Merge branch 'net-sysctl-races-round-4'

Kuniyuki Iwashima says:

====================
sysctl: Fix data-races around ipv4_net_table (Round 4).

This series fixes data-races around 17 knobs after fib_multipath_use_neigh
in ipv4_net_table.

tcp_fack was skipped because it's obsolete and there's no readers.

So, round 5 will start with tcp_dsack, 2 rounds left for 27 knobs.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

10 files changed, 36 insertions, 33 deletions

diff --git a/net/core/secure_seq.c b/net/core/secure_seq.c
index 5f85e01d4093..b0ff6153be62 100644
--- a/net/core/secure_seq.c
+++ b/net/core/secure_seq.c
@@ -64,7 +64,7 @@ u32 secure_tcpv6_ts_off(const struct net *net,
 		.daddr = *(struct in6_addr *)daddr,
 	};
 
-	if (net->ipv4.sysctl_tcp_timestamps != 1)
+	if (READ_ONCE(net->ipv4.sysctl_tcp_timestamps) != 1)
 		return 0;
 
 	ts_secret_init();
@@ -120,7 +120,7 @@ EXPORT_SYMBOL(secure_ipv6_port_ephemeral);
 #ifdef CONFIG_INET
 u32 secure_tcp_ts_off(const struct net *net, __be32 saddr, __be32 daddr)
 {
-	if (net->ipv4.sysctl_tcp_timestamps != 1)
+	if (READ_ONCE(net->ipv4.sysctl_tcp_timestamps) != 1)
 		return 0;
 
 	ts_secret_init();
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index d9fdcbae16ee..db7b2503f068 100644
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -2216,7 +2216,7 @@ void fib_select_multipath(struct fib_result *res, int hash)
 	}
 
 	change_nexthops(fi) {
-		if (net->ipv4.sysctl_fib_multipath_use_neigh) {
+		if (READ_ONCE(net->ipv4.sysctl_fib_multipath_use_neigh)) {
 			if (!fib_good_nh(nexthop_nh))
 				continue;
 			if (!first) {
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 91c4f60de75a..4702c61207a8 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -1929,7 +1929,7 @@ static u32 fib_multipath_custom_hash_outer(const struct net *net,
 					   const struct sk_buff *skb,
 					   bool *p_has_inner)
 {
-	u32 hash_fields = net->ipv4.sysctl_fib_multipath_hash_fields;
+	u32 hash_fields = READ_ONCE(net->ipv4.sysctl_fib_multipath_hash_fields);
 	struct flow_keys keys, hash_keys;
 
 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
@@ -1958,7 +1958,7 @@ static u32 fib_multipath_custom_hash_inner(const struct net *net,
 					   const struct sk_buff *skb,
 					   bool has_inner)
 {
-	u32 hash_fields = net->ipv4.sysctl_fib_multipath_hash_fields;
+	u32 hash_fields = READ_ONCE(net->ipv4.sysctl_fib_multipath_hash_fields);
 	struct flow_keys keys, hash_keys;
 
 	/* We assume the packet carries an encapsulation, but if none was
@@ -2018,7 +2018,7 @@ static u32 fib_multipath_custom_hash_skb(const struct net *net,
 static u32 fib_multipath_custom_hash_fl4(const struct net *net,
 					 const struct flowi4 *fl4)
 {
-	u32 hash_fields = net->ipv4.sysctl_fib_multipath_hash_fields;
+	u32 hash_fields = READ_ONCE(net->ipv4.sysctl_fib_multipath_hash_fields);
 	struct flow_keys hash_keys;
 
 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
@@ -2048,7 +2048,7 @@ int fib_multipath_hash(const struct net *net, const struct flowi4 *fl4,
 	struct flow_keys hash_keys;
 	u32 mhash = 0;
 
-	switch (net->ipv4.sysctl_fib_multipath_hash_policy) {
+	switch (READ_ONCE(net->ipv4.sysctl_fib_multipath_hash_policy)) {
 	case 0:
 		memset(&hash_keys, 0, sizeof(hash_keys));
 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
index 9b234b42021e..942d2dfa1115 100644
--- a/net/ipv4/syncookies.c
+++ b/net/ipv4/syncookies.c
@@ -247,12 +247,12 @@ bool cookie_timestamp_decode(const struct net *net,
 		return true;
 	}
 
-	if (!net->ipv4.sysctl_tcp_timestamps)
+	if (!READ_ONCE(net->ipv4.sysctl_tcp_timestamps))
 		return false;
 
 	tcp_opt->sack_ok = (options & TS_OPT_SACK) ? TCP_SACK_SEEN : 0;
 
-	if (tcp_opt->sack_ok && !net->ipv4.sysctl_tcp_sack)
+	if (tcp_opt->sack_ok && !READ_ONCE(net->ipv4.sysctl_tcp_sack))
 		return false;
 
 	if ((options & TS_OPT_WSCALE_MASK) == TS_OPT_WSCALE_MASK)
@@ -261,7 +261,7 @@ bool cookie_timestamp_decode(const struct net *net,
 	tcp_opt->wscale_ok = 1;
 	tcp_opt->snd_wscale = options & TS_OPT_WSCALE_MASK;
 
-	return net->ipv4.sysctl_tcp_window_scaling != 0;
+	return READ_ONCE(net->ipv4.sysctl_tcp_window_scaling) != 0;
 }
 EXPORT_SYMBOL(cookie_timestamp_decode);
 
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index 130e9c130311..5490c285668b 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -84,7 +84,7 @@ static int ipv4_local_port_range(struct ctl_table *table, int write,
 		 * port limit.
 		 */
 		if ((range[1] < range[0]) ||
-		    (range[0] < net->ipv4.sysctl_ip_prot_sock))
+		    (range[0] < READ_ONCE(net->ipv4.sysctl_ip_prot_sock)))
 			ret = -EINVAL;
 		else
 			set_local_port_range(net, range);
@@ -110,7 +110,7 @@ static int ipv4_privileged_ports(struct ctl_table *table, int write,
 		.extra2 = &ip_privileged_port_max,
 	};
 
-	pports = net->ipv4.sysctl_ip_prot_sock;
+	pports = READ_ONCE(net->ipv4.sysctl_ip_prot_sock);
 
 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 
@@ -122,7 +122,7 @@ static int ipv4_privileged_ports(struct ctl_table *table, int write,
 		if (range[0] < pports)
 			ret = -EINVAL;
 		else
-			net->ipv4.sysctl_ip_prot_sock = pports;
+			WRITE_ONCE(net->ipv4.sysctl_ip_prot_sock, pports);
 	}
 
 	return ret;
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index d451248bebec..07dbcbae7782 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -1051,7 +1051,7 @@ static void tcp_check_sack_reordering(struct sock *sk, const u32 low_seq,
 			 tp->undo_marker ? tp->undo_retrans : 0);
 #endif
 		tp->reordering = min_t(u32, (metric + mss - 1) / mss,
-				       sock_net(sk)->ipv4.sysctl_tcp_max_reordering);
+				       READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_max_reordering));
 	}
 
 	/* This exciting event is worth to be remembered. 8) */
@@ -2030,7 +2030,7 @@ static void tcp_check_reno_reordering(struct sock *sk, const int addend)
 		return;
 
 	tp->reordering = min_t(u32, tp->packets_out + addend,
-			       sock_net(sk)->ipv4.sysctl_tcp_max_reordering);
+			       READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_max_reordering));
 	tp->reord_seen++;
 	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRENOREORDER);
 }
@@ -2095,7 +2095,8 @@ static inline void tcp_init_undo(struct tcp_sock *tp)
 
 static bool tcp_is_rack(const struct sock *sk)
 {
-	return sock_net(sk)->ipv4.sysctl_tcp_recovery & TCP_RACK_LOSS_DETECTION;
+	return READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_recovery) &
+		TCP_RACK_LOSS_DETECTION;
 }
 
 /* If we detect SACK reneging, forget all SACK information
@@ -4060,7 +4061,7 @@ void tcp_parse_options(const struct net *net,
 				break;
 			case TCPOPT_WINDOW:
 				if (opsize == TCPOLEN_WINDOW && th->syn &&
-				    !estab && net->ipv4.sysctl_tcp_window_scaling) {
+				    !estab && READ_ONCE(net->ipv4.sysctl_tcp_window_scaling)) {
 					__u8 snd_wscale = *(__u8 *)ptr;
 					opt_rx->wscale_ok = 1;
 					if (snd_wscale > TCP_MAX_WSCALE) {
@@ -4076,7 +4077,7 @@ void tcp_parse_options(const struct net *net,
 			case TCPOPT_TIMESTAMP:
 				if ((opsize == TCPOLEN_TIMESTAMP) &&
 				    ((estab && opt_rx->tstamp_ok) ||
-				     (!estab && net->ipv4.sysctl_tcp_timestamps))) {
+				     (!estab && READ_ONCE(net->ipv4.sysctl_tcp_timestamps)))) {
 					opt_rx->saw_tstamp = 1;
 					opt_rx->rcv_tsval = get_unaligned_be32(ptr);
 					opt_rx->rcv_tsecr = get_unaligned_be32(ptr + 4);
@@ -4084,7 +4085,7 @@ void tcp_parse_options(const struct net *net,
 				break;
 			case TCPOPT_SACK_PERM:
 				if (opsize == TCPOLEN_SACK_PERM && th->syn &&
-				    !estab && net->ipv4.sysctl_tcp_sack) {
+				    !estab && READ_ONCE(net->ipv4.sysctl_tcp_sack)) {
 					opt_rx->sack_ok = TCP_SACK_SEEN;
 					tcp_sack_reset(opt_rx);
 				}
@@ -5571,7 +5572,7 @@ static void tcp_check_urg(struct sock *sk, const struct tcphdr *th)
 	struct tcp_sock *tp = tcp_sk(sk);
 	u32 ptr = ntohs(th->urg_ptr);
 
-	if (ptr && !sock_net(sk)->ipv4.sysctl_tcp_stdurg)
+	if (ptr && !READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_stdurg))
 		ptr--;
 	ptr += ntohl(th->seq);
 
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index 6854bb1fb32b..cb95d88497ae 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -173,7 +173,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
 			 * Oh well... nobody has a sufficient solution to this
 			 * protocol bug yet.
 			 */
-			if (twsk_net(tw)->ipv4.sysctl_tcp_rfc1337 == 0) {
+			if (!READ_ONCE(twsk_net(tw)->ipv4.sysctl_tcp_rfc1337)) {
 kill:
 				inet_twsk_deschedule_put(tw);
 				return TCP_TW_SUCCESS;
@@ -781,7 +781,7 @@ listen_overflow:
 	if (sk != req->rsk_listener)
 		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMIGRATEREQFAILURE);
 
-	if (!sock_net(sk)->ipv4.sysctl_tcp_abort_on_overflow) {
+	if (!READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_abort_on_overflow)) {
 		inet_rsk(req)->acked = 1;
 		return NULL;
 	}
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 3b3552d292a5..c38e07b50639 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -791,18 +791,18 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb,
 	opts->mss = tcp_advertise_mss(sk);
 	remaining -= TCPOLEN_MSS_ALIGNED;
 
-	if (likely(sock_net(sk)->ipv4.sysctl_tcp_timestamps && !*md5)) {
+	if (likely(READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_timestamps) && !*md5)) {
 		opts->options |= OPTION_TS;
 		opts->tsval = tcp_skb_timestamp(skb) + tp->tsoffset;
 		opts->tsecr = tp->rx_opt.ts_recent;
 		remaining -= TCPOLEN_TSTAMP_ALIGNED;
 	}
-	if (likely(sock_net(sk)->ipv4.sysctl_tcp_window_scaling)) {
+	if (likely(READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_window_scaling))) {
 		opts->ws = tp->rx_opt.rcv_wscale;
 		opts->options |= OPTION_WSCALE;
 		remaining -= TCPOLEN_WSCALE_ALIGNED;
 	}
-	if (likely(sock_net(sk)->ipv4.sysctl_tcp_sack)) {
+	if (likely(READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_sack))) {
 		opts->options |= OPTION_SACK_ADVERTISE;
 		if (unlikely(!(OPTION_TS & opts->options)))
 			remaining -= TCPOLEN_SACKPERM_ALIGNED;
@@ -1898,7 +1898,7 @@ static void tcp_cwnd_validate(struct sock *sk, bool is_cwnd_limited)
 		if (tp->packets_out > tp->snd_cwnd_used)
 			tp->snd_cwnd_used = tp->packets_out;
 
-		if (sock_net(sk)->ipv4.sysctl_tcp_slow_start_after_idle &&
+		if (READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_slow_start_after_idle) &&
 		    (s32)(tcp_jiffies32 - tp->snd_cwnd_stamp) >= inet_csk(sk)->icsk_rto &&
 		    !ca_ops->cong_control)
 			tcp_cwnd_application_limited(sk);
@@ -2741,7 +2741,7 @@ bool tcp_schedule_loss_probe(struct sock *sk, bool advancing_rto)
 	if (rcu_access_pointer(tp->fastopen_rsk))
 		return false;
 
-	early_retrans = sock_net(sk)->ipv4.sysctl_tcp_early_retrans;
+	early_retrans = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_early_retrans);
 	/* Schedule a loss probe in 2*RTT for SACK capable connections
 	 * not in loss recovery, that are either limited by cwnd or application.
 	 */
@@ -3105,7 +3105,7 @@ static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *to,
 	struct sk_buff *skb = to, *tmp;
 	bool first = true;
 
-	if (!sock_net(sk)->ipv4.sysctl_tcp_retrans_collapse)
+	if (!READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_retrans_collapse))
 		return;
 	if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_SYN)
 		return;
@@ -3647,7 +3647,7 @@ static void tcp_connect_init(struct sock *sk)
 	 * See tcp_input.c:tcp_rcv_state_process case TCP_SYN_SENT.
 	 */
 	tp->tcp_header_len = sizeof(struct tcphdr);
-	if (sock_net(sk)->ipv4.sysctl_tcp_timestamps)
+	if (READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_timestamps))
 		tp->tcp_header_len += TCPOLEN_TSTAMP_ALIGNED;
 
 #ifdef CONFIG_TCP_MD5SIG
@@ -3683,7 +3683,7 @@ static void tcp_connect_init(struct sock *sk)
 				  tp->advmss - (tp->rx_opt.ts_recent_stamp ? tp->tcp_header_len - sizeof(struct tcphdr) : 0),
 				  &tp->rcv_wnd,
 				  &tp->window_clamp,
-				  sock_net(sk)->ipv4.sysctl_tcp_window_scaling,
+				  READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_window_scaling),
 				  &rcv_wscale,
 				  rcv_wnd);
 
diff --git a/net/ipv4/tcp_recovery.c b/net/ipv4/tcp_recovery.c
index 48f30e7209f2..50abaa941387 100644
--- a/net/ipv4/tcp_recovery.c
+++ b/net/ipv4/tcp_recovery.c
@@ -14,7 +14,8 @@ static u32 tcp_rack_reo_wnd(const struct sock *sk)
 			return 0;
 
 		if (tp->sacked_out >= tp->reordering &&
-		    !(sock_net(sk)->ipv4.sysctl_tcp_recovery & TCP_RACK_NO_DUPTHRESH))
+		    !(READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_recovery) &
+		      TCP_RACK_NO_DUPTHRESH))
 			return 0;
 	}
 
@@ -187,7 +188,8 @@ void tcp_rack_update_reo_wnd(struct sock *sk, struct rate_sample *rs)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 
-	if (sock_net(sk)->ipv4.sysctl_tcp_recovery & TCP_RACK_STATIC_REO_WND ||
+	if ((READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_recovery) &
+	     TCP_RACK_STATIC_REO_WND) ||
 	    !rs->prior_delivered)
 		return;
 
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index ec5277becc6a..50bba370486e 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -578,7 +578,7 @@ out_reset_timer:
 	 * linear-timeout retransmissions into a black hole
 	 */
 	if (sk->sk_state == TCP_ESTABLISHED &&
-	    (tp->thin_lto || net->ipv4.sysctl_tcp_thin_linear_timeouts) &&
+	    (tp->thin_lto || READ_ONCE(net->ipv4.sysctl_tcp_thin_linear_timeouts)) &&
 	    tcp_stream_is_thin(tp) &&
 	    icsk->icsk_retransmits <= TCP_THIN_LINEAR_RETRIES) {
 		icsk->icsk_backoff = 0;