darshan-posix.c 45.8 KB
Newer Older
1
/*
Shane Snyder's avatar
Shane Snyder committed
2
3
4
 * Copyright (C) 2015 University of Chicago.
 * See COPYRIGHT notice in top-level directory.
 *
5
6
 */

7
#include "darshan-runtime-config.h"
8
9
10
11
12
13
14
15
16
17
18
19
20
21
#include <stdio.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <stdarg.h>
#include <string.h>
#include <time.h>
#include <stdlib.h>
#include <errno.h>
#include <sys/uio.h>
#include <sys/mman.h>
#include <search.h>
#include <assert.h>
22
#include <libgen.h>
Philip Carns's avatar
Philip Carns committed
23
#include <aio.h>
24
#define __USE_GNU
25
#include <pthread.h>
26

27
#include "uthash.h"
28
#include "darshan.h"
29
#include "darshan-posix-log-format.h"
30

31
#ifndef HAVE_OFF64_T
32
33
typedef int64_t off64_t;
#endif
34
35
36
#ifndef HAVE_AIOCB64
#define aiocb64 aiocb
#endif
37

38
39
/* TODO: more libc, fgetc, etc etc etc. */

40
41
DARSHAN_FORWARD_DECL(open, int, (const char *path, int flags, ...));
DARSHAN_FORWARD_DECL(open64, int, (const char *path, int flags, ...));
42
43
DARSHAN_FORWARD_DECL(creat, int, (const char* path, mode_t mode));
DARSHAN_FORWARD_DECL(creat64, int, (const char* path, mode_t mode));
44
45
DARSHAN_FORWARD_DECL(fopen, FILE*, (const char *path, const char *mode));
DARSHAN_FORWARD_DECL(fopen64, FILE*, (const char *path, const char *mode));
46
47
48
49
50
51
52
53
DARSHAN_FORWARD_DECL(read, ssize_t, (int fd, void *buf, size_t count));
DARSHAN_FORWARD_DECL(write, ssize_t, (int fd, const void *buf, size_t count));
DARSHAN_FORWARD_DECL(pread, ssize_t, (int fd, void *buf, size_t count, off_t offset));
DARSHAN_FORWARD_DECL(pwrite, ssize_t, (int fd, const void *buf, size_t count, off_t offset));
DARSHAN_FORWARD_DECL(pread64, ssize_t, (int fd, void *buf, size_t count, off64_t offset));
DARSHAN_FORWARD_DECL(pwrite64, ssize_t, (int fd, const void *buf, size_t count, off64_t offset));
DARSHAN_FORWARD_DECL(readv, ssize_t, (int fd, const struct iovec *iov, int iovcnt));
DARSHAN_FORWARD_DECL(writev, ssize_t, (int fd, const struct iovec *iov, int iovcnt));
54
55
56
57
58
DARSHAN_FORWARD_DECL(fread, size_t, (void *ptr, size_t size, size_t nmemb, FILE *stream));
DARSHAN_FORWARD_DECL(fwrite, size_t, (const void *ptr, size_t size, size_t nmemb, FILE *stream));
DARSHAN_FORWARD_DECL(lseek, off_t, (int fd, off_t offset, int whence));
DARSHAN_FORWARD_DECL(lseek64, off64_t, (int fd, off64_t offset, int whence));
DARSHAN_FORWARD_DECL(fseek, int, (FILE *stream, long offset, int whence));
59
60
61
62
63
64
DARSHAN_FORWARD_DECL(__xstat, int, (int vers, const char* path, struct stat *buf));
DARSHAN_FORWARD_DECL(__xstat64, int, (int vers, const char* path, struct stat64 *buf));
DARSHAN_FORWARD_DECL(__lxstat, int, (int vers, const char* path, struct stat *buf));
DARSHAN_FORWARD_DECL(__lxstat64, int, (int vers, const char* path, struct stat64 *buf));
DARSHAN_FORWARD_DECL(__fxstat, int, (int vers, int fd, struct stat *buf));
DARSHAN_FORWARD_DECL(__fxstat64, int, (int vers, int fd, struct stat64 *buf));
Shane Snyder's avatar
Shane Snyder committed
65
66
DARSHAN_FORWARD_DECL(mmap, void*, (void *addr, size_t length, int prot, int flags, int fd, off_t offset));
DARSHAN_FORWARD_DECL(mmap64, void*, (void *addr, size_t length, int prot, int flags, int fd, off64_t offset));
67
68
DARSHAN_FORWARD_DECL(fsync, int, (int fd));
DARSHAN_FORWARD_DECL(fdatasync, int, (int fd));
69
DARSHAN_FORWARD_DECL(close, int, (int fd));
70
DARSHAN_FORWARD_DECL(fclose, int, (FILE *fp));
71
72
73
/* TODO mkstemp */
/* TODO aio */
/* TODO listio */
74

75
76
77
78
79
80
81
82
83
84
85
86
87
88
static void posix_runtime_initialize(void);
static struct posix_file_runtime* posix_file_by_name(const char *name);
static struct posix_file_runtime* posix_file_by_name_setfd(const char* name, int fd);
static struct posix_file_runtime* posix_file_by_fd(int fd);
static void posix_file_close_fd(int fd);

static void posix_disable_instrumentation(void);
static void posix_prepare_for_reduction(darshan_record_id *shared_recs,
    int *shared_rec_count, void **send_buf, void **recv_buf, int *rec_size);
static void posix_record_reduction_op(void* infile_v, void* inoutfile_v,
    int *len, MPI_Datatype *datatype);
static void posix_get_output_data(void **buffer, int *size);
static void posix_shutdown(void);

89
90
91
92
93
94
95
96
97
98
99
100
enum posix_io_type
{
    POSIX_READ = 1,
    POSIX_WRITE = 2,
};

/* The posix_file_runtime structure maintains necessary runtime metadata
 * for the POSIX file record (darshan_posix_file structure, defined in
 * darshan-posix-log-format.h) pointed to by 'file_record'. This metadata
 * assists with the instrumenting of specific statistics in the file record.
 * 'hlink' is a hash table link structure used to add/remove this record
 * from the hash table of POSIX file records for this process. 
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
 *
 * RATIONALE: the POSIX module needs to track some stateful, volatile 
 * information about each open file (like the current file offset, most recent 
 * access time, etc.) to aid in instrumentation, but this information can't be
 * stored in the darshan_posix_file struct because we don't want it to appear in
 * the final darshan log file.  We therefore associate a posix_file_runtime
 * struct with each darshan_posix_file struct in order to track this information.
  *
 * NOTE: There is a one-to-one mapping of posix_file_runtime structs to
 * darshan_posix_file structs.
 *
 * NOTE: The posix_file_runtime struct contains a pointer to a darshan_posix_file
 * struct (see the *file_record member) rather than simply embedding an entire
 * darshan_posix_file struct.  This is done so that all of the darshan_posix_file
 * structs can be kept contiguous in memory as a single array to simplify
 * reduction, compression, and storage.
117
 */
118
struct posix_file_runtime
119
{
120
    struct darshan_posix_file* file_record;
121
122
123
124
    int64_t offset;
    int64_t last_byte_read;
    int64_t last_byte_written;
    enum posix_io_type last_io_type;
125
126
127
    double last_meta_end;
    double last_read_end;
    double last_write_end;
128
    UT_hash_handle hlink;
129
};
130

131
132
133
134
135
136
137
138
139
140
141
/* The posix_file_runtime_ref structure is used to associate a POSIX
 * file descriptor with an already existing POSIX file record. This is
 * necessary as many POSIX I/O functions take only an input file descriptor,
 * but POSIX file records are indexed by their full file paths (i.e., darshan
 * record identifiers for POSIX files are created by hashing the file path).
 * In other words, this structure is necessary as it allows us to look up a
 * file record either by a pathname (posix_file_runtime) or by POSIX file
 * descriptor (posix_file_runtime_ref), depending on which parameters are
 * available. This structure includes another hash table link, since separate
 * hashes are maintained for posix_file_runtime structures and posix_file_runtime_ref
 * structures.
142
143
144
145
146
147
148
149
150
151
152
 *
 * RATIONALE: In theory the fd information could be included in the
 * posix_file_runtime struct rather than in a separate structure here.  The
 * reason we don't do that is because the same file could be opened multiple
 * times by a given process with different file descriptors and thus
 * simulataneously referenced using different file descriptors.  This practice is
 * not common, but we must support it.
 *
 * NOTE: there are potentially multiple posix_file_runtime_ref structures
 * referring to a single posix_file_runtime structure.  Most of the time there is
 * only one, however.
153
 */
154
struct posix_file_runtime_ref
155
{
156
    struct posix_file_runtime* file;
157
158
159
160
    int fd;
    UT_hash_handle hlink;
};

161
162
163
164
/* The posix_runtime structure maintains necessary state for storing
 * POSIX file records and for coordinating with darshan-core at 
 * shutdown time.
 */
165
166
struct posix_runtime
{
167
    struct posix_file_runtime* file_runtime_array;
168
    struct darshan_posix_file* file_record_array;
169
    int file_array_size;
170
    int file_array_ndx;
171
172
    struct posix_file_runtime* file_hash;
    struct posix_file_runtime_ref* fd_hash;
173
174
    void *red_buf;
    int shared_rec_count;
175
176
};

177
static struct posix_runtime *posix_runtime = NULL;
178
static pthread_mutex_t posix_runtime_mutex = PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP;
Shane Snyder's avatar
Shane Snyder committed
179
static int instrumentation_disabled = 0;
180
static int my_rank = -1;
181
182
183

#define POSIX_LOCK() pthread_mutex_lock(&posix_runtime_mutex)
#define POSIX_UNLOCK() pthread_mutex_unlock(&posix_runtime_mutex)
184

185
#define POSIX_RECORD_OPEN(__ret, __path, __mode, __stream_flag, __tm1, __tm2) do { \
186
    struct posix_file_runtime* file; \
187
188
189
    char* exclude; \
    int tmp_index = 0; \
    if(__ret < 0) break; \
190
    while((exclude = darshan_path_exclusions[tmp_index])) { \
191
192
193
194
195
        if(!(strncmp(exclude, __path, strlen(exclude)))) \
            break; \
        tmp_index++; \
    } \
    if(exclude) break; \
196
    file = posix_file_by_name_setfd(__path, __ret); \
197
    if(!file) break; \
198
    /* CP_STAT_FILE(file, __path, __ret); */ \
199
    file->file_record->rank = my_rank; \
200
    if(__mode) \
201
202
203
204
        DARSHAN_COUNTER_SET(file->file_record, POSIX_MODE, __mode); \
    file->offset = 0; \
    file->last_byte_written = 0; \
    file->last_byte_read = 0; \
205
    if(__stream_flag)\
206
        DARSHAN_COUNTER_INC(file->file_record, POSIX_FOPENS, 1); \
207
    else \
208
209
210
211
        DARSHAN_COUNTER_INC(file->file_record, POSIX_OPENS, 1); \
    if(DARSHAN_COUNTER_F_VALUE(file->file_record, POSIX_F_OPEN_TIMESTAMP) == 0) \
        DARSHAN_COUNTER_F_SET(file->file_record, POSIX_F_OPEN_TIMESTAMP, __tm1); \
    DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record, __tm1, __tm2, file->last_meta_end, POSIX_F_META_TIME); \
212
213
214
} while(0)

#define POSIX_RECORD_READ(__ret, __fd, __pread_flag, __pread_offset, __aligned, __stream_flag, __tm1, __tm2) do{ \
215
216
    /* size_t stride; */\
    int64_t this_offset; \
217
218
219
220
221
222
    struct posix_file_runtime* file; \
    /* int64_t file_alignment; */ \
    double __elapsed = __tm2-__tm1; \
    if(__ret < 0) break; \
    file = posix_file_by_fd(__fd); \
    if(!file) break; \
223
    if(__pread_flag) \
224
225
226
        this_offset = __pread_offset; \
    else \
        this_offset = file->offset; \
227
    /* file_alignment = CP_VALUE(file, CP_FILE_ALIGNMENT); */\
228
    if(this_offset > file->last_byte_read) \
229
        DARSHAN_COUNTER_INC(file->file_record, POSIX_SEQ_READS, 1); \
230
    if(this_offset == (file->last_byte_read + 1)) \
231
232
        DARSHAN_COUNTER_INC(file->file_record, POSIX_CONSEC_READS, 1); \
    /* if(this_offset > 0 && this_offset > file->last_byte_read \
233
234
235
        && file->last_byte_read != 0) \
        stride = this_offset - file->last_byte_read - 1; \
    else \
236
        stride = 0; */\
237
238
    file->last_byte_read = this_offset + __ret - 1; \
    file->offset = this_offset + __ret; \
239
240
    DARSHAN_COUNTER_MAX(file->file_record, POSIX_MAX_BYTE_READ, (this_offset + __ret - 1)); \
    DARSHAN_COUNTER_INC(file->file_record, POSIX_BYTES_READ, __ret); \
241
    if(__stream_flag)\
242
        DARSHAN_COUNTER_INC(file->file_record, POSIX_FREADS, 1); \
243
    else\
244
        DARSHAN_COUNTER_INC(file->file_record, POSIX_READS, 1); \
245
246
247
248
249
250
    /* CP_BUCKET_INC(file, CP_SIZE_READ_0_100, __ret); \
    cp_access_counter(file, stride, CP_COUNTER_STRIDE); \
    if(!__aligned) \
        CP_INC(file, CP_MEM_NOT_ALIGNED, 1); \
    if(file_alignment > 0 && (this_offset % file_alignment) != 0) \
        CP_INC(file, CP_FILE_NOT_ALIGNED, 1); \
251
252
253
254
255
256
257
258
259
260
261
    cp_access_counter(file, __ret, CP_COUNTER_ACCESS); */\
    if(file->last_io_type == POSIX_WRITE) \
        DARSHAN_COUNTER_INC(file->file_record, POSIX_RW_SWITCHES, 1); \
    file->last_io_type = POSIX_READ; \
    DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record, __tm1, __tm2, file->last_read_end, POSIX_F_READ_TIME); \
    if(DARSHAN_COUNTER_F_VALUE(file->file_record, POSIX_F_READ_START_TIMESTAMP) == 0) \
        DARSHAN_COUNTER_F_SET(file->file_record, POSIX_F_READ_START_TIMESTAMP, __tm1); \
    DARSHAN_COUNTER_F_SET(file->file_record, POSIX_F_READ_END_TIMESTAMP, __tm2); \
    if(DARSHAN_COUNTER_F_VALUE(file->file_record, POSIX_F_MAX_READ_TIME) < __elapsed){ \
        DARSHAN_COUNTER_F_SET(file->file_record, POSIX_F_MAX_READ_TIME, __elapsed); \
        DARSHAN_COUNTER_SET(file->file_record, POSIX_MAX_READ_TIME_SIZE, __ret); } \
262
263
264
} while(0)

#define POSIX_RECORD_WRITE(__ret, __fd, __pwrite_flag, __pwrite_offset, __aligned, __stream_flag, __tm1, __tm2) do{ \
265
266
    /* size_t stride; */\
    int64_t this_offset; \
267
268
269
270
271
272
    struct posix_file_runtime* file; \
    /* int64_t file_alignment; */ \
    double __elapsed = __tm2-__tm1; \
    if(__ret < 0) break; \
    file = posix_file_by_fd(__fd); \
    if(!file) break; \
273
    if(__pwrite_flag) \
274
275
276
        this_offset = __pwrite_offset; \
    else \
        this_offset = file->offset; \
277
    /* file_alignment = CP_VALUE(file, CP_FILE_ALIGNMENT); */\
278
    if(this_offset > file->last_byte_written) \
279
        DARSHAN_COUNTER_INC(file->file_record, POSIX_SEQ_WRITES, 1); \
280
    if(this_offset == (file->last_byte_written + 1)) \
281
282
        DARSHAN_COUNTER_INC(file->file_record, POSIX_CONSEC_WRITES, 1); \
    /* if(this_offset > 0 && this_offset > file->last_byte_written \
283
284
        && file->last_byte_written != 0) \
        stride = this_offset - file->last_byte_written - 1; \
285
    else \
286
        stride = 0; */\
287
288
    file->last_byte_written = this_offset + __ret - 1; \
    file->offset = this_offset + __ret; \
289
290
    DARSHAN_COUNTER_MAX(file->file_record, POSIX_MAX_BYTE_WRITTEN, (this_offset + __ret - 1)); \
    DARSHAN_COUNTER_INC(file->file_record, POSIX_BYTES_WRITTEN, __ret); \
291
    if(__stream_flag) \
292
        DARSHAN_COUNTER_INC(file->file_record, POSIX_FWRITES, 1); \
293
    else \
294
        DARSHAN_COUNTER_INC(file->file_record, POSIX_WRITES, 1); \
295
296
297
298
299
300
    /* CP_BUCKET_INC(file, CP_SIZE_WRITE_0_100, __ret); \
    cp_access_counter(file, stride, CP_COUNTER_STRIDE); \
    if(!__aligned) \
        CP_INC(file, CP_MEM_NOT_ALIGNED, 1); \
    if(file_alignment > 0 && (this_offset % file_alignment) != 0) \
        CP_INC(file, CP_FILE_NOT_ALIGNED, 1); \
301
302
303
304
305
306
307
308
309
310
311
    cp_access_counter(file, __ret, CP_COUNTER_ACCESS); */\
    if(file->last_io_type == POSIX_READ) \
        DARSHAN_COUNTER_INC(file->file_record, POSIX_RW_SWITCHES, 1); \
    file->last_io_type = POSIX_WRITE; \
    DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record, __tm1, __tm2, file->last_write_end, POSIX_F_WRITE_TIME); \
    if(DARSHAN_COUNTER_F_VALUE(file->file_record, POSIX_F_WRITE_START_TIMESTAMP) == 0) \
        DARSHAN_COUNTER_F_SET(file->file_record, POSIX_F_WRITE_START_TIMESTAMP, __tm1); \
    DARSHAN_COUNTER_F_SET(file->file_record, POSIX_F_WRITE_END_TIMESTAMP, __tm2); \
    if(DARSHAN_COUNTER_F_VALUE(file->file_record, POSIX_F_MAX_WRITE_TIME) < __elapsed){ \
        DARSHAN_COUNTER_F_SET(file->file_record, POSIX_F_MAX_WRITE_TIME, __elapsed); \
        DARSHAN_COUNTER_SET(file->file_record, POSIX_MAX_WRITE_TIME_SIZE, __ret); } \
312
} while(0)
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
#define POSIX_LOOKUP_RECORD_STAT(__path, __statbuf, __tm1, __tm2) do { \
    char* exclude; \
    int tmp_index = 0; \
    struct posix_file_runtime* file; \
    while((exclude = darshan_path_exclusions[tmp_index])) { \
        if(!(strncmp(exclude, __path, strlen(exclude)))) \
            break; \
        tmp_index++; \
    } \
    if(exclude) break; \
    file = posix_file_by_name(__path); \
    if(file) \
    { \
        POSIX_RECORD_STAT(file, __statbuf, __tm1, __tm2); \
    } \
} while(0)

#define POSIX_RECORD_STAT(__file, __statbuf, __tm1, __tm2) do { \
    if(!DARSHAN_COUNTER_VALUE((__file)->file_record, POSIX_STATS) && !DARSHAN_COUNTER_VALUE((__file)->file_record, POSIX_OPENS)){ \
        DARSHAN_COUNTER_SET((__file)->file_record, FILE_ALIGNMENT, (__statbuf)->st_blksize); \
        DARSHAN_COUNTER_SET((__file)->file_record, SIZE_AT_OPEN, (__statbuf)->st_size); \
    }\
    (__file)->file_record->rank = my_rank; \
    DARSHAN_COUNTER_F_INC_NO_OVERLAP((__file)->file_record, __tm1, __tm2, (__file)->last_meta_end, POSIX_F_META_TIME); \
    DARSHAN_COUNTER_INC((__file)->file_record, POSIX_STATS, 1); \
} while(0)

341
342
343
344
/**********************************************************
 *      Wrappers for POSIX I/O functions of interest      * 
 **********************************************************/

345
int DARSHAN_DECL(open)(const char *path, int flags, ...)
346
347
348
349
350
{
    int mode = 0;
    int ret;
    double tm1, tm2;

351
352
    MAP_OR_FAIL(open);

353
    if(flags & O_CREAT) 
354
355
356
357
358
359
    {
        va_list arg;
        va_start(arg, flags);
        mode = va_arg(arg, int);
        va_end(arg);

360
        tm1 = darshan_core_wtime();
361
        ret = __real_open(path, flags, mode);
362
        tm2 = darshan_core_wtime();
363
364
365
    }
    else
    {
366
        tm1 = darshan_core_wtime();
367
        ret = __real_open(path, flags);
368
        tm2 = darshan_core_wtime();
369
370
    }

371
372
    POSIX_LOCK();
    posix_runtime_initialize();
373
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
374
    POSIX_UNLOCK();
375
376
377
378

    return(ret);
}

Shane Snyder's avatar
Shane Snyder committed
379
380
381
382
383
384
int DARSHAN_DECL(open64)(const char *path, int flags, ...)
{
    int mode = 0;
    int ret;
    double tm1, tm2;

385
    MAP_OR_FAIL(open64);
Shane Snyder's avatar
Shane Snyder committed
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406

    if(flags & O_CREAT)
    {
        va_list arg;
        va_start(arg, flags);
        mode = va_arg(arg, int);
        va_end(arg);

        tm1 = darshan_core_wtime();
        ret = __real_open64(path, flags, mode);
        tm2 = darshan_core_wtime();
    }
    else
    {
        tm1 = darshan_core_wtime();
        ret = __real_open64(path, flags);
        tm2 = darshan_core_wtime();
    }

    POSIX_LOCK();
    posix_runtime_initialize();
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
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(creat)(const char* path, mode_t mode)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(creat);

    tm1 = darshan_core_wtime();
    ret = __real_creat(path, mode);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(creat64)(const char* path, mode_t mode)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(creat64);
Shane Snyder's avatar
Shane Snyder committed
438

439
440
441
442
443
444
    tm1 = darshan_core_wtime();
    ret = __real_creat64(path, mode);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
Shane Snyder's avatar
Shane Snyder committed
445
446
447
448
449
450
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
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
498
499
500
FILE* DARSHAN_DECL(fopen)(const char *path, const char *mode)
{
    FILE* ret;
    int fd;
    double tm1, tm2;

    MAP_OR_FAIL(fopen);

    tm1 = darshan_core_wtime();
    ret = __real_fopen(path, mode);
    tm2 = darshan_core_wtime();

    if(ret == NULL)
        fd = -1;
    else
        fd = fileno(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(fd, path, 0, 1, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

FILE* DARSHAN_DECL(fopen64)(const char *path, const char *mode)
{
    FILE* ret;
    int fd;
    double tm1, tm2;

    MAP_OR_FAIL(fopen64);

    tm1 = darshan_core_wtime();
    ret = __real_fopen64(path, mode);
    tm2 = darshan_core_wtime();

    if(ret == NULL)
        fd = -1;
    else
        fd = fileno(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(fd, path, 0, 1, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
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
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
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
648
649
650
651
652
653
ssize_t DARSHAN_DECL(read)(int fd, void *buf, size_t count)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(write);

    /* if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_read(fd, buf, count);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_READ(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(write)(int fd, const void *buf, size_t count)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(write);

    /* if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_write(fd, buf, count);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_WRITE(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pread)(int fd, void *buf, size_t count, off_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pread);

    /* if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_pread(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_READ(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pwrite)(int fd, const void *buf, size_t count, off_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pwrite);

    /* if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_pwrite(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_WRITE(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pread64)(int fd, void *buf, size_t count, off64_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pread64);

    /* if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_pread64(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_READ(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pwrite64)(int fd, const void *buf, size_t count, off64_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pwrite64);

    /* if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_pwrite64(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_WRITE(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(readv)(int fd, const struct iovec *iov, int iovcnt)
{
    ssize_t ret;
    int aligned_flag = 0;
    /* int aligned_flag = 1; */
    int i;
    double tm1, tm2;

    MAP_OR_FAIL(readv);
/*
    for(i=0; i<iovcnt; i++)
    {
        if(((unsigned long)iov[i].iov_base % darshan_mem_alignment) != 0)
            aligned_flag = 0;
    }*/

    tm1 = darshan_core_wtime();
    ret = __real_readv(fd, iov, iovcnt);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
654
    posix_runtime_initialize();
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
    POSIX_RECORD_READ(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(writev)(int fd, const struct iovec *iov, int iovcnt)
{
    ssize_t ret;
    int aligned_flag = 0;
    /* int aligned_flag = 1; */
    int i;
    double tm1, tm2;

    MAP_OR_FAIL(writev);
/*
    for(i=0; i<iovcnt; i++)
    {
        if(((unsigned long)iov[i].iov_base % darshan_mem_alignment) != 0)
            aligned_flag = 0;
    }*/

    tm1 = darshan_core_wtime();
    ret = __real_writev(fd, iov, iovcnt);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
682
    posix_runtime_initialize();
683
684
685
686
687
688
    POSIX_RECORD_WRITE(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

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
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
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
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
size_t DARSHAN_DECL(fread)(void *ptr, size_t size, size_t nmemb, FILE *stream)
{
    size_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(fread);

    /* if((unsigned long)ptr % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_fread(ptr, size, nmemb, stream);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    if(ret > 0)
    {
        POSIX_RECORD_READ(size*ret, fileno(stream), 0, 0,
            aligned_flag, 1, tm1, tm2);
    }
    else
    {
        POSIX_RECORD_READ(ret, fileno(stream), 0, 0,
            aligned_flag, 1, tm1, tm2);
    }
    POSIX_UNLOCK();

    return(ret);
}

size_t DARSHAN_DECL(fwrite)(const void *ptr, size_t size, size_t nmemb, FILE *stream)
{
    size_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(fwrite);

    /* if((unsigned long)ptr % darshan_mem_alignment == 0) aligned_flag = 1; */

    tm1 = darshan_core_wtime();
    ret = __real_fwrite(ptr, size, nmemb, stream);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    if(ret > 0)
    {
        POSIX_RECORD_WRITE(size*ret, fileno(stream), 0, 0,
            aligned_flag, 1, tm1, tm2);
    }
    else
    {
        POSIX_RECORD_WRITE(ret, fileno(stream), 0, 0,
            aligned_flag, 1, tm1, tm2);
    }
    POSIX_UNLOCK();

    return(ret);
}

off_t DARSHAN_DECL(lseek)(int fd, off_t offset, int whence)
{
    off_t ret;
    struct posix_file_runtime* file;
    double tm1, tm2;

    MAP_OR_FAIL(lseek);

    tm1 = darshan_core_wtime();
    ret = __real_lseek(fd, offset, whence);
    tm2 = darshan_core_wtime();

    if(ret >= 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        file = posix_file_by_fd(fd);
        if(file)
        {
            file->offset = ret;
            DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record,
                tm1, tm2, file->last_meta_end, POSIX_F_META_TIME);
            DARSHAN_COUNTER_INC(file->file_record, POSIX_SEEKS, 1);
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

off_t DARSHAN_DECL(lseek64)(int fd, off_t offset, int whence)
{
    off_t ret;
    struct posix_file_runtime* file;
    double tm1, tm2;

    MAP_OR_FAIL(lseek64);

    tm1 = darshan_core_wtime();
    ret = __real_lseek64(fd, offset, whence);
    tm2 = darshan_core_wtime();

    if(ret >= 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        file = posix_file_by_fd(fd);
        if(file)
        {
            file->offset = ret;
            DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record,
                tm1, tm2, file->last_meta_end, POSIX_F_META_TIME);
            DARSHAN_COUNTER_INC(file->file_record, POSIX_SEEKS, 1);
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

int DARSHAN_DECL(fseek)(FILE *stream, long offset, int whence)
{
    int ret;
    struct posix_file_runtime* file;
    double tm1, tm2;

    MAP_OR_FAIL(fseek);

    tm1 = darshan_core_wtime();
    ret = __real_fseek(stream, offset, whence);
    tm2 = darshan_core_wtime();

    if(ret >= 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        file = posix_file_by_fd(fileno(stream));
        if(file)
        {
830
            file->offset = ftell(stream);
831
832
833
834
835
836
837
838
839
840
            DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record,
                tm1, tm2, file->last_meta_end, POSIX_F_META_TIME);
            DARSHAN_COUNTER_INC(file->file_record, POSIX_FSEEKS, 1);
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
int DARSHAN_DECL(__xstat)(int vers, const char *path, struct stat *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__xstat);

    tm1 = darshan_core_wtime();
    ret = __real___xstat(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__xstat64)(int vers, const char *path, struct stat64 *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__xstat64);

    tm1 = darshan_core_wtime();
    ret = __real___xstat64(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__lxstat)(int vers, const char *path, struct stat *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__lxstat);

    tm1 = darshan_core_wtime();
    ret = __real___lxstat(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__lxstat64)(int vers, const char *path, struct stat64 *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__lxstat64);

    tm1 = darshan_core_wtime();
    ret = __real___lxstat64(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__fxstat)(int vers, int fd, struct stat *buf)
{
    int ret;
    struct posix_file_runtime* file;
    double tm1, tm2;

    MAP_OR_FAIL(__fxstat);

    tm1 = darshan_core_wtime();
    ret = __real___fxstat(vers, fd, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    /* TODO */
#if 0
    /* skip logging if this was triggered internally */
    if((void*)buf == (void*)&cp_stat_buf)
        return(ret);
#endif

    POSIX_LOCK();
    posix_runtime_initialize();
    file = posix_file_by_fd(fd);
    if(file)
    {
        POSIX_RECORD_STAT(file, buf, tm1, tm2);
    }
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__fxstat64)(int vers, int fd, struct stat64 *buf)
{
    int ret;
    struct posix_file_runtime* file;
    double tm1, tm2;

    MAP_OR_FAIL(__fxstat64);

    tm1 = darshan_core_wtime();
    ret = __real___fxstat64(vers, fd, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    /* TODO */
#if 0
    /* skip logging if this was triggered internally */
    if((void*)buf == (void*)&cp_stat_buf)
        return(ret);
#endif

    POSIX_LOCK();
    posix_runtime_initialize();
    file = posix_file_by_fd(fd);
    if(file)
    {
        POSIX_RECORD_STAT(file, buf, tm1, tm2);
    }
    POSIX_UNLOCK();

    return(ret);
}

Shane Snyder's avatar
Shane Snyder committed
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
void* DARSHAN_DECL(mmap)(void *addr, size_t length, int prot, int flags,
    int fd, off_t offset)
{
    void* ret;
    struct posix_file_runtime* file;

    MAP_OR_FAIL(mmap);

    ret = __real_mmap(addr, length, prot, flags, fd, offset);
    if(ret == MAP_FAILED)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    file = posix_file_by_fd(fd);
    if(file)
    {
        DARSHAN_COUNTER_INC(file->file_record, POSIX_MMAPS, 1);
    }
    POSIX_UNLOCK();

    return(ret);
}

void* DARSHAN_DECL(mmap64)(void *addr, size_t length, int prot, int flags,
    int fd, off64_t offset)
{
    void* ret;
    struct posix_file_runtime* file;

    MAP_OR_FAIL(mmap64);

    ret = __real_mmap64(addr, length, prot, flags, fd, offset);
    if(ret == MAP_FAILED)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    file = posix_file_by_fd(fd);
    if(file)
    {
        DARSHAN_COUNTER_INC(file->file_record, POSIX_MMAPS, 1);
    }
    POSIX_UNLOCK();

    return(ret);
}

1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
int DARSHAN_DECL(fsync)(int fd)
{
    int ret;
    struct posix_file_runtime* file;
    double tm1, tm2;

    MAP_OR_FAIL(fsync);

    tm1 = darshan_core_wtime();
    ret = __real_fsync(fd);
    tm2 = darshan_core_wtime();

    if(ret < 0)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    file = posix_file_by_fd(fd);
    if(file)
    {
        DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record,
            tm1, tm2, file->last_write_end, POSIX_F_WRITE_TIME);
        DARSHAN_COUNTER_INC(file->file_record, POSIX_FSYNCS, 1);
    }
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(fdatasync)(int fd)
{
    int ret;
    struct posix_file_runtime* file;
    double tm1, tm2;

    MAP_OR_FAIL(fdatasync);

    tm1 = darshan_core_wtime();
    ret = __real_fdatasync(fd);
    tm2 = darshan_core_wtime();

    if(ret < 0)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    file = posix_file_by_fd(fd);
    if(file)
    {
        DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record,
            tm1, tm2, file->last_write_end, POSIX_F_WRITE_TIME); 
        DARSHAN_COUNTER_INC(file->file_record, POSIX_FDSYNCS, 1);
    }
    POSIX_UNLOCK();

    return(ret);
}

1103
1104
int DARSHAN_DECL(close)(int fd)
{
1105
    struct posix_file_runtime* file;
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
    double tm1, tm2;
    int ret;

    MAP_OR_FAIL(close);

    tm1 = darshan_core_wtime();
    ret = __real_close(fd);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
Shane Snyder's avatar
Shane Snyder committed
1117
1118
1119
    file = posix_file_by_fd(fd);
    if(file)
    {
1120
1121
1122
1123
1124
1125
        file->last_byte_written = 0;
        file->last_byte_read = 0;
        DARSHAN_COUNTER_F_SET(file->file_record,
            POSIX_F_CLOSE_TIMESTAMP, darshan_core_wtime());
        DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record,
             tm1, tm2, file->last_meta_end, POSIX_F_META_TIME);
Shane Snyder's avatar
Shane Snyder committed
1126
1127
        posix_file_close_fd(fd);
    }
1128
1129
1130
1131
1132
    POSIX_UNLOCK();    

    return(ret);
}

1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
int DARSHAN_DECL(fclose)(FILE *fp)
{
    struct posix_file_runtime* file;
    int fd = fileno(fp);
    double tm1, tm2;
    int ret;

    MAP_OR_FAIL(fclose);

    tm1 = darshan_core_wtime();
    ret = __real_fclose(fp);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    file = posix_file_by_fd(fd);
    if(file)
    {
        file->last_byte_written = 0;
        file->last_byte_read = 0;
        DARSHAN_COUNTER_F_SET(file->file_record,
            POSIX_F_CLOSE_TIMESTAMP, darshan_core_wtime());
        DARSHAN_COUNTER_F_INC_NO_OVERLAP(file->file_record,
            tm1, tm2, file->last_meta_end, POSIX_F_META_TIME);
        posix_file_close_fd(fd);
    }
    POSIX_UNLOCK();

    return(ret);
}

1164
1165
1166
/**********************************************************
 * Internal functions for manipulating POSIX module state *
 **********************************************************/
1167

1168
/* initialize internal POSIX module data structures and register with darshan-core */
1169
static void posix_runtime_initialize()
1170
{
1171
1172
1173
    int mem_limit;
    struct darshan_module_funcs posix_mod_fns =
    {
Shane Snyder's avatar
Shane Snyder committed
1174
        .disable_instrumentation = &posix_disable_instrumentation,
1175
        .prepare_for_reduction = &posix_prepare_for_reduction,
1176
        .record_reduction_op = &posix_record_reduction_op,
1177
        .get_output_data = &posix_get_output_data,
1178
        .shutdown = &posix_shutdown
1179
    };
1180

Shane Snyder's avatar
Shane Snyder committed
1181
1182
    /* don't do anything if already initialized or instrumenation is disabled */
    if(posix_runtime || instrumentation_disabled)
1183
        return;
1184

1185
1186
    /* register the posix module with darshan core */
    darshan_core_register_module(
1187
        DARSHAN_POSIX_MOD,
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
        &posix_mod_fns,
        &mem_limit);

    /* return if no memory assigned by darshan core */
    if(mem_limit == 0)
        return;

    posix_runtime = malloc(sizeof(*posix_runtime));
    if(!posix_runtime)
        return;
    memset(posix_runtime, 0, sizeof(*posix_runtime));

    /* set maximum number of file records according to max memory limit */
    /* NOTE: maximum number of records is based on the size of a posix file record */
1202
    /* TODO: should we base memory usage off file record or total runtime structure sizes? */
1203
    posix_runtime->file_array_size = mem_limit / sizeof(struct darshan_posix_file);
1204
    posix_runtime->file_array_ndx = 0;
1205
1206

    /* allocate array of runtime file records */
1207
    posix_runtime->file_runtime_array = malloc(posix_runtime->file_array_size *
1208
                                               sizeof(struct posix_file_runtime));
1209
1210
1211
    posix_runtime->file_record_array = malloc(posix_runtime->file_array_size *
                                              sizeof(struct darshan_posix_file));
    if(!posix_runtime->file_runtime_array || !posix_runtime->file_record_array)
1212
1213
1214
1215
    {
        posix_runtime->file_array_size = 0;
        return;
    }
1216
    memset(posix_runtime->file_runtime_array, 0, posix_runtime->file_array_size *
1217
           sizeof(struct posix_file_runtime));
1218
1219
    memset(posix_runtime->file_record_array, 0, posix_runtime->file_array_size *
           sizeof(struct darshan_posix_file));
1220

Shane Snyder's avatar
Shane Snyder committed
1221
1222
    /* TODO: can we move this out of here? perhaps register_module returns rank? */
    DARSHAN_MPI_CALL(PMPI_Comm_rank)(MPI_COMM_WORLD, &my_rank);
1223

1224
1225
    return;
}
1226

1227
/* get a POSIX file record for the given file path */
1228
static struct posix_file_runtime* posix_file_by_name(const char *name)
1229
{
1230
    struct posix_file_runtime *file = NULL;
1231
    char *newname = NULL;
1232
    darshan_record_id file_id;
1233

Shane Snyder's avatar
Shane Snyder committed
1234
    if(!posix_runtime || instrumentation_disabled)
1235
        return(NULL);
1236

1237
    newname = darshan_clean_file_path(name);
1238
    if(!newname)
1239
        newname = (char*)name;
1240

1241
    /* get a unique id for this file from darshan core */
Shane Snyder's avatar
Shane Snyder committed
1242
    darshan_core_register_record(
1243
1244
1245
        (void*)newname,
        strlen(newname),
        1,
1246
        DARSHAN_POSIX_MOD,
1247
        &file_id);
1248

1249
    /* search the hash table for this file record, and return if found */
1250
    HASH_FIND(hlink, posix_runtime->file_hash, &file_id, sizeof(darshan_record_id), file);
1251
    if(file)
1252
    {
1253
        if(newname != name)
1254
            free(newname);
1255
        return(file);
1256
1257
    }

1258
1259
1260
1261
1262
1263
    if(posix_runtime->file_array_ndx < posix_runtime->file_array_size);
    {
        /* no existing record, assign a new file record from the global array */
        file = &(posix_runtime->file_runtime_array[posix_runtime->file_array_ndx]);
        file->file_record = &(posix_runtime->file_record_array[posix_runtime->file_array_ndx]);
        file->file_record->f_id = file_id;
1264

1265
        /* add new record to file hash table */
1266
        HASH_ADD(hlink, posix_runtime->file_hash, file_record->f_id, sizeof(darshan_record_id), file);
1267

1268
1269
        posix_runtime->file_array_ndx++;
    }
1270
1271

    if(newname != name)
1272
        free(newname);
1273
1274
1275
    return(file);
}

1276
1277
1278
/* get a POSIX file record for the given file path, and also create a
 * reference structure using the returned file descriptor
 */
1279
static struct posix_file_runtime* posix_file_by_name_setfd(const char* name, int fd)
1280
{
1281
1282
    struct posix_file_runtime* file;
    struct posix_file_runtime_ref* ref;
1283

Shane Snyder's avatar
Shane Snyder committed
1284
    if(!posix_runtime || instrumentation_disabled)
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
        return(NULL);

    /* find file record by name first */
    file = posix_file_by_name(name);

    if(!file)
        return(NULL);

    /* search hash table for existing file ref for this fd */
    HASH_FIND(hlink, posix_runtime->fd_hash, &fd, sizeof(int), ref);
    if(ref)
    {
        /* we have a reference.  Make sure it points to the correct file
         * and return it
         */
        ref->file = file;
        return(file);
    }

    /* if we hit this point, then we don't have a reference for this fd
     * in the table yet.  Add it.
     */
    ref = malloc(sizeof(*ref));
    if(!ref)
        return(NULL);
    memset(ref, 0, sizeof(*ref));
Shane Snyder's avatar
Shane Snyder committed
1311

1312
1313
1314
1315
1316
1317
1318
    ref->file = file;
    ref->fd = fd;    
    HASH_ADD(hlink, posix_runtime->fd_hash, fd, sizeof(int), ref);

    return(file);
}

1319
/* get a POSIX file record for the given file descriptor */
1320
static struct posix_file_runtime* posix_file_by_fd(int fd)
Shane Snyder's avatar
Shane Snyder committed
1321
{
1322
    struct posix_file_runtime_ref* ref;
Shane Snyder's avatar
Shane Snyder committed
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334

    if(!posix_runtime || instrumentation_disabled)
        return(NULL);

    /* search hash table for existing file ref for this fd */
    HASH_FIND(hlink, posix_runtime->fd_hash, &fd, sizeof(int), ref);
    if(ref)
        return(ref->file);

    return(NULL);
}

1335
/* free up reference data structures for the given file descriptor */
1336
1337
static void posix_file_close_fd(int fd)
{
1338
    struct posix_file_runtime_ref* ref;
1339

Shane Snyder's avatar
Shane Snyder committed
1340
    if(!posix_runtime || instrumentation_disabled)
1341
1342
1343
1344
        return;

    /* search hash table for this fd */
    HASH_FIND(hlink, posix_runtime->fd_hash, &fd, sizeof(int), ref);
1345
    if(ref)
1346
1347
1348
1349
1350
1351
1352
    {
        /* we have a reference, delete it */
        HASH_DELETE(hlink, posix_runtime->fd_hash, ref);
        free(ref);
    }

    return;
1353
1354
}

1355
/* compare function for sorting file records by descending rank */
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
static int posix_file_compare(const void* a, const void* b)
{
    const struct darshan_posix_file* f_a = a;
    const struct darshan_posix_file* f_b = b;

    if(f_a->rank < f_b->rank)
        return 1;
    if(f_a->rank > f_b->rank)
        return -1;

    return 0;
}

1369
1370
1371
/************************************************************************
 * Functions exported by this module for coordinating with darshan-core *
 ************************************************************************/
1372

Shane Snyder's avatar
Shane Snyder committed
1373
static void posix_disable_instrumentation()
1374
{
1375
1376
    assert(posix_runtime);

Shane Snyder's avatar
Shane Snyder committed
1377
1378
1379
    POSIX_LOCK();
    instrumentation_disabled = 1;
    POSIX_UNLOCK();
1380

Shane Snyder's avatar
Shane Snyder committed
1381
1382
1383
    return;
}

1384
1385
1386
1387
1388
1389
1390
static void posix_prepare_for_reduction(
    darshan_record_id *shared_recs,
    int *shared_rec_count,
    void **send_buf,
    void **recv_buf,
    int *rec_size)
{
1391
    struct posix_file_runtime *file;
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
    int i;

    assert(posix_runtime);

    /* necessary initialization of shared records (e.g., change rank to -1) */
    for(i = 0; i < *shared_rec_count; i++)
    {
        HASH_FIND(hlink, posix_runtime->file_hash, &shared_recs[i],
            sizeof(darshan_record_id), file);
        assert(file);

        /* TODO: any initialization before reduction */
        file->file_record->rank = -1;
    }

    /* sort the array of files descending by rank so that we get all of the 
     * shared files (marked by rank -1) in a contiguous portion at end 
     * of the array
     */
    qsort(posix_runtime->file_record_array, posix_runtime->file_array_ndx,
        sizeof(struct darshan_posix_file), posix_file_compare);

    /* make *send_buf point to the shared files at the end of sorted array */
    *send_buf =
        &(posix_runtime->file_record_array[posix_runtime->file_array_ndx-(*shared_rec_count)]);

    /* allocate memory for the reduction output on rank 0 */
    if(my_rank == 0)
    {
        *recv_buf = malloc(*shared_rec_count * sizeof(struct darshan_posix_file));
        if(!(*recv_buf))
            return;
    }

    *rec_size = sizeof(struct darshan_posix_file);

Shane Snyder's avatar
Shane Snyder committed
1428
    /* TODO: cleaner way to do this? */
1429
1430
1431
1432
1433
1434
1435
    if(my_rank == 0)
        posix_runtime->red_buf = *recv_buf;
    posix_runtime->shared_rec_count = *shared_rec_count;

    return;
}

1436
static void posix_record_reduction_op(
1437
1438
1439
1440
    void* infile_v,
    void* inoutfile_v,
    int *len,
    MPI_Datatype *datatype)
Shane Snyder's avatar
Shane Snyder committed
1441
{
1442
1443
1444
1445
    struct darshan_posix_file tmp_file;
    struct darshan_posix_file *infile = infile_v;
    struct darshan_posix_file *inoutfile = inoutfile_v;
    int i;
1446
    int j;
1447
1448
1449

    assert(posix_runtime);

1450
    for(i=0; i<*len; i++)
1451
1452
1453
1454
1455
1456
    {
        memset(&tmp_file, 0, sizeof(struct darshan_posix_file));

        tmp_file.f_id = infile->f_id;
        tmp_file.rank = -1;

1457
        /* sum */
1458
        for(j=POSIX_OPENS; j<=POSIX_FDSYNCS; j++)
1459
1460
1461
1462
1463
1464
        {
            tmp_file.counters[j] = infile->counters[j] + inoutfile->counters[j];
        }

        tmp_file.counters[POSIX_MODE] = infile->counters[POSIX_MODE];

1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
        /* sum */
        for(j=POSIX_BYTES_READ; j<=POSIX_BYTES_WRITTEN; j++)
        {
            tmp_file.counters[j] = infile->counters[j] + inoutfile->counters[j];
        }

        /* max */
        for(j=POSIX_MAX_BYTE_READ; j<=POSIX_MAX_BYTE_WRITTEN; j++)
        {
            tmp_file.counters[j] = (
                (infile->counters[j] > inoutfile->counters[j]) ?
                infile->counters[j] :
                inoutfile->counters[j]);
        }

        /* sum */
        for(j=POSIX_CONSEC_READS; j<=POSIX_RW_SWITCHES; j++)
        {
            tmp_file.counters[j] = infile->counters[j] +
                inoutfile->counters[j];
        }

1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
        /* min non-zero (if available) value */
        for(j=POSIX_F_OPEN_TIMESTAMP; j<=POSIX_F_WRITE_START_TIMESTAMP; j++)
        {
            if(infile->fcounters[j] > inoutfile->fcounters[j] && inoutfile->fcounters[j] > 0)
                tmp_file.fcounters[j] = inoutfile->fcounters[j];
            else
                tmp_file.fcounters[j] = infile->fcounters[j];
        }

        /* max */
        for(j=POSIX_F_READ_END_TIMESTAMP; j<=POSIX_F_CLOSE_TIMESTAMP; j++)
        {
            if(infile->fcounters[j] > inoutfile->fcounters[j])
                tmp_file.fcounters[j] = infile->fcounters[j];
            else
                tmp_file.fcounters[j] = inoutfile->fcounters[j];
        }

        /* sum */
        for(j=POSIX_F_READ_TIME; j<=POSIX_F_META_TIME; j++)
        {
            tmp_file.fcounters[j] = infile->fcounters[j] + inoutfile->fcounters[j];
        }

        /* max (special case) */
        if(infile->fcounters[POSIX_F_MAX_READ_TIME] >
            inoutfile->fcounters[POSIX_F_MAX_READ_TIME])
        {
            tmp_file.fcounters[POSIX_F_MAX_READ_TIME] =
                infile->fcounters[POSIX_F_MAX_READ_TIME];
            tmp_file.counters[POSIX_MAX_READ_TIME_SIZE] =
                infile->counters[POSIX_MAX_READ_TIME_SIZE];
        }
1520
        else
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
        {
            tmp_file.fcounters[POSIX_F_MAX_READ_TIME] =
                inoutfile->fcounters[POSIX_F_MAX_READ_TIME];
            tmp_file.counters[POSIX_MAX_READ_TIME_SIZE] =
                inoutfile->counters[POSIX_MAX_READ_TIME_SIZE];
        }

        if(infile->fcounters[POSIX_F_MAX_WRITE_TIME] >
            inoutfile->fcounters[POSIX_F_MAX_WRITE_TIME])
        {
            tmp_file.fcounters[POSIX_F_MAX_WRITE_TIME] =
                infile->fcounters[POSIX_F_MAX_WRITE_TIME];
            tmp_file.counters[POSIX_MAX_WRITE_TIME_SIZE] =
                infile->counters[POSIX_MAX_WRITE_TIME_SIZE];
        }
1536
        else
1537
1538
1539
1540
1541
1542
        {
            tmp_file.fcounters[POSIX_F_MAX_WRITE_TIME] =
                inoutfile->fcounters[POSIX_F_MAX_WRITE_TIME];
            tmp_file.counters[POSIX_MAX_WRITE_TIME_SIZE] =
                inoutfile->counters[POSIX_MAX_WRITE_TIME_SIZE];
        }
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558

        /* update pointers */
        *inoutfile = tmp_file;
        inoutfile++;
        infile++;
    }

    return;
}

static void posix_get_output_data(
    void **buffer,
    int *size)
{
    assert(posix_runtime);

Shane Snyder's avatar
Shane Snyder committed
1559
    /* TODO: cleaner way to do this? */
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
    /* clean up reduction state */
    if(my_rank == 0)
    {
        int tmp_ndx = posix_runtime->file_array_ndx - posix_runtime->shared_rec_count;
        memcpy(&(posix_runtime->file_record_array[tmp_ndx]), posix_runtime->red_buf,
            posix_runtime->shared_rec_count * sizeof(struct darshan_posix_file));
        free(posix_runtime->red_buf);
    }
    else
    {
        posix_runtime->file_array_ndx -= posix_runtime->shared_rec_count;
    }
1572
1573
1574

    *buffer = (void *)(posix_runtime->file_record_array);
    *size = posix_runtime->file_array_ndx * sizeof(struct darshan_posix_file);
1575

1576
    return;
1577
1578
}

1579
static void posix_shutdown()
1580
{
1581
    struct posix_file_runtime_ref *ref, *tmp;
1582
1583
1584
1585
1586
1587

    HASH_ITER(hlink, posix_runtime->fd_hash, ref, tmp)
    {
        HASH_DELETE(hlink, posix_runtime->fd_hash, ref);
        free(ref);
    }
1588

1589
    HASH_CLEAR(hlink, posix_runtime->file_hash); /* these entries are freed all at once below */
1590
1591
1592
1593
1594

    free(posix_runtime->file_runtime_array);
    free(posix_runtime->file_record_array);
    free(posix_runtime);
    posix_runtime = NULL;
1595
1596

    return;
1597
1598
}

1599
1600
1601
1602
1603
1604
1605
1606
/*
 * Local variables:
 *  c-indent-level: 4
 *  c-basic-offset: 4
 * End:
 *
 * vim: ts=8 sts=4 sw=4 expandtab
 */