summaryrefslogtreecommitdiff
path: root/net/dccp/ccids/lib
diff options
context:
space:
mode:
authorGerrit Renker <gerrit@erg.abdn.ac.uk>2008-09-04 07:30:19 +0200
committerGerrit Renker <gerrit@erg.abdn.ac.uk>2008-09-04 07:45:43 +0200
commita3cbdde8e9c38b66b4f13ac5d6ff1939ded0ff20 (patch)
tree8e66f40579776dbc07fdacb206c4d56e1b351e86 /net/dccp/ccids/lib
parent53ac9570c8145710aaed9e1eb850c2e991a4ebc1 (diff)
dccp ccid-3: Preventing Oscillations
This implements [RFC 3448, 4.5], which performs congestion avoidance behaviour by reducing the transmit rate as the queueing delay (measured in terms of long-term RTT) increases. Oscillation can be turned on/off via a module option (do_osc_prev) and via sysfs (using mode 0644), the default is off. Overflow analysis: ------------------ * oscillation prevention is done after update_x(), so that t_ipi <= 64000; * hence the multiplication "t_ipi * sqrt(R_sample)" needs 64 bits; * done using u64 for sqrt_sample and explicit typecast of t_ipi; * the divisor, R_sqmean, is non-zero because oscillation prevention is first called when receiving the second feedback packet, and tfrc_scaled_rtt() > 0. A detailed discussion of the algorithm (with plots) is on http://www.erg.abdn.ac.uk/users/gerrit/dccp/notes/ccid3/sender_notes/oscillation_prevention/ The algorithm has negative side effects: * when allowing to decrease t_ipi (leads to a large RTT) and * when using it during slow-start; both uses are therefore disabled. Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Diffstat (limited to 'net/dccp/ccids/lib')
-rw-r--r--net/dccp/ccids/lib/tfrc.h15
1 files changed, 15 insertions, 0 deletions
diff --git a/net/dccp/ccids/lib/tfrc.h b/net/dccp/ccids/lib/tfrc.h
index bb47146ac7d1..ede12f53de5a 100644
--- a/net/dccp/ccids/lib/tfrc.h
+++ b/net/dccp/ccids/lib/tfrc.h
@@ -48,6 +48,21 @@ static inline u32 scaled_div32(u64 a, u64 b)
}
/**
+ * tfrc_scaled_sqrt - Compute scaled integer sqrt(x) for 0 < x < 2^22-1
+ * Uses scaling to improve accuracy of the integer approximation of sqrt(). The
+ * scaling factor of 2^10 limits the maximum @sample to 4e6; this is okay for
+ * clamped RTT samples (dccp_sample_rtt).
+ * Should best be used for expressions of type sqrt(x)/sqrt(y), since then the
+ * scaling factor is neutralised. For this purpose, it avoids returning zero.
+ */
+static inline u16 tfrc_scaled_sqrt(const u32 sample)
+{
+ const unsigned long non_zero_sample = sample ? : 1;
+
+ return int_sqrt(non_zero_sample << 10);
+}
+
+/**
* tfrc_ewma - Exponentially weighted moving average
* @weight: Weight to be used as damping factor, in units of 1/10
*/