1/*
2 * Copyright (c) 1992, 1993, 1994, 1995, 1996
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that: (1) source code distributions
7 * retain the above copyright notice and this paragraph in its entirety, (2)
8 * distributions including binary code include the above copyright notice and
9 * this paragraph in its entirety in the documentation or other materials
10 * provided with the distribution, and (3) all advertising materials mentioning
11 * features or use of this software display the following acknowledgement:
12 * ``This product includes software developed by the University of California,
13 * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
14 * the University nor the names of its contributors may be used to endorse
15 * or promote products derived from this software without specific prior
16 * written permission.
17 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
18 * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
19 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
20 *
21 * @(#) $Header: /tcpdump/master/tcpdump/extract.h,v 1.25 2006-01-30 16:20:07 hannes Exp $ (LBL)
22 */
23
24/*
25 * Macros to extract possibly-unaligned big-endian integral values.
26 */
27#ifdef LBL_ALIGN
28/*
29 * The processor doesn't natively handle unaligned loads.
30 */
31#if defined(__GNUC__) && defined(HAVE___ATTRIBUTE__) && \
32    (defined(__alpha) || defined(__alpha__) || \
33     defined(__mips) || defined(__mips__))
34
35/*
36 * This is a GCC-compatible compiler and we have __attribute__, which
37 * we assume that mean we have __attribute__((packed)), and this is
38 * MIPS or Alpha, which has instructions that can help when doing
39 * unaligned loads.
40 *
41 * Declare packed structures containing a u_int16_t and a u_int32_t,
42 * cast the pointer to point to one of those, and fetch through it;
43 * the GCC manual doesn't appear to explicitly say that
44 * __attribute__((packed)) causes the compiler to generate unaligned-safe
45 * code, but it apppears to do so.
46 *
47 * We do this in case the compiler can generate code using those
48 * instructions to do an unaligned load and pass stuff to "ntohs()" or
49 * "ntohl()", which might be better than than the code to fetch the
50 * bytes one at a time and assemble them.  (That might not be the
51 * case on a little-endian platform, such as DEC's MIPS machines and
52 * Alpha machines, where "ntohs()" and "ntohl()" might not be done
53 * inline.)
54 *
55 * We do this only for specific architectures because, for example,
56 * at least some versions of GCC, when compiling for 64-bit SPARC,
57 * generate code that assumes alignment if we do this.
58 *
59 * XXX - add other architectures and compilers as possible and
60 * appropriate.
61 *
62 * HP's C compiler, indicated by __HP_cc being defined, supports
63 * "#pragma unaligned N" in version A.05.50 and later, where "N"
64 * specifies a number of bytes at which the typedef on the next
65 * line is aligned, e.g.
66 *
67 *	#pragma unalign 1
68 *	typedef u_int16_t unaligned_u_int16_t;
69 *
70 * to define unaligned_u_int16_t as a 16-bit unaligned data type.
71 * This could be presumably used, in sufficiently recent versions of
72 * the compiler, with macros similar to those below.  This would be
73 * useful only if that compiler could generate better code for PA-RISC
74 * or Itanium than would be generated by a bunch of shifts-and-ORs.
75 *
76 * DEC C, indicated by __DECC being defined, has, at least on Alpha,
77 * an __unaligned qualifier that can be applied to pointers to get the
78 * compiler to generate code that does unaligned loads and stores when
79 * dereferencing the pointer in question.
80 *
81 * XXX - what if the native C compiler doesn't support
82 * __attribute__((packed))?  How can we get it to generate unaligned
83 * accesses for *specific* items?
84 */
85typedef struct {
86	u_int16_t	val;
87} __attribute__((packed)) unaligned_u_int16_t;
88
89typedef struct {
90	u_int32_t	val;
91} __attribute__((packed)) unaligned_u_int32_t;
92
93static inline u_int16_t
94EXTRACT_16BITS(const void *p)
95{
96	return ((u_int16_t)ntohs(((const unaligned_u_int16_t *)(p))->val));
97}
98
99static inline u_int32_t
100EXTRACT_32BITS(const void *p)
101{
102	return ((u_int32_t)ntohl(((const unaligned_u_int32_t *)(p))->val));
103}
104
105static inline u_int64_t
106EXTRACT_64BITS(const void *p)
107{
108	return ((u_int64_t)(((u_int64_t)ntohl(((const unaligned_u_int32_t *)(p) + 0)->val)) << 32 | \
109		((u_int64_t)ntohl(((const unaligned_u_int32_t *)(p) + 1)->val)) << 0));
110}
111
112#else /* have to do it a byte at a time */
113/*
114 * This isn't a GCC-compatible compiler, we don't have __attribute__,
115 * or we do but we don't know of any better way with this instruction
116 * set to do unaligned loads, so do unaligned loads of big-endian
117 * quantities the hard way - fetch the bytes one at a time and
118 * assemble them.
119 */
120#define EXTRACT_16BITS(p) \
121	((u_int16_t)((u_int16_t)*((const u_int8_t *)(p) + 0) << 8 | \
122		     (u_int16_t)*((const u_int8_t *)(p) + 1)))
123#define EXTRACT_32BITS(p) \
124	((u_int32_t)((u_int32_t)*((const u_int8_t *)(p) + 0) << 24 | \
125		     (u_int32_t)*((const u_int8_t *)(p) + 1) << 16 | \
126		     (u_int32_t)*((const u_int8_t *)(p) + 2) << 8 | \
127		     (u_int32_t)*((const u_int8_t *)(p) + 3)))
128#define EXTRACT_64BITS(p) \
129	((u_int64_t)((u_int64_t)*((const u_int8_t *)(p) + 0) << 56 | \
130		     (u_int64_t)*((const u_int8_t *)(p) + 1) << 48 | \
131		     (u_int64_t)*((const u_int8_t *)(p) + 2) << 40 | \
132		     (u_int64_t)*((const u_int8_t *)(p) + 3) << 32 | \
133	             (u_int64_t)*((const u_int8_t *)(p) + 4) << 24 | \
134		     (u_int64_t)*((const u_int8_t *)(p) + 5) << 16 | \
135		     (u_int64_t)*((const u_int8_t *)(p) + 6) << 8 | \
136		     (u_int64_t)*((const u_int8_t *)(p) + 7)))
137#endif /* must special-case unaligned accesses */
138#else /* LBL_ALIGN */
139/*
140 * The processor natively handles unaligned loads, so we can just
141 * cast the pointer and fetch through it.
142 */
143static inline u_int16_t
144EXTRACT_16BITS(const void *p)
145{
146	return ((u_int16_t)ntohs(*(const u_int16_t *)(p)));
147}
148
149static inline u_int32_t
150EXTRACT_32BITS(const void *p)
151{
152	return ((u_int32_t)ntohl(*(const u_int32_t *)(p)));
153}
154
155static inline u_int64_t
156EXTRACT_64BITS(const void *p)
157{
158	return ((u_int64_t)(((u_int64_t)ntohl(*((const u_int32_t *)(p) + 0))) << 32 | \
159		((u_int64_t)ntohl(*((const u_int32_t *)(p) + 1))) << 0));
160
161}
162
163#endif /* LBL_ALIGN */
164
165#define EXTRACT_24BITS(p) \
166	((u_int32_t)((u_int32_t)*((const u_int8_t *)(p) + 0) << 16 | \
167		     (u_int32_t)*((const u_int8_t *)(p) + 1) << 8 | \
168		     (u_int32_t)*((const u_int8_t *)(p) + 2)))
169
170/*
171 * Macros to extract possibly-unaligned little-endian integral values.
172 * XXX - do loads on little-endian machines that support unaligned loads?
173 */
174#define EXTRACT_LE_8BITS(p) (*(p))
175#define EXTRACT_LE_16BITS(p) \
176	((u_int16_t)((u_int16_t)*((const u_int8_t *)(p) + 1) << 8 | \
177		     (u_int16_t)*((const u_int8_t *)(p) + 0)))
178#define EXTRACT_LE_32BITS(p) \
179	((u_int32_t)((u_int32_t)*((const u_int8_t *)(p) + 3) << 24 | \
180		     (u_int32_t)*((const u_int8_t *)(p) + 2) << 16 | \
181		     (u_int32_t)*((const u_int8_t *)(p) + 1) << 8 | \
182		     (u_int32_t)*((const u_int8_t *)(p) + 0)))
183#define EXTRACT_LE_24BITS(p) \
184	((u_int32_t)((u_int32_t)*((const u_int8_t *)(p) + 2) << 16 | \
185		     (u_int32_t)*((const u_int8_t *)(p) + 1) << 8 | \
186		     (u_int32_t)*((const u_int8_t *)(p) + 0)))
187#define EXTRACT_LE_64BITS(p) \
188	((u_int64_t)((u_int64_t)*((const u_int8_t *)(p) + 7) << 56 | \
189		     (u_int64_t)*((const u_int8_t *)(p) + 6) << 48 | \
190		     (u_int64_t)*((const u_int8_t *)(p) + 5) << 40 | \
191		     (u_int64_t)*((const u_int8_t *)(p) + 4) << 32 | \
192	             (u_int64_t)*((const u_int8_t *)(p) + 3) << 24 | \
193		     (u_int64_t)*((const u_int8_t *)(p) + 2) << 16 | \
194		     (u_int64_t)*((const u_int8_t *)(p) + 1) << 8 | \
195		     (u_int64_t)*((const u_int8_t *)(p) + 0)))
196