summaryrefslogtreecommitdiffstats
path: root/net/iucv/af_iucv.c
diff options
context:
space:
mode:
Diffstat (limited to 'net/iucv/af_iucv.c')
-rw-r--r--net/iucv/af_iucv.c44
1 files changed, 23 insertions, 21 deletions
diff --git a/net/iucv/af_iucv.c b/net/iucv/af_iucv.c
index 026704a..2f13738 100644
--- a/net/iucv/af_iucv.c
+++ b/net/iucv/af_iucv.c
@@ -45,7 +45,8 @@ static struct proto iucv_proto = {
static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
-static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8], u8 ipuser[16]);
+static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
+ u8 ipuser[16]);
static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
static struct iucv_sock_list iucv_sk_list = {
@@ -152,7 +153,7 @@ static void iucv_sock_close(struct sock *sk)
iucv_sock_clear_timer(sk);
lock_sock(sk);
- switch(sk->sk_state) {
+ switch (sk->sk_state) {
case IUCV_LISTEN:
iucv_sock_cleanup_listen(sk);
break;
@@ -164,7 +165,7 @@ static void iucv_sock_close(struct sock *sk)
sk->sk_state = IUCV_CLOSING;
sk->sk_state_change(sk);
- if(!skb_queue_empty(&iucv->send_skb_q)) {
+ if (!skb_queue_empty(&iucv->send_skb_q)) {
if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
timeo = sk->sk_lingertime;
else
@@ -292,7 +293,7 @@ struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
struct iucv_sock *isk, *n;
struct sock *sk;
- list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q){
+ list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
sk = (struct sock *) isk;
lock_sock(sk);
@@ -537,7 +538,7 @@ static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
/* Wait for an incoming connection */
add_wait_queue_exclusive(sk->sk_sleep, &wait);
- while (!(nsk = iucv_accept_dequeue(sk, newsock))){
+ while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
set_current_state(TASK_INTERRUPTIBLE);
if (!timeo) {
err = -EAGAIN;
@@ -618,13 +619,13 @@ static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
goto out;
}
- if (sk->sk_state == IUCV_CONNECTED){
- if(!(skb = sock_alloc_send_skb(sk, len,
- msg->msg_flags & MSG_DONTWAIT,
- &err)))
+ if (sk->sk_state == IUCV_CONNECTED) {
+ if (!(skb = sock_alloc_send_skb(sk, len,
+ msg->msg_flags & MSG_DONTWAIT,
+ &err)))
goto out;
- if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)){
+ if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
err = -EFAULT;
goto fail;
}
@@ -710,7 +711,7 @@ static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
/* Queue backlog skbs */
rskb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
- while(rskb) {
+ while (rskb) {
if (sock_queue_rcv_skb(sk, rskb)) {
skb_queue_head(&iucv_sk(sk)->backlog_skb_q,
rskb);
@@ -731,7 +732,7 @@ static inline unsigned int iucv_accept_poll(struct sock *parent)
struct iucv_sock *isk, *n;
struct sock *sk;
- list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q){
+ list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
sk = (struct sock *) isk;
if (sk->sk_state == IUCV_CONNECTED)
@@ -762,7 +763,7 @@ unsigned int iucv_sock_poll(struct file *file, struct socket *sock,
mask |= POLLHUP;
if (!skb_queue_empty(&sk->sk_receive_queue) ||
- (sk->sk_shutdown & RCV_SHUTDOWN))
+ (sk->sk_shutdown & RCV_SHUTDOWN))
mask |= POLLIN | POLLRDNORM;
if (sk->sk_state == IUCV_CLOSED)
@@ -793,7 +794,7 @@ static int iucv_sock_shutdown(struct socket *sock, int how)
return -EINVAL;
lock_sock(sk);
- switch(sk->sk_state) {
+ switch (sk->sk_state) {
case IUCV_CLOSED:
err = -ENOTCONN;
goto fail;
@@ -809,7 +810,7 @@ static int iucv_sock_shutdown(struct socket *sock, int how)
err = iucv_message_send(iucv->path, &txmsg, IUCV_IPRMDATA, 0,
(void *) prmmsg, 8);
if (err) {
- switch(err) {
+ switch (err) {
case 1:
err = -ENOTCONN;
break;
@@ -912,7 +913,7 @@ static int iucv_callback_connreq(struct iucv_path *path,
/* Create the new socket */
nsk = iucv_sock_alloc(NULL, SOCK_STREAM, GFP_ATOMIC);
- if (!nsk){
+ if (!nsk) {
err = iucv_path_sever(path, user_data);
goto fail;
}
@@ -935,7 +936,7 @@ static int iucv_callback_connreq(struct iucv_path *path,
path->msglim = IUCV_QUEUELEN_DEFAULT;
err = iucv_path_accept(path, &af_iucv_handler, nuser_data, nsk);
- if (err){
+ if (err) {
err = iucv_path_sever(path, user_data);
goto fail;
}
@@ -966,7 +967,7 @@ static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len,
struct sk_buff *nskb;
dataleft = len;
- while(dataleft) {
+ while (dataleft) {
if (dataleft >= sk->sk_rcvbuf / 4)
size = sk->sk_rcvbuf / 4;
else
@@ -989,6 +990,7 @@ static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len,
return 0;
}
+
static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
{
struct sock *sk = path->private;
@@ -1035,7 +1037,7 @@ static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
}
/* Queue the fragmented skb */
fskb = skb_dequeue(&fragmented_skb_q);
- while(fskb) {
+ while (fskb) {
if (!skb_queue_empty(&iucv->backlog_skb_q))
skb_queue_tail(&iucv->backlog_skb_q, fskb);
else if (sock_queue_rcv_skb(sk, fskb))
@@ -1076,7 +1078,7 @@ static void iucv_callback_txdone(struct iucv_path *path,
kfree_skb(this);
}
- if (sk->sk_state == IUCV_CLOSING){
+ if (sk->sk_state == IUCV_CLOSING) {
if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
sk->sk_state = IUCV_CLOSED;
sk->sk_state_change(sk);
@@ -1123,7 +1125,7 @@ static struct net_proto_family iucv_sock_family_ops = {
.create = iucv_sock_create,
};
-static int afiucv_init(void)
+static int __init afiucv_init(void)
{
int err;
OpenPOWER on IntegriCloud