1 /* 2 * Copyright (c) 2016-present, Yann Collet, Facebook, Inc. 3 * All rights reserved. 4 * 5 * This source code is licensed under both the BSD-style license (found in the 6 * LICENSE file in the root directory of this source tree) and the GPLv2 (found 7 * in the COPYING file in the root directory of this source tree). 8 * You may select, at your option, one of the above-listed licenses. 9 */ 10 11 12 /* ************************************* 13 * Compiler Options 14 ***************************************/ 15 #ifdef _MSC_VER /* Visual */ 16 # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ 17 # pragma warning(disable : 4204) /* non-constant aggregate initializer */ 18 #endif 19 #if defined(__MINGW32__) && !defined(_POSIX_SOURCE) 20 # define _POSIX_SOURCE 1 /* disable %llu warnings with MinGW on Windows */ 21 #endif 22 23 /*-************************************* 24 * Includes 25 ***************************************/ 26 #include "platform.h" /* Large Files support, SET_BINARY_MODE */ 27 #include "util.h" /* UTIL_getFileSize, UTIL_isRegularFile, UTIL_isSameFile */ 28 #include <stdio.h> /* fprintf, fopen, fread, _fileno, stdin, stdout */ 29 #include <stdlib.h> /* malloc, free */ 30 #include <string.h> /* strcmp, strlen */ 31 #include <assert.h> 32 #include <errno.h> /* errno */ 33 #include <signal.h> 34 #include "timefn.h" /* UTIL_getTime, UTIL_clockSpanMicro */ 35 36 #if defined (_MSC_VER) 37 # include <sys/stat.h> 38 # include <io.h> 39 #endif 40 41 #include "mem.h" /* U32, U64 */ 42 #include "fileio.h" 43 44 #define ZSTD_STATIC_LINKING_ONLY /* ZSTD_magicNumber, ZSTD_frameHeaderSize_max */ 45 #include "zstd.h" 46 #include "zstd_errors.h" /* ZSTD_error_frameParameter_windowTooLarge */ 47 48 #if defined(ZSTD_GZCOMPRESS) || defined(ZSTD_GZDECOMPRESS) 49 # include <zlib.h> 50 # if !defined(z_const) 51 # define z_const 52 # endif 53 #endif 54 55 #if defined(ZSTD_LZMACOMPRESS) || defined(ZSTD_LZMADECOMPRESS) 56 # include <lzma.h> 57 #endif 58 59 #define LZ4_MAGICNUMBER 0x184D2204 60 #if defined(ZSTD_LZ4COMPRESS) || defined(ZSTD_LZ4DECOMPRESS) 61 # define LZ4F_ENABLE_OBSOLETE_ENUMS 62 # include <lz4frame.h> 63 # include <lz4.h> 64 #endif 65 66 67 /*-************************************* 68 * Constants 69 ***************************************/ 70 #define KB *(1<<10) 71 #define MB *(1<<20) 72 #define GB *(1U<<30) 73 74 #define ADAPT_WINDOWLOG_DEFAULT 23 /* 8 MB */ 75 #define DICTSIZE_MAX (32 MB) /* protection against large input (attack scenario) */ 76 77 #define FNSPACE 30 78 79 80 /*-************************************* 81 * Macros 82 ***************************************/ 83 84 struct FIO_display_prefs_s { 85 int displayLevel; /* 0 : no display; 1: errors; 2: + result + interaction + warnings; 3: + progression; 4: + information */ 86 U32 noProgress; 87 }; 88 89 static FIO_display_prefs_t g_display_prefs = {2, 0}; 90 91 #define DISPLAY(...) fprintf(stderr, __VA_ARGS__) 92 #define DISPLAYOUT(...) fprintf(stdout, __VA_ARGS__) 93 #define DISPLAYLEVEL(l, ...) { if (g_display_prefs.displayLevel>=l) { DISPLAY(__VA_ARGS__); } } 94 95 static const U64 g_refreshRate = SEC_TO_MICRO / 6; 96 static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER; 97 98 #define READY_FOR_UPDATE() (!g_display_prefs.noProgress && UTIL_clockSpanMicro(g_displayClock) > g_refreshRate) 99 #define DELAY_NEXT_UPDATE() { g_displayClock = UTIL_getTime(); } 100 #define DISPLAYUPDATE(l, ...) { \ 101 if (g_display_prefs.displayLevel>=l && !g_display_prefs.noProgress) { \ 102 if (READY_FOR_UPDATE() || (g_display_prefs.displayLevel>=4)) { \ 103 DELAY_NEXT_UPDATE(); \ 104 DISPLAY(__VA_ARGS__); \ 105 if (g_display_prefs.displayLevel>=4) fflush(stderr); \ 106 } } } 107 108 #undef MIN /* in case it would be already defined */ 109 #define MIN(a,b) ((a) < (b) ? (a) : (b)) 110 111 112 #define EXM_THROW(error, ...) \ 113 { \ 114 DISPLAYLEVEL(1, "zstd: "); \ 115 DISPLAYLEVEL(5, "Error defined at %s, line %i : \n", __FILE__, __LINE__); \ 116 DISPLAYLEVEL(1, "error %i : ", error); \ 117 DISPLAYLEVEL(1, __VA_ARGS__); \ 118 DISPLAYLEVEL(1, " \n"); \ 119 exit(error); \ 120 } 121 122 #define CHECK_V(v, f) \ 123 v = f; \ 124 if (ZSTD_isError(v)) { \ 125 DISPLAYLEVEL(5, "%s \n", #f); \ 126 EXM_THROW(11, "%s", ZSTD_getErrorName(v)); \ 127 } 128 #define CHECK(f) { size_t err; CHECK_V(err, f); } 129 130 131 /*-************************************ 132 * Signal (Ctrl-C trapping) 133 **************************************/ 134 static const char* g_artefact = NULL; 135 static void INThandler(int sig) 136 { 137 assert(sig==SIGINT); (void)sig; 138 #if !defined(_MSC_VER) 139 signal(sig, SIG_IGN); /* this invocation generates a buggy warning in Visual Studio */ 140 #endif 141 if (g_artefact) { 142 assert(UTIL_isRegularFile(g_artefact)); 143 remove(g_artefact); 144 } 145 DISPLAY("\n"); 146 exit(2); 147 } 148 static void addHandler(char const* dstFileName) 149 { 150 if (UTIL_isRegularFile(dstFileName)) { 151 g_artefact = dstFileName; 152 signal(SIGINT, INThandler); 153 } else { 154 g_artefact = NULL; 155 } 156 } 157 /* Idempotent */ 158 static void clearHandler(void) 159 { 160 if (g_artefact) signal(SIGINT, SIG_DFL); 161 g_artefact = NULL; 162 } 163 164 165 /*-********************************************************* 166 * Termination signal trapping (Print debug stack trace) 167 ***********************************************************/ 168 #if defined(__has_feature) && !defined(BACKTRACE_ENABLE) /* Clang compiler */ 169 # if (__has_feature(address_sanitizer)) 170 # define BACKTRACE_ENABLE 0 171 # endif /* __has_feature(address_sanitizer) */ 172 #elif defined(__SANITIZE_ADDRESS__) && !defined(BACKTRACE_ENABLE) /* GCC compiler */ 173 # define BACKTRACE_ENABLE 0 174 #endif 175 176 #if !defined(BACKTRACE_ENABLE) 177 /* automatic detector : backtrace enabled by default on linux+glibc and osx */ 178 # if (defined(__linux__) && (defined(__GLIBC__) && !defined(__UCLIBC__))) \ 179 || (defined(__APPLE__) && defined(__MACH__)) 180 # define BACKTRACE_ENABLE 1 181 # else 182 # define BACKTRACE_ENABLE 0 183 # endif 184 #endif 185 186 /* note : after this point, BACKTRACE_ENABLE is necessarily defined */ 187 188 189 #if BACKTRACE_ENABLE 190 191 #include <execinfo.h> /* backtrace, backtrace_symbols */ 192 193 #define MAX_STACK_FRAMES 50 194 195 static void ABRThandler(int sig) { 196 const char* name; 197 void* addrlist[MAX_STACK_FRAMES]; 198 char** symbollist; 199 int addrlen, i; 200 201 switch (sig) { 202 case SIGABRT: name = "SIGABRT"; break; 203 case SIGFPE: name = "SIGFPE"; break; 204 case SIGILL: name = "SIGILL"; break; 205 case SIGINT: name = "SIGINT"; break; 206 case SIGSEGV: name = "SIGSEGV"; break; 207 default: name = "UNKNOWN"; 208 } 209 210 DISPLAY("Caught %s signal, printing stack:\n", name); 211 /* Retrieve current stack addresses. */ 212 addrlen = backtrace(addrlist, MAX_STACK_FRAMES); 213 if (addrlen == 0) { 214 DISPLAY("\n"); 215 return; 216 } 217 /* Create readable strings to each frame. */ 218 symbollist = backtrace_symbols(addrlist, addrlen); 219 /* Print the stack trace, excluding calls handling the signal. */ 220 for (i = ZSTD_START_SYMBOLLIST_FRAME; i < addrlen; i++) { 221 DISPLAY("%s\n", symbollist[i]); 222 } 223 free(symbollist); 224 /* Reset and raise the signal so default handler runs. */ 225 signal(sig, SIG_DFL); 226 raise(sig); 227 } 228 #endif 229 230 void FIO_addAbortHandler() 231 { 232 #if BACKTRACE_ENABLE 233 signal(SIGABRT, ABRThandler); 234 signal(SIGFPE, ABRThandler); 235 signal(SIGILL, ABRThandler); 236 signal(SIGSEGV, ABRThandler); 237 signal(SIGBUS, ABRThandler); 238 #endif 239 } 240 241 242 /*-************************************************************ 243 * Avoid fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW 244 ***************************************************************/ 245 #if defined(_MSC_VER) && _MSC_VER >= 1400 246 # define LONG_SEEK _fseeki64 247 # define LONG_TELL _ftelli64 248 #elif !defined(__64BIT__) && (PLATFORM_POSIX_VERSION >= 200112L) /* No point defining Large file for 64 bit */ 249 # define LONG_SEEK fseeko 250 # define LONG_TELL ftello 251 #elif defined(__MINGW32__) && !defined(__STRICT_ANSI__) && !defined(__NO_MINGW_LFS) && defined(__MSVCRT__) 252 # define LONG_SEEK fseeko64 253 # define LONG_TELL ftello64 254 #elif defined(_WIN32) && !defined(__DJGPP__) 255 # include <windows.h> 256 static int LONG_SEEK(FILE* file, __int64 offset, int origin) { 257 LARGE_INTEGER off; 258 DWORD method; 259 off.QuadPart = offset; 260 if (origin == SEEK_END) 261 method = FILE_END; 262 else if (origin == SEEK_CUR) 263 method = FILE_CURRENT; 264 else 265 method = FILE_BEGIN; 266 267 if (SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, NULL, method)) 268 return 0; 269 else 270 return -1; 271 } 272 static __int64 LONG_TELL(FILE* file) { 273 LARGE_INTEGER off, newOff; 274 off.QuadPart = 0; 275 newOff.QuadPart = 0; 276 SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, &newOff, FILE_CURRENT); 277 return newOff.QuadPart; 278 } 279 #else 280 # define LONG_SEEK fseek 281 # define LONG_TELL ftell 282 #endif 283 284 285 /*-************************************* 286 * Parameters: Typedefs 287 ***************************************/ 288 289 struct FIO_prefs_s { 290 291 /* Algorithm preferences */ 292 FIO_compressionType_t compressionType; 293 U32 sparseFileSupport; /* 0: no sparse allowed; 1: auto (file yes, stdout no); 2: force sparse */ 294 int dictIDFlag; 295 int checksumFlag; 296 int blockSize; 297 int overlapLog; 298 U32 adaptiveMode; 299 int rsyncable; 300 int minAdaptLevel; 301 int maxAdaptLevel; 302 int ldmFlag; 303 int ldmHashLog; 304 int ldmMinMatch; 305 int ldmBucketSizeLog; 306 int ldmHashRateLog; 307 size_t targetCBlockSize; 308 ZSTD_literalCompressionMode_e literalCompressionMode; 309 310 /* IO preferences */ 311 U32 removeSrcFile; 312 U32 overwrite; 313 314 /* Computation resources preferences */ 315 unsigned memLimit; 316 int nbWorkers; 317 }; 318 319 320 /*-************************************* 321 * Parameters: Initialization 322 ***************************************/ 323 324 #define FIO_OVERLAP_LOG_NOTSET 9999 325 #define FIO_LDM_PARAM_NOTSET 9999 326 327 328 FIO_prefs_t* FIO_createPreferences(void) 329 { 330 FIO_prefs_t* const ret = (FIO_prefs_t*)malloc(sizeof(FIO_prefs_t)); 331 if (!ret) EXM_THROW(21, "Allocation error : not enough memory"); 332 333 ret->compressionType = FIO_zstdCompression; 334 ret->overwrite = 0; 335 ret->sparseFileSupport = ZSTD_SPARSE_DEFAULT; 336 ret->dictIDFlag = 1; 337 ret->checksumFlag = 1; 338 ret->removeSrcFile = 0; 339 ret->memLimit = 0; 340 ret->nbWorkers = 1; 341 ret->blockSize = 0; 342 ret->overlapLog = FIO_OVERLAP_LOG_NOTSET; 343 ret->adaptiveMode = 0; 344 ret->rsyncable = 0; 345 ret->minAdaptLevel = -50; /* initializing this value requires a constant, so ZSTD_minCLevel() doesn't work */ 346 ret->maxAdaptLevel = 22; /* initializing this value requires a constant, so ZSTD_maxCLevel() doesn't work */ 347 ret->ldmFlag = 0; 348 ret->ldmHashLog = 0; 349 ret->ldmMinMatch = 0; 350 ret->ldmBucketSizeLog = FIO_LDM_PARAM_NOTSET; 351 ret->ldmHashRateLog = FIO_LDM_PARAM_NOTSET; 352 ret->targetCBlockSize = 0; 353 ret->literalCompressionMode = ZSTD_lcm_auto; 354 return ret; 355 } 356 357 void FIO_freePreferences(FIO_prefs_t* const prefs) 358 { 359 free(prefs); 360 } 361 362 363 /*-************************************* 364 * Parameters: Display Options 365 ***************************************/ 366 367 void FIO_setNotificationLevel(int level) { g_display_prefs.displayLevel=level; } 368 369 void FIO_setNoProgress(unsigned noProgress) { g_display_prefs.noProgress = noProgress; } 370 371 372 /*-************************************* 373 * Parameters: Setters 374 ***************************************/ 375 376 void FIO_setCompressionType(FIO_prefs_t* const prefs, FIO_compressionType_t compressionType) { prefs->compressionType = compressionType; } 377 378 void FIO_overwriteMode(FIO_prefs_t* const prefs) { prefs->overwrite = 1; } 379 380 void FIO_setSparseWrite(FIO_prefs_t* const prefs, unsigned sparse) { prefs->sparseFileSupport = sparse; } 381 382 void FIO_setDictIDFlag(FIO_prefs_t* const prefs, int dictIDFlag) { prefs->dictIDFlag = dictIDFlag; } 383 384 void FIO_setChecksumFlag(FIO_prefs_t* const prefs, int checksumFlag) { prefs->checksumFlag = checksumFlag; } 385 386 void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, unsigned flag) { prefs->removeSrcFile = (flag>0); } 387 388 void FIO_setMemLimit(FIO_prefs_t* const prefs, unsigned memLimit) { prefs->memLimit = memLimit; } 389 390 void FIO_setNbWorkers(FIO_prefs_t* const prefs, int nbWorkers) { 391 #ifndef ZSTD_MULTITHREAD 392 if (nbWorkers > 0) DISPLAYLEVEL(2, "Note : multi-threading is disabled \n"); 393 #endif 394 prefs->nbWorkers = nbWorkers; 395 } 396 397 void FIO_setBlockSize(FIO_prefs_t* const prefs, int blockSize) { 398 if (blockSize && prefs->nbWorkers==0) 399 DISPLAYLEVEL(2, "Setting block size is useless in single-thread mode \n"); 400 prefs->blockSize = blockSize; 401 } 402 403 void FIO_setOverlapLog(FIO_prefs_t* const prefs, int overlapLog){ 404 if (overlapLog && prefs->nbWorkers==0) 405 DISPLAYLEVEL(2, "Setting overlapLog is useless in single-thread mode \n"); 406 prefs->overlapLog = overlapLog; 407 } 408 409 void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, unsigned adapt) { 410 if ((adapt>0) && (prefs->nbWorkers==0)) 411 EXM_THROW(1, "Adaptive mode is not compatible with single thread mode \n"); 412 prefs->adaptiveMode = adapt; 413 } 414 415 void FIO_setRsyncable(FIO_prefs_t* const prefs, int rsyncable) { 416 if ((rsyncable>0) && (prefs->nbWorkers==0)) 417 EXM_THROW(1, "Rsyncable mode is not compatible with single thread mode \n"); 418 prefs->rsyncable = rsyncable; 419 } 420 421 void FIO_setTargetCBlockSize(FIO_prefs_t* const prefs, size_t targetCBlockSize) { 422 prefs->targetCBlockSize = targetCBlockSize; 423 } 424 425 void FIO_setLiteralCompressionMode( 426 FIO_prefs_t* const prefs, 427 ZSTD_literalCompressionMode_e mode) { 428 prefs->literalCompressionMode = mode; 429 } 430 431 void FIO_setAdaptMin(FIO_prefs_t* const prefs, int minCLevel) 432 { 433 #ifndef ZSTD_NOCOMPRESS 434 assert(minCLevel >= ZSTD_minCLevel()); 435 #endif 436 prefs->minAdaptLevel = minCLevel; 437 } 438 439 void FIO_setAdaptMax(FIO_prefs_t* const prefs, int maxCLevel) 440 { 441 prefs->maxAdaptLevel = maxCLevel; 442 } 443 444 void FIO_setLdmFlag(FIO_prefs_t* const prefs, unsigned ldmFlag) { 445 prefs->ldmFlag = (ldmFlag>0); 446 } 447 448 void FIO_setLdmHashLog(FIO_prefs_t* const prefs, int ldmHashLog) { 449 prefs->ldmHashLog = ldmHashLog; 450 } 451 452 void FIO_setLdmMinMatch(FIO_prefs_t* const prefs, int ldmMinMatch) { 453 prefs->ldmMinMatch = ldmMinMatch; 454 } 455 456 void FIO_setLdmBucketSizeLog(FIO_prefs_t* const prefs, int ldmBucketSizeLog) { 457 prefs->ldmBucketSizeLog = ldmBucketSizeLog; 458 } 459 460 461 void FIO_setLdmHashRateLog(FIO_prefs_t* const prefs, int ldmHashRateLog) { 462 prefs->ldmHashRateLog = ldmHashRateLog; 463 } 464 465 466 /*-************************************* 467 * Functions 468 ***************************************/ 469 /** FIO_remove() : 470 * @result : Unlink `fileName`, even if it's read-only */ 471 static int FIO_remove(const char* path) 472 { 473 if (!UTIL_isRegularFile(path)) { 474 DISPLAYLEVEL(2, "zstd: Refusing to remove non-regular file %s \n", path); 475 return 0; 476 } 477 #if defined(_WIN32) || defined(WIN32) 478 /* windows doesn't allow remove read-only files, 479 * so try to make it writable first */ 480 chmod(path, _S_IWRITE); 481 #endif 482 return remove(path); 483 } 484 485 /** FIO_openSrcFile() : 486 * condition : `srcFileName` must be non-NULL. 487 * @result : FILE* to `srcFileName`, or NULL if it fails */ 488 static FILE* FIO_openSrcFile(const char* srcFileName) 489 { 490 assert(srcFileName != NULL); 491 if (!strcmp (srcFileName, stdinmark)) { 492 DISPLAYLEVEL(4,"Using stdin for input \n"); 493 SET_BINARY_MODE(stdin); 494 return stdin; 495 } 496 497 if (!UTIL_fileExist(srcFileName)) { 498 DISPLAYLEVEL(1, "zstd: can't stat %s : %s -- ignored \n", 499 srcFileName, strerror(errno)); 500 return NULL; 501 } 502 503 if (!UTIL_isRegularFile(srcFileName)) { 504 DISPLAYLEVEL(1, "zstd: %s is not a regular file -- ignored \n", 505 srcFileName); 506 return NULL; 507 } 508 509 { FILE* const f = fopen(srcFileName, "rb"); 510 if (f == NULL) 511 DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno)); 512 return f; 513 } 514 } 515 516 /** FIO_openDstFile() : 517 * condition : `dstFileName` must be non-NULL. 518 * @result : FILE* to `dstFileName`, or NULL if it fails */ 519 static FILE* FIO_openDstFile(FIO_prefs_t* const prefs, const char* srcFileName, const char* dstFileName) 520 { 521 assert(dstFileName != NULL); 522 if (!strcmp (dstFileName, stdoutmark)) { 523 DISPLAYLEVEL(4,"Using stdout for output \n"); 524 SET_BINARY_MODE(stdout); 525 if (prefs->sparseFileSupport == 1) { 526 prefs->sparseFileSupport = 0; 527 DISPLAYLEVEL(4, "Sparse File Support is automatically disabled on stdout ; try --sparse \n"); 528 } 529 return stdout; 530 } 531 532 /* ensure dst is not the same as src */ 533 if (srcFileName != NULL && UTIL_isSameFile(srcFileName, dstFileName)) { 534 DISPLAYLEVEL(1, "zstd: Refusing to open an output file which will overwrite the input file \n"); 535 return NULL; 536 } 537 538 if (prefs->sparseFileSupport == 1) { 539 prefs->sparseFileSupport = ZSTD_SPARSE_DEFAULT; 540 } 541 542 if (UTIL_isRegularFile(dstFileName)) { 543 /* Check if destination file already exists */ 544 FILE* const fCheck = fopen( dstFileName, "rb" ); 545 if (!strcmp(dstFileName, nulmark)) { 546 EXM_THROW(40, "%s is unexpectedly categorized as a regular file", 547 dstFileName); 548 } 549 if (fCheck != NULL) { /* dst file exists, authorization prompt */ 550 fclose(fCheck); 551 if (!prefs->overwrite) { 552 if (g_display_prefs.displayLevel <= 1) { 553 /* No interaction possible */ 554 DISPLAY("zstd: %s already exists; not overwritten \n", 555 dstFileName); 556 return NULL; 557 } 558 DISPLAY("zstd: %s already exists; overwrite (y/N) ? ", 559 dstFileName); 560 { int ch = getchar(); 561 if ((ch!='Y') && (ch!='y')) { 562 DISPLAY(" not overwritten \n"); 563 return NULL; 564 } 565 /* flush rest of input line */ 566 while ((ch!=EOF) && (ch!='\n')) ch = getchar(); 567 } } 568 /* need to unlink */ 569 FIO_remove(dstFileName); 570 } } 571 572 { FILE* const f = fopen( dstFileName, "wb" ); 573 if (f == NULL) { 574 DISPLAYLEVEL(1, "zstd: %s: %s\n", dstFileName, strerror(errno)); 575 } else { 576 chmod(dstFileName, 00600); 577 } 578 return f; 579 } 580 } 581 582 583 /*! FIO_createDictBuffer() : 584 * creates a buffer, pointed by `*bufferPtr`, 585 * loads `filename` content into it, up to DICTSIZE_MAX bytes. 586 * @return : loaded size 587 * if fileName==NULL, returns 0 and a NULL pointer 588 */ 589 static size_t FIO_createDictBuffer(void** bufferPtr, const char* fileName) 590 { 591 FILE* fileHandle; 592 U64 fileSize; 593 594 assert(bufferPtr != NULL); 595 *bufferPtr = NULL; 596 if (fileName == NULL) return 0; 597 598 DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName); 599 fileHandle = fopen(fileName, "rb"); 600 if (fileHandle==NULL) EXM_THROW(31, "%s: %s", fileName, strerror(errno)); 601 602 fileSize = UTIL_getFileSize(fileName); 603 if (fileSize > DICTSIZE_MAX) { 604 EXM_THROW(32, "Dictionary file %s is too large (> %u MB)", 605 fileName, DICTSIZE_MAX >> 20); /* avoid extreme cases */ 606 } 607 *bufferPtr = malloc((size_t)fileSize); 608 if (*bufferPtr==NULL) EXM_THROW(34, "%s", strerror(errno)); 609 { size_t const readSize = fread(*bufferPtr, 1, (size_t)fileSize, fileHandle); 610 if (readSize != fileSize) 611 EXM_THROW(35, "Error reading dictionary file %s : %s", 612 fileName, strerror(errno)); 613 } 614 fclose(fileHandle); 615 return (size_t)fileSize; 616 } 617 618 #ifndef ZSTD_NOCOMPRESS 619 620 /* ********************************************************************** 621 * Compression 622 ************************************************************************/ 623 typedef struct { 624 FILE* srcFile; 625 FILE* dstFile; 626 void* srcBuffer; 627 size_t srcBufferSize; 628 void* dstBuffer; 629 size_t dstBufferSize; 630 const char* dictFileName; 631 ZSTD_CStream* cctx; 632 } cRess_t; 633 634 static cRess_t FIO_createCResources(FIO_prefs_t* const prefs, 635 const char* dictFileName, int cLevel, 636 U64 srcSize, 637 ZSTD_compressionParameters comprParams) { 638 cRess_t ress; 639 memset(&ress, 0, sizeof(ress)); 640 641 DISPLAYLEVEL(6, "FIO_createCResources \n"); 642 ress.cctx = ZSTD_createCCtx(); 643 if (ress.cctx == NULL) 644 EXM_THROW(30, "allocation error (%s): can't create ZSTD_CCtx", 645 strerror(errno)); 646 ress.srcBufferSize = ZSTD_CStreamInSize(); 647 ress.srcBuffer = malloc(ress.srcBufferSize); 648 ress.dstBufferSize = ZSTD_CStreamOutSize(); 649 ress.dstBuffer = malloc(ress.dstBufferSize); 650 if (!ress.srcBuffer || !ress.dstBuffer) 651 EXM_THROW(31, "allocation error : not enough memory"); 652 653 /* Advanced parameters, including dictionary */ 654 { void* dictBuffer; 655 size_t const dictBuffSize = FIO_createDictBuffer(&dictBuffer, dictFileName); /* works with dictFileName==NULL */ 656 if (dictFileName && (dictBuffer==NULL)) 657 EXM_THROW(32, "allocation error : can't create dictBuffer"); 658 ress.dictFileName = dictFileName; 659 660 if (prefs->adaptiveMode && !prefs->ldmFlag && !comprParams.windowLog) 661 comprParams.windowLog = ADAPT_WINDOWLOG_DEFAULT; 662 663 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_contentSizeFlag, 1) ); /* always enable content size when available (note: supposed to be default) */ 664 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_dictIDFlag, prefs->dictIDFlag) ); 665 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_checksumFlag, prefs->checksumFlag) ); 666 /* compression level */ 667 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, cLevel) ); 668 /* max compressed block size */ 669 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetCBlockSize, (int)prefs->targetCBlockSize) ); 670 /* long distance matching */ 671 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_enableLongDistanceMatching, prefs->ldmFlag) ); 672 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashLog, prefs->ldmHashLog) ); 673 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmMinMatch, prefs->ldmMinMatch) ); 674 if (prefs->ldmBucketSizeLog != FIO_LDM_PARAM_NOTSET) { 675 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmBucketSizeLog, prefs->ldmBucketSizeLog) ); 676 } 677 if (prefs->ldmHashRateLog != FIO_LDM_PARAM_NOTSET) { 678 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashRateLog, prefs->ldmHashRateLog) ); 679 } 680 /* compression parameters */ 681 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_windowLog, (int)comprParams.windowLog) ); 682 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_chainLog, (int)comprParams.chainLog) ); 683 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_hashLog, (int)comprParams.hashLog) ); 684 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_searchLog, (int)comprParams.searchLog) ); 685 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_minMatch, (int)comprParams.minMatch) ); 686 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetLength, (int)comprParams.targetLength) ); 687 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_strategy, comprParams.strategy) ); 688 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_literalCompressionMode, (int)prefs->literalCompressionMode) ); 689 /* multi-threading */ 690 #ifdef ZSTD_MULTITHREAD 691 DISPLAYLEVEL(5,"set nb workers = %u \n", prefs->nbWorkers); 692 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_nbWorkers, prefs->nbWorkers) ); 693 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_jobSize, prefs->blockSize) ); 694 if (prefs->overlapLog != FIO_OVERLAP_LOG_NOTSET) { 695 DISPLAYLEVEL(3,"set overlapLog = %u \n", prefs->overlapLog); 696 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_overlapLog, prefs->overlapLog) ); 697 } 698 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_rsyncable, prefs->rsyncable) ); 699 #endif 700 /* dictionary */ 701 CHECK( ZSTD_CCtx_setPledgedSrcSize(ress.cctx, srcSize) ); /* set the value temporarily for dictionary loading, to adapt compression parameters */ 702 CHECK( ZSTD_CCtx_loadDictionary(ress.cctx, dictBuffer, dictBuffSize) ); 703 CHECK( ZSTD_CCtx_setPledgedSrcSize(ress.cctx, ZSTD_CONTENTSIZE_UNKNOWN) ); /* reset */ 704 705 free(dictBuffer); 706 } 707 708 return ress; 709 } 710 711 static void FIO_freeCResources(cRess_t ress) 712 { 713 free(ress.srcBuffer); 714 free(ress.dstBuffer); 715 ZSTD_freeCStream(ress.cctx); /* never fails */ 716 } 717 718 719 #ifdef ZSTD_GZCOMPRESS 720 static unsigned long long 721 FIO_compressGzFrame(cRess_t* ress, 722 const char* srcFileName, U64 const srcFileSize, 723 int compressionLevel, U64* readsize) 724 { 725 unsigned long long inFileSize = 0, outFileSize = 0; 726 z_stream strm; 727 int ret; 728 729 if (compressionLevel > Z_BEST_COMPRESSION) 730 compressionLevel = Z_BEST_COMPRESSION; 731 732 strm.zalloc = Z_NULL; 733 strm.zfree = Z_NULL; 734 strm.opaque = Z_NULL; 735 736 ret = deflateInit2(&strm, compressionLevel, Z_DEFLATED, 737 15 /* maxWindowLogSize */ + 16 /* gzip only */, 738 8, Z_DEFAULT_STRATEGY); /* see http://www.zlib.net/manual.html */ 739 if (ret != Z_OK) 740 EXM_THROW(71, "zstd: %s: deflateInit2 error %d \n", srcFileName, ret); 741 742 strm.next_in = 0; 743 strm.avail_in = 0; 744 strm.next_out = (Bytef*)ress->dstBuffer; 745 strm.avail_out = (uInt)ress->dstBufferSize; 746 747 while (1) { 748 if (strm.avail_in == 0) { 749 size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile); 750 if (inSize == 0) break; 751 inFileSize += inSize; 752 strm.next_in = (z_const unsigned char*)ress->srcBuffer; 753 strm.avail_in = (uInt)inSize; 754 } 755 ret = deflate(&strm, Z_NO_FLUSH); 756 if (ret != Z_OK) 757 EXM_THROW(72, "zstd: %s: deflate error %d \n", srcFileName, ret); 758 { size_t const decompBytes = ress->dstBufferSize - strm.avail_out; 759 if (decompBytes) { 760 if (fwrite(ress->dstBuffer, 1, decompBytes, ress->dstFile) != decompBytes) 761 EXM_THROW(73, "Write error : cannot write to output file"); 762 outFileSize += decompBytes; 763 strm.next_out = (Bytef*)ress->dstBuffer; 764 strm.avail_out = (uInt)ress->dstBufferSize; 765 } 766 } 767 if (srcFileSize == UTIL_FILESIZE_UNKNOWN) 768 DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%", 769 (unsigned)(inFileSize>>20), 770 (double)outFileSize/inFileSize*100) 771 else 772 DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%", 773 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20), 774 (double)outFileSize/inFileSize*100); 775 } 776 777 while (1) { 778 ret = deflate(&strm, Z_FINISH); 779 { size_t const decompBytes = ress->dstBufferSize - strm.avail_out; 780 if (decompBytes) { 781 if (fwrite(ress->dstBuffer, 1, decompBytes, ress->dstFile) != decompBytes) 782 EXM_THROW(75, "Write error : %s", strerror(errno)); 783 outFileSize += decompBytes; 784 strm.next_out = (Bytef*)ress->dstBuffer; 785 strm.avail_out = (uInt)ress->dstBufferSize; 786 } } 787 if (ret == Z_STREAM_END) break; 788 if (ret != Z_BUF_ERROR) 789 EXM_THROW(77, "zstd: %s: deflate error %d \n", srcFileName, ret); 790 } 791 792 ret = deflateEnd(&strm); 793 if (ret != Z_OK) 794 EXM_THROW(79, "zstd: %s: deflateEnd error %d \n", srcFileName, ret); 795 *readsize = inFileSize; 796 797 return outFileSize; 798 } 799 #endif 800 801 802 #ifdef ZSTD_LZMACOMPRESS 803 static unsigned long long 804 FIO_compressLzmaFrame(cRess_t* ress, 805 const char* srcFileName, U64 const srcFileSize, 806 int compressionLevel, U64* readsize, int plain_lzma) 807 { 808 unsigned long long inFileSize = 0, outFileSize = 0; 809 lzma_stream strm = LZMA_STREAM_INIT; 810 lzma_action action = LZMA_RUN; 811 lzma_ret ret; 812 813 if (compressionLevel < 0) compressionLevel = 0; 814 if (compressionLevel > 9) compressionLevel = 9; 815 816 if (plain_lzma) { 817 lzma_options_lzma opt_lzma; 818 if (lzma_lzma_preset(&opt_lzma, compressionLevel)) 819 EXM_THROW(71, "zstd: %s: lzma_lzma_preset error", srcFileName); 820 ret = lzma_alone_encoder(&strm, &opt_lzma); /* LZMA */ 821 if (ret != LZMA_OK) 822 EXM_THROW(71, "zstd: %s: lzma_alone_encoder error %d", srcFileName, ret); 823 } else { 824 ret = lzma_easy_encoder(&strm, compressionLevel, LZMA_CHECK_CRC64); /* XZ */ 825 if (ret != LZMA_OK) 826 EXM_THROW(71, "zstd: %s: lzma_easy_encoder error %d", srcFileName, ret); 827 } 828 829 strm.next_in = 0; 830 strm.avail_in = 0; 831 strm.next_out = (BYTE*)ress->dstBuffer; 832 strm.avail_out = ress->dstBufferSize; 833 834 while (1) { 835 if (strm.avail_in == 0) { 836 size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile); 837 if (inSize == 0) action = LZMA_FINISH; 838 inFileSize += inSize; 839 strm.next_in = (BYTE const*)ress->srcBuffer; 840 strm.avail_in = inSize; 841 } 842 843 ret = lzma_code(&strm, action); 844 845 if (ret != LZMA_OK && ret != LZMA_STREAM_END) 846 EXM_THROW(72, "zstd: %s: lzma_code encoding error %d", srcFileName, ret); 847 { size_t const compBytes = ress->dstBufferSize - strm.avail_out; 848 if (compBytes) { 849 if (fwrite(ress->dstBuffer, 1, compBytes, ress->dstFile) != compBytes) 850 EXM_THROW(73, "Write error : %s", strerror(errno)); 851 outFileSize += compBytes; 852 strm.next_out = (BYTE*)ress->dstBuffer; 853 strm.avail_out = ress->dstBufferSize; 854 } } 855 if (srcFileSize == UTIL_FILESIZE_UNKNOWN) 856 DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%", 857 (unsigned)(inFileSize>>20), 858 (double)outFileSize/inFileSize*100) 859 else 860 DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%", 861 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20), 862 (double)outFileSize/inFileSize*100); 863 if (ret == LZMA_STREAM_END) break; 864 } 865 866 lzma_end(&strm); 867 *readsize = inFileSize; 868 869 return outFileSize; 870 } 871 #endif 872 873 #ifdef ZSTD_LZ4COMPRESS 874 875 #if LZ4_VERSION_NUMBER <= 10600 876 #define LZ4F_blockLinked blockLinked 877 #define LZ4F_max64KB max64KB 878 #endif 879 880 static int FIO_LZ4_GetBlockSize_FromBlockId (int id) { return (1 << (8 + (2 * id))); } 881 882 static unsigned long long 883 FIO_compressLz4Frame(cRess_t* ress, 884 const char* srcFileName, U64 const srcFileSize, 885 int compressionLevel, int checksumFlag, 886 U64* readsize) 887 { 888 const size_t blockSize = FIO_LZ4_GetBlockSize_FromBlockId(LZ4F_max64KB); 889 unsigned long long inFileSize = 0, outFileSize = 0; 890 891 LZ4F_preferences_t prefs; 892 LZ4F_compressionContext_t ctx; 893 894 LZ4F_errorCode_t const errorCode = LZ4F_createCompressionContext(&ctx, LZ4F_VERSION); 895 if (LZ4F_isError(errorCode)) 896 EXM_THROW(31, "zstd: failed to create lz4 compression context"); 897 898 memset(&prefs, 0, sizeof(prefs)); 899 900 assert(blockSize <= ress->srcBufferSize); 901 902 prefs.autoFlush = 1; 903 prefs.compressionLevel = compressionLevel; 904 prefs.frameInfo.blockMode = LZ4F_blockLinked; 905 prefs.frameInfo.blockSizeID = LZ4F_max64KB; 906 prefs.frameInfo.contentChecksumFlag = (contentChecksum_t)checksumFlag; 907 #if LZ4_VERSION_NUMBER >= 10600 908 prefs.frameInfo.contentSize = (srcFileSize==UTIL_FILESIZE_UNKNOWN) ? 0 : srcFileSize; 909 #endif 910 assert(LZ4F_compressBound(blockSize, &prefs) <= ress->dstBufferSize); 911 912 { 913 size_t readSize; 914 size_t headerSize = LZ4F_compressBegin(ctx, ress->dstBuffer, ress->dstBufferSize, &prefs); 915 if (LZ4F_isError(headerSize)) 916 EXM_THROW(33, "File header generation failed : %s", 917 LZ4F_getErrorName(headerSize)); 918 if (fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile) != headerSize) 919 EXM_THROW(34, "Write error : %s (cannot write header)", strerror(errno)); 920 outFileSize += headerSize; 921 922 /* Read first block */ 923 readSize = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile); 924 inFileSize += readSize; 925 926 /* Main Loop */ 927 while (readSize>0) { 928 size_t const outSize = LZ4F_compressUpdate(ctx, 929 ress->dstBuffer, ress->dstBufferSize, 930 ress->srcBuffer, readSize, NULL); 931 if (LZ4F_isError(outSize)) 932 EXM_THROW(35, "zstd: %s: lz4 compression failed : %s", 933 srcFileName, LZ4F_getErrorName(outSize)); 934 outFileSize += outSize; 935 if (srcFileSize == UTIL_FILESIZE_UNKNOWN) { 936 DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%", 937 (unsigned)(inFileSize>>20), 938 (double)outFileSize/inFileSize*100) 939 } else { 940 DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%", 941 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20), 942 (double)outFileSize/inFileSize*100); 943 } 944 945 /* Write Block */ 946 { size_t const sizeCheck = fwrite(ress->dstBuffer, 1, outSize, ress->dstFile); 947 if (sizeCheck != outSize) 948 EXM_THROW(36, "Write error : %s", strerror(errno)); 949 } 950 951 /* Read next block */ 952 readSize = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile); 953 inFileSize += readSize; 954 } 955 if (ferror(ress->srcFile)) EXM_THROW(37, "Error reading %s ", srcFileName); 956 957 /* End of Stream mark */ 958 headerSize = LZ4F_compressEnd(ctx, ress->dstBuffer, ress->dstBufferSize, NULL); 959 if (LZ4F_isError(headerSize)) 960 EXM_THROW(38, "zstd: %s: lz4 end of file generation failed : %s", 961 srcFileName, LZ4F_getErrorName(headerSize)); 962 963 { size_t const sizeCheck = fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile); 964 if (sizeCheck != headerSize) 965 EXM_THROW(39, "Write error : %s (cannot write end of stream)", 966 strerror(errno)); 967 } 968 outFileSize += headerSize; 969 } 970 971 *readsize = inFileSize; 972 LZ4F_freeCompressionContext(ctx); 973 974 return outFileSize; 975 } 976 #endif 977 978 979 static unsigned long long 980 FIO_compressZstdFrame(FIO_prefs_t* const prefs, 981 const cRess_t* ressPtr, 982 const char* srcFileName, U64 fileSize, 983 int compressionLevel, U64* readsize) 984 { 985 cRess_t const ress = *ressPtr; 986 FILE* const srcFile = ress.srcFile; 987 FILE* const dstFile = ress.dstFile; 988 U64 compressedfilesize = 0; 989 ZSTD_EndDirective directive = ZSTD_e_continue; 990 991 /* stats */ 992 ZSTD_frameProgression previous_zfp_update = { 0, 0, 0, 0, 0, 0 }; 993 ZSTD_frameProgression previous_zfp_correction = { 0, 0, 0, 0, 0, 0 }; 994 typedef enum { noChange, slower, faster } speedChange_e; 995 speedChange_e speedChange = noChange; 996 unsigned flushWaiting = 0; 997 unsigned inputPresented = 0; 998 unsigned inputBlocked = 0; 999 unsigned lastJobID = 0; 1000 1001 DISPLAYLEVEL(6, "compression using zstd format \n"); 1002 1003 /* init */ 1004 if (fileSize != UTIL_FILESIZE_UNKNOWN) { 1005 CHECK(ZSTD_CCtx_setPledgedSrcSize(ress.cctx, fileSize)); 1006 } 1007 (void)srcFileName; 1008 1009 /* Main compression loop */ 1010 do { 1011 size_t stillToFlush; 1012 /* Fill input Buffer */ 1013 size_t const inSize = fread(ress.srcBuffer, (size_t)1, ress.srcBufferSize, srcFile); 1014 ZSTD_inBuffer inBuff = { ress.srcBuffer, inSize, 0 }; 1015 DISPLAYLEVEL(6, "fread %u bytes from source \n", (unsigned)inSize); 1016 *readsize += inSize; 1017 1018 if ((inSize == 0) || (*readsize == fileSize)) 1019 directive = ZSTD_e_end; 1020 1021 stillToFlush = 1; 1022 while ((inBuff.pos != inBuff.size) /* input buffer must be entirely ingested */ 1023 || (directive == ZSTD_e_end && stillToFlush != 0) ) { 1024 1025 size_t const oldIPos = inBuff.pos; 1026 ZSTD_outBuffer outBuff = { ress.dstBuffer, ress.dstBufferSize, 0 }; 1027 size_t const toFlushNow = ZSTD_toFlushNow(ress.cctx); 1028 CHECK_V(stillToFlush, ZSTD_compressStream2(ress.cctx, &outBuff, &inBuff, directive)); 1029 1030 /* count stats */ 1031 inputPresented++; 1032 if (oldIPos == inBuff.pos) inputBlocked++; /* input buffer is full and can't take any more : input speed is faster than consumption rate */ 1033 if (!toFlushNow) flushWaiting = 1; 1034 1035 /* Write compressed stream */ 1036 DISPLAYLEVEL(6, "ZSTD_compress_generic(end:%u) => input pos(%u)<=(%u)size ; output generated %u bytes \n", 1037 (unsigned)directive, (unsigned)inBuff.pos, (unsigned)inBuff.size, (unsigned)outBuff.pos); 1038 if (outBuff.pos) { 1039 size_t const sizeCheck = fwrite(ress.dstBuffer, 1, outBuff.pos, dstFile); 1040 if (sizeCheck != outBuff.pos) 1041 EXM_THROW(25, "Write error : %s (cannot write compressed block)", 1042 strerror(errno)); 1043 compressedfilesize += outBuff.pos; 1044 } 1045 1046 /* display notification; and adapt compression level */ 1047 if (READY_FOR_UPDATE()) { 1048 ZSTD_frameProgression const zfp = ZSTD_getFrameProgression(ress.cctx); 1049 double const cShare = (double)zfp.produced / (zfp.consumed + !zfp.consumed/*avoid div0*/) * 100; 1050 1051 /* display progress notifications */ 1052 if (g_display_prefs.displayLevel >= 3) { 1053 DISPLAYUPDATE(3, "\r(L%i) Buffered :%4u MB - Consumed :%4u MB - Compressed :%4u MB => %.2f%% ", 1054 compressionLevel, 1055 (unsigned)((zfp.ingested - zfp.consumed) >> 20), 1056 (unsigned)(zfp.consumed >> 20), 1057 (unsigned)(zfp.produced >> 20), 1058 cShare ); 1059 } else { /* summarized notifications if == 2; */ 1060 DISPLAYLEVEL(2, "\rRead : %u ", (unsigned)(zfp.consumed >> 20)); 1061 if (fileSize != UTIL_FILESIZE_UNKNOWN) 1062 DISPLAYLEVEL(2, "/ %u ", (unsigned)(fileSize >> 20)); 1063 DISPLAYLEVEL(2, "MB ==> %2.f%% ", cShare); 1064 DELAY_NEXT_UPDATE(); 1065 } 1066 1067 /* adaptive mode : statistics measurement and speed correction */ 1068 if (prefs->adaptiveMode) { 1069 1070 /* check output speed */ 1071 if (zfp.currentJobID > 1) { /* only possible if nbWorkers >= 1 */ 1072 1073 unsigned long long newlyProduced = zfp.produced - previous_zfp_update.produced; 1074 unsigned long long newlyFlushed = zfp.flushed - previous_zfp_update.flushed; 1075 assert(zfp.produced >= previous_zfp_update.produced); 1076 assert(prefs->nbWorkers >= 1); 1077 1078 /* test if compression is blocked 1079 * either because output is slow and all buffers are full 1080 * or because input is slow and no job can start while waiting for at least one buffer to be filled. 1081 * note : exclude starting part, since currentJobID > 1 */ 1082 if ( (zfp.consumed == previous_zfp_update.consumed) /* no data compressed : no data available, or no more buffer to compress to, OR compression is really slow (compression of a single block is slower than update rate)*/ 1083 && (zfp.nbActiveWorkers == 0) /* confirmed : no compression ongoing */ 1084 ) { 1085 DISPLAYLEVEL(6, "all buffers full : compression stopped => slow down \n") 1086 speedChange = slower; 1087 } 1088 1089 previous_zfp_update = zfp; 1090 1091 if ( (newlyProduced > (newlyFlushed * 9 / 8)) /* compression produces more data than output can flush (though production can be spiky, due to work unit : (N==4)*block sizes) */ 1092 && (flushWaiting == 0) /* flush speed was never slowed by lack of production, so it's operating at max capacity */ 1093 ) { 1094 DISPLAYLEVEL(6, "compression faster than flush (%llu > %llu), and flushed was never slowed down by lack of production => slow down \n", newlyProduced, newlyFlushed); 1095 speedChange = slower; 1096 } 1097 flushWaiting = 0; 1098 } 1099 1100 /* course correct only if there is at least one new job completed */ 1101 if (zfp.currentJobID > lastJobID) { 1102 DISPLAYLEVEL(6, "compression level adaptation check \n") 1103 1104 /* check input speed */ 1105 if (zfp.currentJobID > (unsigned)(prefs->nbWorkers+1)) { /* warm up period, to fill all workers */ 1106 if (inputBlocked <= 0) { 1107 DISPLAYLEVEL(6, "input is never blocked => input is slower than ingestion \n"); 1108 speedChange = slower; 1109 } else if (speedChange == noChange) { 1110 unsigned long long newlyIngested = zfp.ingested - previous_zfp_correction.ingested; 1111 unsigned long long newlyConsumed = zfp.consumed - previous_zfp_correction.consumed; 1112 unsigned long long newlyProduced = zfp.produced - previous_zfp_correction.produced; 1113 unsigned long long newlyFlushed = zfp.flushed - previous_zfp_correction.flushed; 1114 previous_zfp_correction = zfp; 1115 assert(inputPresented > 0); 1116 DISPLAYLEVEL(6, "input blocked %u/%u(%.2f) - ingested:%u vs %u:consumed - flushed:%u vs %u:produced \n", 1117 inputBlocked, inputPresented, (double)inputBlocked/inputPresented*100, 1118 (unsigned)newlyIngested, (unsigned)newlyConsumed, 1119 (unsigned)newlyFlushed, (unsigned)newlyProduced); 1120 if ( (inputBlocked > inputPresented / 8) /* input is waiting often, because input buffers is full : compression or output too slow */ 1121 && (newlyFlushed * 33 / 32 > newlyProduced) /* flush everything that is produced */ 1122 && (newlyIngested * 33 / 32 > newlyConsumed) /* input speed as fast or faster than compression speed */ 1123 ) { 1124 DISPLAYLEVEL(6, "recommend faster as in(%llu) >= (%llu)comp(%llu) <= out(%llu) \n", 1125 newlyIngested, newlyConsumed, newlyProduced, newlyFlushed); 1126 speedChange = faster; 1127 } 1128 } 1129 inputBlocked = 0; 1130 inputPresented = 0; 1131 } 1132 1133 if (speedChange == slower) { 1134 DISPLAYLEVEL(6, "slower speed , higher compression \n") 1135 compressionLevel ++; 1136 if (compressionLevel > ZSTD_maxCLevel()) compressionLevel = ZSTD_maxCLevel(); 1137 if (compressionLevel > prefs->maxAdaptLevel) compressionLevel = prefs->maxAdaptLevel; 1138 compressionLevel += (compressionLevel == 0); /* skip 0 */ 1139 ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel); 1140 } 1141 if (speedChange == faster) { 1142 DISPLAYLEVEL(6, "faster speed , lighter compression \n") 1143 compressionLevel --; 1144 if (compressionLevel < prefs->minAdaptLevel) compressionLevel = prefs->minAdaptLevel; 1145 compressionLevel -= (compressionLevel == 0); /* skip 0 */ 1146 ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel); 1147 } 1148 speedChange = noChange; 1149 1150 lastJobID = zfp.currentJobID; 1151 } /* if (zfp.currentJobID > lastJobID) */ 1152 } /* if (g_adaptiveMode) */ 1153 } /* if (READY_FOR_UPDATE()) */ 1154 } /* while ((inBuff.pos != inBuff.size) */ 1155 } while (directive != ZSTD_e_end); 1156 1157 if (ferror(srcFile)) { 1158 EXM_THROW(26, "Read error : I/O error"); 1159 } 1160 if (fileSize != UTIL_FILESIZE_UNKNOWN && *readsize != fileSize) { 1161 EXM_THROW(27, "Read error : Incomplete read : %llu / %llu B", 1162 (unsigned long long)*readsize, (unsigned long long)fileSize); 1163 } 1164 1165 return compressedfilesize; 1166 } 1167 1168 /*! FIO_compressFilename_internal() : 1169 * same as FIO_compressFilename_extRess(), with `ress.desFile` already opened. 1170 * @return : 0 : compression completed correctly, 1171 * 1 : missing or pb opening srcFileName 1172 */ 1173 static int 1174 FIO_compressFilename_internal(FIO_prefs_t* const prefs, 1175 cRess_t ress, 1176 const char* dstFileName, const char* srcFileName, 1177 int compressionLevel) 1178 { 1179 UTIL_time_t const timeStart = UTIL_getTime(); 1180 clock_t const cpuStart = clock(); 1181 U64 readsize = 0; 1182 U64 compressedfilesize = 0; 1183 U64 const fileSize = UTIL_getFileSize(srcFileName); 1184 DISPLAYLEVEL(5, "%s: %u bytes \n", srcFileName, (unsigned)fileSize); 1185 1186 /* compression format selection */ 1187 switch (prefs->compressionType) { 1188 default: 1189 case FIO_zstdCompression: 1190 compressedfilesize = FIO_compressZstdFrame(prefs, &ress, srcFileName, fileSize, compressionLevel, &readsize); 1191 break; 1192 1193 case FIO_gzipCompression: 1194 #ifdef ZSTD_GZCOMPRESS 1195 compressedfilesize = FIO_compressGzFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize); 1196 #else 1197 (void)compressionLevel; 1198 EXM_THROW(20, "zstd: %s: file cannot be compressed as gzip (zstd compiled without ZSTD_GZCOMPRESS) -- ignored \n", 1199 srcFileName); 1200 #endif 1201 break; 1202 1203 case FIO_xzCompression: 1204 case FIO_lzmaCompression: 1205 #ifdef ZSTD_LZMACOMPRESS 1206 compressedfilesize = FIO_compressLzmaFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize, prefs->compressionType==FIO_lzmaCompression); 1207 #else 1208 (void)compressionLevel; 1209 EXM_THROW(20, "zstd: %s: file cannot be compressed as xz/lzma (zstd compiled without ZSTD_LZMACOMPRESS) -- ignored \n", 1210 srcFileName); 1211 #endif 1212 break; 1213 1214 case FIO_lz4Compression: 1215 #ifdef ZSTD_LZ4COMPRESS 1216 compressedfilesize = FIO_compressLz4Frame(&ress, srcFileName, fileSize, compressionLevel, prefs->checksumFlag, &readsize); 1217 #else 1218 (void)compressionLevel; 1219 EXM_THROW(20, "zstd: %s: file cannot be compressed as lz4 (zstd compiled without ZSTD_LZ4COMPRESS) -- ignored \n", 1220 srcFileName); 1221 #endif 1222 break; 1223 } 1224 1225 /* Status */ 1226 DISPLAYLEVEL(2, "\r%79s\r", ""); 1227 DISPLAYLEVEL(2,"%-20s :%6.2f%% (%6llu => %6llu bytes, %s) \n", 1228 srcFileName, 1229 (double)compressedfilesize / (readsize+(!readsize)/*avoid div by zero*/) * 100, 1230 (unsigned long long)readsize, (unsigned long long) compressedfilesize, 1231 dstFileName); 1232 1233 /* Elapsed Time and CPU Load */ 1234 { clock_t const cpuEnd = clock(); 1235 double const cpuLoad_s = (double)(cpuEnd - cpuStart) / CLOCKS_PER_SEC; 1236 U64 const timeLength_ns = UTIL_clockSpanNano(timeStart); 1237 double const timeLength_s = (double)timeLength_ns / 1000000000; 1238 double const cpuLoad_pct = (cpuLoad_s / timeLength_s) * 100; 1239 DISPLAYLEVEL(4, "%-20s : Completed in %.2f sec (cpu load : %.0f%%)\n", 1240 srcFileName, timeLength_s, cpuLoad_pct); 1241 } 1242 return 0; 1243 } 1244 1245 1246 /*! FIO_compressFilename_dstFile() : 1247 * open dstFileName, or pass-through if ress.dstFile != NULL, 1248 * then start compression with FIO_compressFilename_internal(). 1249 * Manages source removal (--rm) and file permissions transfer. 1250 * note : ress.srcFile must be != NULL, 1251 * so reach this function through FIO_compressFilename_srcFile(). 1252 * @return : 0 : compression completed correctly, 1253 * 1 : pb 1254 */ 1255 static int FIO_compressFilename_dstFile(FIO_prefs_t* const prefs, 1256 cRess_t ress, 1257 const char* dstFileName, 1258 const char* srcFileName, 1259 int compressionLevel) 1260 { 1261 int closeDstFile = 0; 1262 int result; 1263 stat_t statbuf; 1264 int transfer_permissions = 0; 1265 1266 assert(ress.srcFile != NULL); 1267 1268 if (ress.dstFile == NULL) { 1269 closeDstFile = 1; 1270 DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: opening dst: %s", dstFileName); 1271 ress.dstFile = FIO_openDstFile(prefs, srcFileName, dstFileName); 1272 if (ress.dstFile==NULL) return 1; /* could not open dstFileName */ 1273 /* Must only be added after FIO_openDstFile() succeeds. 1274 * Otherwise we may delete the destination file if it already exists, 1275 * and the user presses Ctrl-C when asked if they wish to overwrite. 1276 */ 1277 addHandler(dstFileName); 1278 1279 if ( strcmp (srcFileName, stdinmark) 1280 && UTIL_getFileStat(srcFileName, &statbuf)) 1281 transfer_permissions = 1; 1282 } 1283 1284 result = FIO_compressFilename_internal(prefs, ress, dstFileName, srcFileName, compressionLevel); 1285 1286 if (closeDstFile) { 1287 FILE* const dstFile = ress.dstFile; 1288 ress.dstFile = NULL; 1289 1290 clearHandler(); 1291 1292 if (fclose(dstFile)) { /* error closing dstFile */ 1293 DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno)); 1294 result=1; 1295 } 1296 if ( (result != 0) /* operation failure */ 1297 && strcmp(dstFileName, nulmark) /* special case : don't remove() /dev/null */ 1298 && strcmp(dstFileName, stdoutmark) /* special case : don't remove() stdout */ 1299 ) { 1300 FIO_remove(dstFileName); /* remove compression artefact; note don't do anything special if remove() fails */ 1301 } else if ( strcmp(dstFileName, stdoutmark) 1302 && strcmp(dstFileName, nulmark) 1303 && transfer_permissions) { 1304 UTIL_setFileStat(dstFileName, &statbuf); 1305 } 1306 } 1307 1308 return result; 1309 } 1310 1311 1312 /*! FIO_compressFilename_srcFile() : 1313 * @return : 0 : compression completed correctly, 1314 * 1 : missing or pb opening srcFileName 1315 */ 1316 static int 1317 FIO_compressFilename_srcFile(FIO_prefs_t* const prefs, 1318 cRess_t ress, 1319 const char* dstFileName, 1320 const char* srcFileName, 1321 int compressionLevel) 1322 { 1323 int result; 1324 1325 /* ensure src is not a directory */ 1326 if (UTIL_isDirectory(srcFileName)) { 1327 DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName); 1328 return 1; 1329 } 1330 1331 /* ensure src is not the same as dict (if present) */ 1332 if (ress.dictFileName != NULL && UTIL_isSameFile(srcFileName, ress.dictFileName)) { 1333 DISPLAYLEVEL(1, "zstd: cannot use %s as an input file and dictionary \n", srcFileName); 1334 return 1; 1335 } 1336 1337 ress.srcFile = FIO_openSrcFile(srcFileName); 1338 if (ress.srcFile == NULL) return 1; /* srcFile could not be opened */ 1339 1340 result = FIO_compressFilename_dstFile(prefs, ress, dstFileName, srcFileName, compressionLevel); 1341 1342 fclose(ress.srcFile); 1343 ress.srcFile = NULL; 1344 if ( prefs->removeSrcFile /* --rm */ 1345 && result == 0 /* success */ 1346 && strcmp(srcFileName, stdinmark) /* exception : don't erase stdin */ 1347 ) { 1348 /* We must clear the handler, since after this point calling it would 1349 * delete both the source and destination files. 1350 */ 1351 clearHandler(); 1352 if (FIO_remove(srcFileName)) 1353 EXM_THROW(1, "zstd: %s: %s", srcFileName, strerror(errno)); 1354 } 1355 return result; 1356 } 1357 1358 1359 int FIO_compressFilename(FIO_prefs_t* const prefs, 1360 const char* dstFileName, const char* srcFileName, 1361 const char* dictFileName, int compressionLevel, 1362 ZSTD_compressionParameters comprParams) 1363 { 1364 U64 const fileSize = UTIL_getFileSize(srcFileName); 1365 U64 const srcSize = (fileSize == UTIL_FILESIZE_UNKNOWN) ? ZSTD_CONTENTSIZE_UNKNOWN : fileSize; 1366 1367 cRess_t const ress = FIO_createCResources(prefs, dictFileName, compressionLevel, srcSize, comprParams); 1368 int const result = FIO_compressFilename_srcFile(prefs, ress, dstFileName, srcFileName, compressionLevel); 1369 1370 1371 FIO_freeCResources(ress); 1372 return result; 1373 } 1374 1375 1376 /* FIO_determineCompressedName() : 1377 * create a destination filename for compressed srcFileName. 1378 * @return a pointer to it. 1379 * This function never returns an error (it may abort() in case of pb) 1380 */ 1381 static const char* 1382 FIO_determineCompressedName(const char* srcFileName, const char* suffix) 1383 { 1384 static size_t dfnbCapacity = 0; 1385 static char* dstFileNameBuffer = NULL; /* using static allocation : this function cannot be multi-threaded */ 1386 1387 size_t const sfnSize = strlen(srcFileName); 1388 size_t const suffixSize = strlen(suffix); 1389 1390 if (dfnbCapacity <= sfnSize+suffixSize+1) { 1391 /* resize buffer for dstName */ 1392 free(dstFileNameBuffer); 1393 dfnbCapacity = sfnSize + suffixSize + 30; 1394 dstFileNameBuffer = (char*)malloc(dfnbCapacity); 1395 if (!dstFileNameBuffer) { 1396 EXM_THROW(30, "zstd: %s", strerror(errno)); 1397 } } 1398 assert(dstFileNameBuffer != NULL); 1399 memcpy(dstFileNameBuffer, srcFileName, sfnSize); 1400 memcpy(dstFileNameBuffer+sfnSize, suffix, suffixSize+1 /* Include terminating null */); 1401 1402 return dstFileNameBuffer; 1403 } 1404 1405 1406 /* FIO_compressMultipleFilenames() : 1407 * compress nbFiles files 1408 * into one destination (outFileName) 1409 * or into one file each (outFileName == NULL, but suffix != NULL). 1410 */ 1411 int FIO_compressMultipleFilenames(FIO_prefs_t* const prefs, 1412 const char** inFileNamesTable, unsigned nbFiles, 1413 const char* outFileName, const char* suffix, 1414 const char* dictFileName, int compressionLevel, 1415 ZSTD_compressionParameters comprParams) 1416 { 1417 int error = 0; 1418 U64 const firstFileSize = UTIL_getFileSize(inFileNamesTable[0]); 1419 U64 const firstSrcSize = (firstFileSize == UTIL_FILESIZE_UNKNOWN) ? ZSTD_CONTENTSIZE_UNKNOWN : firstFileSize; 1420 U64 const srcSize = (nbFiles != 1) ? ZSTD_CONTENTSIZE_UNKNOWN : firstSrcSize ; 1421 cRess_t ress = FIO_createCResources(prefs, dictFileName, compressionLevel, srcSize, comprParams); 1422 1423 /* init */ 1424 assert(outFileName != NULL || suffix != NULL); 1425 1426 if (outFileName != NULL) { /* output into a single destination (stdout typically) */ 1427 ress.dstFile = FIO_openDstFile(prefs, NULL, outFileName); 1428 if (ress.dstFile == NULL) { /* could not open outFileName */ 1429 error = 1; 1430 } else { 1431 unsigned u; 1432 for (u=0; u<nbFiles; u++) 1433 error |= FIO_compressFilename_srcFile(prefs, ress, outFileName, inFileNamesTable[u], compressionLevel); 1434 if (fclose(ress.dstFile)) 1435 EXM_THROW(29, "Write error (%s) : cannot properly close %s", 1436 strerror(errno), outFileName); 1437 ress.dstFile = NULL; 1438 } 1439 } else { 1440 unsigned u; 1441 for (u=0; u<nbFiles; u++) { 1442 const char* const srcFileName = inFileNamesTable[u]; 1443 const char* const dstFileName = FIO_determineCompressedName(srcFileName, suffix); /* cannot fail */ 1444 error |= FIO_compressFilename_srcFile(prefs, ress, dstFileName, srcFileName, compressionLevel); 1445 } } 1446 1447 FIO_freeCResources(ress); 1448 return error; 1449 } 1450 1451 #endif /* #ifndef ZSTD_NOCOMPRESS */ 1452 1453 1454 1455 #ifndef ZSTD_NODECOMPRESS 1456 1457 /* ************************************************************************** 1458 * Decompression 1459 ***************************************************************************/ 1460 typedef struct { 1461 void* srcBuffer; 1462 size_t srcBufferSize; 1463 size_t srcBufferLoaded; 1464 void* dstBuffer; 1465 size_t dstBufferSize; 1466 ZSTD_DStream* dctx; 1467 FILE* dstFile; 1468 } dRess_t; 1469 1470 static dRess_t FIO_createDResources(FIO_prefs_t* const prefs, const char* dictFileName) 1471 { 1472 dRess_t ress; 1473 memset(&ress, 0, sizeof(ress)); 1474 1475 /* Allocation */ 1476 ress.dctx = ZSTD_createDStream(); 1477 if (ress.dctx==NULL) 1478 EXM_THROW(60, "Error: %s : can't create ZSTD_DStream", strerror(errno)); 1479 CHECK( ZSTD_DCtx_setMaxWindowSize(ress.dctx, prefs->memLimit) ); 1480 ress.srcBufferSize = ZSTD_DStreamInSize(); 1481 ress.srcBuffer = malloc(ress.srcBufferSize); 1482 ress.dstBufferSize = ZSTD_DStreamOutSize(); 1483 ress.dstBuffer = malloc(ress.dstBufferSize); 1484 if (!ress.srcBuffer || !ress.dstBuffer) 1485 EXM_THROW(61, "Allocation error : not enough memory"); 1486 1487 /* dictionary */ 1488 { void* dictBuffer; 1489 size_t const dictBufferSize = FIO_createDictBuffer(&dictBuffer, dictFileName); 1490 CHECK( ZSTD_initDStream_usingDict(ress.dctx, dictBuffer, dictBufferSize) ); 1491 free(dictBuffer); 1492 } 1493 1494 return ress; 1495 } 1496 1497 static void FIO_freeDResources(dRess_t ress) 1498 { 1499 CHECK( ZSTD_freeDStream(ress.dctx) ); 1500 free(ress.srcBuffer); 1501 free(ress.dstBuffer); 1502 } 1503 1504 1505 /** FIO_fwriteSparse() : 1506 * @return : storedSkips, to be provided to next call to FIO_fwriteSparse() of LZ4IO_fwriteSparseEnd() */ 1507 static unsigned FIO_fwriteSparse(FIO_prefs_t* const prefs, FILE* file, const void* buffer, size_t bufferSize, unsigned storedSkips) 1508 { 1509 const size_t* const bufferT = (const size_t*)buffer; /* Buffer is supposed malloc'ed, hence aligned on size_t */ 1510 size_t bufferSizeT = bufferSize / sizeof(size_t); 1511 const size_t* const bufferTEnd = bufferT + bufferSizeT; 1512 const size_t* ptrT = bufferT; 1513 static const size_t segmentSizeT = (32 KB) / sizeof(size_t); /* 0-test re-attempted every 32 KB */ 1514 1515 if (!prefs->sparseFileSupport) { /* normal write */ 1516 size_t const sizeCheck = fwrite(buffer, 1, bufferSize, file); 1517 if (sizeCheck != bufferSize) 1518 EXM_THROW(70, "Write error : %s (cannot write decoded block)", 1519 strerror(errno)); 1520 return 0; 1521 } 1522 1523 /* avoid int overflow */ 1524 if (storedSkips > 1 GB) { 1525 int const seekResult = LONG_SEEK(file, 1 GB, SEEK_CUR); 1526 if (seekResult != 0) 1527 EXM_THROW(71, "1 GB skip error (sparse file support)"); 1528 storedSkips -= 1 GB; 1529 } 1530 1531 while (ptrT < bufferTEnd) { 1532 size_t seg0SizeT = segmentSizeT; 1533 size_t nb0T; 1534 1535 /* count leading zeros */ 1536 if (seg0SizeT > bufferSizeT) seg0SizeT = bufferSizeT; 1537 bufferSizeT -= seg0SizeT; 1538 for (nb0T=0; (nb0T < seg0SizeT) && (ptrT[nb0T] == 0); nb0T++) ; 1539 storedSkips += (unsigned)(nb0T * sizeof(size_t)); 1540 1541 if (nb0T != seg0SizeT) { /* not all 0s */ 1542 int const seekResult = LONG_SEEK(file, storedSkips, SEEK_CUR); 1543 if (seekResult) EXM_THROW(72, "Sparse skip error ; try --no-sparse"); 1544 storedSkips = 0; 1545 seg0SizeT -= nb0T; 1546 ptrT += nb0T; 1547 { size_t const sizeCheck = fwrite(ptrT, sizeof(size_t), seg0SizeT, file); 1548 if (sizeCheck != seg0SizeT) 1549 EXM_THROW(73, "Write error : cannot write decoded block"); 1550 } } 1551 ptrT += seg0SizeT; 1552 } 1553 1554 { static size_t const maskT = sizeof(size_t)-1; 1555 if (bufferSize & maskT) { 1556 /* size not multiple of sizeof(size_t) : implies end of block */ 1557 const char* const restStart = (const char*)bufferTEnd; 1558 const char* restPtr = restStart; 1559 size_t restSize = bufferSize & maskT; 1560 const char* const restEnd = restStart + restSize; 1561 for ( ; (restPtr < restEnd) && (*restPtr == 0); restPtr++) ; 1562 storedSkips += (unsigned) (restPtr - restStart); 1563 if (restPtr != restEnd) { 1564 int seekResult = LONG_SEEK(file, storedSkips, SEEK_CUR); 1565 if (seekResult) 1566 EXM_THROW(74, "Sparse skip error ; try --no-sparse"); 1567 storedSkips = 0; 1568 { size_t const sizeCheck = fwrite(restPtr, 1, (size_t)(restEnd - restPtr), file); 1569 if (sizeCheck != (size_t)(restEnd - restPtr)) 1570 EXM_THROW(75, "Write error : cannot write decoded end of block"); 1571 } } } } 1572 1573 return storedSkips; 1574 } 1575 1576 static void 1577 FIO_fwriteSparseEnd(FIO_prefs_t* const prefs, FILE* file, unsigned storedSkips) 1578 { 1579 if (storedSkips>0) { 1580 assert(prefs->sparseFileSupport > 0); /* storedSkips>0 implies sparse support is enabled */ 1581 (void)prefs; /* assert can be disabled, in which case prefs becomes unused */ 1582 if (LONG_SEEK(file, storedSkips-1, SEEK_CUR) != 0) 1583 EXM_THROW(69, "Final skip error (sparse file support)"); 1584 /* last zero must be explicitly written, 1585 * so that skipped ones get implicitly translated as zero by FS */ 1586 { const char lastZeroByte[1] = { 0 }; 1587 if (fwrite(lastZeroByte, 1, 1, file) != 1) 1588 EXM_THROW(69, "Write error : cannot write last zero"); 1589 } } 1590 } 1591 1592 1593 /** FIO_passThrough() : just copy input into output, for compatibility with gzip -df mode 1594 @return : 0 (no error) */ 1595 static int FIO_passThrough(FIO_prefs_t* const prefs, 1596 FILE* foutput, FILE* finput, 1597 void* buffer, size_t bufferSize, 1598 size_t alreadyLoaded) 1599 { 1600 size_t const blockSize = MIN(64 KB, bufferSize); 1601 size_t readFromInput = 1; 1602 unsigned storedSkips = 0; 1603 1604 /* assumption : ress->srcBufferLoaded bytes already loaded and stored within buffer */ 1605 { size_t const sizeCheck = fwrite(buffer, 1, alreadyLoaded, foutput); 1606 if (sizeCheck != alreadyLoaded) { 1607 DISPLAYLEVEL(1, "Pass-through write error \n"); 1608 return 1; 1609 } } 1610 1611 while (readFromInput) { 1612 readFromInput = fread(buffer, 1, blockSize, finput); 1613 storedSkips = FIO_fwriteSparse(prefs, foutput, buffer, readFromInput, storedSkips); 1614 } 1615 1616 FIO_fwriteSparseEnd(prefs, foutput, storedSkips); 1617 return 0; 1618 } 1619 1620 /* FIO_highbit64() : 1621 * gives position of highest bit. 1622 * note : only works for v > 0 ! 1623 */ 1624 static unsigned FIO_highbit64(unsigned long long v) 1625 { 1626 unsigned count = 0; 1627 assert(v != 0); 1628 v >>= 1; 1629 while (v) { v >>= 1; count++; } 1630 return count; 1631 } 1632 1633 /* FIO_zstdErrorHelp() : 1634 * detailed error message when requested window size is too large */ 1635 static void FIO_zstdErrorHelp(FIO_prefs_t* const prefs, dRess_t* ress, size_t err, char const* srcFileName) 1636 { 1637 ZSTD_frameHeader header; 1638 1639 /* Help message only for one specific error */ 1640 if (ZSTD_getErrorCode(err) != ZSTD_error_frameParameter_windowTooLarge) 1641 return; 1642 1643 /* Try to decode the frame header */ 1644 err = ZSTD_getFrameHeader(&header, ress->srcBuffer, ress->srcBufferLoaded); 1645 if (err == 0) { 1646 unsigned long long const windowSize = header.windowSize; 1647 unsigned const windowLog = FIO_highbit64(windowSize) + ((windowSize & (windowSize - 1)) != 0); 1648 assert(prefs->memLimit > 0); 1649 DISPLAYLEVEL(1, "%s : Window size larger than maximum : %llu > %u\n", 1650 srcFileName, windowSize, prefs->memLimit); 1651 if (windowLog <= ZSTD_WINDOWLOG_MAX) { 1652 unsigned const windowMB = (unsigned)((windowSize >> 20) + ((windowSize & ((1 MB) - 1)) != 0)); 1653 assert(windowSize < (U64)(1ULL << 52)); /* ensure now overflow for windowMB */ 1654 DISPLAYLEVEL(1, "%s : Use --long=%u or --memory=%uMB\n", 1655 srcFileName, windowLog, windowMB); 1656 return; 1657 } 1658 } 1659 DISPLAYLEVEL(1, "%s : Window log larger than ZSTD_WINDOWLOG_MAX=%u; not supported\n", 1660 srcFileName, ZSTD_WINDOWLOG_MAX); 1661 } 1662 1663 /** FIO_decompressFrame() : 1664 * @return : size of decoded zstd frame, or an error code 1665 */ 1666 #define FIO_ERROR_FRAME_DECODING ((unsigned long long)(-2)) 1667 static unsigned long long FIO_decompressZstdFrame( 1668 FIO_prefs_t* const prefs, 1669 dRess_t* ress, 1670 FILE* finput, 1671 const char* srcFileName, 1672 U64 alreadyDecoded) 1673 { 1674 U64 frameSize = 0; 1675 U32 storedSkips = 0; 1676 1677 size_t const srcFileLength = strlen(srcFileName); 1678 if (srcFileLength>20) srcFileName += srcFileLength-20; /* display last 20 characters only */ 1679 1680 ZSTD_resetDStream(ress->dctx); 1681 1682 /* Header loading : ensures ZSTD_getFrameHeader() will succeed */ 1683 { size_t const toDecode = ZSTD_FRAMEHEADERSIZE_MAX; 1684 if (ress->srcBufferLoaded < toDecode) { 1685 size_t const toRead = toDecode - ress->srcBufferLoaded; 1686 void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded; 1687 ress->srcBufferLoaded += fread(startPosition, 1, toRead, finput); 1688 } } 1689 1690 /* Main decompression Loop */ 1691 while (1) { 1692 ZSTD_inBuffer inBuff = { ress->srcBuffer, ress->srcBufferLoaded, 0 }; 1693 ZSTD_outBuffer outBuff= { ress->dstBuffer, ress->dstBufferSize, 0 }; 1694 size_t const readSizeHint = ZSTD_decompressStream(ress->dctx, &outBuff, &inBuff); 1695 if (ZSTD_isError(readSizeHint)) { 1696 DISPLAYLEVEL(1, "%s : Decoding error (36) : %s \n", 1697 srcFileName, ZSTD_getErrorName(readSizeHint)); 1698 FIO_zstdErrorHelp(prefs, ress, readSizeHint, srcFileName); 1699 return FIO_ERROR_FRAME_DECODING; 1700 } 1701 1702 /* Write block */ 1703 storedSkips = FIO_fwriteSparse(prefs, ress->dstFile, ress->dstBuffer, outBuff.pos, storedSkips); 1704 frameSize += outBuff.pos; 1705 DISPLAYUPDATE(2, "\r%-20.20s : %u MB... ", 1706 srcFileName, (unsigned)((alreadyDecoded+frameSize)>>20) ); 1707 1708 if (inBuff.pos > 0) { 1709 memmove(ress->srcBuffer, (char*)ress->srcBuffer + inBuff.pos, inBuff.size - inBuff.pos); 1710 ress->srcBufferLoaded -= inBuff.pos; 1711 } 1712 1713 if (readSizeHint == 0) break; /* end of frame */ 1714 if (inBuff.size != inBuff.pos) { 1715 DISPLAYLEVEL(1, "%s : Decoding error (37) : should consume entire input \n", 1716 srcFileName); 1717 return FIO_ERROR_FRAME_DECODING; 1718 } 1719 1720 /* Fill input buffer */ 1721 { size_t const toDecode = MIN(readSizeHint, ress->srcBufferSize); /* support large skippable frames */ 1722 if (ress->srcBufferLoaded < toDecode) { 1723 size_t const toRead = toDecode - ress->srcBufferLoaded; /* > 0 */ 1724 void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded; 1725 size_t const readSize = fread(startPosition, 1, toRead, finput); 1726 if (readSize==0) { 1727 DISPLAYLEVEL(1, "%s : Read error (39) : premature end \n", 1728 srcFileName); 1729 return FIO_ERROR_FRAME_DECODING; 1730 } 1731 ress->srcBufferLoaded += readSize; 1732 } } } 1733 1734 FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips); 1735 1736 return frameSize; 1737 } 1738 1739 1740 #ifdef ZSTD_GZDECOMPRESS 1741 static unsigned long long FIO_decompressGzFrame(dRess_t* ress, 1742 FILE* srcFile, const char* srcFileName) 1743 { 1744 unsigned long long outFileSize = 0; 1745 z_stream strm; 1746 int flush = Z_NO_FLUSH; 1747 int decodingError = 0; 1748 1749 strm.zalloc = Z_NULL; 1750 strm.zfree = Z_NULL; 1751 strm.opaque = Z_NULL; 1752 strm.next_in = 0; 1753 strm.avail_in = 0; 1754 /* see http://www.zlib.net/manual.html */ 1755 if (inflateInit2(&strm, 15 /* maxWindowLogSize */ + 16 /* gzip only */) != Z_OK) 1756 return FIO_ERROR_FRAME_DECODING; 1757 1758 strm.next_out = (Bytef*)ress->dstBuffer; 1759 strm.avail_out = (uInt)ress->dstBufferSize; 1760 strm.avail_in = (uInt)ress->srcBufferLoaded; 1761 strm.next_in = (z_const unsigned char*)ress->srcBuffer; 1762 1763 for ( ; ; ) { 1764 int ret; 1765 if (strm.avail_in == 0) { 1766 ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile); 1767 if (ress->srcBufferLoaded == 0) flush = Z_FINISH; 1768 strm.next_in = (z_const unsigned char*)ress->srcBuffer; 1769 strm.avail_in = (uInt)ress->srcBufferLoaded; 1770 } 1771 ret = inflate(&strm, flush); 1772 if (ret == Z_BUF_ERROR) { 1773 DISPLAYLEVEL(1, "zstd: %s: premature gz end \n", srcFileName); 1774 decodingError = 1; break; 1775 } 1776 if (ret != Z_OK && ret != Z_STREAM_END) { 1777 DISPLAYLEVEL(1, "zstd: %s: inflate error %d \n", srcFileName, ret); 1778 decodingError = 1; break; 1779 } 1780 { size_t const decompBytes = ress->dstBufferSize - strm.avail_out; 1781 if (decompBytes) { 1782 if (fwrite(ress->dstBuffer, 1, decompBytes, ress->dstFile) != decompBytes) { 1783 DISPLAYLEVEL(1, "zstd: %s \n", strerror(errno)); 1784 decodingError = 1; break; 1785 } 1786 outFileSize += decompBytes; 1787 strm.next_out = (Bytef*)ress->dstBuffer; 1788 strm.avail_out = (uInt)ress->dstBufferSize; 1789 } 1790 } 1791 if (ret == Z_STREAM_END) break; 1792 } 1793 1794 if (strm.avail_in > 0) 1795 memmove(ress->srcBuffer, strm.next_in, strm.avail_in); 1796 ress->srcBufferLoaded = strm.avail_in; 1797 if ( (inflateEnd(&strm) != Z_OK) /* release resources ; error detected */ 1798 && (decodingError==0) ) { 1799 DISPLAYLEVEL(1, "zstd: %s: inflateEnd error \n", srcFileName); 1800 decodingError = 1; 1801 } 1802 return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize; 1803 } 1804 #endif 1805 1806 1807 #ifdef ZSTD_LZMADECOMPRESS 1808 static unsigned long long FIO_decompressLzmaFrame(dRess_t* ress, FILE* srcFile, const char* srcFileName, int plain_lzma) 1809 { 1810 unsigned long long outFileSize = 0; 1811 lzma_stream strm = LZMA_STREAM_INIT; 1812 lzma_action action = LZMA_RUN; 1813 lzma_ret initRet; 1814 int decodingError = 0; 1815 1816 strm.next_in = 0; 1817 strm.avail_in = 0; 1818 if (plain_lzma) { 1819 initRet = lzma_alone_decoder(&strm, UINT64_MAX); /* LZMA */ 1820 } else { 1821 initRet = lzma_stream_decoder(&strm, UINT64_MAX, 0); /* XZ */ 1822 } 1823 1824 if (initRet != LZMA_OK) { 1825 DISPLAYLEVEL(1, "zstd: %s: %s error %d \n", 1826 plain_lzma ? "lzma_alone_decoder" : "lzma_stream_decoder", 1827 srcFileName, initRet); 1828 return FIO_ERROR_FRAME_DECODING; 1829 } 1830 1831 strm.next_out = (BYTE*)ress->dstBuffer; 1832 strm.avail_out = ress->dstBufferSize; 1833 strm.next_in = (BYTE const*)ress->srcBuffer; 1834 strm.avail_in = ress->srcBufferLoaded; 1835 1836 for ( ; ; ) { 1837 lzma_ret ret; 1838 if (strm.avail_in == 0) { 1839 ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile); 1840 if (ress->srcBufferLoaded == 0) action = LZMA_FINISH; 1841 strm.next_in = (BYTE const*)ress->srcBuffer; 1842 strm.avail_in = ress->srcBufferLoaded; 1843 } 1844 ret = lzma_code(&strm, action); 1845 1846 if (ret == LZMA_BUF_ERROR) { 1847 DISPLAYLEVEL(1, "zstd: %s: premature lzma end \n", srcFileName); 1848 decodingError = 1; break; 1849 } 1850 if (ret != LZMA_OK && ret != LZMA_STREAM_END) { 1851 DISPLAYLEVEL(1, "zstd: %s: lzma_code decoding error %d \n", 1852 srcFileName, ret); 1853 decodingError = 1; break; 1854 } 1855 { size_t const decompBytes = ress->dstBufferSize - strm.avail_out; 1856 if (decompBytes) { 1857 if (fwrite(ress->dstBuffer, 1, decompBytes, ress->dstFile) != decompBytes) { 1858 DISPLAYLEVEL(1, "zstd: %s \n", strerror(errno)); 1859 decodingError = 1; break; 1860 } 1861 outFileSize += decompBytes; 1862 strm.next_out = (BYTE*)ress->dstBuffer; 1863 strm.avail_out = ress->dstBufferSize; 1864 } } 1865 if (ret == LZMA_STREAM_END) break; 1866 } 1867 1868 if (strm.avail_in > 0) 1869 memmove(ress->srcBuffer, strm.next_in, strm.avail_in); 1870 ress->srcBufferLoaded = strm.avail_in; 1871 lzma_end(&strm); 1872 return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize; 1873 } 1874 #endif 1875 1876 #ifdef ZSTD_LZ4DECOMPRESS 1877 static unsigned long long FIO_decompressLz4Frame(dRess_t* ress, 1878 FILE* srcFile, const char* srcFileName) 1879 { 1880 unsigned long long filesize = 0; 1881 LZ4F_errorCode_t nextToLoad; 1882 LZ4F_decompressionContext_t dCtx; 1883 LZ4F_errorCode_t const errorCode = LZ4F_createDecompressionContext(&dCtx, LZ4F_VERSION); 1884 int decodingError = 0; 1885 1886 if (LZ4F_isError(errorCode)) { 1887 DISPLAYLEVEL(1, "zstd: failed to create lz4 decompression context \n"); 1888 return FIO_ERROR_FRAME_DECODING; 1889 } 1890 1891 /* Init feed with magic number (already consumed from FILE* sFile) */ 1892 { size_t inSize = 4; 1893 size_t outSize= 0; 1894 MEM_writeLE32(ress->srcBuffer, LZ4_MAGICNUMBER); 1895 nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &outSize, ress->srcBuffer, &inSize, NULL); 1896 if (LZ4F_isError(nextToLoad)) { 1897 DISPLAYLEVEL(1, "zstd: %s: lz4 header error : %s \n", 1898 srcFileName, LZ4F_getErrorName(nextToLoad)); 1899 LZ4F_freeDecompressionContext(dCtx); 1900 return FIO_ERROR_FRAME_DECODING; 1901 } } 1902 1903 /* Main Loop */ 1904 for (;nextToLoad;) { 1905 size_t readSize; 1906 size_t pos = 0; 1907 size_t decodedBytes = ress->dstBufferSize; 1908 1909 /* Read input */ 1910 if (nextToLoad > ress->srcBufferSize) nextToLoad = ress->srcBufferSize; 1911 readSize = fread(ress->srcBuffer, 1, nextToLoad, srcFile); 1912 if (!readSize) break; /* reached end of file or stream */ 1913 1914 while ((pos < readSize) || (decodedBytes == ress->dstBufferSize)) { /* still to read, or still to flush */ 1915 /* Decode Input (at least partially) */ 1916 size_t remaining = readSize - pos; 1917 decodedBytes = ress->dstBufferSize; 1918 nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &decodedBytes, (char*)(ress->srcBuffer)+pos, &remaining, NULL); 1919 if (LZ4F_isError(nextToLoad)) { 1920 DISPLAYLEVEL(1, "zstd: %s: lz4 decompression error : %s \n", 1921 srcFileName, LZ4F_getErrorName(nextToLoad)); 1922 decodingError = 1; nextToLoad = 0; break; 1923 } 1924 pos += remaining; 1925 1926 /* Write Block */ 1927 if (decodedBytes) { 1928 if (fwrite(ress->dstBuffer, 1, decodedBytes, ress->dstFile) != decodedBytes) { 1929 DISPLAYLEVEL(1, "zstd: %s \n", strerror(errno)); 1930 decodingError = 1; nextToLoad = 0; break; 1931 } 1932 filesize += decodedBytes; 1933 DISPLAYUPDATE(2, "\rDecompressed : %u MB ", (unsigned)(filesize>>20)); 1934 } 1935 1936 if (!nextToLoad) break; 1937 } 1938 } 1939 /* can be out because readSize == 0, which could be an fread() error */ 1940 if (ferror(srcFile)) { 1941 DISPLAYLEVEL(1, "zstd: %s: read error \n", srcFileName); 1942 decodingError=1; 1943 } 1944 1945 if (nextToLoad!=0) { 1946 DISPLAYLEVEL(1, "zstd: %s: unfinished lz4 stream \n", srcFileName); 1947 decodingError=1; 1948 } 1949 1950 LZ4F_freeDecompressionContext(dCtx); 1951 ress->srcBufferLoaded = 0; /* LZ4F will reach exact frame boundary */ 1952 1953 return decodingError ? FIO_ERROR_FRAME_DECODING : filesize; 1954 } 1955 #endif 1956 1957 1958 1959 /** FIO_decompressFrames() : 1960 * Find and decode frames inside srcFile 1961 * srcFile presumed opened and valid 1962 * @return : 0 : OK 1963 * 1 : error 1964 */ 1965 static int FIO_decompressFrames(FIO_prefs_t* const prefs, dRess_t ress, FILE* srcFile, 1966 const char* dstFileName, const char* srcFileName) 1967 { 1968 unsigned readSomething = 0; 1969 unsigned long long filesize = 0; 1970 assert(srcFile != NULL); 1971 1972 /* for each frame */ 1973 for ( ; ; ) { 1974 /* check magic number -> version */ 1975 size_t const toRead = 4; 1976 const BYTE* const buf = (const BYTE*)ress.srcBuffer; 1977 if (ress.srcBufferLoaded < toRead) /* load up to 4 bytes for header */ 1978 ress.srcBufferLoaded += fread((char*)ress.srcBuffer + ress.srcBufferLoaded, 1979 (size_t)1, toRead - ress.srcBufferLoaded, srcFile); 1980 if (ress.srcBufferLoaded==0) { 1981 if (readSomething==0) { /* srcFile is empty (which is invalid) */ 1982 DISPLAYLEVEL(1, "zstd: %s: unexpected end of file \n", srcFileName); 1983 return 1; 1984 } /* else, just reached frame boundary */ 1985 break; /* no more input */ 1986 } 1987 readSomething = 1; /* there is at least 1 byte in srcFile */ 1988 if (ress.srcBufferLoaded < toRead) { 1989 DISPLAYLEVEL(1, "zstd: %s: unknown header \n", srcFileName); 1990 return 1; 1991 } 1992 if (ZSTD_isFrame(buf, ress.srcBufferLoaded)) { 1993 unsigned long long const frameSize = FIO_decompressZstdFrame(prefs, &ress, srcFile, srcFileName, filesize); 1994 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; 1995 filesize += frameSize; 1996 } else if (buf[0] == 31 && buf[1] == 139) { /* gz magic number */ 1997 #ifdef ZSTD_GZDECOMPRESS 1998 unsigned long long const frameSize = FIO_decompressGzFrame(&ress, srcFile, srcFileName); 1999 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; 2000 filesize += frameSize; 2001 #else 2002 DISPLAYLEVEL(1, "zstd: %s: gzip file cannot be uncompressed (zstd compiled without HAVE_ZLIB) -- ignored \n", srcFileName); 2003 return 1; 2004 #endif 2005 } else if ((buf[0] == 0xFD && buf[1] == 0x37) /* xz magic number */ 2006 || (buf[0] == 0x5D && buf[1] == 0x00)) { /* lzma header (no magic number) */ 2007 #ifdef ZSTD_LZMADECOMPRESS 2008 unsigned long long const frameSize = FIO_decompressLzmaFrame(&ress, srcFile, srcFileName, buf[0] != 0xFD); 2009 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; 2010 filesize += frameSize; 2011 #else 2012 DISPLAYLEVEL(1, "zstd: %s: xz/lzma file cannot be uncompressed (zstd compiled without HAVE_LZMA) -- ignored \n", srcFileName); 2013 return 1; 2014 #endif 2015 } else if (MEM_readLE32(buf) == LZ4_MAGICNUMBER) { 2016 #ifdef ZSTD_LZ4DECOMPRESS 2017 unsigned long long const frameSize = FIO_decompressLz4Frame(&ress, srcFile, srcFileName); 2018 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; 2019 filesize += frameSize; 2020 #else 2021 DISPLAYLEVEL(1, "zstd: %s: lz4 file cannot be uncompressed (zstd compiled without HAVE_LZ4) -- ignored \n", srcFileName); 2022 return 1; 2023 #endif 2024 } else if ((prefs->overwrite) && !strcmp (dstFileName, stdoutmark)) { /* pass-through mode */ 2025 return FIO_passThrough(prefs, 2026 ress.dstFile, srcFile, 2027 ress.srcBuffer, ress.srcBufferSize, 2028 ress.srcBufferLoaded); 2029 } else { 2030 DISPLAYLEVEL(1, "zstd: %s: unsupported format \n", srcFileName); 2031 return 1; 2032 } } /* for each frame */ 2033 2034 /* Final Status */ 2035 DISPLAYLEVEL(2, "\r%79s\r", ""); 2036 DISPLAYLEVEL(2, "%-20s: %llu bytes \n", srcFileName, filesize); 2037 2038 return 0; 2039 } 2040 2041 /** FIO_decompressDstFile() : 2042 open `dstFileName`, 2043 or path-through if ress.dstFile is already != 0, 2044 then start decompression process (FIO_decompressFrames()). 2045 @return : 0 : OK 2046 1 : operation aborted 2047 */ 2048 static int FIO_decompressDstFile(FIO_prefs_t* const prefs, 2049 dRess_t ress, FILE* srcFile, 2050 const char* dstFileName, const char* srcFileName) 2051 { 2052 int result; 2053 stat_t statbuf; 2054 int transfer_permissions = 0; 2055 int releaseDstFile = 0; 2056 2057 if (ress.dstFile == NULL) { 2058 releaseDstFile = 1; 2059 2060 ress.dstFile = FIO_openDstFile(prefs, srcFileName, dstFileName); 2061 if (ress.dstFile==0) return 1; 2062 2063 /* Must only be added after FIO_openDstFile() succeeds. 2064 * Otherwise we may delete the destination file if it already exists, 2065 * and the user presses Ctrl-C when asked if they wish to overwrite. 2066 */ 2067 addHandler(dstFileName); 2068 2069 if ( strcmp(srcFileName, stdinmark) /* special case : don't transfer permissions from stdin */ 2070 && UTIL_getFileStat(srcFileName, &statbuf) ) 2071 transfer_permissions = 1; 2072 } 2073 2074 2075 result = FIO_decompressFrames(prefs, ress, srcFile, dstFileName, srcFileName); 2076 2077 if (releaseDstFile) { 2078 FILE* const dstFile = ress.dstFile; 2079 clearHandler(); 2080 ress.dstFile = NULL; 2081 if (fclose(dstFile)) { 2082 DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno)); 2083 result = 1; 2084 } 2085 2086 if ( (result != 0) /* operation failure */ 2087 && strcmp(dstFileName, nulmark) /* special case : don't remove() /dev/null (#316) */ 2088 && strcmp(dstFileName, stdoutmark) /* special case : don't remove() stdout */ 2089 ) { 2090 FIO_remove(dstFileName); /* remove decompression artefact; note: don't do anything special if remove() fails */ 2091 } else { /* operation success */ 2092 if ( strcmp(dstFileName, stdoutmark) /* special case : don't chmod stdout */ 2093 && strcmp(dstFileName, nulmark) /* special case : don't chmod /dev/null */ 2094 && transfer_permissions ) /* file permissions correctly extracted from src */ 2095 UTIL_setFileStat(dstFileName, &statbuf); /* transfer file permissions from src into dst */ 2096 } 2097 } 2098 2099 return result; 2100 } 2101 2102 2103 /** FIO_decompressSrcFile() : 2104 Open `srcFileName`, transfer control to decompressDstFile() 2105 @return : 0 : OK 2106 1 : error 2107 */ 2108 static int FIO_decompressSrcFile(FIO_prefs_t* const prefs, dRess_t ress, const char* dstFileName, const char* srcFileName) 2109 { 2110 FILE* srcFile; 2111 int result; 2112 2113 if (UTIL_isDirectory(srcFileName)) { 2114 DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName); 2115 return 1; 2116 } 2117 2118 srcFile = FIO_openSrcFile(srcFileName); 2119 if (srcFile==NULL) return 1; 2120 ress.srcBufferLoaded = 0; 2121 2122 result = FIO_decompressDstFile(prefs, ress, srcFile, dstFileName, srcFileName); 2123 2124 /* Close file */ 2125 if (fclose(srcFile)) { 2126 DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno)); /* error should not happen */ 2127 return 1; 2128 } 2129 if ( prefs->removeSrcFile /* --rm */ 2130 && (result==0) /* decompression successful */ 2131 && strcmp(srcFileName, stdinmark) ) /* not stdin */ { 2132 /* We must clear the handler, since after this point calling it would 2133 * delete both the source and destination files. 2134 */ 2135 clearHandler(); 2136 if (FIO_remove(srcFileName)) { 2137 /* failed to remove src file */ 2138 DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno)); 2139 return 1; 2140 } } 2141 return result; 2142 } 2143 2144 2145 2146 int FIO_decompressFilename(FIO_prefs_t* const prefs, 2147 const char* dstFileName, const char* srcFileName, 2148 const char* dictFileName) 2149 { 2150 dRess_t const ress = FIO_createDResources(prefs, dictFileName); 2151 2152 int const decodingError = FIO_decompressSrcFile(prefs, ress, dstFileName, srcFileName); 2153 2154 FIO_freeDResources(ress); 2155 return decodingError; 2156 } 2157 2158 2159 /* FIO_determineDstName() : 2160 * create a destination filename from a srcFileName. 2161 * @return a pointer to it. 2162 * @return == NULL if there is an error */ 2163 static const char* 2164 FIO_determineDstName(const char* srcFileName) 2165 { 2166 static size_t dfnbCapacity = 0; 2167 static char* dstFileNameBuffer = NULL; /* using static allocation : this function cannot be multi-threaded */ 2168 2169 size_t const sfnSize = strlen(srcFileName); 2170 size_t suffixSize; 2171 const char* const suffixPtr = strrchr(srcFileName, '.'); 2172 if (suffixPtr == NULL) { 2173 DISPLAYLEVEL(1, "zstd: %s: unknown suffix -- ignored \n", 2174 srcFileName); 2175 return NULL; 2176 } 2177 suffixSize = strlen(suffixPtr); 2178 2179 /* check suffix is authorized */ 2180 if (sfnSize <= suffixSize 2181 || ( strcmp(suffixPtr, ZSTD_EXTENSION) 2182 #ifdef ZSTD_GZDECOMPRESS 2183 && strcmp(suffixPtr, GZ_EXTENSION) 2184 #endif 2185 #ifdef ZSTD_LZMADECOMPRESS 2186 && strcmp(suffixPtr, XZ_EXTENSION) 2187 && strcmp(suffixPtr, LZMA_EXTENSION) 2188 #endif 2189 #ifdef ZSTD_LZ4DECOMPRESS 2190 && strcmp(suffixPtr, LZ4_EXTENSION) 2191 #endif 2192 ) ) { 2193 const char* suffixlist = ZSTD_EXTENSION 2194 #ifdef ZSTD_GZDECOMPRESS 2195 "/" GZ_EXTENSION 2196 #endif 2197 #ifdef ZSTD_LZMADECOMPRESS 2198 "/" XZ_EXTENSION "/" LZMA_EXTENSION 2199 #endif 2200 #ifdef ZSTD_LZ4DECOMPRESS 2201 "/" LZ4_EXTENSION 2202 #endif 2203 ; 2204 DISPLAYLEVEL(1, "zstd: %s: unknown suffix (%s expected) -- ignored \n", 2205 srcFileName, suffixlist); 2206 return NULL; 2207 } 2208 2209 /* allocate enough space to write dstFilename into it */ 2210 if (dfnbCapacity+suffixSize <= sfnSize+1) { 2211 free(dstFileNameBuffer); 2212 dfnbCapacity = sfnSize + 20; 2213 dstFileNameBuffer = (char*)malloc(dfnbCapacity); 2214 if (dstFileNameBuffer==NULL) 2215 EXM_THROW(74, "%s : not enough memory for dstFileName", strerror(errno)); 2216 } 2217 2218 /* return dst name == src name truncated from suffix */ 2219 assert(dstFileNameBuffer != NULL); 2220 memcpy(dstFileNameBuffer, srcFileName, sfnSize - suffixSize); 2221 dstFileNameBuffer[sfnSize-suffixSize] = '\0'; 2222 return dstFileNameBuffer; 2223 2224 /* note : dstFileNameBuffer memory is not going to be free */ 2225 } 2226 2227 2228 int 2229 FIO_decompressMultipleFilenames(FIO_prefs_t* const prefs, 2230 const char* srcNamesTable[], unsigned nbFiles, 2231 const char* outFileName, 2232 const char* dictFileName) 2233 { 2234 int error = 0; 2235 dRess_t ress = FIO_createDResources(prefs, dictFileName); 2236 2237 if (outFileName) { 2238 unsigned u; 2239 ress.dstFile = FIO_openDstFile(prefs, NULL, outFileName); 2240 if (ress.dstFile == 0) EXM_THROW(71, "cannot open %s", outFileName); 2241 for (u=0; u<nbFiles; u++) 2242 error |= FIO_decompressSrcFile(prefs, ress, outFileName, srcNamesTable[u]); 2243 if (fclose(ress.dstFile)) 2244 EXM_THROW(72, "Write error : %s : cannot properly close output file", 2245 strerror(errno)); 2246 } else { 2247 unsigned u; 2248 for (u=0; u<nbFiles; u++) { /* create dstFileName */ 2249 const char* const srcFileName = srcNamesTable[u]; 2250 const char* const dstFileName = FIO_determineDstName(srcFileName); 2251 if (dstFileName == NULL) { error=1; continue; } 2252 2253 error |= FIO_decompressSrcFile(prefs, ress, dstFileName, srcFileName); 2254 } 2255 } 2256 2257 FIO_freeDResources(ress); 2258 return error; 2259 } 2260 2261 2262 2263 /* ************************************************************************** 2264 * .zst file info (--list command) 2265 ***************************************************************************/ 2266 2267 typedef struct { 2268 U64 decompressedSize; 2269 U64 compressedSize; 2270 U64 windowSize; 2271 int numActualFrames; 2272 int numSkippableFrames; 2273 int decompUnavailable; 2274 int usesCheck; 2275 U32 nbFiles; 2276 } fileInfo_t; 2277 2278 typedef enum { 2279 info_success=0, 2280 info_frame_error=1, 2281 info_not_zstd=2, 2282 info_file_error=3, 2283 info_truncated_input=4, 2284 } InfoError; 2285 2286 #define ERROR_IF(c,n,...) { \ 2287 if (c) { \ 2288 DISPLAYLEVEL(1, __VA_ARGS__); \ 2289 DISPLAYLEVEL(1, " \n"); \ 2290 return n; \ 2291 } \ 2292 } 2293 2294 static InfoError 2295 FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile) 2296 { 2297 /* begin analyzing frame */ 2298 for ( ; ; ) { 2299 BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX]; 2300 size_t const numBytesRead = fread(headerBuffer, 1, sizeof(headerBuffer), srcFile); 2301 if (numBytesRead < ZSTD_FRAMEHEADERSIZE_MIN) { 2302 if ( feof(srcFile) 2303 && (numBytesRead == 0) 2304 && (info->compressedSize > 0) 2305 && (info->compressedSize != UTIL_FILESIZE_UNKNOWN) ) { 2306 unsigned long long file_position = (unsigned long long) LONG_TELL(srcFile); 2307 unsigned long long file_size = (unsigned long long) info->compressedSize; 2308 ERROR_IF(file_position != file_size, info_truncated_input, 2309 "Error: seeked to position %llu, which is beyond file size of %llu\n", 2310 file_position, 2311 file_size); 2312 break; /* correct end of file => success */ 2313 } 2314 ERROR_IF(feof(srcFile), info_not_zstd, "Error: reached end of file with incomplete frame"); 2315 ERROR_IF(1, info_frame_error, "Error: did not reach end of file but ran out of frames"); 2316 } 2317 { U32 const magicNumber = MEM_readLE32(headerBuffer); 2318 /* Zstandard frame */ 2319 if (magicNumber == ZSTD_MAGICNUMBER) { 2320 ZSTD_frameHeader header; 2321 U64 const frameContentSize = ZSTD_getFrameContentSize(headerBuffer, numBytesRead); 2322 if ( frameContentSize == ZSTD_CONTENTSIZE_ERROR 2323 || frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN ) { 2324 info->decompUnavailable = 1; 2325 } else { 2326 info->decompressedSize += frameContentSize; 2327 } 2328 ERROR_IF(ZSTD_getFrameHeader(&header, headerBuffer, numBytesRead) != 0, 2329 info_frame_error, "Error: could not decode frame header"); 2330 info->windowSize = header.windowSize; 2331 /* move to the end of the frame header */ 2332 { size_t const headerSize = ZSTD_frameHeaderSize(headerBuffer, numBytesRead); 2333 ERROR_IF(ZSTD_isError(headerSize), info_frame_error, "Error: could not determine frame header size"); 2334 ERROR_IF(fseek(srcFile, ((long)headerSize)-((long)numBytesRead), SEEK_CUR) != 0, 2335 info_frame_error, "Error: could not move to end of frame header"); 2336 } 2337 2338 /* skip all blocks in the frame */ 2339 { int lastBlock = 0; 2340 do { 2341 BYTE blockHeaderBuffer[3]; 2342 ERROR_IF(fread(blockHeaderBuffer, 1, 3, srcFile) != 3, 2343 info_frame_error, "Error while reading block header"); 2344 { U32 const blockHeader = MEM_readLE24(blockHeaderBuffer); 2345 U32 const blockTypeID = (blockHeader >> 1) & 3; 2346 U32 const isRLE = (blockTypeID == 1); 2347 U32 const isWrongBlock = (blockTypeID == 3); 2348 long const blockSize = isRLE ? 1 : (long)(blockHeader >> 3); 2349 ERROR_IF(isWrongBlock, info_frame_error, "Error: unsupported block type"); 2350 lastBlock = blockHeader & 1; 2351 ERROR_IF(fseek(srcFile, blockSize, SEEK_CUR) != 0, 2352 info_frame_error, "Error: could not skip to end of block"); 2353 } 2354 } while (lastBlock != 1); 2355 } 2356 2357 /* check if checksum is used */ 2358 { BYTE const frameHeaderDescriptor = headerBuffer[4]; 2359 int const contentChecksumFlag = (frameHeaderDescriptor & (1 << 2)) >> 2; 2360 if (contentChecksumFlag) { 2361 info->usesCheck = 1; 2362 ERROR_IF(fseek(srcFile, 4, SEEK_CUR) != 0, 2363 info_frame_error, "Error: could not skip past checksum"); 2364 } } 2365 info->numActualFrames++; 2366 } 2367 /* Skippable frame */ 2368 else if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { 2369 U32 const frameSize = MEM_readLE32(headerBuffer + 4); 2370 long const seek = (long)(8 + frameSize - numBytesRead); 2371 ERROR_IF(LONG_SEEK(srcFile, seek, SEEK_CUR) != 0, 2372 info_frame_error, "Error: could not find end of skippable frame"); 2373 info->numSkippableFrames++; 2374 } 2375 /* unknown content */ 2376 else { 2377 return info_not_zstd; 2378 } 2379 } /* magic number analysis */ 2380 } /* end analyzing frames */ 2381 return info_success; 2382 } 2383 2384 2385 static InfoError 2386 getFileInfo_fileConfirmed(fileInfo_t* info, const char* inFileName) 2387 { 2388 InfoError status; 2389 FILE* const srcFile = FIO_openSrcFile(inFileName); 2390 ERROR_IF(srcFile == NULL, info_file_error, "Error: could not open source file %s", inFileName); 2391 2392 info->compressedSize = UTIL_getFileSize(inFileName); 2393 status = FIO_analyzeFrames(info, srcFile); 2394 2395 fclose(srcFile); 2396 info->nbFiles = 1; 2397 return status; 2398 } 2399 2400 2401 /** getFileInfo() : 2402 * Reads information from file, stores in *info 2403 * @return : InfoError status 2404 */ 2405 static InfoError 2406 getFileInfo(fileInfo_t* info, const char* srcFileName) 2407 { 2408 ERROR_IF(!UTIL_isRegularFile(srcFileName), 2409 info_file_error, "Error : %s is not a file", srcFileName); 2410 return getFileInfo_fileConfirmed(info, srcFileName); 2411 } 2412 2413 2414 static void 2415 displayInfo(const char* inFileName, const fileInfo_t* info, int displayLevel) 2416 { 2417 unsigned const unit = info->compressedSize < (1 MB) ? (1 KB) : (1 MB); 2418 const char* const unitStr = info->compressedSize < (1 MB) ? "KB" : "MB"; 2419 double const windowSizeUnit = (double)info->windowSize / unit; 2420 double const compressedSizeUnit = (double)info->compressedSize / unit; 2421 double const decompressedSizeUnit = (double)info->decompressedSize / unit; 2422 double const ratio = (info->compressedSize == 0) ? 0 : ((double)info->decompressedSize)/info->compressedSize; 2423 const char* const checkString = (info->usesCheck ? "XXH64" : "None"); 2424 if (displayLevel <= 2) { 2425 if (!info->decompUnavailable) { 2426 DISPLAYOUT("%6d %5d %7.2f %2s %9.2f %2s %5.3f %5s %s\n", 2427 info->numSkippableFrames + info->numActualFrames, 2428 info->numSkippableFrames, 2429 compressedSizeUnit, unitStr, decompressedSizeUnit, unitStr, 2430 ratio, checkString, inFileName); 2431 } else { 2432 DISPLAYOUT("%6d %5d %7.2f %2s %5s %s\n", 2433 info->numSkippableFrames + info->numActualFrames, 2434 info->numSkippableFrames, 2435 compressedSizeUnit, unitStr, 2436 checkString, inFileName); 2437 } 2438 } else { 2439 DISPLAYOUT("%s \n", inFileName); 2440 DISPLAYOUT("# Zstandard Frames: %d\n", info->numActualFrames); 2441 if (info->numSkippableFrames) 2442 DISPLAYOUT("# Skippable Frames: %d\n", info->numSkippableFrames); 2443 DISPLAYOUT("Window Size: %.2f %2s (%llu B)\n", 2444 windowSizeUnit, unitStr, 2445 (unsigned long long)info->windowSize); 2446 DISPLAYOUT("Compressed Size: %.2f %2s (%llu B)\n", 2447 compressedSizeUnit, unitStr, 2448 (unsigned long long)info->compressedSize); 2449 if (!info->decompUnavailable) { 2450 DISPLAYOUT("Decompressed Size: %.2f %2s (%llu B)\n", 2451 decompressedSizeUnit, unitStr, 2452 (unsigned long long)info->decompressedSize); 2453 DISPLAYOUT("Ratio: %.4f\n", ratio); 2454 } 2455 DISPLAYOUT("Check: %s\n", checkString); 2456 DISPLAYOUT("\n"); 2457 } 2458 } 2459 2460 static fileInfo_t FIO_addFInfo(fileInfo_t fi1, fileInfo_t fi2) 2461 { 2462 fileInfo_t total; 2463 memset(&total, 0, sizeof(total)); 2464 total.numActualFrames = fi1.numActualFrames + fi2.numActualFrames; 2465 total.numSkippableFrames = fi1.numSkippableFrames + fi2.numSkippableFrames; 2466 total.compressedSize = fi1.compressedSize + fi2.compressedSize; 2467 total.decompressedSize = fi1.decompressedSize + fi2.decompressedSize; 2468 total.decompUnavailable = fi1.decompUnavailable | fi2.decompUnavailable; 2469 total.usesCheck = fi1.usesCheck & fi2.usesCheck; 2470 total.nbFiles = fi1.nbFiles + fi2.nbFiles; 2471 return total; 2472 } 2473 2474 static int 2475 FIO_listFile(fileInfo_t* total, const char* inFileName, int displayLevel) 2476 { 2477 fileInfo_t info; 2478 memset(&info, 0, sizeof(info)); 2479 { InfoError const error = getFileInfo(&info, inFileName); 2480 switch (error) { 2481 case info_frame_error: 2482 /* display error, but provide output */ 2483 DISPLAYLEVEL(1, "Error while parsing \"%s\" \n", inFileName); 2484 break; 2485 case info_not_zstd: 2486 DISPLAYOUT("File \"%s\" not compressed by zstd \n", inFileName); 2487 if (displayLevel > 2) DISPLAYOUT("\n"); 2488 return 1; 2489 case info_file_error: 2490 /* error occurred while opening the file */ 2491 if (displayLevel > 2) DISPLAYOUT("\n"); 2492 return 1; 2493 case info_truncated_input: 2494 DISPLAYOUT("File \"%s\" is truncated \n", inFileName); 2495 if (displayLevel > 2) DISPLAYOUT("\n"); 2496 return 1; 2497 case info_success: 2498 default: 2499 break; 2500 } 2501 2502 displayInfo(inFileName, &info, displayLevel); 2503 *total = FIO_addFInfo(*total, info); 2504 assert(error == info_success || error == info_frame_error); 2505 return error; 2506 } 2507 } 2508 2509 int FIO_listMultipleFiles(unsigned numFiles, const char** filenameTable, int displayLevel) 2510 { 2511 /* ensure no specified input is stdin (needs fseek() capability) */ 2512 { unsigned u; 2513 for (u=0; u<numFiles;u++) { 2514 ERROR_IF(!strcmp (filenameTable[u], stdinmark), 2515 1, "zstd: --list does not support reading from standard input"); 2516 } } 2517 2518 if (numFiles == 0) { 2519 if (!IS_CONSOLE(stdin)) { 2520 DISPLAYLEVEL(1, "zstd: --list does not support reading from standard input \n"); 2521 } 2522 DISPLAYLEVEL(1, "No files given \n"); 2523 return 1; 2524 } 2525 2526 if (displayLevel <= 2) { 2527 DISPLAYOUT("Frames Skips Compressed Uncompressed Ratio Check Filename\n"); 2528 } 2529 { int error = 0; 2530 fileInfo_t total; 2531 memset(&total, 0, sizeof(total)); 2532 total.usesCheck = 1; 2533 /* --list each file, and check for any error */ 2534 { unsigned u; 2535 for (u=0; u<numFiles;u++) { 2536 error |= FIO_listFile(&total, filenameTable[u], displayLevel); 2537 } } 2538 if (numFiles > 1 && displayLevel <= 2) { /* display total */ 2539 unsigned const unit = total.compressedSize < (1 MB) ? (1 KB) : (1 MB); 2540 const char* const unitStr = total.compressedSize < (1 MB) ? "KB" : "MB"; 2541 double const compressedSizeUnit = (double)total.compressedSize / unit; 2542 double const decompressedSizeUnit = (double)total.decompressedSize / unit; 2543 double const ratio = (total.compressedSize == 0) ? 0 : ((double)total.decompressedSize)/total.compressedSize; 2544 const char* const checkString = (total.usesCheck ? "XXH64" : ""); 2545 DISPLAYOUT("----------------------------------------------------------------- \n"); 2546 if (total.decompUnavailable) { 2547 DISPLAYOUT("%6d %5d %7.2f %2s %5s %u files\n", 2548 total.numSkippableFrames + total.numActualFrames, 2549 total.numSkippableFrames, 2550 compressedSizeUnit, unitStr, 2551 checkString, (unsigned)total.nbFiles); 2552 } else { 2553 DISPLAYOUT("%6d %5d %7.2f %2s %9.2f %2s %5.3f %5s %u files\n", 2554 total.numSkippableFrames + total.numActualFrames, 2555 total.numSkippableFrames, 2556 compressedSizeUnit, unitStr, decompressedSizeUnit, unitStr, 2557 ratio, checkString, (unsigned)total.nbFiles); 2558 } } 2559 return error; 2560 } 2561 } 2562 2563 2564 #endif /* #ifndef ZSTD_NODECOMPRESS */ 2565