5. Input/output and security

Two last functions complete the picture. The first is input/output management: how a program can use a printer, a disk or a keyboard without knowing how each one works. The second is security: who can open, change or run each file and each shared resource. Both are everywhere in your daily work with a computer.

How to use this page

1. Read sections 1 to 3 (about 35 minutes) and try the tab I/O manager of the Process & CPU Scheduler Lab.
2. Read sections 4 to 7 (about 50 minutes) and have a Linux terminal and a Windows machine (or virtual machines) ready.
3. Do Activity 7.

1. How an application uses a device

Devices are very different: a keyboard sends a few bytes per second, an SSD moves gigabytes per second, a printer can run out of paper. The OS hides these differences with layers:

1. The application asks the OS to do an operation (“print this”, “read this file”).
2. The I/O manager of the kernel checks the request, finds the right device and queues the request if the device is busy.
3. The driver is a program that knows one family of devices. It translates the general request into the commands of that device. Each device needs its own driver, which is why a new device sometimes needs a driver installation.
4. The controller is the electronics of the device. It receives the data and the commands in its registers and runs the mechanics (move the head, heat the paper…).
5. The device does the work, and the answer goes back along the same path. The controller writes the result in a status register that the driver reads.

Linux shows many devices as files in /dev. Disks are block devices (they work with blocks of data, /dev/sda). Keyboards and serial ports are character devices (they work with a stream of bytes). Windows names its devices in the Device Manager (devmgmt.msc).

2. How the CPU knows that a device has finished

The CPU is much faster than any device. There are three ways to deal with the difference.

Polling (programmed I/O). The CPU asks the device again and again “are you ready?”. It is simple, but the CPU is busy waiting: it is working only to wait.

Interrupts. The CPU starts the operation and goes on with other work. When the device is ready, it sends an interrupt request (IRQ), an electrical signal. The CPU stops what it is doing, saves its state, runs a small interrupt handler (also called interrupt service routine) from the driver and then returns to the previous work. Keyboard and mouse use interrupts for every key and movement. A process that asked for the operation waits in the Waiting state, and the interrupt is what moves it back to Ready (see the state diagram in page 1).

DMA (Direct Memory Access). With interrupts the CPU still copies each piece of data. For large transfers this is too much. With DMA, a special controller moves the whole block between the device and RAM by itself. The CPU only prepares the transfer (it gives the controller the address and the size) and receives one single interrupt at the end. Disks, network cards and graphic cards use DMA.

	Polling	Interrupts	DMA
Who copies the data	CPU	CPU	DMA controller
CPU while the device works	Waits (busy)	Free for other processes	Free for other processes
Interruptions	None	One per piece of data	One per transfer
Good for	Very fast devices, simple systems	Keyboards, mice, small transfers	Disks, network, large transfers

Example (simplified). We transfer 4 words from a device that needs 5 time units to prepare each word. Copying a word or running a handler costs 1 unit.

• Polling: 4 × (5 + 1) = 24 units, and the CPU is busy for all of them. 0 units free.
• Interrupts: 24 units in total, but the CPU only works 4 units (4 handlers). 20 units free (83 %).
• DMA: 1 unit to start, 20 units of transfer, 1 unit to finish: 22 units. The CPU works 2 units. 20 units free (91 %).

The tab I/O manager of the lab lets you change the numbers.

3. Cache, buffer and spooling

The OS also uses memory to make I/O faster. The three techniques are different:

Technique	What it is	Example
Cache	A fast copy of data that you use often	The file you read twice comes the second time from RAM, not from the disk
Buffer	A temporary area that connects two devices or processes of different speed	A video player loads a few seconds in advance so that it does not stop
Spooling	A queue of complete jobs for a device that cannot be shared	The print queue

Cache. If a read finds the data in the cache, it is a hit; if not, it is a miss and the OS reads the slow device and copies the data to the cache. Imagine that a hit costs 1 unit and a miss 10 units. With 80 % hits, the average cost is 0.8 × 1 + 0.2 × 10 = 2.8 units, nearly four times faster than always reading the device.

Buffer. While the slow device fills or empties the buffer, the CPU can continue. When you copy a big file to a USB drive, the data first goes to a buffer in RAM, and this is the reason why you must safely remove the drive: the buffer may not be empty yet.

Spooling (Simultaneous Peripheral Operations On-Line). A printer can print only one document at a time, but several programs and users want to print. The jobs are saved in a spool (a folder on the disk) and a service sends them to the printer one by one. The programs are free immediately, and you can see, pause or cancel jobs. In Linux, lpstat -o lists the jobs; in Windows you open the printer queue.

4. Security

The OS must protect the system against errors (a program that writes in memory that is not its own), misuse by a user, unauthorised access (physical, through the network or by malware such as viruses, worms or trojans) and faults. No system is completely secure, but good habits reduce the risk. The aims of security are:

• Confidentiality: only authorised people can read the data.
• Integrity: only authorised people can change it, and changes are not accidental.
• Availability: the data and services work when you need them.

Logical (software) measures	Physical measures
Accounts with strong passwords	Locked room or locked case
Permissions on files and folders	Access control with cards or fingerprints
Updates and antivirus	UPS (uninterruptible power supply)
Firewall	Cable locks, security cameras
Encryption of disks and files	Backups kept in another place

Three ideas are behind every access decision:

1. Authentication: proving who you are (password, fingerprint…).
2. Authorisation: deciding what you may do (permissions).
3. Accounting: keeping a record of what happened (logs).

Least privilege: give each user and each program only the rights they need. For this reason you work with a standard account and use administrator rights only when needed (sudo in Linux, the shield icon and the UAC window in Windows). A process that runs with your account can do only what you can do: a program that you open by mistake cannot change system files.

5. Attributes and permissions

Every file and directory has metadata. Besides name, size and dates, it has attributes and permissions.

Attributes

System	Attributes	How to see or change them
Windows	Read-only, Hidden, System, Archive	Properties → General, or attrib +h file.txt
Linux	A name that begins with a dot is hidden. In ext4, extra flags such as immutable (nobody can modify or delete)	ls -a, lsattr, sudo chattr +i file

Permissions in Linux

Linux gives each file an owner, a group and three sets of permissions, for the owner (u), the group (g) and others (o). Each set has three permissions: r (read), w (write) and x (execute). Read the output of ls -l:

-rwxr-x---  1 student teachers  4096 Oct  5 10:00 backup.sh
│└┬┘└┬┘└┬┘    │        │
│ │  │  │     │        └── group
│ │  │  │     └── owner
│ │  │  └── others: --- (nothing)
│ │  └── group: r-x (read and execute)
│ └── owner: rwx (read, write, execute)
└── type: - file, d directory, l link

Each permission has a number: r = 4, w = 2, x = 1. Add them for each set and you get the octal notation: rwx = 7, r-x = 5, --- = 0, so this file is 750.

Octal	Permissions	Typical use
644	rw-r--r--	A normal file: you edit it, all can read it
600	rw-------	A private file (for example an SSH key)
755	rwxr-xr-x	A program or a public folder
750	rwxr-x---	A script for a group only
700	rwx------	A private folder

chmod 640 notes.txt        # rw-r----- (octal)
chmod u+x backup.sh        # add execute for the owner
chmod g-w,o-rwx report.txt # remove write from the group, all from others
sudo chown student:teachers report.txt   # change the owner and the group

The permissions mean something different for a directory:

	On a file	On a directory
r	Read the content	List the names inside
w	Change the content	Create, delete and rename files inside
x	Run it as a program	Enter it (cd) and reach what is inside

New files are created with permissions that depend on the umask (a typical value, 022, gives 644 to files and 755 to directories).

Permissions in Windows (NTFS)

On an NTFS disk, right click → Properties → Security shows a list of users and groups with their permissions. The basic ones are:

Permission	What it allows
Read	See the content and the attributes
Write	Create files and change the content
Read & execute	Read and run programs (on folders: also List folder contents)
Modify	Read, write, execute and delete
Full control	Everything, including changing permissions and taking ownership

Rules to remember:

• Permissions are inherited from the parent folder, unless you disable inheritance.
• A user receives the sum of the permissions of their own entry and of all their groups.
• A Deny entry is stronger than an Allow one.
• icacls file shows and changes them from the command line.

Linux	NTFS (approximately)
r--	Read
r-x	Read & execute
rw-	Modify (without execute)
rwx	Full control (roughly)
Owner / group / others	Any number of users and groups

NTFS is more flexible because it can give different permissions to many users in the same file. In Linux, ACLs (setfacl) give the same flexibility when you need it.

6. Shared resources

A shared resource is a folder, a printer or a disk that other computers on the network can use. Sharing is useful, but it opens a door, so you must configure it carefully.

Windows

1. Right click the folder → Properties → Sharing → Advanced Sharing.
2. Tick Share this folder, give it a name and press Permissions to choose who can use it and with which level (Read, Change or Full control).
3. In the Security tab, check the NTFS permissions.

When a folder is accessed through the network, Windows applies both the share permissions and the NTFS permissions, and the most restrictive one wins. Other computers open it with \\COMPUTERNAME\sharename. Useful commands: net share (list what you share) and net use Z: \\PC\share (connect a network drive).

Linux

The usual tool for sharing with Windows computers is Samba, which implements the SMB protocol. In its configuration file you define the share:

[documents]
   path = /srv/documents
   read only = no
   valid users = @teachers

After a change the service is restarted. Clients connect with smb://server/documents from a file manager. You will configure network services in other modules.

Good practices

• Share only what is needed, and not the entire disk.
• Give permissions to specific users or groups, not to Everyone with Full control.
• Prefer read-only access when users do not need to write.
• Do not allow anonymous or guest access unless it is needed, and keep the firewall active.
• Check from time to time what is shared (net share in Windows).

7. Putting the functions together

When you print a PDF from a USB drive, at least five functions work: the file system finds the file and the permissions allow you to open it, the memory manager loads the viewer, the scheduler shares the CPU between the viewer and your music, the I/O manager and the driver talk to the printer, and the spooler queues the job. Understanding which function fails is the first step to solving a problem.

Comprueba lo aprendido

1. What is the main advantage of interrupts over polling, and the main advantage of DMA over interrupts?

   Answer

   With interrupts the CPU does not waste time waiting for the device; it works on other things. With DMA the CPU does not even copy the data: it only starts the transfer and receives one interrupt at the end, which is better for large transfers.

2. Explain the difference between a buffer and a spool with an example of each.

   Answer

   A buffer is a temporary memory area that smooths the speed difference between two parts (for example the data of a video before it is played). A spool is a queue of complete jobs for a device that cannot be shared (for example the documents waiting for the printer).

3. A file has the permissions -rw-r----- and belongs to the user ana and the group staff. Write the octal number and say who can write.

   Answer

   rw- = 6, r-- = 4, --- = 0: the octal number is 640. Only ana (the owner) can write. The members of staff can read, and the others cannot do anything.

4. A user has the r-- permission on a directory (no x). Can they see the names of the files? Can they open one of them?

   Answer

   They can list the names (r), but they cannot enter the directory or open the files (x is missing), so they will see errors when they try to read them.

5. A Windows folder is shared with Full control for everybody, but the NTFS permissions allow only Read for the group Students. What can a student do through the network?

   Answer

   Only read. When the share permissions and the NTFS permissions are different, the most restrictive result applies.

Glossary

English	Castellano
input/output (I/O)	entrada/salida (E/S)
device driver	controlador de dispositivo
device controller	controladora del dispositivo
polling	sondeo
interrupt / interrupt request (IRQ)	interrupción / petición de interrupción
interrupt handler	rutina de tratamiento de la interrupción
DMA (direct memory access)	DMA (acceso directo a memoria)
cache / hit / miss	caché / acierto / fallo
buffer	búfer (memoria intermedia)
spooling / spool	spooling / cola de impresión
authentication / authorisation	autenticación / autorización
least privilege	mínimo privilegio
permission	permiso
owner / group / others	propietario / grupo / otros
read / write / execute	lectura / escritura / ejecución
shared resource	recurso compartido
inheritance	herencia

Credits

This page adapts the SOM course notes by Moisés Mas Rocamora, Daniel Álvarez Santamaría and Jana Taboada León, licensed under CC BY-NC-SA 4.0. This page is shared under the same licence.