Updated the readme to reflect the current support.
/*
* Copyright (c) 2004 Nokia Corporation and/or its subsidiary(-ies).
* All rights reserved.
* This component and the accompanying materials are made available
* under the terms of "Eclipse Public License v1.0"
* which accompanies this distribution, and is available
* at the URL "http://www.eclipse.org/legal/epl-v10.html".
*
* Initial Contributors:
* Nokia Corporation - initial contribution.
*
* Contributors:
*
* Description:
*
*/
#ifdef __WINS__
#error - this driver cannot be built for emulation
#endif
#include <e32def.h>
#include <e32cmn.h>
#include <u32std.h>
#include <kernel.h>
#include <kern_priv.h>
#include <nk_trace.h>
//#include <mmboot.h>
#include <arm.h>
#include <cache.h>
#include <platform.h>
#include <nkern.h>
#include <u32hal.h>
#include "TrkKernelDriver.h"
#include "TrkDriver.h"
#include "TrkEventHandler.h"
#define KTrkAppSecurUid 0x200170BB
#define KTrkExeSecurUid 0x200159E2
#define KTrkSrvSecurUid 0x200170B7
// Uncomment the line below for reading kernel thread registers.
// There is a problem using NKern::Lock before calling UserContextType on old releases like S60 3.0
// So for now, disabling the supporting reading kern thread registers as we are not really supporting
// device driver debugging anyway.
//#define SUPPORT_KERNCONTEXT
//
// Static function definitions
//
static TInt Bitcount(TUint32 val)
{
TInt nbits;
for (nbits = 0; val != 0; nbits++)
{
val &= val - 1; // delete rightmost 1-bit in val
}
return nbits;
}
static TUint8 tolower(TUint8 c)
{
if (c >= 'A' && c <= 'Z')
c = c + ('a' - 'A');
return c;
}
static TInt _strnicmp(const TUint8 *s1, const TUint8 *s2, int n)
{
int i;
TUint8 c1, c2;
for (i=0; i<n; i++)
{
c1 = tolower(*s1++);
c2 = tolower(*s2++);
if (c1 < c2) return -1;
if (c1 > c2) return 1;
if (!c1) return 0;
}
return 0;
}
//
//
// DMetroTrkDriverFactory implementation
//
//
//
// DMetroTrkDriverFactory constructor
//
DMetroTrkDriverFactory::DMetroTrkDriverFactory()
{
iVersion = TVersion(KMajorVersionNumber,KMinorVersionNumber,KBuildVersionNumber);
}
//
// DMetroTrkDriverFactory::Create
//
TInt DMetroTrkDriverFactory::Create(DLogicalChannelBase*& aChannel)
{
if (iOpenChannels != 0)
return KErrInUse; // a channel is already open
aChannel = new DMetroTrkChannel(this);
return aChannel ? KErrNone : KErrNoMemory;
}
//
// DMetroTrkDriverFactory::Install
//
TInt DMetroTrkDriverFactory::Install()
{
return(SetName(&KMetroTrkDriverName));
}
//
// DMetroTrkDriverFactory::Install
//
void DMetroTrkDriverFactory::GetCaps(TDes8& aDes) const
{
TCapsMetroTrkDriver b;
b.iVersion = TVersion(KMajorVersionNumber, KMinorVersionNumber, KBuildVersionNumber);
Kern::InfoCopy(aDes,(TUint8*)&b, sizeof(b));
}
//
//
// DMetroTrkChannel implementation
//
//
//
// DMetroTrkChannel constructor
//
DMetroTrkChannel::DMetroTrkChannel(DLogicalDevice* aLogicalDevice)
: iExcludedROMAddressStart(ROM_LINEAR_BASE),
iExcludedROMAddressEnd(0),
iBreakPointList(NUMBER_OF_TEMP_BREAKPOINTS, 0),
iNextBreakId(NUMBER_OF_TEMP_BREAKPOINTS),
iEventInfo(NULL),
iEventQueue(NUMBER_OF_EVENTS_TO_QUEUE, 0),
iTraceEventQueue(NUMBER_OF_EVENTS_TO_QUEUE, 0),
iRequestGetEventStatus(NULL),
iPageSize(0x1000),
iNotifyLibLoadedEvent(ETrue),
iMultipleMemModel(EFalse),
iExcInfoValid(EFalse),
iDebugging(ETrue)
{
LOG_MSG("DMetroTrkChannel::DMetroTrkChannel()");
iDevice = aLogicalDevice;
iClientThread = &Kern::CurrentThread();
TInt err = iClientThread->Open();
iBreakPointList.Reset();
TBreakEntry emptyTempBreak;
for (TInt i = 0; i < NUMBER_OF_TEMP_BREAKPOINTS; i++)
{
emptyTempBreak.iId = i;
if (KErrNone != iBreakPointList.Append(emptyTempBreak))
{
LOG_MSG("Error appending blank temp break entry");
}
}
SEventInfo emptyEvent;
for (TInt i=0; i<NUMBER_OF_EVENTS_TO_QUEUE; i++)
{
if (KErrNone != iEventQueue.Append(emptyEvent))
{
LOG_MSG("Error appending blank event entry");
}
}
for (TInt i=0; i<NUMBER_OF_EVENTS_TO_QUEUE; i++)
{
if (KErrNone != iTraceEventQueue.Append(emptyEvent))
{
LOG_MSG("Error appending blank trace event entry");
}
}
TTrkLibName emptyLib;
for (TInt i=0; i<NUMBER_OF_LIBS_TO_REGISTER; i++)
{
if (KErrNone != iLibraryNotifyList.Append(emptyLib))
{
LOG_MSG("Error appending blank empty lib entry");
}
}
iPageSize = Kern::RoundToPageSize(1);
}
//
// DMetroTrkChannel destructor
//
DMetroTrkChannel::~DMetroTrkChannel()
{
LOG_MSG("DMetroTrkChannel::~DMetroTrkChannel()");
iDebugging = EFalse;
Kern::SafeClose((DObject*&)iClientThread, NULL);
ClearAllBreakPoints();
// close the breakpoint list and free the memory associated with it
iBreakPointList.Close();
// close the event queue and free the memory associated with it
iEventQueue.Close();
// close the trace event queue and free the memory associated with it
iTraceEventQueue.Close();
//close the debug process list
iDebugProcessList.Close();
//close the process notify list
iProcessNotifyList.Close();
//close the code modifier
DebugSupport::CloseCodeModifier();
// PANIC_BACKPORT
// Resume all the frozen threads with semaphores.
for(TInt i=0; i<iFrozenThreadSemaphores.Count(); i++)
{
NKern::FSSignal(iFrozenThreadSemaphores[i]);
NKern::ThreadEnterCS();
delete iFrozenThreadSemaphores[i];
NKern::ThreadLeaveCS();
iFrozenThreadSemaphores.Remove(i);
}
//Reset the array and delete the objects that its members point to
iFrozenThreadSemaphores.ResetAndDestroy();
// END PANIC_BACKPORT
}
//
// DMetroTrkChannel::DoCreate
//
TInt DMetroTrkChannel::DoCreate(TInt /*aUnit*/, const TDesC* anInfo, const TVersion& aVer)
{
LOG_MSG("DMetroTrkChannel::DoCreate()");
if (!Kern::QueryVersionSupported(TVersion(KMajorVersionNumber, KMinorVersionNumber, KBuildVersionNumber), aVer))
return KErrNotSupported;
// Do the security check here so that any arbitrary application doesn't make
// use of Trk kernel driver.
if (!DoSecurityCheck())
{
return KErrPermissionDenied;
}
if (anInfo)
{
// this is the end address of the user library.
// this doesn't seem to be valid for EKA2.
// right now we dont need this for EKA2 since we are not worried
// about kernel being stopped as kernel is multithreaded.
// just retaining this for future use.
TBuf8<32> buf;
TInt err = Kern::ThreadRawRead(iClientThread, anInfo, &buf, 32);
if(err != KErrNone)
return err;
//iExcludedROMAddressEnd = *(TUint32 *)(&(buf.Ptr()[20]));
}
//check whether the memory model is multiple or not.
TUint32 memModelAttrib = (TUint32)Kern::HalFunction(EHalGroupKernel, EKernelHalMemModelInfo, NULL, NULL);
if ((memModelAttrib & EMemModelTypeMask) == EMemModelTypeMultiple)
{
//Multiple memory model
iMultipleMemModel = ETrue;
}
TUint caps; //ignored for now
TInt err = DebugSupport::InitialiseCodeModifier(caps, NUMBER_OF_MAX_BREAKPOINTS);
//if code modifier initializer failed,
//return here, since we can't set an breakpoints
if(err != KErrNone)
return err;
//Setup the driver for receiving client messages
iDFCQue = NULL;
TBuf8<KMaxInfoName> trkDFC = _L8("TRK DFC");
err = Kern::DfcQCreate(iDFCQue, 27, &trkDFC);
if (err == KErrNone)
{
SetDfcQ(iDFCQue);
}
else
{
SetDfcQ(Kern::DfcQue0());
}
iMsgQ.Receive();
iEventHandler = new DMetroTrkEventHandler;
if (!iEventHandler)
return KErrNoMemory;
err = iEventHandler->Create(iDevice, this, iClientThread);
if (err != KErrNone)
return err;
return iEventHandler->Start();
}
//
// DMetroTrkChannel::DoCancel
//
void DMetroTrkChannel::DoCancel(TInt aReqNo)
{
LOG_MSG("DMetroTrkChannel::DoCancel()");
switch(aReqNo)
{
case RMetroTrkDriver::ERequestGetEventCancel:
{
Kern::RequestComplete(iClientThread, iRequestGetEventStatus, KErrCancel);
iEventInfo = NULL;
iRequestGetEventStatus = 0;
}
break;
}
}
//
// DMetroTrkChannel::DoRequest
//
void DMetroTrkChannel::DoRequest(TInt aReqNo, TRequestStatus* aStatus, TAny* a1, TAny* a2)
{
LOG_MSG("DMetroTrkChannel::DoRequest()");
switch(aReqNo)
{
case RMetroTrkDriver::ERequestGetEvent:
{
// check to see if we have any queued up events
for (TInt i=0; i<NUMBER_OF_EVENTS_TO_QUEUE; i++)
{
if (SEventInfo::EUnknown != iEventQueue[i].iEventType)
{
LOG_MSG("DoRequest - slot NOT empty");
// iClientThread is the user side debugger thread, so use it to write the info to it memory
TInt err = Kern::ThreadRawWrite(iClientThread, a1, (TUint8 *)&iEventQueue[i], sizeof(SEventInfo), iClientThread);
if (KErrNone != err)
LOG_MSG2("Error writing event info: %d", err);
// signal the debugger thread
Kern::RequestComplete(iClientThread, aStatus, KErrNone);
iEventQueue[i].Reset();
return;
}
LOG_MSG("DoRequest - slot empty");
}
// check to see if we have any queued up trace events
for (TInt i=0; i<NUMBER_OF_EVENTS_TO_QUEUE; i++)
{
if (SEventInfo::EUnknown != iTraceEventQueue[i].iEventType)
{
LOG_MSG("DoRequest - slot NOT empty");
// iClientThread is the user side debugger thread, so use it to write the info to it memory
TInt err = Kern::ThreadRawWrite(iClientThread, a1, (TUint8 *)&iTraceEventQueue[i], sizeof(SEventInfo), iClientThread);
if (KErrNone != err)
LOG_MSG2("Error writing trace event info: %d", err);
// signal the debugger thread
Kern::RequestComplete(iClientThread, aStatus, KErrNone);
iTraceEventQueue[i].Reset();
return;
}
LOG_MSG("DoRequest - trace slot empty");
}
// store the pointer so we can modify it later
iEventInfo = (SEventInfo *)a1;
iRequestGetEventStatus = aStatus;
break;
}
default:
Kern::RequestComplete(iClientThread, aStatus, KErrNotSupported);
}
}
//
// DMetroTrkChannel::DoControl
//
TInt DMetroTrkChannel::DoControl(TInt aFunction, TAny* a1, TAny* a2)
{
LOG_MSG("DMetroTrkChannel::DoControl()");
LOG_MSG2("DoControl Function %d", aFunction);
TInt err = KErrNone;
switch(aFunction)
{
case RMetroTrkDriver::EControlSetBreak:
{
err = SetBreak((TUint32)a1, (TMetroTrkBreakInfo*)a2);
break;
}
case RMetroTrkDriver::EControlClearBreak:
{
err = DoClearBreak((TInt32)a1);
break;
}
case RMetroTrkDriver::EControlChangeBreakThread:
{
err = DoChangeBreakThread((TUint32)a1, (TInt32)a2);
break;
}
case RMetroTrkDriver::EControlSuspendThread:
{
err = DoSuspendThread(ThreadFromId((TUint32)a1));
break;
}
case RMetroTrkDriver::EControlResumeThread:
{
err = DoResumeThread(ThreadFromId((TUint32)a1));
break;
}
case RMetroTrkDriver::EControlStepRange:
{
err = StepRange(ThreadFromId((TUint32)a1), (TMetroTrkStepInfo*)a2);
break;
}
case RMetroTrkDriver::EControlReadMemory:
{
err = ReadMemory(ThreadFromId((TUint32)a1), (TMetroTrkMemoryInfo*)a2);
break;
}
case RMetroTrkDriver::EControlWriteMemory:
{
err = WriteMemory(ThreadFromId((TUint32)a1), (TMetroTrkMemoryInfo*)a2);
break;
}
case RMetroTrkDriver::EControlReadRegisters:
{
err = ReadRegisters(ThreadFromId((TUint32)a1), (TMetroTrkRegisterInfo*)a2);
break;
}
case RMetroTrkDriver::EControlWriteRegisters:
{
err = WriteRegisters(ThreadFromId((TUint32)a1), (TMetroTrkRegisterInfo*)a2);
break;
}
case RMetroTrkDriver::EControlGetProcessInfo:
{
err = GetProcessInfo((TInt)a1, (TMetroTrkTaskInfo*)a2);
break;
}
case RMetroTrkDriver::EControlGetThreadInfo:
{
err = GetThreadInfo((TInt)a1, (TMetroTrkTaskInfo*)a2);
break;
}
case RMetroTrkDriver::EControlGetProcessAddresses:
{
err = GetProcessAddresses(ThreadFromId((TUint32)a1), (TMetroTrkProcessInfo*)a2);
break;
}
case RMetroTrkDriver::EControlGetStaticLibraryInfo:
{
err = GetStaticLibraryInfo((TInt)a1, (SEventInfo*)a2);
break;
}
case RMetroTrkDriver::EControlEnableLibLoadedEvent:
{
iNotifyLibLoadedEvent = ETrue;
break;
}
case RMetroTrkDriver::EControlDisableLibLoadedEvent:
{
iNotifyLibLoadedEvent = EFalse;
break;
}
case RMetroTrkDriver::EControlGetLibraryInfo:
{
err = GetLibraryInfo((TMetroTrkLibInfo*)a1);
break;
}
case RMetroTrkDriver::EControlGetExeInfo:
{
err = GetExeInfo((TMetroTrkExeInfo*)a1);
break;
}
case RMetroTrkDriver::EControlGetProcUidInfo:
{
err = GetProcUidInfo((TMetroTrkProcUidInfo*)a1);
break;
}
case RMetroTrkDriver::EControlDetachProcess:
{
err = DetachProcess(ProcessFromId((TInt32)a1));
break;
}
default:
{
return KErrGeneral;
}
}
if (KErrNone != err)
{
LOG_MSG2("Error %d from control function", err);
}
return err;
}
void DMetroTrkChannel::HandleMsg(TMessageBase* aMsg)
{
LOG_MSG("DMetroTrkChannel::HandleMsg()");
TThreadMessage& m = *(TThreadMessage*)aMsg;
TInt id = m.iValue;
if (id == (TInt)ECloseMsg)
{
if (iEventHandler)
{
iDebugging = EFalse;
iEventHandler->Stop();
iEventHandler->Close();
iEventHandler = NULL;
}
m.Complete(KErrNone, EFalse);
return;
}
if (id == KMaxTInt)
{
// DoCancel
DoCancel(m.Int0());
m.Complete(KErrNone, ETrue);
return;
}
if (id < 0)
{
// DoRequest
TRequestStatus* pStatus = (TRequestStatus*)m.Ptr0();
DoRequest(~id, pStatus, m.Ptr1(), m.Ptr2());
m.Complete(KErrNone, ETrue);
}
else
{
// DoControl
TInt err = DoControl(id, m.Ptr0(), m.Ptr1());
m.Complete(err, ETrue);
}
}
//
// DMetroTrkChannel::AddProcess
//
void DMetroTrkChannel::AddProcess(DProcess *aProcess, DThread *aThread)
{
LOG_MSG("DMetroTrkChannel::AddProcess()");
// check to see if we are still debugging, otherwise just return
if (!iDebugging)
return;
if (aProcess)
{
if (!aThread)
LOG_MSG("Creator thread not available");
// check to see if this process is being started by debug agent.
// If this is the case, we don't need to notify the agent since the debug agent already knows about it.
if (aThread && aThread->iOwningProcess->iId == iClientThread->iOwningProcess->iId)
return;
SEventInfo processEventInfo;
processEventInfo.iEventType = processEventInfo.EProcessAdded;
processEventInfo.iProcessId = aProcess->iId;
processEventInfo.iFileName.Copy(*aProcess->iName);
processEventInfo.iUid = aProcess->iUids.iUid[2].iUid;
// Kernel hasn't created teh code segment yet for this process,
// so queue this event separately and wait for start thread event to notify.
iProcessNotifyList.Append(processEventInfo);
}
}
//
// DMetroTrkChannel::StartThread
//
void DMetroTrkChannel::StartThread(DThread *aThread)
{
LOG_MSG("DMetroTrkChannel::StartThread()");
// check to see if we are still debugging, otherwise just return
if (!iDebugging)
return;
// Using the lib loaded event flag for notifying processes as well.
// Check if lib loaded event is disabled. Lib loaded event
// is disabled temporarily in some situations, when the engine
// is performing some complex operations like installing sis
// files, in which case, suspending the thread for lib loaded
// events would end up in a dead lock. This is due to the fact
// that the active object that handles the event notification
// wouldn't get a chance to run since the code in TRK engine
// which handles the commands from the host debugger is also
// done in an active object which runs in the same thread.
if (!iNotifyLibLoadedEvent)
return;
if (aThread && aThread->iOwningProcess)
{
DCodeSeg* codeSeg = aThread->iOwningProcess->iCodeSeg;
if (codeSeg)
{
TModuleMemoryInfo processMemoryInfo;
TInt err = codeSeg->GetMemoryInfo(processMemoryInfo, aThread->iOwningProcess);
if (err == KErrNone)
{
for (TInt i = 0; i < iProcessNotifyList.Count(); i++)
{
if (iProcessNotifyList[i].iProcessId == aThread->iOwningProcess->iId)
{
// Suspend the thread so that the host debugger can set breakpoints.
Kern::ThreadSuspend(*aThread, 1);
iProcessNotifyList[i].iThreadId = aThread->iId;
iProcessNotifyList[i].iCodeAddress = processMemoryInfo.iCodeBase;
iProcessNotifyList[i].iDataAddress = processMemoryInfo.iInitialisedDataBase;
// Notify the process added event now that we have the code segment for the process
NotifyEvent(iProcessNotifyList[i]);
// Now remove from the list
iProcessNotifyList.Remove(i);
break;
}
}
}
else
{
LOG_MSG2("Error in getting memory info: %d", err);
}
}
else
{
LOG_MSG2("Invalid code segment found for the started thread: %d", aThread->iId);
}
}
}
//
// DMetroTrkChannel::RemoveProcess
//
void DMetroTrkChannel::RemoveProcess(DProcess *aProcess)
{
LOG_MSG("DMetroTrkChannel::RemoveProcess()");
// check to see if we are still debugging, otherwise just return
if (!iDebugging)
return;
// this is called when a process dies. we want to mark any breakpoints in this
// process space as obsolete. the main reason for this is so we don't return
// an error when the host debugger tries to clear breakpoints for the process
TUint32 codeAddress = 0;
TUint32 codeSize = 0;
LOG_MSG2("Process being removed, Name %S", aProcess->iName);
DCodeSeg* codeSeg = aProcess->iCodeSeg;
if (codeSeg)
{
TModuleMemoryInfo processMemoryInfo;
TInt err = codeSeg->GetMemoryInfo(processMemoryInfo, aProcess);
if (err != KErrNone)
{
codeAddress = processMemoryInfo.iCodeBase;
codeSize = processMemoryInfo.iCodeSize;
}
else
{
LOG_MSG2("Error in getting memory info: %d", err);
}
}
if (!codeAddress || !codeSize)
{
LOG_MSG2("Code segment not available for process %d", aProcess->iId);
// make sure there is not already a breakpoint at this address
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == aProcess->iId)
{
codeAddress = iDebugProcessList[i].iCodeAddress;
codeSize = iDebugProcessList[i].iCodeSize;
//now remove from the list
iDebugProcessList.Remove(i);
break;
}
}
}
if (!codeAddress || !codeSize)
return;
// first invalidate all breakpoints that were set in the library code
for (TInt i=0; i<iBreakPointList.Count(); i++)
{
if ((iBreakPointList[i].iAddress >= codeAddress) && (iBreakPointList[i].iAddress < (codeAddress + codeSize)))
{
LOG_MSG2("Disabling process breakpoint at address %x", iBreakPointList[i].iAddress);
iBreakPointList[i].iObsoleteLibraryBreakpoint = ETrue;
}
}
}
//
// DMetroTrkChannel::AddLibrary
//
void DMetroTrkChannel::AddLibrary(DLibrary *aLibrary, DThread *aThread)
{
LOG_MSG("DMetroTrkChannel::AddLibrary()");
LOG_MSG2(("Lib loaded: %S"), aLibrary->iName);
// check to see if we are still debugging, otherwise just return
if (!iDebugging)
return;
// Check if lib loaded event is disabled. Lib loaded event
// is disabled temporarily in some situations, when the engine
// is performing some complex operations like installing sis
// files, in which case, suspending the thread for lib loaded
// events would end up in a dead lock. This is due to the fact
// that the active object that handles the event notification
// wouldn't get a chance to run since the code in TRK engine
// which handles the commands from the host debugger is also
// done in an active object which runs in the same thread.
if (!iNotifyLibLoadedEvent)
return;
TBool isDebugging = EFalse;
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == aThread->iOwningProcess->iId)
{
isDebugging = ETrue;
break;
}
}
if (isDebugging == EFalse)
return;
if (aThread)
{
// make sure this is not the debugger thread
if ((aThread != iClientThread) && (aThread->iOwningProcess->iId != iClientThread->iOwningProcess->iId))
{
SEventInfo info;
Kern::ThreadSuspend(*aThread, 1);
info.iEventType = SEventInfo::ELibraryLoaded;
info.iProcessId = aThread->iOwningProcess->iId;
info.iThreadId = aThread->iId;
//get the code address
DCodeSeg* codeSeg = aLibrary->iCodeSeg;
if (!codeSeg)
{
LOG_MSG2("Code segment not available for library %S", aLibrary->iName);
return;
}
TModuleMemoryInfo memoryInfo;
TInt err = codeSeg->GetMemoryInfo(memoryInfo, NULL); //NULL for DProcess should be ok;
if (err != KErrNone)
{
LOG_MSG2("Error in getting memory info: %d", err);
return;
}
info.iCodeAddress = memoryInfo.iCodeBase;
info.iDataAddress = memoryInfo.iInitialisedDataBase;
info.iFileName.Copy(*(aLibrary->iName)); //just the name, without uid info.
// now remove this library if its in our notify list
for (TInt i =0; i<iLibraryNotifyList.Count(); i++)
{
if (!iLibraryNotifyList[i].iEmptySlot &&
!_strnicmp(iLibraryNotifyList[i].iName.Ptr(), info.iFileName.Ptr(), info.iFileName.Length()))
{
iLibraryNotifyList[i].iEmptySlot = ETrue;
break;
}
}
// now check to see if any libs are loaded because of this library load event.
CheckLibraryNotifyList(info.iProcessId);
//queue up or complete the event
NotifyEvent(info);
}
}
}
//
// DMetroTrkChannel::RemoveLibrary
//
void DMetroTrkChannel::RemoveLibrary(DLibrary *aLibrary)
{
LOG_MSG("DMetroTrkChannel::RemoveLibrary()");
LOG_MSG2(("Lib unloaded: %S"), aLibrary->iName);
// check to see if we are still debugging, otherwise just return
if (!iDebugging)
return;
// this is called when all handles to this library have been closed. this can happen when a process dies, or when a dll is
// unloaded while the process lives on. in former case, we don't need to notify the host debugger because that process is
// dying anyway. for the latter case, we do need to notify the host so it can unload the symbolics, etc.
DThread* aThread = &Kern::CurrentThread();
if ((aThread) &&
(aThread != iClientThread) &&
(aThread->iOwningProcess->iId != iClientThread->iOwningProcess->iId))
{
//the library gets unloaded only when the mapcount is 0.
if (aLibrary->iMapCount != 0)
return;
DCodeSeg* codeSeg = aLibrary->iCodeSeg;
if (!codeSeg)
{
LOG_MSG2("Code segment not available for library %S", aLibrary->iName);
return;
}
TModuleMemoryInfo processMemoryInfo;
TInt err = codeSeg->GetMemoryInfo(processMemoryInfo, NULL); //passing NULL for the DProcess argument should be ok;
if (err != KErrNone)
{
LOG_MSG2("Error in getting memory info: %d", err);
return;
}
TUint32 codeAddress = processMemoryInfo.iCodeBase;
TUint32 codeSize = processMemoryInfo.iCodeSize;
// first invalidate all breakpoints that were set in the library code
for (TInt i=0; i<iBreakPointList.Count(); i++)
{
if ((iBreakPointList[i].iAddress >= codeAddress) && (iBreakPointList[i].iAddress < (codeAddress + codeSize)))
{
LOG_MSG2("Disabling library breakpoint at address %x", iBreakPointList[i].iAddress);
iBreakPointList[i].iObsoleteLibraryBreakpoint = ETrue;
}
}
DProcess *process = &Kern::CurrentProcess();
if (process)
{
RArray<SCodeSegEntry>* dynamicCode = &(process->iDynamicCode);
if (dynamicCode)
{
for (TInt j=0; j<dynamicCode->Count(); j++)
{
if ((*dynamicCode)[j].iLib == aLibrary)
{
SEventInfo info;
info.iEventType = SEventInfo::ELibraryUnloaded;
info.iFileName.Copy(*(aLibrary->iName)); //lib name without uid info
//info.iFileName.ZeroTerminate();
info.iProcessId = process->iId;
info.iThreadId = 0xFFFFFFFF; // don't care!
//queue up or complete the event
NotifyEvent(info);
}
}
}
}
}
}
//
// DMetroTrkChannel::AddCodeSegment
//
void DMetroTrkChannel::AddCodeSegment(DCodeSeg *aCodeSeg, DProcess *aProcess)
{
LOG_MSG("DMetroTrkChannel::AddCodeSegment()");
// Check if lib loaded event is disabled. Lib loaded event
// is disabled temporarily in some situations, when the engine
// is performing some complex operations like installing sis
// files, in which case, suspending the thread for lib loaded
// events would end up in a dead lock. This is due to the fact
// that the active object that handles the event notification
// wouldn't get a chance to run since the code in TRK engine
// which handles the commands from the host debugger is also
// done in an active object which runs in the same thread.
if (!iNotifyLibLoadedEvent)
return;
TBool isDebugging = EFalse;
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == aProcess->iId)
{
isDebugging = ETrue;
break;
}
}
if (isDebugging == EFalse)
return;
if (aCodeSeg)
{
const TUint8* ptr = aCodeSeg->iFileName->Ptr();
if (aCodeSeg->IsDll())
{
LOG_MSG2("DLL code segment is loaded: %s", ptr);
DThread* mainThread = aProcess->FirstThread();
if (mainThread)
{
// make sure this is not the debugger thread
if ((mainThread->iId != iClientThread->iId) && (aProcess->iId != iClientThread->iOwningProcess->iId))
{
SEventInfo info;
//DoSuspendThread(mainThread);
Kern::ThreadSuspend(*mainThread, 1);
info.iEventType = SEventInfo::ELibraryLoaded;
info.iProcessId = aProcess->iId;
info.iThreadId = mainThread->iId;
TModuleMemoryInfo memoryInfo;
TInt err = aCodeSeg->GetMemoryInfo(memoryInfo, NULL); //NULL for DProcess should be ok;
if (err != KErrNone)
{
LOG_MSG2("Error in getting memory info: %d", err);
return;
}
info.iCodeAddress = memoryInfo.iCodeBase;
info.iDataAddress = memoryInfo.iInitialisedDataBase;
info.iFileName.Copy(aCodeSeg->iRootName); //just the name, without uid info.
//queue up or complete the event
NotifyEvent(info);
}
}
else
{
LOG_MSG2("Invalid main thread for this process: %d", aProcess->iId);
}
}
else
{
if (aCodeSeg->IsExe())
LOG_MSG2("EXE code segment is loaded: %s", ptr);
}
}
}
//
// DMetroTrkChannel::RemoveCodeSegment
//
void DMetroTrkChannel::RemoveCodeSegment(DCodeSeg *aCodeSeg, DProcess *aProcess)
{
LOG_MSG("DMetroTrkChannel::RemoveCodeSegment()");
// We don't do anything right now as we are not using the code segment events.
}
//
// DMetroTrkChannel::HandleEventKillThread
//
TBool DMetroTrkChannel::HandleEventKillThread(DThread* aThread)
{
LOG_MSG("DMetroTrkChannel::HandleEventKillThread");
// check to see if we are still debugging, otherwise just return
if (!iDebugging)
return EFalse;
if (!aThread)
{
LOG_MSG("Invalid thread handle");
return EFalse;
}
DThread* currentThread = &Kern::CurrentThread();
if (!currentThread)
{
LOG_MSG("Error getting current thread");
return EFalse;
}
//Kern::ThreadSuspend(*currentThread, 1);
if (aThread->iExitType != EExitPanic)
{
return EFalse;
}
SEventInfo info;
info.iProcessId = aThread->iOwningProcess->iId;
info.iThreadId = aThread->iId;
info.iCurrentPC = ReadRegister(aThread, 14);//PC_REGISTER);
info.iPanicCategory.Copy(aThread->iExitCategory);
info.iExceptionNumber = 100;//aThread->iExitReason;
info.iPanicReason = aThread->iExitReason;
info.iEventType = SEventInfo::EThreadPanic;
// if its not an invalide opcode exception, check to see if we are debugging
// the process.
TBool isDebugging = EFalse;
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == info.iProcessId)
{
isDebugging = ETrue;
break;
}
}
if (!isDebugging)
return EFalse;
#define DOFREEZE 1
// PANIC_BACKPORT
#if DOFREEZE
NKern::ThreadEnterCS();
NFastSemaphore* sem = new NFastSemaphore();
NKern::ThreadLeaveCS();
// sem->iOwningThread = &(Kern::CurrentThread().iNThread);
sem->iOwningThread = &aThread->iNThread;
iFrozenThreadSemaphores.Append(sem);
// First do the notify, then FSWait
NotifyEvent(info);
NKern::FSWait(sem);
return ETrue;
#else
NotifyEvent(info);
LOG_MSG("DMetroTrkChannel::HandleEventKillThread - Panic Do Nothing");
return EFalse;
#endif
// END PANIC_BACKPORT
}
//
// DMetroTrkChannel::HandleSwException
//
TBool DMetroTrkChannel::HandleSwException(TExcType aExcType)
{
LOG_MSG("DMetroTrkChannel::HandleSwException");
SEventInfo info;
DThread* currentThread = &Kern::CurrentThread();
if (!currentThread)
{
LOG_MSG("Error getting current thread");
return EFalse;
}
info.iProcessId = currentThread->iOwningProcess->iId;
info.iThreadId = currentThread->iId;
info.iCurrentPC = ReadRegister(currentThread, PC_REGISTER);
info.iExceptionNumber = aExcType;
info.iEventType = SEventInfo::EThreadException;
if (info.iExceptionNumber != EExcInvalidOpCode)
{
// if its not an invalide opcode exception, check to see if we are debugging
// the process.
TBool isDebugging = EFalse;
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == info.iProcessId)
{
isDebugging = ETrue;
break;
}
}
if (!isDebugging)
return EFalse;
}
NThread nThread = currentThread->iNThread;
if (nThread.iSuspendCount == 0)
Kern::ThreadSuspend(*currentThread, 1);
HandleException(info, currentThread);
return ETrue;
}
//
// DMetroTrkChannel::HandleHwException
//
TBool DMetroTrkChannel::HandleHwException(TArmExcInfo* aExcInfo)
{
LOG_MSG("DMetroTrkChannel::HandleHwException()");
SEventInfo info;
DThread* currentThread = &Kern::CurrentThread();
if (!currentThread)
{
LOG_MSG("Error getting current thread");
return EFalse;
}
//store the exception info as its needed to read the context for system threads
iExcInfoValid = ETrue;
iCurrentExcInfo = *aExcInfo;
info.iProcessId = currentThread->iOwningProcess->iId;
info.iThreadId = currentThread->iId;
info.iCurrentPC = aExcInfo->iR15;
switch (aExcInfo->iExcCode)
{
case 0:
info.iExceptionNumber = EExcCodeAbort;
break;
case 1:
info.iExceptionNumber = EExcDataAbort;
break;
case 2:
info.iExceptionNumber = EExcInvalidOpCode;
break;
default:
return EFalse;
}
info.iEventType = SEventInfo::EThreadException;
if (info.iExceptionNumber != EExcInvalidOpCode)
{
// if its not an invalide opcode exception, check to see if we are debugging
// the process.
TBool isDebugging = EFalse;
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == info.iProcessId)
{
isDebugging = ETrue;
break;
}
}
if (!isDebugging)
return EFalse;
}
NThread nThread = currentThread->iNThread;
if (nThread.iSuspendCount == 0) //don't suspend the thread if its already suspended.
Kern::ThreadSuspend(*currentThread, 1);
HandleException(info, currentThread);
return ETrue;
}
//
// DMetroTrkChannel::HandleException
//
void DMetroTrkChannel::HandleException(SEventInfo& aEventInfo, DThread* aCurrentThread)
{
TInt err = KErrNone;
// see if it was a breakpoint that stopped the thread
if (((TInt)2 == aEventInfo.iExceptionNumber) || ((TInt)EExcInvalidOpCode == aEventInfo.iExceptionNumber))
{
TUint32 inst = KArmBreakPoint;
TInt instSize = 4;
// change these for thumb mode
if (ReadRegister(aCurrentThread, STATUS_REGISTER) & ECpuThumb)
{
inst = KThumbBreakPoint;
instSize = 2;
}
TUint32 instruction = 0;
err = Kern::ThreadRawRead(aCurrentThread, (TUint32 *)aEventInfo.iCurrentPC, (TUint8 *)&instruction, instSize);
if (KErrNone != err)
LOG_MSG2("Error reading instruction at currentpc: %d", err);
if (!memcompare((TUint8 *)&inst, instSize, (TUint8 *)&instruction, instSize))
{
// the exception was a breakpoint instruction. see if we have a breakpoint at that address
for (TInt i=0; i<iBreakPointList.Count(); i++)
{
TBreakEntry breakEntry = iBreakPointList[i];
if (breakEntry.iAddress == aEventInfo.iCurrentPC)
{
LOG_MSG2("Breakpoint with Id %d has been hit", breakEntry.iId);
//change the event type to breakpoint type
aEventInfo.iEventType = SEventInfo::EThreadBreakPoint;
// enable any breakpoints we had to disable for this thread
err = DoEnableDisabledBreak(aEventInfo.iThreadId);
if (KErrNone != err)
LOG_MSG2("Error %d enabling disabled breakpoints", err);
// see if this is a temp breakpoint
if (i < NUMBER_OF_TEMP_BREAKPOINTS)
{
// this was a temp breakpoint, so we need to clear it now
err = DoClearBreak(i);
if (KErrNone != err)
LOG_MSG2("Error %d clearing temp breakpoint", err);
// if we're not out of range yet, go ahead and single step again
// if we are out of range, either continue or report the event depending
// on the iResumeOnceOutOfRange flag
if ((aEventInfo.iCurrentPC > breakEntry.iRangeStart) && (aEventInfo.iCurrentPC < breakEntry.iRangeEnd))
{
LOG_MSG("PC is still in range, stepping will continue");
err = DoStepRange(aCurrentThread, breakEntry.iRangeStart, breakEntry.iRangeEnd, breakEntry.iSteppingInto, breakEntry.iResumeOnceOutOfRange, breakEntry.iSteppingInto);
if (KErrNone != err)
{
LOG_MSG2("Error in DoStepRange: %d. Resuming thread.", err);
err = DoResumeThread(aCurrentThread);
if (KErrNone != err)
LOG_MSG2("Error in DoResumeThread: %d", err);
}
return;
}
else
{
if (breakEntry.iResumeOnceOutOfRange)
{
LOG_MSG("PC is out of range, continuing thread");
DoResumeThread(aCurrentThread);
return;
}
}
}
// if the breakpoint is thread specific, make sure it's the right thread
// if not, just continue the thread. take special care if it's the debugger
// thread. if it hits a regular breakpoint, we NEVER want to stop at it. if
// it hits a temp breakpoint, we're probably just stepping past a real breakpoint
// and we do need to handle it.
if (((breakEntry.iThreadId != aEventInfo.iThreadId) && breakEntry.iThreadSpecific)/*(breakEntry.iThreadId != 0xFFFFFFFF))*/ ||
((aEventInfo.iProcessId == iClientThread->iOwningProcess->iId) && (breakEntry.iThreadId != aEventInfo.iThreadId)))
{
LOG_MSG("breakpoint does not match threadId, calling DoResumeThread");
err = DoResumeThread(aCurrentThread);
if (KErrNone != err)
LOG_MSG2("Error in DoResumeThread: %d", err);
return;
}
//normal user break point, just notify the event
break;
}
}
}
}
NotifyEvent(aEventInfo);
}
//
// DMetroTrkChannel::HandleUserTrace
//
TBool DMetroTrkChannel::HandleUserTrace(TText* aStr, TInt aLen)
{
LOG_MSG("DMetroTrkChannel::HandleUserTrace()");
// handle user trace events only for app trk.
#ifndef __OEM_TRK__
// check to see if we are still debugging, otherwise just return
if (!iDebugging)
return EFalse;
DThread* currentThread = &Kern::CurrentThread();
if (!currentThread)
{
LOG_MSG("Error getting current thread");
return EFalse;
}
if (currentThread->iOwningProcess->iId == iClientThread->iOwningProcess->iId)
return EFalse;
// only send traces for processes you are debugging
TBool isDebugging = EFalse;
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == currentThread->iOwningProcess->iId)
{
isDebugging = ETrue;
break;
}
}
if (!isDebugging)
return EFalse;
if (aLen > 0)
{
SEventInfo info;
// This is a temporary solution for grabbing the rdebug printfs.
// With the current TRK implementation, we don't have a mechanism for
// using dynamically allocated buffers between the engine and the driver.
// So for now we are limiting the printf strings to 256 bytes.
TUint8 traceStr[260];
info.iTraceDataLen = aLen;
if (info.iTraceDataLen > 256)
info.iTraceDataLen = 256;
XTRAPD(r, XT_DEFAULT, kumemget(traceStr, aStr, info.iTraceDataLen));
if (r == KErrNone)
{
info.iEventType = SEventInfo::EUserTrace;
traceStr[info.iTraceDataLen] = '\r';
traceStr[info.iTraceDataLen+1] = '\n';
info.iTraceDataLen += 2; // account for \r and \n.
info.iTraceData.Copy((TUint8*)traceStr, info.iTraceDataLen);
NotifyEvent(info, ETrue);
}
else
{
return EFalse;
}
}
#endif
return ETrue;
}
//
// DMetroTrkChannel::SetBreak
//
TInt DMetroTrkChannel::SetBreak(TUint32 aThreadId, TMetroTrkBreakInfo* aBreakInfo)
{
LOG_MSG("DMetroTrkChannel::SetBreak()");
TInt err = KErrNone;
if (!aBreakInfo)
return KErrArgument;
//User side memory is not accessible directly
TMetroTrkBreakInfo info(0, 0, 0, 0);
err = Kern::ThreadRawRead(iClientThread, aBreakInfo, (TUint8*)&info, sizeof(TMetroTrkBreakInfo));
if (err != KErrNone)
return err;
if (!info.iId) //first check if the iId address is valid
return KErrArgument;
if (err == KErrNone)
{
TInt32 iId;
err = DoSetBreak(info.iProcessId, aThreadId, info.iAddress, info.iThumbMode, iId);
if (err == KErrNone)
{
err = Kern::ThreadRawWrite(iClientThread, (TUint8 *)info.iId, &iId, sizeof(TInt32), iClientThread);
}
}
return err;
}
//
// DMetroTrkChannel::StepRange
//
TInt DMetroTrkChannel::StepRange(DThread* aThread, TMetroTrkStepInfo* aStepInfo)
{
LOG_MSG("DMetroTrkChannel::StepRange()");
TInt err = KErrNone;
if (!aStepInfo)
return KErrArgument;
TMetroTrkStepInfo info(0, 0, 0);
err = Kern::ThreadRawRead(iClientThread, aStepInfo, (TUint8*)&info, sizeof(TMetroTrkStepInfo));
if (err != KErrNone)
return err;
err = DoStepRange(aThread, info.iStartAddress, info.iStopAddress, info.iStepInto, EFalse, ETrue);
return err;
}
//
// DMetroTrkChannel::ReadMemory
//
TInt DMetroTrkChannel::ReadMemory(DThread* aThread, TMetroTrkMemoryInfo* aMemoryInfo)
{
LOG_MSG("DMetroTrkChannel::ReadMemory()");
TInt err = KErrNone;
if (!aMemoryInfo)
return KErrArgument;
#ifndef __OEM_TRK__
if (!IsBeingDebugged(aThread))
return KErrPermissionDenied;
#endif
TMetroTrkMemoryInfo info(0, 0, 0);
err = Kern::ThreadRawRead(iClientThread, aMemoryInfo, (TUint8*)&info, sizeof(TMetroTrkMemoryInfo));
if (err != KErrNone)
return err;
if (!info.iData)
return KErrArgument;
TUint8 *data = (TUint8*)Kern::Alloc(info.iLength);
if (!data)
return KErrNoMemory;
TPtr8 dataDes(data, info.iLength);
err = DoReadMemory(aThread, info.iAddress, info.iLength, dataDes);
if (err != KErrNone)
return err;
err = Kern::ThreadDesWrite(iClientThread, info.iData, dataDes, 0, KChunkShiftBy0, iClientThread);
Kern::Free(data);
return err;
}
//
// DMetroTrkChannel::WriteMemory
//
TInt DMetroTrkChannel::WriteMemory(DThread* aThread, TMetroTrkMemoryInfo* aMemoryInfo)
{
LOG_MSG("DMetroTrkChannel::WriteMemory()");
TInt err = KErrNone;
if (!aMemoryInfo)
return KErrArgument;
#ifndef __OEM_TRK__
if (!IsBeingDebugged(aThread))
return KErrPermissionDenied;
#endif
TMetroTrkMemoryInfo info(0, 0, 0);
err = Kern::ThreadRawRead(iClientThread, aMemoryInfo, (TUint8*)&info, sizeof(TMetroTrkMemoryInfo));
if (err != KErrNone)
return err;
if (!info.iData)
return KErrArgument;
TUint8 *data = (TUint8*)Kern::Alloc(info.iLength);
if (!data)
return KErrNoMemory;
TPtr8 dataDes(data, info.iLength);
err = Kern::ThreadDesRead(iClientThread, info.iData, dataDes, 0);
if (err != KErrNone)
return err;
err = DoWriteMemory(aThread, info.iAddress, info.iLength, dataDes);
Kern::Free(data);
return err;
}
//
// DMetroTrkChannel::ReadRegisters
//
TInt DMetroTrkChannel::ReadRegisters(DThread* aThread, TMetroTrkRegisterInfo* aRegisterInfo)
{
LOG_MSG("DMetroTrkChannel::ReadRegisters()");
TInt err = KErrNone;
if (!aRegisterInfo)
return KErrArgument;
#ifndef __OEM_TRK__
if (!IsBeingDebugged(aThread))
return KErrPermissionDenied;
#endif
TMetroTrkRegisterInfo info(0, 0, 0);
err = Kern::ThreadRawRead(iClientThread, aRegisterInfo, (TUint8*)&info, sizeof(TMetroTrkRegisterInfo));
if (err != KErrNone)
return err;
if (!info.iValues)
return KErrArgument;
TUint length = (info.iLastRegister - info.iFirstRegister + 1) * 4;
TUint8 *values = (TUint8*)Kern::Alloc(length);
if (!values)
return KErrNoMemory;
TPtr8 valuesDes(values, length);
err = DoReadRegisters(aThread, info.iFirstRegister, info.iLastRegister, valuesDes);
if (err != KErrNone)
return err;
err = Kern::ThreadDesWrite(iClientThread, info.iValues, valuesDes, 0, KChunkShiftBy0, iClientThread);
Kern::Free(values);
return err;
}
//
// DMetroTrkChannel::WriteRegisters
//
TInt DMetroTrkChannel::WriteRegisters(DThread* aThread, TMetroTrkRegisterInfo* aRegisterInfo)
{
LOG_MSG("DMetroTrkChannel::WriteRegisters()");
TInt err = KErrNone;
if (!aRegisterInfo)
return KErrArgument;
#ifndef __OEM_TRK__
if (!IsBeingDebugged(aThread))
return KErrPermissionDenied;
#endif
TMetroTrkRegisterInfo info(0, 0, 0);
err = Kern::ThreadRawRead(iClientThread, aRegisterInfo, (TUint8*)&info, sizeof(TMetroTrkRegisterInfo));
if (err != KErrNone)
return err;
if (!info.iValues)
return KErrArgument;
TUint length = (info.iLastRegister - info.iFirstRegister + 1) * 4;
TUint8 *values = (TUint8*)Kern::Alloc(length);
if (!values)
return KErrNoMemory;
TPtr8 valuesDes(values, length);
err = Kern::ThreadDesRead(iClientThread, info.iValues, valuesDes, 0);
if (err != KErrNone)
return err;
err = DoWriteRegisters(aThread, info.iFirstRegister, info.iLastRegister, valuesDes);
Kern::Free(values);
return err;
}
//
// DMetroTrkChannel::GetProcessInfo
//
TInt DMetroTrkChannel::GetProcessInfo(TInt aIndex, TMetroTrkTaskInfo* aTaskInfo)
{
LOG_MSG("DMetroTrkChannel::GetProcessInfo()");
TInt err = KErrNone;
if (!aTaskInfo)
return KErrArgument;
TMetroTrkTaskInfo info(0);
err = DoGetProcessInfo(aIndex, &info);
if (err != KErrNone)
return err;
err = Kern::ThreadRawWrite(iClientThread, aTaskInfo, &info, sizeof(TMetroTrkTaskInfo), iClientThread);
return err;
}
//
// DMetroTrkChannel::GetThreadInfo
//
TInt DMetroTrkChannel::GetThreadInfo(TInt aIndex, TMetroTrkTaskInfo* aTaskInfo)
{
LOG_MSG("DMetroTrkChannel::GetThreadInfo()");
TInt err = KErrNone;
if (!aTaskInfo)
return KErrArgument;
TMetroTrkTaskInfo info(0);
err = Kern::ThreadRawRead(iClientThread, aTaskInfo, (TUint8*)&info, sizeof(TMetroTrkTaskInfo));
if (err != KErrNone)
return err;
err = DoGetThreadInfo(aIndex, &info);
if (err != KErrNone)
return err;
err = Kern::ThreadRawWrite(iClientThread, aTaskInfo, &info, sizeof(TMetroTrkTaskInfo), iClientThread);
return err;
}
//
// DMetroTrkChannel::GetProcessAddresses
//
TInt DMetroTrkChannel::GetProcessAddresses(DThread* aThread, TMetroTrkProcessInfo* aProcessInfo)
{
LOG_MSG("DMetroTrkChannel::GetProcessAddresses()");
TInt err = KErrNone;
if (!aProcessInfo)
return KErrArgument;
TMetroTrkProcessInfo info(0, 0);
err = Kern::ThreadRawRead(iClientThread, aProcessInfo, (TUint8*)&info, sizeof(TMetroTrkProcessInfo));
if (err != KErrNone)
return err;
if (!info.iCodeAddress || !info.iDataAddress)
return KErrArgument;
TUint32 codeAddress;
TUint32 dataAddress;
err = DoGetProcessAddresses(aThread, codeAddress, dataAddress);
if (err != KErrNone)
return err;
err = Kern::ThreadRawWrite(iClientThread, info.iCodeAddress, (TUint8*)&codeAddress, sizeof(TUint32), iClientThread);
err = Kern::ThreadRawWrite(iClientThread, info.iDataAddress, (TUint8*)&dataAddress, sizeof(TUint32), iClientThread);
return err;
}
//
// DMetroTrkChannel::GetStaticLibraryInfo
//
TInt DMetroTrkChannel::GetStaticLibraryInfo(TInt aIndex, SEventInfo* aEventInfo)
{
LOG_MSG("DMetroTrkChannel::GetStaticLibraryInfo()");
TInt err = KErrNone;
if (!aEventInfo)
return KErrArgument;
SEventInfo info;
err = Kern::ThreadRawRead(iClientThread, aEventInfo, (TUint8*)&info, sizeof(SEventInfo));
if (err != KErrNone)
return err;
err = DoGetStaticLibraryInfo(aIndex, &info);
if (err != KErrNone)
return err;
err = Kern::ThreadRawWrite(iClientThread, aEventInfo, &info, sizeof(SEventInfo), iClientThread);
return err;
}
//
// DMetroTrkChannel::GetLibInfo
//
TInt DMetroTrkChannel::GetLibraryInfo(TMetroTrkLibInfo* aLibInfo)
{
LOG_MSG("DMetroTrkChannel::GetLibraryInfo()");
TInt err = KErrNone;
if (!aLibInfo)
return KErrArgument;
TMetroTrkLibInfo info(0, 0);
err = Kern::ThreadRawRead(iClientThread, aLibInfo, (TUint8*)&info, sizeof(TMetroTrkLibInfo));
if (err != KErrNone)
return err;
if (!info.iFileName)
return KErrArgument;
TUint8 *dllName = (TUint8*)Kern::Alloc(info.iFileNameLength);
if (!dllName)
return KErrNoMemory;
TPtr8 dllNameDes(dllName, info.iFileNameLength);
err = Kern::ThreadDesRead(iClientThread, info.iFileName, dllNameDes, 0);
if (!err)
{
err = DoGetLibraryInfo(dllNameDes, &info);
if (!err)
err = Kern::ThreadRawWrite(iClientThread, aLibInfo, &info, sizeof(TMetroTrkLibInfo), iClientThread);
// we couldn't find the library info, so add to our list to look for
// when a library is loaded or a process is loaded.
if (err != KErrNone)
{
for (TInt i=0; i<NUMBER_OF_LIBS_TO_REGISTER; i++)
{
if (!iLibraryNotifyList[i].iName.Length() && iLibraryNotifyList[i].iEmptySlot)
{
iLibraryNotifyList[i].iName.Copy(dllNameDes);
iLibraryNotifyList[i].iEmptySlot = EFalse;
break;
}
else if (!iLibraryNotifyList[i].iEmptySlot && !_strnicmp(iLibraryNotifyList[i].iName.Ptr(), dllNameDes.Ptr(), dllNameDes.Length()))
{
break;
}
}
}
}
// now free the allocated memory
Kern::Free(dllName);
return err;
}
//
// DMetroTrkChannel::GetExeInfo
//
TInt DMetroTrkChannel::GetExeInfo(TMetroTrkExeInfo* aExeInfo)
{
LOG_MSG("DMetroTrkChannel::GetExeInfo()");
TInt err = KErrNone;
if (!aExeInfo)
return KErrArgument;
TMetroTrkExeInfo info(0, 0, 0);
err = Kern::ThreadRawRead(iClientThread, aExeInfo, (TUint8*)&info, sizeof(TMetroTrkExeInfo));
if (err != KErrNone)
return err;
if (!info.iFileName)
return KErrArgument;
TUint8 *exeName = (TUint8*)Kern::Alloc(info.iFileNameLength);
if (!exeName)
return KErrNoMemory;
TPtr8 exeNameDes(exeName, info.iFileNameLength);
err = Kern::ThreadDesRead(iClientThread, info.iFileName, exeNameDes, 0);
if (KErrNone == err)
{
err = DoGetExeInfo(exeNameDes, &info);
if (KErrNone == err)
{
err = Kern::ThreadRawWrite(iClientThread, aExeInfo, &info, sizeof(TMetroTrkExeInfo), iClientThread);
}
}
Kern::Free(exeName);
return err;
}
//
// DMetroTrkChannel::GetProcUidInfo
//
TInt DMetroTrkChannel::GetProcUidInfo(TMetroTrkProcUidInfo* aProcUidInfo)
{
LOG_MSG("DMetroTrkChannel::GetProcUidInfo()");
TInt err = KErrNone;
if (!aProcUidInfo)
return KErrArgument;
TMetroTrkProcUidInfo info(0);
err = Kern::ThreadRawRead(iClientThread, aProcUidInfo, (TUint8*)&info, sizeof(TMetroTrkProcUidInfo));
if (err != KErrNone)
return err;
err = DoGetProcUidInfo(&info);
if (err != KErrNone)
return err;
err = Kern::ThreadRawWrite(iClientThread, aProcUidInfo, &info, sizeof(TMetroTrkProcUidInfo), iClientThread);
return err;
}
//
//DMetroTrkChannel::DetachProcess
//
TInt DMetroTrkChannel::DetachProcess(DProcess *aProcess)
{
LOG_MSG("DMetroTrkChannel::DetachProcess()");
// check to see if we are still debugging, otherwise just return
if (!iDebugging || !aProcess)
return KErrArgument;
// This is called when a process has been detached.
// We want to mark any breakpoints in this process space as clear.
TUint32 codeAddress = 0;
TUint32 codeSize = 0;
LOG_MSG2("Process being Detached, Name %S", aProcess->iName);
// make sure there is not already a breakpoint at this address
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == aProcess->iId)
{
codeAddress = iDebugProcessList[i].iCodeAddress;
codeSize = iDebugProcessList[i].iCodeSize;
//now remove from the list
iDebugProcessList.Remove(i);
break;
}
}
if (!codeAddress || !codeSize)
return KErrArgument;
// first invalidate all breakpoints that were set in the library code
for (TInt i=0; i<iBreakPointList.Count(); i++)
{
if ((iBreakPointList[i].iAddress >= codeAddress) && (iBreakPointList[i].iAddress < (codeAddress + codeSize)))
{
LOG_MSG2("Clearing process breakpoint at address %x", iBreakPointList[i].iAddress);
// clear the break here
TInt32 err = KErrNone;
err = DoClearBreak(iBreakPointList[i].iId);
//If not able to clear the break point host debugger make obselete.
if (KErrNone != err)
{
LOG_MSG2("Clearing process breakpoint at address %x failed", iBreakPointList[i].iAddress);
}
}
}
return KErrNone;
}
//
// DMetroTrkChannel::DoSetBreak
//
TInt DMetroTrkChannel::DoSetBreak(const TUint32 aProcessId, const TUint32 aThreadId, const TUint32 aAddress, const TBool aThumbMode, TInt32 &aId)
{
LOG_MSG("DMetroTrkChannel::DoSetBreak()");
// do not allow breakpoints in the excluded ROM region
//if ((aAddress >= iExcludedROMAddressStart) && (aAddress < iExcludedROMAddressEnd))
//{
// return KErrNotSupported;
//}
// make sure there is not already a breakpoint at this address
for (TInt i = NUMBER_OF_TEMP_BREAKPOINTS; i < iBreakPointList.Count(); i++)
{
if (iBreakPointList[i].iAddress == aAddress)
{
return KErrAlreadyExists;
}
}
// increment the break id
aId = iNextBreakId++;
// create the new breakpoint entry
TBreakEntry breakEntry(aId, aThreadId, aAddress, aThumbMode);
// IDE currently sends non-thread specific breakpoints (threadId == 0xffffffff)
// if this changes, we will get a real thread ID
// This check must be done before calling DoEnableBreak as it will replace the
// the breakentry threadId with a real one
if (aThreadId == 0xFFFFFFFF)
{
breakEntry.iThreadSpecific = EFalse;
breakEntry.iThreadId = aProcessId+1; //process wide break, just use the main thread id for this process
}
else
{
breakEntry.iThreadSpecific = ETrue;
}
TInt err = DoEnableBreak(breakEntry, ETrue);
if (KErrNone == err)
{
ReturnIfError(iBreakPointList.Append(breakEntry));
}
return err;
}
//
// DMetroTrkChannel::DoEnableBreak
//
TInt DMetroTrkChannel::DoEnableBreak(TBreakEntry &aEntry, TBool aSaveOldInstruction)
{
LOG_MSG("DMetroTrkChannel::DoEnableBreak()");
// default to arm mode
TUint32 inst = KArmBreakPoint;
TInt instSize = 4;
if (aEntry.iThumbMode)
{
LOG_MSG("Thumb Breakpoint");
inst = KThumbBreakPoint;
instSize = 2;
}
TInt err = KErrNone;
// Get thread id from the process that we are debugging
TProcessInfo * proc = NULL;
TUint32 threadId = NULL;
threadId = aEntry.iThreadId;
if (!threadId || threadId==0xFFFFFFFF) //threadId=0xFFFFFFFF is special case with CW debugger.
{
for (TInt i=0; i<iDebugProcessList.Count(); i++)
{
proc = &iDebugProcessList[i];
if ( proc && (proc->iCodeAddress <= aEntry.iAddress) && (aEntry.iAddress <= (proc->iCodeAddress + proc->iCodeSize)))
{
threadId = proc->iId+1;
break;
}
}
}
DThread* threadObj = ThreadFromId(threadId);
//if we don't have the right thread id for the address,
//then try with the thread id of the process that we are debugging
if (!threadObj && iDebugProcessList.Count())
{
proc = &iDebugProcessList[0];
if (proc)
{
threadId = proc->iId+1;
}
threadObj = ThreadFromId(threadId);
}
if (threadObj)
{
if (aSaveOldInstruction)
{
TUint32 instruction;
// read the instruction at the address so we can store it in the break entry for when we clear this breakpoint
// trap exceptions in case the address is invalid
XTRAPD(r, XT_DEFAULT, err = TryToReadMemory(threadObj, (TAny *)aEntry.iAddress, (TAny *)&instruction, instSize));
ReturnIfError((KErrNone == r) ? err : r);
aEntry.iInstruction.Copy((TUint8 *)&instruction, instSize);
}
aEntry.iThreadId = threadId; //set the thread ID here
XTRAPD(r, XT_DEFAULT, err = DebugSupport::ModifyCode(threadObj, aEntry.iAddress, instSize, inst, DebugSupport::EBreakpointGlobal));
err = (DebugSupport::EBreakpointGlobal == r) ? KErrNone : r;
}
else
{
err = KErrBadHandle;
}
return err;
}
//
// DMetroTrkChannel::DoClearBreak
//
TInt DMetroTrkChannel::DoClearBreak(const TInt32 aId)
{
LOG_MSG("DMetroTrkChannel::DoClearBreak()");
// find the break entry matching this id. note that the breakpoints are already sorted in ascending order by id
TBreakEntry entry;
entry.iId = aId;
TInt index = iBreakPointList.FindInSignedKeyOrder(entry);
TInt err = KErrNone;
if (index >= 0)
{
// if this breakpoint was set in a library and that library has already been unloaded, don't try to clear it
if (!iBreakPointList[index].iObsoleteLibraryBreakpoint)
{
LOG_MSG2("Clearing breakpoint at address %x", iBreakPointList[index].iAddress);
DThread* threadObj = ThreadFromId(iBreakPointList[index].iThreadId);
// In case of multiple memory model, if the thread doesn't exist any more, don't try to clear it.
// For example it might be in a static library which was unloaded when the thread exited,
// but we didn't get told so we didn't mark it as iObsoleteLibraryBreakpoint.
if (threadObj != NULL)
{
XTRAPD(r, XT_DEFAULT, err = DebugSupport::RestoreCode(threadObj, iBreakPointList[index].iAddress));
err = (KErrNone == r) ? err : r;
}
else
{
err = KErrBadHandle;
}
}
if (KErrNone == err)
{
// if this is a temp breakpoint, just clear out the values, otherwise remove it from the list
if (index < NUMBER_OF_TEMP_BREAKPOINTS)
{
iBreakPointList[index].Reset();
}
else
{
iBreakPointList.Remove(index);
}
}
else
{
LOG_MSG2("Error clearing breakpoint %d", err);
}
return err;
}
LOG_MSG2("Break Id %d not found", aId);
return KErrNotFound;
}
//
// DMetroTrkChannel::DoChangeBreakThread
//
TInt DMetroTrkChannel::DoChangeBreakThread(TUint32 aThreadId, TInt32 aId)
{
LOG_MSG("DMetroTrkChannel::DoChangeBreakThread()");
// find the break entry matching this id. note that the breakpoints are already sorted in ascending order by id
TBreakEntry entry;
entry.iId = aId;
TInt index = iBreakPointList.FindInSignedKeyOrder(entry);
if (index >=0)
{
// change the thread id for this breakpoint
iBreakPointList[index].iThreadId = aThreadId;
return KErrNone;
}
return KErrNotFound;
}
//
// DMetroTrkChannel::DoSuspendThread
//
TInt DMetroTrkChannel::DoSuspendThread(DThread *aThread)
{
LOG_MSG("DMetroTrkChannel::DoSuspendThread()");
if (!aThread)
{
LOG_MSG("Invalid dthread object");
return KErrArgument;
}
NThread nThread = aThread->iNThread;
if (nThread.iSuspendCount == 0)
Kern::ThreadSuspend(*aThread, 1);
return KErrNone;
}
//
// DMetroTrkChannel::DoResumeThread
//
TInt DMetroTrkChannel::DoResumeThread(DThread *aThread)
{
LOG_MSG("DMetroTrkChannel::DoResumeThread()");
if (!aThread)
return KErrArgument;
// get the current PC
TUint32 currentPC = ReadRegister(aThread, PC_REGISTER);
// if there is a breakpoint at the current PC, we need to single step past it
for (TInt i=NUMBER_OF_TEMP_BREAKPOINTS; i<iBreakPointList.Count(); i++)
{
if (iBreakPointList[i].iAddress == currentPC)
{
return DoStepRange(aThread, currentPC, currentPC+1, ETrue, ETrue);
}
}
// PANIC_BACKPORT
// if frozen use semaphore
for(TInt i=0; i<iFrozenThreadSemaphores.Count(); i++)
{
if(iFrozenThreadSemaphores[i]->iOwningThread == &aThread->iNThread)
{
NKern::FSSignal(iFrozenThreadSemaphores[i]);
NKern::ThreadEnterCS();
delete iFrozenThreadSemaphores[i];
NKern::ThreadLeaveCS();
iFrozenThreadSemaphores.Remove(i);
return KErrNone;
}
}
// END PANIC_BACKPORT
// else use ThreadResume
Kern::ThreadResume(*aThread);
return KErrNone;
}
//
// DMetroTrkChannel::DoStepRange
//
TInt DMetroTrkChannel::DoStepRange(DThread *aThread, const TUint32 aStartAddress, const TUint32 aStopAddress, TBool aStepInto, TBool aResumeOnceOutOfRange, TBool aUserRequest)
{
LOG_MSG("DMetroTrkChannel::DoStepRange()");
if (!aThread)
return KErrArgument;
TUint32 startAddress = (aStartAddress & 0x1) ? aStartAddress + 1 : aStartAddress;
TUint32 stopAddress = (aStopAddress & 0x1) ? aStopAddress + 1 : aStopAddress;
// don't allow the user to step in the excluded ROM region. this could be called
// internally however. for example, the the special breakpoints we set to handle
// panics, exceptions, and library loaded events are in the user library, and we
// will need to step past the breakpoint before continuing the thread.
//if (aUserRequest && (startAddress >= iExcludedROMAddressStart) && (startAddress < iExcludedROMAddressEnd))
//{
// return KErrNotSupported;
//}
// set the temp breakpoint, and disable the breakpoint at the current PC if necessary
// if its not a user request, and we are just trying to resume from a breakpoint,
// then we don't need to check for stubs. The last parameter aUserRequest tells
// ModifyBreaksForStep to check for stubs or not. In some cases, the check for stubs
// is true even if its not a user request.For example, this is true in cases where
// we are doing a step range and the instruction in the range modified PC.
// in this case, DoStepRange will be called from the exception handler where
// we need to check for the stubs for the valid behavior. So truly, we don't need to check
// for stubs only when resuming from a breakpoint.
ReturnIfError(ModifyBreaksForStep(aThread, startAddress, stopAddress, aStepInto, aResumeOnceOutOfRange, aUserRequest));
// PANIC_BACKPORT
// if frozen use semaphore
for(TInt i=0; i<iFrozenThreadSemaphores.Count(); i++)
{
if(iFrozenThreadSemaphores[i]->iOwningThread == &aThread->iNThread)
{
NKern::FSSignal(iFrozenThreadSemaphores[i]);
NKern::ThreadEnterCS();
delete iFrozenThreadSemaphores[i];
NKern::ThreadLeaveCS();
iFrozenThreadSemaphores.Remove(i);
return KErrNone;
}
}
// END PANIC_BACKPORT
// else use ThreadResume
Kern::ThreadResume(*aThread);
return KErrNone;
}
//
// DMetroTrkChannel::DoReadMemory
//
TInt DMetroTrkChannel::DoReadMemory(DThread *aThread, const TUint32 aAddress, const TInt16 aLength, TDes8 &aData)
{
LOG_MSG("DMetroTrkChannel::DoReadMemory()");
// make sure the parameters are valid
if (aLength > aData.MaxSize())
return KErrArgument;
TInt err = KErrNone;
// trap exceptions in case the address is invalid
XTRAPD(r, XT_DEFAULT, err = TryToReadMemory(aThread, (TAny *)aAddress, (TAny *)aData.Ptr(), aLength));
err = (KErrNone == r) ? err : r;
if (KErrNone == err)
{
aData.SetLength(aLength);
TPtr8 data((TUint8 *)aData.Ptr(), aLength, aLength);
// if we have any breakpoints in this range, put the actual instruction in the buffer
for (TInt i = NUMBER_OF_TEMP_BREAKPOINTS; i < iBreakPointList.Count(); i++)
{
if ((iBreakPointList[i].iAddress >= aAddress) && (iBreakPointList[i].iAddress < (aAddress + aLength)))
{
TInt instSize = (TInt)(iBreakPointList[i].iThumbMode ? 2 : 4);
memcpy((TAny*)&data[iBreakPointList[i].iAddress - aAddress], (TAny *)iBreakPointList[i].iInstruction.Ptr(), instSize);
}
}
}
return err;
}
//
// DMetroTrkChannel::DoWriteMemory
//
TInt DMetroTrkChannel::DoWriteMemory(DThread *aThread, const TUint32 aAddress, const TInt16 aLength, TDes8 &aData)
{
LOG_MSG("DMetroTrkChannel::DoWriteMemory()");
// make sure the parameters are valid
if (aLength > aData.Length())
return KErrArgument;
TInt err = KErrNone;
// trap exceptions in case the address is invalid
XTRAPD(r, XT_DEFAULT, err = TryToWriteMemory(aThread, (TAny *)aAddress, (TAny *)aData.Ptr(), aLength));
err = (KErrNone == r) ? err : r;
// reset any breakpoints we may have just overwritten
if (KErrNone == err)
{
TPtr8 data((TUint8 *)aData.Ptr(), aLength, aLength);
for (TInt i = NUMBER_OF_TEMP_BREAKPOINTS; i < iBreakPointList.Count(); i++)
{
if ((iBreakPointList[i].iAddress >= aAddress) && (iBreakPointList[i].iAddress < (aAddress + aLength)))
{
// default to arm mode
TUint32 inst = KArmBreakPoint;
TInt instSize = 4;
if (iBreakPointList[i].iThumbMode)
{
inst = KThumbBreakPoint;
instSize = 2;
}
iBreakPointList[i].iInstruction.Copy(&data[iBreakPointList[i].iAddress - aAddress], instSize);
//memcpy((TAny*)iBreakPointList[i].iAddress, (TAny *)&inst, instSize);
TryToWriteMemory(aThread, (TAny*)iBreakPointList[i].iAddress, (TAny *)&inst, instSize);
}
}
}
return err;
}
//
// DMetroTrkChannel::DoReadRegisters
//
TInt DMetroTrkChannel::DoReadRegisters(DThread *aThread, const TInt16 aFirstRegister, const TInt16 aLastRegister, TDes8 &aValues)
{
LOG_MSG("DMetroTrkChannel::DoReadRegisters()");
// make sure the parameters are valid
if (!aThread || (aFirstRegister < 0) || (aLastRegister >= (TInt16)(sizeof(TArmRegSet)/sizeof(TArmReg))))
return KErrArgument;
// make sure the descriptor is big enough to hold the requested data
if ((TInt)((aLastRegister - aFirstRegister + 1) * sizeof(TArmReg)) > (aValues.MaxSize()))
return KErrArgument;
TArmRegSet regSet;
TUint32 unused;
#ifdef SUPPORT_KERNCONTEXT
NKern::Lock(); // lock the kernel before callin UserContextType as its required by this function
NThread nThread = aThread->iNThread;
NThread::TUserContextType userContextType = nThread.UserContextType();
NKern::Unlock(); //unlock the kernel now
if (userContextType == NThread::EContextNone || userContextType == NThread::EContextKernel)
{
//NKern::ThreadGetSystemContext(&aThread->iNThread, ®Set, unused);
if (!GetSystemThreadRegisters(®Set))
return KErrGeneral;
}
else
#endif
{
NKern::ThreadGetUserContext(&aThread->iNThread, ®Set, unused);
}
TArmReg *reg = ®Set.iR0;
if (!reg)
return KErrGeneral;
for (TInt16 i = aFirstRegister; i <= aLastRegister; i++)
aValues.Append((TUint8 *)®[i], sizeof(TArmReg));
return KErrNone;
}
//
// DMetroTrkChannel::DoWriteRegisters
//
TInt DMetroTrkChannel::DoWriteRegisters(DThread *aThread, const TInt16 aFirstRegister, const TInt16 aLastRegister, TDesC8 &aValues)
{
LOG_MSG("DMetroTrkChannel::DoWriteRegisters()");
// make sure the parameters are valid
if (!aThread || (aFirstRegister < 0) || (aLastRegister >= (TInt16)(sizeof(TArmRegSet)/sizeof(TArmReg))))
return KErrArgument;
// make sure the descriptor is big enough to hold the data to write
if ((TInt)((aLastRegister - aFirstRegister + 1) * sizeof(TArmReg)) > (aValues.Length()))
return KErrArgument;
TArmRegSet regSet;
TUint32 unused;
NKern::ThreadGetUserContext(&aThread->iNThread, ®Set, unused);
TArmReg *reg = ®Set.iR0;
for (TInt16 i = aFirstRegister; i <= aLastRegister; i++)
{
#ifndef __OEM_TRK__
if (IsRegisterSecure(i))
return KErrNotSupported;
#endif
reg[i] = *(TUint32 *)&aValues[(i-aFirstRegister)*sizeof(TArmReg)];
}
NKern::ThreadSetUserContext(&aThread->iNThread, ®Set);
return KErrNone;
}
//
// DMetroTrkChannel::DoGetProcessInfo
//
TInt DMetroTrkChannel::DoGetProcessInfo(const TInt aIndex, TMetroTrkTaskInfo *aInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetProcessInfo()");
DObjectCon *processes = Kern::Containers()[EProcess];
if (!processes)
return KErrGeneral;
TInt err = KErrNone;
NKern::ThreadEnterCS(); // Prevent us from dying or suspending whilst holding a DMutex
processes->Wait(); // Obtain the container mutex so the list does get changed under us
// make sure the index is valid
if ((aIndex >= 0) && (aIndex < processes->Count())) // >= because the index is zero based
{
DProcess *process = (DProcess *)((*processes)[aIndex]);
if (process)
{
process->Name(aInfo->iName);
aInfo->iId = process->iId;
aInfo->iPriority = (TUint32)process->iPriority;
}
else
{
LOG_MSG2("Process %d not found", aIndex);
err = KErrArgument;
}
}
else
{
err = KErrArgument;
}
processes->Signal();
NKern::ThreadLeaveCS();
return err;
}
//
// DMetroTrkChannel::DoGetThreadInfo
//
TInt DMetroTrkChannel::DoGetThreadInfo(const TInt aIndex, TMetroTrkTaskInfo *aInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetThreadInfo()");
DObjectCon *threads = Kern::Containers()[EThread];
if (!threads)
return KErrGeneral;
NKern::ThreadEnterCS(); // Prevent us from dying or suspending whilst holding a DMutex
threads->Wait(); // Obtain the container mutex so the list does get changed under us
TInt err = KErrNone;
if ((aIndex >= 0) && (aIndex < threads->Count()))
{
DThread *thread = NULL;
TInt threadsForProcessCount = 0;
TBool found = EFalse;
for (TInt i=0; i<threads->Count(); i++)
{
thread = (DThread *)((*threads)[i]);
if (thread && (aInfo->iOtherId == thread->iOwningProcess->iId))
{
if (threadsForProcessCount == aIndex)
{
thread->Name(aInfo->iName);
aInfo->iId = thread->iId;
aInfo->iPriority = (TUint32)thread->iThreadPriority;
found = ETrue;
break;
}
threadsForProcessCount++;
}
}
if (!found)
{
LOG_MSG("Thread for process matching index not found");
err = KErrArgument;
}
}
else
{
err = KErrArgument;
}
threads->Signal();
NKern::ThreadLeaveCS();
return err;
}
//
// DMetroTrkChannel::DoGetProcessAddresses
//
TInt DMetroTrkChannel::DoGetProcessAddresses(DThread *aThread, TUint32 &aCodeAddress, TUint32 &aDataAddress)
{
LOG_MSG("DMetroTrkChannel::DoGetProcessAddresses()");
if (!aThread)
return KErrArgument;
#ifndef __OEM_TRK__
if (HasManufacturerCaps(aThread))
return KErrPermissionDenied;
#endif
DProcess *process = (DProcess *)aThread->iOwningProcess;
if (!process)
return KErrArgument;
DCodeSeg* codeSeg = process->iCodeSeg;
if (!codeSeg)
return KErrArgument;
TModuleMemoryInfo processMemoryInfo;
TInt err = codeSeg->GetMemoryInfo(processMemoryInfo, process);
if (err != KErrNone)
return err;
aCodeAddress = processMemoryInfo.iCodeBase;
aDataAddress = processMemoryInfo.iInitialisedDataBase;
//add this process to the list of processes that we are debugging
TProcessInfo processInfo(process->iId, aCodeAddress, processMemoryInfo.iCodeSize, aDataAddress);
iDebugProcessList.Append(processInfo);
return KErrNone;
}
//
// DMetroTrkChannel::DoGetStaticLibraryInfo
//
TInt DMetroTrkChannel::DoGetStaticLibraryInfo(const TInt aIndex, SEventInfo *aInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetStaticLibraryInfo()");
if (!aInfo)
return KErrArgument;
DThread *thread = ThreadFromId(aInfo->iThreadId);
if (!thread)
return KErrArgument;
DProcess *process = (DProcess *)thread->iOwningProcess;
if (!process)
return KErrArgument;
DCodeSeg *processCodeSeg = process->iCodeSeg;
if (!processCodeSeg)
return KErrArgument;
int count = processCodeSeg->iDepCount;
LOG_MSG2("code segment count %d", count);
if (aIndex < 0 || aIndex >= count)
return KErrArgument;
DCodeSeg **codeSegList = processCodeSeg->iDeps;
if (!codeSegList)
return KErrArgument;
DCodeSeg* codeSeg = codeSegList[aIndex];
if (!codeSeg)
{
return KErrArgument;
}
if (!codeSeg->IsDll())
{
LOG_MSG(" -- code segment is not for a dll");
return KErrArgument;
}
TModuleMemoryInfo memoryInfo;
TInt err = codeSeg->GetMemoryInfo(memoryInfo, NULL); //NULL for DProcess should be ok;
if (err != KErrNone)
{
LOG_MSG2("Error in getting TModuleMemoryInfo info: %d", err);
return KErrArgument;
}
aInfo->iEventType = SEventInfo::ELibraryLoaded;
aInfo->iFileName.Copy(*(codeSeg->iFileName)); //just the name, without uid info.
aInfo->iCodeAddress = memoryInfo.iCodeBase;
aInfo->iDataAddress = memoryInfo.iInitialisedDataBase;
//this way, host debugger will not resume the thread, instead engine will resume
//after going through the list of all static libraries..
aInfo->iThreadId = 0xFFFFFFFF;
LOG_MSG2("library name: %S", codeSeg->iFileName);
LOG_MSG2("code address: %x", aInfo->iCodeAddress);
LOG_MSG2("data address: %x", aInfo->iDataAddress);
return KErrNone;
}
//
// DMetroTrkChannel::DoGetLibraryInfo
//
TInt DMetroTrkChannel::DoGetLibraryInfo(TDesC8 &aDllName, TMetroTrkLibInfo *aInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetLibraryInfo()");
TInt err = KErrArgument;
if (!aInfo)
return err;
err = DoGetLibInfoFromCodeSegList(aDllName, aInfo);
return err;
}
//
// DMetroTrkChannel::DoGetExeInfo
//
TInt DMetroTrkChannel::DoGetExeInfo(TDesC8 &aExeName, TMetroTrkExeInfo* aExeInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetExeInfo()");
DObjectCon *processes = Kern::Containers()[EProcess];
if (!processes)
return KErrGeneral;
NKern::ThreadEnterCS(); // Prevent us from dying or suspending whilst holding a DMutex
processes->Wait(); // Obtain the container mutex so the list does get changed under us
TInt err = KErrNotFound; //set err to KErrNotFound
for (TInt i=0; i < processes->Count(); i++) // >= because the index is zero based
{
DProcess *process = (DProcess *)((*processes)[i]);
if (process && (0x0 != aExeInfo->iUid) && (aExeInfo->iUid == process->iUids.iUid[2].iUid))
{
DCodeSeg* codeSeg = process->iCodeSeg;
DThread* mainThread = process->FirstThread();
if (codeSeg && mainThread)
{
aExeInfo->iProcessID = process->iId;
aExeInfo->iThreadID = mainThread->iId;
TModuleMemoryInfo memoryInfo;
err = codeSeg->GetMemoryInfo(memoryInfo, NULL);
if (KErrNone == err)
{
aExeInfo->iCodeAddress = memoryInfo.iCodeBase;
aExeInfo->iDataAddress = memoryInfo.iInitialisedDataBase;
break;
}
}
}
}
processes->Signal();
NKern::ThreadLeaveCS();
return err;
}
//
// DMetroTrkChannel::DoGetProcUidInfo
//
TInt DMetroTrkChannel::DoGetProcUidInfo(TMetroTrkProcUidInfo* aProcUidInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetProcUidInfo()");
TInt err = KErrNotFound;
DProcess* process = ProcessFromId(aProcUidInfo->iProcessID);
if (process)
{
aProcUidInfo->iUid1 = process->iUids.iUid[0].iUid;
aProcUidInfo->iUid2 = process->iUids.iUid[1].iUid;
aProcUidInfo->iUid3 = process->iUids.iUid[2].iUid;
aProcUidInfo->iSecurID = process->iS.iSecureId;
aProcUidInfo->iVendorID = process->iS.iVendorId;
err = KErrNone;
}
return err;
}
//
// DMetroTrkChannel::DoGetLibInfoFromCodeSegList
//
TInt DMetroTrkChannel::DoGetLibInfoFromCodeSegList(TDesC8 &aDllName, TMetroTrkLibInfo *aInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetLibInfoFromCodeSegList()");
TInt err = KErrArgument;
//get global code seg list
SDblQue* codeSegList = Kern::CodeSegList();
if (!codeSegList)
return KErrBadHandle;
//iterate through the list
for (SDblQueLink* codeSegPtr = codeSegList->First(); codeSegPtr!=(SDblQueLink*) (codeSegList); codeSegPtr=codeSegPtr->iNext)
{
DEpocCodeSeg* codeSeg = (DEpocCodeSeg*)_LOFF(codeSegPtr,DCodeSeg, iLink);
if (codeSeg && codeSeg->IsDll())
{
if (codeSeg->iFileName) //If this is valid, the rootname will also be valid.
{
// some dll names from the code segment list have some characters towards the end.
// Not sure why.
// To account for this, we need to compare those strings that are
// atleast as big as the dll we are looking for.
// Also the dll names in the code segment list don't have null terminator and so -1.
if (codeSeg->iRootName.Length() < aDllName.Length()-1)
continue;
if (!_strnicmp(codeSeg->iRootName.Ptr(), aDllName.Ptr(), aDllName.Length()-1))
{
TModuleMemoryInfo memoryInfo;
TInt err = codeSeg->GetMemoryInfo(memoryInfo, NULL);
if (err != KErrNone)
{
//there's been an error so return it
return err;
}
aInfo->iCodeAddress = memoryInfo.iCodeBase;
aInfo->iDataAddress = memoryInfo.iInitialisedDataBase;
if (codeSeg->iAttachProcess) //not valid if dll is used by multiple processes, so not reliable.
{
aInfo->iAttachProcessId = codeSeg->iAttachProcess->iId;
if (codeSeg->iAttachProcess->FirstThread())
aInfo->iAttachThreadId = codeSeg->iAttachProcess->FirstThread()->iId;
}
LOG_MSG2("Code segment found for lib: %s", aDllName.Ptr());
LOG_MSG2("code address: %x", aInfo->iCodeAddress);
LOG_MSG2("data address: %x", aInfo->iDataAddress);
return KErrNone;
}
}
}
}
LOG_MSG2("Code segment not found for lib: %s", aDllName.Ptr());
return err;
}
//
// DMetroTrkChannel::DoGetLibInfoFromKernLibContainer
//
TInt DMetroTrkChannel::DoGetLibInfoFromKernLibContainer(TDesC8 &aDllName, TMetroTrkLibInfo *aInfo)
{
LOG_MSG("DMetroTrkChannel::DoGetLibInfoFromKernLibContainer()");
TInt err = KErrNone;
DObjectCon *libraries = Kern::Containers()[ELibrary];
if (!libraries)
return KErrGeneral;
NKern::ThreadEnterCS(); // Prevent us from dying or suspending whilst holding a DMutex
libraries->Wait(); // Obtain the container mutex so the list does get changed under us
for (TInt i=0; i<libraries->Count(); i++)
{
DLibrary *library = (DLibrary *)((*libraries)[i]);
if (library)
{
TBuf<KMaxPath> libName;
libName.Copy(*(library->iName)); //just copy the name without the UID info.
if (libName.Length() < aDllName.Length()-1)
continue;
if (!_strnicmp(libName.Ptr(), aDllName.Ptr(), aDllName.Length()-1))
{
//get the code address
DCodeSeg* codeSeg = library->iCodeSeg;
if (codeSeg)
{
TModuleMemoryInfo memoryInfo;
TInt err = codeSeg->GetMemoryInfo(memoryInfo, NULL);
if (err == KErrNone)
{
// there's been an error so return it
aInfo->iCodeAddress = memoryInfo.iCodeBase;
aInfo->iDataAddress = memoryInfo.iInitialisedDataBase;
// process id and thread id are in DLibrary object.
}
}
else
{
LOG_MSG2("Code segment not available for library %S", library->iName);
err = KErrNotFound;
}
break;
}
}
}
libraries->Signal();
NKern::ThreadLeaveCS();
return err;
}
//
// DMetroTrkChannel::DoSecurityCheck
//
TBool DMetroTrkChannel::DoSecurityCheck()
{
DProcess* clientProcess = iClientThread->iOwningProcess;
if (clientProcess)
{
// now we also check to make sure that TRK app has ALLFILES capability as well.
if (!iClientThread->HasCapability(ECapabilityAllFiles))
return EFalse;
SSecurityInfo secureInfo = clientProcess->iS;
if ((secureInfo.iSecureId == KTrkSrvSecurUid) || (secureInfo.iSecureId == KTrkAppSecurUid) || (secureInfo.iSecureId == KTrkExeSecurUid))
{
return ETrue;
}
}
return EFalse;
}
//
// DMetroTrkChannel::TryToReadMemory
//
TInt DMetroTrkChannel::TryToReadMemory(DThread *aThread, TAny *aSrc, TAny *aDest, TInt16 aLength)
{
LOG_MSG("DMetroTrkChannel::TryToReadMemory()");
TInt err = KErrNone;
// on some targets reading from 0xFFFFFFFF address causes a kernel fault.
// avoid reading the last 4 bytes in the 32 bit address space.
TUint32 srcAddr = (TUint32)(aSrc);
if ((srcAddr >= 0xFFFFFFFC) || (aLength > (0xFFFFFFFC-srcAddr)))
return KErrAccessDenied;
#ifndef __OEM_TRK__
if (IsAddressInRom((TUint32)(aSrc)) || IsAddressSecure((TUint32)(aSrc)))
return KErrNotSupported;
#endif
//check if we have a valid thread object
if (!aThread)
return KErrBadHandle;
LOG_MSG2("Using Kern::ThreadRawRead to read memory at address %x", aSrc);
err = Kern::ThreadRawRead(aThread, aSrc, aDest, aLength);
return err;
}
//
// DMetroTrkChannel::TryToWriteMemory
//
TInt DMetroTrkChannel::TryToWriteMemory(DThread *aThread, TAny *aDest, TAny *aSrc, TInt16 aLength)
{
LOG_MSG("DMetroTrkChannel::TryToWriteMemory()");
TInt err = KErrNone;
// on some targets writing to 0xFFFFFFFF address causes a kernel fault.
// avoid writing the last 4 bytes in the 32 bit address space.
TUint32 destAddr = (TUint32)(aDest);
if ((destAddr>=0xFFFFFFFC) || (aLength > (0xFFFFFFFC-destAddr)))
return KErrAccessDenied;
#ifndef __OEM_TRK__
if (IsAddressInRom((TUint32)(aDest)) || IsAddressSecure((TUint32)(aDest)))
return KErrNotSupported;
#endif
//check if we have a valid thread object
if (!aThread)
return KErrBadHandle;
LOG_MSG2("Using Kern::ThreadRawWrite to write memory at address %x", (TUint32)aDest);
err = Kern::ThreadRawWrite(aThread, aDest, aSrc, aLength, iClientThread);
return err;
}
//
// DMetroTrkChannel::ReadRegister
//
TInt32 DMetroTrkChannel::ReadRegister(DThread *aThread, TInt aNum)
{
LOG_MSG("DMetroTrkChannel::ReadRegister()");
if (!aThread || (aNum < 0) || (aNum >= (TInt16)(sizeof(TArmRegSet)/sizeof(TArmReg))))
{
LOG_MSG2("Invalid register number (%d) passed to ReadRegister", aNum);
return 0;
}
TArmRegSet regSet;
TUint32 unused;
#ifdef SUPPORT_KERNCONTEXT
NKern::Lock(); // lock the kernel before callin UserContextType as its required by this function
NThread nThread = aThread->iNThread;
NThread::TUserContextType userContextType = nThread.UserContextType();
NKern::Unlock(); // unlock the kernel now
if (userContextType == NThread::EContextNone || userContextType == NThread::EContextKernel)
{
//NKern::ThreadGetSystemContext(&aThread->iNThread, ®Set, unused);
if (!GetSystemThreadRegisters(®Set))
return KErrGeneral;
}
else
#endif
{
NKern::ThreadGetUserContext(&aThread->iNThread, ®Set, unused);
}
TArmReg *reg = ®Set.iR0;
return ((TUint32 *)reg)[aNum];
}
//
// DMetroTrkChannel::ShiftedRegValue
//
TUint32 DMetroTrkChannel::ShiftedRegValue(DThread *aThread, TUint32 aInstruction, TUint32 aCurrentPC, TUint32 aStatusRegister)
{
LOG_MSG("DMetroTrkChannel::ShiftedRegValue()");
TUint32 shift = 0;
if (aInstruction & 0x10) // bit 4
{
shift = (ARM_RS(aInstruction) == PC_REGISTER ? aCurrentPC + 8 : aStatusRegister) & 0xFF;
}
else
{
shift = ARM_DATA_C(aInstruction);
}
TInt rm = ARM_RM(aInstruction);
TUint32 res = (rm == PC_REGISTER ? (aCurrentPC + ((aInstruction & 0x10) ? 12 : 8)) : ReadRegister(aThread, rm));
switch(ARM_DATA_SHIFT(aInstruction))
{
case 0: // LSL
{
res = shift >= 32 ? 0 : res << shift;
break;
}
case 1: // LSR
{
res = shift >= 32 ? 0 : res >> shift;
break;
}
case 2: // ASR
{
if (shift >= 32)
shift = 31;
res = ((res & 0x80000000L) ? ~((~res) >> shift) : res >> shift);
break;
}
case 3: // ROR/RRX
{
shift &= 31;
if (shift == 0)
{
res = (res >> 1) | ((aStatusRegister & ARM_CARRY_BIT) ? 0x80000000L : 0);
}
else
{
res = (res >> shift) | (res << (32 - shift));
}
break;
}
}
return res & 0xFFFFFFFF;
}
//
// DMetroTrkChannel::ModifyBreaksForStep
//
// Set a temporary breakpoint at the next instruction to be executed after the one at the current PC
// Disable the breakpoint at the current PC if one exists
//
TInt DMetroTrkChannel::ModifyBreaksForStep(DThread *aThread, TUint32 aRangeStart, TUint32 aRangeEnd, TBool aStepInto, TBool aResumeOnceOutOfRange, TBool aCheckForStubs)
{
LOG_MSG("DMetroTrkChannel::ModifyBreaksForStep()");
if (!aThread)
return KErrArgument;
LOG_MSG2("Range Start: %x", aRangeStart);
LOG_MSG2("Range End: %x", aRangeEnd);
// get the current PC
TUint32 currentPC = ReadRegister(aThread, PC_REGISTER);
LOG_MSG2("Current PC: %x", currentPC);
// disable breakpoint at the current PC if necessary
ReturnIfError(DisableBreakAtAddress(currentPC));
// get the status register
TUint32 statusRegister = ReadRegister(aThread, STATUS_REGISTER);
LOG_MSG2("Current SR: %x", statusRegister);
TBool thumbMode = (statusRegister & ECpuThumb);
if (thumbMode)
LOG_MSG("Thumb Mode");
TInt instSize = thumbMode ? 2 : 4;
TBool changingModes = EFalse;
TUint32 breakAddress = 0;
TInt rangeSize = aRangeEnd - currentPC;
const TInt KMaxInstructionBuffer = 80;
// scan the memory and see if any instruction might modify the PC. if one does,
// stop scanning and just set a breakpoint at that instruction (with the range set accordingly).
// if none is found just set the breakpoint after the range.
// if there is only one instruction there is no need to parse the memory, just execute it
if ((rangeSize <= KMaxInstructionBuffer) && (rangeSize > instSize))
{
LOG_MSG("Scanning range for instructions that might modify the PC");
// set it to the end of the range by default
breakAddress = aRangeEnd;
// get the instructions in range
// we really should be dynamically allocating this memory, but this could be
// called from another thread, so we can't
TBuf8<KMaxInstructionBuffer> instructions;
ReturnIfError(DoReadMemory(aThread, currentPC, rangeSize, instructions));
for (TInt i = 0; i < (TInt)rangeSize/instSize; i++)
{
if (InstructionModifiesPC(aThread, &instructions[i*instSize], thumbMode, aStepInto))
{
breakAddress = currentPC + i*instSize;
LOG_MSG2("Setting breakpoint at %x inside range", breakAddress);
break;
}
}
}
TUint32 newRangeEnd = aRangeEnd;
if ((breakAddress == 0) || (breakAddress == currentPC))
{
// either the range consists of a single instruction, or the instruction at the currentPC may modify the PC
// decode the instruction and see where we need to set the breakpoint
breakAddress = PCAfterInstructionExecutes(aThread, currentPC, statusRegister, instSize, aStepInto, newRangeEnd, changingModes);
// check to see if this is one of the stubs (found in stubs.s)
if (aStepInto && aCheckForStubs)
{
TBuf8<16> destination;
TInt err = DoReadMemory(aThread, breakAddress, 16, destination);
if (KErrNone == err)
{
TInt offset = 0;
if (0 == destination.Find(KArm4Stub, sizeof(KArm4Stub)))
{
LOG_MSG("Arm4 stub found");
offset = 8;
}
else if ((0 == destination.Find(KArmIStub, sizeof(KArmIStub))) ||
(0 == destination.Find(KFastArmIStub, sizeof(KFastArmIStub))))
{
LOG_MSG("ArmI stub found");
offset = 12;
}
else if ((0 == destination.Find(KThumbStub, sizeof(KThumbStub))) ||
(0 == destination.Find(KFastThumbStub, sizeof(KFastThumbStub))))
{
LOG_MSG("Thumb stub found");
offset = 12;
}
else if ((0 == destination.Find(KThumbStub2, sizeof(KThumbStub2))) ||
(0 == destination.Find(KFastThumbStub2, sizeof(KFastThumbStub2))))
{
LOG_MSG("Thumb stub found");
offset = 8;
}
//check to see if this is the stub generated with RVCT tools.
//Look into genstubs.cpp for more details on this stub
else if(0 == destination.Find(KRvctArm4Stub, sizeof(KRvctArm4Stub)))
{
LOG_MSG("RVCT Arm4 stub found");
offset = 4;
}
if (offset != 0)
{
if (offset == 4)
{
breakAddress = *(TUint32 *)&destination[offset];
}
else
{
err = DoReadMemory(aThread, *(TUint32 *)&destination[offset], 4, destination);
if (KErrNone == err)
breakAddress = *(TUint32 *)destination.Ptr();
}
if (KErrNone == err)
{
if (thumbMode)
{
if ((breakAddress & 0x00000001) == 1)
changingModes = EFalse;
else
changingModes = ETrue;
}
else
{
if ((breakAddress & 0x00000001) == 1)
changingModes = ETrue;
else
changingModes = EFalse;
}
breakAddress &= 0xFFFFFFFE;
}
else
{
LOG_MSG("Error reading destination of stub");
}
}
}
else
{
LOG_MSG("Error reading memory while decoding branch instruction");
}
}
// don't allow the user to step in a function in the excluded ROM region.
//if ((breakAddress >= iExcludedROMAddressStart) && (breakAddress < iExcludedROMAddressEnd))
//{
// breakAddress = currentPC + instSize;
// changingModes = EFalse;
//}
}
// see if there is already a breakpoint at this address. if there is, we do not need to set the temp breakpoint
for (TInt i=NUMBER_OF_TEMP_BREAKPOINTS; i<iBreakPointList.Count(); i++)
{
if (iBreakPointList[i].iAddress == breakAddress)
{
return KErrNone;
}
}
for (TInt i=0; i<NUMBER_OF_TEMP_BREAKPOINTS; i++)
{
if (iBreakPointList[i].iAddress == 0)
{
iBreakPointList[i].iThreadId = aThread->iId;
iBreakPointList[i].iAddress = breakAddress;
iBreakPointList[i].iThumbMode = (thumbMode && !changingModes) || (!thumbMode && changingModes);
iBreakPointList[i].iResumeOnceOutOfRange = aResumeOnceOutOfRange;
iBreakPointList[i].iSteppingInto = aStepInto;
iBreakPointList[i].iRangeStart = aRangeStart;
iBreakPointList[i].iRangeEnd = newRangeEnd;
// Temporary breakpoints are always thread specific
iBreakPointList[i].iThreadSpecific = ETrue;
LOG_MSG2("Adding temp breakpoint with id: %d", i);
LOG_MSG2("Adding temp breakpoint with thread id: %d", aThread->iId);
return DoEnableBreak(iBreakPointList[i], ETrue);
}
}
return KErrNoMemory;
}
//
// DMetroTrkChannel::ClearAllBreakPoints
//
void DMetroTrkChannel::ClearAllBreakPoints()
{
LOG_MSG("DMetroTrkChannel::ClearAllBreakPoints()");
TInt err = KErrNone;
for (TInt i=0; i<iBreakPointList.Count(); i++)
{
if ((iBreakPointList[i].iAddress != 0) && !iBreakPointList[i].iObsoleteLibraryBreakpoint)
{
LOG_MSG2("Clearing breakpoint at address %x", iBreakPointList[i].iAddress);
DThread* threadObj = ThreadFromId(iBreakPointList[i].iThreadId);
if (threadObj != NULL)
{
XTRAPD(r, XT_DEFAULT, err = DebugSupport::RestoreCode(threadObj, iBreakPointList[i].iAddress));
err = (KErrNone == r) ? err : r;
}
else
{
err = KErrBadHandle;
}
if (KErrNone != err)
{
LOG_MSG2("Error %d while clearing breakpoint", err);
}
}
}
iBreakPointList.Reset();
}
//
// DMetroTrkChannel::DisableBreakAtAddress
//
TInt DMetroTrkChannel::DisableBreakAtAddress(TUint32 aAddress)
{
LOG_MSG("DMetroTrkChannel::DisableBreakAtAddress()");
TInt err = KErrNone;
for (TInt i = NUMBER_OF_TEMP_BREAKPOINTS; i < iBreakPointList.Count(); i++)
{
if (iBreakPointList[i].iAddress == aAddress)
{
iBreakPointList[i].iDisabledForStep = ETrue;
LOG_MSG2("Disabling breakpoint at address %x", iBreakPointList[i].iAddress);
DThread* threadObj = ThreadFromId(iBreakPointList[i].iThreadId);
//clear the breakpoint with code modifier
//code modifier will restore the org instruction and also frees the shadow page if necessary
if (threadObj != NULL)
{
XTRAPD(r, XT_DEFAULT, err = DebugSupport::RestoreCode(threadObj, aAddress));
err = (KErrNone == r) ? err : r;
}
else
{
err = KErrBadHandle;
}
break;
}
}
return err;
}
//
// DMetroTrkChannel::InstructionModifiesPC
//
TBool DMetroTrkChannel::InstructionModifiesPC(DThread *aThread, TUint8 *aInstruction, TBool aThumbMode, TBool aStepInto)
{
LOG_MSG("DMetroTrkChannel::InstructionModifiesPC()");
if (aThumbMode)
{
TUint16 inst = *(TUint16 *)aInstruction;
switch(THUMB_OPCODE(inst))
{
case 0x08:
{
if (aStepInto && (THUMB_INST_7_15(inst) == 0x08F))
{
// BLX(2)
return ETrue;
}
else if (THUMB_INST_7_15(inst) == 0x08E)
{
// BX
return ETrue;
}
else if (((THUMB_INST_8_15(inst) == 0x46) || (THUMB_INST_8_15(inst) == 0x44)) && ((inst & 0x87) == 0x87))
{
// ADD or MOV with PC as the destination
return ETrue;
}
break;
}
case 0x13:
{
if (THUMB_INST_8_15(inst) == 0x9F)
{
// LDR(4) with the PC as the destination
return ETrue;
}
break;
}
case 0x17:
{
if (THUMB_INST_8_15(inst) == 0xBD)
{
// POP with the PC in the list
return ETrue;
}
break;
}
case 0x1A:
case 0x1B:
{
if (THUMB_INST_8_15(inst) < 0xDE)
{
// B(1) conditional branch
return ETrue;
}
break;
}
case 0x1C:
{
// B(2) unconditional branch
return ETrue;
}
case 0x1D:
{
// NOTE: Only return true for the suffix since it's the second instruction that actually does the branch
if (aStepInto && !(inst & 0x00000001))
{
// BLX(1)
return ETrue;
}
break;
}
case 0x1F:
{
// NOTE: Only return true for the suffix since it's the second instruction that actually does the branch
if (aStepInto)
{
// BL
return ETrue;
}
break;
}
}
}
else
{
// Arm mode
TUint32 inst = *(TUint32 *)aInstruction;
switch(ARM_OPCODE(inst)) // bits 27-25
{
case 0:
{
switch((inst & 0x00000010) >> 4) // bit 4
{
case 0:
{
switch((inst & 0x01800000) >> 23) // bits 24-23
{
case 2:
{
// move to/from status register. pc updates not allowed
// or TST, TEQ, CMP, CMN which don't modify the PC
break;
}
default:
{
// Data processing immediate shift
if (ARM_RD(inst) == PC_REGISTER)
{
// destination register is the PC
if (IsPreviousInstructionMovePCToLR(aThread))
{
return aStepInto;
}
else
{
return ETrue;
}
}
break;
}
}
break;
}
case 1:
{
switch((inst & 0x00000080) >> 7) // bit 7
{
case 0:
{
switch((inst & 0x01900000) >> 20) // bits 24-23 and bit 20
{
case 0x10:
{
// from figure 3-3
switch((inst & 0x000000F0) >> 4) // bits 7-4
{
case 1:
{
if (((inst & 0x00400000) >> 22) == 0) // bit 22
{
// BX
if (IsPreviousInstructionMovePCToLR(aThread))
{
return aStepInto;
}
else
{
return ETrue;
}
}
break;
}
case 3:
{
// BLX
if (aStepInto)
{
return ETrue;
}
break;
}
default:
{
// either doesn't modify the PC or it is illegal to
break;
}
}
break;
}
default:
{
// Data processing register shift
if (((inst & 0x01800000) >> 23) == 2) // bits 24-23
{
// TST, TEQ, CMP, CMN don't modify the PC
return EFalse;
}
else if (ARM_RD(inst) == PC_REGISTER)
{
// destination register is the PC
if (IsPreviousInstructionMovePCToLR(aThread))
{
return aStepInto;
}
else
{
return ETrue;
}
}
break;
}
}
break;
}
default:
{
// from figure 3-2, updates to the PC illegal
break;
}
}
break;
}
}
break;
}
case 1:
{
if (((inst & 0x01800000) >> 23) == 2) // bits 24-23
{
// cannot modify the PC
break;
}
else if (ARM_RD(inst) == PC_REGISTER)
{
// destination register is the PC
if (IsPreviousInstructionMovePCToLR(aThread))
{
return aStepInto;
}
else
{
return ETrue;
}
}
break;
}
case 2:
{
// load/store immediate offset
if (ARM_LOAD(inst)) // bit 20
{
// loading a register from memory
if (ARM_RD(inst) == PC_REGISTER)
{
// loading the PC register
if (IsPreviousInstructionMovePCToLR(aThread))
{
return aStepInto;
}
else
{
return ETrue;
}
}
}
break;
}
case 3:
{
if (((inst & 0xF0000000) != 0xF) && ((inst & 0x00000010) == 0))
{
// load/store register offset
if (ARM_LOAD(inst)) // bit 20
{
// loading a register from memory
if (ARM_RD(inst) == PC_REGISTER)
{
// loading the PC register
if (IsPreviousInstructionMovePCToLR(aThread))
{
return aStepInto;
}
else
{
return ETrue;
}
}
}
}
break;
}
case 4:
{
if ((inst & 0xF0000000) != 0xF)
{
// load/store multiple
if (ARM_LOAD(inst)) // bit 20
{
// loading a register from memory
if (((inst & 0x00008000) >> 15))
{
// loading the PC register
return ETrue;
}
}
}
break;
}
case 5:
{
if ((inst & 0xF0000000) == 0xF)
{
// BLX
if (aStepInto)
{
return ETrue;
}
}
else
{
if ((inst & 0x01000000)) // bit 24
{
// BL
if (aStepInto)
{
return ETrue;
}
}
else
{
// B
return ETrue;
}
}
break;
}
}
}
return EFalse;
}
//
// DMetroTrkChannel::PCAfterInstructionExecutes
//
TUint32 DMetroTrkChannel::PCAfterInstructionExecutes(DThread *aThread, TUint32 aCurrentPC, TUint32 aStatusRegister, TInt aInstSize, TBool aStepInto, TUint32 &aNewRangeEnd, TBool &aChangingModes)
{
LOG_MSG("DMetroTrkChannel::PCAfterInstructionExecutes()");
// by default we will set the breakpoint at the next instruction
TUint32 breakAddress = aCurrentPC + aInstSize;
// get the instruction at the current PC
TBuf8<4> instruction;
TInt err = DoReadMemory(aThread, aCurrentPC, aInstSize, instruction);
if (KErrNone != err)
{
return breakAddress;
}
if (4 == aInstSize)
{
TUint32 inst = *(TUint32 *)instruction.Ptr();
LOG_MSG2("Current instruction: %x", inst);
// check the conditions to see if this will actually get executed
if (IsExecuted(((inst>>28) & 0x0000000F), aStatusRegister))
{
switch(ARM_OPCODE(inst)) // bits 27-25
{
case 0:
{
switch((inst & 0x00000010) >> 4) // bit 4
{
case 0:
{
switch((inst & 0x01800000) >> 23) // bits 24-23
{
case 2:
{
// move to/from status register. pc updates not allowed
// or TST, TEQ, CMP, CMN which don't modify the PC
break;
}
default:
{
// Data processing immediate shift
if (ARM_RD(inst) == PC_REGISTER)
{
// destination register is the PC
if (IsPreviousInstructionMovePCToLR(aThread) && !aStepInto)
{
return breakAddress;
}
TUint32 rn = aCurrentPC + 8;
if (ARM_RN(inst) != PC_REGISTER) // bits 19-16
{
rn = ReadRegister(aThread, ARM_RN(inst));
}
TUint32 shifter = ShiftedRegValue(aThread, inst, aCurrentPC, aStatusRegister);
DecodeDataProcessingInstruction(((inst & 0x01E00000) >> 21), rn, shifter, aStatusRegister, breakAddress);
}
break;
}
}
break;
}
case 1:
{
switch((inst & 0x00000080) >> 7) // bit 7
{
case 0:
{
switch((inst & 0x01900000) >> 20) // bits 24-23 and bit 20
{
case 0x10:
{
// from figure 3-3
switch((inst & 0x000000F0) >> 4) // bits 7-4
{
case 1:
{
if (((inst & 0x00400000) >> 22) == 0) // bit 22
{
// BX
// this is a strange case. normally this is used in the epilogue to branch the the link
// register. sometimes it is used to call a function, and the LR is stored in the previous
// instruction. since what we want to do is different for the two cases when stepping over,
// we need to read the previous instruction to see what we should do
if (IsPreviousInstructionMovePCToLR(aThread) && !aStepInto)
{
return breakAddress;
}
breakAddress = ReadRegister(aThread, (inst & 0x0000000F));
if ((breakAddress & 0x00000001) == 1)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
break;
}
case 3:
{
// BLX
if (aStepInto)
{
breakAddress = ReadRegister(aThread, (inst & 0x0000000F));
if ((breakAddress & 0x00000001) == 1)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
break;
}
default:
{
// either doesn't modify the PC or it is illegal to
break;
}
}
break;
}
default:
{
// Data processing register shift
if (((inst & 0x01800000) >> 23) == 2) // bits 24-23
{
// TST, TEQ, CMP, CMN don't modify the PC
}
else if (ARM_RD(inst) == PC_REGISTER)
{
// destination register is the PC
if (IsPreviousInstructionMovePCToLR(aThread) && !aStepInto)
{
return breakAddress;
}
TUint32 rn = aCurrentPC + 8;
if (ARM_RN(inst) != PC_REGISTER) // bits 19-16
{
rn = ReadRegister(aThread, ARM_RN(inst));
}
TUint32 shifter = ShiftedRegValue(aThread, inst, aCurrentPC, aStatusRegister);
DecodeDataProcessingInstruction(((inst & 0x01E00000) >> 21), rn, shifter, aStatusRegister, breakAddress);
}
break;
}
}
break;
}
default:
{
// from figure 3-2, updates to the PC illegal
break;
}
}
break;
}
}
break;
}
case 1:
{
if (((inst & 0x01800000) >> 23) == 2) // bits 24-23
{
// cannot modify the PC
break;
}
else if (ARM_RD(inst) == PC_REGISTER)
{
// destination register is the PC
if (IsPreviousInstructionMovePCToLR(aThread) && !aStepInto)
{
return breakAddress;
}
TUint32 rn = ReadRegister(aThread, ARM_RN(inst)); // bits 19-16
TUint32 shifter = ((ARM_DATA_IMM(inst) >> ARM_DATA_ROT(inst)) | (ARM_DATA_IMM(inst) << (32 - ARM_DATA_ROT(inst)))) & 0xffffffff;
DecodeDataProcessingInstruction(((inst & 0x01E00000) >> 21), rn, shifter, aStatusRegister, breakAddress);
}
break;
}
case 2:
{
// load/store immediate offset
if (ARM_LOAD(inst)) // bit 20
{
// loading a register from memory
if (ARM_RD(inst) == PC_REGISTER)
{
// loading the PC register
if (IsPreviousInstructionMovePCToLR(aThread) && !aStepInto)
{
return breakAddress;
}
TUint32 base = ReadRegister(aThread, ARM_RN(inst));
TUint32 offset = 0;
if (ARM_SINGLE_PRE(inst))
{
// Pre-indexing
offset = ARM_SINGLE_IMM(inst);
if (ARM_SINGLE_U(inst))
{
base += offset;
}
else
{
base -= offset;
}
}
TBuf8<4> destination;
TInt err = DoReadMemory(aThread, base, 4, destination);
if (KErrNone == err)
{
breakAddress = *(TUint32 *)destination.Ptr();
if ((breakAddress & 0x00000001) == 1)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
else
{
LOG_MSG("Error reading memory in decoding step instruction");
}
}
}
break;
}
case 3:
{
if (((inst & 0xF0000000) != 0xF) && ((inst & 0x00000010) == 0))
{
// load/store register offset
if (ARM_LOAD(inst)) // bit 20
{
// loading a register from memory
if (ARM_RD(inst) == PC_REGISTER)
{
// loading the PC register
if (IsPreviousInstructionMovePCToLR(aThread) && !aStepInto)
{
return breakAddress;
}
TUint32 base = (ARM_RN(inst) == PC_REGISTER) ? aCurrentPC + 8 : ReadRegister(aThread, ARM_RN(inst));
TUint32 offset = 0;
if (ARM_SINGLE_PRE(inst))
{
offset = ShiftedRegValue(aThread, inst, aCurrentPC, aStatusRegister);
if (ARM_SINGLE_U(inst))
{
base += offset;
}
else
{
base -= offset;
}
}
TBuf8<4> destination;
TInt err = DoReadMemory(aThread, base, 4, destination);
if (KErrNone == err)
{
breakAddress = *(TUint32 *)destination.Ptr();
if ((breakAddress & 0x00000001) == 1)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
else
{
LOG_MSG("Error reading memory in decoding step instruction");
}
}
}
}
break;
}
case 4:
{
if ((inst & 0xF0000000) != 0xF)
{
// load/store multiple
if (ARM_LOAD(inst)) // bit 20
{
// loading a register from memory
if (((inst & 0x00008000) >> 15))
{
// loading the PC register
TInt offset = 0;
if (ARM_BLOCK_U(inst))
{
TUint32 reglist = ARM_BLOCK_REGLIST(inst);
offset = Bitcount(reglist) * 4 - 4;
if (ARM_BLOCK_PRE(inst))
offset += 4;
}
else if (ARM_BLOCK_PRE(inst))
{
offset = -4;
}
TUint32 temp = ReadRegister(aThread, ARM_RN(inst));
temp += offset;
TBuf8<4> destination;
TInt err = DoReadMemory(aThread, temp, 4, destination);
if (KErrNone == err)
{
breakAddress = *(TUint32 *)destination.Ptr();
if ((breakAddress & 0x00000001) == 1)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
else
{
LOG_MSG("Error reading memory in decoding step instruction");
}
}
}
}
break;
}
case 5:
{
if ((inst & 0xF0000000) == 0xF)
{
// BLX
if (aStepInto)
{
breakAddress = (TUint32)ARM_INSTR_B_DEST(inst, aCurrentPC);
if ((breakAddress & 0x00000001) == 1)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
}
else
{
if ((inst & 0x01000000)) // bit 24
{
// BL
if (aStepInto)
{
breakAddress = (TUint32)ARM_INSTR_B_DEST(inst, aCurrentPC);
}
}
else
{
// B
breakAddress = (TUint32)ARM_INSTR_B_DEST(inst, aCurrentPC);
}
}
break;
}
}
}
}
else
{
// Thumb Mode
TUint16 inst = *(TUint16 *)instruction.Ptr();
LOG_MSG2("Current instruction: %x", inst);
switch(THUMB_OPCODE(inst))
{
case 0x08:
{
if (aStepInto && (THUMB_INST_7_15(inst) == 0x08F))
{
// BLX(2)
breakAddress = ReadRegister(aThread, ((inst & 0x0078) >> 3));
if ((breakAddress & 0x00000001) == 0)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
else if (THUMB_INST_7_15(inst) == 0x08E)
{
// BX
breakAddress = ReadRegister(aThread, ((inst & 0x0078) >> 3));
if ((breakAddress & 0x00000001) == 0)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
else if ((THUMB_INST_8_15(inst) == 0x46) && ((inst & 0x87) == 0x87))
{
// MOV with PC as the destination
breakAddress = ReadRegister(aThread, ((inst & 0x0078) >> 3));
}
else if ((THUMB_INST_8_15(inst) == 0x44) && ((inst & 0x87) == 0x87))
{
// ADD with PC as the destination
breakAddress = aCurrentPC + ReadRegister(aThread, ((inst & 0x0078) >> 3));
}
break;
}
case 0x13:
{
//This instruction doesn't modify the PC.
//if (THUMB_INST_8_15(inst) == 0x9F)
//{
// LDR(4) with the PC as the destination
// breakAddress = ReadRegister(aThread, SP_REGISTER) + (4 * (inst & 0x00FF));
//}
break;
}
case 0x17:
{
if (THUMB_INST_8_15(inst) == 0xBD)
{
// POP with the PC in the list
TUint32 regList = (inst & 0x00FF);
TInt offset = ReadRegister(aThread, SP_REGISTER) + (Bitcount(regList) * 4);
TBuf8<4> destination;
TInt err = DoReadMemory(aThread, offset, 4, destination);
if (KErrNone == err)
{
breakAddress = *(TUint32 *)destination.Ptr();
if ((breakAddress & 0x00000001) == 0)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFE;
}
else
{
LOG_MSG("Error reading memory in decoding step instruction");
}
}
break;
}
case 0x1A:
case 0x1B:
{
if (THUMB_INST_8_15(inst) < 0xDE)
{
// B(1) conditional branch
if (IsExecuted(((inst & 0x0F00) >> 8), aStatusRegister))
{
TUint32 offset = ((inst & 0x000000FF) << 1);
if (offset & 0x00000100)
{
offset |= 0xFFFFFF00;
}
breakAddress = aCurrentPC + 4 + offset;
}
}
break;
}
case 0x1C:
{
// B(2) unconditional branch
TUint32 offset = (inst & 0x000007FF) << 1;
if (offset & 0x00000800)
{
offset |= 0xFFFFF800;
}
breakAddress = aCurrentPC + 4 + offset;
break;
}
case 0x1D:
{
if (aStepInto && !(inst & 0x0001))
{
// BLX(1)
breakAddress = (ReadRegister(aThread, LINK_REGISTER) + ((inst & 0x07FF) << 1));
if ((breakAddress & 0x00000001) == 0)
{
aChangingModes = ETrue;
}
breakAddress &= 0xFFFFFFFC;
}
break;
}
case 0x1E:
{
// BL/BLX prefix - destination is encoded in this and the next instruction
aNewRangeEnd += 2;
// BL Stepping Changes
if (aStepInto)
{
TBuf8<2> nextInstruction;
TInt err = DoReadMemory(aThread, aCurrentPC+2, aInstSize, nextInstruction);
if (KErrNone == err)
{
TUint16 nextInst = *(TUint16*)nextInstruction.Ptr();
LOG_MSG2("Next instruction: %x", nextInst);
// base new LR from first instruction
TUint32 newLR = inst & 0x07FF;
newLR <<= 12;
if (newLR & (TUint32)(0x0400 << 12))
{
// sign-extend it
newLR |= 0xFF800000;
}
newLR += aCurrentPC + 4; // pc is pointing to first instruction + 4
TUint32 newPC = newLR + ((nextInst & 0x07FF) << 1);
if (THUMB_OPCODE(nextInst) == 0x1D) // H = 01 changing to ARM (BLX)
{
newPC &= 0xFFFFFFFC;
aChangingModes = ETrue;
}
breakAddress = newPC;
}
}
else // step over
{
breakAddress += 2;
}
// End of BL Stepping Changes
break;
}
case 0x1F:
{
if (aStepInto)
{
// BL
breakAddress = ReadRegister(aThread, LINK_REGISTER) + ((inst & 0x07FF) << 1);
}
break;
}
}
}
return breakAddress;
}
//
// DMetroTrkChannel::DecodeDataProcessingInstruction
//
void DMetroTrkChannel::DecodeDataProcessingInstruction(TUint8 aOpcode, TUint32 aOp1, TUint32 aOp2, TUint32 aStatusRegister, TUint32 &aBreakAddress)
{
LOG_MSG("DMetroTrkChannel::DecodeDataProcessingInstruction()");
switch(aOpcode)
{
case 0:
{
// AND
aBreakAddress = aOp1 & aOp2;
break;
}
case 1:
{
// EOR
aBreakAddress = aOp1 ^ aOp2;
break;
}
case 2:
{
// SUB
aBreakAddress = aOp1 - aOp2;
break;
}
case 3:
{
// RSB
aBreakAddress = aOp2 - aOp1;
break;
}
case 4:
{
// ADD
aBreakAddress = aOp1 + aOp2;
break;
}
case 5:
{
// ADC
aBreakAddress = aOp1 + aOp2 + (aStatusRegister & ARM_CARRY_BIT) ? 1 : 0;
break;
}
case 6:
{
// SBC
aBreakAddress = aOp1 - aOp2 - (aStatusRegister & ARM_CARRY_BIT) ? 0 : 1;
break;
}
case 7:
{
// RSC
aBreakAddress = aOp2 - aOp1 - (aStatusRegister & ARM_CARRY_BIT) ? 0 : 1;
break;
}
case 12:
{
// ORR
aBreakAddress = aOp1 | aOp2;
break;
}
case 13:
{
// MOV
aBreakAddress = aOp2;
break;
}
case 14:
{
// BIC
aBreakAddress = aOp1 & ~aOp2;
break;
}
case 15:
{
// MVN
aBreakAddress = ~aOp2;
break;
}
}
}
//
// DMetroTrkChannel::IsPreviousInstructionMovePCToLR
//
TBool DMetroTrkChannel::IsPreviousInstructionMovePCToLR(DThread *aThread)
{
LOG_MSG("DMetroTrkChannel::IsPreviousInstructionMovePCToLR()");
// there are several types of instructions that modify the PC that aren't
// designated as linked or non linked branches. the way gcc generates the
// code can tell us whether or not these instructions are to be treated as
// linked branches. the main cases are bx and any type of mov or load or
// arithmatic operation that changes the PC. if these are really just
// function calls that will return, gcc will generate a mov lr, pc
// instruction as the previous instruction. note that this is just for arm
// and armi
// get the address of the previous instruction
TUint32 address = ReadRegister(aThread, PC_REGISTER) - 4;
TBuf8<4> previousInstruction;
TInt err = DoReadMemory(aThread, address, 4, previousInstruction);
if (KErrNone != err)
{
LOG_MSG2("Error %d reading memory at address %x", address);
return EFalse;
}
const TUint32 movePCToLRIgnoringCondition = 0x01A0E00F;
TUint32 inst = *(TUint32 *)previousInstruction.Ptr();
if ((inst & 0x0FFFFFFF) == movePCToLRIgnoringCondition)
{
return ETrue;
}
return EFalse;
}
//
// DMetroTrkChannel::DoEnableDisabledBreak
//
// Restore the breakpoint that was disabled for stepping past it if necessary
//
TInt DMetroTrkChannel::DoEnableDisabledBreak(TUint32 aThreadId)
{
LOG_MSG("DMetroTrkChannel::DoEnableDisabledBreak()");
for (TInt i = NUMBER_OF_TEMP_BREAKPOINTS; i < iBreakPointList.Count(); i++)
{
// if (iBreakPointList[i].iDisabledForStep && ((iBreakPointList[i].iThreadId == aThreadId) || (iBreakPointList[i].iThreadId == 0xFFFFFFFF)))
// Always re-enable non-Thread Specific breakpoints
if (iBreakPointList[i].iDisabledForStep && ((iBreakPointList[i].iThreadId == aThreadId) || (iBreakPointList[i].iThreadSpecific == EFalse)))
{
LOG_MSG2("Re-enabling breakpoint at address %x", iBreakPointList[i].iAddress);
iBreakPointList[i].iDisabledForStep = EFalse;
return DoEnableBreak(iBreakPointList[i], EFalse);
}
}
return KErrNone;
}
//
// DMetroTrkChannel::IsExecuted
//
// Determines whether or not an instruction will be executed
//
TBool DMetroTrkChannel::IsExecuted(TUint8 aCondition ,TUint32 aStatusRegister)
{
LOG_MSG("DMetroTrkChannel::IsExecuted()");
TBool N = ((aStatusRegister >> 28) & 0x0000000F) & 0x00000008;
TBool Z = ((aStatusRegister >> 28) & 0x0000000F) & 0x00000004;
TBool C = ((aStatusRegister >> 28) & 0x0000000F) & 0x00000002;
TBool V = ((aStatusRegister >> 28) & 0x0000000F) & 0x00000001;
switch(aCondition)
{
case 0:
return Z;
case 1:
return !Z;
case 2:
return C;
case 3:
return !C;
case 4:
return N;
case 5:
return !N;
case 6:
return V;
case 7:
return !V;
case 8:
return (C && !Z);
case 9:
return (!C || Z);
case 10:
return (N == V);
case 11:
return (N != V);
case 12:
return ((N == V) && !Z);
case 13:
return (Z || (N != V));
case 14:
case 15:
return ETrue;
}
return EFalse;
}
TBool DMetroTrkChannel::IsAddressInRom(TUint32 aAddress)
{
LOG_MSG("DMetroTrkChannel::IsAddressInRom()");
TRomHeader romHeader = Epoc::RomHeader();
if ((aAddress >= romHeader.iRomBase ) && (aAddress < (romHeader.iRomBase + romHeader.iRomSize)))
return ETrue;
return EFalse;
}
TBool DMetroTrkChannel::IsAddressSecure(TUint32 aAddress)
{
LOG_MSG("DMetroTrkChannel::IsAddressInRom()");
// SHORT TERM FIX...
// For now, don't allow access to Kernel memory...
TLinAddr kernDataStartAddr = iMultipleMemModel ? KSuperPageMultipleLinAddr: KSuperPageMovingLinAddr;
TLinAddr kernDataEndAddr = iMultipleMemModel ? KKernDataEndMultipleLinAddr: KKernDataEndMovingLinAddr;
if ((aAddress >= kernDataStartAddr) && (aAddress <= kernDataEndAddr))
return ETrue;
return EFalse;
}
TBool DMetroTrkChannel::IsRegisterSecure(TInt registerIndex)
{
LOG_MSG("DMetroTrkChannel::IsRegisterSecure()");
if (registerIndex == SP_REGISTER || registerIndex == LINK_REGISTER || registerIndex == STATUS_REGISTER)
return ETrue;
return EFalse;
}
//
// DMetroTrkChannel::AllocateShadowPageIfNecessary
//
// Allocate a shadow page if the address is in ROM and no page has been allocated for that range yet
//
TInt DMetroTrkChannel::AllocateShadowPageIfNecessary(TUint32 aAddress, TUint32 &aPageAddress)
{
LOG_MSG("DMetroTrkChannel::AllocateShadowPageIfNecessary()");
// if this is in ROM, we need to shadow memory
TBool inRom = EFalse;
//find out if M::IsRomAddress is available from ekern or any other library
//ReturnIfError(inRom = M::IsRomAddress((TAny *)aAddress));
TRomHeader romHeader = Epoc::RomHeader();
if ((aAddress >= romHeader.iRomBase ) && (aAddress < (romHeader.iRomBase + romHeader.iRomSize)))
inRom = ETrue;
TInt err = KErrNone;
if (inRom) // && (TSuperPage().iRomConfig[0].iType != KBdbBankTypeRamAsRom))
{
// if a shadow page has not already been allocated for this address range, do so now
TUint32 pageAddress = (aAddress & ~(iPageSize-1));
TBool found = EFalse;
for (TInt i=0; i<iBreakPointList.Count(); i++)
{
if (pageAddress == iBreakPointList[i].iPageAddress)
{
LOG_MSG2("Shadow page already allocated at address %x", pageAddress);
found = ETrue;
}
}
if (!found)
{
LOG_MSG2("Allocating shadow page starting at address %x", pageAddress);
err = Epoc::AllocShadowPage(pageAddress);
if (KErrNone != err)
{
LOG_MSG2("Error %d allocating shadow page", err);
return KErrGeneral;
}
//check to see if this is still necessary
//if ((ASSPID() & ASSP_ID_MASK) == XSCALE_ASSP_ID)
//FlushDataCache(); // workaround for cache flush problem on Lubbock/PXA255 base port
}
// return the page address
aPageAddress = pageAddress;
}
return err;
}
//
// DMetroTrkChannel::FreeShadowPageIfNecessary
//
// Free a shadow page if the address is in ROM and no no other breakpoints are set in that page range
//
TInt DMetroTrkChannel::FreeShadowPageIfNecessary(TUint32 aAddress, TUint32 aPageAddress)
{
LOG_MSG("DMetroTrkChannel::FreeShadowPageIfNecessary()");
TInt err = KErrNone;
TBool found = EFalse;
TBool inRom = EFalse;
//find out if M::IsRomAddress is available from ekern or any other library
//ReturnIfError(inRom = M::IsRomAddress((TAny *)aAddress));
TRomHeader romHeader = Epoc::RomHeader();
if ((aAddress >= romHeader.iRomBase ) && (aAddress < (romHeader.iRomBase + romHeader.iRomSize)))
inRom = ETrue;
if (inRom)// && (TSuperPage().iRomConfig[0].iType != KBdbBankTypeRamAsRom))
{
// if there are no other breakpoints in this shadow page, we can free it now
for (TInt i = 0; i < iBreakPointList.Count(); i++)
{
if (aAddress != iBreakPointList[i].iAddress) // ignore the current entry
{
if ((iBreakPointList[i].iAddress >= aPageAddress) &&
(iBreakPointList[i].iAddress < (aPageAddress + iPageSize)))
{
LOG_MSG("Still a breakpoint in this shadow page range");
found = ETrue;
break;
}
}
}
if (!found)
{
LOG_MSG2("Freeing shadow page starting at address %x", aPageAddress);
err = Epoc::FreeShadowPage(aPageAddress);
if (KErrNone != err)
{
LOG_MSG2("Error %d freeing shadow page", err);
return KErrGeneral;
}
}
}
return err;
}
//
// DMetroTrkChannel::NotifyEvent
//
void DMetroTrkChannel::NotifyEvent(SEventInfo aEventInfo, TBool isTraceEvent)
{
LOG_MSG("DMetroTrkChannel::NotifyEvent()");
if (iEventInfo)
{
LOG_MSG2("Completing event, type: %d", aEventInfo.iEventType);
// iThread is the user side debugger thread, so use it to write the info to it memory
TInt err = Kern::ThreadRawWrite(iClientThread, iEventInfo, (TUint8 *)&aEventInfo, sizeof(SEventInfo), iClientThread);
if (KErrNone != err)
LOG_MSG2("Error writing event info: %d", err);
// clear this since we've completed the request
iEventInfo = NULL;
// signal the debugger thread
Kern::RequestComplete(iClientThread, iRequestGetEventStatus, KErrNone);
}
else
{
if (isTraceEvent)
{
LOG_MSG("Queuing trace event\r\n");
for (TInt i=0; i<NUMBER_OF_EVENTS_TO_QUEUE; i++)
{
if (SEventInfo::EUnknown == iTraceEventQueue[i].iEventType)
{
iTraceEventQueue[i] = aEventInfo;
break;
}
}
}
else
{
LOG_MSG2("Queuing event, type: %d", aEventInfo.iEventType);
for (TInt i=0; i<NUMBER_OF_EVENTS_TO_QUEUE; i++)
{
if (SEventInfo::EUnknown == iEventQueue[i].iEventType)
{
iEventQueue[i] = aEventInfo;
break;
}
}
}
}
}
//
// DMetroTrkChannel::ThreadFromId
//
DThread* DMetroTrkChannel::ThreadFromId(TUint32 aId)
{
LOG_MSG("DMetroTrkChannel::ThreadFromId()");
NKern::ThreadEnterCS(); // Prevent us from dying or suspending whilst holding a DMutex
DObjectCon& threads = *Kern::Containers()[EThread]; // Get containing holding threads
threads.Wait(); // Obtain the container mutex so the list does get changed under us
DThread* thread = Kern::ThreadFromId(aId);
threads.Signal(); // Release the container mutex
NKern::ThreadLeaveCS(); // End of critical section
return thread;
}
//
// DMetroTrkChannel::ProcessFromId
//
DProcess* DMetroTrkChannel::ProcessFromId(TUint32 aId)
{
LOG_MSG("DMetroTrkChannel::ProcessFromId()");
NKern::ThreadEnterCS(); // Prevent us from dying or suspending whilst holding a DMutex
DObjectCon& processes = *Kern::Containers()[EProcess]; // Get containing holding threads
processes.Wait(); // Obtain the container mutex so the list does get changed under us
DProcess* process = Kern::ProcessFromId(aId);
processes.Signal(); // Release the container mutex
NKern::ThreadLeaveCS(); // End of critical section
return process;
}
//
// DMetroTrkChannel::GetSystemThreadRegisters
//
TBool DMetroTrkChannel::GetSystemThreadRegisters(TArmRegSet* aArmRegSet)
{
if (iExcInfoValid)
{
aArmRegSet->iR0 = iCurrentExcInfo.iR0;
aArmRegSet->iR1 = iCurrentExcInfo.iR1;
aArmRegSet->iR2 = iCurrentExcInfo.iR2;
aArmRegSet->iR3 = iCurrentExcInfo.iR3;
aArmRegSet->iR4 = iCurrentExcInfo.iR4;
aArmRegSet->iR5 = iCurrentExcInfo.iR5;
aArmRegSet->iR6 = iCurrentExcInfo.iR6;
aArmRegSet->iR7 = iCurrentExcInfo.iR7;
aArmRegSet->iR8 = iCurrentExcInfo.iR8;
aArmRegSet->iR9 = iCurrentExcInfo.iR9;
aArmRegSet->iR10 = iCurrentExcInfo.iR10;
aArmRegSet->iR11 = iCurrentExcInfo.iR11;
aArmRegSet->iR12 = iCurrentExcInfo.iR12;
aArmRegSet->iR13 = iCurrentExcInfo.iR13;
aArmRegSet->iR14 = iCurrentExcInfo.iR14;
aArmRegSet->iR15 = iCurrentExcInfo.iR15;
aArmRegSet->iFlags = iCurrentExcInfo.iCpsr;
return ETrue;
}
return EFalse;
}
TBool DMetroTrkChannel::HasManufacturerCaps(DThread* aThread)
{
if (aThread && (aThread->HasCapability(ECapabilityTCB) ||
aThread->HasCapability(ECapabilityDRM) ||
aThread->HasCapability(ECapabilityAllFiles)))
{
return ETrue;
}
return EFalse;
}
TBool DMetroTrkChannel::IsBeingDebugged(const DThread* aThread)
{
TBool isDebugging = EFalse;
if (aThread)
{
for (TInt i = 0; i < iDebugProcessList.Count(); i++)
{
if (iDebugProcessList[i].iId == aThread->iOwningProcess->iId)
{
isDebugging = ETrue;
break;
}
}
}
return isDebugging;
}
void DMetroTrkChannel::CheckLibraryNotifyList(TUint32 aProcessId)
{
SDblQue* codeSegList = Kern::CodeSegList();
for (TInt i=0; i<iLibraryNotifyList.Count(); i++)
{
if (!iLibraryNotifyList[i].iEmptySlot)
{
//iterate through the list
for (SDblQueLink* codeSegPtr = codeSegList->First(); codeSegPtr != (SDblQueLink*)(codeSegList); codeSegPtr = codeSegPtr->iNext)
{
DEpocCodeSeg* codeSeg = (DEpocCodeSeg*)_LOFF(codeSegPtr, DCodeSeg, iLink);
if (codeSeg && codeSeg->IsDll())
{
SEventInfo info;
if (!_strnicmp(iLibraryNotifyList[i].iName.Ptr(), codeSeg->iRootName.Ptr(), codeSeg->iRootName.Length()))
{
info.iFileName.Copy((codeSeg->iRootName.Ptr()));
LOG_MSG2("library name match: %S", &info.iFileName);
TModuleMemoryInfo memoryInfo;
TInt err = codeSeg->GetMemoryInfo(memoryInfo, NULL);
if (err != KErrNone)
{
break;
}
//
// check the process id from the DCodeSeg instead of using aProcessId
// if the iAttachProcess from DCodeSeg is null, then we should use aProcessId
info.iProcessId = aProcessId;
info.iEventType = SEventInfo::ELibraryLoaded;
info.iCodeAddress = memoryInfo.iCodeBase;
info.iDataAddress = memoryInfo.iInitialisedDataBase;
info.iThreadId = 0xFFFFFFFF;
LOG_MSG2("info.iCodeAddress: %x", info.iCodeAddress);
LOG_MSG2("info.iDataAddress: %x", info.iDataAddress);
LOG_MSG2("info.iProcessId: %d", info.iProcessId);
NotifyEvent(info);
// now reset this entry
iLibraryNotifyList[i].iEmptySlot = ETrue;
break;
}
}
}
}
}
}
DECLARE_STANDARD_LDD()
{
return new DMetroTrkDriverFactory;
}