1 ============================= 1 =============================
2 The Linux Kernel Device Model 2 The Linux Kernel Device Model
3 ============================= 3 =============================
4 4
5 Patrick Mochel <mochel@digitalimplant.org> 5 Patrick Mochel <mochel@digitalimplant.org>
6 6
7 Drafted 26 August 2002 7 Drafted 26 August 2002
8 Updated 31 January 2006 8 Updated 31 January 2006
9 9
10 10
11 Overview 11 Overview
12 ~~~~~~~~ 12 ~~~~~~~~
13 13
14 The Linux Kernel Driver Model is a unification 14 The Linux Kernel Driver Model is a unification of all the disparate driver
15 models that were previously used in the kernel 15 models that were previously used in the kernel. It is intended to augment the
16 bus-specific drivers for bridges and devices b 16 bus-specific drivers for bridges and devices by consolidating a set of data
17 and operations into globally accessible data s 17 and operations into globally accessible data structures.
18 18
19 Traditional driver models implemented some sor 19 Traditional driver models implemented some sort of tree-like structure
20 (sometimes just a list) for the devices they c 20 (sometimes just a list) for the devices they control. There wasn't any
21 uniformity across the different bus types. 21 uniformity across the different bus types.
22 22
23 The current driver model provides a common, un 23 The current driver model provides a common, uniform data model for describing
24 a bus and the devices that can appear under th 24 a bus and the devices that can appear under the bus. The unified bus
25 model includes a set of common attributes whic 25 model includes a set of common attributes which all busses carry, and a set
26 of common callbacks, such as device discovery 26 of common callbacks, such as device discovery during bus probing, bus
27 shutdown, bus power management, etc. 27 shutdown, bus power management, etc.
28 28
29 The common device and bridge interface reflect 29 The common device and bridge interface reflects the goals of the modern
30 computer: namely the ability to do seamless de 30 computer: namely the ability to do seamless device "plug and play", power
31 management, and hot plug. In particular, the m 31 management, and hot plug. In particular, the model dictated by Intel and
32 Microsoft (namely ACPI) ensures that almost ev 32 Microsoft (namely ACPI) ensures that almost every device on almost any bus
33 on an x86-compatible system can work within th 33 on an x86-compatible system can work within this paradigm. Of course,
34 not every bus is able to support all such oper 34 not every bus is able to support all such operations, although most
35 buses support most of those operations. 35 buses support most of those operations.
36 36
37 37
38 Downstream Access 38 Downstream Access
39 ~~~~~~~~~~~~~~~~~ 39 ~~~~~~~~~~~~~~~~~
40 40
41 Common data fields have been moved out of indi 41 Common data fields have been moved out of individual bus layers into a common
42 data structure. These fields must still be acc 42 data structure. These fields must still be accessed by the bus layers,
43 and sometimes by the device-specific drivers. 43 and sometimes by the device-specific drivers.
44 44
45 Other bus layers are encouraged to do what has 45 Other bus layers are encouraged to do what has been done for the PCI layer.
46 struct pci_dev now looks like this:: 46 struct pci_dev now looks like this::
47 47
48 struct pci_dev { 48 struct pci_dev {
49 ... 49 ...
50 50
51 struct device dev; /* Generic devi 51 struct device dev; /* Generic device interface */
52 ... 52 ...
53 }; 53 };
54 54
55 Note first that the struct device dev within t 55 Note first that the struct device dev within the struct pci_dev is
56 statically allocated. This means only one allo 56 statically allocated. This means only one allocation on device discovery.
57 57
58 Note also that that struct device dev is not n 58 Note also that that struct device dev is not necessarily defined at the
59 front of the pci_dev structure. This is to ma 59 front of the pci_dev structure. This is to make people think about what
60 they're doing when switching between the bus d 60 they're doing when switching between the bus driver and the global driver,
61 and to discourage meaningless and incorrect ca 61 and to discourage meaningless and incorrect casts between the two.
62 62
63 The PCI bus layer freely accesses the fields o 63 The PCI bus layer freely accesses the fields of struct device. It knows about
64 the structure of struct pci_dev, and it should 64 the structure of struct pci_dev, and it should know the structure of struct
65 device. Individual PCI device drivers that hav 65 device. Individual PCI device drivers that have been converted to the current
66 driver model generally do not and should not t 66 driver model generally do not and should not touch the fields of struct device,
67 unless there is a compelling reason to do so. 67 unless there is a compelling reason to do so.
68 68
69 The above abstraction prevents unnecessary pai 69 The above abstraction prevents unnecessary pain during transitional phases.
70 If it were not done this way, then when a fiel 70 If it were not done this way, then when a field was renamed or removed, every
71 downstream driver would break. On the other h 71 downstream driver would break. On the other hand, if only the bus layer
72 (and not the device layer) accesses the struct 72 (and not the device layer) accesses the struct device, it is only the bus
73 layer that needs to change. 73 layer that needs to change.
74 74
75 75
76 User Interface 76 User Interface
77 ~~~~~~~~~~~~~~ 77 ~~~~~~~~~~~~~~
78 78
79 By virtue of having a complete hierarchical vi 79 By virtue of having a complete hierarchical view of all the devices in the
80 system, exporting a complete hierarchical view 80 system, exporting a complete hierarchical view to userspace becomes relatively
81 easy. This has been accomplished by implementi 81 easy. This has been accomplished by implementing a special purpose virtual
82 file system named sysfs. 82 file system named sysfs.
83 83
84 Almost all mainstream Linux distros mount this 84 Almost all mainstream Linux distros mount this filesystem automatically; you
85 can see some variation of the following in the 85 can see some variation of the following in the output of the "mount" command::
86 86
87 $ mount 87 $ mount
88 ... 88 ...
89 none on /sys type sysfs (rw,noexec,nosuid,no 89 none on /sys type sysfs (rw,noexec,nosuid,nodev)
90 ... 90 ...
91 $ 91 $
92 92
93 The auto-mounting of sysfs is typically accomp 93 The auto-mounting of sysfs is typically accomplished by an entry similar to
94 the following in the /etc/fstab file:: 94 the following in the /etc/fstab file::
95 95
96 none /sys sysfs defaults 96 none /sys sysfs defaults 0 0
97 97
98 or something similar in the /lib/init/fstab fi 98 or something similar in the /lib/init/fstab file on Debian-based systems::
99 99
100 none /sys sysfs nodev,noexe 100 none /sys sysfs nodev,noexec,nosuid 0 0
101 101
102 If sysfs is not automatically mounted, you can 102 If sysfs is not automatically mounted, you can always do it manually with::
103 103
104 # mount -t sysfs sysfs /sys 104 # mount -t sysfs sysfs /sys
105 105
106 Whenever a device is inserted into the tree, a 106 Whenever a device is inserted into the tree, a directory is created for it.
107 This directory may be populated at each layer 107 This directory may be populated at each layer of discovery - the global layer,
108 the bus layer, or the device layer. 108 the bus layer, or the device layer.
109 109
110 The global layer currently creates two files - 110 The global layer currently creates two files - 'name' and 'power'. The
111 former only reports the name of the device. Th 111 former only reports the name of the device. The latter reports the
112 current power state of the device. It will als 112 current power state of the device. It will also be used to set the current
113 power state. 113 power state.
114 114
115 The bus layer may also create files for the de 115 The bus layer may also create files for the devices it finds while probing the
116 bus. For example, the PCI layer currently crea 116 bus. For example, the PCI layer currently creates 'irq' and 'resource' files
117 for each PCI device. 117 for each PCI device.
118 118
119 A device-specific driver may also export files 119 A device-specific driver may also export files in its directory to expose
120 device-specific data or tunable interfaces. 120 device-specific data or tunable interfaces.
121 121
122 More information about the sysfs directory lay 122 More information about the sysfs directory layout can be found in
123 the other documents in this directory and in t 123 the other documents in this directory and in the file
124 Documentation/filesystems/sysfs.rst. 124 Documentation/filesystems/sysfs.rst.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.