Make the CC framework more VIMAGE friendly by adding the machinery to allow

vnets to select their own default CC algorithm independent of each other and the
base system. If the base system or a vnet has set a default which gets unloaded,
we reset that netstack's default to NewReno.

Sponsored by:	FreeBSD Foundation
Tested by:	Mikolaj Golub <to.my.trociny at gmail com>
Reviewed by:	bz (briefly)
MFC after:	3 months

2 files changed, 34 insertions, 32 deletions

diff --git a/sys/netinet/cc.h b/sys/netinet/cc.h
index 0d1dab7..aaa1d67 100644
--- a/sys/netinet/cc.h
+++ b/sys/netinet/cc.h
@@ -58,6 +58,10 @@ extern STAILQ_HEAD(cc_head, cc_algo) cc_list;
 extern const int tcprexmtthresh;
 extern struct cc_algo newreno_cc_algo;
 
+/* Per-netstack bits. */
+VNET_DECLARE(struct cc_algo *, default_cc_ptr);
+#define	V_default_cc_ptr VNET(default_cc_ptr)
+
 /* Define the new net.inet.tcp.cc sysctl tree. */
 SYSCTL_DECL(_net_inet_tcp_cc);
 
@@ -146,7 +150,7 @@ struct cc_algo {
 #define	CC_DATA(tp)	((tp)->ccv->cc_data)
 
 /* Macro to obtain the system default CC algo's struct ptr. */
-#define	CC_DEFAULT()	STAILQ_FIRST(&cc_list)
+#define	CC_DEFAULT()	V_default_cc_ptr
 
 extern struct rwlock cc_list_lock;
 #define	CC_LIST_LOCK_INIT()	rw_init(&cc_list_lock, "cc_list")
@@ -155,6 +159,6 @@ extern struct rwlock cc_list_lock;
 #define	CC_LIST_RUNLOCK()	rw_runlock(&cc_list_lock)
 #define	CC_LIST_WLOCK()		rw_wlock(&cc_list_lock)
 #define	CC_LIST_WUNLOCK()	rw_wunlock(&cc_list_lock)
-#define	CC_LIST_WLOCK_ASSERT()	rw_assert(&cc_list_lock, RA_WLOCKED)
+#define	CC_LIST_LOCK_ASSERT()	rw_assert(&cc_list_lock, RA_LOCKED)
 
 #endif /* _NETINET_CC_H_ */
diff --git a/sys/netinet/cc/cc.c b/sys/netinet/cc/cc.c
index dc50cda..f075327 100644
--- a/sys/netinet/cc/cc.c
+++ b/sys/netinet/cc/cc.c
@@ -81,24 +81,7 @@ struct cc_head cc_list = STAILQ_HEAD_INITIALIZER(cc_list);
 /* Protects the cc_list TAILQ. */
 struct rwlock cc_list_lock;
 
-/*
- * Set the default CC algorithm to new_default. The default is identified
- * by being the first element in the cc_list TAILQ.
- */
-static void
-cc_set_default(struct cc_algo *new_default)
-{
-	CC_LIST_WLOCK_ASSERT();
-
-	/*
-	 * Make the requested system default CC algorithm the first element in
-	 * the list if it isn't already.
-	 */
-	if (new_default != CC_DEFAULT()) {
-		STAILQ_REMOVE(&cc_list, new_default, cc_algo, entries);
-		STAILQ_INSERT_HEAD(&cc_list, new_default, entries);
-	}
-}
+VNET_DEFINE(struct cc_algo *, default_cc_ptr) = &newreno_cc_algo;
 
 /*
  * Sysctl handler to show and change the default CC algorithm.
@@ -120,15 +103,15 @@ cc_default_algo(SYSCTL_HANDLER_ARGS)
 		err = sysctl_handle_string(oidp, default_cc, 1, req);
 	} else {
 		/* Find algo with specified name and set it to default. */
-		CC_LIST_WLOCK();
+		CC_LIST_RLOCK();
 		STAILQ_FOREACH(funcs, &cc_list, entries) {
 			if (strncmp((char *)req->newptr, funcs->name,
 			    TCP_CA_NAME_MAX) == 0) {
 				found = 1;
-				cc_set_default(funcs);
+				V_default_cc_ptr = funcs;
 			}
 		}
-		CC_LIST_WUNLOCK();
+		CC_LIST_RUNLOCK();
 
 		if (!found)
 			err = ESRCH;
@@ -173,6 +156,28 @@ cc_list_available(SYSCTL_HANDLER_ARGS)
 }
 
 /*
+ * Reset the default CC algo to NewReno for any netstack which is using the algo
+ * that is about to go away as its default.
+ */
+static void
+cc_checkreset_default(struct cc_algo *remove_cc)
+{
+	VNET_ITERATOR_DECL(vnet_iter);
+
+	CC_LIST_LOCK_ASSERT();
+
+	VNET_LIST_RLOCK_NOSLEEP();
+	VNET_FOREACH(vnet_iter) {
+		CURVNET_SET(vnet_iter);
+		if (strncmp(CC_DEFAULT()->name, remove_cc->name,
+		    TCP_CA_NAME_MAX) == 0)
+			V_default_cc_ptr = &newreno_cc_algo;
+		CURVNET_RESTORE();
+	}
+	VNET_LIST_RUNLOCK_NOSLEEP();
+}
+
+/*
  * Initialise CC subsystem on system boot.
  */
 static void
@@ -201,14 +206,7 @@ cc_deregister_algo(struct cc_algo *remove_cc)
 	CC_LIST_WLOCK();
 	STAILQ_FOREACH_SAFE(funcs, &cc_list, entries, tmpfuncs) {
 		if (funcs == remove_cc) {
-			/*
-			 * If we're removing the current system default,
-			 * reset the default to newreno.
-			 */
-			if (strncmp(CC_DEFAULT()->name, remove_cc->name,
-			    TCP_CA_NAME_MAX) == 0)
-				cc_set_default(&newreno_cc_algo);
-
+			cc_checkreset_default(remove_cc);
 			STAILQ_REMOVE(&cc_list, funcs, cc_algo, entries);
 			err = 0;
 			break;
@@ -302,7 +300,7 @@ SYSINIT(cc, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, cc_init, NULL);
 SYSCTL_NODE(_net_inet_tcp, OID_AUTO, cc, CTLFLAG_RW, NULL,
     "congestion control related settings");
 
-SYSCTL_PROC(_net_inet_tcp_cc, OID_AUTO, algorithm, CTLTYPE_STRING|CTLFLAG_RW,
+SYSCTL_VNET_PROC(_net_inet_tcp_cc, OID_AUTO, algorithm, CTLTYPE_STRING|CTLFLAG_RW,
     NULL, 0, cc_default_algo, "A", "default congestion control algorithm");
 
 SYSCTL_PROC(_net_inet_tcp_cc, OID_AUTO, available, CTLTYPE_STRING|CTLFLAG_RD,