diff --git a/src/class/msc/msc_device.c b/src/class/msc/msc_device.c
index c7c926f4e..e3fe5a078 100644
--- a/src/class/msc/msc_device.c
+++ b/src/class/msc/msc_device.c
@@ -78,10 +78,10 @@ typedef struct {
   uint8_t sense_key;
   uint8_t add_sense_code;
   uint8_t add_sense_qualifier;
-#if CFG_TUD_MSC_ASYNC_IO
+
+  // Async IO
   uint8_t next_op;
   uint32_t xferred_bytes;
-#endif
 }mscd_interface_t;
 
 static mscd_interface_t _mscd_itf;
@@ -100,9 +100,7 @@ static void proc_write10_cmd(mscd_interface_t* p_msc);
 static void proc_write10_new_data(mscd_interface_t* p_msc, uint32_t xferred_bytes);
 static void proc_write10_next(mscd_interface_t* p_msc, uint32_t xferred_bytes, int32_t nbytes);
 static bool proc_stage_status(mscd_interface_t* p_msc);
-#if CFG_TUD_MSC_ASYNC_IO
 static void tud_msc_async_io_done_cb(void* bytes_processed);
-#endif
 
 TU_ATTR_ALWAYS_INLINE static inline bool is_data_in(uint8_t dir) {
   return tu_bit_test(dir, 7);
@@ -265,7 +263,6 @@ static inline void set_sense_medium_not_present(uint8_t lun) {
   tud_msc_set_sense(lun, SCSI_SENSE_NOT_READY, 0x3A, 0x00);
 }
 
-#if CFG_TUD_MSC_ASYNC_IO
 void tud_msc_async_io_done(int32_t bytes_processed) {
   // Precheck to avoid queueing multiple RW done callback
   TU_VERIFY(_mscd_itf.next_op != MSC_NEXT_OP_NONE,);
@@ -289,7 +286,6 @@ static void tud_msc_async_io_done_cb(void* bytes_processed) {
     }
   }
 }
-#endif
 
 //--------------------------------------------------------------------+
 // USBD Driver API
@@ -821,13 +817,12 @@ static void proc_read10_cmd(mscd_interface_t* p_msc) {
   // Application can consume smaller bytes
   uint32_t const offset = p_msc->xferred_len % block_sz;
 
-#if CFG_TUD_MSC_ASYNC_IO
   p_msc->next_op = MSC_NEXT_OP_READ10;
-  tud_msc_read10_cb(p_cbw->lun, lba, offset, _mscd_epbuf.buf, (uint32_t)nbytes);
-#else
   nbytes = tud_msc_read10_cb(p_cbw->lun, lba, offset, _mscd_epbuf.buf, (uint32_t)nbytes);
-  proc_read10_next(p_msc, nbytes);
-#endif
+  if (nbytes != TUD_MSC_RET_ASYNC) {
+    p_msc->next_op = MSC_NEXT_OP_NONE;
+    proc_read10_next(p_msc, nbytes);
+  }
 }
 
 static void proc_read10_next(mscd_interface_t* p_msc, int32_t nbytes) {
@@ -885,14 +880,13 @@ static void proc_write10_new_data(mscd_interface_t* p_msc, uint32_t xferred_byte
   // Invoke callback to consume new data
   uint32_t const offset = p_msc->xferred_len % block_sz;
 
-#if CFG_TUD_MSC_ASYNC_IO
   p_msc->next_op = MSC_NEXT_OP_WRITE10;
   p_msc->xferred_bytes = xferred_bytes;
-  tud_msc_write10_cb(p_cbw->lun, lba, offset, _mscd_epbuf.buf, xferred_bytes);
-#else
   int32_t nbytes =  tud_msc_write10_cb(p_cbw->lun, lba, offset, _mscd_epbuf.buf, xferred_bytes);
-  proc_write10_next(p_msc, xferred_bytes, nbytes);
-#endif
+  if (nbytes != TUD_MSC_RET_ASYNC) {
+    p_msc->next_op = MSC_NEXT_OP_NONE;
+    proc_write10_next(p_msc, xferred_bytes, nbytes);
+  }
 }
 
 static void proc_write10_next(mscd_interface_t* p_msc, uint32_t xferred_bytes, int32_t nbytes) {
diff --git a/src/class/msc/msc_device.h b/src/class/msc/msc_device.h
index 407fe241c..7162b11e4 100644
--- a/src/class/msc/msc_device.h
+++ b/src/class/msc/msc_device.h
@@ -48,10 +48,11 @@
   #error CFG_TUD_MSC_EP_BUFSIZE must be defined, value of a block size should work well, the more the better
 #endif
 
-// Enable asynchronous read/write, once operation is finished tud_msc_async_io_done() must be called
-#ifndef CFG_TUD_MSC_ASYNC_IO
-#define CFG_TUD_MSC_ASYNC_IO 0
-#endif
+// Return value of callback functions
+// Error
+#define TUD_MSC_RET_ERROR  -1
+// Asynchronous IO
+#define TUD_MSC_RET_ASYNC  -16
 
 TU_VERIFY_STATIC(CFG_TUD_MSC_EP_BUFSIZE < UINT16_MAX, "Size is not correct");
 
@@ -62,6 +63,11 @@ TU_VERIFY_STATIC(CFG_TUD_MSC_EP_BUFSIZE < UINT16_MAX, "Size is not correct");
 // Set SCSI sense response
 bool tud_msc_set_sense(uint8_t lun, uint8_t sense_key, uint8_t add_sense_code, uint8_t add_sense_qualifier);
 
+// Called once asynchronous read/write operation is done
+// bytes_processed has the same meaning of tud_msc_read10_cb() /
+// tud_msc_write10_cb() return value
+void tud_msc_async_io_done(int32_t bytes_processed);
+
 //--------------------------------------------------------------------+
 // Application Callbacks (WEAK is optional)
 //--------------------------------------------------------------------+
@@ -70,34 +76,40 @@ bool tud_msc_set_sense(uint8_t lun, uint8_t sense_key, uint8_t add_sense_code, u
 // - Address = lba * BLOCK_SIZE + offset
 //   - offset is only needed if CFG_TUD_MSC_EP_BUFSIZE is smaller than BLOCK_SIZE.
 //
-// - Application fill the buffer (up to bufsize) with address contents and return number of read byte. If
-//   - read < bufsize : These bytes are transferred first and callback invoked again for remaining data.
+// - Application fill the buffer (up to bufsize) with address contents and return number of bytes read or status.
 //
-//   - read == 0      : Indicate application is not ready yet e.g disk I/O busy.
-//                      Callback invoked again with the same parameters later on.
+//   - ret < bufsize  : These bytes are transferred first and callback will be invoked again for remaining data.
 //
-//   - read < 0       : Indicate application error e.g invalid address. This request will be STALLed
+//   - ret == 0       : Indicate application is not ready yet e.g disk I/O busy.
+//                      Callback will be invoked again with the same parameters later on.
+//
+//   - ret == TUD_MSC_RET_ERROR (-1)
+//                    : Indicate application error e.g invalid address. This request will be STALLed
 //                      and return failed status in command status wrapper phase.
 //
-// - In case of asynchronous IO enabled, application should passing reading parameters to background IO
-//   task and return immediately. Once reading is done, tud_msc_async_io_done() must be called.
+//   - ret == TUD_MSC_RET_ASYNC (-16)
+//                    : Data reading will be done asynchronously in a background task. Application should return immediately.
+//                      tud_msc_async_io_done() must be called once reading is done to signal completion.
 int32_t tud_msc_read10_cb (uint8_t lun, uint32_t lba, uint32_t offset, void* buffer, uint32_t bufsize);
 
 // Invoked when received SCSI WRITE10 command
 // - Address = lba * BLOCK_SIZE + offset
 //   - offset is only needed if CFG_TUD_MSC_EP_BUFSIZE is smaller than BLOCK_SIZE.
 //
-// - Application write data from buffer to address contents (up to bufsize) and return number of written byte. If
-//   - write < bufsize : callback invoked again with remaining data later on.
+// - Application writes data from buffer to address contents (up to bufsize) and returns the number of bytes written or status.
 //
-//   - write == 0      : Indicate application is not ready yet e.g disk I/O busy.
-//                       Callback invoked again with the same parameters later on.
+//   - ret < bufsize  : Callback will be invoked again with remaining data later on.
 //
-//   - write < 0       : Indicate application error e.g invalid address. This request will be STALLed
-//                       and return failed status in command status wrapper phase.
+//   - ret == 0       : Indicate application is not ready yet e.g disk I/O busy.
+//                      Callback will be invoked again with the same parameters later on.
 //
-// - In case of asynchronous IO enabled, application should passing writing parameters to background IO
-//   task and return immediately. Once writing is done, tud_msc_async_io_done() must be called.
+//   - ret == TUD_MSC_RET_ERROR (-1)
+//                    : Indicate application error e.g invalid address. This request will be STALLed
+//                      and return failed status in command status wrapper phase.
+//
+//   - ret == TUD_MSC_RET_ASYNC (-16)
+//                    : Data writing will be done asynchronously in a background task. Application should return immediately.
+//                      tud_msc_async_io_done() must be called once writing is done to signal completion.
 // TODO change buffer to const uint8_t*
 int32_t tud_msc_write10_cb (uint8_t lun, uint32_t lba, uint32_t offset, uint8_t* buffer, uint32_t bufsize);
 
@@ -131,12 +143,6 @@ void tud_msc_capacity_cb(uint8_t lun, uint32_t* block_count, uint16_t* block_siz
  */
 int32_t tud_msc_scsi_cb (uint8_t lun, uint8_t const scsi_cmd[16], void* buffer, uint16_t bufsize);
 
-#if CFG_TUD_MSC_ASYNC_IO
-// Called once asynchronous read/write operation is done
-// bytes_processed has the same meaning of tud_msc_read10_cb() /
-// tud_msc_write10_cb() return value
-void tud_msc_async_io_done(int32_t bytes_processed);
-#endif
 /*------------- Optional callbacks -------------*/
 
 // Invoked when received GET_MAX_LUN request, required for multiple LUNs implementation