From 12e46bad2c8d04ca817c69f6adb592ee1757504d Mon Sep 17 00:00:00 2001
From: David C Somayajulu <davidcs@FreeBSD.org>
Date: Wed, 19 Jul 2017 19:08:37 +0000
Subject: [PATCH] Raise the watchdog timer interval to 2 ticks, there by
 guaranteeing that it fires between 1ms and 2ms. ` Treat two consecutive
 occurrences of Heartbeat failures as a legitimate Heartbeat failure

MFC after:3 days
---
 sys/dev/qlxgbe/ql_def.h |  2 +-
 sys/dev/qlxgbe/ql_hw.c  | 14 +++++++++++---
 sys/dev/qlxgbe/ql_hw.h  |  1 +
 sys/dev/qlxgbe/ql_os.c  |  2 +-
 4 files changed, 14 insertions(+), 5 deletions(-)

diff --git a/sys/dev/qlxgbe/ql_def.h b/sys/dev/qlxgbe/ql_def.h
index 83774a7d7d7..c03c5c974b3 100644
--- a/sys/dev/qlxgbe/ql_def.h
+++ b/sys/dev/qlxgbe/ql_def.h
@@ -105,7 +105,7 @@ struct qla_ivec {
 
 typedef struct qla_ivec qla_ivec_t;
 
-#define QLA_WATCHDOG_CALLOUT_TICKS	1
+#define QLA_WATCHDOG_CALLOUT_TICKS	2
 
 typedef struct _qla_tx_ring {
 	qla_tx_buf_t	tx_buf[NUM_TX_DESCRIPTORS];
diff --git a/sys/dev/qlxgbe/ql_hw.c b/sys/dev/qlxgbe/ql_hw.c
index 8a8494ce5e7..877dbf30ccd 100644
--- a/sys/dev/qlxgbe/ql_hw.c
+++ b/sys/dev/qlxgbe/ql_hw.c
@@ -3366,7 +3366,7 @@ ql_hw_check_health(qla_host_t *ha)
 
 	ha->hw.health_count++;
 
-	if (ha->hw.health_count < 1000)
+	if (ha->hw.health_count < 500)
 		return 0;
 
 	ha->hw.health_count = 0;
@@ -3385,10 +3385,18 @@ ql_hw_check_health(qla_host_t *ha)
 	if ((val != ha->hw.hbeat_value) &&
 		(!(QL_ERR_INJECT(ha, INJCT_HEARTBEAT_FAILURE)))) {
 		ha->hw.hbeat_value = val;
+		ha->hw.hbeat_failure = 0;
 		return 0;
 	}
-	device_printf(ha->pci_dev, "%s: Heartbeat Failue [0x%08x]\n",
-		__func__, val);
+
+	ha->hw.hbeat_failure++;
+
+	if (ha->hw.hbeat_failure < 2) /* we ignore the first failure */
+		return 0;
+	else
+		device_printf(ha->pci_dev, "%s: Heartbeat Failue [0x%08x]\n",
+			__func__, val);
+
 
 	return -1;
 }
diff --git a/sys/dev/qlxgbe/ql_hw.h b/sys/dev/qlxgbe/ql_hw.h
index 0c71dc2cfe6..6a4e7e9a11e 100644
--- a/sys/dev/qlxgbe/ql_hw.h
+++ b/sys/dev/qlxgbe/ql_hw.h
@@ -1671,6 +1671,7 @@ typedef struct _qla_hw {
 	/* heart beat register value */
 	uint32_t	hbeat_value;
 	uint32_t	health_count;
+	uint32_t	hbeat_failure;
 
 	uint32_t	max_tx_segs;
 	uint32_t	min_lro_pkt_size;
diff --git a/sys/dev/qlxgbe/ql_os.c b/sys/dev/qlxgbe/ql_os.c
index d4a18270dd0..99d4d504ecc 100644
--- a/sys/dev/qlxgbe/ql_os.c
+++ b/sys/dev/qlxgbe/ql_os.c
@@ -276,7 +276,7 @@ qla_watchdog(void *arg)
 		ha->qla_watchdog_paused = 1;
 	}
 
-	ha->watchdog_ticks = ha->watchdog_ticks++ % 1000;
+	ha->watchdog_ticks = ha->watchdog_ticks++ % 500;
 	callout_reset(&ha->tx_callout, QLA_WATCHDOG_CALLOUT_TICKS,
 		qla_watchdog, ha);
 }