[lldb] Fix UB cast when encountering DW_LANG_* >= eNumLanguageTypes

[dsandersllvm]

LanguageType has two kinds of enumerators in it. The first is DWARF-assigned enumerators which must be consecutive and match DW_LANG values. The second is the vendor-assigned enumerators which must be unique and must follow on from the DWARF-assigned values (i.e. the first one is currently eLanguageTypeMojo

• 1. even if that collides with DWARF-assigned values that lldb is not yet aware of

Only the DWARF-assigned enumerators may be static_cast from DW_LANG since their values match. The vendor-assigned enumerators must be explicitly converted since their values do not match. This needs to handle new languages added to DWARF and not yet implemented in lldb.

This fixes a crash when encountering a DW_LANG value >= eNumLanguageTypes and wrong behaviour when encountering DW_LANG values that have not yet been added to LanguageType but happen to coincide with a vendor-assigned enumerator due to the consecutive values requirement described above.

Another way to fix the crash is to add the language to LanguageType (and fill any preceeding gaps in the number space) so that the DW_LANG being encountered is correctly handled but this just moves the problem to a new subset of DW_LANG values.

Also fix an unnecessary static-cast from LanguageType to LanguageType.

[llvmbot]

@llvm/pr-subscribers-lldb

Author: Daniel Sanders (dsandersllvm)

Changes

LanguageType has two kinds of enumerators in it. The first is DWARF-assigned enumerators which must be consecutive and match DW_LANG values. The second is the vendor-assigned enumerators which must be unique and must follow on from the DWARF-assigned values (i.e. the first one must currently be eLanguageTypeMojo

• 1. even if that collides with DWARF-assigned values that lldb is not yet aware of

Only the DWARF-assigned enumerators may be static_cast from DW_LANG since their values match. The vendor-assigned enumerators must be explicitly converted since their values do not match. This needs to handle new languages added to DWARF and not yet implemented in lldb.

This fixes a crash when encountering a DW_LANG value >= eNumLanguageTypes and wrong behaviour when encountering DW_LANG values that have not yet been added to LanguageType but happen to coincide with a vendor-assigned enumerator due to the consecutive values requirement described above.

Another way to fix the crash is to add the language to LanguageType (and fill any preceeding gaps in the number space) so that the DW_LANG being encountered is correctly handled but this just moves the problem to a new subset of DW_LANG values.

Also fix an unnecessary static-cast from LanguageType to LanguageType.

---

Full diff: https://github.com/llvm/llvm-project/pull/150132.diff

3 Files Affected:

• (modified) lldb/include/lldb/lldb-enumerations.h (+1)
• (modified) lldb/source/Plugins/SymbolFile/DWARF/SymbolFileDWARF.cpp (+4-1)
• (modified) lldb/source/Target/Language.cpp (+1-1)

diff --git a/lldb/include/lldb/lldb-enumerations.h b/lldb/include/lldb/lldb-enumerations.h
index e021c7a926cf1..7768b260785e3 100644
--- a/lldb/include/lldb/lldb-enumerations.h
+++ b/lldb/include/lldb/lldb-enumerations.h
@@ -518,6 +518,7 @@ enum LanguageType {
   eLanguageTypeAssembly = 0x0031,
   eLanguageTypeC_sharp = 0x0032,
   eLanguageTypeMojo = 0x0033,
+  eLanguageTypeLastStandardLanguage = eLanguageTypeMojo,
 
   // Vendor Extensions
   // Note: Language::GetNameForLanguageType
diff --git a/lldb/source/Plugins/SymbolFile/DWARF/SymbolFileDWARF.cpp b/lldb/source/Plugins/SymbolFile/DWARF/SymbolFileDWARF.cpp
index 5b16ce5f75138..69d397e1a86f7 100644
--- a/lldb/source/Plugins/SymbolFile/DWARF/SymbolFileDWARF.cpp
+++ b/lldb/source/Plugins/SymbolFile/DWARF/SymbolFileDWARF.cpp
@@ -4330,13 +4330,16 @@ SymbolFileDWARF::GetContainingDeclContext(const DWARFDIE &die) {
 }
 
 LanguageType SymbolFileDWARF::LanguageTypeFromDWARF(uint64_t val) {
+  if (val <= eLanguageTypeLastStandardLanguage)
+    return static_cast<LanguageType>(val);
+
   // Note: user languages between lo_user and hi_user must be handled
   // explicitly here.
   switch (val) {
   case DW_LANG_Mips_Assembler:
     return eLanguageTypeMipsAssembler;
   default:
-    return static_cast<LanguageType>(val);
+    return eLanguageTypeUnknown;
   }
 }
 
diff --git a/lldb/source/Target/Language.cpp b/lldb/source/Target/Language.cpp
index 86754c251cd93..484d9badde397 100644
--- a/lldb/source/Target/Language.cpp
+++ b/lldb/source/Target/Language.cpp
@@ -257,7 +257,7 @@ static uint32_t num_languages =
 LanguageType Language::GetLanguageTypeFromString(llvm::StringRef string) {
   for (const auto &L : language_names) {
     if (string.equals_insensitive(L.name))
-      return static_cast<LanguageType>(L.type);
+      return L.type;
   }
 
   return eLanguageTypeUnknown;