cgroup.c 42.8 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 62 63 64
typedef enum {
    VIR_CGROUP_NONE = 0, /* create subdir under each cgroup if possible. */
    VIR_CGROUP_MEM_HIERACHY = 1 << 0, /* call virCgroupSetMemoryUseHierarchy
                                       * before creating subcgroups and
                                       * attaching tasks
                                       */
    VIR_CGROUP_VCPU = 1 << 1, /* create subdir only under the cgroup cpu,
                               * cpuacct and cpuset if possible. */
} virCgroupFlags;

65 66 67 68 69 70 71
/**
 * virCgroupFree:
 *
 * @group: The group structure to free
 */
void virCgroupFree(virCgroupPtr *group)
{
72 73 74 75 76 77 78 79
    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);
80
    }
81 82 83

    VIR_FREE((*group)->path);
    VIR_FREE(*group);
84 85
}

L
Lai Jiangshan 已提交
86 87 88 89 90 91 92 93 94 95 96 97 98
/**
 * 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;
}

99
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
100 101 102 103 104
/*
 * Process /proc/mounts figuring out what controllers are
 * mounted and where
 */
static int virCgroupDetectMounts(virCgroupPtr group)
105
{
106
    int i;
107
    FILE *mounts = NULL;
108 109 110 111 112
    struct mntent entry;
    char buf[CGROUP_MAX_VAL];

    mounts = fopen("/proc/mounts", "r");
    if (mounts == NULL) {
113
        VIR_ERROR(_("Unable to open /proc/mounts"));
114
        return -ENOENT;
115 116 117
    }

    while (getmntent_r(mounts, &entry, buf, sizeof(buf)) != NULL) {
118 119
        if (STRNEQ(entry.mnt_type, "cgroup"))
            continue;
120

121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
        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;
            }
        }
140 141
    }

142
    VIR_FORCE_FCLOSE(mounts);
143

144
    return 0;
145

146
no_memory:
147
    VIR_FORCE_FCLOSE(mounts);
148
    return -ENOMEM;
149 150
}

151 152 153 154 155

/*
 * Process /proc/self/cgroup figuring out what cgroup
 * sub-path the current process is assigned to. ie not
 * neccessarily in the root
156
 */
157
static int virCgroupDetectPlacement(virCgroupPtr group)
158 159
{
    int i;
160 161
    FILE *mapping  = NULL;
    char line[1024];
162

163 164
    mapping = fopen("/proc/self/cgroup", "r");
    if (mapping == NULL) {
165
        VIR_ERROR(_("Unable to open /proc/self/cgroup"));
166
        return -ENOENT;
167 168
    }

169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
    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;
            }
        }
    }

206
    VIR_FORCE_FCLOSE(mapping);
207

208
    return 0;
209 210

no_memory:
211
    VIR_FORCE_FCLOSE(mapping);
212 213
    return -ENOMEM;

214 215
}

216
static int virCgroupDetect(virCgroupPtr group)
217
{
218 219 220
    int any = 0;
    int rc;
    int i;
221

222 223
    rc = virCgroupDetectMounts(group);
    if (rc < 0) {
224
        VIR_ERROR(_("Failed to detect mounts for %s"), group->path);
225
        return rc;
226 227
    }

228 229 230 231 232 233 234
    /* 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;
235 236


237
    rc = virCgroupDetectPlacement(group);
238

239 240 241 242 243
    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;
244

245
            if (!group->controllers[i].placement) {
246
                VIR_ERROR(_("Could not find placement for controller %s at %s"),
247 248 249 250 251
                          virCgroupControllerTypeToString(i),
                          group->controllers[i].placement);
                rc = -ENOENT;
                break;
            }
252

253 254 255 256 257 258
            VIR_DEBUG("Detected mount/mapping %i:%s at %s in %s", i,
                      virCgroupControllerTypeToString(i),
                      group->controllers[i].mountPoint,
                      group->controllers[i].placement);
        }
    } else {
259
        VIR_ERROR(_("Failed to detect mapping for %s"), group->path);
260 261 262 263
    }

    return rc;
}
D
Daniel P. Berrange 已提交
264
#endif
265

266

267 268 269 270
int virCgroupPathOfController(virCgroupPtr group,
                              int controller,
                              const char *key,
                              char **path)
271
{
272 273 274 275 276 277 278 279 280 281 282 283 284
    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;

285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301
    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;
}


302
static int virCgroupSetValueStr(virCgroupPtr group,
303
                                int controller,
304 305 306 307 308 309
                                const char *key,
                                const char *value)
{
    int rc = 0;
    char *keypath = NULL;

310
    rc = virCgroupPathOfController(group, controller, key, &keypath);
311 312 313
    if (rc != 0)
        return rc;

314
    VIR_DEBUG("Set value '%s' to '%s'", keypath, value);
315
    rc = virFileWriteStr(keypath, value, 0);
316 317
    if (rc < 0) {
        rc = -errno;
318
        VIR_DEBUG("Failed to write value '%s': %m", value);
319 320
    } else {
        rc = 0;
321 322 323 324 325 326 327 328
    }

    VIR_FREE(keypath);

    return rc;
}

static int virCgroupGetValueStr(virCgroupPtr group,
329
                                int controller,
330 331 332 333 334 335
                                const char *key,
                                char **value)
{
    int rc;
    char *keypath = NULL;

336
    *value = NULL;
337

338
    rc = virCgroupPathOfController(group, controller, key, &keypath);
339
    if (rc != 0) {
340
        VIR_DEBUG("No path of %s, %s", group->path, key);
341 342 343
        return rc;
    }

344
    VIR_DEBUG("Get value %s", keypath);
345

346
    rc = virFileReadAll(keypath, 1024, value);
347 348
    if (rc < 0) {
        rc = -errno;
349
        VIR_DEBUG("Failed to read %s: %m\n", keypath);
350
    } else {
351 352 353 354
        /* Terminated with '\n' has sometimes harmful effects to the caller */
        char *p = strchr(*value, '\n');
        if (p) *p = '\0';

355
        rc = 0;
356 357 358 359 360 361 362
    }

    VIR_FREE(keypath);

    return rc;
}

363
static int virCgroupSetValueU64(virCgroupPtr group,
364
                                int controller,
365
                                const char *key,
D
Daniel P. Berrange 已提交
366
                                unsigned long long int value)
367 368 369 370
{
    char *strval = NULL;
    int rc;

D
Daniel P. Berrange 已提交
371
    if (virAsprintf(&strval, "%llu", value) == -1)
372 373
        return -ENOMEM;

374
    rc = virCgroupSetValueStr(group, controller, key, strval);
375 376 377 378 379 380 381

    VIR_FREE(strval);

    return rc;
}


382 383

static int virCgroupSetValueI64(virCgroupPtr group,
384
                                int controller,
385
                                const char *key,
D
Daniel P. Berrange 已提交
386
                                long long int value)
387 388 389 390
{
    char *strval = NULL;
    int rc;

D
Daniel P. Berrange 已提交
391
    if (virAsprintf(&strval, "%lld", value) == -1)
392 393
        return -ENOMEM;

394
    rc = virCgroupSetValueStr(group, controller, key, strval);
395 396 397 398 399 400

    VIR_FREE(strval);

    return rc;
}

401 402
#if 0
/* This is included for completeness, but not yet used */
403
static int virCgroupGetValueI64(virCgroupPtr group,
404
                                int controller,
405
                                const char *key,
D
Daniel P. Berrange 已提交
406
                                long long int *value)
407 408 409 410
{
    char *strval = NULL;
    int rc = 0;

411
    rc = virCgroupGetValueStr(group, controller, key, &strval);
412 413 414
    if (rc != 0)
        goto out;

415
    if (virStrToLong_ll(strval, NULL, 10, value) < 0)
416 417 418 419 420 421
        rc = -EINVAL;
out:
    VIR_FREE(strval);

    return rc;
}
422
#endif
423

424
static int virCgroupGetValueU64(virCgroupPtr group,
425
                                int controller,
426
                                const char *key,
D
Daniel P. Berrange 已提交
427
                                unsigned long long int *value)
428 429 430 431
{
    char *strval = NULL;
    int rc = 0;

432
    rc = virCgroupGetValueStr(group, controller, key, &strval);
433 434 435
    if (rc != 0)
        goto out;

D
Daniel P. Berrange 已提交
436
    if (virStrToLong_ull(strval, NULL, 10, value) < 0)
437 438 439 440 441 442 443 444
        rc = -EINVAL;
out:
    VIR_FREE(strval);

    return rc;
}


445
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
446
static int virCgroupCpuSetInherit(virCgroupPtr parent, virCgroupPtr group)
447 448 449 450 451 452 453 454
{
    int i;
    int rc = 0;
    const char *inherit_values[] = {
        "cpuset.cpus",
        "cpuset.mems",
    };

455 456 457
    VIR_DEBUG("Setting up inheritance %s -> %s", parent->path, group->path);
    for (i = 0; i < ARRAY_CARDINALITY(inherit_values) ; i++) {
        char *value;
458

459 460 461 462
        rc = virCgroupGetValueStr(parent,
                                  VIR_CGROUP_CONTROLLER_CPUSET,
                                  inherit_values[i],
                                  &value);
463
        if (rc != 0) {
464
            VIR_ERROR(_("Failed to get %s %d"), inherit_values[i], rc);
465 466 467 468 469 470 471 472 473
            break;
        }

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

        rc = virCgroupSetValueStr(group,
                                  VIR_CGROUP_CONTROLLER_CPUSET,
                                  inherit_values[i],
                                  value);
474
        VIR_FREE(value);
475 476

        if (rc != 0) {
477
            VIR_ERROR(_("Failed to set %s %d"), inherit_values[i], rc);
478 479 480 481 482 483 484
            break;
        }
    }

    return rc;
}

485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
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,
516
                              int create, unsigned int flags)
517 518 519 520
{
    int i;
    int rc = 0;

521
    VIR_DEBUG("Make group %s", group->path);
522
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
523 524
        char *path = NULL;

525 526
        /* Skip over controllers that aren't mounted */
        if (!group->controllers[i].mountPoint)
527 528
            continue;

529 530 531 532 533 534 535
        /* We need to control cpu bandwidth for each vcpu now */
        if ((flags & VIR_CGROUP_VCPU) && (i != VIR_CGROUP_CONTROLLER_CPU)) {
            /* treat it as unmounted and we can use virCgroupAddTask */
            VIR_FREE(group->controllers[i].mountPoint);
            continue;
        }

536 537 538
        rc = virCgroupPathOfController(group, i, "", &path);
        if (rc < 0)
            return rc;
539 540 541
        /* As of Feb 2011, clang can't see that the above function
         * call did not modify group. */
        sa_assert(group->controllers[i].mountPoint);
542

543
        VIR_DEBUG("Make controller %s", path);
544
        if (access(path, F_OK) != 0) {
545 546
            if (!create ||
                mkdir(path, 0755) < 0) {
547 548 549 550 551 552 553 554 555 556 557 558 559 560
                /* 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;
                }
561
            }
562 563 564 565
            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);
566 567
                if (rc != 0) {
                    VIR_FREE(path);
568
                    break;
569
                }
570
            }
571 572 573 574
            /*
             * Note that virCgroupSetMemoryUseHierarchy should always be
             * called prior to creating subcgroups and attaching tasks.
             */
575 576
            if ((flags & VIR_CGROUP_MEM_HIERACHY) &&
                (group->controllers[VIR_CGROUP_CONTROLLER_MEMORY].mountPoint != NULL) &&
577 578 579 580 581 582 583 584
                (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;
                }
            }
585 586 587 588 589 590 591 592
        }

        VIR_FREE(path);
    }

    return rc;
}

593 594 595

static int virCgroupNew(const char *path,
                        virCgroupPtr *group)
596 597 598 599
{
    int rc = 0;
    char *typpath = NULL;

600 601
    VIR_DEBUG("New group %s", path);
    *group = NULL;
602

603
    if (VIR_ALLOC((*group)) != 0) {
604 605 606 607
        rc = -ENOMEM;
        goto err;
    }

608
    if (!((*group)->path = strdup(path))) {
609 610 611 612
        rc = -ENOMEM;
        goto err;
    }

613 614 615
    rc = virCgroupDetect(*group);
    if (rc < 0)
        goto err;
616 617 618

    return rc;
err:
619 620
    virCgroupFree(group);
    *group = NULL;
621 622 623 624 625 626

    VIR_FREE(typpath);

    return rc;
}

627
static int virCgroupAppRoot(int privileged,
628 629
                            virCgroupPtr *group,
                            int create)
630
{
631 632
    virCgroupPtr rootgrp = NULL;
    int rc;
633

634 635 636
    rc = virCgroupNew("/", &rootgrp);
    if (rc != 0)
        return rc;
637

638 639 640 641 642
    if (privileged) {
        rc = virCgroupNew("/libvirt", group);
    } else {
        char *rootname;
        char *username;
643
        username = virGetUserName(getuid());
644 645 646 647 648 649 650 651 652 653
        if (!username) {
            rc = -ENOMEM;
            goto cleanup;
        }
        rc = virAsprintf(&rootname, "/libvirt-%s", username);
        VIR_FREE(username);
        if (rc < 0) {
            rc = -ENOMEM;
            goto cleanup;
        }
654

655 656
        rc = virCgroupNew(rootname, group);
        VIR_FREE(rootname);
657 658
    }
    if (rc != 0)
659
        goto cleanup;
660

661
    rc = virCgroupMakeGroup(rootgrp, *group, create, VIR_CGROUP_NONE);
662

663 664
cleanup:
    virCgroupFree(&rootgrp);
665 666
    return rc;
}
D
Daniel P. Berrange 已提交
667
#endif
668

669
#if defined _DIRENT_HAVE_D_TYPE
670 671 672 673 674 675 676 677
static int virCgroupRemoveRecursively(char *grppath)
{
    DIR *grpdir;
    struct dirent *ent;
    int rc = 0;

    grpdir = opendir(grppath);
    if (grpdir == NULL) {
678 679
        if (errno == ENOENT)
            return 0;
680
        rc = -errno;
681
        VIR_ERROR(_("Unable to open %s (%d)"), grppath, errno);
682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709
        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);

710
    VIR_DEBUG("Removing cgroup %s", grppath);
711 712 713 714 715 716 717
    if (rmdir(grppath) != 0 && errno != ENOENT) {
        rc = -errno;
        VIR_ERROR(_("Unable to remove %s (%d)"), grppath, errno);
    }

    return rc;
}
718 719 720 721 722 723 724
#else
static int virCgroupRemoveRecursively(char *grppath ATTRIBUTE_UNUSED)
{
    /* Claim no support */
    return -ENXIO;
}
#endif
725

726 727 728 729 730
/**
 * virCgroupRemove:
 *
 * @group: The group to be removed
 *
731 732 733 734 735
 * 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.
 *
736 737 738 739 740 741 742 743
 * Returns: 0 on success
 */
int virCgroupRemove(virCgroupPtr group)
{
    int rc = 0;
    int i;
    char *grppath = NULL;

744
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
745 746
        /* Skip over controllers not mounted */
        if (!group->controllers[i].mountPoint)
747 748
            continue;

749 750 751 752 753
        if (virCgroupPathOfController(group,
                                      i,
                                      NULL,
                                      &grppath) != 0)
            continue;
754

755
        VIR_DEBUG("Removing cgroup %s and all child cgroups", grppath);
756
        rc = virCgroupRemoveRecursively(grppath);
757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775
        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;

776
    for (i = 0 ; i < VIR_CGROUP_CONTROLLER_LAST ; i++) {
777 778 779
        /* Skip over controllers not mounted */
        if (!group->controllers[i].mountPoint)
            continue;
780

781
        rc = virCgroupSetValueU64(group, i, "tasks", (unsigned long long)pid);
782 783 784 785 786 787 788
        if (rc != 0)
            break;
    }

    return rc;
}

789

790
/**
791
 * virCgroupForDriver:
792
 *
793
 * @name: name of this driver (e.g., xen, qemu, lxc)
794 795 796 797
 * @group: Pointer to returned virCgroupPtr
 *
 * Returns 0 on success
 */
798
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
799 800 801 802
int virCgroupForDriver(const char *name,
                       virCgroupPtr *group,
                       int privileged,
                       int create)
803 804
{
    int rc;
805
    char *path = NULL;
806
    virCgroupPtr rootgrp = NULL;
807

808
    rc = virCgroupAppRoot(privileged, &rootgrp, create);
809
    if (rc != 0)
810 811
        goto out;

812 813
    if (virAsprintf(&path, "%s/%s", rootgrp->path, name) < 0) {
        rc = -ENOMEM;
814
        goto out;
815
    }
816

817 818 819
    rc = virCgroupNew(path, group);
    VIR_FREE(path);

820
    if (rc == 0) {
821
        rc = virCgroupMakeGroup(rootgrp, *group, create, VIR_CGROUP_NONE);
822 823 824
        if (rc != 0)
            virCgroupFree(group);
    }
825

826
out:
827
    virCgroupFree(&rootgrp);
828 829 830

    return rc;
}
D
Daniel P. Berrange 已提交
831 832 833 834 835 836 837 838 839 840
#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
841

842 843 844 845 846 847 848 849 850 851

/**
 * virCgroupForDomain:
 *
 * @driver: group for driver owning the domain
 * @name: name of the domain
 * @group: Pointer to returned virCgroupPtr
 *
 * Returns 0 on success
 */
852
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
853 854 855 856 857 858 859 860
int virCgroupForDomain(virCgroupPtr driver,
                       const char *name,
                       virCgroupPtr *group,
                       int create)
{
    int rc;
    char *path;

861 862 863
    if (driver == NULL)
        return -EINVAL;

864 865 866 867 868 869
    if (virAsprintf(&path, "%s/%s", driver->path, name) < 0)
        return -ENOMEM;

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

870
    if (rc == 0) {
871 872 873 874 875 876 877 878 879 880
        /*
         * 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.
         */
881
        rc = virCgroupMakeGroup(driver, *group, create, VIR_CGROUP_MEM_HIERACHY);
882 883 884 885 886 887
        if (rc != 0)
            virCgroupFree(group);
    }

    return rc;
}
D
Daniel P. Berrange 已提交
888 889 890 891 892 893 894 895 896
#else
int virCgroupForDomain(virCgroupPtr driver ATTRIBUTE_UNUSED,
                       const char *name ATTRIBUTE_UNUSED,
                       virCgroupPtr *group ATTRIBUTE_UNUSED,
                       int create ATTRIBUTE_UNUSED)
{
    return -ENXIO;
}
#endif
897

898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942
/**
 * virCgroupForVcpu:
 *
 * @driver: group for the domain
 * @vcpuid: id of the vcpu
 * @group: Pointer to returned virCgroupPtr
 *
 * Returns 0 on success
 */
#if defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
int virCgroupForVcpu(virCgroupPtr driver,
                     int vcpuid,
                     virCgroupPtr *group,
                     int create)
{
    int rc;
    char *path;

    if (driver == NULL)
        return -EINVAL;

    if (virAsprintf(&path, "%s/vcpu%d", driver->path, vcpuid) < 0)
        return -ENOMEM;

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

    if (rc == 0) {
        rc = virCgroupMakeGroup(driver, *group, create, VIR_CGROUP_VCPU);
        if (rc != 0)
            virCgroupFree(group);
    }

    return rc;
}
#else
int virCgroupForVcpu(virCgroupPtr driver ATTRIBUTE_UNUSED,
                     int vcpuid ATTRIBUTE_UNUSED,
                     virCgroupPtr *group ATTRIBUTE_UNUSED,
                     int create ATTRIBUTE_UNUSED)
{
    return -ENXIO;
}
#endif

943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981
/**
 * 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;
}

982 983 984 985 986 987 988 989
/**
 * virCgroupSetMemory:
 *
 * @group: The cgroup to change memory for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
990
int virCgroupSetMemory(virCgroupPtr group, unsigned long long kb)
991
{
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005
    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);
1006 1007
}

R
Ryota Ozaki 已提交
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
/**
 * 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
1018
    long long unsigned int usage_in_bytes;
R
Ryota Ozaki 已提交
1019 1020 1021 1022 1023 1024 1025 1026 1027
    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;
}

1028 1029 1030 1031 1032 1033 1034 1035
/**
 * virCgroupSetMemoryHardLimit:
 *
 * @group: The cgroup to change memory hard limit for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
1036
int virCgroupSetMemoryHardLimit(virCgroupPtr group, unsigned long long kb)
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
{
    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
 */
1049
int virCgroupGetMemoryHardLimit(virCgroupPtr group, unsigned long long *kb)
1050 1051 1052 1053 1054 1055 1056
{
    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)
1057
        *kb = limit_in_bytes >> 10;
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
    return ret;
}

/**
 * virCgroupSetMemorySoftLimit:
 *
 * @group: The cgroup to change memory soft limit for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
1069
int virCgroupSetMemorySoftLimit(virCgroupPtr group, unsigned long long kb)
1070
{
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
    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);
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095
}


/**
 * virCgroupGetMemorySoftLimit:
 *
 * @group: The cgroup to get the memory soft limit for
 * @kb: The memory amount in kilobytes
 *
 * Returns: 0 on success
 */
1096
int virCgroupGetMemorySoftLimit(virCgroupPtr group, unsigned long long *kb)
1097 1098 1099 1100 1101 1102 1103
{
    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)
1104
        *kb = limit_in_bytes >> 10;
1105 1106 1107 1108
    return ret;
}

/**
1109
 * virCgroupSetMemSwapHardLimit:
1110
 *
1111 1112
 * @group: The cgroup to change mem+swap hard limit for
 * @kb: The mem+swap amount in kilobytes
1113 1114 1115
 *
 * Returns: 0 on success
 */
1116
int virCgroupSetMemSwapHardLimit(virCgroupPtr group, unsigned long long kb)
1117
{
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131
    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);
1132 1133 1134
}

/**
1135
 * virCgroupGetMemSwapHardLimit:
1136
 *
1137 1138
 * @group: The cgroup to get mem+swap hard limit for
 * @kb: The mem+swap amount in kilobytes
1139 1140 1141
 *
 * Returns: 0 on success
 */
1142
int virCgroupGetMemSwapHardLimit(virCgroupPtr group, unsigned long long *kb)
1143 1144 1145 1146 1147 1148 1149
{
    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)
1150
        *kb = limit_in_bytes >> 10;
1151 1152 1153
    return ret;
}

1154 1155 1156
/**
 * virCgroupDenyAllDevices:
 *
1157
 * @group: The cgroup to deny all permissions, for all devices
1158 1159 1160 1161 1162 1163
 *
 * Returns: 0 on success
 */
int virCgroupDenyAllDevices(virCgroupPtr group)
{
    return virCgroupSetValueStr(group,
1164 1165 1166
                                VIR_CGROUP_CONTROLLER_DEVICES,
                                "devices.deny",
                                "a");
1167 1168 1169 1170 1171 1172 1173 1174 1175
}

/**
 * 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
1176
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to allow
1177 1178 1179
 *
 * Returns: 0 on success
 */
1180 1181
int virCgroupAllowDevice(virCgroupPtr group, char type, int major, int minor,
                         int perms)
1182 1183 1184 1185
{
    int rc;
    char *devstr = NULL;

1186 1187 1188 1189
    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) {
1190 1191 1192 1193 1194
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
1195
                              VIR_CGROUP_CONTROLLER_DEVICES,
1196 1197 1198 1199 1200 1201 1202
                              "devices.allow",
                              devstr);
out:
    VIR_FREE(devstr);

    return rc;
}
1203

1204 1205 1206 1207 1208 1209
/**
 * 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
1210
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to allow
1211 1212 1213
 *
 * Returns: 0 on success
 */
1214 1215
int virCgroupAllowDeviceMajor(virCgroupPtr group, char type, int major,
                              int perms)
1216 1217 1218 1219
{
    int rc;
    char *devstr = NULL;

1220 1221 1222 1223
    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) {
1224 1225 1226 1227 1228
        rc = -ENOMEM;
        goto out;
    }

    rc = virCgroupSetValueStr(group,
1229
                              VIR_CGROUP_CONTROLLER_DEVICES,
1230 1231 1232 1233 1234 1235 1236 1237
                              "devices.allow",
                              devstr);
 out:
    VIR_FREE(devstr);

    return rc;
}

1238 1239 1240 1241 1242
/**
 * virCgroupAllowDevicePath:
 *
 * @group: The cgroup to allow the device for
 * @path: the device to allow
1243
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to allow
1244 1245 1246 1247
 *
 * Queries the type of device and its major/minor number, and
 * adds that to the cgroup ACL
 *
1248 1249
 * Returns: 0 on success, 1 if path exists but is not a device, or
 * negative errno value on failure
1250
 */
D
Daniel P. Berrange 已提交
1251
#if defined(major) && defined(minor)
1252
int virCgroupAllowDevicePath(virCgroupPtr group, const char *path, int perms)
1253 1254 1255 1256 1257 1258 1259
{
    struct stat sb;

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

    if (!S_ISCHR(sb.st_mode) && !S_ISBLK(sb.st_mode))
1260
        return 1;
1261 1262 1263 1264

    return virCgroupAllowDevice(group,
                                S_ISCHR(sb.st_mode) ? 'c' : 'b',
                                major(sb.st_rdev),
1265 1266
                                minor(sb.st_rdev),
                                perms);
1267
}
D
Daniel P. Berrange 已提交
1268 1269
#else
int virCgroupAllowDevicePath(virCgroupPtr group ATTRIBUTE_UNUSED,
1270 1271
                             const char *path ATTRIBUTE_UNUSED,
                             int perms ATTRIBUTE_UNUSED)
D
Daniel P. Berrange 已提交
1272 1273 1274 1275 1276
{
    return -ENOSYS;
}
#endif

1277 1278 1279 1280 1281 1282 1283 1284

/**
 * 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
1285
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to deny
1286 1287 1288
 *
 * Returns: 0 on success
 */
1289 1290
int virCgroupDenyDevice(virCgroupPtr group, char type, int major, int minor,
                        int perms)
1291 1292 1293 1294
{
    int rc;
    char *devstr = NULL;

1295 1296 1297 1298
    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) {
1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
        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
1319
 * @perms: Bitwise or of VIR_CGROUP_DEVICE permission bits to deny
1320 1321 1322
 *
 * Returns: 0 on success
 */
1323 1324
int virCgroupDenyDeviceMajor(virCgroupPtr group, char type, int major,
                             int perms)
1325 1326 1327 1328
{
    int rc;
    char *devstr = NULL;

1329 1330 1331 1332
    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) {
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346
        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
1347
#if defined(major) && defined(minor)
1348
int virCgroupDenyDevicePath(virCgroupPtr group, const char *path, int perms)
1349 1350 1351 1352 1353 1354 1355
{
    struct stat sb;

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

    if (!S_ISCHR(sb.st_mode) && !S_ISBLK(sb.st_mode))
1356
        return 1;
1357 1358 1359 1360

    return virCgroupDenyDevice(group,
                               S_ISCHR(sb.st_mode) ? 'c' : 'b',
                               major(sb.st_rdev),
1361 1362
                               minor(sb.st_rdev),
                               perms);
1363
}
D
Daniel P. Berrange 已提交
1364 1365
#else
int virCgroupDenyDevicePath(virCgroupPtr group ATTRIBUTE_UNUSED,
1366 1367
                            const char *path ATTRIBUTE_UNUSED,
                            int perms ATTRIBUTE_UNUSED)
D
Daniel P. Berrange 已提交
1368 1369 1370 1371
{
    return -ENOSYS;
}
#endif
1372

1373
int virCgroupSetCpuShares(virCgroupPtr group, unsigned long long shares)
1374
{
1375 1376
    return virCgroupSetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPU,
D
Daniel P. Berrange 已提交
1377
                                "cpu.shares", shares);
1378 1379
}

1380
int virCgroupGetCpuShares(virCgroupPtr group, unsigned long long *shares)
1381
{
1382 1383
    return virCgroupGetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPU,
D
Daniel P. Berrange 已提交
1384
                                "cpu.shares", shares);
1385
}
1386 1387 1388

int virCgroupGetCpuacctUsage(virCgroupPtr group, unsigned long long *usage)
{
1389 1390
    return virCgroupGetValueU64(group,
                                VIR_CGROUP_CONTROLLER_CPUACCT,
D
Daniel P. Berrange 已提交
1391
                                "cpuacct.usage", usage);
1392
}
R
Ryota Ozaki 已提交
1393 1394 1395 1396 1397 1398 1399 1400 1401 1402

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

int virCgroupGetFreezerState(virCgroupPtr group, char **state)
{
1403
    return virCgroupGetValueStr(group,
R
Ryota Ozaki 已提交
1404 1405 1406
                                VIR_CGROUP_CONTROLLER_CPU,
                                "freezer.state", state);
}
1407

1408

E
Eric Blake 已提交
1409
#if defined HAVE_KILL && defined HAVE_MNTENT_H && defined HAVE_GETMNTENT_R
1410 1411 1412 1413 1414 1415
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
1416 1417 1418
    FILE *fp = NULL;
    VIR_DEBUG("group=%p path=%s signum=%d pids=%p",
              group, group->path, signum, pids);
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441

    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
1442 1443
                    VIR_DEBUG("Failed to read %s: %m\n", keypath);
                    goto cleanup;
1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
                }
                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
1460
                ignore_value(virHashAddEntry(pids, (void*)pid, (void*)1));
1461 1462 1463 1464 1465 1466 1467 1468 1469
            }
            VIR_FORCE_FCLOSE(fp);
        }
    }

    rc = killedAny ? 1 : 0;

cleanup:
    VIR_FREE(keypath);
E
Eric Blake 已提交
1470
    VIR_FORCE_FCLOSE(fp);
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 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625

    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;
}
1626

E
Eric Blake 已提交
1627
#else /* !(HAVE_KILL, HAVE_MNTENT_H, HAVE_GETMNTENT_R) */
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642
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
1643
#endif /* HAVE_KILL, HAVE_MNTENT_H, HAVE_GETMNTENT_R */