Working with Linux Driver Modules
You just replaced a failed network card, rebooted your server and now network wouldn’t start and your screen is filled with cryptic error messages about some kernel module doing something it’s not supposed to do. This usually happens on Fridays or a day before you leave on vacation. Below is a summary of some basic Linux commands that can help you to quickly troubleshoot many driver/module-related problems.
First things first: whenever you add new hardware, replace failed hardware, or upgrade firmware, almost always you must install the drivers first. When that new disk controller is in and the system is booting up, it usually helps to have the correct driver already installed. How do you know that a specific module is loaded? Linux hides this information in /proc/modules. Let’s say you are looking for NFS-related modules:
icebox:~ # egrep "^nfs|^lockd" /proc/modules nfsd 211612 13 - Live 0xfa716000 lockd 61748 1 nfsd, Live 0xf97e7000 nfs_acl 2848 1 nfsd, Live 0xf9418000
Another way of getting the same information is by using the “lsmod” command:
icebox:~ # lsmod | egrep "^nfs|^lockd" nfsd 211612 13 lockd 61748 1 nfsd nfs_acl 2848 1 nfsd
Use the “modinfo” command to get a little bit more detail about the module that’s causing your headache:
icebox:~ # modinfo nfsd filename: /lib/modules/2.6.27.7-9-pae/kernel/fs/nfsd/nfsd.ko license: GPL author: Olaf Kirch <okir@monad.swb.de> srcversion: E6DFAD8170CEA2285A8E64C depends: auth_rpcgss,sunrpc,lockd,exportfs,nfs_acl supported: yes vermagic: 2.6.27.7-9-pae SMP mod_unload modversions 586
The particularly useful feature of the “modinfo” command is the list of dependencies. Here is a simple script that will “modinfo” the module of your choice and then check to make sure all dependent modules are loaded:
for module in `modinfo nfsd | grep depends | awk '{print $2}' | sed 's/,/ /g'` do if [ `egrep -c "^${module} " /proc/modules` -gt 0 ] then egrep "^${module} " /proc/modules else echo "ERROR: dependednt module $module not found in /proc/modules" fi done
If you have a driver module that is interfering with another module, you can unload one of them using “rmmod” command. You can use the “insmod” command to insert a module into the system kernel.
Perhaps a better way of removing (and adding) modules is by using the “modprobe” command. A cool feature of this command is the “–dry-run” option, which allows you to test adding/removing modules without actually adding or removing them. This is great if you have a production server that you don’t want to mess up (just yet).
If you support SLES systems and have the misfortune of dealing with Novell’s tech support, there is one tidbit of information you may find useful. Staring with SLES 11 any modules not directly supported by Novell will not load unless the /etc/modprobe.d/unsupported-modules file has this entry: allow_unsupported_modules 1
Popularity: 1% [?]
Related posts:
