[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-Id: <20190320.111912.473489134725484625.davem@davemloft.net>
Date: Wed, 20 Mar 2019 11:19:12 -0700 (PDT)
From: David Miller <davem@...emloft.net>
To: pabeni@...hat.com
Cc: netdev@...r.kernel.org, mike.marciniszyn@...el.com,
dledford@...hat.com, j.vosburgh@...il.com, saeedb@...zon.com,
f.fainelli@...il.com, aelior@...vell.com, vishal@...lsio.com,
jeffrey.t.kirsher@...el.com, salil.mehta@...wei.com,
tariqt@...lanox.com, kys@...rosoft.com, kvalo@...eaurora.org,
wei.liu2@...rix.com, johannes@...solutions.net,
willemdebruijn.kernel@...il.com, eric.dumazet@...il.com
Subject: Re: [PATCH net-next v2 0/3] net: refactor ndo_select_queue()
From: Paolo Abeni <pabeni@...hat.com>
Date: Wed, 20 Mar 2019 11:02:03 +0100
> Currently, on most devices implementing ndo_select_queue(), we get 2
> indirect calls per xmit packet, at least in some scenarios.
>
> We can avoid one of such indirect calls refactoring the ndo_select_queue()
> usage so that we don't need anymore the 'fallback' argument.
>
> The first patch renames a helper used later as a public API, the second one
> changes the af packet implementation so that it uses the common infrastructure
> to select the xmit queue, and the second patch drops the now unneeded argument
> from ndo_select_queue().
>
> Alternatively we could use the INDIRECT_CALL_WRAPPER infrastructure to avoid
> the fallback indirect call in the common case, but this solution allows also
> for some code cleanup.
>
> v1 -> v2:
> - renamed select queue helpers, as per Eric's and David's suggestions
I like it, series applied.
Thanks Paolo.
Powered by blists - more mailing lists