git: b1d63265ac39 - main - Flush remaining routes from the routing table during VNET shutdown.

[Alexander V. Chernikov]

The branch main has been updated by melifaro:

URL: https://cgit.FreeBSD.org/src/commit/?id=b1d63265ac399112b3bca36c3d75df1a3c2c8102

commit b1d63265ac399112b3bca36c3d75df1a3c2c8102
Author:     Alexander V. Chernikov <melifaro at FreeBSD.org>
AuthorDate: 2021-03-08 21:35:41 +0000
Commit:     Alexander V. Chernikov <melifaro at FreeBSD.org>
CommitDate: 2021-03-10 21:10:14 +0000

    Flush remaining routes from the routing table during VNET shutdown.
    
    Summary:
    This fixes rtentry leak for the cloned interfaces created inside the
     VNET.
    
    PR:     253998
    Reported by:    rashey at superbox.pl
    MFC after:      3 days
    
    Loopback teardown order is `SI_SUB_INIT_IF`, which happens after `SI_SUB_PROTO_DOMAIN` (route table teardown).
    Thus, any route table operations are too late to schedule.
    As the intent of the vnet teardown procedures to minimise the amount of effort by doing global cleanups instead of per-interface ones, address this by adding a relatively light-weight routing table cleanup function, `rib_flush_routes()`.
    It removes all remaining routes from the routing table and schedules the deletion, which will happen later, when `rtables_destroy()` waits for the current epoch to finish.
    
    Test Plan:
    ```
    set_skip:set_skip_group_lo  ->  passed  [0.053s]
    tail -n 200 /var/log/messages | grep rtentry
    ```
    
    Reviewers: #network, kp, bz
    
    Reviewed By: kp
    
    Subscribers: imp, ae
    
    Differential Revision: https://reviews.freebsd.org/D29116
---
 sys/net/route.c           | 15 ---------------
 sys/net/route.h           |  2 +-
 sys/net/route/route_ctl.c | 36 ++++++++++++++++++++++++++++++++++++
 sys/netinet/ip_input.c    |  6 +-----
 sys/netinet6/ip6_input.c  |  5 +++--
 5 files changed, 41 insertions(+), 23 deletions(-)

diff --git a/sys/net/route.c b/sys/net/route.c
index d2b405fafcbf..2416aa9a983f 100644
--- a/sys/net/route.c
+++ b/sys/net/route.c
@@ -493,21 +493,6 @@ rt_ifdelroute(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
 	return (1);
 }
 
-/*
- * Delete all remaining routes using this interface
- * Unfortuneatly the only way to do this is to slog through
- * the entire routing table looking for routes which point
- * to this interface...oh well...
- */
-void
-rt_flushifroutes_af(struct ifnet *ifp, int af)
-{
-	KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
-	    __func__, af, AF_MAX));
-
-	rib_foreach_table_walk_del(af, rt_ifdelroute, ifp);
-}
-
 void
 rt_flushifroutes(struct ifnet *ifp)
 {
diff --git a/sys/net/route.h b/sys/net/route.h
index ab6e1aabc5ae..3fdca303596e 100644
--- a/sys/net/route.h
+++ b/sys/net/route.h
@@ -429,7 +429,6 @@ struct sockaddr *rtsock_fix_netmask(const struct sockaddr *dst,
 
 void	rt_updatemtu(struct ifnet *);
 
-void	rt_flushifroutes_af(struct ifnet *, int);
 void	rt_flushifroutes(struct ifnet *ifp);
 
 /* XXX MRT NEW VERSIONS THAT USE FIBs
@@ -442,6 +441,7 @@ int	rib_lookup_info(uint32_t, const struct sockaddr *, uint32_t, uint32_t,
 void	rib_free_info(struct rt_addrinfo *info);
 
 /* New API */
+void rib_flush_routes_family(int family);
 struct nhop_object *rib_lookup(uint32_t fibnum, const struct sockaddr *dst,
 	    uint32_t flags, uint32_t flowid);
 #endif
diff --git a/sys/net/route/route_ctl.c b/sys/net/route/route_ctl.c
index e07a94295b89..d9fd1aa18c92 100644
--- a/sys/net/route/route_ctl.c
+++ b/sys/net/route/route_ctl.c
@@ -1328,6 +1328,42 @@ rib_walk_del(u_int fibnum, int family, rib_filter_f_t *filter_f, void *arg, bool
 	NET_EPOCH_EXIT(et);
 }
 
+static int
+rt_delete_unconditional(struct radix_node *rn, void *arg)
+{
+	struct rtentry *rt = RNTORT(rn);
+	struct rib_head *rnh = (struct rib_head *)arg;
+
+	rn = rnh->rnh_deladdr(rt_key(rt), rt_mask(rt), &rnh->head);
+	if (RNTORT(rn) == rt)
+		rtfree(rt);
+
+	return (0);
+}
+
+/*
+ * Removes all routes from the routing table without executing notifications.
+ * rtentres will be removed after the end of a current epoch.
+ */
+static void
+rib_flush_routes(struct rib_head *rnh)
+{
+	RIB_WLOCK(rnh);
+	rnh->rnh_walktree(&rnh->head, rt_delete_unconditional, rnh);
+	RIB_WUNLOCK(rnh);
+}
+
+void
+rib_flush_routes_family(int family)
+{
+	struct rib_head *rnh;
+
+	for (uint32_t fibnum = 0; fibnum < rt_numfibs; fibnum++) {
+		if ((rnh = rt_tables_get_rnh(fibnum, family)) != NULL)
+			rib_flush_routes(rnh);
+	}
+}
+
 static void
 rib_notify(struct rib_head *rnh, enum rib_subscription_type type,
     struct rib_cmd_info *rc)
diff --git a/sys/netinet/ip_input.c b/sys/netinet/ip_input.c
index be21decff6cb..a85f8ac7b567 100644
--- a/sys/netinet/ip_input.c
+++ b/sys/netinet/ip_input.c
@@ -379,7 +379,6 @@ ip_init(void)
 static void
 ip_destroy(void *unused __unused)
 {
-	struct ifnet *ifp;
 	int error;
 
 #ifdef	RSS
@@ -405,10 +404,7 @@ ip_destroy(void *unused __unused)
 	in_ifscrub_all();
 
 	/* Make sure the IPv4 routes are gone as well. */
-	IFNET_RLOCK();
-	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
-		rt_flushifroutes_af(ifp, AF_INET);
-	IFNET_RUNLOCK();
+	rib_flush_routes_family(AF_INET);
 
 	/* Destroy IP reassembly queue. */
 	ipreass_destroy();
diff --git a/sys/netinet6/ip6_input.c b/sys/netinet6/ip6_input.c
index 80e5acc62548..9ea578f88417 100644
--- a/sys/netinet6/ip6_input.c
+++ b/sys/netinet6/ip6_input.c
@@ -386,11 +386,12 @@ ip6_destroy(void *unused __unused)
 		/* IF_ADDR_UNLOCK(ifp); */
 		in6_ifdetach_destroy(ifp);
 		mld_domifdetach(ifp);
-		/* Make sure any routes are gone as well. */
-		rt_flushifroutes_af(ifp, AF_INET6);
 	}
 	IFNET_RUNLOCK();
 
+	/* Make sure any routes are gone as well. */
+	rib_flush_routes_family(AF_INET6);
+
 	frag6_destroy();
 	nd6_destroy();
 	in6_ifattach_destroy();