iwlwifi: jiffies based tx queues watchdog

This patch replace monitor/recover timer by watchdog based on time
stamp. New code allow to discover hangs more precisely.

Timeout values are currently doubled monitoring period values of
previous timer. This have to be tuned based of firmware timing
capabilities.

Tested on 3945, 4965, 5300, 6300.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

1 files changed, 49 insertions, 62 deletions

diff --git a/drivers/net/wireless/iwlwifi/iwl-core.c b/drivers/net/wireless/iwlwifi/iwl-core.c
index c41f5a87821..d62b9251841 100644
--- a/drivers/net/wireless/iwlwifi/iwl-core.c
+++ b/drivers/net/wireless/iwlwifi/iwl-core.c
@@ -1894,77 +1894,58 @@ int iwl_mac_change_interface(struct ieee80211_hw *hw, struct ieee80211_vif *vif,
 }
 EXPORT_SYMBOL(iwl_mac_change_interface);
 
-/**
- * iwl_bg_monitor_recover - Timer callback to check for stuck queue and recover
- *
- * During normal condition (no queue is stuck), the timer is continually set to
- * execute every monitor_recover_period milliseconds after the last timer
- * expired.  When the queue read_ptr is at the same place, the timer is
- * shorten to 100mSecs.  This is
- *      1) to reduce the chance that the read_ptr may wrap around (not stuck)
- *      2) to detect the stuck queues quicker before the station and AP can
- *      disassociate each other.
- *
- * This function monitors all the tx queues and recover from it if any
- * of the queues are stuck.
- * 1. It first check the cmd queue for stuck conditions.  If it is stuck,
- *      it will recover by resetting the firmware and return.
- * 2. Then, it checks for station association.  If it associates it will check
- *      other queues.  If any queue is stuck, it will recover by resetting
- *      the firmware.
- * Note: It the number of times the queue read_ptr to be at the same place to
- *      be MAX_REPEAT+1 in order to consider to be stuck.
- */
 /*
- * The maximum number of times the read pointer of the tx queue at the
- * same place without considering to be stuck.
+ * On every watchdog tick we check (latest) time stamp. If it does not
+ * change during timeout period and queue is not empty we reset firmware.
  */
-#define MAX_REPEAT      (2)
 static int iwl_check_stuck_queue(struct iwl_priv *priv, int cnt)
 {
-	struct iwl_tx_queue *txq;
-	struct iwl_queue *q;
+	struct iwl_tx_queue *txq = &priv->txq[cnt];
+	struct iwl_queue *q = &txq->q;
+	unsigned long timeout;
+	int ret;
 
-	txq = &priv->txq[cnt];
-	q = &txq->q;
-	/* queue is empty, skip */
-	if (q->read_ptr == q->write_ptr)
+	if (q->read_ptr == q->write_ptr) {
+		txq->time_stamp = jiffies;
 		return 0;
+	}
 
-	if (q->read_ptr == q->last_read_ptr) {
-		/* a queue has not been read from last time */
-		if (q->repeat_same_read_ptr > MAX_REPEAT) {
-			IWL_ERR(priv,
-				"queue %d stuck %d time. Fw reload.\n",
-				q->id, q->repeat_same_read_ptr);
-			q->repeat_same_read_ptr = 0;
-			iwl_force_reset(priv, IWL_FW_RESET, false);
-		} else {
-			q->repeat_same_read_ptr++;
-			IWL_DEBUG_RADIO(priv,
-					"queue %d, not read %d time\n",
-					q->id,
-					q->repeat_same_read_ptr);
-			mod_timer(&priv->monitor_recover,
-				jiffies + msecs_to_jiffies(
-				IWL_ONE_HUNDRED_MSECS));
-			return 1;
-		}
-	} else {
-		q->last_read_ptr = q->read_ptr;
-		q->repeat_same_read_ptr = 0;
+	timeout = txq->time_stamp +
+		  msecs_to_jiffies(priv->cfg->base_params->wd_timeout);
+
+	if (time_after(jiffies, timeout)) {
+		IWL_ERR(priv, "Queue %d stuck for %u ms.\n",
+				q->id, priv->cfg->base_params->wd_timeout);
+		ret = iwl_force_reset(priv, IWL_FW_RESET, false);
+		return (ret == -EAGAIN) ? 0 : 1;
 	}
+
 	return 0;
 }
 
-void iwl_bg_monitor_recover(unsigned long data)
+/*
+ * Making watchdog tick be a quarter of timeout assure we will
+ * discover the queue hung between timeout and 1.25*timeout
+ */
+#define IWL_WD_TICK(timeout) ((timeout) / 4)
+
+/*
+ * Watchdog timer callback, we check each tx queue for stuck, if if hung
+ * we reset the firmware. If everything is fine just rearm the timer.
+ */
+void iwl_bg_watchdog(unsigned long data)
 {
 	struct iwl_priv *priv = (struct iwl_priv *)data;
 	int cnt;
+	unsigned long timeout;
 
 	if (test_bit(STATUS_EXIT_PENDING, &priv->status))
 		return;
 
+	timeout = priv->cfg->base_params->wd_timeout;
+	if (timeout == 0)
+		return;
+
 	/* monitor and check for stuck cmd queue */
 	if (iwl_check_stuck_queue(priv, priv->cmd_queue))
 		return;
@@ -1979,17 +1960,23 @@ void iwl_bg_monitor_recover(unsigned long data)
 				return;
 		}
 	}
-	if (priv->cfg->base_params->monitor_recover_period) {
-		/*
-		 * Reschedule the timer to occur in
-		 * priv->cfg->base_params->monitor_recover_period
-		 */
-		mod_timer(&priv->monitor_recover, jiffies + msecs_to_jiffies(
-			  priv->cfg->base_params->monitor_recover_period));
-	}
+
+	mod_timer(&priv->watchdog, jiffies +
+		  msecs_to_jiffies(IWL_WD_TICK(timeout)));
 }
-EXPORT_SYMBOL(iwl_bg_monitor_recover);
+EXPORT_SYMBOL(iwl_bg_watchdog);
+
+void iwl_setup_watchdog(struct iwl_priv *priv)
+{
+	unsigned int timeout = priv->cfg->base_params->wd_timeout;
 
+	if (timeout)
+		mod_timer(&priv->watchdog,
+			  jiffies + msecs_to_jiffies(IWL_WD_TICK(timeout)));
+	else
+		del_timer(&priv->watchdog);
+}
+EXPORT_SYMBOL(iwl_setup_watchdog);
 
 /*
  * extended beacon time format