diff mbox series

[4/6] net/tcp: Disable TCP-MD5 static key on tcp_md5sig_info destruction

Message ID 20220726201600.1715505-5-dima@arista.com (mailing list archive)
State Changes Requested
Delegated to: Netdev Maintainers
Headers show
Series net/crypto: Introduce crypto_pool | expand

Checks

Context Check Description
netdev/tree_selection success Guessed tree name to be net-next, async
netdev/fixes_present success Fixes tag not required for -next series
netdev/subject_prefix success Link
netdev/cover_letter success Series has a cover letter
netdev/patch_count success Link
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit fail Errors and warnings before: 1008 this patch: 1010
netdev/cc_maintainers success CCed 7 of 7 maintainers
netdev/build_clang fail Errors and warnings before: 125 this patch: 127
netdev/module_param success Was 0 now: 0
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn fail Errors and warnings before: 1015 this patch: 1017
netdev/checkpatch warning CHECK: Alignment should match open parenthesis WARNING: line length of 82 exceeds 80 columns WARNING: line length of 84 exceeds 80 columns
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

Dmitry Safonov July 26, 2022, 8:15 p.m. UTC
To do that, separate two scenarios:
- where it's the first MD5 key on the system, which means that enabling
  of the static key may need to sleep;
- copying of an existing key from a listening socket to the request
  socket upon receiving a signed TCP segment, where static key was
  already enabled (when the key was added to the listening socket).

Now the life-time of the static branch for TCP-MD5 is until:
- last tcp_md5sig_info is destroyed
- last socket in time-wait state with MD5 key is closed.

Which means that after all sockets with TCP-MD5 keys are gone, the
system gets back the performance of disabled md5-key static branch.

Signed-off-by: Dmitry Safonov <dima@arista.com>
---
 include/net/tcp.h        | 10 ++++++---
 net/ipv4/tcp.c           |  5 +----
 net/ipv4/tcp_ipv4.c      | 45 +++++++++++++++++++++++++++++++---------
 net/ipv4/tcp_minisocks.c |  9 +++++---
 net/ipv4/tcp_output.c    |  4 ++--
 net/ipv6/tcp_ipv6.c      | 10 ++++-----
 6 files changed, 55 insertions(+), 28 deletions(-)

Comments

kernel test robot Aug. 14, 2022, 3:49 p.m. UTC | #1
Hi Dmitry,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on 058affafc65a74cf54499fb578b66ad0b18f939b]

url:    https://github.com/intel-lab-lkp/linux/commits/Dmitry-Safonov/net-crypto-Introduce-crypto_pool/20220727-041830
base:   058affafc65a74cf54499fb578b66ad0b18f939b
config: i386-randconfig-a005 (https://download.01.org/0day-ci/archive/20220814/202208142357.bDGLpecB-lkp@intel.com/config)
compiler: gcc-11 (Debian 11.3.0-3) 11.3.0
reproduce (this is a W=1 build):
        # https://github.com/intel-lab-lkp/linux/commit/a4ee3ecdaada036ed6747ed86eaf7270d3f27bab
        git remote add linux-review https://github.com/intel-lab-lkp/linux
        git fetch --no-tags linux-review Dmitry-Safonov/net-crypto-Introduce-crypto_pool/20220727-041830
        git checkout a4ee3ecdaada036ed6747ed86eaf7270d3f27bab
        # save the config file
        mkdir build_dir && cp config build_dir/.config
        make W=1 O=build_dir ARCH=i386 SHELL=/bin/bash net/ipv4/

If you fix the issue, kindly add following tag where applicable
Reported-by: kernel test robot <lkp@intel.com>

All warnings (new ones prefixed by >>):

>> net/ipv4/tcp_ipv4.c:1174:5: warning: no previous prototype for '__tcp_md5_do_add' [-Wmissing-prototypes]
    1174 | int __tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
         |     ^~~~~~~~~~~~~~~~


vim +/__tcp_md5_do_add +1174 net/ipv4/tcp_ipv4.c

  1172	
  1173	/* This can be called on a newly created socket, from other files */
> 1174	int __tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
  1175			     int family, u8 prefixlen, int l3index, u8 flags,
  1176			     const u8 *newkey, u8 newkeylen, gfp_t gfp)
  1177	{
  1178		/* Add Key to the list */
  1179		struct tcp_md5sig_key *key;
  1180		struct tcp_sock *tp = tcp_sk(sk);
  1181		struct tcp_md5sig_info *md5sig;
  1182	
  1183		key = tcp_md5_do_lookup_exact(sk, addr, family, prefixlen, l3index, flags);
  1184		if (key) {
  1185			/* Pre-existing entry - just update that one.
  1186			 * Note that the key might be used concurrently.
  1187			 * data_race() is telling kcsan that we do not care of
  1188			 * key mismatches, since changing MD5 key on live flows
  1189			 * can lead to packet drops.
  1190			 */
  1191			data_race(memcpy(key->key, newkey, newkeylen));
  1192	
  1193			/* Pairs with READ_ONCE() in tcp_md5_hash_key().
  1194			 * Also note that a reader could catch new key->keylen value
  1195			 * but old key->key[], this is the reason we use __GFP_ZERO
  1196			 * at sock_kmalloc() time below these lines.
  1197			 */
  1198			WRITE_ONCE(key->keylen, newkeylen);
  1199	
  1200			return 0;
  1201		}
  1202	
  1203		md5sig = rcu_dereference_protected(tp->md5sig_info,
  1204						   lockdep_sock_is_held(sk));
  1205	
  1206		key = sock_kmalloc(sk, sizeof(*key), gfp | __GFP_ZERO);
  1207		if (!key)
  1208			return -ENOMEM;
  1209		if (!tcp_alloc_md5sig_pool()) {
  1210			sock_kfree_s(sk, key, sizeof(*key));
  1211			return -ENOMEM;
  1212		}
  1213	
  1214		memcpy(key->key, newkey, newkeylen);
  1215		key->keylen = newkeylen;
  1216		key->family = family;
  1217		key->prefixlen = prefixlen;
  1218		key->l3index = l3index;
  1219		key->flags = flags;
  1220		memcpy(&key->addr, addr,
  1221		       (IS_ENABLED(CONFIG_IPV6) && family == AF_INET6) ? sizeof(struct in6_addr) :
  1222									 sizeof(struct in_addr));
  1223		hlist_add_head_rcu(&key->node, &md5sig->head);
  1224		return 0;
  1225	}
  1226
kernel test robot Aug. 14, 2022, 3:49 p.m. UTC | #2
Hi Dmitry,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on 058affafc65a74cf54499fb578b66ad0b18f939b]

url:    https://github.com/intel-lab-lkp/linux/commits/Dmitry-Safonov/net-crypto-Introduce-crypto_pool/20220727-041830
base:   058affafc65a74cf54499fb578b66ad0b18f939b
config: x86_64-defconfig (https://download.01.org/0day-ci/archive/20220814/202208142332.WUqM9sfv-lkp@intel.com/config)
compiler: gcc-11 (Debian 11.3.0-3) 11.3.0
reproduce (this is a W=1 build):
        # https://github.com/intel-lab-lkp/linux/commit/a4ee3ecdaada036ed6747ed86eaf7270d3f27bab
        git remote add linux-review https://github.com/intel-lab-lkp/linux
        git fetch --no-tags linux-review Dmitry-Safonov/net-crypto-Introduce-crypto_pool/20220727-041830
        git checkout a4ee3ecdaada036ed6747ed86eaf7270d3f27bab
        # save the config file
        mkdir build_dir && cp config build_dir/.config
        make W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash net/ipv4/

If you fix the issue, kindly add following tag where applicable
Reported-by: kernel test robot <lkp@intel.com>

All warnings (new ones prefixed by >>):

>> net/ipv4/tcp_ipv4.c:1174:5: warning: no previous prototype for '__tcp_md5_do_add' [-Wmissing-prototypes]
    1174 | int __tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
         |     ^~~~~~~~~~~~~~~~


vim +/__tcp_md5_do_add +1174 net/ipv4/tcp_ipv4.c

  1172	
  1173	/* This can be called on a newly created socket, from other files */
> 1174	int __tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
  1175			     int family, u8 prefixlen, int l3index, u8 flags,
  1176			     const u8 *newkey, u8 newkeylen, gfp_t gfp)
  1177	{
  1178		/* Add Key to the list */
  1179		struct tcp_md5sig_key *key;
  1180		struct tcp_sock *tp = tcp_sk(sk);
  1181		struct tcp_md5sig_info *md5sig;
  1182	
  1183		key = tcp_md5_do_lookup_exact(sk, addr, family, prefixlen, l3index, flags);
  1184		if (key) {
  1185			/* Pre-existing entry - just update that one.
  1186			 * Note that the key might be used concurrently.
  1187			 * data_race() is telling kcsan that we do not care of
  1188			 * key mismatches, since changing MD5 key on live flows
  1189			 * can lead to packet drops.
  1190			 */
  1191			data_race(memcpy(key->key, newkey, newkeylen));
  1192	
  1193			/* Pairs with READ_ONCE() in tcp_md5_hash_key().
  1194			 * Also note that a reader could catch new key->keylen value
  1195			 * but old key->key[], this is the reason we use __GFP_ZERO
  1196			 * at sock_kmalloc() time below these lines.
  1197			 */
  1198			WRITE_ONCE(key->keylen, newkeylen);
  1199	
  1200			return 0;
  1201		}
  1202	
  1203		md5sig = rcu_dereference_protected(tp->md5sig_info,
  1204						   lockdep_sock_is_held(sk));
  1205	
  1206		key = sock_kmalloc(sk, sizeof(*key), gfp | __GFP_ZERO);
  1207		if (!key)
  1208			return -ENOMEM;
  1209		if (!tcp_alloc_md5sig_pool()) {
  1210			sock_kfree_s(sk, key, sizeof(*key));
  1211			return -ENOMEM;
  1212		}
  1213	
  1214		memcpy(key->key, newkey, newkeylen);
  1215		key->keylen = newkeylen;
  1216		key->family = family;
  1217		key->prefixlen = prefixlen;
  1218		key->l3index = l3index;
  1219		key->flags = flags;
  1220		memcpy(&key->addr, addr,
  1221		       (IS_ENABLED(CONFIG_IPV6) && family == AF_INET6) ? sizeof(struct in6_addr) :
  1222									 sizeof(struct in_addr));
  1223		hlist_add_head_rcu(&key->node, &md5sig->head);
  1224		return 0;
  1225	}
  1226
diff mbox series

Patch

diff --git a/include/net/tcp.h b/include/net/tcp.h
index f366626cbbba..aa735f963723 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -1663,7 +1663,11 @@  int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk, const struct sk_buff *skb);
 int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
 		   int family, u8 prefixlen, int l3index, u8 flags,
-		   const u8 *newkey, u8 newkeylen, gfp_t gfp);
+		   const u8 *newkey, u8 newkeylen);
+int tcp_md5_key_copy(struct sock *sk, const union tcp_md5_addr *addr,
+		     int family, u8 prefixlen, int l3index,
+		     struct tcp_md5sig_key *key);
+
 int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
 		   int family, u8 prefixlen, int l3index, u8 flags);
 struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
@@ -1671,7 +1675,7 @@  struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
 
 #ifdef CONFIG_TCP_MD5SIG
 #include <linux/jump_label.h>
-extern struct static_key_false tcp_md5_needed;
+extern struct static_key_false_deferred tcp_md5_needed;
 struct tcp_md5sig_key *__tcp_md5_do_lookup(const struct sock *sk, int l3index,
 					   const union tcp_md5_addr *addr,
 					   int family);
@@ -1679,7 +1683,7 @@  static inline struct tcp_md5sig_key *
 tcp_md5_do_lookup(const struct sock *sk, int l3index,
 		  const union tcp_md5_addr *addr, int family)
 {
-	if (!static_branch_unlikely(&tcp_md5_needed))
+	if (!static_branch_unlikely(&tcp_md5_needed.key))
 		return NULL;
 	return __tcp_md5_do_lookup(sk, l3index, addr, family);
 }
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 635c6782cdbb..300056008e90 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -4404,11 +4404,8 @@  bool tcp_alloc_md5sig_pool(void)
 	if (unlikely(!tcp_md5sig_pool_populated)) {
 		mutex_lock(&tcp_md5sig_mutex);
 
-		if (!tcp_md5sig_pool_populated) {
+		if (!tcp_md5sig_pool_populated)
 			__tcp_alloc_md5sig_pool();
-			if (tcp_md5sig_pool_populated)
-				static_branch_inc(&tcp_md5_needed);
-		}
 
 		mutex_unlock(&tcp_md5sig_mutex);
 	}
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 55e4092209a5..5b0caaea5029 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1044,7 +1044,7 @@  static void tcp_v4_reqsk_destructor(struct request_sock *req)
  * We need to maintain these in the sk structure.
  */
 
-DEFINE_STATIC_KEY_FALSE(tcp_md5_needed);
+DEFINE_STATIC_KEY_DEFERRED_FALSE(tcp_md5_needed, HZ);
 EXPORT_SYMBOL(tcp_md5_needed);
 
 static bool better_md5_match(struct tcp_md5sig_key *old, struct tcp_md5sig_key *new)
@@ -1171,9 +1171,9 @@  static int tcp_md5sig_info_add(struct sock *sk, gfp_t gfp)
 }
 
 /* This can be called on a newly created socket, from other files */
-int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
-		   int family, u8 prefixlen, int l3index, u8 flags,
-		   const u8 *newkey, u8 newkeylen, gfp_t gfp)
+int __tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
+		     int family, u8 prefixlen, int l3index, u8 flags,
+		     const u8 *newkey, u8 newkeylen, gfp_t gfp)
 {
 	/* Add Key to the list */
 	struct tcp_md5sig_key *key;
@@ -1200,9 +1200,6 @@  int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
 		return 0;
 	}
 
-	if (tcp_md5sig_info_add(sk, gfp))
-		return -ENOMEM;
-
 	md5sig = rcu_dereference_protected(tp->md5sig_info,
 					   lockdep_sock_is_held(sk));
 
@@ -1226,8 +1223,36 @@  int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
 	hlist_add_head_rcu(&key->node, &md5sig->head);
 	return 0;
 }
+
+int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
+		   int family, u8 prefixlen, int l3index, u8 flags,
+		   const u8 *newkey, u8 newkeylen)
+{
+	if (tcp_md5sig_info_add(sk, GFP_KERNEL))
+		return -ENOMEM;
+
+	static_branch_inc(&tcp_md5_needed.key);
+
+	return __tcp_md5_do_add(sk, addr, family, prefixlen, l3index, flags,
+				newkey, newkeylen, GFP_KERNEL);
+}
 EXPORT_SYMBOL(tcp_md5_do_add);
 
+int tcp_md5_key_copy(struct sock *sk, const union tcp_md5_addr *addr,
+		     int family, u8 prefixlen, int l3index,
+		     struct tcp_md5sig_key *key)
+{
+	if (tcp_md5sig_info_add(sk, sk_gfp_mask(sk, GFP_ATOMIC)))
+		return -ENOMEM;
+
+	atomic_inc(&tcp_md5_needed.key.key.enabled);
+
+	return __tcp_md5_do_add(sk, addr, family, prefixlen, l3index,
+				key->flags, key->key, key->keylen,
+				sk_gfp_mask(sk, GFP_ATOMIC));
+}
+EXPORT_SYMBOL(tcp_md5_key_copy);
+
 int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family,
 		   u8 prefixlen, int l3index, u8 flags)
 {
@@ -1314,7 +1339,7 @@  static int tcp_v4_parse_md5_keys(struct sock *sk, int optname,
 		return -EINVAL;
 
 	return tcp_md5_do_add(sk, addr, AF_INET, prefixlen, l3index, flags,
-			      cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
+			      cmd.tcpm_key, cmd.tcpm_keylen);
 }
 
 static int tcp_v4_md5_hash_headers(struct tcp_md5sig_pool *hp,
@@ -1571,8 +1596,7 @@  struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
 		 * memory, then we end up not copying the key
 		 * across. Shucks.
 		 */
-		tcp_md5_do_add(newsk, addr, AF_INET, 32, l3index, key->flags,
-			       key->key, key->keylen, GFP_ATOMIC);
+		tcp_md5_key_copy(newsk, addr, AF_INET, 32, l3index, key);
 		sk_gso_disable(newsk);
 	}
 #endif
@@ -2260,6 +2284,7 @@  void tcp_v4_destroy_sock(struct sock *sk)
 		tcp_clear_md5_list(sk);
 		kfree_rcu(rcu_dereference_protected(tp->md5sig_info, 1), rcu);
 		tp->md5sig_info = NULL;
+		static_branch_slow_dec_deferred(&tcp_md5_needed);
 	}
 #endif
 
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index cb95d88497ae..5d475a45a478 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -291,13 +291,14 @@  void tcp_time_wait(struct sock *sk, int state, int timeo)
 		 */
 		do {
 			tcptw->tw_md5_key = NULL;
-			if (static_branch_unlikely(&tcp_md5_needed)) {
+			if (static_branch_unlikely(&tcp_md5_needed.key)) {
 				struct tcp_md5sig_key *key;
 
 				key = tp->af_specific->md5_lookup(sk, sk);
 				if (key) {
 					tcptw->tw_md5_key = kmemdup(key, sizeof(*key), GFP_ATOMIC);
 					BUG_ON(tcptw->tw_md5_key && !tcp_alloc_md5sig_pool());
+					atomic_inc(&tcp_md5_needed.key.key.enabled);
 				}
 			}
 		} while (0);
@@ -337,11 +338,13 @@  EXPORT_SYMBOL(tcp_time_wait);
 void tcp_twsk_destructor(struct sock *sk)
 {
 #ifdef CONFIG_TCP_MD5SIG
-	if (static_branch_unlikely(&tcp_md5_needed)) {
+	if (static_branch_unlikely(&tcp_md5_needed.key)) {
 		struct tcp_timewait_sock *twsk = tcp_twsk(sk);
 
-		if (twsk->tw_md5_key)
+		if (twsk->tw_md5_key) {
 			kfree_rcu(twsk->tw_md5_key, rcu);
+			static_branch_slow_dec_deferred(&tcp_md5_needed);
+		}
 	}
 #endif
 }
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 78b654ff421b..9e12845a8758 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -766,7 +766,7 @@  static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb,
 
 	*md5 = NULL;
 #ifdef CONFIG_TCP_MD5SIG
-	if (static_branch_unlikely(&tcp_md5_needed) &&
+	if (static_branch_unlikely(&tcp_md5_needed.key) &&
 	    rcu_access_pointer(tp->md5sig_info)) {
 		*md5 = tp->af_specific->md5_lookup(sk, sk);
 		if (*md5) {
@@ -922,7 +922,7 @@  static unsigned int tcp_established_options(struct sock *sk, struct sk_buff *skb
 
 	*md5 = NULL;
 #ifdef CONFIG_TCP_MD5SIG
-	if (static_branch_unlikely(&tcp_md5_needed) &&
+	if (static_branch_unlikely(&tcp_md5_needed.key) &&
 	    rcu_access_pointer(tp->md5sig_info)) {
 		*md5 = tp->af_specific->md5_lookup(sk, sk);
 		if (*md5) {
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index e54eee80ce5f..cb891a71db0d 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -658,12 +658,11 @@  static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 	if (ipv6_addr_v4mapped(&sin6->sin6_addr))
 		return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
 				      AF_INET, prefixlen, l3index, flags,
-				      cmd.tcpm_key, cmd.tcpm_keylen,
-				      GFP_KERNEL);
+				      cmd.tcpm_key, cmd.tcpm_keylen);
 
 	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
 			      AF_INET6, prefixlen, l3index, flags,
-			      cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
+			      cmd.tcpm_key, cmd.tcpm_keylen);
 }
 
 static int tcp_v6_md5_hash_headers(struct tcp_md5sig_pool *hp,
@@ -1359,9 +1358,8 @@  static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 		 * memory, then we end up not copying the key
 		 * across. Shucks.
 		 */
-		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
-			       AF_INET6, 128, l3index, key->flags, key->key, key->keylen,
-			       sk_gfp_mask(sk, GFP_ATOMIC));
+		tcp_md5_key_copy(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
+			       AF_INET6, 128, l3index, key);
 	}
 #endif