netdev - Re: [PATCH net v2] lan743x: fix rx_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: <f462a419-aa6f-38f7-573e-c48c2a859f55@gmail.com>
Date:   Fri, 11 Dec 2020 13:50:12 +0100
From:   Heiner Kallweit <hkallweit1@...il.com>
To:     Sven Van Asbroeck <thesven73@...il.com>
Cc:     Bryan Whitehead <bryan.whitehead@...rochip.com>,
        Microchip Linux Driver Support <UNGLinuxDriver@...rochip.com>,
        David S Miller <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>, Andrew Lunn <andrew@...n.ch>,
        netdev <netdev@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH net v2] lan743x: fix rx_napi_poll/interrupt ping-pong

Am 11.12.2020 um 13:43 schrieb Sven Van Asbroeck:
> Hi Heiner,
> 
> On Thu, Dec 10, 2020 at 2:32 AM Heiner Kallweit <hkallweit1@...il.com> wrote:
>>
>>
>> In addition you could play with sysfs attributes
>> /sys/class/net/<if>/gro_flush_timeout
>> /sys/class/net/<if>/napi_defer_hard_irqs
> 
> Interesting, I will look into that.
> 
I run a 1Gbit chip with gro_flush_timeout = 20000 and napi_defer_hard_irqs = 1.
This helped to reduce interrupt load significantly under iperf3
(w/o interrupt coalescing at chip level)

>>> @@ -2407,7 +2409,7 @@ static int lan743x_rx_open(struct lan743x_rx *rx)
>>>
>>>       netif_napi_add(adapter->netdev,
>>>                      &rx->napi, lan743x_rx_napi_poll,
>>> -                    rx->ring_size - 1);
>>> +                    64);
>>
>> This value isn't completely arbitrary.
>> Better use constant NAPI_POLL_WEIGHT.
>>
> 
> Thank you, I will change it in the next patch version.
>