
IDT Usage Models
PES32NT24xG2 User Manual
26 - 25
January 30, 2013
Notes
• PFMODE = SFMODE = 0x5 (i.e., primary and secondary failover mode is set to unattached port)
–
Ports 12 and 16 are configured to respond to primary and secondary failover events by program-
ming fields in the SWPORT12FCTL and SWPORT16FCTL registers as follows:
• PFMODE = SFMODE = 0x1 (i.e., primary and secondary failover mode is set to downstream
port)
• PFSWPART = SFSWPART = 0x0 (i.e., primary and secondary failover partition is set to partition
0)
• PFDEVNUM = SFDEVNUM = 0xC (for port 12) and 0x10 (for port 16)
–
In addition, ports 12 and 16 are configured to be reset during the failover operation by program-
ming the OMA field in the corresponding SWPORTxCTL registers to a value of 0x1.
Note: In this example, primary and secondary failover operations are configured identically. This is
necessary because after a failover occurs and the switch reconfigures itself, restoring the system to
its initial configuration is performed by software executing on the roots, and not through a subse-
quent failover operation.
The serial EEPROM configures port 0 so that the PCI-to-PCI bridge function in this port generates an
interrupt when a failover completed event is detected in the failover capability structure associated with
partition 0. To do this, the FMCC bit is cleared in the P2PINTMSK register of port 0.
Similarly, the serial EEPROM configures port 8 so that the PCI-to-PCI bridge function in this port gener-
ates an interrupt when a failover completed event is detected in the failover capability structure associated
with partition 1. To do this, the FMCC bit is cleared in the P2PINTMSK register of port 8.
Note: The enabling of MSI/INTx is done at a later time by the operating system running on each of the
roots.
At this point the failover mechanism is armed.
Failover capability 0 is associated with partition 0. When a failover is triggered, port 8 is reset and its
operating mode is changed to unattached mode. In addition, ports 12 and 16 are reset and migrated to
partition 0. Failover capability 1 is associated with partition 1. When a failover is triggered, port 0 is reset
and its operating mode is changed to unattached mode. In addition, ports 4 and 6 are reset and migrated to
partition 1.
After EEPROM loading completes, the switch exits quasi-reset mode and the root in each partition can
proceed to enumerate the switch.
–
RC0 enumerates partition 0.
–
RC1 enumerates partition 1.
RC0 and RC1 enable the interrupt mechanism (i.e., INTx or MSI) in the PCI-to-PCI bridge function in the
upstream switch port of their respective partitions, and bind the interrupt to a software interrupt handler. The
interrupt handler is responsible for understanding and notifying other software layers of the re-configuration
of the switch as a result of the failover event.
RC0 and RC1 proceed to configure the NT endpoint, use NT messaging to exchange NT window infor-
mation, and configure the NT lookup and NT Mapping tables appropriately. During normal operation, RC0
and RC1 use the NTB to exchange recovery point information and issue “heart-beats” to each other.
When a root fails to issue a heart-beat, it is assumed to have failed. As a result, the other root triggers a
software initiated failover by setting the FSWTRIG bit in the failover capability structure associated with its
partition. Once the switch completes the failover operation, an interrupt is generated by the upstream port in
the partition indicating that the failover change has completed.
For example, if RC0 fails to issue a heart-beat to RC1, RC1 triggers a failover by setting the FSWTRIG
bit in the FCAP1CTL register. As a result, the switch automatically resets port 0 and changes its operating
mode to unattached mode, and resets ports 4 and 6 and migrates them to partition 1 (see Figure 26.15).
When the switch completes the failover action, the upstream port in partition 1 generates an interrupt to
RC1. The interrupt handler executes in RC1, and notifies other software layers that ports 4 and 6 are now
available in the partition.
Summary of Contents for PCI Express 89HPES32NT24xG2
Page 20: ...IDT Table of Contents PES32NT24xG2 User Manual x January 30 2013 Notes...
Page 24: ...IDT List of Tables PES32NT24xG2 User Manual xiv January 30 2013 Notes...
Page 28: ...IDT List of Figures PES32NT24xG2 User Manual xviii January 30 2013 Notes...
Page 56: ...IDT PES32NT24xG2 Device Overview PES32NT24xG2 User Manual 1 20 January 30 2013 Notes...
Page 100: ...IDT Switch Core PES32NT24xG2 User Manual 4 22 January 30 2013 Notes...
Page 128: ...IDT Failover PES32NT24xG2 User Manual 6 4 January 30 2013 Notes...
Page 148: ...IDT Link Operation PES32NT24xG2 User Manual 7 20 January 30 2013 Notes...
Page 164: ...IDT SerDes PES32NT24xG2 User Manual 8 16 January 30 2013 Notes...
Page 170: ...IDT Power Management PES32NT24xG2 User Manual 9 6 January 30 2013 Notes...
Page 196: ...IDT Transparent Switch Operation PES32NT24xG2 User Manual 10 26 January 30 2013 Notes...
Page 244: ...IDT SMBus Interfaces PES32NT24xG2 User Manual 12 40 January 30 2013 Notes...
Page 247: ...IDT General Purpose I O PES32NT24xG2 User Manual 13 3 January 30 2013 Notes...
Page 248: ...IDT General Purpose I O PES32NT24xG2 User Manual 13 4 January 30 2013 Notes...
Page 330: ...IDT Switch Events PES32NT24xG2 User Manual 16 6 January 30 2013 Notes...
Page 342: ...IDT Multicast PES32NT24xG2 User Manual 17 12 January 30 2013 Notes...
Page 344: ...IDT Temperature Sensor PES32NT24xG2 User Manual 18 2 January 30 2013 Notes...
Page 384: ...IDT Register Organization PES32NT24xG2 User Manual 19 40 January 30 2013...
Page 492: ...IDT Proprietary Port Specific Registers PES32NT24xG2 User Manual 21 44 January 30 2013 Notes...
Page 588: ...IDT NT Endpoint Registers PES32NT24xG2 User Manual 22 96 January 30 2013 Notes...
Page 710: ...IDT JTAG Boundary Scan PES32NT24xG2 User Manual 25 12 January 30 2013 Notes...
Page 743: ...IDT Usage Models PES32NT24xG2 User Manual 26 33 January 30 2013 Notes...
Page 744: ...IDT Usage Models PES32NT24xG2 User Manual 26 34 January 30 2013 Notes...