diff --git a/src/portable/analog/max3421/hcd_max3421.c b/src/portable/analog/max3421/hcd_max3421.c
index 3eb3f796f..1b182e551 100644
--- a/src/portable/analog/max3421/hcd_max3421.c
+++ b/src/portable/analog/max3421/hcd_max3421.c
@@ -184,17 +184,21 @@ enum {
 typedef struct {
   uint8_t daddr;
 
-  struct TU_ATTR_PACKED {
-    uint8_t state         : 4;
-    uint8_t is_setup      : 1; // also bit 4 in HXFR reg (smaller code, no shift necessary)
-    uint8_t ep_dir        : 1; // also bit 5 in HXFR reg (smaller code, no shift necessary)
-    uint8_t is_iso        : 1; // also bit 6 in HXFR reg (smaller code, no shift necessary)
-    uint8_t data_toggle   : 1;
+  union { ;
+    struct TU_ATTR_PACKED {
+      uint8_t ep_num   : 4;
+      uint8_t is_setup : 1;
+      uint8_t is_out   : 1;
+      uint8_t is_iso   : 1;
+    }hxfr_bm;
+
+    uint8_t hxfr;
   };
 
   struct TU_ATTR_PACKED {
-    uint8_t ep_num : 4;
-    uint16_t packet_size : 12;
+    uint8_t state        : 4;
+    uint8_t data_toggle  : 1;
+    uint16_t packet_size : 11;
   };
 
   uint16_t total_len;
@@ -370,10 +374,11 @@ TU_ATTR_ALWAYS_INLINE static inline void sndbc_write(uint8_t rhport, uint8_t dat
 //--------------------------------------------------------------------+
 
 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;
   for(size_t i=1; i<CFG_TUH_MAX3421_ENDPOINT_TOTAL; i++) {
     max3421_ep_t* ep = &_hcd_data.ep[i];
-    // for control endpoint, skip direction check
-    if (daddr == ep->daddr && ep_num == ep->ep_num && (ep_dir == ep->ep_dir || ep_num == 0)) {
+    // control endpoint is bi-direction (skip check)
+    if (daddr == ep->daddr && ep_num == ep->hxfr_bm.ep_num && (ep_num == 0 || is_out == ep->hxfr_bm.is_out)) {
       return ep;
     }
   }
@@ -577,7 +582,6 @@ void hcd_device_close(uint8_t rhport, uint8_t dev_addr) {
 // Open an endpoint
 bool hcd_edpt_open(uint8_t rhport, uint8_t daddr, tusb_desc_endpoint_t const * ep_desc) {
   (void) rhport;
-  (void) daddr;
 
   uint8_t const ep_num = tu_edpt_number(ep_desc->bEndpointAddress);
   tusb_dir_t const ep_dir = tu_edpt_dir(ep_desc->bEndpointAddress);
@@ -589,12 +593,9 @@ bool hcd_edpt_open(uint8_t rhport, uint8_t daddr, tusb_desc_endpoint_t const * e
     ep = allocate_ep();
     TU_ASSERT(ep);
     ep->daddr = daddr;
-    ep->ep_num = (uint8_t) (ep_num & 0x0f);
-    ep->ep_dir = (ep_dir == TUSB_DIR_IN) ? 1 : 0;
-  }
-
-  if ( TUSB_XFER_ISOCHRONOUS == ep_desc->bmAttributes.xfer ) {
-    ep->is_iso = 1;
+    ep->hxfr_bm.ep_num = (uint8_t) (ep_num & 0x0f);
+    ep->hxfr_bm.is_out = (ep_dir == TUSB_DIR_OUT) ? 1 : 0;
+    ep->hxfr_bm.is_iso = (TUSB_XFER_ISOCHRONOUS == ep_desc->bmAttributes.xfer) ? 1 : 0;
   }
 
   ep->packet_size = (uint16_t) (tu_edpt_packet_size(ep_desc) & 0x7ff);
@@ -602,7 +603,7 @@ bool hcd_edpt_open(uint8_t rhport, uint8_t daddr, tusb_desc_endpoint_t const * e
   return true;
 }
 
-void xact_out(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
+static void xact_out(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
   // Page 12: Programming BULK-OUT Transfers
   // TODO double buffered
   if (switch_ep) {
@@ -618,12 +619,10 @@ void xact_out(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
     fifo_write(rhport, SNDFIFO_ADDR, ep->buf, xact_len, in_isr);
   }
   sndbc_write(rhport, xact_len, in_isr);
-
-  uint8_t const hxfr = (uint8_t ) (ep->ep_num | HXFR_OUT_NIN | (ep->is_iso ? HXFR_ISO : 0));
-  hxfr_write(rhport, hxfr, in_isr);
+  hxfr_write(rhport, ep->hxfr, in_isr);
 }
 
-void xact_in(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
+static void xact_in(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
   // Page 13: Programming BULK-IN Transfers
   if (switch_ep) {
     peraddr_write(rhport, ep->daddr, in_isr);
@@ -632,34 +631,36 @@ void xact_in(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
     reg_write(rhport, HCTL_ADDR, hctl, in_isr);
   }
 
-  uint8_t const hxfr = (uint8_t) (ep->ep_num | (ep->is_iso ? HXFR_ISO : 0));
-  hxfr_write(rhport, hxfr, in_isr);
+  hxfr_write(rhport, ep->hxfr, in_isr);
 }
 
-TU_ATTR_ALWAYS_INLINE static inline
-void xact_inout(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
-  if (ep->ep_num == 0 ) {
+static void xact_setup(uint8_t rhport, max3421_ep_t *ep, bool in_isr) {
+  peraddr_write(rhport, ep->daddr, in_isr);
+  fifo_write(rhport, SUDFIFO_ADDR, ep->buf, 8, in_isr);
+  hxfr_write(rhport, HXFR_SETUP, in_isr);
+}
+
+static void xact_generic(uint8_t rhport, max3421_ep_t *ep, bool switch_ep, bool in_isr) {
+  if (ep->hxfr_bm.ep_num == 0 ) {
     // setup
-    if (ep->is_setup) {
-      peraddr_write(rhport, ep->daddr, in_isr);
-      fifo_write(rhport, SUDFIFO_ADDR, ep->buf, 8, in_isr);
-      hxfr_write(rhport, HXFR_SETUP, in_isr);
+    if (ep->hxfr_bm.is_setup) {
+      xact_setup(rhport, ep, in_isr);
       return;
     }
 
     // status
     if (ep->buf == NULL || ep->total_len == 0) {
-      uint8_t const hxfr = HXFR_HS | (ep->ep_dir ? 0 : HXFR_OUT_NIN);
+      uint8_t const hxfr = HXFR_HS | (ep->hxfr_bm.is_out ? HXFR_OUT_NIN : 0);
       peraddr_write(rhport, ep->daddr, in_isr);
       hxfr_write(rhport, hxfr, in_isr);
       return;
     }
   }
 
-  if (ep->ep_dir) {
-    xact_in(rhport, ep, switch_ep, in_isr);
-  }else {
+  if (ep->hxfr_bm.is_out) {
     xact_out(rhport, ep, switch_ep, in_isr);
+  }else {
+    xact_in(rhport, ep, switch_ep, in_isr);
   }
 }
 
@@ -672,21 +673,21 @@ bool hcd_edpt_xfer(uint8_t rhport, uint8_t daddr, uint8_t ep_addr, uint8_t * buf
   TU_VERIFY(ep);
 
   // control transfer can switch direction
-  ep->ep_dir = ep_dir ? 1u : 0u;
+  ep->hxfr_bm.is_out = ep_dir ? 0u : 1u;
 
   ep->buf = buffer;
   ep->total_len = buflen;
   ep->xferred_len = 0;
   ep->state = EP_STATE_ATTEMPT_1;
 
-  if ( ep_num == 0 ) {
-    ep->is_setup = 0;
+  if (ep_num == 0) {
+    ep->hxfr_bm.is_setup = 0;
     ep->data_toggle = 1;
   }
 
   // carry out transfer if not busy
-  if ( !atomic_flag_test_and_set(&_hcd_data.busy) ) {
-    xact_inout(rhport, ep, true, false);
+  if (!atomic_flag_test_and_set(&_hcd_data.busy)) {
+    xact_generic(rhport, ep, true, false);
   }
 
   return true;
@@ -709,16 +710,16 @@ bool hcd_setup_send(uint8_t rhport, uint8_t daddr, uint8_t const setup_packet[8]
   max3421_ep_t* ep = find_opened_ep(daddr, 0, 0);
   TU_ASSERT(ep);
 
-  ep->ep_dir = 0;
-  ep->is_setup = 1;
+  ep->hxfr_bm.is_out = 1;
+  ep->hxfr_bm.is_setup = 1;
   ep->buf = (uint8_t*)(uintptr_t) setup_packet;
   ep->total_len = 8;
   ep->xferred_len = 0;
   ep->state = EP_STATE_ATTEMPT_1;
 
   // carry out transfer if not busy
-  if ( !atomic_flag_test_and_set(&_hcd_data.busy) ) {
-    xact_inout(rhport, ep, true, false);
+  if (!atomic_flag_test_and_set(&_hcd_data.busy)) {
+    xact_setup(rhport, ep, false);
   }
 
   return true;
@@ -783,10 +784,11 @@ 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_addr = tu_edpt_addr(ep->ep_num, ep->ep_dir);
+  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);
 
   // save data toggle
-  if (ep->ep_dir) {
+  if (ep_dir) {
     ep->data_toggle = (hrsl & HRSL_RCVTOGRD) ? 1u : 0u;
   }else {
     ep->data_toggle = (hrsl & HRSL_SNDTOGRD) ? 1u : 0u;
@@ -798,7 +800,7 @@ static void xfer_complete_isr(uint8_t rhport, max3421_ep_t *ep, xfer_result_t re
   // Find next pending endpoint
   max3421_ep_t * next_ep = find_next_pending_ep(ep);
   if (next_ep) {
-    xact_inout(rhport, next_ep, true, in_isr);
+    xact_generic(rhport, next_ep, true, in_isr);
   }else {
     // no more pending
     atomic_flag_clear(&_hcd_data.busy);
@@ -841,7 +843,7 @@ static void handle_xfer_done(uint8_t rhport, bool in_isr) {
           hxfr_write(rhport, _hcd_data.hxfr, in_isr);
         } else if (next_ep) {
           // switch to next pending endpoint TODO could have issue with double buffered if not clear previously out data
-          xact_inout(rhport, next_ep, true, 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);
@@ -945,7 +947,7 @@ 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_inout(rhport, ep_retry, true, in_isr);
+      xact_generic(rhport, ep_retry, true, in_isr);
     }
   }
 
diff --git a/src/tusb_option.h b/src/tusb_option.h
index 18cc8a82c..f90fb7026 100644
--- a/src/tusb_option.h
+++ b/src/tusb_option.h
@@ -34,7 +34,7 @@
 #define TUSB_VERSION_MAJOR     0
 #define TUSB_VERSION_MINOR     16
 #define TUSB_VERSION_REVISION  0
-#define TUSB_VERSION_BUILD     1
+#define TUSB_VERSION_BUILD     2
 
 #define TUSB_VERSION_NUMBER    (TUSB_VERSION_MAJOR << 24 | TUSB_VERSION_MINOR << 16 | TUSB_VERSION_REVISION << 8 | TUSB_VERSION_BUILD)
 #define TUSB_VERSION_STRING    TU_STRING(TUSB_VERSION_MAJOR) "." TU_STRING(TUSB_VERSION_MINOR) "." TU_STRING(TUSB_VERSION_REVISION)
@@ -529,17 +529,6 @@
   #define CFG_TUH_MAX3421  0
 #endif
 
-// MAX3421 Host max. transfer attempts per frame (except control and iso)
-// retry quantity = (CFG_TUH_MAX3421_MAX_ATTEMPS_PER_FRAME - 1)
-// 0 = endless retries. is default to keep compatibility
-//     => this may overload MCU with permanently repeating NAK interrupts
-//     => possible increased USB traffic, increased latencies, reduced data throughput
-// a usual attempt quantity is 3 (reference: the book "USB Complete" by Jan Axelson)
-#ifndef CFG_TUH_MAX3421_MAX_ATTEMPTS_PER_FRAME
-  #define CFG_TUH_MAX3421_MAX_ATTEMPTS_PER_FRAME 0
-#endif
-
-
 //--------------------------------------------------------------------+
 // TypeC Options (Default)
 //--------------------------------------------------------------------+