Disk IO Processing Apr78

This document, "Disk I/O Processing" from April 1978, details how the TOPS10 operating system handles disk input/output operations, including the general flow, dual-port support, and comprehensive error recovery mechanisms.

Key areas covered:

1. General Disk I/O Flow:

   • Describes the sequence for handling disk requests, including calculating the required cylinder, initiating seeks, and queuing requests for busy drives or channels to maximize throughput.
   • Explains the interrupt-driven process for handling completed seeks, drive status changes (e.g., coming online), and data transfer completions.
   • Outlines how errors during data transfer are detected and lead to error recovery procedures.
   • Details the process for starting queued seeks (shortest/closest) and transfers (shortest latency), prioritizing SWAPPER requests.
   • Mentions special considerations for MASSBUS devices and potential conflicts with a "front end" disk unit.

2. Dual Port Handling:

   • Explains that dual-ported disks (identified at system startup by matching serial numbers) use an alternate path if the primary access path is busy.

3. Exception Conditions (Error Handling):

   • Covers a wide range of error types and the system's response:
     • ECC Correctable Errors: Data is reconstructed by software, and the transfer resumes.
     • Non-data Errors (e.g., header errors): Involve extensive retry sequences (up to 30 attempts), including recalibrations and seeks.
     • Data Errors (non-ECC): Utilize detailed retry sequences involving varying the head position around the track centerline.
     • Hard Errors: If retries fail, errors are logged, and an excessive number of hard errors triggers an operator alert.
     • Seek Incomplete/Hung Devices: Detected by timers; repeated failures result in retries or the drive being taken offline.
     • Register Access Errors (RAE): Handled differently depending on the controller (RH10 vs. RH20), often involving retries or being ignored.

4. BAT Blocks (Bad Block Allocation Tables):

   • These tables record up to 63 disk errors (bad blocks) to prevent their future use, ensuring disk integrity. The system alerts if the BAT block has few remaining entries.

5. DSKRAT Utility:

   • A program designed to check disk consistency, identifying four types of errors: corrupt file structure information (RIB errors), lost blocks (allocated but unused), free blocks (used but unallocated), and multiply-defined blocks (owned by multiple files).
   • Recommends backing up, refreshing, and restoring the disk for significant consistency problems.

In essence, the document provides a deep dive into the robust and complex mechanisms the TOPS10 operating system employed for efficient and reliable disk I/O, particularly its detailed error detection and recovery strategies, crucial for data integrity on systems of that era.