1099 * Now that the packet has been run through ether_output(), yank it
1100 * off our own send queue and stick it on the queue for the appropriate
1101 * underlying physical interface. Note that if the interface's send
1102 * queue is full, we save an error status in our private netgraph
1103 * space which will eventually be handed up to ng_fec_output(), which
1104 * will return it to the rest of the IP stack. We need to do this
1105 * in order to duplicate the effect of ether_output() returning ENOBUFS
1106 * when it detects that an interface's send queue is full. There's no
1107 * other way to signal the error status from here since the if_start()
1108 * routine is spec'ed to return void.
1110 * Once the frame is queued, we call ether_output_frame() to initiate
1114 ng_fec_start(struct ifnet
*ifp
)
1116 struct ng_fec_private
*priv
;
1117 struct ng_fec_bundle
*b
;
1118 struct ifnet
*oifp
= NULL
;
1122 priv
= ifp
->if_softc
;
1123 b
= &priv
->fec_bundle
;
1125 IF_DEQUEUE(&ifp
->if_snd
, m0
);
1131 /* Queue up packet on the proper port. */
1132 error
= ng_fec_choose_port(b
, m0
, &oifp
);
1136 priv
->if_error
= ENOBUFS
;
1141 priv
->if_error
= IF_HANDOFF(&oifp
->if_snd
, m0
, oifp
) ? 0 : ENOBUFS
;
1148 * Display an ioctl to the virtual interface
1152 ng_fec_print_ioctl(struct ifnet
*ifp
, int command
, caddr_t data
)
1156 switch (command
& IOC_DIRMASK
) {
1172 log(LOG_DEBUG
, "%s: %s('%c', %d, char[%d])\n",
1177 IOCPARM_LEN(command
));
1181 /************************************************************************
1183 ************************************************************************/
1186 * Constructor for a node
1189 ng_fec_constructor(node_p node
)
1191 char ifname
[NG_FEC_FEC_NAME_MAX
+ 1];
1194 const uint8_t eaddr
[ETHER_ADDR_LEN
] = {0, 0, 0, 0, 0, 0};
1195 struct ng_fec_bundle
*b
;
1198 /* Allocate node and interface private structures */
1199 MALLOC(priv
, priv_p
, sizeof(*priv
), M_NETGRAPH
, M_NOWAIT
| M_ZERO
);
1203 ifp
= priv
->ifp
= if_alloc(IFT_ETHER
);
1205 FREE(priv
, M_NETGRAPH
);
1208 b
= &priv
->fec_bundle
;
1210 /* Link them together */
1211 ifp
->if_softc
= priv
;
1213 /* Get an interface unit number */
1214 if ((error
= ng_fec_get_unit(&priv
->unit
)) != 0) {
1216 FREE(priv
, M_NETGRAPH
);
1220 /* Link together node and private info */
1221 NG_NODE_SET_PRIVATE(node
, priv
);
1224 /* Initialize interface structure */
1225 if_initname(ifp
, NG_FEC_FEC_NAME
, priv
->unit
);
1226 ifp
->if_start
= ng_fec_start
;
1227 ifp
->if_ioctl
= ng_fec_ioctl
;
1228 ifp
->if_init
= ng_fec_init
;
1229 ifp
->if_watchdog
= NULL
;
1230 ifp
->if_snd
.ifq_maxlen
= IFQ_MAXLEN
;
1231 ifp
->if_mtu
= NG_FEC_MTU_DEFAULT
;
1232 ifp
->if_flags
= (IFF_SIMPLEX
|IFF_BROADCAST
|IFF_MULTICAST
);
1233 ifp
->if_addrlen
= 0; /* XXX */
1234 ifp
->if_hdrlen
= 0; /* XXX */
1235 ifp
->if_baudrate
= 100000000; /* XXX */
1236 TAILQ_INIT(&ifp
->if_addrhead
); /* XXX useless - done in if_attach */
1238 /* Give this node the same name as the interface (if possible) */
1239 bzero(ifname
, sizeof(ifname
));
1240 strlcpy(ifname
, ifp
->if_xname
, sizeof(ifname
));
1241 if (ng_name_node(node
, ifname
) != 0)
1242 log(LOG_WARNING
, "%s: can't acquire netgraph name\n", ifname
);
1244 /* Attach the interface */
1245 ether_ifattach(ifp
, eaddr
);
1246 callout_handle_init(&priv
->fec_ch
);
1248 /* Override output method with our own */
1249 ifp
->if_output
= ng_fec_output
;
1251 TAILQ_INIT(&b
->ng_fec_ports
);
1254 ifmedia_init(&priv
->ifmedia
, 0,
1255 ng_fec_ifmedia_upd
, ng_fec_ifmedia_sts
);
1256 ifmedia_add(&priv
->ifmedia
, IFM_ETHER
|IFM_NONE
, 0, NULL
);
1257 ifmedia_set(&priv
->ifmedia
, IFM_ETHER
|IFM_NONE
);
1264 * Receive a control message
1267 ng_fec_rcvmsg(node_p node
, item_p item
, hook_p lasthook
)
1269 const priv_p priv
= NG_NODE_PRIVATE(node
);
1270 struct ng_fec_bundle
*b
;
1271 struct ng_mesg
*resp
= NULL
;
1272 struct ng_mesg
*msg
;
1276 NGI_GET_MSG(item
, msg
);
1277 b
= &priv
->fec_bundle
;
1279 switch (msg
->header
.typecookie
) {
1280 case NGM_FEC_COOKIE
:
1281 switch (msg
->header
.cmd
) {
1282 case NGM_FEC_ADD_IFACE
:
1284 error
= ng_fec_addport(priv
, ifname
);
1286 case NGM_FEC_DEL_IFACE
:
1288 error
= ng_fec_delport(priv
, ifname
);
1290 case NGM_FEC_SET_MODE_MAC
:
1291 b
->fec_btype
= FEC_BTYPE_MAC
;
1294 case NGM_FEC_SET_MODE_INET
:
1295 b
->fec_btype
= FEC_BTYPE_INET
;
1298 case NGM_FEC_SET_MODE_INET6
:
1299 b
->fec_btype
= FEC_BTYPE_INET6
;
1312 NG_RESPOND_MSG(error
, node
, item
, resp
);
1318 * Shutdown and remove the node and its associated interface.
1321 ng_fec_shutdown(node_p node
)
1323 const priv_p priv
= NG_NODE_PRIVATE(node
);
1324 struct ng_fec_bundle
*b
;
1325 struct ng_fec_portlist
*p
;
1327 b
= &priv
->fec_bundle
;
1328 ng_fec_stop(priv
->ifp
);
1330 while (!TAILQ_EMPTY(&b
->ng_fec_ports
)) {
1331 p
= TAILQ_FIRST(&b
->ng_fec_ports
);
1332 ng_fec_ether_cmdmulti(priv
->ifp
, p
, 0);
1333 ng_fec_delport(priv
, p
->fec_if
->if_xname
);
1336 ether_ifdetach(priv
->ifp
);
1337 if_free_type(priv
->ifp
, IFT_ETHER
);
1338 ifmedia_removeall(&priv
->ifmedia
);
1339 ng_fec_free_unit(priv
->unit
);
1340 FREE(priv
, M_NETGRAPH
);
1341 NG_NODE_SET_PRIVATE(node
, NULL
);
1342 NG_NODE_UNREF(node
);
1347 * Handle loading and unloading for this node type.
1350 ng_fec_mod_event(module_t mod
, int event
, void *data
)
1356 mtx_init(&ng_fec_mtx
, "ng_fec", NULL
, MTX_DEF
);
1359 mtx_destroy(&ng_fec_mtx
);