tcp: use dctcp if enabled on the route to the initiator

Currently, the following case doesn't use DCTCP, even if it should:
A responder has f.e. Cubic as system wide default, but for a specific
route to the initiating host, DCTCP is being set in RTAX_CC_ALGO. The
initiating host then uses DCTCP as congestion control, but since the
initiator sets ECT(0), tcp_ecn_create_request() doesn't set ecn_ok,
and we have to fall back to Reno after 3WHS completes.

We were thinking on how to solve this in a minimal, non-intrusive
way without bloating tcp_ecn_create_request() needlessly: lets cache
the CA ecn option flag in RTAX_FEATURES. In other words, when ECT(0)
is set on the SYN packet, set ecn_ok=1 iff route RTAX_FEATURES
contains the unexposed (internal-only) DST_FEATURE_ECN_CA. This allows
to only do a single metric feature lookup inside tcp_ecn_create_request().

Joint work with Florian Westphal.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/include/net/dst.h b/include/net/dst.h
index 4c4801645371d12c5fadc5427c8a029c0ba1f1e3..9261d928303d475a8ef2772144f9f70e0febda49 100644 (file)
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -207,6 +207,12 @@ static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
                p[metric-1] = val;
 }
 
+/* Kernel-internal feature bits that are unallocated in user space. */
+#define DST_FEATURE_ECN_CA     (1 << 31)
+
+#define DST_FEATURE_MASK       (DST_FEATURE_ECN_CA)
+#define DST_FEATURE_ECN_MASK   (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
+
 static inline u32
 dst_feature(const struct dst_entry *dst, u32 feature)
 {

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 4a7b03947a38bc14bc366b5907e8821726c685c5..0cab28cd43a9067e731af2d0b2557172e63fe5e9 100644 (file)
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -888,7 +888,7 @@ void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked);
 extern struct tcp_congestion_ops tcp_reno;
 
 struct tcp_congestion_ops *tcp_ca_find_key(u32 key);
-u32 tcp_ca_get_key_by_name(const char *name);
+u32 tcp_ca_get_key_by_name(const char *name, bool *ecn_ca);
 #ifdef CONFIG_INET
 char *tcp_ca_get_name_by_key(u32 key, char *buffer);
 #else

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 788ceed394636e4a3c3c38e4fbfe383dca8e48df..a466821d1441f22ac244eb3366f53f859a0aac29 100644 (file)
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -678,6 +678,12 @@ int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
                                        continue;
                                if (nla_put_string(skb, i + 1, name))
                                        goto nla_put_failure;
+                       } else if (i == RTAX_FEATURES - 1) {
+                               u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
+
+                               BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
+                               if (nla_put_u32(skb, i + 1, user_features))
+                                       goto nla_put_failure;
                        } else {
                                if (nla_put_u32(skb, i + 1, metrics[i]))
                                        goto nla_put_failure;

diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index 115a08e70d43243215feb8a8f7759493cfa969dd..992a9597daf80d7d734d8e0245c8410dc334511b 100644 (file)
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -879,6 +879,7 @@ static bool fib_valid_prefsrc(struct fib_config *cfg, __be32 fib_prefsrc)
 static int
 fib_convert_metrics(struct fib_info *fi, const struct fib_config *cfg)
 {
+       bool ecn_ca = false;
        struct nlattr *nla;
        int remaining;
 
@@ -898,7 +899,7 @@ fib_convert_metrics(struct fib_info *fi, const struct fib_config *cfg)
                        char tmp[TCP_CA_NAME_MAX];
 
                        nla_strlcpy(tmp, nla, sizeof(tmp));
-                       val = tcp_ca_get_key_by_name(tmp);
+                       val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
                        if (val == TCP_CA_UNSPEC)
                                return -EINVAL;
                } else {
@@ -913,6 +914,9 @@ fib_convert_metrics(struct fib_info *fi, const struct fib_config *cfg)
                fi->fib_metrics[type - 1] = val;
        }
 
+       if (ecn_ca)
+               fi->fib_metrics[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
+
        return 0;
 }
 

diff --git a/net/ipv4/tcp_cong.c b/net/ipv4/tcp_cong.c
index a2ed23c595cf185cadbebcdf19e801012a64250a..93c4dc3ab23fdf224b48247584787f90916a1af5 100644 (file)
--- a/net/ipv4/tcp_cong.c
+++ b/net/ipv4/tcp_cong.c
@@ -114,16 +114,19 @@ void tcp_unregister_congestion_control(struct tcp_congestion_ops *ca)
 }
 EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control);
 
-u32 tcp_ca_get_key_by_name(const char *name)
+u32 tcp_ca_get_key_by_name(const char *name, bool *ecn_ca)
 {
        const struct tcp_congestion_ops *ca;
-       u32 key;
+       u32 key = TCP_CA_UNSPEC;
 
        might_sleep();
 
        rcu_read_lock();
        ca = __tcp_ca_find_autoload(name);
-       key = ca ? ca->key : TCP_CA_UNSPEC;
+       if (ca) {
+               key = ca->key;
+               *ecn_ca = ca->flags & TCP_CONG_NEEDS_ECN;
+       }
        rcu_read_unlock();
 
        return key;

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index dc08e235266523b00faa677811bad3e7c2fe44a0..a8f515bb19c4bbd2d379c6f5c29f9a3d02e8bc96 100644 (file)
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -6003,14 +6003,17 @@ static void tcp_ecn_create_request(struct request_sock *req,
        const struct net *net = sock_net(listen_sk);
        bool th_ecn = th->ece && th->cwr;
        bool ect, ecn_ok;
+       u32 ecn_ok_dst;
 
        if (!th_ecn)
                return;
 
        ect = !INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield);
-       ecn_ok = net->ipv4.sysctl_tcp_ecn || dst_feature(dst, RTAX_FEATURE_ECN);
+       ecn_ok_dst = dst_feature(dst, DST_FEATURE_ECN_MASK);
+       ecn_ok = net->ipv4.sysctl_tcp_ecn || ecn_ok_dst;
 
-       if ((!ect && ecn_ok) || tcp_ca_needs_ecn(listen_sk))
+       if ((!ect && ecn_ok) || tcp_ca_needs_ecn(listen_sk) ||
+           (ecn_ok_dst & DST_FEATURE_ECN_CA))
                inet_rsk(req)->ecn_ok = 1;
 }
 

diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index 8771530df45ea0041d5d7a6402dc6de2ee5d0d92..f45cac6f83563977186732adacedcec55a6b3ce7 100644 (file)
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -1698,6 +1698,7 @@ out:
 static int ip6_convert_metrics(struct mx6_config *mxc,
                               const struct fib6_config *cfg)
 {
+       bool ecn_ca = false;
        struct nlattr *nla;
        int remaining;
        u32 *mp;
@@ -1722,7 +1723,7 @@ static int ip6_convert_metrics(struct mx6_config *mxc,
                        char tmp[TCP_CA_NAME_MAX];
 
                        nla_strlcpy(tmp, nla, sizeof(tmp));
-                       val = tcp_ca_get_key_by_name(tmp);
+                       val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
                        if (val == TCP_CA_UNSPEC)
                                goto err;
                } else {
@@ -1735,8 +1736,12 @@ static int ip6_convert_metrics(struct mx6_config *mxc,
                __set_bit(type - 1, mxc->mx_valid);
        }
 
-       mxc->mx = mp;
+       if (ecn_ca) {
+               __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
+               mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
+       }
 
+       mxc->mx = mp;
        return 0;
  err:
        kfree(mp);