[Spice-devel] [spice-gtk v3 2/3] gtk-session: clipboard: x11: owner-change improvement

[Marc-André Lureau]

Hi

On Mon, Jan 14, 2019 at 3:09 PM Victor Toso <victortoso at redhat.com> wrote:
>
> Hi,
>
> On Fri, Jan 11, 2019 at 01:39:22PM +0400, Marc-André Lureau wrote:
> > Hi
> >
> > On Thu, Jan 10, 2019 at 5:40 PM Victor Toso <victortoso at redhat.com> wrote:
> > >
> > > Hi,
> > >
> > > On Thu, Jan 10, 2019 at 05:18:48PM +0400, Marc-André Lureau wrote:
> > > > On Thu, Jan 10, 2019 at 4:47 PM Victor Toso <victortoso at redhat.com> wrote:
> > > > >
> > > > > From: Victor Toso <me at victortoso.com>
> > > > >
> > > > > While interacting within the guest the VD_AGENT_CLIPBOARD_GRAB message
> > > > > will be sent by the guest agent to notify that agent is holding some
> > > > > clipboard data. When this clipboard data changes, the agent will send
> > > > > VD_AGENT_CLIPBOARD_RELEASE to simply notify that previous clipboard
> > > > > data is not available anymore and a new VD_AGENT_CLIPBOARD_GRAB
> > > > > follows up with the current clipboard data being hold by agent.
> > > > >
> > > > > This patch helps in fixing a state race, in X11, between who is
> > > > > 'holding' the clipboard grab.
> > > > >
> > > > > The bug happens because gtk_clipboard_clear() will set owner to none,
> > > > > making the rest of the code path consider that clipboard data has
> > > > > changed in the *client* and ends up requesting the metadata with
> > > > > gtk_clipboard_request_targets(), handled by clipboard_get_targets().
> > > > >
> > > > > It is possible to have clipboard_get_targets() being called between
> > > > > one VD_AGENT_CLIPBOARD_RELEASE and other VD_AGENT_CLIPBOARD_GRAB.
> > > > >
> > > > > Related: https://gitlab.freedesktop.org/spice/win32/vd_agent/issues/6
> > > > > Related: https://gitlab.freedesktop.org/spice/linux/vd_agent/issues/9
> > > > > Related: https://bugzilla.redhat.com/show_bug.cgi?id=1594876
> > > > >
> > > > > Signed-off-by: Victor Toso <victortoso at redhat.com>
> > > > > Tested-by: James Harvey @jamespharvey20
> > > > > ---
> > > > >  src/spice-gtk-session.c | 16 ++++++++++++++++
> > > > >  1 file changed, 16 insertions(+)
> > > > >
> > > > > diff --git a/src/spice-gtk-session.c b/src/spice-gtk-session.c
> > > > > index adc72a2..85d5880 100644
> > > > > --- a/src/spice-gtk-session.c
> > > > > +++ b/src/spice-gtk-session.c
> > > > > @@ -27,6 +27,9 @@
> > > > >  #include <X11/Xlib.h>
> > > > >  #include <gdk/gdkx.h>
> > > > >  #endif
> > > > > +#ifdef GDK_WINDOWING_WAYLAND
> > > > > +#include <gdk/gdkwayland.h>
> > > > > +#endif
> > > > >  #ifdef G_OS_WIN32
> > > > >  #include <windows.h>
> > > > >  #include <gdk/gdkwin32.h>
> > > > > @@ -674,6 +677,19 @@ static void clipboard_owner_change(GtkClipboard        *clipboard,
> > > > >          return;
> > > > >      }
> > > > >
> > > > > +#ifdef GDK_WINDOWING_X11
> > > > > +    /* In X11, while holding the keyboard-grab we are not interested in this
> > > > > +     * event as it either came by grab or release messages from agent.  */
> > > > > +    if (GDK_IS_X11_DISPLAY(gdk_display_get_default()) &&
> > > > > +        spice_gtk_session_get_keyboard_has_focus(self)) {
> > > > > +        SPICE_DEBUG("clipboard: owner-changed event: not requesting client's target "
> > > > > +                    "while holding focus");
> > > >
> > > > Couldn't this race happen without keyboard focus?
> > >
> > > In X11 we would receive owner-changed every time that clipboard
> > > data has changed so, if guest agent sends grab/release without
> > > user triggering them, it might happen but I don't think that
> > > would be common.
> > >
> > > This patch addresses doing gtk_clipboard_request_targets() while
> > > spice-gtk is the one changing GtkClipboard.
> >
> >
> > I thought that case was avoided with
> > gtk_clipboard_get_owner(clipboard) == G_OBJECT(self)
> >
> > But you are saying that the gtk_clipboard_clear() call ends up
> > triggering gtk_clipboard_request_targets() and that is the source of
> > races.
> >
> > Why not simply?
>
> I thought it should work too but it didn't as explained by Jakub.
>
>     https://lists.freedesktop.org/archives/spice-devel/2019-January/047142.html
>
> I don't mind ignoring 3/3 patch but some fix here would be good
> prior the release.
>
> Earlier on, I had a patch that combined 2/3 and 3/3 which was
> acked but I though later that would be better to split the change
> which is done on onwer-changed event and avoiding the
> clipboard_get() as it could be two different issues IMHO.
>
>     https://lists.freedesktop.org/archives/spice-devel/2019-January/047012.html
>
> So, ignoring the clipboard_get() issue for now, do you have any
> other idea to address this one?

I am a bit lost, can you send an updated patch with a good commit
message of what it is actually fixing?

thanks!

>
> > -    if (gtk_clipboard_get_owner(clipboard) == G_OBJECT(self)) {
> > +    GObject *owner = gtk_clipboard_get_owner(clipboard);
> > +    if (owner == NULL || gtk_clipboard_get_owner(clipboard) ==
> > G_OBJECT(self)) {
> >          return;
> >      }
>
> Cheers,
> Victor



-- 
Marc-André Lureau