Paqe of GC20-1807-7 As Updated April 1, 1981 by TNL GN25-0829
The VM/370 control program destages any changed cylinders on a volume
at detach time when the user count for the entire volume goes to zero.
Destaging is accomplished when CP issues a relinquish order to the MSS via the central server application program (DMKMSS). A relinquish order
is issued at detach time for volids mounted on SYSVIRT virtual unit
addresses and VIRTUAL virtual unit addresses. This is the case only
when the volume was mounted by the control program on behalf of the
guest operating system. Refer to the Generation Guide that identifies VS1/VS2 APARs to be applied to the central-server--virtual machine operating system to support the VM/370 relinquish function (APAR 11344).
172.2 IBM VM/370 System Proarammer's Guide
Asynchronous MSS Mount Processing When an MSS volume mount is required to satisfy a LINK or ATTACH command
or an MDISK or OED directory statement, CP returns control to the
virtual machine as soon as MSC accepts the mount request. The virtual
machine may continue to execute before the virtual device specified en
the MDISK, OED, LINK, or ATTACH is available.
The reasons for asynchronous MSS mount processing are the relatively
long time required to complete the mount, and the chance that an errcr
may occur in the MSS after the mount order is accepted. The virtual
device to be mounted may not be vital to the specific task to be
accomplished. Also, if an error occurs in the MSS (such as a permanent
read error on a cartridge) after the mount is accepted, the errcr
indication is passed from the MSC to the virtual machine. VM/370 cannot
determine that an error has occurred and that the mount will net
complete. If the virtual machine were not dispatchable until the mount completed, it would be locked out until the MSS error was corrected. With asynchronous mount processing, the virtual machine has the
flexibility to either continue processing without the affected virtual
device, or wait until the MSS mount completes. If the virtual machine
issues an SIO instruction to a virtual device that is defined on the
volume being mounted, VM/370 will queue the I/O request until the mount
completes. The virtual machine will be marked I/O wait nondispatchable
until the mount completes and the SIO is started. I VM/370 Processing of MSS Cylinder Faults VM/370 supports 3330V cylinder fault processing in two ways: real
channel programs directed to 3330Vs are constructed so that cylinder
faults can be recognized and the channel program restarted; and the
attention interruption (indicating that the cylinder fault has been
satisfied) is recognized and any I/O for that device restarted. When the VM/370 processor issues a seek CCW to a 3330V device, the
staging adapter must translate the seek argument to the correct cylinder
of staging space. If the cylinder referenced in the seek is staged,
then the SIO is passed to the associated staging DASD drive. If the
referenced cylinder is not staged, the staging adapter initiates
cylinder fault processing. The staging adapter first passes a cylinder
fault indication to the MSC, requesting the cylinder of data to be
staged. It then returns a status modifier to the channel in response to
the seek, which causes the channel to skip one CCW in its CCW fetch
processing. That is, the channel does nQt fetch the next CCW after the
seek.
As a result of the cylinder fault, the MSC allocates staging space
for the requested data and causes it to be staged. The staging adapter
then generates a channel end/device end interruption to indicate that
the cylinder has been staged.
It is possible in error situations that the attention interruption may not be received. Each time an I/O request is queued by VM/370 as a
result of a cylinder fault, a timer is set. If the timer expires before
the interruption is received, a message (DMKSSS074I) is written to the VM/370 system operator and the request is retried. Part 2. Control Program (CP) 173
Previous Page Next Page