]>
Commit | Line | Data |
---|---|---|
11fdf7f2 | 1 | /* |
f67539c2 | 2 | * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc. |
7c673cae FG |
3 | * All rights reserved. |
4 | * | |
11fdf7f2 TL |
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. | |
7c673cae FG |
9 | */ |
10 | ||
11fdf7f2 | 11 | |
7c673cae FG |
12 | /* ************************************* |
13 | * Compiler Options | |
14 | ***************************************/ | |
15 | #ifdef _MSC_VER /* Visual */ | |
11fdf7f2 | 16 | # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ |
7c673cae FG |
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 | ||
7c673cae FG |
23 | /*-************************************* |
24 | * Includes | |
25 | ***************************************/ | |
11fdf7f2 | 26 | #include "platform.h" /* Large Files support, SET_BINARY_MODE */ |
9f95a23c | 27 | #include "util.h" /* UTIL_getFileSize, UTIL_isRegularFile, UTIL_isSameFile */ |
7c673cae FG |
28 | #include <stdio.h> /* fprintf, fopen, fread, _fileno, stdin, stdout */ |
29 | #include <stdlib.h> /* malloc, free */ | |
30 | #include <string.h> /* strcmp, strlen */ | |
9f95a23c | 31 | #include <assert.h> |
7c673cae | 32 | #include <errno.h> /* errno */ |
f67539c2 | 33 | #include <limits.h> /* INT_MAX */ |
9f95a23c TL |
34 | #include <signal.h> |
35 | #include "timefn.h" /* UTIL_getTime, UTIL_clockSpanMicro */ | |
7c673cae | 36 | |
11fdf7f2 TL |
37 | #if defined (_MSC_VER) |
38 | # include <sys/stat.h> | |
39 | # include <io.h> | |
40 | #endif | |
41 | ||
f67539c2 | 42 | #include "../lib/common/mem.h" /* U32, U64 */ |
7c673cae | 43 | #include "fileio.h" |
9f95a23c | 44 | |
7c673cae | 45 | #define ZSTD_STATIC_LINKING_ONLY /* ZSTD_magicNumber, ZSTD_frameHeaderSize_max */ |
f67539c2 TL |
46 | #include "../lib/zstd.h" |
47 | #include "../lib/common/zstd_errors.h" /* ZSTD_error_frameParameter_windowTooLarge */ | |
48 | #include "../lib/compress/zstd_compress_internal.h" | |
9f95a23c | 49 | |
11fdf7f2 TL |
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 | |
9f95a23c | 56 | |
11fdf7f2 TL |
57 | #if defined(ZSTD_LZMACOMPRESS) || defined(ZSTD_LZMADECOMPRESS) |
58 | # include <lzma.h> | |
7c673cae FG |
59 | #endif |
60 | ||
11fdf7f2 TL |
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> | |
7c673cae FG |
66 | #endif |
67 | ||
68 | ||
69 | /*-************************************* | |
70 | * Constants | |
71 | ***************************************/ | |
9f95a23c | 72 | #define ADAPT_WINDOWLOG_DEFAULT 23 /* 8 MB */ |
11fdf7f2 | 73 | #define DICTSIZE_MAX (32 MB) /* protection against large input (attack scenario) */ |
7c673cae FG |
74 | |
75 | #define FNSPACE 30 | |
7c673cae | 76 | |
7c673cae FG |
77 | /*-************************************* |
78 | * Macros | |
79 | ***************************************/ | |
9f95a23c TL |
80 | |
81 | struct FIO_display_prefs_s { | |
82 | int displayLevel; /* 0 : no display; 1: errors; 2: + result + interaction + warnings; 3: + progression; 4: + information */ | |
83 | U32 noProgress; | |
84 | }; | |
85 | ||
86 | static FIO_display_prefs_t g_display_prefs = {2, 0}; | |
87 | ||
7c673cae | 88 | #define DISPLAY(...) fprintf(stderr, __VA_ARGS__) |
11fdf7f2 | 89 | #define DISPLAYOUT(...) fprintf(stdout, __VA_ARGS__) |
9f95a23c TL |
90 | #define DISPLAYLEVEL(l, ...) { if (g_display_prefs.displayLevel>=l) { DISPLAY(__VA_ARGS__); } } |
91 | ||
92 | static const U64 g_refreshRate = SEC_TO_MICRO / 6; | |
93 | static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER; | |
94 | ||
95 | #define READY_FOR_UPDATE() (!g_display_prefs.noProgress && UTIL_clockSpanMicro(g_displayClock) > g_refreshRate) | |
96 | #define DELAY_NEXT_UPDATE() { g_displayClock = UTIL_getTime(); } | |
97 | #define DISPLAYUPDATE(l, ...) { \ | |
f67539c2 | 98 | if (g_display_prefs.displayLevel>=l && !g_display_prefs.noProgress) { \ |
9f95a23c TL |
99 | if (READY_FOR_UPDATE() || (g_display_prefs.displayLevel>=4)) { \ |
100 | DELAY_NEXT_UPDATE(); \ | |
101 | DISPLAY(__VA_ARGS__); \ | |
102 | if (g_display_prefs.displayLevel>=4) fflush(stderr); \ | |
103 | } } } | |
7c673cae | 104 | |
11fdf7f2 | 105 | #undef MIN /* in case it would be already defined */ |
7c673cae FG |
106 | #define MIN(a,b) ((a) < (b) ? (a) : (b)) |
107 | ||
108 | ||
9f95a23c TL |
109 | #define EXM_THROW(error, ...) \ |
110 | { \ | |
111 | DISPLAYLEVEL(1, "zstd: "); \ | |
112 | DISPLAYLEVEL(5, "Error defined at %s, line %i : \n", __FILE__, __LINE__); \ | |
113 | DISPLAYLEVEL(1, "error %i : ", error); \ | |
114 | DISPLAYLEVEL(1, __VA_ARGS__); \ | |
115 | DISPLAYLEVEL(1, " \n"); \ | |
116 | exit(error); \ | |
11fdf7f2 TL |
117 | } |
118 | ||
9f95a23c TL |
119 | #define CHECK_V(v, f) \ |
120 | v = f; \ | |
121 | if (ZSTD_isError(v)) { \ | |
122 | DISPLAYLEVEL(5, "%s \n", #f); \ | |
123 | EXM_THROW(11, "%s", ZSTD_getErrorName(v)); \ | |
124 | } | |
125 | #define CHECK(f) { size_t err; CHECK_V(err, f); } | |
11fdf7f2 TL |
126 | |
127 | ||
128 | /*-************************************ | |
129 | * Signal (Ctrl-C trapping) | |
130 | **************************************/ | |
11fdf7f2 TL |
131 | static const char* g_artefact = NULL; |
132 | static void INThandler(int sig) | |
133 | { | |
134 | assert(sig==SIGINT); (void)sig; | |
135 | #if !defined(_MSC_VER) | |
136 | signal(sig, SIG_IGN); /* this invocation generates a buggy warning in Visual Studio */ | |
137 | #endif | |
9f95a23c TL |
138 | if (g_artefact) { |
139 | assert(UTIL_isRegularFile(g_artefact)); | |
140 | remove(g_artefact); | |
141 | } | |
11fdf7f2 TL |
142 | DISPLAY("\n"); |
143 | exit(2); | |
144 | } | |
9f95a23c TL |
145 | static void addHandler(char const* dstFileName) |
146 | { | |
147 | if (UTIL_isRegularFile(dstFileName)) { | |
148 | g_artefact = dstFileName; | |
149 | signal(SIGINT, INThandler); | |
150 | } else { | |
151 | g_artefact = NULL; | |
152 | } | |
153 | } | |
154 | /* Idempotent */ | |
155 | static void clearHandler(void) | |
156 | { | |
157 | if (g_artefact) signal(SIGINT, SIG_DFL); | |
158 | g_artefact = NULL; | |
159 | } | |
160 | ||
161 | ||
162 | /*-********************************************************* | |
163 | * Termination signal trapping (Print debug stack trace) | |
164 | ***********************************************************/ | |
165 | #if defined(__has_feature) && !defined(BACKTRACE_ENABLE) /* Clang compiler */ | |
166 | # if (__has_feature(address_sanitizer)) | |
167 | # define BACKTRACE_ENABLE 0 | |
168 | # endif /* __has_feature(address_sanitizer) */ | |
169 | #elif defined(__SANITIZE_ADDRESS__) && !defined(BACKTRACE_ENABLE) /* GCC compiler */ | |
170 | # define BACKTRACE_ENABLE 0 | |
171 | #endif | |
172 | ||
173 | #if !defined(BACKTRACE_ENABLE) | |
174 | /* automatic detector : backtrace enabled by default on linux+glibc and osx */ | |
f67539c2 | 175 | # if (defined(__linux__) && (defined(__GLIBC__) && !defined(__UCLIBC__))) \ |
9f95a23c TL |
176 | || (defined(__APPLE__) && defined(__MACH__)) |
177 | # define BACKTRACE_ENABLE 1 | |
178 | # else | |
179 | # define BACKTRACE_ENABLE 0 | |
180 | # endif | |
181 | #endif | |
182 | ||
183 | /* note : after this point, BACKTRACE_ENABLE is necessarily defined */ | |
184 | ||
185 | ||
186 | #if BACKTRACE_ENABLE | |
11fdf7f2 | 187 | |
9f95a23c TL |
188 | #include <execinfo.h> /* backtrace, backtrace_symbols */ |
189 | ||
190 | #define MAX_STACK_FRAMES 50 | |
191 | ||
192 | static void ABRThandler(int sig) { | |
193 | const char* name; | |
194 | void* addrlist[MAX_STACK_FRAMES]; | |
195 | char** symbollist; | |
196 | int addrlen, i; | |
197 | ||
198 | switch (sig) { | |
199 | case SIGABRT: name = "SIGABRT"; break; | |
200 | case SIGFPE: name = "SIGFPE"; break; | |
201 | case SIGILL: name = "SIGILL"; break; | |
202 | case SIGINT: name = "SIGINT"; break; | |
203 | case SIGSEGV: name = "SIGSEGV"; break; | |
204 | default: name = "UNKNOWN"; | |
205 | } | |
206 | ||
207 | DISPLAY("Caught %s signal, printing stack:\n", name); | |
208 | /* Retrieve current stack addresses. */ | |
209 | addrlen = backtrace(addrlist, MAX_STACK_FRAMES); | |
210 | if (addrlen == 0) { | |
211 | DISPLAY("\n"); | |
212 | return; | |
213 | } | |
214 | /* Create readable strings to each frame. */ | |
215 | symbollist = backtrace_symbols(addrlist, addrlen); | |
216 | /* Print the stack trace, excluding calls handling the signal. */ | |
217 | for (i = ZSTD_START_SYMBOLLIST_FRAME; i < addrlen; i++) { | |
218 | DISPLAY("%s\n", symbollist[i]); | |
219 | } | |
220 | free(symbollist); | |
221 | /* Reset and raise the signal so default handler runs. */ | |
222 | signal(sig, SIG_DFL); | |
223 | raise(sig); | |
224 | } | |
225 | #endif | |
226 | ||
227 | void FIO_addAbortHandler() | |
228 | { | |
229 | #if BACKTRACE_ENABLE | |
230 | signal(SIGABRT, ABRThandler); | |
231 | signal(SIGFPE, ABRThandler); | |
232 | signal(SIGILL, ABRThandler); | |
233 | signal(SIGSEGV, ABRThandler); | |
234 | signal(SIGBUS, ABRThandler); | |
235 | #endif | |
236 | } | |
11fdf7f2 | 237 | |
9f95a23c TL |
238 | |
239 | /*-************************************************************ | |
240 | * Avoid fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW | |
11fdf7f2 TL |
241 | ***************************************************************/ |
242 | #if defined(_MSC_VER) && _MSC_VER >= 1400 | |
243 | # define LONG_SEEK _fseeki64 | |
9f95a23c | 244 | # define LONG_TELL _ftelli64 |
11fdf7f2 TL |
245 | #elif !defined(__64BIT__) && (PLATFORM_POSIX_VERSION >= 200112L) /* No point defining Large file for 64 bit */ |
246 | # define LONG_SEEK fseeko | |
9f95a23c | 247 | # define LONG_TELL ftello |
11fdf7f2 TL |
248 | #elif defined(__MINGW32__) && !defined(__STRICT_ANSI__) && !defined(__NO_MINGW_LFS) && defined(__MSVCRT__) |
249 | # define LONG_SEEK fseeko64 | |
9f95a23c | 250 | # define LONG_TELL ftello64 |
11fdf7f2 TL |
251 | #elif defined(_WIN32) && !defined(__DJGPP__) |
252 | # include <windows.h> | |
253 | static int LONG_SEEK(FILE* file, __int64 offset, int origin) { | |
254 | LARGE_INTEGER off; | |
255 | DWORD method; | |
256 | off.QuadPart = offset; | |
257 | if (origin == SEEK_END) | |
258 | method = FILE_END; | |
259 | else if (origin == SEEK_CUR) | |
260 | method = FILE_CURRENT; | |
261 | else | |
262 | method = FILE_BEGIN; | |
263 | ||
264 | if (SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, NULL, method)) | |
265 | return 0; | |
266 | else | |
267 | return -1; | |
268 | } | |
f67539c2 TL |
269 | static __int64 LONG_TELL(FILE* file) { |
270 | LARGE_INTEGER off, newOff; | |
271 | off.QuadPart = 0; | |
272 | newOff.QuadPart = 0; | |
273 | SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, &newOff, FILE_CURRENT); | |
274 | return newOff.QuadPart; | |
275 | } | |
11fdf7f2 TL |
276 | #else |
277 | # define LONG_SEEK fseek | |
9f95a23c | 278 | # define LONG_TELL ftell |
11fdf7f2 TL |
279 | #endif |
280 | ||
281 | ||
7c673cae | 282 | /*-************************************* |
f67539c2 | 283 | * Parameters: FIO_prefs_t |
9f95a23c TL |
284 | ***************************************/ |
285 | ||
f67539c2 | 286 | /* typedef'd to FIO_prefs_t within fileio.h */ |
9f95a23c TL |
287 | struct FIO_prefs_s { |
288 | ||
289 | /* Algorithm preferences */ | |
290 | FIO_compressionType_t compressionType; | |
291 | U32 sparseFileSupport; /* 0: no sparse allowed; 1: auto (file yes, stdout no); 2: force sparse */ | |
292 | int dictIDFlag; | |
293 | int checksumFlag; | |
294 | int blockSize; | |
295 | int overlapLog; | |
296 | U32 adaptiveMode; | |
297 | int rsyncable; | |
298 | int minAdaptLevel; | |
299 | int maxAdaptLevel; | |
300 | int ldmFlag; | |
301 | int ldmHashLog; | |
302 | int ldmMinMatch; | |
303 | int ldmBucketSizeLog; | |
304 | int ldmHashRateLog; | |
f67539c2 TL |
305 | size_t streamSrcSize; |
306 | size_t targetCBlockSize; | |
307 | int srcSizeHint; | |
308 | int testMode; | |
9f95a23c TL |
309 | ZSTD_literalCompressionMode_e literalCompressionMode; |
310 | ||
311 | /* IO preferences */ | |
312 | U32 removeSrcFile; | |
313 | U32 overwrite; | |
314 | ||
315 | /* Computation resources preferences */ | |
316 | unsigned memLimit; | |
317 | int nbWorkers; | |
f67539c2 TL |
318 | |
319 | int excludeCompressedFiles; | |
320 | int patchFromMode; | |
321 | int contentSize; | |
9f95a23c TL |
322 | }; |
323 | ||
324 | ||
325 | /*-************************************* | |
326 | * Parameters: Initialization | |
327 | ***************************************/ | |
328 | ||
329 | #define FIO_OVERLAP_LOG_NOTSET 9999 | |
330 | #define FIO_LDM_PARAM_NOTSET 9999 | |
331 | ||
332 | ||
333 | FIO_prefs_t* FIO_createPreferences(void) | |
334 | { | |
335 | FIO_prefs_t* const ret = (FIO_prefs_t*)malloc(sizeof(FIO_prefs_t)); | |
336 | if (!ret) EXM_THROW(21, "Allocation error : not enough memory"); | |
337 | ||
338 | ret->compressionType = FIO_zstdCompression; | |
339 | ret->overwrite = 0; | |
340 | ret->sparseFileSupport = ZSTD_SPARSE_DEFAULT; | |
341 | ret->dictIDFlag = 1; | |
342 | ret->checksumFlag = 1; | |
343 | ret->removeSrcFile = 0; | |
344 | ret->memLimit = 0; | |
345 | ret->nbWorkers = 1; | |
346 | ret->blockSize = 0; | |
347 | ret->overlapLog = FIO_OVERLAP_LOG_NOTSET; | |
348 | ret->adaptiveMode = 0; | |
349 | ret->rsyncable = 0; | |
350 | ret->minAdaptLevel = -50; /* initializing this value requires a constant, so ZSTD_minCLevel() doesn't work */ | |
351 | ret->maxAdaptLevel = 22; /* initializing this value requires a constant, so ZSTD_maxCLevel() doesn't work */ | |
352 | ret->ldmFlag = 0; | |
353 | ret->ldmHashLog = 0; | |
354 | ret->ldmMinMatch = 0; | |
355 | ret->ldmBucketSizeLog = FIO_LDM_PARAM_NOTSET; | |
356 | ret->ldmHashRateLog = FIO_LDM_PARAM_NOTSET; | |
f67539c2 TL |
357 | ret->streamSrcSize = 0; |
358 | ret->targetCBlockSize = 0; | |
359 | ret->srcSizeHint = 0; | |
360 | ret->testMode = 0; | |
9f95a23c | 361 | ret->literalCompressionMode = ZSTD_lcm_auto; |
f67539c2 | 362 | ret->excludeCompressedFiles = 0; |
9f95a23c TL |
363 | return ret; |
364 | } | |
365 | ||
366 | void FIO_freePreferences(FIO_prefs_t* const prefs) | |
367 | { | |
368 | free(prefs); | |
369 | } | |
370 | ||
371 | ||
372 | /*-************************************* | |
373 | * Parameters: Display Options | |
374 | ***************************************/ | |
375 | ||
376 | void FIO_setNotificationLevel(int level) { g_display_prefs.displayLevel=level; } | |
377 | ||
378 | void FIO_setNoProgress(unsigned noProgress) { g_display_prefs.noProgress = noProgress; } | |
379 | ||
380 | ||
381 | /*-************************************* | |
382 | * Parameters: Setters | |
7c673cae | 383 | ***************************************/ |
9f95a23c TL |
384 | |
385 | void FIO_setCompressionType(FIO_prefs_t* const prefs, FIO_compressionType_t compressionType) { prefs->compressionType = compressionType; } | |
386 | ||
387 | void FIO_overwriteMode(FIO_prefs_t* const prefs) { prefs->overwrite = 1; } | |
388 | ||
389 | void FIO_setSparseWrite(FIO_prefs_t* const prefs, unsigned sparse) { prefs->sparseFileSupport = sparse; } | |
390 | ||
391 | void FIO_setDictIDFlag(FIO_prefs_t* const prefs, int dictIDFlag) { prefs->dictIDFlag = dictIDFlag; } | |
392 | ||
393 | void FIO_setChecksumFlag(FIO_prefs_t* const prefs, int checksumFlag) { prefs->checksumFlag = checksumFlag; } | |
394 | ||
395 | void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, unsigned flag) { prefs->removeSrcFile = (flag>0); } | |
396 | ||
397 | void FIO_setMemLimit(FIO_prefs_t* const prefs, unsigned memLimit) { prefs->memLimit = memLimit; } | |
398 | ||
399 | void FIO_setNbWorkers(FIO_prefs_t* const prefs, int nbWorkers) { | |
11fdf7f2 | 400 | #ifndef ZSTD_MULTITHREAD |
9f95a23c | 401 | if (nbWorkers > 0) DISPLAYLEVEL(2, "Note : multi-threading is disabled \n"); |
11fdf7f2 | 402 | #endif |
9f95a23c | 403 | prefs->nbWorkers = nbWorkers; |
11fdf7f2 | 404 | } |
9f95a23c | 405 | |
f67539c2 TL |
406 | void FIO_setExcludeCompressedFile(FIO_prefs_t* const prefs, int excludeCompressedFiles) { prefs->excludeCompressedFiles = excludeCompressedFiles; } |
407 | ||
9f95a23c TL |
408 | void FIO_setBlockSize(FIO_prefs_t* const prefs, int blockSize) { |
409 | if (blockSize && prefs->nbWorkers==0) | |
11fdf7f2 | 410 | DISPLAYLEVEL(2, "Setting block size is useless in single-thread mode \n"); |
9f95a23c | 411 | prefs->blockSize = blockSize; |
11fdf7f2 | 412 | } |
9f95a23c TL |
413 | |
414 | void FIO_setOverlapLog(FIO_prefs_t* const prefs, int overlapLog){ | |
415 | if (overlapLog && prefs->nbWorkers==0) | |
11fdf7f2 | 416 | DISPLAYLEVEL(2, "Setting overlapLog is useless in single-thread mode \n"); |
9f95a23c TL |
417 | prefs->overlapLog = overlapLog; |
418 | } | |
419 | ||
420 | void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, unsigned adapt) { | |
421 | if ((adapt>0) && (prefs->nbWorkers==0)) | |
422 | EXM_THROW(1, "Adaptive mode is not compatible with single thread mode \n"); | |
423 | prefs->adaptiveMode = adapt; | |
424 | } | |
425 | ||
426 | void FIO_setRsyncable(FIO_prefs_t* const prefs, int rsyncable) { | |
427 | if ((rsyncable>0) && (prefs->nbWorkers==0)) | |
428 | EXM_THROW(1, "Rsyncable mode is not compatible with single thread mode \n"); | |
429 | prefs->rsyncable = rsyncable; | |
430 | } | |
431 | ||
f67539c2 TL |
432 | void FIO_setStreamSrcSize(FIO_prefs_t* const prefs, size_t streamSrcSize) { |
433 | prefs->streamSrcSize = streamSrcSize; | |
434 | } | |
435 | ||
436 | void FIO_setTargetCBlockSize(FIO_prefs_t* const prefs, size_t targetCBlockSize) { | |
437 | prefs->targetCBlockSize = targetCBlockSize; | |
438 | } | |
439 | ||
440 | void FIO_setSrcSizeHint(FIO_prefs_t* const prefs, size_t srcSizeHint) { | |
441 | prefs->srcSizeHint = (int)MIN((size_t)INT_MAX, srcSizeHint); | |
442 | } | |
443 | ||
444 | void FIO_setTestMode(FIO_prefs_t* const prefs, int testMode) { | |
445 | prefs->testMode = (testMode!=0); | |
446 | } | |
447 | ||
9f95a23c TL |
448 | void FIO_setLiteralCompressionMode( |
449 | FIO_prefs_t* const prefs, | |
450 | ZSTD_literalCompressionMode_e mode) { | |
451 | prefs->literalCompressionMode = mode; | |
452 | } | |
453 | ||
454 | void FIO_setAdaptMin(FIO_prefs_t* const prefs, int minCLevel) | |
455 | { | |
456 | #ifndef ZSTD_NOCOMPRESS | |
457 | assert(minCLevel >= ZSTD_minCLevel()); | |
458 | #endif | |
459 | prefs->minAdaptLevel = minCLevel; | |
11fdf7f2 | 460 | } |
9f95a23c TL |
461 | |
462 | void FIO_setAdaptMax(FIO_prefs_t* const prefs, int maxCLevel) | |
463 | { | |
464 | prefs->maxAdaptLevel = maxCLevel; | |
11fdf7f2 | 465 | } |
9f95a23c TL |
466 | |
467 | void FIO_setLdmFlag(FIO_prefs_t* const prefs, unsigned ldmFlag) { | |
468 | prefs->ldmFlag = (ldmFlag>0); | |
11fdf7f2 | 469 | } |
9f95a23c TL |
470 | |
471 | void FIO_setLdmHashLog(FIO_prefs_t* const prefs, int ldmHashLog) { | |
472 | prefs->ldmHashLog = ldmHashLog; | |
11fdf7f2 | 473 | } |
7c673cae | 474 | |
9f95a23c TL |
475 | void FIO_setLdmMinMatch(FIO_prefs_t* const prefs, int ldmMinMatch) { |
476 | prefs->ldmMinMatch = ldmMinMatch; | |
11fdf7f2 | 477 | } |
7c673cae | 478 | |
9f95a23c TL |
479 | void FIO_setLdmBucketSizeLog(FIO_prefs_t* const prefs, int ldmBucketSizeLog) { |
480 | prefs->ldmBucketSizeLog = ldmBucketSizeLog; | |
7c673cae FG |
481 | } |
482 | ||
483 | ||
9f95a23c TL |
484 | void FIO_setLdmHashRateLog(FIO_prefs_t* const prefs, int ldmHashRateLog) { |
485 | prefs->ldmHashRateLog = ldmHashRateLog; | |
486 | } | |
487 | ||
f67539c2 TL |
488 | void FIO_setPatchFromMode(FIO_prefs_t* const prefs, int value) |
489 | { | |
490 | prefs->patchFromMode = value != 0; | |
491 | } | |
492 | ||
493 | void FIO_setContentSize(FIO_prefs_t* const prefs, int value) | |
494 | { | |
495 | prefs->contentSize = value != 0; | |
496 | } | |
11fdf7f2 | 497 | |
7c673cae FG |
498 | /*-************************************* |
499 | * Functions | |
500 | ***************************************/ | |
11fdf7f2 TL |
501 | /** FIO_remove() : |
502 | * @result : Unlink `fileName`, even if it's read-only */ | |
503 | static int FIO_remove(const char* path) | |
504 | { | |
9f95a23c TL |
505 | if (!UTIL_isRegularFile(path)) { |
506 | DISPLAYLEVEL(2, "zstd: Refusing to remove non-regular file %s \n", path); | |
507 | return 0; | |
508 | } | |
11fdf7f2 TL |
509 | #if defined(_WIN32) || defined(WIN32) |
510 | /* windows doesn't allow remove read-only files, | |
511 | * so try to make it writable first */ | |
f67539c2 | 512 | UTIL_chmod(path, _S_IWRITE); |
11fdf7f2 TL |
513 | #endif |
514 | return remove(path); | |
515 | } | |
516 | ||
7c673cae | 517 | /** FIO_openSrcFile() : |
9f95a23c TL |
518 | * condition : `srcFileName` must be non-NULL. |
519 | * @result : FILE* to `srcFileName`, or NULL if it fails */ | |
7c673cae FG |
520 | static FILE* FIO_openSrcFile(const char* srcFileName) |
521 | { | |
9f95a23c | 522 | assert(srcFileName != NULL); |
7c673cae | 523 | if (!strcmp (srcFileName, stdinmark)) { |
9f95a23c | 524 | DISPLAYLEVEL(4,"Using stdin for input \n"); |
7c673cae | 525 | SET_BINARY_MODE(stdin); |
9f95a23c TL |
526 | return stdin; |
527 | } | |
528 | ||
529 | if (!UTIL_fileExist(srcFileName)) { | |
530 | DISPLAYLEVEL(1, "zstd: can't stat %s : %s -- ignored \n", | |
531 | srcFileName, strerror(errno)); | |
532 | return NULL; | |
7c673cae FG |
533 | } |
534 | ||
f67539c2 TL |
535 | if (!UTIL_isRegularFile(srcFileName) |
536 | && !UTIL_isFIFO(srcFileName) | |
537 | ) { | |
9f95a23c TL |
538 | DISPLAYLEVEL(1, "zstd: %s is not a regular file -- ignored \n", |
539 | srcFileName); | |
540 | return NULL; | |
541 | } | |
542 | ||
543 | { FILE* const f = fopen(srcFileName, "rb"); | |
544 | if (f == NULL) | |
545 | DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno)); | |
546 | return f; | |
547 | } | |
7c673cae FG |
548 | } |
549 | ||
550 | /** FIO_openDstFile() : | |
9f95a23c | 551 | * condition : `dstFileName` must be non-NULL. |
7c673cae | 552 | * @result : FILE* to `dstFileName`, or NULL if it fails */ |
f67539c2 TL |
553 | static FILE* |
554 | FIO_openDstFile(FIO_prefs_t* const prefs, | |
555 | const char* srcFileName, const char* dstFileName) | |
7c673cae | 556 | { |
f67539c2 TL |
557 | if (prefs->testMode) return NULL; /* do not open file in test mode */ |
558 | ||
9f95a23c | 559 | assert(dstFileName != NULL); |
7c673cae | 560 | if (!strcmp (dstFileName, stdoutmark)) { |
9f95a23c | 561 | DISPLAYLEVEL(4,"Using stdout for output \n"); |
7c673cae | 562 | SET_BINARY_MODE(stdout); |
9f95a23c TL |
563 | if (prefs->sparseFileSupport == 1) { |
564 | prefs->sparseFileSupport = 0; | |
7c673cae FG |
565 | DISPLAYLEVEL(4, "Sparse File Support is automatically disabled on stdout ; try --sparse \n"); |
566 | } | |
9f95a23c TL |
567 | return stdout; |
568 | } | |
569 | ||
570 | /* ensure dst is not the same as src */ | |
571 | if (srcFileName != NULL && UTIL_isSameFile(srcFileName, dstFileName)) { | |
572 | DISPLAYLEVEL(1, "zstd: Refusing to open an output file which will overwrite the input file \n"); | |
573 | return NULL; | |
574 | } | |
575 | ||
576 | if (prefs->sparseFileSupport == 1) { | |
577 | prefs->sparseFileSupport = ZSTD_SPARSE_DEFAULT; | |
578 | } | |
579 | ||
580 | if (UTIL_isRegularFile(dstFileName)) { | |
581 | /* Check if destination file already exists */ | |
582 | FILE* const fCheck = fopen( dstFileName, "rb" ); | |
f67539c2 TL |
583 | #if !defined(_WIN32) |
584 | /* this test does not work on Windows : | |
585 | * `NUL` and `nul` are detected as regular files */ | |
9f95a23c TL |
586 | if (!strcmp(dstFileName, nulmark)) { |
587 | EXM_THROW(40, "%s is unexpectedly categorized as a regular file", | |
588 | dstFileName); | |
11fdf7f2 | 589 | } |
f67539c2 | 590 | #endif |
9f95a23c TL |
591 | if (fCheck != NULL) { /* dst file exists, authorization prompt */ |
592 | fclose(fCheck); | |
593 | if (!prefs->overwrite) { | |
594 | if (g_display_prefs.displayLevel <= 1) { | |
595 | /* No interaction possible */ | |
596 | DISPLAY("zstd: %s already exists; not overwritten \n", | |
597 | dstFileName); | |
598 | return NULL; | |
599 | } | |
600 | DISPLAY("zstd: %s already exists; overwrite (y/N) ? ", | |
601 | dstFileName); | |
602 | { int ch = getchar(); | |
603 | if ((ch!='Y') && (ch!='y')) { | |
604 | DISPLAY(" not overwritten \n"); | |
7c673cae FG |
605 | return NULL; |
606 | } | |
9f95a23c TL |
607 | /* flush rest of input line */ |
608 | while ((ch!=EOF) && (ch!='\n')) ch = getchar(); | |
609 | } } | |
610 | /* need to unlink */ | |
611 | FIO_remove(dstFileName); | |
612 | } } | |
7c673cae | 613 | |
9f95a23c | 614 | { FILE* const f = fopen( dstFileName, "wb" ); |
f67539c2 | 615 | if (f == NULL) { |
9f95a23c | 616 | DISPLAYLEVEL(1, "zstd: %s: %s\n", dstFileName, strerror(errno)); |
f67539c2 TL |
617 | } else if (srcFileName != NULL |
618 | && strcmp (srcFileName, stdinmark) | |
619 | && strcmp(dstFileName, nulmark) ) { | |
620 | /* reduce rights on newly created dst file while compression is ongoing */ | |
621 | UTIL_chmod(dstFileName, 00600); | |
622 | } | |
9f95a23c TL |
623 | return f; |
624 | } | |
7c673cae FG |
625 | } |
626 | ||
627 | ||
11fdf7f2 TL |
628 | /*! FIO_createDictBuffer() : |
629 | * creates a buffer, pointed by `*bufferPtr`, | |
630 | * loads `filename` content into it, up to DICTSIZE_MAX bytes. | |
9f95a23c | 631 | * @return : loaded size |
11fdf7f2 TL |
632 | * if fileName==NULL, returns 0 and a NULL pointer |
633 | */ | |
f67539c2 | 634 | static size_t FIO_createDictBuffer(void** bufferPtr, const char* fileName, FIO_prefs_t* const prefs) |
7c673cae FG |
635 | { |
636 | FILE* fileHandle; | |
637 | U64 fileSize; | |
638 | ||
9f95a23c | 639 | assert(bufferPtr != NULL); |
7c673cae FG |
640 | *bufferPtr = NULL; |
641 | if (fileName == NULL) return 0; | |
642 | ||
643 | DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName); | |
644 | fileHandle = fopen(fileName, "rb"); | |
9f95a23c TL |
645 | if (fileHandle==NULL) EXM_THROW(31, "%s: %s", fileName, strerror(errno)); |
646 | ||
7c673cae | 647 | fileSize = UTIL_getFileSize(fileName); |
f67539c2 TL |
648 | { |
649 | size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX; | |
650 | if (fileSize > dictSizeMax) { | |
651 | EXM_THROW(32, "Dictionary file %s is too large (> %u bytes)", | |
652 | fileName, (unsigned)dictSizeMax); /* avoid extreme cases */ | |
653 | } | |
7c673cae FG |
654 | } |
655 | *bufferPtr = malloc((size_t)fileSize); | |
11fdf7f2 TL |
656 | if (*bufferPtr==NULL) EXM_THROW(34, "%s", strerror(errno)); |
657 | { size_t const readSize = fread(*bufferPtr, 1, (size_t)fileSize, fileHandle); | |
9f95a23c TL |
658 | if (readSize != fileSize) |
659 | EXM_THROW(35, "Error reading dictionary file %s : %s", | |
660 | fileName, strerror(errno)); | |
11fdf7f2 | 661 | } |
7c673cae FG |
662 | fclose(fileHandle); |
663 | return (size_t)fileSize; | |
664 | } | |
665 | ||
f67539c2 TL |
666 | |
667 | ||
668 | /* FIO_checkFilenameCollisions() : | |
669 | * Checks for and warns if there are any files that would have the same output path | |
670 | */ | |
671 | int FIO_checkFilenameCollisions(const char** filenameTable, unsigned nbFiles) { | |
672 | const char **filenameTableSorted, *c, *prevElem, *filename; | |
673 | unsigned u; | |
674 | ||
675 | #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__) /* windows support */ | |
676 | c = "\\"; | |
677 | #else | |
678 | c = "/"; | |
679 | #endif | |
680 | ||
681 | filenameTableSorted = (const char**) malloc(sizeof(char*) * nbFiles); | |
682 | if (!filenameTableSorted) { | |
683 | DISPLAY("Unable to malloc new str array, not checking for name collisions\n"); | |
684 | return 1; | |
685 | } | |
686 | ||
687 | for (u = 0; u < nbFiles; ++u) { | |
688 | filename = strrchr(filenameTable[u], c[0]); | |
689 | if (filename == NULL) { | |
690 | filenameTableSorted[u] = filenameTable[u]; | |
691 | } else { | |
692 | filenameTableSorted[u] = filename+1; | |
693 | } | |
694 | } | |
695 | ||
696 | qsort((void*)filenameTableSorted, nbFiles, sizeof(char*), UTIL_compareStr); | |
697 | prevElem = filenameTableSorted[0]; | |
698 | for (u = 1; u < nbFiles; ++u) { | |
699 | if (strcmp(prevElem, filenameTableSorted[u]) == 0) { | |
700 | DISPLAY("WARNING: Two files have same filename: %s\n", prevElem); | |
701 | } | |
702 | prevElem = filenameTableSorted[u]; | |
703 | } | |
704 | ||
705 | free((void*)filenameTableSorted); | |
706 | return 0; | |
707 | } | |
708 | ||
709 | static const char* | |
710 | extractFilename(const char* path, char separator) | |
711 | { | |
712 | const char* search = strrchr(path, separator); | |
713 | if (search == NULL) return path; | |
714 | return search+1; | |
715 | } | |
716 | ||
717 | /* FIO_createFilename_fromOutDir() : | |
718 | * Takes a source file name and specified output directory, and | |
719 | * allocates memory for and returns a pointer to final path. | |
720 | * This function never returns an error (it may abort() in case of pb) | |
721 | */ | |
722 | static char* | |
723 | FIO_createFilename_fromOutDir(const char* path, const char* outDirName, const size_t suffixLen) | |
724 | { | |
725 | const char* filenameStart; | |
726 | char separator; | |
727 | char* result; | |
728 | ||
729 | #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__) /* windows support */ | |
730 | separator = '\\'; | |
731 | #else | |
732 | separator = '/'; | |
733 | #endif | |
734 | ||
735 | filenameStart = extractFilename(path, separator); | |
736 | #if defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__) /* windows support */ | |
737 | filenameStart = extractFilename(filenameStart, '/'); /* sometimes, '/' separator is also used on Windows (mingw+msys2) */ | |
738 | #endif | |
739 | ||
740 | result = (char*) calloc(1, strlen(outDirName) + 1 + strlen(filenameStart) + suffixLen + 1); | |
741 | if (!result) { | |
742 | EXM_THROW(30, "zstd: FIO_createFilename_fromOutDir: %s", strerror(errno)); | |
743 | } | |
744 | ||
745 | memcpy(result, outDirName, strlen(outDirName)); | |
746 | if (outDirName[strlen(outDirName)-1] == separator) { | |
747 | memcpy(result + strlen(outDirName), filenameStart, strlen(filenameStart)); | |
748 | } else { | |
749 | memcpy(result + strlen(outDirName), &separator, 1); | |
750 | memcpy(result + strlen(outDirName) + 1, filenameStart, strlen(filenameStart)); | |
751 | } | |
752 | ||
753 | return result; | |
754 | } | |
755 | ||
756 | /* FIO_highbit64() : | |
757 | * gives position of highest bit. | |
758 | * note : only works for v > 0 ! | |
759 | */ | |
760 | static unsigned FIO_highbit64(unsigned long long v) | |
761 | { | |
762 | unsigned count = 0; | |
763 | assert(v != 0); | |
764 | v >>= 1; | |
765 | while (v) { v >>= 1; count++; } | |
766 | return count; | |
767 | } | |
768 | ||
769 | static void FIO_adjustMemLimitForPatchFromMode(FIO_prefs_t* const prefs, | |
770 | unsigned long long const dictSize, | |
771 | unsigned long long const maxSrcFileSize) | |
772 | { | |
773 | unsigned long long maxSize = MAX(prefs->memLimit, MAX(dictSize, maxSrcFileSize)); | |
774 | assert(maxSize != UTIL_FILESIZE_UNKNOWN); | |
775 | if (maxSize > UINT_MAX) | |
776 | EXM_THROW(42, "Can't handle files larger than %u GB\n", UINT_MAX/(1 GB) + 1); | |
777 | FIO_setMemLimit(prefs, (unsigned)maxSize); | |
778 | } | |
779 | ||
7c673cae FG |
780 | #ifndef ZSTD_NOCOMPRESS |
781 | ||
9f95a23c TL |
782 | /* ********************************************************************** |
783 | * Compression | |
784 | ************************************************************************/ | |
7c673cae | 785 | typedef struct { |
11fdf7f2 TL |
786 | FILE* srcFile; |
787 | FILE* dstFile; | |
7c673cae FG |
788 | void* srcBuffer; |
789 | size_t srcBufferSize; | |
790 | void* dstBuffer; | |
791 | size_t dstBufferSize; | |
f67539c2 TL |
792 | void* dictBuffer; |
793 | size_t dictBufferSize; | |
9f95a23c | 794 | const char* dictFileName; |
7c673cae | 795 | ZSTD_CStream* cctx; |
7c673cae FG |
796 | } cRess_t; |
797 | ||
f67539c2 TL |
798 | static void FIO_adjustParamsForPatchFromMode(FIO_prefs_t* const prefs, |
799 | ZSTD_compressionParameters* comprParams, | |
800 | unsigned long long const dictSize, | |
801 | unsigned long long const maxSrcFileSize, | |
802 | int cLevel) | |
803 | { | |
804 | unsigned const fileWindowLog = FIO_highbit64(maxSrcFileSize) + 1; | |
805 | ZSTD_compressionParameters const cParams = ZSTD_getCParams(cLevel, (size_t)maxSrcFileSize, (size_t)dictSize); | |
806 | FIO_adjustMemLimitForPatchFromMode(prefs, dictSize, maxSrcFileSize); | |
807 | if (fileWindowLog > ZSTD_WINDOWLOG_MAX) | |
808 | DISPLAYLEVEL(1, "Max window log exceeded by file (compression ratio will suffer)\n"); | |
809 | comprParams->windowLog = MIN(ZSTD_WINDOWLOG_MAX, fileWindowLog); | |
810 | if (fileWindowLog > ZSTD_cycleLog(cParams.hashLog, cParams.strategy)) { | |
811 | if (!prefs->ldmFlag) | |
812 | DISPLAYLEVEL(1, "long mode automaticaly triggered\n"); | |
813 | FIO_setLdmFlag(prefs, 1); | |
814 | } | |
815 | if (cParams.strategy >= ZSTD_btopt) { | |
816 | DISPLAYLEVEL(1, "[Optimal parser notes] Consider the following to improve patch size at the cost of speed:\n"); | |
817 | DISPLAYLEVEL(1, "- Use --single-thread mode in the zstd cli\n"); | |
818 | DISPLAYLEVEL(1, "- Set a larger targetLength (eg. --zstd=targetLength=4096)\n"); | |
819 | DISPLAYLEVEL(1, "- Set a larger chainLog (eg. --zstd=chainLog=%u)\n", ZSTD_CHAINLOG_MAX); | |
820 | DISPLAYLEVEL(1, "Also consdier playing around with searchLog and hashLog\n"); | |
821 | } | |
822 | } | |
823 | ||
9f95a23c | 824 | static cRess_t FIO_createCResources(FIO_prefs_t* const prefs, |
f67539c2 TL |
825 | const char* dictFileName, unsigned long long const maxSrcFileSize, |
826 | int cLevel, ZSTD_compressionParameters comprParams) { | |
7c673cae FG |
827 | cRess_t ress; |
828 | memset(&ress, 0, sizeof(ress)); | |
829 | ||
9f95a23c | 830 | DISPLAYLEVEL(6, "FIO_createCResources \n"); |
11fdf7f2 TL |
831 | ress.cctx = ZSTD_createCCtx(); |
832 | if (ress.cctx == NULL) | |
9f95a23c TL |
833 | EXM_THROW(30, "allocation error (%s): can't create ZSTD_CCtx", |
834 | strerror(errno)); | |
7c673cae FG |
835 | ress.srcBufferSize = ZSTD_CStreamInSize(); |
836 | ress.srcBuffer = malloc(ress.srcBufferSize); | |
837 | ress.dstBufferSize = ZSTD_CStreamOutSize(); | |
f67539c2 TL |
838 | |
839 | /* need to update memLimit before calling createDictBuffer | |
840 | * because of memLimit check inside it */ | |
841 | if (prefs->patchFromMode) | |
842 | FIO_adjustParamsForPatchFromMode(prefs, &comprParams, UTIL_getFileSize(dictFileName), maxSrcFileSize, cLevel); | |
7c673cae | 843 | ress.dstBuffer = malloc(ress.dstBufferSize); |
f67539c2 | 844 | ress.dictBufferSize = FIO_createDictBuffer(&ress.dictBuffer, dictFileName, prefs); /* works with dictFileName==NULL */ |
11fdf7f2 TL |
845 | if (!ress.srcBuffer || !ress.dstBuffer) |
846 | EXM_THROW(31, "allocation error : not enough memory"); | |
7c673cae | 847 | |
9f95a23c | 848 | /* Advanced parameters, including dictionary */ |
f67539c2 TL |
849 | if (dictFileName && (ress.dictBuffer==NULL)) |
850 | EXM_THROW(32, "allocation error : can't create dictBuffer"); | |
851 | ress.dictFileName = dictFileName; | |
852 | ||
853 | if (prefs->adaptiveMode && !prefs->ldmFlag && !comprParams.windowLog) | |
854 | comprParams.windowLog = ADAPT_WINDOWLOG_DEFAULT; | |
855 | ||
856 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_contentSizeFlag, prefs->contentSize) ); /* always enable content size when available (note: supposed to be default) */ | |
857 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_dictIDFlag, prefs->dictIDFlag) ); | |
858 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_checksumFlag, prefs->checksumFlag) ); | |
859 | /* compression level */ | |
860 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, cLevel) ); | |
861 | /* max compressed block size */ | |
862 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetCBlockSize, (int)prefs->targetCBlockSize) ); | |
863 | /* source size hint */ | |
864 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_srcSizeHint, (int)prefs->srcSizeHint) ); | |
865 | /* long distance matching */ | |
866 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_enableLongDistanceMatching, prefs->ldmFlag) ); | |
867 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashLog, prefs->ldmHashLog) ); | |
868 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmMinMatch, prefs->ldmMinMatch) ); | |
869 | if (prefs->ldmBucketSizeLog != FIO_LDM_PARAM_NOTSET) { | |
870 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmBucketSizeLog, prefs->ldmBucketSizeLog) ); | |
871 | } | |
872 | if (prefs->ldmHashRateLog != FIO_LDM_PARAM_NOTSET) { | |
873 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashRateLog, prefs->ldmHashRateLog) ); | |
874 | } | |
875 | /* compression parameters */ | |
876 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_windowLog, (int)comprParams.windowLog) ); | |
877 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_chainLog, (int)comprParams.chainLog) ); | |
878 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_hashLog, (int)comprParams.hashLog) ); | |
879 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_searchLog, (int)comprParams.searchLog) ); | |
880 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_minMatch, (int)comprParams.minMatch) ); | |
881 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetLength, (int)comprParams.targetLength) ); | |
882 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_strategy, comprParams.strategy) ); | |
883 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_literalCompressionMode, (int)prefs->literalCompressionMode) ); | |
884 | /* multi-threading */ | |
9f95a23c | 885 | #ifdef ZSTD_MULTITHREAD |
f67539c2 TL |
886 | DISPLAYLEVEL(5,"set nb workers = %u \n", prefs->nbWorkers); |
887 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_nbWorkers, prefs->nbWorkers) ); | |
888 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_jobSize, prefs->blockSize) ); | |
889 | if (prefs->overlapLog != FIO_OVERLAP_LOG_NOTSET) { | |
890 | DISPLAYLEVEL(3,"set overlapLog = %u \n", prefs->overlapLog); | |
891 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_overlapLog, prefs->overlapLog) ); | |
892 | } | |
893 | CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_rsyncable, prefs->rsyncable) ); | |
11fdf7f2 | 894 | #endif |
f67539c2 TL |
895 | /* dictionary */ |
896 | if (prefs->patchFromMode) { | |
897 | CHECK( ZSTD_CCtx_refPrefix(ress.cctx, ress.dictBuffer, ress.dictBufferSize) ); | |
898 | } else { | |
899 | CHECK( ZSTD_CCtx_loadDictionary(ress.cctx, ress.dictBuffer, ress.dictBufferSize) ); | |
7c673cae FG |
900 | } |
901 | ||
902 | return ress; | |
903 | } | |
904 | ||
905 | static void FIO_freeCResources(cRess_t ress) | |
906 | { | |
907 | free(ress.srcBuffer); | |
908 | free(ress.dstBuffer); | |
f67539c2 | 909 | free(ress.dictBuffer); |
7c673cae FG |
910 | ZSTD_freeCStream(ress.cctx); /* never fails */ |
911 | } | |
912 | ||
913 | ||
11fdf7f2 | 914 | #ifdef ZSTD_GZCOMPRESS |
9f95a23c | 915 | static unsigned long long |
f67539c2 | 916 | FIO_compressGzFrame(const cRess_t* ress, /* buffers & handlers are used, but not changed */ |
11fdf7f2 TL |
917 | const char* srcFileName, U64 const srcFileSize, |
918 | int compressionLevel, U64* readsize) | |
919 | { | |
920 | unsigned long long inFileSize = 0, outFileSize = 0; | |
921 | z_stream strm; | |
11fdf7f2 TL |
922 | |
923 | if (compressionLevel > Z_BEST_COMPRESSION) | |
924 | compressionLevel = Z_BEST_COMPRESSION; | |
925 | ||
926 | strm.zalloc = Z_NULL; | |
927 | strm.zfree = Z_NULL; | |
928 | strm.opaque = Z_NULL; | |
929 | ||
f67539c2 | 930 | { int const ret = deflateInit2(&strm, compressionLevel, Z_DEFLATED, |
11fdf7f2 TL |
931 | 15 /* maxWindowLogSize */ + 16 /* gzip only */, |
932 | 8, Z_DEFAULT_STRATEGY); /* see http://www.zlib.net/manual.html */ | |
f67539c2 TL |
933 | if (ret != Z_OK) { |
934 | EXM_THROW(71, "zstd: %s: deflateInit2 error %d \n", srcFileName, ret); | |
935 | } } | |
11fdf7f2 TL |
936 | |
937 | strm.next_in = 0; | |
938 | strm.avail_in = 0; | |
939 | strm.next_out = (Bytef*)ress->dstBuffer; | |
940 | strm.avail_out = (uInt)ress->dstBufferSize; | |
941 | ||
942 | while (1) { | |
f67539c2 | 943 | int ret; |
11fdf7f2 TL |
944 | if (strm.avail_in == 0) { |
945 | size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile); | |
946 | if (inSize == 0) break; | |
947 | inFileSize += inSize; | |
948 | strm.next_in = (z_const unsigned char*)ress->srcBuffer; | |
949 | strm.avail_in = (uInt)inSize; | |
950 | } | |
951 | ret = deflate(&strm, Z_NO_FLUSH); | |
952 | if (ret != Z_OK) | |
953 | EXM_THROW(72, "zstd: %s: deflate error %d \n", srcFileName, ret); | |
f67539c2 TL |
954 | { size_t const cSize = ress->dstBufferSize - strm.avail_out; |
955 | if (cSize) { | |
956 | if (fwrite(ress->dstBuffer, 1, cSize, ress->dstFile) != cSize) | |
957 | EXM_THROW(73, "Write error : cannot write to output file : %s ", strerror(errno)); | |
958 | outFileSize += cSize; | |
11fdf7f2 TL |
959 | strm.next_out = (Bytef*)ress->dstBuffer; |
960 | strm.avail_out = (uInt)ress->dstBufferSize; | |
f67539c2 TL |
961 | } } |
962 | if (srcFileSize == UTIL_FILESIZE_UNKNOWN) { | |
963 | DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%% ", | |
9f95a23c | 964 | (unsigned)(inFileSize>>20), |
11fdf7f2 | 965 | (double)outFileSize/inFileSize*100) |
f67539c2 TL |
966 | } else { |
967 | DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%% ", | |
9f95a23c | 968 | (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20), |
11fdf7f2 | 969 | (double)outFileSize/inFileSize*100); |
f67539c2 | 970 | } } |
11fdf7f2 TL |
971 | |
972 | while (1) { | |
f67539c2 TL |
973 | int const ret = deflate(&strm, Z_FINISH); |
974 | { size_t const cSize = ress->dstBufferSize - strm.avail_out; | |
975 | if (cSize) { | |
976 | if (fwrite(ress->dstBuffer, 1, cSize, ress->dstFile) != cSize) | |
977 | EXM_THROW(75, "Write error : %s ", strerror(errno)); | |
978 | outFileSize += cSize; | |
11fdf7f2 TL |
979 | strm.next_out = (Bytef*)ress->dstBuffer; |
980 | strm.avail_out = (uInt)ress->dstBufferSize; | |
981 | } } | |
982 | if (ret == Z_STREAM_END) break; | |
983 | if (ret != Z_BUF_ERROR) | |
984 | EXM_THROW(77, "zstd: %s: deflate error %d \n", srcFileName, ret); | |
985 | } | |
986 | ||
f67539c2 TL |
987 | { int const ret = deflateEnd(&strm); |
988 | if (ret != Z_OK) { | |
989 | EXM_THROW(79, "zstd: %s: deflateEnd error %d \n", srcFileName, ret); | |
990 | } } | |
11fdf7f2 | 991 | *readsize = inFileSize; |
11fdf7f2 TL |
992 | return outFileSize; |
993 | } | |
994 | #endif | |
995 | ||
996 | ||
997 | #ifdef ZSTD_LZMACOMPRESS | |
9f95a23c TL |
998 | static unsigned long long |
999 | FIO_compressLzmaFrame(cRess_t* ress, | |
1000 | const char* srcFileName, U64 const srcFileSize, | |
1001 | int compressionLevel, U64* readsize, int plain_lzma) | |
11fdf7f2 TL |
1002 | { |
1003 | unsigned long long inFileSize = 0, outFileSize = 0; | |
1004 | lzma_stream strm = LZMA_STREAM_INIT; | |
1005 | lzma_action action = LZMA_RUN; | |
1006 | lzma_ret ret; | |
1007 | ||
1008 | if (compressionLevel < 0) compressionLevel = 0; | |
1009 | if (compressionLevel > 9) compressionLevel = 9; | |
1010 | ||
1011 | if (plain_lzma) { | |
1012 | lzma_options_lzma opt_lzma; | |
1013 | if (lzma_lzma_preset(&opt_lzma, compressionLevel)) | |
f67539c2 | 1014 | EXM_THROW(81, "zstd: %s: lzma_lzma_preset error", srcFileName); |
11fdf7f2 TL |
1015 | ret = lzma_alone_encoder(&strm, &opt_lzma); /* LZMA */ |
1016 | if (ret != LZMA_OK) | |
f67539c2 | 1017 | EXM_THROW(82, "zstd: %s: lzma_alone_encoder error %d", srcFileName, ret); |
11fdf7f2 TL |
1018 | } else { |
1019 | ret = lzma_easy_encoder(&strm, compressionLevel, LZMA_CHECK_CRC64); /* XZ */ | |
1020 | if (ret != LZMA_OK) | |
f67539c2 | 1021 | EXM_THROW(83, "zstd: %s: lzma_easy_encoder error %d", srcFileName, ret); |
11fdf7f2 TL |
1022 | } |
1023 | ||
1024 | strm.next_in = 0; | |
1025 | strm.avail_in = 0; | |
1026 | strm.next_out = (BYTE*)ress->dstBuffer; | |
1027 | strm.avail_out = ress->dstBufferSize; | |
1028 | ||
1029 | while (1) { | |
1030 | if (strm.avail_in == 0) { | |
1031 | size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile); | |
1032 | if (inSize == 0) action = LZMA_FINISH; | |
1033 | inFileSize += inSize; | |
1034 | strm.next_in = (BYTE const*)ress->srcBuffer; | |
1035 | strm.avail_in = inSize; | |
1036 | } | |
1037 | ||
1038 | ret = lzma_code(&strm, action); | |
1039 | ||
1040 | if (ret != LZMA_OK && ret != LZMA_STREAM_END) | |
f67539c2 | 1041 | EXM_THROW(84, "zstd: %s: lzma_code encoding error %d", srcFileName, ret); |
11fdf7f2 TL |
1042 | { size_t const compBytes = ress->dstBufferSize - strm.avail_out; |
1043 | if (compBytes) { | |
1044 | if (fwrite(ress->dstBuffer, 1, compBytes, ress->dstFile) != compBytes) | |
f67539c2 | 1045 | EXM_THROW(85, "Write error : %s", strerror(errno)); |
11fdf7f2 TL |
1046 | outFileSize += compBytes; |
1047 | strm.next_out = (BYTE*)ress->dstBuffer; | |
1048 | strm.avail_out = ress->dstBufferSize; | |
1049 | } } | |
9f95a23c | 1050 | if (srcFileSize == UTIL_FILESIZE_UNKNOWN) |
11fdf7f2 | 1051 | DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%", |
9f95a23c | 1052 | (unsigned)(inFileSize>>20), |
11fdf7f2 TL |
1053 | (double)outFileSize/inFileSize*100) |
1054 | else | |
1055 | DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%", | |
9f95a23c | 1056 | (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20), |
11fdf7f2 TL |
1057 | (double)outFileSize/inFileSize*100); |
1058 | if (ret == LZMA_STREAM_END) break; | |
1059 | } | |
1060 | ||
1061 | lzma_end(&strm); | |
1062 | *readsize = inFileSize; | |
1063 | ||
1064 | return outFileSize; | |
1065 | } | |
1066 | #endif | |
1067 | ||
1068 | #ifdef ZSTD_LZ4COMPRESS | |
9f95a23c TL |
1069 | |
1070 | #if LZ4_VERSION_NUMBER <= 10600 | |
1071 | #define LZ4F_blockLinked blockLinked | |
1072 | #define LZ4F_max64KB max64KB | |
1073 | #endif | |
1074 | ||
11fdf7f2 | 1075 | static int FIO_LZ4_GetBlockSize_FromBlockId (int id) { return (1 << (8 + (2 * id))); } |
9f95a23c TL |
1076 | |
1077 | static unsigned long long | |
1078 | FIO_compressLz4Frame(cRess_t* ress, | |
1079 | const char* srcFileName, U64 const srcFileSize, | |
1080 | int compressionLevel, int checksumFlag, | |
1081 | U64* readsize) | |
11fdf7f2 | 1082 | { |
9f95a23c | 1083 | const size_t blockSize = FIO_LZ4_GetBlockSize_FromBlockId(LZ4F_max64KB); |
11fdf7f2 TL |
1084 | unsigned long long inFileSize = 0, outFileSize = 0; |
1085 | ||
1086 | LZ4F_preferences_t prefs; | |
1087 | LZ4F_compressionContext_t ctx; | |
1088 | ||
1089 | LZ4F_errorCode_t const errorCode = LZ4F_createCompressionContext(&ctx, LZ4F_VERSION); | |
1090 | if (LZ4F_isError(errorCode)) | |
1091 | EXM_THROW(31, "zstd: failed to create lz4 compression context"); | |
1092 | ||
1093 | memset(&prefs, 0, sizeof(prefs)); | |
1094 | ||
9f95a23c | 1095 | assert(blockSize <= ress->srcBufferSize); |
11fdf7f2 TL |
1096 | |
1097 | prefs.autoFlush = 1; | |
1098 | prefs.compressionLevel = compressionLevel; | |
9f95a23c TL |
1099 | prefs.frameInfo.blockMode = LZ4F_blockLinked; |
1100 | prefs.frameInfo.blockSizeID = LZ4F_max64KB; | |
1101 | prefs.frameInfo.contentChecksumFlag = (contentChecksum_t)checksumFlag; | |
11fdf7f2 | 1102 | #if LZ4_VERSION_NUMBER >= 10600 |
9f95a23c | 1103 | prefs.frameInfo.contentSize = (srcFileSize==UTIL_FILESIZE_UNKNOWN) ? 0 : srcFileSize; |
11fdf7f2 | 1104 | #endif |
9f95a23c | 1105 | assert(LZ4F_compressBound(blockSize, &prefs) <= ress->dstBufferSize); |
11fdf7f2 TL |
1106 | |
1107 | { | |
11fdf7f2 TL |
1108 | size_t readSize; |
1109 | size_t headerSize = LZ4F_compressBegin(ctx, ress->dstBuffer, ress->dstBufferSize, &prefs); | |
1110 | if (LZ4F_isError(headerSize)) | |
1111 | EXM_THROW(33, "File header generation failed : %s", | |
1112 | LZ4F_getErrorName(headerSize)); | |
9f95a23c TL |
1113 | if (fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile) != headerSize) |
1114 | EXM_THROW(34, "Write error : %s (cannot write header)", strerror(errno)); | |
11fdf7f2 TL |
1115 | outFileSize += headerSize; |
1116 | ||
1117 | /* Read first block */ | |
1118 | readSize = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile); | |
1119 | inFileSize += readSize; | |
1120 | ||
1121 | /* Main Loop */ | |
1122 | while (readSize>0) { | |
9f95a23c TL |
1123 | size_t const outSize = LZ4F_compressUpdate(ctx, |
1124 | ress->dstBuffer, ress->dstBufferSize, | |
1125 | ress->srcBuffer, readSize, NULL); | |
11fdf7f2 TL |
1126 | if (LZ4F_isError(outSize)) |
1127 | EXM_THROW(35, "zstd: %s: lz4 compression failed : %s", | |
1128 | srcFileName, LZ4F_getErrorName(outSize)); | |
1129 | outFileSize += outSize; | |
9f95a23c | 1130 | if (srcFileSize == UTIL_FILESIZE_UNKNOWN) { |
11fdf7f2 | 1131 | DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%", |
9f95a23c | 1132 | (unsigned)(inFileSize>>20), |
11fdf7f2 | 1133 | (double)outFileSize/inFileSize*100) |
9f95a23c | 1134 | } else { |
11fdf7f2 | 1135 | DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%", |
9f95a23c | 1136 | (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20), |
11fdf7f2 | 1137 | (double)outFileSize/inFileSize*100); |
9f95a23c | 1138 | } |
11fdf7f2 TL |
1139 | |
1140 | /* Write Block */ | |
9f95a23c TL |
1141 | { size_t const sizeCheck = fwrite(ress->dstBuffer, 1, outSize, ress->dstFile); |
1142 | if (sizeCheck != outSize) | |
1143 | EXM_THROW(36, "Write error : %s", strerror(errno)); | |
1144 | } | |
11fdf7f2 TL |
1145 | |
1146 | /* Read next block */ | |
1147 | readSize = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile); | |
1148 | inFileSize += readSize; | |
1149 | } | |
1150 | if (ferror(ress->srcFile)) EXM_THROW(37, "Error reading %s ", srcFileName); | |
1151 | ||
1152 | /* End of Stream mark */ | |
1153 | headerSize = LZ4F_compressEnd(ctx, ress->dstBuffer, ress->dstBufferSize, NULL); | |
1154 | if (LZ4F_isError(headerSize)) | |
1155 | EXM_THROW(38, "zstd: %s: lz4 end of file generation failed : %s", | |
1156 | srcFileName, LZ4F_getErrorName(headerSize)); | |
1157 | ||
9f95a23c TL |
1158 | { size_t const sizeCheck = fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile); |
1159 | if (sizeCheck != headerSize) | |
1160 | EXM_THROW(39, "Write error : %s (cannot write end of stream)", | |
1161 | strerror(errno)); | |
1162 | } | |
11fdf7f2 TL |
1163 | outFileSize += headerSize; |
1164 | } | |
1165 | ||
1166 | *readsize = inFileSize; | |
1167 | LZ4F_freeCompressionContext(ctx); | |
1168 | ||
1169 | return outFileSize; | |
1170 | } | |
1171 | #endif | |
1172 | ||
1173 | ||
9f95a23c TL |
1174 | static unsigned long long |
1175 | FIO_compressZstdFrame(FIO_prefs_t* const prefs, | |
1176 | const cRess_t* ressPtr, | |
1177 | const char* srcFileName, U64 fileSize, | |
1178 | int compressionLevel, U64* readsize) | |
1179 | { | |
1180 | cRess_t const ress = *ressPtr; | |
1181 | FILE* const srcFile = ress.srcFile; | |
1182 | FILE* const dstFile = ress.dstFile; | |
1183 | U64 compressedfilesize = 0; | |
1184 | ZSTD_EndDirective directive = ZSTD_e_continue; | |
1185 | ||
1186 | /* stats */ | |
1187 | ZSTD_frameProgression previous_zfp_update = { 0, 0, 0, 0, 0, 0 }; | |
1188 | ZSTD_frameProgression previous_zfp_correction = { 0, 0, 0, 0, 0, 0 }; | |
1189 | typedef enum { noChange, slower, faster } speedChange_e; | |
1190 | speedChange_e speedChange = noChange; | |
1191 | unsigned flushWaiting = 0; | |
1192 | unsigned inputPresented = 0; | |
1193 | unsigned inputBlocked = 0; | |
1194 | unsigned lastJobID = 0; | |
1195 | ||
1196 | DISPLAYLEVEL(6, "compression using zstd format \n"); | |
1197 | ||
1198 | /* init */ | |
1199 | if (fileSize != UTIL_FILESIZE_UNKNOWN) { | |
1200 | CHECK(ZSTD_CCtx_setPledgedSrcSize(ress.cctx, fileSize)); | |
f67539c2 TL |
1201 | } else if (prefs->streamSrcSize > 0) { |
1202 | /* unknown source size; use the declared stream size */ | |
1203 | CHECK( ZSTD_CCtx_setPledgedSrcSize(ress.cctx, prefs->streamSrcSize) ); | |
9f95a23c TL |
1204 | } |
1205 | (void)srcFileName; | |
1206 | ||
1207 | /* Main compression loop */ | |
1208 | do { | |
1209 | size_t stillToFlush; | |
1210 | /* Fill input Buffer */ | |
1211 | size_t const inSize = fread(ress.srcBuffer, (size_t)1, ress.srcBufferSize, srcFile); | |
1212 | ZSTD_inBuffer inBuff = { ress.srcBuffer, inSize, 0 }; | |
1213 | DISPLAYLEVEL(6, "fread %u bytes from source \n", (unsigned)inSize); | |
1214 | *readsize += inSize; | |
1215 | ||
1216 | if ((inSize == 0) || (*readsize == fileSize)) | |
1217 | directive = ZSTD_e_end; | |
1218 | ||
1219 | stillToFlush = 1; | |
1220 | while ((inBuff.pos != inBuff.size) /* input buffer must be entirely ingested */ | |
1221 | || (directive == ZSTD_e_end && stillToFlush != 0) ) { | |
1222 | ||
1223 | size_t const oldIPos = inBuff.pos; | |
1224 | ZSTD_outBuffer outBuff = { ress.dstBuffer, ress.dstBufferSize, 0 }; | |
1225 | size_t const toFlushNow = ZSTD_toFlushNow(ress.cctx); | |
1226 | CHECK_V(stillToFlush, ZSTD_compressStream2(ress.cctx, &outBuff, &inBuff, directive)); | |
1227 | ||
1228 | /* count stats */ | |
1229 | inputPresented++; | |
1230 | if (oldIPos == inBuff.pos) inputBlocked++; /* input buffer is full and can't take any more : input speed is faster than consumption rate */ | |
1231 | if (!toFlushNow) flushWaiting = 1; | |
1232 | ||
1233 | /* Write compressed stream */ | |
1234 | DISPLAYLEVEL(6, "ZSTD_compress_generic(end:%u) => input pos(%u)<=(%u)size ; output generated %u bytes \n", | |
1235 | (unsigned)directive, (unsigned)inBuff.pos, (unsigned)inBuff.size, (unsigned)outBuff.pos); | |
1236 | if (outBuff.pos) { | |
1237 | size_t const sizeCheck = fwrite(ress.dstBuffer, 1, outBuff.pos, dstFile); | |
1238 | if (sizeCheck != outBuff.pos) | |
1239 | EXM_THROW(25, "Write error : %s (cannot write compressed block)", | |
1240 | strerror(errno)); | |
1241 | compressedfilesize += outBuff.pos; | |
1242 | } | |
1243 | ||
1244 | /* display notification; and adapt compression level */ | |
1245 | if (READY_FOR_UPDATE()) { | |
1246 | ZSTD_frameProgression const zfp = ZSTD_getFrameProgression(ress.cctx); | |
1247 | double const cShare = (double)zfp.produced / (zfp.consumed + !zfp.consumed/*avoid div0*/) * 100; | |
1248 | ||
1249 | /* display progress notifications */ | |
1250 | if (g_display_prefs.displayLevel >= 3) { | |
1251 | DISPLAYUPDATE(3, "\r(L%i) Buffered :%4u MB - Consumed :%4u MB - Compressed :%4u MB => %.2f%% ", | |
1252 | compressionLevel, | |
1253 | (unsigned)((zfp.ingested - zfp.consumed) >> 20), | |
1254 | (unsigned)(zfp.consumed >> 20), | |
1255 | (unsigned)(zfp.produced >> 20), | |
1256 | cShare ); | |
1257 | } else { /* summarized notifications if == 2; */ | |
1258 | DISPLAYLEVEL(2, "\rRead : %u ", (unsigned)(zfp.consumed >> 20)); | |
1259 | if (fileSize != UTIL_FILESIZE_UNKNOWN) | |
1260 | DISPLAYLEVEL(2, "/ %u ", (unsigned)(fileSize >> 20)); | |
1261 | DISPLAYLEVEL(2, "MB ==> %2.f%% ", cShare); | |
1262 | DELAY_NEXT_UPDATE(); | |
1263 | } | |
1264 | ||
1265 | /* adaptive mode : statistics measurement and speed correction */ | |
1266 | if (prefs->adaptiveMode) { | |
1267 | ||
1268 | /* check output speed */ | |
1269 | if (zfp.currentJobID > 1) { /* only possible if nbWorkers >= 1 */ | |
1270 | ||
1271 | unsigned long long newlyProduced = zfp.produced - previous_zfp_update.produced; | |
1272 | unsigned long long newlyFlushed = zfp.flushed - previous_zfp_update.flushed; | |
1273 | assert(zfp.produced >= previous_zfp_update.produced); | |
1274 | assert(prefs->nbWorkers >= 1); | |
1275 | ||
1276 | /* test if compression is blocked | |
1277 | * either because output is slow and all buffers are full | |
1278 | * or because input is slow and no job can start while waiting for at least one buffer to be filled. | |
1279 | * note : exclude starting part, since currentJobID > 1 */ | |
1280 | 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)*/ | |
1281 | && (zfp.nbActiveWorkers == 0) /* confirmed : no compression ongoing */ | |
1282 | ) { | |
1283 | DISPLAYLEVEL(6, "all buffers full : compression stopped => slow down \n") | |
1284 | speedChange = slower; | |
1285 | } | |
1286 | ||
1287 | previous_zfp_update = zfp; | |
1288 | ||
1289 | 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) */ | |
1290 | && (flushWaiting == 0) /* flush speed was never slowed by lack of production, so it's operating at max capacity */ | |
1291 | ) { | |
1292 | DISPLAYLEVEL(6, "compression faster than flush (%llu > %llu), and flushed was never slowed down by lack of production => slow down \n", newlyProduced, newlyFlushed); | |
1293 | speedChange = slower; | |
1294 | } | |
1295 | flushWaiting = 0; | |
1296 | } | |
1297 | ||
1298 | /* course correct only if there is at least one new job completed */ | |
1299 | if (zfp.currentJobID > lastJobID) { | |
1300 | DISPLAYLEVEL(6, "compression level adaptation check \n") | |
1301 | ||
1302 | /* check input speed */ | |
1303 | if (zfp.currentJobID > (unsigned)(prefs->nbWorkers+1)) { /* warm up period, to fill all workers */ | |
1304 | if (inputBlocked <= 0) { | |
1305 | DISPLAYLEVEL(6, "input is never blocked => input is slower than ingestion \n"); | |
1306 | speedChange = slower; | |
1307 | } else if (speedChange == noChange) { | |
1308 | unsigned long long newlyIngested = zfp.ingested - previous_zfp_correction.ingested; | |
1309 | unsigned long long newlyConsumed = zfp.consumed - previous_zfp_correction.consumed; | |
1310 | unsigned long long newlyProduced = zfp.produced - previous_zfp_correction.produced; | |
1311 | unsigned long long newlyFlushed = zfp.flushed - previous_zfp_correction.flushed; | |
1312 | previous_zfp_correction = zfp; | |
1313 | assert(inputPresented > 0); | |
1314 | DISPLAYLEVEL(6, "input blocked %u/%u(%.2f) - ingested:%u vs %u:consumed - flushed:%u vs %u:produced \n", | |
1315 | inputBlocked, inputPresented, (double)inputBlocked/inputPresented*100, | |
1316 | (unsigned)newlyIngested, (unsigned)newlyConsumed, | |
1317 | (unsigned)newlyFlushed, (unsigned)newlyProduced); | |
1318 | if ( (inputBlocked > inputPresented / 8) /* input is waiting often, because input buffers is full : compression or output too slow */ | |
1319 | && (newlyFlushed * 33 / 32 > newlyProduced) /* flush everything that is produced */ | |
1320 | && (newlyIngested * 33 / 32 > newlyConsumed) /* input speed as fast or faster than compression speed */ | |
1321 | ) { | |
1322 | DISPLAYLEVEL(6, "recommend faster as in(%llu) >= (%llu)comp(%llu) <= out(%llu) \n", | |
1323 | newlyIngested, newlyConsumed, newlyProduced, newlyFlushed); | |
1324 | speedChange = faster; | |
1325 | } | |
1326 | } | |
1327 | inputBlocked = 0; | |
1328 | inputPresented = 0; | |
1329 | } | |
1330 | ||
1331 | if (speedChange == slower) { | |
1332 | DISPLAYLEVEL(6, "slower speed , higher compression \n") | |
1333 | compressionLevel ++; | |
1334 | if (compressionLevel > ZSTD_maxCLevel()) compressionLevel = ZSTD_maxCLevel(); | |
1335 | if (compressionLevel > prefs->maxAdaptLevel) compressionLevel = prefs->maxAdaptLevel; | |
1336 | compressionLevel += (compressionLevel == 0); /* skip 0 */ | |
1337 | ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel); | |
1338 | } | |
1339 | if (speedChange == faster) { | |
1340 | DISPLAYLEVEL(6, "faster speed , lighter compression \n") | |
1341 | compressionLevel --; | |
1342 | if (compressionLevel < prefs->minAdaptLevel) compressionLevel = prefs->minAdaptLevel; | |
1343 | compressionLevel -= (compressionLevel == 0); /* skip 0 */ | |
1344 | ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel); | |
1345 | } | |
1346 | speedChange = noChange; | |
1347 | ||
1348 | lastJobID = zfp.currentJobID; | |
1349 | } /* if (zfp.currentJobID > lastJobID) */ | |
1350 | } /* if (g_adaptiveMode) */ | |
1351 | } /* if (READY_FOR_UPDATE()) */ | |
1352 | } /* while ((inBuff.pos != inBuff.size) */ | |
1353 | } while (directive != ZSTD_e_end); | |
1354 | ||
1355 | if (ferror(srcFile)) { | |
1356 | EXM_THROW(26, "Read error : I/O error"); | |
1357 | } | |
1358 | if (fileSize != UTIL_FILESIZE_UNKNOWN && *readsize != fileSize) { | |
1359 | EXM_THROW(27, "Read error : Incomplete read : %llu / %llu B", | |
1360 | (unsigned long long)*readsize, (unsigned long long)fileSize); | |
1361 | } | |
1362 | ||
1363 | return compressedfilesize; | |
1364 | } | |
1365 | ||
7c673cae FG |
1366 | /*! FIO_compressFilename_internal() : |
1367 | * same as FIO_compressFilename_extRess(), with `ress.desFile` already opened. | |
1368 | * @return : 0 : compression completed correctly, | |
1369 | * 1 : missing or pb opening srcFileName | |
1370 | */ | |
9f95a23c TL |
1371 | static int |
1372 | FIO_compressFilename_internal(FIO_prefs_t* const prefs, | |
1373 | cRess_t ress, | |
1374 | const char* dstFileName, const char* srcFileName, | |
1375 | int compressionLevel) | |
7c673cae | 1376 | { |
f67539c2 TL |
1377 | UTIL_time_t const timeStart = UTIL_getTime(); |
1378 | clock_t const cpuStart = clock(); | |
7c673cae FG |
1379 | U64 readsize = 0; |
1380 | U64 compressedfilesize = 0; | |
1381 | U64 const fileSize = UTIL_getFileSize(srcFileName); | |
9f95a23c | 1382 | DISPLAYLEVEL(5, "%s: %u bytes \n", srcFileName, (unsigned)fileSize); |
7c673cae | 1383 | |
9f95a23c TL |
1384 | /* compression format selection */ |
1385 | switch (prefs->compressionType) { | |
1386 | default: | |
11fdf7f2 | 1387 | case FIO_zstdCompression: |
9f95a23c | 1388 | compressedfilesize = FIO_compressZstdFrame(prefs, &ress, srcFileName, fileSize, compressionLevel, &readsize); |
11fdf7f2 TL |
1389 | break; |
1390 | ||
1391 | case FIO_gzipCompression: | |
1392 | #ifdef ZSTD_GZCOMPRESS | |
1393 | compressedfilesize = FIO_compressGzFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize); | |
1394 | #else | |
1395 | (void)compressionLevel; | |
1396 | EXM_THROW(20, "zstd: %s: file cannot be compressed as gzip (zstd compiled without ZSTD_GZCOMPRESS) -- ignored \n", | |
1397 | srcFileName); | |
1398 | #endif | |
9f95a23c | 1399 | break; |
11fdf7f2 TL |
1400 | |
1401 | case FIO_xzCompression: | |
1402 | case FIO_lzmaCompression: | |
1403 | #ifdef ZSTD_LZMACOMPRESS | |
9f95a23c | 1404 | compressedfilesize = FIO_compressLzmaFrame(&ress, srcFileName, fileSize, compressionLevel, &readsize, prefs->compressionType==FIO_lzmaCompression); |
11fdf7f2 TL |
1405 | #else |
1406 | (void)compressionLevel; | |
1407 | EXM_THROW(20, "zstd: %s: file cannot be compressed as xz/lzma (zstd compiled without ZSTD_LZMACOMPRESS) -- ignored \n", | |
1408 | srcFileName); | |
1409 | #endif | |
9f95a23c | 1410 | break; |
11fdf7f2 TL |
1411 | |
1412 | case FIO_lz4Compression: | |
1413 | #ifdef ZSTD_LZ4COMPRESS | |
9f95a23c | 1414 | compressedfilesize = FIO_compressLz4Frame(&ress, srcFileName, fileSize, compressionLevel, prefs->checksumFlag, &readsize); |
11fdf7f2 TL |
1415 | #else |
1416 | (void)compressionLevel; | |
1417 | EXM_THROW(20, "zstd: %s: file cannot be compressed as lz4 (zstd compiled without ZSTD_LZ4COMPRESS) -- ignored \n", | |
1418 | srcFileName); | |
1419 | #endif | |
9f95a23c | 1420 | break; |
7c673cae FG |
1421 | } |
1422 | ||
1423 | /* Status */ | |
1424 | DISPLAYLEVEL(2, "\r%79s\r", ""); | |
f67539c2 TL |
1425 | if (readsize == 0) { |
1426 | DISPLAYLEVEL(2,"%-20s : (%6llu => %6llu bytes, %s) \n", | |
1427 | srcFileName, | |
1428 | (unsigned long long)readsize, (unsigned long long) compressedfilesize, | |
1429 | dstFileName); | |
1430 | } else { | |
1431 | DISPLAYLEVEL(2,"%-20s :%6.2f%% (%6llu => %6llu bytes, %s) \n", | |
1432 | srcFileName, | |
1433 | (double)compressedfilesize / readsize * 100, | |
1434 | (unsigned long long)readsize, (unsigned long long) compressedfilesize, | |
1435 | dstFileName); | |
1436 | } | |
7c673cae | 1437 | |
f67539c2 TL |
1438 | /* Elapsed Time and CPU Load */ |
1439 | { clock_t const cpuEnd = clock(); | |
1440 | double const cpuLoad_s = (double)(cpuEnd - cpuStart) / CLOCKS_PER_SEC; | |
1441 | U64 const timeLength_ns = UTIL_clockSpanNano(timeStart); | |
1442 | double const timeLength_s = (double)timeLength_ns / 1000000000; | |
1443 | double const cpuLoad_pct = (cpuLoad_s / timeLength_s) * 100; | |
1444 | DISPLAYLEVEL(4, "%-20s : Completed in %.2f sec (cpu load : %.0f%%)\n", | |
1445 | srcFileName, timeLength_s, cpuLoad_pct); | |
1446 | } | |
7c673cae FG |
1447 | return 0; |
1448 | } | |
1449 | ||
1450 | ||
9f95a23c TL |
1451 | /*! FIO_compressFilename_dstFile() : |
1452 | * open dstFileName, or pass-through if ress.dstFile != NULL, | |
1453 | * then start compression with FIO_compressFilename_internal(). | |
1454 | * Manages source removal (--rm) and file permissions transfer. | |
1455 | * note : ress.srcFile must be != NULL, | |
1456 | * so reach this function through FIO_compressFilename_srcFile(). | |
7c673cae | 1457 | * @return : 0 : compression completed correctly, |
9f95a23c | 1458 | * 1 : pb |
7c673cae | 1459 | */ |
9f95a23c TL |
1460 | static int FIO_compressFilename_dstFile(FIO_prefs_t* const prefs, |
1461 | cRess_t ress, | |
1462 | const char* dstFileName, | |
1463 | const char* srcFileName, | |
1464 | int compressionLevel) | |
7c673cae | 1465 | { |
9f95a23c | 1466 | int closeDstFile = 0; |
7c673cae | 1467 | int result; |
9f95a23c TL |
1468 | stat_t statbuf; |
1469 | int transfer_permissions = 0; | |
9f95a23c | 1470 | assert(ress.srcFile != NULL); |
9f95a23c TL |
1471 | if (ress.dstFile == NULL) { |
1472 | closeDstFile = 1; | |
f67539c2 | 1473 | DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: opening dst: %s \n", dstFileName); |
9f95a23c TL |
1474 | ress.dstFile = FIO_openDstFile(prefs, srcFileName, dstFileName); |
1475 | if (ress.dstFile==NULL) return 1; /* could not open dstFileName */ | |
1476 | /* Must only be added after FIO_openDstFile() succeeds. | |
1477 | * Otherwise we may delete the destination file if it already exists, | |
1478 | * and the user presses Ctrl-C when asked if they wish to overwrite. | |
1479 | */ | |
1480 | addHandler(dstFileName); | |
1481 | ||
1482 | if ( strcmp (srcFileName, stdinmark) | |
1483 | && UTIL_getFileStat(srcFileName, &statbuf)) | |
1484 | transfer_permissions = 1; | |
7c673cae | 1485 | } |
11fdf7f2 | 1486 | |
9f95a23c | 1487 | result = FIO_compressFilename_internal(prefs, ress, dstFileName, srcFileName, compressionLevel); |
7c673cae | 1488 | |
9f95a23c TL |
1489 | if (closeDstFile) { |
1490 | FILE* const dstFile = ress.dstFile; | |
1491 | ress.dstFile = NULL; | |
7c673cae | 1492 | |
9f95a23c TL |
1493 | clearHandler(); |
1494 | ||
f67539c2 | 1495 | DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: closing dst: %s \n", dstFileName); |
9f95a23c TL |
1496 | if (fclose(dstFile)) { /* error closing dstFile */ |
1497 | DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno)); | |
1498 | result=1; | |
1499 | } | |
1500 | if ( (result != 0) /* operation failure */ | |
1501 | && strcmp(dstFileName, nulmark) /* special case : don't remove() /dev/null */ | |
1502 | && strcmp(dstFileName, stdoutmark) /* special case : don't remove() stdout */ | |
1503 | ) { | |
1504 | FIO_remove(dstFileName); /* remove compression artefact; note don't do anything special if remove() fails */ | |
1505 | } else if ( strcmp(dstFileName, stdoutmark) | |
1506 | && strcmp(dstFileName, nulmark) | |
1507 | && transfer_permissions) { | |
f67539c2 | 1508 | DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: transfering permissions into dst: %s \n", dstFileName); |
9f95a23c | 1509 | UTIL_setFileStat(dstFileName, &statbuf); |
f67539c2 TL |
1510 | } else { |
1511 | DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: do not transfer permissions into dst: %s \n", dstFileName); | |
9f95a23c | 1512 | } |
11fdf7f2 | 1513 | } |
9f95a23c | 1514 | |
7c673cae FG |
1515 | return result; |
1516 | } | |
1517 | ||
f67539c2 TL |
1518 | /* List used to compare file extensions (used with --exclude-compressed flag) |
1519 | * Different from the suffixList and should only apply to ZSTD compress operationResult | |
1520 | */ | |
1521 | static const char *compressedFileExtensions[] = { | |
1522 | ZSTD_EXTENSION, | |
1523 | TZSTD_EXTENSION, | |
1524 | GZ_EXTENSION, | |
1525 | TGZ_EXTENSION, | |
1526 | LZMA_EXTENSION, | |
1527 | XZ_EXTENSION, | |
1528 | TXZ_EXTENSION, | |
1529 | LZ4_EXTENSION, | |
1530 | TLZ4_EXTENSION, | |
1531 | NULL | |
1532 | }; | |
7c673cae | 1533 | |
9f95a23c | 1534 | /*! FIO_compressFilename_srcFile() : |
7c673cae | 1535 | * @return : 0 : compression completed correctly, |
9f95a23c | 1536 | * 1 : missing or pb opening srcFileName |
7c673cae | 1537 | */ |
9f95a23c TL |
1538 | static int |
1539 | FIO_compressFilename_srcFile(FIO_prefs_t* const prefs, | |
1540 | cRess_t ress, | |
1541 | const char* dstFileName, | |
1542 | const char* srcFileName, | |
1543 | int compressionLevel) | |
7c673cae FG |
1544 | { |
1545 | int result; | |
f67539c2 | 1546 | DISPLAYLEVEL(6, "FIO_compressFilename_srcFile: %s \n", srcFileName); |
7c673cae | 1547 | |
9f95a23c TL |
1548 | /* ensure src is not a directory */ |
1549 | if (UTIL_isDirectory(srcFileName)) { | |
1550 | DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName); | |
1551 | return 1; | |
1552 | } | |
7c673cae | 1553 | |
9f95a23c TL |
1554 | /* ensure src is not the same as dict (if present) */ |
1555 | if (ress.dictFileName != NULL && UTIL_isSameFile(srcFileName, ress.dictFileName)) { | |
1556 | DISPLAYLEVEL(1, "zstd: cannot use %s as an input file and dictionary \n", srcFileName); | |
1557 | return 1; | |
11fdf7f2 TL |
1558 | } |
1559 | ||
f67539c2 TL |
1560 | /* Check if "srcFile" is compressed. Only done if --exclude-compressed flag is used |
1561 | * YES => ZSTD will skip compression of the file and will return 0. | |
1562 | * NO => ZSTD will resume with compress operation. | |
1563 | */ | |
1564 | if (prefs->excludeCompressedFiles == 1 && UTIL_isCompressedFile(srcFileName, compressedFileExtensions)) { | |
1565 | DISPLAYLEVEL(4, "File is already compressed : %s \n", srcFileName); | |
1566 | return 0; | |
1567 | } | |
1568 | ||
9f95a23c TL |
1569 | ress.srcFile = FIO_openSrcFile(srcFileName); |
1570 | if (ress.srcFile == NULL) return 1; /* srcFile could not be opened */ | |
11fdf7f2 | 1571 | |
9f95a23c | 1572 | result = FIO_compressFilename_dstFile(prefs, ress, dstFileName, srcFileName, compressionLevel); |
11fdf7f2 | 1573 | |
9f95a23c TL |
1574 | fclose(ress.srcFile); |
1575 | ress.srcFile = NULL; | |
1576 | if ( prefs->removeSrcFile /* --rm */ | |
1577 | && result == 0 /* success */ | |
1578 | && strcmp(srcFileName, stdinmark) /* exception : don't erase stdin */ | |
1579 | ) { | |
1580 | /* We must clear the handler, since after this point calling it would | |
1581 | * delete both the source and destination files. | |
1582 | */ | |
1583 | clearHandler(); | |
1584 | if (FIO_remove(srcFileName)) | |
1585 | EXM_THROW(1, "zstd: %s: %s", srcFileName, strerror(errno)); | |
11fdf7f2 | 1586 | } |
7c673cae FG |
1587 | return result; |
1588 | } | |
1589 | ||
f67539c2 TL |
1590 | int FIO_compressFilename(FIO_prefs_t* const prefs, const char* dstFileName, |
1591 | const char* srcFileName, const char* dictFileName, | |
1592 | int compressionLevel, ZSTD_compressionParameters comprParams) | |
7c673cae | 1593 | { |
f67539c2 | 1594 | cRess_t const ress = FIO_createCResources(prefs, dictFileName, UTIL_getFileSize(srcFileName), compressionLevel, comprParams); |
9f95a23c | 1595 | int const result = FIO_compressFilename_srcFile(prefs, ress, dstFileName, srcFileName, compressionLevel); |
7c673cae | 1596 | |
7c673cae FG |
1597 | |
1598 | FIO_freeCResources(ress); | |
1599 | return result; | |
1600 | } | |
1601 | ||
9f95a23c TL |
1602 | /* FIO_determineCompressedName() : |
1603 | * create a destination filename for compressed srcFileName. | |
1604 | * @return a pointer to it. | |
1605 | * This function never returns an error (it may abort() in case of pb) | |
1606 | */ | |
1607 | static const char* | |
f67539c2 | 1608 | FIO_determineCompressedName(const char* srcFileName, const char* outDirName, const char* suffix) |
9f95a23c TL |
1609 | { |
1610 | static size_t dfnbCapacity = 0; | |
1611 | static char* dstFileNameBuffer = NULL; /* using static allocation : this function cannot be multi-threaded */ | |
f67539c2 TL |
1612 | char* outDirFilename = NULL; |
1613 | size_t sfnSize = strlen(srcFileName); | |
1614 | size_t const srcSuffixLen = strlen(suffix); | |
1615 | if (outDirName) { | |
1616 | outDirFilename = FIO_createFilename_fromOutDir(srcFileName, outDirName, srcSuffixLen); | |
1617 | sfnSize = strlen(outDirFilename); | |
1618 | assert(outDirFilename != NULL); | |
1619 | } | |
9f95a23c | 1620 | |
f67539c2 | 1621 | if (dfnbCapacity <= sfnSize+srcSuffixLen+1) { |
9f95a23c TL |
1622 | /* resize buffer for dstName */ |
1623 | free(dstFileNameBuffer); | |
f67539c2 | 1624 | dfnbCapacity = sfnSize + srcSuffixLen + 30; |
9f95a23c TL |
1625 | dstFileNameBuffer = (char*)malloc(dfnbCapacity); |
1626 | if (!dstFileNameBuffer) { | |
1627 | EXM_THROW(30, "zstd: %s", strerror(errno)); | |
f67539c2 TL |
1628 | } |
1629 | } | |
9f95a23c | 1630 | assert(dstFileNameBuffer != NULL); |
9f95a23c | 1631 | |
f67539c2 TL |
1632 | if (outDirFilename) { |
1633 | memcpy(dstFileNameBuffer, outDirFilename, sfnSize); | |
1634 | free(outDirFilename); | |
1635 | } else { | |
1636 | memcpy(dstFileNameBuffer, srcFileName, sfnSize); | |
1637 | } | |
1638 | memcpy(dstFileNameBuffer+sfnSize, suffix, srcSuffixLen+1 /* Include terminating null */); | |
9f95a23c TL |
1639 | return dstFileNameBuffer; |
1640 | } | |
1641 | ||
f67539c2 TL |
1642 | static unsigned long long FIO_getLargestFileSize(const char** inFileNames, unsigned nbFiles) |
1643 | { | |
1644 | size_t i; | |
1645 | unsigned long long fileSize, maxFileSize = 0; | |
1646 | for (i = 0; i < nbFiles; i++) { | |
1647 | fileSize = UTIL_getFileSize(inFileNames[i]); | |
1648 | maxFileSize = fileSize > maxFileSize ? fileSize : maxFileSize; | |
1649 | } | |
1650 | return maxFileSize; | |
1651 | } | |
9f95a23c TL |
1652 | |
1653 | /* FIO_compressMultipleFilenames() : | |
1654 | * compress nbFiles files | |
f67539c2 TL |
1655 | * into either one destination (outFileName), |
1656 | * or into one file each (outFileName == NULL, but suffix != NULL), | |
1657 | * or into a destination folder (specified with -O) | |
9f95a23c TL |
1658 | */ |
1659 | int FIO_compressMultipleFilenames(FIO_prefs_t* const prefs, | |
1660 | const char** inFileNamesTable, unsigned nbFiles, | |
f67539c2 | 1661 | const char* outDirName, |
9f95a23c | 1662 | const char* outFileName, const char* suffix, |
11fdf7f2 | 1663 | const char* dictFileName, int compressionLevel, |
9f95a23c | 1664 | ZSTD_compressionParameters comprParams) |
7c673cae | 1665 | { |
9f95a23c | 1666 | int error = 0; |
f67539c2 TL |
1667 | cRess_t ress = FIO_createCResources(prefs, dictFileName, |
1668 | FIO_getLargestFileSize(inFileNamesTable, nbFiles), | |
1669 | compressionLevel, comprParams); | |
7c673cae FG |
1670 | |
1671 | /* init */ | |
9f95a23c | 1672 | assert(outFileName != NULL || suffix != NULL); |
9f95a23c TL |
1673 | if (outFileName != NULL) { /* output into a single destination (stdout typically) */ |
1674 | ress.dstFile = FIO_openDstFile(prefs, NULL, outFileName); | |
1675 | if (ress.dstFile == NULL) { /* could not open outFileName */ | |
1676 | error = 1; | |
1677 | } else { | |
1678 | unsigned u; | |
1679 | for (u=0; u<nbFiles; u++) | |
1680 | error |= FIO_compressFilename_srcFile(prefs, ress, outFileName, inFileNamesTable[u], compressionLevel); | |
1681 | if (fclose(ress.dstFile)) | |
1682 | EXM_THROW(29, "Write error (%s) : cannot properly close %s", | |
1683 | strerror(errno), outFileName); | |
1684 | ress.dstFile = NULL; | |
1685 | } | |
7c673cae FG |
1686 | } else { |
1687 | unsigned u; | |
1688 | for (u=0; u<nbFiles; u++) { | |
9f95a23c | 1689 | const char* const srcFileName = inFileNamesTable[u]; |
f67539c2 | 1690 | const char* const dstFileName = FIO_determineCompressedName(srcFileName, outDirName, suffix); /* cannot fail */ |
9f95a23c | 1691 | error |= FIO_compressFilename_srcFile(prefs, ress, dstFileName, srcFileName, compressionLevel); |
f67539c2 TL |
1692 | } |
1693 | if (outDirName) | |
1694 | FIO_checkFilenameCollisions(inFileNamesTable ,nbFiles); | |
1695 | } | |
7c673cae | 1696 | |
7c673cae | 1697 | FIO_freeCResources(ress); |
9f95a23c | 1698 | return error; |
7c673cae FG |
1699 | } |
1700 | ||
1701 | #endif /* #ifndef ZSTD_NOCOMPRESS */ | |
1702 | ||
1703 | ||
1704 | ||
1705 | #ifndef ZSTD_NODECOMPRESS | |
1706 | ||
1707 | /* ************************************************************************** | |
11fdf7f2 TL |
1708 | * Decompression |
1709 | ***************************************************************************/ | |
7c673cae FG |
1710 | typedef struct { |
1711 | void* srcBuffer; | |
7c673cae | 1712 | size_t srcBufferSize; |
11fdf7f2 | 1713 | size_t srcBufferLoaded; |
7c673cae FG |
1714 | void* dstBuffer; |
1715 | size_t dstBufferSize; | |
1716 | ZSTD_DStream* dctx; | |
1717 | FILE* dstFile; | |
1718 | } dRess_t; | |
1719 | ||
9f95a23c | 1720 | static dRess_t FIO_createDResources(FIO_prefs_t* const prefs, const char* dictFileName) |
7c673cae FG |
1721 | { |
1722 | dRess_t ress; | |
1723 | memset(&ress, 0, sizeof(ress)); | |
1724 | ||
f67539c2 TL |
1725 | if (prefs->patchFromMode) |
1726 | FIO_adjustMemLimitForPatchFromMode(prefs, UTIL_getFileSize(dictFileName), 0 /* just use the dict size */); | |
1727 | ||
7c673cae FG |
1728 | /* Allocation */ |
1729 | ress.dctx = ZSTD_createDStream(); | |
9f95a23c TL |
1730 | if (ress.dctx==NULL) |
1731 | EXM_THROW(60, "Error: %s : can't create ZSTD_DStream", strerror(errno)); | |
1732 | CHECK( ZSTD_DCtx_setMaxWindowSize(ress.dctx, prefs->memLimit) ); | |
7c673cae FG |
1733 | ress.srcBufferSize = ZSTD_DStreamInSize(); |
1734 | ress.srcBuffer = malloc(ress.srcBufferSize); | |
1735 | ress.dstBufferSize = ZSTD_DStreamOutSize(); | |
1736 | ress.dstBuffer = malloc(ress.dstBufferSize); | |
11fdf7f2 TL |
1737 | if (!ress.srcBuffer || !ress.dstBuffer) |
1738 | EXM_THROW(61, "Allocation error : not enough memory"); | |
7c673cae FG |
1739 | |
1740 | /* dictionary */ | |
1741 | { void* dictBuffer; | |
f67539c2 | 1742 | size_t const dictBufferSize = FIO_createDictBuffer(&dictBuffer, dictFileName, prefs); |
11fdf7f2 | 1743 | CHECK( ZSTD_initDStream_usingDict(ress.dctx, dictBuffer, dictBufferSize) ); |
7c673cae FG |
1744 | free(dictBuffer); |
1745 | } | |
1746 | ||
1747 | return ress; | |
1748 | } | |
1749 | ||
1750 | static void FIO_freeDResources(dRess_t ress) | |
1751 | { | |
11fdf7f2 | 1752 | CHECK( ZSTD_freeDStream(ress.dctx) ); |
7c673cae FG |
1753 | free(ress.srcBuffer); |
1754 | free(ress.dstBuffer); | |
1755 | } | |
1756 | ||
1757 | ||
1758 | /** FIO_fwriteSparse() : | |
f67539c2 TL |
1759 | * @return : storedSkips, |
1760 | * argument for next call to FIO_fwriteSparse() or FIO_fwriteSparseEnd() */ | |
1761 | static unsigned | |
1762 | FIO_fwriteSparse(FILE* file, | |
1763 | const void* buffer, size_t bufferSize, | |
1764 | const FIO_prefs_t* const prefs, | |
1765 | unsigned storedSkips) | |
7c673cae FG |
1766 | { |
1767 | const size_t* const bufferT = (const size_t*)buffer; /* Buffer is supposed malloc'ed, hence aligned on size_t */ | |
1768 | size_t bufferSizeT = bufferSize / sizeof(size_t); | |
1769 | const size_t* const bufferTEnd = bufferT + bufferSizeT; | |
1770 | const size_t* ptrT = bufferT; | |
f67539c2 TL |
1771 | static const size_t segmentSizeT = (32 KB) / sizeof(size_t); /* check every 32 KB */ |
1772 | ||
1773 | if (prefs->testMode) return 0; /* do not output anything in test mode */ | |
7c673cae | 1774 | |
9f95a23c | 1775 | if (!prefs->sparseFileSupport) { /* normal write */ |
7c673cae | 1776 | size_t const sizeCheck = fwrite(buffer, 1, bufferSize, file); |
9f95a23c | 1777 | if (sizeCheck != bufferSize) |
f67539c2 | 1778 | EXM_THROW(70, "Write error : cannot write decoded block : %s", |
9f95a23c | 1779 | strerror(errno)); |
7c673cae FG |
1780 | return 0; |
1781 | } | |
1782 | ||
1783 | /* avoid int overflow */ | |
1784 | if (storedSkips > 1 GB) { | |
f67539c2 TL |
1785 | if (LONG_SEEK(file, 1 GB, SEEK_CUR) != 0) |
1786 | EXM_THROW(91, "1 GB skip error (sparse file support)"); | |
7c673cae FG |
1787 | storedSkips -= 1 GB; |
1788 | } | |
1789 | ||
1790 | while (ptrT < bufferTEnd) { | |
7c673cae FG |
1791 | size_t nb0T; |
1792 | ||
f67539c2 TL |
1793 | /* adjust last segment if < 32 KB */ |
1794 | size_t seg0SizeT = segmentSizeT; | |
7c673cae FG |
1795 | if (seg0SizeT > bufferSizeT) seg0SizeT = bufferSizeT; |
1796 | bufferSizeT -= seg0SizeT; | |
f67539c2 TL |
1797 | |
1798 | /* count leading zeroes */ | |
7c673cae FG |
1799 | for (nb0T=0; (nb0T < seg0SizeT) && (ptrT[nb0T] == 0); nb0T++) ; |
1800 | storedSkips += (unsigned)(nb0T * sizeof(size_t)); | |
1801 | ||
1802 | if (nb0T != seg0SizeT) { /* not all 0s */ | |
f67539c2 TL |
1803 | size_t const nbNon0ST = seg0SizeT - nb0T; |
1804 | /* skip leading zeros */ | |
1805 | if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0) | |
1806 | EXM_THROW(92, "Sparse skip error ; try --no-sparse"); | |
7c673cae | 1807 | storedSkips = 0; |
f67539c2 TL |
1808 | /* write the rest */ |
1809 | if (fwrite(ptrT + nb0T, sizeof(size_t), nbNon0ST, file) != nbNon0ST) | |
1810 | EXM_THROW(93, "Write error : cannot write decoded block : %s", | |
1811 | strerror(errno)); | |
1812 | } | |
7c673cae FG |
1813 | ptrT += seg0SizeT; |
1814 | } | |
1815 | ||
1816 | { static size_t const maskT = sizeof(size_t)-1; | |
11fdf7f2 TL |
1817 | if (bufferSize & maskT) { |
1818 | /* size not multiple of sizeof(size_t) : implies end of block */ | |
7c673cae FG |
1819 | const char* const restStart = (const char*)bufferTEnd; |
1820 | const char* restPtr = restStart; | |
f67539c2 TL |
1821 | const char* const restEnd = (const char*)buffer + bufferSize; |
1822 | assert(restEnd > restStart && restEnd < restStart + sizeof(size_t)); | |
7c673cae FG |
1823 | for ( ; (restPtr < restEnd) && (*restPtr == 0); restPtr++) ; |
1824 | storedSkips += (unsigned) (restPtr - restStart); | |
1825 | if (restPtr != restEnd) { | |
f67539c2 TL |
1826 | /* not all remaining bytes are 0 */ |
1827 | size_t const restSize = (size_t)(restEnd - restPtr); | |
1828 | if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0) | |
1829 | EXM_THROW(92, "Sparse skip error ; try --no-sparse"); | |
1830 | if (fwrite(restPtr, 1, restSize, file) != restSize) | |
1831 | EXM_THROW(95, "Write error : cannot write end of decoded block : %s", | |
1832 | strerror(errno)); | |
7c673cae | 1833 | storedSkips = 0; |
f67539c2 | 1834 | } } } |
7c673cae FG |
1835 | |
1836 | return storedSkips; | |
1837 | } | |
1838 | ||
9f95a23c | 1839 | static void |
f67539c2 | 1840 | FIO_fwriteSparseEnd(const FIO_prefs_t* const prefs, FILE* file, unsigned storedSkips) |
7c673cae | 1841 | { |
f67539c2 | 1842 | if (prefs->testMode) assert(storedSkips == 0); |
9f95a23c TL |
1843 | if (storedSkips>0) { |
1844 | assert(prefs->sparseFileSupport > 0); /* storedSkips>0 implies sparse support is enabled */ | |
1845 | (void)prefs; /* assert can be disabled, in which case prefs becomes unused */ | |
1846 | if (LONG_SEEK(file, storedSkips-1, SEEK_CUR) != 0) | |
1847 | EXM_THROW(69, "Final skip error (sparse file support)"); | |
1848 | /* last zero must be explicitly written, | |
1849 | * so that skipped ones get implicitly translated as zero by FS */ | |
7c673cae | 1850 | { const char lastZeroByte[1] = { 0 }; |
9f95a23c | 1851 | if (fwrite(lastZeroByte, 1, 1, file) != 1) |
f67539c2 | 1852 | EXM_THROW(69, "Write error : cannot write last zero : %s", strerror(errno)); |
7c673cae FG |
1853 | } } |
1854 | } | |
1855 | ||
1856 | ||
1857 | /** FIO_passThrough() : just copy input into output, for compatibility with gzip -df mode | |
1858 | @return : 0 (no error) */ | |
f67539c2 | 1859 | static int FIO_passThrough(const FIO_prefs_t* const prefs, |
9f95a23c TL |
1860 | FILE* foutput, FILE* finput, |
1861 | void* buffer, size_t bufferSize, | |
1862 | size_t alreadyLoaded) | |
7c673cae FG |
1863 | { |
1864 | size_t const blockSize = MIN(64 KB, bufferSize); | |
f67539c2 | 1865 | size_t readFromInput; |
7c673cae FG |
1866 | unsigned storedSkips = 0; |
1867 | ||
1868 | /* assumption : ress->srcBufferLoaded bytes already loaded and stored within buffer */ | |
11fdf7f2 TL |
1869 | { size_t const sizeCheck = fwrite(buffer, 1, alreadyLoaded, foutput); |
1870 | if (sizeCheck != alreadyLoaded) { | |
f67539c2 | 1871 | DISPLAYLEVEL(1, "Pass-through write error : %s\n", strerror(errno)); |
11fdf7f2 TL |
1872 | return 1; |
1873 | } } | |
7c673cae | 1874 | |
f67539c2 | 1875 | do { |
7c673cae | 1876 | readFromInput = fread(buffer, 1, blockSize, finput); |
f67539c2 TL |
1877 | storedSkips = FIO_fwriteSparse(foutput, buffer, readFromInput, prefs, storedSkips); |
1878 | } while (readFromInput == blockSize); | |
1879 | if (ferror(finput)) { | |
1880 | DISPLAYLEVEL(1, "Pass-through read error : %s\n", strerror(errno)); | |
1881 | return 1; | |
7c673cae | 1882 | } |
f67539c2 | 1883 | assert(feof(finput)); |
7c673cae | 1884 | |
9f95a23c | 1885 | FIO_fwriteSparseEnd(prefs, foutput, storedSkips); |
7c673cae FG |
1886 | return 0; |
1887 | } | |
1888 | ||
9f95a23c TL |
1889 | /* FIO_zstdErrorHelp() : |
1890 | * detailed error message when requested window size is too large */ | |
f67539c2 TL |
1891 | static void |
1892 | FIO_zstdErrorHelp(const FIO_prefs_t* const prefs, | |
1893 | const dRess_t* ress, | |
1894 | size_t err, const char* srcFileName) | |
11fdf7f2 TL |
1895 | { |
1896 | ZSTD_frameHeader header; | |
9f95a23c TL |
1897 | |
1898 | /* Help message only for one specific error */ | |
1899 | if (ZSTD_getErrorCode(err) != ZSTD_error_frameParameter_windowTooLarge) | |
11fdf7f2 | 1900 | return; |
9f95a23c | 1901 | |
11fdf7f2 | 1902 | /* Try to decode the frame header */ |
9f95a23c TL |
1903 | err = ZSTD_getFrameHeader(&header, ress->srcBuffer, ress->srcBufferLoaded); |
1904 | if (err == 0) { | |
1905 | unsigned long long const windowSize = header.windowSize; | |
1906 | unsigned const windowLog = FIO_highbit64(windowSize) + ((windowSize & (windowSize - 1)) != 0); | |
1907 | assert(prefs->memLimit > 0); | |
f67539c2 | 1908 | DISPLAYLEVEL(1, "%s : Window size larger than maximum : %llu > %u \n", |
9f95a23c | 1909 | srcFileName, windowSize, prefs->memLimit); |
11fdf7f2 | 1910 | if (windowLog <= ZSTD_WINDOWLOG_MAX) { |
9f95a23c TL |
1911 | unsigned const windowMB = (unsigned)((windowSize >> 20) + ((windowSize & ((1 MB) - 1)) != 0)); |
1912 | assert(windowSize < (U64)(1ULL << 52)); /* ensure now overflow for windowMB */ | |
f67539c2 | 1913 | DISPLAYLEVEL(1, "%s : Use --long=%u or --memory=%uMB \n", |
11fdf7f2 TL |
1914 | srcFileName, windowLog, windowMB); |
1915 | return; | |
f67539c2 TL |
1916 | } } |
1917 | DISPLAYLEVEL(1, "%s : Window log larger than ZSTD_WINDOWLOG_MAX=%u; not supported \n", | |
11fdf7f2 TL |
1918 | srcFileName, ZSTD_WINDOWLOG_MAX); |
1919 | } | |
7c673cae FG |
1920 | |
1921 | /** FIO_decompressFrame() : | |
11fdf7f2 | 1922 | * @return : size of decoded zstd frame, or an error code |
f67539c2 | 1923 | */ |
11fdf7f2 | 1924 | #define FIO_ERROR_FRAME_DECODING ((unsigned long long)(-2)) |
f67539c2 TL |
1925 | static unsigned long long |
1926 | FIO_decompressZstdFrame(dRess_t* ress, FILE* finput, | |
1927 | const FIO_prefs_t* const prefs, | |
1928 | const char* srcFileName, | |
1929 | U64 alreadyDecoded) /* for multi-frames streams */ | |
7c673cae FG |
1930 | { |
1931 | U64 frameSize = 0; | |
1932 | U32 storedSkips = 0; | |
1933 | ||
f67539c2 TL |
1934 | /* display last 20 characters only */ |
1935 | { size_t const srcFileLength = strlen(srcFileName); | |
1936 | if (srcFileLength>20) srcFileName += srcFileLength-20; | |
1937 | } | |
11fdf7f2 | 1938 | |
7c673cae FG |
1939 | ZSTD_resetDStream(ress->dctx); |
1940 | ||
11fdf7f2 TL |
1941 | /* Header loading : ensures ZSTD_getFrameHeader() will succeed */ |
1942 | { size_t const toDecode = ZSTD_FRAMEHEADERSIZE_MAX; | |
1943 | if (ress->srcBufferLoaded < toDecode) { | |
1944 | size_t const toRead = toDecode - ress->srcBufferLoaded; | |
1945 | void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded; | |
1946 | ress->srcBufferLoaded += fread(startPosition, 1, toRead, finput); | |
1947 | } } | |
7c673cae FG |
1948 | |
1949 | /* Main decompression Loop */ | |
1950 | while (1) { | |
1951 | ZSTD_inBuffer inBuff = { ress->srcBuffer, ress->srcBufferLoaded, 0 }; | |
1952 | ZSTD_outBuffer outBuff= { ress->dstBuffer, ress->dstBufferSize, 0 }; | |
1953 | size_t const readSizeHint = ZSTD_decompressStream(ress->dctx, &outBuff, &inBuff); | |
11fdf7f2 TL |
1954 | if (ZSTD_isError(readSizeHint)) { |
1955 | DISPLAYLEVEL(1, "%s : Decoding error (36) : %s \n", | |
1956 | srcFileName, ZSTD_getErrorName(readSizeHint)); | |
9f95a23c | 1957 | FIO_zstdErrorHelp(prefs, ress, readSizeHint, srcFileName); |
11fdf7f2 TL |
1958 | return FIO_ERROR_FRAME_DECODING; |
1959 | } | |
7c673cae FG |
1960 | |
1961 | /* Write block */ | |
f67539c2 | 1962 | storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, outBuff.pos, prefs, storedSkips); |
7c673cae | 1963 | frameSize += outBuff.pos; |
11fdf7f2 | 1964 | DISPLAYUPDATE(2, "\r%-20.20s : %u MB... ", |
9f95a23c | 1965 | srcFileName, (unsigned)((alreadyDecoded+frameSize)>>20) ); |
7c673cae FG |
1966 | |
1967 | if (inBuff.pos > 0) { | |
1968 | memmove(ress->srcBuffer, (char*)ress->srcBuffer + inBuff.pos, inBuff.size - inBuff.pos); | |
1969 | ress->srcBufferLoaded -= inBuff.pos; | |
1970 | } | |
1971 | ||
1972 | if (readSizeHint == 0) break; /* end of frame */ | |
7c673cae FG |
1973 | |
1974 | /* Fill input buffer */ | |
11fdf7f2 TL |
1975 | { size_t const toDecode = MIN(readSizeHint, ress->srcBufferSize); /* support large skippable frames */ |
1976 | if (ress->srcBufferLoaded < toDecode) { | |
1977 | size_t const toRead = toDecode - ress->srcBufferLoaded; /* > 0 */ | |
1978 | void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded; | |
1979 | size_t const readSize = fread(startPosition, 1, toRead, finput); | |
1980 | if (readSize==0) { | |
1981 | DISPLAYLEVEL(1, "%s : Read error (39) : premature end \n", | |
1982 | srcFileName); | |
1983 | return FIO_ERROR_FRAME_DECODING; | |
1984 | } | |
1985 | ress->srcBufferLoaded += readSize; | |
1986 | } } } | |
7c673cae | 1987 | |
9f95a23c | 1988 | FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips); |
7c673cae FG |
1989 | |
1990 | return frameSize; | |
1991 | } | |
1992 | ||
1993 | ||
1994 | #ifdef ZSTD_GZDECOMPRESS | |
f67539c2 TL |
1995 | static unsigned long long |
1996 | FIO_decompressGzFrame(dRess_t* ress, FILE* srcFile, | |
1997 | const FIO_prefs_t* const prefs, | |
1998 | const char* srcFileName) | |
7c673cae FG |
1999 | { |
2000 | unsigned long long outFileSize = 0; | |
2001 | z_stream strm; | |
11fdf7f2 TL |
2002 | int flush = Z_NO_FLUSH; |
2003 | int decodingError = 0; | |
f67539c2 | 2004 | unsigned storedSkips = 0; |
7c673cae FG |
2005 | |
2006 | strm.zalloc = Z_NULL; | |
2007 | strm.zfree = Z_NULL; | |
2008 | strm.opaque = Z_NULL; | |
2009 | strm.next_in = 0; | |
11fdf7f2 TL |
2010 | strm.avail_in = 0; |
2011 | /* see http://www.zlib.net/manual.html */ | |
2012 | if (inflateInit2(&strm, 15 /* maxWindowLogSize */ + 16 /* gzip only */) != Z_OK) | |
2013 | return FIO_ERROR_FRAME_DECODING; | |
2014 | ||
2015 | strm.next_out = (Bytef*)ress->dstBuffer; | |
2016 | strm.avail_out = (uInt)ress->dstBufferSize; | |
2017 | strm.avail_in = (uInt)ress->srcBufferLoaded; | |
7c673cae FG |
2018 | strm.next_in = (z_const unsigned char*)ress->srcBuffer; |
2019 | ||
2020 | for ( ; ; ) { | |
2021 | int ret; | |
2022 | if (strm.avail_in == 0) { | |
2023 | ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile); | |
11fdf7f2 | 2024 | if (ress->srcBufferLoaded == 0) flush = Z_FINISH; |
7c673cae | 2025 | strm.next_in = (z_const unsigned char*)ress->srcBuffer; |
11fdf7f2 TL |
2026 | strm.avail_in = (uInt)ress->srcBufferLoaded; |
2027 | } | |
2028 | ret = inflate(&strm, flush); | |
2029 | if (ret == Z_BUF_ERROR) { | |
2030 | DISPLAYLEVEL(1, "zstd: %s: premature gz end \n", srcFileName); | |
2031 | decodingError = 1; break; | |
2032 | } | |
2033 | if (ret != Z_OK && ret != Z_STREAM_END) { | |
2034 | DISPLAYLEVEL(1, "zstd: %s: inflate error %d \n", srcFileName, ret); | |
2035 | decodingError = 1; break; | |
7c673cae | 2036 | } |
7c673cae FG |
2037 | { size_t const decompBytes = ress->dstBufferSize - strm.avail_out; |
2038 | if (decompBytes) { | |
f67539c2 | 2039 | storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decompBytes, prefs, storedSkips); |
7c673cae | 2040 | outFileSize += decompBytes; |
11fdf7f2 TL |
2041 | strm.next_out = (Bytef*)ress->dstBuffer; |
2042 | strm.avail_out = (uInt)ress->dstBufferSize; | |
7c673cae FG |
2043 | } |
2044 | } | |
2045 | if (ret == Z_STREAM_END) break; | |
2046 | } | |
2047 | ||
11fdf7f2 TL |
2048 | if (strm.avail_in > 0) |
2049 | memmove(ress->srcBuffer, strm.next_in, strm.avail_in); | |
7c673cae | 2050 | ress->srcBufferLoaded = strm.avail_in; |
11fdf7f2 TL |
2051 | if ( (inflateEnd(&strm) != Z_OK) /* release resources ; error detected */ |
2052 | && (decodingError==0) ) { | |
2053 | DISPLAYLEVEL(1, "zstd: %s: inflateEnd error \n", srcFileName); | |
2054 | decodingError = 1; | |
2055 | } | |
f67539c2 | 2056 | FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips); |
11fdf7f2 | 2057 | return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize; |
7c673cae FG |
2058 | } |
2059 | #endif | |
2060 | ||
2061 | ||
11fdf7f2 | 2062 | #ifdef ZSTD_LZMADECOMPRESS |
f67539c2 TL |
2063 | static unsigned long long |
2064 | FIO_decompressLzmaFrame(dRess_t* ress, FILE* srcFile, | |
2065 | const FIO_prefs_t* const prefs, | |
2066 | const char* srcFileName, int plain_lzma) | |
11fdf7f2 TL |
2067 | { |
2068 | unsigned long long outFileSize = 0; | |
2069 | lzma_stream strm = LZMA_STREAM_INIT; | |
2070 | lzma_action action = LZMA_RUN; | |
2071 | lzma_ret initRet; | |
2072 | int decodingError = 0; | |
f67539c2 | 2073 | unsigned storedSkips = 0; |
11fdf7f2 TL |
2074 | |
2075 | strm.next_in = 0; | |
2076 | strm.avail_in = 0; | |
2077 | if (plain_lzma) { | |
2078 | initRet = lzma_alone_decoder(&strm, UINT64_MAX); /* LZMA */ | |
2079 | } else { | |
2080 | initRet = lzma_stream_decoder(&strm, UINT64_MAX, 0); /* XZ */ | |
2081 | } | |
2082 | ||
2083 | if (initRet != LZMA_OK) { | |
2084 | DISPLAYLEVEL(1, "zstd: %s: %s error %d \n", | |
2085 | plain_lzma ? "lzma_alone_decoder" : "lzma_stream_decoder", | |
2086 | srcFileName, initRet); | |
2087 | return FIO_ERROR_FRAME_DECODING; | |
2088 | } | |
2089 | ||
2090 | strm.next_out = (BYTE*)ress->dstBuffer; | |
2091 | strm.avail_out = ress->dstBufferSize; | |
2092 | strm.next_in = (BYTE const*)ress->srcBuffer; | |
2093 | strm.avail_in = ress->srcBufferLoaded; | |
2094 | ||
2095 | for ( ; ; ) { | |
2096 | lzma_ret ret; | |
2097 | if (strm.avail_in == 0) { | |
2098 | ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile); | |
2099 | if (ress->srcBufferLoaded == 0) action = LZMA_FINISH; | |
2100 | strm.next_in = (BYTE const*)ress->srcBuffer; | |
2101 | strm.avail_in = ress->srcBufferLoaded; | |
2102 | } | |
2103 | ret = lzma_code(&strm, action); | |
2104 | ||
2105 | if (ret == LZMA_BUF_ERROR) { | |
2106 | DISPLAYLEVEL(1, "zstd: %s: premature lzma end \n", srcFileName); | |
2107 | decodingError = 1; break; | |
2108 | } | |
2109 | if (ret != LZMA_OK && ret != LZMA_STREAM_END) { | |
2110 | DISPLAYLEVEL(1, "zstd: %s: lzma_code decoding error %d \n", | |
2111 | srcFileName, ret); | |
2112 | decodingError = 1; break; | |
2113 | } | |
2114 | { size_t const decompBytes = ress->dstBufferSize - strm.avail_out; | |
2115 | if (decompBytes) { | |
f67539c2 | 2116 | storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decompBytes, prefs, storedSkips); |
11fdf7f2 TL |
2117 | outFileSize += decompBytes; |
2118 | strm.next_out = (BYTE*)ress->dstBuffer; | |
2119 | strm.avail_out = ress->dstBufferSize; | |
2120 | } } | |
2121 | if (ret == LZMA_STREAM_END) break; | |
2122 | } | |
2123 | ||
2124 | if (strm.avail_in > 0) | |
2125 | memmove(ress->srcBuffer, strm.next_in, strm.avail_in); | |
2126 | ress->srcBufferLoaded = strm.avail_in; | |
2127 | lzma_end(&strm); | |
f67539c2 | 2128 | FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips); |
11fdf7f2 TL |
2129 | return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize; |
2130 | } | |
2131 | #endif | |
2132 | ||
2133 | #ifdef ZSTD_LZ4DECOMPRESS | |
f67539c2 TL |
2134 | static unsigned long long |
2135 | FIO_decompressLz4Frame(dRess_t* ress, FILE* srcFile, | |
2136 | const FIO_prefs_t* const prefs, | |
2137 | const char* srcFileName) | |
7c673cae | 2138 | { |
7c673cae | 2139 | unsigned long long filesize = 0; |
11fdf7f2 TL |
2140 | LZ4F_errorCode_t nextToLoad; |
2141 | LZ4F_decompressionContext_t dCtx; | |
2142 | LZ4F_errorCode_t const errorCode = LZ4F_createDecompressionContext(&dCtx, LZ4F_VERSION); | |
2143 | int decodingError = 0; | |
f67539c2 | 2144 | unsigned storedSkips = 0; |
11fdf7f2 TL |
2145 | |
2146 | if (LZ4F_isError(errorCode)) { | |
2147 | DISPLAYLEVEL(1, "zstd: failed to create lz4 decompression context \n"); | |
2148 | return FIO_ERROR_FRAME_DECODING; | |
2149 | } | |
7c673cae | 2150 | |
11fdf7f2 TL |
2151 | /* Init feed with magic number (already consumed from FILE* sFile) */ |
2152 | { size_t inSize = 4; | |
2153 | size_t outSize= 0; | |
2154 | MEM_writeLE32(ress->srcBuffer, LZ4_MAGICNUMBER); | |
2155 | nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &outSize, ress->srcBuffer, &inSize, NULL); | |
2156 | if (LZ4F_isError(nextToLoad)) { | |
2157 | DISPLAYLEVEL(1, "zstd: %s: lz4 header error : %s \n", | |
2158 | srcFileName, LZ4F_getErrorName(nextToLoad)); | |
2159 | LZ4F_freeDecompressionContext(dCtx); | |
2160 | return FIO_ERROR_FRAME_DECODING; | |
2161 | } } | |
2162 | ||
2163 | /* Main Loop */ | |
2164 | for (;nextToLoad;) { | |
2165 | size_t readSize; | |
2166 | size_t pos = 0; | |
2167 | size_t decodedBytes = ress->dstBufferSize; | |
2168 | ||
2169 | /* Read input */ | |
2170 | if (nextToLoad > ress->srcBufferSize) nextToLoad = ress->srcBufferSize; | |
2171 | readSize = fread(ress->srcBuffer, 1, nextToLoad, srcFile); | |
2172 | if (!readSize) break; /* reached end of file or stream */ | |
2173 | ||
2174 | while ((pos < readSize) || (decodedBytes == ress->dstBufferSize)) { /* still to read, or still to flush */ | |
2175 | /* Decode Input (at least partially) */ | |
2176 | size_t remaining = readSize - pos; | |
2177 | decodedBytes = ress->dstBufferSize; | |
2178 | nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &decodedBytes, (char*)(ress->srcBuffer)+pos, &remaining, NULL); | |
2179 | if (LZ4F_isError(nextToLoad)) { | |
2180 | DISPLAYLEVEL(1, "zstd: %s: lz4 decompression error : %s \n", | |
2181 | srcFileName, LZ4F_getErrorName(nextToLoad)); | |
9f95a23c | 2182 | decodingError = 1; nextToLoad = 0; break; |
11fdf7f2 TL |
2183 | } |
2184 | pos += remaining; | |
2185 | ||
2186 | /* Write Block */ | |
2187 | if (decodedBytes) { | |
f67539c2 | 2188 | storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decodedBytes, prefs, storedSkips); |
11fdf7f2 TL |
2189 | filesize += decodedBytes; |
2190 | DISPLAYUPDATE(2, "\rDecompressed : %u MB ", (unsigned)(filesize>>20)); | |
2191 | } | |
2192 | ||
2193 | if (!nextToLoad) break; | |
2194 | } | |
2195 | } | |
2196 | /* can be out because readSize == 0, which could be an fread() error */ | |
2197 | if (ferror(srcFile)) { | |
2198 | DISPLAYLEVEL(1, "zstd: %s: read error \n", srcFileName); | |
2199 | decodingError=1; | |
7c673cae FG |
2200 | } |
2201 | ||
11fdf7f2 TL |
2202 | if (nextToLoad!=0) { |
2203 | DISPLAYLEVEL(1, "zstd: %s: unfinished lz4 stream \n", srcFileName); | |
2204 | decodingError=1; | |
2205 | } | |
2206 | ||
2207 | LZ4F_freeDecompressionContext(dCtx); | |
2208 | ress->srcBufferLoaded = 0; /* LZ4F will reach exact frame boundary */ | |
f67539c2 | 2209 | FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips); |
11fdf7f2 TL |
2210 | |
2211 | return decodingError ? FIO_ERROR_FRAME_DECODING : filesize; | |
2212 | } | |
2213 | #endif | |
2214 | ||
2215 | ||
2216 | ||
2217 | /** FIO_decompressFrames() : | |
2218 | * Find and decode frames inside srcFile | |
2219 | * srcFile presumed opened and valid | |
2220 | * @return : 0 : OK | |
2221 | * 1 : error | |
2222 | */ | |
f67539c2 TL |
2223 | static int FIO_decompressFrames(dRess_t ress, FILE* srcFile, |
2224 | const FIO_prefs_t* const prefs, | |
2225 | const char* dstFileName, const char* srcFileName) | |
11fdf7f2 TL |
2226 | { |
2227 | unsigned readSomething = 0; | |
2228 | unsigned long long filesize = 0; | |
2229 | assert(srcFile != NULL); | |
7c673cae FG |
2230 | |
2231 | /* for each frame */ | |
2232 | for ( ; ; ) { | |
2233 | /* check magic number -> version */ | |
2234 | size_t const toRead = 4; | |
11fdf7f2 TL |
2235 | const BYTE* const buf = (const BYTE*)ress.srcBuffer; |
2236 | if (ress.srcBufferLoaded < toRead) /* load up to 4 bytes for header */ | |
2237 | ress.srcBufferLoaded += fread((char*)ress.srcBuffer + ress.srcBufferLoaded, | |
2238 | (size_t)1, toRead - ress.srcBufferLoaded, srcFile); | |
7c673cae | 2239 | if (ress.srcBufferLoaded==0) { |
11fdf7f2 TL |
2240 | if (readSomething==0) { /* srcFile is empty (which is invalid) */ |
2241 | DISPLAYLEVEL(1, "zstd: %s: unexpected end of file \n", srcFileName); | |
2242 | return 1; | |
2243 | } /* else, just reached frame boundary */ | |
7c673cae FG |
2244 | break; /* no more input */ |
2245 | } | |
11fdf7f2 TL |
2246 | readSomething = 1; /* there is at least 1 byte in srcFile */ |
2247 | if (ress.srcBufferLoaded < toRead) { | |
2248 | DISPLAYLEVEL(1, "zstd: %s: unknown header \n", srcFileName); | |
2249 | return 1; | |
2250 | } | |
2251 | if (ZSTD_isFrame(buf, ress.srcBufferLoaded)) { | |
f67539c2 | 2252 | unsigned long long const frameSize = FIO_decompressZstdFrame(&ress, srcFile, prefs, srcFileName, filesize); |
11fdf7f2 TL |
2253 | if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; |
2254 | filesize += frameSize; | |
2255 | } else if (buf[0] == 31 && buf[1] == 139) { /* gz magic number */ | |
7c673cae | 2256 | #ifdef ZSTD_GZDECOMPRESS |
f67539c2 | 2257 | unsigned long long const frameSize = FIO_decompressGzFrame(&ress, srcFile, prefs, srcFileName); |
11fdf7f2 TL |
2258 | if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; |
2259 | filesize += frameSize; | |
2260 | #else | |
2261 | DISPLAYLEVEL(1, "zstd: %s: gzip file cannot be uncompressed (zstd compiled without HAVE_ZLIB) -- ignored \n", srcFileName); | |
2262 | return 1; | |
2263 | #endif | |
2264 | } else if ((buf[0] == 0xFD && buf[1] == 0x37) /* xz magic number */ | |
2265 | || (buf[0] == 0x5D && buf[1] == 0x00)) { /* lzma header (no magic number) */ | |
2266 | #ifdef ZSTD_LZMADECOMPRESS | |
f67539c2 | 2267 | unsigned long long const frameSize = FIO_decompressLzmaFrame(&ress, srcFile, prefs, srcFileName, buf[0] != 0xFD); |
11fdf7f2 TL |
2268 | if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; |
2269 | filesize += frameSize; | |
2270 | #else | |
2271 | DISPLAYLEVEL(1, "zstd: %s: xz/lzma file cannot be uncompressed (zstd compiled without HAVE_LZMA) -- ignored \n", srcFileName); | |
2272 | return 1; | |
2273 | #endif | |
2274 | } else if (MEM_readLE32(buf) == LZ4_MAGICNUMBER) { | |
2275 | #ifdef ZSTD_LZ4DECOMPRESS | |
f67539c2 | 2276 | unsigned long long const frameSize = FIO_decompressLz4Frame(&ress, srcFile, prefs, srcFileName); |
11fdf7f2 TL |
2277 | if (frameSize == FIO_ERROR_FRAME_DECODING) return 1; |
2278 | filesize += frameSize; | |
7c673cae | 2279 | #else |
11fdf7f2 | 2280 | DISPLAYLEVEL(1, "zstd: %s: lz4 file cannot be uncompressed (zstd compiled without HAVE_LZ4) -- ignored \n", srcFileName); |
7c673cae FG |
2281 | return 1; |
2282 | #endif | |
9f95a23c TL |
2283 | } else if ((prefs->overwrite) && !strcmp (dstFileName, stdoutmark)) { /* pass-through mode */ |
2284 | return FIO_passThrough(prefs, | |
2285 | ress.dstFile, srcFile, | |
2286 | ress.srcBuffer, ress.srcBufferSize, | |
2287 | ress.srcBufferLoaded); | |
7c673cae | 2288 | } else { |
11fdf7f2 TL |
2289 | DISPLAYLEVEL(1, "zstd: %s: unsupported format \n", srcFileName); |
2290 | return 1; | |
2291 | } } /* for each frame */ | |
7c673cae FG |
2292 | |
2293 | /* Final Status */ | |
2294 | DISPLAYLEVEL(2, "\r%79s\r", ""); | |
2295 | DISPLAYLEVEL(2, "%-20s: %llu bytes \n", srcFileName, filesize); | |
2296 | ||
7c673cae FG |
2297 | return 0; |
2298 | } | |
2299 | ||
9f95a23c TL |
2300 | /** FIO_decompressDstFile() : |
2301 | open `dstFileName`, | |
2302 | or path-through if ress.dstFile is already != 0, | |
2303 | then start decompression process (FIO_decompressFrames()). | |
2304 | @return : 0 : OK | |
2305 | 1 : operation aborted | |
2306 | */ | |
2307 | static int FIO_decompressDstFile(FIO_prefs_t* const prefs, | |
2308 | dRess_t ress, FILE* srcFile, | |
2309 | const char* dstFileName, const char* srcFileName) | |
2310 | { | |
2311 | int result; | |
2312 | stat_t statbuf; | |
2313 | int transfer_permissions = 0; | |
2314 | int releaseDstFile = 0; | |
2315 | ||
f67539c2 | 2316 | if ((ress.dstFile == NULL) && (prefs->testMode==0)) { |
9f95a23c TL |
2317 | releaseDstFile = 1; |
2318 | ||
2319 | ress.dstFile = FIO_openDstFile(prefs, srcFileName, dstFileName); | |
f67539c2 | 2320 | if (ress.dstFile==NULL) return 1; |
9f95a23c TL |
2321 | |
2322 | /* Must only be added after FIO_openDstFile() succeeds. | |
2323 | * Otherwise we may delete the destination file if it already exists, | |
2324 | * and the user presses Ctrl-C when asked if they wish to overwrite. | |
2325 | */ | |
2326 | addHandler(dstFileName); | |
2327 | ||
2328 | if ( strcmp(srcFileName, stdinmark) /* special case : don't transfer permissions from stdin */ | |
2329 | && UTIL_getFileStat(srcFileName, &statbuf) ) | |
2330 | transfer_permissions = 1; | |
2331 | } | |
2332 | ||
f67539c2 | 2333 | result = FIO_decompressFrames(ress, srcFile, prefs, dstFileName, srcFileName); |
9f95a23c TL |
2334 | |
2335 | if (releaseDstFile) { | |
2336 | FILE* const dstFile = ress.dstFile; | |
2337 | clearHandler(); | |
2338 | ress.dstFile = NULL; | |
2339 | if (fclose(dstFile)) { | |
2340 | DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno)); | |
2341 | result = 1; | |
2342 | } | |
2343 | ||
2344 | if ( (result != 0) /* operation failure */ | |
2345 | && strcmp(dstFileName, nulmark) /* special case : don't remove() /dev/null (#316) */ | |
2346 | && strcmp(dstFileName, stdoutmark) /* special case : don't remove() stdout */ | |
2347 | ) { | |
2348 | FIO_remove(dstFileName); /* remove decompression artefact; note: don't do anything special if remove() fails */ | |
2349 | } else { /* operation success */ | |
2350 | if ( strcmp(dstFileName, stdoutmark) /* special case : don't chmod stdout */ | |
2351 | && strcmp(dstFileName, nulmark) /* special case : don't chmod /dev/null */ | |
2352 | && transfer_permissions ) /* file permissions correctly extracted from src */ | |
2353 | UTIL_setFileStat(dstFileName, &statbuf); /* transfer file permissions from src into dst */ | |
2354 | } | |
2355 | } | |
2356 | ||
2357 | return result; | |
2358 | } | |
2359 | ||
7c673cae | 2360 | |
11fdf7f2 | 2361 | /** FIO_decompressSrcFile() : |
9f95a23c | 2362 | Open `srcFileName`, transfer control to decompressDstFile() |
11fdf7f2 | 2363 | @return : 0 : OK |
9f95a23c | 2364 | 1 : error |
11fdf7f2 | 2365 | */ |
9f95a23c | 2366 | static int FIO_decompressSrcFile(FIO_prefs_t* const prefs, dRess_t ress, const char* dstFileName, const char* srcFileName) |
11fdf7f2 TL |
2367 | { |
2368 | FILE* srcFile; | |
2369 | int result; | |
2370 | ||
2371 | if (UTIL_isDirectory(srcFileName)) { | |
2372 | DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName); | |
2373 | return 1; | |
2374 | } | |
2375 | ||
2376 | srcFile = FIO_openSrcFile(srcFileName); | |
2377 | if (srcFile==NULL) return 1; | |
9f95a23c | 2378 | ress.srcBufferLoaded = 0; |
11fdf7f2 | 2379 | |
9f95a23c | 2380 | result = FIO_decompressDstFile(prefs, ress, srcFile, dstFileName, srcFileName); |
11fdf7f2 TL |
2381 | |
2382 | /* Close file */ | |
2383 | if (fclose(srcFile)) { | |
2384 | DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno)); /* error should not happen */ | |
2385 | return 1; | |
2386 | } | |
9f95a23c TL |
2387 | if ( prefs->removeSrcFile /* --rm */ |
2388 | && (result==0) /* decompression successful */ | |
11fdf7f2 | 2389 | && strcmp(srcFileName, stdinmark) ) /* not stdin */ { |
9f95a23c TL |
2390 | /* We must clear the handler, since after this point calling it would |
2391 | * delete both the source and destination files. | |
2392 | */ | |
2393 | clearHandler(); | |
2394 | if (FIO_remove(srcFileName)) { | |
11fdf7f2 TL |
2395 | /* failed to remove src file */ |
2396 | DISPLAYLEVEL(1, "zstd: %s: %s \n", srcFileName, strerror(errno)); | |
2397 | return 1; | |
2398 | } } | |
2399 | return result; | |
2400 | } | |
2401 | ||
2402 | ||
9f95a23c TL |
2403 | |
2404 | int FIO_decompressFilename(FIO_prefs_t* const prefs, | |
2405 | const char* dstFileName, const char* srcFileName, | |
2406 | const char* dictFileName) | |
7c673cae | 2407 | { |
9f95a23c | 2408 | dRess_t const ress = FIO_createDResources(prefs, dictFileName); |
7c673cae | 2409 | |
9f95a23c | 2410 | int const decodingError = FIO_decompressSrcFile(prefs, ress, dstFileName, srcFileName); |
7c673cae | 2411 | |
9f95a23c TL |
2412 | FIO_freeDResources(ress); |
2413 | return decodingError; | |
2414 | } | |
11fdf7f2 | 2415 | |
f67539c2 TL |
2416 | static const char *suffixList[] = { |
2417 | ZSTD_EXTENSION, | |
2418 | TZSTD_EXTENSION, | |
2419 | #ifdef ZSTD_GZDECOMPRESS | |
2420 | GZ_EXTENSION, | |
2421 | TGZ_EXTENSION, | |
2422 | #endif | |
2423 | #ifdef ZSTD_LZMADECOMPRESS | |
2424 | LZMA_EXTENSION, | |
2425 | XZ_EXTENSION, | |
2426 | TXZ_EXTENSION, | |
2427 | #endif | |
2428 | #ifdef ZSTD_LZ4DECOMPRESS | |
2429 | LZ4_EXTENSION, | |
2430 | TLZ4_EXTENSION, | |
2431 | #endif | |
2432 | NULL | |
2433 | }; | |
2434 | ||
2435 | static const char *suffixListStr = | |
2436 | ZSTD_EXTENSION "/" TZSTD_EXTENSION | |
2437 | #ifdef ZSTD_GZDECOMPRESS | |
2438 | "/" GZ_EXTENSION "/" TGZ_EXTENSION | |
2439 | #endif | |
2440 | #ifdef ZSTD_LZMADECOMPRESS | |
2441 | "/" LZMA_EXTENSION "/" XZ_EXTENSION "/" TXZ_EXTENSION | |
2442 | #endif | |
2443 | #ifdef ZSTD_LZ4DECOMPRESS | |
2444 | "/" LZ4_EXTENSION "/" TLZ4_EXTENSION | |
2445 | #endif | |
2446 | ; | |
7c673cae | 2447 | |
9f95a23c TL |
2448 | /* FIO_determineDstName() : |
2449 | * create a destination filename from a srcFileName. | |
2450 | * @return a pointer to it. | |
2451 | * @return == NULL if there is an error */ | |
2452 | static const char* | |
f67539c2 | 2453 | FIO_determineDstName(const char* srcFileName, const char* outDirName) |
9f95a23c TL |
2454 | { |
2455 | static size_t dfnbCapacity = 0; | |
2456 | static char* dstFileNameBuffer = NULL; /* using static allocation : this function cannot be multi-threaded */ | |
f67539c2 TL |
2457 | size_t dstFileNameEndPos; |
2458 | char* outDirFilename = NULL; | |
2459 | const char* dstSuffix = ""; | |
2460 | size_t dstSuffixLen = 0; | |
2461 | ||
2462 | size_t sfnSize = strlen(srcFileName); | |
2463 | ||
2464 | size_t srcSuffixLen; | |
2465 | const char* const srcSuffix = strrchr(srcFileName, '.'); | |
2466 | if (srcSuffix == NULL) { | |
2467 | DISPLAYLEVEL(1, | |
2468 | "zstd: %s: unknown suffix (%s expected). " | |
2469 | "Can't derive the output file name. " | |
2470 | "Specify it with -o dstFileName. Ignoring.\n", | |
2471 | srcFileName, suffixListStr); | |
9f95a23c | 2472 | return NULL; |
11fdf7f2 | 2473 | } |
f67539c2 TL |
2474 | srcSuffixLen = strlen(srcSuffix); |
2475 | ||
2476 | { | |
2477 | const char** matchedSuffixPtr; | |
2478 | for (matchedSuffixPtr = suffixList; *matchedSuffixPtr != NULL; matchedSuffixPtr++) { | |
2479 | if (!strcmp(*matchedSuffixPtr, srcSuffix)) { | |
2480 | break; | |
2481 | } | |
2482 | } | |
2483 | ||
2484 | /* check suffix is authorized */ | |
2485 | if (sfnSize <= srcSuffixLen || *matchedSuffixPtr == NULL) { | |
2486 | DISPLAYLEVEL(1, | |
2487 | "zstd: %s: unknown suffix (%s expected). " | |
2488 | "Can't derive the output file name. " | |
2489 | "Specify it with -o dstFileName. Ignoring.\n", | |
2490 | srcFileName, suffixListStr); | |
2491 | return NULL; | |
2492 | } | |
2493 | ||
2494 | if ((*matchedSuffixPtr)[1] == 't') { | |
2495 | dstSuffix = ".tar"; | |
2496 | dstSuffixLen = strlen(dstSuffix); | |
2497 | } | |
2498 | } | |
2499 | ||
2500 | if (outDirName) { | |
2501 | outDirFilename = FIO_createFilename_fromOutDir(srcFileName, outDirName, 0); | |
2502 | sfnSize = strlen(outDirFilename); | |
2503 | assert(outDirFilename != NULL); | |
11fdf7f2 TL |
2504 | } |
2505 | ||
f67539c2 TL |
2506 | if (dfnbCapacity+srcSuffixLen <= sfnSize+1+dstSuffixLen) { |
2507 | /* allocate enough space to write dstFilename into it */ | |
9f95a23c TL |
2508 | free(dstFileNameBuffer); |
2509 | dfnbCapacity = sfnSize + 20; | |
2510 | dstFileNameBuffer = (char*)malloc(dfnbCapacity); | |
2511 | if (dstFileNameBuffer==NULL) | |
f67539c2 TL |
2512 | EXM_THROW(74, "%s : not enough memory for dstFileName", |
2513 | strerror(errno)); | |
9f95a23c | 2514 | } |
7c673cae | 2515 | |
9f95a23c TL |
2516 | /* return dst name == src name truncated from suffix */ |
2517 | assert(dstFileNameBuffer != NULL); | |
f67539c2 TL |
2518 | dstFileNameEndPos = sfnSize - srcSuffixLen; |
2519 | if (outDirFilename) { | |
2520 | memcpy(dstFileNameBuffer, outDirFilename, dstFileNameEndPos); | |
2521 | free(outDirFilename); | |
2522 | } else { | |
2523 | memcpy(dstFileNameBuffer, srcFileName, dstFileNameEndPos); | |
2524 | } | |
2525 | ||
2526 | /* The short tar extensions tzst, tgz, txz and tlz4 files should have "tar" | |
2527 | * extension on decompression. Also writes terminating null. */ | |
2528 | strcpy(dstFileNameBuffer + dstFileNameEndPos, dstSuffix); | |
9f95a23c | 2529 | return dstFileNameBuffer; |
7c673cae | 2530 | |
9f95a23c | 2531 | /* note : dstFileNameBuffer memory is not going to be free */ |
7c673cae FG |
2532 | } |
2533 | ||
2534 | ||
9f95a23c TL |
2535 | int |
2536 | FIO_decompressMultipleFilenames(FIO_prefs_t* const prefs, | |
f67539c2 TL |
2537 | const char** srcNamesTable, unsigned nbFiles, |
2538 | const char* outDirName, const char* outFileName, | |
9f95a23c | 2539 | const char* dictFileName) |
7c673cae | 2540 | { |
9f95a23c TL |
2541 | int error = 0; |
2542 | dRess_t ress = FIO_createDResources(prefs, dictFileName); | |
7c673cae | 2543 | |
9f95a23c | 2544 | if (outFileName) { |
7c673cae | 2545 | unsigned u; |
f67539c2 TL |
2546 | if (!prefs->testMode) { |
2547 | ress.dstFile = FIO_openDstFile(prefs, NULL, outFileName); | |
2548 | if (ress.dstFile == 0) EXM_THROW(19, "cannot open %s", outFileName); | |
2549 | } | |
7c673cae | 2550 | for (u=0; u<nbFiles; u++) |
9f95a23c | 2551 | error |= FIO_decompressSrcFile(prefs, ress, outFileName, srcNamesTable[u]); |
f67539c2 | 2552 | if ((!prefs->testMode) && (fclose(ress.dstFile))) |
9f95a23c TL |
2553 | EXM_THROW(72, "Write error : %s : cannot properly close output file", |
2554 | strerror(errno)); | |
7c673cae | 2555 | } else { |
7c673cae | 2556 | unsigned u; |
7c673cae FG |
2557 | for (u=0; u<nbFiles; u++) { /* create dstFileName */ |
2558 | const char* const srcFileName = srcNamesTable[u]; | |
f67539c2 | 2559 | const char* const dstFileName = FIO_determineDstName(srcFileName, outDirName); |
9f95a23c TL |
2560 | if (dstFileName == NULL) { error=1; continue; } |
2561 | ||
2562 | error |= FIO_decompressSrcFile(prefs, ress, dstFileName, srcFileName); | |
7c673cae | 2563 | } |
f67539c2 TL |
2564 | if (outDirName) |
2565 | FIO_checkFilenameCollisions(srcNamesTable ,nbFiles); | |
7c673cae FG |
2566 | } |
2567 | ||
2568 | FIO_freeDResources(ress); | |
9f95a23c | 2569 | return error; |
7c673cae FG |
2570 | } |
2571 | ||
11fdf7f2 TL |
2572 | /* ************************************************************************** |
2573 | * .zst file info (--list command) | |
2574 | ***************************************************************************/ | |
2575 | ||
2576 | typedef struct { | |
2577 | U64 decompressedSize; | |
2578 | U64 compressedSize; | |
2579 | U64 windowSize; | |
2580 | int numActualFrames; | |
2581 | int numSkippableFrames; | |
2582 | int decompUnavailable; | |
2583 | int usesCheck; | |
2584 | U32 nbFiles; | |
2585 | } fileInfo_t; | |
2586 | ||
9f95a23c TL |
2587 | typedef enum { |
2588 | info_success=0, | |
2589 | info_frame_error=1, | |
2590 | info_not_zstd=2, | |
2591 | info_file_error=3, | |
2592 | info_truncated_input=4, | |
2593 | } InfoError; | |
2594 | ||
2595 | #define ERROR_IF(c,n,...) { \ | |
2596 | if (c) { \ | |
2597 | DISPLAYLEVEL(1, __VA_ARGS__); \ | |
2598 | DISPLAYLEVEL(1, " \n"); \ | |
2599 | return n; \ | |
2600 | } \ | |
2601 | } | |
11fdf7f2 | 2602 | |
9f95a23c TL |
2603 | static InfoError |
2604 | FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile) | |
2605 | { | |
11fdf7f2 TL |
2606 | /* begin analyzing frame */ |
2607 | for ( ; ; ) { | |
2608 | BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX]; | |
2609 | size_t const numBytesRead = fread(headerBuffer, 1, sizeof(headerBuffer), srcFile); | |
f67539c2 | 2610 | if (numBytesRead < ZSTD_FRAMEHEADERSIZE_MIN(ZSTD_f_zstd1)) { |
11fdf7f2 TL |
2611 | if ( feof(srcFile) |
2612 | && (numBytesRead == 0) | |
9f95a23c TL |
2613 | && (info->compressedSize > 0) |
2614 | && (info->compressedSize != UTIL_FILESIZE_UNKNOWN) ) { | |
2615 | unsigned long long file_position = (unsigned long long) LONG_TELL(srcFile); | |
2616 | unsigned long long file_size = (unsigned long long) info->compressedSize; | |
2617 | ERROR_IF(file_position != file_size, info_truncated_input, | |
2618 | "Error: seeked to position %llu, which is beyond file size of %llu\n", | |
2619 | file_position, | |
2620 | file_size); | |
2621 | break; /* correct end of file => success */ | |
11fdf7f2 | 2622 | } |
9f95a23c TL |
2623 | ERROR_IF(feof(srcFile), info_not_zstd, "Error: reached end of file with incomplete frame"); |
2624 | ERROR_IF(1, info_frame_error, "Error: did not reach end of file but ran out of frames"); | |
11fdf7f2 TL |
2625 | } |
2626 | { U32 const magicNumber = MEM_readLE32(headerBuffer); | |
2627 | /* Zstandard frame */ | |
2628 | if (magicNumber == ZSTD_MAGICNUMBER) { | |
2629 | ZSTD_frameHeader header; | |
2630 | U64 const frameContentSize = ZSTD_getFrameContentSize(headerBuffer, numBytesRead); | |
9f95a23c TL |
2631 | if ( frameContentSize == ZSTD_CONTENTSIZE_ERROR |
2632 | || frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN ) { | |
11fdf7f2 TL |
2633 | info->decompUnavailable = 1; |
2634 | } else { | |
2635 | info->decompressedSize += frameContentSize; | |
2636 | } | |
9f95a23c TL |
2637 | ERROR_IF(ZSTD_getFrameHeader(&header, headerBuffer, numBytesRead) != 0, |
2638 | info_frame_error, "Error: could not decode frame header"); | |
11fdf7f2 TL |
2639 | info->windowSize = header.windowSize; |
2640 | /* move to the end of the frame header */ | |
2641 | { size_t const headerSize = ZSTD_frameHeaderSize(headerBuffer, numBytesRead); | |
9f95a23c TL |
2642 | ERROR_IF(ZSTD_isError(headerSize), info_frame_error, "Error: could not determine frame header size"); |
2643 | ERROR_IF(fseek(srcFile, ((long)headerSize)-((long)numBytesRead), SEEK_CUR) != 0, | |
2644 | info_frame_error, "Error: could not move to end of frame header"); | |
2645 | } | |
2646 | ||
2647 | /* skip all blocks in the frame */ | |
11fdf7f2 TL |
2648 | { int lastBlock = 0; |
2649 | do { | |
2650 | BYTE blockHeaderBuffer[3]; | |
9f95a23c TL |
2651 | ERROR_IF(fread(blockHeaderBuffer, 1, 3, srcFile) != 3, |
2652 | info_frame_error, "Error while reading block header"); | |
11fdf7f2 TL |
2653 | { U32 const blockHeader = MEM_readLE24(blockHeaderBuffer); |
2654 | U32 const blockTypeID = (blockHeader >> 1) & 3; | |
2655 | U32 const isRLE = (blockTypeID == 1); | |
2656 | U32 const isWrongBlock = (blockTypeID == 3); | |
2657 | long const blockSize = isRLE ? 1 : (long)(blockHeader >> 3); | |
9f95a23c | 2658 | ERROR_IF(isWrongBlock, info_frame_error, "Error: unsupported block type"); |
11fdf7f2 | 2659 | lastBlock = blockHeader & 1; |
9f95a23c TL |
2660 | ERROR_IF(fseek(srcFile, blockSize, SEEK_CUR) != 0, |
2661 | info_frame_error, "Error: could not skip to end of block"); | |
2662 | } | |
11fdf7f2 | 2663 | } while (lastBlock != 1); |
11fdf7f2 TL |
2664 | } |
2665 | ||
2666 | /* check if checksum is used */ | |
2667 | { BYTE const frameHeaderDescriptor = headerBuffer[4]; | |
2668 | int const contentChecksumFlag = (frameHeaderDescriptor & (1 << 2)) >> 2; | |
2669 | if (contentChecksumFlag) { | |
11fdf7f2 | 2670 | info->usesCheck = 1; |
9f95a23c TL |
2671 | ERROR_IF(fseek(srcFile, 4, SEEK_CUR) != 0, |
2672 | info_frame_error, "Error: could not skip past checksum"); | |
2673 | } } | |
11fdf7f2 TL |
2674 | info->numActualFrames++; |
2675 | } | |
2676 | /* Skippable frame */ | |
9f95a23c | 2677 | else if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { |
11fdf7f2 TL |
2678 | U32 const frameSize = MEM_readLE32(headerBuffer + 4); |
2679 | long const seek = (long)(8 + frameSize - numBytesRead); | |
9f95a23c TL |
2680 | ERROR_IF(LONG_SEEK(srcFile, seek, SEEK_CUR) != 0, |
2681 | info_frame_error, "Error: could not find end of skippable frame"); | |
11fdf7f2 TL |
2682 | info->numSkippableFrames++; |
2683 | } | |
2684 | /* unknown content */ | |
2685 | else { | |
9f95a23c | 2686 | return info_not_zstd; |
11fdf7f2 | 2687 | } |
9f95a23c TL |
2688 | } /* magic number analysis */ |
2689 | } /* end analyzing frames */ | |
2690 | return info_success; | |
2691 | } | |
2692 | ||
2693 | ||
2694 | static InfoError | |
2695 | getFileInfo_fileConfirmed(fileInfo_t* info, const char* inFileName) | |
2696 | { | |
2697 | InfoError status; | |
2698 | FILE* const srcFile = FIO_openSrcFile(inFileName); | |
2699 | ERROR_IF(srcFile == NULL, info_file_error, "Error: could not open source file %s", inFileName); | |
2700 | ||
2701 | info->compressedSize = UTIL_getFileSize(inFileName); | |
2702 | status = FIO_analyzeFrames(info, srcFile); | |
2703 | ||
11fdf7f2 TL |
2704 | fclose(srcFile); |
2705 | info->nbFiles = 1; | |
9f95a23c TL |
2706 | return status; |
2707 | } | |
2708 | ||
2709 | ||
2710 | /** getFileInfo() : | |
2711 | * Reads information from file, stores in *info | |
2712 | * @return : InfoError status | |
2713 | */ | |
2714 | static InfoError | |
2715 | getFileInfo(fileInfo_t* info, const char* srcFileName) | |
2716 | { | |
2717 | ERROR_IF(!UTIL_isRegularFile(srcFileName), | |
2718 | info_file_error, "Error : %s is not a file", srcFileName); | |
2719 | return getFileInfo_fileConfirmed(info, srcFileName); | |
11fdf7f2 TL |
2720 | } |
2721 | ||
9f95a23c TL |
2722 | |
2723 | static void | |
2724 | displayInfo(const char* inFileName, const fileInfo_t* info, int displayLevel) | |
2725 | { | |
11fdf7f2 TL |
2726 | unsigned const unit = info->compressedSize < (1 MB) ? (1 KB) : (1 MB); |
2727 | const char* const unitStr = info->compressedSize < (1 MB) ? "KB" : "MB"; | |
2728 | double const windowSizeUnit = (double)info->windowSize / unit; | |
2729 | double const compressedSizeUnit = (double)info->compressedSize / unit; | |
2730 | double const decompressedSizeUnit = (double)info->decompressedSize / unit; | |
2731 | double const ratio = (info->compressedSize == 0) ? 0 : ((double)info->decompressedSize)/info->compressedSize; | |
2732 | const char* const checkString = (info->usesCheck ? "XXH64" : "None"); | |
2733 | if (displayLevel <= 2) { | |
2734 | if (!info->decompUnavailable) { | |
2735 | DISPLAYOUT("%6d %5d %7.2f %2s %9.2f %2s %5.3f %5s %s\n", | |
2736 | info->numSkippableFrames + info->numActualFrames, | |
2737 | info->numSkippableFrames, | |
2738 | compressedSizeUnit, unitStr, decompressedSizeUnit, unitStr, | |
2739 | ratio, checkString, inFileName); | |
2740 | } else { | |
2741 | DISPLAYOUT("%6d %5d %7.2f %2s %5s %s\n", | |
2742 | info->numSkippableFrames + info->numActualFrames, | |
2743 | info->numSkippableFrames, | |
2744 | compressedSizeUnit, unitStr, | |
2745 | checkString, inFileName); | |
2746 | } | |
2747 | } else { | |
9f95a23c | 2748 | DISPLAYOUT("%s \n", inFileName); |
11fdf7f2 | 2749 | DISPLAYOUT("# Zstandard Frames: %d\n", info->numActualFrames); |
9f95a23c TL |
2750 | if (info->numSkippableFrames) |
2751 | DISPLAYOUT("# Skippable Frames: %d\n", info->numSkippableFrames); | |
11fdf7f2 TL |
2752 | DISPLAYOUT("Window Size: %.2f %2s (%llu B)\n", |
2753 | windowSizeUnit, unitStr, | |
2754 | (unsigned long long)info->windowSize); | |
2755 | DISPLAYOUT("Compressed Size: %.2f %2s (%llu B)\n", | |
2756 | compressedSizeUnit, unitStr, | |
2757 | (unsigned long long)info->compressedSize); | |
2758 | if (!info->decompUnavailable) { | |
2759 | DISPLAYOUT("Decompressed Size: %.2f %2s (%llu B)\n", | |
2760 | decompressedSizeUnit, unitStr, | |
2761 | (unsigned long long)info->decompressedSize); | |
2762 | DISPLAYOUT("Ratio: %.4f\n", ratio); | |
2763 | } | |
2764 | DISPLAYOUT("Check: %s\n", checkString); | |
2765 | DISPLAYOUT("\n"); | |
2766 | } | |
2767 | } | |
2768 | ||
2769 | static fileInfo_t FIO_addFInfo(fileInfo_t fi1, fileInfo_t fi2) | |
2770 | { | |
2771 | fileInfo_t total; | |
9f95a23c | 2772 | memset(&total, 0, sizeof(total)); |
11fdf7f2 TL |
2773 | total.numActualFrames = fi1.numActualFrames + fi2.numActualFrames; |
2774 | total.numSkippableFrames = fi1.numSkippableFrames + fi2.numSkippableFrames; | |
2775 | total.compressedSize = fi1.compressedSize + fi2.compressedSize; | |
2776 | total.decompressedSize = fi1.decompressedSize + fi2.decompressedSize; | |
2777 | total.decompUnavailable = fi1.decompUnavailable | fi2.decompUnavailable; | |
2778 | total.usesCheck = fi1.usesCheck & fi2.usesCheck; | |
2779 | total.nbFiles = fi1.nbFiles + fi2.nbFiles; | |
2780 | return total; | |
2781 | } | |
2782 | ||
9f95a23c TL |
2783 | static int |
2784 | FIO_listFile(fileInfo_t* total, const char* inFileName, int displayLevel) | |
2785 | { | |
11fdf7f2 TL |
2786 | fileInfo_t info; |
2787 | memset(&info, 0, sizeof(info)); | |
9f95a23c TL |
2788 | { InfoError const error = getFileInfo(&info, inFileName); |
2789 | switch (error) { | |
2790 | case info_frame_error: | |
2791 | /* display error, but provide output */ | |
2792 | DISPLAYLEVEL(1, "Error while parsing \"%s\" \n", inFileName); | |
2793 | break; | |
2794 | case info_not_zstd: | |
2795 | DISPLAYOUT("File \"%s\" not compressed by zstd \n", inFileName); | |
2796 | if (displayLevel > 2) DISPLAYOUT("\n"); | |
2797 | return 1; | |
2798 | case info_file_error: | |
2799 | /* error occurred while opening the file */ | |
2800 | if (displayLevel > 2) DISPLAYOUT("\n"); | |
2801 | return 1; | |
2802 | case info_truncated_input: | |
2803 | DISPLAYOUT("File \"%s\" is truncated \n", inFileName); | |
2804 | if (displayLevel > 2) DISPLAYOUT("\n"); | |
2805 | return 1; | |
2806 | case info_success: | |
2807 | default: | |
2808 | break; | |
11fdf7f2 | 2809 | } |
9f95a23c | 2810 | |
11fdf7f2 TL |
2811 | displayInfo(inFileName, &info, displayLevel); |
2812 | *total = FIO_addFInfo(*total, info); | |
9f95a23c | 2813 | assert(error == info_success || error == info_frame_error); |
f67539c2 | 2814 | return (int)error; |
11fdf7f2 TL |
2815 | } |
2816 | } | |
2817 | ||
9f95a23c TL |
2818 | int FIO_listMultipleFiles(unsigned numFiles, const char** filenameTable, int displayLevel) |
2819 | { | |
2820 | /* ensure no specified input is stdin (needs fseek() capability) */ | |
2821 | { unsigned u; | |
2822 | for (u=0; u<numFiles;u++) { | |
2823 | ERROR_IF(!strcmp (filenameTable[u], stdinmark), | |
2824 | 1, "zstd: --list does not support reading from standard input"); | |
2825 | } } | |
2826 | ||
11fdf7f2 | 2827 | if (numFiles == 0) { |
9f95a23c TL |
2828 | if (!IS_CONSOLE(stdin)) { |
2829 | DISPLAYLEVEL(1, "zstd: --list does not support reading from standard input \n"); | |
2830 | } | |
2831 | DISPLAYLEVEL(1, "No files given \n"); | |
2832 | return 1; | |
11fdf7f2 | 2833 | } |
9f95a23c | 2834 | |
11fdf7f2 TL |
2835 | if (displayLevel <= 2) { |
2836 | DISPLAYOUT("Frames Skips Compressed Uncompressed Ratio Check Filename\n"); | |
2837 | } | |
2838 | { int error = 0; | |
11fdf7f2 TL |
2839 | fileInfo_t total; |
2840 | memset(&total, 0, sizeof(total)); | |
2841 | total.usesCheck = 1; | |
9f95a23c TL |
2842 | /* --list each file, and check for any error */ |
2843 | { unsigned u; | |
2844 | for (u=0; u<numFiles;u++) { | |
2845 | error |= FIO_listFile(&total, filenameTable[u], displayLevel); | |
2846 | } } | |
2847 | if (numFiles > 1 && displayLevel <= 2) { /* display total */ | |
11fdf7f2 TL |
2848 | unsigned const unit = total.compressedSize < (1 MB) ? (1 KB) : (1 MB); |
2849 | const char* const unitStr = total.compressedSize < (1 MB) ? "KB" : "MB"; | |
2850 | double const compressedSizeUnit = (double)total.compressedSize / unit; | |
2851 | double const decompressedSizeUnit = (double)total.decompressedSize / unit; | |
2852 | double const ratio = (total.compressedSize == 0) ? 0 : ((double)total.decompressedSize)/total.compressedSize; | |
2853 | const char* const checkString = (total.usesCheck ? "XXH64" : ""); | |
2854 | DISPLAYOUT("----------------------------------------------------------------- \n"); | |
2855 | if (total.decompUnavailable) { | |
2856 | DISPLAYOUT("%6d %5d %7.2f %2s %5s %u files\n", | |
2857 | total.numSkippableFrames + total.numActualFrames, | |
2858 | total.numSkippableFrames, | |
2859 | compressedSizeUnit, unitStr, | |
9f95a23c | 2860 | checkString, (unsigned)total.nbFiles); |
11fdf7f2 TL |
2861 | } else { |
2862 | DISPLAYOUT("%6d %5d %7.2f %2s %9.2f %2s %5.3f %5s %u files\n", | |
2863 | total.numSkippableFrames + total.numActualFrames, | |
2864 | total.numSkippableFrames, | |
2865 | compressedSizeUnit, unitStr, decompressedSizeUnit, unitStr, | |
9f95a23c TL |
2866 | ratio, checkString, (unsigned)total.nbFiles); |
2867 | } } | |
11fdf7f2 TL |
2868 | return error; |
2869 | } | |
2870 | } | |
2871 | ||
2872 | ||
7c673cae | 2873 | #endif /* #ifndef ZSTD_NODECOMPRESS */ |