Techwizz/kernel-fxtec-pro1x
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

[DCCP] ccid3: Simplify calculation for reverse lookup of p

Browse source 
This simplifies the calculation of a value p for a given fval when the
 first loss interval is computed (RFC 3448, 6.3.1). It makes use of the
 two new functions scaled_div/scaled_div32 to provide overflow protection.

 Additionally, protection against divide-by-zero is extended - in this
 case the function will return the maximally possible value of p=100%.

Background:

 The maximum fval, f(100%), is approximately 244, i.e. the scaled value of fval
 should never exceed 244E6, which fits easily into u32. The problem is the scaling
 by 10^6, since additionally R(TT) is in microseconds.
 This is resolved by breaking the division into two stages: the first stage
 computes fval=(s*10^6)/R, stores that into u64; the second stage computes
 fval = (fval*10^6)/X_recv and complains if overflow is reached for u32.
 This case is safe since the TFRC reverse-lookup routine then returns p=100%.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
This commit is contained in:
Gerrit Renker 2006-12-10 00:03:30 -02:00 committed by David S. Miller
parent b9039a2a8d
commit bfe24a6cc2
1 changed files with 22 additions and 17 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

39
net/dccp/ccids/ccid3.c
View file

@ -785,12 +785,12 @@ static u32 ccid3_hc_rx_calc_first_li(struct sock *sk)
{
struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
struct dccp_rx_hist_entry *entry, *next, *tail = NULL;
u32 rtt, delta, x_recv, fval, p, tmp2;
u32 rtt, delta, x_recv, p;
struct timeval tstamp = { 0, };
int interval = 0;
int win_count = 0;
int step = 0;
u64 tmp1;
u64 fval;
list_for_each_entry_safe(entry, next, &hcrx->ccid3hcrx_hist,
dccphrx_node) {
@ -834,30 +834,35 @@ static u32 ccid3_hc_rx_calc_first_li(struct sock *sk)
ccid3_pr_debug("%s, sk=%p, approximated RTT to %uus\n",
dccp_role(sk), sk, rtt);
if (rtt == 0) {
DCCP_WARN("RTT==0, setting to 1\n");
rtt = 1;
/*
* Determine the length of the first loss interval via inverse lookup.
* Assume that X_recv can be computed by the throughput equation
* s
* X_recv = --------
* R * fval
* Find some p such that f(p) = fval; return 1/p [RFC 3448, 6.3.1].
*/
if (rtt == 0) { /* would result in divide-by-zero */
DCCP_WARN("RTT==0, returning 1/p = 1\n");
return 1000000;
}
dccp_timestamp(sk, &tstamp);
delta = timeval_delta(&tstamp, &hcrx->ccid3hcrx_tstamp_last_feedback);
x_recv = scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
if (x_recv == 0)
x_recv = hcrx->ccid3hcrx_x_recv;
tmp1 = (u64)x_recv * (u64)rtt;
do_div(tmp1,10000000);
tmp2 = (u32)tmp1;
if (!tmp2) {
DCCP_CRIT("tmp2 = 0, x_recv = %u, rtt =%u\n", x_recv, rtt);
return ~0;
if (x_recv == 0) { /* would also trigger divide-by-zero */
DCCP_WARN("X_recv==0\n");
if ((x_recv = hcrx->ccid3hcrx_x_recv) == 0) {
DCCP_BUG("stored value of X_recv is zero");
return 1000000;
}
}
fval = (hcrx->ccid3hcrx_s * 100000) / tmp2;
/* do not alter order above or you will get overflow on 32 bit */
fval = scaled_div(hcrx->ccid3hcrx_s, rtt);
fval = scaled_div32(fval, x_recv);
p = tfrc_calc_x_reverse_lookup(fval);
ccid3_pr_debug("%s, sk=%p, receive rate=%u bytes/s, implied "
"loss rate=%u\n", dccp_role(sk), sk, x_recv, p);