Statistics
| Branch: | Revision:

root / trace / simple.c @ b618c288

History | View | Annotate | Download (12 kB)

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 <stdlib.h>
12
#include <stdint.h>
13
#include <stdio.h>
14
#include <time.h>
15
#ifndef _WIN32
16
#include <signal.h>
17
#include <pthread.h>
18
#endif
19
#include "qemu/timer.h"
20
#include "trace.h"
21
#include "trace/control.h"
22
#include "trace/simple.h"
23

    
24
/** Trace file header event ID */
25
#define HEADER_EVENT_ID (~(uint64_t)0) /* avoids conflicting with TraceEventIDs */
26

    
27
/** Trace file magic number */
28
#define HEADER_MAGIC 0xf2b177cb0aa429b4ULL
29

    
30
/** Trace file version number, bump if format changes */
31
#define HEADER_VERSION 2
32

    
33
/** Records were dropped event ID */
34
#define DROPPED_EVENT_ID (~(uint64_t)0 - 1)
35

    
36
/** Trace record is valid */
37
#define TRACE_RECORD_VALID ((uint64_t)1 << 63)
38

    
39
/*
40
 * Trace records are written out by a dedicated thread.  The thread waits for
41
 * records to become available, writes them out, and then waits again.
42
 */
43
static GStaticMutex trace_lock = G_STATIC_MUTEX_INIT;
44

    
45
/* g_cond_new() was deprecated in glib 2.31 but we still need to support it */
46
#if GLIB_CHECK_VERSION(2, 31, 0)
47
static GCond the_trace_available_cond;
48
static GCond the_trace_empty_cond;
49
static GCond *trace_available_cond = &the_trace_available_cond;
50
static GCond *trace_empty_cond = &the_trace_empty_cond;
51
#else
52
static GCond *trace_available_cond;
53
static GCond *trace_empty_cond;
54
#endif
55

    
56
static bool trace_available;
57
static bool trace_writeout_enabled;
58

    
59
enum {
60
    TRACE_BUF_LEN = 4096 * 64,
61
    TRACE_BUF_FLUSH_THRESHOLD = TRACE_BUF_LEN / 4,
62
};
63

    
64
uint8_t trace_buf[TRACE_BUF_LEN];
65
static volatile gint trace_idx;
66
static unsigned int writeout_idx;
67
static volatile gint dropped_events;
68
static FILE *trace_fp;
69
static char *trace_file_name;
70

    
71
/* * Trace buffer entry */
72
typedef struct {
73
    uint64_t event; /*   TraceEventID */
74
    uint64_t timestamp_ns;
75
    uint32_t length;   /*    in bytes */
76
    uint32_t reserved; /*    unused */
77
    uint64_t arguments[];
78
} TraceRecord;
79

    
80
typedef struct {
81
    uint64_t header_event_id; /* HEADER_EVENT_ID */
82
    uint64_t header_magic;    /* HEADER_MAGIC    */
83
    uint64_t header_version;  /* HEADER_VERSION  */
84
} TraceLogHeader;
85

    
86

    
87
static void read_from_buffer(unsigned int idx, void *dataptr, size_t size);
88
static unsigned int write_to_buffer(unsigned int idx, void *dataptr, size_t size);
89

    
90
static void clear_buffer_range(unsigned int idx, size_t len)
91
{
92
    uint32_t num = 0;
93
    while (num < len) {
94
        if (idx >= TRACE_BUF_LEN) {
95
            idx = idx % TRACE_BUF_LEN;
96
        }
97
        trace_buf[idx++] = 0;
98
        num++;
99
    }
100
}
101
/**
102
 * Read a trace record from the trace buffer
103
 *
104
 * @idx         Trace buffer index
105
 * @record      Trace record to fill
106
 *
107
 * Returns false if the record is not valid.
108
 */
109
static bool get_trace_record(unsigned int idx, TraceRecord **recordptr)
110
{
111
    uint64_t event_flag = 0;
112
    TraceRecord record;
113
    /* read the event flag to see if its a valid record */
114
    read_from_buffer(idx, &record, sizeof(event_flag));
115

    
116
    if (!(record.event & TRACE_RECORD_VALID)) {
117
        return false;
118
    }
119

    
120
    smp_rmb(); /* read memory barrier before accessing record */
121
    /* read the record header to know record length */
122
    read_from_buffer(idx, &record, sizeof(TraceRecord));
123
    *recordptr = malloc(record.length); /* dont use g_malloc, can deadlock when traced */
124
    /* make a copy of record to avoid being overwritten */
125
    read_from_buffer(idx, *recordptr, record.length);
126
    smp_rmb(); /* memory barrier before clearing valid flag */
127
    (*recordptr)->event &= ~TRACE_RECORD_VALID;
128
    /* clear the trace buffer range for consumed record otherwise any byte
129
     * with its MSB set may be considered as a valid event id when the writer
130
     * thread crosses this range of buffer again.
131
     */
132
    clear_buffer_range(idx, record.length);
133
    return true;
134
}
135

    
136
/**
137
 * Kick writeout thread
138
 *
139
 * @wait        Whether to wait for writeout thread to complete
140
 */
141
static void flush_trace_file(bool wait)
142
{
143
    g_static_mutex_lock(&trace_lock);
144
    trace_available = true;
145
    g_cond_signal(trace_available_cond);
146

    
147
    if (wait) {
148
        g_cond_wait(trace_empty_cond, g_static_mutex_get_mutex(&trace_lock));
149
    }
150

    
151
    g_static_mutex_unlock(&trace_lock);
152
}
153

    
154
static void wait_for_trace_records_available(void)
155
{
156
    g_static_mutex_lock(&trace_lock);
157
    while (!(trace_available && trace_writeout_enabled)) {
158
        g_cond_signal(trace_empty_cond);
159
        g_cond_wait(trace_available_cond,
160
                    g_static_mutex_get_mutex(&trace_lock));
161
    }
162
    trace_available = false;
163
    g_static_mutex_unlock(&trace_lock);
164
}
165

    
166
static gpointer writeout_thread(gpointer opaque)
167
{
168
    TraceRecord *recordptr;
169
    union {
170
        TraceRecord rec;
171
        uint8_t bytes[sizeof(TraceRecord) + sizeof(uint64_t)];
172
    } dropped;
173
    unsigned int idx = 0;
174
    int dropped_count;
175
    size_t unused __attribute__ ((unused));
176

    
177
    for (;;) {
178
        wait_for_trace_records_available();
179

    
180
        if (g_atomic_int_get(&dropped_events)) {
181
            dropped.rec.event = DROPPED_EVENT_ID,
182
            dropped.rec.timestamp_ns = get_clock();
183
            dropped.rec.length = sizeof(TraceRecord) + sizeof(uint64_t),
184
            dropped.rec.reserved = 0;
185
            do {
186
                dropped_count = g_atomic_int_get(&dropped_events);
187
            } while (!g_atomic_int_compare_and_exchange(&dropped_events,
188
                                                        dropped_count, 0));
189
            dropped.rec.arguments[0] = dropped_count;
190
            unused = fwrite(&dropped.rec, dropped.rec.length, 1, trace_fp);
191
        }
192

    
193
        while (get_trace_record(idx, &recordptr)) {
194
            unused = fwrite(recordptr, recordptr->length, 1, trace_fp);
195
            writeout_idx += recordptr->length;
196
            free(recordptr); /* dont use g_free, can deadlock when traced */
197
            idx = writeout_idx % TRACE_BUF_LEN;
198
        }
199

    
200
        fflush(trace_fp);
201
    }
202
    return NULL;
203
}
204

    
205
void trace_record_write_u64(TraceBufferRecord *rec, uint64_t val)
206
{
207
    rec->rec_off = write_to_buffer(rec->rec_off, &val, sizeof(uint64_t));
208
}
209

    
210
void trace_record_write_str(TraceBufferRecord *rec, const char *s, uint32_t slen)
211
{
212
    /* Write string length first */
213
    rec->rec_off = write_to_buffer(rec->rec_off, &slen, sizeof(slen));
214
    /* Write actual string now */
215
    rec->rec_off = write_to_buffer(rec->rec_off, (void*)s, slen);
216
}
217

    
218
int trace_record_start(TraceBufferRecord *rec, TraceEventID event, size_t datasize)
219
{
220
    unsigned int idx, rec_off, old_idx, new_idx;
221
    uint32_t rec_len = sizeof(TraceRecord) + datasize;
222
    uint64_t event_u64 = event;
223
    uint64_t timestamp_ns = get_clock();
224

    
225
    do {
226
        old_idx = g_atomic_int_get(&trace_idx);
227
        smp_rmb();
228
        new_idx = old_idx + rec_len;
229

    
230
        if (new_idx - writeout_idx > TRACE_BUF_LEN) {
231
            /* Trace Buffer Full, Event dropped ! */
232
            g_atomic_int_inc(&dropped_events);
233
            return -ENOSPC;
234
        }
235
    } while (!g_atomic_int_compare_and_exchange(&trace_idx, old_idx, new_idx));
236

    
237
    idx = old_idx % TRACE_BUF_LEN;
238

    
239
    rec_off = idx;
240
    rec_off = write_to_buffer(rec_off, &event_u64, sizeof(event_u64));
241
    rec_off = write_to_buffer(rec_off, &timestamp_ns, sizeof(timestamp_ns));
242
    rec_off = write_to_buffer(rec_off, &rec_len, sizeof(rec_len));
243

    
244
    rec->tbuf_idx = idx;
245
    rec->rec_off  = (idx + sizeof(TraceRecord)) % TRACE_BUF_LEN;
246
    return 0;
247
}
248

    
249
static void read_from_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
        data_ptr[x++] = trace_buf[idx++];
258
    }
259
}
260

    
261
static unsigned int write_to_buffer(unsigned int idx, void *dataptr, size_t size)
262
{
263
    uint8_t *data_ptr = dataptr;
264
    uint32_t x = 0;
265
    while (x < size) {
266
        if (idx >= TRACE_BUF_LEN) {
267
            idx = idx % TRACE_BUF_LEN;
268
        }
269
        trace_buf[idx++] = data_ptr[x++];
270
    }
271
    return idx; /* most callers wants to know where to write next */
272
}
273

    
274
void trace_record_finish(TraceBufferRecord *rec)
275
{
276
    TraceRecord record;
277
    read_from_buffer(rec->tbuf_idx, &record, sizeof(TraceRecord));
278
    smp_wmb(); /* write barrier before marking as valid */
279
    record.event |= TRACE_RECORD_VALID;
280
    write_to_buffer(rec->tbuf_idx, &record, sizeof(TraceRecord));
281

    
282
    if (((unsigned int)g_atomic_int_get(&trace_idx) - writeout_idx)
283
        > TRACE_BUF_FLUSH_THRESHOLD) {
284
        flush_trace_file(false);
285
    }
286
}
287

    
288
void st_set_trace_file_enabled(bool enable)
289
{
290
    if (enable == !!trace_fp) {
291
        return; /* no change */
292
    }
293

    
294
    /* Halt trace writeout */
295
    flush_trace_file(true);
296
    trace_writeout_enabled = false;
297
    flush_trace_file(true);
298

    
299
    if (enable) {
300
        static const TraceLogHeader header = {
301
            .header_event_id = HEADER_EVENT_ID,
302
            .header_magic = HEADER_MAGIC,
303
            /* Older log readers will check for version at next location */
304
            .header_version = HEADER_VERSION,
305
        };
306

    
307
        trace_fp = fopen(trace_file_name, "wb");
308
        if (!trace_fp) {
309
            return;
310
        }
311

    
312
        if (fwrite(&header, sizeof header, 1, trace_fp) != 1) {
313
            fclose(trace_fp);
314
            trace_fp = NULL;
315
            return;
316
        }
317

    
318
        /* Resume trace writeout */
319
        trace_writeout_enabled = true;
320
        flush_trace_file(false);
321
    } else {
322
        fclose(trace_fp);
323
        trace_fp = NULL;
324
    }
325
}
326

    
327
/**
328
 * Set the name of a trace file
329
 *
330
 * @file        The trace file name or NULL for the default name-<pid> set at
331
 *              config time
332
 */
333
bool st_set_trace_file(const char *file)
334
{
335
    st_set_trace_file_enabled(false);
336

    
337
    g_free(trace_file_name);
338

    
339
    if (!file) {
340
        trace_file_name = g_strdup_printf(CONFIG_TRACE_FILE, getpid());
341
    } else {
342
        trace_file_name = g_strdup_printf("%s", file);
343
    }
344

    
345
    st_set_trace_file_enabled(true);
346
    return true;
347
}
348

    
349
void st_print_trace_file_status(FILE *stream, int (*stream_printf)(FILE *stream, const char *fmt, ...))
350
{
351
    stream_printf(stream, "Trace file \"%s\" %s.\n",
352
                  trace_file_name, trace_fp ? "on" : "off");
353
}
354

    
355
void st_flush_trace_buffer(void)
356
{
357
    flush_trace_file(true);
358
}
359

    
360
void trace_print_events(FILE *stream, fprintf_function stream_printf)
361
{
362
    unsigned int i;
363

    
364
    for (i = 0; i < trace_event_count(); i++) {
365
        TraceEvent *ev = trace_event_id(i);
366
        stream_printf(stream, "%s [Event ID %u] : state %u\n",
367
                      trace_event_get_name(ev), i, trace_event_get_state_dynamic(ev));
368
    }
369
}
370

    
371
void trace_event_set_state_dynamic_backend(TraceEvent *ev, bool state)
372
{
373
    ev->dstate = state;
374
}
375

    
376
/* Helper function to create a thread with signals blocked.  Use glib's
377
 * portable threads since QEMU abstractions cannot be used due to reentrancy in
378
 * the tracer.  Also note the signal masking on POSIX hosts so that the thread
379
 * does not steal signals when the rest of the program wants them blocked.
380
 */
381
static GThread *trace_thread_create(GThreadFunc fn)
382
{
383
    GThread *thread;
384
#ifndef _WIN32
385
    sigset_t set, oldset;
386

    
387
    sigfillset(&set);
388
    pthread_sigmask(SIG_SETMASK, &set, &oldset);
389
#endif
390

    
391
#if GLIB_CHECK_VERSION(2, 31, 0)
392
    thread = g_thread_new("trace-thread", fn, NULL);
393
#else
394
    thread = g_thread_create(fn, NULL, FALSE, NULL);
395
#endif
396

    
397
#ifndef _WIN32
398
    pthread_sigmask(SIG_SETMASK, &oldset, NULL);
399
#endif
400

    
401
    return thread;
402
}
403

    
404
bool trace_backend_init(const char *events, const char *file)
405
{
406
    GThread *thread;
407

    
408
    if (!g_thread_supported()) {
409
#if !GLIB_CHECK_VERSION(2, 31, 0)
410
        g_thread_init(NULL);
411
#else
412
        fprintf(stderr, "glib threading failed to initialize.\n");
413
        exit(1);
414
#endif
415
    }
416

    
417
#if !GLIB_CHECK_VERSION(2, 31, 0)
418
    trace_available_cond = g_cond_new();
419
    trace_empty_cond = g_cond_new();
420
#endif
421

    
422
    thread = trace_thread_create(writeout_thread);
423
    if (!thread) {
424
        fprintf(stderr, "warning: unable to initialize simple trace backend\n");
425
        return false;
426
    }
427

    
428
    atexit(st_flush_trace_buffer);
429
    trace_backend_init_events(events);
430
    st_set_trace_file(file);
431
    return true;
432
}