bes2600: bus_reset on connection-loss storm to dodge assoc-comeback blackhole

When mac80211 declares connection loss against this AP (typically driven
by inactivity-deauth or beacon-loss), the userspace reauth that follows
sometimes enters a long blackhole: the AP responds to auth with success
but defers assoc with the 802.11v "assoc comeback" timer; ohm retries
faster than the comeback grants permission; the AP eventually fires an
unprotected deauth-reason-6 ("Class 2 frame received from non-
authenticated station"), and recovery only completes via cross-SSID or
cross-channel fallback. Receipts: ~86 s blackhole observed in the
phase-7 rep on 2026-05-07 02:42, with three subsequent BSSIDs returning
assoc comeback timeouts before reason-9 (STA_REQ_ASSOC_WITHOUT_AUTH)
fired. Documented in marfrit/besser:notes/phase4-2026-05-07.md.

When N=3 driver-side connection_loss decisions fire within a 60 s window
on the same vif, skip the ieee80211_connection_loss() path and trigger
the c5.2-introduced bes2600_chrdev_do_bus_reset() instead. The bus
reset removes and re-probes the chip; userspace re-associates with a
fresh chip state, dodging the AP's comeback-timer rejection cycle.

Predicted Phase 7 delta vs current baseline:
- api_connection_loss rate: unchanged (we don't address the trigger)
- conditional probability of >5 s blackhole given event: <= 30 %
- worst-case recovery: 86 s -> < 10 s

Contract pin: bes2600_chrdev_do_bus_reset(sbus_ops, sbus_priv) at
bes2600/bes_chardev.c:455, introduced by c5.2. The function is async-
returning: sbus_ops->bus_reset() schedules an SDIO rescan; the helper
waits up to 3 s for the remove() callback to clear sbus_priv, then
returns. Per-vif state is gone after this point, so the recover work
lives on bes2600_common (hw_priv) and uses the global bes2600_cdev for
the bus_reset call rather than dereferencing per-vif state.

Threshold (3 / 60 s) is well above the steady-state per-vif
connection_loss rate observed in the patch-A phase-7 rep (0.86/h under
sustained load), so a true storm is required to trip it.

Files touched:
- bes2600/bes2600.h: 3 counter fields on struct bes2600_vif, 1
  work_struct on struct bes2600_common, 3 prototypes
- bes2600/sta.c: 3 helpers + storm-account hook in
  bes2600_connection_loss_work + storm-init in bes2600_vif_setup +
  cancel_work_sync in the hw_priv shutdown path; #include bes_chardev.h
  was already pulled in by an earlier c-stack patch
- bes2600/main.c: INIT_WORK alongside other hw_priv work_structs
- bes2600/debug.c: ConnectionLossStormRecoveries seq_printf in the
  per-vif status seq_file output

The cw1200/cw1260 ancestor has no equivalent; this is a clean
addition. checkpatch.pl --no-tree --strict: clean (0/0/0).

Signed-off-by: Claude (noether) <claude@reauktion.de>

bes2600/bes2600.h

@@ -511,6 +511,9 @@ struct bes2600_common {
 	struct list_head coex_event_list;
 	spinlock_t coex_event_lock;
 
+	/* Connection-loss-storm fast-recover (Trigger A). See sta.c. */
+	struct work_struct connection_loss_storm_recover_work;
+
 	/* member for low power */
 	struct bes2600_pwr_t bes_power;
 
@@ -596,6 +599,11 @@ struct bes2600_vif {
         unsigned long           rx_timestamp;
         u32                     cipherType;
 
+	/* Decrypt-storm fast-recover (Trigger B). See txrx.c. */
+	unsigned long			decrypt_storm_window_start;
+	unsigned int			decrypt_storm_count;
+	unsigned int			decrypt_storm_recoveries;
+	struct work_struct		decrypt_storm_recover_work;
 
 	/* AP powersave */
 	u32			link_id_map;
@@ -622,6 +630,10 @@ struct bes2600_vif {
 	/* CQM Implementation */
 	struct delayed_work	bss_loss_work;
 	struct delayed_work	connection_loss_work;
+	/* Connection-loss-storm fast-recover (Trigger A). See sta.c. */
+	unsigned long		connection_loss_storm_window_start;
+	unsigned int		connection_loss_storm_count;
+	unsigned int		connection_loss_storm_recoveries;
 	struct work_struct	tx_failure_work;
 	int			delayed_link_loss;
 	spinlock_t		bss_loss_lock;
@@ -856,4 +868,13 @@ int bes2600_btusb_setup_pipes(struct sbus_priv *sbus_priv);
 void bes2600_btusb_uninit(struct usb_interface *interface);
 #endif
 
+/* Decrypt-storm fast-recover helpers — see txrx.c. */
+void bes2600_decrypt_storm_init(struct bes2600_vif *priv);
+void bes2600_decrypt_storm_account(struct bes2600_vif *priv);
+
+/* Connection-loss-storm fast-recover helpers — see sta.c. */
+void bes2600_connection_loss_storm_init(struct bes2600_vif *priv);
+bool bes2600_connection_loss_storm_account(struct bes2600_vif *priv);
+void bes2600_connection_loss_storm_recover(struct work_struct *work);
+
 #endif /* BES2600_H */

bes2600/bes2600_sdio.c

@@ -16,6 +16,7 @@
 #include <linux/mmc/host.h>
 #include <linux/mmc/sdio_func.h>
 #include <linux/mmc/card.h>
+#include <linux/mmc/core.h>
 #include <linux/mmc/sdio.h>
 #include <linux/spinlock.h>
 #include <net/mac80211.h>
@@ -1388,7 +1389,14 @@ static void bes2600_gpio_wakeup_mcu(struct sbus_priv *self, int flag)
 
 	/* error check */
 	if((self->gpio_wakup_flags & BIT(flag)) != 0) {
-		bes_err(			"repeat set gpio_wake_flag, sub_sys:%d", flag);
+		/*
+		 * Multiple subsystems holding wake is the steady-state case
+		 * (e.g. WIFI + BT both want MCU awake). Demoted from bes_err
+		 * to bes_devel since it isn't an error - the GPIO is already
+		 * asserted high and the subsystem is now also tracked.
+		 */
+		bes_devel("repeat set gpio_wake_flag, sub_sys:%d\n", flag);
+		self->gpio_wakup_flags |= BIT(flag);
 		mutex_unlock(&self->io_mutex);
 		return;
 	}
@@ -1420,7 +1428,11 @@ static void bes2600_gpio_allow_mcu_sleep(struct sbus_priv *self, int flag)
 
 	/* error check */
 	if((self->gpio_wakup_flags & BIT(flag)) == 0) {
-		bes_err(			"repeat clear gpio_wake_flag, sub_sys:%d", flag);
+		/*
+		 * Mirror of the wake path: a clear when the bit is already
+		 * clear is racy bookkeeping, not a hardware error.
+		 */
+		bes_devel("repeat clear gpio_wake_flag, sub_sys:%d\n", flag);
 		mutex_unlock(&self->io_mutex);
 		return;
 	}
@@ -1777,6 +1789,55 @@ static void bes2600_sdio_halt_device(struct sbus_priv *self)
 	sdio_work_debug(self);
 }
 
+/*
+ * Trigger an SDIO bus reset via mmc_hw_reset().
+ *
+ * With multiple SDIO functions probed (PineTab2 binds func 1 for WLAN and
+ * func 2 for the BT-companion path) mmc_sdio_hw_reset() takes the
+ * remove-and-rescan path: it marks the card removed and schedules
+ * mmc_rescan, which tears down the bound function drivers and re-detects
+ * the card on the next sweep, in turn reinvoking bes2600_sdio_probe().
+ *
+ * With a single function probed it instead invokes mmc_power_cycle()
+ * directly, which on PineTab2 toggles the wifi-reset GPIO via sdio_pwrseq.
+ *
+ * In both cases the chip ends up in a freshly reset state, which is the
+ * goal of the recovery path.
+ *
+ * mmc_hw_reset() must be called without holding the SDIO host claim --
+ * the multi-func remove-and-rescan path acquires the host claim via the
+ * mmc workqueue.
+ */
+static int bes2600_sdio_bus_reset(struct sbus_priv *self)
+{
+	struct mmc_host *host;
+	int ret;
+
+	if (!self || !self->func || !self->func->card)
+		return -EINVAL;
+
+	host = self->func->card->host;
+	ret = mmc_hw_reset(self->func->card);
+
+	/*
+	 * On multi-function SDIO cards (BES2600 has WLAN func 1 + BT
+	 * companion func 2), mmc_sdio_hw_reset() removes the card and
+	 * returns 1 to signal "remove happened, caller must trigger
+	 * rescan". The kernel does NOT auto-rescan in this case;
+	 * single-function cards take the rescan path inline and return 0.
+	 * Treat any non-negative return as success and force a rescan if
+	 * mmc_hw_reset signalled the multi-function path - otherwise the
+	 * card stays removed indefinitely after a wedge recovery,
+	 * leaving wifi (and the BT companion) silent until reboot.
+	 */
+	if (ret > 0) {
+		bes_info("multi-func mmc_hw_reset removed card; scheduling rescan\n");
+		mmc_detect_change(host, 0);
+		ret = 0;
+	}
+	return ret;
+}
+
 static bool bes2600_sdio_wakeup_source(struct sbus_priv *self)
 {
 	struct bes2600_platform_data_sdio *pdata = bes2600_get_platform_data();
@@ -1815,6 +1876,7 @@ static struct sbus_ops bes2600_sdio_sbus_ops = {
 	.gpio_sleep		= bes2600_gpio_allow_mcu_sleep,
 	.halt_device		= bes2600_sdio_halt_device,
 	.wakeup_source		= bes2600_sdio_wakeup_source,
+	.bus_reset		= bes2600_sdio_bus_reset,
 };
 
 static void bes2600_sdio_en_lp_cb(struct bes2600_common *hw_priv)

bes2600/bes_chardev.c

@@ -442,6 +442,60 @@ int bes2600_chrdev_do_system_close(const struct sbus_ops *sbus_ops, struct sbus_
 	return ret;
 }
 
+/*
+ * Hard-reset the bus and wait for the bus core to remove the chip.
+ *
+ * Used by the firmware-wedge recovery path on platforms where the normal
+ * power_switch(0) sequence has no effective chip-reset signal. The bus
+ * implementation triggers an asynchronous re-detect; this helper waits for
+ * the resulting remove() callback to clear bes2600_cdev.sbus_priv so that a
+ * subsequent bes2600_switch_wifi(true) sees a clean state and can wait on
+ * the fresh probe.
+ */
+int bes2600_chrdev_do_bus_reset(const struct sbus_ops *sbus_ops, struct sbus_priv *priv)
+{
+	int ret;
+	long status;
+
+	if (!sbus_ops || !priv)
+		return -EINVAL;
+
+	if (!sbus_ops->bus_reset)
+		return -EOPNOTSUPP;
+
+	bes_info("trigger bus reset to recover wedged firmware.\n");
+
+	ret = sbus_ops->bus_reset(priv);
+	if (ret) {
+		bes_err("bus_reset failed: %d\n", ret);
+		return ret;
+	}
+
+	/*
+	 * The bus reset is asynchronous: the bus core schedules a rescan
+	 * which removes the bound function drivers and then re-detects the
+	 * chip. Wait for the remove callback to clear sbus_priv. Do not
+	 * dereference 'priv' after this point -- it may already be freed.
+	 */
+	status = wait_event_timeout(bes2600_cdev.probe_done_wq,
+				    !bes2600_cdev.sbus_priv, HZ * 3);
+	WARN_ON(status <= 0);
+
+	return 0;
+}
+
+/*
+ * Trigger bes2600_chrdev_do_bus_reset() against the file-global
+ * bes2600_cdev. Used by host-side recovery paths outside this
+ * compilation unit (e.g. sta.c connection-loss-storm fast-recover) so
+ * those callers do not need to reach the static bes2600_cdev directly.
+ */
+int bes2600_chrdev_trigger_bus_reset(void)
+{
+	return bes2600_chrdev_do_bus_reset(bes2600_cdev.sbus_ops,
+					   bes2600_cdev.sbus_priv);
+}
+
 bool bes2600_chrdev_is_wifi_opened(void)
 {
 	bool wifi_opened = false;
@@ -540,8 +594,21 @@ static void bes2600_chrdev_wifi_force_close_work(struct work_struct *work)
 		/* unregister wifi */
 		bes2600_switch_wifi(0);
 
-		/* power down device if wifi is only opened */
-		if (bes2600_chrdev_check_system_close()) {
+		/*
+		 * Hard exception with a bus_reset implementation: tear the
+		 * bus down via mmc_hw_reset() (or equivalent) so the next
+		 * bringup probes a freshly reset chip. On PineTab2 this is
+		 * the only effective recovery path -- the existing
+		 * power_switch(0)/(1) sequence has no chip-reset signal of
+		 * its own (sdio_pwrseq owns wifi_reset).
+		 *
+		 * Soft close, or hard close on a board without bus_reset:
+		 * fall back to the legacy power_switch(0) sequence.
+		 */
+		if (bes2600_cdev.halt_dev && bes2600_cdev.sbus_ops->bus_reset) {
+			bes2600_chrdev_do_bus_reset(bes2600_cdev.sbus_ops,
+						    bes2600_cdev.sbus_priv);
+		} else if (bes2600_chrdev_check_system_close()) {
 			bes2600_chrdev_do_system_close(bes2600_cdev.sbus_ops,
 						bes2600_cdev.sbus_priv);
 		}

bes2600/bes_chardev.h

@@ -60,6 +60,8 @@ struct sbus_priv *bes2600_chrdev_get_sbus_priv_data(void);
 /* used to control device power down */
 int bes2600_chrdev_check_system_close(void);
 int bes2600_chrdev_do_system_close(const struct sbus_ops *sbus_ops, struct sbus_priv *priv);
+int bes2600_chrdev_do_bus_reset(const struct sbus_ops *sbus_ops, struct sbus_priv *priv);
+int bes2600_chrdev_trigger_bus_reset(void);
 void bes2600_chrdev_wakeup_bt(void);
 void bes2600_chrdev_wifi_force_close(struct bes2600_common *hw_priv, bool halt_dev);
 void bes2600_chrdev_usb_remove(struct bes2600_common *hw_priv);

bes2600/bes_pwr.c

@@ -467,6 +467,45 @@ static void bes2600_pwr_device_enter_lp_mode(struct bes2600_common *hw_priv)
 	bes_devel("device enter sleep\n");
 }
 
+/*
+ * Number of consecutive bes2600_pwr_enter_lp_mode timeouts (with zero
+ * PM_INDICATIONs received) before we conclude the firmware does not
+ * honor host-driven PSM and switch to a sticky skip path.
+ */
+#define BES2600_PM_UNSUPPORTED_THRESHOLD	3
+
+/*
+ * Latch pm_unsupported = true and force chip_pm_state = ACTIVE so the
+ * c6.2 wake-side skip branch covers bes2600_pwr_device_exit_lp_mode.
+ * Called after BES2600_PM_UNSUPPORTED_THRESHOLD consecutive enter_lp_mode
+ * timeouts with zero PM_INDICATIONs.
+ */
+static void bes2600_pwr_latch_pm_unsupported(struct bes2600_common *hw_priv)
+{
+	bes_warn("PSM not honored (%u timeouts), switching to skip mode\n",
+		 hw_priv->bes_power.pm_consecutive_timeouts);
+	hw_priv->bes_power.pm_unsupported = true;
+	atomic_set(&hw_priv->bes_power.chip_pm_state,
+		   BES2600_CHIP_PM_ACTIVE);
+
+	/*
+	 * Hold the MCU wake-flag bit permanently. Without this, every
+	 * sdio_rx_work invocation hits bes2600_gpio_wakeup_mcu(SDIO_RX)
+	 * when gpio_wakup_flags == 0, drives the GPIO high and msleeps
+	 * 10 ms per RX. With ~50 RX/s of beacons + multicast that's
+	 * ~50%% of the bes_sdio workqueue thread blocked in msleep,
+	 * which directly caps RX throughput. Holding the MCU bit makes
+	 * those calls bit-only bookkeeping (gpio_wakeup = (flags == 0)
+	 * stays false, no GPIO toggle, no msleep). The bit is never
+	 * cleared once pm_unsupported is set because
+	 * bes2600_pwr_device_enter_lp_mode is unreachable under the
+	 * early-return.
+	 */
+	if (hw_priv->sbus_ops->gpio_wake)
+		hw_priv->sbus_ops->gpio_wake(hw_priv->sbus_priv,
+					     GPIO_WAKE_FLAG_MCU);
+}
+
 static int bes2600_pwr_enter_lp_mode(struct bes2600_common *hw_priv)
 {
 	int i = 0;
@@ -476,6 +515,17 @@ static int bes2600_pwr_enter_lp_mode(struct bes2600_common *hw_priv)
 	char ip_str[20];
 	unsigned long status = 0;
 
+	/*
+	 * Sticky early-return when we've previously concluded the firmware
+	 * doesn't honor PSM. Each attempt would otherwise burn 5s on a
+	 * doomed wait_for_completion_timeout and produce a noisy three-line
+	 * cascade in dmesg every time power_down_work retries (every
+	 * ~10s). The chip stays in active mode, which on this firmware is
+	 * the de-facto state anyway.
+	 */
+	if (hw_priv->bes_power.pm_unsupported)
+		return -EOPNOTSUPP;
+
 	/* set interface low power configuration */
 	bes2600_for_each_vif(hw_priv, priv, i) {
 #ifdef P2P_MULTIVIF
@@ -571,6 +621,9 @@ static int bes2600_pwr_enter_lp_mode(struct bes2600_common *hw_priv)
 						atomic_set(&hw_priv->bes_power.chip_pm_state,
 							   BES2600_CHIP_PM_UNKNOWN);
 						timeouts++;
+						if (++hw_priv->bes_power.pm_consecutive_timeouts
+						    >= BES2600_PM_UNSUPPORTED_THRESHOLD)
+							bes2600_pwr_latch_pm_unsupported(hw_priv);
 					}
 				}
 			} else {
@@ -609,7 +662,8 @@ static int bes2600_pwr_enter_lp_mode(struct bes2600_common *hw_priv)
 		 * GPIO stays high and the bit clear here is purely
 		 * bookkeeping (so the next gpio_wake doesn't no-op).
 		 */
-		if (hw_priv->sbus_ops->gpio_sleep)
+		if (!hw_priv->bes_power.pm_unsupported &&
+		    hw_priv->sbus_ops->gpio_sleep)
 			hw_priv->sbus_ops->gpio_sleep(hw_priv->sbus_priv,
 						      GPIO_WAKE_FLAG_MCU);
 		ret = -ETIMEDOUT;
@@ -621,19 +675,61 @@ static int bes2600_pwr_enter_lp_mode(struct bes2600_common *hw_priv)
 static void bes2600_pwr_device_exit_lp_mode(struct bes2600_common *hw_priv)
 {
 	int ret = 0;
+	enum bes2600_chip_pm_state state;
 	struct wsm_operational_mode mode = {
 		.power_mode = wsm_power_mode_active,
 		.disableMoreFlagUsage = true,
 	};
 
-	bes_devel("host lock lmac\n");
-	if(hw_priv->sbus_ops->gpio_wake)
-		hw_priv->sbus_ops->gpio_wake(hw_priv->sbus_priv, GPIO_WAKE_FLAG_MCU);
+	/*
+	 * Consult chip_pm_state set by bes2600_pwr_notify_ps_changed().
+	 * If we last saw the firmware confirm ACTIVE, skip ONLY the
+	 * gpio_wake + sbus_active wake handshake - the GPIO is already
+	 * asserted high and the SDIO MCU subsystem is already running,
+	 * so another sbus_active() round-trip just hits its 200x2ms
+	 * timeout because the firmware has nothing to do.
+	 *
+	 * wsm_set_operational_mode() below is NOT part of the wake
+	 * handshake; it is the operational-mode setter the firmware
+	 * tracks per call. Skipping it leaves the chip's SDIO state
+	 * machine without a fresh operational-mode update, which on
+	 * PineTab2 wedges the bus (-EBUSY on next sdio_rx_work read)
+	 * within a few seconds of probe completion. So it must run
+	 * unconditionally.
+	 */
+	state = atomic_read(&hw_priv->bes_power.chip_pm_state);
+	if (state == BES2600_CHIP_PM_ACTIVE) {
+		bes_devel("device_exit_lp_mode: chip already ACTIVE, skipping wake handshake\n");
+	} else {
+		bes_devel("host lock lmac\n");
+		if (hw_priv->sbus_ops->gpio_wake)
+			hw_priv->sbus_ops->gpio_wake(hw_priv->sbus_priv,
+						      GPIO_WAKE_FLAG_MCU);
 
-	if(hw_priv->sbus_ops->sbus_active) {
-		ret = hw_priv->sbus_ops->sbus_active(hw_priv->sbus_priv, SUBSYSTEM_MCU);
-		if (ret)
-			bes_err("%s, active mcu fail\n", __func__);
+		if (hw_priv->sbus_ops->sbus_active) {
+			ret = hw_priv->sbus_ops->sbus_active(hw_priv->sbus_priv,
+							     SUBSYSTEM_MCU);
+			if (ret) {
+				/*
+				 * MCU_WAKEUP_READY did not arrive within
+				 * the SDIO handshake window. Record state
+				 * as UNKNOWN so the next exit_lp_mode call
+				 * also runs the full wake sequence (no
+				 * skip), but still send operational_mode
+				 * below to match pre-c6 behaviour - the
+				 * WSM may succeed even if the SDIO active
+				 * confirm was lost, and if it fails too,
+				 * we just emit a second devel-level error.
+				 * Repeated UNKNOWN is the signal for the
+				 * LMAC active-monitor to eventually
+				 * escalate to bus_reset (c5.2's
+				 * mmc_hw_reset path).
+				 */
+				bes_err("%s, active mcu fail\n", __func__);
+				atomic_set(&hw_priv->bes_power.chip_pm_state,
+					   BES2600_CHIP_PM_UNKNOWN);
+			}
+		}
 	}
 
 	ret = wsm_set_operational_mode(hw_priv, &mode, 0);
@@ -890,6 +986,8 @@ void bes2600_pwr_init(struct bes2600_common *hw_priv)
         mutex_init(&hw_priv->bes_power.pwr_mutex);
         atomic_set(&hw_priv->bes_power.dev_state, 0);
 	atomic_set(&hw_priv->bes_power.chip_pm_state, BES2600_CHIP_PM_UNKNOWN);
+	hw_priv->bes_power.pm_unsupported = false;
+	hw_priv->bes_power.pm_consecutive_timeouts = 0;
         init_completion(&hw_priv->bes_power.pm_enter_cmpl);
         sema_init(&hw_priv->bes_power.sync_lock, 1);
         device_set_wakeup_capable(hw_priv->pdev, true);
@@ -1279,6 +1377,18 @@ void bes2600_pwr_notify_ps_changed(struct bes2600_common *hw_priv, u8 psmode)
 	 * indication can prime a future wait against a freshly
 	 * reinit_completion()'ed state.
 	 */
+	/*
+	 * Any PM indication, whatever its psmode, proves the firmware is
+	 * actually emitting them. Reset the consecutive-timeout counter
+	 * so a transient stall doesn't permanently disable PSM, and clear
+	 * pm_unsupported if a previous run had latched it.
+	 */
+	hw_priv->bes_power.pm_consecutive_timeouts = 0;
+	if (hw_priv->bes_power.pm_unsupported) {
+		bes_warn("PM indication arrived after pm_unsupported was set; re-enabling PSM transitions\n");
+		hw_priv->bes_power.pm_unsupported = false;
+	}
+
 	if ((psmode & 0x01) != WSM_PSM_ACTIVE) {
 		atomic_set(&hw_priv->bes_power.chip_pm_state,
 			   BES2600_CHIP_PM_LP);

bes2600/bes_pwr.h

@@ -121,6 +121,15 @@ struct bes2600_pwr_t
         struct bes2600_pwr_event_t pwr_events[BES2600_DELAY_EVENT_NUM];
         atomic_t pm_set_in_process;
 	atomic_t chip_pm_state;
+	/*
+	 * Sticky flag set after BES2600_PM_UNSUPPORTED_THRESHOLD
+	 * consecutive enter_lp_mode timeouts with zero PM_INDICATIONs
+	 * received from firmware. Indicates this chip's firmware does
+	 * not honor host-driven PSM transitions; further attempts are
+	 * skipped to avoid the 5s timeout cascade.
+	 */
+	bool pm_unsupported;
+	unsigned int pm_consecutive_timeouts;
 };
 
 #ifdef CONFIG_BES2600_WOWLAN

bes2600/debug.c

@@ -542,6 +542,10 @@ static int bes2600_status_show_priv(struct seq_file *seq, void *v)
 		priv->listening ? " (listening)" : "");
 	seq_printf(seq, "Assoc:      %s\n",
 		bes2600_debug_join_status[priv->join_status]);
+	seq_printf(seq, "DecryptStormRecoveries: %u\n",
+		   priv->decrypt_storm_recoveries);
+	seq_printf(seq, "ConnectionLossStormRecoveries: %u\n",
+		   priv->connection_loss_storm_recoveries);
 	if (priv->rx_filter.promiscuous)
 		seq_puts(seq,   "Filter:     promisc\n");
 	else if (priv->rx_filter.fcs)

bes2600/main.c

@@ -484,6 +484,8 @@ static struct ieee80211_hw *bes2600_init_common(size_t hw_priv_data_len)
 	spin_lock_init(&hw_priv->rtsvalue_lock);
 	INIT_WORK(&hw_priv->dynamic_opt_txrx_work, bes2600_dynamic_opt_txrx_work);
 	INIT_WORK(&hw_priv->tx_policy_upload_work, tx_policy_upload_work);
+	INIT_WORK(&hw_priv->connection_loss_storm_recover_work,
+		  bes2600_connection_loss_storm_recover);
 	spin_lock_init(&hw_priv->event_queue_lock);
 	INIT_LIST_HEAD(&hw_priv->event_queue);
 	INIT_WORK(&hw_priv->event_handler, bes2600_event_handler);

bes2600/sbus.h

@@ -75,6 +75,14 @@ struct sbus_ops {
 	void (*halt_device)(struct sbus_priv *self);
 	bool (*wakeup_source)(struct sbus_priv *self);
 	int (*reboot)(struct sbus_priv *self);
+	/*
+	 * Force the host bus to re-detect and re-probe the chip. Called
+	 * from the firmware-wedge recovery path when power_switch() has no
+	 * effective chip-reset signal of its own (e.g. PineTab2, where the
+	 * wifi-reset GPIO is owned by sdio_pwrseq, not the bes2600 node).
+	 * Returns 0 on success or a negative errno.
+	 */
+	int (*bus_reset)(struct sbus_priv *self);
 };
 
 void bes2600_irq_handler(struct bes2600_common *priv);

bes2600/sta.c

@@ -266,6 +266,7 @@ void bes2600_stop(struct ieee80211_hw *dev, bool suspend)
 	cancel_work_sync(&hw_priv->coex_work);
 	coex_stop(hw_priv);
 #endif
+	cancel_work_sync(&hw_priv->connection_loss_storm_recover_work);
 
 	bes2600_wifi_stop(hw_priv);
 
@@ -448,6 +449,7 @@ void bes2600_remove_interface(struct ieee80211_hw *dev,
 	cancel_delayed_work_sync(&priv->join_timeout);
 	cancel_delayed_work_sync(&priv->set_cts_work);
 	cancel_delayed_work_sync(&priv->pending_offchanneltx_work);
+	cancel_work_sync(&priv->decrypt_storm_recover_work);
 
 	del_timer_sync(&priv->mcast_timeout);
 	/* TODO:COMBO: May be reset of these variables "delayed_link_loss and
@@ -1658,6 +1660,70 @@ report:
 	spin_unlock(&priv->bss_loss_lock);
 }
 
+/*
+ * Connection-loss-storm fast-recover (Trigger A).
+ *
+ * bes2600_connection_loss_work below is the driver's own decision-point
+ * to give up on a BSS (after bss-loss detection accumulates beyond
+ * tolerance) and tell mac80211 via ieee80211_connection_loss(). On the
+ * deployed pinetab2 stack a single ieee80211_connection_loss() event
+ * sometimes triggers a userspace reauth blackhole (assoc-comeback
+ * timeouts followed by AP unprotected-deauth-reason-6) that ends only
+ * via cross-channel/cross-SSID fallback and can take 80+ s. Receipts at
+ * https://git.reauktion.de/marfrit/besser, notes/phase4-2026-05-07.md.
+ *
+ * When N connection-loss decisions land within WINDOW on the same vif,
+ * skip the ieee80211_connection_loss() path and trigger a chip-level
+ * bus_reset (the c5.2-introduced bes2600_chrdev_do_bus_reset). The chip
+ * is removed and re-probed; userspace re-associates from a fresh state,
+ * dodging the assoc-comeback loop.
+ *
+ * Threshold (3 / 60 s) is chosen well above the steady-state per-vif
+ * connection-loss rate observed in the patch-A Phase-7 rep
+ * (0.86/h under sustained load), so a true storm is required.
+ *
+ * The recover work_struct lives on bes2600_common (hw_priv) so that
+ * scheduling it does not race with vif teardown after bus_reset frees
+ * the per-vif state.
+ */
+#define BES2600_CONNECTION_LOSS_STORM_THRESHOLD	3
+#define BES2600_CONNECTION_LOSS_STORM_WINDOW_MS	60000
+
+void bes2600_connection_loss_storm_recover(struct work_struct *work)
+{
+	bes_warn("[bes2600] connection-loss-storm fast-recover: bus_reset\n");
+	bes2600_chrdev_trigger_bus_reset();
+	/*
+	 * After bes2600_chrdev_do_bus_reset() returns, the SDIO core has
+	 * scheduled a remove + rescan; per-vif state may already be gone.
+	 * Do not dereference any per-vif pointer here.
+	 */
+}
+
+void bes2600_connection_loss_storm_init(struct bes2600_vif *priv)
+{
+	priv->connection_loss_storm_window_start = 0;
+	priv->connection_loss_storm_count = 0;
+	priv->connection_loss_storm_recoveries = 0;
+}
+
+bool bes2600_connection_loss_storm_account(struct bes2600_vif *priv)
+{
+	unsigned long now = jiffies;
+	unsigned long window =
+		msecs_to_jiffies(BES2600_CONNECTION_LOSS_STORM_WINDOW_MS);
+
+	if (priv->connection_loss_storm_window_start == 0 ||
+	    time_after(now, priv->connection_loss_storm_window_start + window)) {
+		priv->connection_loss_storm_window_start = now;
+		priv->connection_loss_storm_count = 1;
+		return false;
+	}
+
+	return ++priv->connection_loss_storm_count >=
+	       BES2600_CONNECTION_LOSS_STORM_THRESHOLD;
+}
+
 void bes2600_connection_loss_work(struct work_struct *work)
 {
 	struct bes2600_vif *priv =
@@ -1667,9 +1733,21 @@ void bes2600_connection_loss_work(struct work_struct *work)
 
 	bes_devel("[CQM] Reporting connection loss.\n");
 	bes2600_pwr_clear_busy_event(priv->hw_priv, BES_PWR_LOCK_ON_BSS_LOST);
-	if(bes2600_suspend_status_get(hw_priv)) {
+
+	if (bes2600_connection_loss_storm_account(priv)) {
+		bes_warn("[bes2600] connection-loss storm: %u in %u s, scheduling bus reset\n",
+			 priv->connection_loss_storm_count,
+			 BES2600_CONNECTION_LOSS_STORM_WINDOW_MS / 1000);
+		priv->connection_loss_storm_count = 0;
+		priv->connection_loss_storm_recoveries++;
+		schedule_work(&hw_priv->connection_loss_storm_recover_work);
+		/* bus_reset will tear the chip down; skip the mac80211 path. */
+		return;
+	}
+
+	if (bes2600_suspend_status_get(hw_priv))
 		bes2600_pending_unjoin_set(hw_priv, priv->if_id);
-	} else
+	else
 		ieee80211_connection_loss(priv->vif);
 #ifdef WIFI_BT_COEXIST_EPTA_ENABLE
 	// set disconnected in BSS_CHANGED_ASSOC
@@ -2619,6 +2697,8 @@ int bes2600_vif_setup(struct bes2600_vif *priv)
 
 	/* Setup per vif workitems and locks */
 	spin_lock_init(&priv->vif_lock);
+	bes2600_decrypt_storm_init(priv);
+	bes2600_connection_loss_storm_init(priv);
 	INIT_WORK(&priv->join_work, bes2600_join_work);
 	INIT_DELAYED_WORK(&priv->join_timeout, bes2600_join_timeout);
 	INIT_WORK(&priv->unjoin_work, bes2600_unjoin_work);

bes2600/txrx.c

@@ -25,6 +25,78 @@
 
 #define BES2600_INVALID_RATE_ID (0xFF)
 
+/*
+ * Decrypt-storm fast-recover (Trigger B).
+ *
+ * When the BES2600 firmware reports WSM_STATUS_DECRYPTFAILURE for a
+ * burst of received frames (typically because the host's PTK or GTK
+ * has fallen out of sync with the AP), the AP eventually concludes that
+ * the STA is not authenticated and emits an unprotected deauth-reason-6
+ * ("Class 2 frame received from non-authenticated station"). On the
+ * deployed pinetab2 + bes2600 stack this AP-initiated deauth has been
+ * observed to leave the link blackholed for up to 109 s before
+ * userspace finds a different SSID/channel to recover on. (Receipts at
+ * https://git.reauktion.de/marfrit/besser, notes/phase5-2026-05-06.md.)
+ *
+ * Recovery here pre-empts the AP: when we see THRESHOLD decrypt
+ * failures within WINDOW, we ask mac80211 for a clean reassoc via
+ * ieee80211_connection_loss(), which causes immediate disassociation
+ * and lets userspace auto-reconnect with fresh keys.
+ *
+ * mac80211 contract: ieee80211_connection_loss() may be called
+ * regardless of IEEE80211_HW_CONNECTION_MONITOR; it causes immediate
+ * disassociation without driver-side recovery attempts. See
+ * include/net/mac80211.h for the canonical doc-comment.
+ *
+ * The threshold is set well above the steady-state per-vif
+ * decrypt-fail rate observed in measurement (~1/min even under
+ * sustained 1 MB/s load), so a true storm is required to trip it.
+ */
+#define BES2600_DECRYPT_STORM_THRESHOLD	5
+#define BES2600_DECRYPT_STORM_WINDOW_MS	5000
+
+static void bes2600_decrypt_storm_recover_work(struct work_struct *work)
+{
+	struct bes2600_vif *priv = container_of(work, struct bes2600_vif,
+						decrypt_storm_recover_work);
+
+	if (!priv->vif)
+		return;
+
+	bes_warn("[bes2600] decrypt-storm fast-recover: forcing reassoc\n");
+	ieee80211_connection_loss(priv->vif);
+	priv->decrypt_storm_recoveries++;
+}
+
+void bes2600_decrypt_storm_init(struct bes2600_vif *priv)
+{
+	INIT_WORK(&priv->decrypt_storm_recover_work,
+		  bes2600_decrypt_storm_recover_work);
+	priv->decrypt_storm_window_start = 0;
+	priv->decrypt_storm_count = 0;
+	priv->decrypt_storm_recoveries = 0;
+}
+
+void bes2600_decrypt_storm_account(struct bes2600_vif *priv)
+{
+	unsigned long now = jiffies;
+	unsigned long window = msecs_to_jiffies(BES2600_DECRYPT_STORM_WINDOW_MS);
+
+	if (priv->decrypt_storm_window_start == 0 ||
+	    time_after(now, priv->decrypt_storm_window_start + window)) {
+		priv->decrypt_storm_window_start = now;
+		priv->decrypt_storm_count = 1;
+		return;
+	}
+
+	if (++priv->decrypt_storm_count >= BES2600_DECRYPT_STORM_THRESHOLD) {
+		priv->decrypt_storm_count = 0;
+		/* Skew the window so we don't re-fire on the same storm. */
+		priv->decrypt_storm_window_start = now + window;
+		schedule_work(&priv->decrypt_storm_recover_work);
+	}
+}
+
 #ifdef CONFIG_BES2600_TESTMODE
 #include "bes_nl80211_testmode_msg.h"
 #endif /* CONFIG_BES2600_TESTMODE */
@@ -1672,6 +1744,8 @@ void bes2600_rx_cb(struct bes2600_vif *priv,
 			goto drop;
 		} else {
 			bes_warn("[RX] Receive failure: %d.\n", arg->status);
+			if (arg->status == WSM_STATUS_DECRYPTFAILURE)
+				bes2600_decrypt_storm_account(priv);
 			goto drop;
 		}
 	}