Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH net-next] tcp: add SNMP counter for the number of packets
 pruned from ofo queue
To:     Yafang Shao <laoar.shao@gmail.com>
Cc:     David Miller <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        netdev <netdev@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
References: <1532524010-11855-1-git-send-email-laoar.shao@gmail.com>
 <047bc12f-9eb5-e512-10c5-6334c494c77e@gmail.com>
 <CALOAHbCC=oWgzcA=VhkXq0rVLBa5_TvtbKtTqESGrgxvYTtH6w@mail.gmail.com>
 <096a8a2f-ea32-2107-b2af-c246450cc3b9@gmail.com>
 <CALOAHbASvcEdudqpit-c=RRZ8xVD+G+0jU-ynp9PsH4_mU7xjw@mail.gmail.com>
 <CALOAHbAHgEF_=w_hbd_03NOH3h9m-4W=UJ0HCYGK0BtEyq=5Tw@mail.gmail.com>
 <70428ad9-5d59-faed-4384-36190939a19d@gmail.com>
 <CALOAHbBKSPQdo9bA5VFLvR448Ag30nE2Td1f6204MPZ7Nfy7ag@mail.gmail.com>
From:   Eric Dumazet <eric.dumazet@gmail.com>
Message-ID: <a78450d6-698a-e819-89f4-ebbe03fdf971@gmail.com>
Date:   Wed, 25 Jul 2018 21:36:14 -0700
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.6.0
MIME-Version: 1.0
In-Reply-To: <CALOAHbBKSPQdo9bA5VFLvR448Ag30nE2Td1f6204MPZ7Nfy7ag@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 07/25/2018 09:31 PM, Yafang Shao wrote:
> On Thu, Jul 26, 2018 at 12:06 PM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
>>
>>
>> On 07/25/2018 06:42 PM, Yafang Shao wrote:
>>
>>>
>>> Hi Eric,
>>>
>>> LINUX_MIB_TCPOFOQUEUE, LINUX_MIB_TCPOFODROP and LINUX_MIB_TCPOFOMERGE
>>> are all for the number of SKBs, but only  LINUX_MIB_OFOPRUNED is for
>>> the event, that could lead misunderstading.
>>> So I think introducing a counter for the number of SKB pruned could be
>>> better, that could help us to track the whole behavior of ofo queue.
>>> That is why I submit this patch.
>>
>> Sure, and I said your patch had issues.
>> You mixed 'packets' and 'skbs' but apparently you do not get my point.
>>
> 
> I had noticed that I made this mistake.
> 
>> I would rather not add another SNMP counter, and refine the current one,
>> trying to track something more meaningful.
>>
>> The notion of 'skb' is internal to the kernel, and can not be mapped easily
>> to 'number of network segments' which probably is more useful for the user.
>>
>> I will send this instead :
>>
>> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
>> index d51fa358b2b196d0f9c258b24354813f2128a675..141a062abd0660c8f6d049de1dc7c7ecf7a7230d 100644
>> --- a/net/ipv4/tcp_input.c
>> +++ b/net/ipv4/tcp_input.c
>> @@ -5001,18 +5001,19 @@ static bool tcp_prune_ofo_queue(struct sock *sk)
>>  {
>>         struct tcp_sock *tp = tcp_sk(sk);
>>         struct rb_node *node, *prev;
>> +       unsigned int segs = 0;
>>         int goal;
>>
>>         if (RB_EMPTY_ROOT(&tp->out_of_order_queue))
>>                 return false;
>>
>> -       NET_INC_STATS(sock_net(sk), LINUX_MIB_OFOPRUNED);
>>         goal = sk->sk_rcvbuf >> 3;
>>         node = &tp->ooo_last_skb->rbnode;
>>         do {
>>                 prev = rb_prev(node);
>>                 rb_erase(node, &tp->out_of_order_queue);
>>                 goal -= rb_to_skb(node)->truesize;
>> +               segs += max_t(u16, 1, skb_shinfo(rb_to_skb(node))->gso_segs);
>>                 tcp_drop(sk, rb_to_skb(node));
>>                 if (!prev || goal <= 0) {
>>                         sk_mem_reclaim(sk);
>> @@ -5023,6 +5024,7 @@ static bool tcp_prune_ofo_queue(struct sock *sk)
>>                 }
>>                 node = prev;
>>         } while (node);
>> +       NET_ADD_STATS(sock_net(sk), LINUX_MIB_OFOPRUNED, segs);
>>         tp->ooo_last_skb = rb_to_skb(prev);
>>
>>         /* Reset SACK state.  A conforming SACK implementation will
>>
> 
> You patch will make it more meaningful.
> How about the other ones?

Same thing really.

Note that tcp_collapse() does not currently propagate skb_shinfo(skb)->gso_segs
when rebuilding the skbs, so this will need a fix.