MultiAccessDiskMgmt Feb85

This document, dated February 8, 1985, is a specification for TOPS-20 Multi-Access Disk Management. It outlines how TOPS-20 handles access to disks that can be shared by multiple systems, particularly focusing on scenarios where communication over the CI (Computer Interconnect) may be disrupted.

The core goals of this specification are:

1. No Data Corruption: Ensuring data integrity on all multi-accessed disks.
2. 5.1 Compatibility: Allowing continued use of unsupported, pre-6.0 multi-access configurations by providing a "don't-care" option for specific disks.
3. Minimal Overhead Writing: Reducing disk access by only writing management data when significant events occur or data changes.

Key aspects of the specification include:

• Disk Access Logic: TOPS-20 employs an algorithm to determine if a disk can be accessed. This logic prioritizes single-ported, "don't-care," or dedicated (20F) disks. For dual-ported disks potentially accessed by other KL10 systems, access is granted only if the other system is down, or if it is up, actively responding to "request-IDs," and has a Control File System (CFS) connection established with the local system. Otherwise, the disk is declared offline.
• Data Structures: Critical information is stored in the Processor Data Block (PDB) on disk and copied to the Unit Data Block (UDB) in memory. These include disk serial numbers, a "don't-care" flag (UDBDCF), and a status word (UDBST1) indicating various conditions like a disk being forced offline or a PDB read/write in progress. The Request-ID Status (RIDSTS) table tracks the responsiveness of other CI nodes, which is crucial for access decisions.
• User Interface: Commands like CHECKD (ENABLE/DISABLE DON'T-CARE structure) and SMON% .SFDCD (declare a disk drive "don't-care" to the monitor) allow users to configure multi-access behavior. SETSPD provides a command to set "DONTCARE" for specific units.
• Homeblock Management: TOPS-20 uses three homeblock areas on a disk. These homeblocks are checked during significant CI or disk events (e.g., system startup, CI wire status changes, connection establishment/breakage). A disk remains offline until its homeblock checking is complete to ensure consistent state.
• Restrictions: All disks must have unique serial numbers. Certain dual-ported MASSBUS configurations (e.g., between different TOPS-20 versions or systems on different CIs) are explicitly unsupported.
• Dependencies: The RSX20F system provides disk configuration information to TOPS-20 at startup, which is vital for managing multi-accessed disks.

In essence, the document details a robust mechanism for TOPS-20 to manage shared disk resources, prioritizing data integrity through a detailed access control logic based on network connectivity and the operational status of other connected systems, while offering backward compatibility through user-configurable "don't-care" settings.