Commit de6f2b59 authored by Gerrit Renker's avatar Gerrit Renker

dccp ccid-3: Bug fix for the inter-packet scheduling algorithm

This fixes a subtle bug in the calculation of the inter-packet gap and shows
that t_delta, as it is currently used, is not needed. And hence replaced.

The algorithm from RFC 3448, 4.6 below continually computes a send time t_nom,
which is initialised with the current time t_now; t_gran = 1E6 / HZ specifies
the scheduling granularity, s the packet size, and X the sending rate:

  t_distance = t_nom - t_now;		// in microseconds
  t_delta    = min(t_ipi, t_gran) / 2;	// `delta' parameter in microseconds

  if (t_distance >= t_delta) {
	reschedule after (t_distance / 1000) milliseconds;
  } else {
  	t_ipi  = s / X;			// inter-packet interval in usec
	t_nom += t_ipi;			// compute the next send time
	send packet now;
  }


1) Description of the bug
-------------------------
Rescheduling requires a conversion into milliseconds, due to this call chain:

 * ccid3_hc_tx_send_packet() returns a timeout in milliseconds,
 * this value is converted by msecs_to_jiffies() in dccp_write_xmit(),
 * and finally used as jiffy-expires-value for sk_reset_timer().

The highest jiffy resolution with HZ=1000 is 1 millisecond, so using a higher
granularity does not make much sense here.

As a consequence, values of t_distance < 1000 are truncated to 0. This issue 
has so far been resolved by using instead

  if (t_distance >= t_delta + 1000)
	reschedule after (t_distance / 1000) milliseconds;

The bug is in artificially inflating t_delta to t_delta' = t_delta + 1000. This
is unnecessarily large, a more adequate value is t_delta' = max(t_delta, 1000).


2) Consequences of using the corrected t_delta'
-----------------------------------------------
Since t_delta <= t_gran/2 = 10^6/(2*HZ), we have t_delta <= 1000 as long as
HZ >= 500. This means that t_delta' = max(1000, t_delta) is constant at 1000.

On the other hand, when using a coarse HZ value of HZ < 500, we have three
sub-cases that can all be reduced to using another constant of t_gran/2.

 (a) The first case arises when t_ipi > t_gran. Here t_delta' is the constant
     t_delta' = max(1000, t_gran/2) = t_gran/2.

 (b) If t_ipi <= 2000 < t_gran = 10^6/HZ usec, then t_delta = t_ipi/2 <= 1000,
     so that t_delta' = max(1000, t_delta) = 1000 < t_gran/2. 

 (c) If 2000 < t_ipi <= t_gran, we have t_delta' = max(t_delta, 1000) = t_ipi/2.

In the second and third cases we have delay values less than t_gran/2, which is
in the order of less than or equal to half a jiffy. 

How these are treated depends on how fractions of a jiffy are handled: they
are either always rounded down to 0, or always rounded up to 1 jiffy (assuming
non-zero values). In both cases the error is on average in the order of 50%.

Thus we are not increasing the error when in the second/third case we replace
a value less than t_gran/2 with 0, by setting t_delta' to the constant t_gran/2.


3) Summary
----------
Fixing (1) and considering (2), the patch replaces t_delta with a constant,
whose value depends on CONFIG_HZ, changing the above algorithm to:
 
  if (t_distance >= t_delta')
	reschedule after (t_distance / 1000) milliseconds;

where t_delta' = 10^6/(2*HZ) if HZ < 500, and t_delta' = 1000 otherwise.
Signed-off-by: default avatarGerrit Renker <gerrit@erg.abdn.ac.uk>
parent b2e317f4
......@@ -93,19 +93,16 @@ static inline u64 rfc3390_initial_rate(struct sock *sk)
return scaled_div(w_init << 6, hctx->rtt);
}
/*
* Recalculate t_ipi and delta (should be called whenever X changes)
/**
* ccid3_update_send_interval - Calculate new t_ipi = s / X_inst
* This respects the granularity of X_inst (64 * bytes/second).
*/
static void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hctx)
{
/* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
hctx->t_ipi = scaled_div32(((u64)hctx->s) << 6, hctx->x);
/* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */
hctx->delta = min_t(u32, hctx->t_ipi / 2, TFRC_OPSYS_HALF_TIME_GRAN);
ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n", hctx->t_ipi,
hctx->delta, hctx->s, (unsigned)(hctx->x >> 6));
ccid3_pr_debug("t_ipi=%u, s=%u, X=%u\n", hctx->t_ipi,
hctx->s, (unsigned)(hctx->x >> 6));
}
static u32 ccid3_hc_tx_idle_rtt(struct ccid3_hc_tx_sock *hctx, ktime_t now)
......@@ -340,8 +337,8 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
* else
* // send the packet in (t_nom - t_now) milliseconds.
*/
if (delay - (s64)hctx->delta >= 1000)
return (u32)delay / 1000L;
if (delay >= TFRC_T_DELTA)
return (u32)delay / USEC_PER_MSEC;
ccid3_hc_tx_update_win_count(hctx, now);
break;
......
......@@ -47,12 +47,23 @@
/* Two seconds as per RFC 3448 4.2 */
#define TFRC_INITIAL_TIMEOUT (2 * USEC_PER_SEC)
/* In usecs - half the scheduling granularity as per RFC3448 4.6 */
#define TFRC_OPSYS_HALF_TIME_GRAN (USEC_PER_SEC / (2 * HZ))
/* Parameter t_mbi from [RFC 3448, 4.3]: backoff interval in seconds */
#define TFRC_T_MBI 64
/*
* The t_delta parameter (RFC 3448, 4.6): delays of less than %USEC_PER_MSEC are
* rounded down to 0, since sk_reset_timer() here uses millisecond granularity.
* Hence we can use a constant t_delta = %USEC_PER_MSEC when HZ >= 500. A coarse
* resolution of HZ < 500 means that the error is below one timer tick (t_gran)
* when using the constant t_delta = t_gran / 2 = %USEC_PER_SEC / (2 * HZ).
*/
#if (HZ >= 500)
# define TFRC_T_DELTA USEC_PER_MSEC
#else
# define TFRC_T_DELTA (USEC_PER_SEC / (2 * HZ))
#warning Coarse CONFIG_HZ resolution -- higher value recommended for TFRC.
#endif
enum ccid3_options {
TFRC_OPT_LOSS_EVENT_RATE = 192,
TFRC_OPT_LOSS_INTERVALS = 193,
......@@ -92,7 +103,6 @@ enum ccid3_hc_tx_states {
* @no_feedback_timer - Handle to no feedback timer
* @t_ld - Time last doubled during slow start
* @t_nom - Nominal send time of next packet
* @delta - Send timer delta (RFC 3448, 4.6) in usecs
* @hist - Packet history
* @options_received - Parsed set of retrieved options
*/
......@@ -111,7 +121,6 @@ struct ccid3_hc_tx_sock {
struct timer_list no_feedback_timer;
ktime_t t_ld;
ktime_t t_nom;
u32 delta;
struct tfrc_tx_hist_entry *hist;
struct ccid3_options_received options_received;
};
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment