[PATCH net-next 09/15] mlxsw: pci: Use only one event queue

[Petr Machata <petrm@nvidia.com>]

From: Amit Cohen <amcohen@nvidia.com>

The device supports two event queues. EQ0 is used for command interface
completion events. EQ1 is used for completion events of RDQ or SDQ.

Currently, for each EQE (event queue element), we check the queue number
and handle accordingly. More than that, for each interrupt we schedule
tasklets for both EQs. This is really ineffective, especially because of
the fact that EQ0 is used only as part of driver init/fini, when EMADs are
not available. There is no point to schedule the tasklet for it and check
each EQE.

A previous patch changed the code to poll command interface for each use of
it. It means that now there is no real reason to use EQ0, as we poll the
command interface.

Initialize only one event queue and use it as EQ1 (this is determined by
queue number). Then, for each interrupt we can schedule the tasklet only
for one queue and we do not have to check the queue number. This
simplifies the code and should improve performance. Note that polling
command interface is ok as we use it only as part of driver init/fini.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlxsw/pci.c    | 54 ++++++--------------
 drivers/net/ethernet/mellanox/mlxsw/pci_hw.h |  2 +-
 2 files changed, 16 insertions(+), 40 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci.c b/drivers/net/ethernet/mellanox/mlxsw/pci.c
index cb960917e462..3460a4ef7d9a 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/pci.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/pci.c
@@ -223,10 +223,10 @@ static struct mlxsw_pci_queue *mlxsw_pci_cq_get(struct mlxsw_pci *mlxsw_pci,
 	return __mlxsw_pci_queue_get(mlxsw_pci, MLXSW_PCI_QUEUE_TYPE_CQ, q_num);
 }
 
-static struct mlxsw_pci_queue *mlxsw_pci_eq_get(struct mlxsw_pci *mlxsw_pci,
-						u8 q_num)
+static struct mlxsw_pci_queue *mlxsw_pci_eq_get(struct mlxsw_pci *mlxsw_pci)
 {
-	return __mlxsw_pci_queue_get(mlxsw_pci, MLXSW_PCI_QUEUE_TYPE_EQ, q_num);
+	/* There is only one EQ at index 0. */
+	return __mlxsw_pci_queue_get(mlxsw_pci, MLXSW_PCI_QUEUE_TYPE_EQ, 0);
 }
 
 static void __mlxsw_pci_queue_doorbell_set(struct mlxsw_pci *mlxsw_pci,
@@ -754,16 +754,6 @@ static u8 mlxsw_pci_cq_elem_size(const struct mlxsw_pci_queue *q)
 					       MLXSW_PCI_CQE01_SIZE;
 }
 
-static void mlxsw_pci_eq_cmd_event(struct mlxsw_pci *mlxsw_pci, char *eqe)
-{
-	mlxsw_pci->cmd.comp.status = mlxsw_pci_eqe_cmd_status_get(eqe);
-	mlxsw_pci->cmd.comp.out_param =
-		((u64) mlxsw_pci_eqe_cmd_out_param_h_get(eqe)) << 32 |
-		mlxsw_pci_eqe_cmd_out_param_l_get(eqe);
-	mlxsw_pci->cmd.wait_done = true;
-	wake_up(&mlxsw_pci->cmd.wait);
-}
-
 static char *mlxsw_pci_eq_sw_eqe_get(struct mlxsw_pci_queue *q)
 {
 	struct mlxsw_pci_queue_elem_info *elem_info;
@@ -786,7 +776,6 @@ static void mlxsw_pci_eq_tasklet(struct tasklet_struct *t)
 	struct mlxsw_pci_queue *q = from_tasklet(q, t, tasklet);
 	struct mlxsw_pci *mlxsw_pci = q->pci;
 	int credits = q->count >> 1;
-	bool cq_handle = false;
 	u8 cqn, cq_count;
 	int items = 0;
 	char *eqe;
@@ -794,23 +783,9 @@ static void mlxsw_pci_eq_tasklet(struct tasklet_struct *t)
 	memset(&active_cqns, 0, sizeof(active_cqns));
 
 	while ((eqe = mlxsw_pci_eq_sw_eqe_get(q))) {
+		cqn = mlxsw_pci_eqe_cqn_get(eqe);
+		set_bit(cqn, active_cqns);
 
-		/* Command interface completion events are always received on
-		 * queue MLXSW_PCI_EQ_ASYNC_NUM (EQ0) and completion events
-		 * are mapped to queue MLXSW_PCI_EQ_COMP_NUM (EQ1).
-		 */
-		switch (q->num) {
-		case MLXSW_PCI_EQ_ASYNC_NUM:
-			mlxsw_pci_eq_cmd_event(mlxsw_pci, eqe);
-			break;
-		case MLXSW_PCI_EQ_COMP_NUM:
-			cqn = mlxsw_pci_eqe_cqn_get(eqe);
-			set_bit(cqn, active_cqns);
-			cq_handle = true;
-			break;
-		default:
-			WARN_ON_ONCE(1);
-		}
 		if (++items == credits)
 			break;
 	}
@@ -821,9 +796,6 @@ static void mlxsw_pci_eq_tasklet(struct tasklet_struct *t)
 	mlxsw_pci_queue_doorbell_consumer_ring(mlxsw_pci, q);
 	mlxsw_pci_queue_doorbell_arm_consumer_ring(mlxsw_pci, q);
 
-	if (!cq_handle)
-		return;
-
 	cq_count = mlxsw_pci_cq_count(mlxsw_pci);
 	for_each_set_bit(cqn, active_cqns, cq_count) {
 		q = mlxsw_pci_cq_get(mlxsw_pci, cqn);
@@ -837,6 +809,13 @@ static int mlxsw_pci_eq_init(struct mlxsw_pci *mlxsw_pci, char *mbox,
 	int i;
 	int err;
 
+	/* We expect to initialize only one EQ, which gets num=0 as it is
+	 * located at index zero. We use the EQ as EQ1, so set the number for
+	 * future use.
+	 */
+	WARN_ON_ONCE(q->num);
+	q->num = MLXSW_PCI_EQ_COMP_NUM;
+
 	q->consumer_counter = 0;
 
 	for (i = 0; i < q->count; i++) {
@@ -1077,7 +1056,7 @@ static int mlxsw_pci_aqs_init(struct mlxsw_pci *mlxsw_pci, char *mbox)
 	mlxsw_pci->num_sdq_cqs = num_sdqs;
 
 	err = mlxsw_pci_queue_group_init(mlxsw_pci, mbox, &mlxsw_pci_eq_ops,
-					 num_eqs);
+					 MLXSW_PCI_EQS_COUNT);
 	if (err) {
 		dev_err(&pdev->dev, "Failed to initialize event queues\n");
 		return err;
@@ -1414,12 +1393,9 @@ static irqreturn_t mlxsw_pci_eq_irq_handler(int irq, void *dev_id)
 {
 	struct mlxsw_pci *mlxsw_pci = dev_id;
 	struct mlxsw_pci_queue *q;
-	int i;
 
-	for (i = 0; i < MLXSW_PCI_EQS_MAX; i++) {
-		q = mlxsw_pci_eq_get(mlxsw_pci, i);
-		mlxsw_pci_queue_tasklet_schedule(q);
-	}
+	q = mlxsw_pci_eq_get(mlxsw_pci);
+	mlxsw_pci_queue_tasklet_schedule(q);
 	return IRQ_HANDLED;
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h b/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h
index 32a4f436d24d..6bed495dcf0f 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h
@@ -43,7 +43,7 @@
 
 #define MLXSW_PCI_CQS_MAX	96
 #define MLXSW_PCI_EQS_MAX	2
-#define MLXSW_PCI_EQ_ASYNC_NUM	0
+#define MLXSW_PCI_EQS_COUNT	1
 #define MLXSW_PCI_EQ_COMP_NUM	1
 
 #define MLXSW_PCI_SDQS_MIN	2 /* EMAD and control traffic */
-- 
2.43.0