qede: Better utilize the qede_[rt]x_queue
authorMintz, Yuval <Yuval.Mintz@cavium.com>
Tue, 29 Nov 2016 14:47:08 +0000 (16:47 +0200)
committerDavid S. Miller <davem@davemloft.net>
Wed, 30 Nov 2016 19:32:05 +0000 (14:32 -0500)
Improve the cacheline usage of both queues by reordering -
This reduces the cachelines required for egress datapath processing
from 3 to 2 and those required by ingress datapath processing by 2.

It also changes a couple of datapath related functions that currently
require either the fastpath or the qede_dev, changing them to be based
on the tx/rx queue instead.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/qlogic/qede/qede.h
drivers/net/ethernet/qlogic/qede/qede_ethtool.c
drivers/net/ethernet/qlogic/qede/qede_main.c

index c2135765f8ec2e302ff9602b40de0f488d96e1e1..ec372daeba6b81fde31c62dfde69bfa6372cce74 100644 (file)
@@ -243,27 +243,33 @@ struct qede_agg_info {
 };
 
 struct qede_rx_queue {
-       __le16                  *hw_cons_ptr;
-       struct sw_rx_data       *sw_rx_ring;
-       u16                     sw_rx_cons;
-       u16                     sw_rx_prod;
-       struct qed_chain        rx_bd_ring;
-       struct qed_chain        rx_comp_ring;
-       void __iomem            *hw_rxq_prod_addr;
+       __le16 *hw_cons_ptr;
+       void __iomem *hw_rxq_prod_addr;
 
-       /* GRO */
-       struct qede_agg_info    tpa_info[ETH_TPA_MAX_AGGS_NUM];
+       /* Required for the allocation of replacement buffers */
+       struct device *dev;
+
+       u16 sw_rx_cons;
+       u16 sw_rx_prod;
 
-       int                     rx_buf_size;
-       unsigned int            rx_buf_seg_size;
+       u16 num_rx_buffers; /* Slowpath */
+       u8 rxq_id;
 
-       u16                     num_rx_buffers;
-       u16                     rxq_id;
+       u32 rx_buf_size;
+       u32 rx_buf_seg_size;
 
-       u64                     rcv_pkts;
-       u64                     rx_hw_errors;
-       u64                     rx_alloc_errors;
-       u64                     rx_ip_frags;
+       u64 rcv_pkts;
+
+       struct sw_rx_data *sw_rx_ring;
+       struct qed_chain rx_bd_ring;
+       struct qed_chain rx_comp_ring ____cacheline_aligned;
+
+       /* GRO */
+       struct qede_agg_info tpa_info[ETH_TPA_MAX_AGGS_NUM];
+
+       u64 rx_hw_errors;
+       u64 rx_alloc_errors;
+       u64 rx_ip_frags;
 
        void *handle;
 };
@@ -281,22 +287,28 @@ struct sw_tx_bd {
 };
 
 struct qede_tx_queue {
-       int                     index; /* Queue index */
-       __le16                  *hw_cons_ptr;
-       struct sw_tx_bd         *sw_tx_ring;
-       u16                     sw_tx_cons;
-       u16                     sw_tx_prod;
-       struct qed_chain        tx_pbl;
-       void __iomem            *doorbell_addr;
-       union db_prod           tx_db;
-
-       u16                     num_tx_buffers;
-       u64                     xmit_pkts;
-       u64                     stopped_cnt;
-
-       bool                    is_legacy;
-       void *handle;
+       bool is_legacy;
+       u16 sw_tx_cons;
+       u16 sw_tx_prod;
+       u16 num_tx_buffers; /* Slowpath only */
 
+       u64 xmit_pkts;
+       u64 stopped_cnt;
+
+       __le16 *hw_cons_ptr;
+
+       /* Needed for the mapping of packets */
+       struct device *dev;
+
+       void __iomem *doorbell_addr;
+       union db_prod tx_db;
+       int index; /* Slowpath only */
+
+       struct sw_tx_bd *sw_tx_ring;
+       struct qed_chain tx_pbl;
+
+       /* Slowpath; Should be kept in end [unless missing padding] */
+       void *handle;
 };
 
 #define BD_UNMAP_ADDR(bd)              HILO_U64(le32_to_cpu((bd)->addr.hi), \
@@ -363,8 +375,7 @@ void __qede_lock(struct qede_dev *edev);
 void __qede_unlock(struct qede_dev *edev);
 bool qede_has_rx_work(struct qede_rx_queue *rxq);
 int qede_txq_has_work(struct qede_tx_queue *txq);
-void qede_recycle_rx_bd_ring(struct qede_rx_queue *rxq, struct qede_dev *edev,
-                            u8 count);
+void qede_recycle_rx_bd_ring(struct qede_rx_queue *rxq, u8 count);
 void qede_update_rx_prod(struct qede_dev *edev, struct qede_rx_queue *rxq);
 
 #define RX_RING_SIZE_POW       13
index ef8c3276065e91339bd669e74bd5e97afbe02040..60a2e589785e2080aa1d1949e0149981055b0a5e 100644 (file)
@@ -1337,13 +1337,13 @@ static int qede_selftest_receive_traffic(struct qede_dev *edev)
                                        break;
                                }
 
-                       qede_recycle_rx_bd_ring(rxq, edev, 1);
+                       qede_recycle_rx_bd_ring(rxq, 1);
                        qed_chain_recycle_consumed(&rxq->rx_comp_ring);
                        break;
                }
 
                DP_INFO(edev, "Not the transmitted packet\n");
-               qede_recycle_rx_bd_ring(rxq, edev, 1);
+               qede_recycle_rx_bd_ring(rxq, 1);
                qed_chain_recycle_consumed(&rxq->rx_comp_ring);
        }
 
index c9cae3e28ff493b77908ad9fd0985760a010b520..78beef26d6b73e5130881133be354030f39a7fd2 100644 (file)
@@ -96,8 +96,6 @@ static int qede_probe(struct pci_dev *pdev, const struct pci_device_id *id);
 
 static void qede_remove(struct pci_dev *pdev);
 static void qede_shutdown(struct pci_dev *pdev);
-static int qede_alloc_rx_buffer(struct qede_dev *edev,
-                               struct qede_rx_queue *rxq);
 static void qede_link_update(void *dev, struct qed_link_output *link);
 
 /* The qede lock is used to protect driver state change and driver flows that
@@ -355,8 +353,7 @@ static int qede_free_tx_pkt(struct qede_dev *edev,
 }
 
 /* Unmap the data and free skb when mapping failed during start_xmit */
-static void qede_free_failed_tx_pkt(struct qede_dev *edev,
-                                   struct qede_tx_queue *txq,
+static void qede_free_failed_tx_pkt(struct qede_tx_queue *txq,
                                    struct eth_tx_1st_bd *first_bd,
                                    int nbd, bool data_split)
 {
@@ -378,7 +375,7 @@ static void qede_free_failed_tx_pkt(struct qede_dev *edev,
                nbd--;
        }
 
-       dma_unmap_single(&edev->pdev->dev, BD_UNMAP_ADDR(first_bd),
+       dma_unmap_single(txq->dev, BD_UNMAP_ADDR(first_bd),
                         BD_UNMAP_LEN(first_bd) + split_bd_len, DMA_TO_DEVICE);
 
        /* Unmap the data of the skb frags */
@@ -386,7 +383,7 @@ static void qede_free_failed_tx_pkt(struct qede_dev *edev,
                tx_data_bd = (struct eth_tx_bd *)
                        qed_chain_produce(&txq->tx_pbl);
                if (tx_data_bd->nbytes)
-                       dma_unmap_page(&edev->pdev->dev,
+                       dma_unmap_page(txq->dev,
                                       BD_UNMAP_ADDR(tx_data_bd),
                                       BD_UNMAP_LEN(tx_data_bd), DMA_TO_DEVICE);
        }
@@ -401,8 +398,7 @@ static void qede_free_failed_tx_pkt(struct qede_dev *edev,
        txq->sw_tx_ring[idx].flags = 0;
 }
 
-static u32 qede_xmit_type(struct qede_dev *edev,
-                         struct sk_buff *skb, int *ipv6_ext)
+static u32 qede_xmit_type(struct sk_buff *skb, int *ipv6_ext)
 {
        u32 rc = XMIT_L4_CSUM;
        __be16 l3_proto;
@@ -469,18 +465,16 @@ static void qede_set_params_for_ipv6_ext(struct sk_buff *skb,
        second_bd->data.bitfields2 = cpu_to_le16(bd2_bits2);
 }
 
-static int map_frag_to_bd(struct qede_dev *edev,
+static int map_frag_to_bd(struct qede_tx_queue *txq,
                          skb_frag_t *frag, struct eth_tx_bd *bd)
 {
        dma_addr_t mapping;
 
        /* Map skb non-linear frag data for DMA */
-       mapping = skb_frag_dma_map(&edev->pdev->dev, frag, 0,
+       mapping = skb_frag_dma_map(txq->dev, frag, 0,
                                   skb_frag_size(frag), DMA_TO_DEVICE);
-       if (unlikely(dma_mapping_error(&edev->pdev->dev, mapping))) {
-               DP_NOTICE(edev, "Unable to map frag - dropping packet\n");
+       if (unlikely(dma_mapping_error(txq->dev, mapping)))
                return -ENOMEM;
-       }
 
        /* Setup the data pointer of the frag data */
        BD_SET_UNMAP_ADDR_LEN(bd, mapping, skb_frag_size(frag));
@@ -500,8 +494,7 @@ static u16 qede_get_skb_hlen(struct sk_buff *skb, bool is_encap_pkt)
 
 /* +2 for 1st BD for headers and 2nd BD for headlen (if required) */
 #if ((MAX_SKB_FRAGS + 2) > ETH_TX_MAX_BDS_PER_NON_LSO_PACKET)
-static bool qede_pkt_req_lin(struct qede_dev *edev, struct sk_buff *skb,
-                            u8 xmit_type)
+static bool qede_pkt_req_lin(struct sk_buff *skb, u8 xmit_type)
 {
        int allowed_frags = ETH_TX_MAX_BDS_PER_NON_LSO_PACKET - 1;
 
@@ -565,10 +558,10 @@ static netdev_tx_t qede_start_xmit(struct sk_buff *skb,
 
        WARN_ON(qed_chain_get_elem_left(&txq->tx_pbl) < (MAX_SKB_FRAGS + 1));
 
-       xmit_type = qede_xmit_type(edev, skb, &ipv6_ext);
+       xmit_type = qede_xmit_type(skb, &ipv6_ext);
 
 #if ((MAX_SKB_FRAGS + 2) > ETH_TX_MAX_BDS_PER_NON_LSO_PACKET)
-       if (qede_pkt_req_lin(edev, skb, xmit_type)) {
+       if (qede_pkt_req_lin(skb, xmit_type)) {
                if (skb_linearize(skb)) {
                        DP_NOTICE(edev,
                                  "SKB linearization failed - silently dropping this SKB\n");
@@ -588,11 +581,11 @@ static netdev_tx_t qede_start_xmit(struct sk_buff *skb,
                1 << ETH_TX_1ST_BD_FLAGS_START_BD_SHIFT;
 
        /* Map skb linear data for DMA and set in the first BD */
-       mapping = dma_map_single(&edev->pdev->dev, skb->data,
+       mapping = dma_map_single(txq->dev, skb->data,
                                 skb_headlen(skb), DMA_TO_DEVICE);
-       if (unlikely(dma_mapping_error(&edev->pdev->dev, mapping))) {
+       if (unlikely(dma_mapping_error(txq->dev, mapping))) {
                DP_NOTICE(edev, "SKB mapping failed\n");
-               qede_free_failed_tx_pkt(edev, txq, first_bd, 0, false);
+               qede_free_failed_tx_pkt(txq, first_bd, 0, false);
                qede_update_tx_producer(txq);
                return NETDEV_TX_OK;
        }
@@ -716,12 +709,11 @@ static netdev_tx_t qede_start_xmit(struct sk_buff *skb,
        /* Handle fragmented skb */
        /* special handle for frags inside 2nd and 3rd bds.. */
        while (tx_data_bd && frag_idx < skb_shinfo(skb)->nr_frags) {
-               rc = map_frag_to_bd(edev,
+               rc = map_frag_to_bd(txq,
                                    &skb_shinfo(skb)->frags[frag_idx],
                                    tx_data_bd);
                if (rc) {
-                       qede_free_failed_tx_pkt(edev, txq, first_bd, nbd,
-                                               data_split);
+                       qede_free_failed_tx_pkt(txq, first_bd, nbd, data_split);
                        qede_update_tx_producer(txq);
                        return NETDEV_TX_OK;
                }
@@ -741,12 +733,11 @@ static netdev_tx_t qede_start_xmit(struct sk_buff *skb,
 
                memset(tx_data_bd, 0, sizeof(*tx_data_bd));
 
-               rc = map_frag_to_bd(edev,
+               rc = map_frag_to_bd(txq,
                                    &skb_shinfo(skb)->frags[frag_idx],
                                    tx_data_bd);
                if (rc) {
-                       qede_free_failed_tx_pkt(edev, txq, first_bd, nbd,
-                                               data_split);
+                       qede_free_failed_tx_pkt(txq, first_bd, nbd, data_split);
                        qede_update_tx_producer(txq);
                        return NETDEV_TX_OK;
                }
@@ -903,8 +894,7 @@ static inline void qede_rx_bd_ring_consume(struct qede_rx_queue *rxq)
 /* This function reuses the buffer(from an offset) from
  * consumer index to producer index in the bd ring
  */
-static inline void qede_reuse_page(struct qede_dev *edev,
-                                  struct qede_rx_queue *rxq,
+static inline void qede_reuse_page(struct qede_rx_queue *rxq,
                                   struct sw_rx_data *curr_cons)
 {
        struct eth_rx_bd *rx_bd_prod = qed_chain_produce(&rxq->rx_bd_ring);
@@ -926,27 +916,62 @@ static inline void qede_reuse_page(struct qede_dev *edev,
 /* In case of allocation failures reuse buffers
  * from consumer index to produce buffers for firmware
  */
-void qede_recycle_rx_bd_ring(struct qede_rx_queue *rxq,
-                            struct qede_dev *edev, u8 count)
+void qede_recycle_rx_bd_ring(struct qede_rx_queue *rxq, u8 count)
 {
        struct sw_rx_data *curr_cons;
 
        for (; count > 0; count--) {
                curr_cons = &rxq->sw_rx_ring[rxq->sw_rx_cons & NUM_RX_BDS_MAX];
-               qede_reuse_page(edev, rxq, curr_cons);
+               qede_reuse_page(rxq, curr_cons);
                qede_rx_bd_ring_consume(rxq);
        }
 }
 
-static inline int qede_realloc_rx_buffer(struct qede_dev *edev,
-                                        struct qede_rx_queue *rxq,
+static int qede_alloc_rx_buffer(struct qede_rx_queue *rxq)
+{
+       struct sw_rx_data *sw_rx_data;
+       struct eth_rx_bd *rx_bd;
+       dma_addr_t mapping;
+       struct page *data;
+
+       data = alloc_pages(GFP_ATOMIC, 0);
+       if (unlikely(!data))
+               return -ENOMEM;
+
+       /* Map the entire page as it would be used
+        * for multiple RX buffer segment size mapping.
+        */
+       mapping = dma_map_page(rxq->dev, data, 0,
+                              PAGE_SIZE, DMA_FROM_DEVICE);
+       if (unlikely(dma_mapping_error(rxq->dev, mapping))) {
+               __free_page(data);
+               return -ENOMEM;
+       }
+
+       sw_rx_data = &rxq->sw_rx_ring[rxq->sw_rx_prod & NUM_RX_BDS_MAX];
+       sw_rx_data->page_offset = 0;
+       sw_rx_data->data = data;
+       sw_rx_data->mapping = mapping;
+
+       /* Advance PROD and get BD pointer */
+       rx_bd = (struct eth_rx_bd *)qed_chain_produce(&rxq->rx_bd_ring);
+       WARN_ON(!rx_bd);
+       rx_bd->addr.hi = cpu_to_le32(upper_32_bits(mapping));
+       rx_bd->addr.lo = cpu_to_le32(lower_32_bits(mapping));
+
+       rxq->sw_rx_prod++;
+
+       return 0;
+}
+
+static inline int qede_realloc_rx_buffer(struct qede_rx_queue *rxq,
                                         struct sw_rx_data *curr_cons)
 {
        /* Move to the next segment in the page */
        curr_cons->page_offset += rxq->rx_buf_seg_size;
 
        if (curr_cons->page_offset == PAGE_SIZE) {
-               if (unlikely(qede_alloc_rx_buffer(edev, rxq))) {
+               if (unlikely(qede_alloc_rx_buffer(rxq))) {
                        /* Since we failed to allocate new buffer
                         * current buffer can be used again.
                         */
@@ -955,7 +980,7 @@ static inline int qede_realloc_rx_buffer(struct qede_dev *edev,
                        return -ENOMEM;
                }
 
-               dma_unmap_page(&edev->pdev->dev, curr_cons->mapping,
+               dma_unmap_page(rxq->dev, curr_cons->mapping,
                               PAGE_SIZE, DMA_FROM_DEVICE);
        } else {
                /* Increment refcount of the page as we don't want
@@ -963,7 +988,7 @@ static inline int qede_realloc_rx_buffer(struct qede_dev *edev,
                 * which can be recycled multiple times by the driver.
                 */
                page_ref_inc(curr_cons->data);
-               qede_reuse_page(edev, rxq, curr_cons);
+               qede_reuse_page(rxq, curr_cons);
        }
 
        return 0;
@@ -1026,6 +1051,7 @@ static void qede_set_skb_csum(struct sk_buff *skb, u8 csum_flag)
 
 static inline void qede_skb_receive(struct qede_dev *edev,
                                    struct qede_fastpath *fp,
+                                   struct qede_rx_queue *rxq,
                                    struct sk_buff *skb, u16 vlan_tag)
 {
        if (vlan_tag)
@@ -1068,7 +1094,7 @@ static int qede_fill_frag_skb(struct qede_dev *edev,
                           current_bd->data, current_bd->page_offset,
                           len_on_bd);
 
-       if (unlikely(qede_realloc_rx_buffer(edev, rxq, current_bd))) {
+       if (unlikely(qede_realloc_rx_buffer(rxq, current_bd))) {
                /* Incr page ref count to reuse on allocation failure
                 * so that it doesn't get freed while freeing SKB.
                 */
@@ -1087,7 +1113,8 @@ static int qede_fill_frag_skb(struct qede_dev *edev,
 
 out:
        tpa_info->state = QEDE_AGG_STATE_ERROR;
-       qede_recycle_rx_bd_ring(rxq, edev, 1);
+       qede_recycle_rx_bd_ring(rxq, 1);
+
        return -ENOMEM;
 }
 
@@ -1239,7 +1266,7 @@ static void qede_gro_receive(struct qede_dev *edev,
 
 send_skb:
        skb_record_rx_queue(skb, fp->rxq->rxq_id);
-       qede_skb_receive(edev, fp, skb, vlan_tag);
+       qede_skb_receive(edev, fp, fp->rxq, skb, vlan_tag);
 }
 
 static inline void qede_tpa_cont(struct qede_dev *edev,
@@ -1414,7 +1441,7 @@ static struct sk_buff *qede_rx_allocate_skb(struct qede_dev *edev,
        if (len + pad <= edev->rx_copybreak) {
                memcpy(skb_put(skb, len),
                       page_address(page) + pad + offset, len);
-               qede_reuse_page(edev, rxq, bd);
+               qede_reuse_page(rxq, bd);
                goto out;
        }
 
@@ -1435,7 +1462,7 @@ static struct sk_buff *qede_rx_allocate_skb(struct qede_dev *edev,
        skb->data_len -= pull_len;
        skb->tail += pull_len;
 
-       if (unlikely(qede_realloc_rx_buffer(edev, rxq, bd))) {
+       if (unlikely(qede_realloc_rx_buffer(rxq, bd))) {
                /* Incr page ref count to reuse on allocation failure so
                 * that it doesn't get freed while freeing SKB [as its
                 * already mapped there].
@@ -1477,7 +1504,7 @@ static int qede_rx_build_jumbo(struct qede_dev *edev,
                }
 
                /* We need a replacement buffer for each BD */
-               if (unlikely(qede_alloc_rx_buffer(edev, rxq)))
+               if (unlikely(qede_alloc_rx_buffer(rxq)))
                        goto out;
 
                /* Now that we've allocated the replacement buffer,
@@ -1487,7 +1514,7 @@ static int qede_rx_build_jumbo(struct qede_dev *edev,
                bd = &rxq->sw_rx_ring[bd_cons_idx];
                qede_rx_bd_ring_consume(rxq);
 
-               dma_unmap_page(&edev->pdev->dev, bd->mapping,
+               dma_unmap_page(rxq->dev, bd->mapping,
                               PAGE_SIZE, DMA_FROM_DEVICE);
 
                skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags++,
@@ -1582,7 +1609,7 @@ static int qede_rx_process_cqe(struct qede_dev *edev,
                                  "CQE has error, flags = %x, dropping incoming packet\n",
                                  parse_flag);
                        rxq->rx_hw_errors++;
-                       qede_recycle_rx_bd_ring(rxq, edev, fp_cqe->bd_num);
+                       qede_recycle_rx_bd_ring(rxq, fp_cqe->bd_num);
                        return 0;
                }
        }
@@ -1593,7 +1620,7 @@ static int qede_rx_process_cqe(struct qede_dev *edev,
        skb = qede_rx_allocate_skb(edev, rxq, bd, len, pad);
        if (!skb) {
                rxq->rx_alloc_errors++;
-               qede_recycle_rx_bd_ring(rxq, edev, fp_cqe->bd_num);
+               qede_recycle_rx_bd_ring(rxq, fp_cqe->bd_num);
                return 0;
        }
 
@@ -1605,7 +1632,7 @@ static int qede_rx_process_cqe(struct qede_dev *edev,
                                                         fp_cqe, len);
 
                if (unlikely(unmapped_frags > 0)) {
-                       qede_recycle_rx_bd_ring(rxq, edev, unmapped_frags);
+                       qede_recycle_rx_bd_ring(rxq, unmapped_frags);
                        dev_kfree_skb_any(skb);
                        return 0;
                }
@@ -1618,7 +1645,7 @@ static int qede_rx_process_cqe(struct qede_dev *edev,
        skb_record_rx_queue(skb, rxq->rxq_id);
 
        /* SKB is prepared - pass it to stack */
-       qede_skb_receive(edev, fp, skb, le16_to_cpu(fp_cqe->vlan_tag));
+       qede_skb_receive(edev, fp, rxq, skb, le16_to_cpu(fp_cqe->vlan_tag));
 
        return 1;
 }
@@ -2875,47 +2902,6 @@ static void qede_free_mem_rxq(struct qede_dev *edev, struct qede_rx_queue *rxq)
        edev->ops->common->chain_free(edev->cdev, &rxq->rx_comp_ring);
 }
 
-static int qede_alloc_rx_buffer(struct qede_dev *edev,
-                               struct qede_rx_queue *rxq)
-{
-       struct sw_rx_data *sw_rx_data;
-       struct eth_rx_bd *rx_bd;
-       dma_addr_t mapping;
-       struct page *data;
-
-       data = alloc_pages(GFP_ATOMIC, 0);
-       if (unlikely(!data)) {
-               DP_NOTICE(edev, "Failed to allocate Rx data [page]\n");
-               return -ENOMEM;
-       }
-
-       /* Map the entire page as it would be used
-        * for multiple RX buffer segment size mapping.
-        */
-       mapping = dma_map_page(&edev->pdev->dev, data, 0,
-                              PAGE_SIZE, DMA_FROM_DEVICE);
-       if (unlikely(dma_mapping_error(&edev->pdev->dev, mapping))) {
-               __free_page(data);
-               DP_NOTICE(edev, "Failed to map Rx buffer\n");
-               return -ENOMEM;
-       }
-
-       sw_rx_data = &rxq->sw_rx_ring[rxq->sw_rx_prod & NUM_RX_BDS_MAX];
-       sw_rx_data->page_offset = 0;
-       sw_rx_data->data = data;
-       sw_rx_data->mapping = mapping;
-
-       /* Advance PROD and get BD pointer */
-       rx_bd = (struct eth_rx_bd *)qed_chain_produce(&rxq->rx_bd_ring);
-       WARN_ON(!rx_bd);
-       rx_bd->addr.hi = cpu_to_le32(upper_32_bits(mapping));
-       rx_bd->addr.lo = cpu_to_le32(lower_32_bits(mapping));
-
-       rxq->sw_rx_prod++;
-
-       return 0;
-}
-
 static int qede_alloc_sge_mem(struct qede_dev *edev, struct qede_rx_queue *rxq)
 {
        dma_addr_t mapping;
@@ -3010,7 +2996,7 @@ static int qede_alloc_mem_rxq(struct qede_dev *edev, struct qede_rx_queue *rxq)
 
        /* Allocate buffers for the Rx ring */
        for (i = 0; i < rxq->num_rx_buffers; i++) {
-               rc = qede_alloc_rx_buffer(edev, rxq);
+               rc = qede_alloc_rx_buffer(rxq);
                if (rc) {
                        DP_ERR(edev,
                               "Rx buffers allocation failed at index %d\n", i);
@@ -3151,12 +3137,14 @@ static void qede_init_fp(struct qede_dev *edev)
 
                if (fp->type & QEDE_FASTPATH_RX) {
                        fp->rxq->rxq_id = rxq_index++;
+                       fp->rxq->dev = &edev->pdev->dev;
                }
 
                if (fp->type & QEDE_FASTPATH_TX) {
                        fp->txq->index = txq_index++;
                        if (edev->dev_info.is_legacy)
                                fp->txq->is_legacy = 1;
+                       fp->txq->dev = &edev->pdev->dev;
                }
 
                snprintf(fp->name, sizeof(fp->name), "%s-fp-%d",