Browse Prior Art Database

Hardware Feature Detection and Configuration with Failsafe Checkpointing

IP.com Disclosure Number: IPCOM000117132D
Original Publication Date: 1995-Dec-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 2 page(s) / 102K

Publishing Venue

IBM

Related People

Brew, G: AUTHOR [+2]

Abstract

Disclosed is a method for use by an operating system to identify the manufacturers and models of various hardware features within a computer system through "snooping," in a process having a failsafe recovery mechanism. Within this context, "snooping" is the process of using an algorithm, called a "snooper," to identify the operational characteristics of a hardware feature.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 53% of the total text.

Hardware Feature Detection and Configuration with Failsafe Checkpointing

      Disclosed is a method for use by an operating system to
identify the manufacturers and models of various hardware features
within a computer system through "snooping," in a process having a
failsafe recovery mechanism.  Within this context, "snooping" is the
process of using an algorithm, called a "snooper," to identify the
operational characteristics of a hardware feature.

      Conventional methods using a "snooper" to solve the hardware
configuration problem suffer from the fact that it is generally not
possible to snoop for a hardware feature on an ISA (Industry Standard
Architecture) system without attempting some low-level operation of
the I/O (Input/Output) subsystem.  When such operation is attempted,
there is a risk that a control program will become corrupted, causing
the system to "hang," so that it must be powered down to allow
recovery.  Since it is necessary to determine the operating
characteristics of hundreds of hardware features, snoopers cannot be
effectively used on a comprehensive basis with conventional methods.

      With the presently disclosed method, a failsafe recovery method
is provided for all snooping.  If a snooping operation hangs a
system, the user can resume the installation process at the point
where the previous snooper operation failed.  The installation
process continues, resuming after the point of failure and not
repeating the operation which failed.  The installation process
records the operation which failed, determining that the hardware
feature associated with the failing snooper is not present in the
system, since, if it were present, the failure would not have
occurred.

      For this algorithm to be implemented, the code which performs a
snooping operation must be dynamically loadable by the installation
process, with the installation algorithm having an ability to make a
decision on whether to load and execute the snooping operation.
Also, the installation algorithm must have some nonvolatile memory in
which to record the current state of the installation process.  Write
access to a disk or diskette can serve this purpose.  It must be
possible to checkpoint the machine state before executing a snooper
to prevent any permanent system corrupting.  Specifically, it must be
possible to flush any pending I/O buffers at this time.  When these
prerequisites are satisfied, the installation process can snoop for
hundreds of devices and for all combinations of I/O parameters.  If
system corruption occurs, it is accommodated when the customer turns
the system off, reboots it, and resumes the installation process.

      With this algorithm, features from a list of 1...N features,
called FEATURE_LIST, is snooped.  A variable called STATE is main
tained to track the current state of the installation process.  Both
of t...