mirror of
https://github.com/torvalds/linux.git
synced 2024-11-05 03:21:32 +00:00
8f363b77ee
Reading TCP stats when using TCP Illinois congestion control algorithm can cause a divide by zero kernel oops. The division by zero occur in tcp_illinois_info() at: do_div(t, ca->cnt_rtt); where ca->cnt_rtt can become zero (when rtt_reset is called) Steps to Reproduce: 1. Register tcp_illinois: # sysctl -w net.ipv4.tcp_congestion_control=illinois 2. Monitor internal TCP information via command "ss -i" # watch -d ss -i 3. Establish new TCP conn to machine Either it fails at the initial conn, or else it needs to wait for a loss or a reset. This is only related to reading stats. The function avg_delay() also performs the same divide, but is guarded with a (ca->cnt_rtt > 0) at its calling point in update_params(). Thus, simply fix tcp_illinois_info(). Function tcp_illinois_info() / get_info() is called without socket lock. Thus, eliminate any race condition on ca->cnt_rtt by using a local stack variable. Simply reuse info.tcpv_rttcnt, as its already set to ca->cnt_rtt. Function avg_delay() is not affected by this race condition, as its called with the socket lock. Cc: Petr Matousek <pmatouse@redhat.com> Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> Acked-by: Eric Dumazet <edumazet@google.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
359 lines
8.2 KiB
C
359 lines
8.2 KiB
C
/*
|
|
* TCP Illinois congestion control.
|
|
* Home page:
|
|
* http://www.ews.uiuc.edu/~shaoliu/tcpillinois/index.html
|
|
*
|
|
* The algorithm is described in:
|
|
* "TCP-Illinois: A Loss and Delay-Based Congestion Control Algorithm
|
|
* for High-Speed Networks"
|
|
* http://www.ifp.illinois.edu/~srikant/Papers/liubassri06perf.pdf
|
|
*
|
|
* Implemented from description in paper and ns-2 simulation.
|
|
* Copyright (C) 2007 Stephen Hemminger <shemminger@linux-foundation.org>
|
|
*/
|
|
|
|
#include <linux/module.h>
|
|
#include <linux/skbuff.h>
|
|
#include <linux/inet_diag.h>
|
|
#include <asm/div64.h>
|
|
#include <net/tcp.h>
|
|
|
|
#define ALPHA_SHIFT 7
|
|
#define ALPHA_SCALE (1u<<ALPHA_SHIFT)
|
|
#define ALPHA_MIN ((3*ALPHA_SCALE)/10) /* ~0.3 */
|
|
#define ALPHA_MAX (10*ALPHA_SCALE) /* 10.0 */
|
|
#define ALPHA_BASE ALPHA_SCALE /* 1.0 */
|
|
#define U32_MAX ((u32)~0U)
|
|
#define RTT_MAX (U32_MAX / ALPHA_MAX) /* 3.3 secs */
|
|
|
|
#define BETA_SHIFT 6
|
|
#define BETA_SCALE (1u<<BETA_SHIFT)
|
|
#define BETA_MIN (BETA_SCALE/8) /* 0.125 */
|
|
#define BETA_MAX (BETA_SCALE/2) /* 0.5 */
|
|
#define BETA_BASE BETA_MAX
|
|
|
|
static int win_thresh __read_mostly = 15;
|
|
module_param(win_thresh, int, 0);
|
|
MODULE_PARM_DESC(win_thresh, "Window threshold for starting adaptive sizing");
|
|
|
|
static int theta __read_mostly = 5;
|
|
module_param(theta, int, 0);
|
|
MODULE_PARM_DESC(theta, "# of fast RTT's before full growth");
|
|
|
|
/* TCP Illinois Parameters */
|
|
struct illinois {
|
|
u64 sum_rtt; /* sum of rtt's measured within last rtt */
|
|
u16 cnt_rtt; /* # of rtts measured within last rtt */
|
|
u32 base_rtt; /* min of all rtt in usec */
|
|
u32 max_rtt; /* max of all rtt in usec */
|
|
u32 end_seq; /* right edge of current RTT */
|
|
u32 alpha; /* Additive increase */
|
|
u32 beta; /* Muliplicative decrease */
|
|
u16 acked; /* # packets acked by current ACK */
|
|
u8 rtt_above; /* average rtt has gone above threshold */
|
|
u8 rtt_low; /* # of rtts measurements below threshold */
|
|
};
|
|
|
|
static void rtt_reset(struct sock *sk)
|
|
{
|
|
struct tcp_sock *tp = tcp_sk(sk);
|
|
struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
ca->end_seq = tp->snd_nxt;
|
|
ca->cnt_rtt = 0;
|
|
ca->sum_rtt = 0;
|
|
|
|
/* TODO: age max_rtt? */
|
|
}
|
|
|
|
static void tcp_illinois_init(struct sock *sk)
|
|
{
|
|
struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
ca->alpha = ALPHA_MAX;
|
|
ca->beta = BETA_BASE;
|
|
ca->base_rtt = 0x7fffffff;
|
|
ca->max_rtt = 0;
|
|
|
|
ca->acked = 0;
|
|
ca->rtt_low = 0;
|
|
ca->rtt_above = 0;
|
|
|
|
rtt_reset(sk);
|
|
}
|
|
|
|
/* Measure RTT for each ack. */
|
|
static void tcp_illinois_acked(struct sock *sk, u32 pkts_acked, s32 rtt)
|
|
{
|
|
struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
ca->acked = pkts_acked;
|
|
|
|
/* dup ack, no rtt sample */
|
|
if (rtt < 0)
|
|
return;
|
|
|
|
/* ignore bogus values, this prevents wraparound in alpha math */
|
|
if (rtt > RTT_MAX)
|
|
rtt = RTT_MAX;
|
|
|
|
/* keep track of minimum RTT seen so far */
|
|
if (ca->base_rtt > rtt)
|
|
ca->base_rtt = rtt;
|
|
|
|
/* and max */
|
|
if (ca->max_rtt < rtt)
|
|
ca->max_rtt = rtt;
|
|
|
|
++ca->cnt_rtt;
|
|
ca->sum_rtt += rtt;
|
|
}
|
|
|
|
/* Maximum queuing delay */
|
|
static inline u32 max_delay(const struct illinois *ca)
|
|
{
|
|
return ca->max_rtt - ca->base_rtt;
|
|
}
|
|
|
|
/* Average queuing delay */
|
|
static inline u32 avg_delay(const struct illinois *ca)
|
|
{
|
|
u64 t = ca->sum_rtt;
|
|
|
|
do_div(t, ca->cnt_rtt);
|
|
return t - ca->base_rtt;
|
|
}
|
|
|
|
/*
|
|
* Compute value of alpha used for additive increase.
|
|
* If small window then use 1.0, equivalent to Reno.
|
|
*
|
|
* For larger windows, adjust based on average delay.
|
|
* A. If average delay is at minimum (we are uncongested),
|
|
* then use large alpha (10.0) to increase faster.
|
|
* B. If average delay is at maximum (getting congested)
|
|
* then use small alpha (0.3)
|
|
*
|
|
* The result is a convex window growth curve.
|
|
*/
|
|
static u32 alpha(struct illinois *ca, u32 da, u32 dm)
|
|
{
|
|
u32 d1 = dm / 100; /* Low threshold */
|
|
|
|
if (da <= d1) {
|
|
/* If never got out of low delay zone, then use max */
|
|
if (!ca->rtt_above)
|
|
return ALPHA_MAX;
|
|
|
|
/* Wait for 5 good RTT's before allowing alpha to go alpha max.
|
|
* This prevents one good RTT from causing sudden window increase.
|
|
*/
|
|
if (++ca->rtt_low < theta)
|
|
return ca->alpha;
|
|
|
|
ca->rtt_low = 0;
|
|
ca->rtt_above = 0;
|
|
return ALPHA_MAX;
|
|
}
|
|
|
|
ca->rtt_above = 1;
|
|
|
|
/*
|
|
* Based on:
|
|
*
|
|
* (dm - d1) amin amax
|
|
* k1 = -------------------
|
|
* amax - amin
|
|
*
|
|
* (dm - d1) amin
|
|
* k2 = ---------------- - d1
|
|
* amax - amin
|
|
*
|
|
* k1
|
|
* alpha = ----------
|
|
* k2 + da
|
|
*/
|
|
|
|
dm -= d1;
|
|
da -= d1;
|
|
return (dm * ALPHA_MAX) /
|
|
(dm + (da * (ALPHA_MAX - ALPHA_MIN)) / ALPHA_MIN);
|
|
}
|
|
|
|
/*
|
|
* Beta used for multiplicative decrease.
|
|
* For small window sizes returns same value as Reno (0.5)
|
|
*
|
|
* If delay is small (10% of max) then beta = 1/8
|
|
* If delay is up to 80% of max then beta = 1/2
|
|
* In between is a linear function
|
|
*/
|
|
static u32 beta(u32 da, u32 dm)
|
|
{
|
|
u32 d2, d3;
|
|
|
|
d2 = dm / 10;
|
|
if (da <= d2)
|
|
return BETA_MIN;
|
|
|
|
d3 = (8 * dm) / 10;
|
|
if (da >= d3 || d3 <= d2)
|
|
return BETA_MAX;
|
|
|
|
/*
|
|
* Based on:
|
|
*
|
|
* bmin d3 - bmax d2
|
|
* k3 = -------------------
|
|
* d3 - d2
|
|
*
|
|
* bmax - bmin
|
|
* k4 = -------------
|
|
* d3 - d2
|
|
*
|
|
* b = k3 + k4 da
|
|
*/
|
|
return (BETA_MIN * d3 - BETA_MAX * d2 + (BETA_MAX - BETA_MIN) * da)
|
|
/ (d3 - d2);
|
|
}
|
|
|
|
/* Update alpha and beta values once per RTT */
|
|
static void update_params(struct sock *sk)
|
|
{
|
|
struct tcp_sock *tp = tcp_sk(sk);
|
|
struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
if (tp->snd_cwnd < win_thresh) {
|
|
ca->alpha = ALPHA_BASE;
|
|
ca->beta = BETA_BASE;
|
|
} else if (ca->cnt_rtt > 0) {
|
|
u32 dm = max_delay(ca);
|
|
u32 da = avg_delay(ca);
|
|
|
|
ca->alpha = alpha(ca, da, dm);
|
|
ca->beta = beta(da, dm);
|
|
}
|
|
|
|
rtt_reset(sk);
|
|
}
|
|
|
|
/*
|
|
* In case of loss, reset to default values
|
|
*/
|
|
static void tcp_illinois_state(struct sock *sk, u8 new_state)
|
|
{
|
|
struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
if (new_state == TCP_CA_Loss) {
|
|
ca->alpha = ALPHA_BASE;
|
|
ca->beta = BETA_BASE;
|
|
ca->rtt_low = 0;
|
|
ca->rtt_above = 0;
|
|
rtt_reset(sk);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Increase window in response to successful acknowledgment.
|
|
*/
|
|
static void tcp_illinois_cong_avoid(struct sock *sk, u32 ack, u32 in_flight)
|
|
{
|
|
struct tcp_sock *tp = tcp_sk(sk);
|
|
struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
if (after(ack, ca->end_seq))
|
|
update_params(sk);
|
|
|
|
/* RFC2861 only increase cwnd if fully utilized */
|
|
if (!tcp_is_cwnd_limited(sk, in_flight))
|
|
return;
|
|
|
|
/* In slow start */
|
|
if (tp->snd_cwnd <= tp->snd_ssthresh)
|
|
tcp_slow_start(tp);
|
|
|
|
else {
|
|
u32 delta;
|
|
|
|
/* snd_cwnd_cnt is # of packets since last cwnd increment */
|
|
tp->snd_cwnd_cnt += ca->acked;
|
|
ca->acked = 1;
|
|
|
|
/* This is close approximation of:
|
|
* tp->snd_cwnd += alpha/tp->snd_cwnd
|
|
*/
|
|
delta = (tp->snd_cwnd_cnt * ca->alpha) >> ALPHA_SHIFT;
|
|
if (delta >= tp->snd_cwnd) {
|
|
tp->snd_cwnd = min(tp->snd_cwnd + delta / tp->snd_cwnd,
|
|
(u32) tp->snd_cwnd_clamp);
|
|
tp->snd_cwnd_cnt = 0;
|
|
}
|
|
}
|
|
}
|
|
|
|
static u32 tcp_illinois_ssthresh(struct sock *sk)
|
|
{
|
|
struct tcp_sock *tp = tcp_sk(sk);
|
|
struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
/* Multiplicative decrease */
|
|
return max(tp->snd_cwnd - ((tp->snd_cwnd * ca->beta) >> BETA_SHIFT), 2U);
|
|
}
|
|
|
|
|
|
/* Extract info for Tcp socket info provided via netlink. */
|
|
static void tcp_illinois_info(struct sock *sk, u32 ext,
|
|
struct sk_buff *skb)
|
|
{
|
|
const struct illinois *ca = inet_csk_ca(sk);
|
|
|
|
if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
|
|
struct tcpvegas_info info = {
|
|
.tcpv_enabled = 1,
|
|
.tcpv_rttcnt = ca->cnt_rtt,
|
|
.tcpv_minrtt = ca->base_rtt,
|
|
};
|
|
|
|
if (info.tcpv_rttcnt > 0) {
|
|
u64 t = ca->sum_rtt;
|
|
|
|
do_div(t, info.tcpv_rttcnt);
|
|
info.tcpv_rtt = t;
|
|
}
|
|
nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
|
|
}
|
|
}
|
|
|
|
static struct tcp_congestion_ops tcp_illinois __read_mostly = {
|
|
.flags = TCP_CONG_RTT_STAMP,
|
|
.init = tcp_illinois_init,
|
|
.ssthresh = tcp_illinois_ssthresh,
|
|
.min_cwnd = tcp_reno_min_cwnd,
|
|
.cong_avoid = tcp_illinois_cong_avoid,
|
|
.set_state = tcp_illinois_state,
|
|
.get_info = tcp_illinois_info,
|
|
.pkts_acked = tcp_illinois_acked,
|
|
|
|
.owner = THIS_MODULE,
|
|
.name = "illinois",
|
|
};
|
|
|
|
static int __init tcp_illinois_register(void)
|
|
{
|
|
BUILD_BUG_ON(sizeof(struct illinois) > ICSK_CA_PRIV_SIZE);
|
|
return tcp_register_congestion_control(&tcp_illinois);
|
|
}
|
|
|
|
static void __exit tcp_illinois_unregister(void)
|
|
{
|
|
tcp_unregister_congestion_control(&tcp_illinois);
|
|
}
|
|
|
|
module_init(tcp_illinois_register);
|
|
module_exit(tcp_illinois_unregister);
|
|
|
|
MODULE_AUTHOR("Stephen Hemminger, Shao Liu");
|
|
MODULE_LICENSE("GPL");
|
|
MODULE_DESCRIPTION("TCP Illinois");
|
|
MODULE_VERSION("1.0");
|