darshan-logutils.c 46.5 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_203(darshan_fd file, struct darshan_job *job);
225
static int getjob_internal_201(darshan_fd file, struct darshan_job *job);
226
227
228
229
230
231
232
233
234
235
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);
236
237
static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters);
238
static void shift_missing_1_24(struct darshan_file* file);
239
240
static void shift_missing_1_22(struct darshan_file* file);
static void shift_missing_1_21(struct darshan_file* file);
241

242
243
static int darshan_log_seek(darshan_fd fd, int64_t offset);
static int darshan_log_read(darshan_fd fd, void* buf, int len);
244
static int darshan_log_write(darshan_fd fd, void* buf, int len);
245
246
static const char* darshan_log_error(darshan_fd fd, int* errnum);

247
/* a rather crude API for accessing raw binary darshan files */
248
darshan_fd darshan_log_open(const char *name, const char* mode)
249
{
250
251
252
253
254
255
    int test_fd;
    uint8_t magic[2];
    int ret;
    int try_bz2 = 1;
    int len = strlen(name);

256
257
258
259
    /* we only allows "w" or "r" modes, nothing fancy */
    assert(strlen(mode) == 1);
    assert(mode[0] == 'r' || mode[0] == 'w');

260
261
262
263
264
    darshan_fd tmp_fd = malloc(sizeof(*tmp_fd));
    if(!tmp_fd)
        return(NULL);
    memset(tmp_fd, 0, sizeof(*tmp_fd));

265
266
267
268
269
270
271
272
273
    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);
    }

274
275
276
277
278
279
280
281
    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);
282
        if(test_fd < 0)
283
284
        {
            perror("open");
285
            free(tmp_fd->name);
286
287
288
289
290
291
292
293
            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);
294
            free(tmp_fd->name);
295
296
297
298
299
300
301
302
303
304
305
            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);
    }
306

307
308
309
310
311
312
    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?
         */
313
        if(len >= 3 && name[len-3] == 'b' && name[len-2] == 'z' && name[len-1] == '2')
314
315
316
317
318
            try_bz2 = 1;
        else
            try_bz2 = 0;
    }

319
#ifdef HAVE_LIBBZ2
320
321
322
    if(try_bz2)
    {
        tmp_fd->bzf = BZ2_bzopen(name, mode);
323
        if(!tmp_fd->bzf)
324
        {
325
326
327
            free(tmp_fd->name);
            free(tmp_fd);
            return(NULL);
328
        }
329
        return(tmp_fd);
330
    }
331
332
333
334
335
336
337
#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);
    }
338
339
#endif

340
    tmp_fd->gzf = gzopen(name, mode);
341
342
    if(!tmp_fd->gzf)
    {
343
        free(tmp_fd->name);
344
345
346
347
        free(tmp_fd);
        tmp_fd = NULL;
    }
    return tmp_fd;
348
349
}

350
/* darshan_log_getjob()
351
352
353
 *
 * returns 0 on success, -1 on failure
 */
354
int darshan_log_getjob(darshan_fd file, struct darshan_job *job)
355
356
{
    int ret;
357
    char buffer[DARSHAN_JOB_METADATA_LEN];
358

359
360
361
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
362

363
364
365
    /* read version number first so we know how to digest the rest of the
     * file
     */
366
    ret = darshan_log_read(file, file->version, 10);
367
    if(ret < 10)
368
    {
369
        fprintf(stderr, "Error: invalid log file (failed to read version).\n");
370
371
372
        return(-1);
    }

373
374
375
376
377
378
379
    if(strcmp(file->version, "2.03") == 0)
    {
        getjob_internal = getjob_internal_203;
        getfile_internal = getfile_internal_200;
        file->job_struct_size = sizeof(*job);
    }
    else if(strcmp(file->version, "2.02") == 0)
Philip Carns's avatar
Philip Carns committed
380
381
382
383
384
385
    {
        getjob_internal = getjob_internal_201;
        getfile_internal = getfile_internal_200;
        file->job_struct_size = sizeof(*job);
    }
    else if(strcmp(file->version, "2.01") == 0)
386
    {
387
        getjob_internal = getjob_internal_201;
388
        getfile_internal = getfile_internal_200;
389
        file->job_struct_size = sizeof(*job);
390
    }
391
392
393
394
395
396
    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;
    }
397
    else if(strcmp(file->version, "1.24") == 0)
398
    {
399
400
        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, "1.23") == 0)
404
    {
405
406
407
        /* same as 1.24, except that mnt points may be incorrect */
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
408
        file->job_struct_size = JOB_SIZE_124;
409
    }
410
    else if(strcmp(file->version, "1.22") == 0)
411
    {
412
413
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_122;
414
        file->job_struct_size = JOB_SIZE_124;
415
    }
416
    else if(strcmp(file->version, "1.21") == 0)
417
    {
418
        getjob_internal = getjob_internal_124;
Philip Carns's avatar
Philip Carns committed
419
        getfile_internal = getfile_internal_121;
420
        file->job_struct_size = JOB_SIZE_124;
421
422
    }
    else
423
    {
424
425
426
        fprintf(stderr, "Error: incompatible darshan file.\n");
        fprintf(stderr, "Error: expected version %s, but got %s\n", 
                CP_VERSION, file->version);
427
        return(-1);
428
429
    }

430
    ret = getjob_internal(file, job);
431
432
433

    if (ret == 0)
    {
434
#ifdef HAVE_STRNDUP
435
        char *metadata = strndup(job->metadata, sizeof(job->metadata));
436
#else
Philip Carns's avatar
bug fix    
Philip Carns committed
437
        char *metadata = strdup(job->metadata);
438
#endif
439
440
        char *kv;
        char *key;
441
        char *value;
442
443
        char *save;

444
445
446
        for(kv=strtok_r(metadata, "\n", &save);
            kv != NULL;
            kv=strtok_r(NULL, "\n", &save))
447
        {
448
449
450
451
452
453
454
455
456
457
458
459
            /* 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++;
460
            if (strcmp(key, "prev_ver") == 0)
461
            {
462
                strncpy(job->version_string, value, sizeof(job->version_string));
463
            }
464
465
        }
        free(metadata);
466
467
    }

468
469
    return(ret);
}
470

471
472
473
474
475
476
477
478
/* 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;
479
    char    pv_str[64];
480
    int     ret;
481
    int len;
482

483
484
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
485
486
        return(ret);

487
488
    memset(&job_copy, 0, sizeof(job_copy));
    memcpy(&job_copy, job, sizeof(job_copy));
489
490
491
492
493
494
495
496
497
498
499
    /* 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';
        }
    }

500
    sprintf(pv_str, "prev_ver=%s\n", job->version_string);
501
    sprintf(job_copy.version_string, "%s", CP_VERSION);
Philip Carns's avatar
Philip Carns committed
502
503
504
505
    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);
506
507
    job_copy.magic_nr = CP_MAGIC_NR;

508
    ret = darshan_log_write(file, &job_copy, sizeof(job_copy));
509
510
511
512
513
514
515
516
517
    if (ret != sizeof(job_copy))
    {
        fprintf(stderr, "Error: failed to write job header: %d\n", ret);
        return(-1);
    }

    return(0);
}

518
519
520
521
522
523
524
/* 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;
525

526
527
528
    ret = getfile_internal(fd, job, file);

    return(ret);
529
530
}

531
532
533
534
535
536
537
538
/* 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;

539
540
541
542
543
544
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
545

546
    ret = darshan_log_write(fd, file, sizeof(*file));
547
548
549
550
551
552
553
554
555
    if (ret != sizeof(*file))
    {
        fprintf(stderr, "Error: writing file record failed: %d\n", ret);
        return(-1);
    }

    return(0);
}

556
557
558
559
560
561
/* 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
 */
562
int darshan_log_getmounts(darshan_fd fd, int64_t** devs, char*** mnt_pts, char***
563
    fs_types, int* count)
564
565
566
567
568
569
{
    int ret;
    char* pos;
    int array_index = 0;
    char buf[CP_EXE_LEN+1];

570
571
572
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
573

574
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
575
576
    if (ret < (CP_EXE_LEN + 1))
    {
577
        perror("darshan_log_read");
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
        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 */
597
    *devs = malloc((*count)*sizeof(int64_t));
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
    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
615
        ret = sscanf(++pos, "%" PRId64 "\t%s\t%s", &(*devs)[array_index],
616
617
            (*fs_types)[array_index], (*mnt_pts)[array_index]);

618
619
620
621
622
623
624
625
626
        if(ret != 3)
        {
            fprintf(stderr, "Error: poorly formatted mount table in log file.\n");
            return(-1);
        }
        pos--;
        *pos = '\0';
        array_index++;
    }
627

628
629
630
    return (0);
}

631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
/* 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]);
647
        ret = darshan_log_write(fd, line, strlen(line));
648
649
650
651
652
653
        if (ret != strlen(line))
        {
            fprintf(stderr, "Error: failed to write mount entry: %d\n", ret);
            return(-1);
        }
    }
654
655
656
657
658
659
660
661

    /* 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);
    }

662
663
    return(0);
}
664

665
int darshan_log_getexe(darshan_fd fd, char *buf)
666
667
{
    int ret;
668
    char* newline;
669

670
671
672
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
673

674
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
675
676
    if (ret < (CP_EXE_LEN + 1))
    {
677
        perror("darshan_log_read");
678
679
        return(-1);
    }
680

681
682
683
684
685
686
687
688
    /* 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';

689
690
691
    return (0);
}

692
693
694
695
696
697
698
699
700
701
702
/* 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;

703
704
    ret = darshan_log_seek(fd, sizeof(struct darshan_job));
    if(ret < 0)
705
706
707
708
        return(ret);

    len = strlen(buf);

709
    ret = darshan_log_write(fd, buf, len);
710
711
712
713
714
715
716
717
718
    if (ret != len)
    {
        fprintf(stderr, "Error: failed to write exe info: %d\n", ret);
        ret = -1;
    }

    return(ret);
}

719
void darshan_log_close(darshan_fd file)
720
{
721
722
723
724
725
726
727
728
#ifdef HAVE_LIBBZ2
    if(file->bzf)
        BZ2_bzclose(file->bzf);
#endif

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

729
    free(file->name);
730
    free(file);
731
732
}

733
734
735
736
737
738
/* 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)
{
739
    if(strcmp(job->version_string, "2.03") == 0)
740
741
742
743
    {
        /* current version */
        return;
    }
744

745
746
747
748
749
750
751
    if(strcmp(job->version_string, "2.02") == 0)
    {
        printf("# WARNING: version 2.02 log format does not support the following parameters:\n");
        printf("#   wtime_offset (in job header).\n");
        return;
    }

Philip Carns's avatar
Philip Carns committed
752
753
    if(strcmp(job->version_string, "2.01") == 0)
    {
754
755
        printf("# WARNING: version 2.01 log format does not support the following parameters:\n");
        printf("#   wtime_offset (in job header).\n");
Philip Carns's avatar
Philip Carns committed
756
757
758
759
760
        printf("# WARNING: version 2.01 log format has the following limitations:\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
        return;
    }

761
762
    if(strcmp(job->version_string, "2.00") == 0)
    {
763
764
        printf("# WARNING: version 2.00 log format does not support the following parameters:\n");
        printf("#   wtime_offset (in job header).\n");
Philip Carns's avatar
Philip Carns committed
765
766
        printf("# WARNING: version 2.00 log format has the following limitations:\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
767
768
        return;
    }
769
 
770
    if(strcmp(job->version_string, "1.24") == 0)
771
    {
772
773
774
775
776
777
778
779
780
        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");
781
        printf("#   wtime_offset (in job header).\n");
Philip Carns's avatar
Philip Carns committed
782
783
784
        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");
785
786
        return;
    }
787
    
788
789
    if(strcmp(job->version_string, "1.23") == 0)
    {
790
791
792
793
794
795
796
797
798
        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");
799
        printf("#   wtime_offset (in job header).\n");
Philip Carns's avatar
Philip Carns committed
800
801
802
803
        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");
804
        return;
805
806
    }

807
808
809
810
    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");
811
        printf("#   CP_SIZE_AT_OPEN\n");
812
813
814
815
816
817
818
819
        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");
820
        printf("#   wtime_offset (in job header).\n");
Philip Carns's avatar
Philip Carns committed
821
822
823
824
825
        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");
826
827
        return;
    }
828
829
830
831
832
833
834
835
836

    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");
837
        printf("#   CP_DEVICE\n");
838
        printf("#   CP_SIZE_AT_OPEN\n");
839
840
        printf("#   CP_F_MAX_READ_TIME\n");
        printf("#   CP_F_MAX_WRITE_TIME\n");
841
842
843
844
845
846
847
848
        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");
849
        printf("#   wtime_offset (in job header).\n");
Philip Carns's avatar
Philip Carns committed
850
851
852
853
854
        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");
855
856
857
858
859
860
861
862
        return;
    }

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

863
864
865
866
/* shift_missing_1_21()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.21
867
 */
868
/*******************************
Philip Carns's avatar
Philip Carns committed
869
 * version 1.21 to 2.00 differences 
870
871
872
873
874
875
876
877
 * - 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
878
879
880
881
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
882
883
 *   - CP_F_MAX_READ_TIME
 *   - CP_F_MAX_WRITE_TIME
884
885
886
887
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
888
 * - changed params:
889
890
891
 *   - CP_FILE_RECORD_SIZE: 1184 to 1328
 *   - CP_NUM_INDICES: 133 to 144
 *   - CP_F_NUM_INDICES: 12 to 18
892
 */
893
894
895
896
897
898
899
900
901
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,
902
        CP_DEVICE,
903
        CP_SIZE_AT_OPEN,
904
905
906
907
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
908
909
910
911
        -1};
    int missing_f_counters[] = {
        CP_F_MAX_READ_TIME,
        CP_F_MAX_WRITE_TIME,
912
913
914
915
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
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
        -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;
}
952
953
954
955
956
957

/* shift_missing_1_22()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.22
 */
958
/*******************************
Philip Carns's avatar
Philip Carns committed
959
 * version 1.22 to 2.00 differences
960
961
962
963
 *
 * - added:
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
964
965
966
967
968
969
970
971
 *   - 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
972
 * - changed params:
973
974
975
 *   - CP_FILE_RECORD_SIZE: 1240 to 1328
 *   - CP_NUM_INDICES: 138 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
976
 */
977
978
979
980
981
static void shift_missing_1_22(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_DEVICE,
982
        CP_SIZE_AT_OPEN,
983
984
985
986
987
988
989
990
991
992
        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,
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
        -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;
    }

1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
    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;
    }

1027
1028
    return;
}
1029

1030
1031
1032
1033
1034
1035
1036
1037
1038
/* 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:
1039
1040
1041
1042
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
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
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
 *   - 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;
}

1103
static int getjob_internal_203(darshan_fd file, struct darshan_job *job)
1104
1105
1106
{
    int ret;

1107
1108
1109
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
1110

1111
    ret = darshan_log_read(file, job, sizeof(*job));
1112
1113
    if (ret < sizeof(*job))
    {
1114
        fprintf(stderr, "Error: invalid log file (too short).\n");
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
        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);
    }

1125
1126
1127
1128
1129
1130
1131
1132
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
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
    /* 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);
        DARSHAN_BSWAP64(&job->wtime_offset);
        return(0);
    }

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

static int getjob_internal_201(darshan_fd file, struct darshan_job *job)
{
    int ret;
    struct darshan_job_201
    {
        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;
        char metadata[DARSHAN_JOB_METADATA_LEN];
    } job_201;

    memset(&job_201, 0, sizeof(job_201));
    memset(job, 0, sizeof(*job));

    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);

    ret = darshan_log_read(file, &job_201, sizeof(job_201));
    if (ret < sizeof(job_201))
    {
        fprintf(stderr, "Error: invalid log file (too short).\n");
        return(-1);
    }

    memcpy(job->version_string, job_201.version_string, 8);
    job->magic_nr   = job_201.magic_nr;
    job->uid        = job_201.uid;
    job->start_time = job_201.start_time;
    job->end_time   = job_201.end_time;
    job->nprocs     = job_201.nprocs;
    job->jobid      = job_201.jobid;
    memcpy(job->metadata, job_201.metadata, DARSHAN_JOB_METADATA_LEN);

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

1189
1190
1191
1192
1193
1194
1195
1196
1197
    /* 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);
1198
        DARSHAN_BSWAP64(&job->jobid);
1199
1200
1201
1202
1203
1204
1205
1206
        return(0);
    }

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

1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221

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;

1222
    memset(&job_200, 0, sizeof(job_200));
1223
1224
    memset(job, 0, sizeof(*job));

1225
1226
1227
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
1228

1229
    ret = darshan_log_read(file, &job_200, sizeof(job_200));
1230
1231
    if (ret < sizeof(job_200))
    {
1232
        fprintf(stderr, "Error: invalid log file (too short).\n");
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
        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);
}

1269
1270
1271
1272
1273
1274
static int getfile_internal_200(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;
    const char* err_string;
    int i;
1275

1276
1277
1278
1279
1280
1281
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
1282
1283
1284
1285
1286
1287

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

1288
    ret = darshan_log_read(fd, file, sizeof(*file));
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
    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);
    }

1313
    if(ret == 0)
1314
1315
1316
1317
1318
1319
    {
        /* hit end of file */
        return(0);
    }

    /* all other errors */
1320
    err_string = darshan_log_error(fd, &ret);
1321
1322
1323
1324
    fprintf(stderr, "Error: %s\n", err_string);
    return(-1);
}

1325
1326
1327
/* If we see version 1.24, assume that it is stored in big endian 32 bit
 * format.  Convert up to current format.
 */
1328
1329
static int getjob_internal_124(darshan_fd fd, struct darshan_job *job)
{
1330
1331
1332
1333
1334
1335
1336
    char* buffer;
    int ret;
    uint32_t uid;
    int32_t start_time;
    int32_t end_time;
    int32_t nprocs;

1337
1338
1339
1340
1341
1342
#ifdef WORDS_BIGENDIAN
    fd->swap_flag = 0;
#else
    fd->swap_flag = 1;
#endif

1343
1344
1345
1346
1347
1348
1349
1350
    memset(job, 0, sizeof(*job));

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

1351
1352
1353
    ret = darshan_log_seek(fd, 0);
    if(ret < 0)
        return(ret);
1354

1355
    ret = darshan_log_read(fd, buffer, JOB_SIZE_124);
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
    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);

1374
1375
1376
1377
1378
1379
1380
1381
    if(fd->swap_flag)
    {
        /* byte swap */
        DARSHAN_BSWAP32(&uid);
        DARSHAN_BSWAP32(&start_time);
        DARSHAN_BSWAP32(&end_time);
        DARSHAN_BSWAP32(&nprocs);
    }
1382
1383
1384
1385
1386

    job->uid += uid;
    job->start_time += start_time;
    job->end_time += end_time;
    job->nprocs += nprocs;
1387
    job->jobid = 0; /* old log versions did not have this field */
1388
1389
1390
1391
1392
    
    /* set magic number */
    job->magic_nr = CP_MAGIC_NR;

    return(0);
1393
1394
1395
1396
}

static int getfile_internal_124(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
1397
{
1398
1399
    int ret;

1400
    ret = getfile_internal_1x(fd, job, file, 140, 14);
1401
1402
1403
1404
1405
1406
    if(ret <= 0)
        return(ret);

    shift_missing_1_24(file);

    return(1);
1407
1408
1409
1410
1411
1412
1413
}

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

1414
1415
    ret = getfile_internal_1x(fd, job, file, 138, 14);
    if(ret <= 0)
1416
1417
1418
1419
        return(ret);

    shift_missing_1_22(file);

1420
    return(1);
1421
1422
1423
1424
1425
1426
1427
}

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

1428
1429
    ret = getfile_internal_1x(fd, job, file, 133, 12);
    if(ret <= 0)
1430
1431
1432
1433
        return(ret);

    shift_missing_1_21(file);

1434
    return(1);
1435
1436
1437
1438
}

static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters)
1439
{
1440
1441
1442
1443