diff mbox

[v2,net-next] ixgbe: flush when in xmit_more mode and under descriptor pressure

Message ID 1409074458-17989-1-git-send-email-dborkman@redhat.com
State Accepted, archived
Delegated to: David Miller
Headers show

Commit Message

Daniel Borkmann Aug. 26, 2014, 5:34 p.m. UTC 
When xmit_more mode is being used and the ring is about to
become full or the stack has stopped the ring, enforce a tail
pointer write to the hw. Otherwise, we could risk a TX hang.

Code suggested by Alexander Duyck.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
---
 v1->v2:
  - Worked in Alex' feedback; in accordance w/ Alex, sending out v2
  - Rerun tests, looks good

 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 63 +++++++++++++++------------
 1 file changed, 34 insertions(+), 29 deletions(-)

Comments

David Miller Aug. 28, 2014, 6:18 a.m. UTC | #1 
From: Daniel Borkmann <dborkman@redhat.com>
Date: Tue, 26 Aug 2014 19:34:18 +0200

> When xmit_more mode is being used and the ring is about to
> become full or the stack has stopped the ring, enforce a tail
> pointer write to the hw. Otherwise, we could risk a TX hang.
> 
> Code suggested by Alexander Duyck.
> 
> Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
> Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
> ---
>  v1->v2:
>   - Worked in Alex' feedback; in accordance w/ Alex, sending out v2
>   - Rerun tests, looks good

Applied, thanks Daniel.

I'll make similar mods to igb and vhost_net soon.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Daniel Borkmann Aug. 28, 2014, 11:46 a.m. UTC | #2 
On 08/28/2014 08:18 AM, David Miller wrote:
> From: Daniel Borkmann <dborkman@redhat.com>
> Date: Tue, 26 Aug 2014 19:34:18 +0200
>
>> When xmit_more mode is being used and the ring is about to
>> become full or the stack has stopped the ring, enforce a tail
>> pointer write to the hw. Otherwise, we could risk a TX hang.
>>
>> Code suggested by Alexander Duyck.
>>
>> Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
>> Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
>> ---
>>   v1->v2:
>>    - Worked in Alex' feedback; in accordance w/ Alex, sending out v2
>>    - Rerun tests, looks good
>
> Applied, thanks Daniel.
>
> I'll make similar mods to igb and vhost_net soon.

Cool, at some point later, we might want to put ...

   netif_xmit_stopped(<tx_ring>) || !skb->xmit_more

... into an API perhaps, but that can likely wait right now.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index ba9ceaa..53fbf06 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -6837,6 +6837,36 @@  static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
 	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
 }
 
+static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
+{
+	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
+
+	/* Herbert's original patch had:
+	 *  smp_mb__after_netif_stop_queue();
+	 * but since that doesn't exist yet, just open code it.
+	 */
+	smp_mb();
+
+	/* We need to check again in a case another CPU has just
+	 * made room available.
+	 */
+	if (likely(ixgbe_desc_unused(tx_ring) < size))
+		return -EBUSY;
+
+	/* A reprieve! - use start_queue because it doesn't call schedule */
+	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
+	++tx_ring->tx_stats.restart_queue;
+	return 0;
+}
+
+static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
+{
+	if (likely(ixgbe_desc_unused(tx_ring) >= size))
+		return 0;
+
+	return __ixgbe_maybe_stop_tx(tx_ring, size);
+}
+
 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
 		       IXGBE_TXD_CMD_RS)
 
@@ -6958,10 +6988,13 @@  static void ixgbe_tx_map(struct ixgbe_ring *tx_ring,
 
 	tx_ring->next_to_use = i;
 
-	if (!skb->xmit_more) {
+	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
+
+	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
 		/* notify HW of packet */
 		ixgbe_write_tail(tx_ring, i);
 	}
+
 	return;
 dma_error:
 	dev_err(tx_ring->dev, "TX DMA map failed\n");
@@ -7068,32 +7101,6 @@  static void ixgbe_atr(struct ixgbe_ring *ring,
 					      input, common, ring->queue_index);
 }
 
-static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
-{
-	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
-	/* Herbert's original patch had:
-	 *  smp_mb__after_netif_stop_queue();
-	 * but since that doesn't exist yet, just open code it. */
-	smp_mb();
-
-	/* We need to check again in a case another CPU has just
-	 * made room available. */
-	if (likely(ixgbe_desc_unused(tx_ring) < size))
-		return -EBUSY;
-
-	/* A reprieve! - use start_queue because it doesn't call schedule */
-	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
-	++tx_ring->tx_stats.restart_queue;
-	return 0;
-}
-
-static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
-{
-	if (likely(ixgbe_desc_unused(tx_ring) >= size))
-		return 0;
-	return __ixgbe_maybe_stop_tx(tx_ring, size);
-}
-
 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
 			      void *accel_priv, select_queue_fallback_t fallback)
 {
@@ -7262,8 +7269,6 @@  xmit_fcoe:
 #endif /* IXGBE_FCOE */
 	ixgbe_tx_map(tx_ring, first, hdr_len);
 
-	ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
-
 	return NETDEV_TX_OK;
 
 out_drop: