The error handling path at label 'accept_err' already performs these
'sock_release()' and 'mutex_unlock()' calls. The order is reversed, but
it is not important.
So avoid code duplication and 'goto accept_err'.
While at it, add some missing spaces around a '='.
Signed-off-by: Christophe JAILLET <christophe.jail...@wanadoo.fr>
---
Instead of returning the magic number -1, we could also propagate the
return value of 'addr_to_nodeid()', that is to say -EEXIST
---
fs/dlm/lowcomms.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 4813d0e0cd9b..cd9305b17ad5 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -758,13 +758,12 @@ static int tcp_accept_from_sock(struct connection *con)
/* Get the new node's NODEID */
make_sockaddr(&peeraddr, 0, &len);
if (addr_to_nodeid(&peeraddr, &nodeid)) {
- unsigned char *b=(unsigned char *)&peeraddr;
+ unsigned char *b = (unsigned char *)&peeraddr;
log_print("connect from non cluster node");
print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
b, sizeof(struct sockaddr_storage));
- sock_release(newsock);
- mutex_unlock(&con->sock_mutex);
- return -1;
+ result = -1;
+ goto accept_err;
}
log_print("got connection from %d", nodeid);
--
2.11.0
