From: Eric Dumazet edumazet@google.com
[ Upstream commit 348b81b68b13ebd489a3e6a46aa1c384c731c919 ]
do_tcp_getsockopt() reads tp->tcp_tx_delay while another cpu might change its value.
Fixes: a842fe1425cb ("tcp: add optional per socket transmit delay") Signed-off-by: Eric Dumazet edumazet@google.com Link: https://lore.kernel.org/r/20230719212857.3943972-2-edumazet@google.com Signed-off-by: Jakub Kicinski kuba@kernel.org Signed-off-by: Sasha Levin sashal@kernel.org --- net/ipv4/tcp.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index fdf2ddc4864df..e33abcff56080 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -3177,7 +3177,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level, case TCP_TX_DELAY: if (val) tcp_enable_tx_delay(); - tp->tcp_tx_delay = val; + WRITE_ONCE(tp->tcp_tx_delay, val); break; default: err = -ENOPROTOOPT; @@ -3634,7 +3634,7 @@ static int do_tcp_getsockopt(struct sock *sk, int level, break;
case TCP_TX_DELAY: - val = tp->tcp_tx_delay; + val = READ_ONCE(tp->tcp_tx_delay); break;
case TCP_TIMESTAMP: