lvm partition on lv

Let's say on /dev/vg1/test I had one LVM partition and one ext3 partition. How can I access those separate partitions since it's only one device (/dev/vg1/test)? Normally, a partitioned block device (/dev/hda) would show up like /dev/hda1, dev/hda2, etc.

Quoted 4 lines · Randall Smith

Let's say on /dev/vg1/test I had one LVM partition and one ext3 partition. How can I access those separate partitions since it's only one device (/dev/vg1/test)? Normally, a partitioned block device (/dev/hda) would show up like /dev/hda1, dev/hda2, etc.

With loop-device, you can use only part of the block-device as new block-device (See options -o and -s in man:losetup)

So if you know the positions where the partitions start, and their lengths (both in bytes), you could do something like this: (I havn't tested this, so there might be some errors in these commands. But the basic idea should work)

# make LVM partition available at /dev/loop1
losetup -o $lvm_start -s $lvm_size /dev/loop1 /dev/vg1/test
mkdir /mnt/test
# Mount ext3-partition at /mnt/test using loop-device
mount -t ext3 -o offset=$ext3_start,sizelimit=$ext3_length
/dev/vg1/test /mnt/test

Quoted 14 lines · Markus Laire

Quoted 4 lines · Randall Smith

Let's say on /dev/vg1/test I had one LVM partition and one ext3 partition. How can I access those separate partitions since it's only one device (/dev/vg1/test)? Normally, a partitioned block device (/dev/hda) would show up like /dev/hda1, dev/hda2, etc.

With loop-device, you can use only part of the block-device as new block-device (See options -o and -s in man:losetup)

So if you know the positions where the partitions start, and their lengths (both in bytes), you could do something like this: (I havn't tested this, so there might be some errors in these commands. But the basic idea should work)

Quoted 6 lines

# make LVM partition available at /dev/loop1
losetup -o $lvm_start -s $lvm_size /dev/loop1 /dev/vg1/test
mkdir /mnt/test
# Mount ext3-partition at /mnt/test using loop-device
mount -t ext3 -o offset=$ext3_start,sizelimit=$ext3_length
/dev/vg1/test /mnt/test

Well, last command is missing at least "loop" option - i.e. it should be

mount -t ext3 -o loop,offset=$ext3_start,sizelimit=$ext3_length /dev/vg1/test /mnt/test

or you could use these instead
losetup -o $ext3_start -s $ext3_length /dev/loop2 /dev/vg1/test mount -t ext3 /dev/loop2 /mnt/test

I hope I didn't make other mistakes...

($lvm_size and $ext3_length both means the size of the partition in bytes - I should've used 'size' or 'length' for both for consistency)

Warning: This may be insane.

I like the flexibility of LVM and try to use it wherever it's feasible.
In this case, I'd like a Xen guest OS to have control over it's LV's.
The method I usually see for using LVM with Xen is to create an LV and
a filesystem on it. I would like to instead create an LV and partition it with an LVM partition and maybe other partition types. Just to test this I did the following:

1. lvcreate -L 100M -name test vg1
2. cfdisk /dev/vg1/test
3. create LVM partition on entire device
4. pvcreate /dev/vg1/test
5. vgcreate vg2 /dev/vg1/test
6. vgchange -ay /dev/vg2
7. lvcreate -L 50M -n testlv vg2
8. mkfs.ext3 /dev/vg2/testlv
9. mkdir /mnt/test
10. mount /dev/vg2/testlv /mnt/test

And it worked! Cool!

Let's say on /dev/vg1/test I had one LVM partition and one ext3 partition. How can I access those separate partitions since it's only one device (/dev/vg1/test)? Normally, a partitioned block device (/dev/hda) would show up like /dev/hda1, dev/hda2, etc.

In the example above, I'm partitioning the LV and using the partition on the same system, which is useless. What I will be doing is giving the disk image /dev/vg1/test to a Xen guest so it can have it's own VG and LVs. Are there potential problems I should look out for and/or tweaking I should do to make this work optimally?