1 Rules on how to access information in sysfs
2 ===========================================
3
4 The kernel-exported sysfs exports internal ker
5 and depends on internal kernel structures and
6 by the kernel developers that the Linux kernel
7 internal API. Therefore, there are aspects of
8 may not be stable across kernel releases.
9
10 To minimize the risk of breaking users of sysf
11 low-level userspace applications, with a new k
12 of sysfs must follow some rules to use an as-a
13 access this filesystem. The current udev and H
14 implement this and users are encouraged to plu
15 abstractions these programs provide instead of
16
17 But if you really do want or need to access sy
18 the following rules and then your programs sho
19 versions of the sysfs interface.
20
21 - Do not use libsysfs
22 It makes assumptions about sysfs which are
23 offer any abstraction, it exposes all the
24 implementation details in its own API. The
25 reading directories and opening the files
26 Also, it is not actively maintained, in th
27 current kernel development. The goal of pr
28 to sysfs has failed; it causes more proble
29 violates many of the rules in this documen
30
31 - sysfs is always at ``/sys``
32 Parsing ``/proc/mounts`` is a waste of tim
33 system configuration bug you should not tr
34 possibly support a ``SYSFS_PATH`` environm
35 application's behavior, but never try to s
36 to mount it, if you are not an early boot
37
38 - devices are only "devices"
39 There is no such thing like class-, bus-,
40 interfaces, and such that you can rely on
41 just simply a "device". Class-, bus-, phys
42 kernel implementation details which should
43 applications that look for devices in sysf
44
45 The properties of a device are:
46
47 - devpath (``/devices/pci0000:00/0000:00:1
48
49 - identical to the DEVPATH value in the
50 at device creation and removal
51 - the unique key to the device at that p
52 - the kernel's path to the device direct
53 ``/sys``, and always starting with a s
54 - all elements of a devpath must be real
55 pointing to /sys/devices must always b
56 target and the target path must be use
57 That way the devpath to the device mat
58 kernel used at event time.
59 - using or exposing symlink values as el
60 is a bug in the application
61
62 - kernel name (``sda``, ``tty``, ``0000:00
63
64 - a directory name, identical to the las
65 - applications need to handle spaces and
66 the name
67
68 - subsystem (``block``, ``tty``, ``pci``,
69
70 - simple string, never a path or a link
71 - retrieved by reading the "subsystem"-l
72 last element of the target path
73
74 - driver (``tg3``, ``ata_piix``, ``uhci_hc
75
76 - a simple string, which may contain spa
77 link
78 - it is retrieved by reading the "driver
79 last element of the target path
80 - devices which do not have "driver"-lin
81 driver; copying the driver value in a
82 bug in the application
83
84 - attributes
85
86 - the files in the device directory or f
87 of the same device directory
88 - accessing attributes reached by a syml
89 like the "device"-link, is a bug in th
90
91 Everything else is just a kernel driver-co
92 that should not be assumed to be stable ac
93
94 - Properties of parent devices never belong in
95 Always look at the parent devices themselv
96 context properties. If the device ``eth0``
97 "driver"-link, then this device does not h
98 Never copy any property of the parent-devi
99 device properties may change dynamically w
100 child device.
101
102 - Hierarchy in a single device tree
103 There is only one valid place in sysfs whe
104 and this is below: ``/sys/devices.``
105 It is planned that all device directories
106 below this directory.
107
108 - Classification by subsystem
109 There are currently three places for class
110 ``/sys/block,`` ``/sys/class`` and ``/sys/
111 not contain any device directories themsel
112 symlinks pointing to the unified ``/sys/de
113 All three places have completely different
114 device information. It is planned to merge
115 classification directories into one place
116 following the layout of the bus directorie
117 classes, including the converted block sub
118 there.
119 The devices belonging to a subsystem will
120 "devices" directory at ``/sys/subsystem/<n
121
122 If ``/sys/subsystem`` exists, ``/sys/bus``
123 can be ignored. If it does not exist, you
124 places, as the kernel is free to move a su
125 the other, as long as the devices are stil
126 subsystem name.
127
128 Assuming ``/sys/class/<subsystem>`` and ``
129 ``/sys/block`` and ``/sys/class/block`` ar
130 the application.
131
132 - Block
133 The converted block subsystem at ``/sys/cl
134 ``/sys/subsystem/block`` will contain the
135 at the same level, never in a hierarchy. A
136 contain only disks and not partition devic
137 a bug in the application.
138
139 - "device"-link and <subsystem>:<kernel name>-
140 Never depend on the "device"-link. The "de
141 for the old layout, where class devices ar
142 ``/sys/devices/`` like the bus devices. If
143 device directory does not end in ``/sys/de
144 "device"-link to find the parent devices i
145 single valid use of the "device"-link; it
146 path as an element. Assuming the existence
147 a device in ``/sys/devices/`` is a bug in
148 Accessing ``/sys/class/net/eth0/device`` i
149
150 Never depend on the class-specific links b
151 directory. These links are also a workaro
152 that class devices are not created in ``/s
153 directory does not contain directories for
154 may be used to find the child devices in `
155 valid use of these links; they must never
156 element. Assuming the existence of these l
157 real child device directories in the ``/sy
158 the application.
159
160 It is planned to remove all these links wh
161 directories live in ``/sys/devices.``
162
163 - Position of devices along device chain can c
164 Never depend on a specific parent device p
165 or the chain of parent devices. The kernel
166 the chain. You must always request the par
167 by its subsystem value. You need to walk u
168 the device that matches the expected subsy
169 position of a parent device or exposing re
170 access the chain of parents is a bug in th
171
172 - When reading and writing sysfs device attrib
173 on specific error codes wherever possible.
174 the error handling implementation within t
175
176 In general, failures to read or write sysf
177 propagate errors wherever possible. Common
178 limited to:
179
180 ``-EIO``: The read or store operation
181 returned by the sysfs system itself if
182 is ``NULL``.
183
184 ``-ENXIO``: The read or store operatio
185
186 Error codes will not be changed without go
187 to error codes result in user-space breaka
188 the offending change will be reverted.
189
190 Userspace applications can, however, expec
191 the attribute files to remain consistent i
192 attribute change in the context of a given
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.