1 /*
   2  * Copyright (c) 2017, 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 "utilities/globalDefinitions.hpp"
  29 #include "memory/allocation.hpp"
  30 #include "runtime/globals.hpp"
  31 #include "runtime/os.hpp"
  32 #include "logging/log.hpp"
  33 #include "osContainer_linux.hpp"
  34 
  35 /*
  36  * PER_CPU_SHARES has been set to 1024 because CPU shares' quota
  37  * is commonly used in cloud frameworks like Kubernetes[1],
  38  * AWS[2] and Mesos[3] in a similar way. They spawn containers with
  39  * --cpu-shares option values scaled by PER_CPU_SHARES. Thus, we do
  40  * the inverse for determining the number of possible available
  41  * CPUs to the JVM inside a container. See JDK-8216366.
  42  *
  43  * [1] https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-cpu
  44  *     In particular:
  45  *        When using Docker:
  46  *          The spec.containers[].resources.requests.cpu is converted to its core value, which is potentially
  47  *          fractional, and multiplied by 1024. The greater of this number or 2 is used as the value of the
  48  *          --cpu-shares flag in the docker run command.
  49  * [2] https://docs.aws.amazon.com/AmazonECS/latest/APIReference/API_ContainerDefinition.html
  50  * [3] https://github.com/apache/mesos/blob/3478e344fb77d931f6122980c6e94cd3913c441d/src/docker/docker.cpp#L648
  51  *     https://github.com/apache/mesos/blob/3478e344fb77d931f6122980c6e94cd3913c441d/src/slave/containerizer/mesos/isolators/cgroups/constants.hpp#L30
  52  */
  53 #define PER_CPU_SHARES 1024
  54 
  55 bool  OSContainer::_is_initialized   = false;
  56 bool  OSContainer::_is_containerized = false;
  57 int   OSContainer::_active_processor_count = 1;
  58 julong _unlimited_memory;
  59 
  60 class CgroupSubsystem: CHeapObj<mtInternal> {
  61  friend class OSContainer;
  62 
  63 
  64  private:
  65     volatile jlong _next_check_counter;
  66 
  67     /* mountinfo contents */
  68     char *_root;
  69     char *_mount_point;
  70 
  71     /* Constructed subsystem directory */
  72     char *_path;
  73 
  74  public:
  75     CgroupSubsystem(char *root, char *mountpoint) {
  76       _root = os::strdup(root);
  77       _mount_point = os::strdup(mountpoint);
  78       _path = NULL;
  79       _next_check_counter = min_jlong;
  80     }
  81 
  82     /*
  83      * Set directory to subsystem specific files based
  84      * on the contents of the mountinfo and cgroup files.
  85      */
  86     void set_subsystem_path(char *cgroup_path) {
  87       char buf[MAXPATHLEN+1];
  88       if (_root != NULL && cgroup_path != NULL) {
  89         if (strcmp(_root, "/") == 0) {
  90           int buflen;
  91           strncpy(buf, _mount_point, MAXPATHLEN);
  92           buf[MAXPATHLEN-1] = '\0';
  93           if (strcmp(cgroup_path,"/") != 0) {
  94             buflen = strlen(buf);
  95             if ((buflen + strlen(cgroup_path)) > (MAXPATHLEN-1)) {
  96               return;
  97             }
  98             strncat(buf, cgroup_path, MAXPATHLEN-buflen);
  99             buf[MAXPATHLEN-1] = '\0';
 100           }
 101           _path = os::strdup(buf);
 102         } else {
 103           if (strcmp(_root, cgroup_path) == 0) {
 104             strncpy(buf, _mount_point, MAXPATHLEN);
 105             buf[MAXPATHLEN-1] = '\0';
 106             _path = os::strdup(buf);
 107           } else {
 108             char *p = strstr(cgroup_path, _root);
 109             if (p != NULL && p == _root) {
 110               if (strlen(cgroup_path) > strlen(_root)) {
 111                 int buflen;
 112                 strncpy(buf, _mount_point, MAXPATHLEN);
 113                 buf[MAXPATHLEN-1] = '\0';
 114                 buflen = strlen(buf);
 115                 if ((buflen + strlen(cgroup_path) - strlen(_root)) > (MAXPATHLEN-1)) {
 116                   return;
 117                 }
 118                 strncat(buf, cgroup_path + strlen(_root), MAXPATHLEN-buflen);
 119                 buf[MAXPATHLEN-1] = '\0';
 120                 _path = os::strdup(buf);
 121               }
 122             }
 123           }
 124         }
 125       }
 126     }
 127 
 128     char *subsystem_path() { return _path; }
 129 
 130     bool cache_has_expired() {
 131       return os::elapsed_counter() > _next_check_counter;
 132     }
 133 
 134     void set_cache_expiry_time(jlong timeout) {
 135       _next_check_counter = os::elapsed_counter() + timeout;
 136     }
 137 };
 138 
 139 class CgroupMemorySubsystem: CgroupSubsystem {
 140  friend class OSContainer;
 141 
 142  private:
 143     /* Some container runtimes set limits via cgroup
 144      * hierarchy. If set to true consider also memory.stat
 145      * file if everything else seems unlimited */
 146     bool _uses_mem_hierarchy;
 147     volatile jlong _memory_limit_in_bytes;
 148 
 149  public:
 150     CgroupMemorySubsystem(char *root, char *mountpoint) : CgroupSubsystem::CgroupSubsystem(root, mountpoint) {
 151       _uses_mem_hierarchy = false;
 152       _memory_limit_in_bytes = -1;
 153 
 154     }
 155 
 156     bool is_hierarchical() { return _uses_mem_hierarchy; }
 157     void set_hierarchical(bool value) { _uses_mem_hierarchy = value; }
 158 
 159     jlong memory_limit_in_bytes() { return _memory_limit_in_bytes; }
 160     void set_memory_limit_in_bytes(jlong value) {
 161       _memory_limit_in_bytes = value;
 162       // max memory limit is unlikely to change, but we want to remain
 163       // responsive to configuration changes. A very short grace time
 164       // between re-read avoids excessive overhead during startup without
 165       // significantly reducing the VMs ability to promptly react to reduced
 166       // memory availability
 167       set_cache_expiry_time(OSCONTAINER_CACHE_TIMEOUT);
 168     }
 169 
 170 };
 171 
 172 CgroupMemorySubsystem* memory = NULL;
 173 CgroupSubsystem* cpuset = NULL;
 174 CgroupSubsystem* cpu = NULL;
 175 CgroupSubsystem* cpuacct = NULL;
 176 
 177 typedef char * cptr;
 178 
 179 PRAGMA_DIAG_PUSH
 180 PRAGMA_FORMAT_NONLITERAL_IGNORED
 181 template <typename T> int subsystem_file_line_contents(CgroupSubsystem* c,
 182                                               const char *filename,
 183                                               const char *matchline,
 184                                               const char *scan_fmt,
 185                                               T returnval) {
 186   FILE *fp = NULL;
 187   char *p;
 188   char file[MAXPATHLEN+1];
 189   char buf[MAXPATHLEN+1];
 190   char discard[MAXPATHLEN+1];
 191   bool found_match = false;
 192 
 193   if (c == NULL) {
 194     log_debug(os, container)("subsystem_file_line_contents: CgroupSubsytem* is NULL");
 195     return OSCONTAINER_ERROR;
 196   }
 197   if (c->subsystem_path() == NULL) {
 198     log_debug(os, container)("subsystem_file_line_contents: subsystem path is NULL");
 199     return OSCONTAINER_ERROR;
 200   }
 201 
 202   strncpy(file, c->subsystem_path(), MAXPATHLEN);
 203   file[MAXPATHLEN-1] = '\0';
 204   int filelen = strlen(file);
 205   if ((filelen + strlen(filename)) > (MAXPATHLEN-1)) {
 206     log_debug(os, container)("File path too long %s, %s", file, filename);
 207     return OSCONTAINER_ERROR;
 208   }
 209   strncat(file, filename, MAXPATHLEN-filelen);
 210   log_trace(os, container)("Path to %s is %s", filename, file);
 211   fp = fopen(file, "r");
 212   if (fp != NULL) {
 213     int err = 0;
 214     while ((p = fgets(buf, MAXPATHLEN, fp)) != NULL) {
 215       found_match = false;
 216       if (matchline == NULL) {
 217         // single-line file case
 218         int matched = sscanf(p, scan_fmt, returnval);
 219         found_match = (matched == 1);
 220       } else {
 221         // multi-line file case
 222         if (strstr(p, matchline) != NULL) {
 223           // discard matchline string prefix
 224           int matched = sscanf(p, scan_fmt, discard, returnval);
 225           found_match = (matched == 2);
 226         } else {
 227           continue; // substring not found
 228         }
 229       }
 230       if (found_match) {
 231         fclose(fp);
 232         return 0;
 233       } else {
 234         err = 1;
 235         log_debug(os, container)("Type %s not found in file %s", scan_fmt, file);
 236       }
 237     }
 238     if (err == 0) {
 239       log_debug(os, container)("Empty file %s", file);
 240     }
 241   } else {
 242     log_debug(os, container)("Open of file %s failed, %s", file, os::strerror(errno));
 243   }
 244   if (fp != NULL)
 245     fclose(fp);
 246   return OSCONTAINER_ERROR;
 247 }
 248 PRAGMA_DIAG_POP
 249 
 250 #define GET_CONTAINER_INFO(return_type, subsystem, filename,              \
 251                            logstring, scan_fmt, variable)                 \
 252   return_type variable;                                                   \
 253 {                                                                         \
 254   int err;                                                                \
 255   err = subsystem_file_line_contents(subsystem,                           \
 256                                      filename,                            \
 257                                      NULL,                                \
 258                                      scan_fmt,                            \
 259                                      &variable);                          \
 260   if (err != 0)                                                           \
 261     return (return_type) OSCONTAINER_ERROR;                               \
 262                                                                           \
 263   log_trace(os, container)(logstring, variable);                          \
 264 }
 265 
 266 #define GET_CONTAINER_INFO_CPTR(return_type, subsystem, filename,         \
 267                                logstring, scan_fmt, variable, bufsize)    \
 268   char variable[bufsize];                                                 \
 269 {                                                                         \
 270   int err;                                                                \
 271   err = subsystem_file_line_contents(subsystem,                           \
 272                                      filename,                            \
 273                                      NULL,                                \
 274                                      scan_fmt,                            \
 275                                      variable);                           \
 276   if (err != 0)                                                           \
 277     return (return_type) NULL;                                            \
 278                                                                           \
 279   log_trace(os, container)(logstring, variable);                          \
 280 }
 281 
 282 #define GET_CONTAINER_INFO_LINE(return_type, subsystem, filename,         \
 283                            matchline, logstring, scan_fmt, variable)      \
 284   return_type variable;                                                   \
 285 {                                                                         \
 286   int err;                                                                \
 287   err = subsystem_file_line_contents(subsystem,                           \
 288                                 filename,                                 \
 289                                 matchline,                                \
 290                                 scan_fmt,                                 \
 291                                 &variable);                               \
 292   if (err != 0)                                                           \
 293     return (return_type) OSCONTAINER_ERROR;                               \
 294                                                                           \
 295   log_trace(os, container)(logstring, variable);                          \
 296 }
 297 
 298 /* init
 299  *
 300  * Initialize the container support and determine if
 301  * we are running under cgroup control.
 302  */
 303 void OSContainer::init() {
 304   FILE *mntinfo = NULL;
 305   FILE *cgroup = NULL;
 306   char buf[MAXPATHLEN+1];
 307   char tmproot[MAXPATHLEN+1];
 308   char tmpmount[MAXPATHLEN+1];
 309   char *p;
 310   jlong mem_limit;
 311 
 312   assert(!_is_initialized, "Initializing OSContainer more than once");
 313 
 314   _is_initialized = true;
 315   _is_containerized = false;
 316 
 317   _unlimited_memory = (LONG_MAX / os::vm_page_size()) * os::vm_page_size();
 318 
 319   log_trace(os, container)("OSContainer::init: Initializing Container Support");
 320   if (!UseContainerSupport) {
 321     log_trace(os, container)("Container Support not enabled");
 322     return;
 323   }
 324 
 325   /*
 326    * Find the cgroup mount point for memory and cpuset
 327    * by reading /proc/self/mountinfo
 328    *
 329    * Example for docker:
 330    * 219 214 0:29 /docker/7208cebd00fa5f2e342b1094f7bed87fa25661471a4637118e65f1c995be8a34 /sys/fs/cgroup/memory ro,nosuid,nodev,noexec,relatime - cgroup cgroup rw,memory
 331    *
 332    * Example for host:
 333    * 34 28 0:29 / /sys/fs/cgroup/memory rw,nosuid,nodev,noexec,relatime shared:16 - cgroup cgroup rw,memory
 334    */
 335   mntinfo = fopen("/proc/self/mountinfo", "r");
 336   if (mntinfo == NULL) {
 337       log_debug(os, container)("Can't open /proc/self/mountinfo, %s",
 338                                os::strerror(errno));
 339       return;
 340   }
 341 
 342   while ((p = fgets(buf, MAXPATHLEN, mntinfo)) != NULL) {
 343     char tmpcgroups[MAXPATHLEN+1];
 344     char *cptr = tmpcgroups;
 345     char *token;
 346 
 347     // mountinfo format is documented at https://www.kernel.org/doc/Documentation/filesystems/proc.txt
 348     if (sscanf(p, "%*d %*d %*d:%*d %s %s %*[^-]- cgroup %*s %s", tmproot, tmpmount, tmpcgroups) != 3) {
 349       continue;
 350     }
 351     while ((token = strsep(&cptr, ",")) != NULL) {
 352       if (strcmp(token, "memory") == 0) {
 353         memory = new CgroupMemorySubsystem(tmproot, tmpmount);
 354       } else if (strcmp(token, "cpuset") == 0) {
 355         cpuset = new CgroupSubsystem(tmproot, tmpmount);
 356       } else if (strcmp(token, "cpu") == 0) {
 357         cpu = new CgroupSubsystem(tmproot, tmpmount);
 358       } else if (strcmp(token, "cpuacct") == 0) {
 359         cpuacct= new CgroupSubsystem(tmproot, tmpmount);
 360       }
 361     }
 362   }
 363 
 364   fclose(mntinfo);
 365 
 366   if (memory == NULL) {
 367     log_debug(os, container)("Required cgroup memory subsystem not found");
 368     return;
 369   }
 370   if (cpuset == NULL) {
 371     log_debug(os, container)("Required cgroup cpuset subsystem not found");
 372     return;
 373   }
 374   if (cpu == NULL) {
 375     log_debug(os, container)("Required cgroup cpu subsystem not found");
 376     return;
 377   }
 378   if (cpuacct == NULL) {
 379     log_debug(os, container)("Required cgroup cpuacct subsystem not found");
 380     return;
 381   }
 382 
 383   /*
 384    * Read /proc/self/cgroup and map host mount point to
 385    * local one via /proc/self/mountinfo content above
 386    *
 387    * Docker example:
 388    * 5:memory:/docker/6558aed8fc662b194323ceab5b964f69cf36b3e8af877a14b80256e93aecb044
 389    *
 390    * Host example:
 391    * 5:memory:/user.slice
 392    *
 393    * Construct a path to the process specific memory and cpuset
 394    * cgroup directory.
 395    *
 396    * For a container running under Docker from memory example above
 397    * the paths would be:
 398    *
 399    * /sys/fs/cgroup/memory
 400    *
 401    * For a Host from memory example above the path would be:
 402    *
 403    * /sys/fs/cgroup/memory/user.slice
 404    *
 405    */
 406   cgroup = fopen("/proc/self/cgroup", "r");
 407   if (cgroup == NULL) {
 408     log_debug(os, container)("Can't open /proc/self/cgroup, %s",
 409                              os::strerror(errno));
 410     return;
 411   }
 412 
 413   while ((p = fgets(buf, MAXPATHLEN, cgroup)) != NULL) {
 414     char *controllers;
 415     char *token;
 416     char *base;
 417 
 418     /* Skip cgroup number */
 419     strsep(&p, ":");
 420     /* Get controllers and base */
 421     controllers = strsep(&p, ":");
 422     base = strsep(&p, "\n");
 423 
 424     if (controllers == NULL) {
 425       continue;
 426     }
 427 
 428     while ((token = strsep(&controllers, ",")) != NULL) {
 429       if (strcmp(token, "memory") == 0) {
 430         memory->set_subsystem_path(base);
 431         jlong hierarchy = uses_mem_hierarchy();
 432         if (hierarchy > 0) {
 433           memory->set_hierarchical(true);
 434         }
 435       } else if (strcmp(token, "cpuset") == 0) {
 436         cpuset->set_subsystem_path(base);
 437       } else if (strcmp(token, "cpu") == 0) {
 438         cpu->set_subsystem_path(base);
 439       } else if (strcmp(token, "cpuacct") == 0) {
 440         cpuacct->set_subsystem_path(base);
 441       }
 442     }
 443   }
 444 
 445   fclose(cgroup);
 446 
 447   // We need to update the amount of physical memory now that
 448   // command line arguments have been processed.
 449   if ((mem_limit = memory_limit_in_bytes()) > 0) {
 450     os::Linux::set_physical_memory(mem_limit);
 451     log_info(os, container)("Memory Limit is: " JLONG_FORMAT, mem_limit);
 452   }
 453 
 454   _is_containerized = true;
 455 
 456 }
 457 
 458 const char * OSContainer::container_type() {
 459   if (is_containerized()) {
 460     return "cgroupv1";
 461   } else {
 462     return NULL;
 463   }
 464 }
 465 
 466 /* uses_mem_hierarchy
 467  *
 468  * Return whether or not hierarchical cgroup accounting is being
 469  * done.
 470  *
 471  * return:
 472  *    A number > 0 if true, or
 473  *    OSCONTAINER_ERROR for not supported
 474  */
 475 jlong OSContainer::uses_mem_hierarchy() {
 476   GET_CONTAINER_INFO(jlong, memory, "/memory.use_hierarchy",
 477                     "Use Hierarchy is: " JLONG_FORMAT, JLONG_FORMAT, use_hierarchy);
 478   return use_hierarchy;
 479 }
 480 
 481 
 482 /* memory_limit_in_bytes
 483  *
 484  * Return the limit of available memory for this process.
 485  *
 486  * return:
 487  *    memory limit in bytes or
 488  *    -1 for unlimited
 489  *    OSCONTAINER_ERROR for not supported
 490  */
 491 jlong OSContainer::memory_limit_in_bytes() {
 492   if (!memory->cache_has_expired()) {
 493     return memory->memory_limit_in_bytes();
 494   }
 495   jlong memory_limit = read_memory_limit_in_bytes();
 496   // Update CgroupMemorySubsystem to avoid re-reading container settings too often
 497   memory->set_memory_limit_in_bytes(memory_limit);
 498   return memory_limit;
 499 }
 500 
 501 jlong OSContainer::read_memory_limit_in_bytes() {
 502   GET_CONTAINER_INFO(julong, memory, "/memory.limit_in_bytes",
 503                      "Memory Limit is: " JULONG_FORMAT, JULONG_FORMAT, memlimit);
 504 
 505   if (memlimit >= _unlimited_memory) {
 506     log_trace(os, container)("Non-Hierarchical Memory Limit is: Unlimited");
 507     if (memory->is_hierarchical()) {
 508       const char* matchline = "hierarchical_memory_limit";
 509       const char* format = "%s " JULONG_FORMAT;
 510       GET_CONTAINER_INFO_LINE(julong, memory, "/memory.stat", matchline,
 511                              "Hierarchical Memory Limit is: " JULONG_FORMAT, format, hier_memlimit)
 512       if (hier_memlimit >= _unlimited_memory) {
 513         log_trace(os, container)("Hierarchical Memory Limit is: Unlimited");
 514       } else {
 515         return (jlong)hier_memlimit;
 516       }
 517     }
 518     return (jlong)-1;
 519   }
 520   else {
 521     return (jlong)memlimit;
 522   }
 523 }
 524 
 525 jlong OSContainer::memory_and_swap_limit_in_bytes() {
 526   GET_CONTAINER_INFO(julong, memory, "/memory.memsw.limit_in_bytes",
 527                      "Memory and Swap Limit is: " JULONG_FORMAT, JULONG_FORMAT, memswlimit);
 528   if (memswlimit >= _unlimited_memory) {
 529     log_trace(os, container)("Non-Hierarchical Memory and Swap Limit is: Unlimited");
 530     if (memory->is_hierarchical()) {
 531       const char* matchline = "hierarchical_memsw_limit";
 532       const char* format = "%s " JULONG_FORMAT;
 533       GET_CONTAINER_INFO_LINE(julong, memory, "/memory.stat", matchline,
 534                              "Hierarchical Memory and Swap Limit is : " JULONG_FORMAT, format, hier_memlimit)
 535       if (hier_memlimit >= _unlimited_memory) {
 536         log_trace(os, container)("Hierarchical Memory and Swap Limit is: Unlimited");
 537       } else {
 538         return (jlong)hier_memlimit;
 539       }
 540     }
 541     return (jlong)-1;
 542   } else {
 543     return (jlong)memswlimit;
 544   }
 545 }
 546 
 547 jlong OSContainer::memory_soft_limit_in_bytes() {
 548   GET_CONTAINER_INFO(julong, memory, "/memory.soft_limit_in_bytes",
 549                      "Memory Soft Limit is: " JULONG_FORMAT, JULONG_FORMAT, memsoftlimit);
 550   if (memsoftlimit >= _unlimited_memory) {
 551     log_trace(os, container)("Memory Soft Limit is: Unlimited");
 552     return (jlong)-1;
 553   } else {
 554     return (jlong)memsoftlimit;
 555   }
 556 }
 557 
 558 /* memory_usage_in_bytes
 559  *
 560  * Return the amount of used memory for this process.
 561  *
 562  * return:
 563  *    memory usage in bytes or
 564  *    -1 for unlimited
 565  *    OSCONTAINER_ERROR for not supported
 566  */
 567 jlong OSContainer::memory_usage_in_bytes() {
 568   GET_CONTAINER_INFO(jlong, memory, "/memory.usage_in_bytes",
 569                      "Memory Usage is: " JLONG_FORMAT, JLONG_FORMAT, memusage);
 570   return memusage;
 571 }
 572 
 573 /* memory_max_usage_in_bytes
 574  *
 575  * Return the maximum amount of used memory for this process.
 576  *
 577  * return:
 578  *    max memory usage in bytes or
 579  *    OSCONTAINER_ERROR for not supported
 580  */
 581 jlong OSContainer::memory_max_usage_in_bytes() {
 582   GET_CONTAINER_INFO(jlong, memory, "/memory.max_usage_in_bytes",
 583                      "Maximum Memory Usage is: " JLONG_FORMAT, JLONG_FORMAT, memmaxusage);
 584   return memmaxusage;
 585 }
 586 
 587 /* active_processor_count
 588  *
 589  * Calculate an appropriate number of active processors for the
 590  * VM to use based on these three inputs.
 591  *
 592  * cpu affinity
 593  * cgroup cpu quota & cpu period
 594  * cgroup cpu shares
 595  *
 596  * Algorithm:
 597  *
 598  * Determine the number of available CPUs from sched_getaffinity
 599  *
 600  * If user specified a quota (quota != -1), calculate the number of
 601  * required CPUs by dividing quota by period.
 602  *
 603  * If shares are in effect (shares != -1), calculate the number
 604  * of CPUs required for the shares by dividing the share value
 605  * by PER_CPU_SHARES.
 606  *
 607  * All results of division are rounded up to the next whole number.
 608  *
 609  * If neither shares or quotas have been specified, return the
 610  * number of active processors in the system.
 611  *
 612  * If both shares and quotas have been specified, the results are
 613  * based on the flag PreferContainerQuotaForCPUCount.  If true,
 614  * return the quota value.  If false return the smallest value
 615  * between shares or quotas.
 616  *
 617  * If shares and/or quotas have been specified, the resulting number
 618  * returned will never exceed the number of active processors.
 619  *
 620  * return:
 621  *    number of CPUs
 622  */
 623 int OSContainer::active_processor_count() {
 624   int quota_count = 0, share_count = 0;
 625   int cpu_count, limit_count;
 626   int result;
 627 
 628   // We use a cache with a timeout to avoid performing expensive
 629   // computations in the event this function is called frequently.
 630   // [See 8227006].
 631   if (!cpu->cache_has_expired()) {
 632     log_trace(os, container)("OSContainer::active_processor_count (cached): %d", OSContainer::_active_processor_count);
 633     return OSContainer::_active_processor_count;
 634   }
 635 
 636   cpu_count = limit_count = os::Linux::active_processor_count();
 637   int quota  = cpu_quota();
 638   int period = cpu_period();
 639   int share  = cpu_shares();
 640 
 641   if (quota > -1 && period > 0) {
 642     quota_count = ceilf((float)quota / (float)period);
 643     log_trace(os, container)("CPU Quota count based on quota/period: %d", quota_count);
 644   }
 645   if (share > -1) {
 646     share_count = ceilf((float)share / (float)PER_CPU_SHARES);
 647     log_trace(os, container)("CPU Share count based on shares: %d", share_count);
 648   }
 649 
 650   // If both shares and quotas are setup results depend
 651   // on flag PreferContainerQuotaForCPUCount.
 652   // If true, limit CPU count to quota
 653   // If false, use minimum of shares and quotas
 654   if (quota_count !=0 && share_count != 0) {
 655     if (PreferContainerQuotaForCPUCount) {
 656       limit_count = quota_count;
 657     } else {
 658       limit_count = MIN2(quota_count, share_count);
 659     }
 660   } else if (quota_count != 0) {
 661     limit_count = quota_count;
 662   } else if (share_count != 0) {
 663     limit_count = share_count;
 664   }
 665 
 666   result = MIN2(cpu_count, limit_count);
 667   log_trace(os, container)("OSContainer::active_processor_count: %d", result);
 668 
 669   // Update the value and reset the cache timeout
 670   OSContainer::_active_processor_count = result;
 671   cpu->set_cache_expiry_time(OSCONTAINER_CACHE_TIMEOUT);
 672 
 673   return result;
 674 }
 675 
 676 char * OSContainer::cpu_cpuset_cpus() {
 677   GET_CONTAINER_INFO_CPTR(cptr, cpuset, "/cpuset.cpus",
 678                      "cpuset.cpus is: %s", "%1023s", cpus, 1024);
 679   return os::strdup(cpus);
 680 }
 681 
 682 char * OSContainer::cpu_cpuset_memory_nodes() {
 683   GET_CONTAINER_INFO_CPTR(cptr, cpuset, "/cpuset.mems",
 684                      "cpuset.mems is: %s", "%1023s", mems, 1024);
 685   return os::strdup(mems);
 686 }
 687 
 688 /* cpu_quota
 689  *
 690  * Return the number of milliseconds per period
 691  * process is guaranteed to run.
 692  *
 693  * return:
 694  *    quota time in milliseconds
 695  *    -1 for no quota
 696  *    OSCONTAINER_ERROR for not supported
 697  */
 698 int OSContainer::cpu_quota() {
 699   GET_CONTAINER_INFO(int, cpu, "/cpu.cfs_quota_us",
 700                      "CPU Quota is: %d", "%d", quota);
 701   return quota;
 702 }
 703 
 704 int OSContainer::cpu_period() {
 705   GET_CONTAINER_INFO(int, cpu, "/cpu.cfs_period_us",
 706                      "CPU Period is: %d", "%d", period);
 707   return period;
 708 }
 709 
 710 /* cpu_shares
 711  *
 712  * Return the amount of cpu shares available to the process
 713  *
 714  * return:
 715  *    Share number (typically a number relative to 1024)
 716  *                 (2048 typically expresses 2 CPUs worth of processing)
 717  *    -1 for no share setup
 718  *    OSCONTAINER_ERROR for not supported
 719  */
 720 int OSContainer::cpu_shares() {
 721   GET_CONTAINER_INFO(int, cpu, "/cpu.shares",
 722                      "CPU Shares is: %d", "%d", shares);
 723   // Convert 1024 to no shares setup
 724   if (shares == 1024) return -1;
 725 
 726   return shares;
 727 }