xref: /titanic_50/usr/src/cmd/sgs/tools/common/leb128.c (revision e1d3217b9afde782c4d3e946fda0e6ef36a61306)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #include <stdio.h>
28 #include <dwarf.h>
29 #include <sys/types.h>
30 #include <sys/elf.h>
31 
32 /*
33  * Little Endian Base 128 (LEB128) numbers.
34  * ----------------------------------------
35  *
36  * LEB128 is a scheme for encoding integers densely that exploits the
37  * assumption that most integers are small in magnitude. (This encoding
38  * is equally suitable whether the target machine architecture represents
39  * data in big-endian or little- endian
40  *
41  * Unsigned LEB128 numbers are encoded as follows: start at the low order
42  * end of an unsigned integer and chop it into 7-bit chunks. Place each
43  * chunk into the low order 7 bits of a byte. Typically, several of the
44  * high order bytes will be zero; discard them. Emit the remaining bytes in
45  * a stream, starting with the low order byte; set the high order bit on
46  * each byte except the last emitted byte. The high bit of zero on the last
47  * byte indicates to the decoder that it has encountered the last byte.
48  * The integer zero is a special case, consisting of a single zero byte.
49  *
50  * Signed, 2s complement LEB128 numbers are encoded in a similar except
51  * that the criterion for discarding high order bytes is not whether they
52  * are zero, but whether they consist entirely of sign extension bits.
53  * Consider the 32-bit integer -2. The three high level bytes of the number
54  * are sign extension, thus LEB128 would represent it as a single byte
55  * containing the low order 7 bits, with the high order bit cleared to
56  * indicate the end of the byte stream.
57  *
58  * Note that there is nothing within the LEB128 representation that
59  * indicates whether an encoded number is signed or unsigned. The decoder
60  * must know what type of number to expect.
61  *
62  * DWARF Exception Header Encoding
63  * -------------------------------
64  *
65  * The DWARF Exception Header Encoding is used to describe the type of data
66  * used in the .eh_frame_hdr section. The upper 4 bits indicate how the
67  * value is to be applied. The lower 4 bits indicate the format of the data.
68  *
69  * DWARF Exception Header value format
70  *
71  * Name		Value Meaning
72  * DW_EH_PE_omit	    0xff No value is present.
73  * DW_EH_PE_absptr	    0x00 Value is a void*
74  * DW_EH_PE_uleb128	    0x01 Unsigned value is encoded using the
75  *				 Little Endian Base 128 (LEB128)
76  * DW_EH_PE_udata2	    0x02 A 2 bytes unsigned value.
77  * DW_EH_PE_udata4	    0x03 A 4 bytes unsigned value.
78  * DW_EH_PE_udata8	    0x04 An 8 bytes unsigned value.
79  * DW_EH_PE_signed          0x08 bit on for all signed encodings
80  * DW_EH_PE_sleb128	    0x09 Signed value is encoded using the
81  *				 Little Endian Base 128 (LEB128)
82  * DW_EH_PE_sdata2	    0x0A A 2 bytes signed value.
83  * DW_EH_PE_sdata4	    0x0B A 4 bytes signed value.
84  * DW_EH_PE_sdata8	    0x0C An 8 bytes signed value.
85  *
86  * DWARF Exception Header application
87  *
88  * Name	    Value Meaning
89  * DW_EH_PE_absptr	   0x00 Value is used with no modification.
90  * DW_EH_PE_pcrel	   0x10 Value is reletive to the location of itself
91  * DW_EH_PE_textrel	   0x20
92  * DW_EH_PE_datarel	   0x30 Value is reletive to the beginning of the
93  *				eh_frame_hdr segment ( segment type
94  *			        PT_GNU_EH_FRAME )
95  * DW_EH_PE_funcrel        0x40
96  * DW_EH_PE_aligned        0x50 value is an aligned void*
97  * DW_EH_PE_indirect       0x80 bit to signal indirection after relocation
98  * DW_EH_PE_omit	   0xff No value is present.
99  *
100  */
101 
102 uint64_t
103 uleb_extract(unsigned char *data, uint64_t *dotp)
104 {
105 	uint64_t	dot = *dotp;
106 	uint64_t	res = 0;
107 	int		more = 1;
108 	int		shift = 0;
109 	int		val;
110 
111 	data += dot;
112 
113 	while (more) {
114 		/*
115 		 * Pull off lower 7 bits
116 		 */
117 		val = (*data) & 0x7f;
118 
119 		/*
120 		 * Add prepend value to head of number.
121 		 */
122 		res = res | (val << shift);
123 
124 		/*
125 		 * Increment shift & dot pointer
126 		 */
127 		shift += 7;
128 		dot++;
129 
130 		/*
131 		 * Check to see if hi bit is set - if not, this
132 		 * is the last byte.
133 		 */
134 		more = ((*data++) & 0x80) >> 7;
135 	}
136 	*dotp = dot;
137 	return (res);
138 }
139 
140 int64_t
141 sleb_extract(unsigned char *data, uint64_t *dotp)
142 {
143 	uint64_t	dot = *dotp;
144 	int64_t		res = 0;
145 	int		more = 1;
146 	int		shift = 0;
147 	int		val;
148 
149 	data += dot;
150 
151 	while (more) {
152 		/*
153 		 * Pull off lower 7 bits
154 		 */
155 		val = (*data) & 0x7f;
156 
157 		/*
158 		 * Add prepend value to head of number.
159 		 */
160 		res = res | (val << shift);
161 
162 		/*
163 		 * Increment shift & dot pointer
164 		 */
165 		shift += 7;
166 		dot++;
167 
168 		/*
169 		 * Check to see if hi bit is set - if not, this
170 		 * is the last byte.
171 		 */
172 		more = ((*data++) & 0x80) >> 7;
173 	}
174 	*dotp = dot;
175 
176 	/*
177 	 * Make sure value is properly sign extended.
178 	 */
179 	res = (res << (64 - shift)) >> (64 - shift);
180 
181 	return (res);
182 }
183 
184 /*
185  * Extract a DWARF encoded datum
186  *
187  * entry:
188  *	data - Base of data buffer containing encoded bytes
189  *	dotp - Address of variable containing index within data
190  *		at which the desired datum starts.
191  *	ehe_flags - DWARF encoding
192  *	eident - ELF header e_ident[] array for object being processed
193  *	frame_hdr - Boolean, true if we're extracting from .eh_frame_hdr
194  *	sh_base - Base address of ELF section containing desired datum
195  *	sh_offset - Offset relative to sh_base of desired datum.
196  *	dbase - The base address to which DW_EH_PE_datarel is relative
197  *		(if frame_hdr is false)
198  */
199 uint64_t
200 dwarf_ehe_extract(unsigned char *data, uint64_t *dotp, uint_t ehe_flags,
201     unsigned char *eident, boolean_t frame_hdr, uint64_t sh_base,
202     uint64_t sh_offset, uint64_t dbase)
203 {
204 	uint64_t    dot = *dotp;
205 	uint_t	    lsb;
206 	uint_t	    wordsize;
207 	uint_t	    fsize;
208 	uint64_t    result;
209 
210 	if (eident[EI_DATA] == ELFDATA2LSB)
211 		lsb = 1;
212 	else
213 		lsb = 0;
214 
215 	if (eident[EI_CLASS] == ELFCLASS64)
216 		wordsize = 8;
217 	else
218 		wordsize = 4;
219 
220 	switch (ehe_flags & 0x0f) {
221 	case DW_EH_PE_omit:
222 		return (0);
223 	case DW_EH_PE_absptr:
224 		fsize = wordsize;
225 		break;
226 	case DW_EH_PE_udata8:
227 	case DW_EH_PE_sdata8:
228 		fsize = 8;
229 		break;
230 	case DW_EH_PE_udata4:
231 	case DW_EH_PE_sdata4:
232 		fsize = 4;
233 		break;
234 	case DW_EH_PE_udata2:
235 	case DW_EH_PE_sdata2:
236 		fsize = 2;
237 		break;
238 	case DW_EH_PE_uleb128:
239 		return (uleb_extract(data, dotp));
240 	case DW_EH_PE_sleb128:
241 		return ((uint64_t)sleb_extract(data, dotp));
242 	default:
243 		return (0);
244 	}
245 
246 	if (lsb) {
247 		/*
248 		 * Extract unaligned LSB formated data
249 		 */
250 		uint_t	cnt;
251 
252 		result = 0;
253 		for (cnt = 0; cnt < fsize;
254 		    cnt++, dot++) {
255 			uint64_t val;
256 			val = data[dot];
257 			result |= val << (cnt * 8);
258 		}
259 	} else {
260 		/*
261 		 * Extract unaligned MSB formated data
262 		 */
263 		uint_t	cnt;
264 		result = 0;
265 		for (cnt = 0; cnt < fsize;
266 		    cnt++, dot++) {
267 			uint64_t	val;
268 			val = data[dot];
269 			result |= val << ((fsize - cnt - 1) * 8);
270 		}
271 	}
272 	/*
273 	 * perform sign extension
274 	 */
275 	if ((ehe_flags & DW_EH_PE_signed) &&
276 	    (fsize < sizeof (uint64_t))) {
277 		int64_t	sresult;
278 		uint_t	bitshift;
279 		sresult = result;
280 		bitshift = (sizeof (uint64_t) - fsize) * 8;
281 		sresult = (sresult << bitshift) >> bitshift;
282 		result = sresult;
283 	}
284 
285 	/*
286 	 * If value is relative to a base address, adjust it
287 	 */
288 	switch (ehe_flags & 0xf0) {
289 	case DW_EH_PE_pcrel:
290 		result += sh_base + sh_offset;
291 		break;
292 
293 	/*
294 	 * datarel is relative to .eh_frame_hdr if within .eh_frame,
295 	 * but GOT if not.
296 	 */
297 	case DW_EH_PE_datarel:
298 		if (frame_hdr)
299 			result += sh_base;
300 		else
301 			result += dbase;
302 		break;
303 	}
304 
305 	/* Truncate the result to its specified size */
306 	result = (result << ((sizeof (uint64_t) - fsize) * 8)) >>
307 	    ((sizeof (uint64_t) - fsize) * 8);
308 
309 	*dotp = dot;
310 	return (result);
311 }
312