FIO Windows update

Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
diff --git a/HOWTO b/HOWTO
index 3b1eeb0..35dccfb 100644
--- a/HOWTO
+++ b/HOWTO
@@ -272,11 +272,11 @@
 		':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
 		as the two working files, you would use
 		filename=/dev/sda:/dev/sdb. On Windows, disk devices are accessed
-		as /dev/sda for the first device (i.e. \Device\HardDisk0\Partition0,
-		/dev/sda1 for the first partition on the first disk etc.  If the 
-		wanted filename does need to include a colon, then escape that with 
-		a '\' character. For instance, if the filename is 
-		"/dev/dsk/foo@3,0:c", then you would use filename="/dev/dsk/foo@3,0\:c". 
+		as \\.\PhysicalDrive0 for the first device, \\.\PhysicalDrive1
+		for the second etc.  If the wanted filename does need to 
+		include a colon, then escape that with a '\' character. 
+		For instance, if the filename is "/dev/dsk/foo@3,0:c", 
+		then you would use filename="/dev/dsk/foo@3,0\:c". 
 		'-' is a reserved name, meaning stdin or stdout. Which of the 
 		two depends on the read/write direction set.
 
diff --git a/README b/README
index f1ee9aa..f136bb5 100644
--- a/README
+++ b/README
@@ -54,6 +54,10 @@
 tool (http://www.opencsw.org/get-it/pkgutil/) and then install fio via
 'pkgutil -i fio'.
 
+Windows:
+Bruce Cran <bruce@cran.org.uk> has fio packages for Windows at
+http://www.bluestop.org/fio .
+
 
 Mailing list
 ------------
@@ -79,10 +83,10 @@
 Building
 --------
 
-Just type 'make' and 'make install'. If on FreeBSD, for now you have to
-specify the FreeBSD Makefile with -f and use gmake (not make), eg:
+Just type 'make' and 'make install'. If on BSD, for now you have to
+specify the BSD Makefile with -f and use gmake (not make), eg:
 
-$ gmake -f Makefile.Freebsd && gmake -f Makefile.FreeBSD install
+$ gmake -f Makefile.FreeBSD && gmake -f Makefile.FreeBSD install
 
 Same goes for AIX:
 
@@ -132,7 +136,7 @@
 each job file will be regarded as a separate group and fio will stonewall
 its execution.
 
-The --readonly switch is an extra safety guard to prevent accidentically
+The --readonly switch is an extra safety guard to prevent accidentally
 turning on a write setting when that is not desired. Fio will only write
 if rw=write/randwrite/rw/randrw is given, but this extra safety net can
 be used as an extra precaution. It will also enable a write check in the
@@ -294,8 +298,8 @@
 Platforms
 ---------
 
-Fio works on (at least) Linux, Solaris, AIX, OSX, NetBSD, and FreeBSD. Some
-features and/or options may only be available on some of the platforms,
+Fio works on (at least) Linux, Solaris, AIX, OSX, NetBSD, Windows and FreeBSD.
+Some features and/or options may only be available on some of the platforms,
 typically because those features only apply to that platform (like the
 solarisaio engine, or the splice engine on Linux).
 
diff --git a/engines/windowsaio.c b/engines/windowsaio.c
new file mode 100644
index 0000000..5c9d85e
--- /dev/null
+++ b/engines/windowsaio.c
@@ -0,0 +1,487 @@
+/*
+ * Native Windows async IO engine
+ * Copyright (C) 2010 Bruce Cran <bruce@cran.org.uk>
+ */
+
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <signal.h>
+#include <errno.h>
+#include <windows.h>
+
+#include "../fio.h"
+
+BOOL windowsaio_debug = FALSE;
+
+struct windowsaio_data {
+	struct io_u **aio_events;
+	unsigned int ioFinished;
+	BOOL running;
+	BOOL stopped;
+	HANDLE hThread;
+};
+
+typedef struct {
+ OVERLAPPED o;
+ struct io_u *io_u;
+} FIO_OVERLAPPED;
+
+struct thread_ctx {
+	HANDLE ioCP;
+	struct windowsaio_data *wd;
+};
+
+static void PrintError(LPCSTR lpszFunction);
+static int fio_windowsaio_cancel(struct thread_data *td,
+			       struct io_u *io_u);
+static DWORD GetEndCount(DWORD startCount, struct timespec *t);
+static BOOL TimedOut(DWORD startCount, DWORD endCount);
+static int fio_windowsaio_getevents(struct thread_data *td, unsigned int min,
+				    unsigned int max, struct timespec *t);
+static struct io_u *fio_windowsaio_event(struct thread_data *td, int event);
+static int fio_windowsaio_queue(struct thread_data *td,
+			      struct io_u *io_u);
+static void fio_windowsaio_cleanup(struct thread_data *td);
+static DWORD WINAPI IoCompletionRoutine(LPVOID lpParameter);
+static int fio_windowsaio_init(struct thread_data *td);
+static int fio_windowsaio_open_file(struct thread_data *td, struct fio_file *f);
+static int fio_windowsaio_close_file(struct thread_data fio_unused *td, struct fio_file *f);
+
+static void PrintError(LPCSTR lpszFunction)
+{
+	// Retrieve the system error message for the last-error code
+
+	LPSTR lpMsgBuf;
+	DWORD dw = GetLastError();
+
+	FormatMessage(
+		FORMAT_MESSAGE_ALLOCATE_BUFFER |
+		FORMAT_MESSAGE_FROM_SYSTEM |
+		FORMAT_MESSAGE_IGNORE_INSERTS,
+		NULL,
+		dw,
+		MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT),
+		(LPTSTR)&lpMsgBuf,
+		0, NULL );
+
+	log_err("%s - %s", lpszFunction, lpMsgBuf);
+	LocalFree(lpMsgBuf);
+}
+
+static int fio_windowsaio_cancel(struct thread_data *td,
+			       struct io_u *io_u)
+{
+	BOOL bSuccess;
+	int rc = 0;
+
+	bSuccess = CancelIo(io_u->file->hFile);
+
+	if (!bSuccess)
+		rc = 1;
+
+	return rc;
+}
+
+static DWORD GetEndCount(DWORD startCount, struct timespec *t)
+{
+	DWORD endCount = startCount;
+
+	if (t == NULL)
+		return 0;
+
+	endCount += (t->tv_sec * 1000) + (t->tv_nsec / 1000000);
+	return endCount;
+}
+
+static BOOL TimedOut(DWORD startCount, DWORD endCount)
+{
+	BOOL expired = FALSE;
+	DWORD currentTime;
+
+	if (startCount == 0 || endCount == 0)
+		return FALSE;
+
+	currentTime = GetTickCount();
+
+	if ((endCount > startCount) && currentTime >= endCount)
+		expired = TRUE;
+	else if (currentTime < startCount && currentTime > endCount)
+		expired = TRUE;
+
+	if (windowsaio_debug)
+		printf("windowsaio: timedout = %d\n", expired);
+
+	return expired;
+}
+
+static int fio_windowsaio_getevents(struct thread_data *td, unsigned int min,
+				    unsigned int max, struct timespec *t)
+{
+	struct windowsaio_data *wd = td->io_ops->data;
+	struct flist_head *entry;
+	unsigned int dequeued = 0;
+	struct io_u *io_u;
+	DWORD startCount = 0, endCount = 0;
+	BOOL timedout = FALSE;
+	unsigned int r = 0;
+
+	if (windowsaio_debug)
+		printf("getevents (min %d, max %d)\n", min, max);
+
+	if (t != NULL) {
+		startCount = GetTickCount();
+		endCount = GetEndCount(startCount, t);
+	}
+
+	while (dequeued < min && !timedout) {
+
+		flist_for_each(entry, &td->io_u_busylist) {
+			io_u = flist_entry(entry, struct io_u, list);
+
+			if (io_u->seen == 0)
+				continue;
+
+			dequeued++;
+
+			wd->ioFinished--;
+			wd->aio_events[r] = io_u;
+			r++;
+
+			if (windowsaio_debug)
+				printf("dequeued %d\n", dequeued);
+
+			if (dequeued == max)
+				break;
+		}
+
+		if (TimedOut(startCount, endCount))
+			timedout = TRUE;
+
+		if (dequeued < min && !timedout)
+			Sleep(250);
+	}
+
+	if (windowsaio_debug)
+		printf("leave getevents (%d)\n", dequeued);
+
+	return dequeued;
+}
+
+static struct io_u *fio_windowsaio_event(struct thread_data *td, int event)
+{
+	struct windowsaio_data *wd = td->io_ops->data;
+	return wd->aio_events[event];
+}
+
+static int fio_windowsaio_queue(struct thread_data *td,
+			      struct io_u *io_u)
+{
+	FIO_OVERLAPPED *fov;
+	DWORD ioBytes;
+	BOOL bSuccess = TRUE;
+	int rc;
+
+	fio_ro_check(td, io_u);
+
+	if (windowsaio_debug)
+		printf("enqueue enter\n");
+
+	fov = malloc(sizeof(FIO_OVERLAPPED));
+	ZeroMemory(fov, sizeof(FIO_OVERLAPPED));
+
+	io_u->seen = 0;
+
+	fov->o.Offset = io_u->offset & 0xFFFFFFFF;
+	fov->o.OffsetHigh = io_u->offset >> 32;
+	fov->o.hEvent = CreateEvent(NULL, FALSE, FALSE, NULL);
+	fov->io_u = io_u;
+
+	if (fov->o.hEvent == NULL) {
+		PrintError(__func__);
+		return 1;
+	}
+
+	if (io_u->ddir == DDIR_WRITE)
+		bSuccess = WriteFile(io_u->file->hFile, io_u->xfer_buf, io_u->xfer_buflen, &ioBytes, &fov->o);
+	else if (io_u->ddir == DDIR_READ)
+		bSuccess = ReadFile(io_u->file->hFile, io_u->xfer_buf, io_u->xfer_buflen, &ioBytes, &fov->o);
+	else if (io_u->ddir == DDIR_SYNC     ||
+			 io_u->ddir == DDIR_DATASYNC ||
+			 io_u->ddir == DDIR_SYNC_FILE_RANGE)
+	{
+		FlushFileBuffers(io_u->file->hFile);
+		return FIO_Q_COMPLETED;
+	} else if (io_u->ddir == DDIR_TRIM) {
+		log_info("explicit TRIM isn't supported on Windows");
+		return FIO_Q_COMPLETED;
+	}
+
+	if (bSuccess) {
+		io_u->seen = 1;
+		io_u->resid = io_u->xfer_buflen - fov->o.InternalHigh;
+		io_u->error = 0;
+		rc = FIO_Q_COMPLETED;
+	} else if (!bSuccess && GetLastError() == ERROR_IO_PENDING) {
+		rc = FIO_Q_QUEUED;
+	} else {
+		PrintError(__func__);
+		io_u->error = GetLastError();
+		io_u->resid = io_u->xfer_buflen;
+		rc = FIO_Q_COMPLETED;
+	}
+
+	if (windowsaio_debug)
+		printf("enqueue - leave (offset %llu)\n", io_u->offset);
+
+	return rc;
+}
+
+static void fio_windowsaio_cleanup(struct thread_data *td)
+{
+	struct windowsaio_data *wd;
+
+	if (windowsaio_debug)
+		printf("windowsaio: cleanup - enter\n");
+
+	wd = td->io_ops->data;
+	wd->running = FALSE;
+
+	while (wd->stopped == FALSE)
+		Sleep(5);
+
+	if (wd != NULL) {
+		CloseHandle(wd->hThread);
+		free(wd->aio_events);
+		wd->aio_events = NULL;
+		free(wd);
+		td->io_ops->data = NULL;
+	}
+
+	if (windowsaio_debug)
+		printf("windowsaio: cleanup - leave\n");
+}
+
+static DWORD WINAPI IoCompletionRoutine(LPVOID lpParameter)
+{
+	OVERLAPPED *ovl;
+	FIO_OVERLAPPED *fov;
+	struct io_u *io_u;
+	struct windowsaio_data *wd;
+
+	struct thread_ctx *ctx;
+	ULONG_PTR ulKey = 0;
+	BOOL bSuccess;
+	DWORD bytes;
+
+
+	ctx = (struct thread_ctx*)lpParameter;
+	wd = ctx->wd;
+	bSuccess = TRUE;
+
+	if (windowsaio_debug)
+		printf("windowsaio: IoCompletionRoutine - enter\n");
+
+	while (ctx->wd->running) {
+		bSuccess = GetQueuedCompletionStatus(ctx->ioCP, &bytes, &ulKey, &ovl, 500);
+
+		if (windowsaio_debug)
+			printf("GetQueuedCompletionStatus returned %d\n", bSuccess);
+
+		if (!bSuccess) {
+			if (GetLastError() == WAIT_TIMEOUT) {
+				continue;
+			} else {
+				PrintError(__func__);
+				continue;
+			}
+		}
+
+		fov = CONTAINING_RECORD(ovl, FIO_OVERLAPPED, o);
+		io_u = fov->io_u;
+
+		if (windowsaio_debug) {
+			if (io_u->seen == 1)
+				printf("IoCompletionRoutine - got already completed IO\n");
+			else
+				printf("IoCompletionRoutine - completed %d IO\n", ctx->wd->ioFinished);
+		}
+
+		if (io_u->seen == 1)
+			continue;
+
+		ctx->wd->ioFinished++;
+
+		if (ovl->Internal == ERROR_SUCCESS) {
+			io_u->resid = io_u->xfer_buflen - ovl->InternalHigh;
+			io_u->error = 0;
+		} else {
+			io_u->resid = io_u->xfer_buflen;
+			io_u->error = 1;
+		}
+
+		io_u->seen = 1;
+		CloseHandle(ovl->hEvent);
+		free(ovl);
+	}
+
+	bSuccess = CloseHandle(ctx->ioCP);
+	if (!bSuccess)
+		PrintError(__func__);
+
+	if (windowsaio_debug)
+		printf("windowsaio: IoCompletionRoutine - leave\n");
+
+	ctx->wd->stopped = TRUE;
+	free(ctx);
+	return 0;
+}
+
+static int fio_windowsaio_init(struct thread_data *td)
+{
+	int rc = 0;
+	struct windowsaio_data *wd;
+
+	if (windowsaio_debug)
+		printf("windowsaio: init\n");
+
+	wd = malloc(sizeof(struct windowsaio_data));
+
+	ZeroMemory(wd, sizeof(*wd));
+	wd->aio_events = malloc((td->o.iodepth + 1) * sizeof(struct io_u *));
+	ZeroMemory(wd->aio_events, (td->o.iodepth + 1) * sizeof(struct io_u *));
+
+	td->io_ops->data = wd;
+	return rc;
+}
+
+static int fio_windowsaio_open_file(struct thread_data *td, struct fio_file *f)
+{
+	int rc = 0;
+	HANDLE hFile;
+	DWORD flags = FILE_ATTRIBUTE_TEMPORARY | FILE_FLAG_POSIX_SEMANTICS | FILE_FLAG_OVERLAPPED;
+	DWORD sharemode = FILE_SHARE_READ | FILE_SHARE_WRITE;
+	DWORD openmode = OPEN_ALWAYS;
+	DWORD access;
+
+	dprint(FD_FILE, "fd open %s\n", f->file_name);
+
+	if (windowsaio_debug)
+		printf("windowsaio: open file %s - enter\n", f->file_name);
+
+	if (f->filetype == FIO_TYPE_PIPE) {
+		log_err("fio: windowsaio doesn't support pipes\n");
+		return 1;
+	}
+
+	if (!strcmp(f->file_name, "-")) {
+		log_err("fio: can't read/write to stdin/out\n");
+		return 1;
+	}
+
+	if (td->o.odirect)
+		flags |= FILE_FLAG_NO_BUFFERING;
+	if (td->o.sync_io)
+		flags |= FILE_FLAG_WRITE_THROUGH;
+
+
+	if (td->o.td_ddir == TD_DDIR_READ  ||
+		td->o.td_ddir == TD_DDIR_WRITE ||
+		td->o.td_ddir == TD_DDIR_RANDRW)
+	{
+		flags |= FILE_FLAG_SEQUENTIAL_SCAN;
+	}
+	else
+	{
+		flags |= FILE_FLAG_RANDOM_ACCESS;
+	}
+
+	if (td_read(td) || read_only)
+		access = GENERIC_READ;
+	else
+		access = (GENERIC_READ | GENERIC_WRITE);
+
+	if (td->o.create_on_open > 0)
+		openmode = OPEN_ALWAYS;
+	else
+		openmode = OPEN_EXISTING;
+
+	f->hFile = CreateFile(f->file_name, access, sharemode,
+		NULL, openmode, flags, NULL);
+
+	if (f->hFile == INVALID_HANDLE_VALUE) {
+		log_err("Failed to open %s\n", f->file_name);
+		PrintError(__func__);
+		rc = 1;
+	}
+
+	/* Only set up the competion port and thread if we're not just
+	 * querying the device size */
+	if (!rc && td->io_ops->data != NULL) {
+		struct windowsaio_data *wd;
+		struct thread_ctx *ctx;
+		hFile = CreateIoCompletionPort(f->hFile, NULL, 0, 0);
+
+		wd = td->io_ops->data;
+		wd->running = TRUE;
+		wd->stopped = FALSE;
+
+		ctx = malloc(sizeof(struct thread_ctx));
+		ctx->ioCP = hFile;
+		ctx->wd = wd;
+
+		wd->hThread = CreateThread(NULL, 0, IoCompletionRoutine, ctx, 0, NULL);
+
+		if (wd->hThread == NULL) {
+			PrintError(__func__);
+			rc = 1;
+		}
+	}
+
+	if (windowsaio_debug)
+		printf("windowsaio: open file - leave (%d)\n", rc);
+
+	return rc;
+}
+
+static int fio_windowsaio_close_file(struct thread_data fio_unused *td, struct fio_file *f)
+{
+	BOOL bSuccess;
+
+	if (windowsaio_debug)
+		printf("windowsaio: close file\n");
+
+	if (f->hFile != INVALID_HANDLE_VALUE) {
+		bSuccess = CloseHandle(f->hFile);
+		if (!bSuccess)
+			PrintError(__func__);
+	}
+
+	f->hFile = INVALID_HANDLE_VALUE;
+	return 0;
+}
+
+static struct ioengine_ops ioengine = {
+	.name		= "windowsaio",
+	.version	= FIO_IOOPS_VERSION,
+	.init		= fio_windowsaio_init,
+	.queue		= fio_windowsaio_queue,
+	.cancel		= fio_windowsaio_cancel,
+	.getevents	= fio_windowsaio_getevents,
+	.event		= fio_windowsaio_event,
+	.cleanup	= fio_windowsaio_cleanup,
+	.open_file	= fio_windowsaio_open_file,
+	.close_file	= fio_windowsaio_close_file,
+	.get_file_size	= generic_get_file_size
+};
+
+static void fio_init fio_posixaio_register(void)
+{
+	register_ioengine(&ioengine);
+}
+
+static void fio_exit fio_posixaio_unregister(void)
+{
+	unregister_ioengine(&ioengine);
+}
diff --git a/file.h b/file.h
index 994d5fd..10c5379 100644
--- a/file.h
+++ b/file.h
@@ -1,7 +1,9 @@
 #ifndef FIO_FILE_H
 #define FIO_FILE_H
 
+#include "compiler/compiler.h"
 #include "io_ddir.h"
+#include "flist.h"
 
 /*
  * The type of object we are working on
diff --git a/filesetup.c b/filesetup.c
index 5b96c66..9d42deb 100644
--- a/filesetup.c
+++ b/filesetup.c
@@ -11,6 +11,7 @@
 #include "fio.h"
 #include "smalloc.h"
 #include "filehash.h"
+#include "os/os.h"
 
 static int root_warn;
 
@@ -241,7 +242,7 @@
 		return 1;
 	}
 
-	r = blockdev_size(f->fd, &bytes);
+	r = blockdev_size(f, &bytes);
 	if (r) {
 		td_verror(td, r, "blockdev_size");
 		printf("fd is %d\n", f->fd);
@@ -273,7 +274,7 @@
 		return 1;
 	}
 
-	r = chardev_size(f->fd, &bytes);
+	r = chardev_size(f, &bytes);
 	if (r) {
 		td_verror(td, r, "chardev_size");
 		goto err;
@@ -351,9 +352,9 @@
 		(void) posix_madvise(f->mmap_ptr, f->mmap_sz, FIO_MADV_FREE);
 #endif
 	} else if (f->filetype == FIO_TYPE_FILE) {
-		ret = fadvise(f->fd, off, len, POSIX_FADV_DONTNEED);
+		ret = posix_fadvise(f->fd, off, len, POSIX_FADV_DONTNEED);
 	} else if (f->filetype == FIO_TYPE_BD) {
-		ret = blockdev_invalidate_cache(f->fd);
+		ret = blockdev_invalidate_cache(f);
 		if (ret < 0 && errno == EACCES && geteuid()) {
 			if (!root_warn) {
 				log_err("fio: only root may flush block "
@@ -885,7 +886,9 @@
 		f->filetype = FIO_TYPE_FILE;
 
 	if (!stat(f->file_name, &sb)) {
-		if (S_ISBLK(sb.st_mode))
+		/* \\.\ is the device namespace in Windows, where every file is
+		 * a block device */
+		if (S_ISBLK(sb.st_mode) || strncmp(f->file_name, "\\\\.\\", 4) == 0)
 			f->filetype = FIO_TYPE_BD;
 		else if (S_ISCHR(sb.st_mode))
 			f->filetype = FIO_TYPE_CHAR;
diff --git a/fio.c b/fio.c
index 2420d10..93482f5 100644
--- a/fio.c
+++ b/fio.c
@@ -161,13 +161,8 @@
 	evt.sigev_notify = SIGEV_THREAD;
 	evt.sigev_notify_function = ival_fn;
 
-#ifndef __CYGWIN__
-	if (timer_create(CLOCK_MONOTONIC, &evt, &ival_timer) < 0)
+	if (timer_create(FIO_TIMER_CLOCK, &evt, &ival_timer) < 0)
 		perror("timer_create");
-#else /* Windows (and thus Cygwin) doesn't have a monotonic clock */
-	if (timer_create(CLOCK_REALTIME, &evt, &ival_timer) < 0)
-		perror("timer_create");
-#endif
 
 }
 
diff --git a/fio.h b/fio.h
index c04ad8b..9598858 100644
--- a/fio.h
+++ b/fio.h
@@ -14,6 +14,8 @@
 #include <inttypes.h>
 #include <assert.h>
 
+struct thread_data;
+
 #include "compiler/compiler.h"
 #include "flist.h"
 #include "fifo.h"
diff --git a/init.c b/init.c
index bdbb377..0a3529d 100644
--- a/init.c
+++ b/init.c
@@ -431,7 +431,9 @@
 	if (lstat(filename, &sb) == -1)
 		return 0;
 
-	if (S_ISREG(sb.st_mode))
+	/* \\.\ is the device namespace in Windows, where every file
+	 * is a device node */
+	if (S_ISREG(sb.st_mode) && strncmp(filename, "\\\\.\\", 4) != 0)
 		return 0;
 
 	return 1;
diff --git a/ioengines.c b/ioengines.c
index 35471ad..6b677cb 100644
--- a/ioengines.c
+++ b/ioengines.c
@@ -14,6 +14,7 @@
 #include <unistd.h>
 #include <string.h>
 #include <dlfcn.h>
+#include <fcntl.h>
 #include <assert.h>
 
 #include "fio.h"
@@ -386,7 +387,7 @@
 		else
 			flags = POSIX_FADV_SEQUENTIAL;
 
-		if (fadvise(f->fd, f->file_offset, f->io_size, flags) < 0) {
+		if (posix_fadvise(f->fd, f->file_offset, f->io_size, flags) < 0) {
 			td_verror(td, errno, "fadvise");
 			goto err;
 		}
diff --git a/os/os-aix.h b/os/os-aix.h
index 09a0530..b5ee5bb 100644
--- a/os/os-aix.h
+++ b/os/os-aix.h
@@ -9,6 +9,7 @@
 #define FIO_HAVE_POSIXAIO
 #define FIO_HAVE_ODIRECT
 #define FIO_USE_GENERIC_RAND
+#define FIO_HAVE_CLOCK_MONOTONIC
 
 /*
  * This is broken on AIX if _LARGE_FILES is defined...
@@ -22,16 +23,16 @@
 #define OS_MAP_ANON		MAP_ANON
 #define OS_MSG_DONTWAIT		0
 
-static inline int blockdev_invalidate_cache(int fd)
+static inline int blockdev_invalidate_cache(struct fio_file fio_unused *f)
 {
 	return EINVAL;
 }
 
-static inline int blockdev_size(int fd, unsigned long long *bytes)
+static inline int blockdev_size(struct fio_file *f, unsigned long long *bytes)
 {
 	struct devinfo info;
 
-	if (!ioctl(fd, IOCINFO, &info)) {
+	if (!ioctl(f->fd, IOCINFO, &info)) {
         	*bytes = (unsigned long long)info.un.scdk.numblks *
 				info.un.scdk.blksize;
 		return 0;
diff --git a/os/os-freebsd.h b/os/os-freebsd.h
index 4a50a3c..0cb7ae5 100644
--- a/os/os-freebsd.h
+++ b/os/os-freebsd.h
@@ -7,18 +7,21 @@
 
 #define FIO_HAVE_POSIXAIO
 #define FIO_HAVE_ODIRECT
+#define FIO_HAVE_IOPRIO
+#define FIO_HAVE_STRSEP
 #define FIO_USE_GENERIC_RAND
 #define FIO_HAVE_CHARDEV_SIZE
+#define FIO_HAVE_CLOCK_MONOTONIC
 
 #define OS_MAP_ANON		MAP_ANON
 
 typedef off_t off64_t;
 
-static inline int blockdev_size(int fd, unsigned long long *bytes)
+static inline int blockdev_size(struct fio_file *f, unsigned long long *bytes)
 {
 	off_t size;
 
-	if (!ioctl(fd, DIOCGMEDIASIZE, &size)) {
+	if (!ioctl(f->fd, DIOCGMEDIASIZE, &size)) {
 		*bytes = size;
 		return 0;
 	}
@@ -27,12 +30,12 @@
 	return errno;
 }
 
-static inline int chardev_size(int fd, unsigned long long *bytes)
+static inline int chardev_size(struct fio_file *f, unsigned long long *bytes)
 {
-	return blockdev_size(fd, bytes);
+	return blockdev_size(f->fd, bytes);
 }
 
-static inline int blockdev_invalidate_cache(int fd)
+static inline int blockdev_invalidate_cache(struct fio_file *f)
 {
 	return EINVAL;
 }
diff --git a/os/os-linux.h b/os/os-linux.h
index 20f2a94..aab74e5 100644
--- a/os/os-linux.h
+++ b/os/os-linux.h
@@ -39,6 +39,7 @@
 #define FIO_HAVE_FS_STAT
 #define FIO_HAVE_TRIM
 #define FIO_HAVE_BINJECT
+#define FIO_HAVE_CLOCK_MONOTONIC
 
 #ifdef SYNC_FILE_RANGE_WAIT_BEFORE
 #define FIO_HAVE_SYNC_FILE_RANGE
@@ -192,14 +193,14 @@
 #define BLKDISCARD	_IO(0x12,119)
 #endif
 
-static inline int blockdev_invalidate_cache(int fd)
+static inline int blockdev_invalidate_cache(struct fio_file *fd)
 {
-	return ioctl(fd, BLKFLSBUF);
+	return ioctl(f->fd, BLKFLSBUF);
 }
 
-static inline int blockdev_size(int fd, unsigned long long *bytes)
+static inline int blockdev_size(struct fio_file *f, unsigned long long *bytes)
 {
-	if (!ioctl(fd, BLKGETSIZE64, bytes))
+	if (!ioctl(f->fd, BLKGETSIZE64, bytes))
 		return 0;
 
 	return errno;
diff --git a/os/os-mac.h b/os/os-mac.h
index 1b786f8..39a2f82 100644
--- a/os/os-mac.h
+++ b/os/os-mac.h
@@ -13,6 +13,7 @@
 #endif
 
 #define FIO_HAVE_POSIXAIO
+#define FIO_HAVE_CLOCK_MONOTONIC
 #define FIO_USE_GENERIC_BDEV_SIZE
 #define FIO_USE_GENERIC_RAND
 
@@ -21,7 +22,7 @@
 typedef unsigned int clockid_t;
 typedef off_t off64_t;
 
-static inline int blockdev_invalidate_cache(int fd)
+static inline int blockdev_invalidate_cache(struct fio_file fio_unused *f)
 {
 	return EINVAL;
 }
diff --git a/os/os-netbsd.h b/os/os-netbsd.h
index 70d2958..ecfcddb 100644
--- a/os/os-netbsd.h
+++ b/os/os-netbsd.h
@@ -12,10 +12,10 @@
 
 #define FIO_HAVE_POSIXAIO
 #define FIO_HAVE_FADVISE
-#define	fadvise	posix_fadvise	/* XXX NetBSD doesn't have fadvise */
 #define FIO_HAVE_ODIRECT
 #define FIO_HAVE_STRSEP
 #define FIO_HAVE_FDATASYNC
+#define FIO_HAVE_CLOCK_MONOTONIC
 #define FIO_USE_GENERIC_BDEV_SIZE
 #define FIO_USE_GENERIC_RAND
 
@@ -25,7 +25,7 @@
 
 typedef off_t off64_t;
 
-static inline int blockdev_invalidate_cache(int fd)
+static inline int blockdev_invalidate_cache(struct fio_file fio_unused *f)
 {
 	return EINVAL;
 }
diff --git a/os/os-solaris.h b/os/os-solaris.h
index f8ce1f3..14164b0 100644
--- a/os/os-solaris.h
+++ b/os/os-solaris.h
@@ -15,6 +15,7 @@
 #define FIO_HAVE_PSHARED_MUTEX
 #define FIO_USE_GENERIC_BDEV_SIZE
 #define FIO_HAVE_FDATASYNC
+#define FIO_HAVE_CLOCK_MONOTONIC
 
 #define OS_MAP_ANON		MAP_ANON
 #define OS_RAND_MAX		2147483648UL
@@ -26,7 +27,7 @@
 typedef psetid_t os_cpu_mask_t;
 typedef struct solaris_rand_seed os_random_state_t;
 
-static inline int blockdev_invalidate_cache(int fd)
+static inline int blockdev_invalidate_cache(struct fio_file fio_unused *f)
 {
 	return EINVAL;
 }
diff --git a/os/os-windows.h b/os/os-windows.h
index f5da6a5..74c0f9e 100644
--- a/os/os-windows.h
+++ b/os/os-windows.h
@@ -4,7 +4,11 @@
 

 #include <sys/types.h>

 #include <errno.h>

+#include <windows.h>

 

+#include "../smalloc.h"

+#include "../file.h"

+#include "../log.h"

 

 #define FIO_HAVE_ODIRECT

 #define FIO_USE_GENERIC_RAND

@@ -15,73 +19,64 @@
 #define FIO_HAVE_FDATASYNC

 #define FIO_HAVE_WINDOWSAIO

 

-/* TODO add support for FIO_HAVE_CPU_AFFINITY */

-

 #define OS_MAP_ANON		MAP_ANON

 

+#define OS_CLOCK CLOCK_REALTIME

+

 typedef off_t off64_t;

 

-

-#define FIO_NOTUNIX

-

-#include <windows.h>

-#include <io.h>

-

-typedef void* HANDLE;

-

-BOOL WINAPI GetFileSizeEx(

-   HANDLE hFile,

-   PLARGE_INTEGER lpFileSize

-);

-

-long _get_osfhandle(

-   int fd

-);

-

 typedef struct {

   LARGE_INTEGER Length;

 } GET_LENGTH_INFORMATION;

 

 #define IOCTL_DISK_GET_LENGTH_INFO 0x7405C

 

-

-static inline int blockdev_size(int fd, unsigned long long *bytes)

+static inline int blockdev_size(struct fio_file *f, unsigned long long *bytes)

 {

 	int rc = 0;

-	HANDLE hFile = (HANDLE)_get_osfhandle(fd);

-	if (hFile != INVALID_HANDLE_VALUE)

-	{

-		GET_LENGTH_INFORMATION info;

-		DWORD outBytes;

-		LARGE_INTEGER size;

-		size.QuadPart = 0;

-		if (DeviceIoControl(hFile, IOCTL_DISK_GET_LENGTH_INFO, NULL, 0, &info, sizeof(info), &outBytes, NULL))

-			*bytes = info.Length.QuadPart;

-		else

-			rc = EIO;

+	HANDLE hFile;

+

+	if (f->hFile == NULL) {

+		hFile = CreateFile(f->file_name, (GENERIC_READ | GENERIC_WRITE),

+			(FILE_SHARE_READ | FILE_SHARE_WRITE), NULL, OPEN_EXISTING, 0, NULL);

+	} else {

+		hFile = f->hFile;

 	}

 

-	return 0;

-}

-

-static inline int chardev_size(int fd, unsigned long long *bytes)

-{

-	return blockdev_size(fd, bytes);

-}

-

-static inline int blockdev_invalidate_cache(int fd)

-{

-	int rc = 0;

-	HANDLE hFile = (HANDLE)_get_osfhandle(fd);

-

-	if (hFile != INVALID_HANDLE_VALUE)

-		FlushFileBuffers(hFile);

+	GET_LENGTH_INFORMATION info;

+	DWORD outBytes;

+	LARGE_INTEGER size;

+	size.QuadPart = 0;

+	if (DeviceIoControl(hFile, IOCTL_DISK_GET_LENGTH_INFO, NULL, 0, &info, sizeof(info), &outBytes, NULL))

+		*bytes = info.Length.QuadPart;

 	else

 		rc = EIO;

+	}

+

+	/* If we were passed a POSIX fd,

+	 * close the HANDLE we created via CreateFile */

+	if (hFile != INVALID_HANDLE_VALUE && f->hFile == NULL)

+		CloseHandle(hFile);

 

 	return rc;

 }

 

+static inline int chardev_size(struct fio_file *f, unsigned long long *bytes)

+{

+	return blockdev_size(f, bytes);

+}

+

+{

+

+static inline int blockdev_invalidate_cache(struct fio_file *f)

+{

+	BOOL bSuccess = FlushFileBuffers(f->hFile);

+	if (!bSuccess)

+		log_info("blockdev_invalidate_cache - FlushFileBuffers failed\n");

+

+	return 0;

+}

+

 static inline unsigned long long os_phys_mem(void)

 {

 	SYSTEM_INFO sysInfo;

diff --git a/os/os.h b/os/os.h
index 95f0ff2..92b6950 100644
--- a/os/os.h
+++ b/os/os.h
@@ -4,6 +4,7 @@
 #include <sys/types.h>
 #include <pthread.h>
 #include <unistd.h>
+#include <stdlib.h>
 
 #if defined(__linux__)
 #include "os-linux.h"
@@ -45,7 +46,7 @@
 #endif
 
 #ifndef FIO_HAVE_FADVISE
-#define fadvise(fd, off, len, advice)	(0)
+#define posix_fadvise(fd, off, len, advice)	(0)
 
 #ifndef POSIX_FADV_DONTNEED
 #define POSIX_FADV_DONTNEED	(0)
@@ -91,6 +92,12 @@
 #define OS_RAND_MAX			RAND_MAX
 #endif
 
+#ifdef FIO_HAVE_CLOCK_MONOTONIC
+#define FIO_TIMER_CLOCK CLOCK_MONOTONIC
+#else
+#define FIO_TIMER_CLOCK CLOCK_REALTIME
+#endif
+
 #ifndef FIO_HAVE_RAWBIND
 #define fio_lookup_raw(dev, majdev, mindev)	1
 #endif
@@ -124,13 +131,13 @@
 }
 
 #ifdef FIO_USE_GENERIC_BDEV_SIZE
-static inline int blockdev_size(int fd, unsigned long long *bytes)
+static inline int blockdev_size(struct fio_file *f, unsigned long long *bytes)
 {
 	off_t end;
 
 	*bytes = 0;
 
-	end = lseek(fd, 0, SEEK_END);
+	end = lseek(f->fd, 0, SEEK_END);
 	if (end < 0)
 		return errno;
 
diff --git a/os/windows/version.h b/os/windows/version.h
new file mode 100644
index 0000000..3642342
--- /dev/null
+++ b/os/windows/version.h
@@ -0,0 +1,4 @@
+#define FIO_VERSION_MAJOR 1

+#define FIO_VERSION_MINOR 44

+#define FIO_VERSION_BUILD 3

+#define FIO_VERSION_STRING "1.44.3"

diff --git a/os/windows/version.rc b/os/windows/version.rc
index f13ba06..d699f09 100755
--- a/os/windows/version.rc
+++ b/os/windows/version.rc
@@ -1,7 +1,7 @@
-#include "../../version.h"

+#include "version.h"

 1 VERSIONINFO

-FILEVERSION     VERSION_MAJOR,VERSION_MINOR,VERSION_BUILD,0

-PRODUCTVERSION  VERSION_MAJOR,VERSION_MINOR,VERSION_BUILD,0

+FILEVERSION     FIO_VERSION_MAJOR,FIO_VERSION_MINOR,FIO_VERSION_BUILD,0

+PRODUCTVERSION  FIO_VERSION_MAJOR,FIO_VERSION_MINOR,FIO_VERSION_BUILD,0

 BEGIN

   BLOCK "StringFileInfo"

   BEGIN

@@ -9,12 +9,12 @@
     BEGIN

       VALUE "CompanyName", ""

       VALUE "FileDescription", "Flexible IO Tester"

-      VALUE "FileVersion", "1.44.3"

+      VALUE "FileVersion", FIO_VERSION_STRING

       VALUE "InternalName", "fio"

       VALUE "LegalCopyright", "Copyright (C) 2010"

       VALUE "OriginalFilename", "fio.exe"

       VALUE "ProductName", "FIO"

-      VALUE "ProductVersion", VERSION_STR

+      VALUE "ProductVersion", FIO_VERSION_STRING

     END

   END

   BLOCK "VarFileInfo"