On 03/02/2017 06:16 PM, Daniel P. Berrange wrote:
On Thu, Mar 02, 2017 at 11:10:10AM -0700, Jim Fehlig wrote:
> On 02/08/2017 09:44 AM, Joao Martins wrote:
>> Signed-off-by: Joao Martins <joao.m.martins(a)oracle.com>
>
> Since it wasn't needed before, I think it is helpful for the commit message
> to describe why the change is needed now.
I'm open to being convinced otherwise, but I'd say that if the guest
is echo'ing back commands the host sends, then the guest is broken
and we shouldn't workaround that in libvirt. Or perhaps something
about the hypervisor data channel over which the guest agent is
running is running.
Seeing data echo'd back suggests use of a PTY which has not been
put into raw mode, or something like that.
OK, Thanks for the pointers.
This is one of the biggest reasons this is an RFC, as I haven't investigated
this properly yet. My gut feeling was that this could be on the device model
side (i.e. qemu) or even xenconsole handling side - and this was happening on
both socket and pty channel types. Anyhow, I can't make the argument/case for
inclusion of this patch yet, but being a one-liner I thought of including it for
purposes of a working RFC/prototype.
Joao