qemu_cgroup.c 20.9 KB
Newer Older
1 2 3
/*
 * qemu_cgroup.c: QEMU cgroup management
 *
4
 * Copyright (C) 2006-2011 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307  USA
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include "qemu_cgroup.h"
27
#include "qemu_domain.h"
28 29 30 31 32
#include "cgroup.h"
#include "logging.h"
#include "memory.h"
#include "virterror_internal.h"
#include "util.h"
33
#include "domain_audit.h"
34 35 36 37 38 39 40

#define VIR_FROM_THIS VIR_FROM_QEMU

static const char *const defaultDeviceACL[] = {
    "/dev/null", "/dev/full", "/dev/zero",
    "/dev/random", "/dev/urandom",
    "/dev/ptmx", "/dev/kvm", "/dev/kqemu",
41
    "/dev/rtc", "/dev/hpet",
42 43 44 45 46
    NULL,
};
#define DEVICE_PTY_MAJOR 136
#define DEVICE_SND_MAJOR 116

47 48
bool qemuCgroupControllerActive(struct qemud_driver *driver,
                                int controller)
49 50
{
    if (driver->cgroup == NULL)
51 52 53
        return false;
    if (controller < 0 || controller >= VIR_CGROUP_CONTROLLER_LAST)
        return false;
E
Eric Blake 已提交
54 55
    if (!virCgroupMounted(driver->cgroup, controller))
        return false;
56
    if (driver->cgroupControllers & (1 << controller))
57 58
        return true;
    return false;
59 60
}

61
static int
62
qemuSetupDiskPathAllow(virDomainDiskDefPtr disk,
63 64 65
                       const char *path,
                       size_t depth ATTRIBUTE_UNUSED,
                       void *opaque)
66
{
67
    qemuCgroupData *data = opaque;
68 69 70
    int rc;

    VIR_DEBUG("Process path %s for disk", path);
71 72 73
    rc = virCgroupAllowDevicePath(data->cgroup, path,
                                  (disk->readonly ? VIR_CGROUP_DEVICE_READ
                                   : VIR_CGROUP_DEVICE_RW));
74 75
    virDomainAuditCgroupPath(data->vm, data->cgroup, "allow", path,
                             disk->readonly ? "r" : "rw", rc);
76 77
    if (rc < 0) {
        if (rc == -EACCES) { /* Get this for root squash NFS */
78 79 80 81 82 83 84 85 86 87 88 89 90
            VIR_DEBUG("Ignoring EACCES for %s", path);
        } else {
            virReportSystemError(-rc,
                                 _("Unable to allow access for disk path %s"),
                                 path);
            return -1;
        }
    }
    return 0;
}


int qemuSetupDiskCgroup(struct qemud_driver *driver,
91
                        virDomainObjPtr vm,
92 93 94
                        virCgroupPtr cgroup,
                        virDomainDiskDefPtr disk)
{
95
    qemuCgroupData data = { vm, cgroup };
96 97 98
    return virDomainDiskDefForeachPath(disk,
                                       driver->allowDiskFormatProbing,
                                       true,
99
                                       driver->user, driver->group,
100
                                       qemuSetupDiskPathAllow,
101
                                       &data);
102 103 104
}


105 106 107 108 109
static int
qemuTeardownDiskPathDeny(virDomainDiskDefPtr disk ATTRIBUTE_UNUSED,
                         const char *path,
                         size_t depth ATTRIBUTE_UNUSED,
                         void *opaque)
110
{
111
    qemuCgroupData *data = opaque;
112 113 114
    int rc;

    VIR_DEBUG("Process path %s for disk", path);
115 116
    rc = virCgroupDenyDevicePath(data->cgroup, path,
                                 VIR_CGROUP_DEVICE_RWM);
117
    virDomainAuditCgroupPath(data->vm, data->cgroup, "deny", path, "rwm", rc);
118 119
    if (rc < 0) {
        if (rc == -EACCES) { /* Get this for root squash NFS */
120 121 122 123 124 125 126 127 128 129 130 131 132
            VIR_DEBUG("Ignoring EACCES for %s", path);
        } else {
            virReportSystemError(-rc,
                                 _("Unable to deny access for disk path %s"),
                                 path);
            return -1;
        }
    }
    return 0;
}


int qemuTeardownDiskCgroup(struct qemud_driver *driver,
133
                           virDomainObjPtr vm,
134 135 136
                           virCgroupPtr cgroup,
                           virDomainDiskDefPtr disk)
{
137
    qemuCgroupData data = { vm, cgroup };
138 139 140
    return virDomainDiskDefForeachPath(disk,
                                       driver->allowDiskFormatProbing,
                                       true,
141
                                       driver->user, driver->group,
142
                                       qemuTeardownDiskPathDeny,
143
                                       &data);
144 145 146
}


147 148 149 150
static int
qemuSetupChardevCgroup(virDomainDefPtr def,
                       virDomainChrDefPtr dev,
                       void *opaque)
151
{
152
    qemuCgroupData *data = opaque;
153 154
    int rc;

155
    if (dev->source.type != VIR_DOMAIN_CHR_TYPE_DEV)
156 157 158
        return 0;


159
    VIR_DEBUG("Process path '%s' for disk", dev->source.data.file.path);
160 161
    rc = virCgroupAllowDevicePath(data->cgroup, dev->source.data.file.path,
                                  VIR_CGROUP_DEVICE_RW);
162 163
    virDomainAuditCgroupPath(data->vm, data->cgroup, "allow",
                             dev->source.data.file.path, "rw", rc);
164
    if (rc < 0) {
165 166
        virReportSystemError(-rc,
                             _("Unable to allow device %s for %s"),
167
                             dev->source.data.file.path, def->name);
168 169 170 171 172 173 174 175 176 177 178
        return -1;
    }

    return 0;
}


int qemuSetupHostUsbDeviceCgroup(usbDevice *dev ATTRIBUTE_UNUSED,
                                 const char *path,
                                 void *opaque)
{
179
    qemuCgroupData *data = opaque;
180 181 182
    int rc;

    VIR_DEBUG("Process path '%s' for USB device", path);
183 184
    rc = virCgroupAllowDevicePath(data->cgroup, path,
                                  VIR_CGROUP_DEVICE_RW);
185
    virDomainAuditCgroupPath(data->vm, data->cgroup, "allow", path, "rw", rc);
186
    if (rc < 0) {
187 188 189 190 191 192 193 194 195 196
        virReportSystemError(-rc,
                             _("Unable to allow device %s"),
                             path);
        return -1;
    }

    return 0;
}

int qemuSetupCgroup(struct qemud_driver *driver,
197 198
                    virDomainObjPtr vm,
                    char *nodemask)
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
{
    virCgroupPtr cgroup = NULL;
    int rc;
    unsigned int i;
    const char *const *deviceACL =
        driver->cgroupDeviceACL ?
        (const char *const *)driver->cgroupDeviceACL :
        defaultDeviceACL;

    if (driver->cgroup == NULL)
        return 0; /* Not supported, so claim success */

    rc = virCgroupForDomain(driver->cgroup, vm->def->name, &cgroup, 1);
    if (rc != 0) {
        virReportSystemError(-rc,
                             _("Unable to create cgroup for %s"),
                             vm->def->name);
        goto cleanup;
    }

    if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_DEVICES)) {
220
        qemuCgroupData data = { vm, cgroup };
221
        rc = virCgroupDenyAllDevices(cgroup);
222
        virDomainAuditCgroup(vm, cgroup, "deny", "all", rc == 0);
223 224
        if (rc != 0) {
            if (rc == -EPERM) {
225
                VIR_WARN("Group devices ACL is not accessible, disabling whitelisting");
226 227 228 229 230 231 232 233 234
                goto done;
            }

            virReportSystemError(-rc,
                                 _("Unable to deny all devices for %s"), vm->def->name);
            goto cleanup;
        }

        for (i = 0; i < vm->def->ndisks ; i++) {
235
            if (qemuSetupDiskCgroup(driver, vm, cgroup, vm->def->disks[i]) < 0)
236 237 238
                goto cleanup;
        }

239 240
        rc = virCgroupAllowDeviceMajor(cgroup, 'c', DEVICE_PTY_MAJOR,
                                       VIR_CGROUP_DEVICE_RW);
241 242
        virDomainAuditCgroupMajor(vm, cgroup, "allow", DEVICE_PTY_MAJOR,
                                  "pty", "rw", rc == 0);
243 244 245 246 247 248
        if (rc != 0) {
            virReportSystemError(-rc, "%s",
                                 _("unable to allow /dev/pts/ devices"));
            goto cleanup;
        }

249 250 251 252 253
        if (vm->def->nsounds &&
            (!vm->def->ngraphics ||
             ((vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
               driver->vncAllowHostAudio) ||
              (vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SDL)))) {
254 255
            rc = virCgroupAllowDeviceMajor(cgroup, 'c', DEVICE_SND_MAJOR,
                                           VIR_CGROUP_DEVICE_RW);
256 257
            virDomainAuditCgroupMajor(vm, cgroup, "allow", DEVICE_SND_MAJOR,
                                      "sound", "rw", rc == 0);
258 259 260 261 262 263 264 265
            if (rc != 0) {
                virReportSystemError(-rc, "%s",
                                     _("unable to allow /dev/snd/ devices"));
                goto cleanup;
            }
        }

        for (i = 0; deviceACL[i] != NULL ; i++) {
266 267
            rc = virCgroupAllowDevicePath(cgroup, deviceACL[i],
                                          VIR_CGROUP_DEVICE_RW);
268
            virDomainAuditCgroupPath(vm, cgroup, "allow", deviceACL[i], "rw", rc);
269 270 271 272 273 274 275 276 277 278 279 280
            if (rc < 0 &&
                rc != -ENOENT) {
                virReportSystemError(-rc,
                                     _("unable to allow device %s"),
                                     deviceACL[i]);
                goto cleanup;
            }
        }

        if (virDomainChrDefForeach(vm->def,
                                   true,
                                   qemuSetupChardevCgroup,
281
                                   &data) < 0)
282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
            goto cleanup;

        for (i = 0; i < vm->def->nhostdevs; i++) {
            virDomainHostdevDefPtr hostdev = vm->def->hostdevs[i];
            usbDevice *usb;

            if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
                continue;
            if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB)
                continue;

            if ((usb = usbGetDevice(hostdev->source.subsys.u.usb.bus,
                                    hostdev->source.subsys.u.usb.device)) == NULL)
                goto cleanup;

297 298
            if (usbDeviceFileIterate(usb, qemuSetupHostUsbDeviceCgroup,
                                     &data) < 0)
299 300 301 302
                goto cleanup;
        }
    }

303 304
    if (vm->def->blkio.weight != 0) {
        if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_BLKIO)) {
305 306 307 308 309 310 311
            rc = virCgroupSetBlkioWeight(cgroup, vm->def->blkio.weight);
            if(rc != 0) {
                virReportSystemError(-rc,
                                     _("Unable to set io weight for domain %s"),
                                     vm->def->name);
                goto cleanup;
            }
312 313 314
        } else {
            qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                            _("Block I/O tuning is not available on this host"));
315 316 317 318 319 320 321 322
            goto cleanup;
        }
    }

    if (vm->def->blkio.ndevices) {
        if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_BLKIO)) {
            for (i = 0; i < vm->def->blkio.ndevices; i++) {
                virBlkioDeviceWeightPtr dw = &vm->def->blkio.devices[i];
323 324
                if (!dw->weight)
                    continue;
325 326 327 328 329 330 331 332 333 334 335 336 337 338
                rc = virCgroupSetBlkioDeviceWeight(cgroup, dw->path,
                                                   dw->weight);
                if (rc != 0) {
                    virReportSystemError(-rc,
                                         _("Unable to set io device weight "
                                           "for domain %s"),
                                         vm->def->name);
                    goto cleanup;
                }
            }
        } else {
            qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                            _("Block I/O tuning is not available on this host"));
            goto cleanup;
339 340 341
        }
    }

342 343 344
    if (vm->def->mem.hard_limit != 0 ||
        vm->def->mem.soft_limit != 0 ||
        vm->def->mem.swap_hard_limit != 0) {
345
        if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_MEMORY)) {
346 347 348 349 350 351 352 353
            if (vm->def->mem.hard_limit != 0) {
                rc = virCgroupSetMemoryHardLimit(cgroup, vm->def->mem.hard_limit);
                if (rc != 0) {
                    virReportSystemError(-rc,
                                         _("Unable to set memory hard limit for domain %s"),
                                         vm->def->name);
                    goto cleanup;
                }
354
            }
355 356 357 358 359 360 361 362
            if (vm->def->mem.soft_limit != 0) {
                rc = virCgroupSetMemorySoftLimit(cgroup, vm->def->mem.soft_limit);
                if (rc != 0) {
                    virReportSystemError(-rc,
                                         _("Unable to set memory soft limit for domain %s"),
                                         vm->def->name);
                    goto cleanup;
                }
363 364
            }

365 366 367 368 369 370 371 372
            if (vm->def->mem.swap_hard_limit != 0) {
                rc = virCgroupSetMemSwapHardLimit(cgroup, vm->def->mem.swap_hard_limit);
                if (rc != 0) {
                    virReportSystemError(-rc,
                                         _("Unable to set swap hard limit for domain %s"),
                                         vm->def->name);
                    goto cleanup;
                }
373
            }
374 375 376
        } else {
            qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                            _("Memory cgroup is not available on this host"));
377 378 379
        }
    }

380 381
    if (vm->def->cputune.shares != 0) {
        if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPU)) {
382 383 384 385 386 387 388
            rc = virCgroupSetCpuShares(cgroup, vm->def->cputune.shares);
            if(rc != 0) {
                virReportSystemError(-rc,
                                     _("Unable to set io cpu shares for domain %s"),
                                     vm->def->name);
                goto cleanup;
            }
389 390 391
        } else {
            qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                            _("CPU tuning is not available on this host"));
392 393 394
        }
    }

395 396 397
    if ((vm->def->numatune.memory.nodemask ||
         (vm->def->numatune.memory.placement_mode ==
          VIR_DOMAIN_NUMATUNE_MEM_PLACEMENT_MODE_AUTO)) &&
H
Hu Tao 已提交
398 399
        vm->def->numatune.memory.mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
        qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPUSET)) {
400 401 402 403 404 405 406
        char *mask = NULL;
        if (vm->def->numatune.memory.placement_mode ==
            VIR_DOMAIN_NUMATUNE_MEM_PLACEMENT_MODE_AUTO)
            mask = virDomainCpuSetFormat(nodemask, VIR_DOMAIN_CPUMASK_LEN);
        else
            mask = virDomainCpuSetFormat(vm->def->numatune.memory.nodemask,
                                         VIR_DOMAIN_CPUMASK_LEN);
H
Hu Tao 已提交
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421
        if (!mask) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
                            _("failed to convert memory nodemask"));
            goto cleanup;
        }

        rc = virCgroupSetCpusetMems(cgroup, mask);
        VIR_FREE(mask);
        if (rc != 0) {
            virReportSystemError(-rc,
                                 _("Unable to set cpuset.mems for domain %s"),
                                 vm->def->name);
            goto cleanup;
        }
    }
422 423 424 425 426 427 428 429 430 431 432 433
done:
    virCgroupFree(&cgroup);
    return 0;

cleanup:
    if (cgroup) {
        virCgroupRemove(cgroup);
        virCgroupFree(&cgroup);
    }
    return -1;
}

434 435 436 437 438 439 440 441 442 443 444 445 446 447
int qemuSetupCgroupVcpuBW(virCgroupPtr cgroup, unsigned long long period,
                          long long quota)
{
    int rc;
    unsigned long long old_period;

    if (period == 0 && quota == 0)
        return 0;

    if (period) {
        /* get old period, and we can rollback if set quota failed */
        rc = virCgroupGetCpuCfsPeriod(cgroup, &old_period);
        if (rc < 0) {
            virReportSystemError(-rc,
W
Wen Congyang 已提交
448
                                 "%s", _("Unable to get cpu bandwidth period"));
449 450 451 452 453 454
            return -1;
        }

        rc = virCgroupSetCpuCfsPeriod(cgroup, period);
        if (rc < 0) {
            virReportSystemError(-rc,
W
Wen Congyang 已提交
455
                                 "%s", _("Unable to set cpu bandwidth period"));
456 457 458 459 460 461 462 463
            return -1;
        }
    }

    if (quota) {
        rc = virCgroupSetCpuCfsQuota(cgroup, quota);
        if (rc < 0) {
            virReportSystemError(-rc,
W
Wen Congyang 已提交
464
                                 "%s", _("Unable to set cpu bandwidth quota"));
465 466 467 468 469 470 471 472 473 474 475
            goto cleanup;
        }
    }

    return 0;

cleanup:
    if (period) {
        rc = virCgroupSetCpuCfsPeriod(cgroup, old_period);
        if (rc < 0)
            virReportSystemError(-rc,
476 477
                                 "%s",
                                 _("Unable to rollback cpu bandwidth period"));
478 479 480 481 482 483 484 485 486 487 488 489 490 491
    }

    return -1;
}

int qemuSetupCgroupForVcpu(struct qemud_driver *driver, virDomainObjPtr vm)
{
    virCgroupPtr cgroup = NULL;
    virCgroupPtr cgroup_vcpu = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
    unsigned int i;
    unsigned long long period = vm->def->cputune.period;
    long long quota = vm->def->cputune.quota;
W
Wen Congyang 已提交
492
    long long vm_quota = 0;
493 494 495 496 497 498 499 500 501 502 503 504

    if (driver->cgroup == NULL)
        return 0; /* Not supported, so claim success */

    rc = virCgroupForDomain(driver->cgroup, vm->def->name, &cgroup, 0);
    if (rc != 0) {
        virReportSystemError(-rc,
                             _("Unable to find cgroup for %s"),
                             vm->def->name);
        goto cleanup;
    }

W
Wen Congyang 已提交
505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
    /* Set cpu bandwidth for the vm */
    if (period || quota) {
        if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPU)) {
            /* Ensure that we can multiply by vcpus without overflowing. */
            if (quota > LLONG_MAX / vm->def->vcpus) {
                virReportSystemError(EINVAL,
                                     _("%s"),
                                     "Unable to set cpu bandwidth quota");
                goto cleanup;
            }

            if (quota > 0)
                vm_quota = quota * vm->def->vcpus;
            else
                vm_quota = quota;
            if (qemuSetupCgroupVcpuBW(cgroup, period, vm_quota) < 0)
                goto cleanup;
        }
    }

525 526
    if (priv->nvcpupids == 0 || priv->vcpupids[0] == vm->pid) {
        /* If we does not know VCPU<->PID mapping or all vcpu runs in the same
W
Wen Congyang 已提交
527
         * thread, we cannot control each vcpu.
528
         */
W
Wen Congyang 已提交
529
        virCgroupFree(&cgroup);
530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
        return 0;
    }

    for (i = 0; i < priv->nvcpupids; i++) {
        rc = virCgroupForVcpu(cgroup, i, &cgroup_vcpu, 1);
        if (rc < 0) {
            virReportSystemError(-rc,
                                 _("Unable to create vcpu cgroup for %s(vcpu:"
                                   " %d)"),
                                 vm->def->name, i);
            goto cleanup;
        }

        /* move the thread for vcpu to sub dir */
        rc = virCgroupAddTask(cgroup_vcpu, priv->vcpupids[i]);
        if (rc < 0) {
            virReportSystemError(-rc,
                                 _("unable to add vcpu %d task %d to cgroup"),
                                 i, priv->vcpupids[i]);
            goto cleanup;
        }

        if (period || quota) {
            if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPU)) {
                if (qemuSetupCgroupVcpuBW(cgroup_vcpu, period, quota) < 0)
                    goto cleanup;
            }
        }

        virCgroupFree(&cgroup_vcpu);
    }

    virCgroupFree(&cgroup_vcpu);
    virCgroupFree(&cgroup);
    return 0;

cleanup:
    virCgroupFree(&cgroup_vcpu);
    if (cgroup) {
        virCgroupRemove(cgroup);
        virCgroupFree(&cgroup);
    }

    return -1;
}

576 577 578 579 580 581 582 583 584 585 586 587 588 589 590

int qemuRemoveCgroup(struct qemud_driver *driver,
                     virDomainObjPtr vm,
                     int quiet)
{
    virCgroupPtr cgroup;
    int rc;

    if (driver->cgroup == NULL)
        return 0; /* Not supported, so claim success */

    rc = virCgroupForDomain(driver->cgroup, vm->def->name, &cgroup, 0);
    if (rc != 0) {
        if (!quiet)
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
E
Eric Blake 已提交
591
                            _("Unable to find cgroup for %s"),
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
                            vm->def->name);
        return rc;
    }

    rc = virCgroupRemove(cgroup);
    virCgroupFree(&cgroup);
    return rc;
}

int qemuAddToCgroup(struct qemud_driver *driver,
                    virDomainDefPtr def)
{
    virCgroupPtr cgroup = NULL;
    int ret = -1;
    int rc;

    if (driver->cgroup == NULL)
        return 0; /* Not supported, so claim success */

    rc = virCgroupForDomain(driver->cgroup, def->name, &cgroup, 0);
    if (rc != 0) {
        virReportSystemError(-rc,
                             _("unable to find cgroup for domain %s"),
                             def->name);
        goto cleanup;
    }

    rc = virCgroupAddTask(cgroup, getpid());
    if (rc != 0) {
        virReportSystemError(-rc,
                             _("unable to add domain %s task %d to cgroup"),
                             def->name, getpid());
        goto cleanup;
    }

    ret = 0;

cleanup:
    virCgroupFree(&cgroup);
    return ret;
}