qemu/trace/simple.c
<<
>>
Prefs
   1/*
   2 * Simple trace backend
   3 *
   4 * Copyright IBM, Corp. 2010
   5 *
   6 * This work is licensed under the terms of the GNU GPL, version 2.  See
   7 * the COPYING file in the top-level directory.
   8 *
   9 */
  10
  11#include "qemu/osdep.h"
  12#ifndef _WIN32
  13#include <pthread.h>
  14#endif
  15#include "qemu/timer.h"
  16#include "trace.h"
  17#include "trace/control.h"
  18#include "trace/simple.h"
  19
  20/** Trace file header event ID */
  21#define HEADER_EVENT_ID (~(uint64_t)0) /* avoids conflicting with TraceEventIDs */
  22
  23/** Trace file magic number */
  24#define HEADER_MAGIC 0xf2b177cb0aa429b4ULL
  25
  26/** Trace file version number, bump if format changes */
  27#define HEADER_VERSION 3
  28
  29/** Records were dropped event ID */
  30#define DROPPED_EVENT_ID (~(uint64_t)0 - 1)
  31
  32/** Trace record is valid */
  33#define TRACE_RECORD_VALID ((uint64_t)1 << 63)
  34
  35/*
  36 * Trace records are written out by a dedicated thread.  The thread waits for
  37 * records to become available, writes them out, and then waits again.
  38 */
  39static CompatGMutex trace_lock;
  40static CompatGCond trace_available_cond;
  41static CompatGCond trace_empty_cond;
  42
  43static bool trace_available;
  44static bool trace_writeout_enabled;
  45
  46enum {
  47    TRACE_BUF_LEN = 4096 * 64,
  48    TRACE_BUF_FLUSH_THRESHOLD = TRACE_BUF_LEN / 4,
  49};
  50
  51uint8_t trace_buf[TRACE_BUF_LEN];
  52static volatile gint trace_idx;
  53static unsigned int writeout_idx;
  54static volatile gint dropped_events;
  55static uint32_t trace_pid;
  56static FILE *trace_fp;
  57static char *trace_file_name;
  58
  59/* * Trace buffer entry */
  60typedef struct {
  61    uint64_t event; /*   TraceEventID */
  62    uint64_t timestamp_ns;
  63    uint32_t length;   /*    in bytes */
  64    uint32_t pid;
  65    uint64_t arguments[];
  66} TraceRecord;
  67
  68typedef struct {
  69    uint64_t header_event_id; /* HEADER_EVENT_ID */
  70    uint64_t header_magic;    /* HEADER_MAGIC    */
  71    uint64_t header_version;  /* HEADER_VERSION  */
  72} TraceLogHeader;
  73
  74
  75static void read_from_buffer(unsigned int idx, void *dataptr, size_t size);
  76static unsigned int write_to_buffer(unsigned int idx, void *dataptr, size_t size);
  77
  78static void clear_buffer_range(unsigned int idx, size_t len)
  79{
  80    uint32_t num = 0;
  81    while (num < len) {
  82        if (idx >= TRACE_BUF_LEN) {
  83            idx = idx % TRACE_BUF_LEN;
  84        }
  85        trace_buf[idx++] = 0;
  86        num++;
  87    }
  88}
  89/**
  90 * Read a trace record from the trace buffer
  91 *
  92 * @idx         Trace buffer index
  93 * @record      Trace record to fill
  94 *
  95 * Returns false if the record is not valid.
  96 */
  97static bool get_trace_record(unsigned int idx, TraceRecord **recordptr)
  98{
  99    uint64_t event_flag = 0;
 100    TraceRecord record;
 101    /* read the event flag to see if its a valid record */
 102    read_from_buffer(idx, &record, sizeof(event_flag));
 103
 104    if (!(record.event & TRACE_RECORD_VALID)) {
 105        return false;
 106    }
 107
 108    smp_rmb(); /* read memory barrier before accessing record */
 109    /* read the record header to know record length */
 110    read_from_buffer(idx, &record, sizeof(TraceRecord));
 111    *recordptr = malloc(record.length); /* don't use g_malloc, can deadlock when traced */
 112    /* make a copy of record to avoid being overwritten */
 113    read_from_buffer(idx, *recordptr, record.length);
 114    smp_rmb(); /* memory barrier before clearing valid flag */
 115    (*recordptr)->event &= ~TRACE_RECORD_VALID;
 116    /* clear the trace buffer range for consumed record otherwise any byte
 117     * with its MSB set may be considered as a valid event id when the writer
 118     * thread crosses this range of buffer again.
 119     */
 120    clear_buffer_range(idx, record.length);
 121    return true;
 122}
 123
 124/**
 125 * Kick writeout thread
 126 *
 127 * @wait        Whether to wait for writeout thread to complete
 128 */
 129static void flush_trace_file(bool wait)
 130{
 131    g_mutex_lock(&trace_lock);
 132    trace_available = true;
 133    g_cond_signal(&trace_available_cond);
 134
 135    if (wait) {
 136        g_cond_wait(&trace_empty_cond, &trace_lock);
 137    }
 138
 139    g_mutex_unlock(&trace_lock);
 140}
 141
 142static void wait_for_trace_records_available(void)
 143{
 144    g_mutex_lock(&trace_lock);
 145    while (!(trace_available && trace_writeout_enabled)) {
 146        g_cond_signal(&trace_empty_cond);
 147        g_cond_wait(&trace_available_cond, &trace_lock);
 148    }
 149    trace_available = false;
 150    g_mutex_unlock(&trace_lock);
 151}
 152
 153static gpointer writeout_thread(gpointer opaque)
 154{
 155    TraceRecord *recordptr;
 156    union {
 157        TraceRecord rec;
 158        uint8_t bytes[sizeof(TraceRecord) + sizeof(uint64_t)];
 159    } dropped;
 160    unsigned int idx = 0;
 161    int dropped_count;
 162    size_t unused __attribute__ ((unused));
 163
 164    for (;;) {
 165        wait_for_trace_records_available();
 166
 167        if (g_atomic_int_get(&dropped_events)) {
 168            dropped.rec.event = DROPPED_EVENT_ID,
 169            dropped.rec.timestamp_ns = get_clock();
 170            dropped.rec.length = sizeof(TraceRecord) + sizeof(uint64_t),
 171            dropped.rec.pid = trace_pid;
 172            do {
 173                dropped_count = g_atomic_int_get(&dropped_events);
 174            } while (!g_atomic_int_compare_and_exchange(&dropped_events,
 175                                                        dropped_count, 0));
 176            dropped.rec.arguments[0] = dropped_count;
 177            unused = fwrite(&dropped.rec, dropped.rec.length, 1, trace_fp);
 178        }
 179
 180        while (get_trace_record(idx, &recordptr)) {
 181            unused = fwrite(recordptr, recordptr->length, 1, trace_fp);
 182            writeout_idx += recordptr->length;
 183            free(recordptr); /* don't use g_free, can deadlock when traced */
 184            idx = writeout_idx % TRACE_BUF_LEN;
 185        }
 186
 187        fflush(trace_fp);
 188    }
 189    return NULL;
 190}
 191
 192void trace_record_write_u64(TraceBufferRecord *rec, uint64_t val)
 193{
 194    rec->rec_off = write_to_buffer(rec->rec_off, &val, sizeof(uint64_t));
 195}
 196
 197void trace_record_write_str(TraceBufferRecord *rec, const char *s, uint32_t slen)
 198{
 199    /* Write string length first */
 200    rec->rec_off = write_to_buffer(rec->rec_off, &slen, sizeof(slen));
 201    /* Write actual string now */
 202    rec->rec_off = write_to_buffer(rec->rec_off, (void*)s, slen);
 203}
 204
 205int trace_record_start(TraceBufferRecord *rec, TraceEventID event, size_t datasize)
 206{
 207    unsigned int idx, rec_off, old_idx, new_idx;
 208    uint32_t rec_len = sizeof(TraceRecord) + datasize;
 209    uint64_t event_u64 = event;
 210    uint64_t timestamp_ns = get_clock();
 211
 212    do {
 213        old_idx = g_atomic_int_get(&trace_idx);
 214        smp_rmb();
 215        new_idx = old_idx + rec_len;
 216
 217        if (new_idx - writeout_idx > TRACE_BUF_LEN) {
 218            /* Trace Buffer Full, Event dropped ! */
 219            g_atomic_int_inc(&dropped_events);
 220            return -ENOSPC;
 221        }
 222    } while (!g_atomic_int_compare_and_exchange(&trace_idx, old_idx, new_idx));
 223
 224    idx = old_idx % TRACE_BUF_LEN;
 225
 226    rec_off = idx;
 227    rec_off = write_to_buffer(rec_off, &event_u64, sizeof(event_u64));
 228    rec_off = write_to_buffer(rec_off, &timestamp_ns, sizeof(timestamp_ns));
 229    rec_off = write_to_buffer(rec_off, &rec_len, sizeof(rec_len));
 230    rec_off = write_to_buffer(rec_off, &trace_pid, sizeof(trace_pid));
 231
 232    rec->tbuf_idx = idx;
 233    rec->rec_off  = (idx + sizeof(TraceRecord)) % TRACE_BUF_LEN;
 234    return 0;
 235}
 236
 237static void read_from_buffer(unsigned int idx, void *dataptr, size_t size)
 238{
 239    uint8_t *data_ptr = dataptr;
 240    uint32_t x = 0;
 241    while (x < size) {
 242        if (idx >= TRACE_BUF_LEN) {
 243            idx = idx % TRACE_BUF_LEN;
 244        }
 245        data_ptr[x++] = trace_buf[idx++];
 246    }
 247}
 248
 249static unsigned int write_to_buffer(unsigned int idx, void *dataptr, size_t size)
 250{
 251    uint8_t *data_ptr = dataptr;
 252    uint32_t x = 0;
 253    while (x < size) {
 254        if (idx >= TRACE_BUF_LEN) {
 255            idx = idx % TRACE_BUF_LEN;
 256        }
 257        trace_buf[idx++] = data_ptr[x++];
 258    }
 259    return idx; /* most callers wants to know where to write next */
 260}
 261
 262void trace_record_finish(TraceBufferRecord *rec)
 263{
 264    TraceRecord record;
 265    read_from_buffer(rec->tbuf_idx, &record, sizeof(TraceRecord));
 266    smp_wmb(); /* write barrier before marking as valid */
 267    record.event |= TRACE_RECORD_VALID;
 268    write_to_buffer(rec->tbuf_idx, &record, sizeof(TraceRecord));
 269
 270    if (((unsigned int)g_atomic_int_get(&trace_idx) - writeout_idx)
 271        > TRACE_BUF_FLUSH_THRESHOLD) {
 272        flush_trace_file(false);
 273    }
 274}
 275
 276void st_set_trace_file_enabled(bool enable)
 277{
 278    if (enable == !!trace_fp) {
 279        return; /* no change */
 280    }
 281
 282    /* Halt trace writeout */
 283    flush_trace_file(true);
 284    trace_writeout_enabled = false;
 285    flush_trace_file(true);
 286
 287    if (enable) {
 288        static const TraceLogHeader header = {
 289            .header_event_id = HEADER_EVENT_ID,
 290            .header_magic = HEADER_MAGIC,
 291            /* Older log readers will check for version at next location */
 292            .header_version = HEADER_VERSION,
 293        };
 294
 295        trace_fp = fopen(trace_file_name, "wb");
 296        if (!trace_fp) {
 297            return;
 298        }
 299
 300        if (fwrite(&header, sizeof header, 1, trace_fp) != 1) {
 301            fclose(trace_fp);
 302            trace_fp = NULL;
 303            return;
 304        }
 305
 306        /* Resume trace writeout */
 307        trace_writeout_enabled = true;
 308        flush_trace_file(false);
 309    } else {
 310        fclose(trace_fp);
 311        trace_fp = NULL;
 312    }
 313}
 314
 315/**
 316 * Set the name of a trace file
 317 *
 318 * @file        The trace file name or NULL for the default name-<pid> set at
 319 *              config time
 320 */
 321void st_set_trace_file(const char *file)
 322{
 323    st_set_trace_file_enabled(false);
 324
 325    g_free(trace_file_name);
 326
 327    if (!file) {
 328        /* Type cast needed for Windows where getpid() returns an int. */
 329        trace_file_name = g_strdup_printf(CONFIG_TRACE_FILE, (pid_t)getpid());
 330    } else {
 331        trace_file_name = g_strdup_printf("%s", file);
 332    }
 333
 334    st_set_trace_file_enabled(true);
 335}
 336
 337void st_print_trace_file_status(FILE *stream, int (*stream_printf)(FILE *stream, const char *fmt, ...))
 338{
 339    stream_printf(stream, "Trace file \"%s\" %s.\n",
 340                  trace_file_name, trace_fp ? "on" : "off");
 341}
 342
 343void st_flush_trace_buffer(void)
 344{
 345    flush_trace_file(true);
 346}
 347
 348/* Helper function to create a thread with signals blocked.  Use glib's
 349 * portable threads since QEMU abstractions cannot be used due to reentrancy in
 350 * the tracer.  Also note the signal masking on POSIX hosts so that the thread
 351 * does not steal signals when the rest of the program wants them blocked.
 352 */
 353static GThread *trace_thread_create(GThreadFunc fn)
 354{
 355    GThread *thread;
 356#ifndef _WIN32
 357    sigset_t set, oldset;
 358
 359    sigfillset(&set);
 360    pthread_sigmask(SIG_SETMASK, &set, &oldset);
 361#endif
 362
 363    thread = g_thread_new("trace-thread", fn, NULL);
 364
 365#ifndef _WIN32
 366    pthread_sigmask(SIG_SETMASK, &oldset, NULL);
 367#endif
 368
 369    return thread;
 370}
 371
 372bool st_init(void)
 373{
 374    GThread *thread;
 375
 376    trace_pid = getpid();
 377
 378    thread = trace_thread_create(writeout_thread);
 379    if (!thread) {
 380        fprintf(stderr, "warning: unable to initialize simple trace backend\n");
 381        return false;
 382    }
 383
 384    atexit(st_flush_trace_buffer);
 385    return true;
 386}
 387