IBM 1 Series - Page 49

2-38 GA34-0033
@© Termination and retry sequences (Figure 2-15)
Host IPL retry on Series/1 is not only allowed, but is encouraged to
be supported by a host IPL device. Retry sequences are dependent
upon the type of abnormal termination. The discussion of
asynchronous I/O activity in item 1 of @ applies equally as well in the
following sequence descriptions.
Gf] System reset termination and retry (Figure 2-15).
1.
If a ‘system reset’ occurs after the device has enabled cycle-steal
requests and transfers, the device must deactivate the ‘IPL’ tag within
200 nanoseconds at the device interface, terminate the cycle-steal
transfers, and execute all other system reset functions (refer to “Reset
Sequences Description’’). This ‘system reset’ could be the result of (1)
the operator pressing the Reset key or (2) the operator pressing the
Load key to begin a processor-initiated IPL sequence. Note that this
condition for dropping the ‘IPL’ tag is dependent upon being in an
enabled state for IPL transfer as a result of the deactivation of ‘system
reset’ following activation of the ‘IPL’ tag. This termination must be
executed by the attachment regardless of its capability to support
retry.
If retry is supported, the I/O device may not initiate a retry on its
own. This is essentially a restatement that the device may activate the
‘IPL’ tag only once for each host IPL command successfully executed.
After execution of the host IPL command for retry, the device must
reactivate the ‘IPL’ tag. The activation may not occur until a time, T1,
of 3 microseconds, minimum, after the deactivation of the ‘system
reset’ that terminated the IPL sequence.
Hung IPL Termination and Retry (Figure 2-15).
1.
If, during cycle-steal transfers, an error condition is posted to the
device on the status bus, the device must terminate further requests
and cycle-steal transfers, leave the ‘IPL’ tag active, and not present an
end interrupt. If, during the cycle-steal requests and transfers, a
hardware failure causes a channel time-out, the system remains in a
“hung” condition. The I/O channel IPL device should leave the ‘IPL’
tag active. ‘Halt or MCHK’ tag activation does not occur so that
diagnosis of the problem in the state in which the failure or error
occurred is allowed. This termination must be executed by the device
regardless of its capability to support retry.
If retry is supported, the I/O channel IPL device may not drop the
‘IPL’ tag or initiate a retry on its own.
After execution of the host IPL command for retry, the device
deactivates, and then reactivates the ‘IPL’ tag. The reactivation must
occur within a time, T2, of from 3 microseconds, minimum, to 7
microseconds, maximum, after deactivation of the ‘IPL’ tag, as seen at
the device interface.