Provide new socket option TCP_CCALGOOPT, which stands for TCP congestion

control algorithm options.  The argument is variable length and is opaque
to TCP, forwarded directly to the algorithm's ctl_output method.

Provide new includes directory netinet/cc, where algorithm specific
headers can be installed.

The new API doesn't yet have any in tree consumers.

The original code written by lstewart.
Reviewed by:	rrs, emax
Sponsored by:	Netflix
Differential Revision:	https://reviews.freebsd.org/D711

3 files changed, 31 insertions, 1 deletions

diff --git a/sys/netinet/tcp.h b/sys/netinet/tcp.h
index bfc8073..4703810 100644
--- a/sys/netinet/tcp.h
+++ b/sys/netinet/tcp.h
@@ -165,6 +165,7 @@ struct tcphdr {
 #define TCP_MD5SIG	16	/* use MD5 digests (RFC2385) */
 #define	TCP_INFO	32	/* retrieve tcp_info structure */
 #define	TCP_CONGESTION	64	/* get/set congestion control algorithm */
+#define	TCP_CCALGOOPT	65	/* get/set cc algorithm specific options */
 #define	TCP_KEEPINIT	128	/* N, time to establish connection */
 #define	TCP_KEEPIDLE	256	/* L,N,X start keeplives after this period */
 #define	TCP_KEEPINTVL	512	/* L,N interval between keepalives */
diff --git a/sys/netinet/tcp_cc.h b/sys/netinet/tcp_cc.h
index 4a2b0c8..d90cd19 100644
--- a/sys/netinet/tcp_cc.h
+++ b/sys/netinet/tcp_cc.h
@@ -151,6 +151,9 @@ struct cc_algo {
 	/* Called for an additional ECN processing apart from RFC3168. */
 	void	(*ecnpkt_handler)(struct cc_var *ccv);
 
+	/* Called for {get|set}sockopt() on a TCP socket with TCP_CCALGOOPT. */
+	int     (*ctl_output)(struct cc_var *, struct sockopt *, void *);
+
 	STAILQ_ENTRY (cc_algo) entries;
 };
 
diff --git a/sys/netinet/tcp_usrreq.c b/sys/netinet/tcp_usrreq.c
index 29e92b2..4b3150b 100644
--- a/sys/netinet/tcp_usrreq.c
+++ b/sys/netinet/tcp_usrreq.c
@@ -1480,7 +1480,33 @@ tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp
 	struct	tcp_info ti;
 	struct cc_algo *algo;
 	char	*buf;
-	
+
+	/*
+	 * For TCP_CCALGOOPT forward the control to CC module, for both
+	 * SOPT_SET and SOPT_GET.
+	 */
+	switch (sopt->sopt_name) {
+	case TCP_CCALGOOPT:
+		INP_WUNLOCK(inp);
+		buf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
+		error = sooptcopyin(sopt, buf, sopt->sopt_valsize,
+		    sopt->sopt_valsize);
+		if (error) {
+			free(buf, M_TEMP);
+			return (error);
+		}
+		INP_WLOCK_RECHECK(inp);
+		if (CC_ALGO(tp)->ctl_output != NULL)
+			error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, buf);
+		else
+			error = ENOENT;
+		INP_WUNLOCK(inp);
+		if (error == 0 && sopt->sopt_dir == SOPT_GET)
+			error = sooptcopyout(sopt, buf, sopt->sopt_valsize);
+		free(buf, M_TEMP);
+		return (error);
+	}
+
 	switch (sopt->sopt_dir) {
 	case SOPT_SET:
 		switch (sopt->sopt_name) {