netdev - Re: [PATCH net-next v2] ethtool: add netlink based get rxfh support

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: <9af9586f-b579-95f7-3626-f378da792261@intel.com>
Date:   Mon, 14 Nov 2022 19:46:13 -0600
From:   "Samudrala, Sridhar" <sridhar.samudrala@...el.com>
To:     Jakub Kicinski <kuba@...nel.org>
CC:     "Mogilappagari, Sudheer" <sudheer.mogilappagari@...el.com>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "mkubecek@...e.cz" <mkubecek@...e.cz>,
        "andrew@...n.ch" <andrew@...n.ch>,
        "corbet@....net" <corbet@....net>,
        "Nguyen, Anthony L" <anthony.l.nguyen@...el.com>
Subject: Re: [PATCH net-next v2] ethtool: add netlink based get rxfh support

On 11/14/2022 11:15 AM, Jakub Kicinski wrote:
> On Sun, 13 Nov 2022 22:23:25 -0600 Samudrala, Sridhar wrote:
>>> My worry is that if we go with a more broad name like
>>> "queue group" someone may be mislead to adding controls
>>> unrelated to flow <> queue assignment here.
>> Later we would like to add a per queue-group parameter that would allow
>> reducing/changing the number of napi pollers for a queue group from the default
>> value equal to the number of queues in the queue group. Are you suggesting
>> creating a queue-group object and use devlink API to configure such parameters
>> for a queue-group?
> I was thinking devlink because of scheduler/QoS and resource control.
> For NAPI config not so sure, but either way RSS will not be a place
> for NAPI/IRQ config.

Another option we could go with is via sysfs by exposing
   /sys/class/net/<iface>/queue_groups/qg<0-n>/num_pollers

With this option, it may be possible to also consider making per-netdev
napi parameters like threaded and napi_defer_hard_irqs etc to be
per-queue-group parameters.