[Spice-devel] [RFC PATCH spice-server v4 19/22] cursor-channel: Allow to not handle a QXL device

[Christophe de Dinechin]

> On 25 Aug 2017, at 11:54, Frediano Ziglio <fziglio at redhat.com> wrote:
> 
> QXL device requires a QXL interface which imply all set

“all sorts”?


> of requirements like memory sharing.
> Currently RedCursorChannel uses QXL to release resources
> allocated from such devices.
> However this is quite a limitation as potentially we
> can implement cursors in a virtual device or in a different
> one.

I don’t understand what “one” refers to here.

> This for instance is the case of streaming device.
> 
> Signed-off-by: Frediano Ziglio <fziglio at redhat.com>
> ---
> server/cursor-channel.c | 8 +++++---
> server/cursor-channel.h | 4 ++--
> server/red-worker.c     | 2 +-
> 3 files changed, 8 insertions(+), 6 deletions(-)
> 
> diff --git a/server/cursor-channel.c b/server/cursor-channel.c
> index 63d6886d..18071ec7 100644
> --- a/server/cursor-channel.c
> +++ b/server/cursor-channel.c
> @@ -94,7 +94,9 @@ static void cursor_item_unref(CursorItem *item)
>         return;
> 
>     cursor_cmd = item->red_cursor;
> -    red_qxl_release_resource(item->qxl, cursor_cmd->release_info_ext);
> +    if (item->qxl) {
> +        red_qxl_release_resource(item->qxl, cursor_cmd->release_info_ext);
> +    }
>     red_put_cursor_cmd(cursor_cmd);
>     free(cursor_cmd);
> 
> @@ -289,7 +291,7 @@ static void cursor_channel_send_item(RedChannelClient *rcc, RedPipeItem *pipe_it
>     red_channel_client_begin_send_message(rcc);
> }
> 
> -CursorChannel* cursor_channel_new(RedsState *server, QXLInstance *qxl,
> +CursorChannel* cursor_channel_new(RedsState *server, QXLInstance *qxl, uint32_t id,
>                                   const SpiceCoreInterfaceInternal *core)
> {
>     spice_debug("create cursor channel");
> @@ -297,7 +299,7 @@ CursorChannel* cursor_channel_new(RedsState *server, QXLInstance *qxl,
>                         "spice-server", server,
>                         "core-interface", core,
>                         "channel-type", SPICE_CHANNEL_CURSOR,
> -                        "id", qxl->id,
> +                        "id", qxl ? qxl->id : id,
>                         "migration-flags", 0,
>                         "qxl", qxl,
>                         "handle-acks", TRUE,
> diff --git a/server/cursor-channel.h b/server/cursor-channel.h
> index 9f3ef52e..c173e2ca 100644
> --- a/server/cursor-channel.h
> +++ b/server/cursor-channel.h
> @@ -56,8 +56,8 @@ GType cursor_channel_get_type(void) G_GNUC_CONST;
>  * provided as helper functions and should only be called from the
>  * CursorChannel thread.
>  */
> -CursorChannel*       cursor_channel_new         (RedsState *server, QXLInstance *qxl,
> -                                                 const SpiceCoreInterfaceInternal *core);
> +CursorChannel* cursor_channel_new(RedsState *server, QXLInstance *qxl, uint32_t id,
> +                                  const SpiceCoreInterfaceInternal *core);
> 
> /**
>  * Cause the channel to disconnect all clients
> diff --git a/server/red-worker.c b/server/red-worker.c
> index 03a409cd..c3db477f 100644
> --- a/server/red-worker.c
> +++ b/server/red-worker.c
> @@ -1342,7 +1342,7 @@ RedWorker* red_worker_new(QXLInstance *qxl,
> 
>     worker->event_timeout = INF_EVENT_WAIT;
> 
> -    worker->cursor_channel = cursor_channel_new(reds, qxl,
> +    worker->cursor_channel = cursor_channel_new(reds, qxl, qxl->id,
>                                                 &worker->core);

With the code as written, why pass qxl->id and not 0? Or else do as Jonathon
suggested and get rid of the test in cursor_channel_new.

>     channel = RED_CHANNEL(worker->cursor_channel);
>     red_channel_init_stat_node(channel, &worker->stat, "cursor_channel");
> -- 
> 2.13.5


Except for commit message, LGTM.


Christophe

> 
> _______________________________________________
> Spice-devel mailing list
> Spice-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/spice-devel