Browse Prior Art Database

Context Sensitive Dump Messaging Priority Scheme for Converged Adapters

IP.com Disclosure Number: IPCOM000237296D
Publication Date: 2014-Jun-12
Document File: 4 page(s) / 33K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method for context sensitive dump messaging priority scheme for converged adapters is disclosed.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 32% of the total text.

Page 01 of 4

Context Sensitive Dump Messaging Priority Scheme for Converged Adapters

Disclosed is a method for context sensitive dump messaging priority scheme for converged adapters.

A critical component of successful customer support is the ability to diagnose a problem with the data gathered during an initial failure. This is sometimes referred to as "First Failure Data Capture" (FFDC). Requiring recreates of the problem to gather sufficient data (Second Failure Data Capture, or SFDC) is undesirable, and sometimes not possible.

Modern servers are capable of dumping entire system images for debug analysis upon detection of a problem. Some servers are capable of collecting "live dumps," which involves temporarily halting the system or even just certain components of the system, collecting data, and then resuming normal system operation. Often included in these system images are firmware dumps of the various adapter cards or other peripheral devices installed on the system. This is extremely valuable in diagnosing problems that are a result of either faulty adapter hardware or in diagnosing problems perhaps external to the server that require inspection of adapter firmware data structures or traces to solve.

With the emergence of multi-protocol converged adapters such as Fiber Channel over Ethernet (FCoE) Converged Network Adapters (CNA), complications arise in collecting timely adapter firmware dumps. Triggering and/or collecting a firmware dump on particular function of an adapter such as a CNA may cause disruption to sibling adapter functions. These sibling functions may be running entirely different protocols. For a CNA, for example, invoking a firmware dump on the Fiber Channel (FC) function may cause an unexpected interruption to a network interface controller (NIC) function on the CNA. In addition, each function might be required to collect data that is shared between the host and the adapter (such as adapter firmware queues residing in host memory). A vendor may require both adapter firmware information as well as this host memory information from each driver instance in order to compile a complete "adapter dump image" for debug analysis.

A message passing facility can be used to provide a "dump warning" notification to sibling functions as well as to request that sibling functions collect any necessary host memory information. This message passing facility might be provided by the operating system, or it may be available as a hardware feature for a particular adapter.

Assuming a communication mechanism is in place that allows an adapter function to send messages as well as receive acknowledgements that the messages were received by the siblings, an interesting question arises. Should the "master" function driving the adapter firmware dump wait for acknowledgement from all siblings before proceeding to dump the adapter, or should the adapter firmware dump be initiated immediately while the messages are sent in parallel? Waiting for ac...