diff --git a/drivers/usb/dwc3/drd.c b/drivers/usb/dwc3/drd.c
--- a/drivers/usb/dwc3/drd.c
+++ b/drivers/usb/dwc3/drd.c
@@ -55,7 +55,35 @@
 
 	spin_lock(&dwc->lock);
 	if (dwc->otg_restart_host) {
-		dwc3_otg_host_init(dwc);
+		/*
+		 * dwc3_otg_host_init() is documented just above its definition as
+		 * "should be called before Host controller driver is started": it
+		 * rewrites OCTL/OCFG, touches GUSB2PHYCFG and re-asserts port power
+		 * (OCTL.PrtPwrCtl). Doing that to a controller that is already
+		 * running corrupts an in-flight command.
+		 *
+		 * But otg_restart_host is set in dwc3_otg_irq() ONLY when
+		 * current_otg_role is already DWC3_OTG_ROLE_HOST, so this path
+		 * always ran it against a live host -- the exact condition the
+		 * comment forbids. Observed on the RV1106 86-panel as the xHCI
+		 * controller failing to answer a Stop Endpoint command, after which
+		 * xhci-hcd declares it dead and tears the bus down: an r8152 NIC
+		 * would stall a transfer and the controller died 8 times out of 8
+		 * between 452s and 587s. (flare-edge#160)
+		 *
+		 * On this board the event is always spurious: the dwc3 node
+		 * deliberately carries no extcon (wiring it pins the role to
+		 * peripheral on D1-modded panels) and CONIDSTS reads 0, so the role
+		 * cannot legitimately change while host is current. Skip the
+		 * re-init in that case. dwc3_set_mode() below still re-evaluates
+		 * the role, and if it has genuinely changed dwc3_otg_update() does
+		 * a proper dwc3_host_exit() before re-initialising.
+		 */
+		if (dwc->current_otg_role == DWC3_OTG_ROLE_HOST)
+			dev_dbg(dwc->dev,
+				"OTG event while host is live; skipping host re-init\n");
+		else
+			dwc3_otg_host_init(dwc);
 		dwc->otg_restart_host = false;
 	}