darshan-logutils.c 48.2 KB
Newer Older
1
/*
Shane Snyder's avatar
Shane Snyder committed
2 3 4
 * Copyright (C) 2015 University of Chicago.
 * See COPYRIGHT notice in top-level directory.
 *
5 6
 */

7
#define _GNU_SOURCE
8
#include "darshan-util-config.h"
9 10
#include <stdio.h>
#include <string.h>
11
#include <assert.h>
12
#include <stdlib.h>
13
#include <unistd.h>
14
#include <inttypes.h>
15 16 17
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
18
#include <errno.h>
19

20 21
#include "darshan-logutils.h"

22 23 24
/* default input buffer size for decompression algorithm */
#define DARSHAN_DEF_COMP_BUF_SZ (1024*1024) /* 1 MiB */

25
/* special identifers for referring to header, job, and
26 27 28 29 30 31
 * record map regions of the darshan log file
 */
#define DARSHAN_HEADER_REGION_ID    (-3)
#define DARSHAN_JOB_REGION_ID       (-2)
#define DARSHAN_REC_MAP_REGION_ID   (-1)

32 33
struct darshan_dz_state
{
34 35 36 37 38
    /* pointer to arbitrary data structure used for managing
     * compression/decompression state (e.g., z_stream
     * structure needed for libz)
     */
    void *comp_dat;
39 40 41
    /* buffer for staging compressed data to/from log file */
    unsigned char *buf;
    /* size of staging buffer */
42
    unsigned int size;
43 44
    /* for reading logs, flag indicating end of log file region */
    int eor;
45
    /* the region id we last tried reading/writing */
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64
    int prev_reg_id;
};

/* internal fd data structure */
struct darshan_fd_int_state
{
    /* posix file descriptor for the log file */
    int fildes;
    /* file pointer position */
    int64_t pos;
    /* flag indicating whether log file was created (and written) */
    int creat_flag;
    /* log file path name */
    char logfile_path[PATH_MAX];
    /* pointer to exe & mount data in darshan job data structure */
    char *exe_mnt_data;
    /* whether previous file operations have failed */
    int err;

65
    /* compression/decompression stream read/write state */
66 67 68
    struct darshan_dz_state dz;
};

69 70
static int darshan_log_getheader(darshan_fd fd);
static int darshan_log_putheader(darshan_fd fd);
71 72
static int darshan_log_seek(darshan_fd fd, off_t offset);
static int darshan_log_read(darshan_fd fd, void *buf, int len);
73
static int darshan_log_write(darshan_fd fd, void *buf, int len);
74 75
static int darshan_log_dzinit(darshan_fd fd);
static void darshan_log_dzdestroy(darshan_fd fd);
76 77
static int darshan_log_dzread(darshan_fd fd, int region_id, void *buf, int len);
static int darshan_log_dzwrite(darshan_fd fd, int region_id, void *buf, int len);
78 79 80 81
static int darshan_log_libz_read(darshan_fd fd, struct darshan_log_map map, 
    void *buf, int len, int reset_strm_flag);
static int darshan_log_libz_write(darshan_fd fd, struct darshan_log_map *map_p,
    void *buf, int len, int flush_strm_flag);
82
static int darshan_log_libz_flush(darshan_fd fd, int region_id);
83
#ifdef HAVE_LIBBZ2
84 85 86 87
static int darshan_log_bzip2_read(darshan_fd fd, struct darshan_log_map map, 
    void *buf, int len, int reset_strm_flag);
static int darshan_log_bzip2_write(darshan_fd fd, struct darshan_log_map *map_p,
    void *buf, int len, int flush_strm_flag);
88 89
static int darshan_log_bzip2_flush(darshan_fd fd, int region_id);
#endif
90 91
static int darshan_log_dzload(darshan_fd fd, struct darshan_log_map map);
static int darshan_log_dzunload(darshan_fd fd, struct darshan_log_map *map_p);
92 93
static int darshan_log_noz_read(darshan_fd fd, struct darshan_log_map map,
    void *buf, int len, int reset_strm_flag);
94

Shane Snyder's avatar
Shane Snyder committed
95
/* each module's implementation of the darshan logutil functions */
96
#define X(a, b, c, d) d,
97 98
struct darshan_mod_logutil_funcs *mod_logutils[DARSHAN_MAX_MODS] =
{
Shane Snyder's avatar
Shane Snyder committed
99
    DARSHAN_MODULE_IDS
100
};
Shane Snyder's avatar
Shane Snyder committed
101
#undef X
102 103 104

/* darshan_log_open()
 *
105
 * open an existing darshan log file for reading only
106
 *
107
 * returns file descriptor on success, NULL on failure
108
 */
109
darshan_fd darshan_log_open(const char *name)
110
{
111
    darshan_fd tmp_fd;
112
    int ret;
113

114
    /* allocate a darshan file descriptor */
115 116 117 118
    tmp_fd = malloc(sizeof(*tmp_fd));
    if(!tmp_fd)
        return(NULL);
    memset(tmp_fd, 0, sizeof(*tmp_fd));
119 120 121 122 123 124 125
    tmp_fd->state = malloc(sizeof(struct darshan_fd_int_state));
    if(!tmp_fd->state)
    {
        free(tmp_fd->state);
        return(NULL);
    }
    memset(tmp_fd->state, 0, sizeof(struct darshan_fd_int_state));
126

127
    /* open the log file in read mode */
128 129
    tmp_fd->state->fildes = open(name, O_RDONLY);
    if(tmp_fd->state->fildes < 0)
130
    {
131
        fprintf(stderr, "Error: failed to open darshan log file %s.\n", name);
132
        free(tmp_fd->state);
133 134 135
        free(tmp_fd);
        return(NULL);
    }
136
    strncpy(tmp_fd->state->logfile_path, name, PATH_MAX);
137 138 139 140 141

    /* read the header from the log file to init fd data structures */
    ret = darshan_log_getheader(tmp_fd);
    if(ret < 0)
    {
142
        fprintf(stderr, "Error: failed to read darshan log file header.\n");
143 144
        close(tmp_fd->state->fildes);
        free(tmp_fd->state);
145
        free(tmp_fd);
146 147 148 149
        return(NULL);
    }

    /* initialize compression data structures */
150
    ret = darshan_log_dzinit(tmp_fd);
151 152 153
    if(ret < 0)
    {
        fprintf(stderr, "Error: failed to initialize decompression data structures.\n");
154 155
        close(tmp_fd->state->fildes);
        free(tmp_fd->state);
156 157
        free(tmp_fd);
        return(NULL);
158 159
    }

160 161 162 163 164 165 166
    return(tmp_fd);
}

/* darshan_log_create()
 *
 * create a darshan log file for writing with the given compression method
 *
167
 * returns file descriptor on success, NULL on failure
168
 */
169 170
darshan_fd darshan_log_create(const char *name, enum darshan_comp_type comp_type,
    int partial_flag)
171 172
{
    darshan_fd tmp_fd;
173
    int ret;
174

175
    /* allocate a darshan file descriptor */
176 177 178 179
    tmp_fd = malloc(sizeof(*tmp_fd));
    if(!tmp_fd)
        return(NULL);
    memset(tmp_fd, 0, sizeof(*tmp_fd));
180 181 182 183 184 185 186
    tmp_fd->state = malloc(sizeof(struct darshan_fd_int_state));
    if(!tmp_fd->state)
    {
        free(tmp_fd);
        return(NULL);
    }
    memset(tmp_fd->state, 0, sizeof(struct darshan_fd_int_state));
187
    tmp_fd->comp_type = comp_type;
188

189 190 191
    /* create the log for writing, making sure to not overwrite existing log */
    tmp_fd->state->fildes = creat(name, 0400);
    if(tmp_fd->state->fildes < 0)
192
    {
193
        fprintf(stderr, "Error: failed to open darshan log file %s.\n", name);
194
        free(tmp_fd->state);
195
        free(tmp_fd);
196
        return(NULL);
197
    }
198
    tmp_fd->state->creat_flag = 1;
199
    tmp_fd->partial_flag = partial_flag;
200
    strncpy(tmp_fd->state->logfile_path, name, PATH_MAX);
201

202 203 204 205 206
    /* position file pointer to prealloc space for the log file header
     * NOTE: the header is written at close time, after all internal data
     * structures have been properly set
     */
    ret = darshan_log_seek(tmp_fd, sizeof(struct darshan_header));
207
    if(ret < 0)
208 209
    {
        fprintf(stderr, "Error: unable to seek in darshan log file.\n");
210 211
        close(tmp_fd->state->fildes);
        free(tmp_fd->state);
212 213 214
        free(tmp_fd);
        unlink(name);
        return(NULL);
215 216
    }

217
    /* initialize compression data structures */
218
    ret = darshan_log_dzinit(tmp_fd);
219 220
    if(ret < 0)
    {
221
        fprintf(stderr, "Error: failed to initialize compression data structures.\n");
222 223
        close(tmp_fd->state->fildes);
        free(tmp_fd->state);
224 225 226
        free(tmp_fd);
        unlink(name);
        return(NULL);
227 228
    }

229
    return(tmp_fd);
230 231
}

232
/* darshan_log_getjob()
233 234
 *
 * read job level metadata from the darshan log file
235
 *
236
 * returns 0 on success, -1 on failure
237
 */
238
int darshan_log_getjob(darshan_fd fd, struct darshan_job *job)
239
{
240
    struct darshan_fd_int_state *state = fd->state;
241 242
    char job_buf[DARSHAN_JOB_RECORD_SIZE] = {0};
    int job_buf_sz = DARSHAN_JOB_RECORD_SIZE;
243
    int ret;
244

245
    assert(state);
246 247
    assert(fd->job_map.len > 0 && fd->job_map.off > 0);

248
    /* read the compressed job data from the log file */
249
    ret = darshan_log_dzread(fd, DARSHAN_JOB_REGION_ID, job_buf, job_buf_sz);
250
    if(ret <= (int)sizeof(*job))
251
    {
252
        fprintf(stderr, "Error: failed to read darshan log file job data.\n");
253 254
        return(-1);
    }
255 256

    memcpy(job, job_buf, sizeof(*job));
257

258
    if(fd->swap_flag)
259
    {
260 261 262 263 264 265
        /* swap bytes if necessary */
        DARSHAN_BSWAP64(&job->uid);
        DARSHAN_BSWAP64(&job->start_time);
        DARSHAN_BSWAP64(&job->end_time);
        DARSHAN_BSWAP64(&job->nprocs);
        DARSHAN_BSWAP64(&job->jobid);
266 267
    }

268
    /* save trailing exe & mount information, so it can be retrieved later */
269 270 271
    if(!(state->exe_mnt_data))
        state->exe_mnt_data = malloc(DARSHAN_EXE_LEN+1);
    if(!(state->exe_mnt_data))
272
        return(-1);
273
    memcpy(state->exe_mnt_data, &job_buf[sizeof(*job)], DARSHAN_EXE_LEN+1);
274

275 276 277
    return(0);
}

278 279
/* darshan_log_putjob()
 *
280
 * write job level metadata to darshan log file
281 282 283 284 285
 *
 * returns 0 on success, -1 on failure
 */
int darshan_log_putjob(darshan_fd fd, struct darshan_job *job)
{
286
    struct darshan_fd_int_state *state = fd->state;
287 288 289 290
    struct darshan_job job_copy;
    int len;
    int ret;

291 292
    assert(state);

293 294 295 296 297 298 299 300 301 302 303 304 305 306
    memset(&job_copy, 0, sizeof(*job));
    memcpy(&job_copy, job, sizeof(*job));

    /* 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';
        }
    }

307 308
    /* write the compressed job data to log file */
    ret = darshan_log_dzwrite(fd, DARSHAN_JOB_REGION_ID, &job_copy, sizeof(*job));
309
    if(ret != sizeof(*job))
310
    {
311
        state->err = -1;
312 313 314 315 316 317 318 319 320 321 322 323 324
        fprintf(stderr, "Error: failed to write darshan log file job data.\n");
        return(-1);
    }

    return(0);
}

/* darshan_log_getexe()
 *
 * reads the application exe name from darshan log file
 * 
 * returns 0 on success, -1 on failure 
 */
325 326
int darshan_log_getexe(darshan_fd fd, char *buf)
{
327
    struct darshan_fd_int_state *state = fd->state;
328
    char *newline;
329
    int ret;
330

331 332
    assert(state);

333
    /* if the exe/mount data has not been saved yet, read in the job info */
334
    if(!(state->exe_mnt_data))
335
    {
336 337
        struct darshan_job job;
        ret = darshan_log_getjob(fd, &job);
338

339
        if(ret < 0 || !(state->exe_mnt_data))
340
            return(-1);
341
    }
342

343
    /* exe string is located before the first line break */
344
    newline = strchr(state->exe_mnt_data, '\n');
345 346

    /* copy over the exe string */
347
    if(newline)
348
        memcpy(buf, state->exe_mnt_data, (newline - state->exe_mnt_data));
349 350 351 352

    return (0);
}

353 354 355
/* darshan_log_putexe()
 *
 * wrties the application exe name to darshan log file
356
 * NOTE: this needs to be called immediately following put_job as it
357
 * expects the file pointer to be positioned immediately following
358
 * the darshan job information
359 360 361 362 363
 *
 * returns 0 on success, -1 on failure 
 */
int darshan_log_putexe(darshan_fd fd, char *buf)
{
364 365
    struct darshan_fd_int_state *state = fd->state;
    int len = strlen(buf);
366
    int ret;
367

368
    assert(fd->state);
369

370 371
    ret = darshan_log_dzwrite(fd, DARSHAN_JOB_REGION_ID, buf, len);
    if(ret != len)
372
    {
373
        state->err = -1;
374 375 376 377 378 379 380
        fprintf(stderr, "Error: failed to write exe string to darshan log file.\n");
        return(-1);
    }

    return(0);
}

381 382
/* darshan_log_getmounts()
 * 
383
 * retrieves mount table information from the log. Note that mnt_pts and
384
 * fs_types are arrays that will be allocated by the function and must be
385 386 387
 * freed by the caller. count will indicate the size of the arrays
 *
 * returns 0 on success, -1 on failure
388
 */
389
int darshan_log_getmounts(darshan_fd fd, char*** mnt_pts,
390 391
    char*** fs_types, int* count)
{
392
    struct darshan_fd_int_state *state = fd->state;
393 394
    char *pos;
    int array_index = 0;
395
    int ret;
396

397 398
    assert(state);

399
    /* if the exe/mount data has not been saved yet, read in the job info */
400
    if(!(state->exe_mnt_data))
401
    {
402 403
        struct darshan_job job;
        ret = darshan_log_getjob(fd, &job);
404

405
        if(ret < 0 || !(state->exe_mnt_data))
406
            return(-1);
407
    }
408

409
    /* count entries */
410
    *count = 0;
411
    pos = state->exe_mnt_data;
412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
    while((pos = strchr(pos, '\n')) != NULL)
    {
        pos++;
        (*count)++;
    }

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

    /* allocate output arrays */
    *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)
     */
433
    while((pos = strrchr(state->exe_mnt_data, '\n')) != NULL)
434 435
    {
        /* overestimate string lengths */
436
        (*mnt_pts)[array_index] = malloc(DARSHAN_EXE_LEN);
437
        assert((*mnt_pts)[array_index]);
438
        (*fs_types)[array_index] = malloc(DARSHAN_EXE_LEN);
439 440
        assert((*fs_types)[array_index]);

441 442 443
        ret = sscanf(++pos, "%s\t%s", (*fs_types)[array_index],
            (*mnt_pts)[array_index]);
        if(ret != 2)
444
        {
445
            fprintf(stderr, "Error: poorly formatted mount table in darshan log file.\n");
446 447 448 449 450 451 452
            return(-1);
        }
        pos--;
        *pos = '\0';
        array_index++;
    }

453 454 455
    return(0);
}

456 457 458 459 460 461 462 463 464 465 466
/* darshan_log_putmounts()
 *
 * writes mount information to the darshan log file
 * NOTE: this function call should follow immediately after the call
 * to darshan_log_putexe(), as it assumes the darshan log file pointer
 * is pointing to the offset immediately following the exe string
 *
 * returns 0 on success, -1 on failure
 */
int darshan_log_putmounts(darshan_fd fd, char** mnt_pts, char** fs_types, int count)
{
467
    struct darshan_fd_int_state *state = fd->state;
468 469
    int i;
    char line[1024];
470 471 472
    char mnt_dat[DARSHAN_EXE_LEN] = {0};
    int mnt_dat_sz = 0;
    char *tmp;
473 474
    int ret;

475 476
    assert(state);

477
    /* write each mount entry to file */
478
    tmp = mnt_dat;
479 480 481
    for(i=count-1; i>=0; i--)
    {
        sprintf(line, "\n%s\t%s", fs_types[i], mnt_pts[i]);
482 483 484 485 486 487

        memcpy(tmp, line, strlen(line));
        tmp += strlen(line);
        mnt_dat_sz += strlen(line);
    }

488 489
    ret = darshan_log_dzwrite(fd, DARSHAN_JOB_REGION_ID, mnt_dat, mnt_dat_sz);
    if (ret != mnt_dat_sz)
490
    {
491
        state->err = -1;
492
        fprintf(stderr, "Error: failed to write darshan log mount data.\n");
493 494 495 496 497 498
        return(-1);
    }

    return(0);
}

499
/* darshan_log_gethash()
500
 *
501
 * read the hash of records from the darshan log file
502 503 504
 *
 * returns 0 on success, -1 on failure
 */
505
int darshan_log_gethash(darshan_fd fd, struct darshan_record_ref **hash)
506
{
507
    struct darshan_fd_int_state *state = fd->state;
508
    char *hash_buf;
509
    int hash_buf_sz;
510
    char *buf_ptr;
511 512
    darshan_record_id *rec_id_ptr;
    char *path_ptr;
513
    char *tmp_p;
514
    struct darshan_record_ref *ref;
515 516
    int read;
    int read_req_sz;
517
    int buf_rem = 0;
518

519 520
    assert(state);

521 522 523 524 525 526 527
    /* just return if there is no record mapping data */
    if(fd->rec_map.len == 0)
    {
        *hash = NULL;
        return(0);
    }

528 529 530
    /* default to hash buffer twice as big as default compression buf */
    hash_buf = malloc(DARSHAN_DEF_COMP_BUF_SZ * 2);
    if(!hash_buf)
531
        return(-1);
532 533
    memset(hash_buf, 0, DARSHAN_DEF_COMP_BUF_SZ * 2);
    hash_buf_sz = DARSHAN_DEF_COMP_BUF_SZ * 2;
534

535
    do
536
    {
537 538
        /* read chunks of the darshan record id -> file name mapping from log file,
         * constructing a hash table in the process
539
         */
540
        read_req_sz = hash_buf_sz - buf_rem;
541
        read = darshan_log_dzread(fd, DARSHAN_REC_MAP_REGION_ID,
542
            hash_buf + buf_rem, read_req_sz);
543
        if(read < 0)
544
        {
545 546 547
            fprintf(stderr, "Error: failed to read record hash from darshan log file.\n");
            free(hash_buf);
            return(-1);
548 549
        }

550 551 552 553 554 555
        /* work through the hash buffer -- deserialize the mapping data and
         * add to the output hash table
         * NOTE: these mapping pairs are variable in length, so we have to be able
         * to handle incomplete mappings temporarily here
         */
        buf_ptr = hash_buf;
556 557
        buf_rem += read;
        while(buf_rem > (sizeof(darshan_record_id) + 1))
558
        {
559 560 561 562 563 564 565 566 567
            tmp_p = buf_ptr + sizeof(darshan_record_id);
            while(tmp_p < (buf_ptr + buf_rem))
            {
                /* look for terminating null character for record name */
                if(*tmp_p == '\0')
                    break;
                tmp_p++;
            }
            if(*tmp_p != '\0')
568 569 570 571
                break;

            /* get pointers for each field of this darshan record */
            /* NOTE: darshan record hash serialization method: 
572
             *          ... darshan_record_id | path '\0' ...
573 574 575 576 577 578
             */
            rec_id_ptr = (darshan_record_id *)buf_ptr;
            buf_ptr += sizeof(darshan_record_id);
            path_ptr = (char *)buf_ptr;

            if(fd->swap_flag)
579
            {
580 581
                /* we need to sort out endianness issues before deserializing */
                DARSHAN_BSWAP64(rec_id_ptr);
582
            }
583 584 585

            HASH_FIND(hlink, *hash, rec_id_ptr, sizeof(darshan_record_id), ref);
            if(!ref)
586
            {
587 588 589 590 591 592
                ref = malloc(sizeof(*ref));
                if(!ref)
                {
                    free(hash_buf);
                    return(-1);
                }
593 594
                ref->name = malloc(strlen(path_ptr) + 1);
                if(!ref->name)
595 596 597 598 599 600 601
                {
                    free(ref);
                    free(hash_buf);
                    return(-1);
                }

                /* set the fields for this record */
602 603
                ref->id = *rec_id_ptr;
                strcpy(ref->name, path_ptr);
604 605

                /* add this record to the hash */
606
                HASH_ADD(hlink, *hash, id, sizeof(darshan_record_id), ref);
607 608
            }

609 610
            buf_ptr += strlen(path_ptr) + 1;
            buf_rem -= (sizeof(darshan_record_id) + strlen(path_ptr) + 1);
611
        }
612

613
        /* copy any leftover data to beginning of buffer to parse next */
614
        memcpy(hash_buf, buf_ptr, buf_rem);
615 616 617 618 619

        /* we keep reading until we get a short read informing us we have
         * read all of the record hash
         */
    } while(read == read_req_sz);
620
    assert(buf_rem == 0);
621

622
    free(hash_buf);
623 624 625
    return(0);
}

626 627 628
/* darshan_log_puthash()
 *
 * writes the hash table of records to the darshan log file
629 630 631
 * NOTE: this function call should follow immediately after the call
 * to darshan_log_putmounts(), as it assumes the darshan log file pointer
 * is pointing to the offset immediately following the mount information
632 633 634 635 636
 *
 * returns 0 on success, -1 on failure
 */
int darshan_log_puthash(darshan_fd fd, struct darshan_record_ref *hash)
{
637
    struct darshan_fd_int_state *state = fd->state;
638
    char *hash_buf;
639
    int hash_buf_sz;
640
    struct darshan_record_ref *ref, *tmp;
641 642
    char *buf_ptr;
    int wrote;
643

644 645
    assert(state);

646
    /* allocate memory for largest possible hash record */
647
    hash_buf_sz = sizeof(darshan_record_id) + PATH_MAX + 1;
648 649 650
    hash_buf = malloc(hash_buf_sz);
    if(!hash_buf)
        return(-1);
651
    memset(hash_buf, 0, hash_buf_sz);
652

653
    /* individually serialize each hash record and write to log file */
654 655
    HASH_ITER(hlink, hash, ref, tmp)
    {
656
        buf_ptr = hash_buf;
657

658
        /* the hash buffer has space to serialize this record
659
         * NOTE: darshan record hash serialization method: 
660
         *          ... darshan_record_id | path '\0' ...
661
         */
662
        *((darshan_record_id *)buf_ptr) = ref->id;
663
        buf_ptr += sizeof(darshan_record_id);
664 665
        strcpy(buf_ptr, ref->name);
        buf_ptr += strlen(ref->name) + 1;
666

667 668 669 670 671
        /* write this hash entry to log file */
        wrote = darshan_log_dzwrite(fd, DARSHAN_REC_MAP_REGION_ID,
            hash_buf, (buf_ptr - hash_buf));
        if(wrote != (buf_ptr - hash_buf))
        {
672
            state->err = -1;
673 674 675 676
            fprintf(stderr, "Error: failed to write record hash to darshan log file.\n");
            free(hash_buf);
            return(-1);
        }
677 678 679 680 681 682 683 684
    }

    free(hash_buf);
    return(0);
}

/* darshan_log_getmod()
 *
685 686
 * get a chunk of module data from the darshan log file
 *
687
 * returns number of bytes read on success, -1 on failure
688 689
 */
int darshan_log_getmod(darshan_fd fd, darshan_module_id mod_id,
690
    void *mod_buf, int mod_buf_sz)
691
{
692
    struct darshan_fd_int_state *state = fd->state;
693
    int ret;
694

695 696
    assert(state);

697
    if(mod_id < 0 || mod_id >= DARSHAN_MAX_MODS)
698
    {
699 700
        fprintf(stderr, "Error: invalid Darshan module id.\n");
        return(-1);
701 702
    }

703 704 705 706
    if(fd->mod_map[mod_id].len == 0)
        return(0); /* no data corresponding to this mod_id */

    /* read this module's data from the log file */
707
    ret = darshan_log_dzread(fd, mod_id, mod_buf, mod_buf_sz);
708
    if(ret < 0)
709
    {
710
        fprintf(stderr,
711
            "Error: failed to read module %s data from darshan log file.\n",
712
            darshan_module_names[mod_id]);
713 714 715
        return(-1);
    }

716
    return(ret);
717 718
}

719 720 721
/* darshan_log_putmod()
 *
 * write a chunk of module data to the darshan log file
Shane Snyder's avatar
Shane Snyder committed
722 723 724 725 726 727
 * NOTE: this function call should be called directly after the
 * put_hash() function, as it expects the file pointer to be
 * positioned directly past the record hash location. Also,
 * for a set of modules with data to write to file, this function
 * should be called in order of increasing module identifiers,
 * as the darshan log file format expects this ordering.
728
 *
729
 * returns number of bytes written on success, -1 on failure
730 731
 */
int darshan_log_putmod(darshan_fd fd, darshan_module_id mod_id,
732
    void *mod_buf, int mod_buf_sz, int ver)
733
{
734
    struct darshan_fd_int_state *state = fd->state;
735 736
    int ret;

737 738
    assert(state);

Shane Snyder's avatar
Shane Snyder committed
739
    if(mod_id < 0 || mod_id >= DARSHAN_MAX_MODS)
740
    {
741
        state->err = -1;
Shane Snyder's avatar
Shane Snyder committed
742 743
        fprintf(stderr, "Error: invalid Darshan module id.\n");
        return(-1);
744 745 746
    }

    /* write the module chunk to the log file */
747 748
    ret = darshan_log_dzwrite(fd, mod_id, mod_buf, mod_buf_sz);
    if(ret != mod_buf_sz)
749
    {
750
        state->err = -1;
751 752 753 754 755 756
        fprintf(stderr,
            "Error: failed to write module %s data to darshan log file.\n",
            darshan_module_names[mod_id]);
        return(-1);
    }

757 758 759
    /* set the version number for this module's data */
    fd->mod_ver[mod_id] = ver;

760 761 762
    return(0);
}

763 764
/* darshan_log_close()
 *
765
 * close an open darshan file descriptor, freeing any resources
766 767 768
 *
 */
void darshan_log_close(darshan_fd fd)
769
{
770
    struct darshan_fd_int_state *state = fd->state;
771 772
    int ret;

773 774 775 776
    assert(state);

    /* if the file was created for writing */
    if(state->creat_flag)
777 778
    {
        /* flush the last region of the log to file */
779
        switch(fd->comp_type)
780 781
        {
            case DARSHAN_ZLIB_COMP:
782
                ret = darshan_log_libz_flush(fd, state->dz.prev_reg_id);
783 784
                if(ret == 0)
                    break;
785 786
#ifdef HAVE_LIBBZ2
            case DARSHAN_BZIP2_COMP:
787
                ret = darshan_log_bzip2_flush(fd, state->dz.prev_reg_id);
788 789 790
                if(ret == 0)
                    break;
#endif 
791 792
            default:
                /* if flush fails, remove the output log file */
793
                state->err = -1;
794 795 796 797 798
                fprintf(stderr, "Error: final flush to log file failed.\n");
                break;
        }

        /* if no errors flushing, write the log header before closing */
799
        if(state->err != -1)
800 801 802
        {
            ret = darshan_log_putheader(fd);
            if(ret < 0)
803
                state->err = -1;
804 805 806
        }
    }

807
    close(state->fildes);
808 809

    /* remove output log file if error writing to it */
810
    if((state->creat_flag) && (state->err == -1))
811 812
    {
        fprintf(stderr, "Unlinking darshan log file %s ...\n",
813 814
            state->logfile_path);
        unlink(state->logfile_path);
815
    }
816

817
    darshan_log_dzdestroy(fd);
818 819 820
    if(state->exe_mnt_data)
        free(state->exe_mnt_data);
    free(state);
821
    free(fd);
822 823

    return;
824 825
}

826
/* **************************************************** */
827

828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
/* read the header of the darshan log and set internal fd data structures
 * NOTE: this is the only portion of the darshan log that is uncompressed
 *
 * returns 0 on success, -1 on failure
 */
static int darshan_log_getheader(darshan_fd fd)
{
    struct darshan_header header;
    int i;
    int ret;

    ret = darshan_log_seek(fd, 0);
    if(ret < 0)
    {
        fprintf(stderr, "Error: unable to seek in darshan log file.\n");
        return(-1);
    }

846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
    /* read the version number so we know how to process this log */
    ret = darshan_log_read(fd, &fd->version, 8);
    if(ret < 8)
    {
        fprintf(stderr, "Error: invalid log file (failed to read version).\n");
        return(-1);
    }

    /* other log file versions can be detected and handled here */
    if(strcmp(fd->version, "3.00"))
    {
        fprintf(stderr, "Error: incompatible darshan file.\n");
        fprintf(stderr, "Error: expected version %s\n", DARSHAN_LOG_VERSION);
        return(-1);
    }

    /* seek back so we can read the entire header */
    ret = darshan_log_seek(fd, 0);
    if(ret < 0)
    {
        fprintf(stderr, "Error: unable to seek in darshan log file.\n");
        return(-1);
    }

870 871
    /* read uncompressed header from log file */
    ret = darshan_log_read(fd, &header, sizeof(header));
872
    if(ret != (int)sizeof(header))
873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907
    {
        fprintf(stderr, "Error: failed to read darshan log file header.\n");
        return(-1);
    }

    if(header.magic_nr == DARSHAN_MAGIC_NR)
    {
        /* no byte swapping needed, this file is in host format already */
        fd->swap_flag = 0;
    }
    else
    {
        /* try byte swapping */
        DARSHAN_BSWAP64(&(header.magic_nr));
        if(header.magic_nr == DARSHAN_MAGIC_NR)
        {
            fd->swap_flag = 1;

            /* swap the log map variables in the header */
            DARSHAN_BSWAP64(&(header.rec_map.off));
            DARSHAN_BSWAP64(&(header.rec_map.len));
            for(i = 0; i < DARSHAN_MAX_MODS; i++)
            {
                DARSHAN_BSWAP64(&(header.mod_map[i].off));
                DARSHAN_BSWAP64(&(header.mod_map[i].len));
            }
        }
        else
        {
            /* otherwise this file is just broken */
            fprintf(stderr, "Error: bad magic number in darshan log file.\n");
            return(-1);
        }
    }

908
    /* set some fd fields based on what's stored in the header */
909
    fd->comp_type = header.comp_type;
910
    fd->partial_flag = header.partial_flag;
911
    memcpy(fd->mod_ver, header.mod_ver, DARSHAN_MAX_MODS * sizeof(uint32_t));
912 913 914 915 916

    /* save the mapping of data within log file to this file descriptor */
    memcpy(&fd->rec_map, &(header.rec_map), sizeof(struct darshan_log_map));
    memcpy(&fd->mod_map, &(header.mod_map), DARSHAN_MAX_MODS * sizeof(struct darshan_log_map));

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
    /* there may be nothing following the job data, so safety check map */
    fd->job_map.off = sizeof(struct darshan_header);
    if(fd->rec_map.off == 0)
    {
        for(i = 0; i < DARSHAN_MAX_MODS; i++)
        {
            if(fd->mod_map[i].off != 0)
            {
                fd->job_map.len = fd->mod_map[i].off - fd->job_map.off;
                break;
            }
        }

        if(fd->job_map.len == 0)
        {
            struct stat sbuf;
            if(fstat(fd->state->fildes, &sbuf) != 0)
            {
                fprintf(stderr, "Error: unable to stat darshan log file.\n");
                return(-1);
            }
            fd->job_map.len = sbuf.st_size - fd->job_map.off;
        }
    }
    else
    {
        fd->job_map.len = fd->rec_map.off - fd->job_map.off;
    }

946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967
    return(0);
}

/* write a darshan header to log file
 *
 * returns 0 on success, -1 on failure
 */
static int darshan_log_putheader(darshan_fd fd)
{
    struct darshan_header header;
    int ret;

    ret = darshan_log_seek(fd, 0);
    if(ret < 0)
    {
        fprintf(stderr, "Error: unable to seek in darshan log file.\n");
        return(-1);
    }

    memset(&header, 0, sizeof(header));
    strcpy(header.version_string, DARSHAN_LOG_VERSION);
    header.magic_nr = DARSHAN_MAGIC_NR;
968
    header.comp_type = fd->comp_type;
969
    header.partial_flag = fd->partial_flag;
970 971 972 973 974 975 976

    /* copy the mapping information to the header */
    memcpy(&header.rec_map, &fd->rec_map, sizeof(struct darshan_log_map));
    memcpy(&header.mod_map, &fd->mod_map, DARSHAN_MAX_MODS * sizeof(struct darshan_log_map));

    /* write header to file */
    ret = darshan_log_write(fd, &header, sizeof(header));
977
    if(ret != (int)sizeof(header))
978 979 980 981 982 983 984 985
    {
        fprintf(stderr, "Error: failed to write Darshan log file header.\n");
        return(-1);
    }

    return(0);
}

986 987 988 989
/* return 0 on successful seek to offset, -1 on failure.
 */
static int darshan_log_seek(darshan_fd fd, off_t offset)
{
990
    struct darshan_fd_int_state *state = fd->state;
991
    off_t ret_off;
992

993
    if(state->pos == offset)
994 995
        return(0);

996
    ret_off = lseek(state->fildes, offset, SEEK_SET);
997
    if(ret_off == offset)
998
    {
999
        state->pos = offset;
1000
        return(0);
1001 1002 1003 1004 1005
    }

    return(-1);
}

1006
/* return amount read on success, 0 on EOF, -1 on failure.
1007
 */
1008
static int darshan_log_read(darshan_fd fd, void* buf, int len)
1009
{
1010
    struct darshan_fd_int_state *state = fd->state;
1011
    int ret;
1012
    unsigned int read_so_far = 0;
1013

1014 1015 1016 1017 1018 1019 1020 1021 1022
    do
    {
        ret = read(state->fildes, buf + read_so_far, len - read_so_far);
        if(ret <= 0)
            break;
        read_so_far += ret;
    } while(read_so_far < len);
    if(ret < 0)
        return(-1);
1023

1024 1025
    state->pos += read_so_far;
    return(read_so_far);
1026 1027
}

1028
/* return amount written on success, -1 on failure.
1029
 */
1030
static int darshan_log_write(darshan_fd fd, void* buf, int len)
1031
{
1032
    struct darshan_fd_int_state *state = fd->state;
1033
    int ret;
1034
    unsigned int wrote_so_far = 0;
1035

1036 1037 1038 1039 1040 1041 1042 1043 1044
    do
    {
        ret = write(state->fildes, buf + wrote_so_far, len - wrote_so_far);
        if(ret <= 0)
            break;
        wrote_so_far += ret;
    } while(wrote_so_far < len);
    if(ret < 0)
        return(-1);
1045

1046 1047
    state->pos += wrote_so_far;
    return(wrote_so_far);
1048 1049
}

1050
static int darshan_log_dzinit(darshan_fd fd)
1051
{
1052
    struct darshan_fd_int_state *state = fd->state;
1053 1054
    int ret;

1055 1056 1057
    /* initialize buffers for staging compressed data
     * to/from log file
     */
1058 1059
    state->dz.buf = malloc(DARSHAN_DEF_COMP_BUF_SZ);
    if(state->dz.buf == NULL)
1060
        return(-1);
1061
    state->dz.size = 0;
1062
    state->dz.prev_reg_id = DARSHAN_HEADER_REGION_ID;
1063

1064
    switch(fd->comp_type)
1065 1066
    {
        case DARSHAN_ZLIB_COMP:
1067 1068