The Book

The Book — Extras

Software

Meta

About

elsewhere

Contents

Chapter 1 Origins of Mac OS X

• 1.1   Apple’s Quest for the Operating System
  • 1.1.1   Star Trek
  • 1.1.2   Raptor
  • 1.1.3   NuKernel
  • 1.1.4   TalOS
  • 1.1.5   Copland
  • 1.1.6   Gershwin
  • 1.1.7   BeOS
  • 1.1.8   Plan A
• 1.2   The NeXT Chapter
  • 1.2.1   NEXTSTEP
  • 1.2.2   OPENSTEP
• 1.3   The Mach Factor
  • 1.3.1   Rochester’s Intelligent Gateway
  • 1.3.2   Accent
  • 1.3.3   Mach
  • 1.3.4   MkLinux
  • 1.3.5   Musical Names
• 1.4   Strategies
  • 1.4.1   Mac OS 8 and 9
  • 1.4.2   Rhapsody
    • 1.4.2.1   Blue Box
    • 1.4.2.2   Yellow Box
• 1.5   Toward Mac OS X
  • 1.5.1   Mac OS X Server 1.x
  • 1.5.2   Mac OS X Developer Previews
    • 1.5.2.1   DP1
    • 1.5.2.2   DP2
    • 1.5.2.3   DP3
    • 1.5.2.4   DP4
  • 1.5.3   Mac OS X Public Beta
  • 1.5.4   Mac OS X 10.x
    • 1.5.4.1   Mac OS X 10.0
    • 1.5.4.2   Mac OS X 10.1
    • 1.5.4.3   Mac OS X 10.2
    • 1.5.4.4   Mac OS X 10.3
    • 1.5.4.5   Mac OS X 10.4

Chapter 2 An Overview of Mac OS X

• 2.1   Firmware
• 2.2   Bootloader
• 2.3   Darwin
  • 2.3.1   Darwin Packages
  • 2.3.2   The Benefits of Darwin
  • 2.3.3   Darwin and Mac OS X
• 2.4   The xnu Kernel
  • 2.4.1   Mach
  • 2.4.2   BSD
  • 2.4.3   The I/O Kit
  • 2.4.4   The libkern Library
  • 2.4.5   The libsa Library
  • 2.4.6   The Platform Expert
  • 2.4.7   Kernel Extensions
• 2.5   A User-Space View of the File System
  • 2.5.1   File System Domains
    • 2.5.1.1   The User Domain
    • 2.5.1.2   The Local Domain
    • 2.5.1.3   The Network Domain
    • 2.5.1.4   The System Domain
  • 2.5.2   The /System/Library/ Directory
• 2.6   The Runtime Architecture
  • 2.6.1   Mach-O Files
  • 2.6.2   Fat Binaries
  • 2.6.3   Linking
    • 2.6.3.1   Multiple Binding Styles
    • 2.6.3.2   Two-Level Namespaces
    • 2.6.3.3   Weakly Linked Symbols
    • 2.6.3.4   dyld Interposing
• 2.7   The C Library
• 2.8   Bundles and Frameworks
  • 2.8.1   Bundles
  • 2.8.2   Property List Files
  • 2.8.3   Frameworks
  • 2.8.4   Prebinding
• 2.9   Core Services
• 2.10   Application Services
  • 2.10.1   Graphics and Multimedia Services
    • 2.10.1.1   Quartz
    • 2.10.1.2   QuickDraw 2D
    • 2.10.1.3   OpenGL
    • 2.10.1.4   Core Image and Core Video
    • 2.10.1.5   QuickTime
    • 2.10.1.6   Core Audio
  • 2.10.2   Other Application Services
• 2.11   Application Environments
  • 2.11.1   BSD
  • 2.11.2   The X Window System
  • 2.11.3   Carbon
    • 2.11.3.1   Support for CFM Binaries
    • 2.11.3.2   Carbon APIs
  • 2.11.4   Cocoa
    • 2.11.4.1   Nib Files
    • 2.11.4.2   Core Data
  • 2.11.5   WebObjects
  • 2.11.6   Java
  • 2.11.7   QuickTime
  • 2.11.8   Classic
  • 2.11.9   Rosetta
• 2.12   User Interface
  • 2.12.1   Visual Effects
  • 2.12.2   Resolution-Independent User Interface
  • 2.12.3   Productivity Features
    • 2.12.3.1   Fast User Switching
    • 2.12.3.2   Dashboard
    • 2.12.3.3   Spotlight
  • 2.12.4   Universal Access Support
• 2.13   Programming
  • 2.13.1   Xcode
  • 2.13.2   Compilers and Libraries
  • 2.13.3   Interpreters
    • 2.13.3.1   AppleScript
    • 2.13.3.2   Automator
    • 2.13.3.3   Command-Line Support
  • 2.13.4   Tools
    • 2.13.4.1   Debugging and Analysis Tools
    • 2.13.4.2   Computer Hardware Understanding Development Tools
    • 2.13.4.3   Visual Tools
• 2.14   Security
  • 2.14.1   Kernel-Space Security
  • 2.14.2   User-Space Security
    • 2.14.2.1   CDSA Plug-ins
    • 2.14.2.2   CSSM API
    • 2.14.2.3   Mac OS X Security APIs
    • 2.14.2.4   Security Server and Security Agent
    • 2.14.2.5   Using Authorization Services
    • 2.14.2.6   Miscellaneous Security-Related Features
  • 2.14.3   System Administration
    • 2.14.3.1   Interacting with the Security Framework
    • 2.14.3.2   Interacting with Directory Services
    • 2.14.3.3   Managing System Configuration
  • 2.14.4   The Auditing System
• 2.15   Mac OS X Server
  • 2.15.1   Xgrid
    • 2.15.1.1   Xgrid Architecture
    • 2.15.1.2   Xgrid Software
  • 2.15.2   Xsan
    • 2.15.2.1   Storage in Xsan
    • 2.15.2.2   Metadata Controllers
    • 2.15.2.3   Client Systems
    • 2.15.2.4   Communication Infrastructure
• 2.16   Networking

Chapter 3 Inside an Apple

• 3.1   The Power Mac G5
  • 3.1.1   The U3H System Controller
  • 3.1.2   The K2 I/O Device Controller
  • 3.1.3   PCI-X and PCI Express
    • 3.1.3.1   PCI-X
    • 3.1.3.2   PCI Express
  • 3.1.4   HyperTransport
  • 3.1.5   Elastic I/O Interconnect
• 3.2   The G5: Lineage and Roadmap
  • 3.2.1   Fundamental Aspects of the G5
    • 3.2.1.1   64-bit Processor
    • 3.2.1.2   Superscalar
    • 3.2.1.3   Speculative Execution
    • 3.2.1.4   Out-of-Order Execution
  • 3.2.2   New POWER Generations
  • 3.2.3   The PowerPC 970, 970FX, and 970MP
  • 3.2.4   The Intel Core Duo
• 3.3   The PowerPC 970FX
  • 3.3.1   At a Glance
  • 3.3.2   Caches
    • 3.3.2.1   L1 and L2 Caches
    • 3.3.2.2   Cache Properties
    • 3.3.2.2   Associativity
    • 3.3.2.2   Store Policy
    • 3.3.2.2   MERSI
  • 3.3.3   Memory Management Unit (MMU)
    • 3.3.3.1   SLB and TLB
    • 3.3.3.2   Address Translation
    • 3.3.3.3   Caching the Caches: ERATs
    • 3.3.3.4   Large Pages
    • 3.3.3.5   No Support for Block Address Translation Mechanism
  • 3.3.4   Miscellaneous Internal Buffers and Queues
  • 3.3.5   Prefetching
  • 3.3.6   Registers
    • 3.3.6.1   UISA and VEA Registers
    • 3.3.6.2   OEA Registers
  • 3.3.7   Rename Registers
  • 3.3.8   Instruction Set
    • 3.3.8.1   Fixed-Point Instructions
    • 3.3.8.2   Floating-Point Instructions
    • 3.3.8.3   Vector Instructions
    • 3.3.8.4   Control-Flow Instructions
    • 3.3.8.5   Miscellaneous Instructions
  • 3.3.9   The 970FX Core
    • 3.3.9.1   Instruction Pipeline
    • 3.3.9.2   Branch Prediction
    • 3.3.9.3   Summary
  • 3.3.10   AltiVec
    • 3.3.10.1   Vector Computing
    • 3.3.10.2   The 970FX AltiVec Implementation
    • 3.3.10.3   AltiVec Instructions
  • 3.3.11   Power Management
    • 3.3.11.1   PowerTune
    • 3.3.11.2   Power Mac G5 Thermal and Power Management
  • 3.3.12   64-bit Architecture
    • 3.3.12.1   64-bit Features
    • 3.3.12.2   The 970FX as a 32-bit Processor
  • 3.3.13   Softpatch Facility
• 3.4   Software Conventions
  • 3.4.1   Byte Ordering
  • 3.4.2   Register Usage
    • 3.4.2.1   Indirect Calls
    • 3.4.2.2   Direct Calls
  • 3.4.3   Stack Usage
    • 3.4.3.1   Stack Usage Examples
    • 3.4.3.2   Printing Stack Frames
  • 3.4.4   Function Parameters and Return Values
    • 3.4.4.1   Passing Parameters
    • 3.4.4.2   Returning Values
• 3.5   Examples
  • 3.5.1   A Recursive Factorial Function
  • 3.5.2   An Atomic Compare-and-Store Function
  • 3.5.3   Function Rerouting
    • 3.5.3.1   Instruction Patching
    • 3.5.3.2   Constructing Branch Instructions
  • 3.5.4   Cycle-Accurate Simulation of the 970FX

Chapter 4 The Firmware and the Bootloader

• 4.1   Introduction
  • 4.1.1   Varieties of Firmware
  • 4.1.2   Preferential Storage
• 4.2   A Whole New World
  • 4.2.1   “New” Is Good News
  • 4.2.2   Modern Boot ROM (PowerPC)
• 4.3   Power-On Reset
• 4.4   Open Firmware
  • 4.4.1   Interacting with Open Firmware
    • 4.4.1.1   Forth Shell
    • 4.4.1.2   TELNET
    • 4.4.1.3   TFTP
    • 4.4.1.4   Serial Download
  • 4.4.2   Open Firmware Emulators
• 4.5   Forth
  • 4.5.1   Cells
  • 4.5.2   Stacks
  • 4.5.3   Words
  • 4.5.4   Dictionary
    • 4.5.4.1   A Sampling of Built-in Words
    • 4.5.4.1   Stacks
    • 4.5.4.1   Memory
    • 4.5.4.1   Operators
    • 4.5.4.1   Console I/O
    • 4.5.4.1   Control Flow
    • 4.5.4.2   Searching the Dictionary
  • 4.5.5   Debugging
• 4.6   The Device Tree
  • 4.6.1   Properties
  • 4.6.2   Methods
  • 4.6.3   Data
• 4.7   Open Firmware Interfaces
  • 4.7.1   The User Interface
  • 4.7.2   The Client Interface
  • 4.7.3   The Device Interface
• 4.8   Programming Examples
  • 4.8.1   Dumping NVRAM Contents
  • 4.8.2   Determining Screen Dimensions
  • 4.8.3   Working with Colors
  • 4.8.4   Drawing a Color-Filled Rectangle
  • 4.8.5   Creating an Animated Solution to the “Towers of Hanoi” Problem
  • 4.8.6   Fabricating and Using a Mouse Pointer
  • 4.8.7   Stealing a Font
  • 4.8.8   Implementing a Clock
  • 4.8.9   Drawing Images
  • 4.8.10   Creating Windows
• 4.9   Firmware Boot Sequence
  • 4.9.1   The Script
  • 4.9.2   Snag Keys
• 4.10   BootX
  • 4.10.1   File Format
  • 4.10.2   Structure
  • 4.10.3   Operation
• 4.11   Alternate Booting Scenarios
  • 4.11.1   Booting an Alternate Kernel
    • 4.11.1.1   NVRAM Caveats
  • 4.11.2   Booting from a Software RAID Device
  • 4.11.3   Booting over a Network
• 4.12   Firmware Security
  • 4.12.1   Managing Firmware Security
  • 4.12.2   Recovering the Open Firmware Password
• 4.13   Launching the Kernel
• 4.14   The BootCache Optimization
• 4.15   Boot-Time Kernel Arguments
• 4.16   The Extensible Firmware Interface
  • 4.16.1   Legacy Pains
  • 4.16.2   A New Beginning
  • 4.16.3   EFI
    • 4.16.3.1   EFI Services
    • 4.16.3.1   Boot Services
    • 4.16.3.2   EFI Drivers
  • 4.16.4   A Sampling of EFI
    • 4.16.4.1   EFI NVRAM
    • 4.16.4.2   The Boot Manager
    • 4.16.4.3   The EFI Shell
    • 4.16.4.4   The GUID-Based Partitioning Scheme
    • 4.16.4.5   Universal Graphics Adapter
    • 4.16.4.6   EFI Byte Code
    • 4.16.4.7   Binary Format
  • 4.16.5   The Benefits of EFI

Chapter 5 Kernel and User-Level Startup

• 5.1   Arranging for the Kernel to Execute
  • 5.1.1   Exceptions and Exception Vectors
  • 5.1.2   Kernel Symbols
  • 5.1.3   Run Kernel Run
• 5.2   Low-Level Processor Initialization
  • 5.2.1   Per-Processor Data
  • 5.2.2   Reset Types
  • 5.2.3   Processor Types
  • 5.2.4   Memory Patching
  • 5.2.5   Processor-Specific Initialization
  • 5.2.6   Other Early Initialization
• 5.3   High-Level Processor Initialization
  • 5.3.1   Before Virtual Memory
  • 5.3.2   Low-Level Virtual Memory Initialization
    • 5.3.2.1   Sizing Memory
    • 5.3.2.2   Pmap Initialization
    • 5.3.2.3   Starting Address-Translation
  • 5.3.3   After Virtual Memory
    • 5.3.3.1   Console Initialization
    • 5.3.3.2   Preparing for the Bootstrapping of Kernel Subsystems
• 5.4   Mach Subsystem Initialization
  • 5.4.1   Scheduler Initialization
  • 5.4.2   High-Level Virtual Memory Subsystem Initialization
  • 5.4.3   IPC Initialization
  • 5.4.4   Finishing VM and IPC Initialization
  • 5.4.5   Initializing Miscellaneous Subsystems
  • 5.4.6   Tasks and Threads
  • 5.4.7   Launching the Kernel Bootstrap Thread
• 5.5   The First Thread
• 5.6   I/O Kit Initialization
• 5.7   BSD Initialization
  • 5.7.1   Miscellaneous BSD Initialization (Part 1)
  • 5.7.2   File System Initialization
  • 5.7.3   Miscellaneous BSD Initialization (Part 2)
  • 5.7.4   Networking Subsystem Initialization
  • 5.7.5   Miscellaneous BSD Initialization (Part 3)
  • 5.7.6   Mounting the Root File System
  • 5.7.7   Creating Process 1
  • 5.7.8   Shared Memory Regions
• 5.8   Launching the First User-Space Program
• 5.9   Slave Processors
• 5.10   User-Level Startup
  • 5.10.1   launchd
    • 5.10.1.1   Daemon Configuration and Management
    • 5.10.1.2   Daemon Creation
    • 5.10.1.3   launchd Operation
  • 5.10.2   Multiuser Startup
    • 5.10.2.1   User Login
    • 5.10.2.2   User Logout, System Restart, and System Shutdown
  • 5.10.3   Single-User Startup
  • 5.10.4   Installation Startup

Chapter 6 The xnu Kernel

• 6.1   xnu Source
• 6.2   Mach
  • 6.2.1   Kernel Fundamentals
    • 6.2.1.1   Tasks and Threads
    • 6.2.1.2   Ports
    • 6.2.1.3   Messages
    • 6.2.1.4   Virtual Memory and Memory Objects
  • 6.2.2   Exception Handling
• 6.3   A Flavor of the Mach APIs
  • 6.3.1   Displaying Host Information
  • 6.3.2   Accessing the Kernel’s Clock Services
  • 6.3.3   Using a Clock Service to Ring an Alarm
  • 6.3.4   Displaying Host Statistics
• 6.4   Entering the Kernel
  • 6.4.1   Types of Control Transfer
    • 6.4.1.1   External Hardware Interrupts
    • 6.4.1.2   Processor Traps
    • 6.4.1.3   Software Traps
    • 6.4.1.4   System Calls
  • 6.4.2   Implementing System Entry Mechanisms
    • 6.4.2.1   Exceptions and Exception Vectors
    • 6.4.2.2   Exception-Handling Registers
    • 6.4.2.3   System Linkage Instructions
    • 6.4.2.3   System Call
    • 6.4.2.3   Return from Interrupt
    • 6.4.2.4   Machine-Dependent Thread State
    • 6.4.2.5   Exception Save Areas
• 6.5   Exception Processing
  • 6.5.1   Hardware Interrupts
  • 6.5.2   Miscellaneous Traps
  • 6.5.3   System Calls
• 6.6   System Call Processing
• 6.7   System Call Categories
  • 6.7.1   BSD System Calls
    • 6.7.1.1   Data Structures
    • 6.7.1.2   Argument Munging
    • 6.7.1.3   Kernel Processing of BSD System Calls
    • 6.7.1.4   User Processing of BSD System Calls
  • 6.7.2   Mach Traps
  • 6.7.3   I/O Kit Traps
  • 6.7.4   PowerPC-Only System Calls
  • 6.7.5   Ultra-Fast Traps
    • 6.7.5.1   Fast Traps
    • 6.7.5.2   Blue Box Calls
  • 6.7.6   The Commpage
• 6.8   Kernel Support for Debugging, Diagnostics, and Tracing
  • 6.8.1   GDB (Network-Based or FireWire-Based Debugging)
  • 6.8.2   KDB (Serial-Line-Based Debugging)
  • 6.8.3   CHUD Support
    • 6.8.3.1   Task-Related Functions
    • 6.8.3.2   Thread-Related Functions
    • 6.8.3.3   Memory-Related Functions
    • 6.8.3.4   CPU-Related Functions
    • 6.8.3.5   Callback-Related Functions
  • 6.8.4   Kernel Profiling (kgmon and gprof)
    • 6.8.4.1   Per-Process Profiling (profil(2))
    • 6.8.4.2   Mach Task and Thread Sampling
  • 6.8.5   Per-Process Kernel Tracing (ktrace(2) and kdump)
  • 6.8.6   Auditing Support
  • 6.8.7   Fine-Grained Kernel Event Tracing (kdebug)
    • 6.8.7.1   kdebug Tracing
    • 6.8.7.2   kdebug Entropy Collection
  • 6.8.8   Low-Level Diagnostics and Debugging Interfaces
    • 6.8.8.1   Firmware Call Interface
    • 6.8.8.2   Diagnostics System Call Interface
  • 6.8.9   Low-Level Kernel Tracing
    • 6.8.9.1   Low-Memory Global Data Structures
    • 6.8.9.2   Low Tracing
• 6.9   Virtual Machine Monitor
  • 6.9.1   Features
  • 6.9.2   Using the VMM Facility
  • 6.9.3   Example: Running Code in a Virtual Machine
• 6.10   Compiling the Kernel
  • 6.10.1   Retrieving Prerequisite Packages
  • 6.10.2   Compiling Prerequisite Packages
  • 6.10.3   Compiling the xnu Package
  • 6.10.4   DarwinBuild

Chapter 7 Processes

• 7.1   Processes: From Early UNIX to Mac OS X
  • 7.1.1   Mac OS X Process Limits
  • 7.1.2   Mac OS X Execution Flavors
• 7.2   Mach Abstractions, Data Structures, and APIs
  • 7.2.1   Summary of Relationships
  • 7.2.2   Processor Sets
    • 7.2.2.1   Representation
    • 7.2.2.2   The Processor Set API
  • 7.2.3   Processors
    • 7.2.3.1   Interconnections
    • 7.2.3.2   The Processor API
    • 7.2.3.3   Displaying Processor Information
    • 7.2.3.4   Stopping and Starting a Processor in a Multiprocessor System
  • 7.2.4   Tasks and the Task API
  • 7.2.5   Threads
    • 7.2.5.1   The Thread API
    • 7.2.5.2   Kernel Threads
  • 7.2.6   Thread-Related Abstractions
    • 7.2.6.1   Remote Procedure Call
    • 7.2.6.2   Activation and Shuttle
    • 7.2.6.3   Thread Migration
    • 7.2.6.4   Continuations
• 7.3   Many Threads of a New System
  • 7.3.1   Mach Tasks and Threads
    • 7.3.1.1   Creating a Mach Task
    • 7.3.1.2   Creating a Mach Thread in an Existing Task
    • 7.3.1.3   Displaying Task and Thread Details
  • 7.3.2   BSD Processes
    • 7.3.2.1   The fork() System Call: User-Space Implementation
    • 7.3.2.2   The fork() System Call: Kernel Implementation
    • 7.3.2.3   The vfork() System Call
  • 7.3.3   POSIX Threads (Pthreads)
    • 7.3.3.1   The Pthreads API
    • 7.3.3.2   Implementation of Pthread Creation
  • 7.3.4   Java Threads
  • 7.3.5   The NSTask Cocoa Class
  • 7.3.6   The NSThread Cocoa Class
  • 7.3.7   The Carbon Process Manager
  • 7.3.8   Carbon Multiprocessing Services
  • 7.3.9   The Carbon Thread Manager
• 7.4   Scheduling
  • 7.4.1   Scheduling Infrastructure Initialization
    • 7.4.1.1   Timeslicing Quantum
    • 7.4.1.2   Timing and Clocks
    • 7.4.1.3   Converting between Absolute- and Clock-Time Intervals
    • 7.4.1.4   Starting the Scheduler
    • 7.4.1.5   Retrieving the Value of the Scheduler Tick
    • 7.4.1.6   Some Periodic Kernel Activities
  • 7.4.2   Scheduler Operation
    • 7.4.2.1   Priority Ranges
    • 7.4.2.2   Run Queues
    • 7.4.2.3   Scheduling Information in Tasks and Threads
    • 7.4.2.4   Processor Usage Accounting
  • 7.4.3   Scheduling Policies
    • 7.4.3.1   THREAD_STANDARD_POLICY
    • 7.4.3.2   THREAD_EXTENDED_POLICY
    • 7.4.3.3   THREAD_PRECEDENCE_POLICY
    • 7.4.3.4   THREAD_TIME_CONSTRAINT_POLICY
    • 7.4.3.5   Priority Recomputation on Policy Change
    • 7.4.3.6   Task Roles
• 7.5   The execve() System Call
  • 7.5.1   Mach-O Binaries
    • 7.5.1.1   Preparations for the Execution of a Mach-O File
    • 7.5.1.2   Loading the Mach-O File
    • 7.5.1.3   Handling Setuid and Setgid
    • 7.5.1.4   Execution Notification
    • 7.5.1.5   Configuring the User Stack
    • 7.5.1.6   Finishing Up
  • 7.5.2   Fat (Universal) Binaries
  • 7.5.3   Interpreter Scripts
• 7.6   Launching Applications
  • 7.6.1   Mapping Entities to Handlers
  • 7.6.2   Uniform Type Identifiers

Chapter 8 Memory

• 8.1   Looking Back
  • 8.1.1   Virtual Memory and UNIX
  • 8.1.2   Virtual Memory and Personal Computing
  • 8.1.3   Roots of the Mac OS X Virtual Memory Subsystem
• 8.2   An Overview of Mac OS X Memory Management
  • 8.2.1   Reading Kernel Memory from User Space
    • 8.2.1.1   dd and /dev/kmem
    • 8.2.1.2   The kvm(3) Interface
  • 8.2.2   Querying Physical Memory Size
• 8.3   Mach VM
  • 8.3.1   Overview
  • 8.3.2   Task Address Spaces
  • 8.3.3   VM Maps
  • 8.3.4   VM Map Entries
  • 8.3.5   VM Objects
    • 8.3.5.1   Contents of a VM Object
    • 8.3.5.2   Backing Stores
  • 8.3.6   Pagers
    • 8.3.6.1   External Pagers
    • 8.3.6.2   A Pager’s Port
    • 8.3.6.3   The Mach Pager Interface
  • 8.3.7   Copy-on-Write
  • 8.3.8   The Physical Map (Pmap)
    • 8.3.8.1   The Pmap Interface
• 8.4   Resident Memory
  • 8.4.1   The vm_page Structure
  • 8.4.2   Searching for Resident Pages
  • 8.4.3   Resident Page Queues
  • 8.4.4   Page Replacement
  • 8.4.5   Physical Memory Bookkeeping
  • 8.4.6   Page Faults
• 8.5   Virtual Memory Initialization during Bootstrap
• 8.6   The Mach VM User-Space Interface
  • 8.6.1   mach_vm_map()
  • 8.6.2   mach_vm_remap()
  • 8.6.3   mach_vm_allocate()
  • 8.6.4   mach_vm_deallocate()
  • 8.6.5   mach_vm_protect()
  • 8.6.6   mach_vm_inherit()
  • 8.6.7   mach_vm_read()
  • 8.6.8   mach_vm_write()
  • 8.6.9   mach_vm_copy()
  • 8.6.10   mach_vm_wire()
  • 8.6.11   mach_vm_behavior_set()
  • 8.6.12   mach_vm_msync()
  • 8.6.13   Statistics
• 8.7   Using the Mach VM Interfaces
  • 8.7.1   Controlling Memory Inheritance
  • 8.7.2   Debugging the Mach VM Subsystem
  • 8.7.3   Protecting Memory
  • 8.7.4   Accessing Another Task’s Memory
  • 8.7.5   Naming and Sharing Memory
• 8.8   Kernel and User Address Space Layouts
• 8.9   Universal Page Lists (UPLs)
• 8.10   Unified Buffer Cache (UBC)
  • 8.10.1   The UBC Interface
  • 8.10.2   The NFS Buffer Cache
• 8.11   The Dynamic Pager Program
• 8.12   The Update Daemon
• 8.13   System Shared Memory
  • 8.13.1   Applications of Shared Memory
  • 8.13.2   Implementation of the Shared Memory Server Subsystem
    • 8.13.2.1   shared_region_make_private_np()
    • 8.13.2.2   shared_region_map_file_np()
    • 8.13.2.3   load_shared_file()
    • 8.13.2.4   reset_shared_file()
    • 8.13.2.5   new_system_shared_regions()
  • 8.13.3   The Loading of Shared Object Files by the Dynamic Linker
  • 8.13.4   The Use of shared_region_map_file_np() by a System Application
  • 8.13.5   A Note on Prebinding
• 8.14   Task Working Set Detection and Maintenance
  • 8.14.1   The TWS Mechanism
  • 8.14.2   TWS Implementation
• 8.15   Memory Allocation in User Space
  • 8.15.1   A Historical Break
  • 8.15.2   Memory Allocator Internals
    • 8.15.2.1   Tiny Allocations
    • 8.15.2.2   Small Allocations
    • 8.15.2.3   Large Allocations
    • 8.15.2.4   Huge Allocations
  • 8.15.3   The malloc() Routine
  • 8.15.4   The Largest Single Allocation (32-bit)
  • 8.15.5   The Largest Single Allocation (64-bit)
  • 8.15.6   Enumerating All Pointers
  • 8.15.7   Displaying Scalable Zone Statistics
  • 8.15.8   Logging Malloc Operations
  • 8.15.9   Intercepting the Malloc Layer
• 8.16   Memory Allocation in the Kernel
  • 8.16.1   Page-Level Allocation
  • 8.16.2   kmem_alloc
  • 8.16.3   The Mach Zone Allocator
  • 8.16.4   The Kalloc Family
  • 8.16.5   The OSMalloc Family
  • 8.16.6   Memory Allocation in the I/O Kit
  • 8.16.7   Memory Allocation in the Kernel’s BSD Portion
  • 8.16.8   Memory Allocation in libkern’s C++ Environment
• 8.17   Memory-Mapped Files
• 8.18   64-bit Computing
  • 8.18.1   Reasons for 64-bit Computing
    • 8.18.1.1   32-bit Execution on 64-bit PowerPC
    • 8.18.1.2   Need for Address Space
    • 8.18.1.3   Large-File Support
  • 8.18.2   Mac OS X 10.4: 64-bit User Address Spaces
    • 8.18.2.1   Data Model
    • 8.18.2.2   Implementation
    • 8.18.2.3   Usage and Caveats
  • 8.18.3   Why Not to Use 64-bit Executables
  • 8.18.4   The 64-bit “Scene”

Chapter 9 Interprocess Communication

• 9.1   Introduction
  • 9.1.1   The Evolution of IPC
  • 9.1.2   IPC in Mac OS X
• 9.2   Mach IPC: An Overview
  • 9.2.1   Mach Ports
    • 9.2.1.1   Ports for Communication
    • 9.2.1.2   Port Rights
    • 9.2.1.3   Ports as Objects
    • 9.2.1.4   Mach Port Allocation
  • 9.2.2   Mach IPC Messages
    • 9.2.2.1   Message Header
    • 9.2.2.2   Message Body
    • 9.2.2.3   Message Trailer
• 9.3   Mach IPC: The Mac OS X Implementation
  • 9.3.1   IPC Spaces
    • 9.3.1.1   IPC Entry Table
    • 9.3.1.2   IPC Entry Splay Tree
  • 9.3.2   The Anatomy of a Mach Port
    • 9.3.2.1   What’s in a Port’s Name?
    • 9.3.2.2   Validity of a Port Name
  • 9.3.3   Tasks and IPC
  • 9.3.4   Threads and IPC
  • 9.3.5   Port Allocation
  • 9.3.6   Messaging Implementation
  • 9.3.7   IPC Subsystem Initialization
• 9.4   Name and Bootstrap Servers
  • 9.4.1   The Network Message Server
  • 9.4.2   The Bootstrap Server
    • 9.4.2.1   The Bootstrap Port
    • 9.4.2.2   The Bootstrap Context
    • 9.4.2.3   Debugging the Bootstrap Server
  • 9.4.3   The Bootstrap Server API
    • 9.4.3.1   Displaying Information about All Known Services
    • 9.4.3.2   Creating a Crash-Resistant Server
• 9.5   Using Mach IPC
  • 9.5.1   A Simple Client-Server Example
  • 9.5.2   Dead Names
  • 9.5.3   Port Sets
  • 9.5.4   Interposition
  • 9.5.5   Transferring Out-of-Line Memory and Port Rights
• 9.6   MIG
  • 9.6.1   MIG Specification Files
  • 9.6.2   Using MIG to Create a Client-Server System
  • 9.6.3   MIG in the Kernel
    • 9.6.3.1   Interfaces to Kernel Objects
    • 9.6.3.2   MIG Initialization in the Kernel
• 9.7   Mach Exceptions
  • 9.7.1   Programmer-Visible Aspects of Mach’s Exception-Handling Facility
  • 9.7.2   The Mach Exception-Handling Chain
    • 9.7.2.1   Delivering Exceptions
    • 9.7.2.2   Unresolved Kernel Traps
  • 9.7.3   Example: A Mach Exception Handler
• 9.8   Signals
  • 9.8.1   Reliability
  • 9.8.2   The Number of Signals
  • 9.8.3   Application-Defined Signals
  • 9.8.4   Signal-Based Notification of Asynchronous I/O
  • 9.8.5   Signals and Multithreading
  • 9.8.6   Signal Actions
  • 9.8.7   Signal Generation and Delivery
  • 9.8.8   Mach Exceptions and Unix Signals Living Together
  • 9.8.9   Exceptions, Signals, and Debugging
  • 9.8.10   The ptrace() System Call
• 9.9   Pipes
• 9.10   Named Pipes (Fifos)
• 9.11   File Descriptor Passing
• 9.12   XSI IPC
• 9.13   POSIX IPC
  • 9.13.1   POSIX Semaphores
  • 9.13.2   POSIX Shared Memory
• 9.14   Distributed Objects
• 9.15   Apple Events
  • 9.15.1   Tiling Application Windows Using Apple Events in AppleScript
  • 9.15.2   Building and Sending an Apple Event in a C Program
  • 9.15.3   Causing the System to Sleep by Sending an Apple Event
• 9.16   Notifications
  • 9.16.1   Foundation Notifications
  • 9.16.2   The notify(3) API
  • 9.16.3   Kernel Event Notification Mechanism (kqueue(2))
  • 9.16.4   Core Foundation Notifications
  • 9.16.5   Fsevents
  • 9.16.6   Kauth
• 9.17   Core Foundation IPC
  • 9.17.1   Notifications
  • 9.17.2   The Run Loop
    • 9.17.2.1   Run-Loop Observers
    • 9.17.2.2   Run-Loop Sources
    • 9.17.2.2   CFMachPort
    • 9.17.2.2   CFMessagePort
• 9.18   Synchronization
  • 9.18.1   Interfaces for Atomic Operations
  • 9.18.2   Low-Level Locking
    • 9.18.2.1   Spinlocks
    • 9.18.2.2   Mutexes
    • 9.18.2.3   Read/Write Locks
    • 9.18.2.4   Lock Groups and Attributes
  • 9.18.3   BSD Condition Variables
  • 9.18.4   Mach Lock Sets
  • 9.18.5   Mach Semaphores
  • 9.18.6   Pthreads Synchronization Interfaces
  • 9.18.7   Locking in the I/O Kit
  • 9.18.8   Funnels
    • 9.18.8.1   History
    • 9.18.8.2   Funnels in Mac OS X
  • 9.18.9   SPLs
  • 9.18.10   Advisory-Mode File Locking

Chapter 10 Extending the Kernel

• 10.1   A Driver down the Memory Lane
  • 10.1.1   Driver Programming Considered Difficult
  • 10.1.2   Good Inheritance
  • 10.1.3   Everything Is a File
  • 10.1.4   There Is More to Extending the Kernel Than Driving Devices
• 10.2   The I/O Kit
  • 10.2.1   Embedded C++
  • 10.2.2   I/O Kit Class Hierarchy
  • 10.2.3   I/O Kit Families
  • 10.2.4   I/O Kit Drivers
  • 10.2.5   Nubs
  • 10.2.6   General I/O Kit Classes
    • 10.2.6.1   Classes for Memory-Related Operations
    • 10.2.6.2   Classes for Synchronization and Serialization of Access
    • 10.2.6.3   Miscellaneous Classes
  • 10.2.7   The Work Loop
  • 10.2.8   The I/O Registry
  • 10.2.9   The I/O Catalog
  • 10.2.10   I/O Kit Initialization
  • 10.2.11   Driver Matching in the I/O Kit
• 10.3   DART
• 10.4   Dynamically Extending the Kernel
  • 10.4.1   The Structure of a Kernel Extension
  • 10.4.2   Creation of Kernel Extensions
  • 10.4.3   Management of Kernel Extensions
  • 10.4.4   Automatic Loading of Kernel Extensions
• 10.5   Communicating with the Kernel
• 10.6   Creating Kernel Extensions
  • 10.6.1   A Generic Kernel Extension
  • 10.6.2   Implementing Sysctl Variables Using a Generic Kext
  • 10.6.3   I/O Kit Device Driver Kext
• 10.7   A Programming Tour of the I/O Kit’s Functionality
  • 10.7.1   Rotating a Framebuffer
  • 10.7.2   Accessing Framebuffer Memory
  • 10.7.3   Retrieving the List of Firmware Variables
  • 10.7.4   Retrieving Information about Loaded Kernel Extensions
  • 10.7.5   Retrieving Accelerometer Data from the Sudden Motion Sensor
  • 10.7.6   Listing PCI Devices
  • 10.7.7   Retrieving the Computer’s Serial Number and Model Information
  • 10.7.8   Retrieving Temperature Sensor Readings
  • 10.7.9   Retrieving MAC Addresses of Ethernet Interfaces
  • 10.7.10   Implementing an Encrypted Disk Filter Scheme
• 10.8   Debugging
  • 10.8.1   Kernel Panics
  • 10.8.2   Remote Core Dumps
  • 10.8.3   Logging
  • 10.8.4   Debugging by Using GDB
  • 10.8.5   Debugging by Using KDB
  • 10.8.6   Miscellaneous Debugging Tools
  • 10.8.7   Stabs

Chapter 11 File Systems

• 11.1   Disks and Partitions
  • 11.1.1   The Apple Partitioning Scheme
  • 11.1.2   PC-Style Partitioning
  • 11.1.3   GUID-Based Partitioning
• 11.2   Disk Arbitration
  • 11.2.1   Retrieving a Disk’s Description
  • 11.2.2   Participating in Disk Mounting Decisions
  • 11.2.3   Receiving Media Notifications from the I/O Kit
• 11.3   The Implementation of Disk Devices
• 11.4   Disk Images
  • 11.4.1   Using the hdiutil Program
  • 11.4.2   RAM Disks
  • 11.4.3   The BSD Vnode Disk Driver
  • 11.4.4   Creating a Virtual Disk from Scratch
• 11.5   Files and File Descriptors
• 11.6   The VFS Layer
• 11.7   File System Types
  • 11.7.1   HFS Plus and HFS
  • 11.7.2   ISO 9660
  • 11.7.3   MS-DOS
  • 11.7.4   NTFS
  • 11.7.5   UDF
  • 11.7.6   UFS
  • 11.7.7   AFP
  • 11.7.8   FTP
  • 11.7.9   NFS
  • 11.7.10   SMB/CIFS
  • 11.7.11   WebDAV
  • 11.7.12   cddafs
  • 11.7.13   deadfs
  • 11.7.14   devfs
  • 11.7.15   fdesc
  • 11.7.16   specfs and fifofs
  • 11.7.17   synthfs
  • 11.7.18   union
  • 11.7.19   volfs
• 11.8   Spotlight
  • 11.8.1   Spotlight’s Architecture
  • 11.8.2   The Fsevents Mechanism
  • 11.8.3   Importing Metadata
  • 11.8.4   Querying Spotlight
  • 11.8.5   Spotlight Command-Line Tools
  • 11.8.6   Overcoming Granularity Limitations
• 11.9   Access Control Lists
• 11.10   The Kauth Authorization Subsystem
  • 11.10.1   Kauth Concepts
    • 11.10.1.1   Scopes and Actions
    • 11.10.1.2   Listeners and Authorization Decisions
  • 11.10.2   Implementation
  • 11.10.3   A Vnode-Level File System Activity Monitor

Chapter 12 The HFS Plus File System

• 12.1   Analysis Tools
  • 12.1.1   HFSDebug
  • 12.1.2   Interface for Retrieving File System Attributes
  • 12.1.3   Mac OS X Command-Line Tools
  • 12.1.4   HFS+ Source and Technical Note TN1150
• 12.2   Fundamental Concepts
  • 12.2.1   Volumes
  • 12.2.2   Allocation Blocks
  • 12.2.3   Extents
  • 12.2.4   File Forks
  • 12.2.5   Clumps
  • 12.2.6   B-Trees
    • 12.2.6.1   B+ Trees in HFS+
    • 12.2.6.2   Nodes
    • 12.2.6.3   Records
    • 12.2.6.4   Searching
• 12.3   The Structure of an HFS+ Volume
• 12.4   Reserved Areas
• 12.5   The Volume Header
  • 12.5.1   Viewing the Volume Header
  • 12.5.2   Viewing a Volume Control Block
• 12.6   The HFS Wrapper
• 12.7   Special Files
  • 12.7.1   The Allocation File
    • 12.7.1.1   Viewing the Contents of the Allocation File
    • 12.7.1.2   The Roving Next-Allocation Pointer
  • 12.7.2   The Catalog File
    • 12.7.2.1   Catalog Node IDs
    • 12.7.2.2   Examining the Catalog B-Tree
  • 12.7.3   The Extents Overflow File
    • 12.7.3.1   Examining Fragmentation
    • 12.7.3.2   Examining the Extents Overflow B-Tree
  • 12.7.4   The Attributes File
    • 12.7.4.1   Working with Extended Attributes
    • 12.7.4.2   Examining the Attributes B-Tree
  • 12.7.5   The Startup File
• 12.8   Examining HFS+ Features
  • 12.8.1   Case Sensitivity
  • 12.8.2   Filename Encodings
  • 12.8.3   Permissions
    • 12.8.3.1   Manipulating Volume-Level Ownership Rights
    • 12.8.3.2   Repairing Permissions
  • 12.8.4   Journaling
    • 12.8.4.1   Enabling or Disabling Journaling on a Volume
    • 12.8.4.2   Observing the Journal’s Operation
  • 12.8.5   Quotas
  • 12.8.6   Hard Links
  • 12.8.7   Unlinking Open Files
  • 12.8.8   Symbolic Links
  • 12.8.9   Aliases
  • 12.8.10   Resource Forks
• 12.9   Optimizations
  • 12.9.1   On-the-Fly Defragmentation
  • 12.9.2   The Metadata Zone
  • 12.9.3   Hot File Clustering
    • 12.9.3.1   Hot File Clustering Stages
    • 12.9.3.2   The Hot Files B-Tree
    • 12.9.3.3   The Working of Hot File Clustering
• 12.10   Miscellaneous Features
  • 12.10.1   Special System Calls
  • 12.10.2   Freezing and Thawing a Volume
  • 12.10.3   Extending and Shrinking a Volume
  • 12.10.4   Volume Notifications
  • 12.10.5   Support for Sparse Devices
• 12.11   Comparing Mac OS X File Systems
• 12.12   Comparing HFS+ and NTFS
• 7.6   A New I/O Kit Plane

  Appendix A Mac OS X on x86-Based Macintosh Computers

  • A.1   Hardware Differences
  • A.2   Firmware and Booting
  • A.3   Partitioning
  • A.4   Universal Binaries
  • A.5   Rosetta
  • A.6   Byte Ordering
  • A.7   Miscellaneous Changes
    • A.7.1   No Dual-Mapped Kernel Address Space
    • A.7.2   Nonexecutable Stack
    • A.7.3   Thread Creation
    • A.7.4   System Calls
    • A.7.5   No /dev/mem or /dev/kmem
    • A.7.6   A New I/O Kit Plane

All contents of this site, unless otherwise noted, are ©1994-2010 Amit Singh. All Rights Reserved. Terms of Use