On 23/08/2017 18:19, Marc-André Lureau wrote:
> The following patches is going to remove the custom VUS source.
>
> Signed-off-by: Marc-André Lureau <marcandre.lur...@redhat.com>
> ---
> contrib/vhost-user-scsi/vhost-user-scsi.c | 18 +++++++++---------
> 1 file changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/contrib/vhost-user-scsi/vhost-user-scsi.c
> b/contrib/vhost-user-scsi/vhost-user-scsi.c
> index 596c1effa1..b40009e234 100644
> --- a/contrib/vhost-user-scsi/vhost-user-scsi.c
> +++ b/contrib/vhost-user-scsi/vhost-user-scsi.c
> @@ -136,7 +136,7 @@ static GSourceFuncs vus_gsrc_funcs = {
> };
>
> static void vus_gsrc_new(vhost_scsi_dev_t *vdev_scsi, int fd, GIOCondition
> cond,
> - vu_watch_cb vu_cb, GSourceFunc gsrc_cb, gpointer
> data)
> + vu_watch_cb vu_cb, gpointer data)
> {
> GSource *vus_gsrc;
> vus_gsrc_t *vus_src;
> @@ -144,8 +144,7 @@ static void vus_gsrc_new(vhost_scsi_dev_t *vdev_scsi, int
> fd, GIOCondition cond,
>
> assert(vdev_scsi);
> assert(fd >= 0);
> - assert(vu_cb || gsrc_cb);
> - assert(!(vu_cb && gsrc_cb));
> + assert(vu_cb);
>
> vus_gsrc = g_source_new(&vus_gsrc_funcs, sizeof(vus_gsrc_t));
> vus_src = (vus_gsrc_t *)vus_gsrc;
> @@ -156,7 +155,6 @@ static void vus_gsrc_new(vhost_scsi_dev_t *vdev_scsi, int
> fd, GIOCondition cond,
> vus_src->vu_cb = vu_cb;
>
> g_source_add_poll(vus_gsrc, &vus_src->gfd);
> - g_source_set_callback(vus_gsrc, gsrc_cb, data, NULL);
> id = g_source_attach(vus_gsrc, NULL);
> assert(id);
> g_source_unref(vus_gsrc);
> @@ -450,7 +448,7 @@ static void vus_add_watch_cb(VuDev *vu_dev, int fd, int
> vu_evt, vu_watch_cb cb,
> (void)g_tree_remove(vdev_scsi->fdmap, (gpointer)(uintptr_t)fd);
> }
>
> - vus_gsrc_new(vdev_scsi, fd, vu_evt, cb, NULL, pvt);
> + vus_gsrc_new(vdev_scsi, fd, vu_evt, cb, pvt);
> }
>
> static void vus_del_watch_cb(VuDev *vu_dev, int fd)
> @@ -578,7 +576,8 @@ static const VuDevIface vus_iface = {
> .queue_set_started = vus_queue_set_started,
> };
>
> -static gboolean vus_vhost_cb(gpointer data)
> +static gboolean vus_vhost_cb(GIOChannel *source, GIOCondition condition,
> + gpointer data)
> {
> VuDev *vu_dev = (VuDev *)data;
>
> @@ -679,6 +678,7 @@ static int vdev_scsi_add_iscsi_lun(vhost_scsi_dev_t
> *vdev_scsi,
>
> static int vdev_scsi_run(vhost_scsi_dev_t *vdev_scsi)
> {
> + GIOChannel *chan;
> int cli_sock;
> int ret = 0;
>
> @@ -699,10 +699,10 @@ static int vdev_scsi_run(vhost_scsi_dev_t *vdev_scsi)
> vus_del_watch_cb,
> &vus_iface);
>
> - vus_gsrc_new(vdev_scsi, cli_sock, G_IO_IN, NULL, vus_vhost_cb,
> - &vdev_scsi->vu_dev);
> -
> + chan = g_io_channel_unix_new(cli_sock);
> + g_io_add_watch(chan, G_IO_IN, vus_vhost_cb, &vdev_scsi->vu_dev);
> g_main_loop_run(vdev_scsi->loop);
> + g_io_channel_unref(chan);
>
> vu_deinit(&vdev_scsi->vu_dev);
>
>
It's a bit weird to use GIOChannel here but still operate on the raw
socket elsewhere. Maybe keep on open-coding the watch (using
g_source_new+g_source_add_poll) but remove the "vu_watch_cb vu_cb,
GSourceFunc gsrc_cb" ugliness in vus_gsrc_new, and only support the
vu_watch_cb style? That would be exactly what GIOChannel does with GIOFunc.
Paolo
