Suppose you have an LVM volume /dev/vg0/mylv. You have presented this as a virtual disk to a virtualised or emulated guest system. During installation the guest system sees it as /dev/sda and partitions it into /dev/sda{1,2,5,6} and completes the installation. Now at some point you need to access those filesystems from within the host system, without running the guest system. fdisk sees these partitions just fine:
# fdisk -l /dev/vg0/mylv
Device Boot Start End Blocks Id System
/dev/vg0/mylv1 2048 684031 340992 83 Linux
/dev/vg0/mylv2 686078 20969471 10141697 5 Extended
/dev/vg0/mylv5 686080 8290303 3802112 83 Linux
/dev/vg0/mylv6 8292352 11980799 1844224 83 Linux
However, the devices such as /dev/vg0/mylv1 do not actually exist. I guess that because they're within an LV, the OS does not recognise this type of nesting by default. Is there any way I can prod Linux so that /dev/vg0/mylv1 or equivalent appears and thus becomes mountable within the host system?
I understand that it's possible with qemu-nbd, and will use this if necessary. However, I was hoping for something more direct if possible, rather than simulating a network block device and attaching that.
Answer
I believe the tool you're looking for is kpartx
.
The general procedure is:
List partitions in the disk image:
kpartx -l /dev/vg0/mylv
Add the partitions to device-mapper:
kpartx -a /dev/vg0/mylv
Mount the partition you're interested in:
mount -o ro /dev/mapper/loop0p5 /mnt
No comments:
Post a Comment