unsigned *tail;
unsigned *ring_mask;
unsigned *ring_entries;
+ unsigned *flags;
unsigned *array;
};
struct io_uring_cqe *cqes;
};
-#define DEPTH 32
+#define DEPTH 128
-#define BATCH_SUBMIT 8
-#define BATCH_COMPLETE 8
+#define BATCH_SUBMIT 64
+#define BATCH_COMPLETE 64
#define BS 4096
struct file {
unsigned long max_blocks;
+ unsigned pending_ios;
int fd;
};
unsigned long calls;
unsigned long cachehit, cachemiss;
volatile int finish;
+
struct file files[MAX_FDS];
unsigned nr_files;
unsigned cur_file;
static volatile int finish;
static int polled = 1; /* use IO polling */
-static int fixedbufs = 0; /* use fixed user buffers */
+static int fixedbufs = 1; /* use fixed user buffers */
static int buffered = 0; /* use buffered IO, not O_DIRECT */
static int sq_thread_poll = 0; /* use kernel submission/poller thread */
static int sq_thread_cpu = -1; /* pin above thread to this CPU */
IORING_REGISTER_BUFFERS, ®);
}
+static int io_uring_register_files(struct submitter *s)
+{
+ struct io_uring_register_files reg;
+ int i, ret;
+
+ reg.fds = calloc(s->nr_files, sizeof(int));
+ for (i = 0; i < s->nr_files; i++)
+ reg.fds[i] = s->files[i].fd;
+ reg.nr_fds = s->nr_files;
+
+ ret = syscall(__NR_sys_io_uring_register, s->ring_fd,
+ IORING_REGISTER_FILES, ®);
+ free(reg.fds);
+ return ret;
+}
+
static int io_uring_setup(unsigned entries, struct io_uring_params *p)
{
return syscall(__NR_sys_io_uring_setup, entries, p);
return syscall(__NR_gettid);
}
+static unsigned file_depth(struct submitter *s)
+{
+ return (DEPTH + s->nr_files - 1) / s->nr_files;
+}
+
static void init_io(struct submitter *s, unsigned index)
{
struct io_uring_sqe *sqe = &s->sqes[index];
struct file *f;
long r;
- f = &s->files[s->cur_file];
- s->cur_file++;
- if (s->cur_file == s->nr_files)
- s->cur_file = 0;
+ if (s->nr_files == 1) {
+ f = &s->files[0];
+ } else {
+ f = &s->files[s->cur_file];
+ if (f->pending_ios >= file_depth(s)) {
+ s->cur_file++;
+ if (s->cur_file == s->nr_files)
+ s->cur_file = 0;
+ }
+ }
+ f->pending_ios++;
lrand48_r(&s->rand, &r);
offset = (r % (f->max_blocks - 1)) * BS;
- sqe->flags = 0;
+ sqe->flags = IOSQE_FIXED_FILE;
sqe->opcode = IORING_OP_READV;
if (fixedbufs) {
sqe->addr = s->iovecs[index].iov_base;
sqe->ioprio = 0;
sqe->fd = f->fd;
sqe->off = offset;
+ sqe->data = (unsigned long) f;
}
static int prep_more_ios(struct submitter *s, int max_ios)
head = *ring->head;
do {
+ struct file *f;
+
barrier();
if (head == *ring->tail)
break;
cqe = &ring->cqes[head & cq_ring_mask];
+ f = (struct file *) cqe->data;
+ f->pending_ios--;
if (cqe->res != BS) {
printf("io: unexpected ret=%d\n", cqe->res);
return -1;
static void *submitter_fn(void *data)
{
struct submitter *s = data;
+ struct io_sq_ring *ring = &s->sq_ring;
int ret, prepped;
printf("submitter=%d\n", gettid());
else
to_wait = min(s->inflight + to_submit, BATCH_COMPLETE);
- ret = io_uring_enter(s, to_submit, to_wait,
- IORING_ENTER_GETEVENTS);
- s->calls++;
+ /*
+ * Only need to call io_uring_enter if we're not using SQ thread
+ * poll, or if IORING_SQ_NEED_WAKEUP is set.
+ */
+ if (!sq_thread_poll || (*ring->flags & IORING_SQ_NEED_WAKEUP)) {
+ ret = io_uring_enter(s, to_submit, to_wait,
+ IORING_ENTER_GETEVENTS);
+ s->calls++;
+ }
- this_reap = reap_events(s);
- if (this_reap == -1)
- break;
+ /*
+ * For non SQ thread poll, we already got the events we needed
+ * through the io_uring_enter() above. For SQ thread poll, we
+ * need to loop here until we find enough events.
+ */
+ this_reap = 0;
+ do {
+ int r;
+ r = reap_events(s);
+ if (r == -1)
+ break;
+ else if (r > 0)
+ this_reap += r;
+ } while (sq_thread_poll && this_reap < to_wait);
s->reaps += this_reap;
if (ret >= 0) {
if (fixedbufs) {
ret = io_uring_register_buffers(s);
if (ret < 0) {
- perror("io_uring_register");
+ perror("io_uring_register_buffers");
return 1;
}
}
+ ret = io_uring_register_files(s);
+ if (ret < 0) {
+ perror("io_uring_register_files");
+ return 1;
+ }
+
ptr = mmap(0, p.sq_off.array + p.sq_entries * sizeof(__u32),
PROT_READ | PROT_WRITE, MAP_SHARED | MAP_POPULATE, fd,
IORING_OFF_SQ_RING);
sring->tail = ptr + p.sq_off.tail;
sring->ring_mask = ptr + p.sq_off.ring_mask;
sring->ring_entries = ptr + p.sq_off.ring_entries;
+ sring->flags = ptr + p.sq_off.flags;
sring->array = ptr + p.sq_off.array;
sq_ring_mask = *sring->ring_mask;
return 1;
}
- flags = O_RDONLY;
+ flags = O_RDONLY | O_NOATIME;
if (!buffered)
flags |= O_DIRECT;
if (this_call - calls) {
rpc = (this_done - done) / (this_call - calls);
ipc = (this_reap - reap) / (this_call - calls);
- }
- printf("IOPS=%lu, IOS/call=%lu/%lu, inflight=%u (head=%u tail=%u), Cachehit=%0.2f%%\n",
+ } else
+ rpc = ipc = -1;
+ printf("IOPS=%lu, IOS/call=%ld/%ld, inflight=%u (head=%u tail=%u), Cachehit=%0.2f%%\n",
this_done - done, rpc, ipc, s->inflight,
*s->cq_ring.head, *s->cq_ring.tail, hit);
done = this_done;