On Wed, Jan 1, 2014 at 4:58 AM, David Miller [off-list ref] wrote:

From: David Miller <davem@davemloft.net>
Date: Tue, 31 Dec 2013 15:54:59 -0500 (EST)

quoted

From: Zhi Yong Wu <redacted>
Date: Wed,  1 Jan 2014 04:31:01 +0800

quoted

@@ -846,12 +846,16 @@ static inline void sock_rps_reset_flow_hash(__u32 hash)

 static inline void sock_rps_record_flow(const struct sock *sk)
 {
+#ifdef CONFIG_RPS
     sock_rps_record_flow_hash(sk->sk_rxhash);
+#endif
 }

Why is this necessary?  sock_rps_record_flow_hash() is provided
unconditionally, regardless of CONFIG_RPS.

I'm not applying this, it doesn't make any sense.  Either there is
no build failure, or the existing CONFIG_RPS ifdef inside of the
implementation of sock_rps_record_flow_hash() is bogus.

Ok, if you had actually provided the build failure log message I'd be
able to evaluate this patch much better.

The real issue is the lack of sk_rxhash when RPS is not enabled.

I'm going to apply this with the build failure message mentioned.

I noticed that you have added the build failure log message in your
net-next tree.
That is, i needn't do anything now, thanks.

-- 
Regards,

Zhi Yong Wu