cgroup.c 26.8 KB
Newer Older
1 2 3
/*
 * cgroup.c: Tools for managing cgroups
 *
4
 * Copyright (C) 2010 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15
 * Copyright IBM Corp. 2008
 *
 * See COPYING.LIB for the License of this software
 *
 * Authors:
 *  Dan Smith <danms@us.ibm.com>
 */
#include <config.h>

#include <stdio.h>
#include <stdint.h>
16
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
17
# include <mntent.h>
D
Daniel P. Berrange 已提交
18
#endif
19 20 21 22 23 24 25 26
#include <fcntl.h>
#include <string.h>
#include <errno.h>
#include <stdlib.h>
#include <stdbool.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <libgen.h>
27
#include <dirent.h>
28 29 30 31 32

#include "internal.h"
#include "util.h"
#include "memory.h"
#include "cgroup.h"
33
#include "logging.h"
34 35 36

#define CGROUP_MAX_VAL 512

37
VIR_ENUM_IMPL(virCgroupController, VIR_CGROUP_CONTROLLER_LAST,
R
Ryota Ozaki 已提交
38 39
              "cpu", "cpuacct", "cpuset", "memory", "devices",
              "freezer");
40 41 42 43 44 45

struct virCgroupController {
    int type;
    char *mountPoint;
    char *placement;
};
46 47 48

struct virCgroup {
    char *path;
49 50

    struct virCgroupController controllers[VIR_CGROUP_CONTROLLER_LAST];
51 52 53 54 55 56 57 58 59
};

/**
 * virCgroupFree:
 *
 * @group: The group structure to free
 */
void virCgroupFree(virCgroupPtr *group)
{
60 61 62 63 64 65 66 67
    int i;

    if (*group == NULL)
        return;

    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
        VIR_FREE((*group)->controllers[i].mountPoint);
        VIR_FREE((*group)->controllers[i].placement);
68
    }
69 70 71

    VIR_FREE((*group)->path);
    VIR_FREE(*group);
72 73
}

74
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
75 76 77 78 79
/*
 * Process /proc/mounts figuring out what controllers are
 * mounted and where
 */
static int virCgroupDetectMounts(virCgroupPtr group)
80
{
81
    int i;
82
    FILE *mounts = NULL;
83 84 85 86 87
    struct mntent entry;
    char buf[CGROUP_MAX_VAL];

    mounts = fopen("/proc/mounts", "r");
    if (mounts == NULL) {
88
        VIR_ERROR0(_("Unable to open /proc/mounts"));
89
        return -ENOENT;
90 91 92
    }

    while (getmntent_r(mounts, &entry, buf, sizeof(buf)) != NULL) {
93 94
        if (STRNEQ(entry.mnt_type, "cgroup"))
            continue;
95

96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114
        for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
            const char *typestr = virCgroupControllerTypeToString(i);
            int typelen = strlen(typestr);
            char *tmp = entry.mnt_opts;
            while (tmp) {
                char *next = strchr(tmp, ',');
                int len;
                if (next) {
                    len = next-tmp;
                    next++;
                } else {
                    len = strlen(tmp);
                }
                if (typelen == len && STREQLEN(typestr, tmp, len) &&
                    !(group->controllers[i].mountPoint = strdup(entry.mnt_dir)))
                    goto no_memory;
                tmp = next;
            }
        }
115 116 117 118
    }

    fclose(mounts);

119
    return 0;
120

121 122 123 124
no_memory:
    if (mounts)
        fclose(mounts);
    return -ENOMEM;
125 126
}

127 128 129 130 131

/*
 * Process /proc/self/cgroup figuring out what cgroup
 * sub-path the current process is assigned to. ie not
 * neccessarily in the root
132
 */
133
static int virCgroupDetectPlacement(virCgroupPtr group)
134 135
{
    int i;
136 137
    FILE *mapping  = NULL;
    char line[1024];
138

139 140
    mapping = fopen("/proc/self/cgroup", "r");
    if (mapping == NULL) {
141
        VIR_ERROR0(_("Unable to open /proc/self/cgroup"));
142
        return -ENOENT;
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
    while (fgets(line, sizeof(line), mapping) != NULL) {
        char *controllers = strchr(line, ':');
        char *path = controllers ? strchr(controllers+1, ':') : NULL;
        char *nl = path ? strchr(path, '\n') : NULL;

        if (!controllers || !path)
            continue;

        if (nl)
            *nl = '\0';

        *path = '\0';
        controllers++;
        path++;

        for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
            const char *typestr = virCgroupControllerTypeToString(i);
            int typelen = strlen(typestr);
            char *tmp = controllers;
            while (tmp) {
                char *next = strchr(tmp, ',');
                int len;
                if (next) {
                    len = next-tmp;
                    next++;
                } else {
                    len = strlen(tmp);
                }
                if (typelen == len && STREQLEN(typestr, tmp, len) &&
                    !(group->controllers[i].placement = strdup(STREQ(path, "/") ? "" : path)))
                    goto no_memory;

                tmp = next;
            }
        }
    }

    fclose(mapping);

184
    return 0;
185 186

no_memory:
187
    fclose(mapping);
188 189
    return -ENOMEM;

190 191
}

192
static int virCgroupDetect(virCgroupPtr group)
193
{
194 195 196
    int any = 0;
    int rc;
    int i;
197

198 199
    rc = virCgroupDetectMounts(group);
    if (rc < 0) {
200
        VIR_ERROR(_("Failed to detect mounts for %s"), group->path);
201
        return rc;
202 203
    }

204 205 206 207 208 209 210
    /* Check that at least 1 controller is available */
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
        if (group->controllers[i].mountPoint != NULL)
            any = 1;
    }
    if (!any)
        return -ENXIO;
211 212


213
    rc = virCgroupDetectPlacement(group);
214

215 216 217 218 219
    if (rc == 0) {
        /* Check that for every mounted controller, we found our placement */
        for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
            if (!group->controllers[i].mountPoint)
                continue;
220

221
            if (!group->controllers[i].placement) {
222
                VIR_ERROR(_("Could not find placement for controller %s at %s"),
223 224 225 226 227
                          virCgroupControllerTypeToString(i),
                          group->controllers[i].placement);
                rc = -ENOENT;
                break;
            }
228

229 230 231 232 233 234
            VIR_DEBUG("Detected mount/mapping %i:%s at %s in %s", i,
                      virCgroupControllerTypeToString(i),
                      group->controllers[i].mountPoint,
                      group->controllers[i].placement);
        }
    } else {
235
        VIR_ERROR(_("Failed to detect mapping for %s"), group->path);
236 237 238 239
    }

    return rc;
}
D
Daniel P. Berrange 已提交
240
#endif
241

242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264

static int virCgroupPathOfController(virCgroupPtr group,
                                     int controller,
                                     const char *key,
                                     char **path)
{
    if (group->controllers[controller].mountPoint == NULL)
        return -ENOENT;

    if (group->controllers[controller].placement == NULL)
        return -ENOENT;

    if (virAsprintf(path, "%s%s%s/%s",
                    group->controllers[controller].mountPoint,
                    group->controllers[controller].placement,
                    STREQ(group->path, "/") ? "" : group->path,
                    key ? key : "") == -1)
        return -ENOMEM;

    return 0;
}


265
static int virCgroupSetValueStr(virCgroupPtr group,
266
                                int controller,
267 268 269 270 271 272
                                const char *key,
                                const char *value)
{
    int rc = 0;
    char *keypath = NULL;

273
    rc = virCgroupPathOfController(group, controller, key, &keypath);
274 275 276
    if (rc != 0)
        return rc;

277
    VIR_DEBUG("Set value '%s' to '%s'", keypath, value);
278
    rc = virFileWriteStr(keypath, value);
279 280 281
    if (rc < 0) {
        DEBUG("Failed to write value '%s': %m", value);
        rc = -errno;
282 283
    } else {
        rc = 0;
284 285 286 287 288 289 290 291
    }

    VIR_FREE(keypath);

    return rc;
}

static int virCgroupGetValueStr(virCgroupPtr group,
292
                                int controller,
293 294 295 296 297 298
                                const char *key,
                                char **value)
{
    int rc;
    char *keypath = NULL;

299
    *value = NULL;
300

301
    rc = virCgroupPathOfController(group, controller, key, &keypath);
302 303 304 305 306
    if (rc != 0) {
        DEBUG("No path of %s, %s", group->path, key);
        return rc;
    }

307
    VIR_DEBUG("Get value %s", keypath);
308

309
    rc = virFileReadAll(keypath, 1024, value);
310 311 312
    if (rc < 0) {
        DEBUG("Failed to read %s: %m\n", keypath);
        rc = -errno;
313
    } else {
314 315 316 317
        /* Terminated with '\n' has sometimes harmful effects to the caller */
        char *p = strchr(*value, '\n');
        if (p) *p = '\0';

318
        rc = 0;
319 320 321 322 323 324 325
    }

    VIR_FREE(keypath);

    return rc;
}

326
static int virCgroupSetValueU64(virCgroupPtr group,
327
                                int controller,
328
                                const char *key,
D
Daniel P. Berrange 已提交
329
                                unsigned long long int value)
330 331 332 333
{
    char *strval = NULL;
    int rc;

D
Daniel P. Berrange 已提交
334
    if (virAsprintf(&strval, "%llu", value) == -1)
335 336
        return -ENOMEM;

337
    rc = virCgroupSetValueStr(group, controller, key, strval);
338 339 340 341 342 343 344

    VIR_FREE(strval);

    return rc;
}


345 346 347 348
#if 0
/* This is included for completeness, but not yet used */

static int virCgroupSetValueI64(virCgroupPtr group,
349
                                int controller,
350
                                const char *key,
D
Daniel P. Berrange 已提交
351
                                long long int value)
352 353 354 355
{
    char *strval = NULL;
    int rc;

D
Daniel P. Berrange 已提交
356
    if (virAsprintf(&strval, "%lld", value) == -1)
357 358
        return -ENOMEM;

359
    rc = virCgroupSetValueStr(group, controller, key, strval);
360 361 362 363 364 365 366

    VIR_FREE(strval);

    return rc;
}

static int virCgroupGetValueI64(virCgroupPtr group,
367
                                int controller,
368
                                const char *key,
D
Daniel P. Berrange 已提交
369
                                long long int *value)
370 371 372 373
{
    char *strval = NULL;
    int rc = 0;

374
    rc = virCgroupGetValueStr(group, controller, key, &strval);
375 376 377
    if (rc != 0)
        goto out;

378
    if (virStrToLong_ll(strval, NULL, 10, value) < 0)
379 380 381 382 383 384
        rc = -EINVAL;
out:
    VIR_FREE(strval);

    return rc;
}
385
#endif
386

387
static int virCgroupGetValueU64(virCgroupPtr group,
388
                                int controller,
389
                                const char *key,
D
Daniel P. Berrange 已提交
390
                                unsigned long long int *value)
391 392 393 394
{
    char *strval = NULL;
    int rc = 0;

395
    rc = virCgroupGetValueStr(group, controller, key, &strval);
396 397 398
    if (rc != 0)
        goto out;

D
Daniel P. Berrange 已提交
399
    if (virStrToLong_ull(strval, NULL, 10, value) < 0)
400 401 402 403 404 405 406 407
        rc = -EINVAL;
out:
    VIR_FREE(strval);

    return rc;
}


408
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
409
static int virCgroupCpuSetInherit(virCgroupPtr parent, virCgroupPtr group)
410 411 412 413 414 415 416 417
{
    int i;
    int rc = 0;
    const char *inherit_values[] = {
        "cpuset.cpus",
        "cpuset.mems",
    };

418 419 420
    VIR_DEBUG("Setting up inheritance %s -> %s", parent->path, group->path);
    for (i = 0; i < ARRAY_CARDINALITY(inherit_values) ; i++) {
        char *value;
421

422 423 424 425
        rc = virCgroupGetValueStr(parent,
                                  VIR_CGROUP_CONTROLLER_CPUSET,
                                  inherit_values[i],
                                  &value);
426
        if (rc != 0) {
427
            VIR_ERROR(_("Failed to get %s %d"), inherit_values[i], rc);
428 429 430 431 432 433 434 435 436
            break;
        }

        VIR_DEBUG("Inherit %s = %s", inherit_values[i], value);

        rc = virCgroupSetValueStr(group,
                                  VIR_CGROUP_CONTROLLER_CPUSET,
                                  inherit_values[i],
                                  value);
437
        VIR_FREE(value);
438 439

        if (rc != 0) {
440
            VIR_ERROR(_("Failed to set %s %d"), inherit_values[i], rc);
441 442 443 444 445 446 447
            break;
        }
    }

    return rc;
}

448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
static int virCgroupSetMemoryUseHierarchy(virCgroupPtr group)
{
    int rc = 0;
    unsigned long long value;
    const char *filename = "memory.use_hierarchy";

    rc = virCgroupGetValueU64(group,
                              VIR_CGROUP_CONTROLLER_MEMORY,
                              filename, &value);
    if (rc != 0) {
        VIR_ERROR(_("Failed to read %s/%s (%d)"), group->path, filename, rc);
        return rc;
    }

    /* Setting twice causes error, so if already enabled, skip setting */
    if (value == 1)
        return 0;

    VIR_DEBUG("Setting up %s/%s", group->path, filename);
    rc = virCgroupSetValueU64(group,
                              VIR_CGROUP_CONTROLLER_MEMORY,
                              filename, 1);

    if (rc != 0) {
        VIR_ERROR(_("Failed to set %s/%s (%d)"), group->path, filename, rc);
    }

    return rc;
}

static int virCgroupMakeGroup(virCgroupPtr parent, virCgroupPtr group,
                              int create, bool memory_hierarchy)
480 481 482 483
{
    int i;
    int rc = 0;

484
    VIR_DEBUG("Make group %s", group->path);
485
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
486 487
        char *path = NULL;

488 489
        /* Skip over controllers that aren't mounted */
        if (!group->controllers[i].mountPoint)
490 491
            continue;

492 493 494
        rc = virCgroupPathOfController(group, i, "", &path);
        if (rc < 0)
            return rc;
495

496
        VIR_DEBUG("Make controller %s", path);
497
        if (access(path, F_OK) != 0) {
498 499
            if (!create ||
                mkdir(path, 0755) < 0) {
500 501 502 503
                rc = -errno;
                VIR_FREE(path);
                break;
            }
504 505 506 507
            if (group->controllers[VIR_CGROUP_CONTROLLER_CPUSET].mountPoint != NULL &&
                (i == VIR_CGROUP_CONTROLLER_CPUSET ||
                 STREQ(group->controllers[i].mountPoint, group->controllers[VIR_CGROUP_CONTROLLER_CPUSET].mountPoint))) {
                rc = virCgroupCpuSetInherit(parent, group);
508 509
                if (rc != 0) {
                    VIR_FREE(path);
510
                    break;
511
                }
512
            }
513 514 515 516 517 518 519 520 521 522 523 524 525 526
            /*
             * Note that virCgroupSetMemoryUseHierarchy should always be
             * called prior to creating subcgroups and attaching tasks.
             */
            if (memory_hierarchy &&
                group->controllers[VIR_CGROUP_CONTROLLER_MEMORY].mountPoint != NULL &&
                (i == VIR_CGROUP_CONTROLLER_MEMORY ||
                 STREQ(group->controllers[i].mountPoint, group->controllers[VIR_CGROUP_CONTROLLER_MEMORY].mountPoint))) {
                rc = virCgroupSetMemoryUseHierarchy(group);
                if (rc != 0) {
                    VIR_FREE(path);
                    break;
                }
            }
527 528 529 530 531 532 533 534
        }

        VIR_FREE(path);
    }

    return rc;
}

535 536 537

static int virCgroupNew(const char *path,
                        virCgroupPtr *group)
538 539 540 541
{
    int rc = 0;
    char *typpath = NULL;

542 543
    VIR_DEBUG("New group %s", path);
    *group = NULL;
544

545
    if (VIR_ALLOC((*group)) != 0) {
546 547 548 549
        rc = -ENOMEM;
        goto err;
    }

550
    if (!((*group)->path = strdup(path))) {
551 552 553 554
        rc = -ENOMEM;
        goto err;
    }

555 556 557
    rc = virCgroupDetect(*group);
    if (rc < 0)
        goto err;
558 559 560

    return rc;
err:
561 562
    virCgroupFree(group);
    *group = NULL;
563 564 565 566 567 568

    VIR_FREE(typpath);

    return rc;
}

569
static int virCgroupAppRoot(int privileged,
570 571
                            virCgroupPtr *group,
                            int create)
572
{
573 574
    virCgroupPtr rootgrp = NULL;
    int rc;
575

576 577 578
    rc = virCgroupNew("/", &rootgrp);
    if (rc != 0)
        return rc;
579

580 581 582 583 584
    if (privileged) {
        rc = virCgroupNew("/libvirt", group);
    } else {
        char *rootname;
        char *username;
585
        username = virGetUserName(getuid());
586 587 588 589 590 591 592 593 594 595
        if (!username) {
            rc = -ENOMEM;
            goto cleanup;
        }
        rc = virAsprintf(&rootname, "/libvirt-%s", username);
        VIR_FREE(username);
        if (rc < 0) {
            rc = -ENOMEM;
            goto cleanup;
        }
596

597 598
        rc = virCgroupNew(rootname, group);
        VIR_FREE(rootname);
599 600
    }
    if (rc != 0)
601
        goto cleanup;
602

603
    rc = virCgroupMakeGroup(rootgrp, *group, create, false);
604

605 606
cleanup:
    virCgroupFree(&rootgrp);
607 608
    return rc;
}
D
Daniel P. Berrange 已提交
609
#endif
610

611 612 613 614 615 616 617 618
static int virCgroupRemoveRecursively(char *grppath)
{
    DIR *grpdir;
    struct dirent *ent;
    int rc = 0;

    grpdir = opendir(grppath);
    if (grpdir == NULL) {
619 620
        if (errno == ENOENT)
            return 0;
621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659
        VIR_ERROR(_("Unable to open %s (%d)"), grppath, errno);
        rc = -errno;
        return rc;
    }

    for (;;) {
        char *path;

        errno = 0;
        ent = readdir(grpdir);
        if (ent == NULL) {
            if ((rc = -errno))
                VIR_ERROR(_("Failed to readdir for %s (%d)"), grppath, errno);
            break;
        }

        if (ent->d_name[0] == '.') continue;
        if (ent->d_type != DT_DIR) continue;

        if (virAsprintf(&path, "%s/%s", grppath, ent->d_name) == -1) {
            rc = -ENOMEM;
            break;
        }
        rc = virCgroupRemoveRecursively(path);
        VIR_FREE(path);
        if (rc != 0)
            break;
    }
    closedir(grpdir);

    DEBUG("Removing cgroup %s", grppath);
    if (rmdir(grppath) != 0 && errno != ENOENT) {
        rc = -errno;
        VIR_ERROR(_("Unable to remove %s (%d)"), grppath, errno);
    }

    return rc;
}

660 661 662 663 664
/**
 * virCgroupRemove:
 *
 * @group: The group to be removed
 *
665 666 667 668 669
 * It first removes all child groups recursively
 * in depth first order and then removes @group
 * because the presence of the child groups
 * prevents removing @group.
 *
670 671 672 673 674 675 676 677
 * Returns: 0 on success
 */
int virCgroupRemove(virCgroupPtr group)
{
    int rc = 0;
    int i;
    char *grppath = NULL;

678
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
679 680
        /* Skip over controllers not mounted */
        if (!group->controllers[i].mountPoint)
681 682
            continue;

683 684 685 686 687
        if (virCgroupPathOfController(group,
                                      i,
                                      NULL,
                                      &grppath) != 0)
            continue;
688

689 690
        DEBUG("Removing cgroup %s and all child cgroups", grppath);
        rc = virCgroupRemoveRecursively(grppath);
691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709
        VIR_FREE(grppath);
    }

    return rc;
}

/**
 * virCgroupAddTask:
 *
 * @group: The cgroup to add a task to
 * @pid: The pid of the task to add
 *
 * Returns: 0 on success
 */
int virCgroupAddTask(virCgroupPtr group, pid_t pid)
{
    int rc = 0;
    int i;

710
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
711 712 713
        /* Skip over controllers not mounted */
        if (!group->controllers[i].mountPoint)
            continue;
714

715
        rc = virCgroupSetValueU64(group, i, "tasks", (unsigned long long)pid);
716 717 718 719 720 721 722
        if (rc != 0)
            break;
    }

    return rc;
}

723

724
/**
725
 * virCgroupForDriver:
726
 *
727
 * @name: name of this driver (e.g., xen, qemu, lxc)
728 729 730 731
 * @group: Pointer to returned virCgroupPtr
 *
 * Returns 0 on success
 */
732
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
733 734 735 736
int virCgroupForDriver(const char *name,
                       virCgroupPtr *group,
                       int privileged,
                       int create)
737 738
{
    int rc;
739
    char *path = NULL;
740
    virCgroupPtr rootgrp = NULL;
741

742
    rc = virCgroupAppRoot(privileged, &rootgrp, create);
743
    if (rc != 0)
744 745
        goto out;

746 747
    if (virAsprintf(&path, "%s/%s", rootgrp->path, name) < 0) {
        rc = -ENOMEM;
748
        goto out;
749
    }
750

751 752 753
    rc = virCgroupNew(path, group);
    VIR_FREE(path);

754
    if (rc == 0) {
755
        rc = virCgroupMakeGroup(rootgrp, *group, create, false);
756 757 758
        if (rc != 0)
            virCgroupFree(group);
    }
759

760
out:
761
    virCgroupFree(&rootgrp);
762 763 764

    return rc;
}
D
Daniel P. Berrange 已提交
765 766 767 768 769 770 771 772 773 774
#else
int virCgroupForDriver(const char *name ATTRIBUTE_UNUSED,
                       virCgroupPtr *group ATTRIBUTE_UNUSED,
                       int privileged ATTRIBUTE_UNUSED,
                       int create ATTRIBUTE_UNUSED)
{
    /* Claim no support */
    return -ENXIO;
}
#endif
775

776 777 778 779 780 781 782 783 784 785

/**
 * virCgroupForDomain:
 *
 * @driver: group for driver owning the domain
 * @name: name of the domain
 * @group: Pointer to returned virCgroupPtr
 *
 * Returns 0 on success
 */
786
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
787 788 789 790 791 792 793 794
int virCgroupForDomain(virCgroupPtr driver,
                       const char *name,
                       virCgroupPtr *group,
                       int create)
{
    int rc;
    char *path;

795 796 797
    if (driver == NULL)
        return -EINVAL;

798 799 800 801 802 803
    if (virAsprintf(&path, "%s/%s", driver->path, name) < 0)
        return -ENOMEM;

    rc = virCgroupNew(path, group);
    VIR_FREE(path);

804
    if (rc == 0) {
805 806 807 808 809 810 811 812 813 814 815
        /*
         * Create a cgroup with memory.use_hierarchy enabled to
         * surely account memory usage of lxc with ns subsystem
         * enabled. (To be exact, memory and ns subsystems are
         * enabled at the same time.)
         *
         * The reason why doing it here, not a upper group, say
         * a group for driver, is to avoid overhead to track
         * cumulative usage that we don't need.
         */
        rc = virCgroupMakeGroup(driver, *group, create, true);
816 817 818 819 820 821
        if (rc != 0)
            virCgroupFree(group);
    }

    return rc;
}
D
Daniel P. Berrange 已提交
822 823 824 825 826 827 828 829 830
#else
int virCgroupForDomain(virCgroupPtr driver ATTRIBUTE_UNUSED,
                       const char *name ATTRIBUTE_UNUSED,
                       virCgroupPtr *group ATTRIBUTE_UNUSED,
                       int create ATTRIBUTE_UNUSED)
{
    return -ENXIO;
}
#endif
831

832 833 834 835 836 837 838 839 840 841 842
/**
 * virCgroupSetMemory:
 *
 * @group: The cgroup to change memory for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
int virCgroupSetMemory(virCgroupPtr group, unsigned long kb)
{
    return virCgroupSetValueU64(group,
843
                                VIR_CGROUP_CONTROLLER_MEMORY,
844 845 846 847
                                "memory.limit_in_bytes",
                                kb << 10);
}

R
Ryota Ozaki 已提交
848 849 850 851 852 853 854 855 856 857
/**
 * virCgroupGetMemoryUsage:
 *
 * @group: The cgroup to change memory for
 * @kb: Pointer to returned used memory in kilobytes
 *
 * Returns: 0 on success
 */
int virCgroupGetMemoryUsage(virCgroupPtr group, unsigned long *kb)
{
C
Cole Robinson 已提交
858
    long long unsigned int usage_in_bytes;
R
Ryota Ozaki 已提交
859 860 861 862 863 864 865 866 867
    int ret;
    ret = virCgroupGetValueU64(group,
                               VIR_CGROUP_CONTROLLER_MEMORY,
                               "memory.usage_in_bytes", &usage_in_bytes);
    if (ret == 0)
        *kb = (unsigned long) usage_in_bytes >> 10;
    return ret;
}

868 869 870 871 872 873 874 875 876 877
/**
 * virCgroupDenyAllDevices:
 *
 * @group: The cgroup to deny devices for
 *
 * Returns: 0 on success
 */
int virCgroupDenyAllDevices(virCgroupPtr group)
{
    return virCgroupSetValueStr(group,
878 879 880
                                VIR_CGROUP_CONTROLLER_DEVICES,
                                "devices.deny",
                                "a");
881 882 883 884 885 886 887 888 889 890 891 892
}

/**
 * virCgroupAllowDevice:
 *
 * @group: The cgroup to allow a device for
 * @type: The device type (i.e., 'c' or 'b')
 * @major: The major number of the device
 * @minor: The minor number of the device
 *
 * Returns: 0 on success
 */
893
int virCgroupAllowDevice(virCgroupPtr group, char type, int major, int minor)
894 895 896 897
{
    int rc;
    char *devstr = NULL;

898
    if (virAsprintf(&devstr, "%c %i:%i rwm", type, major, minor) == -1) {
899 900 901 902 903
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
904
                              VIR_CGROUP_CONTROLLER_DEVICES,
905 906 907 908 909 910 911
                              "devices.allow",
                              devstr);
out:
    VIR_FREE(devstr);

    return rc;
}
912

913 914 915 916 917 918 919 920 921
/**
 * virCgroupAllowDeviceMajor:
 *
 * @group: The cgroup to allow an entire device major type for
 * @type: The device type (i.e., 'c' or 'b')
 * @major: The major number of the device type
 *
 * Returns: 0 on success
 */
922
int virCgroupAllowDeviceMajor(virCgroupPtr group, char type, int major)
923 924 925 926
{
    int rc;
    char *devstr = NULL;

927
    if (virAsprintf(&devstr, "%c %i:* rwm", type, major) == -1) {
928 929 930 931 932
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
933
                              VIR_CGROUP_CONTROLLER_DEVICES,
934 935 936 937 938 939 940 941
                              "devices.allow",
                              devstr);
 out:
    VIR_FREE(devstr);

    return rc;
}

942 943 944 945 946 947 948 949 950 951 952
/**
 * virCgroupAllowDevicePath:
 *
 * @group: The cgroup to allow the device for
 * @path: the device to allow
 *
 * Queries the type of device and its major/minor number, and
 * adds that to the cgroup ACL
 *
 * Returns: 0 on success
 */
D
Daniel P. Berrange 已提交
953
#if defined(major) && defined(minor)
954 955 956 957 958 959 960 961 962 963 964 965 966 967 968
int virCgroupAllowDevicePath(virCgroupPtr group, const char *path)
{
    struct stat sb;

    if (stat(path, &sb) < 0)
        return -errno;

    if (!S_ISCHR(sb.st_mode) && !S_ISBLK(sb.st_mode))
        return -EINVAL;

    return virCgroupAllowDevice(group,
                                S_ISCHR(sb.st_mode) ? 'c' : 'b',
                                major(sb.st_rdev),
                                minor(sb.st_rdev));
}
D
Daniel P. Berrange 已提交
969 970 971 972 973 974 975 976
#else
int virCgroupAllowDevicePath(virCgroupPtr group ATTRIBUTE_UNUSED,
                             const char *path ATTRIBUTE_UNUSED)
{
    return -ENOSYS;
}
#endif

977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036

/**
 * virCgroupDenyDevice:
 *
 * @group: The cgroup to deny a device for
 * @type: The device type (i.e., 'c' or 'b')
 * @major: The major number of the device
 * @minor: The minor number of the device
 *
 * Returns: 0 on success
 */
int virCgroupDenyDevice(virCgroupPtr group, char type, int major, int minor)
{
    int rc;
    char *devstr = NULL;

    if (virAsprintf(&devstr, "%c %i:%i rwm", type, major, minor) == -1) {
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
                              VIR_CGROUP_CONTROLLER_DEVICES,
                              "devices.deny",
                              devstr);
out:
    VIR_FREE(devstr);

    return rc;
}

/**
 * virCgroupDenyDeviceMajor:
 *
 * @group: The cgroup to deny an entire device major type for
 * @type: The device type (i.e., 'c' or 'b')
 * @major: The major number of the device type
 *
 * Returns: 0 on success
 */
int virCgroupDenyDeviceMajor(virCgroupPtr group, char type, int major)
{
    int rc;
    char *devstr = NULL;

    if (virAsprintf(&devstr, "%c %i:* rwm", type, major) == -1) {
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
                              VIR_CGROUP_CONTROLLER_DEVICES,
                              "devices.deny",
                              devstr);
 out:
    VIR_FREE(devstr);

    return rc;
}

D
Daniel P. Berrange 已提交
1037
#if defined(major) && defined(minor)
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
int virCgroupDenyDevicePath(virCgroupPtr group, const char *path)
{
    struct stat sb;

    if (stat(path, &sb) < 0)
        return -errno;

    if (!S_ISCHR(sb.st_mode) && !S_ISBLK(sb.st_mode))
        return -EINVAL;

    return virCgroupDenyDevice(group,
                               S_ISCHR(sb.st_mode) ? 'c' : 'b',
                               major(sb.st_rdev),
                               minor(sb.st_rdev));
}
D
Daniel P. Berrange 已提交
1053 1054 1055 1056 1057 1058 1059
#else
int virCgroupDenyDevicePath(virCgroupPtr group ATTRIBUTE_UNUSED,
                            const char *path ATTRIBUTE_UNUSED)
{
    return -ENOSYS;
}
#endif
1060

1061
int virCgroupSetCpuShares(virCgroupPtr group, unsigned long long shares)
1062
{
1063 1064
    return virCgroupSetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPU,
D
Daniel P. Berrange 已提交
1065
                                "cpu.shares", shares);
1066 1067
}

1068
int virCgroupGetCpuShares(virCgroupPtr group, unsigned long long *shares)
1069
{
1070 1071
    return virCgroupGetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPU,
D
Daniel P. Berrange 已提交
1072
                                "cpu.shares", shares);
1073
}
1074 1075 1076

int virCgroupGetCpuacctUsage(virCgroupPtr group, unsigned long long *usage)
{
1077 1078
    return virCgroupGetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPUACCT,
D
Daniel P. Berrange 已提交
1079
                                "cpuacct.usage", usage);
1080
}
R
Ryota Ozaki 已提交
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090

int virCgroupSetFreezerState(virCgroupPtr group, const char *state)
{
    return virCgroupSetValueStr(group,
                                VIR_CGROUP_CONTROLLER_CPU,
                                "freezer.state", state);
}

int virCgroupGetFreezerState(virCgroupPtr group, char **state)
{
1091
    return virCgroupGetValueStr(group,
R
Ryota Ozaki 已提交
1092 1093 1094
                                VIR_CGROUP_CONTROLLER_CPU,
                                "freezer.state", state);
}