igb: optimize/refactor receive path
authorAlexander Duyck <alexander.h.duyck@intel.com>
Fri, 6 Feb 2009 23:15:04 +0000 (23:15 +0000)
committerDavid S. Miller <davem@davemloft.net>
Sat, 7 Feb 2009 10:43:01 +0000 (02:43 -0800)
While cleaning up the skb_over panic with small frames I found there was
room for improvement in the ordering of operations within the rx receive
flow.  These changes will place the prefetch for the next descriptor to a
point earlier in the rx path.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/igb/igb_main.c

index 8b80fe3434352ce1835cb3097c68a9b32f1fe78a..21f0c229b64e6d82fcd5ef1284a77e37c8a5f4d1 100644 (file)
@@ -3803,6 +3803,7 @@ static bool igb_clean_rx_irq_adv(struct igb_ring *rx_ring,
        unsigned int total_bytes = 0, total_packets = 0;
 
        i = rx_ring->next_to_clean;
+       buffer_info = &rx_ring->buffer_info[i];
        rx_desc = E1000_RX_DESC_ADV(*rx_ring, i);
        staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
 
@@ -3810,25 +3811,22 @@ static bool igb_clean_rx_irq_adv(struct igb_ring *rx_ring,
                if (*work_done >= budget)
                        break;
                (*work_done)++;
-               buffer_info = &rx_ring->buffer_info[i];
 
-               /* HW will not DMA in data larger than the given buffer, even
-                * if it parses the (NFS, of course) header to be larger.  In
-                * that case, it fills the header buffer and spills the rest
-                * into the page.
-                */
-               hlen = (le16_to_cpu(rx_desc->wb.lower.lo_dword.hdr_info) &
-                 E1000_RXDADV_HDRBUFLEN_MASK) >> E1000_RXDADV_HDRBUFLEN_SHIFT;
-               if (hlen > adapter->rx_ps_hdr_size)
-                       hlen = adapter->rx_ps_hdr_size;
+               skb = buffer_info->skb;
+               prefetch(skb->data - NET_IP_ALIGN);
+               buffer_info->skb = NULL;
+
+               i++;
+               if (i == rx_ring->count)
+                       i = 0;
+               next_rxd = E1000_RX_DESC_ADV(*rx_ring, i);
+               prefetch(next_rxd);
+               next_buffer = &rx_ring->buffer_info[i];
 
                length = le16_to_cpu(rx_desc->wb.upper.length);
                cleaned = true;
                cleaned_count++;
 
-               skb = buffer_info->skb;
-               prefetch(skb->data - NET_IP_ALIGN);
-               buffer_info->skb = NULL;
                if (!adapter->rx_ps_hdr_size) {
                        pci_unmap_single(pdev, buffer_info->dma,
                                         adapter->rx_buffer_len +
@@ -3838,6 +3836,16 @@ static bool igb_clean_rx_irq_adv(struct igb_ring *rx_ring,
                        goto send_up;
                }
 
+               /* HW will not DMA in data larger than the given buffer, even
+                * if it parses the (NFS, of course) header to be larger.  In
+                * that case, it fills the header buffer and spills the rest
+                * into the page.
+                */
+               hlen = (le16_to_cpu(rx_desc->wb.lower.lo_dword.hdr_info) &
+                 E1000_RXDADV_HDRBUFLEN_MASK) >> E1000_RXDADV_HDRBUFLEN_SHIFT;
+               if (hlen > adapter->rx_ps_hdr_size)
+                       hlen = adapter->rx_ps_hdr_size;
+
                if (!skb_shinfo(skb)->nr_frags) {
                        pci_unmap_single(pdev, buffer_info->dma,
                                         adapter->rx_ps_hdr_size +
@@ -3867,13 +3875,6 @@ static bool igb_clean_rx_irq_adv(struct igb_ring *rx_ring,
 
                        skb->truesize += length;
                }
-send_up:
-               i++;
-               if (i == rx_ring->count)
-                       i = 0;
-               next_rxd = E1000_RX_DESC_ADV(*rx_ring, i);
-               prefetch(next_rxd);
-               next_buffer = &rx_ring->buffer_info[i];
 
                if (!(staterr & E1000_RXD_STAT_EOP)) {
                        buffer_info->skb = next_buffer->skb;
@@ -3882,7 +3883,7 @@ send_up:
                        next_buffer->dma = 0;
                        goto next_desc;
                }
-
+send_up:
                if (staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) {
                        dev_kfree_skb_irq(skb);
                        goto next_desc;
@@ -3909,7 +3910,6 @@ next_desc:
                /* use prefetched values */
                rx_desc = next_rxd;
                buffer_info = next_buffer;
-
                staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
        }