Date: Mon, 8 May 2023 06:30:07 -0400
From: "Michael S. Tsirkin" <mst@...hat.com>
To: Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
Cc: huangml@...ur.tech, zy@...ur.tech, Jason Wang <jasowang@...hat.com>,
	"David S. Miller" <davem@...emloft.net>,
	Eric Dumazet <edumazet@...gle.com>,
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
	"open list:VIRTIO CORE AND NET DRIVERS" <virtualization@...ts.linux-foundation.org>,
	"open list:NETWORKING DRIVERS" <netdev@...r.kernel.org>,
	open list <linux-kernel@...r.kernel.org>,
	Hao Chen <chenh@...ur.tech>, hengqi@...ux.alibaba.com
Subject: Re: [PATCH] virtio_net: set default mtu to 1500 when 'Device maximum
 MTU' bigger than 1500

On Mon, May 08, 2023 at 03:41:56PM +0800, Xuan Zhuo wrote:
> On Mon, 8 May 2023 02:43:24 -0400, "Michael S. Tsirkin" <mst@...hat.com> wrote:
> > On Mon, May 08, 2023 at 02:18:08PM +0800, Xuan Zhuo wrote:
> > > On Mon, 8 May 2023 02:15:46 -0400, "Michael S. Tsirkin" <mst@...hat.com> wrote:
> > > > On Mon, May 08, 2023 at 10:01:59AM +0800, Xuan Zhuo wrote:
> > > > > On Sun, 7 May 2023 04:58:58 -0400, "Michael S. Tsirkin" <mst@...hat.com> wrote:
> > > > > > On Sat, May 06, 2023 at 04:56:35PM +0800, Hao Chen wrote:
> > > > > > >
> > > > > > >
> > > > > > > 在 2023/5/6 10:50, Xuan Zhuo 写道:
> > > > > > > > On Sat,  6 May 2023 10:15:29 +0800, Hao Chen <chenh@...ur.tech> wrote:
> > > > > > > > > When VIRTIO_NET_F_MTU(3) Device maximum MTU reporting is supported.
> > > > > > > > > If offered by the device, device advises driver about the value of its
> > > > > > > > > maximum MTU. If negotiated, the driver uses mtu as the maximum
> > > > > > > > > MTU value. But there the driver also uses it as default mtu,
> > > > > > > > > some devices may have a maximum MTU greater than 1500, this may
> > > > > > > > > cause some large packages to be discarded,
> > > > > > > >
> > > > > > > > You mean tx packet?
> > > > > > > Yes.
> > > > > > > >
> > > > > > > > If yes, I do not think this is the problem of driver.
> > > > > > > >
> > > > > > > > Maybe you should give more details about the discard.
> > > > > > > >
> > > > > > > In the current code, if the maximum MTU supported by the virtio net hardware
> > > > > > > is 9000, the default MTU of the virtio net driver will also be set to 9000.
> > > > > > > When sending packets through "ping -s 5000", if the peer router does not
> > > > > > > support negotiating a path MTU through ICMP packets, the packets will be
> > > > > > > discarded. If the peer router supports negotiating path mtu through ICMP
> > > > > > > packets, the host side will perform packet sharding processing based on the
> > > > > > > negotiated path mtu, which is generally within 1500.
> > > > > > > This is not a bugfix patch, I think setting the default mtu to within 1500
> > > > > > > would be more suitable here.Thanks.
> > > > > >
> > > > > > I don't think VIRTIO_NET_F_MTU is appropriate for support for jumbo packets.
> > > > > > The spec says:
> > > > > > 	The device MUST forward transmitted packets of up to mtu (plus low level ethernet header length) size with
> > > > > > 	gso_type NONE or ECN, and do so without fragmentation, after VIRTIO_NET_F_MTU has been success-
> > > > > > 	fully negotiated.
> > > > > > VIRTIO_NET_F_MTU has been designed for all kind of tunneling devices,
> > > > > > and this is why we set mtu to max by default.
> > > > > >
> > > > > > For things like jumbo frames where MTU might or might not be available,
> > > > > > a new feature would be more appropriate.
> > > > >
> > > > >
> > > > > So for jumbo frame, what is the problem?
> > > > >
> > > > > We are trying to do this. @Heng
> > > > >
> > > > > Thanks.
> > > >
> > > > It is not a problem as such. But VIRTIO_NET_F_MTU will set the
> > > > default MTU not just the maximum one, because spec seems to
> > > > say it can.
> > >
> > > I see.
> > >
> > > In the case of Jumbo Frame, we also hope that the driver will set the default
> > > directly to the max mtu. Just like what you said "Bigger packets = better
> > > performance."
> > >
> > > I don't know, in any scenario, when the hardware supports a large mtu, but we do
> > > not want the user to use it by default.
> >
> > When other devices on the same LAN have mtu set to 1500 and
> > won't accept bigger packets.
> 
> So, that depends on pmtu/tcp-probe-mtu.
> 
> If the os without pmtu/tcp-probe-mtu has a bigger mtu, then it's big packet
> will lost.
> 
> Thanks.
> 

pmtu is designed for routing. LAN is supposed to be configured with
a consistent MTU.

> >
> > > Of course, the scene that this patch
> > > wants to handle does exist, but I have never thought that this is a problem at
> > > the driver level.
> > >
> > > Thanks.
> > >
> > >
> > > >
> > > >
> > > > >
> > > > > >
> > > > > > > > > so I changed the MTU to a more
> > > > > > > > > general 1500 when 'Device maximum MTU' bigger than 1500.
> > > > > > > > >
> > > > > > > > > Signed-off-by: Hao Chen <chenh@...ur.tech>
> > > > > > > > > ---
> > > > > > > > >   drivers/net/virtio_net.c | 5 ++++-
> > > > > > > > >   1 file changed, 4 insertions(+), 1 deletion(-)
> > > > > > > > >
> > > > > > > > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > > > > > > > > index 8d8038538fc4..e71c7d1b5f29 100644
> > > > > > > > > --- a/drivers/net/virtio_net.c
> > > > > > > > > +++ b/drivers/net/virtio_net.c
> > > > > > > > > @@ -4040,7 +4040,10 @@ static int virtnet_probe(struct virtio_device *vdev)
> > > > > > > > >   			goto free;
> > > > > > > > >   		}
> > > > > > > > >
> > > > > > > > > -		dev->mtu = mtu;
> > > > > > > > > +		if (mtu > 1500)
> > > > > > > >
> > > > > > > > s/1500/ETH_DATA_LEN/
> > > > > > > >
> > > > > > > > Thanks.
> > > > > > > >
> > > > > > > > > +			dev->mtu = 1500;
> > > > > > > > > +		else
> > > > > > > > > +			dev->mtu = mtu;
> > > > > > > > >   		dev->max_mtu = mtu;
> > > > > > > > >   	}
> > > > > > > > >
> > > > > > > > > --
> > > > > > > > > 2.27.0
> > > > > > > > >
> > > > > >
> > > >
> >

Powered by blists - more mailing lists