force single buffered for device mode, out endpoint

src/device/usbd.c

@@ -1254,14 +1254,13 @@ bool usbd_edpt_xfer(uint8_t rhport, uint8_t ep_addr, uint8_t * buffer, uint16_t
 
   if ( dcd_edpt_xfer(rhport, ep_addr, buffer, total_bytes) )
   {
-    TU_LOG2("OK\r\n");
     return true;
   }else
   {
     // DCD error, mark endpoint as ready to allow next transfer
     _usbd_dev.ep_status[epnum][dir].busy = false;
     _usbd_dev.ep_status[epnum][dir].claimed = 0;
-    TU_LOG2("failed\r\n");
+    TU_LOG2("FAILED\r\n");
     TU_BREAKPOINT();
     return false;
   }

src/portable/raspberrypi/rp2040/dcd_rp2040.c

@@ -145,6 +145,7 @@ static void _hw_endpoint_init(struct hw_endpoint *ep, uint8_t ep_addr, uint16_t
 
         // Now if it hasn't already been done
         //alloc a buffer and fill in endpoint control register
+        // TODO device may change configuration (dynamic), should clear and reallocate
         if(!(ep->configured))
         {
             _hw_endpoint_alloc(ep);
@@ -194,9 +195,6 @@ static void hw_handle_buff_status(void)
     {
         if (remaining_buffers & bit)
         {
-            uint __unused which = (usb_hw->buf_cpu_should_handle & bit) ? 1 : 0;
-            // Should be single buffered
-            assert(which == 0);
             // clear this in advance
             usb_hw_clear->buf_status = bit;
             // IN transfer for even i, OUT transfer for odd i
@@ -463,7 +461,7 @@ void dcd_edpt0_status_complete(uint8_t rhport, tusb_control_request_t const * re
 
 bool dcd_edpt_open (uint8_t rhport, tusb_desc_endpoint_t const * desc_edpt)
 {
-    pico_info("dcd_edpt_open %d %02x\n", rhport, desc_edpt->bEndpointAddress);
+    pico_info("dcd_edpt_open %02x\n", desc_edpt->bEndpointAddress);
     assert(rhport == 0);
     hw_endpoint_init(desc_edpt->bEndpointAddress, desc_edpt->wMaxPacketSize.size, desc_edpt->bmAttributes.xfer);
     return true;
@@ -478,14 +476,14 @@ bool dcd_edpt_xfer(uint8_t rhport, uint8_t ep_addr, uint8_t * buffer, uint16_t t
 
 void dcd_edpt_stall (uint8_t rhport, uint8_t ep_addr)
 {
-    pico_trace("dcd_edpt_stall %d %02x\n", rhport, ep_addr);
+    pico_trace("dcd_edpt_stall %02x\n", ep_addr);
     assert(rhport == 0);
     hw_endpoint_stall(ep_addr);
 }
 
 void dcd_edpt_clear_stall (uint8_t rhport, uint8_t ep_addr)
 {
-    pico_trace("dcd_edpt_clear_stall %d %02x\n", rhport, ep_addr);
+    pico_trace("dcd_edpt_clear_stall %02x\n", ep_addr);
     assert(rhport == 0);
     hw_endpoint_clear_stall(ep_addr);
 }
@@ -493,8 +491,11 @@ void dcd_edpt_clear_stall (uint8_t rhport, uint8_t ep_addr)
 
 void dcd_edpt_close (uint8_t rhport, uint8_t ep_addr)
 {
+    (void) rhport;
+    (void) ep_addr;
+
     // usbd.c says: In progress transfers on this EP may be delivered after this call
-    pico_trace("dcd_edpt_close %d %02x\n", rhport, ep_addr);
+    pico_trace("dcd_edpt_close %02x\n", ep_addr);
 }
 
 void dcd_int_handler(uint8_t rhport)

src/portable/raspberrypi/rp2040/rp2040_usb.c

@@ -148,10 +148,15 @@ static void _hw_endpoint_start_next_buffer(struct hw_endpoint *ep)
 {
   uint32_t ep_ctrl = *ep->endpoint_control;
 
-  // always compute buffer 0
-  uint32_t buf_ctrl = prepare_ep_buffer(ep, 0);
+  // always compute and start with buffer 0
+  uint32_t buf_ctrl = prepare_ep_buffer(ep, 0) | USB_BUF_CTRL_SEL;
 
-  if(ep->remaining_len)
+  // For now: skip double buffered for Device mode, OUT endpoint since
+  // host could send < 64 bytes and cause short packet on buffer0
+  // NOTE this could happen to Host mode IN endpoint
+  bool const force_single = !(usb_hw->main_ctrl & USB_MAIN_CTRL_HOST_NDEVICE_BITS) && !tu_edpt_dir(ep->ep_addr);
+
+  if(ep->remaining_len && !force_single)
   {
     // Use buffer 1 (double buffered) if there is still data
     // TODO: Isochronous for buffer1 bit-field is different than CBI (control bulk, interrupt)
@@ -181,8 +186,7 @@ static void _hw_endpoint_start_next_buffer(struct hw_endpoint *ep)
 void hw_endpoint_xfer_start(struct hw_endpoint *ep, uint8_t *buffer, uint16_t total_len)
 {
   _hw_endpoint_lock_update(ep, 1);
-  pico_trace("Start transfer of total len %d on ep %d %s\n", total_len, tu_edpt_number(ep->ep_addr),
-             ep_dir_string[tu_edpt_dir(ep->ep_addr)]);
+
   if ( ep->active )
   {
     // TODO: Is this acceptable for interrupt packets?
@@ -251,10 +255,42 @@ static void _hw_endpoint_xfer_sync (struct hw_endpoint *ep)
   // always sync buffer 0
   uint16_t buf0_bytes = sync_ep_buffer(ep, 0);
 
-  // sync buffer 1 if double buffered and buffer 0 is not short packet
-  if ( ((*ep->endpoint_control) & EP_CTRL_DOUBLE_BUFFERED_BITS) && (buf0_bytes == ep->wMaxPacketSize) )
+  // sync buffer 1 if double buffered
+  if ( (*ep->endpoint_control) & EP_CTRL_DOUBLE_BUFFERED_BITS )
   {
-    sync_ep_buffer(ep, 1);
+    if (buf0_bytes == ep->wMaxPacketSize)
+    {
+      // sync buffer 1 if not short packet
+      sync_ep_buffer(ep, 1);
+    }else
+    {
+      // short packet on buffer 0
+      // TODO couldn't figure out how to handle this case which happen with net_lwip_webserver example
+      // At this time (currently trigger per 2 buffer), the buffer1 is probably filled with data from
+      // the next transfer (not current one). For now we disable double buffered for device OUT
+      // NOTE this could happen to Host IN
+#if 0
+      uint8_t const ep_num = tu_edpt_number(ep->ep_addr);
+      uint8_t const dir =  (uint8_t) tu_edpt_dir(ep->ep_addr);
+      uint8_t const ep_id = 2*ep_num + (dir ? 0 : 1);
+
+      // abort queued transfer on buffer 1
+      usb_hw->abort |= TU_BIT(ep_id);
+
+      while ( !(usb_hw->abort_done & TU_BIT(ep_id)) ) {}
+
+      uint32_t ep_ctrl = *ep->endpoint_control;
+      ep_ctrl &= ~(EP_CTRL_DOUBLE_BUFFERED_BITS | EP_CTRL_INTERRUPT_PER_DOUBLE_BUFFER);
+      ep_ctrl |= EP_CTRL_INTERRUPT_PER_BUFFER;
+
+      _hw_endpoint_buffer_control_set_value32(ep, 0);
+
+      usb_hw->abort &= ~TU_BIT(ep_id);
+
+      TU_LOG(3, "----SHORT PACKET buffer0 on EP %02X:\r\n", ep->ep_addr);
+      print_bufctrl32(buf_ctrl);
+#endif
+    }
   }
 }
 

src/portable/raspberrypi/rp2040/rp2040_usb.h

@@ -120,8 +120,8 @@ static inline void print_bufctrl16(uint32_t u16)
       .u16 = u16
   };
 
-  TU_LOG(3, "len = %u, available = %u, stall = %u, reset = %u, toggle = %u, last = %u, full = %u\r\n",
-         bufctrl.xfer_len, bufctrl.available, bufctrl.stall, bufctrl.reset_bufsel, bufctrl.data_toggle, bufctrl.last_buf, bufctrl.full);
+  TU_LOG(3, "len = %u, available = %u, full = %u, last = %u, stall = %u, reset = %u, toggle = %u\r\n",
+         bufctrl.xfer_len, bufctrl.available, bufctrl.full, bufctrl.last_buf, bufctrl.stall, bufctrl.reset_bufsel, bufctrl.data_toggle);
 }
 
 static inline void print_bufctrl32(uint32_t u32)