fcst: Handle frame send failures properly

Retry sending XFER_RDY, data and response frames if the network
driver reports that sending failed (-ENOMEM) instead of reporting
a kernel warning (WARN_ON(1)). If sending XFER_RDY or data frames
failed for another reason, report this to the initiator as a
write error (ASC = 03; ASCQ = 00 which stands for PERIPHERAL
DEVICE WRITE FAULT). If sending a response frame failed with
another error code than -ENOMEM, do not send a response.


git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5607 d57e44dd-8a1f-0410-8b47-8ef2f437770f

fcst/fcst.h

@@ -174,4 +174,26 @@ struct ft_tpg *ft_lport_find_tpg(struct fc_lport *);
 struct ft_node_acl *ft_acl_get(struct ft_tpg *, struct fc_rport_priv *);
 void ft_cmd_dump(struct scst_cmd *, const char *);
 
+/* #define FCST_INJECT_SEND_ERRORS 2 */
+
+#ifdef FCST_INJECT_SEND_ERRORS
+#define FCST_INJ_SEND_ERR(e)						\
+({									\
+	int _error = 0;							\
+									\
+	if (scst_random() % 62929 == 0)					\
+		_error = -ENOMEM;					\
+	if (FCST_INJECT_SEND_ERRORS >= 2 && scst_random() % 69491 == 0)	\
+		_error = -ENXIO;					\
+	if (_error)							\
+		pr_warn("%s: injected seq_send() error %d\n", __func__,	\
+			_error);					\
+	else								\
+		_error = (e);						\
+	_error;								\
+})
+#else
+#define FCST_INJ_SEND_ERR(e) (e)
+#endif
+
 #endif /* __SCSI_FCST_H__ */

fcst/ft_cmd.c

@@ -291,6 +291,7 @@ int ft_send_response(struct scst_cmd *cmd)
 	struct fc_exch *ep;
 	unsigned int slen;
 	size_t len;
+	enum ft_cmd_state prev_state;
 	int resid = 0;
 	int bi_resid = 0;
 	int error;
@@ -303,7 +304,7 @@ int ft_send_response(struct scst_cmd *cmd)
 	lport = ep->lp;
 
 	WARN_ON(fcmd->state != FT_STATE_NEW && fcmd->state != FT_STATE_DATA_IN);
-	ft_set_cmd_state(fcmd, FT_STATE_CMD_RSP_SENT);
+	prev_state = ft_set_cmd_state(fcmd, FT_STATE_CMD_RSP_SENT);
 
 	if (scst_cmd_aborted_on_xmit(cmd)) {
 		FT_IO_DBG("cmd aborted did %x oxid %x\n", ep->did, ep->oxid);
@@ -313,7 +314,8 @@ int ft_send_response(struct scst_cmd *cmd)
 
 	if (!scst_cmd_get_is_send_status(cmd)) {
 		FT_IO_DBG("send status not set.  feature not implemented\n");
-		return SCST_TGT_RES_FATAL_ERROR;
+		error = SCST_TGT_RES_FATAL_ERROR;
+		goto err;
 	}
 
 	status = scst_cmd_get_status(cmd);
@@ -333,7 +335,7 @@ int ft_send_response(struct scst_cmd *cmd)
 		error = ft_send_read_data(cmd);
 		if (error) {
 			FT_ERR("ft_send_read_data returned %d\n", error);
-			return error;
+			goto err;
 		}
 
 		if (dir == SCST_DATA_BIDI) {
@@ -347,8 +349,10 @@ int ft_send_response(struct scst_cmd *cmd)
 	}
 
 	fp = fc_frame_alloc(lport, len);
-	if (!fp)
-		return SCST_TGT_RES_QUEUE_FULL;
+	if (!fp) {
+		error = SCST_TGT_RES_QUEUE_FULL;
+		goto err;
+	}
 
 	fcp = fc_frame_payload_get(fp, len);
 	memset(fcp, 0, sizeof(*fcp));
@@ -384,13 +388,26 @@ int ft_send_response(struct scst_cmd *cmd)
 	fc_fill_fc_hdr(fp, FC_RCTL_DD_CMD_STATUS, ep->did, ep->sid, FC_TYPE_FCP,
 		       FC_FC_EX_CTX | FC_FC_LAST_SEQ | FC_FC_END_SEQ, 0);
 
-	error = lport->tt.seq_send(lport, fcmd->seq, fp);
-	if (error < 0)
+	error = FCST_INJ_SEND_ERR(lport->tt.seq_send(lport, fcmd->seq, fp));
+	if (error < 0) {
 		pr_err("Sending response for exchange with OX_ID %#x and RX_ID"
 		       " %#x failed: %d\n", ep->oxid, ep->rxid, error);
+		error = error == -ENOMEM ? SCST_TGT_RES_QUEUE_FULL :
+			SCST_TGT_RES_FATAL_ERROR;
+		goto err;
+	}
 done:
 	scst_tgt_cmd_done(cmd, SCST_CONTEXT_SAME);
 	return SCST_TGT_RES_SUCCESS;
+
+err:
+	ft_set_cmd_state(fcmd, prev_state);
+	WARN_ONCE(error != SCST_TGT_RES_QUEUE_FULL &&
+		  error != SCST_TGT_RES_FATAL_ERROR,
+		  "%s: invalid error code %d\n",
+		  __func__, error);
+	return error;
+
 }
 
 /*
@@ -451,6 +468,7 @@ int ft_send_xfer_rdy(struct scst_cmd *cmd)
 	struct fcp_txrdy *txrdy;
 	struct fc_lport *lport;
 	struct fc_exch *ep;
+	int error;
 
 	fcmd = scst_cmd_get_tgt_priv(cmd);
 
@@ -471,8 +489,17 @@ int ft_send_xfer_rdy(struct scst_cmd *cmd)
 	fcmd->seq = lport->tt.seq_start_next(fcmd->seq);
 	fc_fill_fc_hdr(fp, FC_RCTL_DD_DATA_DESC, ep->did, ep->sid, FC_TYPE_FCP,
 		       FC_FC_EX_CTX | FC_FC_END_SEQ | FC_FC_SEQ_INIT, 0);
-	lport->tt.seq_send(lport, fcmd->seq, fp);
-	return SCST_TGT_RES_SUCCESS;
+	error = FCST_INJ_SEND_ERR(lport->tt.seq_send(lport, fcmd->seq, fp));
+	switch (error) {
+	case 0:
+		return SCST_TGT_RES_SUCCESS;
+	case -ENOMEM:
+		ft_set_cmd_state(fcmd, FT_STATE_NEW);
+		return SCST_TGT_RES_QUEUE_FULL;
+	default:
+		ft_set_cmd_state(fcmd, FT_STATE_NEW);
+		return SCST_TGT_RES_FATAL_ERROR;
+	}
 }
 
 /*
@@ -708,13 +735,6 @@ static void ft_recv_cmd(struct ft_sess *sess, struct fc_frame *fp)
 	scst_cmd_set_tgt_priv(cmd, fcmd);
 	cmd->state = FT_STATE_NEW;
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 36)
-	sp = fr_seq(fp);
-#else
-	sp = lport->tt.seq_assign(lport, fp);
-	if (!sp)
-		goto busy;
-#endif
 	fcmd->seq = sp;
 	lport->tt.seq_set_resp(sp, ft_recv_seq, cmd);
 

fcst/ft_io.c

@@ -175,10 +175,17 @@ int ft_send_read_data(struct scst_cmd *cmd)
 			       remaining ? (FC_FC_EX_CTX | FC_FC_REL_OFF) :
 			       (FC_FC_EX_CTX | FC_FC_REL_OFF | FC_FC_END_SEQ),
 			       fh_off);
-		error = lport->tt.seq_send(lport, fcmd->seq, fp);
+		error = FCST_INJ_SEND_ERR(lport->tt.seq_send(lport, fcmd->seq,
+							     fp));
 		if (error) {
-			WARN_ON(1);
-			/* XXX For now, initiator will retry */
+			pr_warn("Sending frame with oid %#x oxid %#x resp_len"
+				" %d failed at frame_off %u / remaining %zu"
+				" with error code %d  - %s", ep->oid, ep->oxid,
+				scst_cmd_get_resp_data_len(cmd), frame_off,
+				remaining, error, error == -ENOMEM ?
+				"retrying" : "giving up");
+			return error == -ENOMEM ? SCST_TGT_RES_QUEUE_FULL :
+				SCST_TGT_RES_FATAL_ERROR;
 		} else
 			fcmd->read_data_len = frame_off;
 	}