1 //===-- MinidumpParser.cpp ---------------------------------------*- C++ -*-===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 
10 // Project includes
11 #include "MinidumpParser.h"
12 #include "NtStructures.h"
13 #include "RegisterContextMinidump_x86_32.h"
14 
15 // Other libraries and framework includes
16 #include "lldb/Target/MemoryRegionInfo.h"
17 
18 // C includes
19 // C++ includes
20 #include <map>
21 
22 using namespace lldb_private;
23 using namespace minidump;
24 
25 llvm::Optional<MinidumpParser>
26 MinidumpParser::Create(const lldb::DataBufferSP &data_buf_sp) {
27   if (data_buf_sp->GetByteSize() < sizeof(MinidumpHeader)) {
28     return llvm::None;
29   }
30 
31   llvm::ArrayRef<uint8_t> header_data(data_buf_sp->GetBytes(),
32                                       sizeof(MinidumpHeader));
33   const MinidumpHeader *header = MinidumpHeader::Parse(header_data);
34 
35   if (header == nullptr) {
36     return llvm::None;
37   }
38 
39   lldb::offset_t directory_list_offset = header->stream_directory_rva;
40   // check if there is enough data for the parsing of the directory list
41   if ((directory_list_offset +
42        sizeof(MinidumpDirectory) * header->streams_count) >
43       data_buf_sp->GetByteSize()) {
44     return llvm::None;
45   }
46 
47   const MinidumpDirectory *directory = nullptr;
48   Status error;
49   llvm::ArrayRef<uint8_t> directory_data(
50       data_buf_sp->GetBytes() + directory_list_offset,
51       sizeof(MinidumpDirectory) * header->streams_count);
52   llvm::DenseMap<uint32_t, MinidumpLocationDescriptor> directory_map;
53 
54   for (uint32_t i = 0; i < header->streams_count; ++i) {
55     error = consumeObject(directory_data, directory);
56     if (error.Fail()) {
57       return llvm::None;
58     }
59     directory_map[static_cast<const uint32_t>(directory->stream_type)] =
60         directory->location;
61   }
62 
63   return MinidumpParser(data_buf_sp, header, std::move(directory_map));
64 }
65 
66 MinidumpParser::MinidumpParser(
67     const lldb::DataBufferSP &data_buf_sp, const MinidumpHeader *header,
68     llvm::DenseMap<uint32_t, MinidumpLocationDescriptor> &&directory_map)
69     : m_data_sp(data_buf_sp), m_header(header), m_directory_map(directory_map) {
70 }
71 
72 llvm::ArrayRef<uint8_t> MinidumpParser::GetData() {
73   return llvm::ArrayRef<uint8_t>(m_data_sp->GetBytes(),
74                                  m_data_sp->GetByteSize());
75 }
76 
77 llvm::ArrayRef<uint8_t>
78 MinidumpParser::GetStream(MinidumpStreamType stream_type) {
79   auto iter = m_directory_map.find(static_cast<uint32_t>(stream_type));
80   if (iter == m_directory_map.end())
81     return {};
82 
83   // check if there is enough data
84   if (iter->second.rva + iter->second.data_size > m_data_sp->GetByteSize())
85     return {};
86 
87   return llvm::ArrayRef<uint8_t>(m_data_sp->GetBytes() + iter->second.rva,
88                                  iter->second.data_size);
89 }
90 
91 llvm::Optional<std::string> MinidumpParser::GetMinidumpString(uint32_t rva) {
92   auto arr_ref = m_data_sp->GetData();
93   if (rva > arr_ref.size())
94     return llvm::None;
95   arr_ref = arr_ref.drop_front(rva);
96   return parseMinidumpString(arr_ref);
97 }
98 
99 llvm::ArrayRef<MinidumpThread> MinidumpParser::GetThreads() {
100   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::ThreadList);
101 
102   if (data.size() == 0)
103     return llvm::None;
104 
105   return MinidumpThread::ParseThreadList(data);
106 }
107 
108 llvm::ArrayRef<uint8_t>
109 MinidumpParser::GetThreadContext(const MinidumpThread &td) {
110   if (td.thread_context.rva + td.thread_context.data_size > GetData().size())
111     return {};
112 
113   return GetData().slice(td.thread_context.rva, td.thread_context.data_size);
114 }
115 
116 llvm::ArrayRef<uint8_t>
117 MinidumpParser::GetThreadContextWow64(const MinidumpThread &td) {
118   // On Windows, a 32-bit process can run on a 64-bit machine under
119   // WOW64. If the minidump was captured with a 64-bit debugger, then
120   // the CONTEXT we just grabbed from the mini_dump_thread is the one
121   // for the 64-bit "native" process rather than the 32-bit "guest"
122   // process we care about.  In this case, we can get the 32-bit CONTEXT
123   // from the TEB (Thread Environment Block) of the 64-bit process.
124   auto teb_mem = GetMemory(td.teb, sizeof(TEB64));
125   if (teb_mem.empty())
126     return {};
127 
128   const TEB64 *wow64teb;
129   Status error = consumeObject(teb_mem, wow64teb);
130   if (error.Fail())
131     return {};
132 
133   // Slot 1 of the thread-local storage in the 64-bit TEB points to a
134   // structure that includes the 32-bit CONTEXT (after a ULONG).
135   // See:  https://msdn.microsoft.com/en-us/library/ms681670.aspx
136   auto context =
137       GetMemory(wow64teb->tls_slots[1] + 4, sizeof(MinidumpContext_x86_32));
138   if (context.size() < sizeof(MinidumpContext_x86_32))
139     return {};
140 
141   return context;
142   // NOTE:  We don't currently use the TEB for anything else.  If we
143   // need it in the future, the 32-bit TEB is located according to the address
144   // stored in the first slot of the 64-bit TEB (wow64teb.Reserved1[0]).
145 }
146 
147 const MinidumpSystemInfo *MinidumpParser::GetSystemInfo() {
148   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::SystemInfo);
149 
150   if (data.size() == 0)
151     return nullptr;
152 
153   return MinidumpSystemInfo::Parse(data);
154 }
155 
156 ArchSpec MinidumpParser::GetArchitecture() {
157   ArchSpec arch_spec;
158   const MinidumpSystemInfo *system_info = GetSystemInfo();
159 
160   if (!system_info)
161     return arch_spec;
162 
163   // TODO what to do about big endiand flavors of arm ?
164   // TODO set the arm subarch stuff if the minidump has info about it
165 
166   llvm::Triple triple;
167   triple.setVendor(llvm::Triple::VendorType::UnknownVendor);
168 
169   const MinidumpCPUArchitecture arch =
170       static_cast<const MinidumpCPUArchitecture>(
171           static_cast<const uint32_t>(system_info->processor_arch));
172 
173   switch (arch) {
174   case MinidumpCPUArchitecture::X86:
175     triple.setArch(llvm::Triple::ArchType::x86);
176     break;
177   case MinidumpCPUArchitecture::AMD64:
178     triple.setArch(llvm::Triple::ArchType::x86_64);
179     break;
180   case MinidumpCPUArchitecture::ARM:
181     triple.setArch(llvm::Triple::ArchType::arm);
182     break;
183   case MinidumpCPUArchitecture::ARM64:
184     triple.setArch(llvm::Triple::ArchType::aarch64);
185     break;
186   default:
187     triple.setArch(llvm::Triple::ArchType::UnknownArch);
188     break;
189   }
190 
191   const MinidumpOSPlatform os = static_cast<const MinidumpOSPlatform>(
192       static_cast<const uint32_t>(system_info->platform_id));
193 
194   // TODO add all of the OSes that Minidump/breakpad distinguishes?
195   switch (os) {
196   case MinidumpOSPlatform::Win32S:
197   case MinidumpOSPlatform::Win32Windows:
198   case MinidumpOSPlatform::Win32NT:
199   case MinidumpOSPlatform::Win32CE:
200     triple.setOS(llvm::Triple::OSType::Win32);
201     break;
202   case MinidumpOSPlatform::Linux:
203     triple.setOS(llvm::Triple::OSType::Linux);
204     break;
205   case MinidumpOSPlatform::MacOSX:
206     triple.setOS(llvm::Triple::OSType::MacOSX);
207     break;
208   case MinidumpOSPlatform::Android:
209     triple.setOS(llvm::Triple::OSType::Linux);
210     triple.setEnvironment(llvm::Triple::EnvironmentType::Android);
211     break;
212   default:
213     triple.setOS(llvm::Triple::OSType::UnknownOS);
214     break;
215   }
216 
217   arch_spec.SetTriple(triple);
218 
219   return arch_spec;
220 }
221 
222 const MinidumpMiscInfo *MinidumpParser::GetMiscInfo() {
223   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::MiscInfo);
224 
225   if (data.size() == 0)
226     return nullptr;
227 
228   return MinidumpMiscInfo::Parse(data);
229 }
230 
231 llvm::Optional<LinuxProcStatus> MinidumpParser::GetLinuxProcStatus() {
232   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::LinuxProcStatus);
233 
234   if (data.size() == 0)
235     return llvm::None;
236 
237   return LinuxProcStatus::Parse(data);
238 }
239 
240 llvm::Optional<lldb::pid_t> MinidumpParser::GetPid() {
241   const MinidumpMiscInfo *misc_info = GetMiscInfo();
242   if (misc_info != nullptr) {
243     return misc_info->GetPid();
244   }
245 
246   llvm::Optional<LinuxProcStatus> proc_status = GetLinuxProcStatus();
247   if (proc_status.hasValue()) {
248     return proc_status->GetPid();
249   }
250 
251   return llvm::None;
252 }
253 
254 llvm::ArrayRef<MinidumpModule> MinidumpParser::GetModuleList() {
255   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::ModuleList);
256 
257   if (data.size() == 0)
258     return {};
259 
260   return MinidumpModule::ParseModuleList(data);
261 }
262 
263 std::vector<const MinidumpModule *> MinidumpParser::GetFilteredModuleList() {
264   llvm::ArrayRef<MinidumpModule> modules = GetModuleList();
265   // map module_name -> pair(load_address, pointer to module struct in memory)
266   llvm::StringMap<std::pair<uint64_t, const MinidumpModule *>> lowest_addr;
267 
268   std::vector<const MinidumpModule *> filtered_modules;
269 
270   llvm::Optional<std::string> name;
271   std::string module_name;
272 
273   for (const auto &module : modules) {
274     name = GetMinidumpString(module.module_name_rva);
275 
276     if (!name)
277       continue;
278 
279     module_name = name.getValue();
280 
281     auto iter = lowest_addr.end();
282     bool exists;
283     std::tie(iter, exists) = lowest_addr.try_emplace(
284         module_name, std::make_pair(module.base_of_image, &module));
285 
286     if (exists && module.base_of_image < iter->second.first)
287       iter->second = std::make_pair(module.base_of_image, &module);
288   }
289 
290   filtered_modules.reserve(lowest_addr.size());
291   for (const auto &module : lowest_addr) {
292     filtered_modules.push_back(module.second.second);
293   }
294 
295   return filtered_modules;
296 }
297 
298 const MinidumpExceptionStream *MinidumpParser::GetExceptionStream() {
299   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::Exception);
300 
301   if (data.size() == 0)
302     return nullptr;
303 
304   return MinidumpExceptionStream::Parse(data);
305 }
306 
307 llvm::Optional<minidump::Range>
308 MinidumpParser::FindMemoryRange(lldb::addr_t addr) {
309   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::MemoryList);
310   llvm::ArrayRef<uint8_t> data64 = GetStream(MinidumpStreamType::Memory64List);
311 
312   if (data.empty() && data64.empty())
313     return llvm::None;
314 
315   if (!data.empty()) {
316     llvm::ArrayRef<MinidumpMemoryDescriptor> memory_list =
317         MinidumpMemoryDescriptor::ParseMemoryList(data);
318 
319     if (memory_list.empty())
320       return llvm::None;
321 
322     for (const auto &memory_desc : memory_list) {
323       const MinidumpLocationDescriptor &loc_desc = memory_desc.memory;
324       const lldb::addr_t range_start = memory_desc.start_of_memory_range;
325       const size_t range_size = loc_desc.data_size;
326 
327       if (loc_desc.rva + loc_desc.data_size > GetData().size())
328         return llvm::None;
329 
330       if (range_start <= addr && addr < range_start + range_size) {
331         return minidump::Range(range_start,
332                                GetData().slice(loc_desc.rva, range_size));
333       }
334     }
335   }
336 
337   // Some Minidumps have a Memory64ListStream that captures all the heap
338   // memory (full-memory Minidumps).  We can't exactly use the same loop as
339   // above, because the Minidump uses slightly different data structures to
340   // describe those
341 
342   if (!data64.empty()) {
343     llvm::ArrayRef<MinidumpMemoryDescriptor64> memory64_list;
344     uint64_t base_rva;
345     std::tie(memory64_list, base_rva) =
346         MinidumpMemoryDescriptor64::ParseMemory64List(data64);
347 
348     if (memory64_list.empty())
349       return llvm::None;
350 
351     for (const auto &memory_desc64 : memory64_list) {
352       const lldb::addr_t range_start = memory_desc64.start_of_memory_range;
353       const size_t range_size = memory_desc64.data_size;
354 
355       if (base_rva + range_size > GetData().size())
356         return llvm::None;
357 
358       if (range_start <= addr && addr < range_start + range_size) {
359         return minidump::Range(range_start,
360                                GetData().slice(base_rva, range_size));
361       }
362       base_rva += range_size;
363     }
364   }
365 
366   return llvm::None;
367 }
368 
369 llvm::ArrayRef<uint8_t> MinidumpParser::GetMemory(lldb::addr_t addr,
370                                                   size_t size) {
371   // I don't have a sense of how frequently this is called or how many memory
372   // ranges a Minidump typically has, so I'm not sure if searching for the
373   // appropriate range linearly each time is stupid.  Perhaps we should build
374   // an index for faster lookups.
375   llvm::Optional<minidump::Range> range = FindMemoryRange(addr);
376   if (!range)
377     return {};
378 
379   // There's at least some overlap between the beginning of the desired range
380   // (addr) and the current range.  Figure out where the overlap begins and
381   // how much overlap there is.
382 
383   const size_t offset = addr - range->start;
384 
385   if (addr < range->start || offset >= range->range_ref.size())
386     return {};
387 
388   const size_t overlap = std::min(size, range->range_ref.size() - offset);
389   return range->range_ref.slice(offset, overlap);
390 }
391 
392 llvm::Optional<MemoryRegionInfo>
393 MinidumpParser::GetMemoryRegionInfo(lldb::addr_t load_addr) {
394   MemoryRegionInfo info;
395   llvm::ArrayRef<uint8_t> data = GetStream(MinidumpStreamType::MemoryInfoList);
396   if (data.empty())
397     return llvm::None;
398 
399   std::vector<const MinidumpMemoryInfo *> mem_info_list =
400       MinidumpMemoryInfo::ParseMemoryInfoList(data);
401   if (mem_info_list.empty())
402     return llvm::None;
403 
404   const auto yes = MemoryRegionInfo::eYes;
405   const auto no = MemoryRegionInfo::eNo;
406 
407   const MinidumpMemoryInfo *next_entry = nullptr;
408   for (const auto &entry : mem_info_list) {
409     const auto head = entry->base_address;
410     const auto tail = head + entry->region_size;
411 
412     if (head <= load_addr && load_addr < tail) {
413       info.GetRange().SetRangeBase(
414           (entry->state != uint32_t(MinidumpMemoryInfoState::MemFree))
415               ? head
416               : load_addr);
417       info.GetRange().SetRangeEnd(tail);
418 
419       const uint32_t PageNoAccess =
420           static_cast<uint32_t>(MinidumpMemoryProtectionContants::PageNoAccess);
421       info.SetReadable((entry->protect & PageNoAccess) == 0 ? yes : no);
422 
423       const uint32_t PageWritable =
424           static_cast<uint32_t>(MinidumpMemoryProtectionContants::PageWritable);
425       info.SetWritable((entry->protect & PageWritable) != 0 ? yes : no);
426 
427       const uint32_t PageExecutable = static_cast<uint32_t>(
428           MinidumpMemoryProtectionContants::PageExecutable);
429       info.SetExecutable((entry->protect & PageExecutable) != 0 ? yes : no);
430 
431       const uint32_t MemFree =
432           static_cast<uint32_t>(MinidumpMemoryInfoState::MemFree);
433       info.SetMapped((entry->state != MemFree) ? yes : no);
434 
435       return info;
436     } else if (head > load_addr &&
437                (next_entry == nullptr || head < next_entry->base_address)) {
438       // In case there is no region containing load_addr keep track of the
439       // nearest region after load_addr so we can return the distance to it.
440       next_entry = entry;
441     }
442   }
443 
444   // No containing region found. Create an unmapped region that extends to the
445   // next region or LLDB_INVALID_ADDRESS
446   info.GetRange().SetRangeBase(load_addr);
447   info.GetRange().SetRangeEnd((next_entry != nullptr) ? next_entry->base_address
448                                                       : LLDB_INVALID_ADDRESS);
449   info.SetReadable(no);
450   info.SetWritable(no);
451   info.SetExecutable(no);
452   info.SetMapped(no);
453 
454   // Note that the memory info list doesn't seem to contain ranges in kernel
455   // space, so if you're walking a stack that has kernel frames, the stack may
456   // appear truncated.
457   return info;
458 }
459