Merged revisions 330312 via svnmerge from

https://origsvn.digium.com/svn/asterisk/branches/10

................
  r330312 | irroot | 2011-07-30 17:34:41 +0200 (Sat, 30 Jul 2011) | 15 lines

  Merged revisions 330311 via svnmerge from
  https://origsvn.digium.com/svn/asterisk/branches/1.8

  ........
    r330311 | irroot | 2011-07-30 17:25:16 +0200 (Sat, 30 Jul 2011) | 9 lines

    prevent double masqurading channels when one is been hung up and deadlock avoidance is used.

    There is a race condition in ast_do_masquerade / ast_hangup (at least)

    Reported by me signed off by schmidts with input from David Vossel

    Review: https://reviewboard.asterisk.org/r/1323/
  ........
................

git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@330313 65c4cc65-6c06-0410-ace0-fbb531ad65f3

diff --git a/main/channel.c b/main/channel.c
index 902c79d..f4a04d5 100644 (file)
--- a/main/channel.c
+++ b/main/channel.c
@@ -6518,10 +6518,18 @@ int ast_do_masquerade(struct ast_channel *original)
         * for deadlock avoidance will not result in any sort of masquerade race condition.
         * If masq is called by a different thread while this happens, it will be stuck waiting
         * until we unlock the container. */
-       while (ast_channel_trylock(clonechan)) {
+       while ((clonechan = original->masq) && ast_channel_trylock(clonechan)) {
                CHANNEL_DEADLOCK_AVOIDANCE(original);
        }
 
+       /* recheck if masq has been done the deadlock avoidance above could cause a double masq
+        * this is posible with at least ast_hangup*/
+       if (!original->masq) {
+               ast_channel_unlock(original);
+               ao2_unlock(channels);
+               return 0; /* masq already completed by another thread, or never needed to be done to begin with */
+       }
+
        /* Get any transfer masquerade connected line exchange data. */
        xfer_ds = ast_channel_datastore_find(original, &xfer_ds_info, NULL);
        if (xfer_ds) {