A port's LID is assigned by the SM and is unique within the subnet. Routers do not use a LID address to route between subnets.

The LID address space is divided into reserved, unicast, and multicast address ranges (see “LID Address Space” on page 133), as well as the PLID address.

The SLID and DLID are contained within a packet's LRH field (see Figure 7-1 on page 135).

When a CA or router is injecting a packet into a subnet, the packet's LRH:SLID identifies the CA or router port that injected the packet into the subnet. When a switch is injecting a packet (e.g., a response to an SMP request packet), the SLID is the LID assigned to the switch's management port (port 0).

The SLID is always a unicast LID address.

A unicast DLID identifies the destination port on a CA or router or, on a switch, switch port 0. A multicast DLID refers to the set of destination ports within the subnet that are members of a given multicast group.

When the destination CA port is not in the same subnet, the DLID identifies a router ingress port in the path to the destination CA port.

A given CA or router port may receive packets via multiple physical paths within the subnet. Each physical path is associated with one or more DLIDs. To facilitate multipath operation while minimizing CA complexity, each CA and router port and switch port 0 is assigned a base LID and, for a router or CA, the port may optionally be assigned a range of LID addresses.

The LID space is subdivided as follows:

• LID address 0000h is reserved.

• LID address FFFFh is defined as the permissive LID address (PLID). The PLID is used as the DLID (and possibly as the SLID) during the discovery process.

• The unicast LID range consists of addresses 0001h through BFFFh.

• The multicast LID range consists of addresses C000h through FFFEh.