darshan-logutils.c 40.9 KB
Newer Older
1
2
3
4
5
/*
 *  (C) 2009 by Argonne National Laboratory.
 *      See COPYRIGHT in top-level directory.
 */

Philip Carns's avatar
Philip Carns committed
6
#include "darshan-config.h"
7
8
#include <stdio.h>
#include <string.h>
9
#include <assert.h>
10
#include <stdlib.h>
11
#include <unistd.h>
12
#include <inttypes.h>
13
14
15
16
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
#include "darshan-logutils.h"

/* 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",
44
45
46
    "CP_INDEP_NC_OPENS",
    "CP_COLL_NC_OPENS",
    "CP_HDF5_OPENS",
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
    "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 */
81
82
    "CP_MAX_READ_TIME_SIZE",
    "CP_MAX_WRITE_TIME_SIZE",
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
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
    "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",
159
    "CP_DEVICE",
160
    "CP_SIZE_AT_OPEN",
161
162
163
164
    "CP_FASTEST_RANK",
    "CP_FASTEST_RANK_BYTES",
    "CP_SLOWEST_RANK",
    "CP_SLOWEST_RANK_BYTES",
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182

    "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 */
183
184
    "CP_F_MAX_READ_TIME",
    "CP_F_MAX_WRITE_TIME",
185
186
187
188
189
190
    "CP_F_FASTEST_RANK_TIME",
    "CP_F_SLOWEST_RANK_TIME",
    "CP_F_VARIANCE_RANK_TIME",
    "CP_F_VARIANCE_RANK_BYTES",

    "CP_F_NUM_INDICES"
191
192
};

193
194
195
196
197
198
199
/* 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);
200
#define JOB_SIZE_124 28
201
#define JOB_SIZE_200 56
202

203
/* internal routines for parsing different file versions */
204
static int getjob_internal_201(darshan_fd file, struct darshan_job *job);
205
206
207
208
209
210
211
212
213
214
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);
215
216
static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters);
217
static void shift_missing_1_24(struct darshan_file* file);
218
219
static void shift_missing_1_22(struct darshan_file* file);
static void shift_missing_1_21(struct darshan_file* file);
220

221
222
static int darshan_log_seek(darshan_fd fd, int64_t offset);
static int darshan_log_read(darshan_fd fd, void* buf, int len);
223
static int darshan_log_write(darshan_fd fd, void* buf, int len);
224
225
static const char* darshan_log_error(darshan_fd fd, int* errnum);

226
/* a rather crude API for accessing raw binary darshan files */
227
darshan_fd darshan_log_open(const char *name, const char* mode)
228
{
229
230
231
232
233
234
    int test_fd;
    uint8_t magic[2];
    int ret;
    int try_bz2 = 1;
    int len = strlen(name);

235
236
237
238
    /* we only allows "w" or "r" modes, nothing fancy */
    assert(strlen(mode) == 1);
    assert(mode[0] == 'r' || mode[0] == 'w');

239
240
241
242
243
    darshan_fd tmp_fd = malloc(sizeof(*tmp_fd));
    if(!tmp_fd)
        return(NULL);
    memset(tmp_fd, 0, sizeof(*tmp_fd));

244
245
246
247
248
249
250
251
252
    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);
    }

253
254
255
256
257
258
259
260
261
262
263
264
#ifdef HAVE_LIBBZ2
    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);
        if(!test_fd)
        {
            perror("open");
265
            free(tmp_fd->name);
266
267
268
269
270
271
272
273
            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);
274
            free(tmp_fd->name);
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
            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);
    }
    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?
         */
292
        if(len >= 3 && name[len-3] == 'b' && name[len-2] == 'z' && name[len-1] == '2')
293
294
295
296
297
298
299
300
            try_bz2 = 1;
        else
            try_bz2 = 0;
    }

    if(try_bz2)
    {
        tmp_fd->bzf = BZ2_bzopen(name, mode);
301
        if(!tmp_fd->bzf)
302
        {
303
304
305
            free(tmp_fd->name);
            free(tmp_fd);
            return(NULL);
306
        }
307
        return(tmp_fd);
308
309
310
    }
#endif

311
    tmp_fd->gzf = gzopen(name, mode);
312
313
    if(!tmp_fd->gzf)
    {
314
        free(tmp_fd->name);
315
316
317
318
        free(tmp_fd);
        tmp_fd = NULL;
    }
    return tmp_fd;
319
320
}

321
/* darshan_log_getjob()
322
323
324
 *
 * returns 0 on success, -1 on failure
 */
325
int darshan_log_getjob(darshan_fd file, struct darshan_job *job)
326
327
{
    int ret;
328

329
330
331
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
332

333
334
335
    /* read version number first so we know how to digest the rest of the
     * file
     */
336
    ret = darshan_log_read(file, file->version, 10);
337
    if(ret < 10)
338
    {
339
        fprintf(stderr, "Error: invalid log file (failed to read version).\n");
340
341
342
        return(-1);
    }

343
    if(strcmp(file->version, "2.01") == 0)
344
    {
345
        getjob_internal = getjob_internal_201;
346
        getfile_internal = getfile_internal_200;
347
        file->job_struct_size = sizeof(*job);
348
    }
349
350
351
352
353
354
    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;
    }
355
    else if(strcmp(file->version, "1.24") == 0)
356
    {
357
358
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
359
        file->job_struct_size = JOB_SIZE_124;
360
    }
361
    else if(strcmp(file->version, "1.23") == 0)
362
    {
363
364
365
        /* same as 1.24, except that mnt points may be incorrect */
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
366
        file->job_struct_size = JOB_SIZE_124;
367
    }
368
    else if(strcmp(file->version, "1.22") == 0)
369
    {
370
371
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_122;
372
        file->job_struct_size = JOB_SIZE_124;
373
    }
374
    else if(strcmp(file->version, "1.21") == 0)
375
    {
376
        getjob_internal = getjob_internal_124;
Philip Carns's avatar
Philip Carns committed
377
        getfile_internal = getfile_internal_121;
378
        file->job_struct_size = JOB_SIZE_124;
379
380
    }
    else
381
    {
382
383
384
        fprintf(stderr, "Error: incompatible darshan file.\n");
        fprintf(stderr, "Error: expected version %s, but got %s\n", 
                CP_VERSION, file->version);
385
        return(-1);
386
387
    }

388
    ret = getjob_internal(file, job);
389
390
391

    if (ret == 0)
    {
392
393
        char *metadata = strndup(job->metadata, sizeof(job->metadata));
        char *tokenize = metadata;
394
395
396
397
398
        char *kv;
        char *ptr;
        char *key;
        char *val;
        char *save;
399
        char *save2;
400

401
402
403
        for(kv=strtok_r(metadata, "\n", &save);
            kv != NULL;
            kv=strtok_r(NULL, "\n", &save))
404
        {
405
406
407
            key = strtok_r(kv, "=", &save2);
            val = strtok_r(NULL, "=", &save2);
            if (strcmp(key, "prev_ver") == 0)
408
            {
409
                strncpy(job->version_string, val, sizeof(job->version_string));
410
            }
411
412
        }
        free(metadata);
413
414
    }

415
416
    return(ret);
}
417

418
419
420
421
422
423
424
425
/* 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;
426
    char    pv_str[64];
427
428
    int     ret;

429
430
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
431
432
        return(ret);

433
434
    memset(&job_copy, 0, sizeof(job_copy));
    memcpy(&job_copy, job, sizeof(job_copy));
435
    sprintf(pv_str, "prev_ver=%s\n", job->version_string);
436
437
    sprintf(job_copy.version_string, "%s", CP_VERSION);
    strncat(job_copy.metadata, pv_str, strlen(pv_str));
438
439
    job_copy.magic_nr = CP_MAGIC_NR;

440
    ret = darshan_log_write(file, &job_copy, sizeof(job_copy));
441
442
443
444
445
446
447
448
449
    if (ret != sizeof(job_copy))
    {
        fprintf(stderr, "Error: failed to write job header: %d\n", ret);
        return(-1);
    }

    return(0);
}

450
451
452
453
454
455
456
/* 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;
457

458
459
460
    ret = getfile_internal(fd, job, file);

    return(ret);
461
462
}

463
464
465
466
467
468
469
470
/* 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;

471
472
473
474
475
476
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
477

478
    ret = darshan_log_write(fd, file, sizeof(*file));
479
480
481
482
483
484
485
486
487
    if (ret != sizeof(*file))
    {
        fprintf(stderr, "Error: writing file record failed: %d\n", ret);
        return(-1);
    }

    return(0);
}

488
489
490
491
492
493
/* 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
 */
494
int darshan_log_getmounts(darshan_fd fd, int64_t** devs, char*** mnt_pts, char***
495
    fs_types, int* count)
496
497
498
499
500
501
{
    int ret;
    char* pos;
    int array_index = 0;
    char buf[CP_EXE_LEN+1];

502
503
504
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
505

506
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
507
508
    if (ret < (CP_EXE_LEN + 1))
    {
509
        perror("darshan_log_read");
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
        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 */
529
    *devs = malloc((*count)*sizeof(int64_t));
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
    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
547
        ret = sscanf(++pos, "%" PRId64 "\t%s\t%s", &(*devs)[array_index],
548
549
            (*fs_types)[array_index], (*mnt_pts)[array_index]);

550
551
552
553
554
555
556
557
558
        if(ret != 3)
        {
            fprintf(stderr, "Error: poorly formatted mount table in log file.\n");
            return(-1);
        }
        pos--;
        *pos = '\0';
        array_index++;
    }
559

560
561
562
    return (0);
}

563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
/* 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]);
579
        ret = darshan_log_write(fd, line, strlen(line));
580
581
582
583
584
585
586
587
588
        if (ret != strlen(line))
        {
            fprintf(stderr, "Error: failed to write mount entry: %d\n", ret);
            return(-1);
        }
    }
    
    return(0);
}
589

590
int darshan_log_getexe(darshan_fd fd, char *buf)
591
592
{
    int ret;
593
    char* newline;
594

595
596
597
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
598

599
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
600
601
    if (ret < (CP_EXE_LEN + 1))
    {
602
        perror("darshan_log_read");
603
604
        return(-1);
    }
605

606
607
608
609
610
611
612
613
    /* 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';

614
615
616
    return (0);
}

617
618
619
620
621
622
623
624
625
626
627
/* 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;

628
629
    ret = darshan_log_seek(fd, sizeof(struct darshan_job));
    if(ret < 0)
630
631
632
633
        return(ret);

    len = strlen(buf);

634
    ret = darshan_log_write(fd, buf, len);
635
636
637
638
639
640
641
642
643
    if (ret != len)
    {
        fprintf(stderr, "Error: failed to write exe info: %d\n", ret);
        ret = -1;
    }

    return(ret);
}

644
void darshan_log_close(darshan_fd file)
645
{
646
647
648
649
650
651
652
653
#ifdef HAVE_LIBBZ2
    if(file->bzf)
        BZ2_bzclose(file->bzf);
#endif

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

654
    free(file->name);
655
    free(file);
656
657
}

658
659
660
661
662
663
/* 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)
{
664
665
666
667
668
669
    if(strcmp(job->version_string, "2.00") == 0)
    {
        /* current version */
        return;
    }
 
670
    if(strcmp(job->version_string, "1.24") == 0)
671
    {
672
673
674
675
676
677
678
679
680
681
        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");
        printf("# It also does not store the job id in the log file.\n");
682
683
        return;
    }
684
    
685
686
    if(strcmp(job->version_string, "1.23") == 0)
    {
687
688
689
690
691
692
693
694
695
696
        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");
        printf("# It also may have incorrect mount point mappings for files with rank > 0\n");
697
        printf("# It also does not store the job id in the log file.\n");
698
        return;
699
700
    }

701
702
703
704
    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");
705
        printf("#   CP_SIZE_AT_OPEN\n");
706
707
708
709
710
711
712
713
        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");
714
        printf("# It does not record mounted file systems, mount points, or fs types.\n");
715
716
        printf("# It attributes syncs to cumulative metadata time, rather than cumulative write time.\n");
        printf("# It also does not store the job id in the log file.\n");
717
718
        return;
    }
719
720
721
722
723
724
725
726
727

    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");
728
        printf("#   CP_DEVICE\n");
729
        printf("#   CP_SIZE_AT_OPEN\n");
730
731
        printf("#   CP_F_MAX_READ_TIME\n");
        printf("#   CP_F_MAX_WRITE_TIME\n");
732
733
734
735
736
737
738
739
        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");
740
741
        printf("# It does not record mounted file systems, mount points, or fs types.\n");
        printf("# It attributes syncs to cumulative metadata time, rather than cumulative write time.\n");
742
        printf("#\n");
743
        printf("# It also does not store the job id in the file.\n");
744
745
746
747
748
749
750
751
        return;
    }

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

752
753
754
755
/* shift_missing_1_21()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.21
756
 */
757
/*******************************
Philip Carns's avatar
Philip Carns committed
758
 * version 1.21 to 2.00 differences 
759
760
761
762
763
764
765
766
 * - 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
767
768
769
770
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
771
772
 *   - CP_F_MAX_READ_TIME
 *   - CP_F_MAX_WRITE_TIME
773
774
775
776
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
777
 * - changed params:
778
779
780
 *   - CP_FILE_RECORD_SIZE: 1184 to 1328
 *   - CP_NUM_INDICES: 133 to 144
 *   - CP_F_NUM_INDICES: 12 to 18
781
 */
782
783
784
785
786
787
788
789
790
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,
791
        CP_DEVICE,
792
        CP_SIZE_AT_OPEN,
793
794
795
796
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
797
798
799
800
        -1};
    int missing_f_counters[] = {
        CP_F_MAX_READ_TIME,
        CP_F_MAX_WRITE_TIME,
801
802
803
804
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
        -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;
}
841
842
843
844
845
846

/* shift_missing_1_22()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.22
 */
847
/*******************************
Philip Carns's avatar
Philip Carns committed
848
 * version 1.22 to 2.00 differences
849
850
851
852
 *
 * - added:
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
853
854
855
856
857
858
859
860
 *   - 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
861
 * - changed params:
862
863
864
 *   - CP_FILE_RECORD_SIZE: 1240 to 1328
 *   - CP_NUM_INDICES: 138 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
865
 */
866
867
868
869
870
static void shift_missing_1_22(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_DEVICE,
871
        CP_SIZE_AT_OPEN,
872
873
874
875
876
877
878
879
880
881
        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,
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
        -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;
    }

900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
    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;
    }

916
917
    return;
}
918

919
920
921
922
923
924
925
926
927
/* 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:
928
929
930
931
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
 *   - 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;
}

992
static int getjob_internal_201(darshan_fd file, struct darshan_job *job)
993
994
995
{
    int ret;

996
997
998
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
999

1000
    ret = darshan_log_read(file, job, sizeof(*job));
1001
1002
    if (ret < sizeof(*job))
    {
1003
        fprintf(stderr, "Error: invalid log file (too short).\n");
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
        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);
1023
        DARSHAN_BSWAP64(&job->jobid);
1024
1025
1026
1027
1028
1029
1030
1031
        return(0);
    }

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

1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049

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;

    memset(job, 0, sizeof(job_200));
    memset(job, 0, sizeof(*job));

1050
1051
1052
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
1053

1054
    ret = darshan_log_read(file, &job_200, sizeof(job_200));
1055
1056
    if (ret < sizeof(job_200))
    {
1057
        fprintf(stderr, "Error: invalid log file (too short).\n");
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
        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);
}

1094
1095
1096
1097
1098
1099
static int getfile_internal_200(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
{
    int ret;
    const char* err_string;
    int i;
1100

1101
1102
1103
1104
1105
1106
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
1107
1108
1109
1110
1111
1112

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

1113
    ret = darshan_log_read(fd, file, sizeof(*file));
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
    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);
    }

1138
    if(ret == 0)
1139
1140
1141
1142
1143
1144
    {
        /* hit end of file */
        return(0);
    }

    /* all other errors */
1145
    err_string = darshan_log_error(fd, &ret);
1146
1147
1148
1149
    fprintf(stderr, "Error: %s\n", err_string);
    return(-1);
}

1150
1151
1152
/* If we see version 1.24, assume that it is stored in big endian 32 bit
 * format.  Convert up to current format.
 */
1153
1154
static int getjob_internal_124(darshan_fd fd, struct darshan_job *job)
{
1155
1156
1157
1158
1159
1160
1161
    char* buffer;
    int ret;
    uint32_t uid;
    int32_t start_time;
    int32_t end_time;
    int32_t nprocs;

1162
1163
1164
1165
1166
1167
#ifdef WORDS_BIGENDIAN
    fd->swap_flag = 0;
#else
    fd->swap_flag = 1;
#endif

1168
1169
1170
1171
1172
1173
1174
1175
    memset(job, 0, sizeof(*job));

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

1176
1177
1178
    ret = darshan_log_seek(fd, 0);
    if(ret < 0)
        return(ret);
1179

1180
    ret = darshan_log_read(fd, buffer, JOB_SIZE_124);
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
    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);

1199
1200
1201
1202
1203
1204
1205
1206
    if(fd->swap_flag)
    {
        /* byte swap */
        DARSHAN_BSWAP32(&uid);
        DARSHAN_BSWAP32(&start_time);
        DARSHAN_BSWAP32(&end_time);
        DARSHAN_BSWAP32(&nprocs);
    }
1207
1208
1209
1210
1211

    job->uid += uid;
    job->start_time += start_time;
    job->end_time += end_time;
    job->nprocs += nprocs;
1212
    job->jobid = 0; /* old log versions did not have this field */
1213
1214
1215
1216
1217
    
    /* set magic number */
    job->magic_nr = CP_MAGIC_NR;

    return(0);
1218
1219
1220
1221
}

static int getfile_internal_124(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file)
1222
{
1223
1224
    int ret;

1225
    ret = getfile_internal_1x(fd, job, file, 140, 14);
1226
1227
1228
1229
1230
1231
    if(ret <= 0)
        return(ret);

    shift_missing_1_24(file);

    return(1);
1232
1233
1234
1235
1236
1237
1238
}

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

1239
1240
    ret = getfile_internal_1x(fd, job, file, 138, 14);
    if(ret <= 0)
1241
1242
1243
1244
        return(ret);

    shift_missing_1_22(file);

1245
    return(1);
1246
1247
1248
1249
1250
1251
1252
}

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

1253
1254
    ret = getfile_internal_1x(fd, job, file, 133, 12);
    if(ret <= 0)
1255
1256
1257
1258
        return(ret);

    shift_missing_1_21(file);

1259
    return(1);
1260
1261
1262
1263
}

static int getfile_internal_1x(darshan_fd fd, struct darshan_job *job, 
    struct darshan_file *file, int n_counters, int n_fcounters)
1264
{
1265
1266
1267
1268
1269
1270
1271
1272
1273
    char* buffer;
    int ret;
    const char* err_string;
    int i;
    uint64_t hash;
    int32_t rank;
    int64_t* counters;
    double* fcounters;
    char* name_suffix;
1274
    int FILE_SIZE_1x = (32 + n_counters*8 + n_fcounters*8);
1275

1276
1277
1278
1279
1280
    memset(file, 0, sizeof(*file));

    /* set file pointer if this is the first file record; otherwise pick up
     * where we left off last time
     */
1281
1282
1283
1284
1285
1286
1287
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
    
1288
    /* space for file struct, int64 array, and double array */
1289
    buffer = (char*)malloc(FILE_SIZE_1x);
1290
1291
1292
1293
1294
    if(!buffer)
    {
        return(-1);
    }

1295
    ret = darshan_log_read(fd, buffer, FILE_SIZE_1x);
1296

1297
    if(ret > 0 && ret < FILE_SIZE_1x)
1298
1299
1300
1301
1302
1303
    {
        /* got a short read */
        fprintf(stderr, "Error: invalid file record (too small)\n");
        free(buffer);
        return(-1);
    }
1304
    else if(ret == 0)
1305
1306
1307
1308
1309
1310
1311
1312
    {
        /* hit end of file */
        free(buffer);
        return(0);
    }
    else if(ret <= 0)
    {
        /* all other errors */
1313
        err_string = darshan_log_error(fd, &ret);
1314
1315
1316
1317
1318
1319
1320
1321
1322
        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]);
1323
1324
    fcounters = ((double*)&buffer[16 + n_counters*8]);
    name_suffix = &buffer[16 + n_counters*8 + n_fcounters*8];
1325
1326
1327
1328
1329
1330
1331


    if(fd->swap_flag)
    {
        /* swap bytes if necessary */
        DARSHAN_BSWAP64(&hash);
        DARSHAN_BSWAP32(&rank);
1332
        for(i=0; i<n_counters; i++)
1333
            DARSHAN_BSWAP64(&counters[i]);
1334
        for(i=0; i<n_fcounters; i++)
1335
1336
1337
1338
1339
1340
            DARSHAN_BSWAP64(&fcounters[i]);
    }

    /* assign into new format */
    file->hash = hash;
    file->rank += rank;
1341
1342
    memcpy(file->counters, counters, n_counters*8);
    memcpy(file->fcounters, fcounters, n_fcounters*8);
1343
1344
1345
1346
    memcpy(file->name_suffix, name_suffix, 12);

    free(buffer);
    return(1);
1347
1348
}

1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
/* return amount written on success, -1 on failure.
 */
static int darshan_log_write(darshan_fd fd, void* buf, int len)
{
    int ret;

    if(fd->gzf)
    {
        ret = gzwrite(fd, buf, len);
        if(ret > 0)
            fd->pos += ret;
        return(ret);
    }

#ifdef HAVE_LIBBZ2
    if(fd->bzf)
    {
        ret = BZ2_bzwrite(fd->bzf, buf, len);
        if(ret > 0)
            fd->pos += ret;
        return(ret);
    }
#endif

    return(-1);
}


1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
/* return amount read on success, 0 on EOF, -1 on failure.
 */
static int darshan_log_read(darshan_fd fd, void* buf, int len)
{
    int ret;

    if(fd->gzf)
    {
        ret = gzread(fd->gzf, buf, len);
        if(ret > 0)
            fd->pos += ret;
        return(ret);
    }

#ifdef HAVE_LIBBZ2
    if(fd->bzf)
    {
        ret = BZ2_bzread(fd->bzf, buf, len);
        if(ret > 0)
            fd->pos += ret;
        return(ret);
    }
#endif

    return(-1);
}


static const char* darshan_log_error(darshan_fd fd, int* errnum)
{
    if(fd->gzf)
    {
        return(gzerror(fd->gzf, errnum));
    }

#ifdef HAVE_LIBBZ2
    if(fd->bzf)
    {
        return(BZ2_bzerror(fd->bzf, errnum));
    }
#endif

    *errnum = 0;
    return(NULL);
}

/* return 0 on successful seek to offset, -1 on failure.
 */
static int darshan_log_seek(darshan_fd fd, int64_t offset)
{
    z_off_t zoff = 0;
    z_off_t zoff_ret = 0;

    if(fd->pos == offset)
        return(0);

    if(fd->gzf)
    {
        zoff += offset;
        zoff_ret = gzseek(fd->gzf, zoff, SEEK_SET);
        if(zoff_ret == zoff)
        {
            fd->pos = offset;
            return(0);
        }
        return(-1);
    }

#ifdef HAVE_LIBBZ2
    if(fd->bzf)
    {
        int64_t counter;
        char dummy = '\0';
        int ret;

        /* There is no seek in bzip2.  Just close, reopen, and throw away 
         * data until the correct offset.  Very slow, but we don't expect to
         * do this often.
         */
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
        if(fd->mode[0] == 'r' && offset < fd->pos)
        {
            /* to seek backwards in read-only mode we just close and re-open
             * the file
             */
            BZ2_bzclose(fd->bzf);
            fd->bzf = BZ2_bzopen(fd->name, fd->mode);
            if(!fd->bzf)
                return(-1);

            fd->pos = 0;
        }
        else if(fd->mode[0] == 'w' && offset < fd->pos)
        {
            /* there isn't any convenient way to seek backwards in a
             * write-only bzip2 file, but we shouldn't need that
             * functionality in darshan anyway.
             */
            fprintf(stderr, "Error: seeking backwards in a bzip2 compressed darshan output file is not supported.\n");
1475
            return(-1);
1476
        }
1477

1478
        for(counter=0; counter<(offset-fd->pos); counter++)
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
        {
            if(fd->mode[0] == 'r')
            {
                ret = BZ2_bzread(fd->bzf, &dummy, 1);
            }
            else
            {
                ret = BZ2_bzwrite(fd->bzf, &dummy, 1);
            }
            if(ret != 1)
                return(-1);
        }
1491
        fd->pos += counter;
1492
1493
1494
1495
1496
1497
1498
        return(0);
    }
#endif

    return(-1);
}