diff mbox series

[RFC,v2,60/96] cl8k: add recovery.c

Message ID 20220524113502.1094459-61-viktor.barna@celeno.com
State New
Headers show
Series wireless: cl8k driver for Celeno IEEE 802.11ax devices | expand

Commit Message

Viktor Barna May 24, 2022, 11:34 a.m. UTC
From: Viktor Barna <viktor.barna@celeno.com>

(Part of the split. Please, take a look at the cover letter for more
details).

Signed-off-by: Viktor Barna <viktor.barna@celeno.com>
---
 drivers/net/wireless/celeno/cl8k/recovery.c | 280 ++++++++++++++++++++
 1 file changed, 280 insertions(+)
 create mode 100644 drivers/net/wireless/celeno/cl8k/recovery.c
diff mbox series

Patch

diff --git a/drivers/net/wireless/celeno/cl8k/recovery.c b/drivers/net/wireless/celeno/cl8k/recovery.c
new file mode 100644
index 000000000000..dc0c33be9200
--- /dev/null
+++ b/drivers/net/wireless/celeno/cl8k/recovery.c
@@ -0,0 +1,280 @@ 
+// SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause
+/* Copyright(c) 2019-2022, Celeno Communications Ltd. */
+
+#include "hw.h"
+#include "main.h"
+#include "phy.h"
+#include "vif.h"
+#include "dfs.h"
+#include "maintenance.h"
+#include "vns.h"
+#include "config.h"
+#include "ela.h"
+#include "radio.h"
+#include "recovery.h"
+
+struct cl_recovery_work {
+	struct work_struct ws;
+	struct cl_hw *cl_hw;
+	int reason;
+};
+
+#define RECOVERY_POLL_TIMEOUT 6
+
+static void cl_recovery_poll_completion(struct cl_hw *cl_hw)
+{
+	u8 cntr = 0;
+
+	while (test_bit(CL_DEV_SW_RESTART, &cl_hw->drv_flags)) {
+		msleep(1000);
+
+		if (++cntr == RECOVERY_POLL_TIMEOUT) {
+			cl_dbg_verbose(cl_hw, "\n");
+			cl_dbg_err(cl_hw, "Driver handgup was detected!...");
+			break;
+		}
+	}
+}
+
+static void cl_recovery_start_hw(struct cl_hw *cl_hw)
+{
+	clear_bit(CL_DEV_STOP_HW, &cl_hw->drv_flags);
+
+	/* Restart MAC firmware... */
+	if (cl_main_on(cl_hw)) {
+		cl_dbg_err(cl_hw, "Couldn't turn platform on .. aborting\n");
+		return;
+	}
+
+	if (cl_msg_tx_reset(cl_hw)) {
+		cl_dbg_err(cl_hw, "Failed to send firmware reset .. aborting\n");
+		return;
+	}
+
+	set_bit(CL_DEV_SW_RESTART, &cl_hw->drv_flags);
+	clear_bit(CL_DEV_HW_RESTART, &cl_hw->drv_flags);
+
+	/* Hand over to mac80211 from here */
+	ieee80211_restart_hw(cl_hw->hw);
+	/* Start firmware */
+	if (cl_msg_tx_start(cl_hw)) {
+		cl_dbg_err(cl_hw, "Failed to send firmware start .. aborting\n");
+		return;
+	}
+
+	cl_recovery_poll_completion(cl_hw);
+}
+
+static void cl_recovery_stop_hw(struct cl_hw *cl_hw)
+{
+	/* Start recovery process */
+	ieee80211_stop_queues(cl_hw->hw);
+	cl_hw->recovery_db.in_recovery = true;
+
+	clear_bit(CL_DEV_STARTED, &cl_hw->drv_flags);
+	set_bit(CL_DEV_HW_RESTART, &cl_hw->drv_flags);
+	set_bit(CL_DEV_STOP_HW, &cl_hw->drv_flags);
+	/* Disable interrupts */
+	cl_irq_disable(cl_hw, cl_hw->ipc_e2a_irq.all);
+	cl_maintenance_stop(cl_hw);
+
+	mutex_lock(&cl_hw->dbginfo.mutex);
+
+	cl_main_off(cl_hw);
+
+	cl_hw->fw_active = false;
+	cl_hw->fw_send_start = false;
+
+	mutex_unlock(&cl_hw->dbginfo.mutex);
+
+	/* Reset it so MM_SET_FILTER_REQ will be called during the recovery */
+	cl_hw->rx_filter = 0;
+
+	/*
+	 * Reset channel/frequency parameters so that cl_msg_tx_set_channel()
+	 * will not be skipped in cl_ops_config()
+	 */
+	cl_hw->channel = 0;
+	cl_hw->primary_freq = 0;
+	cl_hw->center_freq = 0;
+}
+
+static void cl_recovery_process(struct cl_hw *cl_hw)
+{
+	int ret;
+	struct cl_chip *chip = cl_hw->chip;
+
+	mutex_lock(&chip->recovery_mutex);
+
+	cl_dbg_verbose(cl_hw, "Start\n");
+
+	cl_recovery_stop_hw(cl_hw);
+
+	if (chip->conf->ci_phy_dev != PHY_DEV_DUMMY) {
+		cl_phy_reset(cl_hw);
+
+		ret = cl_phy_load_recovery(cl_hw);
+		if (ret) {
+			cl_dbg_err(cl_hw, "cl_phy_load_recovery failed %d\n", ret);
+			goto out;
+		}
+	}
+
+	cl_recovery_start_hw(cl_hw);
+
+out:
+	mutex_unlock(&chip->recovery_mutex);
+}
+
+static void cl_recovery_handler(struct cl_hw *cl_hw, int reason)
+{
+	unsigned long recovery_diff = jiffies_to_msecs(jiffies - cl_hw->recovery_db.last_restart);
+
+	cl_hw->recovery_db.restart_cnt++;
+
+	if (recovery_diff > cl_hw->conf->ce_fw_watchdog_limit_time) {
+		cl_hw->recovery_db.restart_cnt = 1;
+	} else if (cl_hw->recovery_db.restart_cnt > cl_hw->conf->ce_fw_watchdog_limit_count) {
+		cl_dbg_verbose(cl_hw, "Too many failures... aborting\n");
+		cl_hw->conf->ce_fw_watchdog_mode = FW_WD_DISABLE;
+		return;
+	}
+
+	cl_hw->recovery_db.last_restart = jiffies;
+
+	/* Count recovery attempts for statistics */
+	cl_hw->fw_recovery_cntr++;
+	cl_dbg_trace(cl_hw, "Recovering from firmware failure, attempt #%i\n",
+		     cl_hw->fw_recovery_cntr);
+
+	cl_recovery_process(cl_hw);
+}
+
+static void cl_recovery_work_do(struct work_struct *ws)
+{
+	/* Worker for restarting hw. */
+	struct cl_recovery_work *recovery_work = container_of(ws, struct cl_recovery_work, ws);
+
+	recovery_work->cl_hw->assert_info.restart_sched = false;
+	cl_recovery_handler(recovery_work->cl_hw, recovery_work->reason);
+	kfree(recovery_work);
+}
+
+static void cl_recovery_work_sched(struct cl_hw *cl_hw, int reason)
+{
+	/*
+	 * Schedule work to restart device and firmware
+	 * This is scheduled when driver detects hw assert storm.
+	 */
+	struct cl_recovery_work *recovery_work;
+
+	if (!cl_hw->ipc_env || cl_hw->is_stop_context) {
+		cl_dbg_warn(cl_hw, "Skip recovery - Running down!\n");
+		return;
+	}
+
+	/* If restart is already scheduled - exit */
+	if (cl_hw->assert_info.restart_sched)
+		return;
+
+	cl_hw->assert_info.restart_sched = true;
+
+	/* Recovery_work will be freed by cl_recovery_work_do */
+	recovery_work = kzalloc(sizeof(*recovery_work), GFP_ATOMIC);
+
+	if (!recovery_work)
+		return;
+
+	INIT_WORK(&recovery_work->ws, cl_recovery_work_do);
+	recovery_work->cl_hw = cl_hw;
+	recovery_work->reason = reason;
+
+	queue_work(cl_hw->drv_workqueue, &recovery_work->ws);
+}
+
+bool cl_recovery_in_progress(struct cl_hw *cl_hw)
+{
+	return cl_hw->recovery_db.in_recovery;
+}
+
+void cl_recovery_reconfig_complete(struct cl_hw *cl_hw)
+{
+	clear_bit(CL_DEV_SW_RESTART, &cl_hw->drv_flags);
+
+	if (cl_ela_is_on(cl_hw->chip)) {
+		cl_ela_lcu_reset(cl_hw->chip);
+		cl_ela_lcu_apply_config(cl_hw->chip);
+	}
+
+#ifdef CONFIG_CL8K_DYN_MCAST_RATE
+	cl_dyn_mcast_rate_recovery(cl_hw);
+
+#endif /* CONFIG_CL8K_DYN_MCAST_RATE */
+#ifdef CONFIG_CL8K_DYN_BCAST_RATE
+	cl_dyn_bcast_rate_recovery(cl_hw);
+
+#endif /* CONFIG_CL8K_DYN_BCAST_RATE */
+	/* DFS recovery */
+	cl_dfs_recovery(cl_hw);
+
+	/* VNS recovery */
+	cl_vns_recovery(cl_hw);
+
+	/* Restore EDCA configuration */
+	cl_edca_recovery(cl_hw);
+
+	/* Temperature  recovery */
+	cl_temperature_recovery(cl_hw);
+
+	/* Sounding recovery */
+	cl_sounding_recovery(cl_hw);
+
+	/*
+	 * Update Tx params for all connected stations to sync firmware after the
+	 * recovery process. Should be called after cl_mu_ofdma_grp_recovery to let
+	 * MU-OFDMA rates in FW be updated successfully
+	 */
+	cl_wrs_api_recovery(cl_hw);
+
+	/* Enable maintenance timers back */
+	cl_maintenance_start(cl_hw);
+	if (cl_radio_is_on(cl_hw)) {
+		/*
+		 * Rearm last_tbtt_ind so that error message will
+		 * not be printed in cl_irq_status_tbtt()
+		 */
+		cl_hw->last_tbtt_irq = jiffies;
+
+		cl_msg_tx_set_idle(cl_hw, MAC_ACTIVE, true);
+	}
+
+	cl_hw->recovery_db.in_recovery = false;
+
+	pr_debug("cl_recovery: complete\n");
+
+	cl_rx_post_recovery(cl_hw);
+}
+
+void cl_recovery_start(struct cl_hw *cl_hw, int reason)
+{
+	/* Prevent new messages to be sent until firmware has recovered */
+	set_bit(CL_DEV_FW_ERROR, &cl_hw->drv_flags);
+
+	switch (cl_hw->conf->ce_fw_watchdog_mode) {
+	case FW_WD_DISABLE:
+		cl_dbg_info(cl_hw, "Skip recovery - Watchdog is off!\n");
+		break;
+
+	case FW_WD_INTERNAL_RECOVERY:
+		cl_recovery_work_sched(cl_hw, reason);
+		break;
+
+	case FW_WD_DRV_RELOAD:
+		/* TODO: Implement netlink hint to the userspace */
+		cl_dbg_info(cl_hw, "RELOAD handler is absent, doing nothing");
+		break;
+
+	default:
+		break;
+	}
+}