On 01/18/2018 12:44 PM, Paolo Bonzini wrote:
On 18/01/2018 10:41, klim wrote:
I propose to install a separate pullhup watcher regardless of
whether the device waits for data or not. After closing the connection,
the guest has a capability to read the data within timeout.
I don't understand the timeout part.
The timeout is important, because of following situation:
client sends to the guest big bunch of data and closes his end of socket.
if we just destroy the connection on the qemu's side, the guest can't
read remaining data in channel.
(which can be much more than guest's buffer), although he would do it
later.
For this case timeout after POLLHUP is added. It is set after receiving
POLLHUP and reset after each reading.
But why would one second be enough?
One second was merely a suggestion. Maybe 10 seconds would be better.
Paolo
