diff mbox series

[net-next,v3,3/6] net: bridge: Track and limit dynamically learned FDB entries

Message ID 20230905-fdb_limit-v3-3-7597cd500a82@avm.de (mailing list archive)
State Deferred
Delegated to: Netdev Maintainers
Headers show
Series bridge: Add a limit on learned FDB entries | expand

Checks

Context Check Description
netdev/series_format success Posting correctly formatted
netdev/tree_selection success Clearly marked for net-next
netdev/fixes_present success Fixes tag not required for -next series
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 1337 this patch: 1337
netdev/cc_maintainers success CCed 8 of 8 maintainers
netdev/build_clang success Errors and warnings before: 1353 this patch: 1353
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/deprecated_api success None detected
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 1360 this patch: 1360
netdev/checkpatch warning WARNING: line length of 92 exceeds 80 columns
netdev/kdoc success Errors and warnings before: 3 this patch: 3
netdev/source_inline success Was 0 now: 0

Commit Message

Johannes Nixdorf Sept. 5, 2023, 11:47 a.m. UTC
A malicious actor behind one bridge port may spam the kernel with packets
with a random source MAC address, each of which will create an FDB entry,
each of which is a dynamic allocation in the kernel.

There are roughly 2^48 different MAC addresses, further limited by the
rhashtable they are stored in to 2^31. Each entry is of the type struct
net_bridge_fdb_entry, which is currently 128 bytes big. This means the
maximum amount of memory allocated for FDB entries is 2^31 * 128B =
256GiB, which is too much for most computers.

Mitigate this by maintaining a per bridge count of those automatically
generated entries in fdb_n_learned_entries, and a limit in
fdb_max_learned_entries. If the limit is hit new entries are not learned
anymore.

For backwards compatibility the default setting of 0 disables the limit.

User-added entries by netlink or from bridge or bridge port addresses
are never blocked and do not count towards that limit.

Introduce a new fdb entry flag BR_FDB_DYNAMIC_LEARNED to keep track of
whether an FDB entry is included in the count. The flag is enabled for
dynamically learned entries, and disabled for all other entries. This
should be equivalent to BR_FDB_ADDED_BY_USER and BR_FDB_LOCAL being unset,
but contrary to the two flags it can be toggled atomically.

Atomicity is required here, as there are multiple callers that modify the
flags, but are not under a common lock (br_fdb_update is the exception
for br->hash_lock, br_fdb_external_learn_add for RTNL).

Signed-off-by: Johannes Nixdorf <jnixdorf-oss@avm.de>
---
 net/bridge/br_fdb.c     | 34 ++++++++++++++++++++++++++++++++--
 net/bridge/br_private.h |  4 ++++
 2 files changed, 36 insertions(+), 2 deletions(-)

Comments

Nikolay Aleksandrov Sept. 14, 2023, 3:15 p.m. UTC | #1
On 9/5/23 14:47, Johannes Nixdorf wrote:
> A malicious actor behind one bridge port may spam the kernel with packets
> with a random source MAC address, each of which will create an FDB entry,
> each of which is a dynamic allocation in the kernel.
> 
> There are roughly 2^48 different MAC addresses, further limited by the
> rhashtable they are stored in to 2^31. Each entry is of the type struct
> net_bridge_fdb_entry, which is currently 128 bytes big. This means the
> maximum amount of memory allocated for FDB entries is 2^31 * 128B =
> 256GiB, which is too much for most computers.
> 
> Mitigate this by maintaining a per bridge count of those automatically
> generated entries in fdb_n_learned_entries, and a limit in
> fdb_max_learned_entries. If the limit is hit new entries are not learned
> anymore.
> 
> For backwards compatibility the default setting of 0 disables the limit.
> 
> User-added entries by netlink or from bridge or bridge port addresses
> are never blocked and do not count towards that limit.
> 
> Introduce a new fdb entry flag BR_FDB_DYNAMIC_LEARNED to keep track of
> whether an FDB entry is included in the count. The flag is enabled for
> dynamically learned entries, and disabled for all other entries. This
> should be equivalent to BR_FDB_ADDED_BY_USER and BR_FDB_LOCAL being unset,
> but contrary to the two flags it can be toggled atomically.
> 
> Atomicity is required here, as there are multiple callers that modify the
> flags, but are not under a common lock (br_fdb_update is the exception
> for br->hash_lock, br_fdb_external_learn_add for RTNL).
> 
> Signed-off-by: Johannes Nixdorf <jnixdorf-oss@avm.de>
> ---
>   net/bridge/br_fdb.c     | 34 ++++++++++++++++++++++++++++++++--
>   net/bridge/br_private.h |  4 ++++
>   2 files changed, 36 insertions(+), 2 deletions(-)
> 
> diff --git a/net/bridge/br_fdb.c b/net/bridge/br_fdb.c
> index 06e28ef8d9ff..f8a96ed9a338 100644
> --- a/net/bridge/br_fdb.c
> +++ b/net/bridge/br_fdb.c
> @@ -329,11 +329,18 @@ static void fdb_delete(struct net_bridge *br, struct net_bridge_fdb_entry *f,
>   	hlist_del_init_rcu(&f->fdb_node);
>   	rhashtable_remove_fast(&br->fdb_hash_tbl, &f->rhnode,
>   			       br_fdb_rht_params);
> +	if (test_bit(BR_FDB_DYNAMIC_LEARNED, &f->flags))

this is racy with br_fdb_update(), the fdb entry is still visible 
because a grace period hasn't passed, so in theory br_fdb_update() can
unset the bit after this test and...
> +		atomic_dec(&br->fdb_n_learned_entries);
... this will decrease once more wrongly

>   	fdb_notify(br, f, RTM_DELNEIGH, swdev_notify);
>   	call_rcu(&f->rcu, fdb_rcu_free);
>   }
>   
> -/* Delete a local entry if no other port had the same address. */
> +/* Delete a local entry if no other port had the same address.
> + *
> + * This function should only be called on entries with BR_FDB_LOCAL set,
> + * so even with BR_FDB_ADDED_BY_USER cleared we never need to increase
> + * the accounting for dynamically learned entries again.
> + */
>   static void fdb_delete_local(struct net_bridge *br,
>   			     const struct net_bridge_port *p,
>   			     struct net_bridge_fdb_entry *f)
> @@ -388,9 +395,20 @@ static struct net_bridge_fdb_entry *fdb_create(struct net_bridge *br,
>   					       __u16 vid,
>   					       unsigned long flags)
>   {
> +	bool learned = !test_bit(BR_FDB_ADDED_BY_USER, &flags) &&
> +		       !test_bit(BR_FDB_LOCAL, &flags);
> +	u32 max_learned = READ_ONCE(br->fdb_max_learned_entries);
>   	struct net_bridge_fdb_entry *fdb;
>   	int err;
>   
> +	if (likely(learned)) {
> +		int n_learned = atomic_read(&br->fdb_n_learned_entries);
> +
> +		if (unlikely(max_learned && n_learned >= max_learned))
> +			return NULL;
> +		__set_bit(BR_FDB_DYNAMIC_LEARNED, &flags);
> +	}
> +
>   	fdb = kmem_cache_alloc(br_fdb_cache, GFP_ATOMIC);
>   	if (!fdb)
>   		return NULL;
> @@ -407,6 +425,9 @@ static struct net_bridge_fdb_entry *fdb_create(struct net_bridge *br,
>   		return NULL;
>   	}
>   
> +	if (likely(learned))
> +		atomic_inc(&br->fdb_n_learned_entries);
> +
>   	hlist_add_head_rcu(&fdb->fdb_node, &br->fdb_list);
>   
>   	return fdb;
> @@ -893,8 +914,11 @@ void br_fdb_update(struct net_bridge *br, struct net_bridge_port *source,
>   					clear_bit(BR_FDB_LOCKED, &fdb->flags);
>   			}
>   
> -			if (unlikely(test_bit(BR_FDB_ADDED_BY_USER, &flags)))
> +			if (unlikely(test_bit(BR_FDB_ADDED_BY_USER, &flags))) {
>   				set_bit(BR_FDB_ADDED_BY_USER, &fdb->flags);
> +				if (test_and_clear_bit(BR_FDB_DYNAMIC_LEARNED, &fdb->flags))
> +					atomic_dec(&br->fdb_n_learned_entries);
> +			}
>   			if (unlikely(fdb_modified)) {
>   				trace_br_fdb_update(br, source, addr, vid, flags);
>   				fdb_notify(br, fdb, RTM_NEWNEIGH, true);
> @@ -1071,6 +1095,8 @@ static int fdb_add_entry(struct net_bridge *br, struct net_bridge_port *source,
>   		}
>   
>   		set_bit(BR_FDB_ADDED_BY_USER, &fdb->flags);
> +		if (test_and_clear_bit(BR_FDB_DYNAMIC_LEARNED, &fdb->flags))
> +			atomic_dec(&br->fdb_n_learned_entries);
>   	}
>   
>   	if (fdb_to_nud(br, fdb) != state) {
> @@ -1445,6 +1471,10 @@ int br_fdb_external_learn_add(struct net_bridge *br, struct net_bridge_port *p,
>   		if (!p)
>   			set_bit(BR_FDB_LOCAL, &fdb->flags);
>   
> +		if ((swdev_notify || !p) &&
> +		    test_and_clear_bit(BR_FDB_DYNAMIC_LEARNED, &fdb->flags))
> +			atomic_dec(&br->fdb_n_learned_entries);
> +
>   		if (modified)
>   			fdb_notify(br, fdb, RTM_NEWNEIGH, swdev_notify);
>   	}
> diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
> index a63b32c1638e..675cc40ae1dc 100644
> --- a/net/bridge/br_private.h
> +++ b/net/bridge/br_private.h
> @@ -274,6 +274,7 @@ enum {
>   	BR_FDB_NOTIFY,
>   	BR_FDB_NOTIFY_INACTIVE,
>   	BR_FDB_LOCKED,
> +	BR_FDB_DYNAMIC_LEARNED,
>   };
>   
>   struct net_bridge_fdb_key {
> @@ -554,6 +555,9 @@ struct net_bridge {
>   	struct kobject			*ifobj;
>   	u32				auto_cnt;
>   
> +	atomic_t			fdb_n_learned_entries;
> +	u32				fdb_max_learned_entries;
> +
>   #ifdef CONFIG_NET_SWITCHDEV
>   	/* Counter used to make sure that hardware domains get unique
>   	 * identifiers in case a bridge spans multiple switchdev instances.
>
diff mbox series

Patch

diff --git a/net/bridge/br_fdb.c b/net/bridge/br_fdb.c
index 06e28ef8d9ff..f8a96ed9a338 100644
--- a/net/bridge/br_fdb.c
+++ b/net/bridge/br_fdb.c
@@ -329,11 +329,18 @@  static void fdb_delete(struct net_bridge *br, struct net_bridge_fdb_entry *f,
 	hlist_del_init_rcu(&f->fdb_node);
 	rhashtable_remove_fast(&br->fdb_hash_tbl, &f->rhnode,
 			       br_fdb_rht_params);
+	if (test_bit(BR_FDB_DYNAMIC_LEARNED, &f->flags))
+		atomic_dec(&br->fdb_n_learned_entries);
 	fdb_notify(br, f, RTM_DELNEIGH, swdev_notify);
 	call_rcu(&f->rcu, fdb_rcu_free);
 }
 
-/* Delete a local entry if no other port had the same address. */
+/* Delete a local entry if no other port had the same address.
+ *
+ * This function should only be called on entries with BR_FDB_LOCAL set,
+ * so even with BR_FDB_ADDED_BY_USER cleared we never need to increase
+ * the accounting for dynamically learned entries again.
+ */
 static void fdb_delete_local(struct net_bridge *br,
 			     const struct net_bridge_port *p,
 			     struct net_bridge_fdb_entry *f)
@@ -388,9 +395,20 @@  static struct net_bridge_fdb_entry *fdb_create(struct net_bridge *br,
 					       __u16 vid,
 					       unsigned long flags)
 {
+	bool learned = !test_bit(BR_FDB_ADDED_BY_USER, &flags) &&
+		       !test_bit(BR_FDB_LOCAL, &flags);
+	u32 max_learned = READ_ONCE(br->fdb_max_learned_entries);
 	struct net_bridge_fdb_entry *fdb;
 	int err;
 
+	if (likely(learned)) {
+		int n_learned = atomic_read(&br->fdb_n_learned_entries);
+
+		if (unlikely(max_learned && n_learned >= max_learned))
+			return NULL;
+		__set_bit(BR_FDB_DYNAMIC_LEARNED, &flags);
+	}
+
 	fdb = kmem_cache_alloc(br_fdb_cache, GFP_ATOMIC);
 	if (!fdb)
 		return NULL;
@@ -407,6 +425,9 @@  static struct net_bridge_fdb_entry *fdb_create(struct net_bridge *br,
 		return NULL;
 	}
 
+	if (likely(learned))
+		atomic_inc(&br->fdb_n_learned_entries);
+
 	hlist_add_head_rcu(&fdb->fdb_node, &br->fdb_list);
 
 	return fdb;
@@ -893,8 +914,11 @@  void br_fdb_update(struct net_bridge *br, struct net_bridge_port *source,
 					clear_bit(BR_FDB_LOCKED, &fdb->flags);
 			}
 
-			if (unlikely(test_bit(BR_FDB_ADDED_BY_USER, &flags)))
+			if (unlikely(test_bit(BR_FDB_ADDED_BY_USER, &flags))) {
 				set_bit(BR_FDB_ADDED_BY_USER, &fdb->flags);
+				if (test_and_clear_bit(BR_FDB_DYNAMIC_LEARNED, &fdb->flags))
+					atomic_dec(&br->fdb_n_learned_entries);
+			}
 			if (unlikely(fdb_modified)) {
 				trace_br_fdb_update(br, source, addr, vid, flags);
 				fdb_notify(br, fdb, RTM_NEWNEIGH, true);
@@ -1071,6 +1095,8 @@  static int fdb_add_entry(struct net_bridge *br, struct net_bridge_port *source,
 		}
 
 		set_bit(BR_FDB_ADDED_BY_USER, &fdb->flags);
+		if (test_and_clear_bit(BR_FDB_DYNAMIC_LEARNED, &fdb->flags))
+			atomic_dec(&br->fdb_n_learned_entries);
 	}
 
 	if (fdb_to_nud(br, fdb) != state) {
@@ -1445,6 +1471,10 @@  int br_fdb_external_learn_add(struct net_bridge *br, struct net_bridge_port *p,
 		if (!p)
 			set_bit(BR_FDB_LOCAL, &fdb->flags);
 
+		if ((swdev_notify || !p) &&
+		    test_and_clear_bit(BR_FDB_DYNAMIC_LEARNED, &fdb->flags))
+			atomic_dec(&br->fdb_n_learned_entries);
+
 		if (modified)
 			fdb_notify(br, fdb, RTM_NEWNEIGH, swdev_notify);
 	}
diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
index a63b32c1638e..675cc40ae1dc 100644
--- a/net/bridge/br_private.h
+++ b/net/bridge/br_private.h
@@ -274,6 +274,7 @@  enum {
 	BR_FDB_NOTIFY,
 	BR_FDB_NOTIFY_INACTIVE,
 	BR_FDB_LOCKED,
+	BR_FDB_DYNAMIC_LEARNED,
 };
 
 struct net_bridge_fdb_key {
@@ -554,6 +555,9 @@  struct net_bridge {
 	struct kobject			*ifobj;
 	u32				auto_cnt;
 
+	atomic_t			fdb_n_learned_entries;
+	u32				fdb_max_learned_entries;
+
 #ifdef CONFIG_NET_SWITCHDEV
 	/* Counter used to make sure that hardware domains get unique
 	 * identifiers in case a bridge spans multiple switchdev instances.