On Thu, Mar 03, 2016 at 07:09:14PM -0600, Mike Christie wrote:
> On 03/03/2016 06:09 PM, Chris Leech wrote:
> > When requests are being failed it's important to abort the TCP
> > connection rather than let TCP wait and attempt a graceful shutdown.
> > 
> > That can be accomplished by setting the SO_LINGER socket option with a
> > linger time of 0 to drop queued data and close the connection with a RST
> > instead of a FIN.
> > 
> > Signed-off-by: Chris Leech <[email protected]>
> > ---
> >  usr/io.c | 15 +++++++++++++++
> >  1 file changed, 15 insertions(+)
> > 
> > diff --git a/usr/io.c b/usr/io.c
> > index f552e1e..48b233c 100644
> > --- a/usr/io.c
> > +++ b/usr/io.c
> > @@ -391,9 +391,24 @@ iscsi_io_tcp_poll(iscsi_conn_t *conn, int timeout_ms)
> >  void
> >  iscsi_io_tcp_disconnect(iscsi_conn_t *conn)
> >  {
> > +   struct linger so_linger = { .l_onoff = 1, .l_linger = 0 };
> > +
> >     if (conn->socket_fd >= 0) {
> >             log_debug(1, "disconnecting conn %p, fd %d", conn,
> >                      conn->socket_fd);
> > +
> > +           /* If the state is not IN_LOGOUT, this isn't a clean shutdown
> > +            * and there's some sort of error handling going on. In that
> > +            * case, set a 0 SO_LINGER to force an abortive close (RST) and
> > +            * free whatever is sitting in the TCP transmit queue. This is
> > +            * done to prevent stale data from being sent should the
> > +            * network connection be restored before TCP times out.
> > +            */
> > +           if (conn->state != ISCSI_CONN_STATE_IN_LOGOUT) {
> > +                   setsockopt(conn->socket_fd, SOL_SOCKET, SO_LINGER,
> > +                              &so_linger, sizeof(so_linger));
> > +           }
> > +
> >             close(conn->socket_fd);
> >             conn->socket_fd = -1;
> >     }
> > 
> 
> Nice.
> 
> For maybe a slightly different problem, but hoping I get lucky and your
> patch fixes it too, I thought the network layer was still accessing
> pages that we tried to send and was causing a oops. I get the part where
> with your patch the network layer will not try to send data anymore, but
> I guess I am asking if the network layer could still be doing some sort
> of delayed cleanup process after close() has returned?

>From what I can tell in the tcp code, the zero linger handling purges
all of the socket queues freeing everything before close returns.

- Chris

-- 
You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/open-iscsi.
For more options, visit https://groups.google.com/d/optout.

Reply via email to