Re: [virtio-comment] Re: [EXT] Re: [virtio-comment] Re: [PATCH] virtio-net: Add equal-sized receive buffers feature

["Michael S. Tsirkin" <mst@redhat.com>]

On Mon, Nov 25, 2019 at 04:22:47PM +0800, Jason Wang wrote:
> 
> On 2019/11/25 äå4:18, Michael S. Tsirkin wrote:
> > > > > > Again I can see how we might want to disallow crazy setups with e.g. 1
> > > > > > byte per buffer. That's just abuse, no guest does that anyway. So asking
> > > > > > e.g. for a minimal buffer size sounds very reasonable.
> > > > > One question here is that, the minimal buffer size should depends on various
> > > > > factors. E.g the ring size. Consider a 256 entries ring, the minimal size
> > > > > should be 64K/256=256 ...
> > > > I guess you are right. We can make this driver programmable I guess?
> > > > Basically pass min_buf_len to the device.
> > > Then it still have the chance to program the min_buf_len to 1?
> > If it wants to - yes - presumably device won't work as well,
> > fetching all kind of extra data e.g. unnecessary descriptors.
> > 
> 
> Technically, device should still work for small packet and drop large
> packet.

I am not sure I agree. How is this different from just not being
sufficient space in the RX ring? We certainly do not want to
drop packets just because there's not enough buffers -
it's a temporary condition.


> Vhost has a bug that will stall RX in the case.

I think we want to spec this first then.

> Will post a fix.
> 
> Thanks