IB/{hfi1, qib, rdmavt}: Put qp in error state when cq is full

When a completion queue is full, the associated queue pairs are not put
into the error state. According to the IBTA specification, this is a
violation.

Quote from IBTA spec:
C9-218: A Requester Class F error occurs when the CQ is inaccessible or
full and an attempt is made to complete a WQE.  The Affected QP shall be
moved to the error state and affiliated asynchronous errors generated as
described in 11.6.3.1 Affiliated Asynchronous Events on page 678. The
current WQE and any subsequent WQEs are left in an unknown state.

C11-37: The CI shall generate a CQ Error when a CQ overrun is
detected. This condition will result in an Affiliated Asynchronous Error
for any associated Work Queues when they attempt to use that
CQ. Completions can no longer be added to the CQ. It is not guaranteed
that completions present in the CQ at the time the error occurred can be
retrieved. Possible causes include a CQ overrun or a CQ protection error.

Put the qp in error state when cq is full. Implement a state called full
to continue to put other associated QPs in error state.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Kamenee Arumugam <kamenee.arumugam@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

11 files changed, 75 insertions, 25 deletions

diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c
index 235bdbc706ac..0477c14633ab 100644
--- a/drivers/infiniband/hw/hfi1/rc.c
+++ b/drivers/infiniband/hw/hfi1/rc.c
@@ -3008,8 +3008,7 @@ send_last:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-			     ib_bth_is_solicited(ohdr));
+		rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr));
 		break;
 
 	case OP(RDMA_WRITE_ONLY):
diff --git a/drivers/infiniband/hw/hfi1/uc.c b/drivers/infiniband/hw/hfi1/uc.c
index 4ed4fcfabd6c..0c77f18120ed 100644
--- a/drivers/infiniband/hw/hfi1/uc.c
+++ b/drivers/infiniband/hw/hfi1/uc.c
@@ -476,8 +476,7 @@ last_imm:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-			     ib_bth_is_solicited(ohdr));
+		rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr));
 		break;
 
 	case OP(RDMA_WRITE_FIRST):
diff --git a/drivers/infiniband/hw/hfi1/ud.c b/drivers/infiniband/hw/hfi1/ud.c
index 4cb0fce5c096..e16d499cfd1e 100644
--- a/drivers/infiniband/hw/hfi1/ud.c
+++ b/drivers/infiniband/hw/hfi1/ud.c
@@ -255,8 +255,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	wc.dlid_path_bits = rdma_ah_get_dlid(ah_attr) & ((1 << ppd->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-		     swqe->wr.send_flags & IB_SEND_SOLICITED);
+	rvt_recv_cq(qp, &wc, swqe->wr.send_flags & IB_SEND_SOLICITED);
 	ibp->rvp.n_loop_pkts++;
 bail_unlock:
 	spin_unlock_irqrestore(&qp->r_lock, flags);
@@ -1061,7 +1060,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 		dlid & ((1 << ppd_from_ibp(ibp)->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, solicited);
+	rvt_recv_cq(qp, &wc, solicited);
 	return;
 
 drop:
diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
index 8d9a94d6f685..1d5e2d4ee257 100644
--- a/drivers/infiniband/hw/qib/qib_rc.c
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -1891,8 +1891,7 @@ send_last:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-			     ib_bth_is_solicited(ohdr));
+		rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr));
 		break;
 
 	case OP(RDMA_WRITE_FIRST):
diff --git a/drivers/infiniband/hw/qib/qib_uc.c b/drivers/infiniband/hw/qib/qib_uc.c
index 30c70ad0f4bf..e17b91e2c22a 100644
--- a/drivers/infiniband/hw/qib/qib_uc.c
+++ b/drivers/infiniband/hw/qib/qib_uc.c
@@ -400,8 +400,7 @@ last_imm:
 		wc.dlid_path_bits = 0;
 		wc.port_num = 0;
 		/* Signal completion event if the solicited bit is set. */
-		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-			     ib_bth_is_solicited(ohdr));
+		rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr));
 		break;
 
 	case OP(RDMA_WRITE_FIRST):
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
index 5cdedba2d164..32ad0b635fc6 100644
--- a/drivers/infiniband/hw/qib/qib_ud.c
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -210,8 +210,7 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 	wc.dlid_path_bits = rdma_ah_get_dlid(ah_attr) & ((1 << ppd->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-		     swqe->wr.send_flags & IB_SEND_SOLICITED);
+	rvt_recv_cq(qp, &wc, swqe->wr.send_flags & IB_SEND_SOLICITED);
 	ibp->rvp.n_loop_pkts++;
 bail_unlock:
 	spin_unlock_irqrestore(&qp->r_lock, flags);
@@ -573,8 +572,7 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct ib_header *hdr,
 		dlid & ((1 << ppd_from_ibp(ibp)->lmc) - 1);
 	wc.port_num = qp->port_num;
 	/* Signal completion event if the solicited bit is set. */
-	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-		     ib_bth_is_solicited(ohdr));
+	rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr));
 	return;
 
 drop:
diff --git a/drivers/infiniband/sw/rdmavt/cq.c b/drivers/infiniband/sw/rdmavt/cq.c
index 2602ad8b8cb0..fac87b13329d 100644
--- a/drivers/infiniband/sw/rdmavt/cq.c
+++ b/drivers/infiniband/sw/rdmavt/cq.c
@@ -60,8 +60,11 @@ static struct workqueue_struct *comp_vector_wq;
  * @solicited: true if @entry is solicited
  *
  * This may be called with qp->s_lock held.
+ *
+ * Return: return true on success, else return
+ * false if cq is full.
  */
-void rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited)
+bool rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited)
 {
 	struct ib_uverbs_wc *uqueue = NULL;
 	struct ib_wc *kqueue = NULL;
@@ -97,7 +100,12 @@ void rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited)
 		next = head + 1;
 	}
 
-	if (unlikely(next == tail)) {
+	if (unlikely(next == tail || cq->cq_full)) {
+		struct rvt_dev_info *rdi = cq->rdi;
+
+		if (!cq->cq_full)
+			rvt_pr_err_ratelimited(rdi, "CQ is full!\n");
+		cq->cq_full = true;
 		spin_unlock_irqrestore(&cq->lock, flags);
 		if (cq->ibcq.event_handler) {
 			struct ib_event ev;
@@ -107,7 +115,7 @@ void rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited)
 			ev.event = IB_EVENT_CQ_ERR;
 			cq->ibcq.event_handler(&ev, cq->ibcq.cq_context);
 		}
-		return;
+		return false;
 	}
 	trace_rvt_cq_enter(cq, entry, head);
 	if (uqueue) {
@@ -146,6 +154,7 @@ void rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited)
 	}
 
 	spin_unlock_irqrestore(&cq->lock, flags);
+	return true;
 }
 EXPORT_SYMBOL(rvt_cq_enter);
 
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index 200b292be63e..17e192a2c8b6 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -3103,8 +3103,7 @@ do_write:
 	wc.sl = rdma_ah_get_sl(&qp->remote_ah_attr);
 	wc.port_num = 1;
 	/* Signal completion event if the solicited bit is set. */
-	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-		     wqe->wr.send_flags & IB_SEND_SOLICITED);
+	rvt_recv_cq(qp, &wc, wqe->wr.send_flags & IB_SEND_SOLICITED);
 
 send_comp:
 	spin_unlock_irqrestore(&qp->r_lock, flags);
diff --git a/drivers/infiniband/sw/rdmavt/vt.h b/drivers/infiniband/sw/rdmavt/vt.h
index 0675ea6c3872..d19ff817c2c7 100644
--- a/drivers/infiniband/sw/rdmavt/vt.h
+++ b/drivers/infiniband/sw/rdmavt/vt.h
@@ -78,6 +78,12 @@
 		     fmt, \
 		     ##__VA_ARGS__)
 
+#define rvt_pr_err_ratelimited(rdi, fmt, ...) \
+	__rvt_pr_err_ratelimited((rdi)->driver_f.get_pci_dev(rdi), \
+				 rvt_get_ibdev_name(rdi), \
+				 fmt, \
+				 ##__VA_ARGS__)
+
 #define __rvt_pr_info(pdev, name, fmt, ...) \
 	dev_info(&pdev->dev, "%s: " fmt, name, ##__VA_ARGS__)
 
@@ -87,6 +93,9 @@
 #define __rvt_pr_err(pdev, name, fmt, ...) \
 	dev_err(&pdev->dev, "%s: " fmt, name, ##__VA_ARGS__)
 
+#define __rvt_pr_err_ratelimited(pdev, name, fmt, ...) \
+	dev_err_ratelimited(&(pdev)->dev, "%s: " fmt, name, ##__VA_ARGS__)
+
 static inline int ibport_num_to_idx(struct ib_device *ibdev, u8 port_num)
 {
 	struct rvt_dev_info *rdi = ib_to_rvt(ibdev);
diff --git a/include/rdma/rdmavt_cq.h b/include/rdma/rdmavt_cq.h
index ab22860a63e2..04c519ef6d71 100644
--- a/include/rdma/rdmavt_cq.h
+++ b/include/rdma/rdmavt_cq.h
@@ -93,6 +93,7 @@ struct rvt_cq {
 	spinlock_t lock; /* protect changes in this struct */
 	u8 notify;
 	u8 triggered;
+	u8 cq_full;
 	int comp_vector_cpu;
 	struct rvt_dev_info *rdi;
 	struct rvt_cq_wc *queue;
@@ -105,6 +106,6 @@ static inline struct rvt_cq *ibcq_to_rvtcq(struct ib_cq *ibcq)
 	return container_of(ibcq, struct rvt_cq, ibcq);
 }
 
-void rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited);
+bool rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited);
 
 #endif          /* DEF_RDMAVT_INCCQH */
diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h
index de5915b244be..e4be869c4f21 100644
--- a/include/rdma/rdmavt_qp.h
+++ b/include/rdma/rdmavt_qp.h
@@ -718,6 +718,48 @@ rvt_qp_swqe_incr(struct rvt_qp *qp, u32 val)
 	return val;
 }
 
+int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err);
+
+/**
+ * rvt_recv_cq - add a new entry to completion queue
+ *			by receive queue
+ * @qp: receive queue
+ * @wc: work completion entry to add
+ * @solicited: true if @entry is solicited
+ *
+ * This is wrapper function for rvt_enter_cq function call by
+ * receive queue. If rvt_cq_enter return false, it means cq is
+ * full and the qp is put into error state.
+ */
+static inline void rvt_recv_cq(struct rvt_qp *qp, struct ib_wc *wc,
+			       bool solicited)
+{
+	struct rvt_cq *cq = ibcq_to_rvtcq(qp->ibqp.recv_cq);
+
+	if (unlikely(!rvt_cq_enter(cq, wc, solicited)))
+		rvt_error_qp(qp, IB_WC_LOC_QP_OP_ERR);
+}
+
+/**
+ * rvt_send_cq - add a new entry to completion queue
+ *                        by send queue
+ * @qp: send queue
+ * @wc: work completion entry to add
+ * @solicited: true if @entry is solicited
+ *
+ * This is wrapper function for rvt_enter_cq function call by
+ * send queue. If rvt_cq_enter return false, it means cq is
+ * full and the qp is put into error state.
+ */
+static inline void rvt_send_cq(struct rvt_qp *qp, struct ib_wc *wc,
+			       bool solicited)
+{
+	struct rvt_cq *cq = ibcq_to_rvtcq(qp->ibqp.send_cq);
+
+	if (unlikely(!rvt_cq_enter(cq, wc, solicited)))
+		rvt_error_qp(qp, IB_WC_LOC_QP_OP_ERR);
+}
+
 /**
  * rvt_qp_complete_swqe - insert send completion
  * @qp - the qp
@@ -768,9 +810,7 @@ rvt_qp_complete_swqe(struct rvt_qp *qp,
 			.qp = &qp->ibqp,
 			.byte_len = byte_len,
 		};
-
-		rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.send_cq), &w,
-			     status != IB_WC_SUCCESS);
+		rvt_send_cq(qp, &w, status != IB_WC_SUCCESS);
 	}
 	return last;
 }
@@ -780,7 +820,6 @@ extern const int  ib_rvt_state_ops[];
 struct rvt_dev_info;
 int rvt_get_rwqe(struct rvt_qp *qp, bool wr_id_only);
 void rvt_comm_est(struct rvt_qp *qp);
-int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err);
 void rvt_rc_error(struct rvt_qp *qp, enum ib_wc_status err);
 unsigned long rvt_rnr_tbl_to_usec(u32 index);
 enum hrtimer_restart rvt_rc_rnr_retry(struct hrtimer *t);