netdev - Re: Choose a default DSA CPU port

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

Date:   Wed, 22 Feb 2023 22:42:54 +0300
From:   Arınç ÜNAL <arinc.unal@...nc9.com>
To:     Vladimir Oltean <olteanv@...il.com>
Cc:     Frank Wunderlich <frank-w@...lic-files.de>,
        netdev <netdev@...r.kernel.org>, erkin.bozoglu@...ont.com,
        Andrew Lunn <andrew@...n.ch>,
        Florian Fainelli <f.fainelli@...il.com>,
        Felix Fietkau <nbd@....name>, John Crispin <john@...ozen.org>,
        Mark Lee <Mark-MC.Lee@...iatek.com>,
        Lorenzo Bianconi <lorenzo@...nel.org>,
        Matthias Brugger <matthias.bgg@...il.com>,
        Landen Chao <Landen.Chao@...iatek.com>,
        Sean Wang <sean.wang@...iatek.com>,
        DENG Qingfang <dqfext@...il.com>
Subject: Re: Choose a default DSA CPU port

On 22.02.2023 22:34, Vladimir Oltean wrote:
> On Wed, Feb 22, 2023 at 09:08:03PM +0300, Arınç ÜNAL wrote:
>> On 22.02.2023 21:06, Vladimir Oltean wrote:
>>> On Wed, Feb 22, 2023 at 06:17:42PM +0100, Frank Wunderlich wrote:
>>>> without Arincs Patch i got 940Mbit on gmac0, so something seems to affect the gmac when port5 is enabled.
>>>
>>> which patch?
>>
>> I believe Frank is referring to the patch series I submitted which adds
>> port@5 to Bananapi BPI-R2. Without the patch, gmac0 is the default DSA
>> master.
>>
>> https://lore.kernel.org/linux-mediatek/20230210182505.24597-1-arinc.unal@arinc9.com/
>>
>> Arınç
> 
> And with your patch + my patch, gmac0 is still the default DSA master,
> but gmac1/port5 are also enabled. The claim is that switch port 6/gmac0
> has lower bandwidth when switch port 5/gmac1 is enabled, than when it isn't?
> 
> Frank's testing is done on the MT7623 SoC (with the MT7530 switch),
> an SoC which you have access to, since you've submitted those device
> tree changes, correct? Do you confirm his result?

I just sent a big patch series, doing some tests on this issue is next 
on my task list. So I can't confirm Frank's result for now.

> 
> The posted ethtool stats are not sufficient to determine the cause of
> the issue. It would be necessary to see all non-zero Ethernet counters
> on both CPU port pairs:
> 
> ethtool -S eth0 | grep -v ': 0'
> ethtool -S eth1 | grep -v ': 0'
> 
> to determine whether the cause of the performance degradation is packet
> loss or just a lossless slowdown of some sorts. For example, the
> degradation might be caused by the added flow control + uncalibrated
> watermarks, not by the activation of the other GMAC.

I'll keep this in mind thanks.

Frank, here's my task page for this issue, for your information.

https://arinc9.notion.site/MT7530-port5-performance-issue-98ac5fa19dc248e0b12fab08dcb2e387

Arınç