[PATCH net] pktgen: be friendly to LLTX devices

[PATCH net] pktgen: be friendly to LLTX devices

[Daniel Borkmann]

Similarly to commit 43279500deca ("packet: respect devices with
LLTX flag in direct xmit"), we can basically apply the very same
to pktgen. This will help testing against LLTX devices such as
dummy driver (or others), which only have a single netdevice txq
and would otherwise require locking their txq from pktgen side
while e.g. in dummy case, we would not need any locking. Fix this
by making use of HARD_TX_{UN,}LOCK API, so that NETIF_F_LLTX will
be respected.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Eric Dumazet <edumazet@google.com>
---
 net/core/pktgen.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/net/core/pktgen.c b/net/core/pktgen.c
index d068ec2..0304f98 100644
--- a/net/core/pktgen.c
+++ b/net/core/pktgen.c
@@ -3338,7 +3338,9 @@ static void pktgen_xmit(struct pktgen_dev *pkt_dev)
 	queue_map = skb_get_queue_mapping(pkt_dev->skb);
 	txq = netdev_get_tx_queue(odev, queue_map);
 
-	__netif_tx_lock_bh(txq);
+	local_bh_disable();
+
+	HARD_TX_LOCK(odev, txq, smp_processor_id());
 
 	if (unlikely(netif_xmit_frozen_or_drv_stopped(txq))) {
 		ret = NETDEV_TX_BUSY;
@@ -3374,7 +3376,9 @@ static void pktgen_xmit(struct pktgen_dev *pkt_dev)
 		pkt_dev->last_ok = 0;
 	}
 unlock:
-	__netif_tx_unlock_bh(txq);
+	HARD_TX_UNLOCK(odev, txq);
+
+	local_bh_enable();
 
 	/* If pkt_dev->count is zero, then run forever */
 	if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
-- 
1.7.11.7

Re: [PATCH net] pktgen: be friendly to LLTX devices

[David Miller]

From: Daniel Borkmann <dborkman@redhat.com>
Date: Fri, 11 Apr 2014 13:22:00 +0200

> Similarly to commit 43279500deca ("packet: respect devices with
> LLTX flag in direct xmit"), we can basically apply the very same
> to pktgen. This will help testing against LLTX devices such as
> dummy driver (or others), which only have a single netdevice txq
> and would otherwise require locking their txq from pktgen side
> while e.g. in dummy case, we would not need any locking. Fix this
> by making use of HARD_TX_{UN,}LOCK API, so that NETIF_F_LLTX will
> be respected.
> 
> Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
> Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>

Applied.