ath6kl: Add support to detect fw error through heart beat

This patch adds support to detect fw error condition by sending
periodic message (heart beat challenge) to firmware. Upon reception
of the message, fw would send a response event to driver. When
there are no reponses from fw for about 5 cmd driver would
trigger the recovery logic assuming that fw has gone into an
error state.

Capable fw will advertise this capability through
ATH6KL_FW_CAPABILITY_HEART_BEAT_POLL bit. This feature
is disabled by default, can be enabled through a modparam
(heart_beat_poll). This modparam also confiures the polling
interval in msecs.

Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

7 files changed, 129 insertions, 9 deletions

diff --git a/drivers/net/wireless/ath/ath6kl/cfg80211.c b/drivers/net/wireless/ath/ath6kl/cfg80211.c
index 8cf146b..c8b6be4 100644
--- a/drivers/net/wireless/ath/ath6kl/cfg80211.c
+++ b/drivers/net/wireless/ath/ath6kl/cfg80211.c
@@ -2517,7 +2517,7 @@ static int __ath6kl_cfg80211_resume(struct wiphy *wiphy)
 	if (err)
 		return err;
 
-	ar->fw_recovery.enable = true;
+	ath6kl_recovery_resume(ar);
 
 	return 0;
 }
diff --git a/drivers/net/wireless/ath/ath6kl/core.c b/drivers/net/wireless/ath/ath6kl/core.c
index adcaa96..fd5dd3a 100644
--- a/drivers/net/wireless/ath/ath6kl/core.c
+++ b/drivers/net/wireless/ath/ath6kl/core.c
@@ -33,6 +33,7 @@ static unsigned int wow_mode;
 static unsigned int uart_debug;
 static unsigned int ath6kl_p2p;
 static unsigned int testmode;
+static unsigned int heart_beat_poll;
 
 module_param(debug_mask, uint, 0644);
 module_param(suspend_mode, uint, 0644);
@@ -40,6 +41,9 @@ module_param(wow_mode, uint, 0644);
 module_param(uart_debug, uint, 0644);
 module_param(ath6kl_p2p, uint, 0644);
 module_param(testmode, uint, 0644);
+module_param(heart_beat_poll, uint, 0644);
+MODULE_PARM_DESC(heart_beat_poll, "Enable fw error detection periodic" \
+		 "polling. This also specifies the polling interval in msecs");
 
 void ath6kl_core_tx_complete(struct ath6kl *ar, struct sk_buff *skb)
 {
@@ -202,6 +206,11 @@ int ath6kl_core_init(struct ath6kl *ar, enum ath6kl_htc_type htc_type)
 	ath6kl_dbg(ATH6KL_DBG_TRC, "%s: name=%s dev=0x%p, ar=0x%p\n",
 		   __func__, wdev->netdev->name, wdev->netdev, ar);
 
+	if (heart_beat_poll &&
+	    test_bit(ATH6KL_FW_CAPABILITY_HEART_BEAT_POLL,
+		     ar->fw_capabilities))
+		ar->fw_recovery.hb_poll = heart_beat_poll;
+
 	ath6kl_recovery_init(ar);
 
 	return ret;
diff --git a/drivers/net/wireless/ath/ath6kl/core.h b/drivers/net/wireless/ath/ath6kl/core.h
index c7dcdad..b2cbecf 100644
--- a/drivers/net/wireless/ath/ath6kl/core.h
+++ b/drivers/net/wireless/ath/ath6kl/core.h
@@ -130,6 +130,12 @@ enum ath6kl_fw_capability {
 	/* Firmware supports sched scan decoupled from host sleep */
 	ATH6KL_FW_CAPABILITY_SCHED_SCAN_V2,
 
+	/*
+	 * Firmware capability for hang detection through heart beat
+	 * challenge messages.
+	 */
+	ATH6KL_FW_CAPABILITY_HEART_BEAT_POLL,
+
 	/* this needs to be last */
 	ATH6KL_FW_CAPABILITY_MAX,
 };
@@ -649,8 +655,11 @@ enum ath6kl_state {
 };
 
 /* Fw error recovery */
+#define ATH6KL_HB_RESP_MISS_THRES	5
+
 enum ath6kl_fw_err {
 	ATH6KL_FW_ASSERT,
+	ATH6KL_FW_HB_RESP_FAILURE,
 };
 
 struct ath6kl {
@@ -800,6 +809,11 @@ struct ath6kl {
 		bool enable;
 		struct work_struct recovery_work;
 		unsigned long err_reason;
+		unsigned long hb_poll;
+		struct timer_list hb_timer;
+		u32 seq_num;
+		bool hb_pending;
+		u8 hb_misscnt;
 	} fw_recovery;
 
 #ifdef CONFIG_ATH6KL_DEBUG
@@ -940,7 +954,9 @@ void ath6kl_core_destroy(struct ath6kl *ar);
 /* Fw error recovery */
 void ath6kl_init_hw_restart(struct ath6kl *ar);
 void ath6kl_recovery_err_notify(struct ath6kl *ar, enum ath6kl_fw_err reason);
+void ath6kl_recovery_hb_event(struct ath6kl *ar, u32 cookie);
 void ath6kl_recovery_init(struct ath6kl *ar);
 void ath6kl_recovery_cleanup(struct ath6kl *ar);
 void ath6kl_recovery_suspend(struct ath6kl *ar);
+void ath6kl_recovery_resume(struct ath6kl *ar);
 #endif /* CORE_H */
diff --git a/drivers/net/wireless/ath/ath6kl/init.c b/drivers/net/wireless/ath/ath6kl/init.c
index 301443c..6e270fa 100644
--- a/drivers/net/wireless/ath/ath6kl/init.c
+++ b/drivers/net/wireless/ath/ath6kl/init.c
@@ -1697,9 +1697,6 @@ int ath6kl_init_hw_stop(struct ath6kl *ar)
 
 void ath6kl_init_hw_restart(struct ath6kl *ar)
 {
-
-	ar->state = ATH6KL_STATE_RECOVERY;
-
 	ath6kl_cfg80211_stop_all(ar);
 
 	if (__ath6kl_init_hw_stop(ar))
@@ -1709,9 +1706,6 @@ void ath6kl_init_hw_restart(struct ath6kl *ar)
 		ath6kl_dbg(ATH6KL_DBG_RECOVERY, "Failed to restart during fw error recovery\n");
 		return;
 	}
-
-	ar->state = ATH6KL_STATE_ON;
-	ar->fw_recovery.err_reason = 0;
 }
 
 /* FIXME: move this to cfg80211.c and rename to ath6kl_cfg80211_vif_stop() */
diff --git a/drivers/net/wireless/ath/ath6kl/recovery.c b/drivers/net/wireless/ath/ath6kl/recovery.c
index c225fc4..4e3f205 100644
--- a/drivers/net/wireless/ath/ath6kl/recovery.c
+++ b/drivers/net/wireless/ath/ath6kl/recovery.c
@@ -23,7 +23,18 @@ static void ath6kl_recovery_work(struct work_struct *work)
 	struct ath6kl *ar = container_of(work, struct ath6kl,
 					 fw_recovery.recovery_work);
 
+	ar->state = ATH6KL_STATE_RECOVERY;
+
+	del_timer_sync(&ar->fw_recovery.hb_timer);
+
 	ath6kl_init_hw_restart(ar);
+
+	ar->state = ATH6KL_STATE_ON;
+	ar->fw_recovery.err_reason = 0;
+
+	if (ar->fw_recovery.enable)
+		mod_timer(&ar->fw_recovery.hb_timer, jiffies +
+			  msecs_to_jiffies(ar->fw_recovery.hb_poll));
 }
 
 void ath6kl_recovery_err_notify(struct ath6kl *ar, enum ath6kl_fw_err reason)
@@ -37,18 +48,72 @@ void ath6kl_recovery_err_notify(struct ath6kl *ar, enum ath6kl_fw_err reason)
 		queue_work(ar->ath6kl_wq, &ar->fw_recovery.recovery_work);
 }
 
+void ath6kl_recovery_hb_event(struct ath6kl *ar, u32 cookie)
+{
+	if (cookie == ar->fw_recovery.seq_num)
+		ar->fw_recovery.hb_pending = false;
+}
+
+static void ath6kl_recovery_hb_timer(unsigned long data)
+{
+	struct ath6kl *ar = (struct ath6kl *) data;
+	int err;
+
+	if (!ar->fw_recovery.enable)
+		return;
+
+	if (ar->fw_recovery.hb_pending)
+		ar->fw_recovery.hb_misscnt++;
+	else
+		ar->fw_recovery.hb_misscnt = 0;
+
+	if (ar->fw_recovery.hb_misscnt > ATH6KL_HB_RESP_MISS_THRES) {
+		ar->fw_recovery.hb_misscnt = 0;
+		ar->fw_recovery.seq_num = 0;
+		ar->fw_recovery.hb_pending = false;
+		ath6kl_recovery_err_notify(ar, ATH6KL_FW_HB_RESP_FAILURE);
+		return;
+	}
+
+	ar->fw_recovery.seq_num++;
+	ar->fw_recovery.hb_pending = true;
+
+	err = ath6kl_wmi_get_challenge_resp_cmd(ar->wmi,
+						ar->fw_recovery.seq_num, 0);
+	if (err)
+		ath6kl_warn("Failed to send hb challenge request, err:%d\n",
+			    err);
+
+	if ((ar->state == ATH6KL_STATE_RECOVERY) || !ar->fw_recovery.enable)
+		return;
+
+	mod_timer(&ar->fw_recovery.hb_timer, jiffies +
+		  msecs_to_jiffies(ar->fw_recovery.hb_poll));
+}
+
 void ath6kl_recovery_init(struct ath6kl *ar)
 {
 	struct ath6kl_fw_recovery *recovery = &ar->fw_recovery;
 
 	recovery->enable = true;
 	INIT_WORK(&recovery->recovery_work, ath6kl_recovery_work);
+	recovery->seq_num = 0;
+	recovery->hb_misscnt = 0;
+	ar->fw_recovery.hb_pending = false;
+	ar->fw_recovery.hb_timer.function = ath6kl_recovery_hb_timer;
+	ar->fw_recovery.hb_timer.data = (unsigned long) ar;
+	init_timer_deferrable(&ar->fw_recovery.hb_timer);
+
+	if (ar->fw_recovery.hb_poll)
+		mod_timer(&ar->fw_recovery.hb_timer, jiffies +
+			  msecs_to_jiffies(ar->fw_recovery.hb_poll));
 }
 
 void ath6kl_recovery_cleanup(struct ath6kl *ar)
 {
 	ar->fw_recovery.enable = false;
 
+	del_timer_sync(&ar->fw_recovery.hb_timer);
 	cancel_work_sync(&ar->fw_recovery.recovery_work);
 }
 
@@ -56,7 +121,27 @@ void ath6kl_recovery_suspend(struct ath6kl *ar)
 {
 	ath6kl_recovery_cleanup(ar);
 
+	if (!ar->fw_recovery.err_reason)
+		return;
+
 	/* Process pending fw error detection */
-	if (ar->fw_recovery.err_reason)
-		ath6kl_init_hw_restart(ar);
+	ar->fw_recovery.err_reason = 0;
+	WARN_ON(ar->state != ATH6KL_STATE_ON);
+	ar->state = ATH6KL_STATE_RECOVERY;
+	ath6kl_init_hw_restart(ar);
+	ar->state = ATH6KL_STATE_ON;
+}
+
+void ath6kl_recovery_resume(struct ath6kl *ar)
+{
+	ar->fw_recovery.enable = true;
+
+	if (!ar->fw_recovery.hb_poll)
+		return;
+
+	ar->fw_recovery.hb_pending = false;
+	ar->fw_recovery.seq_num = 0;
+	ar->fw_recovery.hb_misscnt = 0;
+	mod_timer(&ar->fw_recovery.hb_timer,
+		  jiffies + msecs_to_jiffies(ar->fw_recovery.hb_poll));
 }
diff --git a/drivers/net/wireless/ath/ath6kl/wmi.c b/drivers/net/wireless/ath/ath6kl/wmi.c
index cd2db42..68b46bd 100644
--- a/drivers/net/wireless/ath/ath6kl/wmi.c
+++ b/drivers/net/wireless/ath/ath6kl/wmi.c
@@ -3767,6 +3767,19 @@ int ath6kl_wmi_set_inact_period(struct wmi *wmi, u8 if_idx, int inact_timeout)
 				   NO_SYNC_WMIFLAG);
 }
 
+static void ath6kl_wmi_hb_challenge_resp_event(struct wmi *wmi, u8 *datap,
+					       int len)
+{
+	struct wmix_hb_challenge_resp_cmd *cmd;
+
+	if (len < sizeof(struct wmix_hb_challenge_resp_cmd))
+		return;
+
+	cmd = (struct wmix_hb_challenge_resp_cmd *) datap;
+	ath6kl_recovery_hb_event(wmi->parent_dev,
+				 le32_to_cpu(cmd->cookie));
+}
+
 static int ath6kl_wmi_control_rx_xtnd(struct wmi *wmi, struct sk_buff *skb)
 {
 	struct wmix_cmd_hdr *cmd;
@@ -3791,6 +3804,7 @@ static int ath6kl_wmi_control_rx_xtnd(struct wmi *wmi, struct sk_buff *skb)
 	switch (id) {
 	case WMIX_HB_CHALLENGE_RESP_EVENTID:
 		ath6kl_dbg(ATH6KL_DBG_WMI, "wmi event hb challenge resp\n");
+		ath6kl_wmi_hb_challenge_resp_event(wmi, datap, len);
 		break;
 	case WMIX_DBGLOG_EVENTID:
 		ath6kl_dbg(ATH6KL_DBG_WMI, "wmi event dbglog len %d\n", len);
diff --git a/drivers/net/wireless/ath/ath6kl/wmi.h b/drivers/net/wireless/ath/ath6kl/wmi.h
index e916e57..98b1755 100644
--- a/drivers/net/wireless/ath/ath6kl/wmi.h
+++ b/drivers/net/wireless/ath/ath6kl/wmi.h
@@ -2716,6 +2716,8 @@ int ath6kl_wmi_set_inact_period(struct wmi *wmi, u8 if_idx, int inact_timeout);
 
 void ath6kl_wmi_sscan_timer(unsigned long ptr);
 
+int ath6kl_wmi_get_challenge_resp_cmd(struct wmi *wmi, u32 cookie, u32 source);
+
 struct ath6kl_vif *ath6kl_get_vif_by_index(struct ath6kl *ar, u8 if_idx);
 void *ath6kl_wmi_init(struct ath6kl *devt);
 void ath6kl_wmi_shutdown(struct wmi *wmi);