lxc_cgroup.c 13.1 KB
Newer Older
1
/*
2
 * Copyright (C) 2010-2014 Red Hat, Inc.
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright IBM Corp. 2008
 *
 * lxc_cgroup.c: LXC cgroup helpers
 *
 * 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
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25
 */

#include <config.h>

#include "lxc_cgroup.h"
#include "lxc_container.h"
26
#include "domain_cgroup.h"
27
#include "virfile.h"
28
#include "virerror.h"
29
#include "virlog.h"
30
#include "viralloc.h"
31
#include "virstring.h"
32
#include "virsystemd.h"
33 34 35

#define VIR_FROM_THIS VIR_FROM_LXC

36 37
VIR_LOG_INIT("lxc.lxc_cgroup");

38 39 40
static int virLXCCgroupSetupCpuTune(virDomainDefPtr def,
                                    virCgroupPtr cgroup)
{
41 42 43
    if (def->cputune.sharesSpecified) {
        unsigned long long val;
        if (virCgroupSetCpuShares(cgroup, def->cputune.shares) < 0)
44
            return -1;
45 46

        if (virCgroupGetCpuShares(cgroup, &val) < 0)
47
            return -1;
48 49
        def->cputune.shares = val;
    }
50 51 52

    if (def->cputune.quota != 0 &&
        virCgroupSetCpuCfsQuota(cgroup, def->cputune.quota) < 0)
53
        return -1;
54 55 56

    if (def->cputune.period != 0 &&
        virCgroupSetCpuCfsPeriod(cgroup, def->cputune.period) < 0)
57
        return -1;
58

59
    return 0;
60 61 62
}


63 64 65 66
static int virLXCCgroupSetupCpusetTune(virDomainDefPtr def,
                                       virCgroupPtr cgroup,
                                       virBitmapPtr nodemask)
{
67
    int ret = -1;
68
    char *mask = NULL;
69
    virDomainNumatuneMemMode mode;
70 71 72

    if (def->placement_mode != VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO &&
        def->cpumask) {
73
        if (!(mask = virBitmapFormat(def->cpumask)))
74 75
            return -1;

76
        if (virCgroupSetCpusetCpus(cgroup, mask) < 0)
77
            goto cleanup;
78 79
        /* free mask to make sure we won't use it in a wrong way later */
        VIR_FREE(mask);
80 81
    }

82 83
    if (virDomainNumatuneGetMode(def->numa, -1, &mode) < 0 ||
        mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT) {
84
        ret = 0;
85
        goto cleanup;
86
    }
87

88
    if (virDomainNumatuneMaybeFormatNodeset(def->numa, nodemask,
89
                                            &mask, -1) < 0)
90
        goto cleanup;
91

92 93
    if (mask && virCgroupSetCpusetMems(cgroup, mask) < 0)
        goto cleanup;
94

95
    ret = 0;
96
 cleanup:
97
    VIR_FREE(mask);
98
    return ret;
99 100 101
}


102 103 104
static int virLXCCgroupSetupBlkioTune(virDomainDefPtr def,
                                      virCgroupPtr cgroup)
{
105
    return virDomainCgroupSetupBlkio(cgroup, def->blkio);
106 107 108 109 110 111
}


static int virLXCCgroupSetupMemTune(virDomainDefPtr def,
                                    virCgroupPtr cgroup)
{
112
    if (virCgroupSetMemory(cgroup, virDomainDefGetMemoryInitial(def)) < 0)
113
        return -1;
114

115 116
    if (virMemoryLimitIsSet(def->mem.hard_limit))
        if (virCgroupSetMemoryHardLimit(cgroup, def->mem.hard_limit) < 0)
117
            return -1;
118

119 120
    if (virMemoryLimitIsSet(def->mem.soft_limit))
        if (virCgroupSetMemorySoftLimit(cgroup, def->mem.soft_limit) < 0)
121
            return -1;
122

123 124
    if (virMemoryLimitIsSet(def->mem.swap_hard_limit))
        if (virCgroupSetMemSwapHardLimit(cgroup, def->mem.swap_hard_limit) < 0)
125
            return -1;
126

127
    return 0;
128 129 130
}


131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
static int virLXCCgroupGetMemSwapUsage(virCgroupPtr cgroup,
                                       virLXCMeminfoPtr meminfo)
{
    return virCgroupGetMemSwapUsage(cgroup, &meminfo->swapusage);
}


static int virLXCCgroupGetMemSwapTotal(virCgroupPtr cgroup,
                                       virLXCMeminfoPtr meminfo)
{
    return virCgroupGetMemSwapHardLimit(cgroup, &meminfo->swaptotal);
}


static int virLXCCgroupGetMemUsage(virCgroupPtr cgroup,
                                   virLXCMeminfoPtr meminfo)
{
    int ret;
    unsigned long memUsage;

    ret = virCgroupGetMemoryUsage(cgroup, &memUsage);
152
    meminfo->memusage = (unsigned long long)memUsage;
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167

    return ret;
}


static int virLXCCgroupGetMemTotal(virCgroupPtr cgroup,
                                   virLXCMeminfoPtr meminfo)
{
    return virCgroupGetMemoryHardLimit(cgroup, &meminfo->memtotal);
}


static int virLXCCgroupGetMemStat(virCgroupPtr cgroup,
                                  virLXCMeminfoPtr meminfo)
{
P
Pavel Hrdina 已提交
168 169 170 171 172 173 174
    return virCgroupGetMemoryStat(cgroup,
                                  &meminfo->cached,
                                  &meminfo->inactive_anon,
                                  &meminfo->active_anon,
                                  &meminfo->inactive_file,
                                  &meminfo->active_file,
                                  &meminfo->unevictable);
175 176 177 178 179
}


int virLXCCgroupGetMeminfo(virLXCMeminfoPtr meminfo)
{
180
    int ret = -1;
181 182
    virCgroupPtr cgroup;

183 184
    if (virCgroupNewSelf(&cgroup) < 0)
        return -1;
185

186
    if (virLXCCgroupGetMemStat(cgroup, meminfo) < 0)
187 188
        goto cleanup;

189
    if (virLXCCgroupGetMemTotal(cgroup, meminfo) < 0)
190 191
        goto cleanup;

192
    if (virLXCCgroupGetMemUsage(cgroup, meminfo) < 0)
193 194
        goto cleanup;

195 196 197 198 199
    if (virLXCCgroupGetMemSwapTotal(cgroup, meminfo) < 0)
        goto cleanup;

    if (virLXCCgroupGetMemSwapUsage(cgroup, meminfo) < 0)
        goto cleanup;
200 201

    ret = 0;
202
 cleanup:
203
    virCgroupFree(&cgroup);
204 205 206 207 208
    return ret;
}



209 210 211 212 213 214 215 216 217 218
typedef struct _virLXCCgroupDevicePolicy virLXCCgroupDevicePolicy;
typedef virLXCCgroupDevicePolicy *virLXCCgroupDevicePolicyPtr;

struct _virLXCCgroupDevicePolicy {
    char type;
    int major;
    int minor;
};


219
int
J
Ján Tomko 已提交
220
virLXCSetupHostUSBDeviceCgroup(virUSBDevicePtr dev G_GNUC_UNUSED,
221 222 223 224 225 226
                               const char *path,
                               void *opaque)
{
    virCgroupPtr cgroup = opaque;

    VIR_DEBUG("Process path '%s' for USB device", path);
227
    if (virCgroupAllowDevicePath(cgroup, path,
228
                                 VIR_CGROUP_DEVICE_RWM, false) < 0)
229 230 231 232 233 234 235
        return -1;

    return 0;
}


int
J
Ján Tomko 已提交
236
virLXCTeardownHostUSBDeviceCgroup(virUSBDevicePtr dev G_GNUC_UNUSED,
237 238 239 240 241 242
                                  const char *path,
                                  void *opaque)
{
    virCgroupPtr cgroup = opaque;

    VIR_DEBUG("Process path '%s' for USB device", path);
243
    if (virCgroupDenyDevicePath(cgroup, path,
244
                                VIR_CGROUP_DEVICE_RWM, false) < 0)
245 246 247 248 249
        return -1;

    return 0;
}

250 251 252 253

static int virLXCCgroupSetupDeviceACL(virDomainDefPtr def,
                                      virCgroupPtr cgroup)
{
254
    int capMknod = def->caps_features[VIR_DOMAIN_PROCES_CAPS_FEATURE_MKNOD];
255 256 257 258 259 260 261 262 263
    size_t i;
    static virLXCCgroupDevicePolicy devices[] = {
        {'c', LXC_DEV_MAJ_MEMORY, LXC_DEV_MIN_NULL},
        {'c', LXC_DEV_MAJ_MEMORY, LXC_DEV_MIN_ZERO},
        {'c', LXC_DEV_MAJ_MEMORY, LXC_DEV_MIN_FULL},
        {'c', LXC_DEV_MAJ_MEMORY, LXC_DEV_MIN_RANDOM},
        {'c', LXC_DEV_MAJ_MEMORY, LXC_DEV_MIN_URANDOM},
        {'c', LXC_DEV_MAJ_TTY, LXC_DEV_MIN_TTY},
        {'c', LXC_DEV_MAJ_TTY, LXC_DEV_MIN_PTMX},
G
Gao feng 已提交
264
        {'c', LXC_DEV_MAJ_FUSE, LXC_DEV_MIN_FUSE},
265 266
        {0,   0, 0}};

267
    if (virCgroupDenyAllDevices(cgroup) < 0)
268
        return -1;
269

270
    /* white list mknod if CAP_MKNOD has to be kept */
J
Ján Tomko 已提交
271
    if (capMknod == VIR_TRISTATE_SWITCH_ON) {
272 273
        if (virCgroupAllowAllDevices(cgroup,
                                    VIR_CGROUP_DEVICE_MKNOD) < 0)
274
            return -1;
275 276
    }

277 278
    for (i = 0; devices[i].type != 0; i++) {
        virLXCCgroupDevicePolicyPtr dev = &devices[i];
279 280 281 282 283
        if (virCgroupAllowDevice(cgroup,
                                 dev->type,
                                 dev->major,
                                 dev->minor,
                                 VIR_CGROUP_DEVICE_RWM) < 0)
284
            return -1;
285 286
    }

287
    VIR_DEBUG("Allowing any disk block devs");
288
    for (i = 0; i < def->ndisks; i++) {
289 290
        if (virStorageSourceIsEmpty(def->disks[i]->src) ||
            !virStorageSourceIsBlockLocal(def->disks[i]->src))
291 292
            continue;

293
        if (virCgroupAllowDevicePath(cgroup,
294
                                     virDomainDiskGetSource(def->disks[i]),
295
                                     (def->disks[i]->src->readonly ?
296 297
                                      VIR_CGROUP_DEVICE_READ :
                                      VIR_CGROUP_DEVICE_RW) |
298
                                     VIR_CGROUP_DEVICE_MKNOD, false) < 0)
299
            return -1;
300 301
    }

302
    VIR_DEBUG("Allowing any filesystem block devs");
303
    for (i = 0; i < def->nfss; i++) {
304 305 306
        if (def->fss[i]->type != VIR_DOMAIN_FS_TYPE_BLOCK)
            continue;

307
        if (virCgroupAllowDevicePath(cgroup,
308
                                     def->fss[i]->src->path,
309 310
                                     def->fss[i]->readonly ?
                                     VIR_CGROUP_DEVICE_READ :
311
                                     VIR_CGROUP_DEVICE_RW, false) < 0)
312
            return -1;
313 314
    }

315
    VIR_DEBUG("Allowing any hostdev block devs");
316 317
    for (i = 0; i < def->nhostdevs; i++) {
        virDomainHostdevDefPtr hostdev = def->hostdevs[i];
318
        virDomainHostdevSubsysUSBPtr usbsrc = &hostdev->source.subsys.u.usb;
319
        virUSBDevicePtr usb;
320

321 322 323 324 325 326 327
        switch (hostdev->mode) {
        case VIR_DOMAIN_HOSTDEV_MODE_SUBSYS:
            if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB)
                continue;
            if (hostdev->missing)
                continue;

328
            if ((usb = virUSBDeviceNew(usbsrc->bus, usbsrc->device,
329
                                       NULL)) == NULL)
330
                return -1;
331

332
            if (virUSBDeviceFileIterate(usb, virLXCSetupHostUSBDeviceCgroup,
333 334
                                        cgroup) < 0) {
                virUSBDeviceFree(usb);
335
                return -1;
336
            }
337
            virUSBDeviceFree(usb);
338 339 340 341 342 343 344
            break;
        case VIR_DOMAIN_HOSTDEV_MODE_CAPABILITIES:
            switch (hostdev->source.caps.type) {
            case VIR_DOMAIN_HOSTDEV_CAPS_TYPE_STORAGE:
                if (virCgroupAllowDevicePath(cgroup,
                                             hostdev->source.caps.u.storage.block,
                                             VIR_CGROUP_DEVICE_RW |
345
                                             VIR_CGROUP_DEVICE_MKNOD, false) < 0)
346
                    return -1;
347
                break;
348 349 350 351
            case VIR_DOMAIN_HOSTDEV_CAPS_TYPE_MISC:
                if (virCgroupAllowDevicePath(cgroup,
                                             hostdev->source.caps.u.misc.chardev,
                                             VIR_CGROUP_DEVICE_RW |
352
                                             VIR_CGROUP_DEVICE_MKNOD, false) < 0)
353
                    return -1;
354
                break;
355 356 357 358 359 360
            default:
                break;
            }
        default:
            break;
        }
361 362
    }

363 364
    if (virCgroupAllowDevice(cgroup, 'c', LXC_DEV_MAJ_PTY, -1,
                             VIR_CGROUP_DEVICE_RWM) < 0)
365
        return -1;
366

367 368
    VIR_DEBUG("Device whitelist complete");

369
    return 0;
370 371 372
}


373
virCgroupPtr virLXCCgroupCreate(virDomainDefPtr def,
374 375 376
                                pid_t initpid,
                                size_t nnicindexes,
                                int *nicindexes)
377
{
378
    virCgroupPtr cgroup = NULL;
379
    char *machineName = virLXCDomainGetMachineName(def, 0);
380 381 382

    if (!machineName)
        goto cleanup;
383

384 385 386 387 388
    if (def->resource->partition[0] != '/') {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("Resource partition '%s' must start with '/'"),
                       def->resource->partition);
        goto cleanup;
389
    }
390

391
    if (virCgroupNewMachine(machineName,
392 393 394
                            "lxc",
                            def->uuid,
                            NULL,
395
                            initpid,
396
                            true,
397
                            nnicindexes, nicindexes,
398 399
                            def->resource->partition,
                            -1,
400
                            0,
401
                            &cgroup) < 0)
402
        goto cleanup;
403

404 405 406 407 408 409
    /* setup control group permissions for user namespace */
    if (def->idmap.uidmap) {
        if (virCgroupSetOwner(cgroup,
                              def->idmap.uidmap[0].target,
                              def->idmap.gidmap[0].target,
                              (1 << VIR_CGROUP_CONTROLLER_SYSTEMD)) < 0) {
410
            virCgroupFree(&cgroup);
411 412 413 414 415
            cgroup = NULL;
            goto cleanup;
        }
    }

416
 cleanup:
417 418
    VIR_FREE(machineName);

419 420 421 422 423
    return cgroup;
}


int virLXCCgroupSetup(virDomainDefPtr def,
424 425
                      virCgroupPtr cgroup,
                      virBitmapPtr nodemask)
426
{
427
    if (virLXCCgroupSetupCpuTune(def, cgroup) < 0)
428
        return -1;
429

430
    if (virLXCCgroupSetupCpusetTune(def, cgroup, nodemask) < 0)
431
        return -1;
432

433
    if (virLXCCgroupSetupBlkioTune(def, cgroup) < 0)
434
        return -1;
435 436

    if (virLXCCgroupSetupMemTune(def, cgroup) < 0)
437
        return -1;
438 439

    if (virLXCCgroupSetupDeviceACL(def, cgroup) < 0)
440
        return -1;
441

442
    return 0;
443
}