Extract vm images #16

[pombredanne]

This PR:

• adds support to extract filesystems as files as found in VM images from VirtualBox, VMware and QEMU using libguestfs for Extract VM images #16
• provides a new way to configure third-party binaries (using environment variables, plugins or PATH)

[pombredanne]

All clear. Now merging for release.

[rwmjones]

A few comments about the patch in no particular order:

Instead of using guestfish it may or may not be worth considering using the API directly. We have a solid Python binding that covers the entire API. Example: https://libguestfs.org/guestfs-python.3.html#example-2:-inspect-a-virtual-machine-disk-image

For extracting things like partitions from disk images (ie. without processing the filesystem or individual files), we have the lighter weight tooling of libnbd / nbdkit (https://gitlab.com/nbdkit) / qemu-nbd. For example to copy out a partition from a raw disk image (to stdout), you might use:

nbdcopy -- [ nbdkit --filter=partition file disk.img partition=1 ] -

https://libguestfs.org/nbdkit-partition-filter.1.html#EXAMPLE
https://libguestfs.org/nbdcopy.1.html

To copy out a partition from a qcow2 file (again to stdout):

nbdcopy -- [ \
    nbdkit --filter=partition nbd \
        command=qemu-nbd arg=-f arg=qcow2 arg=/path/to/image.qcow2 \
        partition=1 ] -

https://libguestfs.org/nbdkit-nbd-plugin.1.html#Add-nbdkit-partition-filter-to-qemu-nbd
https://libguestfs.org/nbdcopy.1.html

[pombredanne]

@rwmjones thank you ++ for chiming in!

You wrote:

We have a solid Python binding that covers the entire API. Example: https://libguestfs.org/guestfs-python.3.html#example-2:-inspect-a-virtual-machine-disk-image

I considered it at first but then the uneven packaging and support for various Python versions across Linux distros made switch back to a CLI interface for now.

For extracting things like partitions from disk images (ie. without processing the filesystem or individual files), we have the lighter weight tooling of libnbd / nbdkit (https://gitlab.com/nbdkit) / qemu-nbd.

I looked at it in details too. The partitions are not enough and I would need to mound things afterwards otherwise.
Yet I really want the filesystems out so I can process a whole VM image analysis pipeline from a virtual disk in https://github.com/nexB/scancode.io/blob/main/scanpipe/pipelines/root_filesystems.py the same way we can already process whole Docker container images in https://github.com/nexB/scancode.io/blob/main/scanpipe/pipelines/docker.py

On a side note, I have been hit by https://bugs.launchpad.net/ubuntu/+source/libguestfs/+bug/1813662 and I posted there https://bugs.launchpad.net/ubuntu/+source/libguestfs/+bug/1813662/comments/25 ;)