IEther-00 (iether) B.11.31.1303.01 Ethernet Driver Release Notes (Edition 3, June 2013)

• PCI OLRAD

• PCI Error handling and recovery

• vPars V6

• HP Integrity Virtual Machines (HPVM)

• HPVM AVIO

• HP-UX LAN provider

• MIB and driver statistics

• Configurable using SMH or nwmgr

• nettl tracing/logging

3 New and changed features in this release

3.1 New features

There are no new features included in this version of IEther-00 at the time of this publication.

3.2 Changed features

The iether driver has been enhanced with performance and stability improvements.

4 Problems fixed in this or earlier versions

4.1 Problems fixed in this version

This release introduces the following defect fixes:

Table 1 HP-UX 11i v3 March 2013 fixes — IEther-00 B.11.31.1303.01

DescriptionSR number

On server blades with 2-port NC364m/445978-B21 and 4-port NC360m/447883-B21

PCIe mezzanine 1GbE adapters, with APA aggregation configured on ports connected to

QXCR1001259174

an external switch, if a user changes the external switch status from UP to DOWN or vice

versa, a system panic occurs. The following panic string is seen:

'wait_for_lock: Already owns lock: lan:iether_hw_lockp '

This occurs in rare circumstances and only on the specific adapters noted.

4.2 Problems fixed in previous releases

Table 2 HP-UX 11i v3 March 2013 fixes — IEther-00 B.11.31.1303

DescriptionSR number

Link timeout to PCIe device with Superdome 2 XFabric failover.QXCR1001233383

This problem was reported on Superdome2 servers, involving PCIe dual-port LAN controller

chips (AD221A, AD222A, and AD393A adapters). An error on the Superdome 2 XFabric

link immediately causes a fabric failover from that link to the redundant path. Transmissions

that were in process on the failing link are replayed along the redundant path. During Crossbar

Fabric failovers, an I/O adapter could encounter a completion timeout (CTO) for transactions

being transmitted at the instant of failover. In such a scenario, the adapter might be declared

as defective although it is performing well. The CTO value on the affected I/O adapters is

not high enough to account for the XFabric failover time.