Fix for bug 2283 (RVCT 4.0 support is missing from PDK 3.0.h)
Have multiple extension sections in the bld.inf, one for each version
of the compiler. The RVCT version building the tools will build the
runtime libraries for its version, but make sure we extract all the other
versions from zip archives. Also add the archive for RVCT4.
// Copyright (c) 1995-2009 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\memmodel\epoc\multiple\mcodeseg.cpp
//
//
#include "memmodel.h"
#include <mmubase.inl>
#include "cache_maintenance.h"
#include <demand_paging.h>
DCodeSeg* M::NewCodeSeg(TCodeSegCreateInfo&)
//
// Create a new instance of this class.
//
{
__KTRACE_OPT(KDLL,Kern::Printf("M::NewCodeSeg"));
return new DMemModelCodeSeg;
}
//
// DMemModelCodeSegMemory
//
DEpocCodeSegMemory* DEpocCodeSegMemory::New(DEpocCodeSeg* aCodeSeg)
{
return new DMemModelCodeSegMemory(aCodeSeg);
}
DMemModelCodeSegMemory::DMemModelCodeSegMemory(DEpocCodeSeg* aCodeSeg)
: DMmuCodeSegMemory(aCodeSeg)
{
}
TInt DMemModelCodeSegMemory::Create(TCodeSegCreateInfo& aInfo)
{
TInt r = DMmuCodeSegMemory::Create(aInfo);
if(r!=KErrNone)
return r;
Mmu& m=Mmu::Get();
iOsAsids = TBitMapAllocator::New(m.iNumOsAsids, EFalse);
if(!iOsAsids)
return KErrNoMemory;
TInt totalPages = iPageCount+iDataPageCount;
iPages = (TPhysAddr*)Kern::Alloc(totalPages*sizeof(TPhysAddr));
if(!iPages)
return KErrNoMemory;
TInt i;
for (i=0; i<totalPages; ++i)
iPages[i] = KPhysAddrInvalid;
MmuBase::Wait();
// allocate RAM pages...
__KTRACE_OPT(KDLL,Kern::Printf("Alloc DLL pages %x,%x", iPageCount,iDataPageCount));
TInt startPage = iIsDemandPaged ? iPageCount : 0; // if demand paged, skip pages for code
TInt endPage = iPageCount+iDataPageCount;
r=m.AllocRamPages(iPages+startPage, endPage-startPage, EPageMovable);
// initialise SPageInfo objects for allocated pages...
if (r==KErrNone)
{
NKern::LockSystem();
for (i=startPage; i<endPage; ++i)
{
SPageInfo* info = SPageInfo::FromPhysAddr(iPages[i]);
info->SetCodeSegMemory(this,i);
if((i&15)==15)
NKern::FlashSystem();
}
NKern::UnlockSystem();
}
MmuBase::Signal();
if (r!=KErrNone)
return r;
#ifdef BTRACE_CODESEGS
BTrace8(BTrace::ECodeSegs,BTrace::ECodeSegMemoryAllocated,iCodeSeg,iPageCount<<m.iPageShift);
#endif
DCodeSeg::Wait();
TInt code_alloc=((totalPages<<m.iPageShift)+m.iAliasMask)>>m.iAliasShift;
r=MM::UserCodeAllocator->AllocConsecutive(code_alloc, ETrue);
if (r<0)
r = KErrNoMemory;
else
{
MM::UserCodeAllocator->Alloc(r, code_alloc);
iCodeAllocBase=r;
iRamInfo.iCodeRunAddr=m.iUserCodeBase+(r<<m.iAliasShift);
iRamInfo.iCodeLoadAddr=iRamInfo.iCodeRunAddr;
if (iRamInfo.iDataSize)
{
if(iDataPageCount)
iRamInfo.iDataLoadAddr=iRamInfo.iCodeLoadAddr+Mmu::RoundToPageSize(iRamInfo.iCodeSize);
else
iRamInfo.iDataLoadAddr=iRamInfo.iCodeLoadAddr+iRamInfo.iCodeSize;
}
DMemModelProcess* pP=(DMemModelProcess*)TheCurrentThread->iOwningProcess;
r=pP->MapUserRamCode(this, ETrue);
if (r==KErrNone)
iCreator=pP;
}
DCodeSeg::Signal();
return r;
}
void DMemModelCodeSegMemory::Substitute(TInt aOffset, TPhysAddr aOld, TPhysAddr aNew)
{
__KTRACE_OPT(KMMU,Kern::Printf("DMemModelCodeSegMemory::Substitute %x %08x %08x",aOffset,aOld,aNew));
Mmu& m=Mmu::Get();
if (iPages[aOffset>>KPageShift] != aOld)
MM::Panic(MM::ECodeSegRemapWrongPage);
iPages[aOffset>>KPageShift] = aNew;
m.RemapPageByAsid(iOsAsids, iRamInfo.iCodeRunAddr+aOffset, aOld, aNew, m.PtePermissions(EUserCode));
}
TInt DMemModelCodeSegMemory::Loaded(TCodeSegCreateInfo& aInfo)
{
__NK_ASSERT_DEBUG(iPages);
TInt r = DMmuCodeSegMemory::Loaded(aInfo);
if(r!=KErrNone)
return r;
Mmu& m=Mmu::Get();
if(!iIsDemandPaged)
{
UNLOCK_USER_MEMORY();
CacheMaintenance::CodeChanged(iRamInfo.iCodeLoadAddr, iRamInfo.iCodeSize);
LOCK_USER_MEMORY();
}
else
{
// apply code fixups to pages which have already been loaded...
TInt pageShift = m.iPageShift;
for (TInt i = 0 ; i < iPageCount ; ++i)
{
if (iPages[i] != KPhysAddrInvalid)
{
r = ApplyCodeFixupsOnLoad((TUint32*)(iRamInfo.iCodeLoadAddr+(i<<pageShift)),iRamInfo.iCodeRunAddr+(i<<pageShift));
if(r!=KErrNone)
return r;
}
}
// copy export directory (this will now have fixups applied)...
TInt exportDirSize = iRamInfo.iExportDirCount * sizeof(TLinAddr);
if (exportDirSize > 0 || (exportDirSize==0 && (iCodeSeg->iAttr&ECodeSegAttNmdExpData)) )
{
exportDirSize += sizeof(TLinAddr);
TLinAddr* expDir = (TLinAddr*)Kern::Alloc(exportDirSize);
if (!expDir)
return KErrNoMemory;
iCopyOfExportDir = expDir;
UNLOCK_USER_MEMORY();
memcpy(expDir,(TAny*)(iRamInfo.iExportDir-sizeof(TLinAddr)),exportDirSize);
LOCK_USER_MEMORY();
}
}
// unmap code from loading process...
DMemModelProcess* pP=(DMemModelProcess*)TheCurrentThread->iOwningProcess;
__ASSERT_ALWAYS(iCreator==pP, MM::Panic(MM::ECodeSegLoadedNotCreator));
pP->UnmapUserRamCode(this, ETrue);
iCreator=NULL;
// discard any temporary pages used to store loaded data section...
if(iDataPageCount)
{
MmuBase::Wait();
TPhysAddr* pages = iPages+iPageCount;
m.FreePages(pages,iDataPageCount, EPageMovable);
for (TInt i = 0 ; i < iDataPageCount ; ++i)
pages[i] = KPhysAddrInvalid;
MmuBase::Signal();
// see if we can free any virtual address space now we don't need any for loading data
TInt data_start = ((iPageCount << m.iPageShift) + m.iAliasMask) >> m.iAliasShift;
TInt data_end = (((iPageCount + iDataPageCount) << m.iPageShift) + m.iAliasMask) >> m.iAliasShift;
if (data_end != data_start)
{
DCodeSeg::Wait();
MM::UserCodeAllocator->Free(iCodeAllocBase + data_start, data_end - data_start);
DCodeSeg::Signal();
}
iDataPageCount = 0;
//Reduce the size of the DCodeSeg now the data section has been moved
iCodeSeg->iSize = iPageCount << m.iPageShift;
}
return KErrNone;
}
void DMemModelCodeSegMemory::Destroy()
{
if(iCreator)
iCreator->UnmapUserRamCode(this, ETrue); // remove from creating process if not fully loaded
}
DMemModelCodeSegMemory::~DMemModelCodeSegMemory()
{
__KTRACE_OPT(KDLL,Kern::Printf("DMemModelCodeSegMemory::~DMemModelCodeSegMemory %x", this));
__NK_ASSERT_DEBUG(iAccessCount==0);
__NK_ASSERT_DEBUG(iOsAsids==0 || iOsAsids->Avail()==0); // check not mapped (inverted logic!)
Mmu& m=Mmu::Get();
if(iCodeAllocBase>=0)
{
// free allocated virtual memory space...
TInt size = (iPageCount+iDataPageCount)<<KPageShift;
TInt code_alloc=(size+m.iAliasMask)>>m.iAliasShift;
DCodeSeg::Wait();
MM::UserCodeAllocator->Free(iCodeAllocBase, code_alloc);
DCodeSeg::Signal();
}
if(iPages)
{
#ifdef __DEMAND_PAGING__
if (iIsDemandPaged)
{
// Return any paged memory to the paging system
MmuBase::Wait();
NKern::LockSystem();
DemandPaging& p = *DemandPaging::ThePager;
for (TInt i = 0 ; i < iPageCount ; ++i)
{
if (iPages[i] != KPhysAddrInvalid)
p.NotifyPageFree(iPages[i]);
}
NKern::UnlockSystem();
MmuBase::Signal();
Kern::Free(iCopyOfExportDir);
iCopyOfExportDir = NULL;
}
#endif
MmuBase::Wait();
m.FreePages(iPages,iPageCount+iDataPageCount, EPageMovable);
MmuBase::Signal();
Kern::Free(iPages);
iPages = NULL;
#ifdef BTRACE_CODESEGS
BTrace8(BTrace::ECodeSegs,BTrace::ECodeSegMemoryDeallocated,this,iPageCount<<m.iPageShift);
#endif
}
delete iOsAsids;
}
DMemModelCodeSeg::DMemModelCodeSeg()
//
// Constructor
//
: iCodeAllocBase(-1),
iDataAllocBase(-1)
{
}
DMemModelCodeSeg::~DMemModelCodeSeg()
//
// Destructor
//
{
__KTRACE_OPT(KDLL,Kern::Printf("DMemModelCodeSeg::Destruct %C", this));
Mmu& m=Mmu::Get();
DCodeSeg::Wait();
if (iCodeAllocBase>=0)
{
TBool kernel=( (iAttr&(ECodeSegAttKernel|ECodeSegAttGlobal)) == ECodeSegAttKernel );
TBool global=( (iAttr&(ECodeSegAttKernel|ECodeSegAttGlobal)) == ECodeSegAttGlobal );
TInt r=KErrNone;
if (kernel)
{
DMemModelProcess& kproc=*(DMemModelProcess*)K::TheKernelProcess;
r=kproc.iCodeChunk->Decommit(iCodeAllocBase, iSize);
}
else if (global)
{
r=m.iGlobalCode->Decommit(iCodeAllocBase, iSize);
}
__ASSERT_DEBUG(r==KErrNone, MM::Panic(MM::EDecommitFailed));
r=r; // stop compiler warning
}
if(Memory())
Memory()->Destroy();
if (iDataAllocBase>=0 && !iXIP)
{
SRamCodeInfo& ri=RamInfo();
TInt data_alloc=(ri.iDataSize+ri.iBssSize+m.iPageMask)>>m.iPageShift;
MM::DllDataAllocator->Free(iDataAllocBase, data_alloc);
}
DCodeSeg::Signal();
Kern::Free(iKernelData);
DEpocCodeSeg::Destruct();
}
TInt DMemModelCodeSeg::DoCreateRam(TCodeSegCreateInfo& aInfo, DProcess*)
{
__KTRACE_OPT(KDLL,Kern::Printf("DMemModelCodeSeg::DoCreateRam %C", this));
TBool kernel=( (iAttr&(ECodeSegAttKernel|ECodeSegAttGlobal)) == ECodeSegAttKernel );
TBool global=( (iAttr&(ECodeSegAttKernel|ECodeSegAttGlobal)) == ECodeSegAttGlobal );
Mmu& m=Mmu::Get();
SRamCodeInfo& ri=RamInfo();
iSize = Mmu::RoundToPageSize(ri.iCodeSize+ri.iDataSize);
if (iSize==0)
return KErrCorrupt;
TInt total_data_size=ri.iDataSize+ri.iBssSize;
TInt r=KErrNone;
if (kernel)
{
DMemModelProcess& kproc=*(DMemModelProcess*)K::TheKernelProcess;
if (!kproc.iCodeChunk)
r=kproc.CreateCodeChunk();
if (r!=KErrNone)
return r;
r=kproc.iCodeChunk->Allocate(iSize, 0, m.iAliasShift);
if (r<0)
return r;
iCodeAllocBase=r;
ri.iCodeRunAddr=(TUint32)kproc.iCodeChunk->Base();
ri.iCodeRunAddr+=r;
ri.iCodeLoadAddr=ri.iCodeRunAddr;
if (ri.iDataSize)
ri.iDataLoadAddr=ri.iCodeLoadAddr+ri.iCodeSize;
if (total_data_size)
{
iKernelData=Kern::Alloc(total_data_size);
if (!iKernelData)
return KErrNoMemory;
ri.iDataRunAddr=(TLinAddr)iKernelData;
}
return KErrNone;
}
if (global)
{
if (!m.iGlobalCode)
r=m.CreateGlobalCodeChunk();
if (r==KErrNone)
r=m.iGlobalCode->Allocate(iSize, 0, m.iAliasShift);
if (r<0)
return r;
iCodeAllocBase=r;
ri.iCodeRunAddr=(TUint32)m.iGlobalCode->Base();
ri.iCodeRunAddr+=r;
ri.iCodeLoadAddr=ri.iCodeRunAddr;
ri.iDataLoadAddr=0; // we don't allow static data in global code
ri.iDataRunAddr=0;
TInt loadSize = ri.iCodeSize+ri.iDataSize;
memset((TAny*)(ri.iCodeRunAddr+loadSize), 0x03, iSize-loadSize);
return KErrNone;
}
DCodeSeg::Wait();
if (total_data_size && !IsExe())
{
TInt data_alloc=(total_data_size+m.iPageMask)>>m.iPageShift;
__KTRACE_OPT(KDLL,Kern::Printf("Alloc DLL data %x", data_alloc));
r=MM::DllDataAllocator->AllocConsecutive(data_alloc, ETrue);
if (r<0)
r = KErrNoMemory;
else
{
MM::DllDataAllocator->Alloc(r, data_alloc);
iDataAllocBase=r;
ri.iDataRunAddr=m.iDllDataBase+m.iMaxDllDataSize-((r+data_alloc)<<m.iPageShift);
r = KErrNone;
}
}
DCodeSeg::Signal();
if(r==KErrNone)
r = Memory()->Create(aInfo);
return r;
}
TInt DMemModelCodeSeg::DoCreateXIP(DProcess* aProcess)
{
// __KTRACE_OPT(KDLL,Kern::Printf("DMemModelCodeSeg::DoCreateXIP %C proc %O", this, aProcess));
return KErrNone;
}
TInt DMemModelCodeSeg::Loaded(TCodeSegCreateInfo& aInfo)
{
if(iXIP)
return DEpocCodeSeg::Loaded(aInfo);
TBool kernel=( (iAttr&(ECodeSegAttKernel|ECodeSegAttGlobal)) == ECodeSegAttKernel );
TBool global=( (iAttr&(ECodeSegAttKernel|ECodeSegAttGlobal)) == ECodeSegAttGlobal );
if (Pages())
{
TInt r = Memory()->Loaded(aInfo);
if(r!=KErrNone)
return r;
}
else if (kernel && iExeCodeSeg!=this)
{
Mmu& m=Mmu::Get();
DMemModelProcess& kproc=*(DMemModelProcess*)K::TheKernelProcess;
SRamCodeInfo& ri=RamInfo();
// NOTE: Must do IMB before changing permissions since ARMv6 is very pedantic and
// doesn't let you clean a cache line which is marked as read only.
CacheMaintenance::CodeChanged(ri.iCodeRunAddr, ri.iCodeSize);
TInt offset=ri.iCodeRunAddr-TLinAddr(kproc.iCodeChunk->iBase);
kproc.iCodeChunk->ApplyPermissions(offset, iSize, m.iKernelCodePtePerm);
}
else if (global)
{
Mmu& m=Mmu::Get();
SRamCodeInfo& ri=RamInfo();
CacheMaintenance::CodeChanged(ri.iCodeRunAddr, ri.iCodeSize);
TInt offset=ri.iCodeRunAddr-TLinAddr(m.iGlobalCode->iBase);
m.iGlobalCode->ApplyPermissions(offset, iSize, m.iGlobalCodePtePerm);
}
return DEpocCodeSeg::Loaded(aInfo);
}
void DMemModelCodeSeg::ReadExportDir(TUint32* aDest)
{
__KTRACE_OPT(KDLL,Kern::Printf("DMemModelCodeSeg::ReadExportDir %C %08x",this, aDest));
if (!iXIP)
{
SRamCodeInfo& ri=RamInfo();
TInt size=(ri.iExportDirCount+1)*sizeof(TLinAddr);
if (Memory()->iCopyOfExportDir)
{
kumemput(aDest, Memory()->iCopyOfExportDir, size);
return;
}
NKern::ThreadEnterCS();
Mmu& m=Mmu::Get();
TLinAddr src=ri.iExportDir-sizeof(TLinAddr);
MmuBase::Wait();
TInt offset=src-ri.iCodeRunAddr;
TPhysAddr* physArray = Pages();
TPhysAddr* ppa=physArray+(offset>>m.iPageShift);
while(size)
{
TInt pageOffset = src&m.iPageMask;
TInt l=Min(size, m.iPageSize-pageOffset);
TLinAddr alias_src = m.MapTemp(*ppa++,src-pageOffset)+pageOffset;
// Note, the following memory access isn't XTRAP'ed, because...
// a) This function is only called by the loader thread, so even if
// exceptions were trapped the system is doomed anyway
// b) Any exception will cause the crash debugger/logger to be called
// which will provide more information than if trapped exceptions
// and returned an error code.
kumemput32(aDest, (const TAny*)alias_src, l);
m.UnmapTemp();
size-=l;
src+=l;
aDest+=l/sizeof(TUint32);
}
MmuBase::Signal();
NKern::ThreadLeaveCS();
}
}
TBool DMemModelCodeSeg::OpenCheck(DProcess* aProcess)
{
return FindCheck(aProcess);
}
TBool DMemModelCodeSeg::FindCheck(DProcess* aProcess)
{
__KTRACE_OPT(KDLL,Kern::Printf("CSEG:%08x Compat? proc=%O",this,aProcess));
if (aProcess)
{
DMemModelProcess& p=*(DMemModelProcess*)aProcess;
DCodeSeg* pPSeg=p.CodeSeg();
if (iAttachProcess && iAttachProcess!=aProcess)
return EFalse;
if (iExeCodeSeg && iExeCodeSeg!=pPSeg)
return EFalse;
}
return ETrue;
}
void DMemModelCodeSeg::BTracePrime(TInt aCategory)
{
#ifdef BTRACE_CODESEGS
if (aCategory == BTrace::ECodeSegs || aCategory == -1)
{
DCodeSeg::BTracePrime(aCategory);
DMemModelCodeSegMemory* codeSegMemory = Memory();
if(codeSegMemory && codeSegMemory->iPages && codeSegMemory->iPageCount)
{
BTrace8(BTrace::ECodeSegs,BTrace::ECodeSegMemoryAllocated,this,codeSegMemory->iPageCount<<Mmu::Get().iPageShift);
}
}
#endif
}