chan_sip: more accurate retransmissions

RFC3261 states that Timer A should start at 500ms (T1) by default.
In chan_sip this value initially started at 1000ms and I changed
it to 500ms recently. After doing that I noticed in my packet
captures that it still occasionally retransmitted starting at
1000ms instead of 500ms like I told it to.  This occurs because
the scheduler runs in the do_monitor thread.  If a new retransmission
is added while the do_monitor thread is sleeping then it may not
detect that retransmission for nearly 1000ms.  To fix this I just
poke the do_monitor thread to wake up when a new packet is sent
reliably requiring retransmits.  The thread then detects the new
scheduler entry and adjusts its sleep time to account for it.

Review: https://reviewboard.asterisk.org/r/747

git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@272557 65c4cc65-6c06-0410-ace0-fbb531ad65f3

diff --git a/channels/chan_sip.c b/channels/chan_sip.c
index b3bf2cd..5720f9b 100644 (file)
--- a/channels/chan_sip.c
+++ b/channels/chan_sip.c
@@ -3493,6 +3493,11 @@ static enum sip_result __sip_reliable_xmit(struct sip_pvt *p, int seqno, int res
                ast_free(pkt);
                return AST_FAILURE;
        } else {
+               /* This is odd, but since the retrans timer starts at 500ms and the do_monitor thread
+                * only wakes up every 1000ms by default, we have to poke the thread here to make
+                * sure it successfully detects this must be retransmitted in less time than
+                * it usually sleeps for. Otherwise it might not retransmit this packet for 1000ms. */
+               pthread_kill(monitor_thread, SIGURG);
                return AST_SUCCESS;
        }
 }