netdev - Re: Setting RPS affinities from network driver

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]

Date:	Tue, 24 Feb 2015 10:08:47 -0800
From:	Tom Herbert <therbert@...gle.com>
To:	Sunil Kovvuri <sunil.kovvuri@...il.com>
Cc:	David Miller <davem@...emloft.net>,
	Jonathon Reinhart <jonathon.reinhart@...il.com>,
	Linux Netdev List <netdev@...r.kernel.org>
Subject: Re: Setting RPS affinities from network driver

On Tue, Feb 24, 2015 at 2:12 AM, Sunil Kovvuri <sunil.kovvuri@...il.com> wrote:
> Thanks for the valuable information.
>
> Just to give more info on why i want think it is better to have this
> config from driver,
> - The SOC i am working on has multiple cores and has on-board ethernet
> interface which supports upto 40Gbps.
> - It is highly difficult to achieve any meaningful performance with
> default RPS config.
> - So currently i am setting RPS config such that the CPU which takes
> high amount of interrupts/packets from receive queue doesn't do the
> processing as well.
> - This is improving network performance with single flow.
>
> So thought if its possible to do this from driver itself instead
> relying on user space scripts. As you guys have pointed i will try to
> implement a generic API
> instead of exporting 'rps_needed'. Will try to submit this very soon.
>
Please consider the second part of David's comment, it would be better
to have this sort of management be generic in kernel and driver only
takes commands. The RPS flow limit work is already a good example of
this.

Tom

> Thanks,
> Sunil.
>
>
> On Tue, Feb 24, 2015 at 8:44 AM, David Miller <davem@...emloft.net> wrote:
>> From: Tom Herbert <therbert@...gle.com>
>> Date: Fri, 20 Feb 2015 15:05:11 -0800
>>
>>>> Note that this argument is different from RSS where we're dealing with
>>>> actual hardware queues, so the driver of course has a say in the
>>>> configuration.
>>>
>>> Assuming that all queues are equal and we have a standard way to
>>> influence the indirection table, even RSS configuration really isn't
>>> driver specific configuration. We just need to know how many queues
>>> are available.
>>
>> Agreed.
>>
>> If drivers start doing this, fine, but they must do it with a common
>> piece of infrastructure that each and every driver can (easily) plug
>> into and make use of.
>>
>> Even better would be something that was so generic that the driver's
>> don't actualy implement any part of it other than responding to what
>> the generic networking core asks it to do.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html