On Sun, 28 Feb 2016, Johnny Billquist wrote:
You are right. 802.2 specifies other details of the packet, but not
addressing. On "ethernets", it would be a 802.3 frame that contained the
802.2 packet. (Sorry about my mixing up here.)
However, as far as I could figure out, 802.11 also uses 802.2 packets inside
the frame.
I refreshed my memory and 802.11 indeed uses 802.2 LLC+SNAP.
And 802.11 do not look like an 802.3 frame either, so it don't really make
sense to talk about 802.3 here, I'd say. But let's say that the two last
addresses in the 802.11 frame are the ones then used to recreate a 802.3
frame, which holds a 802.2 packet (ugh).
I guess it matches my observation, that the destination mac address is messed
up on packets sent from the station.
No, the destination address should come out correctly. It is the source
(from STA->DS) that is probably mangled at the recipient. In the other
direction (DS->STA).
In fact, the crux is precisely the slight mismatch between the 802.11
address scopes and the 802.3 address scopes and how the frames are
translated. Only the DS side was originally expected to have 802.3 frames,
and the STA side was expected to only have a network stack that only
claimed one MAC address as its own.
In the STA->AP (STA->DS) direction the 802.3 frame is created by taking
the second 802.11 MAC address as the source address and the third as the
destination.
In the AP->STA (DS->STA) direction the AP does not include the original
802.3 destination address in the 802.11 MAC header since there is no room
for it. As there is no way for a single STA to present more then one
source address there is also no way for the AP to learn about more clients
behind the STA. Once STA == one client == one 802.3 destination MAC.
With the 4-address format the 802.11 link can be transparent to the 802.3
frames, since all elements of the 802.3 MAC header have equivalent and
reversible representations in the 802.11 header.
And I just introduced the DS term. DS is the wire-side (802.3) of an AP.
For the LLC header the translation from 802.3->802.11 either the LLC from
the 802.3 frame is taken directly or the EtherType is translated into a
LLC+SNAP header, which is just a longer way of conveying the same
information these days. In the other direction I think the LLC+SNAP is
mostly stipped and converted to raw EtherType. At least I usually do not
see LLC+SNAP headers on the DS side. Maybe that is configureable.
Johnny
--
Johnny Billquist || "I'm on a bus
| | on a psychedelic trip
email: [email protected] || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
Peter
--
Peter Svensson ! Pgp key available by finger, fingerprint:
<[email protected]> ! 8A E9 20 98 C1 FF 43 E3 07 FD B9 0A 80 72 70 AF
------------------------------------------------------------------------
Remember, Luke, your source will be with you... always...
_______________________________________________
Simh mailing list
[email protected]
http://mailman.trailing-edge.com/mailman/listinfo/simh
