[LoongArch] Fixup value adjustment in applyFixup

[wangleiat]

Summary

A complete implementation of applyFixup for D132323.

Makes LoongArchAsmBackend::shouldForceRelocation to determine
if the relocation types must be forced.

This patch also adds range and alignment checks for b* instructions'
operands, at which point the offset to a label is known.

Depends on D132633

[xen0n]

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp

139–143

This is placed here due to review comment on D132323, does the situation change enough to make this change appropriate?

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchMCCodeEmitter.cpp

251

Do we want to also handle BL here? Given it's the only other insn of the format.

[wangleiat]

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp

139–143

Personal preference, just put the processing of Value together (Increases the readability of the code). Maybe it was put in the front at the time because of performance? If you think this is not good, I can modify it back.

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchMCCodeEmitter.cpp

251

Do we want to also handle BL here? Given it's the only other insn of the format.

It is not needed. This is because BL are handled specifically in both the LowerCall phase and ASMParser.

[xen0n]

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp

139–143

Granted it's just a teensy bit of performance, but it's still @SixWeining's input after all, and after the change it's back to the same as RISCV, and effectively code churn for no objective reason. So I'd personally prefer the 2 lines be moved back. Thanks!

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchMCCodeEmitter.cpp

251

Okay, thanks for the explanation. I have no problem with that.

[wangleiat]

Address @xen0n's comments.

[xen0n]

Looks reasonable now. Thanks!

[MaskRay]

Can some use FirstLiteralRelocationKind + ... instead?

[wangleiat]

In D132818#3760264, @MaskRay wrote:

Can some use FirstLiteralRelocationKind + ... instead?

Great! Will omit a lot of code, I will do that.

To avoid a lot of code removal, I'm going to make changes in the commits it depends on D132633

[wangleiat]

Address @MaskRay's comment.

[wangleiat]

rebase

[wangleiat]

rebase.

[MaskRay]

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp

93

% 4. The compiler can optimize it

100

% 4

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchMCCodeEmitter.cpp

267

break;

llvm/test/MC/LoongArch/Relocations/fixups-diagnostics.s

2

|& => 2>&1 |

Some people's default shell (e.g. some macOS) don't support |&

4

#@LINE

@LINE is deprecated

[wangleiat]

Thanks for all the comments, I will fix them. :)

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp

93

% 4. The compiler can optimize it

100

% 4

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchMCCodeEmitter.cpp

267

break;

llvm/test/MC/LoongArch/Relocations/fixups-diagnostics.s

2

|& => 2>&1 |

Some people's default shell (e.g. some macOS) don't support |&

4

#@LINE

@LINE is deprecated

[wangleiat]

Address @MaskRay's comments.

[wangleiat]

Fix [[@LINE]] what was missed.

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp

 //===-- LoongArchAsmBackend.cpp - LoongArch Assembler Backend -*- C++ -*---===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements the LoongArchAsmBackend class.
 //
 //===----------------------------------------------------------------------===//
 
 #include "LoongArchAsmBackend.h"
+#include "LoongArchFixupKinds.h"
 #include "llvm/MC/MCAsmLayout.h"
 #include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCELFObjectWriter.h"
+#include "llvm/MC/MCValue.h"
 #include "llvm/Support/Endian.h"
 #include "llvm/Support/EndianStream.h"
 
 #define DEBUG_TYPE "loongarch-asmbackend"
 
 using namespace llvm;
 
 Optional<MCFixupKind> LoongArchAsmBackend::getFixupKind(StringRef Name) const {
LoongArchAsmBackend::getFixupKindInfo(MCFixupKind Kind) const {
   if (Kind < FirstTargetFixupKind)
     return MCAsmBackend::getFixupKindInfo(Kind);
 
   assert(unsigned(Kind - FirstTargetFixupKind) < getNumFixupKinds() &&
          "Invalid kind!");
   return Infos[Kind - FirstTargetFixupKind];
 }
 
+static uint64_t adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
+                                 MCContext &Ctx) {
+  switch (Fixup.getTargetKind()) {
+  default:
+    llvm_unreachable("Unknown fixup kind");
+  case FK_Data_1:
+  case FK_Data_2:
+  case FK_Data_4:
+  case FK_Data_8:
+    return Value;
+  case LoongArch::fixup_loongarch_b16: {
+    if (!isInt<18>(Value))
+      Ctx.reportError(Fixup.getLoc(), "fixup value out of range");
+    if (Value % 4)
+      Ctx.reportError(Fixup.getLoc(), "fixup value must be 4-byte aligned");
+    return (Value >> 2) & 0xffff;
+  }
+  case LoongArch::fixup_loongarch_b21: {
+    if (!isInt<23>(Value))
+      Ctx.reportError(Fixup.getLoc(), "fixup value out of range");
+    if (Value % 4)
+      Ctx.reportError(Fixup.getLoc(), "fixup value must be 4-byte aligned");
+    return ((Value & 0x3fffc) << 8) | ((Value >> 18) & 0x1f);
+  }
+  case LoongArch::fixup_loongarch_b26: {
+    if (!isInt<28>(Value))
+      Ctx.reportError(Fixup.getLoc(), "fixup value out of range");
+    if (Value % 4)
+      Ctx.reportError(Fixup.getLoc(), "fixup value must be 4-byte aligned");
+    return ((Value & 0x3fffc) << 8) | ((Value >> 18) & 0x3ff);
+  }
+  case LoongArch::fixup_loongarch_abs_hi20:
+  case LoongArch::fixup_loongarch_tls_le_hi20:
+    return (Value >> 12) & 0xfffff;
+  case LoongArch::fixup_loongarch_abs_lo12:
+  case LoongArch::fixup_loongarch_tls_le_lo12:
+    return Value & 0xfff;
+  case LoongArch::fixup_loongarch_abs64_lo20:
+  case LoongArch::fixup_loongarch_tls_le64_lo20:
+    return (Value >> 32) & 0xfffff;
+  case LoongArch::fixup_loongarch_abs64_hi12:
+  case LoongArch::fixup_loongarch_tls_le64_hi12:
+    return (Value >> 52) & 0xfff;
+  }
+}
+
 void LoongArchAsmBackend::applyFixup(const MCAssembler &Asm,
                                      const MCFixup &Fixup,
                                      const MCValue &Target,
                                      MutableArrayRef<char> Data, uint64_t Value,
                                      bool IsResolved,
                                      const MCSubtargetInfo *STI) const {
   if (!Value)
     return; // Doesn't change encoding.
 
   MCFixupKind Kind = Fixup.getKind();
   if (Kind >= FirstLiteralRelocationKind)
     return;
   MCFixupKindInfo Info = getFixupKindInfo(Kind);
-  // TODO: Apply any target-specific value adjustments.
+  MCContext &Ctx = Asm.getContext();
+
+  // Apply any target-specific value adjustments.
+  Value = adjustFixupValue(Fixup, Value, Ctx);
 
   // Shift the value into position.
   Value <<= Info.TargetOffset;
 
   unsigned Offset = Fixup.getOffset();
   unsigned NumBytes = alignTo(Info.TargetSize + Info.TargetOffset, 8) / 8;
 
   assert(Offset + NumBytes <= Data.size() && "Invalid fixup offset!");
   // For each byte of the fragment that the fixup touches, mask in the
   // bits from the fixup value.
   for (unsigned I = 0; I != NumBytes; ++I) {
     Data[Offset + I] |= uint8_t((Value >> (I * 8)) & 0xff);
   }
 }
 
 bool LoongArchAsmBackend::shouldForceRelocation(const MCAssembler &Asm,
                                                 const MCFixup &Fixup,
                                                 const MCValue &Target) {
   if (Fixup.getKind() >= FirstLiteralRelocationKind)
     return true;
-  // TODO: Determine which relocation require special processing at linking
-  // time.
+  switch (Fixup.getTargetKind()) {
+  default:
     return false;
+  case FK_Data_1:
+  case FK_Data_2:
+  case FK_Data_4:
+  case FK_Data_8:
+    return !Target.isAbsolute();
+  }
 }
 
 bool LoongArchAsmBackend::writeNopData(raw_ostream &OS, uint64_t Count,
                                        const MCSubtargetInfo *STI) const {
   // Check for byte count not multiple of instruction word size
   if (Count % 4 != 0)
     return false;
 

llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchMCCodeEmitter.cpp

case LoongArchMCExpr::VK_LoongArch_TLS_GD_HI20:
       FixupKind = LoongArch::fixup_loongarch_tls_gd_hi20;
       break;
     }
   } else if (Kind == MCExpr::SymbolRef &&
              cast<MCSymbolRefExpr>(Expr)->getKind() ==
                  MCSymbolRefExpr::VK_None) {
     switch (MI.getOpcode()) {
     default:
       break;
     case LoongArch::BEQ:
     case LoongArch::BNE:
     case LoongArch::BLT:
     case LoongArch::BGE:
     case LoongArch::BLTU:
     case LoongArch::BGEU:
       FixupKind = LoongArch::fixup_loongarch_b16;
       break;
     case LoongArch::BEQZ:
     case LoongArch::BNEZ:
     case LoongArch::BCEQZ:
     case LoongArch::BCNEZ:
       FixupKind = LoongArch::fixup_loongarch_b21;
       break;
+    case LoongArch::B:
+      FixupKind = LoongArch::fixup_loongarch_b26;
+      break;
     }
   }
 
   assert(FixupKind != LoongArch::fixup_loongarch_invalid &&
          "Unhandled expression!");
 
   Fixups.push_back(
       MCFixup::create(0, Expr, MCFixupKind(FixupKind), MI.getLoc()));

llvm/test/MC/LoongArch/Relocations/fixups-diagnostics.s

+# RUN: not llvm-mc --triple=loongarch64 --filetype=obj %s -o /dev/null 2>&1 | FileCheck %s
+
+  beq $a0, $a1, far_distant # CHECK: :[[#@LINE]]:3: error: fixup value out of range
+  bne $a0, $a1, unaligned # CHECK: :[[#@LINE]]:3: error: fixup value must be 4-byte aligned
+
+  bnez $a0, unaligned # CHECK: :[[#@LINE]]:3: error: fixup value must be 4-byte aligned
+  beqz $a0, far_distant_bz # CHECK: :[[#@LINE]]:3: error: fixup value out of range
+
+  b unaligned # CHECK: :[[#@LINE]]:3: error: fixup value must be 4-byte aligned
+
+  .byte 0
+unaligned:
+  .byte 0
+  .byte 0
+  .byte 0
+
+  .space 1<<16
+distant:
+  .space 1<<18
+far_distant:
+
+  .byte 0
+unaligned_bz:
+  .byte 0
+  .byte 0
+  .byte 0
+distant_bz:
+  .space 1<<23
+far_distant_bz:

llvm/test/MC/LoongArch/Relocations/fixups-invalid.s

+# RUN: not llvm-mc --filetype=obj %s --triple=loongarch32 -o /dev/null 2>&1 \
+# RUN:     | FileCheck %s
+# RUN: not llvm-mc --filetype=obj %s --triple=loongarch64 -o /dev/null 2>&1 \
+# RUN:     | FileCheck %s
+
+.byte foo   # CHECK: [[#@LINE]]:7: error: 1-byte data relocations not supported
+.2byte foo  # CHECK: [[#@LINE]]:8: error: 2-byte data relocations not supported

llvm/test/MC/LoongArch/Relocations/fixups.s

+# RUN: llvm-mc --triple=loongarch64 %s --show-encoding \
+# RUN:     | FileCheck --check-prefix=CHECK-FIXUP %s
+# RUN: llvm-mc --filetype=obj --triple=loongarch64 %s \
+# RUN:     | llvm-objdump -d - | FileCheck --check-prefix=CHECK-INSTR %s
+# RUN: llvm-mc --filetype=obj --triple=loongarch64 %s \
+# RUN:     | llvm-readobj -r - | FileCheck --check-prefix=CHECK-REL %s
+
+## Checks that fixups that can be resolved within the same object file are
+## applied correctly.
+
+.LBB0:
+lu12i.w $t1, %abs_hi20(val)
+# CHECK-FIXUP: fixup A - offset: 0, value: %abs_hi20(val), kind: fixup_loongarch_abs_hi20
+# CHECK-INSTR: lu12i.w $t1, 74565
+
+ori $t1, $t1, %abs_lo12(val)
+# CHECK-FIXUP: fixup A - offset: 0, value: %abs_lo12(val), kind: fixup_loongarch_abs_lo12
+# CHECK-INSTR: ori $t1, $t1, 1656
+
+b .LBB0
+# CHECK-FIXUP: fixup A - offset: 0, value: .LBB0, kind: fixup_loongarch_b26
+# CHECK-INSTR: b -8
+b .LBB2
+# CHECK-FIXUP: fixup A - offset: 0, value: .LBB2, kind: fixup_loongarch_b26
+# CHECK-INSTR: b 331004
+beq $a0, $a1, .LBB0
+# CHECK-FIXUP: fixup A - offset: 0, value: .LBB0, kind: fixup_loongarch_b16
+# CHECK-INSTR: beq $a0, $a1, -16
+blt $a0, $a1, .LBB1
+# CHECK-FIXUP: fixup A - offset: 0, value: .LBB1, kind: fixup_loongarch_b16
+# CHECK-INSTR: blt $a0, $a1, 1116
+beqz $a0, .LBB0
+# CHECK-FIXUP: fixup A - offset: 0, value: .LBB0, kind: fixup_loongarch_b21
+# CHECK-INSTR: beqz $a0, -24
+bnez $a0, .LBB1
+# CHECK-FIXUP: fixup A - offset: 0, value: .LBB1, kind: fixup_loongarch_b21
+# CHECK-INSTR: bnez $a0, 1108
+
+.fill 1104
+
+.LBB1:
+
+.fill 329876
+nop
+.LBB2:
+
+.set val, 0x12345678
+
+# CHECK-REL-NOT: R_LARCH
+
+## Testing the function call offset could resolved by assembler
+## when the function and the callsite within the same compile unit.
+func:
+.fill 100
+bl func
+# CHECK-FIXUP: fixup A - offset: 0, value: func, kind: fixup_loongarch_b26
+# CHECK-INSTR: bl -100
+
+.fill 10000
+bl func
+# CHECK-FIXUP: fixup A - offset: 0, value: func, kind: fixup_loongarch_b26
+# CHECK-INSTR: bl -10104
+
+.fill 20888
+bl func
+# CHECK-FIXUP: fixup A - offset: 0, value: func, kind: fixup_loongarch_b26
+# CHECK-INSTR: bl -30996