On Tue, 29 May 2007 20:23:45 +0200
"Lior Dotan" [off-list ref] wrote:

NTP was not running. I'm not sure what do you mean by fixing the -1.
The trace shows that vegas_cong_avoid() is called with -1, and the
only way it can happen is from tcp_clean_rtx_queue() and the patch
should eliminate this. Another way of solving this is by checking
vegas_rtt_calc() and see if it gets -1 and handle it there.
Another thing that I don't understand is that some places like
tcp_ack() declare seq_rtt as signed and some vegas_cong_avoid()
declare it as unsigned. Shouldn't it be declared always as signed?

On 5/29/07, Stephen Hemminger [off-list ref] wrote:

quoted

On Tue, 29 May 2007 12:18:19 +0300
"Lior Dotan" [off-list ref] wrote:

quoted

Hi,

I had a divide by zero on kernel 2.4.33 running with Vegas enabled.

I don't think anyone has backported the other vegas fixes from 2.6.21
to 2.4.

quoted

quoted

The KDB back trace is:
kdb> bt
Stack traceback for pid 0
0x403a6000        0        0  1    0   R  0x403a6370 *swapper
EBP        EIP        Function (args)
0x403a7d48 0x4026ae51 vegas_cong_avoid+0x111 (0x5f3bb638, 0x73c92cbb, 0xffffffff
, 0x73c92cbb, 0xf28d0275)
                               kernel .text 0x40100000 0x4026ad40 0x4026aef0
0x403a7d8c 0x4026bb67 tcp_ack+0x307 (0x5f3bb560, 0x581985c0, 0x18e, 0x4023e765,
0x5c369044)
                               kernel .text 0x40100000 0x4026b860 0x4026be20

Seems like fixing the -1 would be better.  Perhaps NTP reset
the clock?



--
Stephen Hemminger [off-list ref]

For 2.6.22, rtt_calc doesn't exist, instead pkts_acked is called. The following
should be added to avoid getting bogus timestamp values from retransmits.

--- a/net/ipv4/tcp_vegas.c	2007-05-02 12:26:35.000000000 -0700
+++ b/net/ipv4/tcp_vegas.c	2007-05-29 14:06:26.000000000 -0700

@@ -117,6 +117,10 @@ void tcp_vegas_pkts_acked(struct sock *s
 	struct vegas *vegas = inet_csk_ca(sk);
 	u32 vrtt;
 
+	/* Ignore retransmits */
+	if (unlikely(cnt == 0))
+		return;
+
 	/* Never allow zero rtt or baseRTT */
 	vrtt = ktime_to_us(net_timedelta(last)) + 1;
 

-- 
Stephen Hemminger <shemminger@linux-foundation.org>