ionic: add heartbeat check

Most of our firmware has a heartbeat feature that the driver
can watch for to see if the FW is still alive and likely to
answer a dev_cmd or AdminQ request.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/pensando/ionic/ionic_dev.c

@@ -84,6 +84,49 @@ void ionic_dev_teardown(struct ionic *ionic)
 }
 
 /* Devcmd Interface */
+int ionic_heartbeat_check(struct ionic *ionic)
+{
+	struct ionic_dev *idev = &ionic->idev;
+	unsigned long hb_time;
+	u32 fw_status;
+	u32 hb;
+
+	/* wait a little more than one second before testing again */
+	hb_time = jiffies;
+	if (time_before(hb_time, (idev->last_hb_time + (HZ * 2))))
+		return 0;
+
+	/* firmware is useful only if fw_status is non-zero */
+	fw_status = ioread32(&idev->dev_info_regs->fw_status);
+	if (!fw_status)
+		return -ENXIO;
+
+	/* early FW has no heartbeat, else FW will return non-zero */
+	hb = ioread32(&idev->dev_info_regs->fw_heartbeat);
+	if (!hb)
+		return 0;
+
+	/* are we stalled? */
+	if (hb == idev->last_hb) {
+		/* only complain once for each stall seen */
+		if (idev->last_hb_time != 1) {
+			dev_info(ionic->dev, "FW heartbeat stalled at %d\n",
+				 idev->last_hb);
+			idev->last_hb_time = 1;
+		}
+
+		return -ENXIO;
+	}
+
+	if (idev->last_hb_time == 1)
+		dev_info(ionic->dev, "FW heartbeat restored at %d\n", hb);
+
+	idev->last_hb = hb;
+	idev->last_hb_time = hb_time;
+
+	return 0;
+}
+
 u8 ionic_dev_cmd_status(struct ionic_dev *idev)
 {
 	return ioread8(&idev->dev_cmd_regs->comp.comp.status);

drivers/net/ethernet/pensando/ionic/ionic_dev.h

@@ -123,6 +123,9 @@ struct ionic_dev {
 	union ionic_dev_info_regs __iomem *dev_info_regs;
 	union ionic_dev_cmd_regs __iomem *dev_cmd_regs;
 
+	unsigned long last_hb_time;
+	u32 last_hb;
+
 	u64 __iomem *db_pages;
 	dma_addr_t phy_db_pages;
 
@@ -295,5 +298,6 @@ void ionic_q_post(struct ionic_queue *q, bool ring_doorbell, ionic_desc_cb cb,
 void ionic_q_rewind(struct ionic_queue *q, struct ionic_desc_info *start);
 void ionic_q_service(struct ionic_queue *q, struct ionic_cq_info *cq_info,
 		     unsigned int stop_index);
+int ionic_heartbeat_check(struct ionic *ionic);
 
 #endif /* _IONIC_DEV_H_ */

drivers/net/ethernet/pensando/ionic/ionic_main.c

@@ -245,6 +245,10 @@ static int ionic_adminq_post(struct ionic_lif *lif, struct ionic_admin_ctx *ctx)
 		goto err_out;
 	}
 
+	err = ionic_heartbeat_check(lif->ionic);
+	if (err)
+		goto err_out;
+
 	memcpy(adminq->head->desc, &ctx->cmd, sizeof(ctx->cmd));
 
 	dev_dbg(&lif->netdev->dev, "post admin queue command:\n");
@@ -305,6 +309,14 @@ int ionic_napi(struct napi_struct *napi, int budget, ionic_cq_cb cb,
 	return work_done;
 }
 
+static void ionic_dev_cmd_clean(struct ionic *ionic)
+{
+	union ionic_dev_cmd_regs *regs = ionic->idev.dev_cmd_regs;
+
+	iowrite32(0, &regs->doorbell);
+	memset_io(&regs->cmd, 0, sizeof(regs->cmd));
+}
+
 int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_seconds)
 {
 	struct ionic_dev *idev = &ionic->idev;
@@ -314,6 +326,7 @@ int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_seconds)
 	int opcode;
 	int done;
 	int err;
+	int hb;
 
 	WARN_ON(in_interrupt());
 
@@ -328,7 +341,8 @@ int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_seconds)
 		if (done)
 			break;
 		msleep(20);
-	} while (!done && time_before(jiffies, max_wait));
+		hb = ionic_heartbeat_check(ionic);
+	} while (!done && !hb && time_before(jiffies, max_wait));
 	duration = jiffies - start_time;
 
 	opcode = idev->dev_cmd_regs->cmd.cmd.opcode;
@@ -336,7 +350,15 @@ int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_seconds)
 		ionic_opcode_to_str(opcode), opcode,
 		done, duration / HZ, duration);
 
+	if (!done && hb) {
+		ionic_dev_cmd_clean(ionic);
+		dev_warn(ionic->dev, "DEVCMD %s (%d) failed - FW halted\n",
+			 ionic_opcode_to_str(opcode), opcode);
+		return -ENXIO;
+	}
+
 	if (!done && !time_before(jiffies, max_wait)) {
+		ionic_dev_cmd_clean(ionic);
 		dev_warn(ionic->dev, "DEVCMD %s (%d) timeout after %ld secs\n",
 			 ionic_opcode_to_str(opcode), opcode, max_seconds);
 		return -ETIMEDOUT;