1 .\" Copyright (c) 2013, 2014 by Michael Kerrisk <mtk.manpages@gmail.com>
2 .\" and Copyright (c) 2012, 2014 by Eric W. Biederman <ebiederm@xmission.com>
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.
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.
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15 .\" manual page may be incorrect or out-of-date.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
22 .\" Formatted or processed versions of this manual, if unaccompanied by
23 .\" the source, must acknowledge the copyright and authors of this work.
27 .TH USER_NAMESPACES 7 2019-08-02 "Linux" "Linux Programmer's Manual"
29 user_namespaces \- overview of Linux user namespaces
31 For an overview of namespaces, see
34 User namespaces isolate security-related identifiers and attributes,
36 user IDs and group IDs (see
41 .\" FIXME: This page says very little about the interaction
42 .\" of user namespaces and keys. Add something on this topic.
44 .BR capabilities (7)).
45 A process's user and group IDs can be different
46 inside and outside a user namespace.
48 a process can have a normal unprivileged user ID outside a user namespace
49 while at the same time having a user ID of 0 inside the namespace;
51 the process has full privileges for operations inside the user namespace,
52 but is unprivileged for operations outside the namespace.
54 .\" ============================================================
56 .SS Nested namespaces, namespace membership
57 User namespaces can be nested;
58 that is, each user namespace\(emexcept the initial ("root")
59 namespace\(emhas a parent user namespace,
60 and can have zero or more child user namespaces.
61 The parent user namespace is the user namespace
62 of the process that creates the user namespace via a call to
70 The kernel imposes (since version 3.11) a limit of 32 nested levels of
71 .\" commit 8742f229b635bf1c1c84a3dfe5e47c814c20b5c8
73 .\" FIXME Explain the rationale for this limit. (What is the rationale?)
78 that would cause this limit to be exceeded fail with the error
81 Each process is a member of exactly one user namespace.
88 flag is a member of the same user namespace as its parent.
89 A single-threaded process can join another user namespace with
94 upon doing so, it gains a full set of capabilities in that namespace.
102 flag makes the new child process (for
106 a member of the new user namespace created by the call.
111 operation can be used to discover the parental relationship
112 between user namespaces; see
115 .\" ============================================================
118 The child process created by
122 flag starts out with a complete set
123 of capabilities in the new user namespace.
124 Likewise, a process that creates a new user namespace using
126 or joins an existing user namespace using
128 gains a full set of capabilities in that namespace.
130 that process has no capabilities in the parent (in the case of
132 or previous (in the case of
137 even if the new namespace is created or joined by the root user
138 (i.e., a process with user ID 0 in the root namespace).
142 will cause a process's capabilities to be recalculated in the usual way (see
143 .BR capabilities (7)).
145 unless the process has a user ID of 0 within the namespace,
146 or the executable file has a nonempty inheritable capabilities mask,
147 the process will lose all capabilities.
148 See the discussion of user and group ID mappings, below.
159 that moves the caller into another user namespace
160 sets the "securebits" flags
162 .BR capabilities (7))
163 to their default values (all flags disabled) in the child (for
169 Note that because the caller no longer has capabilities
170 in its original user namespace after a call to
172 it is not possible for a process to reset its "securebits" flags while
173 retaining its user namespace membership by using a pair of
175 calls to move to another user namespace and then return to
176 its original user namespace.
178 The rules for determining whether or not a process has a capability
179 in a particular user namespace are as follows:
181 A process has a capability inside a user namespace
182 if it is a member of that namespace and
183 it has the capability in its effective capability set.
184 A process can gain capabilities in its effective capability
186 For example, it may execute a set-user-ID program or an
187 executable with associated file capabilities.
189 a process may gain capabilities via the effect of
194 as already described.
195 .\" In the 3.8 sources, see security/commoncap.c::cap_capable():
197 If a process has a capability in a user namespace,
198 then it has that capability in all child (and further removed descendant)
201 .\" * The owner of the user namespace in the parent of the
202 .\" * user namespace has all caps.
203 When a user namespace is created, the kernel records the effective
204 user ID of the creating process as being the "owner" of the namespace.
205 .\" (and likewise associates the effective group ID of the creating process
206 .\" with the namespace).
207 A process that resides
208 in the parent of the user namespace
209 .\" See kernel commit 520d9eabce18edfef76a60b7b839d54facafe1f9 for a fix
211 and whose effective user ID matches the owner of the namespace
212 has all capabilities in the namespace.
213 .\" This includes the case where the process executes a set-user-ID
214 .\" program that confers the effective UID of the creator of the namespace.
215 By virtue of the previous rule,
216 this means that the process has all capabilities in all
217 further removed descendant user namespaces as well.
221 operation can be used to discover the user ID of the owner of the namespace;
225 .\" ============================================================
227 .SS Effect of capabilities within a user namespace
228 Having a capability inside a user namespace
229 permits a process to perform operations (that require privilege)
230 only on resources governed by that namespace.
231 In other words, having a capability in a user namespace permits a process
232 to perform privileged operations on resources that are governed by (nonuser)
233 namespaces owned by (associated with) the user namespace
234 (see the next subsection).
236 On the other hand, there are many privileged operations that affect
237 resources that are not associated with any namespace type,
238 for example, changing the system time (governed by
240 loading a kernel module (governed by
241 .BR CAP_SYS_MODULE ),
242 and creating a device (governed by
244 Only a process with privileges in the
246 user namespace can perform such operations.
250 within the user namespace that owns a process's mount namespace
251 allows that process to create bind mounts
252 and mount the following types of filesystems:
253 .\" fs_flags = FS_USERNS_MOUNT in kernel sources
277 .\" commit b2197755b2633e164a439682fb05a9b5ea48f706
284 within the user namespace that owns a process's cgroup namespace
285 allows (since Linux 4.6)
286 that process to the mount the cgroup version 2 filesystem and
287 cgroup version 1 named hierarchies
288 (i.e., cgroup filesystems mounted with the
294 within the user namespace that owns a process's PID namespace
295 allows (since Linux 3.8)
296 that process to mount
300 Note however, that mounting block-based filesystems can be done
301 only by a process that holds
303 in the initial user namespace.
305 .\" ============================================================
307 .SS Interaction of user namespaces and other types of namespaces
308 Starting in Linux 3.8, unprivileged processes can create user namespaces,
309 and the other types of namespaces can be created with just the
311 capability in the caller's user namespace.
313 When a nonuser namespace is created,
314 it is owned by the user namespace in which the creating process
315 was a member at the time of the creation of the namespace.
316 Privileged operations on resources governed by the nonuser namespace
317 require that the process has the necessary capabilities
318 in the user namespace that owns the nonuser namespace.
322 is specified along with other
328 call, the user namespace is guaranteed to be created first,
333 privileges over the remaining namespaces created by the call.
334 Thus, it is possible for an unprivileged caller to specify this combination
337 When a new namespace (other than a user namespace) is created via
341 the kernel records the user namespace of the creating process as the owner of
343 (This association can't be changed.)
344 When a process in the new namespace subsequently performs
345 privileged operations that operate on global
346 resources isolated by the namespace,
347 the permission checks are performed according to the process's capabilities
348 in the user namespace that the kernel associated with the new namespace.
349 For example, suppose that a process attempts to change the hostname
350 .RB ( sethostname (2)),
351 a resource governed by the UTS namespace.
353 the kernel will determine which user namespace owns
354 the process's UTS namespace, and check whether the process has the
356 .RB ( CAP_SYS_ADMIN )
357 in that user namespace.
362 operation can be used to discover the user namespace
363 that owns a nonuser namespace; see
366 .\" ============================================================
368 .SS User and group ID mappings: uid_map and gid_map
369 When a user namespace is created,
370 it starts out without a mapping of user IDs (group IDs)
371 to the parent user namespace.
373 .IR /proc/[pid]/uid_map
375 .IR /proc/[pid]/gid_map
376 files (available since Linux 3.5)
377 .\" commit 22d917d80e842829d0ca0a561967d728eb1d6303
378 expose the mappings for user and group IDs
379 inside the user namespace for the process
381 These files can be read to view the mappings in a user namespace and
382 written to (once) to define the mappings.
384 The description in the following paragraphs explains the details for
388 but each instance of "user ID" is replaced by "group ID".
392 file exposes the mapping of user IDs from the user namespace
395 to the user namespace of the process that opened
397 (but see a qualification to this point below).
398 In other words, processes that are in different user namespaces
399 will potentially see different values when reading from a particular
401 file, depending on the user ID mappings for the user namespaces
402 of the reading processes.
406 file specifies a 1-to-1 mapping of a range of contiguous
407 user IDs between two user namespaces.
408 (When a user namespace is first created, this file is empty.)
409 The specification in each line takes the form of
410 three numbers delimited by white space.
411 The first two numbers specify the starting user ID in
412 each of the two user namespaces.
413 The third number specifies the length of the mapped range.
414 In detail, the fields are interpreted as follows:
416 The start of the range of user IDs in
417 the user namespace of the process
420 The start of the range of user
421 IDs to which the user IDs specified by field one map.
422 How field two is interpreted depends on whether the process that opened
426 are in the same user namespace, as follows:
429 If the two processes are in different user namespaces:
430 field two is the start of a range of
431 user IDs in the user namespace of the process that opened
434 If the two processes are in the same user namespace:
435 field two is the start of the range of
436 user IDs in the parent user namespace of the process
438 This case enables the opener of
440 (the common case here is opening
441 .IR /proc/self/uid_map )
442 to see the mapping of user IDs into the user namespace of the process
443 that created this user namespace.
446 The length of the range of user IDs that is mapped between the two
449 System calls that return user IDs (group IDs)\(emfor example,
452 and the credential fields in the structure returned by
453 .BR stat (2)\(emreturn
454 the user ID (group ID) mapped into the caller's user namespace.
456 When a process accesses a file, its user and group IDs
457 are mapped into the initial user namespace for the purpose of permission
458 checking and assigning IDs when creating a file.
459 When a process retrieves file user and group IDs via
461 the IDs are mapped in the opposite direction,
462 to produce values relative to the process user and group ID mappings.
464 The initial user namespace has no parent namespace,
465 but, for consistency, the kernel provides dummy user and group
466 ID mapping files for this namespace.
471 is the same) from a shell in the initial namespace shows:
475 $ \fBcat /proc/$$/uid_map\fP
480 This mapping tells us
481 that the range starting at user ID 0 in this namespace
482 maps to a range starting at 0 in the (nonexistent) parent namespace,
483 and the length of the range is the largest 32-bit unsigned integer.
484 This leaves 4294967295 (the 32-bit signed \-1 value) unmapped.
487 is used in several interfaces (e.g.,
489 as a way to specify "no user ID".
492 unmapped and unusable guarantees that there will be no
493 confusion when using these interfaces.
495 .\" ============================================================
497 .SS Defining user and group ID mappings: writing to uid_map and gid_map
499 After the creation of a new user namespace, the
503 of the processes in the namespace may be written to
505 to define the mapping of user IDs in the new user namespace.
506 An attempt to write more than once to a
508 file in a user namespace fails with the error
510 Similar rules apply for
517 must conform to the following rules:
519 The three fields must be valid numbers,
520 and the last field must be greater than 0.
522 Lines are terminated by newline characters.
524 There is a limit on the number of lines in the file.
525 In Linux 4.14 and earlier, this limit was (arbitrarily)
526 .\" 5*12-byte records could fit in a 64B cache line
529 .\" commit 6397fac4915ab3002dc15aae751455da1a852f25
530 the limit is 340 lines.
531 In addition, the number of bytes written to
532 the file must be less than the system page size,
533 and the write must be performed at the start of the file (i.e.,
537 can't be used to write to nonzero offsets in the file).
539 The range of user IDs (group IDs)
540 specified in each line cannot overlap with the ranges
542 In the initial implementation (Linux 3.8), this requirement was
543 satisfied by a simplistic implementation that imposed the further
545 the values in both field 1 and field 2 of successive lines must be
546 in ascending numerical order,
547 which prevented some otherwise valid maps from being created.
549 .\" commit 0bd14b4fd72afd5df41e9fd59f356740f22fceba
550 fix this limitation, allowing any valid set of nonoverlapping maps.
552 At least one line must be written to the file.
554 Writes that violate the above rules fail with the error
557 In order for a process to write to the
558 .I /proc/[pid]/uid_map
559 .RI ( /proc/[pid]/gid_map )
560 file, all of the following requirements must be met:
562 The writing process must have the
565 capability in the user namespace of the process
568 The writing process must either be in the user namespace of the process
570 or be in the parent user namespace of the process
573 The mapped user IDs (group IDs) must in turn have a mapping
574 in the parent user namespace.
576 One of the following two cases applies:
580 the writing process has the
588 No further restrictions apply:
589 the process can make mappings to arbitrary user IDs (group IDs)
590 in the parent user namespace.
594 otherwise all of the following restrictions apply:
600 must consist of a single line that maps
601 the writing process's effective user ID
602 (group ID) in the parent user namespace to a user ID (group ID)
603 in the user namespace.
605 The writing process must have the same effective user ID as the process
606 that created the user namespace.
612 system call must first be denied by writing
615 .I /proc/[pid]/setgroups
616 file (see below) before writing to
621 Writes that violate the above rules fail with the error
624 .\" ============================================================
626 .SS Interaction with system calls that change process UIDs or GIDs
627 In a user namespace where the
629 file has not been written, the system calls that change user IDs will fail.
632 file has not been written, the system calls that change group IDs will fail.
637 files have been written, only the mapped values may be used in
638 system calls that change user and group IDs.
640 For user IDs, the relevant system calls include
646 For group IDs, the relevant system calls include
657 .I /proc/[pid]/setgroups
658 file before writing to
659 .I /proc/[pid]/gid_map
660 .\" Things changed in Linux 3.19
661 .\" commit 9cc46516ddf497ea16e8d7cb986ae03a0f6b92f8
662 .\" commit 66d2f338ee4c449396b6f99f5e75cd18eb6df272
663 .\" http://lwn.net/Articles/626665/
664 will permanently disable
666 in a user namespace and allow writing to
667 .I /proc/[pid]/gid_map
670 capability in the parent user namespace.
672 .\" ============================================================
674 .SS The /proc/[pid]/setgroups file
676 .\" commit 9cc46516ddf497ea16e8d7cb986ae03a0f6b92f8
677 .\" commit 66d2f338ee4c449396b6f99f5e75cd18eb6df272
678 .\" http://lwn.net/Articles/626665/
679 .\" http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2014-8989
682 .I /proc/[pid]/setgroups
683 file displays the string
685 if processes in the user namespace that contains the process
687 are permitted to employ the
689 system call; it displays
693 is not permitted in that user namespace.
694 Note that regardless of the value in the
695 .I /proc/[pid]/setgroups
696 file (and regardless of the process's capabilities), calls to
698 are also not permitted if
699 .IR /proc/[pid]/gid_map
700 has not yet been set.
702 A privileged process (one with the
704 capability in the namespace) may write either of the strings
710 writing a group ID mapping
711 for this user namespace to the file
712 .IR /proc/[pid]/gid_map .
715 prevents any process in the user namespace from employing
718 The essence of the restrictions described in the preceding
719 paragraph is that it is permitted to write to
720 .I /proc/[pid]/setgroups
721 only so long as calling
723 is disallowed because
724 .I /proc/[pid]/gid_map
726 This ensures that a process cannot transition from a state where
728 is allowed to a state where
731 a process can transition only from
737 The default value of this file in the initial user namespace is
741 .IR /proc/[pid]/gid_map
743 (which has the effect of enabling
745 in the user namespace),
746 it is no longer possible to disallow
751 .IR /proc/[pid]/setgroups
752 (the write fails with the error
755 A child user namespace inherits the
756 .IR /proc/[pid]/setgroups
757 setting from its parent.
765 system call can't subsequently be reenabled (by writing
767 to the file) in this user namespace.
768 (Attempts to do so fail with the error
770 This restriction also propagates down to all child user namespaces of
774 .I /proc/[pid]/setgroups
775 file was added in Linux 3.19,
776 but was backported to many earlier stable kernel series,
777 because it addresses a security issue.
778 The issue concerned files with permissions such as "rwx\-\-\-rwx".
779 Such files give fewer permissions to "group" than they do to "other".
780 This means that dropping groups using
782 might allow a process file access that it did not formerly have.
783 Before the existence of user namespaces this was not a concern,
784 since only a privileged process (one with the
786 capability) could call
788 However, with the introduction of user namespaces,
789 it became possible for an unprivileged process to create
790 a new namespace in which the user had all privileges.
791 This then allowed formerly unprivileged
792 users to drop groups and thus gain file access
793 that they did not previously have.
795 .I /proc/[pid]/setgroups
796 file was added to address this security issue,
797 by denying any pathway for an unprivileged process to drop groups with
800 .\" /proc/PID/setgroups
801 .\" [allow == setgroups() is allowed, "deny" == setgroups() is disallowed]
802 .\" * Can write if have CAP_SYS_ADMIN in NS
803 .\" * Must write BEFORE writing to /proc/PID/gid_map
806 .\" * Must already have written to gid_map
807 .\" * /proc/PID/setgroups must be "allow"
809 .\" /proc/PID/gid_map -- writing
810 .\" * Must already have written "deny" to /proc/PID/setgroups
812 .\" ============================================================
814 .SS Unmapped user and group IDs
816 There are various places where an unmapped user ID (group ID)
817 may be exposed to user space.
818 For example, the first process in a new user namespace may call
820 before a user ID mapping has been defined for the namespace.
821 In most such cases, an unmapped user ID is converted
822 .\" from_kuid_munged(), from_kgid_munged()
823 to the overflow user ID (group ID);
824 the default value for the overflow user ID (group ID) is 65534.
825 See the descriptions of
826 .IR /proc/sys/kernel/overflowuid
828 .IR /proc/sys/kernel/overflowgid
832 The cases where unmapped IDs are mapped in this fashion include
833 system calls that return user IDs
837 credentials passed over a UNIX domain socket,
839 credentials returned by
842 and the System V IPC "ctl"
845 credentials exposed by
846 .IR /proc/[pid]/status
848 .IR /proc/sysvipc/* ,
849 credentials returned via the
853 received with a signal (see
855 credentials written to the process accounting file (see
857 and credentials returned with POSIX message queue notifications (see
860 There is one notable case where unmapped user and group IDs are
862 .\" from_kuid(), from_kgid()
863 .\" Also F_GETOWNER_UIDS is an exception
864 converted to the corresponding overflow ID value.
869 file in which there is no mapping for the second field,
870 that field is displayed as 4294967295 (\-1 as an unsigned integer).
872 .\" ============================================================
876 In order to determine permissions when an unprivileged process accesses a file,
877 the process credentials (UID, GID) and the file credentials
878 are in effect mapped back to what they would be in
879 the initial user namespace and then compared to determine
880 the permissions that the process has on the file.
881 The same is also of other objects that employ the credentials plus
882 permissions mask accessibility model, such as System V IPC objects
884 .\" ============================================================
886 .SS Operation of file-related capabilities
888 Certain capabilities allow a process to bypass various
889 kernel-enforced restrictions when performing operations on
890 files owned by other users or groups.
891 These capabilities are:
893 .BR CAP_DAC_OVERRIDE ,
894 .BR CAP_DAC_READ_SEARCH ,
899 Within a user namespace,
900 these capabilities allow a process to bypass the rules
901 if the process has the relevant capability over the file,
904 the process has the relevant effective capability in its user namespace; and
906 the file's user ID and group ID both have valid mappings
907 in the user namespace.
911 capability is treated somewhat exceptionally:
912 .\" These are the checks performed by the kernel function
913 .\" inode_owner_or_capable(). There is one exception to the exception:
914 .\" overriding the directory sticky permission bit requires that
915 .\" the file has a valid mapping for both its UID and GID.
916 it allows a process to bypass the corresponding rules so long as
917 at least the file's user ID has a mapping in the user namespace
918 (i.e., the file's group ID does not need to have a valid mapping).
920 .\" ============================================================
922 .SS Set-user-ID and set-group-ID programs
924 When a process inside a user namespace executes
925 a set-user-ID (set-group-ID) program,
926 the process's effective user (group) ID inside the namespace is changed
927 to whatever value is mapped for the user (group) ID of the file.
928 However, if either the user
930 the group ID of the file has no mapping inside the namespace,
931 the set-user-ID (set-group-ID) bit is silently ignored:
932 the new program is executed,
933 but the process's effective user (group) ID is left unchanged.
934 (This mirrors the semantics of executing a set-user-ID or set-group-ID
935 program that resides on a filesystem that was mounted with the
937 flag, as described in
940 .\" ============================================================
944 When a process's user and group IDs are passed over a UNIX domain socket
945 to a process in a different user namespace (see the description of
949 they are translated into the corresponding values as per the
950 receiving process's user and group ID mappings.
953 Namespaces are a Linux-specific feature.
956 Over the years, there have been a lot of features that have been added
957 to the Linux kernel that have been made available only to privileged users
958 because of their potential to confuse set-user-ID-root applications.
959 In general, it becomes safe to allow the root user in a user namespace to
960 use those features because it is impossible, while in a user namespace,
961 to gain more privilege than the root user of a user namespace has.
963 .\" ============================================================
966 Use of user namespaces requires a kernel that is configured with the
969 User namespaces require support in a range of subsystems across
971 When an unsupported subsystem is configured into the kernel,
972 it is not possible to configure user namespaces support.
974 As at Linux 3.8, most relevant subsystems supported user namespaces,
975 but a number of filesystems did not have the infrastructure needed
976 to map user and group IDs between user namespaces.
977 Linux 3.9 added the required infrastructure support for many of
978 the remaining unsupported filesystems
979 (Plan 9 (9P), Andrew File System (AFS), Ceph, CIFS, CODA, NFS, and OCFS2).
980 Linux 3.12 added support for the last of the unsupported major filesystems,
981 .\" commit d6970d4b726cea6d7a9bc4120814f95c09571fc3
985 The program below is designed to allow experimenting with
986 user namespaces, as well as other types of namespaces.
987 It creates namespaces as specified by command-line options and then executes
988 a command inside those namespaces.
991 function inside the program provide a full explanation of the program.
992 The following shell session demonstrates its use.
994 First, we look at the run-time environment:
998 $ \fBuname \-rs\fP # Need Linux 3.8 or later
1000 $ \fBid \-u\fP # Running as unprivileged user
1007 Now start a new shell in new user
1013 namespaces, with user ID
1017 1000 mapped to 0 inside the user namespace:
1021 $ \fB./userns_child_exec \-p \-m \-U \-M '0 1000 1' \-G '0 1000 1' bash\fP
1025 The shell has PID 1, because it is the first process in the new
1037 filesystem and listing all of the processes visible
1038 in the new PID namespace shows that the shell can't see
1039 any processes outside the PID namespace:
1043 bash$ \fBmount \-t proc proc /proc\fP
1045 PID TTY STAT TIME COMMAND
1047 22 pts/3 R+ 0:00 ps ax
1051 Inside the user namespace, the shell has user and group ID 0,
1052 and a full set of permitted and effective capabilities:
1056 bash$ \fBcat /proc/$$/status | egrep '^[UG]id'\fP
1059 bash$ \fBcat /proc/$$/status | egrep '^Cap(Prm|Inh|Eff)'\fP
1060 CapInh: 0000000000000000
1061 CapPrm: 0000001fffffffff
1062 CapEff: 0000001fffffffff
1068 /* userns_child_exec.c
1070 Licensed under GNU General Public License v2 or later
1072 Create a child process that executes a shell command in new
1073 namespace(s); allow UID and GID mappings to be specified when
1074 creating a user namespace.
1080 #include <sys/wait.h>
1088 /* A simple error\-handling function: print an error message based
1089 on the value in \(aqerrno\(aq and terminate the calling process */
1091 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \e
1095 char **argv; /* Command to be executed by child, with args */
1096 int pipe_fd[2]; /* Pipe used to synchronize parent and child */
1104 fprintf(stderr, "Usage: %s [options] cmd [arg...]\en\en", pname);
1105 fprintf(stderr, "Create a child process that executes a shell "
1106 "command in a new user namespace,\en"
1107 "and possibly also other new namespace(s).\en\en");
1108 fprintf(stderr, "Options can be:\en\en");
1109 #define fpe(str) fprintf(stderr, " %s", str);
1110 fpe("\-i New IPC namespace\en");
1111 fpe("\-m New mount namespace\en");
1112 fpe("\-n New network namespace\en");
1113 fpe("\-p New PID namespace\en");
1114 fpe("\-u New UTS namespace\en");
1115 fpe("\-U New user namespace\en");
1116 fpe("\-M uid_map Specify UID map for user namespace\en");
1117 fpe("\-G gid_map Specify GID map for user namespace\en");
1118 fpe("\-z Map user\(aqs UID and GID to 0 in user namespace\en");
1119 fpe(" (equivalent to: \-M \(aq0 <uid> 1\(aq \-G \(aq0 <gid> 1\(aq)\en");
1120 fpe("\-v Display verbose messages\en");
1122 fpe("If \-z, \-M, or \-G is specified, \-U is required.\en");
1123 fpe("It is not permitted to specify both \-z and either \-M or \-G.\en");
1125 fpe("Map strings for \-M and \-G consist of records of the form:\en");
1127 fpe(" ID\-inside\-ns ID\-outside\-ns len\en");
1129 fpe("A map string can contain multiple records, separated"
1131 fpe("the commas are replaced by newlines before writing"
1132 " to map files.\en");
1137 /* Update the mapping file \(aqmap_file\(aq, with the value provided in
1138 \(aqmapping\(aq, a string that defines a UID or GID mapping. A UID or
1139 GID mapping consists of one or more newline\-delimited records
1142 ID_inside\-ns ID\-outside\-ns length
1144 Requiring the user to supply a string that contains newlines is
1145 of course inconvenient for command\-line use. Thus, we permit the
1146 use of commas to delimit records in this string, and replace them
1147 with newlines before writing the string to the file. */
1150 update_map(char *mapping, char *map_file)
1153 size_t map_len; /* Length of \(aqmapping\(aq */
1155 /* Replace commas in mapping string with newlines */
1157 map_len = strlen(mapping);
1158 for (j = 0; j < map_len; j++)
1159 if (mapping[j] == \(aq,\(aq)
1160 mapping[j] = \(aq\en\(aq;
1162 fd = open(map_file, O_RDWR);
1164 fprintf(stderr, "ERROR: open %s: %s\en", map_file,
1169 if (write(fd, mapping, map_len) != map_len) {
1170 fprintf(stderr, "ERROR: write %s: %s\en", map_file,
1178 /* Linux 3.19 made a change in the handling of setgroups(2) and the
1179 \(aqgid_map\(aq file to address a security issue. The issue allowed
1180 *unprivileged* users to employ user namespaces in order to drop
1181 The upshot of the 3.19 changes is that in order to update the
1182 \(aqgid_maps\(aq file, use of the setgroups() system call in this
1183 user namespace must first be disabled by writing "deny" to one of
1184 the /proc/PID/setgroups files for this namespace. That is the
1185 purpose of the following function. */
1188 proc_setgroups_write(pid_t child_pid, char *str)
1190 char setgroups_path[PATH_MAX];
1193 snprintf(setgroups_path, PATH_MAX, "/proc/%ld/setgroups",
1196 fd = open(setgroups_path, O_RDWR);
1199 /* We may be on a system that doesn\(aqt support
1200 /proc/PID/setgroups. In that case, the file won\(aqt exist,
1201 and the system won\(aqt impose the restrictions that Linux 3.19
1202 added. That\(aqs fine: we don\(aqt need to do anything in order
1203 to permit \(aqgid_map\(aq to be updated.
1205 However, if the error from open() was something other than
1206 the ENOENT error that is expected for that case, let the
1209 if (errno != ENOENT)
1210 fprintf(stderr, "ERROR: open %s: %s\en", setgroups_path,
1215 if (write(fd, str, strlen(str)) == \-1)
1216 fprintf(stderr, "ERROR: write %s: %s\en", setgroups_path,
1222 static int /* Start function for cloned child */
1223 childFunc(void *arg)
1225 struct child_args *args = (struct child_args *) arg;
1228 /* Wait until the parent has updated the UID and GID mappings.
1229 See the comment in main(). We wait for end of file on a
1230 pipe that will be closed by the parent process once it has
1231 updated the mappings. */
1233 close(args\->pipe_fd[1]); /* Close our descriptor for the write
1234 end of the pipe so that we see EOF
1235 when parent closes its descriptor */
1236 if (read(args\->pipe_fd[0], &ch, 1) != 0) {
1238 "Failure in child: read from pipe returned != 0\en");
1242 close(args\->pipe_fd[0]);
1244 /* Execute a shell command */
1246 printf("About to exec %s\en", args\->argv[0]);
1247 execvp(args\->argv[0], args\->argv);
1251 #define STACK_SIZE (1024 * 1024)
1253 static char child_stack[STACK_SIZE]; /* Space for child\(aqs stack */
1256 main(int argc, char *argv[])
1258 int flags, opt, map_zero;
1260 struct child_args args;
1261 char *uid_map, *gid_map;
1262 const int MAP_BUF_SIZE = 100;
1263 char map_buf[MAP_BUF_SIZE];
1264 char map_path[PATH_MAX];
1266 /* Parse command\-line options. The initial \(aq+\(aq character in
1267 the final getopt() argument prevents GNU\-style permutation
1268 of command\-line options. That\(aqs useful, since sometimes
1269 the \(aqcommand\(aq to be executed by this program itself
1270 has command\-line options. We don\(aqt want getopt() to treat
1271 those as options to this program. */
1278 while ((opt = getopt(argc, argv, "+imnpuUM:G:zv")) != \-1) {
1280 case \(aqi\(aq: flags |= CLONE_NEWIPC; break;
1281 case \(aqm\(aq: flags |= CLONE_NEWNS; break;
1282 case \(aqn\(aq: flags |= CLONE_NEWNET; break;
1283 case \(aqp\(aq: flags |= CLONE_NEWPID; break;
1284 case \(aqu\(aq: flags |= CLONE_NEWUTS; break;
1285 case \(aqv\(aq: verbose = 1; break;
1286 case \(aqz\(aq: map_zero = 1; break;
1287 case \(aqM\(aq: uid_map = optarg; break;
1288 case \(aqG\(aq: gid_map = optarg; break;
1289 case \(aqU\(aq: flags |= CLONE_NEWUSER; break;
1290 default: usage(argv[0]);
1294 /* \-M or \-G without \-U is nonsensical */
1296 if (((uid_map != NULL || gid_map != NULL || map_zero) &&
1297 !(flags & CLONE_NEWUSER)) ||
1298 (map_zero && (uid_map != NULL || gid_map != NULL)))
1301 args.argv = &argv[optind];
1303 /* We use a pipe to synchronize the parent and child, in order to
1304 ensure that the parent sets the UID and GID maps before the child
1305 calls execve(). This ensures that the child maintains its
1306 capabilities during the execve() in the common case where we
1307 want to map the child\(aqs effective user ID to 0 in the new user
1308 namespace. Without this synchronization, the child would lose
1309 its capabilities if it performed an execve() with nonzero
1310 user IDs (see the capabilities(7) man page for details of the
1311 transformation of a process\(aqs capabilities during execve()). */
1313 if (pipe(args.pipe_fd) == \-1)
1316 /* Create the child in new namespace(s) */
1318 child_pid = clone(childFunc, child_stack + STACK_SIZE,
1319 flags | SIGCHLD, &args);
1320 if (child_pid == \-1)
1323 /* Parent falls through to here */
1326 printf("%s: PID of child created by clone() is %ld\en",
1327 argv[0], (long) child_pid);
1329 /* Update the UID and GID maps in the child */
1331 if (uid_map != NULL || map_zero) {
1332 snprintf(map_path, PATH_MAX, "/proc/%ld/uid_map",
1335 snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getuid());
1338 update_map(uid_map, map_path);
1341 if (gid_map != NULL || map_zero) {
1342 proc_setgroups_write(child_pid, "deny");
1344 snprintf(map_path, PATH_MAX, "/proc/%ld/gid_map",
1347 snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getgid());
1350 update_map(gid_map, map_path);
1353 /* Close the write end of the pipe, to signal to the child that we
1354 have updated the UID and GID maps */
1356 close(args.pipe_fd[1]);
1358 if (waitpid(child_pid, NULL, 0) == \-1) /* Wait for child */
1362 printf("%s: terminating\en", argv[0]);
1368 .BR newgidmap (1), \" From the shadow package
1369 .BR newuidmap (1), \" From the shadow package
1375 .BR subgid (5), \" From the shadow package
1376 .BR subuid (5), \" From the shadow package
1377 .BR capabilities (7),
1378 .BR cgroup_namespaces (7),
1379 .BR credentials (7),
1381 .BR pid_namespaces (7)
1383 The kernel source file
1384 .IR Documentation/namespaces/resource-control.txt .