stack.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.

package runtime

import (
	"runtime/internal/atomic"
	"runtime/internal/sys"
	"unsafe"
)

/*
Stack layout parameters.
Included both by runtime (compiled via 6c) and linkers (compiled via gcc).

The per-goroutine g->stackguard is set to point StackGuard bytes
above the bottom of the stack.  Each function compares its stack
pointer against g->stackguard to check for overflow.  To cut one
instruction from the check sequence for functions with tiny frames,
the stack is allowed to protrude StackSmall bytes below the stack
guard.  Functions with large frames don't bother with the check and
always call morestack.  The sequences are (for amd64, others are
similar):

	guard = g->stackguard
	frame = function's stack frame size
	argsize = size of function arguments (call + return)

	stack frame size <= StackSmall:
		CMPQ guard, SP
		JHI 3(PC)
		MOVQ m->morearg, $(argsize << 32)
		CALL morestack(SB)

	stack frame size > StackSmall but < StackBig
		LEAQ (frame-StackSmall)(SP), R0
		CMPQ guard, R0
		JHI 3(PC)
		MOVQ m->morearg, $(argsize << 32)
		CALL morestack(SB)

	stack frame size >= StackBig:
		MOVQ m->morearg, $((argsize << 32) | frame)
		CALL morestack(SB)

The bottom StackGuard - StackSmall bytes are important: there has
to be enough room to execute functions that refuse to check for
stack overflow, either because they need to be adjacent to the
actual caller's frame (deferproc) or because they handle the imminent
stack overflow (morestack).

For example, deferproc might call malloc, which does one of the
above checks (without allocating a full frame), which might trigger
a call to morestack.  This sequence needs to fit in the bottom
section of the stack.  On amd64, morestack's frame is 40 bytes, and
deferproc's frame is 56 bytes.  That fits well within the
StackGuard - StackSmall bytes at the bottom.
The linkers explore all possible call traces involving non-splitting
functions to make sure that this limit cannot be violated.
*/

const (
	// StackSystem is a number of additional bytes to add
	// to each stack below the usual guard area for OS-specific
	// purposes like signal handling. Used on Windows, Plan 9,
	// and Darwin/ARM because they do not use a separate stack.
	_StackSystem = sys.GoosWindows*512*sys.PtrSize + sys.GoosPlan9*512 + sys.GoosDarwin*sys.GoarchArm*1024

	// The minimum size of stack used by Go code
	_StackMin = 2048

	// The minimum stack size to allocate.
	// The hackery here rounds FixedStack0 up to a power of 2.
	_FixedStack0 = _StackMin + _StackSystem
	_FixedStack1 = _FixedStack0 - 1
	_FixedStack2 = _FixedStack1 | (_FixedStack1 >> 1)
	_FixedStack3 = _FixedStack2 | (_FixedStack2 >> 2)
	_FixedStack4 = _FixedStack3 | (_FixedStack3 >> 4)
	_FixedStack5 = _FixedStack4 | (_FixedStack4 >> 8)
	_FixedStack6 = _FixedStack5 | (_FixedStack5 >> 16)
	_FixedStack  = _FixedStack6 + 1

	// Functions that need frames bigger than this use an extra
	// instruction to do the stack split check, to avoid overflow
	// in case SP - framesize wraps below zero.
	// This value can be no bigger than the size of the unmapped
	// space at zero.
	_StackBig = 4096

	// The stack guard is a pointer this many bytes above the
	// bottom of the stack.
	_StackGuard = 880*sys.StackGuardMultiplier + _StackSystem

	// After a stack split check the SP is allowed to be this
	// many bytes below the stack guard. This saves an instruction
	// in the checking sequence for tiny frames.
	_StackSmall = 128

	// The maximum number of bytes that a chain of NOSPLIT
	// functions can use.
	_StackLimit = _StackGuard - _StackSystem - _StackSmall
)

// Goroutine preemption request.
// Stored into g->stackguard0 to cause split stack check failure.
// Must be greater than any real sp.
// 0xfffffade in hex.
const (
	_StackPreempt = uintptrMask & -1314
	_StackFork    = uintptrMask & -1234
)

const (
	// stackDebug == 0: no logging
	//            == 1: logging of per-stack operations
	//            == 2: logging of per-frame operations
	//            == 3: logging of per-word updates
	//            == 4: logging of per-word reads
	stackDebug       = 0
	stackFromSystem  = 0 // allocate stacks from system memory instead of the heap
	stackFaultOnFree = 0 // old stacks are mapped noaccess to detect use after free
	stackPoisonCopy  = 0 // fill stack that should not be accessed with garbage, to detect bad dereferences during copy

	stackCache = 1

	// check the BP links during traceback.
	debugCheckBP = false
)

const (
	uintptrMask = 1<<(8*sys.PtrSize) - 1

	// Goroutine preemption request.
	// Stored into g->stackguard0 to cause split stack check failure.
	// Must be greater than any real sp.
	// 0xfffffade in hex.
	stackPreempt = uintptrMask & -1314

	// Thread is forking.
	// Stored into g->stackguard0 to cause split stack check failure.
	// Must be greater than any real sp.
	stackFork = uintptrMask & -1234
)

// Global pool of spans that have free stacks.
// Stacks are assigned an order according to size.
//     order = log_2(size/FixedStack)
// There is a free list for each order.
// TODO: one lock per order?
var stackpool [_NumStackOrders]mSpanList
var stackpoolmu mutex

// Global pool of large stack spans.
var stackLarge struct {
	lock mutex
	free [_MHeapMap_Bits]mSpanList // free lists by log_2(s.npages)
}

func stackinit() {
	if _StackCacheSize&_PageMask != 0 {
		throw("cache size must be a multiple of page size")
	}
	for i := range stackpool {
		stackpool[i].init()
	}
	for i := range stackLarge.free {
		stackLarge.free[i].init()
	}
}

// stacklog2 returns ⌊log_2(n)⌋.
func stacklog2(n uintptr) int {
	log2 := 0
	for n > 1 {
		n >>= 1
		log2++
	}
	return log2
}

// Allocates a stack from the free pool. Must be called with
// stackpoolmu held.
func stackpoolalloc(order uint8) gclinkptr {
	list := &stackpool[order]
	s := list.first
	if s == nil {
		// no free stacks. Allocate another span worth.
		s = mheap_.allocStack(_StackCacheSize >> _PageShift)
		if s == nil {
			throw("out of memory")
		}
		if s.allocCount != 0 {
			throw("bad allocCount")
		}
		if s.stackfreelist.ptr() != nil {
			throw("bad stackfreelist")
		}
		for i := uintptr(0); i < _StackCacheSize; i += _FixedStack << order {
			x := gclinkptr(s.base() + i)
			x.ptr().next = s.stackfreelist
			s.stackfreelist = x
		}
		list.insert(s)
	}
	x := s.stackfreelist
	if x.ptr() == nil {
		throw("span has no free stacks")
	}
	s.stackfreelist = x.ptr().next
	s.allocCount++
	if s.stackfreelist.ptr() == nil {
		// all stacks in s are allocated.
		list.remove(s)
	}
	return x
}

// Adds stack x to the free pool. Must be called with stackpoolmu held.
func stackpoolfree(x gclinkptr, order uint8) {
	s := mheap_.lookup(unsafe.Pointer(x))
	if s.state != _MSpanStack {
		throw("freeing stack not in a stack span")
	}
	if s.stackfreelist.ptr() == nil {
		// s will now have a free stack
		stackpool[order].insert(s)
	}
	x.ptr().next = s.stackfreelist
	s.stackfreelist = x
	s.allocCount--
	if gcphase == _GCoff && s.allocCount == 0 {
		// Span is completely free. Return it to the heap
		// immediately if we're sweeping.
		//
		// If GC is active, we delay the free until the end of
		// GC to avoid the following type of situation:
		//
		// 1) GC starts, scans a SudoG but does not yet mark the SudoG.elem pointer
		// 2) The stack that pointer points to is copied
		// 3) The old stack is freed
		// 4) The containing span is marked free
		// 5) GC attempts to mark the SudoG.elem pointer. The
		//    marking fails because the pointer looks like a
		//    pointer into a free span.
		//
		// By not freeing, we prevent step #4 until GC is done.
		stackpool[order].remove(s)
		s.stackfreelist = 0
		mheap_.freeStack(s)
	}
}

// stackcacherefill/stackcacherelease implement a global pool of stack segments.
// The pool is required to prevent unlimited growth of per-thread caches.
//
//go:systemstack
func stackcacherefill(c *mcache, order uint8) {
	if stackDebug >= 1 {
		print("stackcacherefill order=", order, "\n")
	}

	// Grab some stacks from the global cache.
	// Grab half of the allowed capacity (to prevent thrashing).
	var list gclinkptr
	var size uintptr
	lock(&stackpoolmu)
	for size < _StackCacheSize/2 {
		x := stackpoolalloc(order)
		x.ptr().next = list
		list = x
		size += _FixedStack << order
	}
	unlock(&stackpoolmu)
	c.stackcache[order].list = list
	c.stackcache[order].size = size
}

//go:systemstack
func stackcacherelease(c *mcache, order uint8) {
	if stackDebug >= 1 {
		print("stackcacherelease order=", order, "\n")
	}
	x := c.stackcache[order].list
	size := c.stackcache[order].size
	lock(&stackpoolmu)
	for size > _StackCacheSize/2 {
		y := x.ptr().next
		stackpoolfree(x, order)
		x = y
		size -= _FixedStack << order
	}
	unlock(&stackpoolmu)
	c.stackcache[order].list = x
	c.stackcache[order].size = size
}

//go:systemstack
func stackcache_clear(c *mcache) {
	if stackDebug >= 1 {
		print("stackcache clear\n")
	}
	lock(&stackpoolmu)
	for order := uint8(0); order < _NumStackOrders; order++ {
		x := c.stackcache[order].list
		for x.ptr() != nil {
			y := x.ptr().next
			stackpoolfree(x, order)
			x = y
		}
		c.stackcache[order].list = 0
		c.stackcache[order].size = 0
	}
	unlock(&stackpoolmu)
}

// stackalloc allocates an n byte stack.
//
// stackalloc must run on the system stack because it uses per-P
// resources and must not split the stack.
//
//go:systemstack
func stackalloc(n uint32) (stack, []stkbar) {
	// Stackalloc must be called on scheduler stack, so that we
	// never try to grow the stack during the code that stackalloc runs.
	// Doing so would cause a deadlock (issue 1547).
	thisg := getg()
	if thisg != thisg.m.g0 {
		throw("stackalloc not on scheduler stack")
	}
	if n&(n-1) != 0 {
		throw("stack size not a power of 2")
	}
	if stackDebug >= 1 {
		print("stackalloc ", n, "\n")
	}

	// Compute the size of stack barrier array.
	maxstkbar := gcMaxStackBarriers(int(n))
	nstkbar := unsafe.Sizeof(stkbar{}) * uintptr(maxstkbar)
	var stkbarSlice slice

	if debug.efence != 0 || stackFromSystem != 0 {
		v := sysAlloc(round(uintptr(n), _PageSize), &memstats.stacks_sys)
		if v == nil {
			throw("out of memory (stackalloc)")
		}
		top := uintptr(n) - nstkbar
		if maxstkbar != 0 {
			stkbarSlice = slice{add(v, top), 0, maxstkbar}
		}
		return stack{uintptr(v), uintptr(v) + top}, *(*[]stkbar)(unsafe.Pointer(&stkbarSlice))
	}

	// Small stacks are allocated with a fixed-size free-list allocator.
	// If we need a stack of a bigger size, we fall back on allocating
	// a dedicated span.
	var v unsafe.Pointer
	if stackCache != 0 && n < _FixedStack<<_NumStackOrders && n < _StackCacheSize {
		order := uint8(0)
		n2 := n
		for n2 > _FixedStack {
			order++
			n2 >>= 1
		}
		var x gclinkptr
		c := thisg.m.mcache
		if c == nil || thisg.m.preemptoff != "" || thisg.m.helpgc != 0 {
			// c == nil can happen in the guts of exitsyscall or
			// procresize. Just get a stack from the global pool.
			// Also don't touch stackcache during gc
			// as it's flushed concurrently.
			lock(&stackpoolmu)
			x = stackpoolalloc(order)
			unlock(&stackpoolmu)
		} else {
			x = c.stackcache[order].list
			if x.ptr() == nil {
				stackcacherefill(c, order)
				x = c.stackcache[order].list
			}
			c.stackcache[order].list = x.ptr().next
			c.stackcache[order].size -= uintptr(n)
		}
		v = unsafe.Pointer(x)
	} else {
		var s *mspan
		npage := uintptr(n) >> _PageShift
		log2npage := stacklog2(npage)

		// Try to get a stack from the large stack cache.
		lock(&stackLarge.lock)
		if !stackLarge.free[log2npage].isEmpty() {
			s = stackLarge.free[log2npage].first
			stackLarge.free[log2npage].remove(s)
		}
		unlock(&stackLarge.lock)

		if s == nil {
			// Allocate a new stack from the heap.
			s = mheap_.allocStack(npage)
			if s == nil {
				throw("out of memory")
			}
		}
		v = unsafe.Pointer(s.base())
	}

	if raceenabled {
		racemalloc(v, uintptr(n))
	}
	if msanenabled {
		msanmalloc(v, uintptr(n))
	}
	if stackDebug >= 1 {
		print("  allocated ", v, "\n")
	}
	top := uintptr(n) - nstkbar
	if maxstkbar != 0 {
		stkbarSlice = slice{add(v, top), 0, maxstkbar}
	}
	return stack{uintptr(v), uintptr(v) + top}, *(*[]stkbar)(unsafe.Pointer(&stkbarSlice))
}

// stackfree frees an n byte stack allocation at stk.
//
// stackfree must run on the system stack because it uses per-P
// resources and must not split the stack.
//
//go:systemstack
func stackfree(stk stack, n uintptr) {
	gp := getg()
	v := unsafe.Pointer(stk.lo)
	if n&(n-1) != 0 {
		throw("stack not a power of 2")
	}
	if stk.lo+n < stk.hi {
		throw("bad stack size")
	}
	if stackDebug >= 1 {
		println("stackfree", v, n)
		memclrNoHeapPointers(v, n) // for testing, clobber stack data
	}
	if debug.efence != 0 || stackFromSystem != 0 {
		if debug.efence != 0 || stackFaultOnFree != 0 {
			sysFault(v, n)
		} else {
			sysFree(v, n, &memstats.stacks_sys)
		}
		return
	}
	if msanenabled {
		msanfree(v, n)
	}
	if stackCache != 0 && n < _FixedStack<<_NumStackOrders && n < _StackCacheSize {
		order := uint8(0)
		n2 := n
		for n2 > _FixedStack {
			order++
			n2 >>= 1
		}
		x := gclinkptr(v)
		c := gp.m.mcache
		if c == nil || gp.m.preemptoff != "" || gp.m.helpgc != 0 {
			lock(&stackpoolmu)
			stackpoolfree(x, order)
			unlock(&stackpoolmu)
		} else {
			if c.stackcache[order].size >= _StackCacheSize {
				stackcacherelease(c, order)
			}
			x.ptr().next = c.stackcache[order].list
			c.stackcache[order].list = x
			c.stackcache[order].size += n
		}
	} else {
		s := mheap_.lookup(v)
		if s.state != _MSpanStack {
			println(hex(s.base()), v)
			throw("bad span state")
		}
		if gcphase == _GCoff {
			// Free the stack immediately if we're
			// sweeping.
			mheap_.freeStack(s)
		} else {
			// If the GC is running, we can't return a
			// stack span to the heap because it could be
			// reused as a heap span, and this state
			// change would race with GC. Add it to the
			// large stack cache instead.
			log2npage := stacklog2(s.npages)
			lock(&stackLarge.lock)
			stackLarge.free[log2npage].insert(s)
			unlock(&stackLarge.lock)
		}
	}
}

var maxstacksize uintptr = 1 << 20 // enough until runtime.main sets it for real

var ptrnames = []string{
	0: "scalar",
	1: "ptr",
}

// Stack frame layout
//
// (x86)
// +------------------+
// | args from caller |
// +------------------+ <- frame->argp
// |  return address  |
// +------------------+
// |  caller's BP (*) | (*) if framepointer_enabled && varp < sp
// +------------------+ <- frame->varp
// |     locals       |
// +------------------+
// |  args to callee  |
// +------------------+ <- frame->sp
//
// (arm)
// +------------------+
// | args from caller |
// +------------------+ <- frame->argp
// | caller's retaddr |
// +------------------+ <- frame->varp
// |     locals       |
// +------------------+
// |  args to callee  |
// +------------------+
// |  return address  |
// +------------------+ <- frame->sp

type adjustinfo struct {
	old   stack
	delta uintptr // ptr distance from old to new stack (newbase - oldbase)
	cache pcvalueCache

	// sghi is the highest sudog.elem on the stack.
	sghi uintptr
}

// Adjustpointer checks whether *vpp is in the old stack described by adjinfo.
// If so, it rewrites *vpp to point into the new stack.
func adjustpointer(adjinfo *adjustinfo, vpp unsafe.Pointer) {
	pp := (*uintptr)(vpp)
	p := *pp
	if stackDebug >= 4 {
		print("        ", pp, ":", hex(p), "\n")
	}
	if adjinfo.old.lo <= p && p < adjinfo.old.hi {
		*pp = p + adjinfo.delta
		if stackDebug >= 3 {
			print("        adjust ptr ", pp, ":", hex(p), " -> ", hex(*pp), "\n")
		}
	}
}

// Information from the compiler about the layout of stack frames.
type bitvector struct {
	n        int32 // # of bits
	bytedata *uint8
}

type gobitvector struct {
	n        uintptr
	bytedata []uint8
}

func gobv(bv bitvector) gobitvector {
	return gobitvector{
		uintptr(bv.n),
		(*[1 << 30]byte)(unsafe.Pointer(bv.bytedata))[:(bv.n+7)/8],
	}
}

func ptrbit(bv *gobitvector, i uintptr) uint8 {
	return (bv.bytedata[i/8] >> (i % 8)) & 1
}

// bv describes the memory starting at address scanp.
// Adjust any pointers contained therein.
func adjustpointers(scanp unsafe.Pointer, cbv *bitvector, adjinfo *adjustinfo, f *_func) {
	bv := gobv(*cbv)
	minp := adjinfo.old.lo
	maxp := adjinfo.old.hi
	delta := adjinfo.delta
	num := bv.n
	// If this frame might contain channel receive slots, use CAS
	// to adjust pointers. If the slot hasn't been received into
	// yet, it may contain stack pointers and a concurrent send
	// could race with adjusting those pointers. (The sent value
	// itself can never contain stack pointers.)
	useCAS := uintptr(scanp) < adjinfo.sghi
	for i := uintptr(0); i < num; i++ {
		if stackDebug >= 4 {
			print("        ", add(scanp, i*sys.PtrSize), ":", ptrnames[ptrbit(&bv, i)], ":", hex(*(*uintptr)(add(scanp, i*sys.PtrSize))), " # ", i, " ", bv.bytedata[i/8], "\n")
		}
		if ptrbit(&bv, i) == 1 {
			pp := (*uintptr)(add(scanp, i*sys.PtrSize))
		retry:
			p := *pp
			if f != nil && 0 < p && p < minLegalPointer && debug.invalidptr != 0 {
				// Looks like a junk value in a pointer slot.
				// Live analysis wrong?
				getg().m.traceback = 2
				print("runtime: bad pointer in frame ", funcname(f), " at ", pp, ": ", hex(p), "\n")
				throw("invalid pointer found on stack")
			}
			if minp <= p && p < maxp {
				if stackDebug >= 3 {
					print("adjust ptr ", hex(p), " ", funcname(f), "\n")
				}
				if useCAS {
					ppu := (*unsafe.Pointer)(unsafe.Pointer(pp))
					if !atomic.Casp1(ppu, unsafe.Pointer(p), unsafe.Pointer(p+delta)) {
						goto retry
					}
				} else {
					*pp = p + delta
				}
			}
		}
	}
}

// Note: the argument/return area is adjusted by the callee.
func adjustframe(frame *stkframe, arg unsafe.Pointer) bool {
	adjinfo := (*adjustinfo)(arg)
	targetpc := frame.continpc
	if targetpc == 0 {
		// Frame is dead.
		return true
	}
	f := frame.fn
	if stackDebug >= 2 {
		print("    adjusting ", funcname(f), " frame=[", hex(frame.sp), ",", hex(frame.fp), "] pc=", hex(frame.pc), " continpc=", hex(frame.continpc), "\n")
	}
	if f.entry == systemstack_switchPC {
		// A special routine at the bottom of stack of a goroutine that does an systemstack call.
		// We will allow it to be copied even though we don't
		// have full GC info for it (because it is written in asm).
		return true
	}
	if targetpc != f.entry {
		targetpc--
	}
	pcdata := pcdatavalue(f, _PCDATA_StackMapIndex, targetpc, &adjinfo.cache)
	if pcdata == -1 {
		pcdata = 0 // in prologue
	}

	// Adjust local variables if stack frame has been allocated.
	size := frame.varp - frame.sp
	var minsize uintptr
	switch sys.ArchFamily {
	case sys.ARM64:
		minsize = sys.SpAlign
	default:
		minsize = sys.MinFrameSize
	}
	if size > minsize {
		var bv bitvector
		stackmap := (*stackmap)(funcdata(f, _FUNCDATA_LocalsPointerMaps))
		if stackmap == nil || stackmap.n <= 0 {
			print("runtime: frame ", funcname(f), " untyped locals ", hex(frame.varp-size), "+", hex(size), "\n")
			throw("missing stackmap")
		}
		// Locals bitmap information, scan just the pointers in locals.
		if pcdata < 0 || pcdata >= stackmap.n {
			// don't know where we are
			print("runtime: pcdata is ", pcdata, " and ", stackmap.n, " locals stack map entries for ", funcname(f), " (targetpc=", targetpc, ")\n")
			throw("bad symbol table")
		}
		bv = stackmapdata(stackmap, pcdata)
		size = uintptr(bv.n) * sys.PtrSize
		if stackDebug >= 3 {
			print("      locals ", pcdata, "/", stackmap.n, " ", size/sys.PtrSize, " words ", bv.bytedata, "\n")
		}
		adjustpointers(unsafe.Pointer(frame.varp-size), &bv, adjinfo, f)
	}

	// Adjust saved base pointer if there is one.
	if sys.ArchFamily == sys.AMD64 && frame.argp-frame.varp == 2*sys.RegSize {
		if !framepointer_enabled {
			print("runtime: found space for saved base pointer, but no framepointer experiment\n")
			print("argp=", hex(frame.argp), " varp=", hex(frame.varp), "\n")
			throw("bad frame layout")
		}
		if stackDebug >= 3 {
			print("      saved bp\n")
		}
		if debugCheckBP {
			// Frame pointers should always point to the next higher frame on
			// the Go stack (or be nil, for the top frame on the stack).
			bp := *(*uintptr)(unsafe.Pointer(frame.varp))
			if bp != 0 && (bp < adjinfo.old.lo || bp >= adjinfo.old.hi) {
				println("runtime: found invalid frame pointer")
				print("bp=", hex(bp), " min=", hex(adjinfo.old.lo), " max=", hex(adjinfo.old.hi), "\n")
				throw("bad frame pointer")
			}
		}
		adjustpointer(adjinfo, unsafe.Pointer(frame.varp))
	}

	// Adjust arguments.
	if frame.arglen > 0 {
		var bv bitvector
		if frame.argmap != nil {
			bv = *frame.argmap
		} else {
			stackmap := (*stackmap)(funcdata(f, _FUNCDATA_ArgsPointerMaps))
			if stackmap == nil || stackmap.n <= 0 {
				print("runtime: frame ", funcname(f), " untyped args ", frame.argp, "+", frame.arglen, "\n")
				throw("missing stackmap")
			}
			if pcdata < 0 || pcdata >= stackmap.n {
				// don't know where we are
				print("runtime: pcdata is ", pcdata, " and ", stackmap.n, " args stack map entries for ", funcname(f), " (targetpc=", targetpc, ")\n")
				throw("bad symbol table")
			}
			bv = stackmapdata(stackmap, pcdata)
		}
		if stackDebug >= 3 {
			print("      args\n")
		}
		adjustpointers(unsafe.Pointer(frame.argp), &bv, adjinfo, nil)
	}
	return true
}

func adjustctxt(gp *g, adjinfo *adjustinfo) {
	adjustpointer(adjinfo, unsafe.Pointer(&gp.sched.ctxt))
	if !framepointer_enabled {
		return
	}
	if debugCheckBP {
		bp := gp.sched.bp
		if bp != 0 && (bp < adjinfo.old.lo || bp >= adjinfo.old.hi) {
			println("runtime: found invalid top frame pointer")
			print("bp=", hex(bp), " min=", hex(adjinfo.old.lo), " max=", hex(adjinfo.old.hi), "\n")
			throw("bad top frame pointer")
		}
	}
	adjustpointer(adjinfo, unsafe.Pointer(&gp.sched.bp))
}

func adjustdefers(gp *g, adjinfo *adjustinfo) {
	// Adjust defer argument blocks the same way we adjust active stack frames.
	tracebackdefers(gp, adjustframe, noescape(unsafe.Pointer(adjinfo)))

	// Adjust pointers in the Defer structs.
	// Defer structs themselves are never on the stack.
	for d := gp._defer; d != nil; d = d.link {
		adjustpointer(adjinfo, unsafe.Pointer(&d.fn))
		adjustpointer(adjinfo, unsafe.Pointer(&d.sp))
		adjustpointer(adjinfo, unsafe.Pointer(&d._panic))
	}
}

func adjustpanics(gp *g, adjinfo *adjustinfo) {
	// Panics are on stack and already adjusted.
	// Update pointer to head of list in G.
	adjustpointer(adjinfo, unsafe.Pointer(&gp._panic))
}

func adjustsudogs(gp *g, adjinfo *adjustinfo) {
	// the data elements pointed to by a SudoG structure
	// might be in the stack.
	for s := gp.waiting; s != nil; s = s.waitlink {
		adjustpointer(adjinfo, unsafe.Pointer(&s.elem))
		adjustpointer(adjinfo, unsafe.Pointer(&s.selectdone))
	}
}

func adjuststkbar(gp *g, adjinfo *adjustinfo) {
	for i := int(gp.stkbarPos); i < len(gp.stkbar); i++ {
		adjustpointer(adjinfo, unsafe.Pointer(&gp.stkbar[i].savedLRPtr))
	}
}

func fillstack(stk stack, b byte) {
	for p := stk.lo; p < stk.hi; p++ {
		*(*byte)(unsafe.Pointer(p)) = b
	}
}

func findsghi(gp *g, stk stack) uintptr {
	var sghi uintptr
	for sg := gp.waiting; sg != nil; sg = sg.waitlink {
		p := uintptr(sg.elem) + uintptr(sg.c.elemsize)
		if stk.lo <= p && p < stk.hi && p > sghi {
			sghi = p
		}
		p = uintptr(unsafe.Pointer(sg.selectdone)) + unsafe.Sizeof(sg.selectdone)
		if stk.lo <= p && p < stk.hi && p > sghi {
			sghi = p
		}
	}
	return sghi
}

// syncadjustsudogs adjusts gp's sudogs and copies the part of gp's
// stack they refer to while synchronizing with concurrent channel
// operations. It returns the number of bytes of stack copied.
func syncadjustsudogs(gp *g, used uintptr, adjinfo *adjustinfo) uintptr {
	if gp.waiting == nil {
		return 0
	}

	// Lock channels to prevent concurrent send/receive.
	// It's important that we *only* do this for async
	// copystack; otherwise, gp may be in the middle of
	// putting itself on wait queues and this would
	// self-deadlock.
	var lastc *hchan
	for sg := gp.waiting; sg != nil; sg = sg.waitlink {
		if sg.c != lastc {
			lock(&sg.c.lock)
		}
		lastc = sg.c
	}

	// Adjust sudogs.
	adjustsudogs(gp, adjinfo)

	// Copy the part of the stack the sudogs point in to
	// while holding the lock to prevent races on
	// send/receive slots.
	var sgsize uintptr
	if adjinfo.sghi != 0 {
		oldBot := adjinfo.old.hi - used
		newBot := oldBot + adjinfo.delta
		sgsize = adjinfo.sghi - oldBot
		memmove(unsafe.Pointer(newBot), unsafe.Pointer(oldBot), sgsize)
	}

	// Unlock channels.
	lastc = nil
	for sg := gp.waiting; sg != nil; sg = sg.waitlink {
		if sg.c != lastc {
			unlock(&sg.c.lock)
		}
		lastc = sg.c
	}

	return sgsize
}

// Copies gp's stack to a new stack of a different size.
// Caller must have changed gp status to Gcopystack.
//
// If sync is true, this is a self-triggered stack growth and, in
// particular, no other G may be writing to gp's stack (e.g., via a
// channel operation). If sync is false, copystack protects against
// concurrent channel operations.
func copystack(gp *g, newsize uintptr, sync bool) {
	if gp.syscallsp != 0 {
		throw("stack growth not allowed in system call")
	}
	old := gp.stack
	if old.lo == 0 {
		throw("nil stackbase")
	}
	used := old.hi - gp.sched.sp

	// allocate new stack
	new, newstkbar := stackalloc(uint32(newsize))
	if stackPoisonCopy != 0 {
		fillstack(new, 0xfd)
	}
	if stackDebug >= 1 {
		print("copystack gp=", gp, " [", hex(old.lo), " ", hex(old.hi-used), " ", hex(old.hi), "]/", gp.stackAlloc, " -> [", hex(new.lo), " ", hex(new.hi-used), " ", hex(new.hi), "]/", newsize, "\n")
	}

	// Compute adjustment.
	var adjinfo adjustinfo
	adjinfo.old = old
	adjinfo.delta = new.hi - old.hi

	// Adjust sudogs, synchronizing with channel ops if necessary.
	ncopy := used
	if sync {
		adjustsudogs(gp, &adjinfo)
	} else {
		// sudogs can point in to the stack. During concurrent
		// shrinking, these areas may be written to. Find the
		// highest such pointer so we can handle everything
		// there and below carefully. (This shouldn't be far
		// from the bottom of the stack, so there's little
		// cost in handling everything below it carefully.)
		adjinfo.sghi = findsghi(gp, old)

		// Synchronize with channel ops and copy the part of
		// the stack they may interact with.
		ncopy -= syncadjustsudogs(gp, used, &adjinfo)
	}

	// Copy the stack (or the rest of it) to the new location
	memmove(unsafe.Pointer(new.hi-ncopy), unsafe.Pointer(old.hi-ncopy), ncopy)

	// Disallow sigprof scans of this stack and block if there's
	// one in progress.
	gcLockStackBarriers(gp)

	// Adjust remaining structures that have pointers into stacks.
	// We have to do most of these before we traceback the new
	// stack because gentraceback uses them.
	adjustctxt(gp, &adjinfo)
	adjustdefers(gp, &adjinfo)
	adjustpanics(gp, &adjinfo)
	adjuststkbar(gp, &adjinfo)
	if adjinfo.sghi != 0 {
		adjinfo.sghi += adjinfo.delta
	}

	// copy old stack barriers to new stack barrier array
	newstkbar = newstkbar[:len(gp.stkbar)]
	copy(newstkbar, gp.stkbar)

	// Swap out old stack for new one
	gp.stack = new
	gp.stackguard0 = new.lo + _StackGuard // NOTE: might clobber a preempt request
	gp.sched.sp = new.hi - used
	oldsize := gp.stackAlloc
	gp.stackAlloc = newsize
	gp.stkbar = newstkbar
	gp.stktopsp += adjinfo.delta

	// Adjust pointers in the new stack.
	gentraceback(^uintptr(0), ^uintptr(0), 0, gp, 0, nil, 0x7fffffff, adjustframe, noescape(unsafe.Pointer(&adjinfo)), 0)

	gcUnlockStackBarriers(gp)

	// free old stack
	if stackPoisonCopy != 0 {
		fillstack(old, 0xfc)
	}
	stackfree(old, oldsize)
}

// round x up to a power of 2.
func round2(x int32) int32 {
	s := uint(0)
	for 1<<s < x {
		s++
	}
	return 1 << s
}

// Called from runtime·morestack when more stack is needed.
// Allocate larger stack and relocate to new stack.
// Stack growth is multiplicative, for constant amortized cost.
//
// g->atomicstatus will be Grunning or Gscanrunning upon entry.
// If the GC is trying to stop this g then it will set preemptscan to true.
//
// ctxt is the value of the context register on morestack. newstack
// will write it to g.sched.ctxt.
func newstack(ctxt unsafe.Pointer) {
	thisg := getg()
	// TODO: double check all gp. shouldn't be getg().
	if thisg.m.morebuf.g.ptr().stackguard0 == stackFork {
		throw("stack growth after fork")
	}
	if thisg.m.morebuf.g.ptr() != thisg.m.curg {
		print("runtime: newstack called from g=", hex(thisg.m.morebuf.g), "\n"+"\tm=", thisg.m, " m->curg=", thisg.m.curg, " m->g0=", thisg.m.g0, " m->gsignal=", thisg.m.gsignal, "\n")
		morebuf := thisg.m.morebuf
		traceback(morebuf.pc, morebuf.sp, morebuf.lr, morebuf.g.ptr())
		throw("runtime: wrong goroutine in newstack")
	}

	gp := thisg.m.curg
	// Write ctxt to gp.sched. We do this here instead of in
	// morestack so it has the necessary write barrier.
	gp.sched.ctxt = ctxt

	if thisg.m.curg.throwsplit {
		// Update syscallsp, syscallpc in case traceback uses them.
		morebuf := thisg.m.morebuf
		gp.syscallsp = morebuf.sp
		gp.syscallpc = morebuf.pc
		print("runtime: newstack sp=", hex(gp.sched.sp), " stack=[", hex(gp.stack.lo), ", ", hex(gp.stack.hi), "]\n",
			"\tmorebuf={pc:", hex(morebuf.pc), " sp:", hex(morebuf.sp), " lr:", hex(morebuf.lr), "}\n",
			"\tsched={pc:", hex(gp.sched.pc), " sp:", hex(gp.sched.sp), " lr:", hex(gp.sched.lr), " ctxt:", gp.sched.ctxt, "}\n")

		traceback(morebuf.pc, morebuf.sp, morebuf.lr, gp)
		throw("runtime: stack split at bad time")
	}

	morebuf := thisg.m.morebuf
	thisg.m.morebuf.pc = 0
	thisg.m.morebuf.lr = 0
	thisg.m.morebuf.sp = 0
	thisg.m.morebuf.g = 0

	// NOTE: stackguard0 may change underfoot, if another thread
	// is about to try to preempt gp. Read it just once and use that same
	// value now and below.