From: Hannes Reinecke <hare@suse.de>
Date: Thu, 13 Oct 2016 13:10:38 +0000 (+0200)
Subject: scsi: libfc: spurious I/O error under high load
X-Git-Url: https://git.karo-electronics.de/?a=commitdiff_plain;h=a0452bb45c6bcf523acff635325fee7dd961c6bd;p=linux-beck.git

scsi: libfc: spurious I/O error under high load

If a command times out libfc is sending an REC, which also
might fail (due to frames being lost or something).
If no data has been transferred we can simply retry
the command, but the current code sets a state of FC_ERROR,
which then is being translated into DID_ERROR, resulting
in an I/O error.
So to handle this properly we need to set a separate
state FC_TRANS_RESET and mapping it onto DID_SOFT_RETRY.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
---

diff --git a/drivers/scsi/libfc/fc_fcp.c b/drivers/scsi/libfc/fc_fcp.c
index 5c6c73a80379..9f9eb7b0ba2d 100644
--- a/drivers/scsi/libfc/fc_fcp.c
+++ b/drivers/scsi/libfc/fc_fcp.c
@@ -122,6 +122,7 @@ static void fc_fcp_srr_error(struct fc_fcp_pkt *, struct fc_frame *);
 #define FC_HRD_ERROR		9
 #define FC_CRC_ERROR		10
 #define FC_TIMED_OUT		11
+#define FC_TRANS_RESET		12
 
 /*
  * Error recovery timeout values.
@@ -283,7 +284,7 @@ static int fc_fcp_send_abort(struct fc_fcp_pkt *fsp)
  * fc_io_compl() will notify the SCSI-ml that the I/O is done.
  * The SCSI-ml will retry the command.
  */
-static void fc_fcp_retry_cmd(struct fc_fcp_pkt *fsp)
+static void fc_fcp_retry_cmd(struct fc_fcp_pkt *fsp, int status_code)
 {
 	if (fsp->seq_ptr) {
 		fsp->lp->tt.exch_done(fsp->seq_ptr);
@@ -292,7 +293,7 @@ static void fc_fcp_retry_cmd(struct fc_fcp_pkt *fsp)
 
 	fsp->state &= ~FC_SRB_ABORT_PENDING;
 	fsp->io_status = 0;
-	fsp->status_code = FC_ERROR;
+	fsp->status_code = status_code;
 	fc_fcp_complete_locked(fsp);
 }
 
@@ -1208,7 +1209,7 @@ static void fc_fcp_error(struct fc_fcp_pkt *fsp, struct fc_frame *fp)
 		return;
 
 	if (error == -FC_EX_CLOSED) {
-		fc_fcp_retry_cmd(fsp);
+		fc_fcp_retry_cmd(fsp, FC_ERROR);
 		goto unlock;
 	}
 
@@ -1531,10 +1532,11 @@ static void fc_fcp_rec_resp(struct fc_seq *seq, struct fc_frame *fp, void *arg)
 			 */
 			if (rjt->er_explan == ELS_EXPL_OXID_RXID &&
 			    fsp->xfer_len == 0) {
-				fc_fcp_retry_cmd(fsp);
+				fsp->state |= FC_SRB_ABORTED;
+				fc_fcp_retry_cmd(fsp, FC_TRANS_RESET);
 				break;
 			}
-			fc_fcp_recovery(fsp, FC_ERROR);
+			fc_fcp_recovery(fsp, FC_TRANS_RESET);
 			break;
 		}
 	} else if (opcode == ELS_LS_ACC) {
@@ -1630,7 +1632,7 @@ static void fc_fcp_rec_error(struct fc_fcp_pkt *fsp, struct fc_frame *fp)
 	case -FC_EX_CLOSED:
 		FC_FCP_DBG(fsp, "REC %p fid %6.6x exchange closed\n",
 			   fsp, fsp->rport->port_id);
-		fc_fcp_retry_cmd(fsp);
+		fc_fcp_retry_cmd(fsp, FC_ERROR);
 		break;
 
 	default:
@@ -1729,7 +1731,7 @@ static void fc_fcp_srr(struct fc_fcp_pkt *fsp, enum fc_rctl r_ctl, u32 offset)
 	fc_fcp_pkt_hold(fsp);		/* hold for outstanding SRR */
 	return;
 retry:
-	fc_fcp_retry_cmd(fsp);
+	fc_fcp_retry_cmd(fsp, FC_TRANS_RESET);
 }
 
 /**
@@ -1801,7 +1803,7 @@ static void fc_fcp_srr_error(struct fc_fcp_pkt *fsp, struct fc_frame *fp)
 		FC_FCP_DBG(fsp, "SRR error, exchange closed\n");
 		/* fall through */
 	default:
-		fc_fcp_retry_cmd(fsp);
+		fc_fcp_retry_cmd(fsp, FC_ERROR);
 		break;
 	}
 	fc_fcp_unlock_pkt(fsp);
@@ -2014,6 +2016,11 @@ static void fc_io_compl(struct fc_fcp_pkt *fsp)
 			   "due to FC_CMD_RESET\n");
 		sc_cmd->result = (DID_RESET << 16);
 		break;
+	case FC_TRANS_RESET:
+		FC_FCP_DBG(fsp, "Returning DID_SOFT_ERROR to scsi-ml "
+			   "due to FC_TRANS_RESET\n");
+		sc_cmd->result = (DID_SOFT_ERROR << 16);
+		break;
 	case FC_HRD_ERROR:
 		FC_FCP_DBG(fsp, "Returning DID_NO_CONNECT to scsi-ml "
 			   "due to FC_HRD_ERROR\n");