{ "mode": "man", "parameter": "user_namespaces", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/user_namespaces/7/json", "generated": "2026-06-17T09:09:03Z", "sections": { "NAME": { "content": "usernamespaces - overview of Linux user namespaces\n", "subsections": [] }, "DESCRIPTION": { "content": "For an overview of namespaces, see namespaces(7).\n\nUser namespaces isolate security-related identifiers and attributes, in particular, user IDs\nand group IDs (see credentials(7)), the root directory, keys (see keyrings(7)), and capabili‐\nties (see capabilities(7)). A process's user and group IDs can be different inside and out‐\nside a user namespace. In particular, a process can have a normal unprivileged user ID out‐\nside a user namespace while at the same time having a user ID of 0 inside the namespace; in\nother words, the process has full privileges for operations inside the user namespace, but is\nunprivileged for operations outside the namespace.\n", "subsections": [ { "name": "Nested namespaces, namespace membership", "content": "User namespaces can be nested; that is, each user namespace—except the initial (\"root\") name‐\nspace—has a parent user namespace, and can have zero or more child user namespaces. The par‐\nent user namespace is the user namespace of the process that creates the user namespace via a\ncall to unshare(2) or clone(2) with the CLONENEWUSER flag.\n\nThe kernel imposes (since version 3.11) a limit of 32 nested levels of user namespaces.\nCalls to unshare(2) or clone(2) that would cause this limit to be exceeded fail with the er‐\nror EUSERS.\n\nEach process is a member of exactly one user namespace. A process created via fork(2) or\nclone(2) without the CLONENEWUSER flag is a member of the same user namespace as its parent.\nA single-threaded process can join another user namespace with setns(2) if it has the\nCAPSYSADMIN in that namespace; upon doing so, it gains a full set of capabilities in that\nnamespace.\n\nA call to clone(2) or unshare(2) with the CLONENEWUSER flag makes the new child process (for\nclone(2)) or the caller (for unshare(2)) a member of the new user namespace created by the\ncall.\n\nThe NSGETPARENT ioctl(2) operation can be used to discover the parental relationship be‐\ntween user namespaces; see ioctlns(2).\n" }, { "name": "Capabilities", "content": "The child process created by clone(2) with the CLONENEWUSER flag starts out with a complete\nset of capabilities in the new user namespace. Likewise, a process that creates a new user\nnamespace using unshare(2) or joins an existing user namespace using setns(2) gains a full\nset of capabilities in that namespace. On the other hand, that process has no capabilities\nin the parent (in the case of clone(2)) or previous (in the case of unshare(2) and setns(2))\nuser namespace, even if the new namespace is created or joined by the root user (i.e., a\nprocess with user ID 0 in the root namespace).\n\nNote that a call to execve(2) will cause a process's capabilities to be recalculated in the\nusual way (see capabilities(7)). Consequently, unless the process has a user ID of 0 within\nthe namespace, or the executable file has a nonempty inheritable capabilities mask, the\nprocess will lose all capabilities. See the discussion of user and group ID mappings, below.\n\nA call to clone(2) or unshare(2) using the CLONENEWUSER flag or a call to setns(2) that\nmoves the caller into another user namespace sets the \"securebits\" flags (see capabili‐‐\nties(7)) to their default values (all flags disabled) in the child (for clone(2)) or caller\n(for unshare(2) or setns(2)). Note that because the caller no longer has capabilities in its\noriginal user namespace after a call to setns(2), it is not possible for a process to reset\nits \"securebits\" flags while retaining its user namespace membership by using a pair of\nsetns(2) calls to move to another user namespace and then return to its original user name‐\nspace.\n\nThe rules for determining whether or not a process has a capability in a particular user\nnamespace are as follows:\n\n1. A process has a capability inside a user namespace if it is a member of that namespace and\nit has the capability in its effective capability set. A process can gain capabilities in\nits effective capability set in various ways. For example, it may execute a set-user-ID\nprogram or an executable with associated file capabilities. In addition, a process may\ngain capabilities via the effect of clone(2), unshare(2), or setns(2), as already de‐\nscribed.\n\n2. If a process has a capability in a user namespace, then it has that capability in all\nchild (and further removed descendant) namespaces as well.\n\n3. When a user namespace is created, the kernel records the effective user ID of the creating\nprocess as being the \"owner\" of the namespace. A process that resides in the parent of\nthe user namespace and whose effective user ID matches the owner of the namespace has all\ncapabilities in the namespace. By virtue of the previous rule, this means that the\nprocess has all capabilities in all further removed descendant user namespaces as well.\nThe NSGETOWNERUID ioctl(2) operation can be used to discover the user ID of the owner\nof the namespace; see ioctlns(2).\n" }, { "name": "Effect of capabilities within a user namespace", "content": "Having a capability inside a user namespace permits a process to perform operations (that re‐\nquire privilege) only on resources governed by that namespace. In other words, having a ca‐\npability in a user namespace permits a process to perform privileged operations on resources\nthat are governed by (nonuser) namespaces owned by (associated with) the user namespace (see\nthe next subsection).\n\nOn the other hand, there are many privileged operations that affect resources that are not\nassociated with any namespace type, for example, changing the system (i.e., calendar) time\n(governed by CAPSYSTIME), loading a kernel module (governed by CAPSYSMODULE), and creat‐\ning a device (governed by CAPMKNOD). Only a process with privileges in the initial user\nnamespace can perform such operations.\n\nHolding CAPSYSADMIN within the user namespace that owns a process's mount namespace allows\nthat process to create bind mounts and mount the following types of filesystems:\n\n* /proc (since Linux 3.8)\n* /sys (since Linux 3.8)\n* devpts (since Linux 3.9)\n* tmpfs(5) (since Linux 3.9)\n* ramfs (since Linux 3.9)\n* mqueue (since Linux 3.9)\n* bpf (since Linux 4.4)\n\nHolding CAPSYSADMIN within the user namespace that owns a process's cgroup namespace allows\n(since Linux 4.6) that process to the mount the cgroup version 2 filesystem and cgroup ver‐\nsion 1 named hierarchies (i.e., cgroup filesystems mounted with the \"none,name=\" option).\n\nHolding CAPSYSADMIN within the user namespace that owns a process's PID namespace allows\n(since Linux 3.8) that process to mount /proc filesystems.\n\nNote however, that mounting block-based filesystems can be done only by a process that holds\nCAPSYSADMIN in the initial user namespace.\n" }, { "name": "Interaction of user namespaces and other types of namespaces", "content": "Starting in Linux 3.8, unprivileged processes can create user namespaces, and the other types\nof namespaces can be created with just the CAPSYSADMIN capability in the caller's user\nnamespace.\n\nWhen a nonuser namespace is created, it is owned by the user namespace in which the creating\nprocess was a member at the time of the creation of the namespace. Privileged operations on\nresources governed by the nonuser namespace require that the process has the necessary capa‐\nbilities in the user namespace that owns the nonuser namespace.\n\nIf CLONENEWUSER is specified along with other CLONENEW* flags in a single clone(2) or un‐‐\nshare(2) call, the user namespace is guaranteed to be created first, giving the child\n(clone(2)) or caller (unshare(2)) privileges over the remaining namespaces created by the\ncall. Thus, it is possible for an unprivileged caller to specify this combination of flags.\n\nWhen a new namespace (other than a user namespace) is created via clone(2) or unshare(2), the\nkernel records the user namespace of the creating process as the owner of the new namespace.\n(This association can't be changed.) When a process in the new namespace subsequently per‐\nforms privileged operations that operate on global resources isolated by the namespace, the\npermission checks are performed according to the process's capabilities in the user namespace\nthat the kernel associated with the new namespace. For example, suppose that a process at‐\ntempts to change the hostname (sethostname(2)), a resource governed by the UTS namespace. In\nthis case, the kernel will determine which user namespace owns the process's UTS namespace,\nand check whether the process has the required capability (CAPSYSADMIN) in that user name‐\nspace.\n\nThe NSGETUSERNS ioctl(2) operation can be used to discover the user namespace that owns a\nnonuser namespace; see ioctlns(2).\n\nUser and group ID mappings: uidmap and gidmap\nWhen a user namespace is created, it starts out without a mapping of user IDs (group IDs) to\nthe parent user namespace. The /proc/[pid]/uidmap and /proc/[pid]/gidmap files (available\nsince Linux 3.5) expose the mappings for user and group IDs inside the user namespace for the\nprocess pid. These files can be read to view the mappings in a user namespace and written to\n(once) to define the mappings.\n\nThe description in the following paragraphs explains the details for uidmap; gidmap is ex‐\nactly the same, but each instance of \"user ID\" is replaced by \"group ID\".\n\nThe uidmap file exposes the mapping of user IDs from the user namespace of the process pid\nto the user namespace of the process that opened uidmap (but see a qualification to this\npoint below). In other words, processes that are in different user namespaces will poten‐\ntially see different values when reading from a particular uidmap file, depending on the\nuser ID mappings for the user namespaces of the reading processes.\n\nEach line in the uidmap file specifies a 1-to-1 mapping of a range of contiguous user IDs\nbetween two user namespaces. (When a user namespace is first created, this file is empty.)\nThe specification in each line takes the form of three numbers delimited by white space. The\nfirst two numbers specify the starting user ID in each of the two user namespaces. The third\nnumber specifies the length of the mapped range. In detail, the fields are interpreted as\nfollows:\n\n(1) The start of the range of user IDs in the user namespace of the process pid.\n\n(2) The start of the range of user IDs to which the user IDs specified by field one map. How\nfield two is interpreted depends on whether the process that opened uidmap and the\nprocess pid are in the same user namespace, as follows:\n\na) If the two processes are in different user namespaces: field two is the start of a\nrange of user IDs in the user namespace of the process that opened uidmap.\n\nb) If the two processes are in the same user namespace: field two is the start of the\nrange of user IDs in the parent user namespace of the process pid. This case enables\nthe opener of uidmap (the common case here is opening /proc/self/uidmap) to see the\nmapping of user IDs into the user namespace of the process that created this user\nnamespace.\n\n(3) The length of the range of user IDs that is mapped between the two user namespaces.\n\nSystem calls that return user IDs (group IDs)—for example, getuid(2), getgid(2), and the cre‐\ndential fields in the structure returned by stat(2)—return the user ID (group ID) mapped into\nthe caller's user namespace.\n\nWhen a process accesses a file, its user and group IDs are mapped into the initial user name‐\nspace for the purpose of permission checking and assigning IDs when creating a file. When a\nprocess retrieves file user and group IDs via stat(2), the IDs are mapped in the opposite di‐\nrection, to produce values relative to the process user and group ID mappings.\n\nThe initial user namespace has no parent namespace, but, for consistency, the kernel provides\ndummy user and group ID mapping files for this namespace. Looking at the uidmap file\n(gidmap is the same) from a shell in the initial namespace shows:\n\n$ cat /proc/$$/uidmap\n0 0 4294967295\n\nThis mapping tells us that the range starting at user ID 0 in this namespace maps to a range\nstarting at 0 in the (nonexistent) parent namespace, and the length of the range is the\nlargest 32-bit unsigned integer. This leaves 4294967295 (the 32-bit signed -1 value) un‐\nmapped. This is deliberate: (uidt) -1 is used in several interfaces (e.g., setreuid(2)) as\na way to specify \"no user ID\". Leaving (uidt) -1 unmapped and unusable guarantees that\nthere will be no confusion when using these interfaces.\n\nDefining user and group ID mappings: writing to uidmap and gidmap\nAfter the creation of a new user namespace, the uidmap file of one of the processes in the\nnamespace may be written to once to define the mapping of user IDs in the new user namespace.\nAn attempt to write more than once to a uidmap file in a user namespace fails with the error\nEPERM. Similar rules apply for gidmap files.\n\nThe lines written to uidmap (gidmap) must conform to the following rules:\n\n* The three fields must be valid numbers, and the last field must be greater than 0.\n\n* Lines are terminated by newline characters.\n\n* There is a limit on the number of lines in the file. In Linux 4.14 and earlier, this\nlimit was (arbitrarily) set at 5 lines. Since Linux 4.15, the limit is 340 lines. In ad‐\ndition, the number of bytes written to the file must be less than the system page size,\nand the write must be performed at the start of the file (i.e., lseek(2) and pwrite(2)\ncan't be used to write to nonzero offsets in the file).\n\n* The range of user IDs (group IDs) specified in each line cannot overlap with the ranges in\nany other lines. In the initial implementation (Linux 3.8), this requirement was satis‐\nfied by a simplistic implementation that imposed the further requirement that the values\nin both field 1 and field 2 of successive lines must be in ascending numerical order,\nwhich prevented some otherwise valid maps from being created. Linux 3.9 and later fix\nthis limitation, allowing any valid set of nonoverlapping maps.\n\n* At least one line must be written to the file.\n\nWrites that violate the above rules fail with the error EINVAL.\n\nIn order for a process to write to the /proc/[pid]/uidmap (/proc/[pid]/gidmap) file, all of\nthe following requirements must be met:\n\n1. The writing process must have the CAPSETUID (CAPSETGID) capability in the user namespace\nof the process pid.\n\n2. The writing process must either be in the user namespace of the process pid or be in the\nparent user namespace of the process pid.\n\n3. The mapped user IDs (group IDs) must in turn have a mapping in the parent user namespace.\n\n4. One of the following two cases applies:\n\n* Either the writing process has the CAPSETUID (CAPSETGID) capability in the parent\nuser namespace.\n\n+ No further restrictions apply: the process can make mappings to arbitrary user IDs\n(group IDs) in the parent user namespace.\n\n* Or otherwise all of the following restrictions apply:\n\n+ The data written to uidmap (gidmap) must consist of a single line that maps the\nwriting process's effective user ID (group ID) in the parent user namespace to a\nuser ID (group ID) in the user namespace.\n\n+ The writing process must have the same effective user ID as the process that created\nthe user namespace.\n\n+ In the case of gidmap, use of the setgroups(2) system call must first be denied by\nwriting \"deny\" to the /proc/[pid]/setgroups file (see below) before writing to\ngidmap.\n\nWrites that violate the above rules fail with the error EPERM.\n" }, { "name": "Interaction with system calls that change process UIDs or GIDs", "content": "In a user namespace where the uidmap file has not been written, the system calls that change\nuser IDs will fail. Similarly, if the gidmap file has not been written, the system calls\nthat change group IDs will fail. After the uidmap and gidmap files have been written, only\nthe mapped values may be used in system calls that change user and group IDs.\n\nFor user IDs, the relevant system calls include setuid(2), setfsuid(2), setreuid(2), and se‐‐\ntresuid(2). For group IDs, the relevant system calls include setgid(2), setfsgid(2), se‐‐\ntregid(2), setresgid(2), and setgroups(2).\n\nWriting \"deny\" to the /proc/[pid]/setgroups file before writing to /proc/[pid]/gidmap will\npermanently disable setgroups(2) in a user namespace and allow writing to /proc/[pid]/gidmap\nwithout having the CAPSETGID capability in the parent user namespace.\n" }, { "name": "The /proc/[pid]/setgroups file", "content": "The /proc/[pid]/setgroups file displays the string \"allow\" if processes in the user namespace\nthat contains the process pid are permitted to employ the setgroups(2) system call; it dis‐\nplays \"deny\" if setgroups(2) is not permitted in that user namespace. Note that regardless\nof the value in the /proc/[pid]/setgroups file (and regardless of the process's capabili‐\nties), calls to setgroups(2) are also not permitted if /proc/[pid]/gidmap has not yet been\nset.\n\nA privileged process (one with the CAPSYSADMIN capability in the namespace) may write ei‐\nther of the strings \"allow\" or \"deny\" to this file before writing a group ID mapping for this\nuser namespace to the file /proc/[pid]/gidmap. Writing the string \"deny\" prevents any\nprocess in the user namespace from employing setgroups(2).\n\nThe essence of the restrictions described in the preceding paragraph is that it is permitted\nto write to /proc/[pid]/setgroups only so long as calling setgroups(2) is disallowed because\n/proc/[pid]/gidmap has not been set. This ensures that a process cannot transition from a\nstate where setgroups(2) is allowed to a state where setgroups(2) is denied; a process can\ntransition only from setgroups(2) being disallowed to setgroups(2) being allowed.\n\nThe default value of this file in the initial user namespace is \"allow\".\n\nOnce /proc/[pid]/gidmap has been written to (which has the effect of enabling setgroups(2)\nin the user namespace), it is no longer possible to disallow setgroups(2) by writing \"deny\"\nto /proc/[pid]/setgroups (the write fails with the error EPERM).\n\nA child user namespace inherits the /proc/[pid]/setgroups setting from its parent.\n\nIf the setgroups file has the value \"deny\", then the setgroups(2) system call can't subse‐\nquently be reenabled (by writing \"allow\" to the file) in this user namespace. (Attempts to\ndo so fail with the error EPERM.) This restriction also propagates down to all child user\nnamespaces of this user namespace.\n\nThe /proc/[pid]/setgroups file was added in Linux 3.19, but was backported to many earlier\nstable kernel series, because it addresses a security issue. The issue concerned files with\npermissions such as \"rwx---rwx\". Such files give fewer permissions to \"group\" than they do\nto \"other\". This means that dropping groups using setgroups(2) might allow a process file\naccess that it did not formerly have. Before the existence of user namespaces this was not a\nconcern, since only a privileged process (one with the CAPSETGID capability) could call set‐‐\ngroups(2). However, with the introduction of user namespaces, it became possible for an un‐\nprivileged process to create a new namespace in which the user had all privileges. This then\nallowed formerly unprivileged users to drop groups and thus gain file access that they did\nnot previously have. The /proc/[pid]/setgroups file was added to address this security is‐\nsue, by denying any pathway for an unprivileged process to drop groups with setgroups(2).\n" }, { "name": "Unmapped user and group IDs", "content": "There are various places where an unmapped user ID (group ID) may be exposed to user space.\nFor example, the first process in a new user namespace may call getuid(2) before a user ID\nmapping has been defined for the namespace. In most such cases, an unmapped user ID is con‐\nverted to the overflow user ID (group ID); the default value for the overflow user ID (group\nID) is 65534. See the descriptions of /proc/sys/kernel/overflowuid and /proc/sys/ker‐\nnel/overflowgid in proc(5).\n\nThe cases where unmapped IDs are mapped in this fashion include system calls that return user\nIDs (getuid(2), getgid(2), and similar), credentials passed over a UNIX domain socket, cre‐\ndentials returned by stat(2), waitid(2), and the System V IPC \"ctl\" IPCSTAT operations, cre‐\ndentials exposed by /proc/[pid]/status and the files in /proc/sysvipc/*, credentials returned\nvia the siuid field in the siginfot received with a signal (see sigaction(2)), credentials\nwritten to the process accounting file (see acct(5)), and credentials returned with POSIX\nmessage queue notifications (see mqnotify(3)).\n\nThere is one notable case where unmapped user and group IDs are not converted to the corre‐\nsponding overflow ID value. When viewing a uidmap or gidmap file in which there is no map‐\nping for the second field, that field is displayed as 4294967295 (-1 as an unsigned integer).\n" }, { "name": "Accessing files", "content": "In order to determine permissions when an unprivileged process accesses a file, the process\ncredentials (UID, GID) and the file credentials are in effect mapped back to what they would\nbe in the initial user namespace and then compared to determine the permissions that the\nprocess has on the file. The same is also of other objects that employ the credentials plus\npermissions mask accessibility model, such as System V IPC objects\n" }, { "name": "Operation of file-related capabilities", "content": "Certain capabilities allow a process to bypass various kernel-enforced restrictions when per‐\nforming operations on files owned by other users or groups. These capabilities are:\nCAPCHOWN, CAPDACOVERRIDE, CAPDACREADSEARCH, CAPFOWNER, and CAPFSETID.\n\nWithin a user namespace, these capabilities allow a process to bypass the rules if the\nprocess has the relevant capability over the file, meaning that:\n\n* the process has the relevant effective capability in its user namespace; and\n\n* the file's user ID and group ID both have valid mappings in the user namespace.\n\nThe CAPFOWNER capability is treated somewhat exceptionally: it allows a process to bypass\nthe corresponding rules so long as at least the file's user ID has a mapping in the user\nnamespace (i.e., the file's group ID does not need to have a valid mapping).\n" }, { "name": "Set-user-ID and set-group-ID programs", "content": "When a process inside a user namespace executes a set-user-ID (set-group-ID) program, the\nprocess's effective user (group) ID inside the namespace is changed to whatever value is\nmapped for the user (group) ID of the file. However, if either the user or the group ID of\nthe file has no mapping inside the namespace, the set-user-ID (set-group-ID) bit is silently\nignored: the new program is executed, but the process's effective user (group) ID is left un‐\nchanged. (This mirrors the semantics of executing a set-user-ID or set-group-ID program that\nresides on a filesystem that was mounted with the MSNOSUID flag, as described in mount(2).)\n" }, { "name": "Miscellaneous", "content": "When a process's user and group IDs are passed over a UNIX domain socket to a process in a\ndifferent user namespace (see the description of SCMCREDENTIALS in unix(7)), they are trans‐\nlated into the corresponding values as per the receiving process's user and group ID map‐\npings.\n" } ] }, "CONFORMING TO": { "content": "Namespaces are a Linux-specific feature.\n", "subsections": [] }, "NOTES": { "content": "Over the years, there have been a lot of features that have been added to the Linux kernel\nthat have been made available only to privileged users because of their potential to confuse\nset-user-ID-root applications. In general, it becomes safe to allow the root user in a user\nnamespace to use those features because it is impossible, while in a user namespace, to gain\nmore privilege than the root user of a user namespace has.\n", "subsections": [ { "name": "Availability", "content": "Use of user namespaces requires a kernel that is configured with the CONFIGUSERNS option.\nUser namespaces require support in a range of subsystems across the kernel. When an unsup‐\nported subsystem is configured into the kernel, it is not possible to configure user name‐\nspaces support.\n\nAs at Linux 3.8, most relevant subsystems supported user namespaces, but a number of filesys‐\ntems did not have the infrastructure needed to map user and group IDs between user name‐\nspaces. Linux 3.9 added the required infrastructure support for many of the remaining unsup‐\nported filesystems (Plan 9 (9P), Andrew File System (AFS), Ceph, CIFS, CODA, NFS, and OCFS2).\nLinux 3.12 added support for the last of the unsupported major filesystems, XFS.\n" } ] }, "EXAMPLES": { "content": "The program below is designed to allow experimenting with user namespaces, as well as other\ntypes of namespaces. It creates namespaces as specified by command-line options and then ex‐\necutes a command inside those namespaces. The comments and usage() function inside the pro‐\ngram provide a full explanation of the program. The following shell session demonstrates its\nuse.\n\nFirst, we look at the run-time environment:\n\n$ uname -rs # Need Linux 3.8 or later\nLinux 3.8.0\n$ id -u # Running as unprivileged user\n1000\n$ id -g\n1000\n\nNow start a new shell in new user (-U), mount (-m), and PID (-p) namespaces, with user ID\n(-M) and group ID (-G) 1000 mapped to 0 inside the user namespace:\n\n$ ./usernschildexec -p -m -U -M '0 1000 1' -G '0 1000 1' bash\n\nThe shell has PID 1, because it is the first process in the new PID namespace:\n\nbash$ echo $$\n1\n\nMounting a new /proc filesystem and listing all of the processes visible in the new PID name‐\nspace shows that the shell can't see any processes outside the PID namespace:\n\nbash$ mount -t proc proc /proc\nbash$ ps ax\nPID TTY STAT TIME COMMAND\n1 pts/3 S 0:00 bash\n22 pts/3 R+ 0:00 ps ax\n\nInside the user namespace, the shell has user and group ID 0, and a full set of permitted and\neffective capabilities:\n\nbash$ cat /proc/$$/status | egrep '^[UG]id'\nUid: 0 0 0 0\nGid: 0 0 0 0\nbash$ cat /proc/$$/status | egrep '^Cap(Prm|Inh|Eff)'\nCapInh: 0000000000000000\nCapPrm: 0000001fffffffff\nCapEff: 0000001fffffffff\n", "subsections": [ { "name": "Program source", "content": "/* usernschildexec.c\n\nLicensed under GNU General Public License v2 or later\n\nCreate a child process that executes a shell command in new\nnamespace(s); allow UID and GID mappings to be specified when\ncreating a user namespace.\n*/\n#define GNUSOURCE\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n/* A simple error-handling function: print an error message based\non the value in 'errno' and terminate the calling process */\n\n#define errExit(msg) do { perror(msg); exit(EXITFAILURE); \\\n} while (0)\n\nstruct childargs {\nchar argv; /* Command to be executed by child, with args */\nint pipefd[2]; /* Pipe used to synchronize parent and child */\n};\n\nstatic int verbose;\n\nstatic void\nusage(char *pname)\n{\nfprintf(stderr, \"Usage: %s [options] cmd [arg...]\\n\\n\", pname);\nfprintf(stderr, \"Create a child process that executes a shell \"\n\"command in a new user namespace,\\n\"\n\"and possibly also other new namespace(s).\\n\\n\");\nfprintf(stderr, \"Options can be:\\n\\n\");\n#define fpe(str) fprintf(stderr, \" %s\", str);\nfpe(\"-i New IPC namespace\\n\");\nfpe(\"-m New mount namespace\\n\");\nfpe(\"-n New network namespace\\n\");\nfpe(\"-p New PID namespace\\n\");\nfpe(\"-u New UTS namespace\\n\");\nfpe(\"-U New user namespace\\n\");\nfpe(\"-M uidmap Specify UID map for user namespace\\n\");\nfpe(\"-G gidmap Specify GID map for user namespace\\n\");\nfpe(\"-z Map user's UID and GID to 0 in user namespace\\n\");\nfpe(\" (equivalent to: -M '0 1' -G '0 1')\\n\");\nfpe(\"-v Display verbose messages\\n\");\nfpe(\"\\n\");\nfpe(\"If -z, -M, or -G is specified, -U is required.\\n\");\nfpe(\"It is not permitted to specify both -z and either -M or -G.\\n\");\nfpe(\"\\n\");\nfpe(\"Map strings for -M and -G consist of records of the form:\\n\");\nfpe(\"\\n\");\nfpe(\" ID-inside-ns ID-outside-ns len\\n\");\nfpe(\"\\n\");\nfpe(\"A map string can contain multiple records, separated\"\n\" by commas;\\n\");\nfpe(\"the commas are replaced by newlines before writing\"\n\" to map files.\\n\");\n\nexit(EXITFAILURE);\n}\n\n/* Update the mapping file 'mapfile', with the value provided in\n'mapping', a string that defines a UID or GID mapping. A UID or\nGID mapping consists of one or more newline-delimited records\nof the form:\n\nIDinside-ns ID-outside-ns length\n\nRequiring the user to supply a string that contains newlines is\nof course inconvenient for command-line use. Thus, we permit the\nuse of commas to delimit records in this string, and replace them\nwith newlines before writing the string to the file. */\n\nstatic void\nupdatemap(char *mapping, char *mapfile)\n{\nint fd;\nsizet maplen; /* Length of 'mapping' */\n\n/* Replace commas in mapping string with newlines */\n\nmaplen = strlen(mapping);\nfor (int j = 0; j < maplen; j++)\nif (mapping[j] == ',')\nmapping[j] = '\\n';\n\nfd = open(mapfile, ORDWR);\nif (fd == -1) {\nfprintf(stderr, \"ERROR: open %s: %s\\n\", mapfile,\nstrerror(errno));\nexit(EXITFAILURE);\n}\n\nif (write(fd, mapping, maplen) != maplen) {\nfprintf(stderr, \"ERROR: write %s: %s\\n\", mapfile,\nstrerror(errno));\nexit(EXITFAILURE);\n}\n\nclose(fd);\n}\n\n/* Linux 3.19 made a change in the handling of setgroups(2) and the\n'gidmap' file to address a security issue. The issue allowed\n*unprivileged* users to employ user namespaces in order to drop\nThe upshot of the 3.19 changes is that in order to update the\n'gidmaps' file, use of the setgroups() system call in this\nuser namespace must first be disabled by writing \"deny\" to one of\nthe /proc/PID/setgroups files for this namespace. That is the\npurpose of the following function. */\n\nstatic void\nprocsetgroupswrite(pidt childpid, char *str)\n{\nchar setgroupspath[PATHMAX];\nint fd;\n\nsnprintf(setgroupspath, PATHMAX, \"/proc/%jd/setgroups\",\n(intmaxt) childpid);\n\nfd = open(setgroupspath, ORDWR);\nif (fd == -1) {\n\n/* We may be on a system that doesn't support\n/proc/PID/setgroups. In that case, the file won't exist,\nand the system won't impose the restrictions that Linux 3.19\nadded. That's fine: we don't need to do anything in order\nto permit 'gidmap' to be updated.\n\nHowever, if the error from open() was something other than\nthe ENOENT error that is expected for that case, let the\nuser know. */\n\nif (errno != ENOENT)\nfprintf(stderr, \"ERROR: open %s: %s\\n\", setgroupspath,\nstrerror(errno));\nreturn;\n}\n\nif (write(fd, str, strlen(str)) == -1)\nfprintf(stderr, \"ERROR: write %s: %s\\n\", setgroupspath,\nstrerror(errno));\n\nclose(fd);\n}\n\nstatic int /* Start function for cloned child */\nchildFunc(void *arg)\n{\nstruct childargs *args = arg;\nchar ch;\n\n/* Wait until the parent has updated the UID and GID mappings.\nSee the comment in main(). We wait for end of file on a\npipe that will be closed by the parent process once it has\nupdated the mappings. */\n\nclose(args->pipefd[1]); /* Close our descriptor for the write\nend of the pipe so that we see EOF\nwhen parent closes its descriptor */\nif (read(args->pipefd[0], &ch, 1) != 0) {\nfprintf(stderr,\n\"Failure in child: read from pipe returned != 0\\n\");\nexit(EXITFAILURE);\n}\n\nclose(args->pipefd[0]);\n\n/* Execute a shell command */\n\nprintf(\"About to exec %s\\n\", args->argv[0]);\nexecvp(args->argv[0], args->argv);\nerrExit(\"execvp\");\n}\n\n#define STACKSIZE (1024 * 1024)\n\nstatic char childstack[STACKSIZE]; /* Space for child's stack */\n\nint\nmain(int argc, char *argv[])\n{\nint flags, opt, mapzero;\npidt childpid;\nstruct childargs args;\nchar *uidmap, *gidmap;\nconst int MAPBUFSIZE = 100;\nchar mapbuf[MAPBUFSIZE];\nchar mappath[PATHMAX];\n\n/* Parse command-line options. The initial '+' character in\nthe final getopt() argument prevents GNU-style permutation\nof command-line options. That's useful, since sometimes\nthe 'command' to be executed by this program itself\nhas command-line options. We don't want getopt() to treat\nthose as options to this program. */\n\nflags = 0;\nverbose = 0;\ngidmap = NULL;\nuidmap = NULL;\nmapzero = 0;\nwhile ((opt = getopt(argc, argv, \"+imnpuUM:G:zv\")) != -1) {\nswitch (opt) {\ncase 'i': flags |= CLONENEWIPC; break;\ncase 'm': flags |= CLONENEWNS; break;\ncase 'n': flags |= CLONENEWNET; break;\ncase 'p': flags |= CLONENEWPID; break;\ncase 'u': flags |= CLONENEWUTS; break;\ncase 'v': verbose = 1; break;\ncase 'z': mapzero = 1; break;\ncase 'M': uidmap = optarg; break;\ncase 'G': gidmap = optarg; break;\ncase 'U': flags |= CLONENEWUSER; break;\ndefault: usage(argv[0]);\n}\n}\n\n/* -M or -G without -U is nonsensical */\n\nif (((uidmap != NULL || gidmap != NULL || mapzero) &&\n!(flags & CLONENEWUSER)) ||\n(mapzero && (uidmap != NULL || gidmap != NULL)))\nusage(argv[0]);\n\nargs.argv = &argv[optind];\n\n/* We use a pipe to synchronize the parent and child, in order to\nensure that the parent sets the UID and GID maps before the child\ncalls execve(). This ensures that the child maintains its\ncapabilities during the execve() in the common case where we\nwant to map the child's effective user ID to 0 in the new user\nnamespace. Without this synchronization, the child would lose\nits capabilities if it performed an execve() with nonzero\nuser IDs (see the capabilities(7) man page for details of the\ntransformation of a process's capabilities during execve()). */\n\nif (pipe(args.pipefd) == -1)\nerrExit(\"pipe\");\n\n/* Create the child in new namespace(s) */\n\nchildpid = clone(childFunc, childstack + STACKSIZE,\nflags | SIGCHLD, &args);\nif (childpid == -1)\nerrExit(\"clone\");\n\n/* Parent falls through to here */\n\nif (verbose)\nprintf(\"%s: PID of child created by clone() is %jd\\n\",\nargv[0], (intmaxt) childpid);\n\n/* Update the UID and GID maps in the child */\n\nif (uidmap != NULL || mapzero) {\nsnprintf(mappath, PATHMAX, \"/proc/%jd/uidmap\",\n(intmaxt) childpid);\nif (mapzero) {\nsnprintf(mapbuf, MAPBUFSIZE, \"0 %jd 1\",\n(intmaxt) getuid());\nuidmap = mapbuf;\n}\nupdatemap(uidmap, mappath);\n}\n\nif (gidmap != NULL || mapzero) {\nprocsetgroupswrite(childpid, \"deny\");\n\nsnprintf(mappath, PATHMAX, \"/proc/%jd/gidmap\",\n(intmaxt) childpid);\nif (mapzero) {\nsnprintf(mapbuf, MAPBUFSIZE, \"0 %ld 1\",\n(intmaxt) getgid());\ngidmap = mapbuf;\n}\nupdatemap(gidmap, mappath);\n}\n\n/* Close the write end of the pipe, to signal to the child that we\nhave updated the UID and GID maps */\n\nclose(args.pipefd[1]);\n\nif (waitpid(childpid, NULL, 0) == -1) /* Wait for child */\nerrExit(\"waitpid\");\n\nif (verbose)\nprintf(\"%s: terminating\\n\", argv[0]);\n\nexit(EXITSUCCESS);\n}\n" } ] }, "SEE ALSO": { "content": "newgidmap(1), newuidmap(1), clone(2), ptrace(2), setns(2), unshare(2), proc(5), subgid(5),\nsubuid(5), capabilities(7), cgroupnamespaces(7), credentials(7), namespaces(7), pidname‐‐\nspaces(7)\n\nThe kernel source file Documentation/namespaces/resource-control.txt.\n", "subsections": [] }, "COLOPHON": { "content": "This page is part of release 5.10 of the Linux man-pages project. A description of the\nproject, information about reporting bugs, and the latest version of this page, can be found\nat https://www.kernel.org/doc/man-pages/.\n\n\n\nLinux 2020-11-01 USERNAMESPACES(7)", "subsections": [] } }, "summary": "usernamespaces - overview of Linux user namespaces", "flags": [], "examples": [ "The program below is designed to allow experimenting with user namespaces, as well as other", "types of namespaces. It creates namespaces as specified by command-line options and then ex‐", "ecutes a command inside those namespaces. The comments and usage() function inside the pro‐", "gram provide a full explanation of the program. The following shell session demonstrates its", "use.", "First, we look at the run-time environment:", "$ uname -rs # Need Linux 3.8 or later", "Linux 3.8.0", "$ id -u # Running as unprivileged user", "1000", "$ id -g", "1000", "Now start a new shell in new user (-U), mount (-m), and PID (-p) namespaces, with user ID", "(-M) and group ID (-G) 1000 mapped to 0 inside the user namespace:", "$ ./usernschildexec -p -m -U -M '0 1000 1' -G '0 1000 1' bash", "The shell has PID 1, because it is the first process in the new PID namespace:", "bash$ echo $$", "Mounting a new /proc filesystem and listing all of the processes visible in the new PID name‐", "space shows that the shell can't see any processes outside the PID namespace:", "bash$ mount -t proc proc /proc", "bash$ ps ax", "PID TTY STAT TIME COMMAND", "1 pts/3 S 0:00 bash", "22 pts/3 R+ 0:00 ps ax", "Inside the user namespace, the shell has user and group ID 0, and a full set of permitted and", "effective capabilities:", "bash$ cat /proc/$$/status | egrep '^[UG]id'", "Uid: 0 0 0 0", "Gid: 0 0 0 0", "bash$ cat /proc/$$/status | egrep '^Cap(Prm|Inh|Eff)'", "CapInh: 0000000000000000", "CapPrm: 0000001fffffffff", "CapEff: 0000001fffffffff", "/* usernschildexec.c", "Licensed under GNU General Public License v2 or later", "Create a child process that executes a shell command in new", "namespace(s); allow UID and GID mappings to be specified when", "creating a user namespace.", "*/", "#define GNUSOURCE", "#include ", "#include ", "#include ", "#include ", "#include ", "#include ", "#include ", "#include ", "#include ", "#include ", "#include ", "/* A simple error-handling function: print an error message based", "on the value in 'errno' and terminate the calling process */", "#define errExit(msg) do { perror(msg); exit(EXITFAILURE); \\", "} while (0)", "struct childargs {", "char argv; /* Command to be executed by child, with args */", "int pipefd[2]; /* Pipe used to synchronize parent and child */", "};", "static int verbose;", "static void", "usage(char *pname)", "fprintf(stderr, \"Usage: %s [options] cmd [arg...]\\n\\n\", pname);", "fprintf(stderr, \"Create a child process that executes a shell \"", "\"command in a new user namespace,\\n\"", "\"and possibly also other new namespace(s).\\n\\n\");", "fprintf(stderr, \"Options can be:\\n\\n\");", "#define fpe(str) fprintf(stderr, \" %s\", str);", "fpe(\"-i New IPC namespace\\n\");", "fpe(\"-m New mount namespace\\n\");", "fpe(\"-n New network namespace\\n\");", "fpe(\"-p New PID namespace\\n\");", "fpe(\"-u New UTS namespace\\n\");", "fpe(\"-U New user namespace\\n\");", "fpe(\"-M uidmap Specify UID map for user namespace\\n\");", "fpe(\"-G gidmap Specify GID map for user namespace\\n\");", "fpe(\"-z Map user's UID and GID to 0 in user namespace\\n\");", "fpe(\" (equivalent to: -M '0 1' -G '0 1')\\n\");", "fpe(\"-v Display verbose messages\\n\");", "fpe(\"\\n\");", "fpe(\"If -z, -M, or -G is specified, -U is required.\\n\");", "fpe(\"It is not permitted to specify both -z and either -M or -G.\\n\");", "fpe(\"\\n\");", "fpe(\"Map strings for -M and -G consist of records of the form:\\n\");", "fpe(\"\\n\");", "fpe(\" ID-inside-ns ID-outside-ns len\\n\");", "fpe(\"\\n\");", "fpe(\"A map string can contain multiple records, separated\"", "\" by commas;\\n\");", "fpe(\"the commas are replaced by newlines before writing\"", "\" to map files.\\n\");", "exit(EXITFAILURE);", "/* Update the mapping file 'mapfile', with the value provided in", "'mapping', a string that defines a UID or GID mapping. A UID or", "GID mapping consists of one or more newline-delimited records", "of the form:", "IDinside-ns ID-outside-ns length", "Requiring the user to supply a string that contains newlines is", "of course inconvenient for command-line use. Thus, we permit the", "use of commas to delimit records in this string, and replace them", "with newlines before writing the string to the file. */", "static void", "updatemap(char *mapping, char *mapfile)", "int fd;", "sizet maplen; /* Length of 'mapping' */", "/* Replace commas in mapping string with newlines */", "maplen = strlen(mapping);", "for (int j = 0; j < maplen; j++)", "if (mapping[j] == ',')", "mapping[j] = '\\n';", "fd = open(mapfile, ORDWR);", "if (fd == -1) {", "fprintf(stderr, \"ERROR: open %s: %s\\n\", mapfile,", "strerror(errno));", "exit(EXITFAILURE);", "if (write(fd, mapping, maplen) != maplen) {", "fprintf(stderr, \"ERROR: write %s: %s\\n\", mapfile,", "strerror(errno));", "exit(EXITFAILURE);", "close(fd);", "/* Linux 3.19 made a change in the handling of setgroups(2) and the", "'gidmap' file to address a security issue. The issue allowed", "*unprivileged* users to employ user namespaces in order to drop", "The upshot of the 3.19 changes is that in order to update the", "'gidmaps' file, use of the setgroups() system call in this", "user namespace must first be disabled by writing \"deny\" to one of", "the /proc/PID/setgroups files for this namespace. That is the", "purpose of the following function. */", "static void", "procsetgroupswrite(pidt childpid, char *str)", "char setgroupspath[PATHMAX];", "int fd;", "snprintf(setgroupspath, PATHMAX, \"/proc/%jd/setgroups\",", "(intmaxt) childpid);", "fd = open(setgroupspath, ORDWR);", "if (fd == -1) {", "/* We may be on a system that doesn't support", "/proc/PID/setgroups. In that case, the file won't exist,", "and the system won't impose the restrictions that Linux 3.19", "added. That's fine: we don't need to do anything in order", "to permit 'gidmap' to be updated.", "However, if the error from open() was something other than", "the ENOENT error that is expected for that case, let the", "user know. */", "if (errno != ENOENT)", "fprintf(stderr, \"ERROR: open %s: %s\\n\", setgroupspath,", "strerror(errno));", "return;", "if (write(fd, str, strlen(str)) == -1)", "fprintf(stderr, \"ERROR: write %s: %s\\n\", setgroupspath,", "strerror(errno));", "close(fd);", "static int /* Start function for cloned child */", "childFunc(void *arg)", "struct childargs *args = arg;", "char ch;", "/* Wait until the parent has updated the UID and GID mappings.", "See the comment in main(). We wait for end of file on a", "pipe that will be closed by the parent process once it has", "updated the mappings. */", "close(args->pipefd[1]); /* Close our descriptor for the write", "end of the pipe so that we see EOF", "when parent closes its descriptor */", "if (read(args->pipefd[0], &ch, 1) != 0) {", "fprintf(stderr,", "\"Failure in child: read from pipe returned != 0\\n\");", "exit(EXITFAILURE);", "close(args->pipefd[0]);", "/* Execute a shell command */", "printf(\"About to exec %s\\n\", args->argv[0]);", "execvp(args->argv[0], args->argv);", "errExit(\"execvp\");", "#define STACKSIZE (1024 * 1024)", "static char childstack[STACKSIZE]; /* Space for child's stack */", "int", "main(int argc, char *argv[])", "int flags, opt, mapzero;", "pidt childpid;", "struct childargs args;", "char *uidmap, *gidmap;", "const int MAPBUFSIZE = 100;", "char mapbuf[MAPBUFSIZE];", "char mappath[PATHMAX];", "/* Parse command-line options. The initial '+' character in", "the final getopt() argument prevents GNU-style permutation", "of command-line options. That's useful, since sometimes", "the 'command' to be executed by this program itself", "has command-line options. We don't want getopt() to treat", "those as options to this program. */", "flags = 0;", "verbose = 0;", "gidmap = NULL;", "uidmap = NULL;", "mapzero = 0;", "while ((opt = getopt(argc, argv, \"+imnpuUM:G:zv\")) != -1) {", "switch (opt) {", "case 'i': flags |= CLONENEWIPC; break;", "case 'm': flags |= CLONENEWNS; break;", "case 'n': flags |= CLONENEWNET; break;", "case 'p': flags |= CLONENEWPID; break;", "case 'u': flags |= CLONENEWUTS; break;", "case 'v': verbose = 1; break;", "case 'z': mapzero = 1; break;", "case 'M': uidmap = optarg; break;", "case 'G': gidmap = optarg; break;", "case 'U': flags |= CLONENEWUSER; break;", "default: usage(argv[0]);", "/* -M or -G without -U is nonsensical */", "if (((uidmap != NULL || gidmap != NULL || mapzero) &&", "!(flags & CLONENEWUSER)) ||", "(mapzero && (uidmap != NULL || gidmap != NULL)))", "usage(argv[0]);", "args.argv = &argv[optind];", "/* We use a pipe to synchronize the parent and child, in order to", "ensure that the parent sets the UID and GID maps before the child", "calls execve(). This ensures that the child maintains its", "capabilities during the execve() in the common case where we", "want to map the child's effective user ID to 0 in the new user", "namespace. Without this synchronization, the child would lose", "its capabilities if it performed an execve() with nonzero", "user IDs (see the capabilities(7) man page for details of the", "transformation of a process's capabilities during execve()). */", "if (pipe(args.pipefd) == -1)", "errExit(\"pipe\");", "/* Create the child in new namespace(s) */", "childpid = clone(childFunc, childstack + STACKSIZE,", "flags | SIGCHLD, &args);", "if (childpid == -1)", "errExit(\"clone\");", "/* Parent falls through to here */", "if (verbose)", "printf(\"%s: PID of child created by clone() is %jd\\n\",", "argv[0], (intmaxt) childpid);", "/* Update the UID and GID maps in the child */", "if (uidmap != NULL || mapzero) {", "snprintf(mappath, PATHMAX, \"/proc/%jd/uidmap\",", "(intmaxt) childpid);", "if (mapzero) {", "snprintf(mapbuf, MAPBUFSIZE, \"0 %jd 1\",", "(intmaxt) getuid());", "uidmap = mapbuf;", "updatemap(uidmap, mappath);", "if (gidmap != NULL || mapzero) {", "procsetgroupswrite(childpid, \"deny\");", "snprintf(mappath, PATHMAX, \"/proc/%jd/gidmap\",", "(intmaxt) childpid);", "if (mapzero) {", "snprintf(mapbuf, MAPBUFSIZE, \"0 %ld 1\",", "(intmaxt) getgid());", "gidmap = mapbuf;", "updatemap(gidmap, mappath);", "/* Close the write end of the pipe, to signal to the child that we", "have updated the UID and GID maps */", "close(args.pipefd[1]);", "if (waitpid(childpid, NULL, 0) == -1) /* Wait for child */", "errExit(\"waitpid\");", "if (verbose)", "printf(\"%s: terminating\\n\", argv[0]);", "exit(EXITSUCCESS);" ], "see_also": [ { "name": "newgidmap", "section": "1", "url": "https://www.chedong.com/phpMan.php/man/newgidmap/1/json" }, { "name": "newuidmap", "section": "1", "url": "https://www.chedong.com/phpMan.php/man/newuidmap/1/json" }, { "name": "clone", "section": "2", "url": "https://www.chedong.com/phpMan.php/man/clone/2/json" }, { "name": "ptrace", "section": "2", "url": "https://www.chedong.com/phpMan.php/man/ptrace/2/json" }, { "name": "setns", "section": "2", "url": "https://www.chedong.com/phpMan.php/man/setns/2/json" }, { "name": "unshare", "section": "2", "url": "https://www.chedong.com/phpMan.php/man/unshare/2/json" }, { "name": "proc", "section": "5", "url": "https://www.chedong.com/phpMan.php/man/proc/5/json" }, { "name": "subgid", "section": "5", "url": "https://www.chedong.com/phpMan.php/man/subgid/5/json" }, { "name": "subuid", "section": "5", "url": "https://www.chedong.com/phpMan.php/man/subuid/5/json" }, { "name": "capabilities", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/capabilities/7/json" }, { "name": "cgroupnamespaces", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/cgroupnamespaces/7/json" }, { "name": "credentials", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/credentials/7/json" }, { "name": "namespaces", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/namespaces/7/json" }, { "name": "spaces", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/spaces/7/json" } ] }