cmd/objdump: use cmd/internal/objfile

This removes a bunch of ugly duplicate code.
The end goal is to factor the disassembly code
into cmd/internal/objfile too, so that pprof can use it,
but one step at a time.

LGTM=r, iant
R=r, alex.brainman, iant
CC=golang-codereviews
https://golang.org/cl/149400043

src/cmd/internal/objfile/elf.go

@@ -8,6 +8,7 @@ package objfile
 
 import (
 	"debug/elf"
+	"fmt"
 	"os"
 )
 
@@ -77,3 +78,27 @@ func (f *elfFile) pcln() (textStart uint64, symtab, pclntab []byte, err error) {
 	}
 	return textStart, symtab, pclntab, nil
 }
+
+func (f *elfFile) text() (textStart uint64, text []byte, err error) {
+	sect := f.elf.Section(".text")
+	if sect == nil {
+		return 0, nil, fmt.Errorf("text section not found")
+	}
+	textStart = sect.Addr
+	text, err = sect.Data()
+	return
+}
+
+func (f *elfFile) goarch() string {
+	switch f.elf.Machine {
+	case elf.EM_386:
+		return "386"
+	case elf.EM_X86_64:
+		return "amd64"
+	case elf.EM_ARM:
+		return "arm"
+	case elf.EM_PPC64:
+		return "power64"
+	}
+	return ""
+}

src/cmd/internal/objfile/goobj.go

@@ -79,3 +79,15 @@ func (f *goobjFile) symbols() ([]Sym, error) {
 func (f *goobjFile) pcln() (textStart uint64, symtab, pclntab []byte, err error) {
 	return 0, nil, nil, fmt.Errorf("pcln not available in go object file")
 }
+
+// text does not make sense for Go object files, because
+// each function has a separate section.
+func (f *goobjFile) text() (textStart uint64, text []byte, err error) {
+	return 0, nil, fmt.Errorf("text not available in go object file")
+}
+
+// goarch makes sense but is not exposed in debug/goobj's API,
+// and we don't need it yet for any users of internal/objfile.
+func (f *goobjFile) goarch() string {
+	return "GOARCH unimplemented for debug/goobj files"
+}

src/cmd/internal/objfile/macho.go

@@ -85,6 +85,30 @@ func (f *machoFile) pcln() (textStart uint64, symtab, pclntab []byte, err error)
 	return textStart, symtab, pclntab, nil
 }
 
+func (f *machoFile) text() (textStart uint64, text []byte, err error) {
+	sect := f.macho.Section("__text")
+	if sect == nil {
+		return 0, nil, fmt.Errorf("text section not found")
+	}
+	textStart = sect.Addr
+	text, err = sect.Data()
+	return
+}
+
+func (f *machoFile) goarch() string {
+	switch f.macho.Cpu {
+	case macho.Cpu386:
+		return "386"
+	case macho.CpuAmd64:
+		return "amd64"
+	case macho.CpuArm:
+		return "arm"
+	case macho.CpuPpc64:
+		return "power64"
+	}
+	return ""
+}
+
 type uint64s []uint64
 
 func (x uint64s) Len() int           { return len(x) }

src/cmd/internal/objfile/objfile.go

@@ -14,6 +14,8 @@ import (
 type rawFile interface {
 	symbols() (syms []Sym, err error)
 	pcln() (textStart uint64, symtab, pclntab []byte, err error)
+	text() (textStart uint64, text []byte, err error)
+	goarch() string
 }
 
 // A File is an opened executable file.
@@ -70,3 +72,11 @@ func (f *File) PCLineTable() (*gosym.Table, error) {
 	}
 	return gosym.NewTable(symtab, gosym.NewLineTable(pclntab, textStart))
 }
+
+func (f *File) Text() (uint64, []byte, error) {
+	return f.raw.text()
+}
+
+func (f *File) GOARCH() string {
+	return f.raw.goarch()
+}

src/cmd/internal/objfile/pe.go

@@ -133,6 +133,25 @@ func (f *peFile) pcln() (textStart uint64, symtab, pclntab []byte, err error) {
 	return textStart, symtab, pclntab, nil
 }
 
+func (f *peFile) text() (textStart uint64, text []byte, err error) {
+	var imageBase uint64
+	switch oh := f.pe.OptionalHeader.(type) {
+	case *pe.OptionalHeader32:
+		imageBase = uint64(oh.ImageBase)
+	case *pe.OptionalHeader64:
+		imageBase = oh.ImageBase
+	default:
+		return 0, nil, fmt.Errorf("pe file format not recognized")
+	}
+	sect := f.pe.Section(".text")
+	if sect == nil {
+		return 0, nil, fmt.Errorf("text section not found")
+	}
+	textStart = imageBase + uint64(sect.VirtualAddress)
+	text, err = sect.Data()
+	return
+}
+
 func findPESymbol(f *pe.File, name string) (*pe.Symbol, error) {
 	for _, s := range f.Symbols {
 		if s.Name != name {
@@ -168,3 +187,15 @@ func loadPETable(f *pe.File, sname, ename string) ([]byte, error) {
 	}
 	return data[ssym.Value:esym.Value], nil
 }
+
+func (f *peFile) goarch() string {
+	// Not sure how to get the info we want from PE header.
+	// Look in symbol table for telltale rt0 symbol.
+	if _, err := findPESymbol(f.pe, "_rt0_386_windows"); err == nil {
+		return "386"
+	}
+	if _, err := findPESymbol(f.pe, "_rt0_amd64_windows"); err == nil {
+		return "amd64"
+	}
+	return ""
+}

src/cmd/internal/objfile/plan9obj.go

@@ -88,6 +88,16 @@ func (f *plan9File) pcln() (textStart uint64, symtab, pclntab []byte, err error)
 	return textStart, symtab, pclntab, nil
 }
 
+func (f *plan9File) text() (textStart uint64, text []byte, err error) {
+	sect := f.plan9.Section("text")
+	if sect == nil {
+		return 0, nil, fmt.Errorf("text section not found")
+	}
+	textStart = f.plan9.LoadAddress + f.plan9.HdrSize
+	text, err = sect.Data()
+	return
+}
+
 func findPlan9Symbol(f *plan9obj.File, name string) (*plan9obj.Sym, error) {
 	syms, err := f.Symbols()
 	if err != nil {
@@ -122,3 +132,15 @@ func loadPlan9Table(f *plan9obj.File, sname, ename string) ([]byte, error) {
 	textStart := f.LoadAddress + f.HdrSize
 	return data[ssym.Value-textStart : esym.Value-textStart], nil
 }
+
+func (f *plan9File) goarch() string {
+	switch f.plan9.Magic {
+	case plan9obj.Magic386:
+		return "386"
+	case plan9obj.MagicAMD64:
+		return "amd64"
+	case plan9obj.MagicARM:
+		return "arm"
+	}
+	return ""
+}

src/cmd/objdump/Makefile

-all: x86.go armasm.go
-
-x86.go: bundle
-	./bundle -p main -x x86_ rsc.io/x86/x86asm | gofmt >x86.go
-
-armasm.go: bundle
-	./bundle -p main -x arm_ rsc.io/arm/armasm | gofmt >armasm.go
-
-bundle:
-	go build -o bundle code.google.com/p/rsc/cmd/bundle

src/cmd/objdump/elf.go

-// Copyright 2013 The Go Authors.  All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// Parsing of ELF executables (Linux, FreeBSD, and so on).
-
-package main
-
-import (
-	"debug/elf"
-	"os"
-)
-
-func elfSymbols(f *os.File) (syms []Sym, goarch string) {
-	p, err := elf.NewFile(f)
-	if err != nil {
-		errorf("parsing %s: %v", f.Name(), err)
-		return
-	}
-
-	elfSyms, err := p.Symbols()
-	if err != nil {
-		errorf("parsing %s: %v", f.Name(), err)
-		return
-	}
-
-	switch p.Machine {
-	case elf.EM_X86_64:
-		goarch = "amd64"
-	case elf.EM_386:
-		goarch = "386"
-	case elf.EM_ARM:
-		goarch = "arm"
-	}
-
-	for _, s := range elfSyms {
-		sym := Sym{Addr: s.Value, Name: s.Name, Size: int64(s.Size), Code: '?'}
-		switch s.Section {
-		case elf.SHN_UNDEF:
-			sym.Code = 'U'
-		case elf.SHN_COMMON:
-			sym.Code = 'B'
-		default:
-			i := int(s.Section)
-			if i < 0 || i >= len(p.Sections) {
-				break
-			}
-			sect := p.Sections[i]
-			switch sect.Flags & (elf.SHF_WRITE | elf.SHF_ALLOC | elf.SHF_EXECINSTR) {
-			case elf.SHF_ALLOC | elf.SHF_EXECINSTR:
-				sym.Code = 'T'
-			case elf.SHF_ALLOC:
-				sym.Code = 'R'
-			case elf.SHF_ALLOC | elf.SHF_WRITE:
-				sym.Code = 'D'
-			}
-		}
-		if elf.ST_BIND(s.Info) == elf.STB_LOCAL {
-			sym.Code += 'a' - 'A'
-		}
-		syms = append(syms, sym)
-	}
-
-	return
-}

src/cmd/objdump/macho.go

-// Copyright 2013 The Go Authors.  All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// Parsing of Mach-O executables (OS X).
-
-package main
-
-import (
-	"debug/macho"
-	"os"
-	"sort"
-)
-
-func machoSymbols(f *os.File) (syms []Sym, goarch string) {
-	p, err := macho.NewFile(f)
-	if err != nil {
-		errorf("parsing %s: %v", f.Name(), err)
-		return
-	}
-
-	if p.Symtab == nil {
-		errorf("%s: no symbol table", f.Name())
-		return
-	}
-
-	switch p.Cpu {
-	case macho.Cpu386:
-		goarch = "386"
-	case macho.CpuAmd64:
-		goarch = "amd64"
-	case macho.CpuArm:
-		goarch = "arm"
-	}
-
-	// Build sorted list of addresses of all symbols.
-	// We infer the size of a symbol by looking at where the next symbol begins.
-	var addrs []uint64
-	for _, s := range p.Symtab.Syms {
-		addrs = append(addrs, s.Value)
-	}
-	sort.Sort(uint64s(addrs))
-
-	for _, s := range p.Symtab.Syms {
-		sym := Sym{Name: s.Name, Addr: s.Value, Code: '?'}
-		i := sort.Search(len(addrs), func(x int) bool { return addrs[x] > s.Value })
-		if i < len(addrs) {
-			sym.Size = int64(addrs[i] - s.Value)
-		}
-		if s.Sect == 0 {
-			sym.Code = 'U'
-		} else if int(s.Sect) <= len(p.Sections) {
-			sect := p.Sections[s.Sect-1]
-			switch sect.Seg {
-			case "__TEXT":
-				sym.Code = 'R'
-			case "__DATA":
-				sym.Code = 'D'
-			}
-			switch sect.Seg + " " + sect.Name {
-			case "__TEXT __text":
-				sym.Code = 'T'
-			case "__DATA __bss", "__DATA __noptrbss":
-				sym.Code = 'B'
-			}
-		}
-		syms = append(syms, sym)
-	}
-
-	return
-}
-
-type uint64s []uint64
-
-func (x uint64s) Len() int           { return len(x) }
-func (x uint64s) Swap(i, j int)      { x[i], x[j] = x[j], x[i] }
-func (x uint64s) Less(i, j int) bool { return x[i] < x[j] }

src/cmd/objdump/pe.go

-// Copyright 2013 The Go Authors.  All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// Parsing of PE executables (Microsoft Windows).
-
-package main
-
-import (
-	"debug/pe"
-	"os"
-	"sort"
-)
-
-func peSymbols(f *os.File) (syms []Sym, goarch string) {
-	p, err := pe.NewFile(f)
-	if err != nil {
-		errorf("parsing %s: %v", f.Name(), err)
-		return
-	}
-
-	// Build sorted list of addresses of all symbols.
-	// We infer the size of a symbol by looking at where the next symbol begins.
-	var addrs []uint64
-
-	var imageBase uint64
-	switch oh := p.OptionalHeader.(type) {
-	case *pe.OptionalHeader32:
-		imageBase = uint64(oh.ImageBase)
-		goarch = "386"
-	case *pe.OptionalHeader64:
-		imageBase = oh.ImageBase
-		goarch = "amd64"
-	default:
-		errorf("parsing %s: file format not recognized", f.Name())
-		return
-	}
-
-	for _, s := range p.Symbols {
-		const (
-			N_UNDEF = 0  // An undefined (extern) symbol
-			N_ABS   = -1 // An absolute symbol (e_value is a constant, not an address)
-			N_DEBUG = -2 // A debugging symbol
-		)
-		sym := Sym{Name: s.Name, Addr: uint64(s.Value), Code: '?'}
-		switch s.SectionNumber {
-		case N_UNDEF:
-			sym.Code = 'U'
-		case N_ABS:
-			sym.Code = 'C'
-		case N_DEBUG:
-			sym.Code = '?'
-		default:
-			if s.SectionNumber < 0 {
-				errorf("parsing %s: invalid section number %d", f.Name(), s.SectionNumber)
-				return
-			}
-			if len(p.Sections) < int(s.SectionNumber) {
-				errorf("parsing %s: section number %d is large then max %d", f.Name(), s.SectionNumber, len(p.Sections))
-				return
-			}
-			sect := p.Sections[s.SectionNumber-1]
-			const (
-				text  = 0x20
-				data  = 0x40
-				bss   = 0x80
-				permX = 0x20000000
-				permR = 0x40000000
-				permW = 0x80000000
-			)
-			ch := sect.Characteristics
-			switch {
-			case ch&text != 0:
-				sym.Code = 'T'
-			case ch&data != 0:
-				if ch&permW == 0 {
-					sym.Code = 'R'
-				} else {
-					sym.Code = 'D'
-				}
-			case ch&bss != 0:
-				sym.Code = 'B'
-			}
-			sym.Addr += imageBase + uint64(sect.VirtualAddress)
-		}
-		syms = append(syms, sym)
-		addrs = append(addrs, sym.Addr)
-	}
-
-	sort.Sort(uint64s(addrs))
-	for i := range syms {
-		j := sort.Search(len(addrs), func(x int) bool { return addrs[x] > syms[i].Addr })
-		if j < len(addrs) {
-			syms[i].Size = int64(addrs[j] - syms[i].Addr)
-		}
-	}
-
-	return
-}

src/cmd/objdump/plan9obj.go

-// Copyright 2014 The Go Authors.  All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// Parsing of Plan 9 a.out executables.
-
-package main
-
-import (
-	"debug/plan9obj"
-	"os"
-	"sort"
-)
-
-var validSymType = map[rune]bool{
-	'T': true,
-	't': true,
-	'D': true,
-	'd': true,
-	'B': true,
-	'b': true,
-}
-
-func plan9Symbols(f *os.File) (syms []Sym, goarch string) {
-	p, err := plan9obj.NewFile(f)
-	if err != nil {
-		errorf("parsing %s: %v", f.Name(), err)
-		return
-	}
-
-	plan9Syms, err := p.Symbols()
-	if err != nil {
-		errorf("parsing %s: %v", f.Name(), err)
-		return
-	}
-
-	switch p.Magic {
-	case plan9obj.MagicAMD64:
-		goarch = "amd64"
-	case plan9obj.Magic386:
-		goarch = "386"
-	case plan9obj.MagicARM:
-		goarch = "arm"
-	}
-
-	// Build sorted list of addresses of all symbols.
-	// We infer the size of a symbol by looking at where the next symbol begins.
-	var addrs []uint64
-	for _, s := range plan9Syms {
-		if !validSymType[s.Type] {
-			continue
-		}
-		addrs = append(addrs, s.Value)
-	}
-	sort.Sort(uint64s(addrs))
-
-	for _, s := range plan9Syms {
-		if !validSymType[s.Type] {
-			continue
-		}
-		sym := Sym{Addr: s.Value, Name: s.Name, Code: rune(s.Type)}
-		i := sort.Search(len(addrs), func(x int) bool { return addrs[x] > s.Value })
-		if i < len(addrs) {
-			sym.Size = int64(addrs[i] - s.Value)
-		}
-		syms = append(syms, sym)
-	}
-
-	return
-}