FR-2622 customize /etc/fstab

[jawn-smith]

This PR adds a new fstab section to the customization section of the image definition, and populates the contents of /etc/fstab in the chroot based on the image definition.

[codecov]

Codecov Report

Merging #75 (c3832c3) into feature/classic-image-redesign (a38a5dc) will not change coverage.
The diff coverage is 100.00%.

@@                       Coverage Diff                        @@
##           feature/classic-image-redesign       #75   +/-   ##
================================================================
  Coverage                          100.00%   100.00%           
================================================================
  Files                                   9         9           
  Lines                                1526      1555   +29     
================================================================
+ Hits                                 1526      1555   +29     

Impacted Files	Coverage Δ
internal/statemachine/image_definition.go	100.00% <ø> (ø)
internal/statemachine/classic_states.go	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[sil2100]

This is good! Small inline formatting suggestion, but besides that it's all good. I'm thinking a bit about the image-definition format a bit, but I think this all 'fits' so far.

[sil2100]

One thing to think about: should we maybe provide some fstab default if no fstab is given? Is there a common fstab we can figure out from the gadget yaml and use that as a default?

[jawn-smith]

One thing to think about: should we maybe provide some fstab default if no fstab is given? Is there a common fstab we can figure out from the gadget yaml and use that as a default?

We have some basic logic in populate_rootfs_contents to handle this. It creates an entry of LABEL=writable / ext4 defaults 0 0 by default. Do you think this is sufficient?

It is very non-trivial to add a default for a struct like this. It is currently a pointer value and is therefore not memory safe to add it as a default in the same way we are assigning default values for strings, booleans, etc.

[sil2100]

I think this is more than enough, thanks for the reminder! Let's merge.