static int plist_len = 17;
#ifndef IORING_REGISTER_MAP_BUFFERS
-#define IORING_REGISTER_MAP_BUFFERS 22
+#define IORING_REGISTER_MAP_BUFFERS 26
struct io_uring_map_buffers {
__s32 fd;
__u32 buf_start;
static int io_uring_setup(unsigned entries, struct io_uring_params *p)
{
+ int ret;
+
/*
* Clamp CQ ring size at our SQ ring size, we don't need more entries
* than that.
p->flags |= IORING_SETUP_CQSIZE;
p->cq_entries = entries;
- return syscall(__NR_io_uring_setup, entries, p);
+ p->flags |= IORING_SETUP_COOP_TASKRUN;
+ p->flags |= IORING_SETUP_SINGLE_ISSUER;
+ p->flags |= IORING_SETUP_DEFER_TASKRUN;
+retry:
+ ret = syscall(__NR_io_uring_setup, entries, p);
+ if (!ret)
+ return 0;
+
+ if (errno == EINVAL && p->flags & IORING_SETUP_COOP_TASKRUN) {
+ p->flags &= ~IORING_SETUP_COOP_TASKRUN;
+ goto retry;
+ }
+ if (errno == EINVAL && p->flags & IORING_SETUP_SINGLE_ISSUER) {
+ p->flags &= ~IORING_SETUP_SINGLE_ISSUER;
+ goto retry;
+ }
+ if (errno == EINVAL && p->flags & IORING_SETUP_DEFER_TASKRUN) {
+ p->flags &= ~IORING_SETUP_DEFER_TASKRUN;
+ goto retry;
+ }
+
+ return ret;
}
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));
+ p = calloc(1, 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;
return (depth + s->nr_files - 1) / s->nr_files;
}
-static void init_io(struct submitter *s, unsigned index)
+static unsigned long long get_offset(struct submitter *s, struct file *f)
{
- struct io_uring_sqe *sqe = &s->sqes[index];
- unsigned long offset;
- struct file *f;
+ unsigned long long offset;
long r;
- if (do_nop) {
- sqe->opcode = IORING_OP_NOP;
- return;
+ if (random_io) {
+ unsigned long long block;
+
+ r = __rand64(&s->rand_state);
+ block = r % f->max_blocks;
+ offset = block * (unsigned long long) bs;
+ } else {
+ offset = f->cur_off;
+ f->cur_off += bs;
+ if (f->cur_off + bs > f->max_size)
+ f->cur_off = 0;
}
+ return offset;
+}
+
+static struct file *init_new_io(struct submitter *s)
+{
+ struct file *f;
+
if (s->nr_files == 1) {
f = &s->files[0];
} else {
f = &s->files[s->cur_file];
}
}
+
f->pending_ios++;
+ return f;
+}
- if (random_io) {
- r = __rand64(&s->rand_state);
- offset = (r % (f->max_blocks - 1)) * bs;
- } else {
- offset = f->cur_off;
- f->cur_off += bs;
- if (f->cur_off + bs > f->max_size)
- f->cur_off = 0;
+static void init_io(struct submitter *s, unsigned index)
+{
+ struct io_uring_sqe *sqe = &s->sqes[index];
+ struct file *f;
+
+ if (do_nop) {
+ sqe->opcode = IORING_OP_NOP;
+ return;
}
+ f = init_new_io(s);
+
if (register_files) {
sqe->flags = IOSQE_FIXED_FILE;
sqe->fd = f->fixed_fd;
sqe->buf_index = 0;
}
sqe->ioprio = 0;
- sqe->off = offset;
+ sqe->off = get_offset(s, f);
sqe->user_data = (unsigned long) f->fileno;
if (stats && stats_running)
sqe->user_data |= ((uint64_t)s->clock_index << 32);
struct nvme_uring_cmd *cmd;
unsigned long long slba;
unsigned long long nlb;
- long r;
- 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++;
+ f = init_new_io(s);
- if (random_io) {
- r = __rand64(&s->rand_state);
- offset = (r % (f->max_blocks - 1)) * bs;
- } else {
- offset = f->cur_off;
- f->cur_off += bs;
- if (f->cur_off + bs > f->max_size)
- f->cur_off = 0;
- }
+ offset = get_offset(s, f);
if (register_files) {
sqe->fd = f->fixed_fd;
cmd->cdw12 = nlb;
cmd->addr = (unsigned long) s->iovecs[index].iov_base;
cmd->data_len = bs;
+ if (fixedbufs) {
+ sqe->uring_cmd_flags = IORING_URING_CMD_FIXED;
+ sqe->buf_index = index;
+ }
cmd->nsid = f->nsid;
cmd->opcode = 2;
}
static int prep_more_ios_uring(struct submitter *s, int max_ios)
{
struct io_sq_ring *ring = &s->sq_ring;
- unsigned index, tail, next_tail, prepped = 0;
+ unsigned head, index, tail, next_tail, prepped = 0;
+
+ if (sq_thread_poll)
+ head = atomic_load_acquire(ring->head);
+ else
+ head = *ring->head;
next_tail = tail = *ring->tail;
do {
next_tail++;
- if (next_tail == atomic_load_acquire(ring->head))
+ if (next_tail == head)
break;
index = tail & sq_ring_mask;
init_io_pt(s, index);
else
init_io(s, index);
- ring->array[index] = index;
prepped++;
tail = next_tail;
} while (prepped < max_ios);
bs, lbs);
return -1;
}
- f->max_blocks = nlba / bs;
+ f->max_blocks = nlba;
f->max_size = nlba;
f->lba_shift = ilog2(lbs);
return 0;
do {
struct file *f;
- read_barrier();
if (head == atomic_load_acquire(ring->tail))
break;
cqe = &ring->cqes[head & cq_ring_mask];
do {
struct file *f;
- read_barrier();
if (head == atomic_load_acquire(ring->tail))
break;
index = head & cq_ring_mask;
char str[128];
int ret, fd, node;
- sprintf(str, "/sys/block/%s/device/numa_node", base);
+ if (pt)
+ sprintf(str, "/sys/class/nvme-generic/%s/device/numa_node", base);
+ else
+ sprintf(str, "/sys/block/%s/device/numa_node", base);
fd = open(str, O_RDONLY);
if (fd < 0)
return -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;
+ int ret, fd, i;
void *ptr;
size_t len;
cring->ring_entries = ptr + p.cq_off.ring_entries;
cring->cqes = ptr + p.cq_off.cqes;
cq_ring_mask = *cring->ring_mask;
+
+ for (i = 0; i < p.sq_entries; i++)
+ sring->array[i] = i;
+
return 0;
}
static int prep_more_ios_aio(struct submitter *s, int max_ios, struct iocb *iocbs)
{
uint64_t data;
- long long offset;
struct file *f;
unsigned index;
- long r;
index = 0;
while (index < max_ios) {
struct iocb *iocb = &iocbs[index];
- 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++;
+ f = init_new_io(s);
- r = lrand48();
- offset = (r % (f->max_blocks - 1)) * bs;
io_prep_pread(iocb, f->real_fd, s->iovecs[index].iov_base,
- s->iovecs[index].iov_len, offset);
+ s->iovecs[index].iov_len, get_offset(s, f));
data = f->fileno;
if (stats && stats_running)
do {
uint64_t offset;
struct file *f;
- long r;
- 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++;
-
- if (random_io) {
- r = __rand64(&s->rand_state);
- offset = (r % (f->max_blocks - 1)) * bs;
- } else {
- offset = f->cur_off;
- f->cur_off += bs;
- if (f->cur_off + bs > f->max_size)
- f->cur_off = 0;
- }
+ f = init_new_io(s);
#ifdef ARCH_HAVE_CPU_CLOCK
if (stats)
s->inflight++;
s->calls++;
+ offset = get_offset(s, f);
if (polled)
ret = preadv2(f->real_fd, &s->iovecs[0], 1, offset, RWF_HIPRI);
else