1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
|
/* SPDX-License-Identifier: LGPL-2.1+ */
/***
This file is part of systemd.
Copyright 2015 Lennart Poettering
systemd is free software; you can redistribute it and/or modify it
under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 2.1 of the License, or
(at your option) any later version.
systemd is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License
along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/
#include <sys/mount.h>
#include "alloc-util.h"
#include "fd-util.h"
#include "fileio.h"
#include "mkdir.h"
#include "mount-util.h"
#include "nspawn-cgroup.h"
#include "rm-rf.h"
#include "string-util.h"
#include "strv.h"
#include "util.h"
static int chown_cgroup_path(const char *path, uid_t uid_shift) {
_cleanup_close_ int fd = -1;
const char *fn;
fd = open(path, O_RDONLY|O_CLOEXEC|O_DIRECTORY);
if (fd < 0)
return -errno;
FOREACH_STRING(fn,
".",
"tasks",
"notify_on_release",
"cgroup.procs",
"cgroup.events",
"cgroup.clone_children",
"cgroup.controllers",
"cgroup.subtree_control")
if (fchownat(fd, fn, uid_shift, uid_shift, 0) < 0)
log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_WARNING, errno,
"Failed to chown \"%s/%s\", ignoring: %m", path, fn);
return 0;
}
int chown_cgroup(pid_t pid, uid_t uid_shift) {
_cleanup_free_ char *path = NULL, *fs = NULL;
int r;
r = cg_pid_get_path(NULL, pid, &path);
if (r < 0)
return log_error_errno(r, "Failed to get container cgroup path: %m");
r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, path, NULL, &fs);
if (r < 0)
return log_error_errno(r, "Failed to get file system path for container cgroup: %m");
r = chown_cgroup_path(fs, uid_shift);
if (r < 0)
return log_error_errno(r, "Failed to chown() cgroup %s: %m", fs);
return 0;
}
int sync_cgroup(pid_t pid, CGroupUnified unified_requested, uid_t arg_uid_shift) {
_cleanup_free_ char *cgroup = NULL;
char tree[] = "/tmp/unifiedXXXXXX", pid_string[DECIMAL_STR_MAX(pid) + 1];
bool undo_mount = false;
const char *fn;
int r, unified_controller;
unified_controller = cg_unified_controller(SYSTEMD_CGROUP_CONTROLLER);
if (unified_controller < 0)
return log_error_errno(unified_controller, "Failed to determine whether the systemd hierarchy is unified: %m");
if ((unified_controller > 0) == (unified_requested >= CGROUP_UNIFIED_SYSTEMD))
return 0;
/* When the host uses the legacy cgroup setup, but the
* container shall use the unified hierarchy, let's make sure
* we copy the path from the name=systemd hierarchy into the
* unified hierarchy. Similar for the reverse situation. */
r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, pid, &cgroup);
if (r < 0)
return log_error_errno(r, "Failed to get control group of " PID_FMT ": %m", pid);
/* In order to access the unified hierarchy we need to mount it */
if (!mkdtemp(tree))
return log_error_errno(errno, "Failed to generate temporary mount point for unified hierarchy: %m");
if (unified_controller > 0)
r = mount_verbose(LOG_ERR, "cgroup", tree, "cgroup",
MS_NOSUID|MS_NOEXEC|MS_NODEV, "none,name=systemd,xattr");
else
r = mount_verbose(LOG_ERR, "cgroup", tree, "cgroup2",
MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL);
if (r < 0)
goto finish;
undo_mount = true;
/* If nspawn dies abruptly the cgroup hierarchy created below
* its unit isn't cleaned up. So, let's remove it
* https://github.com/systemd/systemd/pull/4223#issuecomment-252519810 */
fn = strjoina(tree, cgroup);
(void) rm_rf(fn, REMOVE_ROOT|REMOVE_ONLY_DIRECTORIES);
fn = strjoina(tree, cgroup, "/cgroup.procs");
(void) mkdir_parents(fn, 0755);
sprintf(pid_string, PID_FMT, pid);
r = write_string_file(fn, pid_string, 0);
if (r < 0) {
log_error_errno(r, "Failed to move process: %m");
goto finish;
}
fn = strjoina(tree, cgroup);
r = chown_cgroup_path(fn, arg_uid_shift);
if (r < 0)
log_error_errno(r, "Failed to chown() cgroup %s: %m", fn);
finish:
if (undo_mount)
(void) umount_verbose(tree);
(void) rmdir(tree);
return r;
}
int create_subcgroup(pid_t pid, CGroupUnified unified_requested) {
_cleanup_free_ char *cgroup = NULL;
const char *child;
int r;
CGroupMask supported;
/* In the unified hierarchy inner nodes may only contain
* subgroups, but not processes. Hence, if we running in the
* unified hierarchy and the container does the same, and we
* did not create a scope unit for the container move us and
* the container into two separate subcgroups. */
if (unified_requested == CGROUP_UNIFIED_NONE)
return 0;
r = cg_unified_controller(SYSTEMD_CGROUP_CONTROLLER);
if (r < 0)
return log_error_errno(r, "Failed to determine whether the systemd controller is unified: %m");
if (r == 0)
return 0;
r = cg_mask_supported(&supported);
if (r < 0)
return log_error_errno(r, "Failed to determine supported controllers: %m");
r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, 0, &cgroup);
if (r < 0)
return log_error_errno(r, "Failed to get our control group: %m");
child = strjoina(cgroup, "/payload");
r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, child, pid);
if (r < 0)
return log_error_errno(r, "Failed to create %s subcgroup: %m", child);
child = strjoina(cgroup, "/supervisor");
r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, child, 0);
if (r < 0)
return log_error_errno(r, "Failed to create %s subcgroup: %m", child);
/* Try to enable as many controllers as possible for the new payload. */
(void) cg_enable_everywhere(supported, supported, cgroup);
return 0;
}
|
