[X86] Avoid generating invalid R_X86_64_GOTPCRELX relocations

[hvdijk]

Summary

We need to make sure not to emit R_X86_64_GOTPCRELX relocations for instructions that use a REX prefix. If a REX prefix is present, we need to instead use a R_X86_64_REX_GOTPCRELX relocation. The existing logic for CALL64m, JMP64m, etc. already handles this by checking the HasREX parameter and using it to determine which relocation type to use. Do this for all instructions that can use relaxed relocations.

[MaskRay]

llvm/test/MC/X86/gotpcrelx.s

125

LG.

May I bother you to delete ELF/got-relaxed-rex.s and add its test coverage into this file as well?

[MaskRay]

The lld/test/ELF tests need fixing (check-lld-elf)

[hvdijk]

Merge ELF/got-relaxed-rex.s into X86/gotpcrelx.s, update lld/test/ELF/x86-64-gotpc-relax-nopic.s.

[hvdijk]

In D93561#2463891, @MaskRay wrote:

The lld/test/ELF tests need fixing (check-lld-elf)

Good spot, it was a single test that needed updating, done.

[MaskRay]

LGTM.

llvm/test/MC/X86/gotpcrelx.s

17

Consider moving add R_X86_64_REX_GOTPCRELX instructions below R_X86_64_REX_GOTPCRELX ones.

[hvdijk]

llvm/test/MC/X86/gotpcrelx.s

17

I'm assuming you meant below R_X86_64_GOTPCRELX :) Sure, pushed with that change.

lld/test/ELF/x86-64-gotpc-relax-nopic.s

 # DISASM-NEXT: <_start>:
 # DISASM-NEXT:   2011c8:       adcl  {{.*}}(%rip), %eax  # 202240
 # DISASM-NEXT:                 addl  {{.*}}(%rip), %ebx  # 202240
 # DISASM-NEXT:                 andl  {{.*}}(%rip), %ecx  # 202240
 # DISASM-NEXT:                 cmpl  {{.*}}(%rip), %edx  # 202240
 # DISASM-NEXT:                 orl   {{.*}}(%rip), %edi  # 202240
 # DISASM-NEXT:                 sbbl  {{.*}}(%rip), %esi  # 202240
 # DISASM-NEXT:                 subl  {{.*}}(%rip), %ebp  # 202240
-# DISASM-NEXT:                 xorl  {{.*}}(%rip), %r8d  # 202240
-# DISASM-NEXT:                 testl %r15d, {{.*}}(%rip) # 202240
+# DISASM-NEXT:                 xorl  $0x203248, %r8d
+# DISASM-NEXT:                 testl $0x203248, %r15d
 # DISASM-NEXT:   201200:       adcq  $0x203248, %rax
 # DISASM-NEXT:                 addq  $0x203248, %rbx
 # DISASM-NEXT:                 andq  $0x203248, %rcx
 # DISASM-NEXT:                 cmpq  $0x203248, %rdx
 # DISASM-NEXT:                 orq   $0x203248, %rdi
 # DISASM-NEXT:                 sbbq  $0x203248, %rsi
 # DISASM-NEXT:                 subq  $0x203248, %rbp
 # DISASM-NEXT:                 xorq  $0x203248, %r8

llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp

unsigned FixupKind = [&]() {
       if (!(Disp.isExpr() && isa<MCSymbolRefExpr>(Disp.getExpr())))
         return X86::reloc_riprel_4byte;
 
       // Certain loads for GOT references can be relocated against the symbol
       // directly if the symbol ends up in the same linkage unit.
       switch (Opcode) {
       default:
         return X86::reloc_riprel_4byte;
+      case X86::MOV64rm:
+        // movq loads is a subset of reloc_riprel_4byte_relax_rex. It is a
+        // special case because COFF and Mach-O don't support ELF's more
+        // flexible R_X86_64_REX_GOTPCRELX relaxation.
+        assert(HasREX);
+        return X86::reloc_riprel_4byte_movq_load;
       case X86::ADC32rm:
       case X86::ADD32rm:
       case X86::AND32rm:
       case X86::CMP32rm:
       case X86::MOV32rm:
       case X86::OR32rm:
       case X86::SBB32rm:
       case X86::SUB32rm:
       case X86::TEST32mr:
       case X86::XOR32rm:
-        return X86::reloc_riprel_4byte_relax;
-      case X86::MOV64rm:
-        // movq loads is a subset of reloc_riprel_4byte_relax_rex. It is a
-        // special case because COFF and Mach-O don't support ELF's more
-        // flexible R_X86_64_REX_GOTPCRELX relaxation.
-        assert(HasREX);
-        return X86::reloc_riprel_4byte_movq_load;
       case X86::CALL64m:
       case X86::JMP64m:
       case X86::TAILJMPm64:
       case X86::TEST64mr:
       case X86::ADC64rm:
       case X86::ADD64rm:
       case X86::AND64rm:
       case X86::CMP64rm:

llvm/test/MC/ELF/got-relaxed-rex.s

-// RUN: llvm-mc -filetype=obj -triple x86_64-pc-linux %s -o - | llvm-readobj -r - | FileCheck %s
-
-// these should produce R_X86_64_REX_GOTPCRELX
-
-        movq mov@GOTPCREL(%rip), %rax
-        test %rax, test@GOTPCREL(%rip)
-        adc adc@GOTPCREL(%rip), %rax
-        add add@GOTPCREL(%rip), %rax
-        and and@GOTPCREL(%rip), %rax
-        cmp cmp@GOTPCREL(%rip), %rax
-        or  or@GOTPCREL(%rip), %rax
-        sbb sbb@GOTPCREL(%rip), %rax
-        sub sub@GOTPCREL(%rip), %rax
-        xor xor@GOTPCREL(%rip), %rax
-
-.section .norelax,"ax"
-## This expression loads the GOT entry with an offset.
-## Don't emit R_X86_64_REX_GOTPCRELX.
-        movq mov@GOTPCREL+1(%rip), %rax
-
-// CHECK:      Relocations [
-// CHECK-NEXT:   Section ({{.*}}) .rela.text {
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX mov
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX test
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX adc
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX add
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX and
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX cmp
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX or
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX sbb
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX sub
-// CHECK-NEXT:     R_X86_64_REX_GOTPCRELX xor
-// CHECK-NEXT:   }
-// CHECK-NEXT:   Section ({{.*}}) .rela.norelax {
-// CHECK-NEXT:     R_X86_64_GOTPCREL mov
-// CHECK-NEXT:   }

llvm/test/MC/X86/gotpcrelx.s

 # RUN: llvm-mc -filetype=obj -triple=x86_64 %s -o %t.o
 # RUN: llvm-readobj -r %t.o | FileCheck %s --check-prefixes=CHECK,COMMON
 # RUN: llvm-mc -filetype=obj -triple=x86_64 -relax-relocations=false %s -o %t1.o
 # RUN: llvm-readobj -r %t1.o | FileCheck %s --check-prefixes=NORELAX,COMMON
 
 # COMMON:     Relocations [
 # COMMON-NEXT:  Section ({{.*}}) .rela.text {
 # CHECK-NEXT:     R_X86_64_GOTPCRELX mov
 # CHECK-NEXT:     R_X86_64_GOTPCRELX test
 # CHECK-NEXT:     R_X86_64_GOTPCRELX adc
 # CHECK-NEXT:     R_X86_64_GOTPCRELX add
 # CHECK-NEXT:     R_X86_64_GOTPCRELX and
 # CHECK-NEXT:     R_X86_64_GOTPCRELX cmp
 # CHECK-NEXT:     R_X86_64_GOTPCRELX or
 # CHECK-NEXT:     R_X86_64_GOTPCRELX sbb
 # CHECK-NEXT:     R_X86_64_GOTPCRELX sub
 # CHECK-NEXT:     R_X86_64_GOTPCRELX xor
 # CHECK-NEXT:     R_X86_64_GOTPCRELX call
 # CHECK-NEXT:     R_X86_64_GOTPCRELX jmp
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX mov
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX test
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX adc
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX add
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX and
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX cmp
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX or
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX sbb
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX sub
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX xor
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX mov
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX test
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX adc
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX add
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX and
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX cmp
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX or
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX sbb
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX sub
+# CHECK-NEXT:     R_X86_64_REX_GOTPCRELX xor
 # CHECK-NEXT:   }
 
 # NORELAX-NEXT:     R_X86_64_GOTPCREL mov
 # NORELAX-NEXT:     R_X86_64_GOTPCREL test
 # NORELAX-NEXT:     R_X86_64_GOTPCREL adc
 # NORELAX-NEXT:     R_X86_64_GOTPCREL add
 # NORELAX-NEXT:     R_X86_64_GOTPCREL and
 # NORELAX-NEXT:     R_X86_64_GOTPCREL cmp
 # NORELAX-NEXT:     R_X86_64_GOTPCREL or
 # NORELAX-NEXT:     R_X86_64_GOTPCREL sbb
 # NORELAX-NEXT:     R_X86_64_GOTPCREL sub
 # NORELAX-NEXT:     R_X86_64_GOTPCREL xor
 # NORELAX-NEXT:     R_X86_64_GOTPCREL call
 # NORELAX-NEXT:     R_X86_64_GOTPCREL jmp
+# NORELAX-NEXT:     R_X86_64_GOTPCREL mov
+# NORELAX-NEXT:     R_X86_64_GOTPCREL test
+# NORELAX-NEXT:     R_X86_64_GOTPCREL adc
+# NORELAX-NEXT:     R_X86_64_GOTPCREL add
+# NORELAX-NEXT:     R_X86_64_GOTPCREL and
+# NORELAX-NEXT:     R_X86_64_GOTPCREL cmp
+# NORELAX-NEXT:     R_X86_64_GOTPCREL or
+# NORELAX-NEXT:     R_X86_64_GOTPCREL sbb
+# NORELAX-NEXT:     R_X86_64_GOTPCREL sub
+# NORELAX-NEXT:     R_X86_64_GOTPCREL xor
+# NORELAX-NEXT:     R_X86_64_GOTPCREL mov
+# NORELAX-NEXT:     R_X86_64_GOTPCREL test
+# NORELAX-NEXT:     R_X86_64_GOTPCREL adc
+# NORELAX-NEXT:     R_X86_64_GOTPCREL add
+# NORELAX-NEXT:     R_X86_64_GOTPCREL and
+# NORELAX-NEXT:     R_X86_64_GOTPCREL cmp
+# NORELAX-NEXT:     R_X86_64_GOTPCREL or
+# NORELAX-NEXT:     R_X86_64_GOTPCREL sbb
+# NORELAX-NEXT:     R_X86_64_GOTPCREL sub
+# NORELAX-NEXT:     R_X86_64_GOTPCREL xor
 # NORELAX-NEXT:   }
 
 movl mov@GOTPCREL(%rip), %eax
 test %eax, test@GOTPCREL(%rip)
 adc adc@GOTPCREL(%rip), %eax
 add add@GOTPCREL(%rip), %eax
 and and@GOTPCREL(%rip), %eax
 cmp cmp@GOTPCREL(%rip), %eax
 or  or@GOTPCREL(%rip), %eax
 sbb sbb@GOTPCREL(%rip), %eax
 sub sub@GOTPCREL(%rip), %eax
 xor xor@GOTPCREL(%rip), %eax
 call *call@GOTPCREL(%rip)
 jmp *jmp@GOTPCREL(%rip)
 
+movl mov@GOTPCREL(%rip), %r8d
+test %r8d, test@GOTPCREL(%rip)
+adc adc@GOTPCREL(%rip), %r8d
+add add@GOTPCREL(%rip), %r8d
+and and@GOTPCREL(%rip), %r8d
+cmp cmp@GOTPCREL(%rip), %r8d
+or  or@GOTPCREL(%rip), %r8d
+sbb sbb@GOTPCREL(%rip), %r8d
+sub sub@GOTPCREL(%rip), %r8d
+xor xor@GOTPCREL(%rip), %r8d
+
+movq mov@GOTPCREL(%rip), %rax
+test %rax, test@GOTPCREL(%rip)
+adc adc@GOTPCREL(%rip), %rax
+add add@GOTPCREL(%rip), %rax
+and and@GOTPCREL(%rip), %rax
+cmp cmp@GOTPCREL(%rip), %rax
+or  or@GOTPCREL(%rip), %rax
+sbb sbb@GOTPCREL(%rip), %rax
+sub sub@GOTPCREL(%rip), %rax
+xor xor@GOTPCREL(%rip), %rax
+
 # COMMON-NEXT:   Section ({{.*}}) .rela.norelax {
 # COMMON-NEXT:     R_X86_64_GOTPCREL mov 0x0
+# COMMON-NEXT:     R_X86_64_GOTPCREL mov 0xFFFFFFFFFFFFFFFD
 # COMMON-NEXT:     R_X86_64_GOTPCREL mov 0xFFFFFFFFFFFFFFFC
 # COMMON-NEXT:   }
 # COMMON-NEXT: ]
 
 .section .norelax,"ax",@progbits
 ## Clang may emit this expression to load the high 32-bit of the GOT entry.
 ## Don't emit R_X86_64_GOTPCRELX.
 movl mov@GOTPCREL+4(%rip), %eax
+## Don't emit R_X86_64_GOTPCRELX.
+movq mov@GOTPCREL+1(%rip), %rax
 ## We could emit R_X86_64_GOTPCRELX, but it is probably unnecessary.
 movl mov@GOTPCREL+0(%rip), %eax