* [PATCH net-next] macvlan: add a salt to mc_hash()
@ 2013-02-08 2:41 Eric Dumazet
2013-02-08 5:07 ` Maciej Żenczykowski
2013-02-08 22:57 ` David Miller
0 siblings, 2 replies; 3+ messages in thread
From: Eric Dumazet @ 2013-02-08 2:41 UTC (permalink / raw)
To: David Miller; +Cc: netdev, Maciej Żenczykowski
From: Eric Dumazet <edumazet@google.com>
Some multicast addresses are common to all macvlans,
so if a multicast message has a hash value collision, we
have to deliver a copy to all macvlans, adding significant
latency and possible packet drops if netdev_max_backlog
limit is hit.
Having a per macvlan hash function permits to reduce the
impact of hash collisions.
Suggested-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
Should be applied after
"macvlan: broadcast addr should be part of mc_filter"
drivers/net/macvlan.c | 17 +++++++++++++----
1 file changed, 13 insertions(+), 4 deletions(-)
diff --git a/drivers/net/macvlan.c b/drivers/net/macvlan.c
index ce7c926..e4b8078 100644
--- a/drivers/net/macvlan.c
+++ b/drivers/net/macvlan.c
@@ -127,10 +127,18 @@ static int macvlan_broadcast_one(struct sk_buff *skb,
return vlan->receive(skb);
}
-static unsigned int mc_hash(const unsigned char *addr)
+static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
+{
+ return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
+}
+
+
+static unsigned int mc_hash(const struct macvlan_dev *vlan,
+ const unsigned char *addr)
{
u32 val = __get_unaligned_cpu32(addr + 2);
+ val ^= macvlan_hash_mix(vlan);
return hash_32(val, MACVLAN_MC_FILTER_BITS);
}
@@ -145,7 +153,7 @@ static void macvlan_broadcast(struct sk_buff *skb,
struct sk_buff *nskb;
unsigned int i;
int err;
- unsigned int hash = mc_hash(eth->h_dest);
+ unsigned int hash;
if (skb->protocol == htons(ETH_P_PAUSE))
return;
@@ -155,6 +163,7 @@ static void macvlan_broadcast(struct sk_buff *skb,
if (vlan->dev == src || !(vlan->mode & mode))
continue;
+ hash = mc_hash(vlan, eth->h_dest);
if (!test_bit(hash, vlan->mc_filter))
continue;
nskb = skb_clone(skb, GFP_ATOMIC);
@@ -424,10 +433,10 @@ static void macvlan_set_mac_lists(struct net_device *dev)
bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
netdev_for_each_mc_addr(ha, dev) {
- __set_bit(mc_hash(ha->addr), filter);
+ __set_bit(mc_hash(vlan, ha->addr), filter);
}
- __set_bit(mc_hash(dev->broadcast), filter);
+ __set_bit(mc_hash(vlan, dev->broadcast), filter);
bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
}
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [PATCH net-next] macvlan: add a salt to mc_hash()
2013-02-08 2:41 [PATCH net-next] macvlan: add a salt to mc_hash() Eric Dumazet
@ 2013-02-08 5:07 ` Maciej Żenczykowski
2013-02-08 22:57 ` David Miller
1 sibling, 0 replies; 3+ messages in thread
From: Maciej Żenczykowski @ 2013-02-08 5:07 UTC (permalink / raw)
To: Eric Dumazet; +Cc: David Miller, netdev
Signed-off-by: Maciej Żenczykowski <maze@google.com>
On Thu, Feb 7, 2013 at 6:41 PM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> From: Eric Dumazet <edumazet@google.com>
>
> Some multicast addresses are common to all macvlans,
> so if a multicast message has a hash value collision, we
> have to deliver a copy to all macvlans, adding significant
> latency and possible packet drops if netdev_max_backlog
> limit is hit.
>
> Having a per macvlan hash function permits to reduce the
> impact of hash collisions.
>
> Suggested-by: Maciej Żenczykowski <maze@google.com>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> ---
> Should be applied after
> "macvlan: broadcast addr should be part of mc_filter"
>
> drivers/net/macvlan.c | 17 +++++++++++++----
> 1 file changed, 13 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/macvlan.c b/drivers/net/macvlan.c
> index ce7c926..e4b8078 100644
> --- a/drivers/net/macvlan.c
> +++ b/drivers/net/macvlan.c
> @@ -127,10 +127,18 @@ static int macvlan_broadcast_one(struct sk_buff *skb,
> return vlan->receive(skb);
> }
>
> -static unsigned int mc_hash(const unsigned char *addr)
> +static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
> +{
> + return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
> +}
> +
> +
> +static unsigned int mc_hash(const struct macvlan_dev *vlan,
> + const unsigned char *addr)
> {
> u32 val = __get_unaligned_cpu32(addr + 2);
>
> + val ^= macvlan_hash_mix(vlan);
> return hash_32(val, MACVLAN_MC_FILTER_BITS);
> }
>
> @@ -145,7 +153,7 @@ static void macvlan_broadcast(struct sk_buff *skb,
> struct sk_buff *nskb;
> unsigned int i;
> int err;
> - unsigned int hash = mc_hash(eth->h_dest);
> + unsigned int hash;
>
> if (skb->protocol == htons(ETH_P_PAUSE))
> return;
> @@ -155,6 +163,7 @@ static void macvlan_broadcast(struct sk_buff *skb,
> if (vlan->dev == src || !(vlan->mode & mode))
> continue;
>
> + hash = mc_hash(vlan, eth->h_dest);
> if (!test_bit(hash, vlan->mc_filter))
> continue;
> nskb = skb_clone(skb, GFP_ATOMIC);
> @@ -424,10 +433,10 @@ static void macvlan_set_mac_lists(struct net_device *dev)
>
> bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
> netdev_for_each_mc_addr(ha, dev) {
> - __set_bit(mc_hash(ha->addr), filter);
> + __set_bit(mc_hash(vlan, ha->addr), filter);
> }
>
> - __set_bit(mc_hash(dev->broadcast), filter);
> + __set_bit(mc_hash(vlan, dev->broadcast), filter);
>
> bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
> }
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH net-next] macvlan: add a salt to mc_hash()
2013-02-08 2:41 [PATCH net-next] macvlan: add a salt to mc_hash() Eric Dumazet
2013-02-08 5:07 ` Maciej Żenczykowski
@ 2013-02-08 22:57 ` David Miller
1 sibling, 0 replies; 3+ messages in thread
From: David Miller @ 2013-02-08 22:57 UTC (permalink / raw)
To: eric.dumazet; +Cc: netdev, maze
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Thu, 07 Feb 2013 18:41:02 -0800
> From: Eric Dumazet <edumazet@google.com>
>
> Some multicast addresses are common to all macvlans,
> so if a multicast message has a hash value collision, we
> have to deliver a copy to all macvlans, adding significant
> latency and possible packet drops if netdev_max_backlog
> limit is hit.
>
> Having a per macvlan hash function permits to reduce the
> impact of hash collisions.
>
> Suggested-by: Maciej Żenczykowski <maze@google.com>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
Applied.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2013-02-08 22:57 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-02-08 2:41 [PATCH net-next] macvlan: add a salt to mc_hash() Eric Dumazet
2013-02-08 5:07 ` Maciej Żenczykowski
2013-02-08 22:57 ` David Miller
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox