Ch'3 Embedded Linux System

1
??? ????? ? ??

• Ch.3 Embedded Linux System

2
Windows vs. Linux
Internet Explorer
Mozilla
Winamp
Outlook Express
Kmail
Xmms
???? ?? 2002
??2002
Windows Media Player
Mplayer
3
Embedded System

• SoC
• CPU ASIC Software
• Traditional application-specific systems
• Implemented by algorithm-to-architecture
  methodology.
• Modern Trend
• Applications of the embedded systems become more
  complex.
• Design Reuse
• Software plays an important role.
• Real-time operating systems ( RTOS ) is the
  critical component.

4
Embedded System

• A new system-level synthesis flow

Compilation
Algorithm Development
Process Management
System Design
Inter process Communication
RTOS
Memory Management
I/O Management
Architecture
IC Design
5
Embedded System

• Embedded System Characteristics
• Change many of the assumptions underlying
  conventional computer system design
• Resource limitations
• No hard disk, small memory, battery-operated, low
  performance CPU,
• Real-time characteristics
• Real-time system requires finer resolution time
  keeping and execution time prediction because
  they must control when actions occur
• Precise control of events on real-time line
• Execution time predictions are required in many
  cases
• Application-specific and often special purpose
• Application semantics differ widely
• Sometimes, single user, single application but
  with multiple external interactions
• Majority of all computers (80) are embedded
• Increasing number must satisfy real-time
  constraints

6
Embedded/Real-Time OS

• According to Japan ITRON survey in 1998
• 75 of embedded systems use operating systems
• 95 of embedded systems with program size larger
  than 1 MB employ operating systems
• gt System software such as operating systems and
  middleware are must for complex and sophisticated
  embedded systems
• Real-Time System OS Characteristic
• Hardware abstraction
• Common to all operating systems
• Without this, to print Hello, World. in a
  display, application programmers must code for
  video card initialization, memory management, and
  video card manipulation.
• Multitasking concurrency management
• Handle multiple task with independent events and
  action
• Task create, schedule, execute, terminate and
  destroy
• Support for context save and restore
• Task execution image task code, data segment,
  stack, heap
• Segment table, page table, file descriptors, and
  file tables
• Hardware register context CPU registers

7
Real-Time OS

• Predictable latency guarantee for context
  switching
• A real-time OS must meet the following conditions
• Fully preemptible kernel
• Priority-based scheduling
• Both fully preemptible and priority-based
  scheduling are required to assign CPU to an
  important task with hard deadline
• Bounded delay for interrupt latency
• Bounded delay for scheduling latency

8
Linux as a Real-time OS

• Why Linux for Embedded system?
• Main characteristics
• Multi-processor, multi-thread, multi-tasking
• Multi-user support
• Dynamic modules
• Support for different file system (ext2, ..)
• Networking (TCP/IP, SLIP, PPP)
• Support various architecture (X86, Sparc, ARM,
  MIPS, Alpha, SuperH, PowerPC ..)
• SMP support
• Key attractions
• Free no license fee
• Open source
• Direct access to source codes allow several
  optimizations and functional additions which are
  not possible in commercial OS
• Support from many individual programmers
• Reliability, scalability, portability,
  standardization

9
Linux as a Real-time OS

• Why Linux for Embedded system?
• Drawbacks
• Monolithic kernel (performance-oriented)
• Not for beginners (for system programmers)
• Not well structured (not designed on the drawing
  board)
• Not useful for real-time systems
• Non-preemptible kernel
• Preemptible scheduling only in user mode, but not
  in kernel mode
• Fairness scheduling, with limited priority
  scheduling
• Throughput-oriented nature from PCs and servers
• Unpredictable, sometimes high latency
• No bounded delay guarantee for interrupt latency
  as well as scheduling latency
• Non-deterministic behavior
• Coarse timing resolution

10
Linux as a Real-time OS

• Real-Time Linux
• Modifications or extensions to standard Linux to
  allow soft or hard real-time applications to meet
  timing constraints.
• Three approaches both open source and commercial
  versions
• Dual-kernel approach
• One approach uses a separate real-time kernel
  that runs Linux as the lowest priority task.
• No modification of Linux kernel
• RT Linux (FSMLabs, 1995), RTAI (DIAPM)
• Preemption point insertion approach
• Insert preemption point (rescheduling routine) to
  all the locations that may cause long delay
  inside the kernel
• This rescheduling routine inspects whether there
  is any higher-priority task pending
• Low latency patch (Ingo Molnar, Andrew Morton)
• Fully preemptible kernel approach
• Another approach modifies standard Linux for
  preemptibility, latency, timing, and/or
  scheduling.
• Montavista Linux

11
Linux as a Real-time OS

• Dual Kernel Approach
• RTLinux and RTAI
• Standard Linux becomes the lowest priority task
  for the real-time kernel (microkernel)
• All real-time processes are run on the
  microkernel while non real-time processes such as
  editor and GUI are run on the standard Linux
  kernel
• Benefits
• No modifications of the Linux kernel
• Ultra low-latency interrupts and context switches
  (5 - 10 mSecs)
• Guaranteed, deterministic scheduling

12
Linux as a Real-time OS

• Dual Kernel Approach
• Limitations
• All real-time features must be implemented as
  kernel modules
• Real-time application development is much more
  difficult
• Real-time applications cannot utilize system
  services that are provided by the standard Linux,
  such as file systems and network support
• Must be very knowledgeable of the Linux kernel
  and device driver interaction
• No memory-protection available
• Improper coding will crash the kernel
• More difficult to debug, lack of easy-to-use
  tools
• Linux system calls are not preemptible and
  performance cannot be guaranteed

13
Linux as a Real-time OS

• A "fully" preemptible Linux kernel
• Allows the kernel to be preempted at any time
  when it is not locked
• This is supported by SMP lock and unlock
• All the data structures used by the kernel are
  protected by the SMP lock/unlock mechanism when
  more than one processes are executed in the
  kernel mode.
• Benefits
• Real-time application development is much easier
• Memory protection is available
• Real-time programs will not crash the kernel
• Full access to system services (TCP/IP, file I/O,
  X-Windows, etc.)
• Extensive tool support for easier debugging

14
Linux as a Real-time OS

• A "fully" preemptible Linux kernel
• Limitations
• Performance may not be good enough for
  low-latency requirements in the microseconds
• Worst-case interrupt latency is unknown since all
  possible code paths have not been tested
• Reduced overall system throughput
• Substantial source code modification is made to
  the kernel
• Extensive testing is needed with every new
  release of the kernel
• Preemptibility analysis required with all device
  drivers and kernel modules

15
Linux Kernel

• Linux Kernel Architecture

16
Linux Kernel

• Linux Source Tree

17
Linux Kernel

• Main subdirectory (relative pathname to
  /usr/src/linux)
• arch/
• Architecture dependent codes arch/i386,
  arch/alpha, .
• arch/i386/boot/
• bootstrapping
• configure devices, memory, and machine specific
  modules
• arch/i386/kernel/
• kernel entry point handling (trap/interrupt
  handling)
• context switch
• arch/i386/mm/
• machine dependent memory management code
• init/
• All the functions needed to start the kernel
  (start_kernel)
• Hand-made process 0 (init_task or task0)
• Fork process 1, 2, 3, ...

18
Linux Kernel

• Main subdirectory
• kernel/ (arch/i386/kernel)
• Central section of the kernel
• All the routines related to task management such
  as process creation, termination, execution,
  scheduling, signal handling, etc.
• Main system call implementation (fork, exit,
  etc.)
• Time management
• Scheduler
• Signal handling
• mm/
• Virtual memory interface
• Paging, kernel memory management
• fs/
• Virtual file system interface
• Implementations of the various file systems
  (ext2, proc, nfs,...)

19
Linux Kernel

• Main subdirectory
• drivers/
• Device drivers for hardware components
• drivers/block/ block-oriented driver(hard
  disks)
• drivers/cdrom/ proprietary CD-ROM drives
• drivers/char/ character-oriented driver (serial
  ports, tty, modem, ..)
• drivers/net network cards
• drivers/pci/ PCI bus access and control
• drivers/scsi/ SCSI interface
• drivers/sound/ sound card drivers
• ipc/
• Classical inter-process communication
• Semaphores, message queues

20
Linux Kernel

• Main subdirectory
• net/
• Various network protocol implementations
  TCP/IP, ARP, ...
• Code for sockets to the UNIX and Internet domains
• lib/
• Some standard kernel library functions (printk)
• include/
• Commonly included kernel-specific header files
• include/asm-i386/ architecture-dependent header
  files for Intel CPU
• include/linux/ Linux kernel internal structure
  (task, inode)

21
Linux Kernel

• Kernel Configuration and Compile
• New kernel is generated in three steps
• Kernel configuration (see Documentation/Configurat
  ion.help or chapter 3 of The LINUX Network)
• make config (menuconfig, xconfig)
• Informs Linux kernel of hardware characteristics,
  kernel components, network characteristics that
  exist in the system
• Kernel compile
• make dep (make clean)
• Checks dependencies among source files
• make bzImage
• Compile and generate executable kernel image
• Kernel installation (module installation and
  inform LILO (Linux loader) of the new kernel
  location
• To be able to boot the system from the new kernel
• make modules
• make modules_install
• cp arch/i386/boot/bzimage /boot/vmlinx-new
• vi /etc/lilo.conf

22
Process Management

• What is the process?
• An instance of a program in execution
• A dynamic and active entity
• Program
• A passive entity stored as an executable binary
  image consisting of code and data sections in the
  non-volatile memory such as HDD
• The basic unit of execution and scheduling.
• A process is named using its process ID (PID).
• Linux process management
• Responsible for the following activities in
  connection with process management
• Process creation and deletion
• Process state transition
• Process context switching
• THE core component of the Linux Kernel

23
Process Management

• Process manager

24
Process Management

• Process creation
• One process can create another process by calling
  a system call fork()
• Parent-child relationship
• Parent may inherit all or a part of resources and
  privileges for its children
• Child duplicates the parents address space or
  has a program loaded into it.
• All the processes are created by calling a fork()
  from a parent process except pid 0 which is
  created by the kernel during system
  initialization
• This pid 0 forks init process (pid 1) and kflushd
  process (pid 2) using the fork system call
• Execution
• A newly created process can be executed by
  calling a system call execve()
• Parent process may either wait for it to finish,
  or it may continue in parallel.

25
Process Management

• Process termination
• Voluntary
• Normal exit
• Process finished its execution
• Error exit
• Involuntary
• Fatal error
• Exceed allocated resources
• Segmentation fault
• Protection fault, etc.
• Killed by another process
• By receiving a signal

26
Process State and Transition

• Running
• Executing on CPU
• Ready
• Ready to be executed (on a timeout or from a
  wakeup)
• Waiting on a run queue
• Waiting
• Waiting until a specific event occurs (by either
  a sleep or lock request, or IO request)
• Zombie

27
Process Information

• Processs context
• task_struct in Linux
• Each task_struct contains all the information
  about a process.
• Task structure
• Identifier pid, pgrp, session, leader
• State TASK_RUNNING, TASK_ZOMBIE...
• Process relationship p_opptr, p_pptr, p_yptr
• Scheduling information policy, priority,
  counter, rt_priority, need_resched
• Signal information signal_struct , sigpendign,
  signal , blocked
• Memory information page table (address
  translation, access information, privilege
  information, etc)
• Files information fs, files, .. (file
  descriptors and file tables)
• Time information start_time, execution time in
  user/kernel mode, ...
• Resource limits
• Tasks binary executable image
• Text, static global data, heap, stack
• Hardware register context
• PC, general-purpose registers

28
Process Scheduling

• Linux uses two process-scheduling algorithms
• Real-time scheduling for tasks where absolute
  priorities are more important than fairness
• Time-sharing based scheduling for fair preemptive
  scheduling between multiple processes
• Real-time scheduling
• Linux implements the FIFO and round-robin
  real-time scheduling classes
• The scheduler selects the process with the
  highest priority
• SCHED_FIFO
• For equal-priority nonpreemptive processes
• The scheduler selects the longest-waiting one
• SCHED_RR
• For preemptive processes
• The scheduler preempts the currently running
  process when a higher-priority (higher value of
  rt_priority) process appears
• Time-sharing scheduling
• SCHED_OTHER Time sharing scheduling for non-real
  time processes

29
Linux file systems

• Linux file systems
• To the user
• Linux file system appears as a hierarchical
  directory tree
• Linux file system sees a hard disk as a set of
  logical disk blocks
• Linux file appears as a storage for a stream of
  bytes
• History
• Linux file system first uses the Minix file
  system (1991)
• Max. file size 64MB, max. file name 14
  characters
• Ext The first native Linux file system (1992)
• Ext2 The second version of native Linux file
  system (1993)
• The virtual file system (VFS)
• Provide an interface between Linux OS and actual
  file systems used
• Linux VFS can support different file systems
  (i.e. different file formats) used in different
  computer systems
• The Linux VFS is designed around object-oriented
  principles
• A set of definitions
• The inode-object and the file-object structures
  represent individual files
• The file system object represents an entire file
  system
• A layer of software to manipulate those objects.

30
Linux File Systems

• Disk-based file systems
• Ext2 Linux native
• Ufs BSD
• Fat DOS FS
• Vfat Win95
• minix the first file system of Linux
• Iso9660 CD-ROM
• Sysv System V
• Hfs Macintosh
• Affs Amiga
• Ntfs NTs FS
• Adfs Acorn-strongarm
• Network file systems
• Nfs
• Coda
• afs Andrew FS
• Smbfs LanManager
• Ncpfs Novell
• Special ones

31
Linux file systems

• Linux file systems

32
Linux file systems

• inode structure
• inode
• Information about each file
• A file has its inode
• inode information
• The type of the file (i_mode)
• regular, directory, block special, character
  special, etc
• File access permissions
• The file seze (i_size)
• The owner and group ids (i_uid, i_gid)
• The creation, last modification and last access
  times (i_ctime, i_mtime, i_atime)
• Device numbers for special file
• The number of the symbolic links pointing to this
  file
• The location of blocks (index block)
• Non-contiguous block allocation for a file
• 3 direct block addresses
• 12 indirect block addresses
• Assuming 4B pointer and 4KB block, it can support
  a file size up to 48KB 4MB 4GB 4TB

33
Linux file systems

• inode structure

1
2
3
S_IFSOCK S_IFLNK S_IFREG S_IFBLK S_IFDIR S_IFCHR S
_IFIFO
read/write/execution 1 users 2 group 3
other users
u set user id g set group id s sticky
34
Linux file systems

• The Virtual File System (VFS)
• Present a uniform interface to user processes
  regardless of the actual file system used
• System calls sent to the VFS
• VFS references a driver for the actual file
  system on that partition
• File system driver uses the disk cache
• 3 main objects
• Superblock object
• Information about the entire file system mounted
• Inode object
• All the necessary information about a file
• File object
• Information regarding the interaction between a
  process and files opened

35
Linux file systems

• A Logical Diagram of the Virtual File System

User space
Inode Cache
VFS
Directory Cache
NTFS
CryptFS
Ext2
Buffer Cache
Maintained by the kernel
Disk Drivers
36
Linux file systems

• Mounting an FS
• To use a FS
• You need to mount it
• This is the process of telling the VFS enough
  information to create the VFS superblock
• Performance differences occur in this area
• Ext2 has similar structure to VFS and is very
  popular
• Other FS need to map their structure to
  elements VFS can understand
• If a certain block device cannot supply this
  information, or if support is not in the kernel,
  the mount fails.

37
Linux Device Drivers

• Device drivers
• Intended to control a peripheral device
• Has interfaces both to the file system and to
  device hardware
• Provides an abstraction to physical hardware
  devices
• Characteristics of Linux device drivers
• An essential part of a Linux kernel
• Must be integrated into the OS kernel
• A bad device driver can corrupt or destroy the
  entire system
• Loadable
• Implemented as a kernel module
• Loaded when necessary and unloaded if not
• Configurable
• Can be included or excluded at compile time

38
Linux Device Drivers

• Types of device drivers
• Character device drivers
• Process individual bytes and transmit data of any
  arbitrary size
• Process sequentially (Terminal, Keyboard, Sound
  Card, Scanner, Printer , Network Card)
• Block device drivers
• Random access and transmit data in blocks
• Accesses are handled transparently by the buffer
  cache
• Process groups of bytes, i.e. (blocks)(hard
  disks, CD-ROM drives, )
• Network device drivers

39
Linux Device Drivers

• Major Minor Number
• Link to the device
• 1Byte each
• Major
• Identifies the driver associated with the device
• Indexes blkdevs or chrdevs
• Minor
• Only used by the device driver
• Indexes tables within the driver to identify the
  virtual device

40
Linux Device Drivers

• Linux Major Number List

Source code include/linux/major.h
41
Linux Networks

• The Linux TCP/IP Networking Layers

42
Linux Networks

• The BSD Socket Interface
• Linux supports the following socket address
  families or domains
• UNIX
• Unix domain sockets
• INET
• The Internet address family supports
  communications via TCP/IP
• AX25
• Amateur radio X25
• IPX
• Novell IPX
• APPLETALK
• X25

43
Linux Networks

• The INET Socket layer
• Creating a BSD Socket
• Binding an Address to an INET BSD Socket
• Making a Connection on an INET BSD Socket
• Listening on an INET BSD Socket
• Accepting Connection Requests
• TCP/UDP layer
• Offer two types of service
• Transmission Control Protocol (TCP)
• Connection-oriented service
• All packets will be delivered (flow control)
• User datagram Protocol (UDP)
• Connectionless service
• No guarantee
• packets are stored and forwarded one at a time

44
Linux Networks

• The IP Layer
• IP address (4 bytes)
• Socket Buffers
• Receiving IP Packets
• Sending IP Packets
• Data Fragmentation
• The Address Resolution Protocol (ARP)
• the task of the ARP is to convert the abstract IP
  address into real hardware addresses.

45
Linux module programming

• The motivation of modules
• Linux kernel is a monolithic kernel
• The limitation of a monolithic kernel
• When the configuration is changed, the kernel has
  to be compiled.
• Rarely used file systems and drivers occupy
  memory permanently.
• The kernel code modification results in creating
  the new kernel and rebooting the system
• lead to the development of modules

46
Linux module programming

• What is a module
• Linux module
• is a functional unit such as file system or
  driver
• can be dynamically linked into the kernel at any
  point after the system has booted.
  
  
  
  
  
• can be removed from the kernel when it is no
  longer needed.
• When a module is loaded, it is same as a kernel.
  

result in small and compact kernel size and the
ability of trying out new kernel code without
rebuilding and rebooting the kernel
47
Linux module programming

• Module data structure
• The loaded modules are maintained by module data
  structure.
• Each module data structure points to the symbol
  table structure that contains all symbols of the
  module.
• The kernel keeps all of the kernels resources in
  the kernel symbol table
• the loaded modules are able to use all kernel
  resources.
• The kernel symbol table is pointed the first
  module data structure.

48
Linux module programming

• The list of kernel modules

49

Linux module programming

• Loading a module
• Manual way
• Using insmod command.
• automatic way
• the kernel discovers the need for loading a
  module (for example, user mounts a file system)
• requests kerneld daemon to load the needed
  module.
• kerneld loads module using insmod.

50
Linux module programming

• The insmod mechanism
• Reading the module into its virtual memory
• Fixes up unresolved references to kernel routines
  and resources using the exported symbols from the
  kernel.
• Requests the kernel for enough space to hold the
  new kernel
• The kernel
• Allocates a new module data structure
• Allocates enough kernel memory to hold the new
  module
• Puts it at the end of the kernel modules list

51
Linux module programming

• The insmod mechanism
• Copies the module into the allocated space
  relocates it
• it will run from the kernel address that it has
  been allocated
• The new module exports sysmbols to the kernel and
  insmod builds a table of these exported symbols
• If the new module depends on another module
• Module has the reference of the new module.
• The kernel calls the modules initialization
  routine and carries on installing the module

52
Linux module programming

• Unloadig a module
• Manual way
• uses rmmod command
• automatic way
• when idle timer expires
• the kernel calls the service routines for all
  unused loaded modules
• The mechanism of unloading
• If the module can be unloaded
• its cleanup routine is called to free up the
  kernel resources that it has allocated
• The module data structure is unlinked from the
  list of kernel modules
• All of the kernel memory that the module needed
  is deallocated.

53
Linux ??? - ???

• man
• ??? ???? ???? ???, ???, ?? ??, ?? ??, ??, ??? ??
  ??? ? ??.
• user_at_host directory man ???
• cd?? ???? ?? ???? ??? ?? man cd
• ??? ???? ???? ??(section)? ??? ?? ??? ??
• ?? ???? ?? ??? ???? ??? ??, ?? ??? ??? ?? ??? ???
  ??? ?? ?? ??
• man passwd
• man 1 passwd
• man 5 passwd
• info
• GNU ??? ????
• Hypertext? ?? ??? ?? ??
• info passwd

54
Linux ??? ?? ??

• ls
• ?? ??? directory ?? file? directory? ??? ?? ?? ??
• ls
• ls --help or man ls
• cd
• directory ?? ???
• ? ?? ?? ??
• ???? ? ????
• /? ?? ?? ??
• / ??? ?? ?? ??
• cd ???? ?
• cd ???? ? ????? ??
• cd /aaa ? ???? ?? aaa ????? ??
• cd - ?? ?? ????? ??

55
Vi ???

• ????

56
????

• man ???? ?? ? ???? ?? ? ??? ????
• ??? ?? ?? ? ???? ??, ??
• vi ??? ???? Hello World!! ?????