diff options
author | Eric Dumazet <edumazet@google.com> | 2014-10-05 18:38:35 -0700 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2014-10-07 13:22:11 -0400 |
commit | 0287587884b15041203b3a362d485e1ab1f24445 (patch) | |
tree | 675ae57663c1ba3ee8768e65e7fb0e6d0259e04c /drivers/net/team | |
parent | fe971b95c22578456ff7198537827841c726d3f7 (diff) | |
download | op-kernel-dev-0287587884b15041203b3a362d485e1ab1f24445.zip op-kernel-dev-0287587884b15041203b3a362d485e1ab1f24445.tar.gz |
net: better IFF_XMIT_DST_RELEASE support
Testing xmit_more support with netperf and connected UDP sockets,
I found strange dst refcount false sharing.
Current handling of IFF_XMIT_DST_RELEASE is not optimal.
Dropping dst in validate_xmit_skb() is certainly too late in case
packet was queued by cpu X but dequeued by cpu Y
The logical point to take care of drop/force is in __dev_queue_xmit()
before even taking qdisc lock.
As Julian Anastasov pointed out, need for skb_dst() might come from some
packet schedulers or classifiers.
This patch adds new helper to cleanly express needs of various drivers
or qdiscs/classifiers.
Drivers that need skb_dst() in their ndo_start_xmit() should call
following helper in their setup instead of the prior :
dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
->
netif_keep_dst(dev);
Instead of using a single bit, we use two bits, one being
eventually rebuilt in bonding/team drivers.
The other one, is permanent and blocks IFF_XMIT_DST_RELEASE being
rebuilt in bonding/team. Eventually, we could add something
smarter later.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'drivers/net/team')
-rw-r--r-- | drivers/net/team/team.c | 8 |
1 files changed, 5 insertions, 3 deletions
diff --git a/drivers/net/team/team.c b/drivers/net/team/team.c index 2277c36..a94a9df 100644 --- a/drivers/net/team/team.c +++ b/drivers/net/team/team.c @@ -970,7 +970,8 @@ static void __team_compute_features(struct team *team) struct team_port *port; u32 vlan_features = TEAM_VLAN_FEATURES & NETIF_F_ALL_FOR_ALL; unsigned short max_hard_header_len = ETH_HLEN; - unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; + unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE | + IFF_XMIT_DST_RELEASE_PERM; list_for_each_entry(port, &team->port_list, list) { vlan_features = netdev_increment_features(vlan_features, @@ -985,8 +986,9 @@ static void __team_compute_features(struct team *team) team->dev->vlan_features = vlan_features; team->dev->hard_header_len = max_hard_header_len; - flags = team->dev->priv_flags & ~IFF_XMIT_DST_RELEASE; - team->dev->priv_flags = flags | dst_release_flag; + team->dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; + if (dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM)) + team->dev->priv_flags |= IFF_XMIT_DST_RELEASE; netdev_change_features(team->dev); } |