netdev - Re: [PATCH net-next v2 2/5] virtio-net: transmit napi

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <a23e1f04-7f2a-9b9f-3d8d-8d5d2fb8e18d@redhat.com>
Date:   Tue, 25 Apr 2017 16:39:03 +0800
From:   Jason Wang <jasowang@...hat.com>
To:     "Michael S. Tsirkin" <mst@...hat.com>,
        Willem de Bruijn <willemdebruijn.kernel@...il.com>
Cc:     Network Development <netdev@...r.kernel.org>,
        virtualization@...ts.linux-foundation.org,
        David Miller <davem@...emloft.net>,
        Willem de Bruijn <willemb@...gle.com>
Subject: Re: [PATCH net-next v2 2/5] virtio-net: transmit napi



On 2017年04月25日 00:40, Michael S. Tsirkin wrote:
> On Fri, Apr 21, 2017 at 10:50:12AM -0400, Willem de Bruijn wrote:
>>>>> Maybe I was wrong, but according to Michael's comment it looks like he
>>>>> want
>>>>> check affinity_hint_set just for speculative tx polling on rx napi
>>>>> instead
>>>>> of disabling it at all.
>>>>>
>>>>> And I'm not convinced this is really needed, driver only provide affinity
>>>>> hint instead of affinity, so it's not guaranteed that tx and rx interrupt
>>>>> are in the same vcpus.
>>>> You're right. I made the restriction broader than the request, to really
>>>> err
>>>> on the side of caution for the initial merge of napi tx. And enabling
>>>> the optimization is always a win over keeping it off, even without irq
>>>> affinity.
>>>>
>>>> The cycle cost is significant without affinity regardless of whether the
>>>> optimization is used.
>>>
>>> Yes, I noticed this in the past too.
>>>
>>>> Though this is not limited to napi-tx, it is more
>>>> pronounced in that mode than without napi.
>>>>
>>>> 1x TCP_RR for affinity configuration {process, rx_irq, tx_irq}:
>>>>
>>>> upstream:
>>>>
>>>> 1,1,1: 28985 Mbps, 278 Gcyc
>>>> 1,0,2: 30067 Mbps, 402 Gcyc
>>>>
>>>> napi tx:
>>>>
>>>> 1,1,1: 34492 Mbps, 269 Gcyc
>>>> 1,0,2: 36527 Mbps, 537 Gcyc (!)
>>>> 1,0,1: 36269 Mbps, 394 Gcyc
>>>> 1,0,0: 34674 Mbps, 402 Gcyc
>>>>
>>>> This is a particularly strong example. It is also representative
>>>> of most RR tests. It is less pronounced in other streaming tests.
>>>> 10x TCP_RR, for instance:
>>>>
>>>> upstream:
>>>>
>>>> 1,1,1: 42267 Mbps, 301 Gcyc
>>>> 1,0,2: 40663 Mbps, 445 Gcyc
>>>>
>>>> napi tx:
>>>>
>>>> 1,1,1: 42420 Mbps, 303 Gcyc
>>>> 1,0,2:  42267 Mbps, 431 Gcyc
>>>>
>>>> These numbers were obtained with the virtqueue_enable_cb_delayed
>>>> optimization after xmit_skb, btw. It turns out that moving that before
>>>> increases 1x TCP_RR further to ~39 Gbps, at the cost of reducing
>>>> 100x TCP_RR a bit.
>>>
>>> I see, so I think we can leave the affinity hint optimization/check for
>>> future investigation:
>>>
>>> - to avoid endless optimization (e.g we may want to share a single
>>> vector/napi for tx/rx queue pairs in the future) for this series.
>>> - tx napi is disabled by default which means we can do optimization on top.
>> Okay. I'll drop the vi->affinity_hint_set from the patch set for now.
> I kind of like it, let's be conservative. But I'd prefer a comment
> near it explaining why it's there.
>

Another issue for affinity_hint_set is that it could be changed when 
setting channels. I think we've already conservative enough (e.g it was 
disabled by default).

Thanks