IBM System/370 145 Manual page 108

microcode engineering changes can be installed merely by replacing the
current disk cartridge with another cartridge that contains the new
microprogram.
The Two-Channel Switch feature, identical to the same feature for the
2314 facility, can be installed on a 3830 Model 1 control unit to allow
the 3830 to be attached to two channels. The Two-Channel switch,
Additional feature can be added to this configuration to permit the 3830
to be attached to four channels. The four channels involved can be all
on the same system, each on a different system, or some on the same
system and others on different systems. The <::hannels to which these two
features are connected must have one control unit position and if block
multiplexing is to be used, eight unshared subchannels available. The
enable/disable switch can be set to dedicate the 3830 to any subset of
the four channels.
The 3830 also incorporates new error detection, correction, and
logging features, designed to improve its availability and
serviceability. The following features implemented in the 3830 are not
provided for previously announced direct access devices:
• I/O error routine correction of recoverable data errors on read
operations with data supplied by the control unit in sense bytes
• command retry initiated by the control unit to attempt hardware
correction of certain errors without programming assistance
• Error logging by the control unit in its control storage of
successful command retry operations
• Inline diagnostic tests contained on disk cartridges, which can be
run on a single drive to diagnose hardware malfunctions while other
drives in the string continue normal operations. (Inline
diagnostics are provided only for 2314 facilities.)
Recovery of correctable data ~ errors. When the control unit
detects a correctable data error during the reading of the data portion
of a record, i t generates the information necessary to correct the
erroneous bytes. The sense bytes presented by the control unit contain
a pattern of corrective bits and a displacement value to indicate which
of the bytes transferred to processor storage contain the errors. The
disk error recovery program need only EXCLUSIVE OR (logical operation)
the corrective bit pattern with the error bytes in the input area in
processor storage to correct the errors.
Command retry. Error correction (without programming assistance) is
performed by a channel/control unit command retry procedure without an
intervening I/O interruption in the following situations:
1. When a correct~ble data error occurs during a search or read
operation on home address, record count, or record key.
During a search or read operation, the home address, count, or
key read from the disk track is placed in a buffer in control
storage within the control unit. When a correctable data error
occurs, the control unit corrects the data in the buffer and
reissues the command that caused the error. During reorientation
to the record, the control unit disconnects and frees the block
multiplexer channel. When the failing search or read command is
reexecuted, the corrected data in the buffer is used instead of
the data actually on the track.
2. When an uncorrectable data error is detected on any portion of
the record during a read or a search operation.
98 A Guide to the IBM System/370 Model 145