diff --git a/arch/arm/src/sama5/sam_udphs.c b/arch/arm/src/sama5/sam_udphs.c
index 39f72d63ff..fde1365e8a 100644
--- a/arch/arm/src/sama5/sam_udphs.c
+++ b/arch/arm/src/sama5/sam_udphs.c
@@ -2395,8 +2395,9 @@ static void sam_dma_interrupt(struct sam_usbdev_s *priv, int epno)
            * REVISIT: On the SAMV7, I found that you really need to
            * wait for the TX completion interrupt before calling
            * sam_req_write().  For the SAMV7, the logic here just
-           * enables that TX completion interrupt.  The symptom of
-           * the problem was occassional missing zero-length packets.
+           * enables that TX completion interrupt if BYCT > 0.  The
+           * symptom of the problem was occasional missing zero-length
+           * packets because sam_req_write() was called too soon.
            */
 
           DEBUGASSERT(USB_ISEPIN(privep->ep.eplog));
diff --git a/arch/arm/src/samv7/sam_usbdevhs.c b/arch/arm/src/samv7/sam_usbdevhs.c
index 90ead9c8a8..5328727062 100644
--- a/arch/arm/src/samv7/sam_usbdevhs.c
+++ b/arch/arm/src/samv7/sam_usbdevhs.c
@@ -2524,6 +2524,10 @@ static void sam_dma_interrupt(struct sam_usbdev_s *priv, int epno)
               /* All bytes have been sent to the host.  We must call
                * sam_req_write() now in the IDLE state with the number of
                * bytes transferred in 'inflight'
+               *
+               * REVISIT: Isn't there a race condition here?  Could TXIN
+               * have fired just before calculating byct?  Could TXIN be
+               * pending here?
                */
 
               privep->epstate = USBHS_EPSTATE_IDLE;