1CGROUP_NAMESPACES(7) Linux Programmer's Manual CGROUP_NAMESPACES(7)
2
6 cgroup_namespaces - overview of Linux cgroup namespaces
7
9 For an overview of namespaces, see namespaces(7).
10 Cgroup namespaces virtualize the view of a process's cgroups (see
12 cgroups(7)) as seen via /proc/[pid]/cgroup and /proc/[pid]/mountinfo.
13 Each cgroup namespace has its own set of cgroup root directories.
15 These root directories are the base points for the relative locations
16 displayed in the corresponding records in the /proc/[pid]/cgroup file.
17 When a process creates a new cgroup namespace using clone(2) or
18 unshare(2) with the CLONE_NEWCGROUP flag, it enters a new cgroup names‐
19 pace in which its current cgroups directories become the cgroup root
20 directories of the new namespace. (This applies both for the cgroups
21 version 1 hierarchies and the cgroups version 2 unified hierarchy.)
22 When viewing /proc/[pid]/cgroup, the pathname shown in the third field
24 of each record will be relative to the reading process's root directory
25 for the corresponding cgroup hierarchy. If the cgroup directory of the
26 target process lies outside the root directory of the reading process's
27 cgroup namespace, then the pathname will show ../ entries for each
28 ancestor level in the cgroup hierarchy.
29 The following shell session demonstrates the effect of creating a new
31 cgroup namespace. First, (as superuser) we create a child cgroup in
32 the freezer hierarchy, and put the shell into that cgroup:
33 # mkdir -p /sys/fs/cgroup/freezer/sub
35 # echo $$ # Show PID of this shell
36 30655
37 # sh -c 'echo 30655 > /sys/fs/cgroup/freezer/sub/cgroup.procs'
38 # cat /proc/self/cgroup | grep freezer
39 7:freezer:/sub
40 Next, we use unshare(1) to create a process running a new shell in new
42 cgroup and mount namespaces:
43 # unshare -Cm bash
45 We then inspect the /proc/[pid]/cgroup files of, respectively, the new
47 shell process started by the unshare(1) command, a process that is in
48 the original cgroup namespace (init, with PID 1), and a process in a
49 sibling cgroup (sub2):
50 $ cat /proc/self/cgroup | grep freezer
52 7:freezer:/
53 $ cat /proc/1/cgroup | grep freezer
54 7:freezer:/..
55 $ cat /proc/20124/cgroup | grep freezer
56 7:freezer:/../sub2
57 From the output of the first command, we see that the freezer cgroup
59 membership of the new shell (which is in the same cgroup as the initial
60 shell) is shown defined relative to the freezer cgroup root directory
61 that was established when the new cgroup namespace was created. (In
62 absolute terms, the new shell is in the /sub freezer cgroup, and the
63 root directory of the freezer cgroup hierarchy in the new cgroup names‐
64 pace is also /sub. Thus, the new shell's cgroup membership is dis‐
65 played as '/'.)
66 However, when we look in /proc/self/mountinfo we see the following
68 anomaly:
69 # cat /proc/self/mountinfo | grep freezer
71 155 145 0:32 /.. /sys/fs/cgroup/freezer ...
72 The fourth field of this line (/..) should show the directory in the
74 cgroup filesystem which forms the root of this mount. Since by the
75 definition of cgroup namespaces, the process's current freezer cgroup
76 directory became its root freezer cgroup directory, we should see '/'
77 in this field. The problem here is that we are seeing a mount entry
78 for the cgroup filesystem corresponding to our initial shell process's
79 cgroup namespace (whose cgroup filesystem is indeed rooted in the par‐
80 ent directory of sub). We need to remount the freezer cgroup filesys‐
81 tem inside this cgroup namespace, after which we see the expected
82 results:
83 # mount --make-rslave / # Don't propagate mount events
85 # to other namespaces
86 # umount /sys/fs/cgroup/freezer
87 # mount -t cgroup -o freezer freezer /sys/fs/cgroup/freezer
88 # cat /proc/self/mountinfo | grep freezer
89 155 145 0:32 / /sys/fs/cgroup/freezer rw,relatime ...
90 Use of cgroup namespaces requires a kernel that is configured with the
92 CONFIG_CGROUPS option.
93
95 Namespaces are a Linux-specific feature.
96
98 Among the purposes served by the virtualization provided by cgroup
99 namespaces are the following:
100 * It prevents information leaks whereby cgroup directory paths outside
102 of a container would otherwise be visible to processes in the con‐
103 tainer. Such leakages could, for example, reveal information about
104 the container framework to containerized applications.
105 * It eases tasks such as container migration. The virtualization pro‐
107 vided by cgroup namespaces allows containers to be isolated from
108 knowledge of the pathnames of ancestor cgroups. Without such isola‐
109 tion, the full cgroup pathnames (displayed in /proc/self/cgroups)
110 would need to be replicated on the target system when migrating a
111 container; those pathnames would also need to be unique, so that they
112 don't conflict with other pathnames on the target system.
113 * It allows better confinement of containerized processes, because it
115 is possible to mount the container's cgroup filesystems such that the
116 container processes can't gain access to ancestor cgroup directories.
117 Consider, for example, the following scenario:
118 · We have a cgroup directory, /cg/1, that is owned by user ID 9000.
120 · We have a process, X, also owned by user ID 9000, that is names‐
122 paced under the cgroup /cg/1/2 (i.e., X was placed in a new
123 cgroup namespace via clone(2) or unshare(2) with the CLONE_NEWC‐
124 GROUP flag).
125 In the absence of cgroup namespacing, because the cgroup directory
127 /cg/1 is owned (and writable) by UID 9000 and process X is also owned
128 by user ID 9000, then process X would be able to modify the contents
129 of cgroups files (i.e., change cgroup settings) not only in /cg/1/2
130 but also in the ancestor cgroup directory /cg/1. Namespacing process
131 X under the cgroup directory /cg/1/2, in combination with suitable
132 mount operations for the cgroup filesystem (as shown above), prevents
133 it modifying files in /cg/1, since it cannot even see the contents of
134 that directory (or of further removed cgroup ancestor directories).
135 Combined with correct enforcement of hierarchical limits, this pre‐
136 vents process X from escaping the limits imposed by ancestor cgroups.
137
139 unshare(1), clone(2), setns(2), unshare(2), proc(5), cgroups(7), cre‐
140 dentials(7), namespaces(7), user_namespaces(7)
141
143 This page is part of release 4.15 of the Linux man-pages project. A
144 description of the project, information about reporting bugs, and the
145 latest version of this page, can be found at
146 https://www.kernel.org/doc/man-pages/.
147Linux 2017-09-15 CGROUP_NAMESPACES(7)