Hi On Tue, Jul 31, 2018 at 1:30 PM, Markus Armbruster <arm...@redhat.com> wrote: > Marc-André Lureau <marcandre.lur...@redhat.com> writes: > >> A chardev may stop trying to write if the associated can_read() >> callback returned 0. This happens when the monitor is suspended. >> The frontend is supposed to call qemu_chr_fe_accept_input() when it is >> ready to accept data again. >> >> Signed-off-by: Marc-André Lureau <marcandre.lur...@redhat.com> > > Impact?
I have observed the issue with a spice port, all pending commands aren't flushed. Most chardev don't use the accept_input() callback, and instead poll regularly, like char-socket.c:tcp_chr_read_poll() > > Is this to be considered for 3.0? This is not a regression, afaik, and doesn't impact common monitor chardev. So it could be delayed. But it's also a small fix for Spice chardev that shouldn't create regression, so it should be fine for 3.0. thanks > >> --- >> monitor.c | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/monitor.c b/monitor.c >> index d580c5a79c..e1a14e02cf 100644 >> --- a/monitor.c >> +++ b/monitor.c >> @@ -4412,6 +4412,7 @@ void monitor_resume(Monitor *mon) >> assert(mon->rs); >> readline_show_prompt(mon->rs); >> } >> + qemu_chr_fe_accept_input(&mon->chr); >> } >> trace_monitor_suspend(mon, -1); >> } > -- Marc-André Lureau
