Allow setting read/write, UUID and volume name options for EXT4 file systems

[pojntfx]

Currently, file systems generated by the tar2ext4 tool lack a volume name and have their UUID set to 0000-...., despite the superblock in the internal package having the fields available. This PR allows to optionally set these fields with ReadWrite() Option, UUID(uuid [16]byte) Option and VolumeName(volumeName [16]byte) Option. The existing behavior does not change as a result of this.

[pojntfx]

@microsoft-github-policy-service agree company="Loophole Labs"

[katiewasnothere]

Hi @pojntfx, thanks for the contributions!

Could you provide a use case for the disk to be read/write? This tool was originally created for converting Linux container image tar files into VHDs that can be used on windows to run hypervisor isolated Linux containers. In this context, the VHDs shouldn't be read/write (though I have heard of some corner cases 😅).

[pojntfx]

Hi @katiewasnothere, sorry about the long delay. We have almost the exact same use case - we take an OCI image, extract it with umoci into an OCI runtime bundle, put it onto an EXT4 file system with hcsshim, mount it in a VM & start it with runc. In our case, we'd like those file systems to be read-writable. We also need to have a proper UUID & volume name so that we can mount it at an expected path, e.g. with a fstab like this:

LABEL=@OCI_RUNTIME_BUNDLE_MOUNT_LABEL@    /@OCI_RUNTIME_BUNDLE_MOUNT_DIR@    ext4    defaults    0    2

Same goes for #2325 - once the disk is read-writable and has a proper UUID & volume name, we'd like to be able to add extra free space for it.

A workaround for this issue ofc could be to use something like overlayfs, but in our case that's not really an option for performance reasons.