Herbert Poetzl wrote: > On Thu, Dec 14, 2006 at 09:36:31AM -0600, Serge E. Hallyn wrote: > > [lot of stuff zapped here] > >> Quoting Cedric Le Goater (clg at fr.ibm.com): >>> how do you enter only a subset of namespaces of a nsproxy/container >>> and not all of it ? >> one container corresponds to one nsproxy which is one set of >> namespaces. Are you asking how you would only switch your pid >> namespace but keep your network namespace as the original? >> >> I'm not sure that's something we want/need to support. > > I can, we use this for several purposes, one is to > extend or modify the VFS namespace by mounting or > unmounting filesystems _into_ the guest > >> (Can you cite a use case?) > >> But since I haven't specified how to ask for the nsproxy switch >> anyway, it's too early to ask how we add a flag to specify a >> namespace subset :) > > we (Linux-VServer) extended the set/enter_namespace > command to take a flag mask similar to unshare > (only 64bit wide :) to allow for 'selecting' the > proper spaces > > note: the nsproxy a context is referring to can also > be changed this way, and a priviledged process can > enter each guest space separately ... I'll work on a clone64 and unshare64 in January. But did you take a look at the bind_ns syscall to see how it fits with vserver ? C.