[lld-macho] Implement GOT_LOAD relaxation

[int3]

Summary

We can have GOT_LOAD relocations that reference __dso_handle.
However, our binding opcode encoder doesn't support binding to the DSOHandle
symbol. Instead of adding support for that, I decided it would be cleaner to
implement GOT_LOAD relaxation since __dso_handle's location is always
statically known.

[smeenai]

LGTM

lld/MachO/Arch/X86_64.cpp

void X86_64::writeStubHelperEntry(uint8_t *buf, const DylibSymbol &sym,
   write32le(buf + 1, sym.lazyBindOffset);
   writeRipRelative(buf, entryAddr, sizeof(stubHelperEntry),
                    in.stubHelper->addr);
 }
 
 void X86_64::prepareSymbolRelocation(lld::macho::Symbol *sym,
                                      const InputSection *isec, const Reloc &r) {
   switch (r.type) {
-  case X86_64_RELOC_GOT_LOAD:
-    // TODO: implement mov -> lea relaxation for non-dynamic symbols
+  case X86_64_RELOC_GOT_LOAD: {
+    if (sym->isWeakDef() || isa<DylibSymbol>(sym))
+      in.got->addEntry(sym);
+
+    if (sym->isTlv())
+      error("found GOT relocation referencing thread-local variable in " +
+            toString(isec));
+    break;
+  }
   case X86_64_RELOC_GOT: {
     in.got->addEntry(sym);
 
     if (sym->isTlv())
       error("found GOT relocation referencing thread-local variable in " +
             toString(isec));
     break;
   }
void X86_64::prepareSymbolRelocation(lld::macho::Symbol *sym,
   default:
     llvm_unreachable("unexpected relocation type");
   }
 }
 
 uint64_t X86_64::resolveSymbolVA(uint8_t *buf, const lld::macho::Symbol &sym,
                                  uint8_t type) const {
   switch (type) {
-  case X86_64_RELOC_GOT_LOAD:
+  case X86_64_RELOC_GOT_LOAD: {
+    if (!sym.isInGot()) {
+      if (buf[-2] != 0x8b)
+        error("X86_64_RELOC_GOT_LOAD must be used with movq instructions");
+      buf[-2] = 0x8d;
+      return sym.getVA();
+    }
+    LLVM_FALLTHROUGH;
+  }
   case X86_64_RELOC_GOT:
     return in.got->addr + sym.gotIndex * WordSize;
   case X86_64_RELOC_BRANCH: {
     if (sym.isInStubs())
       return in.stubs->addr + sym.stubsIndex * sizeof(stub);
     return sym.getVA();
   }
   case X86_64_RELOC_UNSIGNED:

lld/test/MachO/local-got.s

_main:
   movl $0x2000004, %eax # write() syscall
   mov $1, %rdi # stdout
   movq _hello_its_me@GOTPCREL(%rip), %rsi
   mov $15, %rdx # length of str
   syscall
 
   movl $0x2000004, %eax # write() syscall
   mov $1, %rdi # stdout
-  movq _hello_world@GOTPCREL(%rip), %rsi
+## We use pushq/popq here instead of movq in order to avoid relaxation.
+  pushq _hello_world@GOTPCREL(%rip)
+  popq %rsi
   mov $13, %rdx # length of str
   syscall
 
   movl $0x2000004, %eax # write() syscall
   mov $1, %rdi # stdout
-  movq _goodbye_world@GOTPCREL(%rip), %rsi
+  pushq _goodbye_world@GOTPCREL(%rip)
+  popq %rsi
   mov $15, %rdx # length of str
   syscall
 
   mov $0, %rax
   ret
 
 .section __TEXT,__cstring
 _hello_world:
   .asciz "Hello world!\n"
 
 _goodbye_world:
   .asciz "Goodbye world!\n"

lld/test/MachO/x86-64-reloc-got-load.s

+# REQUIRES: x86
+
+## Check that we perform relaxation for GOT_LOAD relocations to defined symbols.
+## Note: GOT_LOAD relocations to dylib symbols are already tested in dylink.s.
+
+# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-darwin %s -o %t.o
+# RUN: lld -flavor darwinnew -o %t %t.o
+# RUN: llvm-objdump -d --no-show-raw-insn %t | FileCheck %s
+# CHECK: leaq [[#]](%rip), %rax  # {{.*}} <_foo>
+
+.globl _main, _foo
+
+_main:
+  movq _foo@GOTPCREL(%rip), %rax
+  ret
+
+_foo:
+  .space 0