darshan-posix.c 60.6 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 8 9
#define _XOPEN_SOURCE 500
#define _GNU_SOURCE

10
#include "darshan-runtime-config.h"
11 12 13 14 15 16 17 18 19 20 21 22 23 24
#include <stdio.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <stdarg.h>
#include <string.h>
#include <time.h>
#include <stdlib.h>
#include <errno.h>
#include <sys/uio.h>
#include <sys/mman.h>
#include <search.h>
#include <assert.h>
25
#include <libgen.h>
26
#include <aio.h>
27
#include <pthread.h>
28

29 30
#include "utlist.h"

31
#include "darshan.h"
32
#include "darshan-dynamic.h"
33

34
#ifndef HAVE_OFF64_T
35 36
typedef int64_t off64_t;
#endif
37 38 39
#ifndef HAVE_AIOCB64
#define aiocb64 aiocb
#endif
40

41 42
DARSHAN_FORWARD_DECL(open, int, (const char *path, int flags, ...));
DARSHAN_FORWARD_DECL(open64, int, (const char *path, int flags, ...));
43 44
DARSHAN_FORWARD_DECL(creat, int, (const char* path, mode_t mode));
DARSHAN_FORWARD_DECL(creat64, int, (const char* path, mode_t mode));
45 46
DARSHAN_FORWARD_DECL(fopen, FILE*, (const char *path, const char *mode));
DARSHAN_FORWARD_DECL(fopen64, FILE*, (const char *path, const char *mode));
47 48 49 50
DARSHAN_FORWARD_DECL(mkstemp, int, (char *template));
DARSHAN_FORWARD_DECL(mkostemp, int, (char *template, int flags));
DARSHAN_FORWARD_DECL(mkstemps, int, (char *template, int suffixlen));
DARSHAN_FORWARD_DECL(mkostemps, int, (char *template, int suffixlen, int flags));
51 52 53 54 55 56 57 58
DARSHAN_FORWARD_DECL(read, ssize_t, (int fd, void *buf, size_t count));
DARSHAN_FORWARD_DECL(write, ssize_t, (int fd, const void *buf, size_t count));
DARSHAN_FORWARD_DECL(pread, ssize_t, (int fd, void *buf, size_t count, off_t offset));
DARSHAN_FORWARD_DECL(pwrite, ssize_t, (int fd, const void *buf, size_t count, off_t offset));
DARSHAN_FORWARD_DECL(pread64, ssize_t, (int fd, void *buf, size_t count, off64_t offset));
DARSHAN_FORWARD_DECL(pwrite64, ssize_t, (int fd, const void *buf, size_t count, off64_t offset));
DARSHAN_FORWARD_DECL(readv, ssize_t, (int fd, const struct iovec *iov, int iovcnt));
DARSHAN_FORWARD_DECL(writev, ssize_t, (int fd, const struct iovec *iov, int iovcnt));
59 60 61 62 63
DARSHAN_FORWARD_DECL(fread, size_t, (void *ptr, size_t size, size_t nmemb, FILE *stream));
DARSHAN_FORWARD_DECL(fwrite, size_t, (const void *ptr, size_t size, size_t nmemb, FILE *stream));
DARSHAN_FORWARD_DECL(lseek, off_t, (int fd, off_t offset, int whence));
DARSHAN_FORWARD_DECL(lseek64, off64_t, (int fd, off64_t offset, int whence));
DARSHAN_FORWARD_DECL(fseek, int, (FILE *stream, long offset, int whence));
64 65 66 67 68 69
DARSHAN_FORWARD_DECL(__xstat, int, (int vers, const char* path, struct stat *buf));
DARSHAN_FORWARD_DECL(__xstat64, int, (int vers, const char* path, struct stat64 *buf));
DARSHAN_FORWARD_DECL(__lxstat, int, (int vers, const char* path, struct stat *buf));
DARSHAN_FORWARD_DECL(__lxstat64, int, (int vers, const char* path, struct stat64 *buf));
DARSHAN_FORWARD_DECL(__fxstat, int, (int vers, int fd, struct stat *buf));
DARSHAN_FORWARD_DECL(__fxstat64, int, (int vers, int fd, struct stat64 *buf));
Shane Snyder's avatar
Shane Snyder committed
70 71
DARSHAN_FORWARD_DECL(mmap, void*, (void *addr, size_t length, int prot, int flags, int fd, off_t offset));
DARSHAN_FORWARD_DECL(mmap64, void*, (void *addr, size_t length, int prot, int flags, int fd, off64_t offset));
72 73
DARSHAN_FORWARD_DECL(fsync, int, (int fd));
DARSHAN_FORWARD_DECL(fdatasync, int, (int fd));
74
DARSHAN_FORWARD_DECL(close, int, (int fd));
75
DARSHAN_FORWARD_DECL(fclose, int, (FILE *fp));
76 77 78 79 80 81 82 83
DARSHAN_FORWARD_DECL(aio_read, int, (struct aiocb *aiocbp));
DARSHAN_FORWARD_DECL(aio_write, int, (struct aiocb *aiocbp));
DARSHAN_FORWARD_DECL(aio_read64, int, (struct aiocb64 *aiocbp));
DARSHAN_FORWARD_DECL(aio_write64, int, (struct aiocb64 *aiocbp));
DARSHAN_FORWARD_DECL(aio_return, ssize_t, (struct aiocb *aiocbp));
DARSHAN_FORWARD_DECL(aio_return64, ssize_t, (struct aiocb64 *aiocbp));
DARSHAN_FORWARD_DECL(lio_listio, int, (int mode, struct aiocb *const aiocb_list[], int nitems, struct sigevent *sevp));
DARSHAN_FORWARD_DECL(lio_listio64, int, (int mode, struct aiocb64 *const aiocb_list[], int nitems, struct sigevent *sevp));
84

85 86 87 88 89 90 91 92
/* struct to track information about aio operations in flight */
struct posix_aio_tracker
{
    double tm1;
    void *aiocbp;
    struct posix_aio_tracker* next;
};

93
struct posix_file_record_ref
94
{
95
    struct darshan_posix_file* file_rec;
96 97 98
    int64_t offset;
    int64_t last_byte_read;
    int64_t last_byte_written;
Shane Snyder's avatar
Shane Snyder committed
99
    enum darshan_io_type last_io_type;
100 101 102
    double last_meta_end;
    double last_read_end;
    double last_write_end;
103 104 105 106
    void* access_root;
    int access_count;
    void* stride_root;
    int stride_count;
107
    struct posix_aio_tracker* aio_list;
108 109 110 111
};

struct posix_runtime
{
112 113 114
    void *rec_id_hash;
    void *fd_hash;
    int file_rec_count;
115 116
};

117
static struct posix_runtime *posix_runtime = NULL;
118
static pthread_mutex_t posix_runtime_mutex = PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP;
Shane Snyder's avatar
Shane Snyder committed
119
static int instrumentation_disabled = 0;
120
static int my_rank = -1;
121
static int darshan_mem_alignment = 1;
122

123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142
static void posix_runtime_initialize(
    void);
static struct posix_file_record_ref *posix_track_new_file_record(
    darshan_record_id rec_id, const char *path);
static void posix_aio_tracker_add(
    int fd, void *aiocbp);
static struct posix_aio_tracker* posix_aio_tracker_del(
    int fd, void *aiocbp);
static int posix_record_compare(
    const void* a, const void* b);
static void posix_record_reduction_op(
    void* infile_v, void* inoutfile_v, int *len, MPI_Datatype *datatype);
static void posix_shared_record_variance(
    MPI_Comm mod_comm, struct darshan_posix_file *inrec_array,
    struct darshan_posix_file *outrec_array, int shared_rec_count);

static void posix_begin_shutdown(
    void);
static void posix_get_output_data(
    MPI_Comm mod_comm, darshan_record_id *shared_recs,
143
    int shared_rec_count, void **posix_buf, int *posix_buf_sz);
144 145
static void posix_shutdown(
    void);
146

147 148
#define POSIX_LOCK() pthread_mutex_lock(&posix_runtime_mutex)
#define POSIX_UNLOCK() pthread_mutex_unlock(&posix_runtime_mutex)
149

150
#define POSIX_RECORD_OPEN(__ret, __path, __mode, __stream_flag, __tm1, __tm2) do { \
151 152
    struct posix_file_record_ref *rec_ref; \
    darshan_record_id rec_id; \
153
    if(__ret < 0) break; \
154
    if(!posix_runtime || instrumentation_disabled) break; /* XXX: PREMABLE??? */\
155
    if(darshan_core_excluded_path(__path)) break; \
156 157 158 159 160 161 162
    rec_id = darshan_record_id_from_path(__path); \
    rec_ref = darshan_lookup_record_ref(posix_runtime->rec_id_hash, &rec_id, sizeof(darshan_record_id)); \
    if(!rec_ref) { \
        rec_ref = posix_track_new_file_record(rec_id, __path); \
        if(!rec_ref) break; \
    } \
    if(darshan_add_record_ref(&(posix_runtime->fd_hash), &__ret, sizeof(int), rec_ref) == 0) break; \
163
    if(__mode) \
164 165 166 167
        rec_ref->file_rec->counters[POSIX_MODE] = __mode; \
    rec_ref->offset = 0; \
    rec_ref->last_byte_written = 0; \
    rec_ref->last_byte_read = 0; \
168
    if(__stream_flag)\
169
        rec_ref->file_rec->counters[POSIX_FOPENS] += 1; \
170
    else \
171 172 173 174 175
        rec_ref->file_rec->counters[POSIX_OPENS] += 1; \
    if(rec_ref->file_rec->fcounters[POSIX_F_OPEN_TIMESTAMP] == 0) \
        rec_ref->file_rec->fcounters[POSIX_F_OPEN_TIMESTAMP] = __tm1; \
    DARSHAN_TIMER_INC_NO_OVERLAP(rec_ref->file_rec->fcounters[POSIX_F_META_TIME], \
        __tm1, __tm2, rec_ref->last_meta_end); \
176 177
} while(0)

178 179
#define POSIX_RECORD_READ(__ret, __fd, __pread_flag, __pread_offset, __aligned, __stream_flag, __tm1, __tm2) do { \
    struct posix_file_record_ref* rec_ref; \
180
    size_t stride; \
181
    int64_t this_offset; \
182
    int64_t file_alignment; \
183 184
    double __elapsed = __tm2-__tm1; \
    if(__ret < 0) break; \
185 186 187
    if(!posix_runtime || instrumentation_disabled) break; /* XXX: PREMABLE??? */\
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &(__fd), sizeof(int)); \
    if(!rec_ref) break; \
188
    if(__pread_flag) \
189 190
        this_offset = __pread_offset; \
    else \
191 192 193 194 195 196 197 198
        this_offset = rec_ref->offset; \
    if(this_offset > rec_ref->last_byte_read) \
        rec_ref->file_rec->counters[POSIX_SEQ_READS] += 1;  \
    if(this_offset == (rec_ref->last_byte_read + 1)) \
        rec_ref->file_rec->counters[POSIX_CONSEC_READS] += 1;  \
    if(this_offset > 0 && this_offset > rec_ref->last_byte_read \
        && rec_ref->last_byte_read != 0) \
        stride = this_offset - rec_ref->last_byte_read - 1; \
199
    else \
200
        stride = 0; \
201 202 203 204 205
    rec_ref->last_byte_read = this_offset + __ret - 1; \
    rec_ref->offset = this_offset + __ret; \
    if(rec_ref->file_rec->counters[POSIX_MAX_BYTE_READ] < (this_offset + __ret - 1)) \
        rec_ref->file_rec->counters[POSIX_MAX_BYTE_READ] = (this_offset + __ret - 1); \
    rec_ref->file_rec->counters[POSIX_BYTES_READ] += __ret; \
206
    if(__stream_flag) \
207
        rec_ref->file_rec->counters[POSIX_FREADS] += 1; \
208
    else \
209 210 211 212 213 214 215 216
        rec_ref->file_rec->counters[POSIX_READS] += 1; \
    DARSHAN_BUCKET_INC(&(rec_ref->file_rec->counters[POSIX_SIZE_READ_0_100]), __ret); \
    darshan_common_val_counter(&rec_ref->access_root, &rec_ref->access_count, __ret, \
        &(rec_ref->file_rec->counters[POSIX_ACCESS1_ACCESS]), \
        &(rec_ref->file_rec->counters[POSIX_ACCESS1_COUNT])); \
    darshan_common_val_counter(&rec_ref->stride_root, &rec_ref->stride_count, stride, \
        &(rec_ref->file_rec->counters[POSIX_STRIDE1_STRIDE]), \
        &(rec_ref->file_rec->counters[POSIX_STRIDE1_COUNT])); \
217
    if(!__aligned) \
218 219
        rec_ref->file_rec->counters[POSIX_MEM_NOT_ALIGNED] += 1; \
    file_alignment = rec_ref->file_rec->counters[POSIX_FILE_ALIGNMENT]; \
220
    if(file_alignment > 0 && (this_offset % file_alignment) != 0) \
221 222 223 224 225 226 227 228 229 230 231 232
        rec_ref->file_rec->counters[POSIX_FILE_NOT_ALIGNED] += 1; \
    if(rec_ref->last_io_type == DARSHAN_IO_WRITE) \
        rec_ref->file_rec->counters[POSIX_RW_SWITCHES] += 1; \
    rec_ref->last_io_type = DARSHAN_IO_READ; \
    if(rec_ref->file_rec->fcounters[POSIX_F_READ_START_TIMESTAMP] == 0) \
        rec_ref->file_rec->fcounters[POSIX_F_READ_START_TIMESTAMP] = __tm1; \
    rec_ref->file_rec->fcounters[POSIX_F_READ_END_TIMESTAMP] = __tm2; \
    if(rec_ref->file_rec->fcounters[POSIX_F_MAX_READ_TIME] < __elapsed) { \
        rec_ref->file_rec->fcounters[POSIX_F_MAX_READ_TIME] = __elapsed; \
        rec_ref->file_rec->counters[POSIX_MAX_READ_TIME_SIZE] = __ret; } \
    DARSHAN_TIMER_INC_NO_OVERLAP(rec_ref->file_rec->fcounters[POSIX_F_READ_TIME], \
        __tm1, __tm2, rec_ref->last_read_end); \
233 234
} while(0)

235 236
#define POSIX_RECORD_WRITE(__ret, __fd, __pwrite_flag, __pwrite_offset, __aligned, __stream_flag, __tm1, __tm2) do { \
    struct posix_file_record_ref* rec_ref; \
237
    size_t stride; \
238
    int64_t this_offset; \
239
    int64_t file_alignment; \
240 241
    double __elapsed = __tm2-__tm1; \
    if(__ret < 0) break; \
242 243 244
    if(!posix_runtime || instrumentation_disabled) break; /* XXX: PREMABLE??? */\
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &__fd, sizeof(int)); \
    if(!rec_ref) break; \
245
    if(__pwrite_flag) \
246 247
        this_offset = __pwrite_offset; \
    else \
248 249 250 251 252 253 254 255
        this_offset = rec_ref->offset; \
    if(this_offset > rec_ref->last_byte_written) \
        rec_ref->file_rec->counters[POSIX_SEQ_WRITES] += 1; \
    if(this_offset == (rec_ref->last_byte_written + 1)) \
        rec_ref->file_rec->counters[POSIX_CONSEC_WRITES] += 1; \
    if(this_offset > 0 && this_offset > rec_ref->last_byte_written \
        && rec_ref->last_byte_written != 0) \
        stride = this_offset - rec_ref->last_byte_written - 1; \
256
    else \
257
        stride = 0; \
258 259 260 261 262
    rec_ref->last_byte_written = this_offset + __ret - 1; \
    rec_ref->offset = this_offset + __ret; \
    if(rec_ref->file_rec->counters[POSIX_MAX_BYTE_WRITTEN] < (this_offset + __ret - 1)) \
        rec_ref->file_rec->counters[POSIX_MAX_BYTE_WRITTEN] = (this_offset + __ret - 1); \
    rec_ref->file_rec->counters[POSIX_BYTES_WRITTEN] += __ret; \
263
    if(__stream_flag) \
264
        rec_ref->file_rec->counters[POSIX_FWRITES] += 1; \
265
    else \
266 267 268 269 270 271 272 273
        rec_ref->file_rec->counters[POSIX_WRITES] += 1; \
    DARSHAN_BUCKET_INC(&(rec_ref->file_rec->counters[POSIX_SIZE_WRITE_0_100]), __ret); \
    darshan_common_val_counter(&rec_ref->access_root, &rec_ref->access_count, __ret, \
        &(rec_ref->file_rec->counters[POSIX_ACCESS1_ACCESS]), \
        &(rec_ref->file_rec->counters[POSIX_ACCESS1_COUNT])); \
    darshan_common_val_counter(&rec_ref->stride_root, &rec_ref->stride_count, stride, \
        &(rec_ref->file_rec->counters[POSIX_STRIDE1_STRIDE]), \
        &(rec_ref->file_rec->counters[POSIX_STRIDE1_COUNT])); \
274
    if(!__aligned) \
275 276
        rec_ref->file_rec->counters[POSIX_MEM_NOT_ALIGNED] += 1; \
    file_alignment = rec_ref->file_rec->counters[POSIX_FILE_ALIGNMENT]; \
277
    if(file_alignment > 0 && (this_offset % file_alignment) != 0) \
278 279 280 281 282 283 284 285 286 287 288 289
        rec_ref->file_rec->counters[POSIX_FILE_NOT_ALIGNED] += 1; \
    if(rec_ref->last_io_type == DARSHAN_IO_READ) \
        rec_ref->file_rec->counters[POSIX_RW_SWITCHES] += 1; \
    rec_ref->last_io_type = DARSHAN_IO_WRITE; \
    if(rec_ref->file_rec->fcounters[POSIX_F_WRITE_START_TIMESTAMP] == 0) \
        rec_ref->file_rec->fcounters[POSIX_F_WRITE_START_TIMESTAMP] = __tm1; \
    rec_ref->file_rec->fcounters[POSIX_F_WRITE_END_TIMESTAMP] = __tm2; \
    if(rec_ref->file_rec->fcounters[POSIX_F_MAX_WRITE_TIME] < __elapsed) { \
        rec_ref->file_rec->fcounters[POSIX_F_MAX_WRITE_TIME] = __elapsed; \
        rec_ref->file_rec->counters[POSIX_MAX_WRITE_TIME_SIZE] = __ret; } \
    DARSHAN_TIMER_INC_NO_OVERLAP(rec_ref->file_rec->fcounters[POSIX_F_WRITE_TIME], \
        __tm1, __tm2, rec_ref->last_write_end); \
290
} while(0)
291

292
#define POSIX_LOOKUP_RECORD_STAT(__path, __statbuf, __tm1, __tm2) do { \
293 294 295
    struct posix_file_record_ref* rec_ref; \
    darshan_record_id rec_id; \
    if(!posix_runtime || instrumentation_disabled) break; /* XXX: PREMABLE??? */\
296
    if(darshan_core_excluded_path(__path)) break; \
297 298 299 300 301 302 303
    rec_id = darshan_record_id_from_path(__path); \
    rec_ref = darshan_lookup_record_ref(posix_runtime->rec_id_hash, &rec_id, sizeof(darshan_record_id)); \
    if(!rec_ref) { \
        rec_ref = posix_track_new_file_record(rec_id, __path); \
    } \
    if(rec_ref) { \
        POSIX_RECORD_STAT(rec_ref, __statbuf, __tm1, __tm2); \
304 305 306
    } \
} while(0)

307 308 309 310
#define POSIX_RECORD_STAT(__rec_ref, __statbuf, __tm1, __tm2) do { \
    (__rec_ref)->file_rec->counters[POSIX_STATS] += 1; \
    DARSHAN_TIMER_INC_NO_OVERLAP((__rec_ref)->file_rec->fcounters[POSIX_F_META_TIME], \
        __tm1, __tm2, (__rec_ref)->last_meta_end); \
311 312
} while(0)

313

314 315 316 317
/**********************************************************
 *      Wrappers for POSIX I/O functions of interest      * 
 **********************************************************/

318
int DARSHAN_DECL(open)(const char *path, int flags, ...)
319 320 321 322 323
{
    int mode = 0;
    int ret;
    double tm1, tm2;

324 325
    MAP_OR_FAIL(open);

326
    if(flags & O_CREAT) 
327 328 329 330 331 332
    {
        va_list arg;
        va_start(arg, flags);
        mode = va_arg(arg, int);
        va_end(arg);

333
        tm1 = darshan_core_wtime();
334
        ret = __real_open(path, flags, mode);
335
        tm2 = darshan_core_wtime();
336 337 338
    }
    else
    {
339
        tm1 = darshan_core_wtime();
340
        ret = __real_open(path, flags);
341
        tm2 = darshan_core_wtime();
342 343
    }

344 345
    POSIX_LOCK();
    posix_runtime_initialize();
346
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
347
    POSIX_UNLOCK();
348 349 350 351

    return(ret);
}

Shane Snyder's avatar
Shane Snyder committed
352 353 354 355 356 357
int DARSHAN_DECL(open64)(const char *path, int flags, ...)
{
    int mode = 0;
    int ret;
    double tm1, tm2;

358
    MAP_OR_FAIL(open64);
Shane Snyder's avatar
Shane Snyder committed
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379

    if(flags & O_CREAT)
    {
        va_list arg;
        va_start(arg, flags);
        mode = va_arg(arg, int);
        va_end(arg);

        tm1 = darshan_core_wtime();
        ret = __real_open64(path, flags, mode);
        tm2 = darshan_core_wtime();
    }
    else
    {
        tm1 = darshan_core_wtime();
        ret = __real_open64(path, flags);
        tm2 = darshan_core_wtime();
    }

    POSIX_LOCK();
    posix_runtime_initialize();
380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(creat)(const char* path, mode_t mode)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(creat);

    tm1 = darshan_core_wtime();
    ret = __real_creat(path, mode);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(creat64)(const char* path, mode_t mode)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(creat64);
Shane Snyder's avatar
Shane Snyder committed
411

412 413 414 415 416 417
    tm1 = darshan_core_wtime();
    ret = __real_creat64(path, mode);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
Shane Snyder's avatar
Shane Snyder committed
418 419 420 421 422 423
    POSIX_RECORD_OPEN(ret, path, mode, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
FILE* DARSHAN_DECL(fopen)(const char *path, const char *mode)
{
    FILE* ret;
    int fd;
    double tm1, tm2;

    MAP_OR_FAIL(fopen);

    tm1 = darshan_core_wtime();
    ret = __real_fopen(path, mode);
    tm2 = darshan_core_wtime();

    if(ret == NULL)
        fd = -1;
    else
        fd = fileno(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(fd, path, 0, 1, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

FILE* DARSHAN_DECL(fopen64)(const char *path, const char *mode)
{
    FILE* ret;
    int fd;
    double tm1, tm2;

    MAP_OR_FAIL(fopen64);

    tm1 = darshan_core_wtime();
    ret = __real_fopen64(path, mode);
    tm2 = darshan_core_wtime();

    if(ret == NULL)
        fd = -1;
    else
        fd = fileno(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(fd, path, 0, 1, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
int DARSHAN_DECL(mkstemp)(char* template)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(mkstemp);

    tm1 = darshan_core_wtime();
    ret = __real_mkstemp(template);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(ret, template, 0, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(mkostemp)(char* template, int flags)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(mkostemp);

    tm1 = darshan_core_wtime();
    ret = __real_mkostemp(template, flags);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(ret, template, 0, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(mkstemps)(char* template, int suffixlen)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(mkstemps);

    tm1 = darshan_core_wtime();
    ret = __real_mkstemps(template, suffixlen);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(ret, template, 0, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(mkostemps)(char* template, int suffixlen, int flags)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(mkostemps);

    tm1 = darshan_core_wtime();
    ret = __real_mkostemps(template, suffixlen, flags);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_OPEN(ret, template, 0, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

550 551 552 553 554 555
ssize_t DARSHAN_DECL(read)(int fd, void *buf, size_t count)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

556
    MAP_OR_FAIL(read);
557

558
    if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1;
559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579

    tm1 = darshan_core_wtime();
    ret = __real_read(fd, buf, count);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_READ(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(write)(int fd, const void *buf, size_t count)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(write);

580
    if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1;
581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601

    tm1 = darshan_core_wtime();
    ret = __real_write(fd, buf, count);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_WRITE(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pread)(int fd, void *buf, size_t count, off_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pread);

602
    if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1;
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623

    tm1 = darshan_core_wtime();
    ret = __real_pread(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_READ(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pwrite)(int fd, const void *buf, size_t count, off_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pwrite);

624
    if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1;
625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645

    tm1 = darshan_core_wtime();
    ret = __real_pwrite(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_WRITE(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pread64)(int fd, void *buf, size_t count, off64_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pread64);

646
    if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1;
647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667

    tm1 = darshan_core_wtime();
    ret = __real_pread64(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_READ(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(pwrite64)(int fd, const void *buf, size_t count, off64_t offset)
{
    ssize_t ret;
    int aligned_flag = 0;
    double tm1, tm2;

    MAP_OR_FAIL(pwrite64);

668
    if((unsigned long)buf % darshan_mem_alignment == 0) aligned_flag = 1;
669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684

    tm1 = darshan_core_wtime();
    ret = __real_pwrite64(fd, buf, count, offset);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_RECORD_WRITE(ret, fd, 1, offset, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(readv)(int fd, const struct iovec *iov, int iovcnt)
{
    ssize_t ret;
685
    int aligned_flag = 1;
686 687 688 689
    int i;
    double tm1, tm2;

    MAP_OR_FAIL(readv);
690

691 692 693 694
    for(i=0; i<iovcnt; i++)
    {
        if(((unsigned long)iov[i].iov_base % darshan_mem_alignment) != 0)
            aligned_flag = 0;
695
    }
696 697 698 699 700 701

    tm1 = darshan_core_wtime();
    ret = __real_readv(fd, iov, iovcnt);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
702
    posix_runtime_initialize();
703 704 705 706 707 708 709 710 711
    POSIX_RECORD_READ(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(writev)(int fd, const struct iovec *iov, int iovcnt)
{
    ssize_t ret;
712
    int aligned_flag = 1;
713 714 715 716
    int i;
    double tm1, tm2;

    MAP_OR_FAIL(writev);
717

718 719 720 721
    for(i=0; i<iovcnt; i++)
    {
        if(((unsigned long)iov[i].iov_base % darshan_mem_alignment) != 0)
            aligned_flag = 0;
722
    }
723 724 725 726 727 728

    tm1 = darshan_core_wtime();
    ret = __real_writev(fd, iov, iovcnt);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
729
    posix_runtime_initialize();
730 731 732 733 734 735
    POSIX_RECORD_WRITE(ret, fd, 0, 0, aligned_flag, 0, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

736 737 738 739
size_t DARSHAN_DECL(fread)(void *ptr, size_t size, size_t nmemb, FILE *stream)
{
    size_t ret;
    int aligned_flag = 0;
740
    int fd;
741 742 743 744
    double tm1, tm2;

    MAP_OR_FAIL(fread);

745
    if((unsigned long)ptr % darshan_mem_alignment == 0) aligned_flag = 1;
746 747 748 749 750 751 752

    tm1 = darshan_core_wtime();
    ret = __real_fread(ptr, size, nmemb, stream);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
753
    fd = fileno(stream);
754 755
    if(ret > 0)
    {
756
        POSIX_RECORD_READ(size*ret, fd, 0, 0, aligned_flag, 1, tm1, tm2);
757 758 759
    }
    else
    {
760
        POSIX_RECORD_READ(ret, fd, 0, 0, aligned_flag, 1, tm1, tm2);
761 762 763 764 765 766 767 768 769 770
    }
    POSIX_UNLOCK();

    return(ret);
}

size_t DARSHAN_DECL(fwrite)(const void *ptr, size_t size, size_t nmemb, FILE *stream)
{
    size_t ret;
    int aligned_flag = 0;
771
    int fd;
772 773 774 775
    double tm1, tm2;

    MAP_OR_FAIL(fwrite);

776
    if((unsigned long)ptr % darshan_mem_alignment == 0) aligned_flag = 1;
777 778 779 780 781 782 783

    tm1 = darshan_core_wtime();
    ret = __real_fwrite(ptr, size, nmemb, stream);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
784
    fd = fileno(stream);
785 786
    if(ret > 0)
    {
787
        POSIX_RECORD_WRITE(size*ret, fd, 0, 0, aligned_flag, 1, tm1, tm2);
788 789 790
    }
    else
    {
791
        POSIX_RECORD_WRITE(ret, fd, 0, 0, aligned_flag, 1, tm1, tm2);
792 793 794 795 796 797 798 799 800
    }
    POSIX_UNLOCK();

    return(ret);
}

off_t DARSHAN_DECL(lseek)(int fd, off_t offset, int whence)
{
    off_t ret;
801
    struct posix_file_record_ref *rec_ref;
802 803 804 805 806 807 808 809 810 811 812 813
    double tm1, tm2;

    MAP_OR_FAIL(lseek);

    tm1 = darshan_core_wtime();
    ret = __real_lseek(fd, offset, whence);
    tm2 = darshan_core_wtime();

    if(ret >= 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
814 815 816
        if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */\
        rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
        if(rec_ref)
817
        {
818
            rec_ref->offset = ret;
819
            DARSHAN_TIMER_INC_NO_OVERLAP(
820 821 822
                rec_ref->file_rec->fcounters[POSIX_F_META_TIME],
                tm1, tm2, rec_ref->last_meta_end);
            rec_ref->file_rec->counters[POSIX_SEEKS] += 1;
823 824 825 826 827 828 829 830 831 832
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

off_t DARSHAN_DECL(lseek64)(int fd, off_t offset, int whence)
{
    off_t ret;
833
    struct posix_file_record_ref *rec_ref;
834 835 836 837 838 839 840 841 842 843 844 845
    double tm1, tm2;

    MAP_OR_FAIL(lseek64);

    tm1 = darshan_core_wtime();
    ret = __real_lseek64(fd, offset, whence);
    tm2 = darshan_core_wtime();

    if(ret >= 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
846 847 848
        if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
        rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
        if(rec_ref)
849
        {
850
            rec_ref->offset = ret;
851
            DARSHAN_TIMER_INC_NO_OVERLAP(
852 853 854
                rec_ref->file_rec->fcounters[POSIX_F_META_TIME],
                tm1, tm2, rec_ref->last_meta_end);
            rec_ref->file_rec->counters[POSIX_SEEKS] += 1;
855 856 857 858 859 860 861 862 863 864
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

int DARSHAN_DECL(fseek)(FILE *stream, long offset, int whence)
{
    int ret;
865 866
    struct posix_file_record_ref *rec_ref;
    int fd;
867 868 869 870 871 872 873 874 875 876 877 878
    double tm1, tm2;

    MAP_OR_FAIL(fseek);

    tm1 = darshan_core_wtime();
    ret = __real_fseek(stream, offset, whence);
    tm2 = darshan_core_wtime();

    if(ret >= 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
879 880 881 882
        if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
        fd = fileno(stream);
        rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
        if(rec_ref)
883
        {
884
            rec_ref->offset = ftell(stream);
885
            DARSHAN_TIMER_INC_NO_OVERLAP(
886 887 888
                rec_ref->file_rec->fcounters[POSIX_F_META_TIME],
                tm1, tm2, rec_ref->last_meta_end);
            rec_ref->file_rec->counters[POSIX_FSEEKS] += 1;
889 890 891 892 893 894 895
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986
int DARSHAN_DECL(__xstat)(int vers, const char *path, struct stat *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__xstat);

    tm1 = darshan_core_wtime();
    ret = __real___xstat(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__xstat64)(int vers, const char *path, struct stat64 *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__xstat64);

    tm1 = darshan_core_wtime();
    ret = __real___xstat64(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__lxstat)(int vers, const char *path, struct stat *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__lxstat);

    tm1 = darshan_core_wtime();
    ret = __real___lxstat(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__lxstat64)(int vers, const char *path, struct stat64 *buf)
{
    int ret;
    double tm1, tm2;

    MAP_OR_FAIL(__lxstat64);

    tm1 = darshan_core_wtime();
    ret = __real___lxstat64(vers, path, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
    POSIX_LOOKUP_RECORD_STAT(path, buf, tm1, tm2);
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__fxstat)(int vers, int fd, struct stat *buf)
{
    int ret;
987
    struct posix_file_record_ref *rec_ref;
988 989 990 991 992 993 994 995 996 997 998 999 1000
    double tm1, tm2;

    MAP_OR_FAIL(__fxstat);

    tm1 = darshan_core_wtime();
    ret = __real___fxstat(vers, fd, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
1001 1002 1003
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
1004
    {
1005
        POSIX_RECORD_STAT(rec_ref, buf, tm1, tm2);
1006 1007 1008 1009 1010 1011 1012 1013 1014
    }
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(__fxstat64)(int vers, int fd, struct stat64 *buf)
{
    int ret;
1015
    struct posix_file_record_ref *rec_ref;
1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
    double tm1, tm2;

    MAP_OR_FAIL(__fxstat64);

    tm1 = darshan_core_wtime();
    ret = __real___fxstat64(vers, fd, buf);
    tm2 = darshan_core_wtime();

    if(ret < 0 || !S_ISREG(buf->st_mode))
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
1029 1030 1031
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
1032
    {
1033
        POSIX_RECORD_STAT(rec_ref, buf, tm1, tm2);
1034 1035 1036 1037 1038 1039
    }
    POSIX_UNLOCK();

    return(ret);
}

Shane Snyder's avatar
Shane Snyder committed
1040 1041 1042 1043
void* DARSHAN_DECL(mmap)(void *addr, size_t length, int prot, int flags,
    int fd, off_t offset)
{
    void* ret;
1044
    struct posix_file_record_ref *rec_ref;
Shane Snyder's avatar
Shane Snyder committed
1045 1046 1047 1048 1049 1050 1051 1052 1053

    MAP_OR_FAIL(mmap);

    ret = __real_mmap(addr, length, prot, flags, fd, offset);
    if(ret == MAP_FAILED)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
1054 1055 1056
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
Shane Snyder's avatar
Shane Snyder committed
1057
    {
1058
        rec_ref->file_rec->counters[POSIX_MMAPS] += 1;
Shane Snyder's avatar
Shane Snyder committed
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
    }
    POSIX_UNLOCK();

    return(ret);
}

void* DARSHAN_DECL(mmap64)(void *addr, size_t length, int prot, int flags,
    int fd, off64_t offset)
{
    void* ret;
1069
    struct posix_file_record_ref *rec_ref;
Shane Snyder's avatar
Shane Snyder committed
1070 1071 1072 1073 1074 1075 1076 1077 1078

    MAP_OR_FAIL(mmap64);

    ret = __real_mmap64(addr, length, prot, flags, fd, offset);
    if(ret == MAP_FAILED)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
1079 1080 1081
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
Shane Snyder's avatar
Shane Snyder committed
1082
    {
1083
        rec_ref->file_rec->counters[POSIX_MMAPS] += 1;
Shane Snyder's avatar
Shane Snyder committed
1084 1085 1086 1087 1088 1089
    }
    POSIX_UNLOCK();

    return(ret);
}

1090 1091 1092
int DARSHAN_DECL(fsync)(int fd)
{
    int ret;
1093
    struct posix_file_record_ref *rec_ref;
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
    double tm1, tm2;

    MAP_OR_FAIL(fsync);

    tm1 = darshan_core_wtime();
    ret = __real_fsync(fd);
    tm2 = darshan_core_wtime();

    if(ret < 0)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
1107 1108 1109
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
1110
    {
1111
        DARSHAN_TIMER_INC_NO_OVERLAP(
1112 1113 1114
            rec_ref->file_rec->fcounters[POSIX_F_WRITE_TIME],
            tm1, tm2, rec_ref->last_write_end);
        rec_ref->file_rec->counters[POSIX_FSYNCS] += 1;
1115 1116 1117 1118 1119 1120 1121 1122 1123
    }
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(fdatasync)(int fd)
{
    int ret;
1124
    struct posix_file_record_ref *rec_ref;
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
    double tm1, tm2;

    MAP_OR_FAIL(fdatasync);

    tm1 = darshan_core_wtime();
    ret = __real_fdatasync(fd);
    tm2 = darshan_core_wtime();

    if(ret < 0)
        return(ret);

    POSIX_LOCK();
    posix_runtime_initialize();
1138 1139 1140
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
1141
    {
1142
        DARSHAN_TIMER_INC_NO_OVERLAP(
1143 1144 1145
            rec_ref->file_rec->fcounters[POSIX_F_WRITE_TIME],
            tm1, tm2, rec_ref->last_write_end);
        rec_ref->file_rec->counters[POSIX_FDSYNCS] += 1;
1146 1147 1148 1149 1150 1151
    }
    POSIX_UNLOCK();

    return(ret);
}

1152 1153 1154
int DARSHAN_DECL(close)(int fd)
{
    int ret;
1155 1156
    struct posix_file_record_ref *rec_ref;
    double tm1, tm2;
1157 1158 1159 1160 1161 1162 1163 1164 1165

    MAP_OR_FAIL(close);

    tm1 = darshan_core_wtime();
    ret = __real_close(fd);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
1166 1167 1168
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
Shane Snyder's avatar
Shane Snyder committed
1169
    {
1170 1171 1172
        rec_ref->last_byte_written = 0;
        rec_ref->last_byte_read = 0;
        rec_ref->file_rec->fcounters[POSIX_F_CLOSE_TIMESTAMP] =
1173 1174
            darshan_core_wtime();
        DARSHAN_TIMER_INC_NO_OVERLAP(
1175 1176 1177
            rec_ref->file_rec->fcounters[POSIX_F_META_TIME],
            tm1, tm2, rec_ref->last_meta_end);
        darshan_delete_record_ref(&(posix_runtime->fd_hash), &fd, sizeof(int));
Shane Snyder's avatar
Shane Snyder committed
1178
    }
1179 1180 1181 1182 1183
    POSIX_UNLOCK();    

    return(ret);
}

1184 1185
int DARSHAN_DECL(fclose)(FILE *fp)
{
1186 1187
    int ret;
    struct posix_file_record_ref *rec_ref;
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
    int fd = fileno(fp);
    double tm1, tm2;

    MAP_OR_FAIL(fclose);

    tm1 = darshan_core_wtime();
    ret = __real_fclose(fp);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
1199 1200 1201
    if(!posix_runtime || instrumentation_disabled) return(ret); /* XXX: PREMABLE??? */
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
1202
    {
1203 1204 1205
        rec_ref->last_byte_written = 0;
        rec_ref->last_byte_read = 0;
        rec_ref->file_rec->fcounters[POSIX_F_CLOSE_TIMESTAMP] =
1206 1207
            darshan_core_wtime();
        DARSHAN_TIMER_INC_NO_OVERLAP(
1208 1209 1210
            rec_ref->file_rec->fcounters[POSIX_F_META_TIME],
            tm1, tm2, rec_ref->last_meta_end);
        darshan_delete_record_ref(&(posix_runtime->fd_hash), &fd, sizeof(int));
1211 1212 1213 1214 1215 1216
    }
    POSIX_UNLOCK();

    return(ret);
}

1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
int DARSHAN_DECL(aio_read)(struct aiocb *aiocbp)
{
    int ret;

    MAP_OR_FAIL(aio_read);

    ret = __real_aio_read(aiocbp);
    if(ret == 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        posix_aio_tracker_add(aiocbp->aio_fildes, aiocbp);
        POSIX_UNLOCK();
    }

    return(ret);
}

int DARSHAN_DECL(aio_write)(struct aiocb *aiocbp)
{
    int ret;

    MAP_OR_FAIL(aio_write);

    ret = __real_aio_write(aiocbp);
    if(ret == 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        posix_aio_tracker_add(aiocbp->aio_fildes, aiocbp);
        POSIX_UNLOCK();
    }

    return(ret);
}

int DARSHAN_DECL(aio_read64)(struct aiocb64 *aiocbp)
{
    int ret;

    MAP_OR_FAIL(aio_read64);

    ret = __real_aio_read64(aiocbp);
    if(ret == 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        posix_aio_tracker_add(aiocbp->aio_fildes, aiocbp);
        POSIX_UNLOCK();
    }

    return(ret);
}

int DARSHAN_DECL(aio_write64)(struct aiocb64 *aiocbp)
{
    int ret;

    MAP_OR_FAIL(aio_write64);

    ret = __real_aio_write64(aiocbp);
    if(ret == 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        posix_aio_tracker_add(aiocbp->aio_fildes, aiocbp);
        POSIX_UNLOCK();
    }

    return(ret);
}

ssize_t DARSHAN_DECL(aio_return)(struct aiocb *aiocbp)
{
    int ret;
    double tm2;
    struct posix_aio_tracker *tmp;
    int aligned_flag = 0;

    MAP_OR_FAIL(aio_return);

    ret = __real_aio_return(aiocbp);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    tmp = posix_aio_tracker_del(aiocbp->aio_fildes, aiocbp);
1304
    if(tmp)
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
    {
        if((unsigned long)aiocbp->aio_buf % darshan_mem_alignment == 0)
            aligned_flag = 1;
        if(aiocbp->aio_lio_opcode == LIO_WRITE)
        {
            POSIX_RECORD_WRITE(ret, aiocbp->aio_fildes,
                1, aiocbp->aio_offset, aligned_flag, 0,
                tmp->tm1, tm2);
        }
        else if(aiocbp->aio_lio_opcode == LIO_READ)
        {
            POSIX_RECORD_READ(ret, aiocbp->aio_fildes,
                1, aiocbp->aio_offset, aligned_flag, 0,
                tmp->tm1, tm2);
        }
        free(tmp);
    }
    POSIX_UNLOCK();

    return(ret);
}

ssize_t DARSHAN_DECL(aio_return64)(struct aiocb64 *aiocbp)
{
    int ret;
    double tm2;
    struct posix_aio_tracker *tmp;
    int aligned_flag = 0;

    MAP_OR_FAIL(aio_return64);

    ret = __real_aio_return64(aiocbp);
    tm2 = darshan_core_wtime();

    POSIX_LOCK();
    posix_runtime_initialize();
    tmp = posix_aio_tracker_del(aiocbp->aio_fildes, aiocbp);
1342
    if(tmp)
1343 1344 1345 1346 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 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
    {
        if((unsigned long)aiocbp->aio_buf % darshan_mem_alignment == 0)
            aligned_flag = 1;
        if(aiocbp->aio_lio_opcode == LIO_WRITE)
        {
            POSIX_RECORD_WRITE(ret, aiocbp->aio_fildes,
                1, aiocbp->aio_offset, aligned_flag, 0,
                tmp->tm1, tm2);
        }
        else if(aiocbp->aio_lio_opcode == LIO_READ)
        {
            POSIX_RECORD_READ(ret, aiocbp->aio_fildes,
                1, aiocbp->aio_offset, aligned_flag, 0,
                tmp->tm1, tm2);
        }
        free(tmp);
    }
    POSIX_UNLOCK();

    return(ret);
}

int DARSHAN_DECL(lio_listio)(int mode, struct aiocb *const aiocb_list[],
    int nitems, struct sigevent *sevp)
{
    int ret;
    int i;

    MAP_OR_FAIL(lio_listio);

    ret = __real_lio_listio(mode, aiocb_list, nitems, sevp);
    if(ret == 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        for(i = 0; i < nitems; i++)
        {
            posix_aio_tracker_add(aiocb_list[i]->aio_fildes, aiocb_list[i]);
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

int DARSHAN_DECL(lio_listio64)(int mode, struct aiocb64 *const aiocb_list[],
    int nitems, struct sigevent *sevp)
{
    int ret;
    int i;

    MAP_OR_FAIL(lio_listio64);

    ret = __real_lio_listio64(mode, aiocb_list, nitems, sevp);
    if(ret == 0)
    {
        POSIX_LOCK();
        posix_runtime_initialize();
        for(i = 0; i < nitems; i++)
        {
            posix_aio_tracker_add(aiocb_list[i]->aio_fildes, aiocb_list[i]);
        }
        POSIX_UNLOCK();
    }

    return(ret);
}

1411 1412 1413
/**********************************************************
 * Internal functions for manipulating POSIX module state *
 **********************************************************/
1414

1415
/* initialize internal POSIX module data structures and register with darshan-core */
1416
static void posix_runtime_initialize()
1417
{
1418 1419
    struct darshan_module_funcs posix_mod_fns =
    {
1420
        .begin_shutdown = &posix_begin_shutdown,
1421
        .get_output_data = &posix_get_output_data,
1422
        .shutdown = &posix_shutdown
1423
    };
1424
    int psx_buf_size;
1425

Shane Snyder's avatar
Shane Snyder committed
1426 1427
    /* don't do anything if already initialized or instrumenation is disabled */
    if(posix_runtime || instrumentation_disabled)
1428
        return;
1429

1430 1431 1432
    /* try and store the default number of records for this module */
    psx_buf_size = DARSHAN_DEF_MOD_REC_COUNT * sizeof(struct darshan_posix_file);

1433 1434
    /* register the posix module with darshan core */
    darshan_core_register_module(
1435
        DARSHAN_POSIX_MOD,
1436
        &posix_mod_fns,
1437
        &psx_buf_size,
1438
        &my_rank,
1439
        &darshan_mem_alignment);
1440

1441 1442
    /* return if darshan-core does not provide enough module memory */
    if(psx_buf_size < sizeof(struct darshan_posix_file))
1443 1444
    {
        darshan_core_unregister_module(DARSHAN_POSIX_MOD);
1445
        return;
1446
    }
1447 1448 1449

    posix_runtime = malloc(sizeof(*posix_runtime));
    if(!posix_runtime)
1450 1451
    {
        darshan_core_unregister_module(DARSHAN_POSIX_MOD);
1452
        return;
1453
    }
1454 1455 1456 1457
    memset(posix_runtime, 0, sizeof(*posix_runtime));

    return;
}
1458

1459 1460
static struct posix_file_record_ref *posix_track_new_file_record(
    darshan_record_id rec_id, const char *path)
1461
{
1462
    struct darshan_posix_file *file_rec = NULL;
1463 1464
    struct posix_file_record_ref *rec_ref = NULL;
    char *newpath = NULL;
1465
    int file_alignment;
1466
    int ret;
1467

1468 1469
    rec_ref = malloc(sizeof(*rec_ref));
    if(!rec_ref)
1470
        return(NULL);
1471
    memset(rec_ref, 0, sizeof(*rec_ref));
1472

1473 1474 1475 1476
    /* add a reference to this file record based on record id */
    ret = darshan_add_record_ref(&(posix_runtime->rec_id_hash), &rec_id,
        sizeof(darshan_record_id), rec_ref);
    if(ret == 0)
1477
    {
1478
        free(rec_ref);
1479 1480 1481
        return(NULL);
    }

1482 1483 1484 1485
    /* cleanup name and convert to absolute path */
    newpath = darshan_clean_file_path(path);
    if(!newpath)
        newpath = (char *)path;
Shane Snyder's avatar
Shane Snyder committed
1486

1487 1488 1489 1490 1491 1492
    /* register the actual file record with darshan-core so it is persisted
     * in the log file
     */
    file_rec = darshan_core_register_record(rec_id, newpath, DARSHAN_POSIX_MOD,
        sizeof(struct darshan_posix_file), &file_alignment);
    if(!file_rec)
1493
    {
1494 1495 1496 1497 1498 1499
        darshan_delete_record_ref(&(posix_runtime->rec_id_hash),
            &rec_id, sizeof(darshan_record_id));
        free(rec_ref);
        if(newpath != path)
            free(newpath);
        return(NULL);
1500 1501
    }

1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512
    /* registering this file record was successful, so initialize some fields */
    file_rec->base_rec.id = rec_id;
    file_rec->base_rec.rank = my_rank;
    file_rec->counters[POSIX_MEM_ALIGNMENT] = darshan_mem_alignment;
    file_rec->counters[POSIX_FILE_ALIGNMENT] = file_alignment;
    rec_ref->file_rec = file_rec;
    posix_runtime->file_rec_count++;

    if(newpath != path)
        free(newpath);
    return(rec_ref);
1513 1514
}

1515
/* compare function for sorting file records by descending rank */
1516
static int posix_record_compare(const void* a_p, const void* b_p)
1517
{
1518 1519
    const struct darshan_posix_file* a = a_p;
    const struct darshan_posix_file* b = b_p;
1520

1521
    if(a->base_rec.rank < b->base_rec.rank)
1522
        return 1;
1523
    if(a->base_rec.rank > b->base_rec.rank)
1524 1525 1526 1527 1528
        return -1;

    return 0;
}

1529 1530 1531 1532 1533 1534 1535 1536
/* finds the tracker structure for a given aio operation, removes it from
 * the linked list for the darshan_file structure, and returns a pointer.  
 *
 * returns NULL if aio operation not found
 */
static struct posix_aio_tracker* posix_aio_tracker_del(int fd, void *aiocbp)
{
    struct posix_aio_tracker *tracker = NULL, *iter, *tmp;
1537
    struct posix_file_record_ref *rec_ref;
1538

1539 1540
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
1541
    {
1542
        LL_FOREACH_SAFE(rec_ref->aio_list, iter, tmp)
1543
        {
1544
            if(iter->aiocbp == aiocbp)
1545
            {
1546
                LL_DELETE(rec_ref->aio_list, iter);
1547 1548 1549 1550 1551 1552 1553 1554 1555
                tracker = iter;
                break;
            }
        }
    }

    return(tracker);
}

1556 1557 1558 1559
/* adds a tracker for the given aio operation */
static void posix_aio_tracker_add(int fd, void *aiocbp)
{
    struct posix_aio_tracker* tracker;
1560
    struct posix_file_record_ref *rec_ref;
1561

1562 1563
    rec_ref = darshan_lookup_record_ref(posix_runtime->fd_hash, &fd, sizeof(int));
    if(rec_ref)
1564 1565
    {
        tracker = malloc(sizeof(*tracker));
1566
        if(tracker)
1567 1568 1569
        {
            tracker->tm1 = darshan_core_wtime();
            tracker->aiocbp = aiocbp;
1570
            LL_PREPEND(rec_ref->aio_list, tracker);
1571 1572 1573 1574 1575 1576 1577 1578
        }
    }

    return;
}

static void posix_record_reduction_op(void* infile_v, void* inoutfile_v,
    int *len, MPI_Datatype *datatype)
Shane Snyder's avatar
Shane Snyder committed
1579
{
1580 1581 1582
    struct darshan_posix_file tmp_file;
    struct darshan_posix_file *infile = infile_v;
    struct darshan_posix_file *inoutfile = inoutfile_v;
1583
    int i, j, k;
1584

1585
    for(i=0; i<*len; i++)
1586 1587
    {
        memset(&tmp_file, 0, sizeof(struct darshan_posix_file));
1588 1589
        tmp_file.base_rec.id = infile->base_rec.id;
        tmp_file.base_rec.rank = -1;
1590

1591
        /* sum */
1592
        for(j=POSIX_OPENS; j<=POSIX_FDSYNCS; j++)
1593 1594 1595 1596
        {
            tmp_file.counters[j] = infile->counters[j] + inoutfile->counters[j];
        }

1597
        tmp_file.counters[POSIX_MODE] = infile->counters[POSIX_MODE];
1598

1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614
        /* sum */
        for(j=POSIX_BYTES_READ; j<=POSIX_BYTES_WRITTEN; j++)
        {
            tmp_file.counters[j] = infile->counters[j] + inoutfile->counters[j];
        }

        /* max */
        for(j=POSIX_MAX_BYTE_READ; j<=POSIX_MAX_BYTE_WRITTEN; j++)
        {
            tmp_file.counters[j] = (
                (infile->counters[j] > inoutfile->counters[j]) ?
                infile->counters[j] :
                inoutfile->counters[j]);
        }

        /* sum */
1615 1616
        for(j=POSIX_CONSEC_READS; j<=POSIX_MEM_NOT_ALIGNED; j++)
        {
1617
            tmp_file.counters[j] = infile->counters[j] + inoutfile->counters[j];
1618 1619
        }

1620
        tmp_file.counters[POSIX_MEM_ALIGNMENT] = infile->counters[POSIX_MEM_ALIGNMENT];
1621 1622 1623

        /* sum */
        for(j=POSIX_FILE_NOT_ALIGNED; j<=POSIX_FILE_NOT_ALIGNED; j++)
1624
        {
1625
            tmp_file.counters[j] = infile->counters[j] + inoutfile->counters[j];
1626 1627
        }

1628
        tmp_file.counters[POSIX_FILE_ALIGNMENT] = infile->counters[POSIX_FILE_ALIGNMENT];
1629

1630 1631 1632 1633
        /* skip POSIX_MAX_*_TIME_SIZE; handled in floating point section */

        for(j=POSIX_SIZE_READ_0_100; j<=POSIX_SIZE_WRITE_1G_PLUS; j++)
        {
1634
            tmp_file.counters[j] = infile->counters[j] + inoutfile->counters[j];
1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653
        }

        /* first collapse any duplicates */
        for(j=POSIX_STRIDE1_STRIDE; j<=POSIX_STRIDE4_STRIDE; j++)
        {
            for(k=POSIX_STRIDE1_STRIDE; k<=POSIX_STRIDE4_STRIDE; k++)
            {
                if(infile->counters[j] == inoutfile->counters[k])
                {
                    infile->counters[j+4] += inoutfile->counters[k+4];
                    inoutfile->counters[k] = 0;
                    inoutfile->counters[k+4] = 0;
                }
            }
        }

        /* first set */
        for(j=POSIX_STRIDE1_STRIDE; j<=POSIX_STRIDE4_STRIDE; j++)
        {
1654 1655
            DARSHAN_COMMON_VAL_COUNTER_INC(&(tmp_file.counters[POSIX_STRIDE1_STRIDE]),
                &(tmp_file.counters[POSIX_STRIDE1_COUNT]), infile->counters[j],
1656
                infile->counters[j+4], 0);
1657 1658 1659 1660
        }
        /* second set */
        for(j=POSIX_STRIDE1_STRIDE; j<=POSIX_STRIDE4_STRIDE; j++)
        {
1661 1662
            DARSHAN_COMMON_VAL_COUNTER_INC(&(tmp_file.counters[POSIX_STRIDE1_STRIDE]),
                &(tmp_file.counters[POSIX_STRIDE1_COUNT]), inoutfile->counters[j],
1663
                inoutfile->counters[j+4], 0);
1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
        }

        /* same for access counts */

        /* first collapse any duplicates */
        for(j=POSIX_ACCESS1_ACCESS; j<=POSIX_ACCESS4_ACCESS; j++)
        {
            for(k=POSIX_ACCESS1_ACCESS; k<=POSIX_ACCESS4_ACCESS; k++)
            {
                if(infile->counters[j] == inoutfile->counters[k])
                {
                    infile->counters[j+4] += inoutfile->counters[k+4];
                    inoutfile->counters[k] = 0;
                    inoutfile->counters[k+4] = 0;
                }
            }
        }

        /* first set */
        for(j=POSIX_ACCESS1_ACCESS; j<=POSIX_ACCESS4_ACCESS; j++)
        {
1685 1686
            DARSHAN_COMMON_VAL_COUNTER_INC(&(tmp_file.counters[POSIX_ACCESS1_ACCESS]),
                &(tmp_file.counters[POSIX_ACCESS1_COUNT]), infile->counters[j],
1687
                infile->counters[j+4], 0);
1688 1689 1690 1691
        }
        /* second set */
        for(j=POSIX_ACCESS1_ACCESS; j<=POSIX_ACCESS4_ACCESS; j++)
        {
1692 1693
            DARSHAN_COMMON_VAL_COUNTER_INC(&(tmp_file.counters[POSIX_ACCESS1_ACCESS]),
                &(tmp_file.counters[POSIX_ACCESS1_COUNT]), inoutfile->counters[j],
1694
                inoutfile->counters[j+4], 0);
1695 1696
        }

1697 1698 1699
        /* min non-zero (if available) value */
        for(j=POSIX_F_OPEN_TIMESTAMP; j<=POSIX_F_WRITE_START_TIMESTAMP; j++)
        {
Shane Snyder's avatar
Shane Snyder committed
1700 1701
            if(infile->fcounters[j] > inoutfile->fcounters[j] &&
               inoutfile->fcounters[j] > 0)
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
                tmp_file.fcounters[j] = inoutfile->fcounters[j];
            else
                tmp_file.fcounters[j] = infile->fcounters[j];
        }

        /* max */
        for(j=POSIX_F_READ_END_TIMESTAMP; j<=POSIX_F_CLOSE_TIMESTAMP; j++)
        {
            if(infile->fcounters[j] > inoutfile->fcounters[j])
                tmp_file.fcounters[j] = infile->fcounters[j];
            else
                tmp_file.fcounters[j] = inoutfile->fcounters[j];
        }

        /* sum */
        for(j=POSIX_F_READ_TIME; j<=POSIX_F_META_TIME; j++)
        {
            tmp_file.fcounters[j] = infile->fcounters[j] + inoutfile->fcounters[j];
        }

        /* max (special case) */
        if(infile->fcounters[POSIX_F_MAX_READ_TIME] >
            inoutfile->fcounters[POSIX_F_MAX_READ_TIME])
        {
            tmp_file.fcounters[POSIX_F_MAX_READ_TIME] =
                infile->fcounters[POSIX_F_MAX_READ_TIME];
            tmp_file.counters[POSIX_MAX_READ_TIME_SIZE] =
                infile->counters[POSIX_MAX_READ_TIME_SIZE];
        }
1731
        else
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
        {
            tmp_file.fcounters[POSIX_F_MAX_READ_TIME] =
                inoutfile->fcounters[POSIX_F_MAX_READ_TIME];
            tmp_file.counters[POSIX_MAX_READ_TIME_SIZE] =