man7/namespaces.7

   1 .\" Copyright (c) 2013, 2016, 2017 by Michael Kerrisk <mtk.manpages@gmail.com>
   2 .\" and Copyright (c) 2012 by Eric W. Biederman <ebiederm@xmission.com>
   3 .\"
   4 .\" %%%LICENSE_START(VERBATIM)
   5 .\" Permission is granted to make and distribute verbatim copies of this
   6 .\" manual provided the copyright notice and this permission notice are
   7 .\" preserved on all copies.
   8 .\"
   9 .\" Permission is granted to copy and distribute modified versions of this
  10 .\" manual under the conditions for verbatim copying, provided that the
  11 .\" entire resulting derived work is distributed under the terms of a
  12 .\" permission notice identical to this one.
  13 .\"
  14 .\" Since the Linux kernel and libraries are constantly changing, this
  15 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  16 .\" responsibility for errors or omissions, or for damages resulting from
  17 .\" the use of the information contained herein.  The author(s) may not
  18 .\" have taken the same level of care in the production of this manual,
  19 .\" which is licensed free of charge, as they might when working
  20 .\" professionally.
  21 .\"
  22 .\" Formatted or processed versions of this manual, if unaccompanied by
  23 .\" the source, must acknowledge the copyright and authors of this work.
  24 .\" %%%LICENSE_END
  25 .\"
  26 .\"
  27 .TH NAMESPACES 7 2019-03-06 "Linux" "Linux Programmer's Manual"
  28 .SH NAME
  29 namespaces \- overview of Linux namespaces
  30 .SH DESCRIPTION
  31 A namespace wraps a global system resource in an abstraction that
  32 makes it appear to the processes within the namespace that they
  33 have their own isolated instance of the global resource.
  34 Changes to the global resource are visible to other processes
  35 that are members of the namespace, but are invisible to other processes.
  36 One use of namespaces is to implement containers.
  37 .PP
  38 Linux provides the following namespaces:
  39 .TS
  40 lB lB lB
  41 l lB l.
  42 Namespace       Constant        Isolates
  43 Cgroup  CLONE_NEWCGROUP Cgroup root directory
  44 IPC     CLONE_NEWIPC    System V IPC, POSIX message queues
  45 Network CLONE_NEWNET    Network devices, stacks, ports, etc.
  46 Mount   CLONE_NEWNS     Mount points
  47 PID     CLONE_NEWPID    Process IDs
  48 User    CLONE_NEWUSER   User and group IDs
  49 UTS     CLONE_NEWUTS    Hostname and NIS domain name
  50 .TE
  51 .PP
  52 This page describes the various namespaces and the associated
  53 .I /proc
  54 files, and summarizes the APIs for working with namespaces.
  55 .\"
  56 .\" ==================== The namespaces API ====================
  57 .\"
  58 .SS The namespaces API
  59 As well as various
  60 .I /proc
  61 files described below,
  62 the namespaces API includes the following system calls:
  63 .TP
  64 .BR clone (2)
  65 The
  66 .BR clone (2)
  67 system call creates a new process.
  68 If the
  69 .I flags
  70 argument of the call specifies one or more of the
  71 .B CLONE_NEW*
  72 flags listed below, then new namespaces are created for each flag,
  73 and the child process is made a member of those namespaces.
  74 (This system call also implements a number of features
  75 unrelated to namespaces.)
  76 .TP
  77 .BR setns (2)
  78 The
  79 .BR setns (2)
  80 system call allows the calling process to join an existing namespace.
  81 The namespace to join is specified via a file descriptor that refers to
  82 one of the
  83 .IR /proc/[pid]/ns
  84 files described below.
  85 .TP
  86 .BR unshare (2)
  87 The
  88 .BR unshare (2)
  89 system call moves the calling process to a new namespace.
  90 If the
  91 .I flags
  92 argument of the call specifies one or more of the
  93 .B CLONE_NEW*
  94 flags listed below, then new namespaces are created for each flag,
  95 and the calling process is made a member of those namespaces.
  96 (This system call also implements a number of features
  97 unrelated to namespaces.)
  98 .TP
  99 .BR ioctl (2)
 100 Various
 101 .BR ioctl (2)
 102 operations can be used to discover information about namespaces.
 103 These operations are described in
 104 .BR ioctl_ns (2).
 105 .PP
 106 Creation of new namespaces using
 107 .BR clone (2)
 108 and
 109 .BR unshare (2)
 110 in most cases requires the
 111 .BR CAP_SYS_ADMIN
 112 capability, since, in the new namespace,
 113 the creator will have the power to change global resources
 114 that are visible to other processes that are subsequently created in,
 115 or join the namespace.
 116 User namespaces are the exception: since Linux 3.8,
 117 no privilege is required to create a user namespace.
 118 .\"
 119 .\" ==================== The /proc/[pid]/ns/ directory ====================
 120 .\"
 121 .SS The /proc/[pid]/ns/ directory
 122 Each process has a
 123 .IR /proc/[pid]/ns/
 124 .\" See commit 6b4e306aa3dc94a0545eb9279475b1ab6209a31f
 125 subdirectory containing one entry for each namespace that
 126 supports being manipulated by
 127 .BR setns (2):
 128 .PP
 129 .in +4n
 130 .EX
 131 $ \fBls \-l /proc/$$/ns\fP
 132 total 0
 133 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 cgroup \-> cgroup:[4026531835]
 134 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 ipc \-> ipc:[4026531839]
 135 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 mnt \-> mnt:[4026531840]
 136 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 net \-> net:[4026531969]
 137 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid \-> pid:[4026531836]
 138 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid_for_children \-> pid:[4026531834]
 139 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 user \-> user:[4026531837]
 140 lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 uts \-> uts:[4026531838]
 141 .EE
 142 .in
 143 .PP
 144 Bind mounting (see
 145 .BR mount (2))
 146 one of the files in this directory
 147 to somewhere else in the filesystem keeps
 148 the corresponding namespace of the process specified by
 149 .I pid
 150 alive even if all processes currently in the namespace terminate.
 151 .PP
 152 Opening one of the files in this directory
 153 (or a file that is bind mounted to one of these files)
 154 returns a file handle for
 155 the corresponding namespace of the process specified by
 156 .IR pid .
 157 As long as this file descriptor remains open,
 158 the namespace will remain alive,
 159 even if all processes in the namespace terminate.
 160 The file descriptor can be passed to
 161 .BR setns (2).
 162 .PP
 163 In Linux 3.7 and earlier, these files were visible as hard links.
 164 Since Linux 3.8,
 165 .\" commit bf056bfa80596a5d14b26b17276a56a0dcb080e5
 166 they appear as symbolic links.
 167 If two processes are in the same namespace,
 168 then the device IDs and inode numbers of their
 169 .IR /proc/[pid]/ns/xxx
 170 symbolic links will be the same; an application can check this using the
 171 .I stat.st_dev
 172 and
 173 .I stat.st_ino
 174 fields returned by
 175 .BR stat (2).
 176 The content of this symbolic link is a string containing
 177 the namespace type and inode number as in the following example:
 178 .PP
 179 .in +4n
 180 .EX
 181 $ \fBreadlink /proc/$$/ns/uts\fP
 182 uts:[4026531838]
 183 .EE
 184 .in
 185 .PP
 186 The symbolic links in this subdirectory are as follows:
 187 .TP
 188 .IR /proc/[pid]/ns/cgroup " (since Linux 4.6)"
 189 This file is a handle for the cgroup namespace of the process.
 190 .TP
 191 .IR /proc/[pid]/ns/ipc " (since Linux 3.0)"
 192 This file is a handle for the IPC namespace of the process.
 193 .TP
 194 .IR /proc/[pid]/ns/mnt " (since Linux 3.8)"
 195 .\" commit 8823c079ba7136dc1948d6f6dcb5f8022bde438e
 196 This file is a handle for the mount namespace of the process.
 197 .TP
 198 .IR /proc/[pid]/ns/net " (since Linux 3.0)"
 199 This file is a handle for the network namespace of the process.
 200 .TP
 201 .IR /proc/[pid]/ns/pid " (since Linux 3.8)"
 202 .\" commit 57e8391d327609cbf12d843259c968b9e5c1838f
 203 This file is a handle for the PID namespace of the process.
 204 This handle is permanent for the lifetime of the process
 205 (i.e., a process's PID namespace membership never changes).
 206 .TP
 207 .IR /proc/[pid]/ns/pid_for_children " (since Linux 4.12)"
 208 .\" commit eaa0d190bfe1ed891b814a52712dcd852554cb08
 209 This file is a handle for the PID namespace of
 210 child processes created by this process.
 211 This can change as a consequence of calls to
 212 .BR unshare (2)
 213 and
 214 .BR setns (2)
 215 (see
 216 .BR pid_namespaces (7)),
 217 so the file may differ from
 218 .IR /proc/[pid]/ns/pid .
 219 The symbolic link gains a value only after the first child process
 220 is created in the namespace.
 221 (Beforehand,
 222 .BR readlink (2)
 223 of the symbolic link will return an empty buffer.)
 224 .TP
 225 .IR /proc/[pid]/ns/user " (since Linux 3.8)"
 226 .\" commit cde1975bc242f3e1072bde623ef378e547b73f91
 227 This file is a handle for the user namespace of the process.
 228 .TP
 229 .IR /proc/[pid]/ns/uts " (since Linux 3.0)"
 230 This file is a handle for the UTS namespace of the process.
 231 .PP
 232 Permission to dereference or read
 233 .RB ( readlink (2))
 234 these symbolic links is governed by a ptrace access mode
 235 .B PTRACE_MODE_READ_FSCREDS
 236 check; see
 237 .BR ptrace (2).
 238 .\"
 239 .\" ==================== The /proc/sys/user directory ====================
 240 .\"
 241 .SS The /proc/sys/user directory
 242 The files in the
 243 .I /proc/sys/user
 244 directory (which is present since Linux 4.9) expose limits
 245 on the number of namespaces of various types that can be created.
 246 The files are as follows:
 247 .TP
 248 .IR max_cgroup_namespaces
 249 The value in this file defines a per-user limit on the number of
 250 cgroup namespaces that may be created in the user namespace.
 251 .TP
 252 .IR max_ipc_namespaces
 253 The value in this file defines a per-user limit on the number of
 254 ipc namespaces that may be created in the user namespace.
 255 .TP
 256 .IR max_mnt_namespaces
 257 The value in this file defines a per-user limit on the number of
 258 mount namespaces that may be created in the user namespace.
 259 .TP
 260 .IR max_net_namespaces
 261 The value in this file defines a per-user limit on the number of
 262 network namespaces that may be created in the user namespace.
 263 .TP
 264 .IR max_pid_namespaces
 265 The value in this file defines a per-user limit on the number of
 266 pid namespaces that may be created in the user namespace.
 267 .TP
 268 .IR max_user_namespaces
 269 The value in this file defines a per-user limit on the number of
 270 user namespaces that may be created in the user namespace.
 271 .TP
 272 .IR max_uts_namespaces
 273 The value in this file defines a per-user limit on the number of
 274 uts namespaces that may be created in the user namespace.
 275 .PP
 276 Note the following details about these files:
 277 .IP * 3
 278 The values in these files are modifiable by privileged processes.
 279 .IP *
 280 The values exposed by these files are the limits for the user namespace
 281 in which the opening process resides.
 282 .IP *
 283 The limits are per-user.
 284 Each user in the same user namespace
 285 can create namespaces up to the defined limit.
 286 .IP *
 287 The limits apply to all users, including UID 0.
 288 .IP *
 289 These limits apply in addition to any other per-namespace
 290 limits (such as those for PID and user namespaces) that may be enforced.
 291 .IP *
 292 Upon encountering these limits,
 293 .BR clone (2)
 294 and
 295 .BR unshare (2)
 296 fail with the error
 297 .BR ENOSPC .
 298 .IP *
 299 For the initial user namespace,
 300 the default value in each of these files is half the limit on the number
 301 of threads that may be created
 302 .RI ( /proc/sys/kernel/threads-max ).
 303 In all descendant user namespaces, the default value in each file is
 304 .BR MAXINT .
 305 .IP *
 306 When a namespace is created, the object is also accounted
 307 against ancestor namespaces.
 308 More precisely:
 309 .RS
 310 .IP + 3
 311 Each user namespace has a creator UID.
 312 .IP +
 313 When a namespace is created,
 314 it is accounted against the creator UIDs in each of the
 315 ancestor user namespaces,
 316 and the kernel ensures that the corresponding namespace limit
 317 for the creator UID in the ancestor namespace is not exceeded.
 318 .IP +
 319 The aforementioned point ensures that creating a new user namespace
 320 cannot be used as a means to escape the limits in force
 321 in the current user namespace.
 322 .RE
 323 .\"
 324 .\" ==================== Cgroup namespaces ====================
 325 .\"
 326 .SS Cgroup namespaces (CLONE_NEWCGROUP)
 327 See
 328 .BR cgroup_namespaces (7).
 329 .\"
 330 .\" ==================== IPC namespaces ====================
 331 .\"
 332 .SS IPC namespaces (CLONE_NEWIPC)
 333 IPC namespaces isolate certain IPC resources,
 334 namely, System V IPC objects (see
 335 .BR sysvipc (7))
 336 and (since Linux 2.6.30)
 337 .\" commit 7eafd7c74c3f2e67c27621b987b28397110d643f
 338 .\" https://lwn.net/Articles/312232/
 339 POSIX message queues (see
 340 .BR mq_overview (7)).
 341 The common characteristic of these IPC mechanisms is that IPC
 342 objects are identified by mechanisms other than filesystem
 343 pathnames.
 344 .PP
 345 Each IPC namespace has its own set of System V IPC identifiers and
 346 its own POSIX message queue filesystem.
 347 Objects created in an IPC namespace are visible to all other processes
 348 that are members of that namespace,
 349 but are not visible to processes in other IPC namespaces.
 350 .PP
 351 The following
 352 .I /proc
 353 interfaces are distinct in each IPC namespace:
 354 .IP * 3
 355 The POSIX message queue interfaces in
 356 .IR /proc/sys/fs/mqueue .
 357 .IP *
 358 The System V IPC interfaces in
 359 .IR /proc/sys/kernel ,
 360 namely:
 361 .IR msgmax ,
 362 .IR msgmnb  ,
 363 .IR msgmni ,
 364 .IR sem ,
 365 .IR shmall ,
 366 .IR shmmax ,
 367 .IR shmmni ,
 368 and
 369 .IR shm_rmid_forced .
 370 .IP *
 371 The System V IPC interfaces in
 372 .IR /proc/sysvipc .
 373 .PP
 374 When an IPC namespace is destroyed
 375 (i.e., when the last process that is a member of the namespace terminates),
 376 all IPC objects in the namespace are automatically destroyed.
 377 .PP
 378 Use of IPC namespaces requires a kernel that is configured with the
 379 .B CONFIG_IPC_NS
 380 option.
 381 .\"
 382 .\" ==================== Network namespaces ====================
 383 .\"
 384 .SS Network namespaces (CLONE_NEWNET)
 385 See
 386 .BR network_namespaces (7).
 387 .\"
 388 .\" ==================== Mount namespaces ====================
 389 .\"
 390 .SS Mount namespaces (CLONE_NEWNS)
 391 See
 392 .BR mount_namespaces (7).
 393 .\"
 394 .\" ==================== PID namespaces ====================
 395 .\"
 396 .SS PID namespaces (CLONE_NEWPID)
 397 See
 398 .BR pid_namespaces (7).
 399 .\"
 400 .\" ==================== User namespaces ====================
 401 .\"
 402 .SS User namespaces (CLONE_NEWUSER)
 403 See
 404 .BR user_namespaces (7).
 405 .\"
 406 .\" ==================== UTS namespaces ====================
 407 .\"
 408 .SS UTS namespaces (CLONE_NEWUTS)
 409 UTS namespaces provide isolation of two system identifiers:
 410 the hostname and the NIS domain name.
 411 These identifiers are set using
 412 .BR sethostname (2)
 413 and
 414 .BR setdomainname (2),
 415 and can be retrieved using
 416 .BR uname (2),
 417 .BR gethostname (2),
 418 and
 419 .BR getdomainname (2).
 420 .PP
 421 .PP
 422 When a process creates a new UTS namespace using
 423 .BR clone (2)
 424 or
 425 .BR unshare (2)
 426 with the
 427 .BR CLONE_NEWUTS
 428 flag, the hostname and domain of the new UTS namespace are copied
 429 from the corresponding values in the caller's UTS namespace.
 430 .PP
 431 Use of UTS namespaces requires a kernel that is configured with the
 432 .B CONFIG_UTS_NS
 433 option.
 434 .\"
 435 .SS Namespace lifetime
 436 Absent any other factors,
 437 a namespace is automatically torn down when the last process in
 438 the namespace terminates or leaves the namespace.
 439 However, there are a number of other factors that may pin
 440 a namespace into existence even though it has no member processes.
 441 These factors include the following:
 442 .IP * 3
 443 An open file descriptor or a bind mount exists for the corresponding
 444 .IR /proc/[pid]/ns/*
 445 file.
 446 .IP *
 447 The namespace is hierarchical (i.e., a PID or user namespace),
 448 and has a child namespace.
 449 .IP *
 450 It is a user namespace that owns one or more nonuser namespaces.
 451 .IP *
 452 It is a PID namespace,
 453 and there is a process that refers to the namespace via a
 454 .IR /proc/[pid]/ns/pid_for_children
 455 symbolic link.
 456 .IP *
 457 It is an IPC namespace, and a corresponding mount of an
 458 .I mqueue
 459 filesystem (see
 460 .BR mq_overview (7))
 461 refers to this namespace.
 462 .IP *
 463 It is a PID namespace, and a corresponding mount of a
 464 .BR proc (5)
 465 filesystem refers to this namespace.
 466 .SH EXAMPLE
 467 See
 468 .BR clone (2)
 469 and
 470 .BR user_namespaces (7).
 471 .SH SEE ALSO
 472 .BR nsenter (1),
 473 .BR readlink (1),
 474 .BR unshare (1),
 475 .BR clone (2),
 476 .BR ioctl_ns (2),
 477 .BR setns (2),
 478 .BR unshare (2),
 479 .BR proc (5),
 480 .BR capabilities (7),
 481 .BR cgroup_namespaces (7),
 482 .BR cgroups (7),
 483 .BR credentials (7),
 484 .BR network_namespaces (7),
 485 .BR pid_namespaces (7),
 486 .BR user_namespaces (7),
 487 .BR lsns (8),
 488 .BR pam_namespace (8),
 489 .BR switch_root (8)