firebird-mirror/src/jrd/os/win32/winnt.cpp

/*
 *	PROGRAM:	JRD Access Method
 *	MODULE:		winnt.cpp
 *	DESCRIPTION:	Windows NT specific physical IO
 *
 * The contents of this file are subject to the Interbase Public
 * License Version 1.0 (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.Inprise.com/IPL.html
 *
 * Software distributed under the License is distributed on an
 * "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, either express
 * or implied. See the License for the specific language governing
 * rights and limitations under the License.
 *
 * The Original Code was created by Inprise Corporation
 * and its predecessors. Portions created by Inprise Corporation are
 * Copyright (C) Inprise Corporation.
 *
 * All Rights Reserved.
 * Contributor(s): ______________________________________.
 * 2001.07.06 Sean Leyne - Code Cleanup, removed "#ifdef READONLY_DATABASE"
 *                         conditionals, as the engine now fully supports
 *                         readonly databases.
 *
 *
 * 17-Oct-2001 Mike Nordell: Non-shared file access
 *
 * 02-Nov-2001 Mike Nordell: Synch with FB1 changes.
 *
 * 20-Nov-2001 Ann Harrison: Make page count work on db with forced write
 *
 * 21-Nov-2001 Ann Harrison: Allow read sharing so gstat works
 */

#include "firebird.h"
#include <string.h>
#include "../jrd/common.h"
#include "../jrd/jrd.h"
#include "../jrd/os/pio.h"
#include "../jrd/ods.h"
#include "../jrd/lck.h"
#include "../jrd/cch.h"
#include "gen/iberror.h"
#include "../jrd/cch_proto.h"
#include "../jrd/err_proto.h"
#include "../jrd/isc_proto.h"
#include "../jrd/isc_f_proto.h"

#include "../jrd/lck_proto.h"
#include "../jrd/mov_proto.h"
#include "../jrd/os/pio_proto.h"
#include "../common/classes/init.h"
#include "../common/config/config.h"

#include <windows.h>

#ifdef WIN9X_SUPPORT
#include "win9x_nt.h"
#endif

namespace Jrd {

class FileExtendLockGuard
{
public:
	FileExtendLockGuard(Firebird::RWLock* lock, bool exclusive) :
	  m_lock(lock), m_exclusive(exclusive)
	{
		if (m_exclusive) {
			fb_assert(m_lock);
		}
		if (m_lock)
		{
			if (m_exclusive)
				m_lock->beginWrite();
			else
				m_lock->beginRead();
		}
	}

	~FileExtendLockGuard()
	{
		if (m_lock)
		{
			if (m_exclusive)
				m_lock->endWrite();
			else
				m_lock->endRead();
		}
	}

private:
	// copying is prohibited
	FileExtendLockGuard(const FileExtendLockGuard&);
	FileExtendLockGuard& operator=(const FileExtendLockGuard&);

	Firebird::RWLock* const m_lock;
	const bool m_exclusive;
};


} // namespace Jrd

using namespace Jrd;
using namespace Firebird;

#ifdef TEXT
#undef TEXT
#endif
#define TEXT		SCHAR

#ifdef SUPERSERVER_V2
static void release_io_event(jrd_file*, OVERLAPPED*);
#endif
static bool	maybeCloseFile(HANDLE&);
static jrd_file* seek_file(jrd_file*, BufferDesc*, ISC_STATUS*, OVERLAPPED*, OVERLAPPED**);
static jrd_file* setup_file(Database*, const Firebird::PathName&, HANDLE, bool);
static bool nt_error(const TEXT*, const jrd_file*, ISC_STATUS, ISC_STATUS* const);
static void adjustFileSystemCacheSize();

struct AdjustFsCache
{
	static void init() { adjustFileSystemCacheSize(); }
};

static InitMutex<AdjustFsCache> adjustFsCacheOnce;

#ifdef SUPERSERVER_V2
static const DWORD g_dwShareFlags = FILE_SHARE_READ;	// no write sharing
static const DWORD g_dwExtraFlags = FILE_FLAG_OVERLAPPED |
									FILE_FLAG_NO_BUFFERING |
									FILE_FLAG_RANDOM_ACCESS;
#else
#ifdef SUPERSERVER
static const DWORD g_dwShareFlags = FILE_SHARE_READ;	// no write sharing
static const DWORD g_dwExtraFlags = FILE_FLAG_RANDOM_ACCESS;
#else
static const DWORD g_dwShareFlags = FILE_SHARE_READ | FILE_SHARE_WRITE;
static const DWORD g_dwExtraFlags = FILE_FLAG_RANDOM_ACCESS;
#endif
#endif

static const DWORD g_dwShareTempFlags = FILE_SHARE_READ;
static const DWORD g_dwExtraTempFlags = FILE_ATTRIBUTE_TEMPORARY |
										FILE_FLAG_DELETE_ON_CLOSE;


int PIO_add_file(Database* dbb, jrd_file* main_file, const Firebird::PathName& file_name, SLONG start)
{
/**************************************
 *
 *	P I O _ a d d _ f i l e
 *
 **************************************
 *
 * Functional description
 *	Add a file to an existing database.  Return the sequence
 *	number of the new file.  If anything goes wrong, return a
 *	sequence of 0.
 *
 **************************************/
	jrd_file* const new_file = PIO_create(dbb, file_name, false, false, false);
	if (!new_file) {
		return 0;
	}

	new_file->fil_min_page = start;
	USHORT sequence = 1;

	jrd_file* file;
	for (file = main_file; file->fil_next; file = file->fil_next) {
		++sequence;
	}

	file->fil_max_page = start - 1;
	file->fil_next = new_file;

	return sequence;
}


void PIO_close(jrd_file* main_file)
{
/**************************************
 *
 *	P I O _ c l o s e
 *
 **************************************
 *
 * Functional description
 *
 **************************************/
	for (jrd_file* file = main_file; file; file = file->fil_next)
	{
		if (maybeCloseFile(file->fil_desc))
		{
#ifdef SUPERSERVER_V2
			for (int i = 0; i < MAX_FILE_IO; i++)
			{
				if (file->fil_io_events[i])
				{
					CloseHandle((HANDLE) file->fil_io_events[i]);
					file->fil_io_events[i] = 0;
				}
			}
#endif
		}
	}
}


jrd_file* PIO_create(Database* dbb, const Firebird::PathName& string,
					 const bool overwrite, const bool temporary, const bool share_delete)
{
/**************************************
 *
 *	P I O _ c r e a t e
 *
 **************************************
 *
 * Functional description
 *	Create a new database file.
 *
 **************************************/
	adjustFsCacheOnce.init();

	const TEXT* file_name = string.c_str();

	DWORD dwShareMode = (temporary ? g_dwShareTempFlags : g_dwShareFlags);
	if (share_delete)
		dwShareMode |= FILE_SHARE_DELETE;

	DWORD dwFlagsAndAttributes = FILE_ATTRIBUTE_NORMAL | g_dwExtraFlags;
	if (temporary)
		dwFlagsAndAttributes |= g_dwExtraTempFlags;

	const HANDLE desc = CreateFile(file_name,
					  GENERIC_READ | GENERIC_WRITE,
					  dwShareMode,
					  NULL,
					  (overwrite ? CREATE_ALWAYS : CREATE_NEW),
					  dwFlagsAndAttributes,
					  0);

	if (desc == INVALID_HANDLE_VALUE)
	{
		ERR_post(Arg::Gds(isc_io_error) << Arg::Str("CreateFile (create)") << Arg::Str(string) <<
				 Arg::Gds(isc_io_create_err) << Arg::Windows(GetLastError()));
	}

	// File open succeeded.  Now expand the file name.
	// workspace is the expanded name here

	Firebird::PathName workspace(string);
	ISC_expand_filename(workspace, false);

	return setup_file(dbb, workspace, desc, false);
}


bool PIO_expand(const TEXT* file_name, USHORT file_length, TEXT* expanded_name, size_t len_expanded)
{
/**************************************
 *
 *	P I O _ e x p a n d
 *
 **************************************
 *
 * Functional description
 *	Fully expand a file name.  If the file doesn't exist, do something
 *	intelligent.
 *
 **************************************/

	return ISC_expand_filename(file_name, file_length, expanded_name, len_expanded, false);
}


void PIO_extend(Database* dbb, jrd_file* main_file, const ULONG extPages, const USHORT pageSize)
{
/**************************************
 *
 *	P I O _ e x t e n d
 *
 **************************************
 *
 * Functional description
 *	Extend file by extPages pages of pageSize size.
 *
 **************************************/

	// hvlad: prevent other reading\writing threads from changing file pointer.
	// As we open file without FILE_FLAG_OVERLAPPED, ReadFile\WriteFile calls
	// will change file pointer we set here and file truncation instead of file
	// extension will occurs.
	// It solved issue CORE-1468 (database file corruption when file extension
	// and read\write activity performed simultaneously)

	// if file have no extend lock it is better to not extend file than corrupt it
	if (!main_file->fil_ext_lock)
		return;

	Database::Checkout dcoHolder(dbb);
	FileExtendLockGuard extLock(main_file->fil_ext_lock, true);

	ULONG leftPages = extPages;
	for (jrd_file* file = main_file; file && leftPages; file = file->fil_next)
	{
		const ULONG filePages = PIO_get_number_of_pages(file, pageSize);
		const ULONG fileMaxPages = (file->fil_max_page == MAX_ULONG) ?
									MAX_ULONG : file->fil_max_page - file->fil_min_page + 1;
		if (filePages < fileMaxPages)
		{
			const ULONG extendBy = MIN(fileMaxPages - filePages + file->fil_fudge, leftPages);

			HANDLE hFile = file->fil_desc;

			LARGE_INTEGER newSize;
			newSize.QuadPart = ((ULONGLONG) filePages + extendBy) * pageSize;

			const DWORD ret = SetFilePointer(hFile, newSize.LowPart, &newSize.HighPart, FILE_BEGIN);
			if (ret == INVALID_SET_FILE_POINTER && GetLastError() != NO_ERROR) {
				nt_error("SetFilePointer", file, isc_io_write_err, 0);
			}
			if (!SetEndOfFile(hFile)) {
				nt_error("SetEndOfFile", file, isc_io_write_err, 0);
			}

			leftPages -= extendBy;
		}
	}
}


void PIO_flush(Database* dbb, jrd_file* main_file)
{
/**************************************
 *
 *	P I O _ f l u s h
 *
 **************************************
 *
 * Functional description
 *	Flush the operating system cache back to good, solid oxide.
 *
 **************************************/
	Database::Checkout dcoHolder(dbb);
	for (jrd_file* file = main_file; file; file = file->fil_next)
	{
		FlushFileBuffers(file->fil_desc);
	}
}


void PIO_force_write(jrd_file* file, const bool forceWrite, const bool notUseFSCache)
{
/**************************************
 *
 *	P I O _ f o r c e _ w r i t e
 *
 **************************************
 *
 * Functional description
 *	Set (or clear) force write, if possible, for the database.
 *
 **************************************/

	const bool oldForce = (file->fil_flags & FIL_force_write) != 0;
	const bool oldNotUseCache = (file->fil_flags & FIL_no_fs_cache) != 0;

	if (forceWrite != oldForce || notUseFSCache != oldNotUseCache)
	{
		const int force = forceWrite ? FILE_FLAG_WRITE_THROUGH : 0;
		const int fsCache = notUseFSCache ? FILE_FLAG_NO_BUFFERING : 0;
		const int writeMode = (file->fil_flags & FIL_readonly) ? 0 : GENERIC_WRITE;

        HANDLE& hFile = file->fil_desc;
		maybeCloseFile(hFile);
		hFile = CreateFile(file->fil_string,
						  GENERIC_READ | writeMode,
						  g_dwShareFlags,
						  NULL,
						  OPEN_EXISTING,
						  FILE_ATTRIBUTE_NORMAL | force | fsCache | g_dwExtraFlags,
						  0);

		if (hFile == INVALID_HANDLE_VALUE)
		{
			ERR_post(Arg::Gds(isc_io_error) << Arg::Str("CreateFile (force write)") <<
											   Arg::Str(file->fil_string) <<
					 Arg::Gds(isc_io_access_err) << Arg::Windows(GetLastError()));
		}

		if (forceWrite) {
			file->fil_flags |= FIL_force_write;
		}
		else {
			file->fil_flags &= ~FIL_force_write;
		}
		if (notUseFSCache) {
			file->fil_flags |= FIL_no_fs_cache;
		}
		else {
			file->fil_flags &= ~FIL_no_fs_cache;
		}
	}
}


void PIO_header(Database* dbb, SCHAR* address, int length)
{
/**************************************
 *
 *	P I O _ h e a d e r
 *
 **************************************
 *
 * Functional description
 *	Read the page header.  This assumes that the file has not been
 *	repositioned since the file was originally mapped.
 *  The detail of Win32 implementation is that it doesn't assume
 *  this fact as seeks to first byte of file initially, but external
 *  callers should not rely on this behavior
 *
 **************************************/
	PageSpace* pageSpace = dbb->dbb_page_manager.findPageSpace(DB_PAGE_SPACE);
	jrd_file* file = pageSpace->file;
	HANDLE desc = file->fil_desc;

	OVERLAPPED overlapped;
	OVERLAPPED* overlapped_ptr;

	memset(&overlapped, 0, sizeof(OVERLAPPED));
	overlapped_ptr = &overlapped;
#ifdef SUPERSERVER_V2
	if (!(overlapped.hEvent = CreateEvent(NULL, TRUE, FALSE, NULL)))
		nt_error("Overlapped event", file, isc_io_read_err, 0);
	ResetEvent(overlapped.hEvent);
#endif

    DWORD actual_length;
	if (dbb->dbb_encrypt_key.hasData())
	{
		SLONG spare_buffer[MAX_PAGE_SIZE / sizeof(SLONG)];

		if (!ReadFile(desc, spare_buffer, length, &actual_length, overlapped_ptr) ||
			actual_length != (DWORD) length)
		{
			nt_error("ReadFile", file, isc_io_read_err, 0);
		}

		(*dbb->dbb_decrypt) (dbb->dbb_encrypt_key.c_str(), spare_buffer, length, address);
	}
	else
	{
		if (!ReadFile(desc, address, length, &actual_length, overlapped_ptr) ||
			actual_length != (DWORD) length)
		{
#ifdef SUPERSERVER_V2
			if (!GetOverlappedResult(desc, overlapped_ptr, &actual_length, TRUE) ||
				actual_length != length)
			{
				CloseHandle(overlapped.hEvent);
				nt_error("GetOverlappedResult", file, isc_io_read_err, 0);
			}
#else
			nt_error("ReadFile", file, isc_io_read_err, 0);
#endif
		}
	}

#ifdef SUPERSERVER_V2
	CloseHandle(overlapped.hEvent);
#endif
}

// we need a class here only to return memory on shutdown and avoid
// false memory leak reports
static Firebird::InitInstance<ZeroBuffer> zeros;


USHORT PIO_init_data(Database* dbb, jrd_file* main_file, ISC_STATUS* status_vector,
					 ULONG startPage, USHORT initPages)
{
/**************************************
 *
 *	P I O _ i n i t _ d a t a
 *
 **************************************
 *
 * Functional description
 *	Initialize tail of file with zeros
 *
 **************************************/
	const char* const zero_buff = zeros().getBuffer();
	const size_t zero_buff_size = zeros().getSize();

	Database::Checkout dcoHolder(dbb);
	FileExtendLockGuard extLock(main_file->fil_ext_lock, false);

	// Fake buffer, used in seek_file. Page space ID doesn't matter there
	// as we already know file to work with
	BufferDesc bdb;
	bdb.bdb_dbb = dbb;
	bdb.bdb_page = PageNumber(0, startPage);

	OVERLAPPED overlapped;
	OVERLAPPED* overlapped_ptr;
	jrd_file* file = seek_file(main_file, &bdb, status_vector, &overlapped, &overlapped_ptr);

	if (!file)
		return 0;

	if (file->fil_min_page + 8 > startPage)
		return 0;

	USHORT leftPages = initPages;
	const ULONG initBy = MIN(file->fil_max_page - startPage, leftPages);
	if (initBy < leftPages)
		leftPages = initBy;

	for (ULONG i = startPage; i < startPage + initBy; )
	{
		bdb.bdb_page = PageNumber(0, i);
		USHORT write_pages = zero_buff_size / dbb->dbb_page_size;
		if (write_pages > leftPages)
			write_pages = leftPages;

		seek_file(main_file, &bdb, status_vector, &overlapped, &overlapped_ptr);

		const DWORD to_write = (DWORD) write_pages * dbb->dbb_page_size;
		DWORD written;

		if (!WriteFile(file->fil_desc, zero_buff, to_write, &written, overlapped_ptr) ||
			to_write != written)
		{
			nt_error("WriteFile", file, isc_io_write_err, status_vector);
			break;
		}

		leftPages -= write_pages;
		i += write_pages;
	}

	return (initPages - leftPages);
}


jrd_file* PIO_open(Database* dbb,
				   const Firebird::PathName& string,
				   const Firebird::PathName& file_name,
				   const bool share_delete)
{
/**************************************
 *
 *	P I O _ o p e n
 *
 **************************************
 *
 * Functional description
 *	Open a database file.
 *
 **************************************/
	const TEXT* const ptr = (string.hasData() ? string : file_name).c_str();
	bool readOnly = false;

	adjustFsCacheOnce.init();

	HANDLE desc = CreateFile(ptr,
					  GENERIC_READ | GENERIC_WRITE,
					  g_dwShareFlags | (share_delete ? FILE_SHARE_DELETE : 0),
					  NULL,
					  OPEN_EXISTING,
					  FILE_ATTRIBUTE_NORMAL |
					  g_dwExtraFlags,
					  0);

	if (desc == INVALID_HANDLE_VALUE)
	{
		// Try opening the database file in ReadOnly mode.
		// The database file could be on a RO medium (CD-ROM etc.).
		// If this fileopen fails, return error.
		desc = CreateFile(ptr,
						  GENERIC_READ,
						  FILE_SHARE_READ,
						  NULL,
						  OPEN_EXISTING,
						  FILE_ATTRIBUTE_NORMAL |
						  g_dwExtraFlags, 0);

		if (desc == INVALID_HANDLE_VALUE)
		{
			ERR_post(Arg::Gds(isc_io_error) << Arg::Str("CreateFile (open)") << Arg::Str(file_name) <<
					 Arg::Gds(isc_io_open_err) << Arg::Windows(GetLastError()));
		}
		else
		{
			// If this is the primary file, set Database flag to indicate that it is
			// being opened ReadOnly. This flag will be used later to compare with
			// the Header Page flag setting to make sure that the database is set ReadOnly.
			readOnly = true;
			PageSpace* pageSpace = dbb->dbb_page_manager.findPageSpace(DB_PAGE_SPACE);
			if (!pageSpace->file)
				dbb->dbb_flags |= DBB_being_opened_read_only;
		}
	}

	return setup_file(dbb, string, desc, readOnly);
}


bool PIO_read(jrd_file* file, BufferDesc* bdb, Ods::pag* page, ISC_STATUS* status_vector)
{
/**************************************
 *
 *	P I O _ r e a d
 *
 **************************************
 *
 * Functional description
 *	Read a data page.
 *
 **************************************/
	Database* const dbb = bdb->bdb_dbb;
	const DWORD size = dbb->dbb_page_size;

	Database::Checkout dcoHolder(dbb);
	FileExtendLockGuard extLock(file->fil_ext_lock, false);

	OVERLAPPED overlapped, *overlapped_ptr;
	if (!(file = seek_file(file, bdb, status_vector, &overlapped, &overlapped_ptr)))
		return false;

	HANDLE desc = file->fil_desc;

	if (dbb->dbb_encrypt_key.hasData())
	{
		SLONG spare_buffer[MAX_PAGE_SIZE / sizeof(SLONG)];
        DWORD actual_length;

		if (!ReadFile(desc, spare_buffer, size, &actual_length, overlapped_ptr) ||
			actual_length != size)
		{
			return nt_error("ReadFile", file, isc_io_read_err, status_vector);
		}

		(*dbb->dbb_decrypt) (dbb->dbb_encrypt_key.c_str(), spare_buffer, size, page);
	}
	else
	{
		DWORD actual_length;
		if (!ReadFile(desc, page, size, &actual_length, overlapped_ptr) ||
			actual_length != size)
		{
#ifdef SUPERSERVER_V2
			if (!GetOverlappedResult(desc, overlapped_ptr, &actual_length, TRUE) ||
				actual_length != size)
			{
				release_io_event(file, overlapped_ptr);
				return nt_error("GetOverlappedResult", file, isc_io_read_err, status_vector);
			}
#else
			return nt_error("ReadFile", file, isc_io_read_err, status_vector);
#endif
		}
	}

#ifdef SUPERSERVER_V2
	release_io_event(file, overlapped_ptr);
#endif

	return true;
}


#ifdef SUPERSERVER_V2
bool PIO_read_ahead(Database*		dbb,
				   SLONG	start_page,
				   SCHAR*	buffer,
				   SLONG	pages,
				   phys_io_blk*		piob,
				   ISC_STATUS*	status_vector)
{
/**************************************
 *
 *	P I O _ r e a d _ a h e a d
 *
 **************************************
 *
 * Functional description
 *	Read a contiguous set of pages. The only
 *	tricky part is to segment the I/O when crossing
 *	file boundaries.
 *
 **************************************/
	OVERLAPPED overlapped, *overlapped_ptr;

	Database::Checkout dcoHolder(dbb);

	// If an I/O status block was passed the caller wants to queue an asynchronous I/O.

	if (!piob) {
		overlapped_ptr = &overlapped;
	}
	else
	{
		overlapped_ptr = (OVERLAPPED*) &piob->piob_io_event;
		piob->piob_flags = 0;
	}

	BufferDesc bdb;
	while (pages)
	{
		// Setup up a dummy buffer descriptor block for seeking file.

		bdb.bdb_dbb = dbb;
		bdb.bdb_page = start_page;

		jrd_file* file = seek_file(dbb->dbb_file, &bdb, status_vector, overlapped_ptr, &overlapped_ptr);
		if (!file) {
			return false;
		}

		// Check that every page within the set resides in the same database
		// file. If not read what you can and loop back for the rest.

		DWORD segmented_length = 0;
		while (pages && start_page >= file->fil_min_page && start_page <= file->fil_max_page)
		{
			segmented_length += dbb->dbb_page_size;
			++start_page;
			--pages;
		}

		HANDLE desc = file->fil_desc;

		DWORD actual_length;
		if (ReadFile(desc, buffer, segmented_length, &actual_length, overlapped_ptr) &&
			actual_length == segmented_length)
		{
			if (piob && !pages) {
				piob->piob_flags = PIOB_success;
			}
		}
		else if (piob && !pages)
		{
			piob->piob_flags = PIOB_pending;
			piob->piob_desc = reinterpret_cast<SLONG>(desc);
			piob->piob_file = file;
			piob->piob_io_length = segmented_length;
		}
		else if (!GetOverlappedResult(desc, overlapped_ptr, &actual_length, TRUE) ||
			actual_length != segmented_length)
		{
			if (piob) {
					piob->piob_flags = PIOB_error;
			}
			release_io_event(file, overlapped_ptr);
			return nt_error("GetOverlappedResult", file, isc_io_read_err, status_vector);
		}

		if (!piob || (piob->piob_flags & (PIOB_success | PIOB_error)))
			release_io_event(file, overlapped_ptr);

		buffer += segmented_length;
	}

	return true;
}
#endif


#ifdef SUPERSERVER_V2
bool PIO_status(Database* dbb, phys_io_blk* piob, ISC_STATUS* status_vector)
{
/**************************************
 *
 *	P I O _ s t a t u s
 *
 **************************************
 *
 * Functional description
 *	Check the status of an asynchronous I/O.
 *
 **************************************/
	Database::Checkout dcoHolder(dbb);

	if (!(piob->piob_flags & PIOB_success))
	{
		if (piob->piob_flags & PIOB_error) {
			return false;
		}
		DWORD actual_length;
		if (!GetOverlappedResult((HANDLE) piob->piob_desc,
								 (OVERLAPPED*) &piob->piob_io_event,
								 &actual_length,
								 piob->piob_wait) ||
			actual_length != piob->piob_io_length)
		{
			release_io_event(piob->piob_file, (OVERLAPPED*) &piob->piob_io_event);
			return nt_error("GetOverlappedResult", piob->piob_file, isc_io_error, status_vector);
			// io_error is wrong here as primary & secondary error.
		}
	}

	release_io_event(piob->piob_file, (OVERLAPPED*) &piob->piob_io_event);
	return true;
}
#endif


bool PIO_write(jrd_file* file, BufferDesc* bdb, Ods::pag* page, ISC_STATUS* status_vector)
{
/**************************************
 *
 *	P I O _ w r i t e
 *
 **************************************
 *
 * Functional description
 *	Write a data page.
 *
 **************************************/
	OVERLAPPED overlapped, *overlapped_ptr;

	Database* const dbb = bdb->bdb_dbb;
	const DWORD size = dbb->dbb_page_size;

	Database::Checkout dcoHolder(dbb);
	FileExtendLockGuard extLock(file->fil_ext_lock, false);

	file = seek_file(file, bdb, status_vector, &overlapped, &overlapped_ptr);
	if (!file) {
		return false;
	}

	HANDLE desc = file->fil_desc;

	if (dbb->dbb_encrypt_key.hasData())
	{
		SLONG spare_buffer[MAX_PAGE_SIZE / sizeof(SLONG)];

		(*dbb->dbb_encrypt) (dbb->dbb_encrypt_key.c_str(), page, size, spare_buffer);

		DWORD actual_length;
		if (!WriteFile(desc, spare_buffer, size, &actual_length, overlapped_ptr) ||
			actual_length != size)
		{
			return nt_error("WriteFile", file, isc_io_write_err, status_vector);
		}
	}
	else
	{
		DWORD actual_length;
		if (!WriteFile(desc, page, size, &actual_length, overlapped_ptr) || actual_length != size )
		{
#ifdef SUPERSERVER_V2
			if (!GetOverlappedResult(desc, overlapped_ptr, &actual_length, TRUE) ||
				actual_length != size)
			{
				release_io_event(file, overlapped_ptr);
				return nt_error("GetOverlappedResult", file, isc_io_write_err, status_vector);
			}
#else
			return nt_error("WriteFile", file, isc_io_write_err, status_vector);
#endif
		}
	}

#ifdef SUPERSERVER_V2
	release_io_event(file, overlapped_ptr);
#endif

	return true;
}


ULONG PIO_get_number_of_pages(const jrd_file* file, const USHORT pagesize)
{
/**************************************
 *
 *	P I O _ g e t _ n u m b e r _ o f _ p a g e s
 *
 **************************************
 *
 * Functional description
 *	Compute number of pages in file, based only on file size.
 *
 **************************************/
	HANDLE hFile = file->fil_desc;

	DWORD dwFileSizeHigh;
	const DWORD dwFileSizeLow = GetFileSize(hFile, &dwFileSizeHigh);

	if (dwFileSizeLow == (DWORD) -1) {
		nt_error("GetFileSize", file, isc_io_access_err, 0);
	}

    const ULONGLONG ullFileSize = (((ULONGLONG) dwFileSizeHigh) << 32) + dwFileSizeLow;
	return (ULONG) ((ullFileSize + pagesize - 1) / pagesize);
}


void PIO_get_unique_file_id(const Jrd::jrd_file* file, Firebird::UCharBuffer& id)
{
/**************************************
 *
 *	P I O _ g e t _ u n i q u e _ f i l e _ i d
 *
 **************************************
 *
 * Functional description
 *	Return a binary string that uniquely identifies the file.
 *
 **************************************/
	BY_HANDLE_FILE_INFORMATION file_info;
	GetFileInformationByHandle(file->fil_desc, &file_info);

	// The identifier is [nFileIndexHigh, nFileIndexLow]
	// MSDN says: After a process opens a file, the identifier is constant until
	// the file is closed. An application can use this identifier and the
	// volume serial number to determine whether two handles refer to the same file.
	const size_t len1 = sizeof(file_info.dwVolumeSerialNumber);
	const size_t len2 = sizeof(file_info.nFileIndexHigh);
	const size_t len3 = sizeof(file_info.nFileIndexLow);

	UCHAR* p = id.getBuffer(len1 + len2 + len3);

	memcpy(p, &file_info.dwVolumeSerialNumber, len1);
	p += len1;
	memcpy(p, &file_info.nFileIndexHigh, len2);
	p += len2;
	memcpy(p, &file_info.nFileIndexLow, len3);
}


#ifdef SUPERSERVER_V2
static void release_io_event(jrd_file* file, OVERLAPPED* overlapped)
{
/**************************************
 *
 *	r e l e a s e _ i o _ e v e n t
 *
 **************************************
 *
 * Functional description
 *	Release an overlapped I/O event
 *	back to the file block.
 *
 **************************************/
	if (!overlapped || !overlapped->hEvent)
		return;

	Firebird::MutexLockGuard guard(file->fil_mutex);

	for (int i = 0; i < MAX_FILE_IO; i++)
	{
		if (!file->fil_io_events[i])
		{
			file->fil_io_events[i] = overlapped->hEvent;
			overlapped->hEvent = NULL;
			break;
		}
	}

	if (overlapped->hEvent)
		CloseHandle(overlapped->hEvent);
}
#endif


static jrd_file* seek_file(jrd_file*	file,
					 	   BufferDesc*	bdb,
					 	   ISC_STATUS*	/*status_vector*/,
					 	   OVERLAPPED*	overlapped,
					 	   OVERLAPPED**	overlapped_ptr)
{
/**************************************
 *
 *	s e e k _ f i l e
 *
 **************************************
 *
 * Functional description
 *	Given a buffer descriptor block, find the appropriate
 *	file block and seek to the proper page in that file.
 *
 **************************************/
	Database* const dbb = bdb->bdb_dbb;
	ULONG page = bdb->bdb_page.getPageNum();

	for (;; file = file->fil_next)
	{
		if (!file) {
			CORRUPT(158);		// msg 158 database file not available
		}
		else if (page >= file->fil_min_page && page <= file->fil_max_page) {
			break;
		}
	}

	page -= file->fil_min_page - file->fil_fudge;

    LARGE_INTEGER liOffset;
	liOffset.QuadPart = UInt32x32To64((DWORD) page, (DWORD) dbb->dbb_page_size);

	overlapped->Offset = liOffset.LowPart;
	overlapped->OffsetHigh = liOffset.HighPart;
	overlapped->Internal = 0;
	overlapped->InternalHigh = 0;
	overlapped->hEvent = (HANDLE) 0;

	*overlapped_ptr = overlapped;

	fb_assert(dbb == bdb->bdb_dbb);

#ifdef SUPERSERVER_V2
	Firebird::MutexLockGuard guard(file->fil_mutex);

	for (USHORT i = 0; i < MAX_FILE_IO; i++)
	{
		if (overlapped->hEvent = (HANDLE) file->fil_io_events[i])
		{
			file->fil_io_events[i] = 0;
			break;
		}
	}

	if (!overlapped->hEvent && !(overlapped->hEvent = CreateEvent(NULL, TRUE, FALSE, NULL)))
	{
		nt_error("CreateEvent", file, isc_io_access_err, status_vector);
		return 0;
	}

	ResetEvent(overlapped->hEvent);
#endif

	return file;
}


static jrd_file* setup_file(Database* dbb,
							const Firebird::PathName& file_name,
							HANDLE desc,
							bool read_only)
{
/**************************************
 *
 *	s e t u p _ f i l e
 *
 **************************************
 *
 * Functional description
 *	Set up file and lock blocks for a file.
 *
 **************************************/
	jrd_file* file = NULL;

	try
	{
		file = FB_NEW_RPT(*dbb->dbb_permanent, file_name.length() + 1) jrd_file();
		file->fil_desc = desc;
		file->fil_max_page = MAX_ULONG;
		strcpy(file->fil_string, file_name.c_str());
#ifdef SUPERSERVER_V2
		memset(file->fil_io_events, 0, MAX_FILE_IO * sizeof(void*));
#endif

		if (read_only)
			file->fil_flags |= FIL_readonly;

		// If this isn't the primary file, we're done

		const PageSpace* const pageSpace = dbb->dbb_page_manager.findPageSpace(DB_PAGE_SPACE);
		if (pageSpace && pageSpace->file)
			return file;

#ifdef WIN9X_SUPPORT
		// Disable sophisticated file extension when running on 9X
		if (ISC_is_WinNT())
#endif
		{
			file->fil_ext_lock = FB_NEW(*dbb->dbb_permanent) Firebird::RWLock();
		}
	}
	catch (const Firebird::Exception&)
	{
		CloseHandle(desc);
		delete file;
		throw;
	}

	fb_assert(file);
	return file;
}

static bool maybeCloseFile(HANDLE& hFile)
{
/**************************************
 *
 *	M a y b e C l o s e F i l e
 *
 **************************************
 *
 * Functional description
 *	If the file is open, close it.
 *
 **************************************/

	if (hFile != INVALID_HANDLE_VALUE)
	{
		CloseHandle(hFile);
		hFile = INVALID_HANDLE_VALUE;
		return true;
	}
	return false;
}

static bool nt_error(const TEXT* string,
					 const jrd_file* file, ISC_STATUS operation,
					 ISC_STATUS* const status_vector)
{
/**************************************
 *
 *	n t _ e r r o r
 *
 **************************************
 *
 * Functional description
 *	Somebody has noticed a file system error and expects error
 *	to do something about it.  Harumph!
 *
 **************************************/
	const DWORD lastError = GetLastError();
	Arg::StatusVector status;
	status << Arg::Gds(isc_io_error) << Arg::Str(string) << Arg::Str(file->fil_string) <<
			  Arg::Gds(operation);
	if (lastError != ERROR_SUCCESS)
		status << Arg::Windows(lastError);

	if (!status_vector)
		ERR_post(status);

	ERR_build_status(status_vector, status);
	gds__log_status(0, status_vector);

	return false;
}

// These are defined in Windows Server 2008 SDK
#ifndef FILE_CACHE_FLAGS_DEFINED
#define FILE_CACHE_MAX_HARD_ENABLE      0x00000001
#define FILE_CACHE_MAX_HARD_DISABLE     0x00000002
#define FILE_CACHE_MIN_HARD_ENABLE      0x00000004
#define FILE_CACHE_MIN_HARD_DISABLE     0x00000008
#endif // FILE_CACHE_FLAGS_DEFINED

BOOL SetPrivilege(
	HANDLE hToken,			// access token handle
	LPCTSTR lpszPrivilege,	// name of privilege to enable/disable
	BOOL bEnablePrivilege)	// to enable or disable privilege
{
	TOKEN_PRIVILEGES tp;
	LUID luid;

	if (!LookupPrivilegeValue(
			NULL,			// lookup privilege on local system
			lpszPrivilege,	// privilege to lookup
			&luid))			// receives LUID of privilege
	{
		// gds__log("LookupPrivilegeValue error: %u", GetLastError() );
		return FALSE;
	}

	tp.PrivilegeCount = 1;
	tp.Privileges[0].Luid = luid;
	if (bEnablePrivilege)
		tp.Privileges[0].Attributes = SE_PRIVILEGE_ENABLED;
	else
		tp.Privileges[0].Attributes = 0;

	// Enable or disable the privilege

	if (!AdjustTokenPrivileges(hToken, FALSE, &tp, sizeof(TOKEN_PRIVILEGES),
			(PTOKEN_PRIVILEGES) NULL, (PDWORD) NULL))
	{
		//gds__log("AdjustTokenPrivileges error: %u", GetLastError() );
		return FALSE;
	}

	if (GetLastError() == ERROR_NOT_ALL_ASSIGNED)
	{
		//gds__log("The token does not have the specified privilege");
		return FALSE;
	}

	return TRUE;
}

static void adjustFileSystemCacheSize()
{
	int percent = Config::getFileSystemCacheSize();

	// firebird.conf asks to do nothing
	if (percent == 0)
		return;

	// Ensure that the setting has a sensible value
	if (percent > 95 || percent < 10)
	{
		gds__log("Incorrect FileSystemCacheSize setting %d. Using default (30 percent).", percent);
		percent = 30;
	}

	HMODULE hmodKernel32 = GetModuleHandle("kernel32.dll");

	// This one requires 64-bit XP or Windows Server 2003 SP1
	typedef BOOL (WINAPI *PFnSetSystemFileCacheSize)(SIZE_T, SIZE_T, DWORD);

	typedef BOOL (WINAPI *PFnGetSystemFileCacheSize)(PSIZE_T, PSIZE_T, PDWORD);

	// This one needs any NT, but load it dynamically anyways just in case
	typedef BOOL (WINAPI *PFnGlobalMemoryStatusEx)(LPMEMORYSTATUSEX);

	PFnSetSystemFileCacheSize pfnSetSystemFileCacheSize =
		(PFnSetSystemFileCacheSize) GetProcAddress(hmodKernel32, "SetSystemFileCacheSize");
	PFnGetSystemFileCacheSize pfnGetSystemFileCacheSize =
		(PFnGetSystemFileCacheSize) GetProcAddress(hmodKernel32, "GetSystemFileCacheSize");
	PFnGlobalMemoryStatusEx pfnGlobalMemoryStatusEx =
		(PFnGlobalMemoryStatusEx) GetProcAddress(hmodKernel32, "GlobalMemoryStatusEx");

	// If we got too old OS and functions are not there - do not bother
	if (!pfnGetSystemFileCacheSize || !pfnSetSystemFileCacheSize || !pfnGlobalMemoryStatusEx)
		return;

	MEMORYSTATUSEX msex;
	msex.dwLength = sizeof(msex);

	// This should work
	if (!pfnGlobalMemoryStatusEx(&msex))
		system_call_failed::raise("GlobalMemoryStatusEx", GetLastError());

	SIZE_T origMinimumFileCacheSize, origMaximumFileCacheSize;
	DWORD origFlags;

	BOOL result = pfnGetSystemFileCacheSize(&origMinimumFileCacheSize,
		&origMaximumFileCacheSize, &origFlags);

	if (!result)
	{
		const DWORD error = GetLastError();
#ifndef _WIN64
		// This error is returned on 64-bit Windows when the file cache size
		// overflows the ULONG limit restricted by the 32-bit Windows API.
		// Let's avoid writing it into the log as it's not a critical failure.
		if (error != ERROR_ARITHMETIC_OVERFLOW)
#endif
		gds__log("GetSystemFileCacheSize error %d", error);
		return;
	}

	// Somebody has already configured maximum cache size limit
	// Hope it is a sensible one - do not bother to adjust it
	if ((origFlags & FILE_CACHE_MAX_HARD_ENABLE) != 0)
		return;

	DWORDLONG maxMem = (msex.ullTotalPhys / 100) * percent;

#ifndef _WIN64
	// If we are trying to set the limit so high that it doesn't fit
	// in 32-bit API - leave settings alone and write a message to log file
	if (maxMem > (SIZE_T)(-2))
	{
		gds__log("Could not use 32-bit SetSystemFileCacheSize API to set cache size limit to %I64d."
				" Please use 64-bit engine or configure cache size limit externally", maxMem);
		return;
	}
#endif

	HANDLE hToken;
	if (!OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES, &hToken))
	{
		gds__log("OpenProcessToken error %d", GetLastError());
		return;
	}

	SetPrivilege(hToken, "SeIncreaseQuotaPrivilege", TRUE);
	result = pfnSetSystemFileCacheSize(0, maxMem, FILE_CACHE_MAX_HARD_ENABLE);
	const DWORD error = GetLastError();
	SetPrivilege(hToken, "SeIncreaseQuotaPrivilege", FALSE);
	CloseHandle(hToken);

	if (!result)
	{
		// If we do not have enough permissions - silently ignore the error
		gds__log("SetSystemFileCacheSize error %d. "
			"The engine will continue to operate, but the system "
			"performance may degrade significantly when working with "
			"large databases", error);
	}
}