Prevent unaligned memory read in parseMinidumpString

[teemperor]

Summary

It's possible to hit an unaligned memory read when reading source_length as the data array is only aligned with 2 bytes (it's actually a UTF16 array). This patch memcpy's source_length into a local variable to prevent this:

MinidumpTypes.cpp:49:23: runtime error: load of misaligned address 0x7f0f4792692a for type 'const uint32_t' (aka 'const unsigned int'), which requires 4 byte alignment

[davide]

I'll be happy to review all your patch set tomorrow, Raphael. Do you mind to add me as reviewer so I don't lose track of your work?

[teemperor]

Thanks, I think only this and the first GoParser patch ( D42339 ) needs some double checking, the others ( D42346, D42338 ) are trivial and I'll merge in the next few hours as zturner suggested.

[davide]

LGTM. Did you run the testsuite? If it's clean, feel free to merge.

lldb/trunk/source/Plugins/Process/minidump/MinidumpTypes.cpp

const MinidumpHeader *MinidumpHeader::Parse(llvm::ArrayRef<uint8_t> &data) {
   return header;
 }
 
 // Minidump string
 llvm::Optional<std::string>
 lldb_private::minidump::parseMinidumpString(llvm::ArrayRef<uint8_t> &data) {
   std::string result;
 
-  const uint32_t *source_length;
-  Status error = consumeObject(data, source_length);
-  if (error.Fail() || *source_length > data.size() || *source_length % 2 != 0)
+  const uint32_t *source_length_ptr;
+  Status error = consumeObject(data, source_length_ptr);
+
+  // Copy non-aligned source_length data into aligned memory.
+  uint32_t source_length;
+  std::memcpy(&source_length, source_length_ptr, sizeof(source_length));
+
+  if (error.Fail() || source_length > data.size() || source_length % 2 != 0)
     return llvm::None;
 
   auto source_start = reinterpret_cast<const llvm::UTF16 *>(data.data());
   // source_length is the length of the string in bytes
   // we need the length of the string in UTF-16 characters/code points (16 bits
   // per char)
   // that's why it's divided by 2
-  const auto source_end = source_start + (*source_length) / 2;
+  const auto source_end = source_start + source_length / 2;
   // resize to worst case length
-  result.resize(UNI_MAX_UTF8_BYTES_PER_CODE_POINT * (*source_length) / 2);
+  result.resize(UNI_MAX_UTF8_BYTES_PER_CODE_POINT * source_length / 2);
   auto result_start = reinterpret_cast<llvm::UTF8 *>(&result[0]);
   const auto result_end = result_start + result.size();
   llvm::ConvertUTF16toUTF8(&source_start, source_end, &result_start, result_end,
                            llvm::strictConversion);
   const auto result_size =
       std::distance(reinterpret_cast<llvm::UTF8 *>(&result[0]), result_start);
   result.resize(result_size); // shrink to actual length