[SCSI] ibmvfc: Fix for lost async events
Brian King [Fri, 15 Aug 2008 15:59:23 +0000 (10:59 -0500)]
If the client virtual fibre channel adapter is already logged into the
server and does an NPIV Login again, the async queue, which is used for
reporting Link Up/Link Down type of events, does not get reset on the
server side. Fix up the client driver so that we also do not reset it.
This fixes a problem of lost async events following relogins.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

drivers/scsi/ibmvscsi/ibmvfc.c

index 406ab27..58f8c9e 100644 (file)
@@ -556,11 +556,12 @@ static void ibmvfc_link_down(struct ibmvfc_host *vhost,
 /**
  * ibmvfc_init_host - Start host initialization
  * @vhost:             ibmvfc host struct
+ * @relogin:   is this a re-login?
  *
  * Return value:
  *     nothing
  **/
-static void ibmvfc_init_host(struct ibmvfc_host *vhost)
+static void ibmvfc_init_host(struct ibmvfc_host *vhost, int relogin)
 {
        struct ibmvfc_target *tgt;
 
@@ -574,6 +575,11 @@ static void ibmvfc_init_host(struct ibmvfc_host *vhost)
        }
 
        if (!ibmvfc_set_host_state(vhost, IBMVFC_INITIALIZING)) {
+               if (!relogin) {
+                       memset(vhost->async_crq.msgs, 0, PAGE_SIZE);
+                       vhost->async_crq.cur = 0;
+               }
+
                list_for_each_entry(tgt, &vhost->targets, queue)
                        tgt->need_login = 1;
                scsi_block_requests(vhost->host);
@@ -2084,11 +2090,11 @@ static void ibmvfc_handle_async(struct ibmvfc_async_crq *crq,
        case IBMVFC_AE_LINK_UP:
        case IBMVFC_AE_RESUME:
                vhost->events_to_log |= IBMVFC_AE_LINKUP;
-               ibmvfc_init_host(vhost);
+               ibmvfc_init_host(vhost, 1);
                break;
        case IBMVFC_AE_SCN_FABRIC:
                vhost->events_to_log |= IBMVFC_AE_RSCN;
-               ibmvfc_init_host(vhost);
+               ibmvfc_init_host(vhost, 1);
                break;
        case IBMVFC_AE_SCN_NPORT:
        case IBMVFC_AE_SCN_GROUP:
@@ -2134,13 +2140,13 @@ static void ibmvfc_handle_crq(struct ibmvfc_crq *crq, struct ibmvfc_host *vhost)
                        /* Send back a response */
                        rc = ibmvfc_send_crq_init_complete(vhost);
                        if (rc == 0)
-                               ibmvfc_init_host(vhost);
+                               ibmvfc_init_host(vhost, 0);
                        else
                                dev_err(vhost->dev, "Unable to send init rsp. rc=%ld\n", rc);
                        break;
                case IBMVFC_CRQ_INIT_COMPLETE:
                        dev_info(vhost->dev, "Partner initialization complete\n");
-                       ibmvfc_init_host(vhost);
+                       ibmvfc_init_host(vhost, 0);
                        break;
                default:
                        dev_err(vhost->dev, "Unknown crq message type: %d\n", crq->format);
@@ -3358,8 +3364,6 @@ static void ibmvfc_npiv_login(struct ibmvfc_host *vhost)
        mad->buffer.va = vhost->login_buf_dma;
        mad->buffer.len = sizeof(*vhost->login_buf);
 
-       memset(vhost->async_crq.msgs, 0, PAGE_SIZE);
-       vhost->async_crq.cur = 0;
        ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_INIT_WAIT);
 
        if (!ibmvfc_send_event(evt, vhost, default_timeout))