diff --git a/Inc/gex_gateway.h b/Inc/gex_gateway.h
index b6c96dd..85e5564 100644
--- a/Inc/gex_gateway.h
+++ b/Inc/gex_gateway.h
@@ -7,5 +7,6 @@
 
 #include "main.h"
 void gw_process(void);
+void gw_handle_usb_out(uint8_t *buffer);
 
 #endif //GEX_NRF_GEX_GATEWAY_H
diff --git a/Src/gex_gateway.c b/Src/gex_gateway.c
index 4bb1267..c9aa5e0 100644
--- a/Src/gex_gateway.c
+++ b/Src/gex_gateway.c
@@ -42,74 +42,79 @@ void start_slave_cmd(uint8_t slave_addr, uint16_t frame_len, uint8_t cksum)
     msg4slave_cksum = cksum;
 }
 
-/** called from the main loop, periodically */
-void gw_process(void)
+void gw_handle_usb_out(uint8_t *buffer)
 {
-    static uint8_t buffer[MQ_SLOT_LEN];
-    while (mq_read(&usb_rxq, buffer)) { // greedy - handle as many as possible
-        dbg("Handling frame.");
-        if (urx_state == URXS_IDLE) {
-            PayloadParser pp = pp_start(buffer, MQ_SLOT_LEN, NULL);
-
-            // handle binary commands for the gateway
-
-            // magic twice, one inverted - denotes a gateway command
-            uint16_t magic1 = pp_u8(&pp);
-            uint16_t magic2 = pp_u8(&pp);
-            if (magic1 == MAGIC_GW_COMMAND && magic2 == (0xFFU & (~MAGIC_GW_COMMAND))) {
-                // third byte is the command code
-                switch (pp_u8(&pp)) {
-                    case CMD_GET_ID:
-                        respond_gw_id();
+    dbg("Handling frame.");
+    if (urx_state == URXS_IDLE) {
+        PayloadParser pp = pp_start(buffer, MQ_SLOT_LEN, NULL);
+
+        // handle binary commands for the gateway
+
+        // magic twice, one inverted - denotes a gateway command
+        uint16_t magic1 = pp_u8(&pp);
+        uint16_t magic2 = pp_u8(&pp);
+        if (magic1 == MAGIC_GW_COMMAND && magic2 == (0xFFU & (~MAGIC_GW_COMMAND))) {
+            // third byte is the command code
+            switch (pp_u8(&pp)) {
+                case CMD_GET_ID:
+                    respond_gw_id();
+                    break;
+
+                case CMD_MSG4SLAVE:;
+                    uint8_t slave_addr = pp_u8(&pp);
+                    uint16_t frame_len = pp_u16(&pp);
+                    uint8_t cksum = pp_u8(&pp);
+
+                    if (frame_len == 0 || frame_len > MAX_FRAME_LEN) {
+                        dbg("Frame too big!");
                         break;
+                    }
 
-                    case CMD_MSG4SLAVE:;
-                        uint8_t slave_addr = pp_u8(&pp);
-                        uint16_t frame_len = pp_u16(&pp);
-                        uint8_t cksum = pp_u8(&pp);
-
-                        if (frame_len == 0 || frame_len > MAX_FRAME_LEN) {
-                            dbg("Frame too big!");
-                            break;
-                        }
-
-                        start_slave_cmd(slave_addr, frame_len, cksum);
-                        dbg("Collecting frame for slave %02x: %d bytes", (int)slave_addr, (int)frame_len);
-                        urx_state = URXS_MSG4SLAVE;
-                        break;
+                    start_slave_cmd(slave_addr, frame_len, cksum);
+                    dbg("Collecting frame for slave %02x: %d bytes", (int)slave_addr, (int)frame_len);
+                    urx_state = URXS_MSG4SLAVE;
+                    break;
 
-                    default:
-                        dbg("Bad cmd");
-                }
-            } else {
-                // Bad frame??
-                dbg("Bad USB frame, starts %x,%x", buffer[0],buffer[1]);
+                default:
+                    dbg("Bad cmd");
             }
+        } else {
+            // Bad frame??
+            dbg("Bad USB frame, starts %x,%x", buffer[0],buffer[1]);
         }
-        else if (urx_state == URXS_MSG4SLAVE) {
-            uint32_t wanted = MIN(msg4slave_len-msg4slave_already, MQ_SLOT_LEN);
-            memcpy(&msg4slave[msg4slave_already], buffer, wanted);
-            msg4slave_already += wanted;
-
-            if (wanted < MQ_SLOT_LEN) {
-                // this was the end
-                uint8_t ck = 0;
-                for (int i = 0; i < msg4slave_len; i++) {
-                    ck ^= msg4slave[i];
-                }
-                ck = ~ck;
-
-                if (ck != msg4slave_cksum) {
-                    dbg("Checksum mismatch!");
-                }
-                else {
-                    dbg("Verified, sending a %d B frame to slave.", (int) msg4slave_len);
-                    // TODO send to slave
-                }
-
-                urx_state = URXS_IDLE;
+    }
+    else if (urx_state == URXS_MSG4SLAVE) {
+        uint32_t wanted = MIN(msg4slave_len-msg4slave_already, MQ_SLOT_LEN);
+        memcpy(&msg4slave[msg4slave_already], buffer, wanted);
+        msg4slave_already += wanted;
+
+        if (wanted < MQ_SLOT_LEN) {
+            // this was the end
+            uint8_t ck = 0;
+            for (int i = 0; i < msg4slave_len; i++) {
+                ck ^= msg4slave[i];
+            }
+            ck = ~ck;
+
+            if (ck != msg4slave_cksum) {
+                dbg("Checksum mismatch!");
             }
+            else {
+                dbg("Verified, sending a %d B frame to slave.", (int) msg4slave_len);
+                // TODO send to slave
+            }
+
+            urx_state = URXS_IDLE;
         }
     }
 }
 
+/** called from the main loop, periodically */
+void gw_process(void)
+{
+//    static uint8_t buffer[MQ_SLOT_LEN];
+//    while (mq_read(&usb_rxq, buffer)) { // greedy - handle as many as possible
+//        gw_handle_usb_out(buffer);
+//    }
+}
+
diff --git a/Src/usbd_cdc_if.c b/Src/usbd_cdc_if.c
index bde0dfa..a5b6603 100644
--- a/Src/usbd_cdc_if.c
+++ b/Src/usbd_cdc_if.c
@@ -50,6 +50,7 @@
 /* Includes ------------------------------------------------------------------*/
 #include <debug.h>
 #include <msg_queue.h>
+#include <gex_gateway.h>
 #include "usbd_cdc_if.h"
 
 /* USER CODE BEGIN INCLUDE */
@@ -298,10 +299,15 @@ static int8_t CDC_Receive_FS(uint8_t* Buf, uint32_t *Len)
 
   // queue the frame
   dbg("USB rx frame");
-  bool suc = mq_post(&usb_rxq, Buf, *Len);
-  if (!suc) {
-    dbg("USB rxq overrun!");
-  }
+  gw_handle_usb_out(Buf); // this expects full 64byte frames always starting with magic or useful data
+
+  // handled immediately in the interrupt to block USB (NAK)
+
+//
+//  bool suc = mq_post(&usb_rxq, Buf, *Len);
+//  if (!suc) {
+//    dbg("USB rxq overrun!");
+//  }
 
   return (USBD_OK);
   /* USER CODE END 6 */
@@ -336,6 +342,8 @@ uint8_t CDC_Transmit_FS(uint8_t* Buf, uint16_t Len)
 
 /* USER CODE END PRIVATE_FUNCTIONS_IMPLEMENTATION */
 
+// TODO add Tx Done notify
+
 /**
   * @}
   */