Mounter: AmigaOS Generic Mounter

Generic autoboot / automount RDB parser and mounter.

This is the mounter code from a4091.device.

1. Introduction

This Mounter software is a generic, autoboot/automount Rigid Disk Block (RDB) parser and mounter for AmigaOS. Originally developed by Toni Wilen and extended by Stefan Reinauer and Matt Harlum. It is used by the a4091.device and 'lide.device' drivers. This software is engineered to be a robust and highly compatible solution for device drivers that need to mount partitions and filesystems at boot time.

It is designed to be highly portable, 68000-compatible, and capable of operating in diverse environments, from a Kickstart 1.3 Boot ROM to a modern AmigaOS system. Its primary function is to scan for storage devices, interpret their partition maps (RDB, MBR, GPT), load the necessary filesystems, and make them available to the operating system.

---

2. Core Features

The Mounter provides a comprehensive set of features making it a versatile solution:

• Broad OS Compatibility: Full support for Kickstart 1.3 and newer, including its specific autoboot mechanisms.
• CPU Compatibility: Compatible with the Motorola 68000 processor, ensuring it runs on all classic Amiga models.
• Autoboot Capability: Can participate in the Amiga's autoconfig boot process to mount bootable partitions
• Full Automount Support: Automatically finds and mounts all valid partitions it discovers.
• Rich Filesystem Support:
  • RDB: Full support for the Amiga Rigid Disk Block (RDB) standard.
  • MBR & GPT: Support for PC-style Master Boot Record and GUID Partition Table layouts, enhancing cross-platform compatibility.
  • CD-ROM: Capable of booting from CD-ROMs following the ISO 9660 standard with Amiga-specific "Amiga Boot" or "CDTV" system identifiers.
• LUN Support: Can scan for and mount devices on multiple Logical Unit Numbers (LUNs).

---

3. High-Level Architecture

The Mounter's architecture is centered around a single primary entry point, MountDrive(), which orchestrates the entire mounting process. It operates by opening a specified device driver, scanning its units, and attempting to identify and mount partitions.

3.1. Key Data Structures

• struct MountStruct (mounter.h): This is the main input structure passed to the MountDrive function. It defines the parameters for a mounting session.

  • deviceName: The name of the device driver to use (e.g., scsi.device).
  • unitNum: A pointer to an array of unit numbers to scan.
  • creatorName: A string to identify the creator of the filesystem entries.
  • configDev: A pointer to the ConfigDev structure for an autoconfig board, essential for autobooting.
  • luns, slowSpinup, cdBoot, ignoreLast: Boolean flags to control behavior like LUN scanning, spin-up delays, CD booting, and handling of the RDB RDBFF_LAST flag.
  • SysBase: A pointer to the Exec library base.

• struct MountData (mounter.c): An internal state-management structure used during the MountDrive execution. It holds pointers to opened libraries, the I/O request, device geometry, and state variables for the scanning process.

3.2. Control Flow

The logical flow of the MountDrive function is as follows:

1. Initialization:

   • Allocate a MountData structure to hold operational state.
   • Open required system libraries (expansion.library, dos.library).
   • Create a message port and an I/O request for communicating with the device driver.

2. Device Scanning Loop:

   • Iterate through the specified SCSI targets (and LUNs, if enabled).
   • For each unit, attempt to OpenDevice().

3. Partition Scheme Identification:

   • If a device is opened successfully, get its geometry using TD_GETGEOMETRY.
   • Based on the device type (DG_DIRECT_ACCESS, DG_CDROM), determine the scanning strategy.
   • For Direct Access Devices:
     1. Attempt to find an RDB by scanning the first RDB_LOCATION_LIMIT blocks (ScanRDSK).
     2. If no RDB is found, check for a GUID Partition Table (GPT) at block 1 (ScanMBR).
     3. If no GPT is found, check for a Master Boot Record (MBR) at block 0 (ScanMBR).
   • For CD-ROM Devices:
     1. Check if the unit is ready and contains a data CD (ScanCDROM).
     2. Check the Primary Volume Descriptor (PVD) for "AMIGA BOOT" or "CDTV" identifiers (CheckPVD).
     3. If not an Amiga bootable CD, fall back to scanning for an RDB (ScanRDSK).

4. Partition Processing:

   • Call the appropriate parsing function (ParseRDSK, ParseGPT, ParseMBR, ScanCDROM).
   • These functions iterate through partition entries.

5. Filesystem Loading & Mounting:

   • For each valid partition, ParsePART is called.
   • It reads the partition's DosEnvec to determine the required filesystem DosType.
   • It calls ParseFSHD to find or load the required filesystem. ParseFSHD searches FileSystem.resource and, if not found or if the version is older, loads the filesystem from the disk's FileSystem Header Blocks (FSHD).
   • The loaded filesystem code is relocated in memory using the fsrelocate function.
   • A DeviceNode is created using MakeDosNode() and populated with the partition's parameters.
   • The DeviceNode is added to the system's mount list using AddDosNode() or AddBootNode() (for bootable partitions).

6. Cleanup:

   • The device is closed.
   • The loop continues to the next unit/target.
   • After the loop, all allocated resources (I/O requests, ports, library bases) are freed.

---

4. Core Functionality Details

4.1. Filesystem Relocation (fsrelocate)

A critical function is fsrelocate, which loads Amiga HUNK-formatted filesystem binaries from disk into memory.

1. It starts by reading a HUNK_HEADER.
2. It pre-allocates memory for all hunks (HUNK_CODE, HUNK_DATA, HUNK_BSS) defined in the header.
3. It reads each hunk's data from the LSEG (Load Segment) blocks on disk.
4. It processes relocation hunks (HUNK_RELOC32) to resolve memory address pointers within the loaded code and data, making the filesystem executable from its new location in RAM.
5. Finally, it performs a CacheClearU() to ensure instruction caches are flushed before the OS attempts to execute the newly loaded code.

4.2. Filesystem Resource Management (FSHDProcess, FSHDAdd)

The mounter interacts carefully with the central FileSystem.resource.

• FSHDProcess is responsible for checking if a required filesystem (identified by DosType) is already present in the resource list.
• If a filesystem is not present, or if the version on disk is newer than the one in memory, it allocates a new FileSysEntry.
• After a filesystem is successfully loaded and relocated by fsrelocate, FSHDAdd adds the new FileSysEntry to FileSystem.resource, making it available for all subsequent mounting operations system-wide.
• If FileSystem.resource does not exist (common on KS 1.3), it is created.

4.3. Kickstart 1.3 Compatibility

To maintain compatibility with Kickstart 1.3, which lacks certain OS functions, the mounter includes its own implementations:

• W_CreateMsgPort / W_DeleteMsgPort
• W_CreateIORequest / W_DeleteIORequest

These functions use AllocMem() to create the necessary structures and manage signals manually, mimicking the behavior of their modern counterparts. For autobooting, it manually creates and inserts a BootNode into the expansion.library's MountList, as AddBootNode is not fully featured on KS 1.3.

4.4. Autoboot Process

When a bootable partition is found and the mounter is operating in an autoboot context (i.e., a configDev is provided), it uses AddBootNode() instead of AddDosNode(). This function links the mountable partition directly to the autoconfig hardware board, signaling to the OS that this is a candidate for booting. If no configDev is provided, the mounter can create a "fake" one to enable autobooting from devices that are not on a standard autoconfig chain.

---

5. Dependencies

The Mounter relies on the following standard AmigaOS libraries:

• exec.library (v34+): For memory management, tasking, ports, and core system functions.
• expansion.library (v34+): For adding boot nodes and interacting with the autoconfig process.
• dos.library (v34+): For creating DeviceNode structures and adding them to the DOS list.

---

6. License

Copyright 2021-2022 Toni Wilen

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.