#include <errno.h>
#include "../fio.h"
+#include "../optgroup.h"
typedef BOOL (WINAPI *CANCELIOEX)(HANDLE hFile, LPOVERLAPPED lpOverlapped);
struct windowsaio_data *wd;
};
-static BOOL timeout_expired(DWORD start_count, DWORD end_count);
-static int fio_windowsaio_getevents(struct thread_data *td, unsigned int min,
- unsigned int max, const 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);
+struct windowsaio_options {
+ struct thread_data *td;
+ unsigned int no_completion_thread;
+};
+
+static struct fio_option options[] = {
+ {
+ .name = "no_completion_thread",
+ .lname = "No completion polling thread",
+ .type = FIO_OPT_STR_SET,
+ .off1 = offsetof(struct windowsaio_options, no_completion_thread),
+ .help = "Use to avoid separate completion polling thread",
+ .category = FIO_OPT_C_ENGINE,
+ .group = FIO_OPT_G_WINDOWSAIO,
+ },
+ {
+ .name = NULL,
+ },
+};
+
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 int fio_windowsaio_init(struct thread_data *td)
{
}
}
- td->io_ops->data = wd;
+ td->io_ops_data = wd;
if (!rc) {
struct thread_ctx *ctx;
struct windowsaio_data *wd;
HANDLE hFile;
+ struct windowsaio_options *o = td->eo;
hFile = CreateIoCompletionPort(INVALID_HANDLE_VALUE, NULL, 0, 0);
if (hFile == INVALID_HANDLE_VALUE) {
rc = 1;
}
- wd = td->io_ops->data;
+ wd = td->io_ops_data;
wd->iothread_running = TRUE;
wd->iocp = hFile;
- if (!rc)
- ctx = malloc(sizeof(struct thread_ctx));
+ if (o->no_completion_thread == 0) {
+ if (!rc)
+ ctx = malloc(sizeof(struct thread_ctx));
- if (!rc && ctx == NULL)
- {
- log_err("windowsaio: failed to allocate memory for thread context structure\n");
- CloseHandle(hFile);
- rc = 1;
- }
+ if (!rc && ctx == NULL) {
+ log_err("windowsaio: failed to allocate memory for thread context structure\n");
+ CloseHandle(hFile);
+ rc = 1;
+ }
- if (!rc)
- {
- ctx->iocp = hFile;
- ctx->wd = wd;
- wd->iothread = CreateThread(NULL, 0, IoCompletionRoutine, ctx, 0, NULL);
- if (wd->iothread == NULL)
- log_err("windowsaio: failed to create io completion thread\n");
- }
+ if (!rc) {
+ DWORD threadid;
- if (rc || wd->iothread == NULL)
- rc = 1;
+ ctx->iocp = hFile;
+ ctx->wd = wd;
+ wd->iothread = CreateThread(NULL, 0, IoCompletionRoutine, ctx, 0, &threadid);
+ if (!wd->iothread)
+ log_err("windowsaio: failed to create io completion thread\n");
+ else if (fio_option_is_set(&td->o, cpumask))
+ fio_setaffinity(threadid, td->o.cpumask);
+ }
+ if (rc || wd->iothread == NULL)
+ rc = 1;
+ }
}
return rc;
{
struct windowsaio_data *wd;
- wd = td->io_ops->data;
+ wd = td->io_ops_data;
if (wd != NULL) {
wd->iothread_running = FALSE;
free(wd->aio_events);
free(wd);
- td->io_ops->data = NULL;
+ td->io_ops_data = NULL;
}
}
+static int windowsaio_invalidate_cache(struct fio_file *f)
+{
+ DWORD error;
+ DWORD isharemode = (FILE_SHARE_DELETE | FILE_SHARE_READ |
+ FILE_SHARE_WRITE);
+ HANDLE ihFile;
+ int rc = 0;
+
+ /*
+ * Encourage Windows to drop cached parts of a file by temporarily
+ * opening it for non-buffered access. Note: this will only work when
+ * the following is the only thing with the file open on the whole
+ * system.
+ */
+ dprint(FD_IO, "windowaio: attempt invalidate cache for %s\n",
+ f->file_name);
+ ihFile = CreateFile(f->file_name, 0, isharemode, NULL, OPEN_EXISTING,
+ FILE_FLAG_NO_BUFFERING, NULL);
+
+ if (ihFile != INVALID_HANDLE_VALUE) {
+ if (!CloseHandle(ihFile)) {
+ error = GetLastError();
+ log_info("windowsaio: invalidation fd close %s failed: error %lu\n",
+ f->file_name, error);
+ rc = 1;
+ }
+ } else {
+ error = GetLastError();
+ if (error != ERROR_FILE_NOT_FOUND) {
+ log_info("windowsaio: cache invalidation of %s failed: error %lu\n",
+ f->file_name, error);
+ rc = 1;
+ }
+ }
+
+ return rc;
+}
static int fio_windowsaio_open_file(struct thread_data *td, struct fio_file *f)
{
/*
* Inform Windows whether we're going to be doing sequential or
- * random io so it can tune the Cache Manager
+ * random IO so it can tune the Cache Manager
*/
- if (td->o.td_ddir == TD_DDIR_READ ||
- td->o.td_ddir == TD_DDIR_WRITE)
- flags |= FILE_FLAG_SEQUENTIAL_SCAN;
- else
+ switch (td->o.fadvise_hint) {
+ case F_ADV_TYPE:
+ if (td_random(td))
+ flags |= FILE_FLAG_RANDOM_ACCESS;
+ else
+ flags |= FILE_FLAG_SEQUENTIAL_SCAN;
+ break;
+ case F_ADV_RANDOM:
flags |= FILE_FLAG_RANDOM_ACCESS;
+ break;
+ case F_ADV_SEQUENTIAL:
+ flags |= FILE_FLAG_SEQUENTIAL_SCAN;
+ break;
+ case F_ADV_NONE:
+ break;
+ default:
+ log_err("fio: unknown fadvise type %d\n", td->o.fadvise_hint);
+ }
- if (!td_write(td) || read_only)
+ if ((!td_write(td) && !(td->flags & TD_F_SYNCS)) || read_only)
access = GENERIC_READ;
else
access = (GENERIC_READ | GENERIC_WRITE);
else
openmode = OPEN_EXISTING;
+ /* If we're going to use direct I/O, Windows will try and invalidate
+ * its cache at that point so there's no need to do it here */
+ if (td->o.invalidate_cache && !td->o.odirect)
+ windowsaio_invalidate_cache(f);
+
f->hFile = CreateFile(f->file_name, access, sharemode,
NULL, openmode, flags, NULL);
/* Only set up the completion port and thread if we're not just
* querying the device size */
- if (!rc && td->io_ops->data != NULL) {
+ if (!rc && td->io_ops_data != NULL) {
struct windowsaio_data *wd;
- wd = td->io_ops->data;
+ wd = td->io_ops_data;
if (CreateIoCompletionPort(f->hFile, wd->iocp, 0, 0) == NULL) {
log_err("windowsaio: failed to create io completion port\n");
static struct io_u* fio_windowsaio_event(struct thread_data *td, int event)
{
- struct windowsaio_data *wd = td->io_ops->data;
+ struct windowsaio_data *wd = td->io_ops_data;
return wd->aio_events[event];
}
-static int fio_windowsaio_getevents(struct thread_data *td, unsigned int min,
- unsigned int max,
- const struct timespec *t)
+/* dequeue completion entrees directly (no separate completion thread) */
+static int fio_windowsaio_getevents_nothread(struct thread_data *td, unsigned int min,
+ unsigned int max, const struct timespec *t)
{
- struct windowsaio_data *wd = td->io_ops->data;
+ struct windowsaio_data *wd = td->io_ops_data;
+ unsigned int dequeued = 0;
+ struct io_u *io_u;
+ DWORD start_count = 0;
+ DWORD end_count = 0;
+ DWORD mswait = 250;
+ struct fio_overlapped *fov;
+
+ if (t != NULL) {
+ mswait = (t->tv_sec * 1000) + (t->tv_nsec / 1000000);
+ start_count = GetTickCount();
+ end_count = start_count + (t->tv_sec * 1000) + (t->tv_nsec / 1000000);
+ }
+
+ do {
+ BOOL ret;
+ OVERLAPPED *ovl;
+
+ ULONG entries = min(16, max-dequeued);
+ OVERLAPPED_ENTRY oe[16];
+ ret = GetQueuedCompletionStatusEx(wd->iocp, oe, 16, &entries, mswait, 0);
+ if (ret && entries) {
+ int entry_num;
+
+ for (entry_num=0; entry_num<entries; entry_num++) {
+ ovl = oe[entry_num].lpOverlapped;
+ fov = CONTAINING_RECORD(ovl, struct fio_overlapped, o);
+ io_u = fov->io_u;
+
+ 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 = win_to_posix_error(GetLastError());
+ }
+
+ fov->io_complete = FALSE;
+ wd->aio_events[dequeued] = io_u;
+ dequeued++;
+ }
+ }
+
+ if (dequeued >= min ||
+ (t != NULL && timeout_expired(start_count, end_count)))
+ break;
+ } while (1);
+ return dequeued;
+}
+
+/* dequeue completion entrees creates by separate IoCompletionRoutine thread */
+static int fio_windowaio_getevents_thread(struct thread_data *td, unsigned int min,
+ unsigned int max, const struct timespec *t)
+{
+ struct windowsaio_data *wd = td->io_ops_data;
unsigned int dequeued = 0;
struct io_u *io_u;
int i;
wd->aio_events[dequeued] = io_u;
dequeued++;
}
-
}
if (dequeued >= min)
break;
break;
}
- if (dequeued >= min || (t != NULL && timeout_expired(start_count, end_count)))
+ if (dequeued >= min ||
+ (t != NULL && timeout_expired(start_count, end_count)))
break;
} while (1);
return dequeued;
}
-static int fio_windowsaio_queue(struct thread_data *td, struct io_u *io_u)
+static int fio_windowsaio_getevents(struct thread_data *td, unsigned int min,
+ unsigned int max, const struct timespec *t)
+{
+ struct windowsaio_options *o = td->eo;
+
+ if (o->no_completion_thread)
+ return fio_windowsaio_getevents_nothread(td, min, max, t);
+ return fio_windowaio_getevents_thread(td, min, max, t);
+}
+
+static enum fio_q_status fio_windowsaio_queue(struct thread_data *td,
+ struct io_u *io_u)
{
struct fio_overlapped *o = io_u->engine_data;
LPOVERLAPPED lpOvl = &o->o;
switch (io_u->ddir) {
case DDIR_WRITE:
- success = WriteFile(io_u->file->hFile, io_u->xfer_buf, io_u->xfer_buflen, NULL, lpOvl);
+ success = WriteFile(io_u->file->hFile, io_u->xfer_buf,
+ io_u->xfer_buflen, NULL, lpOvl);
break;
case DDIR_READ:
- success = ReadFile(io_u->file->hFile, io_u->xfer_buf, io_u->xfer_buflen, NULL, lpOvl);
+ success = ReadFile(io_u->file->hFile, io_u->xfer_buf,
+ io_u->xfer_buflen, NULL, lpOvl);
break;
case DDIR_SYNC:
case DDIR_DATASYNC:
}
return FIO_Q_COMPLETED;
- break;
case DDIR_TRIM:
log_err("windowsaio: manual TRIM isn't supported on Windows\n");
io_u->error = 1;
io_u->resid = io_u->xfer_buflen;
return FIO_Q_COMPLETED;
- break;
default:
assert(0);
break;
wd = ctx->wd;
do {
- if (!GetQueuedCompletionStatus(ctx->iocp, &bytes, &ulKey, &ovl, 250) && ovl == NULL)
+ BOOL ret;
+
+ ret = GetQueuedCompletionStatus(ctx->iocp, &bytes, &ulKey,
+ &ovl, 250);
+ if (!ret && ovl == NULL)
continue;
fov = CONTAINING_RECORD(ovl, struct fio_overlapped, o);
.get_file_size = generic_get_file_size,
.io_u_init = fio_windowsaio_io_u_init,
.io_u_free = fio_windowsaio_io_u_free,
+ .options = options,
+ .option_struct_size = sizeof(struct windowsaio_options),
};
static void fio_init fio_windowsaio_register(void)