Extend CDROM provider to support any common type of disk/partition with ISO format

[errordeveloper]

No description provided.

[errordeveloper]

I've tested it with linuxkit run -disk file=metadata-test.iso and it worked. If anyone can suggests other tests, I'm happy to look into it.

[rn]

The diff in GH and git does not look right. Can you do a git mv provider_cdrom.go provider_iso.go and then make the changes. That way one actually sees the differences. Or even split this up into two commits. One which does the rename and one which adds the new functionality? That way one can see the actual changes.

It would also be good if the commit message explains more of the rationale for this change.

[ijc]

git mv is no different to git rm && git add in terms of what is stored in the repo, there is no specific metadata recorded about such things. The "move" and "copy" display is entirely a presentation layer thing. IOW git mv wouldn't help here -- it's just that the old and new files differ enough that GH didn't guess there was a move involved while rendering the diff.

For the CLI you can control the heuristics a bit with the -M and -C options to git show, I suppose that isn't possible in the GH webui though.

Making it two commits would indeed be one way to avoid the problem though. and +1 to better commit messages...

[errordeveloper]

Sounds good, will do! I have indeed use 'git mv'.

…

On Wed, 30 Jan 2019, 10:45 am Ian Campbell ***@***.*** wrote: git mv is no different to git rm && git add in terms of what is stored in the repo, there is no specific metadata recorded about such things. The "move" and "copy" display is entirely a presentation layer thing. IOW git mv wouldn't help here -- it's just that the old and new files differ enough that GH didn't guess there was a move involved while rendering the diff. For the CLI you can control the heuristics a bit with the -M and -C options to git show, I suppose that isn't possible in the GH webui though. Making it two commits would indeed be one way to avoid the problem though. and +1 to better commit messages... — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#3280 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAPWS6H_48NUd_lBorYFRVLZVLQA9oHTks5vIXe3gaJpZM4aZMe-> .

[GordonTheTurtle]

Please sign your commits following these rules:
https://github.com/moby/moby/blob/master/CONTRIBUTING.md#sign-your-work
The easiest way to do this is to amend the last commit:

$ git clone -b "metadata-iso-disk" git@github.com:errordeveloper/linuxkit.git somewhere
$ cd somewhere
$ git rebase -i HEAD~842353890016
editor opens
change each 'pick' to 'edit'
save the file and quit
$ git commit --amend -s --no-edit
$ git rebase --continue # and repeat the amend for each commit
$ git push -f

Amending updates the existing PR. You DO NOT need to open a new one.

[errordeveloper]

I've split the commits as advised.

[rn]

Thanks for the update. the diff is not much easier to read.

I'm not sure I like this approach. This seems to mount/unmount all disk type devices instead of stopping when one suitable device is found.

[errordeveloper]

the diff is not much easier to read.

I thought 5c3bde1 fairly trivial...

I'm not sure I like this approach. This seems to mount/unmount all disk type devices instead of stopping when one suitable device is found.

That's just what it did before, only limited to all /dev/sr* devices. If we believe that behaviour should change, I'm more than happy to implement that in this PR, or perhaps separate PR would make sense.
Perhaps the current behaviour was somewhat satisfactory, considering that systems with many optical drives aren't very common. I agree that trying to mount each and every drive is not a good idea, perhaps pre-filtering on filesystem type would be a good idea? We could even match on a well-known ISO volume ID, e.g. something like 'LinuxKit metadata'.

[rn]

Sorry, I meant that the diff is now much easier to read.

But the code is mounting every disk in the system. And only later, during probe, it goes through the providers and stops on the first suitable on. This doesn't look nice. Why not mount only during probe and stop mounting once a suitable device is found. I dn't think we want to make it more complicated with filesystem probing in whatnot, just stop mounting once you found a suitable device.

[errordeveloper]

Sure, that makes sense. Happy to adjust the behaviour that way.

[errordeveloper]

@rn please take a look at the latest commit.

[errordeveloper]

There is a challenge with USB devices, I ended up having to do this:

onboot:
  - name: modprobe-usb-storage
    image: linuxkit/modprobe:v0.6
    command: ["modprobe", "usb-storage"]
  - name: wait-usb-storage
    image: linuxkit/alpine:3683c9a66cd4da40bd7d6c7da599b2dcd738b559
    binds:
      - /dev:/dev
    command:
      - sh
      - -c
      - until test -b /dev/sdc ; do sleep 1 ; done
  - name: metadata
    image: errordeveloper/metadata:9250ccfc8a4ba879989281b5f782bae2551521b9

I might find a better way, but this is what I have to do at the moment.