From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <netdev-owner@vger.kernel.org>
Received: from mx3-rdu2.redhat.com ([66.187.233.73]:38730 "EHLO mx1.redhat.com"
        rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP
        id S1754923AbeCSKLE (ORCPT <rfc822;netdev@vger.kernel.org>);
        Mon, 19 Mar 2018 06:11:04 -0400
Date: Mon, 19 Mar 2018 11:10:57 +0100
From: Jesper Dangaard Brouer <brouer@redhat.com>
To: Jason Wang <jasowang@redhat.com>
Cc: netdev@vger.kernel.org,
        =?UTF-8?B?QmrDtnJuVMO2cGVs?= <bjorn.topel@intel.com>,
        magnus.karlsson@intel.com, eugenia@mellanox.com,
        John Fastabend <john.fastabend@gmail.com>,
        Eran Ben Elisha <eranbe@mellanox.com>,
        Saeed Mahameed <saeedm@mellanox.com>, galp@mellanox.com,
        Daniel Borkmann <borkmann@iogearbox.net>,
        Alexei Starovoitov <alexei.starovoitov@gmail.com>,
        Tariq Toukan <tariqt@mellanox.com>, brouer@redhat.com
Subject: Re: [bpf-next V3 PATCH 00/15] XDP redirect memory return API
Message-ID: <20180319111057.12d3ee71@redhat.com>
In-Reply-To: <0cb14140-5632-8b07-9088-2adb7dfedc0b@redhat.com>
References: <152062887576.27458.8590966896888512270.stgit@firesoul>
        <0cb14140-5632-8b07-9088-2adb7dfedc0b@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On Fri, 16 Mar 2018 17:04:17 +0800
Jason Wang <jasowang@redhat.com> wrote:

> Looks like the series forget to register memory model for tun and 
> virtio-net.

Well, no.  It is actually not strictly necessary to invoke
xdp_rxq_info_reg_mem_model() because enum MEM_TYPE_PAGE_SHARED == 0.
And if not passing an allocator pointer to the call, then an mem_id is
not registered... and __xdp_return_frame() skips the rhashtable_lookup.

I designed the API this way, because I want to support later adding an
allocator even for the refcnt scheme MEM_TYPE_PAGE_SHARED.  (As it
would be a performance optimization to return the pages to the
originating RX-CPU, and move the page refcnt dec back to that orig CPU).

I did add an xdp_rxq_info_reg_mem_model() call to ixgbe, for human
programmer "documentation" even-though it isn't strickly necessary.  I
guess, I could add similar calls to tun and virtio_net, and then we
avoid any implicit assumptions. And makes it more clear that
XDP_REDIRECT support use the memory model return API.

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer