namespaces - overview of Linux namespaces

DESCRIPTION         top


       A namespace wraps a global system resource in an abstraction that
       makes it appear to the processes within the namespace that they have
       their own isolated instance of the global resource.  Changes to the
       global resource are visible to other processes that are members of
       the namespace, but are invisible to other processes.  One use of
       namespaces is to implement containers.


       Linux provides the following namespaces:


       Namespace   Constant          Isolates
       Cgroup      CLONE_NEWCGROUP   Cgroup root directory
       IPC         CLONE_NEWIPC      System V IPC, POSIX message queues
       Network     CLONE_NEWNET      Network devices, stacks, ports, etc.
       Mount       CLONE_NEWNS       Mount points
       PID         CLONE_NEWPID      Process IDs
       User        CLONE_NEWUSER     User and group IDs
       UTS         CLONE_NEWUTS      Hostname and NIS domain name


       This page describes the various namespaces and the associated /proc
       files, and summarizes the APIs for working with namespaces.


   The namespaces API
       As well as various /proc files described below, the namespaces API
       includes the following system calls:


       clone(2)
              The clone(2) system call creates a new process.  If the flags
              argument of the call specifies one or more of the CLONE_NEW*
              flags listed below, then new namespaces are created for each
              flag, and the child process is made a member of those
              namespaces.  (This system call also implements a number of
              features unrelated to namespaces.)


       setns(2)
              The setns(2) system call allows the calling process to join an
              existing namespace.  The namespace to join is specified via a
              file descriptor that refers to one of the /proc/[pid]/ns files
              described below.


       unshare(2)
              The unshare(2) system call moves the calling process to a new
              namespace.  If the flags argument of the call specifies one or
              more of the CLONE_NEW* flags listed below, then new namespaces
              are created for each flag, and the calling process is made a
              member of those namespaces.  (This system call also implements
              a number of features unrelated to namespaces.)


       Creation of new namespaces using clone(2) and unshare(2) in most
       cases requires the CAP_SYS_ADMIN capability.  User namespaces are the
       exception: since Linux 3.8, no privilege is required to create a user
       namespace.


   The /proc/[pid]/ns/ directory
       Each process has a /proc/[pid]/ns/ subdirectory containing one entry
       for each namespace that supports being manipulated by setns(2):


           $ ls -l /proc/$$/ns
           total 0
           lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 cgroup -> cgroup:[4026531835]
           lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 ipc -> ipc:[4026531839]
           lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 mnt -> mnt:[4026531840]
           lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 net -> net:[4026531969]
           lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid -> pid:[4026531836]
           lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 user -> user:[4026531837]
           lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 uts -> uts:[4026531838]


       Bind mounting (see mount(2)) one of the files in this directory to
       somewhere else in the filesystem keeps the corresponding namespace of
       the process specified by pid alive even if all processes currently in
       the namespace terminate.


       Opening one of the files in this directory (or a file that is bind
       mounted to one of these files) returns a file handle for the
       corresponding namespace of the process specified by pid.  As long as
       this file descriptor remains open, the namespace will remain alive,
       even if all processes in the namespace terminate.  The file
       descriptor can be passed to setns(2).


       In Linux 3.7 and earlier, these files were visible as hard links.
       Since Linux 3.8, they appear as symbolic links.  If two processes are
       in the same namespace, then the inode numbers of their
       /proc/[pid]/ns/xxx symbolic links will be the same; an application
       can check this using the stat.st_ino field returned by stat(2).  The
       content of this symbolic link is a string containing the namespace
       type and inode number as in the following example:


           $ readlink /proc/$$/ns/uts
           uts:[4026531838]


       The files in this subdirectory are as follows:


       /proc/[pid]/ns/cgroup (since Linux 4.6)
              This file is a handle for the cgroup namespace of the process.


       /proc/[pid]/ns/ipc (since Linux 3.0)
              This file is a handle for the IPC namespace of the process.


       /proc/[pid]/ns/mnt (since Linux 3.8)
              This file is a handle for the mount namespace of the process.


       /proc/[pid]/ns/net (since Linux 3.0)
              This file is a handle for the network namespace of the
              process.


       /proc/[pid]/ns/pid (since Linux 3.8)
              This file is a handle for the PID namespace of the process.


       /proc/[pid]/ns/user (since Linux 3.8)
              This file is a handle for the user namespace of the process.


       /proc/[pid]/ns/uts (since Linux 3.0)
              This file is a handle for the UTS namespace of the process.


   Cgroup namespaces (CLONE_NEWCGROUP)
       See cgroup_namespaces(7).


   IPC namespaces (CLONE_NEWIPC)
       IPC namespaces isolate certain IPC resources, namely, System V IPC
       objects (see svipc(7)) and (since Linux 2.6.30) POSIX message queues
       (see mq_overview(7)).  The common characteristic of these IPC
       mechanisms is that IPC objects are identified by mechanisms other
       than filesystem pathnames.


       Each IPC namespace has its own set of System V IPC identifiers and
       its own POSIX message queue filesystem.  Objects created in an IPC
       namespace are visible to all other processes that are members of that
       namespace, but are not visible to processes in other IPC namespaces.


       The following /proc interfaces are distinct in each IPC namespace:


       *  The POSIX message queue interfaces in /proc/sys/fs/mqueue.


       *  The System V IPC interfaces in /proc/sys/kernel, namely: msgmax,
          msgmnb, msgmni, sem, shmall, shmmax, shmmni, and shm_rmid_forced.


       *  The System V IPC interfaces in /proc/sysvipc.


       When an IPC namespace is destroyed (i.e., when the last process that
       is a member of the namespace terminates), all IPC objects in the
       namespace are automatically destroyed.


       Use of IPC namespaces requires a kernel that is configured with the
       CONFIG_IPC_NS option.


   Network namespaces (CLONE_NEWNET)
       Network namespaces provide isolation of the system resources
       associated with networking: network devices, IPv4 and IPv6 protocol
       stacks, IP routing tables, firewalls, the /proc/net directory, the
       /sys/class/net directory, port numbers (sockets), and so on.  A
       physical network device can live in exactly one network namespace.  A
       virtual network device ("veth") pair provides a pipe-like abstraction
       that can be used to create tunnels between network namespaces, and
       can be used to create a bridge to a physical network device in
       another namespace.


       When a network namespace is freed (i.e., when the last process in the
       namespace terminates), its physical network devices are moved back to
       the initial network namespace (not to the parent of the process).


       Use of network namespaces requires a kernel that is configured with
       the CONFIG_NET_NS option.


   Mount namespaces (CLONE_NEWNS)
       Mount namespaces isolate the set of filesystem mount points, meaning
       that processes in different mount namespaces can have different views
       of the filesystem hierarchy.  The set of mounts in a mount namespace
       is modified using mount(2) and umount(2).


       The /proc/[pid]/mounts file (present since Linux 2.4.19) lists all
       the filesystems currently mounted in the process's mount namespace.
       The format of this file is documented in fstab(5).  Since kernel
       version 2.6.15, this file is pollable: after opening the file for
       reading, a change in this file (i.e., a filesystem mount or unmount)
       causes select(2) to mark the file descriptor as readable, and poll(2)
       and epoll_wait(2) mark the file as having an error condition.


       The /proc/[pid]/mountstats file (present since Linux 2.6.17) exports
       information (statistics, configuration information) about the mount
       points in the process's mount namespace.  This file is readable only
       by the owner of the process.  Lines in this file have the form:


            device /dev/sda7 mounted on /home with fstype ext3 [statistics]
            (       1      )            ( 2 )             (3 ) (4)


              The fields in each line are:


              (1)  The name of the mounted device (or "nodevice" if there is
                   no corresponding device).


              (2)  The mount point within the filesystem tree.


              (3)  The filesystem type.


              (4)  Optional statistics and configuration information.
                   Currently (as at Linux 2.6.26), only NFS filesystems
                   export information via this field.


   PID namespaces (CLONE_NEWPID)
       See pid_namespaces(7).


   User namespaces (CLONE_NEWUSER)
       See user_namespaces(7).


   UTS namespaces (CLONE_NEWUTS)
       UTS namespaces provide isolation of two system identifiers: the
       hostname and the NIS domain name.  These identifiers are set using
       sethostname(2) and setdomainname(2), and can be retrieved using
       uname(2), gethostname(2), and getdomainname(2).


       Use of UTS namespaces requires a kernel that is configured with the
       CONFIG_UTS_NS option.


CONFORMING TO         top


       Namespaces are a Linux-specific feature.


EXAMPLE         top


       See user_namespaces(7).


SEE ALSO         top


       lsns(1), nsenter(1), readlink(1), unshare(1), clone(2), setns(2),
       unshare(2), proc(5), capabilities(7), cgroup_namespaces(7),
       cgroups(7), credentials(7), pid_namespaces(7), user_namespaces(7),
       switch_root(8)