1335640Shselasky/*
2335640Shselasky * Copyright (c) 1992, 1993, 1994, 1995, 1996
3335640Shselasky *	The Regents of the University of California.  All rights reserved.
4335640Shselasky *
5335640Shselasky * Redistribution and use in source and binary forms, with or without
6335640Shselasky * modification, are permitted provided that: (1) source code distributions
7335640Shselasky * retain the above copyright notice and this paragraph in its entirety, (2)
8335640Shselasky * distributions including binary code include the above copyright notice and
9335640Shselasky * this paragraph in its entirety in the documentation or other materials
10335640Shselasky * provided with the distribution, and (3) all advertising materials mentioning
11335640Shselasky * features or use of this software display the following acknowledgement:
12335640Shselasky * ``This product includes software developed by the University of California,
13335640Shselasky * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
14335640Shselasky * the University nor the names of its contributors may be used to endorse
15335640Shselasky * or promote products derived from this software without specific prior
16335640Shselasky * written permission.
17335640Shselasky * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
18335640Shselasky * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
19335640Shselasky * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
20335640Shselasky */
21335640Shselasky
22335640Shselasky#ifndef _WIN32
23335640Shselasky#include <arpa/inet.h>
24335640Shselasky#endif
25335640Shselasky
26335640Shselasky#include <pcap/pcap-inttypes.h>
27335640Shselasky#include <pcap/compiler-tests.h>
28335640Shselasky
29335640Shselasky/*
30335640Shselasky * Macros to extract possibly-unaligned big-endian integral values.
31335640Shselasky */
32335640Shselasky#ifdef LBL_ALIGN
33335640Shselasky/*
34335640Shselasky * The processor doesn't natively handle unaligned loads.
35335640Shselasky */
36335640Shselasky#if PCAP_IS_AT_LEAST_GNUC_VERSION(2,0) && \
37335640Shselasky    (defined(__alpha) || defined(__alpha__) || \
38335640Shselasky     defined(__mips) || defined(__mips__))
39335640Shselasky/*
40335640Shselasky * This is MIPS or Alpha, which don't natively handle unaligned loads,
41335640Shselasky * but which have instructions that can help when doing unaligned
42335640Shselasky * loads, and this is GCC 2.0 or later or a compiler that claims to
43335640Shselasky * be GCC 2.0 or later, which we assume that mean we have
44335640Shselasky * __attribute__((packed)), which we can use to convince the compiler
45335640Shselasky * to generate those instructions.
46335640Shselasky *
47335640Shselasky * Declare packed structures containing a uint16_t and a uint32_t,
48335640Shselasky * cast the pointer to point to one of those, and fetch through it;
49335640Shselasky * the GCC manual doesn't appear to explicitly say that
50335640Shselasky * __attribute__((packed)) causes the compiler to generate unaligned-safe
51335640Shselasky * code, but it apppears to do so.
52335640Shselasky *
53335640Shselasky * We do this in case the compiler can generate code using those
54335640Shselasky * instructions to do an unaligned load and pass stuff to "ntohs()" or
55335640Shselasky * "ntohl()", which might be better than than the code to fetch the
56335640Shselasky * bytes one at a time and assemble them.  (That might not be the
57335640Shselasky * case on a little-endian platform, such as DEC's MIPS machines and
58335640Shselasky * Alpha machines, where "ntohs()" and "ntohl()" might not be done
59335640Shselasky * inline.)
60335640Shselasky *
61335640Shselasky * We do this only for specific architectures because, for example,
62335640Shselasky * at least some versions of GCC, when compiling for 64-bit SPARC,
63335640Shselasky * generate code that assumes alignment if we do this.
64335640Shselasky *
65335640Shselasky * XXX - add other architectures and compilers as possible and
66335640Shselasky * appropriate.
67335640Shselasky *
68335640Shselasky * HP's C compiler, indicated by __HP_cc being defined, supports
69335640Shselasky * "#pragma unaligned N" in version A.05.50 and later, where "N"
70335640Shselasky * specifies a number of bytes at which the typedef on the next
71335640Shselasky * line is aligned, e.g.
72335640Shselasky *
73335640Shselasky *	#pragma unalign 1
74335640Shselasky *	typedef uint16_t unaligned_uint16_t;
75335640Shselasky *
76335640Shselasky * to define unaligned_uint16_t as a 16-bit unaligned data type.
77335640Shselasky * This could be presumably used, in sufficiently recent versions of
78335640Shselasky * the compiler, with macros similar to those below.  This would be
79335640Shselasky * useful only if that compiler could generate better code for PA-RISC
80335640Shselasky * or Itanium than would be generated by a bunch of shifts-and-ORs.
81335640Shselasky *
82335640Shselasky * DEC C, indicated by __DECC being defined, has, at least on Alpha,
83335640Shselasky * an __unaligned qualifier that can be applied to pointers to get the
84335640Shselasky * compiler to generate code that does unaligned loads and stores when
85335640Shselasky * dereferencing the pointer in question.
86335640Shselasky *
87335640Shselasky * XXX - what if the native C compiler doesn't support
88335640Shselasky * __attribute__((packed))?  How can we get it to generate unaligned
89335640Shselasky * accesses for *specific* items?
90335640Shselasky */
91335640Shselaskytypedef struct {
92335640Shselasky	uint16_t	val;
93335640Shselasky} __attribute__((packed)) unaligned_uint16_t;
94335640Shselasky
95335640Shselaskytypedef struct {
96335640Shselasky	uint32_t	val;
97335640Shselasky} __attribute__((packed)) unaligned_uint32_t;
98335640Shselasky
99335640Shselaskystatic inline uint16_t
100335640ShselaskyEXTRACT_16BITS(const void *p)
101335640Shselasky{
102335640Shselasky	return ((uint16_t)ntohs(((const unaligned_uint16_t *)(p))->val));
103335640Shselasky}
104335640Shselasky
105335640Shselaskystatic inline uint32_t
106335640ShselaskyEXTRACT_32BITS(const void *p)
107335640Shselasky{
108335640Shselasky	return ((uint32_t)ntohl(((const unaligned_uint32_t *)(p))->val));
109335640Shselasky}
110335640Shselasky
111335640Shselaskystatic inline uint64_t
112335640ShselaskyEXTRACT_64BITS(const void *p)
113335640Shselasky{
114335640Shselasky	return ((uint64_t)(((uint64_t)ntohl(((const unaligned_uint32_t *)(p) + 0)->val)) << 32 | \
115335640Shselasky		((uint64_t)ntohl(((const unaligned_uint32_t *)(p) + 1)->val)) << 0));
116335640Shselasky}
117335640Shselasky
118335640Shselasky#else /* have to do it a byte at a time */
119335640Shselasky/*
120335640Shselasky * This isn't a GCC-compatible compiler, we don't have __attribute__,
121335640Shselasky * or we do but we don't know of any better way with this instruction
122335640Shselasky * set to do unaligned loads, so do unaligned loads of big-endian
123335640Shselasky * quantities the hard way - fetch the bytes one at a time and
124335640Shselasky * assemble them.
125335640Shselasky */
126335640Shselasky#define EXTRACT_16BITS(p) \
127335640Shselasky	((uint16_t)(((uint16_t)(*((const uint8_t *)(p) + 0)) << 8) | \
128335640Shselasky	            ((uint16_t)(*((const uint8_t *)(p) + 1)) << 0)))
129335640Shselasky#define EXTRACT_32BITS(p) \
130335640Shselasky	((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 0)) << 24) | \
131335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 1)) << 16) | \
132335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 2)) << 8) | \
133335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 3)) << 0)))
134335640Shselasky#define EXTRACT_64BITS(p) \
135335640Shselasky	((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 56) | \
136335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 1)) << 48) | \
137335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 2)) << 40) | \
138335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 3)) << 32) | \
139335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 4)) << 24) | \
140335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 5)) << 16) | \
141335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 6)) << 8) | \
142335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 7)) << 0)))
143335640Shselasky#endif /* must special-case unaligned accesses */
144335640Shselasky#else /* LBL_ALIGN */
145335640Shselasky/*
146335640Shselasky * The processor natively handles unaligned loads, so we can just
147335640Shselasky * cast the pointer and fetch through it.
148335640Shselasky */
149335640Shselaskystatic inline uint16_t
150335640ShselaskyEXTRACT_16BITS(const void *p)
151335640Shselasky{
152335640Shselasky	return ((uint16_t)ntohs(*(const uint16_t *)(p)));
153335640Shselasky}
154335640Shselasky
155335640Shselaskystatic inline uint32_t
156335640ShselaskyEXTRACT_32BITS(const void *p)
157335640Shselasky{
158335640Shselasky	return ((uint32_t)ntohl(*(const uint32_t *)(p)));
159335640Shselasky}
160335640Shselasky
161335640Shselaskystatic inline uint64_t
162335640ShselaskyEXTRACT_64BITS(const void *p)
163335640Shselasky{
164335640Shselasky	return ((uint64_t)(((uint64_t)ntohl(*((const uint32_t *)(p) + 0))) << 32 | \
165335640Shselasky		((uint64_t)ntohl(*((const uint32_t *)(p) + 1))) << 0));
166335640Shselasky
167335640Shselasky}
168335640Shselasky
169335640Shselasky#endif /* LBL_ALIGN */
170335640Shselasky
171335640Shselasky#define EXTRACT_24BITS(p) \
172335640Shselasky	((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 0)) << 16) | \
173335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \
174335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 2)) << 0)))
175335640Shselasky
176335640Shselasky#define EXTRACT_40BITS(p) \
177335640Shselasky	((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 32) | \
178335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 1)) << 24) | \
179335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 2)) << 16) | \
180335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 3)) << 8) | \
181335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 4)) << 0)))
182335640Shselasky
183335640Shselasky#define EXTRACT_48BITS(p) \
184335640Shselasky	((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 40) | \
185335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 1)) << 32) | \
186335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 2)) << 24) | \
187335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 3)) << 16) | \
188335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 4)) << 8) | \
189335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 5)) << 0)))
190335640Shselasky
191335640Shselasky#define EXTRACT_56BITS(p) \
192335640Shselasky	((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 48) | \
193335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 1)) << 40) | \
194335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 2)) << 32) | \
195335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 3)) << 24) | \
196335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 4)) << 16) | \
197335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 5)) << 8) | \
198335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 6)) << 0)))
199335640Shselasky
200335640Shselasky/*
201335640Shselasky * Macros to extract possibly-unaligned little-endian integral values.
202335640Shselasky * XXX - do loads on little-endian machines that support unaligned loads?
203335640Shselasky */
204335640Shselasky#define EXTRACT_LE_8BITS(p) (*(p))
205335640Shselasky#define EXTRACT_LE_16BITS(p) \
206335640Shselasky	((uint16_t)(((uint16_t)(*((const uint8_t *)(p) + 1)) << 8) | \
207335640Shselasky	            ((uint16_t)(*((const uint8_t *)(p) + 0)) << 0)))
208335640Shselasky#define EXTRACT_LE_32BITS(p) \
209335640Shselasky	((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 3)) << 24) | \
210335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 2)) << 16) | \
211335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \
212335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 0)) << 0)))
213335640Shselasky#define EXTRACT_LE_24BITS(p) \
214335640Shselasky	((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 2)) << 16) | \
215335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \
216335640Shselasky	            ((uint32_t)(*((const uint8_t *)(p) + 0)) << 0)))
217335640Shselasky#define EXTRACT_LE_64BITS(p) \
218335640Shselasky	((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 7)) << 56) | \
219335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 6)) << 48) | \
220335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 5)) << 40) | \
221335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 4)) << 32) | \
222335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 3)) << 24) | \
223335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 2)) << 16) | \
224335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 1)) << 8) | \
225335640Shselasky	            ((uint64_t)(*((const uint8_t *)(p) + 0)) << 0)))
226