Commit fde56eea authored by Paolo Abeni's avatar Paolo Abeni Committed by David S. Miller

mptcp: refine mptcp_cleanup_rbuf

The current cleanup rbuf tries a bit too hard to avoid acquiring
the subflow socket lock. We may end-up delaying the needed ack,
or skip acking a blocked subflow.

Address the above extending the conditions used to trigger the cleanup
to reflect more closely what TCP does and invoking tcp_cleanup_rbuf()
on all the active subflows.

Note that we can't replicate the exact tests implemented in
tcp_cleanup_rbuf(), as MPTCP lacks some of the required info - e.g.
ping-pong mode.
Signed-off-by: default avatarPaolo Abeni <pabeni@redhat.com>
Signed-off-by: default avatarMat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent d8e336f7
...@@ -442,49 +442,46 @@ static void mptcp_send_ack(struct mptcp_sock *msk) ...@@ -442,49 +442,46 @@ static void mptcp_send_ack(struct mptcp_sock *msk)
} }
} }
static bool mptcp_subflow_cleanup_rbuf(struct sock *ssk) static void mptcp_subflow_cleanup_rbuf(struct sock *ssk)
{ {
bool slow; bool slow;
int ret;
slow = lock_sock_fast(ssk); slow = lock_sock_fast(ssk);
ret = tcp_can_send_ack(ssk); if (tcp_can_send_ack(ssk))
if (ret)
tcp_cleanup_rbuf(ssk, 1); tcp_cleanup_rbuf(ssk, 1);
unlock_sock_fast(ssk, slow); unlock_sock_fast(ssk, slow);
return ret; }
static bool mptcp_subflow_could_cleanup(const struct sock *ssk, bool rx_empty)
{
const struct inet_connection_sock *icsk = inet_csk(ssk);
bool ack_pending = READ_ONCE(icsk->icsk_ack.pending);
const struct tcp_sock *tp = tcp_sk(ssk);
return (ack_pending & ICSK_ACK_SCHED) &&
((READ_ONCE(tp->rcv_nxt) - READ_ONCE(tp->rcv_wup) >
READ_ONCE(icsk->icsk_ack.rcv_mss)) ||
(rx_empty && ack_pending &
(ICSK_ACK_PUSHED2 | ICSK_ACK_PUSHED)));
} }
static void mptcp_cleanup_rbuf(struct mptcp_sock *msk) static void mptcp_cleanup_rbuf(struct mptcp_sock *msk)
{ {
struct sock *ack_hint = READ_ONCE(msk->ack_hint);
int old_space = READ_ONCE(msk->old_wspace); int old_space = READ_ONCE(msk->old_wspace);
struct mptcp_subflow_context *subflow; struct mptcp_subflow_context *subflow;
struct sock *sk = (struct sock *)msk; struct sock *sk = (struct sock *)msk;
bool cleanup; int space = __mptcp_space(sk);
bool cleanup, rx_empty;
/* this is a simple superset of what tcp_cleanup_rbuf() implements cleanup = (space > 0) && (space >= (old_space << 1));
* so that we don't have to acquire the ssk socket lock most of the time rx_empty = !atomic_read(&sk->sk_rmem_alloc);
* to do actually nothing
*/
cleanup = __mptcp_space(sk) - old_space >= max(0, old_space);
if (!cleanup)
return;
/* if the hinted ssk is still active, try to use it */
if (likely(ack_hint)) {
mptcp_for_each_subflow(msk, subflow) { mptcp_for_each_subflow(msk, subflow) {
struct sock *ssk = mptcp_subflow_tcp_sock(subflow); struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
if (ack_hint == ssk && mptcp_subflow_cleanup_rbuf(ssk)) if (cleanup || mptcp_subflow_could_cleanup(ssk, rx_empty))
return; mptcp_subflow_cleanup_rbuf(ssk);
} }
}
/* otherwise pick the first active subflow */
mptcp_for_each_subflow(msk, subflow)
if (mptcp_subflow_cleanup_rbuf(mptcp_subflow_tcp_sock(subflow)))
return;
} }
static bool mptcp_check_data_fin(struct sock *sk) static bool mptcp_check_data_fin(struct sock *sk)
...@@ -629,7 +626,6 @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk, ...@@ -629,7 +626,6 @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
break; break;
} }
} while (more_data_avail); } while (more_data_avail);
WRITE_ONCE(msk->ack_hint, ssk);
*bytes += moved; *bytes += moved;
return done; return done;
...@@ -1910,7 +1906,6 @@ static bool __mptcp_move_skbs(struct mptcp_sock *msk) ...@@ -1910,7 +1906,6 @@ static bool __mptcp_move_skbs(struct mptcp_sock *msk)
__mptcp_update_rmem(sk); __mptcp_update_rmem(sk);
done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved); done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
mptcp_data_unlock(sk); mptcp_data_unlock(sk);
tcp_cleanup_rbuf(ssk, moved);
if (unlikely(ssk->sk_err)) if (unlikely(ssk->sk_err))
__mptcp_error_report(sk); __mptcp_error_report(sk);
...@@ -1926,7 +1921,6 @@ static bool __mptcp_move_skbs(struct mptcp_sock *msk) ...@@ -1926,7 +1921,6 @@ static bool __mptcp_move_skbs(struct mptcp_sock *msk)
ret |= __mptcp_ofo_queue(msk); ret |= __mptcp_ofo_queue(msk);
__mptcp_splice_receive_queue(sk); __mptcp_splice_receive_queue(sk);
mptcp_data_unlock(sk); mptcp_data_unlock(sk);
mptcp_cleanup_rbuf(msk);
} }
if (ret) if (ret)
mptcp_check_data_fin((struct sock *)msk); mptcp_check_data_fin((struct sock *)msk);
...@@ -2175,9 +2169,6 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk, ...@@ -2175,9 +2169,6 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk,
if (ssk == msk->last_snd) if (ssk == msk->last_snd)
msk->last_snd = NULL; msk->last_snd = NULL;
if (ssk == msk->ack_hint)
msk->ack_hint = NULL;
if (ssk == msk->first) if (ssk == msk->first)
msk->first = NULL; msk->first = NULL;
...@@ -2392,7 +2383,6 @@ static int __mptcp_init_sock(struct sock *sk) ...@@ -2392,7 +2383,6 @@ static int __mptcp_init_sock(struct sock *sk)
msk->rmem_released = 0; msk->rmem_released = 0;
msk->tx_pending_data = 0; msk->tx_pending_data = 0;
msk->ack_hint = NULL;
msk->first = NULL; msk->first = NULL;
inet_csk(sk)->icsk_sync_mss = mptcp_sync_mss; inet_csk(sk)->icsk_sync_mss = mptcp_sync_mss;
WRITE_ONCE(msk->csum_enabled, mptcp_is_checksum_enabled(sock_net(sk))); WRITE_ONCE(msk->csum_enabled, mptcp_is_checksum_enabled(sock_net(sk)));
......
...@@ -243,7 +243,6 @@ struct mptcp_sock { ...@@ -243,7 +243,6 @@ struct mptcp_sock {
bool use_64bit_ack; /* Set when we received a 64-bit DSN */ bool use_64bit_ack; /* Set when we received a 64-bit DSN */
bool csum_enabled; bool csum_enabled;
spinlock_t join_list_lock; spinlock_t join_list_lock;
struct sock *ack_hint;
struct work_struct work; struct work_struct work;
struct sk_buff *ooo_last_skb; struct sk_buff *ooo_last_skb;
struct rb_root out_of_order_queue; struct rb_root out_of_order_queue;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment