Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 9bffc4ac authored by Neil Horman's avatar Neil Horman Committed by David S. Miller
Browse files

[SCTP]: Fix sctp to not return erroneous POLLOUT events. 



Make sctp_writeable() use sk_wmem_alloc rather than sk_wmem_queued to
determine the sndbuf space available. It also removes all the modifications
to sk_wmem_queued as it is not currently used in SCTP.

Signed-off-by: default avatarNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: default avatarSridhar Samudrala <sri@us.ibm.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 399c180a

net/sctp/socket.c

+3 −11
Original line number Diff line number Diff line
@@ -156,10 +156,6 @@ static inline void sctp_set_owner_w(struct sctp_chunk *chunk) 
				sizeof(struct sk_buff) +

				sizeof(struct sctp_chunk);



	sk->sk_wmem_queued += SCTP_DATA_SNDSIZE(chunk) +

				sizeof(struct sk_buff) +

				sizeof(struct sctp_chunk);



	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);

}
@@ -4426,7 +4422,7 @@ cleanup: 
 * tcp_poll().  Note that, based on these implementations, we don't

 * lock the socket in this function, even though it seems that,

 * ideally, locking or some other mechanisms can be used to ensure

 * the integrity of the counters (sndbuf and wmem_queued) used

 * the integrity of the counters (sndbuf and wmem_alloc) used

 * in this place.  We assume that we don't need locks either until proven

 * otherwise.

 *
@@ -4833,10 +4829,6 @@ static void sctp_wfree(struct sk_buff *skb) 
				sizeof(struct sk_buff) +

				sizeof(struct sctp_chunk);



	sk->sk_wmem_queued -= SCTP_DATA_SNDSIZE(chunk) +

				sizeof(struct sk_buff) +

				sizeof(struct sctp_chunk);



	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);



	sock_wfree(skb);
@@ -4920,7 +4912,7 @@ void sctp_write_space(struct sock *sk) 


/* Is there any sndbuf space available on the socket?

 *

 * Note that wmem_queued is the sum of the send buffers on all of the

 * Note that sk_wmem_alloc is the sum of the send buffers on all of the

 * associations on the same socket.  For a UDP-style socket with

 * multiple associations, it is possible for it to be "unwriteable"

 * prematurely.  I assume that this is acceptable because
@@ -4933,7 +4925,7 @@ static int sctp_writeable(struct sock *sk) 
{

	int amt = 0;



	amt = sk->sk_sndbuf - sk->sk_wmem_queued;

	amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);

	if (amt < 0)

		amt = 0;

	return amt;