[Lldb-commits] [lldb] Add support for reading the dynamic symbol table from PT_DYNAMIC (PR #112596)

[Jacob Lalonde via lldb-commits]

================
@@ -3895,3 +3924,103 @@ std::optional<lldb_private::DataExtractor> ObjectFileELF::GetDynamicData() {
   }
   return std::nullopt;
 }
+
+
+std::optional<DataExtractor>
+ObjectFileELF::GetDynsymDataFromDynamic(uint32_t &num_symbols) {
+  // Every ELF file which represents an executable or shared library has
+  // mandatory .dynamic entries. The DT_SYMTAB value contains a pointer to the
+  // symbol table, and DT_SYMENT contains the size of a symbol table entry.
+  // We then can use either the DT_HASH or DT_GNU_HASH to find the number of
+  // symbols in the symbol table as the symbol count is not stored in the
+  // .dynamic section as a key/value pair.
+  //
+  // When loading and ELF file from memory, only the program headers end up
+  // being mapped into memory, and we can find these values in the PT_DYNAMIC
+  // segment.
+  num_symbols = 0;
+  // Get the process in case this is an in memory ELF file.
+  ProcessSP process_sp(m_process_wp.lock());
+  const ELFDynamic *symtab = FindDynamicSymbol(DT_SYMTAB);
+  const ELFDynamic *syment = FindDynamicSymbol(DT_SYMENT);
+  const ELFDynamic *hash = FindDynamicSymbol(DT_HASH);
+  const ELFDynamic *gnu_hash = FindDynamicSymbol(DT_GNU_HASH);
+  // DT_SYMTAB and DT_SYMENT are mandatory.
+  if (symtab == nullptr || syment == nullptr)
+    return std::nullopt;
+  // We must have either a DT_HASH or a DT_GNU_HASH.
+  if (hash == nullptr && gnu_hash == nullptr)
+    return std::nullopt;
+  // The number of symbols in the symbol table is the number of entries in the
+  // symbol table divided by the size of each symbol table entry.
+  // We must figure out the number of symbols in the symbol table using the
+  // DT_HASH or the DT_GNU_HASH as the number of symbols isn't stored anywhere
+  // in the .dynamic section.
+
+  lldb::offset_t offset;
+  if (hash) {
+    // The DT_HASH header contains the number of symbols in the "nchain"
+    // member. The header looks like this:
+    // struct DT_HASH_HEADER {
+    //   uint32_t nbucket;
+    //   uint32_t nchain;
+    // };
+    if (auto data = ReadDataFromDynamic(hash, 8)) {
+      offset = 4;
----------------
Jlalond wrote:

Should we make a comment why we're ignoring the first field

https://github.com/llvm/llvm-project/pull/112596