xref: /freebsd/share/man/man4/multicast.4 (revision b3e7694832e81d7a904a10f525f8797b753bf0d3)
1.\" Copyright (c) 2001-2003 International Computer Science Institute
2.\"
3.\" Permission is hereby granted, free of charge, to any person obtaining a
4.\" copy of this software and associated documentation files (the "Software"),
5.\" to deal in the Software without restriction, including without limitation
6.\" the rights to use, copy, modify, merge, publish, distribute, sublicense,
7.\" and/or sell copies of the Software, and to permit persons to whom the
8.\" Software is furnished to do so, subject to the following conditions:
9.\"
10.\" The above copyright notice and this permission notice shall be included in
11.\" all copies or substantial portions of the Software.
12.\"
13.\" The names and trademarks of copyright holders may not be used in
14.\" advertising or publicity pertaining to the software without specific
15.\" prior permission. Title to copyright in this software and any associated
16.\" documentation will at all times remain with the copyright holders.
17.\"
18.\" THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19.\" IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20.\" FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
21.\" AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22.\" LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
23.\" FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
24.\" DEALINGS IN THE SOFTWARE.
25.\"
26.\" $FreeBSD$
27.\"
28.Dd May 27, 2009
29.Dt MULTICAST 4
30.Os
31.\"
32.Sh NAME
33.Nm multicast
34.Nd Multicast Routing
35.\"
36.Sh SYNOPSIS
37.Cd "options MROUTING"
38.Pp
39.In sys/types.h
40.In sys/socket.h
41.In netinet/in.h
42.In netinet/ip_mroute.h
43.In netinet6/ip6_mroute.h
44.Ft int
45.Fn getsockopt "int s" IPPROTO_IP MRT_INIT "void *optval" "socklen_t *optlen"
46.Ft int
47.Fn setsockopt "int s" IPPROTO_IP MRT_INIT "const void *optval" "socklen_t optlen"
48.Ft int
49.Fn getsockopt "int s" IPPROTO_IPV6 MRT6_INIT "void *optval" "socklen_t *optlen"
50.Ft int
51.Fn setsockopt "int s" IPPROTO_IPV6 MRT6_INIT "const void *optval" "socklen_t optlen"
52.Sh DESCRIPTION
53.Tn "Multicast routing"
54is used to efficiently propagate data
55packets to a set of multicast listeners in multipoint networks.
56If unicast is used to replicate the data to all listeners,
57then some of the network links may carry multiple copies of the same
58data packets.
59With multicast routing, the overhead is reduced to one copy
60(at most) per network link.
61.Pp
62All multicast-capable routers must run a common multicast routing
63protocol.
64It is recommended that either
65Protocol Independent Multicast - Sparse Mode (PIM-SM),
66or Protocol Independent Multicast - Dense Mode (PIM-DM)
67are used, as these are now the generally accepted protocols
68in the Internet community.
69The
70.Sx HISTORY
71section discusses previous multicast routing protocols.
72.Pp
73To start multicast routing,
74the user must enable multicast forwarding in the kernel
75(see
76.Sx SYNOPSIS
77about the kernel configuration options),
78and must run a multicast routing capable user-level process.
79From developer's point of view,
80the programming guide described in the
81.Sx "Programming Guide"
82section should be used to control the multicast forwarding in the kernel.
83.\"
84.Ss Programming Guide
85This section provides information about the basic multicast routing API.
86The so-called
87.Dq advanced multicast API
88is described in the
89.Sx "Advanced Multicast API Programming Guide"
90section.
91.Pp
92First, a multicast routing socket must be open.
93That socket would be used
94to control the multicast forwarding in the kernel.
95Note that most operations below require certain privilege
96(i.e., root privilege):
97.Bd -literal
98/* IPv4 */
99int mrouter_s4;
100mrouter_s4 = socket(AF_INET, SOCK_RAW, IPPROTO_IGMP);
101.Ed
102.Bd -literal
103int mrouter_s6;
104mrouter_s6 = socket(AF_INET6, SOCK_RAW, IPPROTO_ICMPV6);
105.Ed
106.Pp
107Note that if the router needs to open an IGMP or ICMPv6 socket
108(in case of IPv4 and IPv6 respectively)
109for sending or receiving of IGMP or MLD multicast group membership messages,
110then the same
111.Va mrouter_s4
112or
113.Va mrouter_s6
114sockets should be used
115for sending and receiving respectively IGMP or MLD messages.
116In case of
117.Bx Ns
118-derived kernel, it may be possible to open separate sockets
119for IGMP or MLD messages only.
120However, some other kernels (e.g.,
121.Tn Linux )
122require that the multicast
123routing socket must be used for sending and receiving of IGMP or MLD
124messages.
125Therefore, for portability reason the multicast
126routing socket should be reused for IGMP and MLD messages as well.
127.Pp
128After the multicast routing socket is open, it can be used to enable
129or disable multicast forwarding in the kernel:
130.Bd -literal
131/* IPv4 */
132int v = 1;        /* 1 to enable, or 0 to disable */
133setsockopt(mrouter_s4, IPPROTO_IP, MRT_INIT, (void *)&v, sizeof(v));
134.Ed
135.Bd -literal
136/* IPv6 */
137int v = 1;        /* 1 to enable, or 0 to disable */
138setsockopt(mrouter_s6, IPPROTO_IPV6, MRT6_INIT, (void *)&v, sizeof(v));
139\&...
140/* If necessary, filter all ICMPv6 messages */
141struct icmp6_filter filter;
142ICMP6_FILTER_SETBLOCKALL(&filter);
143setsockopt(mrouter_s6, IPPROTO_ICMPV6, ICMP6_FILTER, (void *)&filter,
144           sizeof(filter));
145.Ed
146.Pp
147After multicast forwarding is enabled, the multicast routing socket
148can be used to enable PIM processing in the kernel if we are running PIM-SM or
149PIM-DM
150(see
151.Xr pim 4 ) .
152.Pp
153For each network interface (e.g., physical or a virtual tunnel)
154that would be used for multicast forwarding, a corresponding
155multicast interface must be added to the kernel:
156.Bd -literal
157/* IPv4 */
158struct vifctl vc;
159memset(&vc, 0, sizeof(vc));
160/* Assign all vifctl fields as appropriate */
161vc.vifc_vifi = vif_index;
162vc.vifc_flags = vif_flags;
163vc.vifc_threshold = min_ttl_threshold;
164vc.vifc_rate_limit = 0;
165memcpy(&vc.vifc_lcl_addr, &vif_local_address, sizeof(vc.vifc_lcl_addr));
166setsockopt(mrouter_s4, IPPROTO_IP, MRT_ADD_VIF, (void *)&vc,
167           sizeof(vc));
168.Ed
169.Pp
170The
171.Va vif_index
172must be unique per vif.
173The
174.Va vif_flags
175contains the
176.Dv VIFF_*
177flags as defined in
178.In netinet/ip_mroute.h .
179The
180.Dv VIFF_TUNNEL
181flag is no longer supported by
182.Fx .
183Users who wish to forward multicast datagrams over a tunnel should consider
184configuring a
185.Xr gif 4
186or
187.Xr gre 4
188tunnel and using it as a physical interface.
189.Pp
190The
191.Va min_ttl_threshold
192contains the minimum TTL a multicast data packet must have to be
193forwarded on that vif.
194Typically, it would have value of 1.
195.Pp
196The
197.Va max_rate_limit
198argument is no longer supported in
199.Fx
200and should be set to 0.
201Users who wish to rate-limit multicast datagrams should consider the use of
202.Xr dummynet 4
203or
204.Xr altq 4 .
205.Pp
206The
207.Va vif_local_address
208contains the local IP address of the corresponding local interface.
209The
210.Va vif_remote_address
211contains the remote IP address in case of DVMRP multicast tunnels.
212.Bd -literal
213/* IPv6 */
214struct mif6ctl mc;
215memset(&mc, 0, sizeof(mc));
216/* Assign all mif6ctl fields as appropriate */
217mc.mif6c_mifi = mif_index;
218mc.mif6c_flags = mif_flags;
219mc.mif6c_pifi = pif_index;
220setsockopt(mrouter_s6, IPPROTO_IPV6, MRT6_ADD_MIF, (void *)&mc,
221           sizeof(mc));
222.Ed
223.Pp
224The
225.Va mif_index
226must be unique per vif.
227The
228.Va mif_flags
229contains the
230.Dv MIFF_*
231flags as defined in
232.In netinet6/ip6_mroute.h .
233The
234.Va pif_index
235is the physical interface index of the corresponding local interface.
236.Pp
237A multicast interface is deleted by:
238.Bd -literal
239/* IPv4 */
240vifi_t vifi = vif_index;
241setsockopt(mrouter_s4, IPPROTO_IP, MRT_DEL_VIF, (void *)&vifi,
242           sizeof(vifi));
243.Ed
244.Bd -literal
245/* IPv6 */
246mifi_t mifi = mif_index;
247setsockopt(mrouter_s6, IPPROTO_IPV6, MRT6_DEL_MIF, (void *)&mifi,
248           sizeof(mifi));
249.Ed
250.Pp
251After the multicast forwarding is enabled, and the multicast virtual
252interfaces are
253added, the kernel may deliver upcall messages (also called signals
254later in this text) on the multicast routing socket that was open
255earlier with
256.Dv MRT_INIT
257or
258.Dv MRT6_INIT .
259The IPv4 upcalls have
260.Vt "struct igmpmsg"
261header (see
262.In netinet/ip_mroute.h )
263with field
264.Va im_mbz
265set to zero.
266Note that this header follows the structure of
267.Vt "struct ip"
268with the protocol field
269.Va ip_p
270set to zero.
271The IPv6 upcalls have
272.Vt "struct mrt6msg"
273header (see
274.In netinet6/ip6_mroute.h )
275with field
276.Va im6_mbz
277set to zero.
278Note that this header follows the structure of
279.Vt "struct ip6_hdr"
280with the next header field
281.Va ip6_nxt
282set to zero.
283.Pp
284The upcall header contains field
285.Va im_msgtype
286and
287.Va im6_msgtype
288with the type of the upcall
289.Dv IGMPMSG_*
290and
291.Dv MRT6MSG_*
292for IPv4 and IPv6 respectively.
293The values of the rest of the upcall header fields
294and the body of the upcall message depend on the particular upcall type.
295.Pp
296If the upcall message type is
297.Dv IGMPMSG_NOCACHE
298or
299.Dv MRT6MSG_NOCACHE ,
300this is an indication that a multicast packet has reached the multicast
301router, but the router has no forwarding state for that packet.
302Typically, the upcall would be a signal for the multicast routing
303user-level process to install the appropriate Multicast Forwarding
304Cache (MFC) entry in the kernel.
305.Pp
306An MFC entry is added by:
307.Bd -literal
308/* IPv4 */
309struct mfcctl mc;
310memset(&mc, 0, sizeof(mc));
311memcpy(&mc.mfcc_origin, &source_addr, sizeof(mc.mfcc_origin));
312memcpy(&mc.mfcc_mcastgrp, &group_addr, sizeof(mc.mfcc_mcastgrp));
313mc.mfcc_parent = iif_index;
314for (i = 0; i < maxvifs; i++)
315    mc.mfcc_ttls[i] = oifs_ttl[i];
316setsockopt(mrouter_s4, IPPROTO_IP, MRT_ADD_MFC,
317           (void *)&mc, sizeof(mc));
318.Ed
319.Bd -literal
320/* IPv6 */
321struct mf6cctl mc;
322memset(&mc, 0, sizeof(mc));
323memcpy(&mc.mf6cc_origin, &source_addr, sizeof(mc.mf6cc_origin));
324memcpy(&mc.mf6cc_mcastgrp, &group_addr, sizeof(mf6cc_mcastgrp));
325mc.mf6cc_parent = iif_index;
326for (i = 0; i < maxvifs; i++)
327    if (oifs_ttl[i] > 0)
328        IF_SET(i, &mc.mf6cc_ifset);
329setsockopt(mrouter_s6, IPPROTO_IPV6, MRT6_ADD_MFC,
330           (void *)&mc, sizeof(mc));
331.Ed
332.Pp
333The
334.Va source_addr
335and
336.Va group_addr
337are the source and group address of the multicast packet (as set
338in the upcall message).
339The
340.Va iif_index
341is the virtual interface index of the multicast interface the multicast
342packets for this specific source and group address should be received on.
343The
344.Va oifs_ttl[]
345array contains the minimum TTL (per interface) a multicast packet
346should have to be forwarded on an outgoing interface.
347If the TTL value is zero, the corresponding interface is not included
348in the set of outgoing interfaces.
349Note that in case of IPv6 only the set of outgoing interfaces can
350be specified.
351.Pp
352An MFC entry is deleted by:
353.Bd -literal
354/* IPv4 */
355struct mfcctl mc;
356memset(&mc, 0, sizeof(mc));
357memcpy(&mc.mfcc_origin, &source_addr, sizeof(mc.mfcc_origin));
358memcpy(&mc.mfcc_mcastgrp, &group_addr, sizeof(mc.mfcc_mcastgrp));
359setsockopt(mrouter_s4, IPPROTO_IP, MRT_DEL_MFC,
360           (void *)&mc, sizeof(mc));
361.Ed
362.Bd -literal
363/* IPv6 */
364struct mf6cctl mc;
365memset(&mc, 0, sizeof(mc));
366memcpy(&mc.mf6cc_origin, &source_addr, sizeof(mc.mf6cc_origin));
367memcpy(&mc.mf6cc_mcastgrp, &group_addr, sizeof(mf6cc_mcastgrp));
368setsockopt(mrouter_s6, IPPROTO_IPV6, MRT6_DEL_MFC,
369           (void *)&mc, sizeof(mc));
370.Ed
371.Pp
372The following method can be used to get various statistics per
373installed MFC entry in the kernel (e.g., the number of forwarded
374packets per source and group address):
375.Bd -literal
376/* IPv4 */
377struct sioc_sg_req sgreq;
378memset(&sgreq, 0, sizeof(sgreq));
379memcpy(&sgreq.src, &source_addr, sizeof(sgreq.src));
380memcpy(&sgreq.grp, &group_addr, sizeof(sgreq.grp));
381ioctl(mrouter_s4, SIOCGETSGCNT, &sgreq);
382.Ed
383.Bd -literal
384/* IPv6 */
385struct sioc_sg_req6 sgreq;
386memset(&sgreq, 0, sizeof(sgreq));
387memcpy(&sgreq.src, &source_addr, sizeof(sgreq.src));
388memcpy(&sgreq.grp, &group_addr, sizeof(sgreq.grp));
389ioctl(mrouter_s6, SIOCGETSGCNT_IN6, &sgreq);
390.Ed
391.Pp
392The following method can be used to get various statistics per
393multicast virtual interface in the kernel (e.g., the number of forwarded
394packets per interface):
395.Bd -literal
396/* IPv4 */
397struct sioc_vif_req vreq;
398memset(&vreq, 0, sizeof(vreq));
399vreq.vifi = vif_index;
400ioctl(mrouter_s4, SIOCGETVIFCNT, &vreq);
401.Ed
402.Bd -literal
403/* IPv6 */
404struct sioc_mif_req6 mreq;
405memset(&mreq, 0, sizeof(mreq));
406mreq.mifi = vif_index;
407ioctl(mrouter_s6, SIOCGETMIFCNT_IN6, &mreq);
408.Ed
409.Ss Advanced Multicast API Programming Guide
410If we want to add new features in the kernel, it becomes difficult
411to preserve backward compatibility (binary and API),
412and at the same time to allow user-level processes to take advantage of
413the new features (if the kernel supports them).
414.Pp
415One of the mechanisms that allows us to preserve the backward
416compatibility is a sort of negotiation
417between the user-level process and the kernel:
418.Bl -enum
419.It
420The user-level process tries to enable in the kernel the set of new
421features (and the corresponding API) it would like to use.
422.It
423The kernel returns the (sub)set of features it knows about
424and is willing to be enabled.
425.It
426The user-level process uses only that set of features
427the kernel has agreed on.
428.El
429.\"
430.Pp
431To support backward compatibility, if the user-level process does not
432ask for any new features, the kernel defaults to the basic
433multicast API (see the
434.Sx "Programming Guide"
435section).
436.\" XXX: edit as appropriate after the advanced multicast API is
437.\" supported under IPv6
438Currently, the advanced multicast API exists only for IPv4;
439in the future there will be IPv6 support as well.
440.Pp
441Below is a summary of the expandable API solution.
442Note that all new options and structures are defined
443in
444.In netinet/ip_mroute.h
445and
446.In netinet6/ip6_mroute.h ,
447unless stated otherwise.
448.Pp
449The user-level process uses new
450.Fn getsockopt Ns / Ns Fn setsockopt
451options to
452perform the API features negotiation with the kernel.
453This negotiation must be performed right after the multicast routing
454socket is open.
455The set of desired/allowed features is stored in a bitset
456(currently, in
457.Vt uint32_t ;
458i.e., maximum of 32 new features).
459The new
460.Fn getsockopt Ns / Ns Fn setsockopt
461options are
462.Dv MRT_API_SUPPORT
463and
464.Dv MRT_API_CONFIG .
465Example:
466.Bd -literal
467uint32_t v;
468getsockopt(sock, IPPROTO_IP, MRT_API_SUPPORT, (void *)&v, sizeof(v));
469.Ed
470.Pp
471would set in
472.Va v
473the pre-defined bits that the kernel API supports.
474The eight least significant bits in
475.Vt uint32_t
476are same as the
477eight possible flags
478.Dv MRT_MFC_FLAGS_*
479that can be used in
480.Va mfcc_flags
481as part of the new definition of
482.Vt "struct mfcctl"
483(see below about those flags), which leaves 24 flags for other new features.
484The value returned by
485.Fn getsockopt MRT_API_SUPPORT
486is read-only; in other words,
487.Fn setsockopt MRT_API_SUPPORT
488would fail.
489.Pp
490To modify the API, and to set some specific feature in the kernel, then:
491.Bd -literal
492uint32_t v = MRT_MFC_FLAGS_DISABLE_WRONGVIF;
493if (setsockopt(sock, IPPROTO_IP, MRT_API_CONFIG, (void *)&v, sizeof(v))
494    != 0) {
495    return (ERROR);
496}
497if (v & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
498    return (OK);	/* Success */
499else
500    return (ERROR);
501.Ed
502.Pp
503In other words, when
504.Fn setsockopt MRT_API_CONFIG
505is called, the
506argument to it specifies the desired set of features to
507be enabled in the API and the kernel.
508The return value in
509.Va v
510is the actual (sub)set of features that were enabled in the kernel.
511To obtain later the same set of features that were enabled, then:
512.Bd -literal
513getsockopt(sock, IPPROTO_IP, MRT_API_CONFIG, (void *)&v, sizeof(v));
514.Ed
515.Pp
516The set of enabled features is global.
517In other words,
518.Fn setsockopt MRT_API_CONFIG
519should be called right after
520.Fn setsockopt MRT_INIT .
521.Pp
522Currently, the following set of new features is defined:
523.Bd -literal
524#define	MRT_MFC_FLAGS_DISABLE_WRONGVIF (1 << 0) /* disable WRONGVIF signals */
525#define	MRT_MFC_FLAGS_BORDER_VIF   (1 << 1)  /* border vif              */
526#define MRT_MFC_RP                 (1 << 8)  /* enable RP address	*/
527#define MRT_MFC_BW_UPCALL          (1 << 9)  /* enable bw upcalls	*/
528.Ed
529.\" .Pp
530.\" In the future there might be:
531.\" .Bd -literal
532.\" #define MRT_MFC_GROUP_SPECIFIC     (1 << 10) /* allow (*,G) MFC entries */
533.\" .Ed
534.\" .Pp
535.\" to allow (*,G) MFC entries (i.e., group-specific entries) in the kernel.
536.\" For now this is left-out until it is clear whether
537.\" (*,G) MFC support is the preferred solution instead of something more generic
538.\" solution for example.
539.\"
540.\" 2. The newly defined struct mfcctl2.
541.\"
542.Pp
543The advanced multicast API uses a newly defined
544.Vt "struct mfcctl2"
545instead of the traditional
546.Vt "struct mfcctl" .
547The original
548.Vt "struct mfcctl"
549is kept as is.
550The new
551.Vt "struct mfcctl2"
552is:
553.Bd -literal
554/*
555 * The new argument structure for MRT_ADD_MFC and MRT_DEL_MFC overlays
556 * and extends the old struct mfcctl.
557 */
558struct mfcctl2 {
559        /* the mfcctl fields */
560        struct in_addr  mfcc_origin;       /* ip origin of mcasts       */
561        struct in_addr  mfcc_mcastgrp;     /* multicast group associated*/
562        vifi_t          mfcc_parent;       /* incoming vif              */
563        u_char          mfcc_ttls[MAXVIFS];/* forwarding ttls on vifs   */
564
565        /* extension fields */
566        uint8_t         mfcc_flags[MAXVIFS];/* the MRT_MFC_FLAGS_* flags*/
567        struct in_addr  mfcc_rp;            /* the RP address           */
568};
569.Ed
570.Pp
571The new fields are
572.Va mfcc_flags[MAXVIFS]
573and
574.Va mfcc_rp .
575Note that for compatibility reasons they are added at the end.
576.Pp
577The
578.Va mfcc_flags[MAXVIFS]
579field is used to set various flags per
580interface per (S,G) entry.
581Currently, the defined flags are:
582.Bd -literal
583#define	MRT_MFC_FLAGS_DISABLE_WRONGVIF (1 << 0) /* disable WRONGVIF signals */
584#define	MRT_MFC_FLAGS_BORDER_VIF       (1 << 1) /* border vif          */
585.Ed
586.Pp
587The
588.Dv MRT_MFC_FLAGS_DISABLE_WRONGVIF
589flag is used to explicitly disable the
590.Dv IGMPMSG_WRONGVIF
591kernel signal at the (S,G) granularity if a multicast data packet
592arrives on the wrong interface.
593Usually, this signal is used to
594complete the shortest-path switch in case of PIM-SM multicast routing,
595or to trigger a PIM assert message.
596However, it should not be delivered for interfaces that are not in
597the outgoing interface set, and that are not expecting to
598become an incoming interface.
599Hence, if the
600.Dv MRT_MFC_FLAGS_DISABLE_WRONGVIF
601flag is set for some of the
602interfaces, then a data packet that arrives on that interface for
603that MFC entry will NOT trigger a WRONGVIF signal.
604If that flag is not set, then a signal is triggered (the default action).
605.Pp
606The
607.Dv MRT_MFC_FLAGS_BORDER_VIF
608flag is used to specify whether the Border-bit in PIM
609Register messages should be set (in case when the Register encapsulation
610is performed inside the kernel).
611If it is set for the special PIM Register kernel virtual interface
612(see
613.Xr pim 4 ) ,
614the Border-bit in the Register messages sent to the RP will be set.
615.Pp
616The remaining six bits are reserved for future usage.
617.Pp
618The
619.Va mfcc_rp
620field is used to specify the RP address (in case of PIM-SM multicast routing)
621for a multicast
622group G if we want to perform kernel-level PIM Register encapsulation.
623The
624.Va mfcc_rp
625field is used only if the
626.Dv MRT_MFC_RP
627advanced API flag/capability has been successfully set by
628.Fn setsockopt MRT_API_CONFIG .
629.Pp
630.\"
631.\" 3. Kernel-level PIM Register encapsulation
632.\"
633If the
634.Dv MRT_MFC_RP
635flag was successfully set by
636.Fn setsockopt MRT_API_CONFIG ,
637then the kernel will attempt to perform
638the PIM Register encapsulation itself instead of sending the
639multicast data packets to user level (inside
640.Dv IGMPMSG_WHOLEPKT
641upcalls) for user-level encapsulation.
642The RP address would be taken from the
643.Va mfcc_rp
644field
645inside the new
646.Vt "struct mfcctl2" .
647However, even if the
648.Dv MRT_MFC_RP
649flag was successfully set, if the
650.Va mfcc_rp
651field was set to
652.Dv INADDR_ANY ,
653then the
654kernel will still deliver an
655.Dv IGMPMSG_WHOLEPKT
656upcall with the
657multicast data packet to the user-level process.
658.Pp
659In addition, if the multicast data packet is too large to fit within
660a single IP packet after the PIM Register encapsulation (e.g., if
661its size was on the order of 65500 bytes), the data packet will be
662fragmented, and then each of the fragments will be encapsulated
663separately.
664Note that typically a multicast data packet can be that
665large only if it was originated locally from the same hosts that
666performs the encapsulation; otherwise the transmission of the
667multicast data packet over Ethernet for example would have
668fragmented it into much smaller pieces.
669.\"
670.\" Note that if this code is ported to IPv6, we may need the kernel to
671.\" perform MTU discovery to the RP, and keep those discoveries inside
672.\" the kernel so the encapsulating router may send back ICMP
673.\" Fragmentation Required if the size of the multicast data packet is
674.\" too large (see "Encapsulating data packets in the Register Tunnel"
675.\" in Section 4.4.1 in the PIM-SM spec
676.\" draft-ietf-pim-sm-v2-new-05.{txt,ps}).
677.\" For IPv4 we may be able to get away without it, but for IPv6 we need
678.\" that.
679.\"
680.\" 4. Mechanism for "multicast bandwidth monitoring and upcalls".
681.\"
682.Pp
683Typically, a multicast routing user-level process would need to know the
684forwarding bandwidth for some data flow.
685For example, the multicast routing process may want to timeout idle MFC
686entries, or in case of PIM-SM it can initiate (S,G) shortest-path switch if
687the bandwidth rate is above a threshold for example.
688.Pp
689The original solution for measuring the bandwidth of a dataflow was
690that a user-level process would periodically
691query the kernel about the number of forwarded packets/bytes per
692(S,G), and then based on those numbers it would estimate whether a source
693has been idle, or whether the source's transmission bandwidth is above a
694threshold.
695That solution is far from being scalable, hence the need for a new
696mechanism for bandwidth monitoring.
697.Pp
698Below is a description of the bandwidth monitoring mechanism.
699.Bl -bullet
700.It
701If the bandwidth of a data flow satisfies some pre-defined filter,
702the kernel delivers an upcall on the multicast routing socket
703to the multicast routing process that has installed that filter.
704.It
705The bandwidth-upcall filters are installed per (S,G).
706There can be
707more than one filter per (S,G).
708.It
709Instead of supporting all possible comparison operations
710(i.e., < <= == != > >= ), there is support only for the
711<= and >= operations,
712because this makes the kernel-level implementation simpler,
713and because practically we need only those two.
714Further, the missing operations can be simulated by secondary
715user-level filtering of those <= and >= filters.
716For example, to simulate !=, then we need to install filter
717.Dq bw <= 0xffffffff ,
718and after an
719upcall is received, we need to check whether
720.Dq measured_bw != expected_bw .
721.It
722The bandwidth-upcall mechanism is enabled by
723.Fn setsockopt MRT_API_CONFIG
724for the
725.Dv MRT_MFC_BW_UPCALL
726flag.
727.It
728The bandwidth-upcall filters are added/deleted by the new
729.Fn setsockopt MRT_ADD_BW_UPCALL
730and
731.Fn setsockopt MRT_DEL_BW_UPCALL
732respectively (with the appropriate
733.Vt "struct bw_upcall"
734argument of course).
735.El
736.Pp
737From application point of view, a developer needs to know about
738the following:
739.Bd -literal
740/*
741 * Structure for installing or delivering an upcall if the
742 * measured bandwidth is above or below a threshold.
743 *
744 * User programs (e.g. daemons) may have a need to know when the
745 * bandwidth used by some data flow is above or below some threshold.
746 * This interface allows the userland to specify the threshold (in
747 * bytes and/or packets) and the measurement interval. Flows are
748 * all packet with the same source and destination IP address.
749 * At the moment the code is only used for multicast destinations
750 * but there is nothing that prevents its use for unicast.
751 *
752 * The measurement interval cannot be shorter than some Tmin (currently, 3s).
753 * The threshold is set in packets and/or bytes per_interval.
754 *
755 * Measurement works as follows:
756 *
757 * For >= measurements:
758 * The first packet marks the start of a measurement interval.
759 * During an interval we count packets and bytes, and when we
760 * pass the threshold we deliver an upcall and we are done.
761 * The first packet after the end of the interval resets the
762 * count and restarts the measurement.
763 *
764 * For <= measurement:
765 * We start a timer to fire at the end of the interval, and
766 * then for each incoming packet we count packets and bytes.
767 * When the timer fires, we compare the value with the threshold,
768 * schedule an upcall if we are below, and restart the measurement
769 * (reschedule timer and zero counters).
770 */
771
772struct bw_data {
773        struct timeval  b_time;
774        uint64_t        b_packets;
775        uint64_t        b_bytes;
776};
777
778struct bw_upcall {
779        struct in_addr  bu_src;         /* source address            */
780        struct in_addr  bu_dst;         /* destination address       */
781        uint32_t        bu_flags;       /* misc flags (see below)    */
782#define BW_UPCALL_UNIT_PACKETS (1 << 0) /* threshold (in packets)    */
783#define BW_UPCALL_UNIT_BYTES   (1 << 1) /* threshold (in bytes)      */
784#define BW_UPCALL_GEQ          (1 << 2) /* upcall if bw >= threshold */
785#define BW_UPCALL_LEQ          (1 << 3) /* upcall if bw <= threshold */
786#define BW_UPCALL_DELETE_ALL   (1 << 4) /* delete all upcalls for s,d*/
787        struct bw_data  bu_threshold;   /* the bw threshold          */
788        struct bw_data  bu_measured;    /* the measured bw           */
789};
790
791/* max. number of upcalls to deliver together */
792#define BW_UPCALLS_MAX				128
793/* min. threshold time interval for bandwidth measurement */
794#define BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC	3
795#define BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC	0
796.Ed
797.Pp
798The
799.Vt bw_upcall
800structure is used as an argument to
801.Fn setsockopt MRT_ADD_BW_UPCALL
802and
803.Fn setsockopt MRT_DEL_BW_UPCALL .
804Each
805.Fn setsockopt MRT_ADD_BW_UPCALL
806installs a filter in the kernel
807for the source and destination address in the
808.Vt bw_upcall
809argument,
810and that filter will trigger an upcall according to the following
811pseudo-algorithm:
812.Bd -literal
813 if (bw_upcall_oper IS ">=") {
814    if (((bw_upcall_unit & PACKETS == PACKETS) &&
815         (measured_packets >= threshold_packets)) ||
816        ((bw_upcall_unit & BYTES == BYTES) &&
817         (measured_bytes >= threshold_bytes)))
818       SEND_UPCALL("measured bandwidth is >= threshold");
819  }
820  if (bw_upcall_oper IS "<=" && measured_interval >= threshold_interval) {
821    if (((bw_upcall_unit & PACKETS == PACKETS) &&
822         (measured_packets <= threshold_packets)) ||
823        ((bw_upcall_unit & BYTES == BYTES) &&
824         (measured_bytes <= threshold_bytes)))
825       SEND_UPCALL("measured bandwidth is <= threshold");
826  }
827.Ed
828.Pp
829In the same
830.Vt bw_upcall
831the unit can be specified in both BYTES and PACKETS.
832However, the GEQ and LEQ flags are mutually exclusive.
833.Pp
834Basically, an upcall is delivered if the measured bandwidth is >= or
835<= the threshold bandwidth (within the specified measurement
836interval).
837For practical reasons, the smallest value for the measurement
838interval is 3 seconds.
839If smaller values are allowed, then the bandwidth
840estimation may be less accurate, or the potentially very high frequency
841of the generated upcalls may introduce too much overhead.
842For the >= operation, the answer may be known before the end of
843.Va threshold_interval ,
844therefore the upcall may be delivered earlier.
845For the <= operation however, we must wait
846until the threshold interval has expired to know the answer.
847.Pp
848Example of usage:
849.Bd -literal
850struct bw_upcall bw_upcall;
851/* Assign all bw_upcall fields as appropriate */
852memset(&bw_upcall, 0, sizeof(bw_upcall));
853memcpy(&bw_upcall.bu_src, &source, sizeof(bw_upcall.bu_src));
854memcpy(&bw_upcall.bu_dst, &group, sizeof(bw_upcall.bu_dst));
855bw_upcall.bu_threshold.b_data = threshold_interval;
856bw_upcall.bu_threshold.b_packets = threshold_packets;
857bw_upcall.bu_threshold.b_bytes = threshold_bytes;
858if (is_threshold_in_packets)
859    bw_upcall.bu_flags |= BW_UPCALL_UNIT_PACKETS;
860if (is_threshold_in_bytes)
861    bw_upcall.bu_flags |= BW_UPCALL_UNIT_BYTES;
862do {
863    if (is_geq_upcall) {
864        bw_upcall.bu_flags |= BW_UPCALL_GEQ;
865        break;
866    }
867    if (is_leq_upcall) {
868        bw_upcall.bu_flags |= BW_UPCALL_LEQ;
869        break;
870    }
871    return (ERROR);
872} while (0);
873setsockopt(mrouter_s4, IPPROTO_IP, MRT_ADD_BW_UPCALL,
874          (void *)&bw_upcall, sizeof(bw_upcall));
875.Ed
876.Pp
877To delete a single filter, then use
878.Dv MRT_DEL_BW_UPCALL ,
879and the fields of bw_upcall must be set
880exactly same as when
881.Dv MRT_ADD_BW_UPCALL
882was called.
883.Pp
884To delete all bandwidth filters for a given (S,G), then
885only the
886.Va bu_src
887and
888.Va bu_dst
889fields in
890.Vt "struct bw_upcall"
891need to be set, and then just set only the
892.Dv BW_UPCALL_DELETE_ALL
893flag inside field
894.Va bw_upcall.bu_flags .
895.Pp
896The bandwidth upcalls are received by aggregating them in the new upcall
897message:
898.Bd -literal
899#define IGMPMSG_BW_UPCALL  4  /* BW monitoring upcall */
900.Ed
901.Pp
902This message is an array of
903.Vt "struct bw_upcall"
904elements (up to
905.Dv BW_UPCALLS_MAX
906= 128).
907The upcalls are
908delivered when there are 128 pending upcalls, or when 1 second has
909expired since the previous upcall (whichever comes first).
910In an
911.Vt "struct upcall"
912element, the
913.Va bu_measured
914field is filled-in to
915indicate the particular measured values.
916However, because of the way
917the particular intervals are measured, the user should be careful how
918.Va bu_measured.b_time
919is used.
920For example, if the
921filter is installed to trigger an upcall if the number of packets
922is >= 1, then
923.Va bu_measured
924may have a value of zero in the upcalls after the
925first one, because the measured interval for >= filters is
926.Dq clocked
927by the forwarded packets.
928Hence, this upcall mechanism should not be used for measuring
929the exact value of the bandwidth of the forwarded data.
930To measure the exact bandwidth, the user would need to
931get the forwarded packets statistics with the
932.Fn ioctl SIOCGETSGCNT
933mechanism
934(see the
935.Sx Programming Guide
936section) .
937.Pp
938Note that the upcalls for a filter are delivered until the specific
939filter is deleted, but no more frequently than once per
940.Va bu_threshold.b_time .
941For example, if the filter is specified to
942deliver a signal if bw >= 1 packet, the first packet will trigger a
943signal, but the next upcall will be triggered no earlier than
944.Va bu_threshold.b_time
945after the previous upcall.
946.\"
947.Sh SEE ALSO
948.Xr getsockopt 2 ,
949.Xr recvfrom 2 ,
950.Xr recvmsg 2 ,
951.Xr setsockopt 2 ,
952.Xr socket 2 ,
953.Xr sourcefilter 3 ,
954.Xr altq 4 ,
955.Xr dummynet 4 ,
956.Xr gif 4 ,
957.Xr gre 4 ,
958.Xr icmp6 4 ,
959.Xr igmp 4 ,
960.Xr inet 4 ,
961.Xr inet6 4 ,
962.Xr intro 4 ,
963.Xr ip 4 ,
964.Xr ip6 4 ,
965.Xr mld 4 ,
966.Xr pim 4
967.\"
968.Sh HISTORY
969The Distance Vector Multicast Routing Protocol (DVMRP)
970was the first developed multicast routing protocol.
971Later, other protocols such as Multicast Extensions to OSPF (MOSPF)
972and Core Based Trees (CBT), were developed as well.
973Routers at autonomous system boundaries may now exchange multicast
974routes with peers via the Border Gateway Protocol (BGP).
975Many other routing protocols are able to redistribute multicast routes
976for use with
977.Dv PIM-SM
978and
979.Dv PIM-DM .
980.Sh AUTHORS
981.An -nosplit
982The original multicast code was written by
983.An David Waitzman
984(BBN Labs),
985and later modified by the following individuals:
986.An Steve Deering
987(Stanford),
988.An Mark J. Steiglitz
989(Stanford),
990.An Van Jacobson
991(LBL),
992.An Ajit Thyagarajan
993(PARC),
994.An Bill Fenner
995(PARC).
996The IPv6 multicast support was implemented by the KAME project
997.Pq Pa https://www.kame.net ,
998and was based on the IPv4 multicast code.
999The advanced multicast API and the multicast bandwidth
1000monitoring were implemented by
1001.An Pavlin Radoslavov
1002(ICSI)
1003in collaboration with
1004.An Chris Brown
1005(NextHop).
1006The IGMPv3 and MLDv2 multicast support was implemented by
1007.An Bruce Simpson .
1008.Pp
1009This manual page was written by
1010.An Pavlin Radoslavov
1011(ICSI).
1012