@@ -1570,20 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
extern int drbd_connected(struct drbd_peer_device *);
-static inline void drbd_tcp_cork(struct socket *sock)
-{
- int val = 1;
- (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
- (char*)&val, sizeof(val));
-}
-
-static inline void drbd_tcp_uncork(struct socket *sock)
-{
- int val = 0;
- (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
- (char*)&val, sizeof(val));
-}
-
static inline void drbd_tcp_nodelay(struct socket *sock)
{
int val = 1;
@@ -6162,7 +6162,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
rcu_read_unlock();
if (tcp_cork)
- drbd_tcp_cork(connection->meta.socket);
+ tcp_sock_set_cork(connection->meta.socket->sk, true);
err = drbd_finish_peer_reqs(device);
kref_put(&device->kref, drbd_destroy_device);
@@ -6175,7 +6175,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
}
if (tcp_cork)
- drbd_tcp_uncork(connection->meta.socket);
+ tcp_sock_set_cork(connection->meta.socket->sk, false);
return;
}
@@ -2098,7 +2098,7 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
if (uncork) {
mutex_lock(&connection->data.mutex);
if (connection->data.socket)
- drbd_tcp_uncork(connection->data.socket);
+ tcp_sock_set_cork(connection->data.socket->sk, false);
mutex_unlock(&connection->data.mutex);
}
@@ -2153,9 +2153,9 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
mutex_lock(&connection->data.mutex);
if (connection->data.socket) {
if (cork)
- drbd_tcp_cork(connection->data.socket);
+ tcp_sock_set_cork(connection->data.socket->sk, true);
else if (!uncork)
- drbd_tcp_uncork(connection->data.socket);
+ tcp_sock_set_cork(connection->data.socket->sk, false);
}
mutex_unlock(&connection->data.mutex);
}
@@ -325,7 +325,6 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
size_t total_len = 0, sent, size;
struct socket *ssocket = server->ssocket;
struct msghdr smb_msg;
- int val = 1;
__be32 rfc1002_marker;
if (cifs_rdma_enabled(server)) {
@@ -345,8 +344,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
}
/* cork the socket */
- kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
- (char *)&val, sizeof(val));
+ tcp_sock_set_cork(ssocket->sk, true);
for (j = 0; j < num_rqst; j++)
send_length += smb_rqst_len(server, &rqst[j]);
@@ -435,9 +433,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
}
/* uncork it */
- val = 0;
- kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
- (char *)&val, sizeof(val));
+ tcp_sock_set_cork(ssocket->sk, false);
if ((total_len > 0) && (total_len != send_length)) {
cifs_dbg(FYI, "partial send (wanted=%u sent=%zu): terminating session\n",
@@ -497,4 +497,6 @@ static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
int shiftlen);
+void tcp_sock_set_cork(struct sock *sk, bool on);
+
#endif /* _LINUX_TCP_H */
@@ -2801,6 +2801,37 @@ static void tcp_enable_tx_delay(void)
}
}
+/* When set indicates to always queue non-full frames. Later the user clears
+ * this option and we transmit any pending partial frames in the queue. This is
+ * meant to be used alongside sendfile() to get properly filled frames when the
+ * user (for example) must write out headers with a write() call first and then
+ * use sendfile to send out the data parts.
+ *
+ * TCP_CORK can be set together with TCP_NODELAY and it is stronger than
+ * TCP_NODELAY.
+ */
+static void __tcp_sock_set_cork(struct sock *sk, bool on)
+{
+ struct tcp_sock *tp = tcp_sk(sk);
+
+ if (on) {
+ tp->nonagle |= TCP_NAGLE_CORK;
+ } else {
+ tp->nonagle &= ~TCP_NAGLE_CORK;
+ if (tp->nonagle & TCP_NAGLE_OFF)
+ tp->nonagle |= TCP_NAGLE_PUSH;
+ tcp_push_pending_frames(sk);
+ }
+}
+
+void tcp_sock_set_cork(struct sock *sk, bool on)
+{
+ lock_sock(sk);
+ __tcp_sock_set_cork(sk, on);
+ release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_cork);
+
/*
* Socket option code for TCP.
*/
@@ -2979,25 +3010,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
break;
case TCP_CORK:
- /* When set indicates to always queue non-full frames.
- * Later the user clears this option and we transmit
- * any pending partial frames in the queue. This is
- * meant to be used alongside sendfile() to get properly
- * filled frames when the user (for example) must write
- * out headers with a write() call first and then use
- * sendfile to send out the data parts.
- *
- * TCP_CORK can be set together with TCP_NODELAY and it is
- * stronger than TCP_NODELAY.
- */
- if (val) {
- tp->nonagle |= TCP_NAGLE_CORK;
- } else {
- tp->nonagle &= ~TCP_NAGLE_CORK;
- if (tp->nonagle&TCP_NAGLE_OFF)
- tp->nonagle |= TCP_NAGLE_PUSH;
- tcp_push_pending_frames(sk);
- }
+ __tcp_sock_set_cork(sk, val);
break;
case TCP_KEEPIDLE:
@@ -38,23 +38,18 @@
#include "rds.h"
#include "tcp.h"
-static void rds_tcp_cork(struct socket *sock, int val)
-{
- kernel_setsockopt(sock, SOL_TCP, TCP_CORK, (void *)&val, sizeof(val));
-}
-
void rds_tcp_xmit_path_prepare(struct rds_conn_path *cp)
{
struct rds_tcp_connection *tc = cp->cp_transport_data;
- rds_tcp_cork(tc->t_sock, 1);
+ tcp_sock_set_cork(tc->t_sock->sk, true);
}
void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
{
struct rds_tcp_connection *tc = cp->cp_transport_data;
- rds_tcp_cork(tc->t_sock, 0);
+ tcp_sock_set_cork(tc->t_sock->sk, false);
}
/* the core send_sem serializes this with other xmit and shutdown */
Add a helper to directly set the TCP_CORK sockopt from kernel space without going through a fake uaccess. Cleanup the callers to avoid pointless wrappers now that this is a simple function call. Signed-off-by: Christoph Hellwig <hch@lst.de> --- drivers/block/drbd/drbd_int.h | 14 -------- drivers/block/drbd/drbd_receiver.c | 4 +-- drivers/block/drbd/drbd_worker.c | 6 ++-- fs/cifs/transport.c | 8 ++--- include/linux/tcp.h | 2 ++ net/ipv4/tcp.c | 51 +++++++++++++++++++----------- net/rds/tcp_send.c | 9 ++---- 7 files changed, 43 insertions(+), 51 deletions(-)