idpf: split &idpf_queue into 4 strictly-typed queue structures

Currently, sizeof(struct idpf_queue) is 32 Kb.
This is due to the 12-bit hashtable declaration at the end of the queue.
This HT is needed only for Tx queues when the flow scheduling mode is
enabled. But &idpf_queue is unified for all of the queue types,
provoking excessive memory usage.
The unified structure in general makes the code less effective via
suboptimal fields placement. You can't avoid that unless you make unions
each 2 fields. Even then, different field alignment etc., doesn't allow
you to optimize things to the limit.
Split &idpf_queue into 4 structures corresponding to the queue types:
RQ (Rx queue), SQ (Tx queue), FQ (buffer queue), and CQ (completion
queue). Place only needed fields there and shortcuts handy for hotpath.
Allocate the abovementioned hashtable dynamically and only when needed,
keeping &idpf_tx_queue relatively short (192 bytes, same as Rx). This HT
is used only for OOO completions, which aren't really hotpath anyway.
Note that this change must be done atomically, otherwise it's really
easy to get lost and miss something.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

drivers/net/ethernet/intel/idpf/idpf.h

@@ -17,7 +17,6 @@ struct idpf_vport_max_q;
 #include <linux/sctp.h>
 #include <linux/ethtool_netlink.h>
 #include <net/gro.h>
-#include <linux/dim.h>
 
 #include "virtchnl2.h"
 #include "idpf_txrx.h"
@@ -301,7 +300,7 @@ struct idpf_vport {
 	u16 num_txq_grp;
 	struct idpf_txq_group *txq_grps;
 	u32 txq_model;
-	struct idpf_queue **txqs;
+	struct idpf_tx_queue **txqs;
 	bool crc_enable;
 
 	u16 num_rxq;

drivers/net/ethernet/intel/idpf/idpf_lib.c

@@ -1318,14 +1318,14 @@ static void idpf_rx_init_buf_tail(struct idpf_vport *vport)
 
 		if (idpf_is_queue_model_split(vport->rxq_model)) {
 			for (j = 0; j < vport->num_bufqs_per_qgrp; j++) {
-				struct idpf_queue *q =
+				const struct idpf_buf_queue *q =
 					&grp->splitq.bufq_sets[j].bufq;
 
 				writel(q->next_to_alloc, q->tail);
 			}
 		} else {
 			for (j = 0; j < grp->singleq.num_rxq; j++) {
-				struct idpf_queue *q =
+				const struct idpf_rx_queue *q =
 					grp->singleq.rxqs[j];
 
 				writel(q->next_to_alloc, q->tail);
@@ -1855,7 +1855,7 @@ int idpf_initiate_soft_reset(struct idpf_vport *vport,
 	enum idpf_vport_state current_state = np->state;
 	struct idpf_adapter *adapter = vport->adapter;
 	struct idpf_vport *new_vport;
-	int err, i;
+	int err;
 
 	/* If the system is low on memory, we can end up in bad state if we
 	 * free all the memory for queue resources and try to allocate them
@@ -1929,46 +1929,6 @@ int idpf_initiate_soft_reset(struct idpf_vport *vport,
 	 */
 	memcpy(vport, new_vport, offsetof(struct idpf_vport, link_speed_mbps));
 
-	/* Since idpf_vport_queues_alloc was called with new_port, the queue
-	 * back pointers are currently pointing to the local new_vport. Reset
-	 * the backpointers to the original vport here
-	 */
-	for (i = 0; i < vport->num_txq_grp; i++) {
-		struct idpf_txq_group *tx_qgrp = &vport->txq_grps[i];
-		int j;
-
-		tx_qgrp->vport = vport;
-		for (j = 0; j < tx_qgrp->num_txq; j++)
-			tx_qgrp->txqs[j]->vport = vport;
-
-		if (idpf_is_queue_model_split(vport->txq_model))
-			tx_qgrp->complq->vport = vport;
-	}
-
-	for (i = 0; i < vport->num_rxq_grp; i++) {
-		struct idpf_rxq_group *rx_qgrp = &vport->rxq_grps[i];
-		struct idpf_queue *q;
-		u16 num_rxq;
-		int j;
-
-		rx_qgrp->vport = vport;
-		for (j = 0; j < vport->num_bufqs_per_qgrp; j++)
-			rx_qgrp->splitq.bufq_sets[j].bufq.vport = vport;
-
-		if (idpf_is_queue_model_split(vport->rxq_model))
-			num_rxq = rx_qgrp->splitq.num_rxq_sets;
-		else
-			num_rxq = rx_qgrp->singleq.num_rxq;
-
-		for (j = 0; j < num_rxq; j++) {
-			if (idpf_is_queue_model_split(vport->rxq_model))
-				q = &rx_qgrp->splitq.rxq_sets[j]->rxq;
-			else
-				q = rx_qgrp->singleq.rxqs[j];
-			q->vport = vport;
-		}
-	}
-
 	if (reset_cause == IDPF_SR_Q_CHANGE)
 		idpf_vport_alloc_vec_indexes(vport);