Conversation
When m.Type starts with either fuse. or fuse3, the mount helper binary mount.fuse or mount.fuse3 is executed. Similar to the changes made in containerd: > This is expected to be used by fuse-overlayfs plugin: https://github.com/AkihiroSuda/containerd-fuse-overlayfs > > Motivation > > The purpose of the containerd fuse-overlayfs snapshotter plugin is to provide > OverlayFS functionality for rootless mode without depending on the Ubuntu/Debian > kernel patch. > Although fuse-overlayfs provides shiftfs functionality and supports CRFS plugin, > these functionalities are not planned to be supported by the containerd fuse-overlayfs Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
|
Opening as draft for now, depending on how much sense it will make to replace the mounting part that's in containerd with moby/sys as well. Containerd also has some mount/unmount code for Windows, but that may be very container specify, so TBD if we want to keep this repository as "generic" or "container specific" |
| } | ||
|
|
||
| // fuseSuperMagic is defined in statfs(2) | ||
| const fuseSuperMagic = 0x65735546 |
There was a problem hiding this comment.
This needs to be added to golang.org/x/sys/unix...
| return nil | ||
| } | ||
| } | ||
| return err |
There was a problem hiding this comment.
Does fusermount return an error if target is not mounted? If yes, we should filter it out somehow.
| } | ||
|
|
||
| func unmount(target string, flags int) error { | ||
| if isFUSE(target) { |
There was a problem hiding this comment.
Maybe move it to after a normal mount (need to be checked carefully how it would work).
There was a problem hiding this comment.
Otherwise we're making unmount do one extra syscall every time.
|
The mounter implementation seems not committed |
|
Ah, yes, I think I started looking at differences between containerd and this repo, and was not finished yet; thought I'd open to at least discuss if this would be in or out of scope for this repo |
I think as long as it's not introducing any noticeable overhead to "normal" mounts it's fine to have it. The question is what do we call "noticeable" :) |
|
Any chance of this moving forward? If I understand correctly this would allow mounting of FUSE filesystems like glusterfs using the docker's |
4 participants
When m.Type starts with either fuse. or fuse3, the mount helper binary mount.fuse
or mount.fuse3 is executed.
Similar to the changes made in containerd: