netdev - Re: [net-next PATCH v3 2/3] e1000: add initial XDP support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20160913175926.GA38737@ast-mbp.thefacebook.com>
Date:   Tue, 13 Sep 2016 10:59:28 -0700
From:   Alexei Starovoitov <alexei.starovoitov@...il.com>
To:     Eric Dumazet <eric.dumazet@...il.com>
Cc:     Tom Herbert <tom@...bertland.com>,
        John Fastabend <john.fastabend@...il.com>,
        Brenden Blanco <bblanco@...mgrid.com>,
        Jeff Kirsher <jeffrey.t.kirsher@...el.com>,
        Jesper Dangaard Brouer <brouer@...hat.com>,
        "David S. Miller" <davem@...emloft.net>,
        Cong Wang <xiyou.wangcong@...il.com>,
        intel-wired-lan <intel-wired-lan@...ts.osuosl.org>,
        William Tu <u9012063@...il.com>,
        Linux Kernel Network Developers <netdev@...r.kernel.org>
Subject: Re: [net-next PATCH v3 2/3] e1000: add initial XDP support

On Tue, Sep 13, 2016 at 10:37:32AM -0700, Eric Dumazet wrote:
> On Tue, 2016-09-13 at 10:13 -0700, Alexei Starovoitov wrote:
> 
> > I'm afraid the point 'only for debugging' still didn't make it across.
> > xdp+e1k is for development (and debugging) of xdp-type of bpf
> > programs and _not_ for debugging of xdp itself, kernel or anything else.
> > The e1k provided interfaces and behavior needs to match exactly
> > what real hw nics (like mlx4, mlx5, igxbe, i40e) will do.
> > Doing special hacks are not acceptable. Therefore your
> > 'proposed fix' misses the mark, since:
> > 1. ignoring bql/qdisc is not a bug, but the requirement
> > 2. such 'fix' goes against the goal above since behaviors will be
> > different and xdp developer won't be able to build something like
> > xdp loadbalancer in the kvm.
> > 
> 
> 
> Is e1k the only way a VM can receive and send packets ?
> 
> Instead of adding more cruft to a legacy driver, risking breaking real
> old machines, 

agree that it is the concern.

> I am sure we can find modern alternative.

I've looked through qemu and it appears only emulate e1k and tg3.
The latter is still used in the field, so the risk of touching
it is higher.
The other alternative is virtio, but it doesn't have dma and/or pages,
so it looks to me even messier hack.
The last alternative considered was to invent xdp-only fake 'hw' nic,
but it's too much work to get it into qemu then ask the world
to upgrade qemu.
At that point I ran out of ideas and settled on hacking e1k :(
Not proud of this hack at all.