xpress_decompress: allocate decode table on heap instead of stack

[wimlib] / src / xpress_decompress.c

diff --git a/src/xpress_decompress.c b/src/xpress_decompress.c
index 51e781a33bf6bd0057419b5d26bc49e95a4c6766..623ea187fef32ececc0669afdb247ca51e162726 100644 (file)
--- a/src/xpress_decompress.c
+++ b/src/xpress_decompress.c
@@ -6,7 +6,7 @@
 
 /*
  *
- * Copyright (C) 2012, 2013, 2015 Eric Biggers
+ * Copyright (C) 2012-2016 Eric Biggers
  *
  * This file is free software; you can redistribute it and/or modify it under
  * the terms of the GNU Lesser General Public License as published by the Free
@@ -70,37 +70,43 @@
 #include "wimlib/decompressor_ops.h"
 #include "wimlib/decompress_common.h"
 #include "wimlib/error.h"
+#include "wimlib/util.h"
 #include "wimlib/xpress_constants.h"
 
 /* This value is chosen for fast decompression.  */
-#define XPRESS_TABLEBITS 12
+#define XPRESS_TABLEBITS 11
+
+struct xpress_decompressor {
+       union {
+               DECODE_TABLE(decode_table, XPRESS_NUM_SYMBOLS,
+                            XPRESS_TABLEBITS, XPRESS_MAX_CODEWORD_LEN);
+               u8 lens[XPRESS_NUM_SYMBOLS];
+       };
+} _aligned_attribute(DECODE_TABLE_ALIGNMENT);
 
 static int
-xpress_decompress(const void *compressed_data, size_t compressed_size,
-                 void *uncompressed_data, size_t uncompressed_size, void *_ctx)
+xpress_decompress(const void *restrict compressed_data, size_t compressed_size,
+                 void *restrict uncompressed_data, size_t uncompressed_size,
+                 void *restrict _d)
 {
+       struct xpress_decompressor *d  = _d;
        const u8 * const in_begin = compressed_data;
        u8 * const out_begin = uncompressed_data;
        u8 *out_next = out_begin;
        u8 * const out_end = out_begin + uncompressed_size;
-       union {
-               u16 decode_table[(1 << XPRESS_TABLEBITS) + 2 * XPRESS_NUM_SYMBOLS]
-                               _aligned_attribute(DECODE_TABLE_ALIGNMENT);
-               u8 lens[XPRESS_NUM_SYMBOLS];
-       } u;
        struct input_bitstream is;
 
        /* Read the Huffman codeword lengths.  */
        if (compressed_size < XPRESS_NUM_SYMBOLS / 2)
                return -1;
        for (int i = 0; i < XPRESS_NUM_SYMBOLS / 2; i++) {
-               u.lens[2 * i + 0] = in_begin[i] & 0xf;
-               u.lens[2 * i + 1] = in_begin[i] >> 4;
+               d->lens[2 * i + 0] = in_begin[i] & 0xf;
+               d->lens[2 * i + 1] = in_begin[i] >> 4;
        }
 
        /* Build a decoding table for the Huffman code.  */
-       if (make_huffman_decode_table(u.decode_table, XPRESS_NUM_SYMBOLS,
-                                     XPRESS_TABLEBITS, u.lens,
+       if (make_huffman_decode_table(d->decode_table, XPRESS_NUM_SYMBOLS,
+                                     XPRESS_TABLEBITS, d->lens,
                                      XPRESS_MAX_CODEWORD_LEN))
                return -1;
 
@@ -115,7 +121,7 @@ xpress_decompress(const void *compressed_data, size_t compressed_size,
                u32 length;
                u32 offset;
 
-               sym = read_huffsym(&is, u.decode_table,
+               sym = read_huffsym(&is, d->decode_table,
                                   XPRESS_TABLEBITS, XPRESS_MAX_CODEWORD_LEN);
                if (sym < XPRESS_NUM_CHARS) {
                        /* Literal  */
@@ -137,15 +143,11 @@ xpress_decompress(const void *compressed_data, size_t compressed_size,
                        }
                        length += XPRESS_MIN_MATCH_LEN;
 
-                       if (unlikely(offset > out_next - out_begin))
-                               return -1;
-
-                       if (unlikely(length > out_end - out_next))
+                       if (unlikely(lz_copy(length, offset,
+                                            out_begin, out_next, out_end,
+                                            XPRESS_MIN_MATCH_LEN)))
                                return -1;
 
-                       lz_copy(out_next, length, offset, out_end,
-                               XPRESS_MIN_MATCH_LEN);
-
                        out_next += length;
                }
        }
@@ -153,15 +155,29 @@ xpress_decompress(const void *compressed_data, size_t compressed_size,
 }
 
 static int
-xpress_create_decompressor(size_t max_block_size, void **dec_ret)
+xpress_create_decompressor(size_t max_block_size, void **d_ret)
 {
+       struct xpress_decompressor *d;
+
        if (max_block_size > XPRESS_MAX_OFFSET + 1)
                return WIMLIB_ERR_INVALID_PARAM;
 
+       d = ALIGNED_MALLOC(sizeof(*d), DECODE_TABLE_ALIGNMENT);
+       if (!d)
+               return WIMLIB_ERR_NOMEM;
+
+       *d_ret = d;
        return 0;
 }
 
+static void
+xpress_free_decompressor(void *_d)
+{
+       ALIGNED_FREE(_d);
+}
+
 const struct decompressor_ops xpress_decompressor_ops = {
        .create_decompressor = xpress_create_decompressor,
        .decompress          = xpress_decompress,
+       .free_decompressor   = xpress_free_decompressor,
 };