Transcript A6_presentation3_RTOS

Real-Time Operating
Systems - QNX
Brett O’Neill
CSE 8343 – Group A6
1
Overview

Real-Time Operating Systems



What is a real-time operating system?
Who needs real-time systems?
QNX


What is QNX?
Microkernel



Process Manager





IPC
Process scheduling
Overview
Process Life Cycle
I/O Namespace
File Manager
Device Manager
2
Real-Time Operating Systems
What is a real-time operating system?
From comp.realtime newsgroup faq:
“A realtime system is one in which the correctness of the
computations not only depends upon the logical correctness of the
computation but also upon the time at which the result is
produced. If the timing constraints are not met, system failure is
said to have occurred.”
3
Real-Time Operating Systems (cont.)

The value of the computation depends on the
timeliness of the answer is provided.
Computations finished late have diminished value
 Computations finished early have no extra value


Problems arise when resources are shared
among several computations – real-time systems
use schedules of activities so all activities will be
completed in time
4
Real-Time Operating Systems (cont.)


Hard Real-Time – A system constraint in which late
computations have NO value, and the effects of late
computations can be catastrophic. In hard real-time
systems, activities MUST be completed on time.
Soft Real-Time – Late computations have some value,
albeit diminished. Soft real-time systems can tolerate
some late computations, as long as the value has not
diminished to zero.

Meta requirements such as a stochastic model of the
acceptable frequency of late computations are often used
5
Real-Time Operating Systems (cont.)

Real-Time Systems have different types of
activities:
Those that can be scheduled
 Those that cannot be scheduled
 Non real-time activities

6
Real-Time Operating Systems (cont.)
Who needs real-time systems?
 Mission-critical environments requiring hard real-time,
where timely performance failures can result in harm to
people or property
 Quality/Timeliness of service guarantee environments,
particularly when failure to meet guarantees can result
in financial penalties
 Consumer devices that demand reliability –ie. live video
stream players in which failure to deliver content results
in dropped frame rates
7
QNX
What is QNX?
 A commercial real-time operating system
developed by QNX Software Systems Ltd. of
Canada (QSSL)
 Runs on X86 machines and clones: AMD, Nat
Semiconductor, Cyrix, SGS Semiconductor
 Provides multitasking, priority-driven
preemptive scheduling and fast context
switching
8
QNX - Microkernel

“Bare bones” QNX consists only of a small
kernel in charge of managing cooperating
processes
9
QNX - Microkernel

The microkernel has 4 main functions:
Routing messages between processes
 Low-level network communication
 Process scheduling
 First-level interrupt handling

10
QNX - Microkernel
Message-based Inter-process Communication



A message is a packet of bytes synchronously transferred from one
process to another
QNX does not attach meaning to the context of the message – the data
in the message only has meaning to the sender and the receiver
C language functions Send(), Receive(), and Reply() are used
11
QNX - Microkernel


Message passing synchronizes the execution of several
cooperating processes. If Process A issues a Send()
request, it cannot resume execution until Process B
replies. And once Process B has issued its Receive()
request, it cannot continue execution until it receives
another message.
Processes that are not allowed to continue execution are
blocked.
12
QNX - Microkernel
Proxy-based Inter-process Communication


A proxy is a non-blocking message used for event notification in which
the sending process does not need to interact with the recipient
Proxies are used when:




A process wants to notify another process of an event, but does not want to
risk sending a blocking message
A process wants to send data to another process but does not need
acknowledgment of delivery
An interrupt handler wants to tell a process that some data is available
Proxies can queue up to 65,535 messages for delivery. They can be
triggered more than once, sending a message for each trigger
13
QNX - Microkernel
Signal-based Inter-process Communication


Signals are a method of asynchronous communication
QNX supports POSIX-compliant signals, UNIX signals, and QNXspecific signals
Network Inter-process Communication


Applications can communicate over a network transparently. QNX treats
all processes the same, whether they are local or remote
Virtual circuits are used – these are paths provided by the Network
Manager to transmit messages, proxies and signals
14
QNX - Microkernel

Virtual Circuits:





The sending process is responsible for setting up a VC
between itself and the receiving process.
Upon creation, the VC is given the ability to handle messages
up to a specified size limit. If a message larger than the size
limit is sent, the VC adjusts on the fly.
Two processes can communicate via multiple VC’s, combined
into one logical VC.
Upon process termination, VC’s are automatically released.
Virtual proxies are also possible
15
QNX - Microkernel
Process Scheduling

The microkernel’s scheduler makes scheduling decisions at
three points:




When a process becomes unblocked
When the time quantum for a running process expires
When a running process is preempted
Every QNX process is assigned a priority. The scheduler
selects the process with the highest priority from the ready
pool to run next.

Priorities range from 0 (lowest) to 31 (highest). Processes receive
initial priorities from their parent processes. The default value is 10.
16
QNX - Microkernel

QNX has 3 scheduling
algorithms. These algorithms
are of course only used when
2 or more processes that
share the same priority are in
ready status. If a higherpriority process becomes
available, it immediately
preempts all lower-priority
processes:

FIFO Scheduling – A process
continues to execute until it a)
voluntarily relinquishes control
or b) is preempted by a higherpriority process
17
QNX - Microkernel

Round-Robin Scheduling – A process continues to
execute until it a) voluntarily relinquishes control, b)
is preempted by a higher-priority process, or c)
consumes its time slice. A QNX time slice is 50
milliseconds
18
QNX - Microkernel

Adaptive Scheduling – If a process consumes its
time slice entirely, its priority is reduced by 1. This is
called priority decay. Processes only decay once; if a
process consumes more than one time slice, total
reduction is only 1. If a process blocks, it
immediately regains its original priority.
19
QNX – Process Manager
Overview


The process manager works with the microkernel to provide
operating system services. It shares the same address space as the
process manager, but runs as a unique process, scheduled by the
microkernel. It uses the same types of IPC’s as all other QNX
processes.
The process manager is responsible for creating new processes
using C functions:



Fork()
Exec()
Spawn()
20
QNX – Process Manager
Process Life Cycle

Each process goes through 4 stages in its lifetime:




Creation – A unique process ID is created, and basic
information is defined for the process’ environment.
Loading – Loader code in the process manager creates a
loader thread that runs under the process ID of the new
process.
Execution – The process is in direct competition with other
processes for execution time.
Termination – A signal causes process termination, or the
process invokes an Exit() function.
21
QNX – Process Manager
Process Life Cycle

Processes are always in one of the following states:


READY – capable of being executed
BLOCKED – the process is in one of these states:








SEND
RECEIVE
REPLY
SIGNAL
SEMAPHORE
HELD – the process has received a hold signal
WAIT – a child process has requested that the parent process wait
DEAD – the process has terminated but its parent is still running
22
QNX – I/O Namespace


I/O resources are not built into the microkernel as in
most operating systems. They are started dynamically
while the system is running. Pathname space is not built
into the file system.
Pathname space is divided into regions of authority.
Any processes that required I/O must register a prefix
with the process manager defining the portion of
namespace it wants to administer. These prefixes are
part of a prefix tree maintained on the computer.
23
QNX – I/O Namespace



When a process opens a file, the file’s pathname is
compared to the prefix tree to direct the Open()
function to the correct I/O resource manager.
When an I/O resource is opened, the Open() function
returns an integer called the file descriptor. All further
I/O requests are directed by the file descriptor to the
correct file manager.
The file descriptor namespace is completely local to
each process.
24
QNX – File Manager


The File Manager provides a standardized means of storing and
accessing data on disks.
In QNX, a file is an object that can be read from, written to, or
both. There are 6 types of files:





Regular – randomly accessible sequences of bytes with no predefined
structure
Directories – contain the information needed to locate regular files
Symbolic links – contain a pathname to a file or directory that is to be
accessed in place of the symbolic link file
Pipes and FIFO’s – I/O channels between cooperating processes
Block special – refer to devices; accessed in a manner that hides the
hardware characteristics of the device from applications
25
QNX – File Manager


The file manager maintains four times for each file: a)
date of last access, b) date of last write, c) date of last
modification, and d) date of creation
Access to files is controlled by bits called inodes. Inodes
permit read, write and execute permissions by user, by
group, and by other specifications.
26
QNX – File Manager

The file manager has several means of maintaining highperformance disk accesses:



Elevator seeking – outstanding I/O requests are ordered so they can all
be performed with one sweep of the disk head assembly, from lowest to
highest disk address.
Buffer cache – a buffer between the file manager and the disk driver, the
buffer cache tries to store all file system blocks to minimize the number
of times the file manager needs to access the disk.
Multi-threading – the file manager is multi-threaded, therefore it can
manage several I/O requests simultaneously. Several devices can be
accessed in parallel, and I/O requests can be processed from the buffer
cache while other I/O requests are accessing physical disks.
27
QNX – File Manager
Client-driven priority – when the file manager
receives a message, its priority is set to that of the
process that sent the message.
 Temporary files – data blocks are kept in cache to
avoid writing blocks to a physical disk unless
absolutely necessary.
 Ramdisks – up to 8M of memory can be used as a
simulated disk. Data moves directly from the
ramdisk into application buffers.

28
QNX – Device Manager



QNX’s device manager is an interface between
processes and terminal devices.
Programs access terminal devices using the C
functions Read(), Write(), Open() and Close().
The device manager regulates the flow of data
between applications and devices.
29
Questions?
30