From mboxrd@z Thu Jan  1 00:00:00 1970
From: David Miller <davem@davemloft.net>
Subject: Re: [PATCH net] tcp: TCP Small Queues and strange attractors
Date: Tue, 14 Oct 2014 17:16:52 -0400 (EDT)
Message-ID: <20141014.171652.711587177378371980.davem@davemloft.net>
References: <1413206867.9362.100.camel@edumazet-glaptop2.roam.corp.google.com>
	<20141014.153308.2221538197341630790.davem@davemloft.net>
	<1413321018.17109.7.camel@edumazet-glaptop2.roam.corp.google.com>
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Cc: netdev@vger.kernel.org
To: eric.dumazet@gmail.com
Return-path: <netdev-owner@vger.kernel.org>
Received: from shards.monkeyblade.net ([149.20.54.216]:44847 "EHLO
	shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S932269AbaJNVQz (ORCPT
	<rfc822;netdev@vger.kernel.org>); Tue, 14 Oct 2014 17:16:55 -0400
In-Reply-To: <1413321018.17109.7.camel@edumazet-glaptop2.roam.corp.google.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Tue, 14 Oct 2014 14:10:18 -0700

> On Tue, 2014-10-14 at 15:33 -0400, David Miller wrote:
>> From: Eric Dumazet <eric.dumazet@gmail.com>
>> Date: Mon, 13 Oct 2014 06:27:47 -0700
>> 
>> > diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
>> > index 8d4eac793700..4a7e97811d71 100644
>> > --- a/net/ipv4/tcp_output.c
>> > +++ b/net/ipv4/tcp_output.c
>> > @@ -839,26 +839,38 @@ void tcp_wfree(struct sk_buff *skb)
>> >  {
>>  ...
>> >  		local_irq_restore(flags);
>> > -	} else {
>> > -		sock_wfree(skb);
>> > +		return;
>> >  	}
>> > +out:
>> > +	sk_free(sk);
>> >  }
>> >  
>> 
>> Why do we need to release the socket here?
> 
> Thats because we had to keep a reference at the very beginning :
> 
> +       /* Keep one reference on sk_wmem_alloc.
> +        * Will be released by sk_free() from here or tcp_tasklet_func()
> +        */
> +       wmem = atomic_sub_return(skb->truesize - 1, &sk->sk_wmem_alloc);
> +
> 
> 
> If we find that we do not arm the tasklet (and keep the reference), we
> then can remove the last ref we kept.
> 
> sk_free() is essentially doing :
> 
> if (atomic_dec_and_test(&sk->sk_wmem_alloc))
>     __sk_free(sk)

Gotcha, now it makes sense.

Patch applied, thanks Eric.