delve/pkg/proc/breakpoints.go

619 lines
18 KiB
Go
Raw Normal View History

2015-06-12 19:49:23 +00:00
package proc
2015-01-14 02:37:10 +00:00
import (
"errors"
"fmt"
"go/ast"
"go/constant"
"go/parser"
"go/token"
"reflect"
)
2015-01-14 02:37:10 +00:00
const (
// UnrecoveredPanic is the name given to the unrecovered panic breakpoint.
UnrecoveredPanic = "unrecovered-panic"
// FatalThrow is the name given to the breakpoint triggered when the target
// process dies because of a fatal runtime error.
FatalThrow = "runtime-fatal-throw"
unrecoveredPanicID = -1
fatalThrowID = -2
)
// Breakpoint represents a physical breakpoint. Stores information on the break
2015-01-14 02:37:10 +00:00
// point including the byte of data that originally was stored at that
// address.
2015-06-12 19:32:32 +00:00
type Breakpoint struct {
// File & line information for printing.
2015-01-14 02:37:10 +00:00
FunctionName string
File string
Line int
Addr uint64 // Address breakpoint is set for.
OriginalData []byte // If software breakpoint, the data we replace with breakpoint instruction.
Name string // User defined name of the breakpoint
LogicalID int // ID of the logical breakpoint that owns this physical breakpoint
WatchExpr string
WatchType WatchType
HWBreakIndex uint8 // hardware breakpoint index
// Kind describes whether this is an internal breakpoint (for next'ing or
// stepping).
// A single breakpoint can be both a UserBreakpoint and some kind of
// internal breakpoint, but it can not be two different kinds of internal
// breakpoint.
Kind BreakpointKind
// Breakpoint information
Tracepoint bool // Tracepoint flag
TraceReturn bool
Goroutine bool // Retrieve goroutine information
Stacktrace int // Number of stack frames to retrieve
Variables []string // Variables to evaluate
LoadArgs *LoadConfig
LoadLocals *LoadConfig
HitCount map[int]uint64 // Number of times a breakpoint has been reached in a certain goroutine
TotalHitCount uint64 // Number of times a breakpoint has been reached
// DeferReturns: when kind == NextDeferBreakpoint this breakpoint
// will also check if the caller is runtime.gopanic or if the return
// address is in the DeferReturns array.
// Next uses NextDeferBreakpoints for the breakpoint it sets on the
// deferred function, DeferReturns is populated with the
// addresses of calls to runtime.deferreturn in the current
// function. This ensures that the breakpoint on the deferred
// function only triggers on panic or on the defer call to
// the function, not when the function is called directly
DeferReturns []uint64
// Cond: if not nil the breakpoint will be triggered only if evaluating Cond returns true
Cond ast.Expr
// internalCond is the same as Cond but used for the condition of internal breakpoints
internalCond ast.Expr
// HitCond: if not nil the breakpoint will be triggered only if the evaluated HitCond returns
// true with the TotalHitCount.
HitCond *struct {
Op token.Token
Val int
}
// ReturnInfo describes how to collect return variables when this
// breakpoint is hit as a return breakpoint.
returnInfo *returnBreakpointInfo
}
// BreakpointKind determines the behavior of delve when the
// breakpoint is reached.
type BreakpointKind uint16
const (
// UserBreakpoint is a user set breakpoint
UserBreakpoint BreakpointKind = (1 << iota)
// NextBreakpoint is a breakpoint set by Next, Continue
// will stop on it and delete it
NextBreakpoint
// NextDeferBreakpoint is a breakpoint set by Next on the
// first deferred function. In addition to checking their condition
// breakpoints of this kind will also check that the function has been
// called by runtime.gopanic or through runtime.deferreturn.
NextDeferBreakpoint
// StepBreakpoint is a breakpoint set by Step on a CALL instruction,
// Continue will set a new breakpoint (of NextBreakpoint kind) on the
// destination of CALL, delete this breakpoint and then continue again
StepBreakpoint
)
// WatchType is the watchpoint type
type WatchType uint8
const (
WatchRead WatchType = 1 << iota
WatchWrite
)
// Read returns true if the hardware breakpoint should trigger on memory reads.
func (wtype WatchType) Read() bool {
return wtype&WatchRead != 0
}
// Write returns true if the hardware breakpoint should trigger on memory writes.
func (wtype WatchType) Write() bool {
return wtype&WatchWrite != 0
}
// Size returns the size in bytes of the hardware breakpoint.
func (wtype WatchType) Size() int {
return int(wtype >> 4)
}
// withSize returns a new HWBreakType with the size set to the specified value
func (wtype WatchType) withSize(sz uint8) WatchType {
return WatchType((sz << 4) | uint8(wtype&0xf))
}
var ErrHWBreakUnsupported = errors.New("hardware breakpoints not implemented")
2015-06-12 19:32:32 +00:00
func (bp *Breakpoint) String() string {
return fmt.Sprintf("Breakpoint %d at %#v %s:%d (%d)", bp.LogicalID, bp.Addr, bp.File, bp.Line, bp.TotalHitCount)
2015-01-14 02:37:10 +00:00
}
2016-01-10 08:57:52 +00:00
// BreakpointExistsError is returned when trying to set a breakpoint at
2015-01-14 02:37:10 +00:00
// an address that already has a breakpoint set for it.
2015-06-12 19:32:32 +00:00
type BreakpointExistsError struct {
File string
Line int
Addr uint64
2015-01-14 02:37:10 +00:00
}
2015-06-12 19:32:32 +00:00
func (bpe BreakpointExistsError) Error() string {
return fmt.Sprintf("Breakpoint exists at %s:%d at %x", bpe.File, bpe.Line, bpe.Addr)
2015-01-14 02:37:10 +00:00
}
// InvalidAddressError represents the result of
// attempting to set a breakpoint at an invalid address.
type InvalidAddressError struct {
Address uint64
2015-01-14 02:37:10 +00:00
}
func (iae InvalidAddressError) Error() string {
return fmt.Sprintf("Invalid address %#v\n", iae.Address)
}
type returnBreakpointInfo struct {
retFrameCond ast.Expr
fn *Function
frameOffset int64
spOffset int64
}
// CheckCondition evaluates bp's condition on thread.
func (bp *Breakpoint) CheckCondition(thread Thread) BreakpointState {
bpstate := BreakpointState{Breakpoint: bp, Active: false, Internal: false, CondError: nil}
bpstate.checkCond(thread)
// Update the breakpoint hit counts.
if bpstate.Breakpoint != nil && bpstate.Active {
if g, err := GetG(thread); err == nil {
bpstate.HitCount[g.ID]++
}
bpstate.TotalHitCount++
}
bpstate.checkHitCond(thread)
return bpstate
}
func (bpstate *BreakpointState) checkCond(thread Thread) {
if bpstate.Cond == nil && bpstate.internalCond == nil {
bpstate.Active = true
bpstate.Internal = bpstate.IsInternal()
return
}
nextDeferOk := true
if bpstate.Kind&NextDeferBreakpoint != 0 {
var err error
frames, err := ThreadStacktrace(thread, 2)
if err == nil {
nextDeferOk = isPanicCall(frames)
if !nextDeferOk {
nextDeferOk, _ = isDeferReturnCall(frames, bpstate.DeferReturns)
}
}
}
if bpstate.IsInternal() {
// Check internalCondition if this is also an internal breakpoint
bpstate.Active, bpstate.CondError = evalBreakpointCondition(thread, bpstate.internalCond)
bpstate.Active = bpstate.Active && nextDeferOk
if bpstate.Active || bpstate.CondError != nil {
bpstate.Internal = true
return
}
}
if bpstate.IsUser() {
// Check normal condition if this is also a user breakpoint
bpstate.Active, bpstate.CondError = evalBreakpointCondition(thread, bpstate.Cond)
}
}
// checkHitCond evaluates bp's hit condition on thread.
func (bpstate *BreakpointState) checkHitCond(thread Thread) {
if bpstate.HitCond == nil || !bpstate.Active || bpstate.Internal {
return
}
// Evaluate the breakpoint condition.
switch bpstate.HitCond.Op {
case token.EQL:
bpstate.Active = int(bpstate.TotalHitCount) == bpstate.HitCond.Val
case token.NEQ:
bpstate.Active = int(bpstate.TotalHitCount) != bpstate.HitCond.Val
case token.GTR:
bpstate.Active = int(bpstate.TotalHitCount) > bpstate.HitCond.Val
case token.LSS:
bpstate.Active = int(bpstate.TotalHitCount) < bpstate.HitCond.Val
case token.GEQ:
bpstate.Active = int(bpstate.TotalHitCount) >= bpstate.HitCond.Val
case token.LEQ:
bpstate.Active = int(bpstate.TotalHitCount) <= bpstate.HitCond.Val
case token.REM:
bpstate.Active = int(bpstate.TotalHitCount)%bpstate.HitCond.Val == 0
}
}
func isPanicCall(frames []Stackframe) bool {
return len(frames) >= 3 && frames[2].Current.Fn != nil && frames[2].Current.Fn.Name == "runtime.gopanic"
}
func isDeferReturnCall(frames []Stackframe, deferReturns []uint64) (bool, uint64) {
if len(frames) >= 1 {
for _, pc := range deferReturns {
if frames[0].Ret == pc {
return true, pc
}
}
}
return false, 0
}
// IsInternal returns true if bp is an internal breakpoint.
// User-set breakpoints can overlap with internal breakpoints, in that case
// both IsUser and IsInternal will be true.
func (bp *Breakpoint) IsInternal() bool {
return bp.Kind != UserBreakpoint
}
// IsUser returns true if bp is a user-set breakpoint.
// User-set breakpoints can overlap with internal breakpoints, in that case
// both IsUser and IsInternal will be true.
func (bp *Breakpoint) IsUser() bool {
return bp.Kind&UserBreakpoint != 0
}
func evalBreakpointCondition(thread Thread, cond ast.Expr) (bool, error) {
if cond == nil {
return true, nil
}
scope, err := GoroutineScope(thread)
if err != nil {
scope, err = ThreadScope(thread)
if err != nil {
return true, err
}
}
v, err := scope.evalAST(cond)
if err != nil {
return true, fmt.Errorf("error evaluating expression: %v", err)
}
if v.Kind != reflect.Bool {
return true, errors.New("condition expression not boolean")
}
v.loadValue(loadFullValue)
if v.Unreadable != nil {
return true, fmt.Errorf("condition expression unreadable: %v", v.Unreadable)
}
return constant.BoolVal(v.Value), nil
}
2016-01-10 08:57:52 +00:00
// NoBreakpointError is returned when trying to
// clear a breakpoint that does not exist.
2015-06-12 19:32:32 +00:00
type NoBreakpointError struct {
Addr uint64
}
2015-06-12 19:32:32 +00:00
func (nbp NoBreakpointError) Error() string {
return fmt.Sprintf("no breakpoint at %#v", nbp.Addr)
}
// BreakpointMap represents an (address, breakpoint) map.
type BreakpointMap struct {
M map[uint64]*Breakpoint
breakpointIDCounter int
internalBreakpointIDCounter int
}
// NewBreakpointMap creates a new BreakpointMap.
func NewBreakpointMap() BreakpointMap {
return BreakpointMap{
M: make(map[uint64]*Breakpoint),
}
}
// SetBreakpoint sets a breakpoint at addr, and stores it in the process wide
// break point table.
func (t *Target) SetBreakpoint(addr uint64, kind BreakpointKind, cond ast.Expr) (*Breakpoint, error) {
return t.setBreakpointInternal(addr, kind, 0, cond)
}
// SetWatchpoint sets a data breakpoint at addr and stores it in the
// process wide break point table.
func (t *Target) SetWatchpoint(scope *EvalScope, expr string, wtype WatchType, cond ast.Expr) (*Breakpoint, error) {
if (wtype&WatchWrite == 0) && (wtype&WatchRead == 0) {
return nil, errors.New("at least one of read and write must be set for watchpoint")
}
n, err := parser.ParseExpr(expr)
if err != nil {
return nil, err
}
xv, err := scope.evalAST(n)
if err != nil {
return nil, err
}
if xv.Addr == 0 || xv.Flags&VariableFakeAddress != 0 || xv.DwarfType == nil {
return nil, fmt.Errorf("can not watch %q", expr)
}
if xv.Unreadable != nil {
return nil, fmt.Errorf("expression %q is unreadable: %v", expr, xv.Unreadable)
}
if xv.Kind == reflect.UnsafePointer || xv.Kind == reflect.Invalid {
return nil, fmt.Errorf("can not watch variable of type %s", xv.Kind.String())
}
sz := xv.DwarfType.Size()
if sz <= 0 || sz > int64(t.BinInfo().Arch.PtrSize()) {
//TODO(aarzilli): it is reasonable to expect to be able to watch string
//and interface variables and we could support it by watching certain
//member fields here.
return nil, fmt.Errorf("can not watch variable of type %s", xv.DwarfType.String())
}
if xv.Addr >= scope.g.stack.lo && xv.Addr < scope.g.stack.hi {
//TODO(aarzilli): support watching stack variables
return nil, errors.New("can not watch stack allocated variable")
}
bp, err := t.setBreakpointInternal(xv.Addr, UserBreakpoint, wtype.withSize(uint8(sz)), cond)
if bp != nil {
bp.WatchExpr = expr
}
return bp, err
}
func (t *Target) setBreakpointInternal(addr uint64, kind BreakpointKind, wtype WatchType, cond ast.Expr) (*Breakpoint, error) {
if valid, err := t.Valid(); !valid {
return nil, err
}
bpmap := t.Breakpoints()
if bp, ok := bpmap.M[addr]; ok {
// We can overlap one internal breakpoint with one user breakpoint, we
// need to support this otherwise a conditional breakpoint can mask a
// breakpoint set by next or step.
if (kind != UserBreakpoint && bp.Kind != UserBreakpoint) || (kind == UserBreakpoint && bp.IsUser()) {
return bp, BreakpointExistsError{bp.File, bp.Line, bp.Addr}
}
bp.Kind |= kind
if kind != UserBreakpoint {
bp.internalCond = cond
} else {
bp.Cond = cond
}
return bp, nil
}
f, l, fn := t.BinInfo().PCToLine(uint64(addr))
fnName := ""
if fn != nil {
fnName = fn.Name
}
hwidx := uint8(0)
if wtype != 0 {
m := make(map[uint8]bool)
for _, bp := range bpmap.M {
if bp.WatchType != 0 {
m[bp.HWBreakIndex] = true
}
}
for hwidx = 0; true; hwidx++ {
if !m[hwidx] {
break
}
}
}
newBreakpoint := &Breakpoint{
FunctionName: fnName,
WatchType: wtype,
HWBreakIndex: hwidx,
File: f,
Line: l,
Addr: addr,
Kind: kind,
HitCount: map[int]uint64{},
}
err := t.proc.WriteBreakpoint(newBreakpoint)
if err != nil {
return nil, err
}
if kind != UserBreakpoint {
bpmap.internalBreakpointIDCounter++
newBreakpoint.LogicalID = bpmap.internalBreakpointIDCounter
newBreakpoint.internalCond = cond
} else {
bpmap.breakpointIDCounter++
newBreakpoint.LogicalID = bpmap.breakpointIDCounter
newBreakpoint.Cond = cond
}
bpmap.M[addr] = newBreakpoint
return newBreakpoint, nil
}
// SetBreakpointWithID creates a breakpoint at addr, with the specified logical ID.
func (t *Target) SetBreakpointWithID(id int, addr uint64) (*Breakpoint, error) {
bpmap := t.Breakpoints()
bp, err := t.SetBreakpoint(addr, UserBreakpoint, nil)
if err == nil {
bp.LogicalID = id
bpmap.breakpointIDCounter--
}
return bp, err
}
// ClearBreakpoint clears the breakpoint at addr.
func (t *Target) ClearBreakpoint(addr uint64) (*Breakpoint, error) {
if valid, err := t.Valid(); !valid {
return nil, err
}
bpmap := t.Breakpoints()
bp, ok := bpmap.M[addr]
if !ok {
return nil, NoBreakpointError{Addr: addr}
}
bp.Kind &= ^UserBreakpoint
bp.Cond = nil
if bp.Kind != 0 {
return bp, nil
}
if err := t.proc.EraseBreakpoint(bp); err != nil {
return nil, err
}
delete(bpmap.M, addr)
return bp, nil
}
// ClearInternalBreakpoints removes all internal breakpoints from the map,
// calling clearBreakpoint on each one.
func (t *Target) ClearInternalBreakpoints() error {
bpmap := t.Breakpoints()
threads := t.ThreadList()
for addr, bp := range bpmap.M {
bp.Kind = bp.Kind & UserBreakpoint
bp.internalCond = nil
bp.returnInfo = nil
if bp.Kind != 0 {
continue
}
if err := t.proc.EraseBreakpoint(bp); err != nil {
return err
}
for _, thread := range threads {
if thread.Breakpoint().Breakpoint == bp {
thread.Breakpoint().Clear()
}
}
delete(bpmap.M, addr)
}
return nil
}
// HasInternalBreakpoints returns true if bpmap has at least one internal
// breakpoint set.
func (bpmap *BreakpointMap) HasInternalBreakpoints() bool {
for _, bp := range bpmap.M {
if bp.IsInternal() {
return true
}
}
return false
}
// HasHWBreakpoints returns true if there are hardware breakpoints.
func (bpmap *BreakpointMap) HasHWBreakpoints() bool {
for _, bp := range bpmap.M {
if bp.WatchType != 0 {
return true
}
}
return false
}
// BreakpointState describes the state of a breakpoint in a thread.
type BreakpointState struct {
*Breakpoint
// Active is true if the breakpoint condition was met.
Active bool
// Internal is true if the breakpoint was matched as an internal
// breakpoint.
Internal bool
// CondError contains any error encountered while evaluating the
// breakpoint's condition.
CondError error
}
// Clear zeros the struct.
func (bpstate *BreakpointState) Clear() {
bpstate.Breakpoint = nil
bpstate.Active = false
bpstate.Internal = false
bpstate.CondError = nil
}
func (bpstate *BreakpointState) String() string {
s := bpstate.Breakpoint.String()
if bpstate.Active {
s += " active"
}
if bpstate.Internal {
s += " internal"
}
return s
}
func configureReturnBreakpoint(bi *BinaryInfo, bp *Breakpoint, topframe *Stackframe, retFrameCond ast.Expr) {
if topframe.Current.Fn == nil {
return
}
bp.returnInfo = &returnBreakpointInfo{
retFrameCond: retFrameCond,
fn: topframe.Current.Fn,
frameOffset: topframe.FrameOffset(),
spOffset: topframe.FrameOffset() - int64(bi.Arch.PtrSize()), // must be the value that SP had at the entry point of the function
}
}
func (rbpi *returnBreakpointInfo) Collect(thread Thread) []*Variable {
if rbpi == nil {
return nil
}
g, err := GetG(thread)
if err != nil {
proc/*: remove proc.Thread.Blocked, refactor memory access (#2206) On linux we can not read memory if the thread we use to do it is occupied doing certain system calls. The exact conditions when this happens have never been clear. This problem was worked around by using the Blocked method which recognized the most common circumstances where this would happen. However this is a hack: Blocked returning true doesn't mean that the problem will manifest and Blocked returning false doesn't necessarily mean the problem will not manifest. A side effect of this is issue #2151 where sometimes we can't read the memory of a thread and find its associated goroutine. This commit fixes this problem by always reading memory using a thread we know to be good for this, specifically the one returned by ContinueOnce. In particular the changes are as follows: 1. Remove (ProcessInternal).CurrentThread and (ProcessInternal).SetCurrentThread, the "current thread" becomes a field of Target, CurrentThread becomes a (*Target) method and (*Target).SwitchThread basically just sets a field Target. 2. The backends keep track of their own internal idea of what the current thread is, to use it to read memory, this is the thread they return from ContinueOnce as trapthread 3. The current thread in the backend and the current thread in Target only ever get synchronized in two places: when the backend creates a Target object the currentThread field of Target is initialized with the backend's current thread and when (*Target).Restart gets called (when a recording is rewound the currentThread used by Target might not exist anymore). 4. We remove the MemoryReadWriter interface embedded in Thread and instead add a Memory method to Process that returns a MemoryReadWriter. The backends will return something here that will read memory using the current thread saved by the backend. 5. The Thread.Blocked method is removed One possible problem with this change is processes that have threads with different memory maps. As far as I can determine this could happen on old versions of linux but this option was removed in linux 2.5. Fixes #2151
2020-11-09 19:28:40 +00:00
return returnInfoError("could not get g", err, thread.ProcessMemory())
}
scope, err := GoroutineScope(thread)
if err != nil {
proc/*: remove proc.Thread.Blocked, refactor memory access (#2206) On linux we can not read memory if the thread we use to do it is occupied doing certain system calls. The exact conditions when this happens have never been clear. This problem was worked around by using the Blocked method which recognized the most common circumstances where this would happen. However this is a hack: Blocked returning true doesn't mean that the problem will manifest and Blocked returning false doesn't necessarily mean the problem will not manifest. A side effect of this is issue #2151 where sometimes we can't read the memory of a thread and find its associated goroutine. This commit fixes this problem by always reading memory using a thread we know to be good for this, specifically the one returned by ContinueOnce. In particular the changes are as follows: 1. Remove (ProcessInternal).CurrentThread and (ProcessInternal).SetCurrentThread, the "current thread" becomes a field of Target, CurrentThread becomes a (*Target) method and (*Target).SwitchThread basically just sets a field Target. 2. The backends keep track of their own internal idea of what the current thread is, to use it to read memory, this is the thread they return from ContinueOnce as trapthread 3. The current thread in the backend and the current thread in Target only ever get synchronized in two places: when the backend creates a Target object the currentThread field of Target is initialized with the backend's current thread and when (*Target).Restart gets called (when a recording is rewound the currentThread used by Target might not exist anymore). 4. We remove the MemoryReadWriter interface embedded in Thread and instead add a Memory method to Process that returns a MemoryReadWriter. The backends will return something here that will read memory using the current thread saved by the backend. 5. The Thread.Blocked method is removed One possible problem with this change is processes that have threads with different memory maps. As far as I can determine this could happen on old versions of linux but this option was removed in linux 2.5. Fixes #2151
2020-11-09 19:28:40 +00:00
return returnInfoError("could not get scope", err, thread.ProcessMemory())
}
v, err := scope.evalAST(rbpi.retFrameCond)
if err != nil || v.Unreadable != nil || v.Kind != reflect.Bool {
// This condition was evaluated as part of the breakpoint condition
// evaluation, if the errors happen they will be reported as part of the
// condition errors.
return nil
}
if !constant.BoolVal(v.Value) {
// Breakpoint not hit as a return breakpoint.
return nil
}
oldFrameOffset := rbpi.frameOffset + int64(g.stack.hi)
oldSP := uint64(rbpi.spOffset + int64(g.stack.hi))
err = fakeFunctionEntryScope(scope, rbpi.fn, oldFrameOffset, oldSP)
if err != nil {
proc/*: remove proc.Thread.Blocked, refactor memory access (#2206) On linux we can not read memory if the thread we use to do it is occupied doing certain system calls. The exact conditions when this happens have never been clear. This problem was worked around by using the Blocked method which recognized the most common circumstances where this would happen. However this is a hack: Blocked returning true doesn't mean that the problem will manifest and Blocked returning false doesn't necessarily mean the problem will not manifest. A side effect of this is issue #2151 where sometimes we can't read the memory of a thread and find its associated goroutine. This commit fixes this problem by always reading memory using a thread we know to be good for this, specifically the one returned by ContinueOnce. In particular the changes are as follows: 1. Remove (ProcessInternal).CurrentThread and (ProcessInternal).SetCurrentThread, the "current thread" becomes a field of Target, CurrentThread becomes a (*Target) method and (*Target).SwitchThread basically just sets a field Target. 2. The backends keep track of their own internal idea of what the current thread is, to use it to read memory, this is the thread they return from ContinueOnce as trapthread 3. The current thread in the backend and the current thread in Target only ever get synchronized in two places: when the backend creates a Target object the currentThread field of Target is initialized with the backend's current thread and when (*Target).Restart gets called (when a recording is rewound the currentThread used by Target might not exist anymore). 4. We remove the MemoryReadWriter interface embedded in Thread and instead add a Memory method to Process that returns a MemoryReadWriter. The backends will return something here that will read memory using the current thread saved by the backend. 5. The Thread.Blocked method is removed One possible problem with this change is processes that have threads with different memory maps. As far as I can determine this could happen on old versions of linux but this option was removed in linux 2.5. Fixes #2151
2020-11-09 19:28:40 +00:00
return returnInfoError("could not read function entry", err, thread.ProcessMemory())
}
vars, err := scope.Locals()
if err != nil {
proc/*: remove proc.Thread.Blocked, refactor memory access (#2206) On linux we can not read memory if the thread we use to do it is occupied doing certain system calls. The exact conditions when this happens have never been clear. This problem was worked around by using the Blocked method which recognized the most common circumstances where this would happen. However this is a hack: Blocked returning true doesn't mean that the problem will manifest and Blocked returning false doesn't necessarily mean the problem will not manifest. A side effect of this is issue #2151 where sometimes we can't read the memory of a thread and find its associated goroutine. This commit fixes this problem by always reading memory using a thread we know to be good for this, specifically the one returned by ContinueOnce. In particular the changes are as follows: 1. Remove (ProcessInternal).CurrentThread and (ProcessInternal).SetCurrentThread, the "current thread" becomes a field of Target, CurrentThread becomes a (*Target) method and (*Target).SwitchThread basically just sets a field Target. 2. The backends keep track of their own internal idea of what the current thread is, to use it to read memory, this is the thread they return from ContinueOnce as trapthread 3. The current thread in the backend and the current thread in Target only ever get synchronized in two places: when the backend creates a Target object the currentThread field of Target is initialized with the backend's current thread and when (*Target).Restart gets called (when a recording is rewound the currentThread used by Target might not exist anymore). 4. We remove the MemoryReadWriter interface embedded in Thread and instead add a Memory method to Process that returns a MemoryReadWriter. The backends will return something here that will read memory using the current thread saved by the backend. 5. The Thread.Blocked method is removed One possible problem with this change is processes that have threads with different memory maps. As far as I can determine this could happen on old versions of linux but this option was removed in linux 2.5. Fixes #2151
2020-11-09 19:28:40 +00:00
return returnInfoError("could not evaluate return variables", err, thread.ProcessMemory())
}
vars = filterVariables(vars, func(v *Variable) bool {
return (v.Flags & VariableReturnArgument) != 0
})
return vars
}
func returnInfoError(descr string, err error, mem MemoryReadWriter) []*Variable {
v := newConstant(constant.MakeString(fmt.Sprintf("%s: %v", descr, err.Error())), mem)
v.Name = "return value read error"
return []*Variable{v}
}