[v6,12/13] regex/mlx5: implement dequeue function

Message ID 1595226378-81144-13-git-send-email-orika@mellanox.com (mailing list archive)
State Accepted, archived
Delegated to: Thomas Monjalon
Headers
Series add Mellanox RegEx PMD |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK

Commit Message

Ori Kam July 20, 2020, 6:26 a.m. UTC
From: Yuval Avnery <yuvalav@mellanox.com>

Implement dequeue function for the regex API.

Signed-off-by: Yuval Avnery <yuvalav@mellanox.com>
Acked-by: Ori Kam <orika@mellanox.com>

---
 drivers/common/mlx5/mlx5_prm.h           |  10 +++
 drivers/regex/mlx5/mlx5_regex.c          |   1 +
 drivers/regex/mlx5/mlx5_regex.h          |   4 ++
 drivers/regex/mlx5/mlx5_regex_control.c  |   1 +
 drivers/regex/mlx5/mlx5_regex_fastpath.c | 108 ++++++++++++++++++++++++++++++-
 5 files changed, 123 insertions(+), 1 deletion(-)
  

Patch

diff --git a/drivers/common/mlx5/mlx5_prm.h b/drivers/common/mlx5/mlx5_prm.h
index 6a3f22e..8fab962 100644
--- a/drivers/common/mlx5/mlx5_prm.h
+++ b/drivers/common/mlx5/mlx5_prm.h
@@ -441,6 +441,12 @@  struct mlx5_ifc_regexp_metadata_bits {
 	uint8_t reserved[0x80];
 };
 
+struct mlx5_ifc_regexp_match_tuple_bits {
+	uint8_t length[0x10];
+	uint8_t start_ptr[0x10];
+	uint8_t rule_id[0x20];
+};
+
 /* Adding direct verbs to data-path. */
 
 /* CQ sequence number mask. */
@@ -637,6 +643,10 @@  struct mlx5_modification_cmd {
 				  __mlx5_16_bit_off(typ, fld))); \
 	} while (0)
 
+#define MLX5_GET_VOLATILE(typ, p, fld) \
+	((rte_be_to_cpu_32(*((volatile __be32 *)(p) +\
+	__mlx5_dw_off(typ, fld))) >> __mlx5_dw_bit_off(typ, fld)) & \
+	__mlx5_mask(typ, fld))
 #define MLX5_GET(typ, p, fld) \
 	((rte_be_to_cpu_32(*((__be32 *)(p) +\
 	__mlx5_dw_off(typ, fld))) >> __mlx5_dw_bit_off(typ, fld)) & \
diff --git a/drivers/regex/mlx5/mlx5_regex.c b/drivers/regex/mlx5/mlx5_regex.c
index 607ca4d..fd457bf 100644
--- a/drivers/regex/mlx5/mlx5_regex.c
+++ b/drivers/regex/mlx5/mlx5_regex.c
@@ -168,6 +168,7 @@ 
 	}
 	priv->regexdev->dev_ops = &mlx5_regexdev_ops;
 	priv->regexdev->enqueue = mlx5_regexdev_enqueue;
+	priv->regexdev->dequeue = mlx5_regexdev_dequeue;
 	priv->regexdev->device = (struct rte_device *)pci_dev;
 	priv->regexdev->data->dev_private = priv;
 	priv->regexdev->state = RTE_REGEXDEV_READY;
diff --git a/drivers/regex/mlx5/mlx5_regex.h b/drivers/regex/mlx5/mlx5_regex.h
index fb81cb8..217a1a4 100644
--- a/drivers/regex/mlx5/mlx5_regex.h
+++ b/drivers/regex/mlx5/mlx5_regex.h
@@ -34,6 +34,7 @@  struct mlx5_regex_cq {
 	uint32_t dbr_umem; /* Door bell record umem id. */
 	volatile struct mlx5_cqe *cqe; /* The CQ ring buffer. */
 	struct mlx5dv_devx_umem *cqe_umem; /* CQ buffer umem. */
+	size_t ci;
 	uint32_t *dbr;
 };
 
@@ -105,4 +106,7 @@  int mlx5_regex_qp_setup(struct rte_regexdev *dev, uint16_t qp_ind,
 int mlx5_regexdev_setup_fastpath(struct mlx5_regex_priv *priv, uint32_t qp_id);
 uint16_t mlx5_regexdev_enqueue(struct rte_regexdev *dev, uint16_t qp_id,
 		       struct rte_regex_ops **ops, uint16_t nb_ops);
+uint16_t mlx5_regexdev_dequeue(struct rte_regexdev *dev, uint16_t qp_id,
+		       struct rte_regex_ops **ops, uint16_t nb_ops);
+
 #endif /* MLX5_REGEX_H */
diff --git a/drivers/regex/mlx5/mlx5_regex_control.c b/drivers/regex/mlx5/mlx5_regex_control.c
index f144c69..faafb76 100644
--- a/drivers/regex/mlx5/mlx5_regex_control.c
+++ b/drivers/regex/mlx5/mlx5_regex_control.c
@@ -120,6 +120,7 @@ 
 	cq->cqe_umem = mlx5_glue->devx_umem_reg(priv->ctx, buf,
 						sizeof(struct mlx5_cqe) *
 						cq_size, 7);
+	cq->ci = 0;
 	if (!cq->cqe_umem) {
 		DRV_LOG(ERR, "Can't register cqe mem.");
 		rte_errno  = ENOMEM;
diff --git a/drivers/regex/mlx5/mlx5_regex_fastpath.c b/drivers/regex/mlx5/mlx5_regex_fastpath.c
index ff6386b..fc5a7fc 100644
--- a/drivers/regex/mlx5/mlx5_regex_fastpath.c
+++ b/drivers/regex/mlx5/mlx5_regex_fastpath.c
@@ -35,6 +35,7 @@ 
 #define MLX5_REGEX_WQE_METADATA_OFFSET 16
 #define MLX5_REGEX_WQE_GATHER_OFFSET 32
 #define MLX5_REGEX_WQE_SCATTER_OFFSET 48
+#define MLX5_REGEX_METADATA_OFF 32
 
 #define MLX5_REGEX_WQE_METADATA_OFFSET 16
 #define MLX5_REGEX_WQE_GATHER_OFFSET 32
@@ -159,7 +160,7 @@  struct mlx5_regex_job {
 
 static inline uint32_t
 job_id_get(uint32_t qid, size_t sq_size, size_t index) {
-	return qid*sq_size + index % sq_size;
+	return qid * sq_size + (index & (sq_size - 1));
 }
 
 uint16_t
@@ -192,6 +193,111 @@  struct mlx5_regex_job {
 	return i;
 }
 
+#define MLX5_REGEX_RESP_SZ 8
+
+static inline void
+extract_result(struct rte_regex_ops *op, struct mlx5_regex_job *job)
+{
+	size_t j, offset;
+	op->user_id = job->user_id;
+	op->nb_matches = MLX5_GET_VOLATILE(regexp_metadata, job->metadata +
+					   MLX5_REGEX_METADATA_OFF,
+					   match_count);
+	op->nb_actual_matches = MLX5_GET_VOLATILE(regexp_metadata,
+						  job->metadata +
+						  MLX5_REGEX_METADATA_OFF,
+						  detected_match_count);
+	for (j = 0; j < op->nb_matches; j++) {
+		offset = MLX5_REGEX_RESP_SZ * j;
+		op->matches[j].rule_id =
+			MLX5_GET_VOLATILE(regexp_match_tuple,
+					  (job->output + offset), rule_id);
+		op->matches[j].start_offset =
+			MLX5_GET_VOLATILE(regexp_match_tuple,
+					  (job->output +  offset), start_ptr);
+		op->matches[j].len =
+			MLX5_GET_VOLATILE(regexp_match_tuple,
+					  (job->output +  offset), length);
+	}
+}
+
+static inline volatile struct mlx5_cqe *
+poll_one(struct mlx5_regex_cq *cq)
+{
+	volatile struct mlx5_cqe *cqe;
+	size_t next_cqe_offset;
+
+	next_cqe_offset =  (cq->ci & (cq_size_get(cq) - 1));
+	cqe = (volatile struct mlx5_cqe *)(cq->cqe + next_cqe_offset);
+	rte_cio_wmb();
+
+	int ret = check_cqe(cqe, cq_size_get(cq), cq->ci);
+
+	if (unlikely(ret == MLX5_CQE_STATUS_ERR)) {
+		DRV_LOG(ERR, "Completion with error on qp 0x%x",  0);
+		return NULL;
+	}
+
+	if (unlikely(ret != MLX5_CQE_STATUS_SW_OWN))
+		return NULL;
+
+	return cqe;
+}
+
+
+/**
+ * DPDK callback for dequeue.
+ *
+ * @param dev
+ *   Pointer to the regex dev structure.
+ * @param qp_id
+ *   The queue to enqueue the traffic to.
+ * @param ops
+ *   List of regex ops to dequeue.
+ * @param nb_ops
+ *   Number of ops in ops parameter.
+ *
+ * @return
+ *   Number of packets successfully dequeued (<= pkts_n).
+ */
+uint16_t
+mlx5_regexdev_dequeue(struct rte_regexdev *dev, uint16_t qp_id,
+		      struct rte_regex_ops **ops, uint16_t nb_ops)
+{
+	struct mlx5_regex_priv *priv = dev->data->dev_private;
+	struct mlx5_regex_qp *queue = &priv->qps[qp_id];
+	struct mlx5_regex_cq *cq = &queue->cq;
+	volatile struct mlx5_cqe *cqe;
+	size_t i = 0;
+
+	while ((cqe = poll_one(cq))) {
+		uint16_t wq_counter
+			= (rte_be_to_cpu_16(cqe->wqe_counter) + 1) &
+			  MLX5_REGEX_MAX_WQE_INDEX;
+		size_t sqid = cqe->rsvd3[2];
+		struct mlx5_regex_sq *sq = &queue->sqs[sqid];
+		while (sq->ci != wq_counter) {
+			if (unlikely(i == nb_ops)) {
+				/* Return without updating cq->ci */
+				goto out;
+			}
+			uint32_t job_id = job_id_get(sqid, sq_size_get(sq),
+						     sq->ci);
+			extract_result(ops[i], &queue->jobs[job_id]);
+			sq->ci = (sq->ci + 1) & MLX5_REGEX_MAX_WQE_INDEX;
+			i++;
+		}
+		cq->ci = (cq->ci + 1) & 0xffffff;
+		rte_wmb();
+		cq->dbr[0] = rte_cpu_to_be_32(cq->ci);
+		queue->free_sqs |= (1 << sqid);
+	}
+
+out:
+	queue->ci += i;
+	return i;
+}
+
 static void
 setup_sqs(struct mlx5_regex_qp *queue)
 {