rx: Enter loss recovery when we retransmit

Since I mistakenly wrote commit 36e2d13b, RX hasn't entered congestion
avoidance when a loss event occurs. This is bad, because on todays
networks the majority of packet losses are due to some form of
congestion.

Now that the timeout code has been restructured, the chances of entering
the retransmit routine in error are much much smaller, so this code
needs to be restored.

This change reverts 36e2d13b55085c996d38b30d003296c602ef8ee3. However,
the original RX code has the problem that it assumes that all forms of
fast recovery are the same - in particular, that the call settings that
result from entering fast recovery due to a fast retransmit are
identical to those resulting from a timeout. This is not the case, and
this will be fixed in a later change.

Reviewed-on: http://gerrit.openafs.org/4866
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit e8c4fc4f37912a5d30694caef45d27374e9201e7)

Change-Id: I5ee6aa255a0b923f445dde2f1e9f0abf98dbe623
Reviewed-on: http://gerrit.openafs.org/4938
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

diff --git a/src/rx/rx.c b/src/rx/rx.c
index 538ce706f354ca2c143a539f63e464d026144725..9612892c05793b80ab0bd1b3e1e7ffd8d870159e 100644 (file)
--- a/src/rx/rx.c
+++ b/src/rx/rx.c
@@ -5949,10 +5949,14 @@ static void
 rxi_Resend(struct rxevent *event, void *arg0, void *arg1, int istack)
 {
     struct rx_call *call = arg0;
+    struct rx_peer *peer;
     struct rx_packet *p, *nxp;
     struct clock maxTimeout = { 60, 0 };
 
     MUTEX_ENTER(&call->lock);
+
+    peer = call->conn->peer;
+
     /* Make sure that the event pointer is removed from the call
      * structure, since there is no longer a per-call retransmission
      * event pending. */
@@ -5974,6 +5978,21 @@ rxi_Resend(struct rxevent *event, void *arg0, void *arg1, int istack)
        goto out;
     }
 
+#ifdef AFS_GLOBAL_RXLOCK_KERNEL
+    if (call->flags & RX_CALL_FAST_RECOVER_WAIT) {
+       /* Someone else is waiting to start recovery */
+       goto out;
+    }
+    call->flags |= RX_CALL_FAST_RECOVER_WAIT;
+    rxi_WaitforTQBusy(call);
+    call->flags &= ~RX_CALL_FAST_RECOVER_WAIT;
+    if (call->error)
+       goto out;
+#endif
+
+    /* We're in loss recovery */
+    call->flags |= RX_CALL_FAST_RECOVER;
+
     /* Mark all of the pending packets in the queue as being lost */
     for (queue_Scan(&call->tq, p, nxp, rx_packet)) {
        if (!(p->flags & RX_PKTFLAG_ACKED))
@@ -5983,12 +6002,32 @@ rxi_Resend(struct rxevent *event, void *arg0, void *arg1, int istack)
     /* We're resending, so we double the timeout of the call. This will be
      * dropped back down by the first successful ACK that we receive.
      *
-     * We apply a maximum value here of 60 second
+     * We apply a maximum value here of 60 seconds
      */
     clock_Add(&call->rto, &call->rto);
     if (clock_Gt(&call->rto, &maxTimeout))
        call->rto = maxTimeout;
 
+    /* Packet loss is most likely due to congestion, so drop our window size
+     * and start again from the beginning */
+    if (peer->maxDgramPackets >1) {
+       call->MTU = RX_JUMBOBUFFERSIZE + RX_HEADER_SIZE;
+        call->MTU = MIN(peer->natMTU, peer->maxMTU);
+    }
+    call->ssthresh = MAX(4, MIN((int)call->cwind, (int)call->twind)) >> 1;
+    call->nDgramPackets = 1;
+    call->cwind = 1;
+    call->nextCwind = 1;
+    call->nAcks = 0;
+    call->nNacks = 0;
+    MUTEX_ENTER(&peer->peer_lock);
+    peer->MTU = call->MTU;
+    peer->cwind = call->cwind;
+    peer->nDgramPackets = 1;
+    peer->congestSeq++;
+    call->congestSeq = peer->congestSeq;
+    MUTEX_EXIT(&peer->peer_lock);
+
     rxi_Start(call, istack);
 
 out: