#include <unistd.h>
#include <errno.h>
#include <libaio.h>
+#include <sys/time.h>
+#include <sys/resource.h>
#include "../fio.h"
#include "../lib/pow2.h"
#include "../optgroup.h"
#include "../lib/memalign.h"
-#ifndef IOCB_FLAG_HIPRI
-#define IOCB_FLAG_HIPRI (1 << 2)
+/* Should be defined in newest aio_abi.h */
+#ifndef IOCB_FLAG_IOPRIO
+#define IOCB_FLAG_IOPRIO (1 << 1)
#endif
-#ifndef IOCTX_FLAG_USERIOCB
-#define IOCTX_FLAG_USERIOCB (1 << 0)
-#endif
-#ifndef IOCTX_FLAG_IOPOLL
-#define IOCTX_FLAG_IOPOLL (1 << 1)
-#endif
-#ifndef IOCTX_FLAG_FIXEDBUFS
-#define IOCTX_FLAG_FIXEDBUFS (1 << 2)
+/* Hack for libaio < 0.3.111 */
+#ifndef CONFIG_LIBAIO_RW_FLAGS
+#define aio_rw_flags __pad2
#endif
static int fio_libaio_commit(struct thread_data *td);
+static int fio_libaio_init(struct thread_data *td);
struct libaio_data {
io_context_t aio_ctx;
struct iocb **iocbs;
struct io_u **io_us;
- struct iocb *user_iocbs;
struct io_u **io_u_index;
/*
struct libaio_options {
void *pad;
unsigned int userspace_reap;
- unsigned int hipri;
- unsigned int useriocb;
- unsigned int fixedbufs;
+ unsigned int cmdprio_percentage;
+ unsigned int nowait;
};
static struct fio_option options[] = {
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_LIBAIO,
},
+#ifdef FIO_HAVE_IOPRIO_CLASS
{
- .name = "hipri",
- .lname = "High Priority",
- .type = FIO_OPT_STR_SET,
- .off1 = offsetof(struct libaio_options, hipri),
- .help = "Use polled IO completions",
+ .name = "cmdprio_percentage",
+ .lname = "high priority percentage",
+ .type = FIO_OPT_INT,
+ .off1 = offsetof(struct libaio_options, cmdprio_percentage),
+ .minval = 1,
+ .maxval = 100,
+ .help = "Send high priority I/O this percentage of the time",
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_LIBAIO,
},
+#else
{
- .name = "useriocb",
- .lname = "User IOCBs",
- .type = FIO_OPT_STR_SET,
- .off1 = offsetof(struct libaio_options, useriocb),
- .help = "Use user mapped IOCBs",
- .category = FIO_OPT_C_ENGINE,
- .group = FIO_OPT_G_LIBAIO,
+ .name = "cmdprio_percentage",
+ .lname = "high priority percentage",
+ .type = FIO_OPT_UNSUPPORTED,
+ .help = "Your platform does not support I/O priority classes",
},
+#endif
{
- .name = "fixedbufs",
- .lname = "Fixed (pre-mapped) IO buffers",
- .type = FIO_OPT_STR_SET,
- .off1 = offsetof(struct libaio_options, fixedbufs),
- .help = "Pre map IO buffers",
+ .name = "nowait",
+ .lname = "RWF_NOWAIT",
+ .type = FIO_OPT_BOOL,
+ .off1 = offsetof(struct libaio_options, nowait),
+ .help = "Set RWF_NOWAIT for reads/writes",
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_LIBAIO,
},
*val = (*val + add) % ld->entries;
}
-static int fio_libaio_prep(struct thread_data fio_unused *td, struct io_u *io_u)
+static int fio_libaio_prep(struct thread_data *td, struct io_u *io_u)
{
- struct libaio_data *ld = td->io_ops_data;
- struct fio_file *f = io_u->file;
struct libaio_options *o = td->eo;
- struct iocb *iocb;
-
- if (o->useriocb)
- iocb = &ld->user_iocbs[io_u->index];
- else
- iocb = &io_u->iocb;
+ struct fio_file *f = io_u->file;
+ struct iocb *iocb = &io_u->iocb;
if (io_u->ddir == DDIR_READ) {
- if (o->fixedbufs) {
- iocb->aio_fildes = f->fd;
- iocb->aio_lio_opcode = IO_CMD_PREAD;
- iocb->u.c.offset = io_u->offset;
- } else {
- io_prep_pread(iocb, f->fd, io_u->xfer_buf, io_u->xfer_buflen, io_u->offset);
- if (o->hipri)
- iocb->u.c.flags |= IOCB_FLAG_HIPRI;
- }
+ io_prep_pread(iocb, f->fd, io_u->xfer_buf, io_u->xfer_buflen, io_u->offset);
+ if (o->nowait)
+ iocb->aio_rw_flags |= RWF_NOWAIT;
} else if (io_u->ddir == DDIR_WRITE) {
- if (o->fixedbufs) {
- iocb->aio_fildes = f->fd;
- iocb->aio_lio_opcode = IO_CMD_PWRITE;
- iocb->u.c.offset = io_u->offset;
- } else {
- io_prep_pwrite(iocb, f->fd, io_u->xfer_buf, io_u->xfer_buflen, io_u->offset);
- if (o->hipri)
- iocb->u.c.flags |= IOCB_FLAG_HIPRI;
- }
+ io_prep_pwrite(iocb, f->fd, io_u->xfer_buf, io_u->xfer_buflen, io_u->offset);
+ if (o->nowait)
+ iocb->aio_rw_flags |= RWF_NOWAIT;
} else if (ddir_sync(io_u->ddir))
io_prep_fsync(iocb, f->fd);
return 0;
}
+static void fio_libaio_prio_prep(struct thread_data *td, struct io_u *io_u)
+{
+ struct libaio_options *o = td->eo;
+ if (rand_between(&td->prio_state, 0, 99) < o->cmdprio_percentage) {
+ io_u->iocb.aio_reqprio = IOPRIO_CLASS_RT << IOPRIO_CLASS_SHIFT;
+ io_u->iocb.u.c.flags |= IOCB_FLAG_IOPRIO;
+ io_u->flags |= IO_U_F_PRIORITY;
+ }
+ return;
+}
+
static struct io_u *fio_libaio_event(struct thread_data *td, int event)
{
struct libaio_data *ld = td->io_ops_data;
- struct libaio_options *o = td->eo;
struct io_event *ev;
struct io_u *io_u;
ev = ld->aio_events + event;
- if (o->useriocb) {
- int index = (int) (uintptr_t) ev->obj;
- io_u = ld->io_u_index[index];
- } else
- io_u = container_of(ev->obj, struct io_u, iocb);
+ io_u = container_of(ev->obj, struct io_u, iocb);
if (ev->res != io_u->xfer_buflen) {
if (ev->res > io_u->xfer_buflen)
} else {
/* There is another completion to reap */
events[i] = ring->events[head];
- read_barrier();
- ring->head = (head + 1) % ring->nr;
+ atomic_store_release(&ring->head,
+ (head + 1) % ring->nr);
i++;
}
}
return FIO_Q_COMPLETED;
}
- if (o->useriocb)
- ld->iocbs[ld->head] = (struct iocb *) (uintptr_t) io_u->index;
- else
- ld->iocbs[ld->head] = &io_u->iocb;
+ if (o->cmdprio_percentage)
+ fio_libaio_prio_prep(td, io_u);
+ ld->iocbs[ld->head] = &io_u->iocb;
ld->io_us[ld->head] = io_u;
ring_inc(ld, &ld->head, 1);
ld->queued++;
free(ld->aio_events);
free(ld->iocbs);
free(ld->io_us);
- if (ld->user_iocbs) {
- size_t size = td->o.iodepth * sizeof(struct iocb);
- fio_memfree(ld->user_iocbs, size, false);
- }
free(ld);
}
}
-static int fio_libaio_old_queue_init(struct libaio_data *ld, unsigned int depth,
- bool hipri, bool useriocb, bool fixedbufs)
-{
- if (hipri) {
- log_err("fio: polled aio not available on your platform\n");
- return 1;
- }
- if (useriocb) {
- log_err("fio: user mapped iocbs not available on your platform\n");
- return 1;
- }
- if (fixedbufs) {
- log_err("fio: fixed buffers not available on your platform\n");
- return 1;
- }
-
- return io_queue_init(depth, &ld->aio_ctx);
-}
-
-static int fio_libaio_queue_init(struct libaio_data *ld, unsigned int depth,
- bool hipri, bool useriocb, bool fixedbufs)
-{
-#ifdef __NR_sys_io_setup2
- int ret, flags = 0;
-
- if (hipri)
- flags |= IOCTX_FLAG_IOPOLL;
- if (useriocb)
- flags |= IOCTX_FLAG_USERIOCB;
- if (fixedbufs)
- flags |= IOCTX_FLAG_FIXEDBUFS;
-
- ret = syscall(__NR_sys_io_setup2, depth, flags, ld->user_iocbs,
- &ld->aio_ctx);
- if (!ret)
- return 0;
- /* fall through to old syscall */
-#endif
- return fio_libaio_old_queue_init(ld, depth, hipri, useriocb, fixedbufs);
-}
-
static int fio_libaio_post_init(struct thread_data *td)
{
struct libaio_data *ld = td->io_ops_data;
- struct libaio_options *o = td->eo;
- struct io_u *io_u;
- struct iocb *iocb;
- int err = 0;
-
- if (o->fixedbufs) {
- int i;
-
- for (i = 0; i < td->o.iodepth; i++) {
- io_u = ld->io_u_index[i];
- iocb = &ld->user_iocbs[i];
- iocb->u.c.buf = io_u->buf;
- iocb->u.c.nbytes = td_max_bs(td);
-
- iocb->u.c.flags = 0;
- if (o->hipri)
- iocb->u.c.flags |= IOCB_FLAG_HIPRI;
- }
- }
+ int err;
- err = fio_libaio_queue_init(ld, td->o.iodepth, o->hipri, o->useriocb,
- o->fixedbufs);
+ err = io_queue_init(td->o.iodepth, &ld->aio_ctx);
if (err) {
td_verror(td, -err, "io_queue_init");
return 1;
static int fio_libaio_init(struct thread_data *td)
{
- struct libaio_options *o = td->eo;
struct libaio_data *ld;
+ struct thread_options *to = &td->o;
+ struct libaio_options *o = td->eo;
ld = calloc(1, sizeof(*ld));
- if (o->useriocb) {
- size_t size;
-
- ld->io_u_index = calloc(td->o.iodepth, sizeof(struct io_u *));
- size = td->o.iodepth * sizeof(struct iocb);
- ld->user_iocbs = fio_memalign(page_size, size, false);
- memset(ld->user_iocbs, 0, size);
- }
-
ld->entries = td->o.iodepth;
ld->is_pow2 = is_power_of_2(ld->entries);
ld->aio_events = calloc(ld->entries, sizeof(struct io_event));
ld->io_us = calloc(ld->entries, sizeof(struct io_u *));
td->io_ops_data = ld;
- return 0;
-}
-
-static int fio_libaio_io_u_init(struct thread_data *td, struct io_u *io_u)
-{
- struct libaio_options *o = td->eo;
-
- if (o->useriocb) {
- struct libaio_data *ld = td->io_ops_data;
-
- ld->io_u_index[io_u->index] = io_u;
+ /*
+ * Check for option conflicts
+ */
+ if ((fio_option_is_set(to, ioprio) || fio_option_is_set(to, ioprio_class)) &&
+ o->cmdprio_percentage != 0) {
+ log_err("%s: cmdprio_percentage option and mutually exclusive "
+ "prio or prioclass option is set, exiting\n", to->name);
+ td_verror(td, EINVAL, "fio_libaio_init");
+ return 1;
}
-
return 0;
}
-static struct ioengine_ops ioengine = {
+FIO_STATIC struct ioengine_ops ioengine = {
.name = "libaio",
.version = FIO_IOOPS_VERSION,
+ .flags = FIO_ASYNCIO_SYNC_TRIM,
.init = fio_libaio_init,
.post_init = fio_libaio_post_init,
- .io_u_init = fio_libaio_io_u_init,
.prep = fio_libaio_prep,
.queue = fio_libaio_queue,
.commit = fio_libaio_commit,