virperf.c 10.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
/*
 * virperf.c: methods for managing perf events
 *
 * 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, see
 * <http://www.gnu.org/licenses/>.
 *
 * Authors:
 *  Ren Qiaowei <qiaowei.ren@intel.com>
 */
#include <config.h>

#include <sys/ioctl.h>
#if defined HAVE_SYS_SYSCALL_H
# include <sys/syscall.h>
#endif

#include "virperf.h"
#include "viralloc.h"
#include "virerror.h"
#include "virlog.h"
#include "virfile.h"
#include "virstring.h"
#include "virtypedparam.h"

VIR_LOG_INIT("util.perf");

#define VIR_FROM_THIS VIR_FROM_PERF

VIR_ENUM_IMPL(virPerfEvent, VIR_PERF_EVENT_LAST,
Q
Qiaowei Ren 已提交
41 42
              "cmt", "mbmt", "mbml",
              "cpu_cycles", "instructions",
43
              "cache_references", "cache_misses",
44
              "branch_instructions", "branch_misses",
45
              "bus_cycles", "stalled_cycles_frontend",
46
              "stalled_cycles_backend", "ref_cpu_cycles",
47
              "cpu_clock", "task_clock", "page_faults",
48
              "context_switches", "cpu_migrations",
49
              "page_faults_min", "page_faults_maj",
50
              "alignment_faults", "emulation_faults");
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71

struct virPerfEvent {
    int fd;
    bool enabled;
    union {
        /* cmt */
        struct {
            int scale;
        } cmt;
    } efields;
};
typedef struct virPerfEvent *virPerfEventPtr;

struct virPerf {
    struct virPerfEvent events[VIR_PERF_EVENT_LAST];
};

#if defined(__linux__) && defined(HAVE_SYS_SYSCALL_H)

# include <linux/perf_event.h>

72 73 74 75 76 77
struct virPerfEventAttr {
    unsigned int attrType;
    unsigned long long attrConfig;
};

static struct virPerfEventAttr attrs[] = {
78 79 80 81 82 83 84 85
    [VIR_PERF_EVENT_CMT] = {
        .attrType = 0,
        .attrConfig = 1
    },
    [VIR_PERF_EVENT_MBMT] = {
        .attrType = 0,
        .attrConfig = 2
    },
86
    [VIR_PERF_EVENT_MBML] = {
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
        .attrType = 0,
        .attrConfig = 3
    },
    [VIR_PERF_EVENT_CPU_CYCLES] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_CPU_CYCLES
    },
    [VIR_PERF_EVENT_INSTRUCTIONS] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_INSTRUCTIONS
    },
    [VIR_PERF_EVENT_CACHE_REFERENCES] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_CACHE_REFERENCES
    },
    [VIR_PERF_EVENT_CACHE_MISSES] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_CACHE_MISSES
    },
    [VIR_PERF_EVENT_BRANCH_INSTRUCTIONS] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_BRANCH_INSTRUCTIONS
    },
    [VIR_PERF_EVENT_BRANCH_MISSES] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_BRANCH_MISSES
    },
    [VIR_PERF_EVENT_BUS_CYCLES] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_BUS_CYCLES
    },
    [VIR_PERF_EVENT_STALLED_CYCLES_FRONTEND] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_STALLED_CYCLES_FRONTEND
    },
    [VIR_PERF_EVENT_STALLED_CYCLES_BACKEND] = {
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_STALLED_CYCLES_BACKEND
    },
    [VIR_PERF_EVENT_REF_CPU_CYCLES] = {
127
# ifdef PERF_COUNT_HW_REF_CPU_CYCLES
128 129
        .attrType = PERF_TYPE_HARDWARE,
        .attrConfig = PERF_COUNT_HW_REF_CPU_CYCLES
130
# else
131 132
        .attrType = 0,
        .attrConfig = 0,
133 134
# endif
    },
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
    [VIR_PERF_EVENT_CPU_CLOCK] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_CPU_CLOCK
    },
    [VIR_PERF_EVENT_TASK_CLOCK] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_TASK_CLOCK
    },
    [VIR_PERF_EVENT_PAGE_FAULTS] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_PAGE_FAULTS
    },
    [VIR_PERF_EVENT_CONTEXT_SWITCHES] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_CONTEXT_SWITCHES
    },
    [VIR_PERF_EVENT_CPU_MIGRATIONS] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_CPU_MIGRATIONS
    },
    [VIR_PERF_EVENT_PAGE_FAULTS_MIN] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_PAGE_FAULTS_MIN
    },
    [VIR_PERF_EVENT_PAGE_FAULTS_MAJ] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_PAGE_FAULTS_MAJ
    },
    [VIR_PERF_EVENT_ALIGNMENT_FAULTS] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_ALIGNMENT_FAULTS
    },
    [VIR_PERF_EVENT_EMULATION_FAULTS] = {
        .attrType = PERF_TYPE_SOFTWARE,
        .attrConfig = PERF_COUNT_SW_EMULATION_FAULTS
    },
171
};
172
verify(ARRAY_CARDINALITY(attrs) == VIR_PERF_EVENT_LAST);
173 174
typedef struct virPerfEventAttr *virPerfEventAttrPtr;

175

176
static int
177
virPerfRdtAttrInit(void)
178 179 180
{
    char *buf = NULL;
    char *tmp = NULL;
181
    unsigned int attr_type = 0;
182

183
    if (virFileReadAllQuiet("/sys/devices/intel_cqm/type", 10, &buf) < 0)
184
        goto error;
185 186 187 188

    if ((tmp = strchr(buf, '\n')))
        *tmp = '\0';

189
    if (virStrToLong_ui(buf, NULL, 10, &attr_type) < 0) {
190
        virReportSystemError(errno, "%s",
191
                             _("failed to get rdt event type"));
192
        goto error;
193 194 195
    }
    VIR_FREE(buf);

196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
    attrs[VIR_PERF_EVENT_CMT].attrType = attr_type;
    attrs[VIR_PERF_EVENT_MBMT].attrType = attr_type;
    attrs[VIR_PERF_EVENT_MBML].attrType = attr_type;

    return 0;

 error:
    VIR_FREE(buf);
    return -1;
}


int
virPerfEventEnable(virPerfPtr perf,
                   virPerfEventType type,
                   pid_t pid)
{
    char *buf = NULL;
    struct perf_event_attr attr;
215
    virPerfEventPtr event = &(perf->events[type]);
216
    virPerfEventAttrPtr event_attr = &attrs[type];
217

218 219 220
    if (event->enabled)
        return 0;

221 222 223
    if (event_attr->attrType == 0 && (type == VIR_PERF_EVENT_CMT ||
                                       type == VIR_PERF_EVENT_MBMT ||
                                       type == VIR_PERF_EVENT_MBML)) {
224
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
225
                       _("unable to enable host cpu perf event for %s"),
226
                       virPerfEventTypeToString(type));
227 228 229 230
        return -1;
    }

    if (type == VIR_PERF_EVENT_CMT) {
231 232 233 234
        if (virFileReadAll("/sys/devices/intel_cqm/events/llc_occupancy.scale",
                           10, &buf) < 0)
            goto error;

235
        if (virStrToLong_i(buf, NULL, 10, &event->efields.cmt.scale) < 0) {
236 237 238 239
            virReportSystemError(errno, "%s",
                                 _("failed to get cmt scaling factor"));
            goto error;
        }
240 241

        VIR_FREE(buf);
242 243
    }

244 245 246 247 248 249 250 251 252
    memset(&attr, 0, sizeof(attr));
    attr.size = sizeof(attr);
    attr.inherit = 1;
    attr.disabled = 1;
    attr.enable_on_exec = 0;
    attr.type = event_attr->attrType;
    attr.config = event_attr->attrConfig;

    event->fd = syscall(__NR_perf_event_open, &attr, pid, -1, -1, 0);
253 254
    if (event->fd < 0) {
        virReportSystemError(errno,
255
                             _("unable to open host cpu perf event for %s"),
256
                             virPerfEventTypeToString(type));
257
        goto error;
258 259 260
    }

    if (ioctl(event->fd, PERF_EVENT_IOC_ENABLE) < 0) {
261
        virReportSystemError(errno,
262
                             _("unable to enable host cpu perf event for %s"),
263
                             virPerfEventTypeToString(type));
264
        goto error;
265 266 267 268 269
    }

    event->enabled = true;
    return 0;

270
 error:
271 272 273 274 275 276 277 278 279
    VIR_FORCE_CLOSE(event->fd);
    VIR_FREE(buf);
    return -1;
}

int
virPerfEventDisable(virPerfPtr perf,
                    virPerfEventType type)
{
280
    virPerfEventPtr event = &(perf->events[type]);
281

282 283 284
    if (!event->enabled)
        return 0;

285 286
    if (ioctl(event->fd, PERF_EVENT_IOC_DISABLE) < 0) {
        virReportSystemError(errno,
287
                             _("unable to disable host cpu perf event for %s"),
288
                             virPerfEventTypeToString(type));
289 290 291 292 293 294 295 296 297 298 299
        return -1;
    }

    event->enabled = false;
    VIR_FORCE_CLOSE(event->fd);
    return 0;
}

bool virPerfEventIsEnabled(virPerfPtr perf,
                           virPerfEventType type)
{
300
    return perf->events[type].enabled;
301 302 303 304 305 306 307
}

int
virPerfReadEvent(virPerfPtr perf,
                 virPerfEventType type,
                 uint64_t *value)
{
308 309
    virPerfEventPtr event = &perf->events[type];
    if (!event->enabled)
310 311
        return -1;

312
    if (saferead(event->fd, value, sizeof(uint64_t)) < 0) {
313 314 315 316 317 318 319 320 321 322 323 324
        virReportSystemError(errno, "%s",
                             _("Unable to read cache data"));
        return -1;
    }

    if (type == VIR_PERF_EVENT_CMT)
        *value *= event->efields.cmt.scale;

    return 0;
}

#else
325 326 327 328 329 330 331
static int
virPerfRdtAttrInit(void)
{
    return 0;
}


332 333
int
virPerfEventEnable(virPerfPtr perf ATTRIBUTE_UNUSED,
R
Roman Bogorodskiy 已提交
334
                   virPerfEventType type ATTRIBUTE_UNUSED,
335 336 337 338 339 340 341 342 343
                   pid_t pid ATTRIBUTE_UNUSED)
{
    virReportSystemError(ENXIO, "%s",
                         _("Perf not supported on this platform"));
    return -1;
}

int
virPerfEventDisable(virPerfPtr perf ATTRIBUTE_UNUSED,
R
Roman Bogorodskiy 已提交
344
                    virPerfEventType type ATTRIBUTE_UNUSED)
345 346 347 348 349 350 351
{
    virReportSystemError(ENXIO, "%s",
                         _("Perf not supported on this platform"));
    return -1;
}

bool
R
Roman Bogorodskiy 已提交
352 353
virPerfEventIsEnabled(virPerfPtr perf ATTRIBUTE_UNUSED,
                      virPerfEventType type ATTRIBUTE_UNUSED)
354 355 356 357 358
{
    return false;
}

int
R
Roman Bogorodskiy 已提交
359 360 361
virPerfReadEvent(virPerfPtr perf ATTRIBUTE_UNUSED,
                 virPerfEventType type ATTRIBUTE_UNUSED,
                 uint64_t *value ATTRIBUTE_UNUSED)
362 363 364 365 366 367 368
{
    virReportSystemError(ENXIO, "%s",
                         _("Perf not supported on this platform"));
    return -1;
}

#endif
369 370 371 372 373 374 375 376 377 378 379 380 381 382 383

virPerfPtr
virPerfNew(void)
{
    size_t i;
    virPerfPtr perf;

    if (VIR_ALLOC(perf) < 0)
        return NULL;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
        perf->events[i].fd = -1;
        perf->events[i].enabled = false;
    }

384 385 386
    if (virPerfRdtAttrInit() < 0)
        virResetLastError();

387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404
    return perf;
}

void
virPerfFree(virPerfPtr perf)
{
    size_t i;

    if (perf == NULL)
        return;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
        if (perf->events[i].enabled)
            virPerfEventDisable(perf, i);
    }

    VIR_FREE(perf);
}