I'm writing a device driver that prints error message into ring buffer dmesg output.
I want to see the output of dmesg as it changes.
How can I do this?
asked Oct 13, 2013 at 8:19
-
1superuser.com/questions/289239/…– Ciro Santilli OurBigBook.comAug 9, 2015 at 20:07
Add a comment
|
7 Answers
Relatively recent dmesg versions provide a follow option (-w, --follow) which works analogously to tail -f.
Thus, just use following command:
$ dmesg -wH
(-H, --human enables user-friendly features like colors, relative time)
Those options are available for example in Fedora 19.
answered Oct 13, 2013 at 11:01
-
3
-
1
-
5Wow a sysadmin answer that uses --descriptively-named-flags instead of cryptic single-character flags. BRAVO, SIR. BRAVO. Apr 2, 2016 at 4:59
-
2
-wshould work in every version of Ubuntu from Utopic (14.10) onwards. (launchpad.net/ubuntu/+source/util-linux/+publishinghistory suggests early Utopic packages were for 2.20, but reached 2.25 by the time it was released.) Aug 20, 2017 at 13:58 -
1@Abdull or simply
journalctl -kf- however, keep in mind that journalctl usually requires root privileges for displaying system logs while dmesg doesn't. In addition, dmesg also works when journald isn't running. Jan 22 at 14:13
You can use the watch command which is intended exactly for things like this
watch -n 0.1 "dmesg | tail -n $((LINES-6))"
the $((LINES-6)) part should make it fit nicely into your terminal.
-
7If you use single quotes instead of double quotes (or escape the dollar sign), you'll to get a fresh value of $LINES for every invocation, so it will adjust if you change your terminal size.– P DaddyMar 31, 2016 at 15:22
-
Single quotes will inhibit variable expansion. Moreover, the variable is expanded only once in this example - upon invocation of
watch. Thus it won't really change between invocations ofdmesg. One would need to use a wrapper that would query the terminal state.– peterphApr 2, 2016 at 14:29 -
4That's the whole point. The single quotes prohibit the variable expansion when the shell passes the arguments to
watch, therefore watch will execute the command with the variable expression in there, which will be expanded by the shell it invokes. Each time. Try it, it works.– P DaddyApr 2, 2016 at 14:37 -
Hmm, you are right - I assume
watchusespopen(), which means another shell is spawned and the environment variable is then supplied by it (and thus updated on every execute run). Nice find.– peterphApr 2, 2016 at 15:14 -
The only option for macOS at this point as it the provided
dmsegdoes not implement the-wflag.– ijosephSep 27, 2020 at 20:59
You can't really monitor the output of dmesg directly.
However, chances are high that your module is not printing directly into the ring-buffer of dmesg, but instead uses the kernel logging facilities (which will then be displayed by dmesg). If your syslog has some sane (e.g. default) settings, these messages will most likely also show up in the kern.log logfile.
So you can do something like:
tail -f /var/log/kern.log
-
/var/log/kern.logis fairly Linux-specific. For OpenBSD (and possibly others) the dmesg stuff gets logged in /var/log/messages. Granted, some other stuff goes there too though.– kurtmOct 13, 2013 at 14:27 -
2
tail -f /var/log/{messages,kernel,dmesg,syslog}via superuser: is-it-possible-to-tail-f-the-output-of-dmesg– hereMay 2, 2014 at 7:55
You use dmesg to get log messages of the kernel.
The kernel itself logs into a ring buffer, i.e. just in memory.
Now all dmesg does is output the content of that ring buffer.
If you do dmesg -c it will also delete the ring buffer afterwards.
Therefore you could do something like while true; do dmesg -c; sleep 1; done to have something like the equivalent of a not working dmesg|tail. But this deletes the ring buffer and therefore needs root powers.
The other way is the file /proc/kmsg which allows a view on the ring buffer. You could do tail -f /proc/kmsg, but this is only allow to one process, and this is usually your logging daemon. - It's job is to read the messages and write it to real files (usually in /var/log) where they can be read. It can be configured to output all messages to a single file or different parts into different files. (But configuration depends on the logging daemon of your system.)
Therefore have a look at /var/log if there is any file which suits your needs and configure your logging daemon otherwise.
-
// , On the CEntOS 6 system I'm using, tailing and viewing /proc/kmsg produces no output. [~]$ sudo tail -f /proc/kmsg ♥% [~]$ sudo cat /proc/kmsg ♥% [~]$ Sep 4, 2015 at 19:36
-
If you are using an embedded system, like BusyBox that is common on systems like OpenWrt, it has very limited functionality and only 2-3 flags are supported.
If you want a quick and dirty way of printing dmesg output on the screen continually as events change, a simple Bash loop works fine. It's not ideal, but as I mentioned the BusyBox dmesg is missing a lot of functions. I find the following has the same effect when entered in to the command line:
$ while true; do dmesg -c ; sleep 1 ; done
You can quit the loop with Ctrl + C. The sleep 1 is to stop it battering the CPU needlessly, and the -c flag clears the buffer on each call so you don't see repeated output every second.
-
1
On systems that use systemd you can also:
# journalctl -kf
-
2Ubuntu 14.10: ➜ ~ journalctl -kf No journal files were found. May 6, 2015 at 11:46
Use these two commands from separate terminals:
while true; do dmesg -c >> test.txt;sleep 1; donetail -f test.txt
It will achieve a similar result.
-
// , This basically instructs dmesg to append output to test.txt, right? And the second command just watches that test.txt file? Sep 4, 2015 at 19:31
-
Thanks .. yes.. the 2nd command watches the changes to dmesg. cat /proc/kmsg can achieve the similar output but it doesn't save the logs to a file.– K_KSep 7, 2015 at 6:56