IB/ipath: Fix RC and UC error handling
When errors are detected in RC, the QP should transition to the
IB_QPS_ERR state, not the IB_QPS_SQE state. Also, when the error is on
the responder side, the receive work completion error was incorrect
(remote vs. local).
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
diff --git a/drivers/infiniband/hw/ipath/ipath_ruc.c b/drivers/infiniband/hw/ipath/ipath_ruc.c
index 9e3fe61..c716a03 100644
--- a/drivers/infiniband/hw/ipath/ipath_ruc.c
+++ b/drivers/infiniband/hw/ipath/ipath_ruc.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2006, 2007 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2006, 2007, 2008 QLogic Corporation. All rights reserved.
* Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
*
* This software is available to you under a choice of one of two
@@ -140,20 +140,11 @@
goto bail;
bad_lkey:
+ memset(&wc, 0, sizeof(wc));
wc.wr_id = wqe->wr_id;
wc.status = IB_WC_LOC_PROT_ERR;
wc.opcode = IB_WC_RECV;
- wc.vendor_err = 0;
- wc.byte_len = 0;
- wc.imm_data = 0;
wc.qp = &qp->ibqp;
- wc.src_qp = 0;
- wc.wc_flags = 0;
- wc.pkey_index = 0;
- wc.slid = 0;
- wc.sl = 0;
- wc.dlid_path_bits = 0;
- wc.port_num = 0;
/* Signal solicited completion event. */
ipath_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
ret = 0;
@@ -270,6 +261,7 @@
struct ib_wc wc;
u64 sdata;
atomic64_t *maddr;
+ enum ib_wc_status send_status;
qp = ipath_lookup_qpn(&dev->qp_table, sqp->remote_qpn);
if (!qp) {
@@ -300,8 +292,8 @@
wqe = get_swqe_ptr(sqp, sqp->s_last);
spin_unlock_irqrestore(&sqp->s_lock, flags);
- wc.wc_flags = 0;
- wc.imm_data = 0;
+ memset(&wc, 0, sizeof wc);
+ send_status = IB_WC_SUCCESS;
sqp->s_sge.sge = wqe->sg_list[0];
sqp->s_sge.sg_list = wqe->sg_list + 1;
@@ -313,75 +305,33 @@
wc.imm_data = wqe->wr.ex.imm_data;
/* FALLTHROUGH */
case IB_WR_SEND:
- if (!ipath_get_rwqe(qp, 0)) {
- rnr_nak:
- /* Handle RNR NAK */
- if (qp->ibqp.qp_type == IB_QPT_UC)
- goto send_comp;
- if (sqp->s_rnr_retry == 0) {
- wc.status = IB_WC_RNR_RETRY_EXC_ERR;
- goto err;
- }
- if (sqp->s_rnr_retry_cnt < 7)
- sqp->s_rnr_retry--;
- dev->n_rnr_naks++;
- sqp->s_rnr_timeout =
- ib_ipath_rnr_table[qp->r_min_rnr_timer];
- ipath_insert_rnr_queue(sqp);
- goto done;
- }
+ if (!ipath_get_rwqe(qp, 0))
+ goto rnr_nak;
break;
case IB_WR_RDMA_WRITE_WITH_IMM:
- if (unlikely(!(qp->qp_access_flags &
- IB_ACCESS_REMOTE_WRITE))) {
- wc.status = IB_WC_REM_INV_REQ_ERR;
- goto err;
- }
+ if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
+ goto inv_err;
wc.wc_flags = IB_WC_WITH_IMM;
wc.imm_data = wqe->wr.ex.imm_data;
if (!ipath_get_rwqe(qp, 1))
goto rnr_nak;
/* FALLTHROUGH */
case IB_WR_RDMA_WRITE:
- if (unlikely(!(qp->qp_access_flags &
- IB_ACCESS_REMOTE_WRITE))) {
- wc.status = IB_WC_REM_INV_REQ_ERR;
- goto err;
- }
+ if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
+ goto inv_err;
if (wqe->length == 0)
break;
if (unlikely(!ipath_rkey_ok(qp, &qp->r_sge, wqe->length,
wqe->wr.wr.rdma.remote_addr,
wqe->wr.wr.rdma.rkey,
- IB_ACCESS_REMOTE_WRITE))) {
- acc_err:
- wc.status = IB_WC_REM_ACCESS_ERR;
- err:
- wc.wr_id = wqe->wr.wr_id;
- wc.opcode = ib_ipath_wc_opcode[wqe->wr.opcode];
- wc.vendor_err = 0;
- wc.byte_len = 0;
- wc.qp = &sqp->ibqp;
- wc.src_qp = sqp->remote_qpn;
- wc.pkey_index = 0;
- wc.slid = sqp->remote_ah_attr.dlid;
- wc.sl = sqp->remote_ah_attr.sl;
- wc.dlid_path_bits = 0;
- wc.port_num = 0;
- spin_lock_irqsave(&sqp->s_lock, flags);
- ipath_sqerror_qp(sqp, &wc);
- spin_unlock_irqrestore(&sqp->s_lock, flags);
- goto done;
- }
+ IB_ACCESS_REMOTE_WRITE)))
+ goto acc_err;
break;
case IB_WR_RDMA_READ:
- if (unlikely(!(qp->qp_access_flags &
- IB_ACCESS_REMOTE_READ))) {
- wc.status = IB_WC_REM_INV_REQ_ERR;
- goto err;
- }
+ if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
+ goto inv_err;
if (unlikely(!ipath_rkey_ok(qp, &sqp->s_sge, wqe->length,
wqe->wr.wr.rdma.remote_addr,
wqe->wr.wr.rdma.rkey,
@@ -394,11 +344,8 @@
case IB_WR_ATOMIC_CMP_AND_SWP:
case IB_WR_ATOMIC_FETCH_AND_ADD:
- if (unlikely(!(qp->qp_access_flags &
- IB_ACCESS_REMOTE_ATOMIC))) {
- wc.status = IB_WC_REM_INV_REQ_ERR;
- goto err;
- }
+ if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
+ goto inv_err;
if (unlikely(!ipath_rkey_ok(qp, &qp->r_sge, sizeof(u64),
wqe->wr.wr.atomic.remote_addr,
wqe->wr.wr.atomic.rkey,
@@ -415,7 +362,8 @@
goto send_comp;
default:
- goto done;
+ send_status = IB_WC_LOC_QP_OP_ERR;
+ goto serr;
}
sge = &sqp->s_sge.sge;
@@ -458,14 +406,11 @@
wc.opcode = IB_WC_RECV;
wc.wr_id = qp->r_wr_id;
wc.status = IB_WC_SUCCESS;
- wc.vendor_err = 0;
wc.byte_len = wqe->length;
wc.qp = &qp->ibqp;
wc.src_qp = qp->remote_qpn;
- wc.pkey_index = 0;
wc.slid = qp->remote_ah_attr.dlid;
wc.sl = qp->remote_ah_attr.sl;
- wc.dlid_path_bits = 0;
wc.port_num = 1;
/* Signal completion event if the solicited bit is set. */
ipath_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
@@ -473,9 +418,63 @@
send_comp:
sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
- ipath_send_complete(sqp, wqe, IB_WC_SUCCESS);
+ ipath_send_complete(sqp, wqe, send_status);
goto again;
+rnr_nak:
+ /* Handle RNR NAK */
+ if (qp->ibqp.qp_type == IB_QPT_UC)
+ goto send_comp;
+ /*
+ * Note: we don't need the s_lock held since the BUSY flag
+ * makes this single threaded.
+ */
+ if (sqp->s_rnr_retry == 0) {
+ send_status = IB_WC_RNR_RETRY_EXC_ERR;
+ goto serr;
+ }
+ if (sqp->s_rnr_retry_cnt < 7)
+ sqp->s_rnr_retry--;
+ spin_lock_irqsave(&sqp->s_lock, flags);
+ if (!(ib_ipath_state_ops[sqp->state] & IPATH_PROCESS_RECV_OK))
+ goto unlock;
+ dev->n_rnr_naks++;
+ sqp->s_rnr_timeout = ib_ipath_rnr_table[qp->r_min_rnr_timer];
+ ipath_insert_rnr_queue(sqp);
+ goto unlock;
+
+inv_err:
+ send_status = IB_WC_REM_INV_REQ_ERR;
+ wc.status = IB_WC_LOC_QP_OP_ERR;
+ goto err;
+
+acc_err:
+ send_status = IB_WC_REM_ACCESS_ERR;
+ wc.status = IB_WC_LOC_PROT_ERR;
+err:
+ /* responder goes to error state */
+ ipath_rc_error(qp, wc.status);
+
+serr:
+ spin_lock_irqsave(&sqp->s_lock, flags);
+ ipath_send_complete(sqp, wqe, send_status);
+ if (sqp->ibqp.qp_type == IB_QPT_RC) {
+ int lastwqe = ipath_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
+
+ sqp->s_flags &= ~IPATH_S_BUSY;
+ spin_unlock_irqrestore(&sqp->s_lock, flags);
+ if (lastwqe) {
+ struct ib_event ev;
+
+ ev.device = sqp->ibqp.device;
+ ev.element.qp = &sqp->ibqp;
+ ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
+ sqp->ibqp.event_handler(&ev, sqp->ibqp.qp_context);
+ }
+ goto done;
+ }
+unlock:
+ spin_unlock_irqrestore(&sqp->s_lock, flags);
done:
if (atomic_dec_and_test(&qp->refcount))
wake_up(&qp->wait);
@@ -651,21 +650,15 @@
status != IB_WC_SUCCESS) {
struct ib_wc wc;
+ memset(&wc, 0, sizeof wc);
wc.wr_id = wqe->wr.wr_id;
wc.status = status;
wc.opcode = ib_ipath_wc_opcode[wqe->wr.opcode];
- wc.vendor_err = 0;
- wc.byte_len = wqe->length;
- wc.imm_data = 0;
wc.qp = &qp->ibqp;
- wc.src_qp = 0;
- wc.wc_flags = 0;
- wc.pkey_index = 0;
- wc.slid = 0;
- wc.sl = 0;
- wc.dlid_path_bits = 0;
- wc.port_num = 0;
- ipath_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
+ if (status == IB_WC_SUCCESS)
+ wc.byte_len = wqe->length;
+ ipath_cq_enter(to_icq(qp->ibqp.send_cq), &wc,
+ status != IB_WC_SUCCESS);
}
spin_lock_irqsave(&qp->s_lock, flags);