...
Run Format

Source file src/runtime/symtab.go

Documentation: runtime

  // 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.
  
  package runtime
  
  import (
  	"runtime/internal/atomic"
  	"runtime/internal/sys"
  	"unsafe"
  )
  
  // Frames may be used to get function/file/line information for a
  // slice of PC values returned by Callers.
  type Frames struct {
  	callers []uintptr
  
  	// If previous caller in iteration was a panic, then
  	// ci.callers[0] is the address of the faulting instruction
  	// instead of the return address of the call.
  	wasPanic bool
  
  	// Frames to return for subsequent calls to the Next method.
  	// Used for non-Go frames.
  	frames *[]Frame
  }
  
  // Frame is the information returned by Frames for each call frame.
  type Frame struct {
  	// Program counter for this frame; multiple frames may have
  	// the same PC value.
  	PC uintptr
  
  	// Func for this frame; may be nil for non-Go code or fully
  	// inlined functions.
  	Func *Func
  
  	// Function name, file name, and line number for this call frame.
  	// May be the empty string or zero if not known.
  	// If Func is not nil then Function == Func.Name().
  	Function string
  	File     string
  	Line     int
  
  	// Entry point for the function; may be zero if not known.
  	// If Func is not nil then Entry == Func.Entry().
  	Entry uintptr
  }
  
  // CallersFrames takes a slice of PC values returned by Callers and
  // prepares to return function/file/line information.
  // Do not change the slice until you are done with the Frames.
  func CallersFrames(callers []uintptr) *Frames {
  	return &Frames{callers: callers}
  }
  
  // Next returns frame information for the next caller.
  // If more is false, there are no more callers (the Frame value is valid).
  func (ci *Frames) Next() (frame Frame, more bool) {
  	if ci.frames != nil {
  		// We have saved up frames to return.
  		f := (*ci.frames)[0]
  		if len(*ci.frames) == 1 {
  			ci.frames = nil
  		} else {
  			*ci.frames = (*ci.frames)[1:]
  		}
  		return f, ci.frames != nil || len(ci.callers) > 0
  	}
  
  	if len(ci.callers) == 0 {
  		ci.wasPanic = false
  		return Frame{}, false
  	}
  	pc := ci.callers[0]
  	ci.callers = ci.callers[1:]
  	more = len(ci.callers) > 0
  	f := FuncForPC(pc)
  	if f == nil {
  		ci.wasPanic = false
  		if cgoSymbolizer != nil {
  			return ci.cgoNext(pc, more)
  		}
  		return Frame{}, more
  	}
  
  	entry := f.Entry()
  	xpc := pc
  	if xpc > entry && !ci.wasPanic {
  		xpc--
  	}
  	file, line := f.FileLine(xpc)
  
  	function := f.Name()
  	ci.wasPanic = entry == sigpanicPC
  
  	frame = Frame{
  		PC:       xpc,
  		Func:     f,
  		Function: function,
  		File:     file,
  		Line:     line,
  		Entry:    entry,
  	}
  
  	return frame, more
  }
  
  // cgoNext returns frame information for pc, known to be a non-Go function,
  // using the cgoSymbolizer hook.
  func (ci *Frames) cgoNext(pc uintptr, more bool) (Frame, bool) {
  	arg := cgoSymbolizerArg{pc: pc}
  	callCgoSymbolizer(&arg)
  
  	if arg.file == nil && arg.funcName == nil {
  		// No useful information from symbolizer.
  		return Frame{}, more
  	}
  
  	var frames []Frame
  	for {
  		frames = append(frames, Frame{
  			PC:       pc,
  			Func:     nil,
  			Function: gostring(arg.funcName),
  			File:     gostring(arg.file),
  			Line:     int(arg.lineno),
  			Entry:    arg.entry,
  		})
  		if arg.more == 0 {
  			break
  		}
  		callCgoSymbolizer(&arg)
  	}
  
  	// No more frames for this PC. Tell the symbolizer we are done.
  	// We don't try to maintain a single cgoSymbolizerArg for the
  	// whole use of Frames, because there would be no good way to tell
  	// the symbolizer when we are done.
  	arg.pc = 0
  	callCgoSymbolizer(&arg)
  
  	if len(frames) == 1 {
  		// Return a single frame.
  		return frames[0], more
  	}
  
  	// Return the first frame we saw and store the rest to be
  	// returned by later calls to Next.
  	rf := frames[0]
  	frames = frames[1:]
  	ci.frames = new([]Frame)
  	*ci.frames = frames
  	return rf, true
  }
  
  // NOTE: Func does not expose the actual unexported fields, because we return *Func
  // values to users, and we want to keep them from being able to overwrite the data
  // with (say) *f = Func{}.
  // All code operating on a *Func must call raw to get the *_func instead.
  
  // A Func represents a Go function in the running binary.
  type Func struct {
  	opaque struct{} // unexported field to disallow conversions
  }
  
  func (f *Func) raw() *_func {
  	return (*_func)(unsafe.Pointer(f))
  }
  
  // funcdata.h
  const (
  	_PCDATA_StackMapIndex       = 0
  	_FUNCDATA_ArgsPointerMaps   = 0
  	_FUNCDATA_LocalsPointerMaps = 1
  	_ArgsSizeUnknown            = -0x80000000
  )
  
  // moduledata records information about the layout of the executable
  // image. It is written by the linker. Any changes here must be
  // matched changes to the code in cmd/internal/ld/symtab.go:symtab.
  // moduledata is stored in read-only memory; none of the pointers here
  // are visible to the garbage collector.
  type moduledata struct {
  	pclntable    []byte
  	ftab         []functab
  	filetab      []uint32
  	findfunctab  uintptr
  	minpc, maxpc uintptr
  
  	text, etext           uintptr
  	noptrdata, enoptrdata uintptr
  	data, edata           uintptr
  	bss, ebss             uintptr
  	noptrbss, enoptrbss   uintptr
  	end, gcdata, gcbss    uintptr
  	types, etypes         uintptr
  
  	textsectmap []textsect
  	typelinks   []int32 // offsets from types
  	itablinks   []*itab
  
  	ptab []ptabEntry
  
  	pluginpath string
  	pkghashes  []modulehash
  
  	modulename   string
  	modulehashes []modulehash
  
  	gcdatamask, gcbssmask bitvector
  
  	typemap map[typeOff]*_type // offset to *_rtype in previous module
  
  	next *moduledata
  }
  
  // A modulehash is used to compare the ABI of a new module or a
  // package in a new module with the loaded program.
  //
  // For each shared library a module links against, the linker creates an entry in the
  // moduledata.modulehashes slice containing the name of the module, the abi hash seen
  // at link time and a pointer to the runtime abi hash. These are checked in
  // moduledataverify1 below.
  //
  // For each loaded plugin, the the pkghashes slice has a modulehash of the
  // newly loaded package that can be used to check the plugin's version of
  // a package against any previously loaded version of the package.
  // This is done in plugin.lastmoduleinit.
  type modulehash struct {
  	modulename   string
  	linktimehash string
  	runtimehash  *string
  }
  
  // pinnedTypemaps are the map[typeOff]*_type from the moduledata objects.
  //
  // These typemap objects are allocated at run time on the heap, but the
  // only direct reference to them is in the moduledata, created by the
  // linker and marked SNOPTRDATA so it is ignored by the GC.
  //
  // To make sure the map isn't collected, we keep a second reference here.
  var pinnedTypemaps []map[typeOff]*_type
  
  var firstmoduledata moduledata  // linker symbol
  var lastmoduledatap *moduledata // linker symbol
  var modulesSlice unsafe.Pointer // see activeModules
  
  // activeModules returns a slice of active modules.
  //
  // A module is active once its gcdatamask and gcbssmask have been
  // assembled and it is usable by the GC.
  func activeModules() []*moduledata {
  	p := (*[]*moduledata)(atomic.Loadp(unsafe.Pointer(&modulesSlice)))
  	if p == nil {
  		return nil
  	}
  	return *p
  }
  
  // modulesinit creates the active modules slice out of all loaded modules.
  //
  // When a module is first loaded by the dynamic linker, an .init_array
  // function (written by cmd/link) is invoked to call addmoduledata,
  // appending to the module to the linked list that starts with
  // firstmoduledata.
  //
  // There are two times this can happen in the lifecycle of a Go
  // program. First, if compiled with -linkshared, a number of modules
  // built with -buildmode=shared can be loaded at program initialization.
  // Second, a Go program can load a module while running that was built
  // with -buildmode=plugin.
  //
  // After loading, this function is called which initializes the
  // moduledata so it is usable by the GC and creates a new activeModules
  // list.
  //
  // Only one goroutine may call modulesinit at a time.
  func modulesinit() {
  	modules := new([]*moduledata)
  	for md := &firstmoduledata; md != nil; md = md.next {
  		*modules = append(*modules, md)
  		if md.gcdatamask == (bitvector{}) {
  			md.gcdatamask = progToPointerMask((*byte)(unsafe.Pointer(md.gcdata)), md.edata-md.data)
  			md.gcbssmask = progToPointerMask((*byte)(unsafe.Pointer(md.gcbss)), md.ebss-md.bss)
  		}
  	}
  
  	// Modules appear in the moduledata linked list in the order they are
  	// loaded by the dynamic loader, with one exception: the
  	// firstmoduledata itself the module that contains the runtime. This
  	// is not always the first module (when using -buildmode=shared, it
  	// is typically libstd.so, the second module). The order matters for
  	// typelinksinit, so we swap the first module with whatever module
  	// contains the main function.
  	//
  	// See Issue #18729.
  	mainText := funcPC(main_main)
  	for i, md := range *modules {
  		if md.text <= mainText && mainText <= md.etext {
  			(*modules)[0] = md
  			(*modules)[i] = &firstmoduledata
  			break
  		}
  	}
  
  	atomicstorep(unsafe.Pointer(&modulesSlice), unsafe.Pointer(modules))
  }
  
  type functab struct {
  	entry   uintptr
  	funcoff uintptr
  }
  
  // Mapping information for secondary text sections
  
  type textsect struct {
  	vaddr    uintptr // prelinked section vaddr
  	length   uintptr // section length
  	baseaddr uintptr // relocated section address
  }
  
  const minfunc = 16                 // minimum function size
  const pcbucketsize = 256 * minfunc // size of bucket in the pc->func lookup table
  
  // findfunctab is an array of these structures.
  // Each bucket represents 4096 bytes of the text segment.
  // Each subbucket represents 256 bytes of the text segment.
  // To find a function given a pc, locate the bucket and subbucket for
  // that pc. Add together the idx and subbucket value to obtain a
  // function index. Then scan the functab array starting at that
  // index to find the target function.
  // This table uses 20 bytes for every 4096 bytes of code, or ~0.5% overhead.
  type findfuncbucket struct {
  	idx        uint32
  	subbuckets [16]byte
  }
  
  func moduledataverify() {
  	for datap := &firstmoduledata; datap != nil; datap = datap.next {
  		moduledataverify1(datap)
  	}
  }
  
  const debugPcln = false
  
  func moduledataverify1(datap *moduledata) {
  	// See golang.org/s/go12symtab for header: 0xfffffffb,
  	// two zero bytes, a byte giving the PC quantum,
  	// and a byte giving the pointer width in bytes.
  	pcln := *(**[8]byte)(unsafe.Pointer(&datap.pclntable))
  	pcln32 := *(**[2]uint32)(unsafe.Pointer(&datap.pclntable))
  	if pcln32[0] != 0xfffffffb || pcln[4] != 0 || pcln[5] != 0 || pcln[6] != sys.PCQuantum || pcln[7] != sys.PtrSize {
  		println("runtime: function symbol table header:", hex(pcln32[0]), hex(pcln[4]), hex(pcln[5]), hex(pcln[6]), hex(pcln[7]))
  		throw("invalid function symbol table\n")
  	}
  
  	// ftab is lookup table for function by program counter.
  	nftab := len(datap.ftab) - 1
  	var pcCache pcvalueCache
  	for i := 0; i < nftab; i++ {
  		// NOTE: ftab[nftab].entry is legal; it is the address beyond the final function.
  		if datap.ftab[i].entry > datap.ftab[i+1].entry {
  			f1 := (*_func)(unsafe.Pointer(&datap.pclntable[datap.ftab[i].funcoff]))
  			f2 := (*_func)(unsafe.Pointer(&datap.pclntable[datap.ftab[i+1].funcoff]))
  			f2name := "end"
  			if i+1 < nftab {
  				f2name = funcname(f2)
  			}
  			println("function symbol table not sorted by program counter:", hex(datap.ftab[i].entry), funcname(f1), ">", hex(datap.ftab[i+1].entry), f2name)
  			for j := 0; j <= i; j++ {
  				print("\t", hex(datap.ftab[j].entry), " ", funcname((*_func)(unsafe.Pointer(&datap.pclntable[datap.ftab[j].funcoff]))), "\n")
  			}
  			throw("invalid runtime symbol table")
  		}
  
  		if debugPcln || nftab-i < 5 {
  			// Check a PC near but not at the very end.
  			// The very end might be just padding that is not covered by the tables.
  			// No architecture rounds function entries to more than 16 bytes,
  			// but if one came along we'd need to subtract more here.
  			// But don't use the next PC if it corresponds to a foreign object chunk
  			// (no pcln table, f2.pcln == 0). That chunk might have an alignment
  			// more than 16 bytes.
  			f := (*_func)(unsafe.Pointer(&datap.pclntable[datap.ftab[i].funcoff]))
  			end := f.entry
  			if i+1 < nftab {
  				f2 := (*_func)(unsafe.Pointer(&datap.pclntable[datap.ftab[i+1].funcoff]))
  				if f2.pcln != 0 {
  					end = f2.entry - 16
  					if end < f.entry {
  						end = f.entry
  					}
  				}
  			}
  			pcvalue(f, f.pcfile, end, &pcCache, true)
  			pcvalue(f, f.pcln, end, &pcCache, true)
  			pcvalue(f, f.pcsp, end, &pcCache, true)
  		}
  	}
  
  	if datap.minpc != datap.ftab[0].entry ||
  		datap.maxpc != datap.ftab[nftab].entry {
  		throw("minpc or maxpc invalid")
  	}
  
  	for _, modulehash := range datap.modulehashes {
  		if modulehash.linktimehash != *modulehash.runtimehash {
  			println("abi mismatch detected between", datap.modulename, "and", modulehash.modulename)
  			throw("abi mismatch")
  		}
  	}
  }
  
  // FuncForPC returns a *Func describing the function that contains the
  // given program counter address, or else nil.
  func FuncForPC(pc uintptr) *Func {
  	return (*Func)(unsafe.Pointer(findfunc(pc)))
  }
  
  // Name returns the name of the function.
  func (f *Func) Name() string {
  	return funcname(f.raw())
  }
  
  // Entry returns the entry address of the function.
  func (f *Func) Entry() uintptr {
  	return f.raw().entry
  }
  
  // FileLine returns the file name and line number of the
  // source code corresponding to the program counter pc.
  // The result will not be accurate if pc is not a program
  // counter within f.
  func (f *Func) FileLine(pc uintptr) (file string, line int) {
  	// Pass strict=false here, because anyone can call this function,
  	// and they might just be wrong about targetpc belonging to f.
  	file, line32 := funcline1(f.raw(), pc, false)
  	return file, int(line32)
  }
  
  func findmoduledatap(pc uintptr) *moduledata {
  	for datap := &firstmoduledata; datap != nil; datap = datap.next {
  		if datap.minpc <= pc && pc < datap.maxpc {
  			return datap
  		}
  	}
  	return nil
  }
  
  func findfunc(pc uintptr) *_func {
  	datap := findmoduledatap(pc)
  	if datap == nil {
  		return nil
  	}
  	const nsub = uintptr(len(findfuncbucket{}.subbuckets))
  
  	x := pc - datap.minpc
  	b := x / pcbucketsize
  	i := x % pcbucketsize / (pcbucketsize / nsub)
  
  	ffb := (*findfuncbucket)(add(unsafe.Pointer(datap.findfunctab), b*unsafe.Sizeof(findfuncbucket{})))
  	idx := ffb.idx + uint32(ffb.subbuckets[i])
  
  	// If the idx is beyond the end of the ftab, set it to the end of the table and search backward.
  	// This situation can occur if multiple text sections are generated to handle large text sections
  	// and the linker has inserted jump tables between them.
  
  	if idx >= uint32(len(datap.ftab)) {
  		idx = uint32(len(datap.ftab) - 1)
  	}
  	if pc < datap.ftab[idx].entry {
  
  		// With multiple text sections, the idx might reference a function address that
  		// is higher than the pc being searched, so search backward until the matching address is found.
  
  		for datap.ftab[idx].entry > pc && idx > 0 {
  			idx--
  		}
  		if idx == 0 {
  			throw("findfunc: bad findfunctab entry idx")
  		}
  	} else {
  
  		// linear search to find func with pc >= entry.
  		for datap.ftab[idx+1].entry <= pc {
  			idx++
  		}
  	}
  	return (*_func)(unsafe.Pointer(&datap.pclntable[datap.ftab[idx].funcoff]))
  }
  
  type pcvalueCache struct {
  	entries [16]pcvalueCacheEnt
  }
  
  type pcvalueCacheEnt struct {
  	// targetpc and off together are the key of this cache entry.
  	targetpc uintptr
  	off      int32
  	// val is the value of this cached pcvalue entry.
  	val int32
  }
  
  func pcvalue(f *_func, off int32, targetpc uintptr, cache *pcvalueCache, strict bool) int32 {
  	if off == 0 {
  		return -1
  	}
  
  	// Check the cache. This speeds up walks of deep stacks, which
  	// tend to have the same recursive functions over and over.
  	//
  	// This cache is small enough that full associativity is
  	// cheaper than doing the hashing for a less associative
  	// cache.
  	if cache != nil {
  		for _, ent := range cache.entries {
  			// We check off first because we're more
  			// likely to have multiple entries with
  			// different offsets for the same targetpc
  			// than the other way around, so we'll usually
  			// fail in the first clause.
  			if ent.off == off && ent.targetpc == targetpc {
  				return ent.val
  			}
  		}
  	}
  
  	datap := findmoduledatap(f.entry) // inefficient
  	if datap == nil {
  		if strict && panicking == 0 {
  			print("runtime: no module data for ", hex(f.entry), "\n")
  			throw("no module data")
  		}
  		return -1
  	}
  	p := datap.pclntable[off:]
  	pc := f.entry
  	val := int32(-1)
  	for {
  		var ok bool
  		p, ok = step(p, &pc, &val, pc == f.entry)
  		if !ok {
  			break
  		}
  		if targetpc < pc {
  			// Replace a random entry in the cache. Random
  			// replacement prevents a performance cliff if
  			// a recursive stack's cycle is slightly
  			// larger than the cache.
  			if cache != nil {
  				ci := fastrand() % uint32(len(cache.entries))
  				cache.entries[ci] = pcvalueCacheEnt{
  					targetpc: targetpc,
  					off:      off,
  					val:      val,
  				}
  			}
  
  			return val
  		}
  	}
  
  	// If there was a table, it should have covered all program counters.
  	// If not, something is wrong.
  	if panicking != 0 || !strict {
  		return -1
  	}
  
  	print("runtime: invalid pc-encoded table f=", funcname(f), " pc=", hex(pc), " targetpc=", hex(targetpc), " tab=", p, "\n")
  
  	p = datap.pclntable[off:]
  	pc = f.entry
  	val = -1
  	for {
  		var ok bool
  		p, ok = step(p, &pc, &val, pc == f.entry)
  		if !ok {
  			break
  		}
  		print("\tvalue=", val, " until pc=", hex(pc), "\n")
  	}
  
  	throw("invalid runtime symbol table")
  	return -1
  }
  
  func cfuncname(f *_func) *byte {
  	if f == nil || f.nameoff == 0 {
  		return nil
  	}
  	datap := findmoduledatap(f.entry) // inefficient
  	if datap == nil {
  		return nil
  	}
  	return &datap.pclntable[f.nameoff]
  }
  
  func funcname(f *_func) string {
  	return gostringnocopy(cfuncname(f))
  }
  
  func funcline1(f *_func, targetpc uintptr, strict bool) (file string, line int32) {
  	datap := findmoduledatap(f.entry) // inefficient
  	if datap == nil {
  		return "?", 0
  	}
  	fileno := int(pcvalue(f, f.pcfile, targetpc, nil, strict))
  	line = pcvalue(f, f.pcln, targetpc, nil, strict)
  	if fileno == -1 || line == -1 || fileno >= len(datap.filetab) {
  		// print("looking for ", hex(targetpc), " in ", funcname(f), " got file=", fileno, " line=", lineno, "\n")
  		return "?", 0
  	}
  	file = gostringnocopy(&datap.pclntable[datap.filetab[fileno]])
  	return
  }
  
  func funcline(f *_func, targetpc uintptr) (file string, line int32) {
  	return funcline1(f, targetpc, true)
  }
  
  func funcspdelta(f *_func, targetpc uintptr, cache *pcvalueCache) int32 {
  	x := pcvalue(f, f.pcsp, targetpc, cache, true)
  	if x&(sys.PtrSize-1) != 0 {
  		print("invalid spdelta ", funcname(f), " ", hex(f.entry), " ", hex(targetpc), " ", hex(f.pcsp), " ", x, "\n")
  	}
  	return x
  }
  
  func pcdatavalue(f *_func, table int32, targetpc uintptr, cache *pcvalueCache) int32 {
  	if table < 0 || table >= f.npcdata {
  		return -1
  	}
  	off := *(*int32)(add(unsafe.Pointer(&f.nfuncdata), unsafe.Sizeof(f.nfuncdata)+uintptr(table)*4))
  	return pcvalue(f, off, targetpc, cache, true)
  }
  
  func funcdata(f *_func, i int32) unsafe.Pointer {
  	if i < 0 || i >= f.nfuncdata {
  		return nil
  	}
  	p := add(unsafe.Pointer(&f.nfuncdata), unsafe.Sizeof(f.nfuncdata)+uintptr(f.npcdata)*4)
  	if sys.PtrSize == 8 && uintptr(p)&4 != 0 {
  		if uintptr(unsafe.Pointer(f))&4 != 0 {
  			println("runtime: misaligned func", f)
  		}
  		p = add(p, 4)
  	}
  	return *(*unsafe.Pointer)(add(p, uintptr(i)*sys.PtrSize))
  }
  
  // step advances to the next pc, value pair in the encoded table.
  func step(p []byte, pc *uintptr, val *int32, first bool) (newp []byte, ok bool) {
  	p, uvdelta := readvarint(p)
  	if uvdelta == 0 && !first {
  		return nil, false
  	}
  	if uvdelta&1 != 0 {
  		uvdelta = ^(uvdelta >> 1)
  	} else {
  		uvdelta >>= 1
  	}
  	vdelta := int32(uvdelta)
  	p, pcdelta := readvarint(p)
  	*pc += uintptr(pcdelta * sys.PCQuantum)
  	*val += vdelta
  	return p, true
  }
  
  // readvarint reads a varint from p.
  func readvarint(p []byte) (newp []byte, val uint32) {
  	var v, shift uint32
  	for {
  		b := p[0]
  		p = p[1:]
  		v |= (uint32(b) & 0x7F) << shift
  		if b&0x80 == 0 {
  			break
  		}
  		shift += 7
  	}
  	return p, v
  }
  
  type stackmap struct {
  	n        int32   // number of bitmaps
  	nbit     int32   // number of bits in each bitmap
  	bytedata [1]byte // bitmaps, each starting on a byte boundary
  }
  
  //go:nowritebarrier
  func stackmapdata(stkmap *stackmap, n int32) bitvector {
  	if n < 0 || n >= stkmap.n {
  		throw("stackmapdata: index out of range")
  	}
  	return bitvector{stkmap.nbit, (*byte)(add(unsafe.Pointer(&stkmap.bytedata), uintptr(n*((stkmap.nbit+7)/8))))}
  }
  

View as plain text