[llvm-nm] Introduce synthetic flag

Compatible with GNU nm --syntethic option is used to show special
symbols created by the linker. Current implementation is limited to show
plt entries in the form of symbol@plt and plt entry address. Currently
it would be used for BOLT testing purposes
(#135867)
in order to eliminate external GNU nm dependency.

Author: yota9

llvm/docs/CommandGuide/llvm-nm.rst

@@ -257,6 +257,11 @@ OPTIONS
 
  Do not filter special symbols from the output.
 
+.. option:: --synthetic
+ Include synthetic symbols in the output. Synthetic symbols are not present in
+ the original object file but are generated by the tool to represent relevant
+ runtime or linking information.
+
 .. option:: --undefined-only, -u
 
  Print only undefined symbols.

llvm/test/tools/llvm-nm/AArch64/synthetic.test

@@ -0,0 +1,90 @@
+## Test --synthetic flag.
+
+# RUN: yaml2obj %s -o %t
+
+# RUN: llvm-nm %t | count 0
+# RUN: llvm-nm %t --synthetic | FileCheck %s
+
+--- !ELF
+FileHeader:
+  Class:           ELFCLASS64
+  Data:            ELFDATA2LSB
+  Type:            ET_DYN
+  Machine:         EM_AARCH64
+Sections:
+  - Name:            .rela.plt
+    Type:            SHT_RELA
+    Flags:           [ SHF_ALLOC, SHF_INFO_LINK ]
+    Address:         0x540
+    Link:            .dynsym
+    AddressAlign:    0x8
+    Info:            .got
+    Relocations:
+      - Offset:          0x10FA8
+        Symbol:          __libc_start_main
+        Type:            R_AARCH64_JUMP_SLOT
+      - Offset:          0x10FB0
+        Symbol:          __cxa_finalize
+        Type:            R_AARCH64_JUMP_SLOT
+      - Offset:          0x10FB8
+        Symbol:          __gmon_start__
+        Type:            R_AARCH64_JUMP_SLOT
+      - Offset:          0x10FC0
+        Symbol:          abort
+        Type:            R_AARCH64_JUMP_SLOT
+      - Offset:          0x10FC8
+        Symbol:          puts
+        Type:            R_AARCH64_JUMP_SLOT
+  - Name:            .plt
+    Type:            SHT_PROGBITS
+    Flags:           [ SHF_ALLOC, SHF_EXECINSTR ]
+    Address:         0x5D0
+    AddressAlign:    0x10
+    Content:         F07BBFA99000009011D247F910823E9120021FD61F2003D51F2003D51F2003D59000009011D647F910A23E9120021FD69000009011DA47F910C23E9120021FD69000009011DE47F910E23E9120021FD69000009011E247F910023F9120021FD69000009011E647F910223F9120021FD6
+  - Name:            .text
+    Type:            SHT_PROGBITS
+    Flags:           [ SHF_ALLOC, SHF_EXECINSTR ]
+    Address:         0x640
+  - Name:            .dynamic
+    Type:            SHT_DYNAMIC
+    Flags:           [ SHF_WRITE, SHF_ALLOC ]
+    Address:         0x10DA0
+    Link:            .dynstr
+    AddressAlign:    0x8
+    Entries:
+      - Tag:             DT_JMPREL
+        Value:           0x540
+      - Tag:             DT_PLTRELSZ
+        Value:           0x78
+      - Tag:             DT_PLTGOT
+        Value:           0x10F90
+      - Tag:             DT_NULL
+        Value:           0x0
+  - Name:            .got
+    Type:            SHT_PROGBITS
+    Flags:           [ SHF_WRITE, SHF_ALLOC ]
+    Address:         0x10F90
+    AddressAlign:    0x8
+    EntSize:         0x8
+    Content:         000000000000000000000000000000000000000000000000D005000000000000D005000000000000D005000000000000D005000000000000D005000000000000A00D01000000000000000000000000000000000000000000000000000000000054070000000000000000000000000000
+DynamicSymbols:
+  - Name:            __libc_start_main
+    Type:            STT_FUNC
+    Binding:         STB_GLOBAL
+  - Name:            __cxa_finalize
+    Type:            STT_FUNC
+    Binding:         STB_WEAK
+  - Name:            __gmon_start__
+    Binding:         STB_WEAK
+  - Name:            abort
+    Type:            STT_FUNC
+    Binding:         STB_GLOBAL
+  - Name:            puts
+    Type:            STT_FUNC
+    Binding:         STB_GLOBAL
+
+# CHECK:      0000000000000600 T __cxa_finalize@plt
+# CHECK-NEXT: 0000000000000610 T __gmon_start__@plt
+# CHECK-NEXT: 00000000000005f0 T __libc_start_main@plt
+# CHECK-NEXT: 0000000000000620 T abort@plt
+# CHECK-NEXT: 0000000000000630 T puts@plt

llvm/tools/llvm-nm/Opts.td

@@ -35,6 +35,7 @@ defm radix : Eq<"radix", "Radix (o/d/x) for printing symbol Values">, MetaVarNam
 def reverse_sort : FF<"reverse-sort", "Sort in reverse order">;
 def size_sort : FF<"size-sort", "Sort symbols by size">;
 def special_syms : FF<"special-syms", "Do not filter special symbols from the output">;
+def synthetic_syms : FF<"synthetic", "Include synthetic symbols in the output.  These are special symbols created by the linker for various purposes">;
 def undefined_only : FF<"undefined-only", "Show only undefined symbols">;
 def version : FF<"version", "Display the version">;
 def without_aliases : FF<"without-aliases", "Exclude aliases from output">, Flags<[HelpHidden]>;

llvm/tools/llvm-nm/llvm-nm.cpp

@@ -23,6 +23,7 @@
 #include "llvm/Demangle/Demangle.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/LLVMContext.h"
+#include "llvm/MC/TargetRegistry.h"
 #include "llvm/Object/Archive.h"
 #include "llvm/Object/COFF.h"
 #include "llvm/Object/COFFImportFile.h"
@@ -110,6 +111,7 @@ static bool PrintSize;
 static bool Quiet;
 static bool ReverseSort;
 static bool SpecialSyms;
+static bool SyntheticSyms;
 static bool SizeSort;
 static bool UndefinedOnly;
 static bool WithoutAliases;
@@ -1783,6 +1785,55 @@ getDynamicSyms(SymbolicFile &Obj) {
   return E->getDynamicSymbolIterators();
 }
 
+// Returns false if there is error found or true otherwise.
+static bool getPltSyms(SymbolicFile &Obj, std::vector<NMSymbol> &SymbolList) {
+  const auto *ELFObj = dyn_cast<ELFObjectFileBase>(&Obj);
+  if (!ELFObj)
+    return true;
+
+  std::string Err;
+  Triple TT;
+  TT.setArch(ELFObj->getArch());
+  TT.setOS(ELFObj->getOS());
+  const Target *TheTarget = TargetRegistry::lookupTarget(TT, Err);
+  if (!TheTarget) {
+    error("unable to find target for " + Obj.getFileName() + ": " + Err);
+    return false;
+  }
+
+  std::unique_ptr<const MCSubtargetInfo> STI;
+  STI.reset(TheTarget->createMCSubtargetInfo(
+      TT.getTriple(), ELFObj->tryGetCPUName().value_or("").str(), ""));
+  if (!STI) {
+    error("unable to create subtarget info for " + Obj.getFileName());
+    return false;
+  }
+
+  for (auto Plt : ELFObj->getPltEntries(*STI)) {
+    if (Plt.Symbol) {
+      SymbolRef Symbol(*Plt.Symbol, ELFObj);
+      if (Expected<StringRef> NameOrErr = Symbol.getName()) {
+        if (!NameOrErr->empty()) {
+          NMSymbol S = {};
+          S.Address = Plt.Address;
+          S.Name = NameOrErr->str() + "@plt";
+          S.TypeChar = 'T';
+          S.SectionName = Plt.Section;
+          SymbolList.push_back(S);
+        }
+      } else {
+        consumeError(NameOrErr.takeError());
+      }
+    } else {
+      WithColor::warning(errs(), ToolName)
+          << "PLT entry at 0x" + Twine::utohexstr(Plt.Address)
+          << " references an invalid symbol";
+    }
+  }
+
+  return true;
+}
+
 // Returns false if there is error found or true otherwise.
 static bool getSymbolNamesFromObject(SymbolicFile &Obj,
                                      std::vector<NMSymbol> &SymbolList) {
@@ -1807,6 +1858,12 @@ static bool getSymbolNamesFromObject(SymbolicFile &Obj,
             << toString(VersionsOrErr.takeError()) << "\n";
     }
   }
+
+  if (SyntheticSyms) {
+    if (!getPltSyms(Obj, SymbolList))
+      return false;
+  }
+
   // If a "-s segname sectname" option was specified and this is a Mach-O
   // file get the section number for that section in this object file.
   unsigned int Nsect = 0;
@@ -2474,6 +2531,7 @@ int llvm_nm_main(int argc, char **argv, const llvm::ToolContext &) {
           "(hexadecimal)");
   SizeSort = Args.hasArg(OPT_size_sort);
   SpecialSyms = Args.hasArg(OPT_special_syms);
+  SyntheticSyms = Args.hasArg(OPT_synthetic_syms);
   UndefinedOnly = Args.hasArg(OPT_undefined_only);
   WithoutAliases = Args.hasArg(OPT_without_aliases);