cmd/nm: reimplement in Go

The immediate goal is to support the new object file format,
which libmach (nm's support library) does not understand.
Rather than add code to libmach or reengineer liblink to
support this new use, just write it in Go.

The C version of nm reads the Plan 9 symbol table stored in
Go binaries, now otherwise unused.

This reimplementation uses the standard symbol table for
the corresponding file format instead, bringing us one step
closer to removing the Plan 9 symbol table from Go binaries.

Tell cmd/dist not to build cmd/nm anymore.
Tell cmd/go to install cmd/nm in the tool directory.

R=golang-dev, r, iant, alex.brainman
CC=golang-dev
https://codereview.appspot.com/40600043

8 files changed, 454 insertions, 425 deletions

diff --git a/src/cmd/nm/Makefile b/src/cmd/nm/Makefile
deleted file mode 100644
index 3f528d751..000000000
--- a/src/cmd/nm/Makefile
+++ /dev/null
@@ -1,5 +0,0 @@
-# Copyright 2012 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.
-
-include ../../Make.dist
diff --git a/src/cmd/nm/doc.go b/src/cmd/nm/doc.go
index 8e88e2e63..619c2d12d 100644
--- a/src/cmd/nm/doc.go
+++ b/src/cmd/nm/doc.go
@@ -1,23 +1,37 @@
-// Copyright 2009 The Go Authors. All rights reserved.
+// 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.
 
-// +build ignore
-
-/*
-
-Nm is a version of the Plan 9 nm command.  The original is documented at
-
-	http://plan9.bell-labs.com/magic/man2html/1/nm
-
-It prints the name list (symbol table) for programs compiled by gc as well as the
-Plan 9 C compiler.
-
-This implementation adds the flag -S, which prints each symbol's size
-in decimal after its address.
-
-Usage:
-	go tool nm [-aghnsSTu] file
-
-*/
+// Nm lists the symbols defined or used by an object file, archive, or executable.
+//
+// Usage:
+//	go tool nm [options] file...
+//
+// The default output prints one line per symbol, with three space-separated
+// fields giving the address (in hexadecimal), type (a character), and name of
+// the symbol. The types are:
+//
+//	T	text (code) segment symbol
+//	t	static text segment symbol
+//	D	data segment symbol
+//	d	static data segment symbol
+//	B	bss segment symbol
+//	b	static bss segment symbol
+//	U	referenced but undefined symbol
+//
+// Following established convention, the address is omitted for undefined
+// symbols (type U).
+//
+// The options control the printed output:
+//
+//	-n
+//		an alias for -sort address (numeric),
+//		for compatiblity with other nm commands
+//	-size
+//		print symbol size in decimal between address and type
+//	-sort {address,name,none}
+//		sort output in the given order (default name)
+//	-type
+//		print symbol type after name
+//
 package main
diff --git a/src/cmd/nm/elf.go b/src/cmd/nm/elf.go
new file mode 100644
index 000000000..6dc8d442b
--- /dev/null
+++ b/src/cmd/nm/elf.go
@@ -0,0 +1,54 @@
+// 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) []Sym {
+	p, err := elf.NewFile(f)
+	if err != nil {
+		errorf("parsing %s: %v", f.Name(), err)
+		return nil
+	}
+
+	elfSyms, err := p.Symbols()
+	if err != nil {
+		errorf("parsing %s: %v", f.Name(), err)
+		return nil
+	}
+
+	var syms []Sym
+	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-1]
+			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'
+			}
+		}
+		syms = append(syms, sym)
+	}
+
+	return syms
+}
diff --git a/src/cmd/nm/goobj.go b/src/cmd/nm/goobj.go
new file mode 100644
index 000000000..1ceefff9c
--- /dev/null
+++ b/src/cmd/nm/goobj.go
@@ -0,0 +1,68 @@
+// 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 Go intermediate object files and archives.
+
+package main
+
+import (
+	"debug/goobj"
+	"fmt"
+	"os"
+)
+
+func goobjName(id goobj.SymID) string {
+	if id.Version == 0 {
+		return id.Name
+	}
+	return fmt.Sprintf("%s<%d>", id.Name, id.Version)
+}
+
+func goobjSymbols(f *os.File) []Sym {
+	pkg, err := goobj.Parse(f, `""`)
+	if err != nil {
+		errorf("parsing %s: %v", f.Name(), err)
+		return nil
+	}
+
+	seen := make(map[goobj.SymID]bool)
+
+	var syms []Sym
+	for _, s := range pkg.Syms {
+		seen[s.SymID] = true
+		sym := Sym{Addr: uint64(s.Data.Offset), Name: goobjName(s.SymID), Size: int64(s.Size), Type: s.Type.Name, Code: '?'}
+		switch s.Kind {
+		case goobj.STEXT, goobj.SELFRXSECT:
+			sym.Code = 'T'
+		case goobj.STYPE, goobj.SSTRING, goobj.SGOSTRING, goobj.SGOFUNC, goobj.SRODATA, goobj.SFUNCTAB, goobj.STYPELINK, goobj.SSYMTAB, goobj.SPCLNTAB, goobj.SELFROSECT:
+			sym.Code = 'R'
+		case goobj.SMACHOPLT, goobj.SELFSECT, goobj.SMACHO, goobj.SMACHOGOT, goobj.SNOPTRDATA, goobj.SINITARR, goobj.SDATA, goobj.SWINDOWS:
+			sym.Code = 'D'
+		case goobj.SBSS, goobj.SNOPTRBSS, goobj.STLSBSS:
+			sym.Code = 'B'
+		case goobj.SXREF, goobj.SMACHOSYMSTR, goobj.SMACHOSYMTAB, goobj.SMACHOINDIRECTPLT, goobj.SMACHOINDIRECTGOT, goobj.SFILE, goobj.SFILEPATH, goobj.SCONST, goobj.SDYNIMPORT, goobj.SHOSTOBJ:
+			sym.Code = 'X'
+		}
+		if s.Version != 0 {
+			sym.Code += 'a' - 'A'
+		}
+		syms = append(syms, sym)
+	}
+
+	for _, s := range pkg.Syms {
+		for _, r := range s.Reloc {
+			if !seen[r.Sym] {
+				seen[r.Sym] = true
+				sym := Sym{Name: goobjName(r.Sym), Code: 'U'}
+				if s.Version != 0 {
+					// should not happen but handle anyway
+					sym.Code = 'u'
+				}
+				syms = append(syms, sym)
+			}
+		}
+	}
+
+	return syms
+}
diff --git a/src/cmd/nm/macho.go b/src/cmd/nm/macho.go
new file mode 100644
index 000000000..c60bde55b
--- /dev/null
+++ b/src/cmd/nm/macho.go
@@ -0,0 +1,69 @@
+// 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) []Sym {
+	p, err := macho.NewFile(f)
+	if err != nil {
+		errorf("parsing %s: %v", f.Name(), err)
+		return nil
+	}
+
+	if p.Symtab == nil {
+		errorf("%s: no symbol table", f.Name())
+		return nil
+	}
+
+	// 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))
+
+	var syms []Sym
+	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 syms
+}
+
+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] }
diff --git a/src/cmd/nm/nm.c b/src/cmd/nm/nm.c
deleted file mode 100644
index 820942426..000000000
--- a/src/cmd/nm/nm.c
+++ /dev/null
@@ -1,401 +0,0 @@
-// Inferno utils/nm/nm.c
-// http://code.google.com/p/inferno-os/source/browse/utils/nm/nm.c
-//
-//	Copyright © 1994-1999 Lucent Technologies Inc.  All rights reserved.
-//	Portions Copyright © 1995-1997 C H Forsyth (forsyth@terzarima.net)
-//	Portions Copyright © 1997-1999 Vita Nuova Limited
-//	Portions Copyright © 2000-2007 Vita Nuova Holdings Limited (www.vitanuova.com)
-//	Portions Copyright © 2004,2006 Bruce Ellis
-//	Portions Copyright © 2005-2007 C H Forsyth (forsyth@terzarima.net)
-//	Revisions Copyright © 2000-2007 Lucent Technologies Inc. and others
-//	Portions Copyright © 2009 The Go Authors. All rights reserved.
-//
-// Permission is hereby granted, free of charge, to any person obtaining a copy
-// of this software and associated documentation files (the "Software"), to deal
-// in the Software without restriction, including without limitation the rights
-// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-// copies of the Software, and to permit persons to whom the Software is
-// furnished to do so, subject to the following conditions:
-//
-// The above copyright notice and this permission notice shall be included in
-// all copies or substantial portions of the Software.
-//
-// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
-// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
-// THE SOFTWARE.
-
-/*
- * nm.c -- drive nm
- */
-#include <u.h>
-#include <libc.h>
-#include <ar.h>
-#include <bio.h>
-#include <mach.h>
-
-enum{
-	CHUNK	=	256	/* must be power of 2 */
-};
-
-char	*errs;			/* exit status */
-char	*filename;		/* current file */
-char	symname[]="__.GOSYMDEF";	/* table of contents file name */
-int	multifile;		/* processing multiple files */
-int	aflag;
-int	gflag;
-int	hflag;
-int	nflag;
-int	sflag;
-int	Sflag;
-int	uflag;
-int	Tflag;
-int	tflag;
-
-Sym	**fnames;		/* file path translation table */
-Sym	**symptr;
-int	nsym;
-Biobuf	bout;
-
-int	cmp(void*, void*);
-void	error(char*, ...);
-void	execsyms(int);
-void	psym(Sym*, void*);
-void	printsyms(Sym**, long);
-void	doar(Biobuf*);
-void	dofile(Biobuf*);
-void	zenter(Sym*);
-
-void
-usage(void)
-{
-	fprint(2, "usage: nm [-aghnsSTu] file ...\n");
-	exits("usage");
-}
-
-void
-main(int argc, char *argv[])
-{
-	int i;
-	Biobuf	*bin;
-
-	Binit(&bout, 1, OWRITE);
-	argv0 = argv[0];
-	ARGBEGIN {
-	default:	usage();
-	case 'a':	aflag = 1; break;
-	case 'g':	gflag = 1; break;
-	case 'h':	hflag = 1; break;
-	case 'n':	nflag = 1; break;
-	case 's':	sflag = 1; break;
-	case 'S':	nflag = Sflag = 1; break;
-	case 'u':	uflag = 1; break;
-	case 't':	tflag = 1; break;
-	case 'T':	Tflag = 1; break;
-	} ARGEND
-	if (argc == 0)
-		usage();
-	if (argc > 1)
-		multifile++;
-	for(i=0; i<argc; i++){
-		filename = argv[i];
-		bin = Bopen(filename, OREAD);
-		if(bin == 0){
-			error("cannot open %s", filename);
-			continue;
-		}
-		if (isar(bin))
-			doar(bin);
-		else{
-			Bseek(bin, 0, 0);
-			dofile(bin);
-		}
-		Bterm(bin);
-	}
-	exits(errs);
-}
-
-/*
- * read an archive file,
- * processing the symbols for each intermediate file in it.
- */
-void
-doar(Biobuf *bp)
-{
-	int offset, size, obj;
-	char name[SARNAME];
-
-	multifile = 1;
-	for (offset = Boffset(bp);;offset += size) {
-		size = nextar(bp, offset, name);
-		if (size < 0) {
-			error("phase error on ar header %d", offset);
-			return;
-		}
-		if (size == 0)
-			return;
-		if (strcmp(name, symname) == 0)
-			continue;
-		obj = objtype(bp, 0);
-		if (obj < 0) {
-			// perhaps foreign object
-			if(strlen(name) > 2 && strcmp(name+strlen(name)-2, ".o") == 0)
-				return;
-			error("inconsistent file %s in %s",
-					name, filename);
-			return;
-		}
-		if (!readar(bp, obj, offset+size, 1)) {
-			error("invalid symbol reference in file %s",
-					name);
-			return;
-		}
-		filename = name;
-		nsym=0;
-		objtraverse(psym, 0);
-		printsyms(symptr, nsym);
-	}
-}
-
-/*
- * process symbols in a file
- */
-void
-dofile(Biobuf *bp)
-{
-	int obj;
-
-	obj = objtype(bp, 0);
-	if (obj < 0)
-		execsyms(Bfildes(bp));
-	else
-	if (readobj(bp, obj)) {
-		nsym = 0;
-		objtraverse(psym, 0);
-		printsyms(symptr, nsym);
-	}
-}
-
-/*
- * comparison routine for sorting the symbol table
- *	this screws up on 'z' records when aflag == 1
- */
-int
-cmp(void *vs, void *vt)
-{
-	Sym **s, **t;
-
-	s = vs;
-	t = vt;
-	if(nflag)	// sort on address (numeric) order
-		if((*s)->value < (*t)->value)
-			return -1;
-		else
-			return (*s)->value > (*t)->value;
-	if(sflag)	// sort on file order (sequence)
-		return (*s)->sequence - (*t)->sequence;
-	return strcmp((*s)->name, (*t)->name);
-}
-/*
- * enter a symbol in the table of filename elements
- */
-void
-zenter(Sym *s)
-{
-	static int maxf = 0;
-
-	if (s->value > maxf) {
-		maxf = (s->value+CHUNK-1) &~ (CHUNK-1);
-		fnames = realloc(fnames, (maxf+1)*sizeof(*fnames));
-		if(fnames == 0) {
-			error("out of memory", argv0);
-			exits("memory");
-		}
-	}
-	fnames[s->value] = s;
-}
-
-/*
- * get the symbol table from an executable file, if it has one
- */
-void
-execsyms(int fd)
-{
-	Fhdr f;
-	Sym *s;
-	int32 n;
-
-	seek(fd, 0, 0);
-	if (crackhdr(fd, &f) == 0) {
-		error("Can't read header for %s", filename);
-		return;
-	}
-	if (syminit(fd, &f) < 0)
-		return;
-	s = symbase(&n);
-	nsym = 0;
-	while(n--)
-		psym(s++, 0);
-
-	printsyms(symptr, nsym);
-}
-
-void
-psym(Sym *s, void* p)
-{
-	USED(p);
-	switch(s->type) {
-	case 'T':
-	case 'L':
-	case 'D':
-	case 'B':
-		if (uflag)
-			return;
-		if (!aflag && ((s->name[0] == '.' || s->name[0] == '$')))
-			return;
-		break;
-	case 'b':
-	case 'd':
-	case 'l':
-	case 't':
-		if (uflag || gflag)
-			return;
-		if (!aflag && ((s->name[0] == '.' || s->name[0] == '$')))
-			return;
-		break;
-	case 'U':
-		if (gflag)
-			return;
-		break;
-	case 'Z':
-		if (!aflag)
-			return;
-		break;
-	case 'm':
-		if(!aflag || uflag || gflag)
-			return;
-		break;
-	case 'f':	/* we only see a 'z' when the following is true*/
-		if(!aflag || uflag || gflag)
-			return;
-		zenter(s);
-		break;
-	case 'a':
-	case 'p':
-	case 'z':
-	default:
-		if(!aflag || uflag || gflag)
-			return;
-		break;
-	}
-	symptr = realloc(symptr, (nsym+1)*sizeof(Sym*));
-	if (symptr == 0) {
-		error("out of memory");
-		exits("memory");
-	}
-	symptr[nsym++] = s;
-}
-
-const char *skipnames[] = {
-	"bss",
-	"data",
-	"ebss",
-	"edata",
-	"egcbss",
-	"egcdata",
-	"enoptrbss",
-	"enoptrdata",
-	"epclntab",
-	"erodata",
-	"esymtab",
-	"etext",
-	"etypelink",
-	"noptrbss",
-	"noptrdata",
-	"rodata",
-	"text",
-};
-
-int
-skipsize(char *name)
-{
-	int i;
-	
-	for(i=0; i<nelem(skipnames); i++)
-		if(strcmp(skipnames[i], name) == 0)
-			return 1;
-	return 0;
-}
-
-void
-printsyms(Sym **symptr, long nsym)
-{
-	int i, j, wid;
-	Sym *s;
-	char *cp;
-	char path[512];
-
-	qsort(symptr, nsym, sizeof(*symptr), (void*)cmp);
-
-	wid = 0;
-	for (i=0; i<nsym; i++) {
-		s = symptr[i];
-		if (s->value && wid == 0)
-			wid = 8;
-		else if (s->value >= 0x100000000LL && wid == 8)
-			wid = 16;
-	}
-	for (i=0; i<nsym; i++) {
-		s = symptr[i];
-		if (multifile && !hflag)
-			Bprint(&bout, "%s:", filename);
-		if (s->type == 'z') {
-			fileelem(fnames, (uchar *) s->name, path, 512);
-			cp = path;
-		} else
-			cp = s->name;
-		if (Tflag)
-			Bprint(&bout, "%8ux ", s->sig);
-		if (s->value || s->type == 'a' || s->type == 'p')
-			Bprint(&bout, "%*llux ", wid, s->value);
-		else
-			Bprint(&bout, "%*s ", wid, "");
-		if(Sflag && !skipsize(cp)) {
-			vlong siz;
-
-			siz = 0;
-			for(j=i+1; j<nsym; j++) {
-				if(!skipsize(symptr[j]->name) && symptr[j]->type != 'a' && symptr[j]->type != 'p') {
-					siz = symptr[j]->value - s->value;
-					break;
-				}
-			}
-			if(siz > 0)
-				Bprint(&bout, "%*llud ", wid, siz);
-		}
-		Bprint(&bout, "%c %s", s->type, cp);
-		if(tflag && s->gotype)
-			Bprint(&bout, " %*llux", wid, s->gotype);
-		Bprint(&bout, "\n");
-	}
-}
-
-void
-error(char *fmt, ...)
-{
-	Fmt f;
-	char buf[128];
-	va_list arg;
-
-	fmtfdinit(&f, 2, buf, sizeof buf);
-	fmtprint(&f, "%s: ", argv0);
-	va_start(arg, fmt);
-	fmtvprint(&f, fmt, arg);
-	va_end(arg);
-	fmtprint(&f, "\n");
-	fmtfdflush(&f);
-	errs = "errors";
-}
diff --git a/src/cmd/nm/nm.go b/src/cmd/nm/nm.go
new file mode 100644
index 000000000..c37d1cc00
--- /dev/null
+++ b/src/cmd/nm/nm.go
@@ -0,0 +1,176 @@
+// 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.
+
+package main
+
+import (
+	"bufio"
+	"bytes"
+	"flag"
+	"fmt"
+	"io"
+	"log"
+	"os"
+	"sort"
+)
+
+func usage() {
+	fmt.Fprintf(os.Stderr, "usage: go tool nm [-n] [-size] [-sort order] [-type] file...\n")
+	os.Exit(2)
+}
+
+var (
+	sortOrder = flag.String("sort", "name", "")
+	printSize = flag.Bool("size", false, "")
+	printType = flag.Bool("type", false, "")
+)
+
+func init() {
+	flag.Var(nflag(0), "n", "") // alias for -sort address
+}
+
+type nflag int
+
+func (nflag) IsBoolFlag() bool {
+	return true
+}
+
+func (nflag) Set(value string) error {
+	if value == "true" {
+		*sortOrder = "address"
+	}
+	return nil
+}
+
+func (nflag) String() string {
+	if *sortOrder == "address" {
+		return "true"
+	}
+	return "false"
+}
+
+func main() {
+	log.SetFlags(0)
+	flag.Usage = usage
+	flag.Parse()
+
+	switch *sortOrder {
+	case "address", "name", "none":
+		// ok
+	default:
+		fmt.Fprintf(os.Stderr, "nm: unknown sort order %q\n", *sortOrder)
+		os.Exit(2)
+	}
+
+	args := flag.Args()
+	if len(args) == 0 {
+		flag.Usage()
+	}
+
+	for _, file := range args {
+		nm(file)
+	}
+
+	os.Exit(exitCode)
+}
+
+var exitCode = 0
+
+func errorf(format string, args ...interface{}) {
+	log.Printf(format, args...)
+	exitCode = 1
+}
+
+type Sym struct {
+	Addr uint64
+	Size int64
+	Code rune
+	Name string
+	Type string
+}
+
+var parsers = []struct {
+	prefix []byte
+	parse  func(*os.File) []Sym
+}{
+	{[]byte("!<arch>\n"), goobjSymbols},
+	{[]byte("go object "), goobjSymbols},
+	{[]byte("\x7FELF"), elfSymbols},
+	{[]byte("\xFE\xED\xFA\xCE"), machoSymbols},
+	{[]byte("\xFE\xED\xFA\xCF"), machoSymbols},
+	{[]byte("\xCE\xFA\xED\xFE"), machoSymbols},
+	{[]byte("\xCF\xFA\xED\xFE"), machoSymbols},
+	{[]byte("MZ"), peSymbols},
+}
+
+func nm(file string) {
+	f, err := os.Open(file)
+	if err != nil {
+		errorf("%v", err)
+		return
+	}
+	defer f.Close()
+
+	buf := make([]byte, 16)
+	io.ReadFull(f, buf)
+	f.Seek(0, 0)
+
+	var syms []Sym
+	for _, p := range parsers {
+		if bytes.HasPrefix(buf, p.prefix) {
+			syms = p.parse(f)
+			goto HaveSyms
+		}
+	}
+	errorf("%v: unknown file format", file)
+	return
+
+HaveSyms:
+	switch *sortOrder {
+	case "address":
+		sort.Sort(byAddr(syms))
+	case "name":
+		sort.Sort(byName(syms))
+	}
+
+	w := bufio.NewWriter(os.Stdout)
+	for _, sym := range syms {
+		if sym.Code == 'U' {
+			fmt.Fprintf(w, "%8s", "")
+		} else {
+			fmt.Fprintf(w, "%8x", sym.Addr)
+		}
+		if *printSize {
+			fmt.Fprintf(w, " %10d", sym.Size)
+		}
+		fmt.Fprintf(w, " %c %s", sym.Code, sym.Name)
+		if *printType && sym.Type != "" {
+			fmt.Fprintf(w, " %s", sym.Type)
+		}
+		fmt.Fprintf(w, "\n")
+	}
+	w.Flush()
+}
+
+func filter(syms []Sym, ok func(Sym) bool) []Sym {
+	out := syms[:0]
+	for _, sym := range syms {
+		if ok(sym) {
+			out = append(out, sym)
+		}
+	}
+	return out
+}
+
+type byAddr []Sym
+
+func (x byAddr) Len() int           { return len(x) }
+func (x byAddr) Swap(i, j int)      { x[i], x[j] = x[j], x[i] }
+func (x byAddr) Less(i, j int) bool { return x[i].Addr < x[j].Addr }
+
+type byName []Sym
+
+func (x byName) Len() int           { return len(x) }
+func (x byName) Swap(i, j int)      { x[i], x[j] = x[j], x[i] }
+func (x byName) Less(i, j int) bool { return x[i].Name < x[j].Name }
diff --git a/src/cmd/nm/pe.go b/src/cmd/nm/pe.go
new file mode 100644
index 000000000..bf78238d6
--- /dev/null
+++ b/src/cmd/nm/pe.go
@@ -0,0 +1,54 @@
+// 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"
+)
+
+func peSymbols(f *os.File) []Sym {
+	p, err := pe.NewFile(f)
+	if err != nil {
+		errorf("parsing %s: %v", f.Name(), err)
+		return nil
+	}
+
+	var syms []Sym
+	for _, s := range p.Symbols {
+		sym := Sym{Name: s.Name, Addr: uint64(s.Value), Code: '?'}
+		if s.SectionNumber == 0 {
+			sym.Code = 'U'
+		} else if int(s.SectionNumber) <= len(p.Sections) {
+			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'
+			}
+		}
+		syms = append(syms, sym)
+	}
+
+	return syms
+}