qsp.c 23.4 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
/*
 * qsp.c - QEMU Synchronization Profiler
 *
 * Copyright (C) 2018, Emilio G. Cota <cota@braap.org>
 *
 * License: GNU GPL, version 2 or later.
 *   See the COPYING file in the top-level directory.
 *
 * QSP profiles the time spent in synchronization primitives, which can
 * help diagnose performance problems, e.g. scalability issues when
 * contention is high.
 *
 * The primitives currently supported are mutexes, recursive mutexes and
 * condition variables. Note that not all related functions are intercepted;
 * instead we profile only those functions that can have a performance impact,
 * either due to blocking (e.g. cond_wait, mutex_lock) or cache line
 * contention (e.g. mutex_lock, mutex_trylock).
 *
 * QSP's design focuses on speed and scalability. This is achieved
 * by having threads do their profiling entirely on thread-local data.
 * The appropriate thread-local data is found via a QHT, i.e. a concurrent hash
 * table. To aggregate data in order to generate a report, we iterate over
 * all entries in the hash table. Depending on the number of threads and
 * synchronization objects this might be expensive, but note that it is
 * very rarely called -- reports are generated only when requested by users.
 *
 * Reports are generated as a table where each row represents a call site. A
 * call site is the triplet formed by the __file__ and __LINE__ of the caller
 * as well as the address of the "object" (i.e. mutex, rec. mutex or condvar)
30 31 32
 * being operated on. Optionally, call sites that operate on different objects
 * of the same type can be coalesced, which can be particularly useful when
 * profiling dynamically-allocated objects.
33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
 *
 * Alternative designs considered:
 *
 * - Use an off-the-shelf profiler such as mutrace. This is not a viable option
 *   for us because QEMU has __malloc_hook set (by one of the libraries it
 *   uses); leaving this hook unset is required to avoid deadlock in mutrace.
 *
 * - Use a glib HT for each thread, protecting each HT with its own lock.
 *   This isn't simpler than the current design, and is 10% slower in the
 *   atomic_add-bench microbenchmark (-m option).
 *
 * - For reports, just use a binary tree as we aggregate data, instead of having
 *   an intermediate hash table. This would simplify the code only slightly, but
 *   would perform badly if there were many threads and objects to track.
 *
E
Emilio G. Cota 已提交
48 49 50 51 52
 * - Wrap operations on qsp entries with RCU read-side critical sections, so
 *   that qsp_reset() can delete entries. Unfortunately, the overhead of calling
 *   rcu_read_lock/unlock slows down atomic_add-bench -m by 24%. Having
 *   a snapshot that is updated on qsp_reset() avoids this overhead.
 *
53 54 55 56 57 58 59 60 61 62
 * Related Work:
 * - Lennart Poettering's mutrace: http://0pointer.de/blog/projects/mutrace.html
 * - Lozi, David, Thomas, Lawall and Muller. "Remote Core Locking: Migrating
 *   Critical-Section Execution to Improve the Performance of Multithreaded
 *   Applications", USENIX ATC'12.
 */
#include "qemu/osdep.h"
#include "qemu/thread.h"
#include "qemu/timer.h"
#include "qemu/qht.h"
E
Emilio G. Cota 已提交
63
#include "qemu/rcu.h"
64 65 66 67
#include "exec/tb-hash-xx.h"

enum QSPType {
    QSP_MUTEX,
68
    QSP_BQL_MUTEX,
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
    QSP_REC_MUTEX,
    QSP_CONDVAR,
};

struct QSPCallSite {
    const void *obj;
    const char *file; /* i.e. __FILE__; shortened later */
    int line;
    enum QSPType type;
};
typedef struct QSPCallSite QSPCallSite;

struct QSPEntry {
    void *thread_ptr;
    const QSPCallSite *callsite;
    uint64_t n_acqs;
    uint64_t ns;
86
    unsigned int n_objs; /* count of coalesced objs; only used for reporting */
87 88 89
};
typedef struct QSPEntry QSPEntry;

E
Emilio G. Cota 已提交
90 91 92 93 94 95
struct QSPSnapshot {
    struct rcu_head rcu;
    struct qht ht;
};
typedef struct QSPSnapshot QSPSnapshot;

96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114
/* initial sizing for hash tables */
#define QSP_INITIAL_SIZE 64

/* If this file is moved, QSP_REL_PATH should be updated accordingly */
#define QSP_REL_PATH "util/qsp.c"

/* this file's full path. Used to present all call sites with relative paths */
static size_t qsp_qemu_path_len;

/* the address of qsp_thread gives us a unique 'thread ID' */
static __thread int qsp_thread;

/*
 * Call sites are the same for all threads, so we track them in a separate hash
 * table to save memory.
 */
static struct qht qsp_callsite_ht;

static struct qht qsp_ht;
E
Emilio G. Cota 已提交
115
static QSPSnapshot *qsp_snapshot;
116 117 118 119
static bool qsp_initialized, qsp_initializing;

static const char * const qsp_typenames[] = {
    [QSP_MUTEX]     = "mutex",
120
    [QSP_BQL_MUTEX] = "BQL mutex",
121 122 123 124
    [QSP_REC_MUTEX] = "rec_mutex",
    [QSP_CONDVAR]   = "condvar",
};

125
QemuMutexLockFunc qemu_bql_mutex_lock_func = qemu_mutex_lock_impl;
126 127 128 129 130 131 132 133 134 135 136 137
QemuMutexLockFunc qemu_mutex_lock_func = qemu_mutex_lock_impl;
QemuMutexTrylockFunc qemu_mutex_trylock_func = qemu_mutex_trylock_impl;
QemuRecMutexLockFunc qemu_rec_mutex_lock_func = qemu_rec_mutex_lock_impl;
QemuRecMutexTrylockFunc qemu_rec_mutex_trylock_func =
    qemu_rec_mutex_trylock_impl;
QemuCondWaitFunc qemu_cond_wait_func = qemu_cond_wait_impl;

/*
 * It pays off to _not_ hash callsite->file; hashing a string is slow, and
 * without it we still get a pretty unique hash.
 */
static inline
138
uint32_t do_qsp_callsite_hash(const QSPCallSite *callsite, uint64_t ab)
139
{
140
    uint64_t cd = (uint64_t)(uintptr_t)callsite->obj;
141 142 143
    uint32_t e = callsite->line;
    uint32_t f = callsite->type;

144
    return qemu_xxhash6(ab, cd, e, f);
145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167
}

static inline
uint32_t qsp_callsite_hash(const QSPCallSite *callsite)
{
    return do_qsp_callsite_hash(callsite, 0);
}

static inline uint32_t do_qsp_entry_hash(const QSPEntry *entry, uint64_t a)
{
    return do_qsp_callsite_hash(entry->callsite, a);
}

static uint32_t qsp_entry_hash(const QSPEntry *entry)
{
    return do_qsp_entry_hash(entry, (uint64_t)(uintptr_t)entry->thread_ptr);
}

static uint32_t qsp_entry_no_thread_hash(const QSPEntry *entry)
{
    return do_qsp_entry_hash(entry, 0);
}

168 169 170 171
/* without the objects we need to hash the file name to get a decent hash */
static uint32_t qsp_entry_no_thread_obj_hash(const QSPEntry *entry)
{
    const QSPCallSite *callsite = entry->callsite;
172 173
    uint64_t ab = g_str_hash(callsite->file);
    uint64_t cd = callsite->line;
174 175
    uint32_t e = callsite->type;

176
    return qemu_xxhash5(ab, cd, e);
177 178
}

179 180 181 182 183 184 185 186 187 188 189 190
static bool qsp_callsite_cmp(const void *ap, const void *bp)
{
    const QSPCallSite *a = ap;
    const QSPCallSite *b = bp;

    return a == b ||
        (a->obj == b->obj &&
         a->line == b->line &&
         a->type == b->type &&
         (a->file == b->file || !strcmp(a->file, b->file)));
}

191 192 193 194 195 196 197 198 199 200 201
static bool qsp_callsite_no_obj_cmp(const void *ap, const void *bp)
{
    const QSPCallSite *a = ap;
    const QSPCallSite *b = bp;

    return a == b ||
        (a->line == b->line &&
         a->type == b->type &&
         (a->file == b->file || !strcmp(a->file, b->file)));
}

202 203 204 205 206 207 208 209
static bool qsp_entry_no_thread_cmp(const void *ap, const void *bp)
{
    const QSPEntry *a = ap;
    const QSPEntry *b = bp;

    return qsp_callsite_cmp(a->callsite, b->callsite);
}

210 211 212 213 214 215 216 217
static bool qsp_entry_no_thread_obj_cmp(const void *ap, const void *bp)
{
    const QSPEntry *a = ap;
    const QSPEntry *b = bp;

    return qsp_callsite_no_obj_cmp(a->callsite, b->callsite);
}

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
static bool qsp_entry_cmp(const void *ap, const void *bp)
{
    const QSPEntry *a = ap;
    const QSPEntry *b = bp;

    return a->thread_ptr == b->thread_ptr &&
        qsp_callsite_cmp(a->callsite, b->callsite);
}

/*
 * Normally we'd call this from a constructor function, but we want it to work
 * via libutil as well.
 */
static void qsp_do_init(void)
{
    /* make sure this file's path in the tree is up to date with QSP_REL_PATH */
    g_assert(strstr(__FILE__, QSP_REL_PATH));
    qsp_qemu_path_len = strlen(__FILE__) - strlen(QSP_REL_PATH);

    qht_init(&qsp_ht, qsp_entry_cmp, QSP_INITIAL_SIZE,
             QHT_MODE_AUTO_RESIZE | QHT_MODE_RAW_MUTEXES);
    qht_init(&qsp_callsite_ht, qsp_callsite_cmp, QSP_INITIAL_SIZE,
             QHT_MODE_AUTO_RESIZE | QHT_MODE_RAW_MUTEXES);
}

static __attribute__((noinline)) void qsp_init__slowpath(void)
{
    if (atomic_cmpxchg(&qsp_initializing, false, true) == false) {
        qsp_do_init();
        atomic_set(&qsp_initialized, true);
    } else {
        while (!atomic_read(&qsp_initialized)) {
            cpu_relax();
        }
    }
}

/* qsp_init() must be called from _all_ exported functions */
static inline void qsp_init(void)
{
    if (likely(atomic_read(&qsp_initialized))) {
        return;
    }
    qsp_init__slowpath();
}

static QSPCallSite *qsp_callsite_find(const QSPCallSite *orig)
{
    QSPCallSite *callsite;
    uint32_t hash;

    hash = qsp_callsite_hash(orig);
    callsite = qht_lookup(&qsp_callsite_ht, orig, hash);
    if (callsite == NULL) {
        void *existing = NULL;

        callsite = g_new(QSPCallSite, 1);
        memcpy(callsite, orig, sizeof(*callsite));
        qht_insert(&qsp_callsite_ht, callsite, hash, &existing);
        if (unlikely(existing)) {
            g_free(callsite);
            callsite = existing;
        }
    }
    return callsite;
}

static QSPEntry *
qsp_entry_create(struct qht *ht, const QSPEntry *entry, uint32_t hash)
{
    QSPEntry *e;
    void *existing = NULL;

    e = g_new0(QSPEntry, 1);
    e->thread_ptr = entry->thread_ptr;
    e->callsite = qsp_callsite_find(entry->callsite);

    qht_insert(ht, e, hash, &existing);
    if (unlikely(existing)) {
        g_free(e);
        e = existing;
    }
    return e;
}

static QSPEntry *
qsp_entry_find(struct qht *ht, const QSPEntry *entry, uint32_t hash)
{
    QSPEntry *e;

    e = qht_lookup(ht, entry, hash);
    if (e == NULL) {
        e = qsp_entry_create(ht, entry, hash);
    }
    return e;
}

/*
 * Note: Entries are never removed, so callers do not have to be in an RCU
 * read-side critical section.
 */
static QSPEntry *qsp_entry_get(const void *obj, const char *file, int line,
                               enum QSPType type)
{
    QSPCallSite callsite = {
        .obj = obj,
        .file = file,
        .line = line,
        .type = type,
    };
    QSPEntry orig;
    uint32_t hash;

    qsp_init();

    orig.thread_ptr = &qsp_thread;
    orig.callsite = &callsite;

    hash = qsp_entry_hash(&orig);
    return qsp_entry_find(&qsp_ht, &orig, hash);
}

/*
 * @e is in the global hash table; it is only written to by the current thread,
 * so we write to it atomically (as in "write once") to prevent torn reads.
 */
static inline void do_qsp_entry_record(QSPEntry *e, int64_t delta, bool acq)
{
346
    atomic_set_u64(&e->ns, e->ns + delta);
347
    if (acq) {
348
        atomic_set_u64(&e->n_acqs, e->n_acqs + 1);
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
    }
}

static inline void qsp_entry_record(QSPEntry *e, int64_t delta)
{
    do_qsp_entry_record(e, delta, true);
}

#define QSP_GEN_VOID(type_, qsp_t_, func_, impl_)                       \
    static void func_(type_ *obj, const char *file, int line)           \
    {                                                                   \
        QSPEntry *e;                                                    \
        int64_t t0, t1;                                                 \
                                                                        \
        t0 = get_clock();                                               \
        impl_(obj, file, line);                                         \
        t1 = get_clock();                                               \
                                                                        \
        e = qsp_entry_get(obj, file, line, qsp_t_);                     \
        qsp_entry_record(e, t1 - t0);                                   \
    }

#define QSP_GEN_RET1(type_, qsp_t_, func_, impl_)                       \
    static int func_(type_ *obj, const char *file, int line)            \
    {                                                                   \
        QSPEntry *e;                                                    \
        int64_t t0, t1;                                                 \
        int err;                                                        \
                                                                        \
        t0 = get_clock();                                               \
        err = impl_(obj, file, line);                                   \
        t1 = get_clock();                                               \
                                                                        \
        e = qsp_entry_get(obj, file, line, qsp_t_);                     \
        do_qsp_entry_record(e, t1 - t0, !err);                          \
        return err;                                                     \
    }

387
QSP_GEN_VOID(QemuMutex, QSP_BQL_MUTEX, qsp_bql_mutex_lock, qemu_mutex_lock_impl)
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
QSP_GEN_VOID(QemuMutex, QSP_MUTEX, qsp_mutex_lock, qemu_mutex_lock_impl)
QSP_GEN_RET1(QemuMutex, QSP_MUTEX, qsp_mutex_trylock, qemu_mutex_trylock_impl)

QSP_GEN_VOID(QemuRecMutex, QSP_REC_MUTEX, qsp_rec_mutex_lock,
             qemu_rec_mutex_lock_impl)
QSP_GEN_RET1(QemuRecMutex, QSP_REC_MUTEX, qsp_rec_mutex_trylock,
             qemu_rec_mutex_trylock_impl)

#undef QSP_GEN_RET1
#undef QSP_GEN_VOID

static void
qsp_cond_wait(QemuCond *cond, QemuMutex *mutex, const char *file, int line)
{
    QSPEntry *e;
    int64_t t0, t1;

    t0 = get_clock();
    qemu_cond_wait_impl(cond, mutex, file, line);
    t1 = get_clock();

    e = qsp_entry_get(cond, file, line, QSP_CONDVAR);
    qsp_entry_record(e, t1 - t0);
}

bool qsp_is_enabled(void)
{
    return atomic_read(&qemu_mutex_lock_func) == qsp_mutex_lock;
}

void qsp_enable(void)
{
    atomic_set(&qemu_mutex_lock_func, qsp_mutex_lock);
    atomic_set(&qemu_mutex_trylock_func, qsp_mutex_trylock);
422
    atomic_set(&qemu_bql_mutex_lock_func, qsp_bql_mutex_lock);
423 424 425 426 427 428 429 430 431
    atomic_set(&qemu_rec_mutex_lock_func, qsp_rec_mutex_lock);
    atomic_set(&qemu_rec_mutex_trylock_func, qsp_rec_mutex_trylock);
    atomic_set(&qemu_cond_wait_func, qsp_cond_wait);
}

void qsp_disable(void)
{
    atomic_set(&qemu_mutex_lock_func, qemu_mutex_lock_impl);
    atomic_set(&qemu_mutex_trylock_func, qemu_mutex_trylock_impl);
432
    atomic_set(&qemu_bql_mutex_lock_func, qemu_mutex_lock_impl);
433 434 435 436 437 438 439 440 441
    atomic_set(&qemu_rec_mutex_lock_func, qemu_rec_mutex_lock_impl);
    atomic_set(&qemu_rec_mutex_trylock_func, qemu_rec_mutex_trylock_impl);
    atomic_set(&qemu_cond_wait_func, qemu_cond_wait_impl);
}

static gint qsp_tree_cmp(gconstpointer ap, gconstpointer bp, gpointer up)
{
    const QSPEntry *a = ap;
    const QSPEntry *b = bp;
442
    enum QSPSortBy sort_by = *(enum QSPSortBy *)up;
443 444 445
    const QSPCallSite *ca;
    const QSPCallSite *cb;

446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464
    switch (sort_by) {
    case QSP_SORT_BY_TOTAL_WAIT_TIME:
        if (a->ns > b->ns) {
            return -1;
        } else if (a->ns < b->ns) {
            return 1;
        }
        break;
    case QSP_SORT_BY_AVG_WAIT_TIME:
    {
        double avg_a = a->n_acqs ? a->ns / a->n_acqs : 0;
        double avg_b = b->n_acqs ? b->ns / b->n_acqs : 0;

        if (avg_a > avg_b) {
            return -1;
        } else if (avg_a < avg_b) {
            return 1;
        }
        break;
465
    }
466 467 468 469
    default:
        g_assert_not_reached();
    }

470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497
    ca = a->callsite;
    cb = b->callsite;
    /* Break the tie with the object's address */
    if (ca->obj < cb->obj) {
        return -1;
    } else if (ca->obj > cb->obj) {
        return 1;
    } else {
        int cmp;

        /* same obj. Break the tie with the callsite's file */
        cmp = strcmp(ca->file, cb->file);
        if (cmp) {
            return cmp;
        }
        /* same callsite file. Break the tie with the callsite's line */
        g_assert(ca->line != cb->line);
        if (ca->line < cb->line) {
            return -1;
        } else if (ca->line > cb->line) {
            return 1;
        } else {
            /* break the tie with the callsite's type */
            return cb->type - ca->type;
        }
    }
}

498
static void qsp_sort(void *p, uint32_t h, void *userp)
499 500 501 502 503 504 505
{
    QSPEntry *e = p;
    GTree *tree = userp;

    g_tree_insert(tree, e, NULL);
}

506
static void qsp_aggregate(void *p, uint32_t h, void *up)
507 508 509 510 511 512 513 514
{
    struct qht *ht = up;
    const QSPEntry *e = p;
    QSPEntry *agg;
    uint32_t hash;

    hash = qsp_entry_no_thread_hash(e);
    agg = qsp_entry_find(ht, e, hash);
515 516 517 518 519 520
    /*
     * The entry is in the global hash table; read from it atomically (as in
     * "read once").
     */
    agg->ns += atomic_read_u64(&e->ns);
    agg->n_acqs += atomic_read_u64(&e->n_acqs);
521 522
}

523
static void qsp_iter_diff(void *p, uint32_t hash, void *htp)
E
Emilio G. Cota 已提交
524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552
{
    struct qht *ht = htp;
    QSPEntry *old = p;
    QSPEntry *new;

    new = qht_lookup(ht, old, hash);
    /* entries are never deleted, so we must have this one */
    g_assert(new != NULL);
    /* our reading of the stats happened after the snapshot was taken */
    g_assert(new->n_acqs >= old->n_acqs);
    g_assert(new->ns >= old->ns);

    new->n_acqs -= old->n_acqs;
    new->ns -= old->ns;

    /* No point in reporting an empty entry */
    if (new->n_acqs == 0 && new->ns == 0) {
        bool removed = qht_remove(ht, new, hash);

        g_assert(removed);
        g_free(new);
    }
}

static void qsp_diff(struct qht *orig, struct qht *new)
{
    qht_iter(orig, qsp_iter_diff, new);
}

553
static void qsp_iter_callsite_coalesce(void *p, uint32_t h, void *htp)
554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571
{
    struct qht *ht = htp;
    QSPEntry *old = p;
    QSPEntry *e;
    uint32_t hash;

    hash = qsp_entry_no_thread_obj_hash(old);
    e = qht_lookup(ht, old, hash);
    if (e == NULL) {
        e = qsp_entry_create(ht, old, hash);
        e->n_objs = 1;
    } else if (e->callsite->obj != old->callsite->obj) {
        e->n_objs++;
    }
    e->ns += old->ns;
    e->n_acqs += old->n_acqs;
}

572
static void qsp_ht_delete(void *p, uint32_t h, void *htp)
E
Emilio G. Cota 已提交
573 574 575 576
{
    g_free(p);
}

577
static void qsp_mktree(GTree *tree, bool callsite_coalesce)
578
{
E
Emilio G. Cota 已提交
579
    QSPSnapshot *snap;
580 581
    struct qht ht, coalesce_ht;
    struct qht *htp;
582

E
Emilio G. Cota 已提交
583 584 585 586 587 588 589 590 591 592 593
    /*
     * First, see if there's a prior snapshot, so that we read the global hash
     * table _after_ the snapshot has been created, which guarantees that
     * the entries we'll read will be a superset of the snapshot's entries.
     *
     * We must remain in an RCU read-side critical section until we're done
     * with the snapshot.
     */
    rcu_read_lock();
    snap = atomic_rcu_read(&qsp_snapshot);

594 595 596 597 598
    /* Aggregate all results from the global hash table into a local one */
    qht_init(&ht, qsp_entry_no_thread_cmp, QSP_INITIAL_SIZE,
             QHT_MODE_AUTO_RESIZE | QHT_MODE_RAW_MUTEXES);
    qht_iter(&qsp_ht, qsp_aggregate, &ht);

E
Emilio G. Cota 已提交
599 600 601 602 603 604 605
    /* compute the difference wrt the snapshot, if any */
    if (snap) {
        qsp_diff(&snap->ht, &ht);
    }
    /* done with the snapshot; RCU can reclaim it */
    rcu_read_unlock();

606 607 608 609 610 611 612 613 614 615 616 617
    htp = &ht;
    if (callsite_coalesce) {
        qht_init(&coalesce_ht, qsp_entry_no_thread_obj_cmp, QSP_INITIAL_SIZE,
                 QHT_MODE_AUTO_RESIZE | QHT_MODE_RAW_MUTEXES);
        qht_iter(&ht, qsp_iter_callsite_coalesce, &coalesce_ht);

        /* free the previous hash table, and point htp to coalesce_ht */
        qht_iter(&ht, qsp_ht_delete, NULL);
        qht_destroy(&ht);
        htp = &coalesce_ht;
    }

618
    /* sort the hash table elements by using a tree */
619
    qht_iter(htp, qsp_sort, tree);
620 621

    /* free the hash table, but keep the elements (those are in the tree now) */
622
    qht_destroy(htp);
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647
}

/* free string with g_free */
static char *qsp_at(const QSPCallSite *callsite)
{
    GString *s = g_string_new(NULL);
    const char *shortened;

    /* remove the absolute path to qemu */
    if (unlikely(strlen(callsite->file) < qsp_qemu_path_len)) {
        shortened = callsite->file;
    } else {
        shortened = callsite->file + qsp_qemu_path_len;
    }
    g_string_append_printf(s, "%s:%u", shortened, callsite->line);
    return g_string_free(s, FALSE);
}

struct QSPReportEntry {
    const void *obj;
    char *callsite_at;
    const char *typename;
    double time_s;
    double ns_avg;
    uint64_t n_acqs;
648
    unsigned int n_objs;
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
};
typedef struct QSPReportEntry QSPReportEntry;

struct QSPReport {
    QSPReportEntry *entries;
    size_t n_entries;
    size_t max_n_entries;
};
typedef struct QSPReport QSPReport;

static gboolean qsp_tree_report(gpointer key, gpointer value, gpointer udata)
{
    const QSPEntry *e = key;
    QSPReport *report = udata;
    QSPReportEntry *entry;

    if (report->n_entries == report->max_n_entries) {
        return TRUE;
    }
    entry = &report->entries[report->n_entries];
    report->n_entries++;

    entry->obj = e->callsite->obj;
672
    entry->n_objs = e->n_objs;
673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716
    entry->callsite_at = qsp_at(e->callsite);
    entry->typename = qsp_typenames[e->callsite->type];
    entry->time_s = e->ns * 1e-9;
    entry->n_acqs = e->n_acqs;
    entry->ns_avg = e->n_acqs ? e->ns / e->n_acqs : 0;
    return FALSE;
}

static void
pr_report(const QSPReport *rep, FILE *f, fprintf_function pr)
{
    char *dashes;
    size_t max_len = 0;
    int callsite_len = 0;
    int callsite_rspace;
    int n_dashes;
    size_t i;

    /* find out the maximum length of all 'callsite' fields */
    for (i = 0; i < rep->n_entries; i++) {
        const QSPReportEntry *e = &rep->entries[i];
        size_t len = strlen(e->callsite_at);

        if (len > max_len) {
            max_len = len;
        }
    }

    callsite_len = MAX(max_len, strlen("Call site"));
    /* white space to leave to the right of "Call site" */
    callsite_rspace = callsite_len - strlen("Call site");

    pr(f, "Type               Object  Call site%*s  Wait Time (s)  "
       "       Count  Average (us)\n", callsite_rspace, "");

    /* build a horizontal rule with dashes */
    n_dashes = 79 + callsite_rspace;
    dashes = g_malloc(n_dashes + 1);
    memset(dashes, '-', n_dashes);
    dashes[n_dashes] = '\0';
    pr(f, "%s\n", dashes);

    for (i = 0; i < rep->n_entries; i++) {
        const QSPReportEntry *e = &rep->entries[i];
717
        GString *s = g_string_new(NULL);
718

719 720 721 722 723 724 725 726 727 728 729 730
        g_string_append_printf(s, "%-9s  ", e->typename);
        if (e->n_objs > 1) {
            g_string_append_printf(s, "[%12u]", e->n_objs);
        } else {
            g_string_append_printf(s, "%14p", e->obj);
        }
        g_string_append_printf(s, "  %s%*s  %13.5f  %12" PRIu64 "  %12.2f\n",
                               e->callsite_at,
                               callsite_len - (int)strlen(e->callsite_at), "",
                               e->time_s, e->n_acqs, e->ns_avg * 1e-3);
        pr(f, "%s", s->str);
        g_string_free(s, TRUE);
731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748
    }

    pr(f, "%s\n", dashes);
    g_free(dashes);
}

static void report_destroy(QSPReport *rep)
{
    size_t i;

    for (i = 0; i < rep->n_entries; i++) {
        QSPReportEntry *e = &rep->entries[i];

        g_free(e->callsite_at);
    }
    g_free(rep->entries);
}

749
void qsp_report(FILE *f, fprintf_function cpu_fprintf, size_t max,
750
                enum QSPSortBy sort_by, bool callsite_coalesce)
751
{
752
    GTree *tree = g_tree_new_full(qsp_tree_cmp, &sort_by, g_free, NULL);
753 754 755 756 757 758 759 760
    QSPReport rep;

    qsp_init();

    rep.entries = g_new0(QSPReportEntry, max);
    rep.n_entries = 0;
    rep.max_n_entries = max;

761
    qsp_mktree(tree, callsite_coalesce);
762 763 764 765 766 767
    g_tree_foreach(tree, qsp_tree_report, &rep);
    g_tree_destroy(tree);

    pr_report(&rep, f, cpu_fprintf);
    report_destroy(&rep);
}
E
Emilio G. Cota 已提交
768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794

static void qsp_snapshot_destroy(QSPSnapshot *snap)
{
    qht_iter(&snap->ht, qsp_ht_delete, NULL);
    qht_destroy(&snap->ht);
    g_free(snap);
}

void qsp_reset(void)
{
    QSPSnapshot *new = g_new(QSPSnapshot, 1);
    QSPSnapshot *old;

    qsp_init();

    qht_init(&new->ht, qsp_entry_cmp, QSP_INITIAL_SIZE,
             QHT_MODE_AUTO_RESIZE | QHT_MODE_RAW_MUTEXES);

    /* take a snapshot of the current state */
    qht_iter(&qsp_ht, qsp_aggregate, &new->ht);

    /* replace the previous snapshot, if any */
    old = atomic_xchg(&qsp_snapshot, new);
    if (old) {
        call_rcu(old, qsp_snapshot_destroy, rcu);
    }
}
新手
引导
客服 返回
顶部