Hi Eric et al.,
All: The attached page aims to provide a fairly complete overview of
user namespaces. I'm looking for review comments (corrections,
improvements, additions, etc.) on this man page. I've provided it in
two forms inline below, and reviewers can comment on whichever form
they are most comfortable with:
1) The rendered page as plain text
2) The *roff source (also attached); rendering that source will enable
readers to see proper formatting for the page.
Note that the namespaces(7) page referred to in this page is not yet
finished; I'll send it out for review at a future time.
Main change since v1 is to address Serge's comments here:
http://thread.gmane.org/gmane.linux.man/3745/focus=1457720
Cheers,
Michael
=====
USER_NAMESPACES(7) Linux Programmer's Manual USER_NAMESPACES(7)
NAME
user_namespaces - overview of Linux user_namespaces
DESCRIPTION
For an overview of namespaces, see namespaces(7).
User namespaces isolate security-related identifiers, in par‐
ticular, user IDs and group IDs (see credentials(7), keys (see
keyctl(2)), and capabilities (see capabilities(7)). A
process's user and group IDs can be different inside and out‐
side a user namespace. In particular, a process can have a
normal unprivileged user ID outside a user namespace while at
the same time having a user ID of 0 inside the namespace; in
other words, the process has full privileges for operations
inside the user namespace, but is unprivileged for operations
outside the namespace.
Nested namespaces, namespace membership
User namespaces can be nested; that is, each user namespace—
except the initial ("root") namespace—has a parent user names‐
pace, and can have zero or more child user namespaces. The
parent user namespace is the user namespace of the process that
creates the user namespace via a call to unshare(2) or clone(2)
with the CLONE_NEWUSER flag.
Each process is a member of exactly one user namespace. A
process created via fork(2) or clone(2) without the
CLONE_NEWUSER flag is a member of the same user namespace as
its parent. A process can join another user namespace with
setns(2) if it has the CAP_SYS_ADMIN in that namespace; upon
doing so, it gains a full set of capabilities in that names‐
pace.
A call to clone(2) or unshare(2) with the CLONE_NEWUSER flag
makes the new child process (for clone(2)) or the caller (for
unshare(2)) a member of the new user namespace created by the
call.
Capabilities
The child process created by clone(2) with the CLONE_NEWUSER
flag starts out with a complete set of capabilities in the new
user namespace. Likewise, a process that creates a new user
namespace using unshare(2) or joins an existing user namespace
using setns(2) gains a full set of capabilities in that names‐
pace. On the other hand, that process has no capabilities in
the parent (in the case of clone(2)) or previous (in the case
of unshare(2) and setns(2)) user namespace, even if the new
namespace is created or joined by the root user (i.e., a
process with user ID 0 in the root namespace). Nevertheless, a
process owned by the root user will be able to access resources
such as files that are owned by user ID 0, and will be able to
do things such as sending signals to processes belonging to
user ID 0.
Note that a call to execve(2) will cause a process to lose any
capabilities that it has, unless it has a user ID of 0 within
the namespace. Thus, before calling execve(2), a user ID map‐
ping for ID 0 must be defined, and the caller may also need to
use setuid(2) or similar to set its user ID to 0.
A call to clone(2), unshare(2), or setns(2) using the
CLONE_NEWUSER flag sets the "securebits" flags (see capabili‐
ties(7)) to their default values (all flags disabled) in the
child (for clone(2)) or caller (for unshare(2), or setns(2)).
Note that because the caller no longer has capabilities in its
original user namespace after a call to setns(2), it is not
possible for a process to reset its "securebits" flags while
retaining its user namespace membership by using a pair of
setns(2) calls to move to another user namespace and then
return to its original user namespace.
Having a capability inside a user namespace permits a process
to perform operations (that require privilege) only on
resources governed by that namespace. The rules for determin‐
ing whether or not a process has a capability in a particular
user namespace are as follows:
1. A process has a capability inside a user namespace if it is
a member of that namespace and it has the capability in its
effective capability set. A process can gain capabilities
in its effective capability set in various ways. For exam‐
ple, it may execute a set-user-ID program or an executable
with associated file capabilities. In addition, a process
may gain capabilities via the effect of clone(2),
unshare(2), or setns(2), as already described.
2. If a process has a capability in a user namespace, then it
has that capability in all child (and further removed
descendant) namespaces as well.
3. When a user namespace is created, the kernel records the
effective user ID of the creating process as being the
"owner" of the namespace. A process that resides in the
parent of the user namespace and whose effective user ID
matches the owner of the namespace has all capabilities in
the namespace. By virtue of the previous rule, this means
that the process has all capabilities in all further removed
descendant user namespaces as well.
Interaction of user namespaces and other types of namespaces
Starting in Linux 3.8, unprivileged processes can create user
namespaces, and mount, PID, IPC, network, and UTS namespaces
can be created with just the CAP_SYS_ADMIN capability in the
caller's user namespace.
If CLONE_NEWUSER is specified along with other CLONE_NEW* flags
in a single clone(2) or unshare(2) call, the user namespace is
guaranteed to be created first, giving the child (clone(2)) or
caller (unshare(2)) privileges over the remaining namespaces
created by the call. Thus, it is possible for an unprivileged
caller to specify this combination of flags.
When a new IPC, mount, network, PID, or UTS namespace is cre‐
ated via clone(2) or unshare(2), the kernel records the user
namespace of the creating process against the new namespace.
(This association can't be changed.) When a process in the new
namespace subsequently performs privileged operations that
operate on global resources isolated by the namespace, the per‐
mission checks are performed according to the process's capa‐
bilities in the user namespace that the kernel associated with
the new namespace.
User and group ID mappings: uid_map and gid_map
When a user namespace is created, it starts out without a map‐
ping of user IDs (group IDs) to the parent user namespace. The
/proc/[pid]/uid_map and /proc/[pid]/gid_map files (available
since Linux 3.5) expose the mappings for user and group IDs
inside the user namespace for the process pid. These files can
be read to view the mappings in a user namespace and written to
(once) to define the mappings.
The description in the following paragraphs explains the
details for uid_map; gid_map is exactly the same, but each
instance of "user ID" is replaced by "group ID".
The uid_map file exposes the mapping of user IDs from the user
namespace of the process pid to the user namespace of the
process that opened uid_map (but see a qualification to this
point below). In other words, processes that are in different
user namespaces will potentially see different values when
reading from a particular uid_map file, depending on the user
ID mappings for the user namespaces of the reading processes.
Each line in the uid_map file specifies a 1-to-1 mapping of a
range of contiguous user IDs between two user namespaces.
(When a user namespace is first created, this file is empty.)
The specification in each line takes the form of three numbers
delimited by white space. The first two numbers specify the
starting user ID in each of the two user namespaces. The third
number specifies the length of the mapped range. In detail,
the fields are interpreted as follows:
(1) The start of the range of user IDs in the user namespace of
the process pid.
(2) The start of the range of user IDs to which the user IDs
specified by field one map. How field two is interpreted
depends on whether the process that opened uid_map and the
process pid are in the same user namespace, as follows:
a) If the two processes are in different user namespaces:
field two is the start of a range of user IDs in the
user namespace of the process that opened uid_map.
b) If the two processes are in the same user namespace:
field two is the start of the range of user IDs in the
parent user namespace of the process pid. This case
enables the opener of uid_map (the common case here is
opening /proc/self/uid_map) to see the mapping of user
IDs into the user namespace of the process that created
this user namespace.
(3) The length of the range of user IDs that is mapped between
the two user namespaces.
System calls that return user IDs (group IDs)—for example,
getuid(2), getgid(2), and the credential fields in the struc‐
ture returned by stat(2)—return the user ID (group ID) mapped
into the caller's user namespace.
When a process accesses a file, its user and group IDs are
mapped into the initial user namespace for the purpose of per‐
mission checking and assigning IDs when creating a file. When
a process retrieves file user and group IDs via stat(2), the
IDs are mapped in the opposite direction, to produce values
relative to the process user and group ID mappings.
The initial user namespace has no parent namespace, but, for
consistency, the kernel provides dummy user and group ID map‐
ping files for this namespace. Looking at the uid_map file
(gid_map is the same) from a shell in the initial namespace
shows:
$ cat /proc/$$/uid_map
0 0 4294967295
This mapping tells us that the range starting at user ID 0 in
this namespace maps to a range starting at 0 in the (nonexis‐
tent) parent namespace, and the length of the range is the
largest 32-bit unsigned integer.
Defining user and group ID mappings: writing to uid_map and gid_map
After the creation of a new user namespace, the uid_map file of
one of the processes in the namespace may be written to once to
define the mapping of user IDs in the new user namespace. An
attempt to write more than once to a uid_map file in a user
namespace fails with the error EPERM. Similar rules apply for
gid_map files.
The lines written to uid_map (gid_map) must conform to the fol‐
lowing rules:
* The three fields must be valid numbers, and the last field
must be greater than 0.
* Lines are terminated by newline characters.
* There is an (arbitrary) limit on the number of lines in the
file. As at Linux 3.8, the limit is five lines. In addi‐
tion, the number of bytes written to the file must be less
than the system page size, and the write must be performed
at the start of the file (i.e., lseek(2) and pwrite(2) can't
be used to write to nonzero offsets in the file).
* The range of user IDs (group IDs) specified in each line
cannot overlap with the ranges in any other lines. In the
initial implementation (Linux 3.8), this requirement was
satisfied by a simplistic implementation that imposed the
further requirement that the values in both field 1 and
field 2 of successive lines must be in ascending numerical
order, which prevented some otherwise valid maps from being
created. Linux 3.9 and later fix this limitation, allowing
any valid set of nonoverlapping maps.
* At least one line must be written to the file.
Writes that violate the above rules fail with the error EINVAL.
In order for a process to write to the /proc/[pid]/uid_map
(/proc/[pid]/gid_map) file, all of the following requirements
must be met:
1. The writing process must have the CAP_SETUID (CAP_SETGID)
capability in the user namespace of the process pid.
2. The writing process must be in either the user namespace of
the process pid or inside the parent user namespace of the
process pid.
3. The mapped user IDs (group IDs) must in turn have a mapping
in the parent user namespace.
4. One of the following is true:
* The data written to uid_map (gid_map) consists of a sin‐
gle line that maps the writing process's file system user
ID (group ID) in the parent user namespace to a user ID
(group ID) in the user namespace. The usual case here is
that this single line provides a mapping for user ID of
the process that created the namespace.
* The process has the CAP_SETUID (CAP_SETGID) capability in
the parent user namespace. Thus, a privileged process
can make mappings to arbitrary user IDs (group IDs) in
the parent user namespace.
Writes that violate the above rules fail with the error EPERM.
Unmapped user and group IDs
There are various places where an unmapped user ID (group ID)
may be exposed to user space. For example, the first process
in a new user namespace may call getuid() before a user ID map‐
ping has been defined for the namespace. In most such cases,
an unmapped user ID is converted to the overflow user ID (group
ID); the default value for the overflow user ID (group ID) is
65534. See the descriptions of /proc/sys/kernel/overflowuid
and /proc/sys/kernel/overflowgid in proc(5).
The cases where unmapped IDs are mapped in this fashion include
system calls that return user IDs (getuid(2) getgid(2), and
similar), credentials passed over a UNIX domain socket, creden‐
tials returned by stat(2), waitid(2), and the System V IPC
"ctl" IPC_STAT operations, credentials exposed by
/proc/PID/status and the files in /proc/sysvipc/*, credentials
returned via the si_uid field in the siginfo_t received with a
signal (see sigaction(2)), credentials written to the process
accounting file (see acct(5)), and credentials returned with
POSIX message queue notifications (see mq_notify(3)).
There is one notable case where unmapped user and group IDs are
not converted to the corresponding overflow ID value. When
viewing a uid_map or gid_map file in which there is no mapping
for the second field, that field is displayed as 4294967295 (-1
as an unsigned integer);
Set-user-ID and set-group-ID programs
When a process inside a user namespace executes a set-user-ID
(set-group-ID) program, the process's effective user (group) ID
inside the namespace is changed to whatever value is mapped for
the user (group) ID of the file. However, if either the user
or the group ID of the file has no mapping inside the names‐
pace, the set-user-ID (set-group-ID) bit is silently ignored:
the new program is executed, but the process's effective user
(group) ID is left unchanged. (This mirrors the semantics of
executing a set-user-ID or set-group-ID program that resides on
a file system that was mounted with the MS_NOSUID flag, as
described in mount(2).)
Miscellaneous
When a process's user and group IDs are passed over a UNIX
domain socket to a process in a different user namespace (see
the description of SCM_CREDENTIALS in unix(7)), they are trans‐
lated into the corresponding values as per the receiving
process's user and group ID mappings.
CONFORMING TO
Namespaces are a Linux-specific feature.
NOTES
Over the years, there have been a lot of features that have
been added to the Linux kernel that have been made available
only to privileged users because of their potential to confuse
set-user-ID-root applications. In general, it becomes safe to
allow the root user in a user namespace to use those features
because it is impossible, while in a user namespace, to gain
more privilege than the root user of a user namespace has.
Availability
Use of user namespaces requires a kernel that is configured
with the CONFIG_USER_NS option. User namespaces require sup‐
port in a range of subsystems across the kernel. When an
unsupported subsystem is configured into the kernel, it is not
possible to configure user namespaces support. As at Linux
3.8, most relevant subsystems support user namespaces, but
there are a number of file systems that do not. Linux 3.9
added user namespaces support for many of the remaining unsup‐
ported file systems: Plan 9 (9P), Andrew File System (AFS),
Ceph, CIFS, CODA, NFS, and OCFS2. XFS support for user names‐
paces is not yet available.
EXAMPLE
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 exe‐
cutes a command inside those namespaces. The comments and
usage() function inside the program 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:
$ ./userns_child_exec -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 $$
1
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
Mounting a new /proc file system and listing all of the pro‐
cesses visible in the new PID namespace 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
Program source
/* userns_child_exec.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 _GNU_SOURCE
#include <sched.h>
#include <unistd.h>
#include <stdlib.h>
#include <sys/wait.h>
#include <signal.h>
#include <fcntl.h>
#include <stdio.h>
#include <string.h>
#include <limits.h>
#include <errno.h>
/* 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(EXIT_FAILURE); \
} while (0)
struct child_args {
char **argv; /* Command to be executed by child, with args */
int pipe_fd[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 uid_map Specify UID map for user namespace\n");
fpe("-G gid_map 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 <uid> 1' -G '0 <gid> 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(EXIT_FAILURE);
}
/* Update the mapping file 'map_file', 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:
ID_inside-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
update_map(char *mapping, char *map_file)
{
int fd, j;
size_t map_len; /* Length of 'mapping' */
/* Replace commas in mapping string with newlines */
map_len = strlen(mapping);
for (j = 0; j < map_len; j++)
if (mapping[j] == ',')
mapping[j] = '\n';
fd = open(map_file, O_RDWR);
if (fd == -1) {
fprintf(stderr, "ERROR: open %s: %s\n", map_file,
strerror(errno));
return;
//exit(EXIT_FAILURE);
}
if (write(fd, mapping, map_len) != map_len) {
fprintf(stderr, "ERROR: write %s: %s\n", map_file,
strerror(errno));
//exit(EXIT_FAILURE);
}
close(fd);
}
static int /* Start function for cloned child */
childFunc(void *arg)
{
struct child_args *args = (struct child_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->pipe_fd[1]); /* Close our descriptor for the write
end of the pipe so that we see EOF
when parent closes its descriptor */
if (read(args->pipe_fd[0], &ch, 1) != 0) {
fprintf(stderr,
"Failure in child: read from pipe returned != 0\n");
exit(EXIT_FAILURE);
}
/* Execute a shell command */
printf("About to exec %s\n", args->argv[0]);
execvp(args->argv[0], args->argv);
errExit("execvp");
}
#define STACK_SIZE (1024 * 1024)
static char child_stack[STACK_SIZE]; /* Space for child's stack */
int
main(int argc, char *argv[])
{
int flags, opt, map_zero;
pid_t child_pid;
struct child_args args;
char *uid_map, *gid_map;
const int MAP_BUF_SIZE = 100;
char map_buf[MAP_BUF_SIZE];
char map_path[PATH_MAX];
/* 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;
gid_map = NULL;
uid_map = NULL;
map_zero = 0;
while ((opt = getopt(argc, argv, "+imnpuUM:G:zv")) != -1) {
switch (opt) {
case 'i': flags |= CLONE_NEWIPC; break;
case 'm': flags |= CLONE_NEWNS; break;
case 'n': flags |= CLONE_NEWNET; break;
case 'p': flags |= CLONE_NEWPID; break;
case 'u': flags |= CLONE_NEWUTS; break;
case 'v': verbose = 1; break;
case 'z': map_zero = 1; break;
case 'M': uid_map = optarg; break;
case 'G': gid_map = optarg; break;
case 'U': flags |= CLONE_NEWUSER; break;
default: usage(argv[0]);
}
}
/* -M or -G without -U is nonsensical */
if (((uid_map != NULL || gid_map != NULL || map_zero) &&
!(flags & CLONE_NEWUSER)) ||
(map_zero && (uid_map != NULL || gid_map != 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.pipe_fd) == -1)
errExit("pipe");
/* Create the child in new namespace(s) */
child_pid = clone(childFunc, child_stack + STACK_SIZE,
flags | SIGCHLD, &args);
if (child_pid == -1)
errExit("clone");
/* Parent falls through to here */
if (verbose)
printf("%s: PID of child created by clone() is %ld\n",
argv[0], (long) child_pid);
/* Update the UID and GID maps in the child */
if (uid_map != NULL || map_zero) {
snprintf(map_path, PATH_MAX, "/proc/%ld/uid_map",
(long) child_pid);
if (map_zero) {
snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getuid());
uid_map = map_buf;
}
update_map(uid_map, map_path);
}
if (gid_map != NULL || map_zero) {
snprintf(map_path, PATH_MAX, "/proc/%ld/gid_map",
(long) child_pid);
if (map_zero) {
snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getgid());
gid_map = map_buf;
}
update_map(gid_map, map_path);
}
/* Close the write end of the pipe, to signal to the child that we
have updated the UID and GID maps */
close(args.pipe_fd[1]);
if (waitpid(child_pid, NULL, 0) == -1) /* Wait for child */
errExit("waitpid");
if (verbose)
printf("%s: terminating\n", argv[0]);
exit(EXIT_SUCCESS);
}
SEE ALSO
newgidmap(1), newuidmap(1), clone(2), setns(2), unshare(2),
proc(5), subgid(5), subuid(5), credentials(7), capabilities(7),
namespaces(7), pid_namespaces(7)
The kernel source file Documentation/namespaces/resource-con‐
trol.txt.
Linux 2013-01-14 USER_NAMESPACES(7)
========== *roff source ==========
.\" Copyright (c) 2013 by Michael Kerrisk <[email protected]>
.\" and Copyright (c) 2012 by Eric W. Biederman <[email protected]>
.\"
.\" Permission is granted to make and distribute verbatim copies of this
.\" manual provided the copyright notice and this permission notice are
.\" preserved on all copies.
.\"
.\" Permission is granted to copy and distribute modified versions of this
.\" manual under the conditions for verbatim copying, provided that the
.\" entire resulting derived work is distributed under the terms of a
.\" permission notice identical to this one.
.\"
.\" Since the Linux kernel and libraries are constantly changing, this
.\" manual page may be incorrect or out-of-date. The author(s) assume no
.\" responsibility for errors or omissions, or for damages resulting from
.\" the use of the information contained herein. The author(s) may not
.\" have taken the same level of care in the production of this manual,
.\" which is licensed free of charge, as they might when working
.\" professionally.
.\"
.\" Formatted or processed versions of this manual, if unaccompanied by
.\" the source, must acknowledge the copyright and authors of this work.
.\"
.\"
.TH USER_NAMESPACES 7 2013-01-14 "Linux" "Linux Programmer's Manual"
.SH NAME
user_namespaces \- overview of Linux user_namespaces
.SH DESCRIPTION
For an overview of namespaces, see
.BR namespaces (7).
User namespaces isolate security-related identifiers, in particular,
user IDs and group IDs (see
.BR credentials (7),
keys (see
.BR keyctl (2)),
.\" FIXME: This page says very little about the interaction
.\" of user namespaces and keys. Add something on this topic.
and capabilities (see
.BR capabilities (7)).
A process's user and group IDs can be different
inside and outside a user namespace.
In particular,
a process can have a normal unprivileged user ID outside a user namespace
while at the same time having a user ID of 0 inside the namespace;
in other words,
the process has full privileges for operations inside the user namespace,
but is unprivileged for operations outside the namespace.
.\"
.\" ============================================================
.\"
.SS Nested namespaces, namespace membership
User namespaces can be nested;
that is, each user namespace\(emexcept the initial ("root")
namespace\(emhas a parent user namespace,
and can have zero or more child user namespaces.
The parent user namespace is the user namespace
of the process that creates the user namespace via a call to
.BR unshare (2)
or
.BR clone (2)
with the
.BR CLONE_NEWUSER
flag.
Each process is a member of exactly one user namespace.
A process created via
.BR fork (2)
or
.BR clone (2)
without the
.BR CLONE_NEWUSER
flag is a member of the same user namespace as its parent.
A process can join another user namespace with
.BR setns (2)
if it has the
.BR CAP_SYS_ADMIN
in that namespace;
upon doing so, it gains a full set of capabilities in that namespace.
A call to
.BR clone (2)
or
.BR unshare (2)
with the
.BR CLONE_NEWUSER
flag makes the new child process (for
.BR clone (2))
or the caller (for
.BR unshare (2))
a member of the new user namespace created by the call.
.\"
.\" ============================================================
.\"
.SS Capabilities
The child process created by
.BR clone (2)
with the
.BR CLONE_NEWUSER
flag starts out with a complete set
of capabilities in the new user namespace.
Likewise, a process that creates a new user namespace using
.BR unshare (2)
or joins an existing user namespace using
.BR setns (2)
gains a full set of capabilities in that namespace.
On the other hand,
that process has no capabilities in the parent (in the case of
.BR clone (2))
or previous (in the case of
.BR unshare (2)
and
.BR setns (2))
user namespace,
even if the new namespace is created or joined by the root user
(i.e., a process with user ID 0 in the root namespace).
Nevertheless, a process owned by the root user
will be able to access resources such as
files that are owned by user ID 0,
and will be able to do things such as sending signals
to processes belonging to user ID 0.
Note that a call to
.BR execve (2)
will cause a process to lose any capabilities that it has,
unless it has a user ID of 0 within the namespace.
Thus, before calling
.BR execve (2),
a user ID mapping for ID 0 must be defined,
and the caller may also need to use
.BR setuid (2)
or similar to set its user ID to 0.
A call to
.BR clone (2),
.BR unshare (2),
or
.BR setns (2)
using the
.BR CLONE_NEWUSER
flag sets the "securebits" flags
(see
.BR capabilities (7))
to their default values (all flags disabled) in the child (for
.BR clone (2))
or caller (for
.BR unshare (2),
or
.BR setns (2)).
Note that because the caller no longer has capabilities
in its original user namespace after a call to
.BR setns (2),
it is not possible for a process to reset its "securebits" flags while
retaining its user namespace membership by using a pair of
.BR setns (2)
calls to move to another user namespace and then return to
its original user namespace.
Having a capability inside a user namespace
permits a process to perform operations (that require privilege)
only on resources governed by that namespace.
The rules for determining whether or not a process has a capability
in a particular user namespace are as follows:
.IP 1. 3
A process has a capability inside a user namespace
if it is a member of that namespace and
it has the capability in its effective capability set.
A process can gain capabilities in its effective capability
set in various ways.
For example, it may execute a set-user-ID program or an
executable with associated file capabilities.
In addition,
a process may gain capabilities via the effect of
.BR clone (2),
.BR unshare (2),
or
.BR setns (2),
as already described.
.\" In the 3.8 sources, see security/commoncap.c::cap_capable():
.IP 2.
If a process has a capability in a user namespace,
then it has that capability in all child (and further removed descendant)
namespaces as well.
.IP 3.
.\" * The owner of the user namespace in the parent of the
.\" * user namespace has all caps.
When a user namespace is created, the kernel records the effective
user ID of the creating process as being the "owner" of the namespace.
.\" (and likewise associates the effective group ID of the creating process
.\" with the namespace).
A process that resides
in the parent of the user namespace
.\" See kernel commit 520d9eabce18edfef76a60b7b839d54facafe1f9 for a fix
.\" on this point
and whose effective user ID matches the owner of the namespace
has all capabilities in the namespace.
.\" This includes the case where the process executes a set-user-ID
.\" program that confers the effective UID of the creator of the namespace.
By virtue of the previous rule,
this means that the process has all capabilities in all
further removed descendant user namespaces as well.
.\"
.\" ============================================================
.\"
.SS Interaction of user namespaces and other types of namespaces
Starting in Linux 3.8, unprivileged processes can create user namespaces,
and mount, PID, IPC, network, and UTS namespaces can be created with just the
.B CAP_SYS_ADMIN
capability in the caller's user namespace.
If
.BR CLONE_NEWUSER
is specified along with other
.B CLONE_NEW*
flags in a single
.BR clone (2)
or
.BR unshare (2)
call, the user namespace is guaranteed to be created first,
giving the child
.RB ( clone (2))
or caller
.RB ( unshare (2))
privileges over the remaining namespaces created by the call.
Thus, it is possible for an unprivileged caller to specify this combination
of flags.
When a new IPC, mount, network, PID, or UTS namespace is created via
.BR clone (2)
or
.BR unshare (2),
the kernel records the user namespace of the creating process against
the new namespace.
(This association can't be changed.)
When a process in the new namespace subsequently performs
privileged operations that operate on global
resources isolated by the namespace,
the permission checks are performed according to the process's capabilities
in the user namespace that the kernel associated with the new namespace.
.\"
.\" ============================================================
.\"
.SS User and group ID mappings: uid_map and gid_map
When a user namespace is created,
it starts out without a mapping of user IDs (group IDs)
to the parent user namespace.
The
.IR /proc/[pid]/uid_map
and
.IR /proc/[pid]/gid_map
files (available since Linux 3.5)
.\" commit 22d917d80e842829d0ca0a561967d728eb1d6303
expose the mappings for user and group IDs
inside the user namespace for the process
.IR pid .
These files can be read to view the mappings in a user namespace and
written to (once) to define the mappings.
The description in the following paragraphs explains the details for
.IR uid_map ;
.IR gid_map
is exactly the same,
but each instance of "user ID" is replaced by "group ID".
The
.I uid_map
file exposes the mapping of user IDs from the user namespace
of the process
.IR pid
to the user namespace of the process that opened
.IR uid_map
(but see a qualification to this point below).
In other words, processes that are in different user namespaces
will potentially see different values when reading from a particular
.I uid_map
file, depending on the user ID mappings for the user namespaces
of the reading processes.
Each line in the
.I uid_map
file specifies a 1-to-1 mapping of a range of contiguous
user IDs between two user namespaces.
(When a user namespace is first created, this file is empty.)
The specification in each line takes the form of
three numbers delimited by white space.
The first two numbers specify the starting user ID in
each of the two user namespaces.
The third number specifies the length of the mapped range.
In detail, the fields are interpreted as follows:
.IP (1) 4
The start of the range of user IDs in
the user namespace of the process
.IR pid .
.IP (2)
The start of the range of user
IDs to which the user IDs specified by field one map.
How field two is interpreted depends on whether the process that opened
.I uid_map
and the process
.IR pid
are in the same user namespace, as follows:
.RS
.IP a) 3
If the two processes are in different user namespaces:
field two is the start of a range of
user IDs in the user namespace of the process that opened
.IR uid_map .
.IP b)
If the two processes are in the same user namespace:
field two is the start of the range of
user IDs in the parent user namespace of the process
.IR pid .
This case enables the opener of
.I uid_map
(the common case here is opening
.IR /proc/self/uid_map )
to see the mapping of user IDs into the user namespace of the process
that created this user namespace.
.RE
.IP (3)
The length of the range of user IDs that is mapped between the two
user namespaces.
.PP
System calls that return user IDs (group IDs)\(emfor example,
.BR getuid (2),
.BR getgid (2),
and the credential fields in the structure returned by
.BR stat (2)\(emreturn
the user ID (group ID) mapped into the caller's user namespace.
When a process accesses a file, its user and group IDs
are mapped into the initial user namespace for the purpose of permission
checking and assigning IDs when creating a file.
When a process retrieves file user and group IDs via
.BR stat (2),
the IDs are mapped in the opposite direction,
to produce values relative to the process user and group ID mappings.
The initial user namespace has no parent namespace,
but, for consistency, the kernel provides dummy user and group
ID mapping files for this namespace.
Looking at the
.I uid_map
file
.RI ( gid_map
is the same) from a shell in the initial namespace shows:
.in +4n
.nf
$ \fBcat /proc/$$/uid_map\fP
0 0 4294967295
.fi
.in
This mapping tells us
that the range starting at user ID 0 in this namespace
maps to a range starting at 0 in the (nonexistent) parent namespace,
and the length of the range is the largest 32-bit unsigned integer.
.\"
.\" ============================================================
.\"
.SS Defining user and group ID mappings: writing to uid_map and gid_map
.PP
After the creation of a new user namespace, the
.I uid_map
file of
.I one
of the processes in the namespace may be written to
.I once
to define the mapping of user IDs in the new user namespace.
An attempt to write more than once to a
.I uid_map
file in a user namespace fails with the error
.BR EPERM .
Similar rules apply for
.I gid_map
files.
The lines written to
.IR uid_map
.RI ( gid_map )
must conform to the following rules:
.IP * 3
The three fields must be valid numbers,
and the last field must be greater than 0.
.IP *
Lines are terminated by newline characters.
.IP *
There is an (arbitrary) limit on the number of lines in the file.
As at Linux 3.8, the limit is five lines.
In addition, the number of bytes written to
the file must be less than the system page size,
.\" FIXME(Eric): the restriction "less than" rather than "less than or equal"
.\" seems strangely arbitrary. Furthermore, the comment does not agree
.\" with the code in kernel/user_namespace.c. Which is correct.
and the write must be performed at the start of the file (i.e.,
.BR lseek (2)
and
.BR pwrite (2)
can't be used to write to nonzero offsets in the file).
.IP *
The range of user IDs (group IDs)
specified in each line cannot overlap with the ranges
in any other lines.
In the initial implementation (Linux 3.8), this requirement was
satisfied by a simplistic implementation that imposed the further
requirement that
the values in both field 1 and field 2 of successive lines must be
in ascending numerical order,
which prevented some otherwise valid maps from being created.
Linux 3.9 and later
.\" commit 0bd14b4fd72afd5df41e9fd59f356740f22fceba
fix this limitation, allowing any valid set of nonoverlapping maps.
.IP *
At least one line must be written to the file.
.PP
Writes that violate the above rules fail with the error
.BR EINVAL .
In order for a process to write to the
.I /proc/[pid]/uid_map
.RI ( /proc/[pid]/gid_map )
file, all of the following requirements must be met:
.IP 1. 3
The writing process must have the
.BR CAP_SETUID
.RB ( CAP_SETGID )
capability in the user namespace of the process
.IR pid .
.IP 2.
The writing process must be in either the user namespace of the process
.I pid
or inside the parent user namespace of the process
.IR pid .
.IP 3.
The mapped user IDs (group IDs) must in turn have a mapping
in the parent user namespace.
.IP 4.
One of the following is true:
.RS
.IP * 3
The data written to
.I uid_map
.RI ( gid_map )
consists of a single line that maps the writing process's file system user ID
(group ID) in the parent user namespace to a user ID (group ID)
in the user namespace.
The usual case here is that this single line provides a mapping for user ID
of the process that created the namespace.
.IP * 3
The process has the
.BR CAP_SETUID
.RB ( CAP_SETGID )
capability in the parent user namespace.
Thus, a privileged process can make mappings to arbitrary user IDs (group IDs)
in the parent user namespace.
.RE
.PP
Writes that violate the above rules fail with the error
.BR EPERM .
.\"
.\" ============================================================
.\"
.SS Unmapped user and group IDs
.PP
There are various places where an unmapped user ID (group ID)
may be exposed to user space.
For example, the first process in a new user namespace may call
.BR getuid ()
before a user ID mapping has been defined for the namespace.
In most such cases, an unmapped user ID is converted
.\" from_kuid_munged(), from_kgid_munged()
to the overflow user ID (group ID);
the default value for the overflow user ID (group ID) is 65534.
See the descriptions of
.IR /proc/sys/kernel/overflowuid
and
.IR /proc/sys/kernel/overflowgid
in
.BR proc (5).
The cases where unmapped IDs are mapped in this fashion include
system calls that return user IDs
.RB ( getuid (2)
.BR getgid (2),
and similar),
credentials passed over a UNIX domain socket,
.\" also SO_PEERCRED
credentials returned by
.BR stat (2),
.BR waitid (2),
and the System V IPC "ctl"
.B IPC_STAT
operations,
credentials exposed by
.IR /proc/PID/status
and the files in
.IR /proc/sysvipc/* ,
credentials returned via the
.I si_uid
field in the
.I siginfo_t
received with a signal (see
.BR sigaction (2)),
credentials written to the process accounting file (see
.BR acct (5)),
and credentials returned with POSIX message queue notifications (see
.BR mq_notify (3)).
There is one notable case where unmapped user and group IDs are
.I not
.\" from_kuid(), from_kgid()
.\" Also F_GETOWNER_UIDS is an exception
converted to the corresponding overflow ID value.
When viewing a
.I uid_map
or
.I gid_map
file in which there is no mapping for the second field,
that field is displayed as 4294967295 (\-1 as an unsigned integer);
.\"
.\" ============================================================
.\"
.SS Set-user-ID and set-group-ID programs
.PP
When a process inside a user namespace executes
a set-user-ID (set-group-ID) program,
the process's effective user (group) ID inside the namespace is changed
to whatever value is mapped for the user (group) ID of the file.
However, if either the user
.I or
the group ID of the file has no mapping inside the namespace,
the set-user-ID (set-group-ID) bit is silently ignored:
the new program is executed,
but the process's effective user (group) ID is left unchanged.
(This mirrors the semantics of executing a set-user-ID or set-group-ID
program that resides on a file system that was mounted with the
.BR MS_NOSUID
flag, as described in
.BR mount (2).)
.\"
.\" ============================================================
.\"
.SS Miscellaneous
.PP
When a process's user and group IDs are passed over a UNIX domain socket
to a process in a different user namespace (see the description of
.B SCM_CREDENTIALS
in
.BR unix (7)),
they are translated into the corresponding values as per the
receiving process's user and group ID mappings.
.\"
.SH CONFORMING TO
Namespaces are a Linux-specific feature.
.\"
.SH NOTES
Over the years, there have been a lot of features that have been added
to the Linux kernel that have been made available only to privileged users
because of their potential to confuse set-user-ID-root applications.
In general, it becomes safe to allow the root user in a user namespace to
use those features because it is impossible, while in a user namespace,
to gain more privilege than the root user of a user namespace has.
.SS Availability
Use of user namespaces requires a kernel that is configured with the
.B CONFIG_USER_NS
option.
User namespaces require support in a range of subsystems across
the kernel.
When an unsupported subsystem is configured into the kernel,
it is not possible to configure user namespaces support.
As at Linux 3.8, most relevant subsystems support user namespaces,
but there are a number of file systems that do not.
Linux 3.9 added user namespaces support for many of the remaining
unsupported file systems:
Plan 9 (9P), Andrew File System (AFS), Ceph, CIFS, CODA, NFS, and OCFS2.
XFS support for user namespaces is not yet available.
.\"
.SH EXAMPLE
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 executes
a command inside those namespaces.
The comments and
.I usage()
function inside the program provide a full explanation of the program.
The following shell session demonstrates its use.
First, we look at the run-time environment:
.in +4n
.nf
$ \fBuname -rs\fP # Need Linux 3.8 or later
Linux 3.8.0
$ \fBid -u\fP # Running as unprivileged user
1000
$ \fBid -g\fP
1000
.fi
.in
Now start a new shell in new user
.RI ( \-U ),
mount
.RI ( \-m ),
and PID
.RI ( \-p )
namespaces, with user ID
.RI ( \-M )
and group ID
.RI ( \-G )
1000 mapped to 0 inside the user namespace:
.in +4n
.nf
$ \fB./userns_child_exec -p -m -U -M '0 1000 1' -G '0 1000 1' bash\fP
.fi
.in
The shell has PID 1, because it is the first process in the new
PID namespace:
.in +4n
.nf
bash$ \fBecho $$\fP
1
.fi
.in
Inside the user namespace, the shell has user and group ID 0,
and a full set of permitted and effective capabilities:
.in +4n
.nf
bash$ \fBcat /proc/$$/status | egrep '^[UG]id'\fP
Uid: 0 0 0 0
Gid: 0 0 0 0
bash$ \fBcat /proc/$$/status | egrep '^Cap(Prm|Inh|Eff)'\fP
CapInh: 0000000000000000
CapPrm: 0000001fffffffff
CapEff: 0000001fffffffff
.fi
.in
Mounting a new
.I /proc
file system and listing all of the processes visible
in the new PID namespace shows that the shell can't see
any processes outside the PID namespace:
.in +4n
.nf
bash$ \fBmount -t proc proc /proc\fP
bash$ \fBps ax\fP
PID TTY STAT TIME COMMAND
1 pts/3 S 0:00 bash
22 pts/3 R+ 0:00 ps ax
.fi
.in
.SS Program source
\&
.nf
/* userns_child_exec.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 _GNU_SOURCE
#include <sched.h>
#include <unistd.h>
#include <stdlib.h>
#include <sys/wait.h>
#include <signal.h>
#include <fcntl.h>
#include <stdio.h>
#include <string.h>
#include <limits.h>
#include <errno.h>
/* A simple error\-handling function: print an error message based
on the value in \(aqerrno\(aq and terminate the calling process */
#define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \\
} while (0)
struct child_args {
char **argv; /* Command to be executed by child, with args */
int pipe_fd[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 uid_map Specify UID map for user namespace\\n");
fpe("\-G gid_map Specify GID map for user namespace\\n");
fpe("\-z Map user\(aqs UID and GID to 0 in user namespace\\n");
fpe(" (equivalent to: \-M \(aq0 <uid> 1\(aq \-G \(aq0
<gid> 1\(aq)\\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(EXIT_FAILURE);
}
/* Update the mapping file \(aqmap_file\(aq, with the value provided in
\(aqmapping\(aq, 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:
ID_inside\-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
update_map(char *mapping, char *map_file)
{
int fd, j;
size_t map_len; /* Length of \(aqmapping\(aq */
/* Replace commas in mapping string with newlines */
map_len = strlen(mapping);
for (j = 0; j < map_len; j++)
if (mapping[j] == \(aq,\(aq)
mapping[j] = \(aq\\n\(aq;
fd = open(map_file, O_RDWR);
if (fd == \-1) {
fprintf(stderr, "ERROR: open %s: %s\\n", map_file, strerror(errno));
return;
//exit(EXIT_FAILURE);
}
if (write(fd, mapping, map_len) != map_len) {
fprintf(stderr, "ERROR: write %s: %s\\n", map_file, strerror(errno));
//exit(EXIT_FAILURE);
}
close(fd);
}
static int /* Start function for cloned child */
childFunc(void *arg)
{
struct child_args *args = (struct child_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\->pipe_fd[1]); /* Close our descriptor for the write
end of the pipe so that we see EOF
when parent closes its descriptor */
if (read(args\->pipe_fd[0], &ch, 1) != 0) {
fprintf(stderr,
"Failure in child: read from pipe returned != 0\\n");
exit(EXIT_FAILURE);
}
/* Execute a shell command */
printf("About to exec %s\\n", args\->argv[0]);
execvp(args\->argv[0], args\->argv);
errExit("execvp");
}
#define STACK_SIZE (1024 * 1024)
static char child_stack[STACK_SIZE]; /* Space for child\(aqs stack */
int
main(int argc, char *argv[])
{
int flags, opt, map_zero;
pid_t child_pid;
struct child_args args;
char *uid_map, *gid_map;
const int MAP_BUF_SIZE = 100;
char map_buf[MAP_BUF_SIZE];
char map_path[PATH_MAX];
/* Parse command\-line options. The initial \(aq+\(aq character in
the final getopt() argument prevents GNU\-style permutation
of command\-line options. That\(aqs useful, since sometimes
the \(aqcommand\(aq to be executed by this program itself
has command\-line options. We don\(aqt want getopt() to treat
those as options to this program. */
flags = 0;
verbose = 0;
gid_map = NULL;
uid_map = NULL;
map_zero = 0;
while ((opt = getopt(argc, argv, "+imnpuUM:G:zv")) != \-1) {
switch (opt) {
case \(aqi\(aq: flags |= CLONE_NEWIPC; break;
case \(aqm\(aq: flags |= CLONE_NEWNS; break;
case \(aqn\(aq: flags |= CLONE_NEWNET; break;
case \(aqp\(aq: flags |= CLONE_NEWPID; break;
case \(aqu\(aq: flags |= CLONE_NEWUTS; break;
case \(aqv\(aq: verbose = 1; break;
case \(aqz\(aq: map_zero = 1; break;
case \(aqM\(aq: uid_map = optarg; break;
case \(aqG\(aq: gid_map = optarg; break;
case \(aqU\(aq: flags |= CLONE_NEWUSER; break;
default: usage(argv[0]);
}
}
/* \-M or \-G without \-U is nonsensical */
if (((uid_map != NULL || gid_map != NULL || map_zero) &&
!(flags & CLONE_NEWUSER)) ||
(map_zero && (uid_map != NULL || gid_map != 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\(aqs 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\(aqs capabilities during execve()). */
if (pipe(args.pipe_fd) == \-1)
errExit("pipe");
/* Create the child in new namespace(s) */
child_pid = clone(childFunc, child_stack + STACK_SIZE,
flags | SIGCHLD, &args);
if (child_pid == \-1)
errExit("clone");
/* Parent falls through to here */
if (verbose)
printf("%s: PID of child created by clone() is %ld\\n",
argv[0], (long) child_pid);
/* Update the UID and GID maps in the child */
if (uid_map != NULL || map_zero) {
snprintf(map_path, PATH_MAX, "/proc/%ld/uid_map",
(long) child_pid);
if (map_zero) {
snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getuid());
uid_map = map_buf;
}
update_map(uid_map, map_path);
}
if (gid_map != NULL || map_zero) {
snprintf(map_path, PATH_MAX, "/proc/%ld/gid_map",
(long) child_pid);
if (map_zero) {
snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getgid());
gid_map = map_buf;
}
update_map(gid_map, map_path);
}
/* Close the write end of the pipe, to signal to the child that we
have updated the UID and GID maps */
close(args.pipe_fd[1]);
if (waitpid(child_pid, NULL, 0) == \-1) /* Wait for child */
errExit("waitpid");
if (verbose)
printf("%s: terminating\\n", argv[0]);
exit(EXIT_SUCCESS);
}
.fi
.SH SEE ALSO
.BR newgidmap (1), \" From the shadow package
.BR newuidmap (1), \" From the shadow package
.BR clone (2),
.BR setns (2),
.BR unshare (2),
.BR proc (5),
.BR subgid (5), \" From the shadow package
.BR subuid (5), \" From the shadow package
.BR credentials (7),
.BR capabilities (7),
.BR namespaces (7),
.BR pid_namespaces (7)
.sp
The kernel source file
.IR Documentation/namespaces/resource-control.txt .
"Michael Kerrisk (man-pages)" <[email protected]> writes:
> Hi Eric et al.,
>
> All: The attached page aims to provide a fairly complete overview of
> user namespaces. I'm looking for review comments (corrections,
> improvements, additions, etc.) on this man page. I've provided it in
> two forms inline below, and reviewers can comment on whichever form
> they are most comfortable with:
>
> 1) The rendered page as plain text
> 2) The *roff source (also attached); rendering that source will enable
> readers to see proper formatting for the page.
>
> Note that the namespaces(7) page referred to in this page is not yet
> finished; I'll send it out for review at a future time.
>
> Main change since v1 is to address Serge's comments here:
> http://thread.gmane.org/gmane.linux.man/3745/focus=1457720
Overall it looks pretty good. Thanks for doing all of the work to put
this together.
I have a few comments below.
> Cheers,
>
> Michael
>
> =====
>
> USER_NAMESPACES(7) Linux Programmer's Manual USER_NAMESPACES(7)
>
>
>
> NAME
> user_namespaces - overview of Linux user_namespaces
>
> DESCRIPTION
> For an overview of namespaces, see namespaces(7).
>
> User namespaces isolate security-related identifiers, in par‐
> ticular, user IDs and group IDs (see credentials(7), keys (see
> keyctl(2)), and capabilities (see capabilities(7)).
And the root directory.
> A
> process's user and group IDs can be different inside and out‐
> side a user namespace. In particular, a process can have a
> normal unprivileged user ID outside a user namespace while at
> the same time having a user ID of 0 inside the namespace; in
> other words, the process has full privileges for operations
> inside the user namespace, but is unprivileged for operations
> outside the namespace.
>
> Nested namespaces, namespace membership
> User namespaces can be nested; that is, each user namespace—
> except the initial ("root") namespace—has a parent user names‐
> pace, and can have zero or more child user namespaces. The
> parent user namespace is the user namespace of the process that
> creates the user namespace via a call to unshare(2) or clone(2)
> with the CLONE_NEWUSER flag.
>
> Each process is a member of exactly one user namespace. A
> process created via fork(2) or clone(2) without the
> CLONE_NEWUSER flag is a member of the same user namespace as
> its parent. A process can join another user namespace with
> setns(2) if it
is single threaded and
> has the CAP_SYS_ADMIN in that namespace; upon
> doing so, it gains a full set of capabilities in that names‐
> pace.
It is important to implement I don't know if it is important to document
that you may not join your current user namespace with setns(2). This
prevents a process with just CAP_SYS_ADMIN in the current user namespace
from gaining all of the other caps simply by calling setns(2).
> A call to clone(2) or unshare(2) with the CLONE_NEWUSER flag
> makes the new child process (for clone(2)) or the caller (for
> unshare(2)) a member of the new user namespace created by the
> call.
>
> Capabilities
> The child process created by clone(2) with the CLONE_NEWUSER
> flag starts out with a complete set of capabilities in the new
> user namespace. Likewise, a process that creates a new user
> namespace using unshare(2) or joins an existing user namespace
> using setns(2) gains a full set of capabilities in that names‐
> pace. On the other hand, that process has no capabilities in
> the parent (in the case of clone(2)) or previous (in the case
> of unshare(2) and setns(2)) user namespace, even if the new
> namespace is created or joined by the root user (i.e., a
> process with user ID 0 in the root namespace).
> Nevertheless, a
> process owned by the root user will be able to access resources
> such as files that are owned by user ID 0, and will be able to
> do things such as sending signals to processes belonging to
> user ID 0.
I don't understand what you are trying to say in the sentence above.
I think you are trying to say that unprivielged processes in the parent
user namespace with the UID of the creator of the namespace can do the
things the root user can do in the user namespace.
> Note that a call to execve(2) will cause a process to lose any
> capabilities that it has, unless it has a user ID of 0 within
> the namespace. Thus, before calling execve(2), a user ID map‐
> ping for ID 0 must be defined, and the caller may also need to
> use setuid(2) or similar to set its user ID to 0.
I think the above sentence could use a little clarification.
If the users uid in the outer namespace mapes to uid 0 then once the
the mapping is established getuid(2) will return 0, and everything will
treat all processes of that user in the user namespace as uid 0.
If the users uid in the outer namespace is not mapped to uid 0 then
setuid(0) needs to be called if you the process to have uid 0 in the
user namespace.
> A call to clone(2), unshare(2), or setns(2) using the
> CLONE_NEWUSER flag sets the "securebits" flags (see capabili‐
> ties(7)) to their default values (all flags disabled) in the
> child (for clone(2)) or caller (for unshare(2), or setns(2)).
> Note that because the caller no longer has capabilities in its
> original user namespace after a call to setns(2), it is not
> possible for a process to reset its "securebits" flags while
> retaining its user namespace membership by using a pair of
> setns(2) calls to move to another user namespace and then
> return to its original user namespace.
>
> Having a capability inside a user namespace permits a process
> to perform operations (that require privilege) only on
> resources governed by that namespace. The rules for determin‐
> ing whether or not a process has a capability in a particular
> user namespace are as follows:
>
> 1. A process has a capability inside a user namespace if it is
> a member of that namespace and it has the capability in its
> effective capability set. A process can gain capabilities
> in its effective capability set in various ways. For exam‐
> ple, it may execute a set-user-ID program or an executable
> with associated file capabilities. In addition, a process
> may gain capabilities via the effect of clone(2),
> unshare(2), or setns(2), as already described.
>
> 2. If a process has a capability in a user namespace, then it
> has that capability in all child (and further removed
> descendant) namespaces as well.
>
> 3. When a user namespace is created, the kernel records the
> effective user ID of the creating process as being the
> "owner" of the namespace. A process that resides in the
> parent of the user namespace and whose effective user ID
> matches the owner of the namespace has all capabilities in
> the namespace. By virtue of the previous rule, this means
> that the process has all capabilities in all further removed
> descendant user namespaces as well.
>
> Interaction of user namespaces and other types of namespaces
> Starting in Linux 3.8, unprivileged processes can create user
> namespaces, and mount, PID, IPC, network, and UTS namespaces
> can be created with just the CAP_SYS_ADMIN capability in the
> caller's user namespace.
>
> If CLONE_NEWUSER is specified along with other CLONE_NEW* flags
> in a single clone(2) or unshare(2) call, the user namespace is
> guaranteed to be created first, giving the child (clone(2)) or
> caller (unshare(2)) privileges over the remaining namespaces
> created by the call. Thus, it is possible for an unprivileged
> caller to specify this combination of flags.
>
> When a new IPC, mount, network, PID, or UTS namespace is cre‐
> ated via clone(2) or unshare(2), the kernel records the user
> namespace of the creating process against the new namespace.
> (This association can't be changed.) When a process in the new
> namespace subsequently performs privileged operations that
> operate on global resources isolated by the namespace, the per‐
> mission checks are performed according to the process's capa‐
> bilities in the user namespace that the kernel associated with
> the new namespace.
>
> User and group ID mappings: uid_map and gid_map
> When a user namespace is created, it starts out without a map‐
> ping of user IDs (group IDs) to the parent user namespace. The
> /proc/[pid]/uid_map and /proc/[pid]/gid_map files (available
> since Linux 3.5) expose the mappings for user and group IDs
> inside the user namespace for the process pid. These files can
> be read to view the mappings in a user namespace and written to
> (once) to define the mappings.
>
> The description in the following paragraphs explains the
> details for uid_map; gid_map is exactly the same, but each
> instance of "user ID" is replaced by "group ID".
>
> The uid_map file exposes the mapping of user IDs from the user
> namespace of the process pid to the user namespace of the
> process that opened uid_map (but see a qualification to this
> point below). In other words, processes that are in different
> user namespaces will potentially see different values when
> reading from a particular uid_map file, depending on the user
> ID mappings for the user namespaces of the reading processes.
>
> Each line in the uid_map file specifies a 1-to-1 mapping of a
> range of contiguous user IDs between two user namespaces.
> (When a user namespace is first created, this file is empty.)
> The specification in each line takes the form of three numbers
> delimited by white space. The first two numbers specify the
> starting user ID in each of the two user namespaces. The third
> number specifies the length of the mapped range. In detail,
> the fields are interpreted as follows:
>
> (1) The start of the range of user IDs in the user namespace of
> the process pid.
>
> (2) The start of the range of user IDs to which the user IDs
> specified by field one map. How field two is interpreted
> depends on whether the process that opened uid_map and the
> process pid are in the same user namespace, as follows:
>
> a) If the two processes are in different user namespaces:
> field two is the start of a range of user IDs in the
> user namespace of the process that opened uid_map.
>
> b) If the two processes are in the same user namespace:
> field two is the start of the range of user IDs in the
> parent user namespace of the process pid. This case
> enables the opener of uid_map (the common case here is
> opening /proc/self/uid_map) to see the mapping of user
> IDs into the user namespace of the process that created
> this user namespace.
>
> (3) The length of the range of user IDs that is mapped between
> the two user namespaces.
>
> System calls that return user IDs (group IDs)—for example,
> getuid(2), getgid(2), and the credential fields in the struc‐
> ture returned by stat(2)—return the user ID (group ID) mapped
> into the caller's user namespace.
>
> When a process accesses a file, its user and group IDs are
> mapped into the initial user namespace for the purpose of per‐
> mission checking and assigning IDs when creating a file. When
> a process retrieves file user and group IDs via stat(2), the
> IDs are mapped in the opposite direction, to produce values
> relative to the process user and group ID mappings.
>
> The initial user namespace has no parent namespace, but, for
> consistency, the kernel provides dummy user and group ID map‐
> ping files for this namespace. Looking at the uid_map file
> (gid_map is the same) from a shell in the initial namespace
> shows:
>
> $ cat /proc/$$/uid_map
> 0 0 4294967295
>
> This mapping tells us that the range starting at user ID 0 in
> this namespace maps to a range starting at 0 in the (nonexis‐
> tent) parent namespace, and the length of the range is the
> largest 32-bit unsigned integer.
Which winds up meaning that 4294967295 is not mapped even in the inital
user namespace. Documenting that (uid_t)-1 == 4294967295 is not a valid
uid, which reserves (uid_t)-1 for the other uses (uid_t)-1 serves in uid
based apis.
> Defining user and group ID mappings: writing to uid_map and gid_map
> After the creation of a new user namespace, the uid_map file of
> one of the processes in the namespace may be written to once to
> define the mapping of user IDs in the new user namespace. An
> attempt to write more than once to a uid_map file in a user
> namespace fails with the error EPERM. Similar rules apply for
> gid_map files.
>
> The lines written to uid_map (gid_map) must conform to the fol‐
> lowing rules:
>
> * The three fields must be valid numbers, and the last field
> must be greater than 0.
>
> * Lines are terminated by newline characters.
>
> * There is an (arbitrary) limit on the number of lines in the
> file. As at Linux 3.8, the limit is five lines. In addi‐
> tion, the number of bytes written to the file must be less
> than the system page size, and the write must be performed
> at the start of the file (i.e., lseek(2) and pwrite(2) can't
> be used to write to nonzero offsets in the file).
>
> * The range of user IDs (group IDs) specified in each line
> cannot overlap with the ranges in any other lines. In the
> initial implementation (Linux 3.8), this requirement was
> satisfied by a simplistic implementation that imposed the
> further requirement that the values in both field 1 and
> field 2 of successive lines must be in ascending numerical
> order, which prevented some otherwise valid maps from being
> created. Linux 3.9 and later fix this limitation, allowing
> any valid set of nonoverlapping maps.
>
> * At least one line must be written to the file.
>
> Writes that violate the above rules fail with the error EINVAL.
>
> In order for a process to write to the /proc/[pid]/uid_map
> (/proc/[pid]/gid_map) file, all of the following requirements
> must be met:
>
> 1. The writing process must have the CAP_SETUID (CAP_SETGID)
> capability in the user namespace of the process pid.
>
> 2. The writing process must be in either the user namespace of
> the process pid or inside the parent user namespace of the
> process pid.
>
> 3. The mapped user IDs (group IDs) must in turn have a mapping
> in the parent user namespace.
>
> 4. One of the following is true:
>
> * The data written to uid_map (gid_map) consists of a sin‐
> gle line that maps the writing process's file system user
> ID (group ID) in the parent user namespace to a user ID
> (group ID) in the user namespace. The usual case here is
> that this single line provides a mapping for user ID of
> the process that created the namespace.
>
> * The process has the CAP_SETUID (CAP_SETGID) capability in
> the parent user namespace. Thus, a privileged process
> can make mappings to arbitrary user IDs (group IDs) in
> the parent user namespace.
>
> Writes that violate the above rules fail with the error EPERM.
>
> Unmapped user and group IDs
> There are various places where an unmapped user ID (group ID)
> may be exposed to user space. For example, the first process
> in a new user namespace may call getuid() before a user ID map‐
> ping has been defined for the namespace. In most such cases,
> an unmapped user ID is converted to the overflow user ID (group
> ID); the default value for the overflow user ID (group ID) is
> 65534. See the descriptions of /proc/sys/kernel/overflowuid
> and /proc/sys/kernel/overflowgid in proc(5).
>
> The cases where unmapped IDs are mapped in this fashion include
> system calls that return user IDs (getuid(2) getgid(2), and
> similar), credentials passed over a UNIX domain socket, creden‐
> tials returned by stat(2), waitid(2), and the System V IPC
> "ctl" IPC_STAT operations, credentials exposed by
> /proc/PID/status and the files in /proc/sysvipc/*, credentials
> returned via the si_uid field in the siginfo_t received with a
> signal (see sigaction(2)), credentials written to the process
> accounting file (see acct(5)), and credentials returned with
> POSIX message queue notifications (see mq_notify(3)).
>
> There is one notable case where unmapped user and group IDs are
> not converted to the corresponding overflow ID value. When
> viewing a uid_map or gid_map file in which there is no mapping
> for the second field, that field is displayed as 4294967295 (-1
> as an unsigned integer);
>
> Set-user-ID and set-group-ID programs
> When a process inside a user namespace executes a set-user-ID
> (set-group-ID) program, the process's effective user (group) ID
> inside the namespace is changed to whatever value is mapped for
> the user (group) ID of the file. However, if either the user
> or the group ID of the file has no mapping inside the names‐
> pace, the set-user-ID (set-group-ID) bit is silently ignored:
> the new program is executed, but the process's effective user
> (group) ID is left unchanged. (This mirrors the semantics of
> executing a set-user-ID or set-group-ID program that resides on
> a file system that was mounted with the MS_NOSUID flag, as
> described in mount(2).)
>
> Miscellaneous
> When a process's user and group IDs are passed over a UNIX
> domain socket to a process in a different user namespace (see
> the description of SCM_CREDENTIALS in unix(7)), they are trans‐
> lated into the corresponding values as per the receiving
> process's user and group ID mappings.
>
> CONFORMING TO
> Namespaces are a Linux-specific feature.
>
> NOTES
> Over the years, there have been a lot of features that have
> been added to the Linux kernel that have been made available
> only to privileged users because of their potential to confuse
> set-user-ID-root applications. In general, it becomes safe to
> allow the root user in a user namespace to use those features
> because it is impossible, while in a user namespace, to gain
> more privilege than the root user of a user namespace has.
>
> Availability
> Use of user namespaces requires a kernel that is configured
> with the CONFIG_USER_NS option. User namespaces require sup‐
> port in a range of subsystems across the kernel. When an
> unsupported subsystem is configured into the kernel, it is not
> possible to configure user namespaces support. As at Linux
> 3.8, most relevant subsystems support user namespaces, but
> there are a number of file systems that do not. Linux 3.9
> added user namespaces support for many of the remaining unsup‐
> ported file systems: Plan 9 (9P), Andrew File System (AFS),
> Ceph, CIFS, CODA, NFS, and OCFS2. XFS support for user names‐
> paces is not yet available.
I have a conceptual problem with this description of filesystems
supporting user namespaces. There are two levels of support a
filesystem may have.
A filesystem may be like extN and have deal with kuid and kgids and have
all of the necessary mappings of uids and gids to/from userspace in
place. Making the filesystem safe to use in a user namespace enabled
kernel.
A filesystem may be like tmpfs and actually support being mounted from
inside of a user namespace. I suspect nfs may join tmpfs in the not too
distant future, and possibly some of the block based filesystems.
Mostly it is a question of how much do we trust the network or disk
facing side of their implementation. And hopefully at some point fuse
will join in but so far fuse is weird.
At the point you can safely mount a filesystem inside a userns is when I
feel like the filesystem really supports user namespaces. The case for
most of them is that they simply don't break the system when user
namespaces are enabled.
> EXAMPLE
> 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 exe‐
> cutes a command inside those namespaces. The comments and
> usage() function inside the program 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:
>
> $ ./userns_child_exec -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 $$
> 1
>
> 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
>
> Mounting a new /proc file system and listing all of the pro‐
> cesses visible in the new PID namespace 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
>
> Program source
>
> /* userns_child_exec.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 _GNU_SOURCE
> #include <sched.h>
> #include <unistd.h>
> #include <stdlib.h>
> #include <sys/wait.h>
> #include <signal.h>
> #include <fcntl.h>
> #include <stdio.h>
> #include <string.h>
> #include <limits.h>
> #include <errno.h>
>
> /* 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(EXIT_FAILURE); \
> } while (0)
>
> struct child_args {
> char **argv; /* Command to be executed by child, with args */
> int pipe_fd[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 uid_map Specify UID map for user namespace\n");
> fpe("-G gid_map 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 <uid> 1' -G '0 <gid> 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(EXIT_FAILURE);
> }
>
> /* Update the mapping file 'map_file', 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:
>
> ID_inside-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
> update_map(char *mapping, char *map_file)
> {
> int fd, j;
> size_t map_len; /* Length of 'mapping' */
>
> /* Replace commas in mapping string with newlines */
>
> map_len = strlen(mapping);
> for (j = 0; j < map_len; j++)
> if (mapping[j] == ',')
> mapping[j] = '\n';
>
> fd = open(map_file, O_RDWR);
> if (fd == -1) {
> fprintf(stderr, "ERROR: open %s: %s\n", map_file,
> strerror(errno));
> return;
> //exit(EXIT_FAILURE);
> }
>
> if (write(fd, mapping, map_len) != map_len) {
> fprintf(stderr, "ERROR: write %s: %s\n", map_file,
> strerror(errno));
> //exit(EXIT_FAILURE);
> }
>
> close(fd);
> }
>
> static int /* Start function for cloned child */
> childFunc(void *arg)
> {
> struct child_args *args = (struct child_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->pipe_fd[1]); /* Close our descriptor for the write
> end of the pipe so that we see EOF
> when parent closes its descriptor */
> if (read(args->pipe_fd[0], &ch, 1) != 0) {
> fprintf(stderr,
> "Failure in child: read from pipe returned != 0\n");
> exit(EXIT_FAILURE);
> }
>
> /* Execute a shell command */
>
> printf("About to exec %s\n", args->argv[0]);
> execvp(args->argv[0], args->argv);
> errExit("execvp");
> }
>
> #define STACK_SIZE (1024 * 1024)
>
> static char child_stack[STACK_SIZE]; /* Space for child's stack */
>
> int
> main(int argc, char *argv[])
> {
> int flags, opt, map_zero;
> pid_t child_pid;
> struct child_args args;
> char *uid_map, *gid_map;
> const int MAP_BUF_SIZE = 100;
> char map_buf[MAP_BUF_SIZE];
> char map_path[PATH_MAX];
>
> /* 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;
> gid_map = NULL;
> uid_map = NULL;
> map_zero = 0;
> while ((opt = getopt(argc, argv, "+imnpuUM:G:zv")) != -1) {
> switch (opt) {
> case 'i': flags |= CLONE_NEWIPC; break;
> case 'm': flags |= CLONE_NEWNS; break;
> case 'n': flags |= CLONE_NEWNET; break;
> case 'p': flags |= CLONE_NEWPID; break;
> case 'u': flags |= CLONE_NEWUTS; break;
> case 'v': verbose = 1; break;
> case 'z': map_zero = 1; break;
> case 'M': uid_map = optarg; break;
> case 'G': gid_map = optarg; break;
> case 'U': flags |= CLONE_NEWUSER; break;
> default: usage(argv[0]);
> }
> }
>
> /* -M or -G without -U is nonsensical */
>
> if (((uid_map != NULL || gid_map != NULL || map_zero) &&
> !(flags & CLONE_NEWUSER)) ||
> (map_zero && (uid_map != NULL || gid_map != 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.pipe_fd) == -1)
> errExit("pipe");
>
> /* Create the child in new namespace(s) */
>
> child_pid = clone(childFunc, child_stack + STACK_SIZE,
> flags | SIGCHLD, &args);
> if (child_pid == -1)
> errExit("clone");
>
> /* Parent falls through to here */
>
> if (verbose)
> printf("%s: PID of child created by clone() is %ld\n",
> argv[0], (long) child_pid);
>
> /* Update the UID and GID maps in the child */
>
> if (uid_map != NULL || map_zero) {
> snprintf(map_path, PATH_MAX, "/proc/%ld/uid_map",
> (long) child_pid);
> if (map_zero) {
> snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getuid());
> uid_map = map_buf;
> }
> update_map(uid_map, map_path);
> }
> if (gid_map != NULL || map_zero) {
> snprintf(map_path, PATH_MAX, "/proc/%ld/gid_map",
> (long) child_pid);
> if (map_zero) {
> snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getgid());
> gid_map = map_buf;
> }
> update_map(gid_map, map_path);
> }
>
> /* Close the write end of the pipe, to signal to the child that we
> have updated the UID and GID maps */
>
> close(args.pipe_fd[1]);
>
> if (waitpid(child_pid, NULL, 0) == -1) /* Wait for child */
> errExit("waitpid");
>
> if (verbose)
> printf("%s: terminating\n", argv[0]);
>
> exit(EXIT_SUCCESS);
> }
>
> SEE ALSO
> newgidmap(1), newuidmap(1), clone(2), setns(2), unshare(2),
> proc(5), subgid(5), subuid(5), credentials(7), capabilities(7),
> namespaces(7), pid_namespaces(7)
>
> The kernel source file Documentation/namespaces/resource-con‐
> trol.txt.
>
>
>
> Linux 2013-01-14 USER_NAMESPACES(7)
>
>
>
> ========== *roff source ==========
>
> .\" Copyright (c) 2013 by Michael Kerrisk <[email protected]>
> .\" and Copyright (c) 2012 by Eric W. Biederman <[email protected]>
> .\"
> .\" Permission is granted to make and distribute verbatim copies of this
> .\" manual provided the copyright notice and this permission notice are
> .\" preserved on all copies.
> .\"
> .\" Permission is granted to copy and distribute modified versions of this
> .\" manual under the conditions for verbatim copying, provided that the
> .\" entire resulting derived work is distributed under the terms of a
> .\" permission notice identical to this one.
> .\"
> .\" Since the Linux kernel and libraries are constantly changing, this
> .\" manual page may be incorrect or out-of-date. The author(s) assume no
> .\" responsibility for errors or omissions, or for damages resulting from
> .\" the use of the information contained herein. The author(s) may not
> .\" have taken the same level of care in the production of this manual,
> .\" which is licensed free of charge, as they might when working
> .\" professionally.
> .\"
> .\" Formatted or processed versions of this manual, if unaccompanied by
> .\" the source, must acknowledge the copyright and authors of this work.
> .\"
> .\"
> .TH USER_NAMESPACES 7 2013-01-14 "Linux" "Linux Programmer's Manual"
> .SH NAME
> user_namespaces \- overview of Linux user_namespaces
> .SH DESCRIPTION
> For an overview of namespaces, see
> .BR namespaces (7).
>
> User namespaces isolate security-related identifiers, in particular,
> user IDs and group IDs (see
> .BR credentials (7),
> keys (see
> .BR keyctl (2)),
> .\" FIXME: This page says very little about the interaction
> .\" of user namespaces and keys. Add something on this topic.
> and capabilities (see
> .BR capabilities (7)).
> A process's user and group IDs can be different
> inside and outside a user namespace.
> In particular,
> a process can have a normal unprivileged user ID outside a user namespace
> while at the same time having a user ID of 0 inside the namespace;
> in other words,
> the process has full privileges for operations inside the user namespace,
> but is unprivileged for operations outside the namespace.
> .\"
> .\" ============================================================
> .\"
> .SS Nested namespaces, namespace membership
> User namespaces can be nested;
> that is, each user namespace\(emexcept the initial ("root")
> namespace\(emhas a parent user namespace,
> and can have zero or more child user namespaces.
> The parent user namespace is the user namespace
> of the process that creates the user namespace via a call to
> .BR unshare (2)
> or
> .BR clone (2)
> with the
> .BR CLONE_NEWUSER
> flag.
>
> Each process is a member of exactly one user namespace.
> A process created via
> .BR fork (2)
> or
> .BR clone (2)
> without the
> .BR CLONE_NEWUSER
> flag is a member of the same user namespace as its parent.
> A process can join another user namespace with
> .BR setns (2)
> if it has the
> .BR CAP_SYS_ADMIN
> in that namespace;
> upon doing so, it gains a full set of capabilities in that namespace.
>
> A call to
> .BR clone (2)
> or
> .BR unshare (2)
> with the
> .BR CLONE_NEWUSER
> flag makes the new child process (for
> .BR clone (2))
> or the caller (for
> .BR unshare (2))
> a member of the new user namespace created by the call.
> .\"
> .\" ============================================================
> .\"
> .SS Capabilities
> The child process created by
> .BR clone (2)
> with the
> .BR CLONE_NEWUSER
> flag starts out with a complete set
> of capabilities in the new user namespace.
> Likewise, a process that creates a new user namespace using
> .BR unshare (2)
> or joins an existing user namespace using
> .BR setns (2)
> gains a full set of capabilities in that namespace.
> On the other hand,
> that process has no capabilities in the parent (in the case of
> .BR clone (2))
> or previous (in the case of
> .BR unshare (2)
> and
> .BR setns (2))
> user namespace,
> even if the new namespace is created or joined by the root user
> (i.e., a process with user ID 0 in the root namespace).
> Nevertheless, a process owned by the root user
> will be able to access resources such as
> files that are owned by user ID 0,
> and will be able to do things such as sending signals
> to processes belonging to user ID 0.
>
> Note that a call to
> .BR execve (2)
> will cause a process to lose any capabilities that it has,
> unless it has a user ID of 0 within the namespace.
> Thus, before calling
> .BR execve (2),
> a user ID mapping for ID 0 must be defined,
> and the caller may also need to use
> .BR setuid (2)
> or similar to set its user ID to 0.
>
> A call to
> .BR clone (2),
> .BR unshare (2),
> or
> .BR setns (2)
> using the
> .BR CLONE_NEWUSER
> flag sets the "securebits" flags
> (see
> .BR capabilities (7))
> to their default values (all flags disabled) in the child (for
> .BR clone (2))
> or caller (for
> .BR unshare (2),
> or
> .BR setns (2)).
> Note that because the caller no longer has capabilities
> in its original user namespace after a call to
> .BR setns (2),
> it is not possible for a process to reset its "securebits" flags while
> retaining its user namespace membership by using a pair of
> .BR setns (2)
> calls to move to another user namespace and then return to
> its original user namespace.
>
> Having a capability inside a user namespace
> permits a process to perform operations (that require privilege)
> only on resources governed by that namespace.
> The rules for determining whether or not a process has a capability
> in a particular user namespace are as follows:
> .IP 1. 3
> A process has a capability inside a user namespace
> if it is a member of that namespace and
> it has the capability in its effective capability set.
> A process can gain capabilities in its effective capability
> set in various ways.
> For example, it may execute a set-user-ID program or an
> executable with associated file capabilities.
> In addition,
> a process may gain capabilities via the effect of
> .BR clone (2),
> .BR unshare (2),
> or
> .BR setns (2),
> as already described.
> .\" In the 3.8 sources, see security/commoncap.c::cap_capable():
> .IP 2.
> If a process has a capability in a user namespace,
> then it has that capability in all child (and further removed descendant)
> namespaces as well.
> .IP 3.
> .\" * The owner of the user namespace in the parent of the
> .\" * user namespace has all caps.
> When a user namespace is created, the kernel records the effective
> user ID of the creating process as being the "owner" of the namespace.
> .\" (and likewise associates the effective group ID of the creating process
> .\" with the namespace).
> A process that resides
> in the parent of the user namespace
> .\" See kernel commit 520d9eabce18edfef76a60b7b839d54facafe1f9 for a fix
> .\" on this point
> and whose effective user ID matches the owner of the namespace
> has all capabilities in the namespace.
> .\" This includes the case where the process executes a set-user-ID
> .\" program that confers the effective UID of the creator of the namespace.
> By virtue of the previous rule,
> this means that the process has all capabilities in all
> further removed descendant user namespaces as well.
> .\"
> .\" ============================================================
> .\"
> .SS Interaction of user namespaces and other types of namespaces
> Starting in Linux 3.8, unprivileged processes can create user namespaces,
> and mount, PID, IPC, network, and UTS namespaces can be created with just the
> .B CAP_SYS_ADMIN
> capability in the caller's user namespace.
>
> If
> .BR CLONE_NEWUSER
> is specified along with other
> .B CLONE_NEW*
> flags in a single
> .BR clone (2)
> or
> .BR unshare (2)
> call, the user namespace is guaranteed to be created first,
> giving the child
> .RB ( clone (2))
> or caller
> .RB ( unshare (2))
> privileges over the remaining namespaces created by the call.
> Thus, it is possible for an unprivileged caller to specify this combination
> of flags.
>
> When a new IPC, mount, network, PID, or UTS namespace is created via
> .BR clone (2)
> or
> .BR unshare (2),
> the kernel records the user namespace of the creating process against
> the new namespace.
> (This association can't be changed.)
> When a process in the new namespace subsequently performs
> privileged operations that operate on global
> resources isolated by the namespace,
> the permission checks are performed according to the process's capabilities
> in the user namespace that the kernel associated with the new namespace.
> .\"
> .\" ============================================================
> .\"
> .SS User and group ID mappings: uid_map and gid_map
> When a user namespace is created,
> it starts out without a mapping of user IDs (group IDs)
> to the parent user namespace.
> The
> .IR /proc/[pid]/uid_map
> and
> .IR /proc/[pid]/gid_map
> files (available since Linux 3.5)
> .\" commit 22d917d80e842829d0ca0a561967d728eb1d6303
> expose the mappings for user and group IDs
> inside the user namespace for the process
> .IR pid .
> These files can be read to view the mappings in a user namespace and
> written to (once) to define the mappings.
>
> The description in the following paragraphs explains the details for
> .IR uid_map ;
> .IR gid_map
> is exactly the same,
> but each instance of "user ID" is replaced by "group ID".
>
> The
> .I uid_map
> file exposes the mapping of user IDs from the user namespace
> of the process
> .IR pid
> to the user namespace of the process that opened
> .IR uid_map
> (but see a qualification to this point below).
> In other words, processes that are in different user namespaces
> will potentially see different values when reading from a particular
> .I uid_map
> file, depending on the user ID mappings for the user namespaces
> of the reading processes.
>
> Each line in the
> .I uid_map
> file specifies a 1-to-1 mapping of a range of contiguous
> user IDs between two user namespaces.
> (When a user namespace is first created, this file is empty.)
> The specification in each line takes the form of
> three numbers delimited by white space.
> The first two numbers specify the starting user ID in
> each of the two user namespaces.
> The third number specifies the length of the mapped range.
> In detail, the fields are interpreted as follows:
> .IP (1) 4
> The start of the range of user IDs in
> the user namespace of the process
> .IR pid .
> .IP (2)
> The start of the range of user
> IDs to which the user IDs specified by field one map.
> How field two is interpreted depends on whether the process that opened
> .I uid_map
> and the process
> .IR pid
> are in the same user namespace, as follows:
> .RS
> .IP a) 3
> If the two processes are in different user namespaces:
> field two is the start of a range of
> user IDs in the user namespace of the process that opened
> .IR uid_map .
> .IP b)
> If the two processes are in the same user namespace:
> field two is the start of the range of
> user IDs in the parent user namespace of the process
> .IR pid .
> This case enables the opener of
> .I uid_map
> (the common case here is opening
> .IR /proc/self/uid_map )
> to see the mapping of user IDs into the user namespace of the process
> that created this user namespace.
> .RE
> .IP (3)
> The length of the range of user IDs that is mapped between the two
> user namespaces.
> .PP
> System calls that return user IDs (group IDs)\(emfor example,
> .BR getuid (2),
> .BR getgid (2),
> and the credential fields in the structure returned by
> .BR stat (2)\(emreturn
> the user ID (group ID) mapped into the caller's user namespace.
>
> When a process accesses a file, its user and group IDs
> are mapped into the initial user namespace for the purpose of permission
> checking and assigning IDs when creating a file.
> When a process retrieves file user and group IDs via
> .BR stat (2),
> the IDs are mapped in the opposite direction,
> to produce values relative to the process user and group ID mappings.
>
> The initial user namespace has no parent namespace,
> but, for consistency, the kernel provides dummy user and group
> ID mapping files for this namespace.
> Looking at the
> .I uid_map
> file
> .RI ( gid_map
> is the same) from a shell in the initial namespace shows:
>
> .in +4n
> .nf
> $ \fBcat /proc/$$/uid_map\fP
> 0 0 4294967295
> .fi
> .in
>
> This mapping tells us
> that the range starting at user ID 0 in this namespace
> maps to a range starting at 0 in the (nonexistent) parent namespace,
> and the length of the range is the largest 32-bit unsigned integer.
> .\"
> .\" ============================================================
> .\"
> .SS Defining user and group ID mappings: writing to uid_map and gid_map
> .PP
> After the creation of a new user namespace, the
> .I uid_map
> file of
> .I one
> of the processes in the namespace may be written to
> .I once
> to define the mapping of user IDs in the new user namespace.
> An attempt to write more than once to a
> .I uid_map
> file in a user namespace fails with the error
> .BR EPERM .
> Similar rules apply for
> .I gid_map
> files.
>
> The lines written to
> .IR uid_map
> .RI ( gid_map )
> must conform to the following rules:
> .IP * 3
> The three fields must be valid numbers,
> and the last field must be greater than 0.
> .IP *
> Lines are terminated by newline characters.
> .IP *
> There is an (arbitrary) limit on the number of lines in the file.
> As at Linux 3.8, the limit is five lines.
> In addition, the number of bytes written to
> the file must be less than the system page size,
> .\" FIXME(Eric): the restriction "less than" rather than "less than or equal"
> .\" seems strangely arbitrary. Furthermore, the comment does not agree
> .\" with the code in kernel/user_namespace.c. Which is correct.
> and the write must be performed at the start of the file (i.e.,
> .BR lseek (2)
> and
> .BR pwrite (2)
> can't be used to write to nonzero offsets in the file).
> .IP *
> The range of user IDs (group IDs)
> specified in each line cannot overlap with the ranges
> in any other lines.
> In the initial implementation (Linux 3.8), this requirement was
> satisfied by a simplistic implementation that imposed the further
> requirement that
> the values in both field 1 and field 2 of successive lines must be
> in ascending numerical order,
> which prevented some otherwise valid maps from being created.
> Linux 3.9 and later
> .\" commit 0bd14b4fd72afd5df41e9fd59f356740f22fceba
> fix this limitation, allowing any valid set of nonoverlapping maps.
> .IP *
> At least one line must be written to the file.
> .PP
> Writes that violate the above rules fail with the error
> .BR EINVAL .
>
> In order for a process to write to the
> .I /proc/[pid]/uid_map
> .RI ( /proc/[pid]/gid_map )
> file, all of the following requirements must be met:
> .IP 1. 3
> The writing process must have the
> .BR CAP_SETUID
> .RB ( CAP_SETGID )
> capability in the user namespace of the process
> .IR pid .
> .IP 2.
> The writing process must be in either the user namespace of the process
> .I pid
> or inside the parent user namespace of the process
> .IR pid .
> .IP 3.
> The mapped user IDs (group IDs) must in turn have a mapping
> in the parent user namespace.
> .IP 4.
> One of the following is true:
> .RS
> .IP * 3
> The data written to
> .I uid_map
> .RI ( gid_map )
> consists of a single line that maps the writing process's file system user ID
> (group ID) in the parent user namespace to a user ID (group ID)
> in the user namespace.
> The usual case here is that this single line provides a mapping for user ID
> of the process that created the namespace.
> .IP * 3
> The process has the
> .BR CAP_SETUID
> .RB ( CAP_SETGID )
> capability in the parent user namespace.
> Thus, a privileged process can make mappings to arbitrary user IDs (group IDs)
> in the parent user namespace.
> .RE
> .PP
> Writes that violate the above rules fail with the error
> .BR EPERM .
> .\"
> .\" ============================================================
> .\"
> .SS Unmapped user and group IDs
> .PP
> There are various places where an unmapped user ID (group ID)
> may be exposed to user space.
> For example, the first process in a new user namespace may call
> .BR getuid ()
> before a user ID mapping has been defined for the namespace.
> In most such cases, an unmapped user ID is converted
> .\" from_kuid_munged(), from_kgid_munged()
> to the overflow user ID (group ID);
> the default value for the overflow user ID (group ID) is 65534.
> See the descriptions of
> .IR /proc/sys/kernel/overflowuid
> and
> .IR /proc/sys/kernel/overflowgid
> in
> .BR proc (5).
>
> The cases where unmapped IDs are mapped in this fashion include
> system calls that return user IDs
> .RB ( getuid (2)
> .BR getgid (2),
> and similar),
> credentials passed over a UNIX domain socket,
> .\" also SO_PEERCRED
> credentials returned by
> .BR stat (2),
> .BR waitid (2),
> and the System V IPC "ctl"
> .B IPC_STAT
> operations,
> credentials exposed by
> .IR /proc/PID/status
> and the files in
> .IR /proc/sysvipc/* ,
> credentials returned via the
> .I si_uid
> field in the
> .I siginfo_t
> received with a signal (see
> .BR sigaction (2)),
> credentials written to the process accounting file (see
> .BR acct (5)),
> and credentials returned with POSIX message queue notifications (see
> .BR mq_notify (3)).
>
> There is one notable case where unmapped user and group IDs are
> .I not
> .\" from_kuid(), from_kgid()
> .\" Also F_GETOWNER_UIDS is an exception
> converted to the corresponding overflow ID value.
> When viewing a
> .I uid_map
> or
> .I gid_map
> file in which there is no mapping for the second field,
> that field is displayed as 4294967295 (\-1 as an unsigned integer);
> .\"
> .\" ============================================================
> .\"
> .SS Set-user-ID and set-group-ID programs
> .PP
> When a process inside a user namespace executes
> a set-user-ID (set-group-ID) program,
> the process's effective user (group) ID inside the namespace is changed
> to whatever value is mapped for the user (group) ID of the file.
> However, if either the user
> .I or
> the group ID of the file has no mapping inside the namespace,
> the set-user-ID (set-group-ID) bit is silently ignored:
> the new program is executed,
> but the process's effective user (group) ID is left unchanged.
> (This mirrors the semantics of executing a set-user-ID or set-group-ID
> program that resides on a file system that was mounted with the
> .BR MS_NOSUID
> flag, as described in
> .BR mount (2).)
> .\"
> .\" ============================================================
> .\"
> .SS Miscellaneous
> .PP
> When a process's user and group IDs are passed over a UNIX domain socket
> to a process in a different user namespace (see the description of
> .B SCM_CREDENTIALS
> in
> .BR unix (7)),
> they are translated into the corresponding values as per the
> receiving process's user and group ID mappings.
> .\"
> .SH CONFORMING TO
> Namespaces are a Linux-specific feature.
> .\"
> .SH NOTES
> Over the years, there have been a lot of features that have been added
> to the Linux kernel that have been made available only to privileged users
> because of their potential to confuse set-user-ID-root applications.
> In general, it becomes safe to allow the root user in a user namespace to
> use those features because it is impossible, while in a user namespace,
> to gain more privilege than the root user of a user namespace has.
> .SS Availability
> Use of user namespaces requires a kernel that is configured with the
> .B CONFIG_USER_NS
> option.
> User namespaces require support in a range of subsystems across
> the kernel.
> When an unsupported subsystem is configured into the kernel,
> it is not possible to configure user namespaces support.
> As at Linux 3.8, most relevant subsystems support user namespaces,
> but there are a number of file systems that do not.
> Linux 3.9 added user namespaces support for many of the remaining
> unsupported file systems:
> Plan 9 (9P), Andrew File System (AFS), Ceph, CIFS, CODA, NFS, and OCFS2.
> XFS support for user namespaces is not yet available.
> .\"
> .SH EXAMPLE
> 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 executes
> a command inside those namespaces.
> The comments and
> .I usage()
> function inside the program provide a full explanation of the program.
> The following shell session demonstrates its use.
>
> First, we look at the run-time environment:
>
> .in +4n
> .nf
> $ \fBuname -rs\fP # Need Linux 3.8 or later
> Linux 3.8.0
> $ \fBid -u\fP # Running as unprivileged user
> 1000
> $ \fBid -g\fP
> 1000
> .fi
> .in
>
> Now start a new shell in new user
> .RI ( \-U ),
> mount
> .RI ( \-m ),
> and PID
> .RI ( \-p )
> namespaces, with user ID
> .RI ( \-M )
> and group ID
> .RI ( \-G )
> 1000 mapped to 0 inside the user namespace:
>
> .in +4n
> .nf
> $ \fB./userns_child_exec -p -m -U -M '0 1000 1' -G '0 1000 1' bash\fP
> .fi
> .in
>
> The shell has PID 1, because it is the first process in the new
> PID namespace:
>
> .in +4n
> .nf
> bash$ \fBecho $$\fP
> 1
> .fi
> .in
>
> Inside the user namespace, the shell has user and group ID 0,
> and a full set of permitted and effective capabilities:
>
> .in +4n
> .nf
> bash$ \fBcat /proc/$$/status | egrep '^[UG]id'\fP
> Uid: 0 0 0 0
> Gid: 0 0 0 0
> bash$ \fBcat /proc/$$/status | egrep '^Cap(Prm|Inh|Eff)'\fP
> CapInh: 0000000000000000
> CapPrm: 0000001fffffffff
> CapEff: 0000001fffffffff
> .fi
> .in
>
> Mounting a new
> .I /proc
> file system and listing all of the processes visible
> in the new PID namespace shows that the shell can't see
> any processes outside the PID namespace:
>
> .in +4n
> .nf
> bash$ \fBmount -t proc proc /proc\fP
> bash$ \fBps ax\fP
> PID TTY STAT TIME COMMAND
> 1 pts/3 S 0:00 bash
> 22 pts/3 R+ 0:00 ps ax
> .fi
> .in
> .SS Program source
> \&
> .nf
> /* userns_child_exec.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 _GNU_SOURCE
> #include <sched.h>
> #include <unistd.h>
> #include <stdlib.h>
> #include <sys/wait.h>
> #include <signal.h>
> #include <fcntl.h>
> #include <stdio.h>
> #include <string.h>
> #include <limits.h>
> #include <errno.h>
>
> /* A simple error\-handling function: print an error message based
> on the value in \(aqerrno\(aq and terminate the calling process */
>
> #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \\
> } while (0)
>
> struct child_args {
> char **argv; /* Command to be executed by child, with args */
> int pipe_fd[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 uid_map Specify UID map for user namespace\\n");
> fpe("\-G gid_map Specify GID map for user namespace\\n");
> fpe("\-z Map user\(aqs UID and GID to 0 in user namespace\\n");
> fpe(" (equivalent to: \-M \(aq0 <uid> 1\(aq \-G \(aq0
> <gid> 1\(aq)\\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(EXIT_FAILURE);
> }
>
> /* Update the mapping file \(aqmap_file\(aq, with the value provided in
> \(aqmapping\(aq, 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:
>
> ID_inside\-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
> update_map(char *mapping, char *map_file)
> {
> int fd, j;
> size_t map_len; /* Length of \(aqmapping\(aq */
>
> /* Replace commas in mapping string with newlines */
>
> map_len = strlen(mapping);
> for (j = 0; j < map_len; j++)
> if (mapping[j] == \(aq,\(aq)
> mapping[j] = \(aq\\n\(aq;
>
> fd = open(map_file, O_RDWR);
> if (fd == \-1) {
> fprintf(stderr, "ERROR: open %s: %s\\n", map_file, strerror(errno));
> return;
> //exit(EXIT_FAILURE);
> }
>
> if (write(fd, mapping, map_len) != map_len) {
> fprintf(stderr, "ERROR: write %s: %s\\n", map_file, strerror(errno));
> //exit(EXIT_FAILURE);
> }
>
> close(fd);
> }
>
> static int /* Start function for cloned child */
> childFunc(void *arg)
> {
> struct child_args *args = (struct child_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\->pipe_fd[1]); /* Close our descriptor for the write
> end of the pipe so that we see EOF
> when parent closes its descriptor */
> if (read(args\->pipe_fd[0], &ch, 1) != 0) {
> fprintf(stderr,
> "Failure in child: read from pipe returned != 0\\n");
> exit(EXIT_FAILURE);
> }
>
> /* Execute a shell command */
>
> printf("About to exec %s\\n", args\->argv[0]);
> execvp(args\->argv[0], args\->argv);
> errExit("execvp");
> }
>
> #define STACK_SIZE (1024 * 1024)
>
> static char child_stack[STACK_SIZE]; /* Space for child\(aqs stack */
>
> int
> main(int argc, char *argv[])
> {
> int flags, opt, map_zero;
> pid_t child_pid;
> struct child_args args;
> char *uid_map, *gid_map;
> const int MAP_BUF_SIZE = 100;
> char map_buf[MAP_BUF_SIZE];
> char map_path[PATH_MAX];
>
> /* Parse command\-line options. The initial \(aq+\(aq character in
> the final getopt() argument prevents GNU\-style permutation
> of command\-line options. That\(aqs useful, since sometimes
> the \(aqcommand\(aq to be executed by this program itself
> has command\-line options. We don\(aqt want getopt() to treat
> those as options to this program. */
>
> flags = 0;
> verbose = 0;
> gid_map = NULL;
> uid_map = NULL;
> map_zero = 0;
> while ((opt = getopt(argc, argv, "+imnpuUM:G:zv")) != \-1) {
> switch (opt) {
> case \(aqi\(aq: flags |= CLONE_NEWIPC; break;
> case \(aqm\(aq: flags |= CLONE_NEWNS; break;
> case \(aqn\(aq: flags |= CLONE_NEWNET; break;
> case \(aqp\(aq: flags |= CLONE_NEWPID; break;
> case \(aqu\(aq: flags |= CLONE_NEWUTS; break;
> case \(aqv\(aq: verbose = 1; break;
> case \(aqz\(aq: map_zero = 1; break;
> case \(aqM\(aq: uid_map = optarg; break;
> case \(aqG\(aq: gid_map = optarg; break;
> case \(aqU\(aq: flags |= CLONE_NEWUSER; break;
> default: usage(argv[0]);
> }
> }
>
> /* \-M or \-G without \-U is nonsensical */
>
> if (((uid_map != NULL || gid_map != NULL || map_zero) &&
> !(flags & CLONE_NEWUSER)) ||
> (map_zero && (uid_map != NULL || gid_map != 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\(aqs 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\(aqs capabilities during execve()). */
>
> if (pipe(args.pipe_fd) == \-1)
> errExit("pipe");
>
> /* Create the child in new namespace(s) */
>
> child_pid = clone(childFunc, child_stack + STACK_SIZE,
> flags | SIGCHLD, &args);
> if (child_pid == \-1)
> errExit("clone");
>
> /* Parent falls through to here */
>
> if (verbose)
> printf("%s: PID of child created by clone() is %ld\\n",
> argv[0], (long) child_pid);
>
> /* Update the UID and GID maps in the child */
>
> if (uid_map != NULL || map_zero) {
> snprintf(map_path, PATH_MAX, "/proc/%ld/uid_map",
> (long) child_pid);
> if (map_zero) {
> snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getuid());
> uid_map = map_buf;
> }
> update_map(uid_map, map_path);
> }
> if (gid_map != NULL || map_zero) {
> snprintf(map_path, PATH_MAX, "/proc/%ld/gid_map",
> (long) child_pid);
> if (map_zero) {
> snprintf(map_buf, MAP_BUF_SIZE, "0 %ld 1", (long) getgid());
> gid_map = map_buf;
> }
> update_map(gid_map, map_path);
> }
>
> /* Close the write end of the pipe, to signal to the child that we
> have updated the UID and GID maps */
>
> close(args.pipe_fd[1]);
>
> if (waitpid(child_pid, NULL, 0) == \-1) /* Wait for child */
> errExit("waitpid");
>
> if (verbose)
> printf("%s: terminating\\n", argv[0]);
>
> exit(EXIT_SUCCESS);
> }
> .fi
> .SH SEE ALSO
> .BR newgidmap (1), \" From the shadow package
> .BR newuidmap (1), \" From the shadow package
> .BR clone (2),
> .BR setns (2),
> .BR unshare (2),
> .BR proc (5),
> .BR subgid (5), \" From the shadow package
> .BR subuid (5), \" From the shadow package
> .BR credentials (7),
> .BR capabilities (7),
> .BR namespaces (7),
> .BR pid_namespaces (7)
> .sp
> The kernel source file
> .IR Documentation/namespaces/resource-control.txt .
Over the last little while I have been working to correct a design
oversight in user namespaces, that probably needs to be documented
somewhere, and the fixes for the worst of the oversight have been
merged.
The problem was I forgot to consider what when there are shared
resources and root uses things like chroot and mounts as access policy
controls, and not as a mechanism to prevent the gaining of privilege.
This has led to the realization that the root directory is one of the
privileged identifiers that is controlled by the user namespace.
So now there is a restriction that user namespaces can not be created
if you are chrooted.
Beyond that there are restrictions on what you can do in a mount
namespace created inside a user namespace. Read-only bind mounts
may not be remounted to read-write. The mqueue filesystem may only be
mounted if you have CAP_SYS_ADMIN over it's ipc namespace. proc and
sysfs may only be mounted if they are already somewhere in the mount
namespace.
There is a remaining open question on what to allow in the context of
unmounting and bind mounts. In the normal case unmounting something is
safe because mounts almost always happen on an empty directory. The
only significant case that I can think of where this is different are
union mounts and union filesystems. However the general principle of
following the restrictions of the root user makes suggests that unmounts
should not happen.
In the grand scheme of things these are small little things but they are
details we need to get right so that enabling user namespaces is no
worse that adding any other feature to the kernel. In the worst case
just adding more attack surface for the bad guys, but not a matter of
risk semantically.
Eric
On Wed, Mar 27, 2013 at 10:26 PM, Michael Kerrisk (man-pages)
<[email protected]> wrote:
> 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
I've tried your demo program, but inside the new ns I'm automatically nobody.
As Eric said, setuid(0)/setgid(0) are missing.
Eric, maybe you can help me. How can I drop capabilities within a user
namespace?
In childFunc() I did add prctl(PR_CAPBSET_DROP, CAP_NET_ADMIN) but it always
returns ENOPERM.
What that? I thought I get a completely fresh set of cap which I can modify.
I don't want that uid 0 inside the container has all caps.
And why does /proc/*/loginuid always contain 4294967295 in a new user namespace?
Writing to it also fails. (Noticed that because pam_loginuid.so does not work).
Final question, is it by design that uid 0 within a namespace in not
allowed to write to
/proc/*/oom_score_adj?
Thanks,
//richard
P.s: I've used 3.9-rc8 for my tests...
richard -rw- weinberger <[email protected]> writes:
> On Wed, Mar 27, 2013 at 10:26 PM, Michael Kerrisk (man-pages)
> <[email protected]> wrote:
>> 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
>
> I've tried your demo program, but inside the new ns I'm automatically nobody.
> As Eric said, setuid(0)/setgid(0) are missing.
Is it the setuid/setgid or not setting up the uid/gid map?
> Eric, maybe you can help me. How can I drop capabilities within a user
> namespace?
> In childFunc() I did add prctl(PR_CAPBSET_DROP, CAP_NET_ADMIN) but it always
> returns ENOPERM.
> What that? I thought I get a completely fresh set of cap which I can modify.
> I don't want that uid 0 inside the container has all caps.
There are weird things that happen with exec and the user namespace. If
you have exec'd as an unmapped user all of your capabilities have
already been droped.
> And why does /proc/*/loginuid always contain 4294967295 in a new user namespace?
> Writing to it also fails. (Noticed that because pam_loginuid.so does not work).
Almost certainly because the loginuid has already been set. Yes. It
looks like I am simply using from_kuid instead of from_kuid_munged on
the read. So an unmapped loginuid will be reported as 4294967295.
For some circumstances 65534 (nobody) is definitely better in some it is
a toss up, and most of the time no one really cares. So I have tried to
do something but in this case I don't know which was the best policy.
> Final question, is it by design that uid 0 within a namespace in not
> allowed to write to
> /proc/*/oom_score_adj?
Essentially. It is by design that uid 0 within a namespace be mapped to
some other uid outside the namespace, and that the permissions on writes
should use the permission needed outside of the user namespace.
Which means there are all kinds of things only uid 0 can write to, that
you can't touch in a user namespace. Some of those things the policy
may need to be reconsidered. A lot of those things the default policy
is good. Regardless we are now defaulting to not letting root in a
container do risky things which is a good thing.
Eric
On Fri, Apr 26, 2013 at 2:54 AM, Eric W. Biederman
<[email protected]> wrote:
> richard -rw- weinberger <[email protected]> writes:
>
>> On Wed, Mar 27, 2013 at 10:26 PM, Michael Kerrisk (man-pages)
>> <[email protected]> wrote:
>>> 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
>>
>> I've tried your demo program, but inside the new ns I'm automatically nobody.
>> As Eric said, setuid(0)/setgid(0) are missing.
>
> Is it the setuid/setgid or not setting up the uid/gid map?
uid/git mapping are set up.
>> Eric, maybe you can help me. How can I drop capabilities within a user
>> namespace?
>
>> In childFunc() I did add prctl(PR_CAPBSET_DROP, CAP_NET_ADMIN) but it always
>> returns ENOPERM.
>> What that? I thought I get a completely fresh set of cap which I can modify.
>> I don't want that uid 0 inside the container has all caps.
>
> There are weird things that happen with exec and the user namespace. If
> you have exec'd as an unmapped user all of your capabilities have
> already been droped.
I've setup the mappings. If I look into /proc/*/status I see that my process has
all caps.
So, in general it is possible to drop cap within a user namespace?
I really want to drop CAP_NET_ADMIN and some others.
root within my container must not change any networking settings.
>> And why does /proc/*/loginuid always contain 4294967295 in a new user namespace?
>> Writing to it also fails. (Noticed that because pam_loginuid.so does not work).
>
> Almost certainly because the loginuid has already been set. Yes. It
> looks like I am simply using from_kuid instead of from_kuid_munged on
> the read. So an unmapped loginuid will be reported as 4294967295.
>
> For some circumstances 65534 (nobody) is definitely better in some it is
> a toss up, and most of the time no one really cares. So I have tried to
> do something but in this case I don't know which was the best policy.
Hmm, I hoped that loginuid will be reset upon entering a user namespace.
>> Final question, is it by design that uid 0 within a namespace in not
>> allowed to write to
>> /proc/*/oom_score_adj?
>
> Essentially. It is by design that uid 0 within a namespace be mapped to
> some other uid outside the namespace, and that the permissions on writes
> should use the permission needed outside of the user namespace.
Okay, I've asked because systemd is a heavy user of this file and
fails due to this
within a user namespace.
Luckily it is possible to remove all the score changes from the .service files.
--
Thanks,
//richard
On Thu, Apr 25, 2013 at 10:48 PM, richard -rw- weinberger
<[email protected]> wrote:
> On Fri, Apr 26, 2013 at 2:54 AM, Eric W. Biederman
> <[email protected]> wrote:
>> richard -rw- weinberger <[email protected]> writes:
>>
>>> On Wed, Mar 27, 2013 at 10:26 PM, Michael Kerrisk (man-pages)
>>> <[email protected]> wrote:
>>>> 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
>>>
>>> I've tried your demo program, but inside the new ns I'm automatically nobody.
>>> As Eric said, setuid(0)/setgid(0) are missing.
>>
>> Is it the setuid/setgid or not setting up the uid/gid map?
>
> uid/git mapping are set up.
>
>>> Eric, maybe you can help me. How can I drop capabilities within a user
>>> namespace?
>>
>>> In childFunc() I did add prctl(PR_CAPBSET_DROP, CAP_NET_ADMIN) but it always
>>> returns ENOPERM.
>>> What that? I thought I get a completely fresh set of cap which I can modify.
>>> I don't want that uid 0 inside the container has all caps.
>>
>> There are weird things that happen with exec and the user namespace. If
>> you have exec'd as an unmapped user all of your capabilities have
>> already been droped.
>
> I've setup the mappings. If I look into /proc/*/status I see that my process has
> all caps.
> So, in general it is possible to drop cap within a user namespace?
> I really want to drop CAP_NET_ADMIN and some others.
> root within my container must not change any networking settings.
You may have the common issue that uid 0 tends to regain capabilities
on exec due to "legacy" capability emulation. Try playing with
securebits and/or the bounding set. (The setpriv command in very new
util-linux-ng makes this easy to play with.)
Note that you almost certainly want to set no_new_privs if anything
other than uid 0 is running with non-default securebits.
--Andy