1// Copyright 2014 The Go Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style
3// license that can be found in the LICENSE file.
4
5package objfile
6
7import (
8	"bufio"
9	"bytes"
10	"cmd/internal/src"
11	"container/list"
12	"debug/gosym"
13	"encoding/binary"
14	"fmt"
15	"io"
16	"io/ioutil"
17	"os"
18	"path/filepath"
19	"regexp"
20	"sort"
21	"strings"
22	"text/tabwriter"
23
24	"golang.org/x/arch/arm/armasm"
25	"golang.org/x/arch/arm64/arm64asm"
26	"golang.org/x/arch/ppc64/ppc64asm"
27	"golang.org/x/arch/x86/x86asm"
28)
29
30// Disasm is a disassembler for a given File.
31type Disasm struct {
32	syms      []Sym            //symbols in file, sorted by address
33	pcln      Liner            // pcln table
34	text      []byte           // bytes of text segment (actual instructions)
35	textStart uint64           // start PC of text
36	textEnd   uint64           // end PC of text
37	goarch    string           // GOARCH string
38	disasm    disasmFunc       // disassembler function for goarch
39	byteOrder binary.ByteOrder // byte order for goarch
40}
41
42// Disasm returns a disassembler for the file f.
43func (e *Entry) Disasm() (*Disasm, error) {
44	syms, err := e.Symbols()
45	if err != nil {
46		return nil, err
47	}
48
49	pcln, err := e.PCLineTable()
50	if err != nil {
51		return nil, err
52	}
53
54	textStart, textBytes, err := e.Text()
55	if err != nil {
56		return nil, err
57	}
58
59	goarch := e.GOARCH()
60	disasm := disasms[goarch]
61	byteOrder := byteOrders[goarch]
62	if disasm == nil || byteOrder == nil {
63		return nil, fmt.Errorf("unsupported architecture")
64	}
65
66	// Filter out section symbols, overwriting syms in place.
67	keep := syms[:0]
68	for _, sym := range syms {
69		switch sym.Name {
70		case "runtime.text", "text", "_text", "runtime.etext", "etext", "_etext":
71			// drop
72		default:
73			keep = append(keep, sym)
74		}
75	}
76	syms = keep
77	d := &Disasm{
78		syms:      syms,
79		pcln:      pcln,
80		text:      textBytes,
81		textStart: textStart,
82		textEnd:   textStart + uint64(len(textBytes)),
83		goarch:    goarch,
84		disasm:    disasm,
85		byteOrder: byteOrder,
86	}
87
88	return d, nil
89}
90
91// lookup finds the symbol name containing addr.
92func (d *Disasm) lookup(addr uint64) (name string, base uint64) {
93	i := sort.Search(len(d.syms), func(i int) bool { return addr < d.syms[i].Addr })
94	if i > 0 {
95		s := d.syms[i-1]
96		if s.Addr != 0 && s.Addr <= addr && addr < s.Addr+uint64(s.Size) {
97			return s.Name, s.Addr
98		}
99	}
100	return "", 0
101}
102
103// base returns the final element in the path.
104// It works on both Windows and Unix paths,
105// regardless of host operating system.
106func base(path string) string {
107	path = path[strings.LastIndex(path, "/")+1:]
108	path = path[strings.LastIndex(path, `\`)+1:]
109	return path
110}
111
112// CachedFile contains the content of a file split into lines.
113type CachedFile struct {
114	FileName string
115	Lines    [][]byte
116}
117
118// FileCache is a simple LRU cache of file contents.
119type FileCache struct {
120	files  *list.List
121	maxLen int
122}
123
124// NewFileCache returns a FileCache which can contain up to maxLen cached file contents.
125func NewFileCache(maxLen int) *FileCache {
126	return &FileCache{
127		files:  list.New(),
128		maxLen: maxLen,
129	}
130}
131
132// Line returns the source code line for the given file and line number.
133// If the file is not already cached, reads it, inserts it into the cache,
134// and removes the least recently used file if necessary.
135// If the file is in cache, it is moved to the front of the list.
136func (fc *FileCache) Line(filename string, line int) ([]byte, error) {
137	if filepath.Ext(filename) != ".go" {
138		return nil, nil
139	}
140
141	// Clean filenames returned by src.Pos.SymFilename()
142	// or src.PosBase.SymFilename() removing
143	// the leading src.FileSymPrefix.
144	filename = strings.TrimPrefix(filename, src.FileSymPrefix)
145
146	// Expand literal "$GOROOT" rewritten by obj.AbsFile()
147	filename = filepath.Clean(os.ExpandEnv(filename))
148
149	var cf *CachedFile
150	var e *list.Element
151
152	for e = fc.files.Front(); e != nil; e = e.Next() {
153		cf = e.Value.(*CachedFile)
154		if cf.FileName == filename {
155			break
156		}
157	}
158
159	if e == nil {
160		content, err := ioutil.ReadFile(filename)
161		if err != nil {
162			return nil, err
163		}
164
165		cf = &CachedFile{
166			FileName: filename,
167			Lines:    bytes.Split(content, []byte{'\n'}),
168		}
169		fc.files.PushFront(cf)
170
171		if fc.files.Len() >= fc.maxLen {
172			fc.files.Remove(fc.files.Back())
173		}
174	} else {
175		fc.files.MoveToFront(e)
176	}
177
178	return cf.Lines[line-1], nil
179}
180
181// Print prints a disassembly of the file to w.
182// If filter is non-nil, the disassembly only includes functions with names matching filter.
183// If printCode is true, the disassembly includs corresponding source lines.
184// The disassembly only includes functions that overlap the range [start, end).
185func (d *Disasm) Print(w io.Writer, filter *regexp.Regexp, start, end uint64, printCode bool) {
186	if start < d.textStart {
187		start = d.textStart
188	}
189	if end > d.textEnd {
190		end = d.textEnd
191	}
192	printed := false
193	bw := bufio.NewWriter(w)
194
195	var fc *FileCache
196	if printCode {
197		fc = NewFileCache(8)
198	}
199
200	tw := tabwriter.NewWriter(bw, 18, 8, 1, '\t', tabwriter.StripEscape)
201	for _, sym := range d.syms {
202		symStart := sym.Addr
203		symEnd := sym.Addr + uint64(sym.Size)
204		relocs := sym.Relocs
205		if sym.Code != 'T' && sym.Code != 't' ||
206			symStart < d.textStart ||
207			symEnd <= start || end <= symStart ||
208			filter != nil && !filter.MatchString(sym.Name) {
209			continue
210		}
211		if printed {
212			fmt.Fprintf(bw, "\n")
213		}
214		printed = true
215
216		file, _, _ := d.pcln.PCToLine(sym.Addr)
217		fmt.Fprintf(bw, "TEXT %s(SB) %s\n", sym.Name, file)
218
219		if symEnd > end {
220			symEnd = end
221		}
222		code := d.text[:end-d.textStart]
223
224		var lastFile string
225		var lastLine int
226
227		d.Decode(symStart, symEnd, relocs, func(pc, size uint64, file string, line int, text string) {
228			i := pc - d.textStart
229
230			if printCode {
231				if file != lastFile || line != lastLine {
232					if srcLine, err := fc.Line(file, line); err == nil {
233						fmt.Fprintf(tw, "%s%s%s\n", []byte{tabwriter.Escape}, srcLine, []byte{tabwriter.Escape})
234					}
235
236					lastFile, lastLine = file, line
237				}
238
239				fmt.Fprintf(tw, "  %#x\t", pc)
240			} else {
241				fmt.Fprintf(tw, "  %s:%d\t%#x\t", base(file), line, pc)
242			}
243
244			if size%4 != 0 || d.goarch == "386" || d.goarch == "amd64" {
245				// Print instruction as bytes.
246				fmt.Fprintf(tw, "%x", code[i:i+size])
247			} else {
248				// Print instruction as 32-bit words.
249				for j := uint64(0); j < size; j += 4 {
250					if j > 0 {
251						fmt.Fprintf(tw, " ")
252					}
253					fmt.Fprintf(tw, "%08x", d.byteOrder.Uint32(code[i+j:]))
254				}
255			}
256			fmt.Fprintf(tw, "\t%s\t\n", text)
257		})
258		tw.Flush()
259	}
260	bw.Flush()
261}
262
263// Decode disassembles the text segment range [start, end), calling f for each instruction.
264func (d *Disasm) Decode(start, end uint64, relocs []Reloc, f func(pc, size uint64, file string, line int, text string)) {
265	if start < d.textStart {
266		start = d.textStart
267	}
268	if end > d.textEnd {
269		end = d.textEnd
270	}
271	code := d.text[:end-d.textStart]
272	lookup := d.lookup
273	for pc := start; pc < end; {
274		i := pc - d.textStart
275		text, size := d.disasm(code[i:], pc, lookup, d.byteOrder)
276		file, line, _ := d.pcln.PCToLine(pc)
277		sep := "\t"
278		for len(relocs) > 0 && relocs[0].Addr < i+uint64(size) {
279			text += sep + relocs[0].Stringer.String(pc-start)
280			sep = " "
281			relocs = relocs[1:]
282		}
283		f(pc, uint64(size), file, line, text)
284		pc += uint64(size)
285	}
286}
287
288type lookupFunc = func(addr uint64) (sym string, base uint64)
289type disasmFunc func(code []byte, pc uint64, lookup lookupFunc, ord binary.ByteOrder) (text string, size int)
290
291func disasm_386(code []byte, pc uint64, lookup lookupFunc, _ binary.ByteOrder) (string, int) {
292	return disasm_x86(code, pc, lookup, 32)
293}
294
295func disasm_amd64(code []byte, pc uint64, lookup lookupFunc, _ binary.ByteOrder) (string, int) {
296	return disasm_x86(code, pc, lookup, 64)
297}
298
299func disasm_x86(code []byte, pc uint64, lookup lookupFunc, arch int) (string, int) {
300	inst, err := x86asm.Decode(code, arch)
301	var text string
302	size := inst.Len
303	if err != nil || size == 0 || inst.Op == 0 {
304		size = 1
305		text = "?"
306	} else {
307		text = x86asm.GoSyntax(inst, pc, lookup)
308	}
309	return text, size
310}
311
312type textReader struct {
313	code []byte
314	pc   uint64
315}
316
317func (r textReader) ReadAt(data []byte, off int64) (n int, err error) {
318	if off < 0 || uint64(off) < r.pc {
319		return 0, io.EOF
320	}
321	d := uint64(off) - r.pc
322	if d >= uint64(len(r.code)) {
323		return 0, io.EOF
324	}
325	n = copy(data, r.code[d:])
326	if n < len(data) {
327		err = io.ErrUnexpectedEOF
328	}
329	return
330}
331
332func disasm_arm(code []byte, pc uint64, lookup lookupFunc, _ binary.ByteOrder) (string, int) {
333	inst, err := armasm.Decode(code, armasm.ModeARM)
334	var text string
335	size := inst.Len
336	if err != nil || size == 0 || inst.Op == 0 {
337		size = 4
338		text = "?"
339	} else {
340		text = armasm.GoSyntax(inst, pc, lookup, textReader{code, pc})
341	}
342	return text, size
343}
344
345func disasm_arm64(code []byte, pc uint64, lookup lookupFunc, byteOrder binary.ByteOrder) (string, int) {
346	inst, err := arm64asm.Decode(code)
347	var text string
348	if err != nil || inst.Op == 0 {
349		text = "?"
350	} else {
351		text = arm64asm.GoSyntax(inst, pc, lookup, textReader{code, pc})
352	}
353	return text, 4
354}
355
356func disasm_ppc64(code []byte, pc uint64, lookup lookupFunc, byteOrder binary.ByteOrder) (string, int) {
357	inst, err := ppc64asm.Decode(code, byteOrder)
358	var text string
359	size := inst.Len
360	if err != nil || size == 0 {
361		size = 4
362		text = "?"
363	} else {
364		text = ppc64asm.GoSyntax(inst, pc, lookup)
365	}
366	return text, size
367}
368
369var disasms = map[string]disasmFunc{
370	"386":     disasm_386,
371	"amd64":   disasm_amd64,
372	"arm":     disasm_arm,
373	"arm64":   disasm_arm64,
374	"ppc64":   disasm_ppc64,
375	"ppc64le": disasm_ppc64,
376}
377
378var byteOrders = map[string]binary.ByteOrder{
379	"386":     binary.LittleEndian,
380	"amd64":   binary.LittleEndian,
381	"arm":     binary.LittleEndian,
382	"arm64":   binary.LittleEndian,
383	"ppc64":   binary.BigEndian,
384	"ppc64le": binary.LittleEndian,
385	"s390x":   binary.BigEndian,
386}
387
388type Liner interface {
389	// Given a pc, returns the corresponding file, line, and function data.
390	// If unknown, returns "",0,nil.
391	PCToLine(uint64) (string, int, *gosym.Func)
392}
393