diff mbox series

[for-next,v2,10/18] RDMA/rxe: Handle qp error in rxe_resp.c

Message ID 20221021200118.2163-11-rpearsonhpe@gmail.com (mailing list archive)
State Superseded
Delegated to: Jason Gunthorpe
Headers show
Series Implement work queues for rdma_rxe | expand

Commit Message

Bob Pearson Oct. 21, 2022, 8:01 p.m. UTC 
Split rxe_drain_req_pkts() into two subroutines which perform
separate tasks. Change qp error and reset states and !qp->valid
in the same way as rxe_comp.c. Flush recv wqes for qp in error.

Signed-off-by: Ian Ziemba <ian.ziemba@hpe.com>
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
---
 drivers/infiniband/sw/rxe/rxe_resp.c | 64 ++++++++++++++++++++--------
 1 file changed, 47 insertions(+), 17 deletions(-)

Comments

kernel test robot Oct. 21, 2022, 9:22 p.m. UTC | #1 
Hi Bob,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on 9abf2313adc1ca1b6180c508c25f22f9395cc780]

url:    https://github.com/intel-lab-lkp/linux/commits/Bob-Pearson/Implement-work-queues-for-rdma_rxe/20221022-040425
base:   9abf2313adc1ca1b6180c508c25f22f9395cc780
patch link:    https://lore.kernel.org/r/20221021200118.2163-11-rpearsonhpe%40gmail.com
patch subject: [PATCH for-next v2 10/18] RDMA/rxe: Handle qp error in rxe_resp.c
config: parisc-allyesconfig
compiler: hppa-linux-gcc (GCC) 12.1.0
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # https://github.com/intel-lab-lkp/linux/commit/89164294e45c7434c93b957bc31354f8057ed6b5
        git remote add linux-review https://github.com/intel-lab-lkp/linux
        git fetch --no-tags linux-review Bob-Pearson/Implement-work-queues-for-rdma_rxe/20221022-040425
        git checkout 89164294e45c7434c93b957bc31354f8057ed6b5
        # save the config file
        mkdir build_dir && cp config build_dir/.config
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-12.1.0 make.cross W=1 O=build_dir ARCH=parisc SHELL=/bin/bash drivers/infiniband/sw/rxe/

If you fix the issue, kindly add following tag where applicable
| Reported-by: kernel test robot <lkp@intel.com>

All warnings (new ones prefixed by >>):

>> drivers/infiniband/sw/rxe/rxe_resp.c:1253:5: warning: no previous prototype for 'complete_flush' [-Wmissing-prototypes]
    1253 | int complete_flush(struct rxe_qp *qp, struct rxe_recv_wqe *wqe)
         |     ^~~~~~~~~~~~~~


vim +/complete_flush +1253 drivers/infiniband/sw/rxe/rxe_resp.c

  1252	
> 1253	int complete_flush(struct rxe_qp *qp, struct rxe_recv_wqe *wqe)
  1254	{
  1255		struct rxe_cqe cqe;
  1256		struct ib_wc *wc = &cqe.ibwc;
  1257		struct ib_uverbs_wc *uwc = &cqe.uibwc;
  1258	
  1259		memset(&cqe, 0, sizeof(cqe));
  1260	
  1261		if (qp->rcq->is_user) {
  1262			uwc->status		= IB_WC_WR_FLUSH_ERR;
  1263			uwc->qp_num		= qp->ibqp.qp_num;
  1264			uwc->wr_id		= wqe->wr_id;
  1265		} else {
  1266			wc->status		= IB_WC_WR_FLUSH_ERR;
  1267			wc->qp			= &qp->ibqp;
  1268			wc->wr_id		= wqe->wr_id;
  1269		}
  1270	
  1271		if (rxe_cq_post(qp->rcq, &cqe, 0))
  1272			return -ENOMEM;
  1273	
  1274		return 0;
  1275	}
  1276
diff mbox series

Patch

diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index dd11dea70bbf..0bcdd1154641 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -1025,7 +1025,6 @@  static enum resp_states do_complete(struct rxe_qp *qp,
 		return RESPST_CLEANUP;
 }
 
-
 static int send_common_ack(struct rxe_qp *qp, u8 syndrome, u32 psn,
 				  int opcode, const char *msg)
 {
@@ -1240,22 +1239,56 @@  static enum resp_states do_class_d1e_error(struct rxe_qp *qp)
 	}
 }
 
-static void rxe_drain_req_pkts(struct rxe_qp *qp, bool notify)
+static void rxe_drain_req_pkts(struct rxe_qp *qp)
 {
 	struct sk_buff *skb;
-	struct rxe_queue *q = qp->rq.queue;
 
 	while ((skb = skb_dequeue(&qp->req_pkts))) {
 		rxe_put(qp);
 		kfree_skb(skb);
 		ib_device_put(qp->ibqp.device);
 	}
+}
+
+int complete_flush(struct rxe_qp *qp, struct rxe_recv_wqe *wqe)
+{
+	struct rxe_cqe cqe;
+	struct ib_wc *wc = &cqe.ibwc;
+	struct ib_uverbs_wc *uwc = &cqe.uibwc;
+
+	memset(&cqe, 0, sizeof(cqe));
 
-	if (notify)
-		return;
+	if (qp->rcq->is_user) {
+		uwc->status		= IB_WC_WR_FLUSH_ERR;
+		uwc->qp_num		= qp->ibqp.qp_num;
+		uwc->wr_id		= wqe->wr_id;
+	} else {
+		wc->status		= IB_WC_WR_FLUSH_ERR;
+		wc->qp			= &qp->ibqp;
+		wc->wr_id		= wqe->wr_id;
+	}
 
-	while (!qp->srq && q && queue_head(q, q->type))
+	if (rxe_cq_post(qp->rcq, &cqe, 0))
+		return -ENOMEM;
+
+	return 0;
+}
+
+/* drain the receive queue. Complete each wqe with a flush error
+ * if notify is true or until a cq overflow occurs.
+ */
+static void rxe_drain_recv_queue(struct rxe_qp *qp, bool notify)
+{
+	struct rxe_recv_wqe *wqe;
+	struct rxe_queue *q = qp->rq.queue;
+
+	while ((wqe = queue_head(q, q->type))) {
+		if (notify && complete_flush(qp, wqe))
+			notify = 0;
 		queue_advance_consumer(q, q->type);
+	}
+
+	qp->resp.wqe = NULL;
 }
 
 int rxe_responder(void *arg)
@@ -1264,6 +1297,7 @@  int rxe_responder(void *arg)
 	struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
 	enum resp_states state;
 	struct rxe_pkt_info *pkt = NULL;
+	bool notify;
 	int ret;
 
 	if (!rxe_get(qp))
@@ -1271,20 +1305,16 @@  int rxe_responder(void *arg)
 
 	qp->resp.aeth_syndrome = AETH_ACK_UNLIMITED;
 
-	if (!qp->valid)
-		goto exit;
-
-	switch (qp->resp.state) {
-	case QP_STATE_RESET:
-		rxe_drain_req_pkts(qp, false);
-		qp->resp.wqe = NULL;
+	if (!qp->valid || qp->resp.state == QP_STATE_ERROR ||
+	    qp->resp.state == QP_STATE_RESET) {
+		notify = qp->valid && (qp->resp.state == QP_STATE_ERROR);
+		rxe_drain_req_pkts(qp);
+		rxe_drain_recv_queue(qp, notify);
 		goto exit;
-
-	default:
-		state = RESPST_GET_REQ;
-		break;
 	}
 
+	state = RESPST_GET_REQ;
+
 	while (1) {
 		pr_debug("qp#%d state = %s\n", qp_num(qp),
 			 resp_state_name[state]);