1 Rules on how to access information in the Linux kernel sysfs
3 The kernel-exported sysfs exports internal kernel implementation details
4 and depends on internal kernel structures and layout. It is agreed upon
5 by the kernel developers that the Linux kernel does not provide a stable
6 internal API. Therefore, there are aspects of the sysfs interface that
7 may not be stable across kernel releases.
9 To minimize the risk of breaking users of sysfs, which are in most cases
10 low-level userspace applications, with a new kernel release, the users
11 of sysfs must follow some rules to use an as-abstract-as-possible way to
12 access this filesystem. The current udev and HAL programs already
13 implement this and users are encouraged to plug, if possible, into the
14 abstractions these programs provide instead of accessing sysfs directly.
16 But if you really do want or need to access sysfs directly, please follow
17 the following rules and then your programs should work with future
18 versions of the sysfs interface.
21 It makes assumptions about sysfs which are not true. Its API does not
22 offer any abstraction, it exposes all the kernel driver-core
23 implementation details in its own API. Therefore it is not better than
24 reading directories and opening the files yourself.
25 Also, it is not actively maintained, in the sense of reflecting the
26 current kernel development. The goal of providing a stable interface
27 to sysfs has failed; it causes more problems than it solves. It
28 violates many of the rules in this document.
30 - sysfs is always at /sys
31 Parsing /proc/mounts is a waste of time. Other mount points are a
32 system configuration bug you should not try to solve. For test cases,
33 possibly support a SYSFS_PATH environment variable to overwrite the
34 application's behavior, but never try to search for sysfs. Never try
35 to mount it, if you are not an early boot script.
37 - devices are only "devices"
38 There is no such thing like class-, bus-, physical devices,
39 interfaces, and such that you can rely on in userspace. Everything is
40 just simply a "device". Class-, bus-, physical, ... types are just
41 kernel implementation details which should not be expected by
42 applications that look for devices in sysfs.
44 The properties of a device are:
45 o devpath (/devices/pci0000:00/0000:00:1d.1/usb2/2-2/2-2:1.0)
46 - identical to the DEVPATH value in the event sent from the kernel
47 at device creation and removal
48 - the unique key to the device at that point in time
49 - the kernel's path to the device directory without the leading
50 /sys, and always starting with with a slash
51 - all elements of a devpath must be real directories. Symlinks
52 pointing to /sys/devices must always be resolved to their real
53 target and the target path must be used to access the device.
54 That way the devpath to the device matches the devpath of the
55 kernel used at event time.
56 - using or exposing symlink values as elements in a devpath string
57 is a bug in the application
59 o kernel name (sda, tty, 0000:00:1f.2, ...)
60 - a directory name, identical to the last element of the devpath
61 - applications need to handle spaces and characters like '!' in
64 o subsystem (block, tty, pci, ...)
65 - simple string, never a path or a link
66 - retrieved by reading the "subsystem"-link and using only the
67 last element of the target path
69 o driver (tg3, ata_piix, uhci_hcd)
70 - a simple string, which may contain spaces, never a path or a
72 - it is retrieved by reading the "driver"-link and using only the
73 last element of the target path
74 - devices which do not have "driver"-link just do not have a
75 driver; copying the driver value in a child device context is a
76 bug in the application
79 - the files in the device directory or files below subdirectories
80 of the same device directory
81 - accessing attributes reached by a symlink pointing to another device,
82 like the "device"-link, is a bug in the application
84 Everything else is just a kernel driver-core implementation detail
85 that should not be assumed to be stable across kernel releases.
87 - Properties of parent devices never belong into a child device.
88 Always look at the parent devices themselves for determining device
89 context properties. If the device 'eth0' or 'sda' does not have a
90 "driver"-link, then this device does not have a driver. Its value is empty.
91 Never copy any property of the parent-device into a child-device. Parent
92 device properties may change dynamically without any notice to the
95 - Hierarchy in a single device tree
96 There is only one valid place in sysfs where hierarchy can be examined
97 and this is below: /sys/devices.
98 It is planned that all device directories will end up in the tree
101 - Classification by subsystem
102 There are currently three places for classification of devices:
103 /sys/block, /sys/class and /sys/bus. It is planned that these will
104 not contain any device directories themselves, but only flat lists of
105 symlinks pointing to the unified /sys/devices tree.
106 All three places have completely different rules on how to access
107 device information. It is planned to merge all three
108 classification directories into one place at /sys/subsystem,
109 following the layout of the bus directories. All buses and
110 classes, including the converted block subsystem, will show up
112 The devices belonging to a subsystem will create a symlink in the
113 "devices" directory at /sys/subsystem/<name>/devices.
115 If /sys/subsystem exists, /sys/bus, /sys/class and /sys/block can be
116 ignored. If it does not exist, you always have to scan all three
117 places, as the kernel is free to move a subsystem from one place to
118 the other, as long as the devices are still reachable by the same
121 Assuming /sys/class/<subsystem> and /sys/bus/<subsystem>, or
122 /sys/block and /sys/class/block are not interchangeable is a bug in
126 The converted block subsystem at /sys/class/block or
127 /sys/subsystem/block will contain the links for disks and partitions
128 at the same level, never in a hierarchy. Assuming the block subsystem to
129 contain only disks and not partition devices in the same flat list is
130 a bug in the application.
132 - "device"-link and <subsystem>:<kernel name>-links
133 Never depend on the "device"-link. The "device"-link is a workaround
134 for the old layout, where class devices are not created in
135 /sys/devices/ like the bus devices. If the link-resolving of a
136 device directory does not end in /sys/devices/, you can use the
137 "device"-link to find the parent devices in /sys/devices/. That is the
138 single valid use of the "device"-link; it must never appear in any
139 path as an element. Assuming the existence of the "device"-link for
140 a device in /sys/devices/ is a bug in the application.
141 Accessing /sys/class/net/eth0/device is a bug in the application.
143 Never depend on the class-specific links back to the /sys/class
144 directory. These links are also a workaround for the design mistake
145 that class devices are not created in /sys/devices. If a device
146 directory does not contain directories for child devices, these links
147 may be used to find the child devices in /sys/class. That is the single
148 valid use of these links; they must never appear in any path as an
149 element. Assuming the existence of these links for devices which are
150 real child device directories in the /sys/devices tree is a bug in
153 It is planned to remove all these links when all class device
154 directories live in /sys/devices.
156 - Position of devices along device chain can change.
157 Never depend on a specific parent device position in the devpath,
158 or the chain of parent devices. The kernel is free to insert devices into
159 the chain. You must always request the parent device you are looking for
160 by its subsystem value. You need to walk up the chain until you find
161 the device that matches the expected subsystem. Depending on a specific
162 position of a parent device or exposing relative paths using "../" to
163 access the chain of parents is a bug in the application.