1SYSKLOGD(8) Linux System Administration SYSKLOGD(8)
2
6 sysklogd - Linux system logging utilities.
7
9 syslogd [ -a socket ] [ -d ] [ -f config file ] [ -h ] [ -l hostlist ]
10 [ -m interval ] [ -n ] [ -p socket ] [ -r ] [ -s domainlist ] [ -v ]
11
13 Sysklogd provides two system utilities which provide support for system
14 logging and kernel message trapping. Support of both internet and unix
15 domain sockets enables this utility package to support both local and
16 remote logging.
17 System logging is provided by a version of syslogd(8) derived from the
19 stock BSD sources. Support for kernel logging is provided by the
20 klogd(8) utility which allows kernel logging to be conducted in either
21 a standalone fashion or as a client of syslogd.
22 Syslogd provides a kind of logging that many modern programs use.
24 Every logged message contains at least a time and a hostname field,
25 normally a program name field, too, but that depends on how trusty the
26 logging program is.
27 While the syslogd sources have been heavily modified a couple of notes
29 are in order. First of all there has been a systematic attempt to
30 insure that syslogd follows its default, standard BSD behavior. The
31 second important concept to note is that this version of syslogd inter‐
32 acts transparently with the version of syslog found in the standard
33 libraries. If a binary linked to the standard shared libraries fails
34 to function correctly we would like an example of the anomalous behav‐
35 ior.
36 The main configuration file /etc/syslog.conf or an alternative file,
38 given with the -f option, is read at startup. Any lines that begin
39 with the hash mark (``#'') and empty lines are ignored. If an error
40 occurs during parsing the whole line is ignored.
41
43 -a socket
44 Using this argument you can specify additional sockets from that
45 syslogd has to listen to. This is needed if you're going to let
46 some daemon run within a chroot() environment. You can use up
47 to 19 additional sockets. If your environment needs even more,
48 you have to increase the symbol MAXFUNIX within the syslogd.c
49 source file. An example for a chroot() daemon is described by
50 the people from OpenBSD at <http://www.guides.sk/psionic/dns/>.
51 -d Turns on debug mode. Using this the daemon will not proceed a
53 fork(2) to set itself in the background, but opposite to that
54 stay in the foreground and write much debug information on the
55 current tty. See the DEBUGGING section for more information.
56 -f config file
58 Specify an alternative configuration file instead of /etc/sys‐
59 log.conf, which is the default.
60 -h By default syslogd will not forward messages it receives from
62 remote hosts. Specifying this switch on the command line will
63 cause the log daemon to forward any remote messages it receives
64 to forwarding hosts which have been defined. This can cause
65 syslog loops that fill up hard disks quite fast and thus needs
66 to be used with caution.
67 -l hostlist
69 Specify a hostname that should be logged only with its simple
70 hostname and not the fqdn. Multiple hosts may be specified
71 using the colon (``:'') separator.
72 -m interval
74 The syslogd logs a mark timestamp regularly. The default inter‐
75 val between two -- MARK -- lines is 20 minutes. This can be
76 changed with this option. Setting the interval to zero turns it
77 off entirely. Depending on other log messages generated these
78 lines may not be written consecutively.
79 -n Avoid auto-backgrounding. This is needed especially if the sys‐
81 logd is started and controlled by init(8).
82 -p socket
84 You can specify an alternative unix domain socket instead of
85 /dev/log.
86 -r This option will enable the facility to receive message from the
88 network using an internet domain socket with the syslog service
89 (see services(5)). The default is to not receive any messages
90 from the network.
91 This option is introduced in version 1.3 of the sysklogd pack‐
93 age. Please note that the default behavior is the opposite of
94 how older versions behave, so you might have to turn this on.
95 -s domainlist
97 Specify a domainname that should be stripped off before logging.
98 Multiple domains may be specified using the colon (``:'') sepa‐
99 rator. Please be advised that no sub-domains may be specified
100 but only entire domains. For example if -s north.de is speci‐
101 fied and the host logging resolves to satu.infodrom.north.de no
102 domain would be cut, you will have to specify two domains like:
103 -s north.de:infodrom.north.de.
104 -v Print version and exit.
106
108 Syslogd reacts to a set of signals. You may easily send a signal to
109 syslogd using the following:
110 kill -SIGNAL `cat /var/run/syslogd.pid`
112 SIGHUP This lets syslogd perform a re-initialization. All open files
114 are closed, the configuration file (default is /etc/syslog.conf)
115 will be reread and the syslog(3) facility is started again.
116 SIGTERM
118 The syslogd will die.
119 SIGINT, SIGQUIT
121 If debugging is enabled these are ignored, otherwise syslogd
122 will die.
123 SIGUSR1
125 Switch debugging on/off. This option can only be used if sys‐
126 logd is started with the -d debug option.
127 SIGCHLD
129 Wait for childs if some were born, because of wall'ing messages.
130
132 Syslogd uses a slightly different syntax for its configuration file
133 than the original BSD sources. Originally all messages of a specific
134 priority and above were forwarded to the log file.
135 For example the following line caused ALL output from daemons
137 using the daemon facilities (debug is the lowest priority, so
138 every higher will also match) to go into /usr/adm/daemons:
139 # Sample syslog.conf
141 daemon.debug /usr/adm/daemons
142 Under the new scheme this behavior remains the same. The difference is
144 the addition of four new specifiers, the asterisk (*) wildcard, the
145 equation sign (=), the exclamation mark (!), and the minus sign (-).
146 The * specifies that all messages for the specified facility are to be
148 directed to the destination. Note that this behavior is degenerate
149 with specifying a priority level of debug. Users have indicated that
150 the asterisk notation is more intuitive.
151 The = wildcard is used to restrict logging to the specified priority
153 class. This allows, for example, routing only debug messages to a par‐
154 ticular logging source.
155 For example the following line in syslog.conf would direct debug
157 messages from all sources to the /usr/adm/debug file.
158 # Sample syslog.conf
160 *.=debug /usr/adm/debug
161 The ! is used to exclude logging of the specified priorities. This
163 affects all (!) possibilities of specifying priorities.
164 For example the following lines would log all messages of the
166 facility mail except those with the priority info to the
167 /usr/adm/mail file. And all messages from news.info (including)
168 to news.crit (excluding) would be logged to the /usr/adm/news
169 file.
170 # Sample syslog.conf
172 mail.*;mail.!=info /usr/adm/mail
173 news.info;news.!crit /usr/adm/news
174 You may use it intuitively as an exception specifier. The above men‐
176 tioned interpretation is simply inverted. Doing that you may use
177 mail.none
179 or
180 mail.!*
181 or
182 mail.!debug
183 to skip every message that comes with a mail facility. There is much
185 room to play with it. :-)
186 The - may only be used to prefix a filename if you want to omit
188 sync'ing the file after every write to it.
189 This may take some acclimatization for those individuals used to the
191 pure BSD behavior but testers have indicated that this syntax is some‐
192 what more flexible than the BSD behavior. Note that these changes
193 should not affect standard syslog.conf(5) files. You must specifically
194 modify the configuration files to obtain the enhanced behavior.
195
197 These modifications provide network support to the syslogd facility.
198 Network support means that messages can be forwarded from one node run‐
199 ning syslogd to another node running syslogd where they will be actu‐
200 ally logged to a disk file.
201 To enable this you have to specify the -r option on the command line.
203 The default behavior is that syslogd won't listen to the network.
204 The strategy is to have syslogd listen on a unix domain socket for
206 locally generated log messages. This behavior will allow syslogd to
207 inter-operate with the syslog found in the standard C library. At the
208 same time syslogd listens on the standard syslog port for messages for‐
209 warded from other hosts. To have this work correctly the services(5)
210 files (typically found in /etc) must have the following entry:
211 syslog 514/udp
213 If this entry is missing syslogd neither can receive remote messages
215 nor send them, because the UDP port cant be opened. Instead syslogd
216 will die immediately, blowing out an error message.
217 To cause messages to be forwarded to another host replace the normal
219 file line in the syslog.conf file with the name of the host to which
220 the messages is to be sent prepended with an @.
221 For example, to forward ALL messages to a remote host use the
223 following syslog.conf entry:
224 # Sample syslogd configuration file to
226 # messages to a remote host forward all.
227 *.* @hostname
228 To forward all kernel messages to a remote host the configura‐
230 tion file would be as follows:
231 # Sample configuration file to forward all kernel
233 # messages to a remote host.
234 kern.* @hostname
235 If the remote hostname cannot be resolved at startup, because the name-
237 server might not be accessible (it may be started after syslogd) you
238 don't have to worry. Syslogd will retry to resolve the name ten times
239 and then complain. Another possibility to avoid this is to place the
240 hostname in /etc/hosts.
241 With normal syslogds you would get syslog-loops if you send out mes‐
243 sages that were received from a remote host to the same host (or more
244 complicated to a third host that sends it back to the first one, and so
245 on). In my domain (Infodrom Oldenburg) we accidently got one and our
246 disks filled up with the same single message. :-(
247 To avoid this no messages received from a remote host are sent out to
249 another (or the same) remote host anymore. If you experience are setup
250 in which you need this behaviour, please use the -h command line
251 switch. However, this option needs to be handled with caution since a
252 syslog loop can fill up hard disks quite fast.
253 If the remote host is located in the same domain as the host, syslogd
255 is running on, only the simple hostname will be logged instead of the
256 whole fqdn.
257 In a local network you may provide a central log server to have all the
259 important information kept on one machine. If the network consists of
260 different domains you don't have to complain about logging fully quali‐
261 fied names instead of simple hostnames. You may want to use the strip-
262 domain feature -s of this server. You can tell the syslogd to strip
263 off several domains other than the one the server is located in and
264 only log simple hostnames.
265 Using the -l option there's also a possibility to define single hosts
267 as local machines. This, too, results in logging only their simple
268 hostnames and not the fqdns.
269 The UDP socket used to forward messages to remote hosts or to receive
271 messages from them is only opened when it is needed. In releases prior
272 to 1.3-23 it was opened every time but not opened for reading or for‐
273 warding respectively.
274
276 This version of syslogd has support for logging output to named pipes
277 (fifos). A fifo or named pipe can be used as a destination for log
278 messages by prepending a pipy symbol (``|'') to the name of the file.
279 This is handy for debugging. Note that the fifo must be created with
280 the mkfifo command before syslogd is started.
281 The following configuration file routes debug messages from the
283 kernel to a fifo:
284 # Sample configuration to route kernel debugging
286 # messages ONLY to /usr/adm/debug which is a
287 # named pipe.
288 kern.=debug |/usr/adm/debug
289
291 There is probably one important consideration when installing this ver‐
292 sion of syslogd. This version of syslogd is dependent on proper for‐
293 matting of messages by the syslog function. The functioning of the
294 syslog function in the shared libraries changed somewhere in the region
295 of libc.so.4.[2-4].n. The specific change was to null-terminate the
296 message before transmitting it to the /dev/log socket. Proper func‐
297 tioning of this version of syslogd is dependent on null-termination of
298 the message.
299 This problem will typically manifest itself if old statically linked
301 binaries are being used on the system. Binaries using old versions of
302 the syslog function will cause empty lines to be logged followed by the
303 message with the first character in the message removed. Relinking
304 these binaries to newer versions of the shared libraries will correct
305 this problem.
306 Both the syslogd(8) and the klogd(8) can either be run from init(8) or
308 started as part of the rc.* sequence. If it is started from init the
309 option -n must be set, otherwise you'll get tons of syslog daemons
310 started. This is because init(8) depends on the process ID.
311
313 There is the potential for the syslogd daemon to be used as a conduit
314 for a denial of service attack. Thanks go to John Morrison (jmor‐
315 riso@rflab.ee.ubc.ca) for alerting me to this potential. A rogue pro‐
316 gram(mer) could very easily flood the syslogd daemon with syslog mes‐
317 sages resulting in the log files consuming all the remaining space on
318 the filesystem. Activating logging over the inet domain sockets will
319 of course expose a system to risks outside of programs or individuals
320 on the local machine.
321 There are a number of methods of protecting a machine:
323 1. Implement kernel firewalling to limit which hosts or networks
325 have access to the 514/UDP socket.
326 2. Logging can be directed to an isolated or non-root filesystem
328 which, if filled, will not impair the machine.
329 3. The ext2 filesystem can be used which can be configured to limit
331 a certain percentage of a filesystem to usage by root only.
332 NOTE that this will require syslogd to be run as a non-root
333 process. ALSO NOTE that this will prevent usage of remote log‐
334 ging since syslogd will be unable to bind to the 514/UDP socket.
335 4. Disabling inet domain sockets will limit risk to the local
337 machine.
338 5. Use step 4 and if the problem persists and is not secondary to a
340 rogue program/daemon get a 3.5 ft (approx. 1 meter) length of
341 sucker rod* and have a chat with the user in question.
342 Sucker rod def. — 3/4, 7/8 or 1in. hardened steel rod, male
344 threaded on each end. Primary use in the oil industry in West‐
345 ern North Dakota and other locations to pump 'suck' oil from oil
346 wells. Secondary uses are for the construction of cattle feed
347 lots and for dealing with the occasional recalcitrant or bel‐
348 ligerent individual.
349
351 When debugging is turned on using -d option then syslogd will be very
352 verbose by writing much of what it does on stdout. Whenever the con‐
353 figuration file is reread and re-parsed you'll see a tabular, corre‐
354 sponding to the internal data structure. This tabular consists of four
355 fields:
356 number This field contains a serial number starting by zero. This num‐
358 ber represents the position in the internal data structure (i.e.
359 the array). If one number is left out then there might be an
360 error in the corresponding line in /etc/syslog.conf.
361 pattern
363 This field is tricky and represents the internal structure
364 exactly. Every column stands for a facility (refer to sys‐
365 log(3)). As you can see, there are still some facilities left
366 free for former use, only the left most are used. Every field
367 in a column represents the priorities (refer to syslog(3)).
368 action This field describes the particular action that takes place
370 whenever a message is received that matches the pattern. Refer
371 to the syslog.conf(5) manpage for all possible actions.
372 arguments
374 This field shows additional arguments to the actions in the last
375 field. For file-logging this is the filename for the logfile;
376 for user-logging this is a list of users; for remote logging
377 this is the hostname of the machine to log to; for console-log‐
378 ging this is the used console; for tty-logging this is the spec‐
379 ified tty; wall has no additional arguments.
380
382 /etc/syslog.conf
383 Configuration file for syslogd. See syslog.conf(5) for exact
384 information.
385 /dev/log
386 The Unix domain socket to from where local syslog messages are
387 read.
388 /var/run/syslogd.pid
389 The file containing the process id of syslogd.
390
392 If an error occurs in one line the whole rule is ignored.
393 Syslogd doesn't change the filemode of opened logfiles at any stage of
395 process. If a file is created it is world readable. If you want to
396 avoid this, you have to create it and change permissions on your own.
397 This could be done in combination with rotating logfiles using the
398 savelog(8) program that is shipped in the smail 3.x distribution.
399 Remember that it might be a security hole if everybody is able to read
400 auth.* messages as these might contain passwords.
401
403 syslog.conf(5), klogd(8), logger(1), syslog(2), syslog(3), services(5),
404 savelog(8).
405
407 The system log daemon syslogd is originally taken from BSD sources,
408 Greg Wettstein <greg@wind.enjellic.com> performed the port to Linux,
409 Martin Schulze <joey@infodrom.org> fixed some bugs, added several new
410 features and took over maintenance.
411Version 1.5 27 May 2007 SYSKLOGD(8)