How to wait for kernel messages?

How to wait for kernel messages?

[Sebastian Piecha]

I have some problems with one NIC. Due to lack of time as an 
workaround I'd like to wait for the kernel message "NETDEV WATCHDOG: 
eth0: transmit timed out" and ifconfig down/up the NIC.

How can I trigger any action by such a kernel message? Do I have to 
grep the kernel log?

Mit freundlichen Gruessen/Best regards,
Sebastian Piecha

EMail: spi@gmxpro.de

Re: How to wait for kernel messages?

[Richard B. Johnson]

On Mon, 13 Oct 2003, Sebastian Piecha wrote:

> I have some problems with one NIC. Due to lack of time as an
> workaround I'd like to wait for the kernel message "NETDEV WATCHDOG:
> eth0: transmit timed out" and ifconfig down/up the NIC.
>
> How can I trigger any action by such a kernel message? Do I have to
> grep the kernel log?
>
> Mit freundlichen Gruessen/Best regards,
> Sebastian Piecha

`cat /proc/kmsg`

...or to wait forever for each message, just open /proc/kmsg and
post a read. Note that kernel logger is also waiting for such
messages so you might not get them all.

Of course the 'Unix-way' is to just do `tail -f /var/log/messages`,
or whatever your log-file name is.


Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (797.90 BogoMips).
            Note 96.31% of all statistics are fiction.

Re: How to wait for kernel messages?

[Jim Keniston]

On Mon, 13 Oct 2003, Sebastian Piecha wrote:

> I have some problems with one NIC. Due to lack of time as an
> workaround I'd like to wait for the kernel message "NETDEV WATCHDOG:
> eth0: transmit timed out" and ifconfig down/up the NIC.
>
> How can I trigger any action by such a kernel message? Do I have to
> grep the kernel log?
>
> Mit freundlichen Gruessen/Best regards,
> Sebastian Piecha

If you had LTC's Event Logging installed, and you had your kernel
configured to forward printk messages to the event log, you could
do something like this:

evlnotify -f 'data~"NETDEV WATCHDOG: eth0: transmit timed out"' \
	-a 'ifconfig eth0 down; ifconfig eth0 up'

That would cause the ifconfig down/up to run each time the indicated
printk happened.

Jim Keniston
IBM Linux Technology Center

Re: How to wait for kernel messages?

[Sebastian Piecha]

[-- Attachment #1: Mail message body --]
[-- Type: text/plain, Size: 641 bytes --]

> On Mon, 13 Oct 2003, Sebastian Piecha wrote:
> 
> > I have some problems with one NIC. Due to lack of time as an
> > workaround I'd like to wait for the kernel message "NETDEV WATCHDOG:
> > eth0: transmit timed out" and ifconfig down/up the NIC.
> >
> > How can I trigger any action by such a kernel message? Do I have to
> > grep the kernel log?
> >

Thanks a lot for all the hints. I'm using now a fifo to which syslogd 
is logging kernel messages. A shell script greps the right message.

I attach the shell script inline for whom it may be interesting.


Mit freundlichen Gruessen/Best regards,
Sebastian Piecha

EMail: spi@gmxpro.de


[-- Attachment #2: Attachment information. --]
[-- Type: text/plain, Size: 466 bytes --]

The following section of this message contains a file attachment
prepared for transmission using the Internet MIME message format.
If you are using Pegasus Mail, or any other MIME-compliant system,
you should be able to save it or view it from within your mailer.
If you cannot, please ask your system administrator for assistance.

   ---- File information -----------
     File:  check_nic.txt
     Date:  15 Oct 2003, 0:56
     Size:  771 bytes.
     Type:  Text

[-- Attachment #3: check_nic.txt --]
[-- Type: Application/Octet-stream, Size: 735 bytes --]

#!/bin/bash

# Signalling
trap 'exit 0' SIGTERM
trap '' SIGHUP

# grep
arg1="NETDEV WATCHDOG: eth0: transmit timed out"
#arg2="dhcpd: receive_packet failed on eth0: Network is down"

# Syslog
tag=$0
prio="kern.err"

fifo=/var/log/check_nic

IFCONFIG=$(which ifconfig)
SLEEP=$(which sleep)
EGREP=$(which egrep)
DATE=$(which date)
HOSTNAME=$(which hostname)
LOGGER=$(which logger)

while true
do
        read line < $fifo
        echo $line | \
          $EGREP -e "$arg1" \
          > /dev/null 2>&1
        if [ $? -eq 0 ] ; then
                $IFCONFIG eth0 down
                $SLEEP 1
                $IFCONFIG eth0 up
                $LOGGER -t $tag -p $prio "eth0 down, restarting..." 
                $SLEEP 5
        fi
done