(root)/
xz-5.4.5/
src/
liblzma/
simple/
simple_coder.c
       1  ///////////////////////////////////////////////////////////////////////////////
       2  //
       3  /// \file       simple_coder.c
       4  /// \brief      Wrapper for simple filters
       5  ///
       6  /// Simple filters don't change the size of the data i.e. number of bytes
       7  /// in equals the number of bytes out.
       8  //
       9  //  Author:     Lasse Collin
      10  //
      11  //  This file has been put into the public domain.
      12  //  You can do whatever you want with this file.
      13  //
      14  ///////////////////////////////////////////////////////////////////////////////
      15  
      16  #include "simple_private.h"
      17  
      18  
      19  /// Copied or encodes/decodes more data to out[].
      20  static lzma_ret
      21  copy_or_code(lzma_simple_coder *coder, const lzma_allocator *allocator,
      22  		const uint8_t *restrict in, size_t *restrict in_pos,
      23  		size_t in_size, uint8_t *restrict out,
      24  		size_t *restrict out_pos, size_t out_size, lzma_action action)
      25  {
      26  	assert(!coder->end_was_reached);
      27  
      28  	if (coder->next.code == NULL) {
      29  		lzma_bufcpy(in, in_pos, in_size, out, out_pos, out_size);
      30  
      31  		// Check if end of stream was reached.
      32  		if (coder->is_encoder && action == LZMA_FINISH
      33  				&& *in_pos == in_size)
      34  			coder->end_was_reached = true;
      35  
      36  	} else {
      37  		// Call the next coder in the chain to provide us some data.
      38  		const lzma_ret ret = coder->next.code(
      39  				coder->next.coder, allocator,
      40  				in, in_pos, in_size,
      41  				out, out_pos, out_size, action);
      42  
      43  		if (ret == LZMA_STREAM_END) {
      44  			assert(!coder->is_encoder
      45  					|| action == LZMA_FINISH);
      46  			coder->end_was_reached = true;
      47  
      48  		} else if (ret != LZMA_OK) {
      49  			return ret;
      50  		}
      51  	}
      52  
      53  	return LZMA_OK;
      54  }
      55  
      56  
      57  static size_t
      58  call_filter(lzma_simple_coder *coder, uint8_t *buffer, size_t size)
      59  {
      60  	const size_t filtered = coder->filter(coder->simple,
      61  			coder->now_pos, coder->is_encoder,
      62  			buffer, size);
      63  	coder->now_pos += filtered;
      64  	return filtered;
      65  }
      66  
      67  
      68  static lzma_ret
      69  simple_code(void *coder_ptr, const lzma_allocator *allocator,
      70  		const uint8_t *restrict in, size_t *restrict in_pos,
      71  		size_t in_size, uint8_t *restrict out,
      72  		size_t *restrict out_pos, size_t out_size, lzma_action action)
      73  {
      74  	lzma_simple_coder *coder = coder_ptr;
      75  
      76  	// TODO: Add partial support for LZMA_SYNC_FLUSH. We can support it
      77  	// in cases when the filter is able to filter everything. With most
      78  	// simple filters it can be done at offset that is a multiple of 2,
      79  	// 4, or 16. With x86 filter, it needs good luck, and thus cannot
      80  	// be made to work predictably.
      81  	if (action == LZMA_SYNC_FLUSH)
      82  		return LZMA_OPTIONS_ERROR;
      83  
      84  	// Flush already filtered data from coder->buffer[] to out[].
      85  	if (coder->pos < coder->filtered) {
      86  		lzma_bufcpy(coder->buffer, &coder->pos, coder->filtered,
      87  				out, out_pos, out_size);
      88  
      89  		// If we couldn't flush all the filtered data, return to
      90  		// application immediately.
      91  		if (coder->pos < coder->filtered)
      92  			return LZMA_OK;
      93  
      94  		if (coder->end_was_reached) {
      95  			assert(coder->filtered == coder->size);
      96  			return LZMA_STREAM_END;
      97  		}
      98  	}
      99  
     100  	// If we get here, there is no filtered data left in the buffer.
     101  	coder->filtered = 0;
     102  
     103  	assert(!coder->end_was_reached);
     104  
     105  	// If there is more output space left than there is unfiltered data
     106  	// in coder->buffer[], flush coder->buffer[] to out[], and copy/code
     107  	// more data to out[] hopefully filling it completely. Then filter
     108  	// the data in out[]. This step is where most of the data gets
     109  	// filtered if the buffer sizes used by the application are reasonable.
     110  	const size_t out_avail = out_size - *out_pos;
     111  	const size_t buf_avail = coder->size - coder->pos;
     112  	if (out_avail > buf_avail || buf_avail == 0) {
     113  		// Store the old position so that we know from which byte
     114  		// to start filtering.
     115  		const size_t out_start = *out_pos;
     116  
     117  		// Flush data from coder->buffer[] to out[], but don't reset
     118  		// coder->pos and coder->size yet. This way the coder can be
     119  		// restarted if the next filter in the chain returns e.g.
     120  		// LZMA_MEM_ERROR.
     121  		//
     122  		// Do the memcpy() conditionally because out can be NULL
     123  		// (in which case buf_avail is always 0). Calling memcpy()
     124  		// with a null-pointer is undefined even if the third
     125  		// argument is 0.
     126  		if (buf_avail > 0)
     127  			memcpy(out + *out_pos, coder->buffer + coder->pos,
     128  					buf_avail);
     129  
     130  		*out_pos += buf_avail;
     131  
     132  		// Copy/Encode/Decode more data to out[].
     133  		{
     134  			const lzma_ret ret = copy_or_code(coder, allocator,
     135  					in, in_pos, in_size,
     136  					out, out_pos, out_size, action);
     137  			assert(ret != LZMA_STREAM_END);
     138  			if (ret != LZMA_OK)
     139  				return ret;
     140  		}
     141  
     142  		// Filter out[] unless there is nothing to filter.
     143  		// This way we avoid null pointer + 0 (undefined behavior)
     144  		// when out == NULL.
     145  		const size_t size = *out_pos - out_start;
     146  		const size_t filtered = size == 0 ? 0 : call_filter(
     147  				coder, out + out_start, size);
     148  
     149  		const size_t unfiltered = size - filtered;
     150  		assert(unfiltered <= coder->allocated / 2);
     151  
     152  		// Now we can update coder->pos and coder->size, because
     153  		// the next coder in the chain (if any) was successful.
     154  		coder->pos = 0;
     155  		coder->size = unfiltered;
     156  
     157  		if (coder->end_was_reached) {
     158  			// The last byte has been copied to out[] already.
     159  			// They are left as is.
     160  			coder->size = 0;
     161  
     162  		} else if (unfiltered > 0) {
     163  			// There is unfiltered data left in out[]. Copy it to
     164  			// coder->buffer[] and rewind *out_pos appropriately.
     165  			*out_pos -= unfiltered;
     166  			memcpy(coder->buffer, out + *out_pos, unfiltered);
     167  		}
     168  	} else if (coder->pos > 0) {
     169  		memmove(coder->buffer, coder->buffer + coder->pos, buf_avail);
     170  		coder->size -= coder->pos;
     171  		coder->pos = 0;
     172  	}
     173  
     174  	assert(coder->pos == 0);
     175  
     176  	// If coder->buffer[] isn't empty, try to fill it by copying/decoding
     177  	// more data. Then filter coder->buffer[] and copy the successfully
     178  	// filtered data to out[]. It is probable, that some filtered and
     179  	// unfiltered data will be left to coder->buffer[].
     180  	if (coder->size > 0) {
     181  		{
     182  			const lzma_ret ret = copy_or_code(coder, allocator,
     183  					in, in_pos, in_size,
     184  					coder->buffer, &coder->size,
     185  					coder->allocated, action);
     186  			assert(ret != LZMA_STREAM_END);
     187  			if (ret != LZMA_OK)
     188  				return ret;
     189  		}
     190  
     191  		coder->filtered = call_filter(
     192  				coder, coder->buffer, coder->size);
     193  
     194  		// Everything is considered to be filtered if coder->buffer[]
     195  		// contains the last bytes of the data.
     196  		if (coder->end_was_reached)
     197  			coder->filtered = coder->size;
     198  
     199  		// Flush as much as possible.
     200  		lzma_bufcpy(coder->buffer, &coder->pos, coder->filtered,
     201  				out, out_pos, out_size);
     202  	}
     203  
     204  	// Check if we got everything done.
     205  	if (coder->end_was_reached && coder->pos == coder->size)
     206  		return LZMA_STREAM_END;
     207  
     208  	return LZMA_OK;
     209  }
     210  
     211  
     212  static void
     213  simple_coder_end(void *coder_ptr, const lzma_allocator *allocator)
     214  {
     215  	lzma_simple_coder *coder = coder_ptr;
     216  	lzma_next_end(&coder->next, allocator);
     217  	lzma_free(coder->simple, allocator);
     218  	lzma_free(coder, allocator);
     219  	return;
     220  }
     221  
     222  
     223  static lzma_ret
     224  simple_coder_update(void *coder_ptr, const lzma_allocator *allocator,
     225  		const lzma_filter *filters_null lzma_attribute((__unused__)),
     226  		const lzma_filter *reversed_filters)
     227  {
     228  	lzma_simple_coder *coder = coder_ptr;
     229  
     230  	// No update support, just call the next filter in the chain.
     231  	return lzma_next_filter_update(
     232  			&coder->next, allocator, reversed_filters + 1);
     233  }
     234  
     235  
     236  extern lzma_ret
     237  lzma_simple_coder_init(lzma_next_coder *next, const lzma_allocator *allocator,
     238  		const lzma_filter_info *filters,
     239  		size_t (*filter)(void *simple, uint32_t now_pos,
     240  			bool is_encoder, uint8_t *buffer, size_t size),
     241  		size_t simple_size, size_t unfiltered_max,
     242  		uint32_t alignment, bool is_encoder)
     243  {
     244  	// Allocate memory for the lzma_simple_coder structure if needed.
     245  	lzma_simple_coder *coder = next->coder;
     246  	if (coder == NULL) {
     247  		// Here we allocate space also for the temporary buffer. We
     248  		// need twice the size of unfiltered_max, because then it
     249  		// is always possible to filter at least unfiltered_max bytes
     250  		// more data in coder->buffer[] if it can be filled completely.
     251  		coder = lzma_alloc(sizeof(lzma_simple_coder)
     252  				+ 2 * unfiltered_max, allocator);
     253  		if (coder == NULL)
     254  			return LZMA_MEM_ERROR;
     255  
     256  		next->coder = coder;
     257  		next->code = &simple_code;
     258  		next->end = &simple_coder_end;
     259  		next->update = &simple_coder_update;
     260  
     261  		coder->next = LZMA_NEXT_CODER_INIT;
     262  		coder->filter = filter;
     263  		coder->allocated = 2 * unfiltered_max;
     264  
     265  		// Allocate memory for filter-specific data structure.
     266  		if (simple_size > 0) {
     267  			coder->simple = lzma_alloc(simple_size, allocator);
     268  			if (coder->simple == NULL)
     269  				return LZMA_MEM_ERROR;
     270  		} else {
     271  			coder->simple = NULL;
     272  		}
     273  	}
     274  
     275  	if (filters[0].options != NULL) {
     276  		const lzma_options_bcj *simple = filters[0].options;
     277  		coder->now_pos = simple->start_offset;
     278  		if (coder->now_pos & (alignment - 1))
     279  			return LZMA_OPTIONS_ERROR;
     280  	} else {
     281  		coder->now_pos = 0;
     282  	}
     283  
     284  	// Reset variables.
     285  	coder->is_encoder = is_encoder;
     286  	coder->end_was_reached = false;
     287  	coder->pos = 0;
     288  	coder->filtered = 0;
     289  	coder->size = 0;
     290  
     291  	return lzma_next_filter_init(&coder->next, allocator, filters + 1);
     292  }