Re: [PATCH v3] rfs: Receive Flow Steering

[Tom Herbert <therbert@google.com>]

>>   * @mc_ttl - Multicasting TTL
>>   * @is_icsk - is this an inet_connection_sock?
>> @@ -124,6 +126,9 @@ struct inet_sock {
>>       __u16                   cmsg_flags;
>>       __be16                  inet_sport;
>>       __u16                   inet_id;
>> +#ifdef CONFIG_RPS
>> +     __u32                   rxhash;
>> +#endif
>
> I am a bit worried, because dirtying this cache line might hurt non RPS
> setups (if network interrupts are balanced to all cpus)
>
The rxhash should only be written when it changes.  So as long as
device or lower stack provide a consistent rxhash for a connection
this should be okay.

> Best place would be to put rxhash close to sk_refcnt (because we dirty
> it to get a reference on rcu sk lookups)
>
In sock_common?... I don't know if we need this in every socket yet.

> I believe we have a 32bits hole on 64bit arches for this :)
>
>
> While testint latest net-nex-2.6 on my nehalem machine, I got a crash
> (in RPS I am afraid...)
>
> I am going to correct this crash before testing RFS and let you know the
> results.

Thanks for doing that.

>
> Thanks
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>