openkylin
/
linux
mirror of https://gitee.com/openkylin/linux.git
9
0
Fork 0
Code Issues Projects Releases Wiki Activity

Revert "RDS: IB: split the mr registration and invalidation path" 

This reverts commit 5601245931.

RDS kept spinning inside function "rds_ib_post_reg_frmr", waiting for
"i_fastreg_wrs" to become incremented:
         while (atomic_dec_return(&ibmr->ic->i_fastreg_wrs) <= 0) {
                 atomic_inc(&ibmr->ic->i_fastreg_wrs);
                 cpu_relax();
         }

Looking at the original commit:

commit 5601245931 ("RDS: IB: split the mr registration and
invalidation path")

In there, the "rds_ib_mr_cqe_handler" was changed in the following
way:

 void rds_ib_mr_cqe_handler(struct
 rds_ib_connection *ic,
 struct ib_wc *wc)
        if (frmr->fr_inv) {
                  frmr->fr_state = FRMR_IS_FREE;
                  frmr->fr_inv = false;
                atomic_inc(&ic->i_fastreg_wrs);
        } else {
                atomic_inc(&ic->i_fastunreg_wrs);
        }

It looks like it's got it exactly backwards:

Function "rds_ib_post_reg_frmr" keeps track of the outstanding
requests via "i_fastreg_wrs".

Function "rds_ib_post_inv" keeps track of the outstanding requests
via "i_fastunreg_wrs" (post original commit). It also sets:
         frmr->fr_inv = true;

However the completion handler "rds_ib_mr_cqe_handler" adjusts
"i_fastreg_wrs" when "fr_inv" had been true, and adjusts
"i_fastunreg_wrs" otherwise.

The original commit was done in the name of performance:
to remove the performance bottleneck

No performance benefit could be observed with a fixed-up version
of the original commit measured between two Oracle X7 servers,
both equipped with Mellanox Connect-X5 HCAs.

The prudent course of action is to revert this commit.

Signed-off-by: Gerd Rausch <gerd.rausch@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
This commit is contained in:
Gerd Rausch 2019-06-28 17:31:19 -07:00 committed by Santosh Shilimkar
parent 616d37a070
commit a552078847
3 changed files with 8 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
net/rds/ib.h
View File

@ -15,8 +15,7 @@
#define RDS_IB_DEFAULT_RECV_WR 1024
#define RDS_IB_DEFAULT_SEND_WR 256
#define RDS_IB_DEFAULT_FR_WR 256
#define RDS_IB_DEFAULT_FR_INV_WR 256
#define RDS_IB_DEFAULT_FR_WR 512
#define RDS_IB_DEFAULT_RETRY_COUNT 1
@ -157,7 +156,6 @@ struct rds_ib_connection {
/* To control the number of wrs from fastreg */
atomic_t i_fastreg_wrs;
atomic_t i_fastunreg_wrs;
/* interrupt handling */
struct tasklet_struct i_send_tasklet;

9
net/rds/ib_cm.c
View File

@ -460,10 +460,7 @@ static int rds_ib_setup_qp(struct rds_connection *conn)
* completion queue and send queue. This extra space is used for FRMR
* registration and invalidation work requests
*/
fr_queue_space = rds_ibdev->use_fastreg ?
(RDS_IB_DEFAULT_FR_WR + 1) +
(RDS_IB_DEFAULT_FR_INV_WR + 1)
: 0;
fr_queue_space = (rds_ibdev->use_fastreg ? RDS_IB_DEFAULT_FR_WR : 0);
/* add the conn now so that connection establishment has the dev */
rds_ib_add_conn(rds_ibdev, conn);
@ -530,7 +527,6 @@ static int rds_ib_setup_qp(struct rds_connection *conn)
attr.send_cq = ic->i_send_cq;
attr.recv_cq = ic->i_recv_cq;
atomic_set(&ic->i_fastreg_wrs, RDS_IB_DEFAULT_FR_WR);
atomic_set(&ic->i_fastunreg_wrs, RDS_IB_DEFAULT_FR_INV_WR);
/*
* XXX this can fail if max_*_wr is too large? Are we supposed
@ -1009,8 +1005,7 @@ void rds_ib_conn_path_shutdown(struct rds_conn_path *cp)
wait_event(rds_ib_ring_empty_wait,
rds_ib_ring_empty(&ic->i_recv_ring) &&
(atomic_read(&ic->i_signaled_sends) == 0) &&
(atomic_read(&ic->i_fastreg_wrs) == RDS_IB_DEFAULT_FR_WR) &&
(atomic_read(&ic->i_fastunreg_wrs) == RDS_IB_DEFAULT_FR_INV_WR));
(atomic_read(&ic->i_fastreg_wrs) == RDS_IB_DEFAULT_FR_WR));
tasklet_kill(&ic->i_send_tasklet);
tasklet_kill(&ic->i_recv_tasklet);

11
net/rds/ib_frmr.c
View File

@ -239,8 +239,8 @@ static int rds_ib_post_inv(struct rds_ib_mr *ibmr)
if (frmr->fr_state != FRMR_IS_INUSE)
goto out;
while (atomic_dec_return(&ibmr->ic->i_fastunreg_wrs) <= 0) {
atomic_inc(&ibmr->ic->i_fastunreg_wrs);
while (atomic_dec_return(&ibmr->ic->i_fastreg_wrs) <= 0) {
atomic_inc(&ibmr->ic->i_fastreg_wrs);
cpu_relax();
}
@ -257,7 +257,7 @@ static int rds_ib_post_inv(struct rds_ib_mr *ibmr)
if (unlikely(ret)) {
frmr->fr_state = FRMR_IS_STALE;
frmr->fr_inv = false;
atomic_inc(&ibmr->ic->i_fastunreg_wrs);
atomic_inc(&ibmr->ic->i_fastreg_wrs);
pr_err("RDS/IB: %s returned error(%d)\n", __func__, ret);
goto out;
}
@ -285,10 +285,9 @@ void rds_ib_mr_cqe_handler(struct rds_ib_connection *ic, struct ib_wc *wc)
if (frmr->fr_inv) {
frmr->fr_state = FRMR_IS_FREE;
frmr->fr_inv = false;
atomic_inc(&ic->i_fastreg_wrs);
} else {
atomic_inc(&ic->i_fastunreg_wrs);
}
atomic_inc(&ic->i_fastreg_wrs);
}
void rds_ib_unreg_frmr(struct list_head *list, unsigned int *nfreed,