[elfabi] Add support for reading DT_NEEDED from binaries

[amontanez]

Summary

This patch gives elfabi the ability to read DT_NEEDED entries from ELF binaries to populate NeededLibs in TextAPI's ELFStub.

[jakehehrlich]

Other than me realizing I made a mistake in previous reviews and a minor little issue (NeededLibCount), this LGTM.

llvm/tools/llvm-elfabi/ELFObjHandler.cpp

132

You never use the NeededLibCount. Why do you have it?

llvm/tools/llvm-elfabi/ELFObjHandler.h

31 ↗

(On Diff #178779)

ditto (below)

37 ↗

(On Diff #178779)

ditto (below)

45 ↗

(On Diff #178779)

ditto (below)

56 ↗

(On Diff #178779)

Just realized something I should have caught in previous reviews. You don't need these in the headers. You can make them static template functions in just the file they're used in. That reduces .o size and improves the compilers ability to perform optimizations.

69 ↗

(On Diff #178779)

Ditto here.

[ruiu]

llvm/test/tools/llvm-elfabi/binary-read-neededlibs.test

36

nit: remove extraneous space characters?

llvm/tools/llvm-elfabi/ELFObjHandler.cpp

140

Is NeededLibs empty on function entry? If so, you can return TargetStub.NeededLibs.count() and remove NeededLibCount variable.

[amontanez]

Rebase onto previous patch resolves most of the comments for this patch.

[ruiu]

llvm/tools/llvm-elfabi/ELFObjHandler.cpp

126

nit: I believe you don't need std::string as std::string has an implicit constructor from const char *.

[amontanez]

Rebase and use terminatedSubstr() to fetch DT_NEEDED strings from .dynstr string table.

[ruiu]

llvm/tools/llvm-elfabi/ELFObjHandler.cpp

36–37

Indentation should be 2 spaces.

126

I think all strings in a string table should be terminated with '\0' in any ELF file. If the last string "overruns" the string table, it is an error, and the last string should not automatically be terminated at the end of the string table.

[amontanez]

Rebase on D55629 change that addresses strings that overrun string table bounds.

[amontanez]

Rebase again.

[amontanez]

Rebase, update error message for DT_NEEDED entries that are outside the string table, and add another test.

[jakehehrlich]

LGTM

llvm/test/tools/llvm-elfabi/binary-read-neededlibs.test

+# RUN: yaml2obj %s > %t
+# RUN: llvm-elfabi %t --emit-tbe=- | FileCheck %s
+
+!ELF
+FileHeader:
+  Class:           ELFCLASS64
+  Data:            ELFDATA2LSB
+  Type:            ET_DYN
+  Machine:         EM_X86_64
+Sections:
+  - Name:            .dynamic
+    Type:            SHT_DYNAMIC
+    Flags:           [ SHF_ALLOC ]
+    Address:         0x1000
+    Content:         "010000000000000001000000000000000e00000000000000150000000000000001000000000000000b000000000000000a000000000000001c000000000000000500000000000000d00200000000000000000000000000000000000000000000"
+      # DT_NEEDED     1 (0x01)
+      # DT_SONAME     21 (0x15)
+      # DT_NEEDED     11 (0x0b)
+      # DT_STRSZ      36 (0x24)
+      # DT_STRTAB     0x2b0
+      # DT_NULL       0x0
+    Size:            96
+  - Name:            .dynsym
+    Type:            SHT_DYNSYM
+    Flags:           [ SHF_ALLOC ]
+    Address:         0x2000
+    EntSize:         24
+    Size:            0
+    Link:            .dynstr
+  - Name:            .dynstr
+    Type:            SHT_STRTAB
+    Flags:           [ SHF_ALLOC ]
+    Address:         0x3000
+    #                "\0libfoo.so\0libbar.so\0somelib.so\0foo\0"
+    Content:         "006c6962666f6f2e736f006c69626261722e736f00736f6d656c69622e736f00666f6f00"
+ProgramHeaders:
+  - Type: PT_LOAD
+    VAddr: 0x0000
+    PAddr: 0x0000
+  - Type: PT_DYNAMIC
+    Flags: [ PF_X, PF_R ]
+    VAddr: 0x1000
+    PAddr: 0x1000
+    Sections:
+      - Section: .dynamic
+
+# CHECK:      --- !tapi-tbe
+# CHECK-NEXT: TbeVersion: {{[1-9]\d*\.(0|([1-9]\d*))}}
+# CHECK-NEXT: SoName: somelib.so
+# CHECK-NEXT: Arch: x86_64
+# CHECK-NEXT: NeededLibs:
+# CHECK-NEXT:   - libfoo.so
+# CHECK-NEXT:   - libbar.so
+# CHECK-NEXT: Symbols: {}
+# CHECK-NEXT: ...

llvm/tools/llvm-elfabi/ELFObjHandler.cpp

 
 namespace {
 
 // Simple struct to hold relevant .dynamic entries.
 struct DynamicEntries {
   Optional<uint64_t> StrTabOffset;
   Optional<uint64_t> StrSize;
   Optional<uint64_t> SONameOffset;
+  std::vector<uint64_t> NeededLibNames;
 };
 
 }
 
 namespace llvm {
 namespace elfabi {
 
 /// This function populates a DynamicEntries struct using a binary
 /// ELFT::DynRange. After populating the struct, the members are validated with
 /// some basic sanity checks.
 ///
case DT_SONAME:
       Dyn.SONameOffset = Entry.d_un.d_val;
       break;
     case DT_STRTAB:
       Dyn.StrTabOffset = Entry.d_un.d_ptr;
       break;
     case DT_STRSZ:
       Dyn.StrSize = Entry.d_un.d_val;
       break;
+    case DT_NEEDED:
+      Dyn.NeededLibNames.push_back(Entry.d_un.d_val);
+      break;
     }
   }
 
   if (!Dyn.StrTabOffset.hasValue()) {
     return createError(
         "Couldn't locate dynamic string table (no DT_STRTAB entry)");
   }
   if (!Dyn.StrSize.hasValue()) {
     return createError(
         "Couldn't determine dynamic string table size (no DT_STRSZ entry)");
   }
   if (Dyn.SONameOffset.hasValue() && *Dyn.SONameOffset >= *Dyn.StrSize) {
     return createError(
         "DT_SONAME string offset outside of dynamic string table");
   }
+  for (uint64_t Offset : Dyn.NeededLibNames) {
+    if (Offset >= *Dyn.StrSize){
+      return createError(
+          "DT_NEEDED string offset outside of dynamic string table");
+    }
+  }
 
   return Error::success();
 }
 
 /// Returns a new ELFStub with all members populated from an ELFObjectFile.
 /// @param ElfObj Source ELFObjectFile.
 template <class ELFT>
 Expected<std::unique_ptr<ELFStub>>
static buildStub(const ELFObjectFile<ELFT> &ElfObj) {
   // Populate Arch from ELF header.
   DestStub->Arch = ElfFile->getHeader()->e_machine;
 
   // Populate SoName from .dynamic entries and dynamic string table.
   if (DynEnt.SONameOffset.hasValue()) {
     DestStub->SoName = std::string(DynStrPtr + *DynEnt.SONameOffset);
   }
 
-  // TODO: Populate NeededLibs from .dynamic entries and linked string table.
+  // Populate NeededLibs from .dynamic entries and dynamic string table.
+  for (uint64_t NeededStrOffset : DynEnt.NeededLibNames) {
+    DestStub->NeededLibs.push_back(std::string(DynStrPtr + NeededStrOffset));
+  }
+
   // TODO: Populate Symbols from .dynsym table and linked string table.
 
   return std::move(DestStub);
 }
 
 Expected<std::unique_ptr<ELFStub>> readELFFile(MemoryBufferRef Buf) {
   Expected<std::unique_ptr<Binary>> BinOrErr = createBinary(Buf);
   if (!BinOrErr) {
     return BinOrErr.takeError();
   }
 
   Binary *Bin = BinOrErr->get();
   if (auto Obj = dyn_cast<ELFObjectFile<ELF32LE>>(Bin)) {
     return buildStub(*Obj);
   } else if (auto Obj = dyn_cast<ELFObjectFile<ELF64LE>>(Bin)) {
     return buildStub(*Obj);
   } else if (auto Obj = dyn_cast<ELFObjectFile<ELF32BE>>(Bin)) {
     return buildStub(*Obj);
   } else if (auto Obj = dyn_cast<ELFObjectFile<ELF64BE>>(Bin)) {
     return buildStub(*Obj);
   }
 
   return createStringError(errc::not_supported, "Unsupported binary format");
 }
 
 } // end namespace elfabi
 } // end namespace llvm