1 ///////////////////////////////////////////////////////////////////////////////
4 /// \brief Compresses or uncompresses a file
6 // Copyright (C) 2007 Lasse Collin
8 // This program is free software; you can redistribute it and/or
9 // modify it under the terms of the GNU Lesser General Public
10 // License as published by the Free Software Foundation; either
11 // version 2.1 of the License, or (at your option) any later version.
13 // This program is distributed in the hope that it will be useful,
14 // but WITHOUT ANY WARRANTY; without even the implied warranty of
15 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 // Lesser General Public License for more details.
18 ///////////////////////////////////////////////////////////////////////////////
23 enum operation_mode opt_mode = MODE_COMPRESS;
25 enum format_type opt_format = FORMAT_AUTO;
28 /// Stream used to communicate with liblzma
29 static lzma_stream strm = LZMA_STREAM_INIT;
31 /// Filters needed for all encoding all formats, and also decoding in raw data
32 static lzma_filter filters[LZMA_FILTERS_MAX + 1];
34 /// Number of filters. Zero indicates that we are using a preset.
35 static size_t filters_count = 0;
37 /// Number of the preset (0-9)
38 static size_t preset_number = 6;
40 /// True if we should auto-adjust the compression settings to use less memory
41 /// if memory usage limit is too low for the original settings.
42 static bool auto_adjust = true;
44 /// Indicate if no preset has been explicitly given. In that case, if we need
45 /// to auto-adjust for lower memory usage, we won't print a warning.
46 static bool preset_default = true;
48 /// If a preset is used (no custom filter chain) and preset_extreme is true,
49 /// a significantly slower compression is used to achieve slightly better
50 /// compression ratio.
51 static bool preset_extreme = false;
53 /// Integrity check type
54 #ifdef HAVE_CHECK_CRC64
55 static lzma_check check = LZMA_CHECK_CRC64;
57 static lzma_check check = LZMA_CHECK_CRC32;
62 coder_set_check(lzma_check new_check)
70 coder_set_preset(size_t new_preset)
72 preset_number = new_preset;
73 preset_default = false;
79 coder_set_extreme(void)
81 preset_extreme = true;
87 coder_add_filter(lzma_vli id, void *options)
89 if (filters_count == LZMA_FILTERS_MAX)
90 message_fatal(_("Maximum number of filters is four"));
92 filters[filters_count].id = id;
93 filters[filters_count].options = options;
100 static void lzma_attribute((noreturn))
101 memlimit_too_small(uint64_t memory_usage, uint64_t memory_limit)
103 message_fatal(_("Memory usage limit (%" PRIu64 " MiB) is too small "
104 "for the given filter setup (%" PRIu64 " MiB)"),
105 memory_limit >> 20, memory_usage >> 20);
110 coder_set_compression_settings(void)
112 // Options for LZMA1 or LZMA2 in case we are using a preset.
113 static lzma_options_lzma opt_lzma;
115 if (filters_count == 0) {
116 // We are using a preset. This is not a good idea in raw mode
117 // except when playing around with things. Different versions
118 // of this software may use different options in presets, and
119 // thus make uncompressing the raw data difficult.
120 if (opt_format == FORMAT_RAW) {
121 // The message is shown only if warnings are allowed
122 // but the exit status isn't changed.
123 message(V_WARNING, _("Using a preset in raw mode "
125 message(V_WARNING, _("The exact options of the "
126 "presets may vary between software "
130 // Get the preset for LZMA1 or LZMA2.
132 preset_number |= LZMA_PRESET_EXTREME;
134 if (lzma_lzma_preset(&opt_lzma, preset_number))
137 // Use LZMA2 except with --format=lzma we use LZMA1.
138 filters[0].id = opt_format == FORMAT_LZMA
139 ? LZMA_FILTER_LZMA1 : LZMA_FILTER_LZMA2;
140 filters[0].options = &opt_lzma;
143 preset_default = false;
146 // Terminate the filter options array.
147 filters[filters_count].id = LZMA_VLI_UNKNOWN;
149 // If we are using the LZMA_Alone format, allow exactly one filter
150 // which has to be LZMA.
151 if (opt_format == FORMAT_LZMA && (filters_count != 1
152 || filters[0].id != LZMA_FILTER_LZMA1))
153 message_fatal(_("With --format=lzma only the LZMA1 filter "
156 // Print the selected filter chain.
157 message_filters(V_DEBUG, filters);
159 // If using --format=raw, we can be decoding. The memusage function
160 // also validates the filter chain and the options used for the
162 uint64_t memory_usage;
163 uint64_t memory_limit;
164 if (opt_mode == MODE_COMPRESS) {
165 memory_usage = lzma_raw_encoder_memusage(filters);
166 memory_limit = hardware_memlimit_encoder();
168 memory_usage = lzma_raw_decoder_memusage(filters);
169 memory_limit = hardware_memlimit_decoder();
172 if (memory_usage == UINT64_MAX)
173 message_fatal("Unsupported filter chain or filter options");
175 // Print memory usage info.
176 message(V_DEBUG, _("%'" PRIu64 " MiB (%'" PRIu64 " B) of memory is "
177 "required per thread, "
178 "limit is %'" PRIu64 " MiB (%'" PRIu64 " B)"),
179 memory_usage >> 20, memory_usage,
180 memory_limit >> 20, memory_limit);
182 if (memory_usage > memory_limit) {
183 // If --no-auto-adjust was used or we didn't find LZMA1 or
184 // LZMA2 as the last filter, give an error immediatelly.
185 // --format=raw implies --no-auto-adjust.
186 if (!auto_adjust || opt_format == FORMAT_RAW)
187 memlimit_too_small(memory_usage, memory_limit);
189 assert(opt_mode == MODE_COMPRESS);
191 // Look for the last filter if it is LZMA2 or LZMA1, so
192 // we can make it use less RAM. With other filters we don't
195 while (filters[i].id != LZMA_FILTER_LZMA2
196 && filters[i].id != LZMA_FILTER_LZMA1) {
197 if (filters[i].id == LZMA_VLI_UNKNOWN)
198 memlimit_too_small(memory_usage, memory_limit);
203 // Decrease the dictionary size until we meet the memory
204 // usage limit. First round down to full mebibytes.
205 lzma_options_lzma *opt = filters[i].options;
206 const uint32_t orig_dict_size = opt->dict_size;
207 opt->dict_size &= ~((UINT32_C(1) << 20) - 1);
209 // If it is below 1 MiB, auto-adjusting failed. We
210 // could be more sophisticated and scale it down even
211 // more, but let's see if many complain about this
214 // FIXME: Displays the scaled memory usage instead
216 if (opt->dict_size < (UINT32_C(1) << 20))
217 memlimit_too_small(memory_usage, memory_limit);
219 memory_usage = lzma_raw_encoder_memusage(filters);
220 if (memory_usage == UINT64_MAX)
223 // Accept it if it is low enough.
224 if (memory_usage <= memory_limit)
227 // Otherwise 1 MiB down and try again. I hope this
228 // isn't too slow method for cases where the original
229 // dict_size is very big.
230 opt->dict_size -= UINT32_C(1) << 20;
233 // Tell the user that we decreased the dictionary size.
234 // However, omit the message if no preset or custom chain
235 // was given. FIXME: Always warn?
237 message(V_WARNING, "Adjusted LZMA%c dictionary size "
238 "from %'" PRIu32 " MiB to "
239 "%'" PRIu32 " MiB to not exceed "
240 "the memory usage limit of "
242 filters[i].id == LZMA_FILTER_LZMA2
244 orig_dict_size >> 20,
245 opt->dict_size >> 20,
250 // Limit the number of worker threads so that memory usage
251 // limit isn't exceeded.
252 assert(memory_usage > 0);
253 size_t thread_limit = memory_limit / memory_usage;
254 if (thread_limit == 0)
257 if (opt_threads > thread_limit)
258 opt_threads = thread_limit;
268 lzma_ret ret = LZMA_PROG_ERROR;
270 if (opt_mode == MODE_COMPRESS) {
271 switch (opt_format) {
273 // args.c ensures this.
278 ret = lzma_stream_encoder(&strm, filters, check);
282 ret = lzma_alone_encoder(&strm, filters[0].options);
286 ret = lzma_raw_encoder(&strm, filters);
290 const uint32_t flags = LZMA_TELL_UNSUPPORTED_CHECK
293 switch (opt_format) {
295 ret = lzma_auto_decoder(&strm,
296 hardware_memlimit_decoder(), flags);
300 ret = lzma_stream_decoder(&strm,
301 hardware_memlimit_decoder(), flags);
305 ret = lzma_alone_decoder(&strm,
306 hardware_memlimit_decoder());
310 // Memory usage has already been checked in
311 // coder_set_compression_settings().
312 ret = lzma_raw_decoder(&strm, filters);
317 if (ret != LZMA_OK) {
318 if (ret == LZMA_MEM_ERROR)
319 message_error("%s", message_strm(LZMA_MEM_ERROR));
331 coder_run(file_pair *pair)
333 // Buffers to hold input and output data.
334 uint8_t in_buf[IO_BUFFER_SIZE];
335 uint8_t out_buf[IO_BUFFER_SIZE];
337 // Initialize the progress indicator.
338 const uint64_t in_size = pair->src_st.st_size <= (off_t)(0)
339 ? 0 : (uint64_t)(pair->src_st.st_size);
340 message_progress_start(pair->src_name, in_size);
342 lzma_action action = LZMA_RUN;
346 strm.next_out = out_buf;
347 strm.avail_out = IO_BUFFER_SIZE;
349 while (!user_abort) {
350 // Fill the input buffer if it is empty and we haven't reached
352 if (strm.avail_in == 0 && !pair->src_eof) {
353 strm.next_in = in_buf;
354 strm.avail_in = io_read(pair, in_buf, IO_BUFFER_SIZE);
356 if (strm.avail_in == SIZE_MAX)
359 // Encoder needs to know when we have given all the
360 // input to it. The decoders need to know it too when
361 // we are using LZMA_CONCATENATED.
363 action = LZMA_FINISH;
366 // Let liblzma do the actual work.
367 ret = lzma_code(&strm, action);
369 // Write out if the output buffer became full.
370 if (strm.avail_out == 0) {
371 if (opt_mode != MODE_TEST && io_write(pair, out_buf,
372 IO_BUFFER_SIZE - strm.avail_out))
375 strm.next_out = out_buf;
376 strm.avail_out = IO_BUFFER_SIZE;
379 if (ret != LZMA_OK) {
380 // Determine if the return value indicates that we
381 // won't continue coding.
382 const bool stop = ret != LZMA_NO_CHECK
383 && ret != LZMA_UNSUPPORTED_CHECK;
386 // First print the final progress info.
387 // This way the user sees more accurately
388 // where the error occurred. Note that we
389 // print this *before* the possible error
392 // FIXME: What if something goes wrong
394 message_progress_end(strm.total_in,
396 ret == LZMA_STREAM_END);
398 // Write the remaining bytes even if something
399 // went wrong, because that way the user gets
400 // as much data as possible, which can be good
401 // when trying to get at least some useful
402 // data out of damaged files.
403 if (opt_mode != MODE_TEST && io_write(pair,
404 out_buf, IO_BUFFER_SIZE
409 if (ret == LZMA_STREAM_END) {
410 // Check that there is no trailing garbage.
411 // This is needed for LZMA_Alone and raw
413 if (strm.avail_in == 0 && (pair->src_eof
414 || io_read(pair, in_buf, 1)
416 assert(pair->src_eof);
420 // FIXME: What about io_read() failing?
422 // We hadn't reached the end of the file.
423 ret = LZMA_DATA_ERROR;
427 // If we get here and stop is true, something went
428 // wrong and we print an error. Otherwise it's just
429 // a warning and coding can continue.
431 message_error("%s: %s", pair->src_name,
434 message_warning("%s: %s", pair->src_name,
437 // When compressing, all possible errors set
439 assert(opt_mode != MODE_COMPRESS);
442 if (ret == LZMA_MEMLIMIT_ERROR) {
443 // Figure out how much memory it would have
445 uint64_t memusage = lzma_memusage(&strm);
447 = hardware_memlimit_decoder();
449 // Round the memory limit down and usage up.
450 // This way we don't display a ridiculous
451 // message like "Limit was 9 MiB, but 9 MiB
452 // would have been needed".
453 memusage = (memusage + 1024 * 1024 - 1)
455 memlimit /= 1024 * 1024;
457 message_error(_("Limit was %'" PRIu64 " MiB, "
458 "but %'" PRIu64 " MiB would "
467 // Show progress information if --verbose was specified and
468 // stderr is a terminal.
469 message_progress_update(strm.total_in, strm.total_out);
477 process_file(const char *filename)
479 // First try initializing the coder. If it fails, it's useless to try
480 // opening the file. Check also for user_abort just in case if we had
481 // got a signal while initializing the coder.
482 if (coder_init() || user_abort)
485 // Try to open the input and output files.
486 file_pair *pair = io_open(filename);
490 // Do the actual coding.
491 const bool success = coder_run(pair);
493 // Close the file pair. It needs to know if coding was successful to
494 // know if the source or target file should be unlinked.
495 io_close(pair, success);