xiaomi-sm8450/android_kernel_xiaomi_sm8450
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

IB/qib: Use rdmavt version of post_send

Browse Source 
This patch removes the post_send and post_one_send from the qib driver.
The "posting" of sends will be done by rdmavt which will walk a WQE and
queue work. This patch will still provide the capability to schedule that
work as well as kick the progress. These are provided to the rdmavt layer.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
This commit is contained in:
Harish Chegondi
2016-01-22 13:07:42 -08:00
committed by Doug Ledford
parent 4bb88e5f84
commit db3ef0eb84
10 changed files with 81 additions and 291 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
drivers/infiniband/hw/qib/qib_ruc.c
View File

@@ -158,7 +158,7 @@ int qib_get_rwqe(struct rvt_qp *qp, int wr_id_only)
}
spin_lock_irqsave(&rq->lock, flags);
if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)) {
if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)) {
ret = 0;
goto unlock;
}
@@ -379,7 +379,7 @@ static void qib_ruc_loopback(struct rvt_qp *sqp)
/* Return if we are already busy processing a work request. */
if ((sqp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT)) ||
!(ib_qib_state_ops[sqp->state] & QIB_PROCESS_OR_FLUSH_SEND))
!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_OR_FLUSH_SEND))
goto unlock;
sqp->s_flags |= RVT_S_BUSY;
@@ -387,11 +387,11 @@ static void qib_ruc_loopback(struct rvt_qp *sqp)
again:
if (sqp->s_last == sqp->s_head)
goto clr_busy;
wqe = get_swqe_ptr(sqp, sqp->s_last);
wqe = rvt_get_swqe_ptr(sqp, sqp->s_last);
/* Return if it is not OK to start a new work reqeust. */
if (!(ib_qib_state_ops[sqp->state] & QIB_PROCESS_NEXT_SEND_OK)) {
if (!(ib_qib_state_ops[sqp->state] & QIB_FLUSH_SEND))
if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
if (!(ib_rvt_state_ops[sqp->state] & RVT_FLUSH_SEND))
goto clr_busy;
/* We are in the error state, flush the work request. */
send_status = IB_WC_WR_FLUSH_ERR;
@@ -409,7 +409,7 @@ again:
}
spin_unlock_irqrestore(&sqp->s_lock, flags);
if (!qp || !(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) ||
if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) ||
qp->ibqp.qp_type != sqp->ibqp.qp_type) {
ibp->rvp.n_pkt_drops++;
/*
@@ -590,7 +590,7 @@ rnr_nak:
if (sqp->s_rnr_retry_cnt < 7)
sqp->s_rnr_retry--;
spin_lock_irqsave(&sqp->s_lock, flags);
if (!(ib_qib_state_ops[sqp->state] & QIB_PROCESS_RECV_OK))
if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK))
goto clr_busy;
sqp->s_flags |= RVT_S_WAIT_RNR;
sqp->s_timer.function = qib_rc_rnr_retry;
@@ -711,19 +711,26 @@ void qib_make_ruc_header(struct rvt_qp *qp, struct qib_other_headers *ohdr,
this_cpu_inc(ibp->pmastats->n_unicast_xmit);
}
void _qib_do_send(struct work_struct *work)
{
struct qib_qp_priv *priv = container_of(work, struct qib_qp_priv,
s_work);
struct rvt_qp *qp = priv->owner;
qib_do_send(qp);
}
/**
* qib_do_send - perform a send on a QP
* @work: contains a pointer to the QP
* @qp: pointer to the QP
*
* Process entries in the send work queue until credit or queue is
* exhausted. Only allow one CPU to send a packet per QP (tasklet).
* Otherwise, two threads could send packets out of order.
*/
void qib_do_send(struct work_struct *work)
void qib_do_send(struct rvt_qp *qp)
{
struct qib_qp_priv *priv = container_of(work, struct qib_qp_priv,
s_work);
struct rvt_qp *qp = priv->owner;
struct qib_qp_priv *priv = qp->priv;
struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
struct qib_pportdata *ppd = ppd_from_ibp(ibp);
int (*make_req)(struct rvt_qp *qp);
@@ -780,7 +787,7 @@ void qib_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
u32 old_last, last;
unsigned i;
if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_OR_FLUSH_SEND))
if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
return;
for (i = 0; i < wqe->wr.num_sge; i++) {