[BPF] Report an error if comparison imm operand cannot fit in 32bit

Ihor Solodrai reported a case ([1]) where gcc reports an error but clang
ignores that error and proceeds to generate incorrect code. More
specifically, the problematic code looks like:
  if r1 == 0xcafefeeddeadbeef goto <label>

Here, 0xcafefeeddeadbeef needs to be encoded in a 32-bit imm field
of the insns and the 32-bit imm allows sign extenstion to 64-bit imm.
Obviously, 0xcafefeeddeadbeef cannot encode properly.

The compilation failed for gcc with the following error:
  Error: immediate out of range, shall fit in 32 bits

Given a 64-bit imm value, converting to the proper 32-bit imm value
must satisfy the following 64-bit patterns:
  00000000 00000000 00000000 00000000 xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
  11111111 11111111 11111111 11111111 1xxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx

So if the top 32-bits is 0 or the top 33-bits is 0x1ffffffff, then the 64-bit imm
value can be truncated into proper 32-bit imm. Otherwise, a warning
message, the same as gcc will be issued. If -Werror is enabled during
compilation, the warning will turn into an error.

  [1] https://lore.kernel.org/bpf/[email protected]/

Author: Yonghong Song

llvm/lib/Target/BPF/MCTargetDesc/BPFMCCodeEmitter.cpp

@@ -14,6 +14,7 @@
 #include "MCTargetDesc/BPFMCTargetDesc.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/MC/MCCodeEmitter.h"
+#include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCFixup.h"
 #include "llvm/MC/MCInst.h"
@@ -33,11 +34,12 @@ namespace {
 class BPFMCCodeEmitter : public MCCodeEmitter {
   const MCRegisterInfo &MRI;
   bool IsLittleEndian;
+  MCContext &Ctx;
 
 public:
   BPFMCCodeEmitter(const MCInstrInfo &, const MCRegisterInfo &mri,
-                   bool IsLittleEndian)
-      : MRI(mri), IsLittleEndian(IsLittleEndian) { }
+                   bool IsLittleEndian, MCContext &ctx)
+      : MRI(mri), IsLittleEndian(IsLittleEndian), Ctx(ctx) {}
   BPFMCCodeEmitter(const BPFMCCodeEmitter &) = delete;
   void operator=(const BPFMCCodeEmitter &) = delete;
   ~BPFMCCodeEmitter() override = default;
@@ -67,12 +69,12 @@ class BPFMCCodeEmitter : public MCCodeEmitter {
 
 MCCodeEmitter *llvm::createBPFMCCodeEmitter(const MCInstrInfo &MCII,
                                             MCContext &Ctx) {
-  return new BPFMCCodeEmitter(MCII, *Ctx.getRegisterInfo(), true);
+  return new BPFMCCodeEmitter(MCII, *Ctx.getRegisterInfo(), true, Ctx);
 }
 
 MCCodeEmitter *llvm::createBPFbeMCCodeEmitter(const MCInstrInfo &MCII,
                                               MCContext &Ctx) {
-  return new BPFMCCodeEmitter(MCII, *Ctx.getRegisterInfo(), false);
+  return new BPFMCCodeEmitter(MCII, *Ctx.getRegisterInfo(), false, Ctx);
 }
 
 unsigned BPFMCCodeEmitter::getMachineOpValue(const MCInst &MI,
@@ -81,8 +83,16 @@ unsigned BPFMCCodeEmitter::getMachineOpValue(const MCInst &MI,
                                              const MCSubtargetInfo &STI) const {
   if (MO.isReg())
     return MRI.getEncodingValue(MO.getReg());
-  if (MO.isImm())
-    return static_cast<unsigned>(MO.getImm());
+  if (MO.isImm()) {
+    uint64_t Imm = MO.getImm();
+    uint64_t High32Bits = Imm >> 32, High33Bits = Imm >> 31;
+    if (MI.getOpcode() != BPF::LD_imm64 && High32Bits != 0 &&
+        High33Bits != 0x1FFFFFFFFULL) {
+      Ctx.reportWarning(MI.getLoc(),
+                        "immediate out of range, shall fit in 32 bits");
+    }
+    return static_cast<unsigned>(Imm);
+  }
 
   assert(MO.isExpr());
 

llvm/test/CodeGen/BPF/warn-cmp.ll

@@ -0,0 +1,15 @@
+; RUN: llc -mtriple=bpfel -filetype=obj < %s 2>&1 >/dev/null | FileCheck %s
+
+; CHECK: warning: immediate out of range, shall fit in 32 bits
+define dso_local void @test_1() naked {
+  tail call void asm sideeffect
+    "r1 = 40; if r1 == 0x1deadbeef goto +0; r0 = 0; exit;", "~{r0},~{r1}"()
+  unreachable
+}
+
+; CHECK: warning: immediate out of range, shall fit in 32 bits
+define dso_local void @test_2() naked {
+  tail call void asm sideeffect
+    "r1 = 40; if r1 == 0xffffffff00000000 goto +0; r0 = 0; exit;", "~{r0},~{r1}"()
+  unreachable
+}