1============================= 2The Linux Kernel Device Model 3============================= 4 5Patrick Mochel <mochel@digitalimplant.org> 6 7Drafted 26 August 2002 8Updated 31 January 2006 9 10 11Overview 12~~~~~~~~ 13 14The Linux Kernel Driver Model is a unification of all the disparate driver 15models that were previously used in the kernel. It is intended to augment the 16bus-specific drivers for bridges and devices by consolidating a set of data 17and operations into globally accessible data structures. 18 19Traditional driver models implemented some sort of tree-like structure 20(sometimes just a list) for the devices they control. There wasn't any 21uniformity across the different bus types. 22 23The current driver model provides a common, uniform data model for describing 24a bus and the devices that can appear under the bus. The unified bus 25model includes a set of common attributes which all busses carry, and a set 26of common callbacks, such as device discovery during bus probing, bus 27shutdown, bus power management, etc. 28 29The common device and bridge interface reflects the goals of the modern 30computer: namely the ability to do seamless device "plug and play", power 31management, and hot plug. In particular, the model dictated by Intel and 32Microsoft (namely ACPI) ensures that almost every device on almost any bus 33on an x86-compatible system can work within this paradigm. Of course, 34not every bus is able to support all such operations, although most 35buses support most of those operations. 36 37 38Downstream Access 39~~~~~~~~~~~~~~~~~ 40 41Common data fields have been moved out of individual bus layers into a common 42data structure. These fields must still be accessed by the bus layers, 43and sometimes by the device-specific drivers. 44 45Other bus layers are encouraged to do what has been done for the PCI layer. 46struct pci_dev now looks like this:: 47 48 struct pci_dev { 49 ... 50 51 struct device dev; /* Generic device interface */ 52 ... 53 }; 54 55Note first that the struct device dev within the struct pci_dev is 56statically allocated. This means only one allocation on device discovery. 57 58Note also that that struct device dev is not necessarily defined at the 59front of the pci_dev structure. This is to make people think about what 60they're doing when switching between the bus driver and the global driver, 61and to discourage meaningless and incorrect casts between the two. 62 63The PCI bus layer freely accesses the fields of struct device. It knows about 64the structure of struct pci_dev, and it should know the structure of struct 65device. Individual PCI device drivers that have been converted to the current 66driver model generally do not and should not touch the fields of struct device, 67unless there is a compelling reason to do so. 68 69The above abstraction prevents unnecessary pain during transitional phases. 70If it were not done this way, then when a field was renamed or removed, every 71downstream driver would break. On the other hand, if only the bus layer 72(and not the device layer) accesses the struct device, it is only the bus 73layer that needs to change. 74 75 76User Interface 77~~~~~~~~~~~~~~ 78 79By virtue of having a complete hierarchical view of all the devices in the 80system, exporting a complete hierarchical view to userspace becomes relatively 81easy. This has been accomplished by implementing a special purpose virtual 82file system named sysfs. 83 84Almost all mainstream Linux distros mount this filesystem automatically; you 85can see some variation of the following in the output of the "mount" command:: 86 87 $ mount 88 ... 89 none on /sys type sysfs (rw,noexec,nosuid,nodev) 90 ... 91 $ 92 93The auto-mounting of sysfs is typically accomplished by an entry similar to 94the following in the /etc/fstab file:: 95 96 none /sys sysfs defaults 0 0 97 98or something similar in the /lib/init/fstab file on Debian-based systems:: 99 100 none /sys sysfs nodev,noexec,nosuid 0 0 101 102If sysfs is not automatically mounted, you can always do it manually with:: 103 104 # mount -t sysfs sysfs /sys 105 106Whenever a device is inserted into the tree, a directory is created for it. 107This directory may be populated at each layer of discovery - the global layer, 108the bus layer, or the device layer. 109 110The global layer currently creates two files - 'name' and 'power'. The 111former only reports the name of the device. The latter reports the 112current power state of the device. It will also be used to set the current 113power state. 114 115The bus layer may also create files for the devices it finds while probing the 116bus. For example, the PCI layer currently creates 'irq' and 'resource' files 117for each PCI device. 118 119A device-specific driver may also export files in its directory to expose 120device-specific data or tunable interfaces. 121 122More information about the sysfs directory layout can be found in 123the other documents in this directory and in the file 124Documentation/filesystems/sysfs.txt. 125