IBM 1 Series - Interrupt-Service Sequence Description

Interrupt-Service Sequence Description
2-22 GA34-0033
Refer to Figure 2-9. An interrupt-service sequence is executed as follows:
1.
‘Service gate’ is activated. The device detecting the first leading edge
of ‘service gate’ activation following a poll capture is the selected
device for the service sequence. This is called service gate capture.
‘Service gate’ is propagated outbound from the processor so that the
priority is determined by the sequence from the processor for each
interrupt level.
Upon service gate capture, the device activates the ‘condition code in
bus,’ the ‘data bus,’ and then ‘service gate return.’ ‘Condition code in
bus’ and ‘data bus’ must be activated prior to ‘service gate return.’
The permissible delay, T2, from ‘service gate’ to ‘service gate return,’
as seen at the device interface, is 3 microseconds, maximum. There is
no specific time-out on this delay. As with DPC sequences, this delay
is provided for attachment convenience; however, for performance
reasons, this delay should be held to a minimum.
‘Data strobe’ is activated. The time between ‘service gate return’ and
‘data strobe,’ CT1, is 100 nanoseconds, minimum, as seen at the
device interface. The duration of ‘data strobe,’ CT2, is 200
nanoseconds, minimum, as seen at the device interface.
‘Service gate’ and ‘data strobe’ are deactivated simultaneously at the
processor channel output. As denoted by the relationship of CT2 and
CT3 in Figure 2-9, ‘data strobe’ may extend beyond the active
envelope of ‘service gate’ by 100 nanoseconds, maximum, but the
overlap of ‘data strobe’ and ‘service gate’ is 100 nanoseconds,
minimum, as seen at the device interface.
Upon the deactivation of both ‘service gate’ and ‘data strobe,’ the
device deactivates the ‘condition code in bus’ and ‘service gate return.’
The ‘condition code in bus’ and ‘data bus’ must be deactivated prior to
the deactivation of ‘service gate return,’ as seen at the device interface.
The permissible delay, T4, from the deactivation of both ‘service gate’
and ‘data strobe’ to the deactivation of ‘service gate return’ is 3
microseconds, maximum, as seen at the device interface. Again, this
delay is for attachment convenience; however, the delay should be
held to a minimum. AIl device action for the service must take place
prior to dropping ‘service gate return.’
The total duration of the interrupt-service sequence is timed out by the
channel for error-detection purposes. The total duration is measured
from the activation of ‘service gate’ to the deactivation of ‘service gate
return.’ If, within the time-out, ‘service gate return’ either: (1) does
not become active with ‘service gate’ active or (2) does not deactivate
after it has become active, a machine check occurs and the channel
activates the ‘halt or MCHK’ line. Under normal operation, if the
device attachment adheres to the specified times over which it has
control, the total duration of the sequence is within this channel
time-out.