On Tue, Apr 26, 2016 at 12:29:25PM +0200, Karel Zak wrote:

On Mon, Apr 25, 2016 at 09:42:07PM -0500, Serge E. Hallyn wrote:

quoted

quoted

Looking at it another way...  The value we're talking about shows us
the path of the root dentry of a cgroup mount.  If a task in cgns2
rooted at /a/b/c mounts a cgroupfs, it will see '/' as the root dentry.
If a task in cgns1 rooted at /a/b looks at that mountinfo, '/' would
be misleading.  It really should be '/c'.

So I think that for cgroup mount entries in mountinfo to be useful (i.e.
to criu) we either need the root dentry path to be given as relative to
the reader's cgroup namespace (as I have it in this patchset), or else
we need to add another piece of information in the mountinfo entry, such
as the nsfd inode number of the cgroup namespace in which it was
mounted.

In the ideal world there is no mountinfo file, but /proc/self/mountinfo/<id>/
directory with individual files, so every subsystem and filesystem has 
absolute freedom to store there all relevant information. The result will 
be also lucky kernel that does not have to always generate entire huge 
mountinfo file for all mountpoins... etc. :-)

Yeah mountinfo does seem like a big stick to swing around every time I want
one little piece of information.  Also mght be good to have per-fstype
directories so we can just look under /proc/self/mountsdir/cgroupfs/ for
only cgroupfs <id>s.

There we might also find open fds for source and mountdir, kinda fitting
in with previous discussions of separating bdev_open() and mountat().

BTW, assuming this would in fact report source and mountpoint location
with fds, these would really (through realpath) be reported relative to
the reader's namespace, as I'm doing and advocating here.

-serge