From: Haiyang Zhang
Date: Wed, 16 Dec 2015 19:20:44 +
> In our test, we have bisected and found the following patch introduced big
> overhead into skb_flow_dissect_flow_keys(), and caused performance
> regression:
> commit: d34af823
> net: Add VLAN ID to flow_keys
NEVER _EVER_ work around
On Wed, 2015-12-16 at 19:20 +, Haiyang Zhang wrote:
> > -Original Message-
> > From: Eric Dumazet [mailto:eric.duma...@gmail.com]
> > Sent: Wednesday, December 16, 2015 12:08 PM
> >
> > This looks very very wrong to me.
> >
> > How many times this is called per second, for the 'one fl
> -Original Message-
> From: Eric Dumazet [mailto:eric.duma...@gmail.com]
> Sent: Wednesday, December 16, 2015 12:08 PM
>
> This looks very very wrong to me.
>
> How many times this is called per second, for the 'one flow' case ?
>
> Don't you use TSO in this driver ?
>
> What about enc
On 12/16/2015 09:34 PM, Sergei Shtylyov wrote:
To avoid performance overhead when using skb_flow_dissect_flow_keys(),
we switch to the simple parsers to get the IP and port numbers.
Performance comparison: throughput (Gbps):
Number of connections, before patch, after patch
18.56
Hello.
On 12/16/2015 09:03 PM, Haiyang Zhang wrote:
To avoid performance overhead when using skb_flow_dissect_flow_keys(),
we switch to the simple parsers to get the IP and port numbers.
Performance comparison: throughput (Gbps):
Number of connections, before patch, after patch
1
On Wed, 2015-12-16 at 10:03 -0800, Haiyang Zhang wrote:
> To avoid performance overhead when using skb_flow_dissect_flow_keys(),
> we switch to the simple parsers to get the IP and port numbers.
>
> Performance comparison: throughput (Gbps):
> Number of connections, before patch, after patch
> 1
