1 // Copyright (c) 2010 Google Inc. All Rights Reserved.
2 //
3 // Redistribution and use in source and binary forms, with or without
4 // modification, are permitted provided that the following conditions are
5 // met:
6 //
7 //     * Redistributions of source code must retain the above copyright
8 // notice, this list of conditions and the following disclaimer.
9 //     * Redistributions in binary form must reproduce the above
10 // copyright notice, this list of conditions and the following disclaimer
11 // in the documentation and/or other materials provided with the
12 // distribution.
13 //     * Neither the name of Google Inc. nor the names of its
14 // contributors may be used to endorse or promote products derived from
15 // this software without specific prior written permission.
16 //
17 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 
29 #include <assert.h>
30 #include <stdlib.h>
31 
32 #include "common/dwarf/bytereader-inl.h"
33 #include "common/dwarf/bytereader.h"
34 
35 namespace dwarf2reader {
36 
ByteReader(enum Endianness endian)37 ByteReader::ByteReader(enum Endianness endian)
38     :offset_reader_(NULL), address_reader_(NULL), endian_(endian),
39      address_size_(0), offset_size_(0),
40      have_section_base_(), have_text_base_(), have_data_base_(),
41      have_function_base_() { }
42 
~ByteReader()43 ByteReader::~ByteReader() { }
44 
SetOffsetSize(uint8 size)45 void ByteReader::SetOffsetSize(uint8 size) {
46   offset_size_ = size;
47   assert(size == 4 || size == 8);
48   if (size == 4) {
49     this->offset_reader_ = &ByteReader::ReadFourBytes;
50   } else {
51     this->offset_reader_ = &ByteReader::ReadEightBytes;
52   }
53 }
54 
SetAddressSize(uint8 size)55 void ByteReader::SetAddressSize(uint8 size) {
56   address_size_ = size;
57   assert(size == 4 || size == 8);
58   if (size == 4) {
59     this->address_reader_ = &ByteReader::ReadFourBytes;
60   } else {
61     this->address_reader_ = &ByteReader::ReadEightBytes;
62   }
63 }
64 
ReadInitialLength(const char * start,size_t * len)65 uint64 ByteReader::ReadInitialLength(const char* start, size_t* len) {
66   const uint64 initial_length = ReadFourBytes(start);
67   start += 4;
68 
69   // In DWARF2/3, if the initial length is all 1 bits, then the offset
70   // size is 8 and we need to read the next 8 bytes for the real length.
71   if (initial_length == 0xffffffff) {
72     SetOffsetSize(8);
73     *len = 12;
74     return ReadOffset(start);
75   } else {
76     SetOffsetSize(4);
77     *len = 4;
78   }
79   return initial_length;
80 }
81 
ValidEncoding(DwarfPointerEncoding encoding) const82 bool ByteReader::ValidEncoding(DwarfPointerEncoding encoding) const {
83   if (encoding == DW_EH_PE_omit) return true;
84   if (encoding == DW_EH_PE_aligned) return true;
85   if ((encoding & 0x7) > DW_EH_PE_udata8)
86     return false;
87   if ((encoding & 0x70) > DW_EH_PE_funcrel)
88     return false;
89   return true;
90 }
91 
UsableEncoding(DwarfPointerEncoding encoding) const92 bool ByteReader::UsableEncoding(DwarfPointerEncoding encoding) const {
93   switch (encoding & 0x70) {
94     case DW_EH_PE_absptr:  return true;
95     case DW_EH_PE_pcrel:   return have_section_base_;
96     case DW_EH_PE_textrel: return have_text_base_;
97     case DW_EH_PE_datarel: return have_data_base_;
98     case DW_EH_PE_funcrel: return have_function_base_;
99     default:               return false;
100   }
101 }
102 
ReadEncodedPointer(const char * buffer,DwarfPointerEncoding encoding,size_t * len) const103 uint64 ByteReader::ReadEncodedPointer(const char *buffer,
104                                       DwarfPointerEncoding encoding,
105                                       size_t *len) const {
106   // UsableEncoding doesn't approve of DW_EH_PE_omit, so we shouldn't
107   // see it here.
108   assert(encoding != DW_EH_PE_omit);
109 
110   // The Linux Standards Base 4.0 does not make this clear, but the
111   // GNU tools (gcc/unwind-pe.h; readelf/dwarf.c; gdb/dwarf2-frame.c)
112   // agree that aligned pointers are always absolute, machine-sized,
113   // machine-signed pointers.
114   if (encoding == DW_EH_PE_aligned) {
115     assert(have_section_base_);
116 
117     // We don't need to align BUFFER in *our* address space. Rather, we
118     // need to find the next position in our buffer that would be aligned
119     // when the .eh_frame section the buffer contains is loaded into the
120     // program's memory. So align assuming that buffer_base_ gets loaded at
121     // address section_base_, where section_base_ itself may or may not be
122     // aligned.
123 
124     // First, find the offset to START from the closest prior aligned
125     // address.
126     uint64 skew = section_base_ & (AddressSize() - 1);
127     // Now find the offset from that aligned address to buffer.
128     uint64 offset = skew + (buffer - buffer_base_);
129     // Round up to the next boundary.
130     uint64 aligned = (offset + AddressSize() - 1) & -AddressSize();
131     // Convert back to a pointer.
132     const char *aligned_buffer = buffer_base_ + (aligned - skew);
133     // Finally, store the length and actually fetch the pointer.
134     *len = aligned_buffer - buffer + AddressSize();
135     return ReadAddress(aligned_buffer);
136   }
137 
138   // Extract the value first, ignoring whether it's a pointer or an
139   // offset relative to some base.
140   uint64 offset;
141   switch (encoding & 0x0f) {
142     case DW_EH_PE_absptr:
143       // DW_EH_PE_absptr is weird, as it is used as a meaningful value for
144       // both the high and low nybble of encoding bytes. When it appears in
145       // the high nybble, it means that the pointer is absolute, not an
146       // offset from some base address. When it appears in the low nybble,
147       // as here, it means that the pointer is stored as a normal
148       // machine-sized and machine-signed address. A low nybble of
149       // DW_EH_PE_absptr does not imply that the pointer is absolute; it is
150       // correct for us to treat the value as an offset from a base address
151       // if the upper nybble is not DW_EH_PE_absptr.
152       offset = ReadAddress(buffer);
153       *len = AddressSize();
154       break;
155 
156     case DW_EH_PE_uleb128:
157       offset = ReadUnsignedLEB128(buffer, len);
158       break;
159 
160     case DW_EH_PE_udata2:
161       offset = ReadTwoBytes(buffer);
162       *len = 2;
163       break;
164 
165     case DW_EH_PE_udata4:
166       offset = ReadFourBytes(buffer);
167       *len = 4;
168       break;
169 
170     case DW_EH_PE_udata8:
171       offset = ReadEightBytes(buffer);
172       *len = 8;
173       break;
174 
175     case DW_EH_PE_sleb128:
176       offset = ReadSignedLEB128(buffer, len);
177       break;
178 
179     case DW_EH_PE_sdata2:
180       offset = ReadTwoBytes(buffer);
181       // Sign-extend from 16 bits.
182       offset = (offset ^ 0x8000) - 0x8000;
183       *len = 2;
184       break;
185 
186     case DW_EH_PE_sdata4:
187       offset = ReadFourBytes(buffer);
188       // Sign-extend from 32 bits.
189       offset = (offset ^ 0x80000000ULL) - 0x80000000ULL;
190       *len = 4;
191       break;
192 
193     case DW_EH_PE_sdata8:
194       // No need to sign-extend; this is the full width of our type.
195       offset = ReadEightBytes(buffer);
196       *len = 8;
197       break;
198 
199     default:
200       abort();
201   }
202 
203   // Find the appropriate base address.
204   uint64 base;
205   switch (encoding & 0x70) {
206     case DW_EH_PE_absptr:
207       base = 0;
208       break;
209 
210     case DW_EH_PE_pcrel:
211       assert(have_section_base_);
212       base = section_base_ + (buffer - buffer_base_);
213       break;
214 
215     case DW_EH_PE_textrel:
216       assert(have_text_base_);
217       base = text_base_;
218       break;
219 
220     case DW_EH_PE_datarel:
221       assert(have_data_base_);
222       base = data_base_;
223       break;
224 
225     case DW_EH_PE_funcrel:
226       assert(have_function_base_);
227       base = function_base_;
228       break;
229 
230     default:
231       abort();
232   }
233 
234   uint64 pointer = base + offset;
235 
236   // Remove inappropriate upper bits.
237   if (AddressSize() == 4)
238     pointer = pointer & 0xffffffff;
239   else
240     assert(AddressSize() == sizeof(uint64));
241 
242   return pointer;
243 }
244 
245 }  // namespace dwarf2reader
246