#include "../arch/arch.h"
#include "../lib/types.h"
-#include "../os/io_uring.h"
-
-#define barrier() __asm__ __volatile__("": : :"memory")
+#include "../os/linux/io_uring.h"
#define min(a, b) ((a < b) ? (a) : (b))
unsigned *tail;
unsigned *ring_mask;
unsigned *ring_entries;
+ unsigned *flags;
unsigned *array;
};
struct io_uring_cqe *cqes;
};
-#define DEPTH 32
-
-#define BATCH_SUBMIT 8
-#define BATCH_COMPLETE 8
-
+#define DEPTH 128
+#define BATCH_SUBMIT 32
+#define BATCH_COMPLETE 32
#define BS 4096
+#define MAX_FDS 16
+
static unsigned sq_ring_mask, cq_ring_mask;
+struct file {
+ unsigned long max_blocks;
+ unsigned pending_ios;
+ int real_fd;
+ int fixed_fd;
+};
+
struct submitter {
pthread_t thread;
- unsigned long max_blocks;
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;
- char filename[128];
+
+ __s32 *fds;
+
+ struct file files[MAX_FDS];
+ unsigned nr_files;
+ unsigned cur_file;
+ struct iovec iovecs[];
};
-static struct submitter submitters[1];
+static struct submitter *submitter;
static volatile int finish;
+static int depth = DEPTH;
+static int batch_submit = BATCH_SUBMIT;
+static int batch_complete = BATCH_COMPLETE;
+static int bs = BS;
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 register_files = 1; /* use fixed files */
static int buffered = 0; /* use buffered IO, not O_DIRECT */
-static int sq_thread = 0; /* use kernel submission thread */
-static int sq_thread_cpu = 0; /* pin above thread to this CPU */
+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 do_nop = 0; /* no-op SQ ring commands */
+
+static int vectored = 1;
+
+static int io_uring_register_buffers(struct submitter *s)
+{
+ if (do_nop)
+ return 0;
+
+ return syscall(__NR_io_uring_register, s->ring_fd,
+ IORING_REGISTER_BUFFERS, s->iovecs, depth);
+}
-static int io_uring_setup(unsigned entries, struct iovec *iovecs,
- struct io_uring_params *p)
+static int io_uring_register_files(struct submitter *s)
{
- return syscall(__NR_sys_io_uring_setup, entries, iovecs, p);
+ int i;
+
+ if (do_nop)
+ return 0;
+
+ s->fds = calloc(s->nr_files, sizeof(__s32));
+ for (i = 0; i < s->nr_files; i++) {
+ s->fds[i] = s->files[i].real_fd;
+ s->files[i].fixed_fd = i;
+ }
+
+ return syscall(__NR_io_uring_register, s->ring_fd,
+ IORING_REGISTER_FILES, s->fds, s->nr_files);
+}
+
+static int io_uring_setup(unsigned entries, struct io_uring_params *p)
+{
+ return syscall(__NR_io_uring_setup, entries, p);
+}
+
+static void io_uring_probe(int fd)
+{
+ struct io_uring_probe *p;
+ int ret;
+
+ p = malloc(sizeof(*p) + 256 * sizeof(struct io_uring_probe_op));
+ if (!p)
+ return;
+
+ memset(p, 0, sizeof(*p) + 256 * sizeof(struct io_uring_probe_op));
+ ret = syscall(__NR_io_uring_register, fd, IORING_REGISTER_PROBE, p, 256);
+ if (ret < 0)
+ goto out;
+
+ if (IORING_OP_READ > p->ops_len)
+ goto out;
+
+ if ((p->ops[IORING_OP_READ].flags & IO_URING_OP_SUPPORTED))
+ vectored = 0;
+out:
+ free(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);
+ return syscall(__NR_io_uring_enter, s->ring_fd, to_submit, min_complete,
+ flags, NULL, 0);
}
+#ifndef CONFIG_HAVE_GETTID
static int gettid(void)
{
return syscall(__NR_gettid);
}
+#endif
+
+static unsigned file_depth(struct submitter *s)
+{
+ return (depth + s->nr_files - 1) / s->nr_files;
+}
-static void init_io(struct submitter *s, int fd, unsigned index)
+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;
- lrand48_r(&s->rand, &r);
- offset = (r % (s->max_blocks - 1)) * BS;
+ if (do_nop) {
+ sqe->opcode = IORING_OP_NOP;
+ return;
+ }
+ 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 = &s->files[s->cur_file];
+ }
+ }
+ f->pending_ios++;
+
+ r = lrand48();
+ offset = (r % (f->max_blocks - 1)) * bs;
+
+ if (register_files) {
+ sqe->flags = IOSQE_FIXED_FILE;
+ sqe->fd = f->fixed_fd;
+ } else {
+ sqe->flags = 0;
+ sqe->fd = f->real_fd;
+ }
if (fixedbufs) {
sqe->opcode = IORING_OP_READ_FIXED;
- sqe->addr = s->iovecs[index].iov_base;
- sqe->len = BS;
+ sqe->addr = (unsigned long) s->iovecs[index].iov_base;
+ sqe->len = bs;
+ sqe->buf_index = index;
+ } else if (!vectored) {
+ sqe->opcode = IORING_OP_READ;
+ sqe->addr = (unsigned long) s->iovecs[index].iov_base;
+ sqe->len = bs;
+ sqe->buf_index = 0;
} else {
sqe->opcode = IORING_OP_READV;
- sqe->addr = &s->iovecs[index];
+ sqe->addr = (unsigned long) &s->iovecs[index];
sqe->len = 1;
+ sqe->buf_index = 0;
}
- sqe->flags = 0;
sqe->ioprio = 0;
- sqe->fd = fd;
sqe->off = offset;
+ sqe->user_data = (unsigned long) f;
}
-static int prep_more_ios(struct submitter *s, int fd, int max_ios)
+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)
+ if (next_tail == atomic_load_acquire(ring->head))
break;
index = tail & sq_ring_mask;
- init_io(s, fd, index);
+ 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();
- }
+ if (prepped)
+ atomic_store_release(ring->tail, tail);
return prepped;
}
-static int get_file_size(int fd, unsigned long *blocks)
+static int get_file_size(struct file *f)
{
struct stat st;
- if (fstat(fd, &st) < 0)
+ if (fstat(f->real_fd, &st) < 0)
return -1;
if (S_ISBLK(st.st_mode)) {
unsigned long long bytes;
- if (ioctl(fd, BLKGETSIZE64, &bytes) != 0)
+ if (ioctl(f->real_fd, BLKGETSIZE64, &bytes) != 0)
return -1;
- *blocks = bytes / BS;
+ f->max_blocks = bytes / bs;
return 0;
} else if (S_ISREG(st.st_mode)) {
- *blocks = st.st_size / BS;
+ f->max_blocks = st.st_size / bs;
return 0;
}
head = *ring->head;
do {
- barrier();
- if (head == *ring->tail)
+ struct file *f;
+
+ read_barrier();
+ if (head == atomic_load_acquire(ring->tail))
break;
cqe = &ring->cqes[head & cq_ring_mask];
- if (cqe->res != BS) {
- struct io_uring_sqe *sqe = &s->sqes[cqe->index];
-
- printf("io: unexpected ret=%d\n", cqe->res);
- printf("offset=%lu, size=%lu\n",
- (unsigned long) sqe->off,
- (unsigned long) sqe->len);
- return -1;
+ if (!do_nop) {
+ f = (struct file *) (uintptr_t) cqe->user_data;
+ f->pending_ios--;
+ if (cqe->res != bs) {
+ printf("io: unexpected ret=%d\n", cqe->res);
+ if (polled && cqe->res == -EOPNOTSUPP)
+ printf("Your filesystem/driver/kernel doesn't support polled IO\n");
+ return -1;
+ }
}
- if (cqe->flags & IOCQE_FLAG_CACHEHIT)
- s->cachehit++;
- else
- s->cachemiss++;
reaped++;
head++;
} while (1);
- s->inflight -= reaped;
- *ring->head = head;
- barrier();
+ if (reaped) {
+ s->inflight -= reaped;
+ atomic_store_release(ring->head, head);
+ }
return reaped;
}
static void *submitter_fn(void *data)
{
struct submitter *s = data;
- int fd, ret, prepped, flags;
+ struct io_sq_ring *ring = &s->sq_ring;
+ int ret, prepped;
printf("submitter=%d\n", gettid());
- flags = O_RDONLY;
- if (!buffered)
- flags |= O_DIRECT;
- fd = open(s->filename, flags);
- if (fd < 0) {
- perror("open");
- goto done;
- }
-
- if (get_file_size(fd, &s->max_blocks)) {
- printf("failed getting size of device/file\n");
- goto err;
- }
- if (s->max_blocks <= 1) {
- printf("Zero file/device size?\n");
- goto err;
- }
- s->max_blocks--;
-
- srand48_r(pthread_self(), &s->rand);
+ srand48(pthread_self());
prepped = 0;
do {
int to_wait, to_submit, this_reap, to_prep;
+ unsigned ring_flags = 0;
- if (!prepped && s->inflight < DEPTH) {
- to_prep = min(DEPTH - s->inflight, BATCH_SUBMIT);
- prepped = prep_more_ios(s, fd, 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)
+ if (to_submit && (s->inflight + to_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++;
+ to_wait = min(s->inflight + to_submit, batch_complete);
+
+ /*
+ * 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 = atomic_load_acquire(ring->flags);
+ if (!sq_thread_poll || ring_flags & IORING_SQ_NEED_WAKEUP) {
+ unsigned flags = 0;
+
+ if (to_wait)
+ flags = IORING_ENTER_GETEVENTS;
+ if (ring_flags & IORING_SQ_NEED_WAKEUP)
+ flags |= IORING_ENTER_SQ_WAKEUP;
+ ret = io_uring_enter(s, to_submit, to_wait, flags);
+ s->calls++;
+ } else {
+ /* for SQPOLL, we submitted it all effectively */
+ ret = to_submit;
+ }
- 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) {
+ s->finish = 1;
+ break;
+ } else if (r > 0)
+ this_reap += r;
+ } while (sq_thread_poll && this_reap < to_wait);
s->reaps += this_reap;
if (ret >= 0) {
break;
}
} while (!s->finish);
-err:
- close(fd);
-done:
+
finish = 1;
return NULL;
}
static void sig_int(int sig)
{
printf("Exiting on signal %d\n", sig);
- submitters[0].finish = 1;
+ submitter->finish = 1;
finish = 1;
}
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;
- int fd;
memset(&p, 0, sizeof(p));
- if (polled)
+ if (polled && !do_nop)
p.flags |= IORING_SETUP_IOPOLL;
- if (buffered)
- p.flags |= IORING_SETUP_SQWQ;
- else if (sq_thread) {
- p.flags |= IORING_SETUP_SQTHREAD;
- p.sq_thread_cpu = sq_thread_cpu;
+ if (sq_thread_poll) {
+ p.flags |= IORING_SETUP_SQPOLL;
+ if (sq_thread_cpu != -1) {
+ p.flags |= IORING_SETUP_SQ_AFF;
+ p.sq_thread_cpu = sq_thread_cpu;
+ }
}
- if (fixedbufs)
- fd = io_uring_setup(DEPTH, s->iovecs, &p);
- else
- fd = io_uring_setup(DEPTH, NULL, &p);
+ fd = io_uring_setup(depth, &p);
if (fd < 0) {
perror("io_uring_setup");
return 1;
}
-
s->ring_fd = fd;
+
+ io_uring_probe(fd);
+
+ if (fixedbufs) {
+ ret = io_uring_register_buffers(s);
+ if (ret < 0) {
+ perror("io_uring_register_buffers");
+ return 1;
+ }
+ }
+
+ if (register_files) {
+ 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 0;
}
+static void file_depths(char *buf)
+{
+ struct submitter *s = submitter;
+ char *p;
+ int i;
+
+ buf[0] = '\0';
+ p = buf;
+ for (i = 0; i < s->nr_files; i++) {
+ struct file *f = &s->files[i];
+
+ if (i + 1 == s->nr_files)
+ p += sprintf(p, "%d", f->pending_ios);
+ else
+ p += sprintf(p, "%d, ", f->pending_ios);
+ }
+}
+
+static void usage(char *argv)
+{
+ printf("%s [options] -- [filenames]\n"
+ " -d <int> : IO Depth, default %d\n"
+ " -s <int> : Batch submit, default %d\n"
+ " -c <int> : Batch complete, default %d\n"
+ " -b <int> : Block size, default %d\n"
+ " -p <bool> : Polled IO, default %d\n",
+ argv, DEPTH, BATCH_SUBMIT, BATCH_COMPLETE, BS, polled);
+ exit(0);
+}
+
int main(int argc, char *argv[])
{
- struct submitter *s = &submitters[0];
- unsigned long done, calls, reap, cache_hit, cache_miss;
- int err, i;
- struct rlimit rlim;
+ struct submitter *s;
+ unsigned long done, calls, reap;
+ int err, i, flags, fd, opt;
+ char *fdepths;
void *ret;
- if (argc < 2) {
- printf("%s: filename\n", argv[0]);
+ if (!do_nop && argc < 2) {
+ printf("%s: filename [options]\n", argv[0]);
return 1;
}
- rlim.rlim_cur = RLIM_INFINITY;
- rlim.rlim_max = RLIM_INFINITY;
- if (setrlimit(RLIMIT_MEMLOCK, &rlim) < 0) {
- perror("setrlimit");
- return 1;
+ while ((opt = getopt(argc, argv, "d:s:c:b:p:B:F:h?")) != -1) {
+ switch (opt) {
+ case 'd':
+ depth = atoi(optarg);
+ break;
+ case 's':
+ batch_submit = atoi(optarg);
+ break;
+ case 'c':
+ batch_complete = atoi(optarg);
+ break;
+ case 'b':
+ bs = atoi(optarg);
+ break;
+ case 'p':
+ polled = !!atoi(optarg);
+ break;
+ case 'B':
+ fixedbufs = !!atoi(optarg);
+ break;
+ case 'F':
+ register_files = !!atoi(optarg);
+ break;
+ case 'h':
+ case '?':
+ default:
+ usage(argv[0]);
+ break;
+ }
+ }
+
+ submitter = malloc(sizeof(*submitter) + depth * sizeof(struct iovec));
+ memset(submitter, 0, sizeof(*submitter) + depth * sizeof(struct iovec));
+ s = submitter;
+
+ flags = O_RDONLY | O_NOATIME;
+ if (!buffered)
+ flags |= O_DIRECT;
+
+ i = optind;
+ while (!do_nop && i < argc) {
+ struct file *f;
+
+ if (s->nr_files == MAX_FDS) {
+ printf("Max number of files (%d) reached\n", MAX_FDS);
+ break;
+ }
+ fd = open(argv[i], flags);
+ if (fd < 0) {
+ perror("open");
+ return 1;
+ }
+
+ f = &s->files[s->nr_files];
+ f->real_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++;
+ }
+
+ if (fixedbufs) {
+ struct rlimit rlim;
+
+ 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++) {
+ for (i = 0; i < depth; i++) {
void *buf;
- if (posix_memalign(&buf, BS, BS)) {
+ if (posix_memalign(&buf, bs, bs)) {
printf("failed alloc\n");
return 1;
}
s->iovecs[i].iov_base = buf;
- s->iovecs[i].iov_len = BS;
+ s->iovecs[i].iov_len = bs;
}
err = setup_ring(s);
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);
- strcpy(s->filename, argv[1]);
+ printf("polled=%d, fixedbufs=%d, register_files=%d, buffered=%d", polled, fixedbufs, register_files, 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;
+ fdepths = malloc(8 * s->nr_files);
+ 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",
+ } else
+ rpc = ipc = -1;
+ file_depths(fdepths);
+ printf("IOPS=%lu, IOS/call=%ld/%ld, inflight=%u (%s)\n",
this_done - done, rpc, ipc, s->inflight,
- *s->cq_ring.head, *s->cq_ring.tail, hit);
+ fdepths);
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);
+ free(fdepths);
return 0;
}