[PATCH 0/8] Critical bug fixes for RDMA/cxgb4

[PATCH 0/8] Critical bug fixes for RDMA/cxgb4

[Vipul Pandya]

This patch series fixes critical bugs for RDMA/cxgb4. It fixes bugs in QP flush
path, handles newer firmware changes, deals with wrap around issue of the queue,
sets arp error handler etc.

The patch-series is based on Roland's infiniband tree for-next branch.

This series is on top of previously submitted series "Add IPv6 support for
iWARP" which can be found at following location:
http://comments.gmane.org/gmane.linux.network/272124
We request Roland to merge above series first and then apply this series on top
of it.

Please review the series and let us know in case of any review comments.

Thanks,
Vipul Pandya

Steve Wise (8):
  RDMA/cxgb4: Use correct bit shift macros for vlan filter tuples
  RDMA/cxgb4: Handle newer firmware changes
  RDMA/cxgb4: Fix QP flush logic
  RDMA/cxgb4: Account for unsignled SQ WRs that fails to deal with wrap
  RDMA/cxgb4: Set arp error handler for PASS_ACCEPT_RPL messages.
  RDMA/cxgb4: Always do GTS write if cidx_inc == CIDXINC_MASK.
  RDMA/cxgb4: Advertise ~0ULL as max MR size.
  RDMA/cxgb4: Issue RI.FINI before closing when entering TERM

 drivers/infiniband/hw/cxgb4/cm.c       |   9 +-
 drivers/infiniband/hw/cxgb4/cq.c       | 329 +++++++++++++++++++++------------
 drivers/infiniband/hw/cxgb4/ev.c       |  10 -
 drivers/infiniband/hw/cxgb4/iw_cxgb4.h |   5 +-
 drivers/infiniband/hw/cxgb4/qp.c       |  41 ++--
 drivers/infiniband/hw/cxgb4/t4.h       |  29 ++-
 6 files changed, 270 insertions(+), 153 deletions(-)

-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 1/8] RDMA/cxgb4: Use correct bit shift macros for vlan filter tuples

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/cm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
index 9423441..31dee27 100644
--- a/drivers/infiniband/hw/cxgb4/cm.c
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -542,7 +542,7 @@ static unsigned int select_ntuple(struct c4iw_dev *dev, struct dst_entry *dst,
 			ntuple |= FILTER_SEL_VLAN_NONE << FILTER_SEL_WIDTH_P_FC;
 		else {
 			ntuple |= l2t->vlan << FILTER_SEL_WIDTH_P_FC;
-			ntuple |= 1 << FILTER_SEL_WIDTH_VLD_TAG_P_FC;
+			ntuple |= 1 << FILTER_SEL_WIDTH_TAG_P_FC;
 		}
 		ntuple |= l2t->lport << S_PORT | IPPROTO_TCP <<
 			  FILTER_SEL_WIDTH_VLD_TAG_P_FC;
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 2/8] RDMA/cxgb4: Handle newer firmware changes

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

Move QP to TERMINATE instead to allow the peer to get the TERM message. This
bug wasn't detectable until newer FW that moves connections out of RDMA mode as
soon as an error is detected.

QP can exit RTS before the last AE arrives. This was introduced by changes in
the FW to kick connections out of RDMA mode as soon as an error is detected. A
side effect of this is that the driver can move the QP out of RTS before the AE
causing the connection to get kicked out of RDMA mode is processed. Fix for
this is to always post async errors even if the QP is out of RTS.

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
Signed-off-by: Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/cm.c |  5 ++---
 drivers/infiniband/hw/cxgb4/ev.c | 10 ----------
 2 files changed, 2 insertions(+), 13 deletions(-)

diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
index 31dee27..1b0fd72 100644
--- a/drivers/infiniband/hw/cxgb4/cm.c
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -1568,10 +1568,9 @@ static int rx_data(struct c4iw_dev *dev, struct sk_buff *skb)
 			       " qpid %u ep %p state %d tid %u status %d\n",
 			       __func__, ep->com.qp->wq.sq.qid, ep,
 			       state_read(&ep->com), ep->hwtid, status);
-		attrs.next_state = C4IW_QP_STATE_ERROR;
+		attrs.next_state = C4IW_QP_STATE_TERMINATE;
 		c4iw_modify_qp(ep->com.qp->rhp, ep->com.qp,
-			       C4IW_QP_ATTR_NEXT_STATE, &attrs, 1);
-		c4iw_ep_disconnect(ep, 1, GFP_KERNEL);
+			       C4IW_QP_ATTR_NEXT_STATE, &attrs, 0);
 		break;
 	}
 	default:
diff --git a/drivers/infiniband/hw/cxgb4/ev.c b/drivers/infiniband/hw/cxgb4/ev.c
index 1a840b2..d61d0a1 100644
--- a/drivers/infiniband/hw/cxgb4/ev.c
+++ b/drivers/infiniband/hw/cxgb4/ev.c
@@ -44,16 +44,6 @@ static void post_qp_event(struct c4iw_dev *dev, struct c4iw_cq *chp,
 	struct c4iw_qp_attributes attrs;
 	unsigned long flag;
 
-	if ((qhp->attr.state == C4IW_QP_STATE_ERROR) ||
-	    (qhp->attr.state == C4IW_QP_STATE_TERMINATE)) {
-		pr_err("%s AE after RTS - qpid 0x%x opcode %d status 0x%x "\
-		       "type %d wrid.hi 0x%x wrid.lo 0x%x\n",
-		       __func__, CQE_QPID(err_cqe), CQE_OPCODE(err_cqe),
-		       CQE_STATUS(err_cqe), CQE_TYPE(err_cqe),
-		       CQE_WRID_HI(err_cqe), CQE_WRID_LOW(err_cqe));
-		return;
-	}
-
 	printk(KERN_ERR MOD "AE qpid 0x%x opcode %d status 0x%x "
 	       "type %d wrid.hi 0x%x wrid.lo 0x%x\n",
 	       CQE_QPID(err_cqe), CQE_OPCODE(err_cqe),
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 3/8] RDMA/cxgb4: Fix QP flush logic

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

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 appropraite 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-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
Signed-off-by: Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/cq.c       | 326 +++++++++++++++++++++------------
 drivers/infiniband/hw/cxgb4/iw_cxgb4.h |   5 +-
 drivers/infiniband/hw/cxgb4/qp.c       |  34 ++--
 drivers/infiniband/hw/cxgb4/t4.h       |  25 ++-
 4 files changed, 254 insertions(+), 136 deletions(-)

diff --git a/drivers/infiniband/hw/cxgb4/cq.c b/drivers/infiniband/hw/cxgb4/cq.c
index 0f1607c..d84b2c0 100644
--- a/drivers/infiniband/hw/cxgb4/cq.c
+++ b/drivers/infiniband/hw/cxgb4/cq.c
@@ -225,43 +225,186 @@ static void insert_sq_cqe(struct t4_wq *wq, struct t4_cq *cq,
 	t4_swcq_produce(cq);
 }
 
-int c4iw_flush_sq(struct t4_wq *wq, struct t4_cq *cq, int count)
+static void advance_oldest_read(struct t4_wq *wq);
+
+int c4iw_flush_sq(struct c4iw_qp *qhp)
 {
 	int flushed = 0;
-	struct t4_swsqe *swsqe = &wq->sq.sw_sq[wq->sq.cidx + count];
-	int in_use = wq->sq.in_use - count;
-
-	BUG_ON(in_use < 0);
-	while (in_use--) {
-		swsqe->signaled = 0;
-		insert_sq_cqe(wq, cq, swsqe);
-		swsqe++;
-		if (swsqe == (wq->sq.sw_sq + wq->sq.size))
-			swsqe = wq->sq.sw_sq;
-		flushed++;
+	struct t4_wq *wq = &qhp->wq;
+	struct c4iw_cq *chp = to_c4iw_cq(qhp->ibqp.send_cq);
+	struct t4_cq *cq = &chp->cq;
+	int idx;
+	struct t4_swsqe *swsqe;
+	int error = (qhp->attr.state != C4IW_QP_STATE_CLOSING &&
+			qhp->attr.state != C4IW_QP_STATE_IDLE);
+
+	if (wq->sq.flush_cidx == -1)
+		wq->sq.flush_cidx = wq->sq.cidx;
+	idx = wq->sq.flush_cidx;
+	BUG_ON(idx >= wq->sq.size);
+	while (idx != wq->sq.pidx) {
+		if (error) {
+			swsqe = &wq->sq.sw_sq[idx];
+			BUG_ON(swsqe->flushed);
+			swsqe->flushed = 1;
+			insert_sq_cqe(wq, cq, swsqe);
+			if (wq->sq.oldest_read == swsqe) {
+				BUG_ON(swsqe->opcode != FW_RI_READ_REQ);
+				advance_oldest_read(wq);
+			}
+			flushed++;
+		} else {
+			t4_sq_consume(wq);
+		}
+		if (++idx == wq->sq.size)
+			idx = 0;
 	}
+	wq->sq.flush_cidx += flushed;
+	if (wq->sq.flush_cidx >= wq->sq.size)
+		wq->sq.flush_cidx -= wq->sq.size;
 	return flushed;
 }
 
+static void flush_completed_wrs(struct t4_wq *wq, struct t4_cq *cq)
+{
+	struct t4_swsqe *swsqe;
+	int cidx;
+
+	if (wq->sq.flush_cidx == -1)
+		wq->sq.flush_cidx = wq->sq.cidx;
+	cidx = wq->sq.flush_cidx;
+	BUG_ON(cidx > wq->sq.size);
+
+	while (cidx != wq->sq.pidx) {
+		swsqe = &wq->sq.sw_sq[cidx];
+		if (!swsqe->signaled) {
+			if (++cidx == wq->sq.size)
+				cidx = 0;
+		} else if (swsqe->complete) {
+
+			BUG_ON(swsqe->flushed);
+
+			/*
+			 * Insert this completed cqe into the swcq.
+			 */
+			PDBG("%s moving cqe into swcq sq idx %u cq idx %u\n",
+					__func__, cidx, cq->sw_pidx);
+			swsqe->cqe.header |= htonl(V_CQE_SWCQE(1));
+			cq->sw_queue[cq->sw_pidx] = swsqe->cqe;
+			t4_swcq_produce(cq);
+			swsqe->flushed = 1;
+			if (++cidx == wq->sq.size)
+				cidx = 0;
+			wq->sq.flush_cidx = cidx;
+		} else
+			break;
+	}
+}
+
+static void create_read_req_cqe(struct t4_wq *wq, struct t4_cqe *hw_cqe,
+		struct t4_cqe *read_cqe)
+{
+	read_cqe->u.scqe.cidx = wq->sq.oldest_read->idx;
+	read_cqe->len = (__force __be32) ntohl(wq->sq.oldest_read->read_len);
+	read_cqe->header = htonl(V_CQE_QPID(CQE_QPID(hw_cqe)) |
+			V_CQE_SWCQE(SW_CQE(hw_cqe)) |
+			V_CQE_OPCODE(FW_RI_READ_REQ) |
+			V_CQE_TYPE(1));
+	read_cqe->bits_type_ts = hw_cqe->bits_type_ts;
+}
+
+static void advance_oldest_read(struct t4_wq *wq)
+{
+
+	u32 rptr = wq->sq.oldest_read - wq->sq.sw_sq + 1;
+
+	if (rptr == wq->sq.size)
+		rptr = 0;
+	while (rptr != wq->sq.pidx) {
+		wq->sq.oldest_read = &wq->sq.sw_sq[rptr];
+
+		if (wq->sq.oldest_read->opcode == FW_RI_READ_REQ)
+			return;
+		if (++rptr == wq->sq.size)
+			rptr = 0;
+	}
+	wq->sq.oldest_read = NULL;
+}
+
 /*
  * Move all CQEs from the HWCQ into the SWCQ.
+ * Deal with out-of-order and/or completions that complete
+ * prior unsignalled WRs.
  */
-void c4iw_flush_hw_cq(struct t4_cq *cq)
+void c4iw_flush_hw_cq(struct c4iw_cq *chp)
 {
-	struct t4_cqe *cqe = NULL, *swcqe;
+	struct t4_cqe *hw_cqe, *swcqe, read_cqe;
+	struct c4iw_qp *qhp;
+	struct t4_swsqe *swsqe;
 	int ret;
 
-	PDBG("%s cq %p cqid 0x%x\n", __func__, cq, cq->cqid);
-	ret = t4_next_hw_cqe(cq, &cqe);
+	PDBG("%s  cqid 0x%x\n", __func__, chp->cq.cqid);
+	ret = t4_next_hw_cqe(&chp->cq, &hw_cqe);
+
+	/*
+	 * This logic is similar to poll_cq(), but not quite the same
+	 * unfortunately.  Need to move pertinent HW CQEs to the SW CQ but
+	 * also do any translation magic that poll_cq() normally does.
+	 */
 	while (!ret) {
-		PDBG("%s flushing hwcq cidx 0x%x swcq pidx 0x%x\n",
-		     __func__, cq->cidx, cq->sw_pidx);
-		swcqe = &cq->sw_queue[cq->sw_pidx];
-		*swcqe = *cqe;
-		swcqe->header |= cpu_to_be32(V_CQE_SWCQE(1));
-		t4_swcq_produce(cq);
-		t4_hwcq_consume(cq);
-		ret = t4_next_hw_cqe(cq, &cqe);
+		qhp = get_qhp(chp->rhp, CQE_QPID(hw_cqe));
+
+		/*
+		 * drop CQEs with no associated QP
+		 */
+		if (qhp == NULL)
+			goto next_cqe;
+
+		if (CQE_OPCODE(hw_cqe) == FW_RI_TERMINATE)
+			goto next_cqe;
+
+		if (CQE_OPCODE(hw_cqe) == FW_RI_READ_RESP) {
+
+			/*
+			 * drop peer2peer RTR reads.
+			 */
+			if (CQE_WRID_STAG(hw_cqe) == 1)
+				goto next_cqe;
+
+			/*
+			 * Eat completions for unsignaled read WRs.
+			 */
+			if (!qhp->wq.sq.oldest_read->signaled) {
+				advance_oldest_read(&qhp->wq);
+				goto next_cqe;
+			}
+
+			/*
+			 * Don't write to the HWCQ, create a new read req CQE
+			 * in local memory and move it into the swcq.
+			 */
+			create_read_req_cqe(&qhp->wq, hw_cqe, &read_cqe);
+			hw_cqe = &read_cqe;
+			advance_oldest_read(&qhp->wq);
+		}
+
+		/* if its a SQ completion, then do the magic to move all the
+		 * unsignaled and now in-order completions into the swcq.
+		 */
+		if (SQ_TYPE(hw_cqe)) {
+			swsqe = &qhp->wq.sq.sw_sq[CQE_WRID_SQ_IDX(hw_cqe)];
+			swsqe->cqe = *hw_cqe;
+			swsqe->complete = 1;
+			flush_completed_wrs(&qhp->wq, &chp->cq);
+		} else {
+			swcqe = &chp->cq.sw_queue[chp->cq.sw_pidx];
+			*swcqe = *hw_cqe;
+			swcqe->header |= cpu_to_be32(V_CQE_SWCQE(1));
+			t4_swcq_produce(&chp->cq);
+		}
+next_cqe:
+		t4_hwcq_consume(&chp->cq);
+		ret = t4_next_hw_cqe(&chp->cq, &hw_cqe);
 	}
 }
 
@@ -281,25 +424,6 @@ static int cqe_completes_wr(struct t4_cqe *cqe, struct t4_wq *wq)
 	return 1;
 }
 
-void c4iw_count_scqes(struct t4_cq *cq, struct t4_wq *wq, int *count)
-{
-	struct t4_cqe *cqe;
-	u32 ptr;
-
-	*count = 0;
-	ptr = cq->sw_cidx;
-	while (ptr != cq->sw_pidx) {
-		cqe = &cq->sw_queue[ptr];
-		if ((SQ_TYPE(cqe) || ((CQE_OPCODE(cqe) == FW_RI_READ_RESP) &&
-				      wq->sq.oldest_read)) &&
-		    (CQE_QPID(cqe) == wq->sq.qid))
-			(*count)++;
-		if (++ptr == cq->size)
-			ptr = 0;
-	}
-	PDBG("%s cq %p count %d\n", __func__, cq, *count);
-}
-
 void c4iw_count_rcqes(struct t4_cq *cq, struct t4_wq *wq, int *count)
 {
 	struct t4_cqe *cqe;
@@ -319,70 +443,6 @@ void c4iw_count_rcqes(struct t4_cq *cq, struct t4_wq *wq, int *count)
 	PDBG("%s cq %p count %d\n", __func__, cq, *count);
 }
 
-static void flush_completed_wrs(struct t4_wq *wq, struct t4_cq *cq)
-{
-	struct t4_swsqe *swsqe;
-	u16 ptr = wq->sq.cidx;
-	int count = wq->sq.in_use;
-	int unsignaled = 0;
-
-	swsqe = &wq->sq.sw_sq[ptr];
-	while (count--)
-		if (!swsqe->signaled) {
-			if (++ptr == wq->sq.size)
-				ptr = 0;
-			swsqe = &wq->sq.sw_sq[ptr];
-			unsignaled++;
-		} else if (swsqe->complete) {
-
-			/*
-			 * Insert this completed cqe into the swcq.
-			 */
-			PDBG("%s moving cqe into swcq sq idx %u cq idx %u\n",
-			     __func__, ptr, cq->sw_pidx);
-			swsqe->cqe.header |= htonl(V_CQE_SWCQE(1));
-			cq->sw_queue[cq->sw_pidx] = swsqe->cqe;
-			t4_swcq_produce(cq);
-			swsqe->signaled = 0;
-			wq->sq.in_use -= unsignaled;
-			break;
-		} else
-			break;
-}
-
-static void create_read_req_cqe(struct t4_wq *wq, struct t4_cqe *hw_cqe,
-				struct t4_cqe *read_cqe)
-{
-	read_cqe->u.scqe.cidx = wq->sq.oldest_read->idx;
-	read_cqe->len = cpu_to_be32(wq->sq.oldest_read->read_len);
-	read_cqe->header = htonl(V_CQE_QPID(CQE_QPID(hw_cqe)) |
-				 V_CQE_SWCQE(SW_CQE(hw_cqe)) |
-				 V_CQE_OPCODE(FW_RI_READ_REQ) |
-				 V_CQE_TYPE(1));
-	read_cqe->bits_type_ts = hw_cqe->bits_type_ts;
-}
-
-/*
- * Return a ptr to the next read wr in the SWSQ or NULL.
- */
-static void advance_oldest_read(struct t4_wq *wq)
-{
-
-	u32 rptr = wq->sq.oldest_read - wq->sq.sw_sq + 1;
-
-	if (rptr == wq->sq.size)
-		rptr = 0;
-	while (rptr != wq->sq.pidx) {
-		wq->sq.oldest_read = &wq->sq.sw_sq[rptr];
-
-		if (wq->sq.oldest_read->opcode == FW_RI_READ_REQ)
-			return;
-		if (++rptr == wq->sq.size)
-			rptr = 0;
-	}
-	wq->sq.oldest_read = NULL;
-}
-
 /*
  * poll_cq
  *
@@ -427,6 +487,22 @@ static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
 	}
 
 	/*
+	* skip hw cqe's if the wq is flushed.
+	*/
+	if (wq->flushed && !SW_CQE(hw_cqe)) {
+		ret = -EAGAIN;
+		goto skip_cqe;
+	}
+
+	/*
+	 * skip TERMINATE cqes...
+	 */
+	if (CQE_OPCODE(hw_cqe) == FW_RI_TERMINATE) {
+		ret = -EAGAIN;
+		goto skip_cqe;
+	}
+
+	/*
 	 * Gotta tweak READ completions:
 	 *	1) the cqe doesn't contain the sq_wptr from the wr.
 	 *	2) opcode not reflected from the wr.
@@ -440,7 +516,7 @@ static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
 		 * was generated by the kernel driver as part of peer-2-peer
 		 * connection setup.  So ignore the completion.
 		 */
-		if (!wq->sq.oldest_read) {
+		if (CQE_WRID_STAG(hw_cqe) == 1) {
 			if (CQE_STATUS(hw_cqe))
 				t4_set_wq_in_error(wq);
 			ret = -EAGAIN;
@@ -448,6 +524,15 @@ static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
 		}
 
 		/*
+		 * Eat completions for unsignaled read WRs.
+		 */
+		if (!wq->sq.oldest_read->signaled) {
+			advance_oldest_read(wq);
+			ret = -EAGAIN;
+			goto skip_cqe;
+		}
+
+		/*
 		 * Don't write to the HWCQ, so create a new read req CQE
 		 * in local memory.
 		 */
@@ -457,14 +542,8 @@ static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
 	}
 
 	if (CQE_STATUS(hw_cqe) || t4_wq_in_error(wq)) {
-		*cqe_flushed = t4_wq_in_error(wq);
+		*cqe_flushed = (CQE_STATUS(hw_cqe) == T4_ERR_SWFLUSH);
 		t4_set_wq_in_error(wq);
-		goto proc_cqe;
-	}
-
-	if (CQE_OPCODE(hw_cqe) == FW_RI_TERMINATE) {
-		ret = -EAGAIN;
-		goto skip_cqe;
 	}
 
 	/*
@@ -523,7 +602,21 @@ proc_cqe:
 	 * completion.
 	 */
 	if (SQ_TYPE(hw_cqe)) {
-		wq->sq.cidx = CQE_WRID_SQ_IDX(hw_cqe);
+		int idx = CQE_WRID_SQ_IDX(hw_cqe);
+		BUG_ON(idx > wq->sq.size);
+
+		/*
+		* Account for any unsignaled completions completed by
+		* this signaled completion.  In this case, cidx points
+		* to the first unsignaled one, and idx points to the
+		* signaled one.  So adjust in_use based on this delta.
+		* if this is not completing any unsigned wrs, then the
+		* delta will be 0.
+		*/
+		wq->sq.in_use -= idx - wq->sq.cidx;
+		BUG_ON(wq->sq.in_use < 0 && wq->sq.in_use < wq->sq.size);
+
+		wq->sq.cidx = (uint16_t)idx;
 		PDBG("%s completing sq idx %u\n", __func__, wq->sq.cidx);
 		*cookie = wq->sq.sw_sq[wq->sq.cidx].wr_id;
 		t4_sq_consume(wq);
@@ -532,6 +625,7 @@ proc_cqe:
 		*cookie = wq->rq.sw_rq[wq->rq.cidx].wr_id;
 		BUG_ON(t4_rq_empty(wq));
 		t4_rq_consume(wq);
+		goto skip_cqe;
 	}
 
 flush_wq:
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
index b3cae9f..23eaeab 100644
--- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
+++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
@@ -917,12 +917,11 @@ void c4iw_pblpool_free(struct c4iw_rdev *rdev, u32 addr, int size);
 u32 c4iw_ocqp_pool_alloc(struct c4iw_rdev *rdev, int size);
 void c4iw_ocqp_pool_free(struct c4iw_rdev *rdev, u32 addr, int size);
 int c4iw_ofld_send(struct c4iw_rdev *rdev, struct sk_buff *skb);
-void c4iw_flush_hw_cq(struct t4_cq *cq);
+void c4iw_flush_hw_cq(struct c4iw_cq *chp);
 void c4iw_count_rcqes(struct t4_cq *cq, struct t4_wq *wq, int *count);
-void c4iw_count_scqes(struct t4_cq *cq, struct t4_wq *wq, int *count);
 int c4iw_ep_disconnect(struct c4iw_ep *ep, int abrupt, gfp_t gfp);
 int c4iw_flush_rq(struct t4_wq *wq, struct t4_cq *cq, int count);
-int c4iw_flush_sq(struct t4_wq *wq, struct t4_cq *cq, int count);
+int c4iw_flush_sq(struct c4iw_qp *qhp);
 int c4iw_ev_handler(struct c4iw_dev *rnicp, u32 qid);
 u16 c4iw_rqes_posted(struct c4iw_qp *qhp);
 int c4iw_post_terminate(struct c4iw_qp *qhp, struct t4_cqe *err_cqe);
diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c
index a4975e1..22d1f01 100644
--- a/drivers/infiniband/hw/cxgb4/qp.c
+++ b/drivers/infiniband/hw/cxgb4/qp.c
@@ -737,6 +737,7 @@ int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		swsqe->idx = qhp->wq.sq.pidx;
 		swsqe->complete = 0;
 		swsqe->signaled = (wr->send_flags & IB_SEND_SIGNALED);
+		swsqe->flushed = 0;
 		swsqe->wr_id = wr->wr_id;
 
 		init_wr_hdr(wqe, qhp->wq.sq.pidx, fw_opcode, fw_flags, len16);
@@ -1006,7 +1007,15 @@ static void __flush_qp(struct c4iw_qp *qhp, struct c4iw_cq *rchp,
 	/* locking hierarchy: cq lock first, then qp lock. */
 	spin_lock_irqsave(&rchp->lock, flag);
 	spin_lock(&qhp->lock);
-	c4iw_flush_hw_cq(&rchp->cq);
+
+	if (qhp->wq.flushed) {
+		spin_unlock(&qhp->lock);
+		spin_unlock_irqrestore(&rchp->lock, flag);
+		return;
+	}
+	qhp->wq.flushed = 1;
+
+	c4iw_flush_hw_cq(rchp);
 	c4iw_count_rcqes(&rchp->cq, &qhp->wq, &count);
 	flushed = c4iw_flush_rq(&qhp->wq, &rchp->cq, count);
 	spin_unlock(&qhp->lock);
@@ -1020,9 +1029,9 @@ static void __flush_qp(struct c4iw_qp *qhp, struct c4iw_cq *rchp,
 	/* locking hierarchy: cq lock first, then qp lock. */
 	spin_lock_irqsave(&schp->lock, flag);
 	spin_lock(&qhp->lock);
-	c4iw_flush_hw_cq(&schp->cq);
-	c4iw_count_scqes(&schp->cq, &qhp->wq, &count);
-	flushed = c4iw_flush_sq(&qhp->wq, &schp->cq, count);
+	if (schp != rchp)
+		c4iw_flush_hw_cq(schp);
+	flushed = c4iw_flush_sq(qhp);
 	spin_unlock(&qhp->lock);
 	spin_unlock_irqrestore(&schp->lock, flag);
 	if (flushed) {
@@ -1037,11 +1046,11 @@ static void flush_qp(struct c4iw_qp *qhp)
 	struct c4iw_cq *rchp, *schp;
 	unsigned long flag;
 
-	rchp = get_chp(qhp->rhp, qhp->attr.rcq);
-	schp = get_chp(qhp->rhp, qhp->attr.scq);
+	rchp = to_c4iw_cq(qhp->ibqp.recv_cq);
+	schp = to_c4iw_cq(qhp->ibqp.send_cq);
 
+	t4_set_wq_in_error(&qhp->wq);
 	if (qhp->ibqp.uobject) {
-		t4_set_wq_in_error(&qhp->wq);
 		t4_set_cq_in_error(&rchp->cq);
 		spin_lock_irqsave(&rchp->comp_handler_lock, flag);
 		(*rchp->ibcq.comp_handler)(&rchp->ibcq, rchp->ibcq.cq_context);
@@ -1330,8 +1339,7 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 				disconnect = 1;
 				c4iw_get_ep(&qhp->ep->com);
 			}
-			if (qhp->ibqp.uobject)
-				t4_set_wq_in_error(&qhp->wq);
+			t4_set_wq_in_error(&qhp->wq);
 			ret = rdma_fini(rhp, qhp, ep);
 			if (ret)
 				goto err;
@@ -1340,8 +1348,7 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 			set_state(qhp, C4IW_QP_STATE_TERMINATE);
 			qhp->attr.layer_etype = attrs->layer_etype;
 			qhp->attr.ecode = attrs->ecode;
-			if (qhp->ibqp.uobject)
-				t4_set_wq_in_error(&qhp->wq);
+			t4_set_wq_in_error(&qhp->wq);
 			ep = qhp->ep;
 			if (!internal)
 				terminate = 1;
@@ -1350,8 +1357,7 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 			break;
 		case C4IW_QP_STATE_ERROR:
 			set_state(qhp, C4IW_QP_STATE_ERROR);
-			if (qhp->ibqp.uobject)
-				t4_set_wq_in_error(&qhp->wq);
+			t4_set_wq_in_error(&qhp->wq);
 			if (!internal) {
 				abort = 1;
 				disconnect = 1;
@@ -1552,12 +1558,12 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 
 	ucontext = pd->uobject ? to_c4iw_ucontext(pd->uobject->context) : NULL;
 
-
 	qhp = kzalloc(sizeof(*qhp), GFP_KERNEL);
 	if (!qhp)
 		return ERR_PTR(-ENOMEM);
 	qhp->wq.sq.size = sqsize;
 	qhp->wq.sq.memsize = (sqsize + 1) * sizeof *qhp->wq.sq.queue;
+	qhp->wq.sq.flush_cidx = -1;
 	qhp->wq.rq.size = rqsize;
 	qhp->wq.rq.memsize = (rqsize + 1) * sizeof *qhp->wq.rq.queue;
 
diff --git a/drivers/infiniband/hw/cxgb4/t4.h b/drivers/infiniband/hw/cxgb4/t4.h
index ebcb03b..3a6a289 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;
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 4/8] RDMA/cxgb4: Account for unsignled SQ WRs that fails to deal with wrap

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

When determining how many WRs are completed with a signaled CQE,
correctly deal with queue wraps.

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
Signed-off-by: Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/cq.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/cxgb4/cq.c b/drivers/infiniband/hw/cxgb4/cq.c
index d84b2c0..4e91168 100644
--- a/drivers/infiniband/hw/cxgb4/cq.c
+++ b/drivers/infiniband/hw/cxgb4/cq.c
@@ -611,9 +611,12 @@ proc_cqe:
 		* to the first unsignaled one, and idx points to the
 		* signaled one.  So adjust in_use based on this delta.
 		* if this is not completing any unsigned wrs, then the
-		* delta will be 0.
+		* delta will be 0. Handle wrapping also!
 		*/
-		wq->sq.in_use -= idx - wq->sq.cidx;
+		if (idx < wq->sq.cidx)
+			wq->sq.in_use -= wq->sq.size + idx - wq->sq.cidx;
+		else
+			wq->sq.in_use -= idx - wq->sq.cidx;
 		BUG_ON(wq->sq.in_use < 0 && wq->sq.in_use < wq->sq.size);
 
 		wq->sq.cidx = (uint16_t)idx;
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 5/8] RDMA/cxgb4: Set arp error handler for PASS_ACCEPT_RPL messages.

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

accept_cr() failed to set the arp error handler on a reused skb.
This results in a kernel crash if the arp does indeed time out.

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
Signed-off-by: Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/cm.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
index 1b0fd72..e95c6ba 100644
--- a/drivers/infiniband/hw/cxgb4/cm.c
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -332,6 +332,7 @@ static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
 	} else {
 		skb = alloc_skb(len, gfp);
 	}
+	t4_set_arp_err_handler(skb, NULL, NULL);
 	return skb;
 }
 
@@ -2018,6 +2019,7 @@ static void accept_cr(struct c4iw_ep *ep, struct sk_buff *skb,
 	rpl->opt0 = cpu_to_be64(opt0);
 	rpl->opt2 = cpu_to_be32(opt2);
 	set_wr_txq(skb, CPL_PRIORITY_SETUP, ep->ctrlq_idx);
+	t4_set_arp_err_handler(skb, NULL, arp_failure_discard);
 	c4iw_l2t_send(&ep->com.dev->rdev, skb, ep->l2t);
 
 	return;
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 6/8] RDMA/cxgb4: Always do GTS write if cidx_inc == CIDXINC_MASK.

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

When polling, we do a GTS update if the accumulated cidx_inc == the CQ
depth / 16.  However, if the CQ is large enough, Cq depth / 16 exceeds the size
of the field in the GTS word. So we also need to update if cidx_inc hits
CIDXINC_MASK to avoid overflowing the field.

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
Signed-off-by: Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/t4.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/cxgb4/t4.h b/drivers/infiniband/hw/cxgb4/t4.h
index 3a6a289..1272dfe 100644
--- a/drivers/infiniband/hw/cxgb4/t4.h
+++ b/drivers/infiniband/hw/cxgb4/t4.h
@@ -531,7 +531,7 @@ static inline void t4_swcq_consume(struct t4_cq *cq)
 static inline void t4_hwcq_consume(struct t4_cq *cq)
 {
 	cq->bits_type_ts = cq->queue[cq->cidx].bits_type_ts;
-	if (++cq->cidx_inc == (cq->size >> 4)) {
+	if (++cq->cidx_inc == (cq->size >> 4) || cq->cidx_inc == CIDXINC_MASK) {
 		u32 val;
 
 		val = SEINTARM(0) | CIDXINC(cq->cidx_inc) | TIMERREG(7) |
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 7/8] RDMA/cxgb4: Advertise ~0ULL as max MR size.

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

Lustre uses a advertised max MR size of ~0ULL to indicate it should use
a dma_mr. Hence advertise max MR size as ~0ULL.

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
Signed-off-by: Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/t4.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/cxgb4/t4.h b/drivers/infiniband/hw/cxgb4/t4.h
index 1272dfe..e73ace7 100644
--- a/drivers/infiniband/hw/cxgb4/t4.h
+++ b/drivers/infiniband/hw/cxgb4/t4.h
@@ -47,7 +47,7 @@
 #define T4_MAX_QP_DEPTH (T4_MAX_RQ_SIZE - 1)
 #define T4_MAX_CQ_DEPTH (T4_MAX_IQ_SIZE - 1)
 #define T4_MAX_NUM_STAG (1<<15)
-#define T4_MAX_MR_SIZE (~0ULL - 1)
+#define T4_MAX_MR_SIZE (~0ULL)
 #define T4_PAGESIZE_MASK 0xffff000  /* 4KB-128MB */
 #define T4_STAG_UNSET 0xffffffff
 #define T4_FW_MAJ 0
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 8/8] RDMA/cxgb4: Issue RI.FINI before closing when entering TERM

[Vipul Pandya]

From: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>

Signed-off-by: Steve Wise <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>
Signed-off-by: Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
---
 drivers/infiniband/hw/cxgb4/qp.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c
index 22d1f01..5829367 100644
--- a/drivers/infiniband/hw/cxgb4/qp.c
+++ b/drivers/infiniband/hw/cxgb4/qp.c
@@ -1350,9 +1350,14 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 			qhp->attr.ecode = attrs->ecode;
 			t4_set_wq_in_error(&qhp->wq);
 			ep = qhp->ep;
+			disconnect = 1;
 			if (!internal)
 				terminate = 1;
-			disconnect = 1;
+			else {
+				ret = rdma_fini(rhp, qhp, ep);
+				if (ret)
+					goto err;
+			}
 			c4iw_get_ep(&qhp->ep->com);
 			break;
 		case C4IW_QP_STATE_ERROR:
-- 
1.8.0

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 0/8] Critical bug fixes for RDMA/cxgb4

[Roland Dreier]

On Tue, Aug 6, 2013 at 8:34 AM, Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org> wrote:

> This series is on top of previously submitted series "Add IPv6 support for
> iWARP" which can be found at following location:
> http://comments.gmane.org/gmane.linux.network/272124
> We request Roland to merge above series first and then apply this series on top
> of it.

I can do that, and in fact I should be able to start queueing 3.12
material soon.  However, if these are really critical bug fixes, do
you want them to wait for 3.12 as well?

Which are really critical bug fixes?  For example, just from the
subject line, I can't see advertising the wrong MR size as a critical
bug.

 - R.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 0/8] Critical bug fixes for RDMA/cxgb4

[Vipul Pandya]

On 06-08-2013 23:12, Roland Dreier wrote:
> On Tue, Aug 6, 2013 at 8:34 AM, Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org> wrote:
> 
>> This series is on top of previously submitted series "Add IPv6 support for
>> iWARP" which can be found at following location:
>> http://comments.gmane.org/gmane.linux.network/272124
>> We request Roland to merge above series first and then apply this series on top
>> of it.
> 
> I can do that, and in fact I should be able to start queueing 3.12
> material soon.  However, if these are really critical bug fixes, do
> you want them to wait for 3.12 as well?
> 
> Which are really critical bug fixes?  For example, just from the
> subject line, I can't see advertising the wrong MR size as a critical
> bug.
> 

Hi Roland,

We would like to wait for iwarp/ipv6 and this series to get merge in
3.12. 3.11 is at RC4 and we think it won't take long for 3.12-rc.

Thanks,
Vipul

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 0/8] Critical bug fixes for RDMA/cxgb4

[Or Gerlitz]

On Tue, Aug 6, 2013 at 6:34 PM, Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org> wrote:

[...]

> The patch-series is based on Roland's infiniband tree for-next branch.
> This series is on top of previously submitted series "Add IPv6 support for iWARP"
> We request Roland to merge above series first and then apply this series on top of it.

[...]

> This patch series fixes critical bugs for RDMA/cxgb4. It fixes bugs in QP flush
> path, handles newer firmware changes, deals with wrap around issue of the queue,
> sets arp error handler etc.

Hey Vipul, its unclear if this series has fixes to bugs introduced in
the iwarp/ipv6 series, or unrelated fixes which are just rebased on
the other series. But this way or another, in this form it  can't work
for the current cycle, since the iwarp/ipv6 series suits merge window
and can't be accepted just to let few other fixes to get it. What you
can do is wait for the iwarp/ipv6 + this series to get into 3.12-rc
and then backport (adjust) the fixes which are really critical to 3.11
and get them into 3.11.y (and to 3.x.y for x < 11 as well if you want
that) through the -stable process. I think Divy/Steve should be
familiar with all this stuff, isn't that? but anyway, just for the
sake of you and others learning something from this little 101

Or.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 0/8] Critical bug fixes for RDMA/cxgb4

[Vipul Pandya]

On 07-08-2013 01:45, Or Gerlitz wrote:
> On Tue, Aug 6, 2013 at 6:34 PM, Vipul Pandya <vipul-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org> wrote:
> 
> [...]
> 
>> The patch-series is based on Roland's infiniband tree for-next branch.
>> This series is on top of previously submitted series "Add IPv6 support for iWARP"
>> We request Roland to merge above series first and then apply this series on top of it.
> 
> [...]
> 
>> This patch series fixes critical bugs for RDMA/cxgb4. It fixes bugs in QP flush
>> path, handles newer firmware changes, deals with wrap around issue of the queue,
>> sets arp error handler etc.
> 
> Hey Vipul, its unclear if this series has fixes to bugs introduced in
> the iwarp/ipv6 series, or unrelated fixes which are just rebased on
> the other series. But this way or another, in this form it  can't work
> for the current cycle, since the iwarp/ipv6 series suits merge window
> and can't be accepted just to let few other fixes to get it. What you
> can do is wait for the iwarp/ipv6 + this series to get into 3.12-rc
> and then backport (adjust) the fixes which are really critical to 3.11
> and get them into 3.11.y (and to 3.x.y for x < 11 as well if you want
> that) through the -stable process. I think Divy/Steve should be
> familiar with all this stuff, isn't that? but anyway, just for the
> sake of you and others learning something from this little 101
> 
> Or.
> 

Thanks Or Gerlitz for the insights. This series has unrelated fixes
which are just based on IPv6 series. As you have suggested we would like
to wait for iwarp/ipv6 and this series to get merge in 3.12-rc.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html