determing lvm support

is there any fool proof way of knowing whether or not a system has lvm support?

right now i determine it this way:

lvm1:
look for "vgcreate" command and
look for "lvm" in /proc/devices or
look for "lvm-mod" in /lib/modules/[release]/kernel/drivers/md/*

lvm2:
look for "vgcreate" command and
look for "device-mapper" in /proc/devices or
look for "dm-mod" in /lib/modules/[release]/kernel/drivers/md/*

i think i am fairly certain with lvm1, but with lvm2 i am not sure if i have device mapper support and a lvm command that is enough.

any ideas?

rich turner

Better to check for the new 'lvm' base command: other commands are links to it: check that running 'lvm version' produces complete output. That will also confirm that device-mapper is available in the live kernel and that /dev/mapper/control is set up correctly. (See device-mapper/scripts/devmap_mknod.sh)

It's better to check for that in /proc/misc, which will tell you whether or not device-mapper support is present in the running kernel.

Better to attempt to load the module with modprobe and then do the tests above to check it worked.

Alasdair

[email protected]

Besides that is there any way to reliably check if dm is loaded from a c program, besides calling dm_task_run(), which is noisy if the driver is not loaded? I don't want the user to see an error if she's not running device-mapper.
it the moment i try to open dm_dir()/control, is that ok?

L.

Luca Berra -- [email protected]
       Communication Media & Services S.r.l.
/"\
\ /     ASCII RIBBON CAMPAIGN
 X        AGAINST HTML MAIL
/ \

The only problem with this step is that many kernels (any production-level server of mine at least) do not use loadable modules, but instead build LVM support striaght into the kernel. In these cases your tests for the modules will fail.

Look at using dm_log_init() to provide a function that suppresses all output.

That should work OK bypassing libdevmapper.
Otherwise, try dm_task_get_driver_version() after disabling log output.

Alasdair

[email protected]

I'll look at it, but it seems i cannot restore the previous log function then, i think it's easier to bypass the lib for my particular case. Thanks a lot for your answers.

Regards,
L.

Luca Berra -- [email protected]
       Communication Media & Services S.r.l.
/"\
\ /     ASCII RIBBON CAMPAIGN
 X        AGAINST HTML MAIL
/ \

I'll fix it so people can.

Alasdair

[email protected]

if i have 'device-mapper' and 'lvm version' produces complete output, is there anything else to check if lvm is supported on the system?

is 'device-mapper' used for anything else besides lvm?

does the kernel need to be built with 'lvm' support (which i believe was true with lvm1-created 'lvm-mod'), or does building a kernel for 'lvm' support just create 'dm-mod' loadable module/builtin 'dm-mod' kernel support?

why is it better to check in /proc/misc? when device-mapper support is loaded, it shows up in both /proc/misc as well as /proc/devices.

You can only guarantee its appearance in /proc/devices while you have a device active.

Alasdair

[email protected]