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 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
|
/*
* %CopyrightBegin%
*
* Copyright Ericsson AB 2014-2018. All Rights Reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
* %CopyrightEnd%
*/
/*
* Description: Microstate accounting.
*
* We keep track of the different states that the
* Erlang VM threads are in, in order to provide
* performance/debugging statistics. There is a
* small overhead in enabling this, but in the big
* scheme of things it should be negligible.
*/
#ifdef HAVE_CONFIG_H
# include "config.h"
#endif
#define ERTS_MSACC_STATE_STRINGS 1
#include "sys.h"
#include "global.h"
#include "erl_threads.h"
#include "erl_bif_unique.h"
#include "erl_map.h"
#include "erl_msacc.h"
#include "erl_bif_table.h"
#if ERTS_ENABLE_MSACC
static Eterm erts_msacc_gather_stats(ErtsMsAcc *msacc, ErtsHeapFactory *factory);
static void erts_msacc_reset(ErtsMsAcc *msacc);
static ErtsMsAcc* get_msacc(void);
erts_tsd_key_t ERTS_WRITE_UNLIKELY(erts_msacc_key);
#ifndef ERTS_MSACC_ALWAYS_ON
int ERTS_WRITE_UNLIKELY(erts_msacc_enabled);
#endif
static Eterm *erts_msacc_state_atoms = NULL;
static erts_rwmtx_t msacc_mutex;
static ErtsMsAcc *msacc_managed = NULL;
static ErtsMsAcc *msacc_unmanaged = NULL;
static Uint msacc_unmanaged_count = 0;
#if ERTS_MSACC_STATE_COUNT < MAP_SMALL_MAP_LIMIT
#define DEFAULT_MSACC_MSG_SIZE (3 + 1 + ERTS_MSACC_STATE_COUNT * 2 + 3 + ERTS_REF_THING_SIZE)
#else
#define DEFAULT_MSACC_MSG_SIZE (3 + ERTS_MSACC_STATE_COUNT * 3 + 3 + ERTS_REF_THING_SIZE)
#endif
/* we have to split initiation as atoms are not inited in early init */
void erts_msacc_early_init(void) {
#ifndef ERTS_MSACC_ALWAYS_ON
erts_msacc_enabled = 0;
#endif
erts_rwmtx_init(&msacc_mutex, "msacc_list_mutex", NIL,
ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
erts_tsd_key_create(&erts_msacc_key,"erts_msacc_key");
}
void erts_msacc_init(void) {
int i;
erts_msacc_state_atoms = erts_alloc(ERTS_ALC_T_MSACC,
sizeof(Eterm)*ERTS_MSACC_STATE_COUNT);
for (i = 0; i < ERTS_MSACC_STATE_COUNT; i++) {
erts_msacc_state_atoms[i] = am_atom_put(erts_msacc_states[i],
sys_strlen(erts_msacc_states[i]));
}
}
void erts_msacc_init_thread(char *type, int id, int managed) {
ErtsMsAcc *msacc;
msacc = erts_alloc(ERTS_ALC_T_MSACC, sizeof(ErtsMsAcc) +
sizeof(ErtsMsAccPerfCntr) * ERTS_MSACC_STATE_COUNT);
msacc->type = strdup(type);
msacc->id = make_small(id);
msacc->unmanaged = !managed;
msacc->tid = erts_thr_self();
msacc->perf_counter = 0;
erts_rwmtx_rwlock(&msacc_mutex);
if (!managed) {
erts_mtx_init(&msacc->mtx, "msacc_unmanaged_mutex", NIL,
ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
msacc->next = msacc_unmanaged;
msacc_unmanaged = msacc;
msacc_unmanaged_count++;
ERTS_MSACC_TSD_SET(msacc);
} else {
msacc->next = msacc_managed;
msacc_managed = msacc;
}
erts_rwmtx_rwunlock(&msacc_mutex);
erts_msacc_reset(msacc);
#ifdef ERTS_MSACC_ALWAYS_ON
ERTS_MSACC_TSD_SET(msacc);
msacc->perf_counter = erts_sys_perf_counter();
msacc->state = ERTS_MSACC_STATE_OTHER;
#endif
}
#ifdef ERTS_MSACC_EXTENDED_STATES
void erts_msacc_set_bif_state(ErtsMsAcc *__erts_msacc_cache, Eterm mod, void *fn) {
#ifdef ERTS_MSACC_EXTENDED_BIFS
#define BIF_LIST(Mod,Func,Arity,BifFuncAddr,FuncAddr,Num) \
if (fn == &BifFuncAddr) { \
ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATIC_STATE_COUNT + Num); \
} else
#include "erl_bif_list.h"
#undef BIF_LIST
{ /* The last else in the macro expansion,
this happens for internal bifs, i.e. traps etc */
ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_BIF);
}
#else
if (mod == am_ets) {
ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_ETS);
} else {
ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_BIF);
}
#endif
}
#endif
/*
* Creates a structure looking like this
* #{ type => scheduler, id => 1, counters => #{ State1 => Counter1 ... StateN => CounterN}}
*/
static
Eterm erts_msacc_gather_stats(ErtsMsAcc *msacc, ErtsHeapFactory *factory) {
Uint sz = 0;
Eterm *hp, cvs[ERTS_MSACC_STATE_COUNT];
Eterm key, state_map;
int i;
flatmap_t *map;
hp = erts_produce_heap(factory, 4, 0);
key = TUPLE3(hp,am_counters,am_id,am_type);
for (i = 0; i < ERTS_MSACC_STATE_COUNT; i++) {
cvs[i] = erts_bld_sint64(NULL, &sz,(Sint64)msacc->counters[i].pc);
#ifdef ERTS_MSACC_STATE_COUNTERS
erts_bld_uint64(NULL,&sz,msacc->counters[i].sc);
sz += 3;
#endif
}
hp = erts_produce_heap(factory, sz, 0);
for (i = 0; i < ERTS_MSACC_STATE_COUNT; i++) {
cvs[i] = erts_bld_sint64(&hp,NULL,(Sint64)msacc->counters[i].pc);
#ifdef ERTS_MSACC_STATE_COUNTERS
Eterm counter__ = erts_bld_uint64(&hp,NULL,msacc->counters[i].sc);
cvs[i] = TUPLE2(hp,cvs[i],counter__);
hp += 3;
#endif
}
state_map = erts_map_from_ks_and_vs(factory, erts_msacc_state_atoms, cvs,
ERTS_MSACC_STATE_COUNT);
hp = erts_produce_heap(factory, MAP_HEADER_FLATMAP_SZ + 3, 0);
map = (flatmap_t*)hp;
hp += MAP_HEADER_FLATMAP_SZ;
map->thing_word = MAP_HEADER_FLATMAP;
map->size = 3;
map->keys = key;
hp[0] = state_map;
hp[1] = msacc->id;
hp[2] = am_atom_put(msacc->type,sys_strlen(msacc->type));
return make_flatmap(map);
}
typedef struct {
int action;
Process *proc;
Eterm ref;
Eterm ref_heap[ERTS_REF_THING_SIZE];
Uint req_sched;
erts_atomic32_t refc;
} ErtsMSAccReq;
static ErtsMsAcc* get_msacc(void) {
ErtsMsAcc *msacc;
erts_rwmtx_rlock(&msacc_mutex);
msacc = msacc_managed;
while (!erts_equal_tids(msacc->tid,erts_thr_self())) {
msacc = msacc->next;
ASSERT(msacc != NULL);
}
erts_rwmtx_runlock(&msacc_mutex);
return msacc;
}
static void send_reply(ErtsMsAcc *msacc, ErtsMSAccReq *msaccrp) {
ErtsSchedulerData *esdp = erts_get_scheduler_data();
Process *rp = msaccrp->proc;
ErtsMessage *msgp = NULL;
Eterm *hp;
Eterm ref_copy = NIL, msg;
ErtsProcLocks rp_locks = (esdp && msaccrp->req_sched == esdp->no
? ERTS_PROC_LOCK_MAIN : 0);
ErtsHeapFactory factory;
if (msaccrp->action == ERTS_MSACC_GATHER) {
msgp = erts_factory_message_create(&factory, rp, &rp_locks, DEFAULT_MSACC_MSG_SIZE);
if (msacc->unmanaged) erts_mtx_lock(&msacc->mtx);
hp = erts_produce_heap(&factory, ERTS_REF_THING_SIZE + 3 /* tuple */, 0);
ref_copy = STORE_NC(&hp, &msgp->hfrag.off_heap, msaccrp->ref);
msg = erts_msacc_gather_stats(msacc, &factory);
msg = TUPLE2(hp, ref_copy, msg);
if (msacc->unmanaged) erts_mtx_unlock(&msacc->mtx);
erts_factory_close(&factory);
} else {
ErlOffHeap *ohp = NULL;
msgp = erts_alloc_message_heap(rp, &rp_locks, ERTS_REF_THING_SIZE, &hp, &ohp);
msg = STORE_NC(&hp, &msgp->hfrag.off_heap, msaccrp->ref);
}
erts_queue_message(rp, rp_locks, msgp, msg, am_system);
if (esdp && msaccrp->req_sched == esdp->no)
rp_locks &= ~ERTS_PROC_LOCK_MAIN;
if (rp_locks)
erts_proc_unlock(rp, rp_locks);
}
static void
reply_msacc(void *vmsaccrp)
{
ErtsMsAcc *msacc = ERTS_MSACC_TSD_GET();
ErtsMSAccReq *msaccrp = (ErtsMSAccReq *) vmsaccrp;
ASSERT(!msacc || !msacc->unmanaged);
if (msaccrp->action == ERTS_MSACC_ENABLE && !msacc) {
msacc = get_msacc();
msacc->perf_counter = erts_sys_perf_counter();
msacc->state = ERTS_MSACC_STATE_OTHER;
ERTS_MSACC_TSD_SET(msacc);
} else if (msaccrp->action == ERTS_MSACC_DISABLE && msacc) {
ERTS_MSACC_TSD_SET(NULL);
} else if (msaccrp->action == ERTS_MSACC_RESET) {
msacc = msacc ? msacc : get_msacc();
erts_msacc_reset(msacc);
} else if (msaccrp->action == ERTS_MSACC_GATHER && !msacc) {
msacc = get_msacc();
}
ASSERT(!msacc || !msacc->unmanaged);
send_reply(msacc, msaccrp);
erts_proc_dec_refc(msaccrp->proc);
if (erts_atomic32_dec_read_nob(&msaccrp->refc) == 0)
erts_free(ERTS_ALC_T_MSACC, vmsaccrp);
}
static void erts_msacc_reset(ErtsMsAcc *msacc) {
int i;
if (msacc->unmanaged) erts_mtx_lock(&msacc->mtx);
for (i = 0; i < ERTS_MSACC_STATE_COUNT; i++) {
msacc->counters[i].pc = 0;
#ifdef ERTS_MSACC_STATE_COUNTERS
msacc->counters[i].sc = 0;
#endif
}
if (msacc->unmanaged) erts_mtx_unlock(&msacc->mtx);
}
#endif /* ERTS_ENABLE_MSACC */
/*
* This function is responsible for enabling, disabling, resetting and
* gathering data related to microstate accounting.
*
* Managed threads and unmanaged threads are handled differently.
* - managed threads get a misc_aux job telling them to switch on msacc
* - unmanaged have some fields protected by a mutex that has to be taken
* before any values can be updated
*
* For performance reasons there is also a global value erts_msacc_enabled
* that controls the state of all threads. Statistics gathering is only on
* if erts_msacc_enabled && msacc is true.
*/
Eterm
erts_msacc_request(Process *c_p, int action, Eterm *threads)
{
#ifdef ERTS_ENABLE_MSACC
ErtsMsAcc *msacc = ERTS_MSACC_TSD_GET();
ErtsSchedulerData *esdp = erts_proc_sched_data(c_p);
Eterm ref;
ErtsMSAccReq *msaccrp;
Eterm *hp;
#ifdef ERTS_MSACC_ALWAYS_ON
if (action == ERTS_MSACC_ENABLE || action == ERTS_MSACC_DISABLE)
return THE_NON_VALUE;
#else
/* take care of double enable, and double disable here */
if (msacc && action == ERTS_MSACC_ENABLE) {
return THE_NON_VALUE;
} else if (!msacc && action == ERTS_MSACC_DISABLE) {
return THE_NON_VALUE;
}
#endif
ref = erts_make_ref(c_p);
msaccrp = erts_alloc(ERTS_ALC_T_MSACC, sizeof(ErtsMSAccReq));
hp = &msaccrp->ref_heap[0];
msaccrp->action = action;
msaccrp->proc = c_p;
msaccrp->ref = STORE_NC(&hp, NULL, ref);
msaccrp->req_sched = esdp->no;
*threads = erts_no_schedulers;
*threads += 1; /* aux thread */
erts_atomic32_init_nob(&msaccrp->refc,(erts_aint32_t)*threads);
erts_proc_add_refc(c_p, *threads);
if (erts_no_schedulers > 1)
erts_schedule_multi_misc_aux_work(1,
erts_no_schedulers,
reply_msacc,
(void *) msaccrp);
/* aux thread */
erts_schedule_misc_aux_work(0, reply_msacc, (void *) msaccrp);
/* Manage unmanaged threads */
switch (action) {
case ERTS_MSACC_GATHER: {
Uint unmanaged_count;
ErtsMsAcc *msacc, **unmanaged;
int i = 0;
/* we copy a list of pointers here so that we do not have to have
the msacc_mutex when sending messages */
erts_rwmtx_rlock(&msacc_mutex);
unmanaged_count = msacc_unmanaged_count;
unmanaged = erts_alloc(ERTS_ALC_T_MSACC,
sizeof(ErtsMsAcc*)*unmanaged_count);
for (i = 0, msacc = msacc_unmanaged;
i < unmanaged_count;
i++, msacc = msacc->next) {
unmanaged[i] = msacc;
}
erts_rwmtx_runlock(&msacc_mutex);
for (i = 0; i < unmanaged_count; i++) {
erts_mtx_lock(&unmanaged[i]->mtx);
if (unmanaged[i]->perf_counter) {
ErtsSysPerfCounter perf_counter;
/* if enabled update stats */
perf_counter = erts_sys_perf_counter();
unmanaged[i]->counters[unmanaged[i]->state].pc +=
perf_counter - unmanaged[i]->perf_counter;
unmanaged[i]->perf_counter = perf_counter;
}
erts_mtx_unlock(&unmanaged[i]->mtx);
send_reply(unmanaged[i],msaccrp);
}
erts_free(ERTS_ALC_T_MSACC,unmanaged);
/* We have just sent unmanaged_count messages, so bump no of threads */
*threads += unmanaged_count;
break;
}
case ERTS_MSACC_RESET: {
ErtsMsAcc *msacc;
erts_rwmtx_rlock(&msacc_mutex);
for (msacc = msacc_unmanaged; msacc != NULL; msacc = msacc->next)
erts_msacc_reset(msacc);
erts_rwmtx_runlock(&msacc_mutex);
break;
}
case ERTS_MSACC_ENABLE: {
erts_rwmtx_rlock(&msacc_mutex);
for (msacc = msacc_unmanaged; msacc != NULL; msacc = msacc->next) {
erts_mtx_lock(&msacc->mtx);
msacc->perf_counter = erts_sys_perf_counter();
/* we assume the unmanaged thread is sleeping */
msacc->state = ERTS_MSACC_STATE_SLEEP;
erts_mtx_unlock(&msacc->mtx);
}
erts_rwmtx_runlock(&msacc_mutex);
break;
}
case ERTS_MSACC_DISABLE: {
ErtsSysPerfCounter perf_counter;
erts_rwmtx_rlock(&msacc_mutex);
/* make sure to update stats with latest results */
for (msacc = msacc_unmanaged; msacc != NULL; msacc = msacc->next) {
erts_mtx_lock(&msacc->mtx);
perf_counter = erts_sys_perf_counter();
msacc->counters[msacc->state].pc += perf_counter - msacc->perf_counter;
msacc->perf_counter = 0;
erts_mtx_unlock(&msacc->mtx);
}
erts_rwmtx_runlock(&msacc_mutex);
break;
}
default: { ASSERT(0); }
}
*threads = make_small(*threads);
reply_msacc((void *) msaccrp);
#ifndef ERTS_MSACC_ALWAYS_ON
/* enable/disable the global value */
if (action == ERTS_MSACC_ENABLE) {
erts_msacc_enabled = 1;
} else if (action == ERTS_MSACC_DISABLE) {
erts_msacc_enabled = 0;
}
#endif
return ref;
#else
return THE_NON_VALUE;
#endif
}
|