On Tue, 5 Jun 2018 15:30:01 -0300
Eduardo Habkost <ehabk...@redhat.com> wrote:
> On Tue, Jun 05, 2018 at 04:00:43PM +0200, Igor Mammedov wrote:
> > When using --daemonize, the initial lead process will fork a child and
> > then wait to be notified that setup is complete via a pipe, before it
> > exits. When using --preconfig there is an extra call to main_loop()
> > before the notification is done from os_setup_post(). Thus the parent
> > process won't exit until the mgmt application connects to the monitor
> > and tells QEMU to leave the RUN_STATE_PRECONFIG. The mgmt application
> > won't connect to the monitor until daemonizing has completed though.
> >
> > This is a chicken and egg problem, leading to deadlock at startup.
> >
> > The only viable way to fix this is to call os_setup_post() before
> > the early main_loop_wait() call when in RUN_STATE_PRECONFIG. This has
> > the downside that any errors from this point onwards won't be handled
> > well by the mgmt application, because it will think QEMU has started
> > successfully, so not be expecting an abrupt exit. The only way to
> > deal with that is to move as much user input validation as possible
> > to before the main_loop() call. This is left as an exercise for
> > future interested developers.
> >
> > Based on:
> > From: Daniel P. Berrangé <berra...@redhat.com>
> > Subject: [PATCH v2 2/2] vl: fix use of --daemonize with --preconfig
> > Message-Id: <20180604120345.12955-3-berra...@redhat.com>
> >
> > Signed-off-by: Igor Mammedov <imamm...@redhat.com>
> > ---
> > v3:
> > - rewrite to apply on top of 1/2
> > ---
> > os-posix.c | 6 ++++++
> > vl.c | 2 +-
> > 2 files changed, 7 insertions(+), 1 deletion(-)
> >
> > diff --git a/os-posix.c b/os-posix.c
> > index 9ce6f74..ee06a8d 100644
> > --- a/os-posix.c
> > +++ b/os-posix.c
> > @@ -309,8 +309,14 @@ void os_daemonize(void)
> >
> > void os_setup_post(void)
> > {
> > + static bool os_setup_post_done = false;
> > int fd = 0;
> >
> > + if (os_setup_post_done) {
> > + return;
> > + }
> > + os_setup_post_done = true;
> > +
>
> This part is nice because it allows the os_setup_post() call in
> the second main loop to be unconditional, but:
>
> > if (daemonize) {
> > if (chdir("/")) {
> > error_report("not able to chdir to /: %s", strerror(errno));
> > diff --git a/vl.c b/vl.c
> > index fa44138..d6fa67f 100644
> > --- a/vl.c
> > +++ b/vl.c
> > @@ -1960,6 +1960,7 @@ static void main_loop(void)
> > #ifdef CONFIG_PROFILER
> > ti = profile_getclock();
> > #endif
> > + os_setup_post();
> > main_loop_wait(false);
>
> Ensuring the correctness of this os_setup_post() call depends on
> reading the whole body of main_loop_should_exit(), which is a
> complex and large function. I think this is too fragile for my
> taste.
Fragility was the reason why I moved it into main_loop(),
as os_setup_post() was already overlooked once, since
one would have to make very non-obvious connection with
libvirt requirement to call it before main_loop_wait()
This way call to os_setup_post() will not be forgotten,
and would get an attention whenever main_loop() is concerned.
> I prefer Daniel's approach where we have two
> os_setup_post()/main_loop() call sites, and the first one is
> conditional on --preconfig.
Considering we are unlikely to add one more invocation of main_loop().
I'll post here Daniel's version that applies on top of 1/2 with
a comment so we won't forget about libvirt's requirement
(not the best way to write something robust but better then nothing).
So pick whatever variant would seem the best.
> > #ifdef CONFIG_PROFILER
> > dev_time += profile_getclock() - ti;
> > @@ -4707,7 +4708,6 @@ int main(int argc, char **argv, char **envp)
> > }
> >
> > accel_setup_post(current_machine);
> > - os_setup_post();
> >
> > main_loop();
> >
> > --
> > 2.7.4
> >
>
