RDMA/cxgb4: Fix QP flush logic

This patch makes following fixes in QP flush logic:

- correctly flushes unsignaled WRs followed by a signaled WR
- supports for flushing a CQ bound to multiple QPs
- resets cidx_flush if a active queue starts getting HW CQEs again
- marks WQ in error when we leave RTS. This was only being done for
  user queues, but we need it for kernel queues too so that
  post_send/post_recv will start returning the appropriate error
  synchronously
- eats unsignaled read resp CQEs. HW always inserts CQEs so we must
  silently discard them if the read work request was unsignaled.
- handles QP flushes with pending SW CQEs. The flush and out of order
  completion logic has a bug where if out of order completions are
  flushed but not yet polled by the consumer and the qp is then
  flushed then we end up inserting duplicate completions.
- c4iw_flush_sq() should only flush wrs that have not already been
  flushed.  Since we already track where in the SQ we've flushed via
  sq.cidx_flush, just start at that point and flush any remaining.
  This bug only caused a problem in the presence of unsignaled work
  requests.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Vipul Pandya <vipul@chelsio.com>

[ Fixed sparse warning due to htonl/ntohl confusion.  - Roland ]

Signed-off-by: Roland Dreier <roland@purestorage.com>

1 files changed, 22 insertions, 3 deletions

diff --git a/drivers/infiniband/hw/cxgb4/t4.h b/drivers/infiniband/hw/cxgb4/t4.h
index ebcb03bd1b72..3a6a289b9d3e 100644
--- a/drivers/infiniband/hw/cxgb4/t4.h
+++ b/drivers/infiniband/hw/cxgb4/t4.h
@@ -36,9 +36,9 @@
 #include "t4_msg.h"
 #include "t4fw_ri_api.h"
 
-#define T4_MAX_NUM_QP (1<<16)
-#define T4_MAX_NUM_CQ (1<<15)
-#define T4_MAX_NUM_PD (1<<15)
+#define T4_MAX_NUM_QP 65536
+#define T4_MAX_NUM_CQ 65536
+#define T4_MAX_NUM_PD 65536
 #define T4_EQ_STATUS_ENTRIES (L1_CACHE_BYTES > 64 ? 2 : 1)
 #define T4_MAX_EQ_SIZE (65520 - T4_EQ_STATUS_ENTRIES)
 #define T4_MAX_IQ_SIZE (65520 - 1)
@@ -269,6 +269,7 @@ struct t4_swsqe {
 	int			complete;
 	int			signaled;
 	u16			idx;
+	int                     flushed;
 };
 
 static inline pgprot_t t4_pgprot_wc(pgprot_t prot)
@@ -300,6 +301,7 @@ struct t4_sq {
 	u16 pidx;
 	u16 wq_pidx;
 	u16 flags;
+	short flush_cidx;
 };
 
 struct t4_swrqe {
@@ -330,6 +332,7 @@ struct t4_wq {
 	void __iomem *db;
 	void __iomem *gts;
 	struct c4iw_rdev *rdev;
+	int flushed;
 };
 
 static inline int t4_rqes_posted(struct t4_wq *wq)
@@ -412,6 +415,9 @@ static inline void t4_sq_produce(struct t4_wq *wq, u8 len16)
 
 static inline void t4_sq_consume(struct t4_wq *wq)
 {
+	BUG_ON(wq->sq.in_use < 1);
+	if (wq->sq.cidx == wq->sq.flush_cidx)
+		wq->sq.flush_cidx = -1;
 	wq->sq.in_use--;
 	if (++wq->sq.cidx == wq->sq.size)
 		wq->sq.cidx = 0;
@@ -505,12 +511,18 @@ static inline int t4_arm_cq(struct t4_cq *cq, int se)
 static inline void t4_swcq_produce(struct t4_cq *cq)
 {
 	cq->sw_in_use++;
+	if (cq->sw_in_use == cq->size) {
+		PDBG("%s cxgb4 sw cq overflow cqid %u\n", __func__, cq->cqid);
+		cq->error = 1;
+		BUG_ON(1);
+	}
 	if (++cq->sw_pidx == cq->size)
 		cq->sw_pidx = 0;
 }
 
 static inline void t4_swcq_consume(struct t4_cq *cq)
 {
+	BUG_ON(cq->sw_in_use < 1);
 	cq->sw_in_use--;
 	if (++cq->sw_cidx == cq->size)
 		cq->sw_cidx = 0;
@@ -552,6 +564,7 @@ static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
 		ret = -EOVERFLOW;
 		cq->error = 1;
 		printk(KERN_ERR MOD "cq overflow cqid %u\n", cq->cqid);
+		BUG_ON(1);
 	} else if (t4_valid_cqe(cq, &cq->queue[cq->cidx])) {
 		*cqe = &cq->queue[cq->cidx];
 		ret = 0;
@@ -562,6 +575,12 @@ static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
 
 static inline struct t4_cqe *t4_next_sw_cqe(struct t4_cq *cq)
 {
+	if (cq->sw_in_use == cq->size) {
+		PDBG("%s cxgb4 sw cq overflow cqid %u\n", __func__, cq->cqid);
+		cq->error = 1;
+		BUG_ON(1);
+		return NULL;
+	}
 	if (cq->sw_in_use)
 		return &cq->sw_queue[cq->sw_cidx];
 	return NULL;