On Wed, Jan 04, 2023 at 06:45:25PM +0100, Stefano Brivio wrote:
> On Fri,  9 Dec 2022 16:42:27 +1100
> David Gibson <david@gibson.dropbear.id.au> wrote:
> 
> > To send frames on the tap interface, the UDP uses a fairly complicated two
> > level batching.  First multiple frames are gathered into a single "message"
> > for the qemu stream socket, then multiple messages are send with
> > sendmmsg().  We now have tap_send_frames() which already deals with sending
> > a number of frames, including batching and handling partial sends.  Use
> > that to considerably simplify things.
> > 
> > This does make a couple of behavioural changes:
> >   * We no longer split messages to keep them under 64kiB, which comments
> >     say is necessary.  However the TCP code didn't have equivalent code, so
> >     either this isn't actually needed, or we should implement for both
> >     (which is now easier since it can be done in one place).
> 
> That used to be SHRT_MAX (32 KiB - 1), not 64 KiB. For some reason, if
> I sent messages with multiple frames, with any one of them larger than
> 32 KiB, qemu used to close the connection.
> 
> But it looks like this was caused by some issue in the handling of short
> sends rather than an issue in qemu, especially given that, with this
> series applied, I don't hit the original issue (and I could predictably
> reproduce it earlier).

Right, I already updated the commit messages and some comments based
on the discussion we had last night.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson