[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3fbe534d-b816-4d0b-b1c4-de63a02aba0f@redhat.com>
Date: Tue, 30 Jul 2024 18:01:58 +0200
From: Paolo Abeni <pabeni@...hat.com>
To: Jiri Pirko <jiri@...nulli.us>
Cc: netdev@...r.kernel.org, Jakub Kicinski <kuba@...nel.org>,
Madhu Chittim <madhu.chittim@...el.com>,
Sridhar Samudrala <sridhar.samudrala@...el.com>,
Simon Horman <horms@...nel.org>, John Fastabend <john.fastabend@...il.com>,
Sunil Kovvuri Goutham <sgoutham@...vell.com>,
Jamal Hadi Salim <jhs@...atatu.com>
Subject: Re: [PATCH RFC v2 00/11] net: introduce TX shaping H/W offload API
On 7/29/24 17:13, Jiri Pirko wrote:
> Mon, Jul 29, 2024 at 04:42:19PM CEST, pabeni@...hat.com wrote:
>> The general idea is, with time, to leverage this API to replace others H/W
>> shaping related in-kernel interfaces.
>>
>> At least ndo_set_tx_maxrate() should be quite straight-forward, after that
>> the relevant device drivers have implemented (very limited) support for this
>> API.
>
> Could you try to draft at least one example per each user? I mean, this
> is likely to be the tricky part of this work, would be great to make
> that click from very beginning.
I think we need to clarify that we are not going to replace all the
existing in-kernel interfaces that somewhat configure shapers on the
network devices.
Trying to achieve the above would be a (not so) nice way to block this
effort forever.
ndo_set_tx_maxrate() is IMHO a good, feasible example. Others could be
ieee_setmaxrate() or ndo_set_vf_rate() - ignoring the “obsoleted” argument.
>> The latter will need some effort from the drivers' owners.
>
> Let me know what you need exactly. Will try to do my best to help.
Something like what is implemented in this series for the iavf driver
would suffice.
Thanks,
Paolo
Powered by blists - more mailing lists