blob: ac4a7a737e430207a22be59b2d817b309eefa828 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001The Linux Kernel Device Model
2
Linas Vepstasab11f892006-02-03 03:03:38 -08003Patrick Mochel <mochel@digitalimplant.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -07004
Linas Vepstasab11f892006-02-03 03:03:38 -08005Drafted 26 August 2002
6Updated 31 January 2006
Linus Torvalds1da177e2005-04-16 15:20:36 -07007
8
9Overview
10~~~~~~~~
11
Linas Vepstasab11f892006-02-03 03:03:38 -080012The Linux Kernel Driver Model is a unification of all the disparate driver
13models that were previously used in the kernel. It is intended to augment the
Linus Torvalds1da177e2005-04-16 15:20:36 -070014bus-specific drivers for bridges and devices by consolidating a set of data
15and operations into globally accessible data structures.
16
Linas Vepstasab11f892006-02-03 03:03:38 -080017Traditional driver models implemented some sort of tree-like structure
18(sometimes just a list) for the devices they control. There wasn't any
19uniformity across the different bus types.
Linus Torvalds1da177e2005-04-16 15:20:36 -070020
Linas Vepstasab11f892006-02-03 03:03:38 -080021The current driver model provides a comon, uniform data model for describing
22a bus and the devices that can appear under the bus. The unified bus
23model includes a set of common attributes which all busses carry, and a set
24of common callbacks, such as device discovery during bus probing, bus
25shutdown, bus power management, etc.
Linus Torvalds1da177e2005-04-16 15:20:36 -070026
Linas Vepstasab11f892006-02-03 03:03:38 -080027The common device and bridge interface reflects the goals of the modern
28computer: namely the ability to do seamless device "plug and play", power
29management, and hot plug. In particular, the model dictated by Intel and
30Microsoft (namely ACPI) ensures that almost every device on almost any bus
31on an x86-compatible system can work within this paradigm. Of course,
32not every bus is able to support all such operations, although most
33buses support a most of those operations.
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
35
36Downstream Access
37~~~~~~~~~~~~~~~~~
38
39Common data fields have been moved out of individual bus layers into a common
Linas Vepstasab11f892006-02-03 03:03:38 -080040data structure. These fields must still be accessed by the bus layers,
Linus Torvalds1da177e2005-04-16 15:20:36 -070041and sometimes by the device-specific drivers.
42
43Other bus layers are encouraged to do what has been done for the PCI layer.
44struct pci_dev now looks like this:
45
46struct pci_dev {
47 ...
48
Linas Vepstasab11f892006-02-03 03:03:38 -080049 struct device dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070050};
51
52Note first that it is statically allocated. This means only one allocation on
53device discovery. Note also that it is at the _end_ of struct pci_dev. This is
54to make people think about what they're doing when switching between the bus
55driver and the global driver; and to prevent against mindless casts between
56the two.
57
58The PCI bus layer freely accesses the fields of struct device. It knows about
59the structure of struct pci_dev, and it should know the structure of struct
Linas Vepstasab11f892006-02-03 03:03:38 -080060device. Individual PCI device drivers that have been converted the the current
61driver model generally do not and should not touch the fields of struct device,
62unless there is a strong compelling reason to do so.
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64This abstraction is prevention of unnecessary pain during transitional phases.
65If the name of the field changes or is removed, then every downstream driver
66will break. On the other hand, if only the bus layer (and not the device
67layer) accesses struct device, it is only that layer that needs to change.
68
69
70User Interface
71~~~~~~~~~~~~~~
72
73By virtue of having a complete hierarchical view of all the devices in the
74system, exporting a complete hierarchical view to userspace becomes relatively
75easy. This has been accomplished by implementing a special purpose virtual
76file system named sysfs. It is hence possible for the user to mount the
77whole sysfs filesystem anywhere in userspace.
78
79This can be done permanently by providing the following entry into the
80/etc/fstab (under the provision that the mount point does exist, of course):
81
82none /sys sysfs defaults 0 0
83
84Or by hand on the command line:
85
86# mount -t sysfs sysfs /sys
87
88Whenever a device is inserted into the tree, a directory is created for it.
89This directory may be populated at each layer of discovery - the global layer,
90the bus layer, or the device layer.
91
92The global layer currently creates two files - 'name' and 'power'. The
93former only reports the name of the device. The latter reports the
94current power state of the device. It will also be used to set the current
95power state.
96
97The bus layer may also create files for the devices it finds while probing the
98bus. For example, the PCI layer currently creates 'irq' and 'resource' files
99for each PCI device.
100
101A device-specific driver may also export files in its directory to expose
102device-specific data or tunable interfaces.
103
104More information about the sysfs directory layout can be found in
105the other documents in this directory and in the file
106Documentation/filesystems/sysfs.txt.
107