[PATCH net-next] dp83640: don't hold spinlock while calling netif_rx_ni

[PATCH net-next] dp83640: don't hold spinlock while calling netif_rx_ni

[Stefan Sørensen]

We should not hold a spinlock while pushing the skb into the networking
stack, so move the call to netif_rx_ni out of the critical region to where
we have dropped the spinlock.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
---
 drivers/net/phy/dp83640.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/net/phy/dp83640.c b/drivers/net/phy/dp83640.c
index c3065236ffcc..cbd629822f04 100644
--- a/drivers/net/phy/dp83640.c
+++ b/drivers/net/phy/dp83640.c
@@ -874,7 +874,6 @@ static void decode_rxts(struct dp83640_private *dp83640,
 			shhwtstamps = skb_hwtstamps(skb);
 			memset(shhwtstamps, 0, sizeof(*shhwtstamps));
 			shhwtstamps->hwtstamp = ns_to_ktime(rxts->ns);
-			netif_rx_ni(skb);
 			list_add(&rxts->list, &dp83640->rxpool);
 			break;
 		}
@@ -885,6 +884,9 @@ static void decode_rxts(struct dp83640_private *dp83640,
 		list_add_tail(&rxts->list, &dp83640->rxts);
 out:
 	spin_unlock_irqrestore(&dp83640->rx_lock, flags);
+
+	if (shhwtstamps)
+		netif_rx_ni(skb);
 }
 
 static void decode_txts(struct dp83640_private *dp83640,
@@ -1425,7 +1427,6 @@ static bool dp83640_rxtstamp(struct phy_device *phydev,
 			shhwtstamps = skb_hwtstamps(skb);
 			memset(shhwtstamps, 0, sizeof(*shhwtstamps));
 			shhwtstamps->hwtstamp = ns_to_ktime(rxts->ns);
-			netif_rx_ni(skb);
 			list_del_init(&rxts->list);
 			list_add(&rxts->list, &dp83640->rxpool);
 			break;
@@ -1438,6 +1439,8 @@ static bool dp83640_rxtstamp(struct phy_device *phydev,
 		skb_info->tmo = jiffies + SKB_TIMESTAMP_TIMEOUT;
 		skb_queue_tail(&dp83640->rx_queue, skb);
 		schedule_delayed_work(&dp83640->ts_work, SKB_TIMESTAMP_TIMEOUT);
+	} else {
+		netif_rx_ni(skb);
 	}
 
 	return true;
-- 
2.13.5

Re: [PATCH net-next] dp83640: don't hold spinlock while calling netif_rx_ni

[David Miller]

From: Stefan Sørensen <stefan.sorensen@spectralink.com>
Date: Wed, 30 Aug 2017 08:58:47 +0200

> We should not hold a spinlock while pushing the skb into the networking
> stack, so move the call to netif_rx_ni out of the critical region to where
> we have dropped the spinlock.
> 
> Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>

Looks good, applied, thanks.

Re: [PATCH net-next] dp83640: don't hold spinlock while calling netif_rx_ni

[Richard Cochran]

On Wed, Aug 30, 2017 at 02:50:31PM -0700, David Miller wrote:
> From: Stefan Sørensen <stefan.sorensen@spectralink.com>
> Date: Wed, 30 Aug 2017 08:58:47 +0200
> 
> > We should not hold a spinlock while pushing the skb into the networking
> > stack, so move the call to netif_rx_ni out of the critical region to where
> > we have dropped the spinlock.
> > 
> > Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
> 
> Looks good, applied, thanks.

And thanks, Stefan, for following up on this.  For the record, we
discussed this fix back in April.

    https://patchwork.ozlabs.org/patch/752201/

Cheers,
Richard