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 static lzma_check check = LZMA_CHECK_CRC64;
58 coder_set_check(lzma_check new_check)
66 coder_set_preset(size_t new_preset)
68 preset_number = new_preset;
69 preset_default = false;
75 coder_set_extreme(void)
77 preset_extreme = true;
83 coder_add_filter(lzma_vli id, void *options)
85 if (filters_count == LZMA_FILTERS_MAX)
86 message_fatal(_("Maximum number of filters is four"));
88 filters[filters_count].id = id;
89 filters[filters_count].options = options;
96 static void lzma_attribute((noreturn))
97 memlimit_too_small(uint64_t memory_usage, uint64_t memory_limit)
99 message_fatal(_("Memory usage limit (%" PRIu64 " MiB) is too small "
100 "for the given filter setup (%" PRIu64 " MiB)"),
101 memory_limit >> 20, memory_usage >> 20);
106 coder_set_compression_settings(void)
108 // Options for LZMA1 or LZMA2 in case we are using a preset.
109 static lzma_options_lzma opt_lzma;
111 if (filters_count == 0) {
112 // We are using a preset. This is not a good idea in raw mode
113 // except when playing around with things. Different versions
114 // of this software may use different options in presets, and
115 // thus make uncompressing the raw data difficult.
116 if (opt_format == FORMAT_RAW) {
117 // The message is shown only if warnings are allowed
118 // but the exit status isn't changed.
119 message(V_WARNING, _("Using a preset in raw mode "
121 message(V_WARNING, _("The exact options of the "
122 "presets may vary between software "
126 // Get the preset for LZMA1 or LZMA2.
128 preset_number |= LZMA_PRESET_EXTREME;
130 if (lzma_lzma_preset(&opt_lzma, preset_number))
133 // Use LZMA2 except with --format=lzma we use LZMA1.
134 filters[0].id = opt_format == FORMAT_LZMA
135 ? LZMA_FILTER_LZMA1 : LZMA_FILTER_LZMA2;
136 filters[0].options = &opt_lzma;
139 preset_default = false;
142 // Terminate the filter options array.
143 filters[filters_count].id = LZMA_VLI_UNKNOWN;
145 // If we are using the LZMA_Alone format, allow exactly one filter
146 // which has to be LZMA.
147 if (opt_format == FORMAT_LZMA && (filters_count != 1
148 || filters[0].id != LZMA_FILTER_LZMA1))
149 message_fatal(_("With --format=lzma only the LZMA1 filter "
152 // Print the selected filter chain.
153 message_filters(V_DEBUG, filters);
155 // If using --format=raw, we can be decoding. The memusage function
156 // also validates the filter chain and the options used for the
158 uint64_t memory_usage;
159 uint64_t memory_limit;
160 if (opt_mode == MODE_COMPRESS) {
161 memory_usage = lzma_memusage_encoder(filters);
162 memory_limit = hardware_memlimit_encoder();
164 memory_usage = lzma_memusage_decoder(filters);
165 memory_limit = hardware_memlimit_decoder();
168 if (memory_usage == UINT64_MAX)
169 message_fatal("Unsupported filter chain or filter options");
171 // Print memory usage info.
172 message(V_DEBUG, _("%'" PRIu64 " MiB (%'" PRIu64 " B) of memory is "
173 "required per thread, "
174 "limit is %'" PRIu64 " MiB (%'" PRIu64 " B)"),
175 memory_usage >> 20, memory_usage,
176 memory_limit >> 20, memory_limit);
178 if (memory_usage > memory_limit) {
179 // If --no-auto-adjust was used or we didn't find LZMA1 or
180 // LZMA2 as the last filter, give an error immediatelly.
181 // --format=raw implies --no-auto-adjust.
182 if (!auto_adjust || opt_format == FORMAT_RAW)
183 memlimit_too_small(memory_usage, memory_limit);
185 assert(opt_mode == MODE_COMPRESS);
187 // Look for the last filter if it is LZMA2 or LZMA1, so
188 // we can make it use less RAM. With other filters we don't
191 while (filters[i].id != LZMA_FILTER_LZMA2
192 && filters[i].id != LZMA_FILTER_LZMA1) {
193 if (filters[i].id == LZMA_VLI_UNKNOWN)
194 memlimit_too_small(memory_usage, memory_limit);
199 // Decrease the dictionary size until we meet the memory
200 // usage limit. First round down to full mebibytes.
201 lzma_options_lzma *opt = filters[i].options;
202 const uint32_t orig_dict_size = opt->dict_size;
203 opt->dict_size &= ~((UINT32_C(1) << 20) - 1);
205 // If it is below 1 MiB, auto-adjusting failed. We
206 // could be more sophisticated and scale it down even
207 // more, but let's see if many complain about this
210 // FIXME: Displays the scaled memory usage instead
212 if (opt->dict_size < (UINT32_C(1) << 20))
213 memlimit_too_small(memory_usage, memory_limit);
215 memory_usage = lzma_memusage_encoder(filters);
216 if (memory_usage == UINT64_MAX)
219 // Accept it if it is low enough.
220 if (memory_usage <= memory_limit)
223 // Otherwise 1 MiB down and try again. I hope this
224 // isn't too slow method for cases where the original
225 // dict_size is very big.
226 opt->dict_size -= UINT32_C(1) << 20;
229 // Tell the user that we decreased the dictionary size.
230 // However, omit the message if no preset or custom chain
231 // was given. FIXME: Always warn?
233 message(V_WARNING, "Adjusted LZMA%c dictionary size "
234 "from %'" PRIu32 " MiB to "
235 "%'" PRIu32 " MiB to not exceed "
236 "the memory usage limit of "
238 filters[i].id == LZMA_FILTER_LZMA2
240 orig_dict_size >> 20,
241 opt->dict_size >> 20,
245 // Limit the number of worker threads so that memory usage
246 // limit isn't exceeded.
247 assert(memory_usage > 0);
248 size_t thread_limit = memory_limit / memory_usage;
249 if (thread_limit == 0)
252 if (opt_threads > thread_limit)
253 opt_threads = thread_limit;
262 lzma_ret ret = LZMA_PROG_ERROR;
264 if (opt_mode == MODE_COMPRESS) {
265 switch (opt_format) {
267 // args.c ensures this.
272 ret = lzma_stream_encoder(&strm, filters, check);
276 ret = lzma_alone_encoder(&strm, filters[0].options);
280 ret = lzma_raw_encoder(&strm, filters);
284 const uint32_t flags = LZMA_TELL_UNSUPPORTED_CHECK
287 switch (opt_format) {
289 ret = lzma_auto_decoder(&strm,
290 hardware_memlimit_decoder(), flags);
294 ret = lzma_stream_decoder(&strm,
295 hardware_memlimit_decoder(), flags);
299 ret = lzma_alone_decoder(&strm,
300 hardware_memlimit_decoder());
304 // Memory usage has already been checked in
305 // coder_set_compression_settings().
306 ret = lzma_raw_decoder(&strm, filters);
311 if (ret != LZMA_OK) {
312 if (ret == LZMA_MEM_ERROR)
313 message_error("%s", message_strm(LZMA_MEM_ERROR));
325 coder_run(file_pair *pair)
327 // Buffers to hold input and output data.
328 uint8_t in_buf[IO_BUFFER_SIZE];
329 uint8_t out_buf[IO_BUFFER_SIZE];
331 // Initialize the progress indicator.
332 const uint64_t in_size = pair->src_st.st_size <= (off_t)(0)
333 ? 0 : (uint64_t)(pair->src_st.st_size);
334 message_progress_start(pair->src_name, in_size);
336 lzma_action action = LZMA_RUN;
340 strm.next_out = out_buf;
341 strm.avail_out = IO_BUFFER_SIZE;
343 while (!user_abort) {
344 // Fill the input buffer if it is empty and we haven't reached
346 if (strm.avail_in == 0 && !pair->src_eof) {
347 strm.next_in = in_buf;
348 strm.avail_in = io_read(pair, in_buf, IO_BUFFER_SIZE);
350 if (strm.avail_in == SIZE_MAX)
353 // Encoder needs to know when we have given all the
354 // input to it. The decoders need to know it too when
355 // we are using LZMA_CONCATENATED.
357 action = LZMA_FINISH;
360 // Let liblzma do the actual work.
361 ret = lzma_code(&strm, action);
363 // Write out if the output buffer became full.
364 if (strm.avail_out == 0) {
365 if (opt_mode != MODE_TEST && io_write(pair, out_buf,
366 IO_BUFFER_SIZE - strm.avail_out))
369 strm.next_out = out_buf;
370 strm.avail_out = IO_BUFFER_SIZE;
373 if (ret != LZMA_OK) {
374 // Determine if the return value indicates that we
375 // won't continue coding.
376 const bool stop = ret != LZMA_NO_CHECK
377 && ret != LZMA_UNSUPPORTED_CHECK;
380 // First print the final progress info.
381 // This way the user sees more accurately
382 // where the error occurred. Note that we
383 // print this *before* the possible error
386 // FIXME: What if something goes wrong
388 message_progress_end(strm.total_in,
390 ret == LZMA_STREAM_END);
392 // Write the remaining bytes even if something
393 // went wrong, because that way the user gets
394 // as much data as possible, which can be good
395 // when trying to get at least some useful
396 // data out of damaged files.
397 if (opt_mode != MODE_TEST && io_write(pair,
398 out_buf, IO_BUFFER_SIZE
403 if (ret == LZMA_STREAM_END) {
404 // Check that there is no trailing garbage.
405 // This is needed for LZMA_Alone and raw
407 if (strm.avail_in == 0 && (pair->src_eof
408 || io_read(pair, in_buf, 1)
410 assert(pair->src_eof);
414 // FIXME: What about io_read() failing?
416 // We hadn't reached the end of the file.
417 ret = LZMA_DATA_ERROR;
421 // If we get here and stop is true, something went
422 // wrong and we print an error. Otherwise it's just
423 // a warning and coding can continue.
425 message_error("%s: %s", pair->src_name,
428 message_warning("%s: %s", pair->src_name,
431 // When compressing, all possible errors set
433 assert(opt_mode != MODE_COMPRESS);
436 if (ret == LZMA_MEMLIMIT_ERROR) {
437 // Figure out how much memory it would have
439 uint64_t memusage = lzma_memusage(&strm);
441 = hardware_memlimit_decoder();
443 // Round the memory limit down and usage up.
444 // This way we don't display a ridiculous
445 // message like "Limit was 9 MiB, but 9 MiB
446 // would have been needed".
447 memusage = (memusage + 1024 * 1024 - 1)
449 memlimit /= 1024 * 1024;
451 message_error(_("Limit was %'" PRIu64 " MiB, "
452 "but %'" PRIu64 " MiB would "
461 // Show progress information if --verbose was specified and
462 // stderr is a terminal.
463 message_progress_update(strm.total_in, strm.total_out);
471 process_file(const char *filename)
473 // First try initializing the coder. If it fails, it's useless to try
474 // opening the file. Check also for user_abort just in case if we had
475 // got a signal while initializing the coder.
476 if (coder_init() || user_abort)
479 // Try to open the input and output files.
480 file_pair *pair = io_open(filename);
484 // Do the actual coding.
485 const bool success = coder_run(pair);
487 // Close the file pair. It needs to know if coding was successful to
488 // know if the source or target file should be unlinked.
489 io_close(pair, success);