FUSE: report real filesystem stats in Statfs to fix df/kubelet 0-capa…

[wswsmao]

The FUSE layer previously returned zeroed Statfs, causing df to show 0 and kubelet to warn InvalidDiskCapacity on the image filesystem.

0s (x255 over 45h)      Warning   InvalidDiskCapacity   Node/10.255.64.239                      invalid capacity 0 on image filesystem
0s (x256 over 45h)      Warning   InvalidDiskCapacity   Node/10.255.64.239                      invalid capacity 0 on image filesystem
0s (x257 over 45h)      Warning   InvalidDiskCapacity   Node/10.255.64.239                      invalid capacity 0 on image filesystem

we can get this log:

df -h /var/lib/containerd-stargz-grpc/snapshotter/snapshots/113/fs
Filesystem      Size  Used Avail Use% Mounted on
stargz             0     0     0    - /var/lib/containerd-stargz-grpc/snapshotter/snapshots/113/fs

This change plumbs a base path and fills Statfs from unix.Statfs, with a safe fallback to zeros. A minimal test ensures zero when base is empty and non-zero when provided.

After this change:

df -h /var/lib/containerd-stargz-grpc/snapshotter/snapshots/1/fs
Filesystem      Size  Used Avail Use% Mounted on
stargz           99G   61G   34G  65% /var/lib/containerd-stargz-grpc/snapshotter/snapshots/1/fs

df -lh
...
stargz           99G   61G   34G  65% /var/lib/containerd-stargz-grpc/snapshotter/snapshots/1/fs
stargz           99G   61G   34G  65% /var/lib/containerd-stargz-grpc/snapshotter/snapshots/2/fs

[wswsmao]

@ktock

[ktock]

Could you separate the commit(or PR) if it contains both of the statfs fix and the config reaload?

[wswsmao]

done

[ktock]

invalid capacity 0 on image filesystem

I thought it was related to the wrong proxy_plugins.stargz.exports.root configuration for the CRI plugin (related PR: #1893, related issue: #1349 (comment) ). Can't this be solve just by setting proxy_plugins.stargz.exports.root to "/var/lib/containerd-stargz-grpc/" in containerd TOML file, or is it still an issue even with that configuration?