llvm-symbolizer: Untag addresses in object files by default.

Any addresses that we pass to llvm-symbolizer are going to be untagged,
while any HWASAN instrumented globals are going to be tagged in the
symbol table. Therefore we need to untag the addresses before using them.

Differential Revision: https://reviews.llvm.org/D65769

llvm-svn: 367926

include/llvm/DebugInfo/Symbolize/Symbolize.h

@@ -39,6 +39,7 @@ public:
     bool UseSymbolTable = true;
     bool Demangle = true;
     bool RelativeAddresses = false;
+    bool UntagAddresses = false;
     std::string DefaultArch;
     std::vector<std::string> DsymHints;
     std::string FallbackDebugPath;

lib/DebugInfo/Symbolize/SymbolizableObjectFile.cpp

@@ -43,10 +43,11 @@ getDILineInfoSpecifier(FunctionNameKind FNKind) {
 
 ErrorOr<std::unique_ptr<SymbolizableObjectFile>>
 SymbolizableObjectFile::create(const object::ObjectFile *Obj,
-                               std::unique_ptr<DIContext> DICtx) {
+                               std::unique_ptr<DIContext> DICtx,
+                               bool UntagAddresses) {
   assert(DICtx);
   std::unique_ptr<SymbolizableObjectFile> res(
-      new SymbolizableObjectFile(Obj, std::move(DICtx)));
+      new SymbolizableObjectFile(Obj, std::move(DICtx), UntagAddresses));
   std::unique_ptr<DataExtractor> OpdExtractor;
   uint64_t OpdAddress = 0;
   // Find the .opd (function descriptor) section if any, for big-endian
@@ -103,8 +104,10 @@ SymbolizableObjectFile::create(const object::ObjectFile *Obj,
 }
 
 SymbolizableObjectFile::SymbolizableObjectFile(const ObjectFile *Obj,
-                                               std::unique_ptr<DIContext> DICtx)
-    : Module(Obj), DebugInfoContext(std::move(DICtx)) {}
+                                               std::unique_ptr<DIContext> DICtx,
+                                               bool UntagAddresses)
+    : Module(Obj), DebugInfoContext(std::move(DICtx)),
+      UntagAddresses(UntagAddresses) {}
 
 namespace {
 
@@ -172,6 +175,11 @@ std::error_code SymbolizableObjectFile::addSymbol(const SymbolRef &Symbol,
   if (!SymbolAddressOrErr)
     return errorToErrorCode(SymbolAddressOrErr.takeError());
   uint64_t SymbolAddress = *SymbolAddressOrErr;
+  if (UntagAddresses) {
+    // For kernel addresses, bits 56-63 need to be set, so we sign extend bit 55
+    // into bits 56-63 instead of masking them out.
+    SymbolAddress = (int64_t(SymbolAddress) << 8) >> 8;
+  }
   if (OpdExtractor) {
     // For big-endian PowerPC64 ELF, symbols in the .opd section refer to
     // function descriptors. The first word of the descriptor is a pointer to

lib/DebugInfo/Symbolize/SymbolizableObjectFile.h

@@ -31,7 +31,8 @@ namespace symbolize {
 class SymbolizableObjectFile : public SymbolizableModule {
 public:
   static ErrorOr<std::unique_ptr<SymbolizableObjectFile>>
-  create(const object::ObjectFile *Obj, std::unique_ptr<DIContext> DICtx);
+  create(const object::ObjectFile *Obj, std::unique_ptr<DIContext> DICtx,
+         bool UntagAddresses);
 
   DILineInfo symbolizeCode(object::SectionedAddress ModuleOffset,
                            FunctionNameKind FNKind,
@@ -70,6 +71,7 @@ private:
 
   const object::ObjectFile *Module;
   std::unique_ptr<DIContext> DebugInfoContext;
+  bool UntagAddresses;
 
   struct SymbolDesc {
     uint64_t Addr;
@@ -85,7 +87,8 @@ private:
   std::vector<std::pair<SymbolDesc, StringRef>> Objects;
 
   SymbolizableObjectFile(const object::ObjectFile *Obj,
-                         std::unique_ptr<DIContext> DICtx);
+                         std::unique_ptr<DIContext> DICtx,
+                         bool UntagAddresses);
 };
 
 } // end namespace symbolize

lib/DebugInfo/Symbolize/Symbolize.cpp

@@ -418,8 +418,8 @@ Expected<SymbolizableModule *>
 LLVMSymbolizer::createModuleInfo(const ObjectFile *Obj,
                                  std::unique_ptr<DIContext> Context,
                                  StringRef ModuleName) {
-  auto InfoOrErr =
-      SymbolizableObjectFile::create(Obj, std::move(Context));
+  auto InfoOrErr = SymbolizableObjectFile::create(Obj, std::move(Context),
+                                                  Opts.UntagAddresses);
   std::unique_ptr<SymbolizableModule> SymMod;
   if (InfoOrErr)
     SymMod = std::move(*InfoOrErr);

test/tools/llvm-symbolizer/untag-addresses.test

@@ -0,0 +1,16 @@
+# RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o
+# RUN: echo DATA %t.o 0 | llvm-symbolizer | FileCheck --check-prefix=UNTAG %s
+# RUN: echo DATA %t.o 0 | llvm-symbolizer -untag-addresses=0 | FileCheck --check-prefix=NOUNTAG %s
+# RUN: echo DATA %t.o 0 | llvm-addr2line | FileCheck --check-prefix=NOUNTAG %s
+
+# UNTAG: foo
+# UNTAG: 0 4
+# NOUNTAG: ??
+# NOUNTAG: 0 0
+
+.data
+.globl foo
+.type foo, @object
+.size foo, 4
+foo = . + 0x1100000000000000
+.4byte 1

tools/llvm-symbolizer/llvm-symbolizer.cpp

@@ -55,6 +55,10 @@ static cl::opt<bool>
                          cl::desc("Interpret addresses as relative addresses"),
                          cl::ReallyHidden);
 
+static cl::opt<bool> ClUntagAddresses(
+    "untag-addresses", cl::init(true),
+    cl::desc("Remove memory tags from addresses before symbolization"));
+
 static cl::opt<bool>
     ClPrintInlining("inlining", cl::init(true),
                     cl::desc("Print all inlined frames for a given address"));
@@ -274,6 +278,7 @@ int main(int argc, char **argv) {
     ClDemangle.setInitialValue(false);
     ClPrintFunctions.setInitialValue(FunctionNameKind::None);
     ClPrintInlining.setInitialValue(false);
+    ClUntagAddresses.setInitialValue(false);
     ClOutputStyle.setInitialValue(DIPrinter::OutputStyle::GNU);
   }
 
@@ -290,6 +295,7 @@ int main(int argc, char **argv) {
   Opts.UseSymbolTable = ClUseSymbolTable;
   Opts.Demangle = ClDemangle;
   Opts.RelativeAddresses = ClUseRelativeAddress;
+  Opts.UntagAddresses = ClUntagAddresses;
   Opts.DefaultArch = ClDefaultArch;
   Opts.FallbackDebugPath = ClFallbackDebugPath;
   Opts.DWPName = ClDwpName;