2
0
mirror of https://github.com/edk2-porting/linux-next.git synced 2025-01-09 06:04:05 +08:00

RDMA/rtrs-srv: Replace atomic_t with percpu_ref for ids_inflight

ids_inflight is used to track the inflight IOs. But the use of atomic_t
variable can cause performance drops and can also become a performance
bottleneck.

This commit replaces the use of atomic_t with a percpu_ref structure. The
advantage it offers is, it doesn't check if the reference has fallen to 0,
until the user explicitly signals it to; and that is done by the
percpu_ref_kill() function call. After that, the percpu_ref structure
behaves like an atomic_t and for every put call, checks whether the
reference has fallen to 0 or not.

rtrs_srv_stats_rdma_to_str shows the count of ids_inflight as 0
for user-mode tools not to be confused.

Fixes: 9cb8374804 ("RDMA/rtrs: server: main functionality")
Link: https://lore.kernel.org/r/20210528113018.52290-14-jinpu.wang@ionos.com
Signed-off-by: Md Haris Iqbal <haris.iqbal@ionos.com>
Signed-off-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Gioh Kim <gi-oh.kim@ionos.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
This commit is contained in:
Md Haris Iqbal 2021-05-28 13:30:11 +02:00 committed by Jason Gunthorpe
parent 41db63a7ef
commit 0cdfb3b207
3 changed files with 35 additions and 24 deletions

View File

@ -27,12 +27,10 @@ ssize_t rtrs_srv_stats_rdma_to_str(struct rtrs_srv_stats *stats,
char *page, size_t len)
{
struct rtrs_srv_stats_rdma_stats *r = &stats->rdma_stats;
struct rtrs_srv_sess *sess = stats->sess;
return scnprintf(page, len, "%lld %lld %lld %lld %u\n",
(s64)atomic64_read(&r->dir[READ].cnt),
(s64)atomic64_read(&r->dir[READ].size_total),
(s64)atomic64_read(&r->dir[WRITE].cnt),
(s64)atomic64_read(&r->dir[WRITE].size_total),
atomic_read(&sess->ids_inflight));
return scnprintf(page, len, "%lld %lld %lld %lldn %u\n",
(s64)atomic64_read(&r->dir[READ].cnt),
(s64)atomic64_read(&r->dir[READ].size_total),
(s64)atomic64_read(&r->dir[WRITE].cnt),
(s64)atomic64_read(&r->dir[WRITE].size_total), 0);
}

View File

@ -111,7 +111,6 @@ static void rtrs_srv_free_ops_ids(struct rtrs_srv_sess *sess)
struct rtrs_srv *srv = sess->srv;
int i;
WARN_ON(atomic_read(&sess->ids_inflight));
if (sess->ops_ids) {
for (i = 0; i < srv->queue_depth; i++)
free_id(sess->ops_ids[i]);
@ -126,11 +125,19 @@ static struct ib_cqe io_comp_cqe = {
.done = rtrs_srv_rdma_done
};
static inline void rtrs_srv_inflight_ref_release(struct percpu_ref *ref)
{
struct rtrs_srv_sess *sess = container_of(ref, struct rtrs_srv_sess, ids_inflight_ref);
percpu_ref_exit(&sess->ids_inflight_ref);
complete(&sess->complete_done);
}
static int rtrs_srv_alloc_ops_ids(struct rtrs_srv_sess *sess)
{
struct rtrs_srv *srv = sess->srv;
struct rtrs_srv_op *id;
int i;
int i, ret;
sess->ops_ids = kcalloc(srv->queue_depth, sizeof(*sess->ops_ids),
GFP_KERNEL);
@ -144,8 +151,14 @@ static int rtrs_srv_alloc_ops_ids(struct rtrs_srv_sess *sess)
sess->ops_ids[i] = id;
}
init_waitqueue_head(&sess->ids_waitq);
atomic_set(&sess->ids_inflight, 0);
ret = percpu_ref_init(&sess->ids_inflight_ref,
rtrs_srv_inflight_ref_release, 0, GFP_KERNEL);
if (ret) {
pr_err("Percpu reference init failed\n");
goto err;
}
init_completion(&sess->complete_done);
return 0;
@ -156,21 +169,14 @@ err:
static inline void rtrs_srv_get_ops_ids(struct rtrs_srv_sess *sess)
{
atomic_inc(&sess->ids_inflight);
percpu_ref_get(&sess->ids_inflight_ref);
}
static inline void rtrs_srv_put_ops_ids(struct rtrs_srv_sess *sess)
{
if (atomic_dec_and_test(&sess->ids_inflight))
wake_up(&sess->ids_waitq);
percpu_ref_put(&sess->ids_inflight_ref);
}
static void rtrs_srv_wait_ops_ids(struct rtrs_srv_sess *sess)
{
wait_event(sess->ids_waitq, !atomic_read(&sess->ids_inflight));
}
static void rtrs_srv_reg_mr_done(struct ib_cq *cq, struct ib_wc *wc)
{
struct rtrs_srv_con *con = to_srv_con(wc->qp->qp_context);
@ -1479,8 +1485,15 @@ static void rtrs_srv_close_work(struct work_struct *work)
rdma_disconnect(con->c.cm_id);
ib_drain_qp(con->c.qp);
}
/* Wait for all inflights */
rtrs_srv_wait_ops_ids(sess);
/*
* Degrade ref count to the usual model with a single shared
* atomic_t counter
*/
percpu_ref_kill(&sess->ids_inflight_ref);
/* Wait for all completion */
wait_for_completion(&sess->complete_done);
/* Notify upper layer if we are the last path */
rtrs_srv_sess_down(sess);

View File

@ -81,8 +81,8 @@ struct rtrs_srv_sess {
spinlock_t state_lock;
int cur_cq_vector;
struct rtrs_srv_op **ops_ids;
atomic_t ids_inflight;
wait_queue_head_t ids_waitq;
struct percpu_ref ids_inflight_ref;
struct completion complete_done;
struct rtrs_srv_mr *mrs;
unsigned int mrs_num;
dma_addr_t *dma_addr;