i40evf: enforce descriptor write-back mechanism for VF

The current driver mode is to use a write-back mechanism for the head
register which indicates transmit completions. The VF driver needs to be
able to work on hardware that exclusively uses descriptor write-back, so
change the default driver mode of operation to descriptor write-back for
VF. In our analysis, performance wasn't significantly different with
either write-back method.

Change-ID: Ia92e4ec77c2df8dc4515c71d53746d57d77759af
Signed-off-by: Preethi Banala <preethi.banala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Orabug: 26132523
(cherry picked from commit b1cb07db6e2b3c982ec858b06d42d24c7e267fdc)
Signed-off-by: Jack Vogel <jack.vogel@oracle.com>
Reviewed-by: Brian Maly <brian.maly@oracle.com>

diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 78a1af277fc63b28599a2118bd27296fda1f00d6..790fdfe42f012c115c0cde0fcc6d91f7245e9fc4 100644 (file)
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -138,10 +138,7 @@ u32 i40evf_get_tx_pending(struct i40e_ring *ring, bool in_sw)
 {
        u32 head, tail;
 
-       if (!in_sw)
-               head = i40e_get_head(ring);
-       else
-               head = ring->next_to_clean;
+       head = ring->next_to_clean;
        tail = readl(ring->tail);
 
        if (head != tail)
@@ -164,7 +161,6 @@ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget)
 {
        u16 i = tx_ring->next_to_clean;
        struct i40e_tx_buffer *tx_buf;
-       struct i40e_tx_desc *tx_head;
        struct i40e_tx_desc *tx_desc;
        unsigned int total_packets = 0;
        unsigned int total_bytes = 0;
@@ -173,8 +169,6 @@ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget)
        tx_desc = I40E_TX_DESC(tx_ring, i);
        i -= tx_ring->count;
 
-       tx_head = I40E_TX_DESC(tx_ring, i40e_get_head(tx_ring));
-
        do {
                struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
 
@@ -185,8 +179,9 @@ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget)
                /* prevent any other reads prior to eop_desc */
                read_barrier_depends();
 
-               /* we have caught up to head, no work left to do */
-               if (tx_head == tx_desc)
+               /* if the descriptor isn't done, no work yet to do */
+               if (!(eop_desc->cmd_type_offset_bsz &
+                     cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
                        break;
 
                /* clear next_to_watch to prevent false hangs */
@@ -463,10 +458,6 @@ int i40evf_setup_tx_descriptors(struct i40e_ring *tx_ring)
 
        /* round up to nearest 4K */
        tx_ring->size = tx_ring->count * sizeof(struct i40e_tx_desc);
-       /* add u32 for head writeback, align after this takes care of
-        * guaranteeing this is at least one cache line in size
-        */
-       tx_ring->size += sizeof(u32);
        tx_ring->size = ALIGN(tx_ring->size, 4096);
        tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
                                           &tx_ring->dma, GFP_KERNEL);
@@ -1968,7 +1959,6 @@ static inline void i40evf_tx_map(struct i40e_ring *tx_ring, struct sk_buff *skb,
        u16 i = tx_ring->next_to_use;
        u32 td_tag = 0;
        dma_addr_t dma;
-       u16 desc_count = 1;
 
        if (tx_flags & I40E_TX_FLAGS_HW_VLAN) {
                td_cmd |= I40E_TX_DESC_CMD_IL2TAG1;
@@ -2004,7 +1994,6 @@ static inline void i40evf_tx_map(struct i40e_ring *tx_ring, struct sk_buff *skb,
 
                        tx_desc++;
                        i++;
-                       desc_count++;
 
                        if (i == tx_ring->count) {
                                tx_desc = I40E_TX_DESC(tx_ring, 0);
@@ -2026,7 +2015,6 @@ static inline void i40evf_tx_map(struct i40e_ring *tx_ring, struct sk_buff *skb,
 
                tx_desc++;
                i++;
-               desc_count++;
 
                if (i == tx_ring->count) {
                        tx_desc = I40E_TX_DESC(tx_ring, 0);
@@ -2052,46 +2040,8 @@ static inline void i40evf_tx_map(struct i40e_ring *tx_ring, struct sk_buff *skb,
 
        i40e_maybe_stop_tx(tx_ring, DESC_NEEDED);
 
-       /* write last descriptor with EOP bit */
-       td_cmd |= I40E_TX_DESC_CMD_EOP;
-
-       /* We can OR these values together as they both are checked against
-        * 4 below and at this point desc_count will be used as a boolean value
-        * after this if/else block.
-        */
-       desc_count |= ++tx_ring->packet_stride;
-
-       /* Algorithm to optimize tail and RS bit setting:
-        * if queue is stopped
-        *      mark RS bit
-        *      reset packet counter
-        * else if xmit_more is supported and is true
-        *      advance packet counter to 4
-        *      reset desc_count to 0
-        *
-        * if desc_count >= 4
-        *      mark RS bit
-        *      reset packet counter
-        * if desc_count > 0
-        *      update tail
-        *
-        * Note: If there are less than 4 descriptors
-        * pending and interrupts were disabled the service task will
-        * trigger a force WB.
-        */
-       if (netif_xmit_stopped(txring_txq(tx_ring))) {
-               goto do_rs;
-       } else if (skb->xmit_more) {
-               /* set stride to arm on next packet and reset desc_count */
-               tx_ring->packet_stride = WB_STRIDE;
-               desc_count = 0;
-       } else if (desc_count >= WB_STRIDE) {
-do_rs:
-               /* write last descriptor with RS bit set */
-               td_cmd |= I40E_TX_DESC_CMD_RS;
-               tx_ring->packet_stride = 0;
-       }
-
+       /* write last descriptor with RS and EOP bits */
+       td_cmd |= I40E_TXD_CMD;
        tx_desc->cmd_type_offset_bsz =
                        build_ctob(td_cmd, td_offset, size, td_tag);
 
@@ -2107,7 +2057,7 @@ do_rs:
        first->next_to_watch = tx_desc;
 
        /* notify HW of packet */
-       if (desc_count) {
+       if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
                writel(i, tx_ring->tail);
 
                /* we need this if more than one processor can write to our tail

diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
index 8274ba68bd32a6583538f7af6fd80417dc25e6ec..17c84930f216193fce819acc6cd7996121e6a895 100644 (file)
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
@@ -384,20 +384,6 @@ u32 i40evf_get_tx_pending(struct i40e_ring *ring, bool in_sw);
 int __i40evf_maybe_stop_tx(struct i40e_ring *tx_ring, int size);
 bool __i40evf_chk_linearize(struct sk_buff *skb);
 
-/**
- * i40e_get_head - Retrieve head from head writeback
- * @tx_ring: Tx ring to fetch head of
- *
- * Returns value of Tx ring head based on value stored
- * in head write-back location
- **/
-static inline u32 i40e_get_head(struct i40e_ring *tx_ring)
-{
-       void *head = (struct i40e_tx_desc *)tx_ring->desc + tx_ring->count;
-
-       return le32_to_cpu(*(volatile __le32 *)head);
-}
-
 /**
  * i40e_xmit_descriptor_count - calculate number of Tx descriptors needed
  * @skb:     send buffer

diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c b/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c
index bee58af390e1c633006ddc38efddf256a307161a..519b5cf2c48ab98ef9d594158b888c8a46c92382 100644 (file)
--- a/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c
@@ -259,10 +259,6 @@ void i40evf_configure_queues(struct i40evf_adapter *adapter)
                vqpi->txq.queue_id = i;
                vqpi->txq.ring_len = adapter->tx_rings[i].count;
                vqpi->txq.dma_ring_addr = adapter->tx_rings[i].dma;
-               vqpi->txq.headwb_enabled = 1;
-               vqpi->txq.dma_headwb_addr = vqpi->txq.dma_ring_addr +
-                   (vqpi->txq.ring_len * sizeof(struct i40e_tx_desc));
-
                vqpi->rxq.vsi_id = vqci->vsi_id;
                vqpi->rxq.queue_id = i;
                vqpi->rxq.ring_len = adapter->rx_rings[i].count;