Ytdyklly

I would like to know how Linux, such us Ubuntu, "know" how and what drivers to install when installing it from scratch.
For example, I buy a new computer without any system and I install Ubuntu. Inside my PC I have a GPU, HDDs, etc., also some peripherals, like mouse, keyboard, etc. Obviously, a fresh install does not have drivers needed for the system to control and communicate with the hardware so how does Ubuntu "know" what drivers to install/download and how does it do that?

All peripherals identify themselves with mostly unique IDs. Some IDs are for generic interfaces (HDD/mouse,etc). Linux has most drivers built in, and the generic drivers have compatibility lists for IDs that support a limited feature set.It gets more complicated but dmesg, lscpu, hwinfo, lshw, dmidecode, lspci, etc will list the IDs if you want to look

(Based on Google-cached copy of http://people.skolelinux.org/pere/blog/Modalias_strings___a_practical_way_to_map__stuff__to_hardware.html by Petter Reinholdtsen.)

In the hardware, there are certain standard device identifiers that can be accessed as long as you know the standard access method for that particular I/O bus or subsystem, without having any further knowledge about the actual device. In Linux, these identifiers are used to build up modalias strings, which are then used to find the correct driver for each device.

The source code of each driver module can include MODULE_DEVICE_TABLE structures, which are used by the depmod command to create module alias wildcard entries that will match the modalias strings of the hardware supported by that particular module.

When the kernel detects a piece of hardware with no matching driver loaded yet, it will create a modalias string from the identifiers of the hardware, and use it to request a module to be autoloaded. The modprobe command will then use the /lib/modules/$(uname -r)/modules.alias[.bin] file created by depmod to see if a matching module exists. If it does, that module is loaded and gets to probe the hardware for further details if necessary.

For example, I have a DVB TV card:

$ lspci -v -nn -s 07:00.0

07:00.0 Multimedia video controller [0400]: Conexant Systems, Inc. CX23885 PCI Video and Audio Decoder [14f1:8852] (rev 04)

    Subsystem: Hauppauge computer works Inc. CX23885 PCI Video and Audio Decoder [0070:6a28]

This results in a modalias string like this:

pci:v000014F1d00008852sv00000070sd00006A28bc04sc00i00

The cx23885 module has these aliases based on MODULE_DEVICE_TABLE in its source code:

# modinfo cx23885

...

alias:          pci:v000014F1d00008880sv*sd*bc*sc*i*

alias:          pci:v000014F1d00008852sv*sd*bc*sc*i*

...

When the kernel detects the card, it effectively runs the modprobe pci:v000014F1d00008852sv00000070sd00006A28bc04sc00i00 command. The second alias of the cx23885 module matches, and so that module gets loaded.

PCI/PCI-X/PCIe bus devices

This is the "PCI subtype". It uses modalias strings like this:

pci:v00008086d00002770sv00001028sd000001ADbc06sc00i00

This can be decoded as follows:

v   00008086  (vendor)

d   00002770  (device)

sv  00001028  (subvendor)

sd  000001AD  (subdevice)

bc  06        (bus class)

sc  00        (bus subclass)

i   00        (interface)

With lspci -nn, you can see the class, subclass, vendor and device IDs. If you add the -v option, you can also see the subvendor:subdevice IDs.

USB devices

With USB devices, the modalias strings look like this:

usb:v1D6Bp0001d0206dc09dsc00dp00ic09isc00ip00

This unpacks to:

v    1D6B  (device vendor)

p    0001  (device product)

d    0206  (bcddevice)

dc     09  (device class)

dsc    00  (device subclass)

dp     00  (device protocol)

ic     09  (interface class)

isc    00  (interface subclass)

ip     00  (interface protocol)

With the lsusb command, you can see the vendor and product IDs. If you use the -v option, you can see the other IDs too.

ACPI devices

These use the ACPI PNP identifiers, prefixed with acpi: and separated with colons:

acpi:IBM0071:PNP0511:

DMI devices

This can be a very long modalias string:

dmi:bvnIBM:bvr1UETB6WW(1.66):bd06/15/2005:svnIBM:pn2371H4G:pvrThinkPadX40:rvnIBM:rn2371H4G:rvrNotAvailable:cvnIBM:ct10:cvrNotAvailable:

This unpacks to:

bvn  IBM            (BIOS vendor)

bvr  1UETB6WW(1.66) (BIOS version)

bd   06/15/2005     (BIOS date)

svn  IBM            (system vendor)

pn   2371H4G        (product name)

pvr  ThinkPadX40    (product version)

rvn  IBM            (board vendor)

rn   2371H4G        (board name)

rvr  NotAvailable   (board version)

cvn  IBM            (chassis vendor)

ct   10             (chassis type)

cvr  NotAvailable   (chassis version)

SerIO devices, i.e. mostly PS/2 mice

The modalias string will look like this:

serio:ty01pr00id00ex00

The values here are:

ty  01  (type)

pr  00  (prototype)

id  00  (id)

ex  00  (extra)

Other bus/device types

There are many other bus types recognized by the Linux kernel. Studying the contents of the kernel source file file2alias.c might be helpful in deciphering the meaning of the components of each type of modalias string.