1 /*
2 * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #include <string.h>
26 #include <math.h>
27 #include <errno.h>
28 #include "cgroupSubsystem_linux.hpp"
29 #include "cgroupV1Subsystem_linux.hpp"
30 #include "logging/log.hpp"
31 #include "memory/allocation.hpp"
32 #include "runtime/globals.hpp"
33 #include "runtime/os.hpp"
34 #include "utilities/globalDefinitions.hpp"
35
36 CgroupSubsystem* CgroupSubsystemFactory::create() {
37 CgroupV1MemoryController* memory = NULL;
38 CgroupV1Controller* cpuset = NULL;
39 CgroupV1Controller* cpu = NULL;
40 CgroupV1Controller* cpuacct = NULL;
41 FILE *mntinfo = NULL;
42 FILE *cgroup = NULL;
43 char buf[MAXPATHLEN+1];
44 char tmproot[MAXPATHLEN+1];
45 char tmpmount[MAXPATHLEN+1];
46 char *p;
47
48 /*
49 * Find the cgroup mount point for memory and cpuset
50 * by reading /proc/self/mountinfo
51 *
52 * Example for docker:
53 * 219 214 0:29 /docker/7208cebd00fa5f2e342b1094f7bed87fa25661471a4637118e65f1c995be8a34 /sys/fs/cgroup/memory ro,nosuid,nodev,noexec,relatime - cgroup cgroup rw,memory
54 *
55 * Example for host:
56 * 34 28 0:29 / /sys/fs/cgroup/memory rw,nosuid,nodev,noexec,relatime shared:16 - cgroup cgroup rw,memory
57 */
58 mntinfo = fopen("/proc/self/mountinfo", "r");
59 if (mntinfo == NULL) {
60 log_debug(os, container)("Can't open /proc/self/mountinfo, %s",
61 os::strerror(errno));
62 return NULL;
63 }
64
65 while ((p = fgets(buf, MAXPATHLEN, mntinfo)) != NULL) {
66 char tmpcgroups[MAXPATHLEN+1];
67 char *cptr = tmpcgroups;
68 char *token;
69
70 // mountinfo format is documented at https://www.kernel.org/doc/Documentation/filesystems/proc.txt
71 if (sscanf(p, "%*d %*d %*d:%*d %s %s %*[^-]- cgroup %*s %s", tmproot, tmpmount, tmpcgroups) != 3) {
72 continue;
73 }
74 while ((token = strsep(&cptr, ",")) != NULL) {
75 if (strcmp(token, "memory") == 0) {
76 memory = new CgroupV1MemoryController(tmproot, tmpmount);
77 } else if (strcmp(token, "cpuset") == 0) {
78 cpuset = new CgroupV1Controller(tmproot, tmpmount);
79 } else if (strcmp(token, "cpu") == 0) {
80 cpu = new CgroupV1Controller(tmproot, tmpmount);
81 } else if (strcmp(token, "cpuacct") == 0) {
82 cpuacct= new CgroupV1Controller(tmproot, tmpmount);
83 }
84 }
85 }
86
87 fclose(mntinfo);
88
89 if (memory == NULL) {
90 log_debug(os, container)("Required cgroup memory subsystem not found");
91 return NULL;
92 }
93 if (cpuset == NULL) {
94 log_debug(os, container)("Required cgroup cpuset subsystem not found");
95 return NULL;
96 }
97 if (cpu == NULL) {
98 log_debug(os, container)("Required cgroup cpu subsystem not found");
99 return NULL;
100 }
101 if (cpuacct == NULL) {
102 log_debug(os, container)("Required cgroup cpuacct subsystem not found");
103 return NULL;
104 }
105
106 /*
107 * Read /proc/self/cgroup and map host mount point to
108 * local one via /proc/self/mountinfo content above
109 *
110 * Docker example:
111 * 5:memory:/docker/6558aed8fc662b194323ceab5b964f69cf36b3e8af877a14b80256e93aecb044
112 *
113 * Host example:
114 * 5:memory:/user.slice
115 *
116 * Construct a path to the process specific memory and cpuset
117 * cgroup directory.
118 *
119 * For a container running under Docker from memory example above
120 * the paths would be:
121 *
122 * /sys/fs/cgroup/memory
123 *
124 * For a Host from memory example above the path would be:
125 *
126 * /sys/fs/cgroup/memory/user.slice
127 *
128 */
129 cgroup = fopen("/proc/self/cgroup", "r");
130 if (cgroup == NULL) {
131 log_debug(os, container)("Can't open /proc/self/cgroup, %s",
132 os::strerror(errno));
133 return NULL;
134 }
135
136 while ((p = fgets(buf, MAXPATHLEN, cgroup)) != NULL) {
137 char *controllers;
138 char *token;
139 char *base;
140
141 /* Skip cgroup number */
142 strsep(&p, ":");
143 /* Get controllers and base */
144 controllers = strsep(&p, ":");
145 base = strsep(&p, "\n");
146
147 if (controllers == NULL) {
148 continue;
149 }
150
151 while ((token = strsep(&controllers, ",")) != NULL) {
152 if (strcmp(token, "memory") == 0) {
153 memory->set_subsystem_path(base);
154 } else if (strcmp(token, "cpuset") == 0) {
155 cpuset->set_subsystem_path(base);
156 } else if (strcmp(token, "cpu") == 0) {
157 cpu->set_subsystem_path(base);
158 } else if (strcmp(token, "cpuacct") == 0) {
159 cpuacct->set_subsystem_path(base);
160 }
161 }
162 }
163
164 fclose(cgroup);
165 return new CgroupV1Subsystem(cpuset, cpu, cpuacct, memory);
166 }
167
168 /* active_processor_count
169 *
170 * Calculate an appropriate number of active processors for the
171 * VM to use based on these three inputs.
172 *
173 * cpu affinity
174 * cgroup cpu quota & cpu period
175 * cgroup cpu shares
176 *
177 * Algorithm:
178 *
179 * Determine the number of available CPUs from sched_getaffinity
180 *
181 * If user specified a quota (quota != -1), calculate the number of
182 * required CPUs by dividing quota by period.
183 *
184 * If shares are in effect (shares != -1), calculate the number
185 * of CPUs required for the shares by dividing the share value
186 * by PER_CPU_SHARES.
187 *
188 * All results of division are rounded up to the next whole number.
189 *
190 * If neither shares or quotas have been specified, return the
191 * number of active processors in the system.
192 *
193 * If both shares and quotas have been specified, the results are
194 * based on the flag PreferContainerQuotaForCPUCount. If true,
195 * return the quota value. If false return the smallest value
196 * between shares or quotas.
197 *
198 * If shares and/or quotas have been specified, the resulting number
199 * returned will never exceed the number of active processors.
200 *
201 * return:
202 * number of CPUs
203 */
204 int CgroupSubsystem::active_processor_count(int physical_proc_count) {
205 int quota_count = 0, share_count = 0;
206 int cpu_count, limit_count;
207 int result;
208
209 cpu_count = limit_count = physical_proc_count;
210 int quota = cpu_quota();
211 int period = cpu_period();
212 int share = cpu_shares();
213
214 if (quota > -1 && period > 0) {
215 quota_count = ceilf((float)quota / (float)period);
216 log_trace(os, container)("CPU Quota count based on quota/period: %d", quota_count);
217 }
218 if (share > -1) {
219 share_count = ceilf((float)share / (float)PER_CPU_SHARES);
220 log_trace(os, container)("CPU Share count based on shares: %d", share_count);
221 }
222
223 // If both shares and quotas are setup results depend
224 // on flag PreferContainerQuotaForCPUCount.
225 // If true, limit CPU count to quota
226 // If false, use minimum of shares and quotas
227 if (quota_count !=0 && share_count != 0) {
228 if (PreferContainerQuotaForCPUCount) {
229 limit_count = quota_count;
230 } else {
231 limit_count = MIN2(quota_count, share_count);
232 }
233 } else if (quota_count != 0) {
234 limit_count = quota_count;
235 } else if (share_count != 0) {
236 limit_count = share_count;
237 }
238
239 result = MIN2(cpu_count, limit_count);
240 log_trace(os, container)("OSContainer::active_processor_count: %d", result);
241 return result;
242 }