Browse Prior Art Database

VM/370 V=R Virtual Machine Recovery Survival

IP.com Disclosure Number: IPCOM000045888D
Original Publication Date: 1983-Apr-01
Included in the Prior Art Database: 2005-Feb-07
Document File: 2 page(s) / 13K

Publishing Venue

IBM

Related People

Reynolds, RO: AUTHOR

Abstract

The programming support to be described can be added to VM/370 to preserve the status of a V=R guest across a CP abend, thus allowing the guest to continue running once CP has reIPLed (initial program loaded) itself, or if the V=R guest is using the preferred machine assist (PMA), instead of CP loading a disabled wait PSW in the event of certain system failures, the V=R guest will be given control in native mode.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 2

VM/370 V=R Virtual Machine Recovery Survival

The programming support to be described can be added to VM/370 to preserve the status of a V=R guest across a CP abend, thus allowing the guest to continue running once CP has reIPLed (initial program loaded) itself, or if the V=R guest is using the preferred machine assist (PMA), instead of CP loading a disabled wait PSW in the event of certain system failures, the V=R guest will be given control in native mode.

The new support captures and preserves the V=R virtual machine environment when a CP software failure occurs and restores this environment when VM/370 has been automatically reIPLed. Since the CP nucleus is reloaded into storage through DMKSAV, the next page after DMKSAV is used as a data buffer. The saving of the VMBLOK and ECBLOK is fundamental. Pending I/O interrupt data contained in the virtual I/O control blocks is preserved. The code for constructing the I/O interrupt already existed in DMKDSP. The only change required was to save them in the data buffer rather than reflect them to the virtual machine. The existing dump and reIPL process purged and discarded pending I/O interrupts. This had to be modified to preserve those interrupts which were not generated by the dump or reIPL process itself.

After DMKCPI is given control, the I/O devices belonging to the V=R virtual machine must not be disturbed. This is accomplished by marking those RDEVBLOKs and not issuing any I/O operations to those devices. When the V=R virtual machine has been logged on by DMKCPI, the new VMBLOK and ECBLOK are updated with data from the old blocks. The existing anchor, DMKDSPQI provided a place to chain IOBLOKs queued for the V=R virtual machine. If the V=R user was running in PMAMODE, then the appropriate IOBLOKs are queued on DMKPMAIO.

In SPMODE, since the second processor was stopped when the...