[SCSI] lpfc 8.1.2: Modify RSCN handling to unregister rpis on lost FCP_TARGETs immedi...
authorJamie Wellnitz <Jamie.Wellnitz@emulex.com>
Wed, 1 Mar 2006 00:25:38 +0000 (19:25 -0500)
committerJames Bottomley <jejb@mulgrave.il.steeleye.com>
Wed, 1 Mar 2006 01:05:42 +0000 (19:05 -0600)
Modify RSCN handling to unregister rpis on lost FCP_TARGETs immediately

Signed-off-by: Jamie Wellnitz <Jamie.Wellnitz@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
drivers/scsi/lpfc/lpfc_ct.c

index eab087bf826ed28d4b17204e3b92ddde8f9fa29b..0c982bbc4c77a6d9325d7dc300ea60b4c0f2bec8 100644 (file)
@@ -323,6 +323,7 @@ lpfc_ns_rsp(struct lpfc_hba * phba, struct lpfc_dmabuf * mp, uint32_t Size)
        struct lpfc_sli_ct_request *Response =
                (struct lpfc_sli_ct_request *) mp->virt;
        struct lpfc_nodelist *ndlp = NULL;
+       struct lpfc_nodelist *next_ndlp;
        struct lpfc_dmabuf *mlast, *next_mp;
        uint32_t *ctptr = (uint32_t *) & Response->un.gid.PortType;
        uint32_t Did;
@@ -391,8 +392,36 @@ lpfc_ns_rsp(struct lpfc_hba * phba, struct lpfc_dmabuf * mp, uint32_t Size)
 nsout1:
        list_del(&head);
 
-       /* Here we are finished in the case RSCN */
+       /*
+        * The driver has cycled through all Nports in the RSCN payload.
+        * Complete the handling by cleaning up and marking the
+        * current driver state.
+        */
        if (phba->hba_state == LPFC_HBA_READY) {
+
+               /*
+                * Switch ports that connect a loop of multiple targets need
+                * special consideration.  The driver wants to unregister the
+                * rpi only on the target that was pulled from the loop.  On
+                * RSCN, the driver wants to rediscover an NPort only if the
+                * driver flagged it as NLP_NPR_2B_DISC.  Provided adisc is
+                * not enabled and the NPort is not capable of retransmissions
+                * (FC Tape) prevent timing races with the scsi error handler by
+                * unregistering the Nport's RPI.  This action causes all
+                * outstanding IO to flush back to the midlayer.
+                */
+               list_for_each_entry_safe(ndlp, next_ndlp, &phba->fc_npr_list,
+                                        nlp_listp) {
+                       if (!(ndlp->nlp_flag & NLP_NPR_2B_DISC) &&
+                           (lpfc_rscn_payload_check(phba, ndlp->nlp_DID))) {
+                               if ((phba->cfg_use_adisc == 0) &&
+                                   !(ndlp->nlp_fcp_info &
+                                     NLP_FCP_2_DEVICE)) {
+                                       lpfc_unreg_rpi(phba, ndlp);
+                                       ndlp->nlp_flag &= ~NLP_NPR_ADISC;
+                               }
+                       }
+               }
                lpfc_els_flush_rscn(phba);
                spin_lock_irq(phba->host->host_lock);
                phba->fc_flag |= FC_RSCN_MODE; /* we are still in RSCN mode */