darshan-logutils.c 43.8 KB
Newer Older
1
2
3
4
5
/*
 *  (C) 2009 by Argonne National Laboratory.
 *      See COPYRIGHT in top-level directory.
 */

6
#define _GNU_SOURCE
7
#include "darshan-util-config.h"
8
9
#include <stdio.h>
#include <string.h>
10
#include <assert.h>
11
#include <stdlib.h>
12
#include <unistd.h>
13
#include <inttypes.h>
14
15
16
17
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

18
19
20
21
22
#include <zlib.h>
#ifdef HAVE_LIBBZ2
#include <bzlib.h>
#endif

23
24
#include "darshan-logutils.h"

25
26
27
28
29
30
31
32
33
34
35
36
37
38
struct darshan_fd_s
{
    gzFile gzf;
#ifdef HAVE_LIBBZ2
    BZFILE* bzf;
#endif
    int64_t pos;
    char mode[2];
    int swap_flag;
    char version[10];
    int job_struct_size;
    char* name;
};

39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
/* isn't there a clever c way to avoid this? */
char *darshan_names[] = {
    "CP_INDEP_OPENS",
    "CP_COLL_OPENS",               /* count of MPI collective opens */
    "CP_INDEP_READS",              /* count of independent MPI reads */
    "CP_INDEP_WRITES",             /* count of independent MPI writes */
    "CP_COLL_READS",               /* count of collective MPI reads */
    "CP_COLL_WRITES",              /* count of collective MPI writes */
    "CP_SPLIT_READS",              /* count of split collective MPI reads */
    "CP_SPLIT_WRITES",             /* count of split collective MPI writes */
    "CP_NB_READS",                 /* count of nonblocking MPI reads */
    "CP_NB_WRITES",                /* count of nonblocking MPI writes */
    "CP_SYNCS",                    /* count of MPI_File_sync */
    "CP_POSIX_READS",              /* count of posix reads */
    "CP_POSIX_WRITES",             /* count of posix writes */
    "CP_POSIX_OPENS",              /* count of posix opens */
    "CP_POSIX_SEEKS",              /* count of posix seeks */
    "CP_POSIX_STATS",              /* count of posix stat/lstat/fstats */
    "CP_POSIX_MMAPS",              /* count of posix mmaps */
    "CP_POSIX_FREADS",
    "CP_POSIX_FWRITES",
    "CP_POSIX_FOPENS",
    "CP_POSIX_FSEEKS",
    "CP_POSIX_FSYNCS",
    "CP_POSIX_FDSYNCS",
64
65
66
    "CP_INDEP_NC_OPENS",
    "CP_COLL_NC_OPENS",
    "CP_HDF5_OPENS",
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
    "CP_COMBINER_NAMED",           /* count of each MPI datatype category */
    "CP_COMBINER_DUP",
    "CP_COMBINER_CONTIGUOUS",
    "CP_COMBINER_VECTOR",
    "CP_COMBINER_HVECTOR_INTEGER",
    "CP_COMBINER_HVECTOR",
    "CP_COMBINER_INDEXED",
    "CP_COMBINER_HINDEXED_INTEGER",
    "CP_COMBINER_HINDEXED",
    "CP_COMBINER_INDEXED_BLOCK",
    "CP_COMBINER_STRUCT_INTEGER",
    "CP_COMBINER_STRUCT",
    "CP_COMBINER_SUBARRAY",
    "CP_COMBINER_DARRAY",
    "CP_COMBINER_F90_REAL",
    "CP_COMBINER_F90_COMPLEX",
    "CP_COMBINER_F90_INTEGER",
    "CP_COMBINER_RESIZED",
    "CP_HINTS",                     /* count of MPI hints used */
    "CP_VIEWS",                     /* count of MPI set view calls */
    "CP_MODE",                      /* mode of file */
    "CP_BYTES_READ",                /* total bytes read */
    "CP_BYTES_WRITTEN",             /* total bytes written */
    "CP_MAX_BYTE_READ",             /* highest offset byte read */
    "CP_MAX_BYTE_WRITTEN",          /* highest offset byte written */
    "CP_CONSEC_READS",              /* count of consecutive reads */
    "CP_CONSEC_WRITES",             /* count of consecutive writes */
    "CP_SEQ_READS",                 /* count of sequential reads */
    "CP_SEQ_WRITES",                /* count of sequential writes */
    "CP_RW_SWITCHES",
    "CP_MEM_NOT_ALIGNED",           /* count of accesses not mem aligned */
    "CP_MEM_ALIGNMENT",             /* mem alignment in bytes */
    "CP_FILE_NOT_ALIGNED",          /* count of accesses not file aligned */
    "CP_FILE_ALIGNMENT",            /* file alignment in bytes */
101
102
    "CP_MAX_READ_TIME_SIZE",
    "CP_MAX_WRITE_TIME_SIZE",
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
    "CP_SIZE_READ_0_100",           /* count of posix read size ranges */
    "CP_SIZE_READ_100_1K",
    "CP_SIZE_READ_1K_10K",
    "CP_SIZE_READ_10K_100K",
    "CP_SIZE_READ_100K_1M",
    "CP_SIZE_READ_1M_4M",
    "CP_SIZE_READ_4M_10M",
    "CP_SIZE_READ_10M_100M",
    "CP_SIZE_READ_100M_1G",
    "CP_SIZE_READ_1G_PLUS",
    "CP_SIZE_WRITE_0_100",          /* count of posix write size ranges */
    "CP_SIZE_WRITE_100_1K",
    "CP_SIZE_WRITE_1K_10K",
    "CP_SIZE_WRITE_10K_100K",
    "CP_SIZE_WRITE_100K_1M",
    "CP_SIZE_WRITE_1M_4M",
    "CP_SIZE_WRITE_4M_10M",
    "CP_SIZE_WRITE_10M_100M",
    "CP_SIZE_WRITE_100M_1G",
    "CP_SIZE_WRITE_1G_PLUS",
    "CP_SIZE_READ_AGG_0_100",       /* count of MPI read size ranges */
    "CP_SIZE_READ_AGG_100_1K",
    "CP_SIZE_READ_AGG_1K_10K",
    "CP_SIZE_READ_AGG_10K_100K",
    "CP_SIZE_READ_AGG_100K_1M",
    "CP_SIZE_READ_AGG_1M_4M",
    "CP_SIZE_READ_AGG_4M_10M",
    "CP_SIZE_READ_AGG_10M_100M",
    "CP_SIZE_READ_AGG_100M_1G",
    "CP_SIZE_READ_AGG_1G_PLUS",
    "CP_SIZE_WRITE_AGG_0_100",      /* count of MPI write size ranges */
    "CP_SIZE_WRITE_AGG_100_1K",
    "CP_SIZE_WRITE_AGG_1K_10K",
    "CP_SIZE_WRITE_AGG_10K_100K",
    "CP_SIZE_WRITE_AGG_100K_1M",
    "CP_SIZE_WRITE_AGG_1M_4M",
    "CP_SIZE_WRITE_AGG_4M_10M",
    "CP_SIZE_WRITE_AGG_10M_100M",
    "CP_SIZE_WRITE_AGG_100M_1G",
    "CP_SIZE_WRITE_AGG_1G_PLUS",
    "CP_EXTENT_READ_0_100",          /* count of MPI read extent ranges */
    "CP_EXTENT_READ_100_1K",
    "CP_EXTENT_READ_1K_10K",
    "CP_EXTENT_READ_10K_100K",
    "CP_EXTENT_READ_100K_1M",
    "CP_EXTENT_READ_1M_4M",
    "CP_EXTENT_READ_4M_10M",
    "CP_EXTENT_READ_10M_100M",
    "CP_EXTENT_READ_100M_1G",
    "CP_EXTENT_READ_1G_PLUS",
    "CP_EXTENT_WRITE_0_100",         /* count of MPI write extent ranges */
    "CP_EXTENT_WRITE_100_1K",
    "CP_EXTENT_WRITE_1K_10K",
    "CP_EXTENT_WRITE_10K_100K",
    "CP_EXTENT_WRITE_100K_1M",
    "CP_EXTENT_WRITE_1M_4M",
    "CP_EXTENT_WRITE_4M_10M",
    "CP_EXTENT_WRITE_10M_100M",
    "CP_EXTENT_WRITE_100M_1G",
    "CP_EXTENT_WRITE_1G_PLUS",
    "CP_STRIDE1_STRIDE",             /* the four most frequently appearing strides */
    "CP_STRIDE2_STRIDE",
    "CP_STRIDE3_STRIDE",
    "CP_STRIDE4_STRIDE",
    "CP_STRIDE1_COUNT",              /* count of each of the most frequent strides */
    "CP_STRIDE2_COUNT",
    "CP_STRIDE3_COUNT",
    "CP_STRIDE4_COUNT",
    "CP_ACCESS1_ACCESS",
    "CP_ACCESS2_ACCESS",
    "CP_ACCESS3_ACCESS",
    "CP_ACCESS4_ACCESS",
    "CP_ACCESS1_COUNT",
    "CP_ACCESS2_COUNT",
    "CP_ACCESS3_COUNT",
    "CP_ACCESS4_COUNT",
179
    "CP_DEVICE",
180
    "CP_SIZE_AT_OPEN",
181
182
183
184
    "CP_FASTEST_RANK",
    "CP_FASTEST_RANK_BYTES",
    "CP_SLOWEST_RANK",
    "CP_SLOWEST_RANK_BYTES",
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202

    "CP_NUM_INDICES"
};

/* isn't there a clever c way to avoid this? */
char *darshan_f_names[] = {
    "CP_F_OPEN_TIMESTAMP",        /* timestamp of first open */
    "CP_F_READ_START_TIMESTAMP",  /* timestamp of first read */
    "CP_F_WRITE_START_TIMESTAMP", /* timestamp of first write */
    "CP_F_CLOSE_TIMESTAMP",       /* timestamp of last close */
    "CP_F_READ_END_TIMESTAMP",    /* timestamp of last read */
    "CP_F_WRITE_END_TIMESTAMP",   /* timestamp of last write */
    "CP_F_POSIX_READ_TIME",       /* cumulative posix read time */
    "CP_F_POSIX_WRITE_TIME",      /* cumulative posix write time */
    "CP_F_POSIX_META_TIME",       /* cumulative posix meta time */
    "CP_F_MPI_META_TIME",         /* cumulative mpi-io metadata time */
    "CP_F_MPI_READ_TIME",         /* cumulative mpi-io read time */
    "CP_F_MPI_WRITE_TIME",        /* cumulative mpi-io write time */
203
204
    "CP_F_MAX_READ_TIME",
    "CP_F_MAX_WRITE_TIME",
205
206
207
208
209
210
    "CP_F_FASTEST_RANK_TIME",
    "CP_F_SLOWEST_RANK_TIME",
    "CP_F_VARIANCE_RANK_TIME",
    "CP_F_VARIANCE_RANK_BYTES",

    "CP_F_NUM_INDICES"
211
212
};

213
214
215
216
217
218
219
/* function pointers so that we can switch functions depending on what file
 * version is detected
 */
int (*getjob_internal)(darshan_fd file, struct darshan_job *job);
int (*getfile_internal)(darshan_fd fd, 
    struct darshan_job *job, 
    struct darshan_file *file);
220
#define JOB_SIZE_124 28
221
#define JOB_SIZE_200 56
222

223
/* internal routines for parsing different file versions */
224
static int getjob_internal_201(darshan_fd file, struct darshan_job *job);
225
226
227
228
229
230
231
232
233
234
static int getjob_internal_200(darshan_fd file, struct darshan_job *job);
static int getfile_internal_200(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
static int getjob_internal_124(darshan_fd file, struct darshan_job *job);
static int getfile_internal_124(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
static int getfile_internal_122(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
static int getfile_internal_121(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file);
235
236
static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters);
237
static void shift_missing_1_24(struct darshan_file* file);
238
239
static void shift_missing_1_22(struct darshan_file* file);
static void shift_missing_1_21(struct darshan_file* file);
240

241
242
static int darshan_log_seek(darshan_fd fd, int64_t offset);
static int darshan_log_read(darshan_fd fd, void* buf, int len);
243
static int darshan_log_write(darshan_fd fd, void* buf, int len);
244
245
static const char* darshan_log_error(darshan_fd fd, int* errnum);

246
/* a rather crude API for accessing raw binary darshan files */
247
darshan_fd darshan_log_open(const char *name, const char* mode)
248
{
249
250
251
252
253
254
    int test_fd;
    uint8_t magic[2];
    int ret;
    int try_bz2 = 1;
    int len = strlen(name);

255
256
257
258
    /* we only allows "w" or "r" modes, nothing fancy */
    assert(strlen(mode) == 1);
    assert(mode[0] == 'r' || mode[0] == 'w');

259
260
261
262
263
    darshan_fd tmp_fd = malloc(sizeof(*tmp_fd));
    if(!tmp_fd)
        return(NULL);
    memset(tmp_fd, 0, sizeof(*tmp_fd));

264
265
266
267
268
269
270
271
272
    tmp_fd->mode[0] = mode[0];
    tmp_fd->mode[1] = mode[1];
    tmp_fd->name  = strdup(name);
    if(!tmp_fd->name)
    {
        free(tmp_fd);
        return(NULL);
    }

273
274
275
276
277
278
279
280
    if(strcmp(mode, "r") == 0)
    {
        /* Try to detect if existing file is a bzip2 file or not.  Both 
         * libbz2 and libz will fall back to normal I/O (without compression) 
         * automatically, so we need to do some detection manually up front 
         * in order to get a chance to try both compression formats.
         */
        test_fd = open(name, O_RDONLY);
281
        if(test_fd < 0)
282
283
        {
            perror("open");
284
            free(tmp_fd->name);
285
286
287
288
289
290
291
292
            free(tmp_fd);
            return(NULL);
        }
        ret = read(test_fd, &magic, 2);
        if(ret != 2)
        {
            fprintf(stderr, "Error: failed to read any data from %s.\n", 
                name);
293
            free(tmp_fd->name);
294
295
296
297
298
299
300
301
302
303
304
            free(tmp_fd);
            close(test_fd);
            return(NULL);
        }
        /* header magic for bz2 */
        if(magic[0] != 0x42 && magic[1] != 0x5A)
        {
            try_bz2 = 0;
        }
        close(test_fd);
    }
305

306
307
308
309
310
311
    if(strcmp(mode, "w") == 0)
    {
        /* TODO: is this the behavior that we want? */
        /* if we are writing a new file, go by the file extension to tell
         * whether to use bz2 or not?
         */
312
        if(len >= 3 && name[len-3] == 'b' && name[len-2] == 'z' && name[len-1] == '2')
313
314
315
316
317
            try_bz2 = 1;
        else
            try_bz2 = 0;
    }

318
#ifdef HAVE_LIBBZ2
319
320
321
    if(try_bz2)
    {
        tmp_fd->bzf = BZ2_bzopen(name, mode);
322
        if(!tmp_fd->bzf)
323
        {
324
325
326
            free(tmp_fd->name);
            free(tmp_fd);
            return(NULL);
327
        }
328
        return(tmp_fd);
329
    }
330
331
332
333
334
335
336
#else
    if(try_bz2)
    {
        fprintf(stderr, "Error: this Darshan build does not support bz2 files.\n");
        fprintf(stderr, "Error: please install libbz2-dev and reconfigure.\n");
        return(NULL);
    }
337
338
#endif

339
    tmp_fd->gzf = gzopen(name, mode);
340
341
    if(!tmp_fd->gzf)
    {
342
        free(tmp_fd->name);
343
344
345
346
        free(tmp_fd);
        tmp_fd = NULL;
    }
    return tmp_fd;
347
348
}

349
/* darshan_log_getjob()
350
351
352
 *
 * returns 0 on success, -1 on failure
 */
353
int darshan_log_getjob(darshan_fd file, struct darshan_job *job)
354
355
{
    int ret;
356
    char buffer[DARSHAN_JOB_METADATA_LEN];
357

358
359
360
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
361

362
363
364
    /* read version number first so we know how to digest the rest of the
     * file
     */
365
    ret = darshan_log_read(file, file->version, 10);
366
    if(ret < 10)
367
    {
368
        fprintf(stderr, "Error: invalid log file (failed to read version).\n");
369
370
371
        return(-1);
    }

372
    if(strcmp(file->version, "2.02") == 0)
Philip Carns's avatar
Philip Carns committed
373
374
375
376
377
378
    {
        getjob_internal = getjob_internal_201;
        getfile_internal = getfile_internal_200;
        file->job_struct_size = sizeof(*job);
    }
    else if(strcmp(file->version, "2.01") == 0)
379
    {
380
        getjob_internal = getjob_internal_201;
381
        getfile_internal = getfile_internal_200;
382
        file->job_struct_size = sizeof(*job);
383
    }
384
385
386
387
388
389
    else if(strcmp(file->version, "2.00") == 0)
    {
        getjob_internal = getjob_internal_200;
        getfile_internal = getfile_internal_200;
        file->job_struct_size = JOB_SIZE_200;
    }
390
    else if(strcmp(file->version, "1.24") == 0)
391
    {
392
393
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
394
        file->job_struct_size = JOB_SIZE_124;
395
    }
396
    else if(strcmp(file->version, "1.23") == 0)
397
    {
398
399
400
        /* same as 1.24, except that mnt points may be incorrect */
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
401
        file->job_struct_size = JOB_SIZE_124;
402
    }
403
    else if(strcmp(file->version, "1.22") == 0)
404
    {
405
406
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_122;
407
        file->job_struct_size = JOB_SIZE_124;
408
    }
409
    else if(strcmp(file->version, "1.21") == 0)
410
    {
411
        getjob_internal = getjob_internal_124;
Philip Carns's avatar
Philip Carns committed
412
        getfile_internal = getfile_internal_121;
413
        file->job_struct_size = JOB_SIZE_124;
414
415
    }
    else
416
    {
417
418
419
        fprintf(stderr, "Error: incompatible darshan file.\n");
        fprintf(stderr, "Error: expected version %s, but got %s\n", 
                CP_VERSION, file->version);
420
        return(-1);
421
422
    }

423
    ret = getjob_internal(file, job);
424
425
426

    if (ret == 0)
    {
427
#ifdef HAVE_STRNDUP
428
        char *metadata = strndup(job->metadata, sizeof(job->metadata));
429
#else
Philip Carns's avatar
bug fix    
Philip Carns committed
430
        char *metadata = strdup(job->metadata);
431
#endif
432
433
        char *kv;
        char *key;
434
        char *value;
435
436
        char *save;

437
438
439
        for(kv=strtok_r(metadata, "\n", &save);
            kv != NULL;
            kv=strtok_r(NULL, "\n", &save))
440
        {
441
442
443
444
445
446
447
448
449
450
451
452
            /* NOTE: we intentionally only split on the first = character.
             * There may be additional = characters in the value portion
             * (for example, when storing mpi-io hints).
             */
            strcpy(buffer, kv);
            key = buffer;
            value = index(buffer, '=');
            if(!value)
                continue;
            /* convert = to a null terminator to split key and value */
            value[0] = '\0';
            value++;
453
            if (strcmp(key, "prev_ver") == 0)
454
            {
455
                strncpy(job->version_string, value, sizeof(job->version_string));
456
            }
457
458
        }
        free(metadata);
459
460
    }

461
462
    return(ret);
}
463

464
465
466
467
468
469
470
471
/* darshan_putjob()
 * write job header in gzfile
 *
 * return 0 on success, -1 on failure.
 */
int darshan_log_putjob(darshan_fd file, struct darshan_job *job)
{
    struct darshan_job job_copy;
472
    char    pv_str[64];
473
    int     ret;
474
    int len;
475

476
477
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
478
479
        return(ret);

480
481
    memset(&job_copy, 0, sizeof(job_copy));
    memcpy(&job_copy, job, sizeof(job_copy));
482
483
484
485
486
487
488
489
490
491
492
    /* check for newline in existing metadata, add if needed */
    len = strlen(job_copy.metadata);
    if(len > 0 && len < DARSHAN_JOB_METADATA_LEN)
    {
        if(job_copy.metadata[len-1] != '\n')
        {
            job_copy.metadata[len] = '\n';
            job_copy.metadata[len+1] = '\0';
        }
    }

493
    sprintf(pv_str, "prev_ver=%s\n", job->version_string);
494
    sprintf(job_copy.version_string, "%s", CP_VERSION);
Philip Carns's avatar
Philip Carns committed
495
496
497
498
    if(strlen(job_copy.metadata) + strlen(pv_str) < DARSHAN_JOB_METADATA_LEN)
        strncat(job_copy.metadata, pv_str, strlen(pv_str));
    else
        sprintf(job_copy.metadata, "%s", pv_str);
499
500
    job_copy.magic_nr = CP_MAGIC_NR;

501
    ret = darshan_log_write(file, &job_copy, sizeof(job_copy));
502
503
504
505
506
507
508
509
510
    if (ret != sizeof(job_copy))
    {
        fprintf(stderr, "Error: failed to write job header: %d\n", ret);
        return(-1);
    }

    return(0);
}

511
512
513
514
515
516
517
/* darshan_log_getfile()
 *
 * return 1 if file record found, 0 on eof, and -1 on error
 */
int darshan_log_getfile(darshan_fd fd, struct darshan_job *job, struct darshan_file *file)
{
    int ret;
518

519
520
521
    ret = getfile_internal(fd, job, file);

    return(ret);
522
523
}

524
525
526
527
528
529
530
531
/* darshan_log_putfile()
 *
 * return 0 if file record written, -1 on error.
 */
int darshan_log_putfile(darshan_fd fd, struct darshan_job *job, struct darshan_file *file)
{
    int     ret;

532
533
534
535
536
537
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
538

539
    ret = darshan_log_write(fd, file, sizeof(*file));
540
541
542
543
544
545
546
547
548
    if (ret != sizeof(*file))
    {
        fprintf(stderr, "Error: writing file record failed: %d\n", ret);
        return(-1);
    }

    return(0);
}

549
550
551
552
553
554
/* darshan_log_getmounts()
 * 
 * retrieves mount table information from the log.  Note that devs, mnt_pts,
 * and fs_types are arrays that will be allocated by the function and must
 * be freed by the caller.  count will indicate the size of the arrays
 */
555
int darshan_log_getmounts(darshan_fd fd, int64_t** devs, char*** mnt_pts, char***
556
    fs_types, int* count)
557
558
559
560
561
562
{
    int ret;
    char* pos;
    int array_index = 0;
    char buf[CP_EXE_LEN+1];

563
564
565
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
566

567
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
568
569
    if (ret < (CP_EXE_LEN + 1))
    {
570
        perror("darshan_log_read");
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
        return(-1);
    }

    /* count entries */
    *count = 0;
    pos = buf;
    while((pos = strchr(pos, '\n')) != NULL)
    {
        pos++;
        (*count)++;
    }

    if(*count == 0)
    {
        /* no mount entries present */
        return(0);
    }

    /* allocate output arrays */
590
    *devs = malloc((*count)*sizeof(int64_t));
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
    assert(*devs);
    *mnt_pts = malloc((*count)*sizeof(char*));
    assert(*mnt_pts);
    *fs_types = malloc((*count)*sizeof(char*));
    assert(*fs_types);
    
    /* work backwards through the table and parse each line (except for
     * first, which holds command line information)
     */
    while((pos = strrchr(buf, '\n')) != NULL)
    {
        /* overestimate string lengths */
        (*mnt_pts)[array_index] = malloc(CP_EXE_LEN);
        assert((*mnt_pts)[array_index]);
        (*fs_types)[array_index] = malloc(CP_EXE_LEN);
        assert((*fs_types)[array_index]);
        
Philip Carns's avatar
Philip Carns committed
608
        ret = sscanf(++pos, "%" PRId64 "\t%s\t%s", &(*devs)[array_index],
609
610
            (*fs_types)[array_index], (*mnt_pts)[array_index]);

611
612
613
614
615
616
617
618
619
        if(ret != 3)
        {
            fprintf(stderr, "Error: poorly formatted mount table in log file.\n");
            return(-1);
        }
        pos--;
        *pos = '\0';
        array_index++;
    }
620

621
622
623
    return (0);
}

624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
/* darshan_log_putmounts
 *
 * encode mount information back into mtab format.
 *
 * returns 0 on success, -1 on failure.
 */
int darshan_log_putmounts(darshan_fd fd, int64_t* devs, char** mnt_pts, char** fs_types, int count)
{
    int     ret;
    char    line[1024];
    int     i;

    for(i=count-1; i>=0; i--)
    {
        sprintf(line, "\n%" PRId64 "\t%s\t%s",
                devs[i], fs_types[i], mnt_pts[i]);
640
        ret = darshan_log_write(fd, line, strlen(line));
641
642
643
644
645
646
        if (ret != strlen(line))
        {
            fprintf(stderr, "Error: failed to write mount entry: %d\n", ret);
            return(-1);
        }
    }
647
648
649
650
651
652
653
654

    /* seek ahead to end of exe region, will be zero filled */
    ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
    if (ret)
    {
        fprintf(stderr, "Error: forward seek failed: %d\n", CP_JOB_RECORD_SIZE);
    }

655
656
    return(0);
}
657

658
int darshan_log_getexe(darshan_fd fd, char *buf)
659
660
{
    int ret;
661
    char* newline;
662

663
664
665
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
666

667
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
668
669
    if (ret < (CP_EXE_LEN + 1))
    {
670
        perror("darshan_log_read");
671
672
        return(-1);
    }
673

674
675
676
677
678
679
680
681
    /* this call is only supposed to return the exe string, but starting in
     * log format 1.23 there could be a table of mount entry information
     * after the exe.  Look for newline character and truncate there.
     */
    newline = strchr(buf, '\n');
    if(newline)
        *newline = '\0';

682
683
684
    return (0);
}

685
686
687
688
689
690
691
692
693
694
695
/* darshan_log_putexe()
 *
 * Write the exe string to the log.
 *
 * return 0 on success, -1 on failure.
 */
int darshan_log_putexe(darshan_fd fd, char *buf)
{
    int     ret;
    int     len;

696
697
    ret = darshan_log_seek(fd, sizeof(struct darshan_job));
    if(ret < 0)
698
699
700
701
        return(ret);

    len = strlen(buf);

702
    ret = darshan_log_write(fd, buf, len);
703
704
705
706
707
708
709
710
711
    if (ret != len)
    {
        fprintf(stderr, "Error: failed to write exe info: %d\n", ret);
        ret = -1;
    }

    return(ret);
}

712
void darshan_log_close(darshan_fd file)
713
{
714
715
716
717
718
719
720
721
#ifdef HAVE_LIBBZ2
    if(file->bzf)
        BZ2_bzclose(file->bzf);
#endif

    if(file->gzf)
        gzclose(file->gzf);

722
    free(file->name);
723
    free(file);
724
725
}

726
727
728
729
730
731
/* darshan_log_print_version_warnings()
 *
 * Print summary of any problems with the detected log format
 */
void darshan_log_print_version_warnings(struct darshan_job *job)
{
732
733
    if(strcmp(job->version_string, "2.02") == 0)
    {
734
        /* current version */
735
736
737
        return;
    }

Philip Carns's avatar
Philip Carns committed
738
739
740
741
742
743
744
    if(strcmp(job->version_string, "2.01") == 0)
    {
        printf("# WARNING: version 2.01 log format has the following limitations:\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
        return;
    }

745
746
    if(strcmp(job->version_string, "2.00") == 0)
    {
Philip Carns's avatar
Philip Carns committed
747
748
        printf("# WARNING: version 2.00 log format has the following limitations:\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
749
750
        return;
    }
751
 
752
    if(strcmp(job->version_string, "1.24") == 0)
753
    {
754
755
756
757
758
759
760
761
762
        printf("# WARNING: version 1.24 log format does not support the following parameters:\n");
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
763
764
765
        printf("# WARNING: version 1.24 log format has the following limitations:\n");
        printf("# - does not store the job id in the file.\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
766
767
        return;
    }
768
    
769
770
    if(strcmp(job->version_string, "1.23") == 0)
    {
771
772
773
774
775
776
777
778
779
        printf("# WARNING: version 1.23 log format does not support the following parameters:\n");
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
780
781
782
783
        printf("# WARNING: version 1.23 log format has the following limitations:\n");
        printf("# - may have incorrect mount point mappings for files with rank > 0.\n");
        printf("# - does not store the job id in the file.\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
784
        return;
785
786
    }

787
788
789
790
    if(strcmp(job->version_string, "1.22") == 0)
    {
        printf("# WARNING: version 1.22 log format does not support the following parameters:\n");
        printf("#   CP_DEVICE\n");
791
        printf("#   CP_SIZE_AT_OPEN\n");
792
793
794
795
796
797
798
799
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
800
801
802
803
804
        printf("# WARNING: version 1.22 log format has the following limitations:\n");
        printf("# - does not record mounted file systems, mount points, or fs types.\n");
        printf("# - attributes syncs to cumulative metadata time, rather than cumulative write time.\n");
        printf("# - does not store the job id in the file.\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
805
806
        return;
    }
807
808
809
810
811
812
813
814
815

    if(strcmp(job->version_string, "1.21") == 0)
    {
        printf("# WARNING: version 1.21 log format does not support the following parameters:\n");
        printf("#   CP_INDEP_NC_OPENS\n");
        printf("#   CP_COLL_NC_OPENS\n");
        printf("#   CP_HDF5_OPENS\n");
        printf("#   CP_MAX_READ_TIME_SIZE\n");
        printf("#   CP_MAX_WRITE_TIME_SIZE\n");
816
        printf("#   CP_DEVICE\n");
817
        printf("#   CP_SIZE_AT_OPEN\n");
818
819
        printf("#   CP_F_MAX_READ_TIME\n");
        printf("#   CP_F_MAX_WRITE_TIME\n");
820
821
822
823
824
825
826
827
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
828
829
830
831
832
        printf("# WARNING: version 1.21 log format has the following limitations:\n");
        printf("# - does not record mounted file systems, mount points, or fs types.\n");
        printf("# - attributes syncs to cumulative metadata time, rather than cumulative write time.\n");
        printf("# - does not store the job id in the file.\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
833
834
835
836
837
838
839
840
        return;
    }

    fprintf(stderr, "Error: version %s not supported by parser.\n",
        job->version_string);
    return;
}

841
842
843
844
/* shift_missing_1_21()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.21
845
 */
846
/*******************************
Philip Carns's avatar
Philip Carns committed
847
 * version 1.21 to 2.00 differences 
848
849
850
851
852
853
854
855
 * - added:
 *   - CP_INDEP_NC_OPENS
 *   - CP_COLL_NC_OPENS
 *   - CP_HDF5_OPENS
 *   - CP_MAX_READ_TIME_SIZE
 *   - CP_MAX_WRITE_TIME_SIZE
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
856
857
858
859
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
860
861
 *   - CP_F_MAX_READ_TIME
 *   - CP_F_MAX_WRITE_TIME
862
863
864
865
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
866
 * - changed params:
867
868
869
 *   - CP_FILE_RECORD_SIZE: 1184 to 1328
 *   - CP_NUM_INDICES: 133 to 144
 *   - CP_F_NUM_INDICES: 12 to 18
870
 */
871
872
873
874
875
876
877
878
879
static void shift_missing_1_21(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_INDEP_NC_OPENS,
        CP_COLL_NC_OPENS,
        CP_HDF5_OPENS,
        CP_MAX_READ_TIME_SIZE,
        CP_MAX_WRITE_TIME_SIZE,
880
        CP_DEVICE,
881
        CP_SIZE_AT_OPEN,
882
883
884
885
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
886
887
888
889
        -1};
    int missing_f_counters[] = {
        CP_F_MAX_READ_TIME,
        CP_F_MAX_WRITE_TIME,
890
891
892
893
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
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
        -1};

    c_index = 0;
    while(missing_counters[c_index] != -1)
    {
        int missing_counter = missing_counters[c_index];
        c_index++;
        if(missing_counter < (CP_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->counters[missing_counter+1],
                &file->counters[missing_counter],
                (CP_NUM_INDICES-missing_counter-1)*sizeof(int64_t));
        }
        /* zero out missing counter */
        file->counters[missing_counter] = 0;
    }

    c_index = 0;
    while(missing_f_counters[c_index] != -1)
    {
        int missing_counter = missing_f_counters[c_index];
        c_index++;
        if(missing_counter < (CP_F_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->fcounters[missing_counter+1],
                &file->fcounters[missing_counter],
                (CP_F_NUM_INDICES-missing_counter-1)*sizeof(double));
        }
        /* zero out missing counter */
        file->fcounters[missing_counter] = 0;
    }

    return;
}
930
931
932
933
934
935

/* shift_missing_1_22()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.22
 */
936
/*******************************
Philip Carns's avatar
Philip Carns committed
937
 * version 1.22 to 2.00 differences
938
939
940
941
 *
 * - added:
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
942
943
944
945
946
947
948
949
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
950
 * - changed params:
951
952
953
 *   - CP_FILE_RECORD_SIZE: 1240 to 1328
 *   - CP_NUM_INDICES: 138 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
954
 */
955
956
957
958
959
static void shift_missing_1_22(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_DEVICE,
960
        CP_SIZE_AT_OPEN,
961
962
963
964
965
966
967
968
969
970
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
        -1};
    int missing_f_counters[] = {
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
        -1};

    c_index = 0;
    while(missing_counters[c_index] != -1)
    {
        int missing_counter = missing_counters[c_index];
        c_index++;
        if(missing_counter < (CP_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->counters[missing_counter+1],
                &file->counters[missing_counter],
                (CP_NUM_INDICES-missing_counter-1)*sizeof(int64_t));
        }
        /* zero out missing counter */
        file->counters[missing_counter] = 0;
    }

989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
    c_index = 0;
    while(missing_f_counters[c_index] != -1)
    {
        int missing_counter = missing_f_counters[c_index];
        c_index++;
        if(missing_counter < (CP_F_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->fcounters[missing_counter+1],
                &file->fcounters[missing_counter],
                (CP_F_NUM_INDICES-missing_counter-1)*sizeof(double));
        }
        /* zero out missing counter */
        file->fcounters[missing_counter] = 0;
    }

1005
1006
    return;
}
1007

1008
1009
1010
1011
1012
1013
1014
1015
1016
/* shift_missing_1_24()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.24
 */
/*******************************
 * version 1.24 to 2.00 differences
 *
 * - added:
1017
1018
1019
1020
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
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
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
 * - changed params:
 *   - CP_FILE_RECORD_SIZE: ? to 1328
 *   - CP_NUM_INDICES: 140 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
 */
static void shift_missing_1_24(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
        -1};
    int missing_f_counters[] = {
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
        -1};

    c_index = 0;
    while(missing_counters[c_index] != -1)
    {
        int missing_counter = missing_counters[c_index];
        c_index++;
        if(missing_counter < (CP_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->counters[missing_counter+1],
                &file->counters[missing_counter],
                (CP_NUM_INDICES-missing_counter-1)*sizeof(int64_t));
        }
        /* zero out missing counter */
        file->counters[missing_counter] = 0;
    }

    c_index = 0;
    while(missing_f_counters[c_index] != -1)
    {
        int missing_counter = missing_f_counters[c_index];
        c_index++;
        if(missing_counter < (CP_F_NUM_INDICES - 1))
        {
            /* shift down */
            memmove(&file->fcounters[missing_counter+1],
                &file->fcounters[missing_counter],
                (CP_F_NUM_INDICES-missing_counter-1)*sizeof(double));
        }
        /* zero out missing counter */
        file->fcounters[missing_counter] = 0;
    }

    return;
}

1081
static int getjob_internal_201(darshan_fd file, struct darshan_job *job)
1082
1083
1084
{
    int ret;

1085
1086
1087
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
1088

1089
    ret = darshan_log_read(file, job, sizeof(*job));
1090
1091
    if (ret < sizeof(*job))
    {
1092
        fprintf(stderr, "Error: invalid log file (too short).\n");
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
        return(-1);
    }

    if(job->magic_nr == CP_MAGIC_NR)
    {
        /* no byte swapping needed, this file is in host format already */
        file->swap_flag = 0;
        return(0);
    }

    /* try byte swapping */
    DARSHAN_BSWAP64(&job->magic_nr);
    if(job->magic_nr == CP_MAGIC_NR)
    {
        file->swap_flag = 1;
        DARSHAN_BSWAP64(&job->uid);
        DARSHAN_BSWAP64(&job->start_time);
        DARSHAN_BSWAP64(&job->end_time);
        DARSHAN_BSWAP64(&job->nprocs);
1112
        DARSHAN_BSWAP64(&job->jobid);
1113
1114
1115
1116
1117
1118
1119
1120
        return(0);
    }

    /* otherwise this file is just broken */
    fprintf(stderr, "Error: bad magic number in darshan file.\n");
    return(-1);
}

1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135

static int getjob_internal_200(darshan_fd file, struct darshan_job *job)
{
    int ret;
    struct darshan_job_200
    {
        char version_string[8];
        int64_t magic_nr;
        int64_t uid;
        int64_t start_time;
        int64_t end_time;
        int64_t nprocs;
        int64_t jobid;
    } job_200;

1136
    memset(job, 0, sizeof(job_200));
1137
1138
    memset(job, 0, sizeof(*job));

1139
1140
1141
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
1142

1143
    ret = darshan_log_read(file, &job_200, sizeof(job_200));
1144
1145
    if (ret < sizeof(job_200))
    {
1146
        fprintf(stderr, "Error: invalid log file (too short).\n");
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
        return(-1);
    }

    memcpy(job->version_string, job_200.version_string, 8);
    job->magic_nr   = job_200.magic_nr;
    job->uid        = job_200.uid;
    job->start_time = job_200.start_time;
    job->end_time   = job_200.end_time;
    job->nprocs     = job_200.nprocs;
    job->jobid      = job_200.jobid;

    if(job->magic_nr == CP_MAGIC_NR)
    {
        /* no byte swapping needed, this file is in host format already */
        file->swap_flag = 0;
        return(0);
    }

    /* try byte swapping */
    DARSHAN_BSWAP64(&job->magic_nr);
    if(job->magic_nr == CP_MAGIC_NR)
    {
        file->swap_flag = 1;
        DARSHAN_BSWAP64(&job->uid);
        DARSHAN_BSWAP64(&job->start_time);
        DARSHAN_BSWAP64(&job->end_time);
        DARSHAN_BSWAP64(&job->nprocs);
        DARSHAN_BSWAP64(&job->jobid);
        return(0);
    }

    /* otherwise this file is just broken */
    fprintf(stderr, "Error: bad magic number in darshan file.\n");
    return(-1);
}

1183
1184
1185
1186
1187
1188
static int getfile_internal_200(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;
    const char* err_string;
    int i;
1189

1190
1191
1192
1193
1194
1195
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
1196
1197
1198
1199
1200
1201

    /* reset file record, so that diff compares against a zero'd out record
     * if file is missing
     */
    memset(file, 0, sizeof(&file));

1202
    ret = darshan_log_read(fd, file, sizeof(*file));
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
    if(ret == sizeof(*file))
    {
        /* got exactly one, correct size record */
        if(fd->swap_flag)
        {
            /* swap bytes if necessary */
            DARSHAN_BSWAP64(&file->hash);
            DARSHAN_BSWAP64(&file->rank);
            for(i=0; i<CP_NUM_INDICES; i++)
                DARSHAN_BSWAP64(&file->counters[i]);
            for(i=0; i<CP_F_NUM_INDICES; i++)
                DARSHAN_BSWAP64(&file->fcounters[i]);
        }
        return(1);
    }


    if(ret > 0)
    {
        /* got a short read */
        fprintf(stderr, "Error: invalid file record (too small)\n");
        return(-1);
    }

1227
    if(ret == 0)
1228
1229
1230
1231
1232
1233
    {
        /* hit end of file */
        return(0);
    }

    /* all other errors */
1234
    err_string = darshan_log_error(fd, &ret);
1235
1236
1237
1238
    fprintf(stderr, "Error: %s\n", err_string);
    return(-1);
}

1239
1240
1241
/* If we see version 1.24, assume that it is stored in big endian 32 bit
 * format.  Convert up to current format.
 */
1242
1243
static int getjob_internal_124(darshan_fd fd, struct darshan_job *job)
{
1244
1245
1246
1247
1248
1249
1250
    char* buffer;
    int ret;
    uint32_t uid;
    int32_t start_time;
    int32_t end_time;
    int32_t nprocs;

1251
1252
1253
1254
1255
1256
#ifdef WORDS_BIGENDIAN
    fd->swap_flag = 0;
#else
    fd->swap_flag = 1;
#endif

1257
1258
1259
1260
1261
1262
1263
1264
    memset(job, 0, sizeof(*job));

    buffer = (char*)malloc(JOB_SIZE_124);
    if(!buffer)
    {
        return(-1);
    }

1265
1266
1267
    ret = darshan_log_seek(fd, 0);
    if(ret < 0)
        return(ret);
1268

1269
    ret = darshan_log_read(fd, buffer, JOB_SIZE_124);
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
    if (ret < JOB_SIZE_124)
    {
        fprintf(stderr, "Error: invalid log file (could not read file record).\n");
        free(buffer);
        return(-1);
    }

    /* pull job header information out of specific bytes in case struct
     * padding is off
     */
    strncpy(job->version_string, buffer, 8);
    uid = *((uint32_t*)&buffer[12]);
    start_time = *((int32_t*)&buffer[16]);
    end_time = *((int32_t*)&buffer[20]);
    nprocs = *((int32_t*)&buffer[24]);

    free(buffer);

1288
1289
1290
1291
1292
1293
1294
1295
    if(fd->swap_flag)
    {
        /* byte swap */
        DARSHAN_BSWAP32(&uid);
        DARSHAN_BSWAP32(&start_time);
        DARSHAN_BSWAP32(&end_time);
        DARSHAN_BSWAP32(&nprocs);
    }
1296
1297
1298
1299
1300

    job->uid += uid;
    job->start_time += start_time;
    job->end_time += end_time;
    job->nprocs += nprocs;
1301
    job->jobid = 0; /* old log versions did not have this field */
1302
1303
1304
1305
1306
    
    /* set magic number */
    job->magic_nr = CP_MAGIC_NR;

    return(0);
1307
1308
1309
1310
}

static int getfile_internal_124(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
1311
{
1312
1313
    int ret;

1314
    ret = getfile_internal_1x(fd, job, file, 140, 14);
1315
1316
1317
1318
1319
1320
    if(ret <= 0)
        return(ret);

    shift_missing_1_24(file);

    return(1);
1321
1322
1323
1324
1325
1326
1327
}

static int getfile_internal_122(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;

1328
1329
    ret = getfile_internal_1x(fd, job, file, 138, 14);
    if(ret <= 0)
1330
1331
1332
1333
        return(ret);

    shift_missing_1_22(file);

1334
    return(1);
1335
1336
1337
1338
1339
1340
1341
}

static int getfile_internal_121(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;

1342
1343
    ret = getfile_internal_1x(fd, job, file, 133, 12);
    if(ret <= 0)
1344
1345
1346
1347
        return(ret);

    shift_missing_1_21(file);

1348
    return(1);
1349
1350
1351
1352
}

static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters)
1353
{
1354
1355
1356
1357
1358
1359
1360
1361
1362
    char* buffer;
    int ret;
    const char* err_string;
    int i;
    uint64_t hash;
    int32_t rank;
    int64_t* counters;
    double* fcounters;
    char* name_suffix;
1363
    int FILE_SIZE_1x = (32 + n_counters*8 + n_fcounters*8);
1364

1365
1366
1367
1368
1369
    memset(file, 0, sizeof(*file));

    /* set file pointer if this is the first file record; otherwise pick up
     * where we left off last time
     */
1370
1371
1372
1373
1374
1375
1376
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
    
1377
    /* space for file struct, int64 array, and double array */
1378
    buffer = (char*)malloc(FILE_SIZE_1x);
1379
1380
1381
1382
1383
    if(!buffer)
    {
        return(-1);
    }

1384
    ret = darshan_log_read(fd, buffer, FILE_SIZE_1x);
1385

1386
    if(ret > 0 && ret < FILE_SIZE_1x)
1387
1388
1389
1390
1391
1392
    {
        /* got a short read */
        fprintf(stderr, "Error: invalid file record (too small)\n");
        free(buffer);
        return(-1);
    }
1393
    else if(ret == 0)
1394
1395
1396
1397
1398
1399
1400
1401
    {
        /* hit end of file */
        free(buffer);
        return(0);
    }
    else if(ret <= 0)
    {
        /* all other errors */
1402
        err_string = darshan_log_error(fd, &ret);
1403
1404
1405
1406
1407
1408
1409
1410
1411
        fprintf(stderr, "Error: %s\n", err_string);
        free(buffer);
        return(-1);
    }

    /* got exactly one, correct size record */
    hash = *((int64_t*)&buffer[0]);
    rank = *((int32_t*)&buffer[8]);
    counters = ((int64_t*)&buffer[16]);
1412
1413
    fcounters = ((double*)&buffer[16 + n_counters*8]);
    name_suffix = &buffer[16 + n_counters*8 + n_fcounters*8];
1414
1415
1416
1417
1418
1419
1420


    if(fd->swap_flag)
    {
        /* swap bytes if necessary */
        DARSHAN_BSWAP64(&hash);
        DARSHAN_BSWAP32(&rank);
1421
        for(i=0; i<n_counters; i++)
1422
            DARSHAN_BSWAP64(&counters[i]);
1423
        for(i=0; i<n_fcounters; i++)
1424
1425
1426
1427
1428
1429
            DARSHAN_BSWAP64(&fcounters[i]);
    }

    /* assign into new format */
    file->hash = hash;
    file->rank += rank;</