igb: optimize/refactor receive path
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>
diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
index 8b80fe3..21f0c22 100644
--- a/drivers/net/igb/igb_main.c
+++ b/drivers/net/igb/igb_main.c
@@ -3803,6 +3803,7 @@
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);
@@ -3810,7 +3811,30 @@
if (*work_done >= budget)
break;
(*work_done)++;
- buffer_info = &rx_ring->buffer_info[i];
+
+ 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++;
+
+ if (!adapter->rx_ps_hdr_size) {
+ pci_unmap_single(pdev, buffer_info->dma,
+ adapter->rx_buffer_len +
+ NET_IP_ALIGN,
+ PCI_DMA_FROMDEVICE);
+ skb_put(skb, length);
+ 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
@@ -3822,22 +3846,6 @@
if (hlen > adapter->rx_ps_hdr_size)
hlen = adapter->rx_ps_hdr_size;
- 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 +
- NET_IP_ALIGN,
- PCI_DMA_FROMDEVICE);
- skb_put(skb, length);
- goto send_up;
- }
-
if (!skb_shinfo(skb)->nr_frags) {
pci_unmap_single(pdev, buffer_info->dma,
adapter->rx_ps_hdr_size +
@@ -3867,13 +3875,6 @@
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 @@
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 @@
/* use prefetched values */
rx_desc = next_rxd;
buffer_info = next_buffer;
-
staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
}