feat(asm): add .debug_frame CFI section for stack unwinding

Assisted-by: MiMo V2.5 Pro

asm/elf_dwarf.go

@@ -63,6 +63,7 @@ type dwarfSections struct {
 	debugInfo    []byte
 	debugLine    []byte
 	debugLineStr []byte
+	debugFrame   []byte
 	// Relocations for .debug_info: (offset, symbol name, addend).
 	infoRelocs []dwarfReloc
 	// Relocations for .debug_line: (offset, symbol name, addend).
@@ -90,6 +91,9 @@ func emitDWARF(img *Image, srcFile string) *dwarfSections {
 
 	// Build .debug_info.
 	ds.debugInfo = dwarfBuildInfoSection(img, srcFile, ds)
+
+	// Build .debug_frame.
+	ds.debugFrame = dwarfBuildFrameSection(img)
 	return ds
 }
 
@@ -248,3 +252,68 @@ func dwarfBuildInfoSection(img *Image, srcFile string, ds *dwarfSections) []byte
 func appendUleb(b []byte, v uint64) []byte {
 	return binary.AppendUvarint(b, v)
 }
+
+func appendSleb(b []byte, v int64) []byte {
+	return binary.AppendVarint(b, v)
+}
+
+// dwarfBuildFrameSection builds a .debug_frame section with CFI for stack
+// unwinding.  It emits one CIE and one FDE per function, encoding the
+// CFA (Canonical Frame Address) rule changes at each stack-adjustment
+// boundary recorded in FuncLayout.Spadj.
+func dwarfBuildFrameSection(img *Image) []byte {
+	var b []byte
+	le := binary.LittleEndian
+
+	// CIE (Common Information Entry).
+	cieStart := len(b)
+	b = append(b, 0, 0, 0, 0)          // length (placeholder)
+	b = le.AppendUint32(b, 0xFFFFFFFF) // CIE marker
+	b = append(b, 3)                   // version (DWARF3, widely supported)
+	b = append(b, 0)                   // augmentation (empty)
+	b = appendUleb(b, 1)               // code alignment
+	b = appendSleb(b, -8)              // data alignment (-8 for 64-bit)
+	b = appendUleb(b, 16)              // return address register (LR on arm64, RIP on amd64)
+	// Initial CFA rule: DW_CFA_def_cfa (SP, 0)
+	b = append(b, 0x0c)   // DW_CFA_def_cfa
+	b = appendUleb(b, 31) // register: SP (RSP=7 on amd64, SP=31 on arm64)
+	b = appendUleb(b, 0)  // offset: 0
+	b = append(b, 0)      // DW_CFA_nop (padding)
+	// Patch CIE length.
+	le.PutUint32(b[cieStart:], uint32(len(b)-cieStart-4))
+
+	// FDEs (Frame Description Entries) — one per function.
+	for _, fn := range img.Funcs {
+		fdeStart := len(b)
+		b = append(b, 0, 0, 0, 0)                // length (placeholder)
+		b = le.AppendUint32(b, uint32(cieStart)) // CIE pointer (offset from start)
+		// Initial location: function offset in .text (relocated by linker).
+		b = le.AppendUint64(b, uint64(fn.Offset))
+		// Address range: function size.
+		b = le.AppendUint64(b, uint64(fn.Size))
+
+		// Emit CFA rule changes at each Spadj boundary.
+		for _, step := range fn.Spadj {
+			if step.Value == 0 {
+				continue
+			}
+			// DW_CFA_def_cfa_offset: set CFA = SP + |delta|.
+			// The delta is negative (stack grows down), so CFA offset = -delta.
+			offset := -step.Value
+			if offset > 0 {
+				b = append(b, 0x0e) // DW_CFA_def_cfa_offset
+				b = appendUleb(b, uint64(offset))
+			}
+		}
+
+		// Pad to alignment.
+		for len(b)%4 != 0 {
+			b = append(b, 0) // DW_CFA_nop
+		}
+
+		// Patch FDE length.
+		le.PutUint32(b[fdeStart:], uint32(len(b)-fdeStart-4))
+	}
+
+	return b
+}