--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/kernel/eka/debug/crashMonitor/src/scmonitor.cpp Mon Oct 19 15:55:17 2009 +0100
@@ -0,0 +1,1032 @@
+// Copyright (c) 2008-2009 Nokia Corporation and/or its subsidiary(-ies).
+// All rights reserved.
+// This component and the accompanying materials are made available
+// under the terms of the License "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:
+// e32\debug\crashMonitor\src\scmonitor.cpp
+// Core dump server - Kernel side crash monitor
+//
+//
+
+/**
+ @file
+ @internalTechnology
+*/
+
+#include <scmonitor.h>
+#include <kernel/monitor.h>
+#include <assp.h>
+#include <drivers/crashflash.h>
+#include <kernel/klib.h>
+#include <crashlogwalker.h>
+#include <scmconfigitem.h>
+
+#include "scmdatasave.h"
+
+GLDEF_D SCMonitor TheSCMonitor; //global definition of SCMonitor
+
+//keep things 4 byte aligned
+const TInt KRestartType = SCMonitor::ESoftRestart;
+
+/**
+SCMonitor constructor
+*/
+SCMonitor::SCMonitor()
+ : iMultiCrashInfo(NULL)
+ {
+ }
+
+SCMonitor::~SCMonitor()
+ {
+ delete iMultiCrashInfo;
+ }
+
+/**
+ Print data to the corresponding output channel. Derived from monitor
+ @param aDes the buffer containing the data
+ */
+void SCMonitor::Print (const TDesC8& aDes )
+ {
+ //intended to do nothing
+ }
+
+/**
+ * Allocates resources for SCMonitor
+ * cant fully construct in constructor as we are a kernel extension and resources are limited when we are created
+ */
+void SCMonitor::StableConstruction()
+ {
+ LOG_CONTEXT
+ iDataSave = new SCMDataSave(this, TheSCMonitor.iFlash);
+
+ //Configuration object for use upon crash
+ iScmConfig = new SCMConfiguration();
+ TInt err = iScmConfig->SetDefaultConfig();
+ if(KErrNone != err)
+ {
+ CLTRACE1("SCMonitor::StableConstruction - Unable to set default config err = %d", err);
+ }
+
+
+#ifdef NO_MULTICRASHINFO
+ iMultiCrashInfo = NULL;
+#else
+
+ //We need to take a look at the flash map from variant_norflash_layout.h
+ iMultiCrashInfo = new SCMMultiCrashInfo();
+
+ TUint numberBlocks = KCrashLogSize / KCrashLogBlockSize;
+ for(TUint32 cnt = 0; cnt < numberBlocks; cnt++)
+ {
+ iMultiCrashInfo->AddBlock(new SCMCrashBlockEntry(cnt, cnt * KCrashLogBlockSize, KCrashLogBlockSize));
+ }
+#endif
+ }
+
+/**
+ * Start a secondary DFC queue for the Flash and Init the flash in the variant(h4)
+ * @param aAny
+ */
+void StartSecondary (TAny* )
+ {
+ LOG_CONTEXT
+ //InitFlash is implemented in the variant as it creates a variant
+ //specific derived CrashFlash
+ TheSCMonitor.InitFlash ( );
+ TheSCMonitor.StableConstruction();
+ }
+
+/**
+ * Global method to create a dfc queue
+ * @param Method to intialise the flash.
+ * @param Null
+ * @param Gets the address of the supervisor thread DFC queue
+ * @param TDfcQ priority number
+ * @return a DFC object
+ */
+GLDEF_C TDfc StartSecondaryDfc(&StartSecondary, NULL, Kern::SvMsgQue(), KMaxDfcPriority-1);
+
+/**
+ * Kernel Main module entry - Own implementation( similar to crash logger)
+ * @param aReason reason to enter to the method
+ * @return One of the system wide codes
+ */
+GLDEF_C TInt KernelModuleEntry(TInt aReason)
+ {
+ if(aReason==KModuleEntryReasonVariantInit0)
+ {
+ new(&TheSCMonitor) SCMonitor;
+ // We are going to register the system Crash monitor here so that the order
+ // the monitor modules are placed in rom is preserved.
+ // The monitor is not fully intialised here.
+ //the variant target is missing as we still have to finalise on the crash flash
+ //implementation. H2 & H4 doesnt support currently.
+ LOG_CONTEXT
+ CLTRACE("Installing System Crash Monitor");
+ Monitor::RegisterMonitorImpl (&TheSCMonitor );
+ return KErrNone;
+ }
+ else if (aReason==KModuleEntryReasonExtensionInit0 )
+ {
+ return KErrNone;
+ }
+ else if (aReason==KModuleEntryReasonExtensionInit1 )
+ {
+ LOG_CONTEXT
+ CLTRACE("Enqueing dfc to init crash flash for System Crash Monitor after all modules loaded");
+ StartSecondaryDfc.Enque ( );
+ return KErrNone;
+ }
+ return KErrArgument;
+ }
+
+/**
+ Method to intialize the system crash monitor
+ @param aCategory the fault category type
+ @param aReason the reason for crash
+ @return restart type
+ */
+TInt SCMonitor::Init2 (TAny* aCategory, TInt aReason )
+ {
+ LOG_CONTEXT
+ __KTRACE_OPT(KALWAYS, Kern::Printf("\n\nSystem Crash Monitor Launched: To Analyse Crash Produced Use Core Dump Server\n"));
+
+ //Start logging the data:
+ //Need to lock kernel to access object containers (it technically is anyway, but flag isnt set)
+ NKern::Lock();
+ DoCrash(aCategory, aReason);
+ NKern::Unlock();
+
+ __KTRACE_OPT(KALWAYS, Kern::Printf("System Crash Monitor Finished: Log Size = [%d]\n", iDataSave->iCrashInf.iLogSize));
+
+ return KRestartType;
+ }
+
+/**
+ * This is responsible for setting up any structures required for processing of the crash
+ * @param aCategory the fault category type
+ * @param aReason
+ */
+void SCMonitor::DoCrash(TAny* aCategory, TInt aReason )
+ {
+ // get debug mask
+ TInt dbgMask = Kern::SuperPage().iDebugMask[0];
+
+ // if we are writing to the comm port then we need to turn off other debug messages
+ if( iDataSave->GetWriteSelect() == SCMDataSave::EWriteComm)
+ {
+ Kern::SuperPage().iDebugMask[0] = 0;
+ }
+
+ if(!aCategory)
+ {
+ CLTRACE("\tNULL category retrieved and returning");
+ TheSCMonitor.iFlash->EndTransaction();
+ return;
+ }
+
+ iFrame = NULL;
+
+ CLTRACE("\tAbout to set category -- note: can occasionaly crash board");
+ iFaultCategory = *(const TDesC8*)aCategory; // this crashes the board sometimes
+ iFaultReason = aReason;
+ Epoc::SetMonitorExceptionHandler ((TLinAddr)HandleException );
+
+ // get the first start block
+ // will retieve start of flash by default
+ SCMCrashBlockEntry block;
+ TInt err = GetNextCrashStartPoint(block); // will also attempt to read iScmConfig
+
+ if(KErrNone == err)
+ {
+ CLTRACE2("SCMonitor::DoCrash next crash will be written at blocknumber = %d offset %d"
+ , block.iBlockNumber, block.iBlockOffset);
+ }
+ else
+ {
+ CLTRACE1("SCMonitor::DoCrash Failed to find a valid block to write to, can not continue. err = [%d]", err);
+ return;
+ }
+
+ TUint crashId = block.iBlockNumber;
+ iDataSave->iWriter->ResetBytesWritten();
+
+ //Write the crash (1st pass is to gather header data)
+ TInt spaceRequired = ProcessCrash(block, crashId, EFalse);
+
+ // now do the real write
+ // prepare flash for data
+ TheSCMonitor.iFlash->StartTransaction();
+ TheSCMonitor.iFlash->SetWritePos(block.iBlockOffset);
+
+ //write the crash this time
+ ProcessCrash(block, crashId, ETrue);
+
+ TheSCMonitor.iFlash->EndTransaction();
+
+ // restore debug mask
+ Kern::SuperPage().iDebugMask[0] = dbgMask;
+ }
+
+
+/**
+ * This walks the existing crash log and finds out where current crashes finish
+ * @param aBlockEntry Block to use. Only valid if KErrNone is returned.
+ * @return One of the OS wide codes
+ */
+TInt SCMonitor::GetNextCrashStartPoint(SCMCrashBlockEntry& aBlockEntry)
+ {
+ LOG_CONTEXT
+
+ //First thing is to try and read the config
+ TBool configFound = (iDataSave->ReadConfig(*iScmConfig) == KErrNone);
+
+ if( iMultiCrashInfo)
+ {
+ /**
+ * data save has been configured to use multicrash info to find the next block we are on we need to scan each
+ * block to see if it contains a valid header. if we find an empty block in our block list then that is the
+ * one we will use if we find no empty blocks then we have no room left
+ */
+ iMultiCrashInfo->Reset();
+ SCMCrashBlockEntry* block = iMultiCrashInfo->GetNextBlock();
+ TBool blockFound = EFalse;
+
+ //For any crashes in flash, we need to record where they end, so that we can then go to the next
+ //block after the one in which it ends
+ TInt crashEndPoint = 0;
+
+ while(block)
+ {
+ CLTRACE1("SCMonitor::GetNextCrashStartPoint Processing block number %d", block->iBlockNumber );
+
+ //If we have already found our block, we should erase subsequent ones for use
+ if(blockFound)
+ {
+ TInt err = EraseFlashBlock(*block);
+ if(err != KErrNone)
+ {
+ return err;
+ }
+
+ block = iMultiCrashInfo->GetNextBlock();
+ continue;
+ }
+
+ //is this block before a crash end? if it is, we cant use it as a crash can span multiple blocks
+ if(block->iBlockOffset >= crashEndPoint)
+ {
+ //special condition if we have a config
+ TUint startPos = block->iBlockOffset;
+ TUint skipBytes = 0;
+ if(configFound && block->iBlockOffset == 0)
+ {
+ startPos+=iScmConfig->GetSize();
+
+ //must align to flash for read
+ skipBytes = startPos % KFlashAlignment;
+ startPos -= skipBytes;
+ }
+
+ // try and read an info header at these flash coords
+ TBuf8<TCrashInfoHeader::KSCMCrashInfoMaxSize + KFlashAlignment> buf;
+ buf.SetLength(TCrashInfoHeader::KSCMCrashInfoMaxSize + KFlashAlignment);
+
+ CLTRACE1("(SCMonitor::GetNextCrashStartPoint) reading at offset %d", block->iBlockOffset);
+
+ TheSCMonitor.iFlash->SetReadPos(startPos);
+ TheSCMonitor.iFlash->Read(buf);
+
+ // create the buffer applying the offset of bytes skipped
+ TByteStreamReader reader(const_cast<TUint8*> (buf.Ptr() + skipBytes));
+
+ TCrashInfoHeader header;
+ TInt err = header.Deserialize(reader);
+
+ if(err == KErrCorrupt)
+ {
+ CLTRACE2("(SCMonitor::GetNextCrashStartPoint) Found empty block blocknumber %d blockoffset = %d"
+ , block->iBlockNumber, block->iBlockOffset);
+
+ blockFound = ETrue;
+ aBlockEntry = *block;
+
+ continue; //Dont get next block, as next run will erase this current block for use
+ }
+ else
+ {
+ crashEndPoint = header.iLogSize + startPos;
+ CLTRACE3("(SCMonitor::GetNextCrashStartPoint) In block [%d] we found a valid crash header. This crash finishes at [%d] [0x%X]", block->iBlockNumber, crashEndPoint, crashEndPoint);
+ }
+ }
+
+ block = iMultiCrashInfo->GetNextBlock();
+ }
+
+ if(blockFound)
+ {
+ return KErrNone;
+ }
+ else
+ {
+ //CLTRACE("(SCMonitor::GetNextCrashStartPoint) No available blocks TREATING as NO MULTICRASH INFO will write to default block");
+ //In this case should we just overwrite old crashes and return the first block as the comment above suggests
+ //return blockFound;
+ }
+ }
+
+ // no multi crash info supplied - use default first block settings
+ TInt err = EraseEntireFlashPartition();
+ if(err != KErrNone)
+ {
+ CLTRACE1("Unable to delete area required to log to flash. Aborting. Error - [%d]", err);
+ return err;
+ }
+
+ aBlockEntry = SCMCrashBlockEntry(0,0,0);
+ return KErrNone;
+ }
+
+/**
+ * Handles the processing of the crash
+ * @return The size of the crash log (including header) that has been/will be written
+ */
+TInt SCMonitor::ProcessCrash(const SCMCrashBlockEntry& aBlock, TUint aCrashId, TBool aCommit)
+ {
+ LOG_CONTEXT
+ CLTRACE5("aBlock.iBlockOffset = [%d] [0x%X] aBlock.iBlockNumber = %d aBlock.iBlockSize = [%d] [0x%X]",
+ aBlock.iBlockOffset, aBlock.iBlockOffset, aBlock.iBlockNumber, aBlock.iBlockSize, aBlock.iBlockSize);
+
+ // reset writer for start of each crash
+ iDataSave->iWriter->ResetBytesWritten();
+ TInt logLevel = 0;
+
+ if(aCommit)
+ {
+ logLevel = KALWAYS;
+ iDataSave->iWriter->EnablePhysicalWriting();
+ }
+ else
+ {
+#if defined(_DEBUG)
+ logLevel = KDEBUGGER;
+#else
+ logLevel = KALWAYS; //Doesnt matter, KTRACE OPT is empty for rel builds
+ if(logLevel != KALWAYS)
+ {
+ //This is to avoid warning
+ }
+#endif
+
+ iDataSave->iWriter->DisablePhysicalWriting();
+ }
+
+ iDataSave->SetByteCount(aBlock.iBlockOffset);
+ if(aBlock.iBlockOffset == 0 && aBlock.iBlockNumber == 0)
+ {
+ // this is the first crash - we need to save the config here first
+ CLTRACE("(SCMonitor::ProcessCrash) - this is block 0 - WRITING CONFIG");
+ iDataSave->LogConfig(*iScmConfig);
+
+ //Config is not part of crash so reset bytes written
+ iDataSave->SetCrashStartingPoint(iDataSave->iWriter->GetBytesWritten());
+ }
+ else
+ {
+ iDataSave->SetCrashStartingPoint(aBlock.iBlockOffset);
+ }
+
+ iDataSave->iWriter->ResetBytesWritten();
+
+ TUint32 logSize = 0;
+ TUint sizeOfObjectDumped = 0;
+
+ TInt err = iDataSave->LogCrashHeader(iFaultCategory, iFaultReason, aCrashId, sizeOfObjectDumped);
+ if(KErrNone != err)
+ {
+ CLTRACE("System Crash Monitor: Failed to create crash info header - (TCrashInfo)");
+ return KRestartType;
+ }
+
+ logSize += sizeOfObjectDumped;
+
+ //Now we must read the configuration to use. This is held at the start of our flash partition
+ //and managed by the iConfig object
+ iScmConfig->ResetToHighestPriority();
+
+ //Always want the crash context
+ iDataSave->iHdr.iCTFullRegOffset = logSize + iDataSave->GetCrashStartingPoint();
+
+ err = iDataSave->LogCPURegisters(sizeOfObjectDumped);
+ if(KErrNone != err)
+ {
+ CLTRACE1("\tError logging full registers = %d", err);
+ }
+
+ logSize += sizeOfObjectDumped;
+
+ CLTRACE("\tAbout to enter processing loop");
+ SCMDataSave::TDataToDump dump;
+
+ for(;;)
+ {
+ //now we get each item by priority from the configuration
+ TConfigItem* configItem = iScmConfig->GetNextItem();
+
+ if(!configItem)
+ {
+ // end of list
+ break;
+ }
+
+ CLTRACE1("\nLooking at item type [%d]", configItem->GetDataType());
+ if(configItem->GetSpaceRequired() > iDataSave->SpaceRemaining())
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\t\tFor Item Type [%d]: Unable to log [0x%X] [%d] bytes because we only have [0x%X] [%d] bytes left", configItem->GetDataType(), configItem->GetSpaceRequired(), configItem->GetSpaceRequired(), iDataSave->SpaceRemaining(), iDataSave->SpaceRemaining()));
+ continue;
+ }
+ else
+ {
+ CLTRACE1("Will require [%d] bytes for this item", configItem->GetSpaceRequired());
+ }
+
+ // only interested in logging items with priority > 0
+ if( configItem->GetPriority() <= 0)
+ {
+ CLTRACE1("\tIgnored config item type %d priority 0", configItem->GetDataType());
+ continue;
+ }
+
+ //there are a lot of TUints in the hdr to record where we wrote this item.
+ //This will point to the one of interest for this configItem
+ TUint32* offsetPointer = NULL;
+
+ //now we check the type of data we wish to dump
+ switch(configItem->GetDataType())
+ {
+ case TConfigItem::ECrashedThreadMetaData:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ECrashedThreadMetaData at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = LogThreadMetaData(SCMDataSave::EThreadSpecific, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iCTMetaOffset);
+
+ break;
+ }
+ case TConfigItem::EThreadsMetaData:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EThreadsMetaData at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //record location we are writing to in the header
+ iDataSave->iHdr.iTLstOffset = iDataSave->iWriter->GetBytesWritten();
+ err = LogThreadMetaData(SCMDataSave::ESystemWide, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iTLstOffset);
+
+ break;
+ }
+ case TConfigItem::ECrashedProcessMetaData:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ECrashedProcessMetaData at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = LogProcessMetaData(SCMDataSave::EProcessSpecific, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iCPMetaOffset);
+
+ break;
+ }
+ case TConfigItem::EProcessMetaData:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EProcessMetaData at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = LogProcessMetaData(SCMDataSave::ESystemWide, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iPLstOffset);
+
+ break;
+ }
+ case TConfigItem::ECrashedProcessUsrStacks:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ECrashedProcessUsrStacks at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = EFalse;
+ dump.iStk = SCMDataSave::EUsrStack;
+ dump.iReg = SCMDataSave::ERegSetNone;
+ err = LogObjectContainers(EThread, SCMDataSave::EProcessSpecific, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iCTUsrStkOffset);
+
+ break;
+ }
+ case TConfigItem::EThreadsUsrStack:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EThreadsUsrStack at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = EFalse;
+ dump.iStk = SCMDataSave::EUsrStack;
+ dump.iReg = SCMDataSave::ERegSetNone;
+
+ err = LogObjectContainers(EThread, SCMDataSave::ESystemWide, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iSysSvrStkOffset);
+
+ break;
+ }
+ case TConfigItem::ECrashedProcessSvrStacks:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ECrashedProcessSvrStacks at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = EFalse;
+ dump.iStk = SCMDataSave::ESvrStack;
+ dump.iReg = SCMDataSave::ERegSetNone;
+
+ err = LogObjectContainers(EThread, SCMDataSave::EProcessSpecific, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iCTSvrStkOffset);
+
+ break;
+ }
+ case TConfigItem::EThreadsSvrStack:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EThreadsSvrStack at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = EFalse;
+ dump.iStk = SCMDataSave::ESvrStack;
+ dump.iReg = SCMDataSave::ERegSetNone;
+
+ err = LogObjectContainers(EThread, SCMDataSave::ESystemWide, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iSysSvrStkOffset);
+
+ break;
+ }
+ case TConfigItem::EThreadsUsrRegisters:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EThreadsUsrRegisters at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = EFalse;
+ dump.iStk = SCMDataSave::EStackTypeNone;
+ dump.iReg = SCMDataSave::EUserRegisters;
+
+ err = LogObjectContainers(EThread, SCMDataSave::ESystemWide, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iSysUsrRegOffset);
+
+ break;
+ }
+ case TConfigItem::EThreadsSvrRegisters:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EThreadsSvrRegisters at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = EFalse;
+ dump.iStk = SCMDataSave::EStackTypeNone;
+ dump.iReg = SCMDataSave::ESupervisorRegisters;
+
+ err = LogObjectContainers(EThread, SCMDataSave::ESystemWide, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iSysSvrRegOffset);
+
+ break;
+ }
+ case TConfigItem::EExceptionStacks:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EExceptionStacks at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = iDataSave->LogExceptionStacks(sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iExcStkOffset);
+
+ break;
+ }
+ case TConfigItem::ECrashedProcessCodeSegs:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ECrashedProcessCodeSegs at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = ETrue;
+ dump.iStk = SCMDataSave::EStackTypeNone;
+ dump.iReg = SCMDataSave::ERegSetNone;
+
+ err = LogObjectContainers(EProcess, SCMDataSave::EProcessSpecific, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iCPCodeSegOffset);
+
+ break;
+ }
+ case TConfigItem::EProcessCodeSegs:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EProcessCodeSegs at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ //define what we wish to dump
+ dump.iMetaData = EFalse;
+ dump.iCodeSegs = ETrue;
+ dump.iStk = SCMDataSave::EStackTypeNone;
+ dump.iReg = SCMDataSave::ERegSetNone;
+ err = LogObjectContainers(EProcess, SCMDataSave::ESystemWide, dump, sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iSysCodeSegOffset);
+
+ break;
+ }
+ case TConfigItem::ETraceData:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ETraceData at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = iDataSave->LogTraceBuffer(configItem->GetSizeToDump(), sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iTraceOffset);
+
+ break;
+ }
+ case TConfigItem::ELocks:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ELocks at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = iDataSave->LogLocks(sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iScmLocksOffset);
+
+ break;
+ }
+ case TConfigItem::EKernelHeap:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EKernelHeap at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = iDataSave->LogKernelHeap(sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iKernelHeapOffset);
+
+ break;
+ }
+ case TConfigItem::EVariantSpecificData:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: EVariantSpecificData at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = iDataSave->LogVariantSpecificData(sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iVarSpecInfOffset);
+
+ break;
+ }
+ case TConfigItem::ERomInfo:
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tDoing: ERomInfo at [%d] offset from [%d]", iDataSave->iWriter->GetBytesWritten(), iDataSave->GetCrashStartingPoint()));
+
+ err = iDataSave->LogRomInfo(sizeOfObjectDumped);
+ offsetPointer = &(iDataSave->iHdr.iRomInfoOffset);
+
+ break;
+ }
+ //unknown configuration type - something bad is going on
+ default: return 0;
+ }
+
+ if(KErrNone != err)
+ {
+ __KTRACE_OPT(logLevel, Kern::Printf("\tError logging data: [%d] Type = [%d]", err, aBlock.iBlockOffset));
+ continue;
+ }
+
+ //Set the space required so next time around we will know in advance how much space we need
+ configItem->SetSpaceRequired(sizeOfObjectDumped);
+
+ //Note: the following steps are only required for the first time we call process crash. The second time,
+ //when physical writing is enabled, these will have been written already and so they dont matter
+
+ //update the offset and logsize if we are going to dump this item
+ TUint32 absoluteLogPos = logSize + iDataSave->GetCrashStartingPoint();
+ if(absoluteLogPos+sizeOfObjectDumped < iDataSave->MaxLogSize())
+ {
+ //now, we must record where in the crash log this item will be dumped
+ *offsetPointer = absoluteLogPos;
+ logSize += sizeOfObjectDumped;
+ }
+ }
+
+ iDataSave->iCrashInf.iLogSize = logSize;
+ iDataSave->iWriter->FlushCache();
+
+ return iDataSave->iCrashInf.iLogSize;
+ }
+
+/**
+ * Logs the meta data for processes
+ * @param aCurrentProcess - scope to dump
+ * @return one of the OS wide codes
+ */
+TInt SCMonitor::LogProcessMetaData(SCMDataSave::TDumpScope aScope, TUint& aSizeDumped) const
+ {
+ LOG_CONTEXT
+
+ SCMDataSave::TDataToDump dump;
+ dump.iMetaData = ETrue;
+
+ return LogObjectContainers(EProcess, aScope, dump, aSizeDumped);
+ }
+
+/**
+ *
+ * @param aCurrentThread - to only do the current (crashed thread) or to do all the others
+ * @return one of the OS wide codes
+ */
+TInt SCMonitor::LogThreadMetaData(SCMDataSave::TDumpScope aScope, TUint& aSizeDumped) const
+ {
+ LOG_CONTEXT
+
+ SCMDataSave::TDataToDump dump;
+ dump.iMetaData = ETrue;
+
+ return LogObjectContainers(EThread, aScope, dump, aSizeDumped);
+ }
+
+/**
+ * Generic method that looks at all kernel objects of aObjectType
+ * @param aObjectType
+ * @param aDumpScope - if you wish to dump for the the current process, current thread or entire system
+ * @param aDataToDump - data you wish to dump
+ * @param aSizeDumped - records how much was dumped
+ * @return
+ */
+TInt SCMonitor::LogObjectContainers(TObjectType aObjectType, SCMDataSave::TDumpScope aDumpScope, const SCMDataSave::TDataToDump& aDataToDump, TUint& aSizeDumped) const
+ {
+ aSizeDumped = 0;
+
+ if(aObjectType >= ENumObjectTypes)
+ {
+ return KErrArgument;
+ }
+
+ //Get the object container for the given object type
+ DObjectCon* objectContainer = Kern::Containers()[aObjectType];
+ if(objectContainer == NULL)
+ {
+ CLTRACE("tFailed to get object container");
+ return KErrNotFound;
+ }
+
+ //Must check the mutex on this is ok otherwise the data will be in an inconsistent state
+ if(objectContainer->iMutex->iHoldCount)
+ {
+ CLTRACE("\tObject Container is in an inconsistant state");
+ return KErrCorrupt;
+ }
+
+ TInt numObjects = objectContainer->Count();
+ TInt err = KErrNone;
+
+ for(TInt cnt = 0; cnt< numObjects; cnt ++)
+ {
+ DObject* object = (*objectContainer)[cnt];
+
+ //Are we interested in the object? scope only relevant for thread and process objects, for others, the scope is implicit
+ if(aObjectType == EThread)
+ {
+ switch(aDumpScope)
+ {
+ case SCMDataSave::EThreadSpecific :
+ {
+ //if we are interested in the current thread and this is not it, continue
+ if(((DThread*)object) != &Kern::CurrentThread())
+ continue;
+ break;
+ }
+ case SCMDataSave::EProcessSpecific :
+ {
+ //if we are interested in the current proc and this is not it, continue
+ if(((DThread*)object)->iOwningProcess != &Kern::CurrentProcess())
+ continue;
+ break;
+ }
+ case SCMDataSave::ESystemWide :
+ default:
+ break;
+ }
+ }
+ else if(aObjectType == EProcess)
+ {
+ switch(aDumpScope)
+ {
+ case SCMDataSave::EProcessSpecific :
+ {
+ if((DProcess*)object != &Kern::CurrentProcess())
+ continue;
+ break;
+ }
+ case SCMDataSave::EThreadSpecific : //thread specific process doesnt make sense
+ return KErrArgument;
+ case SCMDataSave::ESystemWide :
+ default:
+ break;
+ }
+ }
+
+ //Now we look at the data we have been asked to dump
+ if(aDataToDump.iMetaData)
+ {
+ TUint dumped = 0;
+ err = HelpDumpMetaData(object, aObjectType, dumped);
+ if(KErrNone != err)
+ {
+ CLTRACE1("Failed to meta data: [%d]", err);
+ return err;
+ }
+ aSizeDumped += dumped;
+ }
+
+ if(aDataToDump.iCodeSegs)
+ {
+ if(aObjectType != EProcess)
+ {
+ return KErrArgument;
+ }
+
+ TUint dumped = 0;
+ err = iDataSave->LogCodeSegments((DProcess*)object, dumped);
+ if(KErrNone != err)
+ {
+ CLTRACE1("Failed to log code segments: [%d]", err);
+ return err;
+ }
+ aSizeDumped += dumped;
+ }
+
+ if(aDataToDump.iStk != SCMDataSave::EStackTypeNone)
+ {
+ TUint dumped = 0;
+ err = HelpDumpStacks(object, aObjectType, dumped, aDataToDump.iStk);
+ if(KErrNone != err)
+ {
+ CLTRACE1("Failed to log stacks: [%d]", err);
+ return err;
+ }
+ aSizeDumped += dumped;
+ }
+
+ if(aDataToDump.iReg != SCMDataSave::ERegSetNone)
+ {
+ if(aObjectType != EThread)
+ {
+ return KErrArgument;
+ }
+ TUint dumped = 0;
+ err = iDataSave->LogRegisters((DThread*)object, aDataToDump.iReg, dumped);
+ if(KErrNone != err && KErrNotSupported !=err) //we expect to send down a KErrNotSupported when we ask for Full CPU set for the non crashed thread - thats fine
+ {
+ CLTRACE1("Failed to log registers: [%d]", err);
+ return err;
+ }
+ aSizeDumped += dumped;
+ }
+ }
+
+ return KErrNone;
+ }
+
+/**
+ * Helper method for dumping stacks. Looks to see what type of stack we want and then calls
+ * appropriate method
+ * @param aObject The DThread object whose stack we want
+ * @param aObjectType The object type of this aObject. Anything other than EThread will give KErrArgument
+ * @param aSizeDumped Holds the size of the stack dumped after processing
+ * @param aStkType The type of stack to be dumped
+ * @see TObjectType
+ * @see SCMDataSave::TStackType
+ * @return One of the system wide codes
+ */
+TInt SCMonitor::HelpDumpStacks(DObject* aObject, TObjectType aObjectType, TUint& aSizeDumped, SCMDataSave::TStackType aStkType) const
+ {
+ //verify args
+ if(aObjectType != EThread)
+ {
+ return KErrArgument;
+ }
+
+ switch(aStkType)
+ {
+ case SCMDataSave::EUsrStack:
+ {
+ return iDataSave->LogThreadUserStack((DThread*)aObject, ETrue, aSizeDumped);
+ }
+ case SCMDataSave::ESvrStack:
+ {
+ return iDataSave->LogThreadSupervisorStack((DThread*)aObject, ETrue, aSizeDumped);
+ }
+ default: return KErrArgument;
+ }
+ }
+
+/**
+ * Helper method to dump meta data about a DThread or a DProcess object
+ * @param aObject DObject to use
+ * @param aObjectType Type of DObject. Must be EThread or EProcess
+ * @param aSizeDumped Holds the size of the stack dumped after processing
+ * @return
+ */
+TInt SCMonitor::HelpDumpMetaData(DObject* aObject, TObjectType aObjectType, TUint& aSizeDumped) const
+ {
+ aSizeDumped = 0;
+
+ switch(aObjectType)
+ {
+ case EThread:
+ {
+ return iDataSave->LogThreadData((DThread*)aObject, aSizeDumped);
+ }
+ case EProcess:
+ {
+ return iDataSave->LogProcessData((DProcess*)aObject, aSizeDumped);
+ }
+ default: return KErrArgument;
+ }
+ }
+
+/**
+ * Wrapper method around the flash erase block fundtion to determine if the erase was succesful.
+ * If the erase was not succesful we can't continue as we cannot write.
+ * @param aBlockOffset Block to erase
+ * @return One of the OS wide codes
+ */
+TInt SCMonitor::EraseFlashBlock(const SCMCrashBlockEntry& aBlock)
+ {
+ iFlash->StartTransaction();
+
+ TInt numAttempts = 0;
+ while(numAttempts < KFlashEraseAttempts)
+ {
+ iFlash->SetWritePos(aBlock.iBlockOffset);
+ iFlash->EraseFlashBlock(aBlock.iBlockOffset);
+
+ //we will read the flash to make sure that it set the block to all 1's (well not all, just the start)
+ TBuf8<sizeof(TUint32)> buf;
+ buf.SetLength(sizeof(TUint32));
+
+ iFlash->SetReadPos(aBlock.iBlockOffset);
+ iFlash->Read(buf);
+
+ volatile TUint32* result = (TUint32*)buf.Ptr();
+ if(*result == 0xFFFFFFFF)
+ {
+ __KTRACE_OPT(KALWAYS, Kern::Printf("Erase of block [0x%X] succesful after [%d] attempts", aBlock.iBlockOffset, numAttempts+1))
+ iFlash->EndTransaction();
+ return KErrNone;
+ }
+
+ numAttempts++;
+
+ //Sometimes a write to the block helps the next erase
+ TUint32 bytesWritten = 0;
+ while(bytesWritten < aBlock.iBlockSize)
+ {
+ TBuf8<sizeof(TUint8)> num;
+ num.Append(0x0);
+ iFlash->Write(num);
+ bytesWritten++;
+ }
+ }
+
+ __KTRACE_OPT(KALWAYS, Kern::Printf("After %d attempts, we were unable to erase the flash block at [0x%X]. This could be because "
+ "the driver is defective or because the flash has gone past its lifetime. Whatever it is though, "
+ "we cannot continue.", KFlashEraseAttempts, aBlock.iBlockOffset));
+
+ iFlash->EndTransaction();
+ return KErrAbort;
+ }
+
+/**
+ * This erases each block in the flash partition
+ * @return One of the system wide codes
+ */
+TInt SCMonitor::EraseEntireFlashPartition()
+ {
+ if(iMultiCrashInfo)
+ {
+ iMultiCrashInfo->Reset();
+
+ SCMCrashBlockEntry* block = iMultiCrashInfo->GetNextBlock();
+ while(block)
+ {
+ TInt err = EraseFlashBlock(*block);
+ if(KErrNone != err)
+ {
+ return err;
+ }
+
+ block = iMultiCrashInfo->GetNextBlock();
+ }
+
+ return KErrNone;
+ }
+
+ CLTRACE("SCMonitor::EraseEntireFlashPartition() -- No Flash MAP available, trying to use the raw driver to delete.");
+ TheSCMonitor.iFlash->EraseLogArea();
+
+ return KErrNone;
+ }
+
+//eof scmonitor.cpp
+