Merge branch 'master' into fork/HiFiPhile/h7rs

25 files changed, 499 insertions, 119 deletions

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 55c52033c..99d3059fc 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -1,9 +1,6 @@
-# TODO more docs and example on how to use this file
-# TINYUSB_TARGET_PREFIX and TINYUSB_TARGET_SUFFIX can be used to change the name of the target
-
 cmake_minimum_required(VERSION 3.20)
 
-# Add tinyusb to a existing target
+# Add tinyusb to a existing target, DCD and HCD drivers are not included
 function(tinyusb_target_add TARGET)
   target_sources(${TARGET} PRIVATE
     # common
diff --git a/src/class/net/ecm_rndis_device.c b/src/class/net/ecm_rndis_device.c
index a54e6d662..299eb97c8 100644
--- a/src/class/net/ecm_rndis_device.c
+++ b/src/class/net/ecm_rndis_device.c
@@ -81,6 +81,7 @@ typedef struct {
 static netd_interface_t _netd_itf;
 CFG_TUD_MEM_SECTION static netd_epbuf_t _netd_epbuf;
 static bool can_xmit;
+static bool ecm_link_is_up = true;  // Store link state for ECM mode
 
 void tud_network_recv_renew(void) {
   usbd_edpt_xfer(0, _netd_itf.ep_out, _netd_epbuf.rx, NETD_PACKET_SIZE);
@@ -95,7 +96,11 @@ void netd_report(uint8_t *buf, uint16_t len) {
   const uint8_t rhport = 0;
   len = tu_min16(len, sizeof(ecm_notify_t));
 
-  TU_VERIFY(usbd_edpt_claim(rhport, _netd_itf.ep_notif), );
+  if (!usbd_edpt_claim(rhport, _netd_itf.ep_notif)) {
+    TU_LOG1("ECM: Failed to claim notification endpoint\n");
+    return;
+  }
+
   memcpy(_netd_epbuf.notify, buf, len);
   usbd_edpt_xfer(rhport, _netd_itf.ep_notif, _netd_epbuf.notify, len);
 }
@@ -181,8 +186,6 @@ uint16_t netd_open(uint8_t rhport, tusb_desc_interface_t const * itf_desc, uint1
     // Open endpoint pair for RNDIS
     TU_ASSERT(usbd_open_edpt_pair(rhport, p_desc, 2, TUSB_XFER_BULK, &_netd_itf.ep_out, &_netd_itf.ep_in), 0);
 
-    tud_network_init_cb();
-
     // we are ready to transmit a packet
     can_xmit = true;
 
@@ -196,11 +199,11 @@ uint16_t netd_open(uint8_t rhport, tusb_desc_interface_t const * itf_desc, uint1
 }
 
 static void ecm_report(bool nc) {
-  const ecm_notify_t ecm_notify_nc = {
+  ecm_notify_t ecm_notify_nc = {
     .header = {
       .bmRequestType = 0xA1,
       .bRequest = 0, /* NETWORK_CONNECTION aka NetworkConnection */
-      .wValue = 1,   /* Connected */
+      .wValue = ecm_link_is_up ? 1 : 0,   /* Use current link state */
       .wLength = 0,
     },
   };
@@ -259,7 +262,6 @@ bool netd_control_xfer_cb (uint8_t rhport, uint8_t stage, tusb_control_request_t
 
                 // TODO should be merge with RNDIS's after endpoint opened
                 // Also should have opposite callback for application to disable network !!
-                tud_network_init_cb();
                 can_xmit = true; // we are ready to transmit a packet
                 tud_network_recv_renew(); // prepare for incoming packets
               }
@@ -286,7 +288,10 @@ bool netd_control_xfer_cb (uint8_t rhport, uint8_t stage, tusb_control_request_t
           /* the only required CDC-ECM Management Element Request is SetEthernetPacketFilter */
           if (0x43 /* SET_ETHERNET_PACKET_FILTER */ == request->bRequest) {
             tud_control_xfer(rhport, request, NULL, 0);
-            ecm_report(true);
+            // Only send connection notification if link is up
+            if (ecm_link_is_up) {
+              ecm_report(true);
+            }
           }
         } else {
           if (request->bmRequestType_bit.direction == TUSB_DIR_IN) {
@@ -363,9 +368,8 @@ bool netd_xfer_cb(uint8_t rhport, uint8_t ep_addr, xfer_result_t result, uint32_
   }
 
   if (_netd_itf.ecm_mode && (ep_addr == _netd_itf.ep_notif)) {
-    if (sizeof(tusb_control_request_t) == xferred_bytes) {
-      ecm_report(false);
-    }
+    // Notification transfer complete - endpoint is now free
+    // Don't automatically send speed change notification after link state changes
   }
 
   return true;
@@ -398,4 +402,31 @@ void tud_network_xmit(void *ref, uint16_t arg) {
   do_in_xfer(_netd_epbuf.tx, len);
 }
 
+// Set the network link state (up/down) and notify the host
+void tud_network_link_state(uint8_t rhport, bool is_up) {
+  (void)rhport;
+
+  if (_netd_itf.ecm_mode) {
+    ecm_link_is_up = is_up;
+
+    // For ECM mode, send network connection notification only
+    // Don't trigger speed change notification for link state changes
+    ecm_notify_t notify = {
+      .header = {
+        .bmRequestType = 0xA1,
+        .bRequest = 0,        /* NETWORK_CONNECTION */
+        .wValue = is_up ? 1 : 0,  /* 0 = disconnected, 1 = connected */
+        .wLength = 0,
+      },
+    };
+    notify.header.wIndex = _netd_itf.itf_num;
+    netd_report((uint8_t *)&notify, sizeof(notify.header));
+  } else {
+    // For RNDIS mode, we would need to implement RNDIS status indication
+    // This is more complex and requires RNDIS_INDICATE_STATUS_MSG
+    // For now, RNDIS doesn't support dynamic link state changes
+    (void)is_up;
+  }
+}
+
 #endif
diff --git a/src/class/net/ncm_device.c b/src/class/net/ncm_device.c
index f9fda0698..02833c5f1 100644
--- a/src/class/net/ncm_device.c
+++ b/src/class/net/ncm_device.c
@@ -110,6 +110,7 @@ typedef struct {
     NOTIFICATION_DONE
   } notification_xmit_state;                            // state of notification transmission
   bool notification_xmit_is_running;                    // notification is currently transmitted
+  bool link_is_up;                                      // current link state
 
   // misc
   bool tud_network_recv_renew_active;                   // tud_network_recv_renew() is active (avoid recursive invocations)
@@ -218,7 +219,7 @@ static void notification_xmit(uint8_t rhport, bool force_next) {
           .direction = TUSB_DIR_IN
         },
         .bRequest = CDC_NOTIF_NETWORK_CONNECTION,
-        .wValue = 1 /* Connected */,
+        .wValue = ncm_interface.link_is_up ? 1 : 0,  /* Dynamic link state */
         .wIndex = ncm_interface.itf_num,
         .wLength = 0,
       },
@@ -232,6 +233,7 @@ static void notification_xmit(uint8_t rhport, bool force_next) {
     ncm_interface.notification_xmit_is_running = true;
   } else {
     TU_LOG_DRV("  NOTIFICATION_FINISHED\n");
+    ncm_interface.notification_xmit_is_running = false;
   }
 } // notification_xmit
 
@@ -755,6 +757,32 @@ static void tud_network_recv_renew_r(uint8_t rhport) {
   tud_network_recv_renew();
 } // tud_network_recv_renew
 
+/**
+ * Set the link state and send notification to host
+ */
+void tud_network_link_state(uint8_t rhport, bool is_up) {
+  TU_LOG_DRV("tud_network_link_state(%d, %d)\n", rhport, is_up);
+
+  if (ncm_interface.link_is_up == is_up) {
+    // No change in link state
+    return;
+  }
+
+  ncm_interface.link_is_up = is_up;
+
+  // Only send notification if we have an active data interface
+  if (ncm_interface.itf_data_alt != 1) {
+    TU_LOG_DRV("  link state notification skipped (interface not active)\n");
+    return;
+  }
+
+  // Reset notification state to send link state update
+  ncm_interface.notification_xmit_state = NOTIFICATION_CONNECTED;
+
+  // Trigger notification transmission
+  notification_xmit(rhport, false);
+}
+
 //-----------------------------------------------------------------------------
 //
 // all the netd_*() stuff (interface TinyUSB -> driver)
@@ -774,6 +802,12 @@ void netd_init(void) {
   for (int i = 0; i < RECV_NTB_N; ++i) {
     ncm_interface.recv_free_ntb[i] = &ncm_epbuf.recv[i].ntb;
   }
+  // Default link state - can be configured via CFG_TUD_NCM_DEFAULT_LINK_UP
+  #ifdef CFG_TUD_NCM_DEFAULT_LINK_UP
+  ncm_interface.link_is_up = CFG_TUD_NCM_DEFAULT_LINK_UP;
+  #else
+  ncm_interface.link_is_up = true; // Default to link up if not set.
+  #endif
 } // netd_init
 
 /**
diff --git a/src/class/net/net_device.h b/src/class/net/net_device.h
index 4c9a92f2d..fff2623b7 100644
--- a/src/class/net/net_device.h
+++ b/src/class/net/net_device.h
@@ -87,6 +87,11 @@ void tud_network_init_cb(void);
 // TODO removed later since it is not part of tinyusb stack
 extern uint8_t tud_network_mac_address[6];
 
+//------------- NCM -------------//
+
+// Set the network link state (up/down) and notify the host
+void tud_network_link_state(uint8_t rhport, bool is_up);
+
 //--------------------------------------------------------------------+
 // INTERNAL USBD-CLASS DRIVER API
 //--------------------------------------------------------------------+
diff --git a/src/class/vendor/vendor_device.c b/src/class/vendor/vendor_device.c
index 2fc0ac944..7f1fd8c41 100644
--- a/src/class/vendor/vendor_device.c
+++ b/src/class/vendor/vendor_device.c
@@ -196,8 +196,8 @@ void vendord_reset(uint8_t rhport) {
 
 uint16_t vendord_open(uint8_t rhport, const tusb_desc_interface_t* desc_itf, uint16_t max_len) {
   TU_VERIFY(TUSB_CLASS_VENDOR_SPECIFIC == desc_itf->bInterfaceClass, 0);
+  const uint8_t* desc_end = (const uint8_t*)desc_itf + max_len;
   const uint8_t* p_desc = tu_desc_next(desc_itf);
-  const uint8_t* desc_end = (uint8_t const*)desc_itf + max_len;
 
   // Find available interface
   vendord_interface_t* p_vendor = NULL;
@@ -210,26 +210,26 @@ uint16_t vendord_open(uint8_t rhport, const tusb_desc_interface_t* desc_itf, uin
   TU_VERIFY(p_vendor, 0);
 
   p_vendor->itf_num = desc_itf->bInterfaceNumber;
-  uint8_t found_ep = 0;
-  while (found_ep < desc_itf->bNumEndpoints) {
-    // skip non-endpoint descriptors
-    while ( (TUSB_DESC_ENDPOINT != tu_desc_type(p_desc)) && (p_desc < desc_end) ) {
-      p_desc = tu_desc_next(p_desc);
-    }
-    if (p_desc >= desc_end) {
-      break;
-    }
-
-    const tusb_desc_endpoint_t* desc_ep = (const tusb_desc_endpoint_t*) p_desc;
-    TU_ASSERT(usbd_edpt_open(rhport, desc_ep));
-    found_ep++;
+  while (tu_desc_is_valid(p_desc, desc_end)) {
+    const uint8_t desc_type = tu_desc_type(p_desc);
+    if (desc_type == TUSB_DESC_INTERFACE || desc_type == TUSB_DESC_INTERFACE_ASSOCIATION) {
+      break; // end of this interface
+    } else if (desc_type == TUSB_DESC_ENDPOINT) {
+      const tusb_desc_endpoint_t* desc_ep = (const tusb_desc_endpoint_t*) p_desc;
+      TU_ASSERT(usbd_edpt_open(rhport, desc_ep));
 
-    if (tu_edpt_dir(desc_ep->bEndpointAddress) == TUSB_DIR_IN) {
-      tu_edpt_stream_open(&p_vendor->tx.stream, desc_ep);
-      tud_vendor_n_write_flush((uint8_t)(p_vendor - _vendord_itf));
-    } else {
-      tu_edpt_stream_open(&p_vendor->rx.stream, desc_ep);
-      TU_ASSERT(tu_edpt_stream_read_xfer(rhport, &p_vendor->rx.stream) > 0, 0); // prepare for incoming data
+      // open endpoint stream, skip if already opened
+      if (tu_edpt_dir(desc_ep->bEndpointAddress) == TUSB_DIR_IN) {
+        if (p_vendor->tx.stream.ep_addr == 0) {
+          tu_edpt_stream_open(&p_vendor->tx.stream, desc_ep);
+          tud_vendor_n_write_flush((uint8_t)(p_vendor - _vendord_itf));
+        }
+      } else {
+        if (p_vendor->rx.stream.ep_addr == 0) {
+          tu_edpt_stream_open(&p_vendor->rx.stream, desc_ep);
+          TU_ASSERT(tu_edpt_stream_read_xfer(rhport, &p_vendor->rx.stream) > 0, 0); // prepare for incoming data
+        }
+      }
     }
 
     p_desc = tu_desc_next(p_desc);
diff --git a/src/common/tusb_mcu.h b/src/common/tusb_mcu.h
index a0175d664..6678265b5 100644
--- a/src/common/tusb_mcu.h
+++ b/src/common/tusb_mcu.h
@@ -369,6 +369,10 @@
   #define TUP_DCD_ENDPOINT_MAX    7 // only 5 TX FIFO for endpoint IN
   #define CFG_TUSB_OS_INC_PATH_DEFAULT   freertos/
 
+  #if CFG_TUSB_MCU == OPT_MCU_ESP32S3
+    #define TUP_MCU_MULTIPLE_CORE 1
+  #endif
+
   // Disable slave if DMA is enabled
   #define CFG_TUD_DWC2_SLAVE_ENABLE_DEFAULT  !CFG_TUD_DWC2_DMA_ENABLE
   #define CFG_TUH_DWC2_SLAVE_ENABLE_DEFAULT  !CFG_TUH_DWC2_DMA_ENABLE
@@ -381,6 +385,8 @@
 
   #define CFG_TUSB_OS_INC_PATH_DEFAULT   freertos/
 
+  #define TUP_MCU_MULTIPLE_CORE   1
+
   // Disable slave if DMA is enabled
   #define CFG_TUD_DWC2_SLAVE_ENABLE_DEFAULT  !CFG_TUD_DWC2_DMA_ENABLE
   #define CFG_TUH_DWC2_SLAVE_ENABLE_DEFAULT  !CFG_TUH_DWC2_DMA_ENABLE
@@ -410,6 +416,7 @@
 #elif TU_CHECK_MCU(OPT_MCU_RP2040)
   #define TUP_DCD_EDPT_ISO_ALLOC
   #define TUP_DCD_ENDPOINT_MAX    16
+  #define TUP_MCU_MULTIPLE_CORE   1
 
   #define TU_ATTR_FAST_FUNC       __attribute__((section(".time_critical.tinyusb")))
 
diff --git a/src/common/tusb_types.h b/src/common/tusb_types.h
index e000a4bd3..fd7f01b67 100644
--- a/src/common/tusb_types.h
+++ b/src/common/tusb_types.h
@@ -586,6 +586,12 @@ TU_ATTR_ALWAYS_INLINE static inline uint8_t tu_desc_subtype(void const* desc) {
   return ((uint8_t const*) desc)[DESC_OFFSET_SUBTYPE];
 }
 
+TU_ATTR_ALWAYS_INLINE static inline uint8_t tu_desc_is_valid(void const* desc, uint8_t const* desc_end) {
+  const uint8_t* desc8 = (uint8_t const*) desc;
+  return (desc8 < desc_end) && (tu_desc_next(desc) <= desc_end);
+}
+
+
 // find descriptor that match byte1 (type)
 uint8_t const * tu_desc_find(uint8_t const* desc, uint8_t const* end, uint8_t byte1);
 
diff --git a/src/device/usbd.c b/src/device/usbd.c
index 9c381d5e0..6e5fcf3b6 100644
--- a/src/device/usbd.c
+++ b/src/device/usbd.c
@@ -340,15 +340,16 @@ TU_ATTR_ALWAYS_INLINE static inline usbd_class_driver_t const * get_driver(uint8
 enum { RHPORT_INVALID = 0xFFu };
 tu_static uint8_t _usbd_rhport = RHPORT_INVALID;
 
-// Event queue
-// usbd_int_set() is used as mutex in OS NONE config
+static OSAL_SPINLOCK_DEF(_usbd_spin, usbd_int_set);
+
+// Event queue: usbd_int_set() is used as mutex in OS NONE config
 OSAL_QUEUE_DEF(usbd_int_set, _usbd_qdef, CFG_TUD_TASK_QUEUE_SZ, dcd_event_t);
-tu_static osal_queue_t _usbd_q;
+static osal_queue_t _usbd_q;
 
 // Mutex for claiming endpoint
 #if OSAL_MUTEX_REQUIRED
-  tu_static osal_mutex_def_t _ubsd_mutexdef;
-  tu_static osal_mutex_t _usbd_mutex;
+  static osal_mutex_def_t _ubsd_mutexdef;
+  static osal_mutex_t _usbd_mutex;
 #else
   #define _usbd_mutex   NULL
 #endif
@@ -466,7 +467,7 @@ bool tud_rhport_init(uint8_t rhport, const tusb_rhport_init_t* rh_init) {
   TU_ASSERT(rh_init);
 
   TU_LOG_USBD("USBD init on controller %u, speed = %s\r\n", rhport,
-    rh_init->speed == TUSB_SPEED_HIGH ? "High" : "Full");
+              rh_init->speed == TUSB_SPEED_HIGH ? "High" : "Full");
   TU_LOG_INT(CFG_TUD_LOG_LEVEL, sizeof(usbd_device_t));
   TU_LOG_INT(CFG_TUD_LOG_LEVEL, sizeof(dcd_event_t));
   TU_LOG_INT(CFG_TUD_LOG_LEVEL, sizeof(tu_fifo_t));
@@ -475,6 +476,8 @@ bool tud_rhport_init(uint8_t rhport, const tusb_rhport_init_t* rh_init) {
   tu_varclr(&_usbd_dev);
   _usbd_queued_setup = 0;
 
+  osal_spin_init(&_usbd_spin);
+
 #if OSAL_MUTEX_REQUIRED
   // Init device mutex
   _usbd_mutex = osal_mutex_create(&_ubsd_mutexdef);
@@ -1242,17 +1245,21 @@ TU_ATTR_FAST_FUNC void dcd_event_handler(dcd_event_t const* event, bool in_isr)
 // USBD API For Class Driver
 //--------------------------------------------------------------------+
 
-void usbd_int_set(bool enabled)
-{
-  if (enabled)
-  {
+void usbd_int_set(bool enabled) {
+  if (enabled) {
     dcd_int_enable(_usbd_rhport);
-  }else
-  {
+  } else {
     dcd_int_disable(_usbd_rhport);
   }
 }
 
+void usbd_spin_lock(bool in_isr) {
+  osal_spin_lock(&_usbd_spin, in_isr);
+}
+void usbd_spin_unlock(bool in_isr) {
+  osal_spin_unlock(&_usbd_spin, in_isr);
+}
+
 // Parse consecutive endpoint descriptors (IN & OUT)
 bool usbd_open_edpt_pair(uint8_t rhport, uint8_t const* p_desc, uint8_t ep_count, uint8_t xfer_type, uint8_t* ep_out, uint8_t* ep_in)
 {
diff --git a/src/device/usbd_pvt.h b/src/device/usbd_pvt.h
index 190d6fd7f..5c6f9dbee 100644
--- a/src/device/usbd_pvt.h
+++ b/src/device/usbd_pvt.h
@@ -68,6 +68,8 @@ usbd_class_driver_t const* usbd_app_driver_get_cb(uint8_t* driver_count) TU_ATTR
 typedef bool (*usbd_control_xfer_cb_t)(uint8_t rhport, uint8_t stage, tusb_control_request_t const * request);
 
 void usbd_int_set(bool enabled);
+void usbd_spin_lock(bool in_isr);
+void usbd_spin_unlock(bool in_isr);
 
 //--------------------------------------------------------------------+
 // USBD Endpoint API
diff --git a/src/host/usbh.c b/src/host/usbh.c
index b7d5a05f2..f2e5c1f0e 100644
--- a/src/host/usbh.c
+++ b/src/host/usbh.c
@@ -147,6 +147,9 @@ static osal_mutex_t _usbh_mutex;
 #define _usbh_mutex   NULL
 #endif
 
+// Spinlock for interrupt handler
+static OSAL_SPINLOCK_DEF(_usbh_spin, usbh_int_set);
+
 // Event queue: usbh_int_set() is used as mutex in OS NONE config
 OSAL_QUEUE_DEF(usbh_int_set, _usbh_qdef, CFG_TUH_TASK_QUEUE_SZ, hcd_event_t);
 static osal_queue_t _usbh_q;
@@ -424,6 +427,8 @@ bool tuh_rhport_init(uint8_t rhport, const tusb_rhport_init_t* rh_init) {
     TU_LOG_INT_USBH(sizeof(tu_fifo_t));
     TU_LOG_INT_USBH(sizeof(tu_edpt_stream_t));
 
+    osal_spin_init(&_usbh_spin);
+
     // Event queue
     _usbh_q = osal_queue_create(&_usbh_qdef);
     TU_ASSERT(_usbh_q != NULL);
@@ -895,6 +900,14 @@ void usbh_int_set(bool enabled) {
   }
 }
 
+void usbh_spin_lock(bool in_isr) {
+  osal_spin_lock(&_usbh_spin, in_isr);
+}
+
+void usbh_spin_unlock(bool in_isr) {
+  osal_spin_unlock(&_usbh_spin, in_isr);
+}
+
 void usbh_defer_func(osal_task_func_t func, void *param, bool in_isr) {
   hcd_event_t event = { 0 };
   event.event_id = USBH_EVENT_FUNC_CALL;
diff --git a/src/host/usbh.h b/src/host/usbh.h
index 6f34d8bb3..13eede869 100644
--- a/src/host/usbh.h
+++ b/src/host/usbh.h
@@ -42,7 +42,7 @@
 //--------------------------------------------------------------------+
 
 // Endpoint Bulk size depending on host mx speed
-#define TUH_EPSIZE_BULK_MPS   (TUD_OPT_HIGH_SPEED ? TUSB_EPSIZE_BULK_HS : TUSB_EPSIZE_BULK_FS)
+#define TUH_EPSIZE_BULK_MPS   (TUH_OPT_HIGH_SPEED ? TUSB_EPSIZE_BULK_HS : TUSB_EPSIZE_BULK_FS)
 
 // forward declaration
 struct tuh_xfer_s;
diff --git a/src/host/usbh_pvt.h b/src/host/usbh_pvt.h
index 61b012493..cb092e5f3 100644
--- a/src/host/usbh_pvt.h
+++ b/src/host/usbh_pvt.h
@@ -71,6 +71,9 @@ void usbh_int_set(bool enabled);
 
 void usbh_defer_func(osal_task_func_t func, void *param, bool in_isr);
 
+void usbh_spin_lock(bool in_isr);
+void usbh_spin_unlock(bool in_isr);
+
 //--------------------------------------------------------------------+
 // USBH Endpoint API
 //--------------------------------------------------------------------+
diff --git a/src/osal/osal.h b/src/osal/osal.h
index 38d45da44..a33280425 100644
--- a/src/osal/osal.h
+++ b/src/osal/osal.h
@@ -75,6 +75,10 @@ typedef void (*osal_task_func_t)( void * );
 // OSAL Porting API
 // Should be implemented as static inline function in osal_port.h header
 /*
+   void osal_spin_init(osal_spinlock_t *ctx);
+   void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr)
+   void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr);
+
    osal_semaphore_t osal_semaphore_create(osal_semaphore_def_t* semdef);
    bool osal_semaphore_delete(osal_semaphore_t semd_hdl);
    bool osal_semaphore_post(osal_semaphore_t sem_hdl, bool in_isr);
diff --git a/src/osal/osal_freertos.h b/src/osal/osal_freertos.h
index a3a0f3a3f..bde5ec010 100644
--- a/src/osal/osal_freertos.h
+++ b/src/osal/osal_freertos.h
@@ -42,20 +42,20 @@ extern "C" {
 //--------------------------------------------------------------------+
 
 #if configSUPPORT_STATIC_ALLOCATION
-  typedef StaticSemaphore_t osal_semaphore_def_t;
-  typedef StaticSemaphore_t osal_mutex_def_t;
+typedef StaticSemaphore_t osal_semaphore_def_t;
+typedef StaticSemaphore_t osal_mutex_def_t;
 #else
-  // not used therefore defined to smallest possible type to save space
-  typedef uint8_t osal_semaphore_def_t;
-  typedef uint8_t osal_mutex_def_t;
+
+// not used therefore defined to the smallest possible type to save space
+typedef uint8_t osal_semaphore_def_t;
+typedef uint8_t osal_mutex_def_t;
 #endif
 
 typedef SemaphoreHandle_t osal_semaphore_t;
 typedef SemaphoreHandle_t osal_mutex_t;
 typedef QueueHandle_t osal_queue_t;
 
-typedef struct
-{
+typedef struct {
   uint16_t depth;
   uint16_t item_sz;
   void*    buf;
@@ -83,16 +83,14 @@ typedef struct
 //--------------------------------------------------------------------+
 // TASK API
 //--------------------------------------------------------------------+
-
 TU_ATTR_ALWAYS_INLINE static inline uint32_t _osal_ms2tick(uint32_t msec) {
-  if ( msec == OSAL_TIMEOUT_WAIT_FOREVER ) return portMAX_DELAY;
-  if ( msec == 0 ) return 0;
+  if (msec == OSAL_TIMEOUT_WAIT_FOREVER) { return portMAX_DELAY; }
+  if (msec == 0) { return 0; }
 
   uint32_t ticks = pdMS_TO_TICKS(msec);
 
-  // configTICK_RATE_HZ is less than 1000 and 1 tick > 1 ms
-  // we still need to delay at least 1 tick
-  if ( ticks == 0 ) ticks = 1;
+  // If configTICK_RATE_HZ is less than 1000 and 1 tick > 1 ms, we still need to delay at least 1 tick
+  if (ticks == 0) { ticks = 1; }
 
   return ticks;
 }
@@ -102,9 +100,70 @@ TU_ATTR_ALWAYS_INLINE static inline void osal_task_delay(uint32_t msec) {
 }
 
 //--------------------------------------------------------------------+
-// Semaphore API
+// Spinlock API
 //--------------------------------------------------------------------+
+#define OSAL_SPINLOCK_DEF(_name, _int_set) \
+  osal_spinlock_t _name
+
+#if TUSB_MCU_VENDOR_ESPRESSIF
+// Espressif critical take spinlock as argument and does not use in_isr
+typedef portMUX_TYPE osal_spinlock_t;
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+  spinlock_initialize(ctx);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  portENTER_CRITICAL(ctx);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  portEXIT_CRITICAL(ctx);
+}
+
+#else
+
+typedef UBaseType_t osal_spinlock_t;
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+  (void) ctx;
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  if (in_isr) {
+    if (!TUP_MCU_MULTIPLE_CORE) {
+      (void) ctx;
+      return; // single core MCU does not need to lock in ISR
+    }
+    *ctx = taskENTER_CRITICAL_FROM_ISR();
+  } else {
+    taskENTER_CRITICAL();
+  }
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  if (in_isr) {
+    if (!TUP_MCU_MULTIPLE_CORE) {
+      (void) ctx;
+      return; // single core MCU does not need to lock in ISR
+    }
+    taskEXIT_CRITICAL_FROM_ISR(*ctx);
+  } else {
+    taskEXIT_CRITICAL();
+  }
+}
 
+#endif
+
+//--------------------------------------------------------------------+
+// Semaphore API
+//--------------------------------------------------------------------+
 TU_ATTR_ALWAYS_INLINE static inline osal_semaphore_t osal_semaphore_create(osal_semaphore_def_t *semdef) {
 #if configSUPPORT_STATIC_ALLOCATION
   return xSemaphoreCreateBinaryStatic(semdef);
@@ -120,19 +179,12 @@ TU_ATTR_ALWAYS_INLINE static inline bool osal_semaphore_delete(osal_semaphore_t
 }
 
 TU_ATTR_ALWAYS_INLINE static inline bool osal_semaphore_post(osal_semaphore_t sem_hdl, bool in_isr) {
-  if ( !in_isr ) {
+  if (!in_isr) {
     return xSemaphoreGive(sem_hdl) != 0;
   } else {
     BaseType_t xHigherPriorityTaskWoken = pdFALSE;
     BaseType_t res = xSemaphoreGiveFromISR(sem_hdl, &xHigherPriorityTaskWoken);
-
-#if CFG_TUSB_MCU == OPT_MCU_ESP32S2 || CFG_TUSB_MCU == OPT_MCU_ESP32S3
-    // not needed after https://github.com/espressif/esp-idf/commit/c5fd79547ac9b7bae06fa660e9f814d18d3390b7
-    if ( xHigherPriorityTaskWoken ) portYIELD_FROM_ISR();
-#else
     portYIELD_FROM_ISR(xHigherPriorityTaskWoken);
-#endif
-
     return res != 0;
   }
 }
@@ -148,7 +200,6 @@ TU_ATTR_ALWAYS_INLINE static inline void osal_semaphore_reset(osal_semaphore_t c
 //--------------------------------------------------------------------+
 // MUTEX API (priority inheritance)
 //--------------------------------------------------------------------+
-
 TU_ATTR_ALWAYS_INLINE static inline osal_mutex_t osal_mutex_create(osal_mutex_def_t *mdef) {
 #if configSUPPORT_STATIC_ALLOCATION
   return xSemaphoreCreateMutexStatic(mdef);
@@ -174,7 +225,6 @@ TU_ATTR_ALWAYS_INLINE static inline bool osal_mutex_unlock(osal_mutex_t mutex_hd
 //--------------------------------------------------------------------+
 // QUEUE API
 //--------------------------------------------------------------------+
-
 TU_ATTR_ALWAYS_INLINE static inline osal_queue_t osal_queue_create(osal_queue_def_t* qdef) {
   osal_queue_t q;
 
@@ -201,19 +251,12 @@ TU_ATTR_ALWAYS_INLINE static inline bool osal_queue_receive(osal_queue_t qhdl, v
 }
 
 TU_ATTR_ALWAYS_INLINE static inline bool osal_queue_send(osal_queue_t qhdl, void const *data, bool in_isr) {
-  if ( !in_isr ) {
+  if (!in_isr) {
     return xQueueSendToBack(qhdl, data, OSAL_TIMEOUT_WAIT_FOREVER) != 0;
   } else {
     BaseType_t xHigherPriorityTaskWoken = pdFALSE;
     BaseType_t res = xQueueSendToBackFromISR(qhdl, data, &xHigherPriorityTaskWoken);
-
-#if CFG_TUSB_MCU == OPT_MCU_ESP32S2 || CFG_TUSB_MCU == OPT_MCU_ESP32S3
-    // not needed after https://github.com/espressif/esp-idf/commit/c5fd79547ac9b7bae06fa660e9f814d18d3390b7 (IDF v5)
-    if ( xHigherPriorityTaskWoken ) portYIELD_FROM_ISR();
-#else
     portYIELD_FROM_ISR(xHigherPriorityTaskWoken);
-#endif
-
     return res != 0;
   }
 }
diff --git a/src/osal/osal_mynewt.h b/src/osal/osal_mynewt.h
index 16def0d2a..ee95e684f 100644
--- a/src/osal/osal_mynewt.h
+++ b/src/osal/osal_mynewt.h
@@ -41,6 +41,32 @@ TU_ATTR_ALWAYS_INLINE static inline void osal_task_delay(uint32_t msec) {
 }
 
 //--------------------------------------------------------------------+
+// Spinlock API
+//--------------------------------------------------------------------+
+typedef os_sr_t osal_spinlock_t;
+
+#define OSAL_SPINLOCK_DEF(_name, _int_set) \
+  osal_spinlock_t _name
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+ (void) ctx;
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  OS_ENTER_CRITICAL(*ctx);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  OS_ENTER_CRITICAL(*ctx);
+}
+
+//--------------------------------------------------------------------+
 // Semaphore API
 //--------------------------------------------------------------------+
 typedef struct os_sem  osal_semaphore_def_t;
diff --git a/src/osal/osal_none.h b/src/osal/osal_none.h
index 40e9bb83a..a8eb1042b 100644
--- a/src/osal/osal_none.h
+++ b/src/osal/osal_none.h
@@ -41,6 +41,33 @@ TU_ATTR_WEAK void osal_task_delay(uint32_t msec);
 #endif
 
 //--------------------------------------------------------------------+
+// Spinlock API
+//--------------------------------------------------------------------+
+typedef struct {
+  void (* interrupt_set)(bool);
+} osal_spinlock_t;
+
+// For SMP, spinlock must be locked by hardware, cannot just use interrupt
+#define OSAL_SPINLOCK_DEF(_name, _int_set) \
+  osal_spinlock_t _name = { .interrupt_set = _int_set }
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+  (void) ctx;
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!in_isr) {
+    ctx->interrupt_set(false);
+  }
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!in_isr) {
+    ctx->interrupt_set(true);
+  }
+}
+
+//--------------------------------------------------------------------+
 // Binary Semaphore API
 //--------------------------------------------------------------------+
 typedef struct {
diff --git a/src/osal/osal_pico.h b/src/osal/osal_pico.h
index 315de0950..ace5907d7 100644
--- a/src/osal/osal_pico.h
+++ b/src/osal/osal_pico.h
@@ -44,6 +44,27 @@ TU_ATTR_ALWAYS_INLINE static inline void osal_task_delay(uint32_t msec) {
 }
 
 //--------------------------------------------------------------------+
+// Spinlock API
+//--------------------------------------------------------------------+
+typedef critical_section_t osal_spinlock_t; // pico implement critical section with spinlock
+#define OSAL_SPINLOCK_DEF(_name, _int_set) \
+  osal_spinlock_t _name
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+  critical_section_init(ctx);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  (void) in_isr;
+  critical_section_enter_blocking(ctx);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  (void) in_isr;
+  critical_section_exit(ctx);
+}
+
+//--------------------------------------------------------------------+
 // Binary Semaphore API
 //--------------------------------------------------------------------+
 typedef struct semaphore osal_semaphore_def_t, * osal_semaphore_t;
diff --git a/src/osal/osal_rtthread.h b/src/osal/osal_rtthread.h
index c27814835..a778f5425 100644
--- a/src/osal/osal_rtthread.h
+++ b/src/osal/osal_rtthread.h
@@ -43,6 +43,32 @@ TU_ATTR_ALWAYS_INLINE static inline void osal_task_delay(uint32_t msec) {
 }
 
 //--------------------------------------------------------------------+
+// Spinlock API
+//--------------------------------------------------------------------+
+typedef struct rt_spinlock osal_spinlock_t;
+
+#define OSAL_SPINLOCK_DEF(_name, _int_set) \
+  osal_spinlock_t _name
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+  rt_spin_lock_init(ctx);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  rt_spin_lock(ctx);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  rt_spin_unlock(ctx);
+}
+
+//--------------------------------------------------------------------+
 // Semaphore API
 //--------------------------------------------------------------------+
 typedef struct rt_semaphore osal_semaphore_def_t;
diff --git a/src/osal/osal_rtx4.h b/src/osal/osal_rtx4.h
index 35909e4d6..35860ddd5 100644
--- a/src/osal/osal_rtx4.h
+++ b/src/osal/osal_rtx4.h
@@ -57,6 +57,25 @@ TU_ATTR_ALWAYS_INLINE static inline uint16_t msec2wait(uint32_t msec) {
 }
 
 //--------------------------------------------------------------------+
+// Spinlock API, stub not implemented
+//--------------------------------------------------------------------+
+typedef uint8_t osal_spinlock_t;
+#define OSAL_SPINLOCK_DEF(_name, _int_set) \
+  osal_spinlock_t _name
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+  (void) ctx;
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  (void) ctx; (void) in_isr;
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  (void) ctx; (void) in_isr;
+}
+
+//--------------------------------------------------------------------+
 // Semaphore API
 //--------------------------------------------------------------------+
 typedef OS_SEM osal_semaphore_def_t;
diff --git a/src/osal/osal_zephyr.h b/src/osal/osal_zephyr.h
index 8ecb13c6d..91f225f79 100644
--- a/src/osal/osal_zephyr.h
+++ b/src/osal/osal_zephyr.h
@@ -36,6 +36,35 @@ TU_ATTR_ALWAYS_INLINE static inline void osal_task_delay(uint32_t msec) {
 }
 
 //--------------------------------------------------------------------+
+// Spinlock API
+//--------------------------------------------------------------------+
+typedef struct {
+  struct k_spinlock lock;
+  k_spinlock_key_t key;
+} osal_spinlock_t;
+
+#define OSAL_SPINLOCK_DEF(_name, _int_set) \
+  osal_spinlock_t _name
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_init(osal_spinlock_t *ctx) {
+  (void) ctx;
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_lock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  ctx->key = k_spin_lock(&ctx->lock);
+}
+
+TU_ATTR_ALWAYS_INLINE static inline void osal_spin_unlock(osal_spinlock_t *ctx, bool in_isr) {
+  if (!TUP_MCU_MULTIPLE_CORE && in_isr) {
+    return; // single core MCU does not need to lock in ISR
+  }
+  k_spin_unlock(&ctx->lock, ctx->key);
+}
+
+//--------------------------------------------------------------------+
 // Binary Semaphore API
 //--------------------------------------------------------------------+
 typedef struct k_sem osal_semaphore_def_t, * osal_semaphore_t;
diff --git a/src/portable/analog/max3421/hcd_max3421.c b/src/portable/analog/max3421/hcd_max3421.c
index bb33200f2..971dbd62e 100644
--- a/src/portable/analog/max3421/hcd_max3421.c
+++ b/src/portable/analog/max3421/hcd_max3421.c
@@ -28,9 +28,9 @@
 
 #if CFG_TUH_ENABLED && defined(CFG_TUH_MAX3421) && CFG_TUH_MAX3421
 
-#include <stdatomic.h>
 #include "host/hcd.h"
 #include "host/usbh.h"
+#include "host/usbh_pvt.h"
 
 //--------------------------------------------------------------------+
 //
@@ -233,7 +233,7 @@ typedef struct {
     uint8_t hxfr;
   }sndfifo_owner;
 
-  atomic_flag busy; // busy transferring
+  bool busy_lock; // busy transferring
 
 #if OSAL_MUTEX_REQUIRED
   OSAL_MUTEX_DEF(spi_mutexdef);
@@ -327,7 +327,9 @@ TU_ATTR_ALWAYS_INLINE static inline void mode_write(uint8_t rhport, uint8_t data
 }
 
 TU_ATTR_ALWAYS_INLINE static inline void peraddr_write(uint8_t rhport, uint8_t data, bool in_isr) {
-  if ( _hcd_data.peraddr == data ) return; // no need to change address
+  if (_hcd_data.peraddr == data) {
+    return; // no need to change address
+  }
 
   _hcd_data.peraddr = data;
   reg_write(rhport, PERADDR_ADDR, data, in_isr);
@@ -373,7 +375,7 @@ TU_ATTR_ALWAYS_INLINE static inline void hwfifo_setup(uint8_t rhport, const uint
 
 static void hwfifo_receive(uint8_t rhport, uint8_t * buffer, uint16_t len, bool in_isr) {
   uint8_t hirq;
-  uint8_t const reg = RCVVFIFO_ADDR;
+  const uint8_t reg = RCVVFIFO_ADDR;
 
   max3421_spi_lock(rhport, in_isr);
 
@@ -389,7 +391,7 @@ static void hwfifo_receive(uint8_t rhport, uint8_t * buffer, uint16_t len, bool
 //--------------------------------------------------------------------+
 
 static max3421_ep_t* find_ep_not_addr0(uint8_t daddr, uint8_t ep_num, uint8_t ep_dir) {
-  uint8_t const is_out = 1-ep_dir;
+  const uint8_t is_out = 1-ep_dir;
   for(size_t i=1; i<CFG_TUH_MAX3421_ENDPOINT_TOTAL; i++) {
     max3421_ep_t* ep = &_hcd_data.ep[i];
     // control endpoint is bi-direction (skip check)
@@ -727,8 +729,8 @@ static void xact_generic(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool
 
 // Submit a transfer, when complete hcd_event_xfer_complete() must be invoked
 bool hcd_edpt_xfer(uint8_t rhport, uint8_t daddr, uint8_t ep_addr, uint8_t * buffer, uint16_t buflen) {
-  uint8_t const ep_num = tu_edpt_number(ep_addr);
-  uint8_t const ep_dir = (uint8_t) tu_edpt_dir(ep_addr);
+  const uint8_t ep_num = tu_edpt_number(ep_addr);
+  const uint8_t ep_dir = (uint8_t) tu_edpt_dir(ep_addr);
   max3421_ep_t* ep = find_opened_ep(daddr, ep_num, ep_dir);
   TU_VERIFY(ep);
 
@@ -744,8 +746,17 @@ bool hcd_edpt_xfer(uint8_t rhport, uint8_t daddr, uint8_t ep_addr, uint8_t * buf
   ep->xferred_len = 0;
   ep->state = EP_STATE_ATTEMPT_1;
 
+  bool has_xfer = false;
+
+  usbh_spin_lock(false);
+  if (!_hcd_data.busy_lock) {
+    _hcd_data.busy_lock = true;
+    has_xfer = true;
+  }
+  usbh_spin_unlock(false);
+
   // carry out transfer if not busy
-  if (!atomic_flag_test_and_set(&_hcd_data.busy)) {
+  if (has_xfer) {
     xact_generic(rhport, ep, true, false);
   }
 
@@ -781,8 +792,17 @@ bool hcd_setup_send(uint8_t rhport, uint8_t daddr, uint8_t const setup_packet[8]
   ep->xferred_len = 0;
   ep->state = EP_STATE_ATTEMPT_1;
 
+  bool has_xfer = false;
+
+  usbh_spin_lock(false);
+  if (!_hcd_data.busy_lock) {
+    _hcd_data.busy_lock = true;
+    has_xfer = true;
+  }
+  usbh_spin_unlock(false);
+
   // carry out transfer if not busy
-  if (!atomic_flag_test_and_set(&_hcd_data.busy)) {
+  if (has_xfer) {
     xact_setup(rhport, ep, false);
   }
 
@@ -848,8 +868,8 @@ static void handle_connect_irq(uint8_t rhport, bool in_isr) {
 }
 
 static void xfer_complete_isr(uint8_t rhport, max3421_ep_t *ep, xfer_result_t result, uint8_t hrsl, bool in_isr) {
-  uint8_t const ep_dir = 1-ep->hxfr_bm.is_out;
-  uint8_t const ep_addr = tu_edpt_addr(ep->hxfr_bm.ep_num, ep_dir);
+  const uint8_t ep_dir = 1 - ep->hxfr_bm.is_out;
+  const uint8_t ep_addr = tu_edpt_addr(ep->hxfr_bm.ep_num, ep_dir);
 
   // save data toggle
   if (ep_dir) {
@@ -867,7 +887,9 @@ static void xfer_complete_isr(uint8_t rhport, max3421_ep_t *ep, xfer_result_t re
     xact_generic(rhport, next_ep, true, in_isr);
   }else {
     // no more pending
-    atomic_flag_clear(&_hcd_data.busy);
+    usbh_spin_lock(in_isr);
+    _hcd_data.busy_lock = false;
+    usbh_spin_unlock(in_isr);
   }
 }
 
@@ -906,7 +928,9 @@ static void handle_xfer_done(uint8_t rhport, bool in_isr) {
         xact_generic(rhport, next_ep, true, in_isr);
       } else {
         // no more pending in this frame -> clear busy
-        atomic_flag_clear(&_hcd_data.busy);
+        usbh_spin_lock(in_isr);
+        _hcd_data.busy_lock = false;
+        usbh_spin_unlock(in_isr);
       }
       return;
 
@@ -997,8 +1021,8 @@ void print_hirq(uint8_t hirq) {
 // Interrupt handler
 void hcd_int_handler(uint8_t rhport, bool in_isr) {
   uint8_t hirq = reg_read(rhport, HIRQ_ADDR, in_isr) & _hcd_data.hien;
-  if (!hirq) return;
-//  print_hirq(hirq);
+  if (!hirq) { return; }
+  //  print_hirq(hirq);
 
   if (hirq & HIRQ_FRAME_IRQ) {
     _hcd_data.frame_count++;
@@ -1017,8 +1041,19 @@ void hcd_int_handler(uint8_t rhport, bool in_isr) {
     }
 
     // start usb transfer if not busy
-    if (ep_retry != NULL && !atomic_flag_test_and_set(&_hcd_data.busy)) {
-      xact_generic(rhport, ep_retry, true, in_isr);
+    if (ep_retry != NULL) {
+      bool has_xfer = false;
+
+      usbh_spin_lock(in_isr);
+      if (!_hcd_data.busy_lock) {
+        _hcd_data.busy_lock = true;
+        has_xfer = true;
+      }
+      usbh_spin_unlock(in_isr);
+
+      if (has_xfer) {
+        xact_generic(rhport, ep_retry, true, in_isr);
+      }
     }
   }
 
diff --git a/src/portable/chipidea/ci_hs/dcd_ci_hs.c b/src/portable/chipidea/ci_hs/dcd_ci_hs.c
index a716dc24c..244f5a2d4 100644
--- a/src/portable/chipidea/ci_hs/dcd_ci_hs.c
+++ b/src/portable/chipidea/ci_hs/dcd_ci_hs.c
@@ -239,7 +239,11 @@ bool dcd_init(uint8_t rhport, const tusb_rhport_init_t* rh_init) {
   usbmode |= USBMODE_CM_DEVICE;
   dcd_reg->USBMODE = usbmode;
 
+#ifdef CFG_TUD_CI_HS_VBUS_CHARGE
+  dcd_reg->OTGSC = OTGSC_VBUS_CHARGE | OTGSC_OTG_TERMINATION;
+#else
   dcd_reg->OTGSC = OTGSC_VBUS_DISCHARGE | OTGSC_OTG_TERMINATION;
+#endif
 
 #if !TUD_OPT_HIGH_SPEED
   dcd_reg->PORTSC1 = PORTSC1_FORCE_FULL_SPEED;
diff --git a/src/portable/synopsys/dwc2/dcd_dwc2.c b/src/portable/synopsys/dwc2/dcd_dwc2.c
index 52d675611..865c51894 100644
--- a/src/portable/synopsys/dwc2/dcd_dwc2.c
+++ b/src/portable/synopsys/dwc2/dcd_dwc2.c
@@ -39,6 +39,7 @@
 #define DWC2_DEBUG    2
 
 #include "device/dcd.h"
+#include "device/usbd_pvt.h"
 #include "dwc2_common.h"
 
 //--------------------------------------------------------------------+
@@ -52,6 +53,7 @@ typedef struct {
   uint8_t interval;
 } xfer_ctl_t;
 
+// This variable is modified from ISR context, so it must be protected by critical section
 static xfer_ctl_t xfer_status[DWC2_EP_MAX][2];
 #define XFER_CTL_BASE(_ep, _dir) (&xfer_status[_ep][_dir])
 
@@ -321,6 +323,9 @@ static void edpt_disable(uint8_t rhport, uint8_t ep_addr, bool stall) {
   }
 }
 
+// Since this function returns void, it is not possible to return a boolean success message
+// We must make sure that this function is not called when the EP is disabled
+// Must be called from critical section
 static void edpt_schedule_packets(uint8_t rhport, const uint8_t epnum, const uint8_t dir) {
   dwc2_regs_t* dwc2 = DWC2_REG(rhport);
   xfer_ctl_t* const xfer = XFER_CTL_BASE(epnum, dir);
@@ -531,6 +536,8 @@ void dcd_edpt_close_all(uint8_t rhport) {
   dwc2_regs_t* dwc2 = DWC2_REG(rhport);
   uint8_t const ep_count = _dwc2_controller[rhport].ep_count;
 
+  usbd_spin_lock(false);
+
   _dcd_data.allocated_epin_count = 0;
 
   // Disable non-control interrupt
@@ -548,8 +555,9 @@ void dcd_edpt_close_all(uint8_t rhport) {
 
   dfifo_flush_tx(dwc2, 0x10); // all tx fifo
   dfifo_flush_rx(dwc2);
-
   dfifo_device_init(rhport); // re-init dfifo
+
+  usbd_spin_unlock(false);
 }
 
 bool dcd_edpt_iso_alloc(uint8_t rhport, uint8_t ep_addr, uint16_t largest_packet_size) {
@@ -567,21 +575,31 @@ bool dcd_edpt_iso_activate(uint8_t rhport,  tusb_desc_endpoint_t const * p_endpo
 bool dcd_edpt_xfer(uint8_t rhport, uint8_t ep_addr, uint8_t* buffer, uint16_t total_bytes) {
   uint8_t const epnum = tu_edpt_number(ep_addr);
   uint8_t const dir = tu_edpt_dir(ep_addr);
-
   xfer_ctl_t* xfer = XFER_CTL_BASE(epnum, dir);
-  xfer->buffer = buffer;
-  xfer->ff = NULL;
-  xfer->total_len = total_bytes;
+  bool ret;
 
-  // EP0 can only handle one packet
-  if (epnum == 0) {
-    _dcd_data.ep0_pending[dir] = total_bytes;
+  usbd_spin_lock(false);
+
+  if (xfer->max_size == 0) {
+    ret = false;  // Endpoint is closed
+  } else {
+    xfer->buffer = buffer;
+    xfer->ff = NULL;
+    xfer->total_len = total_bytes;
+
+    // EP0 can only handle one packet
+    if (epnum == 0) {
+      _dcd_data.ep0_pending[dir] = total_bytes;
+    }
+
+    // Schedule packets to be sent within interrupt
+    edpt_schedule_packets(rhport, epnum, dir);
+    ret = true;
   }
 
-  // Schedule packets to be sent within interrupt
-  edpt_schedule_packets(rhport, epnum, dir);
+  usbd_spin_unlock(false);
 
-  return true;
+  return ret;
 }
 
 // The number of bytes has to be given explicitly to allow more flexible control of how many
@@ -594,17 +612,27 @@ bool dcd_edpt_xfer_fifo(uint8_t rhport, uint8_t ep_addr, tu_fifo_t* ff, uint16_t
 
   uint8_t const epnum = tu_edpt_number(ep_addr);
   uint8_t const dir = tu_edpt_dir(ep_addr);
-
   xfer_ctl_t* xfer = XFER_CTL_BASE(epnum, dir);
-  xfer->buffer = NULL;
-  xfer->ff = ff;
-  xfer->total_len = total_bytes;
+  bool ret;
 
-  // Schedule packets to be sent within interrupt
-  // TODO xfer fifo may only available for slave mode
-  edpt_schedule_packets(rhport, epnum, dir);
+  usbd_spin_lock(false);
 
-  return true;
+  if (xfer->max_size == 0) {
+    ret = false;  // Endpoint is closed
+  } else {
+    xfer->buffer = NULL;
+    xfer->ff = ff;
+    xfer->total_len = total_bytes;
+
+    // Schedule packets to be sent within interrupt
+    // TODO xfer fifo may only available for slave mode
+    edpt_schedule_packets(rhport, epnum, dir);
+    ret = true;
+  }
+
+  usbd_spin_unlock(false);
+
+  return ret;
 }
 
 void dcd_edpt_stall(uint8_t rhport, uint8_t ep_addr) {
@@ -631,6 +659,7 @@ void dcd_edpt_clear_stall(uint8_t rhport, uint8_t ep_addr) {
 //--------------------------------------------------------------------
 
 // 7.4.1 Initialization on USB Reset
+// Must be called from critical section
 static void handle_bus_reset(uint8_t rhport) {
   dwc2_regs_t *dwc2 = DWC2_REG(rhport);
   const uint8_t ep_count =  dwc2_ep_count(dwc2);
@@ -983,14 +1012,16 @@ static void handle_ep_irq(uint8_t rhport, uint8_t dir) {
  */
 void dcd_int_handler(uint8_t rhport) {
   dwc2_regs_t* dwc2 = DWC2_REG(rhport);
-
   const uint32_t gintmask = dwc2->gintmsk;
   const uint32_t gintsts = dwc2->gintsts & gintmask;
 
   if (gintsts & GINTSTS_USBRST) {
     // USBRST is start of reset.
     dwc2->gintsts = GINTSTS_USBRST;
+
+    usbd_spin_lock(true);
     handle_bus_reset(rhport);
+    usbd_spin_unlock(true);
   }
 
   if (gintsts & GINTSTS_ENUMDNE) {
diff --git a/src/portable/synopsys/dwc2/dwc2_esp32.h b/src/portable/synopsys/dwc2/dwc2_esp32.h
index 3309760ff..49b8c54cb 100644
--- a/src/portable/synopsys/dwc2/dwc2_esp32.h
+++ b/src/portable/synopsys/dwc2/dwc2_esp32.h
@@ -55,8 +55,8 @@ static const dwc2_controller_t _dwc2_controller[] = {
 // On ESP32 for consistency we associate
 // - Port0 to OTG_FS, and Port1 to OTG_HS
 static const dwc2_controller_t _dwc2_controller[] = {
-{ .reg_base = DWC2_FS_REG_BASE, .irqnum = ETS_USB_OTG11_CH0_INTR_SOURCE, .ep_count = 7, .ep_in_count = 5, .ep_fifo_size = 1024 },
-{ .reg_base = DWC2_HS_REG_BASE, .irqnum = ETS_USB_OTG_INTR_SOURCE, .ep_count = 16, .ep_in_count = 8, .ep_fifo_size = 4096 }
+  { .reg_base = DWC2_FS_REG_BASE, .irqnum = ETS_USB_OTG11_CH0_INTR_SOURCE, .ep_count = 7, .ep_in_count = 5, .ep_fifo_size = 1024 },
+  { .reg_base = DWC2_HS_REG_BASE, .irqnum = ETS_USB_OTG_INTR_SOURCE, .ep_count = 16, .ep_in_count = 8, .ep_fifo_size = 4096 }
 };
 #endif
 
diff --git a/src/tusb_option.h b/src/tusb_option.h
index 98f1a91b5..104f669c9 100644
--- a/src/tusb_option.h
+++ b/src/tusb_option.h
@@ -195,6 +195,7 @@
 
 // Analog Devices
 #define OPT_MCU_MAX32690         2400  ///< ADI MAX32690
+#define OPT_MCU_MAX32665         2401  ///< ADI MAX32666/5
 #define OPT_MCU_MAX32666         2401  ///< ADI MAX32666/5
 #define OPT_MCU_MAX32650         2402  ///< ADI MAX32650/1/2
 #define OPT_MCU_MAX78002         2403  ///< ADI MAX78002
@@ -267,6 +268,15 @@
   #define CFG_TUD_DWC2_DMA_ENABLE CFG_TUD_DWC2_DMA_ENABLE_DEFAULT
 #endif
 
+// Enable CI_HS VBUS Charge. Set this to 1 if the USB_VBUS pin is not connected to 5V VBUS (note: 3.3V is insufficient).
+#ifndef CFG_TUD_CI_HS_VBUS_CHARGE
+  #ifndef CFG_TUD_CI_HS_VBUS_CHARGE_DEFAULT
+  #define CFG_TUD_CI_HS_VBUS_CHARGE_DEFAULT 0
+  #endif
+
+  #define CFG_TUD_CI_HS_VBUS_CHARGE CFG_TUD_CI_HS_VBUS_CHARGE_DEFAULT
+#endif
+
 // Enable DWC2 Slave mode for host
 #ifndef CFG_TUH_DWC2_SLAVE_ENABLE
   #ifndef CFG_TUH_DWC2_SLAVE_ENABLE_DEFAULT