Chapter 18: The Linux System

 Linux History
 Design Principles
Chapter 18:  Kernel Modules

The Linux System 


Process Management
Scheduling
 Memory Management
 File Systems
 Input and Output
 Interprocess Communication
 Network Structure
 Security

Operating System Concepts – 9th Edition Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.2 Silberschatz, Galvin and Gagne ©2013

Objectives History
 Linux is a modern, free operating system based on UNIX
 To explore the history of the UNIX operating system from standards
which Linux is derived and the principles upon which Linux’s
design is based  First developed as a small but self-contained kernel in 1991
by Linus Torvalds, with the major design goal of UNIX
 To examine the Linux process model and illustrate how Linux compatibility, released as open source
schedules processes and provides interprocess  Its history has been one of collaboration by many users from
communication all around the world, corresponding almost exclusively over
 To look at memory management in Linux the Internet
 To explore how Linux implements file systems and manages  It has been designed to run efficiently and reliably on
I/O devices common PC hardware, but also runs on a variety of other
platforms
 The core Linux operating system kernel is entirely original,
but it can run much existing free UNIX software, resulting in
an entire UNIX-compatible operating system free from
proprietary code
 Linux system has many, varying Linux distributions
including the kernel, applications, and management tools

Operating System Concepts – 9th Edition 18.3 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.4 Silberschatz, Galvin and Gagne ©2013
 The Linux Kernel Linux 2.0
 Version 0.01 (May 1991) had no networking, ran only on 80386-  Released in June 1996, 2.0 added two major new capabilities:
compatible Intel processors and on PC hardware, had extremely  Support for multiple architectures, including a fully 64-bit native Alpha
limited device-drive support, and supported only the Minix file port
system  Support for multiprocessor architectures
 Linux 1.0 (March 1994) included these new features:  Other new features included:
 Support for UNIX’s standard TCP/IP networking protocols  Improved memory-management code
 BSD-compatible socket interface for networking programming  Improved TCP/IP performance
 Device-driver support for running IP over an Ethernet  Support for internal kernel threads, for handling dependencies between
loadable modules, and for automatic loading of modules on demand
 Enhanced file system  Standardized configuration interface
Support for a range of SCSI controllers for

 Available for Motorola 68000-series processors, Sun Sparc
high-performance disk access systems, and for PC and PowerMac systems
 Extra hardware support
 2.4 and 2.6 increased SMP support, added journaling file system,
 Version 1.2 (March 1995) was the final PC-only Linux kernel preemptive kernel, 64-bit memory support
 Kernels with odd version numbers are development kernels,  3.0 released in 2011, 20th anniversary of Linux, improved
those with even numbers are production kernels virtualization support, new page write-back facility, improved
memory management, new Completely Fair Scheduler

Operating System Concepts – 9th Edition 18.5 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.6 Silberschatz, Galvin and Gagne ©2013

The Linux System Linux Distributions

 Linux uses many tools developed as part of Berkeley’s BSD
 Standard, precompiled sets of packages, or distributions,
operating system, MIT’s X Window System, and the Free
include the basic Linux system, system installation and
Software Foundation's GNU project
management utilities, and ready-to-install packages of common
 The main system libraries were started by the GNU project, with UNIX tools
improvements provided by the Linux community
 The first distributions managed these packages by simply
 Linux networking-administration tools were derived from 4.3BSD providing a means of unpacking all the files into the appropriate
code; recent BSD derivatives such as Free BSD have borrowed places; modern distributions include advanced package
code from Linux in return management
 The Linux system is maintained by a loose network of developers  Early distributions included SLS and Slackware
collaborating over the Internet, with a small number of public ftp
 Red Hat and Debian are popular distributions from
sites acting as de facto standard repositories
commercial and noncommercial sources, respectively,
 File System Hierarchy Standard document maintained by the others include Canonical and SuSE
Linux community to ensure compatibility across the various
 The RPM Package file format permits compatibility among the
system components
various Linux distributions
 Specifies overall layout of a standard Linux file system, determines
under which directory names configuration files, libraries, system
binaries, and run-time data files should be stored

Operating System Concepts – 9th Edition 18.7 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.8 Silberschatz, Galvin and Gagne ©2013
 Linux Licensing Design Principles
 The Linux kernel is distributed under the GNU General Public  Linux is a multiuser, multitasking system with a full set of
License (GPL), the terms of which are set out by the Free UNIX-compatible tools
Software Foundation
 Its file system adheres to traditional UNIX semantics, and it
 Not public domain, in that not all rights are waived fully implements the standard UNIX networking model
 Anyone using Linux, or creating their own derivative of Linux,  Main design goals are speed, efficiency, and standardization
may not make the derived product proprietary; software
 Linux is designed to be compliant with the relevant POSIX
released under the GPL may not be redistributed as a binary-
documents; at least two Linux distributions have achieved
only product
official POSIX certification
 Can sell distributions, but must offer the source code too
 Supports Pthreads and a subset of POSIX real-time
process control
 The Linux programming interface adheres to the SVR4 UNIX
semantics, rather than to BSD behavior

Operating System Concepts – 9th Edition 18.9 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.10 Silberschatz, Galvin and Gagne ©2013

Components of a Linux System Components of a Linux System

 Like most UNIX implementations, Linux is composed of three

main bodies of code; the most important distinction between
the kernel and all other components.
 The kernel is responsible for maintaining the important
abstractions of the operating system
 Kernel code executes in kernel mode with full access to all
the physical resources of the computer
 All kernel code and data structures are kept in the same
single address space

Operating System Concepts – 9th Edition 18.11 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.12 Silberschatz, Galvin and Gagne ©2013
 Components of a Linux System (Cont.) Kernel Modules
 The system libraries define a standard set of functions  Sections of kernel code that can be compiled, loaded, and
through which applications interact with the kernel, and which unloaded independent of the rest of the kernel.
implement much of the operating-system functionality that
 A kernel module may typically implement a device driver, a file
does not need the full privileges of kernel code
system, or a networking protocol
 The system utilities perform individual specialized
 The module interface allows third parties to write and distribute, on
management tasks
their own terms, device drivers or file systems that could not be
 User-mode programs rich and varied, including multiple distributed under the GPL.
shells like the bourne-again (bash)
 Kernel modules allow a Linux system to be set up with a standard,
minimal kernel, without any extra device drivers built in.
 Four components to Linux module support:
 module-management system
 module loader and unloader
 driver-registration system
 conflict-resolution mechanism

Operating System Concepts – 9th Edition 18.13 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.14 Silberschatz, Galvin and Gagne ©2013

Module Management Driver Registration

 Supports loading modules into memory and letting them talk  Allows modules to tell the rest of the kernel that a new driver
to the rest of the kernel has become available
 Module loading is split into two separate sections:  The kernel maintains dynamic tables of all known drivers, and
provides a set of routines to allow drivers to be added to or
 Managing sections of module code in kernel memory
removed from these tables at any time
 Handling symbols that modules are allowed to reference
 Registration tables include the following items:
 The module requestor manages loading requested, but
 Device drivers
currently unloaded, modules; it also regularly queries the
kernel to see whether a dynamically loaded module is still in  File systems
use, and will unload it when it is no longer actively needed  Network protocols
 Binary format

Operating System Concepts – 9th Edition 18.15 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.16 Silberschatz, Galvin and Gagne ©2013
 Conflict Resolution Process Management
 A mechanism that allows different device drivers to reserve  UNIX process management separates the creation of
hardware resources and to protect those resources from processes and the running of a new program into two distinct
accidental use by another driver. operations.
 The conflict resolution module aims to:  The fork() system call creates a new process
 Prevent modules from clashing over access to hardware  A new program is run after a call to exec()
resources
 Under UNIX, a process encompasses all the information that
 Prevent autoprobes from interfering with existing device the operating system must maintain to track the context of a
drivers single execution of a single program
 Resolve conflicts with multiple drivers trying to access the  Under Linux, process properties fall into three groups: the
same hardware: process’s identity, environment, and context
1. Kernel maintains list of allocated HW resources

2. Driver reserves resources with kernel database first

3. Reservation request rejected if resource not available

Operating System Concepts – 9th Edition 18.17 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.18 Silberschatz, Galvin and Gagne ©2013

Process Identity Process Environment

 Process ID (PID) - The unique identifier for the process; used to  The process’s environment is inherited from its parent, and is
specify processes to the operating system when an application composed of two null-terminated vectors:
makes a system call to signal, modify, or wait for another process
 The argument vector lists the command-line arguments
 Credentials - Each process must have an associated user ID used to invoke the running program; conventionally starts
and one or more group IDs that determine the process’s rights to
with the name of the program itself.
access system resources and files
 Personality - Not traditionally found on UNIX systems, but under  The environment vector is a list of “NAME=VALUE” pairs
Linux each process has an associated personality identifier that that associates named environment variables with arbitrary
can slightly modify the semantics of certain system calls textual values.
 Used primarily by emulation libraries to request that system  Passing environment variables among processes and inheriting
calls be compatible with certain specific flavors of UNIX variables by a process’s children are flexible means of passing
 Namespace – Specific view of file system hierarchy information to components of the user-mode system software.
 Most processes share common namespace and operate on a  The environment-variable mechanism provides a customization
shared file-system hierarchy of the operating system that can be set on a per-process basis,
 But each can have unique file-system hierarchy with its own rather than being configured for the system as a whole.
root directory and set of mounted file systems

Operating System Concepts – 9th Edition 18.19 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.20 Silberschatz, Galvin and Gagne ©2013
 Process Context Process Context (Cont.)

 The (constantly changing) state of a running program at any  Whereas the file table lists the existing open files, the
point in time file-system context applies to requests to open new files
 The scheduling context is the most important part of the  The current root and default directories to be used for new
process context; it is the information that the scheduler needs to file searches are stored here
suspend and restart the process  The signal-handler table defines the routine in the process’s
 The kernel maintains accounting information about the address space to be called when specific signals arrive
resources currently being consumed by each process, and the  The virtual-memory context of a process describes the full
total resources consumed by the process in its lifetime so far contents of the its private address space
 The file table is an array of pointers to kernel file structures
 When making file I/O system calls, processes refer to files by
their index into this table, the file descriptor (fd)

Operating System Concepts – 9th Edition 18.21 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.22 Silberschatz, Galvin and Gagne ©2013

Processes and Threads Scheduling

 Linux uses the same internal representation for processes and threads; a  The job of allocating CPU time to different tasks within an
thread is simply a new process that happens to share the same address operating system
space as its parent
 While scheduling is normally thought of as the running and
 Both are called tasks by Linux
interrupting of processes, in Linux, scheduling also includes the
 A distinction is only made when a new thread is created by the clone()
running of the various kernel tasks
system call
 fork() creates a new task with its own entirely new task context  Running kernel tasks encompasses both tasks that are
requested by a running process and tasks that execute internally
 clone() creates a new task with its own identity, but that is allowed
on behalf of a device driver
to share the data structures of its parent
 Using clone() gives an application fine-grained control over exactly what  As of 2.5, new scheduling algorithm – preemptive, priority-based,
is shared between two threads known as O(1)
 Real-time range
 nice value
 Had challenges with interactive performance
 2.6 introduced Completely Fair Scheduler (CFS)

Operating System Concepts – 9th Edition 18.23 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.24 Silberschatz, Galvin and Gagne ©2013
 CFS CFS (Cont.)

 Eliminates traditional, common idea of time slice  Then each task run with for time proportional to task’s weight
 Instead all tasks allocated portion of processor’s time divided by total weight of all runnable tasks

 CFS calculates how long a process should run as a function  Configurable variable target latency is desired interval during
of total number of tasks which each task should run at least once

 N runnable tasks means each gets 1/N of processor’s time  Consider simple case of 2 runnable tasks with equal weight
and target latency of 10ms – each then runs for 5ms
 Then weights each task with its nice value
 If 10 runnable tasks, each runs for 1ms
 Smaller nice value -> higher weight (higher priority)
 Minimum granularity ensures each run has reasonable
amount of time (which actually violates fairness idea)

Operating System Concepts – 9th Edition 18.25 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.26 Silberschatz, Galvin and Gagne ©2013

Kernel Synchronization Kernel Synchronization (Cont.)

 A request for kernel-mode execution can occur in two ways:  Linux uses two techniques to protect critical sections:
 A running program may request an operating system 1. Normal kernel code is nonpreemptible (until 2.6)
service, either explicitly via a system call, or implicitly, for – when a time interrupt is received while a process is
executing a kernel system service routine, the kernel’s
example, when a page fault occurs
need_resched flag is set so that the scheduler will run
 A device driver may deliver a hardware interrupt that once the system call has completed and control is
causes the CPU to start executing a kernel-defined about to be returned to user mode
handler for that interrupt 2. The second technique applies to critical sections that occur in an
interrupt service routines
 Kernel synchronization requires a framework that will allow
the kernel’s critical sections to run without interruption by – By using the processor’s interrupt control hardware to disable
another critical section interrupts during a critical section, the kernel guarantees that it can
proceed without the risk of concurrent access of shared data structures
 Provides spin locks, semaphores, and reader-writer versions of both
 Behavior modified if on single processor or multi:

Operating System Concepts – 9th Edition 18.27 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.28 Silberschatz, Galvin and Gagne ©2013
 Kernel Synchronization (Cont.) Interrupt Protection Levels
 To avoid performance penalties, Linux’s kernel uses a
synchronization architecture that allows long critical sections to
run without having interrupts disabled for the critical section’s
entire duration
 Interrupt service routines are separated into a top half and a
bottom half
 The top half is a normal interrupt service routine, and runs
with recursive interrupts disabled
 The bottom half is run, with all interrupts enabled, by a  Each level may be interrupted by code running at a higher
miniature scheduler that ensures that bottom halves never level, but will never be interrupted by code running at the
interrupt themselves same or a lower level
 This architecture is completed by a mechanism for disabling  User processes can always be preempted by another
selected bottom halves while executing normal, foreground process when a time-sharing scheduling interrupt occurs
kernel code

Operating System Concepts – 9th Edition 18.29 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.30 Silberschatz, Galvin and Gagne ©2013

Symmetric Multiprocessing Memory Management

 Linux 2.0 was the first Linux kernel to support SMP hardware;  Linux’s physical memory-management system deals with
separate processes or threads can execute in parallel on allocating and freeing pages, groups of pages, and small blocks
separate processors of memory
 Until version 2.2, to preserve the kernel’s nonpreemptible  It has additional mechanisms for handling virtual memory,
synchronization requirements, SMP imposes the restriction, via a memory mapped into the address space of running processes
single kernel spinlock, that only one processor at a time may  Splits memory into four different zones due to hardware
execute kernel-mode code characteristics
 Later releases implement more scalability by splitting single  Architecture specific, for example on x86:
spinlock into multiple locks, each protecting a small subset of
kernel data structures
 Version 3.0 adds even more fine-grained locking, processor
affinity, and load-balancing

Operating System Concepts – 9th Edition 18.31 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.32 Silberschatz, Galvin and Gagne ©2013
 Managing Physical Memory Managing Physical Memory (Cont.)

 The page allocator allocates and frees all physical pages; it  Memory allocations in the Linux kernel occur either statically
can allocate ranges of physically-contiguous pages on (drivers reserve a contiguous area of memory during system
request boot time) or dynamically (via the page allocator)
 The allocator uses a buddy-heap algorithm to keep track of  Also uses slab allocator for kernel memory
available physical pages  Page cache and virtual memory system also manage
 Each allocatable memory region is paired with an physical memory
adjacent partner  Page cache is kernel’s main cache for files and main
 Whenever two allocated partner regions are both freed mechanism for I/O to block devices
up they are combined to form a larger region  Page cache stores entire pages of file contents for local
 If a small memory request cannot be satisfied by and network file I/O
allocating an existing small free region, then a larger free
region will be subdivided into two partners to satisfy the
request

Operating System Concepts – 9th Edition 18.33 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.34 Silberschatz, Galvin and Gagne ©2013

Splitting of Memory in a Buddy Heap Slab Allocator in Linux

Operating System Concepts – 9th Edition 18.35 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.36 Silberschatz, Galvin and Gagne ©2013
 Virtual Memory Virtual Memory (Cont.)

 The VM system maintains the address space visible to each  Virtual memory regions are characterized by:
process: It creates pages of virtual memory on demand, and
 The backing store, which describes from where the pages for
manages the loading of those pages from disk or their swapping
a region come; regions are usually backed by a file or by
back out to disk as required.
nothing (demand-zero memory)
 The VM manager maintains two separate views of a process’s
address space:
 The region’s reaction to writes (page sharing or copy-on-
write
 A logical view describing instructions concerning the layout of
 The kernel creates a new virtual address space
the address space
1. When a process runs a new program with the exec()
 The address space consists of a set of non-overlapping
system call
regions, each representing a continuous, page-aligned
subset of the address space 2. Upon creation of a new process by the fork() system call
 A physical view of each address space which is stored in the
hardware page tables for the process

Operating System Concepts – 9th Edition 18.37 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.38 Silberschatz, Galvin and Gagne ©2013

Virtual Memory (Cont.) Swapping and Paging

 The VM paging system relocates pages of memory from

 On executing a new program, the process is given a new,
physical memory out to disk when the memory is needed for
completely empty virtual-address space; the program-loading
something else
routines populate the address space with virtual-memory regions
 The VM paging system can be divided into two sections:
 Creating a new process with fork() involves creating a
complete copy of the existing process’s virtual address space  The pageout-policy algorithm decides which pages to
write out to disk, and when
 The kernel copies the parent process’s VMA descriptors,
then creates a new set of page tables for the child  The paging mechanism actually carries out the transfer,
and pages data back into physical memory as needed
 The parent’s page tables are copied directly into the child’s,
with the reference count of each page covered being  Can page out to either swap device or normal files
incremented  Bitmap used to track used blocks in swap space kept in
 After the fork, the parent and child share the same physical physical memory
pages of memory in their address spaces  Allocator uses next-fit algorithm to try to write contiguous
runs

Operating System Concepts – 9th Edition 18.39 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.40 Silberschatz, Galvin and Gagne ©2013
 Kernel Virtual Memory Executing and Loading User Programs

 The Linux kernel reserves a constant, architecture-dependent  Linux maintains a table of functions for loading programs; it gives
region of the virtual address space of every process for its own each function the opportunity to try loading the given file when an
internal use exec system call is made
 This kernel virtual-memory area contains two regions:  The registration of multiple loader routines allows Linux to support
 A static area that contains page table references to every both the ELF and a.out binary formats
available physical page of memory in the system, so that  Initially, binary-file pages are mapped into virtual memory
there is a simple translation from physical to virtual  Only when a program tries to access a given page will a page
addresses when running kernel code fault result in that page being loaded into physical memory
 The reminder of the reserved section is not reserved for  An ELF-format binary file consists of a header followed by several
any specific purpose; its page-table entries can be modified page-aligned sections
to point to any other areas of memory
 The ELF loader works by reading the header and mapping the
sections of the file into separate regions of virtual memory

Operating System Concepts – 9th Edition 18.41 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.42 Silberschatz, Galvin and Gagne ©2013

Memory Layout for ELF Programs Static and Dynamic Linking

 A program whose necessary library functions are embedded
directly in the program’s executable binary file is statically
linked to its libraries
 The main disadvantage of static linkage is that every program
generated must contain copies of exactly the same common
system library functions
 Dynamic linking is more efficient in terms of both physical
memory and disk-space usage because it loads the system
libraries into memory only once

Operating System Concepts – 9th Edition 18.43 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.44 Silberschatz, Galvin and Gagne ©2013
 Static and Dynamic Linking (Cont.) File Systems
 To the user, Linux’s file system appears as a hierarchical directory tree
 Linux implements dynamic linking in user mode through special
obeying UNIX semantics
linker library
 Internally, the kernel hides implementation details and manages the
 Every dynamically linked program contains small statically multiple different file systems via an abstraction layer, that is, the virtual
linked function called when process starts file system (VFS)
 Maps the link library into memory  The Linux VFS is designed around object-oriented principles and is
composed of four components:
 Link library determines dynamic libraries required by process
and names of variables and functions needed  A set of definitions that define what a file object is allowed to look like
 The inode object structure represent an individual file
 Maps libraries into middle of virtual memory and resolves
references to symbols contained in the libraries  The file object represents an open file

 Shared libraries compiled to be position-independent code  The superblock object represents an entire file system
(PIC) so can be loaded anywhere  A dentry object represents an individual directory entry

Operating System Concepts – 9th Edition 18.45 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.46 Silberschatz, Galvin and Gagne ©2013

File Systems (Cont.) The Linux ext3 File System

 To the user, Linux’s file system appears as a hierarchical  ext3 is standard on disk file system for Linux
directory tree obeying UNIX semantics
 Uses a mechanism similar to that of BSD Fast File
 Internally, the kernel hides implementation details and manages System (FFS) for locating data blocks belonging to a
the multiple different file systems via an abstraction layer, that is, specific file
the virtual file system (VFS)
 Supersedes older extfs, ext2 file systems
 The Linux VFS is designed around object-oriented principles and
 Work underway on ext4 adding features like extents
layer of software to manipulate those objects with a set of
operations on the objects  Of course, many other file system choices with Linux
distros
 For example for the file object operations include (from struct
file_operations in /usr/include/linux/fs.h
int open(. . .) — Open a file
ssize t read(. . .) — Read from a file
ssize t write(. . .) — Write to a file
int mmap(. . .) — Memory-map a file

Operating System Concepts – 9th Edition 18.47 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.48 Silberschatz, Galvin and Gagne ©2013
 The Linux ext3 File System (Cont.) Ext2fs Block-Allocation Policies
 The main differences between ext2fs and FFS concern their disk
allocation policies
 In ffs, the disk is allocated to files in blocks of 8Kb, with blocks being
subdivided into fragments of 1Kb to store small files or partially filled
blocks at the end of a file
 ext3 does not use fragments; it performs its allocations in smaller
units
 The default block size on ext3 varies as a function of total size of
file system with support for 1, 2, 4 and 8 KB blocks
 ext3 uses cluster allocation policies designed to place logically
adjacent blocks of a file into physically adjacent blocks on disk, so
that it can submit an I/O request for several disk blocks as a single
operation on a block group
 Maintains bit map of free blocks in a block group, searches for free
byte to allocate at least 8 blocks at a time

Operating System Concepts – 9th Edition 18.49 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.50 Silberschatz, Galvin and Gagne ©2013

Journaling The Linux Proc File System

 ext3 implements journaling, with file system updates first  The proc file system does not store data, rather, its contents
written to a log file in the form of transactions are computed on demand according to user file I/O requests
 Once in log file, considered committed  proc must implement a directory structure, and the file contents
within; it must then define a unique and persistent inode
 Over time, log file transactions replayed over file system to
number for each directory and files it contains
put changes in place
 It uses this inode number to identify just what operation is
 On system crash, some transactions might be in journal but not
required when a user tries to read from a particular file
yet placed into file system
inode or perform a lookup in a particular directory inode
 Must be completed once system recovers
 When data is read from one of these files, proc collects the
 No other consistency checking is needed after a crash appropriate information, formats it into text form and places
(much faster than older methods) it into the requesting process’s read buffer
 Improves write performance on hard disks by turning random
I/O into sequential I/O

Operating System Concepts – 9th Edition 18.51 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.52 Silberschatz, Galvin and Gagne ©2013
 Input and Output Block Devices
 Provide the main interface to all disk devices in a system
 The Linux device-oriented file system accesses disk storage
through two caches:  The block buffer cache serves two main purposes:
 Data is cached in the page cache, which is unified with the  it acts as a pool of buffers for active I/O
virtual memory system  it serves as a cache for completed I/O
 Metadata is cached in the buffer cache, a separate cache  The request manager manages the reading and writing of buffer
indexed by the physical disk block contents to and from a block device driver
 Linux splits all devices into three classes:  Kernel 2.6 introduced Completely Fair Queueing (CFQ)
 block devices allow random access to completely  Now the default scheduler
independent, fixed size blocks of data
 Fundamentally different from elevator algorithms
 character devices include most other devices; they don’t
need to support the functionality of regular files  Maintains set of lists, one for each process by default

 network devices are interfaced via the kernel’s networking  Uses C-SCAN algorithm, with round robin between all
subsystem outstanding I/O from all processes
 Four blocks from each process put on at once

Operating System Concepts – 9th Edition 18.53 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.54 Silberschatz, Galvin and Gagne ©2013

Device-Driver Block Structure Character Devices

 A device driver which does not offer random access to fixed
blocks of data
 A character device driver must register a set of functions which
implement the driver’s various file I/O operations
 The kernel performs almost no preprocessing of a file read or
write request to a character device, but simply passes on the
request to the device
 The main exception to this rule is the special subset of character
device drivers which implement terminal devices, for which the
kernel maintains a standard interface

Operating System Concepts – 9th Edition 18.55 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.56 Silberschatz, Galvin and Gagne ©2013
 Character Devices (Cont.) Interprocess Communication

 Line discipline is an interpreter for the information from the  Like UNIX, Linux informs processes that an event has occurred
terminal device via signals
 The most common line discipline is tty discipline, which glues  There is a limited number of signals, and they cannot carry
the terminal’s data stream onto standard input and output information: Only the fact that a signal occurred is available to a
streams of user’s running processes, allowing processes to process
communicate directly with the user’s terminal
 The Linux kernel does not use signals to communicate with
 Several processes may be running simultaneously, tty line processes with are running in kernel mode, rather,
discipline responsible for attaching and detaching terminal’s communication within the kernel is accomplished via scheduling
input and output from various processes connected to it as states and wait_queue structures
processes are suspended or awakened by user
 Also implements System V Unix semaphores
 Other line disciplines also are implemented have nothing to
 Process can wait for a signal or a semaphore
do with I/O to user process – i.e. PPP and SLIP networking
protocols  Semaphores scale better
 Operations on multiple semaphores can be atomic

Operating System Concepts – 9th Edition 18.57 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.58 Silberschatz, Galvin and Gagne ©2013

Passing Data Between Processes Network Structure

 Networking is a key area of functionality for Linux
 The pipe mechanism allows a child process to inherit a  It supports the standard Internet protocols for UNIX to UNIX
communication channel to its parent, data written to one end communications
of the pipe can be read a the other
 It also implements protocols native to non-UNIX operating systems, in
 Shared memory offers an extremely fast way of particular, protocols used on PC networks, such as Appletalk and IPX
communicating; any data written by one process to a shared  Internally, networking in the Linux kernel is implemented by three
memory region can be read immediately by any other layers of software:
process that has mapped that region into its address space
 The socket interface
 To obtain synchronization, however, shared memory must
 Protocol drivers
be used in conjunction with another Interprocess-
communication mechanism  Network device drivers
 Most important set of protocols in the Linux networking system is the
internet protocol suite
 It implements routing between different hosts anywhere on the network
 On top of the routing protocol are built the UDP, TCP and ICMP protocols
 Packets also pass to firewall management for filtering based on
firewall chains of rules

Operating System Concepts – 9th Edition 18.59 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.60 Silberschatz, Galvin and Gagne ©2013
 Security Security (Cont.)

 The pluggable authentication modules (PAM) system is  Linux augments the standard UNIX setuid mechanism in two
available under Linux ways:
 PAM is based on a shared library that can be used by any  It implements the POSIX specification’s saved user-id
system component that needs to authenticate users mechanism, which allows a process to repeatedly drop and
 Access control under UNIX systems, including Linux, is reacquire its effective uid
performed through the use of unique numeric identifiers (uid  It has added a process characteristic that grants just a
and gid) subset of the rights of the effective uid
 Access control is performed by assigning objects a protections  Linux provides another mechanism that allows a client to
mask, which specifies which access modes—read, write, or selectively pass access to a single file to some server process
execute—are to be granted to processes with owner, group, or without granting it any other privileges
world access

Operating System Concepts – 9th Edition 18.61 Silberschatz, Galvin and Gagne ©2013 Operating System Concepts – 9th Edition 18.62 Silberschatz, Galvin and Gagne ©2013

End of Chapter 18

Operating System Concepts – 9th Edition Silberschatz, Galvin and Gagne ©2013