1//===-- DWARFDebugArangeSet.cpp -------------------------------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8
9#include "DWARFDebugArangeSet.h"
10#include "DWARFDataExtractor.h"
11#include "LogChannelDWARF.h"
12#include "llvm/Object/Error.h"
13#include <cassert>
14
15using namespace lldb_private;
16using namespace lldb_private::plugin::dwarf;
17
18DWARFDebugArangeSet::DWARFDebugArangeSet()
19    : m_offset(DW_INVALID_OFFSET), m_next_offset(DW_INVALID_OFFSET) {}
20
21void DWARFDebugArangeSet::Clear() {
22  m_offset = DW_INVALID_OFFSET;
23  m_next_offset = DW_INVALID_OFFSET;
24  m_header.length = 0;
25  m_header.version = 0;
26  m_header.cu_offset = 0;
27  m_header.addr_size = 0;
28  m_header.seg_size = 0;
29  m_arange_descriptors.clear();
30}
31
32llvm::Error DWARFDebugArangeSet::extract(const DWARFDataExtractor &data,
33                                         lldb::offset_t *offset_ptr) {
34  assert(data.ValidOffset(*offset_ptr));
35
36  m_arange_descriptors.clear();
37  m_offset = *offset_ptr;
38
39  // 7.20 Address Range Table
40  //
41  // Each set of entries in the table of address ranges contained in the
42  // .debug_aranges section begins with a header consisting of: a 4-byte
43  // length containing the length of the set of entries for this compilation
44  // unit, not including the length field itself; a 2-byte version identifier
45  // containing the value 2 for DWARF Version 2; a 4-byte offset into
46  // the.debug_infosection; a 1-byte unsigned integer containing the size in
47  // bytes of an address (or the offset portion of an address for segmented
48  // addressing) on the target system; and a 1-byte unsigned integer
49  // containing the size in bytes of a segment descriptor on the target
50  // system. This header is followed by a series of tuples. Each tuple
51  // consists of an address and a length, each in the size appropriate for an
52  // address on the target architecture.
53  m_header.length = data.GetDWARFInitialLength(offset_ptr);
54  // The length could be 4 bytes or 12 bytes, so use the current offset to
55  // determine the next offset correctly.
56  if (m_header.length > 0)
57    m_next_offset = *offset_ptr + m_header.length;
58  else
59    m_next_offset = DW_INVALID_OFFSET;
60  m_header.version = data.GetU16(offset_ptr);
61  m_header.cu_offset = data.GetDWARFOffset(offset_ptr);
62  m_header.addr_size = data.GetU8(offset_ptr);
63  m_header.seg_size = data.GetU8(offset_ptr);
64
65  // Try to avoid reading invalid arange sets by making sure:
66  // 1 - the version looks good
67  // 2 - the address byte size looks plausible
68  // 3 - the length seems to make sense
69  // 4 - size looks plausible
70  // 5 - the arange tuples do not contain a segment field
71  if (m_header.version < 2 || m_header.version > 5)
72    return llvm::make_error<llvm::object::GenericBinaryError>(
73        "Invalid arange header version");
74
75  if (m_header.addr_size != 4 && m_header.addr_size != 8)
76    return llvm::make_error<llvm::object::GenericBinaryError>(
77        "Invalid arange header address size");
78
79  if (m_header.length == 0)
80    return llvm::make_error<llvm::object::GenericBinaryError>(
81        "Invalid arange header length");
82
83  if (!data.ValidOffset(m_offset + sizeof(m_header.length) + m_header.length -
84                        1))
85    return llvm::make_error<llvm::object::GenericBinaryError>(
86        "Invalid arange header length");
87
88  if (m_header.seg_size)
89    return llvm::make_error<llvm::object::GenericBinaryError>(
90        "segmented arange entries are not supported");
91
92  // The first tuple following the header in each set begins at an offset
93  // that is a multiple of the size of a single tuple (that is, twice the
94  // size of an address). The header is padded, if necessary, to the
95  // appropriate boundary.
96  const uint32_t header_size = *offset_ptr - m_offset;
97  const uint32_t tuple_size = m_header.addr_size << 1;
98  uint32_t first_tuple_offset = 0;
99  while (first_tuple_offset < header_size)
100    first_tuple_offset += tuple_size;
101
102  *offset_ptr = m_offset + first_tuple_offset;
103
104  Descriptor arangeDescriptor;
105
106  static_assert(sizeof(arangeDescriptor.address) ==
107                    sizeof(arangeDescriptor.length),
108                "DWARFDebugArangeSet::Descriptor.address and "
109                "DWARFDebugArangeSet::Descriptor.length must have same size");
110
111  const lldb::offset_t next_offset = GetNextOffset();
112  assert(next_offset != DW_INVALID_OFFSET);
113  uint32_t num_terminators = 0;
114  bool last_was_terminator = false;
115  while (*offset_ptr < next_offset) {
116    arangeDescriptor.address = data.GetMaxU64(offset_ptr, m_header.addr_size);
117    arangeDescriptor.length = data.GetMaxU64(offset_ptr, m_header.addr_size);
118
119    // Each set of tuples is terminated by a 0 for the address and 0 for
120    // the length. Some linkers can emit .debug_aranges with multiple
121    // terminator pair entries that are still withing the length of the
122    // DWARFDebugArangeSet. We want to be sure to parse all entries for
123    // this DWARFDebugArangeSet so that we don't stop parsing early and end up
124    // treating addresses as a header of the next DWARFDebugArangeSet. We also
125    // need to make sure we parse all valid address pairs so we don't omit them
126    // from the aranges result, so we can't stop at the first terminator entry
127    // we find.
128    if (arangeDescriptor.address == 0 && arangeDescriptor.length == 0) {
129      ++num_terminators;
130      last_was_terminator = true;
131    } else {
132      last_was_terminator = false;
133      // Only add .debug_aranges address entries that have a non zero size.
134      // Some linkers will zero out the length field for some .debug_aranges
135      // entries if they were stripped. We also could watch out for multiple
136      // entries at address zero and remove those as well.
137      if (arangeDescriptor.length > 0)
138        m_arange_descriptors.push_back(arangeDescriptor);
139    }
140  }
141  if (num_terminators > 1) {
142    Log *log = GetLog(DWARFLog::DebugInfo);
143    LLDB_LOG(log,
144             "warning: DWARFDebugArangeSet at %#" PRIx64 " contains %u "
145             "terminator entries",
146             m_offset, num_terminators);
147  }
148  if (last_was_terminator)
149    return llvm::ErrorSuccess();
150
151  return llvm::make_error<llvm::object::GenericBinaryError>(
152      "arange descriptors not terminated by null entry");
153}
154
155class DescriptorContainsAddress {
156public:
157  DescriptorContainsAddress(dw_addr_t address) : m_address(address) {}
158  bool operator()(const DWARFDebugArangeSet::Descriptor &desc) const {
159    return (m_address >= desc.address) &&
160           (m_address < (desc.address + desc.length));
161  }
162
163private:
164  const dw_addr_t m_address;
165};
166
167dw_offset_t DWARFDebugArangeSet::FindAddress(dw_addr_t address) const {
168  DescriptorConstIter end = m_arange_descriptors.end();
169  DescriptorConstIter pos =
170      std::find_if(m_arange_descriptors.begin(), end,   // Range
171                   DescriptorContainsAddress(address)); // Predicate
172  if (pos != end)
173    return m_header.cu_offset;
174
175  return DW_INVALID_OFFSET;
176}
177