xref: /linux/Documentation/ABI/testing/dev-kmsg (revision bc885f1ab6de0d38c6956a71b0126543b64875b0)
1What:		/dev/kmsg
2Date:		Mai 2012
3KernelVersion:	3.5
4Contact:	Kay Sievers <kay@vrfy.org>
5Description:	The /dev/kmsg character device node provides userspace access
6		to the kernel's printk buffer.
7
8		Injecting messages:
9		Every write() to the opened device node places a log entry in
10		the kernel's printk buffer.
11
12		The logged line can be prefixed with a <N> syslog prefix, which
13		carries the syslog priority and facility. The single decimal
14		prefix number is composed of the 3 lowest bits being the syslog
15		priority and the next 8 bits the syslog facility number.
16
17		If no prefix is given, the priority number is the default kernel
18		log priority and the facility number is set to LOG_USER (1). It
19		is not possible to inject messages from userspace with the
20		facility number LOG_KERN (0), to make sure that the origin of
21		the messages can always be reliably determined.
22
23		Accessing the buffer:
24		Every read() from the opened device node receives one record
25		of the kernel's printk buffer.
26
27		The first read() directly following an open() always returns
28		first message in the buffer; there is no kernel-internal
29		persistent state; many readers can concurrently open the device
30		and read from it, without affecting other readers.
31
32		Every read() will receive the next available record. If no more
33		records are available read() will block, or if O_NONBLOCK is
34		used -EAGAIN returned.
35
36		Messages in the record ring buffer get overwritten as whole,
37		there are never partial messages received by read().
38
39		In case messages get overwritten in the circular buffer while
40		the device is kept open, the next read() will return -EPIPE,
41		and the seek position be updated to the next available record.
42		Subsequent reads() will return available records again.
43
44		Unlike the classic syslog() interface, the 64 bit record
45		sequence numbers allow to calculate the amount of lost
46		messages, in case the buffer gets overwritten. And they allow
47		to reconnect to the buffer and reconstruct the read position
48		if needed, without limiting the interface to a single reader.
49
50		The device supports seek with the following parameters:
51		SEEK_SET, 0
52		  seek to the first entry in the buffer
53		SEEK_END, 0
54		  seek after the last entry in the buffer
55		SEEK_DATA, 0
56		  seek after the last record available at the time
57		  the last SYSLOG_ACTION_CLEAR was issued.
58
59		Due to the record nature of this interface with a "read all"
60		behavior and the specific positions each seek operation sets,
61		SEEK_CUR is not supported, returning -ESPIPE (invalid seek) to
62		errno whenever requested.
63
64		Other seek operations or offsets are not supported because of
65		the special behavior this device has. The device allows to read
66		or write only whole variable length messages (records) that are
67		stored in a ring buffer.
68
69		Because of the non-standard behavior also the error values are
70		non-standard. -ESPIPE is returned for non-zero offset. -EINVAL
71		is returned for other operations, e.g. SEEK_CUR. This behavior
72		and values are historical and could not be modified without the
73		risk of breaking userspace.
74
75		The output format consists of a prefix carrying the syslog
76		prefix including priority and facility, the 64 bit message
77		sequence number and the monotonic timestamp in microseconds,
78		and a flag field. All fields are separated by a ','.
79
80		Future extensions might add more comma separated values before
81		the terminating ';'. Unknown fields and values should be
82		gracefully ignored.
83
84		The human readable text string starts directly after the ';'
85		and is terminated by a '\n'. Untrusted values derived from
86		hardware or other facilities are printed, therefore
87		all non-printable characters and '\' itself in the log message
88		are escaped by "\x00" C-style hex encoding.
89
90		A line starting with ' ', is a continuation line, adding
91		key/value pairs to the log message, which provide the machine
92		readable context of the message, for reliable processing in
93		userspace.
94
95		Example:
96		7,160,424069,-;pci_root PNP0A03:00: host bridge window [io  0x0000-0x0cf7] (ignored)
97		 SUBSYSTEM=acpi
98		 DEVICE=+acpi:PNP0A03:00
99		6,339,5140900,-;NET: Registered protocol family 10
100		30,340,5690716,-;udevd[80]: starting version 181
101
102		The DEVICE= key uniquely identifies devices the following way:
103		  b12:8        - block dev_t
104		  c127:3       - char dev_t
105		  n8           - netdev ifindex
106		  +sound:card0 - subsystem:devname
107
108		The flags field carries '-' by default. A 'c' indicates a
109		fragment of a line. Note, that these hints about continuation
110		lines are not necessarily correct, and the stream could be
111		interleaved with unrelated messages, but merging the lines in
112		the output usually produces better human readable results. A
113		similar logic is used internally when messages are printed to
114		the console, /proc/kmsg or the syslog() syscall.
115
116		By default, kernel tries to avoid fragments by concatenating
117		when it can and fragments are rare; however, when extended
118		console support is enabled, the in-kernel concatenation is
119		disabled and /dev/kmsg output will contain more fragments. If
120		the log consumer performs concatenation, the end result
121		should be the same. In the future, the in-kernel concatenation
122		may be removed entirely and /dev/kmsg users are recommended to
123		implement fragment handling.
124
125Users:		dmesg(1), userspace kernel log consumers
126