darshan-logutils.c 44.8 KB
Newer Older
1 2 3 4 5
/*
 *  (C) 2009 by Argonne National Laboratory.
 *      See COPYRIGHT in top-level directory.
 */

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

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

23 24
#include "darshan-logutils.h"

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

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

    "CP_NUM_INDICES"
};

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

    "CP_F_NUM_INDICES"
211 212
};

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

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

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

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

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

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

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

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

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

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

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

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

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

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

372 373 374 375 376 377 378
    if(strcmp(file->version, "2.03") == 0)
    {
        getjob_internal = getjob_internal_201;
        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
379 380 381 382 383 384
    {
        getjob_internal = getjob_internal_201;
        getfile_internal = getfile_internal_200;
        file->job_struct_size = sizeof(*job);
    }
    else if(strcmp(file->version, "2.01") == 0)
385
    {
386
        getjob_internal = getjob_internal_201;
387
        getfile_internal = getfile_internal_200;
388
        file->job_struct_size = sizeof(*job);
389
    }
390 391 392 393 394 395
    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;
    }
396
    else if(strcmp(file->version, "1.24") == 0)
397
    {
398 399
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
400
        file->job_struct_size = JOB_SIZE_124;
401
    }
402
    else if(strcmp(file->version, "1.23") == 0)
403
    {
404 405 406
        /* same as 1.24, except that mnt points may be incorrect */
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_124;
407
        file->job_struct_size = JOB_SIZE_124;
408
    }
409
    else if(strcmp(file->version, "1.22") == 0)
410
    {
411 412
        getjob_internal = getjob_internal_124;
        getfile_internal = getfile_internal_122;
413
        file->job_struct_size = JOB_SIZE_124;
414
    }
415
    else if(strcmp(file->version, "1.21") == 0)
416
    {
417
        getjob_internal = getjob_internal_124;
Philip Carns's avatar
Philip Carns committed
418
        getfile_internal = getfile_internal_121;
419
        file->job_struct_size = JOB_SIZE_124;
420 421
    }
    else
422
    {
423 424 425
        fprintf(stderr, "Error: incompatible darshan file.\n");
        fprintf(stderr, "Error: expected version %s, but got %s\n", 
                CP_VERSION, file->version);
426
        return(-1);
427 428
    }

429
    ret = getjob_internal(file, job);
430 431 432

    if (ret == 0)
    {
433
#ifdef HAVE_STRNDUP
434
        char *metadata = strndup(job->metadata, sizeof(job->metadata));
435
#else
Philip Carns's avatar
bug fix  
Philip Carns committed
436
        char *metadata = strdup(job->metadata);
437
#endif
438 439
        char *kv;
        char *key;
440
        char *value;
441 442
        char *save;

443 444 445
        for(kv=strtok_r(metadata, "\n", &save);
            kv != NULL;
            kv=strtok_r(NULL, "\n", &save))
446
        {
447 448 449 450 451 452 453 454 455 456 457 458
            /* 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++;
459
            if (strcmp(key, "prev_ver") == 0)
460
            {
461
                strncpy(job->version_string, value, sizeof(job->version_string));
462
            }
463 464
        }
        free(metadata);
465 466
    }

467 468
    return(ret);
}
469

470 471 472 473 474 475 476 477
/* 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;
478
    char    pv_str[64];
479
    int     ret;
480
    int len;
481

482 483
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
484 485
        return(ret);

486 487
    memset(&job_copy, 0, sizeof(job_copy));
    memcpy(&job_copy, job, sizeof(job_copy));
488 489 490 491 492 493 494 495 496 497 498
    /* 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';
        }
    }

499
    sprintf(pv_str, "prev_ver=%s\n", job->version_string);
500
    sprintf(job_copy.version_string, "%s", CP_VERSION);
Philip Carns's avatar
Philip Carns committed
501 502 503 504
    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);
505 506
    job_copy.magic_nr = CP_MAGIC_NR;

507
    ret = darshan_log_write(file, &job_copy, sizeof(job_copy));
508 509 510 511 512 513 514 515 516
    if (ret != sizeof(job_copy))
    {
        fprintf(stderr, "Error: failed to write job header: %d\n", ret);
        return(-1);
    }

    return(0);
}

517 518 519 520 521 522 523
/* 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;
524

525 526 527
    ret = getfile_internal(fd, job, file);

    return(ret);
528 529
}

530 531 532 533 534 535 536 537
/* 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;

538 539 540 541 542 543
    if(fd->pos < CP_JOB_RECORD_SIZE)
    {
        ret = darshan_log_seek(fd, CP_JOB_RECORD_SIZE);
        if(ret < 0)
            return(ret);
    }
544

545
    ret = darshan_log_write(fd, file, sizeof(*file));
546 547 548 549 550 551 552 553 554
    if (ret != sizeof(*file))
    {
        fprintf(stderr, "Error: writing file record failed: %d\n", ret);
        return(-1);
    }

    return(0);
}

555 556 557 558 559 560
/* 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
 */
561
int darshan_log_getmounts(darshan_fd fd, int64_t** devs, char*** mnt_pts, char***
562
    fs_types, int* count)
563 564 565 566 567 568
{
    int ret;
    char* pos;
    int array_index = 0;
    char buf[CP_EXE_LEN+1];

569 570 571
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
572

573
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
574 575
    if (ret < (CP_EXE_LEN + 1))
    {
576
        perror("darshan_log_read");
577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
        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 */
596
    *devs = malloc((*count)*sizeof(int64_t));
597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
    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
614
        ret = sscanf(++pos, "%" PRId64 "\t%s\t%s", &(*devs)[array_index],
615 616
            (*fs_types)[array_index], (*mnt_pts)[array_index]);

617 618 619 620 621 622 623 624 625
        if(ret != 3)
        {
            fprintf(stderr, "Error: poorly formatted mount table in log file.\n");
            return(-1);
        }
        pos--;
        *pos = '\0';
        array_index++;
    }
626

627 628 629
    return (0);
}

630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
/* 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]);
646
        ret = darshan_log_write(fd, line, strlen(line));
647 648 649 650 651 652
        if (ret != strlen(line))
        {
            fprintf(stderr, "Error: failed to write mount entry: %d\n", ret);
            return(-1);
        }
    }
653 654 655 656 657 658 659 660

    /* 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);
    }

661 662
    return(0);
}
663

664
int darshan_log_getexe(darshan_fd fd, char *buf)
665 666
{
    int ret;
667
    char* newline;
668

669 670 671
    ret = darshan_log_seek(fd, fd->job_struct_size);
    if(ret < 0)
        return(ret);
672

673
    ret = darshan_log_read(fd, buf, (CP_EXE_LEN + 1));
674 675
    if (ret < (CP_EXE_LEN + 1))
    {
676
        perror("darshan_log_read");
677 678
        return(-1);
    }
679

680 681 682 683 684 685 686 687
    /* 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';

688 689 690
    return (0);
}

691 692 693 694 695 696 697 698 699 700 701
/* 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;

702 703
    ret = darshan_log_seek(fd, sizeof(struct darshan_job));
    if(ret < 0)
704 705 706 707
        return(ret);

    len = strlen(buf);

708
    ret = darshan_log_write(fd, buf, len);
709 710 711 712 713 714 715 716 717
    if (ret != len)
    {
        fprintf(stderr, "Error: failed to write exe info: %d\n", ret);
        ret = -1;
    }

    return(ret);
}

718
void darshan_log_close(darshan_fd file)
719
{
720 721 722 723 724 725 726 727
#ifdef HAVE_LIBBZ2
    if(file->bzf)
        BZ2_bzclose(file->bzf);
#endif

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

728
    free(file->name);
729
    free(file);
730 731
}

732 733 734 735 736 737
/* 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)
{
738
    if(strcmp(job->version_string, "2.03") == 0)
739
    {
740
        /* current version */
741 742 743
        return;
    }

744 745 746 747 748 749 750
    if(strcmp(job->version_string, "2.02") == 0)
    {
        printf("# WARNING: version 2.01 log format has the following limitations:\n");
        printf("# - *_TIMESTAMP fields are not normalized relative to MPI_Init() time.\n");
        return;
    }

Philip Carns's avatar
Philip Carns committed
751 752 753
    if(strcmp(job->version_string, "2.01") == 0)
    {
        printf("# WARNING: version 2.01 log format has the following limitations:\n");
754
        printf("# - *_TIMESTAMP fields are not normalized relative to MPI_Init() time.\n");
Philip Carns's avatar
Philip Carns committed
755 756 757 758
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
        return;
    }

759 760
    if(strcmp(job->version_string, "2.00") == 0)
    {
Philip Carns's avatar
Philip Carns committed
761
        printf("# WARNING: version 2.00 log format has the following limitations:\n");
762
        printf("# - *_TIMESTAMP fields are not normalized relative to MPI_Init() time.\n");
Philip Carns's avatar
Philip Carns committed
763
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
764 765
        return;
    }
766
 
767
    if(strcmp(job->version_string, "1.24") == 0)
768
    {
769 770 771 772 773 774 775 776 777
        printf("# WARNING: version 1.24 log format does not support the following parameters:\n");
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
778
        printf("# WARNING: version 1.24 log format has the following limitations:\n");
779
        printf("# - *_TIMESTAMP fields are not normalized relative to MPI_Init() time.\n");
Philip Carns's avatar
Philip Carns committed
780 781
        printf("# - does not store the job id in the file.\n");
        printf("# - inaccurate statistics in some multi-threaded cases.\n");
782 783
        return;
    }
784
    
785 786
    if(strcmp(job->version_string, "1.23") == 0)
    {
787 788 789 790 791 792 793 794 795
        printf("# WARNING: version 1.23 log format does not support the following parameters:\n");
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
796
        printf("# WARNING: version 1.23 log format has the following limitations:\n");
797
        printf("# - *_TIMESTAMP fields are not normalized relative to MPI_Init() time.\n");
Philip Carns's avatar
Philip Carns committed
798 799 800
        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");
801
        return;
802 803
    }

804 805 806 807
    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");
808
        printf("#   CP_SIZE_AT_OPEN\n");
809 810 811 812 813 814 815 816
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
817
        printf("# WARNING: version 1.22 log format has the following limitations:\n");
818
        printf("# - *_TIMESTAMP fields are not normalized relative to MPI_Init() time.\n");
Philip Carns's avatar
Philip Carns committed
819 820 821 822
        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");
823 824
        return;
    }
825 826 827 828 829 830 831 832 833

    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");
834
        printf("#   CP_DEVICE\n");
835
        printf("#   CP_SIZE_AT_OPEN\n");
836 837
        printf("#   CP_F_MAX_READ_TIME\n");
        printf("#   CP_F_MAX_WRITE_TIME\n");
838 839 840 841 842 843 844 845
        printf("#   CP_FASTEST_RANK\n");
        printf("#   CP_FASTEST_RANK_BYTES\n");
        printf("#   CP_SLOWEST_RANK\n");
        printf("#   CP_SLOWEST_RANK_BYTES\n");
        printf("#   CP_F_FASTEST_RANK_TIME\n");
        printf("#   CP_F_SLOWEST_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_TIME\n");
        printf("#   CP_F_VARIANCE_RANK_BYTES\n");
Philip Carns's avatar
Philip Carns committed
846
        printf("# WARNING: version 1.21 log format has the following limitations:\n");
847
        printf("# - *_TIMESTAMP fields are not normalized relative to MPI_Init() time.\n");
Philip Carns's avatar
Philip Carns committed
848 849 850 851
        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");
852 853 854 855 856 857 858 859
        return;
    }

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

860 861 862 863
/* shift_missing_1_21()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.21
864
 */
865
/*******************************
Philip Carns's avatar
Philip Carns committed
866
 * version 1.21 to 2.00 differences 
867 868 869 870 871 872 873 874
 * - 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
875 876 877 878
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
879 880
 *   - CP_F_MAX_READ_TIME
 *   - CP_F_MAX_WRITE_TIME
881 882 883 884
 *   - CP_F_FASTEST_RANK_TIME
 *   - CP_F_SLOWEST_RANK_TIME
 *   - CP_F_VARIANCE_RANK_TIME
 *   - CP_F_VARIANCE_RANK_BYTES
885
 * - changed params:
886 887 888
 *   - CP_FILE_RECORD_SIZE: 1184 to 1328
 *   - CP_NUM_INDICES: 133 to 144
 *   - CP_F_NUM_INDICES: 12 to 18
889
 */
890 891 892 893 894 895 896 897 898
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,
899
        CP_DEVICE,
900
        CP_SIZE_AT_OPEN,
901 902 903 904
        CP_FASTEST_RANK,
        CP_FASTEST_RANK_BYTES,
        CP_SLOWEST_RANK,
        CP_SLOWEST_RANK_BYTES,
905 906 907 908
        -1};
    int missing_f_counters[] = {
        CP_F_MAX_READ_TIME,
        CP_F_MAX_WRITE_TIME,
909 910 911 912
        CP_F_FASTEST_RANK_TIME,
        CP_F_SLOWEST_RANK_TIME,
        CP_F_VARIANCE_RANK_TIME,
        CP_F_VARIANCE_RANK_BYTES,
913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948
        -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;
}
949 950 951 952 953 954

/* shift_missing_1_22()
 *
 * translates indices to account for counters that weren't present in log
 * format 1.22
 */
955
/*******************************
Philip Carns's avatar
Philip Carns committed
956
 * version 1.22 to 2.00 differences
957 958 959 960
 *
 * - added:
 *   - CP_DEVICE
 *   - CP_SIZE_AT_OPEN
961 962 963 964 965 966 967 968
 *   - 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
969
 * - changed params:
970 971 972
 *   - CP_FILE_RECORD_SIZE: 1240 to 1328
 *   - CP_NUM_INDICES: 138 to 144
 *   - CP_F_NUM_INDICES: 14 to 18
973
 */
974 975 976 977 978
static void shift_missing_1_22(struct darshan_file* file)
{
    int c_index = 0;
    int missing_counters[] = {
        CP_DEVICE,
979
        CP_SIZE_AT_OPEN,
980 981 982 983 984 985 986 987 988 989
        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,
990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007
        -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;
    }

1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023
    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;
    }

1024 1025
    return;
}
1026

1027 1028 1029 1030 1031 1032 1033 1034 1035
/* 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:
1036 1037 1038 1039
 *   - CP_FASTEST_RANK
 *   - CP_FASTEST_RANK_BYTES
 *   - CP_SLOWEST_RANK
 *   - CP_SLOWEST_RANK_BYTES
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
 *   - 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;
}

1100
static int getjob_internal_201(darshan_fd file, struct darshan_job *job)
1101 1102 1103
{
    int ret;

1104 1105 1106
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
1107

1108
    ret = darshan_log_read(file, job, sizeof(*job));
1109 1110
    if (ret < sizeof(*job))
    {
1111
        fprintf(stderr, "Error: invalid log file (too short).\n");
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130
        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);
1131
        DARSHAN_BSWAP64(&job->jobid);
1132 1133 1134 1135 1136 1137 1138 1139
        return(0);
    }

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

1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154

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;

1155
    memset(job, 0, sizeof(job_200));
1156 1157
    memset(job, 0, sizeof(*job));

1158 1159 1160
    ret = darshan_log_seek(file, 0);
    if(ret < 0)
        return(ret);
1161

1162
    ret = darshan_log_read(file, &job_200, sizeof(job_200));
1163 1164
    if (ret < sizeof(job_200))
    {
1165
        fprintf(stderr, "Error: invalid log file (too short).\n");
1166