LZX, LZMS: Annotate unaligned memory accesses in x86 filtering

[wimlib] / src / lzx-common.c

diff --git a/src/lzx-common.c b/src/lzx-common.c
index 827b1b29657d436d0472fae98487b6d2d7f50f5c..9f55f171f220689bb61dc9e50006507a210ad66c 100644 (file)
--- a/src/lzx-common.c
+++ b/src/lzx-common.c
@@ -5,20 +5,18 @@
 /*
  * Copyright (C) 2012, 2013 Eric Biggers
  *
- * This file is part of wimlib, a library for working with WIM files.
+ * 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
+ * Software Foundation; either version 3 of the License, or (at your option) any
+ * later version.
  *
- * wimlib is free software; you can redistribute it and/or modify it under the
- * terms of the GNU General Public License as published by the Free
- * Software Foundation; either version 3 of the License, or (at your option)
- * any later version.
- *
- * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
- * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
- * A PARTICULAR PURPOSE. See the GNU General Public License for more
+ * This file is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
  * details.
  *
- * You should have received a copy of the GNU General Public License
- * along with wimlib; if not, see http://www.gnu.org/licenses/.
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this file; if not, see http://www.gnu.org/licenses/.
  */
 
 #ifdef HAVE_CONFIG_H
@@ -27,6 +25,7 @@
 
 #include "wimlib/endianness.h"
 #include "wimlib/lzx.h"
+#include "wimlib/unaligned.h"
 #include "wimlib/util.h"
 
 #ifdef __SSE2__
@@ -73,12 +72,12 @@ lzx_get_window_order(size_t max_block_size)
 {
        unsigned order;
 
-       if (max_block_size == 0 || max_block_size > (1 << LZX_MAX_WINDOW_ORDER))
+       if (max_block_size == 0 || max_block_size > LZX_MAX_WINDOW_SIZE)
                return 0;
 
        order = bsr32(max_block_size);
 
-       if ((1 << order) != max_block_size)
+       if (((u32)1 << order) != max_block_size)
                order++;
 
        return max(order, LZX_MIN_WINDOW_ORDER);
@@ -89,7 +88,7 @@ lzx_get_window_order(size_t max_block_size)
 unsigned
 lzx_get_num_main_syms(unsigned window_order)
 {
-       u32 window_size = 1 << window_order;
+       u32 window_size = (u32)1 << window_order;
 
        /* NOTE: the calculation *should* be as follows:
         *
@@ -122,12 +121,11 @@ lzx_get_num_main_syms(unsigned window_order)
 }
 
 static void
-do_translate_target(sle32 *target, s32 input_pos)
+do_translate_target(void *target, s32 input_pos)
 {
        s32 abs_offset, rel_offset;
 
-       /* XXX: This assumes unaligned memory accesses are okay.  */
-       rel_offset = le32_to_cpu(*target);
+       rel_offset = le32_to_cpu(load_le32_unaligned(target));
        if (rel_offset >= -input_pos && rel_offset < LZX_WIM_MAGIC_FILESIZE) {
                if (rel_offset < LZX_WIM_MAGIC_FILESIZE - input_pos) {
                        /* "good translation" */
@@ -136,30 +134,29 @@ do_translate_target(sle32 *target, s32 input_pos)
                        /* "compensating translation" */
                        abs_offset = rel_offset - LZX_WIM_MAGIC_FILESIZE;
                }
-               *target = cpu_to_le32(abs_offset);
+               store_le32_unaligned(cpu_to_le32(abs_offset), target);
        }
 }
 
 static void
-undo_translate_target(sle32 *target, s32 input_pos)
+undo_translate_target(void *target, s32 input_pos)
 {
        s32 abs_offset, rel_offset;
 
-       /* XXX: This assumes unaligned memory accesses are okay.  */
-       abs_offset = le32_to_cpu(*target);
+       abs_offset = le32_to_cpu(load_le32_unaligned(target));
        if (abs_offset >= 0) {
                if (abs_offset < LZX_WIM_MAGIC_FILESIZE) {
                        /* "good translation" */
                        rel_offset = abs_offset - input_pos;
 
-                       *target = cpu_to_le32(rel_offset);
+                       store_le32_unaligned(cpu_to_le32(rel_offset), target);
                }
        } else {
                if (abs_offset >= -input_pos) {
                        /* "compensating translation" */
                        rel_offset = abs_offset + LZX_WIM_MAGIC_FILESIZE;
 
-                       *target = cpu_to_le32(rel_offset);
+                       store_le32_unaligned(cpu_to_le32(rel_offset), target);
                }
        }
 }
@@ -196,7 +193,7 @@ inline  /* Although inlining the 'process_target' function still speeds up the
           SSE2 case, it bloats the binary more.  */
 #endif
 void
-lzx_e8_filter(u8 *data, u32 size, void (*process_target)(sle32 *, s32))
+lzx_e8_filter(u8 *data, u32 size, void (*process_target)(void *, s32))
 {
 #ifdef __SSE2__
        /* SSE2 vectorized implementation for x86_64.  This speeds up LZX
@@ -251,7 +248,7 @@ lzx_e8_filter(u8 *data, u32 size, void (*process_target)(sle32 *, s32))
 
                                        /* Do (or undo) the e8 translation.  */
                                        u8 *p8 = (u8 *)p128 + bit;
-                                       (*process_target)((sle32 *)(p8 + 1),
+                                       (*process_target)(p8 + 1,
                                                          p8 - data);
 
                                        /* Don't start an e8 translation in the
@@ -281,7 +278,7 @@ lzx_e8_filter(u8 *data, u32 size, void (*process_target)(sle32 *, s32))
                u8 *p8_end = data + size - 10;
                do {
                        if (*p8 == 0xe8) {
-                               (*process_target)((sle32 *)(p8 + 1), p8 - data);
+                               (*process_target)(p8 + 1, p8 - data);
                                p8 += 5;
                        } else {
                                p8++;