dslreports logo
 
    All Forums Hot Topics Gallery
spc
uniqs
13

rchandra
Stargate Universe fan
Premium Member
join:2000-11-09
14225-2105
ARRIS ONT1000GJ4
EnGenius EAP1250

2 recommendations

rchandra to Da Geek Kid

Premium Member

to Da Geek Kid

Re: [ipv6] why do cable modems need ipv6 support ?

First...ummmm....what? Sorry, but just throwing some words together does not necessarily a comprehensible sentence make.

But more to the point...Switches are not hubs. Hubs always forward all frames to all ports except the originator. Some hubs have collision control, but it's only of the form of shutting down a port which seems to be receiving too many frames per unit of time, sometimes until administratively unblocked, sometimes for a configured time interval (depends on the code in that hub and configuration).

Switches on the other hand can buffer frames in an attempt to avoid collisions. Also, your typical switch firmware doesn't care about ARP or any other protocol. All it cares about are the MAC addresses it has received and on what port that frame with that MAC address has arrived. Therefore, if the destination MAC address of a frame is in its memory (either discovered or configured), the frame is only forwarded to that particular port. If the port is not known, it falls back to operating as a hub, and floods the frame out to all ports (except the originator). If a switch is able to determine whether a link partner is half duplex, it will attempt to buffer frames destined for that port until that port is not transmitting in order to try to avoid collisions. It doesn't always succeed, because the link partner can always attempt to transmit at any time.

Some switches with more sophisticated firmware do look at higher layer protocols, and do something more intelligent with frames. For example, it could snoop on IGMP, and deliver multicast frames only to those ports which have joined some multicast group. It's only in that case where the frame content matters, and where any IPv4 or IPv6 compatibility comes into play. But even in that case, just as unknown destination MAC address switching devolves into hub behavior, if the switch does not recognize the upper level protocol (such as IPv6), it should just devolve into switching based on MAC address. Typically (but not always) these switches are of the managed type, the kind where one may log into them and configure them, or at the very least send them SNMP messages.

But...circling back to the question at hand (cable modem IPv6 compatibility), chances are cable modem firmware will not forward frames it cannot identify onto the radio, in order to try to protect the integrity of the cable system. As IPv4 has its own Ethertype ID, cable modems are probably not going to forward frames not having that Ethertype. I haven't read the specs all that deeply, but it also may be possible to use other frame types, such as PPPoE (although I'm going to guess DSL is pretty much the only place that's been used). So that's why I wrote in another post it's possible (though I would think unlikely) that IPv6 frames could be sent to the CMTS; the firmware will likely just block such frames from going out.

And for another poster: DOCSIS < 3 modems do not necessarily need to be replaced in order to be IPv6 compatible; it's possible all they need is a firmware update. It would all depend on whether it can be updated, whether it has the capacity (e.g. flash ROM space) to accept the necessary firmware, and whether something implemented in hardware (such as an Ethertype filter) would prevent such bridging/forwarding.

whfsdude
Premium Member
join:2003-04-05
Washington, DC

1 recommendation

whfsdude

Premium Member

said by rchandra:

For example, it could snoop on IGMP

or MLD snooping since this is the IPv6 forum.

DOCSIS < 3 modems do not necessarily need to be replaced in order to be IPv6 compatible; it's possible all they need is a firmware update. It would all depend on whether it can be updated, whether it has the capacity (e.g. flash ROM space) to accept the necessary firmware, and whether something implemented in hardware (such as an Ethertype filter) would prevent such bridging/forwarding.

or an ISP could make 6rd available for their legacy footprint. Of course you'd have to have customers pop a 6rd server into their router. AT&T has gone with this model for uverse rather than deploying all new CPE gear.

rchandra
Stargate Universe fan
Premium Member
join:2000-11-09
14225-2105
ARRIS ONT1000GJ4
EnGenius EAP1250

2 recommendations

rchandra

Premium Member

said by whfsdude:

or MLD snooping since this is the IPv6 forum.

aw, shucks, you caught me. I had seen something about that in my SMC6128 manual, but I didn't remember that when posting before. I had no idea what it was at the time I first saw it, so researched it a bit, just to know what that feature did. Thanks for reminding me/jogging my memory.
cramer
Premium Member
join:2007-04-10
Raleigh, NC
Westell 6100
Cisco PIX 501

1 recommendation

cramer to rchandra

Premium Member

to rchandra
(storm control -- excessive number of broadcasts, jabber -- causing an excessive number of collisions)

Most unmanaged switches are blindingly stupid. They care about two things... dst mac (first 6 bytes) and src mac (next 6 bytes). They don't even pay attention to the rest of the L2 frame. DST is used to find the port to forward the packet. SRC is used to learn what's on the port. This is not "arp snooping". This type of free learning is simple, fast, and done with minimal logic. (arp snooping is done by a CPU, which only exists in managed switches.)

It's only a small step (the next 2 bytes) to add protocol checking. For IP, that would be 0x0800 (IP) and 0x0806 (ARP.) And maybe PPPoE (0x8863, 0x8864), but I don't know of any US cable ISPs using PPPoE ('tho they did at one point, long ago.) IPv6 is 0x86dd. It's possible a D2 modem filters it out, but that's not something we users can test... if the headend isn't setup for IPv6, it doesn't matter if it gets past the modem or not.