From 8dbd76e79a16b45b2ccb01d2f2e08dbf64e71e40 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 13 Dec 2019 18:20:41 -0800 Subject: tcp/dccp: fix possible race __inet_lookup_established() Michal Kubecek and Firo Yang did a very nice analysis of crashes happening in __inet_lookup_established(). Since a TCP socket can go from TCP_ESTABLISH to TCP_LISTEN (via a close()/socket()/listen() cycle) without a RCU grace period, I should not have changed listeners linkage in their hash table. They must use the nulls protocol (Documentation/RCU/rculist_nulls.txt), so that a lookup can detect a socket in a hash list was moved in another one. Since we added code in commit d296ba60d8e2 ("soreuseport: Resolve merge conflict for v4/v6 ordering fix"), we have to add hlist_nulls_add_tail_rcu() helper. Fixes: 3b24d854cb35 ("tcp/dccp: do not touch listener sk_refcnt under synflood") Signed-off-by: Eric Dumazet Reported-by: Michal Kubecek Reported-by: Firo Yang Reviewed-by: Michal Kubecek Link: https://lore.kernel.org/netdev/20191120083919.GH27852@unicorn.suse.cz/ Signed-off-by: Jakub Kicinski --- include/net/sock.h | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 87d54ef57f00..04c274a20620 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -722,6 +722,11 @@ static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_h hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); } +static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list) +{ + hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list); +} + static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) { sock_hold(sk); -- cgit v1.2.3 From 7c68fa2bddda6d942bd387c9ba5b4300737fd991 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Mon, 16 Dec 2019 18:51:03 -0800 Subject: net: annotate lockless accesses to sk->sk_pacing_shift sk->sk_pacing_shift can be read and written without lock synchronization. This patch adds annotations to document this fact and avoid future syzbot complains. This might also avoid unexpected false sharing in sk_pacing_shift_update(), as the compiler could remove the conditional check and always write over sk->sk_pacing_shift : if (sk->sk_pacing_shift != val) sk->sk_pacing_shift = val; Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 4 ++-- net/core/sock.c | 2 +- net/ipv4/tcp_bbr.c | 3 ++- net/ipv4/tcp_output.c | 4 ++-- 4 files changed, 7 insertions(+), 6 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 04c274a20620..22be668457bf 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -2588,9 +2588,9 @@ static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto) */ static inline void sk_pacing_shift_update(struct sock *sk, int val) { - if (!sk || !sk_fullsock(sk) || sk->sk_pacing_shift == val) + if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val) return; - sk->sk_pacing_shift = val; + WRITE_ONCE(sk->sk_pacing_shift, val); } /* if a socket is bound to a device, check that the given device diff --git a/net/core/sock.c b/net/core/sock.c index 043db3ce023e..8459ad579f73 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -2916,7 +2916,7 @@ void sock_init_data(struct socket *sock, struct sock *sk) sk->sk_max_pacing_rate = ~0UL; sk->sk_pacing_rate = ~0UL; - sk->sk_pacing_shift = 10; + WRITE_ONCE(sk->sk_pacing_shift, 10); sk->sk_incoming_cpu = -1; sk_rx_queue_clear(sk); diff --git a/net/ipv4/tcp_bbr.c b/net/ipv4/tcp_bbr.c index 32772d6ded4e..a6545ef0d27b 100644 --- a/net/ipv4/tcp_bbr.c +++ b/net/ipv4/tcp_bbr.c @@ -306,7 +306,8 @@ static u32 bbr_tso_segs_goal(struct sock *sk) /* Sort of tcp_tso_autosize() but ignoring * driver provided sk_gso_max_size. */ - bytes = min_t(unsigned long, sk->sk_pacing_rate >> sk->sk_pacing_shift, + bytes = min_t(unsigned long, + sk->sk_pacing_rate >> READ_ONCE(sk->sk_pacing_shift), GSO_MAX_SIZE - 1 - MAX_TCP_HEADER); segs = max_t(u32, bytes / tp->mss_cache, bbr_min_tso_segs(sk)); diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 36902d08473e..1f7735ca8f22 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -1725,7 +1725,7 @@ static u32 tcp_tso_autosize(const struct sock *sk, unsigned int mss_now, u32 bytes, segs; bytes = min_t(unsigned long, - sk->sk_pacing_rate >> sk->sk_pacing_shift, + sk->sk_pacing_rate >> READ_ONCE(sk->sk_pacing_shift), sk->sk_gso_max_size - 1 - MAX_TCP_HEADER); /* Goal is to send at least one packet per ms, @@ -2260,7 +2260,7 @@ static bool tcp_small_queue_check(struct sock *sk, const struct sk_buff *skb, limit = max_t(unsigned long, 2 * skb->truesize, - sk->sk_pacing_rate >> sk->sk_pacing_shift); + sk->sk_pacing_rate >> READ_ONCE(sk->sk_pacing_shift)); if (sk->sk_pacing_status == SK_PACING_NONE) limit = min_t(unsigned long, limit, sock_net(sk)->ipv4.sysctl_tcp_limit_output_bytes); -- cgit v1.2.3