After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 561241 - metadatademux pad activation deadlock
metadatademux pad activation deadlock
Status: RESOLVED FIXED
Product: GStreamer
Classification: Platform
Component: gst-plugins-bad
git master
Other Linux
: Normal normal
: 0.10.14
Assigned To: GStreamer Maintainers
GStreamer Maintainers
Depends on:
Blocks:
 
 
Reported: 2008-11-17 16:58 UTC by Alessandro Decina
Modified: 2009-01-09 23:23 UTC
See Also:
GNOME target: ---
GNOME version: ---


Description Alessandro Decina 2008-11-17 16:58:50 UTC 
I found a bug similar to http://bugzilla.gnome.org/show_bug.cgi?id=560380 in metadatademux. I'm seeing it frequently running Elisa but I can't isolate it in a script.

Again the problem is that filesrc is activated in push mode while downstream is pulling. The source of the problem is in gstbasemetadata.c:gst_base_metadata_sink_activate

static gboolean
gst_base_metadata_sink_activate (GstPad * pad)
{
  GstBaseMetadata *filter = NULL;
  gboolean ret = TRUE;


  filter = GST_BASE_METADATA (GST_PAD_PARENT (pad));

  if (!gst_pad_check_pull_range (pad) ||
      !gst_pad_activate_pull (filter->sinkpad, TRUE)) {
    /* Fail here ? nothing to be done by now, activate push mode */
    return gst_pad_activate_push (pad, TRUE);
  }

  /* try to base */
  if (filter->state == MT_STATE_NULL) {
    ret = gst_base_metadata_pull_range_parse (filter);
  }

  if (ret) {
    gst_pad_activate_pull (pad, FALSE);
    gst_pad_activate_push (filter->srcpad, FALSE);
    if (!gst_pad_is_active (pad)) {
      ret = gst_pad_activate_push (filter->srcpad, TRUE);
      ret = ret && gst_pad_activate_push (pad, TRUE);
    }
  }

  return ret;

}

It tries to do something like typefind does, but looks really wrong.

Here's the stack trace I get when it deadlocks:

(gdb) thr a a bt

+ Trace 209936







Comment 1


Alessandro Decina





          2008-11-17 17:54:35 UTC
        



This seems to solve my issue:

--- ext/metadata/gstbasemetadata.c	18 Jun 2008 06:31:11 -0000	1.9
+++ ext/metadata/gstbasemetadata.c	17 Nov 2008 17:50:52 -0000
@@ -1807,16 +1807,24 @@
   }
 
   if (ret) {
-    gst_pad_activate_pull (pad, FALSE);
-    gst_pad_activate_push (filter->srcpad, FALSE);
-    if (!gst_pad_is_active (pad)) {
-      ret = gst_pad_activate_push (filter->srcpad, TRUE);
-      ret = ret && gst_pad_activate_push (pad, TRUE);
+    GstActivateMode mode;
+
+    /* in gst_base_metadata_pull_range_parse() we could have triggered
+     * negotiation and plugged new downstream elements.
+     * If GST_PAD_ACTIVATE_MODE (filter->srcpad) is GST_ACTIVATE_PULL it means
+     * that downstream is active in pull mode so we don't deactivate pull mode.
+     */
+    GST_OBJECT_LOCK (filter->srcpad);
+    mode = GST_PAD_ACTIVATE_MODE (filter->srcpad);
+    GST_OBJECT_UNLOCK (filter->srcpad);
+
+    if (mode != GST_ACTIVATE_PULL) {
+      /* change from PULL to PUSH */
+      gst_pad_activate_push (pad, TRUE);
     }
   }
 
   return ret;
-
 }