[fio.git] / t / io_uring.c

#include <stdio.h>
#include <errno.h>
#include <assert.h>
#include <stdlib.h>
#include <stddef.h>
#include <signal.h>
#include <inttypes.h>

#include <sys/types.h>
#include <sys/stat.h>
#include <sys/ioctl.h>
#include <sys/syscall.h>
#include <sys/resource.h>
#include <sys/mman.h>
#include <sys/uio.h>
#include <linux/fs.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <pthread.h>
#include <sched.h>

#include "../arch/arch.h"
#include "../lib/types.h"
#include "../os/io_uring.h"

#define barrier()	__asm__ __volatile__("": : :"memory")

#define min(a, b)		((a < b) ? (a) : (b))

struct io_sq_ring {
	unsigned *head;
	unsigned *tail;
	unsigned *ring_mask;
	unsigned *ring_entries;
	unsigned *array;
};

struct io_cq_ring {
	unsigned *head;
	unsigned *tail;
	unsigned *ring_mask;
	unsigned *ring_entries;
	struct io_uring_cqe *cqes;
};

#define DEPTH			32

#define BATCH_SUBMIT		8
#define BATCH_COMPLETE		8

#define BS			4096

#define MAX_FDS			16

static unsigned sq_ring_mask, cq_ring_mask;

struct file {
	unsigned long max_blocks;
	int fd;
};

struct submitter {
	pthread_t thread;
	int ring_fd;
	struct drand48_data rand;
	struct io_sq_ring sq_ring;
	struct io_uring_sqe *sqes;
	struct iovec iovecs[DEPTH];
	struct io_cq_ring cq_ring;
	int inflight;
	unsigned long reaps;
	unsigned long done;
	unsigned long calls;
	unsigned long cachehit, cachemiss;
	volatile int finish;
	struct file files[MAX_FDS];
	unsigned nr_files;
	unsigned cur_file;
};

static struct submitter submitters[1];
static volatile int finish;

static int polled = 1;		/* use IO polling */
static int fixedbufs = 0;	/* 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 */

static int io_uring_register_buffers(struct submitter *s)
{
	struct io_uring_register_buffers reg = {
		.iovecs = s->iovecs,
		.nr_iovecs = DEPTH
	};

	return syscall(__NR_sys_io_uring_register, s->ring_fd,
			IORING_REGISTER_BUFFERS, &reg);
}

static int io_uring_setup(unsigned entries, struct io_uring_params *p)
{
	return syscall(__NR_sys_io_uring_setup, entries, p);
}

static int io_uring_enter(struct submitter *s, unsigned int to_submit,
			  unsigned int min_complete, unsigned int flags)
{
	return syscall(__NR_sys_io_uring_enter, s->ring_fd, to_submit,
			min_complete, flags);
}

static int gettid(void)
{
	return syscall(__NR_gettid);
}

static void init_io(struct submitter *s, unsigned index)
{
	struct io_uring_sqe *sqe = &s->sqes[index];
	unsigned long offset;
	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;

	lrand48_r(&s->rand, &r);
	offset = (r % (f->max_blocks - 1)) * BS;

	sqe->flags = 0;
	sqe->opcode = IORING_OP_READV;
	if (fixedbufs) {
		sqe->addr = s->iovecs[index].iov_base;
		sqe->len = BS;
		sqe->buf_index = index;
		sqe->flags |= IOSQE_FIXED_BUFFER;
	} else {
		sqe->addr = &s->iovecs[index];
		sqe->len = 1;
		sqe->buf_index = 0;
	}
	sqe->ioprio = 0;
	sqe->fd = f->fd;
	sqe->off = offset;
}

static int prep_more_ios(struct submitter *s, int max_ios)
{
	struct io_sq_ring *ring = &s->sq_ring;
	unsigned index, tail, next_tail, prepped = 0;

	next_tail = tail = *ring->tail;
	do {
		next_tail++;
		barrier();
		if (next_tail == *ring->head)
			break;

		index = tail & sq_ring_mask;
		init_io(s, index);
		ring->array[index] = index;
		prepped++;
		tail = next_tail;
	} while (prepped < max_ios);

	if (*ring->tail != tail) {
		/* order tail store with writes to sqes above */
		barrier();
		*ring->tail = tail;
		barrier();
	}
	return prepped;
}

static int get_file_size(struct file *f)
{
	struct stat st;

	if (fstat(f->fd, &st) < 0)
		return -1;
	if (S_ISBLK(st.st_mode)) {
		unsigned long long bytes;

		if (ioctl(f->fd, BLKGETSIZE64, &bytes) != 0)
			return -1;

		f->max_blocks = bytes / BS;
		return 0;
	} else if (S_ISREG(st.st_mode)) {
		f->max_blocks = st.st_size / BS;
		return 0;
	}

	return -1;
}

static int reap_events(struct submitter *s)
{
	struct io_cq_ring *ring = &s->cq_ring;
	struct io_uring_cqe *cqe;
	unsigned head, reaped = 0;

	head = *ring->head;
	do {
		barrier();
		if (head == *ring->tail)
			break;
		cqe = &ring->cqes[head & cq_ring_mask];
		if (cqe->res != BS) {
			printf("io: unexpected ret=%d\n", cqe->res);
			return -1;
		}
		if (cqe->flags & IOCQE_FLAG_CACHEHIT)
			s->cachehit++;
		else
			s->cachemiss++;
		reaped++;
		head++;
	} while (1);

	s->inflight -= reaped;
	*ring->head = head;
	barrier();
	return reaped;
}

static void *submitter_fn(void *data)
{
	struct submitter *s = data;
	int ret, prepped;

	printf("submitter=%d\n", gettid());

	srand48_r(pthread_self(), &s->rand);

	prepped = 0;
	do {
		int to_wait, to_submit, this_reap, to_prep;

		if (!prepped && s->inflight < DEPTH) {
			to_prep = min(DEPTH - s->inflight, BATCH_SUBMIT);
			prepped = prep_more_ios(s, to_prep);
		}
		s->inflight += prepped;
submit_more:
		to_submit = prepped;
submit:
		if (s->inflight + BATCH_SUBMIT < DEPTH)
			to_wait = 0;
		else
			to_wait = min(s->inflight + to_submit, BATCH_COMPLETE);

		ret = io_uring_enter(s, to_submit, to_wait,
					IORING_ENTER_GETEVENTS);
		s->calls++;

		this_reap = reap_events(s);
		if (this_reap == -1)
			break;
		s->reaps += this_reap;

		if (ret >= 0) {
			if (!ret) {
				to_submit = 0;
				if (s->inflight)
					goto submit;
				continue;
			} else if (ret < to_submit) {
				int diff = to_submit - ret;

				s->done += ret;
				prepped -= diff;
				goto submit_more;
			}
			s->done += ret;
			prepped = 0;
			continue;
		} else if (ret < 0) {
			if (errno == EAGAIN) {
				if (s->finish)
					break;
				if (this_reap)
					goto submit;
				to_submit = 0;
				goto submit;
			}
			printf("io_submit: %s\n", strerror(errno));
			break;
		}
	} while (!s->finish);

	finish = 1;
	return NULL;
}

static void sig_int(int sig)
{
	printf("Exiting on signal %d\n", sig);
	submitters[0].finish = 1;
	finish = 1;
}

static void arm_sig_int(void)
{
	struct sigaction act;

	memset(&act, 0, sizeof(act));
	act.sa_handler = sig_int;
	act.sa_flags = SA_RESTART;
	sigaction(SIGINT, &act, NULL);
}

static int setup_ring(struct submitter *s)
{
	struct io_sq_ring *sring = &s->sq_ring;
	struct io_cq_ring *cring = &s->cq_ring;
	struct io_uring_params p;
	int ret, fd;
	void *ptr;

	memset(&p, 0, sizeof(p));

	if (sq_thread_poll) {
		p.flags |= IORING_SETUP_SQPOLL;
		if (sq_thread_cpu != -1)
			p.flags |= IORING_SETUP_SQ_AFF;
	}

	fd = io_uring_setup(DEPTH, &p);
	if (fd < 0) {
		perror("io_uring_setup");
		return 1;
	}
	s->ring_fd = fd;

	if (fixedbufs) {
		ret = io_uring_register_buffers(s);
		if (ret < 0) {
			perror("io_uring_register");
			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);
	printf("sq_ring ptr = 0x%p\n", ptr);
	sring->head = ptr + p.sq_off.head;
	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->array = ptr + p.sq_off.array;
	sq_ring_mask = *sring->ring_mask;

	s->sqes = mmap(0, p.sq_entries * sizeof(struct io_uring_sqe),
			PROT_READ | PROT_WRITE, MAP_SHARED | MAP_POPULATE, fd,
			IORING_OFF_SQES);
	printf("sqes ptr    = 0x%p\n", s->sqes);

	ptr = mmap(0, p.cq_off.cqes + p.cq_entries * sizeof(struct io_uring_cqe),
			PROT_READ | PROT_WRITE, MAP_SHARED | MAP_POPULATE, fd,
			IORING_OFF_CQ_RING);
	printf("cq_ring ptr = 0x%p\n", ptr);
	cring->head = ptr + p.cq_off.head;
	cring->tail = ptr + p.cq_off.tail;
	cring->ring_mask = ptr + p.cq_off.ring_mask;
	cring->ring_entries = ptr + p.cq_off.ring_entries;
	cring->cqes = ptr + p.cq_off.cqes;
	cq_ring_mask = *cring->ring_mask;
	return 0;
}

int main(int argc, char *argv[])
{
	struct submitter *s = &submitters[0];
	unsigned long done, calls, reap, cache_hit, cache_miss;
	int err, i, flags, fd;
	struct rlimit rlim;
	void *ret;

	if (argc < 2) {
		printf("%s: filename\n", argv[0]);
		return 1;
	}

	flags = O_RDONLY;
	if (!buffered)
		flags |= O_DIRECT;

	i = 1;
	while (i < argc) {
		struct file *f = &s->files[s->nr_files];

		fd = open(argv[i], flags);
		if (fd < 0) {
			perror("open");
			return 1;
		}
		f->fd = fd;
		if (get_file_size(f)) {
			printf("failed getting size of device/file\n");
			return 1;
		}
		if (f->max_blocks <= 1) {
			printf("Zero file/device size?\n");
			return 1;
		}
		f->max_blocks--;

		printf("Added file %s\n", argv[i]);
		s->nr_files++;
		i++;
	}

	rlim.rlim_cur = RLIM_INFINITY;
	rlim.rlim_max = RLIM_INFINITY;
	if (setrlimit(RLIMIT_MEMLOCK, &rlim) < 0) {
		perror("setrlimit");
		return 1;
	}

	arm_sig_int();

	for (i = 0; i < DEPTH; i++) {
		void *buf;

		if (posix_memalign(&buf, BS, BS)) {
			printf("failed alloc\n");
			return 1;
		}
		s->iovecs[i].iov_base = buf;
		s->iovecs[i].iov_len = BS;
	}

	err = setup_ring(s);
	if (err) {
		printf("ring setup failed: %s, %d\n", strerror(errno), err);
		return 1;
	}
	printf("polled=%d, fixedbufs=%d, buffered=%d", polled, fixedbufs, buffered);
	printf(" QD=%d, sq_ring=%d, cq_ring=%d\n", DEPTH, *s->sq_ring.ring_entries, *s->cq_ring.ring_entries);

	pthread_create(&s->thread, NULL, submitter_fn, s);

	cache_hit = cache_miss = reap = calls = done = 0;
	do {
		unsigned long this_done = 0;
		unsigned long this_reap = 0;
		unsigned long this_call = 0;
		unsigned long this_cache_hit = 0;
		unsigned long this_cache_miss = 0;
		unsigned long rpc = 0, ipc = 0;
		double hit = 0.0;

		sleep(1);
		this_done += s->done;
		this_call += s->calls;
		this_reap += s->reaps;
		this_cache_hit += s->cachehit;
		this_cache_miss += s->cachemiss;
		if (this_cache_hit && this_cache_miss) {
			unsigned long hits, total;

			hits = this_cache_hit - cache_hit;
			total = hits + this_cache_miss - cache_miss;
			hit = (double) hits / (double) total;
			hit *= 100.0;
		}
		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",
				this_done - done, rpc, ipc, s->inflight,
				*s->cq_ring.head, *s->cq_ring.tail, hit);
		done = this_done;
		calls = this_call;
		reap = this_reap;
		cache_hit = s->cachehit;
		cache_miss = s->cachemiss;
	} while (!finish);

	pthread_join(s->thread, &ret);
	close(s->ring_fd);
	return 0;
}
Commit	Line	Data
c9fb4c5b JA	1	#include <stdio.h>
	2	#include <errno.h>
	3	#include <assert.h>
	4	#include <stdlib.h>
	5	#include <stddef.h>
	6	#include <signal.h>
	7	#include <inttypes.h>
	8
	9	#include <sys/types.h>
	10	#include <sys/stat.h>
	11	#include <sys/ioctl.h>
	12	#include <sys/syscall.h>
	13	#include <sys/resource.h>
c3e2fc25	14	#include <sys/mman.h>
e31b8288	15	#include <sys/uio.h>
c9fb4c5b JA	16	#include <linux/fs.h>
	17	#include <fcntl.h>
	18	#include <unistd.h>
c9fb4c5b JA	19	#include <string.h>
	20	#include <pthread.h>
	21	#include <sched.h>
	22
74efb029	23	#include "../arch/arch.h"
57fa61f0	24	#include "../lib/types.h"
e31b8288	25	#include "../os/io_uring.h"
ac122fea	26
e31b8288	27	#define barrier() __asm__ __volatile__("": : :"memory")
c3e2fc25	28
e31b8288	29	#define min(a, b) ((a < b) ? (a) : (b))
c3e2fc25	30
e31b8288	31	struct io_sq_ring {
e2239016 JA	32	unsigned *head;
	33	unsigned *tail;
	34	unsigned *ring_mask;
	35	unsigned *ring_entries;
	36	unsigned *array;
c9fb4c5b JA	37	};
c9fb4c5b JA	38
e31b8288	39	struct io_cq_ring {
e2239016 JA	40	unsigned *head;
	41	unsigned *tail;
	42	unsigned *ring_mask;
	43	unsigned *ring_entries;
f0403f94	44	struct io_uring_cqe *cqes;
c9fb4c5b JA	45	};
c9fb4c5b JA	46
c9fb4c5b	47	#define DEPTH 32
c9fb4c5b JA	48
	49	#define BATCH_SUBMIT 8
	50	#define BATCH_COMPLETE 8
	51
	52	#define BS 4096
	53
a7086591 JA	54	#define MAX_FDS 16
a7086591 JA	55
c3e2fc25	56	static unsigned sq_ring_mask, cq_ring_mask;
e39c34dc	57
a7086591 JA	58	struct file {
	59	unsigned long max_blocks;
	60	int fd;
	61	};
	62
c9fb4c5b JA	63	struct submitter {
c9fb4c5b JA	64	pthread_t thread;
f310970e	65	int ring_fd;
c9fb4c5b	66	struct drand48_data rand;
e31b8288	67	struct io_sq_ring sq_ring;
f0403f94	68	struct io_uring_sqe *sqes;
c3e2fc25	69	struct iovec iovecs[DEPTH];
e31b8288	70	struct io_cq_ring cq_ring;
c9fb4c5b JA	71	int inflight;
	72	unsigned long reaps;
	73	unsigned long done;
	74	unsigned long calls;
305c06ce	75	unsigned long cachehit, cachemiss;
c9fb4c5b	76	volatile int finish;
a7086591 JA	77	struct file files[MAX_FDS];
	78	unsigned nr_files;
	79	unsigned cur_file;
c9fb4c5b JA	80	};
	81
	82	static struct submitter submitters[1];
	83	static volatile int finish;
	84
f0403f94	85	static int polled = 1; /* use IO polling */
c3e2fc25	86	static int fixedbufs = 0; /* use fixed user buffers */
f0403f94	87	static int buffered = 0; /* use buffered IO, not O_DIRECT */
3d7d00a3 JA	88	static int sq_thread_poll = 0; /* use kernel submission/poller thread */
3d7d00a3 JA	89	static int sq_thread_cpu = -1; /* pin above thread to this CPU */
c9fb4c5b	90
2ea53ca3	91	static int io_uring_register_buffers(struct submitter *s)
c9fb4c5b	92	{
2ea53ca3 JA	93	struct io_uring_register_buffers reg = {
	94	.iovecs = s->iovecs,
	95	.nr_iovecs = DEPTH
	96	};
	97
	98	return syscall(__NR_sys_io_uring_register, s->ring_fd,
	99	IORING_REGISTER_BUFFERS, &reg);
	100	}
	101
	102	static int io_uring_setup(unsigned entries, struct io_uring_params *p)
	103	{
	104	return syscall(__NR_sys_io_uring_setup, entries, p);
c9fb4c5b JA	105	}
c9fb4c5b JA	106
c3e2fc25 JA	107	static int io_uring_enter(struct submitter *s, unsigned int to_submit,
c3e2fc25 JA	108	unsigned int min_complete, unsigned int flags)
c9fb4c5b	109	{
f310970e JA	110	return syscall(__NR_sys_io_uring_enter, s->ring_fd, to_submit,
f310970e JA	111	min_complete, flags);
c9fb4c5b JA	112	}
	113
	114	static int gettid(void)
	115	{
	116	return syscall(__NR_gettid);
	117	}
	118
a7086591	119	static void init_io(struct submitter *s, unsigned index)
c9fb4c5b	120	{
f0403f94	121	struct io_uring_sqe *sqe = &s->sqes[index];
c9fb4c5b	122	unsigned long offset;
a7086591	123	struct file *f;
c9fb4c5b JA	124	long r;
c9fb4c5b JA	125
a7086591 JA	126	f = &s->files[s->cur_file];
	127	s->cur_file++;
	128	if (s->cur_file == s->nr_files)
	129	s->cur_file = 0;
	130
c9fb4c5b	131	lrand48_r(&s->rand, &r);
a7086591	132	offset = (r % (f->max_blocks - 1)) * BS;
c9fb4c5b	133
2ea53ca3 JA	134	sqe->flags = 0;
2ea53ca3 JA	135	sqe->opcode = IORING_OP_READV;
f0403f94	136	if (fixedbufs) {
f0403f94 JA	137	sqe->addr = s->iovecs[index].iov_base;
f0403f94 JA	138	sqe->len = BS;
2ea53ca3 JA	139	sqe->buf_index = index;
2ea53ca3 JA	140	sqe->flags \|= IOSQE_FIXED_BUFFER;
f0403f94	141	} else {
f0403f94 JA	142	sqe->addr = &s->iovecs[index];
f0403f94 JA	143	sqe->len = 1;
2ea53ca3	144	sqe->buf_index = 0;
f0403f94	145	}
f0403f94	146	sqe->ioprio = 0;
a7086591	147	sqe->fd = f->fd;
f0403f94	148	sqe->off = offset;
c9fb4c5b JA	149	}
c9fb4c5b JA	150
a7086591	151	static int prep_more_ios(struct submitter *s, int max_ios)
c9fb4c5b	152	{
e31b8288	153	struct io_sq_ring *ring = &s->sq_ring;
e2239016	154	unsigned index, tail, next_tail, prepped = 0;
c9fb4c5b	155
c3e2fc25	156	next_tail = tail = *ring->tail;
c9fb4c5b JA	157	do {
c9fb4c5b JA	158	next_tail++;
c9fb4c5b	159	barrier();
c3e2fc25	160	if (next_tail == *ring->head)
c9fb4c5b JA	161	break;
c9fb4c5b JA	162
e39c34dc	163	index = tail & sq_ring_mask;
a7086591	164	init_io(s, index);
c3e2fc25	165	ring->array[index] = index;
c9fb4c5b JA	166	prepped++;
	167	tail = next_tail;
	168	} while (prepped < max_ios);
	169
c3e2fc25	170	if (*ring->tail != tail) {
f0403f94	171	/* order tail store with writes to sqes above */
c9fb4c5b	172	barrier();
c3e2fc25	173	*ring->tail = tail;
c9fb4c5b JA	174	barrier();
	175	}
	176	return prepped;
	177	}
	178
a7086591	179	static int get_file_size(struct file *f)
c9fb4c5b JA	180	{
	181	struct stat st;
	182
a7086591	183	if (fstat(f->fd, &st) < 0)
c9fb4c5b JA	184	return -1;
	185	if (S_ISBLK(st.st_mode)) {
	186	unsigned long long bytes;
	187
a7086591	188	if (ioctl(f->fd, BLKGETSIZE64, &bytes) != 0)
c9fb4c5b JA	189	return -1;
c9fb4c5b JA	190
a7086591	191	f->max_blocks = bytes / BS;
c9fb4c5b JA	192	return 0;
c9fb4c5b JA	193	} else if (S_ISREG(st.st_mode)) {
a7086591	194	f->max_blocks = st.st_size / BS;
c9fb4c5b JA	195	return 0;
	196	}
	197
	198	return -1;
	199	}
	200
	201	static int reap_events(struct submitter *s)
	202	{
e31b8288	203	struct io_cq_ring *ring = &s->cq_ring;
f0403f94	204	struct io_uring_cqe *cqe;
e2239016	205	unsigned head, reaped = 0;
c9fb4c5b	206
c3e2fc25	207	head = *ring->head;
c9fb4c5b JA	208	do {
c9fb4c5b JA	209	barrier();
c3e2fc25	210	if (head == *ring->tail)
c9fb4c5b	211	break;
f0403f94 JA	212	cqe = &ring->cqes[head & cq_ring_mask];
f0403f94 JA	213	if (cqe->res != BS) {
f0403f94	214	printf("io: unexpected ret=%d\n", cqe->res);
c9fb4c5b JA	215	return -1;
c9fb4c5b JA	216	}
f0403f94	217	if (cqe->flags & IOCQE_FLAG_CACHEHIT)
305c06ce JA	218	s->cachehit++;
	219	else
	220	s->cachemiss++;
c9fb4c5b JA	221	reaped++;
c9fb4c5b JA	222	head++;
c9fb4c5b JA	223	} while (1);
	224
	225	s->inflight -= reaped;
c3e2fc25	226	*ring->head = head;
c9fb4c5b JA	227	barrier();
	228	return reaped;
	229	}
	230
	231	static void submitter_fn(void data)
	232	{
	233	struct submitter *s = data;
a7086591	234	int ret, prepped;
c9fb4c5b JA	235
	236	printf("submitter=%d\n", gettid());
	237
c9fb4c5b JA	238	srand48_r(pthread_self(), &s->rand);
	239
	240	prepped = 0;
	241	do {
f310970e	242	int to_wait, to_submit, this_reap, to_prep;
c9fb4c5b	243
f310970e JA	244	if (!prepped && s->inflight < DEPTH) {
f310970e JA	245	to_prep = min(DEPTH - s->inflight, BATCH_SUBMIT);
a7086591	246	prepped = prep_more_ios(s, to_prep);
f310970e	247	}
c9fb4c5b JA	248	s->inflight += prepped;
	249	submit_more:
	250	to_submit = prepped;
	251	submit:
	252	if (s->inflight + BATCH_SUBMIT < DEPTH)
	253	to_wait = 0;
	254	else
	255	to_wait = min(s->inflight + to_submit, BATCH_COMPLETE);
	256
f310970e JA	257	ret = io_uring_enter(s, to_submit, to_wait,
f310970e JA	258	IORING_ENTER_GETEVENTS);
c9fb4c5b JA	259	s->calls++;
	260
	261	this_reap = reap_events(s);
	262	if (this_reap == -1)
	263	break;
	264	s->reaps += this_reap;
	265
	266	if (ret >= 0) {
	267	if (!ret) {
	268	to_submit = 0;
	269	if (s->inflight)
	270	goto submit;
	271	continue;
	272	} else if (ret < to_submit) {
	273	int diff = to_submit - ret;
	274
	275	s->done += ret;
	276	prepped -= diff;
	277	goto submit_more;
	278	}
	279	s->done += ret;
	280	prepped = 0;
	281	continue;
	282	} else if (ret < 0) {
ac122fea	283	if (errno == EAGAIN) {
c9fb4c5b JA	284	if (s->finish)
	285	break;
	286	if (this_reap)
	287	goto submit;
c9fb4c5b JA	288	to_submit = 0;
	289	goto submit;
	290	}
ac122fea	291	printf("io_submit: %s\n", strerror(errno));
c9fb4c5b JA	292	break;
	293	}
	294	} while (!s->finish);
a7086591	295
c9fb4c5b JA	296	finish = 1;
	297	return NULL;
	298	}
	299
	300	static void sig_int(int sig)
	301	{
	302	printf("Exiting on signal %d\n", sig);
	303	submitters[0].finish = 1;
	304	finish = 1;
	305	}
	306
	307	static void arm_sig_int(void)
	308	{
	309	struct sigaction act;
	310
	311	memset(&act, 0, sizeof(act));
	312	act.sa_handler = sig_int;
	313	act.sa_flags = SA_RESTART;
	314	sigaction(SIGINT, &act, NULL);
	315	}
	316
c3e2fc25 JA	317	static int setup_ring(struct submitter *s)
c3e2fc25 JA	318	{
e31b8288 JA	319	struct io_sq_ring *sring = &s->sq_ring;
	320	struct io_cq_ring *cring = &s->cq_ring;
	321	struct io_uring_params p;
2ea53ca3	322	int ret, fd;
c3e2fc25	323	void *ptr;
c3e2fc25 JA	324
	325	memset(&p, 0, sizeof(p));
	326
3d7d00a3	327	if (sq_thread_poll) {
2ea53ca3	328	p.flags \|= IORING_SETUP_SQPOLL;
3d7d00a3 JA	329	if (sq_thread_cpu != -1)
3d7d00a3 JA	330	p.flags \|= IORING_SETUP_SQ_AFF;
c3e2fc25 JA	331	}
c3e2fc25 JA	332
2ea53ca3	333	fd = io_uring_setup(DEPTH, &p);
c3e2fc25 JA	334	if (fd < 0) {
	335	perror("io_uring_setup");
	336	return 1;
	337	}
f310970e	338	s->ring_fd = fd;
2ea53ca3 JA	339
	340	if (fixedbufs) {
	341	ret = io_uring_register_buffers(s);
	342	if (ret < 0) {
	343	perror("io_uring_register");
	344	return 1;
	345	}
	346	}
	347
e2239016	348	ptr = mmap(0, p.sq_off.array + p.sq_entries * sizeof(__u32),
f310970e JA	349	PROT_READ \| PROT_WRITE, MAP_SHARED \| MAP_POPULATE, fd,
f310970e JA	350	IORING_OFF_SQ_RING);
c3e2fc25 JA	351	printf("sq_ring ptr = 0x%p\n", ptr);
	352	sring->head = ptr + p.sq_off.head;
	353	sring->tail = ptr + p.sq_off.tail;
	354	sring->ring_mask = ptr + p.sq_off.ring_mask;
	355	sring->ring_entries = ptr + p.sq_off.ring_entries;
ac122fea	356	sring->array = ptr + p.sq_off.array;
c3e2fc25 JA	357	sq_ring_mask = *sring->ring_mask;
c3e2fc25 JA	358
f0403f94	359	s->sqes = mmap(0, p.sq_entries * sizeof(struct io_uring_sqe),
f310970e	360	PROT_READ \| PROT_WRITE, MAP_SHARED \| MAP_POPULATE, fd,
f0403f94 JA	361	IORING_OFF_SQES);
f0403f94 JA	362	printf("sqes ptr = 0x%p\n", s->sqes);
c3e2fc25	363
f0403f94	364	ptr = mmap(0, p.cq_off.cqes + p.cq_entries * sizeof(struct io_uring_cqe),
f310970e JA	365	PROT_READ \| PROT_WRITE, MAP_SHARED \| MAP_POPULATE, fd,
f310970e JA	366	IORING_OFF_CQ_RING);
c3e2fc25 JA	367	printf("cq_ring ptr = 0x%p\n", ptr);
	368	cring->head = ptr + p.cq_off.head;
	369	cring->tail = ptr + p.cq_off.tail;
	370	cring->ring_mask = ptr + p.cq_off.ring_mask;
	371	cring->ring_entries = ptr + p.cq_off.ring_entries;
f0403f94	372	cring->cqes = ptr + p.cq_off.cqes;
c3e2fc25 JA	373	cq_ring_mask = *cring->ring_mask;
	374	return 0;
	375	}
	376
c9fb4c5b JA	377	int main(int argc, char *argv[])
	378	{
	379	struct submitter *s = &submitters[0];
305c06ce	380	unsigned long done, calls, reap, cache_hit, cache_miss;
a7086591	381	int err, i, flags, fd;
c9fb4c5b	382	struct rlimit rlim;
c3e2fc25	383	void *ret;
c9fb4c5b JA	384
	385	if (argc < 2) {
	386	printf("%s: filename\n", argv[0]);
	387	return 1;
	388	}
	389
a7086591 JA	390	flags = O_RDONLY;
	391	if (!buffered)
	392	flags \|= O_DIRECT;
	393
	394	i = 1;
	395	while (i < argc) {
	396	struct file *f = &s->files[s->nr_files];
	397
	398	fd = open(argv[i], flags);
	399	if (fd < 0) {
	400	perror("open");
	401	return 1;
	402	}
	403	f->fd = fd;
	404	if (get_file_size(f)) {
	405	printf("failed getting size of device/file\n");
	406	return 1;
	407	}
	408	if (f->max_blocks <= 1) {
	409	printf("Zero file/device size?\n");
	410	return 1;
	411	}
	412	f->max_blocks--;
	413
	414	printf("Added file %s\n", argv[i]);
	415	s->nr_files++;
	416	i++;
	417	}
	418
c9fb4c5b JA	419	rlim.rlim_cur = RLIM_INFINITY;
	420	rlim.rlim_max = RLIM_INFINITY;
	421	if (setrlimit(RLIMIT_MEMLOCK, &rlim) < 0) {
	422	perror("setrlimit");
	423	return 1;
	424	}
	425
	426	arm_sig_int();
	427
c3e2fc25 JA	428	for (i = 0; i < DEPTH; i++) {
c3e2fc25 JA	429	void *buf;
c9fb4c5b	430
c3e2fc25	431	if (posix_memalign(&buf, BS, BS)) {
c9fb4c5b JA	432	printf("failed alloc\n");
	433	return 1;
	434	}
c3e2fc25 JA	435	s->iovecs[i].iov_base = buf;
c3e2fc25 JA	436	s->iovecs[i].iov_len = BS;
c9fb4c5b JA	437	}
c9fb4c5b JA	438
c3e2fc25	439	err = setup_ring(s);
c9fb4c5b	440	if (err) {
c3e2fc25	441	printf("ring setup failed: %s, %d\n", strerror(errno), err);
c9fb4c5b JA	442	return 1;
c9fb4c5b JA	443	}
c3e2fc25 JA	444	printf("polled=%d, fixedbufs=%d, buffered=%d", polled, fixedbufs, buffered);
c3e2fc25 JA	445	printf(" QD=%d, sq_ring=%d, cq_ring=%d\n", DEPTH, s->sq_ring.ring_entries, s->cq_ring.ring_entries);
c9fb4c5b JA	446
	447	pthread_create(&s->thread, NULL, submitter_fn, s);
	448
305c06ce	449	cache_hit = cache_miss = reap = calls = done = 0;
c9fb4c5b JA	450	do {
	451	unsigned long this_done = 0;
	452	unsigned long this_reap = 0;
	453	unsigned long this_call = 0;
305c06ce JA	454	unsigned long this_cache_hit = 0;
305c06ce JA	455	unsigned long this_cache_miss = 0;
c9fb4c5b	456	unsigned long rpc = 0, ipc = 0;
305c06ce	457	double hit = 0.0;
c9fb4c5b JA	458
	459	sleep(1);
	460	this_done += s->done;
	461	this_call += s->calls;
	462	this_reap += s->reaps;
305c06ce JA	463	this_cache_hit += s->cachehit;
	464	this_cache_miss += s->cachemiss;
	465	if (this_cache_hit && this_cache_miss) {
	466	unsigned long hits, total;
	467
	468	hits = this_cache_hit - cache_hit;
	469	total = hits + this_cache_miss - cache_miss;
	470	hit = (double) hits / (double) total;
	471	hit *= 100.0;
	472	}
c9fb4c5b JA	473	if (this_call - calls) {
	474	rpc = (this_done - done) / (this_call - calls);
	475	ipc = (this_reap - reap) / (this_call - calls);
	476	}
10c4d131	477	printf("IOPS=%lu, IOS/call=%lu/%lu, inflight=%u (head=%u tail=%u), Cachehit=%0.2f%%\n",
c9fb4c5b	478	this_done - done, rpc, ipc, s->inflight,
c3e2fc25	479	s->cq_ring.head, s->cq_ring.tail, hit);
c9fb4c5b JA	480	done = this_done;
	481	calls = this_call;
	482	reap = this_reap;
305c06ce JA	483	cache_hit = s->cachehit;
305c06ce JA	484	cache_miss = s->cachemiss;
c9fb4c5b JA	485	} while (!finish);
	486
	487	pthread_join(s->thread, &ret);
e31b8288	488	close(s->ring_fd);
c9fb4c5b JA	489	return 0;
c9fb4c5b JA	490	}