Date:	Wed, 29 Jun 2011 09:57:04 -0400
From:	Thomas Graf <tgraf@...radead.org>
To:	netdev@...r.kernel.org
Cc:	davem@...emloft.net, Wei Yongjun <yjwei@...fujitsu.com>,
	Vlad Yasevich <vladislav.yasevich@...com>,
	Sridhar Samudrala <sri@...ibm.com>, linux-sctp@...r.kernel.org
Subject: [PATCH] sctp: Enforce maximum retransmissions during shutdown

When initiating a graceful shutdown while having data chunks
on the retransmission queue with a peer which is in zero
window mode the shutdown is never completed because the
retransmission error count is reset periodically by the
following two rules:

 - Do not timeout association while doing zero window probe.
 - Reset overall error count when a heartbeat request has
   been acknowledged.

The graceful shutdown will wait for all outstanding TSN to be
acknowledged before sending the SHUTDOWN request. This never
happens due to the peer's zero window not acknowledging the
continuously retransmitted data chunks. Although the error
counter is incremented for each failed retransmission done
via the T3-rtx timer, the receiving of the SACK sent in return
to the retransmission, announcing the zero window, clears the
error count again immediately.

Also heartbeat requests continue to be sent periodically. The
peer acknowledges these requests causing the error counter to
be reset as well.

This patch changes behaviour to only reset the overall error
counter for the above rules while not in shutdown. This means
that if already queued data can't be transmitted in max_retrans
attempts we ABORT because a graceful shutdown is obviously not
possible.

The issue can be easily reproduced by establishing a sctp
association over the loopback device, constantly queueing data
at the sender while not reading any at the receiver.  Wait for
the window to reach zero, then initiate a shutdown by killing
both processes simultaneously. The association will never be
freed and the chunks on the retransmission queue will be
retransmitted indefinitely.

Signed-off-by: Thomas Graf <tgraf@...radead.org>

diff --git a/net/sctp/outqueue.c b/net/sctp/outqueue.c
index 1c88c89..14a5295 100644
--- a/net/sctp/outqueue.c
+++ b/net/sctp/outqueue.c
@@ -1629,10 +1629,15 @@ static void sctp_check_transmitted(struct sctp_outq *q,
 			 * A sender is doing zero window probing when the
 			 * receiver's advertised window is zero, and there is
 			 * only one data chunk in flight to the receiver.
+			 *
+			 * Allow the association to timeout if SHUTDOWN is
+			 * pending. We have no interest in keeping the
+			 * association around forever.
 			 */
 			if (!q->asoc->peer.rwnd &&
 			    !list_empty(&tlist) &&
-			    (sack_ctsn+2 == q->asoc->next_tsn)) {
+			    (sack_ctsn+2 == q->asoc->next_tsn) &&
+			    !(q->asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)) {
 				SCTP_DEBUG_PRINTK("%s: SACK received for zero "
 						  "window probe: %u\n",
 						  __func__, sack_ctsn);
diff --git a/net/sctp/sm_sideeffect.c b/net/sctp/sm_sideeffect.c
index 534c2e5..fa92f4d6 100644
--- a/net/sctp/sm_sideeffect.c
+++ b/net/sctp/sm_sideeffect.c
@@ -670,10 +670,21 @@ static void sctp_cmd_transport_on(sctp_cmd_seq_t *cmds,
 	/* 8.3 Upon the receipt of the HEARTBEAT ACK, the sender of the
 	 * HEARTBEAT should clear the error counter of the destination
 	 * transport address to which the HEARTBEAT was sent.
-	 * The association's overall error count is also cleared.
 	 */
 	t->error_count = 0;
-	t->asoc->overall_error_count = 0;
+
+	/*
+	 * Although RFC2960 and RFC4460 specify that the overall error
+	 * count must be cleared when a HEARTBEAT ACK is received this
+	 * behaviour may prevent the maximum retransmission count from
+	 * being reached while in SHUTDOWN. If the peer keeps its window
+	 * closed not acknowledging any outstanding TSN we may rely on
+	 * reaching the max_retrans limit via the T3-rtx timer to close
+	 * the association which will never happen if the error count is
+	 * reset every heartbeat interval.
+	 */
+	if (!(t->asoc->state >= SCTP_STATE_SHUTDOWN_PENDING))
+		t->asoc->overall_error_count = 0;
 
 	/* Clear the hb_sent flag to signal that we had a good
 	 * acknowledgement.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists