Merge branch 'ipv6-data-races'

Eric Dumazet says:

====================
ipv6: round of data-races fixes

This series is inspired by one related syzbot report.

Many inet6_sk(sk) fields reads or writes are racy.

Move 1-bit fields to inet->inet_flags to provide
atomic safety. inet6_{test|set|clear|assign}_bit() helpers
could be changed later if we need to make room in inet_flags.

Also add missing READ_ONCE()/WRITE_ONCE() when
lockless readers need access to specific fields.

np->srcprefs will be handled separately to avoid merge conflicts
because a prior patch was posted for net tree.
====================

Signed-off-by: David S. Miller <[email protected]>

6 files changed, 44 insertions, 49 deletions

diff --git a/include/linux/ipv6.h b/include/linux/ipv6.h
index af8a771a053c..09253825c99c 100644
--- a/include/linux/ipv6.h
+++ b/include/linux/ipv6.h
@@ -213,28 +213,9 @@ struct ipv6_pinfo {
 	__be32			flow_label;
 	__u32			frag_size;
 
-	/*
-	 * Packed in 16bits.
-	 * Omit one shift by putting the signed field at MSB.
-	 */
-#if defined(__BIG_ENDIAN_BITFIELD)
-	__s16			hop_limit:9;
-	__u16			__unused_1:7;
-#else
-	__u16			__unused_1:7;
-	__s16			hop_limit:9;
-#endif
+	s16			hop_limit;
+	u8			mcast_hops;
 
-#if defined(__BIG_ENDIAN_BITFIELD)
-	/* Packed in 16bits. */
-	__s16			mcast_hops:9;
-	__u16			__unused_2:6,
-				mc_loop:1;
-#else
-	__u16			mc_loop:1,
-				__unused_2:6;
-	__s16			mcast_hops:9;
-#endif
 	int			ucast_oif;
 	int			mcast_oif;
 
@@ -262,21 +243,11 @@ struct ipv6_pinfo {
 	} rxopt;
 
 	/* sockopt flags */
-	__u16			recverr:1,
-	                        sndflow:1,
-				repflow:1,
-				pmtudisc:3,
-				padding:1,	/* 1 bit hole */
-				srcprefs:3,	/* 001: prefer temporary address
+	__u8			srcprefs:3;	/* 001: prefer temporary address
 						 * 010: prefer public address
 						 * 100: prefer care-of address
 						 */
-				dontfrag:1,
-				autoflowlabel:1,
-				autoflowlabel_set:1,
-				mc_all:1,
-				recverr_rfc4884:1,
-				rtalert_isolate:1;
+	__u8			pmtudisc;
 	__u8			min_hopcount;
 	__u8			tclass;
 	__be32			rcv_flowinfo;
@@ -293,6 +264,18 @@ struct ipv6_pinfo {
 	struct inet6_cork	cork;
 };
 
+/* We currently use available bits from inet_sk(sk)->inet_flags,
+ * this could change in the future.
+ */
+#define inet6_test_bit(nr, sk)			\
+	test_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
+#define inet6_set_bit(nr, sk)			\
+	set_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
+#define inet6_clear_bit(nr, sk)			\
+	clear_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
+#define inet6_assign_bit(nr, sk, val)		\
+	assign_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags, val)
+
 /* WARNING: don't change the layout of the members in {raw,udp,tcp}6_sock! */
 struct raw6_sock {
 	/* inet_sock has to be the first member of raw6_sock */
diff --git a/include/net/inet_sock.h b/include/net/inet_sock.h
index 2de0e4d4a027..98e11958cdff 100644
--- a/include/net/inet_sock.h
+++ b/include/net/inet_sock.h
@@ -268,6 +268,16 @@ enum {
 	INET_FLAGS_NODEFRAG	= 17,
 	INET_FLAGS_BIND_ADDRESS_NO_PORT = 18,
 	INET_FLAGS_DEFER_CONNECT = 19,
+	INET_FLAGS_MC6_LOOP	= 20,
+	INET_FLAGS_RECVERR6_RFC4884 = 21,
+	INET_FLAGS_MC6_ALL	= 22,
+	INET_FLAGS_AUTOFLOWLABEL_SET = 23,
+	INET_FLAGS_AUTOFLOWLABEL = 24,
+	INET_FLAGS_DONTFRAG	= 25,
+	INET_FLAGS_RECVERR6	= 26,
+	INET_FLAGS_REPFLOW	= 27,
+	INET_FLAGS_RTALERT_ISOLATE = 28,
+	INET_FLAGS_SNDFLOW	= 29,
 };
 
 /* cmsg flags for inet */
diff --git a/include/net/ip6_route.h b/include/net/ip6_route.h
index b32539bb0fb0..b1ea49900b4a 100644
--- a/include/net/ip6_route.h
+++ b/include/net/ip6_route.h
@@ -266,7 +266,7 @@ static inline unsigned int ip6_skb_dst_mtu(const struct sk_buff *skb)
 	const struct dst_entry *dst = skb_dst(skb);
 	unsigned int mtu;
 
-	if (np && np->pmtudisc >= IPV6_PMTUDISC_PROBE) {
+	if (np && READ_ONCE(np->pmtudisc) >= IPV6_PMTUDISC_PROBE) {
 		mtu = READ_ONCE(dst->dev->mtu);
 		mtu -= lwtunnel_headroom(dst->lwtstate, mtu);
 	} else {
@@ -277,14 +277,18 @@ static inline unsigned int ip6_skb_dst_mtu(const struct sk_buff *skb)
 
 static inline bool ip6_sk_accept_pmtu(const struct sock *sk)
 {
-	return inet6_sk(sk)->pmtudisc != IPV6_PMTUDISC_INTERFACE &&
-	       inet6_sk(sk)->pmtudisc != IPV6_PMTUDISC_OMIT;
+	u8 pmtudisc = READ_ONCE(inet6_sk(sk)->pmtudisc);
+
+	return pmtudisc != IPV6_PMTUDISC_INTERFACE &&
+	       pmtudisc != IPV6_PMTUDISC_OMIT;
 }
 
 static inline bool ip6_sk_ignore_df(const struct sock *sk)
 {
-	return inet6_sk(sk)->pmtudisc < IPV6_PMTUDISC_DO ||
-	       inet6_sk(sk)->pmtudisc == IPV6_PMTUDISC_OMIT;
+	u8 pmtudisc = READ_ONCE(inet6_sk(sk)->pmtudisc);
+
+	return pmtudisc < IPV6_PMTUDISC_DO ||
+	       pmtudisc == IPV6_PMTUDISC_OMIT;
 }
 
 static inline const struct in6_addr *rt6_nexthop(const struct rt6_info *rt,
diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index c6932d1a3fa8..bd115980809f 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -373,12 +373,12 @@ static inline void ipcm6_init(struct ipcm6_cookie *ipc6)
 }
 
 static inline void ipcm6_init_sk(struct ipcm6_cookie *ipc6,
-				 const struct ipv6_pinfo *np)
+				 const struct sock *sk)
 {
 	*ipc6 = (struct ipcm6_cookie) {
 		.hlimit = -1,
-		.tclass = np->tclass,
-		.dontfrag = np->dontfrag,
+		.tclass = inet6_sk(sk)->tclass,
+		.dontfrag = inet6_test_bit(DONTFRAG, sk),
 	};
 }
 
@@ -428,7 +428,7 @@ int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
 			   int flags);
 int ip6_flowlabel_init(void);
 void ip6_flowlabel_cleanup(void);
-bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np);
+bool ip6_autoflowlabel(struct net *net, const struct sock *sk);
 
 static inline void fl6_sock_release(struct ip6_flowlabel *fl)
 {
@@ -914,9 +914,9 @@ static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
 	int hlimit;
 
 	if (ipv6_addr_is_multicast(&fl6->daddr))
-		hlimit = np->mcast_hops;
+		hlimit = READ_ONCE(np->mcast_hops);
 	else
-		hlimit = np->hop_limit;
+		hlimit = READ_ONCE(np->hop_limit);
 	if (hlimit < 0)
 		hlimit = ip6_dst_hoplimit(dst);
 	return hlimit;
@@ -1303,9 +1303,7 @@ static inline int ip6_sock_set_v6only(struct sock *sk)
 
 static inline void ip6_sock_set_recverr(struct sock *sk)
 {
-	lock_sock(sk);
-	inet6_sk(sk)->recverr = true;
-	release_sock(sk);
+	inet6_set_bit(RECVERR6, sk);
 }
 
 static inline int __ip6_sock_set_addr_preferences(struct sock *sk, int val)
diff --git a/include/net/sock.h b/include/net/sock.h
index 676146e9d181..56ac1abadea5 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2238,7 +2238,7 @@ static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
 	}
 }
 
-bool sk_mc_loop(struct sock *sk);
+bool sk_mc_loop(const struct sock *sk);
 
 static inline bool sk_can_gso(const struct sock *sk)
 {
diff --git a/include/net/xfrm.h b/include/net/xfrm.h
index 363c7d510554..98d7aa78adda 100644
--- a/include/net/xfrm.h
+++ b/include/net/xfrm.h
@@ -2166,7 +2166,7 @@ static inline bool xfrm6_local_dontfrag(const struct sock *sk)
 
 	proto = sk->sk_protocol;
 	if (proto == IPPROTO_UDP || proto == IPPROTO_RAW)
-		return inet6_sk(sk)->dontfrag;
+		return inet6_test_bit(DONTFRAG, sk);
 
 	return false;
 }