mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-12-30 04:00:37 +00:00
ext/jack/gstjackaudioclient.c: Don't need to take the connection lock, it will not be used and could cause deadlocks.
Original commit message from CVS: Based on patch by: Paul Davis <paul at linuxaudiosystems dot com> * ext/jack/gstjackaudioclient.c: (gst_jack_audio_unref_connection): Don't need to take the connection lock, it will not be used and could cause deadlocks.
This commit is contained in:
parent
5b3587b7f5
commit
0abc869e3e
2 changed files with 27 additions and 7 deletions
|
@ -1,3 +1,11 @@
|
|||
2007-03-18 Wim Taymans <wim@fluendo.com>
|
||||
|
||||
Based on patch by: Paul Davis <paul at linuxaudiosystems dot com>
|
||||
|
||||
* ext/jack/gstjackaudioclient.c: (gst_jack_audio_unref_connection):
|
||||
Don't need to take the connection lock, it will not be used and could
|
||||
cause deadlocks.
|
||||
|
||||
2007-03-16 Edward Hervey <edward@fluendo.com>
|
||||
|
||||
* sys/osxvideo/osxvideosink.m:
|
||||
|
|
|
@ -283,18 +283,32 @@ static void
|
|||
gst_jack_audio_unref_connection (GstJackAudioConnection * conn)
|
||||
{
|
||||
gint res;
|
||||
gboolean zero;
|
||||
|
||||
GST_DEBUG ("unref connection %p", conn);
|
||||
GST_DEBUG ("unref connection %p refcnt %d", conn, conn->refcount);
|
||||
|
||||
G_LOCK (connections_lock);
|
||||
conn->refcount--;
|
||||
if (conn->refcount == 0) {
|
||||
if ((zero = (conn->refcount == 0))) {
|
||||
GST_DEBUG ("closing connection %p", conn);
|
||||
/* remove from list */
|
||||
/* remove from list, we can release the mutex after removing the connection
|
||||
* from the list because after that, nobody can access the connection anymore. */
|
||||
connections = g_list_remove (connections, conn);
|
||||
}
|
||||
G_UNLOCK (connections_lock);
|
||||
|
||||
/* grab lock to be sure that we are not in one of the callbacks */
|
||||
g_mutex_lock (conn->lock);
|
||||
/* if we are zero, close and cleanup the connection */
|
||||
if (zero) {
|
||||
/* don't use conn->lock here. two reasons:
|
||||
*
|
||||
* 1) its not necessary: jack_deactivate() will not return until the JACK thread
|
||||
* associated with this connection is cleaned up by a thread join, hence
|
||||
* no more callbacks can occur or be in progress.
|
||||
*
|
||||
* 2) it would deadlock anyway, because jack_deactivate() will sleep
|
||||
* waiting for the JACK thread, and can thus cause deadlock in
|
||||
* jack_process_cb()
|
||||
*/
|
||||
if ((res = jack_deactivate (conn->client))) {
|
||||
/* we only warn, this means the server is probably shut down and the client
|
||||
* is gone anyway. */
|
||||
|
@ -305,7 +319,6 @@ gst_jack_audio_unref_connection (GstJackAudioConnection * conn)
|
|||
/* we assume the client is gone. */
|
||||
GST_WARNING ("close failed (%d)", res);
|
||||
}
|
||||
g_mutex_unlock (conn->lock);
|
||||
|
||||
/* free resources */
|
||||
g_mutex_free (conn->lock);
|
||||
|
@ -313,7 +326,6 @@ gst_jack_audio_unref_connection (GstJackAudioConnection * conn)
|
|||
g_free (conn->server);
|
||||
g_free (conn);
|
||||
}
|
||||
G_UNLOCK (connections_lock);
|
||||
}
|
||||
|
||||
static void
|
||||
|
|
Loading…
Reference in a new issue