On Thu, 2015-01-29 at 11:46 -0500, David L Stevens wrote:
quoted hunk ↗ jump to hunk
[resending this since it never appeared on netdev yesterday]
This patch fixes a bug where vnet_skb_shape() didn't set the already-selected
queue mapping when a packet copy was required. This results in using the
wrong queue index for stops/starts, hung tx queues and watchdog timeouts
under heavy load.
Signed-off-by: David L Stevens <redacted>
Acked-by: Sowmini Varadhan <redacted>
---
drivers/net/ethernet/sun/sunvnet.c | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/drivers/net/ethernet/sun/sunvnet.c b/drivers/net/ethernet/sun/sunvnet.c
index 2b719cc..2b10b85 100644
--- a/drivers/net/ethernet/sun/sunvnet.c
+++ b/drivers/net/ethernet/sun/sunvnet.c
@@ -1123,6 +1123,7 @@ static inline struct sk_buff *vnet_skb_shape(struct sk_buff *skb, int ncookies)
skb_shinfo(nskb)->gso_size = skb_shinfo(skb)->gso_size;
skb_shinfo(nskb)->gso_type = skb_shinfo(skb)->gso_type;
}
+ nskb->queue_mapping = skb->queue_mapping;
dev_kfree_skb(skb);
skb = nskb;
}
Note that vnet_skb_shape() also drops the skb -> sk assocation.
-> That removes flow control, a single socket can fill TX ring buffer.
TCP Small queue is also disabled.
I would try add before the dev_kfree_skb(skb) :
swap(nskb->sk, skb->sk);
swap(nskb->truesize, skb->truesize);