IDT PCI Express 89HPES32NT24xG2, User Manual, Page: 325 / 744

Notes
PES32NT24xG2 User Manual 16 - 1 January 30, 2013
®
Chapter 16
Switch Events
Overview
As described in section Switch Events on page 1-16, in a PCI Express switch with multiple partitions a
need may exist to signal the occurrence of significant global events to a switch management agent. A need
may also exist for communication between roots associated with different partitions as well as for communi-
cation between these roots and a management agent. This section describes mechanisms provided by the
PES32NT24xG2 to facilitate these forms of communication.
Switch Events
A switch event represents the occurrence of a significant global switch event or an event within a switch
partition. Switch event status is logged in global switch control and status registers and may be used to
generate an interrupt within selected partitions.
The following events are considered switch events:
– A switch port link going up (i.e., a transition from DL_Down to DL_Up)
– A switch port link going down (i.e., a transition from DL_Up to DL_Down)
– A switch port detecting an AER error
– A fundamental reset in a partition
– A hot reset in a switch partition
– Failover mode change initiated
– Failover mode change completed
– A global signal from a switch partition
Corresponding to each of these events is a status bit in the Switch Event Status (SESTS) register and a
mask bit in the Switch Event Mask (SEMSK) register. When a switch event is detected, the event is
signaled to all partitions not masked by the Switch Event Partition Mask (SEPMSK) register. Associated
with each of these events are further status and mask registers that provide fine grain status and masking
control. These are described in section Link Up on page 16-2 through section Port AER Errors on page 16-
5.
When a switch event is signaled to a partition, the event may be used to generate an MSI or INTx inter-
rupt within the partition.
– For transparent partitions (i.e., those without an NT endpoint), an interrupt may be reported by the
upstream switch port of the partition (i.e., the PCI-to-PCI bridge function in the partition’s upstream
port). See section Interrupts on page 10-4 for details.
– For partitions consisting only of an NT endpoint port (i.e., a port configured to operate as an NT
function or NT with DMA function), an interrupt may be reported by the NT function. See section
Interrupts on page 14-20 for details.
– For partitions consisting of both a transparent switch and an NT endpoint (i.e., the upstream port
contains a PCI-to-PCI bridge function and an NT function), the event is signaled to both the
upstream PCI-to-PCI bridge function and the NT function. An interrupt may be reported by one or
both of these functions.
Figure 16.1 shows a simplified representation of the switch event detection and signaling mechanism.
As shown, for each type of switch event (e.g., switch port link-up, switch port link-down, failover change initi-
ated, etc.) there is logic that detects the occurrence of that event. The SESTS register logs the occurrence
of the event, the SEMSK register controls which events are signaled to the partitions, and the SEPMSK