tcp: no longer hold ehash lock while calling tcp_get_info()

We had various problems in the past in tcp_get_info() and used
specific synchronization to avoid deadlocks.

We would like to add more instrumentation points for TCP, and
avoiding grabing socket lock in tcp_getinfo() was too costly.

Being able to lock the socket allows to provide consistent set
of fields.

inet_diag_dump_icsk() can make sure ehash locks are not
held any more when tcp_get_info() is called.

We can remove syncp added in commit d654976cbf85
("tcp: fix a potential deadlock in tcp_get_info()"), but we need
to use lock_sock_fast() instead of spin_lock_bh() since TCP input
path can now be run from process context.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 9 insertions, 11 deletions

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 117982be0cab..a7d54cbcdabb 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -405,7 +405,6 @@ void tcp_init_sock(struct sock *sk)
 	tp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
 	tp->snd_cwnd_clamp = ~0;
 	tp->mss_cache = TCP_MSS_DEFAULT;
-	u64_stats_init(&tp->syncp);
 
 	tp->reordering = sock_net(sk)->ipv4.sysctl_tcp_reordering;
 	tcp_enable_early_retrans(tp);
@@ -2710,9 +2709,8 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 	const struct tcp_sock *tp = tcp_sk(sk); /* iff sk_type == SOCK_STREAM */
 	const struct inet_connection_sock *icsk = inet_csk(sk);
 	u32 now = tcp_time_stamp, intv;
-	unsigned int start;
-	int notsent_bytes;
 	u64 rate64;
+	bool slow;
 	u32 rate;
 
 	memset(info, 0, sizeof(*info));
@@ -2792,17 +2790,17 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 
 	info->tcpi_total_retrans = tp->total_retrans;
 
-	do {
-		start = u64_stats_fetch_begin_irq(&tp->syncp);
-		put_unaligned(tp->bytes_acked, &info->tcpi_bytes_acked);
-		put_unaligned(tp->bytes_received, &info->tcpi_bytes_received);
-	} while (u64_stats_fetch_retry_irq(&tp->syncp, start));
+	slow = lock_sock_fast(sk);
+
+	put_unaligned(tp->bytes_acked, &info->tcpi_bytes_acked);
+	put_unaligned(tp->bytes_received, &info->tcpi_bytes_received);
+	info->tcpi_notsent_bytes = max_t(int, 0, tp->write_seq - tp->snd_nxt);
+
+	unlock_sock_fast(sk, slow);
+
 	info->tcpi_segs_out = tp->segs_out;
 	info->tcpi_segs_in = tp->segs_in;
 
-	notsent_bytes = READ_ONCE(tp->write_seq) - READ_ONCE(tp->snd_nxt);
-	info->tcpi_notsent_bytes = max(0, notsent_bytes);
-
 	info->tcpi_min_rtt = tcp_min_rtt(tp);
 	info->tcpi_data_segs_in = tp->data_segs_in;
 	info->tcpi_data_segs_out = tp->data_segs_out;