Re: [Qemu-devel] [PATCH 0/4] Fix socket chardev regression

[Peter Xu <peterx@redhat.com>]

On Fri, Aug 17, 2018 at 03:52:20PM +0200, Marc-André Lureau wrote:
> Hi,
> 
> In commit 25679e5d58e "chardev: tcp: postpone async connection setup"
> (and its follow up 99f2f54174a59), Peter moved chardev socket
> connection to machine_done event. However, chardev created later will
> no longer attempt to connect, and chardev created in tests do not have
> machine_done event (breaking some of vhost-user-test).
> 
> The goal was to move the "connect" source to the chardev frontend
> context (the monitor thread context in his case). chr->gcontext is set
> with qemu_chr_fe_set_handlers(). But there is no guarantee that the
> function will be called in general,

Could you hint a case where we didn't use qemu_chr_fe_set_handlers()
upon a chardev backend?  I thought it was always used in chardev
frontends, and what the backend could do if without a frontend?

[1]

> so we can't delay connection until
> then: the chardev should still attempt to connect during open(), using
> the main context.
> 
> An alternative would be to specify the iothread during chardev
> creation. Setting up monitor OOB would be quite different too, it
> would take the same iothread as argument.
> 
> 99f2f54174a595e is also a bit problematic, since it will behave
> differently before and after machine_done (the first case gives a
> chance to use a different context reliably, the second looks racy)
> 
> In the end, I am not sure this is all necessary, as chardev callbacks
> are called after qemu_chr_fe_set_handlers(), at which point the
> context of sources are updated. In "char-socket: update all ioc
> handlers when changing context", I moved also the hup handler to the
> updated context. So unless the main thread is already stuck, we can
> setup a different context for the chardev at that time. Or not?

IMHO the two patches that you reverted are special-cases for reasons.

The TLS handshake is carried out with an TLS internal GSource which is
not owned by the chardev code, so the qemu_chr_fe_set_handlers() won't
update that GSource (please refer to qio_channel_tls_handshake_task).

The async connection is carried out in a standalone thread that calls
connect().  IMHO we'd better not update the gcontext bound to the
async task since otherwise there'll be a race (IIRC I proposed
something before using a mutex to update the gcontext, but Dan would
prefer not to, and I followed with the suggestion which makes sense to
me).

Could we just postpone these machine done tasks into
qemu_chr_fe_set_handlers() (or say, chr_update_read_handler() hook,
just like what I mentioned in the other thread)?  Though we'll be sure
qemu_chr_fe_set_handlers() will be called for all chardev backends
hence I asked question [1] above.

Regards,

-- 
Peter Xu