cgroup.c 40.9 KB
Newer Older
1 2 3
/*
 * cgroup.c: Tools for managing cgroups
 *
4
 * Copyright (C) 2010-2011 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
#include <fcntl.h>
#include <string.h>
#include <errno.h>
#include <stdlib.h>
#include <sys/stat.h>
#include <sys/types.h>
25
#include <signal.h>
26
#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
#include "files.h"
35
#include "hash.h"
36 37 38

#define CGROUP_MAX_VAL 512

39
VIR_ENUM_IMPL(virCgroupController, VIR_CGROUP_CONTROLLER_LAST,
R
Ryota Ozaki 已提交
40
              "cpu", "cpuacct", "cpuset", "memory", "devices",
41
              "freezer", "blkio");
42 43 44 45 46 47

struct virCgroupController {
    int type;
    char *mountPoint;
    char *placement;
};
48 49 50

struct virCgroup {
    char *path;
51 52

    struct virCgroupController controllers[VIR_CGROUP_CONTROLLER_LAST];
53 54 55 56 57 58 59 60 61
};

/**
 * virCgroupFree:
 *
 * @group: The group structure to free
 */
void virCgroupFree(virCgroupPtr *group)
{
62 63 64 65 66 67 68 69
    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);
70
    }
71 72 73

    VIR_FREE((*group)->path);
    VIR_FREE(*group);
74 75
}

L
Lai Jiangshan 已提交
76 77 78 79 80 81 82 83 84 85 86 87 88
/**
 * virCgroupMounted: query whether a cgroup subsystem is mounted or not
 *
 * @cgroup: The group structure to be queried
 * @controller: cgroup subsystem id
 *
 * Returns true if a cgroup is subsystem is mounted.
 */
bool virCgroupMounted(virCgroupPtr cgroup, int controller)
{
    return cgroup->controllers[controller].mountPoint != NULL;
}

89
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
90 91 92 93 94
/*
 * Process /proc/mounts figuring out what controllers are
 * mounted and where
 */
static int virCgroupDetectMounts(virCgroupPtr group)
95
{
96
    int i;
97
    FILE *mounts = NULL;
98 99 100 101 102
    struct mntent entry;
    char buf[CGROUP_MAX_VAL];

    mounts = fopen("/proc/mounts", "r");
    if (mounts == NULL) {
103
        VIR_ERROR(_("Unable to open /proc/mounts"));
104
        return -ENOENT;
105 106 107
    }

    while (getmntent_r(mounts, &entry, buf, sizeof(buf)) != NULL) {
108 109
        if (STRNEQ(entry.mnt_type, "cgroup"))
            continue;
110

111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
        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;
            }
        }
130 131
    }

132
    VIR_FORCE_FCLOSE(mounts);
133

134
    return 0;
135

136
no_memory:
137
    VIR_FORCE_FCLOSE(mounts);
138
    return -ENOMEM;
139 140
}

141 142 143 144 145

/*
 * Process /proc/self/cgroup figuring out what cgroup
 * sub-path the current process is assigned to. ie not
 * neccessarily in the root
146
 */
147
static int virCgroupDetectPlacement(virCgroupPtr group)
148 149
{
    int i;
150 151
    FILE *mapping  = NULL;
    char line[1024];
152

153 154
    mapping = fopen("/proc/self/cgroup", "r");
    if (mapping == NULL) {
155
        VIR_ERROR(_("Unable to open /proc/self/cgroup"));
156
        return -ENOENT;
157 158
    }

159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195
    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;
            }
        }
    }

196
    VIR_FORCE_FCLOSE(mapping);
197

198
    return 0;
199 200

no_memory:
201
    VIR_FORCE_FCLOSE(mapping);
202 203
    return -ENOMEM;

204 205
}

206
static int virCgroupDetect(virCgroupPtr group)
207
{
208 209 210
    int any = 0;
    int rc;
    int i;
211

212 213
    rc = virCgroupDetectMounts(group);
    if (rc < 0) {
214
        VIR_ERROR(_("Failed to detect mounts for %s"), group->path);
215
        return rc;
216 217
    }

218 219 220 221 222 223 224
    /* 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;
225 226


227
    rc = virCgroupDetectPlacement(group);
228

229 230 231 232 233
    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;
234

235
            if (!group->controllers[i].placement) {
236
                VIR_ERROR(_("Could not find placement for controller %s at %s"),
237 238 239 240 241
                          virCgroupControllerTypeToString(i),
                          group->controllers[i].placement);
                rc = -ENOENT;
                break;
            }
242

243 244 245 246 247 248
            VIR_DEBUG("Detected mount/mapping %i:%s at %s in %s", i,
                      virCgroupControllerTypeToString(i),
                      group->controllers[i].mountPoint,
                      group->controllers[i].placement);
        }
    } else {
249
        VIR_ERROR(_("Failed to detect mapping for %s"), group->path);
250 251 252 253
    }

    return rc;
}
D
Daniel P. Berrange 已提交
254
#endif
255

256

257 258 259 260
int virCgroupPathOfController(virCgroupPtr group,
                              int controller,
                              const char *key,
                              char **path)
261
{
262 263 264 265 266 267 268 269 270 271 272 273 274
    if (controller == -1) {
        int i;
        for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
            if (group->controllers[i].mountPoint &&
                group->controllers[i].placement) {
                controller = i;
                break;
            }
        }
    }
    if (controller == -1)
        return -ENOSYS;

275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
    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;
}


292
static int virCgroupSetValueStr(virCgroupPtr group,
293
                                int controller,
294 295 296 297 298 299
                                const char *key,
                                const char *value)
{
    int rc = 0;
    char *keypath = NULL;

300
    rc = virCgroupPathOfController(group, controller, key, &keypath);
301 302 303
    if (rc != 0)
        return rc;

304
    VIR_DEBUG("Set value '%s' to '%s'", keypath, value);
305
    rc = virFileWriteStr(keypath, value, 0);
306 307
    if (rc < 0) {
        rc = -errno;
308
        VIR_DEBUG("Failed to write value '%s': %m", value);
309 310
    } else {
        rc = 0;
311 312 313 314 315 316 317 318
    }

    VIR_FREE(keypath);

    return rc;
}

static int virCgroupGetValueStr(virCgroupPtr group,
319
                                int controller,
320 321 322 323 324 325
                                const char *key,
                                char **value)
{
    int rc;
    char *keypath = NULL;

326
    *value = NULL;
327

328
    rc = virCgroupPathOfController(group, controller, key, &keypath);
329
    if (rc != 0) {
330
        VIR_DEBUG("No path of %s, %s", group->path, key);
331 332 333
        return rc;
    }

334
    VIR_DEBUG("Get value %s", keypath);
335

336
    rc = virFileReadAll(keypath, 1024, value);
337 338
    if (rc < 0) {
        rc = -errno;
339
        VIR_DEBUG("Failed to read %s: %m\n", keypath);
340
    } else {
341 342 343 344
        /* Terminated with '\n' has sometimes harmful effects to the caller */
        char *p = strchr(*value, '\n');
        if (p) *p = '\0';

345
        rc = 0;
346 347 348 349 350 351 352
    }

    VIR_FREE(keypath);

    return rc;
}

353
static int virCgroupSetValueU64(virCgroupPtr group,
354
                                int controller,
355
                                const char *key,
D
Daniel P. Berrange 已提交
356
                                unsigned long long int value)
357 358 359 360
{
    char *strval = NULL;
    int rc;

D
Daniel P. Berrange 已提交
361
    if (virAsprintf(&strval, "%llu", value) == -1)
362 363
        return -ENOMEM;

364
    rc = virCgroupSetValueStr(group, controller, key, strval);
365 366 367 368 369 370 371

    VIR_FREE(strval);

    return rc;
}


372 373

static int virCgroupSetValueI64(virCgroupPtr group,
374
                                int controller,
375
                                const char *key,
D
Daniel P. Berrange 已提交
376
                                long long int value)
377 378 379 380
{
    char *strval = NULL;
    int rc;

D
Daniel P. Berrange 已提交
381
    if (virAsprintf(&strval, "%lld", value) == -1)
382 383
        return -ENOMEM;

384
    rc = virCgroupSetValueStr(group, controller, key, strval);
385 386 387 388 389 390

    VIR_FREE(strval);

    return rc;
}

391 392
#if 0
/* This is included for completeness, but not yet used */
393
static int virCgroupGetValueI64(virCgroupPtr group,
394
                                int controller,
395
                                const char *key,
D
Daniel P. Berrange 已提交
396
                                long long int *value)
397 398 399 400
{
    char *strval = NULL;
    int rc = 0;

401
    rc = virCgroupGetValueStr(group, controller, key, &strval);
402 403 404
    if (rc != 0)
        goto out;

405
    if (virStrToLong_ll(strval, NULL, 10, value) < 0)
406 407 408 409 410 411
        rc = -EINVAL;
out:
    VIR_FREE(strval);

    return rc;
}
412
#endif
413

414
static int virCgroupGetValueU64(virCgroupPtr group,
415
                                int controller,
416
                                const char *key,
D
Daniel P. Berrange 已提交
417
                                unsigned long long int *value)
418 419 420 421
{
    char *strval = NULL;
    int rc = 0;

422
    rc = virCgroupGetValueStr(group, controller, key, &strval);
423 424 425
    if (rc != 0)
        goto out;

D
Daniel P. Berrange 已提交
426
    if (virStrToLong_ull(strval, NULL, 10, value) < 0)
427 428 429 430 431 432 433 434
        rc = -EINVAL;
out:
    VIR_FREE(strval);

    return rc;
}


435
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
436
static int virCgroupCpuSetInherit(virCgroupPtr parent, virCgroupPtr group)
437 438 439 440 441 442 443 444
{
    int i;
    int rc = 0;
    const char *inherit_values[] = {
        "cpuset.cpus",
        "cpuset.mems",
    };

445 446 447
    VIR_DEBUG("Setting up inheritance %s -> %s", parent->path, group->path);
    for (i = 0; i < ARRAY_CARDINALITY(inherit_values) ; i++) {
        char *value;
448

449 450 451 452
        rc = virCgroupGetValueStr(parent,
                                  VIR_CGROUP_CONTROLLER_CPUSET,
                                  inherit_values[i],
                                  &value);
453
        if (rc != 0) {
454
            VIR_ERROR(_("Failed to get %s %d"), inherit_values[i], rc);
455 456 457 458 459 460 461 462 463
            break;
        }

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

        rc = virCgroupSetValueStr(group,
                                  VIR_CGROUP_CONTROLLER_CPUSET,
                                  inherit_values[i],
                                  value);
464
        VIR_FREE(value);
465 466

        if (rc != 0) {
467
            VIR_ERROR(_("Failed to set %s %d"), inherit_values[i], rc);
468 469 470 471 472 473 474
            break;
        }
    }

    return rc;
}

475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
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)
507 508 509 510
{
    int i;
    int rc = 0;

511
    VIR_DEBUG("Make group %s", group->path);
512
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
513 514
        char *path = NULL;

515 516
        /* Skip over controllers that aren't mounted */
        if (!group->controllers[i].mountPoint)
517 518
            continue;

519 520 521
        rc = virCgroupPathOfController(group, i, "", &path);
        if (rc < 0)
            return rc;
522 523 524
        /* As of Feb 2011, clang can't see that the above function
         * call did not modify group. */
        sa_assert(group->controllers[i].mountPoint);
525

526
        VIR_DEBUG("Make controller %s", path);
527
        if (access(path, F_OK) != 0) {
528 529
            if (!create ||
                mkdir(path, 0755) < 0) {
530 531 532 533 534 535 536 537 538 539 540 541 542 543
                /* With a kernel that doesn't support multi-level directory
                 * for blkio controller, libvirt will fail and disable all
                 * other controllers even though they are available. So
                 * treat blkio as unmounted if mkdir fails. */
                if (i == VIR_CGROUP_CONTROLLER_BLKIO) {
                    rc = 0;
                    VIR_FREE(group->controllers[i].mountPoint);
                    VIR_FREE(path);
                    continue;
                } else {
                    rc = -errno;
                    VIR_FREE(path);
                    break;
                }
544
            }
545 546 547 548
            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);
549 550
                if (rc != 0) {
                    VIR_FREE(path);
551
                    break;
552
                }
553
            }
554 555 556 557 558 559 560 561 562 563 564 565 566 567
            /*
             * 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;
                }
            }
568 569 570 571 572 573 574 575
        }

        VIR_FREE(path);
    }

    return rc;
}

576 577 578

static int virCgroupNew(const char *path,
                        virCgroupPtr *group)
579 580 581 582
{
    int rc = 0;
    char *typpath = NULL;

583 584
    VIR_DEBUG("New group %s", path);
    *group = NULL;
585

586
    if (VIR_ALLOC((*group)) != 0) {
587 588 589 590
        rc = -ENOMEM;
        goto err;
    }

591
    if (!((*group)->path = strdup(path))) {
592 593 594 595
        rc = -ENOMEM;
        goto err;
    }

596 597 598
    rc = virCgroupDetect(*group);
    if (rc < 0)
        goto err;
599 600 601

    return rc;
err:
602 603
    virCgroupFree(group);
    *group = NULL;
604 605 606 607 608 609

    VIR_FREE(typpath);

    return rc;
}

610
static int virCgroupAppRoot(int privileged,
611 612
                            virCgroupPtr *group,
                            int create)
613
{
614 615
    virCgroupPtr rootgrp = NULL;
    int rc;
616

617 618 619
    rc = virCgroupNew("/", &rootgrp);
    if (rc != 0)
        return rc;
620

621 622 623 624 625
    if (privileged) {
        rc = virCgroupNew("/libvirt", group);
    } else {
        char *rootname;
        char *username;
626
        username = virGetUserName(getuid());
627 628 629 630 631 632 633 634 635 636
        if (!username) {
            rc = -ENOMEM;
            goto cleanup;
        }
        rc = virAsprintf(&rootname, "/libvirt-%s", username);
        VIR_FREE(username);
        if (rc < 0) {
            rc = -ENOMEM;
            goto cleanup;
        }
637

638 639
        rc = virCgroupNew(rootname, group);
        VIR_FREE(rootname);
640 641
    }
    if (rc != 0)
642
        goto cleanup;
643

644
    rc = virCgroupMakeGroup(rootgrp, *group, create, false);
645

646 647
cleanup:
    virCgroupFree(&rootgrp);
648 649
    return rc;
}
D
Daniel P. Berrange 已提交
650
#endif
651

652
#if defined _DIRENT_HAVE_D_TYPE
653 654 655 656 657 658 659 660
static int virCgroupRemoveRecursively(char *grppath)
{
    DIR *grpdir;
    struct dirent *ent;
    int rc = 0;

    grpdir = opendir(grppath);
    if (grpdir == NULL) {
661 662
        if (errno == ENOENT)
            return 0;
663
        rc = -errno;
664
        VIR_ERROR(_("Unable to open %s (%d)"), grppath, errno);
665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
        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);

693
    VIR_DEBUG("Removing cgroup %s", grppath);
694 695 696 697 698 699 700
    if (rmdir(grppath) != 0 && errno != ENOENT) {
        rc = -errno;
        VIR_ERROR(_("Unable to remove %s (%d)"), grppath, errno);
    }

    return rc;
}
701 702 703 704 705 706 707
#else
static int virCgroupRemoveRecursively(char *grppath ATTRIBUTE_UNUSED)
{
    /* Claim no support */
    return -ENXIO;
}
#endif
708

709 710 711 712 713
/**
 * virCgroupRemove:
 *
 * @group: The group to be removed
 *
714 715 716 717 718
 * 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.
 *
719 720 721 722 723 724 725 726
 * Returns: 0 on success
 */
int virCgroupRemove(virCgroupPtr group)
{
    int rc = 0;
    int i;
    char *grppath = NULL;

727
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
728 729
        /* Skip over controllers not mounted */
        if (!group->controllers[i].mountPoint)
730 731
            continue;

732 733 734 735 736
        if (virCgroupPathOfController(group,
                                      i,
                                      NULL,
                                      &grppath) != 0)
            continue;
737

738
        VIR_DEBUG("Removing cgroup %s and all child cgroups", grppath);
739
        rc = virCgroupRemoveRecursively(grppath);
740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
        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;

759
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
760 761 762
        /* Skip over controllers not mounted */
        if (!group->controllers[i].mountPoint)
            continue;
763

764
        rc = virCgroupSetValueU64(group, i, "tasks", (unsigned long long)pid);
765 766 767 768 769 770 771
        if (rc != 0)
            break;
    }

    return rc;
}

772

773
/**
774
 * virCgroupForDriver:
775
 *
776
 * @name: name of this driver (e.g., xen, qemu, lxc)
777 778 779 780
 * @group: Pointer to returned virCgroupPtr
 *
 * Returns 0 on success
 */
781
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
782 783 784 785
int virCgroupForDriver(const char *name,
                       virCgroupPtr *group,
                       int privileged,
                       int create)
786 787
{
    int rc;
788
    char *path = NULL;
789
    virCgroupPtr rootgrp = NULL;
790

791
    rc = virCgroupAppRoot(privileged, &rootgrp, create);
792
    if (rc != 0)
793 794
        goto out;

795 796
    if (virAsprintf(&path, "%s/%s", rootgrp->path, name) < 0) {
        rc = -ENOMEM;
797
        goto out;
798
    }
799

800 801 802
    rc = virCgroupNew(path, group);
    VIR_FREE(path);

803
    if (rc == 0) {
804
        rc = virCgroupMakeGroup(rootgrp, *group, create, false);
805 806 807
        if (rc != 0)
            virCgroupFree(group);
    }
808

809
out:
810
    virCgroupFree(&rootgrp);
811 812 813

    return rc;
}
D
Daniel P. Berrange 已提交
814 815 816 817 818 819 820 821 822 823
#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
824

825 826 827 828 829 830 831 832 833 834

/**
 * virCgroupForDomain:
 *
 * @driver: group for driver owning the domain
 * @name: name of the domain
 * @group: Pointer to returned virCgroupPtr
 *
 * Returns 0 on success
 */
835
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
836 837 838 839 840 841 842 843
int virCgroupForDomain(virCgroupPtr driver,
                       const char *name,
                       virCgroupPtr *group,
                       int create)
{
    int rc;
    char *path;

844 845 846
    if (driver == NULL)
        return -EINVAL;

847 848 849 850 851 852
    if (virAsprintf(&path, "%s/%s", driver->path, name) < 0)
        return -ENOMEM;

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

853
    if (rc == 0) {
854 855 856 857 858 859 860 861 862 863 864
        /*
         * 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);
865 866 867 868 869 870
        if (rc != 0)
            virCgroupFree(group);
    }

    return rc;
}
D
Daniel P. Berrange 已提交
871 872 873 874 875 876 877 878 879
#else
int virCgroupForDomain(virCgroupPtr driver ATTRIBUTE_UNUSED,
                       const char *name ATTRIBUTE_UNUSED,
                       virCgroupPtr *group ATTRIBUTE_UNUSED,
                       int create ATTRIBUTE_UNUSED)
{
    return -ENXIO;
}
#endif
880

881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
/**
 * virCgroupSetBlkioWeight:
 *
 * @group: The cgroup to change io weight for
 * @weight: The Weight for this cgroup
 *
 * Returns: 0 on success
 */
int virCgroupSetBlkioWeight(virCgroupPtr group, unsigned int weight)
{
    if (weight > 1000 || weight < 100)
        return -EINVAL;

    return virCgroupSetValueU64(group,
                                VIR_CGROUP_CONTROLLER_BLKIO,
                                "blkio.weight",
                                weight);
}

/**
 * virCgroupGetBlkioWeight:
 *
 * @group: The cgroup to get weight for
 * @Weight: Pointer to returned weight
 *
 * Returns: 0 on success
 */
int virCgroupGetBlkioWeight(virCgroupPtr group, unsigned int *weight)
{
    unsigned long long tmp;
    int ret;
    ret = virCgroupGetValueU64(group,
                               VIR_CGROUP_CONTROLLER_BLKIO,
                               "blkio.weight", &tmp);
    if (ret == 0)
        *weight = tmp;
    return ret;
}

920 921 922 923 924 925 926 927
/**
 * virCgroupSetMemory:
 *
 * @group: The cgroup to change memory for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
928
int virCgroupSetMemory(virCgroupPtr group, unsigned long long kb)
929
{
930 931 932 933 934 935 936 937 938 939 940 941 942 943
    unsigned long long maxkb = VIR_DOMAIN_MEMORY_PARAM_UNLIMITED;

    if (kb > maxkb)
        return -EINVAL;
    else if (kb == maxkb)
        return virCgroupSetValueI64(group,
                                    VIR_CGROUP_CONTROLLER_MEMORY,
                                    "memory.limit_in_bytes",
                                    -1);
    else
        return virCgroupSetValueU64(group,
                                    VIR_CGROUP_CONTROLLER_MEMORY,
                                    "memory.limit_in_bytes",
                                    kb << 10);
944 945
}

R
Ryota Ozaki 已提交
946 947 948 949 950 951 952 953 954 955
/**
 * 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
956
    long long unsigned int usage_in_bytes;
R
Ryota Ozaki 已提交
957 958 959 960 961 962 963 964 965
    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;
}

966 967 968 969 970 971 972 973
/**
 * virCgroupSetMemoryHardLimit:
 *
 * @group: The cgroup to change memory hard limit for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
974
int virCgroupSetMemoryHardLimit(virCgroupPtr group, unsigned long long kb)
975 976 977 978 979 980 981 982 983 984 985 986
{
    return virCgroupSetMemory(group, kb);
}

/**
 * virCgroupGetMemoryHardLimit:
 *
 * @group: The cgroup to get the memory hard limit for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
987
int virCgroupGetMemoryHardLimit(virCgroupPtr group, unsigned long long *kb)
988 989 990 991 992 993 994
{
    long long unsigned int limit_in_bytes;
    int ret;
    ret = virCgroupGetValueU64(group,
                               VIR_CGROUP_CONTROLLER_MEMORY,
                               "memory.limit_in_bytes", &limit_in_bytes);
    if (ret == 0)
995
        *kb = limit_in_bytes >> 10;
996 997 998 999 1000 1001 1002 1003 1004 1005 1006
    return ret;
}

/**
 * virCgroupSetMemorySoftLimit:
 *
 * @group: The cgroup to change memory soft limit for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
1007
int virCgroupSetMemorySoftLimit(virCgroupPtr group, unsigned long long kb)
1008
{
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
    unsigned long long maxkb = VIR_DOMAIN_MEMORY_PARAM_UNLIMITED;

    if (kb > maxkb)
        return -EINVAL;
    else if (kb == maxkb)
        return virCgroupSetValueI64(group,
                                    VIR_CGROUP_CONTROLLER_MEMORY,
                                    "memory.soft_limit_in_bytes",
                                    -1);
    else
        return virCgroupSetValueU64(group,
                                    VIR_CGROUP_CONTROLLER_MEMORY,
                                    "memory.soft_limit_in_bytes",
                                    kb << 10);
1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
}


/**
 * virCgroupGetMemorySoftLimit:
 *
 * @group: The cgroup to get the memory soft limit for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
1034
int virCgroupGetMemorySoftLimit(virCgroupPtr group, unsigned long long *kb)
1035 1036 1037 1038 1039 1040 1041
{
    long long unsigned int limit_in_bytes;
    int ret;
    ret = virCgroupGetValueU64(group,
                               VIR_CGROUP_CONTROLLER_MEMORY,
                               "memory.soft_limit_in_bytes", &limit_in_bytes);
    if (ret == 0)
1042
        *kb = limit_in_bytes >> 10;
1043 1044 1045 1046
    return ret;
}

/**
1047
 * virCgroupSetMemSwapHardLimit:
1048
 *
1049 1050
 * @group: The cgroup to change mem+swap hard limit for
 * @kb: The mem+swap amount in kilobytes
1051 1052 1053
 *
 * Returns: 0 on success
 */
1054
int virCgroupSetMemSwapHardLimit(virCgroupPtr group, unsigned long long kb)
1055
{
1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069
    unsigned long long maxkb = VIR_DOMAIN_MEMORY_PARAM_UNLIMITED;

    if (kb > maxkb)
        return -EINVAL;
    else if (kb == maxkb)
        return virCgroupSetValueI64(group,
                                    VIR_CGROUP_CONTROLLER_MEMORY,
                                    "memory.memsw.limit_in_bytes",
                                    -1);
    else
        return virCgroupSetValueU64(group,
                                    VIR_CGROUP_CONTROLLER_MEMORY,
                                    "memory.memsw.limit_in_bytes",
                                    kb << 10);
1070 1071 1072
}

/**
1073
 * virCgroupGetMemSwapHardLimit:
1074
 *
1075 1076
 * @group: The cgroup to get mem+swap hard limit for
 * @kb: The mem+swap amount in kilobytes
1077 1078 1079
 *
 * Returns: 0 on success
 */
1080
int virCgroupGetMemSwapHardLimit(virCgroupPtr group, unsigned long long *kb)
1081 1082 1083 1084 1085 1086 1087
{
    long long unsigned int limit_in_bytes;
    int ret;
    ret = virCgroupGetValueU64(group,
                               VIR_CGROUP_CONTROLLER_MEMORY,
                               "memory.memsw.limit_in_bytes", &limit_in_bytes);
    if (ret == 0)
1088
        *kb = limit_in_bytes >> 10;
1089 1090 1091
    return ret;
}

1092 1093 1094
/**
 * virCgroupDenyAllDevices:
 *
1095
 * @group: The cgroup to deny all permissions, for all devices
1096 1097 1098 1099 1100 1101
 *
 * Returns: 0 on success
 */
int virCgroupDenyAllDevices(virCgroupPtr group)
{
    return virCgroupSetValueStr(group,
1102 1103 1104
                                VIR_CGROUP_CONTROLLER_DEVICES,
                                "devices.deny",
                                "a");
1105 1106 1107 1108 1109 1110 1111 1112 1113
}

/**
 * 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
1114
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to allow
1115 1116 1117
 *
 * Returns: 0 on success
 */
1118 1119
int virCgroupAllowDevice(virCgroupPtr group, char type, int major, int minor,
                         int perms)
1120 1121 1122 1123
{
    int rc;
    char *devstr = NULL;

1124 1125 1126 1127
    if (virAsprintf(&devstr, "%c %i:%i %s%s%s", type, major, minor,
                    perms & VIR_CGROUP_DEVICE_READ ? "r" : "",
                    perms & VIR_CGROUP_DEVICE_WRITE ? "w" : "",
                    perms & VIR_CGROUP_DEVICE_MKNOD ? "m" : "") == -1) {
1128 1129 1130 1131 1132
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
1133
                              VIR_CGROUP_CONTROLLER_DEVICES,
1134 1135 1136 1137 1138 1139 1140
                              "devices.allow",
                              devstr);
out:
    VIR_FREE(devstr);

    return rc;
}
1141

1142 1143 1144 1145 1146 1147
/**
 * 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
1148
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to allow
1149 1150 1151
 *
 * Returns: 0 on success
 */
1152 1153
int virCgroupAllowDeviceMajor(virCgroupPtr group, char type, int major,
                              int perms)
1154 1155 1156 1157
{
    int rc;
    char *devstr = NULL;

1158 1159 1160 1161
    if (virAsprintf(&devstr, "%c %i:* %s%s%s", type, major,
                    perms & VIR_CGROUP_DEVICE_READ ? "r" : "",
                    perms & VIR_CGROUP_DEVICE_WRITE ? "w" : "",
                    perms & VIR_CGROUP_DEVICE_MKNOD ? "m" : "") == -1) {
1162 1163 1164 1165 1166
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
1167
                              VIR_CGROUP_CONTROLLER_DEVICES,
1168 1169 1170 1171 1172 1173 1174 1175
                              "devices.allow",
                              devstr);
 out:
    VIR_FREE(devstr);

    return rc;
}

1176 1177 1178 1179 1180
/**
 * virCgroupAllowDevicePath:
 *
 * @group: The cgroup to allow the device for
 * @path: the device to allow
1181
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to allow
1182 1183 1184 1185
 *
 * Queries the type of device and its major/minor number, and
 * adds that to the cgroup ACL
 *
1186 1187
 * Returns: 0 on success, 1 if path exists but is not a device, or
 * negative errno value on failure
1188
 */
D
Daniel P. Berrange 已提交
1189
#if defined(major) && defined(minor)
1190
int virCgroupAllowDevicePath(virCgroupPtr group, const char *path, int perms)
1191 1192 1193 1194 1195 1196 1197
{
    struct stat sb;

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

    if (!S_ISCHR(sb.st_mode) && !S_ISBLK(sb.st_mode))
1198
        return 1;
1199 1200 1201 1202

    return virCgroupAllowDevice(group,
                                S_ISCHR(sb.st_mode) ? 'c' : 'b',
                                major(sb.st_rdev),
1203 1204
                                minor(sb.st_rdev),
                                perms);
1205
}
D
Daniel P. Berrange 已提交
1206 1207
#else
int virCgroupAllowDevicePath(virCgroupPtr group ATTRIBUTE_UNUSED,
1208 1209
                             const char *path ATTRIBUTE_UNUSED,
                             int perms ATTRIBUTE_UNUSED)
D
Daniel P. Berrange 已提交
1210 1211 1212 1213 1214
{
    return -ENOSYS;
}
#endif

1215 1216 1217 1218 1219 1220 1221 1222

/**
 * 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
1223
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to deny
1224 1225 1226
 *
 * Returns: 0 on success
 */
1227 1228
int virCgroupDenyDevice(virCgroupPtr group, char type, int major, int minor,
                        int perms)
1229 1230 1231 1232
{
    int rc;
    char *devstr = NULL;

1233 1234 1235 1236
    if (virAsprintf(&devstr, "%c %i:%i %s%s%s", type, major, minor,
                    perms & VIR_CGROUP_DEVICE_READ ? "r" : "",
                    perms & VIR_CGROUP_DEVICE_WRITE ? "w" : "",
                    perms & VIR_CGROUP_DEVICE_MKNOD ? "m" : "") == -1) {
1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
        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
1257
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to deny
1258 1259 1260
 *
 * Returns: 0 on success
 */
1261 1262
int virCgroupDenyDeviceMajor(virCgroupPtr group, char type, int major,
                             int perms)
1263 1264 1265 1266
{
    int rc;
    char *devstr = NULL;

1267 1268 1269 1270
    if (virAsprintf(&devstr, "%c %i:* %s%s%s", type, major,
                    perms & VIR_CGROUP_DEVICE_READ ? "r" : "",
                    perms & VIR_CGROUP_DEVICE_WRITE ? "w" : "",
                    perms & VIR_CGROUP_DEVICE_MKNOD ? "m" : "") == -1) {
1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
        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
1285
#if defined(major) && defined(minor)
1286
int virCgroupDenyDevicePath(virCgroupPtr group, const char *path, int perms)
1287 1288 1289 1290 1291 1292 1293
{
    struct stat sb;

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

    if (!S_ISCHR(sb.st_mode) && !S_ISBLK(sb.st_mode))
1294
        return 1;
1295 1296 1297 1298

    return virCgroupDenyDevice(group,
                               S_ISCHR(sb.st_mode) ? 'c' : 'b',
                               major(sb.st_rdev),
1299 1300
                               minor(sb.st_rdev),
                               perms);
1301
}
D
Daniel P. Berrange 已提交
1302 1303
#else
int virCgroupDenyDevicePath(virCgroupPtr group ATTRIBUTE_UNUSED,
1304 1305
                            const char *path ATTRIBUTE_UNUSED,
                            int perms ATTRIBUTE_UNUSED)
D
Daniel P. Berrange 已提交
1306 1307 1308 1309
{
    return -ENOSYS;
}
#endif
1310

1311
int virCgroupSetCpuShares(virCgroupPtr group, unsigned long long shares)
1312
{
1313 1314
    return virCgroupSetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPU,
D
Daniel P. Berrange 已提交
1315
                                "cpu.shares", shares);
1316 1317
}

1318
int virCgroupGetCpuShares(virCgroupPtr group, unsigned long long *shares)
1319
{
1320 1321
    return virCgroupGetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPU,
D
Daniel P. Berrange 已提交
1322
                                "cpu.shares", shares);
1323
}
1324 1325 1326

int virCgroupGetCpuacctUsage(virCgroupPtr group, unsigned long long *usage)
{
1327 1328
    return virCgroupGetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPUACCT,
D
Daniel P. Berrange 已提交
1329
                                "cpuacct.usage", usage);
1330
}
R
Ryota Ozaki 已提交
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340

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

int virCgroupGetFreezerState(virCgroupPtr group, char **state)
{
1341
    return virCgroupGetValueStr(group,
R
Ryota Ozaki 已提交
1342 1343 1344
                                VIR_CGROUP_CONTROLLER_CPU,
                                "freezer.state", state);
}
1345

1346

E
Eric Blake 已提交
1347
#if defined HAVE_KILL && defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
1348 1349 1350 1351 1352 1353
static int virCgroupKillInternal(virCgroupPtr group, int signum, virHashTablePtr pids)
{
    int rc;
    int killedAny = 0;
    char *keypath = NULL;
    bool done = false;
E
Eric Blake 已提交
1354 1355 1356
    FILE *fp = NULL;
    VIR_DEBUG("group=%p path=%s signum=%d pids=%p",
              group, group->path, signum, pids);
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379

    rc = virCgroupPathOfController(group, -1, "tasks", &keypath);
    if (rc != 0) {
        VIR_DEBUG("No path of %s, tasks", group->path);
        return rc;
    }

    /* PIDs may be forking as we kill them, so loop
     * until there are no new PIDs found
     */
    while (!done) {
        done = true;
        if (!(fp = fopen(keypath, "r"))) {
            rc = -errno;
            VIR_DEBUG("Failed to read %s: %m\n", keypath);
            goto cleanup;
        } else {
            while (!feof(fp)) {
                unsigned long pid;
                if (fscanf(fp, "%lu", &pid) != 1) {
                    if (feof(fp))
                        break;
                    rc = -errno;
E
Eric Blake 已提交
1380 1381
                    VIR_DEBUG("Failed to read %s: %m\n", keypath);
                    goto cleanup;
1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
                }
                if (virHashLookup(pids, (void*)pid))
                    continue;

                VIR_DEBUG("pid=%lu", pid);
                if (kill((pid_t)pid, signum) < 0) {
                    if (errno != ESRCH) {
                        rc = -errno;
                        goto cleanup;
                    }
                    /* Leave RC == 0 since we didn't kill one */
                } else {
                    killedAny = 1;
                    done = false;
                }

E
Eric Blake 已提交
1398
                ignore_value(virHashAddEntry(pids, (void*)pid, (void*)1));
1399 1400 1401 1402 1403 1404 1405 1406 1407
            }
            VIR_FORCE_FCLOSE(fp);
        }
    }

    rc = killedAny ? 1 : 0;

cleanup:
    VIR_FREE(keypath);
E
Eric Blake 已提交
1408
    VIR_FORCE_FCLOSE(fp);
1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563

    return rc;
}


static unsigned long virCgroupPidCode(const void *name)
{
    return (unsigned long)name;
}
static bool virCgroupPidEqual(const void *namea, const void *nameb)
{
    return namea == nameb;
}
static void *virCgroupPidCopy(const void *name)
{
    return (void*)name;
}

/*
 * Returns
 *   < 0 : errno that occurred
 *     0 : no PIDs killed
 *     1 : at least one PID killed
 */
int virCgroupKill(virCgroupPtr group, int signum)
{
    VIR_DEBUG("group=%p path=%s signum=%d", group, group->path, signum);
    int rc;
    /* The 'tasks' file in cgroups can contain duplicated
     * pids, so we use a hash to track which we've already
     * killed.
     */
    virHashTablePtr pids = virHashCreateFull(100,
                                             NULL,
                                             virCgroupPidCode,
                                             virCgroupPidEqual,
                                             virCgroupPidCopy,
                                             NULL);

    rc = virCgroupKillInternal(group, signum, pids);

    virHashFree(pids);

    return rc;
}


static int virCgroupKillRecursiveInternal(virCgroupPtr group, int signum, virHashTablePtr pids, bool dormdir)
{
    int rc;
    int killedAny = 0;
    char *keypath = NULL;
    DIR *dp;
    virCgroupPtr subgroup = NULL;
    struct dirent *ent;
    VIR_DEBUG("group=%p path=%s signum=%d pids=%p", group, group->path, signum, pids);

    rc = virCgroupPathOfController(group, -1, "", &keypath);
    if (rc != 0) {
        VIR_DEBUG("No path of %s, tasks", group->path);
        return rc;
    }

    if ((rc = virCgroupKillInternal(group, signum, pids)) != 0)
        return rc;

    VIR_DEBUG("Iterate over children of %s", keypath);
    if (!(dp = opendir(keypath))) {
        rc = -errno;
        return rc;
    }

    while ((ent = readdir(dp))) {
        char *subpath;

        if (STREQ(ent->d_name, "."))
            continue;
        if (STREQ(ent->d_name, ".."))
            continue;
        if (ent->d_type != DT_DIR)
            continue;

        VIR_DEBUG("Process subdir %s", ent->d_name);
        if (virAsprintf(&subpath, "%s/%s", group->path, ent->d_name) < 0) {
            rc = -ENOMEM;
            goto cleanup;
        }

        if ((rc = virCgroupNew(subpath, &subgroup)) != 0)
            goto cleanup;

        if ((rc = virCgroupKillRecursiveInternal(subgroup, signum, pids, true)) < 0)
            goto cleanup;
        if (rc == 1)
            killedAny = 1;

        if (dormdir)
            virCgroupRemove(subgroup);

        virCgroupFree(&subgroup);
    }

    rc = killedAny;

cleanup:
    virCgroupFree(&subgroup);
    closedir(dp);

    return rc;
}

int virCgroupKillRecursive(virCgroupPtr group, int signum)
{
    int rc;
    VIR_DEBUG("group=%p path=%s signum=%d", group, group->path, signum);
    virHashTablePtr pids = virHashCreateFull(100,
                                             NULL,
                                             virCgroupPidCode,
                                             virCgroupPidEqual,
                                             virCgroupPidCopy,
                                             NULL);

    rc = virCgroupKillRecursiveInternal(group, signum, pids, false);

    virHashFree(pids);

    return rc;
}


int virCgroupKillPainfully(virCgroupPtr group)
{
    int i;
    int rc;
    VIR_DEBUG("cgroup=%p path=%s", group, group->path);
    for (i = 0 ; i < 15 ; i++) {
        int signum;
        if (i == 0)
            signum = SIGTERM;
        else if (i == 8)
            signum = SIGKILL;
        else
            signum = 0; /* Just check for existance */

        rc = virCgroupKillRecursive(group, signum);
        VIR_DEBUG("Iteration %d rc=%d", i, rc);
        /* If rc == -1 we hit error, if 0 we ran out of PIDs */
        if (rc <= 0)
            break;

        usleep(200 * 1000);
    }
    VIR_DEBUG("Complete %d", rc);
    return rc;
}
1564

E
Eric Blake 已提交
1565
#else /* !(HAVE_KILL, HAVE_MNTENT_H, HAVE_GETMNTENT_R) */
1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
int virCgroupKill(virCgroupPtr group ATTRIBUTE_UNUSED,
                  int signum ATTRIBUTE_UNUSED)
{
    return -ENOSYS;
}
int virCgroupKillRecursive(virCgroupPtr group ATTRIBUTE_UNUSED,
                           int signum ATTRIBUTE_UNUSED)
{
    return -ENOSYS;
}

int virCgroupKillPainfully(virCgroupPtr group ATTRIBUTE_UNUSED)
{
    return -ENOSYS;
}
E
Eric Blake 已提交
1581
#endif /* HAVE_KILL, HAVE_MNTENT_H, HAVE_GETMNTENT_R */