[ELF] Don't relax R_X86_64_GOTPCRELX if addend != -4

clang may produce `movl x@GOTPCREL+4(%rip), %eax` when loading the high 32 bits
of the address of a global variable in -fpic/-fpie mode.

If assembled by GNU as, the fixup emits an R_X86_64_GOTPCRELX with an
addend != -4. The instruction loads from the GOT entry with an offset
and thus it is incorrect to relax the instruction.

If assembled by the integrated assembler, we emit R_X86_64_GOTPCREL for
relocations that definitely cannot be relaxed (D92114), so this patch is not
needed.

This patch disables the relaxation, which is compatible with the implementation in GNU ld
("Add R_X86_64_[REX_]GOTPCRELX support to gas and ld").

Reviewed By: grimar, jhenderson

Differential Revision: https://reviews.llvm.org/D91993

lld/ELF/Arch/X86_64.cpp

@@ -730,7 +730,12 @@ void X86_64::relocate(uint8_t *loc, const Relocation &rel, uint64_t val) const {
 
 RelExpr X86_64::adjustGotPcExpr(RelType type, int64_t addend,
                                 const uint8_t *loc) const {
-  if (type != R_X86_64_GOTPCRELX && type != R_X86_64_REX_GOTPCRELX)
+  // Only R_X86_64_[REX_]GOTPCRELX can be relaxed. GNU as may emit GOTPCRELX
+  // with addend != -4. Such an instruction does not load the full GOT entry, so
+  // we cannot relax the relocation. E.g. movl x@GOTPCREL+4(%rip), %rax
+  // (addend=0) loads the high 32 bits of the GOT entry.
+  if ((type != R_X86_64_GOTPCRELX && type != R_X86_64_REX_GOTPCRELX) ||
+      addend != -4)
     return R_GOT_PC;
   const uint8_t op = loc[-2];
   const uint8_t modRm = loc[-1];

lld/test/ELF/x86-64-gotpc-offset.s

@@ -0,0 +1,30 @@
+# REQUIRES: x86
+# RUN: llvm-mc -filetype=obj -triple=x86_64 %s -o %t.o
+# RUN: ld.lld %t.o -o %t
+# RUN: llvm-objdump -s -d %t | FileCheck %s
+
+# CHECK:      Contents of section .got:
+# CHECK-NEXT: {{^}} [[#%x,ADDR:]] {{.*}} 00000000
+
+# CHECK:      leal {{.*}}(%rip), %eax  # {{.*}} <foo>
+# CHECK-NEXT: movl {{.*}}(%rip), %eax  # [[#ADDR+4]]
+# CHECK-NEXT: movq {{.*}}(%rip), %rax  # [[#ADDR+1]]
+
+## movl foo@GOTPCREL(%rip), %eax
+  movl 0(%rip), %eax
+  .reloc .-4, R_X86_64_GOTPCRELX, foo-4
+
+## The instruction has an offset (addend!=-4). It is incorrect to relax movl to leal.
+## movl foo@GOTPCREL+4(%rip), %eax
+  movl 0(%rip), %eax
+  .reloc .-4, R_X86_64_GOTPCRELX, foo
+
+## This does not make sense because it loads one byte past the GOT entry.
+## It is just to demonstrate the behavior.
+## movq foo@GOTPCREL+1(%rip), %rax
+  movq 0(%rip), %rax
+  .reloc .-4, R_X86_64_REX_GOTPCRELX, foo-3
+
+.globl foo
+foo:
+  nop