From: Quinn Tran <qutran@marvell.com>
Date: Fri, 18 Jun 2021 22:24:27 -0700
Subject: scsi: qla2xxx: Add heartbeat check
Patch-mainline: v5.14-rc1
Git-commit: d94d8158e1841813624e9fecf93a12e64e004dd8
References: bsc#1189392

Use "no-op" mailbox command to check if the adapter firmware is still
responsive.

Link: https://lore.kernel.org/r/20210619052427.6440-1-njavali@marvell.com
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Acked-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/scsi/qla2xxx/qla_def.h  |    4 ++
 drivers/scsi/qla2xxx/qla_gbl.h  |    1 
 drivers/scsi/qla2xxx/qla_init.c |    6 ++-
 drivers/scsi/qla2xxx/qla_iocb.c |    4 ++
 drivers/scsi/qla2xxx/qla_isr.c  |    4 ++
 drivers/scsi/qla2xxx/qla_mbx.c  |   27 +++++++++++++++
 drivers/scsi/qla2xxx/qla_nvme.c |    4 ++
 drivers/scsi/qla2xxx/qla_os.c   |   68 ++++++++++++++++++++++++++++++++++++++++
 8 files changed, 117 insertions(+), 1 deletion(-)

--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -3661,6 +3661,8 @@ struct qla_qpair {
 	struct qla_tgt_counters tgt_counters;
 	uint16_t cpuid;
 	struct qla_fw_resources fwres ____cacheline_aligned;
+	u32	cmd_cnt;
+	u32	cmd_completion_cnt;
 };
 
 /* Place holder for FW buffer parameters */
@@ -4617,6 +4619,7 @@ struct qla_hw_data {
 
 	struct qla_hw_data_stat stat;
 	pci_error_state_t pci_error_state;
+	u64 prev_cmd_cnt;
 };
 
 struct active_regions {
@@ -4744,6 +4747,7 @@ typedef struct scsi_qla_host {
 #define SET_ZIO_THRESHOLD_NEEDED 32
 #define ISP_ABORT_TO_ROM	33
 #define VPORT_DELETE		34
+#define HEARTBEAT_CHK		38
 
 #define PROCESS_PUREX_IOCB	63
 
--- a/drivers/scsi/qla2xxx/qla_gbl.h
+++ b/drivers/scsi/qla2xxx/qla_gbl.h
@@ -550,6 +550,7 @@ extern int qla2xxx_read_remote_register(
     uint32_t *);
 extern int qla2xxx_write_remote_register(scsi_qla_host_t *, uint32_t,
     uint32_t);
+void qla_no_op_mb(struct scsi_qla_host *vha);
 
 /*
  * Global Function Prototypes in qla_isr.c source file.
--- a/drivers/scsi/qla2xxx/qla_init.c
+++ b/drivers/scsi/qla2xxx/qla_init.c
@@ -6871,10 +6871,14 @@ qla2x00_abort_isp_cleanup(scsi_qla_host_
 	ha->flags.fw_init_done = 0;
 	ha->chip_reset++;
 	ha->base_qpair->chip_reset = ha->chip_reset;
+	ha->base_qpair->cmd_cnt = ha->base_qpair->cmd_completion_cnt = 0;
 	for (i = 0; i < ha->max_qpairs; i++) {
-		if (ha->queue_pair_map[i])
+		if (ha->queue_pair_map[i]) {
 			ha->queue_pair_map[i]->chip_reset =
 				ha->base_qpair->chip_reset;
+			ha->queue_pair_map[i]->cmd_cnt =
+			    ha->queue_pair_map[i]->cmd_completion_cnt = 0;
+		}
 	}
 
 	/* purge MBox commands */
--- a/drivers/scsi/qla2xxx/qla_iocb.c
+++ b/drivers/scsi/qla2xxx/qla_iocb.c
@@ -1711,6 +1711,7 @@ qla24xx_start_scsi(srb_t *sp)
 	} else
 		req->ring_ptr++;
 
+	sp->qpair->cmd_cnt++;
 	sp->flags |= SRB_DMA_VALID;
 
 	/* Set chip new ring index. */
@@ -1913,6 +1914,7 @@ qla24xx_dif_start_scsi(srb_t *sp)
 	} else
 		req->ring_ptr++;
 
+	sp->qpair->cmd_cnt++;
 	/* Set chip new ring index. */
 	wrt_reg_dword(req->req_q_in, req->ring_index);
 
@@ -2069,6 +2071,7 @@ qla2xxx_start_scsi_mq(srb_t *sp)
 	} else
 		req->ring_ptr++;
 
+	sp->qpair->cmd_cnt++;
 	sp->flags |= SRB_DMA_VALID;
 
 	/* Set chip new ring index. */
@@ -2285,6 +2288,7 @@ qla2xxx_dif_start_scsi_mq(srb_t *sp)
 	} else
 		req->ring_ptr++;
 
+	sp->qpair->cmd_cnt++;
 	/* Set chip new ring index. */
 	wrt_reg_dword(req->req_q_in, req->ring_index);
 
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -2323,6 +2323,8 @@ static void qla24xx_nvme_iocb_entry(scsi
 
 	if (unlikely(iocb->u.nvme.aen_op))
 		atomic_dec(&sp->vha->hw->nvme_active_aen_cnt);
+	else
+		sp->qpair->cmd_completion_cnt++;
 
 	if (unlikely(comp_status != CS_COMPLETE))
 		logit = 1;
@@ -2977,6 +2979,8 @@ qla2x00_status_entry(scsi_qla_host_t *vh
 		return;
 	}
 
+	sp->qpair->cmd_completion_cnt++;
+
 	/* Fast path completion. */
 	if (comp_status == CS_COMPLETE && scsi_status == 0) {
 		qla2x00_process_completed_request(vha, req, handle);
--- a/drivers/scsi/qla2xxx/qla_mbx.c
+++ b/drivers/scsi/qla2xxx/qla_mbx.c
@@ -6940,3 +6940,30 @@ ql26xx_led_config(scsi_qla_host_t *vha,
 
 	return rval;
 }
+
+/**
+ * qla_no_op_mb(): This MB is used to check if FW is still alive and
+ * able to generate an interrupt. Otherwise, a timeout will trigger
+ * FW dump + reset
+ * @vha: host adapter pointer
+ * Return: None
+ */
+void qla_no_op_mb(struct scsi_qla_host *vha)
+{
+	mbx_cmd_t mc;
+	mbx_cmd_t *mcp = &mc;
+	int rval;
+
+	memset(&mc, 0, sizeof(mc));
+	mcp->mb[0] = 0; // noop cmd= 0
+	mcp->out_mb = MBX_0;
+	mcp->in_mb = MBX_0;
+	mcp->tov = 5;
+	mcp->flags = 0;
+	rval = qla2x00_mailbox_command(vha, mcp);
+
+	if (rval) {
+		ql_dbg(ql_dbg_async, vha, 0x7071,
+			"Failed %s %x\n", __func__, rval);
+	}
+}
--- a/drivers/scsi/qla2xxx/qla_nvme.c
+++ b/drivers/scsi/qla2xxx/qla_nvme.c
@@ -548,6 +548,10 @@ static inline int qla2x00_start_nvme_mq(
 		req->ring_ptr++;
 	}
 
+	/* ignore nvme async cmd due to long timeout */
+	if (!nvme->u.nvme.aen_op)
+		sp->qpair->cmd_cnt++;
+
 	/* Set chip new ring index. */
 	wrt_reg_dword(req->req_q_in, req->ring_index);
 
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -6983,6 +6983,17 @@ qla2x00_do_dpc(void *data)
 			qla2x00_lip_reset(base_vha);
 		}
 
+		if (test_bit(HEARTBEAT_CHK, &base_vha->dpc_flags)) {
+			/*
+			 * if there is a mb in progress then that's
+			 * enough of a check to see if fw is still ticking.
+			 */
+			if (!ha->flags.mbox_busy && base_vha->flags.init_done)
+				qla_no_op_mb(base_vha);
+
+			clear_bit(HEARTBEAT_CHK, &base_vha->dpc_flags);
+		}
+
 		ha->dpc_active = 0;
 end_loop:
 		set_current_state(TASK_INTERRUPTIBLE);
@@ -7039,6 +7050,61 @@ qla2x00_rst_aen(scsi_qla_host_t *vha)
 	}
 }
 
+static bool qla_do_heartbeat(struct scsi_qla_host *vha)
+{
+	u64 cmd_cnt, prev_cmd_cnt;
+	bool do_hb = false;
+	struct qla_hw_data *ha = vha->hw;
+	int i;
+
+	/* if cmds are still pending down in fw, then do hb */
+	if (ha->base_qpair->cmd_cnt != ha->base_qpair->cmd_completion_cnt) {
+		do_hb = true;
+		goto skip;
+	}
+
+	for (i = 0; i < ha->max_qpairs; i++) {
+		if (ha->queue_pair_map[i] &&
+		    ha->queue_pair_map[i]->cmd_cnt !=
+		    ha->queue_pair_map[i]->cmd_completion_cnt) {
+			do_hb = true;
+			break;
+		}
+	}
+
+skip:
+	prev_cmd_cnt = ha->prev_cmd_cnt;
+	cmd_cnt = ha->base_qpair->cmd_cnt;
+	for (i = 0; i < ha->max_qpairs; i++) {
+		if (ha->queue_pair_map[i])
+			cmd_cnt += ha->queue_pair_map[i]->cmd_cnt;
+	}
+	ha->prev_cmd_cnt = cmd_cnt;
+
+	if (!do_hb && ((cmd_cnt - prev_cmd_cnt) > 50))
+		/*
+		 * IOs are completing before periodic hb check.
+		 * IOs seems to be running, do hb for sanity check.
+		 */
+		do_hb = true;
+
+	return do_hb;
+}
+
+static void qla_heart_beat(struct scsi_qla_host *vha)
+{
+	if (vha->vp_idx)
+		return;
+
+	if (vha->hw->flags.eeh_busy || qla2x00_chip_is_down(vha))
+		return;
+
+	if (qla_do_heartbeat(vha)) {
+		set_bit(HEARTBEAT_CHK, &vha->dpc_flags);
+		qla2xxx_wake_dpc(vha);
+	}
+}
+
 /**************************************************************************
 *   qla2x00_timer
 *
@@ -7256,6 +7322,8 @@ qla2x00_timer(scsi_qla_host_t *vha)
 		qla2xxx_wake_dpc(vha);
 	}
 
+	qla_heart_beat(vha);
+
 	qla2x00_restart_timer(vha, WATCH_INTERVAL);
 }