[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <1459315764.2953.83.camel@intel.com>
Date: Tue, 29 Mar 2016 22:29:24 -0700
From: Jeff Kirsher <jeffrey.t.kirsher@...el.com>
To: Alexander Duyck <alexander.duyck@...il.com>,
Jesse Brandeburg <jesse.brandeburg@...il.com>
Cc: Dave Miller <davem@...emloft.net>,
Alexander Duyck <aduyck@...antis.com>,
NetDEV list <netdev@...r.kernel.org>,
Neil Horman <nhorman@...hat.com>, sassmann@...hat.com,
John Greene <jogreene@...hat.com>,
Jesse Brandeburg <jesse.brandeburg@...el.com>
Subject: Re: [net-next 02/16] i40e/i40evf: Rewrite logic for 8 descriptor
per packet check
On Tue, 2016-03-29 at 22:19 -0700, Alexander Duyck wrote:
> Minor correction here. This is 72 hex, which is 114 in decimal. The
> interesting bit I believe is the fact that we have both header and
> payload data in the same descriptor.
>
> You should probably check with your hardware team on this but I have
> a
> working theory on the issue. I'm thinking that because both header
> and payload are coming out of the same descriptor this must count as
> 2
> descriptors and not 1 when we compute the usage for the first
> descriptor. As such I do probably need to start testing at frag 0
> because the first payload can actually come out of the header region
> if the first descriptor includes both payload and data.
>
> I'll try to submit a patch for it tonight. If you can test it
> tomorrow I would appreciate it. Otherwise I will try setting up an
> environment to try and reproduce the issue tomorrow.
If you get it submitted tonight, I can have it ready in my tree for
testing for Jesse and the team when they get into the office tomorrow.
Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)
Powered by blists - more mailing lists