[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CANn89iK9pDX=dA78Bk-sm8p4xxSno6XgHT=s0epSes=WLwxOZA@mail.gmail.com>
Date: Fri, 5 Apr 2024 16:49:07 +0200
From: Eric Dumazet <edumazet@...gle.com>
To: Jason Xing <kerneljasonxing@...il.com>
Cc: "David S . Miller" <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>, netdev@...r.kernel.org, eric.dumazet@...il.com
Subject: Re: [PATCH net-next] tcp: annotate data-races around tp->window_clamp
On Fri, Apr 5, 2024 at 4:29 PM Jason Xing <kerneljasonxing@...il.com> wrote:
>
> On Thu, Apr 4, 2024 at 7:53 PM Eric Dumazet <edumazet@...gle.com> wrote:
> >
> > tp->window_clamp can be read locklessly, add READ_ONCE()
> > and WRITE_ONCE() annotations.
> >
> > Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> > ---
> > net/ipv4/syncookies.c | 3 ++-
> > net/ipv4/tcp.c | 8 ++++----
> > net/ipv4/tcp_input.c | 17 ++++++++++-------
> > net/ipv4/tcp_output.c | 18 ++++++++++--------
> > net/ipv6/syncookies.c | 2 +-
> > net/mptcp/protocol.c | 2 +-
> > net/mptcp/sockopt.c | 2 +-
> > 7 files changed, 29 insertions(+), 23 deletions(-)
> >
> > diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
> > index 500f665f98cbce4a3d681f8e39ecd368fe4013b1..b61d36810fe3fd62b1e5c5885bbaf20185f1abf0 100644
> > --- a/net/ipv4/syncookies.c
> > +++ b/net/ipv4/syncookies.c
> > @@ -462,7 +462,8 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)
> > }
> >
> > /* Try to redo what tcp_v4_send_synack did. */
> > - req->rsk_window_clamp = tp->window_clamp ? :dst_metric(&rt->dst, RTAX_WINDOW);
> > + req->rsk_window_clamp = READ_ONCE(tp->window_clamp) ? :
> > + dst_metric(&rt->dst, RTAX_WINDOW);
> > /* limit the window selection if the user enforce a smaller rx buffer */
> > full_space = tcp_full_space(sk);
> > if (sk->sk_userlocks & SOCK_RCVBUF_LOCK &&
> > diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> > index e767721b3a588b5d56567ae7badf5dffcd35a76a..92ee60492314a1483cfbfa2f73d32fcad5632773 100644
> > --- a/net/ipv4/tcp.c
> > +++ b/net/ipv4/tcp.c
> > @@ -1721,7 +1721,7 @@ int tcp_set_rcvlowat(struct sock *sk, int val)
> > space = tcp_space_from_win(sk, val);
> > if (space > sk->sk_rcvbuf) {
> > WRITE_ONCE(sk->sk_rcvbuf, space);
> > - tcp_sk(sk)->window_clamp = val;
> > + WRITE_ONCE(tcp_sk(sk)->window_clamp, val);
> > }
> > return 0;
> > }
> > @@ -3379,7 +3379,7 @@ int tcp_set_window_clamp(struct sock *sk, int val)
> > if (!val) {
> > if (sk->sk_state != TCP_CLOSE)
> > return -EINVAL;
> > - tp->window_clamp = 0;
> > + WRITE_ONCE(tp->window_clamp, 0);
> > } else {
> > u32 new_rcv_ssthresh, old_window_clamp = tp->window_clamp;
> > u32 new_window_clamp = val < SOCK_MIN_RCVBUF / 2 ?
> > @@ -3388,7 +3388,7 @@ int tcp_set_window_clamp(struct sock *sk, int val)
> > if (new_window_clamp == old_window_clamp)
> > return 0;
> >
> > - tp->window_clamp = new_window_clamp;
> > + WRITE_ONCE(tp->window_clamp, new_window_clamp);
> > if (new_window_clamp < old_window_clamp) {
> > /* need to apply the reserved mem provisioning only
> > * when shrinking the window clamp
> > @@ -4057,7 +4057,7 @@ int do_tcp_getsockopt(struct sock *sk, int level,
> > TCP_RTO_MAX / HZ);
> > break;
> > case TCP_WINDOW_CLAMP:
> > - val = tp->window_clamp;
> > + val = READ_ONCE(tp->window_clamp);
> > break;
> > case TCP_INFO: {
> > struct tcp_info info;
> > diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> > index 1b6cd384001202df5f8e8e8c73adff0db89ece63..8d44ab5671eacd4bc06647c7cca387a79e346618 100644
> > --- a/net/ipv4/tcp_input.c
> > +++ b/net/ipv4/tcp_input.c
> > @@ -563,19 +563,20 @@ static void tcp_init_buffer_space(struct sock *sk)
> > maxwin = tcp_full_space(sk);
> >
> > if (tp->window_clamp >= maxwin) {
>
> I wonder if it is necessary to locklessly protect the above line with
> READ_ONCE() because I saw the full reader protection in the
> tcp_select_initial_window()? There are some other places like this.
> Any special reason?
We hold the socket lock at this point.
READ_ONCE() is only needed if another thread can potentially change
the value under us.
Powered by blists - more mailing lists