AVB Specifications

P802.1Qav

This specification defines the mechanism for queuing and forwarding AV traffic from a talker to a listener across the network. This can involve several network hops (network bridge devices that the data must pass through).

P802.1Qav is also responsible for enforcing the 75% maximum bandwidth restriction across each link of the network that can be reserved for the AV traffic.

Only a subset of the P802.1Qav requirements for an Endpoint is implemented in the

Ethernet AVB Endpoint core, with the following assumptions for talkers and listeners:

Talker Assumptions

AV traffic Ethernet frames that are input to the Ethernet AVB Endpoint use the VLAN priority values that the Bridges in the network recognize as being associated with SR classes exclusively for transmitting stream data.

Legacy traffic Ethernet frames that are input to the Ethernet AVB Endpoint do not use the VLAN priority values that the Bridges in the network recognize as being associated with SR classes exclusively for transmitting stream data.

The credit shaping algorithm operates on the AV traffic port; so in order to comply with the transmission selection rules for P802.1Qav, all Ethernet frames input on the AV traffic port are assumed to be of the same SR Class. However, the Ethernet AVB Endpoint does not enforce this rule and it is acceptable to send a mix of SR Class A and SR Class B Ethernet frames on the AV traffic port. In this case the Ethernet AVB Endpoint will not prioritize SR Class A Ethernet frames over SR Class B Ethernet frames; instead it will apply the credit-based shaper algorithm to all of the Ethernet frames that are input on the AV traffic port.

The Ethernet AVB Endpoint assumes that any per-stream traffic management has been done prior to AV traffic being input on the AV traffic port. To comply with the transmission selection rules for P802.1Qav it is assumed that if multiple streams are input to the Ethernet AVB Endpoint via the AV traffic port, that the credit-based shaper algorithm has been used per stream as the transmission selection mechanism, prior to the AV traffic being input on the AV traffic port.

If multiple AV streams are input to the Ethernet AVB Endpoint via the AV traffic port, it is assumed that the IdleSlope/SendSlope control registers (See “Tx Arbiter Send Slope Control Register” and “Tx Arbiter Idle Slope Control Register”) are programmed correctly to be the sum of the IdleSlope /SendSlope values for all the streams that are input on the AV traffic port. The credit-based shaper algorithm used on the AV traffic port will enforce a hiLimit/loLimit on the credits to ensure that this interface is not misused.

Listener Assumptions

The Ethernet AVB Endpoint provides a mechanism for identifying received AV traffic for either one or two SR classes (see “Rx Filtering Control Register”); however, it does not provide any buffering for AV traffic Ethernet frames. Buffering is expected to be done outside the Ethernet AVB Endpoint, after it has separated out the AV traffic Ethernet frames, as the buffering requirements are expected to be application-specific.

Ethernet AVB Endpoint User Guide

www.xilinx.com

31

UG492 July 23, 2010

Page 31
Image 31
Xilinx UG492 manual P802.1Qav, Talker Assumptions, Listener Assumptions, AVB Specifications