work around possible RP2040 erratum

RP2040 device controller does not seem to clear pending transactions
configured in EP0 buffer controls when the host aborts a control
transfer. This causes assertion failures, including when a buffer
AVAILABLE flag set for a previous transfer causes an unexpected
transaction completion.

src/portable/raspberrypi/rp2040/dcd_rp2040.c

@@ -217,14 +217,33 @@ static void __tusb_irq_path_func(hw_handle_buff_status)(void)
 
 TU_ATTR_ALWAYS_INLINE static inline void reset_ep0_pid(void)
 {
+    // Abort any transactions from a prior control transfer (HW erratum?)
+    usb_hw_set->abort = 0x3;
+    while (usb_hw->abort_done & 0x3 != 0x3)
+        ;
     // If we have finished this transfer on EP0 set pid back to 1 for next
     // setup transfer. Also clear a stall in case
     uint8_t addrs[] = {0x0, 0x80};
     for (uint i = 0 ; i < TU_ARRAY_SIZE(addrs); i++)
     {
         struct hw_endpoint *ep = hw_endpoint_get_by_addr(addrs[i]);
         ep->next_pid = 1u;
+        // Reset the buffer control now to minimize race conditions
+        _hw_endpoint_buffer_control_set_value32(ep, USB_BUF_CTRL_DATA1_PID | USB_BUF_CTRL_SEL);
+        // Explicit delay, because the one in
+        // _hw_endpoint_buffer_control_set_value32 is only to set AVAILABLE
+        __asm volatile (
+            "b 1f\n"
+            "1: b 1f\n"
+            "1: b 1f\n"
+            "1: b 1f\n"
+            "1: b 1f\n"
+            "1: b 1f\n"
+            "1:\n"
+            : : : "memory");
+        hw_endpoint_reset_transfer(ep);
     }
+    usb_hw_clear->abort = 0x3;
 }
 
 static void __tusb_irq_path_func(reset_non_control_endpoints)(void)