[SCSI] libfc: rport retry on LS_RJT from certain ELS
authorChris Leech <christopher.leech@intel.com>
Fri, 27 Feb 2009 18:55:02 +0000 (10:55 -0800)
committerJames Bottomley <James.Bottomley@HansenPartnership.com>
Fri, 6 Mar 2009 21:39:34 +0000 (15:39 -0600)
This allows any rport ELS to retry on LS_RJT.

The rport error handling would only retry on resource allocation failures
and exchange timeouts.  I have a target that will occasionally reject PLOGI
when we do a quick LOGO/PLOGI.  When a critical ELS was rejected, libfc would
fail silently leaving the rport in a dead state.

The retry count and delay are managed by fc_rport_error_retry.  If the retry
count is exceeded fc_rport_error will be called.  When retrying is not the
correct course of action, fc_rport_error can be called directly.

Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
drivers/scsi/libfc/fc_exch.c
drivers/scsi/libfc/fc_rport.c
include/scsi/fc/fc_fs.h

index dd269e5f953e7240d5e24020065bfb4c2e3f418d..8c4018956d4cd516e7c8409e621516290b24b02c 100644 (file)
@@ -32,8 +32,6 @@
 #include <scsi/libfc.h>
 #include <scsi/fc_encode.h>
 
-#define          FC_DEF_R_A_TOV      (10 * 1000) /* resource allocation timeout */
-
 /*
  * fc_exch_debug can be set in debugger or at compile time to get more logs.
  */
index 7175759341523087a2546cbf7ad44f90c4b5dad6..600a8fffa940985dc5d086df7526e33c68cf7fa0 100644 (file)
@@ -81,6 +81,7 @@ static void fc_rport_recv_logo_req(struct fc_rport *,
                                   struct fc_seq *, struct fc_frame *);
 static void fc_rport_timeout(struct work_struct *);
 static void fc_rport_error(struct fc_rport *, struct fc_frame *);
+static void fc_rport_error_retry(struct fc_rport *, struct fc_frame *);
 static void fc_rport_work(struct work_struct *);
 
 static const char *fc_rport_state_names[] = {
@@ -410,57 +411,73 @@ static void fc_rport_timeout(struct work_struct *work)
 }
 
 /**
- * fc_rport_error - Handler for any errors
+ * fc_rport_error - Error handler, called once retries have been exhausted
  * @rport: The fc_rport object
  * @fp: The frame pointer
  *
- * If the error was caused by a resource allocation failure
- * then wait for half a second and retry, otherwise retry
- * immediately.
- *
  * Locking Note: The rport lock is expected to be held before
  * calling this routine
  */
 static void fc_rport_error(struct fc_rport *rport, struct fc_frame *fp)
 {
        struct fc_rport_libfc_priv *rdata = rport->dd_data;
-       unsigned long delay = 0;
 
        FC_DEBUG_RPORT("Error %ld in state %s, retries %d\n",
                       PTR_ERR(fp), fc_rport_state(rport), rdata->retries);
 
-       if (!fp || PTR_ERR(fp) == -FC_EX_TIMEOUT) {
-               /*
-                * Memory allocation failure, or the exchange timed out.
-                *  Retry after delay
-                */
-               if (rdata->retries < rdata->local_port->max_retry_count) {
-                       rdata->retries++;
-                       if (!fp)
-                               delay = msecs_to_jiffies(500);
-                       get_device(&rport->dev);
-                       schedule_delayed_work(&rdata->retry_work, delay);
-               } else {
-                       switch (rdata->rp_state) {
-                       case RPORT_ST_PLOGI:
-                       case RPORT_ST_PRLI:
-                       case RPORT_ST_LOGO:
-                               rdata->event = RPORT_EV_FAILED;
-                               queue_work(rport_event_queue,
-                                          &rdata->event_work);
-                               break;
-                       case RPORT_ST_RTV:
-                               fc_rport_enter_ready(rport);
-                               break;
-                       case RPORT_ST_NONE:
-                       case RPORT_ST_READY:
-                       case RPORT_ST_INIT:
-                               break;
-                       }
-               }
+       switch (rdata->rp_state) {
+       case RPORT_ST_PLOGI:
+       case RPORT_ST_PRLI:
+       case RPORT_ST_LOGO:
+               rdata->event = RPORT_EV_FAILED;
+               queue_work(rport_event_queue,
+                          &rdata->event_work);
+               break;
+       case RPORT_ST_RTV:
+               fc_rport_enter_ready(rport);
+               break;
+       case RPORT_ST_NONE:
+       case RPORT_ST_READY:
+       case RPORT_ST_INIT:
+               break;
        }
 }
 
+/**
+ * fc_rport_error_retry - Error handler when retries are desired
+ * @rport: The fc_rport object
+ * @fp: The frame pointer
+ *
+ * If the error was an exchange timeout retry immediately,
+ * otherwise wait for E_D_TOV.
+ *
+ * Locking Note: The rport lock is expected to be held before
+ * calling this routine
+ */
+static void fc_rport_error_retry(struct fc_rport *rport, struct fc_frame *fp)
+{
+       struct fc_rport_libfc_priv *rdata = rport->dd_data;
+       unsigned long delay = FC_DEF_E_D_TOV;
+
+       /* make sure this isn't an FC_EX_CLOSED error, never retry those */
+       if (PTR_ERR(fp) == -FC_EX_CLOSED)
+               return fc_rport_error(rport, fp);
+
+       if (rdata->retries < rdata->local_port->max_retry_count) {
+               FC_DEBUG_RPORT("Error %ld in state %s, retrying\n",
+                              PTR_ERR(fp), fc_rport_state(rport));
+               rdata->retries++;
+               /* no additional delay on exchange timeouts */
+               if (PTR_ERR(fp) == -FC_EX_TIMEOUT)
+                       delay = 0;
+               get_device(&rport->dev);
+               schedule_delayed_work(&rdata->retry_work, delay);
+               return;
+       }
+
+       return fc_rport_error(rport, fp);
+}
+
 /**
  * fc_rport_plogi_recv_resp - Handle incoming ELS PLOGI response
  * @sp: current sequence in the PLOGI exchange
@@ -495,7 +512,7 @@ static void fc_rport_plogi_resp(struct fc_seq *sp, struct fc_frame *fp,
        }
 
        if (IS_ERR(fp)) {
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
                goto err;
        }
 
@@ -527,7 +544,7 @@ static void fc_rport_plogi_resp(struct fc_seq *sp, struct fc_frame *fp,
                else
                        fc_rport_enter_prli(rport);
        } else
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
 
 out:
        fc_frame_free(fp);
@@ -557,14 +574,14 @@ static void fc_rport_enter_plogi(struct fc_rport *rport)
        rport->maxframe_size = FC_MIN_MAX_PAYLOAD;
        fp = fc_frame_alloc(lport, sizeof(struct fc_els_flogi));
        if (!fp) {
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
                return;
        }
        rdata->e_d_tov = lport->e_d_tov;
 
        if (!lport->tt.elsct_send(lport, rport, fp, ELS_PLOGI,
                                  fc_rport_plogi_resp, rport, lport->e_d_tov))
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
        else
                get_device(&rport->dev);
 }
@@ -604,7 +621,7 @@ static void fc_rport_prli_resp(struct fc_seq *sp, struct fc_frame *fp,
        }
 
        if (IS_ERR(fp)) {
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
                goto err;
        }
 
@@ -662,7 +679,7 @@ static void fc_rport_logo_resp(struct fc_seq *sp, struct fc_frame *fp,
                       rport->port_id);
 
        if (IS_ERR(fp)) {
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
                goto err;
        }
 
@@ -712,13 +729,13 @@ static void fc_rport_enter_prli(struct fc_rport *rport)
 
        fp = fc_frame_alloc(lport, sizeof(*pp));
        if (!fp) {
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
                return;
        }
 
        if (!lport->tt.elsct_send(lport, rport, fp, ELS_PRLI,
                                  fc_rport_prli_resp, rport, lport->e_d_tov))
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
        else
                get_device(&rport->dev);
 }
@@ -809,13 +826,13 @@ static void fc_rport_enter_rtv(struct fc_rport *rport)
 
        fp = fc_frame_alloc(lport, sizeof(struct fc_els_rtv));
        if (!fp) {
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
                return;
        }
 
        if (!lport->tt.elsct_send(lport, rport, fp, ELS_RTV,
                                     fc_rport_rtv_resp, rport, lport->e_d_tov))
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
        else
                get_device(&rport->dev);
 }
@@ -840,13 +857,13 @@ static void fc_rport_enter_logo(struct fc_rport *rport)
 
        fp = fc_frame_alloc(lport, sizeof(struct fc_els_logo));
        if (!fp) {
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
                return;
        }
 
        if (!lport->tt.elsct_send(lport, rport, fp, ELS_LOGO,
                                  fc_rport_logo_resp, rport, lport->e_d_tov))
-               fc_rport_error(rport, fp);
+               fc_rport_error_retry(rport, fp);
        else
                get_device(&rport->dev);
 }
index 3e4801d2bdbb93332d5ceccef588130457213bdb..1b7af3a64c7c4cc065081885dbfddd227df421d0 100644 (file)
@@ -337,4 +337,9 @@ enum fc_pf_rjt_reason {
        FC_RJT_VENDOR =         0xff,   /* vendor specific reject */
 };
 
+/* default timeout values */
+
+#define FC_DEF_E_D_TOV 2000UL
+#define FC_DEF_R_A_TOV 10000UL
+
 #endif /* _FC_FS_H_ */