Page of GC20-1807-7 As Updated April 1, 1981 by TNL GN25-0829
In most cases, the I/O devices and control units on a channel are shared
among many virtual machines as minidisks and dedicated devices, and
shared with CP system functions such as paging and spooling. Because of
this sharing, CP has to schedule all the I/O requests to achieve a
balance between virtual machines. In addition! CP must reflect the results of the subsequent I/O interruption to the appropriate storage
areas of each virtual machine.
By specifying a dedicated (or channels) for a virtual machine
via the Class B ATTACH CHANNEL command, the CP channel scheduling
function is bypassed for that virtual machine. A virtual machine
assigned a dedicated channel has that channel and all of its devices for
its own exclusive use. CP translates the virtual storage locations
specified in channel commands to real locations and performs any
necessary paging operations, but does not perform any device address
translations. The virtual device addresses on the dedicated channel
must match the real device addresses; thus, a minidisk cannot be used. SPOOLING FUNCTIONS A virtual unit record which is mapped directly to a real unit
record device, is said to be dedicated. The real device is then
controlled completely by the virtual machine's operating system. CP facilities allow multiple virtual machines to share unit record
devices. Since virtual machines controlled by CKS ordinarily have
modest requirements for unit record input/output devices, such device
sharing is advantageous, and it is the standard mode of system
operation. Spooling operations cease if the direct access storage space assigned
to spooling is exhausted, and the virtual unit record devices appear in
a not-ready status. The system operator may make additional spooling
space available by purging existing spool files or by assigning
additional direct access storage space to the spooling fUnction. Specific files can be transferred from the spooled card punch or
printer of a virtual machine to the card reader of the same or another
virtual machine. Files transferred between virtual unit record devices
by the spooling routines are not physically punched or printed. With this method, files can be made available to multiple virtual machines,
or to different operating systems executing at different times in the
same virtual machine.
Files may also be spooled to remote stations via the Remote Spooling Communications Subsystem (RSCS), a component of VK/370. For a
description of RSCS and the remote stations that it supports, see "Part
5. Remote Spooling communications Subsystem (RSCS)." CP spooling includes many desirable options for the virtual machine
user and the real machine operator. These options include printing
multiple copies of a single spool file, backspacing any number of
printer pages, and defining spooling classes for the scheduling of real
output. Each output spool file has, associated with it, a 136-byte area
known as the spool file tag. The information contained in this area and
its syntax are determined by the originator and receiver of the file.
For example, whenever an output spool file is destined for transmission 80 IBM VM/370 System Proqrammer's Guide
to a remote location via the Remote Spooling Communications Subsystem,
RSCS expects to find the destination identification in the file tag. Tag
data is set, changed, and queried using the CP TAG It to spool terminal input and output. All data sent to
the terminal, whether it be from the virtual machine, the control
program or the virtual machine operator, can be spooled. Spooling is
particularly desirable when a virtual machine is run with its console
disconnected. Console spooling is usually started via the command SPOOL CONSOLE START An exception to this is when a system operator logs on using a graphics
device. In this instance, console spooling is automatically started and
continues in effect even if the system operator should disconnect from
the graphics device and log on to a nongraphic device. In order to stop
automatic console spooling, the system operator must issue the command SPOOL CONSOLE STOP SPOOL FILE RECOVERY If the system should suffer an abnormal termination, there are three
degrees of recovery for the system spool files; warm start (WARft), checkpoint start (CKPT), and force start (FORCE). Warm start is
automatically invoked if SET DUMP AUTO is in effect. Otherwise, the
choice of recovery method is selected when the following message is
issued;
hh: mm: ss START «COLD I WARM I CKPT I FORCE) (DRAIN}) I (SHUTDOWN) :
Bote that a cold (COLD) start does not recover any spool files. After a system failure, the warm start procedure copies spool file,
accounting, and system message data to warm start cylinders on an
auxiliary DASD. When the system is reloaded, this information is
retrieved and the spool file chains and other system data are restored
to their original status. If the warm start procedure cannot be
implemented because certain required areas of storage are invalid, the
operator is notified to take other recovery procedures.
Any new or revised status of spool file blocks, spooling devices, and
spool hold queue blocks is dynamically copied to checkpoint cylinders on
an auxiliary DASD as they occur. When a checkpoint (CKPT) start is
requested, this is the information that is used to recreate the spocl
file It differs from warm start data in that only spool file
data is restored; accounting and system messages information is not
recovered. Also, the order of spool files on any particular restored
chain is not the original sequence but a random one. Part 2. Control Program (CP) 81
Previous Page Next Page