xref: /freebsd/sys/contrib/zstd/programs/fileio.c (revision 5ff13fbc199bdf5f0572845351c68ee5ca828e71)
1 /*
2  * Copyright (c) Yann Collet, Facebook, Inc.
3  * All rights reserved.
4  *
5  * This source code is licensed under both the BSD-style license (found in the
6  * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7  * in the COPYING file in the root directory of this source tree).
8  * You may select, at your option, one of the above-listed licenses.
9  */
10 
11 
12 /* *************************************
13 *  Compiler Options
14 ***************************************/
15 #ifdef _MSC_VER   /* Visual */
16 #  pragma warning(disable : 4127)  /* disable: C4127: conditional expression is constant */
17 #  pragma warning(disable : 4204)  /* non-constant aggregate initializer */
18 #endif
19 #if defined(__MINGW32__) && !defined(_POSIX_SOURCE)
20 #  define _POSIX_SOURCE 1          /* disable %llu warnings with MinGW on Windows */
21 #endif
22 
23 /*-*************************************
24 *  Includes
25 ***************************************/
26 #include "platform.h"   /* Large Files support, SET_BINARY_MODE */
27 #include "util.h"       /* UTIL_getFileSize, UTIL_isRegularFile, UTIL_isSameFile */
28 #include <stdio.h>      /* fprintf, open, fdopen, fread, _fileno, stdin, stdout */
29 #include <stdlib.h>     /* malloc, free */
30 #include <string.h>     /* strcmp, strlen */
31 #include <fcntl.h>      /* O_WRONLY */
32 #include <assert.h>
33 #include <errno.h>      /* errno */
34 #include <limits.h>     /* INT_MAX */
35 #include <signal.h>
36 #include "timefn.h"     /* UTIL_getTime, UTIL_clockSpanMicro */
37 
38 #if defined (_MSC_VER)
39 #  include <sys/stat.h>
40 #  include <io.h>
41 #endif
42 
43 #include "../lib/common/mem.h"     /* U32, U64 */
44 #include "fileio.h"
45 
46 #define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_magicNumber, ZSTD_frameHeaderSize_max */
47 #include "../lib/zstd.h"
48 #include "../lib/zstd_errors.h"  /* ZSTD_error_frameParameter_windowTooLarge */
49 
50 #if defined(ZSTD_GZCOMPRESS) || defined(ZSTD_GZDECOMPRESS)
51 #  include <zlib.h>
52 #  if !defined(z_const)
53 #    define z_const
54 #  endif
55 #endif
56 
57 #if defined(ZSTD_LZMACOMPRESS) || defined(ZSTD_LZMADECOMPRESS)
58 #  include <lzma.h>
59 #endif
60 
61 #define LZ4_MAGICNUMBER 0x184D2204
62 #if defined(ZSTD_LZ4COMPRESS) || defined(ZSTD_LZ4DECOMPRESS)
63 #  define LZ4F_ENABLE_OBSOLETE_ENUMS
64 #  include <lz4frame.h>
65 #  include <lz4.h>
66 #endif
67 
68 
69 /*-*************************************
70 *  Constants
71 ***************************************/
72 #define ADAPT_WINDOWLOG_DEFAULT 23   /* 8 MB */
73 #define DICTSIZE_MAX (32 MB)   /* protection against large input (attack scenario) */
74 
75 #define FNSPACE 30
76 
77 /* Default file permissions 0666 (modulated by umask) */
78 #if !defined(_WIN32)
79 /* These macros aren't defined on windows. */
80 #define DEFAULT_FILE_PERMISSIONS (S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP|S_IROTH|S_IWOTH)
81 #else
82 #define DEFAULT_FILE_PERMISSIONS (0666)
83 #endif
84 
85 /*-*************************************
86 *  Macros
87 ***************************************/
88 #define KB *(1 <<10)
89 #define MB *(1 <<20)
90 #define GB *(1U<<30)
91 #undef MAX
92 #define MAX(a,b) ((a)>(b) ? (a) : (b))
93 
94 struct FIO_display_prefs_s {
95     int displayLevel;   /* 0 : no display;  1: errors;  2: + result + interaction + warnings;  3: + progression;  4: + information */
96     FIO_progressSetting_e progressSetting;
97 };
98 
99 static FIO_display_prefs_t g_display_prefs = {2, FIO_ps_auto};
100 
101 #define DISPLAY(...)         fprintf(stderr, __VA_ARGS__)
102 #define DISPLAYOUT(...)      fprintf(stdout, __VA_ARGS__)
103 #define DISPLAYLEVEL(l, ...) { if (g_display_prefs.displayLevel>=l) { DISPLAY(__VA_ARGS__); } }
104 
105 static const U64 g_refreshRate = SEC_TO_MICRO / 6;
106 static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
107 
108 #define READY_FOR_UPDATE() ((g_display_prefs.progressSetting != FIO_ps_never) && UTIL_clockSpanMicro(g_displayClock) > g_refreshRate)
109 #define DELAY_NEXT_UPDATE() { g_displayClock = UTIL_getTime(); }
110 #define DISPLAYUPDATE(l, ...) {                              \
111         if (g_display_prefs.displayLevel>=l && (g_display_prefs.progressSetting != FIO_ps_never)) { \
112             if (READY_FOR_UPDATE() || (g_display_prefs.displayLevel>=4)) { \
113                 DELAY_NEXT_UPDATE();                         \
114                 DISPLAY(__VA_ARGS__);                        \
115                 if (g_display_prefs.displayLevel>=4) fflush(stderr);       \
116     }   }   }
117 
118 #undef MIN  /* in case it would be already defined */
119 #define MIN(a,b)    ((a) < (b) ? (a) : (b))
120 
121 
122 #define EXM_THROW(error, ...)                                             \
123 {                                                                         \
124     DISPLAYLEVEL(1, "zstd: ");                                            \
125     DISPLAYLEVEL(5, "Error defined at %s, line %i : \n", __FILE__, __LINE__); \
126     DISPLAYLEVEL(1, "error %i : ", error);                                \
127     DISPLAYLEVEL(1, __VA_ARGS__);                                         \
128     DISPLAYLEVEL(1, " \n");                                               \
129     exit(error);                                                          \
130 }
131 
132 #define CHECK_V(v, f)                                \
133     v = f;                                           \
134     if (ZSTD_isError(v)) {                           \
135         DISPLAYLEVEL(5, "%s \n", #f);                \
136         EXM_THROW(11, "%s", ZSTD_getErrorName(v));   \
137     }
138 #define CHECK(f) { size_t err; CHECK_V(err, f); }
139 
140 
141 /*-************************************
142 *  Signal (Ctrl-C trapping)
143 **************************************/
144 static const char* g_artefact = NULL;
INThandler(int sig)145 static void INThandler(int sig)
146 {
147     assert(sig==SIGINT); (void)sig;
148 #if !defined(_MSC_VER)
149     signal(sig, SIG_IGN);  /* this invocation generates a buggy warning in Visual Studio */
150 #endif
151     if (g_artefact) {
152         assert(UTIL_isRegularFile(g_artefact));
153         remove(g_artefact);
154     }
155     DISPLAY("\n");
156     exit(2);
157 }
addHandler(char const * dstFileName)158 static void addHandler(char const* dstFileName)
159 {
160     if (UTIL_isRegularFile(dstFileName)) {
161         g_artefact = dstFileName;
162         signal(SIGINT, INThandler);
163     } else {
164         g_artefact = NULL;
165     }
166 }
167 /* Idempotent */
clearHandler(void)168 static void clearHandler(void)
169 {
170     if (g_artefact) signal(SIGINT, SIG_DFL);
171     g_artefact = NULL;
172 }
173 
174 
175 /*-*********************************************************
176 *  Termination signal trapping (Print debug stack trace)
177 ***********************************************************/
178 #if defined(__has_feature) && !defined(BACKTRACE_ENABLE) /* Clang compiler */
179 #  if (__has_feature(address_sanitizer))
180 #    define BACKTRACE_ENABLE 0
181 #  endif /* __has_feature(address_sanitizer) */
182 #elif defined(__SANITIZE_ADDRESS__) && !defined(BACKTRACE_ENABLE) /* GCC compiler */
183 #  define BACKTRACE_ENABLE 0
184 #endif
185 
186 #if !defined(BACKTRACE_ENABLE)
187 /* automatic detector : backtrace enabled by default on linux+glibc and osx */
188 #  if (defined(__linux__) && (defined(__GLIBC__) && !defined(__UCLIBC__))) \
189      || (defined(__APPLE__) && defined(__MACH__))
190 #    define BACKTRACE_ENABLE 1
191 #  else
192 #    define BACKTRACE_ENABLE 0
193 #  endif
194 #endif
195 
196 /* note : after this point, BACKTRACE_ENABLE is necessarily defined */
197 
198 
199 #if BACKTRACE_ENABLE
200 
201 #include <execinfo.h>   /* backtrace, backtrace_symbols */
202 
203 #define MAX_STACK_FRAMES    50
204 
ABRThandler(int sig)205 static void ABRThandler(int sig) {
206     const char* name;
207     void* addrlist[MAX_STACK_FRAMES];
208     char** symbollist;
209     int addrlen, i;
210 
211     switch (sig) {
212         case SIGABRT: name = "SIGABRT"; break;
213         case SIGFPE: name = "SIGFPE"; break;
214         case SIGILL: name = "SIGILL"; break;
215         case SIGINT: name = "SIGINT"; break;
216         case SIGSEGV: name = "SIGSEGV"; break;
217         default: name = "UNKNOWN";
218     }
219 
220     DISPLAY("Caught %s signal, printing stack:\n", name);
221     /* Retrieve current stack addresses. */
222     addrlen = backtrace(addrlist, MAX_STACK_FRAMES);
223     if (addrlen == 0) {
224         DISPLAY("\n");
225         return;
226     }
227     /* Create readable strings to each frame. */
228     symbollist = backtrace_symbols(addrlist, addrlen);
229     /* Print the stack trace, excluding calls handling the signal. */
230     for (i = ZSTD_START_SYMBOLLIST_FRAME; i < addrlen; i++) {
231         DISPLAY("%s\n", symbollist[i]);
232     }
233     free(symbollist);
234     /* Reset and raise the signal so default handler runs. */
235     signal(sig, SIG_DFL);
236     raise(sig);
237 }
238 #endif
239 
FIO_addAbortHandler()240 void FIO_addAbortHandler()
241 {
242 #if BACKTRACE_ENABLE
243     signal(SIGABRT, ABRThandler);
244     signal(SIGFPE, ABRThandler);
245     signal(SIGILL, ABRThandler);
246     signal(SIGSEGV, ABRThandler);
247     signal(SIGBUS, ABRThandler);
248 #endif
249 }
250 
251 
252 /*-************************************************************
253 * Avoid fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW
254 ***************************************************************/
255 #if defined(_MSC_VER) && _MSC_VER >= 1400
256 #   define LONG_SEEK _fseeki64
257 #   define LONG_TELL _ftelli64
258 #elif !defined(__64BIT__) && (PLATFORM_POSIX_VERSION >= 200112L) /* No point defining Large file for 64 bit */
259 #  define LONG_SEEK fseeko
260 #  define LONG_TELL ftello
261 #elif defined(__MINGW32__) && !defined(__STRICT_ANSI__) && !defined(__NO_MINGW_LFS) && defined(__MSVCRT__)
262 #   define LONG_SEEK fseeko64
263 #   define LONG_TELL ftello64
264 #elif defined(_WIN32) && !defined(__DJGPP__)
265 #   include <windows.h>
LONG_SEEK(FILE * file,__int64 offset,int origin)266     static int LONG_SEEK(FILE* file, __int64 offset, int origin) {
267         LARGE_INTEGER off;
268         DWORD method;
269         off.QuadPart = offset;
270         if (origin == SEEK_END)
271             method = FILE_END;
272         else if (origin == SEEK_CUR)
273             method = FILE_CURRENT;
274         else
275             method = FILE_BEGIN;
276 
277         if (SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, NULL, method))
278             return 0;
279         else
280             return -1;
281     }
LONG_TELL(FILE * file)282     static __int64 LONG_TELL(FILE* file) {
283         LARGE_INTEGER off, newOff;
284         off.QuadPart = 0;
285         newOff.QuadPart = 0;
286         SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, &newOff, FILE_CURRENT);
287         return newOff.QuadPart;
288     }
289 #else
290 #   define LONG_SEEK fseek
291 #   define LONG_TELL ftell
292 #endif
293 
294 
295 /*-*************************************
296 *  Parameters: FIO_prefs_t
297 ***************************************/
298 
299 /* typedef'd to FIO_prefs_t within fileio.h */
300 struct FIO_prefs_s {
301 
302     /* Algorithm preferences */
303     FIO_compressionType_t compressionType;
304     U32 sparseFileSupport;   /* 0: no sparse allowed; 1: auto (file yes, stdout no); 2: force sparse */
305     int dictIDFlag;
306     int checksumFlag;
307     int blockSize;
308     int overlapLog;
309     U32 adaptiveMode;
310     U32 useRowMatchFinder;
311     int rsyncable;
312     int minAdaptLevel;
313     int maxAdaptLevel;
314     int ldmFlag;
315     int ldmHashLog;
316     int ldmMinMatch;
317     int ldmBucketSizeLog;
318     int ldmHashRateLog;
319     size_t streamSrcSize;
320     size_t targetCBlockSize;
321     int srcSizeHint;
322     int testMode;
323     ZSTD_paramSwitch_e literalCompressionMode;
324 
325     /* IO preferences */
326     U32 removeSrcFile;
327     U32 overwrite;
328 
329     /* Computation resources preferences */
330     unsigned memLimit;
331     int nbWorkers;
332 
333     int excludeCompressedFiles;
334     int patchFromMode;
335     int contentSize;
336     int allowBlockDevices;
337 };
338 
339 /*-*************************************
340 *  Parameters: FIO_ctx_t
341 ***************************************/
342 
343 /* typedef'd to FIO_ctx_t within fileio.h */
344 struct FIO_ctx_s {
345 
346     /* file i/o info */
347     int nbFilesTotal;
348     int hasStdinInput;
349     int hasStdoutOutput;
350 
351     /* file i/o state */
352     int currFileIdx;
353     int nbFilesProcessed;
354     size_t totalBytesInput;
355     size_t totalBytesOutput;
356 };
357 
358 
359 /*-*************************************
360 *  Parameters: Initialization
361 ***************************************/
362 
363 #define FIO_OVERLAP_LOG_NOTSET 9999
364 #define FIO_LDM_PARAM_NOTSET 9999
365 
366 
FIO_createPreferences(void)367 FIO_prefs_t* FIO_createPreferences(void)
368 {
369     FIO_prefs_t* const ret = (FIO_prefs_t*)malloc(sizeof(FIO_prefs_t));
370     if (!ret) EXM_THROW(21, "Allocation error : not enough memory");
371 
372     ret->compressionType = FIO_zstdCompression;
373     ret->overwrite = 0;
374     ret->sparseFileSupport = ZSTD_SPARSE_DEFAULT;
375     ret->dictIDFlag = 1;
376     ret->checksumFlag = 1;
377     ret->removeSrcFile = 0;
378     ret->memLimit = 0;
379     ret->nbWorkers = 1;
380     ret->blockSize = 0;
381     ret->overlapLog = FIO_OVERLAP_LOG_NOTSET;
382     ret->adaptiveMode = 0;
383     ret->rsyncable = 0;
384     ret->minAdaptLevel = -50;   /* initializing this value requires a constant, so ZSTD_minCLevel() doesn't work */
385     ret->maxAdaptLevel = 22;   /* initializing this value requires a constant, so ZSTD_maxCLevel() doesn't work */
386     ret->ldmFlag = 0;
387     ret->ldmHashLog = 0;
388     ret->ldmMinMatch = 0;
389     ret->ldmBucketSizeLog = FIO_LDM_PARAM_NOTSET;
390     ret->ldmHashRateLog = FIO_LDM_PARAM_NOTSET;
391     ret->streamSrcSize = 0;
392     ret->targetCBlockSize = 0;
393     ret->srcSizeHint = 0;
394     ret->testMode = 0;
395     ret->literalCompressionMode = ZSTD_ps_auto;
396     ret->excludeCompressedFiles = 0;
397     ret->allowBlockDevices = 0;
398     return ret;
399 }
400 
FIO_createContext(void)401 FIO_ctx_t* FIO_createContext(void)
402 {
403     FIO_ctx_t* const ret = (FIO_ctx_t*)malloc(sizeof(FIO_ctx_t));
404     if (!ret) EXM_THROW(21, "Allocation error : not enough memory");
405 
406     ret->currFileIdx = 0;
407     ret->hasStdinInput = 0;
408     ret->hasStdoutOutput = 0;
409     ret->nbFilesTotal = 1;
410     ret->nbFilesProcessed = 0;
411     ret->totalBytesInput = 0;
412     ret->totalBytesOutput = 0;
413     return ret;
414 }
415 
FIO_freePreferences(FIO_prefs_t * const prefs)416 void FIO_freePreferences(FIO_prefs_t* const prefs)
417 {
418     free(prefs);
419 }
420 
FIO_freeContext(FIO_ctx_t * const fCtx)421 void FIO_freeContext(FIO_ctx_t* const fCtx)
422 {
423     free(fCtx);
424 }
425 
426 
427 /*-*************************************
428 *  Parameters: Display Options
429 ***************************************/
430 
FIO_setNotificationLevel(int level)431 void FIO_setNotificationLevel(int level) { g_display_prefs.displayLevel=level; }
432 
FIO_setProgressSetting(FIO_progressSetting_e setting)433 void FIO_setProgressSetting(FIO_progressSetting_e setting) { g_display_prefs.progressSetting = setting; }
434 
435 
436 /*-*************************************
437 *  Parameters: Setters
438 ***************************************/
439 
440 /* FIO_prefs_t functions */
441 
FIO_setCompressionType(FIO_prefs_t * const prefs,FIO_compressionType_t compressionType)442 void FIO_setCompressionType(FIO_prefs_t* const prefs, FIO_compressionType_t compressionType) { prefs->compressionType = compressionType; }
443 
FIO_overwriteMode(FIO_prefs_t * const prefs)444 void FIO_overwriteMode(FIO_prefs_t* const prefs) { prefs->overwrite = 1; }
445 
FIO_setSparseWrite(FIO_prefs_t * const prefs,unsigned sparse)446 void FIO_setSparseWrite(FIO_prefs_t* const prefs, unsigned sparse) { prefs->sparseFileSupport = sparse; }
447 
FIO_setDictIDFlag(FIO_prefs_t * const prefs,int dictIDFlag)448 void FIO_setDictIDFlag(FIO_prefs_t* const prefs, int dictIDFlag) { prefs->dictIDFlag = dictIDFlag; }
449 
FIO_setChecksumFlag(FIO_prefs_t * const prefs,int checksumFlag)450 void FIO_setChecksumFlag(FIO_prefs_t* const prefs, int checksumFlag) { prefs->checksumFlag = checksumFlag; }
451 
FIO_setRemoveSrcFile(FIO_prefs_t * const prefs,unsigned flag)452 void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, unsigned flag) { prefs->removeSrcFile = (flag>0); }
453 
FIO_setMemLimit(FIO_prefs_t * const prefs,unsigned memLimit)454 void FIO_setMemLimit(FIO_prefs_t* const prefs, unsigned memLimit) { prefs->memLimit = memLimit; }
455 
FIO_setNbWorkers(FIO_prefs_t * const prefs,int nbWorkers)456 void FIO_setNbWorkers(FIO_prefs_t* const prefs, int nbWorkers) {
457 #ifndef ZSTD_MULTITHREAD
458     if (nbWorkers > 0) DISPLAYLEVEL(2, "Note : multi-threading is disabled \n");
459 #endif
460     prefs->nbWorkers = nbWorkers;
461 }
462 
FIO_setExcludeCompressedFile(FIO_prefs_t * const prefs,int excludeCompressedFiles)463 void FIO_setExcludeCompressedFile(FIO_prefs_t* const prefs, int excludeCompressedFiles) { prefs->excludeCompressedFiles = excludeCompressedFiles; }
464 
FIO_setAllowBlockDevices(FIO_prefs_t * const prefs,int allowBlockDevices)465 void FIO_setAllowBlockDevices(FIO_prefs_t* const prefs, int allowBlockDevices) { prefs->allowBlockDevices = allowBlockDevices; }
466 
FIO_setBlockSize(FIO_prefs_t * const prefs,int blockSize)467 void FIO_setBlockSize(FIO_prefs_t* const prefs, int blockSize) {
468     if (blockSize && prefs->nbWorkers==0)
469         DISPLAYLEVEL(2, "Setting block size is useless in single-thread mode \n");
470     prefs->blockSize = blockSize;
471 }
472 
FIO_setOverlapLog(FIO_prefs_t * const prefs,int overlapLog)473 void FIO_setOverlapLog(FIO_prefs_t* const prefs, int overlapLog){
474     if (overlapLog && prefs->nbWorkers==0)
475         DISPLAYLEVEL(2, "Setting overlapLog is useless in single-thread mode \n");
476     prefs->overlapLog = overlapLog;
477 }
478 
FIO_setAdaptiveMode(FIO_prefs_t * const prefs,unsigned adapt)479 void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, unsigned adapt) {
480     if ((adapt>0) && (prefs->nbWorkers==0))
481         EXM_THROW(1, "Adaptive mode is not compatible with single thread mode \n");
482     prefs->adaptiveMode = adapt;
483 }
484 
FIO_setUseRowMatchFinder(FIO_prefs_t * const prefs,int useRowMatchFinder)485 void FIO_setUseRowMatchFinder(FIO_prefs_t* const prefs, int useRowMatchFinder) {
486     prefs->useRowMatchFinder = useRowMatchFinder;
487 }
488 
FIO_setRsyncable(FIO_prefs_t * const prefs,int rsyncable)489 void FIO_setRsyncable(FIO_prefs_t* const prefs, int rsyncable) {
490     if ((rsyncable>0) && (prefs->nbWorkers==0))
491         EXM_THROW(1, "Rsyncable mode is not compatible with single thread mode \n");
492     prefs->rsyncable = rsyncable;
493 }
494 
FIO_setStreamSrcSize(FIO_prefs_t * const prefs,size_t streamSrcSize)495 void FIO_setStreamSrcSize(FIO_prefs_t* const prefs, size_t streamSrcSize) {
496     prefs->streamSrcSize = streamSrcSize;
497 }
498 
FIO_setTargetCBlockSize(FIO_prefs_t * const prefs,size_t targetCBlockSize)499 void FIO_setTargetCBlockSize(FIO_prefs_t* const prefs, size_t targetCBlockSize) {
500     prefs->targetCBlockSize = targetCBlockSize;
501 }
502 
FIO_setSrcSizeHint(FIO_prefs_t * const prefs,size_t srcSizeHint)503 void FIO_setSrcSizeHint(FIO_prefs_t* const prefs, size_t srcSizeHint) {
504     prefs->srcSizeHint = (int)MIN((size_t)INT_MAX, srcSizeHint);
505 }
506 
FIO_setTestMode(FIO_prefs_t * const prefs,int testMode)507 void FIO_setTestMode(FIO_prefs_t* const prefs, int testMode) {
508     prefs->testMode = (testMode!=0);
509 }
510 
FIO_setLiteralCompressionMode(FIO_prefs_t * const prefs,ZSTD_paramSwitch_e mode)511 void FIO_setLiteralCompressionMode(
512         FIO_prefs_t* const prefs,
513         ZSTD_paramSwitch_e mode) {
514     prefs->literalCompressionMode = mode;
515 }
516 
FIO_setAdaptMin(FIO_prefs_t * const prefs,int minCLevel)517 void FIO_setAdaptMin(FIO_prefs_t* const prefs, int minCLevel)
518 {
519 #ifndef ZSTD_NOCOMPRESS
520     assert(minCLevel >= ZSTD_minCLevel());
521 #endif
522     prefs->minAdaptLevel = minCLevel;
523 }
524 
FIO_setAdaptMax(FIO_prefs_t * const prefs,int maxCLevel)525 void FIO_setAdaptMax(FIO_prefs_t* const prefs, int maxCLevel)
526 {
527     prefs->maxAdaptLevel = maxCLevel;
528 }
529 
FIO_setLdmFlag(FIO_prefs_t * const prefs,unsigned ldmFlag)530 void FIO_setLdmFlag(FIO_prefs_t* const prefs, unsigned ldmFlag) {
531     prefs->ldmFlag = (ldmFlag>0);
532 }
533 
FIO_setLdmHashLog(FIO_prefs_t * const prefs,int ldmHashLog)534 void FIO_setLdmHashLog(FIO_prefs_t* const prefs, int ldmHashLog) {
535     prefs->ldmHashLog = ldmHashLog;
536 }
537 
FIO_setLdmMinMatch(FIO_prefs_t * const prefs,int ldmMinMatch)538 void FIO_setLdmMinMatch(FIO_prefs_t* const prefs, int ldmMinMatch) {
539     prefs->ldmMinMatch = ldmMinMatch;
540 }
541 
FIO_setLdmBucketSizeLog(FIO_prefs_t * const prefs,int ldmBucketSizeLog)542 void FIO_setLdmBucketSizeLog(FIO_prefs_t* const prefs, int ldmBucketSizeLog) {
543     prefs->ldmBucketSizeLog = ldmBucketSizeLog;
544 }
545 
546 
FIO_setLdmHashRateLog(FIO_prefs_t * const prefs,int ldmHashRateLog)547 void FIO_setLdmHashRateLog(FIO_prefs_t* const prefs, int ldmHashRateLog) {
548     prefs->ldmHashRateLog = ldmHashRateLog;
549 }
550 
FIO_setPatchFromMode(FIO_prefs_t * const prefs,int value)551 void FIO_setPatchFromMode(FIO_prefs_t* const prefs, int value)
552 {
553     prefs->patchFromMode = value != 0;
554 }
555 
FIO_setContentSize(FIO_prefs_t * const prefs,int value)556 void FIO_setContentSize(FIO_prefs_t* const prefs, int value)
557 {
558     prefs->contentSize = value != 0;
559 }
560 
561 /* FIO_ctx_t functions */
562 
FIO_setHasStdoutOutput(FIO_ctx_t * const fCtx,int value)563 void FIO_setHasStdoutOutput(FIO_ctx_t* const fCtx, int value) {
564     fCtx->hasStdoutOutput = value;
565 }
566 
FIO_setNbFilesTotal(FIO_ctx_t * const fCtx,int value)567 void FIO_setNbFilesTotal(FIO_ctx_t* const fCtx, int value)
568 {
569     fCtx->nbFilesTotal = value;
570 }
571 
FIO_determineHasStdinInput(FIO_ctx_t * const fCtx,const FileNamesTable * const filenames)572 void FIO_determineHasStdinInput(FIO_ctx_t* const fCtx, const FileNamesTable* const filenames) {
573     size_t i = 0;
574     for ( ; i < filenames->tableSize; ++i) {
575         if (!strcmp(stdinmark, filenames->fileNames[i])) {
576             fCtx->hasStdinInput = 1;
577             return;
578         }
579     }
580 }
581 
582 /*-*************************************
583 *  Functions
584 ***************************************/
585 /** FIO_removeFile() :
586  * @result : Unlink `fileName`, even if it's read-only */
FIO_removeFile(const char * path)587 static int FIO_removeFile(const char* path)
588 {
589     stat_t statbuf;
590     if (!UTIL_stat(path, &statbuf)) {
591         DISPLAYLEVEL(2, "zstd: Failed to stat %s while trying to remove it\n", path);
592         return 0;
593     }
594     if (!UTIL_isRegularFileStat(&statbuf)) {
595         DISPLAYLEVEL(2, "zstd: Refusing to remove non-regular file %s\n", path);
596         return 0;
597     }
598 #if defined(_WIN32) || defined(WIN32)
599     /* windows doesn't allow remove read-only files,
600      * so try to make it writable first */
601     if (!(statbuf.st_mode & _S_IWRITE)) {
602         UTIL_chmod(path, &statbuf, _S_IWRITE);
603     }
604 #endif
605     return remove(path);
606 }
607 
608 /** FIO_openSrcFile() :
609  *  condition : `srcFileName` must be non-NULL. `prefs` may be NULL.
610  * @result : FILE* to `srcFileName`, or NULL if it fails */
FIO_openSrcFile(const FIO_prefs_t * const prefs,const char * srcFileName)611 static FILE* FIO_openSrcFile(const FIO_prefs_t* const prefs, const char* srcFileName)
612 {
613     stat_t statbuf;
614     int allowBlockDevices = prefs != NULL ? prefs->allowBlockDevices : 0;
615     assert(srcFileName != NULL);
616     if (!strcmp (srcFileName, stdinmark)) {
617         DISPLAYLEVEL(4,"Using stdin for input \n");
618         SET_BINARY_MODE(stdin);
619         return stdin;
620     }
621 
622     if (!UTIL_stat(srcFileName, &statbuf)) {
623         DISPLAYLEVEL(1, "zstd: can't stat %s : %s -- ignored \n",
624                         srcFileName, strerror(errno));
625         return NULL;
626     }
627 
628     if (!UTIL_isRegularFileStat(&statbuf)
629      && !UTIL_isFIFOStat(&statbuf)
630      && !(allowBlockDevices && UTIL_isBlockDevStat(&statbuf))
631     ) {
632         DISPLAYLEVEL(1, "zstd: %s is not a regular file -- ignored \n",
633                         srcFileName);
634         return NULL;
635     }
636 
637     {   FILE* const f = fopen(srcFileName, "rb");
638         if (f == NULL)
639             DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno));
640         return f;
641     }
642 }
643 
644 /** FIO_openDstFile() :
645  *  condition : `dstFileName` must be non-NULL.
646  * @result : FILE* to `dstFileName`, or NULL if it fails */
647 static FILE*
FIO_openDstFile(FIO_ctx_t * fCtx,FIO_prefs_t * const prefs,const char * srcFileName,const char * dstFileName,const int mode)648 FIO_openDstFile(FIO_ctx_t* fCtx, FIO_prefs_t* const prefs,
649                 const char* srcFileName, const char* dstFileName,
650                 const int mode)
651 {
652     if (prefs->testMode) return NULL;  /* do not open file in test mode */
653 
654     assert(dstFileName != NULL);
655     if (!strcmp (dstFileName, stdoutmark)) {
656         DISPLAYLEVEL(4,"Using stdout for output \n");
657         SET_BINARY_MODE(stdout);
658         if (prefs->sparseFileSupport == 1) {
659             prefs->sparseFileSupport = 0;
660             DISPLAYLEVEL(4, "Sparse File Support is automatically disabled on stdout ; try --sparse \n");
661         }
662         return stdout;
663     }
664 
665     /* ensure dst is not the same as src */
666     if (srcFileName != NULL && UTIL_isSameFile(srcFileName, dstFileName)) {
667         DISPLAYLEVEL(1, "zstd: Refusing to open an output file which will overwrite the input file \n");
668         return NULL;
669     }
670 
671     if (prefs->sparseFileSupport == 1) {
672         prefs->sparseFileSupport = ZSTD_SPARSE_DEFAULT;
673     }
674 
675     if (UTIL_isRegularFile(dstFileName)) {
676         /* Check if destination file already exists */
677 #if !defined(_WIN32)
678         /* this test does not work on Windows :
679          * `NUL` and `nul` are detected as regular files */
680         if (!strcmp(dstFileName, nulmark)) {
681             EXM_THROW(40, "%s is unexpectedly categorized as a regular file",
682                         dstFileName);
683         }
684 #endif
685         if (!prefs->overwrite) {
686             if (g_display_prefs.displayLevel <= 1) {
687                 /* No interaction possible */
688                 DISPLAY("zstd: %s already exists; not overwritten  \n",
689                         dstFileName);
690                 return NULL;
691             }
692             DISPLAY("zstd: %s already exists; ", dstFileName);
693             if (UTIL_requireUserConfirmation("overwrite (y/n) ? ", "Not overwritten  \n", "yY", fCtx->hasStdinInput))
694                 return NULL;
695         }
696         /* need to unlink */
697         FIO_removeFile(dstFileName);
698     }
699 
700     {
701 #if defined(_WIN32)
702         /* Windows requires opening the file as a "binary" file to avoid
703          * mangling. This macro doesn't exist on unix. */
704         const int openflags = O_WRONLY|O_CREAT|O_TRUNC|O_BINARY;
705         const int fd = _open(dstFileName, openflags, mode);
706         FILE* f = NULL;
707         if (fd != -1) {
708             f = _fdopen(fd, "wb");
709         }
710 #else
711         const int openflags = O_WRONLY|O_CREAT|O_TRUNC;
712         const int fd = open(dstFileName, openflags, mode);
713         FILE* f = NULL;
714         if (fd != -1) {
715             f = fdopen(fd, "wb");
716         }
717 #endif
718         if (f == NULL) {
719             DISPLAYLEVEL(1, "zstd: %s: %s\n", dstFileName, strerror(errno));
720         }
721         return f;
722     }
723 }
724 
725 /*! FIO_createDictBuffer() :
726  *  creates a buffer, pointed by `*bufferPtr`,
727  *  loads `filename` content into it, up to DICTSIZE_MAX bytes.
728  * @return : loaded size
729  *  if fileName==NULL, returns 0 and a NULL pointer
730  */
FIO_createDictBuffer(void ** bufferPtr,const char * fileName,FIO_prefs_t * const prefs)731 static size_t FIO_createDictBuffer(void** bufferPtr, const char* fileName, FIO_prefs_t* const prefs)
732 {
733     FILE* fileHandle;
734     U64 fileSize;
735     stat_t statbuf;
736 
737     assert(bufferPtr != NULL);
738     *bufferPtr = NULL;
739     if (fileName == NULL) return 0;
740 
741     DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
742 
743     if (!UTIL_stat(fileName, &statbuf)) {
744         EXM_THROW(31, "Stat failed on dictionary file %s: %s", fileName, strerror(errno));
745     }
746 
747     if (!UTIL_isRegularFileStat(&statbuf)) {
748         EXM_THROW(32, "Dictionary %s must be a regular file.", fileName);
749     }
750 
751     fileHandle = fopen(fileName, "rb");
752 
753     if (fileHandle == NULL) {
754         EXM_THROW(33, "Couldn't open dictionary %s: %s", fileName, strerror(errno));
755     }
756 
757     fileSize = UTIL_getFileSizeStat(&statbuf);
758     {
759         size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX;
760         if (fileSize >  dictSizeMax) {
761             EXM_THROW(34, "Dictionary file %s is too large (> %u bytes)",
762                             fileName,  (unsigned)dictSizeMax);   /* avoid extreme cases */
763         }
764     }
765     *bufferPtr = malloc((size_t)fileSize);
766     if (*bufferPtr==NULL) EXM_THROW(34, "%s", strerror(errno));
767     {   size_t const readSize = fread(*bufferPtr, 1, (size_t)fileSize, fileHandle);
768         if (readSize != fileSize) {
769             EXM_THROW(35, "Error reading dictionary file %s : %s",
770                     fileName, strerror(errno));
771         }
772     }
773     fclose(fileHandle);
774     return (size_t)fileSize;
775 }
776 
777 
778 
779 /* FIO_checkFilenameCollisions() :
780  * Checks for and warns if there are any files that would have the same output path
781  */
FIO_checkFilenameCollisions(const char ** filenameTable,unsigned nbFiles)782 int FIO_checkFilenameCollisions(const char** filenameTable, unsigned nbFiles) {
783     const char **filenameTableSorted, *prevElem, *filename;
784     unsigned u;
785 
786     filenameTableSorted = (const char**) malloc(sizeof(char*) * nbFiles);
787     if (!filenameTableSorted) {
788         DISPLAY("Unable to malloc new str array, not checking for name collisions\n");
789         return 1;
790     }
791 
792     for (u = 0; u < nbFiles; ++u) {
793         filename = strrchr(filenameTable[u], PATH_SEP);
794         if (filename == NULL) {
795             filenameTableSorted[u] = filenameTable[u];
796         } else {
797             filenameTableSorted[u] = filename+1;
798         }
799     }
800 
801     qsort((void*)filenameTableSorted, nbFiles, sizeof(char*), UTIL_compareStr);
802     prevElem = filenameTableSorted[0];
803     for (u = 1; u < nbFiles; ++u) {
804         if (strcmp(prevElem, filenameTableSorted[u]) == 0) {
805             DISPLAY("WARNING: Two files have same filename: %s\n", prevElem);
806         }
807         prevElem = filenameTableSorted[u];
808     }
809 
810     free((void*)filenameTableSorted);
811     return 0;
812 }
813 
814 static const char*
extractFilename(const char * path,char separator)815 extractFilename(const char* path, char separator)
816 {
817     const char* search = strrchr(path, separator);
818     if (search == NULL) return path;
819     return search+1;
820 }
821 
822 /* FIO_createFilename_fromOutDir() :
823  * Takes a source file name and specified output directory, and
824  * allocates memory for and returns a pointer to final path.
825  * This function never returns an error (it may abort() in case of pb)
826  */
827 static char*
FIO_createFilename_fromOutDir(const char * path,const char * outDirName,const size_t suffixLen)828 FIO_createFilename_fromOutDir(const char* path, const char* outDirName, const size_t suffixLen)
829 {
830     const char* filenameStart;
831     char separator;
832     char* result;
833 
834 #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__) /* windows support */
835     separator = '\\';
836 #else
837     separator = '/';
838 #endif
839 
840     filenameStart = extractFilename(path, separator);
841 #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__) /* windows support */
842     filenameStart = extractFilename(filenameStart, '/');  /* sometimes, '/' separator is also used on Windows (mingw+msys2) */
843 #endif
844 
845     result = (char*) calloc(1, strlen(outDirName) + 1 + strlen(filenameStart) + suffixLen + 1);
846     if (!result) {
847         EXM_THROW(30, "zstd: FIO_createFilename_fromOutDir: %s", strerror(errno));
848     }
849 
850     memcpy(result, outDirName, strlen(outDirName));
851     if (outDirName[strlen(outDirName)-1] == separator) {
852         memcpy(result + strlen(outDirName), filenameStart, strlen(filenameStart));
853     } else {
854         memcpy(result + strlen(outDirName), &separator, 1);
855         memcpy(result + strlen(outDirName) + 1, filenameStart, strlen(filenameStart));
856     }
857 
858     return result;
859 }
860 
861 /* FIO_highbit64() :
862  * gives position of highest bit.
863  * note : only works for v > 0 !
864  */
FIO_highbit64(unsigned long long v)865 static unsigned FIO_highbit64(unsigned long long v)
866 {
867     unsigned count = 0;
868     assert(v != 0);
869     v >>= 1;
870     while (v) { v >>= 1; count++; }
871     return count;
872 }
873 
FIO_adjustMemLimitForPatchFromMode(FIO_prefs_t * const prefs,unsigned long long const dictSize,unsigned long long const maxSrcFileSize)874 static void FIO_adjustMemLimitForPatchFromMode(FIO_prefs_t* const prefs,
875                                     unsigned long long const dictSize,
876                                     unsigned long long const maxSrcFileSize)
877 {
878     unsigned long long maxSize = MAX(prefs->memLimit, MAX(dictSize, maxSrcFileSize));
879     unsigned const maxWindowSize = (1U << ZSTD_WINDOWLOG_MAX);
880     if (maxSize == UTIL_FILESIZE_UNKNOWN)
881         EXM_THROW(42, "Using --patch-from with stdin requires --stream-size");
882     assert(maxSize != UTIL_FILESIZE_UNKNOWN);
883     if (maxSize > maxWindowSize)
884         EXM_THROW(42, "Can't handle files larger than %u GB\n", maxWindowSize/(1 GB));
885     FIO_setMemLimit(prefs, (unsigned)maxSize);
886 }
887 
888 /* FIO_removeMultiFilesWarning() :
889  * Returns 1 if the console should abort, 0 if console should proceed.
890  * This function handles logic when processing multiple files with -o, displaying the appropriate warnings/prompts.
891  *
892  * If -f is specified, or there is just 1 file, zstd will always proceed as usual.
893  * If --rm is specified, there will be a prompt asking for user confirmation.
894  *         If -f is specified with --rm, zstd will proceed as usual
895  *         If -q is specified with --rm, zstd will abort pre-emptively
896  *         If neither flag is specified, zstd will prompt the user for confirmation to proceed.
897  * If --rm is not specified, then zstd will print a warning to the user (which can be silenced with -q).
898  * However, if the output is stdout, we will always abort rather than displaying the warning prompt.
899  */
FIO_removeMultiFilesWarning(FIO_ctx_t * const fCtx,const FIO_prefs_t * const prefs,const char * outFileName,int displayLevelCutoff)900 static int FIO_removeMultiFilesWarning(FIO_ctx_t* const fCtx, const FIO_prefs_t* const prefs, const char* outFileName, int displayLevelCutoff)
901 {
902     int error = 0;
903     if (fCtx->nbFilesTotal > 1 && !prefs->overwrite) {
904         if (g_display_prefs.displayLevel <= displayLevelCutoff) {
905             if (prefs->removeSrcFile) {
906                 DISPLAYLEVEL(1, "zstd: Aborting... not deleting files and processing into dst: %s\n", outFileName);
907                 error =  1;
908             }
909         } else {
910             if (!strcmp(outFileName, stdoutmark)) {
911                 DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into stdout. \n");
912             } else {
913                 DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into a single output file: %s \n", outFileName);
914             }
915             DISPLAYLEVEL(2, "The concatenated output CANNOT regenerate the original directory tree. \n")
916             if (prefs->removeSrcFile) {
917                 if (fCtx->hasStdoutOutput) {
918                     DISPLAYLEVEL(1, "Aborting. Use -f if you really want to delete the files and output to stdout\n");
919                     error = 1;
920                 } else {
921                     error = g_display_prefs.displayLevel > displayLevelCutoff && UTIL_requireUserConfirmation("This is a destructive operation. Proceed? (y/n): ", "Aborting...", "yY", fCtx->hasStdinInput);
922                 }
923             }
924         }
925     }
926     return error;
927 }
928 
929 #ifndef ZSTD_NOCOMPRESS
930 
931 /* **********************************************************************
932  *  Compression
933  ************************************************************************/
934 typedef struct {
935     FILE* srcFile;
936     FILE* dstFile;
937     void*  srcBuffer;
938     size_t srcBufferSize;
939     void*  dstBuffer;
940     size_t dstBufferSize;
941     void* dictBuffer;
942     size_t dictBufferSize;
943     const char* dictFileName;
944     ZSTD_CStream* cctx;
945 } cRess_t;
946 
947 /** ZSTD_cycleLog() :
948  *  condition for correct operation : hashLog > 1 */
ZSTD_cycleLog(U32 hashLog,ZSTD_strategy strat)949 static U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
950 {
951     U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);
952     assert(hashLog > 1);
953     return hashLog - btScale;
954 }
955 
FIO_adjustParamsForPatchFromMode(FIO_prefs_t * const prefs,ZSTD_compressionParameters * comprParams,unsigned long long const dictSize,unsigned long long const maxSrcFileSize,int cLevel)956 static void FIO_adjustParamsForPatchFromMode(FIO_prefs_t* const prefs,
957                                     ZSTD_compressionParameters* comprParams,
958                                     unsigned long long const dictSize,
959                                     unsigned long long const maxSrcFileSize,
960                                     int cLevel)
961 {
962     unsigned const fileWindowLog = FIO_highbit64(maxSrcFileSize) + 1;
963     ZSTD_compressionParameters const cParams = ZSTD_getCParams(cLevel, (size_t)maxSrcFileSize, (size_t)dictSize);
964     FIO_adjustMemLimitForPatchFromMode(prefs, dictSize, maxSrcFileSize);
965     if (fileWindowLog > ZSTD_WINDOWLOG_MAX)
966         DISPLAYLEVEL(1, "Max window log exceeded by file (compression ratio will suffer)\n");
967     comprParams->windowLog = MAX(ZSTD_WINDOWLOG_MIN, MIN(ZSTD_WINDOWLOG_MAX, fileWindowLog));
968     if (fileWindowLog > ZSTD_cycleLog(cParams.chainLog, cParams.strategy)) {
969         if (!prefs->ldmFlag)
970             DISPLAYLEVEL(1, "long mode automatically triggered\n");
971         FIO_setLdmFlag(prefs, 1);
972     }
973     if (cParams.strategy >= ZSTD_btopt) {
974         DISPLAYLEVEL(1, "[Optimal parser notes] Consider the following to improve patch size at the cost of speed:\n");
975         DISPLAYLEVEL(1, "- Use --single-thread mode in the zstd cli\n");
976         DISPLAYLEVEL(1, "- Set a larger targetLength (eg. --zstd=targetLength=4096)\n");
977         DISPLAYLEVEL(1, "- Set a larger chainLog (eg. --zstd=chainLog=%u)\n", ZSTD_CHAINLOG_MAX);
978         DISPLAYLEVEL(1, "Also consider playing around with searchLog and hashLog\n");
979     }
980 }
981 
FIO_createCResources(FIO_prefs_t * const prefs,const char * dictFileName,unsigned long long const maxSrcFileSize,int cLevel,ZSTD_compressionParameters comprParams)982 static cRess_t FIO_createCResources(FIO_prefs_t* const prefs,
983                                     const char* dictFileName, unsigned long long const maxSrcFileSize,
984                                     int cLevel, ZSTD_compressionParameters comprParams) {
985     cRess_t ress;
986     memset(&ress, 0, sizeof(ress));
987 
988     DISPLAYLEVEL(6, "FIO_createCResources \n");
989     ress.cctx = ZSTD_createCCtx();
990     if (ress.cctx == NULL)
991         EXM_THROW(30, "allocation error (%s): can't create ZSTD_CCtx",
992                     strerror(errno));
993     ress.srcBufferSize = ZSTD_CStreamInSize();
994     ress.srcBuffer = malloc(ress.srcBufferSize);
995     ress.dstBufferSize = ZSTD_CStreamOutSize();
996 
997     /* need to update memLimit before calling createDictBuffer
998      * because of memLimit check inside it */
999     if (prefs->patchFromMode) {
1000         unsigned long long const ssSize = (unsigned long long)prefs->streamSrcSize;
1001         FIO_adjustParamsForPatchFromMode(prefs, &comprParams, UTIL_getFileSize(dictFileName), ssSize > 0 ? ssSize : maxSrcFileSize, cLevel);
1002     }
1003     ress.dstBuffer = malloc(ress.dstBufferSize);
1004     ress.dictBufferSize = FIO_createDictBuffer(&ress.dictBuffer, dictFileName, prefs);   /* works with dictFileName==NULL */
1005     if (!ress.srcBuffer || !ress.dstBuffer)
1006         EXM_THROW(31, "allocation error : not enough memory");
1007 
1008     /* Advanced parameters, including dictionary */
1009     if (dictFileName && (ress.dictBuffer==NULL))
1010         EXM_THROW(32, "allocation error : can't create dictBuffer");
1011     ress.dictFileName = dictFileName;
1012 
1013     if (prefs->adaptiveMode && !prefs->ldmFlag && !comprParams.windowLog)
1014         comprParams.windowLog = ADAPT_WINDOWLOG_DEFAULT;
1015 
1016     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_contentSizeFlag, prefs->contentSize) );  /* always enable content size when available (note: supposed to be default) */
1017     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_dictIDFlag, prefs->dictIDFlag) );
1018     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_checksumFlag, prefs->checksumFlag) );
1019     /* compression level */
1020     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, cLevel) );
1021     /* max compressed block size */
1022     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetCBlockSize, (int)prefs->targetCBlockSize) );
1023     /* source size hint */
1024     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_srcSizeHint, (int)prefs->srcSizeHint) );
1025     /* long distance matching */
1026     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_enableLongDistanceMatching, prefs->ldmFlag) );
1027     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashLog, prefs->ldmHashLog) );
1028     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmMinMatch, prefs->ldmMinMatch) );
1029     if (prefs->ldmBucketSizeLog != FIO_LDM_PARAM_NOTSET) {
1030         CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmBucketSizeLog, prefs->ldmBucketSizeLog) );
1031     }
1032     if (prefs->ldmHashRateLog != FIO_LDM_PARAM_NOTSET) {
1033         CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashRateLog, prefs->ldmHashRateLog) );
1034     }
1035     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_useRowMatchFinder, prefs->useRowMatchFinder));
1036     /* compression parameters */
1037     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_windowLog, (int)comprParams.windowLog) );
1038     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_chainLog, (int)comprParams.chainLog) );
1039     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_hashLog, (int)comprParams.hashLog) );
1040     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_searchLog, (int)comprParams.searchLog) );
1041     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_minMatch, (int)comprParams.minMatch) );
1042     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetLength, (int)comprParams.targetLength) );
1043     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_strategy, (int)comprParams.strategy) );
1044     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_literalCompressionMode, (int)prefs->literalCompressionMode) );
1045     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_enableDedicatedDictSearch, 1) );
1046     /* multi-threading */
1047 #ifdef ZSTD_MULTITHREAD
1048     DISPLAYLEVEL(5,"set nb workers = %u \n", prefs->nbWorkers);
1049     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_nbWorkers, prefs->nbWorkers) );
1050     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_jobSize, prefs->blockSize) );
1051     if (prefs->overlapLog != FIO_OVERLAP_LOG_NOTSET) {
1052         DISPLAYLEVEL(3,"set overlapLog = %u \n", prefs->overlapLog);
1053         CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_overlapLog, prefs->overlapLog) );
1054     }
1055     CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_rsyncable, prefs->rsyncable) );
1056 #endif
1057     /* dictionary */
1058     if (prefs->patchFromMode) {
1059         CHECK( ZSTD_CCtx_refPrefix(ress.cctx, ress.dictBuffer, ress.dictBufferSize) );
1060     } else {
1061         CHECK( ZSTD_CCtx_loadDictionary(ress.cctx, ress.dictBuffer, ress.dictBufferSize) );
1062     }
1063 
1064     return ress;
1065 }
1066 
FIO_freeCResources(const cRess_t * const ress)1067 static void FIO_freeCResources(const cRess_t* const ress)
1068 {
1069     free(ress->srcBuffer);
1070     free(ress->dstBuffer);
1071     free(ress->dictBuffer);
1072     ZSTD_freeCStream(ress->cctx);   /* never fails */
1073 }
1074 
1075 
1076 #ifdef ZSTD_GZCOMPRESS
1077 static unsigned long long
FIO_compressGzFrame(const cRess_t * ress,const char * srcFileName,U64 const srcFileSize,int compressionLevel,U64 * readsize)1078 FIO_compressGzFrame(const cRess_t* ress,  /* buffers & handlers are used, but not changed */
1079                     const char* srcFileName, U64 const srcFileSize,
1080                     int compressionLevel, U64* readsize)
1081 {
1082     unsigned long long inFileSize = 0, outFileSize = 0;
1083     z_stream strm;
1084 
1085     if (compressionLevel > Z_BEST_COMPRESSION)
1086         compressionLevel = Z_BEST_COMPRESSION;
1087 
1088     strm.zalloc = Z_NULL;
1089     strm.zfree = Z_NULL;
1090     strm.opaque = Z_NULL;
1091 
1092     {   int const ret = deflateInit2(&strm, compressionLevel, Z_DEFLATED,
1093                         15 /* maxWindowLogSize */ + 16 /* gzip only */,
1094                         8, Z_DEFAULT_STRATEGY); /* see http://www.zlib.net/manual.html */
1095         if (ret != Z_OK) {
1096             EXM_THROW(71, "zstd: %s: deflateInit2 error %d \n", srcFileName, ret);
1097     }   }
1098 
1099     strm.next_in = 0;
1100     strm.avail_in = 0;
1101     strm.next_out = (Bytef*)ress->dstBuffer;
1102     strm.avail_out = (uInt)ress->dstBufferSize;
1103 
1104     while (1) {
1105         int ret;
1106         if (strm.avail_in == 0) {
1107             size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile);
1108             if (inSize == 0) break;
1109             inFileSize += inSize;
1110             strm.next_in = (z_const unsigned char*)ress->srcBuffer;
1111             strm.avail_in = (uInt)inSize;
1112         }
1113         ret = deflate(&strm, Z_NO_FLUSH);
1114         if (ret != Z_OK)
1115             EXM_THROW(72, "zstd: %s: deflate error %d \n", srcFileName, ret);
1116         {   size_t const cSize = ress->dstBufferSize - strm.avail_out;
1117             if (cSize) {
1118                 if (fwrite(ress->dstBuffer, 1, cSize, ress->dstFile) != cSize)
1119                     EXM_THROW(73, "Write error : cannot write to output file : %s ", strerror(errno));
1120                 outFileSize += cSize;
1121                 strm.next_out = (Bytef*)ress->dstBuffer;
1122                 strm.avail_out = (uInt)ress->dstBufferSize;
1123         }   }
1124         if (srcFileSize == UTIL_FILESIZE_UNKNOWN) {
1125             DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%% ",
1126                             (unsigned)(inFileSize>>20),
1127                             (double)outFileSize/inFileSize*100)
1128         } else {
1129             DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%% ",
1130                             (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
1131                             (double)outFileSize/inFileSize*100);
1132     }   }
1133 
1134     while (1) {
1135         int const ret = deflate(&strm, Z_FINISH);
1136         {   size_t const cSize = ress->dstBufferSize - strm.avail_out;
1137             if (cSize) {
1138                 if (fwrite(ress->dstBuffer, 1, cSize, ress->dstFile) != cSize)
1139                     EXM_THROW(75, "Write error : %s ", strerror(errno));
1140                 outFileSize += cSize;
1141                 strm.next_out = (Bytef*)ress->dstBuffer;
1142                 strm.avail_out = (uInt)ress->dstBufferSize;
1143         }   }
1144         if (ret == Z_STREAM_END) break;
1145         if (ret != Z_BUF_ERROR)
1146             EXM_THROW(77, "zstd: %s: deflate error %d \n", srcFileName, ret);
1147     }
1148 
1149     {   int const ret = deflateEnd(&strm);
1150         if (ret != Z_OK) {
1151             EXM_THROW(79, "zstd: %s: deflateEnd error %d \n", srcFileName, ret);
1152     }   }
1153     *readsize = inFileSize;
1154     return outFileSize;
1155 }
1156 #endif
1157 
1158 
1159 #ifdef ZSTD_LZMACOMPRESS
1160 static unsigned long long
FIO_compressLzmaFrame(cRess_t * ress,const char * srcFileName,U64 const srcFileSize,int compressionLevel,U64 * readsize,int plain_lzma)1161 FIO_compressLzmaFrame(cRess_t* ress,
1162                       const char* srcFileName, U64 const srcFileSize,
1163                       int compressionLevel, U64* readsize, int plain_lzma)
1164 {
1165     unsigned long long inFileSize = 0, outFileSize = 0;
1166     lzma_stream strm = LZMA_STREAM_INIT;
1167     lzma_action action = LZMA_RUN;
1168     lzma_ret ret;
1169 
1170     if (compressionLevel < 0) compressionLevel = 0;
1171     if (compressionLevel > 9) compressionLevel = 9;
1172 
1173     if (plain_lzma) {
1174         lzma_options_lzma opt_lzma;
1175         if (lzma_lzma_preset(&opt_lzma, compressionLevel))
1176             EXM_THROW(81, "zstd: %s: lzma_lzma_preset error", srcFileName);
1177         ret = lzma_alone_encoder(&strm, &opt_lzma); /* LZMA */
1178         if (ret != LZMA_OK)
1179             EXM_THROW(82, "zstd: %s: lzma_alone_encoder error %d", srcFileName, ret);
1180     } else {
1181         ret = lzma_easy_encoder(&strm, compressionLevel, LZMA_CHECK_CRC64); /* XZ */
1182         if (ret != LZMA_OK)
1183             EXM_THROW(83, "zstd: %s: lzma_easy_encoder error %d", srcFileName, ret);
1184     }
1185 
1186     strm.next_in = 0;
1187     strm.avail_in = 0;
1188     strm.next_out = (BYTE*)ress->dstBuffer;
1189     strm.avail_out = ress->dstBufferSize;
1190 
1191     while (1) {
1192         if (strm.avail_in == 0) {
1193             size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile);
1194             if (inSize == 0) action = LZMA_FINISH;
1195             inFileSize += inSize;
1196             strm.next_in = (BYTE const*)ress->srcBuffer;
1197             strm.avail_in = inSize;
1198         }
1199 
1200         ret = lzma_code(&strm, action);
1201 
1202         if (ret != LZMA_OK && ret != LZMA_STREAM_END)
1203             EXM_THROW(84, "zstd: %s: lzma_code encoding error %d", srcFileName, ret);
1204         {   size_t const compBytes = ress->dstBufferSize - strm.avail_out;
1205             if (compBytes) {
1206                 if (fwrite(ress->dstBuffer, 1, compBytes, ress->dstFile) != compBytes)
1207                     EXM_THROW(85, "Write error : %s", strerror(errno));
1208                 outFileSize += compBytes;
1209                 strm.next_out = (BYTE*)ress->dstBuffer;
1210                 strm.avail_out = ress->dstBufferSize;
1211         }   }
1212         if (srcFileSize == UTIL_FILESIZE_UNKNOWN)
1213             DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%",
1214                             (unsigned)(inFileSize>>20),
1215                             (double)outFileSize/inFileSize*100)
1216         else
1217             DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%",
1218                             (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
1219                             (double)outFileSize/inFileSize*100);
1220         if (ret == LZMA_STREAM_END) break;
1221     }
1222 
1223     lzma_end(&strm);
1224     *readsize = inFileSize;
1225 
1226     return outFileSize;
1227 }
1228 #endif
1229 
1230 #ifdef ZSTD_LZ4COMPRESS
1231 
1232 #if LZ4_VERSION_NUMBER <= 10600
1233 #define LZ4F_blockLinked blockLinked
1234 #define LZ4F_max64KB max64KB
1235 #endif
1236 
FIO_LZ4_GetBlockSize_FromBlockId(int id)1237 static int FIO_LZ4_GetBlockSize_FromBlockId (int id) { return (1 << (8 + (2 * id))); }
1238 
1239 static unsigned long long
FIO_compressLz4Frame(cRess_t * ress,const char * srcFileName,U64 const srcFileSize,int compressionLevel,int checksumFlag,U64 * readsize)1240 FIO_compressLz4Frame(cRess_t* ress,
1241                      const char* srcFileName, U64 const srcFileSize,
1242                      int compressionLevel, int checksumFlag,
1243                      U64* readsize)
1244 {
1245     const size_t blockSize = FIO_LZ4_GetBlockSize_FromBlockId(LZ4F_max64KB);
1246     unsigned long long inFileSize = 0, outFileSize = 0;
1247 
1248     LZ4F_preferences_t prefs;
1249     LZ4F_compressionContext_t ctx;
1250 
1251     LZ4F_errorCode_t const errorCode = LZ4F_createCompressionContext(&ctx, LZ4F_VERSION);
1252     if (LZ4F_isError(errorCode))
1253         EXM_THROW(31, "zstd: failed to create lz4 compression context");
1254 
1255     memset(&prefs, 0, sizeof(prefs));
1256 
1257     assert(blockSize <= ress->srcBufferSize);
1258 
1259     prefs.autoFlush = 1;
1260     prefs.compressionLevel = compressionLevel;
1261     prefs.frameInfo.blockMode = LZ4F_blockLinked;
1262     prefs.frameInfo.blockSizeID = LZ4F_max64KB;
1263     prefs.frameInfo.contentChecksumFlag = (contentChecksum_t)checksumFlag;
1264 #if LZ4_VERSION_NUMBER >= 10600
1265     prefs.frameInfo.contentSize = (srcFileSize==UTIL_FILESIZE_UNKNOWN) ? 0 : srcFileSize;
1266 #endif
1267     assert(LZ4F_compressBound(blockSize, &prefs) <= ress->dstBufferSize);
1268 
1269     {
1270         size_t readSize;
1271         size_t headerSize = LZ4F_compressBegin(ctx, ress->dstBuffer, ress->dstBufferSize, &prefs);
1272         if (LZ4F_isError(headerSize))
1273             EXM_THROW(33, "File header generation failed : %s",
1274                             LZ4F_getErrorName(headerSize));
1275         if (fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile) != headerSize)
1276             EXM_THROW(34, "Write error : %s (cannot write header)", strerror(errno));
1277         outFileSize += headerSize;
1278 
1279         /* Read first block */
1280         readSize  = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile);
1281         inFileSize += readSize;
1282 
1283         /* Main Loop */
1284         while (readSize>0) {
1285             size_t const outSize = LZ4F_compressUpdate(ctx,
1286                                         ress->dstBuffer, ress->dstBufferSize,
1287                                         ress->srcBuffer, readSize, NULL);
1288             if (LZ4F_isError(outSize))
1289                 EXM_THROW(35, "zstd: %s: lz4 compression failed : %s",
1290                             srcFileName, LZ4F_getErrorName(outSize));
1291             outFileSize += outSize;
1292             if (srcFileSize == UTIL_FILESIZE_UNKNOWN) {
1293                 DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%",
1294                                 (unsigned)(inFileSize>>20),
1295                                 (double)outFileSize/inFileSize*100)
1296             } else {
1297                 DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%",
1298                                 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
1299                                 (double)outFileSize/inFileSize*100);
1300             }
1301 
1302             /* Write Block */
1303             {   size_t const sizeCheck = fwrite(ress->dstBuffer, 1, outSize, ress->dstFile);
1304                 if (sizeCheck != outSize)
1305                     EXM_THROW(36, "Write error : %s", strerror(errno));
1306             }
1307 
1308             /* Read next block */
1309             readSize  = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile);
1310             inFileSize += readSize;
1311         }
1312         if (ferror(ress->srcFile)) EXM_THROW(37, "Error reading %s ", srcFileName);
1313 
1314         /* End of Stream mark */
1315         headerSize = LZ4F_compressEnd(ctx, ress->dstBuffer, ress->dstBufferSize, NULL);
1316         if (LZ4F_isError(headerSize))
1317             EXM_THROW(38, "zstd: %s: lz4 end of file generation failed : %s",
1318                         srcFileName, LZ4F_getErrorName(headerSize));
1319 
1320         {   size_t const sizeCheck = fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile);
1321             if (sizeCheck != headerSize)
1322                 EXM_THROW(39, "Write error : %s (cannot write end of stream)",
1323                             strerror(errno));
1324         }
1325         outFileSize += headerSize;
1326     }
1327 
1328     *readsize = inFileSize;
1329     LZ4F_freeCompressionContext(ctx);
1330 
1331     return outFileSize;
1332 }
1333 #endif
1334 
1335 
1336 static unsigned long long
FIO_compressZstdFrame(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,const cRess_t * ressPtr,const char * srcFileName,U64 fileSize,int compressionLevel,U64 * readsize)1337 FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
1338                       FIO_prefs_t* const prefs,
1339                       const cRess_t* ressPtr,
1340                       const char* srcFileName, U64 fileSize,
1341                       int compressionLevel, U64* readsize)
1342 {
1343     cRess_t const ress = *ressPtr;
1344     FILE* const srcFile = ress.srcFile;
1345     FILE* const dstFile = ress.dstFile;
1346     U64 compressedfilesize = 0;
1347     ZSTD_EndDirective directive = ZSTD_e_continue;
1348     U64 pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;
1349 
1350     /* stats */
1351     ZSTD_frameProgression previous_zfp_update = { 0, 0, 0, 0, 0, 0 };
1352     ZSTD_frameProgression previous_zfp_correction = { 0, 0, 0, 0, 0, 0 };
1353     typedef enum { noChange, slower, faster } speedChange_e;
1354     speedChange_e speedChange = noChange;
1355     unsigned flushWaiting = 0;
1356     unsigned inputPresented = 0;
1357     unsigned inputBlocked = 0;
1358     unsigned lastJobID = 0;
1359     UTIL_HumanReadableSize_t const file_hrs = UTIL_makeHumanReadableSize(fileSize);
1360 
1361     DISPLAYLEVEL(6, "compression using zstd format \n");
1362 
1363     /* init */
1364     if (fileSize != UTIL_FILESIZE_UNKNOWN) {
1365         pledgedSrcSize = fileSize;
1366         CHECK(ZSTD_CCtx_setPledgedSrcSize(ress.cctx, fileSize));
1367     } else if (prefs->streamSrcSize > 0) {
1368       /* unknown source size; use the declared stream size */
1369       pledgedSrcSize = prefs->streamSrcSize;
1370       CHECK( ZSTD_CCtx_setPledgedSrcSize(ress.cctx, prefs->streamSrcSize) );
1371     }
1372 
1373     {
1374         int windowLog;
1375         UTIL_HumanReadableSize_t windowSize;
1376         CHECK(ZSTD_CCtx_getParameter(ress.cctx, ZSTD_c_windowLog, &windowLog));
1377         if (windowLog == 0) {
1378             const ZSTD_compressionParameters cParams = ZSTD_getCParams(compressionLevel, fileSize, 0);
1379             windowLog = cParams.windowLog;
1380         }
1381         windowSize = UTIL_makeHumanReadableSize(MAX(1ULL, MIN(1ULL << windowLog, pledgedSrcSize)));
1382         DISPLAYLEVEL(4, "Decompression will require %.*f%s of memory\n", windowSize.precision, windowSize.value, windowSize.suffix);
1383     }
1384     (void)srcFileName;
1385 
1386     /* Main compression loop */
1387     do {
1388         size_t stillToFlush;
1389         /* Fill input Buffer */
1390         size_t const inSize = fread(ress.srcBuffer, (size_t)1, ress.srcBufferSize, srcFile);
1391         ZSTD_inBuffer inBuff = { ress.srcBuffer, inSize, 0 };
1392         DISPLAYLEVEL(6, "fread %u bytes from source \n", (unsigned)inSize);
1393         *readsize += inSize;
1394 
1395         if ((inSize == 0) || (*readsize == fileSize))
1396             directive = ZSTD_e_end;
1397 
1398         stillToFlush = 1;
1399         while ((inBuff.pos != inBuff.size)   /* input buffer must be entirely ingested */
1400             || (directive == ZSTD_e_end && stillToFlush != 0) ) {
1401 
1402             size_t const oldIPos = inBuff.pos;
1403             ZSTD_outBuffer outBuff = { ress.dstBuffer, ress.dstBufferSize, 0 };
1404             size_t const toFlushNow = ZSTD_toFlushNow(ress.cctx);
1405             CHECK_V(stillToFlush, ZSTD_compressStream2(ress.cctx, &outBuff, &inBuff, directive));
1406 
1407             /* count stats */
1408             inputPresented++;
1409             if (oldIPos == inBuff.pos) inputBlocked++;  /* input buffer is full and can't take any more : input speed is faster than consumption rate */
1410             if (!toFlushNow) flushWaiting = 1;
1411 
1412             /* Write compressed stream */
1413             DISPLAYLEVEL(6, "ZSTD_compress_generic(end:%u) => input pos(%u)<=(%u)size ; output generated %u bytes \n",
1414                             (unsigned)directive, (unsigned)inBuff.pos, (unsigned)inBuff.size, (unsigned)outBuff.pos);
1415             if (outBuff.pos) {
1416                 size_t const sizeCheck = fwrite(ress.dstBuffer, 1, outBuff.pos, dstFile);
1417                 if (sizeCheck != outBuff.pos)
1418                     EXM_THROW(25, "Write error : %s (cannot write compressed block)",
1419                                     strerror(errno));
1420                 compressedfilesize += outBuff.pos;
1421             }
1422 
1423             /* display notification; and adapt compression level */
1424             if (READY_FOR_UPDATE()) {
1425                 ZSTD_frameProgression const zfp = ZSTD_getFrameProgression(ress.cctx);
1426                 double const cShare = (double)zfp.produced / (double)(zfp.consumed + !zfp.consumed/*avoid div0*/) * 100;
1427                 UTIL_HumanReadableSize_t const buffered_hrs = UTIL_makeHumanReadableSize(zfp.ingested - zfp.consumed);
1428                 UTIL_HumanReadableSize_t const consumed_hrs = UTIL_makeHumanReadableSize(zfp.consumed);
1429                 UTIL_HumanReadableSize_t const produced_hrs = UTIL_makeHumanReadableSize(zfp.produced);
1430 
1431                 /* display progress notifications */
1432                 if (g_display_prefs.displayLevel >= 3) {
1433                     DISPLAYUPDATE(3, "\r(L%i) Buffered :%6.*f%4s - Consumed :%6.*f%4s - Compressed :%6.*f%4s => %.2f%% ",
1434                                 compressionLevel,
1435                                 buffered_hrs.precision, buffered_hrs.value, buffered_hrs.suffix,
1436                                 consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix,
1437                                 produced_hrs.precision, produced_hrs.value, produced_hrs.suffix,
1438                                 cShare );
1439                 } else if (g_display_prefs.displayLevel >= 2 || g_display_prefs.progressSetting == FIO_ps_always) {
1440                     /* Require level 2 or forcibly displayed progress counter for summarized updates */
1441                     DISPLAYLEVEL(1, "\r%79s\r", "");    /* Clear out the current displayed line */
1442                     if (fCtx->nbFilesTotal > 1) {
1443                         size_t srcFileNameSize = strlen(srcFileName);
1444                         /* Ensure that the string we print is roughly the same size each time */
1445                         if (srcFileNameSize > 18) {
1446                             const char* truncatedSrcFileName = srcFileName + srcFileNameSize - 15;
1447                             DISPLAYLEVEL(1, "Compress: %u/%u files. Current: ...%s ",
1448                                         fCtx->currFileIdx+1, fCtx->nbFilesTotal, truncatedSrcFileName);
1449                         } else {
1450                             DISPLAYLEVEL(1, "Compress: %u/%u files. Current: %*s ",
1451                                         fCtx->currFileIdx+1, fCtx->nbFilesTotal, (int)(18-srcFileNameSize), srcFileName);
1452                         }
1453                     }
1454                     DISPLAYLEVEL(1, "Read:%6.*f%4s ", consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix);
1455                     if (fileSize != UTIL_FILESIZE_UNKNOWN)
1456                         DISPLAYLEVEL(2, "/%6.*f%4s", file_hrs.precision, file_hrs.value, file_hrs.suffix);
1457                     DISPLAYLEVEL(1, " ==> %2.f%%", cShare);
1458                     DELAY_NEXT_UPDATE();
1459                 }
1460 
1461                 /* adaptive mode : statistics measurement and speed correction */
1462                 if (prefs->adaptiveMode) {
1463 
1464                     /* check output speed */
1465                     if (zfp.currentJobID > 1) {  /* only possible if nbWorkers >= 1 */
1466 
1467                         unsigned long long newlyProduced = zfp.produced - previous_zfp_update.produced;
1468                         unsigned long long newlyFlushed = zfp.flushed - previous_zfp_update.flushed;
1469                         assert(zfp.produced >= previous_zfp_update.produced);
1470                         assert(prefs->nbWorkers >= 1);
1471 
1472                         /* test if compression is blocked
1473                          * either because output is slow and all buffers are full
1474                          * or because input is slow and no job can start while waiting for at least one buffer to be filled.
1475                          * note : exclude starting part, since currentJobID > 1 */
1476                         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)*/
1477                           && (zfp.nbActiveWorkers == 0)                       /* confirmed : no compression ongoing */
1478                           ) {
1479                             DISPLAYLEVEL(6, "all buffers full : compression stopped => slow down \n")
1480                             speedChange = slower;
1481                         }
1482 
1483                         previous_zfp_update = zfp;
1484 
1485                         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) */
1486                           && (flushWaiting == 0)                        /* flush speed was never slowed by lack of production, so it's operating at max capacity */
1487                           ) {
1488                             DISPLAYLEVEL(6, "compression faster than flush (%llu > %llu), and flushed was never slowed down by lack of production => slow down \n", newlyProduced, newlyFlushed);
1489                             speedChange = slower;
1490                         }
1491                         flushWaiting = 0;
1492                     }
1493 
1494                     /* course correct only if there is at least one new job completed */
1495                     if (zfp.currentJobID > lastJobID) {
1496                         DISPLAYLEVEL(6, "compression level adaptation check \n")
1497 
1498                         /* check input speed */
1499                         if (zfp.currentJobID > (unsigned)(prefs->nbWorkers+1)) {   /* warm up period, to fill all workers */
1500                             if (inputBlocked <= 0) {
1501                                 DISPLAYLEVEL(6, "input is never blocked => input is slower than ingestion \n");
1502                                 speedChange = slower;
1503                             } else if (speedChange == noChange) {
1504                                 unsigned long long newlyIngested = zfp.ingested - previous_zfp_correction.ingested;
1505                                 unsigned long long newlyConsumed = zfp.consumed - previous_zfp_correction.consumed;
1506                                 unsigned long long newlyProduced = zfp.produced - previous_zfp_correction.produced;
1507                                 unsigned long long newlyFlushed  = zfp.flushed  - previous_zfp_correction.flushed;
1508                                 previous_zfp_correction = zfp;
1509                                 assert(inputPresented > 0);
1510                                 DISPLAYLEVEL(6, "input blocked %u/%u(%.2f) - ingested:%u vs %u:consumed - flushed:%u vs %u:produced \n",
1511                                                 inputBlocked, inputPresented, (double)inputBlocked/inputPresented*100,
1512                                                 (unsigned)newlyIngested, (unsigned)newlyConsumed,
1513                                                 (unsigned)newlyFlushed, (unsigned)newlyProduced);
1514                                 if ( (inputBlocked > inputPresented / 8)     /* input is waiting often, because input buffers is full : compression or output too slow */
1515                                   && (newlyFlushed * 33 / 32 > newlyProduced)  /* flush everything that is produced */
1516                                   && (newlyIngested * 33 / 32 > newlyConsumed) /* input speed as fast or faster than compression speed */
1517                                 ) {
1518                                     DISPLAYLEVEL(6, "recommend faster as in(%llu) >= (%llu)comp(%llu) <= out(%llu) \n",
1519                                                     newlyIngested, newlyConsumed, newlyProduced, newlyFlushed);
1520                                     speedChange = faster;
1521                                 }
1522                             }
1523                             inputBlocked = 0;
1524                             inputPresented = 0;
1525                         }
1526 
1527                         if (speedChange == slower) {
1528                             DISPLAYLEVEL(6, "slower speed , higher compression \n")
1529                             compressionLevel ++;
1530                             if (compressionLevel > ZSTD_maxCLevel()) compressionLevel = ZSTD_maxCLevel();
1531                             if (compressionLevel > prefs->maxAdaptLevel) compressionLevel = prefs->maxAdaptLevel;
1532                             compressionLevel += (compressionLevel == 0);   /* skip 0 */
1533                             ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
1534                         }
1535                         if (speedChange == faster) {
1536                             DISPLAYLEVEL(6, "faster speed , lighter compression \n")
1537                             compressionLevel --;
1538                             if (compressionLevel < prefs->minAdaptLevel) compressionLevel = prefs->minAdaptLevel;
1539                             compressionLevel -= (compressionLevel == 0);   /* skip 0 */
1540                             ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
1541                         }
1542                         speedChange = noChange;
1543 
1544                         lastJobID = zfp.currentJobID;
1545                     }  /* if (zfp.currentJobID > lastJobID) */
1546                 }  /* if (g_adaptiveMode) */
1547             }  /* if (READY_FOR_UPDATE()) */
1548         }  /* while ((inBuff.pos != inBuff.size) */
1549     } while (directive != ZSTD_e_end);
1550 
1551     if (ferror(srcFile)) {
1552         EXM_THROW(26, "Read error : I/O error");
1553     }
1554     if (fileSize != UTIL_FILESIZE_UNKNOWN && *readsize != fileSize) {
1555         EXM_THROW(27, "Read error : Incomplete read : %llu / %llu B",
1556                 (unsigned long long)*readsize, (unsigned long long)fileSize);
1557     }
1558 
1559     return compressedfilesize;
1560 }
1561 
1562 /*! FIO_compressFilename_internal() :
1563  *  same as FIO_compressFilename_extRess(), with `ress.desFile` already opened.
1564  *  @return : 0 : compression completed correctly,
1565  *            1 : missing or pb opening srcFileName
1566  */
1567 static int
FIO_compressFilename_internal(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,cRess_t ress,const char * dstFileName,const char * srcFileName,int compressionLevel)1568 FIO_compressFilename_internal(FIO_ctx_t* const fCtx,
1569                               FIO_prefs_t* const prefs,
1570                               cRess_t ress,
1571                               const char* dstFileName, const char* srcFileName,
1572                               int compressionLevel)
1573 {
1574     UTIL_time_t const timeStart = UTIL_getTime();
1575     clock_t const cpuStart = clock();
1576     U64 readsize = 0;
1577     U64 compressedfilesize = 0;
1578     U64 const fileSize = UTIL_getFileSize(srcFileName);
1579     DISPLAYLEVEL(5, "%s: %llu bytes \n", srcFileName, (unsigned long long)fileSize);
1580 
1581     /* compression format selection */
1582     switch (prefs->compressionType) {
1583         default:
1584         case FIO_zstdCompression:
1585             compressedfilesize = FIO_compressZstdFrame(fCtx, prefs, &ress, srcFileName, fileSize, compressionLevel, &readsize);
1586             break;
1587 
1588         case FIO_gzipCompression:
1589 #ifdef ZSTD_GZCOMPRESS
1590             compressedfilesize = FIO_compressGzFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize);
1591 #else
1592             (void)compressionLevel;
1593             EXM_THROW(20, "zstd: %s: file cannot be compressed as gzip (zstd compiled without ZSTD_GZCOMPRESS) -- ignored \n",
1594                             srcFileName);
1595 #endif
1596             break;
1597 
1598         case FIO_xzCompression:
1599         case FIO_lzmaCompression:
1600 #ifdef ZSTD_LZMACOMPRESS
1601             compressedfilesize = FIO_compressLzmaFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize, prefs->compressionType==FIO_lzmaCompression);
1602 #else
1603             (void)compressionLevel;
1604             EXM_THROW(20, "zstd: %s: file cannot be compressed as xz/lzma (zstd compiled without ZSTD_LZMACOMPRESS) -- ignored \n",
1605                             srcFileName);
1606 #endif
1607             break;
1608 
1609         case FIO_lz4Compression:
1610 #ifdef ZSTD_LZ4COMPRESS
1611             compressedfilesize = FIO_compressLz4Frame(&ress, srcFileName, fileSize, compressionLevel, prefs->checksumFlag, &readsize);
1612 #else
1613             (void)compressionLevel;
1614             EXM_THROW(20, "zstd: %s: file cannot be compressed as lz4 (zstd compiled without ZSTD_LZ4COMPRESS) -- ignored \n",
1615                             srcFileName);
1616 #endif
1617             break;
1618     }
1619 
1620     /* Status */
1621     fCtx->totalBytesInput += (size_t)readsize;
1622     fCtx->totalBytesOutput += (size_t)compressedfilesize;
1623     DISPLAYLEVEL(2, "\r%79s\r", "");
1624     if (g_display_prefs.displayLevel >= 2 &&
1625         !fCtx->hasStdoutOutput &&
1626         (g_display_prefs.displayLevel >= 3 || fCtx->nbFilesTotal <= 1)) {
1627         UTIL_HumanReadableSize_t hr_isize = UTIL_makeHumanReadableSize((U64) readsize);
1628         UTIL_HumanReadableSize_t hr_osize = UTIL_makeHumanReadableSize((U64) compressedfilesize);
1629         if (readsize == 0) {
1630             DISPLAYLEVEL(2,"%-20s :  (%6.*f%4s => %6.*f%4s, %s) \n",
1631                 srcFileName,
1632                 hr_isize.precision, hr_isize.value, hr_isize.suffix,
1633                 hr_osize.precision, hr_osize.value, hr_osize.suffix,
1634                 dstFileName);
1635         } else {
1636             DISPLAYLEVEL(2,"%-20s :%6.2f%%   (%6.*f%4s => %6.*f%4s, %s) \n",
1637                 srcFileName,
1638                 (double)compressedfilesize / (double)readsize * 100,
1639                 hr_isize.precision, hr_isize.value, hr_isize.suffix,
1640                 hr_osize.precision, hr_osize.value, hr_osize.suffix,
1641                 dstFileName);
1642         }
1643     }
1644 
1645     /* Elapsed Time and CPU Load */
1646     {   clock_t const cpuEnd = clock();
1647         double const cpuLoad_s = (double)(cpuEnd - cpuStart) / CLOCKS_PER_SEC;
1648         U64 const timeLength_ns = UTIL_clockSpanNano(timeStart);
1649         double const timeLength_s = (double)timeLength_ns / 1000000000;
1650         double const cpuLoad_pct = (cpuLoad_s / timeLength_s) * 100;
1651         DISPLAYLEVEL(4, "%-20s : Completed in %.2f sec  (cpu load : %.0f%%)\n",
1652                         srcFileName, timeLength_s, cpuLoad_pct);
1653     }
1654     return 0;
1655 }
1656 
1657 
1658 /*! FIO_compressFilename_dstFile() :
1659  *  open dstFileName, or pass-through if ress.dstFile != NULL,
1660  *  then start compression with FIO_compressFilename_internal().
1661  *  Manages source removal (--rm) and file permissions transfer.
1662  *  note : ress.srcFile must be != NULL,
1663  *  so reach this function through FIO_compressFilename_srcFile().
1664  *  @return : 0 : compression completed correctly,
1665  *            1 : pb
1666  */
FIO_compressFilename_dstFile(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,cRess_t ress,const char * dstFileName,const char * srcFileName,int compressionLevel)1667 static int FIO_compressFilename_dstFile(FIO_ctx_t* const fCtx,
1668                                         FIO_prefs_t* const prefs,
1669                                         cRess_t ress,
1670                                         const char* dstFileName,
1671                                         const char* srcFileName,
1672                                         int compressionLevel)
1673 {
1674     int closeDstFile = 0;
1675     int result;
1676     stat_t statbuf;
1677     int transferMTime = 0;
1678     assert(ress.srcFile != NULL);
1679     if (ress.dstFile == NULL) {
1680         int dstFilePermissions = DEFAULT_FILE_PERMISSIONS;
1681         if ( strcmp (srcFileName, stdinmark)
1682           && strcmp (dstFileName, stdoutmark)
1683           && UTIL_stat(srcFileName, &statbuf)
1684           && UTIL_isRegularFileStat(&statbuf) ) {
1685             dstFilePermissions = statbuf.st_mode;
1686             transferMTime = 1;
1687         }
1688 
1689         closeDstFile = 1;
1690         DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: opening dst: %s \n", dstFileName);
1691         ress.dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFilePermissions);
1692         if (ress.dstFile==NULL) return 1;  /* could not open dstFileName */
1693         /* Must only be added after FIO_openDstFile() succeeds.
1694          * Otherwise we may delete the destination file if it already exists,
1695          * and the user presses Ctrl-C when asked if they wish to overwrite.
1696          */
1697         addHandler(dstFileName);
1698     }
1699 
1700     result = FIO_compressFilename_internal(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
1701 
1702     if (closeDstFile) {
1703         FILE* const dstFile = ress.dstFile;
1704         ress.dstFile = NULL;
1705 
1706         clearHandler();
1707 
1708         DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: closing dst: %s \n", dstFileName);
1709         if (fclose(dstFile)) { /* error closing dstFile */
1710             DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno));
1711             result=1;
1712         }
1713         if (transferMTime) {
1714             UTIL_utime(dstFileName, &statbuf);
1715         }
1716         if ( (result != 0)  /* operation failure */
1717           && strcmp(dstFileName, stdoutmark)  /* special case : don't remove() stdout */
1718           ) {
1719             FIO_removeFile(dstFileName); /* remove compression artefact; note don't do anything special if remove() fails */
1720         }
1721     }
1722 
1723     return result;
1724 }
1725 
1726 /* List used to compare file extensions (used with --exclude-compressed flag)
1727 * Different from the suffixList and should only apply to ZSTD compress operationResult
1728 */
1729 static const char *compressedFileExtensions[] = {
1730     ZSTD_EXTENSION,
1731     TZSTD_EXTENSION,
1732     GZ_EXTENSION,
1733     TGZ_EXTENSION,
1734     LZMA_EXTENSION,
1735     XZ_EXTENSION,
1736     TXZ_EXTENSION,
1737     LZ4_EXTENSION,
1738     TLZ4_EXTENSION,
1739     NULL
1740 };
1741 
1742 /*! FIO_compressFilename_srcFile() :
1743  *  @return : 0 : compression completed correctly,
1744  *            1 : missing or pb opening srcFileName
1745  */
1746 static int
FIO_compressFilename_srcFile(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,cRess_t ress,const char * dstFileName,const char * srcFileName,int compressionLevel)1747 FIO_compressFilename_srcFile(FIO_ctx_t* const fCtx,
1748                              FIO_prefs_t* const prefs,
1749                              cRess_t ress,
1750                              const char* dstFileName,
1751                              const char* srcFileName,
1752                              int compressionLevel)
1753 {
1754     int result;
1755     DISPLAYLEVEL(6, "FIO_compressFilename_srcFile: %s \n", srcFileName);
1756 
1757     /* ensure src is not a directory */
1758     if (UTIL_isDirectory(srcFileName)) {
1759         DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName);
1760         return 1;
1761     }
1762 
1763     /* ensure src is not the same as dict (if present) */
1764     if (ress.dictFileName != NULL && UTIL_isSameFile(srcFileName, ress.dictFileName)) {
1765         DISPLAYLEVEL(1, "zstd: cannot use %s as an input file and dictionary \n", srcFileName);
1766         return 1;
1767     }
1768 
1769     /* Check if "srcFile" is compressed. Only done if --exclude-compressed flag is used
1770     * YES => ZSTD will skip compression of the file and will return 0.
1771     * NO => ZSTD will resume with compress operation.
1772     */
1773     if (prefs->excludeCompressedFiles == 1 && UTIL_isCompressedFile(srcFileName, compressedFileExtensions)) {
1774         DISPLAYLEVEL(4, "File is already compressed : %s \n", srcFileName);
1775         return 0;
1776     }
1777 
1778     ress.srcFile = FIO_openSrcFile(prefs, srcFileName);
1779     if (ress.srcFile == NULL) return 1;   /* srcFile could not be opened */
1780 
1781     result = FIO_compressFilename_dstFile(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
1782 
1783     fclose(ress.srcFile);
1784     ress.srcFile = NULL;
1785     if ( prefs->removeSrcFile   /* --rm */
1786       && result == 0       /* success */
1787       && strcmp(srcFileName, stdinmark)   /* exception : don't erase stdin */
1788       ) {
1789         /* We must clear the handler, since after this point calling it would
1790          * delete both the source and destination files.
1791          */
1792         clearHandler();
1793         if (FIO_removeFile(srcFileName))
1794             EXM_THROW(1, "zstd: %s: %s", srcFileName, strerror(errno));
1795     }
1796     return result;
1797 }
1798 
checked_index(const char * options[],size_t length,size_t index)1799 static const char* checked_index(const char* options[], size_t length, size_t index) {
1800     assert(index < length);
1801     // Necessary to avoid warnings since -O3 will omit the above `assert`
1802     (void) length;
1803     return options[index];
1804 }
1805 
1806 #define INDEX(options, index) checked_index((options), sizeof(options)  / sizeof(char*), (index))
1807 
FIO_displayCompressionParameters(const FIO_prefs_t * prefs)1808 void FIO_displayCompressionParameters(const FIO_prefs_t* prefs) {
1809     static const char* formatOptions[5] = {ZSTD_EXTENSION, GZ_EXTENSION, XZ_EXTENSION,
1810                                            LZMA_EXTENSION, LZ4_EXTENSION};
1811     static const char* sparseOptions[3] = {" --no-sparse", "", " --sparse"};
1812     static const char* checkSumOptions[3] = {" --no-check", "", " --check"};
1813     static const char* rowMatchFinderOptions[3] = {"", " --no-row-match-finder", " --row-match-finder"};
1814     static const char* compressLiteralsOptions[3] = {"", " --compress-literals", " --no-compress-literals"};
1815 
1816     assert(g_display_prefs.displayLevel >= 4);
1817 
1818     DISPLAY("--format=%s", formatOptions[prefs->compressionType]);
1819     DISPLAY("%s", INDEX(sparseOptions, prefs->sparseFileSupport));
1820     DISPLAY("%s", prefs->dictIDFlag ? "" : " --no-dictID");
1821     DISPLAY("%s", INDEX(checkSumOptions, prefs->checksumFlag));
1822     DISPLAY(" --block-size=%d", prefs->blockSize);
1823     if (prefs->adaptiveMode)
1824         DISPLAY(" --adapt=min=%d,max=%d", prefs->minAdaptLevel, prefs->maxAdaptLevel);
1825     DISPLAY("%s", INDEX(rowMatchFinderOptions, prefs->useRowMatchFinder));
1826     DISPLAY("%s", prefs->rsyncable ? " --rsyncable" : "");
1827     if (prefs->streamSrcSize)
1828         DISPLAY(" --stream-size=%u", (unsigned) prefs->streamSrcSize);
1829     if (prefs->srcSizeHint)
1830         DISPLAY(" --size-hint=%d", prefs->srcSizeHint);
1831     if (prefs->targetCBlockSize)
1832         DISPLAY(" --target-compressed-block-size=%u", (unsigned) prefs->targetCBlockSize);
1833     DISPLAY("%s", INDEX(compressLiteralsOptions, prefs->literalCompressionMode));
1834     DISPLAY(" --memory=%u", prefs->memLimit ? prefs->memLimit : 128 MB);
1835     DISPLAY(" --threads=%d", prefs->nbWorkers);
1836     DISPLAY("%s", prefs->excludeCompressedFiles ? " --exclude-compressed" : "");
1837     DISPLAY(" --%scontent-size", prefs->contentSize ? "" : "no-");
1838     DISPLAY("\n");
1839 }
1840 
1841 #undef INDEX
1842 
FIO_compressFilename(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,const char * dstFileName,const char * srcFileName,const char * dictFileName,int compressionLevel,ZSTD_compressionParameters comprParams)1843 int FIO_compressFilename(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs, const char* dstFileName,
1844                          const char* srcFileName, const char* dictFileName,
1845                          int compressionLevel, ZSTD_compressionParameters comprParams)
1846 {
1847     cRess_t const ress = FIO_createCResources(prefs, dictFileName, UTIL_getFileSize(srcFileName), compressionLevel, comprParams);
1848     int const result = FIO_compressFilename_srcFile(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
1849 
1850 #define DISPLAY_LEVEL_DEFAULT 2
1851 
1852     FIO_freeCResources(&ress);
1853     return result;
1854 }
1855 
1856 /* FIO_determineCompressedName() :
1857  * create a destination filename for compressed srcFileName.
1858  * @return a pointer to it.
1859  * This function never returns an error (it may abort() in case of pb)
1860  */
1861 static const char*
FIO_determineCompressedName(const char * srcFileName,const char * outDirName,const char * suffix)1862 FIO_determineCompressedName(const char* srcFileName, const char* outDirName, const char* suffix)
1863 {
1864     static size_t dfnbCapacity = 0;
1865     static char* dstFileNameBuffer = NULL;   /* using static allocation : this function cannot be multi-threaded */
1866     char* outDirFilename = NULL;
1867     size_t sfnSize = strlen(srcFileName);
1868     size_t const srcSuffixLen = strlen(suffix);
1869     if (outDirName) {
1870         outDirFilename = FIO_createFilename_fromOutDir(srcFileName, outDirName, srcSuffixLen);
1871         sfnSize = strlen(outDirFilename);
1872         assert(outDirFilename != NULL);
1873     }
1874 
1875     if (dfnbCapacity <= sfnSize+srcSuffixLen+1) {
1876         /* resize buffer for dstName */
1877         free(dstFileNameBuffer);
1878         dfnbCapacity = sfnSize + srcSuffixLen + 30;
1879         dstFileNameBuffer = (char*)malloc(dfnbCapacity);
1880         if (!dstFileNameBuffer) {
1881             EXM_THROW(30, "zstd: %s", strerror(errno));
1882         }
1883     }
1884     assert(dstFileNameBuffer != NULL);
1885 
1886     if (outDirFilename) {
1887         memcpy(dstFileNameBuffer, outDirFilename, sfnSize);
1888         free(outDirFilename);
1889     } else {
1890         memcpy(dstFileNameBuffer, srcFileName, sfnSize);
1891     }
1892     memcpy(dstFileNameBuffer+sfnSize, suffix, srcSuffixLen+1 /* Include terminating null */);
1893     return dstFileNameBuffer;
1894 }
1895 
FIO_getLargestFileSize(const char ** inFileNames,unsigned nbFiles)1896 static unsigned long long FIO_getLargestFileSize(const char** inFileNames, unsigned nbFiles)
1897 {
1898     size_t i;
1899     unsigned long long fileSize, maxFileSize = 0;
1900     for (i = 0; i < nbFiles; i++) {
1901         fileSize = UTIL_getFileSize(inFileNames[i]);
1902         maxFileSize = fileSize > maxFileSize ? fileSize : maxFileSize;
1903     }
1904     return maxFileSize;
1905 }
1906 
1907 /* FIO_compressMultipleFilenames() :
1908  * compress nbFiles files
1909  * into either one destination (outFileName),
1910  * or into one file each (outFileName == NULL, but suffix != NULL),
1911  * or into a destination folder (specified with -O)
1912  */
FIO_compressMultipleFilenames(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,const char ** inFileNamesTable,const char * outMirroredRootDirName,const char * outDirName,const char * outFileName,const char * suffix,const char * dictFileName,int compressionLevel,ZSTD_compressionParameters comprParams)1913 int FIO_compressMultipleFilenames(FIO_ctx_t* const fCtx,
1914                                   FIO_prefs_t* const prefs,
1915                                   const char** inFileNamesTable,
1916                                   const char* outMirroredRootDirName,
1917                                   const char* outDirName,
1918                                   const char* outFileName, const char* suffix,
1919                                   const char* dictFileName, int compressionLevel,
1920                                   ZSTD_compressionParameters comprParams)
1921 {
1922     int status;
1923     int error = 0;
1924     cRess_t ress = FIO_createCResources(prefs, dictFileName,
1925         FIO_getLargestFileSize(inFileNamesTable, (unsigned)fCtx->nbFilesTotal),
1926         compressionLevel, comprParams);
1927 
1928     /* init */
1929     assert(outFileName != NULL || suffix != NULL);
1930     if (outFileName != NULL) {   /* output into a single destination (stdout typically) */
1931         if (FIO_removeMultiFilesWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
1932             FIO_freeCResources(&ress);
1933             return 1;
1934         }
1935         ress.dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
1936         if (ress.dstFile == NULL) {  /* could not open outFileName */
1937             error = 1;
1938         } else {
1939             for (; fCtx->currFileIdx < fCtx->nbFilesTotal; ++fCtx->currFileIdx) {
1940                 status = FIO_compressFilename_srcFile(fCtx, prefs, ress, outFileName, inFileNamesTable[fCtx->currFileIdx], compressionLevel);
1941                 if (!status) fCtx->nbFilesProcessed++;
1942                 error |= status;
1943             }
1944             if (fclose(ress.dstFile))
1945                 EXM_THROW(29, "Write error (%s) : cannot properly close %s",
1946                             strerror(errno), outFileName);
1947             ress.dstFile = NULL;
1948         }
1949     } else {
1950         if (outMirroredRootDirName)
1951             UTIL_mirrorSourceFilesDirectories(inFileNamesTable, (unsigned)fCtx->nbFilesTotal, outMirroredRootDirName);
1952 
1953         for (; fCtx->currFileIdx < fCtx->nbFilesTotal; ++fCtx->currFileIdx) {
1954             const char* const srcFileName = inFileNamesTable[fCtx->currFileIdx];
1955             const char* dstFileName = NULL;
1956             if (outMirroredRootDirName) {
1957                 char* validMirroredDirName = UTIL_createMirroredDestDirName(srcFileName, outMirroredRootDirName);
1958                 if (validMirroredDirName) {
1959                     dstFileName = FIO_determineCompressedName(srcFileName, validMirroredDirName, suffix);
1960                     free(validMirroredDirName);
1961                 } else {
1962                     DISPLAYLEVEL(2, "zstd: --output-dir-mirror cannot compress '%s' into '%s' \n", srcFileName, outMirroredRootDirName);
1963                     error=1;
1964                     continue;
1965                 }
1966             } else {
1967                 dstFileName = FIO_determineCompressedName(srcFileName, outDirName, suffix);  /* cannot fail */
1968             }
1969             status = FIO_compressFilename_srcFile(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
1970             if (!status) fCtx->nbFilesProcessed++;
1971             error |= status;
1972         }
1973 
1974         if (outDirName)
1975             FIO_checkFilenameCollisions(inFileNamesTable , (unsigned)fCtx->nbFilesTotal);
1976     }
1977 
1978     if (fCtx->nbFilesProcessed >= 1 && fCtx->nbFilesTotal > 1 && fCtx->totalBytesInput != 0) {
1979         UTIL_HumanReadableSize_t hr_isize = UTIL_makeHumanReadableSize((U64) fCtx->totalBytesInput);
1980         UTIL_HumanReadableSize_t hr_osize = UTIL_makeHumanReadableSize((U64) fCtx->totalBytesOutput);
1981 
1982         DISPLAYLEVEL(2, "\r%79s\r", "");
1983         DISPLAYLEVEL(2, "%3d files compressed :%.2f%%   (%6.*f%4s => %6.*f%4s)\n",
1984                         fCtx->nbFilesProcessed,
1985                         (double)fCtx->totalBytesOutput/((double)fCtx->totalBytesInput)*100,
1986                         hr_isize.precision, hr_isize.value, hr_isize.suffix,
1987                         hr_osize.precision, hr_osize.value, hr_osize.suffix);
1988     }
1989 
1990     FIO_freeCResources(&ress);
1991     return error;
1992 }
1993 
1994 #endif /* #ifndef ZSTD_NOCOMPRESS */
1995 
1996 
1997 
1998 #ifndef ZSTD_NODECOMPRESS
1999 
2000 /* **************************************************************************
2001  *  Decompression
2002  ***************************************************************************/
2003 typedef struct {
2004     void*  srcBuffer;
2005     size_t srcBufferSize;
2006     size_t srcBufferLoaded;
2007     void*  dstBuffer;
2008     size_t dstBufferSize;
2009     ZSTD_DStream* dctx;
2010     FILE*  dstFile;
2011 } dRess_t;
2012 
FIO_createDResources(FIO_prefs_t * const prefs,const char * dictFileName)2013 static dRess_t FIO_createDResources(FIO_prefs_t* const prefs, const char* dictFileName)
2014 {
2015     dRess_t ress;
2016     memset(&ress, 0, sizeof(ress));
2017 
2018     if (prefs->patchFromMode)
2019         FIO_adjustMemLimitForPatchFromMode(prefs, UTIL_getFileSize(dictFileName), 0 /* just use the dict size */);
2020 
2021     /* Allocation */
2022     ress.dctx = ZSTD_createDStream();
2023     if (ress.dctx==NULL)
2024         EXM_THROW(60, "Error: %s : can't create ZSTD_DStream", strerror(errno));
2025     CHECK( ZSTD_DCtx_setMaxWindowSize(ress.dctx, prefs->memLimit) );
2026     CHECK( ZSTD_DCtx_setParameter(ress.dctx, ZSTD_d_forceIgnoreChecksum, !prefs->checksumFlag));
2027 
2028     ress.srcBufferSize = ZSTD_DStreamInSize();
2029     ress.srcBuffer = malloc(ress.srcBufferSize);
2030     ress.dstBufferSize = ZSTD_DStreamOutSize();
2031     ress.dstBuffer = malloc(ress.dstBufferSize);
2032     if (!ress.srcBuffer || !ress.dstBuffer)
2033         EXM_THROW(61, "Allocation error : not enough memory");
2034 
2035     /* dictionary */
2036     {   void* dictBuffer;
2037         size_t const dictBufferSize = FIO_createDictBuffer(&dictBuffer, dictFileName, prefs);
2038         CHECK( ZSTD_initDStream_usingDict(ress.dctx, dictBuffer, dictBufferSize) );
2039         free(dictBuffer);
2040     }
2041 
2042     return ress;
2043 }
2044 
FIO_freeDResources(dRess_t ress)2045 static void FIO_freeDResources(dRess_t ress)
2046 {
2047     CHECK( ZSTD_freeDStream(ress.dctx) );
2048     free(ress.srcBuffer);
2049     free(ress.dstBuffer);
2050 }
2051 
2052 
2053 /** FIO_fwriteSparse() :
2054 *  @return : storedSkips,
2055 *            argument for next call to FIO_fwriteSparse() or FIO_fwriteSparseEnd() */
2056 static unsigned
FIO_fwriteSparse(FILE * file,const void * buffer,size_t bufferSize,const FIO_prefs_t * const prefs,unsigned storedSkips)2057 FIO_fwriteSparse(FILE* file,
2058                  const void* buffer, size_t bufferSize,
2059                  const FIO_prefs_t* const prefs,
2060                  unsigned storedSkips)
2061 {
2062     const size_t* const bufferT = (const size_t*)buffer;   /* Buffer is supposed malloc'ed, hence aligned on size_t */
2063     size_t bufferSizeT = bufferSize / sizeof(size_t);
2064     const size_t* const bufferTEnd = bufferT + bufferSizeT;
2065     const size_t* ptrT = bufferT;
2066     static const size_t segmentSizeT = (32 KB) / sizeof(size_t);   /* check every 32 KB */
2067 
2068     if (prefs->testMode) return 0;  /* do not output anything in test mode */
2069 
2070     if (!prefs->sparseFileSupport) {  /* normal write */
2071         size_t const sizeCheck = fwrite(buffer, 1, bufferSize, file);
2072         if (sizeCheck != bufferSize)
2073             EXM_THROW(70, "Write error : cannot write decoded block : %s",
2074                             strerror(errno));
2075         return 0;
2076     }
2077 
2078     /* avoid int overflow */
2079     if (storedSkips > 1 GB) {
2080         if (LONG_SEEK(file, 1 GB, SEEK_CUR) != 0)
2081             EXM_THROW(91, "1 GB skip error (sparse file support)");
2082         storedSkips -= 1 GB;
2083     }
2084 
2085     while (ptrT < bufferTEnd) {
2086         size_t nb0T;
2087 
2088         /* adjust last segment if < 32 KB */
2089         size_t seg0SizeT = segmentSizeT;
2090         if (seg0SizeT > bufferSizeT) seg0SizeT = bufferSizeT;
2091         bufferSizeT -= seg0SizeT;
2092 
2093         /* count leading zeroes */
2094         for (nb0T=0; (nb0T < seg0SizeT) && (ptrT[nb0T] == 0); nb0T++) ;
2095         storedSkips += (unsigned)(nb0T * sizeof(size_t));
2096 
2097         if (nb0T != seg0SizeT) {   /* not all 0s */
2098             size_t const nbNon0ST = seg0SizeT - nb0T;
2099             /* skip leading zeros */
2100             if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0)
2101                 EXM_THROW(92, "Sparse skip error ; try --no-sparse");
2102             storedSkips = 0;
2103             /* write the rest */
2104             if (fwrite(ptrT + nb0T, sizeof(size_t), nbNon0ST, file) != nbNon0ST)
2105                 EXM_THROW(93, "Write error : cannot write decoded block : %s",
2106                             strerror(errno));
2107         }
2108         ptrT += seg0SizeT;
2109     }
2110 
2111     {   static size_t const maskT = sizeof(size_t)-1;
2112         if (bufferSize & maskT) {
2113             /* size not multiple of sizeof(size_t) : implies end of block */
2114             const char* const restStart = (const char*)bufferTEnd;
2115             const char* restPtr = restStart;
2116             const char* const restEnd = (const char*)buffer + bufferSize;
2117             assert(restEnd > restStart && restEnd < restStart + sizeof(size_t));
2118             for ( ; (restPtr < restEnd) && (*restPtr == 0); restPtr++) ;
2119             storedSkips += (unsigned) (restPtr - restStart);
2120             if (restPtr != restEnd) {
2121                 /* not all remaining bytes are 0 */
2122                 size_t const restSize = (size_t)(restEnd - restPtr);
2123                 if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0)
2124                     EXM_THROW(92, "Sparse skip error ; try --no-sparse");
2125                 if (fwrite(restPtr, 1, restSize, file) != restSize)
2126                     EXM_THROW(95, "Write error : cannot write end of decoded block : %s",
2127                         strerror(errno));
2128                 storedSkips = 0;
2129     }   }   }
2130 
2131     return storedSkips;
2132 }
2133 
2134 static void
FIO_fwriteSparseEnd(const FIO_prefs_t * const prefs,FILE * file,unsigned storedSkips)2135 FIO_fwriteSparseEnd(const FIO_prefs_t* const prefs, FILE* file, unsigned storedSkips)
2136 {
2137     if (prefs->testMode) assert(storedSkips == 0);
2138     if (storedSkips>0) {
2139         assert(prefs->sparseFileSupport > 0);  /* storedSkips>0 implies sparse support is enabled */
2140         (void)prefs;   /* assert can be disabled, in which case prefs becomes unused */
2141         if (LONG_SEEK(file, storedSkips-1, SEEK_CUR) != 0)
2142             EXM_THROW(69, "Final skip error (sparse file support)");
2143         /* last zero must be explicitly written,
2144          * so that skipped ones get implicitly translated as zero by FS */
2145         {   const char lastZeroByte[1] = { 0 };
2146             if (fwrite(lastZeroByte, 1, 1, file) != 1)
2147                 EXM_THROW(69, "Write error : cannot write last zero : %s", strerror(errno));
2148     }   }
2149 }
2150 
2151 
2152 /** FIO_passThrough() : just copy input into output, for compatibility with gzip -df mode
2153     @return : 0 (no error) */
FIO_passThrough(const FIO_prefs_t * const prefs,FILE * foutput,FILE * finput,void * buffer,size_t bufferSize,size_t alreadyLoaded)2154 static int FIO_passThrough(const FIO_prefs_t* const prefs,
2155                            FILE* foutput, FILE* finput,
2156                            void* buffer, size_t bufferSize,
2157                            size_t alreadyLoaded)
2158 {
2159     size_t const blockSize = MIN(64 KB, bufferSize);
2160     size_t readFromInput;
2161     unsigned storedSkips = 0;
2162 
2163     /* assumption : ress->srcBufferLoaded bytes already loaded and stored within buffer */
2164     {   size_t const sizeCheck = fwrite(buffer, 1, alreadyLoaded, foutput);
2165         if (sizeCheck != alreadyLoaded) {
2166             DISPLAYLEVEL(1, "Pass-through write error : %s\n", strerror(errno));
2167             return 1;
2168     }   }
2169 
2170     do {
2171         readFromInput = fread(buffer, 1, blockSize, finput);
2172         storedSkips = FIO_fwriteSparse(foutput, buffer, readFromInput, prefs, storedSkips);
2173     } while (readFromInput == blockSize);
2174     if (ferror(finput)) {
2175         DISPLAYLEVEL(1, "Pass-through read error : %s\n", strerror(errno));
2176         return 1;
2177     }
2178     assert(feof(finput));
2179 
2180     FIO_fwriteSparseEnd(prefs, foutput, storedSkips);
2181     return 0;
2182 }
2183 
2184 /* FIO_zstdErrorHelp() :
2185  * detailed error message when requested window size is too large */
2186 static void
FIO_zstdErrorHelp(const FIO_prefs_t * const prefs,const dRess_t * ress,size_t err,const char * srcFileName)2187 FIO_zstdErrorHelp(const FIO_prefs_t* const prefs,
2188                   const dRess_t* ress,
2189                   size_t err, const char* srcFileName)
2190 {
2191     ZSTD_frameHeader header;
2192 
2193     /* Help message only for one specific error */
2194     if (ZSTD_getErrorCode(err) != ZSTD_error_frameParameter_windowTooLarge)
2195         return;
2196 
2197     /* Try to decode the frame header */
2198     err = ZSTD_getFrameHeader(&header, ress->srcBuffer, ress->srcBufferLoaded);
2199     if (err == 0) {
2200         unsigned long long const windowSize = header.windowSize;
2201         unsigned const windowLog = FIO_highbit64(windowSize) + ((windowSize & (windowSize - 1)) != 0);
2202         assert(prefs->memLimit > 0);
2203         DISPLAYLEVEL(1, "%s : Window size larger than maximum : %llu > %u \n",
2204                         srcFileName, windowSize, prefs->memLimit);
2205         if (windowLog <= ZSTD_WINDOWLOG_MAX) {
2206             unsigned const windowMB = (unsigned)((windowSize >> 20) + ((windowSize & ((1 MB) - 1)) != 0));
2207             assert(windowSize < (U64)(1ULL << 52));   /* ensure now overflow for windowMB */
2208             DISPLAYLEVEL(1, "%s : Use --long=%u or --memory=%uMB \n",
2209                             srcFileName, windowLog, windowMB);
2210             return;
2211     }   }
2212     DISPLAYLEVEL(1, "%s : Window log larger than ZSTD_WINDOWLOG_MAX=%u; not supported \n",
2213                     srcFileName, ZSTD_WINDOWLOG_MAX);
2214 }
2215 
2216 /** FIO_decompressFrame() :
2217  *  @return : size of decoded zstd frame, or an error code
2218  */
2219 #define FIO_ERROR_FRAME_DECODING   ((unsigned long long)(-2))
2220 static unsigned long long
FIO_decompressZstdFrame(FIO_ctx_t * const fCtx,dRess_t * ress,FILE * finput,const FIO_prefs_t * const prefs,const char * srcFileName,U64 alreadyDecoded)2221 FIO_decompressZstdFrame(FIO_ctx_t* const fCtx, dRess_t* ress, FILE* finput,
2222                         const FIO_prefs_t* const prefs,
2223                         const char* srcFileName,
2224                         U64 alreadyDecoded)  /* for multi-frames streams */
2225 {
2226     U64 frameSize = 0;
2227     U32 storedSkips = 0;
2228 
2229     /* display last 20 characters only */
2230     {   size_t const srcFileLength = strlen(srcFileName);
2231         if (srcFileLength>20) srcFileName += srcFileLength-20;
2232     }
2233 
2234     ZSTD_DCtx_reset(ress->dctx, ZSTD_reset_session_only);
2235 
2236     /* Header loading : ensures ZSTD_getFrameHeader() will succeed */
2237     {   size_t const toDecode = ZSTD_FRAMEHEADERSIZE_MAX;
2238         if (ress->srcBufferLoaded < toDecode) {
2239             size_t const toRead = toDecode - ress->srcBufferLoaded;
2240             void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded;
2241             ress->srcBufferLoaded += fread(startPosition, 1, toRead, finput);
2242     }   }
2243 
2244     /* Main decompression Loop */
2245     while (1) {
2246         ZSTD_inBuffer  inBuff = { ress->srcBuffer, ress->srcBufferLoaded, 0 };
2247         ZSTD_outBuffer outBuff= { ress->dstBuffer, ress->dstBufferSize, 0 };
2248         size_t const readSizeHint = ZSTD_decompressStream(ress->dctx, &outBuff, &inBuff);
2249         const int displayLevel = (g_display_prefs.progressSetting == FIO_ps_always) ? 1 : 2;
2250         UTIL_HumanReadableSize_t const hrs = UTIL_makeHumanReadableSize(alreadyDecoded+frameSize);
2251         if (ZSTD_isError(readSizeHint)) {
2252             DISPLAYLEVEL(1, "%s : Decoding error (36) : %s \n",
2253                             srcFileName, ZSTD_getErrorName(readSizeHint));
2254             FIO_zstdErrorHelp(prefs, ress, readSizeHint, srcFileName);
2255             return FIO_ERROR_FRAME_DECODING;
2256         }
2257 
2258         /* Write block */
2259         storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, outBuff.pos, prefs, storedSkips);
2260         frameSize += outBuff.pos;
2261         if (fCtx->nbFilesTotal > 1) {
2262             size_t srcFileNameSize = strlen(srcFileName);
2263             if (srcFileNameSize > 18) {
2264                 const char* truncatedSrcFileName = srcFileName + srcFileNameSize - 15;
2265                 DISPLAYUPDATE(displayLevel, "\rDecompress: %2u/%2u files. Current: ...%s : %.*f%s...    ",
2266                                 fCtx->currFileIdx+1, fCtx->nbFilesTotal, truncatedSrcFileName, hrs.precision, hrs.value, hrs.suffix);
2267             } else {
2268                 DISPLAYUPDATE(displayLevel, "\rDecompress: %2u/%2u files. Current: %s : %.*f%s...    ",
2269                             fCtx->currFileIdx+1, fCtx->nbFilesTotal, srcFileName, hrs.precision, hrs.value, hrs.suffix);
2270             }
2271         } else {
2272             DISPLAYUPDATE(displayLevel, "\r%-20.20s : %.*f%s...     ",
2273                             srcFileName, hrs.precision, hrs.value, hrs.suffix);
2274         }
2275 
2276         if (inBuff.pos > 0) {
2277             memmove(ress->srcBuffer, (char*)ress->srcBuffer + inBuff.pos, inBuff.size - inBuff.pos);
2278             ress->srcBufferLoaded -= inBuff.pos;
2279         }
2280 
2281         if (readSizeHint == 0) break;   /* end of frame */
2282 
2283         /* Fill input buffer */
2284         {   size_t const toDecode = MIN(readSizeHint, ress->srcBufferSize);  /* support large skippable frames */
2285             if (ress->srcBufferLoaded < toDecode) {
2286                 size_t const toRead = toDecode - ress->srcBufferLoaded;   /* > 0 */
2287                 void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded;
2288                 size_t const readSize = fread(startPosition, 1, toRead, finput);
2289                 if (readSize==0) {
2290                     DISPLAYLEVEL(1, "%s : Read error (39) : premature end \n",
2291                                     srcFileName);
2292                     return FIO_ERROR_FRAME_DECODING;
2293                 }
2294                 ress->srcBufferLoaded += readSize;
2295     }   }   }
2296 
2297     FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
2298 
2299     return frameSize;
2300 }
2301 
2302 
2303 #ifdef ZSTD_GZDECOMPRESS
2304 static unsigned long long
FIO_decompressGzFrame(dRess_t * ress,FILE * srcFile,const FIO_prefs_t * const prefs,const char * srcFileName)2305 FIO_decompressGzFrame(dRess_t* ress, FILE* srcFile,
2306                       const FIO_prefs_t* const prefs,
2307                       const char* srcFileName)
2308 {
2309     unsigned long long outFileSize = 0;
2310     z_stream strm;
2311     int flush = Z_NO_FLUSH;
2312     int decodingError = 0;
2313     unsigned storedSkips = 0;
2314 
2315     strm.zalloc = Z_NULL;
2316     strm.zfree = Z_NULL;
2317     strm.opaque = Z_NULL;
2318     strm.next_in = 0;
2319     strm.avail_in = 0;
2320     /* see http://www.zlib.net/manual.html */
2321     if (inflateInit2(&strm, 15 /* maxWindowLogSize */ + 16 /* gzip only */) != Z_OK)
2322         return FIO_ERROR_FRAME_DECODING;
2323 
2324     strm.next_out = (Bytef*)ress->dstBuffer;
2325     strm.avail_out = (uInt)ress->dstBufferSize;
2326     strm.avail_in = (uInt)ress->srcBufferLoaded;
2327     strm.next_in = (z_const unsigned char*)ress->srcBuffer;
2328 
2329     for ( ; ; ) {
2330         int ret;
2331         if (strm.avail_in == 0) {
2332             ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile);
2333             if (ress->srcBufferLoaded == 0) flush = Z_FINISH;
2334             strm.next_in = (z_const unsigned char*)ress->srcBuffer;
2335             strm.avail_in = (uInt)ress->srcBufferLoaded;
2336         }
2337         ret = inflate(&strm, flush);
2338         if (ret == Z_BUF_ERROR) {
2339             DISPLAYLEVEL(1, "zstd: %s: premature gz end \n", srcFileName);
2340             decodingError = 1; break;
2341         }
2342         if (ret != Z_OK && ret != Z_STREAM_END) {
2343             DISPLAYLEVEL(1, "zstd: %s: inflate error %d \n", srcFileName, ret);
2344             decodingError = 1; break;
2345         }
2346         {   size_t const decompBytes = ress->dstBufferSize - strm.avail_out;
2347             if (decompBytes) {
2348                 storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decompBytes, prefs, storedSkips);
2349                 outFileSize += decompBytes;
2350                 strm.next_out = (Bytef*)ress->dstBuffer;
2351                 strm.avail_out = (uInt)ress->dstBufferSize;
2352             }
2353         }
2354         if (ret == Z_STREAM_END) break;
2355     }
2356 
2357     if (strm.avail_in > 0)
2358         memmove(ress->srcBuffer, strm.next_in, strm.avail_in);
2359     ress->srcBufferLoaded = strm.avail_in;
2360     if ( (inflateEnd(&strm) != Z_OK)  /* release resources ; error detected */
2361       && (decodingError==0) ) {
2362         DISPLAYLEVEL(1, "zstd: %s: inflateEnd error \n", srcFileName);
2363         decodingError = 1;
2364     }
2365     FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
2366     return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize;
2367 }
2368 #endif
2369 
2370 
2371 #ifdef ZSTD_LZMADECOMPRESS
2372 static unsigned long long
FIO_decompressLzmaFrame(dRess_t * ress,FILE * srcFile,const FIO_prefs_t * const prefs,const char * srcFileName,int plain_lzma)2373 FIO_decompressLzmaFrame(dRess_t* ress, FILE* srcFile,
2374                         const FIO_prefs_t* const prefs,
2375                         const char* srcFileName, int plain_lzma)
2376 {
2377     unsigned long long outFileSize = 0;
2378     lzma_stream strm = LZMA_STREAM_INIT;
2379     lzma_action action = LZMA_RUN;
2380     lzma_ret initRet;
2381     int decodingError = 0;
2382     unsigned storedSkips = 0;
2383 
2384     strm.next_in = 0;
2385     strm.avail_in = 0;
2386     if (plain_lzma) {
2387         initRet = lzma_alone_decoder(&strm, UINT64_MAX); /* LZMA */
2388     } else {
2389         initRet = lzma_stream_decoder(&strm, UINT64_MAX, 0); /* XZ */
2390     }
2391 
2392     if (initRet != LZMA_OK) {
2393         DISPLAYLEVEL(1, "zstd: %s: %s error %d \n",
2394                         plain_lzma ? "lzma_alone_decoder" : "lzma_stream_decoder",
2395                         srcFileName, initRet);
2396         return FIO_ERROR_FRAME_DECODING;
2397     }
2398 
2399     strm.next_out = (BYTE*)ress->dstBuffer;
2400     strm.avail_out = ress->dstBufferSize;
2401     strm.next_in = (BYTE const*)ress->srcBuffer;
2402     strm.avail_in = ress->srcBufferLoaded;
2403 
2404     for ( ; ; ) {
2405         lzma_ret ret;
2406         if (strm.avail_in == 0) {
2407             ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile);
2408             if (ress->srcBufferLoaded == 0) action = LZMA_FINISH;
2409             strm.next_in = (BYTE const*)ress->srcBuffer;
2410             strm.avail_in = ress->srcBufferLoaded;
2411         }
2412         ret = lzma_code(&strm, action);
2413 
2414         if (ret == LZMA_BUF_ERROR) {
2415             DISPLAYLEVEL(1, "zstd: %s: premature lzma end \n", srcFileName);
2416             decodingError = 1; break;
2417         }
2418         if (ret != LZMA_OK && ret != LZMA_STREAM_END) {
2419             DISPLAYLEVEL(1, "zstd: %s: lzma_code decoding error %d \n",
2420                             srcFileName, ret);
2421             decodingError = 1; break;
2422         }
2423         {   size_t const decompBytes = ress->dstBufferSize - strm.avail_out;
2424             if (decompBytes) {
2425                 storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decompBytes, prefs, storedSkips);
2426                 outFileSize += decompBytes;
2427                 strm.next_out = (BYTE*)ress->dstBuffer;
2428                 strm.avail_out = ress->dstBufferSize;
2429         }   }
2430         if (ret == LZMA_STREAM_END) break;
2431     }
2432 
2433     if (strm.avail_in > 0)
2434         memmove(ress->srcBuffer, strm.next_in, strm.avail_in);
2435     ress->srcBufferLoaded = strm.avail_in;
2436     lzma_end(&strm);
2437     FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
2438     return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize;
2439 }
2440 #endif
2441 
2442 #ifdef ZSTD_LZ4DECOMPRESS
2443 static unsigned long long
FIO_decompressLz4Frame(dRess_t * ress,FILE * srcFile,const FIO_prefs_t * const prefs,const char * srcFileName)2444 FIO_decompressLz4Frame(dRess_t* ress, FILE* srcFile,
2445                        const FIO_prefs_t* const prefs,
2446                        const char* srcFileName)
2447 {
2448     unsigned long long filesize = 0;
2449     LZ4F_errorCode_t nextToLoad;
2450     LZ4F_decompressionContext_t dCtx;
2451     LZ4F_errorCode_t const errorCode = LZ4F_createDecompressionContext(&dCtx, LZ4F_VERSION);
2452     int decodingError = 0;
2453     unsigned storedSkips = 0;
2454 
2455     if (LZ4F_isError(errorCode)) {
2456         DISPLAYLEVEL(1, "zstd: failed to create lz4 decompression context \n");
2457         return FIO_ERROR_FRAME_DECODING;
2458     }
2459 
2460     /* Init feed with magic number (already consumed from FILE* sFile) */
2461     {   size_t inSize = 4;
2462         size_t outSize= 0;
2463         MEM_writeLE32(ress->srcBuffer, LZ4_MAGICNUMBER);
2464         nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &outSize, ress->srcBuffer, &inSize, NULL);
2465         if (LZ4F_isError(nextToLoad)) {
2466             DISPLAYLEVEL(1, "zstd: %s: lz4 header error : %s \n",
2467                             srcFileName, LZ4F_getErrorName(nextToLoad));
2468             LZ4F_freeDecompressionContext(dCtx);
2469             return FIO_ERROR_FRAME_DECODING;
2470     }   }
2471 
2472     /* Main Loop */
2473     for (;nextToLoad;) {
2474         size_t readSize;
2475         size_t pos = 0;
2476         size_t decodedBytes = ress->dstBufferSize;
2477 
2478         /* Read input */
2479         if (nextToLoad > ress->srcBufferSize) nextToLoad = ress->srcBufferSize;
2480         readSize = fread(ress->srcBuffer, 1, nextToLoad, srcFile);
2481         if (!readSize) break;   /* reached end of file or stream */
2482 
2483         while ((pos < readSize) || (decodedBytes == ress->dstBufferSize)) {  /* still to read, or still to flush */
2484             /* Decode Input (at least partially) */
2485             size_t remaining = readSize - pos;
2486             decodedBytes = ress->dstBufferSize;
2487             nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &decodedBytes, (char*)(ress->srcBuffer)+pos, &remaining, NULL);
2488             if (LZ4F_isError(nextToLoad)) {
2489                 DISPLAYLEVEL(1, "zstd: %s: lz4 decompression error : %s \n",
2490                                 srcFileName, LZ4F_getErrorName(nextToLoad));
2491                 decodingError = 1; nextToLoad = 0; break;
2492             }
2493             pos += remaining;
2494 
2495             /* Write Block */
2496             if (decodedBytes) {
2497                 UTIL_HumanReadableSize_t hrs;
2498                 storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decodedBytes, prefs, storedSkips);
2499                 filesize += decodedBytes;
2500                 hrs = UTIL_makeHumanReadableSize(filesize);
2501                 DISPLAYUPDATE(2, "\rDecompressed : %.*f%s  ", hrs.precision, hrs.value, hrs.suffix);
2502             }
2503 
2504             if (!nextToLoad) break;
2505         }
2506     }
2507     /* can be out because readSize == 0, which could be an fread() error */
2508     if (ferror(srcFile)) {
2509         DISPLAYLEVEL(1, "zstd: %s: read error \n", srcFileName);
2510         decodingError=1;
2511     }
2512 
2513     if (nextToLoad!=0) {
2514         DISPLAYLEVEL(1, "zstd: %s: unfinished lz4 stream \n", srcFileName);
2515         decodingError=1;
2516     }
2517 
2518     LZ4F_freeDecompressionContext(dCtx);
2519     ress->srcBufferLoaded = 0; /* LZ4F will reach exact frame boundary */
2520     FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
2521 
2522     return decodingError ? FIO_ERROR_FRAME_DECODING : filesize;
2523 }
2524 #endif
2525 
2526 
2527 
2528 /** FIO_decompressFrames() :
2529  *  Find and decode frames inside srcFile
2530  *  srcFile presumed opened and valid
2531  * @return : 0 : OK
2532  *           1 : error
2533  */
FIO_decompressFrames(FIO_ctx_t * const fCtx,dRess_t ress,FILE * srcFile,const FIO_prefs_t * const prefs,const char * dstFileName,const char * srcFileName)2534 static int FIO_decompressFrames(FIO_ctx_t* const fCtx,
2535                           dRess_t ress, FILE* srcFile,
2536                           const FIO_prefs_t* const prefs,
2537                           const char* dstFileName, const char* srcFileName)
2538 {
2539     unsigned readSomething = 0;
2540     unsigned long long filesize = 0;
2541     assert(srcFile != NULL);
2542 
2543     /* for each frame */
2544     for ( ; ; ) {
2545         /* check magic number -> version */
2546         size_t const toRead = 4;
2547         const BYTE* const buf = (const BYTE*)ress.srcBuffer;
2548         if (ress.srcBufferLoaded < toRead)  /* load up to 4 bytes for header */
2549             ress.srcBufferLoaded += fread((char*)ress.srcBuffer + ress.srcBufferLoaded,
2550                                           (size_t)1, toRead - ress.srcBufferLoaded, srcFile);
2551         if (ress.srcBufferLoaded==0) {
2552             if (readSomething==0) {  /* srcFile is empty (which is invalid) */
2553                 DISPLAYLEVEL(1, "zstd: %s: unexpected end of file \n", srcFileName);
2554                 return 1;
2555             }  /* else, just reached frame boundary */
2556             break;   /* no more input */
2557         }
2558         readSomething = 1;   /* there is at least 1 byte in srcFile */
2559         if (ress.srcBufferLoaded < toRead) {
2560             DISPLAYLEVEL(1, "zstd: %s: unknown header \n", srcFileName);
2561             return 1;
2562         }
2563         if (ZSTD_isFrame(buf, ress.srcBufferLoaded)) {
2564             unsigned long long const frameSize = FIO_decompressZstdFrame(fCtx, &ress, srcFile, prefs, srcFileName, filesize);
2565             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2566             filesize += frameSize;
2567         } else if (buf[0] == 31 && buf[1] == 139) { /* gz magic number */
2568 #ifdef ZSTD_GZDECOMPRESS
2569             unsigned long long const frameSize = FIO_decompressGzFrame(&ress, srcFile, prefs, srcFileName);
2570             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2571             filesize += frameSize;
2572 #else
2573             DISPLAYLEVEL(1, "zstd: %s: gzip file cannot be uncompressed (zstd compiled without HAVE_ZLIB) -- ignored \n", srcFileName);
2574             return 1;
2575 #endif
2576         } else if ((buf[0] == 0xFD && buf[1] == 0x37)  /* xz magic number */
2577                 || (buf[0] == 0x5D && buf[1] == 0x00)) { /* lzma header (no magic number) */
2578 #ifdef ZSTD_LZMADECOMPRESS
2579             unsigned long long const frameSize = FIO_decompressLzmaFrame(&ress, srcFile, prefs, srcFileName, buf[0] != 0xFD);
2580             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2581             filesize += frameSize;
2582 #else
2583             DISPLAYLEVEL(1, "zstd: %s: xz/lzma file cannot be uncompressed (zstd compiled without HAVE_LZMA) -- ignored \n", srcFileName);
2584             return 1;
2585 #endif
2586         } else if (MEM_readLE32(buf) == LZ4_MAGICNUMBER) {
2587 #ifdef ZSTD_LZ4DECOMPRESS
2588             unsigned long long const frameSize = FIO_decompressLz4Frame(&ress, srcFile, prefs, srcFileName);
2589             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
2590             filesize += frameSize;
2591 #else
2592             DISPLAYLEVEL(1, "zstd: %s: lz4 file cannot be uncompressed (zstd compiled without HAVE_LZ4) -- ignored \n", srcFileName);
2593             return 1;
2594 #endif
2595         } else if ((prefs->overwrite) && !strcmp (dstFileName, stdoutmark)) {  /* pass-through mode */
2596             return FIO_passThrough(prefs,
2597                                    ress.dstFile, srcFile,
2598                                    ress.srcBuffer, ress.srcBufferSize,
2599                                    ress.srcBufferLoaded);
2600         } else {
2601             DISPLAYLEVEL(1, "zstd: %s: unsupported format \n", srcFileName);
2602             return 1;
2603     }   }  /* for each frame */
2604 
2605     /* Final Status */
2606     fCtx->totalBytesOutput += (size_t)filesize;
2607     DISPLAYLEVEL(2, "\r%79s\r", "");
2608     /* No status message in pipe mode (stdin - stdout) or multi-files mode */
2609     if ((g_display_prefs.displayLevel >= 2 && fCtx->nbFilesTotal <= 1) ||
2610         g_display_prefs.displayLevel >= 3 ||
2611         g_display_prefs.progressSetting == FIO_ps_always) {
2612         DISPLAYLEVEL(1, "\r%-20s: %llu bytes \n", srcFileName, filesize);
2613     }
2614 
2615     return 0;
2616 }
2617 
2618 /** FIO_decompressDstFile() :
2619     open `dstFileName`,
2620     or path-through if ress.dstFile is already != 0,
2621     then start decompression process (FIO_decompressFrames()).
2622     @return : 0 : OK
2623               1 : operation aborted
2624 */
FIO_decompressDstFile(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,dRess_t ress,FILE * srcFile,const char * dstFileName,const char * srcFileName)2625 static int FIO_decompressDstFile(FIO_ctx_t* const fCtx,
2626                                  FIO_prefs_t* const prefs,
2627                                  dRess_t ress, FILE* srcFile,
2628                                  const char* dstFileName, const char* srcFileName)
2629 {
2630     int result;
2631     stat_t statbuf;
2632     int releaseDstFile = 0;
2633     int transferMTime = 0;
2634 
2635     if ((ress.dstFile == NULL) && (prefs->testMode==0)) {
2636         int dstFilePermissions = DEFAULT_FILE_PERMISSIONS;
2637         if ( strcmp(srcFileName, stdinmark)   /* special case : don't transfer permissions from stdin */
2638           && strcmp(dstFileName, stdoutmark)
2639           && UTIL_stat(srcFileName, &statbuf)
2640           && UTIL_isRegularFileStat(&statbuf) ) {
2641             dstFilePermissions = statbuf.st_mode;
2642             transferMTime = 1;
2643         }
2644 
2645         releaseDstFile = 1;
2646 
2647         ress.dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFilePermissions);
2648         if (ress.dstFile==NULL) return 1;
2649 
2650         /* Must only be added after FIO_openDstFile() succeeds.
2651          * Otherwise we may delete the destination file if it already exists,
2652          * and the user presses Ctrl-C when asked if they wish to overwrite.
2653          */
2654         addHandler(dstFileName);
2655     }
2656 
2657     result = FIO_decompressFrames(fCtx, ress, srcFile, prefs, dstFileName, srcFileName);
2658 
2659     if (releaseDstFile) {
2660         FILE* const dstFile = ress.dstFile;
2661         clearHandler();
2662         ress.dstFile = NULL;
2663         if (fclose(dstFile)) {
2664             DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno));
2665             result = 1;
2666         }
2667 
2668         if (transferMTime) {
2669             UTIL_utime(dstFileName, &statbuf);
2670         }
2671 
2672         if ( (result != 0)  /* operation failure */
2673           && strcmp(dstFileName, stdoutmark)  /* special case : don't remove() stdout */
2674           ) {
2675             FIO_removeFile(dstFileName);  /* remove decompression artefact; note: don't do anything special if remove() fails */
2676         }
2677     }
2678 
2679     return result;
2680 }
2681 
2682 
2683 /** FIO_decompressSrcFile() :
2684     Open `srcFileName`, transfer control to decompressDstFile()
2685     @return : 0 : OK
2686               1 : error
2687 */
FIO_decompressSrcFile(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,dRess_t ress,const char * dstFileName,const char * srcFileName)2688 static int FIO_decompressSrcFile(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs, dRess_t ress, const char* dstFileName, const char* srcFileName)
2689 {
2690     FILE* srcFile;
2691     int result;
2692 
2693     if (UTIL_isDirectory(srcFileName)) {
2694         DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName);
2695         return 1;
2696     }
2697 
2698     srcFile = FIO_openSrcFile(prefs, srcFileName);
2699     if (srcFile==NULL) return 1;
2700     ress.srcBufferLoaded = 0;
2701 
2702     result = FIO_decompressDstFile(fCtx, prefs, ress, srcFile, dstFileName, srcFileName);
2703 
2704     /* Close file */
2705     if (fclose(srcFile)) {
2706         DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno));  /* error should not happen */
2707         return 1;
2708     }
2709     if ( prefs->removeSrcFile  /* --rm */
2710       && (result==0)      /* decompression successful */
2711       && strcmp(srcFileName, stdinmark) ) /* not stdin */ {
2712         /* We must clear the handler, since after this point calling it would
2713          * delete both the source and destination files.
2714          */
2715         clearHandler();
2716         if (FIO_removeFile(srcFileName)) {
2717             /* failed to remove src file */
2718             DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno));
2719             return 1;
2720     }   }
2721     return result;
2722 }
2723 
2724 
2725 
FIO_decompressFilename(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,const char * dstFileName,const char * srcFileName,const char * dictFileName)2726 int FIO_decompressFilename(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs,
2727                            const char* dstFileName, const char* srcFileName,
2728                            const char* dictFileName)
2729 {
2730     dRess_t const ress = FIO_createDResources(prefs, dictFileName);
2731 
2732     int const decodingError = FIO_decompressSrcFile(fCtx, prefs, ress, dstFileName, srcFileName);
2733 
2734     FIO_freeDResources(ress);
2735     return decodingError;
2736 }
2737 
2738 static const char *suffixList[] = {
2739     ZSTD_EXTENSION,
2740     TZSTD_EXTENSION,
2741 #ifndef ZSTD_NODECOMPRESS
2742     ZSTD_ALT_EXTENSION,
2743 #endif
2744 #ifdef ZSTD_GZDECOMPRESS
2745     GZ_EXTENSION,
2746     TGZ_EXTENSION,
2747 #endif
2748 #ifdef ZSTD_LZMADECOMPRESS
2749     LZMA_EXTENSION,
2750     XZ_EXTENSION,
2751     TXZ_EXTENSION,
2752 #endif
2753 #ifdef ZSTD_LZ4DECOMPRESS
2754     LZ4_EXTENSION,
2755     TLZ4_EXTENSION,
2756 #endif
2757     NULL
2758 };
2759 
2760 static const char *suffixListStr =
2761     ZSTD_EXTENSION "/" TZSTD_EXTENSION
2762 #ifdef ZSTD_GZDECOMPRESS
2763     "/" GZ_EXTENSION "/" TGZ_EXTENSION
2764 #endif
2765 #ifdef ZSTD_LZMADECOMPRESS
2766     "/" LZMA_EXTENSION "/" XZ_EXTENSION "/" TXZ_EXTENSION
2767 #endif
2768 #ifdef ZSTD_LZ4DECOMPRESS
2769     "/" LZ4_EXTENSION "/" TLZ4_EXTENSION
2770 #endif
2771 ;
2772 
2773 /* FIO_determineDstName() :
2774  * create a destination filename from a srcFileName.
2775  * @return a pointer to it.
2776  * @return == NULL if there is an error */
2777 static const char*
FIO_determineDstName(const char * srcFileName,const char * outDirName)2778 FIO_determineDstName(const char* srcFileName, const char* outDirName)
2779 {
2780     static size_t dfnbCapacity = 0;
2781     static char* dstFileNameBuffer = NULL;   /* using static allocation : this function cannot be multi-threaded */
2782     size_t dstFileNameEndPos;
2783     char* outDirFilename = NULL;
2784     const char* dstSuffix = "";
2785     size_t dstSuffixLen = 0;
2786 
2787     size_t sfnSize = strlen(srcFileName);
2788 
2789     size_t srcSuffixLen;
2790     const char* const srcSuffix = strrchr(srcFileName, '.');
2791     if (srcSuffix == NULL) {
2792         DISPLAYLEVEL(1,
2793             "zstd: %s: unknown suffix (%s expected). "
2794             "Can't derive the output file name. "
2795             "Specify it with -o dstFileName. Ignoring.\n",
2796             srcFileName, suffixListStr);
2797         return NULL;
2798     }
2799     srcSuffixLen = strlen(srcSuffix);
2800 
2801     {
2802         const char** matchedSuffixPtr;
2803         for (matchedSuffixPtr = suffixList; *matchedSuffixPtr != NULL; matchedSuffixPtr++) {
2804             if (!strcmp(*matchedSuffixPtr, srcSuffix)) {
2805                 break;
2806             }
2807         }
2808 
2809         /* check suffix is authorized */
2810         if (sfnSize <= srcSuffixLen || *matchedSuffixPtr == NULL) {
2811             DISPLAYLEVEL(1,
2812                 "zstd: %s: unknown suffix (%s expected). "
2813                 "Can't derive the output file name. "
2814                 "Specify it with -o dstFileName. Ignoring.\n",
2815                 srcFileName, suffixListStr);
2816             return NULL;
2817         }
2818 
2819         if ((*matchedSuffixPtr)[1] == 't') {
2820             dstSuffix = ".tar";
2821             dstSuffixLen = strlen(dstSuffix);
2822         }
2823     }
2824 
2825     if (outDirName) {
2826         outDirFilename = FIO_createFilename_fromOutDir(srcFileName, outDirName, 0);
2827         sfnSize = strlen(outDirFilename);
2828         assert(outDirFilename != NULL);
2829     }
2830 
2831     if (dfnbCapacity+srcSuffixLen <= sfnSize+1+dstSuffixLen) {
2832         /* allocate enough space to write dstFilename into it */
2833         free(dstFileNameBuffer);
2834         dfnbCapacity = sfnSize + 20;
2835         dstFileNameBuffer = (char*)malloc(dfnbCapacity);
2836         if (dstFileNameBuffer==NULL)
2837             EXM_THROW(74, "%s : not enough memory for dstFileName",
2838                       strerror(errno));
2839     }
2840 
2841     /* return dst name == src name truncated from suffix */
2842     assert(dstFileNameBuffer != NULL);
2843     dstFileNameEndPos = sfnSize - srcSuffixLen;
2844     if (outDirFilename) {
2845         memcpy(dstFileNameBuffer, outDirFilename, dstFileNameEndPos);
2846         free(outDirFilename);
2847     } else {
2848         memcpy(dstFileNameBuffer, srcFileName, dstFileNameEndPos);
2849     }
2850 
2851     /* The short tar extensions tzst, tgz, txz and tlz4 files should have "tar"
2852      * extension on decompression. Also writes terminating null. */
2853     strcpy(dstFileNameBuffer + dstFileNameEndPos, dstSuffix);
2854     return dstFileNameBuffer;
2855 
2856     /* note : dstFileNameBuffer memory is not going to be free */
2857 }
2858 
2859 int
FIO_decompressMultipleFilenames(FIO_ctx_t * const fCtx,FIO_prefs_t * const prefs,const char ** srcNamesTable,const char * outMirroredRootDirName,const char * outDirName,const char * outFileName,const char * dictFileName)2860 FIO_decompressMultipleFilenames(FIO_ctx_t* const fCtx,
2861                                 FIO_prefs_t* const prefs,
2862                                 const char** srcNamesTable,
2863                                 const char* outMirroredRootDirName,
2864                                 const char* outDirName, const char* outFileName,
2865                                 const char* dictFileName)
2866 {
2867     int status;
2868     int error = 0;
2869     dRess_t ress = FIO_createDResources(prefs, dictFileName);
2870 
2871     if (outFileName) {
2872         if (FIO_removeMultiFilesWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
2873             FIO_freeDResources(ress);
2874             return 1;
2875         }
2876         if (!prefs->testMode) {
2877             ress.dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
2878             if (ress.dstFile == 0) EXM_THROW(19, "cannot open %s", outFileName);
2879         }
2880         for (; fCtx->currFileIdx < fCtx->nbFilesTotal; fCtx->currFileIdx++) {
2881             status = FIO_decompressSrcFile(fCtx, prefs, ress, outFileName, srcNamesTable[fCtx->currFileIdx]);
2882             if (!status) fCtx->nbFilesProcessed++;
2883             error |= status;
2884         }
2885         if ((!prefs->testMode) && (fclose(ress.dstFile)))
2886             EXM_THROW(72, "Write error : %s : cannot properly close output file",
2887                         strerror(errno));
2888     } else {
2889         if (outMirroredRootDirName)
2890             UTIL_mirrorSourceFilesDirectories(srcNamesTable, (unsigned)fCtx->nbFilesTotal, outMirroredRootDirName);
2891 
2892         for (; fCtx->currFileIdx < fCtx->nbFilesTotal; fCtx->currFileIdx++) {   /* create dstFileName */
2893             const char* const srcFileName = srcNamesTable[fCtx->currFileIdx];
2894             const char* dstFileName = NULL;
2895             if (outMirroredRootDirName) {
2896                 char* validMirroredDirName = UTIL_createMirroredDestDirName(srcFileName, outMirroredRootDirName);
2897                 if (validMirroredDirName) {
2898                     dstFileName = FIO_determineDstName(srcFileName, validMirroredDirName);
2899                     free(validMirroredDirName);
2900                 } else {
2901                     DISPLAYLEVEL(2, "zstd: --output-dir-mirror cannot decompress '%s' into '%s'\n", srcFileName, outMirroredRootDirName);
2902                 }
2903             } else {
2904                 dstFileName = FIO_determineDstName(srcFileName, outDirName);
2905             }
2906             if (dstFileName == NULL) { error=1; continue; }
2907             status = FIO_decompressSrcFile(fCtx, prefs, ress, dstFileName, srcFileName);
2908             if (!status) fCtx->nbFilesProcessed++;
2909             error |= status;
2910         }
2911         if (outDirName)
2912             FIO_checkFilenameCollisions(srcNamesTable , (unsigned)fCtx->nbFilesTotal);
2913     }
2914 
2915     if (fCtx->nbFilesProcessed >= 1  && fCtx->nbFilesTotal > 1 && fCtx->totalBytesOutput != 0)
2916         DISPLAYLEVEL(2, "%d files decompressed : %6zu bytes total \n", fCtx->nbFilesProcessed, fCtx->totalBytesOutput);
2917 
2918     FIO_freeDResources(ress);
2919     return error;
2920 }
2921 
2922 /* **************************************************************************
2923  *  .zst file info (--list command)
2924  ***************************************************************************/
2925 
2926 typedef struct {
2927     U64 decompressedSize;
2928     U64 compressedSize;
2929     U64 windowSize;
2930     int numActualFrames;
2931     int numSkippableFrames;
2932     int decompUnavailable;
2933     int usesCheck;
2934     U32 nbFiles;
2935 } fileInfo_t;
2936 
2937 typedef enum {
2938   info_success=0,
2939   info_frame_error=1,
2940   info_not_zstd=2,
2941   info_file_error=3,
2942   info_truncated_input=4,
2943 } InfoError;
2944 
2945 #define ERROR_IF(c,n,...) {             \
2946     if (c) {                           \
2947         DISPLAYLEVEL(1, __VA_ARGS__);  \
2948         DISPLAYLEVEL(1, " \n");        \
2949         return n;                      \
2950     }                                  \
2951 }
2952 
2953 static InfoError
FIO_analyzeFrames(fileInfo_t * info,FILE * const srcFile)2954 FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile)
2955 {
2956     /* begin analyzing frame */
2957     for ( ; ; ) {
2958         BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
2959         size_t const numBytesRead = fread(headerBuffer, 1, sizeof(headerBuffer), srcFile);
2960         if (numBytesRead < ZSTD_FRAMEHEADERSIZE_MIN(ZSTD_f_zstd1)) {
2961             if ( feof(srcFile)
2962               && (numBytesRead == 0)
2963               && (info->compressedSize > 0)
2964               && (info->compressedSize != UTIL_FILESIZE_UNKNOWN) ) {
2965                 unsigned long long file_position = (unsigned long long) LONG_TELL(srcFile);
2966                 unsigned long long file_size = (unsigned long long) info->compressedSize;
2967                 ERROR_IF(file_position != file_size, info_truncated_input,
2968                   "Error: seeked to position %llu, which is beyond file size of %llu\n",
2969                   file_position,
2970                   file_size);
2971                 break;  /* correct end of file => success */
2972             }
2973             ERROR_IF(feof(srcFile), info_not_zstd, "Error: reached end of file with incomplete frame");
2974             ERROR_IF(1, info_frame_error, "Error: did not reach end of file but ran out of frames");
2975         }
2976         {   U32 const magicNumber = MEM_readLE32(headerBuffer);
2977             /* Zstandard frame */
2978             if (magicNumber == ZSTD_MAGICNUMBER) {
2979                 ZSTD_frameHeader header;
2980                 U64 const frameContentSize = ZSTD_getFrameContentSize(headerBuffer, numBytesRead);
2981                 if ( frameContentSize == ZSTD_CONTENTSIZE_ERROR
2982                   || frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN ) {
2983                     info->decompUnavailable = 1;
2984                 } else {
2985                     info->decompressedSize += frameContentSize;
2986                 }
2987                 ERROR_IF(ZSTD_getFrameHeader(&header, headerBuffer, numBytesRead) != 0,
2988                         info_frame_error, "Error: could not decode frame header");
2989                 info->windowSize = header.windowSize;
2990                 /* move to the end of the frame header */
2991                 {   size_t const headerSize = ZSTD_frameHeaderSize(headerBuffer, numBytesRead);
2992                     ERROR_IF(ZSTD_isError(headerSize), info_frame_error, "Error: could not determine frame header size");
2993                     ERROR_IF(fseek(srcFile, ((long)headerSize)-((long)numBytesRead), SEEK_CUR) != 0,
2994                             info_frame_error, "Error: could not move to end of frame header");
2995                 }
2996 
2997                 /* skip all blocks in the frame */
2998                 {   int lastBlock = 0;
2999                     do {
3000                         BYTE blockHeaderBuffer[3];
3001                         ERROR_IF(fread(blockHeaderBuffer, 1, 3, srcFile) != 3,
3002                                 info_frame_error, "Error while reading block header");
3003                         {   U32 const blockHeader = MEM_readLE24(blockHeaderBuffer);
3004                             U32 const blockTypeID = (blockHeader >> 1) & 3;
3005                             U32 const isRLE = (blockTypeID == 1);
3006                             U32 const isWrongBlock = (blockTypeID == 3);
3007                             long const blockSize = isRLE ? 1 : (long)(blockHeader >> 3);
3008                             ERROR_IF(isWrongBlock, info_frame_error, "Error: unsupported block type");
3009                             lastBlock = blockHeader & 1;
3010                             ERROR_IF(fseek(srcFile, blockSize, SEEK_CUR) != 0,
3011                                     info_frame_error, "Error: could not skip to end of block");
3012                         }
3013                     } while (lastBlock != 1);
3014                 }
3015 
3016                 /* check if checksum is used */
3017                 {   BYTE const frameHeaderDescriptor = headerBuffer[4];
3018                     int const contentChecksumFlag = (frameHeaderDescriptor & (1 << 2)) >> 2;
3019                     if (contentChecksumFlag) {
3020                         info->usesCheck = 1;
3021                         ERROR_IF(fseek(srcFile, 4, SEEK_CUR) != 0,
3022                                 info_frame_error, "Error: could not skip past checksum");
3023                 }   }
3024                 info->numActualFrames++;
3025             }
3026             /* Skippable frame */
3027             else if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
3028                 U32 const frameSize = MEM_readLE32(headerBuffer + 4);
3029                 long const seek = (long)(8 + frameSize - numBytesRead);
3030                 ERROR_IF(LONG_SEEK(srcFile, seek, SEEK_CUR) != 0,
3031                         info_frame_error, "Error: could not find end of skippable frame");
3032                 info->numSkippableFrames++;
3033             }
3034             /* unknown content */
3035             else {
3036                 return info_not_zstd;
3037             }
3038         }  /* magic number analysis */
3039     }  /* end analyzing frames */
3040     return info_success;
3041 }
3042 
3043 
3044 static InfoError
getFileInfo_fileConfirmed(fileInfo_t * info,const char * inFileName)3045 getFileInfo_fileConfirmed(fileInfo_t* info, const char* inFileName)
3046 {
3047     InfoError status;
3048     FILE* const srcFile = FIO_openSrcFile(NULL, inFileName);
3049     ERROR_IF(srcFile == NULL, info_file_error, "Error: could not open source file %s", inFileName);
3050 
3051     info->compressedSize = UTIL_getFileSize(inFileName);
3052     status = FIO_analyzeFrames(info, srcFile);
3053 
3054     fclose(srcFile);
3055     info->nbFiles = 1;
3056     return status;
3057 }
3058 
3059 
3060 /** getFileInfo() :
3061  *  Reads information from file, stores in *info
3062  * @return : InfoError status
3063  */
3064 static InfoError
getFileInfo(fileInfo_t * info,const char * srcFileName)3065 getFileInfo(fileInfo_t* info, const char* srcFileName)
3066 {
3067     ERROR_IF(!UTIL_isRegularFile(srcFileName),
3068             info_file_error, "Error : %s is not a file", srcFileName);
3069     return getFileInfo_fileConfirmed(info, srcFileName);
3070 }
3071 
3072 
3073 static void
displayInfo(const char * inFileName,const fileInfo_t * info,int displayLevel)3074 displayInfo(const char* inFileName, const fileInfo_t* info, int displayLevel)
3075 {
3076     UTIL_HumanReadableSize_t const window_hrs = UTIL_makeHumanReadableSize(info->windowSize);
3077     UTIL_HumanReadableSize_t const compressed_hrs = UTIL_makeHumanReadableSize(info->compressedSize);
3078     UTIL_HumanReadableSize_t const decompressed_hrs = UTIL_makeHumanReadableSize(info->decompressedSize);
3079     double const ratio = (info->compressedSize == 0) ? 0 : ((double)info->decompressedSize)/(double)info->compressedSize;
3080     const char* const checkString = (info->usesCheck ? "XXH64" : "None");
3081     if (displayLevel <= 2) {
3082         if (!info->decompUnavailable) {
3083             DISPLAYOUT("%6d  %5d  %6.*f%4s  %8.*f%4s  %5.3f  %5s  %s\n",
3084                     info->numSkippableFrames + info->numActualFrames,
3085                     info->numSkippableFrames,
3086                     compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3087                     decompressed_hrs.precision, decompressed_hrs.value, decompressed_hrs.suffix,
3088                     ratio, checkString, inFileName);
3089         } else {
3090             DISPLAYOUT("%6d  %5d  %6.*f%4s                       %5s  %s\n",
3091                     info->numSkippableFrames + info->numActualFrames,
3092                     info->numSkippableFrames,
3093                     compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3094                     checkString, inFileName);
3095         }
3096     } else {
3097         DISPLAYOUT("%s \n", inFileName);
3098         DISPLAYOUT("# Zstandard Frames: %d\n", info->numActualFrames);
3099         if (info->numSkippableFrames)
3100             DISPLAYOUT("# Skippable Frames: %d\n", info->numSkippableFrames);
3101         DISPLAYOUT("Window Size: %.*f%s (%llu B)\n",
3102                    window_hrs.precision, window_hrs.value, window_hrs.suffix,
3103                    (unsigned long long)info->windowSize);
3104         DISPLAYOUT("Compressed Size: %.*f%s (%llu B)\n",
3105                     compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3106                     (unsigned long long)info->compressedSize);
3107         if (!info->decompUnavailable) {
3108             DISPLAYOUT("Decompressed Size: %.*f%s (%llu B)\n",
3109                     decompressed_hrs.precision, decompressed_hrs.value, decompressed_hrs.suffix,
3110                     (unsigned long long)info->decompressedSize);
3111             DISPLAYOUT("Ratio: %.4f\n", ratio);
3112         }
3113         DISPLAYOUT("Check: %s\n", checkString);
3114         DISPLAYOUT("\n");
3115     }
3116 }
3117 
FIO_addFInfo(fileInfo_t fi1,fileInfo_t fi2)3118 static fileInfo_t FIO_addFInfo(fileInfo_t fi1, fileInfo_t fi2)
3119 {
3120     fileInfo_t total;
3121     memset(&total, 0, sizeof(total));
3122     total.numActualFrames = fi1.numActualFrames + fi2.numActualFrames;
3123     total.numSkippableFrames = fi1.numSkippableFrames + fi2.numSkippableFrames;
3124     total.compressedSize = fi1.compressedSize + fi2.compressedSize;
3125     total.decompressedSize = fi1.decompressedSize + fi2.decompressedSize;
3126     total.decompUnavailable = fi1.decompUnavailable | fi2.decompUnavailable;
3127     total.usesCheck = fi1.usesCheck & fi2.usesCheck;
3128     total.nbFiles = fi1.nbFiles + fi2.nbFiles;
3129     return total;
3130 }
3131 
3132 static int
FIO_listFile(fileInfo_t * total,const char * inFileName,int displayLevel)3133 FIO_listFile(fileInfo_t* total, const char* inFileName, int displayLevel)
3134 {
3135     fileInfo_t info;
3136     memset(&info, 0, sizeof(info));
3137     {   InfoError const error = getFileInfo(&info, inFileName);
3138         switch (error) {
3139             case info_frame_error:
3140                 /* display error, but provide output */
3141                 DISPLAYLEVEL(1, "Error while parsing \"%s\" \n", inFileName);
3142                 break;
3143             case info_not_zstd:
3144                 DISPLAYOUT("File \"%s\" not compressed by zstd \n", inFileName);
3145                 if (displayLevel > 2) DISPLAYOUT("\n");
3146                 return 1;
3147             case info_file_error:
3148                 /* error occurred while opening the file */
3149                 if (displayLevel > 2) DISPLAYOUT("\n");
3150                 return 1;
3151             case info_truncated_input:
3152                 DISPLAYOUT("File \"%s\" is truncated \n", inFileName);
3153                 if (displayLevel > 2) DISPLAYOUT("\n");
3154                 return 1;
3155             case info_success:
3156             default:
3157                 break;
3158         }
3159 
3160         displayInfo(inFileName, &info, displayLevel);
3161         *total = FIO_addFInfo(*total, info);
3162         assert(error == info_success || error == info_frame_error);
3163         return (int)error;
3164     }
3165 }
3166 
FIO_listMultipleFiles(unsigned numFiles,const char ** filenameTable,int displayLevel)3167 int FIO_listMultipleFiles(unsigned numFiles, const char** filenameTable, int displayLevel)
3168 {
3169     /* ensure no specified input is stdin (needs fseek() capability) */
3170     {   unsigned u;
3171         for (u=0; u<numFiles;u++) {
3172             ERROR_IF(!strcmp (filenameTable[u], stdinmark),
3173                     1, "zstd: --list does not support reading from standard input");
3174     }   }
3175 
3176     if (numFiles == 0) {
3177         if (!IS_CONSOLE(stdin)) {
3178             DISPLAYLEVEL(1, "zstd: --list does not support reading from standard input \n");
3179         }
3180         DISPLAYLEVEL(1, "No files given \n");
3181         return 1;
3182     }
3183 
3184     if (displayLevel <= 2) {
3185         DISPLAYOUT("Frames  Skips  Compressed  Uncompressed  Ratio  Check  Filename\n");
3186     }
3187     {   int error = 0;
3188         fileInfo_t total;
3189         memset(&total, 0, sizeof(total));
3190         total.usesCheck = 1;
3191         /* --list each file, and check for any error */
3192         {   unsigned u;
3193             for (u=0; u<numFiles;u++) {
3194                 error |= FIO_listFile(&total, filenameTable[u], displayLevel);
3195         }   }
3196         if (numFiles > 1 && displayLevel <= 2) {   /* display total */
3197             UTIL_HumanReadableSize_t const compressed_hrs = UTIL_makeHumanReadableSize(total.compressedSize);
3198             UTIL_HumanReadableSize_t const decompressed_hrs = UTIL_makeHumanReadableSize(total.decompressedSize);
3199             double const ratio = (total.compressedSize == 0) ? 0 : ((double)total.decompressedSize)/(double)total.compressedSize;
3200             const char* const checkString = (total.usesCheck ? "XXH64" : "");
3201             DISPLAYOUT("----------------------------------------------------------------- \n");
3202             if (total.decompUnavailable) {
3203                 DISPLAYOUT("%6d  %5d  %6.*f%4s                       %5s  %u files\n",
3204                         total.numSkippableFrames + total.numActualFrames,
3205                         total.numSkippableFrames,
3206                         compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3207                         checkString, (unsigned)total.nbFiles);
3208             } else {
3209                 DISPLAYOUT("%6d  %5d  %6.*f%4s  %8.*f%4s  %5.3f  %5s  %u files\n",
3210                         total.numSkippableFrames + total.numActualFrames,
3211                         total.numSkippableFrames,
3212                         compressed_hrs.precision, compressed_hrs.value, compressed_hrs.suffix,
3213                         decompressed_hrs.precision, decompressed_hrs.value, decompressed_hrs.suffix,
3214                         ratio, checkString, (unsigned)total.nbFiles);
3215         }   }
3216         return error;
3217     }
3218 }
3219 
3220 
3221 #endif /* #ifndef ZSTD_NODECOMPRESS */
3222