[Spice-devel] [PATCH 17/26] server/red_channel: add red_channel_{, no_}item_being_sent

Alon Levy alevy at redhat.com
Sat Feb 19 10:48:36 PST 2011


On Tue, Feb 15, 2011 at 02:29:23AM +0100, Marc-André Lureau wrote:
> Ok, I am not sure if the imposed synchronization between the channels
> is the right choice, but perhaps this is only a first approach.

This patch doesn't introduce/change the behavior, it just uses some common functions
for both display and cursor channels. The idea was to hide the access to the implementation
(send_data.item) so it can be later easily changed (i.e. I experimented with various pipe
schemes, having a single pipe for all clients, having a pipe per client (always per channel
too of course)).

> 
> On Fri, Feb 11, 2011 at 6:49 PM, Alon Levy <alevy at redhat.com> wrote:
> > ---
> >  server/red_channel.c |   10 ++++++++++
> >  server/red_channel.h |    8 ++++++++
> >  server/red_worker.c  |   10 +++++-----
> >  3 files changed, 23 insertions(+), 5 deletions(-)
> >
> > diff --git a/server/red_channel.c b/server/red_channel.c
> > index 061c9c3..f570960 100644
> > --- a/server/red_channel.c
> > +++ b/server/red_channel.c
> > @@ -727,3 +727,13 @@ int red_channel_send_message_pending(RedChannel *channel)
> >     return channel->send_data.header->type != 0;
> >  }
> >
> > +int red_channel_item_being_sent(RedChannel *channel, PipeItem *item)
> > +{
> > +    return channel->send_data.item == item;
> > +}
> > +
> > +int red_channel_no_item_being_sent(RedChannel *channel)
> > +{
> > +    return channel->send_data.item == NULL;
> > +}
> > +
> > diff --git a/server/red_channel.h b/server/red_channel.h
> > index 5c776fe..02f81fc 100644
> > --- a/server/red_channel.h
> > +++ b/server/red_channel.h
> > @@ -262,6 +262,14 @@ int red_channel_any_blocked(RedChannel *channel);
> >  /* helper for channels that have complex logic that can possibly ready a send */
> >  int red_channel_send_message_pending(RedChannel *channel);
> >
> > +/* returns TRUE if item is being sent by one of the channel clients. This will
> > + * be true if someone called init_send_data but send has not completed (or perhaps
> > + * hasn't even begun, i.e. no one called begin_send_)
> > + * */
> > +int red_channel_item_being_sent(RedChannel *channel, PipeItem *item);
> > +
> > +int red_channel_no_item_being_sent(RedChannel *channel);
> > +
> >  // TODO: unstaticed for display/cursor channels. they do some specific pushes not through
> >  // adding elements or on events. but not sure if this is actually required (only result
> >  // should be that they ""try"" a little harder, but if the event system is correct it
> > diff --git a/server/red_worker.c b/server/red_worker.c
> > index e5117cd..89fede2 100644
> > --- a/server/red_worker.c
> > +++ b/server/red_worker.c
> > @@ -9438,7 +9438,7 @@ static void red_wait_pipe_item_sent(RedChannel *channel, PipeItem *item)
> >         red_printf("timeout");
> >         channel->disconnect(channel);
> >     } else {
> > -        if (channel->send_data.item == item) {
> > +        if (red_channel_item_being_sent(channel, item)) {
> >             red_wait_outgoing_item(channel);
> >         }
> >     }
> > @@ -9524,7 +9524,7 @@ static inline void destroy_surface_wait(RedWorker *worker, int surface_id)
> >     // there is one during sending.
> >     red_wait_outgoing_item((RedChannel *)worker->display_channel);
> >     if (worker->display_channel) {
> > -        ASSERT(!worker->display_channel->common.base.send_data.item);
> > +        ASSERT(red_channel_no_item_being_sent(&worker->display_channel->common.base));
> >     }
> >  }
> >
> > @@ -9572,7 +9572,7 @@ static inline void handle_dev_destroy_surfaces(RedWorker *worker)
> >         if (!worker->cursor_channel->common.base.migrate) {
> >             red_pipe_add_verb(&worker->cursor_channel->common.base, SPICE_MSG_CURSOR_RESET);
> >         }
> > -        ASSERT(!worker->cursor_channel->common.base.send_data.item);
> > +        ASSERT(red_channel_no_item_being_sent(&worker->cursor_channel->common.base));
> >     }
> >
> >     if (worker->display_channel) {
> > @@ -9648,7 +9648,7 @@ static inline void handle_dev_destroy_primary_surface(RedWorker *worker)
> >         if (!worker->cursor_channel->common.base.migrate) {
> >             red_pipe_add_verb(&worker->cursor_channel->common.base, SPICE_MSG_CURSOR_RESET);
> >         }
> > -        ASSERT(!worker->cursor_channel->common.base.send_data.item);
> > +        ASSERT(red_channel_no_item_being_sent(&worker->cursor_channel->common.base));
> >     }
> >
> >     flush_all_qxl_commands(worker);
> > @@ -9710,7 +9710,7 @@ static void handle_dev_input(EventListener *listener, uint32_t events)
> >             if (!cursor_red_channel->migrate) {
> >                 red_pipe_add_verb(cursor_red_channel, SPICE_MSG_CURSOR_RESET);
> >             }
> > -            ASSERT(!cursor_red_channel->send_data.item);
> > +            ASSERT(red_channel_no_item_being_sent(cursor_red_channel));
> >
> >             worker->cursor_visible = TRUE;
> >             worker->cursor_position.x = worker->cursor_position.y = 0;
> > --
> > 1.7.4
> >
> > _______________________________________________
> > Spice-devel mailing list
> > Spice-devel at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/spice-devel
> >
> 
> 
> 
> -- 
> Marc-André Lureau


More information about the Spice-devel mailing list