Files
linuxkit/projects/shiftfs
Ian Campbell 5833d1b6bc init: replace ctr with a custom client using the containerd client library
Currently it supports only `service start <SERVICE>`, but it could grow e.g.
`stop`, `exec` etc in the future (although you can still use `ctr` for those).

In order to be able to use go-compile.sh the containerd build needs to move
from /root/go to /go as the GOPATH.

The vendoring situation is not ideal, but since this tool wants to be an exact
match for the containerd it seems tollerable to reuse its vendoring.

Signed-off-by: Ian Campbell <ian.campbell@docker.com>
2017-06-16 11:48:53 +01:00
..
2017-06-13 11:08:29 -06:00
2017-06-13 11:08:29 -06:00
2017-06-13 14:05:42 -06:00

shiftfs

Shiftfs is a virtual filesystem for mapping mountpoints across user namespaces. The idea is that it would be useful for dockerds spawning containers: they can keep filesystems on the host disk in terms of real root, but mount the container roots via shiftfs, allowing containers to share a particular filesystem with different uid maps, while not having to uidshift every file on disk (and thus destroying some of the sharing properties).

The version included here is the v2 version of shiftfs, using the superblock's user namespace instead of mountopts to figure out mappings. Thus, an extra step of "marking" mounts is needed. For example:

# mkdir source
# touch source/foo  # a root owned file
# mount -t shiftfs -o mark source source
# chmod 777 source

Now, let's make a user namespace:

# setuid 1000 unshare -rm
# cat /proc/self/uidmap
         0       1000          1
# mkdir dest
# mount -t shiftfs source dest
# stat dest/foo | grep Uid
Access: (0644/-rw-r--r--)  Uid: (    0/    root)   Gid: (    0/    root)

And thanks to the magic of shiftfs, the file is root owned in the user namespace.