From: Chas Williams <chas@cmf.nrl.navy.mil>
Date: Mon, 12 Jun 2006 05:23:54 +0000 (+0000)
Subject: recvackpkt-count-acks-and-open-cong-window-20060612
X-Git-Tag: BP-openafs-windows-kdfs-ifs~1272
X-Git-Url: https://git.michaelhowe.org/gitweb/?a=commitdiff_plain;h=b365b06d34a9450dc53ad22e8ecb3f6d922af5f4;p=packages%2Fo%2Fopenafs.git

recvackpkt-count-acks-and-open-cong-window-20060612


"the non AFS_GLOBAL_RXLOCK_KERNEL version of
rxi_ReceiveAckPacket() wasnt counting 'hard' acks and not opening the
congestion window.  using rx_Readv() directs the incoming packets into
the target iov, and sends a hard ack (skipping soft acks).  "
---

diff --git a/src/rx/rx.c b/src/rx/rx.c
index 7545b085d..32e70d3d6 100644
--- a/src/rx/rx.c
+++ b/src/rx/rx.c
@@ -3776,6 +3776,9 @@ rxi_ReceiveAckPacket(register struct rx_call *call, struct rx_packet *np,
 	if (serial
 	    && (tp->header.serial == serial || tp->firstSerial == serial))
 	    rxi_ComputePeerNetStats(call, tp, ap, np);
+	if (!(tp->flags & RX_PKTFLAG_ACKED)) {
+	    newAckCount++;
+	}
 #ifdef	AFS_GLOBAL_RXLOCK_KERNEL
 	/* XXX Hack. Because we have to release the global rx lock when sending
 	 * packets (osi_NetSend) we drop all acks while we're traversing the tq
@@ -3786,9 +3789,6 @@ rxi_ReceiveAckPacket(register struct rx_call *call, struct rx_packet *np,
 	 * set the ack bits in the packets and have rxi_Start remove the packets
 	 * when it's done transmitting.
 	 */
-	if (!(tp->flags & RX_PKTFLAG_ACKED)) {
-	    newAckCount++;
-	}
 	if (call->flags & RX_CALL_TQ_BUSY) {
 #ifdef RX_ENABLE_LOCKS
 	    tp->flags |= RX_PKTFLAG_ACKED;