On 01/06/2011 12:54 AM, Or Gerlitz wrote:
Mike Christie wrote:
I went a different way. In the attached patch we detect the problem when
binding and will force a disconnect of the old ep before binding a new one.
Try it out and let me know.

--- a/drivers/scsi/iscsi_tcp.c
+++ b/drivers/scsi/iscsi_tcp.c
@@ -651,8 +651,7 @@ free_addr:

  static int
  iscsi_sw_tcp_conn_bind(struct iscsi_cls_session *cls_session,
-                      struct iscsi_cls_conn *cls_conn, uint64_t transport_eph,
-                      int is_leading)
+                      struct iscsi_cls_conn *cls_conn, uint64_t transport_eph)
  {
        struct Scsi_Host *shost = iscsi_session_to_shost(cls_session);
        struct iscsi_host *ihost = shost_priv(shost);
@@ -685,7 +684,7 @@ iscsi_sw_tcp_conn_bind(struct iscsi_cls_session 
*cls_session,
        if (err)
                goto free_socket;

-       err = iscsi_conn_bind(cls_session, cls_conn, is_leading);
+       err = iscsi_conn_bind(cls_session, cls_conn, NULL);

should this be transport_eph instead of NULL?


No. iscsi_tcp does not use the iscsi_endpoint abstraction becuase it actually uses normal old sockets that are created in userspace.


--- a/drivers/scsi/libiscsi.c
+++ b/drivers/scsi/libiscsi.c
@@ -3137,16 +3137,18 @@ void iscsi_conn_stop(struct iscsi_cls_conn *cls_conn, 
int flag)
  EXPORT_SYMBOL_GPL(iscsi_conn_stop);

  int iscsi_conn_bind(struct iscsi_cls_session *cls_session,
-                   struct iscsi_cls_conn *cls_conn, int is_leading)
+                   struct iscsi_cls_conn *cls_conn,
+                   struct iscsi_endpoint *ep)
  {
        struct iscsi_session *session = cls_session->dd_data;
        struct iscsi_conn *conn = cls_conn->dd_data;

        spin_lock_bh(&session->lock);
-       if (is_leading)
-               session->leadconn = conn;
+       session->leadconn = conn;
        spin_unlock_bh(&session->lock);

+       ep->conn = cls_conn;
+       cls_conn->ep = ep;

if not, it doesn't look like ep can be null here...

Will fix. Tested offload and did not test software iscsi.



--- a/include/scsi/libiscsi.h
+++ b/include/scsi/libiscsi.h
@@ -388,7 +388,7 @@ extern void iscsi_conn_teardown(struct iscsi_cls_conn *);
  extern int iscsi_conn_bind(struct iscsi_cls_session *, struct iscsi_cls_conn 
*,
-                          int);
+                          struct iscsi_endpoint *);

--- a/include/scsi/scsi_transport_iscsi.h
+++ b/include/scsi/scsi_transport_iscsi.h
@@ -95,7 +95,7 @@ struct iscsi_transport {
                                uint32_t cid);
        int (*bind_conn) (struct iscsi_cls_session *session,
                          struct iscsi_cls_conn *cls_conn,
-                         uint64_t transport_eph, int is_leading);
+                         uint64_t transport_eph);

so we're implicitly casting from u64 to ep pointer? I guess this should be fine.


No. The transport->bind_conn takes the u64 id. The driver then looks that up and passes iscsi_bind_conn a pointer to the ep struct.

--
You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To post to this group, send email to open-is...@googlegroups.com.
To unsubscribe from this group, send email to 
open-iscsi+unsubscr...@googlegroups.com.
For more options, visit this group at 
http://groups.google.com/group/open-iscsi?hl=en.

Reply via email to