[PATCH net-next v3 0/2] Remove expired routes with a separated list of routes.

[PATCH net-next v3 0/2] Remove expired routes with a separated list of routes.

[Kui-Feng Lee]

FIB6 GC walks trees of fib6_tables to remove expired routes. Walking a tree
can be expensive if the number of routes in a table is big, even if most of
them are permanent. Checking routes in a separated list of routes having
expiration will avoid this potential issue.

Background
==========

The size of a Linux IPv6 routing table can become a big problem if not
managed appropriately.  Now, Linux has a garbage collector to remove
expired routes periodically.  However, this may lead to a situation in
which the routing path is blocked for a long period due to an
excessive number of routes.

For example, years ago, there is a commit c7bb4b89033b ("ipv6: tcp: drop
silly ICMPv6 packet too big messages") about "ICMPv6 Packet too big
messages". The root cause is that malicious ICMPv6 packets were sent back
for every small packet sent to them. These packets add routes with an
expiration time that prompts the GC to periodically check all routes in the
tables, including permanent ones.

Why Route Expires
=================

Users can add IPv6 routes with an expiration time manually. However,
the Neighbor Discovery protocol may also generate routes that can
expire.  For example, Router Advertisement (RA) messages may create a
default route with an expiration time. [RFC 4861] For IPv4, it is not
possible to set an expiration time for a route, and there is no RA, so
there is no need to worry about such issues.

Create Routes with Expires
==========================

You can create routes with expires with the  command.

For example,

    ip -6 route add 2001:b000:591::3 via fe80::5054:ff:fe12:3457 \ 
        dev enp0s3 expires 30

The route that has been generated will be deleted automatically in 30
seconds.

GC of FIB6
==========

The function called fib6_run_gc() is responsible for performing
garbage collection (GC) for the Linux IPv6 stack. It checks for the
expiration of every route by traversing the trees of routing
tables. The time taken to traverse a routing table increases with its
size. Holding the routing table lock during traversal is particularly
undesirable. Therefore, it is preferable to keep the lock for the
shortest possible duration.

Solution
========

The cause of the issue is keeping the routing table locked during the
traversal of large trees. To solve this problem, we can create a separate
list of routes that have expiration. This will prevent GC from checking
permanent routes.

Result
======

We conducted a test to measure the execution times of fib6_gc_timer_cb()
and observed that it enhances the GC of FIB6. During the test, we added
permanent routes with the following numbers: 1000, 3000, 6000, and
9000. Additionally, we added a route with an expiration time.

Here are the average execution times for the kernel without the patch.
 - 120020 ns with 1000 permanent routes
 - 308920 ns with 3000 ...
 - 581470 ns with 6000 ...
 - 855310 ns with 9000 ...

The kernel with the patch consistently takes around 14000 ns to execute,
regardless of the number of permanent routes that are installed.

Major changes from v2:

 - Remove unnecessary and incorrect sysctl restoring in the test case.

Major changes from v1:

 - Moved gc_link to avoid creating a hole in fib6_info.

 - Moved fib6_set_expires*() and fib6_clean_expires*() to the header
   file and inlined. And removed duplicated lines.

 - Added a test case.

---
v1: https://lore.kernel.org/all/20230710203609.520720-1-kuifeng@meta.com/
v2: https://lore.kernel.org/all/20230718180321.294721-1-kuifeng@meta.com/

Kui-Feng Lee (2):
  net/ipv6: Remove expired routes with a separated list of routes.
  selftests: fib_tests: Add a test case for IPv6 garbage collection

 include/net/ip6_fib.h                    | 65 +++++++++++++++++++-----
 net/ipv6/ip6_fib.c                       | 53 +++++++++++++++++--
 net/ipv6/route.c                         |  6 +--
 tools/testing/selftests/net/fib_tests.sh | 46 ++++++++++++++++-
 4 files changed, 149 insertions(+), 21 deletions(-)

-- 
2.34.1

[PATCH net-next v3 1/2] net/ipv6: Remove expired routes with a separated list of routes.

[Kui-Feng Lee]

FIB6 GC walks trees of fib6_tables to remove expired routes. Walking a tree
can be expensive if the number of routes in a table is big, even if most of
them are permanent. Checking routes in a separated list of routes having
expiration will avoid this potential issue.

Signed-off-by: Kui-Feng Lee <kuifeng@meta.com>
---
 include/net/ip6_fib.h | 65 ++++++++++++++++++++++++++++++++++---------
 net/ipv6/ip6_fib.c    | 53 ++++++++++++++++++++++++++++++++---
 net/ipv6/route.c      |  6 ++--
 3 files changed, 104 insertions(+), 20 deletions(-)

diff --git a/include/net/ip6_fib.h b/include/net/ip6_fib.h
index 05e6f756feaf..e6f4d986fb63 100644
--- a/include/net/ip6_fib.h
+++ b/include/net/ip6_fib.h
@@ -179,6 +179,9 @@ struct fib6_info {
 
 	refcount_t			fib6_ref;
 	unsigned long			expires;
+
+	struct hlist_node		gc_link;
+
 	struct dst_metrics		*fib6_metrics;
 #define fib6_pmtu		fib6_metrics->metrics[RTAX_MTU-1]
 
@@ -247,19 +250,6 @@ static inline bool fib6_requires_src(const struct fib6_info *rt)
 	return rt->fib6_src.plen > 0;
 }
 
-static inline void fib6_clean_expires(struct fib6_info *f6i)
-{
-	f6i->fib6_flags &= ~RTF_EXPIRES;
-	f6i->expires = 0;
-}
-
-static inline void fib6_set_expires(struct fib6_info *f6i,
-				    unsigned long expires)
-{
-	f6i->expires = expires;
-	f6i->fib6_flags |= RTF_EXPIRES;
-}
-
 static inline bool fib6_check_expired(const struct fib6_info *f6i)
 {
 	if (f6i->fib6_flags & RTF_EXPIRES)
@@ -267,6 +257,11 @@ static inline bool fib6_check_expired(const struct fib6_info *f6i)
 	return false;
 }
 
+static inline bool fib6_has_expires(const struct fib6_info *f6i)
+{
+	return f6i->fib6_flags & RTF_EXPIRES;
+}
+
 /* Function to safely get fn->fn_sernum for passed in rt
  * and store result in passed in cookie.
  * Return true if we can get cookie safely
@@ -388,6 +383,7 @@ struct fib6_table {
 	struct inet_peer_base	tb6_peers;
 	unsigned int		flags;
 	unsigned int		fib_seq;
+	struct hlist_head       tb6_gc_hlist;	/* GC candidates */
 #define RT6_TABLE_HAS_DFLT_ROUTER	BIT(0)
 };
 
@@ -504,6 +500,49 @@ void fib6_gc_cleanup(void);
 
 int fib6_init(void);
 
+/* fib6_info must be locked by the caller, and fib6_info->fib6_table can be
+ * NULL.
+ */
+static inline void fib6_set_expires_locked(struct fib6_info *f6i, unsigned long expires)
+{
+	struct fib6_table *tb6;
+
+	tb6 = f6i->fib6_table;
+	f6i->expires = expires;
+	if (tb6 && !fib6_has_expires(f6i))
+		hlist_add_head(&f6i->gc_link, &tb6->tb6_gc_hlist);
+	f6i->fib6_flags |= RTF_EXPIRES;
+}
+
+/* fib6_info must be locked by the caller, and fib6_info->fib6_table can be
+ * NULL.  If fib6_table is NULL, the fib6_info will no be inserted into the
+ * list of GC candidates until it is inserted into a table.
+ */
+static inline void fib6_set_expires(struct fib6_info *f6i, unsigned long expires)
+{
+	spin_lock_bh(&f6i->fib6_table->tb6_lock);
+	fib6_set_expires_locked(f6i, expires);
+	spin_unlock_bh(&f6i->fib6_table->tb6_lock);
+}
+
+static inline void fib6_clean_expires_locked(struct fib6_info *f6i)
+{
+	struct fib6_table *tb6;
+
+	tb6 = f6i->fib6_table;
+	if (tb6 && fib6_has_expires(f6i))
+		hlist_del_init(&f6i->gc_link);
+	f6i->fib6_flags &= ~RTF_EXPIRES;
+	f6i->expires = 0;
+}
+
+static inline void fib6_clean_expires(struct fib6_info *f6i)
+{
+	spin_lock_bh(&f6i->fib6_table->tb6_lock);
+	fib6_clean_expires_locked(f6i);
+	spin_unlock_bh(&f6i->fib6_table->tb6_lock);
+}
+
 struct ipv6_route_iter {
 	struct seq_net_private p;
 	struct fib6_walker w;
diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
index bac768d36cc1..a4422d513d4d 100644
--- a/net/ipv6/ip6_fib.c
+++ b/net/ipv6/ip6_fib.c
@@ -160,6 +160,8 @@ struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh)
 	INIT_LIST_HEAD(&f6i->fib6_siblings);
 	refcount_set(&f6i->fib6_ref, 1);
 
+	INIT_HLIST_NODE(&f6i->gc_link);
+
 	return f6i;
 }
 
@@ -246,6 +248,7 @@ static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
 				   net->ipv6.fib6_null_entry);
 		table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
 		inet_peer_base_init(&table->tb6_peers);
+		INIT_HLIST_HEAD(&table->tb6_gc_hlist);
 	}
 
 	return table;
@@ -1057,6 +1060,11 @@ static void fib6_purge_rt(struct fib6_info *rt, struct fib6_node *fn,
 				    lockdep_is_held(&table->tb6_lock));
 		}
 	}
+
+	if (fib6_has_expires(rt)) {
+		hlist_del_init(&rt->gc_link);
+		rt->fib6_flags &= ~RTF_EXPIRES;
+	}
 }
 
 /*
@@ -1118,9 +1126,9 @@ static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
 				if (!(iter->fib6_flags & RTF_EXPIRES))
 					return -EEXIST;
 				if (!(rt->fib6_flags & RTF_EXPIRES))
-					fib6_clean_expires(iter);
+					fib6_clean_expires_locked(iter);
 				else
-					fib6_set_expires(iter, rt->expires);
+					fib6_set_expires_locked(iter, rt->expires);
 
 				if (rt->fib6_pmtu)
 					fib6_metric_set(iter, RTAX_MTU,
@@ -1480,6 +1488,9 @@ int fib6_add(struct fib6_node *root, struct fib6_info *rt,
 			list_add(&rt->nh_list, &rt->nh->f6i_list);
 		__fib6_update_sernum_upto_root(rt, fib6_new_sernum(info->nl_net));
 		fib6_start_gc(info->nl_net, rt);
+
+		if (fib6_has_expires(rt))
+			hlist_add_head(&rt->gc_link, &table->tb6_gc_hlist);
 	}
 
 out:
@@ -2295,7 +2306,7 @@ static int fib6_age(struct fib6_info *rt, void *arg)
 	 *	Routes are expired even if they are in use.
 	 */
 
-	if (rt->fib6_flags & RTF_EXPIRES && rt->expires) {
+	if (fib6_has_expires(rt) && rt->expires) {
 		if (time_after(now, rt->expires)) {
 			RT6_TRACE("expiring %p\n", rt);
 			return -1;
@@ -2312,6 +2323,40 @@ static int fib6_age(struct fib6_info *rt, void *arg)
 	return 0;
 }
 
+static void fib6_gc_table(struct net *net,
+			  struct fib6_table *tb6,
+			  void *arg)
+{
+	struct fib6_info *rt;
+	struct hlist_node *n;
+	struct nl_info info = {
+		.nl_net = net,
+		.skip_notify = false,
+	};
+
+	hlist_for_each_entry_safe(rt, n, &tb6->tb6_gc_hlist, gc_link)
+		if (fib6_age(rt, arg) == -1)
+			fib6_del(rt, &info);
+}
+
+static void fib6_gc_all(struct net *net, void *arg)
+{
+	struct fib6_table *table;
+	struct hlist_head *head;
+	unsigned int h;
+
+	rcu_read_lock();
+	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
+		head = &net->ipv6.fib_table_hash[h];
+		hlist_for_each_entry_rcu(table, head, tb6_hlist) {
+			spin_lock_bh(&table->tb6_lock);
+			fib6_gc_table(net, table, arg);
+			spin_unlock_bh(&table->tb6_lock);
+		}
+	}
+	rcu_read_unlock();
+}
+
 void fib6_run_gc(unsigned long expires, struct net *net, bool force)
 {
 	struct fib6_gc_args gc_args;
@@ -2327,7 +2372,7 @@ void fib6_run_gc(unsigned long expires, struct net *net, bool force)
 			  net->ipv6.sysctl.ip6_rt_gc_interval;
 	gc_args.more = 0;
 
-	fib6_clean_all(net, fib6_age, &gc_args);
+	fib6_gc_all(net, &gc_args);
 	now = jiffies;
 	net->ipv6.ip6_rt_last_gc = now;
 
diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index 64e873f5895f..a69083563689 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -3760,10 +3760,10 @@ static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
 		rt->dst_nocount = true;
 
 	if (cfg->fc_flags & RTF_EXPIRES)
-		fib6_set_expires(rt, jiffies +
-				clock_t_to_jiffies(cfg->fc_expires));
+		fib6_set_expires_locked(rt, jiffies +
+					clock_t_to_jiffies(cfg->fc_expires));
 	else
-		fib6_clean_expires(rt);
+		fib6_clean_expires_locked(rt);
 
 	if (cfg->fc_protocol == RTPROT_UNSPEC)
 		cfg->fc_protocol = RTPROT_BOOT;
-- 
2.34.1

[PATCH net-next v3 2/2] selftests: fib_tests: Add a test case for IPv6 garbage collection

[Kui-Feng Lee]

Add 10 IPv6 routes with expiration time.  Wait for a few seconds
to make sure they are removed correctly.

Signed-off-by: Kui-Feng Lee <kuifeng@meta.com>
---
 tools/testing/selftests/net/fib_tests.sh | 46 +++++++++++++++++++++++-
 1 file changed, 45 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/net/fib_tests.sh b/tools/testing/selftests/net/fib_tests.sh
index 35d89dfa6f11..87c871cae8c3 100755
--- a/tools/testing/selftests/net/fib_tests.sh
+++ b/tools/testing/selftests/net/fib_tests.sh
@@ -9,7 +9,7 @@ ret=0
 ksft_skip=4
 
 # all tests in this script. Can be overridden with -t option
-TESTS="unregister down carrier nexthop suppress ipv6_notify ipv4_notify ipv6_rt ipv4_rt ipv6_addr_metric ipv4_addr_metric ipv6_route_metrics ipv4_route_metrics ipv4_route_v6_gw rp_filter ipv4_del_addr ipv4_mangle ipv6_mangle ipv4_bcast_neigh"
+TESTS="unregister down carrier nexthop suppress ipv6_notify ipv4_notify ipv6_rt ipv4_rt ipv6_addr_metric ipv4_addr_metric ipv6_route_metrics ipv4_route_metrics ipv4_route_v6_gw rp_filter ipv4_del_addr ipv4_mangle ipv6_mangle ipv4_bcast_neigh fib6_gc_test"
 
 VERBOSE=0
 PAUSE_ON_FAIL=no
@@ -747,6 +747,49 @@ fib_notify_test()
 	cleanup &> /dev/null
 }
 
+fib6_gc_test()
+{
+	setup
+
+	echo
+	echo "Fib6 garbage collection test"
+	set -e
+
+	# Check expiration of routes every 3 seconds (GC)
+	$NS_EXEC sysctl -wq net.ipv6.route.gc_interval=3
+
+	$IP link add dummy_10 type dummy
+	$IP link set dev dummy_10 up
+	$IP -6 address add 2001:10::1/64 dev dummy_10
+
+	for i in 0 1 2 3 4 5 6 7 8 9; do
+	    # Expire route after 2 seconds
+	    $IP -6 route add 2001:20::1$i \
+		via 2001:10::2 dev dummy_10 expires 2
+	done
+	N_EXP=$($IP -6 route list |grep expires|wc -l)
+	if [ $N_EXP -ne 10 ]; then
+		echo "FAIL: expected 10 routes with expires, got $N_EXP"
+		ret=1
+	else
+	    sleep 4
+	    N_EXP_s20=$($IP -6 route list |grep expires|wc -l)
+
+	    if [ $N_EXP_s20 -ne 0 ]; then
+		echo "FAIL: expected 0 routes with expires, got $N_EXP_s20"
+		ret=1
+	    else
+		ret=0
+	    fi
+	fi
+
+	set +e
+
+	log_test $ret 0 "ipv6 route garbage collection"
+
+	cleanup &> /dev/null
+}
+
 fib_suppress_test()
 {
 	echo
@@ -2217,6 +2260,7 @@ do
 	ipv4_mangle)			ipv4_mangle_test;;
 	ipv6_mangle)			ipv6_mangle_test;;
 	ipv4_bcast_neigh)		ipv4_bcast_neigh_test;;
+	fib6_gc_test|ipv6_gc)		fib6_gc_test;;
 
 	help) echo "Test names: $TESTS"; exit 0;;
 	esac
-- 
2.34.1

Re: [PATCH net-next v3 0/2] Remove expired routes with a separated list of routes.

[Paolo Abeni]

On Tue, 2023-07-18 at 11:33 -0700, Kui-Feng Lee wrote:
> FIB6 GC walks trees of fib6_tables to remove expired routes. Walking a tree
> can be expensive if the number of routes in a table is big, even if most of
> them are permanent. Checking routes in a separated list of routes having
> expiration will avoid this potential issue.
> 
> Background
> ==========
> 
> The size of a Linux IPv6 routing table can become a big problem if not
> managed appropriately.  Now, Linux has a garbage collector to remove
> expired routes periodically.  However, this may lead to a situation in the routing path is blocked for a long period due to an
> excessive number of routes.
> 
> For example, years ago, there is a commit c7bb4b89033b ("ipv6: tcp: drop
> silly ICMPv6 packet too big messages") about "ICMPv6 Packet too big
> messages". The root cause is that malicious ICMPv6 packets were sent back
> for every small packet sent to them. These packets add routes with an
> expiration time that prompts the GC to periodically check all routes in the
> tables, including permanent ones.
> 
> Why Route Expires
> =================
> 
> Users can add IPv6 routes with an expiration time manually. However,
> the Neighbor Discovery protocol may also generate routes that can
> expire.  For example, Router Advertisement (RA) messages may create a
> default route with an expiration time. [RFC 4861] For IPv4, it is not
> possible to set an expiration time for a route, and there is no RA, so
> there is no need to worry about such issues.
> 
> Create Routes with Expires
> ==========================
> 
> You can create routes with expires with the  command.
> 
> For example,
> 
>     ip -6 route add 2001:b000:591::3 via fe80::5054:ff:fe12:3457 \ 
>         dev enp0s3 expires 30
> 
> The route that has been generated will be deleted automatically in 30
> seconds.
> 
> GC of FIB6
> ==========
> 
> The function called fib6_run_gc() is responsible for performing
> garbage collection (GC) for the Linux IPv6 stack. It checks for the
> expiration of every route by traversing the trees of routing
> tables. The time taken to traverse a routing table increases with its
> size. Holding the routing table lock during traversal is particularly
> undesirable. Therefore, it is preferable to keep the lock for the
> shortest possible duration.
> 
> Solution
> ========
> 
> The cause of the issue is keeping the routing table locked during the
> traversal of large trees. To solve this problem, we can create a separate
> list of routes that have expiration. This will prevent GC from checking
> permanent routes.
> 
> Result
> ======
> 
> We conducted a test to measure the execution times of fib6_gc_timer_cb()
> and observed that it enhances the GC of FIB6. During the test, we added
> permanent routes with the following numbers: 1000, 3000, 6000, and
> 9000. Additionally, we added a route with an expiration time.
> 
> Here are the average execution times for the kernel without the patch.
>  - 120020 ns with 1000 permanent routes
>  - 308920 ns with 3000 ...
>  - 581470 ns with 6000 ...
>  - 855310 ns with 9000 ...
> 
> The kernel with the patch consistently takes around 14000 ns to execute,
> regardless of the number of permanent routes that are installed.
> 
> Major changes from v2:
> 
>  - Remove unnecessary and incorrect sysctl restoring in the test case.
> 
> Major changes from v1:
> 
>  - Moved gc_link to avoid creating a hole in fib6_info.
> 
>  - Moved fib6_set_expires*() and fib6_clean_expires*() to the header
>    file and inlined. And removed duplicated lines.
> 
>  - Added a test case.
> 
> ---
> v1: https://lore.kernel.org/all/20230710203609.520720-1-kuifeng@meta.com/
> v2: https://lore.kernel.org/all/20230718180321.294721-1-kuifeng@meta.com/

Too bad I did not notice v3 before starting reviewing v2.

When posting a new version you must wait the 24h quarantine period,
see:

https://elixir.bootlin.com/linux/v6.4/source/Documentation/process/maintainer-netdev.rst#L15

I assume this does not cope with the feedback on previous version ;)

/P