[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <dcdnyuvjksvebfgcavogszlcoro3gwinzc6fzfjjtijadyg3km@7spc2j4v2ci6>
Date: Thu, 3 Oct 2024 13:09:47 +0200
From: Michal Kubecek <mkubecek@...e.cz>
To: Vladimir Oltean <vladimir.oltean@....com>
Cc: Jacob Keller <jacob.e.keller@...el.com>,
"Mogilappagari, Sudheer" <sudheer.mogilappagari@...el.com>, "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
Jakub Kicinski <kuba@...nel.org>
Subject: Re: [PATCH ethtool] netlink: rss: retrieve ring count using
ETHTOOL_GRXRINGS ioctl
On Thu, Oct 03, 2024 at 12:18:10PM +0300, Vladimir Oltean wrote:
> Hello Jake,
>
> On Thu, Sep 19, 2024 at 10:58:48AM -0700, Jacob Keller wrote:
> > We could extend the netlink interface to add this over netlink instead
> > of ioctl, but that hasn't been done yet.
>
> Let me understand how this will play out. So you're proposing that we
> add a new netlink attribute in ETHTOOL_MSG_RSS_GET_REPLY, with Fixes:
> 7112a04664bf ("ethtool: add netlink based get rss support"), and this
> gets backported to all stable kernels which include this commit?
>
> And then also 'fix' ethtool to parse this new netlink attribute instead
> of ETHTOOL_A_CHANNELS_RX_COUNT? Thus introducing another breakage: when
> the new ethtool program runs with the old (aka unpatched stable) kernel?
I'm afraid we will have to keep the unfortunate ioctl fallback for quite
long. The only other option would be to only use netlink for RSS against
kernel which provides full information and use only ioctl against those
which don't.
Michal
Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)
Powered by blists - more mailing lists