summaryrefslogtreecommitdiffstats
path: root/net/ipv4/tcp_cong.c
diff options
context:
space:
mode:
authorStephen Hemminger <shemminger@osdl.org>2005-06-23 12:19:55 -0700
committerDavid S. Miller <davem@davemloft.net>2005-06-23 12:19:55 -0700
commit317a76f9a44b437d6301718f4e5d08bd93f98da7 (patch)
treecaeba9839dee264f59b035b81c3d13d6c61b638e /net/ipv4/tcp_cong.c
parenta8ad86f2dc46356f87be1327dabc18bdbda32f50 (diff)
downloadop-kernel-dev-317a76f9a44b437d6301718f4e5d08bd93f98da7.zip
op-kernel-dev-317a76f9a44b437d6301718f4e5d08bd93f98da7.tar.gz
[TCP]: Add pluggable congestion control algorithm infrastructure.
Allow TCP to have multiple pluggable congestion control algorithms. Algorithms are defined by a set of operations and can be built in or modules. The legacy "new RENO" algorithm is used as a starting point and fallback. Signed-off-by: Stephen Hemminger <shemminger@osdl.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/ipv4/tcp_cong.c')
-rw-r--r--net/ipv4/tcp_cong.c195
1 files changed, 195 insertions, 0 deletions
diff --git a/net/ipv4/tcp_cong.c b/net/ipv4/tcp_cong.c
new file mode 100644
index 0000000..665394a
--- /dev/null
+++ b/net/ipv4/tcp_cong.c
@@ -0,0 +1,195 @@
+/*
+ * Plugable TCP congestion control support and newReno
+ * congestion control.
+ * Based on ideas from I/O scheduler suport and Web100.
+ *
+ * Copyright (C) 2005 Stephen Hemminger <shemminger@osdl.org>
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/mm.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <net/tcp.h>
+
+static DEFINE_SPINLOCK(tcp_cong_list_lock);
+static LIST_HEAD(tcp_cong_list);
+
+/* Simple linear search, don't expect many entries! */
+static struct tcp_congestion_ops *tcp_ca_find(const char *name)
+{
+ struct tcp_congestion_ops *e;
+
+ list_for_each_entry(e, &tcp_cong_list, list) {
+ if (strcmp(e->name, name) == 0)
+ return e;
+ }
+
+ return NULL;
+}
+
+/*
+ * Attach new congestion control algorthim to the list
+ * of available options.
+ */
+int tcp_register_congestion_control(struct tcp_congestion_ops *ca)
+{
+ int ret = 0;
+
+ /* all algorithms must implement ssthresh and cong_avoid ops */
+ if (!ca->ssthresh || !ca->cong_avoid || !ca->min_cwnd) {
+ printk(KERN_ERR "TCP %s does not implement required ops\n",
+ ca->name);
+ return -EINVAL;
+ }
+
+ spin_lock(&tcp_cong_list_lock);
+ if (tcp_ca_find(ca->name)) {
+ printk(KERN_NOTICE "TCP %s already registered\n", ca->name);
+ ret = -EEXIST;
+ } else {
+ list_add_rcu(&ca->list, &tcp_cong_list);
+ printk(KERN_INFO "TCP %s registered\n", ca->name);
+ }
+ spin_unlock(&tcp_cong_list_lock);
+
+ return ret;
+}
+EXPORT_SYMBOL_GPL(tcp_register_congestion_control);
+
+/*
+ * Remove congestion control algorithm, called from
+ * the module's remove function. Module ref counts are used
+ * to ensure that this can't be done till all sockets using
+ * that method are closed.
+ */
+void tcp_unregister_congestion_control(struct tcp_congestion_ops *ca)
+{
+ spin_lock(&tcp_cong_list_lock);
+ list_del_rcu(&ca->list);
+ spin_unlock(&tcp_cong_list_lock);
+}
+EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control);
+
+/* Assign choice of congestion control. */
+void tcp_init_congestion_control(struct tcp_sock *tp)
+{
+ struct tcp_congestion_ops *ca;
+
+ rcu_read_lock();
+ list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
+ if (try_module_get(ca->owner)) {
+ tp->ca_ops = ca;
+ break;
+ }
+
+ }
+ rcu_read_unlock();
+
+ if (tp->ca_ops->init)
+ tp->ca_ops->init(tp);
+}
+
+/* Manage refcounts on socket close. */
+void tcp_cleanup_congestion_control(struct tcp_sock *tp)
+{
+ if (tp->ca_ops->release)
+ tp->ca_ops->release(tp);
+ module_put(tp->ca_ops->owner);
+}
+
+/* Used by sysctl to change default congestion control */
+int tcp_set_default_congestion_control(const char *name)
+{
+ struct tcp_congestion_ops *ca;
+ int ret = -ENOENT;
+
+ spin_lock(&tcp_cong_list_lock);
+ ca = tcp_ca_find(name);
+#ifdef CONFIG_KMOD
+ if (!ca) {
+ spin_unlock(&tcp_cong_list_lock);
+
+ request_module("tcp_%s", name);
+ spin_lock(&tcp_cong_list_lock);
+ ca = tcp_ca_find(name);
+ }
+#endif
+
+ if (ca) {
+ list_move(&ca->list, &tcp_cong_list);
+ ret = 0;
+ }
+ spin_unlock(&tcp_cong_list_lock);
+
+ return ret;
+}
+
+/* Get current default congestion control */
+void tcp_get_default_congestion_control(char *name)
+{
+ struct tcp_congestion_ops *ca;
+ /* We will always have reno... */
+ BUG_ON(list_empty(&tcp_cong_list));
+
+ rcu_read_lock();
+ ca = list_entry(tcp_cong_list.next, struct tcp_congestion_ops, list);
+ strncpy(name, ca->name, TCP_CA_NAME_MAX);
+ rcu_read_unlock();
+}
+
+/*
+ * TCP Reno congestion control
+ * This is special case used for fallback as well.
+ */
+/* This is Jacobson's slow start and congestion avoidance.
+ * SIGCOMM '88, p. 328.
+ */
+void tcp_reno_cong_avoid(struct tcp_sock *tp, u32 ack, u32 rtt, u32 in_flight,
+ int flag)
+{
+ if (in_flight < tp->snd_cwnd)
+ return;
+
+ if (tp->snd_cwnd <= tp->snd_ssthresh) {
+ /* In "safe" area, increase. */
+ if (tp->snd_cwnd < tp->snd_cwnd_clamp)
+ tp->snd_cwnd++;
+ } else {
+ /* In dangerous area, increase slowly.
+ * In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd
+ */
+ if (tp->snd_cwnd_cnt >= tp->snd_cwnd) {
+ if (tp->snd_cwnd < tp->snd_cwnd_clamp)
+ tp->snd_cwnd++;
+ tp->snd_cwnd_cnt = 0;
+ } else
+ tp->snd_cwnd_cnt++;
+ }
+}
+EXPORT_SYMBOL_GPL(tcp_reno_cong_avoid);
+
+/* Slow start threshold is half the congestion window (min 2) */
+u32 tcp_reno_ssthresh(struct tcp_sock *tp)
+{
+ return max(tp->snd_cwnd >> 1U, 2U);
+}
+EXPORT_SYMBOL_GPL(tcp_reno_ssthresh);
+
+/* Lower bound on congestion window. */
+u32 tcp_reno_min_cwnd(struct tcp_sock *tp)
+{
+ return tp->snd_ssthresh/2;
+}
+EXPORT_SYMBOL_GPL(tcp_reno_min_cwnd);
+
+struct tcp_congestion_ops tcp_reno = {
+ .name = "reno",
+ .owner = THIS_MODULE,
+ .ssthresh = tcp_reno_ssthresh,
+ .cong_avoid = tcp_reno_cong_avoid,
+ .min_cwnd = tcp_reno_min_cwnd,
+};
+
+EXPORT_SYMBOL_GPL(tcp_reno);
OpenPOWER on IntegriCloud