From: Nicolas Dichtel > Sent: 29 September 2023 10:46 > > Le 29/09/2023 à 11:25, Christian Brauner a écrit : > >> I fear that creating a new mount ns for each net ns will introduce more problems. > > > > Not sure if we're talking past each other but that is what's happening > > now. Each new ip netns exec invocation will allocate a _new_ mount > > namespace. In other words, if you have 300 ip netns exec commands > > running then there will be 300 individual mount namespaces active. > > > > What I tried to say is that ip netns exec could be changed to > > _optionally_ allocate a prepared mount namespace that is shared between > > ip netns exec commands. And yeah, that would need to be a new command > > line addition to ip netns exec. > > Ok, you talked about changing 'ip netns exec', not adding an option, thus I > thought that you suggested adding this unconditionally ;-) > > I was asking myself how to propagate mount points between the parent and 'ip > netns exec' (both way), but this may be another use case than Toke's use case. I had a different problem. I have a system with two network namespaces (to separate public and private network data) and some programs that really want to read the '/sys/class/net' nodes in both namespaces - so I'd like to have the 'namespace' copy mounted at the same time on a different mount point. But I'm not at all sure that is possible. (It is possible to pass an open directory fd through 'ip netns exec' but that is all a bit cludgy.) Clearing all the mounts also 'lost' the root of a chroot (if not an actual mount point) causing 'pwd -P' (etc) to generate the full path instead of the chroot-relative one. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
