2 * Copyright (c) Meta Platforms, Inc. and affiliates.
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.
12 /* *************************************
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 */
19 #if defined(__MINGW32__) && !defined(_POSIX_SOURCE)
20 # define _POSIX_SOURCE 1 /* disable %llu warnings with MinGW on Windows */
23 /*-*************************************
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, open, fdopen, fread, _fileno, stdin, stdout */
29 #include <stdlib.h> /* malloc, free */
30 #include <string.h> /* strcmp, strlen */
31 #include <time.h> /* clock_t, to measure process time */
32 #include <fcntl.h> /* O_WRONLY */
34 #include <errno.h> /* errno */
35 #include <limits.h> /* INT_MAX */
37 #include "timefn.h" /* UTIL_getTime, UTIL_clockSpanMicro */
39 #if defined (_MSC_VER)
40 # include <sys/stat.h>
45 #include "fileio_asyncio.h"
46 #include "fileio_common.h"
48 FIO_display_prefs_t g_display_prefs = {2, FIO_ps_auto};
49 UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
51 #define ZSTD_STATIC_LINKING_ONLY /* ZSTD_magicNumber, ZSTD_frameHeaderSize_max */
52 #include "../lib/zstd.h"
53 #include "../lib/zstd_errors.h" /* ZSTD_error_frameParameter_windowTooLarge */
55 #if defined(ZSTD_GZCOMPRESS) || defined(ZSTD_GZDECOMPRESS)
57 # if !defined(z_const)
62 #if defined(ZSTD_LZMACOMPRESS) || defined(ZSTD_LZMADECOMPRESS)
66 #define LZ4_MAGICNUMBER 0x184D2204
67 #if defined(ZSTD_LZ4COMPRESS) || defined(ZSTD_LZ4DECOMPRESS)
68 # define LZ4F_ENABLE_OBSOLETE_ENUMS
69 # include <lz4frame.h>
73 char const* FIO_zlibVersion(void)
75 #if defined(ZSTD_GZCOMPRESS) || defined(ZSTD_GZDECOMPRESS)
82 char const* FIO_lz4Version(void)
84 #if defined(ZSTD_LZ4COMPRESS) || defined(ZSTD_LZ4DECOMPRESS)
85 /* LZ4_versionString() added in v1.7.3 */
86 # if LZ4_VERSION_NUMBER >= 10703
87 return LZ4_versionString();
89 # define ZSTD_LZ4_VERSION LZ4_VERSION_MAJOR.LZ4_VERSION_MINOR.LZ4_VERSION_RELEASE
90 # define ZSTD_LZ4_VERSION_STRING ZSTD_EXPAND_AND_QUOTE(ZSTD_LZ4_VERSION)
91 return ZSTD_LZ4_VERSION_STRING;
98 char const* FIO_lzmaVersion(void)
100 #if defined(ZSTD_LZMACOMPRESS) || defined(ZSTD_LZMADECOMPRESS)
101 return lzma_version_string();
103 return "Unsupported";
108 /*-*************************************
110 ***************************************/
111 #define ADAPT_WINDOWLOG_DEFAULT 23 /* 8 MB */
112 #define DICTSIZE_MAX (32 MB) /* protection against large input (attack scenario) */
116 /* Default file permissions 0666 (modulated by umask) */
117 /* Temporary restricted file permissions are used when we're going to
118 * chmod/chown at the end of the operation. */
120 /* These macros aren't defined on windows. */
121 #define DEFAULT_FILE_PERMISSIONS (S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP|S_IROTH|S_IWOTH)
122 #define TEMPORARY_FILE_PERMISSIONS (S_IRUSR|S_IWUSR)
124 #define DEFAULT_FILE_PERMISSIONS (0666)
125 #define TEMPORARY_FILE_PERMISSIONS (0600)
128 /*-************************************
129 * Signal (Ctrl-C trapping)
130 **************************************/
131 static const char* g_artefact = NULL;
132 static void INThandler(int sig)
134 assert(sig==SIGINT); (void)sig;
135 #if !defined(_MSC_VER)
136 signal(sig, SIG_IGN); /* this invocation generates a buggy warning in Visual Studio */
139 assert(UTIL_isRegularFile(g_artefact));
145 static void addHandler(char const* dstFileName)
147 if (UTIL_isRegularFile(dstFileName)) {
148 g_artefact = dstFileName;
149 signal(SIGINT, INThandler);
155 static void clearHandler(void)
157 if (g_artefact) signal(SIGINT, SIG_DFL);
162 /*-*********************************************************
163 * Termination signal trapping (Print debug stack trace)
164 ***********************************************************/
165 #if defined(__has_feature) && !defined(BACKTRACE_ENABLE) /* Clang compiler */
166 # if (__has_feature(address_sanitizer))
167 # define BACKTRACE_ENABLE 0
168 # endif /* __has_feature(address_sanitizer) */
169 #elif defined(__SANITIZE_ADDRESS__) && !defined(BACKTRACE_ENABLE) /* GCC compiler */
170 # define BACKTRACE_ENABLE 0
173 #if !defined(BACKTRACE_ENABLE)
174 /* automatic detector : backtrace enabled by default on linux+glibc and osx */
175 # if (defined(__linux__) && (defined(__GLIBC__) && !defined(__UCLIBC__))) \
176 || (defined(__APPLE__) && defined(__MACH__))
177 # define BACKTRACE_ENABLE 1
179 # define BACKTRACE_ENABLE 0
183 /* note : after this point, BACKTRACE_ENABLE is necessarily defined */
188 #include <execinfo.h> /* backtrace, backtrace_symbols */
190 #define MAX_STACK_FRAMES 50
192 static void ABRThandler(int sig) {
194 void* addrlist[MAX_STACK_FRAMES];
199 case SIGABRT: name = "SIGABRT"; break;
200 case SIGFPE: name = "SIGFPE"; break;
201 case SIGILL: name = "SIGILL"; break;
202 case SIGINT: name = "SIGINT"; break;
203 case SIGSEGV: name = "SIGSEGV"; break;
204 default: name = "UNKNOWN";
207 DISPLAY("Caught %s signal, printing stack:\n", name);
208 /* Retrieve current stack addresses. */
209 addrlen = backtrace(addrlist, MAX_STACK_FRAMES);
214 /* Create readable strings to each frame. */
215 symbollist = backtrace_symbols(addrlist, addrlen);
216 /* Print the stack trace, excluding calls handling the signal. */
217 for (i = ZSTD_START_SYMBOLLIST_FRAME; i < addrlen; i++) {
218 DISPLAY("%s\n", symbollist[i]);
221 /* Reset and raise the signal so default handler runs. */
222 signal(sig, SIG_DFL);
227 void FIO_addAbortHandler(void)
230 signal(SIGABRT, ABRThandler);
231 signal(SIGFPE, ABRThandler);
232 signal(SIGILL, ABRThandler);
233 signal(SIGSEGV, ABRThandler);
234 signal(SIGBUS, ABRThandler);
238 /*-*************************************
239 * Parameters: FIO_ctx_t
240 ***************************************/
242 /* typedef'd to FIO_ctx_t within fileio.h */
252 int nbFilesProcessed;
253 size_t totalBytesInput;
254 size_t totalBytesOutput;
257 static int FIO_shouldDisplayFileSummary(FIO_ctx_t const* fCtx)
259 return fCtx->nbFilesTotal <= 1 || g_display_prefs.displayLevel >= 3;
262 static int FIO_shouldDisplayMultipleFileSummary(FIO_ctx_t const* fCtx)
264 int const shouldDisplay = (fCtx->nbFilesProcessed >= 1 && fCtx->nbFilesTotal > 1);
265 assert(shouldDisplay || FIO_shouldDisplayFileSummary(fCtx) || fCtx->nbFilesProcessed == 0);
266 return shouldDisplay;
270 /*-*************************************
271 * Parameters: Initialization
272 ***************************************/
274 #define FIO_OVERLAP_LOG_NOTSET 9999
275 #define FIO_LDM_PARAM_NOTSET 9999
278 FIO_prefs_t* FIO_createPreferences(void)
280 FIO_prefs_t* const ret = (FIO_prefs_t*)malloc(sizeof(FIO_prefs_t));
281 if (!ret) EXM_THROW(21, "Allocation error : not enough memory");
283 ret->compressionType = FIO_zstdCompression;
285 ret->sparseFileSupport = ZSTD_SPARSE_DEFAULT;
287 ret->checksumFlag = 1;
288 ret->removeSrcFile = 0;
292 ret->overlapLog = FIO_OVERLAP_LOG_NOTSET;
293 ret->adaptiveMode = 0;
295 ret->minAdaptLevel = -50; /* initializing this value requires a constant, so ZSTD_minCLevel() doesn't work */
296 ret->maxAdaptLevel = 22; /* initializing this value requires a constant, so ZSTD_maxCLevel() doesn't work */
299 ret->ldmMinMatch = 0;
300 ret->ldmBucketSizeLog = FIO_LDM_PARAM_NOTSET;
301 ret->ldmHashRateLog = FIO_LDM_PARAM_NOTSET;
302 ret->streamSrcSize = 0;
303 ret->targetCBlockSize = 0;
304 ret->srcSizeHint = 0;
306 ret->literalCompressionMode = ZSTD_ps_auto;
307 ret->excludeCompressedFiles = 0;
308 ret->allowBlockDevices = 0;
309 ret->asyncIO = AIO_supported();
310 ret->passThrough = -1;
314 FIO_ctx_t* FIO_createContext(void)
316 FIO_ctx_t* const ret = (FIO_ctx_t*)malloc(sizeof(FIO_ctx_t));
317 if (!ret) EXM_THROW(21, "Allocation error : not enough memory");
319 ret->currFileIdx = 0;
320 ret->hasStdinInput = 0;
321 ret->hasStdoutOutput = 0;
322 ret->nbFilesTotal = 1;
323 ret->nbFilesProcessed = 0;
324 ret->totalBytesInput = 0;
325 ret->totalBytesOutput = 0;
329 void FIO_freePreferences(FIO_prefs_t* const prefs)
334 void FIO_freeContext(FIO_ctx_t* const fCtx)
340 /*-*************************************
341 * Parameters: Display Options
342 ***************************************/
344 void FIO_setNotificationLevel(int level) { g_display_prefs.displayLevel=level; }
346 void FIO_setProgressSetting(FIO_progressSetting_e setting) { g_display_prefs.progressSetting = setting; }
349 /*-*************************************
350 * Parameters: Setters
351 ***************************************/
353 /* FIO_prefs_t functions */
355 void FIO_setCompressionType(FIO_prefs_t* const prefs, FIO_compressionType_t compressionType) { prefs->compressionType = compressionType; }
357 void FIO_overwriteMode(FIO_prefs_t* const prefs) { prefs->overwrite = 1; }
359 void FIO_setSparseWrite(FIO_prefs_t* const prefs, int sparse) { prefs->sparseFileSupport = sparse; }
361 void FIO_setDictIDFlag(FIO_prefs_t* const prefs, int dictIDFlag) { prefs->dictIDFlag = dictIDFlag; }
363 void FIO_setChecksumFlag(FIO_prefs_t* const prefs, int checksumFlag) { prefs->checksumFlag = checksumFlag; }
365 void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, int flag) { prefs->removeSrcFile = (flag!=0); }
367 void FIO_setMemLimit(FIO_prefs_t* const prefs, unsigned memLimit) { prefs->memLimit = memLimit; }
369 void FIO_setNbWorkers(FIO_prefs_t* const prefs, int nbWorkers) {
370 #ifndef ZSTD_MULTITHREAD
371 if (nbWorkers > 0) DISPLAYLEVEL(2, "Note : multi-threading is disabled \n");
373 prefs->nbWorkers = nbWorkers;
376 void FIO_setExcludeCompressedFile(FIO_prefs_t* const prefs, int excludeCompressedFiles) { prefs->excludeCompressedFiles = excludeCompressedFiles; }
378 void FIO_setAllowBlockDevices(FIO_prefs_t* const prefs, int allowBlockDevices) { prefs->allowBlockDevices = allowBlockDevices; }
380 void FIO_setBlockSize(FIO_prefs_t* const prefs, int blockSize) {
381 if (blockSize && prefs->nbWorkers==0)
382 DISPLAYLEVEL(2, "Setting block size is useless in single-thread mode \n");
383 prefs->blockSize = blockSize;
386 void FIO_setOverlapLog(FIO_prefs_t* const prefs, int overlapLog){
387 if (overlapLog && prefs->nbWorkers==0)
388 DISPLAYLEVEL(2, "Setting overlapLog is useless in single-thread mode \n");
389 prefs->overlapLog = overlapLog;
392 void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, int adapt) {
393 if ((adapt>0) && (prefs->nbWorkers==0))
394 EXM_THROW(1, "Adaptive mode is not compatible with single thread mode \n");
395 prefs->adaptiveMode = adapt;
398 void FIO_setUseRowMatchFinder(FIO_prefs_t* const prefs, int useRowMatchFinder) {
399 prefs->useRowMatchFinder = useRowMatchFinder;
402 void FIO_setRsyncable(FIO_prefs_t* const prefs, int rsyncable) {
403 if ((rsyncable>0) && (prefs->nbWorkers==0))
404 EXM_THROW(1, "Rsyncable mode is not compatible with single thread mode \n");
405 prefs->rsyncable = rsyncable;
408 void FIO_setStreamSrcSize(FIO_prefs_t* const prefs, size_t streamSrcSize) {
409 prefs->streamSrcSize = streamSrcSize;
412 void FIO_setTargetCBlockSize(FIO_prefs_t* const prefs, size_t targetCBlockSize) {
413 prefs->targetCBlockSize = targetCBlockSize;
416 void FIO_setSrcSizeHint(FIO_prefs_t* const prefs, size_t srcSizeHint) {
417 prefs->srcSizeHint = (int)MIN((size_t)INT_MAX, srcSizeHint);
420 void FIO_setTestMode(FIO_prefs_t* const prefs, int testMode) {
421 prefs->testMode = (testMode!=0);
424 void FIO_setLiteralCompressionMode(
425 FIO_prefs_t* const prefs,
426 ZSTD_paramSwitch_e mode) {
427 prefs->literalCompressionMode = mode;
430 void FIO_setAdaptMin(FIO_prefs_t* const prefs, int minCLevel)
432 #ifndef ZSTD_NOCOMPRESS
433 assert(minCLevel >= ZSTD_minCLevel());
435 prefs->minAdaptLevel = minCLevel;
438 void FIO_setAdaptMax(FIO_prefs_t* const prefs, int maxCLevel)
440 prefs->maxAdaptLevel = maxCLevel;
443 void FIO_setLdmFlag(FIO_prefs_t* const prefs, unsigned ldmFlag) {
444 prefs->ldmFlag = (ldmFlag>0);
447 void FIO_setLdmHashLog(FIO_prefs_t* const prefs, int ldmHashLog) {
448 prefs->ldmHashLog = ldmHashLog;
451 void FIO_setLdmMinMatch(FIO_prefs_t* const prefs, int ldmMinMatch) {
452 prefs->ldmMinMatch = ldmMinMatch;
455 void FIO_setLdmBucketSizeLog(FIO_prefs_t* const prefs, int ldmBucketSizeLog) {
456 prefs->ldmBucketSizeLog = ldmBucketSizeLog;
460 void FIO_setLdmHashRateLog(FIO_prefs_t* const prefs, int ldmHashRateLog) {
461 prefs->ldmHashRateLog = ldmHashRateLog;
464 void FIO_setPatchFromMode(FIO_prefs_t* const prefs, int value)
466 prefs->patchFromMode = value != 0;
469 void FIO_setContentSize(FIO_prefs_t* const prefs, int value)
471 prefs->contentSize = value != 0;
474 void FIO_setAsyncIOFlag(FIO_prefs_t* const prefs, int value) {
475 #ifdef ZSTD_MULTITHREAD
476 prefs->asyncIO = value;
480 DISPLAYLEVEL(2, "Note : asyncio is disabled (lack of multithreading support) \n");
484 void FIO_setPassThroughFlag(FIO_prefs_t* const prefs, int value) {
485 prefs->passThrough = (value != 0);
488 void FIO_setMMapDict(FIO_prefs_t* const prefs, ZSTD_paramSwitch_e value)
490 prefs->mmapDict = value;
493 /* FIO_ctx_t functions */
495 void FIO_setHasStdoutOutput(FIO_ctx_t* const fCtx, int value) {
496 fCtx->hasStdoutOutput = value;
499 void FIO_setNbFilesTotal(FIO_ctx_t* const fCtx, int value)
501 fCtx->nbFilesTotal = value;
504 void FIO_determineHasStdinInput(FIO_ctx_t* const fCtx, const FileNamesTable* const filenames) {
506 for ( ; i < filenames->tableSize; ++i) {
507 if (!strcmp(stdinmark, filenames->fileNames[i])) {
508 fCtx->hasStdinInput = 1;
514 /*-*************************************
516 ***************************************/
517 /** FIO_removeFile() :
518 * @result : Unlink `fileName`, even if it's read-only */
519 static int FIO_removeFile(const char* path)
522 if (!UTIL_stat(path, &statbuf)) {
523 DISPLAYLEVEL(2, "zstd: Failed to stat %s while trying to remove it\n", path);
526 if (!UTIL_isRegularFileStat(&statbuf)) {
527 DISPLAYLEVEL(2, "zstd: Refusing to remove non-regular file %s\n", path);
531 /* windows doesn't allow remove read-only files,
532 * so try to make it writable first */
533 if (!(statbuf.st_mode & _S_IWRITE)) {
534 UTIL_chmod(path, &statbuf, _S_IWRITE);
540 /** FIO_openSrcFile() :
541 * condition : `srcFileName` must be non-NULL. `prefs` may be NULL.
542 * @result : FILE* to `srcFileName`, or NULL if it fails */
543 static FILE* FIO_openSrcFile(const FIO_prefs_t* const prefs, const char* srcFileName, stat_t* statbuf)
545 int allowBlockDevices = prefs != NULL ? prefs->allowBlockDevices : 0;
546 assert(srcFileName != NULL);
547 assert(statbuf != NULL);
548 if (!strcmp (srcFileName, stdinmark)) {
549 DISPLAYLEVEL(4,"Using stdin for input \n");
550 SET_BINARY_MODE(stdin);
554 if (!UTIL_stat(srcFileName, statbuf)) {
555 DISPLAYLEVEL(1, "zstd: can't stat %s : %s -- ignored \n",
556 srcFileName, strerror(errno));
560 if (!UTIL_isRegularFileStat(statbuf)
561 && !UTIL_isFIFOStat(statbuf)
562 && !(allowBlockDevices && UTIL_isBlockDevStat(statbuf))
564 DISPLAYLEVEL(1, "zstd: %s is not a regular file -- ignored \n",
569 { FILE* const f = fopen(srcFileName, "rb");
571 DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno));
576 /** FIO_openDstFile() :
577 * condition : `dstFileName` must be non-NULL.
578 * @result : FILE* to `dstFileName`, or NULL if it fails */
580 FIO_openDstFile(FIO_ctx_t* fCtx, FIO_prefs_t* const prefs,
581 const char* srcFileName, const char* dstFileName,
586 if (prefs->testMode) return NULL; /* do not open file in test mode */
588 assert(dstFileName != NULL);
589 if (!strcmp (dstFileName, stdoutmark)) {
590 DISPLAYLEVEL(4,"Using stdout for output \n");
591 SET_BINARY_MODE(stdout);
592 if (prefs->sparseFileSupport == 1) {
593 prefs->sparseFileSupport = 0;
594 DISPLAYLEVEL(4, "Sparse File Support is automatically disabled on stdout ; try --sparse \n");
599 /* ensure dst is not the same as src */
600 if (srcFileName != NULL && UTIL_isSameFile(srcFileName, dstFileName)) {
601 DISPLAYLEVEL(1, "zstd: Refusing to open an output file which will overwrite the input file \n");
605 isDstRegFile = UTIL_isRegularFile(dstFileName); /* invoke once */
606 if (prefs->sparseFileSupport == 1) {
607 prefs->sparseFileSupport = ZSTD_SPARSE_DEFAULT;
609 prefs->sparseFileSupport = 0;
610 DISPLAYLEVEL(4, "Sparse File Support is disabled when output is not a file \n");
615 /* Check if destination file already exists */
617 /* this test does not work on Windows :
618 * `NUL` and `nul` are detected as regular files */
619 if (!strcmp(dstFileName, nulmark)) {
620 EXM_THROW(40, "%s is unexpectedly categorized as a regular file",
624 if (!prefs->overwrite) {
625 if (g_display_prefs.displayLevel <= 1) {
626 /* No interaction possible */
627 DISPLAYLEVEL(1, "zstd: %s already exists; not overwritten \n",
631 DISPLAY("zstd: %s already exists; ", dstFileName);
632 if (UTIL_requireUserConfirmation("overwrite (y/n) ? ", "Not overwritten \n", "yY", fCtx->hasStdinInput))
636 FIO_removeFile(dstFileName);
641 /* Windows requires opening the file as a "binary" file to avoid
642 * mangling. This macro doesn't exist on unix. */
643 const int openflags = O_WRONLY|O_CREAT|O_TRUNC|O_BINARY;
644 const int fd = _open(dstFileName, openflags, mode);
647 f = _fdopen(fd, "wb");
650 const int openflags = O_WRONLY|O_CREAT|O_TRUNC;
651 const int fd = open(dstFileName, openflags, mode);
654 f = fdopen(fd, "wb");
658 DISPLAYLEVEL(1, "zstd: %s: %s\n", dstFileName, strerror(errno));
660 /* An increased buffer size can provide a significant performance
661 * boost on some platforms. Note that providing a NULL buf with a
662 * size that's not 0 is not defined in ANSI C, but is defined in an
663 * extension. There are three possibilities here:
664 * 1. Libc supports the extended version and everything is good.
665 * 2. Libc ignores the size when buf is NULL, in which case
666 * everything will continue as if we didn't call `setvbuf()`.
667 * 3. We fail the call and execution continues but a warning
668 * message might be shown.
669 * In all cases due execution continues. For now, I believe that
670 * this is a more cost-effective solution than managing the buffers
671 * allocations ourselves (will require an API change).
673 if (setvbuf(f, NULL, _IOFBF, 1 MB)) {
674 DISPLAYLEVEL(2, "Warning: setvbuf failed for %s\n", dstFileName);
682 /* FIO_getDictFileStat() :
684 static void FIO_getDictFileStat(const char* fileName, stat_t* dictFileStat) {
685 assert(dictFileStat != NULL);
686 if (fileName == NULL) return;
688 if (!UTIL_stat(fileName, dictFileStat)) {
689 EXM_THROW(31, "Stat failed on dictionary file %s: %s", fileName, strerror(errno));
692 if (!UTIL_isRegularFileStat(dictFileStat)) {
693 EXM_THROW(32, "Dictionary %s must be a regular file.", fileName);
697 /* FIO_setDictBufferMalloc() :
698 * allocates a buffer, pointed by `dict->dictBuffer`,
699 * loads `filename` content into it, up to DICTSIZE_MAX bytes.
700 * @return : loaded size
701 * if fileName==NULL, returns 0 and a NULL pointer
703 static size_t FIO_setDictBufferMalloc(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
707 void** bufferPtr = &dict->dictBuffer;
709 assert(bufferPtr != NULL);
710 assert(dictFileStat != NULL);
712 if (fileName == NULL) return 0;
714 DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
716 fileHandle = fopen(fileName, "rb");
718 if (fileHandle == NULL) {
719 EXM_THROW(33, "Couldn't open dictionary %s: %s", fileName, strerror(errno));
722 fileSize = UTIL_getFileSizeStat(dictFileStat);
724 size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX;
725 if (fileSize > dictSizeMax) {
726 EXM_THROW(34, "Dictionary file %s is too large (> %u bytes)",
727 fileName, (unsigned)dictSizeMax); /* avoid extreme cases */
730 *bufferPtr = malloc((size_t)fileSize);
731 if (*bufferPtr==NULL) EXM_THROW(34, "%s", strerror(errno));
732 { size_t const readSize = fread(*bufferPtr, 1, (size_t)fileSize, fileHandle);
733 if (readSize != fileSize) {
734 EXM_THROW(35, "Error reading dictionary file %s : %s",
735 fileName, strerror(errno));
739 return (size_t)fileSize;
742 #if (PLATFORM_POSIX_VERSION > 0)
743 #include <sys/mman.h>
744 static void FIO_munmap(FIO_Dict_t* dict)
746 munmap(dict->dictBuffer, dict->dictBufferSize);
747 dict->dictBuffer = NULL;
748 dict->dictBufferSize = 0;
750 static size_t FIO_setDictBufferMMap(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
754 void** bufferPtr = &dict->dictBuffer;
756 assert(bufferPtr != NULL);
757 assert(dictFileStat != NULL);
759 if (fileName == NULL) return 0;
761 DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
763 fileHandle = open(fileName, O_RDONLY);
765 if (fileHandle == -1) {
766 EXM_THROW(33, "Couldn't open dictionary %s: %s", fileName, strerror(errno));
769 fileSize = UTIL_getFileSizeStat(dictFileStat);
771 size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX;
772 if (fileSize > dictSizeMax) {
773 EXM_THROW(34, "Dictionary file %s is too large (> %u bytes)",
774 fileName, (unsigned)dictSizeMax); /* avoid extreme cases */
778 *bufferPtr = mmap(NULL, (size_t)fileSize, PROT_READ, MAP_PRIVATE, fileHandle, 0);
779 if (*bufferPtr==NULL) EXM_THROW(34, "%s", strerror(errno));
782 return (size_t)fileSize;
784 #elif defined(_MSC_VER) || defined(_WIN32)
786 static void FIO_munmap(FIO_Dict_t* dict)
788 UnmapViewOfFile(dict->dictBuffer);
789 CloseHandle(dict->dictHandle);
790 dict->dictBuffer = NULL;
791 dict->dictBufferSize = 0;
793 static size_t FIO_setDictBufferMMap(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
795 HANDLE fileHandle, mapping;
797 void** bufferPtr = &dict->dictBuffer;
799 assert(bufferPtr != NULL);
800 assert(dictFileStat != NULL);
802 if (fileName == NULL) return 0;
804 DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
806 fileHandle = CreateFileA(fileName, GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_READONLY, NULL);
808 if (fileHandle == INVALID_HANDLE_VALUE) {
809 EXM_THROW(33, "Couldn't open dictionary %s: %s", fileName, strerror(errno));
812 fileSize = UTIL_getFileSizeStat(dictFileStat);
814 size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX;
815 if (fileSize > dictSizeMax) {
816 EXM_THROW(34, "Dictionary file %s is too large (> %u bytes)",
817 fileName, (unsigned)dictSizeMax); /* avoid extreme cases */
821 mapping = CreateFileMapping(fileHandle, NULL, PAGE_READONLY, 0, 0, NULL);
822 if (mapping == NULL) {
823 EXM_THROW(35, "Couldn't map dictionary %s: %s", fileName, strerror(errno));
826 *bufferPtr = MapViewOfFile(mapping, FILE_MAP_READ, 0, 0, (DWORD)fileSize); /* we can only cast to DWORD here because dictSize <= 2GB */
827 if (*bufferPtr==NULL) EXM_THROW(36, "%s", strerror(errno));
829 dict->dictHandle = fileHandle;
830 return (size_t)fileSize;
833 static size_t FIO_setDictBufferMMap(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
835 return FIO_setDictBufferMalloc(dict, fileName, prefs, dictFileStat);
837 static void FIO_munmap(FIO_Dict_t* dict) {
838 free(dict->dictBuffer);
839 dict->dictBuffer = NULL;
840 dict->dictBufferSize = 0;
844 static void FIO_freeDict(FIO_Dict_t* dict) {
845 if (dict->dictBufferType == FIO_mallocDict) {
846 free(dict->dictBuffer);
847 dict->dictBuffer = NULL;
848 dict->dictBufferSize = 0;
849 } else if (dict->dictBufferType == FIO_mmapDict) {
852 assert(0); /* Should not reach this case */
856 static void FIO_initDict(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat, FIO_dictBufferType_t dictBufferType) {
857 dict->dictBufferType = dictBufferType;
858 if (dict->dictBufferType == FIO_mallocDict) {
859 dict->dictBufferSize = FIO_setDictBufferMalloc(dict, fileName, prefs, dictFileStat);
860 } else if (dict->dictBufferType == FIO_mmapDict) {
861 dict->dictBufferSize = FIO_setDictBufferMMap(dict, fileName, prefs, dictFileStat);
863 assert(0); /* Should not reach this case */
868 /* FIO_checkFilenameCollisions() :
869 * Checks for and warns if there are any files that would have the same output path
871 int FIO_checkFilenameCollisions(const char** filenameTable, unsigned nbFiles) {
872 const char **filenameTableSorted, *prevElem, *filename;
875 filenameTableSorted = (const char**) malloc(sizeof(char*) * nbFiles);
876 if (!filenameTableSorted) {
877 DISPLAYLEVEL(1, "Allocation error during filename collision checking \n");
881 for (u = 0; u < nbFiles; ++u) {
882 filename = strrchr(filenameTable[u], PATH_SEP);
883 if (filename == NULL) {
884 filenameTableSorted[u] = filenameTable[u];
886 filenameTableSorted[u] = filename+1;
890 qsort((void*)filenameTableSorted, nbFiles, sizeof(char*), UTIL_compareStr);
891 prevElem = filenameTableSorted[0];
892 for (u = 1; u < nbFiles; ++u) {
893 if (strcmp(prevElem, filenameTableSorted[u]) == 0) {
894 DISPLAYLEVEL(2, "WARNING: Two files have same filename: %s\n", prevElem);
896 prevElem = filenameTableSorted[u];
899 free((void*)filenameTableSorted);
904 extractFilename(const char* path, char separator)
906 const char* search = strrchr(path, separator);
907 if (search == NULL) return path;
911 /* FIO_createFilename_fromOutDir() :
912 * Takes a source file name and specified output directory, and
913 * allocates memory for and returns a pointer to final path.
914 * This function never returns an error (it may abort() in case of pb)
917 FIO_createFilename_fromOutDir(const char* path, const char* outDirName, const size_t suffixLen)
919 const char* filenameStart;
923 #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__) /* windows support */
929 filenameStart = extractFilename(path, separator);
930 #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__) /* windows support */
931 filenameStart = extractFilename(filenameStart, '/'); /* sometimes, '/' separator is also used on Windows (mingw+msys2) */
934 result = (char*) calloc(1, strlen(outDirName) + 1 + strlen(filenameStart) + suffixLen + 1);
936 EXM_THROW(30, "zstd: FIO_createFilename_fromOutDir: %s", strerror(errno));
939 memcpy(result, outDirName, strlen(outDirName));
940 if (outDirName[strlen(outDirName)-1] == separator) {
941 memcpy(result + strlen(outDirName), filenameStart, strlen(filenameStart));
943 memcpy(result + strlen(outDirName), &separator, 1);
944 memcpy(result + strlen(outDirName) + 1, filenameStart, strlen(filenameStart));
951 * gives position of highest bit.
952 * note : only works for v > 0 !
954 static unsigned FIO_highbit64(unsigned long long v)
959 while (v) { v >>= 1; count++; }
963 static void FIO_adjustMemLimitForPatchFromMode(FIO_prefs_t* const prefs,
964 unsigned long long const dictSize,
965 unsigned long long const maxSrcFileSize)
967 unsigned long long maxSize = MAX(prefs->memLimit, MAX(dictSize, maxSrcFileSize));
968 unsigned const maxWindowSize = (1U << ZSTD_WINDOWLOG_MAX);
969 if (maxSize == UTIL_FILESIZE_UNKNOWN)
970 EXM_THROW(42, "Using --patch-from with stdin requires --stream-size");
971 assert(maxSize != UTIL_FILESIZE_UNKNOWN);
972 if (maxSize > maxWindowSize)
973 EXM_THROW(42, "Can't handle files larger than %u GB\n", maxWindowSize/(1 GB));
974 FIO_setMemLimit(prefs, (unsigned)maxSize);
977 /* FIO_multiFilesConcatWarning() :
978 * This function handles logic when processing multiple files with -o or -c, displaying the appropriate warnings/prompts.
979 * Returns 1 if the console should abort, 0 if console should proceed.
981 * If output is stdout or test mode is active, check that `--rm` disabled.
983 * If there is just 1 file to process, zstd will proceed as usual.
984 * If each file get processed into its own separate destination file, proceed as usual.
986 * When multiple files are processed into a single output,
987 * display a warning message, then disable --rm if it's set.
989 * If -f is specified or if output is stdout, just proceed.
990 * If output is set with -o, prompt for confirmation.
992 static int FIO_multiFilesConcatWarning(const FIO_ctx_t* fCtx, FIO_prefs_t* prefs, const char* outFileName, int displayLevelCutoff)
994 if (fCtx->hasStdoutOutput) {
995 if (prefs->removeSrcFile)
996 /* this should not happen ; hard fail, to protect user's data
997 * note: this should rather be an assert(), but we want to be certain that user's data will not be wiped out in case it nonetheless happen */
998 EXM_THROW(43, "It's not allowed to remove input files when processed output is piped to stdout. "
999 "This scenario is not supposed to be possible. "
1000 "This is a programming error. File an issue for it to be fixed.");
1002 if (prefs->testMode) {
1003 if (prefs->removeSrcFile)
1004 /* this should not happen ; hard fail, to protect user's data
1005 * note: this should rather be an assert(), but we want to be certain that user's data will not be wiped out in case it nonetheless happen */
1006 EXM_THROW(43, "Test mode shall not remove input files! "
1007 "This scenario is not supposed to be possible. "
1008 "This is a programming error. File an issue for it to be fixed.");
1012 if (fCtx->nbFilesTotal == 1) return 0;
1013 assert(fCtx->nbFilesTotal > 1);
1015 if (!outFileName) return 0;
1017 if (fCtx->hasStdoutOutput) {
1018 DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into stdout. \n");
1020 DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into a single output file: %s \n", outFileName);
1022 DISPLAYLEVEL(2, "The concatenated output CANNOT regenerate original file names nor directory structure. \n")
1024 /* multi-input into single output : --rm is not allowed */
1025 if (prefs->removeSrcFile) {
1026 DISPLAYLEVEL(2, "Since it's a destructive operation, input files will not be removed. \n");
1027 prefs->removeSrcFile = 0;
1030 if (fCtx->hasStdoutOutput) return 0;
1031 if (prefs->overwrite) return 0;
1033 /* multiple files concatenated into single destination file using -o without -f */
1034 if (g_display_prefs.displayLevel <= displayLevelCutoff) {
1035 /* quiet mode => no prompt => fail automatically */
1036 DISPLAYLEVEL(1, "Concatenating multiple processed inputs into a single output loses file metadata. \n");
1037 DISPLAYLEVEL(1, "Aborting. \n");
1040 /* normal mode => prompt */
1041 return UTIL_requireUserConfirmation("Proceed? (y/n): ", "Aborting...", "yY", fCtx->hasStdinInput);
1044 static ZSTD_inBuffer setInBuffer(const void* buf, size_t s, size_t pos)
1053 static ZSTD_outBuffer setOutBuffer(void* buf, size_t s, size_t pos)
1062 #ifndef ZSTD_NOCOMPRESS
1064 /* **********************************************************************
1066 ************************************************************************/
1069 const char* dictFileName;
1070 stat_t dictFileStat;
1072 WritePoolCtx_t *writeCtx;
1073 ReadPoolCtx_t *readCtx;
1076 /** ZSTD_cycleLog() :
1077 * condition for correct operation : hashLog > 1 */
1078 static U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
1080 U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);
1081 assert(hashLog > 1);
1082 return hashLog - btScale;
1085 static void FIO_adjustParamsForPatchFromMode(FIO_prefs_t* const prefs,
1086 ZSTD_compressionParameters* comprParams,
1087 unsigned long long const dictSize,
1088 unsigned long long const maxSrcFileSize,
1091 unsigned const fileWindowLog = FIO_highbit64(maxSrcFileSize) + 1;
1092 ZSTD_compressionParameters const cParams = ZSTD_getCParams(cLevel, (size_t)maxSrcFileSize, (size_t)dictSize);
1093 FIO_adjustMemLimitForPatchFromMode(prefs, dictSize, maxSrcFileSize);
1094 if (fileWindowLog > ZSTD_WINDOWLOG_MAX)
1095 DISPLAYLEVEL(1, "Max window log exceeded by file (compression ratio will suffer)\n");
1096 comprParams->windowLog = MAX(ZSTD_WINDOWLOG_MIN, MIN(ZSTD_WINDOWLOG_MAX, fileWindowLog));
1097 if (fileWindowLog > ZSTD_cycleLog(cParams.chainLog, cParams.strategy)) {
1098 if (!prefs->ldmFlag)
1099 DISPLAYLEVEL(2, "long mode automatically triggered\n");
1100 FIO_setLdmFlag(prefs, 1);
1102 if (cParams.strategy >= ZSTD_btopt) {
1103 DISPLAYLEVEL(3, "[Optimal parser notes] Consider the following to improve patch size at the cost of speed:\n");
1104 DISPLAYLEVEL(3, "- Use --single-thread mode in the zstd cli\n");
1105 DISPLAYLEVEL(3, "- Set a larger targetLength (e.g. --zstd=targetLength=4096)\n");
1106 DISPLAYLEVEL(3, "- Set a larger chainLog (e.g. --zstd=chainLog=%u)\n", ZSTD_CHAINLOG_MAX);
1107 DISPLAYLEVEL(3, "Also consider playing around with searchLog and hashLog\n");
1111 static cRess_t FIO_createCResources(FIO_prefs_t* const prefs,
1112 const char* dictFileName, unsigned long long const maxSrcFileSize,
1113 int cLevel, ZSTD_compressionParameters comprParams) {
1114 int useMMap = prefs->mmapDict == ZSTD_ps_enable;
1115 int forceNoUseMMap = prefs->mmapDict == ZSTD_ps_disable;
1116 FIO_dictBufferType_t dictBufferType;
1118 memset(&ress, 0, sizeof(ress));
1120 DISPLAYLEVEL(6, "FIO_createCResources \n");
1121 ress.cctx = ZSTD_createCCtx();
1122 if (ress.cctx == NULL)
1123 EXM_THROW(30, "allocation error (%s): can't create ZSTD_CCtx",
1126 FIO_getDictFileStat(dictFileName, &ress.dictFileStat);
1128 /* need to update memLimit before calling createDictBuffer
1129 * because of memLimit check inside it */
1130 if (prefs->patchFromMode) {
1131 U64 const dictSize = UTIL_getFileSizeStat(&ress.dictFileStat);
1132 unsigned long long const ssSize = (unsigned long long)prefs->streamSrcSize;
1133 useMMap |= dictSize > prefs->memLimit;
1134 FIO_adjustParamsForPatchFromMode(prefs, &comprParams, dictSize, ssSize > 0 ? ssSize : maxSrcFileSize, cLevel);
1137 dictBufferType = (useMMap && !forceNoUseMMap) ? FIO_mmapDict : FIO_mallocDict;
1138 FIO_initDict(&ress.dict, dictFileName, prefs, &ress.dictFileStat, dictBufferType); /* works with dictFileName==NULL */
1140 ress.writeCtx = AIO_WritePool_create(prefs, ZSTD_CStreamOutSize());
1141 ress.readCtx = AIO_ReadPool_create(prefs, ZSTD_CStreamInSize());
1143 /* Advanced parameters, including dictionary */
1144 if (dictFileName && (ress.dict.dictBuffer==NULL))
1145 EXM_THROW(32, "allocation error : can't create dictBuffer");
1146 ress.dictFileName = dictFileName;
1148 if (prefs->adaptiveMode && !prefs->ldmFlag && !comprParams.windowLog)
1149 comprParams.windowLog = ADAPT_WINDOWLOG_DEFAULT;
1151 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_contentSizeFlag, prefs->contentSize) ); /* always enable content size when available (note: supposed to be default) */
1152 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_dictIDFlag, prefs->dictIDFlag) );
1153 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_checksumFlag, prefs->checksumFlag) );
1154 /* compression level */
1155 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, cLevel) );
1156 /* max compressed block size */
1157 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetCBlockSize, (int)prefs->targetCBlockSize) );
1158 /* source size hint */
1159 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_srcSizeHint, (int)prefs->srcSizeHint) );
1160 /* long distance matching */
1161 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_enableLongDistanceMatching, prefs->ldmFlag) );
1162 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashLog, prefs->ldmHashLog) );
1163 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmMinMatch, prefs->ldmMinMatch) );
1164 if (prefs->ldmBucketSizeLog != FIO_LDM_PARAM_NOTSET) {
1165 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmBucketSizeLog, prefs->ldmBucketSizeLog) );
1167 if (prefs->ldmHashRateLog != FIO_LDM_PARAM_NOTSET) {
1168 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashRateLog, prefs->ldmHashRateLog) );
1170 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_useRowMatchFinder, prefs->useRowMatchFinder));
1171 /* compression parameters */
1172 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_windowLog, (int)comprParams.windowLog) );
1173 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_chainLog, (int)comprParams.chainLog) );
1174 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_hashLog, (int)comprParams.hashLog) );
1175 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_searchLog, (int)comprParams.searchLog) );
1176 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_minMatch, (int)comprParams.minMatch) );
1177 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetLength, (int)comprParams.targetLength) );
1178 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_strategy, (int)comprParams.strategy) );
1179 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_literalCompressionMode, (int)prefs->literalCompressionMode) );
1180 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_enableDedicatedDictSearch, 1) );
1181 /* multi-threading */
1182 #ifdef ZSTD_MULTITHREAD
1183 DISPLAYLEVEL(5,"set nb workers = %u \n", prefs->nbWorkers);
1184 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_nbWorkers, prefs->nbWorkers) );
1185 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_jobSize, prefs->blockSize) );
1186 if (prefs->overlapLog != FIO_OVERLAP_LOG_NOTSET) {
1187 DISPLAYLEVEL(3,"set overlapLog = %u \n", prefs->overlapLog);
1188 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_overlapLog, prefs->overlapLog) );
1190 CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_rsyncable, prefs->rsyncable) );
1193 if (prefs->patchFromMode) {
1194 CHECK( ZSTD_CCtx_refPrefix(ress.cctx, ress.dict.dictBuffer, ress.dict.dictBufferSize) );
1196 CHECK( ZSTD_CCtx_loadDictionary_byReference(ress.cctx, ress.dict.dictBuffer, ress.dict.dictBufferSize) );
1202 static void FIO_freeCResources(cRess_t* const ress)
1204 FIO_freeDict(&(ress->dict));
1205 AIO_WritePool_free(ress->writeCtx);
1206 AIO_ReadPool_free(ress->readCtx);
1207 ZSTD_freeCStream(ress->cctx); /* never fails */
1211 #ifdef ZSTD_GZCOMPRESS
1212 static unsigned long long
1213 FIO_compressGzFrame(const cRess_t* ress, /* buffers & handlers are used, but not changed */
1214 const char* srcFileName, U64 const srcFileSize,
1215 int compressionLevel, U64* readsize)
1217 unsigned long long inFileSize = 0, outFileSize = 0;
1219 IOJob_t *writeJob = NULL;
1221 if (compressionLevel > Z_BEST_COMPRESSION)
1222 compressionLevel = Z_BEST_COMPRESSION;
1224 strm.zalloc = Z_NULL;
1225 strm.zfree = Z_NULL;
1226 strm.opaque = Z_NULL;
1228 { int const ret = deflateInit2(&strm, compressionLevel, Z_DEFLATED,
1229 15 /* maxWindowLogSize */ + 16 /* gzip only */,
1230 8, Z_DEFAULT_STRATEGY); /* see https://www.zlib.net/manual.html */
1232 EXM_THROW(71, "zstd: %s: deflateInit2 error %d \n", srcFileName, ret);
1235 writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
1238 strm.next_out = (Bytef*)writeJob->buffer;
1239 strm.avail_out = (uInt)writeJob->bufferSize;
1243 if (strm.avail_in == 0) {
1244 AIO_ReadPool_fillBuffer(ress->readCtx, ZSTD_CStreamInSize());
1245 if (ress->readCtx->srcBufferLoaded == 0) break;
1246 inFileSize += ress->readCtx->srcBufferLoaded;
1247 strm.next_in = (z_const unsigned char*)ress->readCtx->srcBuffer;
1248 strm.avail_in = (uInt)ress->readCtx->srcBufferLoaded;
1252 size_t const availBefore = strm.avail_in;
1253 ret = deflate(&strm, Z_NO_FLUSH);
1254 AIO_ReadPool_consumeBytes(ress->readCtx, availBefore - strm.avail_in);
1258 EXM_THROW(72, "zstd: %s: deflate error %d \n", srcFileName, ret);
1259 { size_t const cSize = writeJob->bufferSize - strm.avail_out;
1261 writeJob->usedBufferSize = cSize;
1262 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
1263 outFileSize += cSize;
1264 strm.next_out = (Bytef*)writeJob->buffer;
1265 strm.avail_out = (uInt)writeJob->bufferSize;
1267 if (srcFileSize == UTIL_FILESIZE_UNKNOWN) {
1268 DISPLAYUPDATE_PROGRESS(
1269 "\rRead : %u MB ==> %.2f%% ",
1270 (unsigned)(inFileSize>>20),
1271 (double)outFileSize/(double)inFileSize*100)
1273 DISPLAYUPDATE_PROGRESS(
1274 "\rRead : %u / %u MB ==> %.2f%% ",
1275 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
1276 (double)outFileSize/(double)inFileSize*100);
1280 int const ret = deflate(&strm, Z_FINISH);
1281 { size_t const cSize = writeJob->bufferSize - strm.avail_out;
1283 writeJob->usedBufferSize = cSize;
1284 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
1285 outFileSize += cSize;
1286 strm.next_out = (Bytef*)writeJob->buffer;
1287 strm.avail_out = (uInt)writeJob->bufferSize;
1289 if (ret == Z_STREAM_END) break;
1290 if (ret != Z_BUF_ERROR)
1291 EXM_THROW(77, "zstd: %s: deflate error %d \n", srcFileName, ret);
1294 { int const ret = deflateEnd(&strm);
1296 EXM_THROW(79, "zstd: %s: deflateEnd error %d \n", srcFileName, ret);
1298 *readsize = inFileSize;
1299 AIO_WritePool_releaseIoJob(writeJob);
1300 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
1306 #ifdef ZSTD_LZMACOMPRESS
1307 static unsigned long long
1308 FIO_compressLzmaFrame(cRess_t* ress,
1309 const char* srcFileName, U64 const srcFileSize,
1310 int compressionLevel, U64* readsize, int plain_lzma)
1312 unsigned long long inFileSize = 0, outFileSize = 0;
1313 lzma_stream strm = LZMA_STREAM_INIT;
1314 lzma_action action = LZMA_RUN;
1316 IOJob_t *writeJob = NULL;
1318 if (compressionLevel < 0) compressionLevel = 0;
1319 if (compressionLevel > 9) compressionLevel = 9;
1322 lzma_options_lzma opt_lzma;
1323 if (lzma_lzma_preset(&opt_lzma, compressionLevel))
1324 EXM_THROW(81, "zstd: %s: lzma_lzma_preset error", srcFileName);
1325 ret = lzma_alone_encoder(&strm, &opt_lzma); /* LZMA */
1327 EXM_THROW(82, "zstd: %s: lzma_alone_encoder error %d", srcFileName, ret);
1329 ret = lzma_easy_encoder(&strm, compressionLevel, LZMA_CHECK_CRC64); /* XZ */
1331 EXM_THROW(83, "zstd: %s: lzma_easy_encoder error %d", srcFileName, ret);
1334 writeJob =AIO_WritePool_acquireJob(ress->writeCtx);
1335 strm.next_out = (BYTE*)writeJob->buffer;
1336 strm.avail_out = writeJob->bufferSize;
1341 if (strm.avail_in == 0) {
1342 size_t const inSize = AIO_ReadPool_fillBuffer(ress->readCtx, ZSTD_CStreamInSize());
1343 if (ress->readCtx->srcBufferLoaded == 0) action = LZMA_FINISH;
1344 inFileSize += inSize;
1345 strm.next_in = (BYTE const*)ress->readCtx->srcBuffer;
1346 strm.avail_in = ress->readCtx->srcBufferLoaded;
1350 size_t const availBefore = strm.avail_in;
1351 ret = lzma_code(&strm, action);
1352 AIO_ReadPool_consumeBytes(ress->readCtx, availBefore - strm.avail_in);
1356 if (ret != LZMA_OK && ret != LZMA_STREAM_END)
1357 EXM_THROW(84, "zstd: %s: lzma_code encoding error %d", srcFileName, ret);
1358 { size_t const compBytes = writeJob->bufferSize - strm.avail_out;
1360 writeJob->usedBufferSize = compBytes;
1361 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
1362 outFileSize += compBytes;
1363 strm.next_out = (BYTE*)writeJob->buffer;
1364 strm.avail_out = writeJob->bufferSize;
1366 if (srcFileSize == UTIL_FILESIZE_UNKNOWN)
1367 DISPLAYUPDATE_PROGRESS("\rRead : %u MB ==> %.2f%%",
1368 (unsigned)(inFileSize>>20),
1369 (double)outFileSize/(double)inFileSize*100)
1371 DISPLAYUPDATE_PROGRESS("\rRead : %u / %u MB ==> %.2f%%",
1372 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
1373 (double)outFileSize/(double)inFileSize*100);
1374 if (ret == LZMA_STREAM_END) break;
1378 *readsize = inFileSize;
1380 AIO_WritePool_releaseIoJob(writeJob);
1381 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
1387 #ifdef ZSTD_LZ4COMPRESS
1389 #if LZ4_VERSION_NUMBER <= 10600
1390 #define LZ4F_blockLinked blockLinked
1391 #define LZ4F_max64KB max64KB
1394 static int FIO_LZ4_GetBlockSize_FromBlockId (int id) { return (1 << (8 + (2 * id))); }
1396 static unsigned long long
1397 FIO_compressLz4Frame(cRess_t* ress,
1398 const char* srcFileName, U64 const srcFileSize,
1399 int compressionLevel, int checksumFlag,
1402 const size_t blockSize = FIO_LZ4_GetBlockSize_FromBlockId(LZ4F_max64KB);
1403 unsigned long long inFileSize = 0, outFileSize = 0;
1405 LZ4F_preferences_t prefs;
1406 LZ4F_compressionContext_t ctx;
1408 IOJob_t* writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
1410 LZ4F_errorCode_t const errorCode = LZ4F_createCompressionContext(&ctx, LZ4F_VERSION);
1411 if (LZ4F_isError(errorCode))
1412 EXM_THROW(31, "zstd: failed to create lz4 compression context");
1414 memset(&prefs, 0, sizeof(prefs));
1416 assert(blockSize <= ress->readCtx->base.jobBufferSize);
1418 /* autoflush off to mitigate a bug in lz4<=1.9.3 for compression level 12 */
1419 prefs.autoFlush = 0;
1420 prefs.compressionLevel = compressionLevel;
1421 prefs.frameInfo.blockMode = LZ4F_blockLinked;
1422 prefs.frameInfo.blockSizeID = LZ4F_max64KB;
1423 prefs.frameInfo.contentChecksumFlag = (contentChecksum_t)checksumFlag;
1424 #if LZ4_VERSION_NUMBER >= 10600
1425 prefs.frameInfo.contentSize = (srcFileSize==UTIL_FILESIZE_UNKNOWN) ? 0 : srcFileSize;
1427 assert(LZ4F_compressBound(blockSize, &prefs) <= writeJob->bufferSize);
1430 size_t headerSize = LZ4F_compressBegin(ctx, writeJob->buffer, writeJob->bufferSize, &prefs);
1431 if (LZ4F_isError(headerSize))
1432 EXM_THROW(33, "File header generation failed : %s",
1433 LZ4F_getErrorName(headerSize));
1434 writeJob->usedBufferSize = headerSize;
1435 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
1436 outFileSize += headerSize;
1438 /* Read first block */
1439 inFileSize += AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
1442 while (ress->readCtx->srcBufferLoaded) {
1443 size_t inSize = MIN(blockSize, ress->readCtx->srcBufferLoaded);
1444 size_t const outSize = LZ4F_compressUpdate(ctx, writeJob->buffer, writeJob->bufferSize,
1445 ress->readCtx->srcBuffer, inSize, NULL);
1446 if (LZ4F_isError(outSize))
1447 EXM_THROW(35, "zstd: %s: lz4 compression failed : %s",
1448 srcFileName, LZ4F_getErrorName(outSize));
1449 outFileSize += outSize;
1450 if (srcFileSize == UTIL_FILESIZE_UNKNOWN) {
1451 DISPLAYUPDATE_PROGRESS("\rRead : %u MB ==> %.2f%%",
1452 (unsigned)(inFileSize>>20),
1453 (double)outFileSize/(double)inFileSize*100)
1455 DISPLAYUPDATE_PROGRESS("\rRead : %u / %u MB ==> %.2f%%",
1456 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
1457 (double)outFileSize/(double)inFileSize*100);
1461 writeJob->usedBufferSize = outSize;
1462 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
1464 /* Read next block */
1465 AIO_ReadPool_consumeBytes(ress->readCtx, inSize);
1466 inFileSize += AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
1469 /* End of Stream mark */
1470 headerSize = LZ4F_compressEnd(ctx, writeJob->buffer, writeJob->bufferSize, NULL);
1471 if (LZ4F_isError(headerSize))
1472 EXM_THROW(38, "zstd: %s: lz4 end of file generation failed : %s",
1473 srcFileName, LZ4F_getErrorName(headerSize));
1475 writeJob->usedBufferSize = headerSize;
1476 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
1477 outFileSize += headerSize;
1480 *readsize = inFileSize;
1481 LZ4F_freeCompressionContext(ctx);
1482 AIO_WritePool_releaseIoJob(writeJob);
1483 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
1489 static unsigned long long
1490 FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
1491 FIO_prefs_t* const prefs,
1492 const cRess_t* ressPtr,
1493 const char* srcFileName, U64 fileSize,
1494 int compressionLevel, U64* readsize)
1496 cRess_t const ress = *ressPtr;
1497 IOJob_t *writeJob = AIO_WritePool_acquireJob(ressPtr->writeCtx);
1499 U64 compressedfilesize = 0;
1500 ZSTD_EndDirective directive = ZSTD_e_continue;
1501 U64 pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;
1504 ZSTD_frameProgression previous_zfp_update = { 0, 0, 0, 0, 0, 0 };
1505 ZSTD_frameProgression previous_zfp_correction = { 0, 0, 0, 0, 0, 0 };
1506 typedef enum { noChange, slower, faster } speedChange_e;
1507 speedChange_e speedChange = noChange;
1508 unsigned flushWaiting = 0;
1509 unsigned inputPresented = 0;
1510 unsigned inputBlocked = 0;
1511 unsigned lastJobID = 0;
1512 UTIL_time_t lastAdaptTime = UTIL_getTime();
1513 U64 const adaptEveryMicro = REFRESH_RATE;
1515 UTIL_HumanReadableSize_t const file_hrs = UTIL_makeHumanReadableSize(fileSize);
1517 DISPLAYLEVEL(6, "compression using zstd format \n");
1520 if (fileSize != UTIL_FILESIZE_UNKNOWN) {
1521 pledgedSrcSize = fileSize;
1522 CHECK(ZSTD_CCtx_setPledgedSrcSize(ress.cctx, fileSize));
1523 } else if (prefs->streamSrcSize > 0) {
1524 /* unknown source size; use the declared stream size */
1525 pledgedSrcSize = prefs->streamSrcSize;
1526 CHECK( ZSTD_CCtx_setPledgedSrcSize(ress.cctx, prefs->streamSrcSize) );
1531 UTIL_HumanReadableSize_t windowSize;
1532 CHECK(ZSTD_CCtx_getParameter(ress.cctx, ZSTD_c_windowLog, &windowLog));
1533 if (windowLog == 0) {
1534 if (prefs->ldmFlag) {
1535 /* If long mode is set without a window size libzstd will set this size internally */
1536 windowLog = ZSTD_WINDOWLOG_LIMIT_DEFAULT;
1538 const ZSTD_compressionParameters cParams = ZSTD_getCParams(compressionLevel, fileSize, 0);
1539 windowLog = (int)cParams.windowLog;
1542 windowSize = UTIL_makeHumanReadableSize(MAX(1ULL, MIN(1ULL << windowLog, pledgedSrcSize)));
1543 DISPLAYLEVEL(4, "Decompression will require %.*f%s of memory\n", windowSize.precision, windowSize.value, windowSize.suffix);
1547 /* Main compression loop */
1549 size_t stillToFlush;
1550 /* Fill input Buffer */
1551 size_t const inSize = AIO_ReadPool_fillBuffer(ress.readCtx, ZSTD_CStreamInSize());
1552 ZSTD_inBuffer inBuff = setInBuffer( ress.readCtx->srcBuffer, ress.readCtx->srcBufferLoaded, 0 );
1553 DISPLAYLEVEL(6, "fread %u bytes from source \n", (unsigned)inSize);
1554 *readsize += inSize;
1556 if ((ress.readCtx->srcBufferLoaded == 0) || (*readsize == fileSize))
1557 directive = ZSTD_e_end;
1560 while ((inBuff.pos != inBuff.size) /* input buffer must be entirely ingested */
1561 || (directive == ZSTD_e_end && stillToFlush != 0) ) {
1563 size_t const oldIPos = inBuff.pos;
1564 ZSTD_outBuffer outBuff = setOutBuffer( writeJob->buffer, writeJob->bufferSize, 0 );
1565 size_t const toFlushNow = ZSTD_toFlushNow(ress.cctx);
1566 CHECK_V(stillToFlush, ZSTD_compressStream2(ress.cctx, &outBuff, &inBuff, directive));
1567 AIO_ReadPool_consumeBytes(ress.readCtx, inBuff.pos - oldIPos);
1571 if (oldIPos == inBuff.pos) inputBlocked++; /* input buffer is full and can't take any more : input speed is faster than consumption rate */
1572 if (!toFlushNow) flushWaiting = 1;
1574 /* Write compressed stream */
1575 DISPLAYLEVEL(6, "ZSTD_compress_generic(end:%u) => input pos(%u)<=(%u)size ; output generated %u bytes \n",
1576 (unsigned)directive, (unsigned)inBuff.pos, (unsigned)inBuff.size, (unsigned)outBuff.pos);
1578 writeJob->usedBufferSize = outBuff.pos;
1579 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
1580 compressedfilesize += outBuff.pos;
1583 /* adaptive mode : statistics measurement and speed correction */
1584 if (prefs->adaptiveMode && UTIL_clockSpanMicro(lastAdaptTime) > adaptEveryMicro) {
1585 ZSTD_frameProgression const zfp = ZSTD_getFrameProgression(ress.cctx);
1587 lastAdaptTime = UTIL_getTime();
1589 /* check output speed */
1590 if (zfp.currentJobID > 1) { /* only possible if nbWorkers >= 1 */
1592 unsigned long long newlyProduced = zfp.produced - previous_zfp_update.produced;
1593 unsigned long long newlyFlushed = zfp.flushed - previous_zfp_update.flushed;
1594 assert(zfp.produced >= previous_zfp_update.produced);
1595 assert(prefs->nbWorkers >= 1);
1597 /* test if compression is blocked
1598 * either because output is slow and all buffers are full
1599 * or because input is slow and no job can start while waiting for at least one buffer to be filled.
1600 * note : exclude starting part, since currentJobID > 1 */
1601 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)*/
1602 && (zfp.nbActiveWorkers == 0) /* confirmed : no compression ongoing */
1604 DISPLAYLEVEL(6, "all buffers full : compression stopped => slow down \n")
1605 speedChange = slower;
1608 previous_zfp_update = zfp;
1610 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) */
1611 && (flushWaiting == 0) /* flush speed was never slowed by lack of production, so it's operating at max capacity */
1613 DISPLAYLEVEL(6, "compression faster than flush (%llu > %llu), and flushed was never slowed down by lack of production => slow down \n", newlyProduced, newlyFlushed);
1614 speedChange = slower;
1619 /* course correct only if there is at least one new job completed */
1620 if (zfp.currentJobID > lastJobID) {
1621 DISPLAYLEVEL(6, "compression level adaptation check \n")
1623 /* check input speed */
1624 if (zfp.currentJobID > (unsigned)(prefs->nbWorkers+1)) { /* warm up period, to fill all workers */
1625 if (inputBlocked <= 0) {
1626 DISPLAYLEVEL(6, "input is never blocked => input is slower than ingestion \n");
1627 speedChange = slower;
1628 } else if (speedChange == noChange) {
1629 unsigned long long newlyIngested = zfp.ingested - previous_zfp_correction.ingested;
1630 unsigned long long newlyConsumed = zfp.consumed - previous_zfp_correction.consumed;
1631 unsigned long long newlyProduced = zfp.produced - previous_zfp_correction.produced;
1632 unsigned long long newlyFlushed = zfp.flushed - previous_zfp_correction.flushed;
1633 previous_zfp_correction = zfp;
1634 assert(inputPresented > 0);
1635 DISPLAYLEVEL(6, "input blocked %u/%u(%.2f) - ingested:%u vs %u:consumed - flushed:%u vs %u:produced \n",
1636 inputBlocked, inputPresented, (double)inputBlocked/inputPresented*100,
1637 (unsigned)newlyIngested, (unsigned)newlyConsumed,
1638 (unsigned)newlyFlushed, (unsigned)newlyProduced);
1639 if ( (inputBlocked > inputPresented / 8) /* input is waiting often, because input buffers is full : compression or output too slow */
1640 && (newlyFlushed * 33 / 32 > newlyProduced) /* flush everything that is produced */
1641 && (newlyIngested * 33 / 32 > newlyConsumed) /* input speed as fast or faster than compression speed */
1643 DISPLAYLEVEL(6, "recommend faster as in(%llu) >= (%llu)comp(%llu) <= out(%llu) \n",
1644 newlyIngested, newlyConsumed, newlyProduced, newlyFlushed);
1645 speedChange = faster;
1652 if (speedChange == slower) {
1653 DISPLAYLEVEL(6, "slower speed , higher compression \n")
1654 compressionLevel ++;
1655 if (compressionLevel > ZSTD_maxCLevel()) compressionLevel = ZSTD_maxCLevel();
1656 if (compressionLevel > prefs->maxAdaptLevel) compressionLevel = prefs->maxAdaptLevel;
1657 compressionLevel += (compressionLevel == 0); /* skip 0 */
1658 ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
1660 if (speedChange == faster) {
1661 DISPLAYLEVEL(6, "faster speed , lighter compression \n")
1662 compressionLevel --;
1663 if (compressionLevel < prefs->minAdaptLevel) compressionLevel = prefs->minAdaptLevel;
1664 compressionLevel -= (compressionLevel == 0); /* skip 0 */
1665 ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
1667 speedChange = noChange;
1669 lastJobID = zfp.currentJobID;
1670 } /* if (zfp.currentJobID > lastJobID) */
1671 } /* if (prefs->adaptiveMode && UTIL_clockSpanMicro(lastAdaptTime) > adaptEveryMicro) */
1673 /* display notification */
1674 if (SHOULD_DISPLAY_PROGRESS() && READY_FOR_UPDATE()) {
1675 ZSTD_frameProgression const zfp = ZSTD_getFrameProgression(ress.cctx);
1676 double const cShare = (double)zfp.produced / (double)(zfp.consumed + !zfp.consumed/*avoid div0*/) * 100;
1677 UTIL_HumanReadableSize_t const buffered_hrs = UTIL_makeHumanReadableSize(zfp.ingested - zfp.consumed);
1678 UTIL_HumanReadableSize_t const consumed_hrs = UTIL_makeHumanReadableSize(zfp.consumed);
1679 UTIL_HumanReadableSize_t const produced_hrs = UTIL_makeHumanReadableSize(zfp.produced);
1681 DELAY_NEXT_UPDATE();
1683 /* display progress notifications */
1684 DISPLAY_PROGRESS("\r%79s\r", ""); /* Clear out the current displayed line */
1685 if (g_display_prefs.displayLevel >= 3) {
1686 /* Verbose progress update */
1688 "(L%i) Buffered:%5.*f%s - Consumed:%5.*f%s - Compressed:%5.*f%s => %.2f%% ",
1690 buffered_hrs.precision, buffered_hrs.value, buffered_hrs.suffix,
1691 consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix,
1692 produced_hrs.precision, produced_hrs.value, produced_hrs.suffix,
1695 /* Require level 2 or forcibly displayed progress counter for summarized updates */
1696 if (fCtx->nbFilesTotal > 1) {
1697 size_t srcFileNameSize = strlen(srcFileName);
1698 /* Ensure that the string we print is roughly the same size each time */
1699 if (srcFileNameSize > 18) {
1700 const char* truncatedSrcFileName = srcFileName + srcFileNameSize - 15;
1701 DISPLAY_PROGRESS("Compress: %u/%u files. Current: ...%s ",
1702 fCtx->currFileIdx+1, fCtx->nbFilesTotal, truncatedSrcFileName);
1704 DISPLAY_PROGRESS("Compress: %u/%u files. Current: %*s ",
1705 fCtx->currFileIdx+1, fCtx->nbFilesTotal, (int)(18-srcFileNameSize), srcFileName);
1708 DISPLAY_PROGRESS("Read:%6.*f%4s ", consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix);
1709 if (fileSize != UTIL_FILESIZE_UNKNOWN)
1710 DISPLAY_PROGRESS("/%6.*f%4s", file_hrs.precision, file_hrs.value, file_hrs.suffix);
1711 DISPLAY_PROGRESS(" ==> %2.f%%", cShare);
1713 } /* if (SHOULD_DISPLAY_PROGRESS() && READY_FOR_UPDATE()) */
1714 } /* while ((inBuff.pos != inBuff.size) */
1715 } while (directive != ZSTD_e_end);
1717 if (fileSize != UTIL_FILESIZE_UNKNOWN && *readsize != fileSize) {
1718 EXM_THROW(27, "Read error : Incomplete read : %llu / %llu B",
1719 (unsigned long long)*readsize, (unsigned long long)fileSize);
1722 AIO_WritePool_releaseIoJob(writeJob);
1723 AIO_WritePool_sparseWriteEnd(ressPtr->writeCtx);
1725 return compressedfilesize;
1728 /*! FIO_compressFilename_internal() :
1729 * same as FIO_compressFilename_extRess(), with `ress.desFile` already opened.
1730 * @return : 0 : compression completed correctly,
1731 * 1 : missing or pb opening srcFileName
1734 FIO_compressFilename_internal(FIO_ctx_t* const fCtx,
1735 FIO_prefs_t* const prefs,
1737 const char* dstFileName, const char* srcFileName,
1738 int compressionLevel)
1740 UTIL_time_t const timeStart = UTIL_getTime();
1741 clock_t const cpuStart = clock();
1743 U64 compressedfilesize = 0;
1744 U64 const fileSize = UTIL_getFileSize(srcFileName);
1745 DISPLAYLEVEL(5, "%s: %llu bytes \n", srcFileName, (unsigned long long)fileSize);
1747 /* compression format selection */
1748 switch (prefs->compressionType) {
1750 case FIO_zstdCompression:
1751 compressedfilesize = FIO_compressZstdFrame(fCtx, prefs, &ress, srcFileName, fileSize, compressionLevel, &readsize);
1754 case FIO_gzipCompression:
1755 #ifdef ZSTD_GZCOMPRESS
1756 compressedfilesize = FIO_compressGzFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize);
1758 (void)compressionLevel;
1759 EXM_THROW(20, "zstd: %s: file cannot be compressed as gzip (zstd compiled without ZSTD_GZCOMPRESS) -- ignored \n",
1764 case FIO_xzCompression:
1765 case FIO_lzmaCompression:
1766 #ifdef ZSTD_LZMACOMPRESS
1767 compressedfilesize = FIO_compressLzmaFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize, prefs->compressionType==FIO_lzmaCompression);
1769 (void)compressionLevel;
1770 EXM_THROW(20, "zstd: %s: file cannot be compressed as xz/lzma (zstd compiled without ZSTD_LZMACOMPRESS) -- ignored \n",
1775 case FIO_lz4Compression:
1776 #ifdef ZSTD_LZ4COMPRESS
1777 compressedfilesize = FIO_compressLz4Frame(&ress, srcFileName, fileSize, compressionLevel, prefs->checksumFlag, &readsize);
1779 (void)compressionLevel;
1780 EXM_THROW(20, "zstd: %s: file cannot be compressed as lz4 (zstd compiled without ZSTD_LZ4COMPRESS) -- ignored \n",
1787 fCtx->totalBytesInput += (size_t)readsize;
1788 fCtx->totalBytesOutput += (size_t)compressedfilesize;
1789 DISPLAY_PROGRESS("\r%79s\r", "");
1790 if (FIO_shouldDisplayFileSummary(fCtx)) {
1791 UTIL_HumanReadableSize_t hr_isize = UTIL_makeHumanReadableSize((U64) readsize);
1792 UTIL_HumanReadableSize_t hr_osize = UTIL_makeHumanReadableSize((U64) compressedfilesize);
1793 if (readsize == 0) {
1794 DISPLAY_SUMMARY("%-20s : (%6.*f%s => %6.*f%s, %s) \n",
1796 hr_isize.precision, hr_isize.value, hr_isize.suffix,
1797 hr_osize.precision, hr_osize.value, hr_osize.suffix,
1800 DISPLAY_SUMMARY("%-20s :%6.2f%% (%6.*f%s => %6.*f%s, %s) \n",
1802 (double)compressedfilesize / (double)readsize * 100,
1803 hr_isize.precision, hr_isize.value, hr_isize.suffix,
1804 hr_osize.precision, hr_osize.value, hr_osize.suffix,
1809 /* Elapsed Time and CPU Load */
1810 { clock_t const cpuEnd = clock();
1811 double const cpuLoad_s = (double)(cpuEnd - cpuStart) / CLOCKS_PER_SEC;
1812 U64 const timeLength_ns = UTIL_clockSpanNano(timeStart);
1813 double const timeLength_s = (double)timeLength_ns / 1000000000;
1814 double const cpuLoad_pct = (cpuLoad_s / timeLength_s) * 100;
1815 DISPLAYLEVEL(4, "%-20s : Completed in %.2f sec (cpu load : %.0f%%)\n",
1816 srcFileName, timeLength_s, cpuLoad_pct);
1822 /*! FIO_compressFilename_dstFile() :
1823 * open dstFileName, or pass-through if ress.file != NULL,
1824 * then start compression with FIO_compressFilename_internal().
1825 * Manages source removal (--rm) and file permissions transfer.
1826 * note : ress.srcFile must be != NULL,
1827 * so reach this function through FIO_compressFilename_srcFile().
1828 * @return : 0 : compression completed correctly,
1831 static int FIO_compressFilename_dstFile(FIO_ctx_t* const fCtx,
1832 FIO_prefs_t* const prefs,
1834 const char* dstFileName,
1835 const char* srcFileName,
1836 const stat_t* srcFileStat,
1837 int compressionLevel)
1839 int closeDstFile = 0;
1841 int transferStat = 0;
1844 assert(AIO_ReadPool_getFile(ress.readCtx) != NULL);
1845 if (AIO_WritePool_getFile(ress.writeCtx) == NULL) {
1846 int dstFileInitialPermissions = DEFAULT_FILE_PERMISSIONS;
1847 if ( strcmp (srcFileName, stdinmark)
1848 && strcmp (dstFileName, stdoutmark)
1849 && UTIL_isRegularFileStat(srcFileStat) ) {
1851 dstFileInitialPermissions = TEMPORARY_FILE_PERMISSIONS;
1855 DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: opening dst: %s \n", dstFileName);
1856 { FILE *dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFileInitialPermissions);
1857 if (dstFile==NULL) return 1; /* could not open dstFileName */
1858 dstFd = fileno(dstFile);
1859 AIO_WritePool_setFile(ress.writeCtx, dstFile);
1861 /* Must only be added after FIO_openDstFile() succeeds.
1862 * Otherwise we may delete the destination file if it already exists,
1863 * and the user presses Ctrl-C when asked if they wish to overwrite.
1865 addHandler(dstFileName);
1868 result = FIO_compressFilename_internal(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
1874 UTIL_setFDStat(dstFd, dstFileName, srcFileStat);
1877 DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: closing dst: %s \n", dstFileName);
1878 if (AIO_WritePool_closeFile(ress.writeCtx)) { /* error closing file */
1879 DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno));
1884 UTIL_utime(dstFileName, srcFileStat);
1887 if ( (result != 0) /* operation failure */
1888 && strcmp(dstFileName, stdoutmark) /* special case : don't remove() stdout */
1890 FIO_removeFile(dstFileName); /* remove compression artefact; note don't do anything special if remove() fails */
1897 /* List used to compare file extensions (used with --exclude-compressed flag)
1898 * Different from the suffixList and should only apply to ZSTD compress operationResult
1900 static const char *compressedFileExtensions[] = {
2017 /*! FIO_compressFilename_srcFile() :
2018 * @return : 0 : compression completed correctly,
2019 * 1 : missing or pb opening srcFileName
2022 FIO_compressFilename_srcFile(FIO_ctx_t* const fCtx,
2023 FIO_prefs_t* const prefs,
2025 const char* dstFileName,
2026 const char* srcFileName,
2027 int compressionLevel)
2032 U64 fileSize = UTIL_FILESIZE_UNKNOWN;
2033 DISPLAYLEVEL(6, "FIO_compressFilename_srcFile: %s \n", srcFileName);
2035 if (strcmp(srcFileName, stdinmark)) {
2036 if (UTIL_stat(srcFileName, &srcFileStat)) {
2037 /* failure to stat at all is handled during opening */
2039 /* ensure src is not a directory */
2040 if (UTIL_isDirectoryStat(&srcFileStat)) {
2041 DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName);
2045 /* ensure src is not the same as dict (if present) */
2046 if (ress.dictFileName != NULL && UTIL_isSameFileStat(srcFileName, ress.dictFileName, &srcFileStat, &ress.dictFileStat)) {
2047 DISPLAYLEVEL(1, "zstd: cannot use %s as an input file and dictionary \n", srcFileName);
2053 /* Check if "srcFile" is compressed. Only done if --exclude-compressed flag is used
2054 * YES => ZSTD will skip compression of the file and will return 0.
2055 * NO => ZSTD will resume with compress operation.
2057 if (prefs->excludeCompressedFiles == 1 && UTIL_isCompressedFile(srcFileName, compressedFileExtensions)) {
2058 DISPLAYLEVEL(4, "File is already compressed : %s \n", srcFileName);
2062 srcFile = FIO_openSrcFile(prefs, srcFileName, &srcFileStat);
2063 if (srcFile == NULL) return 1; /* srcFile could not be opened */
2065 /* Don't use AsyncIO for small files */
2066 if (strcmp(srcFileName, stdinmark)) /* Stdin doesn't have stats */
2067 fileSize = UTIL_getFileSizeStat(&srcFileStat);
2068 if(fileSize != UTIL_FILESIZE_UNKNOWN && fileSize < ZSTD_BLOCKSIZE_MAX * 3) {
2069 AIO_ReadPool_setAsync(ress.readCtx, 0);
2070 AIO_WritePool_setAsync(ress.writeCtx, 0);
2072 AIO_ReadPool_setAsync(ress.readCtx, 1);
2073 AIO_WritePool_setAsync(ress.writeCtx, 1);
2076 AIO_ReadPool_setFile(ress.readCtx, srcFile);
2077 result = FIO_compressFilename_dstFile(
2079 dstFileName, srcFileName,
2080 &srcFileStat, compressionLevel);
2081 AIO_ReadPool_closeFile(ress.readCtx);
2083 if ( prefs->removeSrcFile /* --rm */
2084 && result == 0 /* success */
2085 && strcmp(srcFileName, stdinmark) /* exception : don't erase stdin */
2087 /* We must clear the handler, since after this point calling it would
2088 * delete both the source and destination files.
2091 if (FIO_removeFile(srcFileName))
2092 EXM_THROW(1, "zstd: %s: %s", srcFileName, strerror(errno));
2098 checked_index(const char* options[], size_t length, size_t index) {
2099 assert(index < length);
2100 /* Necessary to avoid warnings since -O3 will omit the above `assert` */
2102 return options[index];
2105 #define INDEX(options, index) checked_index((options), sizeof(options) / sizeof(char*), (size_t)(index))
2107 void FIO_displayCompressionParameters(const FIO_prefs_t* prefs)
2109 static const char* formatOptions[5] = {ZSTD_EXTENSION, GZ_EXTENSION, XZ_EXTENSION,
2110 LZMA_EXTENSION, LZ4_EXTENSION};
2111 static const char* sparseOptions[3] = {" --no-sparse", "", " --sparse"};
2112 static const char* checkSumOptions[3] = {" --no-check", "", " --check"};
2113 static const char* rowMatchFinderOptions[3] = {"", " --no-row-match-finder", " --row-match-finder"};
2114 static const char* compressLiteralsOptions[3] = {"", " --compress-literals", " --no-compress-literals"};
2116 assert(g_display_prefs.displayLevel >= 4);
2118 DISPLAY("--format=%s", formatOptions[prefs->compressionType]);
2119 DISPLAY("%s", INDEX(sparseOptions, prefs->sparseFileSupport));
2120 DISPLAY("%s", prefs->dictIDFlag ? "" : " --no-dictID");
2121 DISPLAY("%s", INDEX(checkSumOptions, prefs->checksumFlag));
2122 DISPLAY(" --block-size=%d", prefs->blockSize);
2123 if (prefs->adaptiveMode)
2124 DISPLAY(" --adapt=min=%d,max=%d", prefs->minAdaptLevel, prefs->maxAdaptLevel);
2125 DISPLAY("%s", INDEX(rowMatchFinderOptions, prefs->useRowMatchFinder));
2126 DISPLAY("%s", prefs->rsyncable ? " --rsyncable" : "");
2127 if (prefs->streamSrcSize)
2128 DISPLAY(" --stream-size=%u", (unsigned) prefs->streamSrcSize);
2129 if (prefs->srcSizeHint)
2130 DISPLAY(" --size-hint=%d", prefs->srcSizeHint);
2131 if (prefs->targetCBlockSize)
2132 DISPLAY(" --target-compressed-block-size=%u", (unsigned) prefs->targetCBlockSize);
2133 DISPLAY("%s", INDEX(compressLiteralsOptions, prefs->literalCompressionMode));
2134 DISPLAY(" --memory=%u", prefs->memLimit ? prefs->memLimit : 128 MB);
2135 DISPLAY(" --threads=%d", prefs->nbWorkers);
2136 DISPLAY("%s", prefs->excludeCompressedFiles ? " --exclude-compressed" : "");
2137 DISPLAY(" --%scontent-size", prefs->contentSize ? "" : "no-");
2143 int FIO_compressFilename(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs, const char* dstFileName,
2144 const char* srcFileName, const char* dictFileName,
2145 int compressionLevel, ZSTD_compressionParameters comprParams)
2147 cRess_t ress = FIO_createCResources(prefs, dictFileName, UTIL_getFileSize(srcFileName), compressionLevel, comprParams);
2148 int const result = FIO_compressFilename_srcFile(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
2150 #define DISPLAY_LEVEL_DEFAULT 2
2152 FIO_freeCResources(&ress);
2156 /* FIO_determineCompressedName() :
2157 * create a destination filename for compressed srcFileName.
2158 * @return a pointer to it.
2159 * This function never returns an error (it may abort() in case of pb)
2162 FIO_determineCompressedName(const char* srcFileName, const char* outDirName, const char* suffix)
2164 static size_t dfnbCapacity = 0;
2165 static char* dstFileNameBuffer = NULL; /* using static allocation : this function cannot be multi-threaded */
2166 char* outDirFilename = NULL;
2167 size_t sfnSize = strlen(srcFileName);
2168 size_t const srcSuffixLen = strlen(suffix);
2170 if(!strcmp(srcFileName, stdinmark)) {
2175 outDirFilename = FIO_createFilename_fromOutDir(srcFileName, outDirName, srcSuffixLen);
2176 sfnSize = strlen(outDirFilename);
2177 assert(outDirFilename != NULL);
2180 if (dfnbCapacity <= sfnSize+srcSuffixLen+1) {
2181 /* resize buffer for dstName */
2182 free(dstFileNameBuffer);
2183 dfnbCapacity = sfnSize + srcSuffixLen + 30;
2184 dstFileNameBuffer = (char*)malloc(dfnbCapacity);
2185 if (!dstFileNameBuffer) {
2186 EXM_THROW(30, "zstd: %s", strerror(errno));
2189 assert(dstFileNameBuffer != NULL);
2191 if (outDirFilename) {
2192 memcpy(dstFileNameBuffer, outDirFilename, sfnSize);
2193 free(outDirFilename);
2195 memcpy(dstFileNameBuffer, srcFileName, sfnSize);
2197 memcpy(dstFileNameBuffer+sfnSize, suffix, srcSuffixLen+1 /* Include terminating null */);
2198 return dstFileNameBuffer;
2201 static unsigned long long FIO_getLargestFileSize(const char** inFileNames, unsigned nbFiles)
2204 unsigned long long fileSize, maxFileSize = 0;
2205 for (i = 0; i < nbFiles; i++) {
2206 fileSize = UTIL_getFileSize(inFileNames[i]);
2207 maxFileSize = fileSize > maxFileSize ? fileSize : maxFileSize;
2212 /* FIO_compressMultipleFilenames() :
2213 * compress nbFiles files
2214 * into either one destination (outFileName),
2215 * or into one file each (outFileName == NULL, but suffix != NULL),
2216 * or into a destination folder (specified with -O)
2218 int FIO_compressMultipleFilenames(FIO_ctx_t* const fCtx,
2219 FIO_prefs_t* const prefs,
2220 const char** inFileNamesTable,
2221 const char* outMirroredRootDirName,
2222 const char* outDirName,
2223 const char* outFileName, const char* suffix,
2224 const char* dictFileName, int compressionLevel,
2225 ZSTD_compressionParameters comprParams)
2229 cRess_t ress = FIO_createCResources(prefs, dictFileName,
2230 FIO_getLargestFileSize(inFileNamesTable, (unsigned)fCtx->nbFilesTotal),
2231 compressionLevel, comprParams);
2234 assert(outFileName != NULL || suffix != NULL);
2235 if (outFileName != NULL) { /* output into a single destination (stdout typically) */
2237 if (FIO_multiFilesConcatWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
2238 FIO_freeCResources(&ress);
2241 dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
2242 if (dstFile == NULL) { /* could not open outFileName */
2245 AIO_WritePool_setFile(ress.writeCtx, dstFile);
2246 for (; fCtx->currFileIdx < fCtx->nbFilesTotal; ++fCtx->currFileIdx) {
2247 status = FIO_compressFilename_srcFile(fCtx, prefs, ress, outFileName, inFileNamesTable[fCtx->currFileIdx], compressionLevel);
2248 if (!status) fCtx->nbFilesProcessed++;
2251 if (AIO_WritePool_closeFile(ress.writeCtx))
2252 EXM_THROW(29, "Write error (%s) : cannot properly close %s",
2253 strerror(errno), outFileName);
2256 if (outMirroredRootDirName)
2257 UTIL_mirrorSourceFilesDirectories(inFileNamesTable, (unsigned)fCtx->nbFilesTotal, outMirroredRootDirName);
2259 for (; fCtx->currFileIdx < fCtx->nbFilesTotal; ++fCtx->currFileIdx) {
2260 const char* const srcFileName = inFileNamesTable[fCtx->currFileIdx];
2261 const char* dstFileName = NULL;
2262 if (outMirroredRootDirName) {
2263 char* validMirroredDirName = UTIL_createMirroredDestDirName(srcFileName, outMirroredRootDirName);
2264 if (validMirroredDirName) {
2265 dstFileName = FIO_determineCompressedName(srcFileName, validMirroredDirName, suffix);
2266 free(validMirroredDirName);
2268 DISPLAYLEVEL(2, "zstd: --output-dir-mirror cannot compress '%s' into '%s' \n", srcFileName, outMirroredRootDirName);
2273 dstFileName = FIO_determineCompressedName(srcFileName, outDirName, suffix); /* cannot fail */
2275 status = FIO_compressFilename_srcFile(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
2276 if (!status) fCtx->nbFilesProcessed++;
2281 FIO_checkFilenameCollisions(inFileNamesTable , (unsigned)fCtx->nbFilesTotal);
2284 if (FIO_shouldDisplayMultipleFileSummary(fCtx)) {
2285 UTIL_HumanReadableSize_t hr_isize = UTIL_makeHumanReadableSize((U64) fCtx->totalBytesInput);
2286 UTIL_HumanReadableSize_t hr_osize = UTIL_makeHumanReadableSize((U64) fCtx->totalBytesOutput);
2288 DISPLAY_PROGRESS("\r%79s\r", "");
2289 if (fCtx->totalBytesInput == 0) {
2290 DISPLAY_SUMMARY("%3d files compressed : (%6.*f%4s => %6.*f%4s)\n",
2291 fCtx->nbFilesProcessed,
2292 hr_isize.precision, hr_isize.value, hr_isize.suffix,
2293 hr_osize.precision, hr_osize.value, hr_osize.suffix);
2295 DISPLAY_SUMMARY("%3d files compressed : %.2f%% (%6.*f%4s => %6.*f%4s)\n",
2296 fCtx->nbFilesProcessed,
2297 (double)fCtx->totalBytesOutput/((double)fCtx->totalBytesInput)*100,
2298 hr_isize.precision, hr_isize.value, hr_isize.suffix,
2299 hr_osize.precision, hr_osize.value, hr_osize.suffix);
2303 FIO_freeCResources(&ress);
2307 #endif /* #ifndef ZSTD_NOCOMPRESS */
2311 #ifndef ZSTD_NODECOMPRESS
2313 /* **************************************************************************
2315 ***************************************************************************/
2319 WritePoolCtx_t *writeCtx;
2320 ReadPoolCtx_t *readCtx;
2323 static dRess_t FIO_createDResources(FIO_prefs_t* const prefs, const char* dictFileName)
2325 int useMMap = prefs->mmapDict == ZSTD_ps_enable;
2326 int forceNoUseMMap = prefs->mmapDict == ZSTD_ps_disable;
2329 memset(&statbuf, 0, sizeof(statbuf));
2330 memset(&ress, 0, sizeof(ress));
2332 FIO_getDictFileStat(dictFileName, &statbuf);
2334 if (prefs->patchFromMode){
2335 U64 const dictSize = UTIL_getFileSizeStat(&statbuf);
2336 useMMap |= dictSize > prefs->memLimit;
2337 FIO_adjustMemLimitForPatchFromMode(prefs, dictSize, 0 /* just use the dict size */);
2341 ress.dctx = ZSTD_createDStream();
2342 if (ress.dctx==NULL)
2343 EXM_THROW(60, "Error: %s : can't create ZSTD_DStream", strerror(errno));
2344 CHECK( ZSTD_DCtx_setMaxWindowSize(ress.dctx, prefs->memLimit) );
2345 CHECK( ZSTD_DCtx_setParameter(ress.dctx, ZSTD_d_forceIgnoreChecksum, !prefs->checksumFlag));
2349 FIO_dictBufferType_t dictBufferType = (useMMap && !forceNoUseMMap) ? FIO_mmapDict : FIO_mallocDict;
2350 FIO_initDict(&ress.dict, dictFileName, prefs, &statbuf, dictBufferType);
2352 CHECK(ZSTD_DCtx_reset(ress.dctx, ZSTD_reset_session_only) );
2354 if (prefs->patchFromMode){
2355 CHECK(ZSTD_DCtx_refPrefix(ress.dctx, ress.dict.dictBuffer, ress.dict.dictBufferSize));
2357 CHECK(ZSTD_DCtx_loadDictionary_byReference(ress.dctx, ress.dict.dictBuffer, ress.dict.dictBufferSize));
2361 ress.writeCtx = AIO_WritePool_create(prefs, ZSTD_DStreamOutSize());
2362 ress.readCtx = AIO_ReadPool_create(prefs, ZSTD_DStreamInSize());
2366 static void FIO_freeDResources(dRess_t ress)
2368 FIO_freeDict(&(ress.dict));
2369 CHECK( ZSTD_freeDStream(ress.dctx) );
2370 AIO_WritePool_free(ress.writeCtx);
2371 AIO_ReadPool_free(ress.readCtx);
2374 /* FIO_passThrough() : just copy input into output, for compatibility with gzip -df mode
2375 * @return : 0 (no error) */
2376 static int FIO_passThrough(dRess_t *ress)
2378 size_t const blockSize = MIN(MIN(64 KB, ZSTD_DStreamInSize()), ZSTD_DStreamOutSize());
2379 IOJob_t *writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
2380 AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
2382 while(ress->readCtx->srcBufferLoaded) {
2384 writeSize = MIN(blockSize, ress->readCtx->srcBufferLoaded);
2385 assert(writeSize <= writeJob->bufferSize);
2386 memcpy(writeJob->buffer, ress->readCtx->srcBuffer, writeSize);
2387 writeJob->usedBufferSize = writeSize;
2388 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
2389 AIO_ReadPool_consumeBytes(ress->readCtx, writeSize);
2390 AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
2392 assert(ress->readCtx->reachedEof);
2393 AIO_WritePool_releaseIoJob(writeJob);
2394 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
2398 /* FIO_zstdErrorHelp() :
2399 * detailed error message when requested window size is too large */
2401 FIO_zstdErrorHelp(const FIO_prefs_t* const prefs,
2402 const dRess_t* ress,
2404 const char* srcFileName)
2406 ZSTD_frameHeader header;
2408 /* Help message only for one specific error */
2409 if (ZSTD_getErrorCode(err) != ZSTD_error_frameParameter_windowTooLarge)
2412 /* Try to decode the frame header */
2413 err = ZSTD_getFrameHeader(&header, ress->readCtx->srcBuffer, ress->readCtx->srcBufferLoaded);
2415 unsigned long long const windowSize = header.windowSize;
2416 unsigned const windowLog = FIO_highbit64(windowSize) + ((windowSize & (windowSize - 1)) != 0);
2417 assert(prefs->memLimit > 0);
2418 DISPLAYLEVEL(1, "%s : Window size larger than maximum : %llu > %u \n",
2419 srcFileName, windowSize, prefs->memLimit);
2420 if (windowLog <= ZSTD_WINDOWLOG_MAX) {
2421 unsigned const windowMB = (unsigned)((windowSize >> 20) + ((windowSize & ((1 MB) - 1)) != 0));
2422 assert(windowSize < (U64)(1ULL << 52)); /* ensure now overflow for windowMB */
2423 DISPLAYLEVEL(1, "%s : Use --long=%u or --memory=%uMB \n",
2424 srcFileName, windowLog, windowMB);
2427 DISPLAYLEVEL(1, "%s : Window log larger than ZSTD_WINDOWLOG_MAX=%u; not supported \n",
2428 srcFileName, ZSTD_WINDOWLOG_MAX);
2431 /** FIO_decompressFrame() :
2432 * @return : size of decoded zstd frame, or an error code
2434 #define FIO_ERROR_FRAME_DECODING ((unsigned long long)(-2))
2435 static unsigned long long
2436 FIO_decompressZstdFrame(FIO_ctx_t* const fCtx, dRess_t* ress,
2437 const FIO_prefs_t* const prefs,
2438 const char* srcFileName,
2439 U64 alreadyDecoded) /* for multi-frames streams */
2442 IOJob_t *writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
2444 /* display last 20 characters only when not --verbose */
2445 { size_t const srcFileLength = strlen(srcFileName);
2446 if ((srcFileLength>20) && (g_display_prefs.displayLevel<3))
2447 srcFileName += srcFileLength-20;
2450 ZSTD_DCtx_reset(ress->dctx, ZSTD_reset_session_only);
2452 /* Header loading : ensures ZSTD_getFrameHeader() will succeed */
2453 AIO_ReadPool_fillBuffer(ress->readCtx, ZSTD_FRAMEHEADERSIZE_MAX);
2455 /* Main decompression Loop */
2457 ZSTD_inBuffer inBuff = setInBuffer( ress->readCtx->srcBuffer, ress->readCtx->srcBufferLoaded, 0 );
2458 ZSTD_outBuffer outBuff= setOutBuffer( writeJob->buffer, writeJob->bufferSize, 0 );
2459 size_t const readSizeHint = ZSTD_decompressStream(ress->dctx, &outBuff, &inBuff);
2460 UTIL_HumanReadableSize_t const hrs = UTIL_makeHumanReadableSize(alreadyDecoded+frameSize);
2461 if (ZSTD_isError(readSizeHint)) {
2462 DISPLAYLEVEL(1, "%s : Decoding error (36) : %s \n",
2463 srcFileName, ZSTD_getErrorName(readSizeHint));
2464 FIO_zstdErrorHelp(prefs, ress, readSizeHint, srcFileName);
2465 AIO_WritePool_releaseIoJob(writeJob);
2466 return FIO_ERROR_FRAME_DECODING;
2470 writeJob->usedBufferSize = outBuff.pos;
2471 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
2472 frameSize += outBuff.pos;
2473 if (fCtx->nbFilesTotal > 1) {
2474 size_t srcFileNameSize = strlen(srcFileName);
2475 if (srcFileNameSize > 18) {
2476 const char* truncatedSrcFileName = srcFileName + srcFileNameSize - 15;
2477 DISPLAYUPDATE_PROGRESS(
2478 "\rDecompress: %2u/%2u files. Current: ...%s : %.*f%s... ",
2479 fCtx->currFileIdx+1, fCtx->nbFilesTotal, truncatedSrcFileName, hrs.precision, hrs.value, hrs.suffix);
2481 DISPLAYUPDATE_PROGRESS("\rDecompress: %2u/%2u files. Current: %s : %.*f%s... ",
2482 fCtx->currFileIdx+1, fCtx->nbFilesTotal, srcFileName, hrs.precision, hrs.value, hrs.suffix);
2485 DISPLAYUPDATE_PROGRESS("\r%-20.20s : %.*f%s... ",
2486 srcFileName, hrs.precision, hrs.value, hrs.suffix);
2489 AIO_ReadPool_consumeBytes(ress->readCtx, inBuff.pos);
2491 if (readSizeHint == 0) break; /* end of frame */
2493 /* Fill input buffer */
2494 { size_t const toDecode = MIN(readSizeHint, ZSTD_DStreamInSize()); /* support large skippable frames */
2495 if (ress->readCtx->srcBufferLoaded < toDecode) {
2496 size_t const readSize = AIO_ReadPool_fillBuffer(ress->readCtx, toDecode);
2498 DISPLAYLEVEL(1, "%s : Read error (39) : premature end \n",
2500 AIO_WritePool_releaseIoJob(writeJob);
2501 return FIO_ERROR_FRAME_DECODING;
2505 AIO_WritePool_releaseIoJob(writeJob);
2506 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
2512 #ifdef ZSTD_GZDECOMPRESS
2513 static unsigned long long
2514 FIO_decompressGzFrame(dRess_t* ress, const char* srcFileName)
2516 unsigned long long outFileSize = 0;
2518 int flush = Z_NO_FLUSH;
2519 int decodingError = 0;
2520 IOJob_t *writeJob = NULL;
2522 strm.zalloc = Z_NULL;
2523 strm.zfree = Z_NULL;
2524 strm.opaque = Z_NULL;
2527 /* see https://www.zlib.net/manual.html */
2528 if (inflateInit2(&strm, 15 /* maxWindowLogSize */ + 16 /* gzip only */) != Z_OK)
2529 return FIO_ERROR_FRAME_DECODING;
2531 writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
2532 strm.next_out = (Bytef*)writeJob->buffer;
2533 strm.avail_out = (uInt)writeJob->bufferSize;
2534 strm.avail_in = (uInt)ress->readCtx->srcBufferLoaded;
2535 strm.next_in = (z_const unsigned char*)ress->readCtx->srcBuffer;
2539 if (strm.avail_in == 0) {
2540 AIO_ReadPool_consumeAndRefill(ress->readCtx);
2541 if (ress->readCtx->srcBufferLoaded == 0) flush = Z_FINISH;
2542 strm.next_in = (z_const unsigned char*)ress->readCtx->srcBuffer;
2543 strm.avail_in = (uInt)ress->readCtx->srcBufferLoaded;
2545 ret = inflate(&strm, flush);
2546 if (ret == Z_BUF_ERROR) {
2547 DISPLAYLEVEL(1, "zstd: %s: premature gz end \n", srcFileName);
2548 decodingError = 1; break;
2550 if (ret != Z_OK && ret != Z_STREAM_END) {
2551 DISPLAYLEVEL(1, "zstd: %s: inflate error %d \n", srcFileName, ret);
2552 decodingError = 1; break;
2554 { size_t const decompBytes = writeJob->bufferSize - strm.avail_out;
2556 writeJob->usedBufferSize = decompBytes;
2557 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
2558 outFileSize += decompBytes;
2559 strm.next_out = (Bytef*)writeJob->buffer;
2560 strm.avail_out = (uInt)writeJob->bufferSize;
2563 if (ret == Z_STREAM_END) break;
2566 AIO_ReadPool_consumeBytes(ress->readCtx, ress->readCtx->srcBufferLoaded - strm.avail_in);
2568 if ( (inflateEnd(&strm) != Z_OK) /* release resources ; error detected */
2569 && (decodingError==0) ) {
2570 DISPLAYLEVEL(1, "zstd: %s: inflateEnd error \n", srcFileName);
2573 AIO_WritePool_releaseIoJob(writeJob);
2574 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
2575 return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize;
2579 #ifdef ZSTD_LZMADECOMPRESS
2580 static unsigned long long
2581 FIO_decompressLzmaFrame(dRess_t* ress,
2582 const char* srcFileName, int plain_lzma)
2584 unsigned long long outFileSize = 0;
2585 lzma_stream strm = LZMA_STREAM_INIT;
2586 lzma_action action = LZMA_RUN;
2588 int decodingError = 0;
2589 IOJob_t *writeJob = NULL;
2594 initRet = lzma_alone_decoder(&strm, UINT64_MAX); /* LZMA */
2596 initRet = lzma_stream_decoder(&strm, UINT64_MAX, 0); /* XZ */
2599 if (initRet != LZMA_OK) {
2600 DISPLAYLEVEL(1, "zstd: %s: %s error %d \n",
2601 plain_lzma ? "lzma_alone_decoder" : "lzma_stream_decoder",
2602 srcFileName, initRet);
2603 return FIO_ERROR_FRAME_DECODING;
2606 writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
2607 strm.next_out = (BYTE*)writeJob->buffer;
2608 strm.avail_out = writeJob->bufferSize;
2609 strm.next_in = (BYTE const*)ress->readCtx->srcBuffer;
2610 strm.avail_in = ress->readCtx->srcBufferLoaded;
2614 if (strm.avail_in == 0) {
2615 AIO_ReadPool_consumeAndRefill(ress->readCtx);
2616 if (ress->readCtx->srcBufferLoaded == 0) action = LZMA_FINISH;
2617 strm.next_in = (BYTE const*)ress->readCtx->srcBuffer;
2618 strm.avail_in = ress->readCtx->srcBufferLoaded;
2620 ret = lzma_code(&strm, action);
2622 if (ret == LZMA_BUF_ERROR) {
2623 DISPLAYLEVEL(1, "zstd: %s: premature lzma end \n", srcFileName);
2624 decodingError = 1; break;
2626 if (ret != LZMA_OK && ret != LZMA_STREAM_END) {
2627 DISPLAYLEVEL(1, "zstd: %s: lzma_code decoding error %d \n",
2629 decodingError = 1; break;
2631 { size_t const decompBytes = writeJob->bufferSize - strm.avail_out;
2633 writeJob->usedBufferSize = decompBytes;
2634 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
2635 outFileSize += decompBytes;
2636 strm.next_out = (BYTE*)writeJob->buffer;
2637 strm.avail_out = writeJob->bufferSize;
2639 if (ret == LZMA_STREAM_END) break;
2642 AIO_ReadPool_consumeBytes(ress->readCtx, ress->readCtx->srcBufferLoaded - strm.avail_in);
2644 AIO_WritePool_releaseIoJob(writeJob);
2645 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
2646 return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize;
2650 #ifdef ZSTD_LZ4DECOMPRESS
2651 static unsigned long long
2652 FIO_decompressLz4Frame(dRess_t* ress, const char* srcFileName)
2654 unsigned long long filesize = 0;
2655 LZ4F_errorCode_t nextToLoad = 4;
2656 LZ4F_decompressionContext_t dCtx;
2657 LZ4F_errorCode_t const errorCode = LZ4F_createDecompressionContext(&dCtx, LZ4F_VERSION);
2658 int decodingError = 0;
2659 IOJob_t *writeJob = NULL;
2661 if (LZ4F_isError(errorCode)) {
2662 DISPLAYLEVEL(1, "zstd: failed to create lz4 decompression context \n");
2663 return FIO_ERROR_FRAME_DECODING;
2666 writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
2669 for (;nextToLoad;) {
2671 size_t decodedBytes = writeJob->bufferSize;
2672 int fullBufferDecoded = 0;
2675 AIO_ReadPool_fillBuffer(ress->readCtx, nextToLoad);
2676 if(!ress->readCtx->srcBufferLoaded) break; /* reached end of file */
2678 while ((pos < ress->readCtx->srcBufferLoaded) || fullBufferDecoded) { /* still to read, or still to flush */
2679 /* Decode Input (at least partially) */
2680 size_t remaining = ress->readCtx->srcBufferLoaded - pos;
2681 decodedBytes = writeJob->bufferSize;
2682 nextToLoad = LZ4F_decompress(dCtx, writeJob->buffer, &decodedBytes, (char*)(ress->readCtx->srcBuffer)+pos,
2684 if (LZ4F_isError(nextToLoad)) {
2685 DISPLAYLEVEL(1, "zstd: %s: lz4 decompression error : %s \n",
2686 srcFileName, LZ4F_getErrorName(nextToLoad));
2687 decodingError = 1; nextToLoad = 0; break;
2690 assert(pos <= ress->readCtx->srcBufferLoaded);
2691 fullBufferDecoded = decodedBytes == writeJob->bufferSize;
2695 UTIL_HumanReadableSize_t hrs;
2696 writeJob->usedBufferSize = decodedBytes;
2697 AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
2698 filesize += decodedBytes;
2699 hrs = UTIL_makeHumanReadableSize(filesize);
2700 DISPLAYUPDATE_PROGRESS("\rDecompressed : %.*f%s ", hrs.precision, hrs.value, hrs.suffix);
2703 if (!nextToLoad) break;
2705 AIO_ReadPool_consumeBytes(ress->readCtx, pos);
2707 if (nextToLoad!=0) {
2708 DISPLAYLEVEL(1, "zstd: %s: unfinished lz4 stream \n", srcFileName);
2712 LZ4F_freeDecompressionContext(dCtx);
2713 AIO_WritePool_releaseIoJob(writeJob);
2714 AIO_WritePool_sparseWriteEnd(ress->writeCtx);
2716 return decodingError ? FIO_ERROR_FRAME_DECODING : filesize;
2722 /** FIO_decompressFrames() :
2723 * Find and decode frames inside srcFile
2724 * srcFile presumed opened and valid
2728 static int FIO_decompressFrames(FIO_ctx_t* const fCtx,
2729 dRess_t ress, const FIO_prefs_t* const prefs,
2730 const char* dstFileName, const char* srcFileName)
2732 unsigned readSomething = 0;
2733 unsigned long long filesize = 0;
2734 int passThrough = prefs->passThrough;
2736 if (passThrough == -1) {
2737 /* If pass-through mode is not explicitly enabled or disabled,
2738 * default to the legacy behavior of enabling it if we are writing
2739 * to stdout with the overwrite flag enabled.
2741 passThrough = prefs->overwrite && !strcmp(dstFileName, stdoutmark);
2743 assert(passThrough == 0 || passThrough == 1);
2745 /* for each frame */
2747 /* check magic number -> version */
2748 size_t const toRead = 4;
2750 AIO_ReadPool_fillBuffer(ress.readCtx, toRead);
2751 buf = (const BYTE*)ress.readCtx->srcBuffer;
2752 if (ress.readCtx->srcBufferLoaded==0) {
2753 if (readSomething==0) { /* srcFile is empty (which is invalid) */
2754 DISPLAYLEVEL(1, "zstd: %s: unexpected end of file \n", srcFileName);
2756 } /* else, just reached frame boundary */
2757 break; /* no more input */
2759 readSomething = 1; /* there is at least 1 byte in srcFile */
2760 if (ress.readCtx->srcBufferLoaded < toRead) { /* not enough input to check magic number */
2762 return FIO_passThrough(&ress);
2764 DISPLAYLEVEL(1, "zstd: %s: unknown header \n", srcFileName);
2767 if (ZSTD_isFrame(buf, ress.readCtx->srcBufferLoaded)) {
2768 unsigned long long const frameSize = FIO_decompressZstdFrame(fCtx, &ress, prefs, srcFileName, filesize);
2769 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2770 filesize += frameSize;
2771 } else if (buf[0] == 31 && buf[1] == 139) { /* gz magic number */
2772 #ifdef ZSTD_GZDECOMPRESS
2773 unsigned long long const frameSize = FIO_decompressGzFrame(&ress, srcFileName);
2774 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2775 filesize += frameSize;
2777 DISPLAYLEVEL(1, "zstd: %s: gzip file cannot be uncompressed (zstd compiled without HAVE_ZLIB) -- ignored \n", srcFileName);
2780 } else if ((buf[0] == 0xFD && buf[1] == 0x37) /* xz magic number */
2781 || (buf[0] == 0x5D && buf[1] == 0x00)) { /* lzma header (no magic number) */
2782 #ifdef ZSTD_LZMADECOMPRESS
2783 unsigned long long const frameSize = FIO_decompressLzmaFrame(&ress, srcFileName, buf[0] != 0xFD);
2784 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2785 filesize += frameSize;
2787 DISPLAYLEVEL(1, "zstd: %s: xz/lzma file cannot be uncompressed (zstd compiled without HAVE_LZMA) -- ignored \n", srcFileName);
2790 } else if (MEM_readLE32(buf) == LZ4_MAGICNUMBER) {
2791 #ifdef ZSTD_LZ4DECOMPRESS
2792 unsigned long long const frameSize = FIO_decompressLz4Frame(&ress, srcFileName);
2793 if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2794 filesize += frameSize;
2796 DISPLAYLEVEL(1, "zstd: %s: lz4 file cannot be uncompressed (zstd compiled without HAVE_LZ4) -- ignored \n", srcFileName);
2799 } else if (passThrough) {
2800 return FIO_passThrough(&ress);
2802 DISPLAYLEVEL(1, "zstd: %s: unsupported format \n", srcFileName);
2804 } } /* for each frame */
2807 fCtx->totalBytesOutput += (size_t)filesize;
2808 DISPLAY_PROGRESS("\r%79s\r", "");
2809 if (FIO_shouldDisplayFileSummary(fCtx))
2810 DISPLAY_SUMMARY("%-20s: %llu bytes \n", srcFileName, filesize);
2815 /** FIO_decompressDstFile() :
2816 open `dstFileName`, or pass-through if writeCtx's file is already != 0,
2817 then start decompression process (FIO_decompressFrames()).
2819 1 : operation aborted
2821 static int FIO_decompressDstFile(FIO_ctx_t* const fCtx,
2822 FIO_prefs_t* const prefs,
2824 const char* dstFileName,
2825 const char* srcFileName,
2826 const stat_t* srcFileStat)
2829 int releaseDstFile = 0;
2830 int transferStat = 0;
2833 if ((AIO_WritePool_getFile(ress.writeCtx) == NULL) && (prefs->testMode == 0)) {
2835 int dstFilePermissions = DEFAULT_FILE_PERMISSIONS;
2836 if ( strcmp(srcFileName, stdinmark) /* special case : don't transfer permissions from stdin */
2837 && strcmp(dstFileName, stdoutmark)
2838 && UTIL_isRegularFileStat(srcFileStat) ) {
2840 dstFilePermissions = TEMPORARY_FILE_PERMISSIONS;
2845 dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFilePermissions);
2846 if (dstFile==NULL) return 1;
2847 dstFd = fileno(dstFile);
2848 AIO_WritePool_setFile(ress.writeCtx, dstFile);
2850 /* Must only be added after FIO_openDstFile() succeeds.
2851 * Otherwise we may delete the destination file if it already exists,
2852 * and the user presses Ctrl-C when asked if they wish to overwrite.
2854 addHandler(dstFileName);
2857 result = FIO_decompressFrames(fCtx, ress, prefs, dstFileName, srcFileName);
2859 if (releaseDstFile) {
2863 UTIL_setFDStat(dstFd, dstFileName, srcFileStat);
2866 if (AIO_WritePool_closeFile(ress.writeCtx)) {
2867 DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno));
2872 UTIL_utime(dstFileName, srcFileStat);
2875 if ( (result != 0) /* operation failure */
2876 && strcmp(dstFileName, stdoutmark) /* special case : don't remove() stdout */
2878 FIO_removeFile(dstFileName); /* remove decompression artefact; note: don't do anything special if remove() fails */
2886 /** FIO_decompressSrcFile() :
2887 Open `srcFileName`, transfer control to decompressDstFile()
2891 static int FIO_decompressSrcFile(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs, dRess_t ress, const char* dstFileName, const char* srcFileName)
2896 U64 fileSize = UTIL_FILESIZE_UNKNOWN;
2898 if (UTIL_isDirectory(srcFileName)) {
2899 DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName);
2903 srcFile = FIO_openSrcFile(prefs, srcFileName, &srcFileStat);
2904 if (srcFile==NULL) return 1;
2906 /* Don't use AsyncIO for small files */
2907 if (strcmp(srcFileName, stdinmark)) /* Stdin doesn't have stats */
2908 fileSize = UTIL_getFileSizeStat(&srcFileStat);
2909 if(fileSize != UTIL_FILESIZE_UNKNOWN && fileSize < ZSTD_BLOCKSIZE_MAX * 3) {
2910 AIO_ReadPool_setAsync(ress.readCtx, 0);
2911 AIO_WritePool_setAsync(ress.writeCtx, 0);
2913 AIO_ReadPool_setAsync(ress.readCtx, 1);
2914 AIO_WritePool_setAsync(ress.writeCtx, 1);
2917 AIO_ReadPool_setFile(ress.readCtx, srcFile);
2919 result = FIO_decompressDstFile(fCtx, prefs, ress, dstFileName, srcFileName, &srcFileStat);
2921 AIO_ReadPool_setFile(ress.readCtx, NULL);
2924 if (fclose(srcFile)) {
2925 DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno)); /* error should not happen */
2928 if ( prefs->removeSrcFile /* --rm */
2929 && (result==0) /* decompression successful */
2930 && strcmp(srcFileName, stdinmark) ) /* not stdin */ {
2931 /* We must clear the handler, since after this point calling it would
2932 * delete both the source and destination files.
2935 if (FIO_removeFile(srcFileName)) {
2936 /* failed to remove src file */
2937 DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno));
2945 int FIO_decompressFilename(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs,
2946 const char* dstFileName, const char* srcFileName,
2947 const char* dictFileName)
2949 dRess_t const ress = FIO_createDResources(prefs, dictFileName);
2951 int const decodingError = FIO_decompressSrcFile(fCtx, prefs, ress, dstFileName, srcFileName);
2955 FIO_freeDResources(ress);
2956 return decodingError;
2959 static const char *suffixList[] = {
2962 #ifndef ZSTD_NODECOMPRESS
2965 #ifdef ZSTD_GZDECOMPRESS
2969 #ifdef ZSTD_LZMADECOMPRESS
2974 #ifdef ZSTD_LZ4DECOMPRESS
2981 static const char *suffixListStr =
2982 ZSTD_EXTENSION "/" TZSTD_EXTENSION
2983 #ifdef ZSTD_GZDECOMPRESS
2984 "/" GZ_EXTENSION "/" TGZ_EXTENSION
2986 #ifdef ZSTD_LZMADECOMPRESS
2987 "/" LZMA_EXTENSION "/" XZ_EXTENSION "/" TXZ_EXTENSION
2989 #ifdef ZSTD_LZ4DECOMPRESS
2990 "/" LZ4_EXTENSION "/" TLZ4_EXTENSION
2994 /* FIO_determineDstName() :
2995 * create a destination filename from a srcFileName.
2996 * @return a pointer to it.
2997 * @return == NULL if there is an error */
2999 FIO_determineDstName(const char* srcFileName, const char* outDirName)
3001 static size_t dfnbCapacity = 0;
3002 static char* dstFileNameBuffer = NULL; /* using static allocation : this function cannot be multi-threaded */
3003 size_t dstFileNameEndPos;
3004 char* outDirFilename = NULL;
3005 const char* dstSuffix = "";
3006 size_t dstSuffixLen = 0;
3008 size_t sfnSize = strlen(srcFileName);
3010 size_t srcSuffixLen;
3011 const char* const srcSuffix = strrchr(srcFileName, '.');
3013 if(!strcmp(srcFileName, stdinmark)) {
3017 if (srcSuffix == NULL) {
3019 "zstd: %s: unknown suffix (%s expected). "
3020 "Can't derive the output file name. "
3021 "Specify it with -o dstFileName. Ignoring.\n",
3022 srcFileName, suffixListStr);
3025 srcSuffixLen = strlen(srcSuffix);
3028 const char** matchedSuffixPtr;
3029 for (matchedSuffixPtr = suffixList; *matchedSuffixPtr != NULL; matchedSuffixPtr++) {
3030 if (!strcmp(*matchedSuffixPtr, srcSuffix)) {
3035 /* check suffix is authorized */
3036 if (sfnSize <= srcSuffixLen || *matchedSuffixPtr == NULL) {
3038 "zstd: %s: unknown suffix (%s expected). "
3039 "Can't derive the output file name. "
3040 "Specify it with -o dstFileName. Ignoring.\n",
3041 srcFileName, suffixListStr);
3045 if ((*matchedSuffixPtr)[1] == 't') {
3047 dstSuffixLen = strlen(dstSuffix);
3052 outDirFilename = FIO_createFilename_fromOutDir(srcFileName, outDirName, 0);
3053 sfnSize = strlen(outDirFilename);
3054 assert(outDirFilename != NULL);
3057 if (dfnbCapacity+srcSuffixLen <= sfnSize+1+dstSuffixLen) {
3058 /* allocate enough space to write dstFilename into it */
3059 free(dstFileNameBuffer);
3060 dfnbCapacity = sfnSize + 20;
3061 dstFileNameBuffer = (char*)malloc(dfnbCapacity);
3062 if (dstFileNameBuffer==NULL)
3063 EXM_THROW(74, "%s : not enough memory for dstFileName",
3067 /* return dst name == src name truncated from suffix */
3068 assert(dstFileNameBuffer != NULL);
3069 dstFileNameEndPos = sfnSize - srcSuffixLen;
3070 if (outDirFilename) {
3071 memcpy(dstFileNameBuffer, outDirFilename, dstFileNameEndPos);
3072 free(outDirFilename);
3074 memcpy(dstFileNameBuffer, srcFileName, dstFileNameEndPos);
3077 /* The short tar extensions tzst, tgz, txz and tlz4 files should have "tar"
3078 * extension on decompression. Also writes terminating null. */
3079 strcpy(dstFileNameBuffer + dstFileNameEndPos, dstSuffix);
3080 return dstFileNameBuffer;
3082 /* note : dstFileNameBuffer memory is not going to be free */
3086 FIO_decompressMultipleFilenames(FIO_ctx_t* const fCtx,
3087 FIO_prefs_t* const prefs,
3088 const char** srcNamesTable,
3089 const char* outMirroredRootDirName,
3090 const char* outDirName, const char* outFileName,
3091 const char* dictFileName)
3095 dRess_t ress = FIO_createDResources(prefs, dictFileName);
3098 if (FIO_multiFilesConcatWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
3099 FIO_freeDResources(ress);
3102 if (!prefs->testMode) {
3103 FILE* dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
3104 if (dstFile == 0) EXM_THROW(19, "cannot open %s", outFileName);
3105 AIO_WritePool_setFile(ress.writeCtx, dstFile);
3107 for (; fCtx->currFileIdx < fCtx->nbFilesTotal; fCtx->currFileIdx++) {
3108 status = FIO_decompressSrcFile(fCtx, prefs, ress, outFileName, srcNamesTable[fCtx->currFileIdx]);
3109 if (!status) fCtx->nbFilesProcessed++;
3112 if ((!prefs->testMode) && (AIO_WritePool_closeFile(ress.writeCtx)))
3113 EXM_THROW(72, "Write error : %s : cannot properly close output file",
3116 if (outMirroredRootDirName)
3117 UTIL_mirrorSourceFilesDirectories(srcNamesTable, (unsigned)fCtx->nbFilesTotal, outMirroredRootDirName);
3119 for (; fCtx->currFileIdx < fCtx->nbFilesTotal; fCtx->currFileIdx++) { /* create dstFileName */
3120 const char* const srcFileName = srcNamesTable[fCtx->currFileIdx];
3121 const char* dstFileName = NULL;
3122 if (outMirroredRootDirName) {
3123 char* validMirroredDirName = UTIL_createMirroredDestDirName(srcFileName, outMirroredRootDirName);
3124 if (validMirroredDirName) {
3125 dstFileName = FIO_determineDstName(srcFileName, validMirroredDirName);
3126 free(validMirroredDirName);
3128 DISPLAYLEVEL(2, "zstd: --output-dir-mirror cannot decompress '%s' into '%s'\n", srcFileName, outMirroredRootDirName);
3131 dstFileName = FIO_determineDstName(srcFileName, outDirName);
3133 if (dstFileName == NULL) { error=1; continue; }
3134 status = FIO_decompressSrcFile(fCtx, prefs, ress, dstFileName, srcFileName);
3135 if (!status) fCtx->nbFilesProcessed++;
3139 FIO_checkFilenameCollisions(srcNamesTable , (unsigned)fCtx->nbFilesTotal);
3142 if (FIO_shouldDisplayMultipleFileSummary(fCtx)) {
3143 DISPLAY_PROGRESS("\r%79s\r", "");
3144 DISPLAY_SUMMARY("%d files decompressed : %6llu bytes total \n",
3145 fCtx->nbFilesProcessed, (unsigned long long)fCtx->totalBytesOutput);
3148 FIO_freeDResources(ress);
3152 /* **************************************************************************
3153 * .zst file info (--list command)
3154 ***************************************************************************/
3157 U64 decompressedSize;
3160 int numActualFrames;
3161 int numSkippableFrames;
3162 int decompUnavailable;
3174 info_truncated_input=4
3177 #define ERROR_IF(c,n,...) { \
3179 DISPLAYLEVEL(1, __VA_ARGS__); \
3180 DISPLAYLEVEL(1, " \n"); \
3186 FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile)
3188 /* begin analyzing frame */
3190 BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
3191 size_t const numBytesRead = fread(headerBuffer, 1, sizeof(headerBuffer), srcFile);
3192 if (numBytesRead < ZSTD_FRAMEHEADERSIZE_MIN(ZSTD_f_zstd1)) {
3194 && (numBytesRead == 0)
3195 && (info->compressedSize > 0)
3196 && (info->compressedSize != UTIL_FILESIZE_UNKNOWN) ) {
3197 unsigned long long file_position = (unsigned long long) LONG_TELL(srcFile);
3198 unsigned long long file_size = (unsigned long long) info->compressedSize;
3199 ERROR_IF(file_position != file_size, info_truncated_input,
3200 "Error: seeked to position %llu, which is beyond file size of %llu\n",
3203 break; /* correct end of file => success */
3205 ERROR_IF(feof(srcFile), info_not_zstd, "Error: reached end of file with incomplete frame");
3206 ERROR_IF(1, info_frame_error, "Error: did not reach end of file but ran out of frames");
3208 { U32 const magicNumber = MEM_readLE32(headerBuffer);
3209 /* Zstandard frame */
3210 if (magicNumber == ZSTD_MAGICNUMBER) {
3211 ZSTD_frameHeader header;
3212 U64 const frameContentSize = ZSTD_getFrameContentSize(headerBuffer, numBytesRead);
3213 if ( frameContentSize == ZSTD_CONTENTSIZE_ERROR
3214 || frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN ) {
3215 info->decompUnavailable = 1;
3217 info->decompressedSize += frameContentSize;
3219 ERROR_IF(ZSTD_getFrameHeader(&header, headerBuffer, numBytesRead) != 0,
3220 info_frame_error, "Error: could not decode frame header");
3221 if (info->dictID != 0 && info->dictID != header.dictID) {
3222 DISPLAY("WARNING: File contains multiple frames with different dictionary IDs. Showing dictID 0 instead");
3225 info->dictID = header.dictID;
3227 info->windowSize = header.windowSize;
3228 /* move to the end of the frame header */
3229 { size_t const headerSize = ZSTD_frameHeaderSize(headerBuffer, numBytesRead);
3230 ERROR_IF(ZSTD_isError(headerSize), info_frame_error, "Error: could not determine frame header size");
3231 ERROR_IF(fseek(srcFile, ((long)headerSize)-((long)numBytesRead), SEEK_CUR) != 0,
3232 info_frame_error, "Error: could not move to end of frame header");
3235 /* skip all blocks in the frame */
3236 { int lastBlock = 0;
3238 BYTE blockHeaderBuffer[3];
3239 ERROR_IF(fread(blockHeaderBuffer, 1, 3, srcFile) != 3,
3240 info_frame_error, "Error while reading block header");
3241 { U32 const blockHeader = MEM_readLE24(blockHeaderBuffer);
3242 U32 const blockTypeID = (blockHeader >> 1) & 3;
3243 U32 const isRLE = (blockTypeID == 1);
3244 U32 const isWrongBlock = (blockTypeID == 3);
3245 long const blockSize = isRLE ? 1 : (long)(blockHeader >> 3);
3246 ERROR_IF(isWrongBlock, info_frame_error, "Error: unsupported block type");
3247 lastBlock = blockHeader & 1;
3248 ERROR_IF(fseek(srcFile, blockSize, SEEK_CUR) != 0,
3249 info_frame_error, "Error: could not skip to end of block");
3251 } while (lastBlock != 1);
3254 /* check if checksum is used */
3255 { BYTE const frameHeaderDescriptor = headerBuffer[4];
3256 int const contentChecksumFlag = (frameHeaderDescriptor & (1 << 2)) >> 2;
3257 if (contentChecksumFlag) {
3258 info->usesCheck = 1;
3259 ERROR_IF(fread(info->checksum, 1, 4, srcFile) != 4,
3260 info_frame_error, "Error: could not read checksum");
3262 info->numActualFrames++;
3264 /* Skippable frame */
3265 else if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
3266 U32 const frameSize = MEM_readLE32(headerBuffer + 4);
3267 long const seek = (long)(8 + frameSize - numBytesRead);
3268 ERROR_IF(LONG_SEEK(srcFile, seek, SEEK_CUR) != 0,
3269 info_frame_error, "Error: could not find end of skippable frame");
3270 info->numSkippableFrames++;
3272 /* unknown content */
3274 return info_not_zstd;
3276 } /* magic number analysis */
3277 } /* end analyzing frames */
3278 return info_success;
3283 getFileInfo_fileConfirmed(fileInfo_t* info, const char* inFileName)
3287 FILE* const srcFile = FIO_openSrcFile(NULL, inFileName, &srcFileStat);
3288 ERROR_IF(srcFile == NULL, info_file_error, "Error: could not open source file %s", inFileName);
3290 info->compressedSize = UTIL_getFileSizeStat(&srcFileStat);
3291 status = FIO_analyzeFrames(info, srcFile);
3300 * Reads information from file, stores in *info
3301 * @return : InfoError status
3304 getFileInfo(fileInfo_t* info, const char* srcFileName)
3306 ERROR_IF(!UTIL_isRegularFile(srcFileName),
3307 info_file_error, "Error : %s is not a file", srcFileName);
3308 return getFileInfo_fileConfirmed(info, srcFileName);
3313 displayInfo(const char* inFileName, const fileInfo_t* info, int displayLevel)
3315 UTIL_HumanReadableSize_t const window_hrs = UTIL_makeHumanReadableSize(info->windowSize);
3316 UTIL_HumanReadableSize_t const compressed_hrs = UTIL_makeHumanReadableSize(info->compressedSize);
3317 UTIL_HumanReadableSize_t const decompressed_hrs = UTIL_makeHumanReadableSize(info->decompressedSize);
3318 double const ratio = (info->compressedSize == 0) ? 0 : ((double)info->decompressedSize)/(double)info->compressedSize;
3319 const char* const checkString = (info->usesCheck ? "XXH64" : "None");
3320 if (displayLevel <= 2) {
3321 if (!info->decompUnavailable) {
3322 DISPLAYOUT("%6d %5d %6.*f%4s %8.*f%4s %5.3f %5s %s\n",
3323 info->numSkippableFrames + info->numActualFrames,
3324 info->numSkippableFrames,
3325 compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3326 decompressed_hrs.precision, decompressed_hrs.value, decompressed_hrs.suffix,
3327 ratio, checkString, inFileName);
3329 DISPLAYOUT("%6d %5d %6.*f%4s %5s %s\n",
3330 info->numSkippableFrames + info->numActualFrames,
3331 info->numSkippableFrames,
3332 compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3333 checkString, inFileName);
3336 DISPLAYOUT("%s \n", inFileName);
3337 DISPLAYOUT("# Zstandard Frames: %d\n", info->numActualFrames);
3338 if (info->numSkippableFrames)
3339 DISPLAYOUT("# Skippable Frames: %d\n", info->numSkippableFrames);
3340 DISPLAYOUT("DictID: %u\n", info->dictID);
3341 DISPLAYOUT("Window Size: %.*f%s (%llu B)\n",
3342 window_hrs.precision, window_hrs.value, window_hrs.suffix,
3343 (unsigned long long)info->windowSize);
3344 DISPLAYOUT("Compressed Size: %.*f%s (%llu B)\n",
3345 compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3346 (unsigned long long)info->compressedSize);
3347 if (!info->decompUnavailable) {
3348 DISPLAYOUT("Decompressed Size: %.*f%s (%llu B)\n",
3349 decompressed_hrs.precision, decompressed_hrs.value, decompressed_hrs.suffix,
3350 (unsigned long long)info->decompressedSize);
3351 DISPLAYOUT("Ratio: %.4f\n", ratio);
3354 if (info->usesCheck && info->numActualFrames == 1) {
3355 DISPLAYOUT("Check: %s %02x%02x%02x%02x\n", checkString,
3356 info->checksum[3], info->checksum[2],
3357 info->checksum[1], info->checksum[0]
3360 DISPLAYOUT("Check: %s\n", checkString);
3367 static fileInfo_t FIO_addFInfo(fileInfo_t fi1, fileInfo_t fi2)
3370 memset(&total, 0, sizeof(total));
3371 total.numActualFrames = fi1.numActualFrames + fi2.numActualFrames;
3372 total.numSkippableFrames = fi1.numSkippableFrames + fi2.numSkippableFrames;
3373 total.compressedSize = fi1.compressedSize + fi2.compressedSize;
3374 total.decompressedSize = fi1.decompressedSize + fi2.decompressedSize;
3375 total.decompUnavailable = fi1.decompUnavailable | fi2.decompUnavailable;
3376 total.usesCheck = fi1.usesCheck & fi2.usesCheck;
3377 total.nbFiles = fi1.nbFiles + fi2.nbFiles;
3382 FIO_listFile(fileInfo_t* total, const char* inFileName, int displayLevel)
3385 memset(&info, 0, sizeof(info));
3386 { InfoError const error = getFileInfo(&info, inFileName);
3388 case info_frame_error:
3389 /* display error, but provide output */
3390 DISPLAYLEVEL(1, "Error while parsing \"%s\" \n", inFileName);
3393 DISPLAYOUT("File \"%s\" not compressed by zstd \n", inFileName);
3394 if (displayLevel > 2) DISPLAYOUT("\n");
3396 case info_file_error:
3397 /* error occurred while opening the file */
3398 if (displayLevel > 2) DISPLAYOUT("\n");
3400 case info_truncated_input:
3401 DISPLAYOUT("File \"%s\" is truncated \n", inFileName);
3402 if (displayLevel > 2) DISPLAYOUT("\n");
3409 displayInfo(inFileName, &info, displayLevel);
3410 *total = FIO_addFInfo(*total, info);
3411 assert(error == info_success || error == info_frame_error);
3416 int FIO_listMultipleFiles(unsigned numFiles, const char** filenameTable, int displayLevel)
3418 /* ensure no specified input is stdin (needs fseek() capability) */
3420 for (u=0; u<numFiles;u++) {
3421 ERROR_IF(!strcmp (filenameTable[u], stdinmark),
3422 1, "zstd: --list does not support reading from standard input");
3425 if (numFiles == 0) {
3426 if (!UTIL_isConsole(stdin)) {
3427 DISPLAYLEVEL(1, "zstd: --list does not support reading from standard input \n");
3429 DISPLAYLEVEL(1, "No files given \n");
3433 if (displayLevel <= 2) {
3434 DISPLAYOUT("Frames Skips Compressed Uncompressed Ratio Check Filename\n");
3438 memset(&total, 0, sizeof(total));
3439 total.usesCheck = 1;
3440 /* --list each file, and check for any error */
3442 for (u=0; u<numFiles;u++) {
3443 error |= FIO_listFile(&total, filenameTable[u], displayLevel);
3445 if (numFiles > 1 && displayLevel <= 2) { /* display total */
3446 UTIL_HumanReadableSize_t const compressed_hrs = UTIL_makeHumanReadableSize(total.compressedSize);
3447 UTIL_HumanReadableSize_t const decompressed_hrs = UTIL_makeHumanReadableSize(total.decompressedSize);
3448 double const ratio = (total.compressedSize == 0) ? 0 : ((double)total.decompressedSize)/(double)total.compressedSize;
3449 const char* const checkString = (total.usesCheck ? "XXH64" : "");
3450 DISPLAYOUT("----------------------------------------------------------------- \n");
3451 if (total.decompUnavailable) {
3452 DISPLAYOUT("%6d %5d %6.*f%4s %5s %u files\n",
3453 total.numSkippableFrames + total.numActualFrames,
3454 total.numSkippableFrames,
3455 compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3456 checkString, (unsigned)total.nbFiles);
3458 DISPLAYOUT("%6d %5d %6.*f%4s %8.*f%4s %5.3f %5s %u files\n",
3459 total.numSkippableFrames + total.numActualFrames,
3460 total.numSkippableFrames,
3461 compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3462 decompressed_hrs.precision, decompressed_hrs.value, decompressed_hrs.suffix,
3463 ratio, checkString, (unsigned)total.nbFiles);
3470 #endif /* #ifndef ZSTD_NODECOMPRESS */