[PATCH 1/6]: Avoid accumulation of large send credit

[PATCH 1/6]: Avoid accumulation of large send credit

[Gerrit Renker]

[CCID 3]: Avoid accumulation of large send credit

Backlogs of packets can accumulate in the CCID3 TX module when for instance
the application does not send for a longer time or sends slowly.
With regard to computing t_nom, this leads to an accumulation of send credits.
The present patch avoids large packet bursts that would result under these
conditions, by setting the backlog window of tardiness to one t_ipi unit.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
---
 net/dccp/ccids/ccid3.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

--- a/net/dccp/ccids/ccid3.c
+++ b/net/dccp/ccids/ccid3.c
@@ -352,7 +352,9 @@ static int ccid3_hc_tx_send_packet(struc
 	case TFRC_SSTATE_NO_FBACK:
 	case TFRC_SSTATE_FBACK:
 		delay = ktime_delta(hctx->ccid3hctx_t_nom, now);
-		ccid3_pr_debug("delay=%ld\n", (long)delay);
+		/* Resynchronise t_nom when sender has been idle or slow */
+		if (delay + (s64)hctx->ccid3hctx_t_ipi  <  0)
+			hctx->ccid3hctx_t_nom = now;
 		/*
 		 *	Scheduling of packet transmissions [RFC 3448, 4.6]
 		 *
@@ -361,7 +363,7 @@ static int ccid3_hc_tx_send_packet(struc
 		 * else
 		 *       // send the packet in (t_nom - t_now) milliseconds.
 		 */
-		if (delay - (s64)hctx->ccid3hctx_delta >= 1000)
+		else if (delay - (s64)hctx->ccid3hctx_delta >= 1000)
 			return (u32)delay / 1000L;
 
 		ccid3_hc_tx_update_win_count(hctx, now);

Re: [PATCH 1/6]: Avoid accumulation of large send credit

[Ian McDonald]

On 6/10/07, Gerrit Renker <gerrit@erg.abdn.ac.uk> wrote:
> [CCID 3]: Avoid accumulation of large send credit
>
> Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>

Agree with Gerrit's comments in patchset description that there is
still discussion around this. Still doing something is better than
doing nothing.

Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
-- 
Web: http://wand.net.nz/~iam4/
Blog: http://iansblog.jandi.co.nz
WAND Network Research Group

Re: [PATCH 1/6]: Avoid accumulation of large send credit

[Gerrit Renker]

|  Agree with Gerrit's comments in patchset description that there is
|  still discussion around this. Still doing something is better than
|  doing nothing.
Also agree with what you are saying. This is really why I would like
to get the experimental test tree for DCCP/CCID3 started soon.

With the mechanism already in place, it is a trivial change to use
a different value. 

Kohler/Floyd currently suggest to use 1 RTT, which derives from the theory of TFRC. 

The value t_ipi derives from the way the packets are currently scheduled.

Neither value may be the last piece of wisdom - we need people testing
this and report their experience + suggestions.