Re: [PATCH] virtio-net: introduce TSO limit feature

On Wed, Oct 15, 2025 at 3:02 PM Michael S. Tsirkin <[email protected]> wrote:
>
> On Wed, Oct 15, 2025 at 12:29:13PM +0800, Jason Wang wrote:
> > On Tue, Oct 14, 2025 at 4:59 PM Michael S. Tsirkin <[email protected]> wrote:
> > >
> > > thanks for the patch! yet something to improve.
> > > I note issues only when encountered 1st time but please
> > > do go and check all patch for each issue.
> > >
> > >
> > > On Tue, Oct 14, 2025 at 12:22:43PM +0800, Jason Wang wrote:
> > > > This patch introduces TSO limit feature which allows the device to
> > > > advertise:
> > > >
> > > > - Maximum TCP length of a TSO packet or inner TSO packet when UDP
> > > >   tunnel is support
> > > > - Maximum number of segment that can be produced by the device after
> > > >   segmentation of TSO or inner TSO packet of a UDP tunnel
> > > >
> > > > This is a must to implement TCP jumbogram, as networking stack needs
> > > > to know the limitation of the device in order to produce TSO packet as
> > > > large as possible.
>
>
> Maybe you can provide an overview of how this all is used?

Sure, this feature does not implement BIG TCP (jumbogram) itself. BIG
TCP requires a little bit more to be implemented:

1) new gso_type for jumbogram
2) when TCP length exceeds 64K, we should mandate ip->tot_len to be
zero and device can judge from the length of the buffer
3) other stuffs

I guess this is something that Viet wants to work on.

But before BIG TCP, the driver needs to know the device limitation of
TSO packets. That is what this patch did. BIG TCP is not the only
user, it could be used in the software datapath as well. Consider a
simple datapath:

virtio-net -> TAP -> bridge -> eth0

If eth0's tso_max_size is less than 64K, we need to advertise this to
virtio-net otherwise 64K gso packets will be segmented by software run
xmit in eth0. We've encountered this in collaboration with mana and
virtio-net/vhost.

> I am looking at tcp_fragment and I just do not see
> where it looks at any device limits.
>

It works like a device advertising tso_max_size/segs, and it is capped
by gso_max_size/segs. The gso_max_size/segs were the ones that are
used by the stack to determine the skb->len.

Thanks

> --
> MST
>