> > From: Marc-André Lureau <mlureau@xxxxxxxxxx> > > Send the fd associated to the last message sent. > > Even if the fd is invalid, the sendfd msg is appended to the protocol, > for 2 reasons: > - trying to send an invalid fd does not have to close the connection (it > would with an invalid fd) > - even if the fd is invalid, the protocol expects an extra byte for the > ancillary data > > Signed-off-by: Marc-André Lureau <marcandre.lureau@xxxxxxxxx> > --- > server/red-channel.c | 16 ++++++++++++++++ > 1 file changed, 16 insertions(+) > > diff --git a/server/red-channel.c b/server/red-channel.c > index 306c87d..b33c91d 100644 > --- a/server/red-channel.c > +++ b/server/red-channel.c > @@ -608,8 +608,24 @@ static inline void > red_channel_client_release_sent_item(RedChannelClient *rcc) > static void red_channel_peer_on_out_msg_done(void *opaque) > { > RedChannelClient *rcc = (RedChannelClient *)opaque; > + int fd = spice_marshaller_get_fd(rcc->send_data.marshaller); > > rcc->send_data.size = 0; > + > + if (fd != -1) { > + if (fcntl(fd, F_GETFD) == -1) { Is not clear what's the reason for this test. Why should fail? > + close(fd); > + fd = -1; > + } > + > + if (reds_stream_send_msgfd(rcc->stream, fd) < 0) { > + perror("sendfd"); > + red_channel_client_disconnect(rcc); > + return; I think here you should close the file descriptor too to avoid leak. > + } > + close(fd); > + } > + > red_channel_client_release_sent_item(rcc); > if (rcc->send_data.blocked) { > rcc->send_data.blocked = FALSE; > -- > 2.5.0 > > Frediano _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel