ixgbe: delay tail write to every 'n' packets

Current XDP implementation hits the tail on every XDP_TX return
code. This patch changes driver behavior to only hit the tail after
packet processing is complete.

With this patch I can run XDP drop programs @ 14+Mpps and XDP_TX
programs are at ~13.5Mpps.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This commit is contained in:
John Fastabend 2017-03-28 09:47:03 -07:00 committed by Jeff Kirsher
parent 33fdc82f08
commit 7379f97a4f
1 changed files with 17 additions and 11 deletions

View File

@ -2283,6 +2283,7 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
unsigned int mss = 0;
#endif /* IXGBE_FCOE */
u16 cleaned_count = ixgbe_desc_unused(rx_ring);
bool xdp_xmit = false;
while (likely(total_rx_packets < budget)) {
union ixgbe_adv_rx_desc *rx_desc;
@ -2322,10 +2323,12 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
}
if (IS_ERR(skb)) {
if (PTR_ERR(skb) == -IXGBE_XDP_TX)
if (PTR_ERR(skb) == -IXGBE_XDP_TX) {
xdp_xmit = true;
ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
else
} else {
rx_buffer->pagecnt_bias++;
}
total_rx_packets++;
total_rx_bytes += size;
} else if (skb) {
@ -2393,6 +2396,16 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
total_rx_packets++;
}
if (xdp_xmit) {
struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
/* Force memory writes to complete before letting h/w
* know there are new descriptors to fetch.
*/
wmb();
writel(ring->next_to_use, ring->tail);
}
u64_stats_update_begin(&rx_ring->syncp);
rx_ring->stats.packets += total_rx_packets;
rx_ring->stats.bytes += total_rx_bytes;
@ -8238,14 +8251,8 @@ static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
tx_desc->read.olinfo_status =
cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
/* Force memory writes to complete before letting h/w know there
* are new descriptors to fetch. (Only applicable for weak-ordered
* memory model archs, such as IA-64).
*
* We also need this memory barrier to make certain all of the
* status bits have been updated before next_to_watch is written.
*/
wmb();
/* Avoid any potential race with xdp_xmit and cleanup */
smp_wmb();
/* set next_to_watch value indicating a packet is present */
i++;
@ -8255,7 +8262,6 @@ static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
tx_buffer->next_to_watch = tx_desc;
ring->next_to_use = i;
writel(i, ring->tail);
return IXGBE_XDP_TX;
}