no internal mount

[bAndie91]

I'm just wondering is it planned to add the feature that ofs wont mount its underlaying filesystem by itself, instead,

• there would be an already mounted fs on eg. /mnt/sshfs (not mandatory a real mountpoint) and
• ofs would be mounted on eg. /mnt/sshfs.ofs and
• ofs would operate against /mnt/sshfs as it does now against fstype://server/share.

[darkdragon-001]

You can try if the following works in this case

sudo mount -t ofs file://mnt/sshfs /mnt/sshfs.ofs

It might lead to problems though when /mnt/sshfs get's unmounted... I don't know a good way to tackle these problems since ofs does not have information about the mount because it is handled by another entity.

[bAndie91]

cool, I missed it handles the file scheme :)

what about this logic?
maybe it'd check underlaying folder's mountpoint, like:

• case 1: source is not a mountpoint (same st_dev as parent directory's inode)
  • mount.ofs file://mnt/sshfs/subfolder /mnt/ofs
  • user is accessing /mnt/ofs/file
  • check source root folder (/mnt/sshfs/subfolder) existance
    • if not exists, assume fs got unmounted, go to offline mode
    • if exists, continue with normal operation
• case 2: source is a mountpoint
  • mount.ofs file://mnt/sshfs /mnt/ofs
  • user is accessing /mnt/ofs/file
  • check whether source root folder (/mnt/sshfs) is in mtab (or /proc/mount)
    • if not, assume fs got unmounted, go to offline mode
    • if yes, continue with normal operation

[darkdragon-001]

Case 1: How to differentiate a deleted folder?

Case 2: It might be possible to add an additional virtual scheme (e. g. mirror or follow), which assumes a mount in the source folder and otherwise go to offline mode. Feel free to implement this feature and send a PR 😉

[bAndie91]

case 1: you're right. better check by looking up the source folder's (/mnt/sshfs/subfolder) mountpoint root (/mnt/sshfs) and perform the is_mountpoint query on that.
in this way the 2 cases are not so different.
have to walk up to the mountpoint root, then lookup in mtab (or similar) – so many lookup and lstat calls, maybe worth to put some caching mechanism into the logic.

ye, I'm happy to implement, but hardly have the chance :)
anyway, let's keep this on wishlist