summaryrefslogtreecommitdiffstats
path: root/sot/source/sdstor/stgstrms.cxx
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-15 05:54:39 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-15 05:54:39 +0000
commit267c6f2ac71f92999e969232431ba04678e7437e (patch)
tree358c9467650e1d0a1d7227a21dac2e3d08b622b2 /sot/source/sdstor/stgstrms.cxx
parentInitial commit. (diff)
downloadlibreoffice-267c6f2ac71f92999e969232431ba04678e7437e.tar.xz
libreoffice-267c6f2ac71f92999e969232431ba04678e7437e.zip
Adding upstream version 4:24.2.0.upstream/4%24.2.0
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'sot/source/sdstor/stgstrms.cxx')
-rw-r--r--sot/source/sdstor/stgstrms.cxx1356
1 files changed, 1356 insertions, 0 deletions
diff --git a/sot/source/sdstor/stgstrms.cxx b/sot/source/sdstor/stgstrms.cxx
new file mode 100644
index 0000000000..61682ead8a
--- /dev/null
+++ b/sot/source/sdstor/stgstrms.cxx
@@ -0,0 +1,1356 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
+/*
+ * This file is part of the LibreOffice project.
+ *
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/.
+ *
+ * This file incorporates work covered by the following license notice:
+ *
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed
+ * with this work for additional information regarding copyright
+ * ownership. The ASF licenses this file to you under the Apache
+ * License, Version 2.0 (the "License"); you may not use this file
+ * except in compliance with the License. You may obtain a copy of
+ * the License at http://www.apache.org/licenses/LICENSE-2.0 .
+ */
+
+#include <algorithm>
+
+#include <string.h>
+#include <sal/log.hxx>
+#include <o3tl/safeint.hxx>
+#include <osl/file.hxx>
+#include <unotools/tempfile.hxx>
+
+#include "stgelem.hxx"
+#include "stgcache.hxx"
+#include "stgstrms.hxx"
+#include "stgdir.hxx"
+#include "stgio.hxx"
+#include <memory>
+
+///////////////////////////// class StgFAT
+
+// The FAT class performs FAT operations on an underlying storage stream.
+// This stream is either the master FAT stream (m == true ) or a normal
+// storage stream, which then holds the FAT for small data allocations.
+
+StgFAT::StgFAT( StgStrm& r, bool m ) : m_rStrm( r )
+{
+ m_bPhys = m;
+ m_nPageSize = m_rStrm.GetIo().GetPhysPageSize();
+ m_nEntries = m_nPageSize >> 2;
+ m_nOffset = 0;
+ m_nMaxPage = 0;
+ m_nLimit = 0;
+}
+
+// Retrieve the physical page for a given byte offset.
+
+rtl::Reference< StgPage > StgFAT::GetPhysPage( sal_Int32 nByteOff )
+{
+ rtl::Reference< StgPage > pPg;
+ // Position within the underlying stream
+ // use the Pos2Page() method of the stream
+ if( m_rStrm.Pos2Page( nByteOff ) )
+ {
+ m_nOffset = m_rStrm.GetOffset();
+ sal_Int32 nPhysPage = m_rStrm.GetPage();
+ // get the physical page (must be present)
+ pPg = m_rStrm.GetIo().Get( nPhysPage, true );
+ }
+ return pPg;
+}
+
+// Get the follow page for a certain FAT page.
+
+sal_Int32 StgFAT::GetNextPage( sal_Int32 nPg )
+{
+ if (nPg >= 0)
+ {
+ if (nPg > (SAL_MAX_INT32 >> 2))
+ return STG_EOF;
+ rtl::Reference< StgPage > pPg = GetPhysPage( nPg << 2 );
+ nPg = pPg.is() ? StgCache::GetFromPage( pPg, m_nOffset >> 2 ) : STG_EOF;
+ }
+ return nPg;
+}
+
+// Find the best fit block for the given size. Return
+// the starting block and its size or STG_EOF and 0.
+// nLastPage is a stopper which tells the current
+// underlying stream size. It is treated as a recommendation
+// to abort the search to inhibit excessive file growth.
+
+sal_Int32 StgFAT::FindBlock( sal_Int32& nPgs )
+{
+ sal_Int32 nMinStart = STG_EOF, nMinLen = 0;
+ sal_Int32 nMaxStart = STG_EOF, nMaxLen = 0x7FFFFFFFL;
+ sal_Int32 nTmpStart = STG_EOF, nTmpLen = 0;
+ sal_Int32 nPages = m_rStrm.GetSize() >> 2;
+ bool bFound = false;
+ rtl::Reference< StgPage > pPg;
+ short nEntry = 0;
+ for( sal_Int32 i = 0; i < nPages; i++, nEntry++ )
+ {
+ if( !( nEntry % m_nEntries ) )
+ {
+ // load the next page for that stream
+ nEntry = 0;
+ pPg = GetPhysPage( i << 2 );
+ if( !pPg.is() )
+ return STG_EOF;
+ }
+ sal_Int32 nCur = StgCache::GetFromPage( pPg, nEntry );
+ if( nCur == STG_FREE )
+ {
+ // count the size of this area
+ if( nTmpLen )
+ nTmpLen++;
+ else
+ {
+ nTmpStart = i;
+ nTmpLen = 1;
+ }
+ if( nTmpLen == nPgs
+ // If we already did find a block, stop when reaching the limit
+ || ( bFound && ( nEntry >= m_nLimit ) ) )
+ break;
+ }
+ else if( nTmpLen )
+ {
+ if( nTmpLen > nPgs && nTmpLen < nMaxLen )
+ {
+ // block > requested size
+ nMaxLen = nTmpLen;
+ nMaxStart = nTmpStart;
+ bFound = true;
+ }
+ else if( nTmpLen >= nMinLen )
+ {
+ // block < requested size
+ nMinLen = nTmpLen;
+ nMinStart = nTmpStart;
+ bFound = true;
+ if( nTmpLen == nPgs )
+ break;
+ }
+ nTmpStart = STG_EOF;
+ nTmpLen = 0;
+ }
+ }
+ // Determine which block to use.
+ if( nTmpLen )
+ {
+ if( nTmpLen > nPgs && nTmpLen < nMaxLen )
+ {
+ // block > requested size
+ nMaxLen = nTmpLen;
+ nMaxStart = nTmpStart;
+ }
+ else if( nTmpLen >= nMinLen )
+ {
+ // block < requested size
+ nMinLen = nTmpLen;
+ nMinStart = nTmpStart;
+ }
+ }
+ if( nMinStart != STG_EOF && nMaxStart != STG_EOF )
+ {
+ // two areas found; return the best fit area
+ sal_Int32 nMinDiff = nPgs - nMinLen;
+ sal_Int32 nMaxDiff = nMaxLen - nPgs;
+ if( nMinDiff > nMaxDiff )
+ nMinStart = STG_EOF;
+ }
+ if( nMinStart != STG_EOF )
+ {
+ nPgs = nMinLen; return nMinStart;
+ }
+ else
+ {
+ return nMaxStart;
+ }
+}
+
+// Set up the consecutive chain for a given block.
+
+bool StgFAT::MakeChain( sal_Int32 nStart, sal_Int32 nPgs )
+{
+ sal_Int32 nPos = nStart << 2;
+ rtl::Reference< StgPage > pPg = GetPhysPage( nPos );
+ if( !pPg.is() || !nPgs )
+ return false;
+ while( --nPgs )
+ {
+ if( m_nOffset >= m_nPageSize )
+ {
+ pPg = GetPhysPage( nPos );
+ if( !pPg.is() )
+ return false;
+ }
+ m_rStrm.GetIo().SetToPage( pPg, m_nOffset >> 2, ++nStart );
+ m_nOffset += 4;
+ nPos += 4;
+ }
+ if( m_nOffset >= m_nPageSize )
+ {
+ pPg = GetPhysPage( nPos );
+ if( !pPg.is() )
+ return false;
+ }
+ m_rStrm.GetIo().SetToPage( pPg, m_nOffset >> 2, STG_EOF );
+ return true;
+}
+
+// Allocate a block of data from the given page number on.
+// It the page number is != STG_EOF, chain the block.
+
+sal_Int32 StgFAT::AllocPages( sal_Int32 nBgn, sal_Int32 nPgs )
+{
+ sal_Int32 nOrig = nBgn;
+ sal_Int32 nLast = nBgn;
+ sal_Int32 nBegin = STG_EOF;
+ sal_Int32 nAlloc;
+ sal_Int32 nPages = m_rStrm.GetSize() >> 2;
+ short nPasses = 0;
+ // allow for two passes
+ while( nPasses < 2 )
+ {
+ // try to satisfy the request from the pool of free pages
+ while( nPgs )
+ {
+ nAlloc = nPgs;
+ nBegin = FindBlock( nAlloc );
+ // no more blocks left in present alloc chain
+ if( nBegin == STG_EOF )
+ break;
+ if( ( nBegin + nAlloc ) > m_nMaxPage )
+ m_nMaxPage = nBegin + nAlloc;
+ if( !MakeChain( nBegin, nAlloc ) )
+ return STG_EOF;
+ if( nOrig == STG_EOF )
+ nOrig = nBegin;
+ else
+ {
+ // Patch the chain
+ rtl::Reference< StgPage > pPg = GetPhysPage( nLast << 2 );
+ if( !pPg.is() )
+ return STG_EOF;
+ m_rStrm.GetIo().SetToPage( pPg, m_nOffset >> 2, nBegin );
+ }
+ nLast = nBegin + nAlloc - 1;
+ nPgs -= nAlloc;
+ }
+ if( nPgs && !nPasses )
+ {
+ // we need new, fresh space, so allocate and retry
+ if( !m_rStrm.SetSize( ( nPages + nPgs ) << 2 ) )
+ return STG_EOF;
+ if( !m_bPhys && !InitNew( nPages ) )
+ return 0;
+ // FIXME: this was originally "FALSE", whether or not that
+ // makes sense (or should be STG_EOF instead, say?)
+ nPages = m_rStrm.GetSize() >> 2;
+ nPasses++;
+ }
+ else
+ break;
+ }
+ // now we should have a chain for the complete block
+ if( nBegin == STG_EOF || nPgs )
+ {
+ m_rStrm.GetIo().SetError( SVSTREAM_FILEFORMAT_ERROR );
+ return STG_EOF; // bad structure
+ }
+ return nOrig;
+}
+
+// Initialize newly allocated pages for a standard FAT stream
+// It can be assumed that the stream size is always on
+// a page boundary
+
+bool StgFAT::InitNew( sal_Int32 nPage1 )
+{
+ sal_Int32 n = ( ( m_rStrm.GetSize() >> 2 ) - nPage1 ) / m_nEntries;
+ if ( n > 0 )
+ {
+ while( n-- )
+ {
+ rtl::Reference< StgPage > pPg;
+ // Position within the underlying stream
+ // use the Pos2Page() method of the stream
+ m_rStrm.Pos2Page( nPage1 << 2 );
+ // Initialize the page
+ pPg = m_rStrm.GetIo().Copy( m_rStrm.GetPage() );
+ if ( !pPg.is() )
+ return false;
+ for( short i = 0; i < m_nEntries; i++ )
+ m_rStrm.GetIo().SetToPage( pPg, i, STG_FREE );
+ nPage1++;
+ }
+ }
+ return true;
+}
+
+// Release a chain
+
+bool StgFAT::FreePages( sal_Int32 nStart, bool bAll )
+{
+ while( nStart >= 0 )
+ {
+ rtl::Reference< StgPage > pPg = GetPhysPage( nStart << 2 );
+ if( !pPg.is() )
+ return false;
+ nStart = StgCache::GetFromPage( pPg, m_nOffset >> 2 );
+ // The first released page is either set to EOF or FREE
+ m_rStrm.GetIo().SetToPage( pPg, m_nOffset >> 2, bAll ? STG_FREE : STG_EOF );
+ bAll = true;
+ }
+ return true;
+}
+
+///////////////////////////// class StgStrm
+
+// The base stream class provides basic functionality for seeking
+// and accessing the data on a physical basis. It uses the built-in
+// FAT class for the page allocations.
+
+StgStrm::StgStrm( StgIo& r )
+ : m_nPos(0),
+ m_bBytePosValid(true),
+ m_rIo(r),
+ m_pEntry(nullptr),
+ m_nStart(STG_EOF),
+ m_nSize(0),
+ m_nPage(STG_EOF),
+ m_nOffset(0),
+ m_nPageSize(m_rIo.GetPhysPageSize())
+{
+}
+
+StgStrm::~StgStrm()
+{
+}
+
+// Attach the stream to the given entry.
+
+void StgStrm::SetEntry( StgDirEntry& r )
+{
+ r.m_aEntry.SetLeaf( STG_DATA, m_nStart );
+ r.m_aEntry.SetSize( m_nSize );
+ m_pEntry = &r;
+ r.SetDirty();
+}
+
+/*
+ * The page chain, is basically a singly linked list of slots each
+ * point to the next page. Instead of traversing the file structure
+ * for this each time build a simple flat in-memory vector list
+ * of pages.
+ */
+sal_Int32 StgStrm::scanBuildPageChainCache()
+{
+ if (m_nSize > 0)
+ {
+ m_aPagesCache.reserve(m_nSize/m_nPageSize);
+ m_aUsedPageNumbers.reserve(m_nSize/m_nPageSize);
+ }
+
+ bool bError = false;
+ sal_Int32 nBgn = m_nStart;
+ sal_Int32 nOptSize = 0;
+
+ // Track already scanned PageNumbers here and use them to
+ // see if an already counted page is re-visited
+ while( nBgn >= 0 && !bError )
+ {
+ m_aPagesCache.push_back(nBgn);
+ nBgn = m_pFat->GetNextPage( nBgn );
+
+ //returned second is false if it already exists
+ if (!m_aUsedPageNumbers.insert(nBgn).second)
+ {
+ SAL_WARN ("sot", "Error: page number " << nBgn << " already in chain for stream");
+ bError = true;
+ }
+
+ nOptSize += m_nPageSize;
+ }
+ if (bError)
+ {
+ SAL_WARN("sot", "returning wrong format error");
+ m_rIo.SetError( ERRCODE_IO_WRONGFORMAT );
+ m_aPagesCache.clear();
+ m_aUsedPageNumbers.clear();
+ }
+ return nOptSize;
+}
+
+// Compute page number and offset for the given byte position.
+// If the position is behind the size, set the stream right
+// behind the EOF.
+bool StgStrm::Pos2Page( sal_Int32 nBytePos )
+{
+ if ( !m_pFat )
+ return false;
+
+ // Values < 0 seek to the end
+ if( nBytePos < 0 || nBytePos >= m_nSize )
+ nBytePos = m_nSize;
+ // Adjust the position back to offset 0
+ m_nPos -= m_nOffset;
+ sal_Int32 nMask = ~( m_nPageSize - 1 );
+ sal_Int32 nOld = m_nPos & nMask;
+ sal_Int32 nNew = nBytePos & nMask;
+ m_nOffset = static_cast<short>( nBytePos & ~nMask );
+ m_nPos = nBytePos;
+ if (nOld == nNew)
+ return m_bBytePosValid;
+
+ // See fdo#47644 for a .doc with a vast amount of pages where seeking around the
+ // document takes a colossal amount of time
+
+ // Please Note: we build the pagescache incrementally as we go if necessary,
+ // so that a corrupted FAT doesn't poison the stream state for earlier reads
+ size_t nIdx = nNew / m_nPageSize;
+ if( nIdx >= m_aPagesCache.size() )
+ {
+ // Extend the FAT cache ! ...
+ size_t nToAdd = nIdx + 1;
+
+ if (m_aPagesCache.empty())
+ {
+ m_aPagesCache.push_back( m_nStart );
+ assert(m_aUsedPageNumbers.empty());
+ m_aUsedPageNumbers.insert(m_nStart);
+ }
+
+ nToAdd -= m_aPagesCache.size();
+
+ sal_Int32 nBgn = m_aPagesCache.back();
+
+ // Start adding pages while we can
+ while (nToAdd > 0 && nBgn >= 0)
+ {
+ sal_Int32 nOldBgn = nBgn;
+ nBgn = m_pFat->GetNextPage(nOldBgn);
+ if( nBgn >= 0 )
+ {
+ //returned second is false if it already exists
+ if (!m_aUsedPageNumbers.insert(nBgn).second)
+ {
+ SAL_WARN ("sot", "Error: page number " << nBgn << " already in chain for stream");
+ break;
+ }
+
+ //very much the normal case
+ m_aPagesCache.push_back(nBgn);
+ --nToAdd;
+ }
+ }
+ }
+
+ if ( nIdx > m_aPagesCache.size() )
+ {
+ SAL_WARN("sot", "seek to index " << nIdx <<
+ " beyond page cache size " << m_aPagesCache.size());
+ // fdo#84229 - handle seek to end and back as eg. XclImpStream expects
+ m_nPage = STG_EOF;
+ m_nOffset = 0;
+ // Intriguingly in the past we didn't reset nPos to match the real
+ // length of the stream thus:
+ // nIdx = m_aPagesCache.size();
+ // nPos = nPageSize * nIdx;
+ // so retain this behavior for now.
+ m_bBytePosValid = false;
+ return false;
+ }
+
+ // special case: seek to 1st byte of new, unallocated page
+ // (in case the file size is a multiple of the page size)
+ if( nBytePos == m_nSize && !m_nOffset && nIdx > 0 && nIdx == m_aPagesCache.size() )
+ {
+ nIdx--;
+ m_nOffset = m_nPageSize;
+ }
+ else if ( nIdx == m_aPagesCache.size() )
+ {
+ m_nPage = STG_EOF;
+ m_bBytePosValid = false;
+ return false;
+ }
+
+ m_nPage = m_aPagesCache[ nIdx ];
+
+ m_bBytePosValid = m_nPage >= 0;
+ return m_bBytePosValid;
+}
+
+// Copy an entire stream. Both streams are allocated in the FAT.
+// The target stream is this stream.
+
+bool StgStrm::Copy( sal_Int32 nFrom, sal_Int32 nBytes )
+{
+ if ( !m_pFat )
+ return false;
+
+ m_aPagesCache.clear();
+ m_aUsedPageNumbers.clear();
+
+ sal_Int32 nTo = m_nStart;
+ sal_Int32 nPgs = ( nBytes + m_nPageSize - 1 ) / m_nPageSize;
+ while( nPgs-- )
+ {
+ if( nTo < 0 )
+ {
+ m_rIo.SetError( SVSTREAM_FILEFORMAT_ERROR );
+ return false;
+ }
+ m_rIo.Copy( nTo, nFrom );
+ if( nFrom >= 0 )
+ {
+ nFrom = m_pFat->GetNextPage( nFrom );
+ if( nFrom < 0 )
+ {
+ m_rIo.SetError( SVSTREAM_FILEFORMAT_ERROR );
+ return false;
+ }
+ }
+ nTo = m_pFat->GetNextPage( nTo );
+ }
+ return true;
+}
+
+bool StgStrm::SetSize( sal_Int32 nBytes )
+{
+ if ( nBytes < 0 || !m_pFat )
+ return false;
+
+ m_aPagesCache.clear();
+ m_aUsedPageNumbers.clear();
+
+ // round up to page size
+ sal_Int32 nOld = ( ( m_nSize + m_nPageSize - 1 ) / m_nPageSize ) * m_nPageSize;
+ sal_Int32 nNew = ( ( nBytes + m_nPageSize - 1 ) / m_nPageSize ) * m_nPageSize;
+ if( nNew > nOld )
+ {
+ if( !Pos2Page( m_nSize ) )
+ return false;
+ sal_Int32 nBgn = m_pFat->AllocPages( m_nPage, ( nNew - nOld ) / m_nPageSize );
+ if( nBgn == STG_EOF )
+ return false;
+ if( m_nStart == STG_EOF )
+ m_nStart = m_nPage = nBgn;
+ }
+ else if( nNew < nOld )
+ {
+ bool bAll = ( nBytes == 0 );
+ if( !Pos2Page( nBytes ) || !m_pFat->FreePages( m_nPage, bAll ) )
+ return false;
+ if( bAll )
+ m_nStart = m_nPage = STG_EOF;
+ }
+ if( m_pEntry )
+ {
+ // change the dir entry?
+ if( !m_nSize || !nBytes )
+ m_pEntry->m_aEntry.SetLeaf( STG_DATA, m_nStart );
+ m_pEntry->m_aEntry.SetSize( nBytes );
+ m_pEntry->SetDirty();
+ }
+ m_nSize = nBytes;
+ m_pFat->SetLimit( GetPages() );
+ return true;
+}
+
+// Return the # of allocated pages
+
+
+//////////////////////////// class StgFATStrm
+
+// The FAT stream class provides physical access to the master FAT.
+// Since this access is implemented as a StgStrm, we can use the
+// FAT allocator.
+
+StgFATStrm::StgFATStrm(StgIo& r, sal_Int32 nFatStrmSize) : StgStrm( r )
+{
+ m_pFat.reset( new StgFAT( *this, true ) );
+ m_nSize = nFatStrmSize;
+}
+
+bool StgFATStrm::Pos2Page( sal_Int32 nBytePos )
+{
+ // Values < 0 seek to the end
+ if( nBytePos < 0 || nBytePos >= m_nSize )
+ nBytePos = m_nSize ? m_nSize - 1 : 0;
+ m_nPage = nBytePos / m_nPageSize;
+ m_nOffset = static_cast<short>( nBytePos % m_nPageSize );
+ m_nPage = GetPage(m_nPage, false);
+ bool bValid = m_nPage >= 0;
+ SetPos(nBytePos, bValid);
+ return bValid;
+}
+
+// Get the page number entry for the given page offset.
+
+sal_Int32 StgFATStrm::GetPage(sal_Int32 nOff, bool bMake, sal_uInt16 *pnMasterAlloc)
+{
+ OSL_ENSURE( nOff >= 0, "The offset may not be negative!" );
+ if( pnMasterAlloc ) *pnMasterAlloc = 0;
+ if( nOff < StgHeader::GetFAT1Size() )
+ return m_rIo.m_aHdr.GetFATPage( nOff );
+ sal_Int32 nMaxPage = m_nSize >> 2;
+ nOff = nOff - StgHeader::GetFAT1Size();
+ // number of master pages that we need to iterate through
+ sal_uInt16 nMasterCount = ( m_nPageSize >> 2 ) - 1;
+ sal_uInt16 nBlocks = nOff / nMasterCount;
+ // offset in the last master page
+ nOff = nOff % nMasterCount;
+
+ rtl::Reference< StgPage > pOldPage;
+ rtl::Reference< StgPage > pMaster;
+ sal_Int32 nFAT = m_rIo.m_aHdr.GetFATChain();
+ for( sal_uInt16 nCount = 0; nCount <= nBlocks; nCount++ )
+ {
+ if( nFAT == STG_EOF || nFAT == STG_FREE )
+ {
+ if( bMake )
+ {
+ m_aPagesCache.clear();
+ m_aUsedPageNumbers.clear();
+
+ // create a new master page
+ nFAT = nMaxPage++;
+ pMaster = m_rIo.Copy( nFAT );
+ if ( pMaster.is() )
+ {
+ for( short k = 0; k < static_cast<short>( m_nPageSize >> 2 ); k++ )
+ m_rIo.SetToPage( pMaster, k, STG_FREE );
+ // chaining
+ if( !pOldPage.is() )
+ m_rIo.m_aHdr.SetFATChain( nFAT );
+ else
+ m_rIo.SetToPage( pOldPage, nMasterCount, nFAT );
+ if( nMaxPage >= m_rIo.GetPhysPages() )
+ if( !m_rIo.SetSize( nMaxPage ) )
+ return STG_EOF;
+ // mark the page as used
+ // make space for Masterpage
+ if( !pnMasterAlloc ) // create space oneself
+ {
+ if( !Pos2Page( nFAT << 2 ) )
+ return STG_EOF;
+ rtl::Reference< StgPage > pPg = m_rIo.Get( m_nPage, true );
+ if( !pPg.is() )
+ return STG_EOF;
+ m_rIo.SetToPage( pPg, m_nOffset >> 2, STG_MASTER );
+ }
+ else
+ (*pnMasterAlloc)++;
+ m_rIo.m_aHdr.SetMasters( nCount + 1 );
+ pOldPage = pMaster;
+ }
+ }
+ }
+ else
+ {
+ pMaster = m_rIo.Get( nFAT, true );
+ if ( pMaster.is() )
+ {
+ nFAT = StgCache::GetFromPage( pMaster, nMasterCount );
+ pOldPage = pMaster;
+ }
+ }
+ }
+ if( pMaster.is() )
+ return StgCache::GetFromPage( pMaster, nOff );
+ m_rIo.SetError( SVSTREAM_GENERALERROR );
+ return STG_EOF;
+}
+
+
+// Set the page number entry for the given page offset.
+
+bool StgFATStrm::SetPage( short nOff, sal_Int32 nNewPage )
+{
+ OSL_ENSURE( nOff >= 0, "The offset may not be negative!" );
+ m_aPagesCache.clear();
+ m_aUsedPageNumbers.clear();
+
+ bool bRes = true;
+ if( nOff < StgHeader::GetFAT1Size() )
+ m_rIo.m_aHdr.SetFATPage( nOff, nNewPage );
+ else
+ {
+ nOff = nOff - StgHeader::GetFAT1Size();
+ // number of master pages that we need to iterate through
+ sal_uInt16 nMasterCount = ( m_nPageSize >> 2 ) - 1;
+ sal_uInt16 nBlocks = nOff / nMasterCount;
+ // offset in the last master page
+ nOff = nOff % nMasterCount;
+
+ rtl::Reference< StgPage > pMaster;
+ sal_Int32 nFAT = m_rIo.m_aHdr.GetFATChain();
+ for( sal_uInt16 nCount = 0; nCount <= nBlocks; nCount++ )
+ {
+ if( nFAT == STG_EOF || nFAT == STG_FREE )
+ {
+ pMaster = nullptr;
+ break;
+ }
+ pMaster = m_rIo.Get( nFAT, true );
+ if ( pMaster.is() )
+ nFAT = StgCache::GetFromPage( pMaster, nMasterCount );
+ }
+ if( pMaster.is() )
+ m_rIo.SetToPage( pMaster, nOff, nNewPage );
+ else
+ {
+ m_rIo.SetError( SVSTREAM_GENERALERROR );
+ bRes = false;
+ }
+ }
+
+ // lock the page against access
+ if( bRes )
+ {
+ Pos2Page( nNewPage << 2 );
+ rtl::Reference< StgPage > pPg = m_rIo.Get( m_nPage, true );
+ if( pPg.is() )
+ m_rIo.SetToPage( pPg, m_nOffset >> 2, STG_FAT );
+ else
+ bRes = false;
+ }
+ return bRes;
+}
+
+bool StgFATStrm::SetSize( sal_Int32 nBytes )
+{
+ if ( nBytes < 0 )
+ return false;
+
+ m_aPagesCache.clear();
+ m_aUsedPageNumbers.clear();
+
+ // Set the number of entries to a multiple of the page size
+ short nOld = static_cast<short>( ( m_nSize + ( m_nPageSize - 1 ) ) / m_nPageSize );
+ short nNew = static_cast<short>(
+ ( nBytes + ( m_nPageSize - 1 ) ) / m_nPageSize ) ;
+ if( nNew < nOld )
+ {
+ // release master pages
+ for( short i = nNew; i < nOld; i++ )
+ SetPage( i, STG_FREE );
+ }
+ else
+ {
+ while( nOld < nNew )
+ {
+ // allocate master pages
+ // find a free master page slot
+ sal_Int32 nPg = 0;
+ sal_uInt16 nMasterAlloc = 0;
+ nPg = GetPage( nOld, true, &nMasterAlloc );
+ if( nPg == STG_EOF )
+ return false;
+ // 4 Bytes have been used for Allocation of each MegaMasterPage
+ nBytes += nMasterAlloc << 2;
+
+ // find a free page using the FAT allocator
+ sal_Int32 n = 1;
+ OSL_ENSURE( m_pFat, "The pointer is always initializer here!" );
+ sal_Int32 nNewPage = m_pFat->FindBlock( n );
+ if( nNewPage == STG_EOF )
+ {
+ // no free pages found; create a new page
+ // Since all pages are allocated, extend
+ // the file size for the next page!
+ nNewPage = m_nSize >> 2;
+ // if a MegaMasterPage was created avoid taking
+ // the same Page
+ nNewPage += nMasterAlloc;
+ // adjust the file size if necessary
+ if( nNewPage >= m_rIo.GetPhysPages() )
+ if( !m_rIo.SetSize( nNewPage + 1 ) )
+ return false;
+ }
+ // Set up the page with empty entries
+ rtl::Reference< StgPage > pPg = m_rIo.Copy( nNewPage );
+ if ( !pPg.is() )
+ return false;
+ for( short j = 0; j < static_cast<short>( m_nPageSize >> 2 ); j++ )
+ m_rIo.SetToPage( pPg, j, STG_FREE );
+
+ // store the page number into the master FAT
+ // Set the size before so the correct FAT can be found
+ m_nSize = ( nOld + 1 ) * m_nPageSize;
+ SetPage( nOld, nNewPage );
+
+ // MegaMasterPages were created, mark it them as used
+
+ sal_uInt32 nMax = m_rIo.m_aHdr.GetMasters( );
+ sal_uInt32 nFAT = m_rIo.m_aHdr.GetFATChain();
+ if( nMasterAlloc )
+ for( sal_uInt32 nCount = 0; nCount < nMax; nCount++ )
+ {
+ if( !Pos2Page( nFAT << 2 ) )
+ return false;
+ if( nMax - nCount <= nMasterAlloc )
+ {
+ rtl::Reference< StgPage > piPg = m_rIo.Get( m_nPage, true );
+ if( !piPg.is() )
+ return false;
+ m_rIo.SetToPage( piPg, m_nOffset >> 2, STG_MASTER );
+ }
+ rtl::Reference< StgPage > pPage = m_rIo.Get( nFAT, true );
+ if( !pPage.is() ) return false;
+ nFAT = StgCache::GetFromPage( pPage, (m_nPageSize >> 2 ) - 1 );
+ }
+
+ nOld++;
+ // We have used up 4 bytes for the STG_FAT entry
+ nBytes += 4;
+ nNew = static_cast<short>(
+ ( nBytes + ( m_nPageSize - 1 ) ) / m_nPageSize );
+ }
+ }
+ m_nSize = nNew * m_nPageSize;
+ m_rIo.m_aHdr.SetFATSize( nNew );
+ return true;
+}
+
+/////////////////////////// class StgDataStrm
+
+// This class is a normal physical stream which can be initialized
+// either with an existing dir entry or an existing FAT chain.
+// The stream has a size increment which normally is 1, but which can be
+// set to any value is you want the size to be incremented by certain values.
+
+StgDataStrm::StgDataStrm( StgIo& r, sal_Int32 nBgn, sal_Int32 nLen ) : StgStrm( r )
+{
+ Init( nBgn, nLen );
+}
+
+StgDataStrm::StgDataStrm( StgIo& r, StgDirEntry& p ) : StgStrm( r )
+{
+ m_pEntry = &p;
+ Init( p.m_aEntry.GetLeaf( STG_DATA ),
+ p.m_aEntry.GetSize() );
+}
+
+void StgDataStrm::Init( sal_Int32 nBgn, sal_Int32 nLen )
+{
+ if ( m_rIo.m_pFAT )
+ m_pFat.reset( new StgFAT( *m_rIo.m_pFAT, true ) );
+
+ OSL_ENSURE( m_pFat, "The pointer should not be empty!" );
+
+ m_nStart = m_nPage = nBgn;
+ m_nSize = nLen;
+ m_nIncr = 1;
+ m_nOffset = 0;
+ if( nLen < 0 && m_pFat )
+ {
+ // determine the actual size of the stream by scanning
+ // the FAT chain and counting the # of pages allocated
+ m_nSize = scanBuildPageChainCache();
+ }
+}
+
+// Set the size of a physical stream.
+
+bool StgDataStrm::SetSize( sal_Int32 nBytes )
+{
+ if ( !m_pFat )
+ return false;
+
+ nBytes = ( ( nBytes + m_nIncr - 1 ) / m_nIncr ) * m_nIncr;
+ sal_Int32 nOldSz = m_nSize;
+ if( nOldSz != nBytes )
+ {
+ if( !StgStrm::SetSize( nBytes ) )
+ return false;
+ sal_Int32 nMaxPage = m_pFat->GetMaxPage();
+ if( nMaxPage > m_rIo.GetPhysPages() )
+ if( !m_rIo.SetSize( nMaxPage ) )
+ return false;
+ // If we only allocated one page or less, create this
+ // page in the cache for faster throughput. The current
+ // position is the former EOF point.
+ if( ( m_nSize - 1 ) / m_nPageSize - ( nOldSz - 1 ) / m_nPageSize == 1 )
+ {
+ Pos2Page( nBytes );
+ if( m_nPage >= 0 )
+ m_rIo.Copy( m_nPage );
+ }
+ }
+ return true;
+}
+
+// Get the address of the data byte at a specified offset.
+// If bForce = true, a read of non-existent data causes
+// a read fault.
+
+void* StgDataStrm::GetPtr( sal_Int32 Pos, bool bDirty )
+{
+ if( Pos2Page( Pos ) )
+ {
+ rtl::Reference< StgPage > pPg = m_rIo.Get( m_nPage, true/*bForce*/ );
+ if (pPg.is() && m_nOffset < pPg->GetSize())
+ {
+ if( bDirty )
+ m_rIo.SetDirty( pPg );
+ return static_cast<sal_uInt8 *>(pPg->GetData()) + m_nOffset;
+ }
+ }
+ return nullptr;
+}
+
+// This could easily be adapted to a better algorithm by determining
+// the amount of consecutable blocks before doing a read. The result
+// is the number of bytes read. No error is generated on EOF.
+
+sal_Int32 StgDataStrm::Read( void* pBuf, sal_Int32 n )
+{
+ if ( n < 0 )
+ return 0;
+
+ const auto nAvailable = m_nSize - GetPos();
+ if (n > nAvailable)
+ n = nAvailable;
+ sal_Int32 nDone = 0;
+ while( n )
+ {
+ short nBytes = m_nPageSize - m_nOffset;
+ rtl::Reference< StgPage > pPg;
+ if( static_cast<sal_Int32>(nBytes) > n )
+ nBytes = static_cast<short>(n);
+ if( nBytes )
+ {
+ short nRes;
+ void *p = static_cast<sal_uInt8 *>(pBuf) + nDone;
+ if( nBytes == m_nPageSize )
+ {
+ pPg = m_rIo.Find( m_nPage );
+ if( pPg.is() )
+ {
+ // data is present, so use the cached data
+ memcpy( p, pPg->GetData(), nBytes );
+ nRes = nBytes;
+ }
+ else
+ // do a direct (unbuffered) read
+ nRes = static_cast<short>(m_rIo.Read( m_nPage, p )) * m_nPageSize;
+ }
+ else
+ {
+ // partial block read through the cache.
+ pPg = m_rIo.Get( m_nPage, false );
+ if( !pPg.is() )
+ break;
+ memcpy( p, static_cast<sal_uInt8*>(pPg->GetData()) + m_nOffset, nBytes );
+ nRes = nBytes;
+ }
+ nDone += nRes;
+ SetPos(GetPos() + nRes, true);
+ n -= nRes;
+ m_nOffset = m_nOffset + nRes;
+ if( nRes != nBytes )
+ break; // read error or EOF
+ }
+ // Switch to next page if necessary
+ if (m_nOffset >= m_nPageSize && !Pos2Page(GetPos()))
+ break;
+ }
+ return nDone;
+}
+
+sal_Int32 StgDataStrm::Write( const void* pBuf, sal_Int32 n )
+{
+ if ( n < 0 )
+ return 0;
+
+ sal_Int32 nDone = 0;
+ if( ( GetPos() + n ) > m_nSize )
+ {
+ sal_Int32 nOld = GetPos();
+ if( !SetSize( nOld + n ) )
+ return 0;
+ Pos2Page( nOld );
+ }
+ while( n )
+ {
+ short nBytes = m_nPageSize - m_nOffset;
+ rtl::Reference< StgPage > pPg;
+ if( static_cast<sal_Int32>(nBytes) > n )
+ nBytes = static_cast<short>(n);
+ if( nBytes )
+ {
+ short nRes;
+ const void *p = static_cast<const sal_uInt8 *>(pBuf) + nDone;
+ if( nBytes == m_nPageSize )
+ {
+ pPg = m_rIo.Find( m_nPage );
+ if( pPg.is() )
+ {
+ // data is present, so use the cached data
+ memcpy( pPg->GetData(), p, nBytes );
+ m_rIo.SetDirty( pPg );
+ nRes = nBytes;
+ }
+ else
+ // do a direct (unbuffered) write
+ nRes = static_cast<short>(m_rIo.Write( m_nPage, p )) * m_nPageSize;
+ }
+ else
+ {
+ // partial block read through the cache.
+ pPg = m_rIo.Get( m_nPage, false );
+ if( !pPg.is() )
+ break;
+ memcpy( static_cast<sal_uInt8*>(pPg->GetData()) + m_nOffset, p, nBytes );
+ m_rIo.SetDirty( pPg );
+ nRes = nBytes;
+ }
+ nDone += nRes;
+ SetPos(GetPos() + nRes, true);
+ n -= nRes;
+ m_nOffset = m_nOffset + nRes;
+ if( nRes != nBytes )
+ break; // read error
+ }
+ // Switch to next page if necessary
+ if( m_nOffset >= m_nPageSize && !Pos2Page(GetPos()) )
+ break;
+ }
+ return nDone;
+}
+
+//////////////////////////// class StgSmallStream
+
+// The small stream class provides access to streams with a size < 4096 bytes.
+// This stream is a StgStream containing small pages. The FAT for this stream
+// is also a StgStream. The start of the FAT is in the header at DataRootPage,
+// the stream itself is pointed to by the root entry (it holds start & size).
+
+StgSmallStrm::StgSmallStrm( StgIo& r, sal_Int32 nBgn ) : StgStrm( r )
+{
+ Init( nBgn, 0 );
+}
+
+StgSmallStrm::StgSmallStrm( StgIo& r, StgDirEntry& p ) : StgStrm( r )
+{
+ m_pEntry = &p;
+ Init( p.m_aEntry.GetLeaf( STG_DATA ),
+ p.m_aEntry.GetSize() );
+}
+
+void StgSmallStrm::Init( sal_Int32 nBgn, sal_Int32 nLen )
+{
+ if ( m_rIo.m_pDataFAT )
+ m_pFat.reset( new StgFAT( *m_rIo.m_pDataFAT, false ) );
+ m_pData = m_rIo.m_pDataStrm;
+ OSL_ENSURE( m_pFat && m_pData, "The pointers should not be empty!" );
+
+ m_nPageSize = m_rIo.GetDataPageSize();
+ m_nStart =
+ m_nPage = nBgn;
+ m_nSize = nLen;
+}
+
+// This could easily be adapted to a better algorithm by determining
+// the amount of consecutable blocks before doing a read. The result
+// is the number of bytes read. No error is generated on EOF.
+
+sal_Int32 StgSmallStrm::Read( void* pBuf, sal_Int32 n )
+{
+ // We can safely assume that reads are not huge, since the
+ // small stream is likely to be < 64 KBytes.
+ sal_Int32 nBytePos = GetPos();
+ if( ( nBytePos + n ) > m_nSize )
+ n = m_nSize - nBytePos;
+ sal_Int32 nDone = 0;
+ while( n )
+ {
+ short nBytes = m_nPageSize - m_nOffset;
+ if( static_cast<sal_Int32>(nBytes) > n )
+ nBytes = static_cast<short>(n);
+ if( nBytes )
+ {
+ if (!m_pData)
+ break;
+ sal_Int32 nPos;
+ if (o3tl::checked_multiply<sal_Int32>(m_nPage, m_nPageSize, nPos))
+ break;
+ if (!m_pData->Pos2Page(nPos + m_nOffset))
+ break;
+ // all reading through the stream
+ short nRes = static_cast<short>(m_pData->Read( static_cast<sal_uInt8*>(pBuf) + nDone, nBytes ));
+ nDone += nRes;
+ SetPos(GetPos() + nRes, true);
+ n -= nRes;
+ m_nOffset = m_nOffset + nRes;
+ // read problem?
+ if( nRes != nBytes )
+ break;
+ }
+ // Switch to next page if necessary
+ if (m_nOffset >= m_nPageSize && !Pos2Page(GetPos()))
+ break;
+ }
+ return nDone;
+}
+
+sal_Int32 StgSmallStrm::Write( const void* pBuf, sal_Int32 n )
+{
+ // you can safely assume that reads are not huge, since the
+ // small stream is likely to be < 64 KBytes.
+ sal_Int32 nDone = 0;
+ sal_Int32 nOldPos = GetPos();
+ if( ( nOldPos + n ) > m_nSize )
+ {
+ if (!SetSize(nOldPos + n))
+ return 0;
+ Pos2Page(nOldPos);
+ }
+ while( n )
+ {
+ short nBytes = m_nPageSize - m_nOffset;
+ if( static_cast<sal_Int32>(nBytes) > n )
+ nBytes = static_cast<short>(n);
+ if( nBytes )
+ {
+ // all writing goes through the stream
+ sal_Int32 nDataPos = m_nPage * m_nPageSize + m_nOffset;
+ if ( !m_pData
+ || ( m_pData->GetSize() < ( nDataPos + nBytes )
+ && !m_pData->SetSize( nDataPos + nBytes ) ) )
+ break;
+ if( !m_pData->Pos2Page( nDataPos ) )
+ break;
+ short nRes = static_cast<short>(m_pData->Write( static_cast<sal_uInt8 const *>(pBuf) + nDone, nBytes ));
+ nDone += nRes;
+ SetPos(GetPos() + nRes, true);
+ n -= nRes;
+ m_nOffset = m_nOffset + nRes;
+ // write problem?
+ if( nRes != nBytes )
+ break;
+ }
+ // Switch to next page if necessary
+ if( m_nOffset >= m_nPageSize && !Pos2Page(GetPos()) )
+ break;
+ }
+ return nDone;
+}
+
+/////////////////////////// class StgTmpStrm
+
+// The temporary stream uses a memory stream if < 32K, otherwise a
+// temporary file.
+
+#define THRESHOLD 32768L
+
+StgTmpStrm::StgTmpStrm( sal_uInt64 nInitSize )
+ : SvMemoryStream( nInitSize > THRESHOLD
+ ? 16
+ : ( nInitSize ? nInitSize : 16 ), 4096 )
+{
+ m_pStrm = nullptr;
+ // this calls FlushData, so all members should be set by this time
+ SetBufferSize( 0 );
+ if( nInitSize > THRESHOLD )
+ SetSize( nInitSize );
+}
+
+bool StgTmpStrm::Copy( StgTmpStrm& rSrc )
+{
+ sal_uInt64 n = rSrc.GetSize();
+ const sal_uInt64 nCur = rSrc.Tell();
+ SetSize( n );
+ if( GetError() == ERRCODE_NONE )
+ {
+ std::unique_ptr<sal_uInt8[]> p(new sal_uInt8[ 4096 ]);
+ rSrc.Seek( 0 );
+ Seek( 0 );
+ while( n )
+ {
+ const sal_uInt64 nn = std::min<sal_uInt64>(n, 4096);
+ if (rSrc.ReadBytes( p.get(), nn ) != nn)
+ break;
+ if (WriteBytes( p.get(), nn ) != nn)
+ break;
+ n -= nn;
+ }
+ p.reset();
+ rSrc.Seek( nCur );
+ Seek( nCur );
+ return n == 0;
+ }
+ else
+ return false;
+}
+
+StgTmpStrm::~StgTmpStrm()
+{
+ if( m_pStrm )
+ {
+ m_pStrm->Close();
+ osl::File::remove( m_aName );
+ delete m_pStrm;
+ }
+}
+
+sal_uInt64 StgTmpStrm::GetSize() const
+{
+ sal_uInt64 n;
+ if( m_pStrm )
+ {
+ n = m_pStrm->TellEnd();
+ }
+ else
+ n = nEndOfData;
+ return n;
+}
+
+void StgTmpStrm::SetSize(sal_uInt64 n)
+{
+ if( m_pStrm )
+ m_pStrm->SetStreamSize( n );
+ else
+ {
+ if( n > THRESHOLD )
+ {
+ m_aName = utl::CreateTempURL();
+ std::unique_ptr<SvFileStream> s(new SvFileStream( m_aName, StreamMode::READWRITE ));
+ const sal_uInt64 nCur = Tell();
+ sal_uInt64 i = nEndOfData;
+ std::unique_ptr<sal_uInt8[]> p(new sal_uInt8[ 4096 ]);
+ if( i )
+ {
+ Seek( 0 );
+ while( i )
+ {
+ const sal_uInt64 nb = std::min<sal_uInt64>(i, 4096);
+ if (ReadBytes(p.get(), nb) == nb
+ && s->WriteBytes(p.get(), nb) == nb)
+ i -= nb;
+ else
+ break;
+ }
+ }
+ if( !i && n > nEndOfData )
+ {
+ // We have to write one byte at the end of the file
+ // if the file is bigger than the memstream to see
+ // if it fits on disk
+ s->Seek(nEndOfData);
+ memset(p.get(), 0x00, 4096);
+ i = n - nEndOfData;
+ while (i)
+ {
+ const sal_uInt64 nb = std::min<sal_uInt64>(i, 4096);
+ if (s->WriteBytes(p.get(), nb) == nb)
+ i -= nb;
+ else
+ break; // error
+ }
+ s->Flush();
+ if( s->GetError() != ERRCODE_NONE )
+ i = 1;
+ }
+ Seek( nCur );
+ s->Seek( nCur );
+ if( i )
+ {
+ SetError( s->GetError() );
+ return;
+ }
+ m_pStrm = s.release();
+ // Shrink the memory to 16 bytes, which seems to be the minimum
+ ReAllocateMemory( - ( static_cast<tools::Long>(nEndOfData) - 16 ) );
+ }
+ else
+ {
+ if( n > nEndOfData )
+ {
+ SvMemoryStream::SetSize(n);
+ }
+ else
+ nEndOfData = n;
+ }
+ }
+}
+
+std::size_t StgTmpStrm::GetData( void* pData, std::size_t n )
+{
+ if( m_pStrm )
+ {
+ n = m_pStrm->ReadBytes( pData, n );
+ SetError( m_pStrm->GetError() );
+ return n;
+ }
+ else
+ return SvMemoryStream::GetData( pData, n );
+}
+
+std::size_t StgTmpStrm::PutData( const void* pData, std::size_t n )
+{
+ sal_uInt64 nCur = Tell();
+ sal_uInt64 nNew = nCur + n;
+ if( nNew > THRESHOLD && !m_pStrm )
+ {
+ SetSize( nNew );
+ if( GetError() != ERRCODE_NONE )
+ return 0;
+ }
+ if( m_pStrm )
+ {
+ nNew = m_pStrm->WriteBytes( pData, n );
+ SetError( m_pStrm->GetError() );
+ }
+ else
+ nNew = SvMemoryStream::PutData( pData, n );
+ return nNew;
+}
+
+sal_uInt64 StgTmpStrm::SeekPos(sal_uInt64 n)
+{
+ // check if a truncated STREAM_SEEK_TO_END was passed
+ assert(n != SAL_MAX_UINT32);
+ if( n == STREAM_SEEK_TO_END )
+ n = GetSize();
+ if( n > THRESHOLD && !m_pStrm )
+ {
+ SetSize( n );
+ if( GetError() != ERRCODE_NONE )
+ return Tell();
+ else
+ return n;
+ }
+ else if( m_pStrm )
+ {
+ n = m_pStrm->Seek( n );
+ SetError( m_pStrm->GetError() );
+ return n;
+ }
+ else
+ return SvMemoryStream::SeekPos( n );
+}
+
+void StgTmpStrm::FlushData()
+{
+ if( m_pStrm )
+ {
+ m_pStrm->Flush();
+ SetError( m_pStrm->GetError() );
+ }
+ else
+ SvMemoryStream::FlushData();
+}
+
+/* vim:set shiftwidth=4 softtabstop=4 expandtab: */